Datasets:

suolyer

/

pile_pubmed-central

like 1

1. Introduction {#sec1-sensors-18-02788} =============== Since the concept of multiple-input and multiple-output (MIMO) radar has been proposed in recent years, it has drawn a great attention of scholars in the field of radar research \[[@B1-sensors-18-02788],[@B2-sensors-18-02788]\]. Since the MIMO radar transmits the orthogonal waveform and has different arrangement of transmit-receive array, MIMO radar can obtain both spatial diversity and waveform diversity at the same time \[[@B3-sensors-18-02788]\]. For angle estimation, MIMO radar has better parameter estimation performance than the conventional phased array radar, especially for the estimation of joint direction-of-departure (DOD) and direction-of-arrival (DOA) \[[@B4-sensors-18-02788]\]. The study of MIMO radar is mainly divided into two categories: (1) statistical MIMO radar \[[@B5-sensors-18-02788],[@B6-sensors-18-02788]\], which can obtain the spatial diversity gain of both transmit and receive arrays for improving the detection and parameter estimation performance; and (2) collocated MIMO radar \[[@B7-sensors-18-02788]\], which uses the orthogonality of the transmitted waveforms to form a large virtual array aperture for obtaining the corresponding waveform diversity gain. Thus, in the latter, the degrees of freedom (DOFs) are raised and the aperture of MIMO radar is enlarged. This literature focuses on the research of collocated MIMO radar. In bistatic MIMO radar, the estimation of joint DOD and DOA is an important problem, and many popular algorithms have been proposed for this issue. Yan et al. \[[@B8-sensors-18-02788]\] proposed the Capon method, and Gao et al. \[[@B9-sensors-18-02788]\] proposed the multiple signal classification (MUSIC) method. These two methods belong to the spatial spectral method. A spatial spectral function is firstly constructed and then the angles can be estimated from the spatial spectrum. The advantage of the spatial spectrum methods is that they can realize the automatic matching between the DOD and DOA, and the accuracy of angle estimation is high. The disadvantage is that the computational complexity is large due to the two-dimensional spatial spectral searching. To remove the process of the spatial spectral searching, the estimation method of signal parameter via rotational invariance techniques (ESPRIT) is utilized to estimate the angles of targets in MIMO radar \[[@B10-sensors-18-02788]\]. This method achieves the rotation invariance features for estimating DOD and DOA by dividing the virtual array into different subarrays, but this method cannot realize the automatic pairing between the transmit and receive angles. Then, the automatically-paired ESPRIT method is proposed in \[[@B11-sensors-18-02788]\]. In Reference \[[@B12-sensors-18-02788]\], the calculation procedure of ESPRIT algorithm is transformed into the real-valued domain by using unitary transformation, which reduces the computational complexity of ESPRIT algorithm without the performance loss. However, the algorithms mentioned above rely on the ideal transmit and receive arrays. Since the receive and transmit arrays cannot be accurately compensated, there exists the mutual coupling between the array elements \[[@B13-sensors-18-02788]\]. In view of the mutual coupling of receive and transmit array elements in MIMO radar, several methods are proposed, such as MUSIC-like and ESPRIT-like algorithms \[[@B14-sensors-18-02788],[@B15-sensors-18-02788]\]. The ESPRIT-like algorithm \[[@B14-sensors-18-02788]\] utilizes the banded symmetric Toeplitz structure of the mutual coupling matrix in the uniform linear array to remove the effect of unknown mutual coupling, but it leads the loss of array aperture. Using inherent characteristics of the signals is considered as a possible way to compensate the aperture loss. Fortunately, the strictly noncircular signals, such as binary phase shift keying (BPSK) and M-ary amplitude shift keying (MASK), have been widely used in the field of communication and radar systems for aperture extension \[[@B16-sensors-18-02788],[@B17-sensors-18-02788]\]. In view of using the noncircular characteristic of signal to improve the accuracy of DOA estimation, many algorithms have been developed \[[@B18-sensors-18-02788],[@B19-sensors-18-02788],[@B20-sensors-18-02788]\]. The robust DOA estimation method with the unknown mutual coupling is investigated in \[[@B18-sensors-18-02788]\], which takes the noncircular characteristic of the signal into account for eliminating the influence of mutual coupling. Then, the loss of array aperture is partly compensated by using the noncircular structure of these signals. On the other hand, all algorithms mentioned above need to stack the received data into a special structure matrix, which ignores the inherence multidimensional structure of signal. To utilize the inherent multidimensional structure of the signals, many methods have been developed \[[@B21-sensors-18-02788],[@B22-sensors-18-02788],[@B23-sensors-18-02788],[@B24-sensors-18-02788],[@B25-sensors-18-02788]\]. A multi-SVD algorithm is developed to estimate DOD and DOA in MIMO radar \[[@B21-sensors-18-02788]\], and the estimation performance is improved remarkably. Considering the mutual coupling in transmit and receive arrays, the subspace estimation method based on unitary tensor decomposition is introduced in \[[@B23-sensors-18-02788]\]. The algorithm converts the tensor subspace into a new real-valued tensor through using the unitary transformation while eliminating the influence of mutual coupling. Then, the estimation of DOD and DOA is obtained by using tensor-based subspace, and it can achieve better angle estimation accuracy with lower computational burden. In addition, the DODs and DOAs can be estimated in the coexistence of mutual coupling and spatial colored noise \[[@B25-sensors-18-02788]\]. According to the above analysis, these algorithms only utilize the noncircularity and inherent multidimensional structure of strictly noncircular signals separately in the case of unknown mutual coupling. In this paper, we develop a tensor-based angle estimation scheme for strictly noncircular sources in the presence of unknown mutual coupling in bistatic MIMO radar. This method not only takes the multidimensional structure of the signals into account, but also uses the noncircular characteristics of the signals. Firstly, the proposed method uses the band symmetric Toeplitz structure of mutual coupling matrix to remove the influence of unknown mutual coupling. Then, a novel augmented tensor is constructed to utilize both the noncircularity and inherent multidimensional structure of strictly noncircular signals. Afterwards, the higher order SVD (HOSVD) technique of tensor decomposition is utilized to formulate a tensor-based signal subspace. Finally, the estimation of DODs and DOAs are obtained by utilizing the rotational invariance technique, where the DODs and DOAs are matched automatically. Due to the exploitation of inherent multidimensional structure and enlarged array aperture, the proposed method has better angle estimation performance than other algorithms in the presence of mutual coupling. The experiment results are carried out to prove the advantages of the proposed method. The summary is as follows. [Section 2](#sec2-sensors-18-02788){ref-type="sec"} gives some basic concepts of tensor and the tensor-based signal model. The proposed method is developed in [Section 3](#sec3-sensors-18-02788){ref-type="sec"}. [Section 4](#sec4-sensors-18-02788){ref-type="sec"} discusses the comments and specific analysis of the proposed method. Simulation results are given in [Section 5](#sec5-sensors-18-02788){ref-type="sec"}. The conclusion of the proposed algorithm is given in [Section 6](#sec6-sensors-18-02788){ref-type="sec"}. Notation: ${( \cdot )}^{H}$, ${( \cdot )}^{T}$, ${( \cdot )}^{- 1}$ and ${( \cdot )}^{*}$ indicate conjugate-transpose, transpose, inverse, and conjugate, respectively. $\otimes$ and $\odot$ represent the Kronecker product and Khatri--Rao product, respectively. ${diag}( \cdot )$ is the diagonalization operation, and ${Toeplitz}(\mathbf{r})$ means the symmetric Toeplitz matrix constructed by the vector $\mathbf{r}$. ${vec}( \cdot )$ indicates the vectorization operation. $\arg(\gamma)$ represents the phase of $\gamma$, $\mathbf{I}_{K}$ denotes a $K \times K$ identity matrix, $0_{L \times K}$ is the $L \times K$ zero matrix, and $\mathbf{\Gamma}_{K}$ represents a matrix with ones on its anti-diagonal and zeros elsewhere. 2. Tensor Basic Concepts and Tensor-Based Signal Model {#sec2-sensors-18-02788} ====================================================== 2.1. Tensor Basic Concepts {#sec2dot1-sensors-18-02788} -------------------------- In this section, the basic concepts and operational rules of tensor are introduced. More information about tensor can be obtained from previous articles \[[@B26-sensors-18-02788],[@B27-sensors-18-02788]\]. *(Mode-n matrix unfolding). Let*$\mathcal{X} \in \mathbb{C}^{I_{1} \times I_{2} \times \cdots \times I_{N}}$*be a tensor, and the mode-n matrix unfolding of a tensor*$\mathcal{X}$*is indicated by*${\lbrack\mathcal{X}\rbrack}_{(n)}$*. The*$(i_{1},i_{2},\ldots,i_{N})$*th element of*$\mathcal{X}$*maps to the*$(i_{n},j)$*th element of*${\lbrack\mathcal{X}\rbrack}_{n}$*, where*$j = 1 + \mathsf{\Sigma}_{k = 1,k \neq n}^{N}(i_{k} - 1)J_{k}$*with*$J_{k} = \mathsf{\Pi}_{m = 1,m \neq n}^{k - 1}I_{m}$. *(Mode-n tensor-matrix product). The mode-n product of*$\mathcal{X} \in \mathbb{C}^{I_{1} \times I_{2} \times \cdots \times I_{N}}$*with a matrix*$\mathbf{A} \in \mathbb{C}^{J_{n} \times I_{n}}$*is denoted by*$\mathcal{Y} = \mathcal{X} \times_{n}\mathbf{A}$*, where*$\mathcal{Y} \in \mathbb{C}^{I_{1} \times I_{2} \times \cdots I_{n - 1} \times J_{n} \times I_{n + 1} \times \cdots \times I_{N}}$*and*${\lbrack\mathcal{Y}\rbrack}_{{}_{i_{1},\cdots,i_{n - 1},j_{n},j_{n + 1,}\cdots i_{N}}} =$$\sum\limits_{i_{n} = 1}^{I_{n}}{{\lbrack\mathcal{Y}\rbrack}_{{}_{i_{1},\cdots,i_{n},\cdots i_{N}}} \cdot {\lbrack\mathbf{A}\rbrack}_{j_{n},i_{n}}}$. *(The properties of the mode product). The mode-n tensor-matrix product satisfies the following properties* $$\left\{ \begin{array}{l} {\mathcal{X} \times_{n}\mathbf{A} \times_{m}\mathbf{B} = \mathcal{X} \times_{m}\mathbf{A} \times_{n}\mathbf{B},\quad m \neq n} \\ {\mathcal{X} \times_{n}\mathbf{A} \times_{n}\mathbf{B} = \mathcal{X} \times_{n}(\mathbf{AB}),\quad m = n} \\ \end{array} \right.$$ $$\left\lbrack {\mathcal{X} \times_{1}\mathbf{A}_{1} \times_{2}\mathbf{A}_{2} \times \cdots \times_{N}\mathbf{A}_{N}~} \right\rbrack_{(n)} = \mathbf{A}_{n} \cdot {\lbrack\mathcal{X}\rbrack}_{(n)} \cdot {\lbrack\mathbf{A}_{N} \otimes \ldots \otimes \mathbf{A}_{n + 1} \otimes \mathbf{A}_{n - 1} \otimes \ldots \otimes \mathbf{A}_{2} \otimes \mathbf{A}_{1}\rbrack}^{T}$$ *(Tensor decomposition). The HOSVD of a tensor*$\mathcal{X} \in \mathbb{C}^{I_{1} \times I_{2} \times \cdots \times I_{N}}$*is given by*$$\mathcal{X} = \mathcal{G} \times_{1}\mathbf{U}_{1} \times_{2}\mathbf{U}_{2} \times_{3}\ldots \times_{N}\mathbf{U}_{N}$$*where*$\mathcal{G} \in \mathbb{C}^{I_{1} \times I_{2} \times \cdot \cdot \cdot \times I_{N}}$*is the core tensor, and*$\mathbf{U}_{n} \in \mathbb{C}^{I_{n} \times I_{n}}(n = 1,2,3, \cdot \cdot \cdot ,N)$*is a unitary matrix, which is consist of the left singular vectors of*${\lbrack\mathcal{X}\rbrack}_{(n)}$. *(Mode-n concatenation of two tensors). The mode-n concatenation of two tensors*$\mathcal{X} \in \mathbb{C}^{I_{1} \times I_{2} \times \cdots \times I_{N}}$*and*$\mathcal{Y} \in \mathbb{C}^{I_{1} \times I_{2} \times \cdots \times I_{N}}$*is denoted as*$\mathcal{F} = \lbrack\mathcal{X}\bot_{n}\mathcal{Y}\rbrack$*, where*$\mathcal{F} \in \mathbb{C}^{I_{1} \times I_{2},\cdots, \times 2I_{N} \times ,\cdots \times I_{N}}$. 2.2. Tensor-Based Signal Model {#sec2dot2-sensors-18-02788} ------------------------------ Consider a narrowband bistatic MIMO radar, which consists of an $M$-element transmit array and an $N$-element receive array, both of which are composed of half-wavelength spaced uniform linear arrays (ULAs). At the transmit array, $M$ transmit antennas emit $M$ mutual orthogonal strictly noncircular signals, such as BPSK modulated signals. Assume that there are $K$ independent Swerling I targets with low speed in the far field. For the transmit and the receive arrays, the transmit and receive angles of the $K$ target are denoted as $\varphi_{k}$ and respectively. The reflected signals are collected by the receive antennas and dealt with matched filters formed by the transmitted orthogonal waveforms. Then, the output of the received signal can be expressed as \[[@B19-sensors-18-02788],[@B20-sensors-18-02788],[@B21-sensors-18-02788]\]. $$\overline{\mathbf{X}}(t_{l}) = \mathbf{A}_{r}\sum(t_{l})\mathbf{A}_{t}^{T} + \mathbf{N}(t_{l}),~l = 1,2,\cdots,L$$ where $\overline{\mathbf{X}}(t_{l}) \in \mathbb{C}^{N \times M}$ is the received data at the $l$th snapshots, $\mathbf{A}_{r} = \left\lbrack \mathbf{a}_{r}(\theta_{1}),\mathbf{a}_{r}(\theta_{2}),\cdots,\mathbf{a}_{r}(\theta_{K})\rbrack \in \mathbb{C}^{N \times K} \right.$ is the receive steering matrix consisting of receive steering vector $\mathbf{a}_{r}(\theta_{k}) = \lbrack 1,\exp(j\pi\sin\theta_{k}),\cdots,$$\exp(j\pi(N - 1)\sin\theta_{k})\rbrack^{T}$, $\mathbf{A}_{t} = \left\lbrack \mathbf{a}_{t}(\varphi_{1}),\mathbf{a}_{t}(\varphi_{2}),\cdots,\mathbf{a}_{t}(\varphi_{K})\rbrack \in \mathbb{C}^{M \times K} \right.$ is the transmit steering matrix consisting of the transmit steering vector $\mathbf{a}_{t}(\varphi_{k}) = {\lbrack 1,\exp(j\pi\sin\varphi_{k}),\cdots,\exp(j\pi(M - 1)\sin\varphi_{k})\rbrack}^{T}$. $\sum(t_{l}) = diag(\mathbf{s}(t_{l})) \in \mathbb{C}^{K \times K}$ is the strictly noncircular signal data with $\mathbf{s}(t_{l}) = {\lbrack\mathbf{s}_{1}(t_{l}),\mathbf{s}_{2}(t_{l}),\cdots,\mathbf{s}_{K}(t_{l})\rbrack}^{T}$ at the $l$th snapshots, and the noncircular signal vector $\mathbf{s}(t_{l})$ satisfies with $\mathbf{s}(t_{l}) = \Delta\mathbf{s}_{c}(t_{l})$ where $\Delta = diag(\lbrack\exp(j\phi_{1}),\exp(j\phi_{2}),\cdots,$$\exp(j\phi_{K})\rbrack)$ with the arbitrary phases $\phi_{k}(k = 1,2,\cdots,K)$ is assumed to be different for each source, and $\mathbf{s}_{c}(t_{l}) = \mathbf{s}_{c}^{*}(t_{l})$, $\mathbf{N}(t_{l}) \in \mathbb{C}^{N \times M}$ is the additional Gaussian white noise matrix. On account of the radiation effect between the antenna elements, the effect of mutual coupling will be produced \[[@B13-sensors-18-02788]\], and the mutual coupling between antenna elements of uniform linear array can be expressed as banding symmetric Toeplitz matrix, which is called mutual coupling matrix. The mutual coupling coefficient between two antennas in a ULA is inversely proportional to the distance between them, and the mutual coupling coefficient decreases with the increase of the distance, and vice versa \[[@B14-sensors-18-02788]\]. Assume that there are $P + 1$ nonzero mutual coupling coefficients for both transmit and receive arrays, and $P$ is satisfied with $\min\left\{ M,N \right\} > 2P$. Taking the influence of mutual coupling into account, the received signal in Equation (4) can be expressed as $$\mathbf{X}(t_{l}) = \lbrack\mathbf{C}_{r}\mathbf{A}_{r}\rbrack\sum(t_{l}){\lbrack\mathbf{C}_{t}\mathbf{A}_{t}\rbrack~}^{T} + \mathbf{N}(t_{l})$$ where $\mathbf{C}_{t} = {toeplitz}(\lbrack\mathbf{c}_{t}^{T},0_{1 \times (M - P - 1)}\rbrack) \in C^{M \times M}$ and $\mathbf{C}_{r} = {toeplitz}(\lbrack\mathbf{c}_{t}^{T},0_{1 \times (N - P - 1)}\rbrack) \in C^{N \times N}$ are the mutual coupling matrices with $\mathbf{c}_{t} = \lbrack c_{t0},c_{t1}, \cdot \cdot \cdot ,c_{tP}\rbrack$ and $\mathbf{c}_{r} = \lbrack c_{r0},c_{r1}, \cdot \cdot \cdot ,c_{rP}\rbrack$, $\mathbf{c}_{ip}(i = r,t;p = 0,1,2, \cdot \cdot \cdot ,P)$ is the $P + 1$ nonzero mutual coupling coefficients, which satisfy with $\left. 0 < \middle| c_{ip} \middle| < , \cdot \cdot \cdot , < \middle| c_{i1} \middle| < \middle| c_{i0} \middle| = 1 \right.$. The mutual coupling matrices have the banded symmetric Toeplitz structure. In other words, the mutual coupling matrix $\mathbf{C}_{i}(i = r,t)$ can be expressed as $$\mathbf{C}_{i} = \left\lbrack \begin{array}{lllllllll} c_{i0} & c_{i1} & \cdots & c_{ip} & & & & & \\ c_{i1} & c_{i0} & c_{i1} & \cdots & c_{ip} & & & 0 & \\ \vdots & \ddots & \ddots & \ddots & \vdots & \ddots & & & \\ c_{ip} & \cdots & c_{i1} & c_{i0} & c_{i1} & \cdots & c_{ip} & & \\ & \ddots & \vdots & \ddots & \ddots & \ddots & & \ddots & \\ & & c_{ip} & \cdots & c_{i1} & c_{i0} & c_{i1} & \cdots & c_{ip} \\ & & & \ddots & \vdots & \ddots & \ddots & \ddots & \vdots \\ & & 0 & & c_{ip} & \cdots & c_{i1} & c_{i0} & c_{i1} \\ & & & & & c_{ip} & \cdots & c_{i1} & c_{i0} \\ \end{array} \right\rbrack$$ According to the basic concepts and operational rules of tensor, the received signal $\mathbf{X}(t_{l})(l = 1,2,\cdots,L)$ can be seen as different slices of a third-order tensor along the direction of snapshot (the third-dimension). By collecting $L$ snapshots, a third-order tensor data $\mathcal{X} \in \mathbb{C}^{N \times M \times L}$ is formed as $$\lbrack\mathcal{X}_{:,:,l~}\rbrack = \mathbf{X}(t_{l}),l = 1,2,\cdots,L$$ where $\lbrack\mathcal{X}_{:,:,l}\rbrack$ is the $l$th slice of the tensor along the third-dimension. According to the definition of Mode-$n$ matrix unfolding, the relationship between the tensor-based data model and matrix-based data model is expressed as $${\lbrack\mathcal{X}\rbrack~}_{(3)}^{T} = \lbrack{\overline{\mathbf{A}}}_{t} \odot {\overline{\mathbf{A}}}_{r}\rbrack\mathbf{S} + \mathbf{N} = \lbrack{\overline{\mathbf{A}}}_{t} \odot {\overline{\mathbf{A}}}_{r}\rbrack\Delta\mathbf{S}_{c} + \mathbf{N}$$ where ${\overline{\mathbf{A}}}_{t} = \mathbf{C}_{t}\mathbf{A}_{t}$, ${\overline{\mathbf{A}}}_{r} = \mathbf{C}_{r}\mathbf{A}_{r}$ is the transmit-receive steering matrix, $\mathbf{S} = \lbrack\mathbf{s}(t_{1}),\mathbf{s}(t_{2}),\cdots,\mathbf{s}(t_{L})\rbrack \in \mathbb{C}^{K \times L}$ is the signal matrix, and $\mathbf{S}_{c} = \lbrack\mathbf{s}_{c}(t_{l}),\mathbf{s}_{c}(t_{2}),\cdots,\mathbf{s}_{c}(t_{L})\rbrack \in \mathbb{R}^{K \times L}$ satisfies with $\mathbf{S}_{c} = \mathbf{S}_{c}^{*}$. $\mathbf{N} = \lbrack{vec}(\mathbf{N}(t_{l})),$ ${vec}(\mathbf{N}(t_{l})),\cdots,{vec}(\mathbf{N}(t_{L}))\rbrack \in \mathbb{C}^{MN \times L}$ is the noise matrix. 3. Tensor-Based Angle Estimation Method with Unknown Mutual Coupling {#sec3-sensors-18-02788} ==================================================================== In this section, a tensor-based angle estimation algorithm is investigated for capturing the noncircularity and inherent multidimensional structure of the received signal to improve the accuracy of angle estimation in the case of unknown mutual coupling. 3.1. Mutual Coupling Elimination {#sec3dot1-sensors-18-02788} -------------------------------- In Equation (7), the mutual coupling affects the transmit direction matrix ${\overline{\mathbf{A}}}_{t}$ and the receive direction matrix ${\overline{\mathbf{A}}}_{r}$, therefore the Vandermonde structure of ${\overline{\mathbf{A}}}_{t}$ and ${\overline{\mathbf{A}}}_{r}$ are destroyed. Fortunately, the mutual coupling matrices are banded symmetric Toeplitz. It can extract two sub-matrices from transmit and receive direction matrices to elimination the influence of mutual coupling. By defining two choice matrices $$\begin{array}{l} {\mathbf{J}_{1} = \lbrack 0_{(N - 2P) \times P~},I_{(N - 2P)},0_{(N - 2P) \times P}\rbrack} \\ {\mathbf{J}_{2} = \lbrack 0_{(M - 2P) \times P},I_{(M - 2P)},0_{(M - 2P) \times P}\rbrack} \\ \end{array}$$ based on the characteristics of the mutual coupling matrix, we have $$\left\{ \begin{array}{l} {{\hat{\mathbf{a}}}_{r}(\theta_{k}) = \mathbf{J}_{1}{\overline{\mathbf{a}}}_{r}(\theta_{k}) = \beta_{rk~}{\widetilde{\mathbf{a}}}_{r}(\theta_{k})} \\ {{\hat{\mathbf{a}}}_{t}(\varphi_{k}) = \mathbf{J}_{2}{\overline{\mathbf{a}}}_{t}(\varphi_{k}) = \beta_{tk}{\widetilde{\mathbf{a}}}_{t}(\varphi_{k})} \\ \end{array} \right.$$ where $\beta_{tk} = 1 + {\sum_{p = 1}^{p}{2c_{tp}\cos(p\pi\sin\varphi_{k})}}$, $\beta_{rk} = 1 + {\sum_{p = 1}^{p}{2c_{rp}\cos(p\pi\sin\theta_{k})}}$. ${\overline{\mathbf{a}}}_{t}(\varphi_{k})$ and ${\overline{\mathbf{a}}}_{r}(\theta_{k})$ are the transmit and receive steering vectors with mutual coupling, respectively. ${\widetilde{\mathbf{a}}}_{r}(\theta_{k})$ and ${\widetilde{\mathbf{a}}}_{t}(\varphi_{k})$ are the column vectors of the first $\widetilde{N} = N - 2P$ and $\widetilde{M} = M - 2P$ elements of $\mathbf{a}_{r}(\theta_{k})$ and $\mathbf{a}_{t}(\varphi_{k})$. It can be clearly seen from Equation (9) that parameters $\beta_{tk}$ and $\beta_{rk}$ are constant for each target, which means that the direction matrices $\mathbf{\hat{A}}_{r}(\theta) = \lbrack{\hat{\mathbf{a}}}_{r}(\theta_{1}),{\hat{\mathbf{a}}}_{r}(\theta_{2}), \cdot \cdot \cdot ,{\hat{\mathbf{a}}}_{r}(\theta_{K})\rbrack$ and $\mathbf{\hat{A}}_{t}(\theta) = \lbrack{\hat{\mathbf{a}}}_{t}(\theta_{1}),{\hat{\mathbf{a}}}_{t}(\theta_{2}), \cdot \cdot \cdot ,{\hat{\mathbf{a}}}_{t}(\theta_{K})\rbrack$ are the Vandermonde matrices \[[@B23-sensors-18-02788]\]. Therefore, the effect of mutual coupling is removed after. The procedure of decoupling in Equation (9) can be extended to the tensor domain in Equation (7):$$\hat{\mathcal{X}} = \mathcal{X}_{\times 1~}J_{1 \times 2}J_{2} = \mathcal{I}_{K \times 1}\mathbf{\hat{A}}_{r \times 2}\mathbf{\hat{A}}_{t \times 3}\mathbf{S} + \hat{\mathcal{N}}$$ where $\hat{\mathcal{N}} = \mathcal{N}_{\times 1}J_{1 \times 2}J_{2}$ is a part of $\mathcal{N}$, the tensor noise $\hat{\mathcal{N}}$ has the same properties as the $\mathcal{N}$. Then, according to the definition Mode-$n$ matrix unfolding, the mode-3 matrix unfolding of $\hat{\mathcal{X}} \in \mathbb{C}^{2\widetilde{N} \times \widetilde{M} \times L}$ is written as $${\lbrack\hat{\mathcal{X}}\rbrack~}_{(3)}^{T} = \lbrack{\hat{\mathbf{A}}}_{t} \odot {\hat{\mathbf{A}}}_{r}\rbrack\mathbf{S} + \hat{\mathbf{N}} = \lbrack{\hat{\mathbf{A}}}_{t} \odot {\hat{\mathbf{A}}}_{r}\rbrack\Delta\mathbf{S}_{c} + \hat{\mathbf{N}}$$ From the above analysis, we can see that ${\hat{\mathbf{A}}}_{t}$ and ${\hat{\mathbf{A}}}_{r}$ have Vandermonde structure. It can be shown that the mutual coupling effect has been removed in the tensor domain. 3.2. Tensor Augmentation and Signal Subspace Estimation {#sec3dot2-sensors-18-02788} ------------------------------------------------------- To utilize the noncircular property of the signal in the tensor domain, a special augmented tensor is constructed by the tensor-based forward and backward smoothing technique:$$\mathcal{Y} = \left\lbrack \hat{\mathcal{X}}\bot_{1}({\hat{\mathcal{X}}}^{*} \times_{1}\mathsf{\Gamma}_{\widetilde{N}} \times_{2}\mathsf{\Gamma}_{\widetilde{M}}) \right\rbrack~$$ Then, according to the definition Mode-$n$ matrix unfolding, the mode-3 matrix unfolding of $\mathcal{Y} \in \mathbb{C}^{2\widetilde{N} \times \widetilde{M} \times L}$ is written as $${\lbrack\mathcal{Y}\rbrack~}_{(3)}^{T} = \lbrack\mathbf{\hat{A}}_{t} \odot \mathbf{\overset{‿}{A}}_{r}\rbrack\mathbf{S}_{c} + \hat{\mathbf{N}}$$ where $\mathbf{\overset{‿}{A}}_{r} = {\lbrack{(\mathbf{\hat{A}}_{r}^{1})}^{T},{(\mathbf{\hat{A}}_{r}^{2})}^{T}\rbrack}^{T} \in \mathbb{C}^{2\widetilde{M}\widetilde{N} \times K}$ denotes the extended steering matrix, where ${\hat{\mathbf{A}}}_{r}^{1} = {\hat{\mathbf{A}}}_{r}\Delta$ and ${\hat{\mathbf{A}}}_{r}^{2} = {\hat{\mathbf{A}}}_{r}\Delta^{*}\mathbf{D}_{t}\mathbf{D}_{r}$ with $\mathbf{D}_{t} = {diag}(\lbrack\exp( - j\pi(\widetilde{M} - 1)\sin\varphi_{1}),\exp( - j\pi(\widetilde{M} - 1)\sin\varphi_{2}),\cdots,\exp( - j\pi(\widetilde{M} - 1)\sin\varphi_{K})\rbrack)$ and $\mathbf{D}_{r} = {diag}(\lbrack\exp( - j\pi(\widetilde{N} - 1)\sin\theta_{1}),\exp( - j\pi(\widetilde{N} - 1)\sin\theta_{2}),\cdots,\exp( - j\pi(\widetilde{N} - 1)\sin\theta_{K})\rbrack)$, $\hat{\mathbf{N}}$ is the modified noise matrix. It is easy to know from Equation (13) that the available array aperture is twice the model in Equation (11). It can be found that the augmented tensor $\mathcal{Y}$ not only considers the multidimensional structure of tensor, but also captures the noncircularity of the signal and enlarges the virtual aperture of the array. Thus, a better performance of parameter estimation is expected to be achieved in the proposed method. Based on the augmented tensor in Equation (12), the HOSVD method is applied to the augmented tensor $\mathcal{Y}$: $$\mathcal{Y} = \mathcal{G} \times_{1}\mathbf{E}_{1} \times_{2}\mathbf{E}_{2} \times_{3}\mathbf{E}_{3}$$ where $\mathbf{E}_{1} \in \mathbb{C}^{2\widetilde{N} \times 2\widetilde{N}}$, $\mathbf{E}_{2} \in \mathbb{C}^{\widetilde{M} \times \widetilde{M}}$ and $\mathbf{E}_{3} \in \mathbb{C}^{L \times L}$ are unitary matrices, which are made up of the left singular of the mode-$n(n = 1,2,3)$ of matrix unfolding of $\mathcal{Y}$ as ${\lbrack\mathcal{Y}\rbrack}_{(n)} = \mathbf{E}_{n}\mathsf{\Lambda}_{n}\mathbf{V}_{n}^{H}$, respectively. $\mathcal{G} \in \mathbb{C}^{2\widetilde{N} \times \widetilde{M} \times L}$ represents the core tensor. Because there are $K$ sources, $\mathcal{Y}$ is rank-$K$ tensor. Then, a subspace tensor is achieved by using the truncated HOSVD of $\mathcal{Y}$, which is shown as $$\mathcal{Y}_{s} = \mathcal{G}_{s} \times_{1}\mathbf{E}_{s1~} \times_{2}\mathbf{E}_{s2}$$ where $\mathbf{E}_{sn}(n = 1,2,3)$ is made up of the column vectors of $\mathbf{E}_{n}$ corresponding to the largest $K$ singular values, and $\mathcal{G}_{s} = \mathcal{Y} \times_{1}\mathbf{E}_{s1}^{H} \times_{2}\mathbf{E}_{s2}^{H} \times_{3}\mathbf{E}_{3s}^{H}$ denotes the signal component of $\mathcal{G}$. Then, according to the definition of Mode-$n$ tensor-matrix product, substituting $\mathcal{G}_{s}$ into Equation (15) yields $$\mathcal{Y}_{s} = \mathcal{Y} \times_{1}(\mathbf{E}_{s1~}\mathbf{E}_{s1}^{H}) \times_{2}(\mathbf{E}_{s2}\mathbf{E}_{s2}^{H}) \times_{3}\mathbf{E}_{3s}^{H}$$ Then, the tensor-based signal subspace is given by using the mode-3 matrix unfolding of $\mathcal{Y}_{s}$, and according to the properties of the mode product, the tensor-based signal subspace is shown as $${\overline{\mathbf{U}}}_{s} = {\lbrack\mathcal{Y}_{s}\rbrack~}_{(3)}^{T} = (\mathbf{E}_{s2}\mathbf{E}_{s2}^{H} \otimes \mathbf{E}_{s1}\mathbf{E}_{s1}^{H}){\lbrack\mathcal{Y}_{s}\rbrack}_{(3)}\mathbf{E}_{3s}^{*}$$ After using some simplification in \[[@B21-sensors-18-02788],[@B23-sensors-18-02788],[@B24-sensors-18-02788]\], the tensor based signal subspace is written as $$\mathbf{\overline{U}}_{s} = (\mathbf{E}_{s2~}\mathbf{E}_{s2}^{H} \otimes \mathbf{E}_{s1}\mathbf{E}_{s1}^{H})\mathbf{U}_{s}$$ where $\mathbf{U}_{s}$ is the signal subspace of ${\lbrack\mathcal{Y}_{s}\rbrack}_{(3)}$, which can be estimated by truncating SVD of ${\lbrack\mathcal{Y}_{s}\rbrack}_{(3)}$ as ${\lbrack\mathcal{Y}_{s}\rbrack}_{(3)} \approx \mathbf{U}_{s}\mathsf{\Lambda}_{s}\mathbf{V}_{s}^{H}$. According to Equation (18), it is shown that the $\mathbf{\overline{U}}_{s}$ and $\mathbf{U}_{s}$ span to the same subspace, which means that the tensor based signal subspace ${\overline{\mathbf{U}}}_{s}$ and augmented steering matrix $\overset{‿}{\mathbf{A}} = {\hat{\mathbf{A}}}_{t} \odot {\overset{‿}{\mathbf{A}}}_{r}$ also span to the same subspace. Thus, there is a nonsingular matrix $\mathbf{T}$ satisfied with $\mathbf{\overline{U}}_{s} = \overset{‿}{\mathbf{A}}\mathbf{T}$, and the estimation of DODs and DOAs can be achieved from this tensor-based signal subspace. 3.3. Joint DOD and DOA Estimation {#sec3dot3-sensors-18-02788} --------------------------------- Noting that $\mathbf{\hat{A}}_{t}$ has Vandermonde structure, there exists the following rotation invariance equation \[[@B10-sensors-18-02788],[@B11-sensors-18-02788]\] $$\prod_{2}\overset{‿}{\mathbf{A}} = \prod_{1}\overset{‿}{\mathbf{A}}\mathbf{\Phi}_{t}~$$ where $\mathbf{\Phi}_{t} = {diag}(\lbrack\exp(j\pi\sin\varphi_{1}),\exp(j\pi\sin\varphi_{2}),\cdots,\exp(j\pi\sin\varphi_{k})\rbrack)$ is s rotational invariance factor matrix that contains the desired information of DOAs. $\prod_{1} = \mathbf{J}_{3} \otimes \mathbf{I}_{2\widetilde{N}}$ and $\prod_{2} = \mathbf{J}_{4} \otimes \mathbf{I}_{2\widetilde{N}}$ are selection matrices with $\mathbf{J}_{3} = \lbrack\mathbf{I}_{\widetilde{M} - 1},\mathbf{O}_{(\widetilde{M} - 1) \times 1}\rbrack$ and $\mathbf{J}_{4} = \lbrack\mathbf{O}_{(\widetilde{M} - 1) \times 1},\mathbf{I}_{\widetilde{M} - 1}\rbrack$, respectively. Simultaneously, both $\mathbf{\hat{A}}_{r}^{1}$ and $\mathbf{\hat{A}}_{r}^{2}$ have Vandermonde-like structures in ${\overset{‿}{\mathbf{A}}}_{r}$. There is another rotation invariance equation:$$\prod_{4}\overset{‿}{\mathbf{A}} = \prod_{3}\overset{‿}{\mathbf{A}}\mathbf{\Phi}_{r}~$$ where $\mathbf{\Phi}_{r} = {diag}(\lbrack\exp(j\pi\sin\theta_{1}),\exp(j\pi\sin\theta_{2}),\cdots,\exp(j\pi\sin\theta_{k})\rbrack)$ contains the desired information of DODs. $\prod_{3} = \mathbf{I}_{2\widetilde{M}} \otimes \mathbf{J}_{5}$ and $\prod_{4} = \mathbf{I}_{2\widetilde{M}} \otimes \mathbf{J}_{6}$ are the selection matrices with $\mathbf{J}_{5} = \lbrack\mathbf{I}_{\widetilde{N} - 1},\mathbf{O}_{(\widetilde{N} - 1) \times 1}\rbrack$ and $\mathbf{J}_{6} = \lbrack\mathbf{O}_{(\widetilde{N} - 1) \times 1},\mathbf{I}_{\widetilde{N} - 1}\rbrack$, respectively. Utilizing the relationship between the augmented steering matrix and tensor-based signal subspace shown as $\mathbf{\overline{U}}_{s} = \overset{‿}{\mathbf{A}}\mathbf{T}$, the following rotational invariance property can be achieved, which is shown as $${\prod_{2}\mathbf{\overline{U}}_{s} = \prod_{1}\mathbf{\overline{U}}_{s}\mathbf{\Psi}_{t}}{\prod_{4}\mathbf{\overline{U}}_{s} = \prod_{3}\mathbf{\overline{U}}_{s}\mathbf{\Psi}_{r}~}$$ where $\mathbf{\Psi}_{t} = \mathbf{T}\mathbf{\Phi}_{t}\mathbf{T}^{- 1}$ and $\mathbf{\Psi}_{r} = \mathbf{T}\mathbf{\Phi}_{r}\mathbf{T}^{- 1}$. The least squares (LS) or the total least squares (TLS) technique is applied to Equation (21) for estimating $\mathbf{\Psi}_{t}$ and $\mathbf{\Psi}_{r}$. Then, the estimation of $\mathbf{\Phi}_{t}$ can be obtained through the EVD of $\mathbf{\Psi}_{t}$, and supposing $\mathbf{\overline{T}}$ be the eigenvector matrix of $\mathbf{\Psi}_{t}$. To achieve the estimation of the DOAs paired with the estimated DODs, calculate the $\mathbf{\Phi}_{r}$ via $\mathbf{\overline{T}}\mathbf{\Psi}_{r}\mathbf{\overline{T}}^{- 1}$. Finally, the DODs and DOAs are derived as $${{\hat{\mathsf{\varphi}}}_{p} = \arcsin\lbrack\arg(u_{k})/\pi\rbrack}{{\hat{\mathsf{\theta}}}_{p} = \arcsin\lbrack\arg(v_{k})/\pi\rbrack~}$$ where $u_{k}$ and $v_{k}$ are the $k$th diagonal elements of $\mathbf{\Phi}_{t}$ and $\mathbf{\Phi}_{r}$, respectively. 4. Remarks and Algorithm Analysis {#sec4-sensors-18-02788} ================================= 4.1. Related Remarks {#sec4dot1-sensors-18-02788} -------------------- *If a signal has only the in-phase component and the orthogonal component is zero, the signal is called a noncircular signal. The difference between a noncircular signal and a circular signal is whether the elliptic covariance is equal to zero. For a noncircular signal, the elliptic covariance is not equal to zero, which means that more information can be used. Therefore, the number of available array elements can be increased by reconstructing the receiving data matrix of noncircular signals.* *The methods proposed in \[[@B14-sensors-18-02788],[@B15-sensors-18-02788]\] are also based on Equation (7), but these two methods ignore the multidimensional and noncircular characteristics of the measurement tensor*$\mathcal{Y}$, *so the estimation performance is not satisfactory. Additionally, the methods proposed in \[[@B19-sensors-18-02788],[@B22-sensors-18-02788]\] utilize the noncircular characteristics and multidimensional structure of signals respectively. However, both methods are completely invalid under the condition of mutual coupling.* *For the tensor-based signal model in Equation (6), the method in \[[@B23-sensors-18-02788]\] investigates the way to remove the influence of mutual coupling in tensor domain for improving the performance. On the other hand, based on the matrix-based signal model in Equation (7), the noncircularity of signals is utilized to enlarge the array aperture after removing the effect of mutual coupling in \[[@B18-sensors-18-02788]\]. It has been shown that the existing methods consider the noncircularity and inherent multidimensional structure of strictly noncircular signals separately with unknown mutual coupling. However, the proposed algorithm utilizes the noncircularity and the inherent multidimensional structure simultaneously, which results in more accurate signal subspace estimation and excellent angle estimation performance. The experiment results will show its advantage.* 4.2. Computation Complexity {#sec4dot2-sensors-18-02788} --------------------------- According to Golub et al. \[[@B28-sensors-18-02788]\], it is known that for a $M \times N$ dimensional matrix, the $K$ rank truncating SVD decomposition requires $O(MNK)$ complexity. The computational complexity of the algorithm proposed in this paper is mostly concentrated on the high order singular value decomposition of the tensor $\mathcal{Y}$. In other words, the three-dimensional SVD decomposition is used for the tensor $\mathcal{Y}$, so the corresponding computational complexity is $O(6\widetilde{M}\widetilde{N}K)$. On the other hand, the computational complexity of the Tensor unitary ESPRIT algorithm in \[[@B23-sensors-18-02788]\] is $O(\widetilde{M}\widetilde{N}K{3/4})$. Thus, the algorithm proposed in this paper has higher computational burden than Tensor unitary ESPRIT algorithm, but it has superior angle estimation performance. 5. Simulation Results {#sec5-sensors-18-02788} ===================== In this part, some numerical experiments are carried out to prove that the proposed algorithm has superior angle estimation performance. ESPRIT-like algorithm \[[@B14-sensors-18-02788]\], Tensor unitary ESPRIT algorithm \[[@B23-sensors-18-02788]\] and Cramer--Rao bound (CRB) \[[@B14-sensors-18-02788]\] were compared with the proposed method. In these simulations, the bistatic MIMO radar is made up of $M = 8$ transmit antennas and $N = 10$ receive antennas, both of which are composed of half-wavelength spaced uniform linear arrays (ULAs). Unless stated otherwise, it was assumed that there are $K = 3$ uncorrelated targets, located at $(\varphi_{1},\theta_{1}) = (5^{{^\circ}}, - 8^{{^\circ}})$, $(\varphi_{2},\theta_{2}) = ( - 5^{{^\circ}},15^{{^\circ}})$ and $(\varphi_{3},\theta_{3}) = (10^{{^\circ}}, - 5^{{^\circ}})$. The root mean square error (RMSE) was utilized to achieve the evaluation of angle estimation performance, which is expressed as $${RMSE} = \sqrt{\frac{1}{2{QK}}{\sum\limits_{k = 1}^{K}{\sum\limits_{i = 1}^{Q}{\lbrack{({\hat{\varphi}}_{k,i} - \varphi_{k})}^{2} + {({\hat{\theta}}_{k,i} - \theta_{k})}^{2}\rbrack}}}}$$ where ${\hat{\varphi}}_{k,i}$ and ${\hat{\theta}}_{k,i}$ are the estimation of DOD $\varphi_{k}$ and DOA $\theta_{k}$ for the *i*th Monte Carlo trial, respectively. the total number of Monte Carlo trials was assumed as $Q$, and $Q = 500$ was used in the following simulations. The other parameter is the probability of the successful detection (PSD) expressed as ${PSD} = \left( D/Q \right) \times 100\%$, where $D$ represents the total number of successful times and a successful trial requires the absolute error of all the experiment results are smaller than $\min\lbrack{({\hat{\varphi}}_{k} - \varphi_{k})}_{k = 1}^{K},{({\hat{\theta}}_{k} - \theta_{k})}_{k = 1}^{K}\rbrack$. For the mutual coupling parameters, there are two cases: (1) $P = 1$ with $c_{t} = \lbrack 1,0.1185 + j0.058\rbrack$ and $c_{r} = \lbrack 1,0.1520 + j0.0248\rbrack$; and (2) $P = 2$ with $c_{t} = \lbrack 1,0.72 + j0.03,0.18 + j0.072\rbrack$ and $c_{r} = \lbrack 1,0.58 + j0.0145,0.13 + j0.0482\rbrack$. In the first simulation, we investigated the estimation results of the proposed method, and the SNR versus RMSE in two different situations ([Figure 1](#sensors-18-02788-f001){ref-type="fig"}, [Figure 2](#sensors-18-02788-f002){ref-type="fig"} and [Figure 3](#sensors-18-02788-f003){ref-type="fig"}). The number of snapshots is $L = 100$. [Figure 1](#sensors-18-02788-f001){ref-type="fig"} shows the estimation results of the proposed algorithm with SNR = 0 dB in Case (1). We can clearly see that DODs and DOAs were correctly identified and matched accurately, which verifies the validity of the proposed algorithm. [Figure 2](#sensors-18-02788-f002){ref-type="fig"} depicts the RMSE versus SNR with different methods in Case (1). At the same time, [Figure 3](#sensors-18-02788-f003){ref-type="fig"} depicts the RMSE versus SNR in Case (2). In [Figure 2](#sensors-18-02788-f002){ref-type="fig"}, the angle estimation performance of the proposed method is clearly superior to the Tensor unitary ESPRIT algorithm and the ESPRIT-like algorithm, and the performance of the proposed algorithm is close to the Cramer--Rao bound (CRB). That is because the proposed method not uses the multidimensional structure of the signal, but also utilizes the noncircularity characteristics. Other methods only consider the noncircular structure or tensor multidimensional structure. In addition, the performance of the Tensor unitary ESPRIT method is better than the ESPRIT-like method, because the Tensor unitary ESPRIT method considers the multidimensional structure of the signal and obtains superior estimation performance with unknown mutual coupling. Similar conclusions can be achieved from [Figure 3](#sensors-18-02788-f003){ref-type="fig"}, which means that the proposed method can obtain superior performance in both cases. In the second simulation, we analyzed the angle estimation performance of different algorithms in the presence of $K = 2$ targets, where the two targets are located at $(\varphi_{1},\theta_{1}) = (10^{{^\circ}}, - 5^{{^\circ}})$ and $(\varphi_{2},\theta_{2}) = (5^{{^\circ}},0^{{^\circ}})$. The number of snapshots was $L = 100$ and the mutual coupling in [Figure 4](#sensors-18-02788-f004){ref-type="fig"} is set as Case (1). At the same time, [Figure 5](#sensors-18-02788-f005){ref-type="fig"} depicts the RMSE versus SNR in Case (2). [Figure 4](#sensors-18-02788-f004){ref-type="fig"} depicts the RMSE versus SNR with different algorithms for two targets. In [Figure 4](#sensors-18-02788-f004){ref-type="fig"}, the performance of our proposed method is still superior to that of Tensor unitary ESPRIT method and ESPRIT-like method. In addition, the estimation performance of Tensor unitary ESPRIT method is superior to that of ESPRIT-like method. Similar conclusions can be achieved from [Figure 5](#sensors-18-02788-f005){ref-type="fig"}, which means that the proposed method can obtain superior performance in both cases. The third simulation indicates the RMSE versus SNR of different transmit--receive array configurations for $K = 3$ targets in Case (1). As shown in [Figure 6](#sensors-18-02788-f006){ref-type="fig"}, the angle estimation performance of all the three algorithms improved with the increasing of the elements of transmit and receive arrays, in which the configuration of transmit--receive array is $M = 6,N = 8$ and $M = 8,N = 10$, respectively. The main reason is that more spatial diversity gain of MIMO radar was obtained with more transmit and receive arrays. Finally, the spatial resolution of the proposed method is improved. The fourth simulation describes the RMSE versus snapshots of different algorithms for $K = 3$ targets, where SNR = 0 dB and the mutual coupling is set as Case (1). In [Figure 7](#sensors-18-02788-f007){ref-type="fig"}, the performance of all algorithms improved with more snapshots. The performance of our proposed algorithm is superior to that of several other methods in general, but the performance of the Tensor unitary ESPRIT method is slightly better than the proposed method under very low snapshots, which is because the Tensor unitary ESPRIT method increases the number of snapshots effectively by spatial smoothing. When the number of snapshots is greater than a specific threshold, the performance of the proposed method is superior to that of Tensor unitary ESPRIT method and ESPRIT-like method, and is very close to CRB. In addition, the performance of the Tensor unitary ESPRIT method is close to the performance of the ESPRIT-like method under the condition of large snapshot number, but it is still inferior to the proposed method. The fifth simulation depicts the probability of successful detection of several algorithms versus SNR for $K = 3$ targets, where the number of snapshots was $L = 100$ and the mutual coupling is set as Case (1). In [Figure 8](#sensors-18-02788-f008){ref-type="fig"}, all algorithms can achieve 100% accuracy at high SNR region, but the accuracy of the proposed method can reach 100% faster at certain SNR. In other words, in the same case of SNR, the proposed algorithm has a higher PSD than other algorithms. That is mainly because the proposed algorithm can reasonably utilize the noncircular characteristics and multidimensional structure characteristics of signals to promote the performance of angle estimation. 6. Conclusions {#sec6-sensors-18-02788} ============== In this paper, a tensor-based angle estimation approach is proposed for strictly noncircular signals with unknown mutual coupling in MIMO radar. The proposed algorithm can capture both noncircularity and multidimensional structure of signals via formulating a novel augmented tensor. Meanwhile, it can remove the influence of unknown mutual coupling in the tensor domain. As a result, the proposed method has superior angle estimation to the existing subspace-based methods. The advantage of the proposed algorithm is clearly demonstrated using numerical experiments. Y.G. wrote the manuscript, and X.W. provided the idea of the manuscript. M.H. and C.S. assisted in performing the experiments and analyzing the results. W.W. gave some useful suggestions to revise the paper. C.C. and G.B. gave some helpful suggestions to improve the performance and presentation. This work was supported by the National Natural Science Foundation of China (61701144, 61661019), Nature Science Foundation of Beijing Municipality (4174084), the Program of Hainan Association for Science and Technology Plans to Youth R&D Innovation (QCXM201706), the scientific research projects of University in Hainan Province (Hnky2018ZD-4), the major Science and Technology Project of Hainan Province (ZDKJ2016015), the Natural Science Foundation of Hainan Province (617024), and the Scientific Research Setup Fund of Hainan University (KYQD(ZR) 1731). The authors declare no conflict of interest. {#sensors-18-02788-f001} {#sensors-18-02788-f002} {#sensors-18-02788-f003} {#sensors-18-02788-f004} {#sensors-18-02788-f005} {#sensors-18-02788-f006} {#sensors-18-02788-f007} {#sensors-18-02788-f008}

Kheiri B, Al Salihi M, Maldonado D, Nakhleh R, Bachuwa G. Warfarin‐induced spontaneous iliopsoas hematoma --- An unusual complication. Clin Case Rep. 2018;6:1639--1640. 10.1002/ccr3.1665 QUIZ QUESTION: WHAT IS THE DIAGNOSIS AND THE MANAGEMENT? {#ccr31665-sec-0001} ======================================================== A 91‐year‐old gentleman with a medical history of atrial fibrillation on warfarin, brought to the hospital for a 1‐week history of worsening right thigh pain, radiating to his right lower leg with loss of ambulation secondary to his increasing weakness. He denies any history of falls/trauma. On examination, he was hemodynamically stable. Neurological examination was significant for wasting of the right quadriceps, reduced power (1/5) at right hip flexors, reduced sensations at right L2‐4 distribution, and right patellar hyporeflexia. Initial investigations showed a supra‐therapeutic INR at 5.31. A contrast‐enhanced CT scan showed a 4.3 × 4.2 × 7.5 cm right psoas hematoma with no evidence of active bleeding (see Figure [1](#ccr31665-fig-0001){ref-type="fig"}). His neurological deficits were attributed to mechanical compression of the right upper lumbar plexus. Therefore, warfarin was initially reversed with vitamin K and he underwent an unsuccessful CT‐guided hematoma drainage. A decision was made for conservative management given his stable hemodynamic status without worsening symptoms or radiographical findings. After undergoing extensive physical therapy, he began to ambulate and recover slowly. {#ccr31665-fig-0001} Spontaneous iliopsoas hematoma is a rare complication of warfarin therapy.[1](#ccr31665-bib-0001){ref-type="ref"} The management depends on the patient\'s hemodynamic status, comorbidities, and the presence of active bleeding.[1](#ccr31665-bib-0001){ref-type="ref"} Treatment strategies could include surgical decompression, intervention radiology, and/or conservative approaches with cessation/reversal of warfarin along with physical therapy.[1](#ccr31665-bib-0001){ref-type="ref"}, [2](#ccr31665-bib-0002){ref-type="ref"} Physicians should be aware of such complication in any patients presenting with lower limb symptoms to avoid catastrophic permanent neurological deficits. CONFLICT OF INTEREST {#ccr31665-sec-0002} ==================== None declared. AUTHORSHIP {#ccr31665-sec-0003} ========== BK: designed, planned, wrote the manuscript, and did the literature review. MA:‎ designed, planned, and revised the manuscript. DM: designed, planned, and revised the manuscript. RN: designed, planned, and revised the manuscript. GB: designed, planned, and revised the manuscript.

Digital neuropathology: Is it really needed? {#s1} ============================================ Digital technology is progressively changing our vision of the practice of neuropathology. The usefulness of digital technology in medicine has already been well demonstrated through consensus reviews, quality assurance, tissue microarrays, education, and proficiency testing. It is foreseeable that neuropathology departments will progressively incorporate this technology into their routine practice. The greatest potential of digital neuropathology will be achieved when neuropathologists decide to make substantial changes in the way they work by introducing digital technology. Certainly, implementation of quantitative image analysis will permit professionals to obtain more comprehensive and precise neuropathological reports. Digital neuropathology already plays an important role in teaching and research, and its uses are growing exponentially \[[@b1]\]. Digital neuropathology is already commonly used in the research field and greatly facilitates interactions among researchers, whereas in teaching it provides an excellent tool for medical students to integrate neuropathology with clinical medicine \[[@b2]\]. The interactive digital approach to studying cases appears to be particularly valuable in neuropathology teaching \[[@b3]\]. However, the implementation of digital neuropathology in clinical practice is just starting, and a great deal of skepticism seems to preclude its introduction in routine diagnostic work. For those who have experience in large-scale pathological diagnoses, the transition from conventional light microscopy to a digital-based workflow for imaging potentially offers improved efficiency and related operational cost savings \[[@b4]\]. In the coming years, the application of image analysis in neuropathology is likely to continue to increase, making this a critical area of development for our specialty. Reasons for a network in neuropathology {#s2} ======================================= Frequently, hospitals in a variety of European countries have neither specialized personnel nor adequate technology to confront clinical neuropathology challenges. Additionally, the problem may be compounded by a low neuropathology case workload and the reluctance of general pathologists to get involved. Other facts that support the introduction of digital neuropathology are the following: Firstly, digital neuropathology permits professionals to properly address the quantitative aspects of diagnosis \[[@b5]\]. Secondly, digital neuropathology makes tissue bank specimens widely accessible to neuropathologists. This is particularly relevant in the field of neurodegeneration. Now, digital slides offer a better selection and classification of tissue, since complete digital slides offer more information than static pictures. Thirdly, digital neuropathology increases the ability to engage and collaborate with experts. Neuropathologists often feel the need to discuss difficult cases with their colleagues, but whereas the traditional system of sending slides is a very slow and laborious procedure, consultation is greatly enhanced by digital neuropathology. Fourthly, digital neuropathology permits professionals to share limited biopsies that can now be submitted as a single slide for scanning and can be presented to all participants in a networking approach. Currently, this technological revolution allows a number of cases to be available for viewing and interpretation by any neuropathologist in the network, regardless of location. Is a digital neuropathology network a good solution? {#s3} ==================================================== With the development of whole-slide imaging (WSI) systems, which allow for the evaluation and interpretation of digital images of entire histologic sections \[[@b7]\], the difficulties in implementing a neuropathology network have been solved. This technology is much more accessible and popular now than ever \[[@b8]\]. Nowadays, the WSI slide is easy to handle. One of the more relevant difficulties has been image standardization. Relevant efforts have been made to solve this problem. Currently, the diagnostic information contained in a digital slide obtained from a given scanner may not be translatable to another scanner. Furthermore, image analysis algorithms may need to be modified substantially to work on images created by different scanner types. To avoid this problem, an open digital image communication protocol defined by the Digital Imaging and Communications in Medicine (DICOM) standard is already a reality. Therefore, the best solution would be to adapt scanners to use this standard for the sake of international consultancy. DICOM has been recommended by the international guidelines of Telepathology and Digital Pathology (Digital Pathology Association, Canadian Association of Pathologists, and Spanish Society of Pathologists). In 2015, DICOM compliance was already announced by at least two scanning companies. It is true that this initial compliance will be limited (the file format and compression ratio may vary from vendor to vendor), but in order to avoid the current limitations of DICOM in pathology, a universal viewer, accepting most available file formats existing today, will be selected for the neuropathology platform. Recently, other technologic improvements have been carried out to reduce timing consumed in telepathology. Firstly, currently scanners are becoming very fast. Now it is possible to perform whole slide scans in less than 1 minute \[[@b5]\] and, consequently, it is possible to scan numerous slides through digital scanners in an acceptable time. Scanning time and quality are issues that need to be addressed and handled with care. Following College of American Pathologists' (CAP) recommendations, the neuropathology network will begin with a validation study of 60 cases, plus 20 cases for each different scanner participating in the network, in order to confirm the diagnosis performance. Secondly, the time invested in sending images is no longer a problem \[[@b10]\]. WSI allows the transmission of data to neuropathologists around the world within a few minutes after a slide is created. This technology allows rapid second opinions and consultations on challenging diagnostic cases. Thirdly, it is possible to share images with a large number of colleagues at the same time \[[@b9]\]. By using computer conferencing tools, images can be viewed by multiple parties simultaneously, similarly to what happens with a multiheaded microscope. Therefore, a WSI on a single server is accessed by multiple users who can dynamically pass control over the slides among themselves and can see digital markings and annotations added to it. Fourthly, it seems possible to go further in the diagnostic field. In fact, it has been demonstrated that these technologies are suitable for performing pathological diagnoses \[[@b6]\]. The WSI system has been validated for primary diagnosis in surgical pathology \[[@b11]\] and, in that regard, the American College of Pathology has elaborated a practical guide for validating WSI systems for diagnostic work. The WSI for primary diagnosis has received approval by the European Union as well \[[@b12]\]. In fact, the United States is so far the only industrialized Western country to not have approved whole-slide imaging for primary diagnosis. Fifthly, the storage amount needed in a digital neuropathology system server to allocate the daily work exceeds the capacity of the current servers in most hospitals. Why establish a neuropathology network in the European context? {#s4} =============================================================== Europe is the expected geographic context for communication and relationship among European neuropathologists, and clinical consultancy should be reinforced as much as possible in this context. Considered as a whole, neuropathology is a very suitable discipline for the launching of convergent cross-border European initiatives related to diagnosis, research, and teaching. There are already relevant European initiatives in the field of telepathology networking. For example, *Euro-telepath* has the goal of developing standards for digital pathology \[[@b13], [@b14], [@b15]\] and Academia and Industry Collaboration for Digital Pathology (AIDPATH), which was created in 2013, has the objective of exploiting emerging digital pathology technologies, including in universities and industry. In addition, a European diagnostic nephropathology network has been established \[[@b16], [@b17]\]. According to its developers, the use of WSI brings advantages such as efficiency, facilitation of pathology review in a clinical trial setting, improved intraobserver and interobserver reproducibility, and web-based consensus meetings. Neuropathology networking should evolve to create a large-scale consortium of interconnected facilities from different hospitals and institutions across Europe: some of them would mostly send cases for consultancy whereas some others would mostly receive cases for diagnosis. With the progressive availability of scanners, it is likely that the development of this transnational network will soon become a reality. Points of discussion in relation to a transnational consultancy network in neuropathology. Its potential host: who and where? {#s5} ============================================================================================================================= Several limitations should be considered in the implementation of a digital neuropathology consultancy network. Firstly, which institutions should start and support this initiative? Certainly, it should be a European organ concerned with neuropathology and endowed with appropriate personnel and financial resources. In that regard, it is possible that different financing options may arise, as for example company sponsorship and research project funding. Financial support coming from Europe would be welcomed. Secondly, is this initiative financially sustainable? The needs of personnel and infrastructure should be kept to a minimum. It should be clear that we are not proposing the implementation of hospital digital neuropathology systems, but the EURO-CNS organization only need to connect neuropathologists or institutions already using digital systems. Incidentally, it has been estimated that the implementation of digital pathology could result in cost savings \[[@b4]\]. However, estimations about how much financial support would be needed are essential and should be considered before launching this digital neuropathology initiative. Certainly, operational costs for digital pathology are high. This project may benefit, at least initially, from the efforts already made by many pathology departments to implement digital pathology scanning solutions. This means that in order to create the European Network of Neuropathology, no initial investment in scanners would be needed. Digital slide storage costs will only become significant if thousands of slides are exchanged using this network, and this will mean that the project has already succeeded, and in this case, different financing options may arise (e.g., company sponsorship, case fee, research project funding). Technical support is also a very important aspect that must be considered. Several easy-to-use digital slide platforms exists, and in case limited funds are available, a digital network could start with a low cost license (3,000 Euro/year) of an internet-based digital slide solutions, that allows digital slide format from most vendors. This annual fee will include all needed technical support. In the first phase of the project, some other operational costs may be even higher than those directly related to digital slides, like user support or case editing and formatting. These costs can also be significantly reduced if the adequate platform, including chat or remote control options, is selected. On the other hand, how should consultants be compensated for their work and the time they will spend solving the diagnostic problems of their colleagues? It seems reasonable that the honor of being a consultant should be complemented with some kind of financial reward. It is likely that, once the system becomes active, there will be an increasing demand. The latter, eventually, might translate into a source of income for the EURO-CNS. Certainly, a large-scale pan-European implementation of digital neuropathology does not have to be an expensive question of creating an attractive system and fomenting interest in the initiative among all those involved. Thirdly, legal, privacy, security, and confidentiality issues must be considered and solved before launching the network \[[@b13]\]. There is no specific European Union (EU) legislation on telemedicine. In Europe, Member States are primarily responsible for the organization, financing, and delivery of healthcare. Currently, does not exist an EU legislation specifically on telemedicine. Telepathology as an aspect of telemedicine falls within the scope of Directive 2011/24/EU on the application of patients' rights in cross-border healthcare. Moreover, the European Commission (EC) Commission Staff Working Document (06.12.2012) was written to develop the existing EU legal framework applicable to telemedicine. In order to provide telemedicine cross-border within the EU, healthcare professionals have to look for the responses to the licensing, data protection, reimbursement, and liability. According to several directives of the EU, in most Member States, the competence to accredit professionals wishing to deliver health services is delegated to an appointed licensing or registration body. The telemedicine provider should comply with the authorization and registration requirements of his or her Member State. If the healthcare professional complies with the legislation applicable to the taking up and provision of an information society service in his or her Member State of establishment, he will in principle be free to provide these services in other Member States. Upon being licensed/registered, the health professional will have to abide by the rules and regulations established by the licensing authority and be subject to disciplinary sanctions in case of nonobservance. Health data are very sensitive. The EU adopted the e-Privacy Directive 2002/58/EC, which was aimed at ensuring the protection of personal data in the field of telecommunications. The EU Data Protection Directive 1995/46/EC is applicable to the automated processing of personal data. Obviously, it is need the professional secrecy and explicit consent from the patient. It is an obligation of the data controller to implement appropriate security measures to protect personal data. The data controller should disclose the purposes for which the data are intended. The data controller is also obliged to implement adequate technical and organizational measures against unlawful access, accidental loss, destruction and alteration of data. With respect to reimbursement, it is up to the Member States to decide whether telemedicine must be reimbursed. According to the mentioned EC Commission Staff Working Document, telemedicine services or patients receiving healthcare in another Member State have to be reimbursed. No EU legislation about medical liability exists. Medical liability is regulated at the Member State level, and the complexity and diversity of liability rules in the Member States regarding the provision of healthcare are considerable. Fourthly, when paraffin blocks are not available, the consultant inform know the case sender as to which new studies should be performed on the paraffin bocks, and then the new scanned slides may be sent to the consultant. Alternatively, the consultant could receive the paraffin blocks to further study the case. Finally, we think that all these shortcuts can be solved and the network can be successfully developed. To that end, a preliminary study by a group of institutions with WSI scanners has already been started. We have a historical opportunity to create a European network in neuropathology. Thanks to digital technology, neuropathologists have to confront challenging clinical cases in isolation. On the contrary, they may be interconnected by a European neuropathology organization *with a true interest in this matter. Inasmuch as we are convinced of the advantages of the digital revolution in medicine, a reliable European digital neuropathology network should be created for the patients' benefit.* Acknowledgment {#s6} ============== We thank Dr. Aurelio Ariza, Department of Pathology, Hospital Germans Trias i Pujol, Autonomous University of Barcelona, Badalona, Barcelona, for his helpful suggestions. Conflict of interest {#s7} ==================== The authors declare that they have no conflicts of interest.

**Question:** Can we treat neonatal hypoxic-ischemic encephalopathy? **Finding & Meaning:** Both HIF-1α and HIF-2α may promote cell survival in animal model of neonatal hypoxia-ischemia. Developmental age, degree and type of injury are all factors in the expression and action of HIF-1α and HIF-2α. Early HIF-2α upregulation precedes apoptotic cell death and may limit necrotic cell death in animal model of neonatal hypoxia-ischemia. **Graphical Abstract** Mechanisms of cell survival mediated by HIF Introduction ============ Neonatal hypoxic-ischemic encephalopathy causes death in twenty percent of affected infants, and 25% of survivors have permanent neurologic deficits such as developmental delay, cerebral palsy, cognitive dysfunction, hearing loss, and visual loss \[[@b1-kjp-2019-00850]\]. Ten percent of moderate encephalopathy infants die and 30% of the survivors are handicapped; 60% of severe encephalopathy infants die and most of the survivors have permanent disabilities \[[@b2-kjp-2019-00850]\]. Treatment of neonatal hypoxic-ischemic encephalopathy is limited to therapeutic hypothermia for those infants that qualify but does not offer complete protection. Several neuroprotective strategies have been suggested such as antioxidants or erythropoietin (EPO) \[[@b3-kjp-2019-00850]\]. We have previously reported that hypoxia-inducible factor (HIF)-1α deficient mice have increased brain injury after neonatal hypoxia-ischemia (HI) \[[@b4-kjp-2019-00850]\]. HIF-1α is a nuclear protein induced by hypoxia, which acts as a transcription factor crucial for activation of many protective hypoxia-inducible genes such as *erythropoietin*, *glucose transporters*, *nducible nitric oxide synthas* (*NOS*), and *vascular endothelial growth factor* (*VEGF*) \[[@b5-kjp-2019-00850],[@b6-kjp-2019-00850]\]. As a result, HIF-1α promotes cell survival, and initiates neuroprotection, angiogenesis and subsequently, cell repair in neonatalHI \[[@b4-kjp-2019-00850],[@b7-kjp-2019-00850]\]. We also have found that copper-zinc superoxide dismutase (SOD)1 overexpression is not beneficial to the neonatal mouse brain with HI injury \[[@b8-kjp-2019-00850]\], unlike the adult brain with similar injury, in which SOD1 overexpression was shown to be neuroprotective after cerebral ischemia \[[@b9-kjp-2019-00850]\]. SOD1 is an endogenous antioxidant, which converts superoxide anion (O~2~^-^) to hydrogen peroxide (H~2~O~2~) in neurons of the HI brain. Glutathione peroxidase (GPx) converts the accumulated H~2~O~2~ to H2O and O2. However, SOD1 overexpression causes high H~2~O~2~ accumulation after HI in the neonatal brain, but not the adult \[[@b10-kjp-2019-00850]\]. The accumulation of H~2~O~2~ and hydroxyl radical (OH^-^) causes nitric oxide radical production by *NOS*, lipid peroxidation, protein oxidation, DNA damage, and finally neuronal death via necrosis or apoptosis \[[@b11-kjp-2019-00850]\]. These data illustrate that neuroprotective mechanisms of the developing neonatal brain are different from mature brain. We explored whether the expression of HIF-1α and HIF-2α were altered in SOD1 overexpressing mouse brain compared to wild-type littermates to explain this increased injury in neonatal brain. Methods ======= This animal study was approved by the University of California San Francisco (UCSF) institutional animal care and use committee in accordance with National Institutes of Health (NIH) guidelines for the care and use of laboratory animals. C57Bl/6 mice and SOD1 overexpressing littermates were subjected to the Vannucci model of neonatal HI \[[@b12-kjp-2019-00850]-[@b14-kjp-2019-00850]\]. At postnatal day (P) 9, pups were anesthetized with isoflurane (3% isoflurane/balance oxygen), the left common carotid artery was dissected after a midline incision of the neck and the artery was ligated with electrical coagulation. Animals were allowed to recover for 1 hour with the dam and then exposed to 50 minutes of hypoxia in a humidified chamber at 36.5°C with 10% oxygen/balance nitrogen. Sham-operated control animals received isoflurane anesthesia and exposure of the left common carotid artery without ligation or hypoxia. Pups were sacrificed at designated times after injury (30 minutes, 4 hours, or 24 hours) by decapitation after intraperitoneal injection of Euthasol. Brains cortices were dissected on a cold pad, placed in labeled prechilled microfuge tubes, flash frozen in methylbutane on dry ice, then stored at -80°C until use. Animals used in each group were 10, 11, 15, 13, 7, 13, 10, and 13, in wild-type sham, SOD1 overexpressor sham, wild-type 30 minutes, SOD1 overexpressor 30 minutes, wild-type 4 hours, SOD1 overexpressor 4 hours, wild-type 24 hours, and SOD1 overexpressor 24 hours after HI, respectively. Western blot analysis was performed on ipsilateral cortex from wild-type and SOD1 overexpressing mice to quantify HIF-1α, HIF-2α, spectrin 120 as a neuronal apoptosis marker, and spectrin 145/ 150 as a neuronal necrosis marker at 30 minutes, 4 hours, and 24 hours after HI injury. Frozen cortices were homogenized and cytoplasmic and nuclear proteins were extracted with NE-PER (nuclear and cytoplasmic extraction reagents, Pierce, Rockford, IL, USA). Protein samples were prepared with 4X loading buffer and sample reducing agent (10X DTT) and 40-μg cytoplasmic protein or 20-μg nuclear protein underwent sodium dodecyl sulfate-polyacrylamide gel electrophoresis at 150 V for 90 minutes. Proteins were transferred to polyvinyl difluoride membranes (Bio-Rad, Hercules, CA, USA) at 35 V overnight at 4°C as described by Chang et al.15) The membranes were blocked in 5% nonfat dry milk in Tris-buffered saline with Tween for at least 1 hour at room temperature with gentle shaking, then incubated with rabbit HIF-1α polyclonal antibody (1:1500; Novus Biologicals, Littleton, CO, USA), rabbit HIF-2α polyclonal antibody (1:500; Novus Biologicals), and rabbit histone monoclonal antibody (1: 8000; Cell Signaling Technology, Boston, MA, USA) or mouse spectrin monoclonal antibody (1:2000; Millipore, Temecula, CA, USA) and mouse beta-actin monoclonal antibody (1:2000; Santa Cruz Biotechnology, Santa Cruz, CA, USA) overnight at 4°C. horseradish peroxidase-conjugated goat anti-rabbit antibodies (Santa Cruz Biotechnology) were used (1:2000 for HIF-1α and HIF-2α, 1:4,000 for histone) as secondary antibodies. Bound antibodies were developed by enhanced chemiluminescence (ECL, Amersham Life Science, Arlington Heights, IL, USA) and exposed to film. Image J software (NIH, Bethesda, MD, USA) was used to measure the optical densities (OD) and areas of protein signal on after scanning. Data were analyzed by nonparametric tests: the Mann-Whitney *U* test for 2 groups, and the analysis of variance with Kruskal-Wallis test for 3 or more groups. All data were analyzed using Prism 7 (GraphPad Software, La Jolla, CA, USA). Data are given as mean± standard error, and *P* values \<0.05 were considered significant. Results ======= HIF-1α protein expression was not significantly altered in wild-type or SOD1 overexpressing mouse cortex after HI injury ([Fig. 1A](#f1-kjp-2019-00850){ref-type="fig"}). However, HIF-2α protein expression increased in both wild-type and SOD1 overexpressing mouse cortex, reaching peak levels of 7-fold for wild-type and 5-fold for SOD1 overexpressors at 30 minutes after HI injury ([Fig. 1B](#f1-kjp-2019-00850){ref-type="fig"}). The level of HIF-2α decreased at 4 hours and returned to baseline at 24 hours ([Fig. 1B](#f1-kjp-2019-00850){ref-type="fig"}). There were no differences between SOD1 overexpressors and wild-type littermates by time point. Spectrin 145/150 protein expression was not significantly altered in wild-type or SOD1 overexpressing mouse cortex after HI injury ([Fig. 2A](#f2-kjp-2019-00850){ref-type="fig"}). However, spectrin 120 protein expression increased in both wild-type and SOD1 overexpressing mouse cortex, reaching peak levels of 3-fold for both wild-type and SOD1 overexpressors 4 hours after HI injury ([Fig. 2B](#f2-kjp-2019-00850){ref-type="fig"}). Discussion ========== The developing neonatal brain, with a high rate of oxygen consumption, high concentrations of unsaturated fatty acids and relative lack of antioxidants is extremely vulnerable to oxidative stress and exposure to exogenous free radicals and iron-mediated free radical injury \[[@b16-kjp-2019-00850]\]. Antioxidants protect the brain by attacking various injury cascades. These are reactive oxygen species scavengers (melatonin, SOD, catalase), inhibitors of lipid peroxidation, EPO (as a free radical reducer), inhibitors of *NOS* and anti-inflammatories \[[@b17-kjp-2019-00850]\]. The mechanism of neuroprotection by an iron chelator such as deferoxamine is that it activates and upregulates HIF-1α expression. The upregulated HIF-1α in turn activates various HIF-1α mediated genes, such as *VEGF*, which initiate a complicated neuroprotective cascade \[[@b18-kjp-2019-00850],[@b19-kjp-2019-00850]\]. HIF-1α, a subunit of HIF-1, is undetectable under normoxic conditions, which is induced by hypoxia and is promptly degraded under normoxia by the ubiquitin-proteasome pathway \[[@b20-kjp-2019-00850]\]. Like HIF-1α, HIF-2α is induced by hypoxia. It is a less well-characterized factor, but has been shown to be expressed much more in astrocytes than neurons \[[@b21-kjp-2019-00850]\] and is the primary mediator of EPO expression induced by hypoxia, which is in turn protective to neurons \[[@b22-kjp-2019-00850]\], suggesting HIF-2α expression may play a bigger role in neuroprotection than HIF-1α. The actions of these 2 HIF isoforms are cell-specific, regional and developmentally regulated, being complementary rather than redundant \[[@b23-kjp-2019-00850]\]. Previously, in P7 mouse cortex, we have seen increased HIF-1α immediately after hypoxia alone and 24 hours after HI \[[@b4-kjp-2019-00850]\]. Similarly, also in P7 mouse cortex, we saw increased HIF-1α immediately after hypoxia as well as at 4 hours and 24 hours after hypoxia \[[@b24-kjp-2019-00850]\]. However, HIF-1α increased only immediately after HI, not at 4 hours or 24 hours after HI \[[@b24-kjp-2019-00850]\]. On the contrary, HIF-1α decreased 4 hours after HI in P9 mouse cortex with no change in the hippocampus \[[@b25-kjp-2019-00850]\]. We also measured HIF-2α in the latter study. HIF-2α did not change after HI in cortex, but declined in hippocampus at 30 minutes, 4 hours, and 24 hours. These results corresponded with increased spectrin 145/150 at 4 hours and 24 hours in cortex, but not hippocampus, and with increased spectrin 120 at 24 hours in cortex and at 4 hours in hippocampus \[[@b25-kjp-2019-00850]\]. suggesting early and prolonged necrotic cell death with additional late stage apoptosis in the cortex, but early apoptotic cell death with a relative lack of necrotic cell death in the hippocampus. Thus, while we did not see significant changes in HIF-1α in the present study, we did see a strong increase in HIF-2α in both wild-type and SOD1 overexpressing cortex after HI. This may be due, in part, to the above-mentioned cell-specific, regional and complementary actions of the HIF isoforms. We measured spectrin here also in order to determine if mechanism of cell death corresponded to levels of HIF-1α or HIF-2α. Since spectrin 145/150 was not elevated, but spectrin 120 was (at 4 hours), this may indicate that apoptosis has a greater role than necrosis in genesis of the HI injury in the critical phase several hours after the insult. The expression of HIF-2α may play a bigger role in neonatal HI injury than HIF-1α. It is conceivable that HIF-2α stimulated production of factors (such as EPO) that protected neurons from necrotic injury while astrocytes (and perhaps neurons) underwent subsequent apoptotic death. While we did not analyze cell type, neurons and astrocytes may have different responses to injury that are reflected in gene expression at the time points studied. The SOD1 overexpressing cortex behaved quite similarly to the wild-type cortex with regard to expression of HIF-1α, HIF-2α, spectrin 145/150, and spectrin 120 at the time points after HI analyzed here. Thus, alterations in expression of these factors do not explainthe increased injury we have previously seen in the SOD1 overexpressing brain compared to wild-type. HIF-2α activates transcription of mitochondrial manganese SOD (SOD2) \[[@b26-kjp-2019-00850]\]. The early increase in HIF-2α seen in both wild-type and SOD1 overexpressing cortex may indicate additional SOD production; a factor that warrants further study. The activation of HIF-1α and the consequence of the activation of its target genes are suggested to be necessary for protection in neonatal HI \[[@b4-kjp-2019-00850],[@b11-kjp-2019-00850]\]. During neonatal brain injury, excitotoxicity (excessive activation of glutamate neurotransmission, oxidative stress, and inflammation) contributes to cell death either by apoptosis or necrosis. Apoptosis, or programmed cell death, seems to play a more important role than necrosis during HI injury \[[@b16-kjp-2019-00850]\]. The role of HIF-2α in neonatal HI brain is relatively unexplored. HIF-2α is activated by hypoxia stimulated mitogen-activated protein kinase (MAPK)/ERK signaling pathway, leading to the increased transcription of various proliferation factors. HIF-2α directly binds the MAPK3 promoter and activates its function through the interaction of the transcription factor and its regulated genes \[[@b27-kjp-2019-00850]\]. In cardiac myocytes in adult mice, myocyte-specific HIF-2α induced the epithelial growth factor amphiregulin (AREG), activating AREG signaling, and finally protecting the myocardium from myocardial ischemia-reperfusion injury \[[@b28-kjp-2019-00850]\]. In the carotid body, intermittent hypoxia leads to increased expression of HIF-1α and decreased levels of HIF-2α along with decreased SOD2 expression \[[@b29-kjp-2019-00850]\]. In mesenchymal stem cells, HIF-2α maintained significant long-term effects, and was more stable in hypoxic conditions than normoxic \[[@b27-kjp-2019-00850]\]. In the brain, HIF-2α is considered to be astrocyte specific, and astrocytes are important for the clearance of H~2~O~2~, resulting in neuroprotection. Wildtype brain with decreased expression of HIF-2α had more severe brain injury compared to GPx overexpressing brain with higher HIF-2α. This may be a reflection of a neuroprotective effect of HIF-2α associated with SOD2 generated H~2~O~2~ clearing \[[@b25-kjp-2019-00850]\]. Increased HIF-1α after HI is reported in many cases \[[@b4-kjp-2019-00850],[@b24-kjp-2019-00850],[@b30-kjp-2019-00850]\], however, HIF-1α is not changed or rather decreased in some cases as previously mentioned \[[@b25-kjp-2019-00850]\]. The expressions of HIF-1α are different according to the time after HI, cell types (neurons vs. astrocyte), region (cortex vs. hippocampus), or developmental age (P7 vs. P9). So, these different experimental conditions partially account for these different results. There are several limitations in this study. First, the grade of brain injury with histological analysis was not checked in the present study. Though we have shown injury severity with HI brain injury scores twice in the past \[[@b24-kjp-2019-00850]\], the experimental conditions such as developmental age (P7 vs. P9) were different from the previous study. It is not technically feasible to do histology and protein expression in the same sample. Also, spectrin is an assessment of injury which is not quite comparable to histology. Further work is needed to clarify the grade of brain injury. Second, SOD1 expression of SOD overexpressors was not checked. However, as these mice are well-characterized, and we have done SOD activity in the past \[[@b31-kjp-2019-00850]\], we didn't check SOD1 expression on the transgenics in the present study. Third, we could not find difference of protein expressions after HI between wild-type and SOD1 overexpressing mouse. Some of this might be due to small number of mice underwent this experiment. In summary, we propose that both HIF-1α and HIF-2α may promote cell survival in neonatal HI, but in a cell-specific and regional fashion. We also suggest that early HIF-2α upregulation precedes apoptotic cell death and may limit necrotic cell death. Developmental age, degree and type of injury are all factors in the expression and action of HIF-1α and HIF-2α. The influence of SOD was not clarified and remains an intriguing factor in neonatal HI. Further work is needed to clarify the roles of HIF-1α and HIF-2α in neonatal brain injury. No potential conflict of interest relevant to this article was reported. This work was supported by a grant from Research year of Inje University in 2016 (20150560) (Ga Won Jeon), and funded by NS097299 (Donna M Ferriero). We thank Ethan Lu and Xiangning Jiang, Department of Neurology, UCSF, for helping with the western blot analysis. We also thank Christine Windsor, and Olatz Arteaga Cabeza, Department of Neurology, UCSF, for helping with this animal study. {#f1-kjp-2019-00850} {#f2-kjp-2019-00850}

Abbreviations ============= miRNAmicroRNAPMParticulate matterPM~2.5~Particulate matter with diameter less than 2.5 μmNO~2~Nitrogen dioxideCIConfidence intervalBMIBody mass indexFDRFalse discovery rateqRT-PCRQuantitative real-time polymerase chain reactionRSReporter assayqPqPCRWBWestern blotIPImmunoprecipitationMAMicroarraysPrProteomicsNGSNext generation sequencingIQRInterquartile range*CCND1*Cyclin D1*CDKN1A/B*Cyclin-dependent kinase inhibitor 1A/B*TGF/TGFBR2*Transforming growth factor / Transforming growth factor beta receptor II*STAT3/5*Signal transducer and activator of transcription 3/5*E2F1*E2F transcription factor 1*KIT*v-kit hardy-zuckerman 4 feline sarcoma viral oncogene homolog*HIF1A*Hypoxia inducible factor 1, alpha subunit*PTEN*Phosphatase and tensin homolog*APAF1*Apoptotic peptidase activating factor 1*CDC25A*Cell Division cycle 25A*BCL2*B-cell CLL/lymphoma 2*TLR*Toll-like receptor*TRAF6*Tumor necrosis factor (TNF) receptor-associated factor 6*NF-kB/ NFKB1*Nuclear factor of kappa light polypeptide gene enhancer in B-cells/1*IRAK1*Interleukin-1 receptor-associated kinase 1*ETS1*V-Ets avian erythroblastosis virus E26 oncogene homolog 1*FOS*FBJ Murine osteosarcoma viral oncogene homolog*MMP1*Matrix metallopeptidase 1*FOXO3*Forkhead box O3PI3K/AKTPhosphatidylinositol-3-kinase/Protein kinase B*VEGF*Vascular endothelial growth factor*GAPDH*Glyceraldehyde-3-phosphate dehydrogenase*IPO8*Importine 8*POLR2A*Polymerase (RNA) II (DNA Directed) Polypeptide A*UBC*Ubiquitin C*TERT*Telomerase reverse transcriptase*TERC*Telomerase RNA template Introduction {#s0001} ============ Particulate matter (PM) is an airborne mixture of solid particles and liquid droplets \[[@cit0001]\], of which fine particles with a diameter less than 2.5 µm (PM~2.5~) can be inhaled deeply into the lungs. This leads to the generation of oxidative stress and the induction of inflammation \[[@cit0002],[@cit0003]\]. PM~2.5~ exposure contributes to the initiation and progression of various diseases affecting the respiratory and cardiovascular system \[[@cit0004]\]. According to Barker\'s hypothesis, early life perturbations are crucial for the development of disease later in life \[[@cit0008],[@cit0009]\]. Exposure to ambient PM~2.5~ pollution during pregnancy is significantly associated with increased risk of low birth weight at term in mother-child cohorts of 12 European countries \[[@cit0010]\] and preterm birth in a very large cohort of singleton pregnancies (≥20 weeks of gestation) from three states of the USA \[[@cit0011]\]. Prenatal PM exposure has been shown to affect placental weight \[[@cit0012]\], function and morphology \[[@cit0012],[@cit0013]\], and gene expression \[[@cit0014],[@cit0015]\]. These processes affect fetal programming and could thereby increase the risk of disease later in life \[[@cit0016]\]. MicroRNAs (miRNAs) are single-stranded small non-coding RNAs of approximately 22 nucleotides that play a key role in the regulation of gene expression at the posttranscriptional level in many cellular processes, including proliferation and apoptosis, which could lead to diseases such as cancer \[[@cit0017]\]. Notably, miRNAs are able to regulate up to 30% of the human genome \[[@cit0022]\], where one single miRNA can affect the expression of hundreds of genes, whereas one gene can be targeted by many miRNAs \[[@cit0023]\]. In healthy adults, the blood-leukocyte expression of miR-146a and miR-222 was found inversely associated with air pollution exposure \[[@cit0024]\], while miR-21 and miR-222 expression was significantly increased in steel plant workers after exposure to metal-rich PM \[[@cit0025]\]. Inhalation of ozone was shown to disrupt miRNA expression profiles in human induced-sputum samples and network analysis of the 10 miRNAs with significantly increased expression levels revealed an association with diverse biological processes, including inflammatory and immune response signaling \[[@cit0026]\]. Interestingly, exposure to environmental agents induces altered miRNA expression patterns both in placental cell lines \[[@cit0027]\] and cord blood \[[@cit0028]\], which could potentially contribute to adverse fetal development and health outcomes later in life. Maccani et al. \[[@cit0029]\] showed that maternal smoking during pregnancy was inversely associated with placental expression of miR-16, miR-21 and miR-146a. Therefore, placenta could be used as an appropriate target organ to assess the impact of air pollution on miRNA expression in the early-life environment. To date, the potential modulation of placental miRNA expression in association with prenatal exposure to air pollution has not been investigated. For this purpose, six candidate miRNAs, namely miR-16, -20a, -21, -34a, -146a and -222, related to important cellular processes \[[@cit0030]\] were selected, based on a systematic review \[[@cit0031]\]. miR-16 and miR-21 are involved in cell cycle, proliferation, and apoptosis \[[@cit0029],[@cit0032]\]. miR-146a has been described as a regulator of inflammation \[[@cit0029],[@cit0038]\]. miR-20a, miR-34a, and miR-222 function in angiogenesis \[[@cit0039]\]. Maternal exposure to air pollution has been suggested to adversely affect pregnancy by inducing oxidative stress and inflammation \[[@cit0005]\], which may result in impaired placental angiogenesis \[[@cit0012]\]. In the current study, we investigate whether *in utero* exposure to particulate matter and nitrogen dioxide during different periods of gestation is associated with placental expression of six candidate miRNAs. We hypothesize that *in utero* PM exposure might induce epigenetic alterations at the placental miRNA level. To assess whether any miRNA expression alterations could have a functional effect, we also measured expression of a downstream mRNA target. Results {#s0002} ======= Characteristics of the study population and air pollution exposure {#s0002-0001} ------------------------------------------------------------------ In the present study, 210 mother-newborns pairs with a mean age of 29.5 years (±4.3) and mean pre-gestational BMI of 24.1 kg/m^2^ (±4.8) were included. As shown in [Table 1](#t0001){ref-type="table"}, 70% of women never smoked, 14.8% smoked during pregnancy (current-smokers), and the remaining 15.2% quit smoking at the start of pregnancy (past-smokers). For approximately half of the mothers, the newborn was their first child, and 56.2% of the mothers were highly educated. One hundred and fifteen (54.8%) newborns were girls, had a mean gestational age of 39.2 weeks (±1.3), and an average birth weight of 3,395 g (±427); 190 (90.5%) of the newborns had European-Caucasian ethnicity. The frequencies of conception of pregnancies were approximately equally distributed into the four seasons, with the highest rate (29.1%) observed in summer. The average apparent temperature during the 3rd trimester of pregnancy was divided into quartiles of the distribution. Table 1.Characteristics of mother-newborn pairs. Mean ± SD / Frequency (%)Characteristics*Original study (n = 210)Validation study (n = 181)*Maternal   Age, years29.5 ± 4.329.4 ± 4.2 Pre-gestational BMI, kg/m^2^24.1 ± 4.824.3 ± 5.0 Smoking status    Never-smoker147 (70.0)129 (71.3)  Past-smoker32 (15.2)27 (14.9)  Current- smoker31 (14.8)25 (13.8) Parity    1106 (50.5)89 (49.2)  284 (40.0)74 (40.9)  ≥320 (9.5)18 (9.9) Education    Low23 (10.9)20 (11.1)  Middle69 (32.9)59 (32.6)  High118 (56.2)102 (56.3)Newborn   Gender    Female115 (54.8)102 (56.3) Gestational age, weeks39.2 ± 1.339.1 ± 1.3 Birth weight, g3,395 ± 4273,384 ± 429 Ethnicity    European-Caucasian190 (90.5)163 (90.1)  Non-European20 (9.5)18 (9.9)Other   Apparent Temperature, ^o^C   Third trimester (quartiles)    \< Q153 (25.2)46 (25.4)  ≥ Q1 and \< Q252 (27.8)45 (24.9)  ≥ Q2 and \< Q352 (27.8)44 (24.3)  ≥ Q353 (25.2)46 (25.4) Seasonality (at conception)    Winter57 (27.1)53 (29.3)  Spring50 (23.8)38 (21.0)  Summer61 (29.1)55 (30.4)  Fall42 (20.0)35 (19.3) The mean outdoor exposures to PM~2.5~ and NO~2~ averaged for each of the three trimesters of pregnancy are presented in [Table 2](#t0002){ref-type="table"}. Table 2.Characteristics of particulate air pollution exposure. Averaged for each mother-newborn pair during the different time windows during pregnancy.Air pollutant (µg/m^3^)Mean ± SDIQR10th Percentile90th Percentile*Original study (n = 210)*    PM~2.5~    Trimester 1 (1-13 w)15.99 ± 5.298.0810.2224.65Trimester 2 (14-26 w)16.38 ± 5.068.1910.3923.00Trimester 3 (27-delivery)16.74 ± 5.829.4310.0725.54NO~2~    Trimester 1 (1-13 w)19.97 ± 5.869.0912.7428.06Trimester 2 (14-26 w)20.69 ± 6.047.9512.9829.10Trimester 3 (27-delivery)20.91 ± 6.468.4912.7429.35*Validation study (n = 181)*    PM~2.5~    Trimester 1 (1-13 w)16.12 ± 5.328.1310.2824.87Trimester 2 (14-26 w)16.49 ± 5.078.3310.5923.04Trimester 3 (27-delivery)16.93 ± 5.9810.0710.0825.77NO~2~    Trimester 1 (1-13 w)20.18 ± 5.859.2312.7928.09Trimester 2 (14-26 w)20.85 ± 6.248.6013.0129.72Trimester 3 (27-delivery)20.99 ± 6.648.2812.6929.84 Association of miRNAs with exposure to air pollution {#s0002-0002} ---------------------------------------------------- [Figure 1](#f0001){ref-type="fig"} shows the change in placental miRNA expression in association with exposure across the three trimesters of pregnancy for PM~2.5~ and NO~2~. Figure 1.Associations of relative miRNA expression with *in utero* exposure to air pollution. Associations are presented as percentage changes in relative miRNA (miR-16, miR-20a, miR-21, miR-34a, miR-146a, and miR-222) expression across the three trimesters of pregnancy, for each 5 µg/m^3^ increase in PM~2.5~ exposure (black square) and in NO~2~ exposure (grey triangle). Estimates were adjusted for newborn\'s gender, gestational age (weeks) and ethnicity (European, non-European), maternal age (years), pre-gestational BMI (kg/m^2^), smoking status (never-, past- or current-smoker), educational status (low, middle or high), parity (1, 2, or ≥3), seasonality at conception and apparent temperature (during the third trimester). Asterisk (\*) indicates statistically significant (*P* \< 0.05). PM~2.5~ exposure during the 2nd trimester of gestation was most significantly associated with miRNA expression changes. Placental miR-16 (−24.7%, 95% CI: −44.4 to 2.1, *P* = 0.069), miR-20a (−26.0%, 95% CI: −45.2 to 0.0, *P* = 0.052), miR-21 (−33.7%, 95% CI: −53.2 to −6.2, *P* = 0.022), miR-146a (−30.9%, 95% CI: −48.0 to −8.1, *P* = 0.012), and miR-222 (−25.4%, 95% CI: −43.0 to −2.4, *P* = 0.034) expression were inversely associated with PM~2.5~ exposure during the 2nd trimester of pregnancy. Additionally, miR-146a expression was inversely associated (−21.8%, 95% CI: −39.7 to 1.5, *P* = 0.066) with 3rd trimester exposure. We found positive associations between 1st trimester particulate air pollution exposure and placental expression of miR-20a (+70.9%, 95% CI: 16.7 to 150.3, *P* = 0.007) and miR-21 (+73.7%, 95% CI: 11.7 to 170.1, *P* = 0.015). All estimates were calculated for an increase in PM~2.5~ exposure of 5 µg/m^3^. We obtained similar findings for NO~2~ exposure, miR-20a (−26.2%, 95% CI: −46.0 to 0.9, *P* = 0.058), miR-21 (−31.3%, 95% CI: −51.9 to −1.6, *P* = 0.042), and miR-146a (−23.8%, 95% CI: −43.3 to 2.3, *P* = 0.072) were inversely associated with NO~2~ exposure during the 2nd trimester, while a positive association was observed for the placental expression of miR-21 (+41.4%, 95% CI: −3.4 to 106.9, *P* = 0.076) at term with NO~2~ exposure during the 1st trimester of pregnancy. Estimates were calculated for an increase in NO~2~ exposure of 5 µg/m^3^. miRNA target prediction and pathway analysis {#s0002-0003} -------------------------------------------- A list with putative predicted targets compiled from mirTarBase and DIANA-TarBase with all relevant information about their function and methods used for target validation is provided in [Table 3](#t0003){ref-type="table"}. Table 3.*In silico* putative mRNA targets for placental miRNAs under study. For each miRNA, the mRNA targets (n = 15), description, function and the experimentally validated methods are indicated.miRNAsTarget mRNAsDescriptionFunctionValidated methodsmiR-16CCNE1Cyclin E1Cell cycleRS, WB, qP, NGS^\[b\]^, MA^\[a,b\]^, IP^\[a\]^ BCL2B-cell CLL/lymphoma 2ApoptosisRS, WB, qP, MA, NGS^\[b\]^, IP^\[a\]^ ARL2ADP-ribosylation factor-like 2Cell cycleRS, WB, qP^\[b\]^, MA^\[a,b\]^, IP^\[a\]^ HMGA1High mobility group AT-hook 1Controls many cellular processesRS, WB, qP, NGS^\[b\]^, IP^\[a\]^ CDK6Cyclin-dependent kinase 6Cell cycleRS, WB, qP, NGS^\[b\]^, IP^\[a\]^ CCND1Cyclin D1Cell cycleRS, WB, qP, NGS^\[b\]^, IP^\[a\]^ CCND3Cyclin D3Cell cycleRS, WB, qP^\[b\]^, IP^\[a\]^ CHUKConserved Helix-Loop-Helix Ubiquitous KinaseNF-kappa-B signaling pathwayRS^\[b\]^, WB, qP^\[a,b\]^, IP^\[a\]^ RECKReversion-Inducing-Cysteine-Rich Protein With Kazal MotifsSuppressor of tumorigenicityRS, WB, qP, NGS^\[b\]^, IP^\[a\]^ CAPRIN1Cell cycle associated protein 1Synaptic plasticity in neurons & cell proliferationRS, WB, qP^\[b\]^, IP^\[a\]^ PPM1DProtein phosphatase, Mg^+2^/Mn^+2^dependent, 1DCell cycleRS, WB, qP^\[b\]^, IP^\[a\]^ HMGA2High Mobility Group AT-Hook 2Cell cycleRS, WB, qP^\[b\]^, IP^\[a\]^ FGFR1Fibroblast Growth Factor Receptor 1Controls many cellular processesRS, WB, qP^\[b\]^, IP^\[a\]^ ZYXZyxinSignal transductionRS, WB, qP^\[b\]^, IP^\[a\]^ VEGFAVascular endothelial growth factor AAngiogenesis & endothelial cell growthRS, WB, qP^\[a,b\]^, NGS^\[b\]^miR-20aTGFBR2Transforming growth factor, beta receptor IIControls many cellular processesRS^\[a,b\]^, WB, qP, MA,NGS^\[b\]^, IP^\[a\]^ E2F1E2F transcription factor 1Cell cycle & DNA replicationRS, WB, qP, MA, NGS^\[b\]^, IP^\[a\]^ CDKN1ACyclin-Dependent Kinase Inhibitor 1ACell cycleRS, WB, qP^\[a,b\]^, NGS^\[b\]^, IP^\[a\]^ STAT3Signal Transducer And Activator Of Transcription 3 (Acute-Phase Response Factor)JAK-STAT signaling cascadeRS, WB, qP, MA, NGS^\[b\]^, IP^\[a\]^ LIMK1LIM Domain Kinase 1Regulation of actin filament dynamics & signal transductionRS, WB, qP, MA^\[b\]^, IP^\[a\]^ DUSP2Dual Specificity Phosphatase 2Regulates mitogenic signal transductionRS, WB, qP, NGS^\[b\]^, IP^\[a\]^ BMPR2Bone morphogenetic protein receptor, type II (serine/threonine kinase)Endochondral bone formation & embryogenesisRS, WB, qP, NGS^\[b\]^, IP^\[a\]^ APPAmyloid beta (A4) precursor proteinNeurite growth, neuronal adhesion & axonogenesisRS, WB, qP^\[b\]^, IP^\[a\]^ RUNX1Runt-related transcription factor 1Development of normal hematopoiesisRS, WB, qP^\[b\]^, IP^\[a\]^ MAP3K5Mitogen-Activated Protein Kinase Kinase Kinase 5In cascades of cellular responsesRS, WB, qP^\[b\]^, IP^\[a\]^ HIF1AHypoxia Inducible Factor 1, Alpha SubunitEnergy metabolism, angiogenesis, apoptosisRS, WB, qP, NGS^\[b\]^ BNIP2BCL2/adenovirus E1B 19kDa interacting protein 2Suppression of cell deathWB, qP, NGS^\[b\]^, IP^\[a\]^ CCND1Cyclin D1Cell cycleRS^\[a,b\]^, WB, qP, NGS^\[b\]^ PTENPhosphatase and tensin homologTumor suppressorRS^\[a,b\]^, WB, qP, NGS^\[b\]^ KITv-kit Hardy-Zuckerman 4 feline sarcoma viral oncogene homologProto-oncogeneRS, WB, qP, MA^\[b\]^miR-21BTG2BTG family, member 2Cell cycleRS, WB, qP, MA,NGS^\[b\]^, IP^\[a\]^ PDCD4Programmed cell death 4 (neoplastic transformation inhibitor)Inhibits translation initiationRS, WB, qP, MA,NGS^\[b\]^, IP^\[a\]^ TGFBR2Transforming growth factor, beta receptor IIControls many cellular responsesRS, WB, qP, MA^\[a,b\]^, IP^\[a\]^ NFIBNuclear Factor I/BTranscription & replicationRS, WB, qP^\[b\]^, IP, MA^\[a\]^ CDC25ACell Division Cycle 25ACell cycleRS, qP, NGS^\[b\]^, MA^\[a,b\]^, IP^\[a\]^ RASGRP1RAS guanyl releasing protein 1 (calcium and DAG-regulated)Regulates T- & B-cells developmentRS, WB, qP, NGS^\[b\]^, MA^\[a,b\]^ JAG1Jagged 1Notch signaling- in cell-fate during hematopoiesis, in early & late stages of mammalian cardiovascular developmentRS, WB, NGS^\[b\]^, IP^\[a\]^ APAF1Apoptotic peptidase activating factor 1Activation of CASP3RS, WB, qP^\[b\]^, MA^\[a,b\]^ TIMP3TIMP metallopeptidase inhibitor 3Inhibits matrix metalloproteinasesRS, WB, qP, MA^\[a,b\]^ SOX5SRY (sex determining region Y)-box 5Transcription factor- embryonic development & cell fateRS, WB, qP, MA^\[b\]^ RECKReversion-inducing-cysteine-rich protein with kazal motifsTumor invasion and metastasisRS, WB, qP, MA^\[a,b\]^ PTENPhosphatase and tensin homologTumor suppressorRS, WB, qP, MA^\[a,b\]^ TPM1Tropomyosin 1 (alpha)Ca^+2^ dependent regulation of striated muscle contractionRS, WB, qP, MA^\[b\]^ BCL2B-cell CLL/lymphoma 2ApoptosisRS, WB, qP, NGS^\[b\]^ E2F1E2F transcription factor 1Cell cycle & DNA replicationRS, WB, qP^\[b\]^miR-34aCDK6Cyclin-dependent kinase 6Cell cycleRS, WB, qP, MA^\[a,b\]^, NGS^\[b\]^, IP^\[a\]^ CCNE2Cyclin E2Cell cycleRS, WB, qP, NGS^\[b\]^, MA^\[a,b\]^, IP^\[a\]^ E2F3E2F transcription factor 3Cell cycle & DNA replicationRS, NGS^\[b\]^, WB, qP, MA^\[a,b\]^, IP^\[a\]^ CDK4Cyclin-dependent kinase 4Cell cycleRS, WB, qP^\[a,b\]^, MA^\[b\]^, IP^\[a\]^ NOTCH1Notch 1Variety of developmental processes by controlling cell fate decisions- developmentRS, WB, qP^\[a,b\]^, MA^\[b\]^, IP^\[a\]^ NOTCH2Notch 2Variety of developmental processes by controlling cell fate decisions- developmentRS, WB, qP, MA^\[a,b\]^, IP^\[a\]^ MYCv-myc avian myelocytomatosis viral oncogene homologCell cycle, apoptosis & cellular transformationRS, WB, qP, MA, NGS^\[b\]^ JAG1Jagged 1Notch signaling- in cell-fate during hematopoiesis, in early & late stages of mammalian cardiovascular developmentRS, WB, qP^\[a,b\]^, MA^\[b\]^ CCND1Cyclin D1Cell cycleRS, NGS^\[b\]^, WB, qP^\[a,b\]^, BCL2B-cell CLL/lymphoma 2ApoptosisRS, WB, qP^\[a,b\]^, MA^\[b\]^ MYBV-Myb Avian Myeloblastosis Viral Oncogene HomologHematopoiesis & tumorigenesisRS, MA^\[a,b\]^, WB, qP^\[b\]^ SIRT1Sirtuin 1Coordination of several separated cellular functions such as cell cycle,RS, WB, qP^\[a,b\]^, MA^\[a\]^ HNF4AHepatocyte nuclear factor 4, alphaDevelopment ofRS, WB, MA^\[a,b\]^, qP^\[b\]^ METMET Proto-Oncogene, Receptor Tyrosine KinaseControls many cellular processesRS^\[b\]^, WB, qP, MA^\[a,b\]^ MYCNV-Myc Avian Myelocytomatosis Viral Oncogene Neuroblastoma Derived HomologTranscription factorRS, WB, qP, MA^\[b\]^miR-146aIRAK1Interleukin-1 Receptor-Associated Kinase 1Innate immune responseRS, qP, MA^\[b\]^, WB^\[a,b\]^, IP^\[a\]^ PTGS2Prostaglandin-Endoperoxide Synthase 2 (Prostaglandin G/H Synthase And Cyclooxygenase)Inflammatory prostaglandinsRS, WB, qP^\[b\]^, IP^\[a\]^ STAT1Signal Transducer And Activator Of Transcription 1, 91kDaMediates cellular responses to interferons & cytokinesRS, WB, qP^\[b\]^, IP^\[a\]^ NFKB1Nuclear factor of kappa light polypeptide gene enhancer in B-cells 1Transcription factor-immune responseRS, WB, qP, MA, NGS^\[b\]^ CXCR4Chemokine receptor 4Maintenance of immune functionRS, qP, MA^\[b\]^, WB^\[a,b\]^ SMAD4SMAD family member 4TGF-beta signalingRS, WB, qP^\[a,b\]^, MA^\[b\]^ BRCA1Breast cancer 1, early onsetTumor suppressor & maintains genomic stabilityRS, WB, qP, MA^\[b\]^ EGFREpidermal growth factor receptorControls many cellular responsesRS, qP^\[b\]^, WB^\[a,b\]^ TLR2Toll-Like Receptor 2Innate immune systemRS,WB^\[b\]^, qP^\[a,b\]^ TRAF6TNF receptor-associated factor 6, E3 ubiquitin protein ligaseImmune responseRS, WB, qP, MA^\[b\]^ TLR4Toll-like receptor 4Innate immune systemRS, WB, qP, MA^\[b\]^ CD40LGCD40 LigandImmune systemRS, WB, qP^\[b\]^ CARD10Caspase Recruitment Domain Family, Member 10ApoptosisRS, WB, qP^\[a,b\]^ NUMBNumb Homolog (Drosophila)Cell fates during development & neurogenesisRS, WB, qP^\[b\]^ ELAVL1ELAV Like RNA Binding Protein 1Variety of biological processes & diseasesRS, WB, qP^\[b\]^miR-222CDKN1BCyclin-Dependent Kinase Inhibitor 1BCell cycleRS, WB^\[a,b\]^, qP, MA, NGS^\[b\]^, IP^\[a\]^ FOSFBJ murine osteosarcoma viral oncogene homologSignal transduction, cell proliferation & differentiationRS, WB, qP, NGS^\[b\]^, IP^\[a\]^ TRPS1Trichorhinophalangeal syndrome ITranscriptional repressorRS, WB, qP, MA^\[b\]^, IP^\[a\]^ ETS1V-Ets Avian Erythroblastosis Virus E26 Oncogene Homolog 1Transcription factor in wide variety of different cellular processesRS^\[a,b\]^, WB, qP^\[b\]^, IP^\[a\]^ KITv-kit Hardy-Zuckerman 4 feline sarcoma viral oncogene homologProto-oncogeneRS^\[a,b\]^, WB, qP, MA^\[b\]^ SOD2Superoxide Dismutase 2, mitochondrialBinds to O~2~^−^RS, WB, qP, MA^\[b\]^ MMP1Matrix metallopeptidase 1Cleaves collagensRS, WB, qP, MA^\[b\]^ PTENPhosphatase and tensin homologTumor suppressorRS, WB, qP, MA^\[b\]^ STAT5ASignal transducer and activator of transcription 5ASignal transduction & activation of transcriptionRS, WB, qP^\[b\]^ FOXO3Forkhead box O3Transcription factor for apoptosisRS, WB, qP^\[b\]^ CDKN1CCyclin-dependent kinase inhibitor 1CNegative regulator of cell proliferationRS^\[a,b\]^, WB, qP^\[b\]^ ESR1Estrogen receptor 1Controls many cellular processesRS, WB, qP^\[b\]^ TMED7Transmembrane Emp24 Protein Transport Domain Containing 7Vesicular protein traffickingWB, qP^\[b\]^, IP^\[a\]^ CERS2Ceramide Synthase 2Regulates cell growthRS, WB, qP^\[b\]^ DKK2Dickkopf WNT Signaling Pathway Inhibitor 2In embryonic development & Wnt signalingRS, WB, qP^\[b\]^ Pathway analysis was performed in MetaCore™ by uploading top 15 experimentally validated miRNA targets of miR-20a, miR-21, miR-146a, and miR-222; the 5 most significant enriched pathways for each set of miRNA targets are indicated in [Table 4](#t0004){ref-type="table"}. Within the top significant identified pathways we identified cell cycle- and cancer-related pathways from the putative targets of miR-20a, apoptosis- and cancer-related pathways for miR-21, immune-related pathways for miR-146a, and hematopoiesis-, cell cycle- and immune-related pathways for miR-222. Table 4.Functional enrichment analysis for the putative target genes of deregulated miRNAs in association with air pollution exposure (miR-20a, miR-21, miR-146a, and miR-222). The top 5 enriched MetaCore™ pathways, the *P*-value, FDR (*P*-value corrected for multiple testing) and the genes present in our dataset and involved in the listed pathways are provided.Pathways*P*-value%FDRGenes involved*miR-20a targets*   Upregulation of MITF in melanoma7.24E-071.00E-04*E2F1, c-Kit, CDKN1A (p21), HIF1A*Cell cycle: Regulation of G1/S transition (part 1)9.05E-071.00E-04*Cyclin D1, CDKN1A (p21), Cyclin D, TGF-beta receptor type II*Transcription Androgen Receptor nuclear signaling1.81E-061.34E-04*Cyclin D1, CDKN1A (p21), TGF-beta receptor type II, STAT3*IL-6 signaling in multiple myeloma3.01E-061.56E-04*Cyclin D1, E2F1, CDKN1A (p21), STAT3*Cell cycle: Influence of Ras and Rho proteins on G1/S Transition3.52E-061.56E-04*Cyclin D1, E2F1, CDKN1A (p21), STAT3miR-21 targets*   Development: Regulation of epithelial-to-mesenchymal transition (EMT)9.98E-071.45E-04*Jagged1, TGF-beta receptor type II, Tropomyosin-1, Bcl-2*Apoptosis and survival: p53-dependent apoptosis6.02E-064.36E-04*E2F1, Apaf-1, Bcl-2*Cell cycle:3.21E-049.37E-03*CDC25A, E2F1*Mitogenic action of Estradiol / ESR1 (nuclear) in breast cancer5.75E-049.37E-03*CDC25A, E2F1*Cell cycle: Role of SCF complex in cell cycle regulation6.17E-049.37E-03*CDC25A, E2F1miR-146a targets*   Signal transduction: NF-kB activation pathways4.91E-121.32E-09*TLR2, TRAF6, NF-kB, NF-kB1 (p105), NF-kB1 (p50), IRAK1, TLR4*Immune response: TLR2 and TLR4 signaling pathways1.11E-111.50E-09*TLR2, TRAF6, NF-kB, NF-kB1 (p105), IRAK1, COX-2 (PTGS2), TLR4*Immune response: Bacterial infections in normal airways4.20E-103.76E-08*STAT1, TLR2, TRAF6, NF-kB, IRAK1/2, TLR4*Immune response: HSP60 and HSP70/ TLR signaling pathway7.71E-105.18E-08*TLR2, TRAF6, NF-kB, NF-kB1 (p105), IRAK1/2, TLR4*Immune response: Role of PKR in stress-induced antiviral cell response8.17E-084.40E-06*STAT1, TLR2, TRAF6, NF-kB, TLR4miR-222 targets*   Development: c-Kit ligand signaling pathway during hemopoiesis2.96E-066.13E-04*FOXO3A, c-Kit, CDKN1B (p27KIP1), STAT5*Immune response: MIF-mediated glucocorticoid regulation6.45E-066.68E-04*ETS1, c-Fos, MMP-1*Cell cycle: ESR1 regulation of G1/S transition2.26E-051.20E-03*CDKN1B (p27KIP1), c-Fos, ESR1 (nuclear)*Immune response: Oncostatin M signaling via MAPK in human cells3.20E-051.20E-03*c-Fos, MMP-13, MMP-1*Immune response: IL-7 signaling in T lymphocytes3.47E-051.20E-03*FOXO3A, STAT5A, STAT5* The common putative pathways regulated by the top predicted targets (n = 15) of the significant associated miRNAs with PM exposure were identified using the pathway map tool in MetaCore™. The top 10 significant common pathways for all PM-related miRNAs are illustrated in [Figure 2](#f0002){ref-type="fig"}. Furthermore, a gene network was generated for the predicted miRNA targets ([Figure 3](#f0003){ref-type="fig"}). Figure 2.Common putative pathways regulated by identified targets of the significant miRNAs. The top 10 shared pathways for targets of miR-20a, miR-21, miR-146a, and miR-222 are ranked based on their minimum *P-*value, provided by MetaCore™. Pathways regulated by miR-20a are indicated with orange bars, miR-21 with blue bars, miR-146a with red bars, and miR-222 with green bars. Size of the bars is indicative of the *P-*value for that respective miRNA. Figure 3.Gene network among the putative miRNA targets. A gene network (MetaCore™) was generated for the potential connections of at least two miRNA-targets. The orange rounded rectangle corresponds to miR-20a, blue to miR-21, red to miR-146a, and green to miR-222 targets. The green arrows show activation, the red arrows indicate inhibition, and the grey arrows are unspecified connections. Details for the genes shown in the figure: *AML1* (*RUNX1*): Runt-related transcription factor 1; *APP*: Amyloid beta (A4) precursor protein; *ASK1* (*MAP3K5*): Mitogen-Activated Protein Kinase Kinase Kinase 5; *BTG2*: BTG family, member 2; *CCND*: Cyclin D; *CDC25*: Cell Division Cycle 25; *CDKN1A* (*p21*): Cyclin-Dependent Kinase Inhibitor 1A; *CDKN1B* (*p21KIP1*): Cyclin-dependent kinase inhibitor 1A/B; *ELAVL1*: ELAV Like RNA Binding Protein 1; *ESR*: Estrogen receptor; *FOXO3A* Forkhead box O 3A; *GPCRs* (*CXCR4*): Chemokine receptor 4 (G Protein-Coupled Receptors); *IRAK1/2*: Interleukin-1 Receptor-Associated Kinase 1/2; *c-KIT*: v-kit Hardy-Zuckerman 4 feline sarcoma viral oncogene homolog; *NFKB*: Nuclear factor of kappa light polypeptide gene enhancer in B-cells; *PTEN*: Phosphatase and tensin homolog; *SMAD4*: SMAD family member 4; *STAT1/5*: Signal Transducer And Activator Of Transcription 1/5; *TGFBR2*: Transforming growth factor, beta receptor II; *TPM1*: Tropomyosin; *TRAF6*: TNF receptor-associated factor 6. Validation of miRNA target {#s0002-0004} -------------------------- *PTEN* is a predicted target for three of the four miRNAs significantly associated with PM exposure in placental tissue ([Figure 3](#f0003){ref-type="fig"}). In order to validate this miRNA target, we measured its gene expression (*PTEN*) by means of qRT-PCR in a subset (n = 181) of our study population. As expected, the placental relative *PTEN* expression was inversely correlated with the three miRNA candidates: the Pearson correlation coefficients were −0.18 (*P* = 0.013), −0.27 (*P* = 0.0003), and −0.25 (*P* = 0.0007), for miR-20a, miR-21, and miR-222, respectively. Placental relative *PTEN* expression was strongly and positively associated with third trimester PM~2.5~ exposure (+59.6% per 5 µg/m³ increment, 95% CI: 26.9 to 100.7, *P* \< 0.0001) and borderline significantly associated with third trimester NO~2~ exposure (+25.7% per 5 µg/m³ increment, 95% CI: −0.9 to 59.4, *P* = 0.061). The detailed estimates of the other trimesters for *PTEN* and the three regulatory miRNAs are given in Supplementary Figure S1. Discussion {#s0003} ========== Epigenetic modifications by miRNAs may provide a plausible link between *in utero* exposure to particulate air pollution and alterations in gene expression that might lead to disease phenotypes related to fetal programming \[[@cit0016]\]. The placenta plays a crucial role in transfer of nutrients and oxygen from the mother to the fetus. Therefore, perturbations in the maternal environment can be transmitted to the fetus by changes in placental functions. Since particulate matter exposure has been shown to affect miRNA expression both in animal \[[@cit0043]\] and human studies \[[@cit0044]\], it is perceivable that miRNAs could be involved in regulating the *in utero* response to PM exposure as well. Therefore, we studied the expression of six candidate miRNAs involved in many biological processes (cell proliferation, cell cycle, apoptosis, inflammation, angiogenesis) related to air pollution or environmental stressors exposure \[[@cit0025],[@cit0029],[@cit0031]\]. We found that the placental expression of several candidate miRNAs involved in important biological processes was inversely associated with *in utero* particulate air pollution exposure, mainly during the second trimester of pregnancy. These molecular epidemiological observations might have important health consequences as downregulation of miR-16 and miR-21 has been shown to be significantly associated with fetal growth restriction \[[@cit0037]\], a condition which may result in many complications including preeclampsia \[[@cit0045]\]. Our inverse association between second trimester PM exposure and placental miR-miR-146a expression at birth is consistent with observations in a recent study \[[@cit0046]\] in which miR-146a expression at birth, among other miRNAs, was found to be associated with placental lead levels. Our observations on PM-induced placental miRNA changes in newborns are parallel to observations in placentas of mothers who smoked during pregnancy. Significant lower miRNAs expression of miR-16, miR-21, and miR-146a, were identified in placentas from smoking mothers \[[@cit0029]\]. However, in our study population, smoking had no significant effect on studied miRNA expression (data not shown). Similarly, investigators of the Normative Aging Study showed in 77-year-old men downregulation of blood leukocyte miRNAs expression in miR-21, miR-146a, and miR-222 in association with 7-day average PM and black carbon exposure \[[@cit0024]\]. However, Bollati et al. \[[@cit0025]\] observed significantly higher expression of miR-21 and miR-222 pre- and post-exposure in peripheral blood leukocytes from 63 steel factory workers. The differences between the direction of the effect in which PM exposure on the studied miRNAs between our observations in newborns and the Normative Ageing Study \[[@cit0024]\] might be attributed to the different tissue, duration, magnitude, and/or composition of the PM exposure. Our study investigated the association between placental expression of miRNA at birth in associations with exposure to ambient particulate air pollution for different time windows of gestation. The sensitivity of the epigenetic system to environmental factors occurs primarily during the period of developmental plasticity, as this is the time point when epigenetic marks undergo critical modifications \[[@cit0047]\]. Previously, we reported that exposure to particulate air pollution from fertilization up to and including embryo implantation was associated with lower global DNA methylation levels in placental tissue at birth \[[@cit0048]\]. In the current study, we observed significant associations between lower miRNA expression and PM exposure during the second trimester of pregnancy, which indicates that this is a critical time window for PM-related epigenetic changes at the level of miRNA. For "miR-20a and -21", we found significant higher placental expression at birth in association with prenatal particulate matter exposure during the 1st trimester, in contrast to an inverse association with PM exposure during the second trimester. It is not uncommon that expression changes are specific to the time window of exposure; for example, miR-9 expression decreased upon ethanol exposure in early development stages in mice and fish \[[@cit0049],[@cit0050]\], whereas it increased at later developmental stages in mice and in adult rats \[[@cit0051],[@cit0052]\]. It is conceivable that a similar mechanism of action could regulate the response to air pollution exposure for the different trimesters of pregnancy, as each developmental time window has its own hallmark physiological events, regulated by different molecular processes \[[@cit0053]\]. Recently, miRNA expression has been also shown to be regulated by telomerase reverse transcriptase (*TERT*) at early stages of miRNA biogenesis. Particularly, suppression of *TERT* decreased the levels of miRNA expression in human cells, including miR-20a, miR-21, and miR-222 \[[@cit0054]\]. *TERT* and telomerase RNA component (*TERC*) are essential elements of telomerase, a ribonucleoprotein complex responsible for the telomere elongation \[[@cit0055]\]. Previously, higher maternal residential traffic exposure has been associated with shorter placental telomere length at birth \[[@cit0056]\], which may be linked to decreased telomerase activity. Hence, our observed placental miRNA alterations could be also mediated by decreased levels of telomerase activity caused by maternal PM exposure during pregnancy. To further understand the biological function of these miRNAs, we identified their putative targets and performed overrepresentation enrichment analysis on the experimentally validated targets of significantly associated miRNAs with PM exposure. Targets of miR-20a were found to potentially regulate pathways involved mainly in cell cycle \[*CCND1*, *CDKN1A(p21)*, *TGFBR2*, *STAT3*\] and cancer \[*E2F1*, *KIT*, *CDKN1A(p21)*, *HIF1A*\]. For miR-21, its putative targets regulated pathways in cancer (*PTEN*, *APAF1*), cell cycle (*E2F1*, *CDC25A*), and apoptosis (*APAF1*, *BCL2*, *E2F1*). For miR-146a, immune-related pathways (*TLR2*, *TRAF6*, *NFKB1*, *IRAK1*, *TLR4*) were predominantly regulated. Lastly, for miR-222, pathways involved in immune responses (*ETS1*, *FOS*, *MMP1*, *FOXO3A*, *STAT5*) and hematopoiesis \[*FOXO3A*, *KIT*, *CDKN1B (p27)*, *STAT5*\] were identified. In addition, cell cycle-related pathway was found to be the most significant among the shared regulated pathways by the putative targets of the significant miRNAs (miR-20a, -21, -146a, and -222) associated with air pollution. A common putative target of the miR-20a, miR-21, and miR-222 ([Figure 3](#f0003){ref-type="fig"}), *PTEN*, is involved in many key cellular processes by negatively regulating PI3K/AKT pathway involving cell survival, cell cycle, angiogenesis, and metabolism \[[@cit0057]\]. Interestingly, in a validation experiment, we demonstrated that *PTEN* expression inversely correlated with miR-20a, miR-21, and miR-222 expression in placental tissue. These findings confirm the miRNA-*PTEN* co-expression in placental tissue, which is an important criterion for the validation of miRNA targets \[[@cit0058]\]. The inverse association observed between air pollution exposure and miRNA expression was accompanied by a positive association between air pollution exposure and *PTEN* expression, as expected. During normal pregnancy, placental *PTEN* expression decreases with the development of the placenta and as pregnancy progresses \[[@cit0059]\]. Maccani et al. have reported that downregulation of miR-21 through induction of *PTEN* in placenta could lead to reduced invasion of maternal decidua, migration and growth of placental cells \[[@cit0037]\]. Aberrant expression of immune-related target genes, such as *TLR4*, has been associated with inflammation-induced preterm delivery \[[@cit0060]\], and the activation of NF-κB with increased oxidative stress resulting in pregnancy complications, e.g., preeclampsia \[[@cit0061]\]. Low expression of angiogenesis-related genes (*MMP2*, *VEGF*, *TGF-β*) and high expression of apoptosis-related genes (caspases) have been associated with recurrent pregnancy loss \[[@cit0062]\]. Placental vascular development is a crucial process for fetal development ensuring an optimal blood flow between fetus and mother, and an increased uterine vascular resistance and reduced blood flow have been associated with pregnancy complications and fetal growth retardation \[[@cit0063]\]. In addition, under normal conditions in early pregnancy, genes regulating cell cycle, differentiation, metabolic process, and angiogenesis are overexpressed, whereas genes involving in metabolic process, stress response, signaling and ion transport are upregulated in late pregnancy \[[@cit0064]\]. However, in our study, we only measured the miRNA expression at birth; thereby, the regulation of miRNA targets across the different time windows of pregnancy cannot be assessed. A limitation of this study is that placental tissue is composed of a complex population of cells (syncytiotrophoblasts/cytotrophoblasts, mesenchymal cells, Hofbauer cells, and fibroblasts). To minimize the impact of regional differences we combined 4 fetal samples taken at four standardized sites across the middle region of the placenta (approximately 4 cm away from the umbilical cord) to extract miRNAs. Regardless of this, the placenta might be used as a proxy for epigenetic changes in the fetus, as it is derived from the outer layer of the blastocyst. The organ has a great plasticity to a range of intrauterine conditions and exposures. We cannot answer whether epigenetic alterations in placental tissue affect the fetus in a direct manner or indirectly by adaptations in its function. Secondly, although our results were robust and independent of other studied factors, we cannot eliminate the possibility of residual confounding by some unknown factor that is associated with both miRNA expression and ambient air pollution. Season and apparent temperature were taken into account as epigenetic adaptive changes to season have been reported \[[@cit0065]\]. Our study was not designed to evaluate temporal changes of miRNA expression during pregnancy and may be hampered by the fact that assays of term placentas may not reflect *in vivo* miRNA expression patterns occurring earlier at critical points of development. In conclusion, we observed significant associations between PM exposure and miRNA (miR-20a, miR-21, miR-146a, and miR-222) and mRNA (*PTEN*) expression. The second trimester was identified as the most significantly affected time window for the analyzed miRNAs. The potential regulation of immune-, cell cycle-, and angiogenesis-related pathways could underlie the observed miRNA expression changes due to early life exposure to particulate matter. Materials and methods {#s0004} ===================== Study population {#s0004-0001} ---------------- The protocols of the ENVIR*ON*AGE (ENVIRonmental influence *ON* AGEing) birth cohort are approved by the Ethics Committees of the University of Hasselt and the South-East-Limburg hospital (ZOL). Participating mothers provided written informed consent when they arrived at the hospital for delivery, and completed study questionnaires in the postnatal ward after delivery to provide detailed information on maternal age, pre-gestational BMI, maternal education, occupation, smoking status, alcohol consumption, place of residence, use of medication, parity, and newborn\'s ethnicity. Past-smokers were defined as those who had quit smoking before pregnancy. Smokers continued smoking during pregnancy. Ethnicity was classified based on the native country of the newborn\'s grandparents as European-Caucasian (when two or more grandparents were European) or non-European (when at least three grandparents were of non-European origin). We asked women whether they occasionally consumed alcohol during pregnancy. Maternal education was coded as low (no diploma or primary school), middle (high school), or high (college or university degree). Placental tissue was collected and deep-frozen within 10 minutes after delivery. Four biopsies at the fetal side of placental villous tissue, shielded by the chorio-amniotic membrane, were obtained, preserved in RNA later overnight at 4^o^C, and then stored at -20^o^C. The biopsies were taken at four standardized locations across the middle point of placenta, at approximately 4 cm distance from the umbilical cord. Our study population (n = 210) within ENVIR*ON*AGE cohort was recruited from September 2011 to January 2014. In 210 placentas, miR-16, miR-20a, miR-21, miR-34a, miR-146a, and miR-222, were measured. To validate the miRNAs putative regulatory role on placental *PTEN*, we performed a validation study. The transcript of the placental *PTEN *gene, which is regulated by miR-20a, miR-21, and miR-222, was measured in 181 (86.2%) of the newborns. To clarify the generalizability of the study, we have compared the characteristics of these 210 mother-newborn pairs with the data of the birth register of Flanders (Northern part of Belgium). This register comprises all births from Flanders (n = 648,711) from 1999--2009 \[[@cit0066]\]. The main characteristics including maternal age, maternal education, parity, ethnicity, and birth weight are in line with the birth register of all births between 1999--2009 in the Northern part of Belgium and therefore our sample of mother-newborn pairs can be considered to be representative for the population in Flanders (Supplementary Table S1). Air pollution exposure {#s0004-0002} ---------------------- Air pollution exposure was assessed as described previously \[[@cit0067]\]. In brief, we interpolated the regional background level of PM~2.5~ for each mother\'s residential address using a spatial temporal interpolation method (Kriging) that employs pollution data collected in the official fixed site monitoring network and land cover data retrieved from satellite images (Corine land cover data set) in combination with a dispersion model. The utilized dispersion model was described previously \[[@cit0068],[@cit0069]\]. This model chain provides daily PM~2.5~ values using data from the Belgian telemetric air quality network, combined with information from point sources and line sources which are interpolated to a high resolution receptor grid. In the Flemish region of Belgium, more than 80% of the temporal and spatial variability (R^2^) could be explained by the interpolation tool \[[@cit0070]\]. To explore potentially critical exposures during pregnancy, individual mean PM~2.5~ concentrations (µg/m^3^) were calculated for various periods, for which the date of conception was estimated based on ultrasound data: each of the three trimesters of pregnancy, with trimesters being defined as: 1--13 weeks (1st trimester), 14--26 weeks (2nd trimester), and 27 weeks to delivery (3rd trimester). Additionally, nitrogen dioxide (NO~2~) exposure was interpolated using the same methods as PM~2.5~ exposure. We have complete residential information during and before pregnancy. For those that moved during pregnancy (n = 20, 9.5%), we calculated exposure windows accounting for the address changes during this period. RNA isolation and DNase treatment {#s0004-0003} --------------------------------- Total RNA and miRNA were isolated from pooled biopsies using the miRNeasy mini kit (Qiagen, KJ Venlo, the Netherlands) according to the manufacturer\'s protocol. Quality control of the extracted total RNA and miRNA was assessed by spectrophotometry (Nanodrop ND-1000; Isogen Life Science, De Meern, the Netherlands). Sample purity was assessed by calculating the A~260/280~ and A~260/230~ ratios. The average (±SD) yield of total RNA per placenta biopsy was 4.4 (±1.2) µg with average A~260/280~ ratio of 1.96 (±0.03) and average A~260/230~ ratio of 1.85 (±0.18). DNase treatment was performed on extracted RNA samples according to the manufacturer\'s instructions (Turbo DNA-free kit, Ambion, Life Technologies, Diegem, Belgium). Extracted RNA was stored at −80 °C until further use. In a pilot experiment, the variability within the four individual biopsies was assessed in a subset of ten placental tissues. The average C~q~ values of miRNAs (miR-21, miR-222, and RNU6) within the four biopsies of each placenta varied between 2--9% (CV). To reduce interplacental differences, we used pooled samples from 4 placental biopsies. Reverse transcription and miRNA expression analysis {#s0004-0004} --------------------------------------------------- RNA was reverse transcribed using the Megaplex reverse transcription (RT) stem-loop primer pool A (Applied Biosystems, Foster City, CA), enabling miRNA specific cDNA synthesis of 380 different human miRNAs and small RNA controls, according to the manufacturer\'s protocol. Briefly, 375 ng total RNA was reverse transcribed as follows: 2 minutes at 16°C, 1 minute at 42°C and 1 minute at 50°C, for 40 cycles (Thermocycler PCR, Techne, Staffordshire, UK). Afterwards, cDNA was stored at −20°C for a maximum of one week until qRT-PCR measurements were performed. miRNA qRT-PCR analysis was performed using Taqman miRNA assays (Applied Biosystems, Foster City, CA), according to the manufacturer\'s protocol. All target sequences of the miRNAs and control RNA are available in Supplementary Table S2. An input of 5 ng cDNA was used for PCR reactions, which were run on a 7900HT Fast Real-Time PCR System (Applied Biosystems, Foster City, CA), as follows: a polymerase activation for 2 minutes at 50°C, a denaturation step for 10 min at 95°C and an anneal/extension step (40 cycles) for 15 seconds at 95°C and for 1 min at 60°C. For normalization the endogenous control RNU6 was used. In order to minimize the technical variation between the different runs of the same miRNA assay, inter-run calibrators (IRCs) were applied. Expression of candidate miRNAs was studied and Cq values were collected with SDS 2.3 software. Amplification efficiencies were between 90--115% for all assays. The relative miRNA expression was calculated by 2^−ΔΔCq^ method using qBase plus software (Biogazelle, Belgium). Data is presented as relative quantities of target miRNA normalized to endogenous control miRNA. All samples were analyzed in triplicate. Replicates were included when the ΔC~q~ was smaller than 0.5. miRNA target prediction and pathway analysis {#s0004-0005} -------------------------------------------- Many *in silico* prediction tools have been developed to identify putative miRNA-target genes. We utilized miRTarBase \[[@cit0071]\] and DIANA--TarBase \[[@cit0072]\] for prediction of targets for those miRNAs that revealed significant associations with *in utero* air pollution exposure, namely miR-16, miR-20a, miR-21, miR-34a miR-146a, and miR-222. miRTarBase v6.0 includes many miRNA-target interactions, retrieved manually from research articles in literature related to functional studies of human miRNAs \[[@cit0071]\]. DIANA-TarBase v7.0 identifies miRNA-target interactions which have been highly curated from published experiments \[[@cit0073]\]. The available experimental evidence on prediction of miRNA targets was used as a determinant for the selection of target genes. We considered reporter assay (RS), qPCR (qP), Western blot (WB), and immunoprecipitation (IP) as strong evidenced assays, while assays included high-throughput analyses such as microarrays (MA), proteomics (Pr), and next generation sequencing (NGS) were considered as less strongly evidenced. The identified putative miRNA targets were ranked based on available strong evidenced assays and, subsequently, the top 15 targets were selected for analysis. MetaCore™ (Thomson Reuters, New York, USA) was used for pathway analysis. We performed pathway analysis by overrepresentation analysis for each set (n = 15) of predicted miRNAs target genes of miR-20a, miR-21, miR-146a, and miR-222. MiR-16 and miR-34a were excluded from pathway analysis, as we did not observe significant associations for these miRNAs with *in utero* air pollution exposure. The obtained *P*-values were corrected for multiple hypotheses testing by applying Benjamini and Hochberg\'s FDR \[[@cit0074]\]. Extended lists of enriched pathway maps (n = 50) are provided in Supplementary Table S3. Additionally, we identified the shared pathways regulated by the miRNAs of interest. The list of common pathways (n = 33) regulated by these miRNAs with their *P*-values and FDR is given in Supplementary Table S4. miRNA target validation by qRT-PCR {#s0004-0006} ---------------------------------- The validation of a common miRNA target was performed in a subset (n = 181, 86.2%). Total RNA (3 μg) were reverse transcribed into cDNA by GoScript Reverse Transcription System (Promega, Madison, WI, USA) using Thermal cycler (TC-5000; Techne, Burlington, NJ, USA). The synthetized cDNA was stored at --20°C for further applications. qRT-PCR analysis was performed using a 7900HT Fast Real-Time PCR System (Applied Biosystems, Foster City, CA), according to the manufacturer\'s protocol. *PTEN*, as target gene (primer assay: *Hs.PT.58.4416071*, RefSeq number: *NM_000314*), and *GAPDH* (primer assay: *Hs.PT.53a.24391631.gs*, RefSeq number: *NM_001256799*), *IPO8* (primer assay: *Hs.PT.56a.40532361*, RefSeq number: *NM_001190995*), *UBC* (primer assay: *Hs.PT.39a.22214853*, RefSeq number: *NM_021009*), and *POLR2A* (primer assay: *Hs.PT.56a.25515089*, RefSeq number: *NM_000937*), as reference genes were measured. An input of 6ng of cDNA was added to TaqMan Fast Advanced Master Mix (Life Technologies) and PrimeTimeTM assay (Integrated DNA Technologies, Coralville, IA, USA). The same cycling conditions as previously mentioned were used. Inter-run calibrators and reference genes were used for normalization. The expression of target and reference genes was measured, and the Cq values were collected using SDS 2.3 software. The raw data were processed to normalized relative gene expression by 2^−ΔΔCq^ method using qBase plus software (Biogazelle, Belgium). The amplification efficiencies for all assays were within the acceptable range (90-115%) (data not shown). All samples were analyzed in triplicate and replicates were included when the ΔC~q~ was smaller than 0.5. Statistical analysis {#s0004-0007} -------------------- For database management and statistical analysis, we used SAS software (Version 9.3 SAS Institute, Cary, NC, USA). We tested the normality of the obtained relative quantities of miRNA expression. Because of non-normal distribution the relative miRNA expressions were log-transformed. Categorical data are presented as frequencies (%) or numbers and continuous data as mean (±SD). We performed multiple linear regression to assess the independent associations between placental miRNA expression and *in utero* exposure to particulate air pollution, while adjusting for maternal age (years), pre-gestational body mass index (BMI) (kg/m^2^), smoking status (never-smoker, past-smoker, or current-smoker), educational status (low, middle or high), parity (1, 2, or ≥3), and newborn\'s gender, gestational age (weeks) and ethnicity, seasonality (at conception) and apparent temperature during the 3rd trimester of pregnancy divided into quartiles of the distribution. Using the same model, the association between relative miRNA expression and air pollution exposure was estimated for each trimesters of pregnancy. Likewise, in a subsequent validation experiment, the relative placental *PTEN* expression was first log-transformed and then associated with air pollution using the same multiple regression model, as described in the previous paragraph. Pearson correlations between miRNAs of interest and *PTEN* expression were evaluated. The effect of air pollutants on miRNA/mRNA expression is presented as percentage of change \[change (%) = (10^(β^\*^5)^ -- 1)\*100\] with 95% confidence intervals (CI), for each 5-µg/m^3^ increment in air pollution exposure at each time window. Supplementary Material ====================== ###### KEPI_A\_1155012_s02.zip Disclosure of potential conflicts of interest ============================================= No potential conflicts of interest were disclosed. Acknowledgments =============== The ENVIR*ON*AGE birth cohort is supported by grants from the European Research Council (ERC-2012-StG 310898) and Flemish Research Council (FWO G073315N). Karen Vrijens is a postdoctoral fellow of the FWO (12D7714N). [^1]: Both authors contributed equally to this work. [^2]: Supplemental data for this article can be accessed at <https://doi.org/10.1080/15592294.2016.1155012>.

Introduction {#sec1} ============ The achievement of a complete inventory of the earth\'s biota remains an urgent priority for biodiversity conservation. One of the main challenges is exploring the wilder regions of the world where intact habitats of high conservation value remain unknown. Arid areas are a major terrestrial habitat among these environments ([@bibr42]). In South America, deserts are the largest macro-habitat, covering more than 57.3% of the surface area ([@bibr35]). The dry neotropics support considerable biological diversity, though they have received little attention in comparison with the wet, tropical forests ([@bibr06]). Patagonia is a large xeric biome located in the southern tip of South America, remarkably understudied despite the fact that some of the original components and functions of this arid ecosystem are still preserved. One of the largest conservation units of arid ecosystems in Argentina is the Natural Protected Area Península Valdés, located in the northeastern zone of this biome. Since 1999, this area has been included in the UNESCO World Heritage List. Invertebrates represent an essential part of ecosystems ([@bibr45]) having great abundances and species richness in almost all habitats ([@bibr31]; [@bibr01]; [@bibr14]), occurring at all levels of the food web ([@bibr43]; [@bibr45]; [@bibr01]), and playing vital roles in the structure and fertility of soils, the pollination of flowering plants, nutrient cycling, and in the decomposition of organic material and predation ([@bibr29]; [@bibr03]). Furthermore, arthropods can be used for monitoring environmental changes because of their high species abundances, richness, and habitat fidelity ([@bibr02]). Terrestrial arthropods are even better monitors than vegetation because of their rapid response to habitat changes and the capability of generating a finer environmental classification than vascular plants or vertebrates ([@bibr43]; [@bibr45]; [@bibr01]). In arid regions, invertebrates are the most abundant animals ([@bibr15]; [@bibr03]). In these habitats, arthropods play key roles (principally in and above the soil) as decomposers, herbivores, granivores, and predators, controlling nutrient and energy flow through trophic levels in the food chain ([@bibr15]; [@bibr42]; [@bibr29]; [@bibr03]). Arthropods fill these important functional roles in deserts because they are less constrained by low water availability and extreme thermal environments than other animals ([@bibr48]; [@bibr01]). The arthropod biomass and species diversity is much greater than all other desert animal biomass and diversity combined ([@bibr42]). The aim of this work was to give a preliminary description of the composition and structure of the arthropod community of Península Valdés, using species abundance models, diversity analysis and a trophic guild approach, based on a planned and intensive sampling effort. The purpose is to contribute to a currently limited knowledge of the ground-dwelling arthropod fauna of Patagonia ([@bibr18]; [@bibr22]; [@bibr12]; [@bibr17]; [@bibr41]). Materials and Methods {#sec2} ===================== Ground-dwelling arthropods were sampled using pitfall traps during the summers of 2005, 2006 and 2007. A total of 648 traps, 12 cm in diameter at the opening and 12 cm deep, were placed (216 traps/year). According to previous optimization studies of the pitfall sampling in the area (Cheli, unpublished observations), each trap was filled with 300 ml of a 30% solution of ethylene glycol used as a preservative, and each trap was opened on-site for two weeks in the middle of February. Traps were located at least 20 m apart from each other, covering the main environmental units of Península Valdés ([Figure 1](#f01){ref-type="fig"}). The two main vegetation units of Península Valdés are: (1) shrub steppe with 67%) of total vegetal cover dominated by *Chuquiraga avellanedae* Lorentz (Asterales: Asteraceae), *Condalia microphylla* Cav. (Rosales: Rhamnaceae), *Paronychia chilensis* DC (Caryophyllales: Caryophyllaceae), *Hoffmanseggia trifoliata* Cav. (Fabales: Fabaceae), *Nassella tenuis* (Phil.) Barkworth (Poales: Poaceae), *Achnatherum speciosa* (Trin, & Rupr.) Barkworth (Poaceae), *Poa ligularis*Nees & Steud. (Poaceae); and (2) shrub-grass steppe with 75%\> of total vegetal cover dominated by *C. avellanedae*, *Hyalis argentea* D. Don ex Hook & Arn (Asteraceae), *H. trifoliata*, *P. chilensis*, *S. tenuis*, *Sporobolus rigens* (Trin.) E. Desv. (Poaceae), *Piptochaetium napostaense* (Speg.) Hack. (Poaceae), *Plantago patagonica* Jacq. (Lamiales: Plantaginaceae) ([@bibr05]). {#f01} All specimens were identified to order and family levels. Additionally, in order to have a good estimation of the community structure at the species level, three representative groups with different abundances were chosen: Formicidae (Hymenoptera) (the most abundant taxa), Coleoptera (a medium to high abundance taxon), and Heteroptera (Hemiptera) (low abundance taxa). In those cases where it was not possible to determine individuals at the species level, the individuals were described as morphospecies for further analysis. Voucher specimens were deposited in the entomological collection of Centro Nacional Patagónico (CENPAT-CONICET), Museo de La Plata and IADIZA (CRICYT-CONICET). Araneae were only analyzed to the order level due to the large numbers of juvenile specimens and of individuals whose small size impeded proper determination. The same level of analysis was used for Psocoptera because of the lack of accurate literature and keys. Finally, flying Hymenoptera, Lepidoptera, and the suborder Auchenorrhyncha (Hemiptera) were excluded from analysis because the sampling protocol used for this study was not suited for these groups. Statistical analysis {#sec2a} -------------------- Abundance analysis: Abundance distribution models were used to describe the structure of the community. To choose which model best described the community, a Bayesian selection was performed for four models. Those models increased in their evenness as follows: (a) Dominance pre-emption model, (b) Logarithmic Series, (c) Logarithmic Normal Distribution, and (d) MacArthur\'s Broken Stick model ([@bibr47], [@bibr46]; [@bibr33]). The decision criterion for choosing a model was the lowest value of the Akaike Information Criterion (AIC) ([@bibr27]). The estimation of parameters was calculated by means of Markov Chain Montecarlo ([@bibr27]) using the pymc library for Bayesian estimation for the python programming language ([@bibr24]). Diversity analysis: Diversity was estimated through the Shannon-Wiener index, the Shannon evenness measure, and the richness of families and species ([@bibr39]; [@bibr33]). The Shannon-Wiener diversity index was calculated using natural log, and differences between groups were tested by the Hutchenson method (a modification of the t-test, see [@bibr34]) using Bio∼DAP software. Guild analysis: To indicate the trophic structure of the arthropod community, species were classified into feeding guilds as herbivores, predators, and scavengers (following [@bibr08]; [@bibr40]; [@bibr13]). The relationship among abundance and richness of feeding guilds was analyzed using the *X^2^* test. All α-values for multiple tests were corrected by Bonferroni\'s correction (α\' = α /3 = 0.0167) ([@bibr49]). Results {#sec3} ======= A total of 28, 111 arthropods belonging to 18 orders, 52 families and 160 species/morphospecies were collected. At the order level, Hymenoptera (Formicidae and Mutillidae) represented 83.2% of the total catch, thus there were very low relative abundances of other orders. ###### Arthropod orders and families collected through pitfall trapping in Península Valdés.  Among the Hymenoptera, 99.3% were ants (Formicidae). As a consequence of their colonial behavior, they fall in the traps in large numbers; therefore, the percentages of capture were calculated excluding Formicidae to better describe the dominance relationships between the captured groups. This revealed a shared sub-dominance between Araneae and Coleoptera, followed in magnitude by Orthoptera, Collembola, and Solifuga ([Table 1](#t01){ref-type="table"}, [Figure 2](#f02){ref-type="fig"}). At the family level, the analysis showed a sub-dominance of six families (Sminthuridae, Tenebrionidae, Acrididae, Phloeothripidae, Carabidae, and Mummusidae) which represents more than 60% of the total catch. A complete description of the community at the order and family levels is given in [Table 1](#t01){ref-type="table"}. Among the Formicidae caught, 75.1% belong to the Myrmicinae subfamily with *Pheidole bergi* Mayr and *Solenopsis patagonica* Emery being the most abundant species, representing more than 50% of the total captures ([Figure 3](#f03){ref-type="fig"}). A complete description of the ant assemblage is given in [Table 2](#t02){ref-type="table"}. The most abundant families of beetles were Tenebrionidae and Carabidae, representing more than 75% of the total captures of this group, while the most numerous species were *Blapstinus punctulatus* Solier, *Trirammatus* (Plagioplatys) *vagans* (Dejean) and *Metius malachiticus* Dejean ([Figure 4](#f04){ref-type="fig"}, [Table 3](#t03){ref-type="table"}). With respect to the true bug assemblage, the most numerous families were Oxicarenidae and Blissidae with more than 54% of the total captures of this group. The most abundant species was *Anomaloptera patagonica* Dellapé & Cheli ([Figure 5](#f05){ref-type="fig"}); also found were *Valdesiana curiosa* Carpintero, Dellapé & Cheli (Miridae). Both taxa were very recently described as new based on specimens collected from this study. A complete description of the true bug community can be found in [Table 4](#t04){ref-type="table"}. Abundance analysis: The distribution abundance model which best described the abundance data, both at the family and species levels, was the logarithmic series model (AIC fam: 202.231; AIC sp: 134.32). Also, this model best described the species abundances of ants (AIC: 138.551) and beetles (AIC: 134.318). The true bug species were equally well described both by the log series (AIC: 41.318) as well as the log normal series (AIC: 39.72) ([Table 5](#t05){ref-type="table"}). In addition, excluding ants from the analysis increased the capacity of the logarithmic series model to describe the species abundance distribution of the community (AIC excluding ants: 513.668; AIC including ants: 652.527). Diversity analysis: There was a significant increase of diversity (Shannon-Wiener index) at both the family and species levels when ants were excluded from the analysis (Hutchenson test: for the family level, t\' = 101.494, p \< 0.0001; for the species level, t\' = 39.928, p \< 0.0001) as well as an increase in the evenness of both taxonomical levels. At the species level, beetles were more diverse than ants (Hutchenson test; t\' = 11.995, p \< 0.0001). True bugs were equally as diverse as beetles (Hutchenson test, t\' = 2.249, p = 0.026) and ants (Hutchenson test, t\' = 1.645, p = 0.103). The Shannon species evenness measure was considerably high and similar among the three groups of species ([Table 6](#t06){ref-type="table"}). {#f02} {#f03} {#f04} ###### Abundance of ant species (Hymenoptera-Formicidae) in Península Valdés.  {#f05} ###### Abundance of beetle species (Coleoptera) in Península Valdés.  ###### Abundance of true bugs species (Hemiptera-Heteroptera) in Península Valdés.  ###### Fit to species abundances models (*p* values), Diversity (Shannon-Wiener index) and evenness values to family and species levels.  {#f06} Guild analysis: There was a significant difference among abundances of trophic guilds (*X^2^*~0.05;\ 2~ = 459.75; p \< 0.001). The abundance of predators was greater than herbivores (*X^2^*~0.05;\ 1~ = 458.34; p \< 0.001) and scavengers (*X^2^*~0.05;\ 1~ = 97.81; p \< 0.001), while the abundances of scavengers were greater than herbivores (*X^2^*~0.05;\ 1~ = 139.64; p \< 0.001). Family richness did not differ significantly among trophic guilds (*X^2^*~0.05;\ 2~ = 5.81; p = 0.0548) ([Figure 6](#f06){ref-type="fig"}). Discussion {#sec4} ========== This is the first community study based on a planned and intensive sampling effort that describes the composition and structure of the ground-dwelling arthropod community of Península Valdés. The most important orders based on abundance were Hymenoptera, Coleoptera, and Araneae. The same community pattern was found in other arid areas of Argentina ([@bibr26]; [@bibr38]; [@bibr32]), as well as in other regions of the world ([@bibr09]; [@bibr45]). The three aforementioned orders are the most diverse and abundant in the world, and several authors considered them "hyper-diverse" taxa ([@bibr28]; [@bibr36]; [@bibr32]). The community was dominated by few abundant taxa at both family and species levels. Also, there were some groups with intermediate abundances and a large proportion of "rare" taxa for which very few individuals were caught. Therefore, the distribution of both species and family abundances were better described by the Logarithmic series model. This model depicts a system where some species could have arrived at an unsaturated habitat at randomly spaced intervals of time in order to occupy the remaining fractions of the niche hyperspace, thus having intermediate levels of niche preferences. Similarly, this model describes systems in which one or a few factors dominate the ecological relationships of the community and in which the intensity of migration between communities is important ([@bibr33]). ###### Diversity values of arthropod assemblages.  It is worth noting that, at the species level, taxa with remarkably different abundance, such as ants, beetles, and true bugs, were equally described by the logs series. Still, in the case of true bugs, which were adequately described both by the log and log normal series, this represents a special case of log normal distribution called "canonical." Such pattern is a consequence of random niche separation every time a new species is incorporated into the assemblage ([@bibr33]). In this sense, these findings increase knowledge on niche segregation in general and on the invertebrate community structure of northeast Patagonia. Ants are a central component of arthropod abundance in the study area, representing more than 80% of total captures. The contribution of *P. bergi* and *S. patagonica*, both well-known recruiting species, may explain such outstanding numbers. Still, excluding ants from analyses of the assemblages of northeast Patagonia lead to similar findings in terms of abundance patterns. Such consistency likely reflects the robustness of the model and its explanatory factors for the Patagonian arthropods. In arid Patagonia, as in most deserts, the factors dominating the insect community structure are probably related to plants. Vegetation cover has shown to be correlated with diversity, dominance, and species abundance of ground-dwelling arthropods in other deserts ([@bibr16]; [@bibr45]). Vegetation structure usually provides the habitat template for the assembly of ground-dwelling arthropods in multitrophic communities by offering shelter, food resources, oviposition micro-sites, or refuge against predators ([@bibr20]; [@bibr45]; [@bibr37]). In turn, in northwest Patagonia, where there is a similar habitat to the one examined in this study, plant spatial structure has been shown to influence the activity of ground-dwelling ants and beetles ([@bibr21]; [@bibr23]; [@bibr37]). In addition, it should be considered that in Península Valdés sheep grazing has occurred since the late 19th century. Sheep grazing appears to have modified the vegetation and accelerated the soil degradation processes ([@bibr04]). These changes are generally referred to as changes in vegetation structure, diminishing their cover and exposing bare soil to erosive effects, which eventually leads to the fragmentation of the preexisting patches into smaller remnant patches ([@bibr07]). Grazing, through its impact on vegetation, could be influencing observed arthropod communities. From a trophic level approach, studies comparing protected areas versus grazed habitats in other arid areas from Argentina have found that arthropod communities were dominated by scavengers in protected sites and by predators in disturbed areas ([@bibr26]; [@bibr38]; [@bibr32]). In Península Valdés, the ground-dwelling arthropod community was dominated by predators, which suggests that sheep grazing could be one of the main variables modeling the arthropod assemblage structure. Predation could probably act as an important factor driving the distribution and abundances of surface-dwelling arthropods in this habitat (i.e., a top-down effect) and as such could be used as a key element in understanding the above-ground desert community structure. This study found that the arthropod community of northern Patagonia had similar diversity values to those recorded in other arid areas of Argentina, such as the Chaco ([@bibr26]; [@bibr38]) and the central Monte Desert ([@bibr32]). However, lower arthropod families and coleopteran species richness were found, as was smaller evenness at family and species levels. Reduced richness could be explained because of the lower temperatures present in Patagonia, which could constrain the number of species living there. In turn, a less even assemblage such as that found in this study suggests that the dominance of some species over others is greater than it is in other arid zones in northern Argentina. Species autoecological features coupled with a restrictive climate could explain why the community is dominated by a few species. For example, the most abundant beetle, *B. punctulatus* (Tenebrionidae), has a small body size that could allow them to hide into the soil fissures during extreme environmental periods. These features can also be observed in the true bug assemblage. For instance, *A. patagonica* is also small size and has wings like the elytra of coleoptera that enable it to tolerate extreme environmental conditions. The adequate description by the same abundance distribution model both at the family and the species level suggests that the former can be a reasonable predictor of the subjacent abundance model in this community. This reduces costs in terms of time dedicated to taxonomic determination and is in accordance with previous work (e.g. [@bibr10]). Using a higher taxonomic category than species level in community analysis has several advantages (see [@bibr25]), but it can be biased if the community has a fauna rich in endemisms ([@bibr44]). The results obtained in this study could be extended to all of arid Patagonia, due to similar environmental conditions in the area. This work not only improves the knowledge of the composition, taxonomy, and trophic structure of ground-dwelling arthropod communities in arid Patagonian habitats, but also increases the taxonomic knowledge of Hemiptera through the discoveries of new genera and two new species very recently described as new based on material recovered from this survey (see [@bibr19]; [@bibr11]). Additionally, it is necessary to place the results of this study within a conservation context because the richness and composition of a community of ground-dwelling arthropods can be taken as a reflection of the biotic and structural diversity of whole terrestrial ecosystems ([@bibr30]). Because of its abundance, diverse behaviors, and ecological interactions, the development of new lines of research to elucidate the variables controlling the main ecological aspects of grounddwelling arthropods will contribute significantly to the knowledge and functioning of arid Patagonian ecosystems. It also may help to create and assess management and conservation tools for the arid terrestrial ecosystem. The authors are grateful to those professional taxonomists that generously dedicated their time to species determination: G. Flores, S. Roig-Juñent, S. Claver, P. Dellapé, D. Carpintero, F. Ocampo, A. Lanteri, N. Cabrera, and M. Kun. We would also like to thank F. Grandi, F. Brusa, G. Pazos, V. Rodriguez, D. Galvan, L. Venerus, A. Bisigato and U. Pardiñas for their invaluable collaboration. We thank deeply Centro Nacional Patagónico and its staff for providing facilities and logistic support, and also Mrs. Amos Chess, Vicente Hueche, Jorge Mendioroz, Victor Huentelaf and Pedro "Perico" Ibarra who allowed access to the study areas. Finally thanks to L. Cella, R. Loizaga de Castro for her language assistance, two anonymous reviewers and Dr. Henry Hagedorn for their valuable comments that improved the manuscript. G. Cheli was supported by a PhD fellowship awarded by CONICET. This work was declared of interest by the Administration of the Natural Protected Area Península Valdés. [^1]: **Associate Editor:** Megha Parajulee was editor of this paper.

INTRODUCTION {#s1} ============ Squat exercise is the most commonly used exercise among various weight training methods because it is easily accessible without using tools[@r1]^)^. Squat exercises strengthen the gluteal, quadriceps, and trunk muscles that are important for running, jumping, and lifting[@r2]^)^. It is a basic exercise that strengthens bone density, ligaments, and tendons, as well as training the lower body. Squat movement, a close kinetic exercise, causes ankle joint flexion, knee joint flexion together with hip joint flexion, and is conducive to joint compression force and co-contraction[@r3]^)^. Therefore, squat movements are more advantageous than non-weight bearing exercises by moving many joints, mobilizing more muscles, and stimulating proprioceptive more strongly[@r4]^)^. However, the correct posture is important because the effect of squat exercise is influenced by posture, such as the lower extremity joints. In this regard, Tang et al.[@r5]^)^ experimented how muscle activation of the vastus medialis and vastus lateralis by squat-to-stand and stand-to-squat tasks was affected while changing the angle of the knee joint from 0 to 90 degrees in 15° increments. They reported the results of the study demonstrating that the highest muscle activation was observed at 60°. Park et al.[@r6]^)^ showed that the muscle activation of the biceps femoris, vastus medialis, vastus lateralis, gastrocnemius, and tibialis anterior increased as the knee flexion angle increased by 45°, 60°, and 90° when squatting against the wall. Since squat movement is also related to maintaining posture, ground conditions can also affect squat movement. In this regard, Lim[@r7]^)^ compared squat movement on air cushions of different air pressures, specifically 0 mbar, 1 mbar, 2 mbar, and on stable ground at a knee flexion angle of 60°. He reported muscle activation of vastus medialis, vastus lateralis, gluteus maximus, and biceps femoris was higher in squatting on the air cushions than on the stable ground. Thus, previous studies have already demonstrated that the angle of the knee flexion and ground conditions affect squat movement. Nevertheless, no studies have yet been performed that consider different knee flexion angles with ground conditions during squat exercises. Therefore, this study researched the most effective knee flexion angle for strengthening low-extremity muscles. At the same time, this study researched the effects of stable ground and unstable ground regarding for improving lower extremity muscle strength in squat exercise. SUBJECTS AND METHODS {#s2} ==================== The subjects of this study were 15 female college students who were enrolled in S university in Busan and were able to perform squat movements and did not have surgery experience, orthopedic disease, or musculoskeletal impairment. The purpose of this study and the experimental method were explained to all study subjects. This study complied with the ethical standards of the Declaration of Helsinki, and written informed consent was obtained from each participant. The general characteristics of the subjects were mean age 21.5 ± 0.7 years, mean weight 59.4 ± 11.0 kg and average height 160.6 ± 10.6 cm. To measure the knee flexion angle, a goniometer (Professional Goniometer Set, PATTERSON MEDICAL, China) was used. Squats were performed from various poses and postures. Escamilla[@r2]^)^ separated squat movements depending on the angle of the knee, as mini squat (30°), semi squat (40−60°), half squat (70−100°), and deep squat (100°). In addition, Park et al.[@r8]^)^ reported that when the knee flexion angle was above 90°, muscle activation of the rectus femoris, the vastus medialis, and the vastus lateralis decreased. Therefore, this study was conducted to examine changes of muscle activation of low-extremity muscles at different joint angles of 70°, 90°, and 100°, 70° being the minimum of the knee flexion angle of the half squat; 90° showing decrease of muscle activation, and 100° is the starting range of the deep squat. The knee flexion angle was measured as the relative angle of the femur to the fibular, as suggested by Lee et al.[@r9]^)^ Electromyography measurements were performed on stable ground and on unstable ground using the surface electromyography (4D-MT & EMD-11, Relive, Korea). The surface of the stable ground was flat with no obstacles and balance pad (Aero Step, TOGU, Germany) was used as unstable ground. Electromyography measurements were performed with the knee flexion angles of 70°, 90°, and 100°, which were maintained for 5 seconds. A total of three measurements were made and a 5 minute break between the measurements was given. The subjects were verbally instructed to keep their arms straight forward, legs at shoulder width, gaze toward the front, and the back straight with knees bent. Measurement was performed at the angle with a paused posture. The joint angle was measured by the goniometer attached to the lateral side of the knee. The sampling rate of the signal of surface electromyography used in this study was 1,000 Hz, and the signals were filtered using 60 Hz analog notch filter and a 20−500 Hz digital band pass filter. The attachment site of the surface electrode was the right leg, the dominant leg of each subject. Measured muscles were vastus medialis, biceps femoris, tibialis anterior, and gastrocnemius. The muscle activation was determined by the root mean square (RMS) average value. The average value of the root mean square (RMS) values for three seconds was calculated by linear filtering the values of three measurement data for five seconds and excluding one second for each of the initial and late phases. Data were analyzed by paired sample t-test to determine the difference of the value which was measured on both unstable ground and stable ground at each knee flexion angle. Then repeated measurement ANOVA test was performed to determine whether there was a difference in muscle activation according to the angle in each ground condition. Post analysis was performed using Bonferroni Correction. The statistical program used in this study was SPSSWIN (ver. 23.0), with a significance level α=0.05. RESULTS {#s3} ======= There was a difference in muscle activation of the vastus medialis according to the change of the knee flexion angle with stable ground (p\<0.05). Post analysis showed that muscle activation was greater at 90° and 100° than at 70° of the knee joint, but no difference was demonstrated between 90° and 100°. On the other hand, on unstable ground, there was no apparent difference in muscle activation according to the angle change of the knee joint. In biceps femoris, there was no difference in muscle activation due to the angle change of the knee joint with both stable ground and unstable ground. In the tibialis anterior muscle, there was a difference in muscle activation according to the angle of the knee joint on stable ground (p\<0.05). Post analysis showed that there was no difference between knee flexion angle 70° and 90°, 90° and 100°, but muscle activation was greater at 100° than at 70°. However, there was no difference in muscle activation according to the angle change with the unstable ground. In the case of the gastrocnemius, there was no difference in muscle activation according to the angle change of the knee joint with both the stable ground and the unstable ground. In addition, differences in muscle activation measured with the stable ground and the unstable ground at the different knee flexion angles were checked. The data show that there is no difference in muscle activation in the case of the vastus medialis, the biceps femoris, the tibialis anterior and the gastrocnemius according to the ground condition ([Table 1](#tbl_001){ref-type="table"}Table 1.The effect of knee flexion angle and ground condition on the muscle activation in the low-extremity muscle (Unit: mV)MuscleCondition70°90°100°Mean ± SDMean ± SDMean ± SDVastus medialisStable ground\*39.7 ± 12.5^a^48.5 ± 12.148.0 ± 13.6Unstable ground42.9 ± 14.248.0 ± 13.047.3 ± 14.7t−1.4000.3240.448p0.1830.7510.661Biceps femorisStable ground34.9 ± 30.539.0 ± 37.941.7 ± 37.6Unstable ground48.3 ± 34.745.7 ± 30.542.9 ± 22.7t−1.243−0.800−0.281p0.2340.4370.783Tibialis anteriorStable ground\*60.3 ± 22.7^b^68.5 ± 25.274.0 ± 23.9Unstable ground72.5 ± 23.372.3 ± 26.373.9 ± 22.4t−2.071−0.4410.008p0.0570.6660.994GastrocnemiusStable ground41.2 ± 29.743.7 ± 31.243.7 ± 27.6Unstable ground51.6 ± 26.754.2 ± 28.549.5 ± 28.4t−1.785−1.496−0.918p0.0960.1570.374\*p\<0.05 with Repeated ANOVA according to knee flexion angle^a^70°\<90°=100° by using Boneferroni Correction as post analysis^b^70°\<100°, 70°=90°, 90°=100° by using Boneferroni Correction as post analysis). DISCUSSION {#s4} ========== In this study, it was confirmed that muscle activation of the vastus medialis measured on stable ground was significantly increased at the knee flexion angles 90° and 100°, compared to 70°. Marchetti et al.[@r10]^)^ studied changes in muscle activation of the vastus medialis at the knee flexion angles 20°, 90°, and 140°. As a result, it reported that muscle activation was the highest at 90 °, which is the same as the result of this study. This result can be explained by the principle of the lever. In general, torque can be described as the amount of force multiplied by the distance from the line of action of the force to the axis of movement. When the knee flexion angle is 90°, the distance from the axis of movement to the line of action is the longest, resulting in maximum torque[@r11]^)^. Therefore, this study and the study of Marchetti et al.[@r10]^)^ report that the knee flexion angle of 90° is a suitable joint angle for generating the maximum torque during squat movement. In addition, muscle activation of the Tibialis anterior was increased at the knee flexion angles of 90° and 100° on stable ground. Park et al.[@r6]^)^ reported that muscle activation of the Tibialis anterior at the knee flexion angle 90° was higher than at 45° and 60°. These results show that as the angle of the knee joint increases, the center of gravity (COG) of the subject gradually moves away from the base of support (BOS) and the subject uses more force to maintain the center of gravity within the base of support for a stable pose[@r11]^)^. At this time, the ankle attempts to maintain balance using the ankle joint strategy in response to postural sway[@r12]^)^, and the balance against the alternating sway caused by the ankle joint strategy is dependent on alternating activation of the tibialis anterior and the gastrocnemius medialis[@r13]^)^. For this reason, it is clear from this study that the tibialis anterior is more activated when the center of gravity (COG) of the body moves backward, and this contributes to balance of the body. In the case of biceps femoris, muscle activation increased as the angle increased, but there was no statistically significant difference. It appears that the quadriceps femoris was more active than biceps femoris because the muscle activation was measured while bending the knee joint for 5 seconds. And the muscle activation of gastrocnemius increased as the angle increased, but it did not show a statistically significant difference. This is because the region of the origin and of the insertion of the gastrocnemius is attached to the calcaneus starting from the back of the tibia and the fibular, so that the angle of the knee joint seems not to affect the gastrocnemius. Hence, in the case of squat movements on stable ground, the knee flexion angle from 90° to 100° is considered to be effective for the exercise of the vastus medialis and tibialis anterior because the knee flexion angle from 90° to 100° has the highest muscle activation in the vastus medialis, and the knee flexion angle of 90° has the highest muscle activation in the tibialis anterior. Checking the difference in muscle activation of the low-extremity muscles according to ground conditions, there was no difference in muscle activation of the vastus medialis, tibialis anterior, biceps femoris, and gastrocnemius. These results suggest that changes in the angle of the knee joint affect the muscle activation of the vastus medialis and tibialis anterior. It was found that as the joint angle increases, the muscle activation increases, while ground conditions do not affect muscle activation. Therefore, it can be said that a squat posture is helpful to increase the muscle power of the vastus medialis and tibialis anterior. On the other hand, if the subject can safely balance on the balance pad, it is suggested that the he or she can squat at an angle of 70° on unstable ground. However, if the subject is unable to balance on the balance pad due to the lack of balance ability, it is reasonable to suggest he or she perform squat movements at the knee flexion angle 100° on stable ground. But for confirmation these facts various study will be needed more in the future.

Microtubules are hollow, cylindrical polymers of αβ-tubulin that display dynamic instability, the apparently random switching between phases of growing and shrinking. The dynamic instability of microtubules is essential for formation of the mitotic spindle that organizes chromosome segregation, and for numerous other processes in cell physiology. In this issue, McIntosh et al. use electron tomography to provide 3D snapshots of growing microtubule ends in cells and in vitro. They report that the structures of growing and shrinking microtubule ends are very similar to each other, challenging the current understanding of microtubule dynamics and regulation by contradicting a long-held view in the field (see [Fig. 1](#fig1){ref-type="fig"}). Microtubule growing and shrinking occurs by the net addition or loss of αβ-tubulins to or from the end of the polymer. Catastrophe, the switch from growing to shrinking, results from the GTPase activity of αβ-tubulins in the polymer. Dynamic instability can be reconstituted in vitro using pure αβ-tubulin and GTP. However, this compositional simplicity belies a sneaky structural complexity: αβ-tubulin subunits adopt a different conformation outside the polymer than they do inside the polymer, and they adopt multiple conformations in the polymer, only three of which are known in atomic detail ([@bib10]). These different conformations of αβ-tubulin contribute to dynamic instability by modulating tubulin--tubulin interactions (reviewed in [@bib3]). The question [@bib8] addressed sounds like a simple one: What does a growing microtubule end look like? It's an important question because distinctive microtubule end structures can selectively recruit different regulatory factors (reviewed in [@bib1]), transitions between different end structures may contribute to force production by microtubules ([@bib6]; [@bib5]), and knowledge of the end structure informs and constrains mechanistic models for microtubule dynamics. If we understood the biochemistry of the αβ-tubulin conformational cycle, we could predict end structure. Conversely, if we knew the end structure, we could infer the biochemistry. But we lack confidence about both the biochemistry and the end structure. It has been known for some time that growing microtubule ends are structurally heterogeneous and characterized by extensions that curve away from the long axis of the microtubule. However, we lack an atomic resolution view of microtubule ends: their structural features are too small to image by light microscopy and too idiosyncratic for the averaging approaches that have allowed cryoEM to deliver high-resolution structures of αβ-tubulin ([@bib10]) in the more regular, lattice-like body of the microtubule. Electron tomography, the technique used by [@bib8], is ideally suited for one-of-a-kind objects like microtubule ends that cannot be averaged together. Electron tomography produces a 3D image of a sample by combining many (2D) transmission electron micrographs in which the same sample is "viewed" from different angles; the different views are obtained by tilting the specimen to different degrees relative to the electron beam. Exposure to the electron beam damages the sample, so the maximum tolerated exposure must be spread among the multiple images required for tomography. As a result of the attendant low signal-to-noise ratio in individual images, and/or because of other cellular components, electron tomograms are often rather noisy, especially when considering "wispy" structures like individual microtubule protofilaments. [@bib8] obtained tomograms of growing microtubules in cells from six different species, frozen and/or fixed for tomography in a variety of ways. They also obtained tomograms from samples of microtubules in vitro that were plunge-frozen to trap them in the act of growing. The authors then applied "rotary sectioning" ([Fig. 1 A](#fig1){ref-type="fig"}) to characterize the structures of individual microtubule ends: they examined "sagittal" sections (parallel to the long axis; [Fig. 1 A](#fig1){ref-type="fig"}) at various angles, through hundreds of microtubules. Their approach is admirably rigorous, and the paper has a wonderfully "old school" feel---considerable attention is devoted to the minute workings of the sample preparation and to considering, testing, and excluding possible sources of experimental artifact. {#fig1} Both in cells and in vitro, [@bib8] observed short, curved extensions on the ends of growing microtubules ([Fig. 1 B](#fig1){ref-type="fig"}, right). Manually tracing these extensions using rotary sectioning identified about as many curved extensions as protofilaments, and the extensions were evenly spaced around the microtubule. [@bib8] conclude that the extensions are in fact curled protofilaments growing independently of one another. The curvature observed was highly variable, which was taken to indicate that the curled protofilaments are flexible in the plane of curvature. The average curvature was comparable in magnitude to that seen in head-to-tail assemblies modeled from atomic structures of unpolymerized αβ-tubulin, so it likely reflects the intrinsic curvature of unpolymerized, GTP-bound αβ-tubulin. That growing and shrinking microtubule ends have markedly different structures is practically axiomatic in the current understanding of microtubule dynamics ([Fig. 1, B and C](#fig1){ref-type="fig"}). But in cells and in vitro, [@bib8] describe remarkably similar structures for growing and shrinking ends. Their results are therefore quite provocative, contradicting a long-held belief and challenging models built on that belief. Their work has potential implications for how we think about the molecular mechanisms of microtubule dynamics and regulation and for understanding processes like kinetochore--microtubule attachment, where assumptions about different end structures figure prominently ([@bib9]). The poor signal-to-noise ratio of tomographic reconstructions makes robust annotation of fine features challenging. The long-held belief that growing and shrinking microtubule ends have very different structures is itself based on a landmark cryoEM study ([@bib4]) that revealed tapered and partially curved sheet-like structures at growing microtubule ends that were obviously distinct from the more curled, independent protofilaments on shrinking ends. Two recent tomography studies also show curved and partially curved extensions at the growing end ([@bib7]; [@bib2]). However, this other work (in which rotary sectioning was not applied) was interpreted as showing that many of the curved protofilaments make lateral contacts to another protofilament (as in a sheet), contrary to the independent protofilaments described by [@bib8]. The McIntosh team has set a new standard in the quest to define the molecular features of the growing microtubule end. In doing so, they introduce a dissenting view about what the growing end looks like and about how and where curved αβ-tubulins are added to the microtubule end. The conflict revolves around whether the curved extensions are laterally connected, and whether "pioneer" elongation from a subset of protofilaments occurs. These different scenarios, which might in principle coexist on the same microtubule end, have implications for our understanding of the biochemistry of microtubule growth: Do curved tubulins interact tightly enough with individual protofilament ends for the protofilaments to elongate independently, or is simultaneous interaction with more than one protofilament (sheet-like growth) required for efficient elongation? A definitive resolution to this conflict has implications for microtubule regulation and for how growing and shrinking ends are recognized, and will likely come in the not too distant future as ongoing developments in the hardware and software for cryoEM and tomography are expected to improve the signal-to-noise ratio achievable. A clearer view of growing microtubule end structure looks to be around the curve! L.M. Rice is the Thomas O. Hicks Scholar in Medical Research. Work in his laboratory is supported by the National Institutes of Health (R01-GM098543), the National Science Foundation (MCB-1615938), and the Robert A. Welch Foundation (I-1908). The author declares no competing financial interests.

INTRODUCTION {#s1} ============ Severe fever with thrombocytopenia syndrome (SFTS) is an emerging infectious disease that is caused by a novel bunyavirus named SFTS virus (SFTSV) in the *Bunyaviridae* family. Since its discovery in 2009 in the middle east of China, SFTSV infection has been documented in 19 provinces, leading to at least 2047 reported cases nationally in 2011 and 2012 \[[@R1]\]. Recently SFTSV infected cases were also reported from South Korea and Japan, indicating the imminent public health impact of this emerging infectious disease \[[@R2]-[@R4]\]. The SFTSV infected patients had wide clinical spectrum, with some experiencing self-limiting clinical course, while approximately 12% might develop fatal disease. The iceberg phenomenon of SFTS is obvious that a seroprevalence of 0.84-6.37% was found among the healthy population residing in endemic areas \[[@R5], [@R6]\]. Although with large amount of individuals who had been exposed to the virus, only a small proportion developed symptomatic disease, suggesting the role of host susceptibility in determining the clinical outcome. Research on person to person transmission events disclosed a higher tendency of having clinical disease and severe disease outcome in the inheritable related individuals, providing further evidence that human genetic background might play roles in the disease development and modulation of disease severity \[[@R7], [@R8]\]. The pathogenesis of SFTS has been partially explicated, indicating remarkable virus replication and a dyregulated immune response after infection contributing to disease progression \[[@R9], [@R10]\]. Among all the immune dyregulation, the cytokine mediated inflammatory response, characterized by cytokine/chemokine production imbalance, was found to be responsible for SFTSV infection and disease progression \[[@R9]-[@R11]\]. It makes sense to hypothesize that some genetic determinants might reside in those immune system genes polymorphisms that regulate host immune responses, in particular cytokine gene polymorphisms. Up to the current stage, although a handful of cytokine/chemokine has been evaluated in SFTS patients, most of the results have been inconsistent \[[@R9]-[@R11]\]. The only one exception was platelet derived growth factor-BB (PDGF-BB), which had been independently and consistently observed to be decreased in the SFTS patients than controls, regardless of the sample size or sampling time in different studies \[[@R10], [@R12]\]. PDGF is a family of cationic homo- and heterodimers of disulfide-bonded A- and B- polypeptide chains, acting as the major growth factor in health and disease. The BB isoform of PDGF (PDGF-BB) is a key regulatory molecule in various physiological processes such as bone homeostasis, repair and regeneration \[[@R13]\]. The effects of PDGF family on endothelial cells have been demonstrated in various studies \[[@R14], [@R15]\]. Since SFTSV can infect vascular endothelial \[[@R16]\], it\'s highly possible that PDGF might play a role in the pathogenesis of the SFTSV infections. For the *PDGF-B* gene, the population based association studies have been focused on two single nucleotide polymorphisms (SNPs), rs1800818 located in the 5b 2-UTR and rs1800817 in first intron \[[@R17]-[@R21]\]. The rs1800818, a isoleucine (Ile) to valine (Val) substitution, was found to affect receptor affinity and cell activation \[[@R17]\]. The rs1800818 and rs1800817 polymorphisms have been demonstrated to alter the individual risk to cardiac allograft vasculopathy \[[@R18]\]. It has been shown that the rs1800818 polymorphism was the leading SNP that predicted 3-year overall survival in patientas with resected colorectal liver metastases who receive bevacizumab-based chemotherapy \[[@R19]\]. An association between the *PDGF-B* rs1800817 polymorphism and severe recurrent HCV infection after liver transplantation was also observed \[[@R21]\]. The role of the *PDGF-B* polymorphisms in SFTS, however, has never been investigated. In the present study, we examined whether these two important *PDGF-B* polymorphisms has any bearing on the risk or severity of SFTS in Chinese populations. RESULTS {#s2} ======= Clinical and laboratory characteristics of study subjects {#s2_1} --------------------------------------------------------- A total of 1020 virologically confirmed SFTS patients and 1353 controls were recruited for the study. By checking the medical records and by interviewing the participants, we determined that all cases and controls were genetically unrelated Han Chinese. The selected characteristics of subjects are shown in Table [1](#T1){ref-type="table"}. Compared with the controls, the SFTS patients had significantly older age (*P* \< 0.001), higher frequency of female (*P* \< 0.001) and more presence of underlying conditions (*P* = 0.013). ###### Selected characteristic of patients with severe fever with thrombocytopenia syndrome and controls Variables Patients Controls (n= 1353) *P* value[^b^](#tfn_002){ref-type="table-fn"} -------------------------------------------------------------------------- ------------- -------------------- ----------------------------------------------- ------------- ------------- ------------ ---------- ---------- ---------- Age, year   Mean (SD) 60.7 (12.0) 59.8 (12.1) 62.7 (11.7) 59.8 (12.0) 67.5 (10.4) 47.8(19.2) \< 0.001 \< 0.001 \< 0.001   ≤ 60, n. (%) 473 (46.4) 361 (49.5) 112 (38.6) 453 (49.7) 20 (18.5) 972 (71.8) \< 0.001 0.002 \< 0.001 Male, n (%) 428 (42.0) 280 (38.4) 148 (51.0) 373 (40.9) 55 (50.9) 694 (51.3) \< 0.001 \< 0.001 0.046 Underlying medical conditions[^a^](#tfn_001){ref-type="table-fn"}, n (%) 278 (27.3) 171 (23.4) 107 (36.9) 234 (25.7) 44 (40.6) 309 (22.8) 0.013 \< 0.001 0.001 Abbreviations: SD, standard deviation. The underlying medical conditions were defined as patients presenting with one of the following: hypertension, diabetes, cancer, active hepatitis, cerebral infarction, et al. *χ*^2^ test for categorical variables and the Mann Whitney U test for continuous variable *PDGF-B* rs1800818 *G* allele conferred increased susceptibility to SFTS disease {#s2_2} -------------------------------------------------------------------------------- The initial small-scale association study was performed in the first cohort of 250 SFTS patients and 250 controls. The genotyping results were presented in [Supplementary Table S2](#SD1){ref-type="supplementary-material"}. The observed genotype frequencies of the rs1800818 and rs1800817 polymorphisms were in Hardy-Weinberg equilibrium in both patients and controls groups (all *P* \> 0.05, data not shown). By using multivariate logistic regression model to adjust for the effect from age, sex, and underlying disease, significant associations with SFTS were observed for the *PDGF-B* rs1800818. The association between rs1800818 and SFTS remained after multiple corrections ([Supplementary Table S2](#SD1){ref-type="supplementary-material"}). There was no association between rs1800817 polymorphism and SFTS ([Supplementary Table S2](#SD1){ref-type="supplementary-material"}). Therefore we focused on rs1800818 for further study. With DNA samples from more patients available over time, we genotyped rs1800818 from additional 770 SFTS patients and 1103 controls by PCR-direct sequencing, which results were combined with those obtained from initial small-scale association study for analysis. In all of the 1020 SFTS patients and 1353 controls, rs1800818 G allele was significantly overrepresented in SFTS patients than controls (9% vs. 5.6%, *P* \< 0.001). After adjustment for age, sex, and underlying medical conditions, the genotypes containing G allele (AG + GG genotypes) were significantly associated with increased susceptibility to SFTS (OR = 1.66, 95% CI = 1.28-2.16; *P* \< 0.001; Table [2](#T2){ref-type="table"}, Figure [1a](#F1){ref-type="fig"}). ###### Association of *PDGF-B* rs1800818 polymorphism with severe fever with thrombocytopenia syndrome Genotypes Patients (n = 1020) Controls (n = 1353) Model OR (95% CI)[^a^](#tfn_003){ref-type="table-fn"} *P* value[^a^](#tfn_003){ref-type="table-fn"} ----------- --------------------- --------------------- -------------- ------------------------------------------------- ----------------------------------------------- AA 799 (83.1) 1115 (82.4) Reference AG 151 (15.7) 135 (10.0) Codominant 1.57 (1.20-2.05) \< 0.001 GG 11 (1.1) 3 (0.2) 6.18 (1.59-23.98) AG + GG 162 (16.9) 138 (10.2) Dominant 1.66 (1.28-2.16) \< 0.001 Recessive 5.80 (1.50-22.48) 0.005 Overdominant 1.55 (1.18-2.03) 0.001 Log-additive 1.68 (1.31-2.15) \< 0.001 Abbreviations: OR, odds ratio; CI, confidence interval; The ORs and *P* values were adjusted for age, sex, and underlying medical conditions. {#F1} PDGF-BB was persistently reduced in SFTSV-infected C57BL/6J mice {#s2_3} ---------------------------------------------------------------- Altogether 50 mice were successfully infected with SFTSV; the viral loads of the serum samples were shown to be maintained at detectable level throughout the whole observation, indicating an effective infection with SFTSV. Post infection, the reduction of PDGF-BB levels was observed at first observation on 3 DPI, which kept decreasing in the serially collected serum samples on 14 and 21 DPI, till the last observation of 28 DPI (*P* \< 0.001, Figure [1b](#F1){ref-type="fig"}). Effects of the rs1800818 polymorphism on PDGF-BB expression in SFTS patients and controls {#s2_4} ----------------------------------------------------------------------------------------- For eight rs1800818 heterozygous (GA genotype) patients who had paired PBMC estimated, at the acute phase, three had mono-allelic (A allele) *PDGF-BB* expression, however, G allele related *PDGF-BB* expression was undetectable for all patient. Furthermore, PDGF-B messenger RNA was increased at the convalescent phase and we observed that bi-allelic *PDGF-B* expression related to rs1800818 polymorphism in the PBMCs from the convalescent phase of all eight patients (Figure [1c](#F1){ref-type="fig"}). Altogether 44 SFTS patients and 27 controls were evaluated for the serum PDGF-BB level. The serum PDGF-BB levels from SFTS patients at acute phase were significantly lower than that obtained from controls (*P* \< 0.001, Figure [1d](#F1){ref-type="fig"}). In addition, the PDGF-BB levels from the recovered SFTS patients were significantly increased to comparable level with controls at 6 moths (*P* = 0.219) and 12 month (*P* = 0.577) after disease. Among the 44 SFTS patients, those carrying the rs1800818 AG+GG genotype (n=26) had significantly lower PDGF-BB level than the AA genotype carriers at acute phase (n=18) (*P* = 0.015, Figure [1f](#F1){ref-type="fig"}). However, no such difference was observed among either controls or recovered SFTS patients at 6 and 12 months after disease (Figure [1e](#F1){ref-type="fig"}). At acute phase, PDGF-BB levels were also negatively correlated with SFTSV loads (*P* = 0.020, Figure [1f](#F1){ref-type="fig"}). *PDGF-B* rs1800818 conferred no susceptibility to severe outcome in SFTS patients {#s2_5} --------------------------------------------------------------------------------- Among the 1020 SFTS patients, 290 developed severe disease outcome, comprising 108 fatal cases. Compared with mild patients, significantly older age, more male gender and presence of underlying medical conditions were found in severe patients (Table [1](#T1){ref-type="table"}). Similar results were found when comparison was made between fatal and nonfatal patients (Table [1](#T1){ref-type="table"}). Comparison between severe patients vs. mild patients (OR = 0.87, 95% CI = 0.59-1.29; *P* = 0.48; Figure [2a](#F2){ref-type="fig"} and [Supplementary Table S3](#SD1){ref-type="supplementary-material"}) and fatal patients vs. nonfatal patients (OR = 0.91, 95% CI = 0.51-1.62; *P* = 0.75; Figure [2b](#F2){ref-type="fig"} and [Supplementary Table S4](#SD1){ref-type="supplementary-material"}) displayed no significant difference of the genotype frequencies for *PDGF-B* rs1800818 polymorphism after adjusting the effect from age, gender and underlying medical conditions by applying multivariate regression analysis. The PDGF-BB levels that were measured at acute infection in severe cases were comparable with mild cases (*P* = 0.45; Figure [2c](#F2){ref-type="fig"}). The PDGF-BB levels that were measured at acute infection in fatal cases were comparable with non-fatal cases as well (*P* = 0.96, Figure [2d](#F2){ref-type="fig"}). {#F2} *PDGF-B* rs1800818 exerted minor effect on clinical recovery in SFTS patients {#s2_6} ----------------------------------------------------------------------------- We also analyzed the effect of rs1800818 polymorphism on clinical recovery of SFTS patients, with important laboratory parameters that were serially evaluated as dependent variables, i.e, PLT (platelets), WBC (white blood cell), ALT (alanine transaminase), AST (aspartate aminotransferase), LDH (lactate dehydrogenase) and viral load ([Supplementary figure S1](#SD1){ref-type="supplementary-material"}). As we have displayed for all patients, serial LDH and viral loads demonstrated significantly differential pattern for the genotypes containing G allele (AG + GG genotypes) compared with the AA genotype ([Supplementary Figure S1a and S1d](#SD1){ref-type="supplementary-material"}). Both evaluations were kept at comparable level between two groups of patients at early phase of infection, however on 7-9 days after disease when patients entered into convalescent phase, patients who carried AG + GG genotypes demonstrated more rapid viral clearance (*P* = 0.009) and recovery of LDH to normal level (*P* = 0.03). No significant difference was observed for other laboratory parameters regarding the rs1800818 genotypes ([Supplementary Figure S1b, S1c, S1e and S1f](#SD1){ref-type="supplementary-material"}). DISCUSSION {#s3} ========== It is well-known that host genetic immunity might determine the outcome of infectious diseases. In this study, we found that PDGF-BB secretion was significantly reduced at acute phase of SFTS patients and *PDGF-B* rs1800818 polymorphism was significantly associated with host susceptibility to SFTS. Genotypes containing G allele confer increased risk of SFTS disease than the AA genotype. The G allele-specific *PDGF-B* RNA expression and PDGF-BB serum levels were significantly lower than those derived from A allele or AA genotype. In mice model, the SFTSV infected mice demonstrated a consistently decreased expression of PDGF-BB. Taken this together, we provided evidence that PDGF-BB might be associated with SFTS disease at both the genetic and serum levels. It\'s suggested that the ability of individuals to respond properly to SFTSV infection may be impaired by SNPs within *PDGF-B* gene, resulting in decreased transcription and secretion of PDGF-BB, eventually leading to an increased susceptibility to SFTS development. The observed genetic association is plausible from a biological perspective. Platelet is the fundamental component of primary hemostasis, which is also known to release many growth factors when aggregated and activated, including PDGF, transforming growth factor-beta, vascular endothelial growth factor and epidermal growth factor. The PDGF released from platelets, on the other hand, serves an autocrine feedback role in control of platelet aggregation. Thrombocytopenia is the major clinical hallmark symptom of SFTSV infection, the underlying pathogenic mechanism is suggested to be clearance of circulating virus-bound platelets by splenic macrophages, as displayed in C57/BL6 mouse model \[[@R9]\]. In case of SFTSV infection, SFTSV adherence on platelets resulted in enhanced release of PDGF from platelets, exerting feedback control effects on platelet aggregation. This further lead to decreased deposition of platelets in tissues and increased platelets in circulation, facilitating the enhanced clearance of virus-bound platelets promoted by splenic macrophages. This process represents the host response of limiting and clearing SFTSV in asymptomatic individuals. In contrast, incapability of PDGF-BB express and release in individuals who carried the rs1800818 *G* allele might be associated with suppressed PDGF-BB regulation on SFTSV clearance, which ultimately lead to active viral replication and thrombocytopenia, manifesting as the SFTS (Figure [3](#F3){ref-type="fig"}). This could be evidenced by depressed PDGF-BB expression from both human specimens and infected mice in the current study. The significant inverse correlation between PDGF-BB and SFTSV viral loads provided further evidence. As the disease progress into convalescence phase, when the virus began to be cleared by activated immunity, the genotype specific effect from SFTSV was suggested to decline, finally eliciting a rapid recovery of PDGF-BB level, as well as a more rapid clinical recovery of these patients. This genotype specific effect on PDGF-BB expression was not observed in healthy control or the recovered patients, which supported our hypothesis that the transcriptional regulation of PDGF-BB might take place at SFTSV viral exposure and early than the development of the thrombocytopenia. Although with this biologically based priori hypothesis, it remained obscure how the regulation of PDGF-BB induction on virus binding occurred at the transcriptional level. The mechanism underlying the potential autocrine feedback effect of the PDGF-BB in control of platelet aggregation remained to be further investigated. {#F3} On the other hand, we failed to find association between rs1800818 polymorphism and severe outcome in SFTS patients, from either genetic level or serum level, indicating minor role of PDGF-BB in determining severe outcome in SFTS patients. As have been indicated in mouse model, virus replication and over exuberant immune responses contribute to the progressive organ damage resembling human SFTS \[[@R9]\]. The effect of PDGF-BB on severe outcome in SFTS patients, if there is any, might be masked by other uninvestigated host immune related factors. On the other hand, the small sample size of fatal case has hindered our efforts in identifying potential associations between genetic factor and disease outcome, which need further investigation on another case cohort. The current study has the advantage of bearing desirable features that are considered as components of an ideal genetic association study. These characteristics include rigorous case selection, associations that make biological sense and allele that affect the gene product in a physiologically meaningful way. An extraordinary advantage of this study is the minimal interference from the recall bias that is inherent to the traditional case control study. As is known, SFTS is an emerging infectious disease that had become endemic since 2010, which had attracted intense attention in local residence due to the high case fatality rate. Therefore the previous medical history that resembled the SFTS is unlikely to be missed by the recalling information of questionnaire interview. The positive IgG detection in endemic region is highly indicative that patients never had the clinical disease after exposure to SFTSV. In conclusion, our data provide strong evidence that the *PDGF-B* rs1800818 secretion regulated by genetic polymorphism may affect the occurrence of SFTS in Chinese population. The screen of this risk allele in SFTS endemic region might help to identify individuals with high-risk of becoming ill after exposure to SFTSV, for strengthened prevention measures to be taken. From clinical perspective, the specifically depressed PDGF-BB in SFTS has been suggested in our study, supporting its potential application as clinical diagnosis marker. Until recently, no specific therapy is available to treat this infection with high morbidity. The current finding, if confirmed in other cohorts, might offer perspective of applying PDGF-BB as a valuable therapeutic target. Moreover, these findings could shed light on the pathogenesis of SFTS, if PDGF-BB and its signaling pathway involvement could be further explored. Although with this advantage, our initial findings should be independently verified in populations of different ancestry, in case of adequate patients sample is accessible. MATERIALS AND METHODS {#s4} ===================== Study subjects and information collection {#s4_1} ----------------------------------------- The study was performed in the SFTS designated hospital (The PLA 154 Hospital and The Shangcheng People\'s Hospital) in Xinyang administrative district of Henan Province between 2014 and 2015. On admission, all clinical diagnosed SFTS patients had sera samples collected and subjected to SFTSV RNA test by a molecular method as described previously (detailed in [supplemental material and methods](#SD1){ref-type="supplementary-material"}) \[[@R22]\]. Totally 1020 patients with a positive result were invited to participate in the study. Severe cases were defined by the presence of hemorrhagic manifestations (epistaxis, hematemesis, and melena), or presence of one or more organ failure or encephalitis development \[[@R23]\]. Altogether 1353 controls were randomly selected from healthy subjects who underwent routine physical examination in the same region during the same period that the cases were recruited. At recruitment, informed consent was obtained from all participants, and personal information on demographic factors and medical history were collected via structured questionnaire. The study was performed with the approval of the Ethical Committee of Beijing Institute of Microbiology and Epidemiology and conducted according to the principles expressed in the Declaration of Helsinki. Genotyping of rs1800818 and rs1800817 polymorphisms {#s4_2} --------------------------------------------------- Two SNPs of *PDGF-B* (rs1800818 and rs1800817) that either have reputed functional significance or been reported to affect disease susceptibility were evaluated in the present study \[[@R18], [@R21]\]. The initial small-scale case-control study involving the determination of rs1800818 and rs1800817 polymorphisms in 250 SFTS patients and 250 controls were analyzed by the MassArray System (Sequenom) with primers (see [Supplementary Materials, Supplementary Table S1](#SD1){ref-type="supplementary-material"}) as described previously \[[@R24]\]. Additional 1873 DNA samples were genotyped for only rs1800818 by polymerase chain reaction (PCR) direct sequencing (detailed in [supplemental material and methods](#SD1){ref-type="supplementary-material"}). Genotyping was done in a blind manner that the performers did not know the subjects\' case and control status. For quality control, a 15% masked, random sample of cases and controls was tested twice by different people and all results were 100% concordance. In addition, genotypes identified by the MassArray System were confirmed by DNA sequencing. PDGF-BB expression in SFTSV-infected C57BL/6J mice {#s4_3} -------------------------------------------------- To evaluate the expression of PDGF-BB after SFTSV infection, C57BL/6J mice of 3- to 4-week-old were challenged by intraperitoneal injection with 3×10^7^ focus forming unit (FFU) of SFTSV. Every 6-7 mice were bleeded on day 3, 14, 21 and 28 post infection (DPI) (detailed in [supplemental material and methods](#SD1){ref-type="supplementary-material"}). The PDGF-BB level at each time point was detected by ELISA Kit (R&D Systems Inc., USA). All measurements were performed in duplicate. All experimental procedures with animals were in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals and approved by the Committee on the Ethics of Animal Experiments of the Academy of Military Medical Sciences. *PDGF-B* messenger RNA expression in SFTS patients {#s4_4} -------------------------------------------------- The rs1800818 polymorphism lies in the 5b 2-UTR of the *PDGF-B* gene. This allowed us to determine allele-specific gene expression of *PDGF-B* messenger RNA in rs1800818 heterozygous (GA genotype) individuals. Paired acute and convalescent peripheral blood mononuclear cells (PBMCs) were collected from 8 SFTS patients and measured for the *PDGF-B* messenger RNA using real-time quantitative PCR with G allele-specific probe and *A* allele-specific probe, respectively (detailed in [Supplementary Materials](#SD1){ref-type="supplementary-material"}). PDGF-BB expression in serum samples of SFTS patients and controls {#s4_5} ----------------------------------------------------------------- To compare the expression patterns of PDGF-BB, acute SFTS patients and age matched controls were selected to measure the PDGF-BB levels by using PDGF-BB ELISA assay (GenWay Biotech, USA). The SFTS patients who were successfully followed up and sampled during the convalescence were also tested for the PDGF-BB evaluation. All measurements were performed in duplicate. Statistical analysis {#s4_6} -------------------- Genotype and allele frequencies for polymorphisms were determined by gene counting. The fitness to the Hardy-Weinberg equilibrium was tested using the *χ*^2^ test. Associations between polymorphisms and risk of SFTS were estimated by use of logistic regression analyses. Odds ratios (ORs) and 95% confidence intervals (CIs) were used to measure the strength of association. In view of the multiple comparisons, the correction factor n (m--1) (n loci with m alleles each) was applied to correct the significance level. The PDGF-BB concentrations were log transformed, and tested for differences between different groups by two-sample Wilcoxon rank-sum test. These analyses were performed using SPSS software (version 17.0, SPSS Inc., Chicago, IL). SUPPLEMENTARY FIGURE AND TABLES {#s5} =============================== We thank all the tested individuals, their families, and collaborating clinicians for their participation. **CONFLICTS OF INTEREST** The authors declare that they have no conflict of interest. **FINANCIAL SUPPORT** This work was supported in part by grants from the National Natural Science Foundation (No. 81222037, 81473023), China Mega-Project on Infectious Disease Prevention (No. 2013ZX10004202) and the Special Fund for Quarantine-Scientific Research in the Public Interest (201310076).

See related research by Meyer *et al*., <http://arthritis-research.com/content/15/6/R210> and related letter by Toussirot, <http://arthritis-research.com/content/16/2/407> We thank Dr Toussirot for his interest \[[@B1]\] in our work demonstrating that serum-level adiponectin is associated with subsequent radiographic progression in early rheumatoid arthritis (RA) \[[@B2]\]. We would like to respond to each comment. First, our objective was not to determine whether the serum adipokine level might reflect RA disease activity cross-sectionally, but was to find surrogate markers able to predict structural radiographic progression. We performed such an analysis and found no correlation between any serum adipokine levels and the disease activity score in 28 joints (data not shown). Concerning the second and third points, the association we found between the total adiponectin concentration and radiographic progression does not provide any direct indications about any functional roles of this adipokine in RA. Despite an anti-inflammatory role of adiponectin, adiponectin isoforms are proinflammatory on RA synovial cells, in accordance with our results \[[@B3]\]. Moreover, although adiponectin may be protective in collagen-induced arthritis, its proinflammatory effect is well known in other inflammation models \[[@B4]\]. Concerning the fourth point about potential discrepancies between a recently published cross-sectional study \[[@B5]\] and our own work, the comparison is challenging since we have not assessed the high molecular weight (HMW) isoform and have not investigated healthy control subjects. Moreover, Toussirot and colleagues did not study the structural progression. Finally, they investigated treated patients with established RA, while we focused on untreated patients with early RA. Furthermore, the use of an enzyme-linked immunosorbent assay for HMW assessment and a radioimmunoassay for total adiponectin assessment may explain the absence of correlation between both measurements. Recently, a high correlation between both isoforms using enzyme-linked immunosorbent assay for both measurements has been reported in RA \[[@B6]\]. Finally, we fully agree on discrepancies between published studies investigating the serum adiponectin level in RA as noted in the fifth point by Toussirot \[[@B1]\]. Although this can be due to the need to assess HMW rather than total adiponectin, the high correlation using an enzyme-linked immunosorbent assay does not support such a hypothesis. The demographic characteristics, the adjustment for confounding factors, and the sample size of the population may explain these divergences. Of note, our study involved the largest group of early RA adipokine measurements to date with multiple adjustments. In conclusion, while our study has emphasized the usefulness of serum total adiponectin measurement as an accurate biomarker predicting radiographic progression, additional studies are necessary to establish whether the serum HMW adiponectin measurement may be more useful for such a purpose. Abbreviations ============= HMW: High molecular weight; RA: Rheumatoid arthritis. Competing interests =================== The authors declare that they have no competing interests. Author's contributions ====================== JS and FB wrote the letter. JC, J-PB and SF gave their advice about adiponectin isoforms and their assessment. All authors reviewed and approved the final manuscript.

Introduction ============ Caustic soda is an alkaline-based substance. Tissue contact with alkaline pH leads to liquefactive necrosis. Distal esophageal injury is more likely after ingestion of liquids, rather than solids, as they travel further.[@b1-tcrm-12-129] The immediate injury consists of acute necrosis, hyperemia, and edema of the aerodigestive tract, leading to vascular thrombosis. In all, 2--4 days post injury, the superficial layer of necrotic mucosa will be sloughed. First, this may lead to either perforations or scarring and, second, contractures due to fibroblast regeneration and production of collagen. Some patients develop long-term problems, principally oral, pharyngeal, and esophageal strictures. Treatment schemes for caustic injuries vary, and no consensus exists. Multiple protocols were formulated based on documented success in preventing impending complications, although a paucity of randomized trials have assessed any of these interventions.[@b2-tcrm-12-129] These patients develop malnutrition if they do not follow a personalized nutritional support, but no guidelines are available. A key point is the fitting choice of a main nutritional support via percutaneous endoscopic gastrostomy/percutaneous endoscopic jejunostomy (PEJ) or nasogastric tube. Another aspect to consider is the use of specific formulas to integrate the fitting nutritional needs for each patient. Hence, the monitoring must be considered to determine the transition from enteral to oral nutrition and the route of administration of the food more adequately. Given this background, this paper reports the nutritional management of an elderly man who swallowed caustic soda in a suicidal attempt, causing the development of strong esophageal and gastric necrosis with subsequent gastrectomy and digiunostomy. Case report =========== A 70-year-old man was hospitalized in the Institute of Geriatric Rehabilitation Santa Margherita, Pavia, on August 1, 2013, with this admission diagnosis: patient with PEJ from April 2013, esophageal and gastric necrosis from caustic soda ingestion with results of gastrectomy and jejunostomy. In conjunction with a clinical evaluation based on patient history, it was take into account a depressive syndrome, total esophagectomy, pharyngolaryngectomy, pharyngeal-colon-ileal anastomosis, bearer of tracheostomy, and with an history of previous cholecystectomy and appendectomy. Written informed consent was obtained from the patient as well as this case report was performed according to the Declaration of Helsinki and approved by Institute of Geriatric Rehabilitation Santa Margherita institutional review board. First stage of nutritional intervention --------------------------------------- At the initial nutritional assessment on admission, the patient was categorized as "patient with a severe malnutrition", with serum albumin of 3.15 mg/dL, body mass index (BMI) of 17.3 kg/m^2^ (50 kg for 1.70 m), Mini Nutritional Assessment (MNA) equal to 10 points, and weight loss of \~5.2% in the previous 6 months. The score of Activities of Daily Living (ADL) was 2 points, which showed that nutritional difficulties restricted the ability to perform self-care activities. In particular, the patient was not able to maintain an independent life, such as needs help in bathing and in moving from bed to chair. Other parameters, at baseline and in the time progression, are shown in [Table 1](#t1-tcrm-12-129){ref-type="table"}. The nutrition support team recommended to continue the use of the PEJ; therefore, the patient received a polymer and high-caloric formula, with an elevated content of soluble fibers from partially hydrolyzed guar gum (PHGG) (500 mL ×2, 1.5 kcal/mL, 6.0 g proteins/100 mL, 18.3 g carbohydrates/100 mL, 5.9 g lipids/100 mL, and 2.2 g PHGG/100 mL), indicated for patients with a sole source of nutrition on medical grounds (eg, short bowel syndrome), following total gastrectomy and disease-related malnutrition. Other nutritional characteristics of this formula are listed in [Table 2](#t2-tcrm-12-129){ref-type="table"}. The elected speed of the enteral nutrition (EN) was 80 mL/h, because the patient was already under artificial nutrition from April 2013. The formula was well tolerated during hospitalization (2 months), there were no episodes of diarrhea or vomiting, and the weight remained stable. Second stage of nutritional intervention ---------------------------------------- The patient was again hospitalized on November 5, 2013, for a medium-/long-term complication, which was local infection, pain, diarrhea, and malnutrition. At the nutritional assessment, the following parameters were detected: serum albumin of 3.76 mg/dL, BMI of 17.3 kg/m^2^ (50 kg for 1.70 m, weight stable), and MNA equal to 13 points. There was no weight loss between the two admissions; then the nutrition support team recommended to continue the use of the polymer and high-caloric formula, with an elevated content of soluble fiber, PHGG (500 mL ×2). The elected speed of the EN was 60 mL/h (initially decreased due to diarrhea) and was recommended to increase the speed of 20 mL/h up to 120 mL/h, given the good tolerance of the patient to the formula in the previous admission. The patient responded well and continued this formula, by increasing the speed for 1 month. Third stage of nutritional intervention --------------------------------------- On December 10, 2013, a computed tomography was performed to exclude the presence of adhesions and the regular passage of the bolus. Hence, the medical team removed the PEJ, and the patient switched from EN to oral nutrition. At the nutritional assessment, the serum albumin was 3.34 g/dL, the BMI was 17.3 kg/m^2^ (50 kg for 1.70 m, weight stable), and the MNA was equal to 16 points. Other parameters are shown in [Table 1](#t1-tcrm-12-129){ref-type="table"}. The patient reported that he began to do breakfast with a cup of milk and rusks, eating fruits and vegetables two times a day, and meat and fish every day. Generally, the risk of decline in ADL increases markedly in this stage, then, the nutrition support team decided to introduce two oral high-caloric supplements: an energy supplement in powder, based on maltodextrin, immediately soluble in foods, hot/cold drinks, and sweet/savory drinks (six servings a day with the measuring cup; 5 g for three measuring cups in the first course at lunch and three measuring cups in the first course at dinner, 120 kcal) and a liquid energy--protein supplement (a formula with high energy, protein, and enriched with arginine, vitamin C, zinc, and antioxidants). After 1 month, on January 15, 2014, the patient was discharged from the institute and was able to eat a regular meal instead of artificial nutrition. The ADL score of 5 points showed that the patient was able to undertake self-care. In summary, the progression in time of the nutritional intervention from artificial to oral nutrition is shown in [Table 3](#t3-tcrm-12-129){ref-type="table"}. Discussion ========== The ingestion of corrosive industrial chemical agents, such as caustic soda, that are mostly used for household cleaning, usually occurs accidentally or sometimes for suicidal purposes. Nutritional interventions in these situations may be important for life expectancy. At the first stage, nutritional management must be set up on EN. In a patient with esophageal and gastric necrosis, the use of a high-protein-specific formula with a similar caloric percentage of fat and carbohydrates brings about an improvement in both nutritional status and biochemical parameters. This formula should bring about at least 150% of recommended daily amount (RDA) intake in vitamin C (systemic inflammation and damage attenuations both in vivo and in vitro);[@b3-tcrm-12-129] 100% of RDA intake in vitamin D (effects on the biomarkers of inflammation and endothelial activation \[high sensitivity C-reactive protein and endothelin-1\] and flow-mediated dilation);[@b4-tcrm-12-129] and 150% of RDA of folate, vitamins B1, B2, and B12. As shown in a recent study, these vitamins decrease the marked inflammation and apoptosis, despite activation of repair machinery.[@b5-tcrm-12-129] In addition, the intake of minerals, zinc in particular, might suppress the generation of inflammatory cytokines and reactive oxygen species damage.[@b6-tcrm-12-129] The advantage of this approach is its safety in terms of infection. For the first 3 months, the EN infusion speed should be stable at \~80 mL/h. In this case report, the best improvements are highlighted 3 months later, when the infusion speed was set up to 120 mL/h, in the absence of diarrhea. Most previous works concerning the characteristics and the EN infusion speed have concentrated on the risk of infectious of diarrhea, due to either the enteral feeding equipment or the sterility of the formula.[@b7-tcrm-12-129] In our study, after 4 months, we removed the PEJ and we decided to introduce two oral, high-caloric daily supplements: an energy supplement in powder, based on maltodextrin, immediately soluble in foods (or in hot/cold drinks) and a high-energy and protein drink, enriched with arginine, vitamin C, zinc, and antioxidant. Previous to our experience and considering the literature, absorption of glucose and fat from the maltodextrin diet was significantly greater than that from the glucose diet, whereas absorption of protein was only slightly enhanced.[@b8-tcrm-12-129] In addition, introducing different antioxidants such as α-, β-, γ-carotene, lycopene, luteina, and zeaxantina may improve tissue repairing process.[@b9-tcrm-12-129] Initially, we considered a supplementation of omega 3, following The American Dietetic Association Guidelines that recommend consuming 500 mg/d of a combination of EPA (eicosapentaenoic acid) and DHA (docosahexaenoic acid).[@b10-tcrm-12-129] However, this procedure caused side effects, including upset stomach, loose stools, and belching. In addition to this, oral administration was well tolerated, thanks to consuming of homogenized fish (four times per week), poultry meal (three times per week), and fruits and vegetables every day, all diluted in water. In scientific literature, there are no guidelines for the nutritional support of patients with gastric and esophageal necrosis due to ingestion of corrosive industrial chemical agents for suicidal purpose. The correct choice of the most appropriate nutritional support varies from patient to patient, and many factors are determined. By using a personalized formula, through PEJ, the weight loss could reduce for some patients, preventing the transition to cachexia. Moreover, the choice of specific nutrients allows to repair tissues with necrosis and to decrease the inflammatory state. In all patients with no adhesions and the regular passage of the bolus, the removal of the PEJ is indicated, with the combination of homogenized meal and dietary supplements. Conclusion ========== In conclusion, this is the first case of a nutritional approach to follow patients with esophageal and gastric necrosis; it allows reducing the risks of malnutrition or infection, which can lead to negative outcome of the patient. **Disclosure** The authors report no conflicts of interest in this work. ###### Progression of clinical and laboratory parameters of the patient Parameters First stage -- August 2013 Second stage -- November 2013 Third stage -- December 2013 Reference standard values ---------------------- ---------------------------- ------------------------------- ------------------------------ --------------------------- Body weight, kg 50 50 50 53.5--72 BMI, kg/m^2^ 17.3 17.3 17.3 18.5--24.9 Total protein, g 6.1 6.9 6.1 6.0--8.2 Serum albumin, g/dL 3.15 3.76 3.34 3.5--4.8 Cholesterol, mg/dL 136 156 144 \<200 Lymphocyte, 10^3^/μL 1.95 1.72 1.93 0.8--3.6 Glycemia, mg/dL 111 93 78 70--110 MNA score 10 13 16 \>23.5 ADL score 2 -- 5 0--6 **Abbreviations:** BMI, body mass index; MNA, Mini Nutritional Assessment; ADL, activities of daily living. ###### Nutritional info concerning enteral nutrition formula Typical values Per 100 mL Per 500 mL ------------------------------ ------------ ------------ General  Energy, kJ/kcal 649/155 3,245/775  Protein (16% kcal), g 6 30  Carbohydrates (47% kcal), g 18.3 91.5   Sugars, g 1.6 8.0   Lactose, g \<0.05 \<0.25  Fat (34% kcal), g 5.9 29.5   Saturates, g 1.9 9.5   MCT, g 1.4 7   Monounsaturates, g 2.5 12.5   Polyunsaturates, g 1.1 5.5  Fiber (3% kcal), g 2.2 11   Soluble, g 2.2 11 Vitamins  A, μg 160 800   β-carotene, μg 38 190  D, μg 2.2 11  K, μg 11 55  C, mg 16 80  B1 (thiamin), mg 0.22 1.1  B2 (riboflavin), mg 0.26 1.3  B6, mg 0.27 1.35  Niacin, mg 3 15  Folic acid, μg 45 225  B12, mg 0.58 2.9  Pantothenic acid, mg 0.92 4.6  Biotin, μg 7 35  E, mg 2.9 14.5 Minerals  Sodium, mg/mmol 120/5.22 600/26.1  Chloride, mg/mmol 150/4.23 750/21.15  Potassium, mg/mmol 135/3.46 675/17.3  Calcium, mg/mmol 80/2 400/10  Phosphorus, mg/mmol 75/2.42 375/12.10  Magnesium, mg/mmol 30/1.25 150/6.25  Iron, mg 1.6 8  Zinc, mg 1.8 9  Copper, μg 240 1,200  Iodine, μg 22 110  Selenium, μg 10 50  Manganese, mg 0.36 1.8  Chromium, μg 15 75  Molybdenum, μg 20 100  Fluoride, mg 0.16 0.8 Other nutrients  Choline, mg 57 285  Water, g 76 -- Osmolarity, mOsm/L 389 -- Osmolarity, mOsm/kg 440 -- **Abbreviation:** MCT, medium chain triglyceride. ###### Progression of nutritional support of the patient Hospital stay Clinical aspects Oral diet Enteral feeding/ONS --------------- ------------------------------------------------------------- --------------------------------------------------------------------------------------------------------------------------------------------- ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- August 2013 No episodes of diarrhea or vomiting; weight remained stable -- Jejunal feeding: polymeric, high-caloric formula with an elevated content of soluble fiber partially hydrolyzed guar gum (PHGG) with a speed of 80 mL/h November 2013 No episodes of diarrhea or vomiting; weight remained stable -- Jejunal feeding: polymeric, high-caloric formula with an elevated content of soluble fiber PHGG with an initial speed of 60 mL/h and up to 120 mL/h December 2013 Subjective well-being Regular meal: through the consume of homogenized fish (four times per week), poultry meal (three times per week), and fruits and vegetables Two high-caloric supplements: an energy supplement in powder based on maltodextrin and a high-energy and protein drink, enriched with arginine, vitamin C, zinc, and antioxidants **Abbreviation:** ONS, oral nutritional supplements.

Background ========== Protein secretion into the extra-cellular environment is one of most desirable strategy to allow a rapid and not expensive recovery of recombinant proteins. Secretion to the culture medium has several advantages over intracellular recombinant protein production. These advantages include simplified downstream processing, enhanced biological activity, higher product stability and solubility, and N-terminal authenticity of the expressed peptide \[[@B1]-[@B3]\]. If the product is secreted to the culture medium, cell disruption is not required for recovery. As bacteria, usually, do not secrete amounts of proteins higher then they have in the intracellular space, recovery of a recombinant gene product can be greatly simplified by a secretion strategy that minimises contamination from host proteins. Additionally, secretion can provide a method to guarantee the N-terminal authenticity of the expressed polypeptide because it often involves the cleavage of a signal sequence \[[@B4]\], thus avoiding the presence of an unwanted initial methionine on a protein that does not normally contain it. This extra methionine can reduce the biological activity and stability of the product \[[@B5]\] or even elicit an immunogenic response in the case of therapeutic proteins. In a previous paper \[[@B6]\], we reported the realization of a \"cold\" recombinant secretion system in the Antarctic Gram-negative bacterium *P. haloplanktis*TAC125. This system efficiently conjugates the obvious advantages of extra-cellular protein targeting with the positive effect of low temperature on the recombinant product solubility. Indeed, low expression temperature can facilitate the correct folding of \"difficult\" products \[[@B7],[@B8]\] and the use of *P. haloplanktis*TAC125 as expression system \[[@B9]\] allowed the efficient production of some \"intractable\" proteins in soluble and active form at temperature as low as 4°C \[[@B10]-[@B12]\]. The cold-adapted secretion system \[[@B6]\] makes use of the psychrophilic α-amylase from *P. haloplanktis*TAB23 \[[@B13],[@B14]\] as secretion carrier. Three chimerical proteins, made of the psychrophilic α-amylase fused to an intra-cellular protein, were translocated in the extra-cellular medium with a secretion yield always higher than 80%. The system also allowed the correct disulphide bond formation of chimera components, secreting a fully active passenger \[[@B6]\]. However, our previous results addressed to a potential limit of this newly set up technology: host extra-cellular medium may contain proteolytic activities which can affect the quality of heterologous products \[[@B6]\]. This feature could hamper the applicability of the cold-adapted secretion system, due to the likely recombinant product degradation. To overcome this host limit, two possible approaches can be pursued thanks to the availability of *P. haloplanktis*TAC125 genome sequencing and annotation \[[@B15]\]: i) the gene disruption of each *P. haloplanktis*TAC125 gene encoding extra-cellular proteases; or ii) the inactivation of the secretion machinery responsible for the proteases extra-cellular targeting. The *in silico*analysis of *P. haloplanktis*TAC125 genome demonstrated that the bacterium possesses only one canonical secretion system, a putative Type II secretion machinery (T2SS) also called General Secretory Pathway (GSP), homologous to GSPs already described in many other Gram-negative bacteria \[[@B16]\]. Since experimental evidences suggested that in *P. haloplanktis*TAC125 the secretion of the cold α-amylase depends on a still uncharacterized pathway (unpublished results from this laboratory), the inactivation of the T2SS machinery seemed a reasonable strategy to develop a *P. haloplanktis*TAC125 mutant strain with reduced extra-cellular proteolytic activity. In this paper we report the set-up of an integrative plasmid and its use for the construction of a *P. haloplanktis*TAC125 strain in which the *gspE*gene \[[@B17],[@B18]\] was knocked-out. This mutation resulted in the inactivation of the psychrophilic T2SS system. The *P. haloplanktis*TAC125 *gspE*mutant displayed a remarkable reduction of the extra-cellular protease secretion, still maintaining its ability to secrete the psychrophilic amylase (the secretion carrier of our recombinant system) as efficiently as the wild type. These features make the *P. haloplanktis*TAC125 *gspE*mutant strain an improved host with a remarkable biotechnological potential in recombinant protein secretion at low temperature. Results ======= *Pseudoalteromonas haloplanktis*TAC125 growth medium contains several secreted proteases ---------------------------------------------------------------------------------------- Wild type *P. haloplanktis*TAC125 cells were grown in TYP medium at 4°C and culture medium samples were withdrawn at different growth phases (at 24, and 32 hours of incubation corresponding to early and medium exponential phase). Concentrated culture supernatants were analyzed for proteolytic activities using Gelatine-SDS-PAGE (10% acrylamide, w/v) as described in Materials and Methods. As shown in Figure [1A](#F1){ref-type="fig"}, the wild type *P. haloplanktis*TAC125 culture supernatants contain a wide range of proteolytic activities, which display an apparent molecular weight ranging between 120 and 33 kDa. {#F1} A protease inhibition assay was performed by treating the 24 hours extra-cellular protein sample with 10 mM PMSF (a serine protease inhibitor) for 20 hours at 15°C, 10 mM EDTA (a metalloprotease inhibitor) for 20 hours at 15°C, and with the combination of the two inhibitors. The differentially treated samples were then subjected to gelatine zymography, and the results are shown in Figure [1B](#F1){ref-type="fig"}. A comparative evaluation of the results demonstrates that the wild type *P. haloplanktis*TAC125 culture supernatant contains extra-cellular proteases inhibited either by PMSF or by EDTA, and proteolytic activities which appear to be resistant to both the inhibitors (Figure [1B](#F1){ref-type="fig"} lane EDTA+PMSF). *Pseudoalteromonas haloplanktis*TAC125 genome encodes a functional Type II protein secretion system --------------------------------------------------------------------------------------------------- Annotation of *P. haloplanktis*TAC125 genome sequence \[[@B15]\] revealed that the psychrophilic bacterium possesses only a canonical extra-cellular protein secretion pathway, i.e. the Type II secretion system (T2SS). The psychrophilic T2SS is located on the larger *P. haloplanktis*TAC125 chromosome and it is made up of twelve genes, from the gene PSHAa0231 to PSHAa0242. The *gsp*gene cluster is likely arranged in several independent transcriptional units, as previously observed in other Gram-negative bacteria \[[@B19],[@B20]\]. Gene expression of *P. haloplanktis*TAC125 T2SS encoding cluster was evaluated by Reverse Transcriptase (RT)-PCR. Total RNA was extracted from cell samples collected at different growth phases, and was used to assess the transcription of *gspE*, *gspC*, and *gspN*. All the tested *gsp*genes resulted to be constitutively expressed in *P. haloplanktis*TAC125 cells (data not shown). Construction and genetic characterization of *Pseudoalteromonas haloplanktis*TAC125 *gspE*mutant strain ------------------------------------------------------------------------------------------------------- Functional inactivation of *P. haloplanktis*TAC125 T2SS system was achieved by insertional mutagenesis of *gspE*gene. This target was selected because it encodes an inner membrane-associated ATP-synthase, which has previously been reported to be essential for the T2SS functioning in other bacteria \[[@B21]\]. Insertional mutagenesis was obtained by using a suicide vector (pVS), suitably constructed for *P. haloplanktis*TAC125. As shown in Figure [2A](#F2){ref-type="fig"}, the pVS vector is characterized by the presence of: i) the pJB3-derived *oriT*\[[@B22]\], a DNA fragment responsible for the initiation of the conjugative transfer between an *Escherichia coli*S17-1 λ*pir*strain (donor) and the psychrophilic cells (acceptor); ii) the *E. coli blaM*gene, encoding a mesophilic β-lactamase which is used as selection gene to isolate the first site-specific integration event; iii) *pheS*^Gly294^, which encodes a mutated version of the *E. coli α*subunit of Phe-tRNA synthase \[[@B23]\], which renders bacteria sensitive to *p-*chlorophenylalanine. This phenylalanine analogue can be used as counterselective agent for the isolation of those strains in which a second recombination event occurred. {#F2} Two *gspE*gene fragments were amplified by PCR using specific oligonucleotides as primers. They correspond to two internal gene fragments and they are not adjacent (Figure [2A](#F2){ref-type="fig"}). The fragments were suitably digested and cloned into the pVS vector. The resulting vector (pVS*gspE*) was mobilized into *P. haloplanktis*TAC125, and clones in which a single recombination event occurred were selected on carbenicellin containing solid medium. Genomic DNA from some carbenicellin-resistant clones was extracted and subjected to different PCR analyses, to characterize the occurred genomic insertion. A positive clone was selected for further characterization, and its genome organization is reported in Figure [2B](#F2){ref-type="fig"}. *P. haloplanktis*TAC125::VS*gspE*mutant (hereafter called *P. haloplanktis*TAC125 *gspE*mutant) resulted to contain two copies of *gspE*gene, both carrying a specific deletion that was checked by sequencing the specific PCR products shown in Figure [2B](#F2){ref-type="fig"}. The first copy of *gspE*gene differs from the wild type one in lacking i) the region (450 bp long) between the two amplified fragments and ii) the 3\' encoding region (Figure [2B](#F2){ref-type="fig"}). A transcriptional analysis demonstrated that this *gspE*copy is transcribed (data not shown), but due to the large deletion, the resulting gene contains a translation frame shift leading to the production of an abnormal protein. The other *gspE*gene copy (Figure [2B](#F2){ref-type="fig"}) lacks of its promoter region and of the 5\' encoding portion, and therefore it resulted to be not transcribed (data not shown). *P. haloplanktis*TAC125::VS*gspE*mutant was subjected to *p-*chlorophenylalanine treatment to select a clean deletion mutant. Although the counter-selection was carried out in several experimental conditions, all the selected clones resulted to be meroploid strains, containing both inserted and wild type alleles (data not shown). *Pseudoalteromonas haloplanktis*TAC125 *gspE*mutant strain displays a reduced extra-cellular protease activity -------------------------------------------------------------------------------------------------------------- Growth behaviour of *P. haloplanktis*TAC125 *gspE*mutant strain in standard conditions was determined, and compared with the wild type one. As shown in Figure [3A](#F3){ref-type="fig"}, the *gspE*mutant strain grows faster and makes a higher cellular biomass than the wild type strain. Concentrated culture supernatants of *P. haloplanktis*TAC125 wild type and *gspE*mutant were analyzed by SDS-PAGE and results are shown in Figure [3B](#F3){ref-type="fig"}. The *gspE*mutant culture supernatant contains a reduced number of proteins compared to the wild type. {#F3} Extra-cellular protease secretion in *P. haloplanktis*TAC125 *gspE*mutant strain was investigated by gelatine zymography of concentrated culture supernatants. As shown in Figure [1C](#F1){ref-type="fig"}, the samples contain a notably reduced number of proteolytic activities as compared to the protease content of wild type samples (Figure [1A](#F1){ref-type="fig"}) at the corresponding growth phases. Interestingly, the combined EDTA/PMSF treatment resulted in the almost complete inhibition of the extra-cellular proteases secreted by *gspE*mutant (Figure [1D](#F1){ref-type="fig"}). *Pseudoalteromonas haloplanktis*TAC125 *gspE*mutant strain is an improved host for the recombinant protein secretion at low temperature --------------------------------------------------------------------------------------------------------------------------------------- Secretion of psychrophilic α-amylase in *P. haloplanktis*TAC125 *gspE*mutant cells was studied. The mutant strain was transformed with pFC*amy*ΔCt, a psychrophilic vector previously constructed for the recombinant secretion of α-amylase in the Antarctic bacterium \[[@B6]\]. The recombinant mutant strain was grown at 4°C till medium exponential phase and the α-amylase secretion was evaluated by Western blotting analysis of cellular (Figure [4](#F4){ref-type="fig"}, lanes 3) and extra-cellular (Figure [4](#F4){ref-type="fig"}, lanes 4) protein samples. The results demonstrated that the *gspE*mutation does not affect secretion of the psychrophilic enzyme. {#F4} The secretion of the chimerical protein AmyΔCt-BlaM, made up of the psychrophilic amylase fused to the mature β-lactamase \[[@B6]\], in *P. haloplanktis*TAC125 *gspE*mutant cells was investigated. *P. haloplanktis*TAC125 *gspE*mutant-(pFC*amy*Δ*Ct-blaM*) recombinant cells were grown in TYP medium at 4°C and samples were collected at different growth phases (early, and medium exponential phase). Culture supernatants were analysed by Western blotting analysis using anti-β-lactamase (anti-*Ec*Bla) and anti-α-amylase (anti*Ph*α-Amy) polyclonal antisera, respectively. As shown in Figure [5B](#F5){ref-type="fig"} lanes 3 and 4, both antisera detected a single product, with an apparent molecular weight of 80 kDa, corresponding to the AmyΔCt-BlaM chimerical protein. When a similar analysis was applied to wild type *P. haloplanktis*TAC125-(pFC*amy*Δ*Ct-blaM*) recombinant cells (Figure [5A](#F5){ref-type="fig"} lane 1 and 2), culture supernatants contained either the chimerical protein or its different proteolysis products, i.e. AmyΔCt, BlaM, and a chimera truncated form due to the action of host-encoded secreted proteases on the passenger protein. {#F5} Discussion ========== In a previous paper \[[@B6]\], we reported the establishing of a versatile gene-expression system for secretion of heterologous proteins in *P. haloplanktis*TAC125. The system uses the psychrophilic α-amylase from *P. haloplanktis*TAB23 as secretion carrier, and allows an efficient extra-cellular addressing of recombinant proteins. However, we realized that the system efficiency was hampered by the presence of host-encoded extra-cellular proteolytic activities \[[@B6]\]. The presence of extra-cellular proteases represents a severe limit to the use of any genetic system for the recombinant protein secretion, since host-encoded proteases can affect the quality of the heterologous secreted products. To enhance the recombinant product quality and stability, we focused our attention on the extra-cellular proteases produced by the psychrophilic *P. haloplanktis*TAC125 with the aim of developing a host mutant strain with reduced extra-cellular proteolytic activity. By in gel activity assay, we demonstrated that several proteases are present in the culture medium of the psychrophilic bacterium (Figure [1A](#F1){ref-type="fig"}). The zymographies presented in Figure [1](#F1){ref-type="fig"} did not allow us to define the exact number of extra-cellular proteases, due to the low resolution of this experimental technique. However, the inhibition assay allowed us to explore at least the diversity in the action mechanisms of the psychrophilic extra-cellular proteases. Indeed, metalloproteases, serine proteases and enzymes belonging to different classes were detected (Figure [1B](#F1){ref-type="fig"}). Considering the multiplicity and heterogeneity of proteases detected in *P. haloplanktis*TAC125 culture medium, the systematic disruption of each extra-cellular protease encoding gene appears a time-consuming strategy. Therefore, we decided to inactivate the molecular machinery responsible for the extra-cellular targeting of proteases. This aim was achieved combining some information deriving from the *in silico*analysis of *P. haloplanktis*TAC125 genome with several experimental evidences. In particular, from the accurate genome annotation, it was known that the psychrophilic bacterium possesses only one canonical secretion machinery (the T2SS pathway), while we have collected evidences indicating that the psychrophilic α-amylase (the secretion carrier of our recombinant system) is likely secreted by another secretion apparatus not yet fully characterized (unpublished results from this laboratory). Thus, the functional inactivation of *P. haloplanktis*TAC125 T2SS seemed a feasible approach to generate a psychrophilic mutant strain possibly secreting a lower protease amount but still able to secrete the recombinant α-amylase and its chimerical derivates. T2SS (also called General Secretory Pathway) is a multi-component machinery encoded by the *gsp*cluster and promoting secretion of Sec- and Tat-dependent exo-enzymes in a two-step process \[[@B16],[@B19],[@B20]\]. Our results demonstrated that the *P. haloplanktis*TAC125 *gsp*gene cluster is actually transcribed and that the *gsp*genes expression seems to be constitutive over the bacterial growth (data not shown). Therefore it was necessary to use a genetic approach aimed at completely abolishing *gsp*-dependent secretion during all growth phases. *gspE*Gene \[[@B17],[@B18]\] encodes a specialized ATP-synthase whose inactivation resulted in the total loss of T2SS functionality in other Gram-negative bacteria \[[@B21],[@B24],[@B25]\]. Therefore the psychrophilic *gspE*gene was selected as target for inactivation by insertional mutagenesis. The *P. haloplanktis*TAC125 *gspE*mutant strain was constructed by applying a typical gene targeting strategy, which makes use of a suitably constructed psychrophilic suicide vector (Figure [2A](#F2){ref-type="fig"}). As described in result section, this mutagenesis strategy allowed the creation of a *P. haloplanktis*TAC125 mutant strain characterized by the absence of GspE function. A preliminary phenotypic analysis, carried out in standard growth conditions, demonstrated that *P. haloplanktis*TAC125 *gspE*mutant displays specific growth rate and biomass productivity higher than the wild type strain, thus surprisingly the mutation does not affect but improve the bacterium fitness (Figure [3A](#F3){ref-type="fig"}). This observed *gspE*mutant behaviour could be justified considering that *gsp*-dependent protein secretion is a high costly metabolic process, and its inhibition may represent an advantage to *gspE*mutant strain, at least in the tested growth condition. The analysis of the extra-cellular protein content demonstrated that *gspE*mutant actually secretes lower amounts of proteins with respect to the wild type strain (Figure [3B](#F3){ref-type="fig"}). Proteins secreted by the *gspE*mutant are translocated by secretion pathways different from T2SS, such as the specialized machineries for type IV pili and curli components secretion \[[@B15]\] and the secretion apparatus responsible for the recombinant α-amylase secretion (unpublished results from this laboratory). Zymographic analysis revealed that the *gspE*mutant strain medium contains a notably reduced number of proteolytic activities (Figure [1C](#F1){ref-type="fig"}) with respect to wild type culture supernatants (Figure [1A](#F1){ref-type="fig"}). In contrast, knocking out of *gspE*gene did not impair for secretion of the cold-adapted amylase (Figure [4](#F4){ref-type="fig"}). In fact, *P. haloplanktis*TAC125 wild type and *gspE*mutant strains resulted to be indistinguishable in terms of cold-adapted α-amylase production and secretion yields. The *gspE*mutant was further tested for its ability to secrete a α-amylase chimerical product, the AmyΔCt-BlaM chimera. When produced by wild type cells, the chimera components (i.e. the psychrophilic α-amylase and the mesophilic β-lactamase) (Figure [5A](#F5){ref-type="fig"}, lanes 1 and 2) are partially separated due to the sub-stoichiometric proteolytic cleavage of the linker which connects the two proteins (see also \[[@B6]\]). This processing does not affect the catalytic activity of each component of the chimera \[[@B6]\]. The host-encoded secreted proteases are also responsible for further chimera degradation (see \"chimera truncated form\" in figure [5A](#F5){ref-type="fig"}, lanes 1 and 2) which accounts for the previously reported and undesired decrease of passenger activity \[[@B6]\]. On the contrary, due to the reduced number of extra-cellular proteases present in culture medium, AmyΔCt-BlaM chimera accumulates as a unique unprocessed form when produced in *P. haloplanktis*TAC125 *gspE*mutant cells (Figure [5B](#F5){ref-type="fig"}, lanes 3 and 4), thus resulting in an enhancement of chimera quality over the whole production process. Conclusion ========== Combining the experimental evidences we collected on the α-amylase secretion machinery in *P. haloplanktis*TAC125 with the careful *in silico*analysis of its genome, we designed an simple and successful experimental approach for the construction of an improved psychrophilic host for the cold α-amylase-dependent recombinant secretion system. In fact, by a single gene disruption, involving the psychrophilic *gspE*gene, we developed a *P. haloplanktis*TAC125 mutant strain which secretes a significantly reduced extra-cellular protease activity while keeping its ability to secrete the recombinant psychrophilic α-amylase as the wild type strain. The *gspE*mutant strain is also characterized by specific growth rate and biomass productivity higher than wild type strain, making it a truly improved host with a remarkable biotechnological potential in recombinant protein secretion at low temperature. Moreover, this work demonstrates that *P. haloplanktis*TAC125 is a versatile psychrophilic host for recombinant protein production since it that can be easily improved by a directed engineering approach. To the best of our knowledge, this is the first described example of a strain improvement strategy applied to an Antarctic bacterium. Methods ======= Strains and plasmids -------------------- *P. haloplanktis*TAC125 was isolated from Antarctic sea water \[[@B15]\]. *Escherichia coli*DH5α \[[@B26]\] was used as host for the gene cloning. *E. coli*strain S17-1(λ*pir*) was used as donor in interspecific conjugation experiments \[[@B27]\]. Growth conditions and analytical procedures ------------------------------------------- *P. haloplanktis*TAC125 was grown in aerobic conditions at 4°C in TYP broth (16 gr/L yeast extract, 16 gr/L bacto tryptone, 10 gr/L marine mix) at pH 7.5, supplemented with ampicillin 200 μg/ml, chloramphenicol 25 μg/ml, or cabenicellin 30 μg/ml, when required. Antarctic bacteria transformation was achieved by intergeneric conjugation as previously reported \[[@B9]\]. *E. coli*cells were routinely grown in Terrific broth \[[@B28]\] at 37°C. When required, antibiotics were added at the following concentrations in liquid cultures: 100 μg/ml of ampicillin, or cloramphenicol at 50 μg/ml final concentration. Genetic manipulations were carried out following standard procedures \[[@B28]\]. *P. haloplanktis*TAC125 DNA genomic purification was performed by ChargeSwitch gDNA Mini Bacteria Kit (Invitrogen). Protein samples were analyzed by Polyacrylamide Gel Electrophoresis (Sodium Dodecyl Sulphate-PAGE) (12% acrylamide, w/v) according to standard methods \[[@B28]\]. For immunoblotting, the proteins were transferred to a polyvinylidene difluoride membrane (Immobilon PSQ, Millipore). For immunodetection of proteins, *P. haloplanktis*TAB23 anti-α-amylase \[[@B29]\] or anti-β-lactamase antisera were diluted in blocking buffer (phosphate buffer saline; 5% skimmed milk). Peroxidase conjugate anti-rabbit IgG (Sigma-Aldrich, USA) was used as secondary antibody. Proteins were detected by chemiluminescence\'s (Pierce, USA). Construction of suicide insertion vector pVS*gspE* -------------------------------------------------- pVS suicide vector was constructed by the insertion of the pJB3-derived *oriT*\[[@B22]\] and *pheS*^Gly294^gene \[[@B23]\] into the pGEM7Z vector. The *oriT*, responsible for the initiation of the conjugative transfer, was amplified on pJB3 vector by using Oligo oriTEcoRIfw and Oligo oriTSacIrv as primers (see Table [1](#T1){ref-type="table"}), and was cloned into *EcoRI*and *SacI*sites of pGEM7Z (pGEM7Z-OriT). *pheS*^Gly294^gene, which encodes a mutated version of the *E. coli α*subunit of Phe-tRNA synthase, was amplified using pKSS \[[@B23]\] vector as template. The PCR reaction was carried out using the oligonucleotide pair PheSSNfw and PheSXrv, designed to introduce *Nde*I and *Xba*I restriction sites. The amplified DNA fragment was subjected to double *Nde*I/*Xba*I digestion and cloned into pPM13 plasmid \[[@B11]\] corresponding sites generating pPM13-*pheS*^Gly294^vector. The DNA fragment, containing P13 promoter and *pheS*^Gly294^gene, was recovered from pPM13-*pheS*^Gly294^vector by *SmaI/EcoRI*digestion, filled in and cloned into the pGEM7Z-*oriT NaeI*restriction site, resulting in the construction of the pVS vector. ###### Plasmids and oligonucleotides used in this work *Plasmids* --------------------- -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ----------- pFC Psychrophilic gene-expression vector, containing the T/R box, the promoter and termination region of the *P. haloplanktis*TAC125 *asp*C gene and the chloramphenicol resistance gene \[6\] pFC*amy*ΔCt pFC containing a truncated version of the *amy*gene devoid of the C-terminal propeptide encoding portion \[6\] pFC*amy*ΔCt-*blaM* pFC containing the *amy*ΔCt-*blaM*gene encoding the chimerical protein Amy-BlaM \[6\] pVS pGEM7Z vector containing a conjugation transfer origin (oriT), the counter selectable marker pheS ^Gly294^and the ampicillin resistance gene This work *Oligonucleotides* Oligo oriT*Eco*RIfw 5\'-TTGAATTCTCGCACGATATACAGG-3\' Oligo oriT*Sac*Irv 5\'-AAGAGCTCTTGAAGACGAAAGGG-3\' Oligo gspE*Sph*Ifw 5\'-TTGCATGCATGCGCATCATCCGG-3\' Oligo gspE*Sac*Irv 5\'-AAGAGCTCCAATATCGAGCTTAGCC-3\' Oligo gspE*Sac*Ifw 5\'-TTGAGCTCCTAAAGTAGGTATGACC-3\' Oligo gspE*Eco*RIrv 5\'-AAGAATTCGTACACGGGCTACAGCC-3\' Oligo gspESrv 5\'-AAGAGCTCCTTCACTGAGCATCG-3\' Oligo gspEfw 5\'-GCAATTTAAGCAGCGCGAAGATG-3\' Oligo gspErv 5\'-ATCTAGGGCACGGTATTCAAATGC-3\' PheSSNfw 5\'-TTGTCGACATATGTCACATCTCGCAGAAC-3\' PheSXrv 5\'-CCTCTAGAGAATTTCATAATCTATTCCTGCC-3\' Two DNA fragments of *P. haloplanktis*TAC125 *gspE*gene were amplified by PCR using bacterial genomic DNA as template. Two primer pairs were designed to amplify a 567 bp region at the 5\' end (Oligo gspE*Sph*Ifw, Oligo gspE*Sac*Irv) and a 621 bp region at the 3\' (Oligo gspE*Sac*Ifw, Oligo gspE*Eco*RIrv) end of the *gspE*gene. The amplified DNA fragments were digested by *SphI/SacI*and *EcoRI/SacI*and cloned into the pVS *Sph*I/*Eco*RI site to generate the pVS*gspE*vector. The resulting vector was mobilized by intergeneric conjugation \[[@B9]\] into *P. haloplanktis*TAC125, and the cells were plated at 4°C on TYP solid medium containing 30 μg/ml carbenicellin to select those clones in which a single recombination event occurred. All PCR amplifications were performed in standard conditions \[[@B28]\]. The amplified fragments were cloned and their nucleotide sequences were checked to rule out the occurrence of any mutation during synthesis. Zymographic assay ----------------- *P. haloplanktis*TAC125 wild type and *gspE*mutant strains were grown in standard conditions and culture samples were collected at different growth phases as reported in the text. Samples were centrifuged at 10000 × *g*for 5 min at 4°C and the upper phase was collected for further analysis. The collected culture media were tenfold concentrated by Centricon (AMICON, exclusion size 5 kDa), and 12 μl were loaded onto a non reducing SDS-PAGE containing gelatine (1.5 mg ml^-1^). After electrophoresis, gel was soaked twice with 2.5% Triton X-100 (v/v) solution for a total of 60 min to remove SDS. The gel was then incubated in a developing buffer (50 mM Tris-HCl, pH 7.5, containing 5 mM CaCl~2~) for 12 or 18 (as indicated) hours at 15°C, rinsed with water, and stained with Coomassie blue R250. Areas of gelatine digestion, corresponding to proteolytic activities, were visualized as unstained regions in the gel. Protease inhibition assay ------------------------- Tenfold concentrated culture supernatants of *P. haloplanktis*TAC125 wild type and *gspE*^-^mutant cells were incubated with no inhibitors, or 10 mM EDTA, or 10 mM PMSF, or a combination of EDTA and PMSF (both at 10 mM final concentration) at 15°C for 20 hours. The samples were then subjected to protease zymographic assay. Authors\' contributions ======================= DDV and CC performed the experiments and helped to draft the manuscript. EP and MLT drafted the manuscript and designed and coordinated the study. All authors read and approved the manuscript Acknowledgements ================ We are grateful to Dr. Gerard Michel (LISM/IBSM CNRS 31 Chemin J. Aiguier 13402 Marseille) for critical reading of the manuscript and very helpful discussions. This work was supported by grants of Ministero dell\'Università e della Ricerca Scientifica (Progetti di Rilevante Interesse Nazionale 2006) and of Programma Nazionale di Ricerche in Antartide 2004.

