id
stringlengths 40
40
| metadata
dict | text
stringlengths 9
2.66M
| created
timestamp[s]date 2025-03-04 00:00:00
2025-03-07 00:00:00
| source
stringclasses 1
value | attributes
dict |
|---|---|---|---|---|---|
38feaa07d637d88b2ca5cae0cc45235a902e0b4b | {
"olmocr-version": "0.1.59",
"pdf-total-pages": 15,
"total-fallback-pages": 0,
"total-input-tokens": 53811,
"total-output-tokens": 19366,
"fieldofstudy": [
"Agricultural And Food Sciences"
]
} | Lilium regale Wilson WRKY2 Regulates Chitinase Gene Expression During the Response to the Root Rot Pathogen Fusarium oxysporum
Shan Li†, Jun Hai†, Zie Wang, Jie Deng, Tingting Liang, Linlin Su and Diqiu Liu*
Faculty of Life Science and Technology, Kunming University of Science and Technology, Kunming, China
Root rot, mainly caused by Fusarium oxysporum, is the most destructive disease affecting lily (Lilium spp.) production. The WRKY transcription factors (TFs) have important roles during plant immune responses. To clarify the effects of WRKY TFs on plant defense responses to pathogens, a WRKY gene (LrWRKY2) was isolated from Lilium regale Wilson, which is a wild lily species highly resistant to F. oxysporum. The expression of LrWRKY2, which encodes a nuclear protein, is induced by various hormones (methyl jasmonate, ethephon, salicylic acid, and hydrogen peroxide) and by F. oxysporum infection. In this study, LrWRKY2-overexpressing transgenic tobacco plants were more resistant to F. oxysporum than the wild-type plants. Moreover, the expression levels of jasmonic acid biosynthetic pathway-related genes (NtAOC, NtAOS, NtKAT, NtPACX, NtJMT, NtOPR, and NtLOX), pathogenesis-related genes (NtCHI, NtGlu2, and NtPR-1), and antioxidant stress-related superoxide dismutase genes (NtSOD, NtCu-ZnSOD, and MnSOD) were significantly up-regulated in LrWRKY2 transgenic tobacco lines. Additionally, the transient expression of a hairpin RNA targeting LrWRKY2 increased the susceptibility of L. regale scales to F. oxysporum. Furthermore, an F. oxysporum resistance gene (LrCHI2) encoding a chitinase was isolated from L. regale. An electrophoretic mobility shift assay demonstrated that LrWRKY2 can bind to the LrCHI2 promoter containing the W-box element. Yeast one-hybrid assay results suggested that LrWRKY2 can activate LrCHI2 transcription. An examination of transgenic tobacco transformed with LrWRKY2 and the LrCHI2 promoter revealed that LrWRKY2 activates the LrCHI2 promoter. Therefore, in L. regale, LrWRKY2 is an important positive regulator that contributes to plant defense responses to F. oxysporum by modulating LrCHI2 expression.
Keywords: Lilium regale, Fusarium oxysporum, WRKY transcription factors, transcriptional regulation, W-box, chitinase
INTRODUCTION
Plants have evolved a series of responsive and adaptive mechanisms to cope with various stresses. The WRKY transcription factors (TFs), which represent one of the largest families of transcriptional regulators in plants, are mainly involved in regulating the plant immune system during responses to stress (Wang P. et al., 2018). These TFs usually contain one
or two conserved domains comprising approximately 60 amino acids (Yan et al., 2019a). The domain with a highly conserved WRKYQGK motif at the N-terminal end and a zinc-finger motif (CX$_4$–5CX$_{22}$–23HXX or CX$_7$–CX$_{23}$HXH) at the C-terminal end was designated as the WRKY domain (Jimmy and Babu, 2019). On the basis of the number of WRKY domains and the structure of the zinc-finger motif, WRKY TFs have been classified into three main groups (I, II, and III) (Dong et al., 2020). The Group I members have two WRKY domains and a C$_2$H$_2$ (CX$_4$–5CX$_{22}$–23HXX) zinc-finger motif, the Group II members contain one WRKY domain and a C$_2$H$_2$ (CX$_4$–5CX$_{22}$–23HXX) zinc-finger motif, and the Group III members contain a single WRKY domain and a C$_2$HC (CX$_7$–CX$_{23}$HXH) motif (Nan and Gao, 2019). Moreover, the Group II WRKY TFs have been further divided into Subgroups IIa, IIb, IIc, IId, and Ile according to their primary amino acid sequences (Huang et al., 2015).
Transcription factors bind to cis-acting elements in the promoters of their downstream target genes to regulate expression (Long et al., 2019). The WRKY TFs are a family of plant-specific DNA-binding proteins (Rushton et al., 2010). More specifically, the WRKY domain binds to the W-box (TTGACC/T) in the promoters of target genes to activate, repress, or de-repress transcription. WRKY TFs can also bind to several W-box-like elements (Choi et al., 2015). It is due to WRKY TFs binds to the TGAC core motif of the W-box, whereas the preferred WRKY TF-binding sites are determined by the sequences adjacent to the TGAC core motif (Choi et al., 2015). The WRKYQGK and zinc-finger motifs are required for the high-affinity binding of WRKY proteins to the consensus cis-element W-box present in the promoters of target genes, including those encoding pathogenesis-related (PR) proteins (Karim et al., 2015). The solution structure of the WRKY domain was first reported by Yamasaki et al. (2005). The WRKY domain of the Arabidopsis thaliana WRKY4 TF comprises a four-stranded β-sheet, with zinc-coordinating Cys/His residues forming a zinc-binding pocket (Yamasaki et al., 2005). The WRKYQGK residues form the most N-terminal β-strand, which partly protrudes from the surface of the protein, enabling it to interact with the DNA major groove. The β-strand including the WRKYQGK motif reportedly interacts with an approximately 6-bp region, which is consistent with the length of the W-box (Jiang et al., 2016). The WRKYQGK motif recognizes DNA mainly through apolar associations with the methyl groups of the thymine bases present in the W-box which was reviewed by Viana et al. (2018).
The WRKY TFs have a very complex regulatory role influencing plant disease resistance. They are central components mediating innate immunity, including molecular pattern-triggered immunity, pathogen-associated molecular pattern-triggered immunity, effector-triggered immunity, basal defense, and systemic acquired resistance (Eulgem and Somssich, 2007). The activation of WRKY TFs under stress conditions leads to a thorough transcriptional reprogramming affecting defense-related genes, including genes for secondary metabolites (phytoalexins such as momilactones and phenylamides) and PR proteins (PR3, PR4c, PR10a, PR10b, and root-specific PR10) (Jimmy and Babu, 2019). The constitutive expression of rice (Oryza sativa) WRKY45 in A. thaliana enhances the resistance of the transgenic plants to the pathogen Pseudomonas syringae pv. tomato DC3000 by up-regulating the expression of some PR genes (Qiu and Yu, 2009). In Capsicum annuum, WRKY40b is a negative regulator that directly affects immunity at multiple levels by modulating the production of signal regulatory proteins, TFs, and PR proteins (Iftian Khan et al., 2018). In a previous study, the overexpression of apple (Malus × domestica) WRKY1N1 in the susceptible apple cultivar “Golden Delicious” resulted in increased resistance to the apple leaf spot fungus Alternaria alternata f. sp. mali, which was related to the significantly up-regulated expression of PR genes (Zhang et al., 2017).
Root rot caused by Fusarium oxysporum seriously threatens the growth and development of lily (Lilium spp.). Several wild Lilium species are highly resistant to F. oxysporum, including L. regale Wilson, L. pumilum, and L. dauricum Ker Gawler (He et al., 2014, 2019; Ramekar et al., 2019). Although the germplasm resources of Lilium species resistant to F. oxysporum have been used for breeding disease-resistant lines, little is known about their mechanisms regulating transcription during an infection by F. oxysporum. In a recent study, 35 WRKY genes were identified in L. regale after the transcriptomes of plants inoculated with F. oxysporum were sequenced (Li et al., 2021b). At the transcriptional level, LrWRKY2 is highly responsive to methyl jasmonate (MeJ), salicylic acid (SA), ethephon (ETH), and hydrogen peroxide (H$_2$O$_2$) treatments as well as F. oxysporum infections. Thus, in this study, we focused on unraveling the LrWRKY2 (accession no. MW125547) regulatory mechanism during the L. regale defense response to F. oxysporum. The tobacco is an important cash crop and model plant used for rapid verification of plant gene functions, moreover, the F. oxysporum causes root rot in tobacco (Clemente, 2006; Ding et al., 2021). In addition, the infection rate of F. oxysporum in tobacco seedlings is faster than the lily, and the obvious disease symptoms can appear within a week after inoculation. Therefore, the effects of LrWRKY2 on the resistance to F. oxysporum was evaluated by overexpressing LrWKY2 in tobacco plants and transiently expressing the hairpin RNA targeting LrWRKY2 in L. regale scales considering the long period of lily genetic transformation and great difficulties to obtain enough transformants for subsequent experiments. Furthermore, LrWRKY2 was characterized regarding its subcellular localization and its ability to activate the promoter of an L. regale PR gene (LrCHI2), which is an F. oxysporum-responsive gene according to transcriptome sequencing data. Additionally, tobacco plants were transformed with the LrCHI2 promoter and LrWRKY2 to analyze the regulatory effects of LrWRKY2 on LrCHI2 expression.
**MATERIALS AND METHODS**
**Plant and Fungal Materials**
Wild L. regale Wilson plants collected in the Minjiang River Basin of Sichuan province, China, were grown in a greenhouse. An F. oxysporum strain isolated from diseased lily plants exhibiting typical symptoms of Fusarium wilt was characterized and preserved by our research group. Several typical plant fungal pathogens (Colletotrichum gloeosporioides, Fusarium solani, and
*Alternaria panax* were also selected for the antifungal activity analysis. *F. oxysporum*, *C. gloeosporioides*, *F. solani*, and *A. panax* stored at 4°C were cultured on potato dextrose agar (PDA) before use. Sterile tobacco seedlings were cultured in a climate-controlled cabinet prior to genetic transformations.
**Gene Cloning**
The total RNA was extracted from 1 g wild *L. regale* roots using the TRIgene kit (Genstar, China). The cDNA was obtained following the reverse transcription of total RNA using the GoScript™ Reverse Transcription System (Promega, USA) and served as the template for isolating *LrWRKY2* and *LrCHI2*. Gene-specific primers (Supplementary Table 1) were designed to amplify the *LrWRKY2* and *LrCHI2* open reading frames (ORFs) using *L. regale* cDNA as the template. The PCR products were cloned into the pMD-18T vector (TaKaRa, Japan) and the resulting recombinant vectors pMD-18T-LrWRKY2 and pMD-18T-LrCHI2 were inserted into competent *Escherichia coli* DH5α cells. The positive clones on a LB plate adding the ampicillin (50 mg/L) were verified by PCR and sequencing in Tsingke Biotechnology Co., Ltd. Bioinformatics analyses were completed as described by Taif et al. (2020).
**Quantitative Real-Time PCR (qRT-PCR)**
The *LrCHI2* expression pattern in *L. regale* was analyzed in a qRT-PCR assay. The roots, leaves, stems, flowers, and scales of healthy *L. regale* plants were used to analyze tissue-specific expression of *LrCHI2* expression. Regarding the fungal inoculation, the root tips were wounded using surgical scissors and inoculated with a fresh *F. oxysporum* spore suspension (5 × 10⁶ spores/mL) for 30 min. The control plants were treated with sterile water instead of the spore suspension. The roots were sampled at 12, 24, 48, and 72 h post-inoculation. Finally, total RNA extracted from each sample was reverse transcribed into cDNA for a qRT-PCR analysis of *LrCHI2* expression levels using the GoScript™ Reverse Transcription System (Promega, USA). The SYBR green I-based qRT-PCR analysis was conducted using the ABI Prism 7500 Sequence Detection System (Applied Biosystems, USA) and Easteq® qPCR Master Mix (Promega, USA) according to an established method (Zhao et al., 2020). The *L. regale* glyceraldehyde-3-phosphate dehydrogenase gene (*LrGAPDH*, GenBank No. KJ543468.1) was used as an internal reference gene to calculate the relative expression values of *LrCHI2*. The *LrCHI2* expression levels were calculated according to the 2⁻ΔΔCt method. The qRT-PCR analysis was completed using three biological replicates, and there were three technical repetitions included in each biological replicate.
**Subcellular Localization**
The subcellular localization of *LrWRKY2* and *LrCHI2* was predicted using the PSORT online program (https://www.genscript.com/psort.html) and then confirmed by transiently expressing green fluorescent protein (GFP)-tagged fusion proteins in onion (*Allium cepa*) epidermal cells. More specifically, the *LrWRKY2* and *LrCHI2* ORFs lacking the stop codon were amplified from pMD-18T-LrWRKY2 and pMD-18T-LrCHI2, respectively. The PCR products were then inserted into the pBIN m0-gfp5-ER vector to generate the *LrWRKY2*-GFP and *LrCHI2*-GFP constructs. The two recombinant vectors were transferred into *Agrobacterium tumefaciens* EHA105 cells using a CaCl₂ freeze-thaw method (Holsters et al., 1978). An empty pBIN m-gfp5-ER vector served as a control. The transformed *A. tumefaciens* cells were used to insert the recombinant or control vector into onion epidermal cells, which were then cultured in darkness for 48 h. To analyze transient gene expression, GFP fluorescence in the onion epidermal cells was monitored using a confocal microscope (Nikon, JPN) as described by Liu et al. (2018).
**Generation and Screening of Transgenic Tobacco Lines Overexpressing *LrWRKY2* or *LrCHI2**
The plasmids pMD-18T-LrWRKY2 and pCAMBIA2300s were digested with BamHI and XbaI, while the pMD-18T-LrCHI2 and pCAMBIA2300s plasmids were digested with BamHI and EcoRI for constructing the overexpression vectors of *LrWRKY2* and *LrCHI2*, respectively. After which the pCAMBIA2300s-LrWRKY2 and pCAMBIA2300s-LrCHI2 recombinant plasmids were generated using the T4 DNA ligase. The recombinant plasmids were inserted into separate competent *A. tumefaciens* LBA4404 cells. The positive clones identified by PCR were used to transform tobacco leaf disks as described by Horsch et al. (1985). Genomic DNA was extracted from T₀ transgenic tobacco plants using cetyltrimethylammonium bromide (CTAB) as described by Allen et al. (2006), and then the genomic DNA was used as the template for a PCR conducted to confirm the transgenic lines carried the correct transgene, with wild-type (WT) plants serving as the negative control. The transgenic tobacco plants were grown in a greenhouse to produce T₂ generation lines.
**Analysis of Gene Expression Levels and Evaluation of the Disease Resistance of T₂ Transgenic Tobacco Plants**
Total RNA was extracted from T₂ generation *LrWRKY2* transgenic tobacco lines and reverse transcribed into cDNA with the forementioned method. The *LrWRKY2* transcription levels in the transgenic tobacco lines were determined by qRT-PCR with the tobacco actin gene (*NtACT*, GenBank No. AB158612.1) as an internal reference gene. Moreover, the roots and leaves of several *LrWRKY2* transgenic tobacco lines and WT plants were inoculated with *F. oxysporum* to evaluate the disease resistance of the transgenic tobacco lines. After the wounded tobacco roots were immersed in an *F. oxysporum* spore suspension (5 × 10⁶ spores/mL) for 30 min, the inoculated tobacco plants were grown under hydroponic conditions in an illumination incubator for 1 week. Regarding the leaf inoculation, leaves were wounded, inoculated with 20 μL *F. oxysporum* spore suspension (5 × 10⁶ spores/mL), and then placed in a humid illumination incubator for 1 week. The leaves of T₂ generation *LrCHI2* transgenic tobacco lines were similarly inoculated. The disease symptoms caused by the *F. oxysporum* infections were examined.
The expression levels of the following defense-related genes in the *LrWRKY2* transgenic tobacco lines were analyzed.
by qRT-PCR with the forementioned method: jasmonic acid (JA) biosynthetic pathway-related genes (NtAOCS, NtAOS, NtKAT, NtPACX, NtJMT, NtOPR, and NtLOX), PR genes (NtCHI, NtGlU2, and NtPR1), and antioxidant stress-related superoxide dismutase (SOD) genes (NtSOD, NtCu-ZnSOD, and MnSOD). Sequence details regarding the defense-related genes were obtained from the NCBI database (https://www.ncbi.nlm.nih.gov/) and used to design gene-specific primers (Supplementary Table 2). Three LrWRKY2 transgenic tobacco lines were randomly selected for an analysis of the expression of these defense-related genes.
### Transient Expression of Hairpin RNA Targeting LrWRKY2 in L. regale
Primers with the attB linker were designed to amplify the LrWRKY2 RNA interference (RNAi) fragment (Supplementary Table 1). The PCR product was incorporated into the RNAi vector pHellsgate2 via a BP recombination reaction using the Gateway® BP Clonase™ II Enzyme Mix kit (Invitrogen, USA). The pHellsgate2-LrWRKY2 recombinant plasmid was inserted into competent E. coli DH10B cells. Positive clones were selected on agar-solidified Luria-Bertani medium containing spectinomycin (90 mg/L). The plasmids of the positive clones were digested with XbaI and XhoI to confirm the LrWRKY2 fragment was correctly recombined.
The pHellsgate2-LrWRKY2 recombinant plasmid and the empty pHellsgate2 vector were inserted into separate A. tumefaciens EHA105 cells, which were then cultured in MGL medium at 28°C for 5 h in a constant-temperature shaker (150 rpm). Fresh L. regale scales were washed with sterile water and rubbed with sandpaper to form uniformly sized wounds. The wounded L. regale scales were transformed with A. tumefaciens liquid containing the pHellsgate2-LrWRKY2 recombinant plasmid or the empty pHellsgate2 vector. The L. regale scales were placed on filter paper moistened with sterile water and incubated in a climate-controlled cabinet set at 25°C for 24 h, and then inoculated with 20 µL F. oxysporum spore suspension (5 × 10⁶ spores/mL). The inoculated scales were collected at 72 h for an analysis of LrWRKY2 expression and disease symptoms. The L. regale scales in which the empty pHellsgate2 vector was transiently expressed were used as the control.
### Expression and Purification of the Recombinant LrWRKY2 and LrCHI2 Proteins
The SignalP 5.0 Server (https://www.cbs.dtu.dk/services/SignalP/) was used for predicting the presence of a signal peptide. The LrWRKY2 ORF amplified by PCR from pMD-18T-LrWRKY2 was digested with HindIII and BamHI and then subcloned into the pET-32a vector. Gene-specific primers (Supplementary Table 1) were designed to amplify the LrCHI2 ORF without the signal peptide-encoding sequence. The PCR product was inserted into pMD-18T. The resulting recombinant plasmid was digested with EcoRI and EcoRV and then the LrCHI2 fragment was incorporated into the pET-32a vector. The pET-32a-LrWRKY2 and pET-32a-LrCHI2 plasmids were transferred into competent E. coli BL21 cells. The empty pET-32a vector was used as a control. The production of the LrWRKY2 and LrCHI2 recombinant proteins was induced and the inclusion body proteins were denatured and renatured as described by Zhao et al. (2020). The recombinant proteins were purified using the Ni-NTA Sepharose Column (Sangon Biotech, China).
### Analysis of the Antifungal Activity of the Recombinant LrCHI2 Protein
The in vitro antifungal effects of the LrCHI2 recombinant protein were examined using F. oxysporum, C. gloeosporioides, F. solani, and A. panax. Agar plugs (1 cm diameter) removed from the outer edge of actively growing fungus on agar-solidified PDA medium in plates were used to inoculate fresh agar-solidified PDA medium. When the diameter of the fungal colonies reached 2 cm, LrCHI2 (5, 10, and 20 µg) was added to the plates. Sterile water and phosphate buffer (pH 8.0) were used as controls. After a 3-day incubation at 28°C, the fungal colonies were photographed and the average growth inhibition zones (mm²) were calculated using Photoshop 7.0.
### Cloning of the LrCHI2 Promoter Fragment
The LrCHI2 promoter region was isolated from L. regale genomic DNA by genome walking using the Universal GenomeWalker™ 2.0 kit (TaKaRa, Japan). Two nested primers were designed specifically for the LrCHI2 unigene sequence (Supplementary Table 1). The sequence upstream of LrCHI2 was amplified by two rounds of PCR. The resulting PCR product was cloned into pMD-18T. The promoter fragment was confirmed by sequencing and an alignment with the LrCHI2 unigene sequence, and a 491-bp LrCHI2 promoter fragment was obtained. Additionally, the promoter cis-elements were predicted using the PlantCARE program (http://bioinformatics.psb.ugent.be/webtools/plantcare/html/).
### Electrophoretic Mobility Shift Assay (EMSA)
A 50-bp LrCHI2 promoter fragment with one W-box was used as the probe. The mutant probe had one mutation in the W-box core sequence (Supplementary Table 1). These two probe sequences were synthesized and labeled with biotin (Sangon Biotech, China). The unlabeled probe served as the competitor. The EMSA reaction mixtures contained 0.5 µg purified LrWRKY2 protein and 7 µL 1 x gel shift binding buffer [100 mM Tris, 500 mM KCl, 10 mM DTT, 50% glycerol, 100 mM MgCl₂, 1% NP-40, 1 M KCl, 200 mM EDTA, and 1 µg/µL poly (dI•dC)] in a total volume of 20 µL. After a 20-min incubation at room temperature, 2 µL biotin-labeled probes were added and the incubation was continued for 20 min. For the competitor probe, the binding reaction components were incubated for 40 min. The DNA–protein complexes were electrophoresed on 6.5% non-denaturing polyacrylamide gels in an ice-cold water bath and then transferred to a nylon membrane. Finally, the membrane was dried and analyzed using the LightShift Chemiluminescent EMSA Kit (Pierce, USA).
Yeast One-Hybrid (Y1H) Assay
A Y1H assay was performed to analyze the LrCHI2 promoter-binding capability of LrWRKY2. The recombinant prey (pGADT7-LrWRKY2) and bait (pLRCHI2-pAbAi) vectors were constructed. The reporter strains were generated by integrating linearized pLRCHI2-pAbAi or empty pAbAi plasmids into the genome of yeast strain Y1HGold. Positive clones identified by PCR were used to inoculate synthetic defined (SD) medium containing different aureobasidin A (AbA) concentrations to determine the appropriate AbA inhibitory concentration. The recombinant prey plasmid (pGADT7-LrWRKY2) was integrated into yeast cells containing pLRCHI2-pAbAi or empty pAbAi, whereas the recombinant prey plasmid (pGADT7-p53) was inserted into Y1HGold yeast cells carrying the p53-pAbAi plasmid (positive control). Positive clones were confirmed by PCR. The protein–DNA interaction was identified on the basis of the activation of the AbA resistance gene when a prey protein from the library bound to the bait sequence. Thus, the three types of Y1HGold yeast cells were used to inoculate SD/-Leu medium supplemented with the appropriate AbA inhibitory concentration. The Y1H assay was performed using the Matchmaker™ Gold Yeast One-Hybrid Library Screening System (Takara, USA).
Transformation of Tobacco With the LrCHI2 Promoter and LrWRKY2
The LrCHI2 promoter was cloned to study whether LrWRKY2 regulates LrCHI2 expression. The pMD-18T-pLRCHI2 plasmid was digested with BamHI and HindIII and then the LrCHI2 promoter fragment was subcloned into the pBI121-GUS vector digested with the same restriction enzymes. The pBI121-pLRCHI2-GUS plasmid was integrated into WT and LrWRKY2 overexpressing tobacco leaf disks via A. tumefaciens-mediated transformation. As controls, WT and LrWRKY2 transgenic tobacco leaf disks were transformed with the empty pBI121-GUS vector. The genomic DNA of the putative transgenic tobacco lines served as the template for a PCR amplification using β-glucuronidase (GUS) gene-specific primers to verify the transgenic plants were transformed correctly (Supplementary Table 1). The GUS enzyme activity (pM 4-MU min⁻¹ µg⁻¹ protein) of the confirmed transgenic tobacco lines was analyzed using a fluorescence spectrophotometer (Hitachi F-4600, Japan) as described by Chen et al. (2017).