1. Introduction {#sec1-sensors-19-04952} =============== To provide new services and offer new features with excellent quality, modern critical infrastructure such as power plants, smart grids, and water plants nowadays use ICT technologies \[[@B1-sensors-19-04952]\]. However, even if ICT technologies have made possible the provision of new services and new features, the types of connectivity have opened the door to a new wave of possible threats to critical installations. Extensive research has been performed in which the details of the vulnerability and security that affect SCADA systems have been analyzed \[[@B2-sensors-19-04952],[@B3-sensors-19-04952],[@B4-sensors-19-04952],[@B5-sensors-19-04952]\]. In Smart Grids, protocols and architectures are designed for very particular functions in SCADA communication systems. The SCADA acronym stands for Supervisor Control and Data Acquisition, and it allows the supervision and control of plants either remotely or locally using hardware and software dedicated to that system. With this system, the system analyzes, collects and processes data in real time. The DNP3 (Distributed Network Protocol) protocol is one of the most widely used network protocols in smart grid communication networks. This protocol is mostly used in academia and industry research projects, as it provides opportunity for customization as it is an open protocol. Based on this characteristics, any company can employ DNP3 developments that are compatible with their equipment. There are other protocols designed to control the operations of technical systems. In the case of the DNP3 protocol, it is based on the three layers model in the OSI 7 layer model as given in [Figure 1](#sensors-19-04952-f001){ref-type="fig"}. The DNP3 plays a vital role in the smart grid, and it is conventionally used in SCADA industrial processing, including both electricity and water distribution. For this reason, in this paper we focus on this protocol due to its importance in smart grids. The DNP3 protocol is based on a three-layered, enhanced performance architecture (EPA) reference model. The EPA defines the basic application functionality for the user layer, which is located between the OSI App Layer and App Program \[[@B6-sensors-19-04952],[@B7-sensors-19-04952]\]. According to [Figure 1](#sensors-19-04952-f001){ref-type="fig"}, the DNP3 protocol mainly consists of three layers: the application layer, the data-link layer, and the pseudo-transport layer. The DNP3 protocol facilitates reliable communication between the SCADA nodes; for example, the last layer is in charge of the transmission of enormous quantities of data \[[@B8-sensors-19-04952],[@B9-sensors-19-04952]\]. The proposed research emphasizes the DNP3 protocol by parsing its structure during the session, and we will develop an algorithm to analyze the vulnerabilities that will include modeling attacks on all layers, and we propose an intrusion detection system to detect those attacks. Normally, a SCADA system is made up of three main layers or components such as Information Technology Network (IT Network), Operation Technology Network (OT Network) and Process Control System Network (or Field Layer), as given in [Figure 2](#sensors-19-04952-f002){ref-type="fig"}. In recent years, the OT Networks were operated as separate networks or stand-alone system without being connected to public communication and infrastructures. Nevertheless, as the Internet nowadays provides data accessibility and connected services, the businesses in Smart Grid have turned to exploit those services. This situation creates a complex architecture where non-secure systems are added to existing systems without strong security. Ultimately, they are both more exposed to attackers, as the separation that had previously protected these systems is decreased \[[@B10-sensors-19-04952]\]. Unfortunately, with the presence of internet connections inside smart grids, the security risk is very high. After what happened in Ukraine, everyone knows that hackers can bring down the energy grid. Attacks often happen by using steps such as reconnaissance, which consists of gathering information about the targeted system; scanning, which is about finding any weakness or vulnerability in the system by looking for any open ports; and running a service through the port. Thirdly, the attacker exploits the system using the discovered vulnerability and then compromises it so that they can gain full control. Finally, the attacker tries to maintain access by which they will steal the data or damage the system whenever they want \[[@B11-sensors-19-04952],[@B12-sensors-19-04952]\]. These types of attacks are not easy to detect using traditional antivirus software, which detects malware using pattern matching or heuristic methods for obfuscated malware. Cyber-attacks have evolved in business-driven situations, because the criminal actors behind these attacks know where the system weaknesses reside, and they employ appropriate malware, especially ransomware. As previous experience shows, whenever there is a hack on a smart grid, people's life is in danger, and sometimes loss of life occurs. To tackle malware threats on the SCADA system, especially on the smart grid, this article presents an intrusion detection system that isn't based on signatures, like traditional antivirus software. The proposed solution is mainly based on the DNP3 protocol, where the engine parses the packet format and then we train it to learn whether the sample from the frame protocol has been compromised or is good. The advantage of this protocol is that it is used in oil/gas and water utilities, as well as in wastewater. Furthermore, it is broadly utilized in electric facilities such as catapults \[[@B13-sensors-19-04952]\]. The research contribution in this article includes the main steps as follows:An overall customized DNP3 protocol vulnerability exposure with reference to the original protocol.A new attack model using the DNP3 protocol that targets all layers; the attack follows three steps: (a) pre-attack step; (b) attack modeling defined on all layers; and (c) attack settings within DNP3 parameters and consequences.An algorithm that includes machine learning methods for data transformation and data process concept on SCADA/DNP3 protocol.A new cyber-attack algorithm targeting the SCADA/DNP3 system, and the visualization and classification process for an intrusion detection system (IDS). In this paper, [Section 2](#sec2-sensors-19-04952){ref-type="sec"} presents a description of existing solutions in the area, while [Section 3](#sec3-sensors-19-04952){ref-type="sec"} presents the proposed solution along with the methods and data collection in an experimental environment. The results of experimental testing and the discussion are provided in [Section 4](#sec4-sensors-19-04952){ref-type="sec"}, before concluding the work in [Section 5](#sec5-sensors-19-04952){ref-type="sec"}. 2. Related Work {#sec2-sensors-19-04952} =============== This section describes the works---including various solutions applied in SCADA/ICS/DNP3---in which IDS and other methods have been discussed. The Distributed Network Protocol (DNP3) is prevalent within critical infrastructure, especially in smart grids. Unfortunately, DNP3 has some vulnerabilities that have been exploited by hackers, and so SCADA systems would face serious problems \[[@B14-sensors-19-04952]\]. However, before exploring solutions in SCADA systems, there is a great deal of research and many excellent resdults in IoT networks that are very promising. Yang et al. \[[@B15-sensors-19-04952]\] proposed a security scheme in IoT-based healthcare systems. In this research, they proposed a self-adaptive access control together with a privacy-preserving smart IoT-based healthcare big data storage system. Further security approaches have been developed for IoT systems, as described in \[[@B16-sensors-19-04952],[@B17-sensors-19-04952],[@B18-sensors-19-04952]\]. With regard to machine learning methods, de Toledo et al. \[[@B19-sensors-19-04952]\] developed a method that encrypts the traffic using the DNP3 protocol. This study used supervised algorithms to classify messages from the same protocol using datasets from the medium voltage of substations using simulation methods. The generated traffic followed two-direction communication using an encryption mode based on the IPsec and ESP (transport mode) with the exclusion UDP mode. This experiment is the most widely used, as it provides a way of privately limiting the cost of IP bandwidth within networks per byte sent \[[@B20-sensors-19-04952]\]. Other techniques that have been proposed as solutions for protecting DNP3 traffic include statistical pattern recognition, classification-based real-time method with HTTP, FTP and SSH flow, TCP and TLS protocols \[[@B21-sensors-19-04952],[@B22-sensors-19-04952]\]. Widely known layered security methods that provide protection in SCADA networks have been developed, but these methods have numerous limitations with respect to their dependency on the protocols. Among others, protocols such as SSH, SSL, IPSec, and TLS offer end-to-end security solutions, in addition to crypto-protocol encryption systems \[[@B23-sensors-19-04952],[@B24-sensors-19-04952]\]. Further research oriented towards the security of the application layer \[[@B25-sensors-19-04952]\] focusing on data integrity and authentication procedures was developed with the aim of providing solutions for known attacks such as modification, spoofing, and flooding \[[@B26-sensors-19-04952]\]. Nevertheless, a certain number of limitations was revealed resulting from mechanisms defined in the DNP3 protocol---in particular, embedded security mechanisms \[[@B27-sensors-19-04952]\]. A solution based on crypto-algorithms that includes known encryption methods like AES and RSA was developed to protect DNP3 protocol at the application layer \[[@B28-sensors-19-04952]\]. In this research, the authors contributed three primary enhancements, including a new security scheme that was implemented together with the DNP3 protocol, a method for constructing the bytes within every layer, and use of the TCP/IP protocol for data exchange. On the other hand, IDS based on different machine learning methods has been developed, whereby attacks can be detected based on highly accurate results of detected attacks. However, more improvements are necessary due to false alarms or false positive from the detection systems. This problem usually leads to the misclassification between good and bad data in the network \[[@B29-sensors-19-04952]\]. In the same category of research, a group of five machine learning algorithms was tested for cybersecurity solutions to protect SCADA systems \[[@B30-sensors-19-04952],[@B31-sensors-19-04952],[@B32-sensors-19-04952],[@B33-sensors-19-04952]\]. After the training process, the models were implemented in a real network environment to capture and analyze online data from network traffic. Both results from the testbed and live traffic revealed that the IDS based on machine learning algorithms was efficient for detecting attacks. Further research developed by Keliris et al. \[[@B34-sensors-19-04952]\] showed that the Support Vector Machine (SVM) algorithm performs well for anomaly detection and classification. They used a supervised learning method to develop a process-aware defense tactic in the ICS accounting for behavior-based attacks. The work done in \[[@B35-sensors-19-04952]\] suggests that a detection system using machine learning techniques in power systems would be feasible for detecting malicious states. Tomin et al. \[[@B35-sensors-19-04952]\] claimed that such techniques, where applied in SCADA/ICS, offer a range of solutions with a satisfactory level of security. In the course of their research, they used an offline training process using a cross-validation method and they applied it to a semi-automated method for online testing purposes. Further research has been developed to provide security for Smart Grid DNP3, through the identification of malicious activities in ICS of IoT based on Deep Learning, IDS for SCDA systems, and Neural Network-based IDS for critical infrastructure. These have shown tremendous results in the development of models for the detection of attacks on power systems \[[@B36-sensors-19-04952],[@B37-sensors-19-04952],[@B38-sensors-19-04952],[@B39-sensors-19-04952],[@B40-sensors-19-04952],[@B41-sensors-19-04952]\]. 3. Proposed Solution: Method and Implementation Experiments {#sec3-sensors-19-04952} =========================================================== This section describes the proposed scheme for the SCADA/DNP3 protocol. The solution requires several steps, referred to as "modules", and each of these plays a specific role in building a holistic cyber-security solution in an IoT-based Smart Grid environment. 3.1. System Model and Description {#sec3dot1-sensors-19-04952} --------------------------------- The proposed solution is based on the following modules: (a) data input system, (b) data analysis system, and (c) classification and detection system, as shown in [Figure 3](#sensors-19-04952-f003){ref-type="fig"}. However, before we could arrive at this holistic solution, we performed additional research on the DNP3 protocol. Firstly, we developed an attack model for each layer of the DNP3 protocol, as shown in [Figure 4](#sensors-19-04952-f004){ref-type="fig"}. These attacks had two mains functions: (1) to collect data for the purpose of building a database to be used in the training and testing model, (2) to assess the vulnerabilities of the DNP3 protocol \[[@B42-sensors-19-04952]\] that attackers are able to leverage in order to carry out cyber-attacks on IoT-based Smart Grids. Secondly, we developed an algorithm for analyzing a modified DNP3 protocol \[[@B43-sensors-19-04952],[@B44-sensors-19-04952]\]. This algorithm uses the original DNP3 protocol as a reference for the purpose of comparison with the common vulnerabilities of the protocol stack. We used four types of attack---modification, interception, interruption, and fabrication---targeting all layers in order to evaluate them. The collected vulnerabilities (based on the attacks on the two protocols) were used with a mapping function to modify the features of the DNP3 protocol. The results provide the vulnerabilities discovered for the customized protocol, as shown in [Figure 5](#sensors-19-04952-f005){ref-type="fig"}. 3.2. Data Input System: Data Generation {#sec3dot2-sensors-19-04952} --------------------------------------- The dataset used in the experiment is from a variety of different sources, but the most important data, which is related to DNP3 packet parsing, was based on the assessment of vulnerabilities and attacks performed on the protocol. Therefore, we will only describe the data obtained from these experiments, as this is the focus of our research. The rest of the data was obtained from an open-source dataset used at the 4SICS industrial cybersecurity conference \[[@B3-sensors-19-04952]\], which is an annual summit. The summit calls on experts in ICS/SCADA/DNP3 cybersecurity from the most critical infrastructures like smart grids, transportation, and so on. As far as our purposes are concerned, we only collected data---PCAP Files---related to the smart grid from the ICS Lab using RTUs, PLCs, and other industrial networks. Additionally, we included some known malware targeting ICS/SCADA systems. As described above, to generate the data, several steps are required, which can be summarized as two main steps: vulnerability assessment and attack modeling on the DNP3 protocol. ### 3.2.1. DNP3 Protocol Vulnerability Assessment {#sec3dot2dot1-sensors-19-04952} Different methods have been proposed in order to analyze the weaknesses in the DNP3 protocol; one of these methods presents an assessment of specific attacks on function code within certain layers of the protocol stack \[[@B45-sensors-19-04952],[@B46-sensors-19-04952],[@B47-sensors-19-04952]\]. In this paper, we used a customized DNP3 protocol to perform the vulnerability analysis, and this is compared with the original one, as shown in [Figure 4](#sensors-19-04952-f004){ref-type="fig"}. The novelty of our method is that we map common vulnerabilities onto the customized database features, with the results showing us the weakness of the protocol, meaning that we can ultimately launch different attacks in order to collect or generate the data to use in our experiments. The proposed algorithm takes the two protocols as input and checks whether they satisfy the protocol stack requirements. If they do, they are parsed into the main layers; otherwise, they go back to the starting point. This process continues, using DNP3 threats such as modification, interruption, interception, and fabrication, where we define the common vulnerabilities of the protocol stack to to be used for the mapping process. The mapping process is carried out based on a database of features from the DNP3 packet. [Table 1](#sensors-19-04952-t001){ref-type="table"} shows selected features from the layers of the DNP3 protocol, and a full account of the features is provided in [Appendix A](#app1-sensors-19-04952){ref-type="app"}. A "Yes" in the column "Subject to Attack" means that they present a potential weakness that makes them vulnerable to DNP3 protocol threats. ### 3.2.2. Attack Modeling on the DNP3 Protocol {#sec3dot2dot2-sensors-19-04952} To launch attacks on the DNP3 protocol, we made an attack model that was specific to the vulnerabilities discovered. As shown in [Figure 5](#sensors-19-04952-f005){ref-type="fig"}, the model is based on three main steps:✓Step 1: Pre-attack. This is where the preliminary is carried out, including obtaining the DNP3 packets from the repository, and preprocessing the packet in order to obtain three layers for the next step.✓Step 2: Attack Modelling. In this step, we define the attacks on the basis of the vulnerabilities discovered in each of the following layers: Data Link Layer, Transportation Layer, and Application Link Layer. For the first layer, we defined three attacks (Length Overflow Attack, DFC Flag Attack and Reset Function Attack), for the second layer, we defined two attacks (Fragmented Message Interruption and Transport Sequence Modification), and in the last layer, we defined two attacks (FC Modification Request and Configuration Capture Attack IIN).✓Step 3: Setting up the attack with DNP3 parameters and Consequences. This step defines the parameters to be used during the attack (payload) and describes the consequences of each attack. As given in the description of the consequences, each attack leads to bad behavior in the smart grid network. The aim is not to have these attacks, but rather to develop countermeasures in order to protect the network, devices, data, and human beings. Both the vulnerabilities and the attacks have several operational impacts that could cause damage to the system or take over the control system \[[@B48-sensors-19-04952],[@B49-sensors-19-04952],[@B50-sensors-19-04952],[@B51-sensors-19-04952],[@B52-sensors-19-04952]\]. The data input system consists of malware and benign data, as already described in the introductory paragraph of this subsection. [Table 2](#sensors-19-04952-t002){ref-type="table"} gives a summary of the dataset used in this paper, where the name column describes the name of the malware or benign data, Qt is the amount of each type and the percentage of the distribution over the total. The overall distribution of malware is 55%, and that of benign data is 45%, which is acceptable for a classification and detection model. Bencsath et al. \[[@B53-sensors-19-04952]\] described the most dangerous malware targeting industrial infrastructure in detail. Stuxnet was discovered in 2010, when it was reported to have destroyed numerous centrifuges in Natanz. The centrifuges had been designed for a uranium enrichment facility in Iran. The infection vector of Stuxnet was the USB, from which the worm was installed on and spread among interconnected computers. It is therefore very important to produce a cyber-security solution based on the IDS and ML techniques in order to protect such critical infrastructures against malware. 3.3. Data Analysis System {#sec3dot3-sensors-19-04952} ------------------------- This module is located in the middle of the other modules, as it takes the input from the various repositories and then transforms the data into a format compatible with the functions of the next module. The data analysis is built up over many steps and requires advanced knowledge of Data Science, with several tools to be used in such work. In this paper, we describe a few steps taken from [Figure 6](#sensors-19-04952-f006){ref-type="fig"}. The data analysis consists of eight steps, from raw data input to the visualization step. Step 1: This is the initial action, where module one feeds raw data to the second module. As described above, 55% of the dataset is made up of malware and 45% is made up of benign data.Steps 2 to 4: After getting the raw data, the engine proceeds to DNP3 protocol extraction with the integration of various fields with pre-processing actions such as contextualization and mapping in order to prepare for loading to the DB. Before that, the engine carries out the data cleaning, removing unwanted fields, carrying out de-noising, and nullifying some fields that match with the DB used in our experiment.Steps 5 to 8: This is where the engine utilizes the DB constructed in Steps 2 through 4. At this stage, the important features are extracted based on their presence in the DB (presence refers to how frequently this feature occurred throughout the whole DB). Because the DB is a mixture of many types of data, the classification process first requires that the data be transformed from categorical and numerical data to a binary data format. Once we have one type of data, it is possible to apply the ML algorithms directly (green arrows) and then execute the classification process. Different algorithms were used for this process of data transformation in order to obtain the final DB and visualization results. Algorithm 1 is the pseudocode where all main steps are called to execute the ML algorithms. As given by the algorithm, we listed from Step 11 to 14 some of the algorithms used for the data transformation and visualization process. As input, we used a mixture of malware and benign data from module one, but for the paper objectives, we are going to focus only on the DNP3 packet analysis for more details. In the case of malware, we will describe Stuxnet and the features selected from the data analysis process. Algorithm 1 has two main parts: the data input and processing part. The second part of the algorithm gives the main steps that implement the data transformation until the visualization step. The algorithm instructs to select all features from the raw data that include DNP3 protocol features, the 5 tuples (Source IP, Destination IP, Source Port, Destination Port, and Protocol), and eventually the features of the malware raw data. Next, it does a format check, which requires removing some unwanted characters that would cause errors in the database. In this case, null fields are not allowed, and categorical data and numerical data have to be mapped too. The cleaned data will then constitute the initial database, where we can make some queries to see the content. The biggest part of the algorithm is where the call of each machine learning algorithm is running for different functions. Feature selection is the most important step in malware classification when using an ML algorithm. Bugra et al. \[[@B54-sensors-19-04952]\] presented a method for malware classification where they applied DL (Deep Learning) methods. The authors performed the classification of malware based on a shallow deep learning network. To realize their experiment, they used a two-layer neural net to process the text, which consisted of turning text into a numerical form that is understandable by deep networks. This is called word2vec, developed by Tomas Mikolov \[[@B55-sensors-19-04952],[@B56-sensors-19-04952]\] at Google and which is available from the Google code archive \[[@B57-sensors-19-04952]\]. The work in \[[@B58-sensors-19-04952]\] gives methods where ML has been used to classify malware and detection, in addition to implementations directions. The main goal of their work is to give a list of best classification methods such as feature selection, representation using Cuckoo Sandbox, k-Nearest-Neighbors (KNN), Decision Tree (DT), Support Vector Machines (SVM), Naive Bayes and Random Forest. Algorithm 1 Data Transformation & Visualization In this paper, we have used many algorithms, such as k-Nearest-Neighbors (KNN), Decision Tree (DT), Support Vector Machines (SVM), Naive Bayes and Random Forest. The results from our experiments and their descriptions are presented in [Section 4](#sec4-sensors-19-04952){ref-type="sec"}. 3.4. Cyber-Attack Algorithm and IDS Solution {#sec3dot4-sensors-19-04952} -------------------------------------------- This subsection describes the cyber-attack algorithm that we created, in addition to the countermeasure (the IDS to detect the attack). This final step leads to the classification and detection processes from [Figure 3](#sensors-19-04952-f003){ref-type="fig"}. After the vulnerability assessment of the DNP3 protocol, the attack modeling, and data collection, we have now all we need to launch the attack and then perform the classification and detection solution. Algorithm 2 gives the steps to launch an attack on DNP3 protocol. Algorithm 2 Cyber-Attack on DL TL and AL \*\* START \*\* 01: Input ← Raw data 02: Output ← Anomaly and Normal Traffic: {Classification and Detection} \*\* PRE-ATTACKS \*\* 03: Procedure: INTERCEPTION (I) 04: Action: INJECTION (Inj) or MODIFICATION (Mod) 05: Packet ← {pre-process, get DNP3 packet (dnp pkt ) } 06: DNP3 protocol ← {DataLink (DL), TransportLink (TL), ApplicationLink (AL)} 07: Attack = { L O V A , D F C , F C A , F M I 1 , F M I 2 , T S M , F C M , C C \_ I I N } \*\*\* SETTING-UP PARAMETERS & ATTACK LAUNCHING \*\*\* 08: LOVA ⟺ I ( d n p p k t ) { M o d ( D L l e n g t h ← D L l e n g t h ± α ) } 09: DFC Flag ⟺ I ( d n p p k t ) { M o d ( D L D F C = 0 ← D L D F C = 1 ) } 10: FCA ⟺ I ( d n p p k t ) { M o d ( D L F C ← D L F C = 1 ) } 11: FMI1 ⟺ I ( d n p p k t ) { I n j ( T L F M I ( F I N ) ← T L F M I ( F I N = 1 ) ) } 12: FMI2 ⟺ I ( d n p p k t ) { I n j ( T L F M I ( F I R ) ← T L F M I ( F I R = 0 ) ) } 13: TSM ⟺ I ( d n p p k t ) { M o d ( T L T S M ( S e q . N u m b e r ) ← T L T S M ( S e q . N u m b e r ± β ) ) } 14: FCM1 ⟺ I ( d n p p k t ) { M o d ( A L F C ← A L F C M . r e q = 0 x 02 ) } 15: FCM2 ⟺ I ( d n p p k t ) { M o d ( A L F C ← A L F C M . r e q = 0 x 0 d ) } 16: FCM3 ⟺ I ( d n p p k t ) { M o d ( A L F C ← A L F C M . r e q = 0 x 0 e ) } 17: FCM4 ⟺ I ( d n p p k t ) { M o d ( A L F C ← A L F C M . r e q = 0 x 14 ) } 18: FCM5 ⟺ I ( d n p p k t ) { M o d ( A L F C ← A L F C M . r e q = 0 x 15 ) } 19: FCM6 ⟺ I ( d n p p k t ) { M o d ( A L F C ← A L F C M . r e q = 0 x 82 ) } 20: CC_INN ⟺ I ( d n p p k t ) { M o d ( A L C C \_ I I N ← A L C C \_ I I N = 1 → 5 t h b i t i n 2 n d b y t e ) } 21: If I ( d n p p k t ) then, //Interception of the packet 22: Attacker I ( d n p p k t ) { ( M o d ∥ I n j ) D L ¬ T L ¬ A L } //Launch the attack on the layers 23: Get anomaly traffic 24: End if 25: End All of these attacks on DNP3 protocols are assumed to occur during data transmission from one station to another. In practical cases, the system uses Master (client) and Slave (server) terminology. In this case, the server is defined as a station or device that holds and processes the information needed by an operator. To the other side, a client is a substation or device that requests information from the server. The DNP3 protocol provides the ability to facilitate data transmission between Master and Slave \[[@B59-sensors-19-04952]\]. [Figure 7](#sensors-19-04952-f007){ref-type="fig"} shows a schematic attack in which the attacker performs interception, modification and injection attacks on the DNP3 packet content as described in Algorithm 2. The abovementioned algorithm was built up over three stages: start, pre-attack and setting parameters up & attack launching. The last stage includes seven major attacks that target all three layers on specific fields. The second stage instructs the algorithm to use the interception method that executes the injection and modification attacks in addition to the three layers and corresponding attacks. The first stage is related to data input and output data format information. The abbreviations used in the algorithm are described in [Table 3](#sensors-19-04952-t003){ref-type="table"}. The main part (stage number 3) of the algorithm describes the attacks that target the layers as follows:From Steps 8 to 10: These are the sets of attacks aiming for the modification of the Data Link Layer parameters by intercepting the DNP3 packet. The actions that are carried out as part of the Length Overflow Attack (LOVA), Data Flow Control Flag (DFC Flag) and Function Code Attack (FCA) are executed using the interception procedure. During the attack, the length is modified by ± α to the original size, the DCF and FCA are modified with 0 or 1.From Steps 11 to 13: The targeted layer is the Transport Link Layer with the injection action of some fault parameters to the DNP3 packet. At this stage, the Fragmented Message Interruption Attack (FMI1 and FMI2) supports the fault parameters by injecting 1 or 0 to the First (FIR) and Final (FIN) Bit Number. The Transport Sequence Modification Attack (TSM) is also one of the TL attacks with sequence modification by ± α to the original order, but it is based on the modification procedure.From Steps 14 to 20: This is a range of attacks on the Application Link Layer with a large number of parameters. After the DNP3 is intercepted, the modification process is performed on the packet at the Application Layer. To do so, the Function Code Modification Attack (FCM1\~6) is called, where the attacker sets up the parameters to be modified. The request to modify this function code at the application layer is based on the selected values (such as 0x02, 0x0d, ..., 0x82 and modification of a byte of the internal indication, such as the 5th bit in the 2nd byte of the DNP3 packet at the Application Layer). After the last step, the whole DPN3 packet (in the current session) is compromised, and it is time that the engine can classify between bad and good traffic. The results from the experiment are detailed in [Section 4](#sec4-sensors-19-04952){ref-type="sec"}. 4. Experimental Results and Discussion {#sec4-sensors-19-04952} ====================================== 4.1. Malware Sample Feature Selection Results {#sec4dot1-sensors-19-04952} --------------------------------------------- The following section discusses the findings after data transformation for the classification process. For the purposes of our paper, we cannot include all of the figures and tables, but we have selected the most important results from among others. As described above in [Section 3](#sec3-sensors-19-04952){ref-type="sec"}, the input data comprised about 10 malware and 80 of the benign dataset, which represents 56% malware and 45% benign data, respectively. For malware analysis and feature extraction, we selected the Stuxnet malware, and we parsed this sample using the Pepper tool, which is an open-source tool for malware static analysis on a portable executable \[[@B60-sensors-19-04952]\]. We extracted the metadata, header, opt header, sections, and import features from the executable file, as shown in [Figure 8](#sensors-19-04952-f008){ref-type="fig"}. The Stuxnet malware PE result shows that many system files are subject to compromises or attacks. [Figure 9](#sensors-19-04952-f009){ref-type="fig"} shows the distribution of the Top 20 process names found after the reverse engineering of the malware using the Pepper tool. We selected only major information, with 50% and high score points as given in [Table 4](#sensors-19-04952-t004){ref-type="table"}. The main reason is that after computing all of the features, it was necessary to statistically pick out only those with a high degree of presence in the original database. Presence refers to how frequently the feature occurs throughout the whole DB. For our experiment we set, 50% as the threshold. The table indicates that the malware target memory process has highest score, with four times, and it can be observed that the file names being compromised are related to the memory processes. The other process is related to the local security system authority service, which is a highly critical system file in Microsoft Windows (the lsass.exe). Most malware targets this file, because it is used to enforce security policies related to sensitive information such as password changes and login access verifications. The malware also targets another executable file with a task of high importance in the Windows Task Manager, and which contains machine code, and this is called vdmdbg. It has also a high score in the below table. [Appendix B](#app2-sensors-19-04952){ref-type="app"} provides all of the feature information from the Stuxnet Portable Executable (PE) file. 4.2. DNP3 Protocol Packet Sample Feature Selection Results {#sec4dot2-sensors-19-04952} ---------------------------------------------------------- The DNP3 packets that include the attack types defined in [Section 3](#sec3-sensors-19-04952){ref-type="sec"} are collected using the Wireshark tool, which is a network packet analyzer that captures network packets and displays the packet contents with the maximum detail possible \[[@B61-sensors-19-04952]\]. In order to generate the packet, we developed an exploit that was specifically designed to carry out a cyber-attack on the DNP3 protocol. This malicious software is real, and we advise the reader of this paper not to try this on a live product. The exploit, as given in [Appendix C](#app3-sensors-19-04952){ref-type="app"}, carries the data (payload) that intercept the traffic and then injects some modified parameters, as described in Algorithm 2. After the attacks, we collected the features from the DNP3 packet where the results revealed that the predicted attacks (as defined in Algorithm 2) achieved the goals. [Table 5](#sensors-19-04952-t005){ref-type="table"} describes our experimental results in detail, along with the impact on the SCADA/DNP3 devices. As can be seen, the impact depends on the attack type, the parameter modified in the original format, and the link layer that is attacked. As described above, it is prohibited to run the provided exploit in a real working environment, because the impact of the attack would be damaging. The rest of the features of DNP3 are given in [Appendix A](#app1-sensors-19-04952){ref-type="app"}. 4.3. Visualization and Classification {#sec4dot3-sensors-19-04952} ------------------------------------- The discussion in this subsection is related to the results of the proposed methods based on the classification of malware, which is displayed in the form of a graph visualization. [Table 6](#sensors-19-04952-t006){ref-type="table"}, with [Figure 10](#sensors-19-04952-f010){ref-type="fig"} and [Figure 11](#sensors-19-04952-f011){ref-type="fig"}, describes the classification results with the following explanation:Login-Time: The field indicates when the event happened. It is, therefore, easy to track down and find out the right moment for the attack on the system when there is a need for an investigation.Source-IP: Every traffic on the network includes the source IP address, which indicates the origin of the data, request, or other type of transaction. In our experiment, this is the IP address of the device that is sending information to the destination device.Source-Port: This is one of the user session parameters that tells the system where to reply to the response. It is always associated with the source-IP and the different applications and protocols used by the sender.Destination-IP: This is where to go. The receiving device in our experiment has a destination IP to which the packet is to be sent. This enables two-way communication in the configuration.Destination-Port: The same explanation as the source port, except that this is for the destination device.Classification: As stated before, the aim is to distinguish between benign and malware groups for elements in the dataset. Hence, after the process, the result results in an "anomaly", as bad packets related to malware or any malicious activity are discovered during the analysis. We only provide those results that identify an anomaly.Field: With this information, we can see what type of feature, attack type or any other field has been targeted. In this case, the system gives "Transport FIR", which indicates the DNP3 protocol feature.Graph Visualization: The 3D graph indicates the classification as either malware or benign data. The red dots indicate the malware sample in our experiment, while the blue ones indicate the benign dataset. Additionally, there other two graphs, which give an overview of the Top 5 source and destination IP addresses. 5. Conclusions {#sec5-sensors-19-04952} ============== This paper discussed cybersecurity solutions based on the Intrusion Detection System in the IoT-based Smart Grid. We described in detail the concept of a system based on the IoT for Smart Grids using the SCADA/DNP3 communication protocol. To achieve the proposed method, we developed and presented a series of algorithms for implementation along with experiments. In this paper, we developed a new method for assessing DNP3 protocol vulnerability, which gave us an idea of where to perform the attack. This assessment was conducted on a modified DNP3 protocol with reference to the original protocol. Next, based on the discovered vulnerabilities, we developed the new attack model aiming at the Data Link Layer, Transport Link Layer and Application Link Layer of the DNP3 protocol. Moreover, we developed two algorithms that helped us perform data transformation using Machine Learning methods. The other algorithm includes all of the steps for the cyber-attack on the DNP3 protocol; this also includes the classification process. Finally, we presented the experimental results, showing that the proposed method was able to detect intrusions to the SCADA system based on an IoT Smart Grid and could classify them with detailed information about the compromised fields from the DNP3 packet. Conceptualization, Writing the Original Draft, Project Administration, and Funding Acquisition, X.C.Y.; Methodology, Writing---Review & Editing, Formal Analysis and Validation, Z.G.L.; Investigation and Data Curation, L.N.; Visualization, Resources, and Supervision, B.N. This research was supported by the Scientific Fund Project of Facility Horticulture Laboratory of Universities in Shandong of China (Grant number: 2018YY016) and the Doctoral Scientific Fund Project of Weifang University of Science & Technology of China (Grant number: 2017BS17), it was also supported by the Innovation Fund of Ministry of Education, Science and Technology Development Center of China (Grant number: 2018A02013). The authors declare no conflict of interest. sensors-19-04952-t0A1_Table A1 ###### In this table '-' means that the feature does not have value. We have 95 features from the DNP3 packets. Original_Format Readable_Format ------------------------------------- ------------------------------------------------------------------------------------- dnp.data_chunk.CRC.status '-' dnp.hdr.CRC.status '-' dnp3.addr '-' dnp3.src '-' Number of Items: '-' Number of Items: 0 '-' dnp3.al.fragment al_fragment dnp3.al.fragment.count al_fragment_count dnp3.al.fragment.reassembled.length al_fragment_reassembled_length dnp3.al.ctl application_layer_control dnp3.al.con application_layer_control_confirm dnp3.al.fin application_layer_control_final dnp3.al.fir application_layer_control_first dnp3.al.func application_layer_control_function_code dnp3.al.iin application_layer_control_internal_indications dnp3.al.iin.bmsg application_layer_control_internal_indications_broadcast\_ msg_rx dnp3.al.iin.cls1d application_layer_control_internal_indications_class1_data\_ available dnp3.al.iin.cls2d application_layer_control_internal_indications_class2_data\_ available dnp3.al.iin.cls3d application_layer_control_internal_indications_class3_data\_ available dnp3.al.iin.cc application_layer_control_internal_indications_configuration_corrupt dnp3.al.iin.rst application_layer_control_internal_indications_device_restart dnp3.al.iin.dt application_layer_control_internal_indications_device_trouble dnp3.al.iin.dol application_layer_control_internal_indications_digital_outputs_in_local dnp3.al.iin.ebo application_layer_control_internal_indications_event_buffer\_ overflow dnp3.al.iin.fcni application_layer_control_internal_indications_function_code \_not_implemented dnp3.al.iin.oae application_layer_control_internal_indications_operation\_ already_executing dnp3.al.iin.pioor application_layer_control_internal_indications_parameters\_ invalid_or_out_of_range dnp3.al.iin.obju application_layer_control_internal_indications_requested_objects_unknown dnp3.al.iin.tsr application_layer_control_internal_indications_time_sync_required dnp3.al.seq application_layer_control_sequence dnp3.al.uns application_layer_control_unsolicited dnp3.al.obj application_layer_object dnp3.al.range.quantity application_layer_object_items_range_quantity dnp3.al.range.start application_layer_object_items_range_start dnp3.al.range.stop application_layer_object_items_range_stop Point Number application_layer_object_point_number dnp3.al.index application_layer_object_point_number_index dnp3.al.point_index application_layer_object_point_number_index dnp3.al.ana.int application_layer_object_point_number_quality_analog_value dnp3.al.aiq.b2 application_layer_object_point_number_quality_comm_fail dnp3.al.biq.b2 application_layer_object_point_number_quality_comm_fail dnp3.al.aiq.b4 application_layer_object_point_number_quality_local_force dnp3.al.biq.b4 application_layer_object_point_number_quality_local_force dnp3.al.aiq.b0 application_layer_object_point_number_quality_online dnp3.al.biq.b0 application_layer_object_point_number_quality_online dnp3.al.aiq.b5 application_layer_object_point_number_quality_over_range dnp3.al.biq.b5 application_layer_object_point_number_quality_over_range dnp3.al.aiq.b6 application_layer_object_point_number_quality_reference dnp3.al.biq.b6 application_layer_object_point_number_quality_reference dnp3.al.reltimestamp application_layer_object_point_number_quality_relative_timestamp dnp3.al.aiq.b3 application_layer_object_point_number_quality_remote_force dnp3.al.biq.b3 application_layer_object_point_number_quality_remote_force dnp3.al.aiq.b7 application_layer_object_point_number_quality_reserved dnp3.al.biq.b7 application_layer_object_point_number_quality_reserved dnp3.al.aiq.b1 application_layer_object_point_number_quality_restart dnp3.al.biq.b1 application_layer_object_point_number_quality_restart dnp3.al.time_delay application_layer_object_point_number_quality_time_delay dnp3.al.bit application_layer_object_point_number_value dnp3.al.objq.prefix application_layer_object_prefix_code dnp3.al.objq.range application_layer_object_range_code dnp3.al.timestamp application_layer_timestamp dnp3.ctl control dnp3.ctl.dir control_direction dnp3.ctl.fcb control_frame_count_bit dnp3.ctl.fcv control_frame_count_valid dnp3.ctl.prifunc control_function_code dnp3.ctl.prm control_primary dnp.data_chunk data_chunk dnp.data_chunk.CRC data_chunk_crc dnp.data_chunk_len data_chunk_length dnp3.hdr.CRC data_link_header_crc dnp3.dst destination \_ws.expert.group expert_info_group dnp3.iin_abnormal expert_info_iin_abnormal \_ws.malformed expert_info_malformed_packet \_ws.expert.message expert_info_message \_ws.expert.severity expert_info_severity dnp3.len length dnp3.start start_bytes dnp3.tr.ctl transport_control dnp3.tr.fin transport_control_final dnp3.tr.fir transport_control_first dnp3.tr.seq transport_control_sequence dnp3.al.2bit '-' dnp3.al.count '-' dnp3.al.ctrlstatus '-' dnp3.al.off_time '-' dnp3.al.on_time '-' dnp3.al.range.abs '-' dnp3.al.size '-' dnp3.al.unknown_data_chunk '-' dnp3.ctl.clr '-' dnp3.ctl.op '-' dnp3.ctl.trip '-' dnp3.num_items_neg '-' sensors-19-04952-t0A2_Table A2 ###### The complete list of Stuxnet PE reverse engineering malware with the names of the processes and the score. Number Name Process Name Score -------- ----------------------------------------- ---------------- ------- 1 memory-mod-pe-0 × 20000000-0 × 10124000 service.exe 95.6 2 kerner32.dll.aslr.0013a1e svchost.exe 92.5 3 kerner32.dll.aslr.0013b86 svchost.exe 91.5 4 memorymod-pe-0 × 00090000-0 × 0010a000 lsass.exe 85 5 memorymod-pe-0 × 00090000-0 × 0010a000 lsass.exe 85 6 lsass.exe lsass.exe 80.5 7 lsass.exe lsass.exe 80.5 8 memorymod-0 × 006b0000-0 × 006b1000 services.exe 75.8 9 vdmdbg.dll taskmgr.exe 74 10 izarccm.dll explorer.exe 73.6 11 ntosknl.exe System 72.6 12 ntdll.dll smss.exe 71.6 13 olepro32.dll explorer.exe 68 14 mysqld-nt.exe mysqld-nt.exe 64 15 mlang.dll explorer.exe 61 16 bhomanger.dll System 61 17 hal.dll explorer.exe 61 18 wuaucpl.cpl wuauclt.exe 61 19 mrxnet.sys System 50 20 vmhgfs.dll explorer.exe 50 21 odbc32.dll explorer.exe 50 22 wuaucpl.cpl explorer.exe 50 23 odbc32.dll wuauclt.exe 50 24 mrxcls.sys winlogon.exe 49 25 natlanman.dll System 46 26 browselc.dll explorer.exe 42 27 ksecdd.sys explorer.exe 41 28 hidphone.tsp System 34 29 cscdll.dll svchost.exe 34 30 cscdll.dll winlogon.exe 34 31 utildll.dll explorer.exe 34 32 taskmgr.exe taskmgr.exe 26 33 sfc_os.dll taskmgr.exe 26 34 sfc_os.dll spoolsv.exe 26 35 duser.dll svchost.exe 26 36 sfc_os.dll explorer.exe 26 37 ntdll.dll wuauclt.exe 26 38 ntdll.dll crss.exe 26 39 ntdll.dll svchost.exe 26 40 ntdll.dll VMwareUser.exe 26 41 mprapi.dll lsass.exe 19 42 mprapi.dll winlogon.exe 19 43 h323.tsp svchost.exe 19 44 tapisrv.dll svchost.exe 19 45 alg.exe alg.exe 19 46 ntdll.dll svchost.exe 19.6 {#sensors-19-04952-f0A1} {#sensors-19-04952-f001} {#sensors-19-04952-f002} {#sensors-19-04952-f003} {#sensors-19-04952-f004} {#sensors-19-04952-f005} {#sensors-19-04952-f006} {#sensors-19-04952-f007} {#sensors-19-04952-f008} {#sensors-19-04952-f009} {#sensors-19-04952-f010} {#sensors-19-04952-f011} sensors-19-04952-t001_Table 1 ###### DNP3 partial features dataset. Features Subject to Attack ------------------------------------------------------- ------------------- **DNP3_START** **DNP3_LENGTH** Yes **DNP3_SOURCE** **DNP3_DESTINATION** **DNP3_CONTROL_DFC** Yes **DNP3_CONTROL_DIR** **DNP3_CONTROL_FCB** Yes **DNP3_CONTROL_FCV** Yes **DNP3_CONTROL_FUNC_CODE_PRI** Yes **DNP3_CONTROL_FUNC_CODE_SEC** **DNP3_CONTROL_PRM** **DNP3_CONTROL_reserved** **DNP3_CRC** **DNP3_Transport_FIN** Yes **DNP3_Transport_FIR** Yes **DNP3_Transport_SEQUENCE** Yes **DNP3_Application_request_Application_control_CON** **DNP3_Application_request_Application_control_UNS** **DNP3_Application_request_FUNC_CODE** Yes **DNP3_Application_response_Application_control_CON** **DNP3_Application_response_IIN_CLASS_3\_EVENTS** **DNP3_Application_response_IIN_CONFIG_CORRUPT** Yes **DNP3_Application_response_IIN_DEVICE_RESTART** sensors-19-04952-t002_Table 2 ###### Dataset of malware and benign data in this paper. Name Malware Qt. \% Benign Qt. \% ------------------------------ --------- ------ ------- -------- ------ ------- **Triton** Yes 1650 15.46 No 0 0.00 **Industroyer** Yes 1521 14.25 No 0 0.00 **BlackEnergy** Yes 650 6.09 No 0 0.00 **Stuxnet** Yes 1120 10.50 No 0 0.00 **Duqu** Yes 944 8.85 No 0 0.00 **Flame** Yes 1062 9.95 No 0 0.00 **Gauss** Yes 1267 11.87 No 0 0.00 **DNP3-Data(Original)** No 0 0.00 Yes 4593 53.58 **DNP3-Packet (Experiment)** Yes 2456 23.02 N o 0 0.00 **4SICS** No 0 0.00 Yes 3980 46.42 Total 10,670 55 8573 45 sensors-19-04952-t003_Table 3 ###### Notations and description used in Algorithm 2. Abbreviations Description --------------- -------------------------------------------------- **DLL** Data Link Layer **TLL** Transport Link Layer **ALL** Application Link Layer **LOVA** Length Overflow Attack **DFCA** Data Flow Control Attack **FCA** Function Control Attack **FMIA** Fragmented Message Interruption Attack **TSMA** Transport Sequence Modification Attack **FCM** Function Code Modification **CCA_IIN** Configuration Capture Attack_Internal INdication sensors-19-04952-t004_Table 4 ###### Stuxnet features and scores. Number File Name Process Name Score % -------- ----------------------------------------- --------------- --------- 1 memory-mod-pe-0 × 20000000-0 × 10124000 service.exe 95.6 2 kerner32.dll.aslr.0013a1e svchost.exe 92.5 3 kerner32.dll.aslr.0013b86 svchost.exe 91.5 4 Memory mod-pe-0 × 00090000-0 × 0010a000 lsass.exe 85 5 Memory mod-pe-0 × 00090000-0 × 0010a000 lsass.exe 85 6 lsass.exe lsass.exe 80.5 7 lsass.exe lsass.exe 80.5 8 memorymod-0 × 006b0000-0 × 006b1000 services.exe 75.8 9 vdmdbg.dll taskmgr.exe 74 10 izarccm.dll explorer.exe 73.6 11 ntosknl.exe System 72.6 12 ntdll.dll smss.exe 71.6 13 olepro32.dll explorer.exe 68 14 mysqld-nt.exe mysqld-nt.exe 64 15 mlang.dll explorer.exe 61 16 bhomanger.dll System 61 17 hal.dll explorer.exe 61 18 wuaucpl.cpl wuauclt.exe 61 19 mrxnet.sys System 50 20 vmhgfs.dll explorer.exe 50 21 odbc32.dll explorer.exe 50 22 wuaucpl.cpl explorer.exe 50 23 odbc32.dll wuauclt.exe 50 sensors-19-04952-t005_Table 5 ###### Feature description of DNP3 protocol attack. DNP3-Features Description Attack Type Parameter Impact Link Layer -------------------------------------------------- ------------------------------------------------------------- -------------------- ------------------------------------------------ ----------------------------- ------------ **DNP3_LENGTH** Length of field LOVA Original length modification Device crashes DL **DNP3_CONTROL_DFC** The DFC tells other devices that the current device is busy DFC Flag = 1 Eternal busy DL **DNP3_CONTROL_FUN_CODE_PRI** Primary Function code User Process reset Code = 1 Unwanted restart DL **DNP3_Transport_FIN** Final bit FMI Flag modification Early message termination TL **DNP3_Transport_FR** First bit FMI Flag modification Message processing error TL **DNP3_Application_request_FUN_CODE** Function Code FCA 0 × 02, 0 × 0d, 0 × 0e, 0 × 14, 0 × 15, 0 × 82 Crash or reboot AL **DNP3-Application_response_IIN_CONFIG_CORRUPT** Configuration File System CC_IIN 5th bit in 2^nd^ byte Configuration file modified AL sensors-19-04952-t006_Table 6 ###### Classification result description. Source-IP Source-Port Destination-IP Destination-Port Classification Field ---------------- ------------- ---------------- ------------------ ---------------- --------------- **172.17.0.1** 59686 172.17.0.2 45000 Anomaly Transport FIR **172.17.0.1** 59686 172.17.0.2 45000 Anomaly Transport FIR **172.17.0.2** 41044 192.168.0.141 45000 Anomaly Transport FIR **172.17.0.2** 41044 192.168.0.141 45000 Anomaly Transport FIR **172.17.0.2** 41044 192.168.0.141 45000 Anomaly Transport FIR **172.17.0.2** 41044 192.168.0.141 45000 Anomaly Transport FIR **172.17.0.2** 41044 192.168.0.141 45000 Anomaly Transport FIR

First we\'d like to thank V. Malkhasyan and Jan Hrbacek \[[@CIT0001], [@CIT0002]\] for their kind comments. We thank Jan Hrbácek for his kind comments regarding our institution, we\'re all very glad your time here was such a positive experience. Jan Hrbacek is correct in pointing out some of the major shortcomings of the study such as not being able to achieve randomization despite our original intentions as evidenced by the difference in demographics between the groups. However, the selection of the groups was a compromise between true randomization, and patient treatment ethics. As with any selection process, including randomization, bias can be introduced if all potential confounding factors aren\'t taken into account from the outset of the study. All must be done within current standard of care. After consulting with all personnel involved, we agreed on the methodology employed. Our selection process resulted in one we were comfortable for finding answers to our questions while maintaining our standards for clinical care. While there was a lack of true randomization and the patient groups differed significantly, the difference in results cannot be solely attributed to patient selection as the stone free rates achieved were not necessarily correlated with the severity of the disease as can be evidenced by the less effective results of achieved by patients eligible for ESWL. Our results still permit gaining a good perspective on (or adds to the current body of knowledge regarding) risk, cost, benefit considerations that occur with each modality in clinical practice. They both raise the issue of lack of comparison with ureteroscopic stone ablation using laser. Both include the EAU Guidelines for Urolithiasis updated in 2011 in their reference list. The guidelines state that Ho:YAG laser lithotripsy is the preferred method when carrying out (flexible) URS with a grade B recommendation. V. Malkhasyan makes a point regarding improvements in URS regarding caliber and additional lithotrypsy URS devices such as laser. These improvements in devices and techniques make comparisons of treatment modalities a moving target and, as newer devices and techniques arrive, the subject of further study. Regarding URS laser ablation, this modality wasn\'t readily available at our institution during the period of study, which is representative of many institutions at the time.

Introduction {#s1} ============ In the general case, peptides are called natural substances consisting of submolecular blocks (20 standard amino acid residues) connected by a peptide bond (Sewald and Jakubke, [@B11]). Obviously, the minimal linear peptide is a chemical structure in which two amino acid residues are connected by one peptide bond (dipeptide). Natural peptide sequences (primary structures) containing many amino acid residues (polypeptides) are commonly called proteins, and long molecules of this type can consist of tens of thousands of amino acid residues. Thus, the size (length) of a molecule of peptide nature can vary over a wide-ranging number of amino acid residues. There is a great variety of amino acid residue sequences as well. This is based on a broad spectrum of possible combinations of 20 standard amino acid residues. With an increase in the length of the peptide sequence *p*, the number of such combinations of *N*~*p*~ rapidly grows in accordance with the well-known formula N p = 2 0 p , (where *p* is a number of amino acid residues in one molecule). In the case of dipeptides, *p* = 2, *N*~2~ = 400, tripeptides *p* = 3, *N*~3~ = 8,000, tetrapeptides *p* = 4, *N*~4~ = 160,000, etc., and at the end of the interval 2 ≤ *p* ≤ 50, characterizing oligopeptides (Zamyatnin, [@B14], [@B19]), i.e., at *p* = 50, this value reaches the value *N*~50~ = \~10^34^. Of course, not all combinations of amino acid residues can exist in nature, but still the variety of possible primary structures should be accepted as gigantic. In this work, we studied the distribution of the number of different (unique) natural peptide sequences along the length, i.e., by the number of amino acid residues *p*. Materials and Methods {#s2} ===================== To obtain information on natural peptide amino acid residue sequences, the UniProt database is most often used, which combines manually annotated and reviewed data (Swiss-Prot database Bairoch and Boeckmann, [@B2] and the TrEMBL database Kneale and Kennard, [@B8] on primary structures obtained as a result of the translation (Tr) of nucleotide sequences into the language of amino acids (automatically annotated and not reviewed). The ability to perform different procedures and analyses is provided on the UniProt database website. The UniProt database program tools were used in our work to highlight all and specific amino acid residue sequences (option: Search), exclusion of sequence fragments (option: Sequence \> Fragment \> Sequence complete), extraction of specific sequences with a given number of amino acid residues *p* (option: Sequence \> Sequence length \> from *p*~1~ to *p*~2~), exclusion of identical sequences (option: Protein page \> Similar proteins \> 100% identity), and sorting protein names and sequences by different characteristics (standard UniProt table sorting). The procedures described in this work were performed on a high-speed server, which allows you to process large amounts of information. The main part of the study described below was carried out on the Swiss-Prot database data. Results {#s3} ======= Protein Length Distribution in UniProt Database ----------------------------------------------- At the time of the study, the UniProt database contained information on 159,022,877 amino acid residue sequences obtained for species of archaea, prokaryotes, and eukaryotes. The minimum number of amino acid residues *p* = 2 contains three oligopeptides of a different origin, and the maximum (*p* = 74,488) contains one translated bacterial polypeptide. Despite such a large range of *p*, most sequences are concentrated in the range 2 ≤ *p* ≤ 1,000 (154,685,385, i.e., more than 97%). The distribution in this interval is shown in [Figure 1A](#F1){ref-type="fig"}. A characteristic feature of this distribution is several sharp peaks, indicating a significantly larger number of sequences with a given number of amino acid residues *p*, compared with neighboring values of *p*, e.g., a peak is especially distinguished at *p* = 252. An examination of primary structures corresponding to this peak represents a large set of very functionally different proteins. {#F1} However, this distribution does not give information on the true number of occurrences of amino acid residue sequences in nature. These data are incorrect, since the UniProt database contains, in addition to complete sequences, data on 15,495,873 incomplete sequences (almost 10%) that are fragments. The distribution shown in [Figure 1B](#F1){ref-type="fig"} characterizes an array of 143,527,004 natural peptide amino acid residue sequences after excluding all fragments from consideration. Several peaks are also detected in this distribution, but their values and position are different from those shown in [Figure 1A](#F1){ref-type="fig"}. As it transpired during the study, the UniProt database also includes sequences containing amino acid residues indicated by letters that are not used to describe standard residues: O (hydroxyproline), U (α-aminobutyric acid), and X (unidentified). Sequences that contain these letters were also excluded from consideration. The UniProt database also contains a large number of completely identical sequences, obtained, as a rule, as representatives of different, but taxonomically close living organisms. In order for the analysis to be limited to considering only different sequences, all duplicates were also excluded from consideration ([Figure 1C](#F1){ref-type="fig"}). It can be seen that the peak at *p* = 252 is absent in [Figure 1C](#F1){ref-type="fig"} because there are many duplicates in all sequences of UniProt database ([Figures 1A,B](#F1){ref-type="fig"}). This peak contains more than 57,000 sequences of viral matrix proteins 1, but many of them are the same, e.g., the protein sequence (V9SYV1) of influenza A virus \[A/Peru/PER334/2011(H3N2)\] is found in more than 10,000 entries of UniProt database. Protein Length Distribution in Swiss-Prot Database -------------------------------------------------- At the time of the study, the Swiss-Prot database contained information on 560,118 amino acid residue sequences obtained as representatives of archaea, prokaryotes, and eukaryotes. The minimum number of amino acid residues *p* = 2 contained two oligopeptides of a different origin, and the maximum (*p* = 35,213) contained the mouse protein titin (Church et al., [@B3]). It transpired that, despite such a large range of *p*, most sequences are concentrated in the range 2 ≤ *p* ≤ 1,000 (542,302, i.e., \~97%). The distribution of all sequences in this interval is shown in [Figure 2A](#F2){ref-type="fig"}. A characteristic feature of this distribution is the many sharp peaks indicating a significantly larger number of sequences with a given number of amino acid residues *p*, compared with neighboring *p*-values. Among them, two peaks are especially distinguished at *p* = 156 and 379. An examination of the protein corresponding to these peaks showed that, at *p* = 156, 2,216, sequences represent a large set of very different proteins. However, in the case *p* = 379, out of 1,889 proteins, 1,048 (two-thirds) are mitochondrial cytochrome *b*. This value is approximately equal to that part of the peak that rises above the total mass of protein lengths in [Figure 2A](#F2){ref-type="fig"}. {#F2} As in the case of the UniProt database, we deleted all incomplete sequences that are fragments in the Swiss-Prot database. The distribution of complete sequences is shown in [Figure 2B](#F2){ref-type="fig"}. This figure does not differ much from [Figure 2A](#F2){ref-type="fig"}, since fragmented sequences from the Swiss-Prot database make up only 1.6% (9,167) of the total number of data in this database. Some difference is noticeable only in the field of oligopeptides, i.e., at small values of *p*. We obtained a significantly changed distribution after removing sequences with non-standard and unidentified amino acid residues, and also after removing duplicates ([Figure 2C](#F2){ref-type="fig"}). Note that the peak value at *p* = 156 decreased significantly, while at *p* = 379, it remained unchanged. Detailed data on the number of amino acid residue sequences taken for the already described and subsequent analyses are collected in [Table 1](#T1){ref-type="table"}. ###### Data on the number of amino acid residue sequences in various taxonomic groups and species on the Swiss-Prot database. **Taxon groups, species (common names)** **Taxon groups, species (Latin names)** **All on the Swiss-Prot database** **Without fragments** **Without duplicates** ***p*~**min**~** ***p*~**max**~** ------------------------------------------ ----------------------------------------- ------------------------------------ ----------------------- ------------------------ ------------------ ------------------ All 560,118 550,951 463,450 2 35,213 Eukaryotes *Eukaryota* 189,697 182,592 174,063 2 35,213 Animals *Metazoa* 106,843 102,155 98,070 2 35,213 Human *Homo sapiens* 20,421 20,421 20,358 2 34,350 Plants *Viridiplantae* 39,930 38,014 35,093 5 5,400 Mouse-ear cress *Arabidopsis thaliana* 15,856 15,829 15,768 5 5,400 Fungi *Fungi* 34,084 33,841 32,431 3 11,842 Yeast *Saccharomyces cerevisiae* 7,919 7,912 7,290 16 4,910 Prokaryotes *Bacteria* 334,009 332,477 255,373 7 10,746 Bacteria gram- *Escherichia coli* 23,138 23,121 10,153 7 3,289 Bacteria gram+ *Staphylococcus aureus* 10,175 10,164 3,171 9 10,746 Archaea *Archaea* 19,554 19,482 18,452 25 9,159 Viruses *Viruses* 16,858 16,400 15,607 11 7,182 We sequentially isolated and analyzed unique sequences of taxonomic groups of various levels from the entire mass of sequences from the Swiss-Prot database. At the first and highest level, the domains of archaea, prokaryotes, and eukaryotes were chosen (Woese et al., [@B13]). Despite the fact that the number of sequences in these domains varies significantly ([Table 1](#T1){ref-type="table"}), the general distribution pattern ([Figure 3](#F3){ref-type="fig"}) is the same for them. Moreover, the overwhelming majority of amino acid residue sequences, as before, are concentrated in the range 2 ≤ *p* ≤ 1,000 (99.1% in archaea, 98.7% in prokaryotes, and 92.7% in eukaryotes). In archaea ([Figure 3A](#F3){ref-type="fig"}), there is only one small visible peak at *p* = 130; in prokaryotes ([Figure 3B](#F3){ref-type="fig"}), there are quite a lot of these peaks, and the peak is especially distinguished at *p* = 156 represent a large set of very different proteins; and in eukaryotes ([Figure 3C](#F3){ref-type="fig"}), most of the peaks are not too large, except for one at *p* = 379. {ref-type="fig"}--[8](#F8){ref-type="fig"}: **blue**, archaea; **red**, prokaryotes; **brown**, all eukaryotes and animals; **violet**, fungi; **green**, plants; **orange**, viruses.](fmolb-07-00102-g0003){#F3} At the next taxonomic level, the unique amino acid residue sequences of the kingdoms (fungi, plants, and animals) were separately analyzed ([Figure 4](#F4){ref-type="fig"}). The general nature of the obtained distributions indicates that noticeable peaks are practically unobservable in fungi ([Figure 4A](#F4){ref-type="fig"}), while in plants, there are quite a lot of them ([Figure 4B](#F4){ref-type="fig"}), and in the case of animals, a number of peaks also stand out, among which the peak at *p* = 379 is brighter than in the case of the distribution for eukaryotes ([Figure 3C](#F3){ref-type="fig"}). {#F4} We also analyzed the unique amino acid residue sequences of individual representatives of fungi, plants, and animals ([Figure 5](#F5){ref-type="fig"}). In the case of yeast (*Saccharomyces cerevisiae*), a distribution was obtained with a noticeable peak at *p* = 440 ([Figure 5A](#F5){ref-type="fig"}). It turned out that most of the sequences with this number of amino acid residues are various proteins known as transposon polyproteins. Several not very large peaks are manifested in mouse-ear cress ([Figure 5B](#F5){ref-type="fig"}). At the same time, two peaks clearly stand out in human ([Figure 5C](#F5){ref-type="fig"}). One of them at *p* = 117 is more than 50% composed of various immunoglobulins; in the case of the other at *p* = 312, also more than half characterizes the presence of a large number of olfactory receptor proteins. {#F5} We also obtained distributions for a combination of sequences of Gram-negative ([Figure 6A](#F6){ref-type="fig"}) and Gram-positive ([Figure 6B](#F6){ref-type="fig"}) bacteria and for all viruses ([Figure 7](#F7){ref-type="fig"}). All these distributions are characterized by a large number of noticeable peaks. An examination of primary structures corresponding to bacterial protein peaks represents a large set of very different proteins except two peaks in the distribution of viral sequences. The peak at *p* = 252 is particularly prominent (see [Figures 1A,B](#F1){ref-type="fig"}), in which a significant part is occupied by the matrix protein of influenza A virus, which plays a key role in the replication of viruses, and the peak at *p* = 498 almost completely corresponds to the nucleoprotein protein (also in influenza A virus), which protects viral RNA from nucleases. {#F6} {#F7} Common Consideration of Protein Length Distribution {#s4} =================================================== Obviously, the nature of any distribution needs explanation, which could best be made on the basis of an analytical mathematical expression derived from general considerations about the nature of peptide sequences. Obtaining such an expression is still faced with great difficulties, since many factors must be taken into account: details of the genesis of proteins, features of its constituent elements (physicochemical diversity of amino acid residues), evolutionary factors, and much more (Zamyatnin, [@B17], [@B18]). However, the fitting mathematical expression, which allows one to obtain curves that are as close as possible to real distributions, has been carried out repeatedly. Thus, for a number of individual organisms, the log-normal distribution function was used, which satisfactorily described the data for 13 species of bacteria, 4 archaea, and 1 eukaryote (Ramakumar, [@B10]). The same approach was applied to describe the protein distributions of 1,302 species of prokaryotes and 140 eukaryotes (Tiessen et al., [@B12]). In this study, in addition to the log-normal distribution, a gamma-type distribution was also employed. The combined use of the log-normal and gamma distributions (Jhang, [@B7]) allowed the authors to conclude that the average length of eukaryotic proteins is greater than that of prokaryotes. A similar conclusion is illustrated in our present work when considering [Figures 3B,C](#F3){ref-type="fig"}. The linguistic model of Menserath--Altmann was also used no less successfully (Menzerath, [@B9]; Altmann, [@B1]). The data for 10 proteomes with help from this model were described (Eroglu, [@B4]). Unique Length Distribution for One Protein Type {#s5} =============================================== However, in all of these works, the main attention was paid to smoothing the distribution curves in order to fit certain mathematical models. At the same time, it follows from our results that numerous peaks in almost all taxonomic groups and individual species of living organisms carry additional information about the sequences that form these distributions. A striking example is the peak traced in the distribution of all amino acid residue sequences on the Swiss-Prot database ([Figure 2C](#F2){ref-type="fig"}), eukaryotes ([Figure 3C](#F3){ref-type="fig"}), and animals ([Figure 4C](#F4){ref-type="fig"}). As already noted, this peak is a collection of mitochondrial cytochromes *b* having the same length of 379 amino acid residues and found only in animals. An additional analysis of the length distribution of animal cytochromes *b* ([Figure 8](#F8){ref-type="fig"}) showed that these proteins have a predominant length of 379 amino acid residues (1,028 proteins). With a slightly smaller number of residues, there are 99 of these proteins and with a large number of 440. {#F8} For a large number of differing amino acid residue sequences, but with the same *p*-value, cytochrome *b* is precisely typical as being representative of animals. Swiss-Prot contains data on nearly 1,700 different sequences of this protein, obtained not only from animals. They are present in bacteria, plants, and fungi. However, all known non-animal cytochromes *b* always contain more than 379 amino acid residues. It can be assumed that the minimum value of this transmembrane protein in the evolutionary process could be achieved as a result of selection, which led to the most optimal sizes of all its eight sites penetrating the membrane (Esposti et al., [@B5]). Such examples may indicate that the most perfect protein lengths were selected in the evolutionary process to perform this function. As a result, many molecules with different sequences of the same length and characterized by the same functions were formed. Obviously, the diversity of the functional properties of proteins is based on the diversity of the primary structures of their molecules. Many examples can be found where small peptide molecules of the same length (oligopeptides) and with the same *p* and different sequences have the same functions (Zamyatnin, [@B15],[@B16], [@B20]), e.g., pentapeptides met-enkephalin YGGFM and leu-enkephalin YGGFL) are the natural ligand for opiate receptor (Hughes et al., [@B6]). In this work, we particularly demonstrated that there are many different primary structures of polypeptides (proteins) of the same length, with the same functions. Future Directions {#s6} ================= The UniProt and Swiss-Prot databases are constantly updated. However, there is already a large amount of information contained on them, which allows for many new and various analyses of the relationship between size, amino acid residue sequences, and many other physicochemical characteristics of natural sequences with the numerous functional properties of these molecules. Data Availability Statement {#s7} =========================== The datasets generated for this study are available on request to the corresponding author. Author Contributions {#s8} ==================== AZ performed all the experiments, treatment of their results, and wrote the manuscript. TB contributed to the biological consideration of the results and in writing the article. All authors contributed to the article and approved the submitted version. Conflict of Interest {#s9} ==================== The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. We thank Dr. Y. Chaikovsky, Prof. A. G. Malygin, Prof. R. V. Polozov, Prof. M. D. Frank-Kamenetskii, and Prof. L. S. Yaguzhinsky for fruitful discussions and comments on the manuscript, and A. S. Borchikov and M. G. Vladimirov for technical assistance. [^1]: Edited by: Qi Zhang, Fudan University, China [^2]: Reviewed by: Jiafeng Geng, Consultant, United States; Yi Yu, Wuhan University, China [^3]: This article was submitted to Protein Chemistry and Enzymology, a section of the journal Frontiers in Molecular Biosciences

1.. Introduction ================ In the past a few decades, long-period fiber gratings (LPFG) have been broadly investigated for both communication \[[@b1-sensors-13-14055]\] and sensing applications \[[@b2-sensors-13-14055]\]. Traditional long period gratings (LPGs) are inscribed in a germanium (Ge)-doped fiber or a hydrogen-loaded single-mode fiber. Compared with fiber Bragg gratings (FBGs), LPGs are fabricated using an amplitude mask instead of an expensive phase mask, thus making the overall cost of LPGs sensing system advantageous to a certain extent. Apart from the amplitude mask method, there are also other techniques for fabricating LPGs in optical fibers such as lithography method which has been reported recently \[[@b3-sensors-13-14055],[@b4-sensors-13-14055]\], or using a CO~2~ laser \[[@b5-sensors-13-14055]\] to create periodic indentations on the fiber surface, and using a femtosecond laser to create refractive index variations \[[@b6-sensors-13-14055]\]. These techniques will cause physical deformation in the fiber. Recently, tapered fibers have attracted increasing attention since they have large evanescent fields and low-loss interconnection with single-mode fibers \[[@b7-sensors-13-14055]\]. These fibers can be used for optical data storage \[[@b8-sensors-13-14055]\] or filters \[[@b9-sensors-13-14055]\], and they were also found to be popular in sensing applications including acceleration \[[@b10-sensors-13-14055]\], temperature \[[@b11-sensors-13-14055]\], and external refractive index (RI) \[[@b12-sensors-13-14055]--[@b14-sensors-13-14055]\] sensing. In particular for RI sensors, researchers have combined LPGs and tapered fibers together in order to increase the sensitivity \[[@b5-sensors-13-14055],[@b15-sensors-13-14055]\]. In this work, we present long period gratings fabricated in a tapered photosensitive fiber using a KrF excimer laser. The fiber has double claddings and the inner cladding also has photosensitivity. The fiber was tapered adiabatically before writing the gratings and the resulting loss is within 1 dB. To the best of our knowledge, there has been no report on LPGs written in such a type of tapered fiber. We noted that for such type of microfiber long period grating (MLPG) sensor, the resonant wavelength of the LPG shifts to the longer wavelength when the external RI increases, in contrast to the normal LPG where the resonant wavelength shifts in the opposite direction to the shorter wavelength. We have successfully fabricated two such long period gratings with taper waist diameters of 60 μm and 55 μm. The sensitivity achieved in the RI region around 1.37 is around 600 nm/RIU which represents a significant improvement compared to normal LPG RI sensors. The criterion for adiabaticity of fiber taper is derived from the physical argument that taper length-scale must be much larger than the coupling length between the fundamental mode and the dominant coupling mode for power loss to be small \[[@b16-sensors-13-14055]\]. As illustrated in [Figure 1](#f1-sensors-13-14055){ref-type="fig"}, the taper angle is defined as: $$\Omega(z) = \tan^{- 1}\left( \frac{d\rho(z)}{dz} \right)$$where ρ(z) is the radius along the z axis. In this criterion, the taper angle is assumed to be sufficient small so that: $$\text{z}_{t} \approx \frac{\rho}{\Omega}$$ Assuming the taper is axisymmetric, as is usually the case with a single fiber taper, then the fundamental LP~01~ mode can couple only to the modes with the same azimuthal symmetry, *i.e.*, to the higher order LP~0~*~m~* cladding modes. If we are aiming to minimize loss from the fundamental mode, it is obvious that coupling will be mainly to the higher-order mode with propagation constant that is closest to the fundamental mode, *i.e.*, the LP~02~ mode. The coupling length between the two modes is taken to be the beat length between the fundamental mode and higher-order mode, and takes the form: $$\text{z}_{b}(z) = \frac{2\pi}{\beta_{1}(z) - \beta_{2}(z)}$$where *β*~1~(*z*). and *β*~2~(*z*) are the propagation constants of the two modes. When *z~t~* ≫ *z~b~*(*z*) everywhere along the taper, the coupling loss will be negligible and the fundamental mode will propagate nearly adiabatically, whereas significant coupling will occur if *z~t~* ≪ *z~b~*(*z*) This indicates that *z~t~* = *z~b~*(*z*) is a condition just in between adiabatic and lossy tapers. Therefore, the taper angle can be derived as: $$\Omega(z) = \frac{\rho(z)(\beta_{1}(z) - \beta_{2}(z))}{2\pi}$$ This criterion indicates that when the core diameter is reduced to a small amount, in order to maintain low loss taper, the taper angle also needs to be sufficient small by means of making the fiber taper vary gradually. For normal LPGs, the phase matching condition of coupling power from higher order mode to fundamental mode is shown to be \[[@b17-sensors-13-14055]\]: $$\left| \beta_{1} - \beta_{2} \middle| = \frac{2\pi}{\Lambda} \right.$$where *β*~1~ and *β*~2~ are the propagation constants of the fundamental mode and higher order mode, Λ is the period of the gratings. It can be easily derived from [Equation (5)](#FD5){ref-type="disp-formula"} that the resonant wavelength under this phase matching condition takes the form: $$\lambda_{\textit{res}} = (n_{co}^{\textit{eff}} - n_{\textit{clad},m}^{\textit{eff}})\Lambda$$ Once the external refractive index increases, the effective refractive index of the mth cladding mode $n_{\textit{clad},m}^{\textit{eff}}$ will increase thus resulting a shift of resonant dip towards shorter wavelength. 2.. Experimental Section ======================== There are several ways of obtaining fiber tapers. The conventional ways are by the flame brushing method whereby, using a fusion splicer or a CO~2~ laser to soften the fiber, coupled with a mechanical setup to stretch it. In our work, we have used a commercial fiber tapering machine (GPX-3000, Vytran, Morganville, NJ, USA) with motorized fiber holding stages. It is observed that the taper transition length, waist length, and waist diameter can be precisely controlled through this process so that the adiabaticity and uniformity can be ensured. An example of the transmission spectra before and after tapering is shown in [Figure 2](#f2-sensors-13-14055){ref-type="fig"}. The taper transition lengths are around 10 mm each and the taper waist length is around 20 mm. It can be seen that the loss is within 1 dB which is nearly negligible. More importantly, the flat response indicates that there is no higher-order excited mode. In this work, the fibers which are tapered to 60 μm and 55 μm were both adiabatically stretched in order to obtain a flat transmission spectrum. It is crucial to ensure the fiber is tapered gradual enough otherwise interference will occur as \[[@b18-sensors-13-14055]\] demonstrates. After tapering, the fiber was exposed to a 248 nm KrF excimer laser with a pulse frequency of 10 Hz to fabricate the LPG. The periods of the amplitude mask for diameters of 60 μm and 55 μm are 320 μm and 450 μm, respectively. [Figure 3](#f3-sensors-13-14055){ref-type="fig"} shows a schematic diagram of the LPG fabrication setup. The single-mode fiber used in the experiment is the photosensitive double-clad fiber with High Ge-doped core (22 mol%) whose diameter is around 3.6 μm. The inner cladding has a diameter of 25 μm and is lightly doped with Ge and boron (B) (1 mol%). The co-doping of boron in germanosilicate fiber enhances the UV photosensitivity of the fiber \[[@b19-sensors-13-14055],[@b20-sensors-13-14055]\]. The remaining area is the outer cladding with diameter of 125 μm. The refractive index distribution profile is W-shape as shown in [Figure 4](#f4-sensors-13-14055){ref-type="fig"}. The core has the highest RI while the inner cladding has the lowest. In the tapered fiber with waist diameter of 60 μm, the core diameter is reduced to 1.73 μm while the inner cladding diameter becomes 12 μm. This can be obtained by knowing that the core and inner cladding\'s sizes are changing proportionally with outer cladding \[[@b21-sensors-13-14055]\], *i.e.*, n~core~ = 3.6 μm (original core diameter)× 60 μm (original inner cladding diameter)/125 μm (original outer cladding diameter) = 1.73 μm and n~ic~ = 25 μm × 60 μm/125 μm = 12 μm. To further enhance the photosensitivity, the tapered fibers were loaded with hydrogen at a temperature of 70 °C and a pressure of 2,000 psi for 192 h. The length of the gratings is 18 mm. The resonant dips of the two LPGs are observed at around irradiation of 2,000 pulses with fluence of 200 mJ/cm^2^. We stopped to adding more pulses until the resonant wavelength positions tended to stabilize. After the gratings were written, the fibers were annealed at 120 °C for 20 h to dispel the hydrogen that may still be present in the fiber. 3.. Results and Discussion ========================== To characterize the refractive index sensitivity of the sensor, the sensing region, which is the taper waist, was surrounded by a few droplets of salt solution. The salt solutions with different refractive index were prepared by diluting the saturated NaCl solution using distilled water. The refractive index of the solution was determined using a refractometer (KEM RA-130, Kyoto Electronics Manufacturing, Kyoto, Japan) with a resolution of 0.0001. The resulting wavelength responses are depicted in [Figure 5](#f5-sensors-13-14055){ref-type="fig"}. In contrast to the expected behavior of the transmission spectrum for conventional LPGs, we noted that the resonant dip of the fabricated LPG shifts towards longer wavelength when the external refractive index increases. The main reason for such phenomenon is due to the W-shaped refractive index profile of the fiber as discussed in the aforementioned section. After tapering, the original core diameter of 3.6 μm would have become smaller and does not contribute significantly to the core mode. In addition, it is believed that bulk of the Ge in the fiber core have dispersed into the inner cladding during the tapering process but hardly into the outer cladding \[[@b21-sensors-13-14055]\]. In other words, for the tapered fiber, the effective refractive index of the core mode after tapering is lower than the refractive index of the outer cladding. The phase matching condition for coupling between the higher-order modes to fundamental mode can still be derived from [Equation (5)](#FD5){ref-type="disp-formula"}, but the resonance wavelength is now given by: $$\lambda_{\textit{res}} = (n_{\textit{clad},m}^{\textit{eff}} - n_{co}^{\textit{eff}})\Lambda$$ According to the information provided by the manufacturer of the specialty fibre, the refractive index of the core is 1.4650, the inner cladding is 1.4403, and the outer cladding is 1.4476. The guiding mechanism after tapering was also simulated. It was found that such a tapered specialty fiber still supports single mode only. The mode field diameter of the new core mode is calculated to be 4.43 μm, which is much larger than the new core\'s diameter 1.73 μm, but much smaller than the new inner cladding\'s diameter 12 μm. Therefore, although the refractive index of the core is still much higher than the inner cladding and outer cladding, the effective index of the core mode will be greatly influenced by the inner cladding. As expected, the effective index of the core mode was calculated as 1.4443 which is smaller than refractive index of the cladding. This can justify the validity of [Equation (7)](#FD7){ref-type="disp-formula"}. For this simulation, we have no way to input the factors of how much Ge in the core is diffused to the inner cladding into consideration. Therefore, the value of the effective index 1.4443 actually will not be accurate enough. However, since after tapering the inner cladding size is much larger than the core size, the increase of the refractive index of the inner cladding caused by diffusion is nearly negligible. Now it can be seen from [Equation (7)](#FD7){ref-type="disp-formula"} that the resonant wavelength will shift to a longer wavelength due to increase of external refractive index. A broader spectrum is shown in [Figure 6](#f6-sensors-13-14055){ref-type="fig"}, from which one can see that the couplings between the fundamental mode and the higher-order modes that have closer propagation constants with fundamental mode are very weak. This is because the power of the modes in the inner cladding is weak due to the higher refractive index of the outer cladding. The resonance dip become smaller as the index of liquid under test increases actually is not related to adiabaticity of the taper. For LPG, the transmission loss for the resonant wavelength will disappear when *n~ext~* = *n~cl~* since the core will see infinitely thick cladding region and without the cladding-surrounding interface, there is an absence of guided region and thus the cladding modes ceases to exist \[[@b22-sensors-13-14055]--[@b24-sensors-13-14055]\]. Therefore when the external refractive index gets closer to that of cladding, more power of cladding modes will be converted to power of radiation modes due to the lack of total internal reflection at the cladding-surrounding interface, resulting in a diminutive loss in the transmission spectrum. The maximum RI sensitivity achieved in our experiments is around 600 nm/RIU. With our OSA\'s resolution of 10 pm, the detection limit for refractive index is 1.67 × 10^−5^. [Figure 7](#f7-sensors-13-14055){ref-type="fig"} illustrates the sensitivity characterization of two sensors with different taper diameter. The slope of the fiber with smaller outer diameter (55 μm), plotted using squares, and linearly fitted by purple line is slightly larger than the slope of the fiber with larger outer diameter, plotted using circles. As the diameter goes thinner, more evanescent field of the modes in outer cladding will penetrate into the surrounding thus the effective index of the outer cladding modes will be more easily impacted by the change of external refractive index. In another word, for same increment of external refractive index, the sensor with smaller diameter will have larger increment of effective index of the outer cladding modes. Therefore, the wavelength shift will be larger which can be obtained from [Equation (7)](#FD7){ref-type="disp-formula"}. 4.. Conclusions =============== In summary, we have investigated long period gratings inscribed in an adiabatically tapered photosensitive specialty fiber with a W-shaped refractive index profile for refractive index sensing. The adiabaticity of the taper is achieved by making the fiber taper gradual enough. We have fabricated two MLPG sensors with different diameters and compared their sensitivity. The one with smaller diameter achieves a high RI resolution of 1.67 × 10^−5^ around the RI range of 1.38. The wavelength shift differs from conventional LPGs that the resonant wavelength shifts towards longer wavelength when external refractive index increases. Although the fiber is tapered into much a smaller diameter, the change in taper diameter is gradual enough to ensure the robustness of the fiber in terms of the mechanical strength. Therefore, such sensor has the potential for applications in environmental detection and biosensing areas. This work was supported by the Ministry of Education *via* Nanyang Technological University, Singapore. The authors declare no conflict of interest. {#f1-sensors-13-14055} {#f2-sensors-13-14055} {#f3-sensors-13-14055} {#f4-sensors-13-14055} {#f5-sensors-13-14055} {#f6-sensors-13-14055} {#f7-sensors-13-14055}

There is current interest in the design and manufacture of various two-dimensional (2D) ultrathin layered nanosheet materials[@b1][@b2]. 2D nanostructures may find use in more efficient energy-producing, energy-absorbing, and energy storage devices, because their ultrahigh fraction of surface atoms could enable nearly full use of active materials[@b3][@b4][@b5][@b6][@b7][@b8][@b9]. 2D nanomaterials can be prepared through either a "bottom-up" approach[@b10], including assembly of nanoclusters[@b11][@b12], coalescence of nanowires[@b13][@b14], and oriented attachment of nanocrystals[@b15][@b16], or a "top-down" method, e.g., exfoliation of layer-structured bulk materials[@b17][@b18][@b19][@b20]. Both "bottom-up" and "top-down" synthesis techniques usually involve the use of organic solvents, polymeric stabilizers, surface capping agents, and burdensome post-processing. Moreover, there is still a lack of effective methods to integrate 2D nanosheets into three-dimensional (3D) hierarchical macroscopic structures with active materials at exposed/available surfaces in the open-framework 3D structure[@b21][@b22]. Nickel oxide (NiO) has been recognized as one of the most promising electrode materials for pseudocapacitor applications in aqueous alkaline electrolytes due to its high theoretical specific capacitance (2,584 F g^−1^ within 0.5 V), low cost, and excellent chemical and thermal stability[@b23][@b24][@b25][@b26][@b27][@b28][@b29][@b30][@b31]. However, the poor electrical conductivity of NiO increases both the sheet resistance and the charge transfer resistance of the electrode, which in work reported to date has led to a capacitance value much less than the theoretical value and a poor rate capability[@b29][@b30]. Consequently, various methods, such as modification with nanostructured carbon[@b24][@b25][@b26], preparation of Ni-NiO nanocomposites[@b27][@b28][@b29][@b30][@b31], and synthesis of nanostructures with a unique morphology[@b32][@b33][@b34][@b35][@b36], have been proposed to improve the NiO supercapacitor performance. In recent years, there have been numerous reports on preparing hierarchical NiO structures that are promising for energy storage systems because of their extraordinary properties, such as high surface area and unique geometrical structure[@b37][@b38][@b39][@b40][@b41][@b42][@b43]. Despite the advancements, very little research has been done to investigate the formation mechanism of such 3D hierarchical structures, and extensive research is still required to explore effective strategies for achieving high-efficiency utilization of NiO with further performance improvement. We herein present a systematic research to the formation of a hierarchical NiO nanostructure with multiple characteristic dimensions, and propose a 'weaving' mechanism for the formation based on abundant experiments. The 3D NiO 'microflowers' are constructed of 2D ultrathin nanosheets that can be visualized as "woven" from one-dimensional (1D) nanowires. Benefiting from such a unique structure, an interconnected nickel (Ni) conductive network can be generated *in situ* on the surface of the NiO nanosheets through a controllable reduction process. The resulting Ni-NiO hybrid has an open hierarchical 3D structure and good electrical conductivity, as shown in [Fig. 1](#f1){ref-type="fig"}, which offers great potential for supercapacitor applications, including easy transport of ions/electrolytes at the interface of the electrode/electrolyte, fast electron transfer on the surface of the nanosheets, and high-efficiency use of the active material NiO. Results ======= Synthesis --------- The overall method is described in in the supplementary material ([Supplementary, Fig. S1](#S1){ref-type="supplementary-material"}). Briefly, the precursor was prepared by dissolving sodium dodecyl sulphate (SDS), urea, and NiCl~2~ in water. The mixed solution was then transferred to an autoclave for hydrothermal treatment. Several parameters, such as the proportion of reactants, temperature, and reaction time, were examined which in turn made it possible to explore the formation mechanisms of the corresponding nanostructures. In our synthesis process, only urea and SDS were used as auxiliary reagents to grow Ni-based nanostructures. The hydrolysis of urea led to the generation of OH^−^ and CO~3~^2−^ which then reacted with Ni^2+^ to form a green NiCO~3~∙Ni(OH)~2~∙xH~2~O precipitate ([Supplementary, Fig. S1b](#S1){ref-type="supplementary-material"} and [Fig. S2a](#S1){ref-type="supplementary-material"}). The resulting precipitate was subsequently decomposed to a black NiO powder upon annealing ([Supplementary, Fig. S1c](#S1){ref-type="supplementary-material"} and [Fig. S2b](#S1){ref-type="supplementary-material"}), as proven by thermogravimetric analysis (TGA). The NiO sample was then partially reduced to produce a grey Ni-NiO sample ([Supplementary, Fig. S1d](#S1){ref-type="supplementary-material"}) under an H~2~/Ar atmosphere. Microflowers constructed from Ni nanosheets ------------------------------------------- When no SDS was involved in the reactions, the morphology of the samples was strongly influenced by the molar ratio of NiCl~2~/urea. Porous Ni-based nanoplates were formed with a high molar ratio NiCl~2~:urea = 5:1 ([Supplementary, Fig. S3a, b](#S1){ref-type="supplementary-material"}); and microspheres constructed from nanowires and visually similar to silkworm cocoons were produced with a molar ratio NiCl~2~:urea = 3:1 ([Supplementary, Fig. S3c, d](#S1){ref-type="supplementary-material"}); while a porous microsphere with well-defined nanowire building blocks was obtained for a molar ratio NiCl~2~:urea = 1:1 ([Supplementary, Fig. S3e, f](#S1){ref-type="supplementary-material"}). For NiCl~2~:urea = 1:2, the microsphere size decreased and the nanowires became less discernible, possibly due to the high agglomeration of nanowires ([Supplementary, Fig. S4a, b](#S1){ref-type="supplementary-material"}). When the molar ratio was decreased to NiCl~2~:urea = 1:10, the 'microspheres' were deformed ([Supplementary, Fig. S4c, d](#S1){ref-type="supplementary-material"}). For no urea and excess urea (e.g., a molar ratio NiCl~2~:urea = 1:30), there was no solid product formed. The influence of the concentration of SDS on the product morphology was also studied with both NiCl~2~ and urea fixed at 5 mmol. For 0.1 g SDS, two types of nanostructures formed ([Supplementary, Fig. S5a-c](#S1){ref-type="supplementary-material"}): one was nanowire-constructed porous microspheres ([Fig. S5b](#S1){ref-type="supplementary-material"}) that are similar to the products ([Fig. S3e, f](#S1){ref-type="supplementary-material"}) obtained without SDS, and the other was 'microflowers' composed of nanosheets that were evidently formed from "woven" nanowires ([Fig. S5c](#S1){ref-type="supplementary-material"}). For 0.3 g SDS, the samples also contained nanowire-constructed porous microspheres ([Fig. S5e](#S1){ref-type="supplementary-material"}) and nanosheet-constructed 'microflowers' ([Fig. S5f](#S1){ref-type="supplementary-material"}), but a larger fraction of microflowers ([Supplementary, Fig. S5d](#S1){ref-type="supplementary-material"}) was observed compared with the samples obtained with 0.1 g SDS. Notably, for 0.6 g SDS added into the reaction system, only microflowers were obtained as shown in [Fig. 2](#f2){ref-type="fig"}, which presents the morphology and the crystalline structure of the as-prepared NiO microflowers. [Figure 2a](#f2){ref-type="fig"} shows a typical scanning electron microscope (SEM) image, confirming that the samples are actually flower-like structures with a size range of 1--10 μm. [Figure 2b](#f2){ref-type="fig"} shows an SEM image of one flower-like part, in which the entire 3D structure is built from a large number of nanosheets. [Figure 2c](#f2){ref-type="fig"} shows the ridge surface on a single nanosheet, in which the 'smooth' surface looks like a densely woven gauze. [Figure 2d](#f2){ref-type="fig"} shows a typical transmission electron microscope (TEM) image of one NiO microflower, again illustrating that the products have a 3D flower-like structure. [Figure 2e](#f2){ref-type="fig"} shows a selected area electron diffraction (SAED) pattern recorded on a single sheet. A set of well-defined dots indicates that the NiO films are single crystal in nature. [Figure 2f](#f2){ref-type="fig"} is a higher magnification TEM image showing that a large number of nanowires are closely "woven" together. [Figure 2g](#f2){ref-type="fig"} shows that the nanowires emanating from the nanosheet have a diameter of \~3 nm. [Figure 2h](#f2){ref-type="fig"} shows a TEM image of the inner surface of one nanosheet, in which the nanowires can be observed. The nanowire diameters can be measured based on the intensity profile along a straight line, as shown in the inset of [Fig. 2h](#f2){ref-type="fig"}, demonstrating that the nanowire diameter in the NiO nanosheet is around 3.0 nm. It should be noted, for the products obtained with 1.0 g SDS, the microflower morphology was somehow maintained and observed with thicker nanosheets ([Supplementary, Fig. S6](#S1){ref-type="supplementary-material"}). Other processing factors, such as reaction time and temperature, were also investigated ([Supplementary, Fig. S7, S8](#S1){ref-type="supplementary-material"}). A shorter time (1 h) seems to be insufficient to form a well-defined nanosheet structure ([Fig. S7a](#S1){ref-type="supplementary-material"}). The microflowers form with a reaction time of 5 hours ([Fig. S7b](#S1){ref-type="supplementary-material"}) or 10 hours ([Fig. S7c](#S1){ref-type="supplementary-material"}), while some nanosheets peeled off from the hierarchical structures for the reaction time of 20 hours ([Fig. S7d](#S1){ref-type="supplementary-material"}). The optimum temperature for forming microflowers was 150 °C. At a temperature of 120 °C, a microsphere structure with nanowires or small nanosheet building blocks was obtained ([Fig. S8a](#S1){ref-type="supplementary-material"}), while only nanoribbon structures were formed at 180 °C ([Fig. S8b](#S1){ref-type="supplementary-material"}). Interconnected Ni-network modified NiO microflowers --------------------------------------------------- While the open hierarchical texture in NiO microflowers facilitates access of ions to the electrode/electrolyte interface (for supercapacitor applications), the poor electrical conductivity may severely retard or even prevent electron transport, leading to a limited use of active material and a significantly lowered specific capacitance. In an attempt to improve electrical conductivity, the as-prepared NiO microflowers were partially reduced to Ni by a controlled heat-treatment under an H~2~/Ar atmosphere (see Methods for details). [Figure 3a](#f3){ref-type="fig"} shows powder X-ray diffraction (XRD) patterns of the as-produced NiO microflowers and the corresponding samples (Ni-NiO) obtained with reduction times of 2, 5, and 10 minutes. All the samples show three peaks at 2θ = 37.2°, 43.2°, and 62.6°, which are indexed as (111), (200), and (220) peaks from the cubic structured NiO (JCPDS No. 47-1049). The samples subjected to H~2~ treatment have peaks at 44.7° and 51.9°, which we index as the (111) and (200) peaks of face-centered-cubic (fcc) Ni. These two Ni peaks increase in intensity relative to the three NiO peaks with increased time of exposure to H~2~(g). As confirmed by SAED pattern, the as-prepared NiO microflower has a single crystal structure. After partial reduction, the single crystal structure was destroyed with the formation of metallic Ni nanoparticles in Ni-NiO-2min, leading to the decrease of NiO XRD peaks in intensity because of the destruction of single crystalline structure of the NiO, and broadening of XRD peaks in width due to the appearance of metallic Ni nanoparticles. With further reaction, more NiO is converted to metallic Ni; consequently the Ni-NiO-5 min shows a remarkable increase in metallic Ni peaks. The Ni-NiO-10 min shows strong metallic Ni peaks and weak NiO peaks, suggesting that most of NiO is converted to metallic Ni. X-ray photoelectron spectroscopy (XPS) was used to study the chemical states of bonded elements in the Ni-NiO-5 min sample (reduction time of 5 min). The survey XPS spectra indicate C1s (285.1 eV), O1s (531.6 eV), and Ni2p peaks ([Fig. 3b](#f3){ref-type="fig"}). The O 1s and C 1s peaks are attributed to the O in NiO and the substrate (carbon tape), respectively. The Ni2p3/2 curve-fitted data for both Ni-NiO-5 min and pristine NiO samples ([Supplementary, Fig. S9a and 9c](#S1){ref-type="supplementary-material"}) indicate the presence of Ni(II) with a main peak at 855.7 eV and a satellite peak at 861.6 eV. A shoulder peak at 852.8 eV is observed for Ni-NiO-5 min, suggesting the presence of metallic nickel. Moreover, the O 1s spectra for both samples were studied ([Supplementary, Fig. S9b and 9d](#S1){ref-type="supplementary-material"}). The fitted peak with a binding energy at 530.8 eV is assigned to NiO and the peak at 532.7 eV is assigned to hydrous species. Obviously, the Ni-NiO-5 min, because of the partial reduction of its surface, is much less hydrated than the pristine NiO. According to a controlled TGA experiment, the content of Ni in Ni-NiO-5 min samples is estimated as 18.4 wt.% ([Supplementary, Fig. S10](#S1){ref-type="supplementary-material"}). The Ni-NiO-5 min sample shows a similar nitrogen adsorption-desorption curve to the pristine NiO ([Fig. 3c](#f3){ref-type="fig"}); the Ni-NiO-5 min sample has a BET surface area of 151.6 m^2^ g^−1^ and the pristine NiO microflower has a value of 159.6 m^2^ g^−1^. Both samples have an average pore size of around 4.0 nm ([Supplementary, Fig. S11](#S1){ref-type="supplementary-material"}). The surfaces and overall structure remain almost unchanged after the hydrogen treatment. SEM and TEM were performed to investigate the morphology and structure of the samples subjected to H~2~ treatment. The **Ni-NiO-2min** maintained the microflower morphology and showed a polycrystalline structure with lots of metallic Ni nanoparticles decorating the surface ([Supplementary, Fig. S12a-c](#S1){ref-type="supplementary-material"}). A typical SEM image of the Ni-NiO-5 min sample ([Fig. 3d](#f3){ref-type="fig"}) shows that the microflower morphology indeed remains essentially the same after partial reduction with H~2~(g). [Figure 3e](#f3){ref-type="fig"} shows the SEM image of a single microflower that has many pores especially at the ridges of the nanosheets; these are due to reduction of the NiO nanosheet surfaces. [Figure 3f](#f3){ref-type="fig"} shows the TEM image of a "petal" in a microflower from the Ni-NiO-5 min sample, and one sees a large number of pores on the nanosheet surface. Two types of polycrystalline rings can be observed based on the SAED analysis (inset of [Fig. 3f](#f3){ref-type="fig"}) and fit for cubic NiO and fcc Ni, consistent with the XRD results. [Figure 3g](#f3){ref-type="fig"} shows a magnified TEM image of the Ni-NiO-5 min sample, in which the 'woven nanowires' with a diameter of \~3 nm can be clearly observed. [Figures 3h,i](#f3){ref-type="fig"} show high-resolution TEM image from different nanosheets in the Ni-NiO-5 min microflower. Many shallow pores with irregular shape are present on the sheet surfaces, and these concave pores are surrounded by an interconnected network that was produced after the partial reduction of the nanosheet surface. The magnified HRTEM images were further studied ([Figs 3j--k](#f3){ref-type="fig"}, and [Supplementary Fig. S13a-f](#S1){ref-type="supplementary-material"}), revealing that the shallow pores and the network actually belong to two different crystalline structures. One is metallic Ni with a lattice spacing of 0.19 nm, which most commonly appears in the interconnected networks; the other is NiO with a lattice spacing of 0.24 nm that is present on the concave pores of the nanosheets. The **Ni-NiO-10** **min** also basically retained the microflower morphology with a polycrystalline structure and possessed a large amount of pores in the nanosheet surface ([Supplementary, Fig. S12d-f](#S1){ref-type="supplementary-material"}), which can be attributed to the fact that most NiO was converted to Ni. Supercapacitor performance -------------------------- A three-electrode system was used to evaluate the electrochemical properties of the NiO microflower material and the series of Ni-NiO materials. [Figure 4a](#f4){ref-type="fig"} shows cyclic voltammogram (CV) curves of different electrodes in a 1.0 M KOH solution at a scan rate of 10 mV s^−1^. A pair of peaks is observed in the range of 0.35--0.55 V for all curves arising from the following redox reaction of NiO: The Ni-NiO-2 min sample had a three times higher peak current density compared with the pristine NiO, and the Ni-NiO-5 min sample had an order of magnitude higher peak current density than the pristine NiO. Because most of the NiO was converted into Ni, the Ni-NiO-10 min sample showed a much smaller peak current density than the pristine NiO. The capacitance was calculated by integrating the area of the CV curve. The pristine NiO microflower material had a specific capacitance of 100.5 F g^−1^, a much smaller value than the theoretical value of NiO (2,584 F g^−1^). The Ni-NiO-2 min sample gravimetric capacitance was 388.9 F g^−1^, the Ni-NiO-5 min sample 964.9 F g^−1^, and the Ni-NiO-10 min sample 45.6 F g^−1^. [Figure 4b](#f4){ref-type="fig"} shows a series of CV curves at various scan rates for the Ni-NiO-5 min sample. At a scan rate of 1 mV s^−1^, the specific capacitance was 1,579 F g^−1^, while that of the pristine NiO was 149.1 F g^−1^ and that of Ni-NiO-2 min was 754.4 F g^−1^. The Ni-NiO-5 min sample had the highest capacitance among the different samples at the same scan rate ([Supplementary, Fig. S14](#S1){ref-type="supplementary-material"}), which delivered a capacitance of 515.4 F g^−1^ at a scan rate of up to 100 mV s^−1^, while the pristine NiO and the Ni-NiO-2 min samples had lower values of 55.3 F g^−1^ and 225 1 F g^−1^, respectively. The specific capacitance was also calculated based on the galvanostatic charge/discharge curves within a potential window of 0.1 \~ 0.67 V (*vs.* NHE) ([Supplementary, Fig. S15](#S1){ref-type="supplementary-material"}); the Ni-NiO-5 min sample had a significantly improved capacitance compared with the other three NiO-based samples at the same current density, as reflected in [Fig. 4c](#f4){ref-type="fig"}. CV and galvanostatic charge-discharge tests were further performed with a symmetrical two-electrode cell, which could provide reliable measurements for evaluating supercapacitors' performance[@b44]. It should be noted that for a two-electrode cell, we did not observe the pair of charge-discharge plateau in the range of 0 \~ 05 V, which should be present in the NiO system except for a charge plateau in the first charge process, suggesting that only oxidation reactions occur in the first scan. Therefore, the electrochemical tests were conducted within a −0.5 to 0.5 V voltage window. [Figure 4d](#f4){ref-type="fig"} shows rate-dependent CV curves of the supercapacitor with Ni-NiO-5 min at various scan rates from 1.0 to 500 mV/s. A pair of redox peaks between −0.1 and 0.1 V can be seen, which correspond to redox conversion reactions between NiO and NiOOH. The specific capacitance for the Ni-NiO-5 min is substantially larger than that based on pristine NiO at the same scan rates ([Supplementary, Fig. S16a](#S1){ref-type="supplementary-material"}). The improved electrochemical performance was confirmed by galvanostatic charge-discharge measurements performed at different current densities in a two-electrode cell. [Figure 4e](#f4){ref-type="fig"} shows the galvanostatic charge/discharge curves of a Ni-NiO-5 min symmetrical supercapacitor at different current densities. Two plateaus corresponding to redox reactions of NiO can be identified in the curves, indicating that the electrode material has a good pseudocapacitive characteristic and excellent electrochemical reversibility. The two-electrode cell based on Ni-NiO-5 min delivered specific capacitances of 1,828, 1,460, 1,266, 999, 848, 696, and 540 F g^−1^ at current densities of 0.5, 1.0, 2.0, 5.0, 10.0, 20.0, and 50.0 A g^−1^, respectively. Further analyses show that the Ni-NiO-5 min has a significantly higher capacitance and better rate capability than a two-electrode supercapacitor based on pristine NiO ([Supplementary, Fig. S16b, c](#S1){ref-type="supplementary-material"}). [Figure 4f](#f4){ref-type="fig"} shows the relation between the current density and the energy density as well as the power density. The Ni-NiO-5min sample had the following characteristics: (i) An energy density as high as 15.9 Wh kg^−1^ at 0.5 A g^−1^, corresponding to 1,828 F g^−1^ and a power density of 0.12 kW kg^−1^; (ii) For 20 A g^−1^, an energy density of 6.0 Wh kg^−1^ and a power density of 5.1 kW kg^−1^; (iii) At 50 A g^−1^, an energy density of 3.1 Wh kg^−1^ and a power density of 13.2 kW kg^−1^. The Ni-NiO-5 min sample, with a bulk-density of 0.74 g cm^−3^, had a volumetric specific capacitance of 1,352.7 F cm^−3^ at 0.5 A g^−1^. The electrochemical stability was evaluated using the galvanostatic charge-discharge method on the Ni-NiO-5 min-based 2-electrode symmetrical supercapacitor at a current density of 5.0 A g^−1^ ([Fig. 4g](#f4){ref-type="fig"}). It retained more than 80% of its initial capacitance after 10,000 cycles, demonstrating that the Ni-NiO-5 min sample has reasonable electrochemical stability and cycle reversibility. In order to further confirm the influence of the reduction treatment on the supercapacitor performance, electrochemical impedance spectroscopy (EIS) was performed on the series of NiO-based supercapacitors. The results are presented in terms of Nyquist plots in [Fig. 4h](#f4){ref-type="fig"}. The equivalent series resistance (ESR), which is a measure of conductivity related to the resistance of an electrode material, can be calculated from the intercept of the corresponding Nyquist plots on the Z real-axis. All supercapacitors exhibit an ESR value lower than 1.5 Ω, suggesting that the nickel foam provides an excellent current collector for active materials. The order of the intercept on Z real-axis for the four supercapacitors is: Ni-NiO-10 min (0.53 Ω) \< Ni-NiO-5 min (0.56 Ω) \< Ni-NiO-2 min (0.78 Ω) \< pristine NiO (1.03 Ω). The ESR value for these samples are consistent with the reduction level, indicating that the H~2~ treatment did contribute to reduce the ESR of NiO microflowers. In addition, in the low frequency region, the **Ni-NiO-5** **min** sample exhibits a near-vertical Nyquist line, while the pristine NiO and Ni-NiO-10 min shows more inclined Nyquist lines. The more vertical curve in a cell suggests easier diffusion of ions and more ideal capacitive behavior. Discussion ========== Although hierarchical 3D NiO microstructures were reported previously[@b37][@b38][@b39][@b40][@b41][@b42][@b43], an understanding of the formation of such unique microstructures is still lacking. Both SEM and TEM analysis prove that the NiO microflowers are actually composed of highly ordered nanowires. SEM images of typical products obtained under different conditions together demonstrate a clear and intuitive vision of the evolution process of the microflowers, as shown in [Fig. 5](#f5){ref-type="fig"}. The morphology changes from a microsphere built of nanowires, to a porous nanowire microsphere framework, and finally to microflowers constructed from nanosheets. These results provide strong evidence that the nanosheet building block in a Ni-based microflower has evolved from a nanowire "weaving" process. Apparently, urea and SDS are two key ingredients in the formation of microflowers constructed from NiO nanosheets. The hydrolysis of urea under hydrothermal conditions generates ions (e.g., OH^−^, CO~3~^2−^) that are indispensable for precipitating Ni^2+^. In addition, the concentration of urea can significantly affect the morphology of the Ni-based products, indicating that urea also functions as a structure-directing agent to confine the growth of microspheres from nanoplates to nanowires. On the other hand, SDS acts as a "binder" for "weaving" these nanowires into large single-crystalline sheets that finally build a microflower. It is well-known that urea has two −NH~2~ groups joined by a carbonyl (C = O) functional group that tends to form a dense and energetically favorable hydrogen-bond network, which is probably critical to the formation of a Ni-based microsphere constructed from nanowires. Additionally, by virtue of its tendency to form a porous framework, urea also has the ability to trap anionic surfactants of SDS compounds and hold them in the channels formed by interpenetrating helices composed of hydrogen-bonded urea molecules[@b45][@b46]. The hydrophilic group of SDS then binds with Ni^2+^ or Ni-based complexes through static interactions and further induces the ultrahigh density parallel assembly of Ni-based nanowires[@b13][@b47] to subsequently "weave" them into nanosheets within the porous microsphere frameworks and finally evolve into a microflower structure ([Supplementary, Fig. S17](#S1){ref-type="supplementary-material"}). It should be noted that the reduction reaction of highly "woven" NiO nanowires results in the formation of a conductive network of metallic Ni, which is most likely due to rearrangement of Ni nanoparticles on the surfaces of the NiO. It is known that the charge storage reactions in NiO-based supercapacitors are highly dependent on the electrode surface layer. Microflowers constructed from thin-layer nanosheets offer an ideal nanostructure to maximize the exposed active surface area at electrode/electrolyte interfaces, which makes it possible to make full use of the NiO. However, a large fraction of "dead surface" inevitably exists due to the poor electrical conductivity of NiO[@b25][@b26][@b27][@b28][@b29], as demonstrated by the low measured capacitance in pristine NiO microflowers. We also confirmed that partial reduction of other metal oxides contributed to an improved utilization efficiency of active materials ([Supplementary, Figs S18 and S19](#S1){ref-type="supplementary-material"}). In this work, partial reduction of NiO (Ni-NiO-5 min) led to the generation of an interconnected Ni network on its surface, which is completely different from the Ni-NiO nanocomposites[@b27][@b28][@b29][@b30] reported previously that are normally core-shell structures and do not have maximum access to electrolytes (by the active NiO material). First of all, the interconnected Ni network provides a network channel for electron transfer on the nanosheet surface and thus significantly improves the electrical conductivity. In addition, such an open hierarchical microflower enables both sides of every nanosheet to easily access ions/electrolyte for the redox reaction between NiO and NiOOH. And as a result, the Ni-NiO-5 min electrodes manifested a higher utilization efficiency, as evidenced by negligible decreases in specific capacitance values with increasing mass loadings ([Supplementary, Fig. S20](#S1){ref-type="supplementary-material"}). Moreover, the microflower structure can be beneficial to prevent agglomeration and to maintain the structural integrity because microflowers can accommodate the mechanical stress resulting from the volume change upon cycling. Therefore, the near-optimum Ni-NiO-5 min sample had a capacitance of around 1,828 F g^−1^ at a current density of 0.5 A g^−1^, with respectable power density and energy density. Notably, the capacitance of Ni-NiO-5 min is calculated as 2,240 F g^−1^ based on the mass of NiO, which almost approaches its theoretical capacitance (2,584 F g^−1^). In fact, the as-developed Ni-NiO-5 min outperforms nearly all NiO nanostructures reported previously in terms of various supercapacitor performance indices ([Supplementary, Table S1](#S1){ref-type="supplementary-material"}). It should be pointed out that both the bare carbon cloth and bare nickel foam current collectors delivered a capacitance less than 1.0 F g^−1^ ([Supplementary, Fig. S21 and S22](#S1){ref-type="supplementary-material"}), confirming that the high capacitance originates from the Ni-NiO-5 min sample itself instead of the Ni-foam current collector. Furthermore, based on results of the two-electrode planar symmetrical supercapacitor (coin cell CR2032), the Ni-NiO-5 min can deliver a capacity of around 70 mAh g^−1^ at a current density of 0.5 A g^−1^ and 19 mAh g^−1^ even at a high current density of 50 mA g^−1^ ([Supplementary, Fig. S23](#S1){ref-type="supplementary-material"}), and holds an energy density approaching those of conventional batteries without sacrificing the high power that a supercapacitor typically offers. To conclude, based on many experiments a 'weaving' mechanism is proposed for the first time to illustrate the formation of the hierarchical NiO structures with multiple characteristic dimensions: the 3D microflower was constructed from 2D nanosheets derived from a weaving process of 1D nanowires. An interconnected conductive Ni network on the surface of the nanosheet 'building blocks' of the microflower was produced through partial reduction of the NiO, rendering the material more electrically conductive and significantly improving the supercapacitor performance. Our approach may pave the way for understanding the formation mechanism of such 'microflower' nanostructures and further advancing other metal oxide-based electrode materials for supercapacitors with outstanding performance. Methods ======= Materials and Synthesis ----------------------- All chemicals were analytical grade and were purchased from Sigma-Aldrich and used without further treatment. The NiO microflower sample was prepared using a hydrothermal method followed by calcination. In a typical experiment, 0.6 g of sodium dodecyl sulfate (SDS) was dissolved in 60 ml water at room temperature to form a homogeneous and clear solution. 5.0 mmol urea was then added to the solution under drastic stirring. Afterwards, 5.0 mmol nickel chloride (NiCl~2~) was slowly added to this solution with vigorous agitation. The solution was then transferred to a 100 ml Teflon-lined stainless steel autoclave and heated at 150 °C for 5 h. The obtained green precipitates were filtered and washed six times with distilled water and absolute alcohol to remove all soluble materials. After drying the products were calcined at 500 °C for two hours. The NiO products were initially heated to 350 °C at a rate of 5 °C/min in argon and maintained at 350 °C with a flow of H~2~/Ar (1:9) gas for 5 min. The partially reduced product was automatically cooled under an Ar atmosphere. Major chemical reactions involved in the synthesis include: Characterization ---------------- The structures and morphologies of the as-prepared samples were obtained using a Hitachi H-800 TEM and a LEO 1530 SEM. XRD was performed on a Bruker D8-Advance X-ray powder diffractometer. Specific surface areas were measured at 77 K by BET nitrogen adsorption-desorption (Shimadzu, Micromeritics ASAP 2010 Instrument). To investigate the decomposition process of NiCO~3~∙Ni(OH)~2~ ∙ xH~2~O, the as-obtained hydrothermal samples were analyzed by thermogravimetric analysis (TGA, Mettler Toledo TGA-SDTA851 analyzer) from 25 to 500 °C under an air atmosphere with a heating rate of 5 °C/min. To analyze the content of metallic Ni in Ni-NiO, TGA analysis was carried out on the Ni-NiO-5 min samples by first heating them to 300 °C with a heating rate of 5 °C/min and maintaining them at 300 °C for 30 minutes under Ar protection, which can eliminate the effect of adsorbed water; the samples were then heated to 450 °C in air to investigate the oxidation reaction of metallic Ni. X-ray photoelectron spectroscopy (XPS) was carried out on VG ESCA 2000 with Mg K*α* as the X-ray source and the C1s peak at 284.6 eV as an internal standard. Electrochemical Measurements ---------------------------- Electrochemical measurements were conducted using a three-electrode system with a saturated calomel electrode (SCE) and Pt gauze as the reference and the counter electrodes, respectively. Working electrodes were prepared by mixing NiO samples, carbon black (5 wt.%), and poly-vinylidene fluoride (PVDF, 5 wt.%) in N-Methylpyrrolidone (NMP) to form a slurry, which was then pasted onto carbon cloth and dried at a temperature of 80 °C in a vacuum oven. Cyclic voltammetry was conducted using a CHI 760 electrochemical station in 1.0 M KOH aqueous electrolytes with a voltage range of 0.1 \~ 0.67 V at various scan rates. Electrochemical impedance spectroscopy (EIS) was performed by applying an AC voltage with an amplitude of 5 mV within the frequency range from 100 kHz to 50 mHz on the series of NiO-based two-electrode supercapacitors. Galvanostatic charge-discharge tests were also conducted in a CHI 760 electrochemical station using the same system as that used for the above CV testing. The specific capacitance (C~m~, F g^−1^) was calculated based on CV (Eq. [1](#eq13){ref-type="disp-formula"}) and galvanostatic charge-discharge tests (Eq. 2): where I is the current (A), V is the voltage, v is the scan rate (V s^−1^), and m is the mass (g) of the active material in the electrode. ΔV is the potential range for the electrochemical process and Δt is the discharge time. A stainless-steel coin cell (CR2032) supercapacitor with two symmetrical electrodes was assembled in air. To prepare the working electrode, a paste mixture consisting of active materials (Ni-NiO, 90%), acetylene black (5%), and PVDF (5%) was compressed onto a nickel foam current collector (Φ = 0.5 inches). After drying, the pellets with an active material weight of around 0.8 mg were assembled into a symmetrical supercapacitor, in which 1.0 M KOH solution was used as the electrolyte and a hydrophilic Millipore PVDF membrane was used as the separator. Cyclic voltammetry and galvanostatic charge-discharge tests were carried out using CHI 760 electrochemical station and a Land battery tester system. The gravimetric capacitance (F g^−1^) based on the CV curve was calculated as: where *I* is the current recorded (unit: A), *dV*/*dt* is the scan rate mV s^−1^, and m is the average mass (g) of the active material in each electrode. With the galvanostatic charge/discharge plots based on two-electrode cells, the specific capacitance (*C*~m~, unit: F g^−1^) of a single electrode, the energy density (*D*~*e*~, unit: Wh kg^−1^), and the power density (*D*~*p*~, unit: kW kg^−1^) of the supercapacitor were calculated based on Eqs [4](#eq10){ref-type="disp-formula"}, [5](#eq11){ref-type="disp-formula"}, [6](#eq12){ref-type="disp-formula"}: where *I* is the constant discharge current (unit: A), ∆*t* is *t*he discharge time (unit: s), and ∆*V* is the voltage window (0.5 V), and m is the average total mass (g) of the active material in both electrodes. The density of our NiO-based microflower products is ∼0.74 g cm^−3^, and the volumetric capacitance was calculated as: where D is the bulk density calculated as the average of the mass of many particles of the material divided by the total volume they occupy. Additional Information ====================== **How to cite this article**: Ci, S. *et al.* NiO-Microflower Formed by Nanowire-weaving Nanosheets with Interconnected Ni-network Decoration as Supercapacitor Electrode. *Sci. Rep.* **5**, 11919; doi: 10.1038/srep11919 (2015). Supplementary Material {#S1} ====================== ###### Supplementary Information This work was financially supported by the U.S. Department of Energy (DE-EE0003208). We thank Prof. Peter A. Thrower (Emeritus Editor-in-Chief, CARBON) for his kind help with manuscript proofreading. The authors declare no competing financial interests. **Author Contributions** Z.W. and J.C. conceived the idea. Z.W. and Y.Q. performed the material fabrication experiments. Z.W. and S.Q.C. conducted the electrochemical tests. S.H.M. and S.M.C. performed SEM and TEM characterizations. Z.W. and J.C. co-wrote the paper. All authors discussed the results and commented on the manuscript. {#f1} {#f2} {#f3} {#f4} {#f5}

Introduction {#S5} ============ Crohn\'s disease (CD), one of the two types of inflammatory bowel disease (IBD), is characterized by inappropriate and continuous activation of the mucosal immune system \[[@R1],[@R2]\]. In the normal small intestine, Paneth cells, an important part of the innate mucosal immune system, maintain microbial homeostasis by secreting defensins, a class of cationic peptides with a broad spectrum of antimicrobial activity \[[@R3]\]. The two broad categories of defensins include α and β-defensins. In the small intestine, α-defensin 5 (DEFA5) is secreted into the lumen of intestinal crypts by Paneth cells after activation of the NOD2 receptor by bacterial endotoxins \[[@R4]-[@R6]\]. The α-defensins are chemotactic for cells of both innate and adaptive immune systems, including macrophages and T cells \[[@R7]\]. They are involved in several processes that maintain intestinal homeostasis, including regulation of gut flora, intestinal inflammation, stem cell protection, and crypt development \[[@R8]-[@R10]\]. In genetically susceptible individuals, changes in intestinal bacterial flora, including commensal bacteria, play a role in IBD pathogenesis. These changes contribute to the initiation and perpetuation of chronic mucosal inflammation \[[@R11]-[@R14]\]. Since defensins maintain the balance between commensal microbes and intestinal mucosa, their dysregulation could contribute to IBD pathogenesis. Defensin expression has been shown to be altered in the setting of NOD2 mutations and changes in certain signaling pathways, but no research has demonstrated the role of microRNAs in defensin regulation. However, SNPs within the 3′ UTR of DEFA5, have been linked with increased susceptibility to IBD, suggesting the possible role of miRNA regulation \[[@R15]\]. Additionally, post-transcriptional regulation of gene expression, such as that by microRNAs, has been shown to play an important role in IBD pathogenesis \[[@R16]-[@R18]\]. Altered miRNA expression profiles exist in active IBD indicating that changes in miRNA expression are associated with disease \[[@R17]-[@R21]\]. Therefore, the purpose of this study is to examine the role of miRNAs in regulating DEFA5 expression using Caco-2 cells as a model of Paneth cells. Materials and Methods {#S6} ===================== Cell lines and cell culture {#S7} --------------------------- Human colonic epithelial cell lines (Caco-2, HT29, and HCT116 cells) were previously obtained and used for this study. The Caco-2 cells were cultured in Dulbecco\'s Modified Eagle Medium (DMEM) (Cellgro, Manassas, VA), while the HT29 and HCT116 cells were cultured in McCoy\'s 5A Medium (Cellgro, Manassas, VA). The culture medium was supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin. All cell lines were cultured at 37°C in a humidified atmosphere containing 5% CO~2~ and the medium was changed every two days. FGF-2 (10 ng/ml, R&D Systems, Minneapolis, MN) was solubilized in culture medium containing heparin sodium salt (1.25 ug/ml; Sigma, St. Louis, MO). FGF-2 was added daily beginning at 24 hours post-plating and until cells were harvested for assay, as described previously \[[@R22]\]. Caco-2 cells have previously been used as a model of Paneth cells, and differentiation into a Paneth cell phenotype has been confirmed with Paneth cell markers, including expression of alpha-defensin 5, alpha-defensin 6, lysozyme, and sPLA2 in previous studies \[[@R22]-[@R24]\]. For experiments in which Caco-2 cells were stimulated with TNFα, Caco-2 cells were first treated with FGF-2 for 72 hours, followed by addition of medium alone (control) or medium containing TNFα (R&D Systems, Minneapolis, MN) at 50 ng/mL. Cells were harvested at four hours for mRNA and 8 hours for protein analyses. Prior to cell harvesting for Western blot, Caco-2 cells were treated with Brefeldin-A (Sigma) at 5 ug/ml for a total of 8 hours (replenished every four hours) to prevent secretion of DEFA5 into the cellular medium. Total RNA {#S8} --------- Total RNA was isolated using TRIzol reagent (Life Technologies, Grand Island, NY) according to the manufacturer\'s protocol. RNA concentrations were determined using a Nano-Drop 1000 spectrophotometer (Thermo Fisher Scientific, Wilmington, DE). The isolated RNA was stored at -80°C until use in Quantitative Reverse Transcription-Polymerase Chain Reaction (qRT-PCR). Quantitative reverse transcription-polymerase chain reaction for mRNA {#S9} --------------------------------------------------------------------- One microgram of total RNA was converted to complementary DNA (cDNA) using the qScript cDNA SuperMix (Quanta Biosciences, Gaithersburg, MD). qRT-PCR was performed using SYBR Green Power PCR Master Mix (Applied Biosystems, Foster City, CA). The qRT-PCR amplifications were performed on the LightCycler 480 realtime PCR system (Roche, Indianapolis, IN). The cycles passing threshold (Ct) were recorded and the expression of GAPDH was used as an internal control. The primers for DEFA5 were 5′-ACCCAGAAGCAGTCTGGGGAAGA-3′ (forward) and 5′ -- GGTGGCTCTTGCCTGAGAACCTGA-3′ (reverse). The primers for GAPDH were 5′ -- CGACCACTTTGTCAAGCTCA-3′ (forward) and 5′-AGGGGAGATTCAGTGTGGTG-3′ (reverse). Quantitative reverse transcription-polymerase chain reaction for miRNA {#S10} ---------------------------------------------------------------------- One microgram of total RNA was converted to cDNA using the NCode VILO miRNA cDNA Synthesis Kit (Life Technologies)followed by qPCR using the NCode EXPRESS SYBR GreenER miRNA qRT-PCR kit (Life Technologies). The cycles passing threshold were recorded and the expression of miRNAs was calculated relative to U6B, a ubiquitously expressed small nuclear RNA. The forward primer for miR-124 was 5′- TAAGGCACGCGGTGAATGCC-3′. The forward primer for miR-924 was 5′-CCTCTGCCCTCTAAAGGTTTGC-3′. The forward primer for U6B was 5′-CGCAAGGATGACACGCAAATTCG-3′. The reverse primer was the NCode miRNA universal qPCR primer (Invitrogen). Data are presented as target miRNA or mRNA expression=2ΔCt, with ΔCt=(U6B or GAPDH Ct -- target miRNA or mRNA Ct). qRT-PCR was carried out in triplicate for each sample for both the U6B or GAPDH control and each miRNA or mRNA. Western blot {#S11} ------------ Cells were lysed in cold RIPA buffer (Thermo Fisher Scientific, Rockford, IL) supplemented with 1% protease and phosphatase inhibitor cocktail (Thermo Fisher Scientific). Protein concentrations were determined with the BCA protein assay (Thermo Fisher Scientific). Cell lysates were suspended in × Laemmli sample buffer (Bio-Rad, Hercules, CA) containing 5% 2-mercaptoethanol and boiled for 5 minutes. After heat denaturation, total protein lysates (30 ug/lane) were subjected to tricine-SDS-PAGE \[[@R25]\]. The proteins were then transferred electrophoretically to 0.2 um PVDF membranes. Membranes were blocked in blocking buffer (LI-COR Biosciences, Lincoln, NE) diluted 1:1 in × PBS for 1 hour at room temperature. The blots were incubated with mouse anti-DEFA5 (Santa Cruz Biotechnologies, Santa Cruz, CA; 1:200 dilution), goat anti-actin (Sigma; 1:1000 dilution), and mouse anti-GAPDH (Life Technologies; 1:1000 dilution) overnight at 4°C. After washing with PBST, blots were incubated with Alexa Fluor 680 (Life Technologies) and IRDye 800CW (LI-COR Biosciences) conjugated secondary antibodies (1:10,000 dilution) for 1 hour at room temperature. The band intensities were quantified using an Odyssey infrared imaging system (LI-COR Biosciences) and analyzed using Image Studio Lite 3.1 (LI-COR Biosciences). DEFA5 3′ UTR construct {#S12} ---------------------- The full-length 3′ UTR of DEFA5 (nucleotides 328-451 of NM_021010) was cloned into the PmeI and SacI sites downstream of the dual firefly and Renilla luciferase reporter vector, pmirGLO construct (Promega, Madison, WI) by GenScript. The full length 3′ UTR was also used as a template to mutate the entire seed region of the putative miR-124 and miR-924 binding sites. miRNA mimic and luciferase construct transfection {#S13} ------------------------------------------------- Synthesized RNA duplexes of miRNA mimics (agomiRs) to miR-124 and miR-924 and the negative control were obtained from Sigma (St. Louis, MO). A miRNA mimic or a luciferase construct was transfected into Caco-2 cells using Lipofectamine 2000 reagent (Life Technologies) according to the manufacturer\'s guidelines. At 24 hours post-transfection, cells were harvested for RNA and protein analyses or were harvested for measurement of luciferase activities. Luciferase reporter assay {#S14} ------------------------- Caco-2 cells were lysed in passive lysis buffer and then analyzed for the firefly and Renilla luciferase activities using the Dual-Luciferase Reporter Assay System (Promega) on the GloMax-Multi Detection System Luminometer (Promega) according to the manufacturer\'s instructions. The firefly luciferase activity was normalized to the renilla luciferase activity. Statistical analysis {#S15} -------------------- All experiments were performed with four biological repeats in triplicate. R (R Development Core Team, Vienna, Austria) was used for statistical analysis \[[@R26]\]. Statistical significance was determined by 2-tailed Student\'s t tests (for comparison of two conditions) and one-way ANOVA for comparison of multiple groups. Data are presented as mean ± standard error of the mean. P\< 0.05 was considered significant. Results {#S16} ======= DEFA5 expression in colonic epithelial cells {#S17} -------------------------------------------- Alpha defensin expression and inducible expression by FGF-2 were ascertained in three different colonic epithelial cell lines: Caco-2, HCT116, and HT29. Both the Caco-2 and HCT116 cell lines expressed DEFA5, while the HT29 cell lines did not express DEFA5. DEFA5 expression was significantly increased in Caco-2 cells following treatment for 72 hours with FGF-2, as previously demonstrated by Brodrick et al. and Tan et al. ([Figure 1](#F1){ref-type="fig"}) \[[@R22]-[@R24]\]. Because Caco-2 cells expressed the α-defensins at a highly inducible level, all subsequent work was conducted in that cell line. Next, we confirmed that DEFA5 mRNA expression was inducible in Caco-2 cells by both FGF-2 and TNFα. Specifically, DEFA5 mRNA expression after treatment with FGF-2 (10 ng/ml) for 72 hours and FGF-2 for 72 hours plus TNFα (50 ng/ml) for 4 hours was measured ([Figure 2A](#F2){ref-type="fig"}). Caco-2 cells stimulated with FGF-2 resulted in a statistically significant 5.3 fold increase and those stimulated with FGF-2 followed by TNFα resulted in a statistically significant 17.5 fold increase ([Figure 2A](#F2){ref-type="fig"}). Treatment of Caco-2 cells with FGF-2 in combination with TNFα for 8 hours resulted in a statistically significant 1.8 fold increase in protein expression as measured by densitometry performed on Western blots ([Figure 2B](#F2){ref-type="fig"}). An example Western blot of Actin and DEFA5 is shown ([Figure 2C](#F2){ref-type="fig"}). Expressions of miRNAs that bind to DEFA5 3′ UTR in Caco-2 cells {#S18} --------------------------------------------------------------- An in silico analysis utilizing miRBase and TargetScan identified fifteen putative miRNA binding sites in the 3′ UTR of DEFA5. We identified the miRNAs with putative binding sites in the DEFA5 3′UTR and searched for miRNAs that were down-regulated in response to FGF-2 and TNFα. Of those miRNAs, we identified miR-124 and mi-924 ([Figure 3](#F3){ref-type="fig"}) \[[@R27]-[@R30]\]. The other thirteen miRNAs were not examined further due to low relative expression. Because we hypothesized that the FGF- and TNFα- induced increase in DEFA5 expression may be influenced by a corresponding down-regulation of regulatory miRNAs, we examined the expression of miR-124 and miR-924 in response to FGF-2 and TNFα ([Figure 4](#F4){ref-type="fig"}). DEFA5 induction was accompanied by a significant decrease in expression of miR-124 and miR-924 compared to unstimulated cells ([Figures 4A and 4B](#F4){ref-type="fig"}). Our data demonstrate an inverse correlation between DEFA5 expression and the expression of miR-124 and miR-924, suggesting that DEFA5 may be regulated by miRNAs. Additionally, these data raise the possibility that the effect of FGF-2 and TNFα may be at least in part mediated by a down-regulation of DEFA5-associated miRNAs. To determine the effect of endogenous miRNAs on DEFA5 expression, a luciferase reporter construct containing the wild-type DEFA5 3′UTR was transfected into unstimulated Caco-2 cells. Transfecting this construct resulted in a 76% decrease in relative luciferase activity compared to transfection of the empty vector, which contained no 3′ UTR ([Figure 5](#F5){ref-type="fig"}). These data indicate that the DEFA5 3′ UTR has functional miRNA binding sites to which endogenous miRNAs can bind and negatively regulate DEFA5 gene expression. Next, a luciferase reporter construct containing the DEFA5 3′ UTR with mutated binding sites of either miR-124 or miR-924 were transfected into Caco-2 cells. Transfection of the mismatched miR-124 construct resulted in a restoration to 66% of the relative luciferase activities of the empty vector ([Figure 5](#F5){ref-type="fig"}). Transfection of the mismatched miR-924 construct restored relative luciferase activities to 85% of the empty vector ([Figure 5](#F5){ref-type="fig"}). These data indicate that mutation of the binding sites led to a loss of regulation by each miRNA. This increase in relative luciferase activity compared to transfection of the empty vector was statistically significant in both miRNAs tested. We next tested whether miR-124 and miR-924 can negatively regulate basal DEFA5 expression. A miR-124 or miR-924 mimic was transiently transfected into Caco-2 cells and DEFA5 mRNA expression assessed ([Figure 6](#F6){ref-type="fig"}). The miR-124 and miR-924 mimics significantly decreased DEFA5 mRNA expression by 47% and 50%, respectively. In addition, transfection of the miR-124 mimic resulted in a 75% decrease in DEFA5 protein expression as demonstrated by densitometry and transfection of the miR-924 mimic resulted in a 60% decrease in DEFA5 protein expression. Overall, these data suggest that miRNAs negatively regulate the mRNA and protein expression of DEFA5. Discussion {#S19} ========== To the best of our knowledge, this is the first study to show miRNA regulation of DEFA5. We hypothesized that miRNAs regulate DEFA5 expression. We compared the expression of DEFA5 and miRNAs with putative binding sites in the 3′ UTR of DEFA5 in Caco-2 cells. We found that the expression of miR-124 and miR-924 was inversely correlated with that of DEFA5. Additionally, transfecting luciferase constructs containing the DEFA5 3′ UTR into Caco-2 cells resulted in decreased relative luciferase activity compared to transfecting empty vectors. Furthermore, creating a mismatch in the seed region of the miR-124 and miR-924 in the 3′UTR of DEFA5 resulted in restoration of luciferase activity. Finally, transfection of the miR-124 and miR-924 mimics significantly decreased DEFA5 mRNA expression and protein expression. MicroRNAs regulate many processes, including differentiation and activation of the cells of the immune system. They have been demonstrated to play a role in a number of autoimmune diseases, including systemic lupus erythematosus, rheumatoid arthritis, and IBD \[[@R31]-[@R34]\]. Overall, the role of miRNAs as key negative regulators of inflammation, innate immunity, and epithelial function is being increasingly recognized \[[@R35]\]. A number of changes in miRNA expression in IBD and inflammation have been described \[[@R16]-[@R18]\]. As a result, miRNA regulation as an important component of intestinal epithelial innate immunity in epithelial cells is logical \[[@R36],[@R37]\]. In a murine model, miR-146a regulated gut inflammation via the NOD2-sonic hedgehog (SHH) signaling pathway by suppressing SHH and ultimately resulting in increased pro-inflammatory cytokine expression \[[@R38]\]. Conversely, in HT29 cells, miR-122 decreased pro-inflammatory cytokines by downregulating LPS-induced NOD2 expression \[[@R39]\]. Specifically, we demonstrated that miRs-124 and -924 influence DEFA5 gene expression. Our data identifies an additional role for miR-124, which previously had been shown to be involved in IBD by regulation of STAT3 and regulation of CNS macrophages \[[@R40],[@R41]\]. In children with UC and mice with experimental colitis, miR-124 levels were significantly decreased while STAT3 and downstream genes were up-regulated \[[@R40]\]. The IL6/STAT3 pathway activation has been shown to play a role in colitis by promoting inappropriate survival of T cells \[[@R42]\]. In a study examining CNS inflammation in vivo and in an experimental model, miR-124 was shown to play a central role in regulating microglial and macrophage quiescence \[[@R41]\]. miR-124 inhibited macrophage activation by binding the C/EBPα transcription factor \[[@R41]\]. This study defines a novel role for miR-924, which has not been previously described to regulate any genes. A growing body of research demonstrates that defensins play a central role in IBD pathogenesis, in part related to changes in Wnt-signaling and mutations in NOD2 \[[@R43]-[@R45]\]. However, these mutations do not explain changes in defensin expression in a large proportion of patients, and these changes in defensin expression could perhaps be accounted for by dysregulation of miRNA regulation, as demonstrated in this study. These changes in miRNA expression may explain alterations in α-defensin expression in IBD. Crohn\'s disease of the terminal ileum is unique because of the large number of Paneth cells and the highest density of microbes in the distal small intestine, which is otherwise low in the normal proximal small intestine \[[@R46],[@R47]\]. In the ileal mucosa of CD, a decrease in antimicrobial activity could be attributed to decreased Paneth cell α-defensin expression \[[@R13]\]. The decreased levels of expression have been suggested to be predisposing factor for development of CD \[[@R13]\]. Additionally, these decreased levels in terminal ileum of CD patients with ileal disease involvement were noted to occur regardless of the presence of inflammation \[[@R14],[@R43],[@R48]\]. Levels of other Paneth cell products were unchanged or increased, indicating that the decrease in DEFA5 could be due to a defect in Paneth cell α-defensin regulation \[[@R14]\]. Alterations in miRNA regulation in IBD may explain defective α-defensin regulation in patients\' whose disease cannot be attributed to mutations in NOD2. One limitation of this study was the absence of an ideal experimental model. Colonic epithelial cell lines express α-defensin at relatively low levels. Because human Paneth cells do not survive under in vitro culture conditions, Caco-2 intestinal epithelial cells were selected since they share characteristics with small intestinal epithelial differentiation in vitro and constitutively express NOD2, much like Paneth cells \[[@R5],[@R49]\]. Caco-2 cells also express fibroblast growth factor receptor-3, which is a critical regulator of Paneth cell differentiation during gut development \[[@R22],[@R50]\]. Additionally, Caco-2 cells treated with FGF-2 provides the most suitable in vitro model available, as demonstrated in previous studies \[[@R22]-[@R24]\]. While the use of Caco-2 cells confirmed the regulation of DEFA5 by miRNAs, it may not exactly reflect small intestine physiology. Overall, this study demonstrates that miRNAs are an important negative regulator of DEFA5. Our data establishes miRNA regulation of defensins and raises the possibility of dysfunctional miRNA regulation contributing to reduced DEFA5 levels seen in CD. Further examination of miRNA regulation of defensins, especially in inflammatory states, may contribute to a better understanding of the pathogenesis and may lead to the development of new diagnostic and therapeutic strategies for IBD patients. We acknowledge Mark Musch PhD for assistance with tissue culture, and Edith Porter MD for assistance in measuring DEFA5 by Western blot. We acknowledge the following sources of support: the National Institute of Health (K08 DK078046 to J.H.K.), the NIH DDRCC core grant (P30 DK42086), and the NIH NIDDK training grant (T32 DK007074 to D.R.B.M.) {#F1} {#F2} {#F3} {#F4} {#F5} {#F6}

It is known that elderly individuals are at increased risk of cognitive decline after general anesthesia[@b1][@b2][@b3]. Anesthesia may also be one of the contributing factors for sporadic Alzheimer disease (AD), a devastating neurodegenerative disease characterized clinically by cognitive deficit and also the most common form of dementia in adults. Some epidemiological studies have shown that general anesthesia may increase the risk of elderly individuals for developing AD[@b4][@b5]. Increased atrophy of the brain, including the cortical gray matter and the hippocampus, together with reduced performance for cognitive tests are detectable 5 to 9 months after general anesthesia/surgery[@b6]. Evidence from animal studies suggests that anesthetic exposure can increase Aβ plaque formation and tau hyperphosphorylation[@b7][@b8][@b9][@b10][@b11], which are believed to cause neurodegeneration in AD. Anesthesia also causes learning and memory deficits at a later time in rodents[@b9][@b12][@b13]. Rats after exposure to isoflurane, a commonly used inhaled anesthetic, show learning impairment that persists for weeks[@b14]. Persistent memory impairment is also seen in aged rodents after exposure to isoflurane, nitrous oxide, or the combination of both[@b15][@b16]. The molecular mechanisms by which general anesthesia causes long-term cognitive impairment remain to be understood. The major known function of insulin is to regulate glucose metabolism in the periphery. The brain was previously thought to be insensitive to insulin. However, recent studies have well demonstrated that insulin actually has neurotrophic and neuroprotective activities, regulates neural development and plasticity, and plays an important role in learning and memory[@b17][@b18][@b19][@b20]. The important role of brain insulin signaling is further supported by clinical studies showing that administration of insulin into the central nervous system promotes memory and improves cognitive function in individuals with AD[@b21][@b22][@b23]. We recently found that general anesthesia disturbs brain insulin signaling and induces abnormal hyperphosphorylation of tau[@b24], which might contribute to the anesthesia-induced cognitive impairment. We therefore hypothesized that administration of insulin into the brain may prevent anesthesia-induced brain changes and cognitive impairment. To test this hypothesis, we treated 3xTg-AD mice, a commonly used triple transgenic mouse model of AD, with insulin via intranasal delivery for consecutive seven days before anesthesia with propofol. Intranasal delivery bypasses the blood-brain barrier and delivers drugs into the brain through several pathways, including olfactory- and trigeminal-associated extracellular pathways and the perivascular pathway[@b25]. We found that the pretreatment of 3xTg-AD mice with intranasal insulin promotes brain insulin signaling and attenuates propofol-induced hyperphosphorylation of tau[@b24]. However, whether the pretreatment with intranasal insulin can also prevent anesthesia-induced cognitive impairment had not been studied. The present study aimed to answer this important question. In our present study, we treated wild-type mice with daily intranasal administration of insulin or, as a control, saline for seven consecutive days before general anesthesia using a combination of propofol and sevoflurane. We found that intranasal insulin prevented anesthesia-induced spatial memory deficits. We also found that hyperphosphorylation of tau induced by anesthesia was temporary, which can be prevented with the insulin pretreatment. Results ======= Intranasal insulin prevents anesthesia-induced spatial memory deficit in mice ----------------------------------------------------------------------------- We recently found that intranasal administration of insulin (1.75 U/day) for seven consecutive days promotes brain insulin signaling and attenuates propofol-induced hyperphosphorylation of tau in adult 3xTg-AD mice[@b24]. To investigate if the intranasal insulin treatment can also prevent anesthesia-induced cognitive impairment, we treated wild-type mice with intranasal insulin daily for seven days before general anesthesia, followed by assessment of spatial learning and memory using Morris water maze beginning at the next day after anesthesia ([Fig. 1A](#f1){ref-type="fig"}). Anesthesia was induced by intraperitoneal injection of propofol (150 mg/kg) and maintained by 2.5% sevoflurane inhalation for 1.0 hr. The combination of these two anesthetics is commonly used in clinic for surgeries. We observed that all mice were able to learn the platform location during the training phase, as evidenced by the reduction in latency to locate the submerged platform from around 80 seconds in day 1 to 35--50 seconds in day 4 of the training ([Fig. 1B](#f1){ref-type="fig"}). However, the anesthesia-treated mice (red curve) took significantly more time (*p* \< 0.05) to locate the platform than the control mice (green curve), suggesting impaired learning in the anesthesia-treated mice. Importantly, we found that prior daily treatment with intranasal insulin for seven consecutive days prevented the anesthesia-induced learning impairment. The learning curve of the insulin-treated anesthesia group (orange curve) was significantly different from that of the untreated anesthesia group (red curve, *p* \< 0.05) and was indistinguishable from the control mice (green curve, *p* \> 0.05) ([Fig. 1B](#f1){ref-type="fig"}). The intranasal insulin treatment did not affect the spatial learning of mice not treated with anesthesia (blue curve vs. green curve, *p* \> 0.05). To assess the spatial memory of the mice, we carried out probe trial 24 hrs after the last training trial ([Fig. 1A](#f1){ref-type="fig"}). We found that the anesthesia group spent significant less time (50 ± 6% as compared to 63 ± 5% in controls, *p* \< 0.05) in the target quadrant ([Fig. 1C](#f1){ref-type="fig"}), took more time (39.5 ± 5.6 seconds as compared to 22.4 ± 2.7 seconds in controls, *p* \< 0.05) to reach the platform location target ([Fig. 1D](#f1){ref-type="fig"}), and crossed the platform location much less times (1.2 ± 0.3 times as compared to 3.3 ±0.5 times in controls, *p* \< 0.05) ([Fig. 1E](#f1){ref-type="fig"}) than the control mice. These results confirmed the spatial memory deficit in the anesthesia-treated mice. We found that prior daily treatment with intranasal insulin prevented the anesthesia-induced spatial memory deficit because no significant differences in these three parameters were found between the anesthesia- and insulin-treated group and the control group ([Fig. 1C--E](#f1){ref-type="fig"}). These findings suggest that intranasal insulin treatment can prevent anesthesia-induced spatial learning and memory in mice. We also determined the swim speed and found no differences among the four mouse groups ([Fig. 1F](#f1){ref-type="fig"}). Intranasal insulin attenuates anesthesia-induced hyperphosphorylation of tau in the mouse brain ----------------------------------------------------------------------------------------------- To understand the biological changes underlying the cognitive impairment induced by anesthesia and the preventive effect of insulin, we sacrificed the mice after Morris water maze test and analyzed tau phosphorylation in the brain because tau hyperphosphorylation is crucial to AD and other tauopathies[@b26]. In contrast to previous reports by us[@b8][@b24] and others[@b27][@b28], we found no detectable change in tau phosphorylation in the brains of anesthetic mice with or without prior treatment with insulin (data not shown). Because the animals were sacrificed within a few hours after anesthesia in all previous studies in which tau hyperphosphorylation was observed post anesthesia, we thus treated three additional groups of mice with intranasal insulin and/or anesthesia, as described above, and sacrificed the mice immediately after anesthesia for one hour and before they awoke from anesthesia. Analysis of brain tau phosphorylation of these mice showed that anesthesia with propofol/sevoflurane induced marked increase in tau phosphorylation at all phosphorylation sites studied ([Fig. 2](#f2){ref-type="fig"}), which are consistent with previous studies using different paradigms of anesthesia[@b27][@b28]. Up-shift of the apparent gel mobility of tau proteins, which is a well-established phenomenon of tau hyperphosphorylation, was also seen in the Western blots ([Fig. 2A](#f2){ref-type="fig"}). Excitingly, we found that prior treatment of the mice with intranasal insulin attenuated anesthesia-induced tau hyperphosphorylation at most of the sites studied, including Thr181, Thr205, Thr212, Thr231, Ser262/Ser356 (12E8 sites), Ser396, and Ser396/Ser404 (PHF-1 sites) ([Fig. 2B,C](#f2){ref-type="fig"}). The rostral halves of the forebrains were first analyzed because higher insulin concentration can be reached to this area than the caudal part of the brain through intranasal administration[@b29]. Because the hippocampus, which is the major brain area responsible for spatial learning and memory, is localized in the caudal half of the mouse forebrain, we also studied tau phosphorylation in the homogenates of the caudal halves of the mouse brains. We found marked increase in tau phosphorylation at all phosphorylation sites studied and an up gel-mobility shift in the caudal halves of the mouse forebrains immediately after anesthesia ([Fig. 2D](#f2){ref-type="fig"}), which is similar to the changes observed in the rostral halves of the mouse brains. Partial prevention of the anesthesia-induced increase in tau phosphorylation was also observed after the pre-treatment of mice with intranasal insulin, but the attenuating effect in the caudal forebrains was less remarkable than in the rostral forebrains. These regional differences are consistent to the higher drug concentration in the rostral forebrain than the caudal forebrain after intranasal administration[@b29]. The effects of anesthesia and intranasal insulin were clearly detectable in the hippocampal neurons through immunohistochemical staining with monoclonal antibodies 12E8 ([Fig. 2E](#f2){ref-type="fig"}) and PHF1 (data not shown), both of which are against phosphorylated tau. Intranasal insulin enhances the level of synaptic proteins in the mouse brain ----------------------------------------------------------------------------- Synapses are the structural basis of memory and cognition, and their alterations usually underlie functional changes of the brain. To learn whether anesthesia and intranasal insulin treatment alter synaptic activity, we determined the levels of major synaptic proteins in the mouse brains after the treatments using Western blots. These synaptic marker proteins include the presynaptic proteins synaptophysin (Syp) and synapsin-1, the postsynaptic marker postsynaptic density 95 (PSD95), the AMPA receptor subunit GluR1. We found that anesthesia with propofol and sevoflurane did not alter the levels of these synaptic proteins significantly, but pretreatment of mice with intranasal insulin significantly increased the levels of synaptophysin, synapsin-1 and PSD95 by approximately 10--40% in the brains of the anesthesia-treated mice ([Fig. 3A](#f3){ref-type="fig"},B). CREB is a transcription factor crucial to neuronal plasticity and long-term memory formation in the brain. Its transcriptional activity is mainly regulated by its phosphorylation[@b30]. We found that both the level of phosphorylated CREB, as evidenced by the p-CREB/GAPDH, and the net phosphorylation of CREB, as evidenced by the p-CREB/CREB, were increased by approximately 70% and 115%, respectively, after anesthesia, but pre-treatment with intranasal insulin did not induce further changes ([Fig. 3C,D](#f3){ref-type="fig"}). These results suggest that anesthesia with propofol and sevoflurane activate CREB activity. Intranasal insulin promotes brain insulin signaling --------------------------------------------------- We investigated the effect of intranasal insulin treatment and anesthesia on brain insulin signaling by determining the level and activation of each component of the signaling pathway, including insulin receptor β (IRβ), insulin-like growth factor-1 receptor β (IGF-1Rβ), insulin receptor substrate-1 (IRS-1), phosphatidylinositide 3-kinases (PI3K), 3-phosphoinositide-dependent protein kinase-1 (PDK1) and protein kinase B (AKT). The activation of these proteins was assessed by measuring their phosphorylation levels at the activity-dependent sites. We found that brain insulin signaling was somewhat disturbed in mice post anesthesia, as evidenced by significant reduction the levels of IGF-1Rβ and AKT in the anesthesia group as compared to the control group ([Fig. 4](#f4){ref-type="fig"}). We also found that the prior treatment of mice with intranasal insulin increased the levels of IRβ, p-IGF-1Rβ, and p-PDK1 (pS241) in the brains of the anesthesia-treated mice ([Fig. 4](#f4){ref-type="fig"}). These results suggest that intranasal insulin treatment can promote brain insulin signaling. Anesthesia and insulin induce transient brain biochemical changes in the brain ------------------------------------------------------------------------------ Because the anesthesia-induced hyperphosphorylation of tau was seen immediately after anesthesia but not 5 days later after completion of the probe test in the Morris water maze, we studied how long the anesthesia-induced brain biochemical changes and the intranasal insulin's effects last by investigating the brains of mice sacrificed 0 hr, 24 hrs, and 5 days after anesthesia. We found increased tau phosphorylation and changes of CREB phosphorylation, as well as insulin's protective role, only when the mice were sacrificed immediately after anesthesia ([Fig. 5](#f5){ref-type="fig"}). None of the anesthesia-induced changes nor insulin's effects were seen in the brains of mice sacrificed 24 hrs or 5 days after anesthesia. To learn whether increased tau phosphorylation is still detectable in selected neurons of mice 5 days after anesthesia, we stained the brain sections immunohistochemically with antibodies 12E8 and PHF-1 against phosphorylated tau. However, we did not find any neurons showing significant increase in the immunostaining (data not shown). These results indicate that although the behavioral impacts of anesthesia and insulin were detectable five days post anesthesia, the biochemical brain changes disappear within 24 hrs after anesthesia. Discussion ========== It is well known that general anesthesia can impair short-term memory and also increase the risk for long-term cognitive decline and dementia, especially for vulnerable individuals such as the elderly[@b1][@b5]. Anesthesia's adverse impact on cognition has also been replicated in animals[@b9][@b12][@b13][@b24]. To our knowledge, there is no treatment available to date to prevent the cognitive impairment that occurs in individuals post anesthesia. In the present study using mice as an experimental model, we found that intranasal administration of insulin can prevent cognitive impairment, as well as biochemical changes in the brain, induced by general anesthesia. These findings indicate a potential to use intranasal insulin for preventing anesthesia-induced cognitive impairment in vulnerable individuals after surgery. Besides epidemiological studies, previous studies have demonstrated that the use of various anesthetics induces hyperphosphorylation of tau in mice[@b8][@b24][@b27][@b28]. Tau hyperphosphorylation is known to promote neurodegeneration in AD[@b26]. A few studies also showed cognitive impairment of rodents after anesthesia[@b9][@b12][@b13][@b28]. In order to make the present study more clinically relevant, we used propofol to induce and sevoflurane to maintain anesthesia because the combinational use of these two anesthetics are commonly used in clinical practice for surgery. Aged mice (17--18-months old) were used in this study because elderly individuals are more vulnerable to anesthesia-induced cognitive impairment. With this anesthetic paradigm, we found that anesthesia induced impairment of spatial learning and memory, which is consistent with previous studies using different anesthetic paradigms[@b9][@b12][@b13][@b28]. In a recent study, we found that treatment of 3xTg-AD mice with daily intranasal administration of insulin for 7 days prior to anesthesia with propofol promotes brain insulin signaling and attenuates propofol-induced hyperphosphorylation of tau[@b24]. We therefore tested whether the same pretreatment with intranasal insulin could also prevent these brain changes induced by a combined use of propofol and inhaled anesthetic in aged non-transgenic mice. More importantly, we were eager to learn whether the insulin pretreatment can prevent anesthesia-induced cognitive impairment. We found in the present study that the pretreatment effectively prevented anesthesia-induced cognitive impairment. It also enhanced the levels of synaptic proteins and brain insulin signaling, as well as partially prevented anesthesia-induced increase of tau phosphorylation in the aged wild-type mice. It is possible administration of daily intranasal insulin for less than seven days may be sufficient to prevent anesthesia-induced cognitive impairment. Previous studies have reported increased tau phosphorylation and other biochemical changes in the brain immediately after general anesthesia[@b8][@b24][@b27][@b28], but how long these anesthesia-induced brain changes last after animals wake from anesthesia remains elusive. In the present study, we found that these brain changes were transient and returned to the normal levels within 24 hrs post anesthesia. These observations are consistent with an early report by Ikeda *et al.*[@b31] showing that anesthesia-induced increase in tau phosphorylation peaks at 0--10 min after ether exposure and is reversed to the normal level in 40--60 min. However, the cognitive impairment induced by anesthesia appears to last much longer, because we clearly found impairment in spatial memory in Morris water maze 5 days after anesthesia, when the brain biochemical changes had disappeared. Apparently, the anesthesia-induced transient molecular changes have a much longer functional impact to cognition. Alternatively, other molecular changes induced by anesthesia but not studied in the present study might last long and underlie the cognitive impairment. To search for a method to prevent anesthesia-induced brain changes and cognitive impairment, we selected insulin in our recent[@b24] and the present study because it is a neurotrophic factor and is important to neuroplasticity and cognition[@b18]. Furthermore, brain insulin signaling is deregulated in the brains of individuals with AD[@b32][@b33], in which progressive cognitive impairment and dementia are the major clinical symptoms. Systemic administration of insulin is obviously undesirable because it is difficult to enter into the brain and would otherwise disturb metabolism in the periphery and may lead to hypoglycemia. Intranasal administration, which was first introduced by W. H. Frey, bypasses the blood brain barrier and has been used successfully in animal studies and clinical trials in humans[@b21][@b34][@b35][@b36][@b37][@b38]. Intranasal administration of insulin does not appear to interfere with the insulin level or glucose metabolism in the periphery[@b21]. This insulin delivery method is simple and non-invasive and readily applicable in clinic. Our findings reported in the present study suggest that simply administering insulin into the nostrils of patients before anesthesia may prevent anesthesia's adverse effect on cognition. Future studies will determine the minimal dose and the best time before anesthesia for intranasal insulin administration. The underlying molecular mechanism by which intranasal insulin prevents anesthesia-induced cognitive impairment remains to be elucidated. Anesthesia might promote cognitive impairment and AD by promoting Aβ production[@b10][@b11][@b39], tau hyperphosphorylation[@b8][@b27][@b28][@b40], and neuroinflammation[@b41]. The present study demonstrates that intranasal insulin can prevent anesthesia-induced tau hyperphosphorylation and promote the expression of synaptic proteins. Although brain Aβ level was not determined in the present study, we found in a recent study that intranasal insulin reduces Aβ level in the brains of 3xTg-AD mice and inhibits microglial activation[@b40]. Therefore, intranasal insulin might prevent anesthesia's adverse effect on cognition through inhibiting tau hyperphosphorylation, neuroinflammation and synaptic damage induced by anesthesia. One mechanism by which anesthesia induces tau hyperphosphorylation is through down-regulation of protein phosphatase 2A (PP2A)[@b7][@b42]. By using 3xTg-AD mice, we recently found that intranasal insulin up-regulates PP2A in the anesthetized mouse brains, as evidenced by increased levels of both the total and the methylated form (more active form) of the catalytic subunit of PP2A[@b24]. In the present study using aged wild-type mice, we also studied the catalytic subunit of PP2A and its methylation by Western blots in the mouse brains and found that intranasal insulin treatment increased PP2A level and its methylation (data not shown). Therefore, intranasal insulin administration could prevent anesthesia-induced tau hyperphosphorylation partially through upregulation of PP2A. In conclusion, we report here that intranasal administration of insulin prior to anesthesia can prevent spatial learning and memory impairment and increased tau phosphorylation induced by general anesthesia induced by propofol and sevoflurane. These findings provide a simple strategy, i.e., administering insulin into the nose before anesthesia, to prevent postoperative cognitive deficit and increased risk for developing AD and dementia induced by general anesthesia. Materials and Methods ===================== Antibodies and reagents ----------------------- Primary antibodies used in this study are listed in [Table 1](#t1){ref-type="table"}. Peroxidase-conjugated anti-mouse and anti-rabbit IgG were obtained from Jackson ImmunoResearch Laboratories (West Grove, PA, USA). The enhanced chemiluminescence (ECL) kit was from Pierce (Rockford, IL, USA). The ABC staining system was from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Propofol was purchased from MP Biomedicals (Solon, OH, USA). Insulin (Humulin R U-100) was from Eli Lily (Indianapolis, IN, USA). Other chemicals were from Sigma-Aldrich (St. Louis, MO, USA). Animals and animal treatments ----------------------------- The breeding pairs of the C57BL6/129 mice were initially obtained from Jackson Laboratory (New Harbor, 124 ME, USA), and the mice were bred in our institutional animal colony. Mice were housed (4 ∼ 5 animals per cage) with a 12/12 h light/dark cycle and with ad libitum access to food and water. The housing, breeding, and animal experiments were in accordance with the approved protocol from the Institutional Animal Care and Use Committee of New York State Institute for Basic Research in Developmental Disabilities, according to the PHS Policy on Human Care and Use of Laboratory animals (revised March 15, 2010). Female mice at the ages of 17--18 months were used for this study. Aged mice were chosen because elderly individuals are more vulnerable to anesthesia-induced cognitive impairment. Intranasal delivery was carried out manually without anesthesia while the mouse head was restrained in a supine position with the neck in extension, as described[@b36]. A total of 1.75 U/17.5 μl insulin or 0.9% saline was delivered over both nares alternatively using a 10 μl Eppendorf pipetter. The mouse was held for an additional 5--10 seconds to ensure the fluid was inhaled. The successful nasal delivery by using this approach was confirmed by examination of ink (Fount India ink, Pelikan, Schindellegi, Switzerland) in the autopsied brains after nasal delivery with ink using the same approach (data not shown). All mice were treated with insulin or, as a control, saline daily for 7 consecutive days ([Fig. 1A](#f1){ref-type="fig"}). On the following day, the mice were injected intraperitoneally (i.p.) with propofol dissolved in intralipid (150 mg/kg body weight) or the equivalent amount of intralipid, followed by inhalation of 2.5% sevoflurane for 1 hr. Mice were sacrificed immediately, 24 hrs or 5 days post anesthesia by cervical dislocation. The mouse forebrains were removed immediately and divided into two hemispheres. The left hemispheres were fixed in 4% paraformaldehyde in 0.1 M phosphate buffer for immunohistochemical studies. The right hemispheres were further divided into rostral and caudal halves (separated coronally at the bregma level), flash frozen in dry ice, and stored at −80 °C for biochemical analyses at a later date. Morris Water Maze ----------------- Morris water maze (MWM) was used to evaluate spatial learning and memory of the mice. The test was performed in a circular white pool (with a diameter of 180 cm and a height of 60 cm) filled with white dye tinted water and maintained at room temperature (20 ± 1 °C). The maze was designated of two principal axes with each line bisecting the maze perpendicular to the other one to divide the maze into four equal quadrants. The end of each line demarcates four cardinal points: north (N), south (S), east (E) and west (W). A platform was positioned in the middle of one of the quadrants submerged 1 cm below water surface. Each mouse performed 4 trials per day for 4 consecutive days from semi-random start positions to find the hidden platform. Each trial was terminated as soon as the mouse climbed onto the hidden platform. If a mouse failed to find the platform within 90 sec, it was gently guided to it. At the end of each trial, the mouse was left on the platform for 20 sec, then dried and returned to its home cage. A 60 sec probe test without platform was performed 24 hr after the last trial. The swim path, swim distance (cm), escape latency (sec), swim speed (cm/sec), time spent in each quadrant (sec), distance traveled in each quadrant (cm), latency to enter the platform site zone (sec), and the number of platform site zone crossings were recorded through an automated tracking system (Smart video tracking system, Panlab; Havard Apparatus). Western blot analysis --------------------- Brain tissue was homogenized in pre-chilled buffer containing 50 mM Tris-HCl (pH7.4), 50 mM GlcNAc, 20 μM UDP, 1.0 mM EGTA, 2 mM Na~3~VO~4~, 100 mM NaF, 0.5 mM AEBSF, 1 μg/ml aprotinin, 10 μg/ml leupeptin, and 2 μg/ml pepstatin A. Protein concentrations of the homogenates were determined by the Pierce 660-nm Protein Assay. The samples were resolved in 10% or 12.5% SDS-PAGE and electrotransferred onto Immobilon-P membrane (Millipore, Bedford, MA, USA). The blots were then probed with primary antibody and developed with the corresponding horseradish peroxidase-conjugated secondary antibody and ECL kit. Immunohistochemistry -------------------- Frozen mouse brain sagittal sections (40-μm thick) were first washed with phosphate-buffered saline (PBS) for three times, 15 min each, followed by incubation in 0.5% Triton X-100 for 20 min. The sections were then washed with PBS for another 10 min and then blocked in PBS containing 5% normal goat serum and 0.1% Triton X-100 for 30 min, followed by incubation with 12E8 or PHF1 in the blocking solution at 4 °C overnight. After washing with PBS, the sections were incubated with Alexa 488-conjugated goat anti-mouse IgG (1:1000) plus TO-PRO-3 in the blocking solution at room temperature for 2 hrs. The sections were finally washed, mounted, and cover slipped using Prolong^®^ gold antifade mountant (Invitrogen, Carlsbad, USA). The immunostaining was analyzed by using a laser scanning confocal microscope (PCM 200, Nikon). Statistical analysis -------------------- For biochemical analyses, data were analyzed by one-way ANOVA followed by Tukey's post hoc tests or unpaired two-tailed *t* tests, using GraphPad. All data are presented as means ± SEM, and *p* \< 0.05 was considered statistically significant. Additional Information ====================== **How to cite this article**: Zhang, Y. *et al.* Intranasal Insulin Prevents Anesthesia-Induced Spatial Learning and Memory Deficit in Mice. *Sci. Rep.* **6**, 21186; doi: 10.1038/srep21186 (2016). This work was supported in part by Nantong University and New York State Office for People with Developmental Disabilities, as well as grants from the National Natural Science Foundation of China (81030059, 81400866), the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD), and the U.S. Alzheimer's Association (2015-NIRG-339945). We thank Drs. J. Goodman and C. Wang (New York State Institute for Basic Research in Developmental Disabilities, Staten Island, NK, USA) for their assistance in gas anesthesia and Drs. D. Schenk (Elan Pharmaceuticals, San Francisco, CA, USA) and P. Davies (Albert Einstein College of Medicine, Bronx, NY, USA) for providing antibodies 12E8 and PHF-1, respectively. **Author Contributions** Y.Z., C.D. and Y.C. performed the experiments. K.I., F.L. and C.X.G. supervised the study. Y.Z. and C.X.G. wrote the paper. {#f1} {#f2} {#f3} {#f4} {#f5} ###### Primary antibodies used in this study. Antibody Type Specificity Phosphorylation sites Source/Reference --------------- ------- --------------- ------------------------------------------- ---------------------------------------- IRβ Poly- IRβ   Cell Signaling Technology, Danvers, MA IGF-1Rβ Poly- IGF-1Rβ   Cell Signaling Technology P-IRβ/IGF-1Rβ Mono- P-IRβ/IGF-1Rβ Tyr1150/1151(IRβ), Tyr1135/1136 (IGF-1Rβ) Cell Signaling Technology IRS1 Poly- IRS1   Cell Signaling Technology IRS1 pS307 Poly- P-IRS1 Ser307 Cell Signaling Technology PI3K p85 Poly- PI3K (p85)   Cell Signaling Technology P-PI3K p85 Poly- P-PI3K (p85) Tyr458/Tyr199 Cell Signaling Technology PDK1 Poly- PDK1   Cell Signaling Technology PDK1 pS241 Poly- P-PDK1 Ser241 Cell Signaling Technology AKT Poly- AKT   Cell Signaling Technology AKT pS473 Poly- P-AKT Ser473 Cell Signaling Technology AKT pT308 Poly- P-AKT Thr308 Cell Signaling Technology R134d Poly- Tau   Tatebayashi *et al.*, 2012[@b43] pT181 Poly- P-tau Thr181 Invitrogen, Grand Island, NY pS199 Poly- P-tau Ser199 Invitrogen pT205 Poly- P-tau Thr205 Invitrogen pT212 Poly- P-tau Thr212 Invitrogen pT231 Poly- P-tau Thr231 Invitrogen pS396 Poly- P-tau Ser396 Invitrogen 12E8 Mono- P-tau Ser262/356 Dr. D. Schenk[@b44] AT8 Mono- P-tau Ser202/T205 Thermo Fisher Scientific, Rockford, IL PHF-1 Mono- P-tau Ser396/404 Dr. P. Davies[@b45] pS409 Poly- P-tau Ser409 Invitrogen Synapsin-1 Poly- Synapsin-1   Santa Cruz Biotechnology Synaptophysin Mono- Synaptophysin   Millipore PSD95 Mono- PSD95   Cell Signaling Technology GluR1 Poly- GluR1   Millipore GFAP Mono- GFAP   Millipore ED1 Mono- CD68   Abcam CREB Mono- CREB   Cell Signaling Technology p-CREB Mono- p-CREB Ser133 Cell Signaling Technology Anti-GAPDH Poly- GAPDH   Santa Cruz Biotechnology [^1]: These authors contributed equally to this work.

**What was known?** FDH is a rare genetic disorder with developmental anomalies of, cutaneous skeletal, dental, and ocular systems, and facial dysmorphism. Introduction {#sec1-1} ============ Focal dermal hypoplasia (FDH) is a rare mesoectodermal disorder inherited by an X-linked dominant gene, which is lethal in homozygous males. Goltz syndrome was first described by Leibermann as *atrophoderma linearis maculosa et papillomatosis congenitalis*.\[[@ref1]\] Subsequently in 1962, Goltz *et al*. reported three cases of FDH and distinguished them as separate entities.\[[@ref2]\] There are several 100 cases described in the literature but only a few cases from India. We describe a rare case of Goltz syndrome along with the unusual feature of the presence of an atrial septal defect and supernumerary nipples which has not been reported earlier in the literature. Case Report {#sec1-2} =========== An eight-month-old term female child, born of non-consanguineous marriage presented with multiple hypopigmented skin lesions on the trunk and extremities from birth. They appeared as linear erosions at birth that healed gradually within a few days, leaving behind hypopigmentation interspersed with hyperpigmentation at the periphery. Perinatal history was uneventful. She had abnormalities of limbs and nails from birth. Her developmental milestones were delayed. Past history was remarkable for exomphalos minor that was operated and atrial septal defect at birth that had undergone spontaneous closure after three months of age. None of the siblings or family members of the child had similar symptoms. General examination revealed microcephaly, hypertelorism, megalopinna, low-set ears, a broad nasal bridge, and a pointed chin. Cutaneous examination revealed multiple hypopigmented atrophic linear streaks, and macules of varying sizes present on the forehead, neck, trunk, thighs, gluteal region, genitalia, and upper and lower limbs along the lines of blaschko \[Figures [1](#F1){ref-type="fig"} and [2](#F2){ref-type="fig"}\]. There was syndactyly of left foot and lobster claw deformity (ectrodactyly) of right foot \[[Figure 3](#F3){ref-type="fig"}\]. She had supernumerary nipples \[[Figure 4](#F4){ref-type="fig"}\]. A few skin-colored growths were present over the proximal nail fold of the right middle finger and right foot. Nails were dystrophic. Hairs and oral mucosa were normal. Systemic examination was normal. {#F1} {#F2} {#F3} {#F4} Complete hemogram and serum chemistry profile were normal. Skin biopsy on the trunk showed focal spongiosis, occasional dyskeratotic keratinocytes, mild papillary edema, and focal papillary dermal fibrosis with superficial perivascular lymphocytic infiltrate \[[Figure 5](#F5){ref-type="fig"}\]. Ophthalmological examination and computed tomography scan of brain were normal. X-ray of long bones was normal. Based on the clinical presentation and characteristic skeletal abnormalities, a diagnosis of Goltz syndrome was made. {#F5} Discussion {#sec1-3} ========== Goltz syndrome (FDH) is a rare inherited disorder characterized by developmental defects of the skin in conjunction with ocular, dental, and skeletal abnormalities. Ninety percent of the reported cases are females, suggesting an X-linked dominant mode of inheritance.\[[@ref3]\] When the disorder occurs in males, it probably represents a mosaic condition. The variation in severity among affected females is due to the mosaicism as a consequence of the Lyon hypothesis.\[[@ref4]\] Mosaicism for mutations in the PORCN gene on chromosome Xp11.23 has been implicated as the genetic basis for FDH. Paller hypothesized that the distribution pattern indicates that Wnt proteins stimulate epidermal-dermal signals to induce changes in the dermis.\[[@ref5]\] The hallmark of FDH is thinning of the dermis, which results in depressed linear lesions and soft, reddish-yellow outpouchings of the skin caused by herniation of subcutaneous fat. At birth, the lesions may present as blisters or erosions that leave behind typical atrophic scars that can mimic incontinentia pigmenti. Red, hypopigmented, or depigmented atrophic macules, arranged in a linear or blaschkoid pattern or in a reticulate grouping can be found on any part of the body, but typically involve the thighs, buttocks, and trunk.\[[@ref6]\] Telangiectasias are commonly seen interspersed between the atrophic plaques. Multiple soft pinkish yellow to brown nodules can be present from birth particularly in the popliteal and cubital fossae.\[[@ref6]\] Another characteristic feature is the development of raspberry-like papillomas around the lips, anus, and genitalia, and on the end of the digits. Other cutaneous changes reported are dryness associated with pruritus, photosensitivity, hyperkeratosis of palms and soles, sweating abnormalities, and dermatoglyphic changes.\[[@ref7]\] The second most common extracutaneous abnormality is skeletal defects seen in 60-70% of the patients. These include syndactyly, polydactyly, ectrodactyly, hypoplasia of the digits, and vertebral abnormalities like scoliosis, kyphosis, vertebral body fusions, and spina bifida. Lobster claw deformity (split hand or split foot) is a major distinct feature of this condition.\[[@ref7]\] In 20% of the cases, there is osteopathia striata, which are longitudinal linear striations in the metaphyses of the long bones seen on radiography. Associated ocular abnormalities are colobomas, microphthalmia, microcornea, and ectopia lentis.\[[@ref6][@ref7]\] Patients with FDH have a typical facial appearance with asymmetry, pointed chin, maxillary hypoplasia, broad nasal tip with a narrow bridge, and notching of alae nasi. Other associated abnormalities include enamel defects, sparse and brittle hair, nail dystrophy, short stature, mental retardation, hearing loss, microcephaly, cleft lip and palate, duodenal atresia, intestinal malrotation, umbilical, inguinal, epigastric, or diaphragmatic hernia, horseshoe kidneys, bicornuate uterus, and cardiac defects like patent ductus arteriosus, ventricular septal defect, and hypoplasia of lungs. Rare cases of absent nipples are also reported.\[[@ref8]\] In our case, the child had supernumerary nipples. Skin biopsy of an atrophic lesion showed a normal or thinned epidermis overlying a severely hypoplastic dermis with adipose tissue impinging on the epidermis. Electron microscopy revealed loosely arranged abnormally scattered collagen bundles.\[[@ref9]\] Gene sequence analysis for detection of the PORCN gene mutation could also be carried out. The most common differential diagnosis for FDH is incontinentia pigmenti, Rothmund-Thomson syndrome, and MIDAS (MIDAS: Microphthalmia, dermal aplasia, and sclerocornea syndrome). The linear pattern, stages of pigmentary changes, eosinophilic spongiosis, and absence of fat herniation rules out incontinentia pigmenti. Other differential diagnosis to be considered is epidermal nevus, Bart syndrome, Adams-Oliver syndrome, ectodermal dysplasia, and congenital erosive and vesicular dermatoses.\[[@ref10]\] There is no effective treatment for this disease and it requires a multidisciplinary approach. Patients have a normal life span. Surgical correction of deformities can be considered, and genetic counseling should be given to parents. Involvement of the skin with skeletal abnormalities supports a clinical diagnosis of FDH and is thus present in all reported cases. In our case, the child had typical features of Goltz syndrome, but the rare association with atrial septal defect and supernumerary nipples has not been described in previous reports. **What is new?** The association of atrial septal defect and supernumerary nipples with FDH was an unusual feature of our case. **Source of support:** Nil **Conflict of Interest:** Nil.

*Cldn4* : *claudin 4* ESC : embryonic stem cell FS : functional state GIP : insulinotropic polypeptide GLP1 : glucagon‐like peptide‐1 TJ : tight junction A large array of critical regulatory molecules such as transcription factors [1](#feb412735-bib-0001){ref-type="ref"} have been well documented to regulate functional state (FS) of pancreatic endocrine islets of Langerhans that sustain normal glucose homeostasis of the body, mainly via hormones from insulin‐secreting β cells and glucagon‐secreting α cells. To regulate this FS, the islets have evolved to have many unique features including being structurally spread throughout the exocrine pancreas. They do not have their own basement membrane [2](#feb412735-bib-0002){ref-type="ref"} but are closely in contact with that of intra‐islet capillaries [3](#feb412735-bib-0003){ref-type="ref"}, [4](#feb412735-bib-0004){ref-type="ref"}. In viewing the unique structural arrangement of the pancreatic endocrine tissue, we hypothesize that undiscovered structural molecules are important in regulating islet FS. This study aims to study such molecules. We have previously demonstrated that functional genes can be identified with the effective bioinformatics analyses of transcriptomic data sets [5](#feb412735-bib-0005){ref-type="ref"}, [6](#feb412735-bib-0006){ref-type="ref"}. Using the transcriptome data sets and other approaches, we demonstrated the involvement of previously understudied claudin (*Cldn*) family genes of tight junctions (TJs) in the islets. TJs between neighbouring epithelial cells have important biochemical and physiological roles in multiple organs by selective and critical permeability to important compounds, in addition to working as a structural barrier by forming strands against unrestricted paracellular passaging [7](#feb412735-bib-0007){ref-type="ref"}. Surprisingly, the expression and function of TJ molecules in the islets of Langerhans are largely unknown. Cldn family molecules are the tetraspan transmembrane proteins of TJs, forming a structural barrier between the apical and basal portions of epithelial cellular sheets [8](#feb412735-bib-0008){ref-type="ref"}. This family consists of at least 28 members in mice and humans. Cldns are classically expressed in epithelium and are categorized as cation‐selective, anion‐selective and water‐permeable channels, and charge‐selective barriers. The distribution of Cldns varies from one tissue type to another [9](#feb412735-bib-0009){ref-type="ref"}: for example, Cldn1 regulates permeability in the epidermis [10](#feb412735-bib-0010){ref-type="ref"}, Cldn5 does so in the blood--brain barrier [11](#feb412735-bib-0011){ref-type="ref"}, Cldn11 in the myelin and Sertoli cells [12](#feb412735-bib-0012){ref-type="ref"}, Cldn14 in inner ears [13](#feb412735-bib-0013){ref-type="ref"} and Cldn18 in the stomach [14](#feb412735-bib-0014){ref-type="ref"}. Cldn4 is a member of charge‐selective Cldns, usually partnering with Cldn8 and Cldn12 [9](#feb412735-bib-0009){ref-type="ref"}. Cldn4 is prominently expressed in the lung, intestinal and kidney‐collecting tubular epithelia [8](#feb412735-bib-0008){ref-type="ref"}, the urinary bladder and skin [15](#feb412735-bib-0015){ref-type="ref"} and the atypical and nonpolarized epithelial cells such as thymic epithelial cells [16](#feb412735-bib-0016){ref-type="ref"}, [17](#feb412735-bib-0017){ref-type="ref"}. Though highly expressed, Cldn4 seems not to play a major role on the physiology of the lung [18](#feb412735-bib-0018){ref-type="ref"}. Cldn4 is functionally involved in the generation of thymus CD4/CD8 single positive T lymphocytes [19](#feb412735-bib-0019){ref-type="ref"}. Gene knockout experiments demonstrate that *Cldn4* is critical for renal chloride (Cl^−^) reabsorption and blood pressure regulation [20](#feb412735-bib-0020){ref-type="ref"}, [21](#feb412735-bib-0021){ref-type="ref"}. Cldn4 was previously detected by immunofluorescence in the rat pancreatic tissue as well as in the islets of Langerhans [22](#feb412735-bib-0022){ref-type="ref"}. However, no major pathophysiological effect on energy metabolism has been documented on any *Cldn* gene. Here, we show with a number of approaches that Cldn4 in the mouse pancreatic islets is associated with regulating FS of the islets, implicating in translational research for better diabetes therapies. Materials and methods {#feb412735-sec-0002} ===================== Mouse lines {#feb412735-sec-0003} ----------- The *Cldn4*‐deleted, floxed *Cldn4* and CAG‐Cre mice [23](#feb412735-bib-0023){ref-type="ref"} were bred onto a C57BL/6 background for at least 10 generations. PCR‐based genotyping for Cldn4^+/−^ and Cldn4^−/−^ mouse lines was described elsewhere [23](#feb412735-bib-0023){ref-type="ref"}. Cldn4^−/−^, Cldn4^+/−^, floxed *Cldn4* and CAG‐Cre (the latter two lines along with the C57BL/6 designated as Cldn4^+/+^) mouse lines and the type 2 diabetes model db/db mice provided by Jackson Laboratory (Mount Desert Island, ME, USA) were maintained in a 22 ± 1 °C, 12:12 light/dark cycle environment with free access to food and water and used at 8--12 weeks of age. Compliance with Ethical Standards {#feb412735-sec-0004} --------------------------------- All applicable international, national and/or institutional guidelines for the care and use of animals were followed, namely the Animal Ethics Committees of the University of Western Australia, Australia and Kyoto University, Japan, approved the use of experimental animals. MIN6 cells {#feb412735-sec-0005} ---------- Culture, maintenance and passage of MIN6 cells were described previously [24](#feb412735-bib-0024){ref-type="ref"}. Isolation of adult islets {#feb412735-sec-0006} ------------------------- Islets of Langerhans were isolated from euthanized (cervical dislocation) 8‐ to 12‐week‐old C57BL/6 mice, 12‐week db/^+^ mice and db/db diabetic mice as described recently [5](#feb412735-bib-0005){ref-type="ref"}. Briefly, the pancreas was injected via the bile duct with collagenase P solution (Sigma, Melbourne, Vic., Australia, 1.2 mg·mL^−1^ dissolved in Hanks' balanced salt solution containing 2 m[m]{.smallcaps} Ca^2+^ and 20 m[m]{.smallcaps} HEPES). Islets and exocrine layers were isolated by density gradient Histopaque (Sigma) centrifugation, washed and hand‐picked islets for RNA. Glucose‐stimulated insulin secretion assay {#feb412735-sec-0007} ------------------------------------------ Glucose‐stimulated insulin secretion assay was performed essentially as described [25](#feb412735-bib-0025){ref-type="ref"}. Briefly, the indicated passaged MIN6 cells were washed twice with warm PBS. After pre‐incubation with the Krebs--Ringer buffer at 37 °C for 90 min, the cells were incubated at 37 °C for 60 min with basal D‐glucose (2.75 m[m]{.smallcaps}) or stimulus D‐glucose (27.5 m[m]{.smallcaps}). Then, each conditioned medium was collected to determine the insulin concentration using a mouse insulin ELISA kit (Mercodia AB, Uppsala, Sweden). Subsequently, the culture was trypsinized and the number of MIN6 cells was determined with a haemocytometer. Oral glucose tolerance test and serum incretin concentrations {#feb412735-sec-0008} ------------------------------------------------------------- After overnight fasting, mice were orally administered 10% glucose (2 g·kg^−1^ body weight) and blood glucose levels were measured with tail vein blood using OneTouch UltraVue (Johnson & Johnson K.K., Nishikanda Chiyoda‐Ku, Japan). Serum glucose‐dependent insulinotropic polypeptide (GIP) and glucagon‐like peptide‐1 (GLP1) concentrations were determined with Bio‐Plex (Bio‐Rad, Shinagawa‐ku, Tokyo, Japan), according to the manufacturer\'s instruction. Generation of endodermal cells {#feb412735-sec-0009} ------------------------------ Endodermal cells were generated from directed differentiation of undifferentiated mouse embryonic stem cell (ESC) W9.5 line as we described previously [6](#feb412735-bib-0006){ref-type="ref"}. Bioinformatics analyses {#feb412735-sec-0010} ----------------------- Bioinformatics analyses of transcriptome data sets were performed on published data sets generated from ESCs and isolated adult mouse islets and during differentiation of islet progenitors [25](#feb412735-bib-0025){ref-type="ref"}. Gene mining was performed as described previously [6](#feb412735-bib-0006){ref-type="ref"}, [26](#feb412735-bib-0026){ref-type="ref"}. Briefly, the differential expression of genes (*P* ≤ 0.05; −1 ≤ log~2 ~≥ 1) based on bioinformatics contrast between ESC and islet data sets and other contrasts were analysed using the Limma package in the '[r]{.smallcaps}' environment (<http://bioinf.wehi.edu.au/limma>). We will provide the data sets on request. Indirect immunofluorescence {#feb412735-sec-0011} --------------------------- Rat anti‐claudin 4 antibodies were purchased from Millipore (Castle Hill, NSW, Australia) and Abcam (Cambridge, UK), respectively. Biotinylated anti‐human insulin and rabbit anti‐glucagon antibodies were purchased from R&D Systems (Minneapolis, MN, USA). The antibody sources and staining procedures for laminin and nidogen 1 were described previously [2](#feb412735-bib-0002){ref-type="ref"}. Cells and pancreases from all genotypes were fixed in 4% paraformaldehyde and the latter processed for histological sections. After dewaxing and rehydration, tissue sections along with cell preparations were stained with primary antibodies and reacted with streptavidin FITC (BD, Bergen County, NJ, USA) and Texas Red anti‐human, anti‐rat or anti‐mouse (Vector Labs, Burlingame, CA, USA) as we described previously [27](#feb412735-bib-0027){ref-type="ref"}. Slides were observed and microphotographed with the inverse IX71 Olympus fluorescence microscope (Olympus, Tokyo, Japan). Gene expression analyses by qRT‐PCR {#feb412735-sec-0012} ----------------------------------- Total RNA was extracted from the epididymal fat, liver, kidney and skeletal muscles in euthanized (cervical dislocation) adult C57BL/6 mice and other indicated cells with RNeasy Plus Mini Kit (Qiagen Science, Melbourne, Vic., Australia) or the TRIzol‐based method and quantified by a NanoDrop ND‐1000 Spectrophotometer (Australian Biolab group, Melbourne, Vic., Australia) as described previously [26](#feb412735-bib-0026){ref-type="ref"}. RNA (200--400 ng) was reverse‐transcribed with reverse transcriptase to cDNA in 40 μL, 1 μL of which (5--10 ng RNA/reaction, without reverse transcriptase as a negative control) was amplified by qRT‐PCR analysis essentially as we described [26](#feb412735-bib-0026){ref-type="ref"}. Primer sequences are presented in Table [1](#feb412735-tbl-0001){ref-type="table"}. ###### Sequences for qRT‐PCR primers. +----------+--------------------------+-----------------------------+----------------------------+--------------+ | Gene | Forward primer | Reverse primer | Annealing temperature (°C) | Product (bp) | | | | | | | | | 5′ → 3′ | 5′ → 3′ | | | +==========+==========================+=============================+============================+==============+ | *Acta1* | `caatcgtgctgtggttgcag` | `ggagcaaaacagaatggctgg` | 60 | 191 | +----------+--------------------------+-----------------------------+----------------------------+--------------+ | *Cldn3* | `ggagtgcttttcctgttggc` | `cgtagtccttgcggtcgtag` | 60 | 295 | +----------+--------------------------+-----------------------------+----------------------------+--------------+ | *Cldn4* | `ccaagtcatggtgtgctgag` | `cactgggctgcttctaggtc` | 60 | 217 | +----------+--------------------------+-----------------------------+----------------------------+--------------+ | *Fev* | `cggcgtctactcttccctgt` | `catctccgacgggatctggc` | 60 | 191 | +----------+--------------------------+-----------------------------+----------------------------+--------------+ | *FoxO1* | `agccgcgcaagaccag` | `ttgaattcttccagcccgcc` | 60 | 195 | +----------+--------------------------+-----------------------------+----------------------------+--------------+ | *Hnf4a* | `ggatatggccgactacagcg` | `agatggggacgtgtcattgc` | 60 | 100 | +----------+--------------------------+-----------------------------+----------------------------+--------------+ | *IA1* | `gccacccgtctgagaataga` | `ggagtcacagcgagaagacc` | 60 | 231 | +----------+--------------------------+-----------------------------+----------------------------+--------------+ | *Lama1* | `cggcgcgtaaagatttccag` | `ctcctgggtcttgcttccag` | 60 | 290 | +----------+--------------------------+-----------------------------+----------------------------+--------------+ | *Lamb1* | `gccgtcctaatgtggttgga` | `agctgggaaagccccaatac` | 60 | 210 | +----------+--------------------------+-----------------------------+----------------------------+--------------+ | *Isl1* | `cccgggggccactatttg` | `cgggcacgcatcacgaa` | 60 | 397 | +----------+--------------------------+-----------------------------+----------------------------+--------------+ | *MafA* | `atcatcactctgcccaccat` | `agtcggatgacctcctcctt` | 60 | 208 | +----------+--------------------------+-----------------------------+----------------------------+--------------+ | *Mmp2* | `ggctctgtcctcctctgtag` | `tgccctcctaagccagtct` | 60 | 296 | +----------+--------------------------+-----------------------------+----------------------------+--------------+ | *Mycl1* | `gagaacggctggagagagtg` | `ttcaccttcagaatcgctggg` | 60 | 200 | +----------+--------------------------+-----------------------------+----------------------------+--------------+ | *NeuroD* | `cttggccaagaactacatctgg` | `ggagtagggatgcaccgggaa` | 60 | 228 | +----------+--------------------------+-----------------------------+----------------------------+--------------+ | *Nid1* | `atcagcaccatccctgaaac` | `tcaataccgctgaactgctg` | 60 | 206 | +----------+--------------------------+-----------------------------+----------------------------+--------------+ | *Pax4* | `ccacctctctgcctgaagac` | `cccacagcatagctgacaga` | 60 | 236 | +----------+--------------------------+-----------------------------+----------------------------+--------------+ | *Pou3f4* | `gctgcctcgaatccctacag` | `cagttgcagatcttcgcgtc` | 60 | 261 | +----------+--------------------------+-----------------------------+----------------------------+--------------+ | *Rfx6* | `gcttgctggtctaccctgag` | `catcatctgcgtgatgctct` | 60 | 233 | +----------+--------------------------+-----------------------------+----------------------------+--------------+ | *Rps18* | `tgtggtgttgaggaaagcag` | `tcccatccttcacatccttc` | 60 | 155 | +----------+--------------------------+-----------------------------+----------------------------+--------------+ | *Snail1* | `agttgactaccgaccttgcg` | `tgcagctcgctatagttggg` | 60 | 128 | +----------+--------------------------+-----------------------------+----------------------------+--------------+ | *Snail2* | `ggaccgttatccgcccg` | `tatggggaaataataactgtgtgtg` | 60 | 130 | +----------+--------------------------+-----------------------------+----------------------------+--------------+ John Wiley & Sons, Ltd The power SYBR Green PCR Master Mix‐based protocol was utilized, and all quantifications were normalized to the internal 18s rRNA level, as described recently [5](#feb412735-bib-0005){ref-type="ref"}. Briefly, cDNA was amplified with PCR: 95 °C for 10 min, followed by 40 cycles of 95 °C for 15 s and 60 °C for 1 min. The number of cycles of threshold (*C* ~t~) was measured with an ABI Prism 7900HT Sequence Detection System (Applied Biosystems, Foster City, CA, USA) or a Rotor‐Gene RG‐3000 (Corbett Research, Sydney, NSW, Australia). Statistical analysis {#feb412735-sec-0013} -------------------- Experiments were performed in at least three biological repeats. Data are expressed as mean ± standard deviation (SD). Statistical differences between groups are analysed with nonparametric, unpaired Mann--Whitney *U* tests or Student\'s *t* tests in samples with numerous biological repeats. Results {#feb412735-sec-0014} ======= Bioinformatics analyses identified unique pancreatic islet genes {#feb412735-sec-0015} ---------------------------------------------------------------- To identify structural molecules that may regulate islet FS, we first conducted bioinformatics analyses to survey unique genes in adult pancreatic islets on the published global transcriptional data sets [6](#feb412735-bib-0006){ref-type="ref"}, [25](#feb412735-bib-0025){ref-type="ref"}. A bioinformatics contrast of the data sets generated from isolated functional islets to undifferentiated pluripotent ESCs (as a baseline) showed that there were 1618 and 1630 genes negatively and positively enriched (Log~2~ scale), respectively (Fig. [1](#feb412735-fig-0001){ref-type="fig"}A). Here, we only focused our analyses on genes that encode structural molecules for TJs, the basement membrane and mesenchymal tissue. Genes for the latter were analysed as the adult pancreatic islets are reported to have mesenchymal features [28](#feb412735-bib-0028){ref-type="ref"}. {ref-type="ref"} during *in vitro* differentiation of islet progenitors for 0 (Day 0, namely undifferentiated islet progenitors)‐, 1 (Day 1, namely 1‐day differentiation of islet progenitors)‐ and 4 (Day 4, namely 4‐day differentiation of islet progenitors)‐day differentiation and using adult islets (islet) as the reference.](FEB4-10-28-g001){#feb412735-fig-0001} Tight junction *claudin 4* gene was highly enriched in adult functional islets {#feb412735-sec-0016} ------------------------------------------------------------------------------ In the bioinformatics contrast, there were 22 single exon TJ *Cldn* genes. Whereas majority of *Cldn* genes (*Cldn1*,*Cldn5*,*Cldns7‐9* and *Cldns12‐18*) were undetectable, *Cldn6* was strikingly negatively enriched (Fig. [1](#feb412735-fig-0001){ref-type="fig"}B). Interestingly, we revealed that several TJ genes *Cldn3* and *Cldn19* and in particular *Cldn4* were significantly enriched (Fig. [1](#feb412735-fig-0001){ref-type="fig"}B). *MarvelD3*, the TJ‐associated member of occludin family, was also enriched (Fig. [1](#feb412735-fig-0001){ref-type="fig"}C), but not essential for the formation of the TJs [29](#feb412735-bib-0029){ref-type="ref"}. To confirm the high enrichment of *Cldn4* in functional islets, qRT‐PCR analyses were conducted with RNA extracted from adult islets compared to that from pancreatic exocrine cells, a typical epithelial tissue and from the lymph nodes, a mesenchymal tissue. *Cldn4* in adult islets was indeed highly enriched by approximately 75‐fold, whereas *Cldn3* only twofold to threefold (Fig. [1](#feb412735-fig-0001){ref-type="fig"}D), compared to the exocrine cells. We also surveyed their expression in peripheral glucose metabolic tissues such as fat, liver and skeletal muscle using the renal tissue as the positive control [8](#feb412735-bib-0008){ref-type="ref"}. *Cldn4* expression was approximately 20‐fold to 30‐fold higher in the islets than in the kidney and the skeletal muscle (Fig. [1](#feb412735-fig-0001){ref-type="fig"}D,E); *Cldn3* was, however, mainly expressed in the liver, whereas both genes were not detected in fat (Fig. [1](#feb412735-fig-0001){ref-type="fig"}E). *Claudin 4* was highly expressed in adult functional islets {#feb412735-sec-0017} ----------------------------------------------------------- To determine at what developmental stage *Cldn4* is highly enriched, we performed bioinformatics contrasts of published transcriptome data sets generated from *in vitro* differentiation of purified mouse islet progenitors into immature insulin‐secreting cells and from isolated adult islets [25](#feb412735-bib-0025){ref-type="ref"}. Indeed, we demonstrated that *Cldn4* and to a less extent *Cldn3* and *Cldn23* were significantly enriched during functional maturation of islet cells (Fig. [1](#feb412735-fig-0001){ref-type="fig"}F). Taken together, we conclude that Cldn4 is the highly expressed TJ molecule amongst the Cldn family in adult islets. Claudin 4 was down‐regulated in functionally compromised dedifferentiated β cells {#feb412735-sec-0018} --------------------------------------------------------------------------------- If the Cldn4 molecule is a critical regulator, we hypothesized that reduction of its expression or genetic deletion will compromise the FS of adult pancreatic islets. We first investigated the dynamic changes of Cldn4 expression in a characterized cellular model of dedifferentiation, the passaged MIN6 β cells [24](#feb412735-bib-0024){ref-type="ref"}. β‐cell dedifferentiation is broadly defined as becoming insulin‐negative cells, losing function of GSIS, re‐expressing markers of islet progenitors and even transdifferentiating to other islet cell types [24](#feb412735-bib-0024){ref-type="ref"}, [30](#feb412735-bib-0030){ref-type="ref"}. The expression of Cldn4 molecule was progressively and significantly (*P* \< 0.001) down‐regulated at mRNA (Fig. [2](#feb412735-fig-0002){ref-type="fig"}A) but not clearly reduced at protein (Fig. [2](#feb412735-fig-0002){ref-type="fig"}B) levels in dedifferentiating MIN6 cells [24](#feb412735-bib-0024){ref-type="ref"}, whereas *Cldn3* was significantly up‐regulated in dedifferentiating MIN6 cells (*P* \< 0.001). The significance of different expression profiles between *Cldn3* and *Cldn4* deserves further investigation. To shed mechanistic lights, we investigated dynamic changes of many progenitor and functional transcription factor genes. The functional transcription factor genes *Hnf4a*,*FoxO1*,*NeuroD*,*Pax4* and *Pou3f4* (also known as *Brn4*) expression was significantly (*P* \< 0.001) down‐regulated at early dedifferentiating stage. The expression of islet progenitor transcription factor genes *Fev* and *Isl1* was progressively (*P* \< 0.001) up‐regulated, whereas *IA1* (also known as *Insm1*), *Mycl1* and *Rfx6* was unchanged (Fig. [2](#feb412735-fig-0002){ref-type="fig"}C). We also confirmed that dedifferentiated MIN6 cells significantly lost their glucose‐stimulated insulin secretion (Fig. [2](#feb412735-fig-0002){ref-type="fig"}D), consistent with previous report [24](#feb412735-bib-0024){ref-type="ref"}. ![*Claudin 4* is down‐regulated during dedifferentiation of functional βcells. (A) qRT‐PCR analysis of *Cldn3* and *Cldn4*. RNA was extracted from MIN6 cells at passages 34, 48, 68 and 82. (A, B) Data presented as mean ± SD, *n* = 3, \*\**P* \< 0.01 compared to all other passages (Mann--Whitney *U* tests). (B) Immunofluorescence analysis. MIN6 cells at passages 34 and 84 were stained with anti‐Cldn4 (red) and the DNA dye DAPI (blue). Scale bar = 20 μm. (C) Glucose‐stimulated insulin secretion analysis. Indicated passaged MIN6 cells were exposed to basal glucose (2.75 m[m]{.smallcaps}) or stimulus glucose (27.5 m[m]{.smallcaps}) for the determination of insulin concentrations. Data presented as mean ± SD, *n* = 3, \*\**P* \< 0.01 compared to the basal glucose condition (Mann--Whitney *U* tests). (D) qRT‐PCR analysis of selected transcription factor genes associated with islet function and differentiation. RNA was extracted from MIN6 cells at passages 36, 48, 69 and 80. Data presented as mean ± SD, *n* = 3, \*\**P* \< 0.01 compared to all other passages (except *FoxO1* and *NeuroD* not vs passage 80) (Mann--Whitney *U* tests). (E) qRT‐PCR analysis of *Cldn3* (grey bars) and *Cldn4* (black bars). RNA was extracted from isolated db/+ and db/db diabetic islets. Data presented as mean ± SD, *n* = 3, \*\**P* \< 0.01 compared to db/db diabetic islets (Mann--Whitney *U* tests).](FEB4-10-28-g002){#feb412735-fig-0002} To provide evidence whether the down‐regulation of *Cldn4* takes place in the dedifferentiated primary β cells, islets were isolated and analysed from the well‐characterized type 2 diabetes model db/db mice that have dedifferentiated β cells [24](#feb412735-bib-0024){ref-type="ref"}, [31](#feb412735-bib-0031){ref-type="ref"}. We showed that *Cldn4* was down‐regulated approximately 60% in dedifferentiated db/db islets (Fig. [2](#feb412735-fig-0002){ref-type="fig"}E). Collectively, these data suggested that Cldn4 involves in regulating FS in mature β cells and reduction of which could mark the dedifferentiation of β cells. *Claudin 4* deletion compromised glucose tolerance {#feb412735-sec-0019} -------------------------------------------------- We hypothesized that if the dynamic changes of Cldn4 expression were the consequence of β‐cell maturation or dedifferentiation, *Cldn4* deletion will not compromise the islet FS. To test this hypothesis, we analysed the well‐established genetic mouse model in which *Cldn4* was globally removed [23](#feb412735-bib-0023){ref-type="ref"}, diagrammatically depicted in Fig. [3](#feb412735-fig-0003){ref-type="fig"}A. The global knockout model was used because the effect of *Cldn4* in liver, fat and skeletal muscles was negligible as its expression was undetectable in liver and fat and approximately 30‐fold lower than in the islets (Fig. [1](#feb412735-fig-0001){ref-type="fig"}D,E). Both Cldn4^+/−^ and Cldn4^−/−^ mice were born in a normal Mendelian ratio and developed physically indistinguishable from Cldn4^+/+^ mice. The deletion of *Cldn4* was previously confirmed by Southern plot analysis in Cldn4^−/−^ mouse ESCs [23](#feb412735-bib-0023){ref-type="ref"}. We here also showed that the expression of Cldn4 protein was undetectable or dramatically reduced in *Cldn4‐*deleted β cells by immunofluorescence analysis (Fig. [3](#feb412735-fig-0003){ref-type="fig"}B), indicating that *Cldn4* was effectively inactivated in the mutant islets. {#feb412735-fig-0003} We then performed oral glucose tolerance test, as the enteroendocrine cells express a high level of Cldn4 [32](#feb412735-bib-0032){ref-type="ref"} and secrete incretins that are also modulators of glucose homeostasis in Cldn4^+/+^ mice [33](#feb412735-bib-0033){ref-type="ref"}. After overnight fasting, Cldn4^−/−^ mice were administered oral glucose at 2.0 g·kg^−1^ and glucose concentrations measured. The Cldn4^−/−^ mice showed a sex difference in responding to glucose challenge. Cldn4^−/−^ males had modest but significantly (*P* \< 0.05) higher blood glucose concentrations at 30 and 90 min (Fig. [3](#feb412735-fig-0003){ref-type="fig"}C). This was confirmed by a highly significantly (*P* \< 0.01) increase of the area under the curve (AUC) in Cldn4^−/−^ males compared to Cldn4^+/+^ males (Fig. [3](#feb412735-fig-0003){ref-type="fig"}D). Surprisingly Cldn4^−/−^ females were modest but significantly (*P* \< 0.05) more sensitive to glucose metabolism 10 and 20 min after the glucose challenge (Fig. [3](#feb412735-fig-0003){ref-type="fig"}E). This was confirmed by a significantly (*P* \< 0.05) decrease of the area under the curve (AUC) in Cldn4^−/−^ females compared to Cldn4^+/+^ females (Fig. [3](#feb412735-fig-0003){ref-type="fig"}F). Hereafter, we only described phenotypes from male Cldn4^−/−^ mice. We hypothesized, if abnormal blood glucose concentrations in these mice are caused by abnormal function of enteroendocrine cells, that the blood concentrations of key gut hormones will be abnormal due to the deficiency of *Cldn4* in these cells [32](#feb412735-bib-0032){ref-type="ref"}. Nevertheless, the concentrations of glucose‐dependent insulinotropic polypeptide (also known as gastric inhibitory polypeptide, GIP, Fig. [3](#feb412735-fig-0003){ref-type="fig"}G) and glucagon‐like polypeptide‐1 (GLP1, Fig. [3](#feb412735-fig-0003){ref-type="fig"}H) were similar between Cldn4^+/+^ and Cldn4^−/−^ mice, suggesting that the deficiency of *Cldn4* does not affect the secretion of these enteroendocrine hormones. Taken together, these data show that Cldn4 deletion in adult male islets is associated with the glucose intolerance and compromised islet FS observed. *Claudin 4* deletion did not affect islet architecture {#feb412735-sec-0020} ------------------------------------------------------ Cldn4 is a key member of the TJ family molecules, deletion of which in male pancreas may disrupt the islet architecture and lead to an inadequate cellular distribution, that could impair the FS of adult islets [34](#feb412735-bib-0034){ref-type="ref"}, [35](#feb412735-bib-0035){ref-type="ref"}. We thus performed morphological and immunofluorescence analyses in adult male *Cldn4*‐deleted pancreases. Interestingly, pancreas morphology and the islet architecture were similar and apparently unchanged in Cldn4^+/−^ and Cldn4^−/−^ compared to Cldn4^+/+^ mice (Fig. [4](#feb412735-fig-0004){ref-type="fig"}A). Furthermore, the cellular distribution was also apparently similar amongst Cldn4^+/+^, Cldn4^+/−^ and Cldn4^−/−^ islets with β cells at the core and α cells at the mantle (Fig. [4](#feb412735-fig-0004){ref-type="fig"}B). Combining with the observations that Cldn4 is up‐regulated during maturation and down‐regulated during dedifferentiation of β cells, we suggest that *Cldn4* deletion compromised islet FS. Finally, we examined whether basement membrane and mesenchymal genes are also involved in sustaining islet FS. {#feb412735-fig-0004} Basement membrane genes were absent in adult islets {#feb412735-sec-0021} --------------------------------------------------- Encoding important structural proteins surrounding the islet cells, basement membrane genes were also analysed. Bioinformatics analyses showed that several laminin chain genes *Lama1*,*Lamb1* and *Lamc1*, and key linkage genes *Fbln1*,*Fbln2*,*Nid1* and *Nid2* were all negatively enriched in mouse islet cells compared to that of ESCs (Fig. [5](#feb412735-fig-0005){ref-type="fig"}A). Supportively, qRT‐PCR did not detect meaningful expression of *Lama1*,*Lamb1* and *Nid1* in isolated mouse islets (Fig. [5](#feb412735-fig-0005){ref-type="fig"}B). Immunofluorescence analyses verified that laminin and laminin linking molecule nidogen 1 proteins were undetectable around islet cells, though present in a variety of endo‐ and epithelial basement membranes (Fig. [5](#feb412735-fig-0005){ref-type="fig"}C--E). Taken together, no basement membrane molecule was likely to play a role in sustaining islet FS. {#feb412735-fig-0005} Fewer mesenchymal genes expressed in adult islets {#feb412735-sec-0022} ------------------------------------------------- Amongst a selective cohort of 21 mesenchymal genes [36](#feb412735-bib-0036){ref-type="ref"}, the expression of the mesenchymal transcription factor genes *Snail3* and *Twist2* and several others *Col5a1*,*Pmp2* and *Vim* was negatively enriched though *Acta1*,*Dcn*,*Ovol2* and *Pmp22* were positively enriched over twofold in adult functional islets compared to that of ESCs (Fig. [6](#feb412735-fig-0006){ref-type="fig"}A). qRT‐PCR analysis confirmed that the expression of *Acta1* and *Snail1* was significantly higher in islets than in the exocrine tissue (Fig. [6](#feb412735-fig-0006){ref-type="fig"}B). Collectively, the modest enrichment of several mesenchymal genes in the adult islets did not clearly support the notion that these genes play a role in regulating islet FS. {#feb412735-fig-0006} Discussion {#feb412735-sec-0023} ========== We provided multiple pieces of evidence for the first time that the highly expressed TJ molecule Cldn4 may be involved in regulating the FS of the pancreatic insulin‐secreting β cells with implications in translational research for better diabetes therapies. First, we showed that *Cldn4* is mostly up‐regulated during which differentiated β cells are functionally maturated. Second, the expression of *Cldn4* is down‐regulated when β cells are functionally compromised and undergo dedifferentiation. Third, the expression of Cldn4 is also down‐regulated when type 2 diabetic db/db islets have a overtly impaired FS. Fourth, a modest but significantly impaired FS is detected when *Cldn4* is genetically deleted in mice without clearly disrupting islet architecture and cellular distributions. Finally, the impaired FS in *Cldn4*‐deleted mice was apparently not associated with the incretin metabolism as GIP and GLP1 plasma concentrations were unaffected. The absence of meaningful expression of basement membrane genes in purified adult mouse islets supports our previous report [2](#feb412735-bib-0002){ref-type="ref"}. In and around mature pancreatic islets, the observed laminin is located at the endothelial basement membrane, consistent with our previous report [2](#feb412735-bib-0002){ref-type="ref"} and produced from endothelial cells and fibroblasts [37](#feb412735-bib-0037){ref-type="ref"}. In summary, the above data collectively suggest that the developmental up‐regulation of Cldn4 involves in islet FS, whereas pathological down‐regulation or genetic deletion of Cldn4 compromises it (Fig. [7](#feb412735-fig-0007){ref-type="fig"}). {#feb412735-fig-0007} However, unlike the kidney‐collecting duct [20](#feb412735-bib-0020){ref-type="ref"}, islet Cldn4 does not structurally partner with its typical partners, *Cldn8* and *Cldn12* [9](#feb412735-bib-0009){ref-type="ref"}, as the latter are undetectable in the functional endocrine pancreas. We speculate that islet Cldn4 also acts as selective and critical physiological ion channels [9](#feb412735-bib-0009){ref-type="ref"}. As reported [38](#feb412735-bib-0038){ref-type="ref"}, β‐cell Cldn4 may interact with the Cldn4 of adjacent β cells for a coordinated signalling. TJs can receive and convert signals from the cell interior to regulate junction assembly and transduce signals to the cell interior to regulate gene expression and cell response [39](#feb412735-bib-0039){ref-type="ref"}. A previous study demonstrated that serum Ca^+2^ concentration in the *Cldn4* null mice was significantly decreased, potentially due to the increased excretion of Ca^+2^ and Cl^−^ in the urine [23](#feb412735-bib-0023){ref-type="ref"}. If a similar Ca^+2^ metabolic disorder compromises β‐cell function, glucose intolerance should occur in both sexes of Cldn4^−/−^ mice, but in this case, glucose intolerance was only detected in males. Future definitive experiments including pancreas‐ or β‐cell‐specific deletion of *Cldn4* are required to confirm or refute the observation that Cldn4 involves in regulating the islet FS. Our study suggests a possibility that the highly up‐regulated TJ Cldn4 molecule works as a maturation biomarker of postnatal insulin‐secreting β cells. Using the biomarker, the fully matured insulin‐secreting cells given rise from pluripotent stem cells would be enriched for a regenerative therapy to high‐risk type 1 diabetic sufferers [40](#feb412735-bib-0040){ref-type="ref"}. A monoclonal antibody targeted the extracellular loop of Cldn4 has indeed enabled the enrichment of mouse enteroendocrine cells [32](#feb412735-bib-0032){ref-type="ref"}. The β‐cell hormone urocortin 3 [41](#feb412735-bib-0041){ref-type="ref"} has been demonstrated to be a β‐cell maturation marker [42](#feb412735-bib-0042){ref-type="ref"} but is difficult to be utilized for the enrichment of matured insulin‐secreting cells. Data presented point to the possibility that the declining expression of Cldn4 works as a novel biomarker of β‐cell dedifferentiation. β‐cell dedifferentiation has been demonstrated to play a critical role in the development of mouse [31](#feb412735-bib-0031){ref-type="ref"} and human [43](#feb412735-bib-0043){ref-type="ref"} type 2 diabetes, which affects 425 million people worldwide. Identification of such biomarkers would facilitate the investigations of molecular mechanisms of β‐cell dedifferentiation and of therapeutic approaches of the dedifferentiation prevention and of redifferentiation. Dedifferentiated β cells in diabetes have abnormally expressed the mitochondrial enzyme aldehyde dehydrogenase 1 isoform 3A [44](#feb412735-bib-0044){ref-type="ref"} or re‐expressed the fetal islet hormone gastrin [45](#feb412735-bib-0045){ref-type="ref"}. β‐cell dedifferentiation can be induced by the genetic deletion of *FoxO1* [31](#feb412735-bib-0031){ref-type="ref"} or *Pax6* [46](#feb412735-bib-0046){ref-type="ref"}, [47](#feb412735-bib-0047){ref-type="ref"} transcription factor genes. We here showed that when β cells undergo dedifferentiation, *Cldn4* expression is down‐regulated, associated with activation of several islet progenitor transcription factor genes including *Fev* and *Isl1*. Dedifferentiated β cells also progressively lose protein content but increase mRNA of the nuclear receptor Vdr expression and treatment with Vdr agonists is able to prevent β‐cell dedifferentiation [48](#feb412735-bib-0048){ref-type="ref"}. Research is underway to understand how the declined expression of functional transcription factor genes such as *Hnf4a* and *Pax4* and/or the increased expression of progenitor transcription factor genes *Fev* and *Isl1* contribute to the decreasing expression of Cldn4 in dedifferentiated β cells. We noted that *Cldn3* was down‐regulated approximately 50% in db/db islets, but significantly up‐regulated in late dedifferentiating MIN6 cells. Its significance on β‐cell dedifferentiation requires further studies. In summary, our study suggested that the previously unappreciated TJ molecule Cldn4 is involved in regulating islet FS in adult ß cells and may act as a biomarker of β‐cell maturation. This may be of significance for translational research in establishing stem cell therapy for the diabetes sufferers [49](#feb412735-bib-0049){ref-type="ref"}. This report also suggested that a reduction in Cldn4 expression in ß cells is associated with their dedifferentiation. Biomarkers of β‐cell dedifferentiation would impact the translational research for redifferentation therapies of the pandemic type 2 diabetes. Conflict of interest {#feb412735-sec-0025} ==================== The authors declare no conflict of interest. Author contributions {#feb412735-sec-0026} ==================== F‐XJ conceived, designed and performed the research and wrote the manuscript; HL, ANJ and TN performed the research and YH partially designed the research and reviewed the manuscript. The authors would like to thank Ms Caroline Rudnicka for technical assistance. This study was supported in part by the Juvenile Diabetes Research Foundation (4‐2006‐1025), Diabetes Australia Research Trust, Diabetes Research Foundation of Western Australia, Medical Research Foundation of Royal Perth Hospital and Telethon Perth Children\'s Hospital Research Fund (TPCHRF) Grant (to F‐XJ).

{#sp1 .23} {#sp2 .24} {#sp3 .25} {#sp4 .26} {#sp5 .27} {#sp6 .28} {#sp7 .29} {#sp8 .30} {#sp9 .31} {#sp10 .32} {#sp11 .33} {#f1 .26} {#f2 .26} {#f3 .27} {#f4 .27} [^1]: Read at a Meeting of the Bristol Medico-Chirurgical Society, January 12th, 1910.

Intraosseous migration of unerupted tooth is a rare and unusual dental anomaly which occurs mostly in the permanent dentition of the lower jaw. When the tooth crosses the midline, this rare phenomenon is known as dental transmigration and it occurs almost exclusively with mandibular canines with an incidence of 0.34%.[@B1] The mandibular permanent lateral incisor is the most frequently displaced tooth which usually migrates distally in the bone, and it crosses the midline very rarely.[@B2] Camilleri[@B2] reported the first case of lower lateral incisor transmigration coupled with lower canine transmigration in 2007. Although maxillary canine impaction is a well-known dental anomaly, there were a few reports of cases which showed migration across the mid-palatal suture to the opposite side of the maxilla recently.[@B1],[@B3]-[@B8] The etiology and exact mechanism of the transmigration is still unclear. Transmigrating teeth can cause pressure resorption of roots or tilting of the adjacent teeth and may migrate to adjacent structures like the coronoid process, causing pain and discomfort to the patient.[@B9] Since impacted or transmigrated teeth are important especially in terms of orthodontic treatment planning, these teeth must be diagnosed clinically and radiographically. Early diagnosis with a timely orthodontic or surgical intervention can help dentists preserve the canines, which play an important role, in both esthetics and function in the human dentition. Due to unfavorable position of the transmigrated canine, orthodontic repositioning is difficult. Until now, successful corrections of transmigrated canines using orthodontic treatment have rarely been documented in the literatures.[@B10],[@B11] To the best of our knowledge, maxillary lateral incisor crossing the mid-palatal suture has never been reported in the literature. Multiple transmigrations are even rarer. This report demonstrates a case of simultaneous transmigration of a lateral incisor and canine in the maxilla. It also reports four unusual cases of unilateral canine transmigration in the maxilla and mandible and successful eruption of one of the mandibular transmigrated canines following orthodontic traction. Etiology of transmigration and its clinical considerations are also discussed. Case Report =========== Case 1 ------ A 14-year-old female reported to the department with a complaint of crowding of the upper teeth. Intraoral examination showed the absence of the maxillary right permanent lateral incisor and canine. A maxillary occlusal radiograph revealed double transmigration of both upper right lateral incisor and canine which were impacted ([Fig. 1](#F1){ref-type="fig"}). In order to evaluate the extent of the transmigration process, axial computed tomographic (CT) images were acquired from the maxillofacial region along with three dimensional reconstructions using CT equipment (Brilliance CT 64-channel scanner, Philips Healthcare, Andover, MA, USA). CT images revealed that the maxillary right lateral incisor was impacted horizontally and was positioned on the palatal side. Its crown had migrated across the mid palatine suture with its cusp tip against the middle third of the root of the upper left central incisor. The maxillary right canine was almost impacted horizontally and transmigrated across the mid palatine suture to the left side, with its crown near the root apex of the left central incisor. CT images showed pericoronal radiolucency in relation to the upper right canine suggestive of a dentigerous cyst, however no associated pathology was found in relation to the transmigrated upper right lateral incisor ([Fig. 2](#F2){ref-type="fig"}). Due to the unfavorable position of the transmigrated teeth and associated pathology, surgical removal was decided. Extractions were performed under general anesthesia ([Fig. 3](#F3){ref-type="fig"}). Healing was uneventful and a follow-up radiograph 8 months after the surgery showed no further abnormality ([Fig. 4](#F4){ref-type="fig"}). Case 2 ------ A 24-year-old female reported to the department with a chief complaint of spacing between the upper right lateral incisor and first premolar. Intraoral examination showed that the upper right permanent canine and all the third molars were absent. Panoramic radiograph revealed a transmigrated upper right canine which was impacted almost horizontally at the level of the apical third of the roots of the upper central incisors. The crown migrated across the mid-palatine suture with its cusp tip against the root of the upper left central incisor. Transmigration was also confirmed on the maxillary occlusal radiograph ([Fig. 5](#F5){ref-type="fig"}). There was no pathologic finding associated with the tooth. Due to the unfavorable position of the canine, surgical removal of the canine was decided. However, the patient declined the surgical intervention. The patient was informed of the consequences and radiographic monitoring was strongly recommended. Case 3 ------ A 22-year-old male patient was referred to our department for an orthodontic consultation. Intraoral examination showed the absence of the permanent canine and lateral incisor in the upper right quadrant. The lower left permanent canine was also absent and the deciduous canine was retained in the lower left quadrant ([Fig. 6A](#F6){ref-type="fig"}). As the patient did not report any history of extraction or traumatic avulsion of teeth, further investigation was done by panoramic, occlusal, and periapical radiographs. Radiographic examination revealed the presence of a migrated lower left canine with its crown located below the apices of the right central incisor. The upper right canine was impacted vertically and the lateral incisor was congenitally absent ([Fig. 6B](#F6){ref-type="fig"}). As both impacted canines were in favorable position, surgical exposure, followed by orthodontic traction was started to bring the canines into the arch. One year later, the upper right canine was well placed into the arch and transmigrated canine erupted in the oral cavity ([Fig. 7](#F7){ref-type="fig"}). A prosthetic implant would be placed for the missing lateral incisor after the completion of orthodontic treatment. Case 4 ------ A 20-year-old male reported to a dental clinic for a routine dental check up. Oral examination revealed a retained lower deciduous canine on the left side. On further examination, panoramic radiograph showed a horizontally impacted lower left permanent canine. It was transmigrated across the midline to the right side with its crown position near the inferior border of the mandible below the root apices of the right second premolar and permanent first molar ([Fig. 8](#F8){ref-type="fig"}). As there was no pathology associated with the tooth, periodic radiographic follow up was decided. Case 5 ------ A 22-year-od female patient was referred to our department for orthodontic consultation. Intraoral examination showed a missing permanent left lower canine. Panoramic radiograph examination revealed a migrated lower left canine with its crown located below the apices of the right premolars ([Fig. 9](#F9){ref-type="fig"}). Enlarged follicular space was found around the migrated canine. As there was no associated symptoms, periodic radiographic follow up was decided. Discussion ========== Tarsitano et al[@B12] defined transmigration as a phenomenon in which an unerupted mandibular canine migrates, crossing the mandibular midline. Javid[@B13] expanded the definition to include the cases in which more than half of the tooth had passed through the midline. Joshi[@B9] and Auluck et al[@B14] suggested that the tendency of a canine to cross the midline suture is a more important consideration than the actual distance of migration after crossing the midline. Moreover, it will also depend on the stage of transmigration when the orthodontist or oral surgeon first finds the patient. Therefore, we adapted their criteria to define the transmigration in our study. Maxillary canine transmigration in Indian population has never been reported in the literature. Here we have presented two cases of maxillary transmigration from Indian people. Case 1 presented here showed many rare features. This unusual phenomenon has mostly been described and documented for the mandibular canine. The first published case of mandibular lateral incisor transmigration was reported in 2007,[@B2] however a case of maxillary incisor transmigration has been never reported in the literature. Transmigration is rarely found in the maxillary arch. Aydin and Yilmaz[@B3] reported the first case of maxillary canine transmigration in the literature in 2003. A chronological resume of the clinical features of the transmigratory maxillary canine, as observed by various earlier authors, is given in [Table 1](#T1){ref-type="table"}. In all of the previously published cases, isolated maxillary canine transmigration was reported. The case 1 was the first report where unilateral maxillary canine transmigration was seen along with the adjacent lateral incisor transmigration. The larger cross-sectional area of the anterior mandible compared with the anterior maxilla may be a reason for the higher frequency of mandibular canine transmigration. In the maxilla, transmigration of canines might be prevented due to the shorter distance between the roots of maxillary incisors and the floor of the nasal fossa and restriction of the path of tooth movement by the roots of adjacent teeth, the maxillary sinus and the mid-palatal suture, which act as a barrier.[@B6] In 2002, Mupparapu[@B15] proposed a classification for transmigrated mandibular canines according to their migratory pattern and position in the jaw and classified these teeth into five groups. This classification can be summarized as follows: Type 1, canine is impacted mesioangularly across the midline, labial or lingual to the anterior teeth, with the crown portion of the tooth crossing the midline. Type 2, canine is horizontally impacted near the inferior border of the mandible below the apices of the incisors. Type 3, canine has erupted either mesial or distal to the opposite canine. Type 4, canine is horizontally impacted near the border of the mandible below the apices of either premolars or molars on the opposite side. Type 5, canine is positioned vertically in the midline with the long axis of the tooth crossing the midline. The most common transmigratory pattern reported in the literature was type I and type 5 being the least common. The radiographic appearance of the transmigrated canines in Case 4 and 5 was in accordance with type 4, whereas type 1 transmigratory pattern was seen in Case 3. Transmigrated maxillary canines could not be classified as there was no classification system proposed until now in the literature. The etiology and exact mechanism of transmigration is still not clear. Although a number of factors have been suggested, abnormal displacement of the dental lamina in embryonic life is a commonly accepted explanation of the cause of the displacement and non-eruption of such canines.[@B9] Marks and Schroeder[@B16] suggested that a regional disturbance in the dental follicle might lead to local defective osteoclastic function with an abnormal eruption pathway being formed. Bruzst[@B17] believed that the canine germ was situated in front of the lower incisors and that facial growth pushed it towards the contralateral side while other authors[@B18] believed that an abnormally strong eruption force or a change which affected the crypt of the tooth germ might lead to erroneous eruption. According to Pippi and Kaitsas,[@B19] a strong and extended eruptive force caused by lasting root formation, and a pericoronal osteolytic area caused by widening of the follicular space played an important role in intraosseous migration of canines. The authors proposed that following impaction, these two factors could cause the anomalous movement of the canine. The osteolytic area is believed to represent a site of less resistance, toward which the tooth moves while root formation takes place. Hence, the simultaneous occurrence of these two events is believed to cause the intraosseous movement of the canine (in the opposite direction to the root) until a mechanical obstacle is reached (such as the cortical bone), or when the pericoronal osteolytic area dissipates. Other local or pathologic factors suggested in the aetiology of transmigration, are premature loss of primary teeth and unavoidable occupation of the space by adjacent teeth, retention of the deciduous canine, discrepancies of tooth size, unfavorable alveolar length, genetics, trauma, tumors, odontomas, cysts and even a very small obstacle, such as a small root fragment, that would be sufficient to divert a tooth from its normal path of eruption.[@B9] According to Auluck and Mupparapu,[@B20] the angulation of the migrating maxillary canine relative to the mid-palatal suture might be an important factor in determining their transmigration. Impacted maxillary canines require a large amount of force to overcome the strong barrier, the mid-palatal suture, in their path. When the maxillary canines are positioned perpendicular to the mid-palatal suture, they might have enough horizontal component of eruptive force to migrate to the contralateral side. However, when they are positioned at an axiocoronal angulation ie, 45° to 90° the eruptive force will have both vertical and horizontal (angular) components. The horizontal component of the eruptive force of such an impacted maxillary canine might not be sufficient to overcome the resistance of the mid-palatal suture, and thus these teeth abruptly abandon their journey at the midline. In the cases presented here, all the maxillary transmigrated teeth were positioned almost perpendicular to the midline, facilitating their transmigration, except in case 2 where the canine was transmigrated to contralateral side with an axio-coronal angulation of less than 90 degree ([Fig. 5](#F5){ref-type="fig"}). Canine just migrates without any pathological entity mostly, however in a few cases a cyst or odontoma accompanies such a tooth. In case 1 presented here, the transmigrated canine was associated with a dentigerous cyst. Canine transmigration is of significant importance in dentistry, creating orthodontic, surgical, and interceptive problems. These cases have been increasing day by day. It is important to diagnose them in earlier stages of migration or abnormality to prevent more complicated situations. Vichi and Franchi[@B21] suggested that, proclination of the lower incisors, increased axial inclination of the unerupted canine, and an enlarged symphyseal cross-sectional area of the chin may be favorable conditions for the transmigration. All of these features found in 8- to 9-year-old patient demand periodic panoramic examination to rule out transmigration. Surgical removal, transplantation, radiographic follow-up, and surgical exposure with orthodontic treatment are suggested as treatment options for transmigrated canines.[@B7] Orthodontic correction of impacted transmigrated canines is difficult due to unfavorable position of the impacted tooth. Until now, successful corrections of transmigrated canines using orthodontic treatment have been rarely documented in the literature.[@B10],[@B11] In Case 3, the impacted transmigrated canine was erupted successfully after orthodontic traction and its alignment is still in progress ([Fig. 7](#F7){ref-type="fig"}). If surgical removal is considered as the choice of treatment, it should be kept in mind that, although the teeth have transmigrated to the other side of the midline, they still maintain their nerve supply from the originating side.[@B9] This is an important factor in planning any surgical procedure under local anesthesia either to expose and bond or extract the transmigrated tooth. In conclusion, the mechanism of transmigrated maxillary canines passing to the contralateral side of the mid-palatal suture needs to be further studied. Clinicians should be encouraged to report more cases of transmigrated maxillary canines to better understand their mechanism of eruption which will aid in the diagnosis, prevention and treatment of these occurrences. {#F1} {#F2} {#F3} {#F4} {#F5} {#F6} {#F7} {#F8} {#F9} ###### Studies related to maxillary canine transmigration in chronological order  L: left, R: right, F: female, M: male, I: impacted, CR: retained primary canine, CE: exfoliated primary canine, Uni/Bi: unilateral/bilateral, NA: Not available.

Introduction ============ *Streptococcus pneumonia* is commonly found in nasopharyngeal mucosa of healthy people and can cause various diseases, such as pneumonia and meningitis.[@b3-idr-11-1043] The bacterium is a relatively uncommon cause of septic arthritis, including pyogenic sacroiliitis, which is a rare manifestation of invasive pneumococcal diseases (IPD).[@b1-idr-11-1043] The prognosis of IPD is influenced by both bacterial factors, such as serotype, and host characteristics, such as age, chronic cardiovascular and pulmonary diseases, diabetes mellitus, alcoholism, smoking, and immunosuppressive states.[@b2-idr-11-1043],[@b3-idr-11-1043] Because serotype 3 *S. pneumonia* have a thick capsule, it is thought that it is highly pathogenic.[@b4-idr-11-1043] In this report, we describe a rare case of pyogenic sacroiliitis and osteomyelitis with a right iliac muscle abscess caused by serotype 3 *S. pneumonia* in a healthy adult. Case report =========== A previously healthy 54-year-old Japanese man was admitted to our hospital with a 3-week history of worsening right-side thigh pain and a 3-day history of fever. Four weeks prior to admission, the patient had a mild cough with sputum. His symptoms improved without any treatment. Three weeks prior to admission, the patient developed mild right thigh pain without trauma that did not interfere with his daily life. One week prior to admission, he came to a local clinic because of his right thigh pain, and loxoprofen was prescribed. His symptoms transiently improved with loxoprofen, but 3 days prior to admission, the patient had chills and his thigh pain worsened. He returned to the local clinic, and amoxicillin/clavulanate was prescribed. On the day of admission, he visited another community hospital and was transferred to our hospital on the diagnosis of right iliac muscle abscess. He denied sore throat, cough, dyspnea, and chest pain. His past history was not remarkable. The patient had no previous medical history, no use of tobacco or illicit drug, and rarely drink alcohol. He denied recent animal exposure, recent travel, or any history of allergies. On physical examination, his blood pressure was 140/60 mmHg, pulse rate was 66 beats per minute, temperature was 38.0°C, and respiratory rate was 20 breaths per minute with oxygen saturation of 97% on room air. The results of the physical examination were unremarkable, except for right buttock tenderness and positive right psoas sign. Laboratory data obtained on admission revealed a white blood cell count of 19,070 μL with 94% neutrophils. Examination of the serum chemistry revealed the following results: blood urea nitrogen 23.9 mg/dL, creatinine 1.6 mg/dL, albumin 2.2 g/dL, total protein 6.6 g/dL, and C-reactive protein 26.0 mg/dL. Based on these findings, the patient was suspected of having right psoas abscess. CT of the pelvic area showed bone destruction of the right sacroiliac joint with perifocal abscess formation in the right iliac muscle ([Figure 1A](#f1-idr-11-1043){ref-type="fig"}). A diagnostic CT-guided percutaneous drainage of the abscess was performed, and Gram staining of the pus revealed Gram-positive diplococcus bacteria and polymorphonuclear leukocytes ([Figure 2](#f2-idr-11-1043){ref-type="fig"}). After an initial work up in the emergency room, his primary diagnosis was bacterial sacroiliitis and osteomyelitis with right iliac muscle abscess. Transthoracic echocardiography was performed, and the results showed no evidence of an infection. The patient was started on intravenous cefmetazole at a dose of 2 g every 12 hours. No bacterium was detected from his blood culture, but the pus culture grew *S. pneumoniae*. Additional examination revealed the following results: 90 mm/h erythrocyte sedimentation rate, 56 mg/dL IgM (within normal range 33--183 mg/dL), 1450 mg/dL IgG (within normal range 861--1,747 mg/dL), 148 mg/dL C3 (within normal range 73--138 mg/dL), 40 mg/dL C4 (within normal range 11--31 mg/dL), and 70 mg/dL CH50 (within normal range 25--48 mg/dL). Human immunodeficiency virus antibody, M protein, and Bence Jones protein were negative. CT of the abdomen showed no evidence of any abdominal finding including spleen. Serum protein fraction came back with an acute inflammation pattern without M protein. Mucoid *S. pneumonia* was obtained by culturing of abscess sample. Serotyping of *S. pneumonia* was performed by the capsular swelling method using pneumococcal antisera (Statens Serum Institute, Copenhagen, Denmark) and the stain was determined to be serotype 3. Susceptibility of *S. pneumonia* to antibiotics was analyzed using the broth microdilution method according to a protocol of the Clinical and Laboratory Standards Institute. Minimal inhibitory concentration breakpoints were defined according to the Clinical and Laboratory Standards Institute (M100) criteria. The *S. pneumonia* strain exhibited susceptibility to penicillin, ampicillin, ceftriaxone, meropenem, tosufloxacin, vancomycin, and trimethoprim/sulfamethoxazole, but was resistant to clindamycin ([Table 1](#t1-idr-11-1043){ref-type="table"}). On day 4, magnetic resonance imaging of the pelvic area revealed increased signal intensity of the right sacroiliac joint and surrounding bone, and diffuse thickening and effusion surrounding the muscle ([Figure 1B](#f1-idr-11-1043){ref-type="fig"}). On day 5, we attempted to perform de-escalation. The patient was switched to treatment with ampicillin at a dose of 2 g every 6 hours on the basis of the susceptibility test because his renal function had improved. He was given levofloxacin 500 mg orally because of the drug eruption due to ampicillin on day 25. The patient was successfully treated with an 8-week course of antibiotics. No evidence of a relapse of the infection was noted at the 6-month follow-up. Ethics statement ================ Written informed consent has been obtained from the patient to obtain their results and to publish the case details with accompanying images. Discussion ========== Pyogenic sacroiliitis is a rare infectious diseases, representing 1%--2% of all cases of septic arthritis, most frequently in children and young adults.[@b5-idr-11-1043] The risk factors for pyogenic sacroiliitis include being intravenous drug user, having infections in other organs systems, such as urinary tract infection, and pregnancy.[@b6-idr-11-1043] According to previous reports, adult patients are more likely to have atypical presentation, concurrent infections, and local complication.[@b6-idr-11-1043],[@b7-idr-11-1043] Delay in the diagnosis can lead to additional complication such as osteomyelitis, abscess formation, and bacteremia. Based on previous studies, bacteremia is documented in 71% of adults with pneumococcal septic arthritis.[@b1-idr-11-1043] In this case, we believe that the cause of the sacroiliitis, osteomyelitis, and iliopsoas abscess was the secondary infection as a result of primary bacteremia. Although the results of the blood cultures were negative, this patient had chills, and amoxicillin/clavulanate were prescribed before admission to our hospital. Treatment options include antibiotic therapy alone, CT-guided drainage, or surgical drainage. The duration of antimicrobial treatment is at least 4--8 weeks.[@b6-idr-11-1043] *S. pneumonia* infrequently causes bacterial sacroiliitis. A review of pneumococcal septic arthritis showed 2.8% of adult patients with pneumococcal septic arthritis have sacroiliac involvement.[@b7-idr-11-1043] In addition, three cases of pneumococcal sacroiliitis in adult have been reported in the literature ([Table 2](#t2-idr-11-1043){ref-type="table"}).[@b8-idr-11-1043]--[@b10-idr-11-1043] These cases include a 31-year-old woman who presented with a fever, depressed level of consciousness, and right buttock and thigh pain, which was later proven to be caused due to *S. pneumonia* by blood cultures;[@b8-idr-11-1043] a 47-year-old woman who presented with a 4-week history of left-sided lower back pain which was proven to be caused due to *S. pneumonia* by CT-guided drainage;[@b9-idr-11-1043] and a 62-year-old who presented with a 3-day history of right buttock pain which was proven to be caused due to *S. pneumonia* by blood cultures, muscle biopsy, and surgical drainage.[@b10-idr-11-1043] In the present case, the patient did not have an immunodeficiency or chronic illness. Interestingly, all of the patients had no risk factor for IPD. *S. pneumonia* is covered by a capsule, which is a very important factor for its pathogenicity. In the case presented herein, serotype 3 *S. pneumonia* was isolated from the lesion and determined as the causative pathogen. Serotype 3 *S. pneumonia* strains always have a characteristically thicker capsule, show mucoid formation on blood agar plates, possess greater virulence, and have a higher mortality rate when compared to other serotype strains.[@b3-idr-11-1043],[@b4-idr-11-1043],[@b11-idr-11-1043] In the present case, the treatment duration was 8 weeks with no evidence of a relapse of the infection during the 6-month follow-up. We believe an adequate treatment duration leads to successful treatment, even in the case of serotype 3 infection. Conclusion ========== In conclusion, we reported a case of serotype 3 pneumococcal sacroiliitis and osteomyelitis with right iliac muscle abscess in a healthy adult. It is important to keep in mind that bacterial sacroiliitis and osteomyelitis due to *S. pneumonia* are rare, but can occur even in a previously healthy adult. Delay in the diagnosis can lead to additional complication such as osteomyelitis, abscess formation, and bacteremia. **Disclosure** The authors report no conflicts of interest in this work. {#f1-idr-11-1043} {#f2-idr-11-1043} ###### Susceptibility testing of the isolated *Streptococcus pneumonia* Antimicrobial agent MIC (µg/mL) ------------------------------- ------------- Penicillin 0.03 Ampicillin 0.06 Ceftriaxone 0.25 Meropenem 0.015 Erythromycin 4 Tosufloxacin ≤0.12 Clindamycin ≥8 Vancomycin 0.5 Sulfamethoxazole/trimethoprim ≤0.25 **Abbreviation:** MIC, minimal inhibitory concentration. ###### Clinical profiles of the patients with pneumococcal sacroiliitis --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Case number Age (yrs) Gender Risk factors HIV antibody Chief complaints Clinical diagnosis Clinical samples Bacteremia Drainage Antibiotic treatment Treatment duration (weeks) Outcome Year Reference ------------- ----------- -------- -------------- -------------- ------------------------------------ ------------------------------------------ -------------------------------------------- ------------ -------------- ------------------------------ ---------------------------- --------- ------ -------------------- 1 31 Female None Negative Fever Disturbance of consciousness Bacterial meningitis Sacroiliitis Blood culture Cerebrospina fluid l Positive None Penicillin for 42 days\ 8 Cure 1997 [@b8-idr-11-1043] Oral penicillin for 14\ days 2 47 Female None Unknown Left-sided lower back pain Sacroiliitis Pneumonia Pus culture of percutaneous drainage Unknown Percutaneous Benzylpenicillin for 5 days\ 12 Cure 1998 [@b9-idr-11-1043] Oral amoxycillin for 79 days 3 62 Male None Unknown Right buttock pain, fever Sacroiliitis Osteomyelitis Pyomyositis Blood culture Culture of the muscle biopsy Positive Surgical Ceftriaxone 7 days\ 6 Cure 2001 [@b10-idr-11-1043] Cephalexin 35 days 4 54 Male None Negative Fever, chills, thigh pain Sacroiliitis Osteomyelitis Psoas abscess Pus culture of percutaneous drainage Negative Percutaneous Cefmetazole IV 4 days\ 8 Cure 2017 Presentcase Ampicillin IV 21 days\ Oral levofloxacin 31 days --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- **Abbreviations:** HIV, human immunodeficiency virus; IV, intravenous; yrs, years.

Background {#Sec1} ========== Hypertrophic cardiomyopathy (HCM) is estimated to affect 0.2 % of the population \[[@CR1], [@CR2]\]. As the most common inherited cardiac disease, HCM is characterized by asymmetrical septal hypertrophy \[[@CR3], [@CR4]\]. The phenotypes are ranging from minor to severe life-threatening status \[[@CR5]\]. Left ventricular outflow tract (LVOT) obstruction is responsible for disabling symptoms in a large proportion of patients and confers a worse prognosis. HCM is the most common cause of sudden cardiac death (SCD) in people aged less than 35 years. In fact, it is more common for HCM patients to develop to severe progressive heart failure than SCD. Septal myectomy is considered as the gold standard for septal reduction therapy in patients with dynamic LVOT obstruction \[[@CR6]\]. The long-term mortality for hypertrophic obstructive cardiomyopathy (HOCM) patients with septal myectomy reduction therapy decreases to 1.4--1.8 % \[[@CR7], [@CR8]\]. Septal myectomy has contributed to the overall reduction in mortality and provided survival equivalent to that of the general population \[[@CR9]\]. The operative technique of HOCM has undergone evolution from classic Morrow surgery to extensive resection. Extensive septal myectomy surgery has evolved into a comprehensive repair technique for LVOT obstruction and abnormality of the mitral apparatus \[[@CR10]\]. Extension of the myectomy area leads to improved surgical effect. Currently, there are limited data addressing long-term outcome of extensive septal myectomy in Chinese patients with HOCM. In the present study, we sought to report the early and late clinical and echocardiographic outcomes of Chinese HOCM patients who underwent extensive septal myectomy surgery. Methods {#Sec2} ======= HOCM patients {#Sec3} ------------- This study was carried out in 139 consecutive HOCM patients treated with extensive surgical myectomy in Cangzhou Central hospital and Shandong Chinese Medical Hospital from Jan 1, 2008 to May 31, 2011. Patients were eligible for inclusion in surgical intervention if they met the following two criteria \[[@CR11]\]: 1) LVOT pressure gradient ≥50 mmHg at rest or with physiological provocation by transthoracic echocardiography; 2) presence of severe symptoms despite prior appropriate medical therapy with beta-receptor blocker and calcium channel blocker. All patients were evaluated by consultant cardiologists. Although they were given medical therapy, they still had severe symptoms attributable to LVOT obstruction. The review of hospital records and analyses of preoperative data, operative reports, postoperative and follow-up echocardiography were carried out. In this study, perioperative period was defined as time period from surgical preparation to 30 days after extensive septal myectomy. The study had full approval from the Ethics Boards of Cangzhou Central hospital and Shandong Chinese Medical Hospital. Surgical procedure {#Sec4} ------------------ Operations were carried out under mild hypothermic cardiopulmonary bypass with total anesthesia as previously described \[[@CR12], [@CR13]\]. The heart and ascending aorta were exposed by a longitudinal median incision in sternum. Cannulas were inserted into superior vena cava, inferior vena cava and ascending aorta to establish cardiopulmonary bypass. Through a transverse aortotomy, the aortic valve was exposed and cardiac asystole was induced by cold blood antegrade cardioplegia. Hypertrophic septum and mitral valve were exposed sufficiently by pulling right coronary aortic valve. A Ross retractor was used to display the muscular septum. The superior borderline of septal resection located at 3 mm below right coronary sinus valve. Resection started from middle of right aortic sinus and moved 10--12 mm horizontally toward commissure of left sinus valve and right sinus valve. Longitudinal resection usually reached the root of mitral papillary muscle, length of which ranged from 45 to 50 mm. In order to reduce LVOT gradient, the thicknesses of the left ventricular wall and interventricular septum need to be nearly normal by visual inspection. The following extended procedures were performed as previously described \[[@CR14], [@CR15]\]: 1) The resection was continued toward the mitral valve annulus and apically to the bases of the papillary muscles. 2) All areas of papillary muscle fusion to the septum or ventricular free wall are divided, and anomalous chordal structures and fibrous attachments of the mitral leaflets to the ventricular septum are excised. 3) Plication of the anterior mitral leaflet (AML) was performed if there were indications by selection criteria including floppy and lax anterior leaflet, AML ≥ 3.0 cm, mitral regurgitation ≥ 2+, systolic anterior motion (SAM), and/or absence of rheumatic or other intrinsic mitral valve disease. Intraoperative transesophageal echocardiography (TEE) was used routinely. TEE assessment was performed after weaning from cardiopulmonary bypass to evaluate adequacy of LVOT and mitral valve function. Postoperative monitor was carried out to maintain proper cardiac preload and appropriate colloid infusion. Beta-receptor blocker or calcium-channel blocker was applied routinely. Low-dose vasoactive drugs were used when hemodynamic circulation fluctuated. Echocardiography {#Sec5} ---------------- During hospital stay, preoperative transthoracic echocardiography was performed for HOCM patients and intraoperative TEE was applied routinely. The adequacy of the resection and the LVOT gradient were assessed immediately by transthoracic echocardiography after surgery and transthoracic echocardiography was repeated before or on the day of hospital discharge. During outpatient follow-up period, patients had transthoracic echocardiography detection at every visit. A careful analysis of septal hypertrophy, abnormalities of the mitral valve, and subvalvular apparatus was performed. LVOT gradient was detected at rest and with provocation by transthoracic echocardiography. Specific details of mitral valve length, SAM, and mitral regurgitation were evaluated. Images were captured on cine loops at the time of the detection. M-mode from long axis view was used to measure interventricualr septal (IVS) thickness and diastole left ventricular inner diameter (LVID~d~). M-mode from short axis view was used to measure thickness of left ventricular wall. Ejection fraction (EF) was automatically calculated by the measurement package. Follow-up {#Sec6} --------- The follow-up period averaged 5.6 ± 0.9 years for HOCM patients. Follow-up data were derived from medical charts and LVOT gradient information was obtained from the record of transthoracic echocardiography detection at every outpatient visit. Our primary endpoints included all-cause death and cardiac death during follow-up. Cardiac death was defined as a death resulting from heart failure or SCD. SCD was defined as an abrupt loss of consciousness within 1 h after the onset of acute symptoms, and the cause of death could not be attributed in the postmortem examination. All patients were assigned New York Heart Association (NYHA) classification based on symptoms. Statistical analysis {#Sec7} -------------------- Continuous variables were expressed as mean ± SD otherwise described as proportion. *χ* ^2^ test was used to compare categorical variables. For continuous variables, one-way ANOVA was used to analyze the data among three groups and then paired *t* test was performed for the comparison between two groups. The Kaplan-Meier method was used to draw survival curve and calculate survival rate and cumulate hazard. Difference was considered statistically significant when *P* \< 0.05. All statistical tests were performed using SPSS software package 19.0 for Windows. Results {#Sec8} ======= Clinical characteristics {#Sec9} ------------------------ Clinical characteristics of 139 HOCM patients with mean age 43 ± 15 years were shown in Table [1](#Tab1){ref-type="table"}. Females were more than males. Body mass index was 28 ± 8 kg/m^2^ and heart rate was 74 ± 5 beat/min. All patients received medical therapy with beta-receptor blocker or calcium channel blocker. Most patients had severe preoperative symptoms including dyspnea, chest pain, and syncope. Family history of HCM was presented in 54.7 % of subjects. Hypertension, diabetes and dyslipidemia were prevalent in the HOCM patients. The majority of HOCM patients belonged to NYHA III.Table 1Clinical characteristicsVariablesValues*n*139Age (y)43 ± 15Male52 (37.4 %)Body mass index (kg/m^2^)28 ± 8Heart rate (beat/min)74 ± 5SBP (mmHg)113 ± 12DBP (mmHg)73 ± 9Symptoms Dyspnea133 (95.7 %) Chest pain73 (52.5 %) Syncope33 (23.7 %)Hypertension44 (31.7 %)Diabetes25 (18 %)Dyslipidemia101 (72.7 %)Family history of HCM76 (54.7 %)Medical therapy Beta-receptor blocker16 (11.5 %) Calcium channel blocker43 (30.9 %)*SBP* systolic blood pressure, *DBP* diastolic blood pressure Perioperative events and complications {#Sec10} -------------------------------------- Adjunctive surgical procedures were summarized in Table [2](#Tab2){ref-type="table"}. Overall, cardiopulmonary bypass time was 133 ± 40 min and aorta cross-clamp time was 85 ± 26 min. Postoperative intensive care unit (ICU) stay time was 3 ± 3 days. Mechanical ventilation time was 24 ± 15 h. Postoperative hospital stay time was 10 ± 5 days.Table 2Adjunctive proceduresProcedure*n* (%)Age (y)CABG11 (7.9 %)56 ± 16LV aneurysmectomy2 (1.4 %)40 ± 17MV replaement7 (5 %)46 ± 14MV repair17 (12.2 %)44 ± 9AV repair4 (2.9 %)61 ± 19TV repair6 (4.3 %)42 ± 16*CABG* coronary artery bypass graft, *LV* left ventricle, *MV* mitral valve, *AV* aortic valve, *TV* tricuspid valve The 139 patients had no early death within 30 days after extensive septal myectomy. Perioperative arrhythmia was shown in Table [3](#Tab3){ref-type="table"}. In this study, perioperative arrhythmia events included atrial fibrillation, atrial premature beat, ventricular premature beat, atrioventricular block, left bundle branch block, left anterior fascicular block and right bundle branch block. There was a preoperative history of atrial fibrillation in 13 patients while postoperative new atrial fibrillation occurred in 11 patients. In total, there were 16 patients suffering from atrial ventricular block after surgery. Permanent pacemaker was implanted for 8 (5.8 %) patients with complete atrioventricular block. None of the patients needed implantable cardioverter defibrillator (ICD) during early postoperative period. In addition, retraction injury to aortic valve leaflets occurred in 5 patients. One patient needed the mechanical support of intra-aortic balloon pump (IABP) immediately following surgery. Postoperative pleural effusion happened in 25 patients. Hemodialysis was used for 2 patients. Two patients were subjected to second intubation and 1 patient underwent tracheotomy. Second transfer to ICU was required for 1 patient on postoperative 7th day.Table 3Postoperative arrhythmiaArrhythmiaNumberPercentageAtrial fibrillation2417.3 %Atrial premature beat75.0 %Ventricular premature beat42.9 %Atrioventricular block1611.5 %Left bundle branch block3424.5 %Left anterior fascicular block107.2 %Right bundle branch block53.6 % Clinical and echocardiographic follow-up {#Sec11} ---------------------------------------- Clinical follow-up was 5.6 ± 0.9 years (minimum 1.2 years and maximum 7.9 years). Four patients were lost to late follow-up. Overall survival was 100.0, 99.3, 99.3, 98.5 and 97.8 % at 1, 2, 3, 4 and 5 years, respectively (Fig. [1](#Fig1){ref-type="fig"}). One patient died of cardiac origin and 2 patients had undiagnosed sudden death.Fig. 1Kaplan-Meier curve for HOCM patients during follow-up period. Survival rate and log survival rate decreased whereas cumulate hazard gradually increased during follow-up time. The survival rate was 100 % at the first year after surgery and then decreased to 99.3 % at the second and the third year. The survival rate was 98.5 % at the fourth year and reduced to 97.8 % at the fifth year Follow-up outcomes indicated the symptoms were significantly attenuated and physical abilities were increased. Improvement in NYHA class was shown in Fig. [2](#Fig2){ref-type="fig"}. The proportion of NYHA III and IV decreased from preoperative 58 to 19 % at 2.5 years after surgery while it reduced to 11 % at 5 years after operation. LVOT gradient decreased form preoperative 84 ± 17 mmHg to 12 ± 3 mmHg at 2.5 years after surgery and it further reduced to 6 ± 3 mmHg at 5 years after surgery (*P* \< 0.05, Fig. [3](#Fig3){ref-type="fig"}). Compared with the level of preoperative IVS thickness, IVS thickness decreased by 32 % at postoperative 2.5 years and maintained the same low level at postoperative 5 years (*P* \< 0.05, Fig. [3](#Fig3){ref-type="fig"}). LVID~d~approximately increased by 10 % at postoperative 2.5 years and 5 years, compared with the level before surgery (*P* \< 0.05, Fig. [3](#Fig3){ref-type="fig"}). Similarly, EF was significantly elevated at 2.5 years and 5 years after extensive septal myectomy (*P* \< 0.05, Fig. [3](#Fig3){ref-type="fig"}). By transthoracic echocardiography detection, mitral regurgitation was ameliorated for patients after surgery. There were 52 % of patients showing moderate mitral regurgitation and 8 % of patients had severe mitral regurgitation before surgery. Preoperative mild mitral regurgitation existed in 40 % of patients. With extensive septal myectomy, both severe mitral regurgitation and moderate mitral regurgitation disappeared. The proportion of postoperative mild mitral regurgitation increased to 89 and 11 % of patients did not have mitral regurgitation any more.Fig. 2Improvement in cardiac function after extensive septal myectomy. The numbers of patients in NYHA Class I-II and NYHA Class III-IV were calculated before surgery, 2.5 years and 5 years after surgery, respectively. *χ* ^2^ test was performed to analyze the differences among groups. There was significant difference in preoperative and postoperative cardiac function (*P* \< 0.05). There was no difference in cardiac functions between 2.5 years after surgery and 5 years after surgery (*P* \> 0.05). *n* = 59 for NYHA Class I-II and *n* = 80 for NYHA Class III-IV before surgery; *n* = 110 for NYHA Class I-II and *n* = 25 for NYHA Class III-IV at 2.5 years after surgery; *n* = 117 for NYHA Class I-II and *n* = 15 for NYHA Class III-IVat 5 years after surgery Fig. 3Changes in echocardiographic parameters during follow-up period. **a**. Changes in LVID~d~ during follow-up. **b**. Changes in IVS thickness during follow-up. **c**. Changes in LVOTG during follow-up. There was significant decrease in LVOTG at 2.5 years and 5 years after surgery. **d**. Changes in EF during follow-up. EF was elevated at 2.5 years and 5 years after extensive septal myectomy. LVID~d~: diastole left ventricular inner diameter; IVS: interventricular septal; LVOTG: left ventricular out tract gradient; EF: ejection fraction. ^\*^ *P* \< 0.05 vs preoperative group, ^\#^ *P* \< 0.05 vs 2.5 years group Discussion {#Sec12} ========== The present study included a relative large of Chinese HOCM patients who were treated by extensive septal myectomy in two hospitals. Our results showed that extensive septal myectomy significantly reduced LVOT obstruction, mitral regurgitation and HOCM-related symptoms. Cardiac function was obviously increased by surgical treatment. Extensive septal myectomy and adjunctive procedures could be an efficacious and low-risk therapy in Chinese HOCM patients. It gives predictable, immediate, and durable remodeling of the LVOT which translates into long-term control of symptoms. Extensive septal myectomy surgery in HOCM {#Sec13} ----------------------------------------- HOCM, as a common type of hypertrophic cardiomyopathy, is mainly characterized by asymmetric septal hypertrophy, LVOT obstruction, diastolic dysfunction, cardiac ischemia as well as arrhythmia \[[@CR16]--[@CR18]\]. The aim of medical therapy is to abolish the catecholamine-induced effects that may exacerbate LVOT obstruction and to decrease heart rate which allows longer time for diastolic filling \[[@CR17], [@CR19], [@CR20]\]. However, the early improvement for HOCM patients is often followed by clinical symptomatic impairments after conservative management with beta-blocker and/or calcium antagonist \[[@CR21]\]. With septal myectomy reduction therapy, the long-term mortality for HOCM patients could significantly decrease to 1.4--1.8 % \[[@CR7], [@CR8]\]. Left ventricular septal myectomy was firstly reported by Cleland in 1963 and Morrow subsequently revealed the good clinical and hemodynamic outcomes of myectomy surgery \[[@CR22]\]. Isolated septal myectomy mainly resected the bulge part of the hypertrophic septum. Some intraventricular anomalies such as mitral apparatus-related anomalies existed in HOCM patients. Under the condition, isolated septal myectomy could not diminish SAM of mitral valve and relieve LVOT obstruction completely \[[@CR10]\]. Marwick et al. reported that up to 20 % of patients with isolated septal myectomy were placed back on cardiopulmonary bypass because of inadequate resection \[[@CR23]\]. An inadequate length of septal excision was the most common reason of recurrent LVOT obstruction after myectomy \[[@CR24]\]. Extensive septal myectomy to the midventricular level, with or without shaving of the papillary muscles, could eliminate the LVOT gradient and SAM-induced mitral regurgitation \[[@CR25]\]. Knyshov G et al. performed a cohort study in HOCM patients and their results showed that LVOT gradient of HOCM patients reduced from 113.3 ± 14.9 mmHg to 17.3 ± 10.2 mmHg after extensive septal myectomy surgery \[[@CR26]\]. Follow-up outcomes of extensive septal myectomy surgery {#Sec14} ------------------------------------------------------- Until now, there are limited data addressing long-term outcomes of extensive septal myectomy in Chinese HOCM patients. Wang et al. showed that extensive septal myectomy provided excellent relief from LVOT obstruction (91.8 ± 25.1 to 14.3 ± 13.4 mmHg, *P* \< 0.05) and satisfactory clinical outcomes for 93 HOCM patients at early and mid-term follow-up \[[@CR27]\]. Their study showed that the 30-day and in-hospital mortality was 0 % \[[@CR27]\]. After surgery, limiting symptoms were decreased while physical abilities were increased \[[@CR27]\]. In our study, the resection size was extended to release adhesion of mitral papillary muscle with left ventricular wall and satisfactory operative effects and clinical outcomes were obtained. The symptoms of HOCM patients undergoing extensive septal myectomy procedure in our study were obviously mitigated after operation. Echocardiography results by our study showed LVOT gradient and septal width of HOCM patients were significantly reduced whereas their LVID~d,~ and EF were increased after surgery. The most common postoperative complication was arrhythmia, which mainly consisted of left bundle branch block, atrial fibrillation, atrioventricular block and so on. In current study, the types and prevalence of perioperative arrhythmia were similar to the previous reports \[[@CR28]--[@CR30]\]. Conclusion {#Sec15} ========== Extensive septal myectomy and adjunctive procedures provide excellent relief of symptoms and improve cardiac function in Chinese HOCM patients with minimal surgical risk. Further studies with larger number and longer follow-up were expected to aim at examining the clinical outcomes of extensive septal myectomy surgery. Ethics approval and consent to participate {#Sec16} ------------------------------------------ The authors stated that the study had full approval from the Ethics Boards of Cangzhou Central hospital and Shandong Chinese Medical Hospital. AML : anterior mitral leaflet EF : ejection fraction HCM : hypertrophic cardiomyopathy HOCM : hypertrophic obstructive cardiomyopathy ICU : intensive care unit IVS : interventricular septal LVID~d~ : diastole left ventricular inner diameter LVOT : left ventricular outflow tract NYHA : New York Heart Association SAM : systolic anterior motion SCD : sudden cardiac death TEE : transesophageal echocardiography **Competing interests** The authors declare that they have no competing interests. **Authors' contributions** Concept/design: LY; Data collection/analysis/statistics: LL, X-JL, X-NK, F-JD; Drafting article: LY, Y-LM; Approval of article: LY. All authors read and approved the final manuscript. Funding {#FPar1} ======= This study was supported by National Science and Technology Supporting Program (No. 2013BAJ57B08) and Hebei Talents Fund (No.2014000021769G283).

Muscle fat infiltration is recognized as a hallmark pathological feature in dystrophin glycoprotein complex (DGC)[^3^](#FN4){ref-type="fn"}-related muscular dystrophies ([@B1]) that include dystrophinopathies ([@B2], [@B3]) and sarcoglycanopathies (LGMD2C-F) ([@B4]). In agreement, magnetic resonance imaging measurements of fat infiltration allow accurate assessments of disease severity in Duchenne muscular dystrophy patients ([@B3]). Association of adipose tissue development with degenerative/regenerative or atrophic changes in skeletal muscle is also supported by the finding that adipogenesis-competent cells within the skeletal muscle are activated during muscle regeneration ([@B5]). However, the molecular mechanism(s) underlying muscle fatty metamorphosis remain unclear. Ectopic fat deposition in skeletal muscles is primarily described in animals and humans with lipodystrophy and sarcopenia. In these conditions, the accumulation of lipids and adipocytes in skeletal muscle is often accompanied by hyperglycemia and insulin resistance ([@B6][@B7][@B8][@B9][@B10]--[@B11]), both of which are strong indicators of muscle metabolic defects ([@B12], [@B13]) and deregulated adipogenesis ([@B14]). Furthermore, both adipose-derived and muscle-derived stem cells differentiate into adipocytes upon exposure to high levels of glucose ([@B15]), linking impaired muscle metabolism with muscle fat deposition. It is long held that the biogenesis of a basement membrane takes place in the earliest steps of adipogenesis and that extensive extracellular matrix (ECM) remodeling occurs throughout adipogenesis ([@B16], [@B17]). The concept that cell surface receptors play a role in the regulation of adipogenesis and thus may underlie metabolic disorders just recently emerged with a study of the integrin complexes ([@B18]). Given that the DGC in its capacity as an ECM receptor is critical for muscle integrity ([@B19], [@B20]) and that white adipocytes and skeletal muscle cells originate from the same mesenchymal precursor cells ([@B21], [@B22]), we set out to determine whether components of the skeletal muscle DGC are expressed in white adipocytes. Herein, we describe a unique adipose sarcoglycan (SG) complex that includes β-, δ-, and ϵ-SG. This complex is tightly associated with sarcospan (Sspn) and dystroglycan (DG). Moreover, we show that DG functions as a novel ECM receptor in white adipocytes. Because adipose tissue and skeletal muscle play critical roles in the maintenance of normal glucose homeostasis and whole body insulin sensitivity ([@B23]), we examined the metabolic consequences of the SG complex disruption in both adipose tissue and skeletal muscle. Using *in vivo* approaches, we observed that the β-SG null mouse ([@B24]), a mouse model of muscular dystrophy, is glucose-intolerant and exhibits whole body insulin resistance specifically due to impaired insulin-stimulated glucose uptake in skeletal muscle. EXPERIMENTAL PROCEDURES ======================= ### #### Animals Animal care and procedures were approved and performed in accordance with the standards set forth by the National Institutes of Health and the Animal Care Use and Review Committee at the University of Iowa. #### Biochemical Analysis White adipocytes were isolated from wild-type gonadal white adipose tissue ([@B25]). Total RNA extraction using RNA-STAT60^TM^ (IsoTex Diagnostics Inc., Friendswood, TX) and preparation of total membrane extracts ([@B26]) were performed from isolated adipocytes. Sucrose gradient purification of the DGC components ([@B27]) was performed from isolated adipocytes and whole WT gonadal adipose tissue. #### Physiological Analysis Glucose tolerance tests (GTTs) were performed on 16-h fasted male mice following intraperitoneal injection of [d]{.smallcaps}-glucose (1 g/kg). Blood glucose was measured from tail vein using OneTouch Ultra test strips (LifeScan). Several measurements were done at each time point to ensure reproducibility, and the lowest value was used. Fasting insulin was quantified in serum using enzyme-linked immunosorbent assay (Millipore Corp.). Euglycemic-hyperinsulinemic (EU) clamps were conducted on male mice according to the slightly modified protocol ([@B28]). Basal rates of whole body glucose clearance were assessed using a continuous infusion of \[^3^H\]glucose for 2 h prior to the start of the clamp. Insulin was infused continuously at 4 milliunits/kg/min. Forty-five min before the end of the clamp, 2-deoxy-[d]{.smallcaps}-\[1-^14^C\]glucose was administrated as a bolus (10 μCi) to estimate insulin-stimulated glucose uptake in individual tissues. #### Statistics GTT data were analyzed with two-way repeated measures analysis of variance, and all pairwise multiple comparison procedures were done with the Bonferroni *t* test. Other data were analyzed with Student\'s unpaired *t* test. RESULTS ======= To determine whether known components of the DGC are expressed in white adipose tissue while ensuring that our results would not reflect contamination from other tissues, we chose freshly isolated white adipocytes as starting material. RT-PCR data revealed the presence of β-, δ-, and ϵ-SG, Sspn and DG transcripts in adipocytes ([Fig. 1](#F1){ref-type="fig"}*A* and [supplemental Table 1s](http://www.jbc.org/cgi/content/full/C109.010728/DC1)). The identity of these transcripts was confirmed by BLAST analysis. In contrast to skeletal muscle, there was a conspicuous absence of α-SG and γ-SG transcript. The detection of a transcript correlated with the detection of the corresponding protein ([Fig. 1](#F1){ref-type="fig"}*B*). Results were confirmed by immunofluorescence analysis of adipose tissue sections ([Fig. 1](#F1){ref-type="fig"}*C* and data not shown). Of the known DGC proteins, DG is the focus of most research interest due to its widespread expression, association with the basal lamina, and the complex and heterogeneous tissue-specific *O*-glycosylation of its extracellular α-subunit. Post-translational modification of α-DG is critical for its interaction with a broad range of ECM proteins, and loss of this interaction is a common factor in the biogenesis of complex diseases ([@B19]). To better characterize the binding properties of α-DG in adipose tissue, we chose to make use of a well characterized mouse model, the myodystrophy mice (Large^myd^ or myd) ([@B29]). These mice have a spontaneous loss-of-function mutation in *Large*, which encodes a putative glycosyltransferase involved in α-DG *O*-glycosylation, leading to brain and skeletal muscle α-DG hypoglycosylation and pathology ([@B30]). We found adipocyte α-DG reactivity using antibodies directed against either the α-DG core (sheep5) or its sugar moieties (IIIH11), albeit at a lower relative molecular mass (120 kDa) than that in skeletal muscle ([supplemental Fig. 1s, *A* and *B*, *lanes 1* and *3*](http://www.jbc.org/cgi/content/full/C109.010728/DC1)). The absence of *Large* led to a loss of α-DG glycosylated epitope, with preservation of the α-DG protein core in adipocytes ([supplemental Fig. 1s, *A* and *B*, *lanes 2* and *4*](http://www.jbc.org/cgi/content/full/C109.010728/DC1)). Agrin and laminin-1 overlay experiments conclusively demonstrated that the adipose-specific glycosylation was sufficient for α-DG to function as an ECM ligand receptor ([supplemental Fig. 1s, *C* and *D*, *lanes 1* and *3*](http://www.jbc.org/cgi/content/full/C109.010728/DC1)). As expected and consistent with the observations made in myd brain and skeletal muscle, laminin and agrin binding to α-DG were abolished in absence of *Large* ([supplemental Fig. 1s, *C* and *D*, *lanes 2* and *4*](http://www.jbc.org/cgi/content/full/C109.010728/DC1)), with the level of expression of the SGs and Sspn being preserved (data not shown). These data demonstrate that Large-mediated post-translational processing of α-DG takes place in white adipocytes. In support of this conclusion, expression of *Large* and other α-DG post-translational processing enzymes, *i.e. Pomt1*, *Pomt2*, *PomGnt1*, and *Fukutin*, was observed in adipocytes by RT-PCR ([supplemental Fig. 2s](http://www.jbc.org/cgi/content/full/C109.010728/DC1)), confirming a previous gene profiling study performed in white adipocytes ([@B31]). {#F1} To test whether the DGC components expressed in the white adipocytes associate in a fashion similar to that described for skeletal muscle ([@B19], [@B20]), we analyzed their co-fractionation on sucrose gradient ([supplemental Fig. 3s*A*](http://www.jbc.org/cgi/content/full/C109.010728/DC1)). We found that Sspn migrated in the denser gradient fractions, along with the β-, δ-, and ϵ-SG and β-DG. The tight association of these proteins upon solubilization indicates that these proteins are integral components of a multisubunit complex that via DG can function as a novel ECM receptor in the white adipocytes. Control experiments in differentiated 3T3-L1 cells supported this conclusion as well (data not shown). To determine whether the adipose SG complex functions as a single unit, we tested whether the loss of a single component will influence the expression of the others and impair α-DG function. We took advantage of existing mouse models of muscular dystrophy, each with a systemic genetic ablation of one skeletal DGC component ([@B32]). [Fig. 2](#F2){ref-type="fig"}*A* shows the composition of the SG complex and its associated Sspn and DG in mdx and mdx3cv mice, utrophin (utrn), β-SG, δ-SG, and α-SG null mice. As expected, loss of either full-length dystrophin (dys) or α-SG, whose expression are restricted to skeletal muscle, had no consequence on the expression of the SGs, Sspn, and DG at the adipocyte membrane nor on the ability of α-DG to bind laminin. The loss of all dys isoforms in mdx3cv mice or of utrn in the utrn null mice was accompanied by modest decreased expression of all the proteins tested and a reduced laminin binding to α-DG. Therefore, although one cannot rule out that a short dys isoform and utrophin could be expressed in the adipocytes, their expression does not appear to be critical for proper SG complex expression at the adipocyte plasma membrane. However, loss of either β-SG or δ-SG led to a complete loss of the other SGs as well as Sspn. Moreover, only residual expression of the glycosylated form of α-DG could be detected, which correlated with weak residual laminin binding. Altogether, our data suggest that the adipose SG complex should be considered as a single unit that relies on all its individual components for full functionality. {#F2} Because both skeletal muscle and adipose tissue play important roles in glucose homeostasis, we took advantage of the β-SG null mice due to the concomitant loss of expression of the entire adipose and skeletal SG complexes. Upon dissection, we observed that the β-SG null mice had virtually no subcutaneous adipose tissue (data not shown). Furthermore, [Table 1](#T1){ref-type="table"} demonstrates that the β-SG null had a significant decrease of their visceral white adipose tissue mass (epididymal and retroperitoneal fat expressed as a percentage of total body mass) as compared with age-matched WT mice. Histological analysis of the white adipose tissue revealed conspicuous changes such as those that characterize lipodystrophy ([supplemental Fig. 4s](http://www.jbc.org/cgi/content/full/C109.010728/DC1)). Specifically, unilocular cells were heterogeneous in size and shape; numerous small multilocular cells were present, and large blood vessels and fibrous strands coursed throughout the adipose tissue. ###### **Organ weights expressed as a percentage of total weight** Organs were carefully dissected and weighed. Results are expressed as mean (S.E.). All experiments were performed in double-blind fashion. IBAT, intrascapular brown adipose tissue. \*, *p* \< 0.001. Organ WT (*n* = 15) β-SG null mice (*n* = 11) --------------------- --------------- --------------------------- Quadriceps muscle 1.54 (0.02) 1.79 (0.06)\* Calf muscle 1.09 (0.03) 1.43 (0.03)\* Cardiac muscle 0.56 (0.01) 0.57 (0.03) Kidney 1.24 (0.02) 1.22 (0.03) Liver 4.14 (0.09) 5.92 (0.11)\* Epididymal fat 1.57 (0.07) 1.01 (0.06)\* Retroperitoneal fat 0.39 (0.02) 0.22 (0.03)\* Lungs 0.56 (0.02) 0.66 (0.05) Spleen 0.27 (0.01) 0.39 (0.04) Pancreas 0.53 (0.02) 0.56 (0.04) IBAT 0.22 (0.02) 0.27 (0.02) Testis 0.69 (0.02) 0.67 (0.03) Functionally, β-SG null mice were glucose-intolerant ([Fig. 2](#F2){ref-type="fig"}*B*) and had a trend to have higher fasting insulin levels than their WT counterparts (data not shown), suggesting an alteration of their whole body insulin sensitivity. To address this last question, we performed conscious EU clamps coupled with \[^3^H\]glucose infusion in β-SG null and WT mice ([Fig. 2](#F2){ref-type="fig"}, *C--E*). EU clamps indicated that the glucose clearance was significantly diminished in the β-SG null mice as compared with the WT controls (22% decrease with *p* = 0.029, data not shown), confirming the glucose tolerance test data. The whole body glucose infusion rate was 30% lower (*p* = 0.03) in the β-SG null mice than in the controls ([Fig. 2](#F2){ref-type="fig"}*C*), demonstrating that the null animals were insulin-resistant. Interestingly, insulin resistance was accompanied by decreased insulin sensitivity in skeletal muscle only (35% decrease in insulin-stimulated glucose uptake in skeletal muscle of β-SG null mice as compared with that of WT age-matched controls, *p* = 0.016) ([Fig. 2](#F2){ref-type="fig"}*D*). Insulin sensitivity remained normal in both white adipose tissue ([Fig. 2](#F2){ref-type="fig"}*D*) and liver ([Fig. 2](#F2){ref-type="fig"}*E*). DISCUSSION ========== We have demonstrated here that white adipocytes express at their plasma membrane an SG complex unique to this cell type. It includes β-, δ-, and ϵ-SG but lacks α- and γ-SG as determined by RT-PCR and immunoblotting. Both β-SG and δ-SG proteins can seed the formation of the SG complexes (α-βδγ and ϵ-βδγ) found in skeletal and smooth muscle, respectively ([@B33], [@B34]), and a recent model proposed that γ-SG links the β/δ-SG precomplex to α-SG in skeletal muscle ([@B35]). As ζ-SG, the latest SG identified is the closest γ-SG homolog ([@B36]), it is possible that ζ-SG functionally replaces γ-SG in white adipocytes. We also showed that Sspn co-purifies with the above SG complex and that its expression is gone in the adipose tissue of SG-null mice. It was previously suggested that the assembly of a complete SG complex was a prerequisite for proper expression of Sspn at the sarcolemma ([@B37]). Likewise, our data suggest that expression of Sspn at the adipocyte membrane is dependent on proper SG complex expression. The role played by the ECM receptors ([@B18]) and the importance of the cell-cell and cell-matrix interactions in adipose tissue are just emerging concepts ([@B38][@B39]--[@B40]). Here, we found that decreased expression of DG and a loss of extracellular ligand binding to α-DG were associated with a loss of SG complex expression in white adipocytes. This suggests that the SG complex is necessary for functional DG expression as an ECM receptor at the adipocyte plasma membrane. The physiological characterization of β-SG null mice suggests that these mice, primarily known as a genetic model of muscular dystrophy ([@B24], [@B32]), could also be a new genetic model of lipodystrophy ([@B9]). It is well established that adipose tissue serves as a crucial integrator of glucose homeostasis ([@B12], [@B28]) and that alterations of its physiology can have a wide range of metabolic consequences ([@B41], [@B42]). Thus, it is tempting to speculate that the alteration of the adipose SG complex could represent a new molecular mechanism underlying phenotypes of altered adiposity, abnormal fat distribution, and metabolic disorders, likely via an alteration of the cell-cell and/or cell-matrix interactions in adipose tissue. However, since the global ablation of β-SG affects both skeletal muscle and adipocytes, future studies using tissue-specific knock-out animals will be necessary to specifically resolve the tissue-specific roles of the SG complex in glucose metabolism. Nevertheless, these findings raise novel questions that have important implications for clinical studies of muscular dystrophy. DGC-related muscular dystrophies are often accompanied by progressive muscle fat replacement ([@B2][@B3]--[@B4]), and the ECM imparts specific characteristics to each tissue. Thus, we speculate that the SG complex, along with Sspn and DG, plays a role in regulating the adipogenesis/myogenesis cell fate decisions potentially involved in the mechanism of muscle fat replacement. Furthermore, muscular dystrophies are complex diseases likely to involve more than the skeletal muscles ([@B19]). However, the efficiency of the therapeutic strategies are so far solely based on the recovery of the skeletal DGC and/or muscle strength ([@B43], [@B44]). Our data indicate that a cell-specific SG complex is expressed in the white adipocytes. Furthermore, we for the first time bring evidence of metabolic defects in the skeletal muscles and whole body of the β-SG null mice. We speculate that the severity of the metabolic symptoms is likely to depend on whether the skeletal SG complex alone (*i.e.* in α-SG null mice) or both adipose and skeletal SG complexes (*i.e.* in β-SG null mice) are disrupted and on the relative functional contribution of the primary missing protein. Although the association of metabolic defects with the progression of muscular dystrophy is not documented yet in patients, it is possible that the recovery of skeletal muscle function in treated patients might uncover new health problems due to defective SG complexes in non-muscle tissues, such as adipose tissue. Supplementary Material ====================== ###### Supplemental Data This work was supported, in whole or in part, by National Institutes of Health Grant R01 AR051199-01 through the NIAMS. This work was also supported by a Paul D. Wellstone Muscular Dystrophy Cooperative Research Center grant and by a grant from the Muscular Dystrophy Association. The on-line version of this article (available at <http://www.jbc.org>) contains [supplemental "Experimental Procedures](http://www.jbc.org/cgi/content/full/C109.010728/DC1)," [supplemental Table 1](http://www.jbc.org/cgi/content/full/C109.010728/DC1), and [supplemental Figs. 1s--4s](http://www.jbc.org/cgi/content/full/C109.010728/DC1). The abbreviations used are: DGdystroglycanDGCdystrophin glycoprotein complexSGsarcoglycanSspnsarcospanRTreverse transcriptaseECMextracellular matrixWTwild typeGTTglucose tolerance testEUeuglycemic-hyperinsulinemicutrnutrophindysdystrophin. We thank Sarah Anderson and Keith Garringer for maintaining the mouse colonies, Sally J. Prouty for technical assistance, and Campbell laboratory members for the critical reading of the manuscript. We also acknowledge The University of Iowa DNA Facility, which is supported in part by the Holden Comprehensive Cancer Center (HCCC), as well as the University of Iowa Hybridoma Facility. [^1]: An American Heart Association Postdoctoral Fellow (0120483Z).

Background {#Sec1} ========== Inorganic and organic wastes produced by human activities have resulted in high volumes of contaminated waters which threaten the health of human beings and other living organisms \[[@CR1]\]. Water pollution is one of the most serious environmental problems nowadays, which hinders the development of human society \[[@CR2], [@CR3]\]. In particular, dye pollutants have attracted wide concerns from the public because of the high visibility and the toxic impact on biological organisms and the ecology \[[@CR4]\]. Organic dyes have been extensively used in various branches such as textile, paper, printing, color photography, pharmaceutical industry, leather, cosmetics, plastic and other industries, which have been the major industrial wastewater sources \[[@CR5]\]. The quantity of dye wastewater is extremely large, generally, the volume of discharged wastewater from each step of a textile operation is approximately at a high rate of between 40 L/kg and 65 L/kg of the product \[[@CR6]\]. In addition, dyes are non-biodegradable substances that remain stable under different conditions due to their synthetic origin and complex aromatic structures \[[@CR7]\]. Therefore, it is necessary to select an appropriate method to remove dyes from wastewater before discharging into the environment. In recent years, a variety of techniques has been used to treat dye wastewater including photocatalytic degradation \[[@CR8]\], coagulation \[[@CR9]\], electrochemical processes \[[@CR10]\], chemical oxidation \[[@CR11]\], membrane filtration \[[@CR12]\], biological treatment \[[@CR13]\] and adsorption \[[@CR14]\]. Among these dye wastewater treatment techniques, adsorption has been widely used due to their merits of simplicity, high efficiency and economy \[[@CR15], [@CR16]\]. Many adsorbents such as activated carbon, kaolin, montmorillonite clay, waste red mud, fullers earth and fired clay have been reported to decolorize wastewater \[[@CR17], [@CR18]\]. Especially, magnetic nanoparticles (MNPs) have attracted considerable attention as adsorbent materials for organic dyes and heavy metals, due to their unique magnetic properties, low cost, biocompatible, easily synthesized, readily recycle, particularly economic and environmental friendly \[[@CR19]\]. Several methods have been developed to synthesize magnetic Fe~3~O~4~ nanoparticles, including i) coprecipitation of ferrous and ferric aqueous solution in the presence of a base \[[@CR20]\]; ii) thermal decomposition of an iron complex \[[@CR21]\]; iii) sonochemical approach \[[@CR22]\]. Because of their high surface energies and intrinsic magnetic interactions, easy aggregation of Fe~3~O~4~ MNPs would reduce their surface/volume ratio and dispersion stability in aqueous solution \[[@CR23]\]. The stabilizers such as surfactants, supporters, oxides or polymeric compounds have been used to modify Fe~3~O~4~ MNPs to increase their stability and improve their dispersity. Zhang et al. synthesized magnetic Fe~3~O~4~/C core shell nanoparticles and used as absorbents performing good adsorption capacity for dye removal \[[@CR24]\]. Wang et al. prepared Fe~3~O~4~ nanoparticles with cetyltrimethylammonium bromide (CTAB) assistant for adsorption removal of congo red (CR) and methylene blue (MB) \[[@CR25]\]. Furthermore, the adsorption capacity of bare Fe~3~O~4~ MNPs is not strong enough. In order to improve the adsorption property, surface functionalization of Fe~3~O~4~ MNPs has been studied. Zhang et al. modified Fe~3~O~4~ MNPs with 3-glycidoxypropyltrimethoxysilane (GPTMS) and glycine (Gly), the magnetic nanocomposites could excellently remove both anionic and cationic dyes in severe environment (highly acidic or strong alkaline) \[[@CR26]\]. Moreover, selective adsorption can be greatly improved for the enrichment of pollutants due to introduction of large numbers of active sites. Pourjavadi et al. reported a new functionalized magnetic nanocomposite of poly(methylacrylate) for the efficient removal of anionic dyes from aqueous media \[[@CR27]\]. Polycatechol, resulting from the polymerization of catechol catalyzed by Fe(III) \[[@CR28]--[@CR30]\], has been exploited in surface modifications as adhesives and coatings over a wide range of both organic and inorganic materials due to their unique thermal, structural properties, and the ability to form stable complexes with metal oxides \[[@CR31], [@CR32]\]. It means that Fe~3~O~4~ MNPs modified with polycatechol will greatly increase the adsorption ability of Fe~3~O~4~ MNPs. However, there is no report about polycatechol modified Fe~3~O~4~ MNPs as an absorbent for dye removal by far. In this work, polycatechol modified Fe~3~O~4~ MNPs (Fe~3~O~4~/PCC MNPs) were prepared by a facile coprecipitation method and used as adsorbents for dye removal. The absorbent was characterized using magnetic hysteresis loops, thermogravimetric analysis and zeta potential analysis technique. Five kinds of cationic dyes, including methylene blue (MB), cationic turquoise blue GB (GB), malachite green (MG), crystal violet (CV) and cationic pink FG (FG), were chosen as the model compounds to expose the adsorption behavior of Fe~3~O~4~/PCC MNPs. The adsorption kinetics, isotherm analyses and the effect of different experimental conditions on the removal of cationic dyes were also investigated. Methods {#Sec2} ======= Materials {#Sec3} --------- Ferric chloride (FeCl~3~·6H~2~O), ferrous sulfate (FeSO~4~·7H~2~O), ammonium hydroxide (NH~3~·H~2~O, 25%), MB, GB, MG, CV, FG, Orange ΙΙ, Fuchsin, methyl orange (MO) and catechol were obtained from Chuandong Chemical Inc., Chengdu, Sichuan, China. All chemicals were analytical grade and used without further purification and all solutions and suspensions were prepared with deionized water. The structures of five cationic dyes, including MB, GB, MG, CV and FG, were shown in Fig. [1](#Fig1){ref-type="fig"}.Fig. 1Molecular structures of (**a**) MB (**b**) GB (**c**) MG (**d**) CV (**e**) FG. As shown as Fig. 1, the structure of five kinds of cationic dyes are described Preparation and characterization of Fe~3~O~4~/PCC MNPs {#Sec4} ------------------------------------------------------ Fe~3~O~4~/PCC MNPs were prepared by a facile chemical coprecipitation method using iron salts and catechol as precursors \[[@CR23]\]. The whole synthesis process was performed at ambient atmosphere. In brief, FeCl~3~·6H~2~O (10 mmol) and FeSO~4~·7H~2~O (5 mmol) were dissolved into 75 mL deionized water, then mixed with 75 mL of catechol (20 mM) sufficiently. When catechol was mixed with iron solution (pH 2.87), the polymerization of catechol catalyzed by Fe^3+^ happened, forming polycatechol, which was black coarse precipitates \[[@CR30]\]. Simultaneously, Fe^3+^ ions were chemically adsorbed on polycatechol through complexation and served as nucleation precursors. The mixture was standing for 30 min and then was added into 100 mL of ammonium hydroxide (3.3 M) rapidly, aging for 120 min under vigorous stirring. The magnetic nanoparticles in situ grew to form aggregations on the surface of polycatechol chains. Couples of Fe~3~O~4~/PCC chains combined with each other tightly to form Fe~3~O~4~/PCC MNPs. The whole synthesis processes were performed at ambient atmosphere, without any protective atmosphere. The black magnetic nanoparticles were separated by an external magnet and washed with deionized water until neutral pH and dried in a vacuum oven at 50 °C for 24 h. Fe~3~O~4~ MNPs were synthesized with the same procedures as mentioned above without adding catechol. All the products were stored in a desiccator under ambient temperature for further experiments. Magnetic properties were measured at room temperature on a magnetic property measurement system (MPMS XL-7, Quantum Design, America). Thermogravimetric analysis (TGA) was carried out for powder samples using a TGA/DSC 1 thermogravimetric analyzer (TGA) (Mettler-Toledo, Switzerland) under N~2~ environment at a heating rate of 5 °C min^− 1^. The zeta potentials of catalyst suspensions at different pH were determined by a Malvern 3000 Zetasizer. Batch adsorption experiments {#Sec5} ---------------------------- Sorption isotherm experiments were carried out by shaking 25 mg Fe~3~O~4~/PCC MNPs in 25 mL solution with varied adsorbates, with initial adsorbate concentration varying from 0.02 mM to 0.4 mM. The mixture was continuously shaken on a shaker at 180 rpm under controlled temperature of 30 °C until reaching equilibrium. The solution pH was adjusted by using 1.0 M H~2~SO~4~ or 1.0 M NaOH solutions. After adsorption, the adsorbent was separated from the solution under magnetism, and then the supernatant liquid was measured at the maximum absorbance of each dye by a UV-vis spectrophotometer. Furthermore, the adsorption kinetics of the processes were studied. 100 mg Fe~3~O~4~/PCC MNPs were suspended into 100 mL 0.1 mM solutions of adsorbates, and then shaken at 180 rpm under pH 6.0 and 30 °C. At different time intervals, 0.5 mL of suspension sample was withdrawn and immediately separated by an external magnetism and the supernatant liquid was collected for analysis. The influences of pH value and temperature on adsorption of cationic dyes were also studied. The typical batch adsorption experiment was carried out as follows: 50.0 mg of Fe~3~O~4~/PCC MNPs was dispersed in 50.0 mL of cationic dyes solution and then was shaken on a shaker with a speed of 180 rpm. All the adsorption experiments were carried out in duplicate. The adsorption capacity of each dye in the adsorption system, q~*e*~, was calculated according to Eq. ([1](#Equ1){ref-type=""}):$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {q}_e=\left({C}_i-{C}_e\right)\ V/{M}_s $$\end{document}$$ Where *q*~*e*~ (mg g^− 1^) is the adsorption capacity, *C*~*e*~ (mM) is the equilibrium concentration in the aqueous phase, *Ci* (mM) is the initial aqueous phase concentration, *V* (L) is the volume of solution and *M*~*s*~ (g) is the mass of solid adsorbent. Results and discussion {#Sec6} ====================== Characterization of Fe~3~O~4~/PCC MNPs {#Sec7} -------------------------------------- Figure [2a](#Fig2){ref-type="fig"} shows the magnetic hysteresis loops determined at room temperature for Fe~3~O~4~ and Fe~3~O~4~/PCC MNPs. The saturation magnetization values of Fe~3~O~4~/PCC MNPs were 53.5 emu g^− 1^, higher than that of Fe~3~O~4~ (49.6 emu g^− 1^), suggesting that they could be easily separated by an external magnetic field \[[@CR33]\]. The particle size, spin canting phenomenon, size effect, and others, are related to the saturation magnetization of the ferrite nanoparticles \[[@CR34]\]. The modification of polycatechol makes the Fe~3~O~4~/PCC MNPs much higher in crystallization, and smaller in particle size than Fe~3~O~4~ MNPs, which could be result in the higher saturation magnetization of Fe~3~O~4~/PCC MNPs than Fe~3~O~4~ MNPs. Furthermore, higher saturation magnetization of the prepared Fe~3~O~4~/PCC MNPs could also be attributed to the surface effect, sometimes called the "dead surface". The dead surface is associated with disorder of surface spin \[[@CR35]\].Fig. 2**a** Magnetization behavior of Fe~3~O~4~ MNPs and Fe~3~O~4~/PCC MNPs at room temperature. **b** Thermogravimetry (TGA) curves of Fe~3~O~4~ MNPs and Fe~3~O~4~/PCC MNPs. **c** zeta potentials of Fe~3~O~4~/PCC magnetic nanoparticles. In Fig. 2, the blank lines represent the nature of Fe~3~O~4~ MNPs, and the red lines are on behalf of the property of Fe~3~O~4~/PCC MNPs The thermal behaviors of Fe~3~O~4~ and Fe~3~O~4~/PCC MNPs were further investigated by thermogravimetric analysis (TGA) (Fig. [2b](#Fig2){ref-type="fig"}). For the TGA curve of Fe~3~O~4~ MNPs, the weight loss (6.5%) below 150 °C was due to the loss of physically adsorbed water. For the curve of Fe~3~O~4~/PCC MNPs, the weight loss (5.2%) below 150 °C was due to the loss of physically adsorbed water, the weight loss (9.4%) from 150 °C to 400 °C was ascribed to the loss of oxygen-containing functional groups, the weight loss (6.8%) from 400 °C to 800 °C was mainly attributed to the burning of carbon, and a slight weight gain (2.3%) over 800 °C was due to the oxidization of Fe~3~O~4~ to γ-Fe~2~O~3~ \[[@CR36]\]. The Fe~3~O~4~/PCC MNPs exhibited a lower thermal stability rather than Fe~3~O~4~, due to the modification of polycatechol \[[@CR37]\]. Figure [2c](#Fig2){ref-type="fig"} shows the zeta potentials of the Fe~3~O~4~ and Fe~3~O~4~/PCC suspensions at various pH values. As shown in Fig. [2c](#Fig2){ref-type="fig"}, the isoelectric point of Fe~3~O~4~ was 4.2, while the surface of Fe~3~O~4~/PCC MNPs possessed negative charges in the pH range of 3.0--10.0, which could be due to the electronegativity of phenolic hydroxyl group in polycatechol. Moreover, the surface charge density of Fe~3~O~4~/PCC MNPs increased as the pH increased from 3.0 to 10.0. The negative charges of Fe~3~O~4~/PCC MNPs also prevented nanoparticles from agglomeration. Selective adsorption of Fe~3~O~4~/PCC MNPs {#Sec8} ------------------------------------------ The adsorption properties of the Fe~3~O~4~/PCC MNPs towards cationic dyes, anionic dyes and phenol from aqueous solution were investigated in detail. Figure [3](#Fig3){ref-type="fig"} shows the removal efficiencies of MB as a kind of cationic dye, MO as a kind of anionic dye and phenol adsorbed onto Fe~3~O~4~/PCC MNPs. It was observed that the removal efficiency of MB was 75.7%. However, the removal efficiencies of MO was 10.9% only, and the removal efficiency of phenol was 1.5% only. The results indicated that the Fe~3~O~4~/PCC MNPs selectively adsorbed cationic dye, due to the electrostatic interaction (Fig. [2c](#Fig2){ref-type="fig"}).Fig. 3The removal efficiency of MB, MO and phone adsorbed by Fe~3~O~4~/PCC MNPs. As shown as Fig. 3, the blank line represent the removal efficiency of MB, the red line reperesent the removal of MO, and the blue line represent the removal of phonel Adsorption kinetics of cationic dyes {#Sec9} ------------------------------------ Adsorption kinetics of five cationic dyes on Fe~3~O~4~/PCC MNPs were studied using two kinetic models, namely the pseudo-first-order model and the pseudo-second-order model (Fig. [3](#Fig3){ref-type="fig"}). In the pseudo-first-order kinetic model, the rate constant of dyes adsorption is expressed as \[[@CR38]\]:$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ Ln\left({q}_e-{q}_t\right)=\mathit{\ln}\left({q}_e\right)\hbox{--} {k}_1\ t $$\end{document}$$ where *q*~*e*~ and *q*~*t*~ are the amounts of dyes adsorbed (mg g^− 1^) at equilibrium and at any instant of time *t* (min), respectively, and *k*~*1*~ is the rate constant of pseudo-first-order adsorption (min^− 1^). The pseudo-second-order kinetic model is described by the formula \[[@CR39]\]:$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ t/{q}_t= 1/{k}_{ad}\ {q_e}^2+ 1/{q}_e $$\end{document}$$ Where *q*~*e*~ and *q*~*t*~ are, respectively, the amount of dyes adsorbed at the equilibrium and time *t* (mg g^− 1^); and *k*~*ad*~ is the pseudo-second-order rate constant for the adsorption process (mg g^− 1^ min^− 1^). The parameter values for each model were calculated from the linear least square method and the correlation coefficients were presented in Table [1](#Tab1){ref-type="table"}. The results showed that all the adsorption kinetics of these five cationic dyes on Fe~3~O~4~/PCC MNPs could be well described by pseudo-second-order kinetics model with high correlation coefficient (R^2^ \> 0.997) and the rate constants (*k*~*ad*~) were calculated to 0.043, 0.047, 0.051, 0.057, 0.052 g mg^− 1^ mL^− 1^, corresponding to MB, GB, MG, CV and FG, respectively (Fig. [4](#Fig4){ref-type="fig"} and Table [1](#Tab1){ref-type="table"}). Moreover, the adsorption capacity of MB on Fe~3~O~4~/PCC MNPs was significantly improved, comparing to that of Fe~3~O~4~ MNPs (Additional file [1](#MOESM1){ref-type="media"}: Figure S1). The main reason was the electrostatic attractions between the positive charge of cationic dyes and the negative charge of Fe~3~O~4~/PCC MNPs.Table 1Kinetics parameters for the adsorption of cationic dyes on Fe~3~O~4~/PCC MNPsCationic dyespseudo-first-order kineticspseudo-second-order kinetics*q* ~*e*~*kR* ^*2*^*q* ~*e*~*kR* ^*2*^MB11.830.00480.85923.120.0430.999GB20.170.00240.79821.260.0470.997MG12.280.00410.86819.610.0510.998CV7.710.00380.68217.570.0570.999FG17.920.00300.87319.380.0520.998Fig. 4Adsorption of cationic dyes on Fe~3~O~4~/PCC MNPs (**a**) pseudo-second-order kinetics (**b**) pseudo-first-order kinetics. As shown as Fig. 4, the blank line represent the adsorption of MB, the red line reperesent the adsorption of GB, the blue line represent the adsorption of MG, the magenta represent of CV adsorption, and the olive is on behalf of the adsorption of FG Adsorption isotherms of different cationic dyes {#Sec10} ----------------------------------------------- The adsorption isotherm played a significant role in evaluating the adsorption properties of Fe~3~O~4~/PCC MNPs \[[@CR40]\]. To depict the adsorption process thoroughly, two well-known isotherm equations, Langmuir and Freundlich equations (Eqs. ([4](#Equ4){ref-type=""}) and ([5](#Equ5){ref-type=""})), were applied \[[@CR41]\]. Langmuir equation:$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {C}_e/{q}_e={C}_e/{q}_m+ 1/{K}_L\ {q}_m $$\end{document}$$ where *q*~*e*~ (mg g^− 1^) is the equilibrium adsorption capacity of dye on the adsorbent; *C*~*e*~ (mg L^− 1^) is the equilibrium dye concentration in solution; *q*~*m*~ (mg g^− 1^), the maximum capacity of the adsorbent; and *K*~*L*~ (L mg^− 1^), the Langmuir constant. Freundlich equation:$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {q}_e={K}_F\ {C_e}^{1/n} $$\end{document}$$ Where *q*~*e*~ and *C*~*e*~ are defined to be the same as above; *K*~*F*~ (L mg^− 1^) is the Freundlich constant; and n is the heterogeneity factor. Figure [5](#Fig5){ref-type="fig"} shows the adsorption isotherms of cationic dyes on Fe~3~O~4~/PCC MNPs. The results indicated that the adsorption of the five cationic dyes all fitted better with Langmuir equation than with Freundlich equation according to the correlation coefficients. The maximum adsorption capacities (*q*~*m*~) for these dyes were worked out by the Langmuir equation which were listed in Table [2](#Tab2){ref-type="table"}. The *q*~*m*~ for cationic dyes: MB, GB, MG, CV and FG were 60.06, 70.97, 66.84, 66.01 and 50.27 mg g^− 1^, respectively. The fitted Langmuir model assumed that the single pollutant bonded to a single site on the adsorbent and that all surface sites on the adsorbents had the same affinity for pollutant and no interactions between pollutant \[[@CR42]\].Fig. 5**a** Isotherms of cationic dyes adsorption on Fe~3~O~4~/PCC MNPs (**b**) Langmuir and (**c**) Freundlich adsorption isotherm models for organic dyes by Fe~3~O~4~/PCC MNPs. As shown as Fig. 5, the blank line represent the adsorption of MB, the red line reperesent the adsorption of GB, the blue line represent the adsorption of MG, the magenta line represent of CV adsorption, and the olive line is on behalf of the adsorption of FGTable 2Parameters of adsorption isotherm models parameters for the adsorption of dyes on Fe~3~O~4~/PCC MNPsCationic dyesLangmuir modelFreundlich modelq~e~bR^2^q~e~nR^2^MB60.060.310.99526.485.540.928GB70.970.470.99928.124.430.943MG66.840.220.99723.814.440.990CV66.010.280.99724.144.390.852FG50.270.340.98331.334.020.796 Effect of temperature on cationic dyes adsorption {#Sec11} ------------------------------------------------- The effect of temperature on the adsorption of cationic dyes was shown in Fig. [6](#Fig6){ref-type="fig"}. As can be seen, the removal efficiency of MB increased with rising temperature (30--45 °C), and it reached up to 84% at 45 °C, which suggested that the adsorption of MB on Fe~3~O~4~/PCC was an endothermic process. While the removal efficiency of GB and CV decreased with rising temperature, suggesting an exothermic reaction for the adsorption of GB and CV, which indicated the sorption processes were mainly physical adsorption. Furthermore, the reaction temperature had little effect on the adsorption of WG and FG. The effect of reaction temperature on the adsorption of five cationic dyes was different, mainly because of the different structure of dyes and the hole of the MNPs. When the holes of the MNPs are too small to be get into, the adsorbate molecules have to go over the high barrier to get into the hole. Since the holes are small and the diffusion is blocked, the adsorption process is more unstable, resulting in higher energy and the process is endothermic. Otherwise the process is exothermic.Fig. 6Effect of temperature on the adsorption of cationic dyes on Fe~3~O~4~/PCC MNPs. As shown as Fig. 6, the blank line represent that the temperature is 30 °C, the red line reperesent that the temperature is 40 °C, and the blue line represent the temperature is 45 °C Effect of pH on cationic dyes adsorption {#Sec12} ---------------------------------------- The pH of the aqueous solution was an important factor that affects the dye-adsorption process, because it influenced the surface charge of an adsorbent and the ionization behavior of both the adsorbent and dye \[[@CR43]\]. The effect of pH on the removal of cationic dyes was studied at a dye concentration of 0.1 mM at 30 °C and at pH values from 3.0 to 9.0. As shown in Fig. [7](#Fig7){ref-type="fig"}, the removal efficiency of cationic dyes increased with increasing pH value. Because the Fe~3~O~4~/PCC MNPs possessed negative charge, and their surface charge density increased with higher pH (Fig. [2c](#Fig2){ref-type="fig"}), cationic dyes were adsorbed on Fe~3~O~4~/PCC MNPs through the electrostatic attractions between the positive charge of cationic dyes molecules and the negative charge of Fe~3~O~4~/PCC MNPs. As the pH increases, the electrostatic attraction between the negatively charged surface of the Fe~3~O~4~/PCC composite and cationic dyes molecule increased, resulting in the increase in the adsorption capacity of cationic dyes. Therefore, the elevated pH helped the removal of cationic dyes by Fe~3~O~4~/PCC MNPs.Fig. 7Effect of initial pH on the removal of cationic dyes on Fe~3~O~4~/PCC MNPs the temperature is 30 °C. As shown as Fig. 7, the blank line represent that the solution pH is 3.0, the red line reperesent that the solution pH is 6.0, and the blue line represent that the solution pH is 9.0 Effect of coexisting cations on MB adsorption {#Sec13} --------------------------------------------- Dye effluents always contained a large variety of coexisting ions, which might affect the dye adsorption process \[[@CR4]\]. In this study, three commonly coexisting salts, NaCl, MgSO~4~ and FeCl~3~ were selected to study the effect of coexisting cations and their ionic strength on MB adsorption onto Fe~3~O~4~/PCC MNPs with the results presented in Fig. [8](#Fig8){ref-type="fig"}. As can be seen, Na^+^, Mg^2+^ and Fe^3+^ all suppressed MB adsorption due to the competition adsorption between cations and MB on the adsorptive sites of Fe~3~O~4~/PCC MNPs. Moreover, the removal efficiency of MB decreased from 63% to 20% with Fe^3+^ concentration increasing from 0.1 mM to 0.5 mM. Such competitive adsorption was widely reported in the literature \[[@CR44]\]. The results further confirmed the electrostatic adsorption of MB on Fe~3~O~4~/PCC MNPs.Fig. 8Effect of coexisting cations and ionic strength on the adsorption of MB on Fe~3~O~4~/PCC MNPs. As shown as Fig. 8, the blank line represent the adsorption of MB without any coexisting ions, the red line reperesent the effect on MB adsorption with 10 mM NaCl, the blue line reperesent the effect on MB adsorption with 50 mM NaCl, the magenta line reperesent the effect on MB adsorption with 10 mM MgSO~4~, the olive line reperesent the effect on MB adsorption with 50 mM MgSO~4~, the navy line reperesent the effect on MB adsorption with 0.1 mM FeCl~3~, the violet line reperesent the effect on MB adsorption with 0.2 mM FeCl~3~, the purple line reperesent the effect on MB adsorption with 0.1 mM FeCl~3~ Recycle of the adsorbent {#Sec14} ------------------------ After adsorption, Fe~3~O~4~/PCC MNPs could be regenerated with ethanol desorption at pH 4.0 for 12 h and washed with deionized water to neutral condition. The Fe~3~O~4~/PCC MNPs could be regenerated and reused for five times. Figure [9](#Fig9){ref-type="fig"} shows the adsorption performance of the regenerated Fe~3~O~4~/PCC MNPs for cationic dyes. The removal efficiency of cationic dyes decreased gradually during the first adsorption-desorption cycle to the fifth cycle. At the sixth cycle, the removal efficiency of MB, GB, MG, CV and FG decreased dramatically to 27%, 23%, 37%, 43% and 39%, respectively. Notably, the presence of magnetic nanoparticles facilitated separation and recovery of the adsorbent. It indicates that the Fe~3~O~4~/PCC MNPs has a certain economic feasibility.Fig. 9The histogram of removal efficiency of cationic dyes by Fe~3~O~4~/PCC MNPs adsorption (\[Fe~3~O~4~/PCC\] = 1.0 g L^− 1^, \[dyes\]~0~ = 0.1 mM, pH 6.0, t = 300 min) Conclusion {#Sec15} ========== In conclusion, a new magnetic nano-adsorbent (Fe~3~O~4~/PCC MNPs) was successfully prepared with active adsorption sites for removing cationic dyes from aqueous solution. The introduction of polycatechol in the structure of Fe~3~O~4~/PCC MNPs performed amazing advantages, including preventing nanoparticles from agglomeration and improving adsorption behavior of the MNPs. The electrostatic interaction was found to be the main force of the adsorption behavior for the cationic dyes. The adsorption process was well-described by pseudo-second-order kinetics and Langmuir isotherm models, respectively. The results demonstrated that Fe~3~O~4~/PCC MNPs showed potential application for cationic dyes removal in industrial effluents. Additional file =============== {#Sec30} Additional file 1:**Figure S1**. Adsorption removal efficiency of MB on Fe3O4 and Fe3O4/PCC MNPs, inset is the adsorption capacity of MB (\[Fe3O4/PCC\] = 1.0 g L-1, \[Fe3O4\] = 1.0 g L-1, \[MB\] = 0.1 mM, pH = 6.0, T = 30 °C). (DOC 45 kb) CR : Congo red CTAB : Cetyltrimethylammonium bromide Fe~3~O~4~/PCC : Fe~3~O~4~/polycatechol Gly : Glycine GPTMS : 3-glycidoxypropyltrimethoxysilane MB : Methylene blue MNPs : Magnetic nanoparticles PCC : Polycatechol TGA : Thermogravimetric analysis **Electronic supplementary material** The online version of this article (10.1186/s11671-018-2476-7) contains supplementary material, which is available to authorized users. The authors include Chuan Wang thank Professor Liu for fruitful discussions. Funding {#FPar1} ======= Not applicable Availability of data and materials {#FPar2} ================================== Not applicable Juan Xiao participated in the materials preparation, data analysis and drafted the manuscript. Qinqin Zhang participated in the sample characterization. Chang Cui conceived and co-wrote the paper. Chuan Wang involved in the experimental design and revised the manuscript. All authors include Chuan Wang read and approved the final manuscript. Competing interests {#FPar3} =================== All the authors include Chuan Wang declare that they have no competing financial interests. Publisher's Note {#FPar4} ================ Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Introduction {#s1} ============ Schizophrenia is a chronic brain disorder, the lifetime prevalence of which is nearly 1% worldwide ([@B1]). The prevalence of violent offenses in schizophrenic patients is higher than that of the general population ([@B2]--[@B4]). In China, meta-analysis from our group showed, the prevance of aggression of inpatients with schizophrenia was 35.4% (95% CI: 29.7%, 41.4%) ([@B5]). Although the etiology of violent and aggressive behavior in schizophrenia is multifactorial, genetic inheritance may strongly contribute to it ([@B4], [@B6], [@B7]). Recent studies suggested that immuno-inflammatory processes are involved in the etiology of schizophrenia ([@B8]--[@B10]) and aggressive behavior in schizophrenia ([@B11]). For example, *Toxoplasma gondii* antibodies have been observed in some schizophrenic populations ([@B12], [@B13]), as well as increases in maternal ([@B14]--[@B16]) and childhood infections ([@B17]). Most importantly, genetic studies also support an immune component to schizophrenia risk, the most statistically significant component is the major histocompatibility complex (MHC) region of chromosome 6 ([@B18]--[@B21]). Activated ([@B22], [@B23]) or impaired ([@B24]) T-cell function in inflammatory and degenerative process has been shown to be a risk factor for schizophrenia. Moreover, biased T-cell receptor repertoires are involved in severe mental disease ([@B25], [@B26]). In the T-cell receptor beta variable chain (TRBV), the complementarity-determining region 3 (CDR3) is the most variable, and defines the different populations of T cells, determines the specificity by recognizing pathogen or autoantigen epitopes ([@B27]). The diversification of T cells occurs during the lymphocyte maturation process, which includes genomic rearrangements in the CDR3 variable (V), diversity (D), and joining (J) gene segments and the addition/subtraction of non-templated bases of the recombinant junctions ([@B28]). It is estimated that the diversity of human TCR αβ pairs is nearly 2.5 × 10^18^ ([@B29]). Traditional sequencing technologies are inadequate for identification and quantification of this many T-cell antigen-receptor clonotypes. Immune repertoire sequencing, which is the large-scale sequencing of TCR repertoires, provides more distinct and detailed molecular characterization of complex sequencing targets ([@B30]). Thus, it enables the examination of the immune system at an unprecedented level ([@B27]). Using this method, we can search potential autoreactive clones and autoantigens, which may provide important information for the classification and monitoring of schizophrenia ([@B27]). It has been proposed that immuno-inflammatory responses may enhance the risk and development of schizophrenia ([@B14], [@B16], [@B17], [@B31]). The immune system may influence neurometabolic, neuroendocrine, neurodevelopmental processes ([@B32], [@B33]), especially the effect of T-cell on cognitive function ([@B34]), learning and memory ([@B35]), social behavior ([@B36]), which are all critical etiological pathways for schizophrenia. Compared to non-violent patients, violent schizophrenic patients show disturbances in response inhibition and emotional processing ([@B37]), impaired neurocognition ([@B38], [@B39]), and stronger neurodevelopmental symptoms ([@B40]). Therefore, in the current study, high-throughput sequencing technology was used to analyze the characteristics and diversity of immune molecules to explore the role of T cells in the immune system-induced pathogenesis of schizophrenia. We hypothesized that schizophrenic patients would have different TCR-repertoire diversity compared to healthy controls and that violent and non-violent schizophrenic patients would have different TCR-repertoire diversity. Materials and methods {#s2} ===================== Patients and sample collection ------------------------------ 10 violent schizophrenic patients and 10 non-violent schizophrenic patients were enrolled in this study. Violent patients who had been accused of homicide were recruited from the forensic psychiatric department of a hospital in Hunan province, China. The non-violent patients, who had never displayed violent behavior, were recruited from the in- or outpatient department of the same psychiatric hospital. All patients were males who met the following criteria: 18--40 years of age, unmarried, drug-naive, no family history of psychiatric illness, and first-onset schizophrenia diagnosis according to the ICD-10. Patients with any comorbid psychiatric disorders or accompanying somatopathy were excluded. Eight healthy subjects matched for age and gender served as controls. There were no significant differences in age or years of education between three groups (Table [1](#T1){ref-type="table"}). ###### Basic demographic and clinical characteristics of violent and non-violent schizophrenia patients. **Violent schizophrenia (*****n*** = **10)** **Non-violent schizophrenia (*****n*** = **10)** **Healthy control (*****n*** = **8)** ***Statistics*** -------------------- ---------------------------------------------- -------------------------------------------------- --------------------------------------- ------------------ ------ ----- --------- -------- ---------- Age 27.6 5.8 24.2 4.1 24.5 4.4 1.4 2,25 0.26 Education (years) 9.5 1.4 11.0 1.9 11.1 1.5 2,25 0.07 ***t*** **df** **p** BPRS     Total score 40.7 10.5 42.1 28.5 0.1 18 0.89 Anxiety/Depression 4.4 0.7 4.5 1.6 0.2 0.86     Anergia 9.2 3.0 10.2 5.6 0.5 0.6     Thought 12.6 3.8 10.1 4.8 1.3 0.2 Disturbance 4.9 2.3 4.2 2.1 0.7 0.5     Activation 11.8 4.1 15.6 22.4 0.5 0.6     Hostility MOAS 25.2 3.3 0.2 0.6 23.3 18 \< 0.001 *BPRS, Brief Psychiatric Rating Scale; MOAS, Modified Overt Aggression Scale*. The study was carried out in accordance with *The Code of Ethics of the World Medical Association*. Written informed consent was obtained from all participants and their families/guardians. The study was carried out in accordance with the details of relevant guidelines and was approved by the Biomedical Ethics Board of the Second Xiangya Hospital (Approval number: 2013068). A clinical interview was performed by a psychiatrist to confirm the diagnosis and to determine whether the patients met the inclusion and exclusion criteria. The Brief Psychiatric Rating Scale (BPRS) was used to evaluate the patients\' psychiatric symptoms. The 18-item BPRS is a 7-point scale that covers a broad range of symptoms including hallucinations, delusions, and mood disturbances. Here we report the total score and the factor scores for symptoms such as anxiety/depression, anergia, thought disturbance, activation, and hostility ([@B41]). The levels of aggression were assessed by the Modified Overt Aggression Scale (MOAS). The MOAS contains 4 differently weighted subscales, including verbal aggression, aggression against property, self-aggression, and physical aggression toward other people ([@B42], [@B43]). The total MOAS score was used as an overall measure of the aggression level according to the different weights of each subscale score. DNA extraction -------------- A sample of peripheral blood was collected from the participants and was immediately frozen at −80°C until DNA extraction. The RelaxGene Blood DNA System (TIANGEN Biotech, Beijing, China) was used to extract DNA following the manufacturer\'s instructions. Multiplex-PCR and high-throughput sequencing of the TRBV CDR3 region -------------------------------------------------------------------- A multiplex-polymerase chain reaction (PCR) system was used to amplify rearranged CDR3 regions. The PCR amplification conditions were 96°C for 3 min; 35 cycles of 96°C denaturation for 25 s, 62°C annealing for 45 s and 72°C extension for 45 s; followed by 72°C extension for 5 min. The remaining primers in the PCR product were digested by Exo I enzyme at 37°C for 15 min, and the enzyme was then denatured at 80°C for 20 min. PCR products were purified by gel extraction and sequenced on an ABI 3730XL machine. Sequencing results were analyzed with the ATF genotyping software. To prepare the TRBV library, 30 forward, and 13 reverse primers were designed to amplify the CDR3 region of the TRBV gene from genomic DNA template. The PCR products were purified with the AMPure XP system (Beckman Coulter, Inc, Indianapolis IN, USA) to remove PCR primers. Sequencing index sequences and adaptors were added to the library in another round of PCR with the conditions of 98°C for 1 min, 25 cycles of 98°C denaturation for 20 s, 65°C annealing for 30 s and 72°C extension for 30 s, with a final extension for 7 min at 72°C. The library was separated on an agarose gel, and the target region was isolated and purified using the QIAquick Gel Extraction kit. The PCR products were sequenced on the Illumina sequencing platform (Figure [1](#F1){ref-type="fig"}). {#F1} Data analysis ------------- Data quality was evaluated using the FastQC software. FLASH software was used to merge overlapping paired-end reads. The miTCR software, developed by MiLaboratory (<http://mitcr.milaboratory.com/downloads/>), was used to extract each clonotype from the CDR3 region. After sequence alignment, the expression level of each clonotype was calculated. The frequency and frequency distribution of clonotypes of the V and J gene segments, as well as the V-J gene pairs in the CDR3 region were also analyzed. To assess the TRBV repertoire diversity, the Shannon--Wiener index, the inverse Simpson diversity index, and the abundance ratios of different clonotypes were calculated ([@B44]). The V/J/VJ value of highly-expanded-clones (HECs) was calculated according to the definition of previous studies that TCR clones with a frequency of over 1% are considered to be HECs ([@B45]). Principal component analysis (PCA) was used to reduce the dimensionality of the highly expressed gene segments as described in a previous study ([@B46]). One-way ANOVA was used to compare the diversity levels among the three groups. A *p*-value of \< 0.05 was considered statistically significant. Results {#s3} ======= Demographic and clinical characteristics ---------------------------------------- The mean MOAS total score of violent patients was significantly higher than that of non-violent patients (*p* \< 0.001). There was no significant difference in either the total BPRS or the factor scores between the violent and non-violent patients (Table [1](#T1){ref-type="table"}). TCR repertoire diversity among different groups ----------------------------------------------- The numbers of total clonotypes and the numbers of unique clonotype in the violent patients, the non-violent patients, and the healthy control group were showed in Table [2](#T2){ref-type="table"}. The number of unique clonotypes in violent schizophrenic patients was significantly higher than that in non-violent schizophrenic patients (*p* \< 0.05). There were no significant differences in diversity indices such as the Shannon--Wiener and inverse Simpson diversity index between three groups (*p* \> 0.05; Table [2](#T2){ref-type="table"}). ###### Diversity of T-cell receptor repertoire in the violent, non-violent and healthy control groups. **Violent schizophrenia** **Non-violent schizophrenia** **Healthy control** --------------------------------------- ------------------------------------------------ ------------------------------- ------------------------------------------------ ----------- ------------ ----------- ------- --- ---------- TotalClonotype 3,016,151[^\*\*\*^](#TN1){ref-type="table-fn"} 1,182,819 3,360,785[^\*\*\*^](#TN1){ref-type="table-fn"} 3,721,625 12,717,786 1,198,254 44.09 2 \< 0.001 UniqClonotype 76,110Δ 38,414 39,534 22,578 48,892 11,297 4.76 2 0.02 TotalHighClonotype 2,981,641[^\*\*\*^](#TN1){ref-type="table-fn"} 1,167,216 3,341,443[^\*\*\*^](#TN1){ref-type="table-fn"} 3,708,148 12,693,699 1,200,168 44.48 2 \< 0.001 UniqHighClonotype 41,600Δ 20,956 20,192 9,356 24,805 5,221 6.35 2 0.01 Shannon\'s Diversity index 12.03 1.63 11.43 2.10 12.05 1.06 0.42 2 0.661 InvSimpsonDiversity 438.14 555.47 847.09 873.13 452.48 428.30 1.19 2 0.32 Normalized Shannon\'s Diversity index 0.75 0.09 0.76 0.14 0.77 0.07 0.16 2 0.86 Compare to healthy control group, p \< 0.001; Δ: Compare to non-violent schizophrenia group, p \< 0.01 *TotalClonotype: the total volume of clonotype expression; UniqClonotype: kinds of clonotype; TotalHighClonotype: the total expression of those clonotype express no \< 2; UniqHighClonotype: kinds of clonotype express no \< 2; Shannon\'s Diversity Index, InvSimpsonDiversity: the diversity index of clonotypes; Normalized Shannon\'s Diversity Index: the normalized Shannon\'s diversity index based on sample sequencing data*. TRBV CDR3 gene transcript abundance ----------------------------------- The top 20 most highly expressed V genes in each participant were listed in Figure [2](#F2){ref-type="fig"}. PCA revealed two principal components (PCs), PC1 explaining 28.88% and PC2 explaining 13.24% of the variation in the V genes (Figure [3](#F3){ref-type="fig"}). There were a significant differences between the schizophrenic patients (both violent and non-violent) and the healthy controls in V gene distribution (*p* \< 0.01; Figure [4](#F4){ref-type="fig"}). In particular, TRBV2 had significantly higher levels of expression in the violent schizophrenic patients than those of both the non-violent schizophrenic patients (*p* \< 0.01) and the healthy controls (*p* \< 0.01), whereas TRBV7-2 had significantly higher levels of expression in the non-violent schizophrenic patients than those of both the violent schizophrenic patients (*p* \< 0.01) and the healthy controls (*p* \< 0.01). Similar patterns were found in the J and V-J genes in all participants, and there were no significant differences in J and V-J gene pair distribution among the three groups. {#F2} {#F3} {#F4} Discussion {#s4} ========== To our knowledge, this is the first study using immune repertoire sequencing to investigate the characteristics and polymorphisms of the TRBV in violent and non-violent schizophrenic patients. The study found that schizophrenic patients (violent and non-violent) have different V gene distributions compared to those of healthy controls. In particular, the number of unique clonotypes in the violent schizophrenic patients was significantly higher than that in the non-violent schizophrenic patients. TRBV2 showed significantly higher expression in the violent schizophrenic patients, and TRBV7-2 showed significantly higher expression in the non-violent schizophrenic patients. Each T-cell bears a unique T-cell receptor in order to recognize a specific antigen-derived peptide. Recognizing MHC-bound peptides through TCRs, T cells mount the immune response in the adaptive immunity process. Some alterations in T-cell function in schizophrenia have been demonstrated, including reduced proliferative responses to stimulation, significant changes in transcripts associated with the cell cycle, intracellular signaling, oxidative stress and metabolis ([@B47]), and activation of T-cell networks ([@B48]). In addition, it has also been reported that altered immune function caused by T-cell molecular changes is associated with first-onset schizophrenia ([@B49]). The finding that V gene distribution in schizophrenic patients is different from healthy controls provides evidence that the dysregulation of T cells is associated with schizophrenia. TRBV provides the diversity and composition of the entire set of antigen receptors on T cells, which has an extraordinary impact on health and disease. Normally, without any antigen stimulation, T cells are in a positive polyclonal state, and TRBV is randomly rearranged. However, in the case of disease, specific antigen stimulation causes targeted TRBV subfamily rearrangement and dramatic clonal expansion. Therefore, the dominant T-cell clone may suppress the cloning of other T cells, resulting in a decrease in immune function ([@B27]). Clarifying TCR repertoire characteristics may help to explore the role of the immune system in the pathogenesis of schizophrenia and aid in the diagnosis and personalized treatment of patients. Recently, Li et al. reported immune factors (C3 and Th17-related cytokines) were related to schizophrenia and aggressive behavior ([@B11]), it suggested some immune factors would be potential biomarkers for aggressive behavior with schizophrenia. The immune repertoire sequencing is a new tool for deciphering mechanisms of autoimmunity. As the variety of TRBV genes have been indentified ([@B50]), and different distribution of TRBV subfamilies may result in decrease or increase some immune function ([@B51]). To our knowledge, there are some studies reported a significant skewed TRBV repertoire occurred, such as in major depressive disorder ([@B26]), and the systemic lupus erythematosus ([@B27]), acute graft-versus-host disease ([@B52]), and the maternal--fetal interface ([@B53]). In this study, the violent schizophrenic patients had higher numbers of unique expressed clonotypes than non-violent patients, and violent patients expressed higher levels of TRBV2, whereas non-violent patients expressed higher levels of TRBV7-2. TRBV subfamilies exhibit pedigree polymorphism and abnormal clonal proliferation under stimulation with specific antigens may cause a specific immune response. The highly oligoclonal nature of T cells in these patients may be associated with the different reactivity of each patient. The results of this study showed that violent schizophrenic patients have a different distribution of TRBV subfamily abnormalities, suggesting the abnormal expression of some TRBV subfamilies in PBMCs may be associated with the immune pathogenesis of schizophrenia. This provides a valuable clue to explore potential violence-specific T-cell antigens involved in immune-related schizophrenia pathogenesis, which may provide insight for the development of novel diagnostics and targeted immunotherapy. Till now, however, few studies reported the relation between immune factors and TRBV. Further study is needed to identify the function of the TRBV subfamilies evaluated *in vivo* and *in vitro*. Limitations {#s5} =========== This study is limited in some respects. First, as a heterogeneous disorder, schizophrenia produces an extraordinary variety and wide array of symptoms. Although, we used the BPRS to control the symptoms, violent and non-violent patients with schizophrenia may not be from uniform subtypes. Second, the sample size was small, and increasing the sample in future studies would provide opportunities for identifying additional oligoclonal or even monoclonal hyperplasia subfamilies. Finally, the function of this process and the clonal proliferation of T-cells in the body is largely unclear; further research is needed to conclusively reveal the molecular mechanisms of T-cell signaling. Author contributions {#s6} ==================== JZ and XW designed the study, QgL and WL wrote the protocol. JZ, XC managed the literature searches and analyses. QgL wrote the first draft of the manuscript. All authors contributed to and have approved the final manuscript. Conflict of interest statement ------------------------------ The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. We would like to thank the participants and their families for contributing to this research. **Funding.** This work was supported and funded by the National Natural Science Foundation of China (NSFC: 81371500, 81571316, 81571341), the Key Projects in the National Science & Technology Pillar Program during the Twelfth Five-Year Plan Period (No. 2012BAK16B00). There are no conflicts of interest to disclose. [^1]: Edited by: Xiang Yang Zhang, University of Texas Health Science Center at Houston, United States [^2]: Reviewed by: Hongsheng Gui, Henry Ford Health System, United States; Qiang WANG, West China Hospital of Sichuan University, China [^3]: This article was submitted to Behavioral and Psychiatric Genetics, a section of the journal Frontiers in Psychiatry

Introduction {#Sec1} ============ Chronic hepatitis B virus (HBV) infection is a major public health concern. Approximately 350 million individuals worldwide are estimated to be HBV carriers, and 650,000 deaths each year are associated with HBV infection^[@CR1],[@CR2]^. Asian populations have a much higher prevalence of HBV than other populations^[@CR3]^. As an endemic region, China has approximately 130 million HBV carriers and 30 million chronic HBV infections^[@CR4]^. HBV can infect people through acceptance of blood products or drug injection^[@CR5]^. However, due to the special immune status in pregnancies that allow mothers to tolerate the semiallogeneic fetus, mother-to-child transmission has become another major pathway for HBV infection, especially for mothers who are hepatitis B e antigen (HBeAg)- and hepatitis B surface antigen (HBsAg)-positive with high viremias^[@CR6],[@CR7]^. Indeed, infants born to HBsAg-positive mothers (0.6 to 5.8% of pregnant women) have an increased risk for potential chronic HBV infection^[@CR8],[@CR9]^. More than 90% of infants born from HBeAg-positive pregnancies will be infected with HBV if immunoprophylaxis is not accepted^[@CR10]^. In addition, HBV-infected infants have a higher risk of developing chronic HBV infection than those at other ages^[@CR11]^. Strikingly, mother-to-child transmission of HBV infection has decreased dramatically by 75--90% due to global immunoprophylaxis^[@CR1],[@CR2],[@CR12]--[@CR14]^. However, immunoprophylaxis still fails in 10--30% of infants, who develop chronic HBV infection through vertical transmission^[@CR15]--[@CR17]^. High viral burdens in mothers have been reported as the major factor underlying this prophylaxis failure and vertical transmission^[@CR2],[@CR10],[@CR18],[@CR19]^. Furthermore, HBV-infected children also have higher potential for liver cirrhosis and hepatocellular carcinoma. Thus, prevention of mother-to-child transmission, such as through antiviral treatment during pregnancy, has become the key issue to achieve the global goal of eliminating HBV infection. Currently, nucleotide analogues (NAs), including lamivudine, telbivudine and tenofovir disoproxil fumarate (TDF), are used to control HBV infection. As a nucleotide reverse transcriptase inhibitor, TDF is the only approved NA with high efficacy against the virus but no detected clinical resistance to date; thus, this drug has been widely used against HIV or HIV/HBV co-infection^[@CR20]^. In 2015, TDF treatment was recommended by the WHO against chronic HBV infection, particularly in pregnant women^[@CR21]^. However, as a result of limited efficacy and safety data, the FDA of the USA still categorizes TDF as a "B" class reagent. Indeed, controversial observations have been achieved for TDF application. For instance, application of TDF at a late stage of pregnancy reduced the serum HBV DNA level and vertical transmission of HBV^[@CR3],[@CR17],[@CR22]--[@CR24]^, but perinatal transmissions or adverse events were still reported in these trials; initiation of TDF treatment during the second trimester of gestation was also shown to block mother-to-child transmission but still had severe adverse effects^[@CR25]--[@CR28]^. Thus, evaluating the efficacy and safety of TDF treatment in pregnancies with HBV-infected mothers is critical. This clinical trial was designed for pregnancies with HBsAg/HBeAg double-positive mothers with serum HBV DNA ≥ 2×10^6^ (6.3 log10) IU/mL. TDF therapy was initiated at 24 weeks of gestation and continued to 4 weeks after delivery. Dynamic changes in the viral loads, vertical transmission and maternal/fetal outcomes were evaluated. Results {#Sec2} ======= As shown in the subject selection flow chart (Fig. [1](#Fig1){ref-type="fig"}), the pregnant women were randomly divided into two groups. In the TDF treatment group, one abortion occurred at 12 weeks of gestation, and the other 59 pregnancies accepted TDF therapy. Of these pregnancies, 1 case of premature delivery occurred at 32 weeks of gestation, and the premature infants died from hypoxic brain disease. In the control group, 8 subjects were lost to follow-up, and the other 52 pregnancies accepted routine nursing and prenatal care until the end of the study. In total, 110 pregnancies successfully participated in the whole trial process, and 110 infants were born. Table [1](#Tab1){ref-type="table"} shows the basic characteristics and clinical features of these pregnant women. No significant differences (p \> 0.05) were observed between the TDF-treated and control groups in terms of age, pregnancy times, treatment periods, basal ALT, HBV genotype, occupation, native place, residence, education level and family income.Figure 1The subject selection flow chart.Table 1The basic information and clinical features of the pregnant women.Groupstp valueTDF treatment (n = 59) Mean ± SD%Control (n = 52) Mean ± SD%Pregnant womenAge (y)28.31 ± 3.5628.06 ± 3.420.3790.705Pregnancy number1.28 ± 0.4511.25 ± 0.4370.3050.761Basal ALT level (U/L)54.62 ± 105.757.5 ± 103.3−0.1450.88HBV genotype0.809  B246.935.8  C25493.14994.2Family statusDomicile0.262  Rural3356.93567.3  City2543.11732.7Residency0.994  Temporary1932.81732.7  Resident3967.23567.3Education0.823  Below college3967.23669.2  College and above1932.81630.8Family incomes (year)0.976  \<40,200 RMB1831.01630.8  \>40,200 RMB4069.03669.2 To evaluate the effect of TDF treatment on viral replication, the serum HBV DNA levels during the pregnancies were dynamically monitored every 4 weeks up to 28 weeks postpartum. As shown in Fig. [2](#Fig2){ref-type="fig"} and the supplementary Table, prior to TDF treatment (24 weeks of gestation), the serum HBV DNA levels of the pregnant women were similar between the TDF-treated and control groups (7.44 ± 0.80 *vs*. 7.66 ± 0.55 log10 IU/mL, p=0.091). The HBV DNA levels in the control group exhibited no marked changes during the whole gestation period (Fig. [2](#Fig2){ref-type="fig"}, supplementary Table). Compared to that of the control group, TDF treatment significantly and gradually reduced the serum HBV DNA level (Fig. [2](#Fig2){ref-type="fig"}, supplementary Table).Figure 2Changes in the serum HBV DNA levels over time. From 24 weeks of gestation, the pregnant women were treated with or without TDF (300 mg/day, oral). The serum HBV DNA levels were monitored every 4 weeks until delivery. \*\*\*p \< 0.001 TDF treatment *vs*. control at the same time point; \#\#\#p \< 0.001 compared to the DNA level at the previous time point in the TDF-treated women. Paired comparisons between the patients who received TDF treatment showed that the serum HBV DNA level decreased by 4.82 ± 0.94 log10 IU/mL from 24 weeks of gestation to delivery (p \< 0.001, supplementary Table). By delivery, 90% of the patients with TDF treatment had a serum HBV DNA level of 2000 IU/mL or less; among them, 50% of the patients had HBV DNA levels less than 200 IU/mL. The results also showed that the serum HBV DNA decreased rapidly after initiating TDF treatment. Four weeks after initiating TDF therapy, a sharp decrease in HBV DNA (2.75 ± 1.19 log10 IU/mL) was observed (7.44 ± 0.8 *vs*. 4.68 ± 1.09, p \< 0.001) (Table [2](#Tab2){ref-type="table"} and supplementary Table). In addition, the serum DNA levels decreased continuously during TDF treatment (Fig. [2](#Fig2){ref-type="fig"}, Table [2](#Tab2){ref-type="table"}). Although the velocity of the decline slowed with time, it was still maintained at approximately 0.5 log10 IU/mL per month after the 3^rd^ month (32 to 36 weeks) of treatment (Table [2](#Tab2){ref-type="table"}), which kept the serum HBV DNA levels low during the remaining gestation period.Table 2The decrease in the HBV DNA levels after TDF treatment.Compared time pointsMeanSTDp value\*24 *vs*. 28 weeks2.751.19\<0.000128 *vs*. 32 weeks1.080.92\<0.000132 *vs*. 36 weeks0.570.59\<0.000136 *vs*. 40 weeks0.490.58\<0.0001^\*^p \< 0.05 indicates that the serum HBV DNA levels changed significantly between two time points. We also evaluated the effect of TDF on mother-to-child transmission. The vertical transmission rate was determined in infants 28 weeks after birth by examining the serum HBV DNA levels (\>20 IU/mL) or detection of HBsAg (positive). As indicated in Table [3](#Tab3){ref-type="table"}, the TDF treatment group had no mother-to-child transmission events with HBV infection, whereas a vertical transmission rate of 13.5% (7/52 individuals) occurred in the control group (p = 0.004). Although one premature infant died at 32 weeks of gestation in the TDF treatment group, this preterm infant was later confirmed not to be infected with HBV.Table 3The mother-to-child transmission rates and the serum HBV DNA recovery of the mothers at 28 weeks postpartum.TDF treatmentControlvaluepCases (n)%Cases (n)%Mother-to-child Transmission8.3380.004  Occurred00%713.5%  Did not occur59100%4586.5%Serum HBV DNA 28 weeks postpartum21.9160.000003  ≤2000 IU/mL2033.9%00%  Recovery to the original level3966.1%52100% To examine HBV recovery after cessation of TDF treatment 4 weeks postpartum, the serum HBV DNA levels during the pregnancies were checked at 28 weeks postpartum. The results showed that 33.9% (20/59 individuals) of the TDF-treated individuals maintained serum HBV DNA levels less than 2000 IU/mL, whereas 66.1% of the patients (39/59 individuals) had recovered to their original HBV level (Table [3](#Tab3){ref-type="table"}). Conversely, the HBV DNA levels in the control individuals did not change significantly before and after delivery and were always maintained at high levels compared to those of the TDF-treated patients, which was similar to their original levels and thus defined as 100% recovery (Table [3](#Tab3){ref-type="table"}). The emergence of drug-resistant variants was also evaluated by monitoring mutations of HBV genes in the TDF treatment and control groups using a direct gene sequencing technique. The results showed no gene mutations in HBV B2 or C2, which are the two major HBV strain genotypes in the Chinese population^[@CR29]^. The present study also investigated the outcomes of the mothers and children. As shown in Table [4](#Tab4){ref-type="table"}, no significant differences were observed between the mothers in the TDF and control groups during pregnancy, delivery and cesarean section. One case of miscarriage (1.72%) occurred at 12 weeks of gestation in the TDF treatment group prior to TDF therapy. As described above, another case of preterm delivery occurred in the TDF-treated group at 32 weeks of gestation (1.69% of the preterm birth rate), and the preterm infant died 3 days after birth from hypoxemic brain disease without HBV infection. In the control group, 2 cases of preterm delivery (3.85%) occurred. However, no significant difference (p = 0.495) in preterm delivery was observed between the TDF and control groups (Table [4](#Tab4){ref-type="table"}). In addition, 10.3% of the mothers from the TDF treatment group and 9.6% from the control group had cholestasis syndrome. No other obstetric complications were observed in either group. Furthermore, 3.45% of the patients from the TDF group (2 of 59 subjects) experienced nausea and vomiting, but they continued the TDF therapy with symptomatic treatment. No one in the control group had similar symptoms. Finally, approximately 10% of the patients had itchy skin in both the TDF and control groups (Table [4](#Tab4){ref-type="table"}). No marked differences were observed between the groups in either digestive system or skin symptoms. No other adverse reactions appeared in the two groups of pregnant women.Table 4Maternal outcomes between the TDF treatment and control groups.TDF Group Mean ± SD or n (%)Control Group Mean ± SD or n (%)p**Mothers**Pregnancy weeks39.46 ± 1.4339.33 ± 1.490.645Cesarean section, n (%)21 (36.2%)17 (32.7%)0.699**Obstetric complications, n (%)**Pregnancy hypertension00Placenta previa00Intrahepatic cholestasis of pregnancy6 (10.3%)5 (9.6%)0.899Postpartum hemorrhage00Limited fetal growth00Preterm birth1 (1.69%)2 (3.85%)0.495Abortion1 (1.72%)00.342**Adverse reactions, n (%)**Digestive tract reaction: vomiting2 (3.45%)00.177Fatigue00Headache00Cough00Fever00Itchy skin6 (10.3%)5 (9.6%)0.899Jaundice00**Laboratory examinations, n (%)**ALT, U/L \>5 × ULN\*2 (3.45%)00.177Renal dysfunction00Decrease of PLT00Anemia00^\*^ULN: upper limit of normal. None of the mothers in the present study breastfed their infants. A total of 58 healthy newborns were delivered in the TDF treatment group and 52 newborns were born in the control group, since 8 pregnancies were lost to follow-up during the study. Among all 110 surviving infants, 2 cases from the TDF treatment and 1 from the control group had mild asphyxia immediately after birth according to the Apgar score at 1 min (Table [5](#Tab5){ref-type="table"}). No significant differences were found between the groups in the average body weight and length. Neither congenital malformations nor adverse symptoms were observed in any of the infants after birth. The serum ALT levels were similar in the infants in both groups. However, among the infants from the control group, 4 were HBsAg-positive (7.7%), 5 were HBeAg-positive (9.6%) and 7 cases (13.5%) had detectable HBV DNA (Table [5](#Tab5){ref-type="table"}). None of the infants from the TDF-treated mothers were positive for HBsAg, HBeAg or HBV DNA, which was a marked difference compared to those of the control group (p \< 0.05).Table 5Infant outcomes between the TDF treatment and control groups.InfantsApgar score 1 min after birth, n (%)0.613  8\~10 normal55 (96.5%)51 (98.1%)  4\~7 mild asphyxia2 (3.5%)1 (1.9%)  0\~3 severe suffocation00Weight (g)3259.47 ± 369.033338.08 ± 363.040.265Height (cm)49.89 ± 0.8249.92 ± 0.840.859Congenital malformations, n (%)00**Adverse reactions, n (%)**Fever00Cough00Vomiting00Jaundice00**Laboratory examinations, n (%)**ALT (U/L)22.25 ± 11.9224.27 ± 18.830.501HBsAg-positive04 (7.7%)0.033HBeAg-positive05 (9.6%)0.017HBV DNA-positive0 (0%)7 (13.5%)0.004 Discussion {#Sec3} ========== Benefiting from discovery of the HBV vaccine, the morbidity of HBV infection has decreased significantly. However, chronic HBV infection remains a large public health concern, particularly in the high endemic regions of Asia. Perinatal transmission is a major cause of this chronic infection. Indeed, HBV vaccine and immunoglobulin therapy have been shown to only prevent infection during the labor and postnatal periods but have no effect on intrauterine infection^[@CR30]^. The HBV DNA level in pregnant women is also closely associated with the vertical transmission rate and immunoprophylaxis failure^[@CR19]^. Furthermore, the repressed immune system during pregnancies^[@CR6],[@CR7]^ also increases the potential risk of HBV proliferation during pregnancy. Thus, pregnant women with HBV infection, especially those who are HBeAg/HBsAg-positive with high viral loads, must control their viremia with antiviral treatment, such as TDF therapy, to decrease the risk of mother-to-child transmission of HBV and to maintain the health of the pregnancies. The EASL 2017 Clinical Practice Guidelines for the management of HBV infection recommended TDF treatment at 24--28 weeks of gestation during pregnancy^[@CR31]^. However, limited data on the efficacy and safety of the application of TDF to control HBV infection during pregnancy are available, especially for the second trimester of gestation. The present study aims to investigate and evaluate the effect of TDF initiated at 24 weeks of gestation on the efficacy and safety, therapy duration, drug resistance and vertical transmission. In the present work, we dynamically monitored the effect of TDF on the serum HBV DNA levels and found a total reduction of 4.82 ± 0.94 log10 IU/mL from 24 weeks of gestation to delivery (p \< 0.001). Interestingly, a decline of 2.75±1.19 log10 IU/mL of HBV DNA was detected 4 weeks after TDF initiation. However, the pregnancies in the control group, which had similar viral loads to the TDF-treated pregnancies before initiating the therapy, had unchanged viral loads during gestation. In line with our finding, a gradual decrease in the serum HBV DNA level induced by TDF was observed in another study^[@CR28]^, although the negative control (pregnancies without TDF treatment) was lacking in their study. All other clinical trials have only monitored the HBV viral burden before TDF treatment and after delivery. Our data are also intriguing because they show that after the third trimester (32 weeks of gestation), 90% of the patients with TDF treatment have serum HBV DNA levels of 2000 IU/mL or less until delivery and 50% have viral loads lower than 200 IU/mL. This finding suggests that TDF therapy initiated at 24 weeks of gestation can maintain the serum HBV DNA at lower levels during pregnancy until delivery, which may be the ideal condition to limit viral replication and ensure fetal growth. Previous studies have demonstrated that initiation of TDF treatment at the third trimester (30--32 weeks) of gestation reduces perinatal transmission, but mother-to-child transmission of HBV infection still occurs^[@CR3],[@CR17],[@CR22]--[@CR24]^. As a prospective investigation, the current trial started TDF therapy at 24 weeks of gestation, and no vertical transmission was observed in the infants. Wang *et al*. showed a similar result, although a small sample of subjects (21 patients) was included in their study^[@CR28]^. In another three trials, TDF was given as early as 5, 12 and 17 weeks of gestation, and no HBV-infected infants were observed^[@CR10],[@CR25],[@CR26]^. However, these studies were retrospective studies that included small numbers of individuals with TDF treatment. These findings encouraged us to evaluate the efficacy and safety of TDF with more patients. The present study and other trials from China^[@CR32],[@CR33]^ have shown an immunoprophylaxis failure rate of up to 15% in infants born to HBsAg-positive mothers in the control group. However. clinical trials from other countries, such as Thailand, showed a lower rate of immunoprophylaxis failure (\<5%)^[@CR34]^. The following factors may account for this discrepancy: 1) the timeframe of immunoprophylaxis administration; indeed, the infants received immunoprophylaxis at approximately 1 hour after birth in the study from Thailand compared to vaccination of infants within 12 h after birth in China; 2) the vaccine dose; the routine vaccine series (3 doses at 0, 1 and 6 months after birth) are applied in China, whereas infants receive 5 doses of the vaccine (0, 1, 2, 4 and 6 months) in Thailand; and 3) HBV DNA levels of the mothers; trials from various countries have shown that immunoprophylaxis failure in infants is closely associated with the mother's viral load at delivery^[@CR35]--[@CR37]^. Jourdain *et al*. also declared that the mothers of HBV-infected infants from Thailand had high viral loads at delivery^[@CR34]^. In addition to the decline in the serum HBV DNA levels induced by TDF treatment, TDF therapy was also reported to result in seroclearance of HBeAg and HBsAg or HBeAg seroconversion^[@CR17],[@CR22],[@CR26]^. Marcellin *et al*. even showed that HBeAg could not be detected in 45% of patients after TDF treatment^[@CR26]^, which might be associated with the HBeAg levels prior to initiation of TDF therapy. However, the present study did not observe any seroclearance/seroconversion for either HBeAg or for HBsAg in either the TDF-treated or control mothers. The potential reasons are that all pregnancies in our study 1) have a high titration of HBeAg and HBsAg with a high viremia and 2) are tolerant to HBV infection since they have high viral loads (\>2×10^6^ IU/mL) and a relatively normal ALT level. Concerning the safety of TDF treatment, we did not detect any congenital malformations during gestation and at 6 months after birth. A previous study revealed that a high dose of TDF but not the dose used for humans affected fetal growth in primates^[@CR38]^. Indeed, TDF treatment of pregnancies with chronic HBV infection has been associated with adverse events, including a low birth weight, premature delivery and congenital abnormalities, regardless of whether TDF is given at the second or third trimester of gestation^[@CR3],[@CR10],[@CR17],[@CR23],[@CR24]^, although these side effects are much lower than those induced by other NAs. The current study also observed two premature deliveries at 33 and 34 weeks of gestation from the control group, of which both infants survived. However, a preterm delivery occurred in the TDF-treated group, and the infant died from hypoxic encephalopathy; we found no evidence of a direct association with TDF treatment, although we could not exclude this possibility. Therefore, a large amount of complementary data are still needed to further evaluate the safety of TDF on pregnancies and infants. Because the children were followed up to 28 weeks after birth in the present study, the long-term or delayed potential effects of TDF on the infants (e.g., growth and intelligence) also need to be addressed in the future. TDF-treated mothers have breastfed their children without reported complications^[@CR26]^. However, according to the guidelines from the Chinese Foundation for Hepatitis Prevention and Control, breastfeeding should not be recommended since the potential risk for the infants is still uncertain; further studies are needed to clarify the safety. Drug resistance or tolerance is a big problem during chronic hepatitis B infection; for instance, a Lamivudine (LAM)-resistant viral variant has been reported after short-term therapy^[@CR39]^. Our present study did not detect resistant variants of HBV even when we started treatment at 24 weeks of gestation and continued it to 4 weeks postpartum. Consistent with our observation, other studies also did not show the emergence of resistant variants of HBV during TDF therapy at the third trimester of gestation of pregnancy^[@CR3],[@CR17],[@CR22],[@CR24]^, which might be due to the high resistance barrier of TDF^[@CR5]^. In the present trial, the women stopped TDF treatment at 4 weeks postpartum. Notably, 33.8% of the subjects maintained lower serum HBV DNA levels under 2000 IU/mL at 28 weeks postpartum, whereas the HBV DNA level recovered to the original level in the remaining individuals. In addition, only 3.4% of the TDF-treated patients had liver function more than 5 times the upper limit of normal without other clinical manifestations. No patients had kidney dysfunction in TDF treatment group. In line with our observations, Nguyen *et al*. reported that regardless of whether TDF was immediately withdrawn after delivery or continuously used, this medicine did not affect the risk of HBV DNA recovery in pregnancies^[@CR40]^. However, another study recommended close monitoring up to 6 months after delivery for women who were HBeAg positive or had stopped TDF treatment^[@CR41]^. Nevertheless, our study only followed up the patients to 28 weeks postpartum, and thus further intensive monitoring and follow-up are needed. In summary, initiation of TDF treatment at the second trimester (we started at 24 weeks) of gestation in HBsAg/HBeAg-positive pregnant women with high viral loads (≥2×10^6^ IU/mL) dynamically and efficiently reduced the serum HBV DNA levels. This reduction was apparent at the first month of TDF therapy, which facilitated maintenance of HBV DNA at the lower level during the following gestation periods of the pregnancies. Treatment apparently reduced the risk of vertical transmission and ensured an optimal uterine microenvironment for the fetus. Our results demonstrated that application of TDF at 24 weeks of gestation in HBV-infected pregnancies was safe and that no obvious complications for the mothers, severe outcomes or adverse effects were observed. These results indicate the efficacy and safety of TDF therapy at the middle stage of gestation in HBV-infected pregnancies, particularly for those who have high viral loads and are HBsAg/HBeAg-positive, to prevent mother-to-child transmission. Methods {#Sec4} ======= Study design and research subjects {#Sec5} ---------------------------------- The multicenter cohort study was performed from January 2013 to December 2016 in different hospitals located in northwest China. The trial was approved by the Ethics Committee of the First Affiliated Hospital of Xi'an Jiaotong University (Xi'an, China), and the date of registration was March 25, 2016 (registration number: NCT02719808). The study was performed in accordance with the guidelines and regulations for the prevention of vertical transmission of hepatitis B in pregnancy (Chinese Medical Association-Society of Obstetrics and Gynecology, 2013). The details of the full protocol are available in the online complementary data section. The flow chart of study subject selection is shown in Fig. [1](#Fig1){ref-type="fig"}. The eligibility criteria were 20--35-year-old pregnant women who were HBeAg/HBsAg double-positive with a serum HBV DNA titer ≥2×10^6^ (6.3 log10) IU/mL. The exclusion criteria for the subjects were ① co-infection with HIV, HCV or HDV, ② a HBV treatment history within 6 months, ③ an abortion history or clinical manifestation of an inevitable abortion, ④ congenital deformity of the fetus, ⑤ evidence of hepatocellular carcinoma, renal or hepatic dysfunction, a creatinine clearance rate \<100 mL/min, ALT \>5 times the upper limit of normal, or bilirubin \>2 mg/dl, ⑥ hemoglobin \<8 g/100 mL, neutrophils \<1000/mm^3^, or albumin \<2.5 g/100 mL, ⑥ special medicine treatment required during the pregnancy, and ⑦ the biological father of the infant has chronic HBV infection. A total of 120 women were finally enrolled in the study according to the criteria. All participants were informed about the study, volunteered and signed the written consent form. The parallel study was designed with TDF-treated and control groups. The sample size was determined using an online sample size calculator for two parallel-sample proportions. A random number table was used to group the pregnancies into each group (60 individuals per group) based on their enrollment time. Simple randomization was performed, and sealed envelopes were used for concealment of the random allocation. TDF treatment (300 mg/day, oral, GSK, China) was initiated at 24 weeks of gestation and continued to 4 weeks postpartum. The control individuals did not receive anti-viral treatment. All subjects were followed up every 4 weeks to 28 weeks postpartum, and the laboratory parameters were monitored at every visit. All infants were given routine immunoprophylaxis, such as hepatitis B immune globulin (HBIG), immediately after birth and the HBV vaccine within 12 hours after birth. The second and third doses of the vaccine were given to the infants at ages 1 and 6 months, respectively. Due to insufficient long-term safety data for infants exposed to TDF through breastmilk^[@CR24]^, none of the infants were breastfed according to the guidelines from the Chinese Foundation for Hepatitis Prevention and Control. The infants were also monitored from their birth to 7 months of age. During the double-blind study, the participants did not know which type of intervention they accepted until the end of the intervention. The participants, care providers and persons who examined the viral DNA loads and evaluated the outcomes of the patients did not know whether the patients had accepted the intervention. Laboratory evaluations {#Sec6} ---------------------- In accordance with the updated Chinese guideline on "management algorithm for interrupting mother-to-child transmission of hepatitis B"^[@CR42]^, serological testing was performed in the infants 28 weeks after birth (one month after the final dose of the HBV vaccine). The occurrence of mother-to-child transmission was the primary outcome and was determined when the serum HBV DNA of the infant was higher than 20 IU/mL or the infant was HBsAg-positive 28 weeks after birth. Congenital malformations were defined as anatomical abnormalities of the fetus before delivery or of infants 28 weeks after birth and were verified by clinical examinations, radiographic examinations or other experimental methods during follow-up. The rate of congenital malformations reflects the proportion of infants with defects out of all safe-born infants. To evaluate the effect of TDF, the serum HBV DNA level was monitored in all pregnancies with/without TDF treatment. The assay was performed before initiating TDF therapy and then continued every 4 weeks until immediately before delivery. After delivery, this procedure continued to 28 weeks postpartum to evaluate recovery of HBV DNA after cessation of TDF treatment, which was defined as the secondary outcome. Total DNA was extracted from the patient's blood using the DNAout kit (Tianenze, Beijing, China). The RT region (NT 54--1278) was amplified (Chinese patent ZL 200910092331.1) as described previously^[@CR43]^. Briefly, the sense and antisense primers for the first-round PCR were 5′-agtcaggaagacagcctactcc-3′ (NT 3146--3167) and 5′-aggtgaagcgaagtgcacac-3′ (NT 1577--1596), respectively. The primers for the second-round PCR were 5′-ttcctgctggt-ggctccagttc-3′ (NT 54--75) and 5′-ttccgcagtat-ggatcggcag-3′ (NT 1258--1278). The PCR products from all patients were sequenced to check whether gene mutations occurred after TDF treatment during the pregnancy period. In the present study, ALT above 5 times the upper limit of normal was identified as a significant adverse event clinically regardless of whether the patients had symptoms. These patients needed to be further monitored and followed up or treated. Statistical analysis {#Sec7} -------------------- The data were presented as the mean±SE, and SPSS 18.0 (SPSS Inc., Chicago, IL, USA) was used to conduct the analysis. Baseline characteristics and safety outcomes were compared between the TDF treatment and control groups using t-tests for continuous variables and the Chi-square test or Fisher's exact test for categorical variables. The mother-to-child transmission rate and the recovery level of the HBV DNA 28 weeks after delivery were compared between the two groups using Fisher's exact test. A mixed model repeated measures analysis of variance was used to analyze whether the serum HBV DNA level changed differently with or without TDF treatment. Planned contrast was used to test the "change in HBV DNA for each group between baseline and delivery time" and "differences from week to week' within the TDF treatment group. A p value less than 0.05 was considered statistically significant. Electronic supplementary material ================================= {#Sec8} Supplementary table Supplementary information **Publisher's note:** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Electronic supplementary material ================================= **Supplementary information** accompanies this paper at 10.1038/s41598-018-33833-w. The authors acknowledge funding from the National Natural Science Foundation of China (National Science Foundation of China) - 81771615 and The General Program of Innovation of Science & Technology of Shaanxi Province (N°2016KTCL03-01). Y.Y.L. managed the data; Y.L. detected gene mutation of HBV serotype; G.F.D. and W.M.W. collected the patients' data; L.T. critically read the manuscript; N.X. collected and managed the data; K.Y.L. performed part of the statistics; L.Y.Z. collected the data; D.J.C. participated in the design of the project; Y.Z.W. wrote the manuscript; G.Q.B. designed, supervised and participated in the project. The protocol used for this study is included in this published article (Supplementary files). All data generated or analyzed during the study are available from the corresponding author on the reasonable request. Competing Interests {#FPar1} =================== The authors declare no competing interests.

The following article is being retracted. Sehgal R, Misra S, Anand N, Sharma M. Microarray in parasitic infections. Trop Parasitol 2012;2:6-12 **Editor, Trop Parasitol**

Introduction ============ As a common type of inflammatory arthritis, gout is typically characterized by an elevated concentration of serum uric acid (hyperuricemia) and a subsequent deposition of monosodium urate (MSU) crystals in and around joints ([@b1-ijmm-45-04-1047]-[@b3-ijmm-45-04-1047]). The prevalence of gout in developed countries has been reported to be 1-3.9% ([@b1-ijmm-45-04-1047]-[@b3-ijmm-45-04-1047]). In addition, due to changes in dietary habits and an increasing aging population, the prevalence of gout has increased over past decades in China ([@b4-ijmm-45-04-1047],[@b5-ijmm-45-04-1047]). Patients with gout often present with metabolic diseases, including diabetes mellitus, hypertriglyceridemia and obesity, in addition to hypertension ([@b6-ijmm-45-04-1047]). In total, 16% of patients with gout exhibit ischemic heart disease and metabolic syndrome ([@b7-ijmm-45-04-1047]). Insulin resistance serves a key function in the pathogenesis of metabolic syndrome. Hyperuricemia in patients with metabolic syndrome is likely caused by insulin resistance, since insulin promotes uric acid and sodium reabsorption in the proximal tubule ([@b8-ijmm-45-04-1047],[@b9-ijmm-45-04-1047]). Gout is etiologically heterogeneous as environmental and genetic components are involved ([@b1-ijmm-45-04-1047]-[@b3-ijmm-45-04-1047]). Although an unhealthy diet and lifestyle are risk factors for hyperuricemia, genetic factors were revealed to serve a greater function in the development of gout compared with environmental factors ([@b10-ijmm-45-04-1047]-[@b12-ijmm-45-04-1047]). Various previous genome-wide association studies (GWASs) have improved the understanding of the genes that regulate serum uric acid levels and increase susceptibility to gout ([@b13-ijmm-45-04-1047]-[@b15-ijmm-45-04-1047]). Among the loci revealed to be associated with gout, solute carrier family 2 member 9 and ATP-binding cassette super-family G member 2 (*ABCG2*) are the two most important genes identified to serve a function in gout development ([@b10-ijmm-45-04-1047]-[@b12-ijmm-45-04-1047]). Two important pathways, including renal and gut excretion of uric acid, have been revealed to regulate hyperuricemia levels, and each of these pathways are associated with glycolysis ([@b16-ijmm-45-04-1047]). However, previous studies based on GWAS research in independent individuals were only able to explain \~7% of the variance in serum urate concentrations, and only a portion of those loci were confirmed to be associated with the risk of gout ([@b17-ijmm-45-04-1047],[@b18-ijmm-45-04-1047]). Gout is heritable and has tendency to cluster in families ([@b12-ijmm-45-04-1047],[@b19-ijmm-45-04-1047],[@b20-ijmm-45-04-1047]). In total, \~20% of patients with gout were reported to have a family history of this disease. The risk of gout is significantly higher in patients who have first-degree relatives affected by this disease compared with the normal population ([@b19-ijmm-45-04-1047]). A previous segregation and linkage analysis of familial gout revealed an autosomal-arbitrary major gene model, which indicated a genetic basis for familial gout ([@b20-ijmm-45-04-1047]). A previous population-based study was performed in a Taiwanese population to estimate the degree of familial aggregation of gout, and it was revealed that genetic factors serve a substantial function in the development of gout ([@b19-ijmm-45-04-1047]). The present study aimed to identify the potential pathogenic genetic causes of 3 pedigrees with a familial aggregation of gout using whole-exome sequencing (WES) technology and diverse bioinformatics analysis, including genetic interaction networks, disease ontology (DO) and gene ontology analysis. Materials and methods ===================== Patient recruitment ------------------- Patients (n=10) who were attending a clinic for arthritis were recruited between July 2016 and June 2017. Additionally, three families were recruited in the present study as: i) Presenting with autosomal dominant inheritance and ii) healthy family members agreed to participation in the present study. Patient information is summarised in [Table I](#tI-ijmm-45-04-1047){ref-type="table"}. The present study was conducted according to the principles outlined in the Declaration of Helsinki. The study protocol was approved by the Ethics Committee of The First Affiliated Hospital, Wenzhou Medical University (Wenzhou, China; approval no. 2018-020). Written informed consent was obtained from all participants or their guardians. Patients with gout were clinically evaluated by physicians based on the 2015 gout classification criteria by the American College of Rheumatology/European League Against Rheumatism Collaborative Initiative ([@b21-ijmm-45-04-1047]). Patients with inherited metabolic disorders, including Lesch-Nyhan syndrome, were excluded from the present study. Autoantibodies and human leukocyte antigen B27 (HLA-B27) tests were negative in all patients and X-ray analyses of the affected joints were performed to exclude other potential diseases. The classical symptoms of patients with gout are characterized by the rapid development of monoarticular arthritis, which is accompanied by swelling and redness of the first metatarsophalangeal joint (MTP1). WES and variant calling ----------------------- The human genome hg19 was used as the reference genome. Genomic DNA was extracted from peripheral blood leukocytes. In total, 2 *µ*g genomic DNA from each sample was sheared to fragment with a length of 150 base pairs (bp) using the Covaris S220. Subsequently, ligation of small fragments of DNA with A-tail and adapters was conducted. A genomic library was constructed subsequent to the amplification of adapter-ligated DNA using an Agilent SureSelect Library Prep kit (Agilent Technologies, Inc.) according to the manufacturer\'s protocol, and the samples from each individual were marked with a unique index. Whole-exome capture was performed using the Agilent SureSelect Human All Exon v5 kit (Agilent Technologies, Inc.) according to the manufacturer\'s protocol. High throughput sequencing was performed using an Illumina HiSeq 4000 sequencer (Illumina, Inc.). All raw sequencing data obtained from these three families were processed in a similar manner, according to a customized bioinformatics pipeline ([@b22-ijmm-45-04-1047]). To remove sequence adapters and low-quality reads, the raw reads were filtered using the FastQC software program, version 1.11.4 (<http://www.bioin-formatics.babraham.ac.uk/projects/fastqc/>). The filtration criteria excluded Phred-scaled quality scores \<30 and read lengths \<80 bp. Subsequent to removing the low-quality reads, the remaining reads were used for further analysis. FastQ reads were aligned to the human reference genome (GRCh37/hg19) using the Burrows-Wheeler alignment tool ([@b23-ijmm-45-04-1047]) and further visualized using SplicingViewer software ([@b24-ijmm-45-04-1047]). Then, Genome Analysis ToolKit (GATK; version 4.0.10.0; <https://gatk.broadinstitute.org/hc/en-us>) was used to remove duplicated reads and reads mapped to multiple genome locations. In addition, local realignment and map quality score recalibration were performed. Candidate variants were then identified using the GATK Unified Genotype (version 4.0.10.0; <https://gatk.broadinstitute.org/hc/en-us>). Variant annotation and prioritization ------------------------------------- mirTrios with an integrated ANNOVAR tool were used to annotate all the detected mutations according to an in-house pipeline ([@b25-ijmm-45-04-1047]). The minor allele frequency (MAF) was obtained for each variant from publicly available databases ([@b25-ijmm-45-04-1047]), including ExAC, UK10K, dbSNP147, 1000 Genomes and ESP6500, and from in-house exome data. The detected variants with a MAF \>0.01% present in any of the aforementioned databases were removed ([@b26-ijmm-45-04-1047]). Subsequently, the effects of the detected variants were predicted using four tools: i) Polymorphism Phenotyping v2 (PolyPhen2) ([@b27-ijmm-45-04-1047]); ii) Likelihood Ratio Test (LRT) (<https://evomics.org/resources/likelihood-ratio-test/>); iii) MutationTaster ([@b28-ijmm-45-04-1047]); and iv) Functional Analysis through Hidden Markov Models (FATHMM) ([@b29-ijmm-45-04-1047]). A missense variant was considered deleterious if the variant was predicted to be deleterious or damaging by ≥3 of the four genetic prediction tools. The remaining variants were considered to be high-confidence causative variants. Protein structural modeling --------------------------- The amino acid sequence of human protein kinase CGMP-dependent 2 (PRKG2) was retrieved from Uniprot (<http://www.uniprot.org/uniprot/Q13237>). Crystal structures of the wild-type PRKG2 protein were obtained from the protein data bank (PDB; <http://www.rcsb.org/structure/5BV6>) and visualized using PyMol software (version 1.8) (<https://pymol.org/2/>). Homology modelling of mutated PRKG2 protein structures was performed using SWISS-MODEL (<https://swissmodelexpasy.org/>.) ([@b30-ijmm-45-04-1047]). In total, 120 residues were modelled in the PRKG2 structure, including 102 residues in the cGMP-binding region. Collection of candidate genes for hyperuricemia and gout -------------------------------------------------------- Various candidate genes for hyperuricemia and gout were reported in previous studies. The literature search was performed in June 2017 by searching \'gout\' AND \'genes\' in PubMed (<https://www.ncbi.nlm.nih.gov/pubmed>). Then all genes in all literature obtained were included. Following the literature search, the candidate genes associated with hyperuricemia and gout were collected and are listed in [Table SI](#SD1-ijmm-45-04-1047){ref-type="supplementary-material"}. Construction of gene co-expression and protein-protein interaction (PPI) networks --------------------------------------------------------------------------------- Temporally rich transcriptome data extracted from the Genotype-Tissue Expression (GTEx) project (<https://gtexportal.org/home/>, accessed January 2018) were used to build the co-expression network. The Pearson correlation coefficients (r) for the gene co-expression levels were calculated for each pairwise combination between different genes. To investigate similarities among the genes forming the PPI network and analyse their functions, significantly enriched DO terms were identified using the R package DOSE (version 2.0) and a hypergeometric test ([@b31-ijmm-45-04-1047]). To assess function similarities between previously reported genes ([Table SI](#SD1-ijmm-45-04-1047){ref-type="supplementary-material"}) associated with gout and the core risk genes identified in the present study, a PPI network was built. PPI data downloaded from STRING V10 (<https://string-db.org>) ([@b32-ijmm-45-04-1047]) were used for PPI network analysis. To further investigate the gene functions in the PPI network, biological processes (BP) analysis was conducted using ClueGO v2.3.3 ([@b33-ijmm-45-04-1047]), a plug-in of Cytoscape. Results ======= Characteristics of the patient cohort ------------------------------------- The patient cohort for the present study was composed of three families ([Figs. 1A](#f1-ijmm-45-04-1047){ref-type="fig"}, [2A](#f2-ijmm-45-04-1047){ref-type="fig"} and [3](#f3-ijmm-45-04-1047){ref-type="fig"}). All patients experienced acute monoarticular arthritis affecting the MTP1 and/or knee. The affected joint made walking difficult, was painful to the touch and was occasionally accompanied by fever. Tophus was observed in some of the patients ([Fig. 1B](#f1-ijmm-45-04-1047){ref-type="fig"}), and metabolic disorders, including hypertension, diabetes and hyperlipemia, were identified. The clinical data are presented in [Table I](#tI-ijmm-45-04-1047){ref-type="table"}. The serum urate level was \>480 *µ*mol/l in most cases, and in numerous cases it was \>600 *µ*mol/l. Synovial fluid from a number of the patients presented MSU crystals and erosions based on conventional radiography of the affected joint. Autoantibodies and HLA-B27 tests were negative in all patients ([Table SII](#SD1-ijmm-45-04-1047){ref-type="supplementary-material"}). All affected patients were diagnosed with gout. Detection of candidate deleterious mutations -------------------------------------------- To determine the genetic etiology of these families, WES was performed in the three probands. In total, \~13.53 Gb high-quality data was obtained on mean subsequent to removing sequencing adapters and low-quality sequences ([Table SIII](#SD1-ijmm-45-04-1047){ref-type="supplementary-material"}). For each sample, ≥98.68% of the high-quality data was matched with the human reference genome Hg19. Effective sequence on target was \>4.91 Gb, with a minimum of 39.10% fraction of effective bases on target following the removal of polymerase chain reaction duplications. The mean sequencing depth for each sample was 110.16-fold, with \>99.00% of target regions being covered at a 4-fold sequencing depth, 98.50% at 10-fold depth and 97.20% at 20-fold depth. Collectively, the quality control data demonstrated a high reliability, which was fundamental for the subsequent analyses. Subsequent to removing low-quality reads, adapters and duplicated reads from the raw sequencing data, a total of 541,954 single nucleotide variations (SNVs) and 84,415 indels were identified using the GATK tool, which included 67,707 SNVs and 2,200 indels in exonic and splicing regions. Subsequent to applying variant filtration against multiple databases, the number of rare SNVs and indels causing protein change with MAF \<0.001 was reduced to 228 and 29, respectively. As a result, following effect prediction by Polyphen2, LRT, Mutation Taster and FATHMM, the variants predicted to be deleterious by \>2 prediction tools were validated by Sanger sequencing. Finally, three SNVs in the coding regions of three gout-associated genes were retained and confirmed following Sanger sequencing validation ([Table II](#tII-ijmm-45-04-1047){ref-type="table"}). Sanger sequencing validation and co-segregation testing ------------------------------------------------------- Analysis of WES data indicated a missense mutation (p.Lys157Glu) in the *ABCG2* gene in the proband F1:II:1 of family 1 presenting typical symptoms of gout ([Fig. 1](#f1-ijmm-45-04-1047){ref-type="fig"} and [Table I](#tI-ijmm-45-04-1047){ref-type="table"}). The ABCG2 protein is involved in the excretion of urate from the intestine and kidney, and its dysfunction causes extrarenal urate underexcretion type ([@b34-ijmm-45-04-1047]) and/or renal urate underexcretion type gout ([@b35-ijmm-45-04-1047]). The heterozygous missense mutation (c.469A\>G) causing a lysine (Lys) to glutamic acid (Glu) substitution was located at the amino acid 157. Subsequently, Sanger sequencing confirmed the presence of this mutation in the patient\'s affected father while his unaffected mother and sister did not present the A to G change at cDNA nucleotide 469 ([Fig. 1A](#f1-ijmm-45-04-1047){ref-type="fig"}). The MAF of p.Lys157Glu was 8.24×10^−6^ and was predicted to be deleterious by all four effect prediction tools ([Table II](#tII-ijmm-45-04-1047){ref-type="table"}). Additionally, the residue 157Lys is highly conserved among different vertebrate species ([Fig. 1C](#f1-ijmm-45-04-1047){ref-type="fig"}). Subsequently, the tissue specificity of the expression of the human *ABCG2* was examined in all major tissues and organs using the Human Protein Atlas database (<https://www.proteinatlas.org>). *ABCG2* exhibited the most abundant protein expression in the small intestine ([Fig. 1D](#f1-ijmm-45-04-1047){ref-type="fig"}). The results reflected abundant mRNA expression in the luminal membrane of the intestine with 136.8 transcripts per million (TPM). Another heterozygous missense mutation (c.752A\>G; p.Asn251Ser) located in exon 5 of *PRKG2* gene was detected in proband F2:II:1 and his affected brother F2:II:9, as well as in the patient III9 ([Fig. 2A](#f2-ijmm-45-04-1047){ref-type="fig"}). cGMP-dependent protein kinase 2 (cGKII)/*PRKG2* is involved in the regulation of aldosterone and renin secretion ([@b36-ijmm-45-04-1047]). In total, nine unaffected family members presented wild-type alleles ([Fig. 2A](#f2-ijmm-45-04-1047){ref-type="fig"}). Importantly, this mutation was not previously observed in any public database. The residue 251Asn was revealed to be evolutionarily conserved ([Fig. 2B](#f2-ijmm-45-04-1047){ref-type="fig"}) and *PRKG2* expression was identified to be enriched in the small intestine (18.36 TPM), mainly in glandular cells ([Fig. 2C](#f2-ijmm-45-04-1047){ref-type="fig"}). In the F3, the proband F3:II:1 and his two affected sons shared the same missense mutation consisting of a G to C substitution (c.12G\>C) in *ADRB3* ([Fig. 3](#f3-ijmm-45-04-1047){ref-type="fig"}). *ADRB3* is part of the adrenergic system, which is involved in the regulation of lipid metabolism and glucose homeostasis ([@b37-ijmm-45-04-1047]). This single-nucleotide change resulted in a non-synonymous substitution (p.Trp4Cys). PolyPhen2 predicted that this variant was likely damaging ([Table II](#tII-ijmm-45-04-1047){ref-type="table"}). Sanger sequencing validated that six healthy family members presented homozygous wild-type alleles for *ADRB3* ([Fig. 3](#f3-ijmm-45-04-1047){ref-type="fig"}). Similarly to the aforementioned two residues, this position was predicted to be evolutionary conserved by the GERP++ tool (<https://omictools.com/gerp-tool>). Protein structural modeling --------------------------- Previous GWAS studies identified single nucleotide polymorphisms (SNPs) in these three genes that were associated with gout ([@b38-ijmm-45-04-1047]-[@b40-ijmm-45-04-1047]). All SNPs and three candidate mutations identified in the present study were mapped to schematic representations of *ABCG2*, *PRKG2* and *ADRB2* genes ([Fig. 4A](#f4-ijmm-45-04-1047){ref-type="fig"}). Except for the two mutations in*PRKG2*, all the other mutations were located in the protein functional domains. Furthermore, p.Lys157Glu mutation in *ABCG2* was located in the same highly conserved ABC transporter domain as the pathogenic missense variant Gln141Lys, which has been revealed to be associated with an increase in serum uric acid levels ([@b41-ijmm-45-04-1047]). The Asn251Ser mutation was in the cGMP-binding region of *PRKG2*; however, rs768867227 and rs10033237 SNPs, which were previously reported to cause gout, were present in non-coding regions ([@b42-ijmm-45-04-1047]). In order to further study the functional defects caused by mutations in the protein structure, the potential structural differences between the wild-type and mutant proteins were investigated. Therefore, numerous differences were identified between the wild-type and mutant *PRKG2* proteins. The wild-type structure (5BV6) was downloaded from the PDB. In the wide-type protein, there were seven hydrogen bonds at residue 251 (Asn); one bond was revealed between Asn251 and Glu191, and the other six connected Asn251 to water molecules ([Fig. 4B](#f4-ijmm-45-04-1047){ref-type="fig"}). The mutated structural modelling revealed the formation of two different hydrogen bonds, one between the mutated Ser251 and Thr250, and another between the mutated Ser251 and Tyr189 ([Fig. 4B](#f4-ijmm-45-04-1047){ref-type="fig"}). Functional analysis of the co-expression and PPI networks --------------------------------------------------------- Previous studies have identified that gout-associated genes are closely associated with urate excretion ([@b10-ijmm-45-04-1047]-[@b12-ijmm-45-04-1047]). To further investigate the expression pattern of the three candidate genes (*ABCG2*, *PRKG2* and *ADRB2*) identified in the present study and to examine whether their expression patterns were similar to other gout-associated genes identified in previous studies, the transcriptomic data in the small intestinal tissue from the GTEx project were analysed. Based on the co-expression network, these three genes demonstrated similar expression patterns with numerous gout-associated genes, with a Pearson correlation coefficient ranging between 0.602 and 0.898 ([Fig. 5A](#f5-ijmm-45-04-1047){ref-type="fig"}). Among these three genes, the number of candidate genes co-expressed with the *ABCG2* gene was the largest (n≤39). Furthermore, to identify functional similarities among these co-expressed genes that may have a function in the development of gout, significantly enriched DO terms were identified using the R package DOSE. Among the top ten most statistically significant DO terms with Bonferroni corrected P\<0.05 ([Fig. 5B](#f5-ijmm-45-04-1047){ref-type="fig"}; [Table SIV](#SD1-ijmm-45-04-1047){ref-type="supplementary-material"}), the majority of the terms were associated with cardiovascular and metabolic diseases, supporting the idea that gout is often accompanied by metabolic syndrome. The two directly associated terms, hyperuricemia \[false discovery rate (FDR)=8.02×10^−14^, hypergeometric test\] and gout (FDR=2.32×10^−12^, hypergeometric test) were the most statistically significant, indicating the significant enrichment of the co-expressed genes in hyperuricemia and gout. To investigate the association between the three candidate genes and other gout candidate genes, data from the STRING v10 database were analysed. The STRING database contains PPIs, including physical and functional associations. As a result, a total of 79 candidate genes were included in the PPI network ([Fig. 6A](#f6-ijmm-45-04-1047){ref-type="fig"}) with the interaction score ranging between 150 and 967. Solute carrier family 22 member 1 exhibited the strongest interaction with *ABCG2*, which is expressed in the kidney and mediates the transport of xenobiotics, endogenous organic anions and urate ([@b43-ijmm-45-04-1047]). *PRKG2* exhibited the strongest interaction with inositol 1,4,5-trisphosphate receptor type 1, which was previously reported to be associated with high serum uric acid concentrations ([@b44-ijmm-45-04-1047]). In the next step, all 79 candidate genes in the PPI network were used in the BP with ClueGO plugin. A total of 40 enriched BP terms were divided into eight groups ([Fig. 6B](#f6-ijmm-45-04-1047){ref-type="fig"}) and terms in the same group had similar biological functions. In total, two groups served direct functions in the metabolism of uric acid, and the terms in the two groups with the most significant statistical significance were \'purine nucleoside monophosphate biosynthetic process\' and \'urate transport\'. \'Glycometabolism\' and \'anion transport\' groups were also identified following data enrichment analysis. The detailed results of the BP analysis are presented in [Table SV](#SD1-ijmm-45-04-1047){ref-type="supplementary-material"}. Discussion ========== Gout is a complex disorder characterised by clinical and genetic heterogeneity, and the genetic mechanism underlying gout remains unclear, mainly because i) previous studies focused on sporadic cases; and ii) genotyping chips were unable to identify rare or novel variants. To address these two critical issues, the present study aimed to reveal candidate rare/novel mutations in large pedigrees with gout aggregation. A previous complex segregation and linkage analysis of familial gout revealed an autosomal-arbitrary major gene model ([@b20-ijmm-45-04-1047]). WES is able to provide insights into the pathogeny of hereditary diseases and extend molecular diagnosis ([@b45-ijmm-45-04-1047]-[@b48-ijmm-45-04-1047]). Therefore, WES was performed in three families with gout. Subsequent to employing previously established filtering strategies, three candidate variants were identified in these three families. In the proband F1:II:1, one novel missense mutation was revealed in *ABCG2* (c.469A\>G, p.Lys157Glu), and it was predicted to be deleterious by all four functional prediction tools.*ABCG2* has been reported to be an important factor involved in the reduction of urate transport rates ([@b40-ijmm-45-04-1047]) and in a pathway regulating fructose metabolism, which is associated with obesity ([@b13-ijmm-45-04-1047],[@b49-ijmm-45-04-1047]). The patient in F1 presenting the *ABCG2* mutation exhibited early onset-gout and was overweight, and these symptoms are in line with the pathological features of gout. The two most commonly reported dysfunctional SNPs are Gln126Ter (rs72552713) and Gln141LysK (rs2231142), which are located in the ABC domain, which is considered to be critical for the interactions between the intracellular loops of the transmembrane portion of the protein ([@b40-ijmm-45-04-1047]). The presence of the Q141K polymorphism in the ABC transporter domain was previously reported to induce a 2-fold decrease in urate efflux ([@b50-ijmm-45-04-1047]). The mutation p.Lys157Glu identified in the present study is located in the ABC transporter domain, indicating that the pathogenic mechanism of this mutation is caused by dysfunctions in this functional domain. Furthermore, a previous study has identified three common and 19 rare non-synonymous variants of *ABCG2* in patients with gout and functional assays were performed to determine the urate transport activity of each*ABCG2* variant ([@b51-ijmm-45-04-1047]). Almost all rare variants identified in the present study exhibited lower urate transport and almost completely inhibited *ABCG2* activity compared with the wild-type protein, as assessed by functional assays performed to determine the urate transport activity of each *ABCG2* variant. The second mutation identified was present in *ADRB3*, which encodes a β-3-adrenergic receptor and has been revealed to serve an important function in the regulation of lipolysis and glucose homeostasis ([@b37-ijmm-45-04-1047],[@b39-ijmm-45-04-1047]). Hyperuricemia and gout are closely associated with metabolic disorders, including obesity, dyslipidaemia, glucose intolerance and hypertension ([@b52-ijmm-45-04-1047],[@b53-ijmm-45-04-1047]). Wang *et al* ([@b20-ijmm-45-04-1047]) and Huang *et al* ([@b54-ijmm-45-04-1047]) reported that a p.Trp64Arg variant was associated with hyperuricemia in Chinese male patients. A similar study investigating Spanish patients identified that a p.Trp64Arg polymorphism in *ADRB3* gene may increase the risk of hyperuricemia ([@b55-ijmm-45-04-1047]). Therefore, the missense mutation c.12G\>C in the *ADRB3* gene revealed in the present study was spatially near the same intracellular loop affected in the p.Trp64Arg mutation, and it is thus expected to be involved in the development of basal metabolic diseases including dyslipidaemia, which was confirmed by the fact that patients with gout in Family 3 had hyperlipemia and hyperbilirubinemia. The third mutation identified in the present study was located in the *PRKG2* gene. Accumulating evidence has demonstrated that the cGMP signalling pathway serves important functions in urate cycles, and *PRKG2* is a cGKII gene ([@b42-ijmm-45-04-1047]). One case-control study revealed that there was a correlation between the polymorphisms rs768867227 and rs10033237 in *PRKG2*, and gout susceptibility ([@b56-ijmm-45-04-1047]). Each of these polymorphisms are located in non-coding regions. Although the biological function of combined rs10033237 and rs7688672 in the *PRKG2* gene has not been elucidated, dysfunctions in *PRKG2* may result in hypertension by destroying the renin-angiotensin-aldosterone system, and may result in hyperuricemia ([@b42-ijmm-45-04-1047]). The renin-angiotensin-aldosterone system in the patients in Family 3 carrying heterozygous missense mutations (c.752A\>G and Asn251Ser) in *PRKG2* was potentially impaired. An impaired renin-angiotensin-aldosterone system may cause hyperuricemia, which was consistent with the presence of hypertension in the patients in Family 3. Furthermore, a previous study ([@b57-ijmm-45-04-1047]) demonstrated that *PRKG2* serves a key function in mediating M1 polarization and phagocytotic activity by regulating the levels of monosodium urate and lipopolysaccharides. Excretion of uric acid requires specialized transporters located in renal tubule cells and intestinal epithelial cells ([@b58-ijmm-45-04-1047]). Accordingly, direct intestinal secretion is considered as a substantial contributor to the extra-renal elimination of uric acid ([@b59-ijmm-45-04-1047]). It is estimated that \~30% of uric acid is excreted by the intestine ([@b60-ijmm-45-04-1047]). Numerous gout-associated genes are highly expressed in the small intestine, including *ABCG2* and *PRKG2* detected in the present study ([@b10-ijmm-45-04-1047]-[@b12-ijmm-45-04-1047]). The co-expression network revealed that gout-associated genes share similar expression patterns with *ABCG2*, *PRKG2* and *ADRB3* in the small intestine. *ABCG2* is a well-characterized urate transporter in the intestine, and the molecular function of *ABCG2* was previously identified ([@b10-ijmm-45-04-1047]-[@b12-ijmm-45-04-1047]). The functional interactions between proteins are crucial for their biological function, and their systematic characterization may increase current understanding of molecular systems biology ([@b61-ijmm-45-04-1047]). The STING network constructed using *ABCG2*, *PRKG2* and *ADRB3* contained a number of gout-associated genes, suggesting that the pathogenic mechanism of these genes may share the same pathway. Gout is a metabolic disorder caused by urate overproduction and/or reduced urate excretion ([@b62-ijmm-45-04-1047]). According to previous studies, gout is associated with five major metabolic syndromes and/or consequences of metabolic syndrome: Hypertension, cardiovascular disease, insulin resistance and diabetes, obesity and hyperlipidaemia ([@b63-ijmm-45-04-1047]-[@b65-ijmm-45-04-1047]). In the present study, a DO analysis was performed on the genes identified in the co-expression network, and three DO terms associated with the cardiovascular system, including cardiovascular system disease, hypertension and artery disease, were in the top 10 significantly enriched terms. By directly impairing the vascular endothelium and the renal system, high serum urate levels may increase blood pressure ([@b66-ijmm-45-04-1047]). Other enriched terms were mainly associated with obesity. Following a data analysis performed on data from 517 participants in the Bogalusa Heart Study, Muntner *et al* ([@b67-ijmm-45-04-1047]) suggested that an elevated body mass index was associated with high levels of uric acid. Furthermore, the BP analysis of the genes in the PPI network contained terms including \'hexose transmembrane transport\' and \'regulation of insulin secretion\', and the two groups involved in glycose metabolism have been reported to affect the transport of uric acid ([@b68-ijmm-45-04-1047]-[@b70-ijmm-45-04-1047]). The majority of the enriched BP terms were revealed to be involved in metabolite transports, which are closely associated with urate transport. Despite recent advancements in the understanding of the mechanism underlying gout, the number of families involved in the present study was small; therefore, the statistical power for the autosomal dominant inheritance model in gout was limited, although the model was previously proposed according to statistical analyses based on large sample sizes ([@b19-ijmm-45-04-1047],[@b20-ijmm-45-04-1047]). In addition, no functional genomics studies were performed in the present study. Therefore, the sample size of the study should be expanded to provide stronger evidence for the Mendelian genetic inheritance of gout. Copy number variations were not analysed in the present study ([@b71-ijmm-45-04-1047],[@b72-ijmm-45-04-1047]). Furthermore, functional experiments are necessary to further determine the function of the mutations detected in gout pathogenesis. Collectively, to the best of our knowledge, the present study is the first to use WES to dissect rare or novel genetic variants using pedigree and family aggregation analyses. Among the three families, three deleterious mutations in three different gout pathogenic genes were identified. The present results increase the current knowledge of the genotypic heterogeneity underlying the phenotypic heterogeneity of gout, which assists not only clinical diagnoses but also potential personalized therapy. Supplementary Data ================== Not applicable. Funding ======= The present study was supported by the Zhejiang Provincial Natural Science Foundation (grant no. LY20H100001) and the Wenzhou Science and Technology Bureau (grant no. Y20180129). Availability of data and materials ================================== All the datasets generated and/or analysed in the present study are available from the corresponding author upon reasonable request. Authors\' contributions ======================= XX and JJ conceived and designed the study. ZC and ZZ performed the experiments. HC and CZ collected and analysed the experimental data. LZ wrote the article. LS helped revise the article. All authors read and approved the final manuscript. Ethics approval and consent to participate ========================================== This study was conducted according to the principles outlined in the Declaration of Helsinki. The study protocol was approved by the Ethics Committee of The First Affiliated Hospital, Wenzhou Medical University (Wenzhou, China; approval no. 2018-020). Written informed consent was obtained from all participants or their guardians. Patient consent for publication =============================== The patients provided written informed consent regarding the publication of the case details and any associated images. Competing interests =================== The authors declare that they have no competing interests. {#f1-ijmm-45-04-1047} {#f2-ijmm-45-04-1047} {#f3-ijmm-45-04-1047} {#f4-ijmm-45-04-1047} {#f5-ijmm-45-04-1047} {#f6-ijmm-45-04-1047} ###### Clinical data of patients with gout in three large pedigrees. Individual ID Sex Age (years) Age at onset (years) Uric acid (mg/l) Hyperuricemia (+/−) Arthritis Tophi Comorbidities --------------- ----- ------------- ---------------------- ------------------ --------------------- ----------- ------- --------------- ---- ---- ---- ---- F1_I:1 M 48 30 579 \+ \+ \+ \+ \+ \+ \+ − F1_II:1 M 20 15 801 \+ \+ \+ \+ \+ − \+ − F2_II:1 M 67 55 617 \+ \+ \+ \+ − − \+ \+ F2_II:7 M 51 38 580 \+ \+ \+ \+ − − \+ \+ F2_II:11 F 68 − 364 − − − − − − \+ − F2_III:1 M 40 35 499 \+ \+ \+ \+ − − − \+ F2_III:9 M 34 28 610 \+ \+ − − − − \+ \+ F3_II:2 F 78 70 647 \+ \+ \+ \+ \+ − − \+ F3_III:2 F 52 52 474 \+ \+ − \+ \+ − \+ \+ F3_III:3 F 62 − 463 \+ − − \+ \+ − \+ \+ BMI, body mass index; F, female; M, male. ###### Summary of mutations detected by whole-exome sequencing of patients with gout. Individual ID Chromosome Gene Position (hg19) Type Protein change ExAC Polyphen2 LRT MutationTaster FATHMM --------------- ------------ --------- ----------------- ---------- ---------------- ------------- ----------------- ------------- ----------------- ----------- F1:II:1 Chr4 *ABCG2* 89052275 Missense p.K157E 8.24×10^−6^ Likely damaging Deleterious Disease causing Damaging F2:II:1 Chr4 *PRKG2* 82090913 Missense p.N251S \- Benign Deleterious Disease causing Damaging F3:II:1 Chr8 *ADRB3* 37823976 Missense p.W4C 6.49×10^−5^ Likely damaging Neutral Disease causing Tolerable ABCG2, ATP-binding cassette super-family G member 2; PRKG2, protein kinase CGMP-dependent 2; ADRB3, adrenoceptor β3; Polyphen2, Polymorphism Phenotyping v2; LRT, Likelihood Ratio Test; FATHMM, Functional Analysis through Hidden Markov Models; ExAC, Exome Aggregation Consortium. [^1]: Contributed equally

Sepsis is a major cause of morbidity and mortality in modern intensive care units (ICUs). Although several studies have provided epidemiological data on sepsis in ICU patients in the developed world \[[@CIT0001]\], there is limited information on the global burden of sepsis worldwide \[[@CIT0007], [@CIT0008]\]. Yet, such data are crucially important to (1) increase awareness of the global impact of sepsis, (2) highlight the need for continued research into potential preventive and therapeutic interventions, and (3) help guide resource allocation \[[@CIT0009]\]. Information on patterns of sepsis around the globe is also of interest, including causative microorganisms, primary source of infection, and associated outcomes. In 2012, the World Federation of Societies of Intensive and Critical Care Medicine conducted a worldwide audit of data from ICUs around the world, providing a large database from which to extract information. We used these data to explore the characteristics of patients with sepsis around the world, including international differences in occurrence rates, causative microorganisms, and outcomes. We also evaluated some factors associated with in-hospital mortality in these patients. METHODS {#s5} ======= The worldwide Intensive Care over Nations (ICON) audit recruited ICUs by open invitation, through national scientific societies, national and international meetings, e-mail lists, and individual contacts. Participation was entirely voluntary, with no financial incentive. Ethics committee approval was obtained by the participating institutions according to local ethical regulations. Informed consent was not required for this observational and anonymous audit. Of the 730 ICUs contributing to the study from 84 countries (see the participants list in Appendix 1), 419 (57.4%) were in university/academic hospitals. The organizational characteristics of these centers have been described previously \[[@CIT0010]\]. Each ICU was asked to prospectively collect data on all adult (\>16 years) patients admitted to their ICU between May 8 and May 18, 2012, except those who stayed in the ICU for \<24 hours for routine postoperative surveillance. Readmissions of previously included patients were not included. Data were collected daily for a maximum of 28 days in the ICU. Outcome data were collected at the time of ICU and hospital discharge or at 60 days. Data were entered anonymously using electronic case report forms via a secured internet-based website. Data collection on admission included demographic data and comorbidities. Clinical and laboratory data for simplified acute physiology (SAPS) II \[[@CIT0011]\] and Acute Physiologic Assessment and Chronic Health Evaluation (APACHE) II \[[@CIT0012]\] scores were reported as the worst values within the first 24 hours after admission. A daily evaluation of organ function was performed according to the sequential organ failure assessment (SOFA) score \[[@CIT0013]\]; organ failure was defined as a SOFA subscore \>2 for the organ in question. Clinical and microbiologic infections were reported daily as well as antimicrobial therapy. Infection was defined according to the criteria of the International Sepsis Forum \[[@CIT0014]\]. Sepsis was defined as the presence of infection with associated organ failure \[[@CIT0015]\]. Septic shock was defined as sepsis associated with cardiovascular failure requiring vasopressor support (SOFA cardiovascular of 3 or 4). Intensive care unit-acquired infection was defined as infection identified at least 48 hours after ICU admission. Non-ICU acquired infection was defined as infection present on admission or within the first 48 hours after ICU admission. Only the first episode of infection was considered in the analysis. Detailed instructions and definitions were available through a secured website for all participants before starting data collection and throughout the study period. Any additional queries were answered on a per case basis. Validity checks were made at the time of electronic data entry, including plausibility checks within each variable and between variables. Data were further reviewed by the coordinating center for completeness and plausibility, and any doubts were clarified with the participating center. There was no on-site monitoring. We did not attempt to verify the pathogenicity of the microorganisms, including the relevance of *Staphylococcus epidermidis* or the distinction between colonization and infection. For the purposes of this audit, we divided the world into 8 geographic regions: North America, South America, Western Europe, Eastern Europe, South Asia, East and Southeast Asia, Oceania, and Africa. Individual countries were also classified into 3 income groups according to the 2011 gross national income (GNI) per capita, calculated using the World Bank Atlas method \[[@CIT0016]\]: GNI \<\$4035 = low and lower middle income; GNI \$4036--12475 = upper middle income; and GNI \>\$12476 = high income. Statistical Analysis {#s6} -------------------- Data are shown as means with standard deviation (SD), mean and 95% confidence intervals (CIs), medians and interquartile ranges (IQRs), numbers, and percentages. Differences between groups in distribution of variables were assessed using analysis of variance, Kruskal-Wallis test, Student's *t* test, Mann-Whitney test, χ^2^ test, or Fisher's exact test as appropriate. To identify the risk factors associated with in-hospital mortality in septic patients, we used a 3-level multilevel technique with the structure of an individual patient (level 1) admitted to a hospital (level 2) within a country (level 3). The explanatory variables considered in the model were as follows: - Individual-level factors: age, sex, SAPS II score, type of admission, source of admission, mechanical ventilation or renal replacement therapy at any time during the ICU stay, comorbidities, onset of infection, site of infection, and the most common microorganisms - Hospital-level factors: type of hospital; ICU specialty; total number of ICU patients in 2011; number of staffed ICU beds - Country-level factors: GNI Individual-level variables to be included in the final model were selected on the basis of a multilevel model including country-level and hospital-level factors and each of the individual-level factors; variables with *P* \< .2 were considered in the final model. Collinearity between variables was checked by inspection of the correlation between them, by looking at the correlation matrix of the estimated parameters, and by looking at the change of parameter estimates and at their estimated standard errors (SEs) \[[@CIT0017]\]. Q-Q plots were drawn to check for normality in the residuals. The results of fixed effects (measures of association) are given as odds ratios (ORs) with their 95% CIs and the 80% interval OR. Random effects (measures of variation) measures included the variance (var) and its SE and the median OR. The statistical significance of covariates was calculated using the Wald test. Data were analyzed using IBM SPSS Statistics software, version 22 for Windows and R software, version 2.0.1 (CRAN project). All reported *P* values are 2-sided, and *P* \< .05 was considered to indicate statistical significance. The results of fixed effects are given as OR with 95% CIs. RESULTS {#s7} ======= Characteristics of the Study Group {#s8} ---------------------------------- A total of 10069 patients were included in the audit; 2973 patients (29.5%) had sepsis, including 1808 (18.0%) with sepsis at admission to the ICU ([Figure 1](#F1){ref-type="fig"}). In the whole cohort, antimicrobials were given to 5975 (59.3%) patients during their ICU stay. Patients with sepsis were older, had higher severity scores on admission to the ICU, had more comorbidities, and were more commonly receiving mechanical ventilation and renal replacement therapy on admission to the ICU than patients without sepsis ([Table 1](#T1){ref-type="table"}). Patients with sepsis also had more organ failures than the other patients (median \[IQ\]: 3 \[1--4\] vs 1 \[0--2\] organs; *P* \< .001). {#F1} ###### Characteristics of the Study Cohort on Admission to the ICU According to the Presence of Sepsis^a^ ----------------------------------------------------------------------------------------------- Characteristics All Patients\ No Sepsis\ Sepsis\ *P* Value N = 10069 N = 7096 N = 2973 --------------------------------------- --------------- ------------- ------------- ----------- Age, years, mean ± SD 60.0 ± 18.0 59.4 ± 18.4 61.5 ± 17.0 \<.001 Male, n (%) 5973 (60.1) 4177 (59.7) 1796 (61.0) .21 Severity scores, mean ± SD  SAPS II score 40.2 ± 18.2 36.4 ± 17.4 49.2 ± 16.6 \<.001  SOFA score 5 \[3--9\] 4 \[2--7\] 8 \[6--11\] \<.001 Type of admission, n (%) \<.001  Surgical 3432 (36.0) 2475 (37.0) 957 (33.7)  Medical 5382 (56.5) 3646 (54.6) 1736 (61.1)  Trauma 643 (6.8) 512 (7.7) 131 (4.6)  Other 66 (0.7) 49 (.7) 17 (.6) Source of admission, n (%) \<.001  ER/ambulance 3814 (37.9) 2780 (39.2) 1034 (34.8)  Hospital floor 2625 (26.1) 1664 (23.4) 961 (32.3)  OR/recovery 1811 (18.0) 1363 (19.2) 448 (15.1)  Other hospital 981 (9.7) 652 (9.2) 329 (11.1)  Other 838 (8.3) 637 (9.0) 201 (6.8) Comorbidities, n (%)  COPD 1240 (12.3) 788 (11.1) 452 (15.2) \<.001  Cancer 1049 (10.4) 710 (10.0) 339 (11.4) .04  Diabetes mellitus, insulin-dependent 972 (9.7) 664 (9.4) 308 (10.4) .12  Heart failure, NYHA III/IV 921 (9.1) 588 (8.3) 333 (11.2) \<.001  Chronic renal failure 912 (9.1) 582 (8.2) 330 (11.1) \<.001  Cirrhosis 349 (3.5) 217 (3.1) 132 (4.4) \<.001  Immunosuppression 346 (3.4) 177 (2.5) 169 (5.7) \<.001  Metastatic cancer 332 (3.3) 221 (3.1) 111 (3.7) .11  Haematologic cancer 212 (2.1) 99 (1.4) 113 (3.8) \<.001  HIV infection 71 (.7) 37 (.5) 34 (1.1) \<.001 Number of comorbidities, n (%) \<.001  None 5512 (54.7) 4145 (58.4) 1367 (46.0)  1 2800 (27.8) 1880 (26.5) 920 (30.9)  2 1207 (12.0) 740 (10.4) 467 (15.7)  3 416 (4.1) 249 (3.5) 167 (5.6)  ≥4 134 (1.3) 82 (1.2) 52 (1.7) Procedures, n (%)  Mechanical ventilation 4776 (47.4) 2755 (38.8) 2021 (68.0) \<.001  Renal replacement therapy 537 (5.3) 264 (3.7) 273 (9.2) \<.001 Antimicrobials, n (%) 5975 (59.3) 3002 (42.3) 2973 (100) \<.001  Antibiotic 5935 (58.9) 2979 (42.0) 2956 (99.4) \<.001  Antifungal 784 (7.8) 202 (2.8) 582 (19.6) \<.001  Antiviral 273 (2.7) 80 (1.1) 193 (6.5) \<.001 ----------------------------------------------------------------------------------------------- Abbreviations: COPD, chronic obstructive pulmonary disease; ER, emergency room; HIV, human immunodeficiency virus; ICU, intensive care unit; NYHA, New York Heart Association Classification; OR, operating room; SAPS, simplified acute physiology; SOFA, sequential organ assessment; SD, standard deviation. ^a^Valid percentages are given after exclusion of missing values (data missing from 546 patients for type of admission). Patterns of Infections {#s9} ---------------------- The most common source of sepsis was the respiratory tract (67.4%) followed by the abdomen (21.8%) ([Supplementary Table E1](#sup1){ref-type="supplementary-material"}). Positive isolates were retrieved in 69.6% (n = 2069) of patients with sepsis; two thirds of these patients had Gram-negative microorganisms isolated and half had Gram-positive microorganisms; 1068 (51.6%) of the sepsis patients with positive isolates had more than 1 microorganism isolated ([Table 2](#T2){ref-type="table"}). Patients with urinary tract (82.6% vs 43.9%), abdominal (77.1% vs 50.8%), and respiratory tract (70.0% vs 51.4%) infections were more likely to have Gram-negative than Gram-positive isolates ([Supplementary Table E1](#sup1){ref-type="supplementary-material"}). Microbiological patterns varied around the globe ([Table 2](#T2){ref-type="table"}), with Gram-positive isolates being much less frequent (21.4%) in South Asia than in other regions. Methicillin-resistant *Staphylococcus aureus* (MRSA) was more common in the Middle East (14.4%) and North America (12.8%) than in Western Europe (6.1%). *Klebsiella* spp isolates were most commonly reported in Africa (31.3%), Eastern Europe (28.5%), and South America (24.7%), and *Pseudomonas* spp was most frequent in Eastern Europe (21.1%) and South America (20.4%). Fungal organisms contributed to 14.5% and 14.8% of isolates in Western and Eastern Europe, respectively, but to only 5.1% of isolates in North America. ###### Distribution of the Most Common Microorganisms in Patients With Positive Isolates and Mortality Rates According to Geographic Region ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Characteristic All Patients Western\ Eastern\ South\ North\ East and\ South\ Oceania Middle\ Africa Europe Europe America America Southeast Asia Asia East ------------------------------------------------- -------------- ------------- ------------ ------------ ------------ ---------------- ------------ ------------ ------------ ----------- Total number of patients, n 10 069 4335 1110 993 730 946 982 439 393 141 Patients with sepsis, n (%) 2973 (29.5) 1357 (31.3) 336 (30.3) 303 (30.5) 147 (20.1) 372 (39.3) 134 (13.6) 135 (30.8) 151 (38.4) 38 (27.0)  Patients with positive isolates, n (%)^a^ 2069 (69.6) 947 (69.8) 256 (76.2) 186 (61.4) 117 (79.6) 240 (64.5) 84 (62.7) 105 (77.8) 118 (78.1) 16 (42.1)  Gram-positive, n (%) 1030 (49.8) 517 (54.6) 144 (56.3) 84 (45.2) 59 (50.4) 86 (35.8) 18 (21.4) 57 (54.3) 58 (49.2) 7 (43.8)   Methicillin-sensitive *Staphylococcus aureus* 257 (12.4) 120 (12.7) 37 (14.5) 29 (15.6) 14 (12.0) 25 (10.4) 6 (7.1) 14 (13.3) 10 (8.5) 2 (12.5)   Methicillin-resistant *S aureus* 151 (7.3) 58 (6.1) 20 (7.8) 15 (8.1) 15 (12.8) 14 (5.8) 2 (2.4) 9 (8.6) 17 (14.4) 1 (6.3)   Coagulase-negative *Staphylococcus* 500 (24.2) 269 (28.4) 79 (30.9) 39 (21.0) 25 (21.4) 25 (10.4) 11 (13.1) 19 (18.1) 27 (22.9) 6 (37.5)   *Streptococcus*, D group 84 (4.1) 52 (5.5) 10 (3.9) 5 (2.7) 5 (4.3) 3 (1.3) \- 6 (5.7) 3 (2.5) \-   *Streptococcus*, Others 222 (10.7) 109 (11.5) 25 (9.8) 12 (6.5) 15 (12.8) 27 (11.3) 1 (1.2) 15 (14.3) 16 (13.6) 2 (12.5)   Other Gram-positive cocci 46 (2.2) 19 (2.0) 11 (4.3) \- 3 (2.6) 7 (2.9) \- 4 (3.8) 2 (1.7) \-  Gram negative, n (%) 1389 (67.1) 610 (64.4) 189 (73.8) 140 (75.3) 65 (55.6) 159 (66.3) 67 (79.8) 66 (62.9) 84 (71.2) 9 (56.3)   *Escherichia coli* 470 (22.7) 236 (24.9) 57 (22.3) 51 (27.4) 22 (18.8) 35 (14.6) 22 (26.2) 25 (23.8) 19 (16.1) 3 (18.8)   *Klebsiella* spp 356 (17.2) 124 (13.1) 73 (28.5) 46 (24.7) 13 (11.1) 45 (18.8) 18 (21.4) 13 (12.4) 19 (16.1) 5 (31.3)   *Pseudomonas* spp 337 (16.3) 147 (15.5) 54 (21.1) 38 (20.4) 18 (15.4) 35 (14.6) 13 (15.5) 12 (11.4) 20 (16.9) \-   *Acinetobacter* spp 243 (11.7) 39 (4.1) 55 (21.5) 36 (19.4) 5 (4.3) 51 (21.3) 24 (28.6) 1 (1.0) 29 (24.6) 3 (18.8)   *Enterobacter* spp 188 (9.1) 91 (9.6) 45 (17.6) 13 (7.0) 3 (2.6) 11 (4.6) 10 (11.9) 8 (7.6) 5 (4.2) 2 (12.5)   *Proteus* spp 121 (5.8) 63 (6.7) 25 (9.8) 6 (3.2) 5 (4.3) 7 (2.9) 1 (1.2) 5 (4.8) 6 (5.1) 3 (18.8)   Gram-negative, others 320 (15.5) 174 (18.4) 35 (13.7) 20 (10.8) 14 (12.0) 32 (13.3) 4 (4.8) 17 (16.2) 22 (18.6) 2 (12.5)  Anaerobes, n (%) 79 (3.8) 45 (4.8) 12 (4.7) 3 (1.6) 8 (6.8) 4 (1.7) \- 4 (3.8) 2 (1.7) 1 (6.3)  Other bacteria, n (%) 18 (0.9) 4 (0.4) 2 (0.8) 2 (1.1) 1 (0.9) 5 (2.1) 1 (1.2) 2 (1.9) 1 (0.8) \-  Fungi, n (%) 266 (12.9) 137 (14.5) 38 (14.8) 18 (9.7) 6 (5.1) 31 (12.9) 8 (9.5) 10 (9.5) 16 (13.6) 2 (12.5)  *Candida albicans* 185 (8.9) 93 (9.8) 31 (12.1) 9 (4.8) 3 (2.6) 23 (9.6) 6 (7.1) 9 (8.6) 10 (8.5) 1 (6.3)  *Candida* non-*albicans* 89 (4.3) 47 (5.0) 8 (3.1) 8 (4.3) 2 (1.7) 11 (4.6) 4 (4.8) 4 (3.8) 4 (3.4) 1 (6.3)  Fungi, others 44 (2.1) 28 (3.0) 4 (1.6) 2 (1.1) 1 (0.9) 5 (2.1) \- \- 4 (3.4) \-  Viruses and parasites, n (%) 59 (2.9) 31 (3.3) 5 (2.0) 6 (3.2) 2 (1.7) 10 (4.2) \- 1 (1.0) 3 (2.5) 1 (6.3) Mortality rates in patients with sepsis, n (%)  Intensive care unit 753 (25.8) 309 (22.9) 118 (35.3) 102 (36.3) 27 (18.5) 76 (21.2) 37 (28.9) 16 (11.9) 53 (35.6) 15 (39.5)  Hospital 1004 (35.3) 439 (33.3) 146 (44.8) 119 (45.4) 37 (25.2) 108 (31) 44 (35.2) 26 (19.3) 68 (46.6) 17 (47.2) ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ^a^In patients with sepsis. Patients with ICU-acquired infections (n = 764) were younger, more likely to be surgical admissions, and had lower SAPS II and SOFA scores on admission to the ICU, compared with those who had infections within the first 48 hours on the ICU ([Table 3](#T3){ref-type="table"} and [Supplementary Table E2](#sup1){ref-type="supplementary-material"}). Respiratory and catheter-associated infections were more frequent and abdominal infections less frequent in patients with ICU-acquired than in those with non-ICU-acquired infections ([Supplementary Table E2](#sup1){ref-type="supplementary-material"}). Patients with ICU-acquired infections were more likely to have positive isolates than patients with non-ICU-acquired infections (79.5% vs 66.2%, *P* \< .001) ([Supplementary Table E3](#sup1){ref-type="supplementary-material"}). ###### Severity Scores on Admission to the ICU, Maximum Number of Organ Failure, and Mortality Rates According to Sepsis Status ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Severity Scores, Mean ± SD\ Mortality Rates, % (95% CI) ------------------------------------- ------ ----------------------------- --------------- ----------------- ----------------------------- ---------------------- ---------------------- Onset of Sepsis  Within the first 48 hours^d^ 2209 50.5 ± 16.8 9.2 ± 3.9 5 \[2--10\] 3 \[2--4\] 26.0 (24.2--27.9) 35.8 (33.8--37.9)  Later 764 45.4 ± 15.5^e^ 7.5 ± 3.8^e^ 12 \[6--19\]^e^ 3 \[2--4\] 25.1 (22.0--28.3) 33.8 (30.3--37.2) Severity of Sepsis on ICU Admission  No sepsis^d^ 8261 37.8 ± 17.5 5.3 ± 4.1 3 \[1--5\] 1 \[0--2\] 13.6 (12.9--14.4) 19.0 (18.1--19.9)  Sepsis without shock 822 46.2 ± 15.4^e^ 7.4 ± 2.9^e^ 5 \[2--9\]^e^ 2 \[1--3\]^e^ 20.1 (17.4--22.9)^e^ 30.3 (27.1--33.6)^e^  Septic shock 986 55.1 ± 17.2^e^ 11.3 ± 3.6^e^ 5 \[2--11\]^e^ 3 \[2--4\]^e^ 33.7 (30.7--36.7)^e^ 43.0 (39.9--46.2)^e^ Severity of Sepsis During ICU Stay  No sepsis^d^ 7096 36.4 ± 17.4 4.9 ± 4.0 2 \[1--4\] 1 \[0--2\] 12.1 (11.3--12.8) 16.7 (15.8--17.6)  Sepsis without shock 1292 44.6 ± 15.3^e^ 7.0 ± 3.2^e^ 6 \[3--11\]^e^ 2 \[1--3\]^e^ 14.3 (12.4--16.2)^f^ 23.6 (21.3--26.0)^e^  Septic shock 1681 52.7 ± 16.7^e^ 10.1 ± 3.9^e^ 7 \[3--14\]^e^ 3 \[2--4\]^e^ 34.6 (32.3--36.9)^e^ 44.2 (41.7--46.6)^e^ ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Abbreviations: CI, confidence interval; ICU, intensive care unit; IQR, interquartile; LOS, length of stay; SAPS, simplified acute physiology score; SD, standard deviation; SOFA, sequential organ failure assessment score. Missing observations: ^a^489, ^b^401, ^c^797. ^d^The reference group. ^e^ *P* \< .01 among groups. ^f^ *P* \< .05 among groups. Outcomes {#s10} -------- Intensive care unit mortality rates were 25.8% in patients with sepsis and 12.1% in those without (*P* \< .001); hospital mortality rates were 35.3% vs 16.7%, *P* \< .001). Intensive care unit and hospital mortality rates varied from 11.9% and 19.3% (Oceania) to 39.5% and 47.2% (Africa), respectively ([Table 2](#T2){ref-type="table"}). Intensive care unit length of stay was longer (6 \[3--13\] vs 2 \[1--4\] days, *P* \< .001) in patients with than in those without sepsis. As expected, there was a stepwise increase in ICU and hospital mortality rates according to the severity of sepsis ([Table 3](#T3){ref-type="table"}). Although patients with ICU-acquired sepsis had longer ICU stays than those who had sepsis within 48 hours of admission to the ICU, they did not have higher mortality rates ([Table 3](#T3){ref-type="table"}). The crude risk of in-hospital death was higher in patients with infections caused by *Pseudomonas* spp, *Acinetobacter* spp, and fungi ([Table 4](#T4){ref-type="table"}). In the multilevel analysis, independent risk factors for in-hospital death in patients with sepsis were older age, higher SAPS II score, cirrhosis, metastatic cancer, chronic heart failure (NYHA III/IV), use of mechanical ventilation or renal replacement therapy at any time during the ICU stay, and infection with *Acinetobacter* spp ([Supplementary Table E4](#sup1){ref-type="supplementary-material"}). The use of mechanical ventilation and presence of comorbid cirrhosis more than doubled the risk of death. The relative risk of death was higher in patients admitted to ICUs in countries with upper middle GNI than in those with high GNI (1.77 \[1.31--2.39\], *P* \< .001). However, although the model suggested significant between-hospital variation (var = 0.28, *P* = .001) in the individual risk of in-hospital death, the between-country variation was not significant. ###### Outcome According to Isolated Microorganisms in Patients With Sepsis (n = 2973) ---------------------------------------------------------------------------------------------------------------------------------------------- Risk Factor ICU Mortality,\ Hospital Mortality, n (%) Crude Risk of In- Hospital Death\ *P* Value n (%) OR (95% CI) ------------------------------------------------ ----------------- --------------------------- ----------------------------------- ----------- Gram-positive 267 (26.2) 360 (36.0) 1.05 (0.89--1.23) .55  *Staphylococcus aureus* methicillin sensitive 71 (28.0) 89 (36.0) 1.04 (0.79--1.36) .80  MRSA 36 (24.2) 51 (34.7) 0.97 (0.69--1.38) .87  *Staphylococcus*, coagulase negative 129 (26.0) 184 (37.9) 1.14 (0.93--1.40) .20  *Streptococcus*, D group 16 (19.3) 22 (27.5) 0.69 (0.42--1.13) .14  *Streptococcus*, others 57 (26.1) 77 (35.8) 1.02 (0.77--1.37) .87  Other Gram-positive cocci 9 (19.6) 13 (29.5) 0.77 (0.40--1.47) .42 Gram negative 364 (26.6) 492 (37.0) 1.15 (0.99--1.34) .07  *Escherichia coli* 114 (24.7) 162 (36.0) 1.04 (0.84--1.28) .74  *Enterobacter* spp 45 (24.1) 67 (36.8) 1.07 (0.79--1.46) .66  *Klebsiella* spp 92 (26.4) 128 (37.9) 1.13 (0.90--1.43) .29  *Acinetobacter* spp 88 (37.0) 110 (48.0) 1.78 (1.36--2.33) \<.01  *Proteus* spp 28 (23.1) 40 (33.6) 0.92 (0.63--1.36) .69  *Pseudomonas* spp 100 (30.1) 131 (40.4) 1.28 (1.01--1.62) .04  Gram negative, others 82 (25.9) 110 (35.7) 1.02 (0.80--1.31) .87 Anaerobes 23 (29.1) 31 (39.7) 1.22 (0.77--1.93) .41 Other bacteria 6 (33.3) 7 (38.9) 1.17 (0.45--3.02) .75 Fungi 77 (29.2) 107 (41.6) 1.34 (1.04--1.75) .03  *Candida albicans* 49 (26.8) 71 (39.9) 1.23 (0.90--1.68) .19  *Candida* non-*albicans* 26 (29.2) 38 (43.7) 1.44 (0.93--2.21) .10  Fungi, others 16 (36.4) 20 (45.5) 1.54 (0.85--2.80) .16 Viruses and parasites 16 (28.1) 21 (36.8) 1.07 (0.62--1.84) .81 ---------------------------------------------------------------------------------------------------------------------------------------------- Abbreviations: CI, confidence interval; ICU, intensive care unit; MRSA, methicillin-resistant *Staphylococcus aureus*; OR, odds ratio. DISCUSSION {#s11} ========== The present audit confirms the considerable burden that sepsis presents in modern ICUs. This large study, including more than 10000 patients from 730 ICUs, indicates that approximately 30% of all ICU patients have sepsis, as defined by the presence of infection and organ dysfunction. This percentage is identical to that (29.5%) reported in the earlier Sepsis Occurrence in Acutely Ill Patients (SOAP) study \[[@CIT0001]\], a large European study that used the same methodology, and in a recent analysis of a large United Kingdom database \[[@CIT0018]\], but somewhat higher than in some other large studies \[[@CIT0004], [@CIT0005], [@CIT0019], [@CIT0020]\]. In addition to possible differences associated with different definitions of sepsis used in the various studies, 2 other major elements may account for these apparent inconsistencies. First, we did not include all patients admitted to the ICU, but only critically ill patients, excluding patients admitted to the ICU for postoperative surveillance without complications. Second, some studies focused on admission data \[[@CIT0020]\]; if we consider only the patients who had sepsis on admission in our study, the rate of sepsis was 18%. More importantly, the percentage of ICU patients with sepsis varied around the globe, with particularly high rates in East and Southeast Asia, confirming the high disease burden in this area \[[@CIT0021], [@CIT0022]\]. Although these data were collected in 2012, we believe they are still relevant, especially given the general lack of global data in this regard. A strength of the present study compared with studies assessing only sepsis on admission or prevalence studies (eg, EPIC II \[[@CIT0002]\]) is that patients were followed throughout the ICU course, enabling evaluation of sepsis that developed during the ICU stay as well as sepsis present on admission. It is interesting to note that patients with ICU-acquired sepsis had similar outcomes to those of patients with sepsis on admission, and ICU-acquired sepsis was not independently associated with a higher risk of mortality after adjusting for confounders in the multilevel analysis. Although we were unable to assess this specifically, van Vught et al \[[@CIT0023]\] recently reported a low attributable mortality of ICU-acquired infections. Shankar-Hari et al \[[@CIT0024]\] reported that the inferred causal link between sepsis and long-term mortality was significantly confounded by age, comorbidity, and preacute illness trajectory. More importantly, in our multivariable regression analysis, all the above-mentioned factors were found to be significant determinants of mortality, suggesting that ICU-acquired sepsis may not on its own be a causative factor for mortality. Nevertheless, nosocomial infections are responsible for prolonged stays in the ICU and increased costs \[[@CIT0025], [@CIT0026]\]. Positive isolates were obtained in 70% of the patients with sepsis, a similar finding to that reported in other studies \[[@CIT0001], [@CIT0019], [@CIT0027], [@CIT0028]\]. Two thirds of these patients had Gram-negative organisms isolated and one half had Gram-positive organisms isolated. The most common Gram-negative microorganisms recovered were *E coli*, *Klebsiella* spp, *Pseudomonas* spp, and *Acinetobacter* spp, as in previous studies \[[@CIT0001], [@CIT0027], [@CIT0028]\]. It is interesting to note that Gram-positive organisms were more common in North America than in other parts of the world; MRSA was also more common in North America than in other parts of the world except the Middle East. These findings are important when using guidelines for management of infection and sepsis, because guidelines developed in one part of the world, for example North America, may not be relevant to other areas. The results also underline the ongoing importance of fungal infections, which were involved in 13% of cases of sepsis overall, although the frequency was lower in the United States (5%), perhaps because more stringent criteria are used to characterize fungal infections in the United States. Finally it is noteworthy that approximately 42% of patients without sepsis received antimicrobial agents. The reasons for this are unclear, but antimicrobials may still be prescribed despite sepsis resolution or exclusion. In a retrospective analysis of 269 patients who were diagnosed with suspected sepsis in the emergency department and started on antibiotic therapy, 29% of the patients were found not to have bacterial disease, but the median duration of antibiotics in these patients was still 7 days (IQR, 4--10) \[[@CIT0029]\]. Intensive care unit mortality rates in patients with sepsis were approximately 26% and were twice as high as those in nonseptic patients. This percentage is lower than the 32% observed in the SOAP study (using their "severe sepsis" definition that is equivalent to our current definition of sepsis) \[[@CIT0001]\] and in other studies \[[@CIT0001], [@CIT0019], [@CIT0027], [@CIT0028]\]. Intensive care unit mortality rates in patients with septic shock were approximately 35%, a percentage that is also lower than that reported in earlier studies \[[@CIT0001], [@CIT0005]\]. Increased awareness of sepsis diagnosis and improved early management may have contributed to improved outcomes over time. Mortality rates varied around the globe, but in multivariable analysis, the between-country variation was not significant. These findings are in contrast to those from the International Multicenter Prevalence Study on Sepsis (IMPreSS) study of 1794 patients with sepsis from 62 countries, in which mortality rates were higher in East Europe and Central/South America compared with North America after adjustment for adjusted for ICU admission, sepsis status, location of diagnosis, origin of sepsis, APACHE II score, and country \[[@CIT0030]\]. As expected, nonsurvivors were older and had more comorbidities. As in previous ICU studies \[[@CIT0001], [@CIT0002]\], *Pseudomonas* and fungal infections were associated with worse outcomes, although only *Acinetobacter* infection was an independent predictor for hospital death in the multilevel analysis. More importantly, our data do not infer a cause-effect relationship, and the presence of *Acinetobacter* may simply be a marker of severity. In a systematic review of 6 matched case-control and cohort studies, Falagas et al \[[@CIT0031]\] reported that *Acinetobacter* infection was associated with increased attributable mortality, although others have suggested no independent link between *Acinetobacter* infection and increased risk of death \[[@CIT0032]\]. Mechanical ventilation at any time during the ICU stay and pre-existing liver cirrhosis were also important prognostic factors, more than doubling the risk of death. Use of renal replacement therapy at any time during the ICU stay was also associated with increased mortality. We also identified significant between-center variation, suggesting that differences in local ICU organization may have an impact on outcomes of patients with sepsis. Some of the potential factors associated with between-center outcomes differences have been identified in the literature. In an international cohort of 13796 ICU patients, Sakr et al \[[@CIT0033]\] reported that a high nurse/patient ratio was independently associated with a lower risk of in-hospital death. Gaieski et al \[[@CIT0034]\] reported that sepsis outcomes were improved in centers with higher sepsis case volumes. In a multicenter study in Canada, Yergens et al \[[@CIT0035]\] reported that ICU occupancy\>90% was associated with an increase in hospital mortality in patients with sepsis admitted from the emergency department. We are unable to identify which particular organizational factors may have influenced outcomes from our data, and this is an area that needs further study. Our database was very large, including considerable data on demographics, organ function, and outcomes. Nevertheless, to successfully collect a large amount of data in many ICUs requires some limitations in the level of detail of the collected data; therefore, we did not collect precise information on all subtypes of microorganisms or their resistance patterns or on the appropriateness of antimicrobial coverage. Moreover, data were collected by ICU doctors or research nurses who may not have specific expertise in infectious diseases, although the significance of this is uncertain. Our study has other limitations. First, although the audit included a large number of ICUs, the purely voluntary nature of the participation may have an impact on the representativeness of the data. Second, data collection was not monitored so small errors could not be corrected; only obvious incongruous data were verified. Third, in some countries, identification of microorganisms may have been incomplete because of the limited availability of microbiological testing. Moreover, the quality of the antimicrobials used in the treatment of infection has also been questioned in low-resource countries \[[@CIT0036]\]. Fourth, there was no means of differentiating between colonization and infection for some organisms, including *Acinetobacter* and coagulase-negative staphylococci. Therefore, microorganisms were weighted equally in the multilevel analysis. The absence of comparative large epidemiologic data that address this issue makes it difficult to judge whether the estimates of microorganisms provided in our study overestimate the frequency of these infections. Fifth, data were collected for the same period in all regions and therefore do not take into account any possible influence of seasonal variation. Sixth, we did not use the exact recent Sepsis-3 definitions \[[@CIT0037]\], which were published after our study, partly because we had no data on the evolution of SOFA scores before ICU admission and blood lactate levels were not available in all patients. Nevertheless, we used a definition based on the presence of organ dysfunction, a key feature of Sepsis-3. Finally, despite adjusting for a large number of variables that may influence outcome, the results of the multilevel analysis could not take into account other unmeasured variables that may have been of potential significance. CONCLUSIONS {#s12} =========== Sepsis, as defined by infection with organ dysfunction, remains a major health problem in ICU patients worldwide, associated with high mortality. There is wide variation in sepsis rates, causative microorganisms, and outcome in ICU patients around the world. A history of liver cirrhosis or metastatic cancer, use of mechanical ventilation or renal replacement therapy, and *Acinetobacter* infection were independently associated with an increased risk of in-hospital death. Global epidemiological data such as these help increase awareness of sepsis and provide crucial information for future healthcare planning. Further studies in this field should be done on a regular basis with standardized methodology to ensure the comparability of the results. Supplementary Data {#s13} ================== Supplementary materials are available at *Open Forum Infectious Diseases* online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author. ###### Click here for additional data file. We thank Dr. Hassane Njimi (Department of Intensive Care, Erasme University Hospital, Brussels, Belgium) for help with the statistical analyses. ***Potential conflicts of interest.*** All authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest.

Self-assembled monolayers (SAM) have been studied intensively due to their technological importance in various fields, from organic electronics to friction and catalysis (see, for example, refs [@b1], [@b2], [@b3], and references therein). Commonly it is assumed that the structure of the molecules within the SAM is only weakly dependent on their supra-molecular assembly. However, this is not necessarily the case, especially when dealing with SAM based on molecules well-known to exhibit structural diversity, such as oligopeptides. Experiments performed more than a decade ago showed that the spin selectivity of photoelectrons transmitted (from a gold substrate) through an oligopeptide monolayer is affected by the direction of the dipole moment of the molecules relative to the electrons\' velocity[@b4] and that both are temperature-dependent[@b5]. These results are summarized in [Fig. 1](#f1){ref-type="fig"} for a gold-adsorbed monolayer of oligoalanine consisting of 22 amino acid units. By monitoring the work function of the sample (by means of its contact potential difference (CPD), with a reference gold surface) as a function of temperature, a change in the sign of the CPD is observed ([Fig. 1a](#f1){ref-type="fig"}). Although the null CPD signal does not necessarily coincide with the dipole flipping point, owing to the Pauli Push-back effect that changes the dipole of a monolayer-adsorbed surface even in the absence of a molecular dipole[@b6][@b7]. this change does reflect a change in surface dipole[@b8]. The dipole flip correlates with a change in the sign of the preferred spin of the transmitted photoelectrons ([Fig. 1b,c](#f1){ref-type="fig"}). While the data of [Fig. 1](#f1){ref-type="fig"} may hint at a structural effect, the origin of the dipole and spin flipping remained unknown when these results were published. Indeed, determining structural changes in a monolayer at different temperatures, and correlating them with dipole and spin-filtering properties, is not a simple task. Here, we augment the traditional characterization tools with a thorough computational investigation using molecular dynamics (MD) and density functional theory (DFT), as well as with a new experimental setup---the magnet-less Hall device, based on the recently discovered chirality-induced spin selectivity (CISS) effect[@b9][@b10][@b11][@b12]. The combined computational and experimental results reveal that the oligopeptide structure is highly temperature-dependent. We find that the surrounding chemical environment in the SAM can induce 'denaturation\' upon cooling, much as in biological systems 'cold denaturation\' can take place due to protein interaction with the surrounding water[@b13], but here the SAM plays the role of the solvent. This dramatic structural change results in a flip in the direction of the dipole moment of the adsorbed molecules, which, in turn, reverses the polarity of the preferred spin in electron transfer experiments. This phenomenon is of importance not only for the basic understanding of the structure of adsorbed molecules, but also opens new possibilities for controlling spin in organic spintronic devices[@b14]. Results ======= Computational results --------------------- Our working hypothesis at the outset of this investigation was that the oligopeptide chains may be entropy-driven to fold with increasing temperature. For investigating this initial idea, classical MD simulations were performed on a dense monolayer of 10-mer oligoalanine peptides attached covalently to a surface using the GROMACS package (see Methods section for details). Because the MD force fields are commonly calibrated at room temperature, simulations at very low temperatures are not necessarily reliable as they cannot be expected to represent the system energetics accurately. This difficulty was circumvented by avoiding the need to study the peptide chain explicitly over a wide temperature range. Instead, we tested the effect of folding directly by starting from two distinct organizations of the peptides in the self-assembled monolayer, an α-helical and an extended conformation. Each system was simulated for several nanoseconds at 300 K and was analyzed by focusing on the conformational energetics and dynamics of the central polyalanine peptide in the simulated monolayer. As shown in [Fig. 2a](#f2){ref-type="fig"}, the two configurations of the polyalanine monolayer exhibited different properties, with the monolayer of the helical peptides being more disordered and less packed. We further used the distribution of the inter-peptide angle ([Fig. 2c](#f2){ref-type="fig"}) to estimate the contribution of this degree of freedom for the configurational entropy. Using the simple relation *S*=−*k p*ln*p*, where *p* is the probability of the angle, we found that at room temperature this degree of freedom contributes 0.33 kcal mol^−1^ for the helical 10-mer monolayer but only 0.005 kcal mol^−1^ for the linear 10-mer monolayer. Confirming our hypothesis, the stability of the α-helix form at room temperature, thus, is indeed due to its relatively high entropy, while the linear form is more stable at low temperature because of its lower enthalpy. For example, the total number of hydrogen bonds for the central peptide is much larger when the peptide is linear than helical, as shown in [Fig. 2b](#f2){ref-type="fig"}. The detailed calculations suggest that polyalanine adsorbed as a self-assembled monolayer is going through 'cold denaturation\' upon cooling and is gradually transformed from the α-helix structure to an extended β-strand shape. To examine the relation between the cold denaturation and electrical properties, we 'extracted\' the central peptide from MD runs with various degrees of stretching and examined the gas-phase dipole and electronic structure obtained from DFT calculations. Importantly, these calculations were performed using the optimally-tuned range-separated hybrid scheme (see Methods section)[@b15][@b16]. This approach was shown to provide accurate results for the outer-valence electronic structure of finite systems in general[@b17][@b18] and in particular predict their ionization potential and electron affinity from the highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO), respectively[@b15][@b16][@b19][@b20][@b21][@b22][@b23], where standard schemes typically fail[@b24][@b25]. This is of unique importance for highly polar systems[@b26], including protein and peptide molecules[@b27][@b28][@b29], where a strongly underestimated gap may lead, under a large internal electric field, to spurious gap closure and qualitative failure of the calculation. Representative calculated structures of the model, 10-mer oligoalanine system are shown in [Fig. 3a](#f3){ref-type="fig"}. Importantly, and in agreement with the early experiments of [Fig. 1](#f1){ref-type="fig"}, we find that as the helical structure gradually unfolds and the terminal groups move further apart, the dipole becomes less negative and eventually changes sign. The flip in dipole direction upon structural change is not unique to this specific length of the peptide and also appeared in calculations of longer (20-mer) and shorter (4-mer) oligoalanines, not shown for brevity. The main effect causing the dipole flip is demonstrated schematically in [Fig. 3b](#f3){ref-type="fig"} for a 10-mer oligoalanine. For a linear conformation, the individual contributions of the local dipole moments formed by each of the peptide bonds add up coherently along the molecular axis, in the direction opposite to the dipole formed between the C-terminus and the N-terminus of the peptide. In the α-helical form, the direction of the same peptide-bond related local dipoles is flipped, due to additional hydrogen bonds that set the directionality of the oxygen atoms and change the relative direction of the O--N pairs in the peptide bond. The contribution of these local dipoles to the net dipole is then in the same direction as the contribution of the C-terminus to N-terminus dipole. This results in a larger net dipole moment in the case of the α-helix structure, which has an opposite sign to that of the linear structure. Accordingly, two different structures of similar length can lead to different dipole moments with different signs, as demonstrated by structures 3 and 4 in [Fig. 3a](#f3){ref-type="fig"}. Both structures are ∼25 Å long, but while in the somewhat shorter structure 3 many peptide bonds (which are closer to the C-terminus) are still within a helical form, in the somewhat longer structure 4 almost all peptide bonds are linearly aligned. This causes the dipole to flip sign along the molecular axis. Furthermore, the change of dipole sign is accompanied by a related flip in the position of the occupied and unoccupied orbitals, as visualized in [Fig. 3a](#f3){ref-type="fig"} by the charge density distributions of the HOMO and LUMO. This can be easily rationalized in terms of quasi-electron and quasi-hole states responding oppositely to the electric field[@b30]. Experimental results -------------------- Encouraged by the above theoretical findings, we sought direct experimental evidence for chain elongation with decreasing temperature, through a combination of ellipsometry and FTIR (Fourier transform infrared spectroscopy see Methods section). Note that here and throughout this article, experiments were performed on {Ala-Aib} oligopeptides. These were chosen because they possess the same structure as the pure alanine-based ones, discussed so far, but exhibit better solubility. As shown below, they exhibit the exact same phenomena as the oligoalanine peptides. The IR spectroscopy showed consistently an increase in the absorption signal upon cooling to 14 K, with no significant changes observed below 100 K. However, while these data are consistent with the denaturation process, they do not constitute a clear indication for it. Ellipsometry data, given in [Table 1](#t1){ref-type="table"}, did provide clear evidence for chain elongation upon cooling. Experimentally, a length change that depends on the cooling time was observed. For very long cooling times the length is approximately doubled, in agreement with the theoretical prediction. The remaining question, then, is the origin of the spin-transfer selectivity ([Fig. 1b](#f1){ref-type="fig"}) and whether cold denaturation can also explain its polarity inversion. Generally, spin-selectivity in electron transmission or tunneling through chiral layers is explained by the CISS effect[@b9][@b10][@b31], which has also been used to explain spin-specific electron transfer through oligopeptides[@b32] and even suggested to play a role in electron transfer through biological systems[@b33]. Briefly, as an electron moves along a chiral molecule it experiences the electrostatic potential of the molecule, which is chiral. In the electron\'s rest frame, the charge that defines the electrostatic potential generates a magnetic field which splits the degeneracy of the electron spin states. The sign of the spin being transferred more efficiently through the chiral molecules depends on the sign of the electric field acting on the electron. This is because spin-selectivity is controlled by spin--orbit coupling, which has the well-known form, , where is the electron velocity and is the electric field. Because the dipole flip induced by the cold denaturation necessarily implies a change in sign of the overall electric field across the oligopeptide, reversal in spin-filtering polarity emerges naturally. While the above explanation is plausible, proving experimentally that indeed the preferred spin inversion is inexorably linked to structural changes is more difficult. To pursue this, we developed a new type of Hall-effect device, shown in [Fig. 4](#f4){ref-type="fig"} (see [Supplementary Fig. 1](#S1){ref-type="supplementary-material"}). The Hall effect is a long-known phenomenon[@b34]. Briefly, when current is flowing in a substrate between two electrodes, a magnetic field applied perpendicular to the current flow induces an electric potential perpendicular both to the current and the magnetic field. This effect is used as a standard tool for the study of properties of semiconductors, notably charge carrier concentration and mobility[@b35]. Here, we demonstrate that the spin-filtering capabilities of the chiral monolayer allow for the observation of the Hall effect without either an externally applied magnetic field or a permanent magnet. Instead, as explained below, the hall voltage arises from the accumulation of spin-transfer. This allows us to probe the relation between structural changes and spin inversion directly. The device, shown schematically in [Fig. 4](#f4){ref-type="fig"}, is based on a two-dimensional electron gas structure, onto which two sets of Hall probe electrodes were evaporated. We have chosen to implement the device on a GaN substrate due to the long spin-lifetime of this material[@b36]. The width and length of the conducting channel were 40 and 700 μm, respectively. On the conduction channel, a SAM of NH~2~-{Ala-Aib}~8~-COCHNH~2~CH~2~PO~4~H~2~ or SHCH~2~CH~2~CO-{Ala-Aib}~5~-COOH oligopeptides were adsorbed, that is, either a phosphonate or a carboxylate group was used to bind the oligopeptide to the GaN surface. The IR spectra of the adsorbed layer are shown in [Supplementary Fig. 2](#S1){ref-type="supplementary-material"}. The oligopeptides chosen possess the same structure as the pure alanine-based ones discussed so far and were chosen because of their better solubility. CdSe nano particles (NPs) of ∼6 or ∼2 nm diameter were then attached to the amine or thiol tail. The SEM image of the layer with the nanoparticles is shown in [Supplementary Fig. 3](#S1){ref-type="supplementary-material"} and additional information is provided in the [Supplementary Methods](#S1){ref-type="supplementary-material"}. A constant current of 10 μA was injected along the structure, with the Hall voltage measured in the transverse direction (see Methods section). The device was calibrated using an external magnetic field as shown in [Supplementary Fig. 4](#S1){ref-type="supplementary-material"}. As expected from standard theory, no Hall voltage is observed in the absence of a magnetic field under dark conditions. However, as shown in [Fig. 5](#f5){ref-type="fig"}, a Hall voltage appears - with either binding group, with two different sizes of NPs, and at both high and low temperature when the CdSe NPs are illuminated (with a 2.3 mW, 514 nm laser). Strikingly, the Hall voltage is of opposite sign at low and high temperatures. Discussion ========== How can these results be explained and how are they related to the spin-polarization results of [Fig. 1](#f1){ref-type="fig"}? A model explaining these results is given in [Fig. 6](#f6){ref-type="fig"}. Upon excitation of the NPs ([Fig. 6a](#f6){ref-type="fig"}), an electron is transferred from the substrate to the hole state on the NP ([Fig. 6b](#f6){ref-type="fig"}). Because the oligopeptides are chiral, it is expected that the electron transfer process is spin selective, according to the above-explained CISS effect. An electron with the opposite spin is then accumulated on the device surface, acting to magnetize the substrate and create an effective magnetic field for the Hall measurement. If spin--orbit coupling in the NP is large (which is the case for CdSe), the spin of the photo-excited electron on the NP is randomized such that it is spin-selected for transfer to the substrate, still within the lifetime of the net spin formed on the substrate side ([Fig. 6c](#f6){ref-type="fig"}). Both substrate and NP are now neutral again, but the net spin on the substrate is further enhanced ([Fig. 6d](#f6){ref-type="fig"}). If the spin-lifetime in the substrate is long relative to that in the NPs (which is indeed the case for a GaN substrate and CdSe NPs), a specific spin is accumulated on the device surface, creating net magnetization[@b37]. A Hall voltage can then be measured. Its magnitude must be proportional to the spin accumulation field and its sign must reflect the type of spin accumulated. Under continuous illumination, the net effect will naturally depend critically on the lifetime of the spin state at the substrate surface, the number of spins being transferred between the substrate and the NP during this time, and the rate of spin-randomization. The device presented here is unique because it allows, in principle, a high spin-polarization without a net charge polarization. Therefore, high sensitivity is achieved here due to spins 'loaded\' onto the device without encountering the difficulty of Coulomb repulsions that one may face when trying to spin-polarize a medium by injecting spin-polarized electrons. In this sense, the situation is similar to the known case of spin--torque transfer used to spin-polarized ferromagnetic materials[@b38]. With the model of [Fig. 6](#f6){ref-type="fig"} in mind, the results of [Fig. 5](#f5){ref-type="fig"} can now be rationalized. First and foremost, the spin transferred at room temperature is aligned parallel to the electron velocity, while at low temperature it is antiparallel to the velocity. This is in full agreement with an inversion of the CISS effect owing to the electrical dipole flip predicted theoretically in [Fig. 3](#f3){ref-type="fig"}. Other insights can also be drawn from the figure. Clearly, the Hall signal is larger in the case of the phosphonate linker than in the case of the carboxylic one. This result is consistent with reports on the stronger binding of the phosphonates to GaAs[@b39] and we assume, in light of these results, that the same applies to GaN. In addition, while at room temperature spin randomization is efficient both in the large and small NPs, due to phonon--spin interactions, at low temperature the signal is larger for small NPs. This can be explained by the quenching of the phonon--spin interaction and by the known faster spin-randomization in smaller NPs[@b40][@b41], by as much as an order of magnitude, due to spin interaction with the surface[@b41]. Therefore, when the spin randomization in the NPs is fast, the NPs can be excited many times during the lifetime of the spin polarization on the GaN (see [Fig. 5](#f5){ref-type="fig"}) and many spin-pairs can be injected into the GaN per NPs. Using a calibration curve (see [Supplementary Fig. 4](#S1){ref-type="supplementary-material"}), we can estimate the effective magnetic field that would be required for obtaining the same Hall voltage from an ordinary Hall measurement. The fields are found to vary between several Gauss to almost 100 G. Because the density of atoms in GaN is ∼1 × 10^23^ atoms per cm^3^ and the volume of the dielectric (40 μm × 700 μm × 22 nm) is ∼6 × 10^−10^ cm^3^, the total number of atoms in this volume is about 6 × 10^13^. A simple organization of 10^12^ spins can create such a magnetic field if they are equally spaced in a box of the size of the dielectric above the conduction channel. Because the number of spins in this volume is significantly smaller than the number of atoms (by more than an order of magnitude), we can safely assume that the exchange interaction between them, which may limit spin-polarization, is not important. This number of spins implies that ∼10^3^--10^4^ spins must be transferred from each adsorbed NPs during the lifetime of spin-polarization in the GaN, which is between tens to hundreds of nanoseconds[@b36]. The spin lifetime on the NPs is of the order of a few psec[@b40][@b41]. This ratio of lifetimes is about an order of magnitude more than that required to produce the above net spin polarization, showing that the system can indeed produce the required effective magnetic field by spin polarization that does not involve charge polarization. Last but not at all least: Beyond the computational evidence, can we establish a direct link between the above-discussed structural changes and the spin-transfer inversion? To probe this, the magnet-less Hall experiment was repeated with a *mixed* monolayer, consisting of 40% NH~2~-{Ala-Aib}~8~-COCHNH~2~CH~2~PO~4~H~2~ and 60% 11-mercaptoundecylphosphoric acid \[HSCH~2~(CH~2~)~9~CH~2~PO~4~H~2~\]. The phosphoric acid comprises alkane chains and therefore cannot form hydrogen bonds with the oligopeptide. Therefore the enthalpy gain expected from cold denaturation is much smaller and the effect should be much smaller, if not completely inhibited. The results are shown in [Fig. 7](#f7){ref-type="fig"}. Indeed, for the mixed monolayers, the Hall voltage does not change sign upon cooling. Some signal reduction is observed at low temperature, which we attribute to the fact that the two molecules do not mix ideally and there may well be small domains that include the oligopeptides alone and exhibit denaturation as above. Still, the results of [Fig. 7](#f7){ref-type="fig"} lend very strong support to the theoretical prediction and close the circle between denaturation, dipole inversion, and spin-transfer inversion. To summarize, using MD and DFT calculations we have been able to explain hitherto baffling experimental results that indicated inversion of the dipole and spin-filtering properties of chiral oligopeptide monolayers upon cooling. The calculations showed that the α-helix structure of the adsorbed oligopeptides 'stretches\' upon cooling, in a manner reminiscent of 'cold denaturation\' in biological systems, but where the monolayer plays the role of the solvent. Furthermore, this structural change induces an inversion of the peptide dipole, owing primarily to rearrangement of hydrogen bonds. Phenomenological theory of the CISS effect then explains that this dipole flip must be accompanied by a concomitant change in the spin that is preferred in electron transfer experiments. We have then supported the structural prediction via ellipsometry measurements and confirmed the link between the structural changes and the spin-filtering properties using a novel magnet-less, solid-state hybrid organic--inorganic Hall device, employed with oligopeptide monolayers based on alanine and aminoisobutyric acid. The results were further supported by control experiments on mixed monolayers. The results presented indicate that inter-molecular interactions may be of importance in defining the structure of adsorbed molecules within a monolayer and its temperature-dependence may be strongly related to monolayer packing. In addition, we introduced here a new device that can monitor structural changes in adsorbed chiral molecules by their spin filtering properties. In the future, this may serve to probe the role of spin in electron transfer through bio-systems and its dependence on system structure. Methods ======= Device fabrication ------------------ A schematic representation of the AlGaN/GaN Hall devices and the setup is given in [Fig. 4](#f4){ref-type="fig"}. The AlGaN/GaN Hall devices were fabricated by standard photolithography. The AlGaN/GaN HEMT Epi wafers on sapphire-substrates were purchased from NTT AT. The structure consist of the following layers, from the bottom up (see [Fig. 4d](#f4){ref-type="fig"}): 1800, nm bulk i-GaN, 20 nm i-AlGaN, and a capping layer of 2 nm i-GaN. Ohmic contacts were achieved by e-beam evaporation of a standard Ti/Al/Ni/Au stack, followed by rapid thermal annealing at 900 °C. Device isolation was done by mesa etching using a BCl~3~/Cl~2~ base ICP-RIE (Plasma Therm). Formation of monolayers ----------------------- Three organic monolayers were adsorbed as SAM on the GaN: 11-mercaptoundecylphosphoric acid (PCI Synthesis), NH~2~-{Ala-Aib}~8~-COCHNH~2~CH~2~PO~4~H~2~ and SHCH~2~CH~2~CO-{Ala-Aib}~5~-COOH (both from Genemed Synthesis Inc.). Solvents were reagent grade or better (Merck, Baker, or Bio-Lab). The GaN devices were sonicated prior to molecular adsorption in hot acetone and ethanol for 10 s each, then etched for 30 s in 6 M HCl, rinsed in water, and dried under a N~2~ stream. The samples were then cleaned and oxidized in UV/ozone oxidation for 30 min and placed in the adsorption solution (1 m[M]{.smallcaps} in toluene) immediately. Vials with the absorption solutions were filled with N~2~ and placed in a desiccator for 19 h in the case of 1-mercaptoundecylphosphoric acid and NH~2~-{Ala-Aib}~8~-COCHNH~2~CH~2~PO~4~H~2~, and for 65 h in the case of SHCH~2~CH~2~CO-{Ala-Aib}~5~-COOH. After adsorption, the samples were rinsed with toluene and dried with nitrogen. In order to form a monolayer of CdSe NPs on the top of the SAM coated substrates, the devices were immersed for 4 h in a solution of anhydrous toluene (99.8%, Aldrich) containing core-only CdSe with a diameter of 6.2--7.7 nm (MK Nano) or 2.4--2.6 nm (NN-Labs). The samples were than sonicated for 5--10 s and washed with toluene to remove excess NPs. More details are provided in the [Supplementary Methods](#S1){ref-type="supplementary-material"}. Monolayer characterization -------------------------- Monolayer formation was confirmed by FTIR in grazing-angle attenuated total reflectance mode (GATR-FTIR), using a ThermoScientific FTIR instrument (Nicolet 6700) equipped with a VariGATR accessory (Harrick Scientific) and with a single reflection Ge crystal. The spectra of the two oligopeptides, SHCH~2~CH~2~CO-{Ala-Aib}~5~-COOH and NH~2~-{Ala-Aib}~8~-COCHNH~2~CH~2~PO~4~H~2~, exhibit the characteristic peaks that include stretching frequencies at 1659 and 1648, cm^−1^, related to the amide I band, and peaks at 1544 and 1539, cm^−1^ due to the amide II band ([Fig. 8a,b](#f8){ref-type="fig"}). The spectrum of 1-mercaptoundecylphosphoric acid monolayer exhibits two large peaks at 2850 and 2917, cm^−1^, attributed to the symmetric and asymmetric CH~2~ stretching, respectively ([Fig. 8c](#f8){ref-type="fig"})[@b42][@b43]. A broad band between 1050 and 1200 cm^−1^ is related to the symmetric and antisymmetric P--O stretch and additional peaks associated with the P=O stretch appear at 1284 and 1315, cm^−1^ (ref. [@b44]). Hall measurements ----------------- The GaN Hall device was attached to a sample holder and electrically connected to the measuring units. A constant DC current of 10 μA was applied along the device using a Keithley 6221 current source and the Hall voltage (*V*~H~) across the device was measured using a Keithley Nanovoltmeter 2182A (see [Fig. 4](#f4){ref-type="fig"}). The resistance of the devices was in a range of 3--6 Ω. Chiral molecules were absorbed directly to the GaN, with CdSe NPs bound on top of them. The device was placed in between the magnetic poles and on a cold finger that could be cooled down to 14 K. A magnetic field of up to 0.5 T could be applied perpendicular to the sample plane by an electromagnet. The temperature of the sample holder was controlled by a temperature controller, with a temperature stability of 0.1% at 14 K and 0.3% at 300 K. For calibration, the conventional Hall voltage, as a function of external magnetic field, was measured for each device at several temperatures, 300, 120, 60 and 14 K. A representative plot taken at 300 K with a device coated with a monolayer of NH~2~-{Ala-Aib}~8~-COCHNH~2~CH~2~PO~4~H~2~ is shown in [Fig. 9](#f9){ref-type="fig"}. Several control experiments were performed with circularly polarized light on devices coated with the different monolayers and with large and small nanoparticles, as shown in [Supplementary Figures 5,6](#S1){ref-type="supplementary-material"} and on a bare device (see [Supplementary Fig. 7](#S1){ref-type="supplementary-material"}) and with a device coated with 1-mercaptoundecylphosphoric acid and illuminated with circularly polarized light (see [Supplementary Fig. 8](#S1){ref-type="supplementary-material"}). Molecular dynamics simulations of a polyalanine monolayer --------------------------------------------------------- A monolayer composed of 25 Ala10 polyalanines was prepared *in silico* by organizing the peptides initially in 5 × 5 arrays. The longitudinal axes of the peptides were oriented along the Z axis of the simulation box. The peptide molecules were neutralized at the N-terminal end (contained a protonated amino group) and at the C-terminal end (contained a protonated carboxyl group). To imitate the monolayer conditions, the N-terminal nitrogen of the amino group of each peptide was constrained in the Z dimension, which results in the ability of the peptides in the brush to move freely only in the XY plane. Any other peptide atoms were completely free to move. The peptides were attached in either an α-helical or a linear (extended β-strand) conformation. The initial distance between neighboring peptides was chosen empirically, such that the neighboring chains of ALA10 were placed as closely as possible, but without inter-chain clashes. During the MD the peptides were free to move in the XY plane and change the distance between neighboring peptides. The designed monolayers were studied using all-atom molecular dynamics simulations using the GROMACS package, Version 4.5.4 (ref. [@b45]) and with the CHARMM27 force field in vacuum conditions to mimic the dry environment of the monolayer[@b46]. In the initial step of the simulation conformations of the studied monolayers, the systems were relaxed using a conjugate gradient method of energy minimization. The LINCS algorithm[@b47] was used to control bonds during the simulation. The Leapfrog algorithm was employed with a step of 2 fs. The 5 × 5 monolayers were simulated for 10 ns (analyses of the trajectories showed that longer simulations did not result in additional conformational changes) with a time step of 2 fs (ref. [@b28]). Temperature was controlled at 300 K using the Berendsen thermostat. The conformations of Ala10 that were studied by DFT were generated by a pulling MD simulation when starting from a helical conformation. In this simulation, a force is applied at each terminal of the peptide at opposite directions. The pulling is done at rate of 0.01 nm/ps and with force constant of 1000, kJ per mol nm^2^. Several conformations were sampled along the transition from the α-helical to the β-strand structures, providing conformations with length varying between 15 and 32 Å. Density functional theory calculations of gas-phase oligopeptides ----------------------------------------------------------------- All gas-phase electronic-structure calculations were performed with Q-Chem[@b48], version 4.0, using the correlation-consistent polarized valence triple zeta basis set. We used the optimally-tuned range-separated hybrid approach[@b15], with a functional containing 80% short-range Perdew-Burke-Ernzerhof[@b49] exchange, 20% short-range Fock exchange, 100% long-range Fock exchange and full Perdew-Burke-Ernzerhof correlation. The range-separation parameter was optimally tuned so as to obey the ionization potential theorem for the peptides in the neutral form in case of an unbound LUMO level, and to obey the same theorem in both neutral and anionic forms for a bound LUMO level. Additional information ====================== **How to cite this article**: Eckshtain-Levi, M. *et al.* Cold denaturation induces inversion of dipole and spin transfer in chiral peptide monolayers. *Nat. Commun.* 7:10744 doi: 10.1038/ncomms10744 (2016). Supplementary Material {#S1} ====================== ###### Supplementary Information Supplementary Figures 1-8 and Supplementary Methods We acknowledge the help of Dr Anup Kumar in performing the ellipsometry measurements. ME, EC, SM and RN acknowledge the support by the ERC-Adv grant and the Isreal Science Foundation. RN and YP acknowledge the support of the VW Foundation and the Israel Ministry of Science. SRA, SS and LK acknowledge support by the European Research Council, the Israel Science Foundation and the Lise Meitner Center for Computational Chemistry. SRA acknowledges an Adams Fellowship of the Israel Academy of Sciences and Humanities. **Author contributions** M.E.-L and S.P.M. performed the experiments. E.C. produced the GaN devices and contributed to the data analysis. S.R.-A and S.S performed the DFT calculations. Y.G. performed the MD calculations. Y. P., Y. L., L. K. and R. N. conceived the project, supervised the experimental and computational work, assisted the analysis of the data and wrote the manuscript. ![Temperature-dependent contact potential (surface work function) and spin-polarized photoemission.\ (**a**) Temperature- dependent contact potential difference (CPD) measurement of gold coated with a self-assembled monolayer of L-polyalanine, bound to the surface through the carbon terminal. Zero voltage indicates the CPD of the bare gold. (**b**,**c**) Energy distribution of photoelectrons ejected from the gold substrate using a 248 nm laser with clockwise (+, blue) or counter clockwise (−, red) circularly polarized light, at temperatures of 297 K (**b**) or 250 K (**c**). Adapted from ref. [@b5], used with permission.](ncomms10744-f1){#f1} {#f2} {#f3} {#f4} {ref-type="fig"}, coated with 6 and 2 nm diameter CdSe nanoparticles that are attached to the GaN surface via NH~2~-{Ala-Aib}~8~-COCHNH~2~CH~2~PO~4~H~2~ or SHCH~2~CH~2~CO-{Ala-Aib}~5~-COOH molecules, at a current of 10 μA. The devices were illuminated with a 2.3 mW, 514 nm laser. Three sets of devices were measured and the variation in the signal among them is about 10%. The signal observed corresponds to effective magnetic fields of 20, 16, -5, -18, 73, 36, -8 and -22 G for panels **a**,**b**,**c**,**d**,**e**,**f**,**g** and **h**, respectively. All plots are presented after subtracting a background signal, measured for the same devices but covered with the non-chiral 11-mercaptoundecylphosphoric acid (see [Supplementary Fig. 8](#S1){ref-type="supplementary-material"}). Pink and light-blue areas correspond to 'light on\' and 'light off\' regimes, respectively.](ncomms10744-f5){#f5} {#f6} {ref-type="fig"} and [5](#f5){ref-type="fig"}, but with a mixed monolayer consisting of 40% NH~2~-{Ala-Aib}~8~-COCHNH~2~CH~2~PO~4~H~2~, 60% 11-mercaptoundecylphosphoric acid \[HSCH~2~(CH~2~)~9~CH~2~PO~4~H~2~\] molecules. The signal observed corresponds to effective magnetic fields of 11 and 3 G for the high and low temperature, respectively.](ncomms10744-f7){#f7} {#f8} {ref-type="fig"} was measured, with the Hall voltage shown as a function of applied magnetic field, measured for a device coated with NH~2~-{Ala-Aib}~8~-COCHNH~2~CH~2~PO~4~H~2~ monolayer.](ncomms10744-f9){#f9} ###### Monolayer thickness as a function of cooling time at 70K extracted from ellipsometry for a monolayer made from NH~2~-{Ala-Aib}~8~-COCHNH~2~CH~2~PO~4~H~2~ and adsorbed on Si/SiOx. **Cooling time (min)** **Thickness of molecule before cooling (Å)** **Thickness of molecule after cooling (Å)** **Difference in thickness (Å)** ------------------------ ---------------------------------------------- --------------------------------------------- --------------------------------- 30 16±1 20±1 4 60 18±1 25±1 7 120 16±1 29±1 13 [^1]: These authors contributed equally to this work.

Citation: Khodayarian M, Mazloomi-Mahmoodabad SS, Lamyian M, Morowatisharifabad MA, Tavangar H. Response costs of mammography adherence: Iranian women's perceptions. Health Promot Perspect. 2016;6(2):85-91. doi:[10.15171/hpp.2016.15](http://dx.doi.org/10.15171/hpp.2016.15). Introduction {#s1} ============ Breast cancer, as a global health problem, is a prevalent cancer among women, worldwide.^[@R1]^ it is, also, the most frequently diagnosed cancer among women in countries of the Eastern Mediterranean Regional Office (EMRO) and the fifth common cause of death among women in Iran. In general, fundamental changes in the socioeconomic status among developing countries have made a shift in the lifestyles of people in these countries to those of the industrialized countries. Such a shift in the lifestyle, which is associated with reproductive, hormonal, and dietary risk factors, has led to an increase in the burden of breast cancer.^[@R2]^ It is noteworthy that the age of onset for breast cancer among Iranian women is reported to be up to ten years earlier than that in the developed countries.^[@R3]^ Epidemiological studies based on the age-specific incidence in Yazd province showed breast cancer as the most prevalent cancer.^[@R4]^ Despite the annual increase in the incidence rate of the disease among Iranian women, there is a lack of recent evidence on its various aspects in Yazd, Iran.^[@R5]^ Mammography screening has been shown to be associated with early detection and, therefore, mortality and morbidity reduction of breast cancer.^[@R6]^ Although mammography screening programs are being implemented for all older than 40 women in several countries like the United Kingdom and the United States,^[@R7]^ there is no formal mammography screening program in Iran. Unfortunately, 70% of Iranian women with breast cancer are diagnosed in the advanced stages, when it is too late for successful treatment.^[@R8],[@R9]^ It is estimated that patients with breast cancer living in the lower-income countries have lower 5-year survival rates in comparison with their counterparts in the higher-income countries at any stage of the diagnosis.^[@R10]^ Lack of mass media education and screening programs, poverty, poor access to health care facilities, cultural barriers, and poor breast awareness are associated with lower breast cancer survival in the developing countries.^[@R11]^ Based on those mentioned above and, also, the recommendation of the World Health Organization (WHO) on mammography as the only effective breast cancer screening method, it seems to be necessary to encourage Iranian women for mammography adherence (MA).^[@R12]^ There are some perceived barriers influencing the mammography behavior; for instance, Donnelly et al in a study on the Qatari women showed that socioeconomic factors and awareness of recommended national guidelines of breast cancer screening may influence breast health seeking behaviors.^[@R13]^ Moreover, Lwin reported the lack of mammography awareness among Singaporean women in a way that they did not believe on the necessity of mammography behavior.^[@R14]^ According to the findings noted by Thomas et al, Iranian women experience several barriers in relation with breast cancer prevention and control such as the lack of knowledge, personal functioning, motivation, availability of information, health communication, and support.^[@R15]^ In the present qualitative study, the framework for interview questions and data analysis was based on the protection motivation theory (PMT). PMT was originally proposed by Rogers in 1975. It is encompasses of two independent appraisal pathways resulting from fear appeals; threat appraisal and coping appraisal.^[@R16]^ people might respond adaptively or maladaptively to the health information.^[@R17]^ In the threat appraisal pathway, an increase in severity, vulnerability, and fear may result in a decrease in the probability of occurrence of the maladaptive response. In the coping appraisal pathway, people evaluate the usefulness of the behaviors (response efficacy) and their ability to perform, successfully, preventive behaviors (also called self-efficacy). Two important factors that might increase the likelihood of maladaptive response are intrinsic and extrinsic rewards (e.g., pleasure and social approval of unhealthy actions), and response costs (barriers). The key point in this theory is adaptive behavioral change, which is reinforced through motivation.^[@R18]^ Protection motivation is the result of these two cognitive pathways (i.e., intention to perform a recommended behavior like mammography).^[@R19],[@R20]^ Because of the differences exist in cultural contexts, the perceptions of women in Yazd about the breast cancer preventive behaviors may be different from those of the other countries. Thus, the findings in the other cultures may not be applicable for Iranian women. This study was designed to explore the perceptions of women about response costs of MA and to provide more insight on this cultural sensitive topic. Methods {#s2} ======= In this qualitative study, directed content analysis was employed to analyze the data obtained by in-depth interview approach. The semi-structured interviews were conducted after designing a flexible framework for interview questions based on the PMT. The method section was provided based on the consolidated criteria for reporting qualitative research (COREQ) guideline. In the current study, the following key question was explored: What are the perceptions of women about the response costs of MA? The study participants were 12 women, one oncology nurse, and a breast cancer survivor, all from Yazd, Iran, who were invited to participate in the study through purposive sampling. The eligibility criteria for the women were considered to be as the following: (*a*) having the age of 35 years or higher, (*b*) not suffering from mental disorders based on their records, and (*c*) the ability to communicate. The inclusion criteria for the nurse were as follow: (*a*) being with at least one year of nursing experience in an oncology unit, (*b*) having, at least, a bachelor degree in nursing, and (*c*) being female. Breast cancer survivor and oncology nurse were elected as critical cases in this study. The critical cases were considered to be interviewed hoping to address the relations that were being studied.^[@R21]^ Since oncology nurses have close interaction with breast cancer patients and their families, they were considered as key informants or experts in the study. Their perceptions and experiences were valuable to arrive a deep understanding of the effective factors on women's breast health-seeking behaviors. Since the real experience of the participants is important in a qualitative study, a breast cancer survivor was chosen as a key informant, as well. The breast cancer survivors may have a good experience and deep understanding about the process of diagnosis and treatment of the disease. Moreover, as factors like nulliparity and the old age at the first live labor could enhance the chance of breast cancer occurrence, two old girls were interviewed as the persons at the higher risk for the disease. Thus, it was possible to get more insight about the subject. Maximum variation in sampling was considered with the participants' age, marital status, education, occupation, and the breast disease history. The interviews were continued until data saturation was appeared. In other words, emerging all possible aspects and domains of the main theme and categories was confirmed by the researchers. The pilot interviews were conducted with three women and the final form of the interview was prepared. The framework for the interview questions was the same in all three different groups of the participants (healthy women, oncology nurse, and breast cancer survivor). Meanwhile it was very flexible because of using semi-structured and in-depth individual interviews. So, when needed, some necessary changes were made in the interview questions by the interviewer. The first researcher was the main interviewer and data analyzer. A set of 14 audio-taped interviews was carried out from September to December 2014. The other researchers, who were faculty members affiliated with the Yazd University of Medical Sciences, Yazd-Iran and Tarbiat Modares University, Tehran-Iran, monitored and revised the data analysis process, step by step, to fulfill the systematic analysis. Fourteen participants aged 35 and older participated in the study. All were living in Yazd city, Iran. [Table 1](#T1){ref-type="table"} presents further demographic data. Table 1Participants′ demographics (n = 14)**Variables**Age, mean (SD)39.42 (4.61)Range35-48Marital status, n (%)Single2 (14.3)Married10 (71.4)Divorced2 (14.3)Education, n (%)Primary school1 (7.1)School diploma6 (42.9)Higher education7 (50.0)Occupation, n (%)Housewife1 (7.1)Employed13 (92.9)Breast disease history, n (%)Yes, Benign diseases4 (28.6)Yes, Breast cancer survived1 (7.1)No9 (64.3) The interviews were performed with prior appointment with the participants and they lasted for 30-90 minutes in a quiet and calm room at a clinic near to the participant's home. The interviews started with open and general questions; however, the questions became gradually more detailed as the interviews were being progressed. The interviewer tried to keep the participants in the main path of the study in order to achieve the research objectives. The main focus of the interview questions were "what is the meaning of cancer in your idea?," "what factors may prevent you from doing mammography?," and "what are the ideas of the persons close to you about MA?" The current study was an attempt to identify the response costs of MA among women in Yazd, as a part of a more comprehensive exploratory mixed method study for the first author's doctoral dissertation. Directed qualitative content analysis based on the constructs of the PMT was used to data analysis, but, in this paper the results of the qualitative data analysis on the construct of response costs were, only, reported. Data analysis {#s2-1} ------------- In this study, qualitative directed content analysis was performed using Burnard's 14-stage approach.^[@R22]^ Each interview was transcribed verbatim. The texts were imported into MAXQDA software version 10 for computer-assisted qualitative data analysis. The researcher read the texts several times to obtain a whole sense on the data. Unit of analysis and meaning units were considered. Open coding technique was adopted. The initial codes were organized into categories and subcategories. Different methods have been used in order to increase the trustworthiness of data analysis. For example, sufficient time was allocated for data collection and analysis. Constant comparative analysis was used throughout the process of analysis. The researcher bracketed her thoughts, preconceptions, and beliefs before and while performing the data analysis to ensure credibility. Some other strategies such as prolonged engagement with participants, faculty member's revisions, and member checks were used to enhance the rigority of the data. Results {#s3} ======= Analysis of the understandings of the Iranian women about their perceived response costs of MA led to formation of one main theme, two categories, nine subcategories, and 143 initial codes. Response costs factor was considered as the main theme. Psychological barriers and maladaptive coping modes were determined as categories. The subcategories underlying the category of psychological barriers were *embarrassment, worry about being diagnosed with cancer, preoccupation with underlying diseases, misconceptions about mammography, need for an accompanying person, and internalizing the experiences of the others*. The subcategories underlying the category of maladaptive coping modes were *religious faith, fatalism,* and*avoidance and denial*. Brief overview of the thematic structure of the results could be found in [Table 2](#T2){ref-type="table"}. The above mentioned categories and their related subcategories were explained separately in the following section. Table 2Thematic structure**Main ThemeCategorySubcategoryExtracted Codes**Response costsPsychological barriersEmbarrassmentEmbarrassment of older unmarried girls because of:\ 1) Narrow territory of to seek mammography\ 2) Late referral to a gynecologist\ 3) Being ashamed of disease follow-up\ 4) Bad impression of others about reasons to go to a gynecologist\ 5) Not being comfortable to announce the gynecologist visit time to the familyWorry about being diagnosed with cancer1) Worry about mammography result\ 2) Worry about hearing bad newsPreoccupation with underlying diseasesSuffering from another physical problemMisconceptions about mammography1) Harm of mammography for breast tissue\ 2) Uncertainty about the harmfulness of mammography raysMaladaptive coping modesNeed for an accompanying person1) Loosing time to find an accompanyInternalization the experiences of the others1) Internalization unpleasant experiences about mammography\ 2) Easy access to unaware and close person to ask questions rather than health care providersReligious faith1) Lack of belief in the possibility of being affected by breast cancer\ 2) Considering breast cancer as a gift of God or God\'s will\ 3) turning to spirituality, praying, and putting the problem in God\'s hand for cancer treatment\ 4) breast cancer: a type of punishment from the God because of errors in the pastFatalism1) chance: a determinant factor\ 2) Having the feeling of lack of control or powerlessness against deathAvoidance and denial\ 1) Unreal optimistism and given immunity as restrictive factors\ 2) Try not to think about breast cancer Psychological barriers {#s3-1} ---------------------- Embarrassment Some participants stated that they were ashamed of mammography screening. In this regard, the old girls with no partner had comments different from the married women. They reported to experience a higher level of embarrassment than the married women because of their narrow territory to seek mammography, late referral to a gynecologist, being ashamed of disease follow-up, misunderstanding of the others about the reasons forgoing to a gynecologist, and, finally, not being comfortable to announce the family about the gynecologist visit. For instance, some participants reflected these concerns as follows: "*In Yazd, when a girl goes to a gynecologist clinic, the others have a reproachful glareat her and think on what has happened that she visits a gynecologist. The family also has the same opinion*", (P8). Another participant told: "*It is mentally difficult for a girl to announce her family about visiting a gynecologist. One can easily speak about visiting a dentist, but the case for a gynecologist appointment is not so easy*" (P7). *Worry about being diagnosed with cancer* Worry about the mammography result is another barrier. In other words, worry about hearing bad news (being diagnosed with cancer) and have no preparedness to confront with breast cancer may reduce the women's motivation for MA. For example, one of the participants said: "*If now I undergo mammography, the result may be OK. But I am afraid that if six months later I did the procedure, the result would be bad. Then, how should I deal with the disease. We always are waiting to hear good news. Nevertheless, I expect to be affected by breast cancer, as the disease has been occurred in my family*" (P9). Preoccupation with underlying diseases Women announced that suffering from another physical health problem may be a competing priority with breast cancer preventive behavior that may lead to a negative attitude toward mammography. For instance, one of the participants with diabetes said: *"Since I have diabetes, I do not like to do anything with my physical health. I am mostly dealing with my blood sugar. Two years ago, I visited a doctor who prescribed me a medication that disturbed the level of my blood sugar. I am afraid that I visit a doctor for the breast issues, and then I had to take some medications which may increase the level of my blood sugar"* (P2). Misconceptions about mammography The most of the women participated in the study believed that mammography might be harmful for breast tissue. Some of them noted that breast cysts could be ruptured during mammography procedure. For example, one participant reflected her concern as it follows: "*Since I have many breast cysts and I have undergone a surgery once, I am worry about the stimulation and rupture of the cysts during mammography procedure. Won't they \[the cysts\] be pressured a lot while mammography procedure?*" (P6). Uncertainty about the harmfulness of the mammography rays was another concern mentioned by women: "*I don't know what kind of radiations mammography has? However, I have heard that it may be harmful for the body and should not be performed frequently*" (P13). Need for an accompanying person In the current study, the participants expressed that they need for a reliable and close person to accompany them while going to the mammography center. Because of this, they might lose the time or quite forget it. One of the participants said: "*I would like to be accompanied with a close person when going to mammography. When I find someone to be accompanied with, a long period of time has been passed*" (P5). *Internalizing the experiences of the others* According to the women's statements, they could be highly influenced by each other. In other words, when women talk about mammography, they may describe their unpleasant experiences about it and thus decide not to uptake the mammography test: "*I am a hairdresser. Once, one of my customers came to me just after undergoing mammography. She was very pale and had nausea. It seems that she was compressed. For that reason, I never want to undergo mammography*" (P6). Also in relation to breast cancer, the women may ask their questions from unaware and close persons rather than health care providers because of having easy access to them. One participant explained this matter as follows: "*Everyone asks an intimate friend; one who understands her and is available. It is not easy for me to visit a doctor and consult her*" (P8). Maladaptive coping modes {#s3-2} ------------------------ *Religious faith* Participants' religious beliefs have been found to be strong barriers for seeking mammography. the belief in the possibility of not being affected by breast cancer and considering it as a gift of God or God's will were other issues discussed by the participants. This consideration has been well expressed by one of the participants: "*It is natural for people to believe in not being affected by breast cancer. Maybe God has created us in this way. If it was in another way, perhaps we were tortured. God may have put it inside us*" (P10). Another woman said: "*It is God's will that people are being affected by breast cancer. God draws some of us for the disease. God is the one who draws*" (P7). Participants mentioned that the best way for dealing with cancer is turning to spirituality, praying, and delegating the problem to God. For instance, one of the participants commented as follows: "*If I were diagnosed with breast cancer, I would pray for God more and more as I believe that the doctors cannot do anything for me and only God can cure me*" (P10). Another woman stated that having breast cancer maybe a type of punishment from God because of mistakes people do in the past. She, also, noted that: "*Being affected by cancer is related to our deeds, and when someone has cancer, God examines her or him"* (P3). Fatalism Women participated in the current study mainly emphasized on the role of chance as an important factor which may limit the women's power to prevent breast cancer. One of the participants quoted: "*All these things occur by chance. I think that chance plays a very important role in every event*" (P3). Another participant expressed that the cause of death for each person is being determined, previously, and an individual is considered to have a limited control over his/her death: "*Everybody must die in a way; one in a car accident, another experiences stroke, and some others may be diagnosed with breast cancer. There must always be a cause for death*" (P4). Avoidance and denial Some of the participants declared that they never think about breast cancer. Issues related to this subcategory were optimistic bias and an unreal assumption of being safe which may be considered as restrictive factors for mammography uptake. For instance, a participant said: "*You know, I think I will never be affected by breast cancer, as in a proverb it is said 'human beings think that all other people will die except themselves'*" (P1). Another participant reflected the fact as follows: "*When I think about something, for example when I have flank pain and think about it, it seems that the pain aggravates. It is really in this way; thinking to pain worsens it. I try not to think about breast cancer and if it comes to my mind, I remove it*" (P11). Discussion {#s4} ========== The findings of the current study demonstrated a wide range of factors including psychological barriers and maladaptive coping modes that may affect the women's adherence to mammography. Some of the aspects are new and there is no evidence about in the literature. These new findings are valuable as they show the effect of cultural context on the breast health-seeking behaviors of the Iranian women. In the present study, according to the participant's statements, embarrassment associated with mammography was considered as a vital response cost. This is consistent with those found by Fleming et al.^[@R23]^ The new and unique aspect of this study was the comments of the old girls about the shyness of referring to a gynecologist clinic or seeking help for breast cancer prevention. To the best of our knowledge, no report is available in the literature in this regard. The interviewer (the first author) is from Yazd city, Iran and notes that the old girls especially in the older ages are not willing to present any abnormal sign of breast or genital organs because of their fear about the opinions of the people around. In some cases, they might lose some suitable opportunities for marriage in the future. Therefore, in such a cultural context, it is considered inappropriate for an old girl to go to a gynecologist; in the other hand, the family may become remarkably worried about. Unfortunately, the old girls in Yazd are dealing with this problem and there is no evidence to compare the aforementioned results and, so, further investigation is needed. Moreover, worry about being diagnosed with cancer was mentioned as another psychological barrier by the participants, which is in line with the findings reported by Whitaker et al.^[@R24]^ Similar with those reported by Lamyian et al,^[@R25]^ another psychological barrier found in the present study was preoccupation with underlying diseases which was considered as a part of the competing concerns. There were, also, some misconceptions about the mammography procedure emerged during the qualitative analysis in the present study. For example, there was a belief that mammography might have severe complications for the breast tissue and, so, some participants especially those who had the history of breast cysts avoided from mammography uptake. The participants declared that they were less likely to undergo mammography because they follow what they hear from the others. Therefore, the impact of subjective norms is recalled as an important factor in the process of personal decision-making.^[@R26]^ Also, satisfaction with medical care was found to be important especially among asymptomatic women who decide to uptake mammography.^[@R27]^ Another factor was being worry about x-rays as a major barrier for mammography, which is consistent with the results of Sadikoglu et al.^[@R28]^ According to the participants' comments, it was difficult for them to go to mammography clinics alone and they prefer to be accompanied by a close person like a sister or a close friend. It was concluded that embarrassment and fear of being diagnosed with cancer may affect the women's self-esteem in MA. This issue could certainly waste the golden time for the early detection of breast cancer. Therefore, proper communication among the health care providers and women referring to the mammography centers may enhance the mutual trust and peace of mind. Accordingly, Meguerditchian et al showed that communication skills could be principal predictors of MA.^[@R29]^ As all the study participants in the present study were Muslim, religious belief was another issue that was emphasized by the participants. This concept was titled as maladaptive coping modes and grouped into three broad sub-categories. In a previous study, Zarghami stated that religion and spirituality were as significant barriers for Iranian women to seek medical screening and care in early stages of breast cancer; nevertheless, the role of such religious factors on cancer screening behaviors among women is not well understood.^[@R30]^ Some of the participants in the present study believed that being affected by breast cancer might be a God's gift. The others expressed that the disease could be a probe or punishment from God, as they may already have committed a sin or guilt in their lifetime. These findings are similar to the results of a qualitative study by Azaiza and Cohen.^[@R31]^ Moreover, Hatefnia et al showed that religious beliefs have significant association with mammography uptake among Iranian women.^[@R32]^ The study participants believed that medical treatment for breast cancer is useless and God could cure them, only. Mitchell et al declared that it is likely for delay to be happened in presenting a self-discovered breast lump among women with strong religious beliefs; especially belief in "religious intervention in place of treatment." The researchers suggested the interaction between clinicians and clergies to correct wrong beliefs to enhance breast health-seeking behaviors.^[@R33]^ Fatalism was conceptualized as "passively denying personal control" and encompasses the dimensions such as personal perceived lack of control over external events, luck, predetermination of a disease, and belief of inevitable death by a serious disease.^[@R34]^ The participants in the present study claimed that human beings have no control on his/her death and cancer diagnosis will, certainly, result in death. Spurlock and Cullins pointed out that fatalism could negatively influence breast cancer screening behaviors.^[@R35]^ Also, the results of the study conducted by Donnelly et al showed that Arab women living in Qatar reported being diagnosed with cancer as a bad luck.^[@R36]^ It was concluded that women's perceptions on cancer fatalism is a cultural issue and should be further investigated especially in countries such as Iran that holds culturally diverse population groups. The most of the women in this study noted that they have never thought about the likelihood of being diagnosed with breast cancer and the role early detection in the disease survival. They always supposed to have immunity against the disease; thus, they feel no need to have a mammogram. This finding is consistent with those of reported by Hasson et al. who revealed that 16.5% of the participants reported no thinking about undergoing mammography.^[@R37]^ Conclusion {#s5} ========== The current results gave us a more detailed picture on how Iranian women perceive response costs of breast health-seeking behaviors. Although, women are responsible for their own health within the context of breast cancer prevention, governmental policy makers are, also, responsible for provision of women's health care facilities. In Iran, despite the high incidence of breast cancer, mammography screening is not an integrated routine service in the health care system. It was concluded that women in Yazd perceive some cultural-based barriers for MA. It is necessary to identify the priorities and the existing gaps in the country through accomplishing interdisciplinary scientific researches and specific interventions. Beside the development of women's health policy in Iran, it is suggested that religious and cultural belief systems be incorporated into the educational programs applying persuasion- based health monitoring approach. Health educators may have good opportunities to design theory-based educational interventions within which the belief systems are being incorporated with the hope to initiate adaptive coping reactions. Ethical approval {#s6} ================ Research protocol was approved by the Research Council and Ethics Committee affiliated by Yazd Shahid Sadoughi University of Medical Sciences, Yazd, Iran. Voluntary participation, anonymity, confidentiality, and withdrawing from the study at any time without any penalty were some issues that were explained for the participants. All the interviews were recorded with a sound recorder after obtaining permission and signing an informed written consent form by the participants. Competing interests {#s7} =================== The authors declare no conflict of interest. Authors' contributions {#s8} ====================== MK was the main investigator and con­tributed to development of research protocol, imple­mentation of the research, and drafted the manuscript. SSMM supervised the study and scientific in­tegrity of data collection and revision of the manuscript. ML was advisor to the qualitative study, par­ticipated in the design of the study and interpretation of data and revising the manuscript. MAM and Ht were advisor to the study, par­ticipated in the interpretation of data and revising the manu­script. All authors have read and approved the final manuscript. Acknowledgements {#s9} ================ This study was financially supported by Shahid Sadoughi University of Medical Sciences, Yazd-Iran. The authors are grateful to all women who participated in the study.