Data Analysis
Relative gene expression levels, fungal growth inhibition zone sizes, tobacco leaf and lily scale lesion sizes, and GUS enzyme activities are herein presented as the mean ± standard deviation. The data were analyzed using the SPSS software (version 17.0) and the significance of any differences was determined according to Student’s t-test and Duncan’s multiple range test. All experiments were performed independently at least three times under the same conditions.
RESULTS
The L. regale Subgroup Ile WRKY2 Gene Encodes a Nuclear Protein
Researchers previously identified 35 WRKY genes by sequencing the wild L. regale transcriptome during an F. oxysporum infection; they also revealed that LrWRKY2 expression is induced by F. oxysporum and is substantially affected by MeJA, SA, ETH, and H₂O₂ treatments (Li et al., 2021a). Thus, LrWRKY2 was functionally characterized. The L. regale LrWRKY2 cDNA was amplified by RT-PCR using gene-specific primers. The full-length LrWRKY2 cDNA comprised 1,302 bp, including a 1,032-bp ORF encoding a Subgroup Ile WRKY protein consisting of 343 amino acid residues. A sequence analysis confirmed the putative LrWRKY2 protein contained a typical WRKY domain with the highly conserved WRKYGQK sequence and a C₂H₂-type (CX₄₋₅CX₂₋₃HXH) zinc-finger motif, which is a typical feature of Group II members (Figure 1A). The deduced LrWRKY2 amino acid sequence was highly similar to the sequences of several WRKY proteins from monocots, including Phoenix dactylifera WRKY14 (accession no. XP_008788702.2), Elaeis guineensis WRKY14 (accession no. XP_010920047.1), and Cocos nucifera WRKY14 (accession no. KAG1354178.1) (Figure 1B).
The PSORT program predicted that LrWRKY2 contains a putative NLS “KRPR” sequence, implying LrWRKY2 may be localized in the nucleus. To experimentally determine the subcellular localization of LrWRKY2, the LrWRKY2 coding sequence was fused to a GFP reporter gene, and the resulting fusion construct was inserted into onion epidermal cells. Green fluorescence was detected throughout the onion cells infected with A. tumefaciens carrying the empty pBIN m-gfp5-ER vector, whereas it was limited to the nucleus in cells infected with A. tumefaciens carrying the LrWRKY2-GFP construct. Moreover, the LrWRKY2-GFP fusion protein colocalized with the nuclear localization marker PI (Figure 1C). These observations indicate LrWRKY2 is a nuclear protein.
The Overexpression of LrWRKY2 in Tobacco Increases the Resistance to F. oxysporum and Activates Some JA Biosynthesis-Related Genes, PR Genes, and SOD Genes
The integration of LrWRKY2 into transgenic tobacco plants was verified by PCR. There were no differences in the phenotype and growth status of the WT and LrWRKY2 transgenic tobacco plants. Twelve confirmed transgenic lines were selected for the subsequent investigation. The analysis of gene expression by qRT-PCR indicated that LrWRKY2 was stably expressed in the transgenic lines (Figure 2A). Among the transgenic lines, LrWRKY2 expression levels were highest in W2-32, W2-38, W2-39, and W2-44. Thus, these four lines were selected for further analyses.
The roots of WT and LrWRKY2 transgenic tobacco plants were inoculated with an F. oxysporum spore suspension (5 × 10⁶ spores/mL). After a 7-day incubation, the inoculated
WT plants had curled, yellow, and withered leaves as well as blackened or rotted roots. In contrast, the LrWRKY2 transgenic tobacco leaves and roots were healthy and growing well (Figure 2B). The leaf inoculation assay produced similar results. More specifically, obvious disease symptoms were detected surrounding the inoculation site of the WT leaves (Figure 2C). The WT leaves were yellow and rotted, which was in sharp contrast to the healthy LrWRKY2 transgenic leaves. The average leaf lesion size for the transgenic lines and WT control was $< 100 \text{ mm}^2$ and $\sim 580 \text{ mm}^2$, respectively (Figure 2D). These results imply that LrWRKY2 overexpression can substantially increase the resistance of transgenic tobacco to F. oxysporum.
Several defense response-related genes were selected to examine the transcriptional changes in the LrWRKY2 transgenic tobacco lines. The JA biosynthetic pathway-related genes (NtAOC, NtAOS, NtKAT, NtPACX, NtJMT, NtOPR, and NtLOX), PR genes (NtCHI, NtGlu2, and NtPR-1), and SOD genes (NtSOD,
Li et al.
**FIGURE 2** | Gene expression and resistance analyses of T<sub>2</sub> generation *LrWRKY2* transgenic tobacco lines. (A) The *LrWRKY2* was stably expressed in the 12 T<sub>2</sub> generation *LrWRKY2* transgenic tobacco lines (W2-1/6/8/13/24/25/29/32/33/38/39/44). (B) The root inoculation assay revealed the enhanced resistance of four T<sub>2</sub> *LrWRKY2* transgenic tobacco lines (W2-32/38/39/44) to *F. oxysporum* infection. (C) The leaf inoculation assay revealed the increased resistance of four T<sub>2</sub> *LrWRKY2* transgenic tobacco lines (W2-32/38/39/44) to *F. oxysporum* infection. (D) The lesion areas in the inoculated leaves of the four T<sub>2</sub> *LrWRKY2* transgenic tobacco lines were significantly smaller than that in the WT (**p < 0.01). Bars represent the standard errors of three biological replicates, and the statistical analysis was performed with the t test (**p < 0.01).
*NtCu-ZnSOD*, and *MnSOD*) showed higher expression levels in the *LrWRKY2* transgenic lines than in the WT plants (Figure 3). The JA biosynthetic pathway-related genes *NtAOC*, *NtKAT*, *NtJMT*, *NtOPR*, and *NtLOX* as well as the PR gene *NtGlu2* were most highly expressed in line W2-1, with the *NtGlu2* expression level 15-fold higher than that in the WT plants. The JA biosynthetic pathway-related genes *NtAOS* and *NtPACX*, the PR genes *NtCHI* and *NtPR-1*, and the SOD genes *NtSOD*, *NtCu-ZnSOD*, and *MnSOD* displayed the highest expression levels in line W2-33, with the *NtPR-1* expression level 17-fold higher than that in the WT plants. The expression levels of all analyzed genes were higher in line W2-8 than in the WT plants. These observations suggest that *LrWRKY2* overexpression in tobacco up-regulates the expression of JA signaling pathway genes and induces the expression of some PR and SOD genes.
The Transient Expression of the Hairpin RNA Targeting *LrWRKY2* in *L. regale* Leads to Increased Susceptibility to *F. oxysporum*
To clarify the effect of silencing *LrWRKY2* expression, the *LrWRKY2*-RNAi construct was transiently expressed in *L. regale* scales, which were then inoculated with *F. oxysporum*. After a 3-day incubation, the scales expressing the *LrWRKY2*-RNAi fragment had dark brown wounds and were rotted, whereas the scales transformed with the empty RNAi vector had light brown wounds and were only slightly rotted (Figure 4A). Moreover, the average lesion size was greater for the scales expressing the *LrWRKY2*-RNAi fragment (∼100 mm<sup>2</sup>) than for the scales transformed with the empty RNAi vector (∼28 mm<sup>2</sup>) (Figure 4B). At 3 days after the inoculation with *F. oxysporum*, the *LrWRKY2* expression level was significantly lower in the *L. regale* scales expressing the *LrWRKY2*-RNAi fragment than in
FIGURE 3 | The expression levels of some JA biosynthesis and defense-related genes including the NtAOC, NtAOS, NtKAT, NtPACX, NtJMT, NtOPR, NtLOX, NtCHI, NtGlu2, NtPR-1, NtSOD, NtCu-ZnSOD, and MnSOD in the three T2 LrWRKY2 transgenic tobacco lines (W2-1/8/33) were evaluated by qRT-PCR. Bars represent the standard errors of three biological replicates. The results were calculated by the $2^{-\Delta\Delta Ct}$ method and analyzed by the t-test (*p < 0.05, **p < 0.01).
transcriptome sequencing data (unpublished). Gene expression of the LrCHI2 gene was measured using transcript per Million mapped reads (FPKM). The FPKM of LrCHI2 for the 96 h uninoculated roots was 1015.59. Therefore, the log2FC (log2 fold changes) was 6.97, indicating a significant upregulation of LrCHI2 gene in L. regale roots during F. oxysporum infection. The obtained full-length LrCHI2 cDNA (accession no. MZ272344) consisted of 1,313-bp, with a 933-bp ORF, a 19-bp 5’ untranslated region, and a 361-bp 3’ untranslated region. The encoded protein comprising 310 amino acid residues was ~32.46 kDa and had an isoelectric point of about 5.65. A sequence analysis indicated the deduced LrCHI2 protein possesses a chitin-binding domain (ChtBD) and a Glyco_hydro_19 (GH19) catalytic domain. Additionally, a 491-bp LrCHI2 promoter fragment (accession no. MZ272345) was obtained by TAIL-PCR. A number of cis-elements were identified in the LrCHI2 promoter region, including ABRELATERD1 (abscisic acid-responsive element), GT1CONSensus (gibberellin acid-responsive element), and the W-box (Supplementary Table 3).
The qRT-PCR analysis indicated LrCHI2 was expressed at relatively low levels in the stems, leaves, flowers, and scales (Figure 5A), but it was highly expressed in the roots. Moreover, the F. oxysporum infection significantly up-regulated LrCHI2 expression in the L. regale roots (Figure 5B). Specifically, its expression level was about 3-fold and 5.5-fold higher than that in the control (not inoculated) at 24 and 48 h post-inoculation, respectively. Accordingly, LrCHI2 appears to be predominantly expressed in the roots and its expression is induced by F. oxysporum.
An N-terminal signal peptide was detected in LrCHI2, indicating this protein is secreted. The subcellular localization of LrCHI2 was determined by expressing a GFP-tagged fusion protein in onion epidermal cells. The fluorescence of LrCHI2-GFP was exclusively detected in the cell wall, suggesting LrCHI2 is a cell wall protein (Figure 5C).
Four T2 transgenic tobacco lines overexpressing LrCHI2 (C6, C11, C16, and C22) were randomly selected and examined regarding their resistance to F. oxysporum. The WT and LrCHI2 transgenic tobacco leaves were inoculated with an F. oxysporum spore suspension (5 × 106 spores/mL). After a 7-day incubation, transgenic tobacco leaves infected with F. oxysporum had relatively small wounds with slightly yellow edges, whereas the infected WT leaves had large wounds and were more obviously decayed (Figure 5D). Hence, LrCHI2 overexpression considerably enhanced the resistance of the transgenic tobacco lines to F. oxysporum. The lesions on the leaves of the four transgenic tobacco lines (C6, C11, C16, and C22) were smaller than 40 mm2, whereas the WT leaf lesions were almost 120 mm2 (Figure 5E).
The LrCHI2 recombinant protein lacking a signal peptide was expressed in E. coli cells. The SDS-PAGE analysis indicated that the His-tagged LrCHI2 protein produced in cells induced by isopropyl-β-D-1-thiogalactopyranoside (IPTG) was ~47 kDa, which was consistent with the predicted size (Figure 6A). The LrCHI2 recombinant protein in inclusion bodies was solubilized and then purified by nickel-affinity chromatography, with most of the protein eluted by the 100 mM imidazole elution buffer. The lack of extra bands during the SDS-PAGE analysis reflected the purity of the obtained LrCHI2 recombinant protein (Figure 6B). The in vitro antifungal activity of the purified LrCHI2 recombinant protein was examined. Compared with the blank controls, the LrCHI2 recombinant protein significantly inhibited the growth of F. oxysporum, C. gloeosporioides, F. solani, and A. panax (Figures 6C–F). Moreover, the inhibitory
FIGURE 4 | Analysis of the hairpin RNA targeting LrWRKY2 transiently expressed in L. regale scales. (A) The symptoms of L. regale scales after F. oxysporum inoculation, in which the LrWRKY2 RNAi vector and the empty RNAi vector were expressed, respectively. (B) The lesion areas in LrWRKY2-RNAi fragment expressed scales were significantly bigger than that in the empty RNAi vector expressed scales (**p < 0.01). (C) The expression levels of LrWRKY2 in LrWRKY2-RNAi fragment expressed scales and the empty RNAi vector expressed scales were evaluated by qRT-PCR. Bars represent the standard errors of three biological replicates. The results was calculated by the 2−ΔΔCt method and analyzed by the t-test (*p < 0.05, **p < 0.01).