Background ========== Lung ischemia/reperfusion injury (LIRI) is a critical problem in research and in clinical practice \[[@b1-medscimonit-24-1251],[@b2-medscimonit-24-1251]\]. It is well known that LIRI can occur after lung transplant, cardiopulmonary bypass, pneumonectomy, and post-enucleation of pulmonary embolism, which can cause pulmonary dysfunction and damage \[[@b3-medscimonit-24-1251],[@b4-medscimonit-24-1251]\]. The mechanism and clinical courses of LIRI have been well investigated. Oxygen radicals, inflammatory cytokines, and leukocytes have recently been identified to play key roles \[[@b5-medscimonit-24-1251]\]. Since the mechanisms involved in LIRI are complicated and cross-linked, demonstrating an explicit mechanism could help to prevent the potentially severe complications \[[@b6-medscimonit-24-1251]\]. Toll-like receptors (TLRs), a family of pattern recognition receptors, provide a first line of defense against bacterial and viral pathogens. It is also reported that TLRs are initial sites for inflammatory signaling activation in the lung during ischemia and reperfusion \[[@b7-medscimonit-24-1251]--[@b10-medscimonit-24-1251]\]. TLR-4, originally regarded as responding only to specific bacterial ligands, is now proved to be activated by many signals induced by stressed, necrotic, or injured cells \[[@b11-medscimonit-24-1251],[@b12-medscimonit-24-1251]\]. TLR-4 has been reported to be a key modulator in a number of models of ischemia-reperfusion. More recently, TLR-4 was reported to be a key factor in the pathogenesis of LIRI, in which the activation of TLR4 of alveolar macrophages was thought to be the initial step \[[@b13-medscimonit-24-1251]\]. Activation of TLR4 results in the nuclear translocation of nuclear factor-kappa B (NF-κB), an essential regulator involved in the productions of several inflammatory cytokines, such as tumor necrosis factor-α (TNF-α), interleukin-6 (IL-6), and IL-1β \[[@b14-medscimonit-24-1251],[@b15-medscimonit-24-1251]\]. Erythropoietin (EPO) is a cytokine responsible for erythropoiesis. EPO has been proved to protect against inflammatory injuries \[[@b16-medscimonit-24-1251],[@b17-medscimonit-24-1251]\]. In addition, it is reported that EPO can reduce the production of inflammatory mediators \[[@b18-medscimonit-24-1251]\]. Recently, it was reported that RhEpo can reduce I/R-induced lung injuries through its anti-inflammatory effects, but the mechanism is still unclear \[[@b19-medscimonit-24-1251]\]. In the present study, we assessed the protective effects on LIRI and the associated mechanism in rat models of LIRI. Particularly, we explored the protective effects of EPO on the expressions of TLR4, NF-κB, and inflammatory cytokines, as well as changes in pulmonary function and structure. Material and Methods ==================== Animals and groups ------------------ All experiments were conducted following the regulations for experimental animal welfare and were approved by the Committee for Animal Experiments at West China Hospital, Sichuan University. Male Sprague-Dawley (SD) rats, weighing 200--250 g, were purchased from the Animal Center of Sichuan University and were kept at a constant temperature (23±1°C) with a 12 h: 12 h light/dark cycle. Recombined human erythropoietin (RhEpo) was purchased from Shenyang Sunshine Pharmaceutical Co. Ltd. (Shenyang, China). SD rats were divided randomly into 3 groups (n=8): a control group, a vehicle+LIRI group, and an EPO+LIRI group. In the control group, animals just received mechanical ventilation for 3.5 h, after which we performed a thoracotomy. In the vehicle group, 2 ml of normal saline was administered by intraperitoneal injection 2 h before the operation, then rats underwent the experimental protocol of 90-min left lung ischemia and 120-min reperfusion \[[@b20-medscimonit-24-1251]\]. In the EPO group, we administered RhEpo (3 kU/kg \[[@b19-medscimonit-24-1251]\] diluted in 1 ml saline solution) by intraperitoneal injection 2 h before the operation, and the rest of the protocol was the same. Collections of blood, bronchoalveolar lavage fluid (BALF), and tissue --------------------------------------------------------------------- Blood was collected from all rats and the BALF was then collected by intratracheally washing the right lungs with a total of 5 ml phosphate buffer solution (PBS). BALF was then centrifuged at 5000 rpm for 10 min. The precipitate and supernatant were collected for the next studies. Lung tissue Wet/Dry (W/D) ratio ------------------------------- The rest of lung tissue was weighed immediately and recorded as the wet weight. The tissue was then dried for 48 h in a 70°C dry-box and weighed as the dry weight. W/D ratio was calculated as an indicator for pulmonary edema. BALF leukocyte, neutropil count and percentage, and PPI ------------------------------------------------------- The BALF precipitate was studied for the number of leukocytes and neutrophil percentage. Protein analysis was performed in the BALF supernatant and serum. PPI was calculated as the ratio of BALF protein level to serum protein level. Oxygenation Index ----------------- PaO~2~ was measured by arterial blood gas analysis, and the ratio of PaO~2~ to FiO~2~ was then calculated as the oxygenation index. Hematoxylin & eosin (HE) staining --------------------------------- Lung tissues were fixed in 4% paraformaldehyde and embedded in paraffin. We made 5-μm sections and stained them with HE using a standard protocol. The slides were examined by a pulmonary pathologist who was blind to animal grouping. Each slide was given a score based on the 3 hallmarks (neutrophils, alveolar edema, and interstitial infiltrate), and the total score was calculated (ranging from 0 to 9, with 0 representing normal lung and 9 the most injured lung) \[[@b21-medscimonit-24-1251]\]. Serum cytokines levels ---------------------- The serum was obtained from blood samples by centrifuged at 5000 rpm for 10 min and analyzed by specific ELISA kits (R&D Systems, Minneapolis, MN) to determine the levels of cytokines TNF-α, IL-6, and IL-1β. mRNA levels of TLR4 and NF-κB ----------------------------- RNA was obtained by using Trizol solution (Invitrogen, USA) and reversed with the specific kit (Invitrogen, USA). Rt-PCR was performed as follows: 95°C for 90 s, followed by 35 cycles (95°C for 5 s and 59°C for 30 s). Primer for TLR4: 1. forward: 5 -TTATCCAGAGCCGTTGGTGT-3, 2. reverse: 5-CCCACTCGAGGTAGGTGTTT-3. 3. Primer for NF-κB: 4. forward: 5 -TTCCTGCTTACGGTGGGATT-3, 5. reverse: 5 -CCCCACAT CCTCTTCCTTGT-3. Primer for GAPDH: forward: 1. 5 -GAGACAGCCGCATCTTCTTG-3, 2. reverse: 5 -TGACTGTGCCGTTGAACTTG-3. Each sample was tested 3 times. Western blotting analysis ------------------------- Protein was extracted from lung tissues with RIPA buffer. We added 50 μg of protein into SDS-PAGE. Protein was transferred to the polyvinylidene fluoride membrane and then blocked with 5% pure albumin. The membrane was incubated with TLR4 and NF-κB p65 primary antibody (Abcam, USA) overnight at 4°C. Anti-GAPDH antibody (Cell Signaling Technology, USA) was used as the loading control. Statistical analysis -------------------- All descriptive data are expressed as mean ±SD. Statistical data were analyzed by one-way ANOVA followed by Bonferroni correction for multiplicity. All analyses were conducted using SPSS 20.0 (Chicago, USA). *P*\<0.05 was regarded as a significant difference. Results ======= Effects of EPO on pulmonary damage ---------------------------------- As shown in [Figure 1](#f1-medscimonit-24-1251){ref-type="fig"}, the W/D ratio was significantly higher in the vehicle group than in the control group (P\<0.05). BALF leukocyte count and neutrophil ratio were also increased in the vehicle group, and PPI in the vehicle group was remarkably elevated compared with the control group. Compared to the vehicle group, those changes in the EPO group were remarkably decreased (P\<0.05, respectively). The oxygenation index was much higher in the control group than in the vehicle group (P\<0.05). Interestingly, the oxygenation index in the EPO group was better than in the vehicle group (P\<0.05). Effects of EPO on lung pathology -------------------------------- To study the effects of EPO on LIRI, we analyzed pathological changes ([Figure 2](#f2-medscimonit-24-1251){ref-type="fig"}). In the control group ([Figure 2A](#f2-medscimonit-24-1251){ref-type="fig"}), the pulmonary alveoli and interstitium was intact and no signs of pulmonary edema and structural damages were observed. In the vehicle group ([Figure 2B](#f2-medscimonit-24-1251){ref-type="fig"}), significant pulmonary edema and structural damages were identified. Excessive inflammatory cells were infiltrated into the lung tissue and increased the thickness of the interstitium. These pathological changes in the EPO group ([Figure 2C](#f2-medscimonit-24-1251){ref-type="fig"}) were improved. Pathological scores in the vehicle group were remarkably higher than in the control group (P\<0.05), but were significantly (P\<0.05) reduced in the EPO group ([Figure 2D](#f2-medscimonit-24-1251){ref-type="fig"}). The Effects of EPO on Levels of TNF-α, IL-6, and IL-1β ------------------------------------------------------ As shown in [Figure 3](#f3-medscimonit-24-1251){ref-type="fig"}, serum TNF-α, t IL-6, and IL-1β levels in the vehicle group were remarkably elevated compared with those in the control group (P\<0.05), while a remarkable decrease was found with EPO treatment (P\<0.05). The Effects of EPO on Expressions of TLR4 and NF-κB --------------------------------------------------- [Figures 4](#f4-medscimonit-24-1251){ref-type="fig"} and [5](#f5-medscimonit-24-1251){ref-type="fig"} show mRNA and protein expressions of TLR4 and NF-κB. Compared to the control group, mRNA and protein expressions of TLR4 and NF-κB in the vehicle group were significantly increased (P\<0.05), and those in the EPO group were significantly decreased compared with the vehicle-treated LIRI group (P\<0.05). Discussion ========== The present study is the first to demonstrate that EPO can attenuate the expressions of TLR4 and NF-κB in lung tissues in an LIRI rat model. Additionally, we found that EPO inhibited the releasing of serum TNF-α, IL-1β, and IL-6 in the LIRI rat model. These results show that EPO exerts its protective effects on LIRI through inhibiting the TLR4/NF-κB signaling pathway. The TLRs family is essential in innate immunity and many inflammatory responses \[[@b22-medscimonit-24-1251]\]. In all TLRs, TLR4 has been extensively studied and is considered to play important roles in mediating inflammatory responses. Activation of TLR4-mediated inflammatory signaling pathways resulted in the translocation of NF-κB, subsequently leading to production of many pro-inflammatory mediators and induction of inflammatory reactions \[[@b23-medscimonit-24-1251]\]. It is reported that immune and inflammatory responses play essential roles in LIRI \[[@b24-medscimonit-24-1251]\]. Accumulating evidence also shows that the TLR4-NF-κB signaling pathway induces inflammatory reactions and consequently intensifies LIRI \[[@b25-medscimonit-24-1251]\]. Moreover, recent reports show that TLR-4 is critical in the development of LIRI, and its activation in the alveolar macrophage may be the first step \[[@b13-medscimonit-24-1251]\]. In the present study, we found that the expressions of TLR4, NF-κB, and inflammatory mediators were elevated at the end of the experiment, demonstrating that TLR4 initiated the releasing of pro-inflammatory cytokines by activating the NF-κB signaling pathway in LIRI. Thus, targeting the modulation of TLR4 and NF-κB expression might be a promising therapeutic strategy for inhibiting inflammatory cascades in LIRI. The pharmacological roles of EPO have been largely extended from a traditionally hematopoietic hormone to a multi-function cytokine. The beneficial effects of EPO on IR-caused injuries in the heart \[[@b26-medscimonit-24-1251]\], kidney \[[@b27-medscimonit-24-1251]\], and brain \[[@b27-medscimonit-24-1251],[@b28-medscimonit-24-1251]\] have been revealed. However, only a few of reports showed that pretreatment with EPO attenuates LIRI \[[@b19-medscimonit-24-1251],[@b29-medscimonit-24-1251]\]. In this study, we investigated the damaged lung structure in a rat IRI model, showing the damaged alveoli wall, dilated capillaries, and infiltrated inflammatory cells in the interstitium by HE staining. Treatment with EPO resulted in less structural damage and infiltration of neutrophils, lower PPI, and increased oxygenation index, indicating the protective effect of EPO in LIRI. It was reported that EPO regulates TLR4-mediated inflammatory responses and NF-κB and PI3K signaling pathways to protect against hypoxic injury \[[@b30-medscimonit-24-1251],[@b31-medscimonit-24-1251]\]. Through activating these critical signaling components, EPO can modulate many downstream inflammatory cytokines, such as TNF-α, IL-1β, and IL-6 \[[@b31-medscimonit-24-1251]\], regulating oxidative stress, apoptosis, inflammatory damage, and regeneration. Our results demonstrate that the structural and functional presentations of lung injury were relieved and the mRNA and protein levels of TLR4 and NF-κB were significantly decreased by treatment with EPO. In addition, after the administration of EPO, the levels of TNF-α, IL-1β, and IL-6 in serum were significantly reduced, revealing that EPO inhibited inflammatory damage through down-regulating the TLR4/NF-κB signaling pathway, thereby protecting against the effects of LIRI. Conclusions =========== In summary, our results indicate that EPO exerts protective effects on lung injuries after LIRI through inhibiting the TLR4-NF-κB signaling pathway and decreasing production of inflammatory cytokines (TNF-α, IL-6, and IL-1β). The present study provides solid support for the use of EPO as a therapeutic agent in treating LIRI. **Source of support:** Departmental sources {#f1-medscimonit-24-1251} {#f2-medscimonit-24-1251} {#f3-medscimonit-24-1251} {#f4-medscimonit-24-1251} {#f5-medscimonit-24-1251} [^1]: Study Design [^2]: Data Collection [^3]: Statistical Analysis [^4]: Data Interpretation [^5]: Manuscript Preparation [^6]: Literature Search [^7]: Funds Collection

Background {#Sec1} ========== Arterial spin labeling of the heart has been shown to estimate myocardial perfusion and perfusion reserve at a single short-axis slice for coronary artery disease assessment. However, current spatial labeling methods suffer from transit delay effects when imaging is extended to more than a single slice. Velocity selective (VS) labeling is a promising alternative that does not suffer from transit delay effects. Methods {#Sec2} ======= Eight healthy volunteers were scanned using a 3T GE Signa Excite HD scanner with an 8-channel cardiac coil. Myocardial ASL measurements were made at a single short axis slice using both VSASL and conventional flow alternating inversion recovery (FAIR) ASL as a reference. VS labeling was performed using an adiabatic BIR4 pulse with bipolar gradients as shown in Figure [1A](#Fig1){ref-type="fig"}. Labeling was achieved by saturating all spins above a cutoff velocity of 10 cm/s to target coronary blood velocity. Triple inversion recovery was used to suppress a range of myocardial T1s between 1250 ms and 1450 ms for background suppression. 6 breath-held labeled/control image pairs were acquired per subject. Myocardial blood flow (MBF), physiological noise (PN), and temporal SNR (TSNR = MBF/PN) were measured within the left ventricular myocardium ROI.Figure 1**A. the Velocity selective pulse is made using a symmetric BIR4 for reduced eddy current sensitivity and bipolar gradients to prevent spatial signal modulation in static tissue**. **B**. Coronary arterial blood (red) is saturated by VS labeling at mid-diastole with a velocity cutoff of 10 cm/s. **Left:** Gz \"off\" (T~2~ weighting only) **Right:** Gz \"on\" (T~2~ weighting + velocity selective saturation). Results {#Sec3} ======= Figure [1B](#Fig1){ref-type="fig"} shows that the VS labeling pulse was successful in saturating blood within the coronary artery. MBF and PN measurements from VSASL and FAIR were 1.69 ± 0.84 ml/g/min and 2.22 ± 0.57 ml/g/min respectively in eight volunteers. This corresponds to a TSNR of 2.01 and 3.89 respectively. VSASL underestimated MBF by 23.8% when compared to FAIR, which may be due to a signal loss of 15% from inversion inefficiency in background suppression combined with a T2 signal loss of 7% from the 12 ms VS pulse. We suspect that higher PN in VSASL is from spurious labeling of myocardium, which can be further reduced by more consistent background suppression. Low TSNR will be addressed by further sequence improvements that explore different cutoff velocities and velocity labeling directions. In a single volunteer, we performed an additional FAIR experiment with a thicker inversion slab (FAIR-TS) to simulate whole heart coverage. MBF and PN measurements from VSASL, FAIR, and FAIR-TS were 1.38 ± 0.58 ml/g/min, 0.88 ± 0.37 ml/g/min, and -0.04 ± 0.24 ml/g/min respectively. FAIR-TS has a large transit delay and is unable to estimate MBF while VSASL does not suffer from transit delay effects. Conclusions {#Sec4} =========== We have successfully labeled coronary blood based on velocity and demonstrated that VSASL is sensitive to myocardial perfusion. We believe that VSASL has the potential to become a more sensitive labeling scheme than spatial labeling sequences because of its insensitivity to transit delay and because it is inherently compatible with whole heart coverage.Figure 2**MBF and PN measurements from VSASL, FAIR, and FAIR-TS with a thicker inversion slab to simulate whole heart coverage in a single volunteer**. Both VSASL and FAIR are able to measure MBF. The region of high MBF on the lateral wall in VSASL is possibly due to spurious myocardial labeling. While VSASL has higher PN than FAIR, it is compatible with whole heart coverage because it does not suffer from transit delay effects. FAIR-TS is unable to measure MBF because the thickened inversion slab imparts a large transit delay.

Introduction {#Sec1} ============ Nanoparticle-based systems containing more than one functional components represent an active research field having a great potential in numerous technological applications^[@CR1]^. Among others, magnetic core-shell nanoparticles offer new opportunities in the biomedical field, catalysis and sensing^[@CR2]--[@CR10]^. In particular, fluorescent-functionalized silica-coated core-shell magnetic nanoparticles attract high attention in imaging and sensing applications. In such multifunctional nanomaterials the fluorescent dye can be covalently anchored either onto the silica surface or doped into the matrix of the silica shell^[@CR2],[@CR3],[@CR11]--[@CR13]^. Electrospinning has been one of the most versatile methods employed for generating nano- and microfibers^[@CR14]--[@CR16]^. Its simplicity, scalability and high versatility renders this method very attractive in many scientific fields. Electrospun polymer-based organic-inorganic fibrous nanocomposites have been developed by many research groups and further evaluated in various fields including biomedicine^[@CR17]--[@CR19]^, catalysis^[@CR20]--[@CR22]^, sensing^[@CR23]^, energy^[@CR24],[@CR25]^ and environmental protection^[@CR26]--[@CR29]^. However, only a few examples appear to date on the fabrication of nanocomposite electrospun fibers with embedded core-shell ferrous nanoparticles^[@CR30],[@CR31]^. In one such example, core-shell Fe/FeO nanoparticles have been incorporated within polyimide fibers aiming to produce fibrous nanocomposites exhibiting enhanced thermal stability and magnetic properties for potential use in high-temperature magnetic sensing and microwave absorption applications^[@CR30]^. The general advantages of combining superparamagnetic Fe~3~O~4~ nanoparticles with fibrous materials designed for use in sensing applications include among others the inherent ability of magnetic Fe~3~O~4~ nanoparticles to act as effective gas sensors^[@CR32]^ and the possibility for employing fiber alignment *via* magnetic field-assisted electrospinning^[@CR33]^. Concerning the latter, it has been demonstrated that fiber alignment results in the enhancement of the sensing properties compared to their randomly oriented analogues^[@CR34],[@CR35]^. Furthermore, the magnetic properties of the produced electrospun mats could provide an additional functionality in their manipulation and processing. More precisely, it may promote the magnetic bonding of the fibrous mats on suitable electromagnetic holding platforms, the controllable remote heating of materials for specific sensing applications or the efficient collection of mats by magnetic means from remote areas in various sensing applications^[@CR36]^. Additionally, magnetic electrospun mats with sensing capabilities could be employed as overlayers on various optical platforms in integrated optics or optical fibers. Such a thin magnetic overlayer on the proximity of an optical waveguide could operate as a sensing element of magnetic measurands that could alter the surrounding refractive index leading to pure photonic interrogation, with the simultaneous monitoring of pH conditions. Other applications may also include the measurement of magnetic fields by the induced deflection of fibers or waveguide cantilevers in the presence of magnetic fields, retaining at the same time the chemical sensing capability. Furthermore, in the presence of electric high frequency fields the resulted remote heating of overplayed mats could alter the refractive index properties of optical waveguides allowing thus the optical sensing of fields or temperature in magnetic hyperthermia applications^[@CR36]^ and in cases where pH monitoring would also be necessary. Herein, ferrous core-shell nanoparticles consisting of a magnetic γ-Fe~2~O~3~ multi-nanoparticle core and an outer SiO~2~ shell have been synthesized and further functionalized with Rhodamine B (RhB) fluorescent molecules (γ-Fe~2~O~3~/SiO~2~/RhB NPs). The latter were covalently bound in the matrix of the silica shell. The resulting γ-Fe~2~O~3~/SiO~2~/RhB NPs were further incorporated within cellulose acetate (CA) electrospun fibers to yield fluorescent multifunctional fibrous nanocomposites. Electrospun fibers with embedded fluorescence moieties designed for use in fluorescence sensing are considered to be highly advantageous compared to their film analogues due to their larger surface-to-volume ratios. In previous reports on fluorescent-functionalized electrospun polymer fibers, the fluorophores were either covalently attached onto the polymer backbone^[@CR37]--[@CR41]^ or incorporated as dopants within the fibers^[@CR42]--[@CR46]^. In one such example referring to the doping of polymer nanofibers with RhB, the fluorescent dye was added into a poly(ether sulfone) solution prepared in N,N-dimethylacetamide and the mixture was electrospun to obtain fluorescent, RhB-doped nanofibers that were further evaluated as metal ion (Cu^2+^) fluorescent sensors in aqueous media^[@CR47]^. Furthermore, electrospun polymer fibers doped with RhB derivatives have been successfully used as highly efficient turn-on fluorescent sensors for the detection of Hg^2+^ ions^[@CR48],[@CR49]^. In the present study, the use of γ-Fe~2~O~3~/SiO~2~/RhB NPs as dopants in electrospun fibers is considered to be advantageous compared to other fabrication routes reported so far, since the leakage of the RhB fluorophore from the core-shell NPs is prevented due to its covalent anchoring onto the nanoparticle surfaces. In contrast, small fluorescent molecules introduced within the polymer fibers as dopants are only held onto the polymer chains *via* weak van der Waals interactions, thus often resulting to their desorption from the polymer matrix and consequently the decrease in the fluorescence efficiency of the fibers. In addition, the covalent anchoring of RhB molecules onto the nanoparticles' surfaces and the blending of the γ-Fe~2~O~3~/SiO~2~/RhB NPs with the fibrous CA matrix, suppress self‐quenching phenomena, whereas by covalently integrating RhB within the silica shell, fluorescence quenching is further prevented by avoiding direct contact with iron oxides. Besides the above, the use of a renewable, naturally-abundant acetylated cellulose derivative as a polymer matrix exhibiting biocompatibility, biodegradability and environmental friendliness, combined with the magnetic character of the inorganic γ-Fe~2~O~3~/SiO~2~/RhB NPs additives providing the possibility for magnetic separation by applying an external magnetic field, are additional benefits of the multifunctional fibrous nanocomposite fluorescent sensor described in this study. Two different fabrication protocols were followed for the preparation of CA fibers doped with the γ-Fe~2~O~3~/SiO~2~/RhB NPs. In the first synthetic route the nanoparticles were sprayed on top of the fibrous mat while in the second approach a nanoparticle suspension in the CA polymer solution was directly electrospun to yield the final product. The dual sensing capability (i.e. ammonia gas and pH sensing) of the developed fibrous nanocomposites was investigated for both types of the produced nanocomposites. In both cases, a good response for gas ammonia sensing was shown for high NH~3~ concentrations. The presented multifunctional electrospun ammonia sensors show high tolerance in poisoning and saturation effects, providing linear response up to 12000 ppm. Consequently, they could be extremely valuable in the detection of high levels of ammonia (concentrations above 1000 ppm) at industrial facilities employing ammonia transfer lines and dense storage spaces. Moreover, low cost sensing elements or materials can then be replaced after logging an ammonia leakage event. In the case of pH sensing, the fibrous nanocomposites in which the NPs were embedded inside the fibers were more robust compared to the fibers having the NPs deposited *via* spraying onto their surfaces, with the former demonstrating a fast and linear response in both alkaline and acidic environments with good reversibility. Experimental {#Sec2} ============ Materials {#Sec3} --------- For the synthesis of the γ-Fe~2~O~3~/SiO~2~/RhB NPs, reagent grade chemicals were used as received from the manufacturers. Iron (III) sulfate hydrate, iron (II) sulfate heptahydrate (ACS, 99%), citric acid (CA, 99%), tetraethoxysilane (TEOS, 99.9%) and NH~4~OH (28--30%) were supplied by Alfa Aesar (Lancashire, UK). Acetone (AppliChem GmbH) and absolute ethanol (Carlo Erba, reagent - USP) were used as received. (3**-**Aminopropyl)triethoxysilane (APS; silane-NH~2~, 99%), tetraethoxysilane (TEOS; 98%), ethyl acetate (EA), dichloromethane (DCM), dimethylformamide (DMF), Rhodamine B isothiocyanate (RhB), poly(acrylic acid) solution (25 wt. % in water) and polyvinyl pyrrolidone (PVP, 40 kDa) were obtained from Sigma-Aldrich (St. Louis, MO, USA). Cellulose acetate (CA, Mn = 30000 g/mol) was obtained from Sigma-Aldrich and used without further purification. Acetone (Technical Grade 99.5% - Panaska Trading CO) was the solvent used in the preparation of CA polymer solutions that were further electrospun. Synthesis of γ-Fe~2~O~3~/SiO~2~ NPs {#Sec4} ----------------------------------- The core-shell γ-Fe~2~O~3~/SiO~2~ NPs were obtained by the self-assembly of primary superparamagnetic maghemite (γ-Fe~2~O~3~) nanoparticles (size ∼ 10 nm) followed by coating of the superparamagnetic nanoparticle clusters with a layer of silica^[@CR50]--[@CR52]^. In brief, maghemite nanoparticles were synthesized by co-precipitation of Fe^3+^/Fe^2+^ from an aqueous solution and transfer to ethyl acetate^[@CR53]^. The nanoparticle clusters were prepared by emulsification of the ethyl acetate suspension (∼5 vol% of the continuous phase) as an inner phase with water as a continuous phase. The poly(acrylic acid) was mainly used as a capping agent/surfactant for nanoparticle stabilization. The residual ethyl acetate was removed using a rotary evaporator. Subsequently, the particles were washed with ethanol and distilled water. As a result, nanoparticle clusters containing ∼80--100 maghemite nanoparticles were formed^[@CR54],[@CR55]^. All procedures were developed exclusively by the company Nanos SCI and they are described in detail in previous publications^[@CR50]--[@CR55]^. Synthesis of γ-Fe~2~O~3~/SiO~2~/RhB NPs {#Sec5} --------------------------------------- Rhodamine-B (RhB) was covalently integrated into the matrix of the silica shell as described below^[@CR4]^. The reaction between RhB and APS was carried out first in the mixture of DCM/DMF = 4/1 overnight at room temperature. RhB (0.00933 mmol) was dissolved in the solvent mixture (0.5 mL) and then APS (0.186 mmol) was added. Subsequently, the volatile solvent was removed using nitrogen flow and the product (RhB-APS) was mixed with TEOS for the silica coating. The general synthetic route is described as follows: First, 100 mg of PVP was dissolved in 200 mL of ethanol containing 450 mg of nanoparticle clusters. Subsequently, 1.2 mL of TEOS and RhB-APS was mixed and then added to the suspension followed by the addition of 2.0 mL of aqueous ammonia solution. The reaction mixture was stirred using a 2-cm-wide glass propeller at 300 rpm for six hours at room temperature and upon completion, the product was washed first with ethanol and then 3-times with distilled water. Finally, an additional thin layer (a few nanometers thick) of non-fluorescent silica was deposited using the same synthetic protocol with the only difference being the amount of TEOS, i.e. 0.15 mL of TEOS in the absence of RhB-APS. The prepared γ-Fe~2~O~3~/SiO~2~/RhB NPs were dispersed in ethanol-based stable colloidal suspensions at a concentration of 26 mg/mL. Fabrication of γ-Fe~2~O~3~/SiO~2~/RhB NPs-functionalized electrospun CA fibers {#Sec6} ------------------------------------------------------------------------------ Initially, CA (2.5 g) was dissolved in acetone (20 mL) upon stirring at ambient conditions and the obtained colourless transparent homogeneous solution was loaded in the syringe of the electrospinning set-up. All electrospinning experiments were performed at room temperature. Equipment included a controlled-flow, four-channel volumetric microdialysis pump (KD Scientific, Model: 789252), a syringe with a connected spinneret needle electrode, a high-voltage power source (10--50 kV) and a custom-designed grounded target collector inside an interlocked Faraday enclosure safety cabinet. Systematic parametric studies were carried out by varying the applied voltage, the needle-to-collector distance, the needle diameter and the flow rate so as to determine the optimum experimental conditions for obtaining CA nanoparticle-free fibers. The electrospinning conditions used for obtaining continuous, nanoparticle-free fibers were the following: Flow rate: 5.9 mL/hr; applied voltage: 15 kV; needle-to-collector distance: 10 cm; Needle diameter: 16G. Fabrication of CA/γ-Fe~2~O~3~/SiO~2~/RhB NPs nanocomposite fibers was carried out by following 2 different experimental protocols. In the first one, a CA fibrous mat (19.5 mg) was placed inside a petri dish and the nanoparticle suspension (1 mL, nanoparticle suspension in ethanol, concentration: 26 mg/mL) was sprayed on top of the fibrous mat. Afterwards, the fibers were placed in a laboratory fume hood until complete drying. The 2^nd^ experimental protocol involved the dropwise addition of 1 mL of the aforementioned nanoparticle suspension in the CA polymer solution prepared in acetone (polymer solution concentration: 12.5% w/v) during stirring at ambient temperature and subsequent electrospinning of the resulting suspension under identical electrospinning conditions applied in the case of the CA fibers, to obtain CA/γ-Fe~2~O~3~/SiO~2~/RhB NPs nanocomposite fibers. Materials characterization {#Sec7} -------------------------- For the TEM investigations, the γ-Fe~2~O~3~/SiO~2~/RhB NPs were deposited by drying dispersion on a copper-grid supported, perforated, transparent carbon foil and analysed by TEM (Jeol, JEM, 2100) which operated at 200 kV. The magnetic properties of the samples were measured at room temperature by Vibrational Sample Magnetometry (VSM) (Lake Shore 7307 VSM). The dry, γ-Fe~2~O~3~/SiO~2~/RhB NPs-decorated CA fibers (10--15 mg) were placed into the VSM system prior to measurements. Magnetic characterization was carried out at room temperature. The hydrodynamic size distributions of the core-shell nanoparticles were measured using dynamic light scattering (DLS, Fritsch, Analysette 12 DynaSyzer, Germany). The morphological characteristics of the produced fibers prepared in the absence and presence of the γ-Fe~2~O~3~/SiO~2~/RhB NPs were determined by scanning electron microscopy (SEM) (Vega TS5136LS Tescan). The samples were gold-sputtered (sputtering system K575X Turbo Sputter Coater -- Emitech) prior to SEM inspection. Fluorescence microscopy was further used for visualizing the pristine and γ-Fe~2~O~3~/SiO~2~/RhB NPs-modified CA electrospun fibers. The samples were placed on glass slides, covered with coverslips and documented under the Olympus fluorescence microscope (U-RLF-T model). The fluorescence intensity of the pure CA fibers was determined by using the FITC filter (Excitation: 490 nm, Emission: 520 nm) whereas for the γ-Fe~2~O~3~/SiO~2~/RhB NPs-modified fibers a CY3 filter was used (Excitation: 552 nm, Emission: 570 nm). Images were analyzed using the cellSens software. Ammonia and pH sensing apparatus {#Sec8} -------------------------------- In order to measure the samples' response to ammonia vapors and different pH values, the γ-Fe~2~O~3~/SiO~2~/RhB NPs-functionalized CA electrospun fibers were placed in a cuvette holder having two perpendicular light paths that were specially designed for free space applications. An all solid state 532 nm laser was used to excite the RhB moiety using a 400 μm core multimode silica optical fiber. The emitted fluorescence was filtered by a Thorlabs FGL550 longpass filter with 550 nm cutoff wavelength that blocks the excitation wavelength. It was then collected by a 600 μm core multimode optical fiber and analyzed by a Thorlabs CCS200 spectrometer. Both SMA terminated optical fibers are connected with the cuvette with SMA fiber adapters having mounted collimators. A schematic representation of the setup is shown in Fig. [1A](#Fig1){ref-type="fig"}.Figure 1Schematic presentation of the experimental setup employed in ammonia gas sensing and pH sensing experiments (**A**) and of the ammonia gas sensing set-up (**B**). For the gas ammonia detection, the above described cuvette holder containing the sample was placed in a custom made sealed testing chamber of 4.3 L volume. A peltier element was used for the evaporation of the 25% w/v ammonia solution drops as schematically depicted in Fig. [1B](#Fig1){ref-type="fig"}. For the pH detection, aquatic solutions of different pH values were used. HCl solutions with pH values ranging from 1--5 and NaOH solutions ranging from 8--13 were inserted in the cuvette where the sample was placed, in order to evaluate the sample's response. Results and Discussion {#Sec9} ====================== Synthesis and characterization of γ-Fe~2~O~3~/SiO~2~/RhB NPs {#Sec10} ------------------------------------------------------------ Fluorescent silica-coated nanocrystal clusters with approx. 20 nm thick silica shell were prepared as fluorescent and magnetic labels for the produced CA electrospun fibers (γ-Fe~2~O~3~/SiO~2~/RhB NPs). The magnetic cores of the γ-Fe~2~O~3~/SiO~2~/RhB NPs were prepared by the self-assembly of approximately a hundred of superparamagnetic maghemite nanocrystals (size ∼10 nm). Darker magnetic nanocrystal cores can be clearly distinguished from the brighter amorphous silica shell in the TEM images (Fig. [2A,B](#Fig2){ref-type="fig"}). The size of the γ-Fe~2~O~3~/SiO~2~/RhB NPs determined from the TEM images (\>100 particles counted) was found to be ∼130 nm ± 30 nm. The γ-Fe~2~O~3~/SiO~2~/RhB NPs showed superparamagnetic properties with a saturation magnetization Ms of ∼37 Am^2^ kg^−1^ (Fig. [2C](#Fig2){ref-type="fig"}). DLS measurements of the γ-Fe~2~O~3~/SiO~2~/RhB NPs in an ethanol suspension (1.0 mg mL^−1^) showed narrow hydrodynamic-size distribution with the average size at ∼151 nm (SD = 2.8%) (Fig. [2D](#Fig2){ref-type="fig"}).Figure 2TEM images of γ-Fe~2~O~3~/SiO~2~/RhB NPs at low (**A**) and high (**B**) magnification, room-temperature measurement of the magnetization as a function of magnetic field strength (**C**), and intensity-weighted distribution of the γ-Fe~2~O~3~/SiO~2~/RhB NPs hydrodynamic diameters obtained from the DLS measurements in ethanol-based suspension at concentration 1 mgmL^−1^ (**D**). Emission spectra of γ-Fe~2~O~3~/SiO~2~/RhB NPs with high (9.33 µmol of RhB) and low (3.11 µmol of RhB) RhB loading (**E**). Emission spectra of the γ-Fe~2~O~3~/SiO~2~/RhB NPs and of the aqueous solution containing free RhB molecules (**F**). The γ-Fe~2~O~3~/SiO~2~/RhB NPs form stable colloidal suspensions and this is verified by the fact that their DLS-determined size is in close agreement with the size determined by TEM. In order to investigate the influence of the RhB-loading on the nanoparticles' fluorescence properties, γ-Fe~2~O~3~/SiO~2~/RhB NPs having a lower RhB amount (i.e. only 3.11 µmol of RhB instead of 9.33 µmol per 450 mg of γ-Fe~2~O~3~/SiO~2~/RhB NPs -- see in section 2.3) were synthesized. As seen in Fig. [2E](#Fig2){ref-type="fig"}, the emission spectra of the γ-Fe~2~O~3~/SiO~2~/RhB NPs with high RhB loading showed higher relative intensity compared to the low RhB loading clusters as expected. Most importantly, no shift in the maximum emission wavelength was observed upon altering the RhB loading. In order to examine whether the optical response of RhB is influenced by the covalent linkage to the silica shell, the maximum emission wavelengths of the γ-Fe~2~O~3~/SiO~2~/RhB NPs and of the RhB aqueous solution (free RhB) were compared. As seen in Fig. [2F](#Fig2){ref-type="fig"}, the maximum emission wavelength of RhB aqueous solution was recorded at 585 nm, while after covalent anchoring with silica, the wavelength was shifted to 580 nm. It is noteworthy to mention at this point that when RhB was simply added to the TEOS (silica precursor) without the pre-formation of a covalent bond with the amino-silane molecule (APS), the product (silica-coated nanoparticle clusters) was not fluorescently labelled and all RhB molecules were removed during vigorous washing. Fabrication of γ-Fe~2~O~3~/SiO~2~/RhB NPs-functionalized electrospun CA fibers {#Sec11} ------------------------------------------------------------------------------ Electrospinning was first employed in the fabrication of pure CA fibers. A schematic of the electrospinning set-up used is provided in Fig. [3](#Fig3){ref-type="fig"}. The experimental conditions employed for obtaining nanoparticle-free CA electrospun fibers are given in the experimental section.Figure 3Left: Schematic of the electrospinning set-up used in the fabrication of electrospun CA fibers. Right: Indicative photographs of the γ-Fe~2~O~3~/SiO~2~/RhB NPs-functionalized CA fibers obtained *via* spray deposition of the γ-Fe~2~O~3~/SiO~2~/RhB NPs onto the fibers' surfaces -- sprayed magnetic fibers -- (up) and by mixing of the γ-Fe~2~O~3~/SiO~2~/RhB NPs dispersion with the polymer solution prepared in acetone, followed by electrospinning -- electrospun magnetic fibers -- (down). The morphological characteristics of the produced CA fibers were investigated by SEM. As seen in the SEM images appearing in Fig. [4](#Fig4){ref-type="fig"}. CA fibers had a belt-like (ribbon-like) morphology, a random orientation and they were characterized by a relatively broad diameter distribution within the micrometer size range. Ribbon-like morphologies in electrospun fibers based on cellulose and cellulose derivatives have been previously reported^[@CR56]^. Based on earlier reports, the CA fiber morphology can be altered upon changing the solvent system and the polymer solution concentration. Under certain experimental conditions (i.e. specific solvent system, polymer solution concentration and optimum electrospinning parameters) the generation of cylindrical CA fibers is also feasible^[@CR57]^.Figure 4SEM images of the as prepared CA fibers (**A**), the sprayed magnetic fibers (**B**) and the electrospun magnetic fibers (**C**). As already mentioned in the experimental section, the fabrication of CA electrospun fibers decorated with γ-Fe~2~O~3~/SiO~2~/RhB NPs was accomplished by following two different synthetic routes involving: (a) the deposition of the nanoparticles onto the fibers' surfaces *via* spraying (denoted as sprayed magnetic fibers) and (b) the mixing of the nanoparticle dispersion prepared in ethanol with the CA acetone solution followed by electrospinning (denoted as electrospun magnetic fibers). RhB was chosen to be covalently linked onto the nanoparticles' surfaces due to its high photostability compared to other fluorescent dyes introduced in previous studies as active moieties in optical sensing applications^[@CR58]^. Actually, earlier studies in our group involved the incorporation of fluorescein (FL)-functionalized core-shell ferrous nanoparticles (γ-Fe~2~O~3~/SiO~2~/FL NPs) within CA electrospun fibers by following the spraying deposition methodology which however demonstrated low photostability, in agreement with previous reports^[@CR59]^. The γ-Fe~2~O~3~/SiO~2~/RhB NPs-functionalized CA fibers were visualized by SEM and fluorescence microscopy. As seen in Fig. [4(B,C)](#Fig4){ref-type="fig"}, in both cases, the presence of the nanoparticles on the fibers' surfaces can be clearly seen. Transmission electron microscopy (TEM) analyses confirmed the presence of γ-Fe~2~O~3~/SiO~2~/RhB NPs along the nanocomposite fibers (Fig. [5](#Fig5){ref-type="fig"}). By comparing the TEM images of the electrospun magnetic fibers (Fig. [5A--C](#Fig5){ref-type="fig"}) with the sprayed magnetic fibers (Fig. [5E--G](#Fig5){ref-type="fig"}), the presence of the Fe~2~O~3~/SiO~2~/RhB NPs exclusively onto the fibers' surfaces can be observed in the second case, in contrary to the electrospun magnetic fibers where the nanoparticles are mainly accumulated within the fibers. Moreover, the synthetic protocol affected the mean fibers' diameter significantly where electrospun magnetic fibers' approach yields thinner while sprayed magnetic fibers' approach yields thicker fibers with diameter of 547 nm and 770 nm, respectively.Figure 5TEM images of the electrospun magnetic fibers (**A**--**C**) and the sprayed magnetic fibers (**E**--**G**). Fibers diameter distributions were determined by analysis of the TEM images corresponding to the electrospun magnetic fibers (**D**) and the sprayed magnetic fibers (**H**). In the case of the electrospun materials, the core-shell nanoparticle morphology could clearly be resolved in images at higher magnifications (Fig. [5B,C](#Fig5){ref-type="fig"}). The Fe~2~O~3~/SiO~2~/RhB NPs are relatively homogeneously distributed along the fibers while on the nanoscale there is some segregation composed of up to dozen nanoparticles present in small aggregates. Fluorescence microscopy was used to verify the fluorescence properties of the multifunctional nanocomposite fibers. Figure [6](#Fig6){ref-type="fig"} provides the fluorescence images of the produced materials. By observing the fluorescence microscopy images it can be seen that fluorescence is not homogeneous for the entire sample, particularly in the case of the electrospun magnetic fibers (Fig. [6B](#Fig6){ref-type="fig"}). According to the TEM data (provided in Fig. [5](#Fig5){ref-type="fig"}), the sprayed magnetic fibers have Fe~2~O~3~/SiO~2~/RhB NPs anchored all over their external surfaces. In contrary, in the case of the electrospun magnetic fiber analogues, the nanoparticles that are embedded within the CA fibers form clusters and the presence of nanoparticle-free regions along the fibers can be clearly observed, resulting to a fluorescence "inhomogeneity" along the fibers.Figure 6Fluorescence microscopy image of the sprayed magnetic fibers (**A**) and the electrospun magnetic fibers (**B**). The fluorescence efficiency of the Fe~2~O~3~/SiO~2~/RhB NPs anchored onto the CA fiber surfaces was investigated by photoluminescence spectroscopy at 520 nm excitation wavelength (Fig. [7](#Fig7){ref-type="fig"}). The emission wavelength was recorded at 574 nm, in agreement with previous reports recording the emission wavelength of RhB within 574--577 nm^[@CR60]--[@CR62]^.Figure 7Photoluminescence spectra of the γ-Fe~2~O~3~/SiO~2~/RhB NPs-functionalized CA fibers obtained *via* spray deposition (via spray deposition and electrospinning) (excitation wavelength: 520 nm). The magnetic behaviour of the nanocomposite membranes was investigated by VSM at room temperature. Figure [8](#Fig8){ref-type="fig"} presents the magnetization *versus* applied magnetic field strength plots for the 2 types of the γ-Fe~2~O~3~/SiO~2~/RhB NPs-functionalized CA nanocomposite fibers. As seen in the plots, both systems exhibited superparamagnetic behavior at ambient temperature, demonstrated by the symmetrical sigmoidal shape of the magnetization curves and the absence of a hysteresis loop. The sprayed magnetic fibers had a higher saturation magnetization value (Ms ∼ 0.76 Am^2^ kg^−1^) compared to the electrospun magnetic fibers (Ms ∼ 0.20 Am^2^ kg^−1^) due to the presence of the non-magnetic coating (cellulose acetate) around the NPs embedded within the fibers (in contrast to the NPs that are deposited onto the fiber surfaces), resulting to the decrease in magnetization due to quenching of surface effect^[@CR63]^. Moreover, based on VSM magnetic measurements, the amounts of the γ-Fe~2~O~3~/SiO~2~/RhB NPs in the functionalized CA nanocomposite fibers were estimated to be ∼2.1 wt. % and ∼0.5 wt. % for the sprayed magnetic fibers and the electrospun magnetic fibers respectively. These differences justify the differences observed in the photoluminescence spectra corresponding to the 2 cases (Fig. [7](#Fig7){ref-type="fig"}).Figure 8A room-temperature measurement of the magnetization as a function of magnetic field strength of the γ-Fe~2~O~3~/SiO~2~/RhB NPs-functionalized CA fibers obtained *via* spray deposition (**A**) and electrospinning (**B**). Gas (ammonia) and pH sensing {#Sec12} ---------------------------- Both electrospun and sprayed magnetic fibers were evaluated for different gas ammonia concentrations and pH values. The emission spectrum of the sample was stable upon continuous illumination, showing no self-quenching mechanisms observed with other fluorescent moieties such as anthracene^[@CR41]^. For the measurements, the integration time of the spectrometer was set to 10 s. The averaging method, rolling average over the last 3 acquisitions, was used. The sample was exposed for 30 s for each data point, showing fast response to the measurands. As each individual measurement corresponds to 30 s, this can be an estimate of the time response which is comparable to other electrospun-based ammonia sensors^[@CR64],[@CR65]^ where the response times range between 50 s--350 s, as well as fluorescence-based sensors^[@CR66]^. In that sense, the response of the materials presented herein can be considered as fast. In order to assure the reliability of measurements for each recorded value, the sample was illuminated for 10 more minutes before changing NH~3~ concentration or pH, showing no further change. This stability demonstrates that at continuous excitation, the properties of the material at constant ammonia concentration were not deteriorated and also that the response is attributed reliably to the specific measuring conditions. When exposed to NH~3~ vapors, RhB undergoes structural changes resulting to the generation of a non-fluorescent lactone (Fig. [9](#Fig9){ref-type="fig"})^[@CR67]^. The latter explains the reduction observed in the fluorescence intensity upon exposure of the nanocomposite γ-Fe~2~O~3~/SiO~2~/RhB NPs -functionalized fibrous mats in NH~3~ (Fig. [10A,C](#Fig10){ref-type="fig"}).Figure 9Sensing (fluorescence quenching) mechanism of RhB molecules undergoing structural changes when exposed to NH~3~ vapours, resulting to the generation of the non-fluorescent lactone form.Figure 10(**A**) Fluorescence spectra of the electrospun magnetic fibers for different concentrations of ammonia gas. (**B**) Response of the electrospun magnetic fibers at 577 nm for different ammonia concentrations. (**C**) Fluorescence spectra of the sprayed magnetic fibers for different concentrations of ammonia gas. (**D**) Response of the sprayed magnetic fibers at 577 nm for different ammonia concentrations. Figure [10A](#Fig10){ref-type="fig"} shows the fluorescence spectrum of the electrospun magnetic fibers taken for different ammonia concentrations. A reference intensity value (*I*~*ref*~) was taken before inserting the ammonia vapors. The fibers' response is defined as:$$\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Response\,( \% )=\frac{{I}_{sig}-{I}_{ref}}{{I}_{ref}}\times 100$$\end{document}$$where *I*~*sig*~ is the intensity measured for an ammonia concentration. Figure [10B](#Fig10){ref-type="fig"} shows the response corresponding to the 577 nm peak, demonstrating a clear ammonia sensing for concentrations up to 11000 ppm. The sprayed magnetic fibers decorated on their external surfaces with γ-Fe~2~O~3~/SiO~2~/RhB NPs also show a clear response to the ammonia vapours as presented in Fig. [10C,D](#Fig10){ref-type="fig"} for NH~3~ concentrations up to \~13000 ppm. The detection of high NH~3~ concentrations can be attributed to the large surface-to-volume ratio of the electrospun fibers. Previous studies have shown that due to their large surface-to-volume ratio, electrospun fiber sensors exhibit enhanced sensitivity^[@CR67]^ as well as quenching efficiency^[@CR38]^ compared to thin films. The overall factor that enables the high concentration detection is actually the total number of sensing elements which are the γ-Fe~2~O~3~/SiO~2~/RhB NPs. Their total number in a specific volume of the material is determined by the three dimensional (3D) fibrous form and the concentration of the functionalised nanoparticles. However, the surface enhancement enabled by the fibrous morphology is the dominant factor which can be considered as a nonlinear scaling factor in a three dimensional space where the nanofibrous material is organised. The arrangement of nanoparticles in a linear shaped fiber scales linearly but in this 3D fibrous form the total effect can be eventually characterised by a nonlinear scaling or enhancement factor. The response of the sprayed magnetic fibers is \~5% lower compared to the electrospun magnetic fibers. This could be attributed partially to the fact that in the former case the NPs can easily detach from the fibers' surfaces before and during the measurements because they are only weakly attached to them. The sample starts to quench above \~12000 ppm and the response curve reaches a plateau. Therefore the sensor can be used for NH~3~ concentrations detection up to \~12000 ppm. Above this limit, the sensor cannot be reliably used as it becomes permanently "poisoned" and irreversible. Furthermore, our experimental data show no consistent reversibility in high concentrations ammonia sensing. This can be attributed to the high ammonia concentration, as heavy loading (with ammonia) tends to destabilize the sensor, resulting in poor reversibility and smaller relative signal changes^[@CR68]^. However, despite the non-reversible nature of the demonstrated behaviour, there are several sensing and detection applications where reversibility is not required as the purpose of the sensing element is to log/register successfully a specific high concentration of ammonia. Such cases are important in ammonia leakage monitoring especially at very high concentrations that can become toxic. At the current stage of development the applications of detection of high level of ammonia concentration (above 1000 ppm) could be used at industrial facilities employing ammonia transfer lines and dense storage spaces, as the presented sensors can withstand the poisoning and saturation effects and can thus provide linear response up to 12000 ppm. Low cost sensing elements or materials can then be replaced after logging an ammonia leakage event. Although the physical form of the electrospun fibrous materials is suitable for high concentration gas detection, it is not considered ideal for efficient collection of fluorescent light that could enable low level of detection. This is attributed to the fact that the material arrangement in the measuring cell could be vulnerable to external factors like vibrations, air flow etc., resulting to possible displacements that may alter significantly the optimal excitation and collection light angles. Therefore, for the optimization of the sensing performance the development of a more robust customised miniaturised measuring apparatus is required in future studies. Furthermore, modification of the electrospun fibrous materials towards the development of electrospun rigid fibrous mats will facilitate in future studies a highly controllable characterization process in lower ammonia concentrations by improving also the optical excitation and collection system. The γ-Fe~2~O~3~/SiO~2~/RhB NPs-functionalized CA fibers were also evaluated as pH sensors in aqueous solutions with different pH values. Both, electrospun and sprayed magnetic fibers were tested but only the former exhibited a consistent response/behaviour in aquatic environments. On the contrary, detachment of the γ-Fe~2~O~3~/SiO~2~/RhB NPs decorating only the surface of the sprayed magnetic fibers was observed when the latter were immersed in aquatic solutions, thus indicating the limited robustness of this system compared to the electrospun analogue. These results may be attributed to the weak nanoparticle/polymer matrix interactions arising from the surface functionalization *via* spray deposition. The latter is further supported by the TEM analysis provided in Fig. [5E--G](#Fig5){ref-type="fig"}. In Fig. [11A](#Fig11){ref-type="fig"}, the fluorescent intensity of the electrospun magnetic fibers immersed in acidic aqueous solutions with different pH values, ranging from 5 to 1, is presented. The fibers were firstly immersed in a pH 5 solution and subsequently in solutions with lower pH values. As the pH decreases, the fluorescent intensity clearly increases.Figure 11(**A**) Fluorescence spectra of the electrospun magnetic fibers for acidic aqueous solutions with different pH values. (**B**) Fluorescence intensity at 580 nm versus pH. **(C**) Fluorescence spectra of the electrospun magnetic fibers for basic aqueous solutions with different pH values. (**D**) Fluorescence intensity at 580 nm versus pH. The electrospun magnetic fibers were also tested as fluorescent sensors in alkaline aqueous solutions with different pH values ranging from 8 to 13. The fibers were firstly immersed in a pH 8 solution and subsequently in solutions with higher pH values. As the pH increases the fluorescent intensity decreases (Fig. [11C,D](#Fig11){ref-type="fig"}). The intensity at the 580 nm peak is linearly depended on the pH value with R-square values of 0.9577 and 0.8817 for the acid and alkaline environments, respectively (Fig. [11B,D](#Fig11){ref-type="fig"}). The values of the slopes for the acid and alkaline environments are −0.00158 and −0.00165, respectively, with corresponding standard deviations of 0.0001917 and 0.0003028. The fact that the standard errors are less than 20% of the slope values strongly indicates that the correlation is linear. It should be noted that as the characterization of response in alkaline and acidic environments was performed by two different set of experiments due to current experimental limitations, the measurements cannot be directly compared or related in terms of the measured intensity, because of different placement conditions of the electrospun material in the measuring apparatus. However, the behaviour of the material is consistent over the entire pH range. The samples were also evaluated towards pH sensing reversibility by alternating solutions with pH values of 2 and 7. The sample exhibited reversible on/off switchable fluorescence emission as presented in Fig. [12](#Fig12){ref-type="fig"} in accordance to the literature^[@CR69],[@CR70]^. It is noteworthy to emphasize that the γ-Fe~2~O~3~/SiO~2~/RhΒ NPs--functionalized electrospun magnetic fibers characterised 6 months after production towards pH sensing, exhibited a measurable response for a relatively long range of pH between 2 and 7, together with a consistent reversibility, demonstrating their stability and long term functionality. The relatively small degradation of response signal could be attributed to either a drift or poisoning effect, or also to combined random changes in the environmental conditions and setup's parameters during the long duration (\~120 min) of the experiment. Despite this degradation the discrimination between the two pH values is still very clear and reliable.Figure 12Reversibility of the pH-dependent on-off-on fluorescence intensity profile of electrospun magnetic fibers. The intensity was normalised to the intensity measured at pH = 7. The response time of the pH sensor is 30 s as estimated by the fluorescence measurements instrumentation. To summarize, concerning the ammonia sensing performance, the sprayed magnetic fibers showed a \~5% lower response compared to the electrospun analogues, partially due to nanoparticle concentration differences in the two samples deriving from the fact that in the case of spray deposition the NPs are detached from the fibers' surfaces before and during the measurements because they are only weakly attached to them. Further to the above logical assumption, it should be underlined that the absolute value of NPs concentration and the resulted photoluminescence (Fig. [7](#Fig7){ref-type="fig"}) of the two different fiber types should not be related to their sensing responses, as the response is calculated to a reference intensity value (Eq. [1](#Equ1){ref-type=""}) connected to the absolute initial photoluminescence. The absolute photoluminescence is reflected only by the absolute values of the measured intensity in the graphs appearing in Fig. [10A,C](#Fig10){ref-type="fig"}, which are indeed higher for the case of the sprayed magnetic fibers (Fig. [10C](#Fig10){ref-type="fig"}). Therefore, it is noteworthy to stress out that the response in the case of the electrospun magnetic fibers (Fig. [10B](#Fig10){ref-type="fig"}) is higher despite the fact that the sensing NPs are embedded in the hosting fibers and are not directly exposed to ammonia gas as in the case of the NPs deposited onto the fibers' surface *via* spraying. The adsorption mechanism of ammonia in the electrospun magnetic fibers is proved equivalent (or even more efficient), compared to the direct sensing on the sprayed NPs because of the minimal dimension of fibers. Furthermore, the stability of the NPs is retained, since these are protected within the fibers and they are not directly exposed to external degradation factors. Based on the above, the sprayed magnetic fibers were found to be less robust as pH fluorescent sensors owned to the fact that the γ-Fe~2~O~3~/SiO~2~/RhB NPs decorating only the fibers' surfaces were detached upon immersion of the fibers in aquatic solutions indicating their ineffectiveness in applications involving aquatic environments. Conclusions {#Sec13} =========== In this work, the fabrication and characterization of cellulose acetate electrospun fibers doped with γ-Fe~2~O~3~/SiO~2~/RhB NPs is reported. Two different fabrication protocols were followed. In the first synthetic approach, the γ-Fe~2~O~3~/SiO~2~/RhB NPs were sprayed on top of the fibrous mat while in the second, a NP suspension prepared in CA acetone solution was directly electrospun. In the latter case, due to the encapsulation of the NPs within the fibers and the covalent anchoring of the RhB fluorophore onto the NP surfaces, the fluorophore's leakage from the fibrous mat is prevented, enabling thus stable fluorescence-based operation of the developed materials. In ammonia sensing performance, the sprayed magnetic fibers show a response \~5% lower compared to the electrospun magnetic fibers. It should be stressed out that the electrospun magnetic fibers exhibit equivalent or even higher response despite the fact that the sensing γ-Fe~2~O~3~/SiO~2~/RhB NPs are embedded in the hosting fibers and not directly exposed to ammonia gas as in the case of the sprayed analogues, while - as mentioned above - the embedded NPs retain also a much better stability since they are protected within the fibrous polymer matrix. The electrospun nanocomposite fibers were evaluated for both gas ammonia and pH sensing. Due to the large surface to volume ratio of the functionalized fibrous mats, high ammonia concentrations up to 12000 ppm were detected. Furthermore, the electrospun magnetic fibers showed fast and linear response to aquatic solutions of long pH range from acidic to alkaline environments. However, the stability of the sprayed magnetic fibers is limited due to the detachment of the Fe~2~O~3~/SiO~2~/RhB NPs from their surfaces. **Publisher's note** Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Authors acknowledge financial support from the Slovenian Research Agency for research core funding No. (P2-0089) and for the projects "Nanotheranostics based on magneto-responsive materials" (No. J1-7302) and "Tunnelling nanotubes for innovative urinary bladder cancer treatments" (No. J3-7494). We acknowledge support of this work by the project "Advanced Materials and Devices" (MIS 5002409) which is implemented under the "Action for the Strategic Development on the Research and Technological Sector", funded by the Operational Programme "Competitiveness, Entrepreneurship and Innovation" (NSRF 2014--2020) and co-financed by Greece and the European Union (European Regional Development Fund). We also thank Dr. Triantafyllos Stylianopoulos (Cancer Biophysics Laboratory, Department of Mechanical and Manufacturing Engineering, University of Cyprus) for providing access to the fluorescence microscope. Finally, the authors also acknowledge support from the COST Action MP1206 "Electrospun Nanofibres for Bioinspired Composite Materials and Innovative Industrial Applications". Theodora Krasia-Christoforou and Christos Riziotis conceived the idea and designed this project. Afroditi Petropoulou performed all experiments related to ammonia and pH sensing. Slavko Kralj synthesized the RhB-functionalized ferrous core-shell nanoparticles and performed their characterization by TEM, VSM and DLS. Moreover, he carried out morphological studies on the produced nanocomposite fibrous membranes by TEM. Xenofon Karagiorgis, Emilios Loizides and Ioanna Savva carried out the fabrication of the electrospun nanocomposites and their morphological characterization by SEM. Finally, Myrofora Panagi performed the analysis of the produced materials by fluorescence microscopy. Theodora Krasia-Christoforou and Christos Riziotis wrote the paper with input from all other authors. The authors declare no competing interests.

The global campaign to eradicate dracunculiasis (or Guinea worm disease \[GWD\]), which began at the Centers for Disease Control and Prevention (CDC) in 1980 and has been led by the Carter Center since 1986, has assisted 17 of 21 affected countries to interrupt transmission. In 1986, an estimated 3.5 million cases occurred annually in 20 countries (the separation of Sudan and South Sudan created a 21st country in 2011). Today, only 4 countries (Chad, Ethiopia, Mali, and South Sudan) still have endemic transmission of GWD. In 2014, only 126 cases were reported from these 4 countries, and only 22 cases were reported during January--November 2015 ([@R1]). Since the beginning of the campaign, all interventions against GWD have focused on preventing contamination of stagnant sources of drinking water by patients with patent GWD and preventing ingestion of infected copepods. However, in 2013, several unusual epidemiologic findings were noted in Chad, including absence of disease outbreaks associated with sources of drinking water common to many residents year-to-year in affected villages and a relatively common infection of domestic dogs with Guinea worms genetically indistinguishable from those from human cases. This finding led to the hypothesis that an aquatic paratenic host (an intermediate host that serves as transport host for parasite larvae) was involved in the transmission of *Dracunculus medinensis* in Chad ([@R2]). Since 2013, the sporadic pattern of human cases in Chad has continued, the number of infections in dogs has continued to increase, and the presence of a paratenic host in the transmission cycle seems more likely. The purpose of our study was to expose, under laboratory conditions, 2 species of fish and 2 species of tadpoles to *D. medinensis*--infected copepods, in an effort to determine whether these species would support viable larvae (and thus serve as a paratenic host). These potential paratenic hosts were then fed to an experimental definitive host (a domestic ferret) to determine if any third-stage larvae (L3) present would undergo further development in ferrets. Materials and Methods ===================== During July 20--22, 2015, batches of first-stage larvae (L1) were recovered from 5 Guinea worms removed from infected dogs resident in villages within the dracunculiasis-endemic zone along the Chari River between the cities of Guelendeng and Bousso in the Mayo-Kebbi Est region of Chad. Copepods were collected locally in the Guelendeng area (dracunculiasis-endemic zone) and exposed to the L1 per standard methods ([@R3]). L1 were mixed with copepods in a ratio of ≈3--5:1 to ensure a high rate of infection but not high enough to cause copepod mortality. Infected copepods were maintained for 5--7 days in 1.5-L water bottles and then transported to laboratories at CDC (Atlanta, GA, USA), where they were maintained in culture per standard practices ([@R3]). Beginning at day 12 after exposure, copepods were dissected to determine level of infection and stage of development of larvae. Many larvae observed at day 12 were early (and infective) L3. To allow full maturation of the infective larvae, no copepods were used in experimental studies before day 14. For experimental trials, 2 species of fish (29 Nile tilapia \[*Oreochromis niloticus*\] and 36 fathead minnows \[*Pimephalis promelas*\]) and 2 species of tadpoles (20 Fowler's toads \[*Bufo fowleri*\] and 10 green frogs \[*Lithobates* (*Rana*) *clamitans*\]) were used. Both fish species were captive-raised and the tadpoles were wild-caught (green frogs raised from egg masses in captivity) in Georgia. Because the toads and some of the ranid tadpoles were very young at the time of exposure, species identity was confirmed by amplification and sequencing of a ≈500-bp region of the 16S rRNA gene with primers Amph-sp-F (5′-CTGTTTACCAAAAACATCG-3′) and Ecto-univ-R (5′-ATCCAACATCGAGGTCGT-3′). Sequences were a 100% match to GenBank accession numbers AY680224 (toads) and DQ283185 (ranids) confirming their identity. Finally, 2 age classes of green frog tadpoles were used. Beginning at day 14 after exposure to *D. medinensis* L1, batches of infected copepods were fed to groups of fish or tadpoles. Fish and tadpoles, separated by species, were exposed to copepods in shallow water in a 500-mL beaker with a bubbling air source. Both fish species immediately began to ingest copepods, all of which had been ingested by 24 hours. In contrast, tadpoles ingested copepods more slowly, and some copepods remained uneaten after 3 days of exposure. Beginning 1 week after exposure, groups of 4--5 fish and tadpoles were examined for *Dracunculus* larvae. Fish and tadpoles were euthanized by overdose of pH-neutral buffered tricaine methane sulfonate (MS-222), followed by decapitation, and then examined grossly under a dissecting scope to see if any larvae could be visualized in fin or tail fin or under skin. Then, the entire body, including head and viscera, was dissected to fully separate all tissues, and macerated tissues were examined under a dissecting microscope. If no larvae were observed, the tissues were allowed to sit in phosphate-buffered saline for 1--2 hours and reexamined under a dissecting or compound microscope. A total of 19 tilapia, 19 fathead minnows, 11 *Bufo* tadpoles, and 9 *Lithobates* tadpoles were examined by dissection. For a subset of 6 fish (3 of each type), tissues were grossly macerated, examined microscopically, and then digested in a 0.5% pepsin solution for 1 hour before examination for larvae. Two colony-reared ferrets (*Mustela putorius furo*) were fed either tadpoles or fish mixed in canned cat food to determine infectivity of L3 from aquatic hosts. All dissected fish and the additional 5 tilapia and 11 fathead minnows that were not dissected were fed to 1 ferret. The other ferret was fed all dissected tadpoles and an additional 9 *Bufo* and 3 *Lithobates* tadpoles that had not been dissected. Ferrets were euthanized and examined at 70--83 days after exposure to fish or tadpoles. Animals were examined per previously proven methods for detecting developing *Dracunculus* ([@R3]). All animal procedures were reviewed and approved by the University of Georgia Institutional Animal Care and Use Committee (no. A2014 11--010). Results ======= No *Dracunculus* larvae were detected in dissected fish or *Bufo* tadpoles, but 4 of 7 *Lithobates* tadpoles were infected with 1--3 *Dracunculus* L3. Infections were noted in both age classes. The exact locations of the larvae were not determined, but they were recovered from the body musculature or head. No larvae were observed in the body cavity or viscera. The larvae recovered from tadpoles were slightly larger and more active than L3 recovered from copepods. No worms were detected in the ferret that had been fed fish, but in the ferret that had ingested tadpoles, 3 young developing worms were recovered. All 3 worms were females and measured 1.4, 2.0, and 2.7 cm in length by 295--350 μm in maximum diameter. Two of the worms were recovered from the right hind leg, and the third worm was recovered from the lower left abdominal wall. The largest worm was coiled under the muscle fascia, whereas the other 2 worms were present in adipose tissue. Discussion ========== These results, although limited in scope, clearly confirm that *D. medinensis*, like *D. insignis*, can use an aquatic paratenic host; specifically, at least 2 species of amphibians ([@R4]--[@R6]). Although tadpoles consumed far fewer copepods, most tadpoles exposed to infected copepods subsequently had infections, which is consistent with previous data that showed a high percentage of adult frogs (*L. pipiens* and *L. clamitans*) acquired infections when a very high dose of infected copepods (n = 200--500) were given by mouth ([@R4]). *D. insignis* larvae recovered from tadpoles previously were stated to be slightly larger and more active than infective larvae recovered from copepods, which is what we noted in our study with *D. medinensis* ([@R4],[@R6]), in which the larvae grew ≈20% after 15--18 days. Previously these *D. insignis* larvae recovered from tadpoles or frogs were infective to a single raccoon or ferrets, respectively, proving that these larger larvae are infectious ([@R4],[@R5]). Similarly, the larger *D. medinensis* larvae we recovered from tadpoles were infectious for a domestic ferret. The absence of larvae in the 2 species of fish we included in this study does not rule out a role of fish as paratenic hosts. Previously, a low percentage of fish exposed to L3 recovered from copepods subsequently had infections ([@R4]). Although sample sizes were low, 2 of 4 white suckers (*Catostomus commersonii*) and 1 of 2 rainbow trout (*Oncorhynchus mykiss*) had 1--2 larvae recovered after exposure to 100--180 L3. Thus, we may not have exposed our fish to enough larvae to become infected ([@R4]). However, fish species variability in susceptibility is probable, given that common shiners (*Luxilus cornutus*) failed to become infected even though 3 were exposed to [\>]{.ul}200 larvae ([@R4]). Future studies should investigate the fate of *Dracunculus*-infected copepods after ingestion by fish hosts, dose required to establish infections in fish, and additional trials with other species of fish (e.g., *Gambusia* \[mosquito fish\]) that are known to predate copepods. Tilapia, 1 type of fish used in this study, is common in Chad and widely used as a food source. However, many other types of fish are present in Chad, and continued study of tilapia and other native fish should also be undertaken before ruling out fish as potential paratenic hosts. Collectively, our data and the findings of previous reports indicate that *Dracunculus* larvae in general, and *D. medinensis* larvae specifically, are well-adapted to using a paratenic host and that tadpoles of *Lithobates* (*Rana*) and *Xenopus* species are appropriate hosts ([@R4],[@R5]). *Lithobates* spp. are members of the family Ranidae, which has a near global distribution that includes more than 180 known species in sub-Saharan Africa ([@R7]). This study also confirms that domestic ferrets, like domestic cats, domestic dogs, and monkeys, can serve as experimental definitive hosts for *D. medinensis*. The recovered worms were of consistent size and development as previously reported for worms of similar age recovered from dogs or monkeys ([@R8],[@R9]). Finally, these results also suggest that a more extensive examination of tadpoles and frogs in Chad is warranted. Although a small number (n = 28) of ranid frogs from Chad were previously examined for *Dracunculus* larvae and all were negative ([@R2]), sample sizes were low. Because natural infections of tadpoles or frogs have not been documented for either *D. medinensis* or *D. insignis,* the prevalence of natural infections is unknown; therefore, larger numbers of wild-caught frogs should be examined in future efforts. In addition to identifying which aquatic animals are acting as paratenic hosts for *D*. *medinensis* in Chad, it would be important to also identify which wild animals are predators of these transport hosts and whether those predators develop patent infections, thus helping maintain transmission of the parasite. *Suggested citation for this article*: Eberhard ML, Yabsley MJ, Zirimwabagabo H, Bishop H, Cleveland CA, Maerz JC, et al. Possible role of fish and frogs as paratenic hosts of *Dracunculus medinensis*, Chad. Emerg Infect Dis. 2016 Aug \[*date cited*\]. <http://dx.doi.org/10.3201/eid2208.160043> During 2008--2015, the Carter Center's work to eradicate Guinea worm disease was made possible by financial and in-kind contributions from many donors. A full listing of supporters is available at <http://www.cartercenter.org/donate/corporate-government-foundation-partners/index.html>. Dr. Eberhard is a retired parasitologist with broad interest in parasite life cycles and transmission dynamics. He has been engaged in the Guinea worm eradication program since 1986, and continues to work with the US Centers for Disease Control and Prevention, the Carter Center, and the World Health Organization on the eradication program.

INTRODUCTION {#sec1-1} ============ Vascular malformations of spinal cord are rare clinical entities.\[[@ref6][@ref8][@ref15]\] Traditionally, they are classified into three categories depending upon the arterial supply and anatomic characteristics.\[[@ref7]\] Spinal dural arteriovenous fistula (SDAVF) constitutes 70--80% of all spinal vascular malformations with an annual incidence of 5-10 cases per million.\[[@ref1]\] The vascular nidus is located on the dural sheath of a spinal nerve root and is supplied by a dural artery, usually a branch of radicular or intercostal artery\[[@ref3]\] \[[Figure 1](#F1){ref-type="fig"}\]. Intradural arteriovenous malformations (AVMs) and cavernous angiomas are less common types of vascular malformations of spinal cord. {#F1} SDAVF are generally thought to be acquired lesions and usually present in mid to late adulthood.\[[@ref20]\] The pathophysiology is thought to be congestion of the spinal cord medullary venous plexus with edema and neuronal ischemia.\[[@ref3]\] The clinical features include radicular or axial back pain, gait imbalance, weakness, numbness, and bladder or bowel disturbances. Aminoff and Logue in their seminal observations described the clinical features and postulated a disability score to classify the severity of SDAVF\[[@ref3][@ref9]\] (Aminoff and Logue Scale, ALS) \[[Table 1](#T1){ref-type="table"}\]. ###### Aminoff and logue scale score of disability  Traditionally, there are two therapeutic options: (1) Surgical approach of laminectomies and ligation of the vascular nidus \[[Figure 2](#F2){ref-type="fig"}\] and (2) An endovascular approach with focus on embolization of feeding vessel/s. \[[Figure 3](#F3){ref-type="fig"}\]\[[@ref1][@ref2][@ref3]\] There is little data on superiority of one approach over the other.\[[@ref3][@ref4][@ref5][@ref6][@ref7]\] We present our single center experience of 27 patients including clinical features, imaging findings, and response to various modalities of treatment, including complication and recurrence rates. {#F2} {#F3} MATERIALS AND METHODS {#sec1-2} ===================== This was a retrospective study of patients diagnosed with SDAVF who underwent evaluation and treatment at Duke University Medical Center. Duke Institutional Review Board (IRB) approved the research protocol for this study. Study population {#sec2-1} ---------------- All adult patients (age more than 18 years at the time of presentation) who were diagnosed with SDAVF and who underwent evaluation and treatment between January 1, 1993 to December 31, 2012 were identified using the appropriate current procedural terminology (CPT) codes and inclusion characteristics using special software database program DEDUCE. DEDUCE is a comprehensive search database of all patients who underwent evaluation at the medical center over past 10 years. The list of patients was further checked manually by a team of health professionals using electronic records (eBrowser™). Patients were excluded from the study if the diagnosis was doubtful or if the feeding vessel for SDAVF was branch of one of the extracranial arteries. A total of 27 patients met inclusion criteria. Clinical details prior to intervention and following intervention at discharge and follow up were used to calculate ALS \[[Table 1](#T1){ref-type="table"}\]. Demographic details, intervention details, follow up, and complication-recurrence details were obtained from electronic medical records (eBrowser™) and from the Duke Decision Support Repository (DSR). DSR is a quality assured custom-built data warehouse containing integrated clinical and financial data of all patients admitted to the Duke Health Care System. Statistical analysis {#sec2-2} -------------------- Mann--Whitney U test or Chi-square tests were used for intergroup comparisons as appropriate. Data was reported as median (interquartile range) for continuous variables and count (percentage, %) for categorical variables. Statistical significance was defined as a *P* \< 0.05. All analysis was done using SPSS Inc. software version 20, Chicago, IL, USA. RESULTS {#sec1-3} ======= Baseline characteristics {#sec2-3} ------------------------ A total of 27 patients were diagnosed with SDAVF; 20 (74%) were male and the majority (22, 88%) patients were older than 50 years at the time of diagnosis. Majority of patients presented with back pain (93%) and lower extremity weakness (85%). Almost half the patients had sensory disturbances. A significant number of patients reported bladder difficulties (59%) while none of the patients in our study reported impotence or sexual dysfunction. All patients underwent conventional digital subtraction spinal angiogram (DSA) for accurate diagnosis and/or intervention. In addition, all patients underwent magnetic resonance imaging (MRI) as part of pretreatment evaluation. MRI scan showed T2 signal hyper intensity involving spinal cord in 24 (89%) patients. In addition, seven (26%) patients underwent magnetic resonance angiography (MRA), which was diagnostic of SDADF in four patients and was suggestive in three patients. The median ALS was 4 at preintervention and 3 at postintervention. Demographic and clinical details of all patients are summarized in [Table 2](#T2){ref-type="table"}. ###### Demographic details of study population  Treatment details and outcome measures {#sec2-4} -------------------------------------- All patients underwent DSA for diagnosis and localization of fistula. After endovascular or surgical correction, a repeat DSA was performed to assess the closure of fistula or any residual feeding vessels. All patients were followed up clinically and a repeat DSA was performed if there was suspicion of recurrence during course of follow up (worsening of residual symptoms or appearance of new symptoms or both). Out of the 27 patients in the study, 10 patients underwent endovascular embolization as the first line therapy, whereas the remaining 17 patients underwent surgical ligation as initial therapeutic modality. The decision to treatment was based on individual fistula characteristics and associated comorbidities. For example, a common segmental artery origin of the artery of Adamkiewicz and the SDAVF feeding vessel was considered a contraindication for endovascular embolization and felt suitable for surgical ligation. Onyx (ethylene vinyl alcohol) was used as an embolization material in five patients and N-butyl 2-cyanoacrylate (NBCA) was used in the remaining five patients. Patients in surgical group underwent removal of vertebral lamina at involved levels in addition to ligation-resection of the fistula. Preoperative steroids were used in one patient in embolization group as compared with four patients in surgical group. Preoperative steroids were used to reduce cord edema. Patients in endovascular group were slightly older (mean age 65 years) as compared with those in surgical group (mean age 57 years). There was no significant difference in the demographic characteristics and clinical features in both these groups \[[Table 2](#T2){ref-type="table"}\]. The average ALS prior to procedure for both these groups were comparable (4.5 in embolization group and 4 in surgical group). The average duration of postprocedure inpatient hospital stay was 3 days in patients who underwent embolization, as compared with 4 days in patients who underwent surgical ligation and resection. The average ALS at discharge improved in both groups by one point (from 4.5 to 3.5 in the embolization group and from 4 to 3 in the surgical group). One patient in embolization group had spinal cord infarction in immediate postprocedure period. None of the patients in embolization group had evidence of local or systemic infection (pneumonia, urine tract infection within 4 weeks of procedure). Two patients in surgical group presented to the hospital within a month of discharge with local wound infection, requiring repeat inpatient admission, intravenous antibiotics, and local wound care. None of the patients in surgical group had postprocedure recurrence on follow up as compared with three patients in embolization group who presented with recurrence. All of the three patients with recurrence required surgical ligation of fistula and none of them had repeat recurrence of fistula. These results are summarized in [Table 3](#T3){ref-type="table"}. ###### Comparison of patients undergoing embolization and surgical treatment  DISCUSSION {#sec1-4} ========== SDAVF is a rare disorder with annual incidence rate of 5-10 per million. It is thought to be an acquired disease, and men are more commonly affected than women, as seen in our patient cohort.\[[@ref7][@ref8]\] The presentation is usually insidious and subacute. The underlying pathophysiology is thought to be medullary venous hypertension secondary to a direct abnormal communication between arterial supply and venous drainage.\[[@ref2][@ref8][@ref9]\] The increased venous pressure leads to venous congestion, dilation, and tortuous medullary veins. This results in compression and edema of spinal cord, which is often visualized as T2 signal hyper intensity on MRI scan\[[@ref10]\] \[[Figure 1](#F1){ref-type="fig"}\]. It is thought that increased venous pressure could contribute to decreased tissue perfusion and hypoxia. Frank infarction or hemorrhage in the substance of spinal cord is extremely rare in cases of SDAVF.\[[@ref4]\] Almost all the patients present with varying degrees of radicular and back pain. In addition, compressive myelopathy due to edema and tissue hypoxia leads to lower extremity weakness, sensory disturbances, as well as bladder symptoms.\[[@ref7][@ref10]\] ALS is widely used for assessing clinical severity of SDAVF. Advances in neuroimaging have led to better diagnostic modalities for SDAVF.\[[@ref11][@ref12][@ref13]\] As seen from our observations, all the patients underwent MRI scan of spine to evaluate for cause of myelopathy. All of the patients had abnormal cord signal change and a significant proportion of patients (24 out of 27, 88%) had abnormal T2 signal hyper intensity suggestive of cord edema. However, these changes are nonspecific in nature and can be seen in wide variety of spinal cord pathologies such as tumors, demyelinating lesions, or infectious etiologies.\[[@ref2]\] One of the patients was wrongly diagnosed as having demyelinating pathology before being diagnosed as SDAVF in our study. Interestingly, seven patients underwent MRA for diagnostic evaluation. MRA was suggestive of SDAVF in all the patients but was able to localize the site of fistula in only three patients. Thus, we suggest that a combination of MRI plus MRA could be a first line set of investigations in a patient suspected to have SDAVF.\[[@ref2]\] These tests have high sensitivity and are noninvasive, without radiation exposure or theoretical risk of contrast induced kidney damage, as compared with conventional digital subtraction angiogram.\[[@ref2][@ref5][@ref21]\] The primary principle in management of SDAVF is obliteration or removal of abnormal communicating vessel/s and to relieve spinal cord edema. Traditionally, surgical laminectomy and ligation of fistula is used as definitive treatment strategy. It is an open spinal surgical procedure with inherent complications of general anesthesia, cerebrospinal fluid (CSF) leak and risk of postoperative wound infection.\[[@ref1][@ref23][@ref6]\] There is an increasing interest in use of minimally invasive endovascular embolization techniques in past few years.\[[@ref4][@ref6]\] This treatment modality consists of advancing a micro catheter in the offending radicular artery and embolizing that artery using various materials. The most commonly used embolization particles are NBCA or onyx.\[[@ref4][@ref7][@ref14]\] The major complications of endovascular approach are recurrence or incomplete embolization and accidental damage to medullary spinal vessels leading to spinal cord infarction.\[[@ref7][@ref15]\] As seen from our data, the endovascular arm has higher incidence of recurrence rates. Recent advances in neuroanesthesia techniques such as awake procedures and close neuromonitering may help to reduce complication rates of both endovascular and surgical procedures.\[[@ref16]\] There is emerging data to support use of neuromonitering and provocative testing to minimize the complication rates.\[[@ref17][@ref18]\] In our observational study, we have compared the functional outcomes and complications following each of the treatment modalities. There is not much data about head to head comparison of these two modalities. The patient population in both groups is comparable in clinical severity (ALS scores) as well as other demographic characteristics \[[Table 3](#T3){ref-type="table"}\]. Both of the treatment modalities are effective in immediate post-operative period for symptomatic as well as functional benefit. (ALS score improved by about one point in each group.) There appears to be some correlation between the duration of symptoms before diagnosis and degree of improvement after surgical or endovascular treatment. As noted in [Table 2](#T2){ref-type="table"}, patients with longer duration of symptoms before being treated were unlikely to show significant change in ALS score after intervention. This underscores the importance of timely diagnosis and urgent intervention to treat the fistula. There was no significant difference in length of postprocedure hospital stay in patients who underwent embolization (3 days) as compared with those who underwent surgical ligation (4 days) \[[Table 3](#T3){ref-type="table"}\]. None of the patients in the surgical group had recurrence of fistula during the follow up period. Three patients (30%) in the embolization group had recurrence of fistula during follow up (at 1, 4, and 26 months after initial embolization). All of these patients underwent successful surgical ligation of fistula without any further recurrence. We tried to compare the recurrence rates with different embolization materials. Five patients underwent embolization with onyx and five with NBCA. Two patients in onyx group (40%) had recurrence of fistula, as compared with one patient in NBCA group (20%). While choice of embolization material (NBCA/onyx) may affect recurrence rates, the number of patients in our study is small and hence no definitive conclusions can be made. Further the choice of embolization material depends upon advances in endovascular materials, which may act as a confounder in predicting the complication rates. Regarding postprocedure complications, one patient in the embolization group had accidental embolization of medullar artery resulting in spinal cord infarction. None of the patients in surgical group had demonstrable ischemia of cord. Regarding postprocedure infections, two patients (12%) in surgical group developed local wound infections necessitating repeat inpatient admission. Patients in both groups were followed up for several months after procedure and showed sustained improvement in ALS even at follow up. Many patients, however, remained with neurological deficits and these results are summarized in [Table 3](#T3){ref-type="table"}. CONCLUSIONS {#sec1-5} =========== SDAVF is a rare acquired disorder of spinal cord. A combination of MRI plus MRA may offer a sensitive screening test in patients suspected to have SDAVF. DSA still remains the gold standard for diagnosis.\[[@ref1]\] Endovascular embolization and surgical ligation are effective treatment strategies and offer immediate improvement in clinical status, which is sustained at several months' follow up. Timely diagnosis and treatment is essential to maximize the chances for recovery. Our data shows that surgical ligation may offer permanent cure without any recurrence. Endovascular approach is associated with higher incidence of recurrence and has potential to cause catastrophic injury such as spinal cord infarct. We believe that further advances in micro catheter techniques and embolization materials may reduce the recurrence rates. Our findings underscore the fact that despite modern and surgical and interventional therapy that cured the fistula, many patients were left with residual neurological deficits. Available FREE in open access from: <http://www.surgicalneurologyint.com/text.asp?2014/5/1/7/125628> Commentary: Surgery or embolization for spinal dural fistula? Burrows Anthony M. 1 Lanzino Giuseppe 1 2 Department of Neurologic Surgery, Mayo Clinic and Mayo Clinic Foundation, Rochester MN, USA Department of Radiology, Mayo Clinic and Mayo Clinic Foundation, Rochester MN, USA E-mail: lanzino.giuseppe\@mayo.edu Surgical ligation of Type I spinal dural arteriovenous fistulas (SDAVFs) remains the gold standard in the treatment of these lesions. However, with advances in endovascular techniques, several authors have proposed embolization as a valid alternative. Embolization has the main advantage of being a less invasive procedure, which is particularly appealing in patients with SDAVFs who often have significant disability by the time they are correctly diagnosed. In this issue of *Surgical Neurology International*, a retrospective observational study of patients presenting with spinal dural arteriovenous fistulas (SDAVF) and seen at Duke University Hospital between 1993 and 2012 is presented. Primary outcome measures included Aminoff--Logue scores (ALS), complications rates, and recurrence rates. Comparison is made between open surgical and compared with endovascular treatment groups, though the authors point out that the small numbers of patients in the series preclude meaningful statistical comparison. Using a standardized outcome measurement for SDAVF treatment, the ALS, functional outcomes improved in both treatment arms, from 4.5 to 3.5 in the surgical cohort and from 4 to 3 in the endovascular cohort. Significant complications included a spinal cord infarction in the endovascular group and two wound infections requiring readmission in the surgical group. Recurrence was higher in the endovascular group with 3/10 patients requiring subsequent open surgical treatment while none of the open surgical arm required retreatment. This thoughtful analysis attempts to directly compare surgical and endovascular treatment of SDAVF by experienced microsurgical and endovascular practitioners, although it appears that the authors have used the two techniques in a complementary fashion and attempted embolization as first choice in older patients (mean age 65 years in the embolization group versus 57 years in the surgical group). Despite the obvious appeal, embolization for the treatment of SDAVFs has significant limitations. Akin to intracranial dural arteriovenous fistulas, the *sine qua non* condition to be fulfilled for the successful and permanent obliteration of the fistula is adequate penetration in the proximal portion of the draining vein and its complete obliteration. This is not always easy to achieve given the small size of the feeding pedicles involved, which preclude distal catheterization close to the fistulous connection. Moreover, because of the small size of the vessels involved, apparent immediate angiographic obliteration may not necessarily result in complete obliteration of the fistula as micro fistulous connections may persist while not being evident any longer on catheter angiography. This is a major limitation of embolization for SDAVFs as many of these patients are already neurologically impaired and need immediate complete and permanent obliteration of the fistula. In series presented, Onyx™(ev3, Irivne CA) was utilized in 5 of the 10 patients treated with embolization first while the remaining ones were treated (probably before Onyx™ introduction in practice) with n-Butyl cyanoacrylate (NBCA). Although some groups have reported good success with Onyx™ embolization for SDAVFs, there are major limitations associated with the use of Onyx for SDAVFs.\[[@ref2a]\] First, this agent may not achieve good enough penetration to assure complete obliteration of the proximal venous drainage.\[[@ref1a]\] Second, Onyx™ penetration into very small critical arterial pedicles, which may participate in the vascularization of the spinal cord, may not be immediately evident during the embolization (and this may have played a role in the major complication observed in this series). Overall, despite the appeal and definitive less invasiveness, embolization for SDAVFs continues to have important limitations, which make surgical ligation a more effective procedure in a significant portion of patients with these lesions.

Squamous cell carcinoma (SCC) or epidermoid carcinoma arises in stratified epithelia including skin, head and neck, oesophagus, or cervix but also in simple epithelia (25--30% of lung cancer; <http://www.cancer.org> <http://www.cancer.gov>). SCCs are often aggressive and have poor prognosis causing significant death numbers. Although clinics and treatment of SCCs vary, they are mostly caused by external mutagenic agents (UV light, smoking, alcohol). Cutaneous SCC is not frequently fatal (4% metastasize)^[@bib1]^ but is a paradigmatic and accessible model for the study of squamous cancer. The two types of cutaneous carcinoma: basal cell carcinoma (BCC; 80%) and SCC (20%) are the most common forms of cancer.^[@bib1],\ [@bib2],\ [@bib3],\ [@bib4]^ The comparison of BCC and SCC provides divergent lesions displaying different molecular alterations and paradoxical contrasting prognosis.^[@bib5]^ While SCCs retain a squamous differentiated phenotype except for very dysplastic lesions,^[@bib6]^ BCCs typically contain undifferentiated small cells reminiscent of proliferative basal keratinocytes although they display slow growth^[@bib7]^ and a very low metastatic rate (0.1%).^[@bib8]^ Why SCCs are more aggressive than skin BCCs, is a relevant clinical and biological question that remains unresolved. We have previously revealed a novel epidermal response that triggers squamous differentiation upon cell cycle deregulation and DNA damage.^[@bib9],\ [@bib10],\ [@bib11],\ [@bib12]^ This DNA damage response (DDR) suppresses cell division but allows extra rounds of DNA replication resulting in polyploidy. This involves mitotic bypass (without intervening mitosis), mitotic slippage (defined as failure to arrest in G2/M)^[@bib13]^ or acytokinetic mitosis.^[@bib12]^ Due to this process, cell cycle stress induced in human keratinocytes by proto-oncogene MYC, the DNA replication regulator Cyclin E, or by inactivation of tumour suppressor p53, does not result in cellular transformation, but in a G2/M block that in turn induces squamous differentiation and polyploidy.^[@bib9],\ [@bib10],\ [@bib14]^ We have proposed that this response constitutes an anti-oncogenic mechanism by which pre-cancerous cells are eliminated from the skin via squamous differentiation.^[@bib10],\ [@bib11]^ Consistently, in a large body of works on mouse skin differentiation prevailed upon overexpression of cell cycle or oncogenic molecules (for example, refs [@bib15], [@bib16], [@bib17] reviewed in ref. [@bib11]). Replication stress by causing DNA damage alerts cell cycle checkpoints of G2 and mitosis.^[@bib18],\ [@bib19]^ In keratinocytes, a mere impairment of the G2/M transitions induces massive terminal squamous differentiation and polyploidy and loss of proliferative potential within 48 h.^[@bib9],\ [@bib14],\ [@bib20]^ This includes the inhibition of mitotic kinases (Aurora Kinase B, Polo-like Kinase, cdk1) or microtubuli (Nocodazole, Taxol) and also S phase defects (Doxorubicin or Bleomycine). We hypothesised that the capacity of SCC cells to divide upon replication stress in spite of polyploidy might contribute to malignancy. This would require alterations in mitosis control.^[@bib10]^ Supporting this notion, overexpression of the global mitotic regulator FOXM1 drives the keratinocyte response to ectopic MYC or loss of p53 from increased differentiation to increased proliferation.^[@bib21]^ We have now investigated the alterations of SCC and BCC cells in their response to cell cycle stress. We find that SCC but not BCC cells retain a partial squamous differentiation response to replication stress induced by Cyclin E. Analyses of a panel of human biopsies of BCC, non-metastatic SCC and metastatic SCC reveal significant differences in the expression of Cyclin E and in levels of DNA damage. In addition, we report that mitotic stress induced by consecutive mitosis blocks drive SCC cells to accumulate DNA damage, to progressively lose the squamous phenotype, to gain mesenchymal features and to become more malignant *in vivo*. We propose a model by which alterations in the squamous differentiation DDR at mitosis contributes to genomic instability and malignant progression of squamous cancer. This might contribute explaining why SCC tends to be aggressive in opposition to the less differentiated BCC. Results ======= Cyclin E accumulates during squamous differentiation and when overexpressed in proliferative keratinocytes it promotes replication stress, DNA damage, mitotic defects and polyploidy.^[@bib9],\ [@bib20],\ [@bib22],\ [@bib23]^ To study this response in carcinoma cells we overexpressed a GFP form of Cyclin E in cells originated from a human facial BCC (BCCP;^[@bib24]^ [Supplementary Figure 1a](#sup1){ref-type="supplementary-material"}) or from a human facial well-differentiated SCC (SCC12F;^[@bib25]^ [Supplementary Figure 1a](#sup1){ref-type="supplementary-material"}). Cyclin E induced DNA replication both in BCCP and SCC12F as monitored by BrdU incorporation ([Figure 1a](#fig1){ref-type="fig"}; [Supplementary Figure 1b](#sup1){ref-type="supplementary-material"}). However, whereas in BCCP this resulted in a boost of S phase, in SCC12F it promoted polyploidy. Cyclin E induced mitotic Cyclin B in BCCP, but not in SCC12F due to mitotic slippage or bypass ([Figure 1b](#fig1){ref-type="fig"}).^[@bib9]^ As expected, Cyclin E also caused increase of the DNA damage marker *γ*H2AX both in BCCP and SCC12F ([Figure 1b](#fig1){ref-type="fig"}) due to replication stress.^[@bib9],\ [@bib22]^ BCCP display the mutated p53 antigen Pab240 (p53mut), whereas SCC12F are reported not to bear p53 mutations and did not express p53mut ([Supplementary Figure 1c](#sup1){ref-type="supplementary-material"}).^[@bib26]^ In line with inactivating mutations which render the protein resistant to degradation,^[@bib27]^ p53 was overexpressed in BCCP cells ([Figure 1b](#fig1){ref-type="fig"}; [Supplementary Figure 1c](#sup1){ref-type="supplementary-material"}). p53 transcriptional target p21CIP (p21) was barely detectable in proliferating BCCP *in vitro* ([Figure 1b](#fig1){ref-type="fig"}). p21 inhibits cdk2 and cell cycle entry or cdk1 and mitosis progression,^[@bib28]^ for instance in response to DNA damage. In keratinocytes, p21 is transiently induced and binds cdk1 in the onset of squamous differentiation.^[@bib9],\ [@bib29],\ [@bib30]^ Overexpression of Cyclin E in SCC12F cells caused a slight induction of p53 typical of DNA damage ([Figure 1b](#fig1){ref-type="fig"}).^[@bib9]^ However, p21 was high both in parental SCC12F cells and upon ectopic Cyclin E, as compared with normal keratinocytes ([Figure 1b](#fig1){ref-type="fig"}). p21 can be expressed independently of p53 and its deregulation in SCC12F might reflect cell cycle alterations. Overexpression of Cyclin E induced at some extent squamous differentiation in SCC12F, but not in BCCP, as measured by the squamous marker involucrin ([Figure 1c](#fig1){ref-type="fig"}; [Supplementary Figure 1d](#sup1){ref-type="supplementary-material"}). As we found no signs of apoptosis ([Figure 1a](#fig1){ref-type="fig"}; [Supplementary Figure 1b](#sup1){ref-type="supplementary-material"}), the induction of terminal differentiation is consistent with the significant loss of clonogenic potential of SCC12F-Cyclin E ([Figure 1d](#fig1){ref-type="fig"}). Growing SCC12F cells overexpressing Cyclin E after three passages continued to show higher DNA damage ([Supplementary Figure 1e](#sup1){ref-type="supplementary-material"}) and reduced clonogenic capacity than parental cells ([Supplementary Figure 1f](#sup1){ref-type="supplementary-material"}). The results above suggest that an excess of Cyclin E by inducing DNA damage and differentiation might be a burden to carcinoma cells. We studied the expression of Cyclin E and *γ*H2AX in a pilot collection of human BCC and SCC biopsies. 15 BCCs, 17 non-metastatic SCCs (NMSCCs) and 18 metastatic SCCs (MSCCs) were compared ([Supplementary Table 1](#sup1){ref-type="supplementary-material"}). Both Cyclin E and *γ*H2AX were barely detectable in BCC lesions, whereas they were prominent in SCC lesions ([Figures 2a and b](#fig2){ref-type="fig"}). *γ*H2AX in BCCs was detectable only in the growing front of the tumour ([Figure 2b](#fig2){ref-type="fig"}; [Supplementary Figure 2a](#sup1){ref-type="supplementary-material"}). Nuclear expression of Cyclin E was significantly lost in MSCCs as compared with NMSCC. In 40% of MSCCs Cyclin E was strongly accumulated in the cytoplasm (cCE; *P*\<0,01; [Figure 2c](#fig2){ref-type="fig"}; [Supplementary Figure 2b](#sup1){ref-type="supplementary-material"}). Although the number of cases is small, cCE was found only in one case out of 16 NMSCCs and none of 10 well-differentiated SCCs. While the mechanism exporting Cyclin E to the cytoplasm and its consequences are unclear, overexpression of a cytoplasmic shorter form of the protein accelerates mitosis (low molecular weight LMW-Cyclin E).^[@bib23],\ [@bib31]^ Interestingly, we detected a shorter protein in SCC12F overexpressing Cyclin E ([Figure 1b](#fig1){ref-type="fig"}). Reducing Cyclin E in the nucleus by sequestering cdk2 in the cytoplasm might reduce DNA replication-associated damage. To note, the DNA damage marker *γ*H2AX was strong in NMSCCs and significantly lost in poorly differentiated MSCCs ([Figure 2d](#fig2){ref-type="fig"}; [Supplementary Figure 2a](#sup1){ref-type="supplementary-material"}). Analyses of serial sections showed coincident Cyclin E and *γ*H2AX in cases where Cyclin E was strongly nuclear and marked loss of *γ*H2AX in cases where it was strongly cytoplasmic ([Supplementary Figure 3a](#sup1){ref-type="supplementary-material"}). Co-accumulation of Cyclin E and p21 reveals cell cycle conflict, occurs in the onset of initiation of keratinocyte differentiation ([Supplementary Figures 3b--d](#sup1){ref-type="supplementary-material"})^[@bib9],\ [@bib20],\ [@bib29],\ [@bib30]^ and was found in well differentiating SCCs displaying nuclear Cyclin E ([Supplementary Figures 3c,d](#sup1){ref-type="supplementary-material"}). p21 was strong in large nuclei or multinucleate cells (arrows in [Supplementary Figure 3d](#sup1){ref-type="supplementary-material"}). No general correlation was found between p21 or p53 and aggressiveness ([Supplementary Figure 4a](#sup1){ref-type="supplementary-material"}). However, deregulated p53 (maximum intensity in 100% of cells), suggestive of inactivating mutations, was found in 44% of cases with cCE *versus* only 13% with no cCE ([Supplementary Figure 4b](#sup1){ref-type="supplementary-material"}). The results above suggest that the axis squamous differentiation/Cyclin E via cell cycle stress might contribute to genomic instability in SCC. Consistently, MSCCs in the biopsy collection significantly displayed more chromosomal alterations than NMSCCs and these in turn more than BCCs ([Figure 3](#fig3){ref-type="fig"}). BCCs not showing signs of squamous differentiation, nor accumulation of Cyclin E, contained small and homogenous nuclei with two chromosomal copies. We investigated whether the capacity to escape the differentiation-associated cell division block in spite of genetic damage may cause genomic instability in SCC cells. To this end, we subjected BCCP and SCC12F cells to mitosis blocks by use of the microtubule-inhibitory drug Nocodazole (Nz), which triggers the squamous differentiation programme in human keratinocytes within 48 h^[@bib9],\ [@bib14]^ ([Supplementary Figure 5](#sup1){ref-type="supplementary-material"}). This response mimics differentiation induced by MYC, Cyclin E, loss of p53 or other inhibitors of mitosis.^[@bib9],\ [@bib10],\ [@bib20]^ A 24 h Nz treatment irreversibly suppressed the clonogenic capacity of normal keratinocytes ([Figure 4a](#fig4){ref-type="fig"}; [Supplementary Figure 5a](#sup1){ref-type="supplementary-material"}). However, SCC12F conserved some of the capacity to proliferate after the mitosis block and the clonogenic capacity of BBCP cells was barely affected ([Figure 4a](#fig4){ref-type="fig"}; [Supplementary Figure 5a](#sup1){ref-type="supplementary-material"}). This suggests that BCCP cells have a more robust G2 arrest and a tighter control of cell growth. Accordingly, while SCC12F cells strikingly increased in cell size upon the mitosis block (high light scattering typical of differentiated keratinocytes),^[@bib32]^ the size of BCCP cells changed very moderately ([Figures 4b and c](#fig4){ref-type="fig"}; [Supplementary Figures 5c and d](#sup1){ref-type="supplementary-material"}). In addition, SCC12F slipped into polyploidy at a greater extent than BCCP ([Figure 4c](#fig4){ref-type="fig"}; [Supplementary Figures 5c and d](#sup1){ref-type="supplementary-material"}). The changes in cellular size and ploidy in SCC12F were associated with an increase of squamous suprabasal markers (involucrin and keratin K16; [Figure 4d](#fig4){ref-type="fig"}; [Supplementary Figures 6a and b](#sup1){ref-type="supplementary-material"}), indicating that these cells conserve a partial differentiation response to mitotic stress. Differentiation likely accounts for the loss of clonogenicity observed, as no signs of apoptosis were found in the DNA content profiles ([Supplementary Figures 5 b-d](#sup1){ref-type="supplementary-material"}). In contrast, squamous markers were undetectable in BCC cells upon Nz treatment ([Supplementary Figure 6b](#sup1){ref-type="supplementary-material"}, not shown) in line with the absence of clonogenic loss. Proliferating SCC12F cells just after the Nz drug release were larger and a high proportion polyploid. Therefore these cells were able to proliferate in spite of being polyploid ([Supplementary Figures 7a and b](#sup1){ref-type="supplementary-material"}). However, after two passages subculture these cells (SCC12R1) displayed higher levels of p53, p21 and DNA damage marker *γ*H2AX and were more proliferative than the parental cells ([Supplementary Figures 7c--f](#sup1){ref-type="supplementary-material"}). p53 was overexpressed in SCC12R1 and the typical mutated conformation was now detected, although the cell morphology was unchanged. In order to test whether consecutive partial mitotic blocks might induce severe changes in SCC12F cells as it might occur *in vivo*, we subjected SCC12R1 cells to a second 48 h Nz treatment. Again, the mitotic block at first produced a high proportion of polyploid cells and significantly reduced their clonogenic potential (SCC12R1Nz; [Figure 5a](#fig5){ref-type="fig"}; [Supplementary Figure 8a](#sup1){ref-type="supplementary-material"}). However, cells stably growing two passages after the second block (SCC12R2) displayed a markedly different phenotype. They were remarkably homogenous in size, small and diploid and displayed a more fibroblastic morphology ([Figure 5a](#fig5){ref-type="fig"}). To study this phenomenon further we performed western blotting, immunofluorescence or RT-PCR for expression of keratin K5, often lost in aggressive carcinomas, keratin K8 and vimentin, often gained in aggressive carcinomas. Interestingly, SCC12R2 lost expression of epidermal keratin K5 and Cyclin E and strongly gained keratin K8 and vimentin with respect to untreated parental cells ([Figures 5b and c](#fig5){ref-type="fig"}; [Supplementary Figure 8b](#sup1){ref-type="supplementary-material"}). Consequently, SCC12R2 cells lost completely the expression of the typical squamous marker involucrin ([Figure 5b](#fig5){ref-type="fig"}). Similar phenotypic results were obtained on R2 cells in three independent experiments and the degree of the changes was proportional to the number of consecutive mitotic blocks applied (R1, R2, R3; [Supplementary Figure 9](#sup1){ref-type="supplementary-material"}). Altogether the results indicate that the phenotypic changes were not sporadic events but a consistent conversion produced by mitotic stress. The proliferative capacity of SCC12R2 cells was significantly enhanced compared with parental cells ([Figure 6a](#fig6){ref-type="fig"}), in spite of a neat accumulation of *γ*H2AX ([Figure 6b](#fig6){ref-type="fig"}). Western blot analyses confirmed the higher levels of *γ*H2AX in SCC12R2 ([Figure 6c](#fig6){ref-type="fig"}). The expression of the DNA repair factor p53-binding protein 1 (53BP) was low in BCCP cells and high in SCC12F cells ([Figure 6b](#fig6){ref-type="fig"}; [Supplementary Figure 8c](#sup1){ref-type="supplementary-material"}). Interestingly, SCC12FR2 cells displayed a strong nuclear spot of 53BP typical of isolated unrepaired DNA damage.^[@bib33]^ In addition, SCC12R2 strongly expressed the mutated p53 antigen Pab240 (p53mut; [Supplementary Figure 8b](#sup1){ref-type="supplementary-material"}) and strongly accumulated p53 ([Figure 6c](#fig6){ref-type="fig"}), indicative of inactivating mutations.^[@bib27]^ Proliferating SCC12R2 cells also lost p21, decreased the expression of Cyclin E and increased the expression of mitotic Cyclin B, resembling BCCP cells ([Figures 6b and c](#fig6){ref-type="fig"}). Interestingly, SCC12R2 accumulated Cyclin E in the cytoplasm. These changes were consistent with the loss of the squamous phenotype, nuclear Cyclin E and *γ*H2AX observed in aggressive MSCCs *in situ*. In order to test whether loss of Cyclin E function might contribute to the loss of the squamous phenotype, we inhibited the endogenous protein in well-differentiated parental SCC12F. To this aim, we made use of lentiviral constructs carrying specific shRNA to Cyclin E (shCE). Interestingly, shCE diminished the levels of *γ*H2AX, potentiated the expansion of SCC12F colonies, induced expression of keratin K8 and vimentin and produced a more fibroblastic morphology ([Figures 6d--f](#fig6){ref-type="fig"}; [Supplementary Figure 10](#sup1){ref-type="supplementary-material"}). To determine whether the double mitotic block rendered SCC cells more malignant, we injected SCC12R2 subcutaneously in nude mice. Parental SCC12F are scarcely tumorigenic.^[@bib25],\ [@bib26]^ As shown in [Figure 7a](#fig7){ref-type="fig"}, the parental cells developed very slowly growing benign tumours that were detected only after 10 weeks. In striking contrast, SCC12R2 cells rapidly generated detectable tumours after only 10 days. The size of the SCC12R2 lesions was also larger at term. Half of the mice injected with the parental cells never developed tumours after 16 weeks, whereas all mice injected with SCC12R2 cells had developed tumours after 4 weeks ([Figure 7b](#fig7){ref-type="fig"}). The histology showed that the tumours generated by parental cells were well-differentiated, whereas those generated by SCC12R2 cells were characterised as poorly differentiated (H/E; [Figure 7c](#fig7){ref-type="fig"}). SCC12R2 tumours were more proliferative (Ki67) and accumulated *γ*H2AX ([Figure 7c](#fig7){ref-type="fig"}). Consistently, these rapidly growing tumours displayed loss of epidermal keratins K5/K10 and E-cadherin and gain of keratins K13, K8 and vimentin ([Figure 7d](#fig7){ref-type="fig"}; [Supplementary Figure 11](#sup1){ref-type="supplementary-material"}), changes typical of aggressive stages of skin carcinogenesis.^[@bib34],\ [@bib35]^ The abundance of *γ*H2AX and the absence of lung metastasis (not shown) suggest that they are in an early stage of invasive conversion. Discussion ========== We hypothesised that alterations in the link between mitosis control and squamous differentiation might contribute to carcinogenesis.^[@bib10]^ It is paradoxical that BCC of the skin is very rarely invasive in spite of losing the squamous phenotype. In our study SCC, not BCC cells, responded to cell cycle stress by initiating squamous differentiation. Cell cycle deregulation/replication stress induces DNA damage and G2/M arrest.^[@bib18],\ [@bib19]^ Moreover, mitotic arrest is sufficient to cause DNA damage^[@bib36]^ and mitotic slippage/bypass can result in chromosomal defects.^[@bib10]^ Moreover, polyploidy often leads to aneuploidy when cells are able to divide.^[@bib37]^ We propose that the reversibility of the cell division block imposed by the initiation of squamous differentiation contributes to genomic instability and malignant progression in SCCs but not in BCCs ([Figure 8](#fig8){ref-type="fig"}). Normal keratinocytes that initiate differentiation accumulate high levels of DNA damage marker *γ*H2AX,^[@bib10]^ as we found in NMSCC cells. In contrast, MSCCs lost nuclear Cyclin E and the *γ*H2AX signal. Accumulation and coexpression of Cyclin E and *γ*H2AX in NMSCCs and their coincident nuclear loss in MSCC suggests that high levels of nuclear Cyclin E via replication stress^[@bib9],\ [@bib22]^ is a burden to malignant carcinoma cells. While normal keratinocytes differentiate terminally in response to cell cycle stress,^[@bib10],\ [@bib11]^ damaged SCC cells that are able to divide are genetically instable. This would increase the probability of more aggressive clones to appear that would be selected for ([Figure 8](#fig8){ref-type="fig"}). Interestingly, genomic instability has been shown to promote evolutionary adaptation.^[@bib38]^ In our study simply overexpressing Cyclin E was insufficient to drive malignant transformation. However, forcing cell division by FOXM1 allows damaged normal keratinocytes with deregulated MYC or p53 to amplify.^[@bib21]^ FOXM1 is a global mitotic regulator and this suggests that mitosis control is a limiting factor in keratinocyte transformation. Squamous cancer progression might require mitotic alterations. Remarkably, in our study two consecutive mitotic blocks were sufficient to render well-differentiated SCC cells highly tumorigenic. These cells initiated an epithelial--mesenchymal conversion typical of invasive SCCs^[@bib34],\ [@bib35],\ [@bib39]^ and proportional to the number of consecutive mitotic blocks (SCC12R1, R2, R3). The phenotypic conversion of SCC12R2 cells, far more aggressive than the parental cells, shows that mitotic stress can contribute to squamous malignant progression. Paradoxically, the SCC12FR2 cells shared features with BCCP cells. As in carcinoma biopsies, both lines displayed low levels of *γ*H2AX. The critical difference might be the degree of genetic alterations caused by genomic instability. Accordingly, MSCCs displayed a higher degree of chromosomal alterations and SCC12R2 cells strongly displayed spots of 53BP, a marker of persistent unrepaired DNA damage.^[@bib33]^ In contrast, no spots 53BP were observed in BCCP cells. The higher stability of the BCC genome might be due to the lack of the squamous pathway and, a more robust G2 arrest and a tighter control of cellular growth. In contrast, SCC12F cells displayed a loose mitotic control. Interestingly, mutations in the Sonic hedgehog (Hh) pathway are frequent in BCC, suppress squamous differentiation in mouse^[@bib40]^ and cause evasion of G2/M checkpoints.^[@bib41]^ The robustness of the G2/M arrest in BCC cells might allow more efficient DNA repair and maintenance of genomic stability. Consistently, *γ*H2AX in BCC biopsies was scarce while it was detected in the tumour growing front and in proliferating BCCP *in vitro*. The epithelial--mesenchymal conversion has recently been associated with loss of *γ*H2AX.^[@bib42]^ The axis squamous differentiation/high Cyclin E constitutes a mitotic barrier^[@bib11]^ and we argue that the most aggressive SCC cells lose the squamous phenotype in order to avoid it. This is well supported by the loss, or cytoplasmic accumulation of Cyclin E (cCE) that we found in metastatic SCCs in our pilot study and in SCC12R2 *in vitro*. Recent works on large biopsy collections also have shown association between cytoplasmic shorter forms of Cyclin E and poor breast cancer prognosis,^[@bib23],\ [@bib31],\ [@bib43]^ suggesting a growth advantage to cancer cells. Although cCE can cause genomic instability, it accelerates mitosis.^[@bib23],\ [@bib31]^ In our study, the loss of *γ*H2AX was especially marked in cases displaying cCE. By reducing the levels of Cyclin E in the nucleus, malignant squamous cells might in part avoid severe deregulation of DNA replication-S phase. Functionally supporting this model, in our study inhibition of Cyclin E in well-differentiated SCC12F reduced DNA damage and enhanced proliferation and expression of mesenchymal markers. Loss of p53, *The Guardian of the Genome,* leads to polyploidy in a variety of cell types.^[@bib44]^ In keratinocytes this loss induces polyploidy and squamous differentiation.^[@bib10]^ The responses of the carcinoma cells studied here do not seem to be mediated by p53: (i) SCC12F cells seemingly bearing intact p53 become polyploid upon Nocodazole; (ii) BCCP displaying mutated p53 were able to efficiently control G2/M and ploidy; (iii) SCC12R2 cells overexpressing mutated p53 displayed no signs of polyploidy. In addition, the levels of p53 in the human biopsies did not indicate a general association with aggressiveness. However, we detected a potential association between cCE and deregulation of p53. In addition, the cell cycle inhibitor p21, target of p53, stayed high in MSCC. Interestingly, Galanos *et al.* now reports a role of chronic and p53-independent expression of p21 in promoting genomic instability through replication stress in carcinomas of lung of head and neck.^[@bib45]^ Moreover, the deregulation of DNA replication licensing protein cdc6 contributes to features of epithelial--mesenchymal transition^[@bib46]^ and deregulated Cyclin E was found to affect licensing.^[@bib47]^ In summary, our model is that the DNA damage-squamous differentiation pathway constitutes first a barrier to undesired proliferation, second a source of genomic instability, thereby driving malignant progression of genetically damaged cells that are able to divide ([Figure 8](#fig8){ref-type="fig"}). The loss of detectable nuclear Cyclin E and *γ*H2AX in MSCCs in the pilot series of biopsies studied was highly significant. These results, together with the cytoplasmic accumulation of Cyclin E, should encourage further studies on larger cohorts of squamous carcinomas. The findings might have application into squamous cancer in locations other than skin, as a growing body of evidence suggests that they might share common mechanisms. It has been proposed that alterations in ploidy contribute to cancer malignancy.^[@bib37]^ Our observations would indicate that squamous cancer cells become malignant not because they are polyploid, but because they are capable to divide in spite of being so. Materials and methods ===================== Cell culture, human biological samples and viral infections ----------------------------------------------------------- Ethical permission for this study was requested, approved, and obtained from the Ethical Committee for Clinical Research of Cantabria Council, Spain. In all cases of primary cell culture, human tissue material discarded after surgery was obtained with written consent presented by clinicians to the patients and it was treated anonymously. The Basal carcinoma cell line (BCCP^[@bib24]^), isolated from a human facial BCC, was kindly provided by Dr. R. Polakowska (Institut pour la Recherche sur le Cancer de Lille \[IRCL\], Lille Cedex, France). The SCC12F line was cultured from a human facial SCC.^[@bib25]^ SCC12B originated from a more aggressive component of the same carcinoma was also analysed. Primary normal keratinocytes (NK) were isolated from neonatal human foreskin. All cells were cultured in presence of a mouse fibroblast feeder layer (inactivated by mitomycin C) in Rheinwald FAD medium as described previously (10% serum and 1.2 mM Ca^+2^).^[@bib48]^ All cell lines and primary normal keratinocytes used were tested for mycoplasma contamination. Cells were treated with Nocodazole for 24 h (Nz; 20 *μ*M; Sigma-Aldrich, St. Louis, MO, USA).^[@bib49]^ Primary normal keratinocytes were treated 24 h with doxorubicin (0.5 *μ*M). Parallel control cultures were always subjected to the DMSO vehicle only. Fresh medium was added 24 h before addition of Nz and again 24 h after. BCCP or SCC12F were subjected to retroviral infection with pBabe-GFP (GFP) and pBabe-GFP-Cyclin E (CEGFP) constructs. SCC12F also were subjected to lentiviral infection as described (see also [Supplementary Materials and Methods](#sup1){ref-type="supplementary-material"})^[@bib9],\ [@bib10]^ with MISSION Sigma-Aldrich plasmids: control (pLKO.1, Ctr) and plKO.1 with a shRNA against Cyclin E (TRCN0000045300, shCE). The studies were carried out either by analysing unselected pools 4 days after retroviral infections or by stably selecting cells expressing retroviral or lentiviral constructions by 1*μ*g/ml puromycin. Clonogenicity assays were made as described previously (see also [Supplementary Materials and Methods](#sup1){ref-type="supplementary-material"}).^[@bib10]^ 50 non-melanoma skin cancer lesions were included in the study ([Supplementary Table 1](#sup1){ref-type="supplementary-material"}): (i) 15 basal cell carcinomas; (ii) 18 primary metastatic squamous cell carcinomas (MSCCs) that had evolved to (histologically confirmed) lymph node metastases (2 well differentiated; 11 moderately differentiated; 5 undifferentiated) and (iii) 17 patients with SCC who had not developed any metastasis (non-metastatic, NMSCCs) in a 5-year follow-up period (8 well differentiated; 9 moderately differentiated; 0 undifferentiated). A control group of 10 samples of elastotic non-tumoral skin was also included in the study. For details of sample collection and characterisation see [Supplementary Materials and Methods](#sup1){ref-type="supplementary-material"}. After histopathological evaluation the invasive edge of the tumour was selected for the construction of tissue microarray (TMAs).^[@bib50]^ Two tissue cylinders with a diameter of 2 mm were punched from the selected areas from each tissue block and brought into a recipient paraffin block using the tissue micro-arrayer (Arrayer Punch set 2.00 mm, ATA200, Advanced Tissue Arrayer, Chemicon International). Antibodies ---------- Primary and secondary antibodies utilised in this study are listed in [Supplementary Materials and Methods](#sup1){ref-type="supplementary-material"}. Flow-cytometry -------------- Cells were harvested, fixed and stained as previously described for DNA synthesis and content (BrdU incorporation and propidium iodide; see [Supplementary Materials and Methods](#sup1){ref-type="supplementary-material"}).^[@bib9],\ [@bib14]^ Histology and immunostaining ---------------------------- For immunofluorescence, cells were grown on glass coverslips, fixed, and stained as previously described.^[@bib9]^ H&E, immunofluorescence and immunohistochemical stainings of carcinoma sections of paraffin embedded formalin fixed tissues were performed on 4*μ*m thick sections. For more details see [Supplementary Materials and Methods](#sup1){ref-type="supplementary-material"}. The score of immunohistochemical (IHC) stainings was given between 0 and 100% of positively stained neoplastic cells and the intensity was measured as: 1 (weak), 2 (medium), 3 (strong). The whole punch (2 mm in diameter, 3.14 mm^2^) was evaluated using a 10 × objective lens and 10 × ocular lens. The histoscore was calculated by multiplying the percent of positive cells by the intensity (from 0 to 3) to give numbers ranging from 0 to 300. p53 was considered deregulated by inactivating mutations when the intensity of the staining was 3 (maximum) in 100% of cells (histoscore 300). Histograms were then plotted using Test of Kruskal--Wallis in combination with Mann--Whitney U test. Immunohistochemical staining was independently evaluated by two observers (A. Toll and D. López). For analyses of protein expression, cells were washed with PBS, lysed and subjected to SDS-PAGE electrophoresis and western blotting as described.^[@bib9]^ *Insoluble* protein fractions were incubated in Urea lysis buffer (10 mM Tris pH 8, 5% SDS, 5% *β*-mercaptoethanol, 4 M Urea). The whole original blots are shown in [Supplementary Figure 12](#sup1){ref-type="supplementary-material"}. RT-PCR ------ Total RNA was isolated using NucleoSpin RNA (Macherey-Nagel, Düren, Germany) and reverse transcribed with the iScript cDNA synthesis kit (Bio-Rad, Hercules, CA, USA).^[@bib10]^ The cDNAs (1 *μ*l) were amplified by real-time PCR (Bio-Rad iQ SYBR green supermix) and normalised to *β*-actin mRNA levels.^[@bib10]^ Primers utilised in this study are listed in [Supplementary Materials and Methods](#sup1){ref-type="supplementary-material"}. Fluorescence *in situ* hybridisation ------------------------------------ To evaluate genomic instability, fluorescence *in situ* hybridisation (FISH) with a specific probe against EGFR was also performed. Dual-colour hybridisation with fluorescent DNA for the centromeric region of chromosome 7 (CEP7, green) and for the specific DNA region for EGFR (7p12, red) was performed (Abbott Molecular, Abbot Park, IL, USA). One hundred nuclei per case were scored to determine the percent of epithelial cells with EGFR gains (three or more signals for EGFR). FISH were evaluated by two observers (A. Toll and D. López). Tumorigenesis ------------- Experiments using animals were performed in compliance with the United Kingdom Coordinating Committee on Cancer Prevention Research\'s Guidelines for the Welfare of Animals in Experimental Neoplasia, and authorised by the Consejería de Medioambiente y Ordenación del Territorio de la Comunidad de Madrid. Further details on mice conditions in [Supplementary Materials and Methods](#sup1){ref-type="supplementary-material"} Keratinocytes were tripsinized and resuspended in a mixture (2:1) of PBS and Matrigel (BD Biosciences, San Jose CA, USA). A volume of 150 *μ*l of this suspension containing 1 × 10^6^ cells was subcutaneously inoculated into the right flank of each mouse. Tumour width (*W*) and length (*L*) were measured twice a week using an external caliper. Tumour volume was calculated using the formula 0.5 × *L* × *W*^2^ (ref. [@bib51]). Statistical analyses -------------------- For cell culture experiments standard deviation and variance were calculated and served as estimates of variation within each group of data. For statistical comparison of groups with similar variance, a homoscedastic *t*-test was performed. For statistical comparison of groups with diverging variance, a heteroscedastic *t*-test was applied. Data sets were compared using an unpaired Student's *t*-test. A *P*-value of less than 0.05 was considered statistically significant. For histology and FISH statistical analyses were performed using Windows Statistical Package for Social Sciences version 17 (SPSS, Chicago, IL, USA). The non-parametric Mann--Whitney U test was used to compare the histoscore of different immunohistochemical markers and EGFR gains by FISH. For contingency tables, the Fisher exact test was used to assess the level of significance. In all cases, a 2-tailed *P*\<0.05 was required for statistical significance. Data were plotted and analysed using Test of Kruskal--Wallis in combination with Mann--Whitney U test. AG is grateful to Jean-Jeaques Guilhou, Jean-Claude Rossi and the INSERM for professional support and to Renata Polakowska for the generous gift of precious BCCP cell line. We thank Lucía Barbier, Tania Lobato, Evelyn Andrades, Alicia Noriega and María Aramburu for technical assistance and Natalia Sanz for critical reading of the MS. To AG: National grants from Instituto de Salud Carlos III, Fondo de Investigación Sanitaria (ISCIII-FIS/FEDER, Spain): PI08/0890, PI11/02070, PI14/00900; Ligue Nationale Contre la Cancer (La Ligue; France). To AT: ISCIII-FIS PI10/00785. To JP: MINECO grant SAF2015-66015-R; AES grant ISCIII-RETIC RD12/0036/0009. VC was funded by a fellowship from La Ligue (France), PA by IDIVAL (Spain), RM and IdP by AG lab and ISCIII-FIS-FEDER PI11/02070 (Spain). [Supplementary Information](#sup1){ref-type="supplementary-material"} accompanies this paper on Cell Death and Disease website (http://www.nature.com/cddis) Edited by R Aqeilan The authors declare no conflict of interest. Supplementary Material {#sup1} ====================== ###### Click here for additional data file. ###### Click here for additional data file. {ref-type="supplementary-material"}. (**b**) Detection by western blot of CEGFP, Cyclin E (CE), Cyclin B (CB), *γ*H2AX, p53, or p21 in normal keratinocytes (NK) untreated (Ctr) or treated 24 h with doxorubicin (DOXO); or in BCC and SSC12F cells Ctr or overexpressing CEGFP. GAPDH is loading control. (**c**) Top: detection by western blot of involucrin in insoluble protein extracts (Invol); same number of cells per lane. Uncropped blots are shown in [Supplementary Figure 12](#sup1){ref-type="supplementary-material"}. Bottom: percent of SCC12F-CEGFP positive cells for Invol relative to SCC12F-GFP (as determined by immunofluorescence; [Supplementary Figure 1d](#sup1){ref-type="supplementary-material"}). (**d**) Top: clonogenic capacity of SCC12F-GFP and SCC12F-CEGFP, first passage after infection (1000 cells per well, wells are representative of triplicate samples; blue is the fibroblast feeder layer, pink the carcinoma cells); bottom: percent of proliferative colonies of SCC12F-CEGFP relative to SCC12F-GFP. Error bars are s.e.m. of duplicate or triplicate samples of at least two independent representative experiments. \**P*\<0.05 and \*\**P*\<0.01](cddis2017259f1){#fig1} {ref-type="supplementary-material"}. Histograms show histoscore values for CE (**a**) or *γ*H2AX (**b**) in the series of BCCs (*n*=15) and SCCs (*n*=35). Scale bars, 200 μm. (**c**) Left histogram: histoscore values for CE in NMSSC (*n*=17) and metastatic SCC (MSCC; *n*=18). Right: bar histogram shows percent of cases displaying nuclear or cytoplasmic localisation of CE (nuclear or cytoplasmic) within NMSCCs (light grey) or MSCCs (dark grey). See also [Supplementary Figure 2b](#sup1){ref-type="supplementary-material"}. (**d**) Histoscore values of *γ*H2AX in NMSCC and MSCC (left) or in SCC well differentiated (Diff; *n*=10), moderately differentiated (Mod; *n*=20) and poorly differentiated (Poor; *n*=5) (right; see also [Supplementary Figure 2a](#sup1){ref-type="supplementary-material"}). Data plotted by tests Kruskal-Wallis/Mann-Whitney U. \**P*\<0.05 and \*\**P*\<0.01](cddis2017259f2){#fig2} {ref-type="fig"}). Data plotted by tests Kruskal-Wallis/Mann-Whitney U. \*\**P*\<0.01](cddis2017259f3){#fig3} {ref-type="supplementary-material"}). (**b**) Representative 3D histograms of light scattering (SCC: Side Scatter; FSC: Forward Scatter; cell size and complexity; white arrow) in BCCP and SCC12F untreated (Ctr) or treated with Nz for 48 h. (**c**) Percent of cells with large size and complexity (High Scatter) or polyploidy (Polyp, DNA content \>4N), as measured by flow-cytometry in BCCP (light grey) or SCC12F (dark grey) treated with Nz for 48 h or untreated (Ctr) as indicated. See also [Supplementary Figures 5b--d](#sup1){ref-type="supplementary-material"}. (**d**) Percent of SCC12F cells treated with Nz for 48 h that express the differentiation markers involucrin (Invol) or keratin K16, relative to untreated cells, as measured by flow-cytometry. Error bars are s.e.m. of duplicate or triplicate samples of at least two independent representative experiments.\**P*\<0.05 and \*\**P*\<0.01. See also [Supplementary Figure 6](#sup1){ref-type="supplementary-material"}](cddis2017259f4){#fig4} {ref-type="supplementary-material"}. Invol lanes for Ctr are the same as in [Figure 1c](#fig1){ref-type="fig"}. Uncropped blots are shown in [Supplementary Figure 12](#sup1){ref-type="supplementary-material"}. (**c**) Quantitation of the expression of keratin K5 or vimentin (Vm) in SCC12F and SCC12R2 by real-time (RT)-PCR. See also [Supplementary Figure 8b](#sup1){ref-type="supplementary-material"}. Error bars are s.e.m. of duplicate samples of at least two independent representative experiments. \**P*\<0.05, \*\**P*\<0.01](cddis2017259f5){#fig5} {ref-type="fig"}. (**d**) Clonogenic capacity of SCC12F infected with the empty vector (Ctr) or with specific shRNA to Cyclin E (shCE; 2 500 cells per well). (**e**) Bar histogram shows the number of proliferative colonies in (**d**). Error bars are s.e.m. of triplicate samples. \**P*\<0.05. (**f**) Western blotting for expression of keratin K5, keratin K8 in Ctr and shCE. See also [Supplementary Figure 10](#sup1){ref-type="supplementary-material"}. Uncropped images of blots in **c** and **f** are shown in [Supplementary Figure 12](#sup1){ref-type="supplementary-material"}. In (**a** and **d**) wells are representative of triplicate samples](cddis2017259f6){#fig6} {ref-type="supplementary-material"}. DAPI for DNA in blue. Scale bars, 200 *μ*m](cddis2017259f7){#fig7} {#fig8} [^1]: These authors contributed equally to this work.

Sake F‐T‐N, Wong K, Bartlett DJ, Saini B. Benzodiazepine usage and patient preference for alternative therapies: A descriptive study. Health Sci Rep. 2019;2:e116 10.1002/hsr2.116 1. INTRODUCTION {#hsr2116-sec-0005} =============== Benzodiazepines are widely prescribed for their hypnotic,[1](#hsr2116-bib-0001){ref-type="ref"} anxiolytic,[2](#hsr2116-bib-0002){ref-type="ref"} muscle relaxant, and antiepileptic indications.[3](#hsr2116-bib-0003){ref-type="ref"} Despite their common use, concerns remain over their long‐term safety. Pharmacoepidemiological data indicate that prolonged use of benzodiazepines is associated with an increased risk of falling,[4](#hsr2116-bib-0004){ref-type="ref"}, [5](#hsr2116-bib-0005){ref-type="ref"} therapeutic dose dependence,[6](#hsr2116-bib-0006){ref-type="ref"} and an increased incidence of dementia.[7](#hsr2116-bib-0007){ref-type="ref"} Benzodiazepine exposure is also associated with an increased risk of physical disability.[8](#hsr2116-bib-0008){ref-type="ref"} Recent cohort studies suggest an increased risk of exacerbations in benzodiazepine users with asthma and a higher likelihood of episodes of pneumonia and related mortality.[9](#hsr2116-bib-0009){ref-type="ref"}, [10](#hsr2116-bib-0010){ref-type="ref"} A recent systemic review also highlighted an overall increased risk of all‐cause mortality in benzodiazepine users.[11](#hsr2116-bib-0011){ref-type="ref"} Inappropriate use of benzodiazepines in Australia has been recently reported to lead to high costs of managing adverse events in residential aged care facilities.[12](#hsr2116-bib-0012){ref-type="ref"} In light of the well‐established side effect profile of benzodiazepines, prescribing guidelines do not recommend their prolonged use. The Royal Australian College of General Practitioners (RACGP) recommends that benzodiazepines should not be prescribed for longer than 4 weeks.[13](#hsr2116-bib-0013){ref-type="ref"} For patients requiring ongoing treatment, behavioural therapies such as cognitive behavioural therapy offer a promising alternative to the benzodiazepine for several conditions. For example, mounting evidence suggests that behavioural treatments produce comparable efficacy with benzodiazepines and have a longer‐lasting effect in patients with insomnia.[14](#hsr2116-bib-0014){ref-type="ref"}, [15](#hsr2116-bib-0015){ref-type="ref"}, [16](#hsr2116-bib-0016){ref-type="ref"} Similarly, research evidence highlights better or equivalent efficacy profiles for behavioural therapies over benzodiazepines for managing anxiety disorder, panic disorder, and dysthymia.[17](#hsr2116-bib-0017){ref-type="ref"} The benefits of de‐prescribing benzodiazepines have been highlighted in several studies. For example, in elderly nursing home residents, benzodiazepine withdrawal significantly improved memory and cognitive functioning compared with those who continued to take benzodiazepines; withdrawal did not give rise to anxiety, agitation, or sleeplessness.[18](#hsr2116-bib-0018){ref-type="ref"} Health care utilization and hospital admission costs resulting from traffic accidents and falls attributable to benzodiazepine use can be reduced through successful discontinuation[19](#hsr2116-bib-0019){ref-type="ref"}; thus, benzodiazepine discontinuation can also have an economic benefit. Despite the prescribing guidelines, the listed adverse effects of prolonged benzodiazepine use and robust evidence supporting behavioural therapies over benzodiazepines, the long‐term use of benzodiazepines remains high. While the pattern of benzodiazepine prescribing varies globally, prolonged use of benzodiazepines is a common phenomenon in the United Kingdom, Europe, and North America.[20](#hsr2116-bib-0020){ref-type="ref"} A US study reported that in 2008, approximately 5.2% of US adults (18‐80 years) had used benzodiazepines.[21](#hsr2116-bib-0021){ref-type="ref"} Reports comparing data on benzodiazepine use from the US National Health and Nutrition Examination Survey (NHANES) between 1999 and 2014 highlight significantly increased use (2.0% of respondents in 1999‐2000 to 4.2% in 2013‐2014), mostly driven by medium to long‐term users.[22](#hsr2116-bib-0022){ref-type="ref"} Although there has been a modest decline in the volume of benzodiazepine prescriptions dispensed annually, overall, there is still a high level of long‐term benzodiazepine use in Australia.[23](#hsr2116-bib-0023){ref-type="ref"}, [24](#hsr2116-bib-0024){ref-type="ref"}, [25](#hsr2116-bib-0025){ref-type="ref"} For example, in the case of insomnia, almost 80% of patient presentations result in a prescription for benzodiazepine derivatives (such as temazepam, oxazepam, diazepam, and nitrazepam), and this figure has remained stable over the past 10 years.[26](#hsr2116-bib-0026){ref-type="ref"}, [27](#hsr2116-bib-0027){ref-type="ref"} Both patient‐ and prescriber‐related factors appear to be linked with prolonged benzodiazepine use. Prescribers\' perceptions and attitudes have been reported to be a key factor leading to prolonged benzodiazepine use.[28](#hsr2116-bib-0028){ref-type="ref"} Recently, an Australian study highlighted that physicians often believe that patients taking benzodiazepines are unlikely to be willing to withdraw their medication and, thus, renew prescriptions without offering discontinuation or withdrawal plans.[29](#hsr2116-bib-0029){ref-type="ref"} Individual patient factors can also affect the length of benzodiazepine use; data indicate that being older, lonely, less educated, as well as having a poorer mental health profile and lower perceived general health status are associated with prolonged benzodiazepine use.[30](#hsr2116-bib-0030){ref-type="ref"}, [31](#hsr2116-bib-0031){ref-type="ref"} Other patient‐related factors that can influence ongoing use of benzodiazepines include beliefs and attitudes of the users towards their medication.[32](#hsr2116-bib-0032){ref-type="ref"} This has been proposed based on the well‐known Health Belief Model (HBM), which suggests that patients\' beliefs about their health issues, perceived benefits of and barriers to action, and self‐efficacy explain engagement in health‐promoting behaviours while the actual action is triggered by cues.[33](#hsr2116-bib-0033){ref-type="ref"} In fact, instruments such as the Beliefs about Medicines Questionnaire‐specific version (BMQ‐specific) have been designed and validated to assess patients\' beliefs and attitudes about their medications.[34](#hsr2116-bib-0034){ref-type="ref"} The BMQ‐specific consists of two 5‐item scales themed as (1) *necessity* and (2) *concerns*. While the necessity theme assesses individuals\' beliefs about the necessity of using their medication, the concerns theme evaluates their concerns about the medication (eg, side effects, fear of dependence, and among others). Thus, in the case of benzodiazepine use, it may be hypothesised that the balance between the concerns about the risk related to continued use of benzodiazepines versus the beliefs about the necessity of the medication may dictate a patients\' past, current, and future decisions related to their medications, eg, attempting withdrawal, seeking safer alternatives, or continuing use. Gauging the patients\' position on these opposing belief sets (necessity vs concerns) can allow health professionals to employ targeted information provision or counselling to enhance effective withdrawal. Where medication discontinuation is desirable, based on the reasoning behind the BMQ‐specific, beliefs that foster medication discontinuation (higher‐risk perception about the medication and a lesser perceived necessity for use) could serve as a triggering point at which patients can be empowered and pragmatic strategies offered to facilitate withdrawal. In support of this notion, a qualitative study in patients with anxiety indicated that strong beliefs around the necessity of using benzodiazepines and a lower level of concern about long‐term use were associated with resistance to benzodiazepine discontinuation.[35](#hsr2116-bib-0035){ref-type="ref"} Similarly, another qualitative study reported that individuals with higher concerns about the risk of taking benzodiazepines were more likely to attempt ceasing the medication of their own volition and were often more interested in trying behavioural interventions.[36](#hsr2116-bib-0036){ref-type="ref"} Exploratory work mapping such beliefs about benzodiazepines and intention to continue use is clearly required for successful discontinuation trials. Although there has been considerable research to characterize such individual sociodemographic and belief factors, there are, to date, few real‐life studies available on patients\' perceptions around benzodiazepine use and withdrawal. Minimal attention has been paid to the complexity of factors affecting the decision making involved in withdrawing benzodiazepines. Successful and sustained withdrawal behaviours require a prior intention, and past behaviour may contribute to this future intention.[37](#hsr2116-bib-0037){ref-type="ref"} In most research reporting trials of benzodiazepine weaning off, the patients\' willingness to withdraw benzodiazepines and intention or readiness to adopt alternative therapies are assumed but not explored.[38](#hsr2116-bib-0038){ref-type="ref"}, [39](#hsr2116-bib-0039){ref-type="ref"}, [40](#hsr2116-bib-0040){ref-type="ref"} It is not surprising that in many of these reported trials, benzodiazepine discontinuation rates remain less than 65%, with low‐response and high‐dropout rates.[38](#hsr2116-bib-0038){ref-type="ref"}, [39](#hsr2116-bib-0039){ref-type="ref"}, [40](#hsr2116-bib-0040){ref-type="ref"} Patients\' willingness to stop the medication and try behavioural therapies might be a key factor that affects successful discontinuation. Therefore, the present study primarily aimed to investigate the association of participants\' sociodemographic variables and beliefs about the benzodiazepine, with their benzodiazepine use behaviours, including ***past*** (previous benzodiazepine withdrawal attempts), ***current*** (long‐term benzodiazepine use, ie, using benzodiazepines for at least 1 year), as well as ***future*** behaviours (willingness to trial behavioural substitutes in the future). Additionally, the study aimed to explore the patients\' source preference for information about alternative behavioural therapies. The findings of this study will potentially inform effective, patient‐centred, and evidence‐based interventions for reducing long‐term benzodiazepine use and disseminating behavioural therapies with greater scope and sustainability. 2. METHODS {#hsr2116-sec-0006} ========== 2.1. Study design {#hsr2116-sec-0007} ----------------- Given that this study aimed to explore real‐life behaviour, a point of purchase survey of benzodiazepine consumers was used. This method allows researchers to study consumer behaviour for informing interventions and policy development.[41](#hsr2116-bib-0041){ref-type="ref"}, [42](#hsr2116-bib-0042){ref-type="ref"} The survey questionnaire was developed by reviewing relevant literature,[35](#hsr2116-bib-0035){ref-type="ref"}, [43](#hsr2116-bib-0043){ref-type="ref"}, [44](#hsr2116-bib-0044){ref-type="ref"} and the survey items included questions related to participants\' demographic characteristics as well as their perspectives about the chronic use of benzodiazepines and alternative therapies. The developed questionnaire was reviewed by psychology, pharmacy practice, and sleep researchers for relevance and wording of the questionnaire items. A validated 10‐item BMQ‐specific was also used to assess participants\' personal views about their benzodiazepines (Figure [1](#hsr2116-fig-0001){ref-type="fig"}).[34](#hsr2116-bib-0034){ref-type="ref"} Figure [2](#hsr2116-fig-0002){ref-type="fig"} outlines the diagrammatic representation and flow of questions used in the survey. This study was approved by The University of Sydney Human Research Ethics Committee (HREC: 2014/1020). {ref-type="ref"} questionnaire items and scoring](HSR2-2-e116-g001){#hsr2116-fig-0001} {#hsr2116-fig-0002} 2.2. Participants {#hsr2116-sec-0008} ----------------- Pharmacies throughout the Australian state of New South Wales (NSW) were invited to participate utilizing a convenience sampling approach (ie, pharmacies known to the research team). The surveys were conducted from November 2016 to July 2017. Consent for a researcher (FTNS) to be present in a pharmacy was obtained from pharmacy owners. Potential participants included adult (18 years old and above) patients who were supplied benzodiazepines upon prescription on the day of the survey. The Anatomical Therapeutic Chemical (ATC) Classification System[45](#hsr2116-bib-0045){ref-type="ref"} was used for the classification of drugs of interest, ie, N03AE (antiepileptics, benzodiazepine derivatives), N05BA (anxiolytics, benzodiazepine derivatives), and N05CD (hypnotics, benzodiazepine derivatives). N05CF (benzodiazepine‐related drugs, Z‐drugs) were not included in this study. After providing informed consent, participants filled out the survey questionnaire (participants had the option to ask questions to the pharmacist or research team member for unclear items), placed the completed survey in a sealable prestamped envelope, and handed this over to the research team member (FTNS) or mailed the completed survey back to the researchers. Participants were given the option to complete the survey questionnaire either at the pharmacy or at home. The time that it would take participants to complete the survey was tested by the researchers and estimated to be between 15 and 20 minutes. 2.3. Analysis {#hsr2116-sec-0009} ------------- Completed survey questionnaires were assigned a code before initiating the analysis. Data were entered into the IBM SPSS Statistics package for Windows, Version 22.0 (Released 2013, Armonk, New York: IBM Corp), and entered data were checked for normality using the Kolmogorov‐Smirnov (K‐S) test[46](#hsr2116-bib-0046){ref-type="ref"} as well as through visual inspection of histograms.[47](#hsr2116-bib-0047){ref-type="ref"} Descriptive analyses were conducted to examine the demographic characteristics of participants. Correlation coefficients (Spearman\'s correlation coefficient or Phi coefficient for two binary variables) were used to describe (1) associations between current long‐term benzodiazepine use (using benzodiazepines for at least 1 year) and the following variables: age, sex, level of education, benzodiazepines being used, length of action of the benzodiazepine, and reason for taking the benzodiazepine; (2) associations between the BMQ score (concerns and necessity themes) and long‐term benzodiazepine use (current behaviours), benzodiazepine withdrawal attempt (past behaviours), and preference for behavioural therapies (future behaviours); and (3) associations between past, current, and future intended behaviours. Finally, binary logistic regression was conducted to determine whether past benzodiazepine withdrawal attempts and future preference for behavioural therapies differed with demographic or beliefs‐related variables that were significant in the correlation analysis or previous literature.[48](#hsr2116-bib-0048){ref-type="ref"}, [49](#hsr2116-bib-0049){ref-type="ref"}, [50](#hsr2116-bib-0050){ref-type="ref"}, [51](#hsr2116-bib-0051){ref-type="ref"} All tests were two‐sided with a significance level of 0.05. 3. RESULTS {#hsr2116-sec-0010} ========== The results are outlined with reference to our study\'s stated aims; these included exploring the association of participant\'s benzodiazepine usage behaviours with their sociodemographic characteristics and medication‐related beliefs. A secondary aim of the study was to confirm participants\' preferences for information sources on alternative behavioural techniques. Accordingly, in the results section, firstly participants\' demographics and then their scores on the BMQ‐specific questionnaire[34](#hsr2116-bib-0034){ref-type="ref"} are presented. This is followed by presenting associations between the sociodemographics and BMQ scores with participants\' benzodiazepine usage behaviours, which are presented in sequence: ***past** (previous withdrawal attempt)*, ***current** (long‐term use of benzodiazepine),* and ***future*** *(intention to try behavioural alternatives)* behaviours. Finally, participants\' preferred sources of information for behavioural therapies are descriptively reported. A total of 75 participants (67% female) out of 107 approached (70% response rate) were recruited for this study from 12 different pharmacies across NSW. Of these 12 pharmacies, 10 were medium‐sized (150‐350 m^2^ floor space) and two were large‐sized (350‐500 m^2^ floor space) pharmacies; eight were independently owned; 10 were from metropolitan, and two were from regional areas. All the participants completed the survey questionnaire at the pharmacy, even though the option to complete at home and mail back was available. The mean (±SD) age of the participants was 54.3 (±16.7), with a range of 23 to 86 years. Seventy‐one percent of the participants had at least high school level education. Sixty‐seven percent of the participants had been taking benzodiazepines for at least 1 year. The reported reasons for benzodiazepine use were sleep disorders (59%), psychiatric disorders (31%), pain management (7%), and other problems (4%). For the BMQ scale, the mean score (±SD) calculated for the necessity theme was 17.7 (±5.6), and the concerns theme was 13.6 (±4.7). While 65% of the sample had strong beliefs in the necessity of their benzodiazepine use as indicated by a BMQ‐specific score of greater than 15, only 40% reported strong concerns about the potential adverse effects of the medication (BMQ‐specific scores greater than 15). Descriptive statistics for the participants are shown in Table [1](#hsr2116-tbl-0001){ref-type="table"}. ###### Descriptive statistics for demographic characteristics of the participants (*n* = 75) Demographic Variables Number (%) ----------------------------------------------------- ------------ Females 50 (67) Level of education No education 1 (1) Primary school education 21 (28) High school education 23 (31) Vocational education 11 (15) Tertiary education (graduate or postgraduate level) 19 (25) Benzodiazepines purchased on the day of the survey Diazepam 43 (57) Temazepam 28 (37) Oxazepam 4 (5) Using benzodiazepines for at least 1 year 50 (67) Reason for benzodiazepine use Using benzodiazepine for sleep disorders 44 (59) Using benzodiazepine for psychiatric disorders 23 (31) Using benzodiazepine for pain 5 (7) Using benzodiazepine for other problems 3 (4) 3.1. *Past* benzodiazepine use behaviours (withdrawal attempts) {#hsr2116-sec-0011} --------------------------------------------------------------- In the sample, 47% (*n* = 35) of the participants had attempted to withdraw the benzodiazepine, and in these cases, sudden cessation of the benzodiazepine was the most commonly reported strategy applied to withdraw the medication. The reported reasons for attempting benzodiazepine withdrawal included concerns about health (49%), fear of developing dependence (40%), doctors\' suggestion (37%), and other reasons (17%). Details about the reasons and strategies for withdrawing the benzodiazepine are presented in Table [2](#hsr2116-tbl-0002){ref-type="table"}. ###### Reasons and strategies for attempting benzodiazepine withdrawal (*n* = 35) Reasons and Strategies for Attempting to Discontinue Benzodiazepines Number (%) ---------------------------------------------------------------------- ------------ Reasons for attempted benzodiazepine withdrawal Concerns about health 17 (49) Fear of developing dependence 14 (40) Doctors\' suggestion 13 (37) Other reason 6 (17) Strategies applied to withdraw benzodiazepines Stopping the benzodiazepine 15 (20) Reducing the dose 5 (7) Exercise 3 (4) Taking alcohol 2 (3) Trying sleep hygiene 2 (3) Using a herbal product 2 (3) Changing lifestyle 2 (3) Trying an antidepressant 1 (1) Trying behavioural strategies 1 (1) Following pharmacists\' advice 1 (1) Taking the benzodiazepine when needed 1 (1) Reasons for attempted benzodiazepine withdrawal were derived by having participants choose from multiple choice options presented to them with the questions. Strategies used for withdrawing the benzodiazepines were derived by having participants write open‐ended answers in space provided after the questions. The variables presented in the table are thematically collated. In a binary logistic regression analysis, none of the variables evaluated (age, sex, level of education, benzodiazepines being used, length of action of the benzodiazepine, and reason for taking the benzodiazepine) were significantly associated with a benzodiazepine withdrawal attempt. There was no significant correlation between the BMQ themes and past withdrawal attempts. 3.2. *Current* long‐term benzodiazepine use {#hsr2116-sec-0012} ------------------------------------------- There was a moderately strong negative correlation (*r* ~s~ \[75\] = −0.390, *p* = 0.001) between current chronic benzodiazepine use and participants\' level of education. The correlation between current long‐term use of benzodiazepines in the sample with either of the BMQ‐specific themes was only moderately strong. There was a positive (*r* ~s~ \[75\] = 0.316, *p* = 0.006) correlation of the scores from the BMQ‐specific theme around necessity and a negative (*r* ~s~ \[75\] = −0.338, *p* = 0.003) correlation of the scores from the concerns theme with current long‐term use of benzodiazepines. Details about the correlations between current long‐term benzodiazepine use, level of education, BMQ themes as well as past and future behaviours are presented in Table [3](#hsr2116-tbl-0003){ref-type="table"}. ###### Correlation coefficients and *p* values between current long‐term use, past use behaviour, future use behaviour, BMQ subscales, and sociodemographics Dependent Variable Independent Variable *N* Correlation Coefficient *P* Value ------------------------------------------------------------------------------------------ ---------------------- -------- ------------------------- ----------- ***Past benzodiazepine use behaviours (withdrawal attempt)*** Level of education 75 0.141 0.228 Concerns score from BMQ‐specific 75 0.166 0.155 Necessity score from BMQ‐specific 75 −0.016 0.891 ***Current*** benzodiazepine use behaviour (use \>1 year) 75 −0.076 0.513 ***Future*** benzodiazepine use behaviour 75 0.078 0.506 ***Current long‐term benzodiazepine use (use \> 1 year)*** Level of education 75 −0.390 0.001 Concerns score from BMQ‐specific 75 −0.338 0.003 Necessity score from BMQ‐specific 75 0.316 0.006 ***Future*** benzodiazepine use behaviour 75 −0.236 0.041 ***Future benzodiazepine use behaviours (willingness to try behavioural alternatives)*** Level of education 75 0.220 0.058 Concerns score from BMQ‐specific 75 0.297 0.010 Necessity score from BMQ‐specific 75 −0.127 0.278 Abbreviation: BMQ‐specific, Beliefs about Medicines Questionnaire‐specific. We observed no significant correlation between participants\' age, sex, benzodiazepines being used, duration of action of the benzodiazepine, or their reason for taking the benzodiazepine with current long‐term benzodiazepine use. 3.3. *Future* benzodiazepine use behaviours {#hsr2116-sec-0013} ------------------------------------------- In the sample, about two thirds (*n* = 48) of the participants were not willing to consider behavioural therapies as a substitution for their benzodiazepines (only 27/75 participants were interested in behavioural therapies). In these cases, participants\' lack of confidence about the efficacy of behavioural therapies and their lack of time to try behavioural therapies were the main two reasons for not considering behavioural therapies. The reported reasons for not considering behavioural therapies are highlighted in Table [4](#hsr2116-tbl-0004){ref-type="table"}. ###### Reported reasons for not preferring behavioural therapies (*n* = 48) Reasons for Not Preferring Behavioural Therapies Number (%) ----------------------------------------------------------------------------------------- ------------ Lack of confidence about behavioural therapies 18 (38) Lack of time 16 (33) Dependency on sleeping pill 15 (31) Participants\' perception that behavioural therapies take longer time to produce effect 11 (23) Participants\' perception that seeing a psychologist is costly 9 (19) Other reason 5 (10) Reasons for not preferring behavioural therapies were derived by having participants choose from multiple choice options presented to them with the questions. We observed no statistically significant association between past benzodiazepine use behaviours (withdrawal attempts) and the future intentions of trialling behavioural therapies (Table [3](#hsr2116-tbl-0003){ref-type="table"}). There was a weak (*r* ~s~ \[75\] = 0.297, *p* = 0.010) relationship between the BMQ concerns theme and future preference for behavioural therapies (Table [3](#hsr2116-tbl-0003){ref-type="table"}). Interestingly, binary logistic regression revealed that participants using benzodiazepines for sleep disorders were more likely to be in the "interested in behavioural therapies" group (odds ratio \[OR\] = 3.138; 95% confidence interval \[CI\], 1.037‐9.492) (Table [5](#hsr2116-tbl-0005){ref-type="table"}). Indeed, of the 27 participants who were interested to consider behavioural therapies, 20 participants were those who reported using the benzodiazepine for managing sleep disorders. ###### Logistic regression model predicting preference for behavioural therapies based on demographic variables Factor OR 95% CI *P* Value ------------------------------------------- ------- --------------- ----------- Age 0.979 (0.947‐1.011) 0.199 Sex 0.711 (0.245‐2.064) 0.530 Using benzodiazepines for sleep 3.138 (1.037‐9.492) 0.043 Using benzodiazepines for at least 1 year 0.496 (0.169‐1.456) 0.202 The dependent variable in this analysis is future preference for behavioural therapies coded as 0 = *not interested in behavioural therapies* and 1 = *interested in behavioural therapies* (target group). Abbreviations: CI, confidence interval; OR: odds ratio. 3.4. Source preference for information about alternative behavioural therapies {#hsr2116-sec-0014} ------------------------------------------------------------------------------ Doctors were rated as the most preferred source of information for learning more about behavioural therapies (76%, *n* = 57). Following doctors, most of the participants preferred pharmacists as the second (53%, *n* = 40), internet as the third (41%, *n* = 31), and psychologists as the fourth choice (37%, *n* = 28) to consult about behavioural therapies. 4. DISCUSSION {#hsr2116-sec-0015} ============= To our knowledge, this is the first study exploring the actual beliefs of benzodiazepine users about their medication, their willingness to withdraw or discontinue the benzodiazepine, and future preferences for alternative behavioural therapies. The present study highlights the divergence between Australian benzodiazepine prescribing guidelines and current practice in Australian primary care, where two thirds of our primary care‐based sample participants reported long‐term use (ie, using benzodiazepines for at least 1 year). A few participants had tried different strategies to withdraw from benzodiazepines. Many of these strategies (such as sudden cessation of the benzodiazepine and consuming alcohol and herbal products instead of the medication) were inappropriate. Participants\' lower level of education and their beliefs (low concerns and high necessity) appeared to contribute to the current high rate of long‐term benzodiazepine use in this Australian sample. Participants\' beliefs appeared not to be associated with their past withdrawal attempts; however, the item used in our questionnaire simply probed whether participants had made any attempt to withdraw their benzodiazepine, not the actual number of attempts made, which may have been a variable associated with the patients concerns about use. We observed, however, an association between concerns about use and future intention to try behavioural alternatives. Using the benzodiazepine for sleep‐related problems was a predictor for willingness to consider alternative behavioural therapies. Only a third of participants expressed an interest in considering behavioural therapies as an alternative to the use of benzodiazepines. Doctors and pharmacists were the two main health professionals of choice with whom benzodiazepine users wanted to discuss potential behavioural therapies. Therefore, it is clear that a collaborative effort in primary care between doctors and pharmacists may help to ensure quality use of benzodiazepines. Based on the study findings, strategies for shifting from ongoing benzodiazepine use to considering behavioural therapies should include counselling patients about the balance between the necessity of using benzodiazepines and the risks associated with ongoing use in the primary care settings. The observed negative association between the participants\' level of education and the current chronic use of benzodiazepines suggests that individuals with a lower level of education perhaps require particular review and targeted information provision. While past withdrawal attempts do indicate previous positive inclination to cease benzodiazepine use, failure at these attempts perhaps reinforces the necessity of continuing the medication as well as reduces the user\'s perception of their ability to be able to cease use. Therefore, based on theories such as those around planned behaviour[52](#hsr2116-bib-0052){ref-type="ref"}, [53](#hsr2116-bib-0053){ref-type="ref"} as well as our study results, counselling directed at current use and future intention to use is perhaps a more judicious use of clinical time when attempting de‐prescribing of benzodiazepines, rather than focussing on past withdrawal attempts. The observed positive relationship of the necessity theme and the negative relationship of the concerns theme with actual long‐term benzodiazepine use (*current* behaviour) highlights that the decision to continue benzodiazepines reflected the balance between these two opposing belief sets. However, we observed that future intentions around continued benzodiazepine use (eg, voiced preference for behavioural therapies in the *future*) were only related with the concerns theme, highlighting that if targeted information provision can help build realistic concerns about ongoing use, benzodiazepine users may be swayed to discontinue and switch to behavioural alternatives in the future. The results of this study, therefore, suggest that the BMQ could be a useful instrument to elicit medication beliefs and increase health care professionals\' understanding of patients\' attitudes towards the medication, which can, in turn, potentially facilitate the de‐prescribing process. Martin et al describe the use of the BMQ‐specific questionnaire to compare patients\' beliefs about their benzodiazepines before and after an educational program designed to facilitate decisions about discontinuing benzodiazepines; they report a significantly lower‐necessity score and higher‐concerns score after the intervention.[54](#hsr2116-bib-0054){ref-type="ref"} In our study, the results indicate that concerns about the medication are related to both lower current long‐term use and higher preference for future behavioural alternatives. In fact, health concerns and the fear of being addicted were the most commonly reported reasons for a benzodiazepine withdrawal attempt, as shown in other studies as well.[55](#hsr2116-bib-0055){ref-type="ref"} Therefore, discussing concerns in an individualised manner to the patient may be effective in helping patients decide to discontinue long‐term benzodiazepine use. Since there is considerable evidence around applying patients\' beliefs about medication to explore or improve their adherence to a set treatment plan,[56](#hsr2116-bib-0056){ref-type="ref"}, [57](#hsr2116-bib-0057){ref-type="ref"} it can be suggested that patients\' beliefs about their benzodiazepine can also be used to predict adherence to a set treatment withdrawal plan. Sudden cessation of benzodiazepines was the most commonly used strategy for withdrawing the benzodiazepine, and this approach is a distinct departure from clinical guidelines recommending gradual dose reduction for patients on benzodiazepines for longer than 3 to 4 weeks to minimise or avoid withdrawal symptoms.[58](#hsr2116-bib-0058){ref-type="ref"}, [59](#hsr2116-bib-0059){ref-type="ref"} Sudden cessation of benzodiazepines in long‐term users may result in life‐threatening seizures.[58](#hsr2116-bib-0058){ref-type="ref"} These disparities between the guidelines and current withdrawal strategies may reflect that either participant opted to self‐withdraw without health professional advice, or participants possibly received only ad‐hoc advice from their GPs (or primary care physicians or family physicians) or dispensing pharmacists to withdraw the benzodiazepine. In the latter instance, a clear step‐by‐step plan may not have been offered during the GP or pharmacist consultation, given that these health professionals would be well aware of the pharmacological consequences of sudden cessation and would be able to refer to published guidelines on how to help patients discontinue benzodiazepines in a tapered manner.[60](#hsr2116-bib-0060){ref-type="ref"}, [61](#hsr2116-bib-0061){ref-type="ref"}, [62](#hsr2116-bib-0062){ref-type="ref"} Given that even minimal interventions by primary health care professionals can be effective,[63](#hsr2116-bib-0063){ref-type="ref"} primary care professionals should be upskilled and upresourced so as to be able to provide concrete direction when recommending discontinuation of benzodiazepines. A meta‐analysis suggests that gradual dose reduction combined with psychological interventions has better outcomes compared with a gradual dose reduction alone.[64](#hsr2116-bib-0064){ref-type="ref"} Appropriate de‐prescribing plans, adjunctive psychotherapy, and careful monitoring (eg, using withdrawal assessment tools) should be key elements in the benzodiazepine discontinuation process initiated by GPs, who are the main prescribers of benzodiazepines.[58](#hsr2116-bib-0058){ref-type="ref"}, [65](#hsr2116-bib-0065){ref-type="ref"}, [66](#hsr2116-bib-0066){ref-type="ref"} Evidence‐based pragmatic guidelines for de‐prescribing benzodiazepines and practice toolkits may be required to facilitate this. A majority of the participants were not interested in behavioural therapies, contrary to previous studies generally demonstrating that patients favour non‐pharmacological therapies over pharmacotherapies.[67](#hsr2116-bib-0067){ref-type="ref"}, [68](#hsr2116-bib-0068){ref-type="ref"} This inconsistency relative to other studies may be due to small sample size and selection bias in our study. The inconsistency might also be related to the fact that the participants may not have been experienced or informed about behavioural treatments earlier.[67](#hsr2116-bib-0067){ref-type="ref"}, [69](#hsr2116-bib-0069){ref-type="ref"} Greater public awareness about the efficacy of behavioural therapies may be a prudent step for addressing this issue, eg, through public health campaigns given that benzodiazepines have been implicated in vehicle accidents.[70](#hsr2116-bib-0070){ref-type="ref"}, [71](#hsr2116-bib-0071){ref-type="ref"} Another strategy to improve the uptake of behavioural therapies could be to introduce internet‐delivered behavioural therapy services. Studies suggest that internet‐based behavioural therapies have the potential to be time‐efficient and cost‐effective as well as quite acceptable for patients.[72](#hsr2116-bib-0072){ref-type="ref"}, [73](#hsr2116-bib-0073){ref-type="ref"} Interestingly, participants with sleep problems in this study were more likely to prefer behavioural treatment, which supports and further extends previous research.[74](#hsr2116-bib-0074){ref-type="ref"}, [75](#hsr2116-bib-0075){ref-type="ref"} These patients can be specifically targeted for behavioural intervention in primary care settings. Since a majority of the participants identified doctors as their first choice for learning more about behavioural therapies, GPs need to take the initiative to make patients familiar with the behavioural therapies. However, physicians cite time paucity, high workload, lack of skills in behavioural treatment, and limited accessibility to behavioural therapy providers as the barriers to reduce benzodiazepine use and introduce behavioural interventions in general practice setting.[76](#hsr2116-bib-0076){ref-type="ref"}, [77](#hsr2116-bib-0077){ref-type="ref"} Introducing other health care professionals co‐located within GP practice centres to support GPs for educating patients about behavioural therapies may help to overcome some of these issues.[29](#hsr2116-bib-0029){ref-type="ref"} In a qualitative Australian study, GPs acknowledged the role of other health professionals in facilitating successful benzodiazepine cessation.[44](#hsr2116-bib-0044){ref-type="ref"} In the current study, following GPs, pharmacists were the second preferred source of information for behavioural therapies. Thus, a collaborative approach with pharmacists could be an option where they can support GPs in withdrawing benzodiazepines and providing behavioural therapies.[78](#hsr2116-bib-0078){ref-type="ref"} Community pharmacists also have the opportunity to counsel patients at the point of purchase; in the case of patients with repeat prescriptions, community pharmacists see benzodiazepine users even more frequently. At these opportunities, pharmacists can discuss the potential risks of long‐term ongoing use, provide information about alternatives, and refer patients to their doctor for a withdrawal or discontinuation plan. Once patients have a plan, in an interdisciplinary model, pharmacists should be apprised of the plan (which could be written out as a prescription) for further facilitating patient adherence to the plan with supportive advice. Indeed, data suggest that community pharmacists do frequently encounter chronic benzodiazepine use in practice, but fewer than half converse with the patient about ongoing use.[79](#hsr2116-bib-0079){ref-type="ref"}, [80](#hsr2116-bib-0080){ref-type="ref"} On the other hand, a recent study in Australia highlighted that pharmacists could be successfully trained to deliver behavioural interventions with positive sleep health outcomes in patients with insomnia,[81](#hsr2116-bib-0081){ref-type="ref"} suggesting the benefit of developing pharmacist roles in this area. Despite the potential to play key roles in benzodiazepine deprescribing,[82](#hsr2116-bib-0082){ref-type="ref"}, [83](#hsr2116-bib-0083){ref-type="ref"} pharmacists, especially community pharmacists, are currently underutilised.[79](#hsr2116-bib-0079){ref-type="ref"} 4.1. Strengths and limitations {#hsr2116-sec-0016} ------------------------------ The small sample size may be a limitation of this study. For moderate strength correlations (eg, 0.30‐0.40), a sample of 75 may be sufficient to demonstrate significance at a 0.05 significance level with 75% power.[84](#hsr2116-bib-0084){ref-type="ref"} However, the study may have been underpowered to demonstrate lower‐strength correlation at the same significance level and power. There is a possibility of selection bias, as individuals who have successfully withdrawn their benzodiazepines were not recruited for this study. Further, the questions used in this study to explore participants\' willingness to withdraw and preferences for behavioural therapies were customised for this study. However, the questionnaire was developed by reviewing the previous literature and was face‐validated by psychology, sleep, and pharmacy practice researchers. Given this part of the questionnaire measured a set of diverse issues (past behaviour of withdrawing, actual daily use of benzodiazepine, and future intentions to try alternatives to benzodiazepine use), internal consistency was not measured, as behaviour and intention constructs are in themselves quite different and these questions also had different response items. Past behaviours appeared less important based on our data, and current behaviours are verifiable through pharmacy prescription records. Therefore, it may be suggested that a fuller set of items with comparable response sets around intended future behaviours with respect to benzodiazepines should be constructed and tested psychometrically. This would be useful for future research and is certainly a limitation in our study. The questionnaire did not ask explicitly about participants\' readiness or current intention to withdraw from the benzodiazepine. Exploring participants\' belief was limited to BMQ‐specific rather than using the more comprehensive exploration of belief sets, for example, using variables included in models such as the HBM. Lastly, there is a possibility for response bias (eg, strong beliefs about the necessity of benzodiazepines, low concerns, and social desirability). 5. CONCLUSION {#hsr2116-sec-0017} ============= Specific characteristics of benzodiazepine users and their beliefs about taking the benzodiazepine can inform the provision of individualised interventions by GPs to help switch patients currently on long‐term benzodiazepines to alternative behavioural therapies. This study highlights the significance of informing patients about the balance between the necessity of use versus the concerns that long‐term use of benzodiazepines raises. Given that GPs are very time pressured, introducing practice pharmacists within general practices could be time efficient and enhance GPs\' capacity for providing behavioural therapies, and this area of collaborative care in de‐prescribing unwarranted use of high‐risk medications warrants future research. FUNDING INFORMATION {#hsr2116-sec-0020} =================== This research did not receive any specific grant from funding agencies in the public commercial or not‐for‐profit sectors. However, the authors acknowledge the PhD scholarship support received by the first author from NeuroSleep, a Centre of Research Excellence supported by a grant from the Australian National Health and Medical Research Council. The funding sources were not involved in study design; collection, analysis, or interpretation of data; writing of the report; or the decision to submit the report for publication. CONFLICTS OF INTEREST {#hsr2116-sec-0019} ===================== None. AUTHOR CONTRIBUTIONS {#hsr2116-sec-0021} ==================== Conceptualization: Fatema‐Tun‐Naher Sake, Keith Wong, Bandana Saini Formal Analysis: Fatema‐Tun‐Naher Sake, Keith Wong, Bandana Saini Writing (original draft): Fatema‐Tun‐Naher Sake Writing (review and editing): Fatema‐Tun‐Naher Sake, Keith Wong, Delwyn Bartlett, Bandana Saini All authors have read and approved the final version of the manuscript. Fatema‐Tun‐Naher Sake had full access to all of the data in this study and takes complete responsibility for the integrity of the data and the accuracy of the data analysis. All participating benzodiazepine users are acknowledged for their valuable time in completing the survey questionnaire.

INTRODUCTION ============ The great benefits of using nanomaterials in modern technologies are no longer questioned. However, the potential negative effects associated with the propagation and accumulation of nanomaterial components, such as nano-particles and nanofibers in the environment, require further study \[[@R1], [@R2]\]. Plants are the major components of ecosystems; subsequently, significant attention should be paid to the effects of various technogenical materials upon them \[3--[@R5]\]. Carbon nanomaterials (CNM), such as fullerenes, multi-walled carbon nanotubes (MWNTs), and single-walled carbon nanotubes (CWNTs), are a matter of special interest, as their industrial production is rapidly developing. Since nanotubes have a fibrillar form, they are compared with asbestos \[[@R6]\]. In light of such an analogy, potential adverse effects on living organisms can be anticipated \[[@R7]\]. Preliminary studies have provided evidence that MWNTs and SWNTs are pathogenic to animals \[[@R8]\], yet they have different effects on plants. MWNTs were shown to considerably increase the growth rate of tomato seedlings \[[@R9]\], have no effect on the growth parameters of wheat \[[@R10]\], and inhibit the growth of rice seedlings \[[@R11]\]. SWNTs have been shown to suppress the growth of tomato roots, but stimulate the root growth of onion and cucumber \[[@R12]\]. In contrast, MWNTs have a toxic effect on *Arabidopsis* cultured cells. \[[@R13]\]. High adsorption of MWNT/SWNT on the roots of seedlings was observed in all the experiments conducted. However, the penetration, uptake and accumulation of MWNTs/SWNTs in plant cells and tissues are not well documented . Furthermore, the mechanism of the development of the physiological changes caused by the exposure of plants to nanotubes also remains unclear. The present work was aimed at studying the effect of the industrial nanomaterial Taunit, containing MWNTs, on *O. arenaria* seedlings, and the ability of MWNTs to penetrate and accumulate in plant cells and tissues. EXPERIMENTAL PART ================= **Object of the study** The object of the present study is industrial CNM Taunit (NanoTechCenter Ltd., Tambov, Russia). This material is a loose black powder, composed of grainy agglomerates with a size of several micrometers. Agglomerates mostly consist of entangled MWNT bundles. MWNTs have a hollow cylindrical structure; at least 2 µm long, with an external diameter of 20--70 nm and an internal diameter of 5--10 nm. Taunit is produced by chemical vapor deposition; its purity is above 98% \[[@R14]\]. **Seed germination and morphometric assessment** The seeds of *Onobrychis arenaria* were germinated in a medium containing a colloidal aqueous solution of CNM Taunit with a concentration of 100 or 1,000 mg/l. Prior to use, CNM was dispersed in distilled water by ultrasonic treatment . Distilled water was used to prepare the control medium. The seeds (50 seeds per dish) were grown for 10 days on filter paper in glass Petri dishes (diameter of 90 mm) with 5 ml of a CNM suspension added. 200 seeds were used in each experiment. The growth conditions followed the requirements of the State Standard procedure GOST 12038-84 (Agricultural seeds. Methods for evaluation of germination ). The effect of CNM on esparcet seedlings was estimated on the basis of the following parameters: the rate of seed germination (%), the energy of germination (%), and the length of the roots and stems . The energy of germination and the rate of germination were determined as the ratio between the number of germinated seeds and the number of plated seeds by day 5 and day 10, respectively (%, in accordance with the State Standard). **Extraction of soluble peroxidases and determination of their activity** The weighed samples (2 g) of *O. arenaria* seedlings tissues were placed into 5 ml of a cold phosphate/citrate buffer (1 M solution of citric acid + 1 M NaH ~2~ PO ~4~ , two solutions combined to achieve pH 5.5) and ground in a porcelain mortar at 4°С \[[@R15]\]. The homogenate was centrifugated at 3,000 g for 15 min. The cleared supernatant was used to determine the activity of soluble peroxidases on the basis of the change rate (time, s) of the optical density at a wavelength of 580 nm in the reaction mixture containing 0.5 ml of 0.1 M solution of the phosphate/citrate buffer (pH 5.5), 0.5 ml of 0.3% Н ~2~ О ~2~ , 0.5 ml of 0.05% guaiacol (Sigma, USA), and 0.5 ml of the sample. Peroxidase activity was measured at 25 ^о^ С immediately after the enzymes were extracted from the samples. Enzymatic activity was calculated by Boyarkin's method \[[@R16]\] and expressed in arbitrary units of activity per gram of fresh tissue weight per second, according to the following formula: *А* = (ε × α × β × γ)/( *d* × *t* ), where ε is the extinction coefficient, α is the ratio between the amount of buffer taken for extract preparation (ml) to fresh tissue weight (g), β is the degree of additional dilution of the extract in the reaction mixture, γ is the degree of constant dilution of the extract in the reaction mixture, *d* is the thickness of the absorbing layer (mm); and *t* is the reaction time (s). **Light and electron microscopy** The bottom of a plastic box (approximate dimensions 40 × 40 × 7 cm) was covered with four gauze layers moistened with a CNM solution or water (in the control sample). 100 *O. arenaria* seeds were placed on the gauze and exposed either to the CNM solution at the concentration 300 mg/l or water without CNM. After 5 and 10 days of exposure to CNM, the seedlings were fixed for light and electron microscopic studies . For light microscopy, the seedlings were fixed in a 3:1 mixture of 96% ethanol and acetic acid for 16 h. After fixation, the plant samples were put into 70% ethanol. The plant parts under study (roots, leaves, coleoptiles) were then placed onto a glass slide into a drop of 45% acetic acid. The preparations of whole mount plant parts were made according to the standard procedure \[[@R17]\]. The preparations were analyzed using a Leica DM1000 light microscope (objectives ×10, ×20, ×40, and ×100). The images were recorded with a Leica DFC 295 digital camera (sensor size 3 × 10 ^6^ pixels). For transmission electron microscopy (TEM), the seedlings were fixed with 2.5% glutaraldehyde on a 0.1 M Na-K-phosphate buffer (pH 7.2) supplemented with sucrose (15 mg/ml). The samples were then dehydrated in a series of solutions of increasing ethanol concentrations and embedded in Epon 812, according to the standard procedure. For optimization of MWNTs detection within plant tissues, we opted not to use additional fixation with OsO ~4~ and staining with uranyl acetate and lead citrate. The sample of pure CNM was prepared for TEM as follows: 25 mg of Taunit was placed onto the surface of unpolymerized resin (Epon 812), poured into a tube. Then, the sample was centrifugated for 3 min at 6,000 g and polymerized, according to the standard procedure. {#F1} Ultrathin sections of the samples embedded into Epon were investigated by TEM and SAED (selected area electron diffraction) using transmission electron microscopes JEM-1011 (JEOL) equipped with a GATAN ES500W digital camera, and LEO 912АВ (Carl Zeiss). RESULTS AND DISCUSSION ====================== **Effect of Taunit on the morphometric and biochemical characteristics of** *O. arenaria* **seedlings** In order to characterize the phytotoxicity of CNM, we used seed germination tests, in which the germination energy, germination rate, length of roots and stems, and peroxidase activity were estimated \[[@R18]\]. The germination of *O. arenaria* seeds in the presence of the colloidal solution of CNM Taunit increased the germination energy by 14% as compared with the control sample. CNM with a concentration of 100 and 1,000 mg/l increased the seeds germination rate by 2 and 7%, respectively ( *[Fig. 1](#F1){ref-type="fig"}a* ). Taunit also stimulated the growth of roots and stems of seedlings . At CNM concentrations of 100 and 1,000 mg/l, the root length increased by 55 and 73%, respectively; the length of seedling stems increased by 84 and 82%, as compared with the control sample ( *[Fig. 1](#F1){ref-type="fig"}b* ). Thus, CNM Taunit slightly increased the germination rate and the germination energy of seeds and considerably increased the length of roots and stems of seedlings . Taunit at the concentrations of 100 and 1,000 mg/l also enhanced peroxidase activity in *O. arenaria* seedlings, respectively, to 0.31 ± 0.01 and 0.19 ± 0.02 au/g fresh weight, which is significantly higher than the control value (0.12 ± 0.01) ( *[Fig. 1](#F1){ref-type="fig"}c* ). It is well-known that plants respond to mechanical stress and injury by changing their morphology or growth rate. This phenomenon has become known as thigmomorphogenesis. Thigmomorphogenetic changes are regarded as the adaptation process to stress in plants, and plant hormones play an important role in this process \[[@R19]\]. Under mechanical stress or after injury, the activity of the plants stress hormone, jasmonic acid, increases, whereas the activity of auxin , which controls the processes of morphogenesis and plant growth, decreases. These changes of plant hormone levels may be associated with the increase of peroxidase activity \[[@R20]-[@R22]\]. Peroxidases are involved in a number of biological processes, such as photosynthesis, respiration, and protein metabolism. It is an antioxidant enzyme with high sensitivity towards external factors, and this allows using peroxidase activitiy assay for testing of the physiological state of plants. In most cases, a high level of peroxidase activity demonstrates the initiation of the mechanism of a nonspecific response of the plant to stress \[[@R23]\]. It can be assumed that increase in peroxidase activity is associated with the oxidative stress caused by CNM. It has been shown that MWNTs, accumulated at the root surface, often pierce cell walls of epidermal cells \[[@R10]\]. Such interaction can be considered as a mechanical injury and thus elevate the level of peroxidase activity. Our results confirmed that the level of peroxidase activity decreases with an increase in CNM concentration. This observation could be explained by the inactivation of peroxidase molecules by nanotubes due to sorption or other chemical interactions. Our studies demonstrated that, along with an increase in the level of peroxidase activity, exposure to CNM stimulated the growth of roots and stems of plants . Further research is needed to explain the mechanisms of the enhanced growth of plants in the presence of CNM. **Analysis of CNM Taunit in** {#F2} {#F3} *O. arenaria* **seedlings using light and electron microscopy** Upon germination of *O. arenaria* in the presence of CNM Taunit, the roots, stems, and leaves of seedlings acquire a characteristic dark gray color. Analysis of whole mount preparations showed that dark gray and black agglomerates are localized on the surface of seedling roots and within particular tissues and cells of the roots, stems, and leaves ( *Figs. 2a-c, 3a-c* ). Ultrathin sections prepared from the plant organs containing aggregations of CNM were analyzed by TEM. Prior to studies of plant material, we had analyzed a pure CNM sample. TEM demonstrated that CNM Taunit ( *[Fig. 4](#F4){ref-type="fig"}a, b* ) contains mostly agglomerates of MWNTs and some inclusions of a nanodispersed electron-dense material (presumably, graphitized carbon). The characteristic features of MWNTs are presented in *[Fig. 4](#F4){ref-type="fig"}b* . The MWNTs without a small amount of inclusions ( *[Fig. 4](#F4){ref-type="fig"}c* ) were characterized by the SAED method . As shown in *[Fig. 4](#F4){ref-type="fig"}d* , due to the regular periodical packing of carbon atoms, MWNTs have an electron diffraction pattern typical for polycrystalline structures. This diffraction pattern was used as a reference sample for the identification of MWNTs in biological material. {#F4} The analysis of ultrathin sections of roots and leaves showed that CNM is present on the surface and inside the seedling organs. The adsorption of MWNTs on the root surface ( *[Fig. 5](#F5){ref-type="fig"}a* ) has also been reported by other researchers \[[@R3], [@R10], [@R11]\]. Furthermore, agglomerates and single MWNT of different lengths are detected in root ( *Figs. 5b,c* ) and leaf cells ( *Figs. 6a-c* ). Thus the MWNTs are unmistakably identified by TEM. However, if MWNTs are located transversely or at some angle to the section plane, it is difficult to distinguish the MWNT fragments from different electron-dense endogenous inclusions. The method SAED can be used to identify MWNT in plant tissues ( *Figs. 5d, 6d* ). Superposition of electron diffraction patterns of MWNTs found in biological samples ( *[Fig. 5](#F5){ref-type="fig"}e, 6e* ) with the reference electron diffraction pattern ( *[Fig. 4](#F4){ref-type="fig"}d* ) confirms the presence of MWNTs in plant cells. Additional diffraction spots on the electron diffraction pattern are accounted for by the presence of endogenous crystalline inclusions in plant tissues. The accumulation of MWNTs on the root surface was reported by many authors \[3, 9--[@R11]\] and it has been suggested that interaction of MWNTs with plant organs affects plant growth and development \[[@R13]\]. Adsorption of a large amount of MWNTs on the root surface may suppress the water flux and uptake of nutrients, thus inhibiting plant growth \[[@R11]\]. Conversely, the stimulation of seed germination may be associated with the fact that nanotubes (SWNTs) pierce the seed cover and increase water uptake, facilitating seed germination and plant growth \[[@R9]\]. However, SWNTs not only pierce the cell wall , but also penetrate inside the cells; this phenomenon was demonstrated using FITC-labeled SWNTs (SWNT/FITC) \[[@R24]\]. Insertion of MWNT into the wall of epidermal cells and root hairs up to 4 µm was observed in wheat seedlings using two-photon excitation microscopy; however, penetration of whole MWNTs into the cytoplasm was not noted \[[@R10]\]. The authors assumed that penetration, uptake and accumulation of MWNTs are less evident due to the larger nanotube diameter as compared with that of SWNT \[[@R10]\]. CONCLUSIONS =========== {#F5} We demonstrated that MWNTs penetrate cell walls, accumulate in the cells and tissues, and most likely are transported via a plant's vascular system from roots to stems and the leaves of *O. arenaria* seedlings. We argue that the stimulation of *O. arenaria* roots and stems growth and the increase in peroxidase activity were induced by the oxidative stress which develops due to the accumulation of MWNTs in plant cells and tissues. {#F6} The authors are grateful to S.S. Abramchuk (Belozersky Institute of Physico-chemical Biology) for assistance with this study. This work was supported by the Federal Program "Scientific and Scientific-Pedagogical Personnel of Innovative Russia" (Government contracts P 208 dated April 23, 2010 and 16.740.11.01-94 dated September 24, 2010) and Government contract №01.648.11.3003.

1. Introduction {#sec0005} =============== Epidemiological studies of birth and death records led to "Barker\'s hypothesis" almost 30 years ago, suggesting the influence of perturbed gestational milieu on the development of diseases later in life ([@bib0060]). This generated a great deal of interest in the fetal origins of adult disorders. Although Barker\'s hypothesis was mainly based on the potential impact of gestational malnutrition to result in permanently altered organs function and structure, a growing body of evidence suggests that a range of maternal complications of pregnancy such as gestational diabetes mellitus, intrauterine growth restriction (IUGR), preeclampsia and maternal stress are associated with adult disorders ([@bib0065], [@bib0295], [@bib0210]). Perturbation of developmental adaptive processes that are known to be involved in permanent changes in physiology, structure and metabolism or early life programming can adversely affect brain development, impacting both brain structure and function ([@bib1065]). Findings from human and animal studies indicate that several environmental factors and physiological mechanisms involving endocrine as well as immune system during gestation play crucial role in early-life programming of later-life brain and behavior by long-term remodeling of the brain ([@bib1065], [@bib0120], [@bib0050]). Evidence toward this notion was derived from neuropathological aberrations including enlarged cerebral ventricles, as well as changes in gray and white matter indicative of impaired neural development during gestation in serious mental illnesses such as schizophrenia ([@bib0545], [@bib0975], [@bib0340]). Importantly, structural brain changes during development of schizophrenia are seemingly determined by genetic components, altered expression of schizophrenia risk genes and epigenetic dysregulation ([@bib0765], [@bib0895], [@bib1110]). The developmental origin of schizophrenia might potentially be a result of prenatal exposure to a diversity of factors such as infection, stress, persistent organic pollutants, smoking and other substance use, maladaptive diet and developmental epigenetic changes ([@bib0185], [@bib0710], [@bib0635]). A two-hit model is one of the leading theories of schizophrenia pathogenesis. The two hit model predominantly implicating the neurodevelopmental theory is mainly based on the assumption that aberrant development during two critical time points (early brain development and adolescence) additively produces risk for schizophrenia-like symptoms. The "first hit" potentially occurs in utero. Although genetic components were primarily considered as the "first hit", recent understanding suggests that along with genetic, environmental factors might also exert a similar function. Animal as well as human studies demonstrated that genetic susceptibility in combination with a developmental insults can prime an individual for a later event that ultimately increase risk for the onset of schizophrenia and this could possibly be mediated by known salient signaling pathways ([@bib0080], [@bib0850]). The "first hit" may disrupt developing neuronal architecture, specific neural networks, establish an abnormal inflammatory response and account for premorbid signs and symptoms in individuals who later develop schizophrenia ([@bib0685], [@bib0435]). Contextually, prenatal infection has become one of the best evidenced environmental risk factors of the "two-hit" hypothesis of schizophrenia as it increases the offspring\'s sensitivity to environmental challenges postnatally and renders them more vulnerable to the pathological effects of a second postnatal stimulus ([@bib0080], [@bib0850]). This "second hit" due to stress, immune re-exposure or even peri-pubertal sex hormonal changes might potentially unmask and exacerbate the hitherto latent inflammatory processes with resultant onset of clinical symptoms ([@bib0880], [@bib0735]). Experimental support for the impact of a second hit is evident from animal studies, where prenatal immune activation elicited by gestational infection and subsequent exposure to stress, cannabinoid use, etc. during adolescence has been shown to unmask the latent neuropathological consequences of prenatal infection and lead to increased risk for the development of schizophrenia ([@bib0305], [@bib0490]). During adolescence, events like excessive elimination of synapses and loss of plasticity might influence the development of the disorder ([Fig. 1](#fig0005){ref-type="fig"} ). Despite this understanding, precisely how the abnormal developmental trajectory of the brain is established during gestation and also how this is causally related to the manifestation of symptoms that appear early in adult life remains a challenging issue. Here we examine recent advances and discuss many emergent factors and mechanisms that might elucidate the mechanistic link between early life programming and adult manifestation of psychiatric conditions.Fig. 1Depiction of 'two-hit' model of schizophrenia. 2. Neurodevelopmental origin of schizophrenia: a review of evidence {#sec0010} =================================================================== Disparate lines of evidence support the view that schizophrenia is a neurodevelopmental disorder. The primary reason is that the onset of schizophrenia has a cumulative age incidence distribution, or developmental function, that is nonlinear with a peak change in slope or acceleration that usually takes to occur during young adulthood. Given the plausibility of the existence of brain abnormalities in schizophrenia at the onset of the illness, it further seems reasonable to conceive schizophrenia as a neurodevelopmental disorder ([@bib1005]). Recent understanding suggests determining role of gene and environmental interactions in the neurodevelopmental trajectories of schizophrenia. The genetic and environmental factors cause structural and functional change not only during pre- and peri-natal periods but also in childhood and early adolescence ([@bib0625], [@bib1230]). Further support has been provided by epidemiological studies showing premorbid intellectual deficits dating back to early development, and neuropathological studies showing altered cerebral cytoarchitecture indicative of a developmental rather than acquired encephalopathy. The main neurodevelopmental hypotheses for schizophrenia set forth in the last 15 years are relatively restricted and share three assumptions ([@bib1255]). Firstly, the primary pathogenic defect is an early derangement of the orderly development of the central nervous system that occurs in the pre- or peri-natal period. Secondly, the period of active operation of the causative agent is of short duration, meaning that it is essentially static and lastly the behavioral consequences of this static process remain relatively latent until long after the primary pathogenic process has run its course. In addition, it is now becoming evident that prenatal adversity induced neurodevelopmental disruption can have physical manifestations. Minor physical abnormalities, neurological soft signs and altered dermatoglyphics pattern represent important markers of disordered neurodevelopment in schizophrenia ([@bib1095], [@bib0460], [@bib0020]). Some of the minor physical abnormalities such as confluent eyebrows, hypertelorism, ear protrusion, low-set ears, palatal abnormalities, tongue furrows and so on represent a set of risk markers for schizophrenia ([@bib1270]). Neurological soft signs correlate with various neuro-cognitive and neuro-anatomical abnormalities in schizophrenia ([@bib1195]). Neurological soft signs are understood as a manifestation of the "cerebello-thalamo-prefrontal" brain network model of schizophrenia ([@bib1290]). Higher levels of neurological soft signs have been reported in healthy children who later develop schizophrenia and in first episode as well as antipsychotic-naïve schizophrenia patients ([@bib0770], [@bib0325], [@bib1190]). Recent examinations of cortex morphology, a marker of brain development in first-episode schizophrenia patients with neurological soft signs revealed a lower global sulcal index (g-SI) in both hemispheres, a lower regional sulcal indexes (r-SI) in left dorsolateral prefrontal and right lateral occipital cortices ([@bib0470]). Taken together, these findings strongly suggest evidence of distinct neurodevelopmental pathways in schizophrenia patients with neurological soft signs. Although various findings including twin studies have highlighted the importance of genetic determination of neurological soft signs, minor physical abnormalities and dermatoglyphic patterns in schizophrenia ([@bib0920], [@bib0430]); the precise genetic determinant of these parameters are yet to be identified. 3. Different risk factors and mechanisms underlying developmental origin of schizophrenia {#sec0015} ========================================================================================= 3.1. Exogenous risk factors {#sec0020} --------------------------- ### 3.1.1. Prenatal teratogens {#sec0025} A range of teratogens and neurotoxic agents such as industrial chemicals (persistent organic pollutants and heavy metals), high levels of ionizing radiation, tobacco smoke, cocaine, alcohol and certain drugs affect early brain development ([@bib0505]). Multiple studies have demonstrated that chemical exposures during early development constitute a new potential class of risk factors for schizophrenia. Prenatal exposure to Pb^2+^ has been found to increase the risk of schizophrenia in the offspring ([@bib0950]). One plausible underlying neurobiological mechanism is the alteration of subunit composition of N-methyl-[d]{.smallcaps}-aspartate receptor (NMDAR) complexes with subsequent effects on calcium-sensitive signaling pathways involved in CREB phosphorylation ([@bib1145]). Prenatal exposure to analgesics in the second trimester of pregnancy conferred a more than four-fold greater risk of schizophrenia ([@bib1105]). Aspirin taken during pregnancy interferes with the prostaglandin pathway, and exposure may influence the risk of schizophrenia in offspring ([@bib0525]). Animal models of schizophrenia have demonstrated that rats exposed to cigarette smoke in the prenatal period exhibited a significant increase of the lipid peroxidation, protein oxidation and DNA damage in adult offspring ([@bib0450]). These effects can lead to several neurochemical changes overlapping with the pathophysiology of schizophrenia. Maternal smoking through pregnancy increases the risk of later schizophrenia among offspring, conferring an increased severity of negative symptoms ([@bib1115]). Interestingly, some of the chemicals and drugs contribute to schizophrenia risk not only as a result of prenatal exposure but also an adolescent exposure. Cadmium and lead for example appear to be associated with both mood and psychotic disorders ([@bib0675], [@bib0115], [@bib0955]). Pre-adolescence exposure to cannabis is shown to be associated with a twofold increase in the risk of schizophrenia ([@bib1175]). ### 3.1.2. Prenatal nutrition {#sec0030} Compelling evidence suggests that prenatal malnutrition leads to long-term brain impairment and increased risk of schizophrenia ([@bib0205], [@bib0180], [@bib1205]). Studies conducted after in utero exposure to hunger/famine during Dutch Hunger Winter of 1944--1945 and 1959--1961 Chinese Famine have indicated a direct association between prenatal starvation and increased risk of schizophrenia ([@bib0570], [@bib1120]). Prenatal exposure to famine affected brain morphology such as decreased intracranial volume, which is seen in schizophrenia ([@bib0605]). More recently, there is increasing evidence that poor quality diet rather than malnutrition per se is a risk factor. The western diet is calorie, fat and sugar rich and nutrient poor, and obesity is currently one of the major public health issues. There is now good prospective evidence in the depression literature that poor quality diet is a risk factor for depression ([@bib1015]), and recent data that maternal prenatal diet impacts the risk of internalizing and externalizing diet in offspring ([@bib0635]). The amount of information on the link between schizophrenia and diet is poorer, but such data is emerging ([@bib0360]). There is emerging evidence of a link between gluten sensitivity and increased levels of antigliadin antibodies and schizophrenia ([@bib1315]). What is known is that diet influences the pathways to neuroprogression in schizophrenia including inflammation and oxidative stress ([@bib0110]). Maternal vitamin deficiency during early pregnancy has been associated with the risk of schizophrenia in the offspring. One of the widely studied vitamins in the context of neurodevelopmental disorders is vitamin D. Vitamin D is arguably the nutritionally related factor that is most deficient in western populations, and some data that this is particularly an issue in psychiatric cohort ([@bib0100]). There is now clear evidence that vitamin D is involved in brain development and being a potent pro-differentiation agent, vitamin D can influence brain functioning via many different pathways ([@bib0420]). Transient prenatal vitamin D deficiency results in abnormal brain development, persistent changes in adult brain structure, neurotransmission, synaptic plasticity, neurochemistry, and behavior and several other biological pathways including oxidative phosphorylation, redox balance, cytoskeleton maintenance, calcium homeostasis, chaperoning and post-translational modifications ([@bib0405], [@bib0410], [@bib0420]). Several studies have demonstrated the immuno-modulatory properties of vitamin D and its implications in a number of autoimmune/inflammatory disorders ([@bib0445]). Importantly, fetal vitamin D plays a crucial role in controlling placental inflammation ([@bib0800]). Vitamin D influences Th1:Th2 balance by inhibiting Th1 pathway and promoting the Th2 pathway through down-regulation of pro-inflammatory cytokines and up-regulation of anti-inflammatory cytokines. Pro-inflammatory cytokine-induced alterations in cognition/behavior could therefore be linked to low levels of vitamin D. This is supported by recent findings that vitamin D mitigates age-related cognitive decline through the modulation of pro-inflammatory cytokine production in rats ([@bib0155]). In addition, developmental vitamin D3 deficiency induces persistent alterations in immune phenotype and function in adult offspring. Furthermore, when stimulated, lymphocytes from developmental vitamin D-deficient rats exhibited a pro-inflammatory phenotype ([@bib0550]). Human population studies have also indicated a significant association between developmental vitamin D deficiency and risk of schizophrenia ([@bib0865]). ### 3.1.3. Prenatal infection {#sec0035} Some of the strongest support toward the developmental etiology of schizophrenia has come from its association with prenatal infection ([@bib0130], [@bib0690]). Accumulating epidemiological data indicates that 38--46% of cases of schizophrenia may have an association with prenatal infection ([@bib0190]). Maternal infection by *Toxoplasma gondii*, influenza, rubella, herpes simplex virus, and cytomegalovirus has consistently been associated with the risk of schizophrenia in adult offspring. Although some of these infectious agents, e.g. rubella, *T. gondii*, cytomegalovirus, etc. can cross the placenta and directly affect the developing fetus, the detrimental effects of infection is known to be elicited by activation of maternal immune system ([@bib0595]). Converging evidence suggests that immuno-inflammatory responses together with downstream oxidative and nitrosative stress pathway are crucial in mediating prenatal infection induced neurodevelopmental abnormality ([@bib0025]). The detrimental effects of redox dysregulation, excessive generation of reactive oxygen species (ROS)/reactive nitrogen species (RNS), etc. on developing brain after prenatal infection are quite well known ([@bib0365], [@bib0760]). Further, intrauterine infection/inflammation evokes the expression of inducible nitric oxide synthase (iNOS) and nitric oxide (NO) in the brain and leads to oligodendrocyte injury in the developing brain ([@bib1085]). Prenatal immune activation leads to changes in the levels of multiple neurotransmitter systems ([@bib1250]). In this context, dopamine abnormalities are one of the most enduring etiological hypotheses of schizophrenia. Emerging findings from animal studies reveal that prenatal exposure to infection and/or inflammation has long-lasting effects on dopaminergic structures and functions ([@bib0015], [@bib0415], [@bib0960], [@bib1220], [@bib1310]). In mice, a viral-like acute phase response during early/mid gestation causes a complex pattern of age-dependent structural abnormalities in the mesoaccumbal and nigrostriatal dopamine system ([@bib1220]). Notably, cytokines are known to modulate dopaminergic neurotransmission and dopamine can also modulate immune response by influencing the cytokine network ([@bib1100], [@bib0255]). Inflammation and oxidative/nitrosative stress strongly influence each other, and increased levels of serum oxidative stress markers and inflammatory cytokines have been reported in schizophrenia patients ([@bib0985]). Based on this understanding, it is likely that prenatal infection induced activation of IO&NS pathway will have an important downstream role in dopaminergic neurotoxicity in schizophrenia. A recent study has demonstrated increased protein oxidation in the dopamine-rich areas of the prefrontal cortex in schizophrenia patients, suggesting the importance of interactions between oxidative stress and dopamine in the pathophysiology of schizophrenia ([@bib0700]). Psychosis is hypothesized to be a hyperdopaminergic state, and models of increased dopamine transmission are linked to increases in ROS ([@bib1265]). Dopamine antagonists can block increased oxidative stress and apoptotic potential ([@bib0945]). Additionally, dopamine metabolism itself, by generating its metabolites like dopamine-semiquinone and dopamine-quinone leads to the generation of ROS within CNS ([@bib0510]). Importantly, in neurodegenerative disorders quinone formation has been shown to be linked with inflammation and oxidative stress. Furthermore, oxidative stress is shown to inhibit the uptake of dopamine through post-translational modification of the dopamine transporter ([@bib0695]). There is crosstalk between other transmitters germane to schizophrenia such as GABA and serotonin and inflammation ([@bib1235]). Prenatal immune activation in mouse with poly I:C is shown to cause maturation-dependent alterations in prefrontal GABAergic gene expression, implying long-term effect of prenatal immune insults on the GABAergic system ([@bib1025]). Agents enhancing GABAergic transmission such as gabapentin down-regulate inflammatory and oxidative stress markers ([@bib0345], [@bib0515], [@bib1295]). Similarly, serotonergic agents cause down-regulation of inflammatory markers in some studies ([@bib1130], [@bib0780]). Further relevant to schizophrenia pathogenesis, prenatal immune activation has shown to modulate hippocampal NMDAR function by interacting with stress and the stress hormone corticosterone at adolescence ([@bib0200]). Taken together, maternal infection induced IO&NS responses can adversely affect pre- and/or peri-natal outcomes and even lead to neurodegeneration or teratogenesis ([@bib1240], [@bib1275]), indicating that inflammation-induced effects may be associated with a direct damage by oxidative/nitrosative stress. ### 3.1.4. Prenatal stress {#sec0040} Prenatal stress-induced changes are found to confer high risk for diverse outcomes in offspring including schizophrenia ([@bib0835]). For example, gestational exposure to prenatal stress or stress hormones contributes to deficits in hippocampal structure and function, as well as the neurotransmitter and immune systems ([@bib0775], [@bib0090], [@bib0840]). The enhanced production of corticosterone or maternal exposure to exogenous glucocorticoids due to chronic exposures to stress has direct influence on fetal brain development and plasticity as well as programming of hypothalamic-pituitary-adrenocortical (HPA) axis ([@bib0715], [@bib0670]). Prenatal stress increases serotonin 2A and decreases mGlu2 expression in frontal cortex, suggesting schizophrenia-like alterations of serotonin 2A and metabotropic glutamate 2 receptors ([@bib0580]). Prenatal stress also affects development of the hippocampal alpha 7 nicotinic acetyl choline receptor (nAChRs) in adult offspring ([@bib1070]). It is interesting to note that repeated variable prenatal stress during critical period of fetal brain development reprograms the response of the HPA axis to acute stress and alters pre- and postsynaptic gene expression that might impact synaptic function in the offspring ([@bib0705]). Studies have demonstrated that both down-regulation of glutamate decarboxylase 67 (GAD67) and maternal exposure to severe stress can increase the risk of schizophrenia in offspring. This is further supported by a study where heterozygous deletion in GAD67 enhances maternal and fetal stress vulnerability ([@bib1150]). ### 3.1.5. Paternal age {#sec0045} Epidemiological studies have provided robust evidence of association between advanced paternal age and the enhanced risk of schizophrenia ([@bib0870], [@bib0170], [@bib0890]). One of the contributing mechanisms that has consistently been implicated in paternal age associated risk of schizophrenia is de novo mutations, which are found to be more abundant in sperm of older men ([@bib1030], [@bib0830]). In a comprehensive study on 78 families, an increase in two de novo mutations per year of advancing paternal age was observed, indicating a linear relationship between paternal age and the number of de novo mutations in offspring ([@bib0720]). Recently, "selfish spermatogonial selection" which means clonal expansion of germs cells carrying pathogenic mutations has been proposed to be a mechanism for association between advanced paternal age and the risk of schizophrenia ([@bib0500]). The clonal expansion leads to relative enrichment of de novo mutations that essentially explain the effect of advanced paternal age on schizophrenia. 3.2. Endogenous risk factors {#sec0050} ---------------------------- ### 3.2.1. Obstetric events {#sec0055} Obstetric complications are well established predictors of risk of schizophrenia ([@bib0230]). Although the term "obstetric events" includes a wide range of complications, the most commonly found complications are fetal growth retardation, fetal hypoxia and prenatal complications. Low birth weight and reduced length at birth are crude indicators of IUGR or abnormally slow fetal growth. Fetal growth restriction increases risk of later schizophrenia ([@bib0390]). Smaller head circumference and small gestational age have been associated with schizophrenia, indicating the implications of fetal growth restriction ([@bib0875]). Although the underlying mechanism of growth restriction and risk of schizophrenia is poorly understood, accumulating evidence suggests that a combination of factors such as environment, placental, and genetic enable expression of a particular mental health outcome. Recent animal data suggests that IUGR leads to metabolic alterations in the fetal brain by influencing neuronal viability, inflammatory regulation, energy metabolism and oxidative stress pathways ([@bib1185]). Suggesting the importance of genetic effects on fetal growth restriction, it has been found that mothers with schizophrenia also have higher rates of low birth weight among offspring ([@bib0630]). However, in a study on same-sex twins discordant for schizophrenia, it was observed that within these twin pairs, low birth weight and smaller head circumference were significantly associated with later development of schizophrenia, indicating that the fetal growth restriction may be independent of familial factors ([@bib0925]). Fetal hypoxia is an environmental risk factor of schizophrenia in the offspring. Hypoxia-associated obstetric complications and the increased risk of schizophrenia is a replicated finding ([@bib1300], [@bib0310]). It is evident that interactions between neuronal genes and molecular regulators of oxygen could lead to hypoxia induced neurodevelopmental abnormality and subsequent risk of neuropsychiatric disorders. Fetal hypoxia also appears to influence the severity of certain neuropathological attributes of schizophrenia, such as hippocampal and cortical gray matter reduction ([@bib0235]). Importantly, such changes in schizophrenia are proposed to be mediated by interactions between schizophrenia susceptibility genes and hypoxia related regulatory processes ([@bib1165], [@bib1060]). A large number of schizophrenia susceptibility genes viz. AKT1, BDNF, COMT, DNTBP1, NOTCH4, NRG1, PRODH, RELN, RGS4 are regulated by hypoxia ([@bib1055]). Furthermore, genes regulated by hypoxia also interact with serious obstetric complications and influence schizophrenia risk ([@bib0910]). ### 3.2.2. Gut microbiota and developmental immune modulation {#sec0060} Microbial colonization of gut is an evolutionary process, which influences both metabolic and immune functions, particularly during early neonatal life. However, recent findings indicate that microbial colonization or contact of the fetus with maternal gut microbiota may start in utero as bacteria from maternal gastrointestinal tract (GIT) have been detected in amniotic fluid, placental and fetal membranes ([@bib0565]). Pregnancy is associated with a profound alteration of maternal gut microbiota, and the changes in gut microbiota composition have been associated with weight gain, altered biochemical and metabolic parameters during pregnancy ([@bib1050], [@bib0725]). Obstetric complications, long known to increase risk for schizophrenia, may in theory be mediated through higher rates of cesarean section, which limits the infants accrual of a normal microbiota from the maternal genitourinary tract through birth ([@bib0125]). This has been hypothesized to also be a factor in autism ([@bib0300]). This implies the possible roles of the gut microbiota composition in mother\'s health during pregnancy as well as on maternal-fetal interactions, influencing the infant\'s health later in life. The perinatal colonization of the gut by microbes contributes to developmental programming of gut homeostasis, angiogenesis and immune competence ([@bib0585], [@bib0815]). There are extensive interactions between the gut microbiota and the host immune system in the newborns and the development of the adaptive immune system is regulated by bacterial colonization of the gut. It is now evident that time variation in microbial colonization of the gut during early life shapes future immune system reactivity. For example, delayed colonization seems to exert permanent changes in the immune system ([@bib0540]). A range of environmental factors including cesarean section, toxins, infectious agents, diet and stress could affect the gut microbiome during this critical developmental period. Disruption of the developing miocrobiome could lead to long-term changes in immune and psychological development. Furthermore, disturbances in the microbiota can also result in dysregulation of adaptive immune cells, intestinal inflammation and potentially enhance the host\'s susceptibility to immune-mediated diseases. The gut microbiota communicates with the brain and affects normal brain development and subsequent adult behavior ([@bib0350]). Gut--brain communication occurs through direct neuronal, immune-related signaling and hormonal pathways. Exposure to many of the afore-mentioned environmental factors during both the pre-and post-natal periods has been established as risk factors for disorders including schizophrenia and autism. One of the widely accepted underlying mechanisms that mediates the interaction between such environmental factors and host as well as gut--brain communication is through immunological signaling involving cytokines. Animal studies suggest that perturbation in the composition of gut microbiota influence the risk of depression and anxiety-like behaviors ([@bib0905], [@bib0355]). In autistic children, there appears to be a distinct and less diverse gut microbe composition ([@bib0665]). A recent study highlighted the importance of gastrointestinal inflammation in schizophrenia pathology ([@bib1075]). Translocation of commensal microbiota such as *Citrobacter koseri*, *Hafnia alvei*, *Pseudomonas aeruginosa*, *Pseudomonas putida*, *Klebsiella pneumonia* and *Morganella morganii* across the gastrointestinal barrier can lead to an antibody response to bacterial translocation, and lysozyme production and consequently leads to persistent low-grade inflammation and increases the risk of depression ([@bib0820], [@bib0825]). Bacterial translocation causes an imbalanced and activated innate immune state ([@bib1080]). Quality data linking these processes to schizophrenia is awaited. ### 3.2.3. Role of placenta in fetal brain development {#sec0065} There is a wide appreciation that perturbations in the maternal hormonal and nutrient environment have deleterious effects on the developing brain and subsequently influence susceptibility to a range of metabolic, neurodevelopmental and psychiatric disorders in adulthood ([@bib0440], [@bib0055]). Importantly, such effects can be transmitted to the fetus by changes in placental function ([@bib0650]), thereby indicating a direct role of placenta in developmental programming of brain and behavior. Placental insufficiency causes intrauterine growth restriction ([@bib0070]). In mice, food deprivation for 24 h on days E12--E13 has been shown to affect placental gene expression, through which placenta provides protection to the fetal brain ([@bib0160]). Interestingly, recent studies suggest that the placenta can convert maternal tryptophan into the neurotransmitter serotonin (5-hydroxytryptamine or 5-HT), which is essential for developing mouse forebrain at mid-gestation ([@bib0140]). This implies a significant role of tryptophan catabolites (TRYCATs) in the placenta in modulating fetal brain development and well as affecting long-term brain function ([@bib0145], [@bib0495]). However, developmental disruption of 5-HT signaling in specific regions of fetal brain causes abnormal wiring of major axonal pathways, altered cell division and laminar organization in the neocortex ([@bib1210], [@bib0135]). Altered brain 5-HT signaling is consistently linked to several psychiatric disorders. Considering the significant role of 5-HT in essential neurodevelopmental processes, genetic or environmental perturbations directly affecting placental tryptophan metabolism can lead to neurodevelopmental abnormality, and therefore contribute to the developmental origin of diverse psychiatric disorders. Prenatal exposure to maternal infection leads to increased expression of pro-inflammatory cytokines such as IL-6 and TNF-α in the placenta and also leads to fetal inflammation which in turn can cause organ damage and potentially downstream developmental deficiencies ([@bib1155], [@bib0245]). Activation of the maternal immune system induces endocrine changes in the placenta, especially IL-6 dependent disruption of the growth hormone-insulin-like growth factor (GH-IGF) axis i.e. decreased levels of GH, IGFI and IGFBP3 ([@bib0595]). Taken together, these observations suggest that the placenta might act as a central mediator of fetal programming of the TRYCAT pathway that may underlie developmental origin of adult psychiatric disorders including schizophrenia ([@bib0495]). ### 3.2.4. Prenatal neuroendocrine pathway {#sec0070} Hormones are established environment-dependent coordinators of the developing "neuro-endocrine-immune network". However, non-physiological concentrations of hormones during crucial phases of embryogenesis can act as 'endogenous functional teratogens'. This is exemplified by fetal and neonatal hyper-insulinism in the offspring of diabetic mothers. Diabetes mellitus is the most common metabolic complication during gestation and is known to have neurodevelopmental sequelae. A number of studies have highlighted relationship between maternal diabetes and the risk of schizophrenia in the offspring ([@bib1180]). Importantly, the risk of schizophrenia in the offspring born to diabetic mothers could be induced by hyperglycemia and mediated by hypoxia, inflammation and oxidative stress. Endocrine disruption as etiological component is theoretically implicated in schizophrenia. Estrogen is proposed to be a potential mediator of brain functions during development and adulthood. Estrogen protects brain cells against injury from oxidative stress, inflammation and apoptosis ([@bib0045], [@bib0085]). Estrogen is relevant to schizophrenia due to its significant effects on synaptogenesis, neurogenesis, neuroendocrine and inflammatory processes. Recently, combinations of both estrogen and selective estrogen receptor modulators with antipsychotics have been shown to decrease positive and negative symptoms significantly in women with chronic schizophrenia ([@bib0750], [@bib0755], [@bib0475]). However, prenatal exposure to excess estrogen could increase the risk of schizophrenia in the offspring through certain mechanisms ([@bib0165]). For example, estrogen has been found to increase susceptibility to certain viral infections by altering innate and adaptive immune responses, such as by reducing CD4 T-cell responses and by inhibiting type 1 interferon production and dendritic cell maturation ([@bib0400], [@bib0980]). Interestingly, many such viruses including borna disease virus, mumps, influenza, corona viruses, etc. are associated with increased risk of schizophrenia. A summary of prenatal endogenous and exogenous risk factors and/or mechanisms leading to psychopathology has been presented in [Fig. 2](#fig0010){ref-type="fig"} .Fig. 2Summary of potential prenatal endogenous and exogenous risk factors and/or mechanisms leading to psychopathology. 3.3. Specificity of the risk factors of schizophrenia {#sec0075} ----------------------------------------------------- Converging evidence from multiple animal studies indicate that prenatal infection can act as a "neurodevelopmental disease primer', and this seems to be a shared environmental risk factor across a number of chronic mental illnesses (as reviewed in [@bib0885], [@bib0745]). It is now widely appreciated that activation of IO& NS pathway due to prenatal infection is one of the underlying mechanisms in the pathogenesis of various neuropsychiatric disorders like schizophrenia, autism, and bipolar disorder ([@bib0025]; [@bib0040]). Although the specificity of prenatal infection on subsequent disease or symptoms is yet to be precisely understood, it is likely to be influenced by the interactions between genetic and additional environmental factors. For example, major histocompatibility complex (MHC) molecules have been shown to play important roles during neurodevelopment ([@bib0250]). Neurons exposed to MIA were found to have higher levels of MHC class I expression and lower synapse density ([@bib0395]), suggesting that such effects of MIA on MHCI expression could affect activity-dependent plasticity and synaptic pruning during crucial phases of neurodevelopment. A plethora of genetic studies have consistently implicated MHC as a risk determinant of schizophrenia ([@bib0335]); however, this has not been observed in bipolar disorder ([@bib0290]). Furthermore, MHC is also found to be associated with various schizophrenia risk factors such as obstetric complications, season of birth, infection and so on ([@bib0335]), thus implying that genetic components along with environmental factors might determine the specificity of disease pathogenesis. 4. Neurodevelopmental origin of schizophrenia: a unifying genetic--epigenetic--environmental pathway {#sec0080} ==================================================================================================== 4.1. Genomic imprinting {#sec0085} ----------------------- The expression of some genes in the human genome depends on whether they are located on the maternal or paternal chromosome. This is accomplished through a highly powerful epigenetic phenomenon called genomic imprinting, which is a significant form of gene regulation. Several of the imprinted genes that are expressed from only one of the parental chromosomes have novel roles in the normal placental development and fetal growth ([@bib0455]). Although most known imprinted genes are expressed during placentation, a considerable number are expressed primarily or entirely in the brain ([@bib0315]). Brain development is strongly influenced by the epigenetic regulation of imprinted genes, and there is increasing evidence that imprinted genes can influence neurogenetic and psychiatric manifestations by affecting neurodevelopmental processes ([@bib0620], [@bib1245]). A recent family based trio study has suggested the occurrence of imprinting of schizophrenia candidate gene, GABRB2 that codes for GABA(A) receptor β(2) subunit ([@bib1010]). Evidence of imprinting in schizophrenia has also been observed in other chromosomal locus, such as DLK1-DIO3 region in chromosome 14q32 ([@bib0465]). 4.2. Feto-maternal genetic conflict {#sec0090} ----------------------------------- Fetal--maternal genetic conflict may potentially contribute to many disorders related to pregnancy including preeclampsia ([@bib0560]). Interestingly, a combination of maternal/fetal genetic predisposition and environmental factors has been implicated as potential risk factors for preeclampsia ([@bib1035]). This may be based on MHC and Rh gene compatibility/incompatibility ([@bib0535]). With respect to HLA antigens, when there is a lack of maternal recognition i.e. if paternally derived HLAs do not differ from maternal HLAs, this may have an adverse effect on reproductive outcomes ([@bib0935]). HLA matching, in couples or between mother and fetus, increases prenatal or obstetric complications including preeclampsia, low birth weight, and can even lead to fetal loss ([@bib0940]). The first indication of the importance of HLA matching in neurodevelopmental disorders came from a study showing that parents of children with autism are more likely to share at least one HLA-A, -B, or -C antigen in common compared to the parents of unaffected children ([@bib1125]). Recently, the significance of maternal--fetal genotype incompatibility at HLA loci in schizophrenia susceptibility has been described, in that maternal--fetal matching at the HLA-B locus increases fetal risk for developing schizophrenia ([@bib0970], [@bib0965]). Maternal--fetal HLA-B matching may increase the risk of schizophrenia only in female offspring ([@bib0275]). Various roles of MHC molecules in neuronal development by in vitro and in vivo studies and risk of schizophrenia conferred by MHC variants have been well supported by genome wide association studies ([@bib0335], [@bib0855], [@bib0290]). These observations led to the hypothesis that maternal--fetal genotype combinations at HLA loci, by modulating feto-maternal immune responses, can affect fetal neurodevelopment and subsequently enhance one\'s risk of developing schizophrenia. 4.3. Developmental epigenetic modifications {#sec0095} ------------------------------------------- Environmental signals during prenatal life lead to adverse long-term effects. Although the mechanisms underlying such risk remain poorly understood, emerging studies both in animals and humans indicate that maternal exposure to infection, stress, drugs or toxins alter epigenetic programming in regulatory as well as growth-related genes ([@bib0485]). There is a growing recognition that prenatal epigenetic dysregulation due to such adverse in utero environments not only affect fetal brain development, but also predispose an individual to neurodevelopmental, behavioral and neurocognitive deficits later in life ([@bib0425], [@bib0995]). A recent study has demonstrated that MIA causes epigenetic changes in adolescent mouse brain ([@bib0075]). The impact of prenatal stress on the modification of epigenetic signatures during critical periods of fetal brain development has been elucidated by several recent studies. Exposure to gestational stress increases DNA methylation at the glucocorticoid receptor (GR) gene promoter and reduces methylation at the corticotrophin-releasing factor (CRF) gene promoter in hypothalamic tissue in adult male mice ([@bib0900]). Prenatal stress down-regulates 11β-hydroxysteroid dehydrogenase type 2 (HSD11β2) that converts cortisone/corticosterone into inactive metabolites through DNA methylation in the placenta and fetal brain ([@bib0655]). Further evidence suggests that the offspring of mice exposed to gestational stress have an altered transcriptomic brain profile of genes related to development, axonal guidance and neuropathology due to up-regulation and down-regulation of certain microRNA (miRNA) ([@bib1305]). In addition, prenatal stress through disruption of DNA methylation network affects GABAergic interneurons associated with schizophrenia-like phenotypes ([@bib0845]). Maternal cigarette smoking during pregnancy is a common hazard affecting key pathways crucial for proper fetal growth and development. Recent data suggests that maternal cigarette smoking during pregnancy can lead to alteration in DNA methylation and expression of microRNA ([@bib1160]). Importantly, maternal smoking during pregnancy has been found to be associated with increased DNA methylation in a key gene involved in brain development, brain-derived neurotrophic factor (BDNF) in adolescent offspring ([@bib1140]). Maternal diet during pregnancy can also affect brain development and function by modifying the fetal epigenome. Persistent epigenetic changes were observed in the offspring exposed to prenatal famine. Peri-conceptional exposure to famine during Dutch Hunger Winter was associated with reduced DNA methylation of Insulin-like growth factor 2 (IGF2), which is a key factor in human growth and development ([@bib0555]). Dietary intake of methyl-group (choline, methionine, and folate) containing nutrients during critical phases of prenatal development can alter the epigenomic profile of the developing offspring, thereby resulting in altered fetal and lifelong changes in gene expression. Unbalanced maternal diet during pregnancy alters DNA methylation in important genes controlling glucocorticoid function and fetal growth ([@bib0380]). Recently, an association between maternal prenatal nutrition and schizophrenia risk in the offspring via epigenetic effects has been highlighted ([@bib0710]). 4.4. Gene--environment interactions {#sec0100} ----------------------------------- A large number of environmental factors have been proposed to cause neurodevelopmental abnormalities and confer enhanced risk of schizophrenia in offspring. It is now becoming apparent that genetic factors potentially modulate the adverse effects of environmental stressors on neurodevelopment during pre- and peri-natal periods as well as behavioral outcomes in the offspring ([@bib0240]). Importantly, the impact of gene--environmental interactions on perinatal programming of schizophrenia is a critical issue in prenatal adversity induced changes in neurodevelopment. The genetic liability of perinatal environmental adversities has been exemplified by various studies. A recent study has shown how maternal cytomegalovirus infection influences the risk of schizophrenia in the offspring by interacting with the genotype of CTNNA3 gene of the progeny ([@bib0150]). The additive or interactive effects of hypoxia with genetic factors in influencing liability to schizophrenia have been reported ([@bib0225]). Animal studies have demonstrated that MIA during pregnancy with polyI:C of mice having mutation in DISC1, a consistently replicated risk gene of schizophrenia exacerbate schizophrenia-like phenotypes ([@bib0005], [@bib0810]). Another study has shown that prenatal immune activation in mice with mutations in Nurr1, a transcription factor crucial for dopaminergic development leads to neuropathological consequences, such as locomotor hyperactivity, deficits in sensorimotor gating, and attentional impairments ([@bib1225]). Furthermore, recent animal studies demonstrated that combined effects of neonatal immune activation and mutant DISC1 on the risk of schizophrenia- like behavioral phenotypes in the offspring ([@bib0610]). One of the important pathways that can be modulated by perinatal adverse conditions is HPA axis activity in offspring. Perinatal exposure to excess glucocorticoids can lead to a persistently altered HPA axis and subsequently enhance the risk of schizophrenia ([@bib0600]). Disturbed genome regulation due to adverse perinatal environmental conditions has been shown to contribute to schizophrenia risk and pathophysiology. Recent observations indicate that prenatal infection increases the associated risk of schizophrenia if occurring in offspring with a family history of psychosis; this implies an interaction of genetic vulnerability with prenatal infection ([@bib0280]). This could possibly be mediated by certain gene loci such as MHC, HLA-G, toll-like receptor (TLR) 3 and 4 that are known to have predominant expression during pregnancy ([@bib0330], [@bib1200]). Taken together, such evidence suggests that fetal immune programming may have profound consequences on brain and behavior, and is associated with the adult presentation of schizophrenia. 5. Prenatal events and neuroprogression in schizophrenia {#sec0105} ======================================================== There is now strong evidence of pathological reorganization of the central nervous system along the course of severe mental disorder. Neuroimaging studies have demonstrated significant volume reduction of certain specific regions in the brain of schizophrenia patients, providing evidence of neuroprogressive processes including neurodegeneration and neuronal apoptosis ([@bib0575]). There is a wide appreciation that neuroprogression is partially mediated by IO&NS pathways, apoptosis and mitochondrial energy dysregulation ([@bib0105], [@bib0025]). The generation of ROS and RNS has both physiologic and pathologic roles in the placenta, embryo and fetus; the developing brain is highly vulnerable to ROS and RNS. Amongst the several causative mechanisms of fetal origin of adult diseases, a significant role of oxidative and nitrosative stress (O&NS) in fetal programming is predominant ([@bib1135]). It is now evident that a number of factors such as prenatal hypoxia, maternal under-and over-nutrition, excessive glucocorticoid exposure can induce O&NS process during pregnancy and subsequently lead to neuronal death/brain injury ([@bib0615]). In a recent study it was observed that chronic fetal hypoxia induced brain injury is associated with altered nitric oxide synthase activity ([@bib0375]). It is now becoming evident that a range of prenatal adversities can lead to heightened maternal inflammation, which can contribute to placental hypoxia and O&NS process underlying altered fetal growth and development. For example, the effects of prenatal malnutrition have also been found to be mediated by pro-inflammatory factors ([@bib1090]). Fetal hypoxemia can cause fetal inflammatory response syndrome as well as fetal brain injury by up-regulating inflammatory cytokine cascade ([@bib0530]). It is interesting to note that melatonin, an endogenously produced indoleamine from the pineal gland, acts as an antioxidant, free radical scavenger and anti-inflammatory molecule ([@bib1020]). Importantly, melatonin has a role in redox modification in fetal programming and reverses oxidative stress during prenatal period. Further, melatonin influences epigenetic modifications associated with developmental programming ([@bib0730]). Melatonin might play a pivotal role in epigenetic modifications induced by maternal stress, maternal under-nutrition or IUGR ([@bib0270]). As melatonin levels and melatonin circadian rhythm are significantly decreased in schizophrenia ([@bib0035]), a role of melatonin in neuroprogressive pathways seems credible. Molecular hydrogen is an odorless and tasteless gas, and this bioactive molecule has various biological attributes including anti-inflammatory, anti-apoptotic and anti-oxidative effects ([@bib0930]). Animal studies have indicated that inhalation of hydrogen decreased acute lung inflammation and hydrogen enriched water decreased the production of ROS ([@bib1260], [@bib0680]). Importantly, a protective role of hydrogen on fetal brain injury during maternal hypoxia has also been recently highlighted ([@bib0805]). Some intestinal bacteria, e.g. *Escherichia coli* can produce a remarkable amount of molecular hydrogen and under certain circumstances can suppress inflammation ([@bib0660]). Considering the above biological properties of molecular hydrogen and their relevance in neurodevelopment and function including neuroprogressive changes, implications of molecular hydrogen have been highlighted recently in bipolar disorder and schizophrenia ([@bib0480]). 6. Implications: prediction and prevention {#sec0110} ========================================== Understanding the pathophysiological basis of schizophrenia from the perspectives of MIA with fetal programming aberrations permits potential translational implications as well. Over the past few years multiple blood-based, imaging, neurophysiologic and neurocognitive biomarkers have been identified for schizophrenia ([@bib0260], [@bib0795], [@bib1280]). However, there seems to be a considerable imprecision in the nosology of biomarkers that have been identified in schizophrenia. Recent conceptualization suggests that a classification of biomarkers based on six categories such as risk, diagnosis/trait, state or acuity, stage, treatment response and prognosis may be applicable to neuropsychiatric disorders ([@bib0320]). Several studies have consistently shown altered levels of markers of inflammation, oxidative stress, metabolism and hormonal status in the first onset schizophrenia patients and have suggested the potential implications of such risk biomarkers of schizophrenia ([@bib0520], [@bib0990]). Considering the ever growing support toward the neurodevelopmental origin of schizophrenia, identification of biomarkers bearing signatures of events from as early as embryonic development or birth would be of considerable interest in predicting risk as well as treatment responses. A recent methylome-wide association study examining blood based biomarkers in schizophrenia has shown changes in genes that reflect the effects of environmental insults related to hypoxia, infection etc. ([@bib0010]), suggesting that pathogenic events might be preserved in the methylome. Such understanding essentially highlights the importance of environmental adversities on fetal epigenome, therefore, early epigenomic profiling might be one of the areas which could offer potential schizophrenia risk predicting biomarker. In addition, elevated levels of inflammatory marker such as maternal C-reactive protein, TNF-α and IL-8 have robustly been associated with increased risk of schizophrenia in adult offspring ([@bib0195], [@bib0175], [@bib0215]). Such maternal inflammatory markers could also be considered as potential risk biomarker in predicting schizophrenia early in life. Maternal stress is an established risk factor of schizophrenia. Recently, O-GlcNAc transferase (OGT) has been demonstrated to be a placental biomarker of maternal stress ([@bib0590]). This biomarker might also be useful in predicting gestational stress exposure and risk of schizophrenia in adult offspring. Impaired sensory gating indicative of aberrant cerebral inhibition is a documented component of the pathogenesis of schizophrenia ([@bib1215]). Significantly, prenatal MIA can result in sensory gating aberrations ([@bib1040]). Hence, within the pathogenetic paradigm of aberrant fetal programming due to prenatal immune disturbances, certain biomarkers like P50, that are indicative of sensory gating deficits, especially in combination with other significant biomarkers like P85, P300, mismatch negativity and eye movement aberrations (smooth pursuit as well as antisaccades) ([@bib0095]), with subsequent analyses of markers for persistent immune activation (cytokines, lymphocyte functions ([@bib0385]) as well as microglia imaging using positron emission tomography ([@bib0370]) might facilitate reliable prediction of risk for schizophrenia. Prevention of psychiatric disorders of neurodevelopmental origin such as schizophrenia is becoming a challenging issue ([@bib0640]). Public health approaches including control of infectious diseases, improvements in obstetric and neonatal care and nutritional supplementation might be beneficial. For instance, choline, an essential nutrient, has been shown to improve immune parameters with associated adaptive modulation of cognitive functions ([@bib0860], [@bib0785], [@bib0790]). Choline plays a critical role in optimal neurodevelopment ([@bib9055]). Interestingly, several maternal factors that elevate schizophrenia risk can lead to decreased availability of choline during fetal development ([@bib1285]). Recently, a randomized placebo-controlled clinical trial examined 100 healthy pregnant women with dietary phosphatidylcholine supplementation that was initiated in the second trimester until the time of delivery. After birth, infants were given 100 mg of phosphatidylcholine in an oral suspension once daily or placebo ([@bib1045]). In comparison with placebo-treated infants, choline-treated infants demonstrated better sensory gating as examined by P50 response ([@bib9045]). Choline supplementation was shown to enhance sensory gating measures in healthy adults as well ([@bib0740]). Availability of safe interventions like choline supplementation is worth exploring with large scale studies to ascertain potential preventive interventions. 7. Limitations of neurodevelopmental theory and future directions {#sec0115} ================================================================= The neurodevelopmental origin of schizophrenia although has become one of the most influential etiologic theories in recent times; however, there are certain limitations of this theory. Amongst the range of factors and mechanisms that are shown to alter the developmental trajectories, the association of certain risk factors/mechanisms remain weak and needs to be replicated. Various risk factors seem to mediate a large number of mechanisms that impair crucial phases of fetal development and result in a similar set of behavioral outcomes in the offspring. It is essential to classify the risk factors that confer predominant risk and delineate their mechanistic basis. The current understanding suggests that developmental modulation during embryogenesis by various factors and mechanisms could potentially predict behavioral outcomes in the offspring during adolescence or early adulthood and these precede the onset of illness. Early onset schizophrenia is a severe form of schizophrenia that occurs during childhood or adolescence and is often chronic and persistently debilitating. However, the frequency of early onset schizophrenia has been reported to be around 4% ([@bib0220]). Although converging functional genomics suggests involvement of multiple genes in neurodevelopment, certain genes are found to confer susceptibility to various neurodevelopmental disorders. Furthermore, most of the risk genes have a variable pattern of expression and different effects at different developmental stages. Some of the genes exhibit preferential expression in the fetal brain and high gene expression occurring during fetal development can be reversed in early postnatal life ([@bib0285]). In addition, progressive brain changes have also been reported in chronic adult patients with schizophrenia ([@bib0265]). Therefore, it is important to study the effect of various risk factors including genes as well as neuroprogressive changes at different time points during the lifespan to gain more insights on the pathogenesis of schizophrenia rather than focussing emphatically on neurodevelopmental insults. The developmental neuroinflammation has also been contradicted by a recent study where neonates who later developed schizophrenia had unaltered levels of 17 inflammatory markers, thus refuting an association between neonatal inflammation and risk of schizophrenia ([@bib0915]). Future research should pay more focus to establish 'missing heritability' aspects and disentangle the different effects of prenatal environmental and genetic exposures. The early-life epigenetic disruption also plays pivotal role in the neurodevelopmental origin of schizophrenia. This notion is experimentally supported by methylomic profiling of human brain tissue of schizophrenia patients, identifying disease-associated differential DNA methylation in the prefrontal cortex and discrete modules of co-methylated loci associated with the disorder that are significantly enriched for genes involved in neurodevelopmental processes ([@bib1000]). Improved understanding of the genetic as well as epigenetic contribution to environmental risk mediation will offer possible candidate for primary prevention and early intervention strategies. Combined studies of prenatal exposures and neurodevelopmental disorders with simultaneous evaluation of neural and immune systems might delineate the mechanisms enhancing individual vulnerability or resilience to neurodevelopmental disorders. The importance of early life programming as a target for prevention of mental disorders in the offspring has recently been highlighted by [@bib0790] and several recommendations were made with respect to public health and clinical implications. 8. Conclusion {#sec0120} ============= There is strong evidence that schizophrenia is a complex multi-factorial disorder and no one factor seems to be of solitary significance in the genesis of schizophrenia. Among the various etiological models tested so far, gene--environmental interactions appear to be most widely appreciated one in schizophrenia. The adverse effects of certain risk factors are found to be linked to genetic components, therefore, it is essential to identify specific genetic components that confer major risk to schizophrenia in combination with environmental stressors. The series of neurological insults rendered by an array of environmental factors during gestation reviewed above strongly level schizophrenia to a disorder of neurodevelopment and significantly highlight the implications of fetal programming of schizophrenia. Such compelling evidence calls upon researchers to translate these findings into interventions designed to prevent mental disorders. Conflict of interest {#sec0125} ==================== The authors declare that they have no conflicts of interest to disclose germane to this paper. MB is supported by a NHMRC Senior Principal Research Fellowship 1059660.