The L. regale Chitinase Gene CHI2 Confers Resistance to F. oxysporum
To explore the regulatory effects of the LrWRKY2 TF on PR gene expression, an L. regale chitinase gene (LrCHI2) responsive to F. oxysporum was cloned on the basis of the L. regale transcriptome sequencing data (unpublished). Gene expression level of LrCHI2 was estimated as Fragments Per Kilobase of transcript per Million mapped reads (FPKM). The FPKM of LrCHI2 in L. regale roots without inoculation with F. oxysporum was 5.29, while its FPKM in L. regale roots inoculated with F. oxysporum for 96 h was 1015.59. Therefore, the log2FC (log2 fold changes) was 6.97, indicating a significant upregulation of LrCHI2 gene in L. regale roots during F. oxysporum infection. The obtained full-length LrCHI2 cDNA (accession no. MZ272344) consisted of 1,313-bp, with a 933-bp ORF, a 19-bp 5’ untranslated region, and a 361-bp 3’ untranslated region. The encoded protein comprising 310 amino acid residues was ~32.46 kDa and had an isoelectric point of about 5.65. A sequence analysis indicated the deduced LrCHI2 protein possesses a chitin-binding domain (ChtBD) and a Glyco_hydro_19 (GH19) catalytic domain. Additionally, a 491-bp LrCHI2 promoter fragment (accession no. MZ272345) was obtained by TAIL-PCR. A number of cis-elements were identified in the LrCHI2 promoter region, including ABRELATERD1 (abscisic acid-responsive element), GT1CONSensus (gibberellin acid-responsive element), and the W-box (Supplementary Table 3).
The qRT-PCR analysis indicated LrCHI2 was expressed at relatively low levels in the stems, leaves, flowers, and scales (Figure 5A), but it was highly expressed in the roots. Moreover, the F. oxysporum infection significantly up-regulated LrCHI2 expression in the L. regale roots (Figure 5B). Specifically, its expression level was about 3-fold and 5.5-fold higher than that in the control (not inoculated) at 24 and 48 h post-inoculation, respectively. Accordingly, LrCHI2 appears to be predominantly expressed in the roots and its expression is induced by F. oxysporum.
An N-terminal signal peptide was detected in LrCHI2, indicating this protein is secreted. The subcellular localization of LrCHI2 was determined by expressing a GFP-tagged fusion protein in onion epidermal cells. The fluorescence of LrCHI2-GFP was exclusively detected in the cell wall, suggesting LrCHI2 is a cell wall protein (Figure 5C).
Four T2 transgenic tobacco lines overexpressing LrCHI2 (C6, C11, C16, and C22) were randomly selected and examined regarding their resistance to F. oxysporum. The WT and LrCHI2 transgenic tobacco leaves were inoculated with an F. oxysporum spore suspension (5 × 106 spores/mL). After a 7-day incubation, transgenic tobacco leaves infected with F. oxysporum had relatively small wounds with slightly yellow edges, whereas the infected WT leaves had large wounds and were more obviously decayed (Figure 5D). Hence, LrCHI2 overexpression considerably enhanced the resistance of the transgenic tobacco lines to F. oxysporum. The lesions on the leaves of the four transgenic tobacco lines (C6, C11, C16, and C22) were smaller than 40 mm2, whereas the WT leaf lesions were almost 120 mm2 (Figure 5E).
The LrCHI2 recombinant protein lacking a signal peptide was expressed in E. coli cells. The SDS-PAGE analysis indicated that the His-tagged LrCHI2 protein produced in cells induced by isopropyl-β-D-1-thiogalactopyranoside (IPTG) was ~47 kDa, which was consistent with the predicted size (Figure 6A). The LrCHI2 recombinant protein in inclusion bodies was solubilized and then purified by nickel-affinity chromatography, with most of the protein eluted by the 100 mM imidazole elution buffer. The lack of extra bands during the SDS-PAGE analysis reflected the purity of the obtained LrCHI2 recombinant protein (Figure 6B). The in vitro antifungal activity of the purified LrCHI2 recombinant protein was examined. Compared with the blank controls, the LrCHI2 recombinant protein significantly inhibited the growth of F. oxysporum, C. gloeosporioides, F. solani, and A. panax (Figures 6C–F). Moreover, the inhibitory
FIGURE 5 | The functional analysis of LrCHI2. (A) The expression levels of LrCHI2 in L. regale roots after inoculation with F. oxysporum were analyzed by qRT-PCR. The L. regale roots were inoculated with F. oxysporum, and then were collected at 4, 24, 48, and 72 hpi. The roots inoculated with sterile water were used as control sample. (B) The expression levels of LrCHI2 in various tissues of L. regale under normal conditions were analyzed by qRT-PCR. (C) Subcellular localization analysis of LrCHI2-GFP fusion protein. The LrCHI2-GFP fusion protein was transiently expressed in cell wall after genetic transform mediated by A. tumefaciens. GFP, fluorescent light; Bright, white light; Merged, overlaid of fluorescent and white light. (D) The leaf inoculation assay revealed the four T2 LrWRKY2 transgenic tobacco lines (C6, C11, C16, and C22) showed enhanced resistance to F. oxysporum infection. (E) The lesion areas in the inoculated leaves of the four T2 generation LrCHI2 transgenic tobacco lines were significantly smaller than that in the WT (**p < 0.01). Bars represent the standard errors of three biological replicates. The statistical analysis was performed with the t test (**p < 0.01).
**LrWRKY2 Binds to the LrCHI2 Promoter Fragment Containing the W-box and Activates Transcription**
The WRKY TFs often target the W-box cis-element to activate or suppress target gene expression. The EMSA results confirmed the LrWRKY2 recombinant protein purified from E. coli cells was able to bind directly to the W-box sequence in the LrCHI2 promoter. The biotin-labeled probes alone revealed a lack of band shifts in the gel (lane 1, Figure 7A). However, LrWRKY2 was able to bind to the biotin-labeled probes at the LrCHI2 promoter fragment, resulting in a mobility shift (lane 2, Figure 7A). The inclusion of both biotin-labeled and unlabeled probes resulted in the detection of a band shift in the gel (lane 3, Figure 7A), reflecting the binding of the two probes to LrWRKY2. Because the unlabeled probes were 50-fold more abundant than the biotin-labeled probes, LrWRKY2 mainly bound to the competitor (unlabeled probe). Thus, the band shift was less intense in lane 3 than in lane 2 (Figure 7A). A band shift was undetectable in lane 4 (Figure 7A), which corresponded to the analysis involving the mutant probe and LrWRKY2, implying LrWRKY2 binds specifically to the W-box in the LrCHI2 promoter.
The ability of LrWRKY2 to activate transcription was evaluated in a Y1H assay, in which Y1HGold yeast cells were transformed with pGADT7-LrWRKY2 and pLrCHI2-pAbAi. The assay results indicated that the yeast cells co-transformed with
FIGURE 6 | Expression, purification, and antifungal assay of LrCHI2 recombinant protein. (A) The induced expression of LrCHI2 recombinant protein under 1 mM IPTG condition. M, protein marker; Line 1, the protein of E. coli with empty vector pET-32a was detected after induction; Line 2, the expression of LrCHI2 protein was detected without induction; Line 3-7, the expression of LrCHI2 protein was detected at 6, 8, 10, 12, and 24 h after induction, respectively. (B) The purification of LrCHI2 recombinant protein. M, protein marker; Line 8, the supernatant after treatment with the inclusion body solubilizing solution; Line 9-11, the purified LrCHI2 recombinant protein with 50, 100, 150 mM imidazole washing buffer, respectively. (C–F) The LrCHI2 recombinant protein has evident antifungal activity to F. oxysporum (C), C. gloeosporioides (D), F. solani (E) and A. panax (F). (G) The fungal growth inhibition areas (mm²). Bars represent the standard errors of three biological replicates. The statistical analysis was performed with the t test (**p < 0.01).
pGADT7-LrWRKY2 and pLrCHI2-pAbAi grew well, whereas the yeast cells transformed with pGADT7-LrWRKY2 and the empty pAbAi vector failed to survive on the SD/–Leu selective medium containing 50 ng/mL AbA (Figure 7B). These observations indicate LrWRKY2 can bind to the LrCHI2 promoter in vivo and activate transcription in yeast cells.
The transcriptional activation of LrCHI2 by LrWRKY2 was verified by inserting an expression construct comprising a GUS-encoding gene under the control of the LrCHI2 promoter into LrWRKY2 transgenic tobacco. A total of 23 transgenic tobacco plants carrying both LrWRKY2 and pLrCHI2-GUS and 27 transgenic tobacco plants transformed with pLrCHI2-GUS alone were identified by PCR. The GUS activity in the pLrCHI2-GUS transgenic tobacco plants was significantly lower than that in the transgenic tobacco carrying both LrWRKY2 and pLrCHI2-GUS (Figure 7C). The activity of the GUS expressed under the control of the LrCHI2 promoter increased by approximately 1.4-fold in the presence of LrWRKY2. There were no significant differences in the GUS activities of the two kinds of transgenic tobacco plants transformed with the empty pH121-GUS vector. This implies that LrWRKY2 does not affect the CaMV 35S promoter. Therefore, LrWRKY2 positively regulates LrCHI2 expression.
**DISCUSSION**
The WRKY proteins are critical plant TFs mediating diverse biological processes. However, little is known about the functions of _L. regale_ WRKY genes. Protein functions are closely related to the distribution of the proteins in cells. The localization of LrWRKY2 in the nucleus is in accordance with its putative function as a TF that must exist in the nucleus to regulate the transcription of target genes (Liu et al., 2013; Wang et al., 2013; Sun et al., 2015). Similarly, maize (_Zea mays_) WRKY106 (Wang C. T. et al., 2018) and cotton (_Gossypium hirsutum_) WRKY33 are also nuclear proteins (Wang et al., 2019). In fact, the vast majority of the WRKY TFs identified to date are localized in the nucleus. Accordingly, LrWRKY2 is a nuclear protein that regulates cellular processes. In contrast, LrCHI2 was localized to the cell wall in this study. The plant cell wall is the first line of defense against a pathogen invasion. It is possible that LrCHI2 participates in chitinase–pathogen interactions. Some _Cucumis sativus_ chitinases are also localized in the cell wall, wherein they interact directly or indirectly with pathogen elicitors to trigger downstream defense pathways (Bartholomew et al., 2019).
The LrWRKY2 TF belongs to Subgroup Ile of the WRKY family (Li et al., 2021a). Previous research proved that many Subgroup Ile WRKY TFs positively regulate plant responses to biotic and abiotic stresses. For example, the overexpression of the Subgroup Ile _CmWRKY10_ chrysanthemum (_Chrysanthemum morifolium_) gene reportedly leads to increased drought tolerance (Jaffar et al., 2016). Transgenic tobacco plants overexpressing the _C. annuum_ WRKY27 gene enhances the resistance to _Ralstonia solanacearum_ (i.e., milder disease symptoms and inhibited pathogen growth) (Dang et al., 2014). Herbaceous peony (_Paeonia lactiflora_) plants in which the Subgroup Ile gene _WRKY65_ is silenced exhibit increased susceptibility to _Alternaria tenuissima_, suggesting PIWRKY65 is a positive regulator of peony defense responses to _A. tenuissima_ (Wang et al., 2020). It is well-known that the lily is difficult to perform the genetic transform, and only a few lily transformations have been achieved thus far (Yan et al., 2019b). The currently established lily transformation system still has some problems, such as the strong genotype dependence, low efficiency of stable transformation, poor genetic stability, and difficult regeneration after transformation (Yan et al., 2019b; Song et al., 2020). It is not so bad that the _Agrobacterium_ -mediated transformation of tobacco using leaf disks has provided a valuable tool for rapid evaluation of function.
of the transgenes in higher plants (Clemente, 2006). Therefore, the stable genetic transformation of LrWRKY2 was completed in the model plant tobacco in order to further understand the function of LrWRKY2 in the present study. The LrWRKY2-overexpressing transgenic tobacco plants were more resistant to F. oxysporum than the WT control plants. Additionally, L. regale scales transiently expressing LrWRKY2-RNAi were more susceptible to F. oxysporum than the control scales. Our results suggest that LrWRKY2 is an important TF influencing L. regale disease resistance.