Influenza-associated acute encephalopathy has been described in children, and results in a high frequency of neurologic sequelae and death. Altered consciousness, disorientation, and seizures occur within a few days after the onset of fever and respiratory symptoms ([@R1]*--*[@R3]). In some patients, symptoms are transient but in others rapid progression to necrotizing encephalitis, deep coma, and death may occur ([@R1]*--*[@R3]). Cases in adults are infrequently reported and remain poorly characterized, although the more complex clinical scenarios in adults may have hindered case recognition ([@R1]*,*[@R4]*--*[@R6]). The pathogenesis is unclear, but a hyperactivated cytokine response, rather than viral invasion, is believed responsible in most childhood cases ([@R1]*--*[@R5]). We describe 3 cases of acute encephalopathy associated with influenza A infection in adults. The clinical, virologic, immunologic findings (cytokines in plasma and cerebrospinal fluid \[CSF\]), and CSF penetration of oseltamivir for these cases are reported. The Study ========= At Prince of Wales Hospital, Hong Kong ([@R7]), from January 2007 through August 2008, influenza infection was diagnosed for \>460 hospitalized adult patients for whom acute febrile respiratory illnesses had been diagnosed. Nasopharyngeal aspiration and immunofluorescence assays (IFA) were used for rapid diagnosis of influenza A and B infection, confirmed by virus isolation. Thirteen (2.8%) patients had signs of confusion or altered consciousness, together with fever and respiratory symptoms (mean ± SD age 77.7 ± 8.8 years). We studied 3 patients from whom CSF was obtained for analysis, and who fulfilled the definition of influenza-associated acute encephalopathy (altered mental status \>24 hours within 5 days of influenza onset and without alternative explanation) ([@R1]*,*[@R2]*,*[@R4]*--*[@R6]). Nasopharyngeal aspirates were subjected to IFA, virus isolation, and subsequent subtyping ([@R7]). CSF specimens were subjected to virus isolation using MDCK cells, and reverse transcription--PCR to detect influenza virus RNA by using H1/H3 subtype-specific primers. Herpes simplex virus, herpes zoster virus, and enterovirus DNA/RNA was detected using PCRs ([Technical Appendix](#SD1){ref-type="local-data"}). CSF and plasma samples collected on the same day were analyzed simultaneously for the concentrations of 11 cytokines/chemokines by bead-based multiplex flow cytometry. Their assay methods and plasma reference ranges (established from \>100 healthy persons) have been described ([Technical Appendix](#SD1){ref-type="local-data"}) ([@R7]). In CSF, in patients without central nervous system (CNS) disease/infection, cytokines/chemokines are either undetectable (e.g., interleukin-6 \[IL-6\], CXCL8/IL-8, CXCL10/IP-10, CXCL9/MIG) or present at low levels (e.g., CCL2/MCP-1) ([@R8]*--*[@R10]). Concentrations of oseltamivir phosphate (OP) and its biologically active metabolite oseltamivir carboxylate (OC) were measured in CSF and plasma taken simultaneously from 1 patient who received concurrent treatment, using tandem mass spectrometry ([@R11]). The assay methods have been described ([Technical Appendix](#SD1){ref-type="local-data"}). The clinical and virologic findings are summarized in [Table 1](#T1){ref-type="table"}. All case-patients were elderly (72--86 years of age), but none were known to have neuropsychiatric illness, dementia, or to be taking psychotropic medication. None had received updated influenza vaccination ([@R6]). Confusion and altered consciousness developed in patients 1 and 2 one to 2 days after the onset of fever and cough. These patients had no meningismus, focal neurologic deficit, hypotension, respiratory distress, or metabolic disturbances. Brain computed tomography (CT) scans showed no acute cerebral lesion. CSF analyses showed no bacterial or viral pathogen or pleocytosis. Oseltamivir was given to patient 2 only when influenza A was later confirmed by nasopharnygeal aspirate/IFA; patient 1 did not receive antiviral treatment. Both patients recovered in the next 2 days. Patient 3 had fever, severe chronic obstructive pulmonary disease exacerbation requiring noninvasive ventilatory support, complicated by acute coronary syndrome. He was given oseltamivir, 75 mg 2×/day, after influenza A infection was confirmed. Agitation and confusion developed in the patient on day 3--4 of illness (onset after the third dose of oseltamivir), despite resolution of the patient's respiratory failure. These symptoms were followed by involuntary, tremulous movements involving all 4 limbs, while at rest and during movement. Brain CT scan was normal. Electroencephalogram showed generalized slowing. Oseltamivir was stopped after the ninth dose, but tremor persisted. CSF analyses showed no pathogen or pleocytosis. The patient's symptoms resolved in the next 3--4 days without sequelae. ###### Clinical and laboratory findings in 3 patients with acute encephalopathy associated with influenza infection, Prince of Wales Hospital, Hong Kong\* Clinical and laboratory findings Patient 1 Patient 2 Patient 3 -------------------------------------- ------------------------------------------------------------------------- --------------------------------------------------------------------------------------- ---------------------------------------------------------------------------------------------- Age, y/sex 76/M 86/F 72/M Concurrent illnesses Ischemic heart disease Diabetes mellitus, hypertension COPD Influenza vaccination within 6 mo None None None Symptoms on examination Fever \>38°C, cough, disorientation, incoherent speech, mental dullness Fever \>38°C, cough, delirious, impaired consciousness, did not follow verbal command Fever \>38°C, cough, disorientation, agitation, incoherent speech, involuntary 4-limb tremor Focal neurologic sign or meningism Absent Absent Absent Chest radiograph, consolidation Absent Absent Absent Antiviral (oseltamvir) None Given Given Outcome (duration of encephalopathy) Recovered (2--3 d) Recovered (3--4 d) Recovered (6--7 d) Brain CT scan (noncontrast) Normal Old ischemic changes; known small, calcified meningioma Normal Virus isolated from NPA Seasonal (H1N1) 2008 Subtype H3N2 Subtype H3N2 CSF testing results Opening pressure, cm H~2~O 11 9 14 Cell count (x 10^6^/L) 1 -- 0 Glucose, mmol/L 4.2 7.4 3.7 Protein, g/L 0.46 0.47 0.16 Virus isolated None None None RT-PCR for H3 and H1 influenza virus Negative Negative Negative Bacterial culture Negative Negative Negative Others HSV, HZV, and enterovirus PCR negative HSV, HZV, and enterovirus PCR negative HSV PCR negative \*COPD, chronic obstructive pulmonary disease; CT, computed tomographic scan; NPA, nasopharyngeal aspirate; CSF, cerebrospinal fluid; RT-PCR, reverse transcription--PCR; HSV, herpes simples virus; HZV, herpes zoster virus. In all cases, there was no hypoglycemia, and liver and renal function test results were normal. C-reactive protein level was elevated in all cases. For patient 3, an electroencephalogram was performed and showed generalized slowing of background consistent with moderate encephalopathic change (similar to that observed in septic encephalopathy) ([@R1],[@R6]). Findings are consistent with previous reports on adult cases of influenza-associated encephalopathy: patients are all unvaccinated, pleocytosis and cerebral imaging abnormalities (even with magnetic resonance imaging) are usually absent, and symptoms are generally self-limiting ([@R1],[@R6]). Most reports have mentioned influenza A as a cause of encephalopathy, and more commonly subtype H3N2 ([@R1]--[@R6]). Despite apparently normal CSF findings, high concentrations of cytokines/chemokines were detected in the CSF and plasma specimens of all patients ([Table 2](#T2){ref-type="table"}). Plasma concentrations of IL-6, CXCL8/IL-8, CXCL10/IP-10, CCL2/MCP-1, and CXCL9/MIG were elevated at median values of 2.0, 2.8, 11.9, 3.7, and 2.1× the upper limits of their respective reference ranges (comparable to or higher than that observed in other hospitalized influenza patients) ([Table 2](#T2){ref-type="table"}) ([@R7]). Other cytokines were not elevated ([@R4]*,*[@R7]). In their CSF, IL-6, CXCL8/IL-8, CXCL10/IP-10, and CCL2/MCP-1 were consistently detected, and were elevated at median values of 2.6, 15.0, 3.4, and 20.0 × the upper limits of their respective plasma reference ranges. The CSF/plasma concentration ratios of CXCL8/IL-8 and CCL2/MCP-1 were \>3 (median CSF/plasma ratio 5.4 and 8.0, respectively). ###### Cytokine and chemokine concentrations in CSF and plasma samples from 3 patients with acute encephalopathy associated with influenza A infection, Prince of Wales Hospital, Hong Kong\* Cytokine or chemokine Reference range, pg/mL CSF/plasma cytokine concentration, pg/mL (ratio) ----------------------- ------------------------ -------------------------------------------------- -------------------- ------------------- IL-6† \<3.1 8.0/6.3 (1.3) 11.6/35.1 (0.3) 2.2/5.9 (0.4) CXCL8/IL-8‡ \<5.0 84.0/15.5 (5.4) 74.8/13.8 (5.4) 21.9/6.3 (3.5) CXCL10/IP-10† 202--1,480 15,374/102,019 (0.2) 5,101/17,594 (0.3) 1,371/1,550 (0.9) CCL2/MCP-1‡ \< 10-57 996/82 (12.1) 1,287/336 (3.8) -- CXCL9/MIG 48--482 11,58/14,001 (0.1) 70/333 (0.2) 145/1,019 (0.1) IFN-γ \<15.6 UD/14.4 4.7/10.1 0.4/2.0 IL-12p70 \<7.8 1.5/UD 1.3/UD UD/UD TNF-α \<10.0 1.7/1.4 UD/1.2 UD/UD IL-10 \<7.8 2.5/2.2 UD/7.3 UD/1.7 IL-1β \<3.9 UD/UD UD/3.7 UD/UD CCL5/RANTES 4,382--18,783 4/2,507 14/1,609 1.3/814 \*CSF, cerebrospinal fluid; --, test not done due to inadequate sample; UD, undetectable (i.e., below the detection limit of the cytokine/chemokine assay). Cytokines: Interleukin (IL)--1β, IL-6, IL-10, IL-12p70, tumor necrosis factor α (TNF-α). Chemokines: CXCL8/IL-8, monokine induced by interferon-γ (IFN- γ) (CXCL9/MIG), IFN-γ--inducible protein-10 (CXCL10/IP-10), monocyte chemoattractant protein--1 (CCL2/MCP-1), and regulated upon activation normal T cell--expressed and secreted (CCL5/RANTES). The plasma reference ranges are established from \>100 healthy adults. The assay sensitivities of IL-1β, IL-6, IL-10, IL-12p70, TNF-α, IL8, MIG, IP-10, MCP-1, RANTES, and IFN-γ are 2.5, 3.3, 3.7, 1.9, 7.2, 0.2, 2.5, 2.8, 2.7, 1.0, and 7.1 pg/mL, respectively. Coefficients of variation are all \<10%. In an earlier study involving 39 adult influenza patients hospitalized with cardio-respiratory complications ([@R8]), the median (interquartile range) plasma concentrations of IL-6, IL-8, IP-10, MCP-1, and MIG were 10.6 (4.2--18.4), 5.4 (2.5--8.7), 7,043.0 (4,025.1--1,2381.1), 76.5 (49.5-97.0), and 992.1 (499.1--1,992.3) pg/mL, respectively. In CSF, in subjects without neurologic disease/infection, these cytokines/chemokines are either undetectable or present at low levels ([@R9]--[@R11]). In a pediatrics influenza cohort, CSF cytokine levels were substantially higher in encephalopathy cases when compared to those with febrile seizure; CSF/plasma concentration was \<1 ([@R9]).
†CSF cytokine concentrations above plasma reference ranges.
‡CSF/plasma cytokine concentration ratio consistently \>3 (3.5--12.1), in addition to CSF cytokine concentrations being above the plasma reference ranges. For IFN-γ, IL-12p70, TNF-α, IL-10, IL-1β and RANTES, because of their low/undetectable levels, the CSF/plasma ratios were not calculated. CSF specimens from patients 1 and 2 were collected at the peak of symptoms, and before antiviral treatment (if given); CSF from patient 3 was collected when persistent tremor developed 18 hours after the ninth dose of oseltamivir; the drug was stopped afterward. Simultaneous CSF and plasma OC and OP concentrations were determined for patient 3, as symptoms progressed at 18 h after oseltamivir. The concentrations (mean ± SD) of OC in duplicate CSF and plasma samples were 18.3 ± 0.9 ng/mL and 143.8 ± 3.3 ng/mL, respectively; the CSF/plasma concentration ratio was 12%--13%. The OP plasma concentration was 1.05 ± 0.03 ng/mL; it was not detectable in the CSF. Conclusions =========== We report 3 adults with acute encephalopathy (altered consciousness, confusion) associated with influenza. High CSF and blood cytokine/chemokine (CXCL8/IL-8, CCL2/MCP-1, IL-6, CXCL10/IP-10) levels were detected. No evidence of direct viral neuroinvasion was found. All patients recovered rapidly without sequelae ([@R1]*,*[@R6]). Our findings agree with studies of influenza-associated encephalopathy in children. Influenza virus is rarely detected in the CSF, and pleocytosis is often absent ([@R1]*,*[@R2]*,*[@R4]*--*[@R6]). High levels of cytokines (e.g., IL-6, soluble tumor necrosis factor receptor 1) can be consistently found in CSF/blood specimens, correlating with disease severity and outcomes (hyperactivated cytokine response is absent in febrile seizure associated with influenza) ([@R2]*--*[@R4]*,*[@R8]). We found a broader range of cytokines/chemokines being activated ([@R7]); for certain cytokines (CXCL8/IL-8, CCL2/MCP-1), the CSF concentrations were 3× those in plasma. IL-6, CXCL8/IL-8, CCL2/MCP-1 and CXCL10/IP-10 have been shown to play pathogenic roles in CNS viral infections, cerebral injury, and acute brain syndrome in susceptible patients ([@R7]*,*[@R9]*,*[@R12]). The high CSF/plasma ratios suggest that for some cytokines, activation within the CNS might have occurred along with respiratory-tract and systemic productions (cytokines are not detected in CSF normally; ([Table 2](#T2){ref-type="table"}) ([@R4]*,*[@R7]*--*[@R10]*,*[@R12]). Resident macrophages/monocytes, astrocytes, microglial and endothelial cells in the CNS are shown to release cytokines/chemokines when stimulated by viral/influenza infection; activation mechanisms without involving overt CNS invasion have been suggested ([@R1]*,*[@R4]*,*[@R9]*,*[@R12]*--*[@R14]). Cytokines may cause direct neurotoxic effects, cerebral metabolism changes, or breakdown of the blood-brain-barrier (endothelial injury) to produce symptoms ([@R1]*--*[@R4]*,*[@R8]*,*[@R12]*--*[@R14]). Whether early viral suppression by antivirals can lead to attenuation of these cytokine responses and better outcomes warrants further study ([@R7]). We measured oseltamivir concentrations because of the concerns over its neuropsychiatric side-effects in children and adolescents. However, only the active metabolite (OC) was detected in the CSF of patient 3; the CSF/plasma concentration ratio was 12%--13% (18.3/143.8 ng/mL) at 18-hours postdose. This degree of CSF penetration is similar to that observed among healthy patients, with a Cmax CSF/plasma concentration ratio of 3.5% (at ≈8 hours), and a ratio of ≈10% at 18 hours (concentration-time profiles for plasma/CSF differ). Assuming a similar ratio, the CSF OP concentration would have fallen below the assay's detection limit (0.25 ng/mL) by 18 hours ([@R11]*,*[@R15]). The low CSF drug-penetration, together with high cytokines in CSF and symptom progression despite drug withdrawal suggest that the manifestations of patient 3 may have been disease-related. Symptoms developed in patients 1 and 2 without antiviral exposure. Further investigations on the CNS effects of oseltamivir in the clinical setting are needed.. Our study is limited by the small patient number and the lack of feasibility in obtaining CSF for study/comparison in influenza patients without neurologic symptoms. Further studies on the clinical spectrum of influenza encephalopathy and encephalitis in adults ([@R1]*,*[@R6]) and their pathogenesis are indicated. In conclusion, acute encephalopathy may occur in adults with influenza. Exuberant cytokine/chemokine response may play an important role in its pathogenesis. Supplementary Material ====================== ###### Technical Appendix Acute Encephalopathy Associated with Influenza A Infection in Adults *Suggested citation for this article*: Lee N, Wong CK, Chan PKS, Lindegardh N, White NJ, Hayden FG, et al. Acute encephalopathy associated with influenza A infection in adults. Emerg Infect Dis \[serial on the Internet\]. 2010 Jan \[*date cited*\]. Available from <http://www.cdc.gov/EID/content/16/1/139.htm> We thank Rity Wong for clerical assistance. This study was supported by a research fund from the Department of Medicine and Therapeutics, The Chinese University of Hong Kong. Dr Lee is an associate professor in infectious diseases at The Chinese University of Hong Kong. His primary research interest is severe emerging infections, including severe acute respiratory syndrome and influenza.

Introduction ============ Children are exposed to diverse types of evaluative feedback regarding their behavior, intelligence, and performance in their social lives. Parents and teachers often provide children with positive feedback to enhance their intrinsic motivation and boost their self-esteem ([@B16]; [@B5]). However, an increasing body of evidence suggests that praise is not always beneficial for children (see [@B9]; [@B16] for a review). For example, unlike process-related feedback (e.g., "you worked hard"), receiving personal feedback (e.g., "you are smart") following success, when children's positive performance is attributed to their personal traits, could foster resistance to subsequent mistakes ([@B30]; [@B19]). Notably, [@B16] suggested that the effects of praise on motivation depend on not only the praise content (e.g., process-related feedback or personal feedback) but also the recipients' characteristics and interpretation of the praise. Evidence accumulated from research conducted thus far has revealed that some variables, such as perceived autonomy and performance standards, affect children's motivation (e.g., [@B7]; see [@B16] for a review). However, few studies have examined insincerity, another key variable, or have assessed the manner in which children interpret and respond to praise that leaves room for doubt. Developmental studies have shown that young children can produce and understand insincere utterances from around the age of 4 or 5 years ([@B1]; [@B3]). Although previous literature has not directly examined the sensitivity of children to insincerity in praise, this evidence leads us to speculate that children around the age of four or five come to notice insincerity in praise that is given in an inappropriate situation (i.e., failure situation), and begin to display negative responses to insincere praise. The question as to how and when children should be praised is one of the main concerns of adults engaged in parenting and education; however, very few attempts have been made to examine the impact of praise in situations where children fail and evaluate their own performance as poor. Indeed, evaluations received from others are not always congruent with children's own self-evaluation; for example, parents and teachers sometimes praise children when the children themselves feel that they did not perform well on a task (i.e., when they have failed). A recent study that examined children's responses to inflated praise found that school-aged children with low self-esteem sought fewer challenges after they received inflated praise ([@B5]). Crucially, this finding indicated that children with low self-esteem might evaluate their own performance as poor compared to children with high self-esteem, and the disagreement between internal low evaluation and external high evaluation (i.e., inflated praise) might cause the negative response to the praise, because they might recognize the praise as insincere. This study also focused on children's theory of mind that could be one of the key factors that serve to constrain their interpretation of and responses to insincere praise. Theory of mind is defined as the ability to attribute mental states to the self and others to interpret behavior ([@B31]). False-belief tasks are commonly used to measure young children's theory of mind development, and a large body of evidence has shown that children pass first-order false-belief tasks between the ages of 3 and 5 years ([@B39]). Individual differences in theory of mind in early childhood influence children's social lives in diverse ways ([@B17]). Theory of mind allows children to comprehend non-literal speech, such as irony (e.g., [@B11]), white lies ([@B4]), and antisocial lies ([@B34]), by understanding the intention behind the words that are spoken. That is, as theory of mind develops, children can infer the intention underlying non-literal speech and detect its true meaning. Children need to interpret the meaning of praise and compliments in situations where they receive insincere praise. Therefore, in situations where children are praised following failure, they may use theory of mind to determine the reason for the praise. Ambiguity in praise, which is characteristic of feedback in Japanese culture, is another key topic that has been overlooked in previous research. The Japanese language is highly contextualized, and the content of the speaker's discourse is ambiguous, rather than explicit ([@B26]). In Japan, ambiguity is observed not only in language but also in parenting and education. Japanese mothers adopt an "absorbent childrearing" approach, where they do not teach children explicitly but encourage them to understand situations and behave autonomously ([@B2]). In addition, Japanese preschool and kindergarten educators use the *machi no hoiku* (childrearing of waiting) strategy, whereby they hold back and wait, allowing children the opportunity to handle problems on their own with minimal assistance from teachers or adults ([@B35]; [@B13]), and this strategy might support both positive and negative self-reflection. Against a cultural background involving this language system and these educational beliefs, Japanese children might be more likely to receive ambiguous evaluative feedback without an apparent reason (e.g., "great," "good," or "okay") relative to Western children. To my knowledge, [@B20] conducted the first study to examine ambiguous praise, where the experimenter suddenly approached 2- to 4-year-old children while they were engaged in free play at preschool and provided them with praise with no indication of the reason for which praise had been provided ("great!" or "*sugoi!*" in Japanese). Children received this feedback in situations where they had not performed activities that was appropriate for evaluation (e.g., success, failure), and their facial and verbal reactions were observed. The results showed that most 2-year-old children and some 3-year-old children smiled and showed agreement with the sudden ambiguous praise, while some 3-year-old children and most 4-year-old children exhibited confusion at the sudden ambiguous praise, with dubiety in their facial expressions, and half of them asked about the reason for the praise. These findings suggest that at approximately 4 years of age, children begin to consider not only praise itself but also the reason for praise. Although [@B20] observation indicated that 4-year-old children were confused by ambiguous praise and cared about the reason for which praise had been offered, children's responses to and interpretation of ambiguous praise in situations involving activities that evaluation is appropriate for (e.g., success, failure) remained unclear. The overall aim of the current study was to examine young Japanese children's interpretations of and responses to insincere praise, which involves ambiguity in the reason of the praise, and explore the association between these responses and the maturity of their theory of mind. The first hypothesis was that participants would display a negative emotional response, low performance rating, and low motivation to insincere praise, similar to that observed in the failure situation without feedback, since they notice that the praise is insincere. The second hypothesis was that most participants would attribute the receipt of insincere praise following failure to their efforts rather than outcomes or their personal attributes. This hypothesis was proposed based upon the Japanese cultural belief system, which emphasizes the importance of hard work, effort, and persistence in self-improvement (e.g., [@B36]; [@B15]). Their tendency to attribute the insincere praise to their efforts may be the basis for low motivation after receiving insincere praise, as this includes the message that "they had already made a sufficient effort." The third hypothesis concerned the relationship between theory of mind and the child's responses to insincere praise. It was hypothesized that children with mature theory of mind would be more likely to report negative responses (i.e., negative emotion, low levels of self-rated performance, and weaker motivation for perseverance) after receiving insincere praise following failure, relative to children with less mature theory of mind. The proposed reason for this was that, whereas children without mature theory of mind could consider the praise as sincere if they take it at face value, children who possess mature theory of mind could interpret the hidden message behind the praise and consider it invalid. Therefore, when children possess mature theory of mind, their positive emotion, performance rating, and motivation for further challenge could be reduced because they speculate the reason of the praise in relation to the failure situation. Methods {#s1} ======= Participants ------------ Seventy-two Japanese children aged 4 years and 7 months to 6 years and 9 months participated (mean age = 5.70 years, *SD* = 0.61; 24 boys, 48 girls). All participants were native Japanese speakers attending one of two preschools. An ethics approval for psychological research was not required at the time the research was planned and conducted as per the author's institution guidelines (Meiji Gakuin University) as well as national regulations. Though there was no opportunity to submit an ethics application at that time, the author carried out this study in accordance with the recommendations of the Japanese Psychological Association Ethical Principles of Psychologists with careful attention to ethical considerations to uphold high ethical standards. All participants' parents gave the author written informed consent in accordance with the Declaration of Helsinki. Although detailed demographic information regarding the children could not be obtained because of school privacy policies, the sample was recruited predominantly from middle-class families. Materials and Procedure ----------------------- The children were tested individually in quiet rooms at the preschools. They performed a failure-praise task, theory of mind tasks, and a verbal ability test. The order of the tasks was counterbalanced. ### Failure-Praise Task A new task designed to measure individual differences in responses to praise following failure was developed for this study. Children's responses were examined by using stories in which they appear as a protagonist puppet; then, they were asked how they felt in the stories. The task structure and procedure were based on sensitivity-to-criticism tasks that have been used to assess children's responses to criticism in both Eastern and Western countries (e.g., [@B6]; [@B21], [@B22]; [@B27], [@B28]; [@B29]). The stories that were used in the tasks are shown in **Appendix [1](#A1){ref-type="app"}**. In the failure-praise task, the experimenter read two puppet-based stories (a "no-feedback story" and a "praise story") aloud and directed the children to act out the role of the main character simultaneously, using a puppet. The task flow is shown in **Figure [1](#F1){ref-type="fig"}**. Before the experimenter read the stories, the children were instructed to choose one of four puppets to represent themselves and introduced it to another puppet that represented the teacher. The children were told that the puppet child (representing the participant) and the puppet teacher were going to play a block design game that involved the use of four bicolored blocks from the Wechsler Preschool and Primary Scale of Intelligence Block Design test ([@B38]). In the block design game, the puppet teacher showed the puppet child a design on a picture card. The puppet child was asked to recreate the design using the blocks, within a limited amount of time. {#F1} To ensure that the children understood the rules of the game, they were provided with similar, but easier, examples and allowed to practice. Upon completion of the two practice trials, they were told the no-feedback story (control story) first, followed by the praise story (test story). This was to avoid the effect of praise on the response in the no-feedback story (control story) and to confirm how they evaluate their failure itself. The two stories contained descriptions of very similar events. In both stories, the child failed to recreate a design, which was shown on a picture design card, within a limited amount of time. The difference between the two stories was that praise was provided in the praise story but not the no-feedback story. The no-feedback story was used as the control story, mainly to confirm whether children were dissatisfied with their performance when they failed. In the no-feedback story, the children were told that the main character (i.e., the child) played a block design game and tried hard to recreate a complex design with the blocks but failed to do so within the time limit. In the praise story, the main character played another block design game with a similar degree of difficulty and tried hard to recreate the design but failed to do so. However, in the praise story, the teacher offered the main character insincere praise without a reason (i.e., "*sugoi-ne!*" in Japanese that means "great!" in English), in response to the work. The two picture design cards used in the stories were counterbalanced. The following three questions were posed after each of the stories: (Q1) Emotional response: "I want to know how you feel about what happened in this block design game" followed by, "Do you feel happy or not? Do you feel sad or not? Do you feel angry or not?" They answered each question after being shown illustrations of each emotional facial expression. To create an index of a general positive emotional response, one point was awarded for each positive emotion (i.e., happy, not sad, and not angry), and a summed score was calculated (scores ranged from 0 to 3). (Q2) Performance rating: "Think again about everything that happened in this block design game. Should you get a circle (good) or a cross (not good) for what you did?" Children were asked to assign a grade to their block design by pointing to a circle (which means "good" in Japan) or a cross. One point was awarded for a positive evaluation (good), and no points were awarded for a negative evaluation (not good). (Q3) Motivation: "Think again about everything that happened with this block design game. If the teacher asked you to try the game again, would you like to play the difficult game or the easier game?" The children were asked this question to measure their motivation for attempting to perform challenging activities in the stories. They were asked to choose between a card depicting a complex design and a card depicting a straightforward design. One point was awarded for choice of the challenging activity, and no points were awarded for choice of the easy activity. The children were asked another question about interpretation of praise in the praise story, as the praise provided in the story was ambiguous, and they could have interpreted it in diverse ways: (Q4) Interpretation of praise: "Why did the teacher say, 'great!' (*sugoi-ne!*)?" Children's responses were classified into three categories: "outcome," "effort," and "other" (i.e., other interpretation, do not know, and no response). If the children did not answer, they were asked the following forced-choice question: "Did she offer the praise because you did good work (outcome) or because you tried hard (effort)?" Six of the 72 children (8.33%) were asked the forced-choice question. The author and a second coder coded participants' interpretation of the praise. Interrater agreement, established using Cohen's kappa, was high (*K* = 0.85). Disagreements were resolved via discussion. In the debriefing, the children and the experimenter played a block design game without a puppet, if the children wanted to try to recreate the design themselves, and they talked about successful experiences in their daily lives. ### Theory of Mind Tasks The children completed theory of mind tasks that included two first-order false-belief tasks ([@B12]) and two second-order false-belief tasks ([@B33]). The example of stories and questions used in the theory of mind tasks are shown in **Appendix [2](#A2){ref-type="app"}**. These tasks have demonstrated good test-related reliability ([@B18]). Each of the first-order false-belief tasks included two memory control questions, a first-order false-belief question, and a reality question. If the children answered all four questions correctly, they received one point. Each of the second-order false-belief tasks included a first-order false-belief question, a reality question, a second-order false-belief question, and two memory control questions. The children answered the second-order false-belief question and memory control questions only if they had answered the first two questions (the first-order false-belief and reality questions) correctly. If children answered both the first-order and reality questions correctly, they received one point for understanding first-order false beliefs, and if they answered the second-order false-belief and memory control questions correctly, they received one additional point for understanding second-order false beliefs. Therefore, overall theory of mind scores ranged from 0 to 6 (i.e., four points for understanding first-order false beliefs and two points for understanding second-order false beliefs). ### Verbal Ability Test Children's vocabulary was assessed to control for verbal ability in the analysis, as this ability is typically positively related to their performance in theory of mind tasks ([@B25]). The Picture Vocabulary Test-Revised (PVT-R; [@B37]), which was standardized for use with Japanese samples, was used to measure children's receptive vocabulary. The PVT-R is the Japanese version of the Peabody Picture Vocabulary Test ([@B8]). In the task, children were shown an array of four pictures on each page of a stimulus book and asked to select an appropriate picture, which was named by the experimenter, from the four pictures. The standardized score was used for the following analyses. Results ======= Data from 52 children (18 boys and 34 girls) who evaluated their work in the no-feedback story as "not good" (low performance rating after failure without feedback) were included in the main analysis examining their responses to the praise in the praise story. This data selection process was essential because young children tend to be unrealistically overconfident of their ability (e.g., [@B32]) and could be unable to recognize their failure in the story. **Table [1](#T1){ref-type="table"}** shows the means and standard deviations for data regarding age, PVT-R scores, theory of mind scores, and scores for each measure in the failure-praise tasks, for the 52 children. ###### Descriptive statistics. Measure *M* (*SD*) Range ----------------------- --------------- -------- Age in months 68.67 (7.42) 55--81 Verbal ability 26.31 (10.52) 6--46 Theory of mind 3.38 (1.92) 0--6 **No-feedback story** Emotional response 0.92 (0.65) 0--3 Performance rating 0.00 (0.00) 0 Motivation 0.73 (0.45) 0--1 **Praise story** Emotional response 2.33 (0.96) 0--3 Performance rating 0.58 (0.50) 0--1 Motivation 0.58 (0.50) 0--1 Note. n = 52. Concerning an overview of the analysis, first, children's responses to the praise story were compared with their responses to the no-feedback story by using Wilcoxon signed-rank tests to test the first hypothesis. To test the second hypothesis, their interpretation of the praise following failure was coded. Lastly, a logistic regression analysis concerning their responses to praise following failure per their theory of mind development was conducted to test the third hypothesis. Children's Response to Insincere Praise --------------------------------------- To test the first hypothesis, which expected that young children would show a negative response to insincere praise, Wilcoxon signed-rank tests for positive emotion, self-rated performance, and motivation were performed. Children's levels of positive emotion (*Z* = −5.53, *p* \< 0.001), and self-rated performance (*Z* = −5.48, *p* \< 0.001) were significantly higher relative to those observed in the no-feedback story. Their motivation was significantly lower relative to those observed in the no-feedback story (*Z* = −2.53, *p* = 0.01). Children's Interpretation of Insincere Praise --------------------------------------------- To test the second hypothesis, which predicted that most participants would attribute the receipt of insincere and ambiguous praise following failure to their own effort in a story, children's interpretation of the praise was coded. The interpretation provided by seven of the 52 children was categorized as "other," and the remaining 45 children were classified into one of two groups based on their attribution of the praise: 27 were classified into the outcome-interpretation group (e.g., children who believed that the teacher had offered praise because the recreated puzzle had been good or they had been able to recreate half of the puzzle) and 18 were classified into the effort-interpretation group (e.g., children who believed that the teacher had offered praise because they had worked hard or done their best to complete the challenging puzzle). Age, *t*(43) = −0.74, *ns*; PVT-R scores, *t*(43) = 1.24, *ns*; and theory of mind scores *t*(43) = 0.13, *ns* did not differ significantly between outcome- and effort-interpretation groups. Theory of Mind and Responses to Insincere Praise ------------------------------------------------ A logistic regression analysis was performed to test the third hypothesis, which predicted that children with mature theory of mind would respond more negatively to the praise following failure, relative to children with less mature theory of mind. Coding of the interpretation of insincere praise following failure revealed that there were two major types of interpretation (i.e., effort and outcome)^[1](#fn01){ref-type="fn"}^. Thus, whether the interaction between theory of mind and the interpretation were significantly associated with their response to insincere praise was also analyzed. First, theory of mind scores were centered, and interpretation of praise was coded into effort (+1) and outcome (−1). To conduct the logistic regression analysis, children's scores for the emotional response question were divided into two categories reflecting higher (2 and 3 points) and lower (0 and 1 points) levels of positive emotion. As dependent variables, children's responses to praise (emotional response, performance rating, and motivation) were examined in each separate analysis. In each analysis, age and PVT-R scores were entered in the first step, theory of mind scores and interpretation of the praise were entered in the second step, and the interaction terms of theory of mind scores and interpretation of the praise were entered in the final step. **Table [2](#T2){ref-type="table"}** shows the results of the logistic regression analysis for the three dependent variables. For emotional response, the model was not significant, *χ*^2^(5) = 5.00, *p* = 0.42. For performance rating, the model was significant, *χ*^2^(5) = 14.56, *p* = 0.01, and the main effect of theory of mind was significant, *B* = −0.87, Wald = 6.90, *p* = 0.01, indicating that mature theory of mind was related to relatively lower performance ratings after insincere praise. Although the model for motivation did not reach significance, *χ*^2^(5) = 9.11, *p* = 0.11, the interaction between theory of mind and interpretation of the praise approached significance, *B* = −0.38, Wald = 3.80, *p* = 0.05. A simple slope analysis revealed that theory of mind score was negatively associated with motivation after praise following failure in the effort-interpretation group, *B* = −0.60, Wald = 3.11, *p* = 0.08, indicating that there was a tendency that children with mature theory of mind displayed lower motivation after insincere praise, relative to children with less mature theory of mind in this group. On the other hand, for children in the outcome-interpretation group, theory of mind was not significantly related to motivation, *B* = 0.15, Wald = 0.22, *p* = 0.64. ###### Results of the logistic regression analysis (final model). Variable *B* *SE* Wald *df* Odds ratio %ΔOdds 95% *CI* ---------------------------------- ------- ------ ---------- ------ ------------ -------- ------------ **Emotional response** Age 0.05 0.08 0.37 1 1.05 +5 0.90--1.23 Verbal ability 0.07 0.06 1.40 1 1.08 +8 0.95--1.22 ToM −0.10 0.29 0.11 1 0.91 −9 0.51--1.61 Interpretation (effort, outcome) −0.04 0.44 0.01 1 0.96 −4 0.41--2.26 ToM × Interpretation −0.07 0.23 0.10 1 0.93 −7 0.60--1.45 **Performance rating** Age −0.12 0.09 1.86 1 0.89 −11 0.75--1.05 Verbal ability 0.17 0.07 6.68^∗∗^ 1 1.19 +19 1.04--1.35 ToM −0.87 0.33 6.90^∗∗^ 1 0.42 −58 0.22--0.80 Interpretation (effort, outcome) 0.50 0.44 1.28 1 1.64 +64 0.70--3.88 ToM × Interpretation −0.26 0.24 1.24 1 0.77 −23 0.49--1.22 **Motivation** Age 0.07 0.07 1.05 1 1.08 +8 0.94--1.24 Verbal ability 0.05 0.05 0.89 1 1.05 +5 0.95--1.16 ToM −0.23 0.27 0.74 1 0.80 −20 0.47--1.34 Interpretation (effort, outcome) −0.33 0.37 0.82 1 0.72 −28 0.35--1.47 ToM × Interpretation −0.38 0.20 3.80^†^ 1 0.68 −32 0.47--1.00 Note 1. n = 45, † p \< 0.10, ∗ p \< 0.05; SE, standard error; CI, confidence interval; ToM, Theory of mind. Note 2. Emotional response: model fit, χ 2 \(5\) = 5.00, p = 0.42, -2 Log likelihood = 40.04, Nagelkerke R 2 = 0.17, goodness-of-fit test (Hosmer--Lemeshow); χ 2 \(7\) = 9.88, p = 0.20. Note 3. Performance rating: model fit, χ 2 \(5\) = 14.56, p = 0.01, -2 Log likelihood = 46.01, Nagelkerke R 2 = 0.37, goodness-of-fit test (Hosmer--Lemeshow); χ 2 \(7\) = 6.13, p = 0.53. Note 4. Motivation: model fit, χ 2 \(5\) = 9.11, p = 0.11, -2 Log likelihood = 52.18, Nagelkerke R 2 = 0.25, goodness-of-fit test (Hosmer--Lemeshow); χ 2 \(7\) = 11.96, p = 0.10. Discussion ========== This study examined young Japanese children's interpretations of and responses to insincere praise following failure and explored the association between these responses and the maturity of their theory of mind. Children heard stories involving the receipt of insincere praise from a teacher following failure and answered questions regarding their interpretation of the praise, emotion, performance rating, and motivation. Regarding children's responses to the praise, the results showed that the children were generally happy to receive insincere praise following failure. Inconsistent with the first hypothesis, they exhibited positive emotions and their self-rated performance levels increased after the praise. The findings indicate that some of the children increased the positivity of their own evaluations to match those of the teacher. However, motivation to accept the challenge of completing a more demanding task was lower in the situation involving failure followed by praise relative to that observed in the situation involving failure without feedback, suggesting that the receipt of insincere praise from a teacher following failure could negatively affect young children's motivation. This finding suggests that young children began to notice insincerity in praise in failure situations to some extent, but they were still affected by its positive face value. Regarding children's interpretation of insincere praise without an apparent reason (i.e., "great!" or "*sugoi-ne!*" in Japanese), the results did not support the second hypothesis, which predicted that most of the children would attribute the praise to their effort rather than outcomes or their personal attributes. Approximately 52% of the children attributed the praise to outcome, and approximately 34% attributed the praise to their effort. It should be noted that theory of mind scores did not differ significantly between the outcome-interpretation and effort-interpretation groups, confirming that the difference in interpretation did not reflect children's ability to understand others' mind; rather, it may reflect one's attribution tendencies. Although there are no empirical data to explain the reason that about half of the children attributed the praise to outcome even in the failure situation, one interpretation was that some young children still have a tendency to interpret their work positively and with optimism; specifically, they focus on the beauty of their work rather than the components that require improvement ([@B23]). The main topic of interest in this study was the association between individual differences in the maturity of children's theory of mind and their responses to insincere praise. The results provided partial support for the third hypothesis, which predicted that children with mature theory of mind would receive and respond to the praise negatively. Children with mature theory of mind reported lower levels of self-rated performance, relative to those reported by children with less mature theory of mind after praise in both interpretation groups. One possible explanation for this finding is that children who attributed the praise to their outcome might have questioned what was in the teacher's mind in a failure situation; for example, "She praised me; however, did she really think I gave a great performance?," and this could have led them to give themselves a low performance rating. Children who attributed the praise to their effort might have interpreted that the teacher had no choice but to give praise for their effort because their performance was unsatisfactory, and this could have led them to give themselves a low performance rating. In the effort-interpretation group, children with mature theory of mind also exhibited lower levels of motivation for perseverance relative to those observed for children with less mature theory of mind, following praise. They might have doubted the teacher's praise; e.g., "She praised me; however, did she really think I tried well?" and might have answered, "Yes, I trust her. She thinks I try hard even if I have failed; therefore, I cannot be expected to do more than this." This kind of thought might have decreased their motivation to perform more challenging tasks. The findings revealed that children's interpretation of insincere praise differed depending on the maturity of their theory of mind. Specifically, although theory of mind did not account for the children's positive emotional responses, it explained the observation of lower levels of self-rated performance following the praise in those with mature theory of mind in both groups. Furthermore, theory of mind tended to be associated with weaker motivation for meeting the challenge of the task following the praise in the effort-interpretation group. It could be that young children with mature theory of mind recognize the uncertainty in the teacher's insincere praise and therefore considered the feedback as unreliable. One could also consider the idea that some children interpreted "great (*sugoi-ne*)" as a sarcastic remark. However, it is unlikely that the participants interpreted the praise as sarcasm. This is because most children attributed the praise to their effort or outcome and did not reported sarcastic intentions, and most of them exhibited positive emotion following the praise. The remark would be more likely be interpreted as sarcastic at the age of approximately 9 years, when children understand irony ([@B10]; [@B24]). Most previous studies examining children's responses to praise have focused on positive feedback following success ([@B30]; [@B19]; [@B16]; see also [@B5]). Within this body of literature, the main strength of this study is its novel focus on insincere praise following failure, and the new finding that process-related praise, which was believed to have positive effect on children's self-evaluation and motivation (cf. [@B19]), has a negative effect in some cases. This implies that praise for effort in a failure situation sometimes conveys the message that adults are sufficiently satisfied with children's effort, thus decreasing children's motivation for challenges. The evidence of the current study indicated that the effects of insincere praise could differ depending on the maturity of children's theory of mind. The findings also have practical implications that there is a need for fine-grained observation of individual children's understanding of mental states prior to the provision of evaluative feedback. As theory of mind dramatically develops in young children, parents and teachers should avoid using thoughtless positive feedback and assess how children evaluate their own work. This study was the first to perform an empirical examination of insincere praise and revealed the association between young children's negative responses to the praise following failure and the development of theory of mind. Although the study demonstrated strengths and important implications, it was subject to some limitations. For example, children's responses to the praise following failure were examined within the context of the scenarios, and the ways that children respond to this type of praise in their everyday lives remains unclear. Another limitation involves the nature of the sample, which included a relatively small sample size and an unbalanced sex distribution. It should also be noted that Japanese children could be motivated by failure, as they have been shown to work harder following failure than they do following success ([@B14]). Additional studies specifically designed to examine the effects of praise in diverse settings and cultures are required to provide evidence for or against the generalization of the current findings. Author Contributions ==================== The author confirms being the sole contributor of this work and approved it for publication. Conflict of Interest Statement ============================== The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. **Funding.** The research was supported by KAKENHI (Grant No. 15K17283). The author thanks the children who participated in this study. Data from seven children whose interpretation was classified as "other" were eliminated from analyses. Instruction =========== You are playing a block design game with *Hanako-sensei* (teacher Hanako). *Hanako-sensei* shows you four blocks; one side of these blocks has two colors---white and red---and the other side is colored red. No-Feedback Story ================= *Hanako-sensei* shows you a picture of a model and says, "Look at this picture and recreate the design yourself using these blocks. Start!" You try hard to recreate the design but fail to do so. Praise Story ============ *Hanako-sensei* shows you a picture of another model and says, "Look at this picture and recreate the design yourself using these blocks. Start!" You try hard to recreate the design but fail to do so. *Hanako-sensei* comes to your side and says, "\[Child's name\], *sugoi-ne* (great)!" First-Order False-Belief Task ============================= I'm going to tell you a little story about *Saru-san* (Monty) and his lunch box. Look! Here's *Saru-san*. He wants *Kaeru-san* (Freddie) to put an apple in his lunch box to take to school, but *Kaeru-san* says there are no apples left, so he'll have to take a pear instead. *Saru-san* doesn't like pears at all. He really wanted an apple! He's so cross about the pear that he stamps his feet all the way up the stairs. 1. Q1. How does Saru-san feel when he gets a pear? (Memory control 1) 2. Q2. How does Saru-san feel when he gets an apple? (Memory control 2) But, look! While *Saru-san* is out of the kitchen, *Kaeru-san* finds one apple left in the cupboard. He decides to give *Saru-san* a nice surprise; so, he takes the pear out of *Saru-san*'s lunchbox and puts the apple in there instead. He puts the lunchbox in *Saru-san*'s bag. *Saru-san* comes back, picks up the bag, and hurries off to school. *Saru-san* doesn't see what's inside his lunchbox. Now it's lunchtime. *Saru-san* takes out his lunchbox. 1. Q3. What does Saru-san think is in the box, an apple or a pear? (First-order false-belief) 2. Q4. What is really in the box, an apple or a pear? (Reality) Second-Order False-Belief Task ============================== This is Kouta. Today is his birthday, and Kouta's mom is going to surprise him by giving him a puppy. She has hidden the puppy in the shed until it's time for Kouta's birthday party. Kouta says "I really hope you've got me a puppy for my birthday, Mom"; but, remember, Mom wants to surprise Kouta; so, instead of telling Kouta she got him a puppy, Mom says, "Sorry, I didn't get you a puppy, Kouta. Actually, I've got you a really good toy for your birthday." 1. Q1. What did Kouta think he was getting for his birthday? (First-order false-belief) 2. Q2. What was his mom really giving him? (Reality) \[If the child answers the two questions correctly, continue with the following story\]. Now Kouta decides to go outside to play. On his way out, he goes into the shed to get his bike, and he finds the birthday puppy! Kouta says to himself, "Wow! Mom didn't get me a toy; she really got me a puppy for my birthday!" Mom didn't see Kouta go to the shed; so, she doesn't know he found the birthday puppy. Inside, the telephone rings. It's Kouta's granny, calling to find out what time the party is. Granny says to Mom, "What does Kouta think you've gotten him for his birthday?" 1. Q3. What does Mom say to Granny? (Second-order false-belief) 2. Q4. Did Mom see Kouta go into the shed? (Memory control 1) 3. Q5. What did Mom really get Kouta for his birthday? (Memory control 2) [^1]: Edited by: Ann Dowker, University of Oxford, United Kingdom [^2]: Reviewed by: Virginia Slaughter, The University of Queensland, Australia; Elisabetta Conte, Università degli Studi di Milano Bicocca, Italy [^3]: This article was submitted to Developmental Psychology, a section of the journal Frontiers in Psychology

1. Introduction {#sec1-materials-12-03635} =============== Laser cutting is a process in which a (continuous or pulsed) laser beam melts, simultaneously melts and evaporates, or melts and/or burns away the material in the cut (the kerf). It is a thermal cutting process that uses a large amount of energy that is focused on a small area \[[@B1-materials-12-03635]\]. Along with the laser beam, a stream of gas is emitted coaxially; the gas might be reactive to the material being cut (oxygen, air) or it might be an inert gas (nitrogen, argon). The presence of a reactive gas in the kerf produces an exothermic reaction, which effectively increases the power of the laser, allowing it to cut material of considerable thickness \[[@B2-materials-12-03635],[@B3-materials-12-03635]\]. An inert gas is used to remove molten material and it also acts as a cooling and protective agent for the optical system of the laser cutter \[[@B3-materials-12-03635],[@B4-materials-12-03635]\]. Laser technologies can be used to section most engineering materials, e.g. unalloyed steel \[[@B1-materials-12-03635],[@B5-materials-12-03635]\], stainless steel \[[@B4-materials-12-03635],[@B6-materials-12-03635]\], aluminum alloys \[[@B7-materials-12-03635]\], nickel alloys \[[@B8-materials-12-03635]\], carbon fiber reinforced polymer composites \[[@B9-materials-12-03635]\], and also natural materials. The laser cutting of metallic materials is widely used in the automotive, chemical, marine, and aviation industries \[[@B10-materials-12-03635]\]. The strengths of this material separation technology include a high cutting speed (many times faster than the speeds used during machining), satisfactory dimensional accuracy, and low roughness of the cut surface, as well as a high level of automation and flexibility \[[@B1-materials-12-03635],[@B11-materials-12-03635],[@B12-materials-12-03635]\]. During laser cutting, the coherent laser beam might generate imperfections on the cut faces and edges. Imperfections, according to the EN ISO 17658: 2015 standard, are irregularities or deviations from the specified shape or location of cut. The laser cutting process directly causes them (no adverse phenomena resulting from external stresses or deformations are taken into account). The quality of the surface and the course of the cutting process depend on many factors with different levels of control. The most important of them include laser power, laser type and operation mode, position of laser beam focus relative to the surface of the workpiece, cutting speed, and assist gas type and pressure \[[@B10-materials-12-03635]\]. The quality of laser-cut edges and surfaces is evaluated quantitatively in accordance with EN ISO 9013:2017. In industrial practice, surface roughness, width of the heat affected zone, material phase changes, residual stresses in the vicinity of the kerf, and the formation of dross on the lower edges of the kerf define the quality of laser-cut surfaces \[[@B1-materials-12-03635],[@B9-materials-12-03635],[@B11-materials-12-03635],[@B13-materials-12-03635]\]. Surface roughness in laser cutting is the result of the overlapping effects of the thermal and hydrodynamic processes that are characteristic of laser processing and vibrations of the workpiece caused by the high-pressure gas jet \[[@B3-materials-12-03635]\]. The surface of the workpiece can be divided into two areas of different shape and quality. These zones are separated from each other by the so-called boundary layer separation (BLS). One zone is the laser beam entrance area and the other is the laser beam exit area; the latter has a much higher surface roughness \[[@B11-materials-12-03635],[@B14-materials-12-03635]\]. A characteristic feature of the laser cutting process is the formation of striations on the cut surface. The striations pattern forms as a consequence of hydrodynamic flow of molten material, laser power fluctuation, gas flow fluctuation, and laser head oscillation \[[@B15-materials-12-03635],[@B16-materials-12-03635]\]. During laser cutting, the heat that is delivered to the material and the high temperature gradient create a thin film of material with variable properties, called the heat affected zone, which is characterized by a small thickness, but locally, especially at the edges and corners, has high hardness. The differences in hardness may be associated with a non-uniform distribution of energy at beam cross-section and differences in heat dissipation rates \[[@B17-materials-12-03635],[@B18-materials-12-03635]\]. The high temperature in the cutting zone gives rise to thermal stresses. Heating of the material followed by rapid cooling with the assist gas leads to material shrinkage and phase changes. The dynamics of phase changes at different cooling rates lead to the formation of tensile residual stresses. The value and distribution of thermal stresses depend on the thermal conductivity of the material being cut \[[@B11-materials-12-03635],[@B13-materials-12-03635]\]. Non-alloy steel characteristically forms a thin surface layer of iron oxide phases FeO and Fe~3~O~4~, which change the color of the surface to bluish \[[@B19-materials-12-03635],[@B20-materials-12-03635],[@B21-materials-12-03635]\]. The resultant particles are about 50 µm in size and they most often have the form of iron spheres that are coated with iron oxide \[[@B3-materials-12-03635]\]. The presence of various imperfections on the surfaces and edges of laser-cut parts and the presence of drops of solidified material at their bottom edges calls for finishing treatment. Scale and outflow can be removed with hand files, wire brushes, deburring machines, and belt sanders. Mechanical grinding \[[@B22-materials-12-03635]\] and smoothing with diamond tools \[[@B23-materials-12-03635]\] can improve the geometrical structure of a laser-cut surface. Mechanical grinding allows for reducing surface roughness and the thickness of the hardened layer and introduces into the workpiece a state of tensile residual stress, which might lead to cracking of the surface. The resultant defects are loci of stress concentration and reduce the bearing capacity of the processed surface \[[@B22-materials-12-03635]\]. Smoothing with diamond tools allows for reducing surface roughness by several times. At the same time, it causes structural and phase changes in the material. As a result of smoothing, residual austenite is transformed into martensite. The surface layer is mechanically hardened. The use of a diamond tool is associated with the occurrence of sliding friction between the tool and the surface of the workpiece, which leads to an increase in temperature in the machining zone. Hardness in the edge zone is higher than that after laser cutting; discontinuities and microcracks also appear \[[@B23-materials-12-03635]\]. The condition of the surface and edges of a laser-cut semi-finished product does not allow or makes its use difficult in subsequent production stages. It is also difficult to apply paint coatings to it because of flaking. Centrifugal shot peening changes the properties of the surface layer of workpieces pre-machined by turning, milling, or grinding. Centrifugal shot peening of a pre-milled surface leads to the formation of a dimpled surface with elongated depressions that can retain additional lubricant \[[@B24-materials-12-03635]\]. The use of optimal centrifugal shot peening conditions during finish machining of turned C45 and 40H steel shafts, allows for obtaining a several-fold reduction in Ra. The positive effects of using shot peening for post-processing of machined parts encourage the use of centrifugal shot peening to modify the surface layer of laser-cut parts. Shot peening as a process used to finish machine parts allows for obtaining a surface that is characterized by a low roughness \[[@B25-materials-12-03635]\] and a large bearing area (material ratio). This is due to the more "streamlined" shape of the micro asperities that form during peening \[[@B25-materials-12-03635],[@B26-materials-12-03635]\]. Indentations imparted to the shot-peened surface serve as "reservoirs" for lubricants, allowing the formation of a protective layer that increases the resistance of these surfaces to abrasive wear \[[@B27-materials-12-03635]\]. The surface texture formed during processing affects the energy state and the adhesive properties of the surface \[[@B28-materials-12-03635],[@B29-materials-12-03635]\]. One positive effect of shot peening, apart from the fact that it improves the properties of the surface layer, is the rounding of the workpiece edges \[[@B30-materials-12-03635]\]. Centrifugal shot peening can also be combined in a single operation with grinding while using hybrid tools that enable the processing of difficult-to-cut materials, such as nickel and titanium alloys \[[@B31-materials-12-03635]\]. During shot peening, the concentration of crystal structure defects changes, as confirmed in tests that were carried out by annihilation techniques \[[@B32-materials-12-03635]\], as a result of which compressive stresses form in the surface layer of the workpiece \[[@B33-materials-12-03635]\]. Changes in the state of residual stress and the increase in the microhardness of the surface layer enhance the resistance of the material to abrasive and fatigue wear \[[@B34-materials-12-03635],[@B35-materials-12-03635]\]. A review of the literature shows that the laser cut parts require finishing steps. The beneficial effect of shot peening on the condition of the surface of machined and ground parts is well-known. The aim of the present study was to assess the impact of centrifugal shot peening on the surface layer of C45 steel parts that were cut with laser. 2. Materials and Methods {#sec2-materials-12-03635} ======================== The tests were performed while using specimens of non-alloy C45 steel (in accordance with EN ISO 683-2:2018). This steel grade is used in the machine industry for the manufacture of medium-load machine and equipment parts, such as spindles, shafts, axles, and unhardened gears. [Table 1](#materials-12-03635-t001){ref-type="table"} shows the chemical composition and strength properties of the steel tested (as per the technical specification sheet). Rectangular specimens, 5 mm × 8 mm × 100 mm in size, were cut in a LASER Amada 3000 W laser cutter from Amada America Inc. (Düsseldorf, Germany), while using oxygen as the assist gas and standard parameters: cutting speed v = 1150 mm/min., power 2.15 kW, frequency f~Hz~ = 1280 Hz, assist gas pressure 0.06 MPa, and laser beam focus position +13 mm. Laser-cut specimens were finish-machined using centrifugal shot peening, which consisted in impacting the surface of a workpiece with shot propelled from openings radially arranged in a rotating peening head. After the head is set into rotational motion, the shot particles hit the surface of the workpiece under the influence of centrifugal force. Upon brief contact with the shot, the surface of the workpiece undergoes plastic deformation. Peening reduces the surface roughness and hardens the surface layer of the processed material. After centrifugal shot peening, the resulting machining marks on the surface are organised in the longitudinal and transverse direction, which allows for obtaining a surface with less roughness. It can be used to process axially symmetrical parts, flat surfaces, and surfaces with complex shapes. Moreover centrifugal shot peening allows for the treatment of elements with low rigidity. Centrifugal shot peening tests were performed on an FV-580a (MOC Mechanicy Pruszków, Pruszków, Poland) vertical machining center while using a special peening head with an outer diameter of 70 mm with a blasting unit containing twelve (z~k~ =12) symmetrically arranged shot balls with a diameter d~k~ = 6.3 mm ([Figure 1](#materials-12-03635-f001){ref-type="fig"}). During shot peening, specimen (1) was clamped in a holder (2) so that the peened surface was parallel to the spindle axis. The holder was secured in the jaws of a vice (3) that was mounted on the table of the machine tool, which moved at speed v~f~ during peening. The body of special head has radial holes, in which the balls move. In the guiding part the balls can move freely along the parting axis. On the body of the head a ring with radially arranged holes in which the balls are located is mounted. On the circumference of the ring the diameter of the holes is smaller than the diameter of the balls, which prevents the balls from falling out. The peening head (4) rotated with speed *n* and simultaneously moved in the cross-feed direction f~p~. When the head was set in rotational motion, shot particles were blasted from it under the influence of centrifugal force and they impacted the surface of the workpiece. Upon impact, the particles bounced off the peened surface and were retracted into the seats in the head at distance *g*, referred to as "head infeed to the workpiece". Centrifugal shot peening was performed while using Mobile Cut cooling lubricant (Mobile, Germany) at a constant infeed g = 0.5 mm and a constant cross-feed f~p~ = 0.08 mm; the number of passes was *i* = 1. The parameters that were varied in the experiments included \- tangential speed of shot peening head: v~g~ = 528--1143 m/min. \- feed rate: v~f~ = 1368--10488 mm/min. The measurements of surface roughness and three-dimensional (3D) topography were performed while using a Hommel-Etamic T8000RC 120--140 device (Jenoptik, Villingen-Schwenningen, Germany). According to EN ISO 9013:2017, measurements of surface roughness should be made at a distance equal to 1/3 of the thickness of the workpiece, from the upper cut edge, when the thickness of the workpiece is above 2 mm. Following the guidelines of the EN ISO 9013:2017 standard, roughness was measured at a distance of 1/3 of workpiece thickness from the upper cutting edge, in the so-called "laser entrance zone" and at a distance of 1/3 of workpiece thickness from the bottom cut edge, in the so-called "laser exit zone". We decided to determine the roughness parameters in this lower area because the laser-cut parts had two characteristic zones with different roughness. [Figure 2](#materials-12-03635-f002){ref-type="fig"} schematically shows the place of roughness measurements. The same measurement scheme was used to measure the roughness of shot-peened specimens. Following the guidelines for roughness measurements outlined in EN ISO 4288:1997, the sampling length was set at lr = 2.5 mm. The material ratio Rmr was c = 2 µm below the highest peak of the roughness profile. The structure of the material was assessed while using a Nikon Eclipse MA 100 metallographic microscope (Nikon, London, UK). Microhardness measurements were made by the Vickers method in accordance with EN ISO 6507-1:2018 using transverse metallographic specimen surfaces that were prepared in a standard way. A LM Leco 700 at microhardness tester (Leco, San Jose, MI, USA) was used at an indenter load of 50 g (0.05 HV). Microhardness measurements were carried out according to the diagram shown in [Figure 3](#materials-12-03635-f003){ref-type="fig"}. A constant step from the surface after laser cutting or laser cutting and centrifugal shot peening was determined during the measurement. Measurements at a given depth were repeated 10 times. The maximum microhardness, which is just below the machined surface, was determined based on the micro hardness distribution. The results of the microhardness measurements served as a basis for calculating the degree of strain hardening of the surface layer e from the Equation (1):$$e = 100 \times \frac{{HV}_{\max} - {HV}_{0}}{{HV}_{0}},~\left( \% \right)$$ where: ${HV}_{\max}$---maximum microhardness of the surface layer after centrifugal shot peening,${HV}_{0}$---microhardness before shot peening. X-ray diffraction tests were performed with a Empyrean diffractometer (Panalytical, Malvern, UK). The experiments were carried out in the symmetrical Bragg-Brentano geometry, in which a divergent beam is used at room temperature and atmospheric pressure. Rectangular 5 × 8 × 15 mm^3^ specimens, which were cleaned in acetone prior to the experiments, were tested. A 3 kW copper anode lamp was used as the x-ray source. Anode radiation was filtered through nickel, which absorbs K~β~ radiation. Data were recorded in the angle range 2θ = 5--110° while using the step scan method. The scan step was 0.05° and the point exposure time was 3--5 s. Residual stresses in the surface layer of laser-cut-and laser-cut and shot-peened specimens were tested while using the mechanical method. They were calculated on the basis of measurements of specimen deformations that occurred when the sequential layers of the material in which these stresses resided were removed. 3. Results {#sec3-materials-12-03635} ========== 3.1. Surface Roughness {#sec3dot1-materials-12-03635} ---------------------- [Figure 4](#materials-12-03635-f004){ref-type="fig"}a shows the surface topography of a laser-cut specimen. The surface formed by cutting is characterized by large differences in height (there are numerous peaks and valleys). In the area where the laser beam entered the material (A), the surface is covered with fine, evenly spaced straight drag lines. The surface roughness parameters for this area are as follows: Ra = 2.49 ± 0.12 µm and Rz = 15.76 ± 1.33 µm. In the laser beam exit zone (B), the drag lines are curvilinear; they deviate from standard slope (the intended path of the beam). This indicates that the material "slips away" from the kerf. The roughness parameters in the exit zone are Ra = 5.04 ± 0.35 µm and Rz = 31.29 ± 2.27 µm. The values of the surface roughness parameters in the exit zone are about twice as high as those in the entrance zone. The same relationship was observed for the parameters of the Abbott--Firestone curve, which were Rpk = 2.43 ± 0.3µm, Rk = 8.14 ± 0.38 µm, and Rvk = 3.72 ± 0.93 µm in the entrance zone and Rpk = 4.98 ± 0.58 µm, Rk = 16.54 ± 1.17 µm, and Rvk = 7.31 ± 0.56 µm in the exit zone. An increase in the viscosity of the workpiece at the bottom cut edge causes the large differences in the quality of the surface. This enhances the thickness of the liquid metal layer, which, as a result, sticks to the surface being cut. This promotes the formation of large irregularities. During centrifugal shot peening, the shot balls impact the surface of the laser-cut workpiece, leading to the deformation of its geometric structure ([Figure 4](#materials-12-03635-f004){ref-type="fig"}b). As a consequence, the height of the drag lines is reduced and micro asperity summits are flattened. The difference between the height of the surface roughness profiles of the beam entrance zone (A) and the beam exit zone (B) is also reduced. [Figure 5](#materials-12-03635-f005){ref-type="fig"} shows the effect of tangential speed of the tool v~g~ and feed rate v~f~ on the roughness parameter Ra. As the rotational speed n, and thus tangential speed v~g~ increase, surface roughness decreases. This is due to the more intense smoothing of micro asperities on laser-cut surfaces as a result of increasing impact energy. The maximum decrease in Ra relative to the value recorded after laser cutting was four-fold. The changes in surface roughness that were obtained in the present study are more prominent when compared to those reported for vibratory-and-rotational shot peening of C45 steel specimens, for which a 30% reduction in surface roughness parameters \[[@B35-materials-12-03635]\] was observed, and for centrifugal shot peening of 41Cr4 steel specimens, where a three-fold reduction in Ra was obtained \[[@B26-materials-12-03635]\]. On the other hand, an increase in feed rate results in a decrease in the number of impacts per unit of exposure area, which leads to an increase in surface roughness. The nature of the changes in the surface roughness parameter Rz as a function of the analyzed technological parameters of centrifugal shot peening is similar to that observed for Ra. An analysis of the changes in the parameter Rz of the roughness profile as a function of feed rate showed that the maximum reduction in Rz after centrifugal shot peening was about three-fold for the beam entrance zone and about four-fold for the exit zone ([Figure 6](#materials-12-03635-f006){ref-type="fig"}). Cooperation between two surfaces is described by the parameters that are associated with the Abbott--Firestone curve (bearing curve). The bearing curve for laser-cut parts is a degressive-progressive curve ([Figure 7](#materials-12-03635-f007){ref-type="fig"}a). It has a large angle of inclination. The surface described by this curve has "sharp" micro-asperity summits and a small bearing area. The Abbott--Firestone curve for parts processed by centrifugal shot peening ([Figure 7](#materials-12-03635-f007){ref-type="fig"}b) is only slightly inclined, which might point to a high abrasion resistance of the tested surfaces. It is a curve with a "thick center" and it can therefore be classified as a progressive curve. The effect of tangential speed v~g~ on the parameters of the bearing curve was similar to that observed for Ra and Rz ([Figure 8](#materials-12-03635-f008){ref-type="fig"}). An increase in speed v~g~ caused more intense smoothing of micro asperities on the laser cut surface, which translated into a decrease in the values of Rpk, Rk, and Rvk parameters. Core roughness depth Rk, depending on the tangential speed of the peening head, decreased by two to seven times after shot peening. This means that centrifugal shot peening improves the so-called bearing capacity of the surface, or, to put it differently, that, after a grinding-in period, a substantial part of the surface will be in contact with the surface of the mating element. The Rpk parameter, which characterizes the behavior of a surface during grinding-in, decreased by 3 to 8.1 times relative to the value after laser cutting. The depth of the valleys (parameter Rvk) was reduced by centrifugal shot peening up to two times, depending on the tangential speed. This means that surfaces processed in this way may have lower oil retention capacity. The increase in the feed rate v~f~ ([Figure 9](#materials-12-03635-f009){ref-type="fig"}) translated into a smaller proportion of elastic-plastic strains, which caused an increase in the parameters of the bearing curve. Feed rate v~f~ had less effect on the Abbott--Firestone curve parameters than tangential speed v~g~. At feed rate v~f~ = 1368 mm/min "(least)", Rpk dropped up to six-fold when compared to the value that was obtained for laser-cut specimens. On the other hand, when a tangential speed v~g~ = 1143 m/min "(largest)" was used, an over eight-fold decrease in Rpk was observed. In the case of the roughness parameter Rk, the influence of tangential speed v~g~ was also more pronounced than the effect of feed rate v~f~. An up to five-fold reduction in Rk was obtained when compared to the value that was recorded for the laser cut specimens. After shot peening, parameter Rvk was reduced more than two-fold, a result that is similar to that obtained for tangential speed v~g~. The material ratio was increased for all of the investigated peening parameters when compared to the laser-cut surface ([Figure 10](#materials-12-03635-f010){ref-type="fig"}). The material ratio of the roughness profile of the laser-cut parts was Rmr~(2)~ = 6.08% ± 0.57% for the beam entrance zone and Rmr~(2)~ = 5.42% ± 0.42% for the beam exit zone. This means that the laser-cut surface had low abrasion resistance. Centrifugal shot peening increased the material ratio by 3.33 to over 14 times for the beam entrance zone and up to almost 12 times for the beam exit zone. This suggests that the process also increased the abrasion resistance of the surface. During centrifugal shot peening, there is a wide angle between the normal velocity vector and the peened surface. This promotes strong friction between the peened surface and the shot, which causes intense abrasion of the micro-asperity summits that previously formed on the hard material, decreasing the height of micro asperities and leading to the formation of micro-cavities on the peened surface. The favorable changes in the stereometric properties of the peened surface are likely to make the surface resistant to abrasive wear \[[@B35-materials-12-03635]\]. The "smoothing" of the surface as a result of deformation leads to an evening out of zones with varying surface roughness, which is likely to have a positive effect on the use of a given part in a next production stage (e.g. application of paint coatings). 3.2. Microstructure and Microhardness {#sec3dot2-materials-12-03635} ------------------------------------- Before the laser cutting process, the C45 steel specimens had a ferritic--pearlitic structure. As a result of strong heating with the laser beam followed by cooling down, the material around the cut edge became hardened. A heat affected zone was created ([Figure 11](#materials-12-03635-f011){ref-type="fig"}a), which differed in the structure and properties from the parent material. In the heat affected zone, martensite needles ([Figure 11](#materials-12-03635-f011){ref-type="fig"}b) are visible next to the edge of the cut, with ferrite, and then martensite located further away from the edge. The laser cutting process lasted too short for ferrite grains to become enriched with carbon, which resulted in the formation of low-carbon martensite. Perlite, on the other hand, transformed into high carbon martensite. The cyclic bombarding of the material with shot particles modifies its structure ([Figure 11](#materials-12-03635-f011){ref-type="fig"}c). The deformation of the material caused by centrifugal shot peening leads to an increase in dislocation density. The gliding dislocations are stopped when they encounter obstacles, such as other dislocations, grain boundaries, or cementite separation. The increase in the dislocation density has a hardening effect on the material. The resulting structure can be classified as fine-needle martensite. In laser-cut specimens, the zone of changes in microhardness is over 0.3 mm wide ([Figure 12](#materials-12-03635-f012){ref-type="fig"}). The microhardness of the surface layer increased more than twice in the area of the cut edge (approx. 560 HV0.05), and then slowly decreased to reach the core microhardness value (approx. 200 HV0.05). The literature only offers reports on the use of centrifugal shot peening for processing parts with hardnesses of 240 HV and 360 HV \[[@B26-materials-12-03635]\]. However, the present study shows that the laser-cut parts can also be treated by centrifugal shot peening. Though the hardness and thickness of the hardened layer only increase slightly during centrifugal shot peening ([Figure 12](#materials-12-03635-f012){ref-type="fig"}), the increase is likely to translate into increased resistance to abrasive wear \[[@B35-materials-12-03635]\]. The degree of strain hardening of the surface layer for laser-cut-and shot-peened C45 steel depend on the technological parameters of shot peening ([Figure 13](#materials-12-03635-f013){ref-type="fig"}). An increase in the tangential speed v~g~ leads to an increase in impact energy and it elevates the number of impacts per unit of area, which translates into a greater proportion of plastic-elastic deformations in the workpiece and thus increases the degree of strain hardening of the surface layer. The thickness of the hardened layer g~h~, in relation to the core, was max. 60 µm, as verified by a statistical test of means. The largest changes in microhardness occurred right next to the treated surface, where the most deformed grains were found ([Figure 11](#materials-12-03635-f011){ref-type="fig"}c). The degree of strain hardening increased as a function of tangential speed v~g~ from about 6% to 15% ([Figure 13](#materials-12-03635-f013){ref-type="fig"}a), with the dynamics of increase the degree of strain hardening declining in the velocity range v~g~ = 989--1143 m/min. A rise in feed rate increases the spacing between the indentations that are imparted by shot, which results in uneven deformation of the treated surface. The changes in the thickness of the hardened layer after centrifugal shot peening (for a variable feed rate v~f~) ranged from 15 µm to 56 µm. At low feed rates, the number of impacts per unit area was higher, which increased the thickness of the hardened layer. The increase in microhardness for v~f~ = 1368 mm/min relative to the value that was recorded after laser cutting was 16.4% ([Figure 13](#materials-12-03635-f013){ref-type="fig"}b). 3.3. XRD Tests {#sec3dot3-materials-12-03635} -------------- The presence of oxygen as the assist gas during cutting promotes an exothermic reaction, which leads to the formation of combustion products on the cut surface. This is confirmed by the qualitative change in the phase composition of the surface layer after laser cutting ([Figure 14](#materials-12-03635-f014){ref-type="fig"}). The XRD tests that we carried out allowed for us to identify the following phases on the surface of C45 steel: Fe~3~O~4~, Fe~2~O~3~, Fe~0.9~O ([Figure 14](#materials-12-03635-f014){ref-type="fig"}). [Table 2](#materials-12-03635-t002){ref-type="table"} shows the lattice parameters and angles for the compounds formed on the surface of metal as a result of laser cutting. XRD tests of a laser-cut-and shot-peened specimen demonstrated the presence of two compounds on the treated surface. The diffractogram ([Figure 15](#materials-12-03635-f015){ref-type="fig"}) shows two main dominant peaks, which correspond to ferrite and martensite in the diffraction database ([Table 3](#materials-12-03635-t003){ref-type="table"}). Shot balls, when impacting the surface of the workpiece, "blast off" and "shear off" oxide phases, removing this way the thin oxide film that had formed on the surface of the workpiece. The removal of combustion products from the treated surface reduces or eliminates the peeling of coatings applied to it. 3.4. Residual Stress {#sec3dot4-materials-12-03635} -------------------- Laser cutting is accompanied by the re-solidification of molten material, which takes place in unstable conditions. This promotes the occurrence of tensile residual stresses that reside in the specimens at a depth of about 0.05 mm from the surface of the workpiece ([Figure 16](#materials-12-03635-f016){ref-type="fig"}). Tensile stresses also originate from the structural changes that take place during the cutting process. In the process of centrifugal shot peening, compressive residual stresses are produced in the surface layer ([Figure 16](#materials-12-03635-f016){ref-type="fig"}), which improve the functional properties of the material, e.g. increase its resistance to fatigue wear \[[@B34-materials-12-03635]\]. Grinding, which has been used so far to finish laser cut parts, leads to the formation of tensile residual stresses in the surface layer \[[@B22-materials-12-03635]\]. An increase in the tangential speed of the tool increased the depth of compressive residual stresses ([Figure 16](#materials-12-03635-f016){ref-type="fig"}a), and also, though to a lesser extent, contributed to an increase in the value of these stresses. The absolute maximum value of compressive residual stress was found at a depth of 0.05--0.1 mm. On the other hand, as the feed rate increased, the absolute maximum value of compressive stresses produced in the peened part decreased. The depth of compressive residual stresses did not change significantly ([Figure 16](#materials-12-03635-f016){ref-type="fig"}b). The lowest residual stress was obtained at v~f~ = 10 488 mm/min and v~g~ = 835 m/min within the range of the technological parameters of the experiment. 4. Conclusions {#sec4-materials-12-03635} ============== The following conclusions can be drawn based on the centrifugal shot peening tests of the laser-cut C45 steel elements. 1. The use of centrifugal shot peening for finish machining of laser-cut C45 steel parts allowed for obtaining a four-fold reduction in the surface roughness parameters Ra and Rz. Centrifugal shot peening diminished the differences in Ra and Rz between the beam entrance zone and the beam exit zone. 2. An increase in the tangential speed of the tool v~g~ resulted in a decrease in roughness parameters, while an increase in feed rate v~f~ had the opposite effect, with the changes being more pronounced for the variable v~g~. 3. Centrifugal shot peening allowed for changing the nature of the bearing curve from a degressive--progressive curve to a degressive curve. A significant decline in Rpk, Rk, and Rvk parameters was observed as compared to the values that were obtained after laser cutting. 4. Centrifugal shot peening resulted in an up to 14-fold increase in the material ratio of the roughness profile as compared to the value obtained after laser cutting. 5. Peened workpieces had an up to 16% higher microhardness (v~g~ = 835 m/min, v~f~ = 1365 mm/min) and a 15 μm to 58 μm deep hardened layer. 6. As the surface of a workpiece was impacted by shot during centrifugal shot peening, oxide phases, which are combustion products, were blasted off and sheared off the surface, which caused the removal of Fe~3~O~4~, Fe~2~O~3~, and Fe~0.9~ oxides. 7. In the surface layer of the specimens, compressive residual stresses were formed during centrifugal shot peening, whose absolute maximum value varied from 450 MPa to 740 MPa. The stresses resided at a depth of 0.25--0.40 mm, depending on the technological parameters of the peening process. 8. As the tangential speed of the peening head v~g~ grew to 835 m/min, an increase was observed in the absolute value of compressive residual stresses and the depth of their accumulation; a further increase in tangential speed only increased the depth of the occurrence of compressive residual stresses. An increase in feed rate v~f~ caused a decrease in the absolute value of compressive residual stresses, but it did not significantly affect the depth of accumulation of residual stresses. 9. The results of the centrifugal shot peening experiments demonstrate that this cold working method can be successfully used with CNC machine tools. The CNC machine tools are equipped with control systems that can be used to guide the tool along a designated path, which allows for processing parts with complex shapes. Conceptualization, A.S. and K.Z.; Methodology, A.S.; formal analysis, K.Z.; investigation, A.S.; data curation, A.S.; writing---original draft preparation, A.S.; writing---review and editing, A.S. and K.Z.; visualization, A.S.; supervision, K.Z. The project/research was financed in the framework of the project Lublin University of Technology-Regional Excellence Initiative, funded by the Polish Ministry of Science and Higher Education (contract no. 030/RID/2018/19). The authors declare no conflict of interest. {#materials-12-03635-f001} {#materials-12-03635-f002} {#materials-12-03635-f003} {#materials-12-03635-f004} {#materials-12-03635-f005} {#materials-12-03635-f006} {#materials-12-03635-f007} {#materials-12-03635-f008} {#materials-12-03635-f009} {#materials-12-03635-f010} {#materials-12-03635-f011} {#materials-12-03635-f012} {#materials-12-03635-f013} {#materials-12-03635-f014} {#materials-12-03635-f015} {#materials-12-03635-f016} materials-12-03635-t001_Table 1 ###### Chemical composition and selected properties of C45 steel. Chemical Composition, (%) --------------------------- ---------------- ------ ------- ------ ------ ------ ------- ------ 0.48 0.74 0.36 0.011 0.01 0.09 0.02 0.002 rest Yield point R~e~ = 430 MPa Tensile strength R~m~ = 740 MPa Hardness 250 HB materials-12-03635-t002_Table 2 ###### Data of identified compounds based on the diffraction pattern obtained after analysis surface of C45 after laser cutting. ------------------------------------------------------------------------------------ Number in the Base\ Chemical Formula Lattice Parameters (nm) Angles (°) ICDD PDF--4+ --------------------- ------------------ ------------------------- ----------------- 04-009-8436 Fe~3~O~4~ a = b = c = 0.8403 α = β = γ = 90° 04-006-9058 Fe~2~O~3~ a = b = 0.5350\ α = β = 90°\ c = 1.3720 γ = 120° 04-001-9267 Fe~0.9~O a = b = c = 0.4298 α = β = γ = 90° ------------------------------------------------------------------------------------ materials-12-03635-t003_Table 3 ###### Data identified phases based on the diffraction patter obtained after analysis surface of C45 after laser cutting and centrifugal shot peening. ------------------------------------------------------------------------------------ Number in the Base\ Chemical Formula Lattice Parameters (nm) Angles (°) ICDD PDF--4+ --------------------- ------------------ ------------------------- ----------------- 04-002-1061 Fe--α a = b = c = 0.2862 α = β = γ = 90° 00-044-1289 C~0.14~Fe~1.86~ a = b = 0.2846\ α = β = γ = 90° c = 0.3053 ------------------------------------------------------------------------------------

All relevant data are within the paper and its Supporting Information files. Introduction {#sec004} ============ A population of pluripotent adult somatic stem cells, well known as "neoblasts", has been extensively characterized and documented for free-living platyhelminthes (flatworms) \[[@pntd.0005418.ref001]--[@pntd.0005418.ref005]\]. Neoblasts represent the only proliferative cell population, responsible for cell renewal during homeostasis, development and regeneration \[[@pntd.0005418.ref006]--[@pntd.0005418.ref011]\]. Like in its free-living relatives, neoblast-like stem cells and similar cell renewal mechanisms also exist in the other two main groups of flatworms, trematodes (flukes) and cestodes (tapeworms), both living a parasitic life \[[@pntd.0005418.ref003], [@pntd.0005418.ref012]--[@pntd.0005418.ref015]\]. In cestodes, a population of undifferentiated cells, called "germinative cells", is the only source for cell proliferation. These germinative cells are totipotent and are thought to drive growth and development throughout the life cycle of cestodes \[[@pntd.0005418.ref012], [@pntd.0005418.ref015]--[@pntd.0005418.ref016]\]. The larval stage of the cestode *Echinococcus multilocularis* is the causative agent of alveolar echinococcosis (AE), one of the most lethal human helminthiasis \[[@pntd.0005418.ref017]\]. An infection is initiated when the intermediate host (rodents, humans) ingests infective eggs produced by adult tapeworms. The oncosphere hatches from the egg and then develops in the liver into cyst-like metacestode vesicles, which grow like tumors and infiltrate host tissue, forming new vesicles and even metastasizes. The metacestode vesicles bud giving rise to brood capsules, which in turn generate protoscoleces by asexual multiplication. Protoscoleces can either mature into adult tapeworms if ingested by the definitive host (canids) or develop into metacestode vesicles when distributed in the intermediate host. This unique proliferation potential of *E*. *multilocularis* metacestode larvae is believed to be based upon the germinative cells, which are totipotent and have the ability for extensive self-renewal \[[@pntd.0005418.ref015]\]. It has been well documented that the maintenance of pluripotency and self-renewal capacity of stem cells requires a continuous input from cell-extrinsic signals \[[@pntd.0005418.ref018]--[@pntd.0005418.ref019]\]. The extrinsic factors initiate various intrinsic signaling cascades which in turn maintain stem cells and regulate their functions. Signaling axes including LIF/gp130/STAT3, BMPs/BMPRs/Smads, Wnt/Frizzled/β-catenin, PI3K/AKT, and FGF/FGFR have been extensively evidenced to participate in controlling the survival, self-renewal, and differentiation of stem cells \[[@pntd.0005418.ref019]--[@pntd.0005418.ref020]\]. Increasing evidence has shown that the epidermal growth factor receptor (EGFR)-dependent signaling pathways also play important roles in the maintenance and regulation of stem cells \[[@pntd.0005418.ref021]--[@pntd.0005418.ref025]\]. The metacestode larvae of *E*. *multilocularis* grow and proliferate in close contact with the intermediate host's tissues, mainly within the liver. The microenvironment for metacestode development involves a number of host-derived hormones and cytokines, such as insulin, bone morphogenetic protein (BMP), fibroblast growth factor (FGF) and epidermal growth factor (EGF) \[[@pntd.0005418.ref026]\]. These host-derived factors are thought to bind to parasite receptors and in turn influence the parasite's growth and development through exhibiting their impacts on the relevant evolutionarily conserved signaling systems within the parasite \[[@pntd.0005418.ref026]--[@pntd.0005418.ref027]\]. A recent study has evidenced that host insulin activates *E*. *multilocularis* PI3K/AKT signaling pathways and stimulates germinative cell proliferation and larval development \[[@pntd.0005418.ref028]\]. In addition, *in vitro* cultivation of *E*. *multilocularis* larvae and primary cells requires continuous presence of host cells as the feeder cells (like stem cell cultivation) or host cell-conditioned medium which contains host-derived growth factors \[[@pntd.0005418.ref029]\]. Together, lines of evidence offer compelling clues that the conserved signaling pathways in *E*. *multilocularis* could respond to host factors and may regulate germinative cells which are fundamental for the larval growth and development of the parasite \[[@pntd.0005418.ref030]\]. It has been shown that the Ras/Raf/MEK/ERK signaling in *E*. *multilocularis* is activated in response to a host-derived EGF signal, which is most probably mediated by the parasite's EGFR-like kinase \[[@pntd.0005418.ref031]--[@pntd.0005418.ref033]\]. Using *in vitro* cultivation systems we show here that *E*. *multilocularis* EGFR/ERK signaling pathway is activated upon addition of human EGF and promotes germinative cell proliferation during the parasite's larval growth and development. Methods {#sec005} ======= Ethics statement {#sec006} ---------------- All animal experiments were conducted in strict accordance with China regulations on the protection of experimental animals (Regulations for the Administration of Affairs Concerning Experimental Animals, version from July-18-2013) and specifically approved by the Institutional Animal Care and Use Committee of Xiamen University (Permit Number: 2013--0053). Parasite in vitro culture, growth and development assay, and treatment {#sec007} ---------------------------------------------------------------------- The parasite isolate used in this study was obtained from Hulunbeier Pasture of Inner Mongolia of China \[[@pntd.0005418.ref034]\] and maintained by *in vivo* propagation of the parasite material in mice (supplied by Xiamen University Laboratory Animals Center, XMULAC). *In vitro* cultivation of metacestode vesicles was performed using host cell conditioned medium according to a previously established protocol \[[@pntd.0005418.ref035]\] unless otherwise indicated. For the growth assay, vesicles (diameter \< 1mm) were manually picked up and cultured in 24-well cell culture plates supplemented with different media as indicated in the text. 100 ng/mL recombinant human EGF (PeproTech, Rocky Hill, NJ) was used for all experiments unless otherwise indicated. Parasite growth was determined by the measurement of vesicle's diameter under inverted microscope weekly. Each group contains at least 3 replicates and more than 150 vesicles in total for each group were analyzed. Two-three independent experiments were performed. For the treatment of inhibitors, the EGFR inhibitors CI-1033 and BIBW2992 or the MEK inhibitor U0126 (Selleck Chemicals, Houston, TX) was added into the culture medium at a final concentration as indicated. All experiments were performed with exchange of the medium containing the same ingredients every three days. Protoscoleces were collected from parasite material and *in vitro* cultured in conditioned medium. The vesicle formation process, in which protoscoleces dilate and vacuolate, were examined after 18 days of culture. 40 mM of hydroxyurea was used to treat metacestode vesicles as described before \[[@pntd.0005418.ref015]\]. BrdU incorporation assay, EdU labeling and EdU-BrdU dual labeling {#sec008} ----------------------------------------------------------------- Vesicles were incubated with BrdU for two days and chromosomal DNA was isolated for BrdU incorporation assay with the colorimetric BrdU ELISA kit (Roche, Mannheim, Germany) as described before \[[@pntd.0005418.ref028]\]. Vesicles were incubated with 50 μM of EdU for 4 hours and whole-mount prepared according to Cheng et al. \[[@pntd.0005418.ref036]\]. Click-iT-EdU Alexa Fluor 555 Imaging Kit (Life Technologies, Shanghai, China) was used for detection of EdU. For the EdU-BrdU dual labeling, vesicles were incubated with 10 μM EdU for 4 hours, washed and then cultured with no labeling for 44 hours. 10 μM BrdU was next used for continuous labeling for another 24 hours. Vesicles were fixed at the end of the labeling period and whole-mount prepared (see also [S2C Fig](#pntd.0005418.s002){ref-type="supplementary-material"}). After a 45-minute 2N HCl treatment, immunofluorescence was performed for BrdU detection using the anti-BrdU antibody (clone MoBU, Life Technologies) followed by EdU detection. DNA was counterstained with 4', 6-diamidino-2-phenylindole (DAPI) (Sigma, St. Louis, MO) for all labeling experiments. For the inhibitor experiments during the recovery from hydroxyurea treatment, vesicles were allowed to recover in the conditioned medium supplemented with EGF. 10 μM CI-1033 or 20 μM U0126 was then added into the medium immediately after the initial EdU pulse. Germinative cell proliferation was analyzed by EdU-BrdU dual labeling after 4 days of recovery. For the quantification of EdU^+^ or BrdU^+^ cells, at least 12 random microscopic fields from 4--6 vesicles were captured and the positive cells were manually counted. 3--5 labeling experiments were performed and analyzed for each control and treatment group. mRNA expression analysis of *E. multilocularis* EGFR members {#sec009} ------------------------------------------------------------ Total RNA was extracted from *in vitro*-cultivated protoscoleces or metacestode vesicles, treated with RNase-free DNase and reverse transcribed into cDNA. cDNAs were processed for RT-PCR analysis using the primers: EmER-qF2 (5'-GCG AAT GTA AGC ATT TCA AGT CA-3') and EmER-qR2 (5'-TTC ACA AAG TAG CAG AAA GCA CAT-3') for *Emer*; 617300-qF (5'-GCC GCA TCT ATG GAC ACGC-3') and 617300-qR (5'-AGT CAT CTT GTG GGA GGA ATCG-3') for Emuj_000617300; 969600-qF (5'-CTC TGG GGT GTC TGC TGT CC-3') and 969600-qR (5'-TCC CAC AGA GTC ACA CCG TAGG-3') for Emuj_000969600. Expression of EmER in *Xenopus* oocytes {#sec010} --------------------------------------- Expression of the parasite EGF receptor EmER in *Xenopus* oocytes was performed according to \[[@pntd.0005418.ref037]\]. Briefly, the full-length coding sequence of EmER was cloned into the *Xenopus* oocyte expression vector pXT7-flag (a gift from Dr. Li Guang in Xiamen University). Linearized plasmids were then used as the templates for capped mRNA (cRNA) synthesis using the T7 mMessage mMachine Kit (Ambion). Oocytes were obtained from *Xenopus laevis* (supported by Stem Cell Bank, Chinese Academy of Sciences) and then microinjected with EmER cRNA or water (noninjected control). Membrane proteins were extracted from 30 oocytes after 48 h of injection and immunoprecipitated by the anti-flag monoclonal antibody (Sigma) and analyzed by western blot. For EGF and CI-1033 treatment, oocytes that had been expressing EmER for 48 h were incubated with 10 μM of CI-1033 or DMSO for 4 h followed by 20 minutes of EGF stimulation. Membrane extracts were immunoprecipitated and analyzed by western blot using the anti-flag or anti-phospho-tyrosine (CST, Beverly, MA) monoclonal antibodies. Induction of GVBD (germinal vesicle breakdown) in EmER-expressing *Xenopus* oocytes was performed according to \[[@pntd.0005418.ref037]\]. Oocytes that had been expressing EmER for 48 h were pretreated with 10 μM of CI-1033 or DMSO for 4 h and then incubated with EGF. GVBD was monitored after 16 h of EGF incubation. As a positive control for GVBD, oocytes were stimulated with progesterone (PG), the natural inducer. 20--30 oocytes were used for each group and three independent experiments were performed. Western blot and immunofluorescence {#sec011} ----------------------------------- Lysates of *in vitro*-cultivated protoscoleces or metacestode vesicles were produced, separated on 12% acrylamide gels and transferred to PVDF membranes. Detection of *E*. *multilocularis* ERK and phosphorylated ERK was performed according to Spiliotis et al. \[[@pntd.0005418.ref032]\] using the polyclonal rabbit anti-ERK1/2 (Stressgen, Victoria, Canada) and anti-ERK1/2 \[pTpY^185/187^\] (Life Technologies) antibodies, respectively. The anti-rabbit IgG antibody conjugated with horseradish peroxidase (Theromo Scientific, Shanghai, China) was used as a secondary antibody. For all western blot experiments, detection of *E*. *multilocularis* actin was performed using the polyclonal anti-β-actin antibody (CST) as loading controls. Immunofluorescence was performed using the whole-mount prepared metacestode vesicles as described before \[[@pntd.0005418.ref036]\]. For Histone H3 detection, the anti-phospho-Histone H3 antibody (Ser10, 1:200) (CST) was used. For all immunofluorescence experiments, an Alexa 488-conjugated second antibody (Life Technologies) was used and DNA was counterstained with DAPI. Data analysis and statistics {#sec012} ---------------------------- Data of three or more experimental repeats are shown as mean ± SD as indicated in the respective figure legend unless otherwise indicated. The mean values of the data from the experimental groups were compared by performing two-tailed Student's t-test and the *P* values were indicated as those: \**P* \< 0.05, \*\**P* \< 0.01, and \*\*\**P* \< 0.001. Results {#sec013} ======= EGF stimulates *E. multilocularis* larval growth and development {#sec014} ---------------------------------------------------------------- To examine the impacts of EGF on the larval growth of *E*. *multilocularis*, metacestode vesicles were incubated in host cell-conditioned medium supplemented with recombinant human EGF. We found that 10 ng/mL or higher concentrations of EGF can greatly promote vesicle's growth ([S1 Fig](#pntd.0005418.s001){ref-type="supplementary-material"}). We then used 100 ng/mL of EGF which showed the most significant effect on parasite's growth *in vitro* for further studies. The results show that addition of EGF can stimulate the growth of metacestode vesicles ([Fig 1A](#pntd.0005418.g001){ref-type="fig"}). Similar results were observed after EGF was added into the Dulbecco's modified eagle medium (DMEM) containing 10% serum ([Fig 1B](#pntd.0005418.g001){ref-type="fig"}). We also found that EGF greatly promoted the vesicle formation process, in which the protoscoleces dilated and vacuolated ([Fig 1C](#pntd.0005418.g001){ref-type="fig"}). These results illustrate that the larval growth and development of *in vitro*-cultivated *E*. *multilocularis* larvae could be stimulated by exogenously added EGF, which is probably mediated by an EGFR-dependent signaling in the parasite. {#pntd.0005418.g001} Considering that either serum or conditioned medium contains complex ingredients, we then incubated vesicles in the serum-free DMEM only supplemented with EGF. Addition of EGF could not obviously stimulate the growth of vesicles, however, it remarkably promoted their survival ([Fig 1D](#pntd.0005418.g001){ref-type="fig"}). Although metacestode vesicles could not survive for long in this situation, the method excludes the influence of other host factors in serum and suggests that the parasite is responsive to host EGF stimulation. Further experiments in this study were all performed using host cell-conditioned medium unless otherwise indicated. EGF promotes the proliferation of *E. multilocularis* germinative cells {#sec015} ----------------------------------------------------------------------- Given that the germinative cells, a population of stem cell-like cells, drive larval growth and development of *E*. *multilocularis*, we then investigated the impacts of EGF on the germinative cells. Vesicles were incubated with 5-bromo-2'-deoxyuridine (BrdU), an analogue of thymidine used for studying cell proliferation by detecting its incorporation into the newly synthesized DNA of replicating cells. The result shows that addition of EGF greatly stimulated the BrdU incorporation in the vesicles ([Fig 2A](#pntd.0005418.g002){ref-type="fig"}). Since germinative cells are the only cells capable of proliferation in metacestode vesicles, this result suggests that EGF promotes germinative cell proliferation. {#pntd.0005418.g002} A dual labeling method through sequential pulses of 5-ethylnyl-2'-deoxyuridine (EdU) and BrdU, which has been utilized to verify the self-renewal capacity of adult somatic stem cells in the human blood fluke *Schistosoma mansoni* \[[@pntd.0005418.ref014]\], was further applied to determine the effects of EGF on germinative cells. EdU is a newly found analogue of thymidine \[[@pntd.0005418.ref038]\], which has been shown to be incorporated by proliferating cells of *E*. *multilocularis* \[[@pntd.0005418.ref015]\]. Under normal *in vitro* cultivation conditions, we found that most of the dividing germinative cells in the vesicles could incorporate EdU before a chase period of 44 h, which was further used for EdU-BrdU dual labeling experiments ([S2A Fig](#pntd.0005418.s002){ref-type="supplementary-material"}). The results show that about 48% of cells that initially incorporate EdU are BrdU^+^ 3 days after an initial EdU pulse (749 EdU^+^BrdU^+^ / 1554 EdU^+^ nuclei, 4 independent labeling experiments) ([Fig 2B](#pntd.0005418.g002){ref-type="fig"}), indicating that these germinative cells divide and produce proliferation-competent daughter cells that initially incorporate EdU can incorporate BrdU during the second replication. It has been shown that depletion of the germinative cells in *E*. *multilocularis* vesicles could be achieved through hydroxyurea treatment for longer periods of time (*e*.*g*. seven days) and that the germinative cells undergo clonal expansion like stem cells after removal of hydroxyurea \[[@pntd.0005418.ref015]\]. We also performed similar experiments and found that the EdU^+^BrdU^+^ cells are highly presented in the clonally growing germinative cells ([S3 Fig](#pntd.0005418.s003){ref-type="supplementary-material"}), suggesting that the EdU^+^BrdU^+^ cells are extensively proliferating germinative cells and a part of them might be undergoing self-renewing divisions. To further investigate the effect of EGF on germinative cells, we treated vesicles with hydroxyurea to eliminate most germinative cells \[[@pntd.0005418.ref015]\]. After removal of hydroxyurea, germinative cell proliferation/self-renewal was allowed for recovery in the medium with addition of EGF or not for 4 days, and the EdU-BrdU sequential pulses began on the second day of the recovery (see also [S2B and S2C Fig](#pntd.0005418.s002){ref-type="supplementary-material"}). At the end of the dual labeling period, the results show that addition of EGF induced significantly more numbers of both EdU^+^ and EdU^+^BrdU^+^ cells in the vesicles compared to the controls ([Fig 2C](#pntd.0005418.g002){ref-type="fig"}). We found that the proportion of EdU^+^BrdU^+^ cells with respect to the number of EdU^+^ cells was also greatly increased (25.6% and 50.5% for the control and EGF-treated groups, respectively, statistical significance *P* = 0.00101), suggesting increased continuous proliferation, and possibly promoted self-renewal of the germinative cells upon EGF stimulation. Together, these results support the findings that addition of EGF can promote the proliferation of germinative cells ([Fig 2A](#pntd.0005418.g002){ref-type="fig"}), which subsequently drives the larval growth and development of *E*. *multilocularis*. These results also suggest that an EGFR-dependent signaling in the parasite may be involved in regulating germinative cell proliferation upon EGF stimulation. EGFR/ERK signaling contributes to germinative cell proliferation {#sec016} ---------------------------------------------------------------- *E*. *multilocularis* possesses an EGFR-like kinase (EmER) which is suggested to interact with host EGF \[[@pntd.0005418.ref031], [@pntd.0005418.ref033]\]. Besides EmER, we also found two additional *E*. *multilocularis* EGF receptor members (Emuj_000617300 and Emuj_000969600) by analyzing *E*. *multilocularis* genome sequence (<http://www.genedb.org/Homepage/Emultilocularis>) and fully cloned and sequenced the respective genes. Comparisons of the putative protein sequences reveal that these EGF receptor members exhibit significant homologies to human EGFR and the EGFR homologue of the closely related schistosome *S*. *mansoni*, especially in the tyrosine kinase domains ([S4A--S4D Fig](#pntd.0005418.s004){ref-type="supplementary-material"}). The results of mRNA expression analysis show that these EGF receptor homologues are constitutively present in *E*. *multilocularis* metacestode vesicles and protoscoleces ([S4E Fig](#pntd.0005418.s004){ref-type="supplementary-material"}). To investigate whether the *E*. *multilocularis* EGF receptor(s) respond to EGF stimulation or not, we utilized the *Xenopus* oocyte expression system, which is a powerful tool for receptor tyrosine kinase research and has been successfully used for studying the EGF receptor (SER) in *S*. *mansoni* \[[@pntd.0005418.ref037]\]. The results show that the parasite EGFR EmER could be efficiently expressed in *Xenopus* oocytes with a molecular weight approximately 200 kDa ([Fig 3A](#pntd.0005418.g003){ref-type="fig"}), and that addition of EGF resulted in the activation of EmER by detection of phosphorylated tyrosine ([Fig 3B](#pntd.0005418.g003){ref-type="fig"}). It has previously been shown that host EGF could induce germinal-vesicle breakdown (GVBD) in *S*. *mansoni* SER-expressing oocytes \[[@pntd.0005418.ref037]\]. Similar results were also observed in the EmER-expressing oocytes ([Fig 3C](#pntd.0005418.g003){ref-type="fig"}), suggesting that addition of exogenous EGF could activate the parasite EGFR in the oocytes and induce GVBD. Using the oocyte system, we also found that CI-1033 (canertinib), an irreversible inhibitor for human EGF receptors \[[@pntd.0005418.ref039]\], could effectively inhibit the EGF-induced activation of EmER and GVBD ([Fig 3B and 3C](#pntd.0005418.g003){ref-type="fig"}). {#pntd.0005418.g003} We wondered if the impaired EGFR activation would impact *E*. *multilocularis* germinative cell behaviors. To this end, we treated vesicles with CI-1033. The results show that CI-1033 significantly reduced the number of EdU^+^ cells in the vesicles ([Fig 4A](#pntd.0005418.g004){ref-type="fig"}). Another EGFR inhibitor, BIBW2992 (afatinib) \[[@pntd.0005418.ref040]\], also exhibited a similar effect to CI-1033 on the germinative cells ([S5A Fig](#pntd.0005418.s005){ref-type="supplementary-material"}). {#pntd.0005418.g004} In the EdU-BrdU dual labeling experiments, vesicles were allowed to recover from the hydroxyurea treatment with addition of EGF, and CI-1033 was administrated to the vesicles immediately after the initial EdU pulse. At the end of the labeling period, only 1.6‰ of total cells were EdU^+^ (11 EdU^+^ / 6737 DAPI nuclei), and the EdU^+^BrdU^+^ cells were hardly detected ([Fig 4B](#pntd.0005418.g004){ref-type="fig"}). There is a strong accumulation of EdU^+^ cells in numerous aggregates in some developing vesicles, which indicates that the active proliferation and extensive self-renewal of germinative cells may drive the development of brood capsule and protoscolex in the vesicles \[[@pntd.0005418.ref015]\]. Our studies show that CI-1033 can also abolish the accumulation of EdU^+^ cells in these cell aggregates ([Fig 4C](#pntd.0005418.g004){ref-type="fig"}). Further investigations showed that CI-1033 and BIBW2992 can significantly inhibit the larval growth and development upon EGF stimulation ([Fig 4D](#pntd.0005418.g004){ref-type="fig"} and [S5B Fig](#pntd.0005418.s005){ref-type="supplementary-material"}). These results suggest that an EGFR-dependent signaling in the parasite is required for the promoted germinative cell proliferation and larval growth, and that the signaling could probably be impaired by the EGFR inhibitors initially designed against human EGF receptors. In *E*. *multilocularis*, structural and functional homologues to mammalian MAP kinase cascade molecules, such as RAF, MEK and ERK, have been identified and characterized \[[@pntd.0005418.ref027]\]. Previous study has shown that host EGF could induce *E*. *multilocularis* ERK activation \[[@pntd.0005418.ref032]\]. Our data show that the basal level of ERK phosphorylation in the vesicles was down-regulated following CI-1033 treatment in a time-dependent manner ([Fig 5A](#pntd.0005418.g005){ref-type="fig"}). We also found that CI-1033 can inhibit the phosphorylation of ERK induced by EGF ([Fig 5B](#pntd.0005418.g005){ref-type="fig"}). These results suggest that the EGFR inhibitors could impair the activations of parasite's EGFR and ERK. {#pntd.0005418.g005} We then treated vesicles with U0126, a MEK inhibitor which effectively suppressed both of the basal and the addition of EGF-induced ERK phosphorylations in the parasite ([Fig 5C and 5D](#pntd.0005418.g005){ref-type="fig"}). Along with the inhibition of MEK/ERK activity, a remarkable decrease in the number of EdU^+^ germinative cells was observed in the vesicles ([Fig 5E](#pntd.0005418.g005){ref-type="fig"}). EdU-BrdU dual labeling experiments also indicate that U0126 has a comparable inhibition effect to CI-1033 on the EGF-promoted germinative cell proliferation (2.9‰ of the total cells were EdU^+^ and the EdU^+^BrdU^+^ cells were hardly detected) ([Fig 5F](#pntd.0005418.g005){ref-type="fig"}). Our further investigation shows that U0126 can significantly inhibit the EGF-stimulated vesicle growth ([Fig 5G](#pntd.0005418.g005){ref-type="fig"}). Taken together, these results suggest that the EGF-mediated EGFR/MEK/ERK signaling contributes to germinative cell proliferation during *E*. *multilocularis* larval growth and development. Discussion {#sec017} ========== Throughout the complex life cycle of *E*. *multilocularis*, the parasite always keeps a population of stem cell-like cells, the germinative cells, which are considered as one of the fundamental underpinnings of its growth and development in the host \[[@pntd.0005418.ref015]\]. Stem cell maintenance and functions are strictly controlled by signals from the local tissue microenvironments known as "niches", which have been widely characterized and elucidated for mammals and invertebrate model animals \[[@pntd.0005418.ref018]\]. Proliferation and differentiation of neoblasts, the stem cells in the free-living flatworm planarian, are also regulated by signals from the surrounding cells \[[@pntd.0005418.ref011]\]. However, the situation is somewhat different for the host liver tissue-dwelling metacestode larvae of *E*. *multilocularis*. Due to the intimate parasite-host contact, the parasite is believed to be able to sense signals derived not only from its own tissue but also from the host-derived hormones and cytokines, as these signaling receptors and downstream signaling cascades are evolutionarily conserved between the mammalian hosts and *E*. *multilocularis* \[[@pntd.0005418.ref027]\]. Thus it is tempting to suggest that *E*. *multilocularis* germinative cells should be regulated by the host-derived factors \[[@pntd.0005418.ref030]\]. In the present study, we show that germinative cell proliferation of the *in vitro*-cultivated *E*. *multilocularis* larvae is promoted upon the addition of human EGF, which in turn drives vesicle growth and vesicle formation from protoscolex (Figs [1](#pntd.0005418.g001){ref-type="fig"} and [2](#pntd.0005418.g002){ref-type="fig"}). These results suggest that the germinative cells are regulated by the signaling pathways within the parasite that could sense the host-derived EGF signal. A physiologically relevant concentration of EGF (1 ng/mL) showed a modest effect on the growth of *in vitro*-cultivated vesicles, while 10 ng/mL or higher concentrations of EGF greatly stimulated the growth ([S1 Fig](#pntd.0005418.s001){ref-type="supplementary-material"}). We then used 100 ng/mL of EGF, which exhibited the most significant effect on parasite's growth, for further *in vitro* studies. This concentration of EGF is considerable and widely used in human cancer cell research, and is relevant to those used in the *in vitro* studies of *E*. *multilocularis* and *S*. *mansoni* \[[@pntd.0005418.ref032], [@pntd.0005418.ref037]\]. However, it could exceed the physiological concentrations in liver. Considering that the larval development of *E*. *multilocularis* causes host liver tissue destruction and regeneration while EGF is continually made available to the liver and strongly produced during regeneration processes \[[@pntd.0005418.ref041]--[@pntd.0005418.ref042]\], it will be interesting to investigate the effects of host EGF observed in this study on the parasite in future using *in vivo* infection models. The promoted proliferation of germinative cells upon EGF stimulation was supported by our EdU-BrdU dual labeling experiments ([Fig 2B and 2C](#pntd.0005418.g002){ref-type="fig"} and [S3 Fig](#pntd.0005418.s003){ref-type="supplementary-material"}), which also suggest that the cell-cycle time for most of the actively proliferating germinative cells is less than three days ([S2 Fig](#pntd.0005418.s002){ref-type="supplementary-material"} and [Fig 2B](#pntd.0005418.g002){ref-type="fig"}), similar to that for *S*. *mansoni* adult somatic stem cells \[[@pntd.0005418.ref014]\]. It has been suggested that there are subpopulations of the germinative cells capable of maintaining their pluripotency and self-renewing like stem cells \[[@pntd.0005418.ref015]\]. Although our data suggest that the increased number of EdU^+^BrdU^+^ cells (as well as the increased ratio of EdU^+^BrdU^+^ cells to EdU^+^ cells) may also result from the promoted self-renewal of these stem-like cell populations, due to the limitations of the dual labeling experiments we still could not distinguish the self-renewing cells from the transit amplifying cells. Specific molecular markers of the stem-like cell populations that work independently of proliferation would be needed for further explorations to clarify the contribution of EGF to germinative cell self-renewal in *E*. *multilocularis*. Addition of host EGF promotes germinative cell proliferation, which then promotes *in vitro*-cultivated protoscoleces to form metacestode vesicles ([Fig 1C](#pntd.0005418.g001){ref-type="fig"}). The formation of vesicle from protoscolex occurs *in vivo* following the rupture of parasite cysts and distribution of protoscoleces, and is thought to contribute to prolonged parasite survival in the intermediate host \[[@pntd.0005418.ref019], [@pntd.0005418.ref028]\], which would result in a poor prognosis after surgery-induced rupture of parasite cysts in human echinococcosis, at least in cystic echinococcosis (CE). Interestingly, we found that addition of host EGF may not only promote this formation process but also initiate it by triggering activation of the germinative cells from a quiescent state in the developed protoscoleces ([S6 Fig](#pntd.0005418.s006){ref-type="supplementary-material"}). Koziol *et al*. \[[@pntd.0005418.ref015]\] indicated that there is a large population of germinative cells capable of proliferation in the developed protoscolex, but they remain in a quiescent state or with slow cell-cycle kinetics for as long as the protoscolex remains resting within the metacestode. These quiescent germinative cells were activated when the protoscoleces were activated by artificially mimicking the ingestion by the definitive host or when the protoscoleces were *in vitro* cultured in serum-containing DMEM. Thus it is tempting to suggest that host factors activate the quiescent germinative cells to re-enter the cell-cycle for proliferation and self-renewal, which may further stimulate the protoscoleces to mature into adults within the definitive host's intestine or to form metacestode vesicles in the intermediate host's liver and other tissues. It is still unclear how protoscoleces alternate between developmental fates: the adult or metacestode vesicle. In any case, this unique development potential for protoscoleces is based on the germinative cells, which may response to different host-derived signals from different host tissue microenvironments. Our data suggest that host factors may play a vital role in host-parasite interaction via mediating the relative signaling pathways in the parasite to regulate germinative cell functions. It has been suggested that host EGF could activate the highly conserved Ras/Raf/MEK/ERK signaling cascade in *E*. *multilocularis*, which is probably mediated by the parasite's EGF-receptor-like kinase \[[@pntd.0005418.ref031]--[@pntd.0005418.ref033]\]. Besides EmER, the first EGFR homologue identified in *E*. *multilocularis* \[[@pntd.0005418.ref031]\], two additional members of the EGFR family could be identified, which display significant homologies with human EGFR and *S*. *mansoni* EGFR in the functional domains ([S4A--S4D Fig](#pntd.0005418.s004){ref-type="supplementary-material"}). Like EmER, these two EGFR homologues are continually expressed in *E*. *multilocularis* metacestode vesicles and protoscoleces ([S4E Fig](#pntd.0005418.s004){ref-type="supplementary-material"}). It will be interesting to clarify their roles as the EGF receptor kinase in the parasite's development within the host in the future work. In this study, using the *Xenopus* oocyte expression system we show that one of the EGFR homologues EmER can be activated by host EGF ([Fig 3](#pntd.0005418.g003){ref-type="fig"}). We also show here that inhibition of the MEK/ERK signaling activation by either the EGFR inhibitors CI-1033 and BIBW2992 or the MEK inhibitor U0126 significantly impaired *E*. *multilocularis* germinative cell proliferation, larval growth and development (Figs [4](#pntd.0005418.g004){ref-type="fig"} and [5](#pntd.0005418.g005){ref-type="fig"} and [S5 Fig](#pntd.0005418.s005){ref-type="supplementary-material"}). In mammals, the MEK/ERK pathway plays a critical role in regulating stem cells. For example, it is required for maintenance of stemness and self-renewal of mouse neural stem/precursor cells \[[@pntd.0005418.ref043]--[@pntd.0005418.ref044]\]. The role of MEK/ERK pathway in mammalian embryonic stem cells (ESCs) is much more complex. The MEK/ERK signaling plays a functional role in promoting differentiation of mouse ESCs, while it promotes self-renewal of human ESCs (reviewed in \[[@pntd.0005418.ref020]\] and \[[@pntd.0005418.ref045]\]). In invertebrates, much of what is known about the role of the MEK/ERK signaling in regulation of stem cells derives from the studies of *Drosophila*. It has been extensively documented that the EGFR-dependent activation of the MEK/ERK signaling pathway is essential for promoting the maintenance and self-renewal of various types of adult somatic stem cells in *Drosophila* \[[@pntd.0005418.ref023]--[@pntd.0005418.ref024], [@pntd.0005418.ref046]\]. Our data also show that EGF-promoted germinative cell proliferation and larval growth rely on the activation of the parasite's EGFR/ERK signaling. However, it still remains unclear that the contribution of EGFR/ERK signaling to the promoted proliferation is attributed to the direct response in the germinative cells or to the indirect response to a second signal produced by their surrounding differentiated cells upon EGF stimulation, or to both. Further analysis of the EGFR activation in proliferating germinative cells would be needed to clarify this issue. While downstream of EGFR lies the PI3K/AKT, MEK/ERK and STAT3 pathways, our findings define here that the MEK/ERK pathway contributes to the role of EGFR signaling in regulating *E*. *multilocularis* germinative cell proliferation. Considering that the PI3K/AKT pathway in *E*. *multilocularis* was recently suggested to be involved in the host insulin-stimulated germinative cell proliferation \[[@pntd.0005418.ref028]\], we also treated metacestode vesicles with the PI3K inhibitor LY294002, which was shown to effectively inhibit *E*. *multilocularis* PI3K activity \[[@pntd.0005418.ref028]\]. We found that LY294002 did not exhibit as obvious an inhibitory effect as U0126 on the proliferation of germinative cells but slightly decreased the number of EdU^+^ cells in metacestode vesicles. Future work would be required for evaluating the contribution of EGFR/AKT/PI3K signaling to the regulation of *E*. *multilocularis* germinative cells. Increasing evidence has shown that the inhibitors originally designed against the human kinases can effectively inhibit the activity of related kinases in *E*. *multilocularis* \[[@pntd.0005418.ref028], [@pntd.0005418.ref047]--[@pntd.0005418.ref049]\]. Based on the evolutionary conservation among the kinases of vertebrates and invertebrates (including invertebrate parasites), it has been widely considered that small molecules that target human kinases are promising drug candidates for treating human helminthiasis, including echinococcosis \[[@pntd.0005418.ref050]--[@pntd.0005418.ref051]\]. Considering our observations that both basal ERK phosphorylation and ERK phosphorylation induced by administered EGF were effectively suppressed by either CI-1033 or U0126 ([Fig 5A--5D](#pntd.0005418.g005){ref-type="fig"}), it is therefore conceivable that the kinase inhibitors used in this study could impair EGFR/ERK signaling in *E*. *multilocularis*. Although these inhibitors were used within the range concentrations required for these compounds to specifically inhibit their respective targets in humans, it is possible that they may also have cellular targets other than EGFR/ERK signaling in *E*. *multilocularis*. Our findings suggest that exogenous EGF-activated EGFR/ERK signaling in the parasite was inhibited by CI-1033. Given that the long-term *in vitro* maintenance of *E*. *multilocularis* larvae and primary cells requires continuous presence of host cell-derived growth factors \[[@pntd.0005418.ref029]\], it is reasonable to assume that EGFR inhibitors could impair the activation of parasite's EGFR upon host EGF stimulation, which might be the main reason for the diminished germinative cell proliferation and the impaired larval growth and development ([Fig 4](#pntd.0005418.g004){ref-type="fig"} and [S5 Fig](#pntd.0005418.s005){ref-type="supplementary-material"}). However, our data could not exclude the possibility that EGFR inhibitors may also impair the parasite's EGFR activation mediated by its own EGF molecules. It has been recently shown that neoblast clonal expansion in the free-living flatworm planarian is regulated by its own EGF-mediated EGFR signaling \[[@pntd.0005418.ref052]\]. Since the parasite also possesses a putative EGF homologue \[[@pntd.0005418.ref053]\], this endogenous EGFR ligand-mediated signaling might also be involved in regulating *E*. *multilocularis* germinative cells. Improved *in vitro* cultivation systems and methods that could avoid/reduce the impacts of host factors will be helpful to investigate the roles of this endogenous signaling in germinative cell regulation. Although stem cell-like germinative cells has been widely described in tapeworms and their roles in the parasite's development within the host are thought to be of fundamental importance, there are still long standing gaps in our knowledge of mechanisms controlling the behavior of these cells. This study defines an essential role for the EGF-mediated EGFR/ERK signaling in promoting germinative cell proliferation in *E*. *multilocularis*. It makes an effort to unravel the mechanisms of regulation of tapeworm germinative cells in response to host-derived growth factors, and helps in understanding the delicate developmental strategies of these parasites within the host. Targeting the signaling pathways involved in regulating germinative cells may provide a novel therapeutic strategy against echinococcosis and other human cestodiasis. Supporting information {#sec018} ====================== ###### Effects of EGF on *E. multilocularis* larval growth. Metacestode vesicles were cultivated in conditioned medium (Control) supplemented with 1--100 ng/mL recombinant human EGF for 49 days. Vesicle growth is shown as the increase of vesicle diameter as compared to day 0 for each group. Data are shown as mean ± SD of triplicates, representative of two independent experiments. (TIF) ###### Click here for additional data file. ###### EdU-BrdU dual labeling experiments. \(A\) EdU labeling and phospho-Histone H3 (Ser10, p-H3) immunofluorescence. Metacestode vesicles were administrated to a 4-h EdU pulse, and after 44 hours of pulse about 88% (365/413) of p-H3^+^ mitotic cells are EdU^+^. The chase period of 44 h was then used for EdU-BrdU dual labeling experiments. Note the low percentage of p-H3^+^ cells, which is consistent with a previous report \[[@pntd.0005418.ref015]\]. (B) Analysis of proliferating germinative cells by EdU labeling in hydroxyurea (HU) treatment experiments. Metacestode vesicles were treated with 40 mM of hydroxyurea for three days and then allowed for recovery in conditioned medium. Representative images are shown as: no treatment control, hydroxyurea treatment, and 4 days of recovery after removal of hydroxyurea (red: EdU; blue: DAPI). Bar = 40 μm. (C) Timeline for hydroxyurea treatment and EdU-BrdU dual labeling. Metacestode vesicles were pretreated with 40 mM of hydroxyurea for three days. EGF was immediately added into the conditioned medium after removal of hydroxyurea. Sequential pulses of EdU and BrdU began at 96 h after removal of hydroxyurea. Dual labeling under normal culture conditions (related to [Fig 2B](#pntd.0005418.g002){ref-type="fig"}) was carried out without hydroxyurea treatment, which is: EdU label for 4 hours, no label for 44 hours, and BrdU label for 24 hours. (TIF) ###### Click here for additional data file. ###### EdU^+^BrdU^+^ cells are highly presented in the clonally proliferating germinative cells. Metacestode vesicles were treated with 40 mM hydroxyurea (HU) for seven days and then transferred to HU-free medium. Samples were administrated to EdU-BrdU dual labeling at the day 3 after HU removal. Dashed line boxes in (A) indicate patches of EdU^+^ cells clonally growing. Bar = 100 μm. The magnified views are shown as in (B). Bar = 20 μm. (TIF) ###### Click here for additional data file. ###### Analysis of amino acid sequence and mRNA expression of the EGF receptor members of *E. multilocularis*. (A)-(C) Amino acid sequence analysis of the receptor-L-domain 1 (A), receptor-L-domain 2 (B) and Pkinase \_Tyr domain (C) of human (Hs), *S*. *mansoni* (Sm) and *E*. *multilocularis* (Em) EGF receptors. Domains are predicted using the online software (<http://scansite3.mit.edu/>). Positions at which all of the residues are conserved are shaded in black. (D) Similarities of *E*. *multilocularis* EGF receptor members to human EGFR. Similarity values to the L-C-L domain (two receptor L domains separated by a cysteine-rich furin-like region) and the kinase domain are indicated below as % identical residues (not bracketed) and % similar residues (bracketed). Further indicated are the similarities of overall protein sequences. (E) RT-PCR analysis of mRNA expression of *E*. *multilocularis* EGF receptor members in protoscoleces (lane 1--3) and metacestode vesicles (lane 4--6). Lane 1 and 4: EmER. Lane 2 and 5: Em_000617300. Lane 3 and 6: Em_000969600. M indicates the DNA marker. (TIF) ###### Click here for additional data file. ###### BIBW2992 impairs *E. multilocularis* germinative cell proliferation, larval growth and development. \(A\) Representative images of EdU^+^ germinative cells in the metacestode vesicles following treatment of 5 μM BIBW2992 or DMSO control for 3 days (red: EdU; blue: DAPI). Bar = 20 μm. (B) Effects of BIBW2992 on the larval growth and development. Vesicles or protoscoleces were cultivated in the DMSO-containing conditioned medium (control) supplemented with the ingredients as indicated. Vesicle growth (left) and vesicle formation from protoscoleces (right) were analyzed after 28 days and 18 days of cultivation, respectively. Data are shown as mean ± SD of triplicates, representative of 2--3 independent experiments. \*\*\* *P* \< 0.001. (TIF) ###### Click here for additional data file. ###### EGF stimulates the quiescent germinative cells in the developed protoscoleces. Protoscoleces freshly isolated from the metacestode material were *in vitro* maintained in PBS supplemented with EGF or not for 12h followed by a 4-hour pulse of EdU. Few EdU^+^ cells presented in the developed protoscoleces (arrows), however, the number of EdU^+^ cells dramatically increased after EGF stimulation. The arrow head indicates a developing protoscolex which possesses plenty of EdU^+^ cells. Bar = 100 μm. (TIF) ###### Click here for additional data file. The authors would like to thank Prof. Tang Chongti for providing the parasite material, Dr. Li Guang and Prof. Zuo Zhenghong for their assistance in *Xenopus* oocyte microinjection. [^1]: The authors have declared that no competing interests exist. [^2]: **Conceptualization:** ZC YW.**Formal analysis:** ZC FL XL.**Funding acquisition:** YW ZC FL HT.**Investigation:** ZC FL XL MD JW XG HT ZH YL XC.**Methodology:** ZC FL XL XG HT ZH YL.**Project administration:** YW ZC.**Resources:** ZC FL XL MD JW XG HT ZH YL XC.**Supervision:** YW.**Validation:** YW.**Writing -- original draft:** ZC FL YW.

All relevant data are within the paper and its Supporting Information files. Introduction {#sec001} ============ Acid hot springs provide unique environments for the evolution and establishment of microbial communities and their response to various biogeochemical and metabolic processes involving hydrogen (H~2~), sulfur (S), iron (Fe) and arsenic (As) \[[@pone.0146331.ref001]--[@pone.0146331.ref006]\]. Generally, all acid hot springs are classified into two different water types: acid sulfate-Cl water and sulfate water with low Cl \[[@pone.0146331.ref007], [@pone.0146331.ref008]\]. Acid sulfate-Cl hot springs are viewed to form by sulfide oxidation after saturated Na-Cl geothermal water is exposed to the Earth\'s surface, and thus have decreased pH values but retain high K, Na, F, Cl, Li, B and As concentrations as geothermal reservoir water \[[@pone.0146331.ref007]\]. The high As concentration of these environments has attracted attention of several groups studying As redox speciation \[[@pone.0146331.ref001], [@pone.0146331.ref009]--[@pone.0146331.ref011]\]. Results of previous studies illustrated that arsenite (As(III)) was predominant in the source water of acid sulfate-Cl hot springs, and was oxidized to arsenate (As(V)) once being discharged along the outlet by bacterial populations such as *Hydrogenobacter*, *Hydrogenobaculum*, *Sulfurihydrogenibium*, and *Thiomonas* \[[@pone.0146331.ref001], [@pone.0146331.ref009], [@pone.0146331.ref011], [@pone.0146331.ref012]\]. No As(III) oxidation in the source water of acid sulfate-Cl hot springs was observed because of sulfide inhibition by inactivating expressed As(III) oxidase enzyme (Aio) \[[@pone.0146331.ref004], [@pone.0146331.ref013]\]. Due to high concentrations of sulfide, Fe and As in those acid sulfate-Cl hot springs, elemental S and Fe depositions rich in As successively appeared along the outflow channel \[[@pone.0146331.ref009], [@pone.0146331.ref010], [@pone.0146331.ref014]\]. Various microbial populations such as *Sulfolobus*, *Sulfobacillus*, *Metallosphaera*, *Sufurihydrogenibium*, *Hydrogenobaculum*, *Thiomonas* and *Acidicaldus* had been found to respond to sulfide and Fe oxidations in those acid hot springs \[[@pone.0146331.ref001], [@pone.0146331.ref009], [@pone.0146331.ref014]--[@pone.0146331.ref016]\]. Comparatively, acid sulfate hot springs with low Cl have not studied systematically on As biogeochemistry. These geothermal features form by separation of the vapor phase rich in H~2~S from the reservoir and subsequent condensation and oxidation in shallow oxygen-rich groundwater or surface water \[[@pone.0146331.ref017]\]. Tengchong, located in southwestern of China, is a typical volcanic geothermal area and has abundant geothermal resources \[[@pone.0146331.ref018]\]. Zhenzhuquan in the Rehai geothermal field of Tengchong is a representative acid vapor-formed sulfate hot spring low in Cl, with 128.2 mg/L sulfate, 39.2 mg/L Cl and 71.1 μg/L As, as well as low concentrations of K, Na, Li and B \[[@pone.0146331.ref019]\]. This geochemistry is distinctly different from the previously studied acid sulfate-Cl hot springs (sulfate: 21.0--144.9 mg/L; Cl: 473.6--1907.0 mg/L; As: 1.8--5.3 mg/L) \[[@pone.0146331.ref001], [@pone.0146331.ref009]--[@pone.0146331.ref011]\]. The As concentrations in the sediments of Zhenzhuquan outflow channel occur at up to 16.44 g/kg, which substantially exceeded the terrestrial abundance of As (1.5--3 mg/kg) and could pose a potential environmental risk \[[@pone.0146331.ref020]\]. Previous studies indicated that different microbial communities inhabited geothermal environments with distinct geochemistry, including different media such as water and sediment at the same sites of hot springs \[[@pone.0146331.ref021], [@pone.0146331.ref022]\]. Though some microbial studies on As in mats or sediments along those acid sulfate-Cl hot springs outlet had been conducted by clone library and metagenome sequencing \[[@pone.0146331.ref001], [@pone.0146331.ref009], [@pone.0146331.ref010], [@pone.0146331.ref023]\], arsenic geochemistry and corresponding microbial communities in acid sulfate hot spring with low Cl have yet to be fully understood. Therefore, the objectives of this study were to: (1) investigate the As geochemistry and microbial community structures both in water and sediments along the outflow channel of a typical acid sulfate hot spring with low Cl; (2) evaluate the potential microbially-mediated As oxidation process; and (3) assess the environmental factors shaping the microbial community structures. Materials and Methods {#sec002} ===================== Site description {#sec003} ---------------- No specific permission was required for the described field studies because no animal or human subjects were involved in this research. The sampling locations are not privately owned or protected in any way. The field studies were not involved in endangered or protected species. As mentioned above, Zhenzhuquan (N24.9511°, E98.4361°) located in the Rehai geothermal field of Tengchong geothermal area in Yunnan, southwestern China, was selected for this study ([Fig 1A](#pone.0146331.g001){ref-type="fig"}). This spring is a heart-shaped acidic pool with a depth of 6--7 cm and a length of 4.36 m and is fed by numerous very small vigorous degassing vents ([Fig 1B](#pone.0146331.g001){ref-type="fig"}) \[[@pone.0146331.ref022]\]. It has a seasonally fluctuating temperature from 89.1 to 93.3°C and pH from 3.50 to 6.42, which results from the dilution of rainfall \[[@pone.0146331.ref017], [@pone.0146331.ref024]\]. Zhenzhuquan has a low discharge rate of 0.2 L/S and abundant reddish-brown sediments occur downstream from the source ([Fig 1C](#pone.0146331.g001){ref-type="fig"}). These sediments contain large amounts of As with concentrations up to 16.44 g/kg, even though As concentration is extremely low (48.16 μg/L) in the pool water \[[@pone.0146331.ref025]\]. Consequently, a transect of six sampling sites was established along Zhenzhuquan's outflow channel and each sampling site was assigned a name according to the relative distance from commenced discharge point (0 m), such as 3 m, 6 m and 9 m downstream and -2 m and -1 m in the pool ([Fig 1B](#pone.0146331.g001){ref-type="fig"}). Parallel water and sediment samples at each site were collected in August, 2014. {#pone.0146331.g001} Field measurements and sample collection {#sec004} ---------------------------------------- Water temperature, pH and dissolved oxygen (DO) were measured in the field at the site of water collection using a hand-held meter. Concentrations of sulfide, ammonium, ferrous iron (Fe(II)) and total iron (Fe~Tot~) were also determined in the field with a Hach spectrophotometer (DR850, Hach Corp., USA) according to the manufacturer's instructions. Water samples for laboratory measurements, e.g., anions, cations and dissolved organic carbon (DOC) were collected into 50 mL acid-washed polypropylene bottles and brown glass bottles respectively by filtration of spring water through 0.22 μm syringe polyethersulfone (PES) membrane filters (Pall Corp., NY, USA). Filtered biomass-containing membranes were placed into 15 mL sterile polypropylene tubes and immediately stored in dry ice. Water samples for cations and DOC were acidified with 1% v/v HNO~3~. Arsenic species separation was done on site, following the method reported by Le et al. \[[@pone.0146331.ref026]\]. Briefly, 10 mL of each water sample was passed through a silica-based strong anion-exchange cartridge (Supelco, USA) preconditioned with 50% methanol and deionized water before use. As(V) was retained in the cartridge and As(III) remained in the filtered solution. Subsequently, the cartridge was eluted with 10 mL 1 M HCl to release the bound As(V) to eluate samples. Sediment samples were collected in sterile 50 mL polypropylene tubes in duplicate by using sterile spoons and stored in ice. All samples for microbial community analysis (sediments and biomass-containing membranes) were stored in dry ice in the field and during transport, and then stored at -80°C in the laboratory until further analyses. Laboratory geochemical analysis {#sec005} ------------------------------- The cation and anion concentrations were measured by inductively coupled plasma-optical emission spectrometry (CAP6300, Thermo, USA) and ion chromatography (ICS1100, Dionex, USA), respectively. As~Tot~ and Fe~Tot~ in the sediments were extracted by 1:1 aqua regia digestion method in a water bath \[[@pone.0146331.ref027]\]. Pre-separated As(III) and As(V) from hot spring waters and extracted As~Tot~ from sediments were determined using liquid chromatography-hydride generation -atomic fluorescence spectrometry (LC-HG-AFS, Haiguang AFS-9780, Beijing) according to Jiang et al. \[[@pone.0146331.ref025]\]. Extracted Fe~Tot~ from sediments was determined by the 1,10-Phenanthroline-based assay: 10 mL extracted solutions was mixed with 5 mL acetate-sodium acetate buffer (pH = 4.6), 2.5 mL 1% hydroxylamine hydrochloride solution and 5 mL 0.1% 1,10-phenanthroline solution in 50 mL volumetric flask. The mixtures were made up to a volume of 50 mL with deionized water and allowed to stand for 10 min. The absorbance of each solution at 510 nm was measured with a spectrophotometer (UV1750, Shimadzu, Japan). Dissolved organic carton (DOC) of water samples and total organic carbon (TOC) of sediment samples were determined using a TOC analyzer (TOC-V~CPH~, Shimadzu, Japan) and a Macro elemental analyzer (Multi EA 4000, Analytik Jena, Germany), respectively. DNA extraction, amplification and sequencing {#sec006} -------------------------------------------- DNA was extracted from biomass-containing filters or from 0.5 g sediment samples using the FastDNA SPIN Kit for Soil (MP Biomedical, OH, USA). DNA concentrations were measured by Pico Green using a FLUOstar OPTIMA fluorescence plate reader (BMG LABTECH, Jena, Germany). The V4 region of 16S rRNA gene was amplified using the normal primer pair 515F (`5’-GTGCCAGCMGCCGCGGTAA-3’`) and 806R (`5’-GGACTACHVGGGTWTCTAAT-3’`) combined with Illumina adapter sequences, a pad and a linker of two bases, as well as barcodes on the reverse primers \[[@pone.0146331.ref028]\]. PCR amplification was carried out in a 25 μL reaction containing 2.5 μL 10× PCR buffer II (including dNTPs) (Invitrogen, Grand Island, NY), 0.4 μM of both forward and reverse primers, 10--15 ng DNA and 0.25 U high fidelity AccuPrime^™^ Taq DNA polymerase (Life Technologies) under the following program: initial denaturation at 94°C for 1 min, followed by 30 cycles of 94°C for 20 s, 53°C for 25 s, and 68°C for 45 s, and then a final extension at 68°C for 10 min. Reactions were performed in triplicate. Amplicons of each sample were combined and confirmed positive PCR products by agarose gel electrophoresis, and then quantified with PicoGreen. Finally, a total of 200 ng PCR product of each sample was pooled together and purified through QIAquick Gel Extraction Kit (Qiagen, Valencia, CA) and then was re-quantified with PicoGreen. Sample 16S rRNA clone libraries for sequencing were prepared according to the MiSeqTM Reagent Kit Preparation Guide (Illumina, San Diego, CA, USA) and the protocol described previously \[[@pone.0146331.ref029]\]. Briefly, sample denaturation was performed by mixing 10 μL of combined PCR products (2 nM) and 10 μL 0.2 M NaOH and incubated for 8 min at room temperature. Denatured DNA was diluted to 15 pM using HT1 buffer and mixed with a PhiX DNA library (final concentration 14.3%). A total of 600 μL sample mixture, together with customized sequencing primers for forward, reverse, and index reads, were loaded into the corresponding wells on the reagent cartridge of a 500-cycle v2 MiSeq kit. Sequencing was performed for 251, 12, and 251 cycles, respectively for forward, index, and reverse reads on a Illumina MiSeq system (Illumina, San Diego, CA). Sequence data preprocessing and statistical analysis {#sec007} ---------------------------------------------------- Raw sequences with perfect matches to barcodes were split to sample 16S rRNA clone libraries and were trimmed using Btrim with threshold of QC higher than 25 over 5 bp window size and the minimum length of 150 bp \[[@pone.0146331.ref030]\]. Forward and reverse reads with at least 50 bp overlap and lower than 5% mismatches were joined using Fast Length Adjustment of SHort reads (FLASH) \[[@pone.0146331.ref031]\]. After trimming of ambiguous bases (i.e. N), joined sequences with lengths between 247 and 258 bp were subjected to chimera removal by Uchime \[[@pone.0146331.ref032]\]. Operational taxonomic units (OTUs) clustering was through Uclust at 97% similarity level \[[@pone.0146331.ref033]\], and taxonomic assignment was through Ribosomal Database Project (RDP) classifier \[[@pone.0146331.ref034]\] with a minimal 50% confidence estimate. The above steps were performed through the Galaxy pipeline of Institute for Environmental Genomics in University of Oklahoma (<http://zhoulab5.rccc.ou.edu/>). Samples were rarefied at 12 955 sequences per sample. Singletons of generated OTU table were removed for downstream analyses. All statistical analysis in this study was performed based on genus-level OTUs at the 97% similarity level under the Vegan package in R (<http://www.r-project.org/>), unless otherwise stated. A variety of alpha diversity indices were calculated including Chao1, Shannon and Equitability. Hierarchical cluster tree using unweighted pair group method with arithmetic means (UPGMA), principal coordinates analysis (PCoA) and non-metric dimensional scaling (NMDS) ordination were built to depict the community composition structure based on the Bray-Curtis dissimilarity matrix of detected OTUs. The Envfit function in the package of Vegan was used to overlay the significant environmental variables on the NMDS ordination. Analyses of similarity (ANOSIM), non-parametric multivariate ANOVA (ADONIS), multi-response permutation procedure (MRPP) and Mantel were performed to test for significant differences in microbial community compositions between sample types (i.e., water vs. sediment) and different locations (i.e., pool vs. downstream). The DNA sequences were deposited to the Short Read Archive database at NCBI (Accession number: SRP056673). Results {#sec008} ======= Water and sediment geochemistry {#sec009} ------------------------------- Three water samples (-2 m, -1 m and 0 m) located in the pool of Zhenzhuquan showed similar physical and chemical conditions, with a slight fluctuation in temperature (84.3--91.3°C), pH (3.58--4.33), DO (0.23--0.33 mg/L), DOC (0.36--0.52 mg/L) and concentrations of various ions ([Table 1](#pone.0146331.t001){ref-type="table"} and [S1 Fig](#pone.0146331.s006){ref-type="supplementary-material"}), which suggested that Zhenzhuquan was derived by vigorous degassing from the same recharged source. The pool with a low DO average of 0.28 mg/L had correspondingly low ratios of Fe(III) and Fe~Tot~ (Fe(III)/Fe~Tot~: \~ 0.15), but contained high ratios of As(V) and As~Tot~ (As(V)/As~Tot~: \~ 0.81). Once the source water was discharged from the pool, most of physical-chemical parameters dramatically changed ([Fig 2](#pone.0146331.g002){ref-type="fig"}). Temperature ranged from 84.3°C to 49.3°C. The pH values showed a slight decline (pH = 3.89 at 0 m and pH = 3.69 at 12 m), which was consistent with acidic Succession Spring in Yellowstone National Park (YNP) \[[@pone.0146331.ref009]\]. The decrease in pH was primarily caused by evaporation, suggested by slight increase of F and Cl concentrations along the drainage ([S1 Fig](#pone.0146331.s006){ref-type="supplementary-material"}). Across sampling intervals, DO, sulfate and ratios of Fe(III) to Fe~Tot~ (Fe(III)/Fe~Tot~) significantly increased from 0.29 to 4.74 mg/L, 120.05 to 158.21 mg/L and 0.09 to 0.34, respectively. Dissolved Fe~Tot~ and As~Tot~ concentrations in water samples gradually declined from 0.64 to 0.45 mg/L and 66.83 to 44.83 μg/L respectively, coupled with significantly elevated concentrations of Fe~Tot~ (0.05--4.55 g/kg), As~Tot~ (10.53--16.44 g/kg) and TOC (1.37%-2.81%) in the downstream sediments. 10.1371/journal.pone.0146331.t001 ###### Geochemistry of water and sediment samples along Zhenzhuquan\'s outflow channel in Tengchong geothermal area. {#pone.0146331.t001g} Distance from discharge (m) Aqueous phase Solid phase ----------------------------- --------------- ------------- ------ ------ ------ ------ ------ -------- ------ ------ ------ ------ ------- ------- ------- ------ ------ ------- ------ -2 89.9 3.58 0.33 0.52 3.00 bdl 0.03 112.66 0.45 0.07 0.51 0.13 17.29 46.55 63.84 0.73 0.05 0.50 0.27 -1 91.3 4.33 0.23 0.36 3.30 bdl 0.03 114.37 0.40 0.11 0.51 0.22 9.27 52.99 62.26 0.85 0.07 0.36 0.09 0 84.3 3.89 0.29 0.52 3.10 bdl 0.05 120.05 0.58 0.06 0.64 0.09 9.20 57.63 66.83 0.86 0.09 0.28 0.57 3 76.2 3.64 0.30 1.00 3.00 0.24 0.02 129.20 0.47 0.06 0.53 0.10 8.10 49.67 57.77 0.86 4.55 16.44 1.37 6 56.8 3.69 2.38 0.60 3.40 bdl 0.03 150.95 0.36 0.10 0.45 0.21 7.33 38.53 45.86 0.84 1.20 10.53 2.72 9 49.3 3.69 4.74 0.47 3.40 0.38 0.01 158.21 0.31 0.16 0.47 0.34 7.06 33.78 40.83 0.83 1.29 11.64 2.81 bdl:below detection limit (1 μg/L) {#pone.0146331.g002} Alpha diversity of microbial communities {#sec010} ---------------------------------------- A total of 304 743 passing sequences were obtained from six-pair parallel water and sediment samples. After rarefaction at 12 955 sequences per sample, OTU clustering at 97% similarity level and removal of singletons, 155 050 sequences remained. A variety of taxa were present, with 123--834 observed and 172--1173 predicted OTUs (based on Chao1) and coverage values ranging from 47.04% to 83.08% ([S1 Table](#pone.0146331.s001){ref-type="supplementary-material"}). Along the outlet, microbial community richness, Shannon diversity and equitability from water samples significantly increased, whereas those indices of sediment samples had a decline with remarkable minimums at 3 m ([Fig 3](#pone.0146331.g003){ref-type="fig"}). Additionally, in the pool of Zhenzhuquan, richness, diversity and equitability in sediment samples were much higher than those in water samples. However, these indices were distinctly lower in sediment samples than those of water samples downstream. These alpha diversity differences between water and sediment samples can also be seen in their correlation with physical-chemical parameters: Shannon diversity and equitability of water samples were significantly correlated with DO, sulfate, Fe(III)/Fe~Tot~, Fe(II), dissolved As~Tot~, temperature and TOC, while these indices of sediment samples were significantly correlated with DOC, solid Fe~Tot~ and As~Tot~ in the sediments ([Table 2](#pone.0146331.t002){ref-type="table"}). {#pone.0146331.g003} 10.1371/journal.pone.0146331.t002 ###### Correlation between alpha diversity indices at 97% similarity OTU level and environment factors. {#pone.0146331.t002g} Environment factors Water samples Sediment samples --------------------- --------------- ----------------------------------------------- ------------------------------------------------- ------------------------------------------------- --------------------------------------------- --------------------------------------------- ----------------------------------------------- Aqueous phase DO +0.492 +0.978[\*\*\*](#t002fn003){ref-type="table-fn"} +0.991[\*\*\*](#t002fn003){ref-type="table-fn"} -0.284 -0.201 -0.134 Sulfate +0.511 +0.911[\*](#t002fn001){ref-type="table-fn"} +0.935[\*\*](#t002fn002){ref-type="table-fn"} -0.456 -0.445 -0.404 Fe(III)/Fe~Tot~ +0.742 +0.896[\*](#t002fn001){ref-type="table-fn"} +0.829[\*](#t002fn001){ref-type="table-fn"} +0.181 +0.127 +0.105 Fe(II) -0.756 -0.895[\*](#t002fn001){ref-type="table-fn"} -0.829[\*](#t002fn001){ref-type="table-fn"} -0.144 -0.045 +0.023 As~Tot~ -0.655 -0.952[\*\*](#t002fn002){ref-type="table-fn"} -0.938[\*\*](#t002fn002){ref-type="table-fn"} +0.233 +0.277 +0.287 T -0.469 -0.938[\*\*](#t002fn002){ref-type="table-fn"} -0.939[\*\*](#t002fn002){ref-type="table-fn"} +0.486 +0.460 +0.411 DOC -0.448 -0.221 -0.122 -0.717 -0.857[\*](#t002fn001){ref-type="table-fn"} -0.896[\*](#t002fn001){ref-type="table-fn"} Solid phase Fe~Tot~ -0.218 +0.031 +0.107 -0.679 -0.895[\*](#t002fn001){ref-type="table-fn"} -0.964[\*\*](#t002fn002){ref-type="table-fn"} As~Tot~ +0.107 +0.480 +0.544 -0.694 -0.856[\*](#t002fn001){ref-type="table-fn"} -0.897[\*](#t002fn001){ref-type="table-fn"} TOC +0.461 +0.863[\*](#t002fn001){ref-type="table-fn"} +0.896[\*](#t002fn001){ref-type="table-fn"} -0.508 -0.504 -0.496 \*p\<0.05, \*\*p\<0.01, \*\*\*p\<0.001 Microbial community compositions {#sec011} -------------------------------- Microbial community compositions were distinctly different between water and sediment samples ([Fig 4](#pone.0146331.g004){ref-type="fig"}). In the pool, water sample 16S rRNA clone libraries were dominated by *Crenarchaeota* (48.90 to 93.92%) with the rest mainly composed of *Actinobacteria* (1.42--36.14%) and *Proteobacteria* (4.12--14.57%), while sediment sample 16S rRNA clone libraries at -2 m and -1 m were predominated by *Proteobacteria* (74.36--83.31%). Along the outlet from 0 m to 9 m, *Crenarchaeota* dominated in water samples gradually declined from 93.92% to 34.79%, which was coupled with slight increase of *Acidobacteria* (0.07--6.06%), *Actinobacteria* (1.42--2.80%), *Armatimonadetes* (0.01--5.13%), *Bacteroidetes* (0.07--8.18%), *Firmicutes* (0.22--4.36%) and *Proteobacteria* (4.12--23.10%). However, initial dominant *Crenarchaeota* in sediment samples was generally substituted by *Chloroflexi* along the outflow channel, ranging from 0.13% to 37.76%. There were some unclassified phyla in downstream samples 16S rRNA clone libraries, especially the sediment at 9 m ([Fig 4](#pone.0146331.g004){ref-type="fig"}). Generally, *Acidobacteria*, *Armatimonadetes*, *Bacteroidetes* and *Chloroflexi* positively correlated with DO, ammonia, sulfate, Fe(III)/Fe~Tot~, dissolved As~Tot~ and TOC, and negatively correlated with temperature, aqueous Fe(II), solid Fe~Tot~ and As~Tot~ ([S2 Table](#pone.0146331.s002){ref-type="supplementary-material"}). {#pone.0146331.g004} At the genus level, water sample 16S rRNA clone libraries in the pool were composed of the *Crenarchaeota Sulfolobus* (41.38--85.95%), the *Actinobacteria Nocardia* (1.17--34.90%) and the *Proteobacteria Ralstonia* (1.66--4.66%), *Delftia* (0.51--2.76%) and *Acinetobacter* (0.58--2.57%) ([Fig 5](#pone.0146331.g005){ref-type="fig"}, [S3 Table](#pone.0146331.s003){ref-type="supplementary-material"}). In contrast, sediment sample 16S rRNA clone libraries in the pool (-2 m and -1 m) dominated by *Proteobacteria* correspondingly were harbored by more *Ralstonia* (14.96--19.76%), *Delftia* (20.13--23.29%), *Acinetobacter* (4.10--6.99%), *Undibacterium* (7.84--9.29%) and *Pseudomonas* (2.89--2.92%), and less *Sulfolobus* (0.30--8.89%). Along the outflow channel, *Sulfolobus* in water samples gradually decreased from 85.95% to 24.94%, coupled with elevated Gp3 (0.02--4.22%) of *Acidobacteria*, *Chthonomonas* (0--5.11%) of *Armatimonadetes*, *Alicyclobacillus* (0.02--1.93%) of *Firmicutes* and *Acidicaldus* (0--10.48%) of *Proteobacteria*. However, in downstream sediment samples, 73.25--96.21% sequences did not belong to any known genus except for minor Gp3 (0.01--5.05%), *Chthonomonas* (0.03--10.77%) and *Acidicaldus* (0.01--4.86%). Most of unknown-genus sequences were from class *Thermoprotei* (86.48--97.18%) of *Crenarchaeota* at 0 m and 3 m and class *Ktedonobacteria* (31.73--37.34%) of *Chloroflexi* at 6 m and 9 m ([S2 Fig](#pone.0146331.s007){ref-type="supplementary-material"}). Moreover, at the class level, *Thermoprotei* was predominant in all water samples and sediment samples at 0 m and 3 m, but replaced by *Ktedonobacteria* in sediment samples at 6 m and 9 m ([S3 Fig](#pone.0146331.s008){ref-type="supplementary-material"}). *Acidicaldus*, *Gp3*, *Acidisoma* and *Chthonomonas* positively correlated with DO, ammonia, sulfate, Fe(III)/Fe~Tot~, dissolved As~Tot~ and TOC, and negatively correlated with temperature, aqueous Fe(II) and As~Tot~ ([S4 Table](#pone.0146331.s004){ref-type="supplementary-material"}). {ref-type="supplementary-material"}.](pone.0146331.g005){#pone.0146331.g005} Microbial community structure statistics {#sec012} ---------------------------------------- Based on Bray-Curtis dissimilarity at the 97% similarity OTU level, an UPGMA cluster tree of the microbial community populations showed that sediment samples were divided into two groups (pool and downstream), and distinctly separated from the water samples ([Fig 6A](#pone.0146331.g006){ref-type="fig"}). A similar result was also revealed by PCoA analysis with explained 68.5% of the observed variation ([Fig 6B](#pone.0146331.g006){ref-type="fig"}). Four complimentary non-parametric multivariate statistical tests including ADONIS, ANOSIM, MRPP and Mantel further confirmed the significant differences of microbial communities between not only pool and downstream samples, but also water and sediment samples ([S5 Table](#pone.0146331.s005){ref-type="supplementary-material"}). Results of Envfit function indicated that six geochemical parameters were significantly correlated (P\<0.05) with microbial community structure along the outlet, including temperature, DO, sulfate, aqueous As~Tot~, solid TOC and As~Tot~ with R^2^ values of 0.63, 0.53, 0.62, 0.60, 0.65 and 0.57, respectively ([Fig 6C](#pone.0146331.g006){ref-type="fig"}). The similar directions of temperature and aqueous As~Tot~, and the opposite directions of DO, sulfate and TOC indicated correlations among these variables and did not necessarily suggest that all environmental factors were responsible driving forces of community structure. Significant environmental factors shaping microbial communities were selected according to general physiological niche of microbial populations. {#pone.0146331.g006} Discussion {#sec013} ========== Potential As, S and Fe oxidation processes {#sec014} ------------------------------------------ In the Zhenzhuquan pool, the remarkably high As(V)/As~Tot~ ratio (0.73--0.86) suggested that As(III) oxidation occurred at the discharge source, which were distinctly different from previous studies that showed As(III) was predominate in the source water of acid sulfate-Cl hot springs, such as Dragon Spring (As(V)/As~Tot~ = 0.05) \[[@pone.0146331.ref011]\], Beowulf Spring (As(V)/As~Tot~ = 0.04) \[[@pone.0146331.ref010]\] and Succession Spring (As(V)/As~Tot~ = 0.04) \[[@pone.0146331.ref009]\] in YNP and Champagne Pool in New Zealand (As(V)/As~Tot~ = 0.04) \[[@pone.0146331.ref001]\]. The As oxidation in the pool might be mediated microbially. Previous studies demonstrated that sulfide was a potent inhibitor of microbially-mediated As(III) oxidation in acid systems by inactivating expressed As(III) oxidase enzyme (Aio) \[[@pone.0146331.ref002], [@pone.0146331.ref004], [@pone.0146331.ref013]\]. Consequently, different from acid sulfate-Cl hot springs with sulfide concentrations of 2.02--12.6 mg/L in those previous studies, the distinctly low concentrations of sulfide (0.03--0.05 mg/L) in the Zhenzhuquan pool allowed for microbial As(III) oxidation. Our previous study on *aioA* genes had demonstrated the presence of several groups of As(III)-oxidizing microorganisms in the pool, including a few unidentified families of *Aquificae* and some postulated archaea \[[@pone.0146331.ref025]\]. And in this study, based on 16S rRNA sequences, some microbial populations in sample 16S rRNA clone libraries were also highly similar with bacteria *Pseudomonas* and *Ralstonia* which were found to be capable of As oxidation in geothermal and other environments \[[@pone.0146331.ref035], [@pone.0146331.ref036]\]. Except for degassing in part, the lack of sulfide (0.03--0.05 mg/L) in the pool was primarily due to microbial sulfide oxidation, suggested by high sulfate concentrations (112.66--120.05 mg/L) and the presence of abundant *Sulfolobus* (41.38--85.95% in water sample libraries), a putative thermoacidophilic surfur-oxidizing archaeon ([Table 1](#pone.0146331.t001){ref-type="table"} and [Fig 5](#pone.0146331.g005){ref-type="fig"}) \[[@pone.0146331.ref022], [@pone.0146331.ref037], [@pone.0146331.ref038]\]. Concurrent with sulfide oxidation, Fe oxidation happened in the outflow channel, indicated by the decline of Fe(II) concentrations and increase of Fe(III)/Fe~Tot~ ([Fig 2](#pone.0146331.g002){ref-type="fig"}). Previous studies displayed that Fe(II) oxidation in acid hot springs was mediated by microorganisms, such as *Metallosphaera* str. MK1, *Sulfobacillus* str. MK2, *Sulfolobus* str. MK3, *Sulfolobales* str. MK5 and *Acidicaldu*s str. MK6 \[[@pone.0146331.ref009], [@pone.0146331.ref014]\]. In this study, some microbial populations belonging to *Thermoprotei* (such as *Sulfolobus*, and *Metallosphaera*) and *Acidicaldus* dominanated in downstream sample 16S rRNA clone libraries and were probably responsible for Fe(II) oxidation in the outflow channel ([Fig 2](#pone.0146331.g002){ref-type="fig"}) \[[@pone.0146331.ref014]--[@pone.0146331.ref016]\]. Due to lack of sulfide, yellow crystalline elemental S deposit presented in those acid sulfate-Cl hot springs did not appear downstream in this study \[[@pone.0146331.ref001], [@pone.0146331.ref009], [@pone.0146331.ref011]\]. Coupled Fe and As concentrations decline in water samples and increase in sediments suggested the co-deposition of As and Fe ([Fig 2](#pone.0146331.g002){ref-type="fig"}). However, extremely high As concentrations (up to 16.44 g/kg) and As/Fe mole ratios (2.70--6.72) in the sediments was significantly different from previous results from the acid sulfate-Cl hot springs (As/Fe mole ratios: 0.60 to 0.74) \[[@pone.0146331.ref009]--[@pone.0146331.ref011], [@pone.0146331.ref039]--[@pone.0146331.ref041]\]. Previous studies documented the clay minerals in geothermal areas, such as smectite and kaolinite, could host As concentrations up to 4 g/kg \[[@pone.0146331.ref008], [@pone.0146331.ref042]--[@pone.0146331.ref044]\]. Coincidentally, smectite and kaolinite were also detected in the downstream sediments of Zhenzhuquan by X-ray diffraction, which suggested that As might be also adsorbed on the clay mineral. Further investigation on mineralogy of the As-rich sediments downstream is warranted. Significant environmental factors shaping the microbial community structure {#sec015} --------------------------------------------------------------------------- Generally, water and sediment sample 16S rRNA clone libraries along the outflow channel of Zhenzhuquan were dominated by *Thermoprotei* (mostly comprised of the genus *Sulfolobus*) ([Fig 4](#pone.0146331.g004){ref-type="fig"} and [S2 Fig](#pone.0146331.s007){ref-type="supplementary-material"}), which was significantly different with sediments or waters from acid sulfate-Cl hot springs, mainly colonized by *Hydrogenobacter*, *Hydrogenobaculum* and *Sulfurihydrogenibium* of *Aquificae* \[[@pone.0146331.ref001], [@pone.0146331.ref009], [@pone.0146331.ref012]\]. This difference was probably derived by their distinct temperature and pH. *Hydrogenobacter* and *Sulfurihydrogenibium* favor circumneutral pH and *Hydrogenobaculum* has an optimal growth temperature of 60--70°C \[[@pone.0146331.ref045], [@pone.0146331.ref046]\], and microaerophilic acidophilic *Sulfolobus* prefer higher temperature of 65--85°C and more acidic pH of 2--3. These are all conditions similar to the Zhenzhuquan pool above at the 3 m site (DO: 0.23--0.33 mg/L; temperature: 76.2--91.3°C; pH: 3.58--4.33) ([Table 1](#pone.0146331.t001){ref-type="table"}) \[[@pone.0146331.ref022]\]. This result was also consistent with distribution of *Sulfolobus* in acidic solfataras and hot springs in YNP \[[@pone.0146331.ref047], [@pone.0146331.ref048]\]. As revealed by non-parametric multivariate statistical analysis, there were significant differences in microbial community structures between not only pool and downstream samples, but also water and sediment samples along the outflow channel ([S5 Table](#pone.0146331.s005){ref-type="supplementary-material"}). Temperature, DO and TOC significantly shaped microbial community structure of upstream and downstream samples ([Fig 6](#pone.0146331.g006){ref-type="fig"}). Responding to the relatively low DO (0.23--0.30 mg/L) and high temperature (76.2--91.3°C) in upstream ([Table 1](#pone.0146331.t001){ref-type="table"}), most microbial populations in 16S rRNA clone libraries were microaerophilic/anaerobic, thermophiles and hyperthermophiles, such as the dominant class *Thermoprotei* (48.90--97.18%, mainly comprised of *Sulfolobus* in water samples) \[[@pone.0146331.ref049]\], *Nocardia* (34.90%) in the water sample at -1 m \[[@pone.0146331.ref050]\], *Fervidicoccus* (4.46%) in sediment sample at 0 m \[[@pone.0146331.ref051]\], *Delftia* (20.13--23.29%) and *Ralstonia* (14.96--19.76%) in sediment samples at -1 m and -2 m \[[@pone.0146331.ref052]\] ([Fig 5](#pone.0146331.g005){ref-type="fig"} and [S2 Fig](#pone.0146331.s007){ref-type="supplementary-material"}). With a decline in temperature (56.8--49.3°C) and increased DO (2.38--4.74 mg/L) as well as TOC (2.72--2.81%), more aerobic heterotrophic mesophiles and thermophiles correspondingly inhabited downstream samples libraries, such as *Ktedonobacteria* (31.72--37.34% in sediment samples) \[[@pone.0146331.ref053]\], *Acidobacteria* Gp3 (0.91--5.05%) \[[@pone.0146331.ref054]\], *Acidicaldu*s (1.98--10.48%) \[[@pone.0146331.ref055]\], *Chthonomonas* (2.12--10.16%) \[[@pone.0146331.ref056]\] and *Sphingobacteria* (4.94--10.40%) \[[@pone.0146331.ref057]\]. A significant positive correlation between sulfate and TOC (r = 0.989, p\<0.001) along the outlet ([Fig 2](#pone.0146331.g002){ref-type="fig"}) implied that sulfide oxidation might be coupled to carbon fixation by facultatively chemoautotrophic archaea, such as above *Sulfolobus* and unclassified genera within the *Thermoprotei*, leading to TOC accumulation in downstream sediments \[[@pone.0146331.ref023]\]. Previous studies demonstrated that microbial community diversity increased with temperature decline and TOC or DOC increase in geothermal environments \[[@pone.0146331.ref021], [@pone.0146331.ref022]\]. However, in this study, the remarkably high As concentrations (up to 16.4 g/kg) and abnormal low alpha diversity in the downstream sediments (much lower than those of water at the same site) suggested that solid As~Tot~ in the sediments probably played a significant role in shaping microbial community structure ([Fig 3](#pone.0146331.g003){ref-type="fig"}) \[[@pone.0146331.ref058], [@pone.0146331.ref059]\]. Otherwise, it should be noted that even at the cutoff of 0.07, most sequences from the sediments 16S rRNA clone libraries (72.41--95.91%) could still not be assigned to a known genus, which suggested that some novel species possibly inhabit the high As sediments downstream ([Fig 5](#pone.0146331.g005){ref-type="fig"} and [S2 Fig](#pone.0146331.s007){ref-type="supplementary-material"}). Conclusions {#sec016} =========== Arsenic oxidation mainly occurred in acid sulfate Zhenzhuquan pool with low chloride. Coupled with iron and sulfur oxidation along the outflow channel, arsenic was substantially accumulated in downstream sediments and appeared to significantly constrain their microbial community diversity. Temperature, total organic carbon and dissolved oxygen significantly shaped the different microbial communities between upstream and downstream samples of Zhenzhuquan. Some putative functional microbial populations were possibly related to arsenic oxidation (*Aquificae* and *Pseudomonas*), sulfur oxidation (*Sulfolobus*) and iron oxidation (*Sulfolobus*, *Metallosphaera* and *Acidicaldus*). A total of 72.41--95.91% unassigned-genus sequences in downstream high arsenic sediment 16S rRNA clone libraries probably implied the presence of some novel genera. Supporting Information {#sec017} ====================== ###### Distribution of alpha diversity indices at the 97% similarity OTU level by re-sampling 12 955 reads in each sample. (DOC) ###### Click here for additional data file. ###### Correlation between phylum at 97% similarity OTU level and environment factors. Only phylum significantly correlated with environment factors were displayed. (DOC) ###### Click here for additional data file. ###### The relative abundances of all genera in different samples. The unit is %. (DOC) ###### Click here for additional data file. ###### Correlation between genera at 97% similarity OTU level and environment factors. Only genera significantly correlated with environment factors were displayed. (DOC) ###### Click here for additional data file. ###### Significance tests of microbial community structures between different groups with four different statistical approaches. (DOC) ###### Click here for additional data file. ###### Normalized main cations and anions concentrations (divided by the sum) variation along the outflow channel of Zhenzhuquan. The values in parenthesis were averaged concentrations of ions with a unit of mg/L. (TIF) ###### Click here for additional data file. ###### Distribution of microbial community compositions from water sediment samples at the class level. The ratios which exceeded 0.5% were displayed in this figure. (TIF) ###### Click here for additional data file. ###### Change of main microbial community compositions at class level from water samples (A) and sediment samples (B) along the outflow channel of Zhenzhuquan. (TIF) ###### Click here for additional data file. We are grateful to the staff from the Yunnan Tengchong Volcano and Spa Tourist Attraction Development Corporation for their assistance. [^1]: **Competing Interests:**The authors have declared that no competing interests exist. [^2]: Conceived and designed the experiments: ZJ PL YXW. Performed the experiments: ZJ PL DJ. Analyzed the data: ZJ PL. Contributed reagents/materials/analysis tools: ZJ PL DJ XD RZ YHW. Wrote the paper: ZJ PL YXW.

INTRODUCTION ============ The ribonucleoprotein enzyme ribonuclease P (RNase P) throughout phylogeny removes 5′ leader sequences of precursor tRNAs (ptRNAs) to generate mature tRNA 5′ ends. The enzyme harbors a single RNA subunit (P RNA) of common ancestry, but only bacterial P RNAs are efficient catalysts *in vitro* in the absence of the single bacterial protein cofactor ([@B1],[@B2]). Very low RNA-alone activity of archaeal, eukaryal and organellar P RNAs ([@B3; @B4; @B5]) correlates with the presence of multiple protein subunits (usually 4 in Archaea and 9--10 in Eukarya). Bacterial P RNAs are composed of two independent folding domains, the catalytic (C-) domain and the specificity (S-) domain (6,7; [Figure 1](#F1){ref-type="fig"}), the latter playing a crucial role in substrate binding via its interaction with the T stem-loop region of ptRNAs (for review, see [@B8]). The single bacterial RNase P protein (P protein), dispensable *in vitro* but essential *in vivo* ([@B9],[@B10]), binds to the P2--J2/3--J3/4--P4--J18/2 region in the C-domain of P RNA ([@B11; @B12; @B13; @B14]). The protein increases substrate affinity through interaction with the 5′-leader ([@B14],[@B15]), which entails tighter binding of key metal ions involved in substrate positioning and catalysis ([@B16]). Figure 1.Secondary structure presentation according to Massire *et al.* ([@B20]) of *M. thermoautotrophicus* ΔH RNase P RNA (MM), *E. coli* P RNA (EE), and chimeras with the C-domain from *M. thermoautotrophicus* and the S-domain from *E. coli* (ME) and vice versa (EM). The 13 nucleotide exchanges and the single 1-nt insertion in P RNA of strain *M. thermoautotrophicus* strain Marburg (MM-Marburg in the tables) relative to *M. thermoautotrophicus* ΔH are indicated along the sequence in the secondary structure of MM RNA. The C-domain is depicted in green and the S-domain is depicted in blue. Grey nucleotides at the 5′- and 3′-ends indicate non-natural additions present in the T7 *in vitro* transcripts. Activity of archaeal P RNA in the absence of the archaeal RNase P protein cofactors Pop5, Rpp21, Rpp29 and Rpp30 was observed solely for one of the two types of archaeal P RNAs, type A, but not type M ([@B3]). Archaeal type A RNAs are structurally related to bacterial P RNAs, but lack the P13/14 stem and helix-P18 among several other deviations from the bacterial consensus (17; [Figure 1](#F1){ref-type="fig"}). Type A archaeal P RNA-alone activity was observed to be low and limited to conditions of very high ionic strength for optimal activity, such as 3 M NH~4~^+^ and 0.3 M Mg^2+^ in the case of P RNA from the euryarchaeon *Methanothermobacter thermoautotrophicus* ([@B3]). Addition of the P protein from *Bacillus subtilis* activated the archaeal P RNA to some extent at lower ionic strength ([@B3]). For bacterial type A RNA, three loop-helix interdomain contacts (L8--P4, L9--P1 and L18--P8) have been identified ([@B18; @B19; @B20; @B21]) and are thought to orient the C- and S-domains toward each other ([Figure 1](#F1){ref-type="fig"}). Only one of these contacts, L8--P4, can form in the *M*. *thermoautotrophicus* P RNA structure, since the P18 element is missing and the reported 5′- and 3′-ends of the RNA ([@B3]) only permit formation of a P1 helix too short to serve as a receptor site for L9 ([Figure 1](#F1){ref-type="fig"}). Previous attempts to increase the RNA-alone activity of archaeal or organellar P RNAs by insertion of structural elements or by *in vitro* selection have failed to identify variants with improved RNA-alone activity relative to the corresponding wild-type RNA ([@B5],[@B17],[@B22]). We have reinvestigated this issue in a more systematic manner by constructing a series of archaeal C-domain variants with confined alterations towards the bacterial P RNA consensus. The aim of this approach was to identify specific deficits of archaeal P RNAs that limit their RNA-alone activity and make them depend so much more on protein cofactors than their bacterial homologs. Combining the archaeal C-domain from *M*. *thermoautotrophicus* P RNA with the S-domain of *Escherichia coli* P RNA we were able to increase activity---up to three orders of magnitude---by simultaneously introducing a few minor alterations into the archaeal C-domain. All variants were screened at two Mg^2+^ concentrations, 100 mM in the RNA-alone reaction and 4.5 mM in the holoenzyme reaction. Observed activity gains may well include contributions from improved Mg^2+^ binding, which in turn could have affected folding, substrate binding, catalysis or binding of the RNase P protein. Thus, activity differences between tested P RNA variants might be mitigated to some extent at Mg^2+^ concentrations higher than those applied here (see Discussion section). We further provide evidence that the genuine 5′-and 3′-ends of P RNA from *Methanothermobacter* are longer than previously determined ([@B3]). This longer P1, which resulted in 100--200-fold higher RNA-alone activity, may exert its effect primarily through formation of the L9--P1 contact. Finally, analysis of ternary complex formation (P RNA, P protein, ptRNA) revealed that the archaeal S-domain has lost the capacity of bacterial S-domains to confer affine and productive interaction with the substrate, thereby masking the catalytic capacity of archaeal C-domains. MATERIAL AND METHODS ==================== Construction of plasmids for complementation and *in vitro* transcription ------------------------------------------------------------------------- Plasmids were constructed by standard cloning and PCR methods as described in ([@B5]). For specific details of plasmid construction and the oligonucleotides used, see [Supplementary Information](http://nar.oxfordjournals.org/cgi/content/full/gkn915/DC1). *In vitro* transcription and 5′ endlabelling of P RNAs and substrate ptRNA -------------------------------------------------------------------------- Run-off transcription with bacteriophage T7 RNA polymerase and 5′ and 3′ endlabelling with T4 polynucleotide kinase and T4 RNA ligase were performed essentially as described ([@B23],[@B24]). Templates used were pSBpt3′HH linearized with BamHI for the *Thermus thermophilus* ptRNA^Gly^ substrate ([@B23]), pDW98 linearized with BsaAI for *E. coli* P RNA ([@B24]) and pUC119_T7_M.th.\_rnpB linearized with BsaI for *M. thermoautotrophicus* wild-type P RNA ([@B3]). Mutants of the latter were transcribed from the pUC19 constructs described in [Supplementary Information](http://nar.oxfordjournals.org/cgi/content/full/gkn915/DC1); all pUC19 derivatives were linearized with EcoRI. Expression and purification of recombinant *E. coli* RNase P protein -------------------------------------------------------------------- *E. coli* RNase P protein (tagged with an N-terminal hexahistidyl peptide leader MRGSHHHHHHGS, encoded in plasmid pQE-30) was recombinantly expressed in *E. coli* JM109 and purifed by Ni-NTA affinity chromatography as described ([@B25]). RNase P processing assays ------------------------- The ptRNA^Gly^ substrate ([@B23]) was preincubated in assay buffer for 5 min at 55°C and 25 min at 37°C, P RNAs (separately) for 5 min at 55°C and 35 min at 37°C. For holoenzyme reactions, *E. coli* P protein was added to the P RNA 5 min before combining enzyme and substrate. Conditions of RNA-alone reactions were single turnover with trace amounts (\<1 nM) of 5′-^32^P-labelled substrate, 300 nM or 10 µM P RNA, 100 mM Mg(OAc)~2~, 100 mM NH~4~OAc, 50 mM MES and 2 mM EDTA, pH 6.0 and 37°C. Holoenzyme reactions were performed as multiple turnover reactions at concentrations of 10 nM P RNA, 50 nM P protein, 100 nM ptRNA^Gly^, and buffer KN (20 mM HEPES-KOH pH 7.4, 4.5 mM Mg\[OAc\]~2~, 150 mM NH~4~OAc, 2 mM spermidine, 0.05 mM spermine and 4 mM β-mercaptoethanol) at 37°C. Analysis of reactions by denaturing PAGE and evaluation of data was done as described ([@B23]); all quantifications were based on at least three independent experiments. Folding analysis by native PAGE ------------------------------- 3′-End-labelled P RNAs were preincubated at a concentration of 10 nM in buffer KN for 5 min at 55°C and then 50 min at 37°C. After addition of 1 volume of gel loading buffer (10% \[v/v\] glycerol, 4.5 mM MgCl~2~, xylene cyanol and bromophenol blue) at room temperature, the samples were immediately loaded onto non-denaturing polyacrylamide gels (11.25% \[v/v\] polyacrylamide, 66 mM HEPES, 33 mM Tris \[pH 7.4\], 0.1 mM EDTA, 100 mM NH~4~OAc and 4.5 mM MgCl~2~). Gels were run and the RNAs visualized as described ([@B25]). UV melting profiles ------------------- UV melting profiles were recorded at 260 nm on a CARY 100 Bio UV-Visible spectrophotometer, using the Cary Win UV software version 3.00. Bacterial strains ----------------- Total RNAs for RACE experiments were prepared from cells of *M. thermoautotrophicus* strain Marburg (DSM 2133, syn. *M. marburgensis*), whereas the *Methanothermobacter* RNase P RNA used throughout this study is that of *M. thermoautotrophicus* strain ΔH (DSM 1053). Determination of equilibrium dissociation constants --------------------------------------------------- Equilibrium dissociation constants (*K*~d~) of complex formation between RNase P holoenzymes and ptRNA substrate were determined by spin column assays essentially as described ([@B26],[@B27]) in a binding buffer containing 50 mM Tris-acetate, pH 7.1, 200 mM NH~4~OAc, 0.05% Nonidet P-40, and 15 mM Ca(OAc)~2~. Ca^2+^ was used instead of Mg^2+^ to prevent substrate cleavage during binding and the concentration of 15 mM was chosen to increase sensitivity of the assay, because the *K*~d~\'s measured at this Ca^2+^ concentration are substantially lower than those obtained at lower concentrations (e.g. 5 mM Ca^2+^). Before the assay, holoenzymes were reconstituted from P RNA variants and *E. coli* P protein by incubating 1 µM P RNA in binding buffer for 1 h at 37°C and then adding *E. coli* P protein to 5 µM, followed by another 15 min at 37°C. Holoenzymes were then diluted in binding buffer to desired concentrations, incubated for 5 min at 37°C, mixed with trace amounts (\<1 nM) of 5′-^32^P-labelled substrate (preincubated for 30 min at 37°C in binding buffer) and incubated at 37°C for 5 min. RESULTS ======= Analysis of *E. coli* and *M. themoautotrophicus* wild-type and chimeric P RNAs ------------------------------------------------------------------------------- Starting from *M. thermoautotrophicus* wild-type P RNA (henceforth called MM RNA) and *E. coli* wild-type P RNA (termed EE RNA; [Figure 1](#F1){ref-type="fig"}), we constructed P RNA chimeras by combining the S-domain of *E. coli* P RNA with the C-domain of *M. thermoautotrophicus* (termed ME RNA) or the *E. coli* C-domain with the *M. thermoautotrophicus* S-domain (termed EM RNA; [Figure 1](#F1){ref-type="fig"}). The chimeric P RNAs allowed us to address if the heterologous domains are able to cooperate and if there is a major asymmetry in the contribution of the archaeal C- and S-domains to the low RNA-alone activity of the archaeal P RNA. RNA-alone single turnover reactions (E\>\>S) at pH 6.0 in the presence of 300 nM P RNA, trace amounts (\<1 nM) ptRNA^Gly^ and 0.1 M each Mg^2+^ and NH~4~^+^ revealed only very weak cleavage by EM RNA (2 × 10^−4^ min^−1^) and no activity at all with ME and MM P RNA ([Table 1](#T1){ref-type="table"}). The activity of EM RNA increased linearly with concentration (tested from 300 nM to 10 µM, data not shown), indicating low substrate affinity. At 10 µM P RNA concentration, activities were detectable for all P RNA variants, with the chimeras EM and ME being 34--44-fold more active than the archaeal MM RNA ([Table 1](#T1){ref-type="table"}). The finding of activity at 10 µM but not at 0.3 µM for the MM and ME variants suggested that weak substrate affinity is also a major cause for the poor performance of these two P RNAs. The lower RNA-alone activity of MM versus EM and ME RNA ([Table 1](#T1){ref-type="table"}) showed that deficits of MM RNA are located in its S- as well as its C-domain. Table 1.ptRNA processing activity of *E. coli*/*M. thermoautotrophicus* P RNA domain swop chimeras in RNA-alone and holoenzyme assaysP RNA aloneHoloenzymeP RNA*k*~obs~ \[min^--1^ × 10^−4^\]P RNA*k*~obs~ \[min^--1^ × 10^−4^\]EE (0.3 μM)51 000 ± 4000EE (10 nM)82 000 ± 2000EM (0.3 μM)2 ± 1EMNot analyzedME (0.3 μM)Not detectableMENot analyzedMM (0.3 μM)Not detectableMMNot analyzedEM (10 μM)88 ± 19EM (10 nM)70 ± 10ME (10 μM)67 ± 17ME (10 nM)2 ± 1MM (10 μM)2 ± 0.1MM (10 nM)2 ± 1Ecat (10 μM)Not detectableEcat (10 nM)77 ± 6[^1] As in the RNA-alone reactions, RNase P holoenzymes reconstituted with the *E. coli* P protein and tested for multiple turnover under near-physiological salt conditions (buffer KN, 4.5 mM Mg^2+^, 10 nM P RNA, 50 nM P protein, 100 nM ptRNA^Gly^) yielded very low cleavage rates for MM RNA and the ME chimera ([Table 1](#T1){ref-type="table"}). The EM variant was 35-fold more active than MM and ME, though still a factor of \>1000 slower than EE RNA. Higher activity of EM RNA, harbouring the *E. coli* C-domain, was anticipated because the primary binding site of the bacterial P protein is in the C-domain ([@B11; @B12; @B13; @B14]). The fact that EM RNA activity equaled that of the isolated C-domain of *E. coli* P RNA (Ecat; [Table 1](#T1){ref-type="table"}) indicates that the archaeal S-domain made no functional contribution under the applied conditions of the holoenzyme assay. Not unexpected, alleles encoding MM, EM and ME RNAs were unable to rescue the mutant phenotype of a bacterial RNase P mutant strain (Table S3). ME RNA variants with specific structural alterations in the archaeal C-domain ----------------------------------------------------------------------------- With its substantial increase in RNA-alone activity compared to MM RNA ([Table 1](#T1){ref-type="table"}), we assumed ME RNA to be more sensitive toward activating changes in the archaeal C-domain. We thus generated an initial set of ME derivatives that were mutated towards the bacterial type A RNase P RNA consensus ([@B28]) by mutations ranging from single nucleotide exchanges to exchange of defined modules ([Figure 2](#F2){ref-type="fig"}). RNA-alone and holoenzyme assays were conducted as specified in [Table 1](#T1){ref-type="table"}, and activities of the ME variants illustrated in [Figure 2](#F2){ref-type="fig"} are summarized in [Table 2](#T2){ref-type="table"}. Table 2.Activity of ME variants in RNA-alone and holoenzyme assaysRNase P RNARNA-aloneHoloenzyme*k*~obs~ \[min^--1^ × 10^−4^\]Change*k*~obs~ \[min^--1^ × 10^−4^\]ChangeME67 ± 17    12.0 ± 1.01ME-mP10.4 ± 0.1↓ 176×1.1 ± 0.1↓ 2×ME-mP29.7 ± 1.2↓ 7×1.1 ± 0.2↓ 2×ME-mP42.8 ± 1.8↓ 24×1.2 ± 0.1↓ 2×ME-mP541 ± 5↓ 2×1.2 ± 0.1↓ 2×ME-mP17Not detectable1.1 ± 0.1↓ 2×ME-mP1839 ± 6↓ 2×4.0 ± 0.3↑ 2×ME-mJ15/18250 ± 45↑ 4×6.5 ± 0.3↑ 3×ME-mJ2/37.7 ± 1.2↓ 9×52 ± 6↑ 26×ME-mJ15/18/P10.5 ± 0.1↓ 143×1.3 ± 0.2↓ 2×ME-mJ15/18/P229 ± 7.2↓ 2×3.6 ± 0.3↑ 2×ME-mJ15/18/P4160 ± 15↑ 2×3.2 ± 0.3↑ 2×ME-mJ15/18/P5100 ± 25    1×3.2 ± 0.4↑ 2×ME-mJ15/18/P171.6 ± 1.2↓ 42×1.3 ± 0.1↓ 2×ME-mJ15/18/P183250 ± 400↑ 49×12 ± 0.2↑ 6×ME-mJ15/18/2/333 ± 3.5↓ 2×320 ± 30↑ 160×ME-mJ2/3/P21350 ± 240↑ 20×900 ± 200↑ 450×ME-mJ2/3/P2-2480 ± 120↑ 7×480 ± 50↑ 240×ME-mJ15/18/2/3/P25400 ± 910↑ 81×930 ± 20↑ 465×ME-mJ15/18/2/3/P2/nP19880 ± 2520↑ 147×3000 ± 500↑1500×[^2] Figure 2.Secondary structure presentation of the C-domain from *M. thermoautotrophicus* according to Massire *et al.* ([@B20]) with mutations introduced during this work. Regions subjected to mutation are circled and highlighted in grey; sequences of the respective mutants (indicated by the prefix 'm', e.g. mP1) are given in the adjacent boxes with the altered nucleotides coloured in red. Throughout the paper, combinations of mutations are indicated by a slash inbetween, e.g. ME-mJ2/3/P2 refers to mutation mJ2/3 as well as mutation mP2 within one construct, with the ME in front indicating that the construct possesses an *M. thermoautotrophicus* C-domain and an *E. coli* S-domain. Grey nucleotides at the 5′- and 3′-ends indicate non-natural additions present in the T7 *in vitro* transcripts. Nucleotide numbering is that of *M. thermoautotrophicus* ΔH P RNA. RNA-alone activity ------------------ Among the initial set of eight variants, only a single variant, with a point mutation in the region that joins P15 and P18 in bacterial P RNA, had increased RNA-alone activity relative to ME RNA (ME-mJ15/18, 4-fold increase; [Table 2](#T2){ref-type="table"}). The other variants displayed RNA-alone cleavage rates decreased from 2-fold to below the level of detection in the case of ME-mP17. The latter alteration introduced the P6/16/17 elements of *E. coli* P RNA, based on the idea that this may help to position the L15 loop for interaction with the tRNA 3′-CCA end. The next most prominent impairments were seen for ME-mP1 (176-fold), which introduced the *E. coli* P1 helix intended to restore the L9--P1 interdomain contact, and for ME-mP4 (24-fold) with three Watson-Crick base pair identities in P4 changed towards the bacterial consensus ([Figure 2](#F2){ref-type="fig"}, [Table 2](#T2){ref-type="table"}). Likewise, insertion of *E. coli* P18 (ME-mP18) to restore the L18--P8 interdomain contact failed to improve activity, in line with a previous study ([@B17]). In a second round, the variant with increased activity, ME-mJ15/18, was combined with each of the other structural alterations. RNA-alone activities of the double mutants either equalled that of ME RNA (within a factor of two) or remained substantially lower (ME-mJ15/18/P1, ME-mJ15/18/P17), with one notable exception: incorporation of *E. coli* P18 (ME-mJ15/18/P18) significantly increased activity compared to ME and also ME-mJ15/18 ([Table 2](#T2){ref-type="table"}). If the individual changes (mJ15/18 and mP18) were independent and additive, one would have expected a roughly 2-fold increase in activity. However, the 49-fold increase for ME-mJ15/18/P18 versus ME indicates positive cooperativity between the two structural alterations. Activity with the bacterial P protein as cofactor ------------------------------------------------- When we assayed P RNA mutants for holoenzyme activity using the *E. coli* P protein as cofactor ([Table 2](#T2){ref-type="table"}, right columns), again only a single variant showed a marked increase in activity (26-fold) over ME RNA (ME-mJ2/3, with J2/3 reduced from 3 to 1 nt; [Figure 2](#F2){ref-type="fig"}). The fact that this small alteration caused a 9-fold activity decrease in the RNA-alone reaction, but 26-fold increase in the holoenzyme reaction suggested an effect on P protein binding, which is consistent with current knowledge about the P RNA binding interface of the bacterial P protein (Figure S1). All other individual changes had little effect (within 3-fold) on the holoenzyme reaction. Combination of the J15/18 mutation with one of the other structural alterations in the second round revealed two variants with substantially improved activity in the holoenzyme reaction: ME-mJ15/18/P18 (6-fold) and ME-mJ15/18/J2/3 (160-fold; [Table 2](#T2){ref-type="table"}). Changes in P2--P3 and the triple mutant ME-J15/18/2/3/P2 -------------------------------------------------------- The activity increase in the holoenzyme reaction owing to deletion of two nucleotides in J2/3 prompted us to combine changes towards the bacterial consensus in J2/3 and P2, aiming at a further improvement of *E. coli* P protein binding to the ME chimera. Helix-P2, which is part of the bacterial P protein binding site (see Introduction section and Figure S1), consists of 6 bp in MM RNA compared with 7 bp in bacterial P RNAs. We thus constructed P RNA variants that combine the J2/3 mutation with a P2 helix extended by 1 bp. In the first mutant an U--A bp was inserted between the fifth and the sixth base pair (counted from the 5′-end; [Figure 2](#F2){ref-type="fig"}) of the archaeal P2 helix. This increased RNA-alone activity 20-fold (ME-mJ2/3/P2), whereas these alterations individually impaired activity 9-fold (mJ2/3) and 7-fold (mP2; [Table 2](#T2){ref-type="table"}). In the holoenzyme reaction, the activity gain for variant ME-mJ2/3/P2 was 450-fold over ME RNA, compared with a 26-fold rate increase for variant ME-mJ2/3 and a 2-fold drop for variant ME-mP2. Thus, we observed positive cooperativity effects for the combined mJ2/3/P2 changes in the RNA-alone and holoenzyme assays, indicating that the P2--J2/3--P3 region not only contributes to bacterial P protein binding, but also to folding of the catalytic core. We further constructed variant ME-mJ2/3/P2-2 with the goal to restore the bacterial consensus with minimal structural alteration, solely deleting 1 nt in J2/3 and introducing a point mutation that creates an additional U--A bp at the end of P2. This variant improved RNA-alone activity of ME RNA 7-fold and raised substrate turnover 240-fold in the holoenzyme reaction ([Table 2](#T2){ref-type="table"}). The 2--3-fold weaker performance of variant ME-mJ2/3/P2-2 relative to ME-mJ2/3/P2 suggests that a terminal U--A bp at the P2/P3 junction has a slightly destabilizing effect compared with a closing C--G bp. We then combined the mJ15/18 point mutation with the mJ2/3/P2 alteration to generate variant ME-J15/18/2/3/P2. This further increased activity roughly 4-fold relative to ME-mJ2/3/P2 in the RNA-alone reaction, but showed little additional effect in the holoenzyme reaction ([Table 2](#T2){ref-type="table"}), for unknown reasons. Length of P1 and the L9--P1 interdomain contact ----------------------------------------------- In some bacterial P RNAs, a long-range tertiary interaction between loop L9 docking onto helix P1 serves as an interdomain strut of major importance for the global RNA fold ([@B20],[@B29]). The mature 5′- and 3′-ends of MM RNA, mapped in a previous study ([@B3]), would be too short to allow formation of such an L9--P1 contact. However, some archaeal P RNAs seem to have an extended P1 helix compatible with formation of the L9--P1 contact, and even the MM RNA gene sequence encodes the potential for an extended P1 helix that, if part of the P RNA transcript, may enable formation of the P1--L9 interaction. ([@B17]). We thus constructed an MM RNA variant with P1 extended by 7 bp based on the sequence of the P RNA (*rnpB*) gene of *M. thermautotrophicus* strain ΔH. This increased the RNA-alone activity of MM RNA 200-fold ([Table 3](#T3){ref-type="table"}). Introduction of this P1 extension into variant ME-mJ15/18/2/3/P2 to generate ME-mJ15/18/2/3/P2/nP1 further increased activity to 147-fold (RNA-alone) and 1500-fold (holoenzyme) over ME RNA ([Table 2](#T2){ref-type="table"}). Table 3.Activity of MM variants in RNA-alone and holoenzyme assaysRNase P RNARNA-aloneHoloenzyme*k*~obs~ \[min^--1^ × 10^−4^\]Change*k*~obs~ \[min^−1^ × 10^−4^\]ChangeMM1.7 ± 0.1    12 ± 1    1MM-nP1340 ± 60↑ 200×1.3 ± 0.1↓ 2×MM-mJ15/18/P182 ± 1    11.2 ± 0.1↓ 2×MM-mJ2/3/P2230 ± 50↑ 135×7 ± 1↑ 4×MM-mJ2/3/P2-2100 ± 20↑ 59×6.7 ± 0.8↑ 4×MM-mJ15/18/2/3/P2270 ± 60↑ 159×80 ± 10↑ 40×MM-mJ15/18/2/ 3/P2/nP1740 ± 60↑ 435×75 ± 4↑ 38×MM-Marburg3.3 ± 0.6Not detectableMM-Marburg-nP1310 ± 202.1 ± 0.9MM-Marburg- mapped ends240 ± 201.0 ± 0.2[^3] The stimulating effect of extension nP1 on MM RNA-alone activity prompted us to reinvestigate the 5′- and 3′-boundaries of MM RNA by RACE. This was done with the closely related *M. thermoautotrophicus* strain Marburg instead of the original source strain *M. thermautotrophicus* ΔH, because we had access to freshly grown cells of the former only. Owing to the switch of strains we performed several control experiments before conducting the RACE experiment: (i) sequencing of the gene encoding the MM-Marburg RNA revealed a few nucleotide changes relative to MM RNA of strain ΔH ([Figure 1](#F1){ref-type="fig"}), but *in vitro* transcribed MM-Marburg RNA with the short P1 helix as in MM RNA ([Figure 1](#F1){ref-type="fig"}) showed RNA-alone activity very similar to that of MM RNA ([Table 3](#T3){ref-type="table"}); (ii) introducing the same 7-bp extension as present in MM-nP1 into the MM-Marburg RNA, resulting in variant MM-Marburg-nP1 (Figure S2B), indeed revealed a similar activity increase (ca. 100-fold) as seen for MM-nP1 RNA ([Table 3](#T3){ref-type="table"}). Having confirmed that MM-Marburg RNA is an appropriate mimic of MM RNA, we carried out 5′- and 3′-RACE (Figure S2A), which unveiled that *in vivo* transcripts of MM-Marburg RNA, and by inference of MM RNA, carry 5′- and 3′-ends that permit formation of an extended P1 helix (Figure S2B) to provide a receptor site for loop L9. Construction and analysis of MM variants ---------------------------------------- Next we addressed the question if the activating structural changes nP1, mJ2/3, mP2 and mJ15/18, initially explored in the context of variant ME, would be beneficial in the natural context of MM RNA as well. Indeed, RNA-alone activity of MM RNA increased 435-fold by introducing these four minor changes ([Table 3](#T3){ref-type="table"}), while the activity gain was only 38-fold in the holoenzyme reaction ([Table 3](#T3){ref-type="table"}, columns on the right). Noteworthy, based on the comparison of cleavage rates (*k*~obs~) for P RNA MM-mJ15/18/2/3/P2/nP1 ([Table 3](#T3){ref-type="table"}) and its counterpart ME-mJ15/18/2/3/P2/nP1 ([Table 2](#T2){ref-type="table"}), the presence of the archaeal S-domain in the MM variant reduced activity 13- and 40-fold in the RNA-alone and holoenzyme reactions, respectively. Measurement of holoenzyme-substrate affinity -------------------------------------------- The relatively low catalytic performance of variant MM-mJ15/18/2/3/P2/nP1, particularly in the holoenzyme reaction, led us to analyze substrate affinities of P RNA variants reconstituted with the *E. coli* P protein. We initially tried direct measurements of affinity between P RNA and protein, but different approaches were unsuccessful in our hands. We thus switched to measuring ternary complex formation between holoenzyme and substrate ([Table 4](#T4){ref-type="table"}). As complex formation depended on P protein binding to P RNA under the applied low salt concentrations, the assay indirectly measured P RNA--protein affinities. Table 4.Affinity of reconstituted RNase P holoenzymes for the ptRNA^Gly^ substrateRNase P RNA*K*~d~ (nM)*K*~d~ change*k*~obs~ change (holoenzymes)MMn.d.    1MM-mJ2/3/P2n.d.↑ 4×MM-mJ15/18/2/3/P2n.d.↑ 40×MM-mJ15/18/2/3/P2/nP1n.d.↑ 38×MEn.d.    1ME-mJ2/3448 ± 30↑ 895×↑ 2×ME-mJ2/3/P247 ± 11↑ 94×↑ 450×ME-mJ15/18/2/348 ± 4↑ 96×↑ 160×ME-mJ15/18/2/3/P24.2 ± 1.2↑ 8×↑ 465×ME-mJ15/18/2/3/P2/nP15.0 ± 1.0↑ 10×↑ 1500×EE0.5 ± 0.1    1↑ 41 000×[^4] For the original ME chimera as well as MM RNA and all variants thereof ptRNA affinity was too low for *K*~d~ determination ([Table 4](#T4){ref-type="table"}). This changed with variant ME-mJ2/3, for which a *K*~d~ of 450 nM could be determined. Additional introduction of the J15/18 mutation increased affinity roughly tenfold. Similarly, the P2 mutation introduced into either the single mutant ME-mJ2/3 or the double mutant ME-mJ15/18/2/3 in either case led to a tenfold increase in affinity. The lowest *K*~d~ (4.2 nM) was obtained for the ME-mJ15/18/2/3/P2 RNA; additional incorporation of the nP1 extension did not further reduce *K*~d~ (5 nM). Thus, ptRNA affinities obtained with the best ME variants were 10-fold lower than that of the *E. coli* holoenzyme (EE RNA, *K*~d~ = 0.5 nM; [Table 4](#T4){ref-type="table"}). For the ME variants, increases in affinity (reductions in *K*~d~) nicely correlated with increases in bacterial holoenzyme activity, with the exception of the nP1 extension which enhanced activity ca. 3-fold when added to variant ME-mJ15/18/2/3/P2, but left substrate affinity essentially unaffected ([Table 4](#T4){ref-type="table"}). Low substrate affinity of the MM variants explains why activity gains in the bacterial holoenzyme assay were moderate (4--40-fold): since the archaeal S-domain is defective in substrate binding based on our *K*~d~ measurements and thus limits the rate of the cleavage reaction, beneficial effects of alterations mJ2/3, mJ15/18 and mP2 in the archaeal C-domain could not be coaxed into activities as high as in the presence of the *E. coli* S-domain. Folding analysis of P RNA variants by native PAGE ------------------------------------------------- Previous studies have shown that increased activity of P RNAs often goes along with compacted conformers becoming more prominent in native PAA gels ([@B5],[@B25],[@B29],[@B30]). Consistent with this observation, MM RNA showed the most diffuse migrational behaviour in native gels, pointing to a multitude of coexisting conformations ([Figure 3](#F3){ref-type="fig"}A). All other P RNAs analyzed, with a gain in activity compared to that of MM, displayed more distinct bands and some additional compaction, at least after the preincubation step to that allows RNA folding. Replacement of either domain of MM with the counterpart of EE RNA (variants ME and EM) resulted in stabilization of distinct folds, although EM RNA suffered from substantial aggregation as inferred from RNA retained in the gel pocket ([Figure 3](#F3){ref-type="fig"}A). Extension of the P1 helix (MM-nP1) obviously supports tertiary folding ([Figure 3](#F3){ref-type="fig"}B), as distinct conformers appeared relative to MM RNA. Interestingly, RNAs MM-nP1 and MM-M-nP1 differ in that one major conformer appears for the latter compared with two conformers in the case of RNA MM-nP1 ([Figure 3](#F3){ref-type="fig"}B). This finding shows that even minor changes between two related RNAs may have profound effects on conformational equilibria. Gel mobility of variants ME-mJ15/18/2/3/P2/nP1 and MM-mJ15/18/2/3/P2/nP1 differed substantially, with the former resembling ME RNA in showing a single major band and the latter displaying at least two major conformers, similar to MM-nP1 RNA. However, the additional changes mJ15/18, mJ2/3 and mP2 further affected conformation relative to RNA MM-nP1, as inferred from weaker relative intensity of the upper conformer and faster migration of the lower conformer in the case of RNA MM-mJ15/18/2/3/P2/nP1. Figure 3.Native 11.25% PAGE analyses to compare the conformations of P RNA variants. (**A**) P RNAs from *M. thermoautotrophicus* (MM), *E. coli* (EE) and *E. coli/M. thermoautotrophicus* chimeras (EM and ME); (**B**) MM and MM-Marburg P RNA variants with P1 extension (MM-nP1, MM-M-nP1), as well as corresponding C-domain mutants of ME and MM RNA. Assay conditions were the same as in the holoenzyme activity assay (see legend to Table 1 and Material and methods section). Samples were loaded onto the gel either after storage for 55 min on ice (−) or after preincubation for 5 min at 55°C and 50 min at 37°C (+). For further details, see Materials and methods section. UV melting profiles of selected P RNA variants ---------------------------------------------- UV melting profiles were analyzed for P RNA variants ME, MM and MM-nP1, as well as the mJ15/18/2/3/P2/nP1 variants of ME and MM RNA. Conditions were 4.5 mM Mg^2+^ and 100 mM NH~4~^+^ ([Figure 4](#F4){ref-type="fig"}), resembling those of our kinetic holoenzyme assay. Variant MM (red curve) had the lowest *T*~m~ value (73.1°C), which increased to 74.4°C when helix P1 was extended to its natural length (MM-nP1, black curve). Thus, helix P1 has an overall stabilizing effect, in line with large activity losses of bacterial P RNA variants with disrupted P1 elements ([@B31]). In addition, there is only one major unfolding transition for MM-nP1, based on the first derivative of the melting profile (dA~260~/dT), whereas more unfolding of substructures occurred in MM RNA (shoulders between 60°C and 65°C and at approx. 77°C) aside from the major transition at 73.1°C. Variant ME (blue curve) showed a major unfolding transition around 76°C, but also substantial unfolding of substructures between 50°C and 70°C. In contrast, much less unfolding of substructures at lower temperatures was seen for variant ME-mJ15/18/2/3/P2/nP1 (gray curve), for which the P1 extension might be the major cause, in analogy to the profiles of MM versus MM-nP1 RNA. Figure 4.UV melting profiles of selected P RNA variants. The shown curves represent the mean of at least three measurements; assay conditions: 250 nM P RNA, 50 mM MES pH 6.0, 2 mM EDTA, 100 mM NH~4~OAc, 4.5 mM Mg(OAc)~2~; for UV absorbance (260 nm) melting profiles (left), the *y*-axis of every species was normalized to 0.0 − 1.0 (normalized Abs); normalized melting curve shapes were not affected by variation of P RNA concentration, indicating that intramolecular unfolding was measured; 'd normalized Abs/dT' (right panel): first derivatives of the melting curves on the left; the measuring program consisted of three steps: step 1 (preincubation): heating of the sample to 55°C at a rate of 10°C/min, incubation for 5 min at 55°C, cooling of the sample to 37°C at a rate of 10°C/min, incubation at 37°C for 35 min; step 2 (measuring of melting profile): heating of the sample from 37 to 90°C at a rate of 0.5°C/min with data recording every 0.1°C; stage 3: immediately after reaching 90°C in step 2, the sample was cooled to 25°C at a rate of 2°C/min and kept at 25°C for 5 min; the next round of measurement was again started with step 1. It is further instructive to compare variant MM-mJ15/18/2/3/P2/nP1 with MM-nP1. The former (green curve) has a similar *T*~m~ and derivative melting profile as RNA MM-nP1, but shows more unfolding of local structures below 65°C than MM-nP1. Thus, introducing the set of minor changes (mJ15/18/2/3/P2) towards the bacterial consensus into the concertedly folding MM-nP1 RNA uncoupled unfolding of some local structures from unstacking of the bulk of RNA structure. Notably, the derivative melting profiles of variants MM-mJ15/18/2/3/P2/nP1 and ME-mJ15/18/2/3/P2/nP1 were identical up to 65°C, but the major unfolding transition for the latter was broader and shifted to higher temperatures (*T*~m~ = 76.3°C versus 75.0°C). Therefore, replacement of the archaeal S-domain with the *E. coli* counterpart increased overall folding stability but reduced the extent of coordinate unstacking. Kinetic parameters of ME-mJ15/18/2/3/P2/nP1 and its MM counterpart ------------------------------------------------------------------ We determined the RNA-alone single turnover kinetic parameters for the best-performing variant ME-mJ15/18/2/3/P2/nP1 to further explore the catalytic capacity of the engineered archaeal C-domain. This analysis revealed a maximum rate (*k*~react~) of approx. 1.4 min^−1^ and a *K*~m(sto)~ of about 4 μM for the P RNA ME-mJ15/18/2/3/P2/nP1 ([Figure 5](#F5){ref-type="fig"}A). For comparison, *k*~react~ and *K*~m(sto)~ were determined as 10 min^−1^ and 0.24 μM, respectively, for *E. coli* wild-type P RNA under the same conditions ([@B32]). Thus, P RNA ME-mJ15/18/2/3/P2/nP1 differs from *E. coli* P RNA by only a 7-fold lower maximum cleavage rate and a 17-fold higher *K*~m(sto)~. In contrast, the rate of cleavage by variant MM-mJ15/18/2/3/P2/nP1 showed a linear dependence on P RNA concentration up to 10 μM ([Figure 5](#F5){ref-type="fig"}B), indicative of enzyme concentration being subsaturating and consistent with low substrate affinity in the presence of the archaeal S-domain. Figure 5.RNA-alone single turnover kinetics of P RNAs (**A**) ME-mJ15/18/2/3/P2/nP1 and (**B**) MM-mJ15/18/2/3/P2/nP1 as a function of the P RNA concentration. Processing assays were performed in the presence of \<1 nM 5′-endlabelled ptRNA, 100 mM Mg(OAc)~2~, 100 mM NH~4~OAc, 50 mM MES and 2 mM EDTA, at pH 6.0 and 37°C. The *k*~obs~ values are mean values from at least three independent experiments, with error bars indicating the standard error of the mean. For P RNA ME-m J15/18/2/3/P2/nP1, the single turnover kinetic parameters could be derived. *K*~m(sto)~ is the single turnover *K*~m~ (describes the enzyme concentration at which the half maximum rate under conditions of \[E\] \>\> \[S\] is achieved), and *k*~react~ is the single turnover *V*~max~. For further details, see Materials and methods section. We further tested P RNA ME-mJ15/18/2/3/P2/nP1 for function in *B. subtilis* and *E. coli rnpB* mutant strains, also under conditions of simultaneous overexpression of bacterial P proteins to compensate for reduced P RNA-P protein affinity. However, P RNA ME-mJ15/18/2/3/P2/nP1 was unable to rescue the mutant phenotypes (Table S3). DISCUSSION ========== We have shown here that a few mutations toward the bacterial P RNA consensus substantially activate the C-domain of archaeal P RNA from *Methanothermobacter*, in the absence and presence of the bacterial RNase P protein. All variants were screened at two Mg^2+^ concentrations, 100 mM in the RNA-alone reaction and 4.5 mM in the holoenzyme reaction. Activity measurements at the two Mg^2+^ concentrations provide snapshots, as the individual variants may have different Mg^2+^ optima. Such potential differences in Mg^2+^ requirements imply that the structural changes introduced into the archaeal C-domain may exert their effects on folding, substrate binding, catalysis or binding of the RNase P protein at least partly through changes in binding of crucial Mg^2+^ ions, which has to be considered when interpreting the data. In a recent study on the RNA-alone reaction catalyzed by *E. coli* RNase P RNA, Kirsebom and coworkers demonstrated that the Mg^2+^ requirements with respect to cleavage increased when substrate contacts to the C-domain (involving substrate functional groups near the cleavage site) as well as contacts to the S-domain (involving the T-stem/loop region of the substrate) are weakened ([@B33]). The improvement in substrate positioning seen for variant ME-mJ15/18/2/3/P2 in our probing experiments (see below) may thus well include contributions from improved Mg^2+^ binding to P RNA regions that mediate contacts to the substrate. Along these lines, we showed previously for *E. coli* and *B. subtilis* P RNAs with mutations in the substrate 3′-CCA binding region that *in vitro* processing defects in the holoenzyme reaction were exacerbated when Mg^2+^ concentrations were lowered from e.g. 4.5 mM to 2 mM ([@B34],[@B35]). Likewise, P RNAs with structural deletions ([@B36]) or substrates with ribose 2′-substitutions at strategic positions were reported to cause an increase in the Mg^2+^ requirement of P RNA-alone reactions ([@B37],[@B38]). On the basis of these findings it is reasonable to predict that low activity P RNA variants in the study presented here will tend to perform relatively better at higher Mg^2+^ concentrations compared to the variants with improved catalytic performance. A central finding of our study was that ribozyme activity of *M. thermoautotrophicus* P RNA is much higher (100 to 200-fold) than previously assumed owing to the presence of an extended P1 helix ([Table 3](#T3){ref-type="table"}, [Figure 2](#F2){ref-type="fig"}). This P1 extension also stabilizes the global fold as inferred from native PAGE analysis ([Figure 3](#F3){ref-type="fig"}) and UV melting profiles ([Figure 4](#F4){ref-type="fig"}). A likely explanation is that the extension creates a receptor site for loop L9. Since the L18--P8 interdomain contact is lacking, the archaeal RNA may strongly depend on the L9--P1 contact for interdomain orientation of its type A architecture. The presence of an extended P1 helix is also likely for other archaeal P RNAs (Figure S3). Interestingly, a common motif of archaeal P1 elements is an A:A or C:A mismatch at identical position in type A and type M archaeal P RNAs proposed to interact with the third G residue of a conserved 5′-GAGA L9 tetraloop ([@B17]). We have recently demonstrated that a special L9--P1 interaction, occurring in some bacterial P RNAs from thermophiles, represents a major determinant for thermostability of the P RNAs, for RNase P holoenzyme activity at physiological Mg^2+^-concentrations and for conformational compaction ([@B29]). Of note, some bacterial P RNAs harbouring this thermostable P1--P9 module also lack the L18--P8 interaction and have apparently compensated for its absence by strengthening the L9--P1 interdomain contact ([@B29],[@B32]). We recently reported that the P1--L9 interaction is degenerate in *E. coli*-like mesophilic type A RNase P RNAs, and thus unlikely forms at all, because mutations intended to disrupt this contact remained without effect on *E. coli* P RNA function *in vitro* as well as *in vivo* ([@B29]). This finding explains why variant ME-mP1, designed to restore the *E. coli*-like L9--P1 contact when its degeneracy was still unknown, failed to improve RNA-alone activity ([Table 2](#T2){ref-type="table"}). However, it remains unclear why the impairment of activity was this severe (176-fold). A pattern emerging from our reactivation study is that mutations within a single structural element were usually insufficient to improve RNA-alone activity. With the exception of the P1 extension, it took the cooperative effect of at least two alterations to be substantially beneficial for ribozyme activity, such as mP18 plus mJ15/18, mP2 plus mJ2/3 or mP2 plus mJ2/3 plus mJ15/18 ([Table 2](#T2){ref-type="table"}). Successful double mutants combined introduction (mP18) or alteration (mP2) of a helix and fine adjustments within adjacent joining regions likely involved in proper positioning of these structural elements. The stimulatory effects of combined changes mJ2/3/P2 on RNA-alone activity were, however, unexpected. Yet, since P2 and P3 are part of the catalytic core, length changes in J2/3 may affect coaxial stacking of P2 and P3 and thereby, combined with the P2 extension, induce subtle changes in folding of the entire core comprising the P1/P4/P5 and P2/P3 stacks as well as the respective joining segments. Surprisingly, in several cases activity increases in the RNA-alone and holoenzyme assays differed substantially. The combined changes in J2/3 and P2 activated RNA-alone activity of variant ME-mJ2/3/P2 20-fold (relative to ME RNA; [Table 2](#T2){ref-type="table"}) and that of variant MM-mJ2/3/P2 135-fold (relative to MM RNA; [Table 3](#T3){ref-type="table"}); however, in the holoenzyme reaction activity stimulation was 450-fold for variant ME-mJ2/3/P2, but only 4-fold for P RNA MM-mJ2/3/P2 ([Table 2](#T2){ref-type="table"} and [3](#T3){ref-type="table"}). A similar situation was seen for the mJ15/18/2/3/P2/nP1 mutation (RNA-alone reaction: 147-fold in the ME, 435-fold in the MM context; holoenzyme reaction: 1500-fold in the ME, 38-fold in the MM context). Such relatively low stimulatory effect on MM variants in the holoenzyme reaction could be attributed to the failure of the archaeal S-domain to confer ternary complex formation at the low metal ion and enzyme concentrations present in the holoenzyme reactions. Indeed, throughout low activities for variants carrying the archaeal instead of the *E. coli* S-domain ([Tables 2](#T2){ref-type="table"} and [3](#T3){ref-type="table"}, [Figure 5](#F5){ref-type="fig"}) were generally paralleled by very little formation of holoenzyme-substrate complexes ([Table 4](#T4){ref-type="table"}). The less severe phenotype of variants harboring the archaeal S-domain in the RNA-alone relative to the holoenzyme reaction can be attributed to the fact that RNA-alone assays were performed at very high P RNA concentrations (10 μM) and high Mg^2+^ (100 mM), which apparently mitigated the effect of low substrate affinity. These findings pinpoint the S-domain as the limiting part of the archaeal P RNA. In contrast, the archaeal C-domain with a few mutations not only displayed substantial catalytic proficiency, but also highly productive interaction with the bacterial P protein. It remains to be shown whether minor mutations will be able to partially cure the S-domain\'s defect, or if larger architectural alterations, for example, the incorporation of additional structural elements such as P13/P14, will be required to restore S-domain function in the absence of archaeal protein cofactors. The more pronounced activity gain of the mJ15/18/2/3/P2/nP1 mutation in the ME versus MM context (1500-fold versus 38-fold; [Table 2](#T2){ref-type="table"} and [3](#T3){ref-type="table"}) is of further interest in view of the fact that RNAs MM and ME had equal activities in the holoenzyme reaction (2 × 10^−4^ min^−1^; [Table 2](#T2){ref-type="table"} and [3](#T3){ref-type="table"}). This discrepancy can be explained by interdomain cooperativity between the *E. coli* S-domain and the engineered archaeal C-domain carrying the mJ15/18/2/3/P2/nP1 mutations. Indeed, our structure probing data indicate changes in the *E. coli* S-domain between variants ME and ME-mJ15/18/2/3/P2 (see below). The C to U exchange in J15/18 was the only single change that had a beneficial effect on the RNA-alone reaction by itself ([Table 2](#T2){ref-type="table"}). The molecular basis of this effect is not immediately clear, since the corresponding nucleotide in the crystal structure of the type A RNase P RNA from *Thermotoga maritima* is not resolved ([@B21]) and type B RNase P RNAs have a purine at this position ([@B28],[@B39]). In *E. coli* P RNA, a 2′-deoxy modification at this U residue and Rp-phosphorothioate as well as inosine modification at the conserved 5′-terminal G of J15/18, two nucleotides upstream of the U residue ([Figure 2](#F2){ref-type="fig"}), interfered with tRNA binding ([@B24],[@B40]). This G residue forms a *trans* Hoogsteen/sugar edge base pair with the 3′-terminal A residue in J5/15 in the crystal structure of type B RNase P RNA from *B. stearothermophilus* ([@B39]). These findings suggest the J15/18 region to be involved in a structurally and functionally important network of interactions. Comparative probing of P RNAs ME and ME-mJ15/18/2/3/P2 with RNase T1 and Pb^2+^ ions under native conditions unveiled distinct structural differences between the two RNA variants, surprisingly spanning the entire subregion P6, J6/16, P16, J16/15, P15, J15/18 and J5/15 ([Supplementary Figure S4](http://nar.oxfordjournals.org/cgi/content/full/gkn915/DC1)). J16/15 is the site of interaction with the tRNA 3′-CCA end. With respect to J15/18 itself, the 5′-terminal G residue was observed to be more accessible to RNase T1 cleavage in ME versus ME-mJ15/18/2/3/P2 RNA, whereas the site of mutation and the following C residue were more accessible to Pb^2+^-induced hydrolysis in ME-mJ15/18/2/3/P2 relative to ME RNA (Figure S4). Surprisingly, differences in accessibility were also seen in L11/12 and P11, although these regions are identical in both P RNA variants. This indicates that changes mJ15/18/2/3/P2 in the C-domain have long-range effects on S-domain structure. L11/12 and P11 are close to the T-loop region in the P RNA-tRNA complex model based on the *T. maritima* P RNA X-ray structure ([@B21]). In conclusion, changes mJ15/18/2/3/P2 seem to affect regions that interact with the substrate (J15/16) or which are close to the substrate binding interface (L11/12 and P11). The beneficial effects of these changes on P RNA activity thus seem to stem from improvements in substrate positioning. The structural probing data raise the question if the mutations introduced in P RNA ME-mJ15/18/2/3/P2 will have similar structural and functional consequences when placed into other archaeal C-domains. With respect to evolutionary implications of our approach, several findings are remarkable. First, only three minor changes, in J2/3, J15/18 and P2, were required to stimulate RNA-alone activity 80-fold in the ME and 160-fold in the MM context (465- and 40-fold, respectively, in the holoenzyme assay). Second, combining the changes in J2/3 and P2 improved activity in the absence as well as presence of the bacterial P protein. Thus, the latter exerts its function by making use of a structural element crucially important for RNA-alone activity. This may suggest an evolutionary scenario according to which the bacterial P protein was recruited to the catalytic RNA and adapted to it without eliciting significant co-evolution of the RNA subunit. In contrast, the strong dependence of archaeal/eukaryal P RNAs on their protein subunits for enzyme function indicates RNA/protein co-evolution, during which the archaeal P RNA subunit adapted to the protein components concomitant with substantial losses of RNA-alone activity. From a mirror-image perspective, it will be intriguing to explore if, for example, P RNA ME-mJ2/3/P2 may be impaired in its capacity to cooperate with the archaeal C-domain binding proteins Pop5 and Rpp30 ([@B41]). Along these lines, the substantial activation of the archaeal C-domain by surprisingly little changes suggests only minor differences in RNA conformation between bacterial and archaeal C-domains, in line with the structure probing results (see above). This may imply that the main function of the archaeal C-domain binding proteins Pop5 and Rpp30 for the C-domain itself is to merely fine-tune the conformation of the catalytic core. Finally, the maximum single turnover cleavage rate of variant ME-mJ15/18/2/3/P2/nP1 was only 7-fold lower and the *K*~m(sto)~ only 17-fold higher than the corresponding values measured for *E. coli* P RNA. It will be interesting to see if restoring the L9--P1 and/or L18--P8 interdomain struts in this ME chimera will further reduce these differences. In conclusion, the profound activation of the archaeal C-domain by minor changes demonstrates the evolutionary closeness of the archaeal and bacterial P RNAs, at least in terms of C-domain structure and function. SUPPLEMENTARY DATA ================== [Supplementary Data](http://nar.oxfordjournals.org/cgi/content/full/gkn915/DC1) are available at NAR Online. FUNDING ======= This work was supported by the Deutsche Forschungsgemeinschaft (HA 1672/14-1 and GK 1384). Funding for open access charge: DFG/GK 1384. *Conflict of interest statement*. None declared. Supplementary Material ====================== ###### \[Supplementary Data\] The gifts of pUC119_T7_M.th.\_rnpB (James W. Brown, North Carolina State University) and cells of *M. thermoautotrophicus* strain Marburg (Rolf Thauer, Max Planck Institute for Terrestrial Microbiology, Marburg) are gratefully acknowledged. [^1]: Assay conditions for the RNA-alone reaction were 50 mM MES pH 6.0 at 37°C, 2 mM EDTA, 100 mM NH~4~OAc, 100 mM Mg(OAc)~2~, trace amounts of ptRNA and 10 µM P RNA unless otherwise indicated. Holoenzyme activity assays were performed at 20 mM HEPES pH 7.4 at 37°C, 2 mM spermidine, 0.05 mM spermine, 4 mM β-mercaptoethanol, 150 mM NH~4~OAc, 4.5 mM Mg(OAc)~2~, 100 nM ptRNA, 10 nM P RNA and 50 nM recombinant *E. coli* P protein with N-terminal His tag. Errors are standard deviations; all values are based on at least three independent experiments. For further experimental details see Materials and methods section. Values for EE and Ecat RNA are taken from ref. ([@B5]). [^2]: For assay conditions and experimental details, see legend to Table 1 and Materials and methods section. [^3]: For assay conditions and experimental details, see legend to Table 1 and Materials and methods section. For structural details on P1 elements in RNAs MM-Marburg, MM-Marburg-nP1 and MM-Marburg-mapped ends, see Figure S2B. [^4]: Errors for *K*~d~ are standard errors of the curve fit; n.d., non-determinable because of affinities too low for *K*~d~ determination by this assay. *K*~d~ changes are given as *n*-fold increase in *K*~d~ relative to the *E. coli* holoenzyme (EE). Note that in our assay for all variants ptRNA binding to P RNA in the absence of protein did not exceed background values (ptRNA alone). Thus, the binding data for the ternary complex indirectly permit to assess P protein binding to P RNA; *k*~obs~ changes listed for comparison are taken from Tables 2 and 3.

Significance StatementThere is no single medical cause for autism, and there are very few controlled studies. This research suggests that stem cells from autologous cord blood are safe and potentially have an impact on socialization for children with autism. The safety and observed changes warrant further investigation into the potential for cellular therapy as an intervention in pediatric neurologic conditions that do not require bone marrow replacement. This research provides insight and suggestions for future research conducted using stem cells from autologous umbilical cord to treat autism. Introduction {#sct312282-sec-0002} ============ Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder that affects 1:68 children in the U.S. [1](#sct312282-bib-0001){ref-type="ref"}. While there are treatments for comorbid conditions as well as behavioral interventions for the core deficits, there is currently no cure. Based on its definition, ASD symptoms include deficits in communication and language as well as social reciprocity, and self‐stimulatory or perseverate type behaviors [2](#sct312282-bib-0002){ref-type="ref"}. These core features may all reflect a lack of integrative communication between the different brain regions that make up normal executive, social/emotional, and communicative behaviors that define normal development. ASD is a heterogeneous disorder and multiple risk factors including genetic predisposition, prenatal and postnatal environmental exposures, and immune dysregulation may contribute to its development [3](#sct312282-bib-0003){ref-type="ref"}, [4](#sct312282-bib-0004){ref-type="ref"}, [5](#sct312282-bib-0005){ref-type="ref"}. Twin studies have reported a higher concordance rate of ASD in monozygotic twins compared with dizygotic twins, however the monozygotic twin concordance rate is far less than 100% suggesting a distinct nongenetic contribution [6](#sct312282-bib-0006){ref-type="ref"}, [7](#sct312282-bib-0007){ref-type="ref"}. Current opinion suggests that ASD may result from the occurrence of several risk factors. Immune factors may be one important component of this theory. Systemic immune dysfunction in individuals with ASD has been reported in several publications; however, the exact dysfunction has been variable [8](#sct312282-bib-0008){ref-type="ref"}. This variability is to be expected in a heterogeneous disease with multiple potential etiologies such as ASD. Maternal immune activation has been suggested as a potential environmental risk factor for developing ASD [3](#sct312282-bib-0003){ref-type="ref"}, [8](#sct312282-bib-0008){ref-type="ref"}. In addition, post‐mortem analysis of brains of individuals with ASD indicated microglial activation and neuroinflammation [9](#sct312282-bib-0009){ref-type="ref"}, [10](#sct312282-bib-0010){ref-type="ref"}. Historically, the brain has been considered to be immunologically privileged, however a recent study suggested that lymphatic endothelial cells line the dural sinuses in mice [11](#sct312282-bib-0011){ref-type="ref"}. This might explain findings from earlier studies suggesting that the central nervous system undergoes constant immune surveillance in the meningeal compartment [12](#sct312282-bib-0012){ref-type="ref"}, [13](#sct312282-bib-0013){ref-type="ref"}. Brain structure and connectivity in ASD may also be abnormal. Post‐mortem and brain MRI imaging studies have suggested that a period of neuronal overgrowth and hyperconnectivity in early childhood may be followed by hypo‐connectivity later in life [9](#sct312282-bib-0009){ref-type="ref"}, [10](#sct312282-bib-0010){ref-type="ref"}. Other studies have suggested that the brains in individuals with ASD might have areas of both hypo and hyperconnectivity [14](#sct312282-bib-0014){ref-type="ref"}. Immune dysregulation including altered cytokine profiles, T‐cell and B‐cell dysfunction, and in utero stress factors may alter the fetal environment and change neuronal interaction and connectivity via glial cell activity [8](#sct312282-bib-0008){ref-type="ref"}, [9](#sct312282-bib-0009){ref-type="ref"}, [15](#sct312282-bib-0015){ref-type="ref"}, [16](#sct312282-bib-0016){ref-type="ref"}, [17](#sct312282-bib-0017){ref-type="ref"}, [18](#sct312282-bib-0018){ref-type="ref"}, [19](#sct312282-bib-0019){ref-type="ref"}, [20](#sct312282-bib-0020){ref-type="ref"}, [21](#sct312282-bib-0021){ref-type="ref"}, [22](#sct312282-bib-0022){ref-type="ref"}. One study found that the ratio of tumor necrosis factor alpha (TNF‐α) in cerebrospinal fluid versus serum may be elevated [19](#sct312282-bib-0019){ref-type="ref"}. Other studies have shown activated microglia and elevation of cytokines such as IL‐6 and MCP‐1 [9](#sct312282-bib-0009){ref-type="ref"}, [18](#sct312282-bib-0018){ref-type="ref"}.^,^ Maternal and ASD subject serum circulating autoantibodies have been described [20](#sct312282-bib-0020){ref-type="ref"}. Prior studies have examined the use of intravenous immunoglobulin (IVIG), steroid therapy, and specific medication for anti‐inflammatory inhibition of cytokines for immune dysfunction associated with ASD [17](#sct312282-bib-0017){ref-type="ref"}, [23](#sct312282-bib-0023){ref-type="ref"}, [24](#sct312282-bib-0024){ref-type="ref"}, [25](#sct312282-bib-0025){ref-type="ref"}, [26](#sct312282-bib-0026){ref-type="ref"}. Experience with immune dysfunction and ASD therapies to date have highlighted the lack of consensus of what is the best approach with standard medications. Direct cytokine inhibition in animal models and in a human pilot study suggests a correlation of reduced TNF‐α and interleukin beta (IL‐1β) with improved outcomes [27](#sct312282-bib-0027){ref-type="ref"}. Umbilical cord blood (UCB) contains a mixed population of cells and is rich in hematopoietic stem cells. Theoretically, UCB has the potential to regulate an abnormally activated immune system which may potentially improve neuronal function [28](#sct312282-bib-0028){ref-type="ref"}, [29](#sct312282-bib-0029){ref-type="ref"}, [30](#sct312282-bib-0030){ref-type="ref"}. Umbilical cord tissue and UCB have also been used to regulate immune dysfunction [31](#sct312282-bib-0031){ref-type="ref"}. Preclinical studies have reported that cells from intravenously infused human UCB can migrate to the parenchyma of injured brain and alleviate neurological impairment in animal models of perinatal brain injury [32](#sct312282-bib-0032){ref-type="ref"}, [33](#sct312282-bib-0033){ref-type="ref"}, [34](#sct312282-bib-0034){ref-type="ref"}. The demonstrated functional improvements were not due to cell differentiation and repopulation, but rather are hypothesized to be a transient paracrine effect of the infused cells upon the host cells. One proposed mechanism of action is that the donor cells secrete factors that might promote endogenous repair and angiogenesis [35](#sct312282-bib-0035){ref-type="ref"}. Currently, UCB has been used as a source of hematopoietic stem cells in over 35,000 transplants worldwide for the treatment of hematological and immunological conditions [36](#sct312282-bib-0036){ref-type="ref"}, [37](#sct312282-bib-0037){ref-type="ref"}, [38](#sct312282-bib-0038){ref-type="ref"}. The use of autologous UCB (AUCB) offers little risk of immune system reactions. A pilot study assessing the safety of AUCB infusions in newborn infants with hypoxic‐ischemic encephalopathy reported that the infusions were well tolerated and no adverse events were reported [39](#sct312282-bib-0039){ref-type="ref"}. In a separate study, 76 children with hydrocephalus received up to four AUCB infusions and no infusion‐related adverse events were reported [40](#sct312282-bib-0040){ref-type="ref"}. A publication describing AUCB infusions in 184 infants and children with cerebral palsy, congenital hydrocephalus, or other brain injuries reported that 1.5% had only mild hypersensitivity reactions (hives and/or wheezing) with no additional adverse events reported up to 3 years later [41](#sct312282-bib-0041){ref-type="ref"}. Considered overall, the research to date investigating AUCB infusions has established safety in a variety of pediatric populations, however, the question of efficacy remains unanswered because of a lack of studies conducted as randomized clinical trials or that include control groups. A recent meta‐review of four randomized‐controlled trials and one nonrandomized clinical trial compared the use of different cell types in patients with cerebral palsy. This analysis suggested a small statistically significant intervention effect on gross motor scales with UCB being the most effective compared with other cell types analyzed [42](#sct312282-bib-0042){ref-type="ref"}. The current research and clinical experience has established a sound rationale to support investigation of AUCB in patients with ASD. [8](#sct312282-bib-0008){ref-type="ref"}, [30](#sct312282-bib-0030){ref-type="ref"}, [31](#sct312282-bib-0031){ref-type="ref"}, [43](#sct312282-bib-0043){ref-type="ref"} The objective of this exploratory study was to assess the safety of a single infusion of AUCB in subjects with ASD and document changes in language, social behavior, and learning. Materials and Methods {#sct312282-sec-0003} ===================== This study was a randomized, blinded, placebo‐controlled exploratory trial with treatment arm crossover at 24 weeks. Participants were infused with either AUCB or placebo, evaluated at baseline, 12, and 24, weeks, infused with the opposite product, then evaluated again at 12 and 24 weeks post infusion. (Fig. [1](#sct312282-fig-0001){ref-type="fig"}) This study was approved by the Sutter Health Sacramento Institutional Review Committee and registered at [clinicaltrials.gov](http://clinicaltrials.gov) number NCT01638819. {#sct312282-fig-0001} Participants {#sct312282-sec-0004} ------------ Eligible subjects were children ages 2 to 7 years with a diagnosis of ASD based on DSM‐IV‐TR and the Autism Diagnostic Observational Schedule (ADOS) administered by the neuropsychologist and corroborated by the Principal Investigator (PI). Participants were required to have AUCB cryopreserved at Cord Blood Registry (CBR, South San Francisco, CA) processed on the AutoXpress (AXP) Platform (Cesca Therapeutics, Rancho Cordova, CA). Children with a diagnosis of Asperger or Pervasive Developmental Disorder‐Not Otherwise Specified were not eligible to participate. Review of the potential subjects' medical histories were completed to rule‐out genetic conditions or other potentially confounding medical diagnoses. Subjects were required to have documentation of a normal karyotype, Fragile X DNA testing, and chromosomal microarray analysis. Subjects were also required to have 24 hours EEG studies within 6 months of baseline in order to define the population better and determine if EEG abnormalities would change during the study. Cord Blood Unit Criteria and Dosing {#sct312282-sec-0005} ----------------------------------- Cord blood units were required to meet the following criteria as documented at the time of storage at CBR: post‐processing total nucleated cell (TNC) count at least 10 × 10^6^/kg, viability \> 85%, negative sterility cultures, maternal infectious disease marker testing performed and negative (HIV, Hepatitis B and C, HLTV, syphilis) or, if not performed at time of unit storage, confirmed negative through prenatal medical record review. The cord blood unit identity was confirmed via HLA typing of a cord blood test aliquot and peripheral blood of the study subject prior to cord blood unit release. ABO and Rh status were also reviewed as part of identity verification. In addition, a cord blood aliquot was tested for TNC, viability, colony forming unit, and CD34+ cell count to confirm potency prior to cord blood unit release. Cord blood units were initially cryopreserved in a bag consisting of two compartments with an 80%/20% volume configuration. Based on the total available TNC count, only a portion of the cord blood unit was required to meet the minimum cell dose requirements of the study. Either the 80% or 20% compartment was released from CBR for use in the trial and the remainder of the unit remains in storage for potential future use. Subject Enrollment and Randomization {#sct312282-sec-0006} ------------------------------------ After informed consent and screening for eligibility, subjects were enrolled in the study and randomized to a treatment arm. The computerized randomization program assigned subjects 1:1 in blocks of 4 to group A (infusion of AUCB first, infusion of 0.9% saline placebo second) or group B (infusion of 0.9% saline placebo first, infusion of AUCB second). Cord Blood Unit Preparation and Administration {#sct312282-sec-0007} ---------------------------------------------- The cryopreserved cord blood units were shipped from CBR to the local cell therapy laboratory in a dry shipper validated to maintain appropriate temperature. Cord blood units remained cryopreserved until the scheduled infusion date as indicated in the randomization scheme. On the day of the AUCB infusion, the AUCB product arrived at the pediatric infusion outpatient clinic via courier from the cryopreservation lab in a temperature monitored dry‐shipper. Preparation of the product for infusion was conducted in a room separate from study subjects and parents. The blinded PI was not present for the infusion, but was present for pre‐infusion monitoring and post‐infusion safety monitoring on the following day. In preparation for the AUCB infusion, authorized study personnel with knowledge and experience in transplant procedures removed the product bag from the dry shipper, washed the sealed product bag with chlorohexidine, and warmed the sample to 37°C in a sterile water bath. Once the AUCB product was thawed, the unblinded transplant physician extracted the AUCB product from the bag using a sterile syringe diluted with sterile saline to a total volume of 50 ml. After standard blood filtration, the diluted product was administered via peripheral IV site by direct injection via syringe. A small sample of the AUCB product was saved for post‐thaw testing. Premedication of 0.5 mg/kg IV diphenhydramine (Benadryl) was administered in 10 ml saline and dispensed as 50 mg/ml. Dimethyl sulfoxide (DMSO) was used as a cryoprotectant agent for the AUCB during freezing and storage. Because DMSO has a garlic‐like odor, all subjects consumed garlic oil orally from a syringe or mixed with food to preserve blinding. In addition, small amounts of DMSO were placed in containers around the room to produce the odor of DMSO during the infusion. The IV bag and line were covered in cloth/tape to maintain blinding. Vital signs were taken within 30 minutes before start of the IV and every 15 minutes once the saline drip started until infusion (AUCB or placebo) was completed, which took between 5 to 15 minutes. Additionally, subjects were monitored for indications of adverse events and vital signs were monitored every 15 minutes for 1 hour following the infusion. Subjects were discharged after completion of the observation period if subjects were asymptomatic and showed no sign of allergic reaction or unstable vital signs. Subjects returned for an outpatient safety visit with the PI the day after the infusion to assess for infusion‐related side effects and obtain routine lab tests. Post‐Thaw Cord Blood Unit Testing {#sct312282-sec-0008} --------------------------------- A small sample of the cord blood unit was sent on ice by courier to the study laboratory to complete the following post‐thaw testing: sterility, % viability, CD34+ cell counts, and for most cases, CFU analysis. Toward the end of the study, the original lab used for all AUCB testing closed which required processing subsequent samples at a second lab. The second lab was unable to conduct CFU analysis. Endpoint Testing {#sct312282-sec-0009} ---------------- ### Safety {#sct312282-sec-0010} Evaluation of subject safety was conducted by obtaining a complete blood count, comprehensive chemistry panel, and urinalysis at the screening, safety, and follow‐up visits. Safety endpoints included documentation of adverse events as reported by the investigator at an office visit the day after the infusion. Parents were also asked to keep a diary of any observed changes in behavior, irritability, and language, which were discussed at the 12 and 24‐week follow‐up visits. Participants who had abnormal EEGs at baseline had repeat EEGs at 12 and 24 weeks. Abnormalities had to show specific bifrontal predominant generalized or focal spike wave activity predominantly in temporal central parietal regions. ### Primary Endpoints {#sct312282-sec-0011} The primary outcome endpoints were the Expressive One Word Picture Vocabulary Test, 4th edition (EOWPVT‐4) and Receptive One Word Picture Vocabulary Test, 4th edition (ROWPVT‐4). The EOWPVT‐4 and ROWPVT‐4 are standardized measures of single word comprehension and expression respectively. These co‐normed tests are administered by a neuropsychologist and allow comparisons of a child\'s receptive and expressive vocabulary skills. ### Secondary Endpoints {#sct312282-sec-0012} Secondary endpoints included: (a) Stanford Binet, 5th edition Fluid Reasoning (SBFR) and Knowledge (SBKN) subtests. These tests are administered by a neuropsychologist and provide a standardized assessment of verbal as well as nonverbal cognitive abilities across the life span; (b) Vineland Adaptive Behavior and Socialization Scales, 2nd edition for Communication, Daily Living Skills, Socialization, and Adaptive Behavior Composite (ABC). These are questionnaires completed by a parent or caregiver. Scores above 80 are classified using approximately the same ranges as IQ tests. Scores below 80 are categorized as borderline adaptive functioning (70--80); mildly deficient adaptive functioning (51--69); moderately deficient adaptive behavior (36--50); severely deficient adaptive behavior; (20--35); and markedly or profoundly deficient adaptive behavior (\<20) [44](#sct312282-bib-0044){ref-type="ref"}. (c) Clinical Global Impression (CGI) subscales for Social, Receptive, and Expressive skills as rated by the PI based on clinical impressions. The baseline CGI is a measure of severity as rated on a Likert scale from 1 to 7. For this study, severity was categorized as minimal deficits (1--2.5), mild to moderate deficits (3--4.5), and severe deficits in social receptive and expressive speech functions (5.0+). At each subsequent visit, the PI rated patients on improvement from the previous visit on a scale of 1 to 7. For this study, improvement was categorized as improved (≤2.5), no change (3.0--3.5), and worse (4.0+). Subjects underwent in‐person testing for these study endpoints at baseline 12, and 24 weeks after infusion of each product. Sample Size and Statistical Analysis {#sct312282-sec-0013} ------------------------------------ Fifteen patients in each treatment group were calculated as necessary to provide 84% power to detect a 7 (SD = 9.1) unit advantage for AUCB over placebo on the primary endpoint of ROWPVT (*α* = 0.05, two‐tailed) [45](#sct312282-bib-0045){ref-type="ref"}. Histograms and normal probability plots were used to screen the data for outliers and assess normality of individual variables. Scatterplots, correlation coefficients, and means were used to assess the association between post thaw CD34 and percent viability with the primary endpoints. Generalized linear models were used to compare differences between responses on primary and secondary endpoints under the AUCB versus placebo conditions using all available observations. Analyses included the repeated effects of the response variables at the 12‐ and 24‐week time periods after infusion of each product (AUCB, placebo). Group of administration (AUCB first or placebo first) and the group X treatment interaction were included in the model as fixed effects. The analysis was designed to compare the mean difference as: \[12 *and* 24 *week AUCB* -- *baseline AUCB*\] − \[12 *and* 24 *week placebo* -- *baseline placebo*\] A statistically significant interaction indicated presence of a carryover effect, meaning that differences in the response variable depended on administration group. For these variables, differences between the subjects that received AUCB first were compared with the subjects who received placebo first and analysis of covariance (ANCOVA) was used to compare the 12‐week and 24‐week scores between the groups after adjusting for baseline scores. Multiplicity was corrected for by controlling the false discovery rate (FDR) at a level of 0.05 for groups of end points where there was at least one statistically significant finding at the *p* \< .05 level [46](#sct312282-bib-0046){ref-type="ref"}. Because this was an exploratory trial, both the raw *p*‐values and FDR *p*‐values were reported. Results {#sct312282-sec-0014} ======= Thirty subjects were randomized and all except one received both infusions. One subject received the baseline infusion and was lost to follow‐up leaving 29 subjects with complete data on primary and secondary endpoints. Table [1](#sct312282-tbl-0001){ref-type="table-wrap"} provides the demographics and information regarding the AUCB that was infused. Ninety five percent of units tested post‐thaw demonstrated CFU growth. ###### Baseline demographic and autologous umbilical cord blood unit characteristics of subjects that completed the study (*N* = 29) *n* (%) Mean (Min--Max) ------------------------------------------------------------------------ ----------- --------------------------- **Age (years)** 4.53 (2.42--6.80) **Sex** Female 4 (13.8) Male 25 (86.2) **Weight (kg)** 19.21 (13--34) **Racial/Ethnicity** Asian 4 (13.8) White 24 (82.8) Hispanic 5 (17.2) Other 1 (3.4) **Abnormal EEG at baseline** 9 (31.0) **ADOS** ** **Social affect 15.14 (6.00--22.00) Restricted and repetitive behavior 3.76 (0.00--8.00) Comparison 7.75 (3.00--10.00) **Compartment infused** 80% 21 (72.4) 20% 8 (27.6) **Viable total nucleated cell (vTNC) count infused ×10^6^ (*n* = 22)** 335.09 (103.00--1,024.10) **TNC dose infused × 10^6^/kg (*n* = 22)** 16.16 (6.20--31.82) **Total colony forming units (per 1 × 10^5^)** 20 137.0 (0--360.00) **Percent viability post‐thaw, (flow cytometry 7‐AAD)** 53.73 (30.00--70.00) **Percent viable CD34 -- post‐thaw (*n* = 28)** 0.47 ± (0.08--1.48) Abbreviations: ADOS, Autism Diagnostic Observational Schedule; EEG, electroencephalogram. There was no association between 12‐ and 24‐week EOWPVT and ROWPVT scores and post thaw total viability or CD34 count based on correlation coefficients and scatterplots. Stratification of the post thaw CD34 at the median of 0.44 showed that subjects who received an infusion above the median had lower baseline scores than subjects who received an infusion that was below the mean. However, there was little change over time between participants within a stratum. For example, ROWPVT scores for participants who received an infusion that was below the median were 82.64 ± 30.55 at baseline, 81.71 ± 26.56 after 12 weeks, and 84.79 ±24.74 after 24 weeks. ROWPVT scores for participants who received an infusion with a CD34 percent above the median were 76.80 ± 21.26 at baseline, 76.40 ± 20.37 after 12 weeks, and 76.40 ±19.23 after 24 weeks. Safety {#sct312282-sec-0015} ------ Table [2](#sct312282-tbl-0002){ref-type="table-wrap"} shows the reported AEs and the likely relatedness of the events to the AUCB infusion. No adverse events required treatment. There were no observed allergic reactions or serious adverse events associated with the administration of AUCB. Nine of the 29 participants had abnormal baseline 24 hours EEG ambulatory results within 6 months of baseline. This reflects about average the percent of abnormal EEG percentage range found in this age group in other studies (30%--60%). There were no observed changes in EEGs at 12 or 24 weeks for subjects when in either group for patients who had abnormal EEGs at baseline. ###### Adverse events and potential relatedness to autologous cord blood infusion Relatedness ------------------------------------------------- -------- ------------- ------- -------- -------- Constitutional symptoms 37 43% 1 2 34 Skin and subcutaneous tissue disorders 3 3% 0 0 3 Gastrointestinal disorders 26 30% 0 8 18 Musculoskeletal and connective tissue disorders 0 0% 0 0 0 Neurological disorders 0 0% 0 0 0 Pain 0 0% 0 0 0 Pulmonary/Upper respiratory disorders 11 13% 0 0 11 Reproductive system and breast disorders 0 0% 0 0 0 Psychiatric disorders 0 0% 0 0 0 Renal and urinary disorders 9 10% 2 4 3 **Total** **86** **100%** **3** **14** **69** Clinical Global Impression {#sct312282-sec-0016} -------------------------- Table [3](#sct312282-tbl-0003){ref-type="table-wrap"} shows the CGI baseline severity scores and number and percent in each improvement category at the 12‐ and 24‐week visits. Baseline CGI severity and expressive, receptive, and social improvement were similar for both groups. ###### Baseline severity Clinical Global Impression and improvement after 12 and 24 weeks under each condition ----------------------------------------------------------------------------------------------------------------------------------------------------------------------- Baseline severity (*N* = 29) 12 week change from baseline\ 24 week\ (*N* = 29) change from baseline\ (*N* = 29) ---------------- ------------------------------ ------------------------------- ----------------------- ---------- --------- ---------- ---------- --------- ---------- **Expressive** Placebo 12(41.4) 8(27.6) 9(31.0) 15(51.7) 3(10.3) 11(37.9) 16(55.2) 3(10.3) 10(34.5) AUCB 14(44.8) 7(24.1) 8(27.6) 15(51.7) 4(13.8) 10(34.5) 16(55.2) 5(17.2) 8(27.6) **Receptive** Placebo 12(41.4 8(27.6 9(31.0) 14(48.3) 5(17.2) 10(34.5) 17(58.6) 5(17.2) 7(24.1) AUCB 13(44.8) 5(17.2) 11(37.9) 17(58.6) 2(6.9) 10(34.5) 18(62.1) 3(10.3) 8(27.6) **Social** Placebo 12(41.4) 9(31.0) 8(27.6) 14(48.3) 3(10.3 12(41.4 16(55.2) 2(6.9) 11(37.9) AUCB 13(44.8) 6(20.7) 10(34.5) 16(55.2) 1(3.4) 12(41.4) 18(62.1) 3(10.3) 8(27.6) ----------------------------------------------------------------------------------------------------------------------------------------------------------------------- Abbreviation: AUCB, autologous umbilical cord blood Primary and Secondary Endpoints {#sct312282-sec-0017} ------------------------------- Figure [2](#sct312282-fig-0002){ref-type="fig"}A (EOWPVT) and 2B (ROWPVT) show the means for each group throughout their time in the study. Figure [3](#sct312282-fig-0003){ref-type="fig"}A through 3E display the means for the EOWPVT (Panel A), ROWPVT (Panel B) Stanford‐Binet Knowledge (Panel C), Stanford‐Binet Fluid Reasoning (Panel D), and Vineland Socialization (Panel E) over the study for all 29 subjects under both conditions. There were no statistically significant carryover effects for these endpoints. This allowed participants to be used as their own control and mean differences for all 29 subjects compared after infusions of AUCB and placebo at 12 and 24 weeks versus baseline. {#sct312282-fig-0002} ![Mean Expressive One Word Picture Vocabulary Test **(A)**, Receptive One Word Picture Vocabulary Test **(B)**, Stanford Binet‐Knowledge **(C)**, and Stanford Binet Fluid Reasoning subscales **(D)** and Vineland Adaptive Behavior Scale for Socialization **(E)** at baseline, 12 weeks, and 24 weeks for all 29 subjects under each condition. Vineland Adaptive Behavior Scale for Socialization: *F* ~1,27~ = 5.08; *p* = .032 (false discovery rate \[FDR\] adjusted *p*‐value = .025) after 12 weeks; *F* ~1,27~ =1.95; *p* = .174 (FDR adjusted *p*‐value = .045) after 24 weeks.](SCT3-7-333-g003){#sct312282-fig-0003} There were no statistically significant changes in the EOWPVT (Fig. [3](#sct312282-fig-0003){ref-type="fig"}A), ROWPVT (Fig. [3](#sct312282-fig-0003){ref-type="fig"}B), or Stanford Binet subscales (Fig. [3](#sct312282-fig-0003){ref-type="fig"}C, [3](#sct312282-fig-0003){ref-type="fig"}D) when subjects received placebo versus AUCB at 12 and 24 weeks post infusion. The score on the Vineland Socialization subscale was significant as scored per guidelines after 12 weeks for the AUCB group, but after applying the false discovery method for adjustment of multiple comparisons the scoring did not reach significance (Fig. [3](#sct312282-fig-0003){ref-type="fig"}E). There were statistically significant (*p* \< .05) group X treatment interactions for the Vineland ABC, Communication, Motor, and Daily subscales which indicated that changes over time between these scores depended on the order that AUCB was administered. For these endpoints, the means scores between the 14 participants who received AUCB first and 15 participants who received placebo first were compared at 12 and 24 weeks and adjusted for differences in baseline scores. The results of these ANCOVAs (Table [4](#sct312282-tbl-0004){ref-type="table-wrap"}) showed that all Vineland subscales had lower mean scores for the AUCB group than the placebo group at both 12 and 24 weeks. None of these differences were statistically significant after adjustment for baseline scores and using the FDR adjustment for multiplicity. ###### Results of analysis of covariance comparing 12‐ and 24‐week placebo and autologous umbilical cord blood (AUBC) scores when adjusted for baseline scores for Vineland subscales with a statistically significant order by treatment interaction. Means ± standard deviations (SD) represent the 15 cases receiving placebo first and the 14 cases receiving AUCB first Baseline Mean ± SD 12 weeks Mean ± SD *p*‐value FDR adjusted *p*‐value 24 weeks Mean ± SD *p*‐value FDR adjusted *p*‐value ------------------- -------------------- -------------------- ----------- ------------------------ -------------------- ----------- ------------------------ **ABC** Placebo 74.27 ± 16.00 74.93 ± 12.40 .025 .015 75.00 ± 11.60 .015 .005 Cord blood 66.00 ± 12.20 66.15 ± 15.05 67.57 ± 16.52 **Communication** Placebo 79.60 ± 17.18 81.47 ± 14.63 .045 .035 83.33 ± 15.46 .044 .030 Cord blood 66.00 ± 15.56 67.79 ± 16.21 68.00 ± 18.82 **Motor** Placebo 77.67 ± 13.57 79.57 ± 11.51 .029 .020 76.67 ± 7.61 .022 .010 Cord blood 73.86 ± 13.17 73.50 ± 15.09 69.71 ± 15.10 **Daily** Placebo 77.13 ± 16.61 78.13 ± 14.21 .203 .050 77.33 ± 15.30 .097 .040 Cord blood 68.14 ± 15.55 68.13 ± 15.32 70.36 ± 17.73 Abbreviation: FDR, false discovery rate. Discussion {#sct312282-sec-0018} ========== This pilot study is the first randomized, double‐blinded, placebo‐controlled trial performed in the U.S. to assess the feasibility of treating children with ASD with AUCB. The results of this study showed that AUCB infusions are safe for this population. There were no reported serious adverse events, similar to other studies of AUCB infusions [39](#sct312282-bib-0039){ref-type="ref"}, [40](#sct312282-bib-0040){ref-type="ref"}, [41](#sct312282-bib-0041){ref-type="ref"}, [47](#sct312282-bib-0047){ref-type="ref"}. There were also no statistically significant differences between scores on the two primary or secondary endpoints after infusion with AUCB versus infusion of placebo. The present study provides further evidence that treatment with AUCB is safe but there was minimal evidence of clinical effectiveness. Although the lack of significant results of this study are in contrast to the findings of an open‐label trial conducted by Dawson et al. [47](#sct312282-bib-0047){ref-type="ref"} who reported improvements in behavior on parent reports of social communication skills, clinician ratings of autism symptoms, and standardized measures of expressive vocabulary in children with ASD after the infusion of AUCB, parents and investigators observed similar subjective improvements. While those findings show promise for AUCB as a potential treatment for ASD, the lack of both blinding and a control group do not allow for definitive conclusions. Both studies assessed change from baseline at 12 and 24 weeks and used some similar endpoints. However, the present study incorporated both blinding and a placebo treatment to minimize the potential for a placebo effect that may occur in open label studies and used standard scores for both objective and subjective endpoints. While raw scores may show more variation, standard scores reflect what is appropriate for children of a specific age and allow conversion of raw scores into meaningful expression of performance relative to age expectations. Language testing in the ASD population is difficult due to often atypical developmental course. Use of raw scores or use of language tests other than the EOWPVT and ROWPVT could have potentially resulted in more precise measurement of function in these areas, specifying gain, stability, or loss. For example, with the current measures, standard scores above the 95th percentile limit the capture of gains in post‐treatment scores (ceiling effect). This, however, was not contributing factor in the current analysis. Standard scores below the 2nd percentile occurred frequently in this investigation and potentially limit capture of function (loss, stability, minor improvements) as well. While there is little room to measure loss, even opportunity for growth can be misleading. For example, stability in raw scores in each scenario (no loss of skill) could result in decreased standard score even with a 3‐month advance in age. In fact, minor improvements could translate into stability or decrease in standard scores upon conversion if the rate of growth is less than anticipated by the standardized data for demographic age‐expectations. This also pertains to a lesser degree to the Vineland and the Stanford‐Binet tests. The subjective measurement of CGI is commonly used in studies of ASD in part due to these challenges. Although not to be interpreted in isolation, the measurable change in CGI scores may be more meaningful than standardized scores. In addition to the difficulty in measuring language in this population, there are other aspects of the present investigation that could have limited the ability to detect significant findings. These include a crossover design and lack of standardization of AUCB dose. The study was designed as a crossover study. Repeated measures designs allow for greater power with fewer subjects than parallel groups trials. One reason for the choice of the crossover design was that there are few treatments for ASD so parents would be less likely to want to enroll children in a trial with a chance of only receiving a placebo, or in a trial with a wait list option where subjects randomized to placebo could potentially be aged out of the trial while waiting to be given AUCB. The design choice resulted in excellent compliance by parents so that complete data were available for all visits, on all endpoints, for 29 subjects. The disadvantage of this design is the potential for carryover of a treatment which means that only the first sequence of treatments can be used for analysis. The present study showed no carryover effect for the two primary endpoints but there were carryover effects for several secondary endpoints. There is the potential that 12 weeks was not a sufficient amount of time to detect potential changes. However, Figure [2](#sct312282-fig-0002){ref-type="fig"}A and [2](#sct312282-fig-0002){ref-type="fig"}B, which show the change over time for the 14 participants who received AUCB first and were followed for the full 24 weeks, suggest that this is likely not the reason for lack of effect. The other reason for the choice of the crossover design is that it allows for subjects to serve as their own control. The ASD population has widely variable levels of functioning and this was seen in the baseline differences in the endpoint scores between the groups who received AUCB first versus those who received placebo first. This is a potential problem in all randomized studies of autism and was one of the advantages of the crossover design where subjects are used as their own control. Perhaps stricter control on the level of initial baseline function required for study inclusion would have led to less variability among subjects and increased the ability for differences to be detected and is suggested for a future investigation. The advantage of strict inclusion criteria is better control and less variability but the disadvantage is that the results become generalizable to a smaller subset of the ASD population. Dosage of AUCB may be a reason for lack of evidence of efficacy. Although all samples met the minimum requirement of a post thaw percent viable CD34+ of 20%, participants varied widely in percentage and number of CD34+ cells in samples infused. Furthermore, 27% of participants received 1/5th of their available AUCB based on parental choice to only use the minor compartment. Parents were reticent to use the entire banked sample on an investigational treatment but an analysis of dose response would be valuable in future studies. Despite the limitations of the study, a core symptom of autism, socialization, showed trends in improvement on the Socialization Subscale of the Vineland. The strengths of this study were successful maintenance of blinding and excellent compliance by parents attending all follow up visits despite some having to travel long distances over several months. We recognize that our results do not corroborate with the results of Dawson et al. [47](#sct312282-bib-0047){ref-type="ref"}. Such discrepant results between investigations of comparable design warrant careful consideration of the variables that potentially contributed to outcomes. Specific attention should be given to blind versus open designs. Lack of significant measurable benefit was observed only in the blinded scenario while significant change of any kind was observed only in the open scenario. There may be significant potential benefit, even if it occurs within a sub‐population of the larger ASD phenotype. As we begin to explore possible benefits from these new technologies we would benefit from not taking a position of controversy, but rather of collaboration and appreciation that these two investigations were vastly different studies. Dawson et al. [47](#sct312282-bib-0047){ref-type="ref"} may well have discovered something that we missed, but any future studies should consider implementing a blinded investigation with a control group. We might also lack, at this time, tools that effectively capture real change in this population. Measurement of behavior is inherently more subjective than biometrics. The goal of this pilot study was to assess safety and design the first U.S. based study of AUCB for treatment of ASD. This study demonstrates that infusion of AUCB in individuals with ASD is safe and feasible when studied in a rigorous trial design and revealed challenges that need to be addressed to advance the study of the AUCB. Conclusion {#sct312282-sec-0019} ========== Infusion of AUCB for children with ASD is safe but efficacy has yet to be determined. Tightly controlled trials are necessary to further progress the study of AUCB for autism. Author Contributions {#sct312282-sec-0022} ==================== M.C.: conception and design, collection of data, manuscript writing, final approval of manuscript; C.L.: conception and design, collection of data, final approval of manuscript; C.P.: conception and design, data analysis and interpretation, manuscript writing, final approval of manuscript; A.D.‐C.: collection of data; A.H.: data analysis and interpretation, manuscript writing; M.C.: conception and design, collection of data. Disclosure of Potential Conflicts of Interest {#sct312282-sec-0021} ============================================= M.C. has received consulting and speaking fees from CBR Systems, Inc, the source of funding of the research grant. The other authors indicated no potential conflicts of interest. This study was funded by CBR Systems, Inc.

Breeding Science 69: 186--188 (2019) In the above article, sample names on an electrophoresis gel in Fig. 2 were wrongly placed. Therefore, we herewith would like to publish the correct arrangement. **False:** Lanes were mislabeled. **True:** Lane two should be MS and lane three should be F. There is no change in figure legends. Corrected figure arrangement is below. {#f1-69_18120e} The online version of the original article can be found at <https://www.jstage.jst.go.jp/article/jsbbs/69/1/69_18120/_pdf>

Binyamin O, Larush L, Frid K, et al. *Int J Nanomedicine* 2015;10:7165--7174. On page 7165, the first affiliation and correspondence details were incorrectly written as "Department of Neurology, The Agnes Ginges Center of Human Neurogenetics, Hadassah University Hospital". It should be "Department of Neurology, The Agnes Ginges Center of Human Neurogenetics, Hebrew University-Hadassah Medical Center".

Molina‐Romero M, Gómez PA, Sperl JI, et al. A diffusion model‐free framework with echo time dependence for free‐water elimination and brain tissue microstructure characterization. Magn Reson Med. 2018;80:2155--2172. 10.1002/mrm.27181 29573009 **Funding information** German Excellence Initiative; European Commission, Grant/Award Number: 605162; New Investigator Award from the Wellcome Trust, Grant/Award Number: 096646/Z/11/Z; Wellcome Trust Strategic Award (104943/Z/14/Z) The copyright line for this article was changed on 28 September 2019 after original online publication. 1. INTRODUCTION {#mrm27181-sec-0005} =============== More than 50 years have passed since Stejskal and Tanner published their early research on pulsed gradient spin‐echo.[1](#mrm27181-bib-0001){ref-type="ref"} Thereafter, diffusion weighted imaging became an essential tool for nondestructive tissue microstructure characterization. The pioneering studies on ex vivo tissue and simulations of Krägger,[2](#mrm27181-bib-0002){ref-type="ref"} Latour et al.,[3](#mrm27181-bib-0003){ref-type="ref"} Szafer et al.,[4](#mrm27181-bib-0004){ref-type="ref"} and Stanisz et al.[5](#mrm27181-bib-0005){ref-type="ref"} established the theoretical basis of the compartmental model of neural tissue. These early contributions were later translated to target specific biomarkers for in vivo human studies. White matter (WM) anisotropy became fiber orientation with the introduction of diffusion tensor imaging (DTI).[6](#mrm27181-bib-0006){ref-type="ref"} The composite hindered and restricted model of diffusion MR imaging (CHARMED)[7](#mrm27181-bib-0007){ref-type="ref"} extended DTI to two compartments with restricted and hindered diffusion behavior. Using the same principles, the neurite orientation dispersion and density imaging (NODDI) model[8](#mrm27181-bib-0008){ref-type="ref"} introduced fiber orientation dispersion metrics and added an isotropic compartment. Additionally, axon diameter was addressed by AxCaliber[9](#mrm27181-bib-0009){ref-type="ref"} and ActiveAx.[10](#mrm27181-bib-0010){ref-type="ref"} These and other approaches rely on diffusion signal representations or a variety of geometric biophysical assumptions about the underlying tissue compartments, producing a wide range of possible configurations.[11](#mrm27181-bib-0011){ref-type="ref"} In parallel with the development of multicomponent diffusion tissue models, relaxometry addressed the compartmental nature of tissue microstructure from a different perspective.[12](#mrm27181-bib-0012){ref-type="ref"} Multi‐echo spin echo (SE) experiments combined with regularized inverse Laplace transforms (ILTs) for multi‐exponential fitting showed the presence of multiple water compartments in the tissue. Nonnegative least squares (NNLS)[13](#mrm27181-bib-0013){ref-type="ref"} is the current gold standard for computing a regularized discrete ILTs for several components.[14](#mrm27181-bib-0014){ref-type="ref"}, [15](#mrm27181-bib-0015){ref-type="ref"} Alternatively, the exponential analysis via system identification using Steiglitz‐McBride (EASI‐SM) for multicomponent estimation was introduced by Stoika et al.[16](#mrm27181-bib-0016){ref-type="ref"}, [17](#mrm27181-bib-0017){ref-type="ref"} Additionally, mcDESPOT,[18](#mrm27181-bib-0018){ref-type="ref"} used a spoiled gradient‐recalled echo and a balanced steady‐state free precession to yield relaxation, volume fraction, and water exchange parameters for three compartments. Nevertheless, the paths of diffusion MRI and MR relaxometry have become entangled over the years. Studies on ex vivo nerves with a diffusion‐weighted Carr‐Purcell‐Meiboom‐Gill (CPMG) sequence[19](#mrm27181-bib-0019){ref-type="ref"}, [20](#mrm27181-bib-0020){ref-type="ref"} showed the relationship that existed between compartmental *T* ~2~ decay and diffusivity. However, diffusion‐weighted CPMG experiments need long acquisition times and high specific absorption rates, which makes them unsuitable for human in vivo studies. Typically, two‐dimensional ILTs were used to fit the data, but this approach is highly ill‐posed and requires large amounts of data for stabilization. Recently, Benjamini et al.[21](#mrm27181-bib-0021){ref-type="ref"} introduced the marginal distributions constrained optimization (MADCO), a nonCPMG compressed‐sensing based solution that reduced the amount of data necessary for NMR diffusion--relaxation correlation experiments. Kim et al. translated diffusion--relaxation correlation spectroscopy (DR‐COSY)[22](#mrm27181-bib-0022){ref-type="ref"}, [23](#mrm27181-bib-0023){ref-type="ref"} into imaging (DR‐CSI)[24](#mrm27181-bib-0024){ref-type="ref"} using spatial regularization to reduce the amount of necessary data and stabilize the ILTs. However, they require specific diffusion protocols with increasing *b*‐values along a unique diffusion direction and repeated echoes or inversion times. Other alternatives combine diffusion models with multicompartmental relaxation. For instance, inversion recovery diffusion weighted imaging has been used to identify fiber populations,[25](#mrm27181-bib-0025){ref-type="ref"}, [26](#mrm27181-bib-0026){ref-type="ref"} and WM integrity has been characterized using the axonal stick model and multiple echo times (TE).[27](#mrm27181-bib-0027){ref-type="ref"} Compartmental analysis of the diffusion signal is intimately related to a recurring issue: cerebrospinal fluid (CSF) or free‐water contamination.[28](#mrm27181-bib-0028){ref-type="ref"}, [29](#mrm27181-bib-0029){ref-type="ref"} All the existing contributions agree on using a bi‐tensor signal model: parenchyma and CSF. However, this is an ill‐posed problem for a single‐shell and ill‐conditioned for multiple‐shell acquisitions.[30](#mrm27181-bib-0030){ref-type="ref"} Spatial regularization was proposed by Pasternak et al.,[31](#mrm27181-bib-0031){ref-type="ref"} relying on the local smoothness of the diffusion tensor. Later, a protocol optimization for multiple shells was presented by Hoy et al.,[32](#mrm27181-bib-0032){ref-type="ref"} eliminating such a constraint. Other solutions regularize the problem by adding priors[33](#mrm27181-bib-0033){ref-type="ref"} or finding the best fit to the model.[34](#mrm27181-bib-0034){ref-type="ref"} Nevertheless, the CSF contribution to the diffusion signal depends on the TE. Thus, disentangling the tissue CSF volume fraction requires an approach that includes *T* ~2~ compartmental dependencies.[33](#mrm27181-bib-0033){ref-type="ref"}, [35](#mrm27181-bib-0035){ref-type="ref"}, [36](#mrm27181-bib-0036){ref-type="ref"} We propose a general framework for studying diffusion and relaxation characteristics in tissue microstructures. We call it general because it does not model the compartmental diffusion behavior. It replaces the ILTs by a blind source separation (BSS) technique, reducing the minimum number of distinct echo times required to the number of compartments in the tissue, less than for ILTs‐based methods. Other than the requirement to measure at more than one echo time, this framework is diffusion protocol‐agnostic, and can be used in combination with any protocol of interest. Our approach quantifies proton density, compartmental volume fractions, and transverse relaxation times. Importantly, it handles diffusion signals from each compartment independently, allowing for individual analyses, and thus performs CSF partial volume correction as a direct application. 2. THEORY {#mrm27181-sec-0006} ========= Following the Bloch--Torrey equation, we describe the diffusion signal as a weighted sum of the signals from the compartments comprising the tissue $$X(\text{TE},b,\mathbf{g}) = S_{0}{\sum\limits_{i = 1}^{M}f_{i}}e^{- \frac{\text{TE}}{T_{2_{i}}}}S_{i}(b,\mathbf{g}).$$ Where *b* summarizes the gradient effects[1](#mrm27181-bib-0001){ref-type="ref"}, [37](#mrm27181-bib-0037){ref-type="ref"} and **g** defines the gradient directions. Here, the compartmental diffusion sources $S_{i}(b,\mathbf{g})$ are weighted by their volume fraction, *f~i~*, TE, and $T_{2_{i}}$. The exponent (the ratio between TE and $T_{2_{i}}$) scales the contribution of each compartment to the acquired signal. Therefore, measuring at different TEs produces distinct diffusion signals[38](#mrm27181-bib-0038){ref-type="ref"} with different weights from the compartmental signal sources. As a result, the signal of a single voxel measured with a protocol that accounts for multiple echoes can be formulated as $$\begin{bmatrix} {X_{1}(\text{TE}_{1},b,\mathbf{g})} \\ \vdots \\ {X_{N}(\text{TE}_{N},b,\mathbf{g})} \\ \end{bmatrix} = S_{0}\begin{bmatrix} {f_{1}e^{\frac{- \text{TE}_{1}}{T_{2_{1}}}}} & \cdots & {f_{M}e^{\frac{- \text{TE}_{1}}{T_{2_{M}}}}} \\ \vdots & \ddots & \vdots \\ {f_{1}e^{\frac{- \text{TE}_{N}}{T_{2_{1}}}}} & \cdots & {f_{M}e^{\frac{- \text{TE}_{N}}{T_{2_{M}}}}} \\ \end{bmatrix}\begin{bmatrix} {S_{1}(b,\mathbf{g})} \\ \vdots \\ {S_{M}(b,\mathbf{g})} \\ \end{bmatrix},$$where *X~j~* (*j* $\in$ $\lbrack 1,N\rbrack$) are the diffusion signals acquired for the *N* TEs. *f~i~* and $T_{2_{i}}$ (*i* $\in$ $\lbrack 1,M\rbrack$) are the volume fraction and *T* ~2~ decay for the *i*th compartment, respectively, and *M* is the number of compartments. Equation [(2)](#mrm27181-disp-0002){ref-type="disp-formula"} can be expressed in matrix form as **X = AS**. This is a matrix factorization of the measurements, $\mathbf{X} \in \mathbb{R}_{\geq 0}^{N \times n}$, into two new matrices: the mixing matrix, $\mathbf{A} \in \mathbb{R}_{\geq 0}^{N \times M}$, which is defined by the experimental TEs, the compartmental volume fractions *f*, and *T* ~2~ decays; and the sources matrix, $\mathbf{S} \in \mathbb{R}_{\geq 0}^{M \times n}$, representing the diffusion sources in each sub‐voxel compartment. Interestingly, we noticed from the definition of **A** that the ratio between the experimental TEs and $T_{2_{i}}$ determines the direction (or slope for *N* = 2) of the *i*th column vector of the mixing matrix. Therefore: $$T_{2_{i}} = \frac{\text{TE}_{k} - \text{TE}_{l}}{\log\left( \frac{a_{li}}{a_{ki}} \right)},$$where TE~*k*~ \< TE~*l*~, and *a~ki~* and *a~li~* are the *k*th and *l*th elements of the *i*th column of the mixing matrix, respectively. Additionally, diffusion is an attenuation contrast and as such, $S(b = 0) = 1$, allowing Equation [(2)](#mrm27181-disp-0002){ref-type="disp-formula"} to be rewritten as $$\begin{bmatrix} {X_{1}(\text{TE}_{1},b = 0,\mathbf{g})} \\ \vdots \\ {X_{N}(\text{TE}_{N},b = 0,\mathbf{g})} \\ \end{bmatrix} = S_{0}\begin{bmatrix} e^{\frac{- \text{TE}_{1}}{T_{2_{1}}}} & \cdots & e^{\frac{- \text{TE}_{1}}{T_{2_{M}}}} \\ \vdots & \ddots & \vdots \\ e^{\frac{- \text{TE}_{N}}{T_{2_{1}}}} & \cdots & e^{\frac{- \text{TE}_{N}}{T_{2_{M}}}} \\ \end{bmatrix}\begin{bmatrix} f_{1} \\ \vdots \\ f_{M} \\ \end{bmatrix},$$which, together with ${\sum\limits_{i = 1}^{M}f_{i}} = 1$, allows us to solve for the volume fractions and proton density (*f~i~* and *S* ~0~) when the number of measurements matches the number of compartments (*M* = *N*). Contrary, when there are more compartments than measurements (*M* \> *N*), Equation [(4)](#mrm27181-disp-0004){ref-type="disp-formula"} is undetermined and *f~i~* and *S* ~0~ cannot be estimated. Factorizing **X** into **A** and **S** is known as BSS[39](#mrm27181-bib-0039){ref-type="ref"} of mixed measurements into their generating sources (Figure [1](#mrm27181-fig-0001){ref-type="fig"}). For BSS to identify these sources, they have to be distinct: $S_{i} \neq S_{j}$ $\forall$ $i \neq j$. Therefore, based on previous work,[19](#mrm27181-bib-0019){ref-type="ref"}, [20](#mrm27181-bib-0020){ref-type="ref"} we assumed them to be different. {ref-type="disp-formula"} and [(4)](#mrm27181-disp-0004){ref-type="disp-formula"}. We showed an exponential case for simplicity, but BSS is not limited to this choice; any signal can be processed in the same manner](MRM-80-2155-g001){#mrm27181-fig-0001} There are four main approaches to BSS: principal component analysis,[40](#mrm27181-bib-0040){ref-type="ref"} independent component analysis,[41](#mrm27181-bib-0041){ref-type="ref"} nonnegative matrix factorization (NMF),[42](#mrm27181-bib-0042){ref-type="ref"} and sparse component analysis.[43](#mrm27181-bib-0043){ref-type="ref"} Principal component analysis is not an applicable solution for this problem because the diffusion sources are not orthogonal. Independent component analysis assumes, as prior knowledge, that the signal sources are statistically independent and have nonGaussian distributions. However, diffusion MRI signals are correlated with the tissue structure and temperature and they present nonGaussian distributions only in restricted compartments, meaning that independent component analysis is not suitable either. We previously explored sparse component analysis[44](#mrm27181-bib-0044){ref-type="ref"} and found that even though the results for simulations and real data for specific diffusion protocols were encouraging, finding a sparse and disjoint domain to meet the method\'s requirements was not always possible for arbitrary protocols. We observed the same issue for a version of NMF that enforces sparsity similarly.[36](#mrm27181-bib-0036){ref-type="ref"} In the present work, we took a BSS approach based on NMF (assuming **X**, **A**, and **S** are nonnegative). Instead of depending on sparsity, we used a popular NMF solver: the alternating least squares algorithm (ALS).[42](#mrm27181-bib-0042){ref-type="ref"}, [45](#mrm27181-bib-0045){ref-type="ref"}, [46](#mrm27181-bib-0046){ref-type="ref"} We chose ALS instead of the multiplicative update algorithm[47](#mrm27181-bib-0047){ref-type="ref"} due to its faster convergence.[48](#mrm27181-bib-0048){ref-type="ref"} We extended ALS to account for physically plausible limitations, resulting in Algorithm 1, which we refer to as constrained ALS (cALS). Compartmental *T* ~2~ values available from the literature[15](#mrm27181-bib-0015){ref-type="ref"} allowed us to limit the solution space of the columns of **A** (Equation [(3)](#mrm27181-disp-0003){ref-type="disp-formula"}). Additionally, for in vivo data, the diffusion behavior of CSF is known to be approximately isotropic with $3 \times 10^{- 3}$ mm^2^/s diffusivity,[28](#mrm27181-bib-0028){ref-type="ref"} adding extra prior information. These constraints and priors make cALS converge toward physically realistic solutions (Figure [1](#mrm27181-fig-0001){ref-type="fig"}). **Algorithm 1** Constrained Alternating Least Squares (cALS) Constrained ALS initializes the column vectors of **A** at the central *T* ~2~ of their given constraints, avoiding random initializations in regions that are not physically feasible and increasing the stability. After each iteration, cALS verifies that the resulting *T* ~2~ of each column vector is between its boundaries, and sets it back to the center of its constrained solution space otherwise. Following the factorization of **A**, we estimated *T* ~2~ and *f* for each compartment, (Equations [(3)](#mrm27181-disp-0003){ref-type="disp-formula"} and [(4)](#mrm27181-disp-0004){ref-type="disp-formula"}), and recalculated the real **A**. This is important since the column norms of the factorized **A** do not tell us about the volume fractions. Then, **S = A^−1^X** is calculated. An iterative algorithm like cALS inverts **A** repeatedly, requiring it to be nonsingular and introducing a new condition. From Equation [(2)](#mrm27181-disp-0002){ref-type="disp-formula"}, **A** is nonsingular when $T_{2_{i}} \neq T_{2_{j}}$ $\forall$ $i \neq j$. Hence, in accordance with the literature,[19](#mrm27181-bib-0019){ref-type="ref"}, [20](#mrm27181-bib-0020){ref-type="ref"} we assumed that the transverse relaxation times for each compartment were distinct. An open source implementation can be found in <https://github.com/mmromero/dwybss>. 3. METHODS {#mrm27181-sec-0007} ========== 3.1. Simulations {#mrm27181-sec-0008} ---------------- NMF is known for converging to local minima.[45](#mrm27181-bib-0045){ref-type="ref"} Thus, it is necessary to assess the impact of the constraints. We ran simulations with Rician noise for signal‐to‐noise ratio (SNR) levels of 50, 100, and 150 at the nondiffusion weighted volume and minimum TE. We accounted for *T* ~2~ values, volume fractions, and diffusivities supported by literature.[15](#mrm27181-bib-0015){ref-type="ref"}, [28](#mrm27181-bib-0028){ref-type="ref"} 3.2. Two compartments {#mrm27181-sec-0009} --------------------- Two compartments were simulated mimicking intra/extra‐axonal (IE) and CSF water. The diffusion protocol included one nondiffusion weighted volume and 30 directions. We modeled diffusion as a Gaussian process (see Supporting Information Figure S4). For all the simulations we used $T_{2_{\text{CSF}}}$ = 2000 ms, and varied $T_{2_{\text{IE}}}$ from 50 to 150 ms in 30 increments.[15](#mrm27181-bib-0015){ref-type="ref"} Values of *f* ~IE~ = 0.25, 0.5 and, 0.75 were used. We fixed TE~1~ = 60 ms, and explored TE~2~ from 70 to 150 ms in 31 increments. We defined ΔTE = TE~2~ − TE~1~. The performance of the cALS algorithm was tested under the following conditions: **Overlapped *T*~2~ constraints**: $T_{2_{\text{IE}}}$ and $T_{2_{\text{CSF}}}$ were bounded from 0--1000 and 0--3000 ms, respectively, and no assumption on *S* ~CSF~ was made (Figure [2](#mrm27181-fig-0002){ref-type="fig"} and Supporting Information Figure S5).Figure 2Convergence for two compartments (IE and CSF) with overlapping *T* ~2~ constraints and no *S* ~CSF~ prior (SNR = 50). The mean of *f* ~IE~ absolute error and its standard error (SEM) (A, B), and the mean of $T_{2_{\text{IE}}}$ (C) and $T_{2_{\text{CSF}}}$ (E) relative errors per unit (p.u.), and their standard error (D, F). Red and white lines mark the 0.2 and 0.1 contour respectively. One thousand simulations were run for each combination of *f* ~IE~, $T_{2_{\text{IE}}}$, and ΔTE. $T_{2_{\text{IE}}}$ and $T_{2_{\text{CSF}}}$ were bounded between 0--1000 ms and 0--3000 ms respectively, and no prior was imposed on *S* ~CSF~. We defined the convergence area as the one with error lower than 0.1 for *f* ~IE~ and $T_{2_{\text{IE}}}$. The bias of *f* ~IE~ and $T_{2_{\text{IE}}}$ decreases for long ΔTEs as *f* ~IE~ increases. See Supporting Information Figure S5 for more SNR levels**Overlapped *T*~2~ constraints and prior *S*~CSF~**: $T_{2_{\text{IE}}}$ and $T_{2_{\text{CSF}}}$ were bounded from 0--1000 and 0--3000 ms, respectively. CSF diffusivity was assumed to be isotropic with value 3 × 10^−3^ mm^2^/s (Supporting Information Figure S10).**Separated *T*~2~ constraints**: $T_{2_{\text{IE}}}$ and $T_{2_{\text{CSF}}}$ were bounded from 0--300 and 300--3000, ms, respectively, and no assumption on *S* ~CSF~ was made (Supporting Information Figure S11).**Separated *T*~2~ and prior *S~CSF~***: $T_{2_{\text{IE}}}$ and $T_{2_{\text{CSF}}}$ were bounded from 0--300 and 300--3000 ms, respectively. CSF diffusivity was assumed to be isotropic with value 3 × 10^−3^ mm^2^/s (Supporting Information Figure S13).**Fixed** $T_{2_{\text{CSF}}}$: $T_{2_{\text{IE}}}$ was bounded from 0 to 300 ms. $T_{2_{\text{CSF}}}$ was fixed to 2000 ms. No assumption on *S* ~CSF~ was made (Supporting Information Figure S12).**Fixed** $T_{2_{\text{CSF}}}$ **and prior *S*~CSF~**: $T_{2_{\text{IE}}}$ was bounded from 0 to 300 ms. $T_{2_{\text{CSF}}}$ was fixed to 2000 ms. CSF diffusivity was assumed to be isotropic with value 3 × 10^−3^ mm^2^/s (Figure [3](#mrm27181-fig-0003){ref-type="fig"} and Supporting Information Figure S6).Figure 3Convergence for two compartments (IE and CSF) with nonoverlapping *T* ~2~ constraints and *S* ~CSF~ prior (SNR = 50). The mean of *f* ~IE~ absolute error and its standard error (SEM) (A, B), and the mean of $T_{2_{\text{IE}}}$ (C) and $T_{2_{\text{CSF}}}$ (E) relative error per unit (p.u.), and their standard errors (D, F). Red and white lines mark the 0.2 and 0.1 contour respectively. One thousand simulations were run for each combination of *f* ~IE~, $T_{2_{\text{IE}}}$, and ΔTE. $T_{2_{\text{IE}}}$ and $T_{2_{\text{CSF}}}$ were bounded between 0--300 ms and 2000 ms, respectively, and *S* ~CSF~ was set to have isotropic diffusivity with value 3 $\times 10^{- 3}$ mm^2^/s. We defined the convergence area as the one with error lower than 0.1 for *f* ~IE~ and $T_{2_{\text{IE}}}$. This area is larger than for Figure [2](#mrm27181-fig-0002){ref-type="fig"} stressing the importance of priors. See Supporting Information Figure S6 for more SNR levels We repeated the last simulation for values of *f* ~IE~ = 0 and 1, accounting only for IE or CSF (Figure [4](#mrm27181-fig-0004){ref-type="fig"} and Supporting Information Figure S7). {#mrm27181-fig-0004} Finally, intra‐cellular (IC) and extra‐cellular (EC) *T* ~2~ values are similar.[15](#mrm27181-bib-0015){ref-type="ref"} We assessed the potential of BSS to separate them. Two diffusion signals were generated (see Supporting Information Figure S14). We used *f* ~IC~ = 0.25, 0.5, and 0.75. The $T_{2_{\text{IC}}}$ vales ranged from 50 to 90 ms in 30 increments, and $T_{2_{\text{EC}}}$ = 100 ms. TE~1~ was fixed to 60 ms and TE~2~ was varied between 70 and 150 ms in 31 increments. No assumption was made on the diffusion signals, and *T* ~2~ constraints were defined between 0--150 and 0--200 ms for IC and EC, respectively (Figure [5](#mrm27181-fig-0005){ref-type="fig"} and Supporting Information Figure S8). {ref-type="fig"}, Supporting Information Figures S5, S10) and the condition of **A** when the *T* ~2~ values are similar. See Supporting Information Figure S8 for more SNR levels](MRM-80-2155-g005){#mrm27181-fig-0005} We simulated 1000 times each combination of parameters, and reported the mean value of the absolute error of *f*, the relative error of *T* ~2~, and their standard errors (SEM). 3.3. Three compartments: Searching for myelin {#mrm27181-sec-0010} --------------------------------------------- We incorporated a fast decaying component to model myelin, and fixed the *T* ~2~ of myelin ( $T_{2_{M}}$) to 15 ms.[15](#mrm27181-bib-0015){ref-type="ref"} $T_{2_{\text{IE}}}$ was varied from 50 to 150 ms in 30 increments, and $T_{2_{\text{CSF}}}$ = 2000 ms. To account for short T2 components we needed to reduce the minimum TE of our simulations (see phantom experiments in the supporting information). Therefore, we fixed TE~1~ = 10 ms, TE~3~ = 150 ms, and varied TE~2~ from 20 to 140 ms in 31 increments. We defined ΔTE = TE~2~ − TE~1~. Three cases were explored: (1) *f* ~M~ = 0.1, *f* ~IE~ = 0.6; (2) *f* ~M~ = 0.2, *f* ~IE~ = 0.5; and (3) *f* ~M~ = 0.3, *f* ~IE~ = 0.4; keeping *f* ~CSF~ = 0.3 for all of them. Simulations were run for two cases: **Overlapped *T*~2~ constraints**: $T_{2_{M}},\, T_{2_{\text{IE}}}$, and $T_{2_{\text{CSF}}}$ were bounded from 0--40, 0--300, and 0--3000 ms, respectively. No assumption on *S* ~CSF~ was made.**Separated *T*~2~ constraints, fixed** $T_{2_{\text{CSF}}}$ **and prior *S~CSF~***: $T_{2_{M}}$ and $T_{2_{\text{IE}}}$ were bounded from 0--40 and 41--300 ms, respectively, while $T_{2_{\text{CSF}}}$ = 2000 ms. CSF diffusivity was assumed to be isotropic with value 3 × 10^−3^ mm^2^/s (Figure [6](#mrm27181-fig-0006){ref-type="fig"} and Supporting Information Figure S9).Figure 6Convergence for three compartments (myelin, IE, and CSF) with nonoverlapping *T* ~2~ constraints and *S* ~CSF~ prior (SNR = 50). The mean absolute errors of the volume fraction estimates and their standard errors (SEM) (A--D); and the mean of $T_{2_{M}}$ (E) and $T_{2_{\text{IE}}}$ (G) relative error per unit (p.u.), and their standard errors (F, H). Red and white lines mark the 0.2 and 0.1 contour respectively. There is a large convergence area when TE~1~ = 10 ms, TE~2~ = 46 ms, and TE~3~ = 150 ms, which is not reachable with current clinical hardware. See Supporting Information Figure S9 for more SNR levels Each combination of parameters was simulated 1000 times. The mean value of the absolute error of *f*, the relative error of *T* ~2~, and their SEM were reported. 3.4. In vivo clinical data: Free‐water elimination {#mrm27181-sec-0011} -------------------------------------------------- We aim to show that BSS has potential applications in clinical settings. To this end, we ran an experiment to analyze its performance for estimating tissue parameters and correcting for CSF contamination. 3.5. Data acquisition {#mrm27181-sec-0012} --------------------- Two volunteers, a male (age 28 years) and a female (age 24 years) were scanned in a 3.0 T GE MR750w (GE Healthcare, Milwaukee, WI). The in vivo study protocol was approved by our institutional review board and prior informed consent was obtained. We acquired seven diffusion pulsed gradient spin‐echo with echo planar imaging volumes for TE values from 75.1 to 135.1 ms in 10 ms increments. The following parameters were constant: FOV = 240 mm; 4 mm slice thickness; TR = 6000 ms; 96 × 96 matrix size; ASSET = 2; and 30 directions. Additionally, we measured fluid‐attenuated inversion recovery (FLAIR) SE echo planar imaging for 17 equally‐spaced TEs ranging from 20 to 260 ms. The same imaging parameters were used as for the diffusion experiments but with no acceleration (ASSET = 0). 3.6. Data analysis {#mrm27181-sec-0013} ------------------ Diffusion data for all TEs were first registered with FSL FLIRT[49](#mrm27181-bib-0049){ref-type="ref"} to the shortest TE volume. We then processed them with BSS in pairs ( $M = N = 2$) with a fixed short TE of 75.1 ms. The long TE was increased from 85.1 to 135.1 ms for a total ΔTE of 60 ms (Figures [7](#mrm27181-fig-0007){ref-type="fig"} and [8](#mrm27181-fig-0008){ref-type="fig"}). We used literature CSF values ( $T_{2_{\text{CSF}}} = 2$ s and $D_{\text{CSF}} = 3 \times 10^{- 3}$ mm^2^/s) as the prior knowledge, and constrained the possible values of $T_{2_{\text{IE}}}$ between 0 and 200 ms.[15](#mrm27181-bib-0015){ref-type="ref"}, [28](#mrm27181-bib-0028){ref-type="ref"} We report maps of the BSS relative factorization error (Figure [7](#mrm27181-fig-0007){ref-type="fig"}A,B,G), CSF volume fraction (Figure [7](#mrm27181-fig-0007){ref-type="fig"}C,H), proton density (Figure [7](#mrm27181-fig-0007){ref-type="fig"}D,I), $T_{2_{\text{IE}}}$ (Figure [7](#mrm27181-fig-0007){ref-type="fig"}E,J), and number of compartments (Figure [7](#mrm27181-fig-0007){ref-type="fig"}F,K). {#mrm27181-fig-0007} {ref-type="ref"} and mean relative error (B). Histograms of the BSS‐estimated $T_{2_{\text{IE}}}$ values are plotted against the reference (C) and (D). High *T* ~2~ values in the ventricles for the reference indicate that the suppression of the CSF signal in the FLAIR experiment was not perfect, although they appeared dark in the raw images. This might have induced a positive bias for the reference. Finally, the BSS‐estimated of $T_{2_{\text{IE}}}$ values for ΔTE above 50 ms showed good agreement with the reference](MRM-80-2155-g008){#mrm27181-fig-0008} For reference, FLAIR multi‐echo echo planar imaging data were also registered with FLIRT to the shortest TE nondiffusion weighted volume. The signal decay for each voxel was then matched to a dictionary of mono‐exponential decays from 0 to 300 ms with a grid of 1 ms. We compared this map against the BSS $T_{2_{\text{IE}}}$ map (Figure [8](#mrm27181-fig-0008){ref-type="fig"}). We defined the relative error of the matrix factorization for the in vivo data as follows: $$\epsilon = \frac{\left| \mathbf{X} - S_{0}\mathbf{A}\mathbf{S} \right|_{2}}{\left| \mathbf{X} \right|_{2}}.$$ This is a measure of the performance of BSS for each voxel. Given that we calculated **S = A^−1^X**, this error formulation is sensitive to: (1) breaches of the BSS conditions due to artifacts, and (2) numerical instabilities due to the condition of **A**. Point one is the result of B~0~ drift, subject motion, flow, and eddy currents. These effects produce a violation of the BSS condition, making the signal sources different between TE measurements. The second point is the error amplification factor. A high ϵ denotes that the factorization could not find a solution within the constrained space and thus, results might not be trustworthy. Finally, BSS does not model the compartmental diffusion signal. However, to demonstrate a simple way to perform compartment‐independent analysis and correct for CSF contamination, we fitted the disentangled signals to the DTI model.[6](#mrm27181-bib-0006){ref-type="ref"} We further fitted the measured diffusion volumes at the shortest TE, and the BSS separated signals for the IE and CSF compartments to a mono‐exponential model using standard linear regression (FSL FDT Toolbox (<http://www.fmrib.ox.ac.uk/fsl>)). For comparison, bi‐exponential models using Pasternak\'s and Collier\'s methods were used (Figures [9](#mrm27181-fig-0009){ref-type="fig"}, [10](#mrm27181-fig-0010){ref-type="fig"}, Supporting Information Figure S15). Fractional anisotropy (FA) and mean diffusivity (MD) maps were derived for each fit. {#mrm27181-fig-0009} {#mrm27181-fig-0010} 4. RESULTS {#mrm27181-sec-0014} ========== 4.1. Simulations {#mrm27181-sec-0015} ---------------- ### 4.1.1. Two compartments {#mrm27181-sec-0016} The convergence area is the region where the mean relative error of $T_{2_{\text{IE}}}$ is lower than 0.1 per unit (p.u). Its shape for all the simulations (Figures [2](#mrm27181-fig-0002){ref-type="fig"}, [3](#mrm27181-fig-0003){ref-type="fig"}, [4](#mrm27181-fig-0004){ref-type="fig"}, [5](#mrm27181-fig-0005){ref-type="fig"}, Supporting Information Figures S5, S6, S7, S8, S10, S11, S12, and S13) follows two effects. First, the condition number of the mixing matrix limits the lower bound of ΔTE: similar TE values produce more linearly dependent column vectors of **A**. And second, the SNR plays a double role, it increases the error regions where **A** is bad‐conditioned (small ΔTE), and limits the maximum ΔTE due to the *T* ~2~ decay of the signals. Thus, when the SNR increases the convergence area grows and the region of minimum SEM, denoting an improvement on the stability of the algorithm. The convergence area also depends on the IE volume fraction. The larger is the contribution of IE, the better is the $T_{2_{\text{IE}}}$ estimate. Adding priors on *S* ~CSF~ improves the $T_{2_{\text{IE}}}$ estimate, even at SNR = 50 (Supporting Information Figure S10). Bounding the solution space into nonoverlapping regions also improves the results of $T_{2_{\text{IE}}}$ (Supporting Information Figure S11), although less than combining it with CSF prior knowledge (Supporting Information Figure S13). The $T_{2_{\text{CSF}}}$ estimate shows a 0.17 p.u. due to the small variation of *S* ~CSF~ along the acquired TEs (4.4%). This is corrected when relaxometry prior is incorporated (Figure [3](#mrm27181-fig-0003){ref-type="fig"} and Supporting Information Figure S12). The comparison between Figures [2](#mrm27181-fig-0002){ref-type="fig"} and [3](#mrm27181-fig-0003){ref-type="fig"}, show the benefit of including prior knowledge into the factorization algorithm, specially at low SNR. Then, the accuracy of the estimates will be influenced by the selection of ΔTE, the *T* ~2~ boundaries, the *S* ~CSF~ prior, and the expected $T_{2_{\text{IE}}}$ and *f* ~IE~ values. We used literature values for $T_{2_{\text{IE}}},\, T_{2_{\text{CSF}}}$ ~,~ [15](#mrm27181-bib-0015){ref-type="ref"} and *S* ~CSF.~ [28](#mrm27181-bib-0028){ref-type="ref"} According to Figure [3](#mrm27181-fig-0003){ref-type="fig"}A,B one needs a minimum ΔTE of 26 ms for an accurate *f* ~IE~ estimate. Interestingly, *f* ~IE~ is a reliable parameter that tell us about the bias of $T_{2_{\text{IE}}}$, the larger *f* ~IE~ is, the more accurate $T_{2_{\text{IE}}}$ becomes (Figure [3](#mrm27181-fig-0003){ref-type="fig"}A,C). For one tissue compartment BSS is able to precisely (SEM \< 0.01) estimate the volume fraction with mean absolute error below 0.1 when ΔTE \> 35 ms (Figure [4](#mrm27181-fig-0004){ref-type="fig"}A,B). When *f* ~IE~ = 1 the area of mean convergence of the $T_{2_{\text{IE}}}$ estimate is almost independent from ΔTE (Figure [4](#mrm27181-fig-0004){ref-type="fig"}C,D). We found an equivalent result for the mean relative error of $T_{2_{\text{CSF}}}$ when *f* ~IE~ = 0 (Figure [4](#mrm27181-fig-0004){ref-type="fig"}E,F), although in this case it comes from the $T_{2_{\text{CSF}}}$ prior. Notice the large error and instability of $T_{2_{\text{IE}}}$ and $T_{2_{\text{CSF}}}$ in the opposite cases, *f* ~IE~ = 0 and *f* ~IE~ = 1, respectively (Figure [4](#mrm27181-fig-0004){ref-type="fig"}C,E). This results when BSS tries to find a component that is not in the tissue and thus, cannot be estimated. For two components with similar *T* ~2~ values and little priors (IC and EC) cALS losses efficiency. The volume fraction estimates are biased (Figure [5](#mrm27181-fig-0005){ref-type="fig"}A), and $T_{2_{IC}}$ shows a narrow convergence region that is almost independent of ΔTE. The lower bound of this region is limited by the proximity of $T_{2_{\text{IC}}}$ and $T_{2_{\text{EC}}}$ that worses the condition of **A**. The upper bound results of the lack of prior on the signal of one of the compartments, in contrast with the *S* ~CSF~ prior used before (compare Figure [2](#mrm27181-fig-0002){ref-type="fig"} and Supporting Information Figure S10) that increased the convergence area toward lower *T* ~2~ values. ### 4.1.2. Three compartments: Searching for myelin {#mrm27181-sec-0017} The convergence area is the one where the errors of *f* ~M~, *f* ~IE~, $T_{2_{M}}$, and $T_{2_{\text{IE}}}$ are lower than 0.1 in absolute value for the volume fractions and per unit for *T* ~2~. Figure [6](#mrm27181-fig-0006){ref-type="fig"}A, C, E, G shows and optimal ΔTE = 36 ms. Notice that when ΔTE increases the error of the myelin parameters grows due to the reduction of the myelin contribution to the second TE, worsening the SNR of that component (Figure [6](#mrm27181-fig-0006){ref-type="fig"}A,E). Since all the volume fractions add up to one, errors on *f* ~M~ increase the error on *f* ~IE~ (Figure [6](#mrm27181-fig-0006){ref-type="fig"}A,C). The estimate of $T_{2_{\text{IE}}}$ is dependent on SNR and its volume fraction, compounding its calculation for SNR \< 50 and *f* ~IE~ \< 0.4 (Supporting Information Figure S9G lower left corner). One should notice that including a third compartment increases the condition number of **A**, rising the instability of the factorization (Figure [6](#mrm27181-fig-0006){ref-type="fig"}F). See the phantom experiments in the Supporting Information. 4.2. In vivo clinical data: Free‐water elimination {#mrm27181-sec-0018} -------------------------------------------------- We observed that the mean relative error for the whole brain ( $\langle\epsilon\rangle$) decreased as ΔTE increased (Figure [7](#mrm27181-fig-0007){ref-type="fig"}A,B,G), in agreement with phantom findings (see supporting information) and the results of the simulations for two compartments. Interestingly, for the maximum ΔTE, we can see that the number of compartments is two in regions next to the ventricles and the cortex, but one inside the ventricles and in some deep WM areas (Figure [7](#mrm27181-fig-0007){ref-type="fig"}K). It is also noteworthy that the pure CSF areas (eg, the ventricles) have been removed from the $T_{2_{\text{IE}}}$ map (Figure [7](#mrm27181-fig-0007){ref-type="fig"}E,J), while the opposite is observed in the CSF volume fraction (Figure [7](#mrm27181-fig-0007){ref-type="fig"}C,H), indicating a successful disentangling effect. We compared the BSS‐estimated $T_{2_{\text{IE}}}$ maps for increasing ΔTE values with the reference map obtained from the FLAIR multi‐echo SE data. We noted how the structural similarity index[50](#mrm27181-bib-0050){ref-type="ref"} increased and the mean relative error decreased as ΔTE grew (Figure [8](#mrm27181-fig-0008){ref-type="fig"}A,B). Additionally, the histograms for both subjects tended toward the reference as the difference between the short and long TEs grew. This reflects an underestimation of $T_{2_{\text{IE}}}$ for small ΔTE values that can be explained by Equation [(3)](#mrm27181-disp-0003){ref-type="disp-formula"} and Supporting Information Figure S1C. Moreover, the FLAIR *T* ~2~ map showed high values in the ventricles, possibly indicating imperfect CSF suppression and, thus, slightly increased reference values (Figure [8](#mrm27181-fig-0008){ref-type="fig"}A,C,D). FA and MD maps and histograms were calculated from the BSS IE and CSF disentangled signals for both subjects (Figures [9](#mrm27181-fig-0009){ref-type="fig"}, [10](#mrm27181-fig-0010){ref-type="fig"}, Supporting Information Figure S15). These maps displayed an overestimation of the CSF volume fraction for low ΔTE values (the low FA peak in Figure [9](#mrm27181-fig-0009){ref-type="fig"}B and Supporting Information Figure S15B was removed). This resulted in a compensation effect for the previously shown underestimation of $T_{2_{\text{IE}}}$. Additionally, the FA histograms (Figure [9](#mrm27181-fig-0009){ref-type="fig"}B and Supporting Information Figure S15B) showed a tendency toward higher FA values and a reduction of the low FA peak associated with free‐water. At long ΔTE values, FA seems to tend toward a stable distribution. We also observed an enlargement of the corpus callosum and a general recovery of peripheral WM tracts and the fornix in the colored FA maps (Figure [9](#mrm27181-fig-0009){ref-type="fig"}A and Supporting Information Figure S15A). Additionally, on the MD histograms for IE water (Figure [9](#mrm27181-fig-0009){ref-type="fig"}D and Supporting Information Figure S15D) we found a reduced number of voxels with diffusivities greater than $1 \times 10^{- 3}$ mm^2^/s. In contrast, the main peak at 0.7 $\times 10^{- 3}$ mm^2^/s, associated with the parenchyma, remained in its original position, indicating that IE water represents a nonCSF tissue. This MD reduction was also visible in the maps (Figure [9](#mrm27181-fig-0009){ref-type="fig"}C and Supporting Information Figure S15C). Finally, the MD histograms for CSF water (Figure [10](#mrm27181-fig-0010){ref-type="fig"}) showed a tendency toward $3 \times 10^{- 3}$ mm^2^/s as ΔTE increased, in agreement with the literature.[28](#mrm27181-bib-0028){ref-type="ref"} All these findings agreed with a disentangling of IE and CSF signals and thus, a correction of the free‐water partial volume effect in the diffusion signal. 5. DISCUSSION {#mrm27181-sec-0019} ============= 5.1. Stability {#mrm27181-sec-0020} -------------- Four main approaches exist for the BSS problem (independent component analysis, principal component analysis, NMF, and sparse component analysis). Choosing the appropriate method depends on the prior knowledge of the signal sources. In our experiments, we relied on NMF, using a constrained version of the ALS algorithm (cALS). Others explored these algorithms before. Pauca et al.[51](#mrm27181-bib-0051){ref-type="ref"} used low‐rank and sparsity constraints to distinguish semantic features in text mining, and later[52](#mrm27181-bib-0052){ref-type="ref"} smoothness regularization to identify space objects from spectral data. Gao and Church[53](#mrm27181-bib-0053){ref-type="ref"} also employed sparseness for cancer class discovery through gene clustering, which was later extended by Kim and Park[54](#mrm27181-bib-0054){ref-type="ref"} improving the balance between accuracy and sparseness through regularization. They also introduced a variation based on the active set method[55](#mrm27181-bib-0055){ref-type="ref"} and low‐rank approximation.[56](#mrm27181-bib-0056){ref-type="ref"} Liu et al.[57](#mrm27181-bib-0057){ref-type="ref"} incorporated label information to create a semi‐supervised matrix decomposition method. Sun and Févotte[58](#mrm27181-bib-0058){ref-type="ref"} introduced a version based on the alternating direction method of multipliers[59](#mrm27181-bib-0059){ref-type="ref"} (ADMM), that was further stabilized by Zhang et al.[60](#mrm27181-bib-0060){ref-type="ref"} Supported by previous work, we presented a biophysical inspired solution to constrain the diffusion‐relaxometry NMF compartmental problem. Essentially, our cALS algorithm imposes two constraints: (1) the rows of **A** must follow exponential relationships (relaxometry); and (2) when the analytical expression of one component is known (ie, CSF) the corresponding row in **S** is fixed (diffusion). The stability of cALS is linked to the condition of **A** and SNR; an ill‐conditioned mixing matrix will lead to error propagation due to numerical instability. We optimized the experimental TEs to reduce the condition number of **A** for literature values of *T* ~2~. However, further research based on ADMM might yield better results. We ran extensive simulations for two compartments at clinical TE values with different priors, and three compartments at lower TEs. These simulations highlighted the importance of choosing literature supported priors to improve the convergence, especially at low SNR. Constrained ALS converges when the number of compartments in tissue is equal or lower than the expected, but it looses performance for species with similar *T* ~2~. Phantom experiments (see supporting information) agreed with simulation results, validating that BSS was able to accurately estimate *T* ~2~ for one compartment and separate diffusion signal sources and estimate *T* ~2~ and *f* for two compartments. However, they also showed that scaling the cALS algorithm to three compartments, including fast *T* ~2~ decaying species, is unstable in the range of the clinically available TE values. Finally, repeatability and reproducibility analyses (see supporting informtaion) show that cALS yield consistent results across repetitions and subjects, highlighting its stability. 5.2. Relaxation time and volume fraction estimates {#mrm27181-sec-0021} -------------------------------------------------- BSS provides the means to estimate *T* ~2~ relaxation values and volume fractions. Interestingly, only a number of TE repetitions equal to the number of compartments that are assumed to be in the tissue is necessary. This results of the substitution of the ILTs by BSS, in comparison to other techniques.[15](#mrm27181-bib-0015){ref-type="ref"}, [17](#mrm27181-bib-0017){ref-type="ref"}, [21](#mrm27181-bib-0021){ref-type="ref"}, [24](#mrm27181-bib-0024){ref-type="ref"} We found a good agreement between the $T_{2_{\text{IE}}}$ estimates of the FLAIR multi‐echo SE for 17 TEs and those of BSS for 2 TEs. In this sense, all the measurements along the diffusion space are considered for both TEs, incorporating redundancy and reinforcing the estimation of *T* ~2~. The SNR for the in vivo data were 147 and 104 for subjects one and two. According to the simulations at ΔTE = 60 ms, the expected absolute error for the volume fraction estimate is below 0.03, meaning that $T_{2_{\text{IE}}}$ is highly reliable in WM areas, and lesser in the CSF borders. 5.3. Myelin detection {#mrm27181-sec-0022} --------------------- Simulations proved that our method has the potential to disentangle three compartments by reducing the minimum TE in diffusion experiments. As a result, myelin water could be incorporated into the model (Figure [6](#mrm27181-fig-0006){ref-type="fig"}). However, we are prevented from conducting such experiments by gradient performance on clinical scanners. 5.4. Disentangling the diffusion sources and free water elimination {#mrm27181-sec-0023} ------------------------------------------------------------------- Unlike other multicompartment diffusion models[2](#mrm27181-bib-0002){ref-type="ref"}, [7](#mrm27181-bib-0007){ref-type="ref"}, [8](#mrm27181-bib-0008){ref-type="ref"}, [11](#mrm27181-bib-0011){ref-type="ref"} or more recent contributions,[27](#mrm27181-bib-0027){ref-type="ref"}, [35](#mrm27181-bib-0035){ref-type="ref"} our approach does not model compartmental diffusion. Our framework instead relies on three assumptions: (1) microstructural water compartments have distinct *T* ~2~ relaxation times;[14](#mrm27181-bib-0014){ref-type="ref"}, [15](#mrm27181-bib-0015){ref-type="ref"} (2) each have different diffusion characteristics;[19](#mrm27181-bib-0019){ref-type="ref"}, [20](#mrm27181-bib-0020){ref-type="ref"} and (3) the effects of the water exchange are negligible on the timescale of our experiments.[9](#mrm27181-bib-0009){ref-type="ref"}, [61](#mrm27181-bib-0061){ref-type="ref"} Furthermore, our solution is diffusion protocol‐agnostic (only two TEs and one nondiffusion weighted volume are necessary), allowing for flexibility in the design of the acquisition protocol, which might include any number of diffusion directions and *b*‐values. This gives it an advantage over diffusion--relaxation correlation techniques based on regularized ILTs.[21](#mrm27181-bib-0021){ref-type="ref"}, [24](#mrm27181-bib-0024){ref-type="ref"} A promising application of the protocol‐agnostics nature of our framework is correcting for free water contamination. Recently Collier et al.[35](#mrm27181-bib-0035){ref-type="ref"} included TE dependence in their bi‐exponential diffusion tensor model to regularize the fitting problem. However, they fitted the bi‐exponential DTI model directly. Contrary, our solution does not assume any particular diffusion model, we instead separated the signal from each compartment, allowing more flexible and independent study. In this regard, analysis of the signal associated with the CSF compartment can be seen as a disentanglement quality assurance metric (Figures [9](#mrm27181-fig-0009){ref-type="fig"}, [10](#mrm27181-fig-0010){ref-type="fig"}, Supporting Information Figure S15), or in brain tissue applications, a general indicator of the goodness‐of‐fit for IE and CSF. We fitted our data to Collier\'s model[35](#mrm27181-bib-0035){ref-type="ref"} without reaching convergence, which resulted due to our single‐shelled dataset. Comparison of BSS with Pasternak\'s free‐water elimination method[31](#mrm27181-bib-0031){ref-type="ref"} is show in Figure [9](#mrm27181-fig-0009){ref-type="fig"} and Supporting Information Figure S15. We observed a good agreement between BSS for ΔTE = 60 ms and Pasternak\'s free‐water elimination for FAs between 0‐0.2 and 0.8‐1. In the middle FA range both methods disagree, BSS shows an homogeneous correction, while Pasternak\'s results follow the standard DTI fit from 0.2 to 0.4 and shows a correcting effect from 0.4 to 1 (Figure [9](#mrm27181-fig-0009){ref-type="fig"}A,B, Supporting Information Figures S15A and S15B). It is impossible to determine which method is better (no ground‐truth). However, there are two indicators that BSS might be performing better: (1) the BSS FA curve runs in parallel to the standard DTI fit from 0.2 to 0.8, denoting an stable correction without favoring any FA range; and (2) Pasternak\'s MD is spatially over‐regularized (Figure [9](#mrm27181-fig-0009){ref-type="fig"}C,D, Supporting Information Figures S15C and S15D), while BSS\'s MD keeps its maximum at 0.7 mm^2^/s, the reference for parenchyma.[28](#mrm27181-bib-0028){ref-type="ref"} Long ΔTE values benefit our framework, which is not surprising and agrees with the findings of Collier et al.[35](#mrm27181-bib-0035){ref-type="ref"} This is not only due to the relationship between **A** and *T* ~2~ (Equation [(3)](#mrm27181-disp-0003){ref-type="disp-formula"} and Supporting Information Figure S1C) but also because longer differences between TEs produce more distinct levels of mixing and thus better codification of the information from each source. That is to say, the short TE contains more information about the fast‐relaxing species, while the long TE is dominated by CSF. 6. CONCLUSIONS {#mrm27181-sec-0024} ============== We have introduced for the first time a BSS framework for expressing the relationships between diffusion signals acquired at different TEs. This new approach does not rely on diffusion modeling or the ILT. Our results show that, with the current hardware, blind source separation allows for disentangling the diffusion signal sources generated by each sub‐voxel compartment up to two compartments, making it a suitable tool for free‐water elimination. Moreover, it simultaneously estimates proton density, volume fractions, relaxation times and the number of compartments in the underlying microstructure, paving the way for tissue microstructure characterization when the hardware constraints are relieved. CONFLICT OF INTEREST {#mrm27181-sec-0026} ==================== Miguel Molina Romero and Pedro A. Gómez receive research support by GE Global Research. Dr. Jonathan I. Sperl was GE Global Research employee during the process of this work and currently is Siemens Healthcare employee. Dr. Marion I. Menzel is GE Healthcare employee. Supporting information ====================== Additional Supporting Information may be found online in the supporting information tab for this article. ###### **FIGURE S1** Evolution of the relative error in the *T* ~2~ estimate with ΔTE for one compartment. The mean relative error of *T* ~2~ estimated using BSS is shown in (a) for NNLS and in (b) for EASI‐SM references. ΔTE goes from 5 ms (darker colors) to 50 ms (lighter colors). The dependence of *T* ~2~ on the direction (slope) of the columns of **A** (Equation 3) is shown in (c), where it can be seen how increasing ΔTE improves the dynamic range of the slope of **A**, resulting in a better estimate for *T* ~2~. Except for ROI~1~ and ROI~11~, the remaining ones reduce the *T* ~2~ mean relative error as ΔTE increases (a and b, lighter colors are closer to zero), in agreement with plot c. **FIGURE S2** Separation of two compartments and parameter estimation for the phantom data. The signal sources of the *simulated* dataset are plotted in (a), and the *measured* data generated from the sources in (b). The resulting mixtures for both datasets are shown in (c). We use the subscripts *M* and *S* to refer to estimates for the *measured* and *simulated* datasets, respectively. Measurement errors are highlighted by the differences between the *measured* and *simulated* signals, shown in (c). BSS disentangled the original sources for both datasets, as shown in (d). We chose a ΔTE of 50 ms to minimize the condition of **A** (shown in (e)) and increase the numerical stability of the framework. Finally, the relative errors in the estimated parameters, ${\hat{T}}_{2_{ROI_{6}}}$ and ${\hat{f}}_{ROI_{6}}$, are plotted in (f) for all possible values of ΔTE. We observed good agreement between the reference signals and those disentangled with BSS. **FIGURE S3** Separation of three compartments and parameter estimation for the phantom data. The *simulated* dataset was generated from the signal sources in (a). The *measured* datasets were calculated from the measured signals for ROI~5~ (b), ROI~6~ (c), and ROI~11~ (d). The mixed signals for both datasets (shown in (e)) show a mismatch due to measurement errors. They were disentangled with BSS, as shown in (f). We fixed TE~1~ = 77.5 ms and TE~3~ = 127.5 ms, and varied TE~2~ to minimize the condition number of **A** (shown in (g)). The relative errors of the estimated parameters are plotted for different values of the TE~2~ in (h). **FIGURE S4** Simulated diffusion signals for IE and CSF. Synthetically generated diffusion signals for 30 directions (*b* = 1000 s/mm^2^) and one non‐diffusion weighted measurement. We modeled diffusion as a Gaussian process with MD of IE and CSF equal to $0.7 \times 10^{- 3}$ and $3 \times 10^{- 3}$ mm^2^/s respectively,[^28^](#mrm27181-bib-0028){ref-type="ref"} and standard deviations of $0.3 \times 10^{- 3}$ and $0.1 \times 10^{- 3}$ mm^2^/s respectively to distinguish between hindered anisotropic (IE) and free isotropic (CSF) diffusivity. **FIGURE S5** Convergence for two compartments (IE and CSF) with overlapping *T* ~2~ constraints and no *S~CSF~* prior. This figure extends the analysis of Figure 2 for SNR = 100 and 150. The stability for *f~IE~* increases with SNR (a and b) and with *f~IE~* for $T_{2_{IE}}$ (c and d). **FIGURE S6** Convergence for two compartments (IE and CSF) with non‐overlapping *T* ~2~ constraints and *S~CSF~* prior. This figure extends the analysis of Figure 3 for SNR = 100 and 150. The size and stability of the convergence area for *f~IE~* and $T_{2_{IE}}$ increase with SNR. **FIGURE S7** Convergence for two compartments (IE and CSF) with non‐overlapping *T* ~2~ constraints and *S~CSF~* prior when only one is actually present in the tissue. This figure extends the analysis of Figure 4 for SNR = 100 and 150. The SNR does not play an important role in the definition of the convergence area. **FIGURE S8** Convergence for two compartments (IC and EC) with overlapping *T* ~2~ constraints and no other priors. This figure extends the analysis of Figure 5 for SNR = 100 and 150. The influence of SNR on *f* and $T_{2_{IC}}$ is small. **FIGURE S9** Convergence for three compartments (myelin, IE, and CSF) with non‐overlapping *T* ~2~ constraints and *S~CSF~* prior. This Figure extends the analysis of Figure 6 for SNR = 100 and 150. **FIGURE S10** Convergence for two compartments (IE and CSF) with overlapping *T* ~2~ constraints and *S~CSF~* prior. The mean and the standard error of *f~IE~* absolute error (a and b), and the mean and the standard error of $T_{2_{IE}}$ (c and d), and $T_{2_{CSF}}$ (e and f) relative error per unit (p.u.). Red and white lines mark the 0.2 and 0.1 contour respectively. One thousand simulations were run for each combination of SNR, *f~IE~*, $T_{2_{IE}}$, and ΔTE. $T_{2_{IE}}$ and $T_{2_{CSF}}$ were bound between 0--1000 ms and 0--3000 ms respectively. *S~CSF~* was set to have isotropic diffusivity with value $3 \times 10^{- 3}$ mm^2^/s. We defined the convergence area as the one with error lower than 0.1 for *f~IE~* and $T_{2_{IE}}$. Notice the growth of the converge area compared to the lack of priors (Figures 2 and S5). **FIGURE S11** Convergence for two compartments (IE and CSF) with non‐overlapping *T* ~2~ constrained and no *S~CSF~* prior. The mean and the standard error of *f~IE~* absolute error (a and b), and the mean and the standard error of $T_{2_{IE}}$ (c and d), and $T_{2_{CSF}}$ (e and f) relative error per unit (p.u.). Red and white lines mark the 0.2 and 0.1 contour respectively. One thousand simulations were run for each combination of SNR, *f~IE~*, $T_{2_{IE}}$, and ΔTE. $T_{2_{IE}}$ and $T_{2_{CSF}}$ were bound between 0--300 ms and 300--3000 ms respectively. No prior was imposed on *S~CSF~*. We defined the convergence area as the one with error lower than 0.1 for *f~IE~* and $T_{2_{IE}}$. Non‐overlapping *T* ~2~ bounds stabilize the factorization, compared to Figures 2 and S5, although not as much as using priors on the signal sources (Figure S10). **FIGURE S12** Convergence for two compartments (IE and CSF) with fixed $T_{2_{CSF}}$ and no *S~CSF~* prior. The mean and the standard error of *f~IE~* absolute error (a and b), and the mean and the standard error of $T_{2_{IE}}$ (c and d), and $T_{2_{CSF}}$ (e and f) relative error per unit (p.u.). Red and white lines mark the 0.2 and 0.1 contour respectively. One thousand simulations were run for each combination of SNR, *f~IE~*, $T_{2_{IE}}$, and ΔTE. $T_{2_{IE}}$ was bound between 0--300 and $T_{2_{CSF}}$ fixed to 2000 ms. No prior was imposed on *S~CSF~*. We defined the convergence area as the one with error lower than 0.1 for *f~IE~* and $T_{2_{IE}}$. Fixing the value of $T_{2_{CSF}}$ does not have any effect on the size of the convergence area, while bounding $T_{2_{IE}}$ does it (see Figure S11). **FIGURE S13** Convergence for two compartments (IE and CSF) with non‐overlapping *T* ~2~ constraints and *S~CSF~* prior. The mean and standard error of *f~IE~* absolute error (a and b), and mean and standard error of $T_{2_{IE}}$ (c and d), and $T_{2_{CSF}}$ (e and f) relative error per unit (p.u.). Red and white lines mark the 0.2 and 0.1 contour respectively. One thousand simulations were run for each combination of SNR, *f~IE~*, $T_{2_{IE}}$, and ΔTE. $T_{2_{IE}}$ and $T_{2_{CSF}}$ were bound between 0--300 ms and 300--3000 ms respectively. *S~CSF~* was set to have isotropic diffusivity with value $3 \times 10^{- 3}$ mm^2^/s. We defined the convergence area as the one with error lower than 0.1 for *f~IE~* and $T_{2_{IE}}$. Incorporating prior knowledge on the behavior of the signal sources (as CSF) improves convergence and stability more than bounding *T* ~2~ (Compare with Figures S10 and S11) **FIGURE S14** Simulated diffusion signals for intra and extra‐cellular water compartments. Synthetically generated diffusion signals for 30 directions (b = 1000 s/mm^2^) and one non‐diffusion weighted measurement. We modeled diffusion as a Gaussian process with MD of intra‐cellular (IC) and extra‐cellular (EC) equal to $0.6 \times 10^{- 3}$ and $0.8 \times 10^{- 3}$ mm^2^/s respectively (to keep the MD of parenchyma equals to $0.7 \times 10^{- 3}$ mm^2^/s)[^28^](#mrm27181-bib-0028){ref-type="ref"} and standard deviations of $0.3 \times 10^{- 3}$ and $0.1 \times 10^{- 3}$ mm^2^/s respectively to distinguish between a more (IC) and less (EC) hindered anisotropic diffusivity. **FIGURE S15** FA and MD of the BSS‐disentangled IE signal against the standard DTI and Pasternak\'s free‐water elimination (FWE) for subject one. Comparisons of the FA (b) and MD (d) histograms calculated from the separated IE signals are plotted against the standard DTI fit and Pasternak\'s method for the short TE measured data. MD (c) and colored FA (a) maps are also included for comparison. We observed a CSF correction effect in the long ΔTE BSS for FA in agreement with Pasternak\'s FWE. However, both method disagree for MD, where Pasternak\'s introduces spatial over‐regularization. See Figure 9 for subject two. **FIGURE S16** Repeatability analysis showing intra‐subject variability. A healthy volunteer was scanned six times. The FA (a) and MD (b) histograms for standard DTI, BSS and Pasternak\'s method are shown. These histograms were fragmented in sectors and the relative changes in number of voxels per sector and repetition for BSS and Pasternak\'s methods were computed. Statistical t‐tests were run per sector to determine the level of significance of the differences between BSS and Pasternak\'s results (d and e). BSS and FLAIR $T_{2_{IE}}$ histograms (c) showed good agreement. Their peak and the full width half maximum (FWHM) were used for t‐test comparison between BSS and FLAIR (f) highlighting the concordance. **FIGURE S17** Reproducibility analysis showing inter‐subject variability. Twenty healthy volunteers were scanned. The FA (a) and MD (b) histograms for standard DTI, BSS and Pasternak\'s method are shown. These histograms were fragmented in sectors and the relative changes in number of voxels per sector and repetition for BSS and Pasternak\'s methods were computed. Statistical t‐tests were run per sector to determine the level of significance of the differences between BSS and Pasternak\'s results (d and e). Notice that the inter‐subject variability is larger than intra‐subject (Figure S16). BSS and FLAIR $T_{2_{IE}}$ histograms (c) were depicted. Their peak and the full width half maximum (FWHM) were used for t‐test comparison between BSS and FLAIR (f). **Table S1** Phantom reference values and BSS estimates. The ROIs in the phantom experiment was built using the concentrations of agar and sucrose shown here. Signal decays along the diffusion dimension were compared to each other to ensure that they were all different, as required by BSS (see supplementary Figure S18). For reference, the *T* ~2~ values were characterized using an NNLS fit. Confidence intervals were taken at the half maxima of the NNLS spectral peaks. In addition, a second method, EASI‐SM,[^17^](#mrm27181-bib-0017){ref-type="ref"} was used to confirm the validity of the fits. Finally, the $T_{2_{BSS}}$ values were estimated for ΔTE = 50 ms and compared with the NNLS and EASI‐SM references (where *ϵ* refers to the relative error). ###### Click here for additional data file. The authors want to thank Dr. Ofer Pasternak for his support in the comparison of the methods. This work was supported by the TUM Institute of Advanced Study, funded by the German Excellence Initiative, and the European Commission (Grant Agreement Number 605162). DKJ was supported by a Wellcome Trust Investigator Award (096646/Z/11/Z) and a Wellcome Trust Strategic Award (104943/Z/14/Z).

INTRODUCTION ============ Robotic-assisted laparoscopic radical prostatectomy (RALP) has gained popularity around the world. This year the number of robotic cases again exceeds that of traditional open prostatectomy cases in the United States, with 85% performed robotically.^[@B1]^ We performed a review of our prostate cancer database to demonstrate trends in biochemical recurrence (BCR) in a large cohort of patients undergoing RALP, as a function of preoperative prostate-specific antigen (PSA) levels, Gleason score, TNM stage, and surgical margins (SM). Our primary objective was to determine prostate cancer BCR rates with respect to SM status for patients undergoing RALP. We also compared our BCR rates to other open and laparoscopic radical prostatectomy series reported in the literature. MATERIALS AND METHODS ===================== A retrospective review of a prospectively maintained, Internal Review Board (IRB)-approved radical prostatectomy database was performed. Patients undergoing RALP between December 1, 2003 and January 1, 2009 were eligible for analysis. All RALP procedures were performed via a transperitoneal approach by 1 of 3 surgeons using a modified Vattikuti Institute technique.^[@B2]^ DaVinci, DaVinci S, and DaVinci S HD Surgical Systems (Intuitive Surgical, Sunnyvale, CA, USA) were used to perform the procedures. Learning curves when present for all surgeons were included. The demographic information, clinical staging, and intraoperative details were prospectively collected and entered into the database. Using preoperative data, patients were stratified as low, intermediate, and high risk according to D'Amico\'s risk classification.^[@B3]^ Specifically, the low-risk group was defined as having serum PSA \<10ng/mL, Gleason score \<7, and clinical stage \<cT2b. The intermediate-risk group was defined as having serum PSA between 10ng/mL and 20ng/mL or Gleason score equal to 7. The high-risk group was defined as having serum PSA \>20ng/mL or Gleason score 8 to 10. Surgical margin status was determined by pathologic evaluation of the specimen at a single institution. AJCC 2002 staging guidelines were consistently used by the pathologists.^[@B4]^ All specimens were whole-mounted and step-sectioned at 3-mm intervals with apex and base being additionally cross-sectioned. Positive surgical margin was defined as the presence of cancer cells at the inked resection margin in the final specimen. Intraoperative biopsy results or additional tissue excisions were not used to determine margin positivity. Margin presence (negative/positive), margin multiplicity (single/multiple), and margin length (≤3mm focal and \>3mm extensive) were noted. Postoperative PSA values were routinely obtained at 1, 3, 6, 9, 12, 18, and 24 mo and annually thereafter. PSA recurrence was defined as ≥0.2ng/mL. Patients receiving adjuvant treatment were included in our analyses and were considered as having recurred at the time of adjuvant treatment. We excluded 278 patients due to lack of postoperative PSA or lack of follow-up at our institution. Demographic comparison was performed between excluded and included patients. Data were extracted from the prostate database and entered into SPSS v 14.0 statistical software (SPSS, Inc., Chicago, IL, USA.) Univariate analysis was performed for the overall cohort, as well as for patients stratified by Gleason score, TNM stage, and D'Amico classification, comparing biochemical recurrence-free survival (BRFS) with respect to SM status. BRFS was defined as time from RALP to BCR. Patients without BCR were censored on the date of their last PSA measurement. Kaplan-Meier plots of BRFS were generated for the above categories. Differences relative to SM status were evaluated using the log-rank test. Multivariate Cox regression analysis was also performed. Pathologic stage (≤pT2N0M0, pT3aN0M0, pT3b/4N0M0, pTxN1Mx), pathologic grade (pathologic Gleason score ≤6, 7, and 8 to 10), SM status (negative/positive), margin multiplicity (single/multiple), and margin length (≤3mm focal and \>3mm extensive) were entered as categorical variables with the first category as the reference group. Preoperative PSA, transformed to its natural logarithm to minimize effects of extreme values, an approach previously described in the literature, was entered as a continuous variable.^[@B5]^ Statistical significance was set at 2-sided *P* \< .05. To evaluate whether nerve-sparing status is a risk factor for positive margins, a univariate analysis was performed for bilateral, unilateral, and non--nerve-sparing procedures. The impact of the learning curve on margin status was examined by performing a univariate analysis on each quartile of cases. RESULTS ======= Between December 1, 2003 and January 2009, 1437 patients underwent RALP procedures performed by 3 surgeons at our institution. Of these, 278 patients were excluded due to lack of follow-up at our institution and lack of postoperative PSA. Sixteen patients received adjuvant therapy and were included in our analysis. Specifically, 9 patients underwent adjuvant radiation therapy, 3 received adjuvant androgen deprivation therapy, and 4 received both. A total of 1159 patients were included in our analysis. Patient demographics are presented in **[Table 1](#T1){ref-type="table"}**. Mean age, body mass index (BMI), and preoperative PSA values were 59.3 y, 28.1kg/m^2^, and 5.9ng/mL, respectively. Patients had a mean follow-up of 15.9 mo (0 to 60). Predictably, preoperative serum PSA demonstrated a statistically significant difference between the 2 groups, with higher values for those with positive surgical margins (*P* \< .001). The distribution of clinical Gleason score (*P* = .003), D'Amico risk stratification (*P* = .001), pathological Gleason (*P* = .001), and pathologic stage (*P* = .001) also demonstrated a statistically significant difference. Patients with positive margins had higher stage, grade, and risk. Neither performance of pelvic lymphadenectomy (*P* = .182) nor presence of positive lymph nodes (*P* = .139) was statistically different between the 2 groups. ###### Demographic Data Negative Margin Positive Margin All Patients P Value ------------------------------------------------------------------------------------------------------- ----------------- ----------------- --------------- --------- Number of patients (%) 843 (72.7%) 316 (27.3%) 1159 (100.0%) Mean age, years (SD[^a^](#TF1-1){ref-type="table-fn"}) 59.3 (6.5) 59.2 (6.4) 59.3 (6.5) .987 Mean BMI, kg/m^2^ (SD[^a^](#TF1-1){ref-type="table-fn"}) 28.1 (7.2) 28.0 (3.9) 28.1 (6.4) .693 Mean preoperative PSA[^a^](#TF1-1){ref-type="table-fn"}, ng/mL (SD[^a^](#TF1-1){ref-type="table-fn"}) 5.4 (2.8) 7.4 (6.7) 5.9 (4.4) \<.001 Clinical Gleason score .003     ≤6 485 (76.6%) 148 (23.4%) 633 (100.0%)     7 295 (68.6%) 135 (31.4%) 430 (100.0%)     8--10 61 (64.9%) 33 (35.1%) 94 (100.0%) Clinical stage .073     cT1 588 (73.2%) 215 (26.8%) 803 (100.0%)     cT2 246 (72.8%) 92 (27.2%) 338 (100.0%)     cT3 7 (46.7%) 8 (53.3%) 15 (100.0%) Risk class (D'Amico) \<.001     Low 439 (77.0%) 131 (23.0%) 570 (100.0%)     Intermediate 329 (71.4%) 132 (28.6%) 461 (100.0%)     High 74 (58.3%) 53 (41.7%) 127 (100.0%) Pathologic Gleason score \<.001     ≤6 294 (82.4%) 63 (17.6%) 357 (100.0%)     7 483 (69.1%) 216 (30.9%) 699 (100.0%)     8--10 60 (61.9%) 37 (38.1%) 97 (100.0%) Pathologic stage \<.001     pT0--pT2N0M0 718 (79.9%) 183 (20.3%) 901 (100.0%)     T3aN0M0 89 (46.8%) 101 (53.2%) 190 (100.0%)     T3b and pT4N0M0 25 (50.0%) 25 (50.0%) 50 (100.0%)     TxN1Mx (positive nodes) 8 (53.3%) 7 (46.7%) 15 (100.0%) Mean follow-up, months (SD[^a^](#TF1-1){ref-type="table-fn"}) 15.6 (13.0) 16.5 (14.3) 15.9(13.4) .337 Learning curve .248 First quartile 201 (69.3%) 89 (30.7%) 290 (100%) Second quartile 217 (74.8%) 73 (25.2%) 290 (100%) Third quartile 219 (75.8%) 70 24.2%) 289 (100%) Fourth quartile 206 (71.0%) 84 (29.0%) 290 (100%) Nerve sparing status .672 Bilateral nerve sparing 564 (72.7%) 212 (27.3%) 776 (100%) Unilateral nerve sparing 217 (73.8%) 77 (26.2%) 294 (100%) Non nerve sparing 60 (69.0%) 27 (31.0%) 87 (100%) SD=standard deviation; BMI=body mass index; PSA=prostate-specific antigen. Excluded patients were compared with the analyzed patients and no statistically significant differences were found in the demographic, operative, or pathologic characteristics, with exception of frequency of pelvic lymphadenectomy. Pelvic lymphadenectomy was performed on 48.6% of excluded patients vs. 58.9% in the analyzed population (*P* = .002). The frequency of positive margins in the overall cohort was 27.3%, 20.3% for pathologic stage T2, and 52.3% for pathologic stage ≥T3a disease. Specifically, in patients with Gleason score ≤6, pathologic stage T2N0M0 and T3aN0M0 had 14.9% and 65% incidence of positive surgical margin (PSM), respectively. In patients with Gleason score 7, the incidence of PSM for stage T2N0M0, T3aN0M0, T3b/T4N0M0, and TxN1Mx was 24.1%, 53.2%, 51.7%, and 12.5%, respectively. In patients with Gleason score 8 to 10, the incidence of PSM for stage T2N0M0, T3aN0M0, T3b/T4N0M0, and TxN1Mx was 17.9%, 45.2%, 50.0%, and 85.7%, respectively. BRFS curves were generated for the overall cohort as well as individual D'Amico risk groups and are depicted in **[Figure 1](#F1){ref-type="fig"}**. Low-risk and intermediate-risk groups achieved a statistically significant difference in BRSF with respect to SM. In the low-risk group, mean BRFS was 56.7 vs. 51.0 mo (*P* = .005), and in the intermediate-risk group it was 55.2 vs. 43.5 mo (*P* \< .001) for negative and positive SM, respectively. The high-risk group did not reach statistical significance with respect to SM. The overall cohort mean BRFS was statistically significant, 54.9 vs. 45.9 mo (*P* \< .001) for negative and positive SM, respectively. {#F1} Similar curves were generated for individual pathologic tumor stage and pathologic Gleason score for positive and negative SM and are depicted in **[Figure 2](#F2){ref-type="fig"}**. Patients with Gleason score of 7 and pathologic stage T2N0M0 had a statistically significant difference in mean BRFS of 55.6 and 48.7 mo (*P* \< .001) for negative and positive margins, respectively. Others did not have a statistically significant difference in BCRF survival with respect to SM. {#F2} When surgical margins were further categorized by multiplicity and length, multiple margins and extensive margins carried a higher rate of BCR (*P* \< .001) **([Figures 3](#F3){ref-type="fig"} and [4](#F4){ref-type="fig"})**. The rates of positive surgical margins did not change significantly between quartile of cases (*P* = .243, [Table 1](#T1){ref-type="table"}). Nerve-sparing surgical status did not significantly affect the incidence of positive surgical margins (*P* = .672, [Table 1](#T1){ref-type="table"}). {#F3} {#F4} The multivariate prediction of time to BRFS showed pathological stage (*P* \< .001), pathologic Gleason grade (*P* \< .001), and SM (*P* = .035) as significant covariates. Data for preoperative PSA was suggestive but missed statistical significance (*P* = .053). While margin multiplicity and margin length were significant in univariate analysis, they were not in multivariate analysis. Hazard ratios (HR) and confidence intervals (CI) for each variable in the equation are listed in **[Table 2](#T2){ref-type="table"}**. ###### Multivariate analysis of biochemical recurrence-free survival Variable HR[^c^](#TF2-3){ref-type="table-fn"} CI[^c^](#TF2-3){ref-type="table-fn"} (95%) P Value ---------------------------------------------------------- -------------------------------------- -------------------------------------------- --------- Pathologic stage \<.001     ≤pT2N0M0 1.0 (reference)     pT3aN0M0 2.30 1.34--3.97 .003     pT3b/T4N0M0 5.22 2.72--9.99 \<.001     pTxN1Mx 13.48 5.61--32.42 \<.001 Pathologic Gleason score \<.001     ≤6 1.0 (reference)     7 3.42 1.34--8.75 .01     8--10 10.13 3.66--28.02 \<.001 Surgical margin status[^a^](#TF2-1){ref-type="table-fn"} .035     Negative 1.0 (reference)     Positive 1.63 1.83--2.56 .035 Margin multiplicity[^b^](#TF2-2){ref-type="table-fn"}     Single 1.0 (reference)     Multiple .98 .48--2.01 .954 Margin length[^b^](#TF2-2){ref-type="table-fn"}     [\<]{.ul}3mm 1.0 (reference)     \>3mm 1.45 .74--2.87 .281 Preoperative PSA[^c^](#TF2-3){ref-type="table-fn"} 1.44 0.995--2.09 .053 Comparing negative vs positive margin for all patients. Comparing length and multiplicity among those with positive margins. HR=hazard ratio; CI = confidence interval); PSA=prostate-specific antigen. DISCUSSION ========== The importance of positive SM, its impact on BRFS, and as a result, its impact on disease-specific survival has been previously established in univariate and multivariate analyses.^[@B6]--[@B8]^ In our series, the frequency of positive SM for the overall cohort was slightly higher than but within the range reported in the literature for both open and robotic series.^[@B5],[@B9]--[@B13]^ While some series have shown that margin rates improve with surgical volume, this was not our experience.^[@B5],[@B14],[@B15]^ Our margin rates were fairly constant over the time period examined, most likely because 2 of our 3 surgeons had gone through their learning curves during fellowship and/or prior to joining our institution; the learning curve of only 1 surgeon is present in our data. With respect to SM status, our analyses demonstrate that the following groups reached statistically significant difference in BRFS: pT2/Gleason 7, D'Amico low-risk and intermediate--risk groups, and the overall cohort. The rest of the groups did not show a statistically significant difference for BCRF survival with respect to SM. One possible explanation is that these groups were underpowered. Alternatively, the lack of statistical difference may be real and represents the minimal impact of SM on BRFS in certain subgroups. It is known that prostate cancer in patients with Gleason score ≤6 tends to not recur, regardless of SM status, as demonstrated by our pT2/Gleason score ≤6 subgroup. This phenomenon is supported by a recent publication by Caire et al.^[@B16]^ where authors concluded that patients with pathologic Gleason score \<7 had a significantly delayed BCR as compared with those with Gleason score ≥7. These authors also found that delayed BCR confers a disease-specific survival advantage. Similarly, Karakiewicz et al.^[@B13]^ determined SM to be a significant predictor of BCR only for Gleason score ≥7. In the multivariate analysis, SM remained a significant predictor of BCR, as did pathologic stage and Gleason score. When positive margins were further categorized by length and multiplicity, multiple positive margins and extensive positive margins carried a higher rate of BCR. Preoperative PSA did not quite reach statistical significance. Pavlovich et al.^[@B17]^ also failed to find significant association between preoperative PSA levels and BCR. Once pathologic stage, Gleason score, and margin status were included in the analysis, the significance of preoperative PSA was lost. Although serum PSA level may play a role in screening and as a tumor marker to detect BCR postoperatively, the prognostic ability of preoperative serum PSA with respect to BCR is not supported by our analysis. Our cumulative BRFS was 93.3%, 90.6%, 86.2%, 79.7%, and 72.0% at 1, 2, 3, 4, and 5 y after RALP, respectively. The 5-y BRFS of 72% is comparable to other series in the literature. Murphy et al.^[@B15]^ reported 74.0% for robotic-assisted, Rassweiler et al.^[@B18]^ 73.1% for laparoscopic, and Karakiewicz et al.^[@B13]^ 75.4% for open radical retropubic prostatectomy series. Sixteen patients in our study received adjuvant therapy and were included in our analysis as failures whether PSA nadired or not. Whether to include or exclude these patients is not well defined in the literature. Swindle et al.^[@B6]^ confirmed the persistent significance of positive surgical margins on BRFS when patients receiving adjuvant therapy were either considered to recur at the time of adjuvant therapy or were excluded. Although our database contains few patients receiving adjuvant therapy, when a similar analysis was performed our outcomes did not change significantly (results not shown). We acknowledge that the retrospective nature of this study may allow for selection bias. Although patients with positive and negative margins differed in pathologic and clinical characteristics, subgroup analyses by D'Amico risk, TNM stage, and Gleason grade attempted to minimize the above differences. Patients who had insufficient clinical and pathological data, who were lost to follow-up, or who chose to follow-up elsewhere were not captured by this study, although every effort was made to do so. Analysis of available data for 278 excluded patients compared with the 1159 included patients did not show any significant difference except for the frequency of pelvic lymph node dissection, which was lower in the excluded group (48.6% vs. 58.9%). Performance of lymph node dissection was at the surgeon\'s discretion, and lower frequency of lymphadenectomy in the excluded patients suggests that this group was at a lower risk for nodal metastasis and recurrence. Thus, if these patients were included, resultant BRFS would likely increase. This study is also limited by the length of follow-up. Early stage/low-grade prostate cancer is indolent and BCR may not be captured within our follow-up interval, thus possibly minimizing the significance of SM. Because we are a tertiary care center, most of our patients receive follow-up care with their local urologists, and we are exploring other avenues for data capture such as on-line surveys, automated calls, and other such things. Future studies with longer follow-up would allow evaluation of more direct metrics, such as disease-specific survival, metastasis-free survival, and overall survival. CONCLUSIONS =========== This study confirms the significance of SM on BRFS in a large cohort of RALP patients. BCR rates for robotic, laparoscopic, and open series appear to be comparable. Our BRFS further validates a robotic-assisted approach for the treatment of prostate cancer. Longer follow-up is needed to better define the impact of margin status on BCR. A multi-institutional approach would further strengthen the role of RALP as comparable to radical retropubic prostatectomy.

1. Introduction {#sec1-cancers-12-00417} =============== Small-molecule inhibitors such as tyrosine kinase inhibitors (TKIs) target intracellular signal pathways and are designed to cross cellular membranes passively due to their size and lipophilicity. Inside the cell they may exert their action by blocking the ATP-binding pocket of their target protein, thereby inhibiting protein signal transduction. Drug distribution within the cell is therefore an important consideration for drug efficacy. Sunitinib is an orally administered, multi-targeted TKI approved by the United States Food and Drug Administration and the European Medicines Agency for the treatment of imatinib-resistant gastrointestinal stromal tumor, advanced renal cell carcinoma and advanced pancreatic neuroendocrine tumors \[[@B1-cancers-12-00417]\]. Sunitinib exerts its anti-cancer effect directly both on the tumor vasculature and tumor cells \[[@B2-cancers-12-00417]\]. In addition, sunitinib has been shown to inhibit tumor growth by inducing anti-tumor immunity, through reduction of both myeloid-derived suppressor cells (MDSCs) and T regulatory cells (Tregs) \[[@B3-cancers-12-00417],[@B4-cancers-12-00417]\]. The main targets of sunitinib are vascular endtothelial growth factor receptors (VEGFRs), platelet-derived growth factor receptors (PDGFRs), fms-related tyrosine kinase 3 (FLT-3), the stem cell factor receptor (KIT) and the colony stimulating factor 1 (CSF-1), which all are inhibited at nanomolar concentrations \[[@B5-cancers-12-00417]\]. Sunitinib is a weak base (pKa = 8.95, [Figure S1](#app1-cancers-12-00417){ref-type="app"}) and is therefore subjected to increased protonation at decreasing pH values in the physiological pH span. Consequently, protonated sunitinib is sequestered in acidic lysosomes, which impact its intracellular distribution and may reduce its target interaction and, hence, drug efficacy \[[@B6-cancers-12-00417],[@B7-cancers-12-00417]\]. Lysosomal sequestration of sunitinib has previously been shown in acquired sunitinib-resistant HT-29 colon cancer cells \[[@B7-cancers-12-00417]\]. The impact of sequestration on sunitinib resistance has, however, not been studied. Lysosomal sequestration is not only limited to sunitinib, but is also observed for other anti-cancer weak-basic drugs such as other TKIs (gefitinib, lapatinib \[[@B8-cancers-12-00417]\]), anthracyclines (doxorubicin \[[@B9-cancers-12-00417]\], daunorubicin \[[@B10-cancers-12-00417]\]) and preclinical imidazocridinones \[[@B6-cancers-12-00417],[@B11-cancers-12-00417],[@B12-cancers-12-00417]\]. Photochemical internalization (PCI) represents a clinically relevant treatment modality for release of drugs that are entrapped in endosomes and lysosomes \[[@B13-cancers-12-00417],[@B14-cancers-12-00417],[@B15-cancers-12-00417]\]. This method is based on amphiphilic photosensitizers (PSs), such as meso-tetraphenylchlorin disulfonate (TPCS~2a~) ([Figure S1](#app1-cancers-12-00417){ref-type="app"}), which localize in the membrane of endocytic vesicles \[[@B14-cancers-12-00417]\]. Upon illumination, reactive oxygen species (ROS) generated during photochemical reactions destabilize the vesicle membrane and induce cytosolic release of the content entrapped in these vesicles \[[@B16-cancers-12-00417],[@B17-cancers-12-00417]\]. Drugs which accumulate in endo/lysosomal compartments can in this way be released to interact with their intracellular target. PCI may be applied with two different protocols; in the "light after" protocol light exposure is applied after administration of the drug to be released, while in the "light first" protocol the photochemical reaction is generated before administration of the drug \[[@B18-cancers-12-00417]\]. PCI was first documented as a delivery technology of hydrophilic macromolecules which do not readily penetrate the plasma membrane. Recent evidence has, however, indicated the technology as efficient also for small molecules entrapped in endocytic vesicles, such as gemcitabine and doxorubicin \[[@B13-cancers-12-00417],[@B19-cancers-12-00417],[@B20-cancers-12-00417],[@B21-cancers-12-00417]\]. Here, we evaluated PCI as a strategy to release lysosomal sequestered sunitinib, thereby enhancing the cytotoxicity of sunitinib in cancer cells. 2. Results {#sec2-cancers-12-00417} ========== 2.1. Lysosomal Localization of Sunitinib and TPCS~2a~ {#sec2dot1-cancers-12-00417} ----------------------------------------------------- PCI is dependent on endo/lysosomal localization of both the PS and the molecule subjected to cytosolic release. Fluorescence microscopy of sunitinib and TPCS~2a~ was therefore employed to evaluate the intracellular localization of both compounds. Granular fluorescence of sunitinib was detected after 24 h incubation in HT-29 cells ([Figure 1](#cancers-12-00417-f001){ref-type="fig"}a). The fluorescence was to a large degree overlapping with LysoTracker Red in agreement with a previous report by Nowak-Sliwinska et al. \[[@B22-cancers-12-00417]\], indicating accumulation of sunitinib in endo/lysosomal vesicles. TPCS~2a~ enters the cell by means of adsorptive endocytosis and accumulates in the membrane of endo/lysosomal vesicles \[[@B13-cancers-12-00417],[@B16-cancers-12-00417]\]. An 18 h incubation of TPCS~2a~ resulted in granular fluorescence, partly overlapping with LysoTracker, indicating endo/lysosomal localization in HT-29 cells ([Figure S2](#app1-cancers-12-00417){ref-type="app"}). Fluorescence was, however, also detected on the plasma membrane indicating non-endocytosed PS. Plasma membrane associated TPCS~2a~ was largely removed by a 4 h incubation in TPCS~2a~-free medium, resulting in an overall granular pattern of fluorescence. Thus, TPCS~2a~ accumulates mainly in endo/lysosomes in HT-29 cells when used as in a standard PCI protocol ([Figure 1](#cancers-12-00417-f001){ref-type="fig"}b, left panel) in agreement with previous reports \[[@B14-cancers-12-00417],[@B16-cancers-12-00417]\]. Photochemical destabilization of the endo/lysosomal membrane and subsequent cytosolic release of the entrapped drug of interest is the basic mechanism behind PCI. Light exposure was here shown to relocalize TPCS~2a~ to the cytosol in HT-29 cells, which corresponded with a reduction of the LysoTracker signal ([Figure 1](#cancers-12-00417-f001){ref-type="fig"}b, right panel). 2.2. No Enhanced Toxicity by PCI "Light After" of Sunitinib {#sec2dot2-cancers-12-00417} ----------------------------------------------------------- The fluorescence images in [Figure 1](#cancers-12-00417-f001){ref-type="fig"}a,b reveal endo/lysosomal localization of both TPCS~2a~ and sunitinib, and it was therefore expected that light activation of the photosensitizer would result in cytosolic release of sunitinib. This PCI protocol was in accordance with the PCI "light after" procedure where light exposure is applied after administration of the drug to be released. No enhanced cytotoxicity was, however, indicated following PCI of sunitinib exposing the cells to blue light from LumiSource™ ([Figure 1](#cancers-12-00417-f001){ref-type="fig"}c). The observed combined effect was found to be slightly higher than the theoretical additive effect, and the synergy/antagonism parameter difference in log (DL) indicated a negative value −0.089 ± 0.075 although not significantly different from additively (*p* = 0.367). The absorption maximum for sunitinib has previously been reported at 429 nm, which is near the maximum emission wave length of blue light source (λ~max~ = 437 nm) \[[@B22-cancers-12-00417]\]. The PS TPCS~2a~ is also activated at its secondary maxima λ = 652 nm, allowing the circumvention of a putative blue-light induced inactivation of sunitinib. However, no increase in cytotoxicity was observed by PCI of sunitinib with the red light source ([Figure 1](#cancers-12-00417-f001){ref-type="fig"}d) yielding a slightly negative DL value −0.023 ± 0.09 (*p* = 0.834, not significant). PCI of the protein toxin gelonin (rGel) was included as a positive control for the PCI procedure, which resulted in synergistic cytotoxicity between rGel and the photochemical treatment, supported with a positive DL 0.262 ± 0.0048 (*p* \< 0.001) ([Figure 1](#cancers-12-00417-f001){ref-type="fig"}e). Hence, PCI "light after" does not enhance the efficacy of sunitinib. 2.3. Sunitinib Is a Target for ROS-Mediated Photodamage {#sec2dot3-cancers-12-00417} ------------------------------------------------------- We investigated if the lack of enhanced cytotoxicity of sunitinib-PCI ("light after") could be explained by ROS mediated photodamage of sunitinib. Singlet oxygen (^1^O~2~) is considered as the most important ROS formed during photochemical treatment as applied in this work \[[@B23-cancers-12-00417],[@B24-cancers-12-00417]\]. The short half-life (\<0.04 µs) and diffusion distance (10--20 nm) of singlet oxygen in cellular membranes \[[@B25-cancers-12-00417]\] implicate that TPCS~2a~ should be in close intracellular vicinity of sunitinib in order to induce photochemical damage of the TKI. Super-resolution microscopy was therefore performed in order to evaluate the subcellular/suborganellar localization of TPCS~2a~ and sunitinib in detail. TPCS~2a~ and sunitinib partially co-localized in ring-like structures in single optical sections, indicating both compounds to be associated with vesicular membranes ([Figure 1](#cancers-12-00417-f001){ref-type="fig"}f). These results are in support for ROS-mediated photochemical damage of sunitinib. Photodamage of sunitinib in the presence of TPCS~2a~ was further evaluated by absorption and fluorescence spectroscopy in solutions at pH 7 containing 1% fetal bovine serum (FBS) to solubilize these compounds. The emission spectra of both sunitinib and TPCS~2a~ prepared without light exposure were attenuated when they were combined ([Figure 1](#cancers-12-00417-f001){ref-type="fig"}g). However, the sunitinib fluorescence was reduced by approximately 50% while that of TPCS~2a~ was only reduced by \~20% ([Figure S3](#app1-cancers-12-00417){ref-type="app"}). The fluorescence spectrum of sunitinib overlaps well the 4 Q-band absorption spectrum of TPCS~2a~ \[[@B16-cancers-12-00417]\]. Thus, these results may be due to Förster resonance energy transfer (FRET) between sunitinib and TPCS~2a~ i.e., emission from sunitinib is absorbed by TPCS~2a~ and added to the directly excited TPCS~2a~. FRET may occur if the distance between the donor (sunitinib) and acceptor (TPCS~2a~) is short enough, typically 1--10 nm and is in line with the close proximity of the drugs in endo/lysosomal membranes \[[@B26-cancers-12-00417]\]. The light exposure of both sunitinib and TPCS~2a~ separately lead to a smaller attenuation of sunitinib fluorescence (28%) than of TPCS~2a~ fluorescence (57%) ([Figure 1](#cancers-12-00417-f001){ref-type="fig"}g, table). When sunitinib and TPCS~2a~ was combined and exposed to light the TPCS~2a~ fluorescence was reduced to the same extent as in the absence of sunitinib, while the reduction in sunitinib fluorescence was much stronger in the presence of TPCS~2a~. These results indicate that the photooxidation of TPCS~2a~ is independent of the presence of sunitinib, while photoactivation of TPCS~2a~ contribute to the photooxidation of sunitinib. Similar results were observed with the absorption spectra where the broad absorption peak, 330--495 nm of sunitinib disappeared after light exposure of TPCS~2a~ ([Figure S4](#app1-cancers-12-00417){ref-type="app"}). Photochemical targeting of the endosomal membrane has previously been shown to release H^+^ from the endosomal lumen and thereby increase the endosomal pH prior to escape of endocytosed molecules \[[@B27-cancers-12-00417]\] and pH 7 was therefore selected as the experimental condition. Nevertheless, the absorption spectra were also collected at pH 5 showing similar results as for pH 7 ([Figure S4](#app1-cancers-12-00417){ref-type="app"}). These results therefore imply that sunitinib is photodamaged after light exposure of TPCS2a, most likely due to single oxygen-induced photooxidation. 2.4. Synergistic Cytotoxicity by "Light First" Sunitinib-PCI {#sec2dot4-cancers-12-00417} ------------------------------------------------------------ PCI has usually been performed as in [Figure 1](#cancers-12-00417-f001){ref-type="fig"}c,d with the photochemical reaction initiated after administration of the drug of interest in combination with PS ("light after" procedure). However, PCI of non-targeting therapeutics has also been shown just as effective when the drug is administrated after the photochemical treatment ("light first" procedure) \[[@B17-cancers-12-00417],[@B18-cancers-12-00417]\]. The suggested mechanism behind the "light first" procedure is fusion between photochemically damaged vesicles and intact vesicles containing the drug of interest \[[@B18-cancers-12-00417]\]. Circumvention of the photochemical destruction of sunitinib during sunitinib-PCI was therefore attempted by applying PCI with the "light first" procedure. Indeed, treatment of HT-29 cells with sunitinib immediately after TPCS~2a~ incubation and light exposure was found to induce a synergistic reduction of viability as measured by the MTT assay ([Figure 2](#cancers-12-00417-f002){ref-type="fig"}a). The survival after sunitinib-PCI was found to be significantly lower than the theoretical additive effect at 8 µM sunitinib (*p* = 0.015) ([Figure 2](#cancers-12-00417-f002){ref-type="fig"}b) supported by a synergy/antagonism parameter DL of 0.84 ± 0.49 (*p* = 0.0035), indicating synergism. The effect of sunitinib-PCI was further verified by clonogenic assay indicating reduction of clonal cell survival by the PCI treatment ([Figure 2](#cancers-12-00417-f002){ref-type="fig"}c). The "light first" procedure was also applied to CT26.WT cells confirming enhanced reduction in cell viability by PCI of sunitinib ([Figure 2](#cancers-12-00417-f002){ref-type="fig"}d). At 2 µM sunitinib in CT26.WT, the DL value was 0.32 ± 0.046 (*p* = 0.019). Thus, PCI with the "light first" procedure enhances sunitinib cytotoxicity in both HT-29 and CT26.WT colon cancer cell lines in a synergistic manner. Fluorescence imaging was performed to confirm cytosolic release of sunitinib ([Figure 2](#cancers-12-00417-f002){ref-type="fig"}e). Cytosolic release of sunitinib should result in a change in fluorescence from granules (endocytic vesicles) to diffuse fluorescence from the cytoplasm. The PCI "light after" protocol, with the photochemical treatment applied after sunitinib incubation, showed little difference in the fluorescence pattern of sunitinib upon light exposure, indicating poor cytosolic release with this procedure. This was in contrast to the PCI "light first" protocol where a pronounced cytosolic fluorescence from sunitinib was observed upon light exposure. Thus, PCI with the "light first" procedure was indicated to release sunitinib from endosomal compartments, while PCI with the "light after" procedure was not. 2.5. Acquired Resistance and Endo/Lysosomal Accumulation Following Prolonged Sunitinib Exposure {#sec2dot5-cancers-12-00417} ----------------------------------------------------------------------------------------------- Lysosomal accumulation of sunitinib has been reported as a mechanism of resistance \[[@B7-cancers-12-00417],[@B28-cancers-12-00417],[@B29-cancers-12-00417],[@B30-cancers-12-00417],[@B31-cancers-12-00417],[@B32-cancers-12-00417]\] and we tested if PCI could counteract this mechanism by releasing endosomal sunitinib sequestered during long-term exposure. Sunitinib resistant HT-29 cells, HT-29/SR, was generated by continuous exposure to 2 µM. The low drug dose of 2 µM sunitinib reduced the viability in parental cells by approximately 20% ([Figure 3](#cancers-12-00417-f003){ref-type="fig"}a). The sunitinib concentration required to reduce the cell viability by 40% was 1.8 ± 0.08 (*p* \< 0.001) fold higher in HT-29/SR cells compared to the parental cells. The sunitinib response in HT-29/SR cells measured by MTT assay was found similar within the time frame of these experiments indicating a stabilized sunitinib response in HT-29/SR cells ([Figure 3](#cancers-12-00417-f003){ref-type="fig"}a). The decreased sensitivity to sunitinib in HT-29/SR cells was demonstrated even more substantial by clonal cell survival as compared to overall cell viability (MTT) ([Figure 3](#cancers-12-00417-f003){ref-type="fig"}b). Overall, sunitinib sensitivity in HT-29 cells is higher than previously reported \[[@B7-cancers-12-00417]\]. This is probably due to shorter incubation times (72 h versus 96 h). The proliferation rate of HT-29/SR and HT-29 in the presence of sunitinib was also investigated ([Figure 3](#cancers-12-00417-f003){ref-type="fig"}c). A 2 µM sunitinib incubation for 120 h resulted in 42% confluence in the HT-29/SR cells while only 31% confluence was observed in the parental HT-29 cell line. The same tendency was shown for all sunitinib concentrations tested. Thus, the proliferation rate of HT-29/SR cells in the presence of 1--10 µM sunitinib was increased compared to the parental cells, further documenting the successful generation of sunitinib-resistant cells. High degree of co-localization of sunitinib with LysoTracker Red confirmed lysosomal sequestering of sunitinib in HT-29/SR cells as verified by fluorescence microscopy ([Figure 3](#cancers-12-00417-f003){ref-type="fig"}d). The fluorescence images indicated higher fluorescence signals from sunitinib in HT-29/SR cells compared to the parental HT-29 cells ([Figure 1](#cancers-12-00417-f001){ref-type="fig"}a). Accumulation of sunitinib in HT-29 cells was further documented by flow cytometry, indicating a\~1.8 fold higher accumulation in HT-29/SR cells that have been continuously incubated with sunitinib compared to parental cells subjected to 72 h sunitinib incubation (*p* = 0.03, one-tailed *p* value) ([Figure 3](#cancers-12-00417-f003){ref-type="fig"}e). Accumulated sunitinib in HT-29/SR cells was released when removing sunitinib from the media for 24 h ([Figure S5](#app1-cancers-12-00417){ref-type="app"}). In agreement with this, there was no difference in sunitinib accumulation between HT-29 and HT-29/SR cells when a 24 h wash in drug-free medium was added before the 72 h sunitinib incubation ([Figure S5](#app1-cancers-12-00417){ref-type="app"}). 2.6. Photochemical Release of Sequestered Sunitinib Does Not Abolish Sunitinib Resistance in HT-29/SR {#sec2dot6-cancers-12-00417} ----------------------------------------------------------------------------------------------------- As sunitinib was found subjected to photochemical damage by TPCS~2a~ and light, it was expected that sunitinib-PCI with "light after" would not induce synergistic effects on cell viability. Indeed, no difference in light-induced toxicity was found between TPCS~2a~-treated HT-29/SR and HT-29 cells ([Figure 3](#cancers-12-00417-f003){ref-type="fig"}f), and PCI with "light after" procedure yielded no statistically significant difference between sunitinib alone and PCI "light after" at 8 µM sunitinib ([Figure 3](#cancers-12-00417-f003){ref-type="fig"}g). The antagonism/synergy DL parameter −0.161 ± 0.154 (*p* = 0.49, not significant) indicated additivity, but towards antagonism similar to the parental HT-29 cells ([Figure 1](#cancers-12-00417-f001){ref-type="fig"}c,d). PCI with the "light first" strategy induced reduction in HT-29/SR cell viability ([Figure 3](#cancers-12-00417-f003){ref-type="fig"}h, *p* = 0.044 at 8 µM sunitinib compared to PCI). The effect was further evaluated using the DL parameter and was at 8 µM sunitinib 0.049 ± 0.094 (*p* = 0.66, not significant) indicating that the effect was additive with this strategy. Thus, the sunitinib resistance was not abolished by the sunitinib concentration applied to reduce viability with neither "light first" nor "light after" PCI. Notably, PCI of rGel with "light after" strategy was found similarly efficient in both HT-29 ([Figure 1](#cancers-12-00417-f001){ref-type="fig"}e) and HT-29/SR implying rGel-PCI to circumvent sunitinib resistance ([Figure 3](#cancers-12-00417-f003){ref-type="fig"}i). 2.7. Modest HT-29 Tumor Growth Delay After Sunitinib-PCI in Athymic Mice {#sec2dot7-cancers-12-00417} ------------------------------------------------------------------------ Based on the in vitro results, PCI with "light first" protocol, was selected for the in vivo evaluation of sunitinib-PCI in HT-29 xenografts in athymic mice. rGel-PCI has previously been reported as least as efficient with the "light first" procedure as with the "light after" procedure in vivo \[[@B17-cancers-12-00417]\]. The "light first" effect is previously indicated up to \~8 h post-photochemical treatment \[[@B18-cancers-12-00417]\] while sunitinib reaches a maximum plasma concentration 3--6 h post-oral administration \[[@B33-cancers-12-00417]\]. In the first in vivo treatment protocol (Sun1-PCI) sunitinib was therefore administrated 3 h prior to and 30 min after light exposure. Fluorescence microscopy of frozen tumor sections revealed the presence of both TPCS~2a~ and sunitinib in the tumor at the point of light exposure ([Figure 4](#cancers-12-00417-f004){ref-type="fig"}a). The granular TPCS~2a~ fluorescence indicated localization in endosomes and lysosomes. The sunitinib fluorescence was, however, mostly weak and too diffuse to make any conclusions on intracellular localizations. However, some overlap in granular sunitinib and TPCS~2a~ fluorescence was detected in the frozen sections indicating in vivo co-localization. Light exposure of the tumors at 15 J/cm^2^ induced a reduction of both TPCS~2a~ and sunitinib fluorescence ([Figure 4](#cancers-12-00417-f004){ref-type="fig"}b), most likely reflecting photochemical damage of sunitinib as observed in vitro and photobleaching of TPCS~2a~ ([Figure 1](#cancers-12-00417-f001){ref-type="fig"}g) \[[@B16-cancers-12-00417]\]. Sun1-PCI did not increase the overall treatment effect compared to PS + light only or Sun1 only, as shown in the Kaplan-Meier plot and in the estimated mean time to reach endpoint (tumor size \< 900 mm^3^) ([Figure 4](#cancers-12-00417-f004){ref-type="fig"}c,e). In an effort to enhance the anti-tumor efficacy, we extended the duration of sunitinib treatment after light exposure by additional administration on day 1, 2, 3 and 4 post-light exposure (Sun2 protocol). The estimated time to reach endpoint was significantly increased compared to untreated controls in all groups receiving sunitinib, and the longest estimated time to reach endpoint was found for the Sun2-PCI-treated animals (25 d compared to 12 d for untreated animals) ([Figure 4](#cancers-12-00417-f004){ref-type="fig"}d,e). The growth curves of each individual animal indicated tumor growth delay the first 10 days after Sun1-PCI and Sun2-PCI ([Figure S6](#app1-cancers-12-00417){ref-type="app"}). The average tumor size at day 6 and 10 after treatment was therefore assessed in all treatment groups ([Figure 4](#cancers-12-00417-f004){ref-type="fig"}f). The one-way ANOVA test revealed significant differences between the treatment groups at both time points (*p* \< 0.001 at day 6, and *p* = 0.013 at day 10). Further pair-wise multiple comparison (Holm-Sidak method) indicated significant PCI-Sun1-, Sun2- and PCI-Sun2-induced tumor growth delay at 6 days compared to untreated controls ([Figure 4](#cancers-12-00417-f004){ref-type="fig"}g). The tumor volume of PCI-Sun1-treated animals was also significantly smaller than in Sun1-treated animals at day 6. However, no significant difference was found in tumor size between Sun1-PCI- and Sun2-PCI-treated animals at day 6 and 10. The only significant difference at day 10 was found between Sun1-PCI and untreated controls ([Figure 4](#cancers-12-00417-f004){ref-type="fig"}g). Thus, even though sunitinib-PCI showed some initial anti-tumor efficacy compared to non-treated controls, the overall treatment effect was less than expected. Since PCI in vitro could not circumvent sunitinib resistance, the HT-29/SR model was not continued in vivo. 2.8. Poor CT26.WT Tumor Growth Delay After Sunitinib-PCI in Immunocompetent Mice {#sec2dot8-cancers-12-00417} -------------------------------------------------------------------------------- We have previously shown that T-cell activation is important to achieve curative effects after PCI \[[@B34-cancers-12-00417],[@B35-cancers-12-00417]\]. In addition, sunitinib is shown to stimulate anticancer immune response \[[@B3-cancers-12-00417],[@B4-cancers-12-00417]\]. Hence, sunitinib-PCI was explored in CT26.WT tumor-bearing thymic mice. The overall tumor responses of sunitinib-PCI was, however, smaller in the CT26.WT model compared to the HT-29 model ([Figure 5](#cancers-12-00417-f005){ref-type="fig"}a,b). The normalized growth curves indicated a response at early time points ([Figure S7](#app1-cancers-12-00417){ref-type="app"}). Although not significant, the estimated time to reach endpoint was found increased in all the treatment groups compared to untreated controls ([Figure 5](#cancers-12-00417-f005){ref-type="fig"}c). Early treatment responses were evaluated by comparing average tumor sizes in the different treatment groups at day 4 and 7 post-light exposure ([Figure 5](#cancers-12-00417-f005){ref-type="fig"}d,e). Significant difference among the treatment groups was found at day 4 (*p* = 0.015, one-way ANOVA), associated with a significant \~50% reduction in tumor volume in the Sun1-PCI group compared to untreated controls (Holm-Sidak method) ([Figure 5](#cancers-12-00417-f005){ref-type="fig"}e). No significant difference was found between any of the other treatment groups at day 4 ([Figure 5](#cancers-12-00417-f005){ref-type="fig"}e). Furthermore, no significant difference among the treatment groups was found at 7 d post-light exposure in this model (*p* = 0.216, one-way ANOVA) ([Figure 5](#cancers-12-00417-f005){ref-type="fig"}e). 2.9. Tumor Tissue Response {#sec2dot9-cancers-12-00417} -------------------------- The poor response of sunitinib-PCI in the two tumor models was surprising in the light of the promising in vitro data on the "light first" procedure, especially in the CT26.WT model in immunocompetent mice where a strong treatment response was expected. The CT26.WT tumors at endpoint (900 mm^3^) revealed irregular shaped tumors in the PCI-treated animals as compared to the spherical shaped controls. Tumors following the Sun2-PCI protocol and corresponding controls were therefore harvested at endpoint and further evaluated for necrotic-, vascular- and immune-mediated treatment responses by hematoxylin and eosin (H&E) and immunohistochemistry (IHC) staining. Treatment-induced necrosis was detected in PS + light and Sun2-PCI-treated tumors compared to untreated and Sun2-treated tumors where no necrosis was observed ([Figure 6](#cancers-12-00417-f006){ref-type="fig"}a). The necrotic area in the Sun2-PCI-treated tumors were larger compared to in the PS + light group indicating more pronounced cancer parenchymal cell death in agreement with the in vitro results. The immunologic anti-tumor response of light activated TPCS~2a~ is dependent on T-cells infiltrating the treated area \[[@B34-cancers-12-00417],[@B36-cancers-12-00417]\]. An increase in tumor-associated-CD3-positive T-cells was observed in both PS + light and Sun2-treated tumors compared to untreated controls ([Figure 6](#cancers-12-00417-f006){ref-type="fig"}b). Quantification of CD3-stained cells revealed a significant 5- and 3-fold increase in PS + light- and Sun2-treated sections respectively compared to untreated tumors ([Figure 6](#cancers-12-00417-f006){ref-type="fig"}c). No increase was, however, detected in Sun2-PCI-treated tumors ([Figure 6](#cancers-12-00417-f006){ref-type="fig"}c), indicating an antagonistic effect on tumor infiltrating T-cells in this treatment group. A strong Ki-67 staining, indicating proliferating cells, was found in all viable areas in agreement with the H&E stains ([Figure 6](#cancers-12-00417-f006){ref-type="fig"}d), further indicating a stronger treatment response in Sun2-PCI-treated tumors compared to all the other treatment groups. TPCS~2a~ and light has previously been shown to target the tumor vasculature, and sunitinib efficacy has been strongly associated with a vascular response \[[@B16-cancers-12-00417],[@B37-cancers-12-00417],[@B38-cancers-12-00417]\]. IHC with an αCD31 antibody was therefore applied to evaluate vascular responses of the combined treatment. A vascular treatment response was found in both photochemical-treated (PS + light), Sun2-treated and Sun2-PCI-treated tumor tissues, as visualized by rounded fragmented vessels compared to in the untreated controls. No additional effect was, however, found in the Sun2-PCI-treated tumors compared to the PS + light and Sun2 monotherapies ([Figure 6](#cancers-12-00417-f006){ref-type="fig"}e). 3. Discussion {#sec3-cancers-12-00417} ============= Lysosomal degradation is a resistance mechanism for several intracellular acting anticancer therapeutics of different chemical and pharmacologic classes \[[@B39-cancers-12-00417],[@B40-cancers-12-00417]\]. Drugs may accumulate in lysosomes through several ways including different forms of endocytosis \[[@B39-cancers-12-00417]\], transporter-mediated uptake \[[@B6-cancers-12-00417]\] and lysosomal sequestration of hydrophobic weak bases \[[@B12-cancers-12-00417]\]. Within the lysosomes the drugs are inhibited to interact with their cytosolic target \[[@B41-cancers-12-00417]\]. PCI is designed to overcome lysosomal entrapment of anticancer therapeutics, and the method has been shown to potentiate the activity of a variety of drugs including protein toxins, RNA, DNA, nanoparticles and also some conventional chemotherapeutic drugs \[[@B13-cancers-12-00417],[@B14-cancers-12-00417]\]. This is the first report demonstrating direct mechanistic evidence of endo/lysosomal membrane localization and light-controlled cytosolic release of a smallmolecule inhibitor sequestered in endo/lysosomal compartments. Enhancement of sunitinib toxicity with PCI was, however, highly dependent on the treatment protocol, as no increase in cytotoxicity was observed when sunitinib was administrated prior to the photochemical treatment ("light after" strategy, [Figure 1](#cancers-12-00417-f001){ref-type="fig"}c,d). Using super-resolution fluorescence microscopy we were able to show subcellular co-localization of sunitinib and TPCS~2a~ ([Figure 1](#cancers-12-00417-f001){ref-type="fig"}f)~.~ Our data indicate that sunitinib is photodamaged at pH 7 ([Figure 1](#cancers-12-00417-f001){ref-type="fig"}g), which is in agreement with Ohtsuki et al. who report increased endosomal pH prior to release of endocytosed molecules \[[@B27-cancers-12-00417]\]. The absorption spectra ([Figure S4](#app1-cancers-12-00417){ref-type="app"}) suggest that photodamage of sunitinib by TPCS~2a~ can also occur at pH 5. However, due to the amphiphilic nature of TPCS~2a~ \[[@B42-cancers-12-00417]\], the PS will localize and accumulate in the cellular membranes in an in vitro and in vivo setting. This is in agreement with our SIM microscopy images ([Figure 1](#cancers-12-00417-f001){ref-type="fig"}f). Hence, the very close vicinity of these two drugs facilitates ROS-mediated photodamage of sunitinib after light-activation of TPCS~2a~ ([Figure 1](#cancers-12-00417-f001){ref-type="fig"}f,g). The light-triggered, ROS-induced damage of sunitinib most likely also took place when sunitinib-PCI was applied to the sunitinib resistant HT-29/SR cells. PCI was therefore unlikely to release and potentiate sunitinib that was accumulated in the in endo/lysosomal compartments in HT-29/SR cells. This was confirmed by the similar response to photochemical treatment in HT-29 and HT-29/SR cells, indicating no enhanced toxicity from cytosolic release of sunitinib in HT-29/SR cells ([Figure 3](#cancers-12-00417-f003){ref-type="fig"}i). Thus, ROS-mediated photodamage of sunitinib is probably one reason why the photochemical treatment applied here failed to potentiate lysosomal sequestrated sunitinib in resistant cells. As observed in the parental HT-29 cells, sunitinib-PCI with the "light first" protocol was also found to increase the cell toxicity in HT-29/SR cells. This effect was however found to be additive at the highest concentration tested, indicating that the mechanism of sunitinib resistance is not related to lysosomal sequestration. Lysosomal sequestering of sunitinib has been associated with resistance by others \[[@B7-cancers-12-00417],[@B29-cancers-12-00417],[@B31-cancers-12-00417]\]. To our knowledge, direct experiments addressing the cytotoxic impact of lysosomal accumulation of sunitinib in resistant cells has only been included in two reports where only modest increase in sunitinib toxicity was reported by combining sunitinib with Leu-Leu-O-Methyl in vitro \[[@B29-cancers-12-00417]\] or chloroquine in vivo in sunitinib-resistant HT-29 xenografts \[[@B30-cancers-12-00417]\]. These reports therefore also indicate other mechanisms than lysosomal sequestering to orchestrate sunitinib resistance. Further analysis on molecular pathways controlling cellular sunitinib-resistance is needed to conclude on the mechanisms of sunitinib-resistance in the HT-29/SR cells. Even though PCI failed to activate lysosomal sequestrated sunitinib in resistant cells, we here show that the technology can be applied to circumvent sunitinib resistance when combined with the protein toxin gelonin. PCI of protein toxins has previously been shown to circumvent resistance mediated through increased expression of P-glycoprotein \[[@B43-cancers-12-00417],[@B44-cancers-12-00417]\] and the present results in the HT-29/SR cells further documents this strategy as highly efficient for the treatment of therapy resistant cancer. Even though sunitinib-PCI with the "light first" protocol was shown highly efficient in vitro, the treatment did not result in synergistic tumor growth delay in mice. Continuous administration of sunitinib has previously been reported to induce a strong tumor growth delay in similar in vivo models \[[@B30-cancers-12-00417],[@B45-cancers-12-00417]\]. Here, a moderate dose of sunitinib (two or six doses only) and TPCS~2a~ + light was selected to reach a therapeutic window for synergy evaluations. The lack of an increased treatment response with sunitinib-PCI may be due to a narrow therapeutic window. On the other hand, the in vitro data suggest PCI to potentiate sunitinib induced cytotoxicity at low sunitinib concentrations and doses of light. The "light first" protocol has been shown most efficient within \~2--3 h after the photochemical treatment \[[@B18-cancers-12-00417]\], while the estimated time for maximum tumor accumulated sunitinib post-oral administration is 3--4 h \[[@B33-cancers-12-00417]\]. There is therefore a possibility that the amount of sunitinib was too low to synergize with PCI. The lack of overall treatment response to sunitinib-PCI in vivo may also be a result of vascular responses. The photochemical treatment alone targets the tumor vasculature as shown by the IHC on CD31 in CT26.WT tumors in agreement with previous studies \[[@B37-cancers-12-00417],[@B38-cancers-12-00417]\]. This vascular response may inhibit sunitinib penetration into the tumor when administrated after light exposure, and thereby counteract the action of sunitinib-PCI. The H&E data and IHC on Ki-67 supplied here demonstrate, however, larger treatment-induced necrosis in CT26.WT tumor bearing thymic animals receiving sunitinib-PCI compared to any of the control groups. This indicates that sunitinib had access to tumor after the photochemical treatment and induced an additive or supra-additive effect that is not reflected in the tumor growth measurements. The lack of overall treatment response following sunitinib-PCI despite the apparent larger treatment-induced necrosis may reflect a balance between growth of viable cells in the tumor rim and rate of removal of necrotic tissue. Resistance towards vascular disruptive agents, including PS and light, has previously been associated with a residual tumor rim \[[@B46-cancers-12-00417],[@B47-cancers-12-00417]\]. The tumors studied here by H&E and IHC were harvested several days after treatment when the tumors reached their endpoints of 1000 mm^3^. The absence of distinct treatment-induced viable rims was therefore not unexpected, and may have been present at earlier time points. Enhanced efficacy of sunitinib-PCI may therefore be expected by adjuvant treatment targeting this viable rim. The treatment response following sunitinib-PCI in HT-29-bearing athymic mice was smaller than expected, and the response was even further reduced in CT26.WT-bearing immunocompetent mice ([Figure 5](#cancers-12-00417-f005){ref-type="fig"}a,b). PCI in combination with several different drugs has been shown to generate immune-mediated anticancer activity and the overall response has in general been better in immunocompetent mice compared to athymic mice \[[@B34-cancers-12-00417],[@B35-cancers-12-00417]\]. Since the tumor models in athymic and immunocompetent mice here are of different cell line origin, the two models cannot be directly compared. Nevertheless, the IHC data demonstrate an antagonistic effect on CD3+ tumor infiltrating cells in CT26.WT tumors following sunitinib-PCI compared to the agonistic effect observed post- sunitinib- or TPCS~2a~ + light- monotherapy. The combination of sunitinib with immunotherapeutic approaches is somewhat controversial. Sunitinib has been shown to enhance intratumoral infiltation of CD8+ T-cells in combination with a CD40 agonist, a member of the tumor necrosis factor (TNF) superfamily \[[@B48-cancers-12-00417]\] and to suppress immune regulatory cells in combination with celecoxib \[[@B49-cancers-12-00417]\]. On the contrary, sunitinib has been reported to inhibit anti-tumor vaccination by decreasing antigen presenting cells \[[@B50-cancers-12-00417]\] and to impair proliferation and function of phytohemagglutinin (PHA) stimulated human T-cells \[[@B51-cancers-12-00417]\], in agreement with our results showing a decrease in tumor infiltrating T-cells following sunitinib-PCI. Jaini et al. also showed that the decreased immunoresponse following sunitinib and tumor vaccination could be avoided by careful scheduling of the applied drugs, and that enhanced vaccination could be achieved by administration of sunitinib after the priming phase of the vaccination \[[@B50-cancers-12-00417]\]. It is therefore possible that the combination of sunitinib and TPCS~2a~+light would be more effective if sunitinib was delivered one week post-photochemical treatment. This procedure will, however, probably not induce cytosolic release of sunitinib from endocytic vesicles since the time-frame between the photochemical treatment and sunitinib administration will be too long. Sunitinib has previously been suggested as a photosensitizer for endo/lysosomal destruction and it was argued that this approach could be used for the release of sunitinib entrapped in endo/lysosomal compartments \[[@B22-cancers-12-00417]\]. As the activity of sunitinib is here shown to be reduced upon blue light exposure ([Figure 1](#cancers-12-00417-f001){ref-type="fig"}g), dual utilization of sunitinib as a TKI and a photosensitizer is probably little effective at the cellular level. The enhanced tumor growth delay observed by this approach may be due to pharmacologic interactions in the tumor and vascular system. The clinical potential of this suggestion is, however, highly limited due to the absorption maximum of sunitinib in the blue region where the tissue light penetration is only a few 100 µm into the tissue. 4. Materials and Methods {#sec4-cancers-12-00417} ======================== 4.1. Cell Lines and Cultivation {#sec4dot1-cancers-12-00417} ------------------------------- The human colorectal adenocarcinoma cell line HT-29 (ATCC HT-38™) and the murine colorectal carcinoma cell line CT26.WT (ATCC CRL-2638™) were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA). The cells were subcultured 2--3 times a week in McCoy's 5a medium (Sigma-Aldrich, St. Louis, MO, USA) and RPMI-1640 (Sigma-Aldrich), respectively. The culture media were supplied with 10% fetal bovine serum (FBS) (Thermo Fisher Scientific, Waltham, MA, USA), 100 U/mL penicillin and 100 U/mL streptomycin (Sigma-Aldrich). Sunitinib-resistant HT-29 cells, named HT-29/SR, were generated by continuous exposure to 2 µM sunitinib for up to 5 months and the resistance was routinely verified during this time frame. Untreated parental HT-29 cells were cultured in parallel with the HT-29/SR cells. The cell lines were maintained at 37 °C in a humidified atmosphere containing 5% CO~2~. 4.2. Drugs and Chemicals {#sec4dot2-cancers-12-00417} ------------------------ The photosensitizer (PS) TPCS~2a~ (PCI Biotech AS, Oslo, Norway) was dissolved at 0.4 mg/mL in 3% Tween 80, 2.8% mannitol, 50 mM Tris, pH 8.5 (all from Sigma-Aldrich), and kept protected from light at 4 °C. Sunitinib malate (PZ0012, Sigma-Aldrich) was for in vitro experiments dissolved in dimetylsulfoxid (DMSO) to a final concentration of 2.5 mM, stored as aliquots at −20 °C, and subjected to maximum two freeze-thaw cycles. Sunitinib malate for in vivo experiments was dissolved at a concentration of 8 mg/mL (20 mM) in a vehicle containing distilled water with 1.8% NaCl, 0.5% carboxymethylcellulose, 0.4% Tween 80 and 0.9% benzylalcohol. The pH of the solution was adjusted to 6.0. The sunitinib mixture was sonicated to achieve stable dispersion \[[@B45-cancers-12-00417]\] and used within 24 h after sonication. All sunitinib reagents were stored protected from light. Recombinant gelonin (rGel) was generously provided by Dr. Michael Rosenblum's laboratory at M.D. Anderson Cancer Center (Houston, TX, USA). Aliquots of rGel were stored in phosphate-buffered saline (PBS) at −20 °C. All experiments in vitro and in vivo with sunitinib and/or TPCS~2a~ were performed under subdued light. 4.3. In Vitro Light Sources {#sec4dot3-cancers-12-00417} --------------------------- Illumination of the cells was performed with an in-house made diode lamp or a LumiSource™ lamp (PCI Biotech AS). The diode lamp delivers red light (E~max~ = 650--652 nm) at a fluence rate of 6 mW/cm^2^. LumiSource™ consists of four 18-W Osram L 18/67 light tubes and delivers blue light (λ= 400--500, λ~max~ = 435 nm) at a fluence rate of 7.1--9.6 mW/cm^2^. 4.4. In Vitro PCI Treatment {#sec4dot4-cancers-12-00417} --------------------------- 4000 HT-29 or HT-29/SR cells/well or 1500 CT26.WT cells/well were seeded in 96-well plates (Nuncon Delta, Thermo Fisher Scientific) and allowed to attach overnight. For clonogenic assay, 500 cells were seeded in 6-well plates (Nuncon Delta, Thermo Fisher Scientific). In the "light after" protocol, the cells were incubated with sunitinib prior to light exposure. The cells were incubated with sunitinib for 48 h, at indicated concentrations, prior to an 18 h co-incubation with 0.4 µg/mL TPCS~2a~. The cells were incubated in total 66 h with sunitinib. At the end of the incubation, cells were washed with PBS twice and chased 4 h in drug-free mediums before light exposure in order to remove plasma membrane bound-TPCS~2a~. In the "light first" protocol, the cells were subjected to light before sunitinib incubation. The cells were incubated with 0.4 µg/mL TPCS~2a~ for 18 h, washed with PBS twice and chased in medium for 4 h before light exposure. Sunitinib at indicated concentrations was then added immediately after illumination and incubated for 72 h. For the MTT-assay, sunitinib was removed and treatment efficacy was assessed as described in [Section 4.5](#sec4dot5-cancers-12-00417){ref-type="sec"}. For the clonogenic assay, the sunitinib incubation was sustained until the end of the experiment. PCI of rGel was only performed with the PCI "light after" protocol. Here, rGel at indicated concentrations was added to the wells during the chase period (4 h) and replaced with fresh medium before light exposure. 4.5. Treatment Efficacy In Vitro; Colony Formation Capacity, Viability and Proliferation {#sec4dot5-cancers-12-00417} ---------------------------------------------------------------------------------------- Cell viability was assessed using the MTT (3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyltetrazolium bromide). MTT was evaluated 48 h post-light exposure in all PCI "light after" protocols, except from experiments in [Figure 1](#cancers-12-00417-f001){ref-type="fig"}c. Experiments using PCI "light first" protocol was evaluated using the MTT assay 72 h post-light. Cells were incubated with 0.25 mg/mL MTT (Sigma-Aldrich) for 3--4 h. The medium was then removed, and the formazan-crystals were solubilized in DMSO. Absorbance was measured at 570 nm using a plate reader (PowerWave XS2 microplate spectrophotometer) with the Gen5 software program (Biotek Instruments Inc., Winooski, VT, USA). The colony formation assay was performed 10--14 d post-treatment, when sufficiently large colonies (\>50 cells/colony) were formed in controls \[[@B52-cancers-12-00417]\]. The medium was removed, and the cells were washed with 0.9% NaCl solution and fixed with absolute ethanol for 10 min. The colonies were stained with methylene blue (Sigma-Aldrich, 12 mg/mL in 0.1% NaOH) for 5 min. Colonies were manually counted under a magnifying glass using an automatic E-Count™ colony counter pen (Heathrow Scientific, Vernon Hills, IL, USA). Real-time monitoring of cell proliferation was measured using IncuCyte FLR kinetic imaging system (Essen Bioscience, MI, USA). 4000 HT-29 or HT-29/SR cells/well were seeded out in 96-well plates. The cells were treated as indicated and monitored for up to a week. Phase-contrast images were acquired every 3 h in each well, processed using by IncuCyte software Rev2 and presented as cell confluence over time. 4.6. Intracellular Localization of TPCS~2a~ and Sunitinib by Fluorescence Microscopy {#sec4dot6-cancers-12-00417} ------------------------------------------------------------------------------------ HT-29 or HT-29/SR cells were seeded on cover slips (No. 1014/10, Assistent, Sondheim, Germany) in 48-well plates (3 × 10^4^ cells/well) and allowed to attach overnight. The cells were then incubated with 0.4 μg/mL TPCS~2a~ for 18 h and either studied directly after the end of incubation or after wash and a 4-h chase in drug-free medium to remove TPCS~2a~ from the plasma membrane as done in the PCI procedures. For intracellular detection of sunitinib, attached cells were incubated for 24 h with 2 µM sunitinib. HT-29/SR cells were seeded out in the presence of 2 µM sunitinib to maintain sunitinib accumulation in the cells. For evaluation of cytosolic release of sunitinib after PCI "light first" and "light after" procedure, 1.5 × 10^4^ HT-29 cells/well were seeded, allowed to attach and treated as described in [Section 4.4](#sec4dot4-cancers-12-00417){ref-type="sec"} to mimic the PCI protocols. For the "light after" procedure, cells were incubated with 2 µM sunitinib for 48 h, and co-incubated with 0.4 µg/mL TPCS~2a~ for 18 h. The cells were washed and chased 4 h in drug-free medium before light exposure. In the "light first" procedure, the cells were incubated with 0.4 µg/mL TPCS~2a~ for 18 h, washed and chased 4 h before light exposure. Sunitinib, 6 µM, was added immediately after light exposure. Image acquisition after light exposure was performed at least an hour after illumination to allow cytosolic release of the endo/lysosomal content. Lysosomes and endosomes were visualized by 30 min incubation with LysoTracker Red DND-99 or LysoTracker Green DND-26 (both from Life Technologies, Carlsbad, CA, USA) at 75 nM. Hoechst 33342 (Sigma-Aldrich) was added at 10 µM and incubated for 15 min to visualize the nucleus. Upon image acquisition, the cover slip was carefully removed from the well, washed twice in ice cold PBS with Ca^2+^ and Mg^2+,^ and inverted on a microscope slide. Image acquisition was performed with a Zeiss Axioplan epifluorescence and phase contrast microscope using 63x/NA1.4 PlanApo objective (Carl Zeiss AG, Oberkochen, Germany). The images were acquired with a cooled charge-coupled device (CCD) camera (AxioCam MRm camera, Carl Zeiss AG). The TPCS~2a~ fluorescence was recorded using a 395--440 nm excitation filter, a 460 nm dichroic mirror and a 620 nm long pass filter. For recording fluorescence from sunitinib or LysoTracker Green, a 450--490 nm band pass excitation filter, a 495 nm dichroic mirror and a 500--550 nm band pass emission filter was used. The LysoTracker Red fluorescence was recorded using a 595 nm excitation filter and a 620 nm emission filter. The AxioVision Analysis (Carl Zeiss AG) software program was used to process and analyze the images. 4.7. Subcellular Localization of Sunitinib and TPCS~2a~ {#sec4dot7-cancers-12-00417} ------------------------------------------------------- Wide-field and structured illumination microscopy (SIM) were performed in order to determine the subcellular localization of sunitinib and TPCS~2a~ within the endocytic vesicles in live cells. 3 × 10^5^ HT-29/SR cells were seeded in glass bottom dishes (Cat. no. P35GC-1.5-10-C, MatTek Corporation, Ashland, MA, USA) and allowed to attach overnight. The cells were co-incubated with 2 µM sunitinib and 0.4 µg/mL TPCS~2a~ for 18 h, washed twice with PBS and chased in serum-free FluoreBrite DMEM (Thermo Fisher Scientific). Trolox (Sigma-Aldrich) was added to the cells 15 min prior to image acquisition. SIM imaging was performed on a DeltaVision OMX V4 Blaze 3D-SIM microscope (GE Healthcare, Chicago, IL, USA) equipped with an Olympus 60× 1.42 NA Plan Apochromat objective and sCMOS cameras. Sunitinib and TPCS~2a~ were excited with a 488 nm and a 647 nm laser, respectively, and imaged sequentially. Z-stacks were recorded with a z-spacing of 125 nm. For each focal plane, 15 raw images (five phases for three different angular orientations of the illumination pattern) were captured. The fluorescence was detected through the band-pass filters 528/48 nm for sunitinib and 683/40 nm for TPCS~2a~. SI-images were reconstructed and aligned using Softworx software (GE Healthcare), and further processed using ImageJ (<https://www.nature.com/articles/nmeth.2089>). 4.8. Cellular Accumulation of Sunitinib {#sec4dot8-cancers-12-00417} --------------------------------------- For quantification of sunitinib accumulation in the HT-29 parental and sunitinib resistant cell line, 1.5 × 10^5^ cells per well were seeded in 6-well plates. The HT-29/SR cells were seeded in the presence of 2 µM sunitinib, whereas the parental cells where allowed to attach overnight in drug-free medium before adding sunitinib. Both HT-29 and HT-29/SR were then incubated with 2 µM sunitinib, and at the end of a 72 hour-incubation, the cells were detached with 0.25% (*w/v*) trypsin-0.53 mM EDTA (Sigma-Aldrich), washed with PBS and filtered through a 5 mL round-bottom tube with a cell strainer cap (Becton, Dickinson and Company, Franklin Lakes, NJ, USA). Sunitinib accumulation in cells was quantified using a BD LSRII flow cytometer (Becton, Dickinson and Company). Live and single cells were gated based on forward (FSC) and side scatter (SSC) parameters. Sunitinib was excited by a 100 mW 406 nm laser. The fluorescence was collected through a 585/42 nm band pass filter combined with a 545 nm long pass dichroic filter. Data were processed by the FlowJo version 10 software (Tree Star Inc., Ashland, OR, USA). 4.9. Absorption and Fluorescence Spectroscopy of TPCS~2a~ and Sunitinib {#sec4dot9-cancers-12-00417} ----------------------------------------------------------------------- Photochemical damage of sunitinib in solution pH 7 was evaluated by absorbance and emission measurements. Sunitinib (0.15 µM) and TPCS~2a~ (0.15 µg/mL) were prepared in PBS without Ca^2+^ and Mg^2+^ (Sigma-Aldrich) and subjected to blue light exposure (LumiSource™). The sunitinib and TPCS~2a~ concentration was selected based on optical density \<0.1 when both compounds are combined to avoid inner filter effect. The samples were illuminated in quartz cuvettes with lid, and sealed with Parafilm^®^ (Thermo Fisher Scientific) to avoid evaporation during illumination. The absorption and emission spectra were recorded immediately after light exposure at ambient temperature. Absorption spectra were recorded using a Shimadzu UV-2550 spectrophotometer connected to a computer with the software program UVProbe 2.62 (Shimadzu Corporation, Kyoto, Japan). All absorption spectra were recorded from 300 to 750 nm. Emission was recorded with a Cary Eclipse spectrofluorimeter (Agilent, Santa Clara, CA, USA) using the Scan Software V1.1. (Agilent) in the range 360--750 nm. The excitation and emission slit widths were 5 nm for TPCS~2a~ and 20 nm for sunitinib. TPCS~2a~ was excited at 420 nm and sunitinib at 432 nm, based on the recorded absorption spectra ([Figure S4](#app1-cancers-12-00417){ref-type="app"}). Relative decrease in peak intensity was calculated at 656 nm and 505 nm for TPCS~2a~ and sunitinib, respectively. Absorption spectra was also collected at pH 5 where sunitinib was prepared in citrate-phosphate buffer containing 1% FBS (Thermo Fischer Scientific). This citrate-phosphate buffer was prepared by dissolving sodium citrate tribasic dehydrate (Sigma-Aldrich) and disodium hydrogen phosphate dehydrate (Merck KGaA, Darmstadt, Germany) in distilled water \[[@B53-cancers-12-00417]\] and pH-adjusted to pH 5. 4.10. Animals {#sec4dot10-cancers-12-00417} ------------- All animal procedures were performed according to protocols approved by the Norwegian Food Safety Authority (FOTS ID 4593), which is the national animal research authority and were conducted according to the regulations of the Federation of European Laboratory Animal Science Association (FELASA). Handling of animals was therefore performed in compliance with EUs Directive 2010/63/EU on the protection of animals used for scientific purposes. Two different strains of female mice were used in this study. HSD athymic Nude-Foxn1^nu^ mice were bred at the Department of Comparative Medicine at the Norwegian Radium Hospital, Oslo University Hospital. BALB/c mice were obtained from Envigo (Horst, The Netherlands). The mice were maintained under specific pathogen-free conditions in a temperature-controlled room. Food and water were supplied ad libitum. The mice were on average 17--20 g (4--8 weeks old) when experiments were initiated. 4.11. Tumor Grafts {#sec4dot11-cancers-12-00417} ------------------ HT-29 cells (2.5 × 10^6^) in 30 µL PBS or CT26.WT cells (1 × 10^5^) in 15 µL PBS were injected subcutaneously on the left flank in athymic Nude-Foxn1^nu^ or thymic BALB/c mice respectively. The tumors and body weight were monitored 2--3 times per week. Tumor size was calculated using the following formula: V = (W^2^ × L)/2, where W is the width and L the length of the tumor measured by a digital caliper. The protocol was designed with two endpoints; tumor size 1000 mm^3^ and weight loss ≥20%. The animals were euthanized by cervical dislocation. 4.12. In Vivo Experimental Design and Methods {#sec4dot12-cancers-12-00417} --------------------------------------------- TPCS~2a~ was administered intravenously through the lateral tail vein at 5 mg/kg five days (CT26.WT) or seven days (HT-29) after tumor inoculation. Seventy-two hours post-TPCS~2a~ administration, the tumors were irradiated using a 652 nm red diode laser (CeramOptec GmbH, Bonn, Germany) at an irradiance of 90 mW/cm^2^ and a total dose of 15 J/cm^2^ for Nude-Foxn1^nu^ and 10 J/cm2 for BALB/c. The mice were anesthetized (sevofluran inhalation) and kept on a 37 °C heating pad during light exposure. The penetration depth of 652 nm red light into tissue is 4--5 mm \[[@B54-cancers-12-00417]\] which is sufficient for the s.c. tumors at \~100 mm^3^ here subjected to light exposure. The selected light doses are in addition previously reported as sufficient for PCI \[[@B35-cancers-12-00417]\]. Before light exposure, the animals were covered with aluminum foil with an opening diameter 2--3 mm larger than the tumor. Two different procedures, one with 2 doses of sunitinib (Sun1) and the other with 6 doses of sunitinib (Sun2) were performed for assessment of therapeutic effects of sunitinib-PCI in vivo. In both procedures sunitinib (40 mg/kg) was administered by oral gavage \[[@B30-cancers-12-00417],[@B45-cancers-12-00417]\] using a feeding needle (22 Gauge, 25mm, \#7901, Angthos AB, Lidingö, Sweden) 3 h prior to \[[@B22-cancers-12-00417]\] and 30 min after light exposure. In the Sun2 procedure administration of sunitinib at 40 mg/kg was continued for four more days, to a total of 6 administrations. Animals receiving TPCS~2a~ were kept in the dark for one week after administration of TPCS~2a~ or three days after sunitinib administration to avoid photo-toxicity. For evaluation of tumor growth delay, animals were randomized into six treatment groups including no treatment (NT), PS and light (PS + light), sunitinib only Sun1-procedure (Sun1), sunitinib only Sun2-procedure (Sun2), Sun1-PCI and Sun2-PCI. The experiment was set up with 3 treatment sessions for each tumor model and NT, PS + light and sunitinib only controls were included in each session. Intratumoral distribution of TPCS~2a~ and sunitinib in vivo was only evaluated in HT-29-bearing athymic Nude-Foxn1^nu^ mice with the Sun1 protocol. Animals were randomized into three groups; no treatment, TPCS~2a~ + Sun1 and Sun1-PCI with two mice in each group. 30 min post-light exposure the animals were sacrificed and the tumors were immediately placed on liquid nitrogen. 8 µm freeze sections were prepared and images acquired using Zeiss microscope as described above with 5×, 20× and 40× magnification immersion objectives (Carl Zeiss AG). TPCS~2a~ and sunitinib fluorescence was recorded using the filter combinations described in [Section 4.6](#sec4dot6-cancers-12-00417){ref-type="sec"}. The AxioVision software program (AxioVs40, version 4.8.0.0, Carl Zeiss AG) was used to process and analyze the images. For IHC, NT-, PS + light-, Sun2- and Sun2-PCI-treated CT26.WT tumors in BALB/c mice reaching 1000 mm^3^ were harvested, fixed in formalin and embedded in paraffin before they were prepared as previously described \[[@B37-cancers-12-00417]\] using αCD3 (A0452, Dako, Agilent Technologies), αKi-67 (ab15580, Abcam, Cambridge, UK) and αCD31(ab28364, Abcam). The tumor sections (2.5--3 µm) were also stained with H&E for evaluation of viable/necrotic tumor regions. Images were acquired using an AxioImager Z1 CellObserver microscope system (Carl Zeiss AG) with a 20x/NA0.8 lens and a 1ccc1 CCD camera (Carl Zeiss AG). 6 tiles were imaged with 10% overlap using an automatic stage. The images were stitched together using the Zen blue software (Carl Zeiss AG). The imaged area was selected to visualize the tumor from the distal and into the central part. Three ROI was defined at distant sites of the imaged area avoiding necrotic tissue. For CD3 stains, the number of positive cells in each ROI was counted and presented as an average. Two tumors were evaluated from each treatment group except from the PCI-Sun2 group, where three tumors were analyzed. 4.13. Evaluation of Combination Therapy and Statistical Analysis {#sec4dot13-cancers-12-00417} ---------------------------------------------------------------- Evaluation of synergy of sunitinib-PCI treatment was determined with a statistical model based on the assumption that PS + light and sunitinib have distinct and independent mechanisms of action \[[@B55-cancers-12-00417],[@B56-cancers-12-00417]\]. The theoretical additive effect in this model is a product of the survival fraction (SF) of each treatment separately calculated as follows: SF~add~ = SF~sunitinib~ × SF~PS+light~ (or log SF~add~ = log SF~sunitinib~ + log SF~PS+light~). The calculated SF~add~ was compared to the observed combined effect (SF~comb~). Synergy was further evaluated using the parameter DL (difference in logarithm) between observed SF~comb~ and the calculated SF~add~. DL = −(log SF~comb~ − log SF~add~) = log SF~add~/log SF~comb~ = log SF~sunitinib~ + log SF~PS+light~ − log SF~comb~. Synergistic effects resulted in positive DL values, antagonistic effects resulted in negative values and additive effects close to zero. Significant deviation from zero was established through one sample *t*-tests. Sigmaplot version 14.0 (Systat Sofware Inc., San Jose, CA, USA) was used for statistical analysis where *p* ≤ 0.05 was considered statistically significant. Two-sided student's *t*-test was performed for in vitro data, unless otherwise stated. For in vivo experiments, one-way ANOVA test and Holm-Sidak post-hoc test were performed to evaluate significant differences in tumor growth between the treatment groups. Statistical differences in survival were evaluated by pairwise log-rank analysis in IBM SPSS Statistics version 25.0 (IBM, Armonk, NY, USA). 5. Conclusions {#sec5-cancers-12-00417} ============== In conclusion, this is the first report demonstrating cytosolic delivery of a small-moleculeinhibitor by PCI. Sunitinib-PCI was found highly promising in vitro when utilizing the "light first" protocol and was also indicated to increase treatment-induced necrosis in vivo. The overall treatment response in our animal models was, however, less than expected which indicate mechanisms in the tumor stroma to attenuate an overall treatment response. Particularly in CT26.WT tumor-bearing thymic animals where an antagonistic effect on infiltrating T-cells was observed. Hence, our work indicates PCI to potentiate sunitinib cytotoxicity although adjuvant therapy aimed at the tumor stroma should be evaluated to improve the therapeutic efficacy. The following are available online at <https://www.mdpi.com/2072-6694/12/2/417/s1>, Figure S1: Chemical structure of disulfonated tetraphenyl chlorin (TPCS~2a~) and sunitinib, Figure S2. TPCS~2a~ localization after 18 h incubation without wash. Representative live cell fluorescence imaging of TPCS~2a~ in HT-29 cells after 18 h incubation with 0.4 µg/mL TPCS~2a~ without wash and chase. TPCS~2a~ (red), LysoTracker Green (green), Hoechst 33342 (blue). Co-localization indicated in yellow. Scale bar: 20 µm, Figure S3. Signals from fluorescence spectroscopy of sunitinib, TPCS~2a~ or the combination at pH\~7 (PBS containing 1% FBS). Fluorescence detected in sunitinib, TPCS~2a~ or the combination without light exposure. Data are mean of three experiments ± S.E, Figure S4. Absorbance spectra of sunitinib and TPCS~2a~. Representative absorbance spectra of sunitinib alone, TPCS~2a~ and the combination before and after blue light exposure at pH 7 and 5, Figure S5. Sunitinib accumulation in HT-29 and HT-29/SR after 72 h incubation. Median sunitinib fluorescence intensities in live and single cells. Cells were subjected to a 24 h wash before incubation with sunitinib. (Mean of three experiments ± S.E.), Figure S6. Tumor growth curves for HT-29 xenografts in athymic Nude-Foxn1^nu^ mice, Figure S7. Tumor growth curves for CT26.WT allografts in BALB/c mice. ###### Click here for additional data file. Conceptualization, A.W.; Methodology, J.J.W.W., M.B.B., S.P., V.S., and A.W.; investigation, J.J.W.W., M.B.B., A.S.V.F., V.S. and A.W.; resources, S.P., K.B., V.S., Q.P. and A.W.; data curation, J.J.W.W. and A.W.; writing-original draft preparation, J.J.W.W. and A.W.; writing-review and editing, J.J.W.W., M.B.B., A.S.V.F., K.B., S.P., V.S., Q.P., P.K.S. and A.W.; Visualization, J.J.W.W., S.P., V.S. and A.W.; supervision, A.W., P.K.S and K.B.; project administration, A.W.; funding acquisition, K.B., P.K.S. and A.W. All authors have read and agreed to the published version of the manuscript. This research was funded by the South-Eastern Norway Regional Health Authority (Helse Sør-Øst), grant number 2016023 (J.J.W.W.). The authors declare no conflict of interest. {#cancers-12-00417-f001} {#cancers-12-00417-f002} {#cancers-12-00417-f003} {#cancers-12-00417-f004} {#cancers-12-00417-f005} {#cancers-12-00417-f006}

**Core tip:** Lymphoma is known to be a cause of syndrome of inappropriate antidiuretic hormone secretion (SIADH). Moreover, Epstein-Barr virus (EBV) has a high infection rate, and in very rare cases, it can lead to extranodal natural killer (NK)/T-cell lymphoma. A very limited number of patients with lymphoma accompanied by SIADH have been reported, but NK/T-cell lymphoma with concomitant SIADH has not yet been reported in PubMed. Here, we present a case of NK/T-cell lymphoma in a 64-year-old woman with EBV infection accompanied by SIADH, suggesting the importance of monitoring serum ions, especially serum sodium, in patients with NK/T-cell lymphoma. INTRODUCTION ============ Epstein-Barr virus (EBV) is a well-recognized carcinogen that has been implicated in the etiology of several malignancies, including nasopharyngeal carcinoma\[[@B1]\]. The viral tropism of EBV is toward the B lymphocyte, but in very rare cases it can infect ectopic T and/or natural killer (NK) cells, leading to extranodal NK/T-cell lymphoma. Syndrome of inappropriate antidiuretic hormone secretion (SIADH) is a common cause of hyponatremia\[[@B2]\], but the symptoms of SIADH are nonspecific, which make it hard to detect and treat promptly. Untreated acute hyponatremia can cause substantial morbidity and mortality\[[@B3]\]. Many cancers can lead to SIADH\[[@B4]\], including lymphoma. According to the 2016 revision of the World Health Organization classification of lymphoid neoplasms\[[@B5]\], lymphomas are classified as either Hodgkin or non-Hodgkin lymphomas, and NK-cell lymphomas are in the latter class. Although non-Hodgkin lymphoma with concomitant SIADH has been reported internationally, NK/T-cell lymphoma with concomitant SIADH has not been reported in PubMed. Herein, we present a case of NK/T-cell lymphoma in a 64-year-old woman with EBV infection accompanied by SIADH, and suggest that monitoring serum ions, especially serum sodium, is vital in patients with NK/T-cell lymphoma. CASE REPORT =========== A 64-year-old woman was admitted with a chief complaint of intermittent fever for 2 mo. Two months previously, the patient had a sinus infection resulting from a cold, accompanied by intermittent fevers in the afternoon, with body temperature reaching as high as 38.9 °C. The patient had a history of nasopharyngeal carcinoma of over 30 years, which was controlled with chemotherapy; there was no recurrence to this point. One year prior, she had traveled to Europe, where she ate local food including sausage and fish. Physical examination after admission indicated a body temperature of 37.1 °C, heart rate of 80 bpm, respiratory rate of 18 breaths per minute, and blood pressure of 14.7/9.3 kPa. A 2.5 cm × 1 cm swollen, hard, unfixed, and painless lymph node was palpated below the left jaw. No blood congestion was found in the throat area. The left abdomen was soft with tender. No other obvious abnormalities were observed. Laboratory tests indicated the following biochemical results: serum sodium, 137.6 mmol/L; white blood cell (WBC) count, 2.4 × 10^9^/L; neutrophil count, 4.9 × 10^9^/L; C-reactive protein, 11.7 mg/L; EBV quantitative DNA test, 2.19E × 10^4^ copies/mL; EBV NA-IgG antibodies, positive (+) (\> 600); EBV VCA-IgG antibodies, positive (+) (\> 750); and EBV-IgM antibodies, negative. No abnormalities were found in the remaining tests. An imaging examination was performed on admission. Computed tomography of the lungs indicated that the bilateral axillary and mediastinal lymph nodes were slightly enlarged; enhanced computed tomography of the whole abdomen indicated swollen lymph nodes in the hepatic portal region and retroperitoneum, which did not rule out lymphoma; and bilateral ultrasound of the submandibular glands indicated several visible bilateral swollen lymph nodes. To determine the characteristics of the lymph nodes, whether lymphoma was present, and the margins of the lesion, we performed positron emission tomography-computed tomography, bone marrow biopsy, and left submandibular lymph node biopsy. Positron emission tomography-computed tomography indicated multiple swollen lymph nodes throughout the body accompanied by increased fluorodeoxyglucose metabolism, consistent with lymphoma (Figure [1](#F1){ref-type="fig"}). Bone marrow biopsy and immunotyping showed a myelogram with hyperplastic activity and a small number of abnormal lymphocytes, accounting for 4.80% of the visible cells other than NK lymphocytes. Biopsy of the left submandibular lymph nodes confirmed NK/T-cell lymphoma, and *in situ* hybridization showed multiple cells with strongly positive signal for EBV-encoded small RNA (Figure [2](#F2){ref-type="fig"}). {#F1} {#F2} At day 10 after admission, the patient developed lethargy, and had a serum Na^+^ level of 116.9 mmol/L. Once daily concentrated sodium chloride solution of 20 mL + 250 mL 0.9% normal saline was administered intravenously. One day later, the serum Na^+^ level decreased rapidly to 109.3 mmol/L. The serum K^+^ level was 3.36 mmol/L, urea level was 2.99 mmol/L, serum creatinine was 47.8 μmol/L, blood glucose level was 7.08 mmol/L, plasma osmolality was 235.39 mOsm/kg, and urine osmolality was 494 mOsm/kg. The patient's blood pressure was normal during that time. The fluid intake of the patient was immediately restricted, and three times daily, two salt capsules and 10 mL of concentrated sodium chloride were administered orally. After 2 d, the patient's blood Na^+^ level gradually increased to 126.5 mmol/L (8.6 mmol/L per day). Her mental state returned to normal, and the endocrinology department was consulted. Based on the patient's decreased plasma osmolality, urine osmolality greater than plasma osmolality, lack of skin swelling, normal blood pressure, normal renal function, no adrenal function detected on serology, and no abnormalities in imaging examination of the adrenal glands, as well as the effect of treatment, the likelihood of SIADH in the patient was high. The supplemental infusion of intravenous concentrated sodium chloride solution was discontinued, and renal sodium secretion was assayed. The hematology department was also consulted, and the patient was administered with epirubicin, vinorelbine sulfate, flumethasone, cyclophosphamide, and asparaginase chemotherapy, along with supportive treatment. We believed that the patient might be involved in stage 4B NK/T-cell lymphoma with concomitant SIADH, although with slightly lower renal sodium levels (20 mmol/L). On day 30 after admission, the patient's white blood cell (WBC) count gradually decreased to 0.1 × 10^9^/L, and the neutrophil count rapidly decreased to 0/L. The immune function of the patient declined, there were severe symptoms of infection, and respiratory function further deteriorated. On day 31 after admission, the WBC count gradually decreased to 0, various vital signs declined, and the patient died after failed resuscitation. DISCUSSION ========== SIADH has a hidden onset and is a syndrome caused by excessive antidiuretic hormone (ADH) secretion by the posterior pituitary\[[@B6]\]. It has a high mortality rate\[[@B3]\]. SIADH has many causes, and many cancers can lead to SIADH. It has been reported that lymphoma cells secrete ADH, and this prolonged ADH production results in SIADH\[[@B7]-[@B9]\]. SIADH secondary to lymphoma is relatively rare, and SIADH secondary to NK/T-cell lymphoma has not yet been reported. The symptoms of SIADH are nonspecific, and are primarily based on diagnostic standards published by Bartter et al\[[@B10]\] in 1967: (1) decreased plasma osmolality; (2) urine osmolality greater than plasma osmolality; (3) increased renal sodium secretion; (4) no skin swelling and normal blood pressure; and (5) normal renal and adrenal function. Recent studies showed that a fractional excretion of uric acid \> 12% has a high sensitivity and specificity for diagnosing SIADH\[[@B11]\] and can serve as a new basis for confirming an SIADH diagnosis. The first-line treatment for SIADH is restriction of fluid intake. If urine osmolality is higher than 500 mOsm/kg H~2~O and fluid restriction is ineffective, demethylchlortetracycline, urea, tolvaptan, and other drugs are instead used for treatment\[[@B12]-[@B14]\]. Intravenous infusion of 20-40 mg furosemide is used to manage volume overload, and 3% hypertonic saline solution by mouth or continuous intravenous infusion can be used as necessary for correcting hyponatremia\[[@B10],[@B15]\]. During sodium solution infusion, the focus should be on changes in blood sodium rather than the rate of sodium solution infusion. In the first 24-48 h, changes in blood sodium should be closely monitored, and treatment should be adjusted accordingly to rapidly correct blood sodium to within the safe range; otherwise, correction to normal ranges will not be achieved. Studies have shown that drugs and malignant tumors are the most common causes of SIADH\[[@B16]\], and the prognosis for SIADH secondary to malignant tumors is poorer than that caused by drugs\[[@B4]\]. To reduce the development of SIADH, regular examinations of blood sodium should be performed when malignant tumors are discovered as well as when beginning to administer drugs that are known to cause SIADH; this can help facilitate the immediate discovery of decreased blood sodium and thus immediately manage symptoms and improve prognosis. According to the 2016 revision of the World Health Organization classification of lymphoid neoplasms\[[@B5]\], lymphomas are classified as either Hodgkin or non-Hodgkin lymphomas. Non-Hodgkin lymphomas are classified as either mature B cell lymphomas or mature T-cell lymphomas, and NK-cell lymphomas are the latter. Currently, 34 cases of non-Hodgkin lymphoma with concomitant SIADH have been reported internationally\[[@B7],[@B8],[@B17]-[@B34]\] (Table [1](#T1){ref-type="table"}). ###### Published cases of non-Hodgkin lymphoma with concomitant syndrome of inappropriate antidiuretic hormone secretion **Classification** **Ref**. **Country** **Gender** **Age** **EBV status** **Lymphoma classification** **Symptoms** **Treatment** **Outcome** --------------------------------- --------------------------------- ------------- ---------------- ------------ ------------------------------------------------------------------------------------------------------------------- ----------------------------------------- --------------------------------------- ------------------------- ------------- T-cell lymphoma Chubachi et al\[[@B18]\] 1995 Japan M/ F 53/78 yr Infected Nasal T-cell lymphoma Fever following adjuvant chemotherapy Fluid restriction Died Demirkan et al\[[@B19]\] 2001 Turkey M 23 yr Unknown Anaplastic large cell lymphoma Weight loss Fluid restriction Died Night sweats Fever Hirata et al\[[@B8]\] 2012 Japan F 40 yr Unknown Primary cutaneous anaplastic large cell lymphoma Erythema Fluid restriction Died Right shoulder ulceration Isotonic saline Nishiwaki et al\[[@B30]\] 2014 Japan F 70 yr Unknown Acute adult T-cell leukemia/lymphoma Rash Fluid restriction Cured Isotonic saline Sun et al\[[@B34]\] 2018 China M 71 yr Uninfected Extranodal nasal type natural killer (NK) /T cell lymphoma Testicular enlargement Fluid restriction Died Multiple rashes with eschar Sodium chloride for oral Obstinate hyponatremia Hypertonic saline infusion Hydrocortisone infusion Phan et al\[[@B25]\] 1998 Indonesia M 53 yr Unknown Burkitt's lymphoma Left maxillary swelling Plasma exchanges Died Methotrexate Cytarabine Hydrocortisone B-cell lymphoma Sica et al\[[@B22]\] 1999 Italy F 41 yr Unknown Primary central nervous system lymphoma None Fluid restriction Cured Normal saline Diuretics Watabe et al\[[@B23]\] 2000 Japan M 70 yr Infected Angiotropic B-cell lymphoma Anemia Fluid restriction Died High LDH Ohara et al\[[@B27]\] 2007 Japan F 65 yr Unknown Diffuse large B-cell lymphoma Abdominal pain Fluid restriction Cured Acyclovir infusion Morimoto et al\[[@B20]\] 2007 Japan M 75 yr Unknown Intravascular large B-cell lymphoma Nothing Fluid restriction Died Hypertonic saline infusion Furosemide Fludrocortisone acetate Brodmann et al\[[@B26]\] 2007 Switzerland F 76 yr Unknown Mantle cell lymphoma Unknown Fluid restriction Cured Sodium chloride Potassium chloride Kobayashi et al\[[@B7]\] 2008 Japan F 84 yr Unknown Diffuse large B-cell lymphoma Left cervical tumor increased Fluid restriction Cured Isotonic saline Polprasert et al\[[@B21]\] 2011 Thailand F 64 yr Unknown Follicular lymphoma Watery diarrhea Ganciclovir Cured Rituximab Cyclophosphamide Vincristine Prednisolone Onishi et al\[[@B28]\] 2011 Japan F/ M/ M 73/80/83 yr Unknown Asian variant of intravascular large B cell lymphoma Unknown Unknown Unknown M 69/83 yr Died Onishi et al\[[@B28]\] 2011 Japan M/M/F 75/60/83 yr Unknown Diffuse large B-cell lymphoma Unknown Unknown Unknown F/M 84/64 yr Died Onishi et al\[[@B28]\] 2011 Japan M/M/F/M 74/69/75/75 yr Unknown Plasmablastic lymphoma/ Primary central nervous system lymphoma/ Mantle cell lymphoma/ Lymphoplasmacytic lymphoma Unknown Unknown Died/Died/Unknown/ Died Bockorny et al\[[@B29]\] 2012 America M 70 yr Unknown Marginal zone lymphoma Fatigue Fluid restriction Cured Confusion Demeclocycline Skin lesions Rituximab Zhu et al\[[@B24]\] 2013 China F 70 yr Unknown Diffuse large B-cell lymphoma Nausea Fluid restriction Cured Vomiting CHOP Left leg radiating pain Akhtar et al\[[@B17]\] 2013 United Kingdom M 75 yr Unknown Intravascular large B-cell lymphoma Weight loss Fluid restriction Died Clammy Demeclocycline Fludrocortisone Incontinent hypoxic Sumiyoshi et al\[[@B31]\] 2014 Japan F 61 yr Unknown Follicular lymphoma Epigastralgia Aciclovir Cured Peritoneal irritation Intestinal pseudo-obstruction Itaya et al\[[@B32]\] 2015 Japan M 81 yr Unknown Diffuse large B-cell lymphoma Fever Hypertonic saline Died Fatigue Hydrocortisone sodium succinate Anorexia Shimizu et al\[[@B33]\] 2017 Japan F 73 yr Unknown Mucosa-associated lymphoid tissue lymphoma Unknown Fluid restriction Cured Hypertonic saline F: Female; M: Male; EBV: Epstein-Barr virus; LDH: Lactate dehydrogenase; CHOP: Cyclophosphamide, doxorubicin, vincristine, prednisone. The male/female ratio of these 34 cases was 19:15, mean age was 69 years (range 23-84 years), and patients aged between 60 and 90 years accounted for 85% (29/34) of the patients. The clinical manifestations of SIADH in these cases were nonspecific, and large B-cell lymphoma made up 47% (16/34) of all the pathological types. Based on the known treatment data, the main treatment was fluid restriction, which was administered to 80% (16/20). The ratio of outcomes died/cured/unknown was 17:10:7, and the mortality rate of known outcomes was 63% (17/27). The clinical characteristics of known cases may be useful for clinicians to remain aware for these characteristics in patients who may be at risk for SIADH. The pathogenesis of SIADH secondary to lymphoma is complex. One potential cause is abnormal secretion of ADH by lymphocytes\[[@B7],[@B10],[@B30]\]. In our case, because of the death of the patient, whether hyponatremia was recurrent could not be determined, and abnormal secretion of ADH by lymphocytes could not be ruled out. Central nervous system injury secondary to non-Hodgkin lymphomas may also play an important role in the pathogenesis of SIADH\[[@B10]\]. Hyponatremia can also lead to nervous system symptoms\[[@B10]\], but the likelihood of central nervous system injury in our patient was low. The use of chemotherapy drugs such as vinca alkaloids and cyclophosphamide can also induce SIADH\[[@B35]\]. However, before SIADH was discovered in our patient, no chemotherapy drugs had been administered, so this did not apply in our case. Hypercytokinemia is another important factor in the development of SIADH secondary to lymphoma. Studies have shown that the levels of epidermal growth factor (EGF), granulocyte colony-stimulating factor (G-CSF), interleukin (IL)-5, IL-6, IL-12, IP-10, sIL-2Rα, membrane immunoglobulin (MIG), IL-1RA, and other cytokines are significantly elevated in T-cell lymphomas\[[@B36],[@B37]\]. In extranodal nasal type NK/T-cell lymphomas, sIL-2Rα, IL-6, and IL-10 are significantly elevated\[[@B38]\]. Increased IL-2, sIL-2R, IL-6, IL-1β, and tumor necrosis factor (TNF)-α can lead to abnormal secretion of ADH\[[@B39]\]. Watabe et al\[[@B23]\] found that patients with high levels of the cytokine IL-6 are more likely to develop SIADH. We did not examine the levels of cytokines in our patient; however, it cannot be ruled out that SIADH in our patient was associated with abnormal secretion of ADH by lymphocytes or that the lymphoma regulated ADH secretion through cytokines. EBV is a member of the human herpesvirus family and its genome consists of a single double-stranded DNA molecule. It has a high infection rate, with potentially 90% or more of all individuals infected worldwide\[[@B40],[@B41]\]. Humans are the only host for EBV and they become lifelong carriers of the virus after infection. Most hosts can live with EBV for a long period without any serious effects, but in some individuals, EBV is closely associated with the development of malignant tumors\[[@B42]\], including nasopharyngeal carcinoma and lymphomas. The viral tropism of EBV is toward B lymphocytes, but in very rare cases, it can infect ectopic T and/or NK cells, leading to chronic active EBV infection, extranodal NK/T-cell lymphoma, or invasive NK cell leukemia\[[@B43]\]. EBV can be divided into three latency programs. Latency II shows expression of EBV nuclear antigen 1 (EBNA1) and latent membrane proteins (LMPs), and is closely associated with the development of peripheral NK/T-cell lymphoma and nasopharyngeal carcinoma in elderly patients\[[@B42],[@B44]-[@B46]\]. Our patient had been diagnosed with both nasopharyngeal carcinoma and NK/T-cell lymphoma, and the possibility of long-term latent EBV infection cannot be ruled out. According to the 2016 revision of the World Health Organization classification of lymphoid neoplasms, NK/T-cell lymphoma is categorized into mature T-cell lymphoma and extranodal nasal type NK/T-cell lymphoma. The latter type is a relatively rare lymphoma characterized by high invasiveness, poor prognosis, and high likelihood of recurrence. According to a 2010 report, extranodal nasal type NK/T-cell lymphoma accounts for 6.9% of non-Hodgkin lymphomas and 28.2% of T-cell and NK cell lymphomas in China\[[@B47],[@B48]\]. Eighty to ninety percent of patients with nasal type NK/T-cell lymphoma report symptoms of nasal congestion, sinus infection, ulceration, and epistaxis\[[@B48]\]. Traditional treatments include chemotherapy, radiotherapy, and multimodal therapy, but even in patients with stage I or II disease, the 3-year survival rate is only 40%-50%\[[@B49]-[@B51]\]. Most published studies related to the efficacy of chemotherapy in extranodal NK/T-cell lymphoma have shown that the recurrence rate is as high as 50%\[[@B52]\]. Our patient was diagnosed with nasopharyngeal carcinoma 30 years earlier without recurrence following control with chemotherapy, but specific pathological examination results were not obtained. In addition, paraffin sections from over 30 years ago cannot be re-stained. Given the limitations of previous pathological examinations, the possibility that the patient developed extranodal nasal type NK/T-cell lymphoma cannot be ruled out. However, based on its biological characteristics, namely, a high recurrence rate, high invasiveness, and poor prognosis, the likelihood that our patient developed extranodal nasal type NK/T-cell lymphoma 30 years ago is small. In summary, NK/T-cell lymphoma with concomitant SIADH is relatively rare. Blood sodium must be closely monitored in lymphoma patients as an indicator of the development of SIADH, and immediate treatment by restricting fluid intake and replenishing sodium should be administered based on blood sodium levels. At the same time, actively searching for the cause of the disease and planning adjunctive therapy with drugs as needed will prevent critical patients from developing hyponatremia and hypo-osmolality, thereby improving their prognosis. ARTICLE HIGHLIGHTS ================== Case characteristics -------------------- A 64-year-old woman was admitted with intermittent fever for 2 mo. Clinical diagnosis ------------------ Stage 4B natural killer (NK)/T-cell lymphoma with concomitant syndrome of inappropriate antidiuretic hormone secretion (SIADH). Differential diagnosis ---------------------- Infection, typhia, brucellosis, *etc*. Laboratory diagnosis -------------------- As determined by blood and urine sampling examination, serum Na^+^ was 109.3 mmol/L, urea was 2.99 mmol/L, serum creatinine was 47.8 μmol/L, plasma osmolality was 235.39 mOsm/kg, and urine osmolality was 494 mOsm/kg. Imaging diagnosis ----------------- Positron emission tomography-computed tomography indicated multiple swollen lymph nodes throughout the body accompanied by increased fluorodeoxyglucose metabolism, consistent with lymphoma. Pathological diagnosis ---------------------- Biopsy of the left submandibular lymph nodes confirmed NK/T-cell lymphoma. Treatment --------- Chemotherapy, fluid restriction, and administration of sodium chloride. Related reports --------------- This is the first known report of NK/T-cell lymphoma with concomitant SIADH in PubMed. Term explanation ---------------- Lymphoma is one of the causes of SIADH; however, NK/T-cell lymphoma with concomitant SIADH has not been reported. Experiences and lessons ----------------------- This case report emphasizes the importance of monitoring serum ions and etiological treatment in patients with NK/T-cell lymphoma. Informed consent statement: The patient was not required to give informed consent to the study because the analysis used anonymous data that were obtained after the patient gave written consent for treatment. Conflict-of-interest statement: The authors declare that they have no conflicts of interest. CARE Checklist (2013) statement: The authors have read the CARE Checklist (2013), and the manuscript was prepared and revised according to the CARE Checklist (2013). Manuscript source: Unsolicited manuscript Peer-review started: July 31, 2018 First decision: August 31, 2018 Article in press: October 11, 2018 Specialty type: Medicine, research and experimental Country of origin: China Peer-review report classification Grade A (Excellent): 0 Grade B (Very good): B Grade C (Good): C, C, C Grade D (Fair): D Grade E (Poor): E P- Reviewer: Akbulut S, Chowdhury FH, Dasgupta S, Ohashi N, Vaudo G, Vidal EIO S- Editor: Ma RY L- Editor: Wang TQ E- Editor: Song H [^1]: Author contributions: Zheng R and Liu QB designed the report; Liu QB collected the patient's clinical data and wrote the paper. Correspondence to: Rui Zheng, MD, PhD, Doctor, Professor, Teacher, Department of Respiratory Medicine, Shengjing Hospital of China Medical University, No. 36, Sanhao Street, Heping District, Shenyang 110004, Liaoning Province, China. <[email protected]> Telephone: +86-2496-61521211 Fax: +86-2496-61572116

Specification TableSubjectBiomedical signal processing, NeuroscienceSpecific subject areaCognitive neuroscience, Steady-state visual evoked potential, EEG acquisition.Type of dataRaw data of 22-channel EEG signals and a single-lead electrocardiogram (ECG) signal before and after the stimulus (consumption of caffeinated coffee).How data were acquiredData were acquired using a 32-channel EEG acquisition system (RMS Maximus 32 EEG machine)Data formatRaw data (.mat file)Parameters for data collection6 healthy participants (22--28 years age range) were considered for the analysis.Description of data collectionThe EEG signals were acquired prior-consumption as well as post-consumption of caffeinated coffee. The recordings were made by providing photic stimuli of frequencies that lie within the range of 3 Hz--30 Hz. The corresponding recorded signals were used to understand the influence of caffeinated coffee consumption on the SSVEP signal generation.Data source locationNational Institute of Technology, Rourkela, IndiaData accessibilityWith the article**Value of the Data**•The current dataset is recorded in the presence of seven photic stimuli frequencies during before and after-consumption of coffee. The pre-consumption data can be used to analyze the effect of SSVEP stimulus frequencies in the different parts of the brain.•The pre- and post-stimulus data combined can be used to assess the effect of caffeinated coffee on altering the SSVEP signals.•The 22-channel EEG data can be used for frequency-domain analysis and mapping of different brain regions using the Fast Fourier transform algorithm.•The single-lead ECG data can be used to study the impact of caffeinated coffee on heart physiology.•The dataset can also be explored with different feature extraction, classification, and SSVEP detection algorithms. 1. Data {#sec1} ======= The steady-state visual evoked potential (SSVEP) is generated in the parieto-occipital regions of the brain whenever a source of light of constant stimuli frequency, is focused on the retinal cells \[[@bib2],[@bib3]\]. Caffeine, an essential constituent of coffee, is believed to have a partial impact on different brain regions such as parietal and occipital cortex areas \[[@bib4],[@bib5]\]. The current data set can be used to evaluate the post-consumption effect of coffee on the SSVEP activity in the brain regions. The EEG signals were recorded using a clinical EEG acquisition machine (RMS Maximus 32 CH). The equipment has 22 EEG channel electrodes, 2 mastoid electrodes, 1 ground electrode and two limb electrodes for ECG recording. The EEG electrodes were placed on the brain following the 10--20 standard positioning ([Fig. 1](#fig1){ref-type="fig"}). The sampling frequency of the recording was maintained at 256 Hz.Fig. 1Placement of electrodes on the scalp using 10--20 international standard.Fig. 1 The recorded EEG data from a single volunteer contains the embedded signal of 7 different photic stimulus frequencies of 20s duration and the resting period of 10-sec in between two stimuli frequency. The useful 20s data for each photic stimulus frequency is then extracted \[[@bib1]\]. The 20s duration data can be divided into different trials of varying length to increase the sample size as per user requirement. The EEG data of each volunteer was stored in a.mat file, represented in the format of "Vn_B" and "Vn_A", where n represents the serial number of volunteers, and B and A represent before coffee and after coffee consumption condition, respectively. Information regarding the data collected is presented in [Table 1](#tbl1){ref-type="table"}. Each.mat file contains a 7X2 cell array. The first column of the cell array contains the EEG signal in a 5120X23 matrix format, and the 2nd column contains the corresponding photic stimuli frequency. The number of rows in each data matrix is equal to the number of data points (256X20 = 5120) and the first 22 columns contains the corresponding amplitude value of each EEG channel (FP1, FP2, PG1, PG2, F7, F3, FZ, F4, F8, C3, CZ, C4, P3, P4, PZ, T3, T4, T5, T6, O1, O2, and OZ) at that data point. The last column contains the amplitudes of a single lead ECG signal ([Fig. 2](#fig2){ref-type="fig"}) during the recording.Table 1The data of recorded EEG signals.Table 1Volunteer IDAge (in years)SexStimulus condition (Caffeinated Coffee).mat file00123MaleBeforeV1_BAfterV1_A00229MaleBeforeV2_BAfterV2_A00321MaleBeforeV3_BAfterV3_A00425MaleBeforeV4_BAfterV4_A00524MaleBeforeV5_BAfterV5_A00626MaleBeforeV6_BAfterV6_AFig. 2Description of the data (.mat file) of a single recording.Fig. 2 2. Experimental design, materials, and method {#sec2} ============================================= This experiment is based on a within-group analysis that consists of two stages. In stage 1, EEG signals were acquired in the presence of seven photic stimuli frequency (3 Hz, 5 Hz, 10 Hz, 15 Hz, 20 Hz, 25 Hz, and 30 Hz). In stage 2, EEG signals were recorded 5 min after consumption of caffeinated coffee following the same protocol as in stage 1. Two types of stimuli were used in this experiment. A primary, photic stimulus of a certain frequency (to generate the SSVEP response) and a secondary stimulus, caffeinated coffee (to find the effect of coffee consumption on SSVEP). The stimuli frequency values used in this experiment were chosen in such a way that it will cover all the EEG bands ([Table 2](#tbl2){ref-type="table"}) \[[@bib6]\].Table 2List of stimuli frequency considered in the experiment.Table 2Frequency BandRange of Frequency (in Hz)Frequency considered in this study (in Hz)Delta0--43Theta4--85Alpha8--1310Low-beta12.5--1615Beta16.5--2020High-beta20.5--2825Low-gamma30--10030 Six healthy male volunteers, who are living a sedentary life and frequent coffee drinkers, were included in this experiment. Prior to recording, permission from the Institute ethical committee (NIT Rourkela) was taken for the recording of the EEG signals vide office order \#NITRKL/IEC/FORM/2/35/4/11/001, Dated 13/12/2013. The participants were explained the detailed procedure and purpose of the experiment. Further, they were urged to sign a consent form as a record of agreement of their voluntary participation in the experiment. The consent form was prepared following the guidelines of WHO\'s (World Health Organisation) informed consent for clinical study as a reference \[[@bib7]\] and is divided into three parts. Part-I contains the information related to the research, part-II contains the Information related to the volunteers and part-III contain the declaration made by the participant. A sample consent form has also been attached as a supplementary file. [Fig. 3](#fig3){ref-type="fig"} shows the different sections of the consent form, used in this experiment. The volunteers were also assured that they can leave the experimental procedure at any time if they feel uncomfortable or change their minds.Fig. 3The different subsections of the consent form.Fig. 3 The EEG signals were recorded using a 32-channel EEG machine, capable of recording 22 EEG signals from different brain regions. The electrodes were placed on the scalp using an electrode cap. The electrodes on the cap were arranged following the 10--20 international standard configurations. After mounting the electrode cap, the contact impedances of the electrodes were adjusted to \<20 KΩ. This was achieved by applying a conducting gel on the electrode scalp interface. The sampling frequency of the recording was maintained at 256 Hz. The volunteers were instructed to abstain from food and beverages before 2 h of recording. During recording, they were asked to sit on a chair inside a dim-lit room in a relaxed position. A set of LED was placed in front of their eyes at a distance of 60 cm. The LEDs were made to flicker at seven different frequencies (3 Hz, 5 Hz, 10 Hz, 15 Hz, 20 Hz, 25 Hz, and 30 Hz). The volunteers were requested to stare at the flickering light source during the process of EEG recording. The LED flickered at each stimulus frequency for a period of 20s with a 10s gap between two consecutive frequency stimulation. In the first phase, the EEG signals were recorded without any stimulus (pre-consumption of coffee). In the next phase, the volunteers were served a hot cup of coffee (120 ml containing 1.5 mg of coffee powder). After 5 min of consumption, the EEG signals were re-recorded following the same protocol. Conflict of Interest {#sec3} ==================== The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Appendix A. Supplementary data {#appsec1} ============================== The following are the Supplementary data to this article:Multimedia component 1Multimedia component 1Multimedia component 2Multimedia component 2Multimedia component 3Multimedia component 3Multimedia component 4Multimedia component 4Multimedia component 5Multimedia component 5Multimedia component 6Multimedia component 6Multimedia component 7Multimedia component 7Multimedia component 8Multimedia component 8Multimedia component 9Multimedia component 9Multimedia component 10Multimedia component 10Multimedia component 11Multimedia component 11Multimedia component 12Multimedia component 12Multimedia component 13Multimedia component 13Multimedia component 14Multimedia component 14 Supplementary data to this article can be found online at <https://doi.org/10.1016/j.dib.2020.105174>.

Background ========== The concept of self-management is not new and can be traced back in the UK to the start of the 20^th^ century \[[@B1]\]. Nevertheless, despite its longevity, defining self-management as a concept and clarifying its meaning for, and use, in practice has been found to be particularly challenging in the field of palliative care. Major obstacles to seeking clarity on self-management as a concept are that in spite of a plethora of available literature on self-management, the dominant focus is on managing chronic disease. However managing chronic illness presents patients and professionals with very different challenges to those found in palliative care \[[@B2]\]. An example of this is a UK government document, which defines self-care as 'the actions people take for themselves, their children and their families to stay fit and maintain good physical and mental health; meet social and psychological needs; prevent illness or accidents; care for minor ailments and long term conditions; and maintain health and wellbeing after an acute illness or discharge from hospital' \[[@B3]\]. Issues which are not the main focus in advanced disease when people are seriously ill, may lack functional capacity, and may be dependent on others for help. These issues and challenges are compounded by the ambiguity found in the terminology surrounding self-management, including the interchangeable use of similar terms such as self-care, self-efficacy, self-help \[[@B2],[@B4]\]. In a thematic analysis of the conceptualisation of self-care, self-management and self-management support in the long term conditions management literature, Jones et al. \[[@B5]\] found that the terms differed regarding the nature of the *imperative for action*. For instance, self-care is inevitable but includes choice; self-management is an inevitable activity whereas self-management support is an essential activity. Self-management support lies within context of coordinated networks derived from the health and social care system. This process is person-centred and any imperative for action is derived from the collaboration between people with long term conditions and thus improving support services. This concept analysis, therefore, is particularly focussed on self-management support. Political interest in self-management has also increased over the last decade, internationally, as governments attempt to contain the costs of health care, alongside addressing goals to improve health outcomes, patient satisfaction ratings and service delivery \[[@B3],[@B6]\]. This shift in the balance of care addresses the challenges associated with the increasing number of people suffering long-term conditions. Professionals, patients and politicians alike are, therefore, part of a trend in seeking health care and health service delivery solutions that are more patient focused and recognise the central role of patients in the care process. A factor that can affect self-care behaviors is health literacy. Health literacy is "cognitive and social skills which determine the motivation and ability of individuals to gain access to, understand and use information in ways which promote and maintain good health" \[[@B7]\]. Health Literacy includes skills and behaviours, that all people need, to for instance, find their way to the right place in a hospital, to fill out medical forms, and to communicate with healthcare providers. Poor literacy and numeracy skills can, therefore, result in difficulties in interpreting and performing self-care activities. This concept analysis was conducted as part of a larger study: Integrating Self-Management and Palliation Concepts (IMPACT), which was commissioned and funded by the ATLANTIS programme (Actions for Transatlantic Links and Academic Networks in Training and Integrated Studies) aimed at facilitating cooperation in higher education and vocational training in the European Union and the United States. The original study involved a partnership of four universities in Scotland, Lithuania, New York and Ohio. Two aims of IMPACT were the development of a comparative framework for policy analysis pertaining to palliative care and self-management and the creation of policy benchmarks for the delivery of palliative care to guide best practice in the management of self-care practices, in palliative care nursing in Europe and the Unites States of America (USA). Why exploring self-management support and palliative nursing is important ------------------------------------------------------------------------- Living with a life limiting condition, be it advanced cancer, or any other non-malignant disease, for which there is no cure, can have a devastating effect on a person. The impact can extend to psychological, social, physical, economic and cultural aspects of people's lives \[[@B8]-[@B10]\]. Individuals tend to cope as well as they can, with the support they have, whether that be from family or other sources, but often they do not have the information, skills or knowledge to make well informed decisions or the appropriate response \[[@B11]\]. Many patients do not understand what health professionals have said to them and do not, therefore, participate in decisions about their care, this leaves them ill-prepared to make daily decisions and take actions that lead to good care management. A collaborative relationship between nurses, health care teams, and patients and their families is, therefore important. Supported self-management in palliative care, by nurses, can, therefore, empower people to acknowledge the impact of their condition on their life, and enable them, where possible, to face the range of challenges they may have, and identify areas where they need further support, help or care \[[@B8]\]. In seeking to address the above issues this paper is concerned with exploring the issue of how self-management support in palliative nursing is conceptualised in the literature. Methods ======= Concept analysis ---------------- Concept analysis is considered a relatively new research approach, method or process \[[@B12]\] and is not an approach which is universally accepted \[[@B13]\]. The term concept analysis refers to the unfolding, exploring and understanding of concepts for the purposes of concept development, delineation, clarification, correction, identification, refinement and validation \[[@B14]-[@B17]\]. Walker and Avant \[[@B16]\] in the method chosen here, propose that; concepts are mental representations of a phenomenon or an idea, of an action or a thing that can accurately represents these occurrences within clinical practice. They advocate that the conceptualisation of concepts and their use in describing nursing practice is a stepping-stone towards the standardization of nursing language. In addition, concepts can be described as efforts to categorize information into meaningful constructs when applied to a phenomenon that occurs within the field of health care. A concept analysis is, therefore, a rigorous and precise process of operationalizing the defining characteristics and attributes of a phenomenon into a communicable understanding, and is undertaken using a structured framework \[[@B16]\]. Concept analysis is a method of conceptual knowledge representation and data analysis that can be used to clarify meanings and develop operational definitions, through considering evidence from multiple disciplines and sources. By applying a recognised methodological framework a more objective approach to concept clarification is accomplished. The systematic framework also means that the process is applicable within diverse scientific disciplines \[[@B16]\]. To guide the process of literature and analysis a modified version of the eight-step model presented by Walker and Avant was applied \[[@B16]\]. The eight steps of the model do not need to be, and were not, used in chronological order and can and were modified as the enquiry progressed. Table [1](#T1){ref-type="table"} outlines the 8 steps and whether and how applied for this concept analysis. ###### Walker and Avant concept analysis steps **Step** **Used in this concept analysis** -------------------------------------------------------------------------------------------------- --------------------------------------------- 1\. Select the concept Yes (dictionary definition-methods) 2\. Determining the aim or purpose of the analysis Yes (research question and aim-methods) 3\. Identifying all the known uses of the concept Yes (literature review-methods and results) 4\. Determining the defining attributes Yes (literature review- results) 5\. Identifying a model case ("real life" example, which contains all of the critical attributes No 6\. Identifying any of the borderline, related, contrary, invented and illegitimate cases No 7\. Identifying antecedents and consequences Yes (literature review- results) 8\. Identifying empirical referents   Selecting the concept --------------------- The concept self-management support and how it relates to palliative nursing will be analysed in this paper. In Walker and Avant's method the first stage is usually a literature review. ### Search strategy Primary data was identified by searching three online electronic databases via EBSCO Host: Medline, CinAHL and PsycINFO which cover literature from disciplines such as medicine, nursing, allied health, sociology and psychology. The key search terms are presented in Table [2](#T2){ref-type="table"}. ###### Search terms and key words **Search string/number** **Keywords** -------------------------- ----------------------------- 1 Self care 2 Self management 3 Self management support 4 1 or 2 or 3 5 Palliative care 6 Terminally ill 7 Terminal care 8 Hospice 9 Life limiting illness 10 End of life care 11 Nurs\$ 12 5 or 6 or 7 or 8 or 9 or 10 13 4 and 11 Search string in Medline, CinAHL and PsycINFO, keywords: 1. self care. The search led to 205 potential articles, with duplicates removed this left 165 to review. Of these 165 12 provided definitions of self-care and palliative care but no definitions of palliative nursing. A 'google' web search was also conducted for the terms palliative nursing, end of life care self-care, and self-management support. According to Web search workshop a UK consultancy service for optimising and marketing of websites users searching on the internet rarely go beyond the top 30 results \[[@B18]\]. Therefore, only the top 30 results were reviewed for keywords which always took longer than the 30 minutes to review the first 30 sites. This search string was then combined using Boolean operator 'OR' and 'AND'. Searches were limited only to the English language. Reference lists of all identified papers were scrutinised, hand searches of international journal of palliative nursing, grey literature and key websites was also conducted, using google and google scholar. The University of Dundee Library Catalogue and google scholar were also searched for key textbooks and book chapters. Additional websites were identified via links identified within the Google search and in collaboration with subject experts revealed additional websites, which yielded 11 potentially useful definitions or concepts. The top 30 results of a Google search were reviewed for the keywords of self-care and self-management support revealed 16 relevant articles. Key terms and words included self-care, self-management, self-efficacy and self-help. To place self-management in a professional context the palliative care and palliative nursing literature was examined to further elicit usage of the terms self-care and self-management (Google search Table [3](#T3){ref-type="table"}). ###### Google search and results **Google search term** **Number of useful results** ------------------------- ------------------------------ Self care 8 Self management support 8 Palliative care 8 Terminally ill 3 Terminal care 4 Hospice 11 Life limiting illness 7 Palliative nursing 5 End of life care 7 Inclusion and exclusion was applied to ensure that only relevant publications were included in the review (Table [4](#T4){ref-type="table"}). All titles and abstracts returned from the initial search were independently appraised by four authors. Full articles were obtained and appraised if they met the inclusion criteria. ###### Inclusion criteria for literature review **Inclusion** **Rationale** ---------------------------------------------------------- --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Published between 1990-2013 It was necessary to put time limits on the review. It was also determined that the majority of relevant literature was published during this period English text Due to budgetary constraints text other than English was excluded Described the results of empirical research publications Opinion or theoretical pieces were not included. It was determined that a more comprehensive review would be obtained if only empirical papers were used Adults only Palliative care and self care issues affecting children are different and palliative care services for adults and children are different, therefore only studies relating to adults were included Defining attributes palliative nurse Supportive Intensive caring, continuous knowing and continuous giving Fostering hope Providing comfort Proving an empathic relationship Being there Acting on the patients behalf Meeting the patients' needs Working together/teamwork Knows what they are doing Knowing the patient Dignity Providing information Results ======= Results reviewed by BJ and ER ----------------------------- ### Dictionary definitions The concept analysis framework identified by Walker and Avant indicates that definitions of terms are first sought as dictionary definitions *as part of identifying the concept.* Palliative ---------- Palliate has its origins in medieval Latin *palliativus*, from the verb *palliare* 'to cloak'. ### *Adjective* (of a medicine or medical care) relieving pain without dealing with the cause of the condition: *orthodox medicines tend to be palliative rather than curative.* (of an action) intended to alleviate a problem without addressing the underlying cause: *short-term palliative measures had been taken.* ### *Noun* A palliative medicine, measure, etc.: *antibiotics and other palliatives social projects presented as palliatives for the urban crisis oxford*\[[@B19]\]. Self-care --------- Self-care has been defined in the dictionary as: "The care of oneself without medical, professional, or other assistance or oversight" \[[@B20]\]. Self management support ----------------------- Self-management can be defined as the decisions and behaviours that patients with chronic illness engage in that affect their health. Self-management support is the care and encouragement provided to people with chronic conditions and their families to help them understand their central role in managing their illness, make informed decisions about care, and engage in healthy behaviours \[[@B21]\]. Self-management is 'the successful outcome of the person and all appropriate individuals and services working together to support him or her to deal with the very real implications of living the rest of their life with one or more long term condition' \[[@B22]\]. Support for self management is what services provide to encourage people to take decisions and make choices that improve their health, wellbeing and health-related behaviours \[[@B22]\]. Literature definitions of palliative care, palliative nursing and self managent support --------------------------------------------------------------------------------------- ### Palliative care Palliative care is among one of the fastest-growing specializations globally in the fields of nursing and medical education and referred to by current International government documents as not only focussing on death and dying but also on improving the quality of life for the patient and family \[[@B23],[@B24]\]. For almost a generation, to varying degrees, palliative care has been associated with total, active, holistic and therapeutic intervention/s, which focus on the quality of life for the patient and his/her family \[[@B24],[@B25]\]. The universal worldwide accepted definition is that proffered by the World Health Organisation (WHO) \[[@B26]\]. The WHO state that palliative care is an approach that improves the quality of life of patients and their families facing the problem associated with life threatening illness through the prevention and relief of suffering by means of early identification and impeccable assessment and treatment of pain and other problems, physical, psychological and spiritual. Palliative care: •provides relief from pain and other distressing symptoms; •affirms life and regards dying as a normal process; •intends neither to hasten or postpone death; •integrates the psychological and spiritual aspects of patient care; •offers a support system to help patients live as actively as possible until death; •offers a support system to help the family cope during the patients illness and in their own bereavement; •uses a team approach to address the needs of patients and their families, including bereavement counselling, if indicated; •will enhance quality of life, and may also positively influence the course of illness; •is applicable early in the course of illness, in conjunction with other therapies that are intended to prolong life, such as chemotherapy or radiation therapy, and includes those investigations needed to better understand and manage distressing clinical complications" \[[@B26]\]. Palliative nursing ------------------ The development of palliative care nursing has been part of a movement that has grown from roots in the nineteenth century, and particularly the second half of the twentieth century through the UK hospice movement and principally Cicely Saunders, who was originally a nurse \[[@B27]\] Seymour \[[@B28]\] argues that one of the clearest definitions of palliative nursing is that of Johnston \[[@B27]\] p. 2): "All life-threatening illnesses -- be they cancer, neurological, cardiac or respiratory disease -- have implications for physical, social, psychological and spiritual health, for both the individual and their family. The role of palliative nursing is therefore to assess needs in each of these areas and to plan, implement and evaluate appropriate interventions. It aims to improve the quality of life and to enable a dignified death" \[[@B29]\]. The palliative nurse, therefore, enters into a unique therapeutic relationship with the patient, which requires excellent communication skills and emphasises role aspects such as educator and information giver \[[@B27],[@B29],[@B30]\] and highlights their key involvement in the delivery of individualised, holistic care \[[@B27],[@B30]\]. The expert palliative nurse is someone who is interpersonally skilled, particularly in terms of the ability to be willing to listen, has personal humane characteristics such as warmth, kindness and compassion, and who helps the patient by meeting their needs, is there for them and provides them with emotional support, knows the patient as person and is knowledgeable, in particular, about pain and symptom control \[[@B27]\]. Self-management support ----------------------- As discussed previously, the political and professional agenda over the last decade has changed favourably in terms in of integrating self-management into the health care agenda. This interest has the potential to benefit greatly palliative care delivery and the development of palliative care services. The challenge remains, however, as to how this is accomplished and, in particular, how best self-management can be implemented in practice. It is recognised that incorporating the concept of self-management into palliative nursing practice brings additional challenges when managing symptoms at the end of life and when there is no known cure \[[@B7]\]. However, it is understood that if self-management can be utilised to a greater degree it will result in a better quality of life for patients and their families and may reduce the financial costs \[[@B7],[@B28]\]. Self-management has for a long time been associated with a process whereby patients deliberately act on their own behalf in health promotion and prevention of illness and the detection and treatment of health deviations \[[@B28]\]. It has, however, historically taken second place to the medicalisation of disease and the patient's passive acceptance of the care given by the medical and nursing professions. In examining the future potential of the concept for palliative care practice two observations are relevant. Firstly, up to the first decade of the twenty-first century most self-management strategies reviewed in the literature were professionally initiated and led \[[@B2]\]. Secondly, from a research perspective, few studies up to 2005 appeared to incorporate the patient's views on palliative nursing care, particularly the concept of the expert palliative nurse \[[@B27]\]. Nevertheless, the concept of self-management is not unique to nursing practice as the concept was identified in the latter half of the twentieth century as part of the development of nursing theory and models to define and support the principles of nursing practice linked to other concepts such as coping \[[@B31],[@B32]\]. Orem in particular championed the concept of self-management and defined it as the supported activities of individuals, in order to maintain health and wellbeing. Her research identified that deficits in self-management were often related to factors such as lack of knowledge, side effects of treatment, or physical, social and psychological aspects related specific to the individual \[[@B33]\]. These factors remain central to the consideration of self-management in the context of palliative nursing and Orem's model has relevance to palliative nursing as it links the concept to aspects of self-management in the contemporary literature, particularly in relation to wellbeing. Self-management in general has been shown to improve health outcomes, promote a feeling of well-being and improve the quality of life for those suffering incurable conditions \[[@B28]\]. A useful broad starting point to clarifying the meaning, relevance and use of the concept self-management in the context of palliative care and palliative nursing, is the definition put forward by Corner \[[@B34]\] (p.516). In palliative care the goal is to 'live with dying' with the focus on the self and not just the physical effects of illness. the following definition is, therefore, appropriate: maintaining ones usual practices of self-care, those things that are important and unique to oneself in maintaining ones sense of self; being given the means to master or deal with problems, rather than relinquish them to others". This definition emphasises the patient; 'being in control' and 'maintaining independence', which are important in end of life care \[[@B27]\]. This definition immediately places the patient at the centre of the care and caring processes. However, it also introduces the idea of self-management as part of the caring process. In countries where the focus of health care has moved from hospital to the community, many patients desire to be cared for at home, whenever possible, and the goal is often concerned with achieving patient and family choice \[[@B35]\]. This provides an ideal opportunity for the individual and his/her family to be involved fully in and have control of, their care. Health care policy has reinforced this objective incorporating self-management as an additional focus with emphasis on enabling patients to *manage* their own health and well-being \[[@B6],[@B35],[@B36]\]. While willingness and ability to self-management will change over time, it is also affected by the unpredictable nature and complexity of health related challenges with the person receiving palliative care \[[@B37]\]. Self-management challenges may be compounded by the plethora of information which is now available in the public domain. This is increasing more recently, with ease of access through information technology. These developments bring with them the potential of patients to access incorrect information \[[@B38]\]. Informed decision-making and knowing the patient's preferred choice \[[@B37]\] stress the importance of open and collaborative dialogue and knowledge of the patient's own story past and present. These recommendations imply that palliative care is a continuous process enabling the patient to cope with and respond appropriately to challenges as and when they arise and make choices and decisions about the future. These key aspects of self-management highlight the importance of 'knowing' the patient as a person \[[@B27]\]. They also highlight that in the field of palliative care, the facilitation of self-management brings additional challenges in managing symptoms and helping patients to live a life focused on quality as opposed to quantity (time). Moreover, an important issue for using self-management in practice is that not all individuals are either able to, or wish to, engage fully in self-management activities and that part of the professional's assessment is to identify the degree of self-management need and capability that is appropriate at any point in time \[[@B2]\]. Degrees of self-management engagement can be identified through robust physical and emotional management that enables the individual to adjust and match their self-management capability to their identified self-management needs, thus enabling them to stay in control of their unique and individual situation. Table [5](#T5){ref-type="table"} identifies the essential themes aimed at initiating and supporting self-management actions \[[@B8]\]. ###### **Supporting self-management: themes and sub themes**\[[@B8]\] **Theme** **Sub-theme** ----------------------------------- ------------------------------------------------------------------------------------------------------------------------- Maintaining normality Goal setting; How others treat you; Maintain normality-taking a break/holiday Preparing for death Euthanasia; Getting worse; Leaving family behind; Planning funeral; Process of dying Support from family/friends Carer support/information; Talking about difficult issues; Respite Self-cares strategies/physical Activities of daily living management; Aids to house; Complementary therapy; Financial help benefits; Managing symptoms Self-care strategies/emotional Accepting; Being positive; Choice; Control; Religion; Support from others with cancer Support from health professionals Clinical nurse specialist fixer/coordinator; Home help carer; Hospice day care; Out-of-hours care Various attempts have been made to clarify the terms self-management. For instance, it has been viewed as a transition and how people incorporate the consequences of illness into their lives \[[@B39]\]. As well as, associating the concept with the professional support and direction patients receive, including how to follow given instructions and manage other aspects of their condition \[[@B28],[@B31]\]. While these definitions allow to patients a semi-passive role in their care; they are also associated with active patient and professional collaboration in decision making, facilitating choice and decisions that support independent patient activities. While many of these factors are relevant to the palliative care context Johnston et al. \[[@B2]\] identify the description by Foster et al. \[[@B39]\] as most appropriate to palliative care as it highlights strategies used by individuals to enhance control and maximize wellbeing and the effects and approaches used by the individual to optimise living as closer than any other to being appropriate for people with advanced disease at the end of life. The capability to manage self-management appears to be associated strongly with the use of effective and robust assessment techniques, tools and processes, which are dependent on the patient with support from the nurse, assessing their own self-management needs and capability \[[@B2],[@B8],[@B28]\]. In addition, self-management in a palliative care context is linked to capability to control pain, manage other symptoms as well as evaluating the effectiveness of interventions. The concept of 'supported self-management' can, therefore, be said to embrace both self-care and self-management. In 'supported self-management' the concept of self-management can be linked closely to the patient's capability, while the professional is facilitating the patient to assess and identify their needs, moreover, self-management can be linked to outcomes of care and the patient's actual and potential capability to act in a way that meets their identified needs. Palliative care, therefore, needs to be underpinned by robust needs assessment, by the nurse, considering the patient's wishes, skills, behaviours and knowledge. The fundamental concepts underpinning palliative nursing assessment are that it is, *"... dynamic, Individualised, patient and family centred, sensitive and appropriate, holistic, therapeutic, contextual, comprehensive, based on reliable, current and valid information, evidence-based, driven by and focussed on process and outcomes"*\[[@B27]\]. This definition of assessment can be used to guide the professional to achieve the goal of supported self-management as a contemporary concept, with strong underpinnings in the effectiveness of the patient/professional relationship and the skills of the professional to support the patient in his/her self-management endeavours \[[@B5]\]. Assessment also brings into sharp focus the effectiveness of the Multi-Disciplinary Team (MDT) in supporting self-management. Johnston \[[@B8],[@B27]\] highlights the importance of effective collaboration and communication in supported self-management, which could be considered as the heart of the Multi-Disciplinary Team's (MDT). Key characteristics of self-management support in palliative nursing are, therefore, presented in Table [6](#T6){ref-type="table"} according to the Walker and Avant theoretical process. ###### Attributes of self-management and nursing role with author **Attribute** **Nursing role** **Author** ----------------------------------- ------------------------------------- ------------------------------------- Maintaining normality Knowing the patient Jarret et al., Skilbeck et al. Preparing for death Support Johnston et al., O'Berle and Davies Being there Zabalgui Comfort   Excellent communication skills Johnston et al. O'Berle and Davies Support from family/friends Emotional support Zabalegui Self-care strategies/physical Promoting independence Johnston et al., Rhodes et al. Good pain and symptom control Rhodes et al. Self-care strategies/emotional Promoting independence Johnston et al., Rhodes et al. support   Support from health professionals Teamwork Johnston et al. Referral role     Collaborating providing information Skilbeck et al. Antecedents and consequences ---------------------------- The identification of antecedents and consequences helps to refine the critical attributes and elucidate the contexts in which the concept is generally used \[[@B16]\]. Antecedents are factors that must be present before the occurrence of the concept whereas consequences are events that occur as a result of the concept. For supportive self-management in palliative nursing to occur we identified four antecedents; *presence of the nurse* and *spending time with the patient*; \[[@B27],[@B40]\]*development of a relationship with the patient,*\[[@B8],[@B27],[@B40]-[@B44]\]. *Skills, knowledge and expertise of the nurse*; \[[@B45]-[@B49]\], *team working and the ability of nurse to recognise when to refer on to other professionals or support services*. In analysing the concept of supportive self-management and palliative nursing this requires the demonstration of events that occur as a result of support being experienced/delivered. These consequences can be either a positive or negative experience for patients. Positive experiences for patients include feeling cared for and having their needs met; \[[@B27],[@B46]\] being informed \[[@B27],[@B47]-[@B50]\] and being supported \[[@B27],[@B48]\]. Negative experiences include pain and symptom control needs not met \[[@B51],[@B52]\] not having their needs met and not being supported, \[[@B27],[@B42],[@B43]\] as wellbeing unable to manage, particular in relation to activities of daily living \[[@B10]\]. Discussion ========== Strengths and limitations of the method --------------------------------------- The purpose of this analysis was to define the concept of self management support in relation to palliative nursing. Whist literature was retrieved from a variety of sources, certain limitations are worth mentioning. Literature in languages other than English were not accessed. A broader search could have produced a more comprehensive definition. Additionally, an alternative concept analysis method may have produced a different outcome. This concept analysis has identified that there is no universal definition of supportive self-management in palliative nursing. A clarified definition for nursing use based on this concept analysis is proposed below. Supportive self-management in palliative nursing is; *assessing, planning, and implementing appropriate care to enable the patient to live until they die and supporting the patient to be given the means to master or deal with their illness or their effects of their illness themselves.* The use of a model such as that proposed by Walker and Avant \[[@B16]\] was found to be beneficial in facilitating a systematic approach to literature retrieval, review and analysis. While stages 1 and 2 of the model were used to clarify the direction of travel and inform key words and terms for the literature review stages 3 -- 5 and 7 informed the focus of the analysis as well as the breadth and depth of literature retrieved. Stage 8 was important to supporting synthesis, which re-define self-management. Stages eight of the model helped to focus attention on the relevance and applicability of self-management to practice. Conclusions =========== For this concept analysis references to self-management were drawn mainly from chronic and disease management literature, as well as, the few articles on self-management available in the palliative care field. Self-management has been conceptualised in relation to how patients can gain and remain in control of their care process and how professionals can support this process. The concept analysis led to consideration of the extended concept of 'supported self-management'. The important role nurses have in supporting patients with palliative care needs to be able to maintain normality and independence and be in control for as long as possible is highlighted and to be encouraged. Figure [1](#F1){ref-type="fig"} is proposed as a guide to aid current discussion, for clinical practice and to aid further research and development of the concept 'supported self-management' in palliative care nursing and education. {#F1} Competing interests =================== The authors declare that they have no competing interests. Authors' contributions ====================== BJ devised the concept analysis and wrote the first draft ER contributed to the first draft and edited future drafts AB and JM and PC helped devise the concept analysis and edited the first draft. All authors read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1472-6955/13/21/prepub> Acknowledgments =============== This concept analysis was carried out as part of an EACEA/FIPSE funded policy orientated measure grant 2010--2013.