The ectopic expression of TF genes facilitated by transgenic technology and the subsequent analysis of the transcriptional changes in the generated transgenic plants may provide new insights regarding transcriptional regulatory networks. As crucial components of plant defense systems, WRKY proteins regulate the expression of some defense-related genes. The overexpression of rice Subgroup IIA WRKY genes (OsWRKY62, OsWRKY28, OsWRKY71, and OsWRKY76) up-regulates the PR10 expression level (Peng et al., 2010). The constitutive overexpression of PtrWRKY18 and PtrWRKY35 in poplar (Populus trichocarpa) enhances the resistance to the biotrophic pathogen Melampsora by inducing PR gene expression (Jiang et al., 2017). The ectopic overexpression of the cotton (G. hirsutum) WRKY44 gene increases the resistance of Nicotiana benthamiana plants to fungal infections and activates the expression of PR genes, including PR1a, PR4, PR5, and NPR1 (Li et al., 2015). The ectopic expression of rice OsWRKY11 results in the constitutive expression of defense-associated genes (CHI2, PR10, and Betv1), whereas down-regulating OsWRKY11 expression adversely affects the expression of defense-related genes during pathogen invasions, suggesting that OsWRKY11 activates defense responses (Lee et al., 2018). The heterologous expression of rice OsWRKY6 in A. thaliana can dramatically induce the expression of defense-related genes, including PR1, PDF1, NPR4, and a glucanase gene (Hwang et al., 2011). In the present study, the expression levels of several PR genes (e.g., NtCHI, NtGlu2, and NtPR-1) were clearly up-regulated in LrWRKY2 transgenic tobacco lines. Additionally, the expression levels of JA biosynthetic pathway-related genes (NiAOC, NiAOS, NiKAT, NiPACX, NiJMT, NiOPR, and NiLOX) were higher in the LrWRKY2-overexpressing transgenic tobacco lines than in the WT tobacco plants, implying that the JA signaling pathway was activated in response to LrWRKY2 overexpression. Consistent with our findings, a previous study revealed that in transgenic A. thaliana lines overexpressing an L. regale WRKY gene, the expression levels of some JA-responsive genes, including AtLOX, AtMYC2, and AtPDF1.2, are up-regulated (Cui et al., 2018). Moreover, overexpressing PtrWRKY40 in A. thaliana activates the expression of JA-related defense genes, ultimately leading to the resistance to the necrotrophic fungous Botrytis cinerea (Karim et al., 2015). Furthermore, the expression levels of SOD genes (NiSOD, NiCu-ZnSOD, and MnSOD) increased significantly in the LrWRKY2-overexpressing tobacco lines, suggesting that LrWRKY2 confers enhanced disease resistance by activating the JA signaling pathway and inducing the expression of defense-related genes.
The PR proteins have important functions related to plant defense responses to pathogens. They accumulate after pathogen invasions, and may act as antimicrobial agents mediating cell wall hydrolysis, contact toxicity, and perhaps defense signaling (Zhang et al., 2017). A recent study demonstrated that Panax notoginseng PR-like proteins can inhibit F. solani and C. gloeosporioides mycelial growth (Li et al., 2021b). Another study proved that overexpressing the moss (Physcomitrella patens) PR10 gene in A. thaliana enhances plant resistance to Pythium irregularare (Castro et al., 2016). In the current study, LrCHI2 isolated from L. regale was characterized as a new PR gene. This gene was mainly expressed in the roots, and its expression was induced by an F. oxysporum infection. Additionally, LrCHI2 exhibited in vitro antifungal activity that inhibited F. oxysporum, C. gloeosporioides, F. solani, and A. panax mycelial growth. The overexpression of LrCHI2 in tobacco enhanced the resistance of the transgenic plants to F. oxysporum. Thus, LrCHI2 expression contributes to the resistance of L. regale to F. oxysporum.
Previous research proved that various PR genes are targeted by WRKY TFs, which bind to the canonical W-box sequence (TTGACC/T) in the gene promoters. A recent Y1H assay confirmed that rice OsWRKY6 can bind directly to the W-box derived from the rice PR1 promoter in yeasts (Hwang et al., 2011). An in vivo chromatin immunoprecipitation assay and in vitro EMSA experiments revealed that C. annuum WRKY40 binds directly to a C. annuum defensin gene promoter and the WRKY33 promoter, both of which contain a W-box; on the contrary, with mutations to the W-box sequence preventing the binding (Chakraborty et al., 2019). The LrWRKY1 TF induces the transcription of the LrPR10-5, which has a promoter containing three W-boxes (Li et al., 2021a). The co-expression of LrWRKY1 and LrPR10-5 in tobacco indicated LrWRKY1 can activate the LrPR10-5 promoter. In this study, the regulatory effect of LrWRKY2 on the LrCHI2 promoter was examined. This promoter contains hormone-responsive, biotic and abiotic stress-related elements as well as one W-box. The EMSA results proved that LrWRKY2 has a high in vitro affinity for the LrCHI2 promoter containing a W-box. Additionally, the Y1H assay results verified the interaction between LrWRKY2 and the LrCHI2 promoter fragment in yeast. These findings suggest that LrWRKY2 binds to the W-box in the LrCHI2 promoter to activate expression. This likely contributes to the LrWRKY2-regulated L. regale defense response to F. oxysporum. Because the transcript levels of several PR genes, including NtCHI, NtGlu2, and NtPR-1, increased in LrWRKY2 transgenic tobacco lines, future studies should investigate whether LrWRKY2 regulates the expression of other PR genes in L. regale.
CONCLUSIONS
The nuclear protein LrWRKY2 belongs to Subgroup Ile of the WRKY family. Transgenic tobacco lines overexpressing LrWRKY2 are highly resistant to F. oxysporum, which may be related to the significantly up-regulated expression of many defense-related genes, including JA biosynthetic pathway-related genes, PR genes, and SOD genes. Additionally, the transient expression of the LrWRKY2-RNAi fragment in L. regale scales leads to increased sensitivity to F. oxysporum. Moreover, LrWRKY2 is a positive regulator that mediates L. regale defense responses to F. oxysporum infection by regulating the chitinase
gen expression, which is a broad-spectrum resistance gene showing antifungal activity to some important phytopathogens including F. oxysporum. The results of this study may be relevant for future research aimed at elucidating the transcriptional regulatory mechanisms underlying the interaction between L. regale and F. oxysporum. In addition, the lily genetic transformation mediated by the Agrobacterium has been carried out in our research group. It is believed that a more in-depth understanding of the defense response regulation mechanism in L. regale against the Fusarium wilt will be obtained in the near future.
DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material.
AUTHOR CONTRIBUTIONS
SL: methodology, data curation, experiment execution, and writing—original draft preparation. JH: methodology, material collection, and supervision. ZW: investigation and software.
REFERENCES
Allen, G. C., Flores-Vergara, M. A., Krasnynski, S., Kumar, S., and Thompson, W. F. (2006). A modified protocol for rapid DNA isolation from plant tissues using cetlytrimethylammonium bromide. Nat. Protoc. 1, 2320–2325. doi: 10.1038/nprot.2006.384
Bartholomew, E. S., Black, K., Feng, Z., Liu, W., Shan, N., Zhang, X., et al. (2019). Comprehensive analysis of the chitinase gene family in cucumber (Cucumis sativus L.) from gene identification and evolution to expression in response to Fusarium oxysporum. Int. J. Mol. Sci. 20:5309. doi: 10.3390/ijms20215309
Castro, A., Vidal, S., and Ponce de León, I. (2016). Moss pathogenesis-related-10 protein enhances resistance to Pythium irregulare in Physcomitrella patens and Arabidopsis thaliana. Front. Plant Sci. 7:580. doi: 10.3389/fpls.2016.00580
Chakraborty, J., Ghosh, P., Sen, S., Nandi, A., and Das, S. (2019). CaMPK9 increases the stability of CaWRKY40 transcription factor which triggers defense response in chickpea upon Fusarium oxysporum f. sp. ciceri Race1 infection. Plant Mol. BioI. 100, 411–431. doi: 10.1016/j.sciuto.2019.08.880-0
Chen, R., He, H., Yang, Y., Qu, Y., Ge, F., and Liu, D. (2017). Functional characterization of a pathogenesis-related protein family 10 gene, LvPR16-5, from Lilium regale Wilson. Australas. Plant Pathol. 46, 251–259. doi: 10.1007/s13313-017-0485-0
Choi, C., Hwang, S. H., Fang, I., Kwon, S. I., Park, S., Ahn, I., et al. (2015). Molecular characterization of Orzya sativa WRKY6, which binds to W-box-like element 1 of the Orzya sativa pathogenesis-related (PR) 10a promoter and confers reduced susceptibility to pathogens. N. Phytol. 208, 846–859. doi: 10.1111/nph.13516
Clemente, T. (2006). Nicotiana (Nicotiana tabacum, Nicotiana benthamiana). Methods Mol. Biol. 343, 143–154. doi: 10.1385/1-59745-130-4:143
Cui, Q., Yan, X., Gao, X., Zhang, D., He, H., and Jia, G. (2018). Analysis of WRKY transcription factors and characterization of two Botryis cinerea -responsive LrWRKY genes from Lilium regale. Plant Physiol. Biochem. 127, 525–536. doi: 10.1016/j.plaphy.2018.04.027
Dang, F., Wang, Y., She, J., Lei, Y., Liu, Z., Eulgem, T., et al. (2014). Overexpression of CaWRKY27, a subgroup Ile WRKY transcription factor of Capsicum annuum, positively regulates tobacco resistance to Ralstonia solanacearum infection. Physiol. Plant. 150, 397–411. doi: 10.1111/ppl.12093
Ding, Y., Chen, Y., Lin, Z., Tuo, Y., Li, H., and Wang, Y. (2021). Differences in soil microbial community composition between suppressive and root rot-conducive in tobacco fields. Curr. Microbiol. 78, 624–633. doi: 10.1007/s00284-020-03218-3
Dong, Q., Zheng, W., Duan, D., Huang, D., Wang, Q., Liu, C., et al. (2020). MdWRKY30, a group IIA WRKY gene from apple, confers tolerance to salinity and osmotic stresses in transgenic apple callus and Arabidopsis seedlings. Plant Sci. 299:110611. doi: 10.1016/j.plantsci.2020.110611
Eulgem, T., and Somssich, I. (2007). Networks of WRKY transcription factors in defense signaling. Curr. Opin. Plant Biol. 10, 366–371. doi: 10.1016/j.pbi.2007.04.020
He, H., Liu, D., Zhang, N., Zheng, W., Han, Q., Ji, B., et al. (2014). The PR10 gene family is highly expressed in Lilium regale Wilson during Fusarium oxysporum f. sp. lili infected. Genes Genom. 36, 497–507. doi: 10.1007/s12358-014-0185-x
He, X., Li, W., Zhang, W., Jin, X., Shenkute, A. G., Aynalem, T., et al. (2019). Transcriptome sequencing analysis provides insights into the response to Fusarium oxysporum in Lilium pumilum. Evol. Bioinform. Online. 15:1176934319838818. doi: 10.1177/1176934319838818
Holsters, M., Waele, D., Depicker, A., Messens, E., Van Montagu, M., and Schell, J. (1978). Transfection and transformation of Agrobacterium tumefaciens. Mol. Gen. Genet. 163, 181–187. doi: 10.1007/BF00267408
Horsch, R. B., Fry, J. E., Hoffmann, N. L., Eichholtz, D., Rogers, S. G., and Fraley, R. T. (1985). A simple and general method for transferring genes into plants. Science. 227, 1229–1231. doi: 10.1126/science.227.4691.1229
Huang, X., Li, K., Xu, X., Yao, Z., Jin, C., and Zhang, S. (2015). Genome-wide analysis of WRKY transcription factors in white pear (Pyrus bretschneideri) reveals evolution and patterns under drought stress. BMC Genomics 16:1104. doi: 10.1186/s12864-015-2233-6
Hwang, S., Yie, S. W., and Hwang, D. (2011). Heterologous expression of OsWRKY6 gene in Arabidopsis activates the expression of defense related genes and enhances resistance to pathogens. Plant Sci. 181, 316–323. doi: 10.1016/j.plantsci.2011.06.007