A 63-year-old male presented to the emergency department being shot with a hunting shotgun from a distance of more than 10 m. He had multiple entry wounds from pellets in the neck, thorax, abdomen, scrotum and upper extremities, with no exit wounds. He was haemodynamically stable and had a whole body CT scan with no intravenous contrast because of chronic renal insufficiency. A pellet adjacent to the left common carotid artery was demonstrated, accompanied by a limited hematoma. Additionally, a pellet was shown next to the trachea, and multiple pellets throughout his torso. Carotid Duplex demonstrated a pellet embedded in the posterior-medial wall of the left common carotid artery, producing a focal intraluminal protrusion of the intima with no tear ([Fig. 1](#F1){ref-type="fig"}). Bronchoscopy was normal. Repeat duplex and bronchoscopy after 48 hours did not show any change. Clopidogrel 75 mg once daily was prescribed and he was discharged after five days. At six months the patient was asymptomatic and repeat Duplex showed no interval changes. Retained intramural pellet has not been previously reported for the carotids. Additionally, no retained intramural missiles or foreign bodies elsewhere in the cardiovascular system have been reported except for the myocardium. There is lack of data regarding the natural history of such a carotid pellet, but the experience from the myocardium is that, in the absence of infection, completely embedded missiles are usually asymptomatic, tolerated well and may be left in place.[1)](#F1){ref-type="fig"} It is prudent for our patient to be on a life-long follow-up with duplex ultrasound in order to reveal possible complications like dissection, erosion and pseudoaneurysm or arteriovenous fistula formation. Another possible late complication could be intimal erosion and pellet migration to intracranial circulation, as has been previously described in acute cases.[@B2] Lead toxicity, another well-recognized late complication can occur months to years after the event.[@B3] {#F1}

1. Introduction {#sec1-ijerph-17-00691} =============== 1.1. The Links between Ageing, Potential Mobility, Choice and Well-Being {#sec1dot1-ijerph-17-00691} ------------------------------------------------------------------------ The importance of having the possibility to participate in society while ageing has been studied and documented in several different contexts \[[@B1-ijerph-17-00691],[@B2-ijerph-17-00691],[@B3-ijerph-17-00691],[@B4-ijerph-17-00691]\], as has the intrinsic value of mobility in itself \[[@B1-ijerph-17-00691]\]. Having the possibility to participate in various out-of-home activities has been found to be central for well-being outcomes during the later stages of life \[[@B5-ijerph-17-00691],[@B6-ijerph-17-00691]\]. This makes exploring the potential means of getting to these activities a key societal concern. Having a range of alternatives from which to choose is widely considered to be beneficial for a person's well-being (\[[@B7-ijerph-17-00691]\], pp. 33--39). In terms of daily mobility, this can relate to the range of everyday activities in which a person can participate \[[@B8-ijerph-17-00691],[@B9-ijerph-17-00691]\]; the time frames during which a person can participate in said activities \[[@B10-ijerph-17-00691],[@B11-ijerph-17-00691]\]; and the mode(s) of transport a person can use to reach said activities \[[@B12-ijerph-17-00691]\]. The importance of having a range of modal options from which to choose (referred to as 'robustness'), and indeed option value, have been emphasized as important elements of accessibility \[[@B13-ijerph-17-00691]\]. This robustness is considered key as it means that an individual is not reliant on a specific mode (with its inherent limitations) but instead has the option to be flexible and free to choose among several modes. This has also been discussed as a multimodal advantage, with those limited to just one mode said to be subject to a corresponding monomodal disadvantage \[[@B12-ijerph-17-00691]\]. The potential for mobility, most often in the form of accessibility (as opposed to mobility itself), is becoming a more prominent focus \[[@B14-ijerph-17-00691]\]. This potential has in turn been found to have strong associations with freedom and autonomy, particularly for older people during the ageing process \[[@B15-ijerph-17-00691]\]. This potential is also said to comprise important elements of well-being in itself \[[@B7-ijerph-17-00691]\]. Accessibility, motility and capability are key concepts which in recent years have been gaining ground as part of discussions surrounding potential mobility see e.g., \[[@B16-ijerph-17-00691],[@B17-ijerph-17-00691],[@B18-ijerph-17-00691]\] for elaborations of each concept. Instead of how, when, where or with whom a person travels, their possibilities or opportunities to do so are becoming the focus within research. 1.2. The Capability Approach as a Conceptual Framework {#sec1dot2-ijerph-17-00691} ------------------------------------------------------ The Capability Approach (CA) has been receiving increasing attention as a conceptual framework in transport research \[[@B19-ijerph-17-00691],[@B20-ijerph-17-00691]\] not least due to its focus on the size of the choice set and how it is tightly linked to equity concerns \[[@B21-ijerph-17-00691]\]. The CA has an emphasis on realizing potential and thus a focus on Eudemonic as opposed to Hedonic aspects of well-being, or becoming, being or living one's 'true self' \[[@B22-ijerph-17-00691]\]. The CA emphasizes the direct links between greater freedom---conceptualized as a larger set of 'capabilities'---and well-being \[[@B23-ijerph-17-00691]\]. Well-being, in turn, is influenced by the capabilities which are operationalized as functionings or the 'beings' and 'doings' of the individual \[[@B23-ijerph-17-00691]\].With respect to daily travel, this is related to the pre-conditions for a person's mobility; the options, capabilities or scope a person has with respect to mobility. The ties between this scope for action and realized mobility is framed as *evaluation* and *selection* of the best alternative, as deemed 'best' by the individual herself. Choice is a central concept in the field of transport and mobility, with choice---particularly modal choice---a key aspect of many studies e.g., \[[@B24-ijerph-17-00691],[@B25-ijerph-17-00691]\]. Revealed behavior ('functionings' within the CA) is the main consideration for many studies within transport. However, revealed behavior only tells us a very limited amount about a person's range of options \[[@B7-ijerph-17-00691]\] (pp. 36--37). If we are to focus on choice, the *range* of options---and the individual's selection from this range---is a more appropriate realm of analysis \[[@B7-ijerph-17-00691]\]. 1.3. Ageing and a Reduced Scope for Action {#sec1dot3-ijerph-17-00691} ------------------------------------------ A person's choice set with respect to mobility is largely considered to reduce in size in tandem with the ageing process \[[@B26-ijerph-17-00691],[@B27-ijerph-17-00691]\], often as a result of declining health, reduced cognitive or physical capacity \[[@B28-ijerph-17-00691],[@B29-ijerph-17-00691]\], changes in household composition (for instance, the death of a spouse) \[[@B30-ijerph-17-00691]\], or changes in personal financial circumstances \[[@B31-ijerph-17-00691]\]. For example, a person's scope for action may decrease in terms of the cognitive or physical capacity required to negotiate the environment(s) in which she moves, or a person's social capital might be diminished if her social network contracts \[[@B27-ijerph-17-00691],[@B32-ijerph-17-00691],[@B33-ijerph-17-00691]\]. However, the scope for action may also increase in some aspects of life as people age. For instance, as one retires, she may enjoy a greater level of flexibility in terms of time \[[@B34-ijerph-17-00691]\]. Some have noted that for older women in particular, reliance on a partner for transportation can ultimately result in a decreased scope for action \[[@B8-ijerph-17-00691],[@B10-ijerph-17-00691]\]. At the same time, caring for a person within one's social network, even just keeping them company, can also mean a reduced scope for action \[[@B30-ijerph-17-00691]\]. Not having access to technology, or not having the possibility to use it can, on the other hand, result in reduced scope for action, particularly as we witness the digitalization of many forms of transport. This can again affect women in particular \[[@B35-ijerph-17-00691],[@B36-ijerph-17-00691]\]. Maintaining the modal choice set in as far as possible could be highly beneficial for older people, in particular older people who have a more limited range of options from which to choose in comparison to what they may have had during earlier stages of life. It has been found that modal options are important in terms of facilitating the fulfilment of travel needs and activity wishes, while a lack thereof can have the opposite effect \[[@B37-ijerph-17-00691],[@B38-ijerph-17-00691]\]. A link has even been found between the use of a narrower range of modes and loneliness \[[@B39-ijerph-17-00691]\], with another study finding that some modes of transport are more closely associated with a greater scope for action than others \[[@B11-ijerph-17-00691]\]. It has even been emphasized that having options to choose from might not appear to be so important during the earlier stages of the ageing process but its importance may become more apparent during the later stages of the ageing process, if or when one has a reduced scope for action and more difficulty participating in everyday out-of-home activities \[[@B40-ijerph-17-00691]\]. While several studies have highlighted the importance of the car as a modal option during later life \[[@B41-ijerph-17-00691],[@B42-ijerph-17-00691]\], the straightforwardness of the relationship between the use of the car and the fulfilment of travel needs has also been questioned, particularly when one considers and accounts for different socio-demographic factors \[[@B43-ijerph-17-00691]\]. While potential mobility (encompassing modal options) and realized mobility (encompassing the mode(s) chosen) among older people have been explored in previous studies e.g., \[[@B37-ijerph-17-00691],[@B38-ijerph-17-00691],[@B39-ijerph-17-00691]\], investigating the process from potential to realization and the different kinds of motivation and reasoning behind the selection of some options over others is less explored. Furthermore, if we are to treat choice as an element of well-being, it is important to analyze the process of choice, and to distinguish between actually having chosen something and doing the same thing without having had a choice. Many empirical studies focusing on modal choice, particularly those based on standard travel survey data, do not make such a distinction. This study examines the process from potential to realized mobility among older people, with a specific focus on modal choice. The differences between the range of self-reported modal option(s), the modal option(s) selected from this range, and the reasons behind the selection of some modal options over others are explored. Links and exchanges between the experienced so-called 'real' options and realized behavior are then analyzed. In this way, we gain a greater insight into the different dynamic elements at play in the process from potential to realized mobility (evaluation as part of the CA), and the different forms of reasoning associated with the selection of different kinds of options. 2. Materials and Methods {#sec2-ijerph-17-00691} ======================== This study draws on a capability-based travel survey among older people (aged 65--79) and living in Sweden's large metropolitan regions; Stockholm, Gothenburg and Malmö. This meant that the survey was purposefully designed and shaped to capture elements of the CA. People within this age group (65--79) tend to have a greater scope for action than those aged 80 and above \[[@B44-ijerph-17-00691]\]. The questionnaire was developed with the intention of capturing three specific elements of the CA: resources (e.g., income, social capital, access to a bicycle, holding a driving license, etc.), capabilities (the range of different possibilities possessed by a person to achieve 'beings' and 'doings' of value) and functionings (the selected 'beings' and 'doings' of value). For a more detailed description of what the different elements of the CA entail, see \[[@B23-ijerph-17-00691]\]. The questionnaire and the Computer-Assisted Telephone Interviewing (CATI) script was formulated, tested and adapted a number of times before interviewing began. Ethical approval for this study was granted by the Regional Ethical Approval Board in Lund, Sweden in September 2015 (Dnr 2015/447). Sweden's three large metropolitan regions (LMRs), Stockholm, Gothenburg and Malmö, comprised the study area. These three LMRs are geographically delimited based on commuter patterns and movement between the central municipality and outer municipalities \[[@B45-ijerph-17-00691]\]. For this study, it was considered that LMRs would include and approximately reflect the relevant geographical areas of respondents' everyday activities. A stratified random sample of 3200 people aged 65--79 and living in Sweden's LMRs was drawn from a population register, with contact details containing a combination of landline and mobile phone numbers. Eleven people opted out before data collection began. A further two people opted out while data collection was underway. Interviewing was conducted by a small group of trained interviewers at a well-established and ISO-certified market research company. The interviews were carried out using CATI, allowing for more interviewer-respondent interaction so as to keep misunderstandings to a minimum and to facilitate a two-way communication process, when compared to other methods of data collection (e.g., compared to a postal survey or web-based survey). The interviews were conducted over the course of a three-week period during November-December 2015. The progression of the data collection was followed in order to gauge whether respondents understood and answered the questions in a consistent manner. From the remaining sample, attempts were made to contact 2119 people. These people were randomly selected but the probability of being selected changed successively with the completion of each additional interview. From those contacted, 1150 agreed to participate in the study, resulting in a response rate of 54% (56% for Stockholm; 52% for Gothenburg; and 55% for Malmö). In total, 1149 valid interviews were conducted (*n* = 383 in each LMR). The reasons given for not participating were (1) the person has a principle of not participating in surveys (29%); (2) the person did not answer (the phone) when they were contacted (18%); and (3) the person did not have time to participate when contacted (13%). All interviews were carried out in Swedish and over the phone. This meant that some people who were contacted could not participate. Among those who could not participate, 5.7% had difficulty speaking, spoke unclearly or did not speak Swedish. See \[[@B46-ijerph-17-00691]\] for a more detailed description of the sample, questionnaire and descriptive statistics. For this study, the focus lay on the process from potential to realized mobility among older people, with a specific focus on modal choice, and the reasoning behind the selection (and use) of specific modes over others. This meant that the elements of the material related to 'capabilities' and 'functionings' were the main focus. The analysis of this material was divided into the following parts:a)A descriptive analysis of the differences between the range of modal options available to respondents and the modes then selected from this range (modal choice).b)A descriptive and explorative analysis of the reasons behind the selection (and use) of said modes and the links between such reasoning and the different forms of selection processes. As part of the survey, respondents were asked to report which modal options they have the possibility to use for their everyday travel within the local region, even if they never use them. For this analysis, driving a car, using public transport (PT), walking and cycling were included as modal options. This material formed the basis for the 'scope for action', or the range of options from which individuals could choose. It is important to note that whether or not the individual considers it possible to use different modes of transport is highly subjective and is purely based on the perceptions of the individual. As such, two people who, for instance, appear to have all the same pre-conditions to cycle (e.g., access to a bicycle, physical capacity, short distances to activities of interest) may report different possibilities to cycle. This could be considered problematic in the sense that comparisons between individuals are not straightforward. However, the individual's perception is the key aspect of interest here, as it reflects the situation as perceived by the individual and thus, her perceived capability with respect to mobility. The respondents were then asked to state which options from this range they actually use, indicating 'realized mobility' and 'modal choice'. Here, although the respondent may have reported having a range of modal options from which to choose, she might perceive that she does not actually have a choice with respect to the mode she then uses. Similarly, the term choice may be perceived differently by respondents, with some perhaps reporting that they have the freedom to choose, and others reporting that they do not, despite being subject to the exact same circumstances. Respondents were then asked to give some reasons for why they opt to use these specific modal options and not the other modal options they could have selected. These reasons were processed in part as pre-codes. These were reasons, based on literature review, that were predicted to be given by respondents in advance of the survey being conducted. Respondents could also give open-ended reasons or accounts for their selection. It was possible to give several different reasons for selection, both in the form of pre-codes and open-ended answers. This meant that those who stated that in some way one mode was 'more suitable' than another often expanded upon their reasoning in the form of an open-ended answer. The actual formulation used in the open-ended answers was analyzed in a qualitative manner, while the information provided was subsequently coded into usually 2--3 different codes. Many respondents gave a combination of several pre-codes as well as an open-ended explanation. The prevalence of the different kinds of codes given by respondents were analyzed and subsequently combined into themes. The unique combinations of different codes provided were also analyzed, as was the actual wording given by respondents (in the open-ended element) for their modal choice. This material formed the basis for the analysis of the motivation and reasoning behind different forms of modal choice. The approach and analysis for this study can be regarded as a combination of quantitative and qualitative methods. The material obtained through the survey allowed for comprehensive, broader, descriptive analyses. However, such analyses are somewhat superficial in nature in terms of the actual meaning for respondents. The subsequent, more qualitative analyses allowed for a deeper and more detailed insight into the nuances behind the respondents' reasoning, allowing us to a gain a clearer understanding of the processes at play. 3. Results {#sec3-ijerph-17-00691} ========== Below is an account of the results for each part of the analysis. 3.1. The Differences between Scope for Action and Modal Choice {#sec3dot1-ijerph-17-00691} -------------------------------------------------------------- An outline of the various combinations of modal options, choice (use) and the frequency of the different choice processes among respondents is presented in [Table 1](#ijerph-17-00691-t001){ref-type="table"}, with the most common selection processes illustrated in [Figure 1](#ijerph-17-00691-f001){ref-type="fig"}. This gives us an insight into the scope for action and the selection of different modal options among respondents. More than one-quarter of respondents (28.8%) belong to the same category; having and using all four modal options. This category can be considered to comprise those at a relative multimodal advantage to the remainder of respondents \[[@B12-ijerph-17-00691]\]. By looking at the selection processes taking place among the second and third largest categories, it is clear to see that cycling is the most 'sensitive' with respect to selection; it being the first option to drop off among all the modal options. Those cycling, walking and using the car, yet choosing not to use public transport despite it being regarded as a modal option is relatively large (7.1% of respondents). Those who consider that they have walking, cycling, public transport and the car as modal options, yet selecting to use just the car is a larger group than expected, although still relatively small at 2.4%. Those with walking and cycling alongside the car as options yet choosing the car comprise a further 1.4%, with those with public transport as an option yet choosing the car making up a further 1%. Those who have the possibility to cycle yet choose the car comprises a further 0.5%, while those with both cycling and public transport as options yet choosing the car makes up another 0.3%. This means that the group actively selecting the car among a range of modal options amounts to 5.6% of respondents. The proportions of respondents actively selecting the other modal options among their respective ranges are much smaller, with the combinations of their respective categories largely only making up a fraction of the combined category actively selecting the car. Not all respondents have options, with 1.0% of respondents stating that they have no modal options whatsoever (or not answering the question). Several respondents state that they have just one or two options, and yet do not use those one or two options. Others state that they have just one option and use that one option. Some of the socio-demographic differences in modal choice processes are presented in [Table 2](#ijerph-17-00691-t002){ref-type="table"}. The variables of income ('lower income' encompasses a monthly household income of up to 16,666 Swedish crowns (SEK)), gender (male and female) and household status (cohabiting or not) were included in the analysis. [Table 2](#ijerph-17-00691-t002){ref-type="table"} indicates the key differences (and similarities) between socio-demographic groups, although no detailed analysis of statistically significant differences has been carried out. Walking and using public transport is spread evenly between income groups, while women are and those not cohabiting are somewhat overrepresented for this modal selection process. While having cycling as an option, yet choosing to walk and use public transport was linked to an overrepresentation of women, suggesting that the sensitivity of cycling for modal selection may be gendered. Actively selecting to use the car despite having other options is associated with an overrepresentation of men, suggesting that this too may be gendered. Those with lower incomes appear to be overrepresented in the use of cycling, walking and public transport as modes of transport (as a unique combination) while those cohabiting appear to be more inclined to actively select the car. However, this is just intended to give a brief overview of the sample. Further, more detailed statistical analyses are required in order to draw any robust conclusions regarding the differences outlined. 3.2. The Reasons Given for Modal Selection {#sec3dot2-ijerph-17-00691} ------------------------------------------ An outline of the different kinds of reasons given by respondents for their modal choice is presented in [Table 3](#ijerph-17-00691-t003){ref-type="table"}. The most common reason given by respondents for choosing the mode(s) they use over the mode(s) they could have used was that they are *more suitable*. Almost one-third (30.6%) of those who answered this question gave this as their one and only reason for choosing the mode(s) they did, while this reason accounted for 36.3% of mentions, either as a sole reason or combined with other reasons. When combined with other reasons or open-ended answers, 'more suitable' appears to be related to matters of convenience, and the mode fitting the purpose intended, be that using public transport in order to avoid difficulty getting parking, or opting to walk because activities lie within walking distance. The explanation attached to the pre-code *more suitable* meant that both positive and negative reasons could be coded in this way i.e., choosing a mode due to it being more efficient in terms of time or *not* choosing another mode because it does not suit the purpose of the trip. The second most common sole reason given was this/these mode(s) being *more comfortable* than the one(s) not chosen (at 15.2%). This reason accounted for 23.1% of the total mentions. Several respondents also mentioned a combination of just these two reasons (more suitable and more comfortable) (7.9%). A rather small proportion of mentions (1.5%) related to choosing mode(s) because they are *better for the environment/for environmental reasons*. However this was usually combined with at least one other reason, and rarely given as the only or main reason, often combined with *enjoyable* or *exercise/wants to move as much as possible*. At the same time, *enjoyable*, *less expensive* and *better for the environment/environmental reasons* also commonly featured as a combined reason for modal selection. These combinations of reasons were often linked to the selection of more active modes (walking and cycling) when the open-ended statements were analyzed. Furthermore, people mainly mentioned exercise as a reason when choosing to walk or cycle. However, this was also given as a reason for choosing to use public transport, with public transport trips sometimes described as comprising an element of exercise. Exercise when walking to/from public transport was even given as a reason for choosing one form of public transport over another e.g., opting for a longer walk to the train station instead of walking to a nearby bus stop. Fewer than expected reported *Health reasons/poor health/no energy* as their sole reason for modal selection (3.2% of respondents, 3.6% of mentions in combination with other options). Having poor health was often combined with another reason and was rather rarely given as a reason on its own. There was, in fact, another more positive relationship with health linked to modal selection, with 1.4% of mentions related to exercise, with several stating that they make this specific selection as exercising is healthy. A rather small proportion (2.2%) mentioned *faster/speed* (being faster than other modes) as their only reason for choosing the mode(s) they do over the modes they do not, with the proportion of mentions accounting for a more considerable 6.4% when combinations with other reasons were included. Being *less expensive/financial reasons* was also mentioned by surprisingly few (3.2% of reasons mentioned). Interestingly, these two reasons (*faster/speed* and *less expensive/financial reasons*) were often given in combination with one another. When focusing on the open-ended reasons, being 'inexpensive' often meant that a person had, for example, already paid for the car and did not want the extra expense of paying for public transport, or indeed, that a person had already bought a monthly or yearly public transport ticket and wanted to make the most of it. This was also related to choosing to walk and cycle because these modes do not cost anything, in the eyes of some respondents. There was only one mention of free public transport and this was by a person living or travelling in Gothenburg. *Proximity* as a sole reason for modal selection represented a larger proportion of respondents than expected (2.4%, 3.0% of total mentions), with 1.0% mentioning a combination of suitability and proximity for their modal selection. The reason *flexible, depending on the situation*, using all options or the strategic use of modal options accounted for 1.7% of sole mentions (1.3% of all mentions), while 2.3% of respondents mentioned having no choice/not much choice or having no alternative(s) as their sole reason for choosing the mode(s) they do. A small proportion (1.9%) explicitly stated that they have no need for alternatives, and gave this as their only reason for modal selection. While 2.4% stated that the sole reason behind their use of these modal options was simply that these were the options available to them. Some of the open-ended answers alluded to this being the life (style) that was chosen; their modal options were limited, but this was a conscious decision. All of the reasons encompassed in these codes allude to having a limited choice set. However, interestingly, the individuals' perceptions of their limited choice set differ considerably, with some expressing implicit dissatisfaction (*no choice/not much choice or having no alternative(s)*) and others expressing implicit satisfaction (*no need for alternatives*) despite perhaps having the same limited range of modal options. Again, the highly subjective nature of the survey means that individuals' perceptions of what is/is not an option could vary to a considerable extent. Some respondents may report not having a choice with respect to which modes they use despite considering that they do have other modal options. Here, there are two aspects to consider, the differences in the scope for action among respondents, as well as the differences in the perceptions of their scope for action. Respondents were inclined to give more positive reasons for selection when selecting more 'sustainable' modes. There was a sense of 'positive' or 'active' selection of the more active modes, with the environment, health, exercise and enjoyment linked to the selection of active or more sustainable modes. Several who chose walking and cycling explicitly mentioned proximity and accessibility to activities, mentioning that they live 'strategically' or 'centrally' and do not have any need to use a car when everything---or most activities---are within walking or cycling distance of their residential location. This is in contrast to the selection of the car, with this selection mechanism characterized as more passive, as reasons for not selecting other options, as opposed to direct reasons for selecting the car. Several talked about there being no alternatives; this was their option, and being locked into a lifestyle with this modal option. More than expected talked about complications related to the car i.e., having to find and pay for parking being the most prominent (0.3% of mentions). Fewer than expected mentioned freedom or wanting to have flexibility as a reason for choosing the car. Although a few open-ended responses did imply these aspects: > *"I come and go without a timetable."* Others took the car as a given or had normalized the use of the car as a sort of 'de facto' option, stating that they would only look into other options if it was not possible to use the car. Others questioned why one would even think about using other options when then the car is the only necessary option: > *"I use my own car as the first choice."* Some mentioned that their lifestyle dictates which mode(s) they use, meaning that having to carry heavy loads or living considerable distances from activities in remote locations result in the car being the only viable option. Fewer referred to not choosing public transport but when they did they talked about the non-existence of public transport, or mentioned that the destinations that could be reached using public transport were very limited, or that the public transport itself (i.e., the bus stop) was located too far away. Reliability was mentioned by fewer than expected and was usually given in combination with at least one or two other reasons. In the literature, reliability is often mentioned as an important aspect with respect to modal choice \[[@B37-ijerph-17-00691],[@B47-ijerph-17-00691]\] but as a theme, it was less apparent here. There were rather clear adaptive preference effects see \[[@B48-ijerph-17-00691]\] (pp. 83--84) at play with some people mentioning that they should not expect to be able to use other modes as they are sick or have poor or declining health, or mentioning that they have no choice, and that the only options that exist for them are the ones they use. Indications of adjustment related to the ageing process was also present. This materialized in the form of respondents discussing having stopped using certain modes, having had a fall, or having begun to experience declining health: > *"I am a pensioner. There is no need for me to travel every day."* The use of the tram became apparent as a very prominent mode of transport in Gothenburg, as several respondents living in this region went out of their way to mention their specific use of the tram, with habit an intertwined theme with the selection of this mode. Mentions of the tram in Gothenburg were much more evident than any other form of transport in any other area: > *" Because the tram goes exactly where I want it to go, and very often. It suits me fine. The tram is the best thing that exists."* The use of the tram seemed to be associated with a form of pride or a deeper association with the central parts of Gothenburg. In this sense, some of the reasons given suggested that the identity of the city seemed to be connected to the identity of the person, which in turn points to the symbolism of the tram as a mode of transport, and indeed its social meaning as a mode of transport \[[@B1-ijerph-17-00691],[@B49-ijerph-17-00691],[@B50-ijerph-17-00691]\]. Walking and the use of the tram were often discussed in tandem, alluding to the complementary nature of the two modes. Several mentioned the needs or wants of another person (most often a spouse) influencing their modal choice, e.g., having to give the spouse lifts, or not being able to use certain modes because they usually travel with the spouse. On the other hand, others talked about choosing the car because they had to give grandchildren lifts and had installed car seats for this purpose. Others talked about not being able to use the car and instead using public transport when travelling with grandchildren, or using public transport to reach their own children's homes to visit them and grandchildren. A number of respondents (1.3% of mentions) discussed tailoring their choice according to their needs. This meant having several options but, for instance, opting to use public transport in the city center or cycling or walking for shorter distances, but when it came to longer distances, using the car or public transport On the other hand, others (rather few) mentioned using the car for shorter distances, simply because they were used to using the car. 3.3. The Links Between Modal Selection Processes and the Reasons Given for Modal Selection {#sec3dot3-ijerph-17-00691} ------------------------------------------------------------------------------------------ The codes presented in [Table 3](#ijerph-17-00691-t003){ref-type="table"} were combined to form overarching themes, as presented in [Table 4](#ijerph-17-00691-t004){ref-type="table"}. Only the first recorded reason given by the respondent was used to form the codes, and overarching themes. This was because many respondents gave a range of different reasons (some gave up to eight reasons) that could not be placed into just one coding category or theme. This was in order to limit and simplify the categories produced by the cross-tabulation with the selection processes, presented in [Table 5](#ijerph-17-00691-t005){ref-type="table"}, and is, unfortunately, a limitation of this part of the analysis. The cross-tabulation of the different codes (reasons for modal selection) with the selection processes is presented in [Table 5](#ijerph-17-00691-t005){ref-type="table"}. Just some categories of modal selection processes were included, again in order to limit the number of categories produced. This also meant limiting the analysis to a number of more interesting groups, or rather, those who had actively selected a specific mode (or modes) when they had the possibility to select other modes. It is also important to note that many of the groups and sub-groups are very small so it is not possible to draw any robust conclusions from these small numbers. They are instead intended to give an illustration of how the respondents answered and reasoned. Those who actively chose the car were also more inclined to give extra benefits as a reason for choosing the car, although in most if not all of these specific cases, respondents were referring to enjoyment and freedom, and not exercise or the environment. As detailed in [Section 3.2](#sec3dot2-ijerph-17-00691){ref-type="sec"}, those giving these reasons generally gave them in combination with other reasons, and were most often not the first reason recorded, and therefore not presented in [Table 5](#ijerph-17-00691-t005){ref-type="table"}. 4. Discussion {#sec4-ijerph-17-00691} ============= The aim of this study was to examine and gain a greater insight into the process of selection from potential to realized mobility among older people aged 65--79, with a specific focus on modal choice. The differences between the range of modal option(s), the modal option(s) selected from this range, and the reasons behind the selection of some modal options over others were explored. One of the main findings was that a sizeable proportion (more than a quarter) of respondents have the option to use and actually use all modes for everyday travel. Although this can be considered a rather positive finding, this group can be considered to comprise those at a multimodal advantage relative to the remainder of respondents, a remainder which in turn comprises the majority. If we are to consider that a greater range of modal options can imply a greater freedom with its inherent links to well-being \[[@B7-ijerph-17-00691]\], then this group certainly fares best, especially when taking the findings from other studies into account \[[@B37-ijerph-17-00691],[@B38-ijerph-17-00691],[@B39-ijerph-17-00691]\]. However, others might have fewer modal options but still feel that their activity needs are met (e.g., those who expressed having no need for more alternatives). Here, it is difficult to decipher whether this is an indication of adaptive preference, or if this group in fact does have all of their mobility needs covered with just one option. Nonetheless, having just one or very few options might imply being at risk later on, if said modal options are no longer viable in later stages of life. Cycling as a modal option appeared to be the most sensitive with respect to selection; it being the first option to drop off among all the modal options. This is in line with previous findings where cycling has been found to be quite difficult for older people, with reports of issues with traffic safety, concerns with cycling environments and the behavior of other road users \[[@B49-ijerph-17-00691],[@B50-ijerph-17-00691]\]. This sensitivity appeared to be gendered. Although further analysis is required to confirm this. At the same time, other studies have pointed to the benefits of cycling for older people \[[@B51-ijerph-17-00691],[@B52-ijerph-17-00691]\]. Here, just like with driving cessation, cycling cessation (no longer having cycling as a modal option, and thus experiencing a reduced scope for action) requires adjustment and reconfiguration of mobility, and in some cases, participation in society \[[@B50-ijerph-17-00691]\]. However, for this study, we do not know whether the exclusion of cycling from the selected modal options is due to the individual having ceased cycling, or simply never having cycled. There were groups who perceived that they had no choice and just used the one or two options available to them. This could either indicate a difference in modal options, or just a difference in perceptions of modal options and choice among respondents, or indeed, both. This is an important aspect and limitation of subjective material and a focus on perceptions. Similarly, there were very different perspectives on having a limited choice set, with some respondents expressing implicit dissatisfaction (*no choice/not much choice or having no alternative(s))* and others expressing implicit satisfaction *(no need for alternatives)* despite perhaps having the same limited range of modal options. This finding calls into question the conceptualization of modal choice in empirical studies and underscores the importance of exploring the notions of choice and how choice is framed and deciphered as part of empirical studies. Moreover, there may be differences between an observer's account of the individual's choice and the individual's account of her choice, which can also be somewhat problematic, particularly when attempting to objectively assess differences among groups and between people \[[@B53-ijerph-17-00691]\]. Here it is rather difficult to decipher wherein the difference lies---a difficulty which has been highlighted by several studies (see \[[@B53-ijerph-17-00691]\] for an overview). In this sense, engaging different methods and different perspectives (both 'objective' and 'subjective') might assist in giving a more comprehensive and accurate picture of choice. Although we can question whether an option can really be considered an option if it is not the individual's perception that this is the case. Suitability and comfort were the two main reasons given for modal choice. There was a sense of 'positive' and 'active' selection of the more active modes, with more positive reasons such as the environment, health, exercise and enjoyment linked to the selection of active or more sustainable modes such as walking and cycling. This is in contrast to the selection of the car, with this selection mechanism characterized as more passive; as reasons for not selecting other options, as opposed to direct reasons for selecting the car were often given. Nonetheless, the car was more inclined to be selected among those who have a range of different modal options, with poor health one of the main reasons given for selecting the car over public transport. This finding is in line with those of previous studies, where it has been found that health issues related to ageing are associated with more negative effects for public transport use than for the use of the car \[[@B8-ijerph-17-00691]\]. The selection of the car in this manner might suggest that the car is a *de facto* option for some---as alluded to in some of the reasons given by respondents. It also links back to the monomodality question, and the normalization of the car as a mode of transport, and how this reliance on the car can be problematic if driving cessation becomes a reality. In such cases, well-being can be significantly compromised, particularly if the cessation is involuntary \[[@B41-ijerph-17-00691],[@B42-ijerph-17-00691],[@B43-ijerph-17-00691]\]. In this sense, linking back to previous research, we can question how the car is often framed as the transport mode of choice \[[@B54-ijerph-17-00691],[@B55-ijerph-17-00691]\], with this study revealing some of the more intricate elements behind this 'choice'. This implies that choice might not be as clear-cut as it seems, particularly when considering the links between choice, scope for action, freedom and well-being. The adjustment of expectations and preferences in accordance with disadvantage (often referred to as 'adaptive preference') \[[@B48-ijerph-17-00691]\] was evident in some parts of the empirical material. Such effects were apparent in the selection processes, with several respondents expressing that they know they have few or no options yet they do not feel that they should expect to have such a wide range of options. In some cases, this was in tandem with some clear adjustment effects, where respondents discussed having adjusted their expectations and actions as they aged. This also materialized in the form of respondents discussing having stopped using certain modes, having had a fall, or having begun to experience declining health \[[@B56-ijerph-17-00691]\]. This study facilitated a greater insight into the process behind modal choice; an insight which is often missing from discussions regarding modal choice. This study allows us to better understand the intricate processes at play and the underlying perceptions of individuals when considering their modal options and modal choice. Linking back to the CA, we can see that there are indeed differences in the ways in which people perceive their options (their capabilities), and in how they evaluate their options and select and operationalize their capabilities (functionings). This is an inherent difficulty with the empirical application of the CA as it means that there are discrepancies when attempting to carry out inter-personal analyses, especially with respect to equity concerns. Nonetheless, the empirical material engaged as part of this study simultaneously allowed for both a broader and more in-depth analysis of the perceptions of individuals with respect to modal options and modal choice. This study did, however, have some limitations. Its cross-sectional perspective did not allow for changes to be traced according to the ageing process. A longitudinal perspective would have aided us in understanding the selection process from an ageing standpoint. No 'objective' indicator was included, which would have allowed for a calibration of the self-reported responses, and would have given us a more nuanced perspective on some reports. This study did not explicitly include a geographical perspective. A lack of focus on geographical aspects may have resulted in some important considerations being masked as part of this study. Moreover, no comprehensive or clear-cut measure of subjective well-being was included in the study. This would have allowed for an empirical analysis of the links between modal options, modal choice and subjective well-being. This, in turn, would have facilitated a tighter connection to the CA and its conceptualization of the links between the size of the capability set and well-being. Nevertheless, this study did give a greater insight into the selection and reasoning processes behind modal choice, through the lens of those aged 65--79 and living in Sweden's LMRs. 5. Conclusions {#sec5-ijerph-17-00691} ============== Through engaging the careful treatment of the process of choice and drawing on information about individuals' reasoning with respect to the choice process, this study gave us a greater insight into the dynamic aspects at play with respect to modal choice for this group. Findings indicated that certain types of reasoning are related to the selection of specific modal options. Drawing on these types of reasoning and tailoring policy accordingly could prove fruitful in encouraging more sustainable travel behavior, and indeed promoting multimodality among this group. This would require a holistic approach by policymakers, encompassing not just transport policy but also community supports where the possibility to use other modes of transport (other than the car) are highlighted and promoted. For instance, training programs for using public transport could be employed. Solutions in the form of promoting multimodal trips (walking and public transport), highlighting the benefits in terms of exercise, experience and the environment could also be developed. In this way, individualized mobility suggestions could be developed so that people do not just use the car out of habit, and instead consider other modes. Not only 'carrot' but also 'stick' measures could prove effective. Several of the respondents remarked that they actively choose not to use the car because it is difficult to find parking. While others mentioned that they use public transport because they have a monthly ticket and want to make the most of it. Reducing parking opportunities alongside more attractive monthly pricing mechanisms could be another effective means of changing perspectives on modal options and intervening in the selection process. Those who had several options were more inclined to choose to use the car. This is despite car use often being framed as a lack of choice. Importantly, ensuring that the car does not reign as the 'automatic' option is key to achieving a modal shift. For instance, making public transport and other modes more usable for people with health problems---through infrastructure, but also increasing awareness of other passengers and drivers---could be a crucial way of establishing such a change. This study showed that the size and form of the choice set is important, and that it does matter whether someone chose from a range of options or instead just had to go with their one and only option. These findings have important implications for the consideration and treatment of modal options and modal choice within research going forward, presenting compelling reasons for including more detailed analyses of the processes behind modal choice. This approach could also be taken further by linking it to measures of subjective well-being, in order to draw clearer links between the scope for action (choice set) and well-being. Here, the focus was on those in the earlier stages of later life. These considerations could become even more tangible during the later stages of the ageing process. By studying and plotting out the process of modal choice, the reasons behind the selection of specific modal options and links between these and other aspects of life, we have a better basis for understanding and planning for older people's mobility. In this way, measures to promote and encourage sustainable mobility can be shaped and implemented, in such a way that the well-being of older people is also supported. The author would like to thank the respondents who generously dedicated their time to participating in this study. The author would like to acknowledge and thank Anders Wretstrand and Steven M. Schmidt for their supervision of the doctoral thesis from which this work stemmed. This research was funded by K2, The Swedish Knowledge Centre for Public Transport. This study was conducted within the context of the Centre for Ageing and Supportive Environments (CASE) at Lund University, funded by the Swedish Research Council for Health, Working Life and Welfare (FORTE). The author declares no conflict of interest. {#ijerph-17-00691-f001} ijerph-17-00691-t001_Table 1 ###### Frequency of the different modal choice processes. Description of Selection Process Frequency Percentage ------------------------------------------------------------- ----------- ------------ Cycling, walking, PT and car used 331 28.8% Cycling option, walking, PT and car used 104 9.1% Walking, PT and car used 104 9.1% Walking and PT used 82 7.1% PT option, cycling, walking and car used 82 7.1% Cycling, walking and PT used 62 5.4% Cycling and PT options, walking and car used 40 3.5% Cycling option, walking and PT used 33 2.9% PT option, walking and car used 32 2.8% Cycling, walking and PT options, car used 28 2.4% Car option, cycling and PT used 17 1.5% Cycling option, PT and car used 17 1.5% Cycling and walking options, car used 16 1.4% Car and PT used 15 1.3% PT used 13 1.1% No modal options 12 1.0% Cycling, PT and car used 11 1.0% PT option, car used 11 1.0% Cycling and walking options, PT and car used 11 1.0% PT option, cycling and walking used 10 0.9% Car option, walking and PT used 10 0.9% PT and car options, cycling and walking used 9 0.8% Car used 8 0.7% PT option, no modal options used 8 0.7% Cycling option, car used 6 0.5% Cycling option, PT used 6 0.5% PT option, cycling and car used 6 0.5% Cycling, walking and car used 6 0.5% Cycling option, walking and car used 5 0.4% Cycling and PT options, walking used 5 0.4% Walking used 5 0.4% Cycling and PT used 4 0.3% PT and car options, walking used 4 0.3% Car option, PT used 3 0.3% Walking and car used 3 0.3% PT option, walking used 3 0.3% Cycling and PT options, car used 3 0.3% Cycling, walking and car options, PT used 3 0.3% Cycling and walking options, PT used 3 0.3% Cycling, walking and PT options, no modal options used 2 0.2% Walking and cycling used 2 0.2% Cycling and car used 2 0.2% Cycling, PT and car options, walking used 2 0.2% Cycling option, no modal options used 1 0.1% Cycling and car options, no modal options used 1 0.1% Walking and cycling options, no modal options used 1 0.1% PT and car options, no modal options used 1 0.1% Car option, no modal options used 1 0.1% PT option, cycling used 1 0.1% PT and car options, cycling used 1 0.1% Cycling and PT options 1 0.1% Walking option, cycling used 1 0.1% Cycling, walking, PT and car options, no modal options used 1 0.1% Total 1149 100.0 Note: PT---indicates public transport. ijerph-17-00691-t002_Table 2 ###### Socio-demographic differences in modal choice processes. Selection Process Description Income % (*n*) Gender % (*n*) Household Status % (*n*) ---------------------------------------------- ---------------- ---------------- -------------------------- ----------- ---------- ----------- Walking and PT used 50% (41) 50% (41) 21% (17) 79% (65) 68% (56) 32% (26) PT option, walking and car used 28% (9) 72% (23) 56% (18) 44% (14) 19% (6) 81% (26) Walking, PT and car used 23% (24) 77% (80) 45% (47) 55% (57) 43% (45) 57% (59) Cycling option, walking and PT used 42% (14) 58% (19) 6% (2) 94% (31) 61% (20) 39% (13) Cycling, walking and PT options, car used 21% (6) 79% (22) 75% (21) 25% (7) 29% (8) 71% (20) Cycling and PT options, walking and car used 17% (7) 83% (33) 73% (29) 27% (11) 10% (4) 90% (36) Cycling option, walking, PT and car used 15% (16) 85% (88) 52% (54) 48% (50) 20% (21) 80% (83) Cycling, walking and PT used 43% (27) 57% (35) 21% (13) 79% (49) 65% (40) 35% (22) PT option, cycling, walking and car used 24% (20) 76% (62) 59% (48) 41% (34) 18% (15) 82% (67) Cycling, walking, PT and car used 19% (63) 81% (268) 53% (174) 47% (157) 24% (79) 76% (252) Total 28% 72% 46% 54% 33% 67% Note: \* A lower income refers to those with a monthly household income of up to 16,666 Swedish crowns (SEK). ijerph-17-00691-t003_Table 3 ###### Reasons for choosing to use one or several modes over another or other options ^1^. ------------------------------------------------------------------------------------------------------------------------------------------------------- Code Description Frequency of Mentions Percentage ------------------------------------------------------------------------------------------------------------------ ----------------------- ------------ More suitable (or less trouble than other modes) 487 36.3% More comfortable (e.g., having a place to sit) 310 23.1% Faster/speed 86 6.4% Health reasons/poor health/no energy 48 3.6% Habit/not a conscious choice 45 3.4% Less expensive/financial reasons 43 3.2% Proximity 40 3.0% Enjoyable (e.g., the experience. the views/less unenjoyable) 31 2.3% No choice/not much choice/no alternatives 28 2.1% Reliable (e.g., punctuality) 26 1.9% These are the options available 24 1.8% No need for other options/alternatives 21 1.6% Better for the environment/environmental reasons 20 1.5% Exercise/wants to move as much as possible 19 1.4% More secure (e.g., lower risk of being exposed to crime) 19 1.4% Safer (traffic/road safety) 19 1.4% Flexible depending on the situation/use all options/ strategic use 18 1.3% Distance/too far to walk 13 1.0% Due to someone else's needs or wants/\ 11 0.8% looking after someone else/someone else decides/\ shared household resources/\ minding dog/grandchildren Adjustment with age or health circumstances/no need to travel/too much difficulty travelling/'too old' to travel 7 0.5% Freedom (general) 6 0.5% The car is a given/normalised 6 0.5% Too much trouble with the car/parking 4 0.3% Satisfied with option(s)/with freedom of choice 3 0.2% Transporting heavy or large items/loads 3 0.2% Flexibility/no timetable/no planning 2 0.2% Public transport not possible/does not exist 2 0.2% **Total** **1341** **100%** ------------------------------------------------------------------------------------------------------------------------------------------------------- Note: ^1^ Frequency of mentions, several reasons given by some respondents (*n* = 961), with some respondents choosing not to give a reason for their modal selection. ijerph-17-00691-t004_Table 4 ###### Theme descriptions and the descriptions of codes included in each theme. ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Theme Description Descriptions of Codes Included ----------------------------------------------------------------------------------------------------- ------------------------------------------------------------------------------------------------------------------- More comfortable (or previous combination of more suitable and more comfortable) More suitable (or less trouble than other modes)\ More comfortable (e.g., having a place to sit) Flexibility and versatility/habit Habit/not a conscious choice\ Distance/too far to walk\ Transporting heavy or large items/loads Flexibility/no timetable/no planning\ The car is a given/normalized Less expensive/financial reasons Less expensive/financial reasons Security and safety More secure (e.g., lower risk of being exposed to crime)\ More secure and more comfortable\ Safer (traffic/road safety) More suitable More suitable Extra benefits (e.g., freedom, enjoyment, exercise, environment) Enjoyable (e.g., the experience. the views/less unenjoyable)\ Exercise/wants to move as much as possible\ Better for the environment/environmental reasons\ Enjoyable and better for the environment\ Freedom (general) Reliability, punctuality and speed Faster/speed\ Reliable (e.g., punctuality) Expressing a limitation with options (e.g., less choice, adjusted expectations or trouble with car) Public transport not possible/does not exist No choice/not much choice/no alternatives\ Adjustment with age or health circumstances/no need to travel/too much difficulty travelling/'too old' to travel\ Too much trouble with the car/parking\ These are the options available Health reasons/poor health/no energy Health reasons/poor health/no energy Expressing satisfaction with options (e.g., use several options, no need for other options) Flexible depending on the situation/use all options/ strategic use\ Satisfied with option(s)/freedom of choice\ No need for other options/alternatives Decision dependent on someone else or the needs of others Due to someone else's needs or wants/\ looking after someone else/someone else decides/\ shared household resources/\ minding dog/grandchildren ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- ijerph-17-00691-t005_Table 5 ###### Cross-tabulation of selection process groups of interest with themes of reasons for modal selection based on the first recorded reason (*n* = 395) \*. -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Selection Processes Examined Further More Suitable\ More Comfortable (or Previous Combination of More Suitable and More Comfortable)\ Reliability, Punctuality and Speed\ Flexibility and Versatility/Habit\ Less Expensive/Financial Reasons\ Expressing a Limitation with Options (e.g., Less Choice, Adjusted Expectations or Trouble with Car)\ Expressing Satisfaction with Options (e.g., Use Several Options, No Need for Other Options)\ Extra Benefits (e.g., Freedom, Enjoyment, Exercise, Environment)\ Security or Safety\ Health Reasons/Poor Health/No Energy\ Decision Dependent on Someone Else or the Needs of Others\ Total\ % (*n*) % (*n*) % (*n*) % (*n*) % (*n*) % (*n*) % (*n*) % (*n*) % (*n*) % (*n*) % (*n*) % (*n*) ---------------------------------------------- ---------------- ----------------------------------------------------------------------------------- ------------------------------------- ------------------------------------ ----------------------------------- ------------------------------------------------------------------------------------------------------ ---------------------------------------------------------------------------------------------- ------------------------------------------------------------------- --------------------- --------------------------------------- ------------------------------------------------------------ -------------- PT option, car used 30.0% (3) 10.0% (1) 10.0% (1) 0.0% (0) 0.0% (0) 10.0% (1) 0.0% (0) 0.0% (0) 0.0% (0) 40.0% (4) 0.0% (0) 100.0% (10) PT option, car and walking used 53.1% (17) 28.1% (9) 9.4% (3) 3.1% (1) 0.0% (0) 3.1% (1) 0.0% (0) 0.0% (0) 0.0% (0) 0.0% (0) 3.1% (1) 100.0% (32) Car option. walking and PT used 70.0% (7) 0.0% (0) 0.0% (0) 10.0% (10) 0.0% (0) 0.0% (0) 10.0% (1) 0.0% (0) 0.0% (0) 0.0% (0) 10.0% (1) 100.0% (10) Cycling option, PT and car used 64.7% (11) 0.0% (0) 5.9% (1) 0.0% (0) 5.9% (1) 0.0% (0) 0.0% (0) 0.0% (0) 0.0% (0) 23.5% (4) 0.0% (0) 100.0% (17) Car option, cycling and PT used 62.5% (10) 0.0% (0) 6.3% (1) 0.0% (0) 6.3% (1) 0.0% (0) 18.8% (3) 0.0% (0) 6.3% (1) 0.0% (0) 0.0% (0) 100.0% (16) Cycling and walking options, car used 43.8% (7) 18.8% (3) 6.3% (1) 12.5% (2) 6.3% (1) 12.5% (2) 0.0% (0) 0.0% (0) 0.0% (0) 0.0% (0) 0.0% (0) 100.0% (16) Cycling option, walking and PT used 45.5% (15) 6.1% (2) 6.1% (2) 12.1% (4) 3.0% (1) 18.2% (3) 0.0% (0) 0.0% (0) 6.1% (2) 3.0% (1) 0.0% (0) 100.0% (33) Cycling, walking and PT options, car used 42.9% (12) 25.0% (7) 7.1% (2) 10.7% (3) 0.0% (0) 3.6% (1) 0.0% (0) 7.1% (2) 0.0% (0) 3.6% (1) 0.0% (0) 100.0% (28) Cycling and PT options, walking and car used 45.0% (18) 35.0% (14) 2.5% (1) 0.0% (0) 0.0% (0) 0.0% (0) 7.5% (3) 7.5% (3) 0.0% (0) 2.5% (1) 0.0% (0) 100.0% (40) Cycling option, walking, PT and car used 55.3% (57) 21.4% (22) 2.9% (3) 2.9% (3) 2.9% (3) 6.8% (7) 3.9% (4) 1.0% (1) 2.9% (3) 0.0% (0) 0.0% (0) 100.0% (103) PT and car options, walking and cycling used 25.0% (2) 37.5% (3) 0.0% (0) 0.0% (0) 0.0% (0) 0.0% (0) 37.5% (3) 0.0% (0) 0.0% (0) 0.0% (0) 0.0% (0) 100.0% (8) PT option, cycling, walking and car used 47.6% (39) 30.5% (25) 6.1% (5) 3.7% (3) 1.2% (1) 1.2% (1) 2.4% (2) 4.9% (4) 0.0% (0) 1.2% (1) 1.2% (1) 100.0% (82) Total 50.1% (198) 21.8% (86) 5.1% (20) 4.3% (17) 2.0% (8) 4.8% (19) 4.1% (16) 2.5% (10) 1.5% (6) 3.0% (12) 0.8% (3) 100.0% (395) -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- \* Please note that this table is simply intended to give an indication of how respondents answered and some of the percentages represent very small numbers. PT---indicates public transport.

***Cite this article as:*** Hosseini SS, Taheri K, Hosseini SM, Gholami M, Kouhsari E, Edalati E, Madani R, Mohammadzadeh R, Zahedi Bialvaei A, Sholeh M, Golchin Far F. Extraction and purification of the H9N2 virus nucleoprotein: A simple and practical method. *Med J Islam Repub Iran*. 2018 (21 Dec);32:128. https://doi.org/10.14196/mjiri.32.128 ↑ What is "already known" in this topic: {#box1} ======================================== The nucleoprotein H9N2 influenza virus functions as a significant adaptor molecule between the virus and host cell processes. The purified nucleoprotein is used to detect new virus species caused by interbreeding and their resultant disease and to provide a vaccine for avian influenza for prevention and control programs of influenza in the industry. → What this article adds: {#box2} ========================= The aim of this study was to purify virus NP in order to further project with a simple, low-cost, fast and practical method. In the current study, firstly the H9N2 virus identified and the locations of the viral proteins were determined. Subsequently the nucleoprotein was purified with adequate and accurate information about the viral proteins. Introduction {#s1} ============ Avian Influenza is a highly contagious disease that causes symptoms in respiratory, digestive and nervous systems in a wide range of birds and can be transmitted to mammals, including humans ([@R1]). Dramatic changes reassortment between human and avian viruses are due to annual epidemics and pandemics ([@R2]). To date, the most appropriate method for fighting avian influenza is control and eradication. An effective human influenza vaccine development is a critical component to control and prevent an influenza pandemic ([@R3]). Influenza virus is an enveloped, single-stranded negative-sense RNA segmented virus ([@R3]). Based on the two proteins on the surface of the virus; hemagglutinin (H), and neuraminidase (N), influenza A viruses are divided into 15H and 9N subtypes ([@R4]). The proteins play key roles in infecting a host cell. In addition, inside the envelope are segmented single-stranded RNA and nucleoprotein (NP) ([@R5]). Avian influenza (AI) H9N2 subtype is generally reported as the primary isolate from turkey flocks in Wisconsin in America in 1966, then the virus has been reported and circulated among various countries ([@R6]). Although H9N2 is a low-pathogenic virus, the disease has been observed in all types of poultry in the field ([@R7]). Because of the wide range of the virus and the great losses it caused, initially a local vaccination program was implemented, but mass vaccination was quickly authorized ([@R8]). The virus has spread to other avian species such as chickens (breeders, broilers, and layers) and ostriches ([@R8]), it has been panzootic in Eurasia and is now the most dominant subtype of influenza virus in poultry industry ([@R9]). The viruses were classified as low pathogenic on the basis of experimental inoculation of chickens by different routes, but under field conditions, they have proved to cause a severe respiratory syndrome in birds, with heavy reductions in egg production and high mortality ([@R5], [@R7]). The H9N2 outbreak has been reported in the world from 1944 to 1996 in Europe, Africa, and Asia. Subsequently, in 1997, severe H9N2 cases were reported in many Asian countries ([@R10]). All viruses with negative-sense RNA genomes encode an ssRNA binding nucleoprotein (NP) ([@R11]). The NP is one of the key determinants of species specificity, RNA encapsulation, transcription, replication, and packaging ([@R11]). The NPs have interaction with different molecules such as; PB1, PB2, M1, Importin α, F-actin, CRM1/exportin-1 and BAT~1~/UAP~56~ ([@R11]). The NP Functions as a key adaptor molecule between the virus and host cell processes ([@R9], [@R12]). The purified nucleoprotein is used to detect new virus species caused by interbreeding and their resultant disease and to provide a vaccine for avian influenza for prevention and control programs of influenza in the industry ([@R13]). The aim of this study was to purify virus NP in order to further project with a simple, low-cost, fast and practical method. Methods {#s2} ======= Viral propagation {#s2-0-1} ----------------- In this experimental study, viral propagation was performed by the method described ([@R14]). Briefly, the isolates of avian influenza A/Chicken/Iran/259/2014/H9N2 were propagated in specific-pathogen-free chicken embryonated eggs by allantoically inoculating 10^3^ to 10^5^ egg-infective doses (EID~50~) for 9 to 11 days. Eggs were incubated at 37°C with relative humidity of 55% and were candled daily for 7 days to embryo viability. Allantoic fluid from embryos that died after 24 h was collected aseptically and tested for hemagglutinating virus by the hemagglutination (HA) test. If no deaths occurred in the eggs after 6 days, all of the eggs were opened aseptically. The allantoic fluids were pooled and inoculated into 5 more embryonated eggs. If no deaths occurred after 7 days in the eggs of the second passage, all the eggs were opened, and the allantoic fluid of each egg was tested for hemagglutinating activity. If deaths occurred in the eggs and the material had a positive HA test, samples were considered positive, and this result was confirmed with the hemagglutinating inhibition (HI) test. Allantoic fluids from embryos showing early mortality after inoculation were tested for bacterial contamination according to routine procedures, and subsequently the infected fluids which contained between 32×10^5^ EID 50 of virus per ml were harvested and pooled. Reference stocks were lyophilized and working stocks were stored in ampoules at -20°C ([@R15]). However, HA and HI tests were performed by standard methods ([@R15]). Extraction and Purification of nucleoprotein {#s2-0-2} -------------------------------------------- The mixture was incubated for 1 hour at 4°C and clarified by centrifugation at a speed of 3500 rpm and 4°C for 15 minutes. The supernatant was collected, and PEG 5% (w/v) was added to it. Then, clarified by centrifugation at a speed of 9000 rpm and 4°C for 60 minutes, precipitate in TNB buffer was created as a homogenous suspension. Finally, a homogenous suspension was prepared for dialysis to remove PEG from the suspension. The HA test was performed on the sample after dialysis. Then, the dialyzed fluid was ultra-centrifuged at a speed of 20000 rpm and 4°C for 2 h. The resulting precipitate was converted into homogeneous suspension in 100 ml of TNE buffer. The HA test was also carried out at this stage. The purification of virus was performed by sucrose density-gradient centrifugation method ([@R16]). Finally, the resulting virus samples were lyophilized. Protein Assay {#s2-0-3} ------------- For this purpose, Bradford method was performed to determine the protein concentration of the sample according to the previously described method with some modifications ([@R17]). In this method, duplicate volumes of bovine serum albumin (BSA) (10-100 mg/ml) was used to prepare the calibration graph. The basic Laemmli SDS-PAGE procedure was used to separate the protein of interest ([@R18]). Afterward, one of the gels was stained with Coomassie blue G-250 (Merck, Germany) and the other gel with silver nitrate. Also, native and functional biological molecules were purified by preparative electrophoresis (model 491 Prep Cell, Bio-Rad, CA), using reduced and non-reduced methods ([@R19]). Finally, to determine protein fractions the optical density was read at 280 nm using a spectrophotometer. The reduced 12% SDS-PAGE was used to analyze the integrity of the purified influenza virus proteins and the detection of fractions containing nucleoprotein. In brief, 200 μl of each protein fraction \[with a high absorption\] was mixed with 30 μl of 2-mercaptoethanol and incubated for 4 h at 37 °C. The protein fractions \[with high absorption\] were dissolved in non-reducing SDS-PAGE loading buffer (without 2-mercaptoethanol) and stained with silver nitrate ([@R20]). The protein concentration was assessed with the Bradford method. Finally, in order to confirm H9N2 presence, the purified sample was loaded and resolved by SDS-PAGE and proteins were transported to polyvinylidene difluoride (PVDF) membranes. Membrane blocking was done using 5% bovine serum albumin (BSA), overnight. Primary specific antibody versus H9N2 was used at a 1:1000 dilution, and secondary antibody was used at a 1:10000 dilution (HRP-conjugated). Immunoreactive bands visualized with chemiluminescence HRP substrate (Abcam). Results {#s3} ======= Protein assay before SDS-PAGE {#s3-0-1} ----------------------------- A protein assay was performed for influenza virus, and at this stage, BSA was used as a standard in volumes of 100, 300, 500 and 1000 mg/ml, and the results indicated that the volume of the viral solution was appropriate. According to [Table 1](#T1){ref-type="table"}, the highest and lowest amounts of protein were the sample F act 3 and the sample F act 5 with 387.07 mg/ml and 212.5 mg/ml, respectively. ###### Results of Protein assay ------------- ---------- --------------- Sample name OD 595nm Protein mg/ml St 100 0.08 100 St 300 0.31 300 St 500 0.53 500 St 1000 0.97 1000 F act 1 0.22 212.9 F act 2 0.32 309.67 F act 3 0.40 387.07 F act 4 0.27 261.29 F act 5 0.17 212.5 ------------- ---------- --------------- SDS-PAGE after protein assay {#s3-0-2} ---------------------------- Gels were prepared with concentrations of 2.5, 5 and 10 μg of each sample. According to our results ([Fig. 1](#F1){ref-type="fig"}), the molecular weight of the nucleoprotein was estimated between 45 and 66.2 kDa. {#F1} Preparative electrophoresis results {#s3-0-3} ----------------------------------- The absorption of the fractions was obtained from the purified nucleoprotein at 280 nm (UV). [Graphs 1](#F5){ref-type="fig"} and [2](#F6){ref-type="fig"} were plotted from the 121 fractions of the reduced and, 123 fractions of the non-reduced methods. Our results showed that 26 fractions from the reduced and 42 fractions from the non-reduced methods had a positive absorption, and these fractions contained viral NP. {#F5} {#F6} SDS-PAGE and Western blotting after protein purification {#s3-0-4} -------------------------------------------------------- SDS-PAGE was performed for the presence of NP after purification with preparative electrophoresis method ([Figs. 2](#F2){ref-type="fig"} and [3](#F3){ref-type="fig"}). According to the molecular weight of NP, fractions 2 and 3 (reduced purification) contained viral NP, but in fractions with non-reduced purification, no band was detected, and no viral protein was isolated. The NP band (H9N2) was verified again by immunoblotting assay ([Fig. 4](#F4){ref-type="fig"}). {#F2} {#F3} {#F4} Discussion {#s4} ========== The purpose of this work was to prepare and purify the influenza virus nucleoprotein in a simple, low-cost, applied and rapid method. In the current study, firstly the H9N2 virus identified and the locations of the viral proteins were determined. Subsequently the nucleoprotein was purified with adequate and accurate information about the viral proteins. The use of a live virus for propagation is necessary, but epidemiological hygiene needs to be considered, and this issue is more important in influenza viruses due to their genetic rearrangement and transmission potential to humans. However, the advantages of using the inactive virus are due to the lack of infectious potential and long-term maintenance and the absence of contamination with other microorganisms. The virus from dialysis was lyophilized to facilitate further protein purifications. By SDS-PAGE analyses, the virus protein bands were examined, and it was determined that a nucleoprotein with a molecular weight \~ 54 kDa is clearly identifiable \[between bands of 45 to 65 kDa of ladder\], and two bands (top and down), which are related to the HA protein, can be identified and separated. However, to ensure that the NP protein can be observed in low amounts of protein and the diluted sample, gels are stained with silver nitrate, which is about 10 to 100 times more accurate than the Coomassie blue. For purity and the presence of the desired protein, we examined the molecular weight and position of the nucleoprotein in terms of movement and detachment in the polyacrylamide gel; samples were run on a preparative electrophoresis machine. The Prep Cell 491 has a special constructional complexity and requires the precision and observance of the basic principles of assembling its parts. The components should be tightly assembled and the gel preparation, spraying, fractionation and droplet adjustment steps should be done correctly, and since the application and operation of the device is long, it is necessary to schedule proper testing. The accuracy of the purification of the influenza virus proteins is very important due to the small size and the number of proteins and types of proteins. By adjusting the fractionation device, the volume of 8 ml and duration of 20 minutes for each fraction, this step lasted 48 h. Initially, the fractions contain a protein-free stain which should be discarded. After preparing the fractions and determining their absorbance at 280 nm, the precise and pure presence of the nucleoprotein in fractions 2 and 3 was observed after complete removal of color, and this result was confirmed by SDS-PAGE, but in the non-reducing procedure, no purification was performed, and no band was found in SDS-PAGE. To increase the volume of pure nucleoprotein fractions, fractions containing the nucleoprotein were lyophilized. After lyophilization of the fractions containing the nucleoprotein, protein assay, SDS-PAGE and staining with silver nitrate were performed again. The significant points about the electrophoresis of the influenza virus are the importance of the preparation stages and purification of the virus and the electrophoretic conditions affecting the pattern of electrophoresis of viral proteins. So that with the slightest change, it is possible to observe a change in the electrophoretic pattern of the virus. The lack of similarity of the electrophoretic pattern of various influenza viruses can be attributed to this cause. To ensure the absence of the influenza virus genome in purified nucleoprotein, (because the nucleoprotein and genome of the virus are in the structure of the virus and the nucleoprotein is the viral genome coverage, and together with the RNA, the RNP forms), the resulting samples were electrophoresed. To check the accuracy of the results, the influenza virus before and after lyophilization, and a purified nucleoprotein after lyophilization were tested. The results of electrophoresis indicated no genomic band in the nucleoprotein sample. Thus, during testing, it has been proven that with the help of the Prep Cell 491, we will be able to completely purify the H9N2 avian influenza virus and save time and cost. Purified nucleoprotein can be used to provide nucleoprotein anti-sera against influenza viruses in future researches. Significant points in this study are the high speed and precision of the test results, which can be expressed as a superior advantage over the success of the research. Studies have been conducted on nucleoprotein and different method used for purification of nucleoprotein. Harmon et al., investigated an immunoassay for serological diagnosis of influenza type using a recombinant DNA producing nucleoprotein antigen and monoclonal human antibody IgG ([@R21]). They used immunoaffinity chromatography method for NP purification. This method is time-consuming and costly, and although the appropriate amount of nucleoprotein was purified, supplementary experiments and a large number of viral samples are required. In the present study, with the least number of the viral samples, nucleoprotein can be obtained. Li et al., (2014), isolated and purified the NP by SDS-PAGE for NP purification and then mass spectrometry was performed ([@R22]). In the SDS-PAGE electrophoresis method, viral proteins are isolated but not completely purified, and when cutting from the gel, there is a possibility of error. On the other hand, in this method, the accuracy and skill of the researcher are of great importance, and the cost of mass spectrometry is significant for confirmation of the protein. Conclusion {#s5} ========== Considering the results of our study, the purified nucleoprotein was obtained very dynamically, with high speed, precision, and purity. Purified nucleoprotein can be used to produce nucleoprotein vaccines, as well as to study structural, molecular and diagnostic and therapeutic materials. Acknowledgments {#s6} =============== We thank our colleagues from Department of Botany, Biotechnology and Bioinformatics, Payame Noor University who provided insight and expertise that greatly assisted the research. Conflict of Interests {#s7} ===================== The authors declare that they have no competing interests.

**BACKGROUND:** Skin flap vascularity is a critical determinant of aesthetic results in autologous ear reconstruction. In this work, we investigate the utility of intraoperative laser-assisted indocyanine green angiography (ICGA) as an adjunctive measure of skin flap vascularity in pediatric autologous ear reconstruction. **METHODS:** 21 consecutive pediatric patients undergoing first stage autologous total ear reconstruction were retrospectively evaluated. The first 10 patients were treated traditionally (non-ICGA) and the latter 11 patients were evaluated with ICGA intraoperatively after implantation of the cartilage construct and administration of suction. Relative and absolute perfusion units in the form of contour maps were generated. Statistical analyses were performed using independent sample Student's t test. **RESULTS:** Statistically significant differences in exposure and infection were not found between the two groups. However, decreased numbers of surgical revisions were required in cases with ICGA versus without ICGA (p=0.03), suggesting that greater certainty in skin flap perfusion correlated to a reduction in revision surgeries. In cases of exposure, we found an average lowest absolute perfusion unit of 14.3 whereas cases without exposure had an average of 26.1 (p=0.02), thereby defining objective parameters for utilizing ICGA data in tailoring surgical decision making for this special population of patients. **CONCLUSIONS:** Defined quantitative parameters for utilizing ICGA in evaluating skin flap vascularity may be a useful adjunctive technique in pediatric autologous ear reconstruction.

Introduction {#Sec1} ============ Neurologic diseases in the elderly, such as dementia and stroke, will pose an ever increasing burden on societies over the next couple of decades \[[@CR1]--[@CR4]\]. However, effective therapeutic or preventive strategies are lacking. In order to develop such strategies, knowledge on the etiology of these diseases is essential. A feature of neurodegenerative diseases is that morphologic brain pathology may be already present years before clinical onset and can be visualized using magnetic resonance imaging \[[@CR5]--[@CR10]\]. Realizing this potential benefit, already in the 1990s various population-based studies implemented neuroimaging to study the preclinical brain changes that ultimately lead to or may indicate an increased risk of developing clinically manifest diseases, such as dementia and stroke \[[@CR7], [@CR11]--[@CR19]\]. However, most studies implemented neuroimaging in only a subset of their population leading to limited sample sizes. More importantly, over the course of the last two decades MR imaging has undergone a dramatic improvement in hardware and software leading to higher field strengths, higher resolution, shorter scanning times, and more sensitive sequences. Furthermore, digital imaging techniques have led to a new field of research aimed at automating and increasing through-put of image processing for better visualization and quantification of imaging findings. Taken together, these developments now allow for performing neuroimaging in larger sample sizes and using state-of-the-art imaging and processing techniques. In turn, this has opened the way for more in depth and thorough investigation of (more subtle) brain changes that can lead to neurological diseases. It was in this light, that in 1995 the Rotterdam Scan Study was initiated to investigate risk factors and risk indicators of neurological diseases in the elderly using MR imaging to visualize the underlying brain changes and brain pathology. In 1995 and 1999, random subsets of the Rotterdam Study underwent neuroimaging off site in clinical scanners. From 2005 onwards, the Rotterdam Scan Study has been embedded within the core protocol of the Rotterdam Study. This was possible due to installation of a dedicate research scanner on site. The purpose of this paper is to give a general outline of the study population, scanning protocol, image post-processing and to discuss the main findings of the Rotterdam Scan Study, with the main focus on the period from 2005 to 2011. Design and study population {#Sec2} =========================== The source population of the Rotterdam Scan Study originates from the Rotterdam Study \[[@CR20]--[@CR23]\], which is a population-based study in the Netherlands and aims to investigate causes and determinants of chronic diseases in the elderly. The Rotterdam Study (RS I) was initiated in 1990 with 7,983 participants aged 55 years and over, who were interviewed and underwent physical exam at baseline and during follow-up visits every 3--4 years. In 2000 the cohort was expanded by 3,011 persons (RS II), who were aged 55 and over at that time. In 2006 the cohort was further expanded by 3,932 persons aged 45 years and over (RS III). The whole cohort undergoes re-examinations every 3--4 years. The total Rotterdam Study population encompasses 14,926 persons. Figure [1](#Fig1){ref-type="fig"} shows an overview of the various Rotterdam Study cohorts, the time of their (re-)examination visits, and the implementation of MRI-scanning in the core protocol of the Rotterdam Study in 2005. Initially, we invited random persons from the second visit of RS II to undergo MRI. Subsequently, we have scanned all eligible and consenting participants from the first visit of RS III and fifth visit of RS I. Currently, persons from the third visit of RS II are undergoing scanning. For 2012, participants from RS III are scheduled to be re-invited. Because the throughput of performing MR imaging has been higher than that of the Rotterdam Study (56 MRI slots per week versus 36 slots for regular study center visits), we were able to invite additional subsets for re-scanning. As such, we re-invited participants from RS I in 2006, RS II in 2008, and RS III in 2010 outside their regular visits for the Rotterdam Study. As a result, some of the participants are already undergoing their third MRI-exam.Fig. 1Overview of the subcohorts and examination visits of the Rotterdam Study, and imaging visits of the Rotterdam Scan Study. *Boxes* indicate examination visits of the three cohorts of the Rotterdam Study. *Boxes* with *solid colors* indicate visits, during which MRI imaging was conducted as part of the core protocol. Examination visits indicated with an '*X*' indicate extra visits during which only MRI was performed. The *red vertical line* indicates the implementation of MRI on site in the core protocol of the Rotterdam Study. In 1995 and 1999 (indicated with \*) 567 persons underwent MRI as part of the Rotterdam Scan Study outside the core protocol of the Rotterdam Study. '*Total*' indicates the total number of persons taking part in that Rotterdam Study examination visit. '*N*' indicates the number of persons that were eligible (non-demented and no MRI contra-indications) and invited to take part in the Rotterdam Scan Study. '*n*' indicates the number of persons taking part in the Rotterdam Scan Study Of all persons taking part in the Rotterdam Study, those who are demented or with MRI contra-indications are considered not eligible for the Rotterdam Scan Study. Furthermore, persons suffering from claustrophobia are also not considered further. Therefore, as of January 2011, a total of 5,886 brain MR-scans have been obtained on the research scanner. This includes multiple scans from the same person. Scan protocol {#Sec3} ============= In 2005, a 1.5T MRI unit (General Electric Healthcare, Milwaukee, USA, software version 11×) dedicated to the research project was installed in the Rotterdam Study research center. Besides the possibility of high throughput image acquisition, this enabled us to maintain acquisition parameters unchanged by excluding software or hardware upgrades in order to ensure data stability and comparability over time. The MRI unit was fitted with a dedicated 8-channel head coil (best coil configuration available at time of installation) and the possibility for parallel imaging using the array spatial sensitivity encoding technique (ASSET). Maximum total examination time (from arrival of one participant in the MRI suite until the next) was set at 45 min, in order to accommodate the MRI acquisition into the generic workflow of the Rotterdam Study. We therefore designed a 30 min brain imaging protocol that balanced the restrictions of time, costs and inconvenience for the participants with the relevance and quality of the acquired imaging data. To ensure participant compliance and reproducible image quality (reduce motion artefacts) an acquisition limit of 6 min per sequence was chosen. To facilitate easier applicability of the current MRI protocol by radiology technicians, we chose to use the standard brain imaging package delivered by the system manufacturer instead of custom developed sequences. The MRI sequences were chosen based on the primary variables of interest, i.e.:quantitative measures of brain tissue volumes and volumes of various neurostructures (e.g. hippocampus)quantitative assessment of white matter lesions,qualitative assessment of brain infarcts (lacunar and cortical) and microbleeds,quantitative assessment of white matter microstructural integrity and connectivity,quantitative assessment of total cerebral blood flow. In designing the protocol, we tried to meet both the time constraint and the contrast and resolution requirements. Where possible, we preferred 3D over 2D sequences because of higher signal-to-noise ratio (SNR), enabling the acquisition of smaller voxel sizes. Yet, acquisition time, sensitivity to motion and blurring artefacts did not allow 3D acquisition in all sequences. For each sequence, we adjusted the imaging parameters during optimization procedures to obtain a specific target resolution with adequate SNR (≥25) for tissues in the center of the brain while scan time was limited to 6 min for each sequence. The resulting protocol is presented in the Table [1](#Tab1){ref-type="table"} with all the relevant imaging parameters and the execution order listed. The protocol starts with a three-plane localizer, executed with the shimming option enabled. For subsequent sequences shimming is turned off to accelerate receiver adjustments. Morphological imaging is performed with T1-weighted (T1w), proton density-weighted (PDw) and fluid-attenuated inversion recovery (FLAIR) sequences. The combination of different MR contrasts provided by these sequences can be used for automated brain tissue and white matter lesion segmentation (see section on processing). For this purpose, the T1w scan is acquired in 3D at high in-plane resolution and with thin slices (voxel size \<1 mm^3^). A 3D T2\*-weighted gradient-recalled echo (GRE) scan is used to image cerebral microbleeds. For this sequence a TE \> 30 ms was selected to obtain stronger T2\*-weighting. For registration purposes, the same slice thickness with a lower in-plane resolution as compared to the 3D T1w scan is used. Parallel imaging is applied for this sequence to stay within the 6 min scan time limit.Table 1The magnetic resonance imaging protocol used in the Rotterdam Scan StudySequenceCommentModeReadout moduleTime (min:s)TR/TE (ms)TI (ms)BW (kHz)Flip angle (degrees)Number of slicesSlice thickness (mm)FOV (cm^2^)MatrixScout (1)Positioning2DGRE0:077.9/1.831.25303430256 × 256Scout (2)Localizer for 2D phase contrast scan; VENC = 60 cm/s2DGRE0:1224/9.08.061016032256 × 160PDw2DFSE6:0912300/17.317.8690--180901.625416 × 2562D Phase ContrastCarotid and basilar flow; VENC = 120 cm/s, NEX = 82DGRE0:5120/4.022.7381519256 × 160T1w3DGRE6:2413.8/2.840012.52096 (192)1.6 (0.8)25416 × 256FLAIR2DFSE6:258,000/120200031.2590--180642.525320 × 224ASSETCoil sensitivity correction data for calibration of parallel imaging2DGRE0:06150/1.831.257039103032 × 32DTI25 directions; b = 1,000mm2/s, b~0~ NEX = 32DEPI3:448,000/74.625090--180393.52164 × 96T2\*w3DGRE5:5545/3114.711396 (192)1.6 (0.8)25320 × 224*PDw* proton-density weighted, *T1w* T1-weighted, *FLAIR* fluid-attenuated inversion recovery, *ASSET* array spatial sensitivity encoding technique, *DTI* diffusion tensor imaging, *T2\*w* T2\*-weighted, *GRE* gradient-recalled echo, *FSE* fast spin echo, *TR* repetition time, *TE* echo time, *TI* inversion time, *BW* bandwith, *FOV* field of view, *VENC* velocity encoding, *NEX* number of excitations Diffusion tensor imaging (DTI) is used to quantitatively assess white matter microstructural integrity \[[@CR24], [@CR25]\]. For this 2D DTI scan, we use an echo planar imaging (EPI) readout with gradients (b = 1,000 s/mm^2^) applied in 25 directions \[[@CR24], [@CR25]\]. The b = 0 s/mm^2^ image is collected with NEX = 3. The number of gradient directions, i.e. 25, was chosen to best fit the optimized protocol by Jones et al. \[[@CR24], [@CR25]\] whilst remaining within time limits and maximum number of slices permitted by the scanner. To minimize geometrical distortions, the number of frequency encoding points was set to 64 and parallel imaging was applied with an acceleration factor of 2, with an imaging matrix of 64 × 96 and providing a voxel size of 3.3 × 2.2 × 3.5 mm^3^. An ungated 2D GRE phase contrast flow measurement is applied for assessment of total cerebral blood flow \[[@CR26]\], which has shown to be fast and accurate \[[@CR26]\]. A 2D thick slab projection phase contrast angiographic localizer (60 mm thick, velocity encoding (VENC) = 60 cm/sec) is positioned sagittally to determine the location of the carotid and basilar arteries. Next a thin slice perpendicular to all three vessels at the level of the precavernous internal carotid artery is positioned (VENC = 120 cm/s, slice thickness 5 mm, NEX = 8). Flow velocity data can be calculated from the phase difference images as described before \[[@CR26]\]. Figure [2](#Fig2){ref-type="fig"} illustrates an example of all the different sequences acquired in the final protocol. The sequence acquisition order was chosen in a way to provide adequate reconstruction speeds and to eliminate delays. The 3D T2\*w GRE scan was the last sequence executed in the protocol pipeline because of the long reconstruction time necessary for parallel imaging and the resulting lag time in scan execution.Fig. 2Depiction of the images acquired using the MRI protocol. First row: T1-weighted (**a**), proton-density-weighted (**b**), and fluid attenuated inversion recovery (**c**) images. Second row: T2\*-weighted (**d**) image, sagittal scout for the 2D phase contrast measurement (**e**), and the resulting flow image (**f**). Third row: example of an image acquired using diffusion tensor imaging (**g**), map of fractional anisotropy (**h**), and map of mean diffusivity (**i**) For quality check, a daily quality assessment (measuring transmit gain, center frequency and SNR) is performed by technicians using a phantom. In addition, weekly measurements of echo-planar stability, isocenter reliability and accuracy of absolute scaling along the cardinal axes (x, y, z) are performed. Regular scanner maintenance is performed by the manufacturer and results are filed in a log. Furthermore, interscan reproducibility measurements have been performed by re-inviting study participants (*n* = 20--30) within on average 2 weeks after initial examination for repeat MRI. Both for brain tissue segmentation as well as for diffusion tensor connectivity analysis, interscan reproducibility has been shown to be very good \[[@CR27]\]. Image processing {#Sec4} ================ Within the context of the Rotterdam Scan Study, a standardized and validated image analysis workflow is being developed to enable the objective, accurate, and reproducible extraction of relevant parameters describing brain anatomy, possible brain pathologies, and brain connectivity from multispectral MRI data. Image processing in the Rotterdam Scan Study has initially been centered around three main goals: First, owing to the sheer size and complexity of the imaging database being generated, we aim to automate the tedious task of manual analysis, both qualitative (e.g. use of rating scales, for example for cortical atrophy) and quantitative (e.g. manual contouring for volume analyses). Second, our goal is to replace qualitative image assessment by quantitative analyses as much as possible. Third, we aim to limit or avoid altogether inter- and intraobserver variability. In recent years, image processing has been expanded to also allow the extraction of relevant image-derived parameters that would not be feasible manually or cannot be assessed visually. In the following paragraphs, we briefly describe the different quantitative image analysis methods that have been developed and/or employed within the Rotterdam Scan Study. Image pre-processing {#Sec5} -------------------- Prior to analysis, a number of pre-processing steps are performed. For multispectral image analysis, the different scans are spatially registered using rigid registration. Subsequently, the brain is extracted from the scan. Hereto a manually segmented brain mask, which excludes a.o. the cerebellum, eyes and skull, is non-rigidly registered to the T1-weighted image using Elastix \[[@CR28]\]. Finally, scans are corrected for intensity non-uniformity using the N3 method \[[@CR29]\]; non-uniformity correction is carried out within the brain mask. Brain tissue segmentation {#Sec6} ------------------------- Automated brain tissue segmentation on MRI has received considerable attention \[[@CR11], [@CR30]--[@CR39]\]. An important distinction that can be made is whether methods are supervised (i.e. they depend on annotated training data), or unsupervised. In the Rotterdam Scan Study we have used two types of supervised approaches, both based on k-nearest neighbour (kNN) segmentation. In kNN segmentation, image voxels are assigned labels (grey matter (GM), white matter (WM), cerebrospinal fluid (CSF) or background (BG)) based on the most similar voxels in the training data. Similarity here depends on the distance in normalized MR image intensities. We have both investigated segmentation based on T1w images, and multispectral MRI data (T1w and PDw images). The two approaches we have used differ in the way that the training data are obtained. In the first approach manual segmentations by two observers of six T1w datasets (the PDw dataset is implicitly segmented after rigid registration to the T1w datasets), that include labels for GM, WM, CSF, and BG, were used as training data \[[@CR30], [@CR40]\]. In the second approach the training data were obtained from the scan to be segmented itself. Hereto, the manually segmented scans are used as atlases and are non-rigidly transformed to the scan, creating probability density maps that indicate the likeliness of each voxel to belong to a certain tissue type. After random selection of likely voxels for each class, and a pruning step to remove outliers, the training samples for WM, GM, CSF and BG are obtained \[[@CR40], [@CR41]\]. Both methods have been extensively evaluated within the context of the Rotterdam Scan Study, showing good accuracy and reproducibility \[[@CR27], [@CR40]\]. An example of the automated tissue segmentation is shown in Fig. [3](#Fig3){ref-type="fig"}.Fig. 3K-nearest neighbor based tissue segmentation into grey matter (*red*), white matter (*yellow*), cerebrospinal fluid (*blue*), and background (*black*) To facilitate more regionalized analysis of total brain, WM and GM volumes, individual lobes have been segmented. This is achieved by non-rigidly registering a template image in which the lobes have been manually outlined \[[@CR42]--[@CR44]\]. White matter lesion classification {#Sec7} ---------------------------------- The brain tissue segmentation methods described above have been complemented with white matter lesion (WML) segmentation \[[@CR45]\]. Hereto, both the brain tissue segmentation, and the FLAIR image is used. In the brain tissue segmentation, possible WMLs are misclassified as GM with a ring of WM voxels. In the FLAIR image the WMLs are hyperintense. We therefore process the histogram from the FLAIR image intensities of all voxels that are classified as GM, to estimate the mean and standard deviation of true GM voxels. Subsequently, WML voxels are extracted by intensity thresholding, where the threshold depends on the estimated GM distribution. False positives are removed by excluding voxels which are not sufficiently connected to the white matter. The different parameters (intensity threshold, and quantitative definition of not being sufficiently connected) have been optimized on large reference dataset. The method has been quantitatively evaluated \[[@CR45]\] and has successfully been applied to the whole cohort \[[@CR46]\]. Visual inspection of the results indicates that the method is robust, with approximately only 4% of the scans containing false positive or false negative WMLs. Figure [4](#Fig4){ref-type="fig"} demonstrates the automated WML segmentation result.Fig. 4White matter lesion segmentation (*purple*), using both the FLAIR image (*left*) and brain tissue segmentation (Fig. [3](#Fig3){ref-type="fig"}) as input Brain structure segmentation and shape analysis {#Sec8} ----------------------------------------------- Within the Rotterdam Scan Study we have developed a graph cut framework for neurostructure segmentation \[[@CR47]\] combining atlas registration and statistical models of image appearance \[[@CR48]\]. As the hippocampus is one of the first regions to be affected by Alzheimer's disease, most work has focused on developing and optimizing the framework for this application. The developed framework requires annotated training data (atlases), which are used both for atlas registration, and for training the statistical image appearance models. Two observers manually outlined the hippocampus in twenty datasets, following a well-described protocol \[[@CR49]\]. The twenty atlases are nonrigidly registered to an image to be segmented, after which by averaging a spatial probability map is obtained which indicates the likeliness of a voxel to belong to the hippocampus. Within the graph cut framework, this information is complemented by the likeliness that a voxel is part of the hippocampus based on intensity information. Hereto voxel intensities are related to the expected intensity distributions of the hippocampus and the surrounding tissue. The method has been shown to improve on existing manual hippocampus segmentation techniques \[[@CR48]\], and has been applied to a number of studies \[[@CR50], [@CR51]\]. An example of hippocampus segmentation is shown in Fig. [5](#Fig5){ref-type="fig"}. The graph cut framework developed for hippocampus segmentation has additionally been used for ventricle segmentation \[[@CR52]\], and segmentation of the cerebellum \[[@CR53]\]. We are working on further extending the applicability to other neurostructures.Fig. 5Hippocampus segmentation, which is used as a basis for both volumetric and shape analysis Currently, our focus extends beyond plain volumetry to incorporating shape-based analysis. Therefore, we have developed a method to quantify hippocampal shape, and investigated its potential to predict dementia. Because our aim is prediction, the most suitable datasets are those obtained in 1995, which have more than 10 years of clinical follow-up. We compared MRI scans of 47 persons who were cognitively normal at scan time and developed dementia during 10-years of follow-up, with 47 age- and gender-matched controls who remained free of dementia. A statistical shape model was built from the 94 segmentation results and was used to predict incident dementia. We found that the combination of hippocampal volume and hippocampal shape performed better than when just using volume \[[@CR54]\]. Diffusion tensor imaging: global and tract-based analysis {#Sec9} --------------------------------------------------------- DTI enables measurement of the microstructural integrity of white matter. Within the Rotterdam Scan Study, a number of image analysis techniques have been employed and developed for the analysis of DTI data. These include conventional global and regional analysis of DTI-derived measures such as Fractional Anisotropy (FA) and Mean Diffusivity (MD) \[[@CR55]\], and tract-based analysis of FA \[[@CR56]\]. Global and regional analysis of FA and MD has been performed using the FSL toolbox \[[@CR57]\], and consisted of Eddy current correction, head motion correction, skull stripping and tensor model fitting. As discussed below, DTI data were registered with the other imaging data to study relations between atrophy, white matter lesions, and DTI-derived measures. Tract-based analysis of DTI enables a more localized comparison of FA and MD between groups. In the Rotterdam Scan Study, tract-based analysis has been achieved using tract-based spatial statistics (TBSS) \[[@CR58]\] a technique that creates a common skeleton of the white matter tracts from a series of images, onto which for each individual the local maximum FA value is projected. This enables robust voxelwise statistical analysis of the microstructural integrity of white matter across persons \[[@CR56]\]. Statistical analysis is performed with multiple linear regressions, adjusting for age, sex and multiple comparisons. In addition, to study relations between macroscopic white matter changes and white matter microstructure, DTI data are aligned with the other MR imaging data. Diffusion tensor imaging: connectivity analysis {#Sec10} ----------------------------------------------- Using deterministic or probabilistic tractography, DTI can also be used to study structural connectivity of the brain. In order to compare structural connectivity across persons in the Rotterdam Scan Study, we developed a novel framework, SAMSCo, which enables construction of weighted structural brain connectivity networks which can be effectively analyzed using statistical methods \[[@CR59], [@CR60]\]. The weighted networks are obtained using a minimum cost path (mcp) method with an anisotropic local cost function based on the locally estimated diffusion tensor weighted images. Start and end regions of the mcp were defined by a FreeSurfer segmentation \[[@CR61], [@CR62]\] of subcortical structures and cortical parcellation. In proof of principle experiments we showed that the brain connectivity maps contain information on age and sex \[[@CR59], [@CR60]\]; we hypothesize that it can also be used for getting valuable insight into normal brain ageing and disease progression. Using a re-scan on 30 persons, good reproducibility of the connectivity maps was shown \[[@CR27]\]. Visual ratings {#Sec11} ============== Scan quality and incidental findings {#Sec12} ------------------------------------ Each MRI scan that is acquired is visually examined within 1 week after acquisition by a research physician in the Rotterdam Scan Study. During this visual inspection, the MRI scan is rated for quality and the presence and severity of motion artefacts or signal inhomogeneity (for example due to metallic implants) is recorded. Furthermore, each scan is evaluated by trained research physicians for presence of incidental findings, i.e. abnormalities of potential clinical relevance that were previously unknown, that are unexpected and that are unrelated to the purpose of the scan \[[@CR63], [@CR64]\]. All potential findings are recorded in a database and are in a later stage evaluated by an experienced neuroradiologist. The most frequent findings in the Rotterdam Scan Study are aneurysms (1.8%) and benign primary tumors (1.6%) \[[@CR63]\]. Referral of participants for further medical examination occurs in accordance with an expert-defined protocol \[[@CR63]\]. Ischemic small vessel disease on MRI {#Sec13} ------------------------------------ Infarcts are rated on FLAIR, proton density-weighted, and T1-weighted sequences. Lacunar infarcts are defined as focal lesions ≥3 and \<15 mm in size with the same signal characteristics as CSF on all sequences, and (when located supratentorially) with a hyperintense rim on the FLAIR sequence \[[@CR63]\]. Lesions ≥15 mm in size, but otherwise similar, are rated as subcortical infarcts. Infarcts showing involvement of cortical gray matter are classified as cortical infarcts. Cerebral microbleeds {#Sec14} -------------------- All 3D T2\* GRE scans are reviewed for the presence, number, and location of cerebral microbleeds. Microbleeds are defined as focal areas of very low signal intensity on T2\*-weighted imaging that are not accompanied by evident signal abnormality on other structural sequences \[[@CR65]\]. Microbleed location is categorized into one of three locations: lobar (cortical gray and subcortical or periventricular white matter), deep (deep gray matter: basal ganglia and thalamus, and the white matter of the corpus callosum, internal, external, and extreme capsule), and infratentorial (brainstem and cerebellum) \[[@CR65]\]. Intraobserver and interobserver reliabilities for microbleed rating are very good (κ = 0.85--0.87 \[[@CR66]\]) and review of the intial ratings by an experienced neuroradiologist yielded very high accordance as well \[[@CR66]\]. Visual check of automated processing performance {#Sec15} ------------------------------------------------ Though post-processing for tissue segmentation and structure segmentation takes places fully automated and without user interaction, all end results are visually checked for performance. For example, small motion artefacts in the FLAIR sequence that do not necessitate exclusion of an MRI scan may interfere with white matter lesion segmentation and cause false positive lesions after automated segmentation; or brain masking may result in minimal inclusion of dura or skull. After visual inspection, manual editing of any errors is needed in less than 10% of scans, depending on the type of post-processing. Furthermore, less than 1% of scans are excluded based on artefacts that are only discovered after automated post-processing (for example motion or susceptibility artefacts in diffusion tensor images, which are not apparent in the raw unprocessed data). Main findings {#Sec16} ============= Vascular factors and brain pathology {#Sec17} ------------------------------------ Previous experience with brain MRI in the Rotterdam Study had learned that markers of small vessel disease such as white matter lesions and lacunar infarcts are highly prevalent in the elderly and that these relate to common cardiovascular risk factors \[[@CR44], [@CR67]--[@CR78]\]. In the MRI scans obtained from 2005 onwards, we confirmed the frequent occurrence of WML and infarcts in the elderly and further extended the prevalence and volume estimates to the middle aged population \[[@CR63]\]. Furthermore, cerebral microbleeds were added as a new marker of vascular brain pathology, and their interrelationship with 'classic' markers of small vessel disease was examined (see further below). In addition to known classic cardiovascular risk factors, we examined cerebral blood flow as a measure of brain perfusion and studied its determinants and association with cognition and vascular brain disease, as further described below. In line with this, we investigated the association between arterial calcification (measured with computed tomography), a relatively new marker of atherosclerosis, and subclinical vascular disease and found arterial calcification in various vessel beds to be related to both WMLs and lacunar infarcts \[[@CR79]\]. Besides our quantitative measures of brain tissue atrophy and lesion volume, DTI added new quantitative information on tissue microstructural integrity, as further elaborated upon below. Finally, in collaboration with research partners in the CHARGE consortium \[[@CR80]\], we performed a genome-wide association study of WML burden and were able to identify 6 novel risk-associated SNPs on chromosome 17q25 \[[@CR81]\], which we subsequently replicated in a separate Rotterdam Scan Study cohort \[[@CR46]\]. After *Apolipoprotein E* (*APOE*) this is the first common genetic variant to be robustly associated with brain pathology on MRI. Similar analyses have been undertaken for brain infarcts \[[@CR82]\]. These are only the first of more projects that will combine genetic information and imaging markers in the Rotterdam Scan Study. Cerebral microbleeds {#Sec18} -------------------- In 200 Rotterdam Scan Study participants, we showed that the 3D high-resolution T2\* GRE sequence detected more microbleeds in more persons, compared with a conventional MRI sequence. When we applied this high-resolution sequence in 1,062 participants, we found that cerebral microbleeds were present in 1 in 5 persons over age of 60 and in over 1 in 3 in persons aged 80 years and older \[[@CR66]\]. This prevalence is much higher than reported previously, which in part may be explained by the use of a more sensitive MRI sequence. We recently updated this prevalence estimate for a larger population (4,006 persons, mean age 60.3 years) and found that it still holds \[[@CR83]\]. Furthermore, we have consistently shown that risk factors for microbleeds vary according to the location of microbleeds in the brain. Cardiovascular risk factors and markers of ischemic small vessel disease were related to deep or infratentorial microbleeds, whilst *APOE* genotype related to strictly lobar microbleeds \[[@CR66], [@CR83]\]. This is indirect evidence that deep or infratentorial microbleeds result from arteriolosclerotic angiopathy, whereas strictly lobar microbleeds are caused by cerebral amyloid angiopathy. This is also corroborated by our finding that spatial distribution of lobar microbleeds in our population follows the known topographic distribution of amyloid angiopathy \[[@CR84]\]. The high prevalence of cerebral microbleeds and their potential link with bleeding-prone microangiopathy raised our interest in how these relate to antithrombotic drug use. We found that persons who had used or were using antithrombotic medication significantly more often had cerebral microbleeds \[[@CR85]\]. Longitudinal research needs to confirm this association and may unravel a potential link with risk of symptomatic intracerebral hemorrhage. Interestingly, we found a potential link between microbleeds and symptomatic intracerebral hemorrhage when investigating lipid factions as risk factor for both \[[@CR86]\]. We also studied the occurrence of incident microbleeds on follow-up MR imaging after an interval of 3--4 years. We found that 10% of persons developed new microbleeds and that this was strongly related to presence of microbleeds at baseline, but also to various risk factors such as hypertension and smoking \[[@CR87]\]. Again, risk factors differed according to the location of new microbleeds, lending further support to the hypothesis on microbleed etiology. Finally, we recently showed that presence of numerous microbleeds, especially in a strictly lobar location, was associated with worse performance on cognitive tests, even after adjustment for vascular risk factors and other imaging markers of small vessel disease \[[@CR88]\]. These results suggest an independent role for microbleed-associated vasculopathy in cognitive impairment. Cerebral blood flow {#Sec19} ------------------- Total cerebral blood flow (tCBF) and total brain perfusion (tCBF per 100 ml brain tissue) were measured with 2D phase contrast imaging as described above. We showed a close relationship between tCBF and markers of the microvasculature, e.g. retinal vessel diameters \[[@CR77]\]. In 892 persons aged 60 years and older, we further showed that determinants of tCBF and total brain perfusion differed largely, due to the large influence of brain volume on tCBF values \[[@CR89]\]. Higher pulse pressure, higher body mass index, and current smoking were all associated with higher total brain perfusion. Furthermore, persons with low total brain perfusion had significantly more white matter lesions compared to those with high total brain perfusion. The role of tCBF with cognitive performance appeared more complex with brain atrophy either confounding or mediating the association \[[@CR90]\]. White matter microstructural integrity {#Sec20} -------------------------------------- We have found DTI parameters within white matter lesions and normal-appearing white matter to be associated with cognitive function, even when taking into account volume of white matter lesions and white matter atrophy \[[@CR55]\]. This indicates that the deleterious effect of white matter changes on cognition not only depends on lesion burden or amount of atrophy, but also on characteristics that are not easily evaluated by conventional MRI. Also, we demonstrated that age-related changes in the normal-appearing white matter are primarily explained by white matter atrophy and white matter lesion formation and not by the aging process in itself. Furthermore, we found that white matter atrophy and white matter lesion formation related to loss of integrity in distinct brain regions, indicating that the two processes are not sequential events but are rather independent and thus pathophysiologically potentially different \[[@CR56]\]. Future perspectives {#Sec21} =================== The Rotterdam Scan Study provides a unique environment to study the etiology of neurological diseases in the elderly. Nearly 6,000 persons have already undergone brain MRI-scanning and in coming years repeated waves of follow-up examination will ensure a wealth of imaging data, both cross-sectional and longitudinal. So far, the focus of research has been on finding risk factors and determinants of brain pathology. In the coming years we hope to expand the scope of research in several ways, which include novel image sequences, novel post-processing techniques, novel risk factors, and study the relationship of imaging markers with clinical outcomes. Currently, we are performing pilot studies to investigate the feasibility of resting state functional MRI in the Rotterdam Scan Study. This will allow us to explore not only structural markers of brain pathology, but also functional correlates thereof. The current planning is to fully incorporate BOLD-fMRI in the core scanning protocol by the end of 2011. In the coming years ongoing hardware developments will necessitate upgrading of the scanner. Although there are no current plans to change the scanner, we foresee upgrading to a 3 Tesla scanner in a few years. Standardized and evaluated automated image processing techniques are crucial in exploiting the rich information that is available in population imaging data. They have enabled a transition from qualitative image interpretation into quantitative imaging. Quantitative imaging is non-trivial; it requires standardization and optimization in all the steps from data acquisition, to data analysis in structured reporting. In the next years, the number of validated quantitative image analysis techniques will further increase, also including more complex measures. Also, the results of these analyses will increasingly be stored in a standardized manner. This will enable a richer characterization of brain anatomy, pathology, connectivity and function. Relating these quantitative image measures to disease status, progression and events, will be a powerful tool for development of novel diagnostic and prognostic quantitative imaging biomarkers. The standardization of image acquisition and processing protocols also implies that high quality reference data are being acquired. These can be used in a clinical setting to contrast findings in an individual with a reference population. Our visual ratings currently include infarcts and microbleeds. We are exploring the possibilities of developing a validated visual rating for enlaraged Virchow Robin-spaces (perivascular spaces). Evidence suggests that these may be a hallmark of either small vessel disease or atrophy \[[@CR91], [@CR92]\]. Finally, the concomitant continuous monitoring of all participants in the Rotterdam Study ensure that we have a wealth of clinical data available, including cognitive performance and the occurrence of dementia and stroke. By the end of 2012, we will already have 6 years of clinical follow-up for the first participants of the Rotterdam Scan Study. In the coming years we hope to investigate how MRI markers of brain pathology relate to these clinical outcomes. It will be especially interesting to investigate how the recently introduced markers, i.e. microbleeds, microstructural integrity, and cerebral blood flow, relate to incident disease, and whether these markers can contribute additional information on top of established markers, such as white matter lesions and infarcts. The authors thank all participants, technicians, and research staff of the Rotterdam Scan Study. We are also grateful to the general practitioners of the Ommoord area for their continued support. We thank Piotr Wielopolski and Henri Vrooman for their continued technical assistance during the study. The Rotterdam Scan Study is supported by the Erasmus MC University Medical Center, the Erasmus University Rotterdam, the Netherlands Organization for Scientific Research (NWO) grant 918-46-615, the Netherlands Organization for Health Research and Development (ZonMW), the Research Institute for Disease in the Elderly (RIDE), the Alzheimer's Association (Dr Vernooij, grant NIRG-09-131680; Dr van der Lugt, grant NIRG-08-91393), the Netherlands Heart Foundation (Dr Ikram, grant 2009B102), and the Internationaal Parkinson Fonds (Dr Ikram). Further support was obtained from General Electric Health Care. **Open Access** This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.

**Correction to**: *Leukemia*; doi:[10.1038/leu.2012.119](/doifinder/10.1038/leu.2012.119); advance online publication, 3 May 2012 Since the publication of this article, the authors have noticed an error in Figure 4a, specifically that the structures of the methyl-pomalidomide enantiomers were missing a carbonyl group. The error has now been rectified, and the correct article (with the correct figure 4) appears in this issue. The html and online pdf versions have also been rectified, and now carry the correct paper. The authors would like to apologise for any inconvenience this may have caused.

1. Introduction {#sec1-pharmaceutics-12-00422} =============== Electroporation is a physical method for exogenous molecule delivery through the plasma membrane that is used to increase the plasma membrane permeability by applying short (ns--ms) but strong electric pulses \[[@B1-pharmaceutics-12-00422]\]. The permeability is increased due to the electric-field-induced transmembrane voltage, resulting in the formation of transient hydrophilic pores in the cell membrane \[[@B2-pharmaceutics-12-00422]\]. Pore formation starts when transmembrane voltage exceeds an electroporation threshold ranging between 0.2 and 1 V \[[@B3-pharmaceutics-12-00422],[@B4-pharmaceutics-12-00422]\]. During the prolonged application of the over-threshold electric field, the electropore density and/or size increases \[[@B5-pharmaceutics-12-00422]\]. However, when the external electric field is switched off, the electric-field-induced transmembrane potential dissipates, initiating the process of pore resealing \[[@B6-pharmaceutics-12-00422]\]. The presence of electropores---either reversible or irreversible---enables bidirectional transport across the cell membrane \[[@B7-pharmaceutics-12-00422]\]. Such increased permeability can be used for the effective delivery of small charged molecules (e.g., fluorescent dyes \[[@B8-pharmaceutics-12-00422]\] or membrane-impermeable chemotherapeutic drugs \[[@B9-pharmaceutics-12-00422]\]) to the cells. The combined application of electroporation and chemotherapeutic agents is termed electrochemotherapy and is a successful method with clinically proven efficiency in treating different tumours \[[@B10-pharmaceutics-12-00422]\]. In the recent years, intracellular calcium ion (Ca^2+^) delivery via electroporation has been suggested as an alternative to electrochemotherapy with bleomycin or cisplatin \[[@B11-pharmaceutics-12-00422]\]. A recent clinical trial has shown that calcium electroporation has comparable efficiency as electrochemotherapy and is feasible and effective in patients with cutaneous metastases \[[@B12-pharmaceutics-12-00422]\]. Ca^2+^ ions are universal signal mediators that regulate many cellular functions \[[@B13-pharmaceutics-12-00422]\]. At physiological conditions, the concentration of Ca^2+^ ions in the cytoplasm ranges between 10^−8^--10^−7^ M and between 10^−3^--10^−2^ M in the extracellular environment. Calcium homeostasis is maintained by the interplay between multiple types of energy-dependent pumps and passive directional transporters located in plasma membranes and organelles \[[@B14-pharmaceutics-12-00422]\]. Calcium ion distribution between extra- and intra-cellular compartments plays a crucial role in cellular response to various external stress conditions, and they are also involved in certain types of cell death, for instance, intracellular Ca^2+^ overload initiated necrotic and apoptotic processes \[[@B11-pharmaceutics-12-00422],[@B15-pharmaceutics-12-00422]\]. Calcium is also known for its role in the resealing of an injured membrane \[[@B16-pharmaceutics-12-00422]\]. There are several ways that Ca^2+^ ions are involved in membrane resealing, further delineating the importance of maintaining membrane integrity \[[@B17-pharmaceutics-12-00422]\]. Therefore, in this study, we aimed to investigate the impact of different CaCl~2~ concentrations on the electrotransfer of small charged molecules. For this purpose, Chinese hamster ovary (CHO) cells were electroporated in media with different CaCl~2~ concentrations using either microsecond or nanosecond electroporation. Microsecond electroporation is the "classic" mode of electroporation and utilizes electric pulses of µs--ms duration. In contrast, the nanosecond duration pulses have their duration below 1 µs, and much higher pulse strengths. Due to the short duration, they can generate transmembrane potential on intracellular membranes instead or in addition to the plasma membrane \[[@B18-pharmaceutics-12-00422],[@B19-pharmaceutics-12-00422]\] and are lucrative for medical applications because, unlike microsecond duration pulses, they do not induce muscle twitching \[[@B20-pharmaceutics-12-00422]\]. Propidium iodide (PI), YO-PRO-1 and ethidium bromide (EtBr) were used as fluorescent probes to investigate the uptake of small charged molecules after electroporation. The obtained results show that CaCl~2~ presence in the electroporation media prior electric field application impedes the uptake of small charged fluorescent molecules through the plasma membrane. 2. Materials and Methods {#sec2-pharmaceutics-12-00422} ======================== 2.1. Cell Line {#sec2dot1-pharmaceutics-12-00422} -------------- Chinese Hamster Ovary (CHO) cells (European Collection of Authenticated Cell Cultures, 85050302) were used for in vitro experiments. CHO cells were cultivated in Dulbecco's Modified Eagle Medium (DMEM) (Sigma, Darmnstadt, Germany) supplemented with 10% Foetal Bovine Serum (FBS) (Sigma), 1% [l]{.smallcaps}-glutamine (Sigma) and 1% penicillin-streptomycin solution (Sigma). The cells were grown in monolayers in 10 cm Petri dishes (Techno Plastic Products (TPP) (Trasadingen, Switzerland) and incubated at 37 °C in 5% CO~2~ atmosphere. To maintain the culture, cells were passed every 2--3 days and 24 h before the experiments. 2.2. Cell Electroporation {#sec2dot2-pharmaceutics-12-00422} ------------------------- After harvesting, cells were washed and diluted in electroporation medium (10 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) (Lonza, Basel, Switzerland), 250 mM sucrose (Sigma), 1 mM MgCl~2~ (Sigma) in sterile water) that was supplemented with different CaCl~2~ (Sigma) concentrations (0, 0.0001, 0.001, 0.01, 0.1, 0.25, 0.5, and 1 mM). The HEPES buffer with 0 mM CaCl~2~ was used as the experimental control. Cell suspensions were prepared at a concentration of 1.8 × 10^6^ cells/mL, using buffers with different CaCl~2~ concentrations. For each experimental point, 6.5 × 10^4^ cells (36 μL) were used. Cuvette electrodes with a 1-mm gap between electrodes were used for cell electroporation. For small molecule electrotransfer experiments, the cell solution was supplemented with 4 μL propidium iodide (PI) (Sigma, Darmnstadt, Germany) at the final concentration of 40 µM, 4 µL ethidium bromide (EtBr) (Carl Roth, Karlsruhe, Germany) at the final concentration of 40 µM or 4 µL YO-PRO-1 (Thermo Fisher Scientific, Dublin, Ireland) at the final concentration of 4 µM. For the experiments to test the medium conductivity effect on the efficiency of PI electrotransfer CaCl~2~ was replaced by MgCl~2~. For the experiments investigating pore resealing dynamics, PI was present in the electroporation medium during electroporation only for positive controls and was introduced into the solution at 15, 30, 60, 120, 360, 480 or 600 s after electroporation. As the PI can only enter the cells with compromised plasma membrane integrity, the percentage of PI positive cells in these experiments reflect the percentage of cells that did not reseal after electroporation. For micro-electroporation, CHO cells were electroporated using 1 square HV pulse (800--1800 V/cm pulse strength, 100 μs pulse duration) using BTX T820 electroporator (Harvard Apparatus, Holliston, MA, USA) For nano-electroporation, cells were electroporated using 10 square HV pulses (10,000--18,000 V/cm pulse strength, 200 ns pulse duration, 1 Hz repetition frequency) using high voltage pulse generator (VGTU, Vilnius, Lithuania). After 15 min incubation at 37 °C in 5% CO~2~ atmosphere, cell permeabilization was evaluated using flow cytometry (BD Accuri C6, BD Biosciences, Franklin Lakes, NJ, USA). A total of 10^4^ cells per sample at 66 µL/min flow rate and 22 µm core size were collected. The cells were excited with 488 nm laser and fluorescence was collected using 585/40- and 533/30-nm bandpass filters. The flow cytometry gating strategy to collect the cells after various conditions and for all three molecules used are depicted in [Figure 1](#pharmaceutics-12-00422-f001){ref-type="fig"}. 2.3. Visualization of Electroporation {#sec2dot3-pharmaceutics-12-00422} ------------------------------------- For the visualization of PI electrotransfer experiments, 22 × 22 mm glass cover slips (Carl Roth, Karlsruhe, Germany) were immersed into 70% ethanol for 15 min. Then, the coverslips were removed from the ethanol and placed into the 4-cm Petri dishes and airdried for 15 min. Then, 40 µL of cell suspension (1.8 million per mL) was placed on the middle of the coverslip in the Petri dish. After an additional 5-min incubation (for the cells to settle down on the surface) and a gentle 2-mL addition of DMEM (supplemented with 10% FBS, 1% [l]{.smallcaps}-glutamine and 1% penicillin-streptomycin solution), the Petri dishes with cells were placed in the incubator (37 °C and 5% of CO~2~). After 5 h of incubation, the coverslips with cells attached (5 h incubation) was taken from the petri dishes and placed on the laboratory made electrodes (copper foil mounted on the objective glass) with a 2-mm gap. Electroporation medium (HEPES buffer supplemented with different concentrations of CaCl~2~ (0, 0.25, 0.5, 1 mM) and supplemented with PI (40 µM final concentration) was added between the gap of electrodes prior to the coverslip being mounted on the top of the electrodes. The cells were electroporated using 1 HV pulse (1400 V/cm or 1800 V/cm pulse strength, 100 μs pulse duration) delivered by a BTX T820 pulse generator. A Motic AE31 fluorescent microscope mounted with a MoticamPro 285B camera was used for cell imaging. For PI fluorescence imaging, a filter cube (D560/40X excitation, dichroic 595DCLP mirror, D630/60 emission) was used. Motic Images Advanced 3.2 software was used to obtain the images. At all conditions, phase contrast images were taken prior to the application of electric fields. For the fluorescence imaging time lapse was switched on before electroporation. Single HV pulse was delivered at time '0 s'. Then, additional fluorescent images were taken at every second for 300 s. Open source image processing software ImageJ (Version 1.52p, National Institute of Health, Bethesda, MD, USA) was used to calculate corrected total cell fluorescence (CTCF) \[[@B21-pharmaceutics-12-00422]\]. The phase contrast image was used to determine the area of the cell, in which the changes of the propidium iodide fluorescence were measured over time. CTCF was calculated by multiplying the cell area (the area in phase contrast image delineated by single cell) with its average fluorescence. The changes in CTCF allowed to monitor the dynamics of PI entry. The CTCF mean values in a corresponding figure were made from the observation of at least 20 different cell images. 2.4. Model for Computation {#sec2dot4-pharmaceutics-12-00422} -------------------------- The model is mathematically defined in Equations (1), (2) and (3) \[[@B22-pharmaceutics-12-00422]\]. $$\Delta\Phi_{m}\left( t \right) = fERcos\theta~\left\lbrack {1 - e^{\lbrack{- \frac{t}{\tau}}\rbrack}} \right\rbrack$$ where Δ*Φ* is transmembrane voltage, *f* is the shape factor, *R* is the cell radius, *θ* is the angle measured from the centre of the cell with respect to the direction of the field, *t* is the time elapsed since the onset of the field, and *τ* is the time constant of membrane charging. $$f = \frac{3\lambda_{ο}\left\lbrack {3dR^{2}\lambda_{i} + \left\lbrack {3d^{2}R - d^{3}} \right\rbrack\left\lbrack {\lambda_{m} - \lambda_{i}} \right\rbrack} \right\rbrack}{2R^{3}\left( {\lambda_{m} + 2\lambda_{ο}} \right)\left( {\lambda_{m} + \frac{1}{2}\lambda_{i}} \right) - 2{(R - d)}^{3}\left( {\lambda_{ο} - \lambda_{m}} \right)\left( {\lambda_{i} - \lambda_{m}} \right)}$$ where *λ~o~*, *λ~m~* and *λ~i~* are the conductivities of the external, membrane and cytoplasm, respectively, *R* is the radius of the cell and *d* is the thickness of the membrane. $$\tau = \frac{Rc_{m}}{\frac{2\lambda_{ο}\lambda_{i}}{2\lambda_{ο} + \lambda_{i}} + \frac{R}{d}\lambda_{m}}$$ where *τ* is the membrane charging time, *R* is the radius of the cell, *C~m~* is the capacitance of the membrane, *λ~o~*, *λ~m~* and *λ~i~* are the conductivities of the external, membrane and cytoplasm, respectively, and *d* is the thickness of the membrane. In all the situations, the transmembrane potential change at cell poles (cos(0°) = 1) was calculated. The thickness of the membrane d was 3 × 10^−9^ m, as described in \[[@B23-pharmaceutics-12-00422]\], the cytoplasm conductivity *λ~i~* was 0.5 S/m, as described in \[[@B24-pharmaceutics-12-00422]\], membrane conductivity *λ~m~* was 3 × 10^−7^ S/m as described in \[[@B22-pharmaceutics-12-00422]\] and membrane capacitance *C~m~* was 10^−2^·F·m^−2^ as described in \[[@B22-pharmaceutics-12-00422]\]. The average cell radius *R* was experimentally evaluated by calculating the radii of \>100 cells in microscopy pictures using open-source image processing program ImageJ. The *R* value determined this way was 9.7 µm. The specific conductivity values of the extracellular media with different CaCl~2~ concentrations were measured and the results are presented in [Table 1](#pharmaceutics-12-00422-t001){ref-type="table"}. 2.5. MTT Assay {#sec2dot5-pharmaceutics-12-00422} -------------- 3-(4,5-Dimethylthiazol-2-yl)-2,5-Diphenyltetrazolium Bromide (MTT) (Carl Roth, Karlsruhe, Germany) assay was performed to evaluate long term cell response after cell electroporation in the presence of CaCl~2~. MTT is used as a colorimetric cell viability assay, which relates the enzymatic activity of the cell with its viability. The colorimetric assay is based on ability of NAD(P)H-dependent cellular oxidoreductase enzymes to reduce the yellow tetrazolium dye MTT to its insoluble purple formazan. After cell treatment with electric pulses in electroporation medium, supplemented with different CaCl~2~ concentrations (0, 0.25, 0.5, 1 mM), 9000 cells in 200 µL of growing media were plated in each well of 96-well microplates (Plastibrand; Wertheim, Germany) and incubated at 37 °C in 5% CO~2~ atmosphere. After 24 h of incubation, 20 µL of growing medium was removed from the wells and 20 µL of MTT salt at concentration of 0.5 mg/mL was added and incubated for additional 2 h. Afterwards, the medium was taken out from the wells and the wells were washed twice with 100 µL of Phosphate-Buffered Saline (PBS) (Carl Roth, Karlsruhe, Germany) Formazan formed in the cells was dissolved by using 100 µL of DMSO (Sigma). A multiwell scanning spectrophotometer (spectro star nano BMG Labtech, Ortenberg, Germany) was used to measure the absorbance of the samples in microplate. Optical density was estimated at 570 nm. All experimental points were normalised to the control (untreated cells). 2.6. Flow Cytometry Assay (FCA) {#sec2dot6-pharmaceutics-12-00422} ------------------------------- The flow cytometer (BD Accuri C6, BD Biosciences, Franklin Lakes, NJ, USA) modality to determine the number of cells in a specific volume was employed in the study to estimate exact number of cells in a specific sample at 15 min time points after the cell treatment with electric pulses in electroporation medium, supplemented with different CaCl~2~ concentrations. A flow cytometer value of obtained cell speed was obtained when 10,000 cells from the sample were measured. A mean of measured cells in control samples was normalised to a 100%. Other samples were normalised according to the control (untreated cells). 2.7. Statistical Analysis {#sec2dot7-pharmaceutics-12-00422} ------------------------- Experiments for each individual experimental point were repeated 3 times on at least two separate days. The results are presented as mean ± standard mean error (SEM). For visualization, 20 cells per field of view were used to calculate the average CTCF. To test significance, one-way ANOVA with a Bonferroni post-hoc test was used for all experiments. The assumption for normality distribution was verified with Shapiro--Wilk normality test by setting the p value for rejection to 0.05. The statistical analysis was performed using Sigma Plot 12.5 software. The percentages were transformed using logit transformation before the ANOVA analysis. 3. Results {#sec3-pharmaceutics-12-00422} ========== The first set of experiments was designed to evaluate whether the addition of CaCl~2~ to the electroporation medium can have any influence on the PI electrotransfer efficiency after microsecond range electroporation. The dependence of PI positive cell percentage on the CaCl~2~ concentration in the electroporation medium is summarized in [Figure 2](#pharmaceutics-12-00422-f002){ref-type="fig"}A, and the total fluorescence of PI positive cells is shown in [Figure 2](#pharmaceutics-12-00422-f002){ref-type="fig"}B. These results show that CaCl~2~ reduce the efficiency of PI electrotransfer. Even the lowest CaCl~2~ concentration used (0.25 mM) has significantly reduced both the amount of PI positive cells and their total fluorescence. For example, using 1600 V/cm electric pulse strength in a medium without added calcium, PI positive cell percentage is around 69%. However, 1600 V/cm electric pulse in the electroporation medium with 0.25 mM CaCl~2~ yields only \~28% of PI positive cells. The increase in CaCl~2~ concentration from 0.25 to 0.5 and 1 mM did not further decrease the amount of PI positive cells or their total fluorescence. The total cell fluorescence depicted in [Figure 2](#pharmaceutics-12-00422-f002){ref-type="fig"}B shows similar trends. Irrespectively of the electric pulse strength, the total fluorescence of PI positive cells that were electroporated with CaCl~2~ (0.25, 0.5, 1 mM) is \~2 times lower than that of the cells electroporated in the control medium (0 mM CaCl~2~). It should also be noted that, after increasing the electric pulse strength from 1600 to 1800 V/cm, the percentage of PI positive cells increases more sharply than the total fluorescence of the cells. Our results indicate that PI entry into the cells electroporated with microsecond electroporation parameters is significantly diminished when CaCl~2~ is added to the electroporation medium prior to the application of electric field. The next set of experiments was performed in order to determine whether the same effect is visible when nanosecond electroporation was used. The obtained results are presented in [Figure 3](#pharmaceutics-12-00422-f003){ref-type="fig"} (A---percentage of PI positive cells, B---total fluorescence of PI positive cells). The effect of the CaCl~2~ concentration on the PI electrotransfer efficiency after nanosecond electroporation can be most clearly seen at 14,000--15,000 V/cm of electric field strength. At these conditions, there are no significant differences in PI positive cell percentage when the cells were electroporated in the media with 0.25 and 0.5 mM CaCl~2~ concentrations. Cells electroporated in either of these media resulted in significantly lower amount of PI positive cells (\~25%) when comparing to the cells electroporated in control medium (0 mM CaCl~2~, \~77%). A further increase in CaCl~2~ concentration to 1 mM leads to the percentage of PI positive cells (\~45%) that is significantly increased from the one observed with 0.25 and 0.5 mM CaCl~2~ concentrations, but significantly lower than the one observed in the control (0 mM CaCl~2~). Therefore, it can be summarized that CaCl~2~ reduces the efficiency of PI electrotransfer using both microsecond ([Figure 2](#pharmaceutics-12-00422-f002){ref-type="fig"}) and nanosecond ([Figure 3](#pharmaceutics-12-00422-f003){ref-type="fig"}) pulses. However, using nanosecond electric pulses, this reduction is lower when a higher concentration of calcium is used, showing the likelihood of a secondary effect of CaCl~2~ on nanosecond electroporation. Interestingly, the effect of CaCl~2~ is the same with all CaCl~2~ concentrations used for microsecond duration pulses and the same with 0.25- and 0.5-mM CaCl~2~ concentrations for nanosecond duration pulses. This shows that the calcium induced reduction in PI electrotransfer efficiency might be based on external CaCl~2~ concentration threshold that is below 0.25 mM. Alternatively, this effect might be dependent on the concentration but has already reached a plateau at 0.25 mM. To elucidate this, we electroporated the cells with 1 × 1400 V/cm strength and 100 µs duration electric pulse or 10 × 1.4 kV/cm strength and 200 ns electric pulses repeated at 1 Hz frequency with an expanded range of CaCl~2~ concentrations (0.0001--1.0 mM). The results of these experiments are presented in [Figure 4](#pharmaceutics-12-00422-f004){ref-type="fig"}. We see a gradual decline in the number of PI positive cells ([Figure 4](#pharmaceutics-12-00422-f004){ref-type="fig"}A) and total fluorescence ([Figure 4](#pharmaceutics-12-00422-f004){ref-type="fig"}B) with microsecond-duration electric pulses as they were well approximated by the exponential function (*R*^2^ = 0.95 and *R*^2^ = 0.86, respectively). However, the approximations failed for the results of PI electrotransfer that were obtained using nanosecond-duration electric pulses ([Figure 4](#pharmaceutics-12-00422-f004){ref-type="fig"}C,D). This indicates gradual continuous PI positive cell dependence on CaCl~2~ concentration for microsecond electroporation and a concentration threshold effect for nanosecond electroporation. The threshold for nanosecond electroporation was determined to be between 0.1 and 0.25 mM CaCl~2~ concentration, as indicated by the results of PI positive cells ([Figure 4](#pharmaceutics-12-00422-f004){ref-type="fig"}C) and total cell fluorescence ([Figure 4](#pharmaceutics-12-00422-f004){ref-type="fig"}D). The next step in our investigations was to assess if CaCl~2~ induced reduction in electrotransfer efficiency is caused by a specific feature of propidium iodide or is it applicable to electrotransfer of other small charged molecules as well. To assess this, we electroporated the cells with 1 × 1400 V/cm strength and a 100-µs duration electric pulse or 10 × 1.4 kV/cm strength and 200-ns electric pulses repeated at 1 Hz of frequency electric pulses in media containing 0, 0.25 or 1 mM CaCl~2~ concentration in the presence of three different small charged molecules: propidium iodide (molecular weight 668.36 g/mol, formal charge +2), YO-PRO-1 (molecular weight 629.32, formal charge +2) and ethidium bromide (EtBr, molecular weight 394.29 g/mol, formal charge +1). The results of these experiments are presented in [Figure 5](#pharmaceutics-12-00422-f005){ref-type="fig"}A (microsecond pulses) and [Figure 5](#pharmaceutics-12-00422-f005){ref-type="fig"}B (nanosecond pulses). It can be seen that, after microsecond duration electric pulses, the electrotransfer efficiency ([Figure 5](#pharmaceutics-12-00422-f005){ref-type="fig"}A) decreases for all three molecules tested. The highest drop in both the electrotransfer efficiency is observed with EtBr. However, a different situation is observed with nanosecond electric pulses ([Figure 5](#pharmaceutics-12-00422-f005){ref-type="fig"}B). It can be seen that while the decrease in electrotransfer efficiency ([Figure 5](#pharmaceutics-12-00422-f005){ref-type="fig"}B) is observed with YO-PRO-1 and PI, the electrotransfer efficiency increases when EtBr is used. In order to obtain a deeper understanding of the phenomenon of the CaCl~2~-mediated inhibition of PI electrotransfer, we visualized the dynamics of PI electrotransfer into the cells after microsecond pulse treatment using electroporation media with different CaCl~2~ concentrations (0, 0.25, 0.5, 1 mM). The results of PI electrotransfer dynamics after electroporation are shown in [Figure 6](#pharmaceutics-12-00422-f006){ref-type="fig"}. These results show that PI electrotransfer into the cells with CaCl~2~ present in electroporation media is disturbed from the first seconds after electroporation when comparing to cells without CaCl~2~ using both 1400 ([Figure 6](#pharmaceutics-12-00422-f006){ref-type="fig"}A,C) and 1800 V/cm ([Figure 6](#pharmaceutics-12-00422-f006){ref-type="fig"}B,D) electric field strengths. It can be seen that PI electrotransfer is significantly lower with all electric parameters and all CaCl~2~ concentrations tested in comparison to PI electrotransfer to cells in media without CaCl~2~. However, using a 1800-V/cm pulse strength, the results show that the medium with 0.25 mM CaCl~2~ yields a significantly higher PI entry in comparison to media with 0.5 or 1 mM CaCl~2~. The representative pictures of cells electroporated at 1400 ([Figure 6](#pharmaceutics-12-00422-f006){ref-type="fig"}C) and 1800 V/cm ([Figure 6](#pharmaceutics-12-00422-f006){ref-type="fig"}D) in the media with different CaCl~2~ concentrations visually illustrate the data presented in [Figure 6](#pharmaceutics-12-00422-f006){ref-type="fig"}A,B, providing visual proof that the fluorescence of the cells decreases with increasing CaCl~2~ concentration. One of the possible ways in which the calcium ions could exert its effect on the electroporation is the increased rate of membrane resealing, leading to diminished molecular transport across the membrane. To test this theory, we decided to investigate the dependence of pore resealing dynamics on the presence of extracellular CaCl~2~ concentration by monitoring the percentage of PI positive cells when PI is added after the application of electric pulses. The results of these experiments are presented in [Figure 7](#pharmaceutics-12-00422-f007){ref-type="fig"}. A single HV pulse with 1400 V/cm ([Figure 7](#pharmaceutics-12-00422-f007){ref-type="fig"}A) or 2800 V/cm ([Figure 7](#pharmaceutics-12-00422-f007){ref-type="fig"}B) strength and 100 µs duration was used. The results after electroporation with 1400 V/cm strength electric pulses showed that PI transfer to the cells is reduced immediately after application of electric fields in the presence of extracellular Ca^2+^ ions. Therefore, it can be assumed that these conditions either cause immediate membrane resealing (pore closure and/or membrane repair) or impede the pore formation. However, this is not the case when electric pulses with double (2800 V/cm) electric field strength are used. In these conditions, a clear dependence of the duration of the membrane resealing on the concentration of extracellular Ca^2+^ is observed. Indeed, it can be seen that 0.25 mM of extracellular CaCl~2~ was enough to significantly diminish the membrane resealing duration. However, it is interesting to note that while we see the calcium-induced decrease in PI electrotransfer, there is a significant difference in the trend of PI electrotransfer between the cells electroporated with micro- and nano-second duration pulses. After a certain threshold in CaCl~2~ concentration, the PI electrotransfer efficiency increases again using nanosecond duration electric pulses, while no such effect is observed using microsecond duration pulses ([Figure 2](#pharmaceutics-12-00422-f002){ref-type="fig"} and [Figure 3](#pharmaceutics-12-00422-f003){ref-type="fig"}). For a better understanding of this inversion, we have utilized the widely accepted model of transmembrane voltage induction by electric field \[[@B22-pharmaceutics-12-00422]\]. The model neglects possible cell deformation during prolonged electric field application \[[@B24-pharmaceutics-12-00422],[@B25-pharmaceutics-12-00422]\]. The results of the modelling ([Figure 8](#pharmaceutics-12-00422-f008){ref-type="fig"}) show that the specific conductivity of the extracellular medium does not significantly change the generated transmembrane potential when microsecond duration pulses were used. However, a dramatic transmembrane potential change is observed when the simulations were performed using nanosecond duration pulses. The modelling results showing the transmembrane potential distribution on the whole surface of the cell are presented in [Figure 8](#pharmaceutics-12-00422-f008){ref-type="fig"}A and the peak transmembrane potential at the electrode-facing poles (cos*θ* = 0) is presented in [Figure 8](#pharmaceutics-12-00422-f008){ref-type="fig"}B (microsecond duration pulses) and [Figure 8](#pharmaceutics-12-00422-f008){ref-type="fig"}C (nanosecond duration pulses). It is clearly observed that, using nanosecond duration pulses, the peak of transmembrane potential increases approximately twice in all parameters used when extracellular the CaCl~2~ concentration changes from 0 to 1 mM. Moreover, one can suggest that the observed effects are related with changes in medium conductivity. To test this possibility, we performed corresponding experiments with medium conductivities ranging from 0.016--0.16 S/m, where CaCl~2~ in the medium was replaced by MgCl~2~. The results showed that medium conductivity, in this range, has no effect on PI electrotransfer efficiency [Figure 9](#pharmaceutics-12-00422-f009){ref-type="fig"}. It is also possible that Ca^2+^ can induce cell death in a very short time range after cell electroporation, and the observed decrease in PI positive cells could be because of the disintegration of those cells. However, cell count by flow cytometry assay (FCA), as described in \[[@B26-pharmaceutics-12-00422]\], shows no cell disintegration in the range of CaCl~2~ concentrations used ([Figure 10](#pharmaceutics-12-00422-f010){ref-type="fig"}). In addition, MTT assay performed 24 h after cell treatment with 1 HV pulse (1400 V/cm, 100 µs) demonstrates no additional cell death at 0.25 and 0.5 mM compared to that at 0 mM CaCl~2~ ([Figure 10](#pharmaceutics-12-00422-f010){ref-type="fig"}). 4. Discussion {#sec4-pharmaceutics-12-00422} ============= In the current research, CHO cells were electro-permeabilized in the presence of different extracellular CaCl~2~ concentration in order to investigate the effect of Ca^2+^ on the electrotransfer efficiency of small charged molecules. The presented results clearly demonstrate that the electrotransfer efficiency of PI is impeded when calcium is present in electroporation medium. This process is dependent on the concentration of calcium ions, although a plateau is apparent at relatively low calcium concentrations. Notably, the phenomenon is observed after electroporation with both microsecond duration and nanosecond duration pulses with some differences. To eliminate possibility that this effect is not specific to PI only, for example, due to a calcium-mediated decrease in PI fluorescence or calcium-mediated inhibition of PI and DNA binding, we performed similar experiment with two other small hydrophilic molecules, namely EtBr and YO-PRO-1. After microsecond electroporation, the reduction in the electrotransfer efficiency is observed with all three different molecules---PI, EtBr and YO-PRO-1---investigated in this study. However, after nanosecond electroporation, both the electrotransfer efficiency and the total fluorescence of the cells markedly decreased for PI and YO-PRO-1 but increased for EtBr. This is likely related to the size and the charge of the molecules. PI and YO-PRO-1 both have similar molecular weights and +2 formal charge. In contrast, EtBr has \~1/3 lower molecular weight and +1 formal charge, which likely changed its behaviour after nanosecond electroporation. In order to understand the observed differences, we performed the simulation of transmembrane voltage generated upon delivery of microsecond and nanosecond duration pulses. Simulation showed that with an increase in CaCl~2~ concentration, a higher transmembrane voltage was induced when nanosecond, but not microsecond, pulses were used. Most probably, this is related with the increase in electroporation medium conductivity (see [Table 1](#pharmaceutics-12-00422-t001){ref-type="table"}) and consequently the decrease in membrane charging time. Indeed, since the cell membrane charging time is below 5 µs \[[@B24-pharmaceutics-12-00422]\], the delivery of ten 200-ns pulses at 1 Hz of frequency might result in incomplete membrane charging. Therefore, although calcium ions decrease the efficiency of small molecule electrotransfer, this decrease can be compensated by increase in transmembrane voltage generated upon the delivery of nanosecond pulses. According to simulation, increases in CaCl~2~ concentration do not have any effect on the generation of transmembrane voltage when microsecond pulses are used. Therefore, the observed effect with microsecond pulses is attributed solely to calcium ions. Calcium ions play multiply roles in cell physiology; however, when analysing the role of CaCl~2~ in the decrease in efficiency of small charged molecules, electrotransfer pore formation and annihilation must be considered. The cell after electroporation faces a critical threat to its viability in the form of uncontrolled transport of essential ions and organic molecules through the electric field affected membrane. Therefore, defense mechanisms that rapidly repair plasma membrane lesions have to be employed in order to maintain cell viability after membrane permeabilization. The plasma membrane repair requires membrane replacements, fusion events and cytoskeleton reorganization \[[@B27-pharmaceutics-12-00422]\]. It can be segmented into passive and active membrane repair, both of which are triggered by Ca^2+^ ion influx at the injury site due to a thousand-fold gradient of calcium that exists across the plasma membrane \[[@B28-pharmaceutics-12-00422]\]. It has been shown that Ca^2+^ ions interact with phospholipid heads, leading to phospholipid bridging via hydrophobic bonding \[[@B29-pharmaceutics-12-00422]\]. The phospholipid bridging initiates the aggregation of the phospholipids, which provides a significantly higher chance for the fusion of the vesicles, thus allowing phospholipids in damaged cell membrane areas to fuse. This way, there is a greater chance for a damaged cell membrane to reseal. Another Ca^2+^-ion-induced phenomenon that was observed in previously published experiments with liposomes is the reduction in the fluidity of the membrane induced by Ca^2+^ interaction with the phospholipid heads \[[@B30-pharmaceutics-12-00422],[@B31-pharmaceutics-12-00422]\]. The membrane fluidity term describes a relative diffusion motion of molecules within membranes \[[@B32-pharmaceutics-12-00422]\]. Some studies had been investigating the relationship between the membrane fluidity and the electroporation \[[@B33-pharmaceutics-12-00422],[@B34-pharmaceutics-12-00422]\]. In these, the membrane fluidity was changed by regulating the cholesterol percentage in the membrane \[[@B35-pharmaceutics-12-00422],[@B36-pharmaceutics-12-00422]\]. The studies have indicated that the electroporation threshold is dependent on the membrane fluidity \[[@B33-pharmaceutics-12-00422],[@B34-pharmaceutics-12-00422]\]. Indeed, the transmembrane potential threshold for electroporation has an inverse correlation with the fluidity of the lipid bilayer \[[@B37-pharmaceutics-12-00422]\]. Therefore, if calcium ions are present in the extracellular medium, then the fluidity of the membrane is decreased, increasing the transmembrane potential threshold required for electroporation in turn. This assumption also goes in agreement with the published simulations addressing the dependence of electropore formation on Ca^2+^ ions \[[@B38-pharmaceutics-12-00422]\]. Here, we show that higher applied electric pulse strength is needed in order to obtain the similar amount of PI positive cells in Ca^2+^-containing medium in comparison to the medium without added Ca^2+^, which is consistent with this hypothesis. Moreover, it has also been demonstrated that phospholipid aggregates induced by Ca^2+^ ion bridging increase the phospholipid aggregate repulsiveness to water molecules \[[@B39-pharmaceutics-12-00422]\]. The change in phospholipid--water interactions becomes relevant when using hydrophilic pore formation model to explain electroporation. According to this model, pores do not immediately close after the electric-field-induced transmembrane potential dissipates \[[@B22-pharmaceutics-12-00422]\]. Instead, after the electric field is switched off, the size of the pores rapidly diminishes to \~0.4--2.3 nm \[[@B22-pharmaceutics-12-00422],[@B40-pharmaceutics-12-00422]\]. Taking the reduced pore size and the model of the phospholipid head lined interior of the pore into the account, the repulsion to water by the Ca^2+^ ion bridging of phospholipid aggregates can make a great impact on the transfer of hydrophilic molecules. Additionally, extracellular calcium, intracellular vesicles and calcium-dependant exocytosis are involved in the active membrane repair \[[@B41-pharmaceutics-12-00422]\]. Specifically, calcium and cytoplasmic vesicles have been identified as a part of repair/resealing machinery \[[@B21-pharmaceutics-12-00422]\]. This repair machinery is initiated via annexins, a family of Ca^2+^ regulated proteins \[[@B42-pharmaceutics-12-00422],[@B43-pharmaceutics-12-00422]\]. However, the exact mechanism of the annexin role in the repair of membrane damage is not yet clearly understood. Nevertheless, it has been shown that annexins 1 and 2 interact with dysferlin, a protein involved in membrane resealing after dramatic damage \[[@B44-pharmaceutics-12-00422],[@B45-pharmaceutics-12-00422]\]. The calcium-triggered assembly of 2D arrays of annexin 5 has been also observed to play a key role in cell membrane repair \[[@B46-pharmaceutics-12-00422]\]. One could also suggest that Ca^2+^ influx can close non-specific ion channels that permit dye uptake. Indeed, it has been shown that Ca^2+^ overload in the cell desensitizes many cation channels \[[@B47-pharmaceutics-12-00422],[@B48-pharmaceutics-12-00422],[@B49-pharmaceutics-12-00422]\]. Nevertheless, these channels do not permit the transmembrane diffusion of dyes like propidium iodide. On the other hand, connexin hemichannels allow for the exchange of small molecules between the cytoplasm and the extracellular space \[[@B50-pharmaceutics-12-00422]\]. This opens the possibility that Ca^2+^ can regulate PI uptake through these hemichannels. Nevertheless, studies on gating of various connexin hemichannels show that these hemichannels open under low extracellular calcium concentrations \[[@B51-pharmaceutics-12-00422],[@B52-pharmaceutics-12-00422]\], i.e., in conditions that are opposite to the ones described above. 5. Conclusions {#sec5-pharmaceutics-12-00422} ============== In conclusion, we report that extracellular calcium induces a negative effect on the small charged molecule electrotransfer into the cells due to electroporation. This effect was demonstrated by using conventional microsecond duration pulse electroporation. All used concentrations of extracellular CaCl~2~ had the negative effect small charged molecule electrotransfer. However, the effect was diminished when CaCl~2~ concentration increased from 0.25 to 1 mM using nanosecond duration pulse electroporation. This can be explained by mathematical modelling, which proves that, with nanosecond electric pulses, the increase in specific conductivity due to the higher concentration of CaCl~2~ results in significantly higher transmembrane potential generated on the cell. However, no change in the transmembrane potential generated on cells in media with different CaCl~2~ concentrations is observed when microsecond electric pulses are used. These results underline the differences between micro- and nano-second pulses used for electrotransfer of small charged molecules in the presence of CaCl~2~. D.N., P.R. and S.S. conceived and designed the experiments; D.N., M.M. and P.R. performed the experiments; D.N., P.R., V.N., R.S., S.S. analysed the data; D.N., P.R., M.J. wrote the paper, S.S. reviewed and edited the paper. All authors have read and agreed to the published version of the manuscript. This research is funded by the European Social Fund according to the activity 'Improvement of researchers' qualification by implementing world-class R&D projects' of Measure No. 09.3.3-LMT-K-712-01-0188. The authors declare no conflict of interest. {#pharmaceutics-12-00422-f001} {#pharmaceutics-12-00422-f002} {#pharmaceutics-12-00422-f003} {#pharmaceutics-12-00422-f004} {#pharmaceutics-12-00422-f005} {#pharmaceutics-12-00422-f006} {#pharmaceutics-12-00422-f007} {#pharmaceutics-12-00422-f008} {#pharmaceutics-12-00422-f009} {#pharmaceutics-12-00422-f010} pharmaceutics-12-00422-t001_Table 1 ###### The measured specific conductivity of used electroporation media containing different concentrations of CaCl~2~. Measurements were done using conductometer (Mettler Toledo S230). CaCl~2~ Concentration in the Medium Specific Conductivity ------------------------------------- ----------------------- 0 mM 0.016 S/m 0.25 mM 0.021 S/m 0.5 mM 0.026 S/m 1 mM 0.035 S/m [^1]: These authors contributed equally to this work.

Background {#Sec1} ========== Adolescent idiopathic scoliosis (AIS) is a complex three-dimensional (3-D) deformity of the spine, with a prevalence of 1.5--3% within the general population, that normally develops in the beginning of the growth spurt of previously healthy adolescents \[[@CR1], [@CR2]\]. For diagnosis, monitoring of progression, and clinical decision-making, periodical radiographic follow-up is traditionally performed using posterior-anterior and lateral upright radiographs. The Scoliosis Research Society defines scoliosis as a lateral curvature of the spine of more than 10° in the coronal plane on upright radiographs, also emphasizing the importance of radiography \[[@CR3]\]. In addition, supine or prone magnetic resonance imaging (MRI) and computed tomography (CT) are frequently used to obtain more in-depth information about neuroaxis and bony architecture abnormalities. Some imaging involves ionizing radiation, and the radiation doses are cumulative, resulting in 9 to 10 times more radiation exposure and a 17 times higher incidence of cancer in the AIS cohort as compared to the general population \[[@CR4], [@CR5]\]. The importance of the 3-D character of the scoliotic deformity has long been recognized, and the upright X-ray, the gold standard, is not able to accurately represent the true 3-D deformity \[[@CR6]--[@CR9]\]. CT scanning can obtain accurate 3-D information of bony structures but relies on radiation and is not obtained upright \[[@CR10]\]. An important step in attempts to visualize this 3-D character has been the development of low-dose upright imaging modalities that allow for 3-D reconstruction such as the EOS apparatus. Alternatively, MRI utilizes no harmful radiation but is considered inferior in visualizing the bone and is usually also not obtained upright. This study was designed to compare the morphology of the scoliotic spine on conventional radiographs in the upright position to those on MRI and CT obtained in supine and prone positions, respectively. Methods {#Sec2} ======= Study population {#Sec3} ---------------- A subsequent series of AIS patients of ten or more years of age scheduled for scoliosis surgery in one of our centers between 2011 and 2014 and had complete standard pre-operative work-up were included in this study. Complete work-up consisted of posterior-anterior and lateral upright radiographs of the spine, supine bending X-rays, T2-weighted MRI (3.0-T MR scanner (Achieva TX; Philips Healthcare, Best, The Netherlands)) of the spinal cord for exclusion of neural axis abnormalities obtained in a supine position, and high-resolution CT (64 Slice Multi-detector CT scanner, GE Healthcare, Chalfont, St. Giles, UK, slice thickness 0.625 mm), obtained in a prone position. The CT scans were made for navigation purposes according to protocol in one of our institutions, in a position mimicking the position at surgery as closely as possible. Children with other spinal pathology than AIS, early onset scoliosis, previous spinal surgery, neurological symptoms or neural axis abnormalities, syndromes associated with disorders of growth, or atypical left convex thoracic curves or right convex (thoraco)lumbar curves were excluded to obtain an as homogeneous a population as possible. Moreover, cases that had undergone the different imaging methods with an interval of more than 6 months in between imaging were also excluded. Curve characteristics (curve type according to the Lenke classification, Cobb end vertebrae, and apical levels) were determined on the conventional radiographs \[[@CR11], [@CR12]\]. Outcome parameters {#Sec4} ------------------ The conventional radiographs were analyzed for main thoracic and (thoraco)lumbar Cobb angle, apical rotation (using Perdriolle's method \[[@CR13]\]), thoracic kyphosis (TK; superior endplate T4--inferior endplate T12), and lumbar lordosis (LL; superior endplate L1--sacral plate), using our picture archiving and communications system (PACS) workstation (Carestream solution working station, Carestream Health, Version 11.0, Rochester, NY, USA). On the MRI and CT images, the main thoracic and (thoraco)lumbar Cobb angle, TK, and LL were measured using the same technique as for the conventional radiographs, by using multiplanar reconstruction technique through the midsection of each vertebral body for the MRI and the digital reconstructed radiograph (DRR) for the CT scan (Fig. [1](#Fig1){ref-type="fig"}). The same levels were used for each patient on the three different imaging methods. Cobb end vertebrae were selected on the radiographs and applied to the other imaging modalities \[[@CR14]\]. For measurement of apical rotation on the MRI and CT scans, complete 3-D reconstructions were acquired using semi-automatic analysis software (ScoliosisAnalysis 4.1, Imaging Division, Utrecht, The Netherlands) and a previously validated imaging method \[[@CR15]\]. The observer selected the upper and lower endplates of the vertebral body. Then, the observer used the sagittal and coronal orientation of the endplates to correct for coronal and sagittal tilt. Thus, each vertebral level was manually positioned in the true transverse plane as accurately as possible. Subsequently, for each endplate, its longitudinal axis was calculated automatically after manual segmentation of the vertebral body and spinal canal. The rotation was defined as the rotation of this axis minus the rotation of the neutral sacral plate (Fig. [2](#Fig2){ref-type="fig"}).Fig. 1On the MRI and CT images, the main thoracic and (thoraco)lumbar Cobb angle, thoracic kyphosis, and lumbar lordosis were measured using the same technique as for the conventional radiographs on the image where the curve and endplates were best visible by using the multiplanar reconstruction (MPR, **a**) for the MRI and the digitally reconstructed radiograph (**b**) for the CT scan. **c** The conventional X-ray Fig. 2The orientation of the upper and lower endplates of each individual vertebra of the computed tomography scans was determined by using the semi-automatic software, correcting for coronal and sagittal (**a** and **b**) tilt, to reconstruct the true transverse sections. The observer drew a contour around the vertebral body (*yellow line* in **c**) and spinal canal (*blue line* in **c**). The software calculated a center of gravity of the vertebral body (*yellow dot* in **c**) and spinal canal (*blue dot* in **c**). For each endplate, its longitudinal axis was calculated as the line between those two points (*purple line* in **c**). The rotation of this axis minus the rotation of the neutral sacral plate represents the rotation of the endplate Intra- and interobserver reliability for measurement of apical rotation using this method was tested in a previous study; intraclass correlation coefficients were 0.92 (95% confidence interval, 0.82--0.97) and 0.89 (0.74--0.95) on the 3-D scans \[[@CR9]\]. In this study, the intra- and interobserver reliability analysis of the rest of the outcome parameters (Cobb angles, TK, and LL on all the three modalities and the vertebral rotation on the X-rays) was studied. Two observers independently analyzed a randomly selected subset of ten X-rays, CT scans, and MRI scans of the subjects. Statistical analysis {#Sec5} -------------------- Statistical analyses were performed using SPSS 22.0 for Windows (SPSS Inc., Chicago, IL, USA). Descriptive statistics were computed providing means, ranges, and standard deviations. Potential outliers were identified. The agreement between the three positions was tested according to the Bland-Altman plot; first, the one-sample *t* test showed if there was a significant difference between the measurements; second, if there was no significant difference, the regression analysis showed if there was agreement between the measurements \[[@CR16]\]. The two-way mixed intraclass correlation coefficient (ICC) was used to evaluate the correlation between the parameters in different body positions. The intra- and interobserver reliability were obtained as intraclass correlation coefficients. The statistical significance level was set at 0.05 for all analyses. Results {#Sec6} ======= Population {#Sec7} ---------- A total of 142 subjects underwent surgery for AIS during the study period. Eighty subjects had to be excluded for several reasons, as shown in Table [1](#Tab1){ref-type="table"}. Ultimately, 62 AIS patients with full documentation were left for the purpose of this study. On average, the subjects were 15.6 ± 2.5 years of age, 56 (90%) were girls, and most of the curves were classified as type Lenke 1 of these moderate to severe AIS patients (thoracic Cobb angle 37°--110°, lumbar Cobb angle 18°--82°; Table [1](#Tab1){ref-type="table"}).Table 1Demographics are shown for all included AIS patients and controls. Also, the excluded patients are shownDemographic parameter*n* = 62 Age at radiograph (years)Range10--23Mean ± sd15.6 ± 2.5 Girls, *n* (%)56 (90.3%) Right convexity of main thoracic curve, *n* (%)Right convex62 (100%) Interval CT--radiograph (days)Range−7 to 130Mean ± sd2.98 ± 17.2 Interval radiograph--MRI (days)Range−46 to 181Mean ± sd81.3 ± 51.4 Interval CT--MRI (days)Range−26 to 181Mean ± sd84.2 ± 47.1Lenke curve type I26 II12 III6 IV4 V5 VI9Exclusion criterian Scan interval \>6 months38 No MRI available14 No CT scan available10 Incomplete radiologic work-up1 Associated congenital or neuromuscular pathologies12 Left convex main thoracic curve4 Prior spinal surgery1*sd* standard deviation Coronal parameters {#Sec8} ------------------ In the coronal plane, the main thoracic Cobb angle was on average 68° ± 15°, 54° ± 15°, and 57° ± 14° on the upright radiographs, prone CT, and supine MRI, respectively, and differed significantly between all the three positions (*P* \< 0.001; Table [2](#Tab2){ref-type="table"}). The average (thoraco)lumbar Cobb angle on the conventional upright radiograph was 44° ± 17° as compared to those on the prone CT (33° ± 15°) and supine MRI (35° ± 16°) (*P* ≤ 0.018, between the three positions). Although the upright angles were larger, the Cobb angles correlated very well between the three positions (ICC: thoracic 0.97 and lumbar 0.96; Table [3](#Tab3){ref-type="table"}; Fig. [3](#Fig3){ref-type="fig"}). Significant linear correlations were found, indicating that with increasing Cobb angle, differences between the body positions increased simultaneously. The conversion equations that resulted from the correlation analyses of the different parameters between the upright X-ray, prone CT scan, and supine MRI could be used for conversion purposes (Table [4](#Tab4){ref-type="table"}).Table 2Differences (mean ± standard deviation) between upright (X), prone (CT), and supine (MRI) positions for Cobb angle, thoracic kyphosis, lumbar lordosis, and apical vertebral rotation in the thoracic as well as lumbar curves. According to the Bland-Altman plot, the *P* value showed if there is agreement by using the *t* test. If this test showed no significant different (*P* \> 0.05), a regression analysis was performed to see is if there is agreement, written in bracketsUprightProneSupine*P* valueX vs. CTX vs. MRICT vs. MRIThoracic Cobb (°)68.2 ± 15.453.9 ± 14.856.7 ± 13.5\<0.001\<0.001\<0.001 Kyphosis (°)25.8 ± 11.422.4 ± 11.617.3 ± 9.80.004\<0.001\<0.001 Vertebral rotation (°)21.6 ± 11.719.9 ± 8.916.3 ± 10.80.161 (0.007)0.0010.002Lumbar Cobb (°)44.3 ± 16.833.1 ± 15.035.2 ± 15.9\<0.001\<0.0010.018 Lordosis (°)48.8 ± 12.045.4 ± 10.843.7 ± 12.40.006\<0.0010.341 (0.620)^a^ Vertebral rotation (°)10.7 ± 12.87.5 ± 11.46.2 ± 13.70.428 (\<0.001)0.663 (0.129)^a^0.679 (0.006)^a^Agreement according to the Bland-Altman plot Table 3Two-way mixed intraclass correlation coefficient (ICC) and 95% confidence interval (CI) between upright, prone, and supine positionsICC (95% CI)*P* valueThoracic Cobb angle0.967 (0.950--0.979)\<0.001Lumbar Cobb angle0.964 (0.945--0.977)\<0.001Thoracic kyphosis0.873 (0.806--0.919)\<0.001Lumbar lordosis0.854 (0.777--0.907)\<0.001Thoracic apical rotation0.815 (0.718--0.882)\<0.001Lumbar apical rotation0.900 (0.848--0.937)\<0.001 Fig. 3In these scatterplots, the relation between thoracic Cobb angle in the upright, prone (*red trend line*), and supine (*blue trend line*) positions is shown. Although the upright Cobb angle was significantly larger, significant linear correlations were found (ICC 0.967; *P* \< 0.001), indicating that with increasing Cobb angle, differences between the body positions increased simultaneously Table 4For translational purposes, the conversion equations that resulted from the linear correlation analyses of the different parameters between the upright X-ray, prone CT scan, and supine MRI are provided for the thoracic (Th) and lumbar (L) Cobb anglesCobb angleUpright X-rayProne CT scanSupine MRICobb angleUpright X-ray--Th: CT (°) = −6.2 + 0.88 \* X-ray (°)\ L: CT (°) = −2.7 + 0.81 \* X-ray (°)Th: MRI (°) = 2.9 + 0.79 \* X-ray (°)\ L: MRI (°) = −2.1 + 0.85 \* X-ray (°)Prone CTTh: X-ray (°) = 16.6 + 0.96 \* CT (°)\ L: X-ray (°) = 11.1 + 1.00 \* CT (°)--Th: MRI (°) = 11.0 + 0.85 \* CT (°)\ L: MRI (°) = 4.9 + 0.92 \* CT (°)Supine MRITh: X-ray (°) = 10.8 + 1.01 \* MRI (°)\ L: X-ray (°) = 9.5 + 0.98 \* MRI (°)Th: CT (°) = −2.8 + 1.00 \* MRI (°)\ L: CT (°) = 2.6 + 0.86 \* MRI (°)-- Axial rotation {#Sec9} -------------- Parallel to the coronal Cobb angles, in both the thoracic curve and the (thoraco)lumbar curve, the mean apical vertebral rotation was larger in the upright position (Table [2](#Tab2){ref-type="table"}). Significant correlations, however, were observed between the apical rotation as measured using the Perdriolle method on upright radiographs and the rotation on the prone CT and supine MRI (ICC: thoracic 0.82 and lumbar 0.90; Tables [3](#Tab3){ref-type="table"} and [4](#Tab4){ref-type="table"}). Sagittal parameters {#Sec10} ------------------- Also in the sagittal plane, the TK in the upright position (26° ± 11°) was significantly larger as compared to that in the prone (22° ± 12°) and supine (17° ± 10°; *P* ≤ 0.004) positions. The upright LL (49° ± 12°) was significantly higher as compared to the prone LL (45° ± 11°) and supine LL (44° ± 12°; *P ≤* 0.006). According to the Bland-Altman method, there was agreement between the LL in the supine and prone positions. The TK and the LL correlated well between all the positions (ICC 0.87 and 0.85; Tables [3](#Tab3){ref-type="table"} and [4](#Tab4){ref-type="table"}). Reliability {#Sec11} ----------- The ICCs for intra- and interobserver reliabilities of the Cobb angles, TK, LL, and vertebral rotation on the three modalities were all excellent (\>0.93 and \>0.74, respectively; Table [5](#Tab5){ref-type="table"}).Table 5Intra- and interobserver reliability analysis and 95% confidence intervalX-rayCT scanMRI scanIntraInterIntraInterIntraInterThoracic Cobb0.993 (0.971--0.998)0.972 (0.888--0.993)0.997 (0.988--0.999)0.995 (0.980--0.999)0.995 (0.982--0.999)0.974 (0.896--0.994)Lumbar Cobb0.999 (0.996--1.00)0.995 (0.980--0.999)0.999 (0.996--1.00)0.995 (0.981--0.999)0.997 (0.990--0.999)0.986 (0.945--0.997)Thoracic kyphosis0.989 (0.954--0.997)0.922 (0.610--0.984)0.931 (0.722--0.983)0.864 (0.454--0.966)0.992 (0.967--0.998)0.940 (0.759--0.985)Lumbar lordosis0.986 (0.944--0.997)0.989 (0.956--0.997)0.995 (0.980--0.999)0.973 (0.890--0.993)0.995 (0.981--0.999)0.971 (0.884--0.993)Thoracic rotation0.979 (0.915--0.995)0.977 (0.906--0.994)^aa^0.939 (0.756--0.985)0.744 (0.409--0.964)Lumbar rotation0.975 (0.899--0.994)0.996 (0.985--0.999)^aa^0.906 (0.620--0.977)0.885 (0.539--0.972)^a^Intra- and interobserver reliability for the rotation on 3-D scans; this method was tested previously (ICC 0.92 and 0.89) \[[@CR9]\] Discussion {#Sec12} ========== X-rays for scoliosis are, by convention, obtained in an upright position, allowing gravity to have its influence on the morphology of the spine. The drawbacks of this X-ray imaging in analyzing the deformity as well as planning treatment are becoming increasingly clear: the deformity has a complex 3-D nature that is hardly appreciated on plain films, and radiation exposure, even with modern day equipment, is becoming a serious concern. Although the use of ultrasound for diagnosis and follow-up of spinal deformities has been explored and seems promising, this technique gives little detail of the anatomy and needs further evaluation \[[@CR17]--[@CR19]\]. Additional imaging studies are frequently obtained in scoliosis; CT scanning is still considered the gold standard for providing accurate and detailed information on bony anatomy (for instance, in cases where congenital malformations are suspected) and can give accurate 3-D reconstructions of complex deformities \[[@CR10]\]. However, CT carries even more radiation exposure and is performed non-weight bearing \[[@CR10]\]. MRI is safe, provides accurate information on the spinal cord and other soft tissues, but is also (usually) performed in a non-weight-bearing manner, and is known to show less detail of bony structures. Therefore, it is important to define where these techniques overlap, in order to reduce costs and radiation exposure. Previous studies have already described the differences in morphology of the spine in AIS between different imaging methods and between different body positions \[[@CR20]--[@CR26]\]. This study is, however, to the best of our knowledge, the first to look into the relationship between the three different positions in all three planes of the body to visualize the scoliotic spine. In this study, we observed that there is underestimation of the deformation of the spine in the supine and prone positions as compared to that in the upright position, which is overall more pronounced in the thoracic curves as compared to the (thoraco)lumbar curves. The lying positions underestimated the thoracic and (thoraco)lumbar Cobb angles for 12°--14° and 9°--11°, respectively; the TK and LL for 3°--9° and 3°--5°, respectively; and the thoracic and lumbar apical vertebral rotations for 2°--5° and 3°--5°, respectively. Therefore, the parameters on supine and prone scans could not directly be compared to the upright radiographs. However, good and excellent linear correlations were observed for the morphological parameters in the coronal (ICC ≥0.964), sagittal (ICC ≥0.854), and axial (ICC ≥0.815) planes between X-ray, CT, and MRI. This implies that reliable conversion of the parameters between the different positions is possible. A limitation of this study is the population that only includes relatively severe curves. From our results, the reliability of conversion of parameters between different positions for patients with mild AIS curves cannot be derived. Shi et al. described the correlation of the coronal Cobb angle between upright and supine positions in mild, moderate, and severe AIS patients and concluded that the correlation coefficients were more reliable in the severe group, probably due to the reduced curve flexibility in the severe group \[[@CR26], [@CR27]\]. As we demonstrated before, evaluation of the true sagittal plane in scoliosis on plain X-rays is notoriously unreliable and differs greatly from the true sagittal plane as may be analyzed more accurately on both CT and MRI \[[@CR28]\]. Conclusions {#Sec13} =========== There is a good to excellent correlation of the morphology of the scoliotic spine in all three planes between standard upright X-ray, MRI, and CT scan in these moderate to severe AIS patients. Apparently, at least part of the information obtained by these different modalities overlaps. Findings of this study suggest that severity of scoliotic deformity in AIS patients can be largely represented by different imaging modalities despite the differences in body position. Future longitudinal studies to demonstrate the practical implications of these findings are planned. 3-D : Three-dimensional AIS : Adolescent idiopathic scoliosis CI : Confidence interval CT : Computed tomography DRR : Digital reconstructed radiograph ICC : Intraclass correlation coefficient LL : Lumbar lordosis MRI : Magnetic resonance imaging PACS : Picture archiving and communications system sd : Standard deviation TK : Thoracic kyphosis None. Funding {#FPar1} ======= Rob C. Brink received funding from the Alexandre Suerman, MD/Ph.D. program, and René M. Castelein from a Medtronic research grant and a K2M research grant. Availability of data and materials {#FPar2} ================================== The data are shared in the "[Results](#Sec6){ref-type="sec"}" section. The X-rays and scans of the subjects will not be shared. Software: ScoliosisAnalysis 4.1 for Windows (Imaging Division, Utrecht, The Netherlands) SPSS 22.0 for Windows (SPSS Inc., Chicago, IL, USA). Authors' contributions {#FPar3} ====================== RCB handled the conception and design, acquisition of the data, analysis and interpretation of the data, drafting of the manuscript, statistical analysis, and obtaining funding. DC handled the acquisition of the data, analysis and interpretation of the data, drafting of the manuscript, and supervision. TPCS handled the conception and design, acquisition of the data, analysis and interpretation of the data, critical revision of the manuscript for important intellectual content, statistical analysis, and supervision. KLV handled the acquisition of the data, analysis and interpretation of the data, and technical and material support. MvS handled the acquisition of the data, analysis and interpretation of the data, and technical and material support. SCNH handled the acquisition of the data and technical and material support. SL handled the acquisition of the data, analysis and interpretation of the data, critical revision of the manuscript for important intellectual content, and technical and material support. WCWC handled the Acquisition of the data, analysis and interpretation of the data, critical revision of the manuscript for important intellectual content, and supervision. JCY handled the acquisition of the data, analysis and interpretation of the data, and critical revision of the manuscript for important intellectual content and supervision. RMC handled the conception and design, analysis and interpretation of the data, critical revision of the manuscript for important intellectual content, obtaining funding, and supervision. All authors read and approved the final manuscript. Competing interests {#FPar4} =================== The authors declare that they have no competing interests. Consent for publication {#FPar5} ======================= All authors have given permission for publication. Consent for publication of the subject not applicable. Ethics approval and consent to participate {#FPar6} ========================================== The Medical Ethics Review Committee confirmed that the Medical Involving Human Subjects Act (WMO) does not apply to the abovementioned study and that therefore an official approval of this study by MERC UMC Utrecht is not required under the WMO.

Introduction ============ Corneal diseases represent the second leading cause of blindness, affecting 4.9 million people worldwide; these individuals could potentially have their sight restored through corneal transplantation \[[@r1],[@r2]\]. Penetrating keratoplasty is the standard procedure used for the treatment of corneal blindness. However, this procedure faces two primary problems: a shortage of graft donors and a decrease in endothelial cell density within 5 years of transplantation \[[@r3]\]. The corneal endothelium (CE) is responsible for maintaining corneal hydration through a pump--leak mechanism \[[@r4]\]. Although CE cells (CECs) are normally arrested in the early G1 phase of the cell cycle, they retain their proliferative capacity \[[@r5]\]. Tissue engineering can take advantage of this capacity to address the lack of available donor tissue. To accomplish this aim, a robust system for the isolation and propagation of CECs is needed. Several studies exploring complex culture media have reported the increased proliferative capacity of CECs \[[@r6]-[@r10]\]. The addition of growth factors to culture media enhances CEC proliferation; however, this effect is associated with changes in cell morphology (from hexagonal to fibroblastic) and alterations in the expression of characteristic molecular markers, which raises questions concerning the CECs' identity \[[@r6],[@r8],[@r11]-[@r13]\]. The use of culture media without growth factors is able to maintain the hexagonal morphology of the CECs; however, it yields low proliferation rates that cannot be propagated beyond the first passage \[[@r10],[@r14]\]. In this study, with the aim of improving the identity of CECs after proliferation, we first used a widely used supplemented culture medium to proliferate CECs \[[@r9]\], which was then followed by a resting step that incorporated basal medium to provide evidence of the development of a convenient CEC expansion strategy. We compared the morphology and transcriptome of CECs in two conditions and validated CEC markers using immunohistochemistry and quantitative PCR. The results suggest that the resting step helps maintain the identity of cultured CECs. Methods ======= This study was approved by the institutional local ethics committee (School of Medicine of Tecnologico de Monterrey), number 2013-Re-002. All animals were treated according to the Guide for the Care and Use of Laboratory Animals adhering to the guidelines for the human treatment and ethical use of animals for vision research stated by the Association for Research in Vision and Ophthalmology. Corneal endothelial tissue isolation ------------------------------------ Eight corneas were obtained from four 3-month-old New Zealand rabbits weighing about 3 kg. The rabbits were euthanized under general anesthesia with 30 mg/kg of ketamine (Pisa Farmaceutica, Guadalajara, México), followed by a lethal intracardiac injection of sodic pentobarbital (Pets Pharma, Estado de Mexico, Mexico). The corneas were excised, rinsed with OptiMEM-I (Gibco®; Thermo Fisher Scientific, Waltham, MA) supplemented with 8% fetal bovine serum (FBS; Cellgro, Manassas, VA) and 1% streptomycin/penicillin antibiotics (Thermo Fisher Scientific), and placed in a sterile tissue culture dish. Two sets of rabbits were used at different times. The first set of two rabbits was used for culture followed by transcriptome analyses while the second was used for culture followed by validation (immunocytochemistry and quantitative PCR \[qPCR\]). Isolation of CECs ----------------- CECs were isolated using the "peel-and-digest" approach. Briefly, using sterile surgical forceps, Descemet's membrane (DM) with the intact endothelium (DM/CE) was carefully dissected from the corneal stroma, and then was washed several times with Dulbecco\'s Modified Eagle Medium: Nutrient Mixture F-12 (DMEM-F12, Gibco®; Thermo Fisher Scientific, Grand Island, NY) supplemented with 10% FBS and 1% antibiotic combination. DM/CE complexes were incubated in OptiMEM-I 8% FBS and the 1% antibiotic combination overnight to stabilize the cells before culture. They were then incubated with 2 mg/ml of collagenase type I (Sigma-Aldrich Co., St. Louis, MO) at 37 °C for 1 h to release the CECs from DM. CEC clusters were treated with trypsin/EDTA (0.25% trypsin/0.53 mM EDTA; Sigma-Aldrich Co.) for 10 min to dissociate aggregates into smaller cell clumps. They were collected following centrifugation at 375 ×g for 10 min. CEC culture ----------- CECs were first cultured in a previously reported medium (MitoM) \[[@r9]\] containing OptiMEM-I supplemented with 8% FBS, 20 ng/ml of nerve growth factor (NGF; Sigma-Aldrich Co.), 5 ng/ml of epidermal growth factor (EGF; Sigma-Aldrich Co.), 100 µg/ml of pituitary extract (Sigma-Aldrich Co.), 200 µg/l of calcium chloride (Sigma-Aldrich Co.), 20 µg/ml of ascorbic acid (Sigma-Aldrich Co.), 0.08% chondroitin sulfate (Sigma-Aldrich Co.), and antibiotics. Isolated cells were incubated at 37 °C in a 5% CO~2~ humidified atmosphere. The medium was changed every third day until 80% confluence. At passage 1, the CECs were subcultured at a 1:2 split ratio. Population 1 continued to be cultured in MitoM, while population 2 was cultured in OptiMEM-I supplemented with only 8% FBS and 1% antibiotics (RestM) up to 80% confluence for an additional passage. We refer to the RestM procedure as "resting" because it lacks growth factors to decrease proliferation rates. An Axiovert 40 CFL contrast microscope (CFL; Carl Zeiss AG, Oberkochen, Germany) featuring a PowerShot A640 digital camera (Canon Inc., Tokyo, Japan) was used to register cell morphology. Morphology analysis ------------------- NIH [Image J](https://imagej.net/) software \[[@r15]\] was used to analyze the morphology of human CECs from three healthy biomicroscopies from a public database \[[@r16]\], basal rabbit CECs (before culture), in MitoM and RestM. A similar scale was set on each photograph. Then about 40 CECs were delimited and analyzed for human MitoM and RestM, whereas about 20 cells were analyzed for basal rabbit CECs with free shape region of interest. Area and perimeter were obtained using Image J (Analysis menu, Measure tool). Circularity was calculated as 4π (area/perimeter\^2), and values near from 1 were taken as high circularity indices, thus near hexagonality \[[@r17]\]. Cellular yield analysis ----------------------- Cellular yield was calculated for CECs cultured in MitoM and RestM. For this analysis, the quotient of cellular concentration (cells/ml) at the end of passage 2 divided by the cellular concentration at the end of passage 1 was calculated for each culture condition (MitoM and RestM). The average and the standard error were calculated. A *t* test was used to analyze statistically significant differences between the calculated yields. RNA isolation ------------- Total RNA was extracted with the RNeasy mini Kit (Qiagen, Hilden, Germany) from CECs before culture, and then after culture in MitoM and RestM. Cells were harvested at 80--90% confluence around day 9 of culture. RNA concentration and purity were determined by spectrophotometry using a NanoDrop ND-1000 UV-VIS spectrophotometer (Waltham, MA); only RNA samples with an A260/A280 ratio ≥1.8 were used for further experiments. The Experion RNA HighSense (Hercules, CA) was used to determine the concentration and integrity of mRNA. Yields were 62--182 ng/µl per confluent dish, and a total of 3 µg/µl was used for sample preparation. Microarray hybridization ------------------------ RNA preparation, labeling, and hybridization were performed according to the manufacturer's recommendations (Agilent Technologies, Santa Clara, CA) using the customized rabbit microarray (G2519F) containing around 44,000 probes. Briefly, cyanine-3- (Cy3-) and cyanine-5 (Cy5)-labeled cRNA were prepared from the total RNA using a labeling kit (Quick Amp; Agilent Technologies). This was followed by column purification (RNeasy Mini Kit; Qiagen). Cy3 was used for the CECs in MitoM, and Cy5 was used for the CECs in RestM. Dye incorporation and cRNA yield were assessed with spectrophotometry (ND-1000; Thermo Fisher Scientific). Equal amounts of the Cy3- or Cy5-labeled cRNA mixture were hybridized to the microarrays (Rabbit Gene Expression Microarrays; Agilent) for 17 h at 65 °C in a rotating hybridization oven (Agilent), followed by washing and scanning. Data were obtained immediately after washing on a microarray scanner (GenePix 4000B; Molecular Devices LLC, Sunnyvale, CA). Microarray data analysis ------------------------ The R statistical environment was used to process and analyze the data (<https://cran.r-project.org/>). Raw data were transformed using log~2~ and subsequently, were quantile-normalized before statistical analysis. The differentially expressed genes were obtained with a paired *t* test that was conducted between the four replicates of MitoM and RestM. A functional analysis of the differentially expressed genes was performed using the Database for Annotation, Visualization, and Integrated Discovery ([DAVID](https://david.ncifcrf.gov)) \[[@r18]\]. Only those probes with human annotations were used for this process. Each microarray probe was aligned to the human transcriptome annotations (hg19) using BLAST (included in Appendix 1). Only those DAVID terms where the p value was less than 0.01, the adjusted false discovery rate p value was less than 0.25, and the gene count was greater than three were examined. Quantitative PCR ---------------- Validation of expression levels for different corneal endothelial markers was performed by quantitative PCR (qPCR) using (m)RNA from CECs before culture, after cultured in MitoM, and RestM. Primers design was conducted in Blast Primer platform (NCBI) and synthetized by T4 oligo company (Guanajuato, Mexico). GAPDH F: CGA GCT GAA CGG GAA ACT CA, R: CCC AGC ATC GAA GGT AGA GG; ATP1A1 F:GAT CCA CGA AGC TGA CAC GA, R: CTG TTA CAG AGG CCT GCG AT; GPC4 F: CGC CAA ATC ATG GCT CTT CG, R: GGC ACT GCT GGT ACT CAC AT; BTG2 F: GGC TTA AGG TTT TCA GCG GG, R: CTT GTG GTT GAT GCG GAT GC; TJP1 F: CTC AAG TTC CTG AAG CCC GT, R: TAG GAT CAC CCG ACG AGG AG. Amplification was performed with the PowerUp SYBR Green Master Mix kit in a Step One 48-well thermocycler (Applied Biosystems, Foster City, CA), under these conditions: initial denaturing at 95 ºC 1 min, 40 cycles: 95 ºC/30 s, 61 ºC/30 s, 72 ºC/30 s; final extension 72 ºC/5 min. Finally, ΔCt method was used to analyze expression levels. Immunocytochemistry ------------------- Immunocytochemistry was performed in CECs before culture, after culture in MitoM, and in RestM to analyze the presence of GPC4 (Abcam, ab150517, Cambridge, UK), CD166 (Abcam, ab78649), ZO-1/TJP1 (ThermoFisher, 61--7300, Waltham, MA), and Na/K-ATPase (Abcam, ab176163, Cambridge, UK). Immunocytochemistry consisted of overnight cell stabilization over coverslips with poly-D lysine (Sigma-Aldrich, P7280), fixation with 4% paraformaldehyde, nonspecific bonding blockage with 5% bovine serum albumin (BSA; Sigma-Aldrich, A-7030), overnight 4 °C incubation with primary antibodies (GPC4 5 µg/ml, CD166 1 µg/ml, ZO-1 5 µg/ml, and Na/K-ATPase 1:100), and incubation with Alexa Fluor 488 secondary antibody (Abcam, ab150077) for 1 h at room temperature. Fluoroshield Mounting Medium with 4′,6-Diamidine-2′-phenylindole dihydrochloride (DAPI; Abcam, ab104139) counterstain was used. For the complete corneas, we followed a previously described protocol for immunostaining on a flat-mounted whole intact cornea \[[@r19]\]. Briefly, corneas were rinsed in Phosphate-Buffered Saline (PBS 1X; 140 mM NaCl, 3 mM KCl, 10 mM NaPO~4~, pH 7.4 at 25 °C), cut into four pie-shaped wedges and immediately fixed. Fixation occurred for 30 min in 0.5% paraformaldehyde (Sigma-Aldrich Co.) in PBS pH 7.45 at 4 °C. Then, cell membranes were permeabilized with 1% Triton X-100 (Sigma-Aldrich Co.) in PBS for 5 min at room temperature. Blockage of non-specific binding sites was performed by incubation for 30 min at 37 °C with 5% BSA (Sigma-Aldrich Co). Primary and secondary antibodies, as well as counterstaining, were used in the same fashion as for CECs immunocytochemistry but, corneal pieces were fully immersed in the corresponding solutions. Epifluorescence was registered with a widefield fluorescence microscope (Zeiss Imager Z1) with an AxioCam HRm (Zeiss) camera (Göttingen, Germany). Results ======= Isolation of corneal endothelial tissue --------------------------------------- The peel-and-digest approach for the isolation of CECs yielded small groups of cells that showed a polygonal morphology, and they were cultured in MitoM (Appendix 1). This evidence also support that the protocol was successfully implemented, and that CECs can be used for further experiments. Effects of culture conditions on the morphology of CECs ------------------------------------------------------- The cultured CECs showed variations in morphology throughout the 5 days in MitoM at P0 ([Figure 1A--E](#f1){ref-type="fig"}). The morphological changes in the CECs in MitoM and RestM culture medium started at around day 2 and became more evident after 5 days of incubation ([Figure1B,C](#f1){ref-type="fig"}). The passaged cells cultured in MitoM acquired a fibroblastic morphology, whereas those in RestM were far less elongated and had formed a monolayer. At day 9, the effect of the MitoM became more evident, particularly when the CECs were compared to those at P0 ([Figure 1D](#f1){ref-type="fig"}). Conversely, the RestM CECs became polygonal ([Figure 1C,E](#f1){ref-type="fig"}). The average of the circularity index of the specular microscopy of the human CECs was 0.79±0.072, of the rabbit CECs before culture was 0.77±0.063, after culture in MitoM was 0.41±0.19, in RestM passage 1 was 0.73±0.09, and in RestM passage 2 was 0.6±0.18 ([Figure 1G](#f1){ref-type="fig"}). There was no statistically significant difference between the circularity of human and basal rabbit CECs or between CECs in RestM P1 and human CECs. The difference in the circularity of human versus MitoM, human versus RestM P2, and basal versus RestM P2 CECs was statistically significant (p\<0.05). These results showed that the resting phase of this culture system enhances the morphology of the CECs, as they take on a corneal--endothelial-like shape. Nevertheless, these results also warn that prolonged passages may also be detrimental. {#f1} Cellular yield analysis ----------------------- CECs in MitoM showed a fold-change increase in the cellular yield of 1.52 from passage 1 to passage 2, whereas CECs in RestM showed an increase of 1.27 ([Figure 2H](#f2){ref-type="fig"}). Although the difference between the yields obtained in the two culture conditions was not statistically significant (p=0.2583), lower proliferation was apparent in RestM. {#f2} Gene expression and functional analysis --------------------------------------- We first tested whether the cell identities were lost during culture by comparing the overall similarity of normalized gene expressions. Hierarchical clustering showed that those cells in the culture were more similar to those of the subject from which the cells were obtained than to the cells in the culture conditions ([Figure 2](#f2){ref-type="fig"}). Nevertheless, a principal component analysis (PCA) revealed that the major source of variation was the culture condition ([Figure 2B](#f2){ref-type="fig"}), which was confirmed with hierarchical clustering, estimated from 5% of the gene expression profiles that had a higher coefficient of variation ([Figure 2C](#f2){ref-type="fig"}). These results support the validity of the assays, treatments, and collected data. We then compared the expression of specific CEC markers between the RestM and MitoM conditions. We assessed 26 selected genes (in gene: probeID from Appendix 1 these are *AQP1*: A_04_P030932, *COL8A1*: A_04_P001586, *ATP1A1*: A_04_P004696, *ATP1B1:* A_04_P002527,*TJP1*: A_04_P086632, *TJP2*: A_04_P067042, *CDH2*: A_04_P031967, *SLC4A11*: A_04_P094717, *GPC4*: A_04_P095302, *CD200*: A_04_P095118, *CLRN1*: A_04_P082648, *GLP1R*: A_04_P079252, *CNTN3*: A_04_P098144, *PCDHB7*, *HTRD1*: A_04_P004265, *GRIP1*: A_04_P098464, *PKD1*: A_04_P012501, *ZP4*: A_04_P003166, *CNTN6*: CNTN6, *SLC3A2*: A_04_P003146, *ALCAM/CD166*: A_04_P020128, ERBB2/CD340: A_04_P049357, *CD9*: A_04_P018111, *CD44*: A_04_P101042, *ITGA5/CD49e*: A_04_P088227, and *NT5E/CD73*: A_04_P084588).which serve as markers for adult CECs, fetal CECs, non-fibroblast CEC phenotypes, and fibroblast CEC phenotypes \[[@r20],[@r21]\]. Of these genes, nine demonstrated expression changes (p\<0.05; [Figure 3A](#f3){ref-type="fig"}). {#f3} The results suggested that adult CECs markers (*ATP1A1, ATP1B1, COL8A2, GPC4, CDH2*, and*TJP1*) were more expressed in the RestM condition, while three non-adult CEC markers (*CD44, CNTN3,*and *CD166*) were more expressed in the MitoM condition. These results showed that markers of specific corneal functions had begun to diverge between the two conditions. Nevertheless, the number of altered genes observed from these CEC markers was too low to compare the overall changes in the biological function. Therefore, to expand the functional analysis, we used the detected differential gene expression profiles to characterize the functional differences in the CECs between the MitoM and RestM conditions. We identified 781 differentially expressed probe genes, as obtained with paired *t* tests between the four replicates of MitoM and RestM at a statistical significance level of p\<0.01, whose fold change was greater than 1, and had an associated known human gene (Appendix 1). A DAVID analysis was performed using the 308 unique genes represented in the 781 differential probes. To summarize these results, we performed hierarchical clustering that resulted from the estimated fold change, and which included the intersection of functional terms and genes. From the results shown in [Figure 3B](#f3){ref-type="fig"}, we identified around 13 functional clusters related to extracellular matrix, collagen type 4, protein modifications, response to a stimulus, apoptosis and antiapoptosis, nuclear lumen, ribosome biogenesis, signal transduction, immune responses, cell proliferation, and wound healing, among others. The details of the biological terms are shown in Appendix 2. From the functional analysis, we chose some functional terms related to relevant biological functions and compared the gene expression levels of the significant genes as the levels related to those functions ([Figure 3C](#f3){ref-type="fig"}). The analysis revealed that CECs in RestM clearly showed a greater expression of genes related to collagen type IV and the extracellular matrix, suggesting remodeling processes. For example, four types of collagen type IV were overexpressed in RestM together with *TIMP3* to inhibit collagenases. Moreover, they were also overexpressed with *LUM*, which binds collagen fibrils; *SMOC2*, which promotes matrix assembly; and *CRTAP*, which is involved in the hydroxylation of fibrillar collagen. Conversely, in MitoM, *MMP1*was more greatly expressed and it is involved in the cleavage of various types of collagen. The function and expression of these genes suggest that they are associated with the different shapes observed under the microscope. Although the term cell proliferation showed that similar numbers of genes were more greatly expressed in MitoM and RestM, some of the terms were clearly related to proliferation states, such as nuclear lumen and ribosome biogenesis, which demonstrated higher expression levels in those genes in the MitoM condition. These results suggested that cells in the MitoM condition are more likely to be related to proliferative processes. In terms of wound healing, CECs in the MitoM condition overexpressed *IL-6*, *F3*, and *ITGB3*, which are related to inflammation, complement cascade, and cell adhesion, respectively. Conversely, CECs in the RestM condition showed higher expression levels of *FBLN5*, which is known to promote endothelial cell adhesion; *insulin-like growth factor (IGF)-1*, which promotes cell growth and development; and *TIMP3*, an inhibitor of matrix metalloproteinases (MMPs). These results suggested that CECs are involved in the remodeling in MitoM, while they are involved during assembly states in RestM. Other functional terms showed a similar number of genes that were more greatly expressed in either condition, such as those related to apoptosis, acetylation/phosphorylation, and response to stimulus. These terms require further analysis. Validation of gene expression ----------------------------- To compare the results with native expression of transcripts and proteins from adult rabbits and to validate the microarray results, we performed qPCR and immunocytochemistry analyses of four genes. *ATP1A1*, *TJP1*,*GPC4*, and *BTG2* were selected for this purpose because they are good markers of CEC functions. We first tested the protein expression of TJP1 (ZO-1), an adult marker of CECs \[[@r20],[@r21]\], which was also overexpressed in RestM, comparing the immunostaining in basal conditions from complete corneas with the growth in both conditions. The fluorescence images shown in [Figure 4A](#f4){ref-type="fig"} support the previous results where the protein expression of ZO-1 was localized more clearly in membranes in cornea and RestM, whereas in MitoM the expression localization was diffuse. For ATP1A1, TJP1, and GPC4, the overall protein expression at P2 seems higher in RestM than in MitoM ([Figure 4B](#f4){ref-type="fig"}), which is consistent with transcriptional measurements where three out of the four genes show a statistically significant increase in expression between RestM and MitoM ([Figure 4C](#f4){ref-type="fig"}). {#f4} Discussion ========== Culture methods for rabbit and human CEC culture have been previously reported. The culture of rabbit CECs using a supplemented medium demonstrated proliferation for up to 67 passages \[[@r22]\]. However, the CECs in the cited study exhibited chromosomal aneuploidy, and they presented a fibroblastic shape \[[@r22]\]. The changes made to a non-supplemented medium for 1 week (MEM 5% BCF) facilitated the recovery of the cells' polygonal shape. In human CECs, it was demonstrated that the combination of vascular endothelial growth factor (VEGF), EGF, fibroblast growth factor (FGF), and IGF yields better cell attachment and cuboidal morphology up to the first passage (approximately 10 days) when compared to the CECs in the basal media \[[@r8]\]. A different study reported that the use of a supplemented culture media, using the same composition of MitoM used in this study, resulted in polygonal CEC proliferation up until the fifth passage, but the expression of specific molecular markers was evident up until the second passage \[[@r6]\]. In a study in which four different media were used, the best outcomes in terms of proliferation, morphology, and molecular marker expression were obtained with the MitoM medium used here, as well as with another medium containing ascorbic acid, insulin--transferrin--sodium selenite (ITS), and bFGF. These two media allowed the CECs to retain their polygonal morphology, and they further expressed ZO-1 and Na/K-ATPase up until the third passage \[[@r10]\]. ZO-1 and Na/K-ATPase are markers of cell identity and the lack of epithelial to mesenchymal transition \[[@r23]\]. The same research group recently reported the use of a supplemented culture medium featuring ascorbic acid, ITS, and bFGF to proliferate CECs. This approach was coupled with a second step in which a maintenance medium (endothelial SFM 5% FBS) was used. In their method, human CECs were cultured for three passages, each consisting of 14 days in the supplemented medium and 7 days in the maintenance medium. This is different from our method, which involved culturing the cells for 5 days in MitoM until confluence, followed by two additional passages in RestM medium (where the composition was the same as in MitoM, but without growth factors). The authors of the previous work reported that the CECs demonstrated a spindle morphology when they were cultured in supplemented medium alone, and they exhibited an enhancement of cell circularity when the non-supplemented medium was added, which persisted throughout the three passages. In the present experiment, the CECs showed a spindle morphology when cultured in MitoM, and they recovered their polygonal morphology after 48 h in RestM, which persisted during the following passage in the same medium. The results of these studies were in accordance with ours with respect to the recovery of the cells' polygonal morphology after the supplemented medium was changed to a basal medium. Further experiments will determine if multiple cycles of proliferation and recuperation retain the characteristics of CECs without demonstrating senescence. Meanwhile, we observed that shape was maintained around the fourth passage. The role of cell senescence and culture media supplementation has been demonstrated in studies using human CECs. The protein tyrosine phosphate *PTP1B*, known to negatively regulate EGF-induced signaling by dephosphorylating the *EGF receptor* (*EGFR*), is more greatly expressed in CECs obtained from younger donors \[[@r14]\]. In this context, the proliferation response to an EGF stimulus in cultured CECs is dependent on the synergy between *PTP1B* and *EGFR*, and it is lower in older donor and senescent cells. This explains why the EGF stimuli response decreased in senescent cells even when the level of EGFR remained relatively constant. Our culture system prevented the CECs from entering into a senescent state by adding a stabilization step. The *PTP1B* gene was not among the differentially expressed genes observed for the CECs in MitoM and RestM, which could be an indicator of the potential of this system for corneal engineering. To better understand the molecular mechanisms underlying the morphological changes recorded throughout the two-culture media, a microarray analysis was performed. The identification of the Gene Ontology (GO) term related to wound healing suggested that CECs in MitoM can act to signal tissue damage, and they further activate a series of events with the objective of restoring cell integrity following an injury. The movement of cells from the G1 to S phase implicates the inhibition of cell--cell contact \[[@r24]\]. Kimura demonstrated that in the presence of harmful stimuli, it is possible to redistribute tight junctions without affecting adherent junctions \[[@r25]\]. This result correlates with the present results, in which the biological process GO term cell adhesion molecule binding, the CC-GO adherens junction, and the pathway for the adherent junction were established, while no changes were observed for tight junctions. GO terms related to the actin cytoskeleton and the regulation of cell projection assembly suggest dynamic actin cytoskeletal organization. Actin cytoskeletal reorganization can be related to the protrusive forces involved in cell migration \[[@r26]\]. The identification of the MF-GO terms related to metalloendopeptidase activity, collagen metabolic processes, and collagen degradation, along with the GO-BP-positive regulation of catalytic activity, suggest that a catalytic process is involved in extracellular collagen composition, which is controlled by MMPs. MMPs are important in the process of connective tissue remodeling \[[@r27]\], and they may be involved in the reorganization of the TJ protein seen in CECs in MitoM. CECs in MitoM did not express MMP inhibitors; thus, the changes in cell morphology could be related to dynamic actin cytoskeletal reorganization, changes in the extracellular matrix composition, and reorganization of the cell junctions. CECs in RestM showed GO terms related to wound healing as well, such as focal adhesion, remodeling in the extracellular matrix, and extracellular matrix receptor interaction, which suggests that RestM also acts as a signal to indicate damage, ultimately activating a series of events related to restoring the monolayer's integrity. The polygonal morphology seen in CECs in RestM is a good index of the progress of endothelial restoration \[[@r28]\], and together with the rearrangement of the cytoskeleton, it recalls the tissue remodeling state. The expression of collagen IV suggests that RestM can promote the reestablishment of the structure and composition of the extracellular matrix, which is essential for the attachment of CECs to DM. qPCR validated the difference in the expression of the *ATP1A1, TJP1,*and*GPC4* genes among the CECs before culture, in MitoM, and in RestM. The CECs in RestM showed higher expression level of these markers. However, the expression levels were higher than those observed in basal CECs. We hypothesize that activation of the proliferative state during MitoM before the RestM condition could lead to these results. Because we analyzed the expression level at 80% cell confluence, expression levels during an increased rate of transcription were obtained, in contrast to basal expression, in which corneal endothelium is in an expression steady-state. Further experiments with a different rate of confluence would confirm these insights. Immunocytochemistry analysis demonstrated similar locations and expression levels of TJP1/ZO-1 in flat CE and in CECs in RestM. In addition, CECs in RestM showed higher expression of adult CE markers GPC4 and Na/K-ATPase compared to CECs in MitoM. The differentially expressed genes obtained in this system are similar to those reported by Peh et al. following the use of a different proliferative medium coupled with a maintenance medium \[[@r29]\]. Cell proliferation and wound healing were among the top pathways evident in both approaches. Although there are methodological differences between their study and the present study, the results demonstrated the effectiveness of a coupled system in the proliferation of CECs, as well as in the maintenance of their morphological and molecular characteristics. Further in vivo analyses will determine the efficacy of CECs cultured with these systems in restoring corneal function. It has been shown that the proliferative capacity of CECs is age-dependent in humans and rabbits \[[@r30],[@r31]\]. In the present experiments, we used 3-month-old rabbits which show high mitotic activity. It would be interesting to compare the behavior of the present culture system at different ages. CONCLUSIONS ----------- We showed a two-phase system with novel medium and transcriptome data, analysis, and validation. Other research groups used a different medium \[[@r29]\] or did not analyze the transcriptome \[[@r32]\]. The differences in the morphology, pathways, and gene expression observed between CECs in RestM and MitoM suggest that although MitoM enhances CEC proliferation, it could result in cell differentiation and drive the culture to exhibit a wound-like state. The resting step facilitated the recovery of the cells' hexagonal shape; it further benefitted the maintenance of pump function, cell-cycle arrest, the cells' barrier function (via junction reorganization), the reconstruction of the extracellular matrix's structural constituent, and the production of collagen IV (a component of DM), all of which are related to the final events involved in remodeling during the wound-healing process. Future experiments focused on analyzing the number of cycles in which the CECs cultured in this system are able to proliferate and recover specific markers will provide additional evidence for this system's potential in regenerative medicine. Cells cultured in this system may ultimately address the shortage of tissue donors available for corneal grafts. English-language editing of this manuscript was provided by Journal Prep. This work was funded by Consejo Nacional de Ciencia y Tecnología (CONACyT) grant PN6558 and Tecnologico de Monterrey. To access the data, click or select the words "[Appendix 1](http://www.molvis.org/molvis/v25/appendices/mv-v25-745-app-1.tif)." A) Tissue conglomerates obtained after collagenase I treatment. B) Cell clusters obtained after trypsin/EDTA treatment. C) Adherent cells after enzymatic digestion. To access the data, click or select the words "[Appendix 2](http://www.molvis.org/molvis/v25/appendices/mv-v25-745-app-2.tif)." The heat map shows the genes (horizontal axis) contained within functional biologic terms (vertical axis). The color represents the fold change in gene expression (cyan is used to represent those genes that were more greatly expressed in MitoM, while purple represents those genes more greatly expressed in RestM). Genes whose *t* test p values were \<0.01, and which demonstrated a fold change \>1, were used. Only over-represented biologic terms with a p value \<0.01, a false discovery rate \<0.25, and those that contained \>3 genes were used in this analysis.

Commentary {#Sec1} ========== "*Complexity is the enemy of transparency*" \[[@CR1]\]. Today, *BMC Medicine* publishes another paper on journalology (publication science) \[[@CR2]\]. Attributing authorship, and authorship order, is complex and often a 'black box' for prospective authors. Professor Marušić and colleagues have tried to peel back the black box concerning the assigning of authorship for industry-sponsored clinical trials. Their methods are good and reported in sufficient detail to allow interested readers to replicate them \[[@CR3]\]. The research team have used an integrated knowledge translation approach to developing their proposed five-steps for transparently disclosing authorship. Participants from pharmaceutical companies that conduct clinical trials, academics, editors, and the Medical Publishing Insights and Practices initiative were involved in the entire process; this facilitates buy-in and support for the process and outcome. These same people are likely to become front-line ambassadors and early adopters for disseminating and implementing the five-step framework within their own working environments, and hopefully, more broadly. What is positive about this research is that the proposed attribution process for authorship is brief, and not complex; it's only five-steps. It is meant to augment the guidance provided by the International Committee of Medical Journal Editors \[[@CR4]\]. To enhance uptake of the framework it will be important for the team, or others, to develop a bank of worked examples for each step in the five-step process. Using worked examples from specific trials will likely facilitate implementation. The authors have started the process with seven case examples included in their publication. A dedicated website for the framework whereby authorship examples can be submitted by pharmaceutical companies and others, vetted and added to a bank of examples, freely accessible to anybody, is worth considering. This will help prospective clinical trialists ensure a transparent process in deciding on authorship. If this authorship initiative is to be successful it requires endorsement and, more importantly, implementation. As Marušić and colleagues note \[[@CR2]\], previous efforts, such as contributorship, have not been broadly implemented. What is less clear is how the proposed framework is going to be endorsed and implemented. Important initial steps should include strong and consistent language of endorsement across all of the pharmaceutical companies involved in the development of the five-step process. Support and endorsement from umbrella groups, such as the Pharmaceutical Research and Manufacturers of America \[[@CR5]\], and others such as CONSORT \[[@CR6]\], is also worth considering. While endorsement is a useful step, it is difficult to measure and likely not the most relevant outcome. More important will be to develop plans based on appropriately developed approaches \[[@CR7],[@CR8]\] to implement the framework. This is likely to be most effective when pharmaceutical companies modify their authorship practices and polices when conducting any clinical trial. An effective policy would require all clinical trials to implement the five-step framework at their inception. Without strong implementation the framework is less likely to affect positive change. This has been observed when trying to implement reporting guidelines in biomedical journals \[[@CR9],[@CR10]\]. Part of any implementation plan also needs to include an evaluation of the framework. It is important to collect data that will inform its usefulness. There is little merit in maintaining policies that are not supported by evidence. Marušić et al. are silent on whether their framework can be used when developing authorship for submitting clinical trial protocols for publication consideration \[[@CR2]\]. Making clinical trial protocols accessible is important and at least one Biomed Central journal -- *Trials* -- regularly publishes them. Additionally, with the requirement of trial registration, this framework could also be used when completing the investigator information part of the registration. Most of us are not born authors. It is an acquired skill that often starts during graduate school. This is where all journalology issues, including those pertaining to authorship issues (e.g., attributing authorship, authorship order, and ghost and guest authorship, author responsibilities) should be formally taught and discussed. Developing such skills early can translate into something useful throughout a researcher's (author's) career. It is unfortunate that almost all universities, and other centres of higher learning, appear to have abdicated their responsibilities regarding formally teaching journalology; the irony is not lost on me. These institutions are the very same places developing the next generation of biomedical researchers. Universities need to set aside appropriate resources to enable and promote such courses, and others, related to journalology \[[@CR11]\]. While authors have rights and privileges, they also have important responsibilities that require much greater attention. Given the opportunity of authorship, it is equally important to assert this responsibility. Authors must ensure that papers baring their name are "fit for purpose" \[[@CR12]\]. Here, authors need to ensure that every report baring their name is a completely reported and transparent account of what was done (methods) and found (results) to enable interested readers to replicate the methods and use the results. Collectively, authors have not performed appropriately with regards to reporting their clinical trials. This avoidable waste is troublesome for shareholders of publicly traded pharmaceutical companies and tax payers of publicly-funded clinical trials. It is not a good return of a fiscal investment when reports of trials are so inadequate that their results cannot be used. For example, Duff et al. \[[@CR13]\] examined 262 reports of trials from the most prominent oncology journals assessing them for 10 essential elements regarding the description of their interventions, such as the drug's name and route of administration. The authors reported that only 11% of the articles reported all 10 characteristics. Although we have seen improvements over time in reporting the unique characteristics of randomized trials -- sequence generation, allocation concealment, and implementation -- these items are adequately reported in less than half of the trial reports \[[@CR14]\]. In some clinical specialties, the situation is much worse \[[@CR15]\]. Disclosing authorship transparently is important for any manuscript being submitted to a biomedical journal for publication consideration. The responsibilities associated with authorship must be taken seriously. This might help increase value and reduce avoidable waste of biomedical research. **Competing interests** The author declares that he has no competing interests.

Baker SC, Arlt VM, Indra R, et al. Differentiation‐associated urothelial cytochrome P450 oxidoreductase predicates the xenobiotic‐metabolizing activity of "luminal" muscle‐invasive bladder cancers. Molecular Carcinogenesis. 2018;57: 606--618. <https://doi.org/10.1002/mc.22784> 29323757 ABS : adult bovine serum AHR : aryl‐hydrocarbon receptor BaP : benzo\[a\]pyrene CYP : cytochrome P450 DMSO : dimethyl sulfoxide EROD : ethoxyresorufin O‐deethylation ITE : 2‐(1′H‐indole‐3′‐carbonyl)‐thiazole‐4‐carboxylic acid methyl ester KSFMc : Keratinocyte Serum‐Free Medium complete MIBC : muscle invasive bladder cancer NHU : normal human urothelial (NHU) cells PAH : polycyclic aromatic hydrocarbon POR : NADPH:P450 oxidoreductase RT‐qPCR : Reverse Transcribed‐quantitative Polymerase Chain Reaction SNP : single nucleotide polymorphism TMA : tissue micro‐array TMS : 2,3′,4,5′‐tetramethoxystilbene 1. INTRODUCTION {#mc22784-sec-0002} =============== The epithelial lining of the bladder and associated urinary tract, the urothelium, functions as a barrier to prevent reabsorption of the concentrated waste products of metabolism. The mechanisms used by the urothelium to survive a lifetime of toxin and toxicant exposure remain underexplored, although preliminary evidence of cytochrome P450 (CYP) transcript/activity supports a potentially important role in the urinary tract (reviewed in Ref. [1](#mc22784-bib-0001){ref-type="ref"}). The aryl‐hydrocarbon receptor (AHR) is a transcription factor expressed constitutively by many epithelial tissues. In epidermal keratinocytes, AHR‐mediated transcription has been implicated in differentiation,[2](#mc22784-bib-0002){ref-type="ref"} although AHR is better known for regulating expression of a battery of *CYP* genes (including *CYP1A1*, *CYP1A2*, and *CYP1B1*) as part of Phase I metabolism in the detoxification of xenobiotics. CYP1‐mediated metabolism is particularly important for aromatic amines, which include the bladder pro‐carcinogens 2‐naphthylamine and 4‐aminobiphenyl, and polycyclic aromatic hydrocarbons (PAH), including benzo\[*a*\]pyrene (BaP; reviewed in Ref. [3](#mc22784-bib-0003){ref-type="ref"}). Phase I metabolism of xenobiotics by CYPs is often the first step in detoxification as it supports the Phase II conjugation reactions. However, Phase I metabolites are frequently reactive and studies have shown CYP1 enzyme function to be activating in the case of PAH pro‐carcinogens.[4](#mc22784-bib-0004){ref-type="ref"} *CYP1A1*, *CYP1B1*, and the AHR nuclear translocator (*ARNT*) have been identified as potential risk factors for human bladder cancer through association of single nucleotide polymorphisms (SNPs) with the disease,[5](#mc22784-bib-0005){ref-type="ref"}, [6](#mc22784-bib-0006){ref-type="ref"} but a robust case is lacking and it remains unclear whether the SNPs affect production of genotoxicants solely in the liver, or whether extra‐hepatic metabolism is involved. Bladder cancer has a high prevalence of somatic mutations, a trait shared with cancers where chronic mutagen exposure plays a causal role (such as lung cancer and melanoma).[7](#mc22784-bib-0007){ref-type="ref"} Smoking is the main risk factor for bladder cancer with a hazard ratio of 2.33 for former smokers and 4.27 for current smokers.[8](#mc22784-bib-0008){ref-type="ref"} Following PAH exposure, incomplete hepatic metabolism leads to the excretion of bladder pro‐carcinogens in the urine[9](#mc22784-bib-0009){ref-type="ref"} where urothelial CYP‐activity could lead to DNA adduct formation and ultimately mutation. BaP is the major PAH in cigarette smoke and the bladder tissue of current smokers contains bulky DNA adducts.[10](#mc22784-bib-0010){ref-type="ref"} Acting in combination with microsomal epoxide hydrolase (EPHX1), the bioactivation of BaP to species capable of forming DNA adducts is performed most efficiently by CYP1A1 and CYP1B1.[11](#mc22784-bib-0011){ref-type="ref"} Studies of *Cyp1*‐knockout mice suggest Cyp1a1 is essential for detoxification of BaP in the epithelium of the gastrointestinal tract, but the bladder remains unstudied (reviewed in Ref. [12](#mc22784-bib-0012){ref-type="ref"}). CYP activity is driven by electron donation from the NADPH:CYP oxidoreductase (POR) and abundance of POR determines metabolic capacity in the CYP system.[13](#mc22784-bib-0013){ref-type="ref"}, [14](#mc22784-bib-0014){ref-type="ref"} POR is one member of the diflavin oxidoreductase family (nitric oxide synthase is the other) and is not specifically a reductase for the CYPs but can donate electrons to heme oxygenase among other enzymes (reviewed in Ref. [15](#mc22784-bib-0015){ref-type="ref"}). POR expression in normal and neoplastic urothelium of humans remains unquantified, although a SNP was recently associated with increased bladder cancer risk.[16](#mc22784-bib-0016){ref-type="ref"} Since the first report of CYP1 activity in rabbit bladders in 1985,[17](#mc22784-bib-0017){ref-type="ref"} there has only been fragmented study of AHR‐mediated metabolism in the urinary tract. Studies of non‐transformed cells have used either normal porcine urothelial cells (which demonstrate no ethoxyresorufin O‐deethylation (EROD), a measure for CYP1 enzyme activity)[18](#mc22784-bib-0018){ref-type="ref"} or human cells cultured from urinary sediments (which most likely originate from the kidney),[19](#mc22784-bib-0019){ref-type="ref"} limiting their relevance to human urothelium. It is therefore timely to establish whether CYP1 enzymes function in human urothelium and whether there is sufficient capacity to activate potential bladder carcinogens. Normal (non‐neoplastic) human urothelium is not available in the quantities required to generate microsomes for the study of enzyme activity. The aim of this study was to establish the potential of bladder CYP1 metabolism; using in vitro cell culture models representing normal and neoplastic human urothelium in undifferentiated and differentiated states (reviewed in Ref. [20](#mc22784-bib-0020){ref-type="ref"}). Immunohistology was used to relate in vitro findings to the biology of human bladder tissue and muscle‐invasive bladder cancers (MIBC). 2. METHODS {#mc22784-sec-0003} ========== 2.1. Tissues and cells {#mc22784-sec-0004} ---------------------- All tissues were collected with appropriate Research Ethics Committee approvals and patient consents as required. In the UK, tissues were collected under the following NHS REC‐approved projects (Leeds East REC projects 99/095, 02/208, 00/22; York REC project 99/04/003, Newcastle & North Tyneside 1 REC 13/NE/0081). In Germany, tissue use was approved by the local research ethics committee of the University of Regensburg (IRB Number: 08/108) and materials (tissue sections) were transferred to York under a Materials Transfer Agreement. Finite (non‐transformed) normal human urothelial (NHU) cell lines were established from surgical specimens (most commonly discarded ureteric tissues from renal cell carcinoma or transplant), as previously detailed.[21](#mc22784-bib-0021){ref-type="ref"} NHU cell lines were propagated as undifferentiated cultures in Keratinocyte Serum‐Free Medium containing recombinant epidermal growth factor and bovine pituitary extract (KSFMc; Invitrogen).[21](#mc22784-bib-0021){ref-type="ref"} For the studies described here, 15 independent cell lines were used between passages one and five. In the figures, each independent donor NHU cell line used has been given a unique symbol consistent between every panel (including supplementary). Differentiation of in vitro‐propagated NHU cell cultures into barrier‐forming stratified urothelial cell sheets was performed as described in detail elsewhere.[22](#mc22784-bib-0022){ref-type="ref"} Briefly, NHU cells propagated in KSFMc were preconditioned for 5 days in 5% adult bovine serum (ABS) before harvesting and reseeding in same. After 24 h, the exogenous calcium concentration was increased to near‐physiological (2 mM) and cultures were maintained for 7 days before performing assays. The endogenous natural ligand 2‐(1′H‐indole‐3′‐carbonyl)‐thiazole‐4‐carboxylic acid methyl ester (ITE; Tocris) was used to activate AHR in cell cultures at a concentration of 1 μM. NHU cell cultures were exposed to BaP (CAS no. 50‐32‐8; purity ≥96%; Sigma) at a concentration of 2 μM for 6 h. CYP1‐activity and BaP metabolism were inhibited by 2,3′,4,5′‐tetramethoxystilbene (TMS; Enzo Life Sciences) at 0.5--12 μM. Both compounds used DMSO as a vehicle and all controls contained a matched concentration of DMSO (not greater than 0.1% v/v). For bladder cancer cell organ cultures,[23](#mc22784-bib-0023){ref-type="ref"} stromal tissue remaining after de‐epithelialization[21](#mc22784-bib-0021){ref-type="ref"} was dissected into 0.5 cm^2^ pieces and placed on Falcon membrane inserts (3 μm pore, Corning) for culture. RT4 (ECACC 91091914, sourced in 2000), RT112 (ECACC RT112/84 85061106, sourced in 2000) T24 (ATCC HTB‐4, sourced in 1999), and SCaBER (ATCC HTB‐3, sourced in 1999) cells were authenticated by short tandem repeat profiling using the PowerPlex16 System (Promega) in September 2016 and within five passages of use in this study (all cell lines were a perfect match to the ATCC genotype records). Bladder cancer cell lines were seeded onto the basement membrane of de‐epithelialized ureters and cultured at the air:liquid interface in DMEM:RPMI 1640 (50:50, v/v) with 5% fetal bovine serum for 4 weeks. Cancer cell organ cultures were fixed in formalin for 24 h, processed into paraffin wax and sectioned at 5 μm for immunoperoxidase labeling. A tissue microarray (TMA) was constructed from formalin‐fixed paraffin wax‐embedded MIBC biopsies, exactly as described[24](#mc22784-bib-0024){ref-type="ref"} using tumor biopsies obtained from 61 non‐consecutive patients who underwent radical cystectomy for muscle‐invasive urothelial carcinoma of the urinary bladder in the Department of Urology, Regensburg University Medical Center, between 1998 and 2008. Median patient age was 71 (range: 55‐87) years. Eighty‐one percent of patients were male. Surgical specimens were assessed histopathologically by a single expert uropathologist for grading and staging based on the criteria of the 1973 WHO classification and 2010 TNM system.[25](#mc22784-bib-0025){ref-type="ref"}, [26](#mc22784-bib-0026){ref-type="ref"} Clinical characteristics of the patients are summarized in Supplementary Table S1. 1.5‐mm donor tissue cores were used, and the representativeness of the TMA was confirmed histopathologically by comparing the TMA and the original tissue sections for each tumor. Immunoperoxidase labeling of the MIBC TMAs was performed on 4‐μm sections mounted on poly‐L‐lysine‐coated glass slides and compared to a "normal" control group of bladders (non‐trigone cold‐cut biopsies) with no history of bladder malignancy. These tissues were from patients biopsied for a range of conditions including prostate cancer, stress incontinence, overactive bladder, and cystitis, some of which have been described previously.[27](#mc22784-bib-0027){ref-type="ref"} 2.2. Indirect‐immunofluorescent labeling {#mc22784-sec-0005} ---------------------------------------- To generate differentiated NHU cell cultures for immunolabeling, NHU cells preconditioned for 5 days in 5% ABS were passaged and seeded onto glass 12‐well slides at 3 × 10^4^ cells/well. After 24 h the calcium concentration was raised to 2 mM and the cultures maintained for 7 days, before treatment with ITE or a vehicle control. After 16 h, slides were fixed in methanol:acetone (1:1) for 30 s, air dried and rabbit anti‐AHR affinity‐purified polyclonal antibody was applied overnight at 4°C (1:500 dilution; Enzo Life Sciences, BML‐SA210). Unbound primary antibody was removed by washing in PBS and secondary antibody (Alexa‐594, Molecular Probes) was applied for 1 h at ambient temperature. Slides were washed in PBS, with 0.1 μg/mL Hoechst 33258 added to the penultimate wash, before mounting in antifade mountant and visualization by epifluorescence on a BX60 microscope (Olympus). Analysis was performed on images collected at a fixed exposure using TissueQuest Software (Tissue Gnostics). 2.3. Western blotting {#mc22784-sec-0006} --------------------- Twenty micrograms of whole cell protein lysate was loaded into each gel track for electrotransfer to PVDF membranes with the full method provided in Supplementary Methods. The test antibodies used were anti‐AHR (1:2000, rabbit, Enzo Life Sciences, BML‐SA210), anti‐CYP1A1 (1:4000, rabbit, generous gift from Prof. F. Peter Guengerich, Vanderbilt University), anti‐POR (1:10 000, rabbit "CH60" a kind gift from Prof. Roland Wolf and Dr Colin Henderson, Dundee University).[28](#mc22784-bib-0028){ref-type="ref"} Homogeneous loading and transfer were evaluated using a β‐actin antibody (Sigma, Clone AC15, Mouse, 1:10 000 dilution). Detection of CYP1A1 and GAPDH protein was performed exactly as described.[29](#mc22784-bib-0029){ref-type="ref"} Densitometry was performed in all cases using Image Studio Lite Ver 5.0 software (LI‐COR). Cropped Western blots are shown in the main figures with full blots provided as Supplementary Figures S1 and S2. 2.4. Reverse transcribed‐quantitative polymerase chain reaction (RT‐qPCR) {#mc22784-sec-0007} ------------------------------------------------------------------------- Method and primers are provided in Supplementary Materials. 2.5. Ethoxyresorufin O‐deethylation (EROD) activity assays {#mc22784-sec-0008} ---------------------------------------------------------- EROD activity assays, used as a measure for CYP1 enzyme activity, were performed as described in detail elsewhere[30](#mc22784-bib-0030){ref-type="ref"} with minor modifications to convert the assay for black‐walled collagen‐coated glass coverslip‐bottomed 96‐well plates (BD Biosciences). Briefly, NHU/RT4/RT112/T24/SCaBER cells seeded at 6 × 10^4^/well were induced with ITE for predetermined times before washing in HEPES‐buffered saline (HBS; 138 mM NaCl, 5 mM KCl, 0.3 mM KH~2~PO~4~, 4 mM NaHCO~3~, 0.3 mM NaHPO~4~, 1 mM MgCl~2~, 2 mM CaCl~2~, and 10 mM HEPES pH 7.4) and treatment with 5 μM ethoxyresorufin (Cambridge Bioscience) in HBS. CYP1 enzyme activity was specifically inhibited by inclusion of TMS in the ethoxyresorufin/HBS. Plates were incubated for 75 min at 37°C; fluorescence was detected with 544 nm excitation and 590 nm emission filters using a POLARstar optima plate reader (BMG Labetch, housed in the Bioscience Technology Facility at York), and EROD activities were calculated using a (19‐625 nM) resorufin standard curve (*R* ^2^ = 0.996), which was corrected for cellular auto‐fluorescence. Following fluorescence measurement, plates were washed twice in phosphate‐buffered saline (PBS) and cells lysed in RIPA buffer (25 mM Hepes pH 7.4, 125 mM NaCl, 10 mM NaF, 10 mM sodium orthovanadate, 10 mM sodium pyrophosphate, 0.2% (w/v) SDS, 0.5% (w/v) sodium deoxycholate acid, 1% (w/v) Triton X--100, 1 mg/mL aprotinin, 10 mg/mL leupeptin, and 100 mg/mL phenylmethylsulphonyl fluoride) for a bicinchoninic acid (BCA) protein assay used for normalization (Fisher). 2.6. High performance liquid chromatography (HPLC) analysis of BaP metabolites {#mc22784-sec-0009} ------------------------------------------------------------------------------ Culture medium from confluent 10 cm dishes of NHU cells was collected on ice, centrifuged at 4°C for 5 min at 300*g* and stored at −80°C until analysis. Per sample, 1 mL of medium was extracted twice with 1 mL of ethyl acetate. Extracts were evaporated and taken up in 30 μL methanol, of which 20 μL aliquots were injected on HPLC. HPLC analysis was performed using a HPLC Agilent 1100 System (Agilent Technologies) with a SunFireTM C18 reverse phase column (250 × 4.6 mm, 5 μm; Waters). The conditions used for the chromatographic separation of BaP metabolites were as follows: mobile phase (A) 50% acetonitrile in water (v/v), mobile phase and (B) 85% acetonitrile in water (v/v). The separation started with an isocratic elution of 1.4% of mobile phase B. Then a linear gradient to 98.5% of mobile phase B in 34.5 min was followed by isocratic elution for 6 min, a linear gradient from 98.5% to 1.4% of mobile phase B in 3 min, followed by an isocratic elution for 1.5 min. Total run time was 45 min at a flow rate of 1 mL/min. The metabolites were analyzed by fluorescence detection (excitation wavelength 381 nm, emission wavelength 431 nm). The two BaP metabolites analyzed, (±)‐*trans*‐7,8‐dihydroxy‐7,8‐dihydrobenzo\[*a*\]pyrene (BaP‐*t*‐7,8‐dihydrodiol) and (±)‐*r*‐7,*t*‐8,*t*‐9,*c*‐10‐tetrahydroxy‐7,8,9,10‐tetrahydrobenzo\[*a*\]pyrene (BaP‐tetrol‐I‐1), were identified using authentic standards which were synthesized as described.[29](#mc22784-bib-0029){ref-type="ref"} 2.7. ^32^p‐postlabeling of BaP‐DNA adducts {#mc22784-sec-0010} ------------------------------------------ Following collection of the medium, cultures were washed twice in cold D‐PBS (Gibco), scraped in 1 mL PBS, centrifuged at 800*g* for 5 min at 4°C and dry cell pellets were stored at −80°C until analysis. DNA was isolated from BaP‐treated cells using a standard phenol/chloroform extraction method. BaP‐DNA adduct formation was determined using the nuclease P1 digestion enrichment version of thin‐layer chromatography (TLC) and ^32^P‐postlabeling assay was carried out as described.[31](#mc22784-bib-0031){ref-type="ref"} Briefly, DNA samples (4 µg) were digested with micrococcal nuclease (288 mU; Sigma) and calf spleen phosphodiesterase (1.2 mU; MP Biomedical) and then enriched and labeled. Solvent conditions for the resolution of ^32^P‐labeled adducts on polyethyleneimine‐cellulose TLC were as described.[31](#mc22784-bib-0031){ref-type="ref"} Subsequently, TLC sheets were scanned using a Packard Instant Imager (Dowers Grove, IL) and DNA adducts (RAL, relative adduct labeling) were quantified from the adduct counts per minute (cpm), the specific activity of \[γ‐^32^P\]ATP (HP601PE; Hartmann Analytic) and the amount of DNA (pmol of DNA‐P) used. Results were expressed as DNA adducts per 10^8^ normal nucleotides. An external BPDE‐DNA standard[32](#mc22784-bib-0032){ref-type="ref"} was employed for identification of adducts in experimental samples. 2.8. Immunoperoxidase labeling {#mc22784-sec-0011} ------------------------------ For immunolabeling of POR (1:4000, Mouse Clone F10, Santa Cruz, sc‐25270) and GATA3 (1:800, Rabbit antibody D13C9, Cell Signalling), heat‐mediated antigen retrieval was used (boiling for 10 min in 10 mM citric acid buffer \[pH 6\]); and labeling was performed using the Impress polymer detection kit according to the manufacturer\'s instructions (VectorLabs). All sections were counterstained with Mayer\'s haematoxylin, dehydrated and mounted in DPX (CellPath). Cytokeratin 5/6 (KRT5/6; 1:50, M7237, Dako) labeling was performed on the Leica Bond 3 platform using Epitope Retrieval Solution 2 (AR9640; Leica Biosystems) for 30 minutes, a primary antibody application of 15 minutes, the Bond Polymer Refine Detection Kit (DS9800; Leica Biosystems) and Bond DAB enhancer (AR9432; Leica Biosystems) for 5 min. Slides were imaged using an AxioScan.Z1 slide scanner (Zeiss). Labeling intensities were quantified using Histoquest software (v3.5, Tissue Gnostics) whereby regions containing \>90% tumor cells were manually identified. For KRT5/6 quantification, the percentage area of tumor tissue labeled above the threshold intensity (of 50 arbitrary units) was calculated. An automated algorithm identified nuclei and cytoplasm based on the haematoxylin counterstain in order to perform the following analyses. Contaminating lymphocytes were removed from the analysis by gating out cells with a nuclear size smaller than 30 μm^2^. For GATA3 labeling, mean nuclear DAB intensity was quantified and a defined threshold (of labeling intensity 14; arbitrary units) was used to generate a labeling index of percentage positive nuclei within each tumor. For POR labeling, identified nuclei were used to support recognition of the surrounding cell body and cytoplasmic DAB intensity was quantified. The basal/luminal classification of the tumors based on GATA3 and KRT5/6 labeling was performed by reproducing the Logistic regression (LRA) and support vector machine (SVM) cut‐off lines derived previously by Dadhania et al.[33](#mc22784-bib-0033){ref-type="ref"} 2.9. Statistical analysis {#mc22784-sec-0012} ------------------------- Data were assessed for statistical significance using InStat 3 or Prism 6 software (Graphpad). On all graphs statistical significance is represented as follows; \**P *\< 0.05, \*\**P *\< 0.01, and \*\*\**P *\< 0.001. 3. RESULTS {#mc22784-sec-0013} ========== 3.1. AHR in urothelial differentiation {#mc22784-sec-0014} -------------------------------------- AHR protein was detected in similar abundance in both undifferentiated and differentiated NHU cell cultures (Figure [1](#mc22784-fig-0001){ref-type="fig"}A). AHR protein was observed to be widely distributed throughout the cell in differentiated NHU cells treated with a vehicle control; however, following exposure to 1 μM ITE for 16 h, AHR was significantly more abundant in the nucleus (Figures [1](#mc22784-fig-0001){ref-type="fig"}B and [1](#mc22784-fig-0001){ref-type="fig"}C). {#mc22784-fig-0001} AHR expression in epidermis has been attributed to a role in differentiation.[2](#mc22784-bib-0002){ref-type="ref"} *Ahr*‐knockout mice showed no perturbation of urothelial morphology or in the expression or distribution of proteins involved in urinary barrier function (including Claudin 5 and Uroplakin 3a, Supplementary Figure S3 and Supplementary Methods). Furthermore, trans‐epithelial electrical resistance monitored as a measure of barrier‐function in differentiated NHU cell cultures showed no significant change in response to ITE exposure throughout differentiation (mean resistance of 5007 vs 4978 Ω . cm^2^ for vehicle control and 1 μM ITE treatment, respectively; Student\'s *t*‐test *P* = 0.96, *n* = 6 cultures; Supplementary Figure S1 and Supplementary Methods). 3.2. Induction of CYP1A1 and CYP1B1 transcripts by AHR {#mc22784-sec-0015} ------------------------------------------------------ The effect of the endogenous AHR ligand "ITE" on *CYP1* gene expression was studied in NHU cell lines from six independent donors. Significant (*P* ≤ 0.001) induction of *CYP1A1* and *CYP1B1* transcripts was observed in response to ITE in both undifferentiated (30 and 49‐fold, respectively) and differentiated (23 and 19‐fold, respectively) cultures compared to their respective vehicle controls (Figure [1](#mc22784-fig-0001){ref-type="fig"}D). Differentiated NHU cells showed preferential upregulation of *CYP1A1* transcript, while undifferentiated cells induced *CYP1B1* to a greater extent (Supplementary Figure S4). Differentiation of NHU cells itself induced a smaller increase in *CYP1A1* and *CYP1B1* transcript expression (4.1 and 1.5‐fold, respectively; Figure [1](#mc22784-fig-0001){ref-type="fig"}D). No expression of *CYP1A2* transcript was detected in NHU cells (data not shown). 3.3. Induction of EROD activity in NHU cells by AHR {#mc22784-sec-0016} --------------------------------------------------- Basal EROD activity was barely detectable in NHU cells and was not inducible in undifferentiated NHU cells (Figure [2](#mc22784-fig-0002){ref-type="fig"}). EROD activity was rapidly induced in differentiated NHU cells by 1 μM ITE, with effects apparent by 8 h and a peak of activity at 16 h (mean 55‐fold increase, Figure [2](#mc22784-fig-0002){ref-type="fig"}). After 16 h (without replenishment of ITE), EROD activity in differentiated NHU cells began returning to baseline (Figure [2](#mc22784-fig-0002){ref-type="fig"}). {#mc22784-fig-0002} 3.4. Inhibition of EROD activity by TMS {#mc22784-sec-0017} --------------------------------------- EROD can be performed by all CYP1 family members with varying degrees of efficiency.[34](#mc22784-bib-0034){ref-type="ref"} As no *CYP1A2* transcript expression was observed in the urothelium, this enzyme was ruled out. Following a 16 h treatment with 1 μM ITE, induced CYP1 enzyme activity was inhibited by TMS in differentiated NHU cells from three donors with a mean IC~50~ of 6.9 μM (Figure [3](#mc22784-fig-0003){ref-type="fig"}A). To further characterize the inhibition of EROD‐activity by TMS, a Ki was derived from differentiated NHU cells following a 16 h treatment with 1 μM ITE. A Michaelis fit Lineweaver‐Burk plot demonstrated a mixed‐type inhibition where TMS had greater affinity for the free enzyme(s) than for the enzyme/substrate complex (Figure [3](#mc22784-fig-0003){ref-type="fig"}B). The mean estimated Ki for TMS against differentiated NHU cell CYP1 enzymes was 0.39 μM (±0.07) using quadratic analysis or 0.26 μM by identifying the X axis intersect of the Km/Vmax trendline (Figure [3](#mc22784-fig-0003){ref-type="fig"}C). {#mc22784-fig-0003} 3.5. CYP1A1 and POR Western blotting before and after BaP exposure {#mc22784-sec-0018} ------------------------------------------------------------------ NHU cells from three independent donors were lysed for whole‐cell Western blots to determine CYP1A1 and POR expression. Induction of CYP1A1 protein by 16 h pre‐treatment ±1 μM ITE confirmed RT‐qPCR data suggesting expression was low for control cells and significantly induced by ITE in both undifferentiated and differentiated states, but to a significantly greater extent in differentiated NHU cell cultures (Figure [4](#mc22784-fig-0004){ref-type="fig"}). {#mc22784-fig-0004} RT‐qPCR showed BaP exposure did not induce *CYP1* transcripts in undifferentiated NHU cells but did significantly in differentiated cultures (Supplementary Figure S5A). This result was confirmed for CYP1A1 at the protein level, where BaP exposure did not induce CYP1A1 expression in undifferentiated NHU cells, but did significantly in the differentiated cultures (Figure [4](#mc22784-fig-0004){ref-type="fig"}). No specific antibody could be found to support CYP1B1 detection. Western blotting for POR established that abundance was not affected by ITE, TMS, or BaP exposure; but was elevated by differentiation of NHU cells (mean 2.1‐fold; Figure [4](#mc22784-fig-0004){ref-type="fig"}). RT‐qPCR of *POR* transcript confirmed differentiation of NHU cells from three donors increased *POR* transcript expression on average by 2.9‐fold; however, 1 μM ITE treatment had no significant effect on *POR* mRNA expression (Supplementary Figure S5B). 3.6. BaP metabolism {#mc22784-sec-0019} ------------------- Cell lines established from three independent donors were monitored before (EROD, Figure [5](#mc22784-fig-0005){ref-type="fig"}A) and after BaP treatment (HPLC and ^32^P‐postlabeling, Figure [5](#mc22784-fig-0005){ref-type="fig"}B‐D). Monitoring EROD‐activity prior to BaP exposure confirmed significant induction of CYP1‐function in all three independent cell lines (mean 42‐fold) and showed that TMS provided significant inhibition of enzyme function in differentiated NHU cells (mean 14% of ITE‐induced activity, Figure [5](#mc22784-fig-0005){ref-type="fig"}A). {#mc22784-fig-0005} To establish whether urothelial CYP1 was capable of activating the known carcinogen BaP, cultures were exposed for 6 h and metabolites were measured in the culture medium by HPLC (Supplementary Figure S6). First, CYP1A1 oxidizes BaP to an epoxide (ie, BaP‐7,8‐epoxide), which is then converted to BaP‐7,8‐dihydrodiol by EPHX1. Further bioactivation by CYP1A1 leads to the reactive species, BaP‐7,8‐dihydrodiol‐9,10‐epoxide (BPDE) and BaP‐tetrol‐I‐1 is formed by spontaneous hydrolysis of BPDE. Peaks, referenced to standards, were observed and quantified for BaP‐7,8‐dihydrodiol (Figure [5](#mc22784-fig-0005){ref-type="fig"}B) and BaP‐tetrol‐I‐1 (Figure [5](#mc22784-fig-0005){ref-type="fig"}C). The formation of BaP metabolites mirrored the trends observed for EROD‐activity (Figure [5](#mc22784-fig-0005){ref-type="fig"}A); with metabolism greatest in differentiated NHU cell cultures, inducible by pre‐treatment with ITE (2.7 and 4.1‐fold for dihydrodiol and tetrol, respectively) and inhibited by TMS (to 46% and 35% for dihydrodiol and tetrol, respectively; Figures [5](#mc22784-fig-0005){ref-type="fig"}B and [5](#mc22784-fig-0005){ref-type="fig"}C). The presence of BaP‐tetrol‐I‐1 in the medium of NHU cell cultures (Figure [5](#mc22784-fig-0005){ref-type="fig"}C) suggested that BPDE might also be forming the pre‐mutagenic DNA adducts (ie, 10‐(deoxyguanosin‐*N* ^2^‐yl)--7,8,9‐trihydroxy‐7,8,9,10‐tetrahydrobenzo\[*a*\]pyrene; "dG‐*N* ^2^‐BPDE" that have previously been observed in vitro and in vivo.[14](#mc22784-bib-0014){ref-type="ref"} Indeed the formation of dG‐*N* ^2^‐BPDE adducts was confirmed by ^32^P‐postlabeling (Supplementary Figure S7). No DNA adducts were detected in control (untreated) samples. Pre‐treatment of differentiated NHU cells with 1 μM ITE significantly increased BaP‐DNA adduct levels compared to controls (3.7‐fold) and 12 μM TMS significantly reduced this (to 21% of induced levels; Figure [5](#mc22784-fig-0005){ref-type="fig"}D). While the changes were not statistically significant, dG‐*N* ^2^‐BPDE adducts were formed in undifferentiated NHU cell cultures and adduct levels were both increased by ITE pre‐treatment and inhibited by TMS (Figure [5](#mc22784-fig-0005){ref-type="fig"}D). The correlation between EROD‐activity and BaP‐DNA adduct levels in the same three cell lines had an *R* ^2^ = 0.757 (Supplementary Figure S8). 3.7. Xenobiotic metabolism in bladder cancer {#mc22784-sec-0020} -------------------------------------------- Immunoperoxidase labeling of MIBC showed significant (*P *\< 0.05) reduction in mean POR expression when compared to normal non‐neoplastic bladder urothelium (Figure [6](#mc22784-fig-0006){ref-type="fig"} A). A sub‐group of MIBC (11.9% of tumors) was noted to over‐express POR, as compared to the normal bladders (open circles in Figure [6](#mc22784-fig-0006){ref-type="fig"}A). Both *CYP1A1* and *POR* transcript expression was significantly higher in luminal MIBC as compared with basal MIBC in The Cancer Genome Atlas cohort (Supplementary Figure S9). MIBCs were stratified into luminal and basal subgroups based on the GATA3 and cytokeratin 5/6 (KRT5/6) histology classifier first described by Dadhania et al[33](#mc22784-bib-0033){ref-type="ref"} that separates the two subtypes with 91% accuracy. The basal group had significantly less POR expression (*P *\< 0.05, Mann‐Whitney Test) and the luminal group contained all the POR over‐expressing tumors identified previously (16.6% of luminal tumors; Figures [6](#mc22784-fig-0006){ref-type="fig"}B and [6](#mc22784-fig-0006){ref-type="fig"}C). The basal (Supplementary Figure S10) POR^lo^ bladder cancer cell lines T24 and SCaBER were used as in vitro models of POR suppression in MIBC (Figures [6](#mc22784-fig-0006){ref-type="fig"}A and [6](#mc22784-fig-0006){ref-type="fig"}C) and showed no/low inducible CYP1 enzyme activity, respectively (Figure [6](#mc22784-fig-0006){ref-type="fig"}D) despite *CYP1A1* and *CYP1B1* transcripts being ITE‐inducible (Supplementary Figure S11). As a model of the POR over‐expressing tumors (Figures [6](#mc22784-fig-0006){ref-type="fig"}A and [6](#mc22784-fig-0006){ref-type="fig"}C), the well‐differentiated/luminal (Supplementary Figure S10) RT4 and RT112 bladder cancer cell lines showed 1 μM ITE‐inducible EROD‐activity (Figure [6](#mc22784-fig-0006){ref-type="fig"}D). Mean peak EROD activity at 16 h was 60.0, 89.3, and 83.3 nM resorufin/min/mg for differentiated NHU, RT4 and RT112 cells, respectively (Figure [6](#mc22784-fig-0006){ref-type="fig"}D). {ref-type="ref"}, reproducing the logistic regression analysis (LRA) and support vector machine (SVM) cut‐off lines derived in that study. All POR over‐expressing tumors (red circles) were classified as the more differentiated luminal type of MIBC. (C) Micrographs illustrating the POR immunoperoxidase labeling of normal bladder, MIBC, and cancer cell line (RT4, RT112, T24, and SCaBER) organ culture samples described in panel A. Scale bar represents 50 μm and applies to all images. (D) T24 and SCaBER cells generated peak EROD‐activity of 1.6 and 19.5 nM resorufin/min/mg, respectively following ITE exposure. By contrast, 1 μM ITE‐induced a peak EROD‐activity of 89.3 and 83.1 nM resorufin/min/mg in the POR over‐expressing RT4 and RT112 cell lines, respectively. RT4 and RT112 EROD was significantly higher than the mean of 60.0 nM resorufin/min/mg observed for differentiated NHU cells (*n* = 8 donors; each data point is the mean of six replicate cultures per donor)](MC-57-606-g006){#mc22784-fig-0006} 4. DISCUSSION {#mc22784-sec-0021} ============= This study provides experimental and clinical evidence that CYP‐activity by normal urothelium is reliant on the differentiation‐dependent expression of POR, thereby defining the CYP‐capacity of different neoplastic programmes. POR abundance has been shown to influence CYP2B6‐mediated bioactivation of cyclophosphamide in patients[13](#mc22784-bib-0013){ref-type="ref"} and total CYP‐mediated metabolism in mice,[14](#mc22784-bib-0014){ref-type="ref"} and combined with this study suggests POR can be used as a biomarker of total CYP‐capacity in tissues. MIBC showed an overall suppression of POR; this was exemplified by the basal‐type T24 and SCaBER cell lines, which showed no EROD‐activity even though *CYP1* transcripts were inducible (Supplementary Figure S11), suggesting an overall loss of functional CYP‐activity in basal MIBC. By contrast, a subset of MIBC was identified that over‐expressed POR relative to normal bladder urothelium. In our series, this subgroup defined 16.6% of the luminal tumors. Using RT4 and RT112 cells as representative luminal POR over‐expressing bladder cancer cell lines, it was confirmed that these cells showed greater induced peak metabolic activity than achieved by differentiated NHU cells. These observations are important in associating functional, drug‐metabolizing activity to histologically‐defined tumor sub‐groups. Several CYP1‐activated therapies are in development to target epithelial tumors[35](#mc22784-bib-0035){ref-type="ref"}, [36](#mc22784-bib-0036){ref-type="ref"}, [37](#mc22784-bib-0037){ref-type="ref"} and POR is thought to be critical to the activation of some hypoxia‐activated prodrugs,[38](#mc22784-bib-0038){ref-type="ref"} offering potential for future trials targeting the POR^hi^ group of MIBC patients we report here. The evidence presented here for CYP1A1 function in human urothelium and its role in BaP metabolism builds on earlier work showing BaP metabolism by organ cultures of human bladder tissue.[39](#mc22784-bib-0039){ref-type="ref"}, [40](#mc22784-bib-0040){ref-type="ref"} These earlier studies demonstrated that the tissue had capacity to metabolize BaP, but without confirming the cell type responsible due to the heterogeneous nature of the cell types present. In particular, it was noted that the bladder was the most active BaP metabolizer of all the human explant tissues tested at that point.[40](#mc22784-bib-0040){ref-type="ref"} Our study has shown *CYP1A1* and *CYP1B1* transcript expression by human urothelium and confirmed CYP1A1 protein, although no suitable antibody could be found for CYP1B1. Studies of purified enzymes show that CYP1A1 is more efficient at EROD (12‐fold) and BaP hydroxylation (2‐7‐fold) than CYP1B1.[34](#mc22784-bib-0034){ref-type="ref"}, [41](#mc22784-bib-0041){ref-type="ref"} Based on the greater induction of transcript in NHU cells and the greater efficiency of CYP1A1 to perform the reactions studied here, we believe that CYP1A1 is the critical enzyme in urothelial activation of BaP. The potential for activation of pro‐carcinogens by human urothelium observed in this study provides a local mechanism for SNPs previously linked to bladder cancer (*CYP1A1*, *CYP1B1*, and *ARNT* [5](#mc22784-bib-0005){ref-type="ref"}, [6](#mc22784-bib-0006){ref-type="ref"} and, more recently, *POR*).[16](#mc22784-bib-0016){ref-type="ref"} Despite strong evidence for the association of smoking and occupational BaP exposure with bladder cancer and the support in this study for urothelial activation of PAHs, the smoking‐associated mutational signature seen in lung cancer has not been observed in bladder cancer.[7](#mc22784-bib-0007){ref-type="ref"}, [42](#mc22784-bib-0042){ref-type="ref"} This may be due to the tissue‐specific nature of extra‐hepatic CYP activity (reviewed in Ref. [1](#mc22784-bib-0001){ref-type="ref"}) and the role of other CYPs that might metabolize BaP[11](#mc22784-bib-0011){ref-type="ref"} in human urothelium. This conclusion is supported by the efficacy of TMS inhibition which reduced EROD activity to 14%, BaP‐7,8‐dihydrodiol formation only to 46%, BaP‐tetrol formation to 35%, and BaP‐DNA adduct formation to 21% relative to ITE‐induced cells; suggesting that other enzymes not inhibited by TMS may play a role in BaP metabolism by the urothelium. Natural AHR ligands, such as the tryptophan metabolites indigo and indirubin (which are structurally similar to ITE), have been detected in urine from healthy patients[43](#mc22784-bib-0043){ref-type="ref"} making them potential drivers of urothelial CYP1 activity. It is therefore theoretically possible that the coincidence of endogenous AHR ligands with urinary pro‐carcinogens in the bladder might contribute to accelerated carcinogenesis in some patients No evidence was found to support a role for AHR in urothelial cytodifferentiation, which is known to be primarily driven by peroxisome proliferator‐activated receptor γ (PPARγ).[44](#mc22784-bib-0044){ref-type="ref"} By contrast, adipocyte differentiation is also mediated by PPARγ and during that process AHR expression is lost[45](#mc22784-bib-0045){ref-type="ref"} suggesting a functional maintenance of AHR by differentiated urothelium. To what extent urothelial (rather than hepatic) metabolism generates the mutagens that drive cancer initiation remains to be established in future studies. However, the temptation to resort to in vivo studies is flawed by the poor homology between human CYP1A1 protein and the rat/mouse/pig orthologs in the BaP interacting region (80.6%, 80.9%, and 82.5%, respectively; Supplementary Figure S12). In particular, Asn‐222 of *CYP1A1* has evolved as a negatively charged aspartic acid in rats, mice, and pigs. Asn‐222 lies centrally in the five‐residue disruption of helix F that is unique to CYP1A1 and appears to modulate substrate movement, binding, and orientation.[46](#mc22784-bib-0046){ref-type="ref"} In addition, Asn‐222 is thought to be within 5Å of BaP when bound by CYP1A1 and involved in an extensive hydrogen bonding network that stabilizes the binding pocket.[46](#mc22784-bib-0046){ref-type="ref"} Taken together with the differences in rodent urinary tract physiology, with short duration of urinary storage, differences in urothelial cell cycle regulation and a low threshold for cancer initiation,[47](#mc22784-bib-0047){ref-type="ref"} this limits the validity of extrapolating rodent in vivo studies to humans. In this study, a more direct human line of evidence was pursued from 2D cultures of normal and malignant urothelium, via organoids to primary MIBC, offering an alternative route to the in vitro‐in vivo paradigm. 5. CONCLUSION {#mc22784-sec-0022} ============= It has been thought for decades that the metabolism of pro‐carcinogens occurred in the liver and that bladder cancer was caused by the hydrolysis of conjugated metabolites in the urine or by enzymatic deconjugation in the urothelium. This study demonstrates the capacity of functionally‐differentiated normal human urothelium to activate the pro‐carcinogen BaP locally to active intermediates capable of forming DNA adducts (ie, dG‐*N* ^2^‐BPDE). The relative contributions of hepatic and urothelial metabolism to carcinogen activation should be re‐evaluated to better understand their relative roles in bladder cancer initiation. Furthermore, the association between differentiation and xenobiotic metabolism is maintained in a sub‐group of POR‐overexpressing luminal MIBC of predicted high metabolic potential, suggesting these patients as candidates for prodrug therapies. AUTHORS\' CONTRIBUTION {#mc22784-sec-0024} ====================== SB, VA, PR, DP, and JS were involved in the conception and design of the study. SB, VA, RI, WO, and MJ acquired the data. All authors were involved in drafting and approving the final manuscript. Supporting information ====================== Additional Supporting Information may be found online in the supporting information tab for this article. ###### **Figure S1**. Full western blots for Figure 4A&B. **Figure S2**. Full western blots for Figure 4A&C. **Figure S3**. Representative images of immunoperoxidase labelling for AHR and the urothelial barrier proteins, claudin 5 and uroplakin 3a. **Figure S4**. Differential regulation of *CYP1A1* and *CYP1B1* transcripts by AHR stimulation of NHU cells in different states. **Figure S5**. (A) RT‐qPCR demonstrated that 6 h 2 μM BaP exposure did not induce CYP1A1 or CYP1B1 transcripts in undifferentiated NHU cells but did significantly in differentiated cultures (p\<0.01 and p\<0.05 for CYP1A1 and CYP1B1, respectively; ANOVA with Tukey‐Kramer post‐tests). **Figure S6**. Representative HPLC chromatograms of the BaP metabolites observed in medium samples exposed to NHU cells. **Figure S7**. Autoradiographic profiles of DNA adducts, measured by ^32^P‐postlabelling, in undifferentiated and differentiated of NHU cells treated with BaP. **Figure S8**. Linear regression analysis of EROD activity (Figure 5A) and dG‐N2‐BPDE adducts (Figure 5D) showing an R2 of 0.757. **Figure S9**. CYP1A1 and POR transcript expression quantified from RNA sequencing data of the The Cancer Genome Atlas consortium and separated into basal and luminal subtypes based on the gene classifier reported by Choi et al. (55). **Figure S10**. Heatmap of "UBC‐40" bladder cancer cell line gene array data (54) for the relative expression of Choi et al. (55) gene classifiers by RT4, RT112, T24 and SCaBER cells. **Figure S11**. RT‐qPCR demonstrated that 24 h 1 μM ITE exposure induced CYP1A1 and CYP1B1 transcript in both RT4 and T24 cells. **Figure S12**. Clustal Omega alignment of Ensembl protein sequences for the BaP interacting region of CYP1A1 in human (amino acids 115‐496 \[49\]; ENST00000379727.7), CL57BL6 mouse (ENSMUST00000216433.1), rat (ENSRNOT00000026473.4) and pig (ENSSSCT00000002135.3). ###### Click here for additional data file. ###### **Table S1**. Clinical characteristics and histological assessment of MIBC specimens included in the study. ###### Click here for additional data file. The authors would like to thank Dr Andrew Leach (University of York) for his help with the EROD assays. The authors would also like to acknowledge the critical support of urology colleagues for providing tissue samples. Research at the University of York was supported by the Wellcome Trust (092430/Z/10/Z and partly C2D2:105624), the Royal Society (JP101086) and the Astellas European Foundation Grant 2010. JS and SCB are funded by York Against Cancer. Work at King\'s College London was supported by Cancer Research UK (Grant C313/A14329), Wellcome Trust (Grants 101126/Z/13/Z and 101126/B/13/Z), Natural Environment Research Council (Grant NE/L006782/1) and in part by the National Institute for Health Research Health Protection Research Unit (NIHR HPRU) in Health Impact of Environmental Hazards at King\'s College London in partnership with Public Health England (PHE). Work of members of Charles University was supported by GACR (Grant 17‐12816S). The views expressed are those of the authors and not necessarily those of the NHS, the NIHR, the Department of Health or PHE.

Osteoporosis is a common skeletal disease, affecting millions of individuals worldwide, and the prevention of fractures is an important public health goal. Currently, the osteoporosis treatments that are used substantially reduce vertebral fracture risk, whereas nonvertebral fracture risk has only moderately been improved by the osteoporosis drugs used, defining an unmet medical need ([@B1], [@B2]). Although cortical bone mass and composition are major determinants of bone strength and fracture risk in humans ([@B1], [@B3], [@B4]), only a limited number of studies have focused on the cellular and molecular mechanism specifically regulating cortical bone mass ([@B2]). However, it is known that the periosteal surface of the cortical bone is regulated by growth hormone (GH), insulin-like growth factor I (IGF-I), and periostin ([@B5][@B6][@B7]--[@B8]). Wingless-type MMTV integration site family members (WNTs) belong to a family of secreted cysteine-rich glycoproteins that signal *via* either the WNT/β-catenin pathway, also termed the canonical WNT pathway, or the noncanonical WNT pathways. A crucial step for WNT signaling is the interaction between WNTs and their transmembrane receptor proteins frizzled. When WNT proteins are synthesized, an acyl group from palmitoleic acid is attached at a serine amino acid residue ([@B9]). This lipid modification of WNT is necessary for its function because the frizzled proteins contain a cleft that specifically interacts with the acyl group. Recent *in vitro* studies have demonstrated that an extracellular enzyme (secreted lipase) called NOTUM renders WNT inactive by specifically removing the acyl group from WNT ([@B9], [@B10]). It has been described that NOTUM is a circulating protein, but the relative importance of circulating *vs.* local NOTUM is unknown ([@B11]). In a previous report, lifelong NOTUM inactivation in mice, using the less precise gene trapping technique, resulted in developmental defects involving the kidneys (unilateral agenesis) and teeth (dental dysplasia) as well as reduced embryonic viability ([@B12]). Referring to this mouse model of NOTUM inactivation, developed in a large-scale screen of multiple mouse models, it was recently reported that the mice surviving to adulthood had increased cortical bone mass, but the underlying mechanisms for the skeletal phenotype in these mice were not evaluated ([@B13]). Because several WNTs and modulators of WNT activity are major regulators of bone mass ([@B1], [@B4], [@B14], [@B15]), we hypothesized that NOTUM might modulate bone mass *via* an inhibition of WNT activity. To characterize the possible role of endogenous NOTUM as a physiologic modulator of bone mass, we developed global, cell-specific, and inducible *Notum*-inactivated mouse models using homologous recombination, allowing precise manipulation of genetic sequences in the mouse. In addition, we determined the cellular and molecular effects of *Notum* using cultured osteoblasts and osteoclasts. Finally, we aimed to determine if genetic variation in the *NOTUM* locus is associated with bone mass in humans. MATERIALS AND METHODS {#s1} ===================== Generation of global *Notum^−/−^ and Notum^+/−^* mice {#s2} ----------------------------------------------------- The global *Notum^−/−^* and *Notum^+/−^* mice were generated by breeding male *Notum^flox/flox^* mice (kindly provided by Prof. Christine Perret, INSERM, Institut Cochin) ([@B16]), with female mice expressing Cre recombinase ubiquitously and from an early embryonic stage under the control of the phosphoglycerate kinase-1 promoter (PGKcre) ([@B17]). In PGKcre-expressing female mice, the Cre activity starts in the diploid phase of oogenesis, and thereby a complete recombination of *LoxP* sites occurs also in *Cre^−^* offspring. In the *Notum^flox^* mice, *LoxP* sites are introduced upstream from *Notum* exon 2 and downstream from *Notum* exon 8 and in the presence of an active Cre recombinase, the DNA fragment from exon 2 to exon 8 of the *Notum* gene is excised ([@B16]). Presence or absence of the *Notum^flox^* allele was determined by use of multiplex real-time PCR analysis (StepOnePlus Real-Time PCR System, Thermo Fisher Scientific, Waltham, MA, USA). The following primers and carboxyfluorescein-labeled probe were used for detection of the *Notum^flox^* allele: forward (5′-CCATGATCCTGTGCCTTTCT-3′), reverse (5′-CGACGCGTGAAGTTCCTATT-3′), and probe (5′-GGCCGCGAAGTTCCTATACT-3′), whereas the following primers and VIC-labeled probe were used for detection of the *Notum*^*+*^ allele: forward (5′-CCATGATCCTGTGCCTTTCT-3′), reverse (5′-AGTGGCATCGGAGACAAATC-3′), and probe (5′-TTGGCAGCCCCAAAATATAG-3′). Genotyping for the presence of PGKcre was done using forward (5′-AACATGCTTCATCGTCGG-3′) and primer (5′-TTCGGATCATCAGCTACACC-3′). Briefly, male *Notum^flox/flox^* mice were bred with PGKcre-expressing female mice generating *Notum^+/−^* offspring. To generate *Notum^−/−^, Notum^+/−^*, and littermate wild-type (WT) control mice for the experiment, *Notum^+/−^* female and male mice were mated. Male and female mice included in the experiment were euthanized at 15 wk of age. The degree of *Notum* deletion at the DNA level in different tissues was determined by multiplex real-time PCR analysis using the following primers and carboxyfluorescein-labeled probe: forward (5′-TGGCTTCCCAGCCCATTAAG-3′), reverse (5′-GGCATGCTACAGACAGGTTG-3′), and probe (5′-AGTCTCACCACCCTGCCCTGT-3′). The following primers and VIC-labeled probe were used as a reference because they detect a sequence present in all mice: forward (5′-AGGTCTCAGACGGCAATCAT-3′), reverse (5′-CCTCCACTGCCTGCTAAGAT-3′), and probe (5′-ACCCAGCATGGTGAGGCCTGCC-3′). Generation of *Runx2cre-Notum^flox/flox^* mice {#s3} ---------------------------------------------- To generate specific inactivation of *Notum* in the osteoblast lineage, *Notum^flox/flox^* mice were bred with the previously described *Runx2-cre* mice ([@B2], [@B18]). These *Runx2-creNotum*^*flox/flox*^ mice display early osteoblast-specific *Cre* expression and have the capacity to recombine *LoxP*-flanked DNA sequences in an early osteoblast-specific manner, and an efficient recombination occurs at all sites of endochondral and intramembranous bone formation (particularly in periosteal cells, osteoblasts, and osteocytes but not osteoclasts) when *Runx2-cre* mice were crossed to a Rosa26 reporter strain ([@B18]). To generate mice depleted of *Notum* in osteoblasts, female mice with 2 floxed *Notum* alleles (*Notum^flox/flox^*) were mated with male mice having 1 floxed *Notum* allele and 1 allele of *Runx2-cre* (*Runx2-creNotum^+/flox^*). The littermate control mice were *Notum^flox/flox^*. *Runx2-cre* mice have an unchanged skeletal phenotype compared to WT mice ([@B2]). Genotyping of *Notum^flox^* mice was performed as described above, whereas the presence of the *Runx2-cre* gene was determined using a 3-primer strategy. The common 5′ primer (5′-CCAGGAAGACTGCAAGAAGG-3′) binds to the *Runx2* promoter of the endogenous gene and the transgene. The 3′ primer (5′-TGGCTTGCAGGTACAGGAG-3′) binds to the *Cre* sequence, and another 3′ primer (5′-GGAGCTGCCGAGTCAATAAC-3′) binds to the endogenous *Runx2*. Male and female *Runx2-creNotum^flox/flox^* mice included in the experiment were euthanized at 17 wk of age. Generation of mice with inducible *Notum* inactivation {#s4} ------------------------------------------------------ In order to study the acute adult effects of *Notum*, avoiding confounding developmental effects, inducible *Notum* knockout mice were created by breeding the *Notum^flox/flox^* mice with the previously described B6.Cg-Tg(CAG-Cre/Esr1\*)5Amc/J (CAGGCre-ER, 004682; The Jackson Laboratory, Bar Harbor, ME, USA) transgenic mice ([@B19]). Briefly, a male transgenic mouse expressing a tamoxifen-inducible Cre-mediated recombinase, and a *Notum*^+/flox^ allele (hereafter called *CAGGCre-ER-Notum^+/flox^*) were bred with female *Notum^flox/flox^* mice, creating *CAGGCre-ER-Notum^flox/flox^* mice. *Notum^flox/flox^* littermates were used as controls. Genotyping of the *Notum^flox/flox^* mice was done as previously described, and the presence of the Cre construct was determined using the following primers: Cre forward: 5′-GCGGTCTGGCAGTAAAAACTATC-3′; Cre reverse: 5′-GTGAAACAGCATTGCTGTCACTT-3′; WT forward: 5′-CTAGGCCACAGAATTGAAAGATCT-3′; and WT reverse: 5′-GTAGGTGGAAATTCTAGCATCATCC-3′. To study the adult effects of NOTUM on cortical bone thickness and bone strength, avoiding confounding developmental effects, a long-term study using female *CAGGCre-ER-Notum^flox/flox^* mice was performed. Female 11-wk-old *CAGGCre-ER-Notum^flox/flox^* and *Notum^flox/flox^* littermate control mice were injected with tamoxifen (50 mg/kg; T5648; MilliporeSigma, Burlington, MA, USA) for 3 consecutive days to activate the Cre construct. The mice were euthanized 31 d after the first tamoxifen injection at the age of 15 wk. To mechanistically study the importance of NOTUM in a short-term study, 20-wk-old male *CAGGCre-ER-Notum^flox^*^/flox^ and *Notum^flox/flox^* littermate control mice were injected with tamoxifen for 3 consecutive days. The mice were euthanized 11 d after the first tamoxifen injection at the age of 22 wk. The experimental procedures involving animals were approved by the Ethics Committee at the University of Gothenburg, and the care of the animals was in accordance with its guidelines. The mice were housed in a standard animal housing facility with 12-h dark/light periods. The temperature was controlled (22°C), and food and water were available *ad libitum*. At termination, the animals were anesthetized with Ketador/Dexdomitor (Richter Pharma, Wels, Austria/Orion Oyj, Espoo, Finland), bled, and euthanized by cervical dislocation. Long bones and vertebrae were dissected and stored for future analyses, and soft tissues were collected, weighed, and snap-frozen in liquid nitrogen. Assessment of murine bone parameters {#s5} ------------------------------------ ### Dual-energy X-ray absorptiometry {#s6} Total body and femur bone mineral density (BMD) were assessed using Lunar PIXImus densitometer (Wipro GE Healthcare, Chicago, IL, USA), or the UltraFocus^DXA^ (Faxitron Bioptics, Tucson, AZ, USA) ([@B2], [@B20]). ### High-resolution micro--computed tomography {#s7} High-resolution micro--computed tomography (μCT) was used to analyze distal femur, proximal tibia, and vertebrae L5 (Skyscan 1172; Bruker MicroCT, Aartselaar, Belgium) ([@B2]). The cortical region was analyzed in the diaphyseal region of femur and tibia, starting at a distance of ≈5.2 mm from the growth plate and continuing a further 134 μm in proximal direction for femur and in the distal direction for tibia. Trabecular bone was analyzed in the metaphyseal region of the femur and tibia, starting ≈650 μm from the growth plate and continuing a further 134 μm in proximal direction for femur and in distal direction for tibia. In the L5 vertebral body, the trabecular bone caudal of the pedicles was selected for analysis commencing at a distance of approx. 4.5 μm caudal of the lower end of the pedicles, and extending a further longitudinal distance of 230 μm in the caudal direction. Within the trabecular region, the trabecular bone was manually separated from the cortical bone by contouring the irregular anatomic structure adjacent to the endocortical boundary on a slice-by-slice basis. The grayscale µCT images were processed by a global threshold segmentation using the CTAn software (Bruker MicroCT) ([@B2]). ### Peripheral quantitative computed tomography {#s8} Computed tomography (CT) scans were performed using the peripheral quantitative computed tomography XCT Research M (v.4.5B; Norland Stratec, Pforzheim, Germany) with a voxel size of 70 μm, as previously described by Windahl *et al.* ([@B21]). To measure trabecular volumetric BMD, the scan was positioned in the metaphysis of the femur at a distance corresponding to 3.4% of the total femur length, proximal from the distal growth plate. This area contained trabecular and cortical bone, and the trabecular region was therefore defined as the inner 45% of the total cross-sectional area. The cortical bone parameters were analyzed in the middiaphyseal region of the femur, which is an area that contains only cortical bone. The threshold for cortical bone was set to 710 mg/cm^3^ ([@B22]). ### Mechanical strength {#s9} After dissection, the humerus was frozen at −20°C. Three-point bending was performed with a span length of 4.5 mm and a loading speed of 0.155 mm/s using an Instron 3366 (Instron, Norwood, MA, USA). Biomechanical parameters, based on the recorded load deformation curves, were calculated from Bluehill 2 software v.2.6 (Instron) with custom-made Excel (Microsoft, Redmond, WA, USA) macros ([@B2]). ### Bone histomorphometry {#s10} For dynamic histomorphometry, the mice were intraperitoneally injected with the fluorochromes calcein and alizarin (Merck GmbH, Darmstadt, Germany) 9 and 2 d prior to termination, respectively ([@B2]). Upon dissection, femur and tibia were fixated in 4% formaldehyde, dehydrated in 70% EtOH, and imbedded in methyl methacrylate. The femur and tibia were cut in 200-μm-thick transverse cross-sectional sections in the middiaphyseal region (for dynamic cortical histomorphometry), as well as in 4-μm-thick longitudinally coronary sections (for static histomorphometry). The analyses of dynamic cortical bone parameters were performed without any staining. For analyses of cellular parameters, the 4-µm-thick sections were stained with Masson-Goldner trichrome. Osteoclasts were defined in Masson-Goldner trichrome--stained sections as bone-resorbing multinucleated cells (2 or more nuclei) in contact with cortical bone surface. The bone-resorbing nature was identified using the shape of the cell (compared to other endosteal cells, including flat mononuclear bone lining cells and cuboidal mononuclear osteoblasts) and the possible presence of eroded surface ([@B23], [@B24]). All parameters were measured using the OsteoMeasure Histomorphometry Analysis System software v.2.2 (Osteometrics, Atlanta, GA, USA), following the guidelines of the American Society for Bone and Mineral Research ([@B25]). Measurement of serum markers {#s11} ---------------------------- To assess bone resorption, serum levels of C-terminal type I collagen (CTX) fragments were measured using an ELISA RatLaps Kit (Immunodiagnostic Systems, East Boldon, United Kingdom). Serum levels of procollagen type I N-terminal propeptide were measured using a Rat/Mouse EIA Kit (Immunodiagnostic Systems) as a marker of bone formation. Cell cultures {#s12} ------------- ### Silencing of Notum in MC3T3-E1 osteoblasts {#s13} Gene expression analyses in primary calvarial cell cultures from WT mice showed that expression of *Notum* mRNA was down-regulated during culture to an expression level similar to that in cells from *Runx2-creNotum^flox/flox^* mice ([Supplemental Fig. S1*A*](#SM1){ref-type="supplementary-material"}). The mRNA expression of *Notum* was also decreased in primary cultures of osteoblasts from long bones ([Supplemental Fig. S1*B*](#SM1){ref-type="supplementary-material"}). This finding, together with the fact that these cells are a mixture of *Runx2*-expressing cells and other cell types present in the bones used for isolation, made us select the murine preosteoblastic cell line MC3T3-E1 for studies on the role of endogenous NOTUM in osteoblast differentiation. An important observation for this selection was the finding that the *Notum* mRNA expression in the MC3T3-E1 cells was not decreased in cell cultures over 11 d ([Supplemental Fig. S2](#SM1){ref-type="supplementary-material"}), in contrast to primary bone cell cultures. MC3T3-E1 (CRL-2593, subclone 4; American Type Culture Collection, Manassas, VA, USA) was maintained in α-minimum essential medium (α-MEM, M0450; Thermo Fisher Scientific; MilliporeSigma) supplemented with 10% heat inactivated fetal bovine serum (16000-044; Thermo Fisher Scientific), 2 mM L-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin. For experiments, cells were seeded at an initial density of 10,000 cells in 500 μl of culture medium in 48-well culture dishes (Corning, Corning, NY, USA). To induce osteogenic differentiation, osteogenic medium was prepared by further supplementation with 50 μg/ml [l]{.smallcaps}-ascorbic acid and 4 mM β-glycerophosphate (MilliporeSigma). Medium was replaced every 2 d for the duration of all experiments. All cells were maintained at 37°C in a humidified 5% CO~2~ environment. To assess the role of NOTUM in osteoblast differentiation, *Notum* mRNA was silenced using a specific small interfering RNA (siRNA) sequence (Invitrogen Silencer Predesigned siRNA 4390771, S95064; Thermo Fisher Scientific). A scrambled sequence (AM4635) was used as negative control. Twenty-four and 72 h after seeding, the cells were transfected with 30 nM of the siRNA using lipofectamine RNAiMax (Thermo Fisher Scientific) in serum-free OptiMEM (Thermo Fisher Scientific). Cells were harvested 24 h after the first and second transfection for gene expression analyses. For quantification of alkaline phosphatase (ALP) activity and ALP staining, cells were exposed to osteogenic media 24 h after the second silencing, with an additional silencing 3 d later. Seven days after exposure to osteogenic media, cells were stained for ALP and enzyme activity was analyzed quantitatively. ### ALP analyses {#s14} At the end of osteogenic treatment, the cells were fixed with 70% ethanol for 10 min and stained for ALPase activity using a commercial kit (86R; MilliporeSigma) following the manufacturer's instructions and photographed. For quantification of ALPase activity, proteins from another plate were extracted using 200 μl of ultrapure water and exposed to 5 thermal shock cycles for 20 min at −20°C and 15 min at 37°C. Proteins were quantified using Bicinchoninic Acid Assay Kit (23235; Thermo Fisher Scientific). For ALPase activity measurement, a commercial kit based on the ability of ALP to hydrolyze thymolphthalein monophosphate releasing thymolphthalein was used (Labtest Diagnostica, Lagoa Santa, Brazil). Briefly, 10 μg of proteins were used in a final volume of 80 μl of the reaction buffer (pH 10.1) containing 2.75 mM of thymolphthalein monophosphate and incubated at 37°C for 90 min in a 96-well ELISA plate (Corning). After the reaction period, 200 μl of color reagent was added to each well and the absorbance was measured at 580 nM using a spectrophotometer (Epoch; Biotek, Winooski, VT, USA). ### Primary bone cell cultures {#s15} Primary calvarial periosteal bone cells from 3- to 5-d-old C57BL/6N *Notum*^*flox/flox*^ and *Runx2-creNotum*^*flox/flox*^ mice were isolated by sequential enzymatic digestion as previously described in refs. [@B26] and [@B27]. Isolated periosteal bone cells were cultured in complete α-MEM (22561-021; Thermo Fisher Scientific) supplemented with 10% heat inactivated fetal bovine serum (F7524; MilliporeSigma), 2 mM GlutaMax (35050-038; Thermo Fisher Scientific), 50 µg/ml gentamicin (15750-037; Thermo Fisher Scientific), 100 U/ml penicillin and 100 µg/ml streptomycin (15140-148; Thermo Fisher Scientific) for 3--5 d prior to experiment start. At the start of the experiment, cells were seeded at 20,000 cells/cm^2^ in 48 well plates and incubated in osteogenic media \[complete α-MEM supplemented with 10 mM β-glycerophosphate disodium salt hydrate (BGP, G9422; MilliporeSigma) and 0.2 mM [l]{.smallcaps}-ascorbic acid 2-phosphate sesquimagnesium salt hydrate (Asc-2P, A8960; MilliporeSigma)\]. Culture media were replenished after 4 d and cells were harvested for RNA analysis after 4 and 7 d of culture. Primary bone cell cultures from femur and tibia from 8- to 12-wk-old C57BL/6N mice were isolated by outgrowth of cells from collagenase-treated diaphyseal cortical bone as previously described by Bakker and Klein-Nulend ([@B27]). Primary bone cells from femur and tibia were isolated in complete α-MEM, and thereafter experiments were performed in osteogenic medium as for calvarial bone cells above. ### Primary osteoclast cell cultures {#s16} Bone marrow cells from 8- to 12-wk-old *Notum^flox/flox^* and *Runx2-creNotum^flox/flox^* mice were cultured in suspension culture discs (Corning) in complete α-MEM with 30 ng/ml macrophage colony-stimulating factor (M-CSF) (416-ML-050; R&D Systems, Minneapolis, MN, USA) for 2 d, and the adherent bone marrow--derived macrophages (BMMs) were used as osteoclast progenitors ([@B28], [@B29]). BMMs were detached and spot seeded in 24-well plates (40,000 cells/well) and cultured in complete α-MEM with 30 ng/ml M-CSF with or without the addition of 4 ng/ml receptor activator of nuclear factor κ-B ligand (RANKL) (462-TEC; R&D Systems) to induce osteoclast differentiation. The medium was changed after 3 d and cells harvested for RNA isolation by lysis in RLT buffer (Qiagen, Germantown, MD, USA) 4 d after seeding. ### Treatment of osteoclasts with recombinant NOTUM {#s17} Osteoclasts were differentiated from BMMs as described above. Cells were cultured in M-CSF and RANKL with or without the addition of 10 or 100 ng/ml recombinant human NOTUM (9118-NM; R&D Systems) for 3--4 d before staining for expression of tartrate-resistant acid phosphatase (TRAP) using a commercial staining kit (387A; MilliporeSigma) or lysis in RLT buffer for RNA purification. To study osteoclast formation and activity on bone, 20,000 BMMs were seeded on bovine bone discs (IDS Nordic, Copenhagen, Denmark) in 96-well plates. Media were changed and saved for analysis every 3--4 d, and staining for TRAP was performed after 10 d. The total amount of TRAP5b in the culture media were analyzed as a measure of osteoclastogenesis using a commercial kit (IDS Nordic). Resorptive activity, as assessed by the release of C-terminal telopeptides of type I collagen (CTX-I) from the bone slices, was measured using commercial CTX ELISA (IDS Nordic). Lipase activity measurement of recombinant NOTUM {#s18} ------------------------------------------------ Functional enzyme activity of recombinant human NOTUM was tested using the fluorogenic substrate 8-Octanoyloxypyrene-1-3-6-trisulfonic acid trisodium salt (74875; MilliporeSigma). Recombinant NOTUM (50 ng) dissolved in 0.1% bovine serum albumin in PBS (25 µl) was mixed with an equal volume of 5 mM 8-Octanoyloxypyrene-1-3-6-trisulfonic acid trisodium salt in 2 times reaction buffer (10 mM CaCl~2~, 1 mM MgCl~2~, 50 mM TrisHCl, pH 7.4). Fluorescence intensity at 535 nm was measured after 10, 20, and 30 min using a SpectraMax i3 (Molecular Devices, Sunnyvale, CA, USA). Wheat germ lipase (L3001, 250 ng; MilliporeSigma) was used as positive control and 0.1% bovine serum albumin in PBS was used as blank. Real-time quantitative PCR {#s19} -------------------------- Total mRNA was prepared from bone using Trizol reagent (15596018; Thermo Fisher Scientific) and RNeasy Mini Kit (74106; Qiagen). For preparation of total mRNA from cultured cells, the RNeasy Micro Kit (74004; Qiagen) or Total RNA Purification Kit (DPK-108L; Cellco, Basking Ridge, NJ, USA) were used. The mRNA was reversed transcribed into cDNA (4368814; Thermo Fisher Scientific), and real-time PCR analyses were performed using the StepOnePlus Real-Time PCR System (Thermo Fisher Scientific) and Assay-on-Demand primer and probe sets (*Notum*: Mm01253273_m1, *Alpl*: Mm00475834_m1, *Acp5 (Trap)*: Mm00475698_m1, *Nfatc1:* Mm00479445_m1, *Ctsk:* Mm00484036_m1; Thermo Fisher Scientific). The expression of each gene was adjusted to 18S ribosomal subunit (4310893E; Thermo Fisher Scientific) or *Actb* custom-made primers: forward (5′-GGACCTGACGGACTACCTCATG-3′), reverse (5′-TCTTTGATGTCACGCACGATTT-3′), used as an internal standard. All assays had 90--100% efficiency, thus the 2^−ΔΔ^*^Ct^* method was used to calculate the relative gene expression. Statistical analyses {#s20} -------------------- Values are given as means ± [sem]{.smallcaps}. The statistical difference between *Runx2-creNotum^flox/flox^* mice and *Notum^flox/flox^* mice, between *CAGGCre-ER-Notum^flox/flox^* and *Notum^flox/flox^* mice, as well as between *Notum^+/−^* mice and WT mice, were calculated using an unpaired, 2-tailed Student's *t* test. Statistical differences between cell culture treatment groups were calculated using an unpaired, 2-tailed Student's *t* test when there were 2 groups (cells from 2 mouse strains or 2 treatment groups) and using 1-way ANOVA and Tukey's multiple comparison test when there were more than 2 treatment groups. A difference was considered statistically significant if *P* \< 0.05. Large-scale human genetic analyses using the UK Biobank {#s21} ------------------------------------------------------- In 2006--2010, the UK Biobank recruited 502,647 individuals aged 37--76 yr from across the United Kingdom. All participants provided information regarding their health and lifestyle *via* touch screen questionnaires, consented to physical measurements, and agreed to have their health followed. They also provided blood for future analysis. UK Biobank has ethical approval from the Northwest Multicenter Research Ethics Committee, and informed consent was obtained from all participants. ### Estimated BMD using ultrasound {#s22} Quantitative ultrasound of the heel was used to obtain a noninvasive estimate of BMD that predicts fracture ([@B30], [@B31]). A Sahara Clinical Bone Sonometer (Hologic, Marlborough, MA, USA) was used for quantitative ultrasound assessment of calcanei in UK Biobank participants. Details of the complete protocol are publicly available on the UK Biobank website (*<https://www.ukbiobank.ac.uk/>*). Estimated BMD (eBMD; g/cm^2^) was derived as a linear combination of speed of sound (SOS in the following equation) and bone ultrasound attenuation (BUA in the following equation) \[eBMD = 0.0025926 × (BUA + SOS) − 3.687\] ([@B31]). ### Genotyping {#s23} Genotype data are available for ≈488,000 participants in the UK Biobank cohort. Genotyping was performed using the Affymetrix UK BiLEVE Axiom array on an initial set of ≈50,000 participants; the remaining ≈450,000 participants were genotyped using the Affymetrix UK Biobank Axiom array. The 2 arrays are extremely similar (with over 95% common content). Quality control and imputation \[to over 39 million Haplotype Reference Consortium single nucleotide polymorphisms (SNPs) including more than 8 million SNPs with a minor allele frequency \>1%\] were performed centrally by UK Biobank. IMPUTE2 (*<https://mathgen.stats.ox.ac.uk/impute/impute_v2.html#home>*) was used for imputation to the Haplotype Reference Consortium v.1.1 panel. ### Model used for genetic associations {#s24} The recently released version of BOLT-LMM Bayesian mixed model association method (*<https://www.biorxiv.org/content/early/2017/09/27/194944>*) was used for the associations between imputed SNPs and eBMD (on a BMD T-score scale). This method was developed to produce highly powered, robust test statistics when run on all European samples (retaining related individuals) in the UK Biobank. The model was adjusted for sex, age, and weight. The genome-wide association study of estimated ultrasound BMD using this methodology was recently made available for public download (*<https://data.broadinstitute.org/alkesgroup/UKBB/>*). For the present study, we used these downloaded data and searched for possible associations specifically between imputed SNPs (minor allele frequency (MAF) \>1%) in the *NOTUM* locus ±25 kb from the genes' start/end limits and eBMD. In total, 125 SNPs with a minor allele frequency \> 1% were tested, and we applied a conservative Bonferroni correction for these SNPs, setting a value of *P* = 0.05/125 = 4.0 × 10^−4^ as a significance threshold. RESULTS {#s25} ======= Heterozygous female *Notum*-inactivated mice have increased cortical bone thickness {#s26} ----------------------------------------------------------------------------------- To determine the role of NOTUM in the skeleton, we developed a mouse model with global *Notum* inactivation (*Notum*^−/−^). This model had exons 2--8 of the *Notum* gene deleted (*Notum*^−/−^; [**Fig. 1*A***](#F1){ref-type="fig"}). A majority of *Notum*^−/−^ mice (≈90%) died during embryonal development, and the few mice surviving until 13 wk of age were severely growth inhibited. However, heterozygous *Notum*^+/−^ mice had normal embryonic development and were born apparently healthy and had a normal body weight growth until 13 wk of age ([Fig. 1*B*](#F1){ref-type="fig"}). In addition, *Notum*^+/−^ mice displayed normal bone length of femur ([Fig. 1*C*](#F1){ref-type="fig"}) and normal weights of several visceral organs ([Fig. 1*D*](#F1){ref-type="fig"}). When comparing the *Notum* expression in different tissues, the highest expression in WT mice was observed for liver and diaphyseal cortical bone ([Fig. 1*E*](#F1){ref-type="fig"}). *Notum*^+/−^ mice had reduced *Notum* mRNA levels in both liver and bone compared with WT mice ([Fig. 1*E*](#F1){ref-type="fig"}), and no *Notum* expression was observed in the few surviving severely growth-inhibited *Notum*^−/−^ mice. As expected, the recombination of the *Notum* allele in liver and bone was complete in the *Notum*^−/−^ mice, and it was ≈50% in these 2 tissues of *Notum*^+/−^ mice ([Fig. 1*F*](#F1){ref-type="fig"}). ![Heterozygous female *Notum*-inactivated mice have increased cortical bone thickness. *A*) Schematic drawing of the global *Notum*-inactivated mouse model. *B*) Normal body weight in both male and female *Notum^+/−^* mice (*n* = 8) when comparing with WT mice (*n* = 7) at 4, 6, and 13 wk of age. *C*) *Notum^+/−^* mice (*n* = 8) have normal femur length compared to WT mice (*n* = 7). *D*) Soft tissue weights over body weight in male and female *Notum^+/−^* (*n* = 8) and WT (*n* = 7) mice. *E*) mRNA expression analyses of *Notum* in liver, cortical bone, vertebral body, spleen, and gonadal fat, as normalized to WT liver, in male *Notum^+/−^* (*n* = 8) and WT (*n* = 7) mice. *F*) Degree of deletion in purified DNA from male WT (*n* = 7), *Notum^+/−^* (*n* = 8), and *Notum^−/−^* (*n* = 2) mice in cortical bone and liver. *G*, *H*) Femur BMD (females, *G*) and total body BMD (*H*) as measured by DXA in *Notum^+/−^* (female, *n* = 8) and WT mice (female, *n* = 7). *I*--*M*) Cortical bone area (*I*), cortical thickness (*J*), mean polar moment of inertia (*K*), cortical porosity (*L*), and cortical volumetric BMD (*M*) of the femur as measured by μCT in female *Notum^+/−^* (*n* = 8) and WT (*n* = 7) mice. *N*) Maximal load (F max) of the humerus as measured by 3-point bending analyses in female *Notum^+/−^* (*n* = 8) and WT (*n* = 7) mice. *O*, *P*) Trabecular bone volume/tissue volume (BV/TV) of the femur (*O*) and of the vertebrae L5 (*P*) as measured by μCT in female *Notum^+/−^* (*n* = 8) and WT (*n* = 7) mice. Unless otherwise stated, the results refer to 13-wk-old mice. N.D., not detectable; Sem Ves, seminal vesicles. All values are given as means ± [sem]{.smallcaps}. (\*)*P* = 0.09, \**P* \< 0.05, \*\**P* \< 0.01, \*\*\**P* \< 0.001 *vs.* WT mice (Student's *t* test[)]{.smallcaps}.](fj.201900707Rf1){#F1} Dual-energy X-ray absorptiometry (DXA) analyses revealed that the total femur BMD was significantly increased in female *Notum*^+/−^ mice compared with WT mice ([Fig. 1*G*](#F1){ref-type="fig"}), and a similar nonsignificant trend was observed for total body BMD ([Fig. 1*H*](#F1){ref-type="fig"}). We next evaluated the cortical and trabecular bone compartments separately using high-resolution µCT analyses. The femur middiaphyseal cortical bone area ([Fig. 1*I*](#F1){ref-type="fig"}), cortical thickness ([Fig. 1*J*](#F1){ref-type="fig"}), and the estimated bone strength polar moment of inertia ([Fig. 1*K*](#F1){ref-type="fig"}) were significantly increased in female *Notum*^+/−^ mice compared with WT mice, whereas cortical porosity ([Fig. 1*L*](#F1){ref-type="fig"}) and cortical volumetric BMD ([Fig. 1*M*](#F1){ref-type="fig"}) were unaffected. Three-point bending analyses of the diaphyseal region of the humerus revealed increased cortical bone strength in female *Notum*^+/−^ mice compared with WT mice ([Fig. 1*N*](#F1){ref-type="fig"}). In contrast, female *Notum*^+/−^ mice had normal trabecular bone volume/tissue volume (BV/TV) fraction in both the distal metaphyseal region of femur ([Fig. 1*O*](#F1){ref-type="fig"}) and in the L5 vertebral body ([Fig. 1*P*](#F1){ref-type="fig"}). Thus, heterozygous female *Notum*-inactivated mice have specifically increased cortical bone thickness and bone strength. There was no significant bone phenotype in the male *Notum*^+/−^ mice compared with male WT mice ([Supplemental Table S1](#SM2){ref-type="supplementary-material"}). Osteoblast-derived NOTUM is the principal source of NOTUM in bone {#s27} ----------------------------------------------------------------- Because *Notum* expression was high in cortical bone, we hypothesized that the cortical bone phenotype in the heterozygote female *Notum*-inactivated mice might be caused by lack of *Notum* expression by osteoblast lineage cells in cortical bone. However, a role of other contributing cell types in the bone compartment or by circulating factors such as liver-derived *Notum* or other indirect systemic effects could not be excluded. To determine the source of NOTUM in cortical bone and the function of osteoblast-derived NOTUM *in vivo*, we generated a conditional *Notum*-inactivated mouse model targeting osteoblasts and osteocytes. A mouse model with exons 2--8 of *Notum* flanked by *LoxP* sites (*Notum*^*flox/flox*^,[**Fig. 2*A***](#F2){ref-type="fig"}) was used ([@B16]). To achieve inactivation of NOTUM early in the osteoblast lineage, we generated *Runx2-creNotum^flox/flox^* mice, which express Cre recombinase driven by the Runx2 promoter, a promoter that is expressed specifically in early osteoblast lineage cells but not in osteoclasts ([Fig. 2*A*](#F2){ref-type="fig"}) ([@B18]). *Runx2-creNotum^flox/flox^* mice had substantial and specific recombination of the *Notum^flox^* allele in bone ([Fig. 2*B*](#F2){ref-type="fig"}), resulting in 98 ± 0.4 and 79 ± 9.5% lower *Notum* mRNA levels in cortical bone of female and male mice, respectively, compared to corresponding *Notum^flox/flox^* mice ([Fig. 2*C*](#F2){ref-type="fig"}). Analyses of isolated calvarial cells, containing a large proportion of osteoblasts, showed that cells from *Runx2-creNotum^flox/flox^* mice had substantially lower levels of *Notum* mRNA expression compared with cells from *Notum^flox/flox^* mice ([Fig. 2*D*](#F2){ref-type="fig"}). Expression of *Notum* mRNA in bone marrow macrophages was close to the detection limit and was not affected by osteoblast lineage--specific *Notum* inactivation ([Fig. 2*D*](#F2){ref-type="fig"}). These findings demonstrate that osteoblast lineage--derived NOTUM is the principal source of NOTUM in cortical bone. ![Osteoblast-derived NOTUM is the principal source of NOTUM in bone. *A*) Schematic drawing of the conditional osteoblast lineage--specific *Notum*-inactivated mouse model. *B*) Degree of deletion of *Notum* DNA in cortical bone and liver in *Runx2-creNotum^flox/flox^* (*n* = 15) and *Notum^flox/flox^* (*n* = 11) male mice. *C*) mRNA expression analyses of *Notum* in the cortical bone and liver in *Runx2-creNotum^flox/flox^* (females, *n* = 11; males, *n* = 15) and *Notum^flox/flox^* (females, *n* = 13; males, *n* = 11) mice. *D*) mRNA expression analyses of *Notum* in cultured calvarial osteoblasts (cOBL) and bone marrow macrophage--derived osteoclasts (BMMOCLs) from *Runx2-creNotum^flox/flox^* and *Notum^flox/flox^* mice. (Representative experiment, *n* = 3 wells/cell type and mouse strain.) *E*) Normal body weight in *Runx2-creNotum^flox/flox^* mice (female, *n* = 11; male, *n* = 15) compared to *Notum^flox/flox^* mice (female, *n* = 13; male, *n* = 11) at 5, 12, and 17 wk of age. *F*) Normal femur length in *Runx2-creNotum^flox/flox^* mice (female, *n* = 11; male, *n* = 15) compared to *Notum^flox/flox^* mice (female, *n* = 13; male, *n* = 11). *G*) Soft tissue weights over body weight in *Runx2-creNotum^flox/flox^* mice (females, *n* = 11; males, *n* = 15) and *Notum^flox/flox^* mice (females, *n* = 13; males, *n* = 11). *H*) Total body BMD as measured by DXA in 5-wk-old and 17-wk-old *Runx2-creNotum^flox/flox^* mice (female, *n* = 11; male, *n* = 15) and *Notum^flox/flox^* mice (female, *n* = 13; male, *n* = 11). *I*) Femur BMD as measured by DXA in 17-wk-old *Runx2-creNotum^flox/flox^* mice (female, *n* = 11; male, *n* = 15) and *Notum^flox/flox^* mice (female, *n* = 13; male, *n* = 11). Unless otherwise stated, the results refer to 17-wk-old mice. All values are given as means ± [sem]{.smallcaps}. \**P* \< 0.05, \*\**P* \< 0.01 *vs.* *Notum^flox/flox^* mice (Student's *t* test).](fj.201900707Rf2){#F2} *Runx2-creNotum^flox/flox^* mice displayed normal embryonic development and were born apparently healthy and had a normal body weight growth through 17 wk of age ([Fig. 2*E*](#F2){ref-type="fig"}). In addition, *Runx2-creNotum^flox/flox^* mice had normal bone length of femur ([Fig. 2*F*](#F2){ref-type="fig"}) and normal weights of several visceral organs ([Fig. 2*G*](#F2){ref-type="fig"}). DXA analyses revealed that both *male and female Runx2-creNotum^flox/flox^* mice developed increased total body BMD during sexual maturation ([Fig. 2*H*](#F2){ref-type="fig"}). Total femur BMD was also significantly increased in both adult male and adult female *Runx2-creNotum^flox/flox^* mice compared with corresponding *Notum^flox/flox^* mice ([Fig. 2*I*](#F2){ref-type="fig"}). Thus, osteoblast-derived NOTUM is the principal source of NOTUM in bone and decreases total body BMD. Osteoblast-derived NOTUM reduces cortical but not trabecular bone mass {#s28} ---------------------------------------------------------------------- Detailed CT analyses of the cortical diaphyseal bone in femur revealed increased cortical bone area ([**Fig. 3*A***](#F3){ref-type="fig"}), cortical bone thickness ([Fig. 3*B*](#F3){ref-type="fig"}), and cortical polar moment of inertia ([Fig. 3*C*](#F3){ref-type="fig"}) in adult *Runx2-creNotum^flox/flox^* mice compared to control mice. However, neither the cortical porosity ([Fig. 3*D*](#F3){ref-type="fig"}) nor the cortical volumetric BMD ([Fig. 3*E*](#F3){ref-type="fig"}) were affected in the *Runx2-creNotum^flox/flox^* mice. Furthermore, the cortical thickness and cortical area of vertebral body L5 were increased in female *Runx2-creNotum^flox/flox^* mice compared to control mice and a similar nonsignificant trend (*P* = 0.09) was observed for the cortical bone area in male mice ([Supplemental Table S2](#SM2){ref-type="supplementary-material"}). The increased amount of cortical bone resulted in a substantial increase in mechanical strength (maximum load; [Fig. 3*F*](#F3){ref-type="fig"}) and whole-bone work-to-fracture ([Fig. 3*G*](#F3){ref-type="fig"}) of the cortical bone, whereas the structural stiffness ([Fig. 3*H*](#F3){ref-type="fig"}) was not significantly affected when evaluated by using 3-point bending in the diaphyseal region of the humerus. *Runx2-creNotum^flox/flox^* male mice had normal BV/TV in both the distal metaphyseal region of femur ([Fig. 3*I*](#F3){ref-type="fig"}) and in the L5 vertebral body ([Fig. 3*J*](#F3){ref-type="fig"}). *Runx2-creNotum^flox/flox^* female mice had also a normal BV/TV in the distal metaphyseal region of femur ([Fig. 3*I*](#F3){ref-type="fig"}) but a slightly decreased BV/TV in the L5 vertebral body ([Fig. 3*J*](#F3){ref-type="fig"}). The increased cortical thickness but not trabecular bone mass is a phenocopy, although slightly more pronounced, of the heterozygote female global *Notum*-inactivated mice. The increased cortical bone thickness in both the male and the female mice with osteoblast lineage--specific *Notum* inactivation was the result of an increased periosteal circumference ([Fig. 3*K*](#F3){ref-type="fig"}), whereas the endocortical circumference was not reduced ([Fig. 3*L*](#F3){ref-type="fig"}). However, cortical histomorphometric analyses performed at 17 wk of age did not identify any significant effects on bone formation rate or on numbers of osteoblasts or osteoclasts per bone surface ([Supplemental Table S3](#SM2){ref-type="supplementary-material"}). ![Osteoblast-derived NOTUM reduces cortical but not trabecular bone mass. *A*--*E*) Cortical bone area (*A*), cortical thickness (*B*), mean polar moment of inertia (*C*), cortical porosity (*D*), and cortical volumetric BMD (*E*) of the femur as measured by μCT in *Runx2-creNotum^flox/flox^* (female, *n* = 11; males, *n* = 15) and *Notum^flox/flox^* (females, *n* = 13; males, *n* = 11) mice. *F*--*H*) Maximal load (F max) (*F*), whole-bone work-to-fracture (*G*), and (*H*) structural stiffness (*H*) of the humerus as measured by 3-point bending in *Runx2-creNotum^flox/flox^* (female, *n* = 11; males, *n* = 15) and *Notum^flox/flox^* (females, *n* = 13; males, *n* = 11) mice. All biomechanical data are whole-bone data. *I*, *J*) Trabecular BV/TV in the femur (*I*) and vertebral body L5 (*J*) in *Runx2-creNotum^flox/flox^* mice (females, *n* = 11; males, *n* = 15) and *Notum^flox/flox^* mice (females, *n* = 13; males, *n* = 11). *K*, *L*) Periosteal (*K*) and endocortical circumference (*L*) of the femur as measured by µCT in *Runx2-creNotum^flox/flox^* (females, *n* = 11; males, *n* = 15) and *Notum^flox/flox^* (females, *n* = 13; males, *n* = 11) mice. Unless otherwise stated, the results refer to 17-wk-old mice. All values are given as means ± [sem]{.smallcaps}. \**P* \< 0.05, \*\**P* \< 0.01 *vs.* *Notum^flox/flox^* mice (Student's *t* test).](fj.201900707Rf3){#F3} Inducible inactivation of *Notum* increases cortical periosteal bone formation {#s29} ------------------------------------------------------------------------------ Although very informative, the studies using osteoblast lineage--specific *Notum* inactivation could not separate between developmental effects of NOTUM and its effects on adult bone metabolism. Therefore, to evaluate the effect of *Notum* specifically on adult cortical bone homeostasis and the underlying mechanisms, we developed a mouse model with tamoxifen-inducible *Notum* inactivation in adult mice. To this end, we bred *Notum^flox/flox^* mice with *CAGGCre-ER* transgenic mice ([@B19]) expressing a tamoxifen-inducible Cre-mediated recombination system ([**Fig. 4*A***](#F4){ref-type="fig"}). We have previously shown that *CAGGCre-ER* mice do not have a skeletal phenotype ([@B32]). All mice evaluated were treated with the same tamoxifen regime, using daily intraperitoneal injections during 3 consecutive days and the phenotype of *CAGGCre-ER-Notum^flox/flox^* and *Notum^flox/flox^* mice were compared ([Fig. 4*A*](#F4){ref-type="fig"}). The *Notum* mRNA levels in cortical bone were substantially lower (−63 ± 5%; *P* \< 0.01) in tamoxifen-treated *CAGGCre-ER-Notum^flox/flox^* mice compared with tamoxifen-treated *Notum^flox/flox^* mice, demonstrating efficient *Notum* inactivation ([Fig. 4*B*](#F4){ref-type="fig"}). To determine the skeletal phenotype in the inducible *Notum* knockout mice, 11-wk-old *Notum^flox/flox^* and *CAGGCre-ER-Notum^flox/flox^* mice were treated with tamoxifen and the phenotype was evaluated 31 d after the first tamoxifen dose. The inducible inactivation of *Notum* did not affect body weight, weights of liver, gonadal fat, kidney, spleen or uterus, or the lengths of femur or tibia ([Supplemental Table S4](#SM2){ref-type="supplementary-material"}). Detailed skeletal analyses using DXA and CT revealed that the total femur BMD (*P* \< 0.05) and several diaphyseal cortical bone mass parameter (cortical area, cortical thickness, and cortical moment of inertia; *P* \< 0.05) but not BV/TV in the metaphyseal region of tibia were significantly increased in tamoxifen-treated *CAGGCre-ER-Notum^flox/flox^* mice compared with tamoxifen-treated *Notum^flox/flox^* mice ([**Table 1**](#T1){ref-type="table"}). ![Inducible inactivation of the *Notum* gene increases cortical periosteal bone formation. *A*) Schematic drawing of the inducible *Notum* inactivation model. *B*) mRNA expression analyses of *Notum* in cortical bone in *CAGGCre-ER-Notum^flox/flox^* mice (22-wk-old male mice, *n* = 10) and *Notum^flox/flox^* mice (*n* = 12). *C*--*E*) Mineralizing surface per bone surface (*C*), mineral apposition rate (*D*), and bone formation rate per bone surface (*E*) on the periosteal side of tibia as measured by histomorphometry in *CAGGCre-ER-Notum^flox/flox^* mice (15-wk-old female mice, *n* = 11) and *Notum^flox/flox^* mice (*n* = 6). *F*--*H*) Mineralizing surface per bone surface (*F*), mineral apposition rate (*G*), and bone formation rate per bone surface (*H*) on the endocortical side of tibia as measured by histomorphometry in *CAGGCre-ER-Notum^flox/flox^* mice (15-wk-old female mice, *n* = 11) and *Notum^flox/flox^* mice (*n* = 6). *I*) Levels of procollagen type I N-terminal propeptide (P1NP, left) and (CTX) fragments (right) measured in serum in *CAGGCre-ER-Notum^flox/flox^* mice (22-wk-old male mice, *n* = 12) and *Notum^flox/flox^* mice (*n* = 10). *J*) mRNA expression analyses of ALP (*Alpl,* left) and TRAP (*Trap/Acp5*, right) in cortical bone of *CAGGCre-ER-Notum^flox/flox^* mice (22-wk-old male mice, *n* = 10) and *Notum^flox/flox^* mice (*n* = 12). All values are given as means ± [sem]{.smallcaps}. \**P* \< 0.05, \*\**P* \< 0.01, \*\*\**P* \< 0.001 *vs.* *Notum^flox/flox^* mice (Student's *t* test).](fj.201900707Rf4){#F4} ###### *Increased cortical bone mass in mice with inducible* Notum *inactivation* Variable *Notum^flox/flox^* *CAGG-Cre-ER-Notum^flox/flox^* ------------------------------------------- -------------------- -------------------------------- DXA  Femur BMD (mg/cm^2^) 104.2 ± 1.8 109.2 ± 1.5\*  Total body BMD (mg/cm^2^) 73.2 ± 0.8 75.4 ± 0.7^\#^ CT (tibia)  Cortical bone area (mm^2^) 0.87 ± 0.02 0.93 ± 0.02\*  Cortical bone thickness (µm) 222 ± 5 234 ± 3\*  Mean polar moment of inertia (mm^4^) 0.33 ± 0.01 0.37 ± 0.02\*  Cortical porosity (%) 6.4 ± 0.8 5.8 ± 0.6  Cortical volumetric BMD (g/cm^3^) 1.22 ± 0.02 1.23 ± 0.02  Trabecular bone volume/tissue volume (%) 20.1 ± 1.5 21.6 ± 0.7 Values are given as means ± [sem]{.smallcaps}. (*Notum^flox/flox^*, *n* = 10; *CAGG-Cre-ER*-*Notum^flox/flox^*, *n* = 12; 15-wk-old female mice; ^\#^*P* = 0.056). \**P* \< 0.05 *vs.* *Notum^flox/flox^* (Student's *t* test). Dynamic cortical histomorphometry was performed to determine if inducible *Notum* inactivation increased cortical bone thickness *via* increased cortical bone formation. Periosteal mineralized surface (+32 ± 4%; *P* \< 0.01; [Fig. 4*C*](#F4){ref-type="fig"}), mineral apposition rate (+65 ± 16%; *P* \< 0.05; [Fig. 4*D*](#F4){ref-type="fig"}), and bone formation rate (+113 ± 25%; *P* \< 0.01; [Fig. 4*E*](#F4){ref-type="fig"}) were substantially increased in tamoxifen-treated *CAGGCre-ER-Notum^flox/flox^* mice compared with tamoxifen-treated *Notum^flox/flox^* mice. In contrast, endocortical bone formation was not significantly affected by inducible *Notum* inactivation ([Fig. 4*F*--*H*](#F4){ref-type="fig"}). Static histomorphometry revealed that the periosteal number of osteoblasts (+183 ± 68%; *P* \< 0.05) and the periosteal surface covered by osteoblasts (+183 ± 66%; *P* \< 0.05) were significantly increased by inducible *Notum* inactivation, whereas the number of osteoblasts on the endocortical side was unchanged ([**Table 2**](#T2){ref-type="table"}). These mechanistic studies using inducible *Notum* inactivation demonstrate that *Notum* deficiency increases bone formation at the periosteal, but not at the endocortical, side of the cortical bone. For further mechanistic studies including analyses of serum bone markers and gene transcripts, *Notum^flox/flox^* and *CAGGCre-ER-Notum^flox/flox^* mice were induced with tamoxifen and evaluated 11 d after the first tamoxifen dose. Although serum levels of the bone resorption marker CTX fragments ([Fig. 4*I*](#F4){ref-type="fig"}) and the mRNA levels of the osteoclast-specific transcript *Acp5* in cortical bone ([Fig. 4*J*](#F4){ref-type="fig"}) were reduced, neither the number of periosteal osteoclasts nor the number of endocortical osteoclasts were significantly altered by *Notum* inactivation ([Table 2](#T2){ref-type="table"}). Thus, inducible *Notum* inactivation does not seem to influence the number of osteoclasts, but we cannot exclude the possibility that an inhibitory effect on osteoclast activity might have contributed to increased cortical bone thickness. ###### *Increased number of periosteal osteoblasts in mice with inducible* Notum *inactivation* Variable *Notum^flox/flox^* *CAGG-Cre-ER-Notum^flox/flox^* ------------------------------------------- -------------------- -------------------------------- Periosteal  Number of osteoblasts/perimeter (mm^−1^) 1.9 ± 0.6 5.6 ± 1.3\*  Osteoblast surface/bone surface (%) 4.0 ± 1.3 11.5 ± 2.7\*  Number of osteoclasts/perimeter (mm^−1^) 0.8 ± 0.1 0.9 ± 0.1  Osteoclast surface/bone surface (%) 2.7 ± 0.4 2.7 ± 0.4 Endocortical  Number of osteoblasts/perimeter (mm^−1^) 23.3 ± 2.7 18.0 ± 2.1  Osteoblast surface/bone surface (%) 46.4 ± 5.8 34.5 ± 4.1  Number of osteoclasts/perimeter (mm^−1^) 1.3 ± 0.2 1.2 ± 0.2  Osteoclast surface/bone surface (%) 4.1 ± 0.7 3.4 ± 0.5 Values are given as means ± [sem]{.smallcaps}. (*Notum^flox/flox^*, *n* = 12; *CAGG-Cre-ER-Notum^flox/flox^*, *n* = 10; 22-wk-old male mice). \**P* \< 0.05 *vs.* *Notum^flox/flox^* (Student's *t* test). Silencing of NOTUM increases osteoblast differentiation {#s30} ------------------------------------------------------- To determine if osteoblast-derived NOTUM has the capacity to exert local effects on osteoblasts, *Notum* expression was silenced in osteoblasts using the preosteoblast cell line MC3T3-E1. The cells were silenced 1 and 3 d after seeding and treated with osteogenic medium for 7 d, starting 24 h after the second silencing. Expression of *Notum* mRNA was decreased by 82 ± 9% (*P* \< 0.001) and 95 ± 2% (*P* \< 0.05) after the first and second silencing ([**Fig. 5*A***](#F5){ref-type="fig"}). Analyses of osteoblast differentiation showed that silencing of *Notum* resulted in enhanced mRNA expression of *Alpl* already 24 h after the second silencing ([Fig. 5*B*](#F5){ref-type="fig"}) and an increased ALP staining ([Fig. 5*C*](#F5){ref-type="fig"}) and a 7-fold increase of ALP activity ([Fig. 5*D*](#F5){ref-type="fig"}) at the end of the cultures, indicating that NOTUM is a negative regulator of osteoblastic differentiation. These findings clearly demonstrate that osteoblast-derived NOTUM has the capacity to exert local effects on osteoblast differentiation. ![Silencing of NOTUM in MC3T3-E1 cells increases osteoblastic differentiation. *A*) mRNA expression analyses of *Notum* in MC3T3-E1 cells 24 h after first (d 2) and second (d 4) silencing using siRNA. *B*) mRNA expression analyses of *Alpl* in MC3T3-E1 cells 24 h after second silencing of *Notum* using siRNA. *C*) ALP staining of MC3T3-E1 cells after silencing of *Notum* and cultured in osteogenic medium for 7 d. *D*) ALP enzymatic activity in MC3T3-E1 cells after silencing of *Notum* and cultured in osteogenic medium for 7 d. *E*) TRAP staining of osteoclasts cultured in the presence or absence of recombinant NOTUM protein for 3 d. *F*--*H*) Expression of *Acp5* (*F*), *Ctsk* (*G*), and *Nfatc1* (*H*) in osteoclasts cultured in the presence or absence of recombinant NOTUM for 3 d. *I*) TRAP5b protein released into osteoclast culture media during d 7--10 of culture on bone discs in the presence or absence of recombinant NOTUM. *J*) CTX released into osteoclast culture medium during d 7--10 of culture on bone discs in the presence or absence of recombinant NOTUM. M, M-CSF; RL, RANKL; siSCR, control scramble siRNA; siNOTUM, NOTUM-specific siRNA. All values are given as means ± [sem]{.smallcaps}. \**P* \< 0.05, \*\**P* \< 0.01, \*\*\**P* \< 0.001 *vs.* siSCR \[Student's *t* test (*A*, *B*, *D*)\]; \*\*\**P* \< 0.001 *vs.* the M treatment \[1-way ANOVA and Tukey's multiple comparison test (*F*--*H*)\]; *n* = 4 wells/treatment. All cell culture experiments were repeated at least twice with similar results.](fj.201900707Rf5){#F5} Recombinant NOTUM protein does not affect osteoclast differentiation or activity {#s31} -------------------------------------------------------------------------------- Because the NOTUM expression is very low in osteoclasts, we wanted to explore the possibility that paracrine or endocrine NOTUM might exert effects on osteoclasts. To determine if preosteoclasts or osteoclasts have the capacity to respond to NOTUM, we treated BMMs with M-CSF/RANKL in the absence and presence of recombinant human NOTUM. First, we demonstrated that the used recombinant human NOTUM had effective lipase activity ([Supplemental Fig. S3](#SM1){ref-type="supplementary-material"}). Recombinant human NOTUM did not affect the formation of TRAP-positive multinucleated osteoclasts ([Fig. 5*E*](#F5){ref-type="fig"}) or M-CSF/RANKL-induced expression of the osteoclastic genes *Acp5*, *Ctsk*, and *Nfatc1* ([Fig. 5*F*--*H*](#F5){ref-type="fig"}), when BMMs were cultured on plastic dishes. Recombinant NOTUM affected neither osteoclast formation on bone disc as assessed by release of TRAP5b ([Fig. 5*I*](#F5){ref-type="fig"}) nor osteoclast resorption activity analyzed by release of CTX ([Fig. 5*J*](#F5){ref-type="fig"}), suggesting that NOTUM does not exert any direct effects on osteoclasts. These data, however, do not exclude the possibility that NOTUM could have indirect effects on osteoclasts *in vivo.* Large-scale human genetic analyses reveal that the *NOTUM* locus is associated with BMD {#s32} --------------------------------------------------------------------------------------- To determine if the *NOTUM* locus is also associated with bone mass in humans, we used the complete UK Biobank data set, currently the most powerful available resource for human genetic analyses of a bone phenotype (*<https://www.biorxiv.org/content/early/2017/09/27/194944>*). We searched for possible associations between imputed SNPs specifically in the *NOTUM* locus (defined as ± 25 kb from the gene start and end limits) and BMD as estimated by quantitative ultrasound of the heel (eBMD; available in 445,921 European subjects). In total, we tested 125 SNPs with a minor allele of \>1% and identified 2 independent genetic signals located upstream of *NOTUM* ([**Table 3**](#T3){ref-type="table"} and [Supplemental Fig. S4*A, B*](#SM1){ref-type="supplementary-material"}) both associated with a *P* value of \<10^−10^, providing compelling statistical evidence that genetic variation arising from the *NOTUM* locus is associated with eBMD in humans. One of the identified SNPs, rs35344256 ([Supplemental Fig. S4*A*](#SM1){ref-type="supplementary-material"}; C allele associated with reduced eBMD), is a common variant with minor allele frequency of 31%, whereas the other SNP, rs147901986 ([Supplemental Fig. S4*B*](#SM1){ref-type="supplementary-material"}; G allele associated with increased eBMD), is a low-frequency variant with minor allele frequency of 2.9% ([Table 3](#T3){ref-type="table"}). ###### Genetic variants in the NOTUM locus are associated with BMD SNP Chr Position A1 A2 AF β SE P N ------------- ----- ---------- ---- ---- ------ --------- -------- --------- -------- rs35344256 17 79934194 C A 0.69 −0.0134 0.0020 1.5E−11 445921 rs147901986 17 79921522 G A 0.97 0.0386 0.0059 4.6E−11 445921 Large-scale human genetic analyses revealed that genetic variants in the *NOTUM* locus are associated with BMD. We specifically searched for possible associations between imputed SNPs (minor allele frequency \>1%) in the *NOTUM* locus ±25 kb (in total 125 SNPs with a minor allele frequency \>1% were evaluated) and BMD as estimated by quantitative ultrasound of the heel. A1, effect allele; A2, other allele; AF, allele frequency of allele 1. Independent SNPs associated with BMD and with *P* \< 1 × 10^−10^. β given as [sd]{.smallcaps} change in eBMD per effect allele. DISCUSSION {#s33} ========== NOTUM is a secreted WNT lipase reducing WNT activity ([@B9], [@B10]). As the role of NOTUM *in vivo* is essentially unknown, we developed global, cell-specific, and inducible *Notum*-inactivated mouse models. We herein demonstrate that osteoblast lineage--derived NOTUM is a crucial determinant specifically of cortical bone mass and strength *via* regulation of periosteal bone formation in mice and that genetic variants in the *NOTUM* locus are robustly associated with BMD in humans. WNT proteins signal through both the canonical WNT pathway and the noncanonical WNT pathways ([@B1], [@B2], [@B14], [@B33], [@B34]). Activation of canonical β-catenin signaling increases bone mass, and rare human genetic mutations affecting bone have been identified in components of the canonical and noncanonical WNT signaling machinery ([@B1], [@B4], [@B35][@B36][@B37][@B38]--[@B39]). Recent *in vitro* studies demonstrate that NOTUM inactivates WNTs by removing their attached lipid group essential for activation of Frizzled receptors ([@B9], [@B10]). To evaluate if NOTUM may be a local factor in bone, we first developed and characterized a mouse model with global *Notum* inactivation. Because a majority of *Notum^−/−^* mice died during embryonal development, it was not possible to determine the physiologic role of NOTUM for any phenotype in adult *Notum^−/−^* mice. However, heterozygous Notum-inactivated mice, having significantly reduced *Notum* expression in multiple tissues, were born apparently healthy and displayed a normal body weight and longitudinal bone growth. The female mice with heterozygous global *Notum* inactivation had increased cortical bone thickness and bone strength but normal BV/TV. Although our studies using heterozygous female mice with global *Notum* inactivation indicated that NOTUM reduces cortical but not trabecular bone mass, these studies were difficult to interpret. First, a majority of homozygous *Notum*-inactivated mice died during embryonic development, demonstrating that NOTUM is required for normal development and that it therefore is not possible to use this mouse model to determine the effect of NOTUM on adult bone metabolism. Second, although female mice with global *Notum* inactivation had increased cortical bone thickness, no significant bone phenotype was observed in heterozygous male mice. Finally, the experiments using global *Notum* inactivation could not determine the source of NOTUM with an impact on cortical bone. Because *Notum* expression was high in the cortical bone, we hypothesized that the cortical bone phenotype in the heterozygote female *Notum*-inactivated mice might be caused by lack of *Notum* expression by osteoblast lineage cells in cortical bone. However, a role of other contributing cell types in the bone compartment or by circulating factors, such as liver-derived Notum, or other indirect systemic effects could not be excluded using this mouse model. To determine if the relatively high *Notum* expression in cortical bone is osteoblast-derived and to determine the *in vivo* role of osteoblast-derived NOTUM, we developed a conditional *Notum*-inactivated mouse model with inactivation of *Notum* in both osteoblasts and osteocytes. Using this model, we clearly demonstrated that osteoblast lineage cells are the principal source of NOTUM in cortical bone, and we next determined the functional role of osteoblast-derived NOTUM. In contrast to homozygous mice with global *Notum* inactivation, those with osteoblast lineage--specific *Notum* inactivation were born healthy and displayed a normal body weight and longitudinal bone growth. These findings demonstrate that the severe developmental effects in the global *Notum*-inactivated homozygous mice are caused by lack of NOTUM in other cells than osteoblast lineage cells. CT analyses, separating the cortical and trabecular bone compartments revealed that both adult male and adult female mice with osteoblast lineage--specific inactivation of *Notum* had increased cortical bone thickness in the long bones, whereas no increase was observed on BV/TV in the vertebrae or in the long bones. Although slightly more pronounced, this specific cortical bone phenotype in mice with osteoblast lineage--specific *Notum* inactivation is very similar to the cortical bone phenotype observed in the female mice with heterozygous global Notum inactivation, supporting the concept that osteoblast-derived NOTUM is mediating the effects on cortical bone thickness. Three-point bending analyses revealed substantially increased whole-bone mechanical bone strength in mice with osteoblast lineage--specific inactivation of *Notum* compared with control mice, demonstrating that the increased cortical bone thickness resulted in increased bone strength. It is a limitation with the present study that the cortical bone quality was only evaluated by cortical porosity measurements, whereas no analyses of the tissue material properties were performed, precluding statements on the cortical bone quality. Another limitation with the present study is that the cortical tissue organization estimated by informative collagen type I orientation analyses was not performed ([@B40][@B41]--[@B42]). The increased cortical bone thickness in both the male and the female mice with osteoblast lineage--specific *Notum* inactivation was the result of an increased periosteal circumference, whereas the endocortical circumference was not reduced, suggesting that the increased cortical bone thickness was caused by cortical periosteal bone formation. However, the lack of significant effects on cortical bone formation rates and number of osteoclasts in 17-wk-old adult mice with lifelong osteoblast lineage--specific *Notum* inactivation indicated that the increased cortical bone thickness of these mice was a result of effects occurring earlier in life. Therefore, to evaluate the effect of NOTUM on adult cortical bone homeostasis and the underlying mechanisms, we developed a mouse model with tamoxifen-inducible *Notum* inactivation in adult mice. An increase specifically of cortical bone mass was observed only 4 wk after the inducible *Notum* inactivation in adult mice, excluding confounding developmental effects and demonstrating that endogenous NOTUM exerts crucial effects on cortical bone homeostasis in adult mice. The mechanistic studies using the mouse model with inducible inactivation of *Notum* revealed that NOTUM deficiency increases periosteal bone formation as a result of a combination of increased periosteal mineralized surface, reflecting the number of active osteoblasts, and increased periosteal mineral apposition rate, reflecting the activity of osteoblasts. An effect on cortical periosteal bone formation was further supported by the observed increased number of osteoblasts on the periosteal surface in mice with inducible *Notum* inactivation. Collectively, the acute mechanistic data, using inducible *Notum* inactivation, demonstrate that the increased periosteal cortical circumference, observed in mice with lifelong chronic osteoblast lineage--specific inactivation of *Notum,* is primarily caused by increased periosteal bone formation. A local effect of osteoblast-derived *Notum* on osteoblasts is supported by our finding that *Notum* silencing enhanced osteoblast differentiation in cultured MC3T3-E1 osteoblasts. In contrast, very low *Notum* expression was observed in osteoclast cultures and recombinant human NOTUM did not exert any significant direct effect on osteoclast formation or activity *in vitro*. Although serum levels of the bone resorption marker CTX and the mRNA levels of the osteoclast-specific transcript *Acp5* in cortical bone were reduced by inducible *Notum* inactivation, we were unable to detect any decrease of the number of osteoclasts in cortical bone. Thus, NOTUM does not seem to exert any effect on osteoclastogenesis in cortical bone. Our findings overall demonstrate that the principal mechanism for the increased cortical bone thickness in *Notum*-inactivated mice is an increased periosteal bone formation *via* depletion of local inhibitory effects of osteoblast-derived NOTUM on osteoblast number and activity. Very similar increases on cortical bone thickness and cortical bone strength were observed in male and female mice for the mouse model with osteoblast lineage--specific *Notum* inactivation. In contrast, for the mouse model with heterozygous global *Notum* inactivation, a significant cortical bone phenotype was only observed in female mice, suggesting that female mice might be more sensitive to changes in *Notum* expression. It is a limitation with the present study that possible sex differences in the effect on cortical bone thickness were not investigated in the inducible *Notum*-inactivated mouse model. Cortical bone mass is a major determinant of bone strength and therefore of susceptibility to fractures ([@B3], [@B43]). With aging, the mass of cortical bone may decrease more than the mass of trabecular bone, and fractures occurring in older persons result to a large extent from cortical bone fragility ([@B43], [@B44]). Although progress has been made in therapeutic approaches to reduce the risk of vertebral fracture (which occurs at sites rich in trabecular bone), currently available treatments do little to reduce the risk of nonvertebral fracture, which results mostly from cortical bone fragility ([@B4], [@B43], [@B45], [@B46]). Thus, there is a medical need to improve the therapy for nonvertebral fractures, and our present finding that osteoblast-derived *Notum* reduces cortical bone mass suggests that methods inhibiting the effect of NOTUM might increase cortical bone mass and thereby reduce nonvertebral fracture risk. Mice with osteoblast- or osteocyte-specific inactivation of *Wntless* (*Wls*, previously known as Gpr177), which is required for the secretion of WNT ligands from cells, have substantially reduced bone mass, illustrating the crucial role of osteoblast lineage--derived WNTs for bone homeostasis ([@B47]). In contrast to mice with osteoblast lineage--specific *Wls* inactivation (displaying a marked reduction of both trabecular and cortical bone mass) ([@B47]), osteoblast lineage--specific *Notum* inactivation resulted specifically in increased cortical bone thickness without affecting trabecular bone mass. Combined, these findings suggest that NOTUM mainly renders WNTs inactive in cortical bone, and this notion is supported by our observation that the *Notum* expression is higher in cortical than in trabecular bone. The fact that patients with Pyle's disease, caused by recessive mutation in the *SFRP4* gene, encoding a soluble WNT inhibitor, have thin cortical but dense trabecular bone, further demonstrates that WNTs and WNT modulators exert site-specific effects in the skeleton ([@B46]). Previous studies have demonstrated that osteoblast-derived WNT16 increases cortical but not trabecular bone mass, and it is therefore possible that osteoblast-derived WNT16 is inactivated by osteoblast-derived NOTUM in cortical bone ([@B2], [@B48]). We recently observed that inducible *Wnt16* inactivation in adult mice reduces periosteal cortical bone formation ([@B32]). In addition, we herein demonstrate that inducible inactivation of the WNT lipase *Notum* in adult mice increases periosteal cortical bone formation, suggesting that there may be a local periosteal interaction between WNT16 and NOTUM for the regulation of bone formation. We propose that in adult mice, WNT16-mediated periosteal bone formation might be inhibited by a local NOTUM-mediated negative feedback. A role of intact palmitoleoylation of WNTs for periosteal bone growth is supported by our recent finding that inhibition of Porcupine, the enzyme responsible for palmitoleoylation of WNTs, reduces periosteal but not endocortical bone formation ([@B49]). The trabecular bone phenotype in mice with osteoblast lineage--specific inactivation of *Wls* most likely is mediated by loss of WNTs not inactivated by NOTUM within the trabecular bone microenvironment ([@B47]). Collectively, these findings demonstrate that WNTs may exert compartment-specific effects on bone homeostasis and that NOTUM most likely inactivates WNTs with a crucial effect specifically on cortical bone. It has recently been reported that substances inhibiting the lipase activity of NOTUM increased cortical bone thickness ([@B13], [@B50], [@B51]). These substances, selected based on their capacity to inhibit lipase activity and canonical signaling specifically by WNT3A, increased cortical bone thickness mainly *via* an increased endocortical bone formation, whereas inactivation of endogenous NOTUM in the present study increased cortical bone thickness mainly *via* increased periosteal bone formation. This difference in the predominant cortical bone surface involved might be caused by differences in the effects of inactivation of endogenous NOTUM and the effects of pharmacological inhibition of lipase activity, resulting in enhanced WNT3A-mediated canonical signaling. First, endogenous NOTUM may exert lipase independent effects; second, the lipase substrate specificity, both regarding WNTs known to control cortical bone homeostasis such as WNT16 and other non-WNT proteins, may differ between endogenous NOTUM and the pharmacological effect of the developed substances. Finally, we evaluated if the *NOTUM* locus is associated with bone mass also in humans. To this end, we used the recently released complete UK Biobank data set (*n* = 445,921 European subjects) to test genetic variation in the *NOTUM* locus for association with eBMD variation in the most powered setting available. We identified 2 independent genetic variants in the *NOTUM* locus that are associated with eBMD. This finding is the first indication that the *NOTUM* locus is involved in the regulation of bone mass also in humans. Subsequent studies assessing whether these genetic variants in the *NOTUM* locus are specifically associated with cortical bone thickness are warranted in deeper phenotyped cohorts assessed by computer tomography. In summary, we demonstrate that osteoblast-derived NOTUM is an essential regulator of cortical bone mass *via* effects on periosteal bone formation in adult mice and that genetic variation in the *NOTUM* locus is associated with BMD in humans. Therapies targeting osteoblast-derived NOTUM may constitute useful and specific treatments to reduce the risk of nonvertebral fractures. Supplementary Material {#s35} ====================== This article includes supplemental data. Please visit *<http://www.fasebj.org>* to obtain this information. ###### Click here for additional data file. ###### Click here for additional data file. This article includes supplemental data. Please visit *<http://www.fasebj.org>* to obtain this information. The authors thank Lotta Uggla, Biljana Aleksic, Anna Westerlund, and Anette Hansevi (all from the University of Gothenburg) for excellent technical assistance, and Carolina Medina-Gomez (Erasmus University Rotterdam) for support in the human genetic analyses. This study was supported by the Swedish Research Council, the Swedish Foundation for Strategic Research, the Swedish state under the agreement between the Swedish government and the county councils, the ALF-agreement in Gothenburg (Grants 238261, 226481, and 237551), the IngaBritt and Arne Lundberg Foundation, the Royal 80 Year Fund of King Gustav V, the Torsten and Ragnar Söderberg's Foundation, the Knut and Alice Wallenberg Foundation, the Novo Nordisk Foundation, and São Paulo Research Foundation (Grant 2014/05283-3). J.T. was supported by the Deutsche Forschungsgemeinschaft (DFG) Collaborative Research Centre (CRC) 1149, and G.C.N. was supported by Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES; Finance Code 001). S.M.-S. and K.H.N. share first authorship. The authors declare no conflicts of interest. α-MEM : α-minimum essential medium μCT : micro--computed tomography ALP : alkaline phosphatase BMD : bone mineral density BMM : bone marrow--derived macrophage BV/TV : bone volume/tissue volume CT : computed tomography CTX : C-terminal type I collagen DXA : dual-energy X-ray absorptiometry eBMD : estimated bone mineral density M-CSF : macrophage colony-stimulating factor PGK-1 : phosphoglycerate kinase 1 RANKL : receptor activator of nuclear factor κ-B ligand siRNA : small interfering RNA TRAP : tartrate-resistant acid phosphatase SNP : single nucleotide polymorphism Wls : Wntless WNT : Wingless-type MMTV integration site family member WT : wild type AUTHOR CONTRIBUTIONS {#s34} ==================== S. Movérare-Skrtic and K. H. Nilsson were responsible for the animal experiments; P. Henning and T. Funck-Brentano assisted with animal experiments; M. Nethander, F. Rivadeneira, and C. Ohlsson performed the human genetic analyses; A. Koskela and J. Tuukkanen performed the 3-point bending experiments; G. Coletto Nunes and P. P. C. Souza performed studies using MC3T3-E1 osteoblasts; P. Henning and U. H. Lerner were responsible for primary cell cultures of osteoblasts and osteoclasts; J. Tuckermann provided the *Runx2-cre* mice; C. Perret provided the *Notum*^*flox*^ mice; S. Movérare-Skrtic, U. H. Lerner, and C. Ohlsson designed and supervised the project; and S. Movérare-Skrtic, K. H. Nilsson, U. H. Lerner, and C. Ohlsson wrote the manuscript. [^1]: These authors contributed equally to this work.

Introduction {#s1} ============ Surgical endodontic retreatment is essential in cases of failed endodontic treatment or when conventional endodontic treatment cannot be undertaken. Endodontic surgery entails raising a mucoperiosteal flap and performing osteotomy, followed by root-end resection, root-end preparation, and root-end filling. A root-end filling material is considered effective when it provides a complete apical seal, preventing the passage of microorganisms into the root canal system and leading to the throwing out of microorganisms and their by-products from the root canal system.^[@R01]^ Various materials have been suggested and used for root end filling. They include zinc oxide eugenol cements, glass ionomer cement, super EBA, polyvinyl resins, composite resins, resin-glass ionomer hybrids, and mineral trioxide aggregate (MTA).^[@R02]-[@R06]^ Microleakage studies have confirmed MTA has the best apical sealing ability. However, despite its excellent sealing ability and biocompatibility when compared with other root-end filling materials, MTA is expensive, has delayed setting time, and poor handling properties.^[@R07],[@R08]^ Asgary et al^[@R09]^ have introduced a novel endodontic cement with sealing ability comparable to MTA. This biomaterial was formulated using different calcium compounds such as calcium hydroxide, calcium oxide, calcium phosphate, calcium sulfate, calcium silicate, and calcium carbonate. Studies have demonstrated that the calcium-enriched mixture (CEM) cement comprises water-soluble calcium and phosphate which immediately forms hydroxyapatite during and after setting.^[@R09]^ The composition and surface characteristics as well as the physical and chemical properties of CEM cement have been examined. CEM cement has been shown to have a sealing ability comparable to MTA and superior to IRM.,^[@R06],[@R10]^ The cement exhibits several advantages including high tissue biocompatibility, hard tissue induction, effective sealing ability against the entry of microorganisms, ability to set in an aqueous environment, antibacterial effects, and resistance to washout.^[@R10]-[@R11],[@R13]^ *In-vitro* studies have documented the equivalence of CEM cement and MTA properties.^[@R10],[@R14],[@R15]^ *In-vivo* studies have also produced successful results with CEM cement.^[@R11],[@R12],[@R16]-[@R20]^ In an *in-vitro* study, the influence of the thickness of mineral trioxide aggregate on the sealing ability of root-end fillings was assessed by Valois et al^[@R21]^ The results revealed a thickness of 4 mm as most adequate when using MTA as a root-end filling material.^[@R21]^ In a previous study by Rahimi et al,^[@R22]^ the microleakage with MTA as root-end filling material was not found to be significantly different among various thicknesses. Investigations of the sealing ability of CEM cement through dye penetration have revealed that the sealing properties of this root-end filling material parallel those of commercial types of MTA.^[@R23]^ The aim of this study was to compare the sealing ability associated with three different thicknesses of CEM cement as a root-end filling material in cavities prepared by ultrasonic retro-tips. Materials and Methods {#s2} ===================== Seventy single-rooted human premolar teeth extracted for periodontal or orthodontic purposes were selected for this study. The teeth were evaluated under stereomicroscope and radiography. Any teeth with caries, cracks, resorption, fractures, morphologic anomalies, and open apices were excluded from the study. The selected teeth were decoronated at cemento-enamel junction (CEJ) level with a diamond disk (D&Z, Darmstadt, Germany) under running water and air spray. Working length was determined with a \#15 K-type file (Mani, Utsunomiya, Japan) 1 mm short of the apical foramen. The canals were prepared up to size \#40 using the step-back technique and the shaping of the middle and coronal thirds was carried out by Gates Glidden burs 1, 2, and 3. During instrumentation procedures, root canals were irrigated with 10 ml of saline solution. All canals were obturated using gutta percha (Diadent, Korea) and AH-26 sealer (Dentsply, Konstanz, Germany) with the lateral compaction technique. The teeth were stored at 37°C and 100% humidity for 48 hours (Heratherm, Thermo Inc., Switzerland). The apical 3 mm of each tooth was resected perpendicular to the long axis of the tooth with a diamond bur under continuous water and air spray. Teeth were randomly divided into three experimental groups each containing 20 samples, and 5 positive and 5 negative controls. In group 1, root-end cavities were prepared to a depth of 1mm perpendicular to the long axis using ultrasonic retrotips Kis-3D (Spartan, Missouri, USA). In samples of groups 2 and 3, the same procedure was done to depths of 2 mm and 3 mm, respectively. In positive controls, the cavities were prepared to a depth of 3 mm. The cavities were then irrigated and dried by paper cones. The CEM cement was mixed according to the manufacturer's instructions on a sterile glass slab and filled into the cavities with the aid of a small condenser (Kerr Hawe, Orange, CA, USA). Any excess material was removed with a sterile cotton swab. The quality of the root-end fillings was confirmed by radiographs in two directions (mesiodistal and buccolingual), while the root-end cavities of controlled groups remained empty. In all experimental and positive control groups, two layers of nail varnish were applied to the surface of the teeth up to the level of the resected root-end. In negative controls, the entire surface of the teeth was covered with two layers of nail varnish. All teeth were kept at 37°C and 100% humidity for 48 hours, and then immersed into synthetic tissue fluid (STF) with pH=7 for 48 hours. The specimens were placed horizontally in 2% Rhodamine B (Merk, Darmstadt, Germany) for 48 hours. Then the samples were rinsed for 10 min under running water. After that, two facial and lingual fissures were created along the long axis of the roots using a diamond disc and the roots were longitudinally resected into two mesial and distal halves. The maximum amount of linear dye penetration was measured under a stereomicroscope (Zeiss, Munich, Germany) at ×16 magnification with 0.1 mm accuracy and using Image J software (35d; National Institutes of Health, USA). The data were analyzed with one-way ANOVA analysis and the Bonferroni post-hoc test was utilized to reveal specific group differences. The level of significance was set at P \< 0.05. Results {#s3} ======= No dye microleakage was noted in the negative control samples. The mean ± standard deviation of dye penetration in the 1, 2, and 3 mm groups were 3395.58 ± 1893.44, 3410.47 ± 1440.58, and 2581.65 ± 1852.90 micrometers, respectively ([Figure 1](#F01){ref-type="fig"}). {#F01} Significant 'Skewness' and 'Kurtosis' tests confirmed non-normal data distribution. Natural logarithmic transformation provided normal distribution of the data. Analysis of variance indicated significant differences among the tested groups (P = 0.000). However, Bonferroni post-hoc test revealed that only the positive control group differed from the experimental groups (P = 0.000; [Table 1](#T1){ref-type="table"}). Consequently, there was no significant difference among sealing ability of 1, 2 and 3 mm CEM cement as a retrofilling material. ###### Mean dye leakage (upper and lower 95% confidence interval) of experimental groups (micrometer) ---------- ------------------- ------------------- ------------------- Groups Variable 1 mm 2 mm 3 mm Leakage 3395.58\ 3410.47\ 2581.65\ (4281.74-2509.42) (4084.68-2736.25) (3448.84-1714.47) ---------- ------------------- ------------------- ------------------- Discussion {#s4} ========== The success of periradicular surgery directly depends on the achievement of a good apical seal, utilizing a well-adapted root-end filling material that prevents the leakage of irritants from the root canal system into the periradicular region.^[@R24]^ In this study, the microleakage of three different thicknesses of CEM cement as a root-end filling material was evaluated by the dye penetration method. Findings indicated no significant differences among 1, 2, and 3 mm of CEM cement as a root end filling material. But in 1-mm and 2-mm depths, the leakage was more than the root-end filling material's depth; in other words, only 3-mm depth was capable of providing adequate seal. Apical microleakage has been assessed using various methods including dye/ink, bacterial/endotoxin leakage, radioisotope tracing, and fluid filtration technique.^[@R25]-[@R31]^ There is no evidence to favor the superiority of any particular method. However, the dye penetration technique is widely used for microleakage studies because dyes are cheap, safe, easily available, and also relatively easy to be stored, used and to have their penetration assessed quantitatively.^[@R32],[@R33]^ Different dyes have been employed for dye penetration test as India ink and methylene blue. In this study, the extension of dye (2% Rhodamine B) penetration was used as the criteria for evaluation. The use of methylene blue in marginal sealing studies has been debated, due to its incompatibility with alkaline substances, which may induce discoloration of the dye.^[@R34]^ It is known that methylene blue dye presents an acid character and Rhodamine B, a basic one. Rhodamine B is a basic intense organic dye, soluble in water at room temperature, also solvent in alcohols and common organic solvents, in addition to being highly stable.^[@R35]^ Since color stability of organic dyes is an important factor that must be observed in microleakage studies, and because of alkali conditions around CEM cement, in this study Rhodamine B dye was used for leakage assessment. An ideal root-end cavity preparation is a class I cavity at least 3 mm deep with parallel walls.^[@R36]^ Achieving this with the classical method in surgical endodontics is held back by several difficulties such as limited access, root anatomy, and tooth angulation. To avoid these problems, Piezoelectric ultrasonic devices for root-end preparation have been developed and are nowadays used as standard tools for retrograde cavity preparation.^[@R37]-[@R39]^ In another *in-vitro* study, the influence of the thickness of mineral trioxide aggregate on the sealing ability of root-end fillings was assessed by Valois et al.^[@R21]^ The 1-mm-thick MTA was the least effective in preventing apical leakage. No significant difference was found between 2- and 3-mm-thick MTA. Four-millimeter-thick MTA was found to be significantly more effective than the other thicknesses tested. These researchers suggested a thickness of 4 mm as most adequate when MTA is used as a root-end filling material.^[@R21]^ Similar findings were obtained in our study and CEM cement in 3-mm thickness present the most effective sealing ability. The apical sealing ability of CEM cement has been reported to be similar to different commercial types of MTA.^[@R23]^ In a previous study, microleakage in the 3-mm and 2-mm root-end cavities was less than that in 1 mm depth cavities, but analysis of variance revealed no significant differences across the three different thicknesses.^[@R22]^ The current study led to similar results but with CEM cement used as root-end filling material. Conclusion {#s5} ========== Based on the findings of this in vitro study, the CEM cement demonstrated adequate root-end sealing ability in 3-mm thickness. Acknowledgement {#s6} =============== The authors would like to extend their appreciation to the Office of the Vice Chancellor for Research, Tabriz University of Medical Sciences, for the financial support of this study.

Study Highlights**WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?**☑ More than half of ulcerative colitis (UC) patients fail to achieve mucosal healing with conventional drug therapy.☑ The efficacy of probiotics in the treatment of UC remains controversial.☑ We previously identified a long‐chain polyphosphate from *Lactobacillus brevis* SB88 and suggested that it was effective for improving the intestinal barrier function and inflammation.**WHAT QUESTION DID THIS STUDY ADDRESS?**☑ All drugs that are clinically available for the treatment of UC are antiinflammatory or immunomodulatory agents; none directly improve the intestinal barrier function.☑ The safety and efficacy of oral enteric capsules containing long‐chain polyphosphates in humans remain unclear.**WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?**☑ No adverse reactions occurred in association with long‐chain polyphosphate treatment in our first‐in‐human trial.☑ Among the 10 enrolled patients for whom conventional drug therapy had failed, 7 acquired a clinical response, 4 of whom achieved endoscopic remission.**HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?**☑ Because of their high safety and efficacy, long‐chain polyphosphates are strongly believed to be useful for the treatment of refractory ulcerative colitis. {#cpt1628-sec-0002} Ulcerative colitis (UC) is a condition of chronic intestinal inflammation, the pathogenesis of which is largely unknown. While steroids and mesalamine are the standard treatments for relieving symptoms, a treatment strategy for curing the disease has not yet been established. Most patients experience relapses repeatedly and suffer from abdominal pain, diarrhea, and hematochezia for more than 50 years. "Mucosal healing," which is defined by a sufficient intestinal barrier function, leads to a significantly long‐term remission;[1](#cpt1628-bib-0001){ref-type="ref"} however, more than half of UC patients fail to achieve mucosal healing, even when using recently developed anti--tumor necrosis factor (TNF)--α treatments.[2](#cpt1628-bib-0002){ref-type="ref"} Probiotics are live microorganisms that confer a health benefit on the host when administered in adequate amounts[3](#cpt1628-bib-0003){ref-type="ref"} and are used as functional foods as well as drugs for the treatment of acute intestinal inflammation and necrotizing enterocolitis.[4](#cpt1628-bib-0004){ref-type="ref"}, [5](#cpt1628-bib-0005){ref-type="ref"}, [6](#cpt1628-bib-0006){ref-type="ref"} Probiotics possess sufficient safety to be used for a long time; however, their efficacy as a treatment for inflammatory bowel diseases remains controversial[7](#cpt1628-bib-0007){ref-type="ref"} because the function of live bacteria is influenced by the intestinal conditions in individual patients. To solve this weak point of probiotic treatment, we have considered using the bioactive molecules produced by probiotics, which enhance the intestinal barrier function regardless of the intestinal conditions, as do drugs. A method for identifying bioactive molecules[8](#cpt1628-bib-0008){ref-type="ref"} was used to identify the probiotic‐derived molecule responsible for improving the intestinal barrier function. Briefly, fraction‐inducing HSP27, which is known to increase the intestinal barrier function, in human Caco2/BBE cells was separated from the conditioned media of *Lactobacillus brevis SBC8803* using spin columns equipped with several molecular weight cutoff membranes, diethylaminoethyl anion‐exchange chromatography, and size‐exclusion chromatography. The amino acid composition and peptidoglycan content of the HSP27‐inducible fractions were shown to be extremely low; however, the fraction contained a large amount of phosphorus (90% weight for weight) and oxygen. Based on these results, we assumed that the HSP27‐inducible fraction was a polyphosphate. Synthesized polyphosphate poly P significantly induced HSP27 in Caco2/BBE cells, and the HSP27 induction by the culture supernatant of *L. brevis* was diminished by the degradation of polyphosphate using polyphosphate kinase, indicating that poly P was the molecule responsible for the induction of HSP27.[9](#cpt1628-bib-0009){ref-type="ref"} We proposed that the administration of long‐chain polyphosphate led to the localization of cell‐adherent molecules along the tight junction in intestinal epithelia and thereby improved the intestinal barrier function and inflammation *in vitro* and in chemical‐induced colitis mice through the epithelial endocytosis of long‐chain polyphosphate‐integrin β1‐caveolin‐1 complex (patent nos. JP5660508, EP2559437).[10](#cpt1628-bib-0010){ref-type="ref"}, [11](#cpt1628-bib-0011){ref-type="ref"}, [12](#cpt1628-bib-0012){ref-type="ref"} This was the first study to show that long‐chain polyphosphate was able to improve the intestinal barrier function and ameliorate intestinal injury under inflammatory conditions, although short‐polyphosphate is known to enhance the proliferation of human fibroblasts by stabilizing fibroblast growth factors[13](#cpt1628-bib-0013){ref-type="ref"} and to exert antibacterial activity against many bacteria and fungi.[14](#cpt1628-bib-0014){ref-type="ref"}, [15](#cpt1628-bib-0015){ref-type="ref"}, [16](#cpt1628-bib-0016){ref-type="ref"}, [17](#cpt1628-bib-0017){ref-type="ref"} A subsequent study also showed that polyphosphate produced by a marine cyanobacterium, *Synechococcus*, exerted an antiinflammatory function.[18](#cpt1628-bib-0018){ref-type="ref"} The present study identified the most suitable chain length of polyphosphate for the enhancement of the intestinal barrier function and confirmed the efficacy of long‐chain polyphosphate in inflammation models (including a genetically induced mouse colitis model and inflammation‐induced human macrophages). Furthermore, the safety of long‐chain polyphosphate was confirmed by preclinical tests, and we conducted the first‐in‐human trial to determine the safety and efficacy of long‐chain polyphosphate in patients with refractory UC. Results {#cpt1628-sec-0003} ======= The reinforcement of the intestinal barrier by polyphosphate was dependent on the chain length {#cpt1628-sec-0004} ---------------------------------------------------------------------------------------------- To determine the chain lengths of polyphosphate that might enhance the intestinal barrier function, an *ex vivo* ^3^H‐mannitol study was performed using 12.5, 65, 300, 450, and 700‐mer polyphosphates. The barrier function was enhanced by the polyphosphates in a concentration‐dependent manner (**Figure** [**S1**](#cpt1628-sup-0001){ref-type="supplementary-material"}). The grade of the barrier function when using the 450‐mer polyphosphate did not differ from that when using the 700‐mer polyphosphate (**Figure** [1](#cpt1628-fig-0001){ref-type="fig"} **a**). A transepithelial electric resistance (TER) test was performed to assess the effect of polyphosphates on the epithelial barrier function. The permeability of the Caco2 cell layer was increased by TNF‐α treatment, and both preventive and simultaneous treatments with the 450‐mer and 700‐mer polyphosphates improved the TER at 7.5 hours (**Figure** [1](#cpt1628-fig-0001){ref-type="fig"} **b**), indicating that treatment with the 450‐mer polyphosphate sufficiently increased the intestinal barrier function in comparison with longer polyphosphates. Thus, the 450‐mer polyphosphate was used as the long‐chain polyphosphate for the subsequent studies. The incubation of long‐chain polyphosphate under acidic conditions (pH 1.2) for 2 hours failed to reduce ^3^H‐mannitol leakage (**Figure** [1](#cpt1628-fig-0001){ref-type="fig"} **c**). A high‐performance liquid chromatography spectrum analysis revealed that the retention time of polyphosphate had shifted from 16.701 to 18.826 minutes, indicating that fragmentation of polyphosphate occurred under acidic conditions (**Table** [1](#cpt1628-tbl-0001){ref-type="table"}). These findings suggest that the protection of long‐chain polyphosphates from gastric juice in the stomach, which has a low pH, is necessary for them to effectively enhance the intestinal barrier function. ![Polyphosphate enhanced the intestinal barrier function in a chain length‐dependent manner. (**a**) An *ex vivo* study revealed that the intestinal permeability induced by oxidative stress was significantly attenuated by treatment with 300‐mer, 450‐mer, and 700‐mer polyphosphates (*n* = 5). (**b**) The TER test is a method of assessing the resistance between the upper and lower sides of the transwell *in vitro*. The TER test showed that the permeability induced by TNF‐α treatment was reduced by long‐chain polyphosphate treatment (*n* = 3). (**c**) Long‐chain polyphosphates incubated under acidic conditions (pH 1.2) for 2 hours did not enhance the intestinal barrier function in an *ex vivo* study. \**P* \< 0.05 by Student\'s *t*‐test. The error bars show the standard deviation. BSA, bovine serum albumin; LCP, long‐chain polyphosphate; NT, no treatment; TER, transepithelial electrical resistance; TNF‐α, tumor‐necrosis factor alpha. \[Colour figure can be viewed at <http://wileyonlinelibrary.com>\]](CPT-107-452-g001){#cpt1628-fig-0001} ###### Shortening of polyphosphates under acidic conditions   R. Time (minutes) LCP/BSA (R. Time) 50%‐width (minutes) 10%‐width (minutes) ----------------------- ------------------- ------------------- --------------------- --------------------- LCP 16.701 0.971 1.358 2.948 LCP treated with acid 18.826 1.095 2.107 3.163 BSA 17.195 --- --- --- 10%‐width, width of the wave 10% from the bottom of the wave; 50%‐width, width of the wave 50% from the bottom of the wave; BSA, bovine serum albumin; LCP, long‐chain polyphosphate; R. Time, retention time. John Wiley & Sons, Ltd Long‐chain polyphosphate exerted an antiinflammatory effect in human‐derived M1 macrophages, an IL‐10 KO transfusion mouse model, and a DSS‐induced colitis rat model {#cpt1628-sec-0005} --------------------------------------------------------------------------------------------------------------------------------------------------------------------- To clarify the antiinflammatory effects in active macrophages, M1 macrophages were treated with the polyphosphate and Lipopolysaccharide (LPS). The LPS‐inducible production of TNF‐α in M1 macrophages was inhibited by treatment with 0.1 μm long‐chain to 3.0 μm long‐chain polyphosphates (≥450 mer) but not short‐chain polyphosphates (≤65 mer) ([**Figure S2**](#cpt1628-sup-0002){ref-type="supplementary-material"}), suggesting that long‐chain polyphosphate exerted an antiinflammatory effect as well as enhanced the intestinal barrier function. To clarify the treatment effect of long‐chain polyphosphate *in vivo*, an interleukin (IL)‐10 knockout (KO) transfusion mouse model was constructed. The diarrhea score in IL‐10 KO transfusion mice transanally treated with long‐chain polyphosphate showed a significant improvement, and the colon weight in the IL‐10 KO transfusion group was significantly increased in comparison with the no‐transfusion group but was significantly reduced by the transanal treatment with long‐chain polyphosphate (**Figure** [2](#cpt1628-fig-0002){ref-type="fig"} **a**). The diarrhea score, but not the colon weight, of IL‐10 KO transfusion mice that received oral long‐chain polyphosphate recovered in comparison with the phosphate‐buffered saline (PBS) group (**Figure** [2](#cpt1628-fig-0002){ref-type="fig"} **b**). These data suggested that, in order to achieve sufficient efficacy, it was important to prevent the reduction of the activity of polyphosphates under acidic conditions. To overcome this issue, an enteric‐coated capsule containing long‐chain polyphosphate was generated, and the treatment effect of the capsule was analyzed in a dextran sodium sulfate (DSS)‐induced colitis rat model. The DSS‐treated rats showed shortening of the colon, and recovered following treatment with capsules in a dose‐dependent manner (**Figure** [2](#cpt1628-fig-0002){ref-type="fig"} **c**). A reverse transcription polymerase chain reaction revealed that the messenger RNA (mRNA) expression of proinflammatory cytokines, including TNF‐α and IL‐1β, and the histopathological scores of intestinal damage were reduced by the administration of the capsules (**Figure** [2](#cpt1628-fig-0002){ref-type="fig"} **d,e**). These findings suggest that an enteric‐coated capsule containing long‐chain polyphosphate may be useful for the treatment of intestinal inflammation. {#cpt1628-fig-0002} Preclinical tests {#cpt1628-sec-0006} ----------------- A nonrodent (beagle) repeated‐dose toxicity study using long‐chain polyphosphate (328 and 2,296 mg/body/day) for 4 weeks, a rodent single‐dose toxicity study using long‐chain polyphosphate (40, 200, and 1,000 mg/kg/day) for 2 weeks, a rodent single‐dose toxicity study using long‐chain polyphosphate (10 and 90 mg/body/day), a mutagenicity test, an *in vitro* micronucleus test of long‐chain polyphosphate in TK6 cells, and safety pharmacology studies to investigate the effects of long‐chain polyphosphate on the central nervous and respiratory systems in rats revealed no adverse events (details are shown in the study protocol). First‐in‐human trial {#cpt1628-sec-0007} -------------------- After confirming the effects and safety of long‐chain polyphosphate with preclinical studies, a first‐in‐human trial was conducted. In this study, no adverse reactions associated with the intake of long‐chain polyphosphate were observed on any electrocardiograms, urinary tests, or blood examinations (including abnormal serum levels of phosphates). In step 1, the partial Mayo scores of four of five patients were decreased. Two of the patients achieved a clinical remission (partial Mayo score ≤2) as well as endoscopic remission (Mayo endoscopic subscore ≤1, corresponding to mucosal healing). These two patients who achieved remission had shown either treatment failure or intolerance to anti‐TNF‐α agents. In step 2, the partial Mayo scores of three of five patients were decreased, including two patients with failure or intolerance of anti‐TNF‐α agents. Among the 10 patients in the study population, long‐chain polyphosphate relieved the disease activity in 7 patients, including 4 who achieved endoscopic remission at the end of the 4‐week drug administration period. Long‐chain polyphosphate was able to induce clinical and endoscopic remission in patients with refractory UC regardless of the failure or intolerance of anti‐TNF‐α agents (**Table** [2](#cpt1628-tbl-0002){ref-type="table"}). The mRNA expression of inflammatory cytokines decreased in six of seven patients, thereby relieving their disease activity. In the three patients who did not respond to long‐chain polyphosphate therapy, the partial Mayo score at one week after starting the treatment did not improve, and the mRNA expression of inflammatory cytokines did not decrease (data not shown). ###### The prestudy evaluation and Mayo scores before and after long‐chain polyphosphate treatment Step Case Age Sex Type Treatment history Before treatment After treatment ------ ------ ----- ----------- ----------- ------------------- ------------------ ----------------- ---- --- --- ---- --- --- ---- ---- --- 1 1 58 M Left side D \+ − − 8 5 3 5 3 2 \+ − 2 72 M Proctitis D \+ − − 5 3 2 5 3 2 − − 3 39 F Total D \+ \+ \+ 6 4 2 3 2 1 \+ \+ 4 36 F Left side D \+ \+ − 6 4 2 2 1 1 \+ \+ 5 55 F Left side R \+ − − 7 5 2 5 3 2 \+ − 2 1 69 M Left side D \+ \+ \+ 8 6 2 5 3 2 \+ − 2 53 F Left side D \+ \+ − 6 4 2 3 2 1 \+ \+ 3 19 F Total D \+ \+ − 7 5 2 10 7 3 − − 4 54 F Total D \+ \+ − 7 5 2 5 4 1 \+ \+ 5 44 F Left side R \+ − − 7 5 2 10 7 3 − − 5‐ASA, aminosalicylic acid (mesalamine); Anti‐TNF, anti‐TNF‐α agents; CI, calcineurin inhibitor; CR, clinical response; D, steroid‐dependent; MES, Mayo endoscopic subscore; MH, mucosal healing; PMS, partial Mayo score; R, steroid‐resistant; TMS, total Mayo score. John Wiley & Sons, Ltd Case presentation (Case 4 in step 1) {#cpt1628-sec-0008} ------------------------------------ A 36‐year‐old woman with an onset of UC at 22 years of age visited our hospital because of recurrence with a high frequency of bowel movements and hematochezia. Despite receiving 12 subcutaneous injections of adalimumab, she failed to achieve clinical remission. Before the administration of long‐chain polyphosphate, her partial Mayo score and Mayo endoscopic subscore had been 4 and 2, respectively. After 4 weeks of long‐chain polyphosphate treatment (300 mg, daily) her partial Mayo score and Mayo endoscopic subscore decreased to 2 and 1, respectively (**Figure** [3](#cpt1628-fig-0003){ref-type="fig"} **a**). Colonoscopy revealed an edematous mucosa and multiple small depressions in the sigmoid colon and rectum before long‐chain polyphosphate treatment. These findings disappeared after the administration of the drug (**Figure** [3](#cpt1628-fig-0003){ref-type="fig"} **b**). A clinical remission has been maintained for more than one year since the treatment. {#cpt1628-fig-0003} Discussion {#cpt1628-sec-0009} ========== Probiotics possess sufficient safety to be used for a long time, but their efficacy in UC treatment remains controversial[7](#cpt1628-bib-0007){ref-type="ref"} due to the diversity of intestinal conditions among patients. To resolve the issue and develop a drug that is effective against UC, regardless of intestinal conditions, we developed a method for identifying bioactive molecules from probiotic‐conditioned media[8](#cpt1628-bib-0008){ref-type="ref"} and isolated the bioactive long‐chain polyphosphate molecule from media conditioned with *L. brevis* SB88.[9](#cpt1628-bib-0009){ref-type="ref"} Our findings suggested that long‐chain polyphosphate treatment enhanced the intestinal barrier function and relieved the intestinal injury under inflammatory conditions both *in vitro* and in an IL‐10 KO transfusion mouse model. The function of polyphosphates is known to depend on their chain length. Short‐chain‐length polyphosphates have been shown to enhance the proliferation of human fibroblasts by stabilizing fibroblast growth factors[13](#cpt1628-bib-0013){ref-type="ref"} and exert antibacterial effects against microorganisms.[14](#cpt1628-bib-0014){ref-type="ref"}, [15](#cpt1628-bib-0015){ref-type="ref"}, [16](#cpt1628-bib-0016){ref-type="ref"}, [17](#cpt1628-bib-0017){ref-type="ref"} The present study showed that long‐chain (chain length ≥ 450) polyphosphates improved the intestinal barrier function under conditions of oxidative stress, while short‐chain polyphosphates (chain length ≤ 65) exerted no significant effect on the intestinal barrier function. We therefore developed a procedure for fashioning long‐chain polyphosphates for clinical trials and conducted the first‐in‐human trial to assess the outcomes of long‐chain polyphosphate therapy for refractory UC in 10 patients who responded poorly to currently available treatments. Seven patients achieved a clinical response, and four achieved endoscopic remission after 4 weeks of treatment with long‐chain polyphosphate. The response and remission induction rates were almost the same as those in previous studies using anti‐TNF‐α agents[2](#cpt1628-bib-0002){ref-type="ref"} or recently developed biological drugs, including Janus kinase inhibitors and vedolizumab.[19](#cpt1628-bib-0019){ref-type="ref"}, [20](#cpt1628-bib-0020){ref-type="ref"} Because our human trial was a single‐arm study, it is difficult to compare the effects of long‐chain polyphosphate therapy with a control group, such as the achievement of spontaneous remission. However, Travis, *et al*.[21](#cpt1628-bib-0021){ref-type="ref"} showed in a phase III study of budesonide that the remission rates at 8 weeks were 4.5% among patients in the placebo group. In a phase III study of budesonide, Sandborn, *et al*.[22](#cpt1628-bib-0022){ref-type="ref"} showed that the remission rates at 8 weeks were 6.2% among patients in the placebo group. These recent studies suggested that the spontaneous remission rate was quite low. Thus, the remission rate of 40% in our study is thought to be higher in comparison with the general spontaneous remission rate. It is notable that efficacy was observed in patients with failure or intolerance to anti‐TNF‐α agent therapy, as well as in naïve patients. Our *ex vivo* intestinal loop study and *in vitro* TER test showed that long‐chain polyphosphate improved the intestinal barrier function impaired by not only TNF‐α but also monochloramine (NH~2~CL), suggesting that long‐chain polyphosphate does not directly target TNF‐α. Thus, long‐chain polyphosphate therapy has a function that is distinct from that of anti‐TNF‐α agents and is a feasible alternative to costly biological drugs for treating refractory UC. The drugs that are currently available for the treatment of UC in the clinical setting are all antiinflammatory or immunomodulating agents, and none have directly improved the intestinal barrier function. We previously reported the mechanisms of action of long‐chain polyphosphates on the improvement of intestinal injuries. Briefly, long‐chain polyphosphates bind to the integrin β1 expressed on the intestinal epithelia,[9](#cpt1628-bib-0009){ref-type="ref"} forming a complex with caveolin‐1 (phosphate‐integrin β1‐caveolli‐1 complex). Thereafter, the complex is endocytosed into the cytosol of the intestinal epithelia and p38 mitogen‐activated protein kinase is activated. Then, cell‐adhesion‐associated molecules are strongly upregulated and the intestinal barrier function is increased.[11](#cpt1628-bib-0011){ref-type="ref"} Based on our results, the effect of long‐chain polyphosphate on the intestinal barrier function was thought to be due to the upregulation of cell‐adhesion‐associated molecules. The impairment of the intestinal barrier function is known to lead to the translocation of bacteria and virulence, thereby aggravating the disease status of intestinal disorders, including inflammatory bowel diseases. A recent study showed that mesalamine targeted polyphosphate kinase in pathogenic bacteria but did not show that mesalamine enhanced the intestinal barrier function at all.[23](#cpt1628-bib-0023){ref-type="ref"} Long‐chain polyphosphates may be the first drugs to help prevent the invasion of bacteria and virulence through the direct enhancement of the intestinal barrier function. We also confirmed that long‐chain polyphosphates reduced the expression of inflammatory cytokines, including TNF‐α and IL‐12, in human M1 macrophages and in the colon of DSS‐induced and IL‐10 KO transfusion mouse models. This suggests that long‐chain polyphosphate possesses an antiinflammatory function both *in vitro* and *in vivo*. Regarding the function in relation to macrophages, we previously showed that long‐chain polyphosphates reduced the expression of TNFα in LPS‐stimulated THP‐1 cells that were pretreated with macrophage colony‐stimulating factor and IFNγ. We also confirmed that long‐chain polyphosphate improved the intestinal injury in a 2,4,6‐Trinitrobenzenesulfonic acid--induced colitis model[10](#cpt1628-bib-0010){ref-type="ref"} as well as a DSS‐induced colitis model. While the mechanisms of the antiinflammatory effect are still being explored, our previous study showed that long‐chain polyphosphates might target M1 macrophages. Based on these findings in experimental models, long‐chain polyphosphates are expected to exhibit an antiinflammatory effect in patients with inflammatory bowel diseases. Regarding the safety of long‐chain polyphosphates, preclinical studies, including nonrodent repeated‐dose toxicity studies, rodent single‐dose toxicity studies, mutagenicity tests, and safety pharmacology tests, revealed no adverse events. In addition, the present study showed no adverse events in 10 UC patients with active inflammation. These data suggest that long‐chain polyphosphates are a safe drug. In conclusion, long‐chain polyphosphate was safe and effective for the treatment of refractory UC in the first‐in‐human study. The next phase of this study has been prepared and will provide stronger evidence concerning the safety and efficacy of long‐chain polyphosphate in the treatment of UC, including appropriate targets for treatment. Materials and Methods {#cpt1628-sec-0010} ===================== Animals {#cpt1628-sec-0011} ------- Details are provided in the [**Supplementary Material**](#cpt1628-sup-0004){ref-type="supplementary-material"}. Polyphosphate {#cpt1628-sec-0012} ------------- Polyphosphates with a length of 12.5 (BENX‐PolyP‐S), 65 (BENX‐PolyP), and 700 (BENX‐EPolyP) mers were purchased from BIOENEX, Higashihiroshima, Japan, while 300‐mer and 450‐mer polyphosphates were obtained by enzymatic reactions.[9](#cpt1628-bib-0009){ref-type="ref"} Calculating the chain length of the polyphosphates {#cpt1628-sec-0013} -------------------------------------------------- Polyphosphates were incubated in reaction buffer containing 50 mM Tris‐HCl pH 7.5, 0.01% N,N‐Bis(4‐sulfobutyl)‐3‐methylaniline disodium salt, 0.01% 4‐aminoantipyrine, 2 mM adenosine diphosphate, 2 mM MgCl~2~, 10 mM glycerol, 10 U/mL glycero‐3‐P oxidase, 5 U/mL peroxidase, 1 U/mL glycerol kinase, and 0.1 mg/mL polyphosphate kinase at 37°C for 10 minutes. Two reaction volumes of 2% sodium dodecyl sulfate were then added, and detection was performed at an optical density of 550 nm. The chain length of the polyphosphates was calculated using a 700‐mer polyphosphate as the standard. The HPLC spectrum analysis of the polyphosphates {#cpt1628-sec-0014} ------------------------------------------------ The polyphosphates were separated using an HPLC system (Shimadzu Corporation, Kyoto, Japan) with a TSKgel GMPW XL (φ7.8 mm × 300 mm) column at 30°C and eluted with 10 mM phosphate buffer containing 0.5 M sodium chloride, at a flow rate of 0.5 mL/minute. The eluent was monitored by ultraviolet spectrophotometry at 260 nm. Bovine serum albumin was used as a standard. TER test {#cpt1628-sec-0015} -------- Details are provided in the [**Supplementary Material**](#cpt1628-sup-0004){ref-type="supplementary-material"}. Human peripheral blood‐derived macrophages {#cpt1628-sec-0016} ------------------------------------------ Details are provided in the [**Supplementary Material**](#cpt1628-sup-0004){ref-type="supplementary-material"}. The *ex vivo* intestinal loop study {#cpt1628-sec-0017} ----------------------------------- C57Bl/6 mice were euthanized, and the small intestine was divided into three pieces, with each end ligated with silk sutures and the loops filled with RPMI 1640 medium containing the sample. The loops were incubated for 2 hours at 37°C in a 5% CO~2~ incubator. To assess the permeability effects, the loops were filled with RPMI 1640 medium containing 1 mCi/mL \[3H\]‐mannitol with or without 0.3 mM freshly prepared NH~2~Cl. The loops were placed into the middle section of the organ culture dish in 4 mL of RPMI 1640 without NH~2~Cl. To assess the permeability, samples were taken at 5, 20, and 35 minutes to determine the flux of mannitol from the lumen to the media outside the loops. Radioactivity was measured using a scintillation spectrometer. Isotope‐labeled polyphosphate {#cpt1628-sec-0018} ----------------------------- Details are provided in the [**Supplementary Material**](#cpt1628-sup-0004){ref-type="supplementary-material"}. Pharmacokinetics study {#cpt1628-sec-0019} ---------------------- Mice were treated with ^32^P‐labeled long‐chain polyphosphate (500 μg) orally or through the duodenum and then euthanized 2, 5, or 24 hours later. The organs, stool, urine, inferior vena cava blood, and portal blood were collected at euthanizing, and the radioactivity was measured using a scintillation spectrometer. Enteric‐coated capsules {#cpt1628-sec-0020} ----------------------- Details are provided in the [**Supplementary Material**](#cpt1628-sup-0004){ref-type="supplementary-material"}. DSS‐induced colitis {#cpt1628-sec-0021} ------------------- Details are provided in the [**Supplementary Material**](#cpt1628-sup-0004){ref-type="supplementary-material"}. The real‐time polymerase chain reaction {#cpt1628-sec-0022} --------------------------------------- Details are provided in the [**Supplementary Material**](#cpt1628-sup-0004){ref-type="supplementary-material"}. The IL‐10 KO‐derived T‐lymphocyte transfusion (IL‐10 KO transfusion) mouse model {#cpt1628-sec-0023} -------------------------------------------------------------------------------- CD4‐positive cells were purified from the spleen, mesenteric lymph node, and sacral lymph node of IL‐10 KO mice. A total of 5 × 10^6^ CD4‐positive cells were suspended in 1 mL PBS, and 0.2 mL/mouse of cell suspension was transferred to SCID mice. Starting on day 1, long‐chain polyphosphate or PBS was transanally administered to the mice every day. The diarrhea score was assessed at day 15. The scores were determined as follows: score 0, solid stool that was not deformed when pushed by finger; score 1, solid stool that was deformed when pushed by finger; score 3, solid stool that was easily collapsed when pushed by finger; score 3, soft stool that was easily deformed and stuck to container; score 4, muddy or watery stool. The mice were euthanized on day 15 under isoflurane anesthesia, the colon was sampled, and the weight of the colon was measured. First‐in‐human trial {#cpt1628-sec-0024} -------------------- We designed an investigator‐initiated, nonrandomized, open‐label pilot study of long‐chain polyphosphate (University Hospital Medical Information Network number, UMIN000021555) involving 10 adult patients with refractory UC. The study was approved by the ethics committee of Asahikawa Medical University (Number, 15189), and each patient provided their written informed consent. We developed a double‐capsule formulation using size 0 and 2 DRcaps capsules (Capsugel Japan, Tokyo, Japan) containing 100 mg of long‐chain polyphosphate in a capsule. Ten UC patients, including two steroid‐resistant and eight steroid‐dependent cases (refractory UC), who visited Asahikawa Medical University Hospital from May 2016 to December 2017 and who showed a disease activity with a Mayo score ranging from 3 to 8 with obvious hematochezia were enrolled in this study. The first five patients orally took 300 mg of long‐chain polyphosphate once daily for 4 weeks (step 1); the second five patients orally took 900 mg of long‐chain polyphosphate once daily for 4 weeks (step 2). Anti‐TNF‐α agent treatments were fixed at 20 weeks before taking the study drugs, and other medications were fixed 2 weeks before taking the study drugs. All treatments used before the enrollment were continued during the study. In the patients enrolled in the present study, biologics were fixed ≥20 weeks before taking the study drugs, and all other medications were fixed ≥4 weeks before taking the study drugs, except for antibiotics, which were fixed 2 weeks before taking the study drugs in one case (case 2 in step 2). Therefore, none of the enrolled patients were excluded due to concomitant medications. The disease activity of the enrolled patients was assessed using the Mayo score (the calculation of the Mayo and partial mayo scores is described in the [**Supplementary Methods**](#cpt1628-sup-0003){ref-type="supplementary-material"}). The partial Mayo score was assessed before starting the study and every week until the administration schedule was completed. The Mayo endoscopic subscore was assessed before and after the study. Before enrollment, pan‐colonoscopy was performed in all patients to determine the most severe inflammation site, which was set as the assessment area. The severity of the inflammation was then assessed based on the Mayo endoscopic subscore. After long‐chain polyphosphate treatment, the subscore of the assessment area evaluated by initial colonoscopy was reassessed. Pan‐colonoscopy was performed as second colonoscopy only if the endoscopist concluded that other sites needed to be evaluated. The mRNA expression of inflammatory cytokines, including TNF‐α, IFN‐γ, and IL‐1β, in biopsy samples obtained from the rectosigmoid colon were measured before and after this study. The primary endpoint was the safety of the long‐chain polyphosphate, and the secondary endpoint was the efficacy of this drug in patients with refractory UC. Monitoring and inspection of this study were performed by a third‐party organization (MRS Planning, Tokyo, Japan). Steroids, mesalamine, anti‐TNF‐α agents, and calcineurin inhibitors were used by 10, 10, 6, and 2 of the enrolled patients, respectively, before their entry in this study. In step 1, the disease activity was classified as follows: Mayo 5, *n* = 1; Mayo 6, *n* = 2 patients; Mayo 7, *n* = 1 patient; and Mayo 8, *n* = 1. In step 2, the activity at the beginning of this study was classified as Mayo 6, *n* = 1; Mayo 7, *n* = 3; and Mayo 8, *n* = 1 (**Table** [1](#cpt1628-tbl-0001){ref-type="table"}). Blood examinations performed before starting this study revealed no abnormalities, with the exception of slight C‐reactive protein elevation in three patients. Statistical analyses {#cpt1628-sec-0025} -------------------- Details are provided in study protocols of the first‐in human trial. Funding {#cpt1628-sec-0027} ======= This study was partially supported by Ministry of Education, Culture, Sports, Science and Technology Grants‐in‐Aid for Scientific Research, No. 26460956 (M.F.), Translational Research Network Program of Japan Agency for Medical Research and Development, No. C40 (M.F.), and Development and Intractable Disease Health and Labour Sciences Research Grants from the Ministry of Health, Labour and Welfare. This study was partially supported as a joint research project of Asahikawa Medical University and EA Pharma Co., Ltd. Conflict of Interest {#cpt1628-sec-0028} ==================== M.F. reports grants from Japanese Grants‐in‐Aid for Scientific Research, grants from Translational Research Network Program of Japan Agency for Medical Research and Development, nonfinancial support from Development and Intractable Disease Health, and Labour Sciences Research Grants from the Ministry of Health, Labour and Welfare. During the conduct of the study: grants and personal fees from Yakult Honsha Co., Ltd, grants and personal fees from Nippon Kayaku Co., Ltd., grants, personal fees, and nonfinancial support from EA Pharma Co., Ltd., personal fees from Novartis Pharmaceuticals, grants from Boehringer Ingelheim GmbH, personal fees from Technical Information Institute Co., Ltd, personal fees from Pfizer Inc., grants and personal fees from Takeda Pharmaceutical Company Limited, grants and personal fees from Daiichi Sankyo Company, Limited, grants from Shionogi & Co., Ltd, personal fees from Olympus Co., Ltd., personal fees from Janssen Pharmaceutical K.K., personal fees from Kyorin Pharmaceutical Co., Ltd., personal fees from Taisho Toyama Pharmaceutical Co., Ltd., grants from AstraZeneca, personal fees from Tomakomai Minpo Co., Ltd., personal fees from Meiji Seika Pharma Co., Ltd., grants and personal fees from Mochida Pharmaceutical Co., Ltd, grants from Astellas Pharma Inc., personal fees from Mitsubishi Tanabe Pharma Corporation, grants from Baxter International Inc, personal fees from Bristol‐Myers Company, grants and personal fees from Chugai Pharmaceutical Co., Ltd., grants from MSD K.K., grants and personal fees from Otsuka Pharmaceutical Co., Ltd, grants and personal fees from Kyowa Hakko Kirin Co., Ltd., grants from Taiho Pharmaceutical Co., Ltd, personal fees from Zeria Pharmaceutical Co., Ltd., grants from Alexion Pharmaceutical Co., Ltd., grants, personal fees, and nonfinancial support from Ajinomoto Pharmaceutical Co., Ltd., personal fees from Asahi Kasei Corporation, grants and nonfinancial support from Sapporo Breweries Ltd., grants and personal fees from Eisai Co., Ltd., grants from GlaxoSmithKline K.K., grants from AbbVie Inc, grants from Sumitomo Dainippon Pharma Co., Ltd., grants from Asuka Pharmaceutical Co., Ltd., grants from Boston Scientific Corporation or its affiliates, grants and nonfinancial support from Kamui Pharma. Inc., outside the submitted work. In addition, M.F. has been issued patents for intestinal protectants (JP 5660508 and EP 2559437). The other authors declared no conflicts of interest in association with the present study. Author Contributions {#cpt1628-sec-0029} ==================== M.F. and N.U. wrote the manuscript; M.F. and N.U. designed the research; M.F, N.U., S.K., K.T, K.A., Ke.M., H.K., N.K., Y.T., Tom.O. and Ka.M. performed the research; M.F, N.U., A.S., M.T., and Tos.O. analyzed the data. Supporting information ====================== ###### **Figure S1.** The effects of several concentrations of long‐chain polyphosphate on the improvement of the intestinal barrier function. ###### Click here for additional data file. ###### **Figure S2.** Long‐chain polyphosphate inhibited the anti--tumor necrosis factor--α production in Lipopolysaccharide‐stimulated M1 macrophages. ###### Click here for additional data file. ###### **Supplementary Methods** ###### Click here for additional data file. ###### **Clinical Trial Protocol** ###### Click here for additional data file. We thank Miyuki Sato, Kuninori Iwayama, Naoya Kamiyama, Takashi Ono, and other members of the Department of Hospital Pharmacy and Pharmacology and the Clinical Research Support Center of Asahikawa Medical University, Professor Hiroshi Funakoshi, Professor Yasuaki Saijyo, and Professor Nobuyuki Sato, and Atsushi Yonezawa and Takayuki Nakagawa and other members of the Department of Clinical Pharmacology and Therapeutics of Kyoto University Hospital for their great support of this study. [^1]: MF and NU contributed equally to this study.

Background ========== The obesity epidemic in Australia is generally monitored by the proportion of the population whose body mass index (BMI) exceeds 30 kg/m^2^\[[@B1]\]. However, what is not as evident is the growing proportion of those who are morbidly obese. This issue is important as adverse health risks and healthcare expenditure amplify as the level of obesity increases \[[@B2]\]. Obesity-related health expenditure in Australia exceeded \$8 billion in 2008, which included expenditure associated with metabolic disease, cardiac disease and surgical complications \[[@B3]\], but it is unclear how much of this expenditure related to different levels of obesity. Service providers are increasingly having to invest in equipment designed to accommodate the morbidly obese. It seems that both direct and indirect costs will amplify as the prevalence of morbid obesity increases. In Australia, few, if any, studies have described the pattern of morbid obesity in the population. Thus, we aimed to document changes in the prevalence of morbid obesity (BMI ≥ 40.0 kg/m^2^) that have occurred over a decade among women living in south-eastern Australia. Methods ======= Subjects -------- The Geelong Osteoporosis Study (GOS) is a population-based cohort study of adults aged 20 years and over, randomly-selected from the Commonwealth electoral rolls for the Barwon Statistical Division in south-eastern Australia \[[@B4]\]. In Australia, the electoral roll is a register of all eligible voters and provides the most comprehensive available list of adult Australian citizens. At baseline, an age-stratified sample was recruited so that there were approximately 100 individuals for each 5-year age-group 20--24, 25--29, 30--34, 35--39, 40--44, 45--49, 50--54, 55--59, 60--64 and 65--69 years, and approximately 200 individuals for each of the age groups 70--79 years and 80 years and older. Baseline assessments for women enrolled in the GOS occurred during the period 1993--7 (1,494 recruited, 77% response) and again approximately a decade later, 2004--8 (882 of the eligible women were assessed at the 10-year follow-up with 82% response). A further sample of 194 women aged 20--29 years was also randomly generated using the same methods employed for the baseline recruitment (2005--8, 82% response) and included in this study. Thus, data from 1,076 women were included in analyses for the 2004--8 time period. Most of the cohort (99%) was Caucasian; details of the study have been provided elsewhere \[[@B4]\]. All participants gave written, informed consent. The Barwon Health Human Research Ethics Committee approved the study. Weight was measured to the nearest 0.1 kg using electronic scales, standing height was measured to the nearest 0.001 m using a wall-mounted stadiometer and BMI expressed as weight/height^2^ (kg/m^2^). An adult with a BMI of 40.0 kg/m^2^ or over is described as being morbidly obese (class III obesity) but the term can also refer to those who have a BMI over 35 kg/m^2^ with obesity-related co-morbidities \[[@B5]\]. We identified morbid obesity using BMI criteria alone. Class II obesity corresponded to BMI 35.0-39.9 kg/m^2^, class I obesity to BMI 30.0-34.9 kg/m^2^, overweight to BMI 25.0-29.9 kg/m^2^, normal weight to BMI 18.5-24.9 kg/m^2^ and underweight to BMI \<18.5 kg/m^2^\[[@B6]\]. Area-based socioeconomic status (SES) was determined using the Socio-Economic Index for Areas (SEIFA) values based on census data (for 1996 for 1993--7; and for 2006 for 2004--8) from the Australian Bureau of Statistics (ABS). SEIFA values were used to assign an Index of Relative Socioeconomic Disadvantage, which was categorised into quintiles according to cut-off points for the study region. Statistical analysis -------------------- In order to compare the prevalence of each BMI category in 1993--7 and 2004--8, both time period datasets were stratified into age-groups and age-standardised to the 2006 census population figures for the Barwon Statistical Division (ABS Catalogue No. 2001.0). The standardized morbidity ratio (SMR) for morbid obesity was calculated using the 2004--8 dataset as the standard. The distribution of BMI was skewed, but was normalised by natural log-transformation. Differences in mean BMI between the two time periods were determined using linear regression models. The variable of interest was the period of assessment (1993--7 or 2004--8) and the models were adjusted for age and SES tested as a potential confounder. Differences in the likelihood of morbid obesity during the two time periods were determined using logistic regression models. The outcome was morbid obesity and the exposure variable of interest was the period of assessment; models were adjusted for age and SES tested as a confounder. This method enables a comparison of the likelihood of morbid obesity between women of the same age and SES in the different time periods. In the statistical models, interaction terms were tested as effect modifiers and retained in the model if p \< 0.05. Statistical analyses were performed using Minitab (version 15; Minitab, State College, PA). Results ======= Subject characteristics are shown in Table  [1](#T1){ref-type="table"}. The calculated age-adjusted geometric mean BMI increased from 26.0 kg/m^2^ (95%CI 25.7-26.3) in 1993--7, to 27.1 kg/m^2^ (95%CI 26.8-27.4) in 2004--8. The increase was observed across the age range, as evidenced by a non-significant age\*time-period interaction in the regression model. Similarly, no interaction was detected between SES and the time period, indicating that mean BMI increased consistently across the range of SES. ###### Subject characteristics for the two periods of assessment, 1993--7 and 2004--8   **1993-7** **2004-8** ------------------------------------------- ------------------ ------------------ Age (years); median (interquartile range) 54.2 (37.4-71.6) 51.1 (34.7-65.9) BMI categories\*; n (%)     morbid (class III) obesity 34 (2.3%) 44 (4.1%) class II obesity 75 (5.0%) 81 (7.5%) class I obesity 227 (15.2%) 191 (17.8%) overweight 494 (33.1%) 333 (31.0%) normal weight 635 (42.5%) 412 (38.3%) underweight 29 (1.9%) 15 (1.4%) \*BMI (kg/m^2^) for morbid (class III) obesity ≥40.0, class II obesity 35.0-39.9, class I obesity 30.0-34.9, overweight 25.0-29.9, normal weight 18.5-24.9, underweight \<18.5. For the period 1993--7, 34 (2.3%) women were identified as morbidly obese (class III obesity), 75 (5.0%) as class II obesity and 227 (15.2%) as class I obesity; there were 494 (33.1%) women classified as overweight; 635 (42.5%) as having normal weight and 29 (1.9%) as underweight (Table  [1](#T1){ref-type="table"}). A decade later, 2004--8, 44 (4.1%) women were morbidly obese (class III obesity), 81 (7.5%) were class II obesity, 191 (17.8%) were class I obesity, and 333 (31.0%) were overweight, 412 (38.3%) were of normal weight and 15 (1.4%) were underweight. The odds for morbid obesity were 76% greater in 2004--8 compared to 1994--7 (OR = 1.76, 95%CI 1.12-2.78). This association was sustained after adjusting for SES (data not shown) and found to be consistent across SES categories. The age-standardised prevalence for morbid obesity (class III obesity) was 2.5% for 1993--7 and 4.2% for 2004--7; prevalence figures for class II obesity were 5.6% and 7.3%, for class I obesity 15.1% and 17.6%, for overweight 31.8% and 30.8%, for normal weight 43.4% and 38.8%, and for underweight 1.6% and 1.3% (Figure  [1](#F1){ref-type="fig"}). The SMR for morbid obesity (class III obesity) was 1.69 (95%CI 1.26-2.27). {#F1} Discussion ========== We report that mean BMI has increased by 4.2% among women during the decade between 1993--7 and 2004--8. A comparison of the prevalence of morbid obesity in 2004--8 with 1993--7 generated a SMR for morbid obesity between 1.26 and 2.27, demonstrating a measurable increase over this ten year time period. Because of the approximate bell-shaped distribution of BMI in the population, a shift of the distribution to higher BMI resulted in a marked proportional increase in the prevalence of the more extreme BMI values. Disproportionate increases in the more extreme categories of BMI have been similarly reported for adults in the USA between 2000 and 2005 \[[@B7]\]. Our results indicate that the increase in the proportion of morbid obesity was not restricted to particular ages or different SES groups. Greater adiposity, and unhealthy lifestyle choices associated with increased risk for obesity, have been observed with increasing social disadvantage \[[@B8]\]. It is curious that our current study showed an increase in the prevalence of morbid obesity to have occurred across the SES spectrum. We may speculate that the increase in morbid obesity affects the population as a whole, thus shifting the entire population distribution toward greater risk for obesity-related disease. However, given the small number of morbidly obese women in our study, there may have been insufficient power to conclusively test for SES differences and it is plausible that a larger sample size could show inequity in the rise of morbid obesity between SES groups. Either way, our data suggest that population-wide approaches that change the underlying conditions in which behaviours occur could be warranted to combat rising rates of morbid obesity for all sub-groups of the population. The strength of this study lies in the random nature of the study population and that the anthropometric measures used to calculate BMI were measured, rather than self-reported which would increase the potential for misclassification \[[@B9],[@B10]\]. However, it is possible that participation bias may have changed over time. It is not unreasonable that social stigma linked with obesity \[[@B11]\] may change as the obesity epidemic progresses, and we cannot exclude the possibility that this might have differentially impacted on participation at the two time periods. Moreover, limitations of BMI as an indicator of adiposity have previously been demonstrated, especially for individuals with muscular body builds and the elderly \[[@B1]\], and this was not considered in the analyses. Our study findings highlight the growing problem of morbid obesity associated with a population-wide increase in BMI. Recently, there has been increased advocacy in Australia for raising awareness of nutrition-related diseases and integrating nutrition into medical curricula \[[@B12]\]. In further recognition of issues surrounding the clinical management of morbidly obese patients, a Morbid Obesity model of care was developed in Western Australia in 2007 \[[@B13]\]. It should be clear, however, that unless the modern obesogenic environment is addressed, obesity-related health problems and healthcare costs are likely to escalate. In 2009, a parliamentary committee developed a series of recommendations designed to combat the rising rates of obesity in Australia \[[@B3]\]. Recommendations included development of the built environment to promote physically active lifestyles, education to encourage healthy eating and to increase the consumption of fresh fruit and vegetables, reformulation of commercially-available foods, more effective food labelling to assist healthy consumer choices and the development of a registry of bariatric surgery. While surgery and other gastrointestinal interventions may be considered appropriate for individuals with severe obesity and co-morbid disease, such as type 2 diabetes \[[@B5]\], such treatment requires a thorough multidisciplinary team assessment of peri-operative risk and possible long-term complications \[[@B14]\] and is unlikely to impact the rising prevalence of morbid obesity at a population level. The challenge is to identify and implement effective strategies that will shift the population distribution towards lower BMI, if the disproportionate increase observed in the extreme levels of obesity is to be curbed. Conclusions =========== We conclude that over a period of a decade, there has been an increase in mean BMI among women residing in south-eastern Australia, resulting in a measurable rise in the prevalence of morbid obesity. Abbreviations ============= ABS: Australian Bureau of Statistics; BMI: Body mass index; GOS: Geelong Osteoporosis Study; SEIFA: Socio-Economic Index for Areas; SES: Socioeconomic status; SMR: Standardized morbidity ratio Competing interests =================== No relevant disclosures. Authors' contributions ====================== JAP conceived of the study, developed the methods, conducted the literature search and led the development of the final manuscript. SLB and MAK contributed to the study conception, design and data interpretation, and provided intellectual content to the final manuscript. All authors read and approved the final manuscript. Pre-publication history ======================= The pre-publication history for this paper can be accessed here: <http://www.biomedcentral.com/1471-2458/13/290/prepub> Acknowledgements ================ The study was funded by the National Health and Medical Research Council (NHMRC) of Australia, the Victorian Health Promotion Foundation (VHPF) and Geelong Regional Medical Foundation, but they played no part in the design or conduct of the study; collection, management, analysis, and interpretation of the data; or in preparation, review, or approval of the manuscript. SLB is supported by NHMRC Early Career Fellowship (1012472).

Introduction ============ Many students in higher education are not successful and encounter academic failure ([@B66]). One of the factors associated with academic failure is academic procrastination ([@B57]; [@B29]), although there are also instances in which the pressure created by procrastination can improve performance ([@B28]). When students procrastinate, they often experience problems with learning activities, for example, with starting on time to prepare for exams and dealing with deadlines for assignments. In the present study, we define procrastination as *the voluntary delay of an intended and necessary and/or (personally) important activity, despite expecting potential negative consequences that outweigh the positive consequences of the delay* ([@B57]; [@B31], p. 26). In the context of the present study, the "important" activities are learning activities. Various studies clarified that academic procrastination is common among college students. [@B51] conducted a study among 278 Dutch undergraduate college students, showing that more than 70% reported academic procrastination. About 20% reported chronic academic procrastination. In a meta-analysis, [@B57] cited research showing that 80--90% of undergraduate college students report that they experience procrastination in some form ([@B10]; [@B39]). [@B40] investigated the prevalence of academic procrastination in 203 Turkish undergraduate college students and found that 52% of them reported frequent academic procrastination. Although the estimated percentages of procrastination in these studies differ, academic procrastination clearly is a serious problem and has negative consequences. Students who procrastinate regularly receive lower course grades and lower final exam grades ([@B69]; [@B57]), and these students are less successful in their degree programs ([@B65]). The present study examines learning characteristics of students with different levels of academic procrastination and how they deal with factors that possibly influence their learning. We conceptualize learning characteristics as those features of learning processes that students themselves indicate as relevant to their procrastination behavior, as we were interested in students' views of factors influencing their learning. Factors Influencing Procrastination ----------------------------------- Academic procrastination can be seen as a prevalent and pernicious form of self-regulatory failure ([@B57]). Self-regulation is a process through which the student activates and maintains thoughts, feelings, and behavior necessary for achieving personal goals ([@B72]). Differences in self-regulation among students contribute considerably to differences in students' academic achievements ([@B71]) and levels of procrastination ([@B57]; [@B30]; [@B21]). In the self-regulation process, three different phases can be distinguished, when a student sets himself or herself to a given learning activity ([@B53]; [@B43]). The first phase is the forethought and planning phase. In this phase, the student plans his or her study activity, while various thoughts regarding motivation, values, and goals are active. The second phase is the monitoring performance and motivation phase. In this phase, the student has already started the study activity and tries to control his or her motivation and academic performance. For example, during the study activity the student can decide to change his or her learning strategy because it does not have the desired effect. The third phase is the phase after completing the study activity. This is the reflection on performance phase. During this reflection phase, the student attempts to understand why a certain result occurred and manages his or her emotions with respect to this result. In each of the self-regulation phases, procrastination problems can occur [@B22]. For example in the first phase, students with low levels of self-efficacy (e.g., [@B68]) and deficient time management and goal-setting skills ([@B34]) have more procrastination problems. During the performance phase, procrastinating students show low perseverance ([@B21]) and high susceptibility to social temptations ([@B9]). In the self-reflection phase, academic procrastinators tend to make external and global attributions regardless of whether their performance was a success or not ([@B14]). ### Personal Factors Procrastination is influenced by several personal factors. Factors that may increase the tendency to procrastinate include anxiety ([@B56]), depression ([@B61]), self-handicapping (*r* = 0.53; [@B15]), the Big Five factor neuroticism (*r* = 0.26; [@B63]), fear of failure (*r* = 0.63; [@B15]), perceived competence ([@B24]), and pessimistic academic attributional style (*r* = 0.14; [@B65]). There are also factors that may decrease the tendency to procrastinate. These factors include self-esteem (*r* = -0.28; [@B62]), self-efficacy (*r* = -0.54; [@B15]), self-control (*r* = -0.58; [@B57]), and the Big Five factors extraversion (*r* = -0.08), agreeableness (*r* = -0.10), openness (*r* = -0.15), and conscientiousness (*r* = -0.65; [@B63]). Volitional control, social relatedness, and task competence are also important aspects influencing students' procrastination ([@B32]). When students have to do a learning activity for their degree program, it is important that they have a certain level of the executive functioning domains of initiation, plan/organize, organization of materials, and task monitoring ([@B44]). There is a higher risk of academic procrastination if students have poor planning skills ([@B44]), a reduced use of cognitive and meta-cognitive learning strategies ([@B26]), or a low level of perseverance, and if the students are easily distracted ([@B9]). It is also important that students are motivated for the learning activity ([@B21]). Motivation is the force that drives a person to engage in activities ([@B2]) and concerns energy, direction, persistence, and all aspects of activation and intention ([@B46]). Student motivation can be distinguished ([@B46]) as intrinsic motivation, that is, motivation resulting from internal drives (doing a learning activity for the pleasure it brings or because of interest) and extrinsic motivation, that is, motivation resulting from external factors (doing something for an external reason; for example, the student wants to maintain a good relationship with the teacher). High intrinsic motivation reduces academic procrastination ([@B35]; [@B57]). Students have a mastery-approach goal orientation ([@B26]) when they focus on learning, mastering the task according to self-set standards, or when they improve themselves and develop new competencies. Mastery goal orientation is an important factor that prevents academic procrastination ([@B54]). When a learning activity does not intrinsically motivate a student, he or she has to actively transform external regulation into internal regulation and shift from external control of his or her behavior to internal control of his or her behavior. This is the so-called process of internalization ([@B46]). ### Situational Factors Academic procrastination can also be influenced by situational factors. An important situational factor is teachers, because if teachers are well-organized, it is easier for students to organize, structure, and plan their work ([@B7]). Unorganized and lax teachers can be a reason for students' procrastination ([@B21]). Procrastination is also promoted by teachers when they expect less, are willing to negotiate deadlines, and are more flexible in their grading ([@B52]). Teachers with high expectations increase students' class enjoyment and interest and diminish students' procrastination ([@B7]). When a teacher provides clear instructions for assignments, students procrastinate less ([@B1]). When students have to do study tasks, procrastination can be evoked and maintained by task characteristics ([@B38]). A task perceived as boring, unpleasant, and/or uninteresting (so-called task aversiveness) strongly predicts students' procrastination ([@B3]; [@B57]). At the start of a project, task aversiveness is related to personal meaning, such as pleasure, fun, enjoyment, and communion. When a task is perceived as interesting or requires students to use a variety of skills, and when students perceive social norms and rewards for starting promptly, students procrastinate less ([@B1]). An optimal degree of task difficulty is reached if a task is sufficiently challenging but also achievable ([@B62]; [@B57]). Differences in How Students Deal With Procrastination ----------------------------------------------------- Thus, various factors could influence students' procrastination behavior. However, there also can be differences between students in how they are influenced by these factors. Interviews with experienced university counselors showed that antecedents and consequences of academic procrastination are largely a reflection of students' characteristics, personal and learning situations, and the environment of the university ([@B42]). There can also be differences between students' delaying behavior and having academic procrastination, because delaying a task can be an intentional decision. For example, students intentionally delay because of other priorities or because they feel they can work better when they set aside a designated period of time ([@B36]). [@B21] showed that students who are pressure-seeking types are not negatively affected by academic delays and have low academic procrastination. An interesting study for insights into differences between students with different levels of academic procrastination was conducted by [@B38]. Their study showed that differences in students' academic procrastination are associated with the manner in which students deal with environmental factors that might provoke procrastination behavior. Academic procrastination (in the form of choosing social activities when intending to do academic work) and the evaluations of other students' procrastination habits were higher in students in the field of humanities compared to the field of natural sciences and medicine. Differences between students were moderated by the dispositional tendency to procrastinate. The humanities students demonstrated a higher level of socially induced academic procrastination and evaluated other students' procrastination habits (peer procrastination) more than students in medicine and natural sciences did. Environmental factors had a negligible impact on academic procrastination on students with a low disposition to procrastinate. For students with medium levels of dispositional procrastination, which are the majority, procrastination-friendly environments facilitate and augment the students' procrastination. Research Gap ------------ Academic procrastination is a highly complex human behavior that involves a combination of affective, cognitive, and behavioral components ([@B4]; [@B6]). Despite the extensive, mainly quantitative, research concerning procrastination, there is still a lack of understanding of why students procrastinate ([@B57]; [@B32]; [@B27]). Because we wanted to get insights into differences among students with high, average, and low procrastination, we selected students with different measured levels of procrastination. This is an important difference compared with previous qualitative studies on academic procrastination. Most previous qualitative studies about academic procrastination (e.g., [@B21]; [@B32]; [@B36]) included students with high academic procrastination. The selection of the respondents in those studies was not based on students' measured levels of academic procrastination, but students were, for example, approached on the university campus restaurant by the interviewers and asked whether they would be willing to participate in an interview study concerning academic procrastination ([@B32]). In other studies, flyers were distributed on the university campus with a call to participate in the study ([@B21]), or students were asked to fill out a questionnaire between lectures or students were contacted through the course forum when they did not attend lectures ([@B38]). The Present Study ----------------- Therefore, we know that factors influencing procrastination can work differently among students from different academic disciplines and among students with different procrastination levels ([@B38]). We do not know how the process behind these differences works. In the present study, we wanted to dig deeper into the process behind students' academic procrastination and how students deal with factors that might influence their learning. We see the present study as an important continuation of [@B38] study in which they concluded that social and environmental factors should receive increased attention and suggested that future studies should include more diverse measures. We conducted the present study from a qualitative perspective, because students' views ([@B32]) could show more about their learning characteristics whether and how they deal with procrastination. With this study, we wanted to provide new insights for theoretical groundwork in procrastination research and provide insights that can be used for developing interventions that teach students how to handle their procrastination. This is important for educational institutions and people in these institutions who work with students, such as teachers, counselors, and educational psychologists. Research Question ----------------- The following research question guided the present study: what are differences between students with low, average, and high levels of academic procrastination in learning characteristics and in factors that might influence students' learning? Materials and Methods {#s1} ===================== To our knowledge, the present study is the first study to explore differences in learning characteristics and in the influential factors on academic procrastination among students with low, average, or high academic procrastination tendencies. We conducted a qualitative interview study because we aimed to explore this topic without preliminary hypotheses and wanted to get insight into students' actual experiences and regulation of learning activities, and how they relate to the students' study practices. Sampling -------- In this study, we interviewed 22 students (7 men, 15 women, age between 16 and 22 years; *M* = 17.7, *SD* = 1.4). All students were freshmen in a full-time 4-year elementary teacher education program at a small teachers college with approximately 1,500 students, situated in the western region of The Netherlands. To select the participants for the interviews, we first wanted to know at what level students actually experienced academic state procrastination in the week before their exam period. Therefore, we measured the students' academic procrastination with the Academic Procrastination State Inventory (APSI; [@B50], unpublished). The APSI is an existing Dutch questionnaire that meets the requirements of internal reliability and validity (α = 0.94) and has been used in previous research (e.g., [@B25]). In the APSI, the student is asked about his or her study behavior during the week before completing the questionnaire. Each of the 31 items begins with the question "How often did you... last week?" On a five-point Likert scale from never (1) to always (5), the student indicates his or her assessment of how often something happened. After measuring all first-year students' (*N* = 215) levels of academic state procrastination, we used the data of the students who completed the APSI questionnaire (*N* = 186; 26 men, 160 women, age between 16 and 22 years; *M* age = 17.62, *SD* age = 1.03; *M* APSI = 81.68, *SD* APSI = 19.38, *α* APSI = 0.93). We ranked the outcomes and classified three groups of students, based on their APSI scores: students with the lowest sum score, students with an average sum score, and students with the highest sum score for academic state procrastination. In the order of the sum score ranking (beginning with the lowest-scoring student of the low procrastinators, the average-scoring student of the average procrastinators, and the highest-scoring student of the high procrastinators), the first author of this article approached students personally. He informed them about this study and asked whether they were willing to participate in the study and to be interviewed about how they gave meaning to their learning activities. As a result, 22 students were interviewed: eight students with low procrastination scores, eight students with average procrastination scores, and six students with high procrastination scores. See **Table [1](#T1){ref-type="table"}** for the APSI scores of all first-year students and the groups of selected participants. ###### Academic Procrastination State Inventory (APSI) scores of the selected participants and all first-year students. Age Academic state procrastination^∗1^ ------------------------------------ ----- ------------------------------------ ------- ------ ----- --------- -------- ------- ***Low procrastinators*** 16 22 17.75 1.83 36 56 49.13 6.15 *N* = 8: three males, five females ***Average procrastinators*** 16 20 17.75 1.28 73 81 78.89 2.75 *N* = 8: two males, six females ***High procrastinators*** 17 19 17.50 0.84 110 129^∗2^ 117.00 7.01 *N* = 6: two males, four females ***All first-year students*** 16 22 17.62 1.03 36 131 81.68 19.38 *N* = 186: 26 males, 160 females ∗ 1 The lowest possible minimum score is 31; the highest possible maximum score is 155. ∗ 2 The APSI maximum of the high procrastinators differs from the APSI maximum of all first-year students, because not all high procrastinators we approached were willing to be interviewed. Interviews ---------- All the interviews were conducted individually by the first author of this article. The interviewer was not familiar with the students and had no relationship with them. At the start of the interview, the interviewer gave information about the interview process. He explained that the interview would take about 45 min and that the interview was about how the student gave meaning to his/her learning activities. The interviewer explained that he wished to make an audio recording and asked the student for permission for this. He stated that the interview would be processed anonymously and that the conversation was an open conversation, in which there would be no right or wrong answers. The interview protocol for this study was structured by seven general questions. The interviewer used the questions as guidance but also probed specific issues that came up during the interview. The interview questions were general questions regarding students' personal experiences with preparing for assignments and exams, as well as questions to better understand students' procrastinating processes and the influential factors (see **Table [2](#T2){ref-type="table"}**). With the interview questions, we wanted to cover the three phases of the self-regulation process ([@B53]; [@B43]): the forethought and planning phase, the monitoring performance and motivation phase, and the phase after completing the study activity, as well as factors influencing the students' learning. During the interviews, the interviewer used the protocol to check whether all interview questions were covered. On average, the interviews lasted 46 min (*SD* = 9). ###### Overview of the interview questions. Interview questions --------------------- ---------------------------------------------------------------------------------------------------------------------------------------------------------------------- \(1\) Why did you choose this degree program? \(2\) If you have to study/complete assignments for your study, how do you get started? Can you describe your approach? \(3\) How is it for you to be engaged in study activities? \(4\) How is it for you if, while performing study activities, you realize that it is going well? What do you think? What do you say to yourself? \(5\) How is it for you if, while performing study activities, you realize that it is not going well? What do you think? What do you say to yourself? \(6\) How do you appreciate yourself in general? \(7\) What expectations do you have about the results, before an examination? Do you expect them to be positive or negative? What explanation do you have for that result? The procedures of this study were carried out in accordance with the recommendations of the local guidelines of the Faculty of Behavioural and Movement Sciences, VU University Amsterdam ([@B67]). This study was approved on these guidelines by the local ethic committee Research Centre of Driestar Christian University. All subjects gave written informed consent in accordance with the Declaration of Helsinki. Analysis -------- The interviews were anonymously transcribed by a research assistant. The first author read the interviews to get an overall impression and then segmented the text into meaningful units that were comprehensive by themselves and contained one idea, episode, or piece of information ([@B60]; [@B49]). The first author coded nine selectively chosen interview transcripts (the highest-, middle-, and lowest-scoring participants of each subgroup of procrastinators) and composed the codebook, consisting of 18 codes. For the coding procedure, the software ATLAS.ti was used. To calculate intercoder reliability, a research assistant independently coded the nine interviews. Cohen's kappa was calculated for the nine interviews, based on the 18 codes and the 193 segments. The intercoder reliability was good (*κ* = 0.82). After this procedure, the first author coded the remainder of the segments (13 interviews, consisting of 324 segments). To further structure the data, we performed a qualitative content analysis using a thematic analysis approach ([@B11]) to identify overarching themes that captured the differences in learning characteristics and the factors that influence students' learning, as described by participants in the interviews. This analysis resulted in six main themes. To promote validity, the authors applied peer debriefing several times until consensus was reached, and they critically discussed the results of the coding process, the formulation of overarching themes, and the selection of the illustrative quotes. When one or more participants gave answers that contradicted one of the themes we described, we did not neglect these quotes but used them to further elaborate the findings and mentioned the quotes in Section "Results." We also counted how often quotes occurred in relation to the total number of the students in a subgroup and showed this in Section "Results." See **Table [3](#T3){ref-type="table"}** for an overview of the codes and themes. ###### Overview of codes and themes. Codes Themes Explanation --------------------------------------------------------------------------------------------------------------- --------------------------------------------- -------------------------------------------------------------------------------- Motivation to be a teacher \(1\) Degree program choice Student's motivation for his or her degree program choice Approach when performing study activities \(2\) Getting started with study activities Student's approach for getting started with study activities Positive feeling(s) when working on or completing study activities \(3\) Engagement in study activities How the student is engaged in study activities Being in the flow: absorbed in the study activity Motivation enhancing the study experience or study activity Postponement behavior \(4\) Ways of reacting to failure How the student reacts to failure or difficult moments during study activities Distracting thoughts during study activities Focus on what is finished or what still needs to be done Need for fun and contact with friends or other people Dealing with setbacks when the activity is not successful Dealing with moments of lack of motivation/boring study tasks/not seeing the usefulness of the study activity Dealing with distracting multimedia (Facebook/WhatsApp) Sense of conscience/discipline/responsibility Attitude of life Belief in oneself or in one's ability to succeed \(5\) View of oneself The way the student sees himself or herself Attitude of acceptance \(6\) Study results Expectations for and evaluations of study results Expectation about exam results/results of assignments/feelings of anxiety about exams Results ======= Six themes emerged from the interview data, which we discuss below. In these six themes, there were clear differences between all three groups of students, or between one group and the other two groups. In the description of each theme, we present between brackets (.../...) how often quotes occurred in relation to the total number of the students in that subgroup. Every theme is illustrated with a representative quote. When one or more students in the subgroup gave answers related to the theme that contradicted the quotes we described, we also give representative examples of these contradictory quotes. The student names cited for each quote are pseudonyms. See **Table [4](#T4){ref-type="table"}** for a summary of the results. ###### Summary of the results. Low procrastination students Average procrastination students High procrastination students --------------------------------------- ----------------------------------------------------------------------------------------------------------------------------------------------------------- ---------------------------------------------------------------------------------------------------------------------------------------------- ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- Degree program choice Intrinsically motivated decision to become a teacher. Intrinsically motivated decision to become a teacher. Have no clear idea of becoming a teacher. Choice confirmed by positive experiences with theoretical and practical parts of the program. Internships are important experiences motivating students to continue with their program when they have doubts about it. Have doubts about the program and consider quitting. Getting started with study activities First focus on the description of the activities content, nature of the material, and assignments requirements, and then plan tasks and goals to achieve. First focus on content or number of activities and then plan tasks and goals to achieve. Plan learning activities, but carrying out the plan depends on certain preconditions. Set no preconditions to start. Set no preconditions to start. Set preconditions to start. If preconditions are not met, the learning activity will be postponed. Engagement in study activities Intrinsically motivated and go for it. Focus on completing study activities and less so on possible takeaways. Focus on the utility of the study activity. When the learning activity appears to be useful and enjoyable, they enjoy doing it. Aware of how the activity is going and the progress they make. Consciousness of gaining insights and general knowledge. Progress of the task is determined by relevance to the profession and applicability during internships. If not, it's hard to remain engaged. When a learning activity is boring and considered stupid, they tend to stop doing it, turning to other non-school activities that are more appealing. Further their knowledge. Reflect on own role as a teacher when learning activity bears on the profession. Way of reacting to failure Remain focused on completing the learning activity when the result is not desirable. Do less and think they are wasting their time when the result of a learning activity is not desirable. When progress is disappointing, then judge themselves negatively. Encourage themselves verbally to keep going. Experience a sense of failure and feel low or moody. Experience negative feelings and low self-esteem. Rely on their capacities to complete the learning activity. Are hopeful that they will manage and expect that, in the end, they will be able to complete. No longer believe that they are up to it. Those negative feelings can also concern situations outside the program View of oneself Satisfied with the person they are. In general satisfied with the person they are, but also critically reflect on themselves seeing points for improvement. Some moments of not being satisfied with the person they are. The esteem others have for them is also important. Positive self-esteem. Positive self-esteem. Frequent moments of negative self-esteem. Study results Confident about results beforehand, which is reinforced by good results in the past. Sometimes doubtful beforehand about the results. Confident beforehand if they know they spend enough time to prepare for exams. Levelheadedness, no stress or nerves for exams. Nervous about exams and feel pressure to do well. No fear or nerves about exams. Passing the exams is explained by their own efforts. Expectations of passing or failing depend on the difficulty of subject. See themselves as the determining factor in passing or not. If they fail, they attribute their failure to not spending enough time preparing for the exam and/or to not attending all lectures. Theme 1: Degree Program Choice: Intrinsically Motivated Versus "I'd Rather Do Something Different" -------------------------------------------------------------------------------------------------- ### Students With Low Levels of Procrastination When considering statements about motivation for doing the degree made by the group of students with low levels of procrastination ("LP students"), all LP students reported this choice was intrinsically motivated, as the logical way to fulfill their dream of becoming a teacher (8/8). 1. I am highly motivated to do this degree. It simply is my dream to become a teacher (...). (Tiffany) All the LP students were intrinsically motivated for the degree they were doing. They described it was the right choice which was confirmed by their positive experiences with the theoretical and practical parts of the program (8/8). 1. The assignments are so interesting and fun to carry out that you really enjoy doing them. You get a positive energy to carry on. This feeling---I can do it, I've submitted it, it looks good. And then, "Okay, up to the next one..." It's just nice to experience it like this. (Lucas) Most of the LP students felt it was self-evident that they should be committed to the program and persevere. They felt they were accountable for the choice they made for their degree (5/8). 1. It is simply perseverance, because I want to. I just want to get those credits and do well. I really think it is important. You decide to take this degree, so you really go for it.... My approach is purposeful. I have a goal in mind, and I feel responsible for it. Because I chose to do this, it is up to me to do well. I want to become a good teacher, so it's important to go to class and make sure you learn from it. (Irma) ### Students With Average Levels of Procrastination All students (8/8) with average levels of procrastination ("AP students") stated they carefully chose their degree program. They had a clear intrinsically motivated goal in mind for doing the degree; they wanted to become a teacher. 1. I've known for a long time, since I was four years old, that this is what I want. So I think like, this is your drive to get things done after all. (Marvin) Most AP students reported they were more focused on the practical part and less on the theoretical part of the degree (6/8). The internships in the field of their profession were very important to most AP students (7/8). When AP students had doubts about their degree program choice, most drew energy from their internship experiences, which provided them with the incentive to continue their degree (5/8). They realized once again why they wanted to do this, namely, to become a teacher. 1. With some assignments, I have that kind of feeling like, what for? Surely, I don't need all that to become a teacher.... Doing the internship confirms my choice, I really enjoy doing that. But at school I don't always have that feeling. I just love the practical work. So then I think, well I've just got to get through those four years. The internship makes that I realize that this is what I really want. (Ailyn) ### Students With High Levels of Procrastination Students with high levels of procrastination ("HP students") described they did not carefully choose their degree program. Finishing the degree or becoming an elementary school teacher was not the primary intrinsically motivated goal for the majority of HP students (4/6). Most of the students reported their interests lay elsewhere (4/6). The students had a wide range of reasons for still opting for an elementary teacher education program. 1. I didn't have that feeling like "this is what I want later in life" when I opted for this degree. Of course, if you think this is super fun, you will be much more dedicated to the program. But that hasn't been the case for me from the start. I'm here because I didn't have a better idea.... Once I'm done here, I should like to continue at university, I'm kind of interested in History. (Jack) Most HP students (5/6) stated they were doubtful about continuing their degree and considered quitting. Some (3/5) of these HP students who were doubtful about continuing their degree and considered quitting their studies, explained that the program content and the learning activities that came with it had little appeal. 1. If all day long all you do is being busy with those stupid assignments, you do start to wonder, why am I doing a degree in elementary teacher education? I'd rather do fun things. I've got that quite often.... So I'm not really that positive about the program. I keep wondering whether this is the right choice for me. (Jennifer) For some HP students (2/6) who carefully chose their degree program, the way things were going at the college gave them the feeling they were not in the right place. Therefore, they wanted to continue their studies at a different teacher training college. 1. This college just isn't the place for me. I find it too strictly Christian.... I've enrolled in a different college for next year. (Ariah) Theme 2: Getting Started With Learning Activities: Just Get Started Versus Get Started Only Under Certain Preconditions ----------------------------------------------------------------------------------------------------------------------- ### Students With Low Levels of Procrastination Most of the LP students (6/8) stated, when they had to perform learning activities for the program, their approach was to focus first on the description of the content of the activity. In this way, they assessed the nature of the material and the assignment's requirements. Based on this orientation, they planned the different tasks, considering the goals they wanted to achieve for completing the activity. Then the students performed the learning activity. They did not set any preconditions for getting started. 1. The first thing I do primarily is to look at the assignment, see what it says. Its requirements, what I need to do. Then I start with the assignment until I'm done. Or at least I get started with it. I see how far I get, how much time I've got.... I make a to-do list for the assignments. (Audree) All LP students (8/8) stated, that when they started with their learning activity, they found out easily whether they had made the correct assumptions about it, and then they expected to see how far they get. 1. I primarily make sure I understand the assignment and if I don't, just ask. And then I just get started with it and see how far I get. If I'm stuck, I just see what I can do about it so that I can get on with it. You just need to know what's expected... I could start straightaway, but then it would go like, I'm in the middle of it, and I realize this should go in and that as well and I shouldn't have done such and such because I don't need it at all. And I don't like that! (Lucas) ### Students With Average Levels of Procrastination Most of the AP students (7/8), described that, when they had to perform learning activities for the program, they did not set any preconditions to get started and just begin. Four of these AP students who did not set any preconditions planned their work after orienting themselves to the content of the activity. The other three planned their work after it was clear how much they needed to do. Their study behavior was guided by making themselves familiar with the demands of the assignment and its deadline. 1. At the start of the term, I read the course guides and note down the assignments I have to do for each course. I put them in a schedule, so that I know, this is what I have to do in this term and those are the deadlines. It's a kind of goal I set myself. I find it quite useful.... So that I have an absolutely clear overview of what needs to be done in terms of assignments and exams. (Juliet) ### Students With High Levels of Procrastination When HP students had to perform learning activities for the program, most reported they made a study plan (5/6). Whether the plan was actually executed depended for most of the HP students on certain preconditions they have (5/6). Examples of such preconditions regarding study activities are: does the activity yield enough credits? Is the activity attractive and fun to do? Is it clear what must be done? Is the activity not too difficult? How nearby or far is the deadline when the activity must be done? Does the activity concern group work? Some conditions have to do with the student, such as they need a clear overview in order to get started or must feel some enthusiasm for it, be in the right mood for it, or be confident about their ability to complete the learning activity successfully. 1. I'm not good at planning my work. I can make a plan but I do not stick to it. I kind of think, what appeals to me now, or more often really, what must be done now or is there any group work that has priority, because with group work you must do your bit. With the other stuff I look at how important it is, how soon I'll get credits for it, things like that. (Jennifer) In the absence of preconditions, most of the HP students described they easily postponed learning activities (4/6). 1. If an assignment isn't clear, and I believe it is difficult, I postpone it. Because I feel that 'I won't be able to do it anyway,' or 'I don't know how to go about it.' (Ariah) Most HP students told they often thought they had enough time to do the learning activity and postponed getting started, thinking "there's still time" (4/6). Such procrastination behavior could occur at different moments: at the start of the term when there was still enough time, but also when the deadline was near, such as the week before the submission of a report or even the evening before the day of an exam. 1. If I read the summary half an hour before it, I usually get a pass. For example, I think, I've got an hour, so if I get started at nine o'clock, I'll get it done. And then it is nine thirty, and I think, oh well, tomorrow is another day. (Jennifer) Theme 3: Engagement in Learning Activities: Interested in the Content for Its Own Sake Versus Usefulness of Content in Practice ------------------------------------------------------------------------------------------------------------------------------- ### Students With Low Levels of Procrastination Reporting about performing learning activities, most LP students (7/8) told they were conscious of their own commitment to the activity and of how the activity was going. When they (5/8) found the content interesting, they looked to further their knowledge. Because they were interested in the content, they actively became more engaged, deepening their knowledge and doing more than was required by the assignment. 1. If I'm interested, I learn from it. What those people say, I then think, hey, that's a good one, I want to find out more about this. I then sometimes search for articles in newspapers or for discussions about it. (Tiffany) In the experience of LP students, performing learning activities was something they enjoyed (5/8). They were aware that they were gaining important skills and knowledge from doing these activities and that these skills contributed to their personal development. 1. Looking at how far I've come and what I've learned. It all contributes. General knowledge. Knowledge about teaching, insight into yourself. Really very useful. (Lucas) Whether certain learning activities did not look interesting or useful to LP students, most students still continued doing them (7/8). The students described they realized that completing the unenjoyable learning activity was necessary to be able to complete their degree. In their minds, the activity was just something that formed part of the curriculum and just has to be done. They showed perseverance, wanting to get it done. 1. I do not always see the usefulness of an assignment. You've still got to do it though, to build your knowledge and be able to do it in practice. This is after all what's expected of you, it's part of your professional attitude. (Brandon) ### Students With Average Levels of Procrastination When AP students reported about performing learning activities, most described they focused on completing the activities (5/8). These students were less interested in the question of what they would take away from it. Most AP students were motivated to do the learning activity if the activity looked inherently interesting (5/8). Half of the AP students reported it was easier to get started and complete the activity when they considered it interesting (4/8). In order to stay engaged with the activity when completing it, it was important to half of the AP students that the activity was interesting and/or fun. If that was the case, they felt it was easier to perform the activity (4/8). 1. Theories are often too remote. If I don't find it interesting, I think, stupid assignment, I'll leave it till later. With practical assignments, I really put myself to it, thinking, 'I'll make sure it's good.' (Amber) Most AP students described they were interested in the learning activity if it was relevant for, and applicable to, the profession (6/8). They asked themselves, "What can or should I do with this as a teacher?" If a learning activity's usefulness was unclear, then it had little appeal, which made starting the activity harder. 1. It's more fun to do if the material is interesting and practice-oriented. Seeing how you could apply it easily makes it more appealing. It is also easier to remember. (Ailyn) When a learning activity was not interesting or appealing, for all AP students it was harder to get started with it and persist (8/8). 1. When studying for my exams, I notice I'm not really into that. So I think, tomorrow is another day. Then I postpone it again. (Marvin) ### Students With High Levels of Procrastination All of the HP students reported it was important for them to perceive the learning activity as useful and enjoyable. In that case, they were motivated to perform it, and experienced it as pleasant to do (6/6). 1. If teachers would just explain the assignments in class, tell you which to do, and about their purpose. But they never do so. Why so many assignments? Just set a few small assignments with a clear purpose. That would be a huge difference for me. Then I'd do them. (Gordon) Most HP students (5/6) described they did not perceive the utility of a learning activity if they considered it boring. They tended to stop doing it and turned to other activities that were more appealing. 1. If all day long all you must do is those stupid assignments. You do start to wonder, "Why am I taking this degree? What am I doing here? I would much rather do something more fun." That happens quite often with me. (Jennifer) Theme 4: Ways of Reacting to Failure: Perseverance and Getting It Done Versus Doubting One's Ability and Giving Up ------------------------------------------------------------------------------------------------------------------ ### Students With Low Levels of Procrastination Most LP students (7/8) reported they did not give up when their efforts for the learning activity did not seem to lead to a desirable result. LP students were aware that the learning activity was not going as well as they would have liked. They were annoyed but still focused on their goal, wanting to complete the learning activity. Most LP students maintained a positive attitude, relying on their ability to eventually complete the learning activity (5/8). 1. If it isn't going well, you often feel kind of, "Come on---just get on with it and then you're done. Just carry on and then it is finished." That's how it works for me. It's like "it's something that simply has to be done, just go for it now so that it won't bother you later." (Whitney) ### Students With Average Levels of Procrastination Most AP students (7/8) described they noticed that an activity was not going well and that they were doing less than they would have liked to, when their efforts for the learning activity did not seem to lead to a desirable result. If it was not going according to their wishes, most AP students (6/8) reported they got negative feelings, thinking they were wasting their time. For example, they felt low and/or moody, or experienced a sense of failure. 1. But if I'm working on it and not managing, I'll feel very negative. Thinking "Oh it isn't going well at all, I'm just wasting my time." I kind of drift off ... And then I feel quite low. I feel a bit of a loser. (Amber) In such situations, when performing the learning activity was not going well, most AP students (6/8) were still hopeful that eventually they would manage, expecting that in the end they would be able to complete the activity satisfactorily. 1. I kind of think, 'Come on, you can do this.' I then keep telling myself 'just get it done, put yourself to it, and when it is finished you are done.' (Juliet) ### Students With High Levels of Procrastination When the efforts of HP students for the learning activity did not seem to lead to a desirable result, these students described they judged themselves negatively (5/6). Their self-esteem seemed to drop, and they no longer believed that they were up to the task. They no longer wanted to persist, could not recover from a downward spiral, and gave up (5/6). 1. So then I think, "I'm stuck." Or when an assignment is returned for rewriting for the third time, I don't even bother. You never get it right. That wears me out. You try, make an effort, but still it's not right. And then you think, "Why bother..." (Jack) Most HP students reported those negative feelings also concerned situations outside the program (5/6). 1. Then I just think I can no longer do it. And if that happens, I have the same feeling with other things, too. I lose interest, no longer feel like it. It doesn't have to be like that, but at such moments everything looks negative. Your self-esteem really drops. (Gordon) Theme 5: View of Oneself: Positive Versus Negative -------------------------------------------------- ### Students With Low Levels of Procrastination All LP students described a positive view of themselves. Their self-esteem was positive and unwavering (8/8). They knew they had the capacity to complete their degree, which was confirmed by the high grades they received. 1. My self-esteem is quite positive. At a certain point, you know you can do it... With negative self-esteem, thinking you aren't up to it, I think it soon becomes harder to do... To me, it isn't surprising really that I pass everything. My grades are quite good actually. (Audree) Most LP students reported they were satisfied with the person they were. They accepted themselves as they were (6/8). 1. I am who I am. I'm okay with that. (Brandon) ```{=html} <!-- --> ``` 1. No complaints. I am who I am and accept who I am. (Lucas) ### Students With Average Levels of Procrastination Most AP students reported they were generally satisfied with the person they were, and their self-esteem was positive (7/8). In addition, they reflected critically on themselves, seeing certain points for improvement. (7/8). 1. I'm happy with the person I am in most respects. Besides, there isn't much you can change really. I'd like to be a little more outgoing with respect to other people. A little more forthcoming in groups. That doesn't happen so often. (Julisa) ### Students With High Levels of Procrastination Most of the HP students (4/6) described they had frequent moments of low self-esteem. 1. Sometimes I think, "Ugh, I don't want to look the mirror." Without cause. I don't know. How I feel also has to do with how I look. After experiencing success or when we're having fun together, I think, hey, I don't look so bad, and I'm okay. But on a day when I fail something, it affects everything. It's all or nothing with me. (Ariah) In judging themselves, half of the HP students (3/6) reported they did not consider only the way they viewed themselves but also how they were appreciated by others. For two of them this view included HP students' view of fellow students or and for one students the view of teachers. 1. What someone else thinks of me is almost more important than what I think myself---that doesn't count really, it's not important. I don't dare relying on my own opinion of myself. (Ariah) Theme 6: Study Results: "I Passed Because I Am Able To" Versus "I Failed Because of Poor Preparation" ----------------------------------------------------------------------------------------------------- ### Students With Low Levels of Procrastination Most (7/8) of the LP students described they were confident about the results and embarked on the exam period with a certain levelheadedness. None of the LP students (8/8) suffered from stress or nerves. 1. If I know 'I did my best," then that's what it is. I'm never nervous beforehand. (Brandon) When LP students (4/6) passed an exam, most reported they believed it was due to their own efforts, because while preparing they did what they had to do. 1. Beforehand, I expect to pass. I simply go for it. I prepare as best as I can... I do what I can, and then it's just a matter of wait and see. (Lucas) Most of the LP students (5/8) described their confidence about passing their exams was reinforced by good results in the past. Previous good grades were a confirmation of the students' abilities, boosting their confidence they would pass their exams. 1. Now, when facing an exam, I expect to pass. In fact, I've never failed an exam. (Teresia) ### Students With Average Levels of Procrastination Most AP students (5/8) reported they were doubtful about their exam results when they embarked on the exam period. For three of them who were doubtful, this doubt depended on the subject. 1. I know that in most cases I'll pass. Sometimes, I'm afraid to fail, because it's difficult. I'm not really good at arithmetic, for example. (Joyce) Several AP students described their confidence in passing their exams was negatively affected by thoughts of expecting to fail (3/8). 1. I often have negative thoughts like 'I won't pass anyway.' (Rebeca) For two AP students, expecting to fail seemed to act as a form of self-protection (2/8). In their view, it was better to say they would fail than state beforehand that they would pass. When they failed, it was a drawback and a disappointment; when they passed, it was something of a windfall. 1. It's often like, "I don't know." Not often like "don't worry, you'll pass." That would be a little scary. In case you fail, and you thought "surely you'll pass." Then if you don't pass, you're really very disappointed.... If you depart from the notion that you'll fail, then it could always be better than expected, right? (Juliet) Some AP students (3/8) reported they were nervous before the exam period. This feeling went with the thought that it was all or nothing during exam week. They felt pressured to perform well. 1. I'm quite nervous always. I'm not sure about myself. Will I manage? The exam period is very demanding, and you have to show your best abilities. (Howard) ### Students With High Levels of Procrastination Most HP students (4/6) described they did not suffer from fear or nerves before exams. They saw themselves as the determining factor in passing or not passing their exams. 1. When I'm well prepared, I expect to pass. (Melani) In practice, however, most HP students (4/6) reported they often prepared poorly. They did not spend enough time preparing for the exam and/or did not attend all lectures. In the students' view, this behavior explained why they failed some of their exams. 1. You do hope to pass, but occasionally, you've got to admit that it's not going to happen. I don't always prepare that well, and I do tend to skip lectures, and then you miss out on quite a lot. (Jack) Discussion ========== We know that different factors might influence procrastination. The importance of this study is that it shows that there are differences between students with low, average, and high levels of academic procrastination, in whether or how factors influencing procrastination play out in practice, and whether or how this influences students' learning. The results show that for low and average procrastinators their self-chosen goal of becoming a teacher works as a strong intrinsically motivational drive to work on study activities and finish them. This seems to be related to what is called the need for autonomy in Self-Determination Theory ([@B46]). High procrastinators lack the intrinsic motivation to become a teacher, and starting and/or continuing study activities is a problem for them. These students seem unable to externally regulate ([@B46]) their behavior and are unable to shift from an external control of their behavior to an internal control of their behavior. The results are in line with [@B70] who showed that the level of procrastination reported by undergraduate students can be viewed as a function of the motivational beliefs important for self-regulated learning. The results are also in line with other studies showing the role of motivation in students' academic procrastination. For example, [@B23] found that the use of motivational regulation strategies had significant positive indirect effects on students' academic performance and affective/cognitive well-being, via academic procrastination. [@B4] showed that procrastination is less likely to occur for intrinsically motivated activities and [@B65] showed that students' lack of motivation influences their levels of academic procrastination. The motivational problems of high procrastinators can be explained with the temporal motivation theory (TMT; [@B58]). Because high procrastinators do not have a clear intrinsically motivated goal, they do not have an expectation of receiving anything of size or value for starting or finishing their study activity. This result of our study adds the insight that average procrastinators can deal with the tendency to procrastinate, if the value of what they do is clear to them. The results show that high procrastinators set certain preconditions to start or continue a task. For these students, task aversiveness is an argument to postpone or give up a task. This finding is in line with previous research ([@B48]; [@B1]; [@B57]) in which task aversiveness was a predictor of procrastination behavior. The present study adds insights into how students deal with the factor of task aversion. While for students with high levels of procrastination task aversiveness is an argument not to start or to stop, for students with low and average levels of procrastination, task aversion does not automatically lead to procrastination behavior. Admittedly, low procrastinators had less task aversion, because they found everything interesting and deepened their learning. This is in line with the study by [@B59] showing less procrastination was associated with a strategic learning approach. This was different for average procrastinators who reported experiences of task aversion. They realized that finishing the task was important to reach their goal of becoming a teacher, which helped them finish the task. This outcome is partly in line with the study by [@B38], which showed that for students with a low disposition to procrastinate, environmental factors had a negligible impact and for high-level procrastinators, environmental factors facilitated and augmented procrastination. The results of the present study differ for the group of students with average levels of procrastination, because in this study environmental factors had a negligible impact on students with average levels of procrastination. The students' drive to become a teacher makes them persevere. The results show how self-control ([@B57]) among students with different levels of procrastination works. This factor plays a role at different moments in which students work on study activities. An important aspect of having self-control is *effort regulation* ([@B45]), which refers to the capacity to persist when confronted with academic challenges, and for example, start unconditionally when a task is perceived as unattractive. The results show that high procrastinators seem not able to regulate their effort. In addition, when a task turns out to be unattractive, a difference in self-control surfaces: high procrastinators then tend to stop, while average procrastinators tend to continue completing the task, as they keep the goal of becoming a teacher in mind. Low procrastinators are able to make the task interesting to themselves. A possible explanation for these results for the different learning characteristics for students with low, average, and high levels of procrastination could be the students' level of executive functioning ([@B44]). Regarding fear of failure, the results show that average procrastinators compared with the other two groups seem to suffer the most from fear of failure regarding exams. The relationship between fear of failure and procrastination behavior appeared to be more complicated than became visible in previous studies ([@B51]; [@B16]; [@B57]). In the present study, low procrastinators did not suffer from fear of failure and had good results whereas high procrastinators also did not suffer from fear of failure, but reported poor results for exams. The high procrastinators' explanation for their results was that they would have had good results if they would have started studying in time. Low and high procrastinators did not doubt their competence whereas average procrastinators had doubts. This result seems to confirm [@B24] findings that the relation between fear of failure and procrastination is moderated by the level of competence. However, in the present study this seems not the case for high procrastinators. High procrastinators' explanation for their poor results indicates an internal attribution style for failure. This finding differs from previous research ([@B4]) which showed that high academic procrastinators make external attributions (to context and luck) for their successes. The explanation for their poor results seem to indicate that high procrastinators have a tendency for self-handicapping behavior, which is in line with [@B12] finding. The results show that self-esteem and self-efficacy make a difference in procrastination behavior. This finding confirms previous studies showing that negative self-esteem ([@B13]) and negative self-efficacy ([@B70]; [@B30]) are related to procrastination, but in the present study, we see differences between the three groups. Low procrastinators had positive self-esteem and a positive sense of self-efficacy, and relied on this characteristic when they experienced difficulties. Average and high procrastinators had much lower self-esteem. For high procrastinators, experiencing negative thoughts or feelings was a reason to stop or not start study activities. This was the case for average procrastinators: they continued and hoped that they would accomplish the task. In addition, average procrastinators doubted their self-efficacy but thought this was no reason to get stuck in procrastination. High procrastinators seemed to have a higher sense of self-efficacy but experienced more procrastination behavior. Looking at how students perform their study activities, the results show differences between the three groups regarding awareness and control of mental thoughts. Low procrastinators seem to be connected in the moment with their study activity and are aware of what happens in the here and now. This indicates that they are in a so-called state of *presence* ([@B47]). They seem to be connected to their capacities and rely on their strengths to complete the task, and they are determined to finish it. Average procrastinators seem to have lower levels of so-called metacognitive awareness, as defined by [@B17]. They are focused on completing the task and less on learning from it, except when a learning activity pertains to the profession. They then reflect on their own role as a teacher. When the results of their effort disappoint them, they cannot handle it, seem to become disconnected from their capacity to change the situation, become overwhelmed by negative thoughts, do less, and postpone the task temporarily. High procrastinators seem unable to view themselves from a metacognitive perspective. When they do not perceive the activity as useful, they experience negative feelings and judge themselves negatively. They cannot overcome the negative situation and therefore, give up and do more appealing activities outside the study activity. These results confirm previous studies showing self-awareness is necessary in order to motivate corrective behavior ([@B5]) and for procrastinators, low mindfulness may be a risk factor for poor physical and emotional well-being ([@B55]). Limitations ----------- The results of the present study should be interpreted against the background of several limitations. A key characteristic of most qualitative studies is that they focus on participants' perspectives and are not intended to generalize to a broader population ([@B8]). The sample comprised first-year Dutch students enrolled in an elementary teacher education program at a small teachers college. Thus, the extent to which these findings can be generalized to other programs or to other cultures is uncertain. A second limitation is that the scores on the APSI form a continuum. Hence, our distinction of three separate groups may have created a bias. Also the number of interviewed students was low, and because students with average or low levels of academic procrastination were more willing to be interviewed than those with high levels of academic procrastination, the group sizes were not equal. The students with high levels of academic procrastination were more difficult to get in touch with and seemed to be less willing to be interviewed. Therefore, it is also conceivable that we did not interview the students with the highest procrastination problems. It is also conceivable that this serves as a behavioral measure, and the results would have been different if we had interviewed more students with high levels of academic procrastination. A final limitation is that the interviewed students talked about their learning experiences from their own perspectives. It is not clear whether how the students talked about their regulation of learning activities always corresponds to their actual behavior. Future Research --------------- With the insights of the present study in mind, it would be interesting for future intervention studies to see whether programs to overcome procrastination have different effects on students with different levels of procrastination. Future research also might show whether average procrastinators are more open to overcoming their procrastination, for example, because they have stronger ideals. Perhaps this group would benefit the most from interventions and therefore, should be targeted most in intervention programs. A question to answer in future research could also be which interventions for overcoming procrastination are more helpful for average procrastinators and which for high procrastinators. The results of the present study show differences in the metacognitive awareness and degree of presence of the students in the different groups. For future research, this difference raises the question whether students with average and high levels of procrastination can be taught to enhance their metacognitive awareness (as defined by [@B17]) and their level of presence as defined by [@B47], and how this affects their levels of academic procrastination. Implications for Practice ------------------------- Because of the important role of intrinsic motivation in dealing with procrastination behavior, it can be helpful for teacher education institutions to determine to what extent students are motivated. This could have consequences for the intake procedure that takes place, as well as for the aspects of this procedure. This study showed that it especially makes sense for average procrastinators that study activities are practice-oriented. Recognizing the relevance and practical foundation of study activities makes students understand why the study activity is important for them. A stronger connection between the theory students have to study and practice can be reached by taking practical experiences as a starting point for enhancing students' motivation and for promoting their willingness to engage in study activities. This so-called pedagogy of realistic teacher education ([@B33]) might make a difference in maintaining students' motivation and improve the attractiveness of study activities and decrease students' procrastination. In the present study, low procrastinators seem to be most aware of their personal strengths and were engaged in their learning activities while working on them. They are also aware of their interest and their curiosity to learn more and have positive belief in themselves. Research in positive psychology ([@B20]) has shown that it is possible to influence people's beliefs about themselves by supporting awareness and the enactment of their character strengths. Character strengths can be defined as positive traits reflected in thoughts, feelings, and behaviors ([@B41]) and are considered an important aspect of people's 'psychological capital' ([@B37]). Examples of character strengths are curiosity, perseverance, willpower, and hope. According to [@B18]; [@B19]) broaden-and-build theory, a focus on character strengths and positive emotions expands people's repertoires of thoughts and actions ([@B20]). When students are aware of their character strengths in study situations in which the students experience negative feelings, the students can use these strengths to promote their belief in their capacities, which may help them overcome their procrastination behaviors. We implemented these ideas in a field experiment in which we trained procrastinators to overcome their procrastination ([@B64]). This field experiment showed diminishing effects on academic procrastination behavior. Although in this article we discussed an exploratory study, the findings of this study among students with low, average, and high levels of academic procrastination deepens the insights into the process of procrastinatory factors that can influence students' learning. Factors influencing academic procrastination do not have a linear effect but are the result of how students deal with procrastination. The present study provides insights that can lead to hopeful perspectives that more is possible than we see now in the area of academic procrastination. Author Contributions ==================== LV: Ph.D.-candidate at VU University Amsterdam, Netherlands. He is the principal investigator and principal author of this article. He collected and analyzed the data, and interviewed the participants. FK and JS: supervisor and co-writer of parts of the manuscript. Conflict of Interest Statement ============================== The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. **Funding.** This research has been made possible in part by the Netherlands Organization for Scientific Research (NWO) (Project No. 023-002-026). [^1]: Edited by: Katrin B. Klingsieck, Paderborn University, Germany [^2]: Reviewed by: Simone Nadine Loeffler, Karlsruhe Institute of Technology, Germany; Meirav Hen, Tel-Hai Academic College, Israel [^3]: This article was submitted to Personality and Social Psychology, a section of the journal Frontiers in Psychology

Background {#Sec1} ========== Osteoarthritis (OA) is a chronic degenerative joint condition characterized by a progressive destruction of the articular cartilage leading to pain and functional loss (Cross et al. [@CR7]). OA affects more than 39 million people in Europe. These changes lead to a degradation of the extracellular matrix and a reduction of tissue cellularity. Furthermore, due to the little proliferative possibility of articular chondrocytes, cartilage has a limited self-repair ability. Consequently, even minor injuries may progress to important joint degenerations (Laadhar et al. [@CR13]). Small lesions are generally repaired by migration of chondrocytes, while larger lesions are repaired by the formation of a fibrocartilage with impaired biomechanical properties Fernandes et al. ([@CR10]). Recent repair treatments have been explored, including microfracturing or osteochondral allograft transplantation, but they are invasive, with variable prognosis (Gomoll et al. [@CR11]). Currently, oral nonsteroidal anti-inflammatory drugs (NSAID), intra-articular corticosteroids or hyaluronic acid are widely used to relieve OA. Unfortunately, these treatments do not have curative effects on the inflammatory process of OA (Nelson et al. [@CR17]). To address this question, new repair strategies have been developed. Platelet-rich plasma (PRP) is a plasma with 3 to 8 times higher platelet concentration than in the blood, which enables higher concentrations of active growth factors such as platelet-derived growth factors, transforming growth factors β and vascular endothelial growth factor. PRP might promote stem cells recruitment and fibroblast collagen production. Studies in human and animal models suggest that PRP would improve cartilage repair (Sakata et al. [@CR21]). In a recent meta-analysis of 10 level I randomized controlled trials, intra-articular PRP injection was found to better relieve pain and to improve function scores at 1 year post-injection compared to saline and hyaluronic acid (Dai et al. [@CR8]). However, contradictory results have been published (Mascarenhas et al. [@CR15]) but these studies were performed with different types of PRP on different joints and included small heterogeneous populations from first to late stages of OA. Notably, the techniques for intra-articular PRP infiltration were not standardized: some were clinically-guided, others imaging-guided. Radiologists are used to adding iodine contrast agent (ICA) to the injected substance (for instance corticosteroids or hyaluronic acid) to improve the precision of fluoroscopy-guided intra-articular infiltrations. Opacification of the symptomatic joint by ICA confirms that the needle is correctly positioned. By analogy, radiologists have come to mix PRP and ICA during the imaging-guided procedures in order to ascertain that PRP is correctly infiltrated. However, the effect of ICA on platelets is debated and may depend on the type of ICA Aspelin et al. ([@CR1]). In practical applications, knowing that the quality of PRP is not altered by ICA would encourage to the use of ICA to better guide the needle. Conversely, knowing that ICA could alter PRP would discourage the use of ICA during PRP-related procedure. Thus, the aim of this study was to investigate the in vitro effect of ICA on platelet concentration, activation and degranulation from autologous PRP. Methods {#Sec2} ======= The institutional ethical committee approved the study. Signed informed consent was obtained for all patients who participated*.* Patients {#Sec3} -------- From May 2016 to January 2017, a total of 137 consecutive patients were prospectively enrolled in the study. Inclusion criteria were: age (\< 18 years old), radiographically documented OA in the patellofemoral compartment of the symptomatic knee (grade 2 and 3 according to the Kellgren-Lawrence classification), diagnosis of primary OA confirmed at least 3 months prior to the study, failure of non-operative treatment performed at least 6 weeks prior to the study (rehabilitation using analgesic physiotherapy and eccentric work), indication for intra-articular PRP injection validated by a sports medicine physician or an orthopaedic surgeon, infiltration performed at our musculo-skeletal interventional radiology institution. Exclusion criteria were: pregnancy, infections, previous corticosteroid treatment, platelet dysfunction and immunodeficiency. None of the 137 patients met the exclusion criteria. The study included 78 men and 59 women (median age: 45-years old, range 28--64). PRP preparation {#Sec4} --------------- Patients were referred to a clinical pathologist. For each patient, 22 mL of venous blood was collected in a syringe containing 1.5 mL of citrate anticoagulant. Blood was centrifuged at 570 g for 8 min and a final volume of 3.5 mL of PRP was recovered in the lower plasma layer. The platelet concentration factor within the final volume was 2.5, as verified by platelets counting under an automated hematology cell analyzer (ABX Pentra 60, Horiba, Kyoto, Japan). Platelets and leucocytes amounts were controlled within each PRP sample Dallaudière et al. ([@CR9]). Finally, 500 μl of autologous PRP per patient were sampled prior to intra-articular PRP injection. In vitro quantitative assessment {#Sec5} -------------------------------- After anonymization of the PRP samples, 500 μl of PRP were randomly mixed with 50 μl of different ICA (PRP + ICA): Iodixanol 270 mg I/mL (Visipaque270®, GE Healthcare, ChICAgo, Illinois, USA) (*n* = 58) or Iopamidol 200 mg I/mL (Iopamiron200®, Bracco imaging, Milan, Italy) (*n* = 69). Visipaque270® is a non-ionic, hexa-iodinated solution (iodixanol) with an osmolality of 290 mOsm/kg whose excipients are trometanol, sodium chlorure, calcium chlorure, sodium edetate calcium, water and chlorohydrin acid. Iopamiron200® is a non-ionic, tri-iodinated solution with an osmolality of 413 mOsm/kg whose excipients are trometanol, sodium hydroxyd, sodium edetate calcium, water and chlorohydrin acid. The platelet concentration (G/L) within PRP alone (*t* = 0mn) and PRP + ICA at several delays of incubation was quantified (namely, 5, 10, 15, 20, 30mn). 'Thirty minutes' was arbitrarily chosen as the last point in time because PRP is always infiltrated within this delay at our institution and because of lack of enough PRP after this delay for statistical comparisons. Additionally, in vivo*,* it is recommended to acquire X-rays within the 30mn following the infiltration of ICA because the concentration in ICA continuously decreases as it is likely absorbed by the synovia (Obermann et al. [@CR18]; Omoumi et al. [@CR19]). Thus, the interaction between PRP and ICA should also decrease after this incubation delay. Then, the pathologist counted the number of platelets within each sample at each delay of incubation (with an automated cell analyser) and analysed the morphology of the platelets (with an optical microscope with a magnification of 40) blinded to the clinical data. PRP ratio was defined as the concentration of PRP within the PRP + ICA mixture at a given time point divided by the initial concentration of PRP (PRP ratio = \[PRP\](t)/\[PRP\](0mn)). In vitro qualitative assessment {#Sec6} ------------------------------- First, our objective was to compare the platelet function between two different ICAs (Visipaque270®, Iopamiron200®) and a control solution (phosphate buffer solution, PBS). Five patients were included in this part of the study. Their PRP samples were divided in three equal sub-samples and mixed with 50 μL of PBS, Visipaque270® and Iopamiron200®. For each mixture (PRP + PBS, PRP + Visipaque270®, PRP + Iopamiron200®), the platelet aggregation was evaluated according to 4 tests after a delay of incubation of 30mn, using: (i) 10 μmol/L adenosine diphosphate (ADP, Sigma Aldrich Chimie, Lyon, France), (ii) 1 mmol/L arachidonic acid (AA, Nu-Chek-Prep, Elysian, Minnesota, USA), (iii) 25 μmol/L thrombin receptor activating peptide (TRAP, NeosystemSA, Strasbourg, France), and (iv) 2 μg/μL of equine tendon collagen (Horm-Chemie collagen, Nycomed Pharma, Munich, Germany). The tests were achieved in an aggregometer (APACT® 4004, ELITechGroup, Salon de Provence, France) according to standard procedures and during an aggregation test time of 300 s. Second, we investigated whether a different concentration in Iode for the same molecule of ICA could change the platelet function. To do so, we compared the platelet function of two mixtures: PRP + PBS and PRP + Visipaque320® (: iodixanol 320 mg I/mL, instead of 270 mg I/mL), with the same aggregation tests on five other patients. Third, cell surface expressions of P-selectin, which is a marker of alpha-granule release, in Visipaque270® and Iopamiron200® mixtures were compared. The measures were performed on the PRP samples of 5 additional patients before and after addition of ICA, using VH10, which is a murine monoclonal antibody produced by our group Cattaneo et al. ([@CR5]). The mean fluorescence intensity of P-selectin was measured on resting platelets and after stimulation with 25 μM TRAP, enabling to calculate the P-selectin expression ratio (defined as the ratio of mean fluorescence intensity of P-selectin after TRAP stimulation and before TRAP stimulation). Statistical analyses {#Sec7} -------------------- Gaussian distribution was tested with the Shapiro-Wilk normality test. The initial PRP concentrations (incubation delay *t* = 0mn) in the presence of different ICA were compared using the Mann-Whitney test. The influences of both ICA and the delay of incubation on PRP ratio were evaluated with a 2-way ANOVA (with post-hoc Tukey test for multiple comparisons). The influences of ICA and the aggregation tests on the percentage of platelet aggregation were evaluated with a repeated-measures 2-way ANOVA (with a post-hoc Sidak test for multiple comparisons). The P-selectin ratios between the PRP + Visipaque270® and PRP + Iopamiron® mixtures were compared using a non-parametric Wilcoxon matched-pairs signed rank test. Data are presented as mean ± standard deviation. A *p*-value of less than 0.05 was deemed significant. Statistical analyses were performed using SPSS statistical software (version 21.0, Chicago, Illinois, USA) and Graphpad Prism (GraphPad Software, Inc., version 7, La Jolla, California, USA). Results {#Sec8} ======= In vitro quantitative assessment {#Sec9} -------------------------------- Initially (*t* = 0mn), there was no significant difference in \[PRP\] between the following mixtures: PRP + Visipaque270® and PRP + Iopamiron200® mixtures (489.3 +/− 118 G/L versus 460.9 +/− 104.7 G/L, respectively, *p* = 0.250) (Fig. [1a](#Fig1){ref-type="fig"}).Fig. 1In Vitro quantitative assessment. **a**Initial PRP concentration in the samples (*t* = 0), in the presence of Visipaque270® and Iopamiron200®. **b** Evolution of PRP ratio as a function of time (5 incubation delays: 5, 10, 15, 20, 30mn), for each contrast agent group (Visipaque270® and Iopamiron200®). **c** Morphological aspect of platelets before and after exposure to iodine contrast agent: no change was seen A 2-way ANOVA was conducted to examine the effect of ICA, the incubation delay (from 5 to 30mn) and their interaction on PRP ratio (Fig. [1b](#Fig1){ref-type="fig"}). There was no significant influence of the incubation delay on PRP ratio (F(4,128) = 0.069, *p* = 0.991), neither was there a significant influence of the ICA (F(1,128) = 0.009, *p* = 0.926), nor a significant influence of their interaction on PRP ratio (F(4,128) = 2.654, *p* = 0.057). Table [1](#Tab1){ref-type="table"} shows the values of PRP ratio as a function of the incubation delay.Table 1PRP ratio in PRP + Visipaque270® and PRP + Iopamiron200® at different incubation delaysIncubation Delay (mn)PRP ratioVisipaque270® (n = 58)Iopamiron200® (n = 69)50.902 ± 0.0280.916 ± 0.009100.814 ± 0.0140.905 ± 0.026150.909 ± 0.0200.910 ± 0.018200.921 ± 0.0250.894 ± 0.018300.900 ± 0.0280.923 ± 0.020NOTE: PRP ratio is defined as the concentration in PRP in the mixture at each incubation delay divided by the initial concentration in PRP. Results are mean ± sd The optical microscope examination at the end of the incubation delay (30mn) did not demonstrate morphological change of the platelets for both ICA (Fig. [1c](#Fig1){ref-type="fig"}). In vitro qualitative assessment {#Sec10} ------------------------------- First, the percentage of platelet aggregation within the different mixtures (PRP + PBS, PRP + Visipaque270® and PRP + Iopamiron200®) was evaluated after an incubation of 30mn in the presence of ADP, Collagen, AA and TRAP (: 4 tests of aggregation). The solution that was added to PRP did not have a significant effect on the percentage of platelet aggregation (F(2,5) = 0.907, *p* = 0.414), neither did the interaction between the aggregation test and the solution added to PRP (F(6,5) = 0.064, *p* = 0.998) (Fig. [2a](#Fig2){ref-type="fig"}).Fig. 2In Vitro qualitative assessment. **a** Platelet aggregation depending on the iodine contrast agent. Three groups were compared: control (PRP + phosphate buffer solution), Visipaque270® (PRP + Visipaque270®) and Iopamiron200® (PRP+ Iopamiron200®) (**b**) Platelet aggregation depending on the iodine concentration of Visipaque®. Two groups were compared: control (PRP + phosphate buffer solution) and Visipaque320®. The percentage of platelet aggregation was assessed by 4 tests: in the presence of adenosine diphosphate (ADP), Collagen, Arachidonic acid (Ar. acid) and thrombin receptor activating peptide (TRAP). **c** Degranulation: Surface expression of human P-selectin in the presence of PRP + Visipaque270® or PRP + Iopamiron200®, before and after exposure to 25 μM of TRAP. P selectin expression is expressed as the mean fluorescence intensity compare to normal. P-selectin expression ratio corresponds to the ratio between expression after TRAP exposure and before TRAP exposure When another iodine concentration of Visipaque® was tested (PRP + Visipaque320® with 320 mg I/mL versus PRP + PBS) using the 4 same tests of aggregation, the solution that was added to PRP did not have a significant effect on the percentage of aggregation (F(1,5) = 2.110, *p* = 0.166), neither did the interaction between the aggregation test and the added solution (F(3,5) = 0.885, *p* = 0.470) (Fig. [2b](#Fig2){ref-type="fig"}). Post-hoc tests are given in Table [2](#Tab2){ref-type="table"}.Table 2Comparisons in percentage of aggregation between different mixtures, assessed by four aggregation testsComparisonsAggregation testsADPCollagenAATRAPPRP + Iopamiron200® vs. PRP + Visipaque270®\> 0.9990.7940.9400.475PRP + PBS vs. PRP + Visipaque270®0.8390.6410.7200.895PRP + PBS vs. PRP + Iopamiron200®0.8170.9930.9630.748PRP + PBS vs PRP + Visipaque320®0.9930.8760.761\> 0.999NOTE. Results are the p-value of the post-hoc tests of the 2-way ANOVAThe different mixtures were: PRP + Iopamiron200**®**, PRP + Visipaque270**®**, PRP + Visipaque320**®**, PRP+ PBS**®**Abbreviations: *AA* arachidonic acid, *ADP* adenosine diphosphate, *PBS* phosphate buffer solution, *TRAP* thrombin recepto activating peptide, *vs*. versus Given the ability of the platelets to release alpha-granule, the mean fluorescence intensity of P-selectin was not significantly modified after adding TRAP; neither was it in the presence of Visipaque270® (*p* \> 0.999) nor in the presence of Iopamiron200® (*p* = 0.500). The ratio of P-selectin expression after adding TRAP and ICA was not significantly different between Iopamiron200® and Visipaque270® (0.931 +/− 0.141 versus 0.984 +/− 0.111, p = 0.500) (Fig. [2c](#Fig2){ref-type="fig"}). Discussion {#Sec11} ========== In this study, we investigated the potential influence of ICA on platelet function through extensive in vitro analyses. We demonstrated that the concentration in PRP was not altered in the presence of ICA during a delay of incubation of 30mn. We did not observe an alteration of the platelet function in the presence of ICA according to several aggregation tests. The ability of the platelets to release alpha-granule was not modified in the presence of ICA. Altogether, our results suggest that platelets from PRP samples were not quantitatively and functionally modified by adding ICA. The technique for intra-articular infiltration with PRP lacks standardization when it comes to the platelet concentration in PRP samples, the volume of PRP to infiltrate, the optimal moment for an injection of PRP in the course of OA, the number of injections to perform and the adjunction of NSAIDs and/or anaesthetics and/or ICA. All these technical points could also be optimized for each joint (Campbell et al. [@CR3]). In our study, we standardized the collection system with a single centrifugation step and we obtained autologous PRP samples with a 2.5--3 times higher concentration of platelets, as recommended in the literature Dallaudière et al. ([@CR9]). Optimizing the chance to correctly target the pathological joint requires guidance by fluoroscopy and validation of the right position of the needle by opacification with ICA. For such a controversial treatment as PRP, we believe that it is crucial to demonstrate that the substance was well delivered in the joint and not in adjacent tissue whether clinical trials or daily routine. The efficacy of PRP may have been ill estimated in previous studies without the use of ICA because the PRP may have been delivered outside of the pathological joint. Radiologists are used to mixing ICA and hyaluronic acid, or ICA and corticosteroids. By analogy, they routinely mix ICA and PRP. Yet, the interaction between ICA and PRP can be questioned. Prior reports suggest that ICA may have an impact on platelet function. In a study that investigated the adverse events of several ICA based on 48,261 reports, the authors highlighted the frequency of bleeding and clotting disorders likely due to alterations in platelet function (Seong et al. [@CR23]). Cavalli et al. showed that prostaglandin E2, which can modify platelet function, was released in the synovial fluid of patients following a knee arthrography with iothalamate and iopamidol, which may explain the occurrence of transient arthritis (Cavalli et al. [@CR6]). In vitro studies showed that platelet activation was not affected by non-ionic ICA (Li et al. [@CR14]), but may modify platelet aggregation and degranulation (Heptinstall et al. [@CR12]). However, these studies should be carefully analysed because of the heterogeneous study design, either on PRP or on blood samples. Herein, our results did not show any in vitro influence of ICA on PRP within a delay of 30mn. The benefits of PRP depend on the release of bioactive compounds at the optimal moment, through platelets activation and aggregation (Mascarenhas et al. [@CR15]). We excluded patients who were treated by NSAID treatment in order to limit inappropriate platelet activation and aggregation. Adding NSAID to PRP decreased the storage of α-granules and inhibited the activation and aggregation of platelets. In vivo, the function of platelets in autologous PRP in patients who were treated by NSAID is impaired, resulting in a lower quality of PRP bioactive compounds (Schippinger et al. [@CR22]). Moreover, the adjunction of anaesthetics or corticosteroids to intra-tendinous PRP injections demonstrated a significant decrease of tenocytes proliferation and cell viability. These results suggest that adding anaesthetics and/or corticosteroids to PRP would compromise the potential benefits of PRP and the cell viability where the tendon was injured (Carofino et al. [@CR4]). Furthermore, Bausset et al. showed that, in vitro, anaesthetics such as Xylocaine® and Naropin® (belonging to the N-alkyproline anilides group) may compromise PRP potential benefit (Bausset et al. [@CR2]). Herein, the potential interaction between anaesthetics, ICA and PRP was not investigated although anaesthetics are always available during the procedures, at the patient's request. In our study, the function of platelets was investigated using aggregation tests and degranulation tests through the measurement of the P-selectin expression (Prüller et al. [@CR20]). The measurement of the P-selectin expression with Flow cytometry is the gold standard method to evaluate α-granule release after platelets stimulation by standardized inductors of platelet activation (Sakata et al. [@CR21]). Bioactive compounds stored in the α-granules such as platelet-derived growth factors or transforming growth factors β, would not have been adequately released if this pathway had been altered by ICA. Our study has limitations. First, as a pilot study, few patients were included in the in vitro qualitative assessment. The same aggregation tests were performed on 5 patients under 3 conditions (PRP + PBS, PRP + visipaque270®, PRP + iopamiron200®) and on 5 other patients under 2 conditions (PRP + PBS, PRP + visipaque320®). Even if the number of patients was low, the post-hoc tests did not show any tendency towards a difference in the percentage of aggregation under either of the two conditions. The percentages of aggregation (and their standard deviation) under all conditions were very similar to the one under the control condition (: PRP + PBS alone). Second, our study groups consisted in patients with heterogeneous levels of activity, from top-athletes to inactive patients. However, we clearly defined our inclusion criteria and we do not believe that the level of activity could have influenced the interactions between ICA and PRP in an in vitro study. Third, the methods of our in vitro qualitative assessment could be questioned. We did not quantify growth factors and cytokines in the PRP samples before and after ICA adjunction. Furthermore, we used citrate as an anticoagulant instead of hirudin though citrate was suspected to mask pro-aggregatory effects of Iopamiron® in blood samples (Heptinstall et al. [@CR12]). However, this effect was not observed with 'hirudinized' PRP samples instead of 'hirudinized' blood samples. Fourth, all the in vitro tests were performed within a short delay of incubation (*t* = 30mn). Even if we did not identify an influence of ICA on PRP properties during this delay, one cannot eliminate a potential late interaction after the delay of 30mn. However, it should be noted that the concentration of ICA within the joints rapidly decreases with time. Obermann et al. showed that the best diagnostic quality for knee arthrography was obtained when radiographs were acquired within 23mn (Obermann et al. [@CR18]). After this delay, ICA is absorbed by the synovia and is eliminated through blood circulation. That is why it is commonly recommended to limit the delay between ICA infiltration and image acquisition (Omoumi et al. [@CR19]). Finally, the clinical benefit of adding ICA to PRP compared to PRP alone has not been studied, neither have the clinical effects on patients from our series once PRP and ICA were finally injected in the joint. A randomized controlled trial could be considered to compare ICA-helped, imaging-guided, intra-articular PRP infiltration versus direct intra-articular PRP infiltrations on knee OA following OARSI clinical trial recommendations (McAlindon et al. [@CR16]). To conclude, our results suggest that platelets from PRP samples were not in vitro quantitatively and qualitatively modified by adding ICA. These results need to be confirmed by in vivo studies with clinical outcome. AA : Arachidonic acid ADP : Adenosine diphosphate ICA : Iodine contrast agent NSAID : Nonsteroidal anti-inflammatory drugs OA : Osteoarthritis PBS : Phosphate buffer saline PRP : Platelet-rich plasma TRAP : Thrombin receptor activating peptide The authors would like to thank Ms Camille Martinerie for medical writing service. Funding {#FPar1} ======= No funding was received for this study. Availability of data and materials {#FPar2} ================================== The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. Declarations {#FPar3} ============ JEXO-D-18-00028: Intra-articular injection of Platelet Rich Plasma in Osteoarthritis: In vitro Influence of Iodine contrast agent on Platelet Function. BD : designed the study, included the patients, analysed the data, wrote the manuscript. AC : analysed the data, performed the statistical analyses, designed the figures and tables, participated in the elaboration of the manuscript, revised the manuscript. APG: performed the in vitro analyses, analysed the data, validated the manuscript. LP: included the patients, validated the manuscript. AP: helped to design the study, prepared the PRP samples, validated the manuscript. CJ: designed and performed the in vitro analyses, analysed the data, validated the manuscript. AS: designed the study, included the patients, analysed the data, wrote the manuscript. All authors read and approved the final manuscript. Ethics approval and consent to participate {#FPar4} ========================================== This propective single-center study was approved by the local Research Ethics Committee according to good clinical practices and applicable laws. No animal data was used for this study. Consent for publication {#FPar5} ======================= Not applicable Competing interests {#FPar6} =================== The authors declare that they have no competing interests. Publisher's Note {#FPar7} ================ Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Introduction ============ Chronic musculoskeletal neck and back pain are prevalent and disabling conditions among adults \[[@ref1]\]. The report on global economic burden of disease published in 2010 indicated low back pain and neck pain ranked sixth and 21st, respectively, out of 291 diseases and injuries \[[@ref2]\] for medical cost. The studies published in 2016 indicated that combined low back and neck pain ranked fourth out of 315 diseases and injuries \[[@ref3]\] for causes of years lived with disability. Thus, there is an urgent need to identify an effective means to promote self-management of these conditions. Recent advancement in information and communication technology (ICT) places strong emphasis on supporting self-management programs for chronic pain through mobile software apps \[[@ref4],[@ref5]\]. Mobile apps have been identified as a way to reduce the financial burden of chronic neck and back pain \[[@ref6]\]. Use of ICT as part of rehabilitation programs has been suggested to increase adherence to self-management programs \[[@ref7]\]. However, published studies have indicated that approximately 65% of the neck and back pain-related mobile apps have no evidence of the involvement of health care professionals \[[@ref8],[@ref9]\] in their development. To date, only one study was found that investigated the benefits of a professionally designed mobile app that had been trialed. This study tested the effect of using a mobile app on and its benefit in self-management of low back pain \[[@ref10]\]. The app was designed with a group of pain professionals and utilized a cognitive behavior approach based on recommendations from the American Pain Society \[[@ref11]\]. A total of 597 participants with chronic low back pain who had not received any intervention for the condition were recruited. They were randomly assigned to intervention, routine care, and control group. Outcome measures included level of pain and pain frequency. Results indicated greater improvement at 4-month follow-up in both parameters compared to the control group. This study demonstrated that a mobile app may be an effective tool in self-management of low back pain. Although this study provided positive results for the use of mobile app, it excluded the majority of people with chronic back pain that had some sort of prior intervention. In addition, the exercise program appeared to be generic exercises rather than ones specific to the users' symptoms. Other studies are also trialing different mobile apps that focus on relaxation therapy \[[@ref12]\], with a combination of mobile app and clinician offline support \[[@ref6]\]. Results of these studies are yet to be published. Currently, there are a large number of health-related mobile apps on the market, but the majority of the neck and back pain apps have focused on pain management education. Other mobile apps claim to offer exercises and education for chronic neck and back pain but provide limited evidence that a health care professional was involved in their development \[[@ref13]\]. Our institute has taken part in developing an artificial intelligence (AI) system embedded into a mobile app called "Well Health." The AI-embedded mobile app was developed in response to the "World Report on Disability" \[[@ref14]\], which suggested that electronic health (eHealth) or telerehabilitation techniques are effective means to enable people to receive appropriate intervention. However, it also stated that telerehabilitation should be tested to assess its feasibility within the local culture. As documented in a survey of the Chinese population with chronic low back pain (N=113) in 2016, it was found that self-management behaviors were poor \[[@ref15]\]. The survey identified that the contributing factors to poor self-management behavior included a lack of disease knowledge, a lack of understanding of the benefits of exercise, and a lack of communication with health professionals. The AI-embedded mobile app was designed to address some of these factors. It allows users to input their symptoms and generates an exercise program accordingly. The exercise program is delivered in the form of a video that plays in real time. The aim of this study was to evaluate the feasibility of the AI-embedded mobile app Well Health to assist people with chronic neck and back pain to self-manage their condition, and the perceived benefits in symptom improvement by its users. This is the first study that evaluates an AI-embedded mobile app for neck and back pain rehabilitation among the Chinese population. Methods ======= Ethical Consideration --------------------- The study was approved by the Medical Ethical Committee for Clinical Research and Animal Trials of the First Affiliated Hospital of Sun Yat-sen University (reg \# \[2016\] 185). Informed consent was obtained electronically. Once the users clicked on the survey invitation icon, they were directed to an informed consent form. Informed consent was given by selecting "Agree to take part" and "Submit." Each participant was assigned a number to maintain anonymity. Recruitment ----------- This study was a retrospective evaluation study of an AI-embedded mobile app that was available to the general public. All existing users of the mobile app were invited to take part via the in-app "Information" section. The invitation also appeared on the home page where it would be displayed among five rolling pictures of different newsfeeds when the app was loaded each time. Interested users could click on the link which directed them to the evaluation questionnaire. All the participants in this study completed the questionnaire one time, with retrospective recall of preintervention pain level. Sample Population ----------------- The inclusion criteria for the study were adults (1) aged between 18 and 65 years, (2) who had experienced neck and low back pain within the past 3 months, and (3) had access to a mobile phone that could play video on the internet. This study excluded patients who were medically unstable and reported having red flags for cervical and lumbar spine pathology. A list of red flag symptoms related to the cervical and lumber spine were displayed when the app was opened the first time and before the start of the questionnaire. The user would not be able to continue with the survey if more than two red flag symptoms were selected. These inclusion criteria were set to include a wide population because the app was designed to be accessed by members of the general public with experience with chronic neck and back pain. Evaluation Questionnaire ------------------------ The questionnaire included 14 questions that were intended to evaluate if using the AI-embedded mobile app may (1) increase adherence to therapeutic exercises, (2) affect pain level, and (3) reduce the need for other interventions. The World Health Organization defined the term "adherence" as "the extent to which patient behavior taking medication, following a diet, and/or executing lifestyle changes, corresponds with recommendations from a health care provider" in 2003 \[[@ref16]\]. ###### The evaluation questions. 1\. Have you experienced neck and low back pain within the past 3 months? *Responses: (1) yes; (2) no* 2\. Are you age between: *Responses: (1) 18-25; (2) 26-30; (3) 31-40; (4) 41-50; (5) 51-60; (6) 60-65?* 3\. What is your gender? *Responses: (1) male; (2) female* 4\. How long have you used the AI-embedded mobile app? Responses: *(1) 1 day; (2) 1 week; (3) 1 month; (4) 3 months; (5) 6 months or over* 5\. On a scale of 0-10 (0 being no pain, 10 being in extreme pain), how would you rate your pain level before using the rehabilitation program on the AI-embedded mobile app? *Responses: 0 to 10 Likert scale* 6\. On a scale of 0-10, how would you rate your pain level after using the AI-embedded mobile app for rehabilitation program? *Responses: 0 to 10 Likert scale* 7\. Prior to using the app, have you participated in any rehabilitation exercise program? *Responses: (1) yes; (2) no* If "Yes," how much time did you spend on rehabilitation exercise a day on average? *Response: (1) 5 minutes or below; (2) 6 to 10 minutes; (3) between 10 and 30 minutes; (4) between 30 and 60 minutes; (5) 60 minutes or above* 8\. How much time in a day have you followed the AI-embedded mobile app to perform the recommended exercise? *Responses: Amount of time entered by the user (value between 0 to 180 minutes)* 9\. On a scale of 0-100 (0 being no improvement, 100 being completely resolved), how would you rate the overall improvement of your symptoms? *Responses: 0 to 10 Likert scale* 10\. Have you received any of the following interventions prior to using the AI-embedded mobile app? *Responses: (1) acupuncture; (2) soft tissue therapy; (3) topical cream, (4) medication; (5) electrotherapy* 11\. Have you received any of the following interventions prior to using the AI-embedded mobile app? *Responses: (1) acupuncture; (2) soft tissue therapy; (3) topical cream, (4) medication; (5) electrotherapy* 12\. While using the AI-embedded mobile app, have you continued with any of the interventions mentioned? *Responses: (1) acupuncture; (2) soft tissue therapy; (3) topical cream, (4) medication; (5) electrotherapy; (6) no, I did not use any other intervention while using the AI-embedded mobile app* 13\. Have you read the education material within the app? *Responses: (1) yes; (2) no.* If "yes," how much time in a day have you spent on reading the educational material? *Responses: amount of time entered by the user (value between 0 and 180 minutes)* 14\. Would those materials encourage you to adhere to the therapeutic exercise? *Responses: (1) yes; (2) no* Based on this original definition, Donkin et al \[[@ref17]\] provided a further definition of "the degree to which the user followed the program as it was designed" to incorporate e-therapies in 2011. A systematic review on the chronic low back pain population also found self-report diaries to be the most common measure of adherence \[[@ref18]\]. A published study indicated a self-reported exercise log has acceptable agreement with objectively assessed exercise adherence for both exercise frequency and exercise duration \[[@ref19]\]. Therefore, the operating definition of adherence in this study was the amount of "self-reported time spent on therapeutic exercises in a day" because the AI-embedded mobile app was designed to encourage daily therapeutic exercise. The evaluation questions are presented in [Textbox 1](#box1){ref-type="boxed-text"}. The first two questions were used to assess if the responses fit the inclusion criteria. In addition to clinical information, participants were asked to fill in the modified System Usability Scale (SUS). The scale had 10 questions that were validated to assess the generic usability of a product or a service \[[@ref20]\]. The questions are as follows: 1. I think that I would like to use this product frequently. 2. I found the product unnecessarily complex. 3. I thought the product was easy to use. 4. I think that I would need the support of a technical person to able to use this product. 5. I found the various functions in this product were well integrated. 6. I thought there was too much inconsistency in this product. 7. I would imagine that most people would learn to use this product very quickly. 8. I found the system very awkward to use. 9. I felt very confident using the product. 10. I needed to learn a lot of things before I could get going with the product. The SUS showed the domains as five scales numbered from 1 (strongly disagree) to 5 (strongly agree). Positively worded domains were equal to score--1, negatively worded domains were equal to 5--score, so the total SUS was calculated by summing the 10 domains and multiplying by 2.5. Intervention: Well Health Mobile App ------------------------------------ ### Credential and Affiliations of the Developers The affiliations of the developers and the expert panel that were involved in developing the AI-embedded mobile app were clearly displayed within the "About us" section within the mobile app. The embedded AI system, including the assessment method, the exercise programs, and exercise instructions, were developed with an expert panel consisting of physiotherapists from Australia (Maxvale Physiotherapy Practice), the United Kingdom (University College London), and China (Sun Yat-sen University), including a clinical scientist and a rehabilitation doctor from China (Sun Yat-sen University). Technical support in developing and maintaining the app and AI coding was done in collaboration with the information technology company "Well Health" (Ying Kang Wei Er Internet Service Co, Ltd, Guangzhou, China). ### Development Process The AI-embedded mobile app underwent 10 months of development. Members of the expert panel met in China for 3 weeks for residential development. During the residential period, the clinical content of the AI-embedded mobile app was determined. This included the development of the AI algorithm (initial training sample preparation, neuron processing methods), production of the exercise videos and their associated descriptions, and the education material. This was followed by 3 months of AI model validation. During this phase, experts went through computer-simulated data to ensure the AI-generated exercise programs were appropriate for the presented symptoms. The development of the mobile app interface took a further 6 months. System usability and interface development of the app were conducted in the form of a semi-structured telephone interview (n=72) and focus group (n=8) prior to this study. The interview and focus groups were designed to gather information on (1) what users like to use the app for (eg, just browse for information or look for rehabilitation exercises for their back pain); (2) what additional functions they would like to have in the app; (3) how do they like the exercise content to be delivered (eg, exercises to be described by text or displayed as picture accompanied by text or in video with verbal guidance; (4) any in-app functions that users felt were missing; and (5) users' feelings on the app interface. No update was made to the content of the AI-embedded mobile app during the data collection phase. The AI-embedded mobile app is available free of charge online to the general public in the US, UK, and China through the Apple app store and MI app store. The app is searchable using the keyword "Well Health." ### Access Access to the mobile app was completely free. The AI-embedded mobile app could be downloaded from any well-known app store. Participants were required to register an account using their mobile phone number or social media account. Technical support was provided by Well Health in registering for a user account, if needed. Intervention Components ----------------------- ### Physical Component Well Health is a multiple-visit mobile app that provides adults with neck and back pain a tailored exercise rehabilitation program based on their presenting symptoms. The AI-embedded mobile app follows the recommendations of the National Institute of Clinical Excellence (NICE) \[[@ref21]\] in providing a combination of physical and self-management advice as well as the pathological causes of low back pain. It allows users to enter their symptoms using a self-report questionnaire. The questionnaire consists of the questions which physiotherapists routinely use to conduct a subjective assessment. The questionnaire asks for information on present condition, history of present condition, past investigations, 24-hour pattern, drug history, and social history. Once the subjective assessment is completed, the AI-embedded mobile app provides an exercise program based on the information provided. Exercise interventions provided by the AI-embedded mobile app were NICE guideline-recommended biomechanical exercises including stretching, (eg, lumbar flexion and neck flexion); motor control exercises, including Pilates exercises since their principles overlap with principles of motor control interventions \[[@ref22]\] (eg, deep neck flexors exercise, transverse abdominis activation exercise); and strengthening exercises (eg, cervical side flexion exercise, side-lying trunk exercise). [Figure 1](#figure1){ref-type="fig"} shows screenshots of these exercises. The suggested exercise duration was between 20 to 30 minutes a day. Each exercise video was accompanied by detailed descriptions of the exercises and annotated diagrams were displayed to visually demonstrate the targeted muscles (see [Multimedia Appendix 1](#app1){ref-type="supplementary-material"}). Users were expected to follow the video clip from start to finish once they started the set of training. The video clip could not be scrolled through, but could be paused/stopped. Points were awarded when the full length of the video was played. The AI-embedded mobile app determined the exercise program based on the subjective information. The AI algorithm used in this study was based on the multilayered perceptron artificial neural network (MLP-ANN). It is the most commonly used AI algorithm in the medical field \[[@ref23]\] and is capable of learning from historical examples, analyzing nonlinear data, and handling imprecise information \[[@ref24]\]. The AI algorithm observed the data input from the subjective assessment (input layer), processed the information through the neurons to select the appropriate therapeutic exercises from the exercise bank (hidden layer), and then gave out the most appropriate therapeutic program (output layer). [Figure 2](#figure2){ref-type="fig"} shows the architecture of the MLP-ANN of the AI used in this study. A total of 300 sets of training samples were initially provided by the experts during the expert meeting. Once the initial weighting of the MLP-ANN was trained by the samples provided by the experts, the back-propagation algorithm \[[@ref25]\] was used to continue the training until an accuracy of at least 80% was achieved. Model validation was performed by comparing the AI-generated exercise program with the expert-generated exercise program. {#figure1} {#figure2} The basic principle of the AI algorithm was based on the following: (1) each symptom was assigned a weighting in accordance to the clinical importance as determined by the expert panel (ie, the pain score had a higher weighting than social history); (2) each therapeutic exercise was assigned a weighting in accordance to their effectiveness to each of the presented symptoms as deemed by the expert panel; (3) the AI algorithm then selected the five highest scoring exercises from the database. Users were prompted to undertake reassessment every 14 days to report their progress and indicate the exercises that they found beneficial. The machine-learning algorithm would adjust the weighting of the exercise according to the individual's feedback and update the exercise program based on the input of the reassessment. In the case of acute symptoms worsening, the algorithm would adjust the exercise plan according to the new symptoms. All AI-embedded mobile app users could contact members of the medical team who were involved in developing the app via the in-app messaging function. As part of the disclaimer of using the AI-embedded mobile app, users were advised that it did not replace medical service and medical support should be sought at any time should they feel the need to do so. A points-based reward was utilized to promote engagement with the AI-embedded mobile app. Incentivized conditions have been shown to be effective in engaging users in goal-directed behavior \[[@ref26]\]. Points were awarded according to the daily task completed (see [Multimedia Appendix 2](#app2){ref-type="supplementary-material"}). The task list included logging onto the app, completing the number of sets of exercise, sharing their exercise progress on the social media account, and posting a comment about the educational material they read. Users needed to collect sufficient points to "unlock" the next level. The use of a point-based system aimed to increase engagement with the AI-embedded mobile app. The AI-embedded mobile app automatically recorded the number of days and the amount of time users spent on playing the exercise videos. An exercise log was kept and users could track their progress at any time. ### Educational Component Educational material was "pushed" to users via a social media platform once every 3 days. This material included information about neck and back pain pathology, the pathological cause of back pain and the natural course of low back pain, pain physiology, principles to use exercise as an intervention to manage neck and back pain, and coping strategies. These materials were stored within the app and users could access these materials at any time. Users received one reminder daily via the app reminder function which the user could turn off. Data Analysis ------------- Data were analyzed in SPSS version 20.0 (IBM, Armonk, NY, USA) for Windows (Microsoft 10). Descriptive statistics were used to describe the sample population and gave a summary on all responses. Changes in pain score and perceived self-improvement were analyzed as a group and as a subgroup based by the total duration of using the AI-embedded mobile app. Wilcoxon rank sum test was used to assess if the changes in Numerical Pain Rating Scale and self-perceived improvements were statistically significant. Results ======= Overview -------- During the data collection period, the AI-embedded mobile app had 461 active users. A total of 161 users responded to the invitation. Three responses were excluded due to exclusion criteria (reported to be younger than 18 years of age). The sample population contained 119 males and 39 females. Of these, 30 and 31 responses reported to be in the age group of 18 to 25 years and 26 to 30 years, respectively. The age group between 31 and 40 years had the greatest number of responses (n=56). Nineteen and 21 responses reported to be in the age group of 41 to 50 years and 51 to 60 years, respectively. One response reported to be older than 60 years. Time Spent on Therapeutic Exercises ----------------------------------- Most respondents reported they used the AI-embedded mobile app for a month. In all, 14.6% (23/158) and 13.3% (21/158) of respondents indicated they used the AI-embedded mobile app for 3 months and 6 months or more, respectively. The responses from question 3 of the evaluation questionnaire were 1 day (n=35), 1 week (n=38), 1 month (n=41), 3 months (n=23), and 6 months or more (n=21). A total of 60 users (37.98%) reported they had never participated in therapeutic exercises prior to using the AI-embedded mobile app. When using the AI-embedded mobile app, the mean time spent on rehabilitation exercises was 25 (SD 4) minutes per day. An increased number of responses was observed in the categories of 10 to 30 minutes and 30 to 60 minutes, but there was a decreased number of responses for the more than 60 minutes category. [Figure 3](#figure3){ref-type="fig"} is a bar graph showing the self-reported differences in time spent on therapeutic exercise before and when using the AI-embedded mobile app. Time Spent on Reading the Education Material -------------------------------------------- Overall, 142 users (89.9%) reported they had read the education material within the mobile app. Of these, 123 users (77.8%) indicated that the educational material had encouraged them to perform the AI-guided therapeutic exercise program. The mean time spent on reading the educational materials was 15 (SD 14) minutes a day. Numerical Pain Rating Scale --------------------------- An overall reduction in the Numerical Pain Rating Scale score was reported from users after using the AI-embedded mobile app for therapeutic exercises. The median Numerical Pain Rating Scale scores were 6 (interquartile range \[IQR\] 5-8) before and 4 (IQR 3-6) after using the AI-embedded mobile app (95% CI 1.18-1.81). A Wilcoxon rank sum test indicated the reduction in Numerical Pain Rating Scale score was statistically significant (*P*=.04). The greatest reduction in Numerical Pain Rating Scale scores was reported from users who used the AI-embedded mobile app for 6 months. [Figure 4](#figure4){ref-type="fig"} shows the changes in self-reported Numerical Pain Rating Scale scores reported by users with different tenures of using the AI-embedded mobile app. {#figure3} {#figure4} Self-Perceived Improvement -------------------------- The mean self-perceived improvement from all users was 65%. Users who utilized the AI-embedded mobile app for 1 day, 1 week, and 1 month reported improvements of 60.2% (SD 27.9%), 49.5% (SD 29.2%), and 47.2% (SD 26.4%), respectively. For users who used the app for over 3 and 6 months, improvements were reported to be 58.6% (SD 30.4%) and 71.1% (SD 25.8%), respectively. Usage of Other Interventions ---------------------------- A total of 11 respondents (7.0%) indicated that they had not received prior intervention from the listed modalities and 38 respondents (24.1%) indicated they did not receive any other intervention while using the AI-embedded mobile app. A reduction in responses was observed in all the listed interventions when using the AI-embedded mobile app. [Figure 5](#figure5){ref-type="fig"} shows the number of responses of who had used other interventions prior to and when using the AI-embedded mobile app. System Usability Scale ---------------------- The median of the SUS from all responses was 73 (IQR 60-85), which would suggest an acceptable level of usability (score \>68). All users, except for those who used the AI-embedded mobile app for a day (median 65, IQR 53-80), reported acceptable usability (score \>68). {#figure5} Discussion ========== Principal Findings ------------------ The main findings of this preliminary evaluation study were (1) users reported having spent more time on therapeutic exercises when using the AI-embedded mobile app, (2) an overall reduction in pain level was reported after using the AI-embedded mobile app, and (3) the usage of other interventions was reported to have been reduced with the use of the AI-embedded mobile app. This is among the first studies that investigated the feasibility and benefits of an AI-embedded mobile app for the management of chronic neck and low back pain. Adherence to Therapeutic Exercise --------------------------------- Exercise therapy had been recommended as a core component in the management of chronic neck and back pain by several guidelines and reviews \[[@ref27]-[@ref29]\]. It was an unexpected finding that 38% of the users indicated that they had never participated in therapeutic exercises before. It was possible that some of the users might live in a remote part of China where rehabilitation service was not widely available. The use of the AI-embedded mobile app may therefore play a role in increasing access to exercise therapy intervention, as supported by the reported increase in time engaged in therapeutic exercises. Adherence to a therapeutic exercise program in the management of chronic neck and low back pain has been a long-standing issue. Early literature showed that between 50% \[[@ref30]\] and 70% \[[@ref31]\] of people with chronic low back pain were nonadherent to a home exercise program. The use of technologies has been suggested to be a way to increase adherence to a home exercise program \[[@ref32]\]. Previous studies that investigated the effectiveness of using mobile phone text message reminders indicated continued adherence to exercises in adults with recurrent low back pain \[[@ref7]\]. The AI-embedded mobile app also had a reminder function that reminded users to perform daily exercise. A survey conducted in 2016 on the Chinese population with chronic low back pain (N=113) reported that self-management behaviors were poor \[[@ref15]\] and contributing factors included lack of disease knowledge and lack of understanding of the benefit of exercise. The educational material may therefore play a role in promoting regular therapeutic exercise because it provides information on these areas. The majority of users indicated the material motivated them to do therapeutic exercise. The increase in self-reported time spent on exercises observed in this study may suggest its potential to be a useful tool for patient adherence to therapeutic exercise programs, which may contribute to improving clinical outcomes. Pain Level ---------- A pain level reduction of 2 points (on a 0-10 Numerical Pain Rating Scale) is considered to be "much improved" \[[@ref33]\] and a "meaningful" change \[[@ref34]\] in people with chronic pain. The reduction in pain score was consistent with the finding of the Cochrane Review that indicated exercise, including motor control, flexibility, and strengthening exercises, may improve pain severity for people with chronic pain \[[@ref35]\]. The difference in self-reported pain ratings was also consistent with a study that investigated the benefit of the combination of manual therapy and exercise therapy alone \[[@ref36]\]. In this study, Alfuth and Welsink \[[@ref36]\] investigated the pain level in people with chronic low back pain who attended physiotherapy practice for interventions. Interventions included a combination of manual therapy, stabilization exercises, and electrotherapy. A total of 85 participants received a mean number of treatment sessions of 42.6 (SD 2.3) within a year. A reduction of 3 points and 4 points on the Numerical Pain Rating Scale was observed after 3 and 6 months of treatment, respectively, when compared to baseline. Compared to the 2- and 3-point difference reported by users who used the AI-embedded mobile app for 3 and 6 months in this study, it would appear that using the AI-embedded mobile app to manage low back pain is feasible. An earlier study by Cambron et al \[[@ref37]\] investigated the effect of a 4-week course of an active trunk extension exercise program at 3 months and 6 months postintervention. They reported a mean pain score reduction of 1.36 and 1.19 at 3 and 6 months postintervention, respectively. The difference in pain score reduction may be related to the type of exercises that participants were asked to do. The AI-embedded mobile app selected from a range of progressive exercises, whereas participants in the Cambron et al study \[[@ref37]\] were only asked to do trunk extension exercises. Users' self-ratings of improvement of interventions is one of the core domains to be used in assessing impact of an intervention on chronic pain \[[@ref38]\]. The self-rating of symptom improvement observed in this study corresponded to the reduction in pain score. The greatest symptom improvement was reported from users who had used the program for 6 months, followed by 3 months. A study published in 2010 reviewed 15 randomized controlled trials that were published between 2001 and 2007 \[[@ref39]\]. The study summarized literature that reported long-term outcome data of physiotherapy exercise programs for adults with chronic low back pain. Trials were reviewed if they reported any outcome related to chronic low back pain measured at a follow-up period. Of the included trials, 13 reported follow-up data of up to 12 months and two trials reported follow-up data of up to 2 years. The results indicated consistent evidence for the effectiveness of an exercise program in pain reduction at the 6 months follow-up period when compared to a control group. Our study also observed the greatest self-reported pain score reduction and self-perceived improvement from users who used the AI-embedded mobile app for 6 months. When compared to the only other trialed mobile app (FitBack) \[[@ref40]\] that focused specifically on spinal pain management \[[@ref10]\], the AI-embedded mobile app used in our study appeared to be at least as effective in reducing pain level. FitBack users reported a reduction in pain score from 2.86 at baseline to 2.64 and 2.16 at follow-up at 2 and 4 months, respectively. A Cochrane Review \[[@ref35]\] and the NICE guideline \[[@ref21]\] indicated that exercise programs should be tailored according to individual needs and capabilities. Therefore, the higher reduction of pain scores observed in this study may be related to the tailoring of the exercise content based on the presenting symptoms. But the higher magnitude of pain score reduction observed in our study should be interpreted with caution. The two studies have different designs, which may contribute to the differences in reported pain levels. It is typical for single-group data to show greater effects than randomized controlled trial data. The pain score data from our study may also contain recall bias \[[@ref39]\], which further complicates the comparison. Needs for Other Interventions ----------------------------- The results of our study indicate that using the app for the self-management of chronic neck and back pain might reduce the need for other interventions. The reported reduction in usage of other interventions was supported by an earlier study by Sculco et al \[[@ref41]\]. Their study investigated the effect of 10 weeks of a prescribed home-based exercise program in people with chronic low back pain compared to a control group. The results suggested that participants in the exercise group had significantly lower numbers of physiotherapy referrals and medication use during the 2.5-year follow-up period. The reported reduction of other intervention usage suggests that the AI-embedded mobile app may at least reduce the ambulatory time to medical appointments, thus reducing the indirect health care costs that are associated with seeking intervention. Further study is recommended to understand its cost benefit within the local health care system. Limitations ----------- The results reported here must be viewed cautiously due to the limitations. This was a single-group study with no control which relies heavily on self-reported subjective data. This is among the first AI-embedded mobile apps developed for the management of low back pain. Thus, the primary goal of this preliminary study was to evaluate the perceived benefits of such a system by its users. With regard to the fidelity of the intervention, it is worth considering that the app was available for download by members of the general public. Users were not advised by anyone to download it; they downloaded it on their own initiative, assuming that they suffered from chronic neck and back pain and were looking for self-help material on the Web. Thus, it is reasonable to believe that users had followed the therapeutic program provided by the AI-embedded mobile app. This study defined the term "adherence" as the amount of self-reported time spent on the exercise generated by the AI-embedded mobile app. Therefore, the interpretation of adherence within this study cannot be generalized to the long-term adherence to therapeutic exercise. The "law of attrition" of eHealth trials \[[@ref42]\] did not appear to be applicable in this study because this study only recorded response rate but not dropout rate. We could not verify whether participants' eligibility criteria information were accurate, which may affect their responses to the intervention. The amount of time participants viewed the video clips were not verified. However, the AI-embedded mobile app has a scoring system which the user had to satisfy by playing the video clip for the exercises. Further study that combines this with medical verification would provide greater confidence in the intervention effects. This study did not evaluate the functional aspect of the intervention. Thus, it was unclear if the changes in pain score may translate to improvement in function. We included active users at the time of the study, which many have affected the response rate. The study design also had sample selection bias because it relied on people who responded to the invitation, thus the sample population would be more likely to participate in the intervention. This was a retrospective evaluation study, which was likely to be associated with recall bias of preintervention symptoms and behaviors. The focus of this preliminary study was to assess the self-perceived benefits from users. Self-reported data on health care utilization are commonly used in health care research \[[@ref43]\]. Self-reported data may be associated with recall bias; however, a study that investigated agreement between self-reported health care service usage and administrative records indicated good agreement between the two \[[@ref44]\]. Conclusions ----------- The key findings of this evaluation study support the perceived beneficiary effects of the AI-embedded mobile app to provide some personalized interventions that are tailored to individual needs for the self-management of chronic neck and back pain. The positive results of this study suggest that using the tool to assist the self-management of chronic spinal pain may be feasible. The tool at least warrants further study to investigate the benefit of the AI-embedded mobile app and how it may compare with routine clinical care. Further study is also recommended to understand if the AI-embedded mobile app may induce long-term sustainable behavioral change. This research project was supported by China Postdoctoral Science Foundation Grant (\#2016M592581) and Guangzhou City Research & Technology Collaborative Innovation Projects (\#2017010160387). Conflicts of Interest: None declared. Screenshots of a detailed description of the exercise (left) and of the annotated diagrams to show the targeted muscles (right). Screenshots of the daily task list (top) and the exercise log where users could track their progress (bottom). AI : artificial intelligence eHealth : electronic health ICT : information and communication technology IQR : interquartile range MLP-ANN : multilayered perceptron artificial neural network NICE : National Institute of Clinical Excellence NPRS : Numerical Pain Rating Scale SUS : System Usability Scale

INTRODUCTION ============ Some groups of pseudomonads, indigenous in the rhizosphere, produce and excrete secondary metabolites that are inhibitory to plant pathogens \[[@R1]\]. They offer a sustainable alternative to pesticides and can provide disease control when the use of pesticides is legally restricted. Among the metabolites produced by *Pseudomonas* spp., the compound 2,4-diacetylphloroglucinol (DAPG) has been intensively studied due to its broad spectrum activity towards the commercially important plant pathogens*Gaeumannomyces graminis* var. *tritici*, *Thielaviopsis basicola*, *Pythium ultimum* and *Rhizoctonia solani* \[[@R2]\]. DAPG-producing pseudomonads have been detected in the rhizosphere of various crops, e.g. cucumber, maize, pea, tobacco, tomato, wheat and oat \[[@R3]-[@R9]\]. The presence of the compound has been confirmed in different cropping systems such as soils and hydroponic systems \[[@R10],[@R11]\] and the importance of DAPG *in situ* has been verified with DAPG- deficient mutants \[[@R12]\]. An obstacle in the development of efficient biocontrol agents is the inconsistency of their performance \[[@R13]\]. The complex *in situ* condition in the natural rhizosphere influences survival, growth and production of secondary metabolites by the biocontrol strains. With respect to DAPG production, mineral and carbon sources as well as metabolites released by indigenous microflora and plants have been identified as being critical factors \[[@R12],[@R14]-[@R16]\]. An important measure in developing reliable biocontrol systems is to identify suitable conditions for optimal metabolite formation by the biocontrol strains and to provide these conditions in plant cultivation systems. For outdoor field cropping systems this is a considerable challenge due to unavoidable variations in environmental conditions. However, greenhouse cropping systems with their more controlled environment offer a niche for successful development of biocontrol \[[@R17]\]. However, it is still not an easy task bearing in mind the multifactorial nature of biocontrol interactions and given that the needs of the plant are the main concern. The present study focuses on the potential to improve the reliability of biological control in soilless greenhouse systems. The objective was to determine the impact of nitrogen source on DAPG production. The experiment was performed *in vitro*, but with hydroponic cultivation of tomato in mind. The well-known biocontrol strain Pf-5 \[[@R18]\] was cultivated in media based on the nutrient solution commonly used in tomato hydroponic systems \[[@R19]\]. The ratio of ammonium to nitrate ions was varied and the media were amended with organic nitrogen sources found in tomato root exudates \[[@R20]\], and the bacterial growth and DAPG production were studied. MATERIALS AND METHODS ===================== Microorganism and Inoculum Preparation -------------------------------------- The bacterial strain Pf-5 \[[@R18]\] was pre-cultured on nutrient agar for 48 hours at room temperature. This strain was kindly provided by Dr. Loper, University of California, USA, and stock cultures of the strain were kept in 40% glycerol (v/v) at -80 °C. A single colony from the nutrient agar was transferred to 15 mL of nutrient broth (Difco® 0003-17) in a 100 mL Erlenmeyer flask and incubated on a rotary shaker (180 rpm, 24 hours at room temperature). The cells were centrifuged (8000 g, 4 °C, 15 minutes; Avanti J-20, Beckman Coulter, CA, USA). The supernatant was discarded and the pellet was washed once in 0.85% NaCl and finally dissolved in 0.85% NaCl to an optical density of 1.0 at 620 nm. Composition of the Media and Culture Conditions ----------------------------------------------- In medium A no organic nitrogen was added, in medium B organic nitrogen was added in a millimolar concentration range and in medium C organic nitrogen was added in a micromolar concentration range. In treatment 1 (A1, B1 and C1) the inorganic nitrogen was provided as both nitrate and ammonium nitrate. In treatment 2 (A2, B2 and C2) the inorganic nitrogen was provided as nitrate only and in treatment 3 (A3, B3 and C3) the inorganic nitrogen was provided as ammonium only. The total concentration of inorganic nitrogen was 17 mM in all media. Glucose (Sigma-Aldrich Co., USA) (2%) was added as a carbon source in all media. The composition of salts and pH of medium A1 corresponded to those in the nutrient solution commonly used for hydroponic cultivation of tomato: KNO~3~ 8.0, NH~4~NO~3~ 0.6, MgSO~4~ 1.9, KH~2~PO~4~1.5, Ca(NO~3~)~2~ 4.2 **(mM)**, Fe-EDTA 18.7, MnSO~4~ 12.5, ZnSO~4~ 6.3, H~3~BO~3~ 31.3, CuCl~2~ 0.94, Na~2~MoO~4~ 0.6 **(μM)**, pH 5.9. Medium A2 contained nitrate as the only nitrogen source and NH~4~NO~3~was excluded and replaced with an equal amount of KNO~3~ and Ca(NO~3~)~2~. Medium A3 contained ammonium (NH~4~Cl) as the sole nitrogen source. In media B1, B2 and B3, the composition of the broth was the same as described above and the amino acids aspartic acid, glutamic acid, isoleucine, leucine and lysine were added at concentrations of 7.5, 6.8, 7.6, 7.6 and 6.8 mM respectively. These amino acids have previously been shown to be present in root exudates from tomato \[[@R20]\]. All amino acids were L-form and purchased from Sigma-Aldrich Co., USA. In media C1, C2 and C3, the composition of the broth was the same as described for treatment B1, B2 and B3, except that the concentration of the amino acids was at micromolar level (concentrations of 7.5, 6.8, 7.6, 7.6 and 6.8 μM respectively). These concentrations are approximately in agreement with the concentrations observed in tomato root exudates \[[@R20]\]. All media were supplemented with adenine, cytosine, guanine, uracil, thymine and vitamins according to Slininger and Shea-Andersh \[[@R21]\]. All salts, amino acids and vitamins were added to the broth after it had been autoclaved and cooled down. After this the pH was set to 5.9 with NaOH and the broth was membrane-filtered (0.22 μm, Sarstedt, Germany) before the start of the experiment. Nutrient broth amended with 2% glucose (NBglu) has previously been shown to be superior for DAPG production by strain Pf-5 \[[@R22]\] and was included as a control. A volume of 15 mL of the respective medium was added to 100 mL Erlenmeyer flasks and an aliquot of 100 μL inoculum was added to each flask. The flasks were incubated on a rotary shaker (180 rpm at room temperature) and samples were withdrawn after 24, 48 and 72 hours. Analyses -------- DAPG was analysed according to Slininger and Shea-Andersh \[[@R21]\] with slight modifications. Samples were taken after 24, 48 and 72 hours of incubation and centrifuged (8000 g at 4 °C for 15 minutes). A volume of 20 μL of the supernatant was applied to a Waters Symmetry C-18 column (WAT054275, 5 μm particle size, 4.6 diam., 250 mm long) and a mobile phase of 65/35 acetonitrile (HPLC grade)/water with 0.1% glacial acetic acid with UV detection at 270 nm, flow 1 mL min^-1^. Concentrations were calculated relative to standards prepared using DAPG (Toronto Research Chemicals Inc., Ontario, Canada, lot: 1-JS-36-1), retention time 5.8 min. The detection limit for DAPG was approximately 0.5 μg mL^-1^. The spectrum of the peak was compared to the standard and published spectrum \[[@R23]\] to ensure the identity of the peak. The optical density at 620 nm was monitored over time and the generation time in the various media was calculated. Protein content in the samples was measured after 72 hours of growth. An aliquot of 3 mL of cell suspension was centrifuged (8000 g at 4 °C for 15 minutes). The pellet was washed twice and then re-suspended in sterile water. The detergent Triton X (Sigma-Aldrich Co., USA) was added to the sample at a concentration of 0.1% (v/v) and the samples were intensively vortexed. Thereafter, the samples were frozen for 24 hours and after thawing they were homogenised using a glass homogeniser (Duall® 21, Kontes Glass Co., Vineland, NJ, USA). The protein content in each sample was determined by the method of Lowry *et al.* \[[@R24]\], using a standard curve prepared with bovine serum albumin (Sigma-Aldrich Co., USA). Experimental set-Up and Statistics ---------------------------------- All experiments were performed with four replicates and repeated once. Data were analysed by analysis of variance followed by Tukey´s multiple comparison test and P\<0.05 was considered significant (Minitab, version 14). RESULTS ======= When Pf-5 was grown in NBglu, a high production of DAPG was observed. After 24 hours the concentration in the medium was 11.0 μg mL^-1^ and the concentration increased during the following 48 hours to 100.7 μg mL^-1^ (Fig. **[1](#F1){ref-type="fig"}**). The generation time was 6.1 hours and the optical density after 72 hours was 1.54, corresponding to 2926 μg mL^-1^ of protein (Table **[1](#T1){ref-type="table"}**). In medium A amended with only inorganic nitrogen sources, nitrate and ammonium nitrate (A1), nitrate only (A2) and ammonium only (A3), no production of DAPG was observed (Table **[1](#T1){ref-type="table"}**). No significant differences were observed between the various nitrogen sources as regards bacterial growth. The generation time, optical density and protein content are presented in Table **[1](#T1){ref-type="table"}**. When amino acids were added in a millimolar concentration range (treatment B), DAPG production was observed in medium B3, in which ammonium was used as the inorganic nitrogen source. After 24 hours the concentration was 19.8 µg mL^-1^ and declined thereafter (Fig. **[1](#F1){ref-type="fig"}**). Only low DAPG production, below the detection limit, was observed in some samples from medium B1 or B2. Treatment B supported considerably higher bacterial growth than treatments A and C, but the bacterial growth was still significantly lower than in NBglu (Table **[1](#T1){ref-type="table"}**, Fig. **[2](#F2){ref-type="fig"}**). No significant differences were observed when the optical density after 72 hours was compared within treatment B (Table **[1](#T1){ref-type="table"}**, Fig. **[2](#F2){ref-type="fig"}**). However, the protein content was significantly higher for treatment B2 compared with treatments B1 and B3 (Table **[1](#T1){ref-type="table"}**). When amino acids were added to a micromolar concentration range (treatment C) no DAPG production was observed. The bacterial growth was low overall and not significantly different from treatment A (Table **[1](#T1){ref-type="table"}**). DISCUSSION ========== The main finding of this *in vitro* study was that the nitrogen source, both inorganic and organic, clearly influences DAPG production by biocontrol strain Pf-5. Screenings for DAPG production *in vitro* have mostly been done using complex bacterial growth media or using a defined media and addition of yeast extract or casamino acids \[[@R3],[@R10],[@R15],[@R16],[@R21],[@R25]\]. In the present study no organic nitrogen was added in treatment A. This resulted in poor growth and no production of DAPG, independent of the inorganic nitrogen source. These results are in contrast to results obtained by Siddiqui and Shaukat \[[@R26]\], who observed a nematicidal activity of filtrate from the DAPG-producing strain CHA0 when cultured on a medium lacking organic nitrogen. However, in their study it was not confirmed that the nematicidal activity was due to DAPG and other metabolites might have caused the biocontrol effect. When the medium in the present study was supplemented with organic nitrogen in a millimolar concentration range, a significantly higher growth rate was observed. However, substantial DAPG production was only observed in medium B3, in which ammonium was included as the inorganic nitrogen source. This result cannot be explained by an increased cellular density in this treatment, as there was no difference in optical density within treatment B. Furthermore, expressing cellular growth as protein yield showed that medium B2 sustained the highest growth despite the fact that DAPG production was only occasionally observed in this medium and then at concentrations below 0.5 µg mL^-1^. The result concerning an increased production of DAPG when ammonium was included as the inorganic nitrogen source is in agreement with other results. Crowley *et al*. \[[@R27]\] observed that DAPG production by strain *Pseudomonas fluorescens* F113 was stimulated when the nitrogen source was in form of ammonium ions. Duffy and Defago \[[@R15]\] also reported an increased production of DAPG by the strain *Pseudomonas fluorescens* CHA0 when ammonium was added. However, the question concerning whether ammonium ions stimulate DAPG production in general or only in selected strains needs to be further investigated. It has previously been shown that the carbon source that stimulates the highest DAPG production is strain-specific \[[@R15]\] and it is possible that a similar situation may apply regarding the inorganic nitrogen source. Previous work has shown that the amino acid concentration in tomato root exudate is in the micromolar concentration range \[[@R20]\]. The results of the present study show that this concentration is to low to support production of DAPG *in vitro*. However, DAPG has been detected in tomato hydroponic systems \[[@R11]\]. In these plant cultivation systems, especially if recirculated, organic nitrogen is accumulated and it is possible that sufficient levels are reached. It is also likely that nutrient-rich microsites in the immediate surroundings of the root provide a niche. Furthermore, microbial products have been suggested to trigger amino acid exudation from roots \[[@R28]\]. Plants naturally cycle amino acids across their root cell plasma membranes and Phillips *et al*. \[[@R28]\] showed that in the presence of DAPG, the amino acid uptake by the plant was blocked. The previously cited study by Simons *et al*. \[[@R20]\], which showed a micromolar concentration range of amino acids in tomato root exudates, was performed on axenically cultured tomato roots. It is possible that higher levels of amino acid exudation would have been observed if those experiments had been performed in the presence of DAPG. The findings of the present study suggest that it would be possible to adapt hydroponic cultivation systems to increase the production of DAPG by added biocontrol strains. The preferred inorganic nitrogen source should be ammonium and the amount of organic carbon should not be below a millimolar concentration range. However, the needs of plants also have to be addressed and possible effects of these changes on plants need to be further investigated. This study was supported by a grant from the Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (FORMAS). {#F1} {#F2} ###### DAPG production, generation time, optical density and protein content when the strain Pf-5 was grown in media with varying nitro-gen content. In treatment A only inorganic nitrogen was added (1= NO~3~^-^ and NH~4~^+^, 2 = NO~3~^-^ only and 3 = NH~4~^+^ only). In treatment B amino acids were added in a millimolar concentration range. In treatment C amino acids were added in a micromolar concentration range. The medium NBglu was included as a control Medium DAPG\* µg mL^-1^ Generation time (h) Optical density\* \* \* \* (620) nm Protein\* \* \* \* µg mL^-1^ -------- ------------------ --------------------- ------------------------------------- ------------------------------ NBglu 11.0±3.1 6.1 1.54±0.02a^1^ 2926±223.8a A1 ND\* \* 23.5 0.09±0.004b 73.9±25.5b A2 ND 23.4 0.09±0.003b 73.9±13.4b A3 ND 24.4 0.09±0.001b 66.9±23.7b B1 BD\* \* \* 5.2 1.03±0.052c 1421.1±125.4c B2 BD 5.3 1.07±0.052c 1998.9±151.9c B3 19.8±4.5 5.2 0.95±0.117c 1607.22±151.2d C1 ND 25.3 0.09±0.01b 66.8±14.2b C2 ND 42.1 0.09±0.007b 55.0±14.1b C3 ND 25.4 0.11±0.004b 74.2±2.2b \* after 24 hours of growth \* \* ND not detected \* \* \* BD below detection limit \* \* \* \* after 72 hours of growth ^1^ Figures within columns followed by different letters are significantly different (P\<0.05).