Iftnan Khan, M., Zhang, Y., Liu, Z., Hu, J., Liu, C., Yang, S., et al. (2018). CaWRKY40b in pepper acts as a negative regulator in response to Ralstonia solanacearum by directly modulating defense genes including CaWRKY40. Int. J. Mol. Sci. 19:1403. doi: 10.3390/ijms19051403
SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpls.2021.741463/full#supplementary-material
Jaffar, M. A., Song, A., Faheem, M., Chen, S., Jiang, J., Liu, C., et al. (2016). Involvement of CmWRKY19 in drought tolerance of chrysanthemum through the ABA-signaling pathway. Int. J. Mol. Sci. 17:693. doi: 10.3390/ijms17050693
Jiang, J., Li, S., Pu, L., Li, X., Liu, D., and Cui, X. (2020). A β-1,3-glucanase gene from *Panax notoginseng* confers resistance in tobacco to *Fusarium solani*. Ind. Crops Prod. 143:111947. doi: 10.1016/j.indcrop.2019.111947
Viana, V., Bussanello, C., da Maia, L., Pegoraro, C., and Costa de Oliveira, A. (2018). Activation of rice WRKY transcription factors: an army of stress fighting soldiers? *Curr. Opin. Plant Biol.* 45, 268–275. doi: 10.1016/j.pbi.2018.07.007
Wang, C., Deng, P., Chen, L., Wang, X., Ma, H., Hu, W., et al. (2013). A wheat WRKY transcription factor TaWRKY10 confers tolerance to multiple abiotic stresses in transgenic tobacco. *PLoS ONE* 8:e65120. doi: 10.1371/journal.pone.0065120
Wang, C. T., Tu, J. N., Liu, Y. W., Li, M., Zhao, D., Yang, J. F., et al. (2018). Maize WRKY transcription factor ZmWRKY106 confers drought and heat tolerance in transgenic plants. *Int. J. Mol. Sci.* 19:3046. doi: 10.3390/ijms19103046
Wang, N. N., Xu, S. W., Sun, Y. L., Liu, D., Zhou, L., Li, Y., et al. (2019). The cotton WRKY transcription factor (GhWRKY33) reduces transgenic Arabidopsis resistance to drought stress. *Sci. Rep.* 9, 724–724. doi: 10.1038/s41598-018-37035-2
Wang, P., Xu, X., Tang, Z., Zhang, W., Huang, X. Y., and Zhao, F. J. (2018). *OsWRKY28* regulates phosphate and arsenate accumulation, root system architecture and fertility in rice. *Front. Plant Sci.* 9:1330. doi: 10.3389/fpls.2018.01330
Wang, X., Li, J., Guo, J., Qiao, Q., Guo, X., and Ma, Y. (2020). The WRKY transcription factor *PIWRKY65* enhances the resistance of *Paonia lactiflora* (herbaceous peony) to *Alternaria tenuissima*. *Hortic. Res.* 7, 57–57. doi: 10.1038/s41438-020-0267-7
Yamasaki, K., Kigawa, T., Inoue, M., Tateno, M., Yamasaki, T., Yabuki, T., et al. (2005). Solution structure of an Arabidopsis WRKY DNA binding domain. *Plant Cell* 17, 944–956. doi: 10.1010/tpc.104.026435
Yan, H., Li, M., Xiong, Y., Wu, J., Teixeira da Silva, J. A., and Ma, G. (2019a). Genome-wide characterization, expression profile analysis of WRKY family genes in *Santalum album* and functional identification of their role in abiotic stress. *Int. J. Mol. Sci.* 20:5676. doi: 10.3390/ijms20225676
Yan, R., Wang, Z., Ren, Y., Li, H., Liu, N., and Sun, H. (2019b). Establishment of efficient genetic transformation systems and application of CRISPR/Cas9 genome editing technology in *Lilium pumilum* DC. *Fisch. and Lilium longiflorum* White Heaven. *Int. J. Mol. Sci.* 20:2920. doi: 10.3390/ijms20122920
Zheng, Q., Li, Y., Zhang, Y., Wu, C., Wang, S., Hao, L., et al. (2017). Md-miR156ab and Md-miR395 target WRKY transcription factors to influence apple resistance to leaf spot disease. *Front. Plant Sci.* 8:526. doi: 10.3389/fpls.2017.00526
Zhao, Q., Qu, B., Li, S., Zhang, Y., Cui, X., and Liu, D. (2020). OsPRP1 regulates phosphate and arsenate accumulation, root system architecture and fertility in rice. *Front. Plant Sci.* 9:1330. doi: 10.3389/fpls.2018.01330
Publisher's Note: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
Copyright © 2021 Li, H., Wang, D., Liang, S., and Liu, D. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms. | 2025-03-05T00:00:00 | olmocr | {
"pdf_page_numbers": [
[
0,
2648,
1
],
[
2648,
9358,
2
],
[
9358,
15594,
3
],
[
15594,
21428,
4
],
[
21428,
27043,
5
],
[
27043,
28087,
6
],
[
28087,
30884,
7
],
[
30884,
31355,
8
],
[
31355,
39940,
9
],
[
39940,
42582,
10
],
[
42582,
43979,
11
],
[
43979,
47627,
12
],
[
47627,
54386,
13
],
[
54386,
60413,
14
],
[
60413,
64489,
15
]
]
} |
391650262426548aeb6fa942a62ea6f8112ad738 | {
"olmocr-version": "0.1.59",
"pdf-total-pages": 16,
"total-fallback-pages": 0,
"total-input-tokens": 40556,
"total-output-tokens": 11775,
"fieldofstudy": [
"Political Science"
]
} | A media visibility analysis of public leadership in Scandinavian responses to pandemics
Olivier Rubin\textsuperscript{a}, Erik Baekkeskov\textsuperscript{b} and PerOla Öberg\textsuperscript{c}
\textsuperscript{a}Department of Social Sciences and Business, Roskilde University, Roskilde, Denmark; \textsuperscript{b}School of Social and Political Sciences, University of Melbourne, Melbourne, VIC, Australia; \textsuperscript{c}Department of Government, Uppsala University, Uppsala, Sweden
\textbf{ABSTRACT}
This paper analyses public leadership in Scandinavia during the latest two pandemics, the swine flu pandemic in 2009 and the coronavirus pandemic in 2020, by compiling and contrasting national proxies of media visibility among pandemic response actors. Concretely, the paper taps into key media databases to develop indicators of how often national leaders and leading health experts are mentioned in Danish, Norwegian, and Swedish media reports about the 2009 and 2020 pandemics.
The study reveals a high degree of continuity of public leadership in Sweden during the two pandemics. In contrast, Norway and in particular Denmark both moved from a predominately expert-driven media presence in 2009 to a much stronger top-down ministerial leadership presence during the 2020 coronavirus pandemic. In addition, Sweden also displays the most balanced media representation of health experts and cabinet ministers during both pandemics. The paper concludes by discussing the pros and cons of the outlined differences in public leadership and the possible practical implications with regards public debate and trust.
\textbf{1. Introduction}
Otherwise very similar Scandinavian countries, Denmark, Norway, and Sweden have had markedly different responses to the SARS-CoV-2 (henceforth “coronavirus”) pandemic. This has gained international attention, both academically (cf. Gordon, Grafton, and Steinshamm 2020; Rubin et al. 2021; Yarmol-Matusiak, Cipriano, and Stranges 2021) and in the media (cf. Ludvigsson 2020; Steinglass 2020). The less restrictive Swedish social distancing responses to the pandemic, in particular, have been contrasted to the more extensive lock-down responses of Denmark and Norway (and indeed most of Europe).
In light of the higher coronavirus fatalities in Sweden compared to neighboring countries, parts of the international media have characterized the Swedish response as a dangerous and reckless experiment (Bjorklund and Ewing 2020; Bjorklund 2020; Münchach 2020; Goodman 2020; Sayers 2020). The policymaking processes underlying these different pandemic responses in Scandinavia have also been subjected to some academic scrutiny (see Rubin and de Vries (2020) for Denmark; Pierre (2020) and Petridou (2020) for Sweden and Christensen & Laegreid (2020) for Norway). Again, Sweden’s decision-making process appears to stand out. Whereas the national health agency had strong influence on the pandemic response in Sweden, the Danish and Norwegian responses appear to be driven more strongly from the PM’s office (Petridou 2020; Rubin et al. 2021). However, the comparative perspective of these studies remains underdeveloped. Hitherto, no study has compared and contrasted public leadership of pandemic responses across the three Scandinavian countries during the latest two pandemics: the H1N1 (henceforth “swine flu”) pandemic in 2009 and the coronavirus pandemic in 2020.
From a gross roster of more than fifty key actors involved in pandemic responses across the three countries, we produce replicable and quantifiable indicators of public leadership based on visibility in national newspaper articles. These indicators are comparable across different countries, agencies and periods. Importantly, the paper makes no normative judgements about policies implemented, focusing instead on measuring and outlining differences in public leadership during the two pandemics. We find that in terms of public leadership, Sweden actually appears the most consistent among the Scandinavian countries across the two pandemics, which stands somewhat in opposition to the international media’s narrative of an experimental and reckless response.
This paper proceeds as follows. First, existing scholarships on the public leadership of pandemics is laid out. Second, the media sources are introduced and the methods for data collection outlined. Third, comparative analyses across the three Scandinavian countries of the extent to which the Prime Minister (PM) or the leading national expert appeared in media during the 2020 and the 2009 pandemics, respectively. Fourth, analysis of more detailed indicators of media dominance across the three Scandinavian countries, looking at cross-country patterns between different ministers and experts. The paper ends by discussing the implication of this media analysis and pointing to future research.
2. Public leadership of pandemics
Public leadership in this paper refers to actors who assume the role of explaining and defending pandemic policies and strategies in the public arena or are otherwise prominently linked to the pandemic response by third parties, independently of whether these actors actually take the formal or informal decisions. Therefore, this paper proxies public leadership by media visibility for key actors across the three Scandinavian countries, most notably health experts and cabinet ministers, during these two 21st century pandemics. The study thus contributes to the academic literature of expert-led leadership during health emergencies (cf. Baekkeskov 2016a, 2016b; Baekkeskov and Rubin 2014; Baekkeskov and Öberg 2017; Christensen and Hesstvedt 2019;
Glenn, Chaumont, and Dintrans (2020). Christensen and Hestvedt 2019 study of Norwegian commissions in the period 1972–2016 found an increasing reliance over time on scientific knowledge and expert engagement in the commissions. This tendency is further supported by studies that analyzed Scandinavian expert engagement during the 2009 swine-flu pandemic response (Baekkeskov 2016a, 2016b; Baekkeskov and Rubin 2014). Baekkeskov and Rubin (2014), for example, analyzed very different countries during the 2009 pandemic, and documented a tendency for experts to dominate the public debate at the expense of politicians. They hypothesized that as a general rule, politicians would be hands-off during pandemic responses, in part because of the strength and authority of the medical profession in matters of health. Focusing on the 2020 pandemic, Glenn, Chaumont, and Dintrans (2020) analyzed public leadership across three countries, Chile, France and the United States, with a focus on health expert involvement in the management of the pandemic. The authors highlighted the need to achieve the right balance between political responsiveness and administrative responsibility in public leadership of the coronavirus pandemic, as tension between political and administrative messengers could undermine trust and support of government policies and institutions. Baekkeskov and Öberg (2017) found that heavy expert involvement in Danish and Swedish swine flu pandemic responses inhibited public deliberations. Despite enacting very different pandemic policies across countries, the debate in both countries could be characterized by expert consensus and limited public deliberation over viable policy alternatives. The authors concluded that strong expert engagement during pandemics might cause “deliberative freezing” in public discourse surrounding responses.
In this paper, we will explore whether some of these outlined theoretical expectations, hands-off political leaders, expert-driven public leadership, frozen public deliberation and implications for the public’s trust in pandemic responses, are supported by our comparative media analysis findings of Scandinavian public leadership during the two pandemics.
3. Data collection
A core component of this study is to compile and analyze proxies of public leadership of the pandemic response. To this end, we develop and analyze indicators of how often national leaders and leading health experts are mentioned in Danish, Norwegian, and Swedish newspaper reports about the 2009 and 2020 pandemics. Media visibility provides a strong indicator of whom the public associates with the pandemic response and therefore apparent leadership, independently of the formal policy responsibilities. There is, for example, a high degree of alignment between our media analysis and previous in-depth studies of the 2009 pandemic: the same experts that were found to be leading the pandemic response in these studies (cf. Baekkeskov 2016a, 2016b) were also found to dominate the media reporting in this study. Data on media visibility is derived from mentions in newspaper reports. These reports were retrieved from comprehensive collections of Scandinavian regional and national newspapers, magazines, and wires (1215 Danish, 946 Swedish, and 395 Norwegian; Infomedia Database 2021; Retriever Database 2021). The sample of media entries was restricted to reports
mentioning the two pandemics by using the Boolean search terms “H1N1 OR swine flu” (for 2009) and “COVID-19 OR corona*” (for 2020), respectively, for Denmark, Norway, and Sweden. To count the number of media report mentions, identified media texts were searched for names of leading health and elected officials. A gross list of more than fifty potential actors was compiled by researchers with familiarity of the different national policymaking processes during the two pandemics (Olivier Rubin and Erik Bækkeskov for Denmark, PerOla Öberg for Sweden and Reidar Staupé-Delgado for Norway). Subsequently, the sample was restricted to only contain the most prominently mentioned health and elected officials. These data allow us to produce indicators that are comparable across different countries and periods. We refrain from comparing absolute numbers because the available databases include different numbers of media outlets across the cases. The scope of the comparisons (with both temporal and geographical dimensions) together with the extensive data source (more than one million newspaper reports pertained to the pandemics) necessitate parsimonious and objective quantitative measures. These measurements can be used to display differences of degree across countries and identify longitudinal changes within countries. While many studies have characterized the 2009 swine flu response in Scandinavia as expert-led, for example, these measures can illuminate differences of magnitude in public leadership. Concretely, the Danish PM was completely absent from the public during the 2009 pandemic whereas Sweden’s was relatively more visible in the media. Another finding that the indicators helped elucidate was Sweden’s high degree of consistency in public leadership across the two pandemics despite the outbreaks’ obvious differences in terms of scale and impact. We will return to this finding in the discussion. The online supplementary material contains the gross list of actors as well as the detailed Boolean search terms in original language related to the subsequent analyses.
4. Media visibility of national leaders and leading health experts during the 2020 coronavirus response
The coronavirus hit Scandinavia relatively similarly and with equal force. In early March 2020, all three countries experienced exponentially increasing daily infection cases, culminating in hundreds. Each country implemented an initial lockdown phase with social distancing initiatives of various intensities in March and April; a reopening phase over the summer, where the initiatives were scaled down as infections receded; and then new social distancing initiatives in the last quarter of 2020. The Danish government was quick to take control of the policymaking process, implementing measures that went beyond those recommended by the health agency, centralizing power in ministries rather than health agencies and epidemic commissions, and ordering the health authorities to work within the paradigm of precaution rather than proportionality (Rubin and de Vries 2020). The Norwegian government also implemented more radical initiatives than those recommended by the national health agency, such as closing schools and banning the use of vacation homes (Christensen and Laegreid 2020). In contrast, observers agree that the Swedish government generally followed health expert advice. Politicians explicitly stated that expert agencies should handle health issues, and that expert recommendations should be the basis for decisions taken by
government (Petridou, 2020). In the daily press briefings, health experts were the main speakers compared to cabinet ministers (Pierre 2020, 5).
Figure 1 compares media exposures of Scandinavia’s PMs and the most prominent government health experts. These two actors alone account for around half of all mentions in our database. In Denmark and Norway, the most exposed health experts were agency heads: the director of the Danish Health Agency and the department director at
the Norwegian Institute of Public Health, respectively. In Sweden, by contrast, the most exposed health expert was the state epidemiologist at the Public Health Agency of Sweden, who was mentioned close to four times as often as the director of the health agency.
In Denmark, the PM clearly dominated the media landscape relative to the director of the national health agency. The PM’s presence in the media exploded in March when she announced extensive lock-down initiatives over several press meetings. Although media exposure for both actors declined over the summer, the PM sustained three times the daily mentions of the health agency director. During the late 2020 second wave and new response initiatives, the media gap widened again. Norway displayed the same basic pattern. The PM was clearly the most visible in the media as she announced the initial national social distancing initiatives. The PMs’ prominence was sustained over the summer (though media visibility for both actors declined). The PM and the health agency director were both somewhat less present in media during the second coronavirus wave (perhaps due to Norway’s relatively milder outbreak and less restrictive social distancing initiatives; Norwegian government 2020).
The longitudinal analysis of public visibility in Sweden, by contrast, shows media giving most prominence to the state epidemiologist during much of 2020. He appears to have been more publicly visible a few weeks earlier than anyone in Denmark and Norway, yet Sweden’s strongest surge in media coverage for both the PM and the state epidemiologist occurred one week later than surges in Denmark and Norway. Notably, however, gaps between Sweden’s media mentions of the Prime Minister and the leading national expert remained markedly smaller than Denmark’s or Norway’s. The PM eventually became more prominent than the state epidemiologist as Sweden implemented social distancing initiatives more similar in scope to Denmark’s and Norway’s against the outbreak’s second wave. This intertemporal shift suggests that despite Sweden’s unique quasi-autonomous health agencies (see discussion section), PMs can take increased public ownership of the pandemic response and implement policies that exceed health expert advice (Löneberg 2020).
In conclusion, it is clear that there were marked differences during 2020 in the media visibility of key actors involved in pandemic response between the Scandinavian countries. A comparison with the previous pandemic in 2009 can help illuminate whether these differences represent continuity or shifts in public leadership.
5. Media visibility of national leaders and leading health experts during the 2009 swine flu response
In 2009, Scandinavia was first hit by the novel swine flu virus in early May and a major wave of infections from October. The major policy response was vaccinations, which kicked off in November. Sweden and Norway offered vaccinations to all citizens whereas Denmark only offering vaccinations to high-risk groups (Cuesta et al. 2015). Comparative case studies of pandemic policymaking in Scandinavia show that the processes were expert-led, raising questions about how scientists advising policy in similar contexts could arrive at markedly different and opposing solutions (Baekkeskov
This difference was also reflected in discourses carried in national media (Baekkeskov and Öberg 2017).
Our analysis suggests that the reporting on the 2009 swine flu pandemic was less intensive than during the 2020 coronavirus pandemic. Between 20 and 25 percent of all Scandinavian media reports in 2020 (including sports, culture, the weather and so forth) referred to the coronavirus. For the swine flu pandemic, the comparable number for 2009 is around one percent. This difference could reflect the fact that, compared to the much more lethal and complex coronavirus outbreak, the swine flu turned out to be milder than initially feared, and could be curbed by existing pharmaceutical interventions that were less disruptive for the economy than lock-down measures.
Figure 2 compares media exposures of Scandinavia’s PMs and most prominent government health experts in 2009 pandemic reporting.
In all three countries, the most prominent national health expert was more visible in the media than the PM. This finding corroborates existing studies that have characterized the 2009 policymaking processes as expert-led (Baekkeskov and Rubin 2014; Baekkeskov 2016a, 2016b). The twin peaks in Sweden and Norway appear to reflect the pandemic dynamics that included initial outbreaks in May and mass vaccination campaigns in late autumn. The single peak in Denmark reflects that its vaccination campaign targeted small groups only and, hence, elicited less public interest. The media analysis also reveals that PM’s visibility differed significantly between countries. The Danish PM was completely absent in the media coverage of the outbreak, leaving the lead epidemiological expert to make statements and answer questions about the pandemic. In Norway, the leading health expert was also more exposed in media than the PM, in particular when mass vaccinations kicked off in November. In Sweden, the leading health expert was also the most mentioned in media. Notably, however, this visibility gap was again much narrower than in Denmark and Norway.
6. Ratios on public leadership of the pandemic responses across Denmark, Norway and Sweden
The above analyses of public leadership during the two pandemics reveal large inter-temporal and between-country differences. This section contributes with three additional indicators of public leadership of the pandemic responses that are disaggregated and focus on more key actors. This includes three ratios of media mentions of key officials: (i) Health experts/Ministers; (ii) Health Minister/PM; and (iii) Lead disease expert/Health agency director.
6.1. Experts/ministers-ratio
The experts-to-ministers-ratio is calculated by relating the number of pandemic articles mentioning the two most prominent health experts with the two most prominent ministers, namely the PM and the Minister of Health. Thus, the indicator captures variation between key cabinet ministers on one side and government health experts on the other. An experts-to-ministers-ratio of 1.0 indicates parity in public visibility between leading government experts and responsible cabinet members. Values above 1.0 suggest
that health experts dominate in media while values below 1.0 suggests that ministers dominate the media discourse about the pandemic.
Table 1 reveals substantial differences across the six pandemic cases, between countries and periods. First, the table supports that 2009 pandemic responses can be characterized as expert-led in all three countries. However, the ratios are markedly different.
Expert prominence differed between 1.4 in Norway and double that in Denmark (2.8). Between periods, Table 1 provides evidence that experts’ prominence in Swedish pandemic reporting has been fairly consistent (1.6 in 2009 and 1.1 in 2020). In sharp contrast, experts-to-ministers-ratios for Norway and Denmark shifted significantly (from 1.4 to 0.2 and from 2.8 to 0.2 respectively). This suggests continuity in Sweden while key officials’ public visibility in Norway and Denmark changed course substantially between the two pandemics.
### 6.2. Health minister/PM-ratio
The Health Minister/PM-ratio measures the number of pandemic articles referring to the Minister of Health vis-à-vis the number of articles mentioning the PM. The indicator provides a proxy for the extent to which pandemic communication by government primarily relies on health ministerial sectoral expertise or more top-down control from the PM’s office.
Table 2 reveals that in all three countries, PMs were more present than Health Ministers in 2020 media reports on the coronavirus. Again, however, the Danish ratio appears more extreme than its neighbors’, with four times more articles mentioning the PM than the Minister of Health. Ministers of Health were more prominent in the media than PMs during the 2009 pandemic in Denmark and Norway. In Sweden, however, PMs were more prominent during both pandemics. Between periods, the previously described pattern of continuity and shifts is repeated with Sweden displaying the most consistency (0.3 in 2009 compared to 0.6 in 2020).
### 6.2.1. Expert/director-ratio
The expert-to-director-ratio expresses whether health agency directors (Director) were more or less prominently exposed than leading officials specializing in public health and infectious diseases (Expert). In all three countries, health agency directors are appointed by the government while lead disease specialists are employed by the agencies.
Table 3 reveals that during 2009, the directors were much less prominent in media on the swine flu than their lead experts across the three Scandinavian countries. In 2020 by contrast, two media articles on the coronavirus mentioned the Danish director.
(a medical specialist in Gynecology and Obstetrics) for every one mentioning the lead expert (a professor in infectious disease epidemiology). The pattern is similar in Norway where the health agency director is mentioned three times as frequently as the agency’s epidemiological expert. Contrast this to Sweden in 2020, where the state epidemiological expert figures in almost four times more media articles than the Swedish health agency director. Again, Sweden displays the most continuity across periods.
7. Discussion and conclusion
The comparative analysis of public leadership (proxied by media visibility) across the three Scandinavian countries and two pandemics has produced some key findings that merit additional attention below.
7.1. Sweden displayed consistency in public leadership across the two pandemics
One finding, often overlooked in the public debate, is the fact that Sweden has displayed much consistency in public leadership across the two pandemics despite their obvious differences in scope. Sweden displayed a high degree of continuity of expert-driven public leadership while Norway and in particular Denmark both moved from a predominately expert-driven media presence in 2009 to a much stronger top-down, ministerially dominated process during the 2020 coronavirus pandemic. Sweden also displayed a public leadership style where both types of actors, politicians and experts, were more equally represented in the media compared to Denmark and Norway. With regards public leadership, therefore, experimentation appears more prominent in Denmark and Norway where clear shifts are evident in public leadership visibility.
7.2. Political leaders in Denmark and Norway embraced public leadership during the coronavirus pandemic
Contrary to the theoretical expectations drawn from the 2009 swine flu pandemic of hands-off political leaders (cf. Baekkeskov & Rubin, 2014), key ministers in Denmark and Norway appeared to have been very hands-on in their public leadership. The political theory of “lightning rods” describes situations where politicians let policy advisors and other government experts take ownership of particular policy initiatives as a way to avoid responsibility and blame (Ellis 1994). In the case of the 2020 pandemic, however, Danish and Norwegian politicians did not appear hide “behind the backs of experts” (Lodge and Boin 2020). It would be too simplistic to attribute this public visibility of political leaders to the severity of the coronavirus pandemic vis-à-vis the swine flu.
Table 3. Expert/Director-ratio.
| | 2009 | 2020 |
|--------|------|------|
| Denmark| 2.9 | 0.6 |
| Sweden | 3.8 | 3.6 |
| Norway | 3.4 | 0.3 |
Legend: Media reports mentioning the national epidemiological expert in relation to the health agency director.
flu pandemic. Such explanation fails to account for the diversity of public leadership across the three Scandinavian countries during the coronavirus. Rather, institutional differences between the Scandinavian countries constitute a more likely explanation, as will be discussed below.
7.3. The degree of expert and political involvement in public leadership was shaped by existing institutional structures
The observed differences in public leadership appear not to be rooted in political party ideology. During the 2009 swine flu pandemic, Denmark and Sweden had center-right governments while Norway had a center-left government. Yet elected leaders in Denmark and Norway had much lower media profiles than their Swedish colleagues. During the 2020 pandemic, Denmark and Sweden were governed by center-left leaning governments while Norway had a center-right government. Yet various types of officials’ public visibilities differed significantly between the three countries, and again Denmark and Norway were more similar (with elected leaders taking high profiles this time). Nor can differences in public leadership be ascribed any one actor. The ratios (Tables 1–3) that include a broader range of actors exhibit great consistency of findings, suggesting that leadership style cannot be attributed the distinct behavior of any one actor. Thus, rather than highlighting individual actors or political ideologies as key explanatory factors behind the different public leadership configurations, it appears prudent to emphasize factors rooted in the distinct existing institutional arrangements across Scandinavia.
The more expert-driven public leadership is consistent with Sweden’s administrative system, which guarantees that the central agencies enjoy a high degree of quasi-decisional autonomy (Öberg and Wockelberg 2016; Christiansen, Niklasson, and Öhberg 2016; Petridou, 2020). It is unconstitutional for politicians to interfere in specific cases where the agency exercises authority vis-à-vis citizens. In contrast, Denmark and Norway have central agencies under ministerial authority.
A clear expression of expert-guided policy processes is the crucial position that Sweden’s health agencies took in coronavirus press briefings. The key health agencies jointly organized daily press briefings where ministers were not present. The agencies explained and defended the Swedish strategy, responding at length to questions from Swedish and international journalists. Political leaders organized fewer, and separate, press meetings to present specific decisions with responsible ministers present. Thus, the state epidemiologist at the Public Health Agency of Sweden quickly became the public face of pandemic strategy.
In contrast, the faces of coronavirus pandemic leadership in Denmark and Norway were the elected leaders. In Denmark, political leaders exercised control over information flows from the health authorities and interfered in their independent assessments. Internal correspondence reveals that the health agency was instructed to withhold key statistics (Findalen and Weichardt 2020), and that it was requested to work on more gloomy pandemic scenarios than it had deemed realistic (Jensen, Birk & Lund, 2020). The PM led most of the press meetings. At times, she would be flanked by health experts. But at other times, health experts were noticeably absent (Rubin and de Vries
Similarly, Norway’s PM and other ministers played central roles in communicating with citizens and the media through daily media briefings (Christensen and Laegreid 2020). Initially, the Norwegian health agency had authority to hold press briefings. But from March 11 2020, the government decided that all communication would be coordinated from the PMs office, which implied a much more visible and active role for the PM (Kvinnsland 2021).
7.4. Negligible impact of proportional versus precautionary strategy
Whether experts or politicians dominate in media coverage might be associated with the extent to which governments approach pandemic responses as generally proportional or precautionary. In Denmark, swine flu risks were relatively limited, and responses were quite focused, suggesting proportionality. In contrast, coronavirus risks were perhaps greater, but the initial 2020 lockdown shows that response policies were certainly socially wide-ranging ahead of any clarity about these risks. Indeed, the government endorsed precautionary strategy. Hence, these Danish cases support that publicly apparent leadership may impact whether overall response strategy is proportionality or precaution. But the Scandinavian comparisons show that such an association is not general. During the swine flu pandemic in 2009, the Norwegian and Swedish expert-led policymaking process resulted in responses where everyone was offered vaccination (with uptake of 60 percent in Sweden and 45 percent in Norway; Mereckiene et al. 2010), suggesting precautionary motives (Cuesta et al. 2015; however, cf Baekkeskov 2016a, 2016b). During the 2020 pandemic, the quite similar public leadership configuration in Sweden resulted in arguably proportional responses, with relatively limited societal lockdowns. As shown, Swedish media focused to similar degrees on national experts and elected leaders in the two events. In Norway, precaution was apparently repeated in 2020, yet media focus flipped (as shown).
7.5. Practical implications for public debate
As previously mentioned, research from the swine flu pandemic suggests that high levels of health expert involvement in media coverage can “freeze” public and policymaker deliberation about policy alternatives (Baekkeskov and Öberg 2017). Indeed, the Swedish expert-driven response to the coronavirus appears to have spurred a technocratic public debate. Public critiques of Swedish responses in the initial phase of the pandemic were mainly articulated by other health experts, most notably large groups of senior health scientists (Elgh et al. 2020; Carlsson et al. 2020). These public critiques were very concrete and proposed stricter social distancing measures than the government had enacted. In Denmark, by contrast, public critique was mainly offered by newspaper editors and opposition leaders, based on principled discussions of dangers of limiting freedoms and relinquishing powers to the executive branch (Serup 2020; Rubin 2020). This critique, where health experts were noticeably absent, problematized the pandemic strategy without actually suggesting alternative viable public health policies. More systematic research is needed to shed light on the deliberative consequences of these different public leadership configurations.
7.6. Practical implications for trust and support
Public leadership plays an essential role in building trust and legitimacy of actors involved and implemented policies (Siegrist and Zingg 2014). As Boin, Lodge, and Luesink (2020, 199) have noted, the coronavirus pandemic made rock stars out of obscure scientists and thrust them into political scenes. In Sweden, key health professionals were positioned in media as experts and as public leaders.
Evidence on whether this dual image had practical implications for public support and trust in the government’s handling of the coronavirus pandemic is ambiguous. Methodologically, disentangling effects of media public leadership from actual policies implemented is difficult when comparing trust or approval levels between these countries. Yet some polls suggest that government support for coronavirus pandemic policies was higher in Denmark and Norway, where political leaders were more visible in the media (Smith 2020). However, differences stand out most between Denmark and Sweden where Swedes appeared less supportive of their national pandemic response than their Danish neighbors (Keldsen 2020; Hope Dashboard 2021; Djoef 2020). In addition, approval ratings for Sweden’s leading health experts and PM waned substantially during the last quarter of 2020 (Reuters, 2020; Henley 2020). A longitudinal poll from March 2020 onwards documents that 80–60 percent of the Danish citizens consistently agreed that the pandemic policies were necessary and approved of the government’s handling of the pandemic, while 55–35 percent of the Swedish citizens expressed agreement on the same questions (Hope Dashboard 2021). Another poll conducted in March/April 2020 showed that compared to their Swedish counterparts, Danish citizens were 24 percent more satisfied with the government’s handling of the pandemic and 11 percent more satisfied with the national health agency (Djoef 2020). The findings are less clear when comparing Norway and Sweden. Some polls do appear to suggest higher trust and approval ratings in Norway (Smith 2020). A longitudinal poll actually puts the Norwegian approval ratings of the government’s handling of the pandemic throughout 2020 on par with Danish approval ratings, hovering around 80–60 percent (Kvinnsland 2021, 186). But other polls suggest somewhat similar approval ratings between Norway and Sweden (Keldsen 2020) or they find that the Swedish citizens actually exhibited higher trust in authorities than citizens in Norway in the initial phases of the pandemic (Helsingin et al., 2020). Additional research, therefore, is needed to elucidate more robustly the impact of different public leadership configurations on citizens’ trust and support.
In conclusion, leaders’ public visibility comparisons within Scandinavia between the 2009 swine flu and 2020 coronavirus show marked contrasts and surprising constituencies in who media portray as response policymakers. This shows the importance of further investigations of not merely pros and cons of different coronavirus policies but also the different public leadership structures and processes. In this paper, we have devised media-based indicators of public leadership that can easily be replicated in other studies and be used for large-n cross-country comparisons or as inputs in more context-specific studies of public leadership. An improved understanding of public leadership is key for effective crisis management. Such understanding might more fundamentally prepare the countries for the next crises that are unlikely to be similar to the present one.
Disclosure statement
The authors declare that they have no conflict of interest.
Funding
The authors received funding from Riksbankens jubileumsfond [P20-0463] for this work.
ORCID
Olivier Rubin http://orcid.org/0000-0002-2364-6782
Erik Baekkeskov http://orcid.org/0000-0001-9028-9570
PerOla Öberg http://orcid.org/0000-0002-3522-4966
References
Baekkeskov, E. 2016a. “Explaining Science-Led Policy-Making: Pandemic Deaths, Epistemic Deliberation and Ideational Trajectories.” Policy Sciences 49 (4): 395–419. doi:10.1007/s11077-016-9264-y.
Baekkeskov, E. 2016b. “Same Threat, Different Responses: Experts Steering Politicians and Stakeholders in 2009 H1N1 Vaccination Policy-Making.” Public Administration 94 (2): 299–315. doi:10.1111/padm.12244.
Baekkeskov, E., and P. Öberg. 2017. “Freezing Deliberation through Public Expert Advice.” Journal of European Public Policy 24 (7): 1006–1026. doi:10.1080/13501763.2016.1170192.
Baekkeskov, E., and O. Rubin. 2014. “Why Pandemic Response is Unique: Powerful Experts and Hands-off Political Leaders.” Disaster Prevention and Management 23 (1): 81–93. doi:10.1108/DPM-05-2012-0060.
Bjorklund, K. and A. Ewing. 2020. “The Swedish COVID-19 Response Is a Disaster. It Shouldn’t Be a Model for the Rest of the World.” Time Magazine, October 14, 2020. Accessed 12 May 2021. https://time.com/5899432/sweden-coronovirus-disaster/
Bjorklund, K. 2020. “The Inside Story of How Sweden Botched Its Coronavirus Response.” Foreign Policy, December 22, 2020. Accessed 12 May 2021. https://foreignpolicy.com/2020/12/22/sweden-coronavirus-covid-response/
Boin, A., M. Lodge, and M. Luesink. 2020. “Learning from the COVID-19 Crisis: An Initial Analysis of National Responses.” Policy Design and Practice 3 (3): 189–204. doi:10.1080/25741292.2020.1823670.
Carlsson, M., L. Einhorn, S. Einhorn 2020., and A. Wahlin. “Folkhâlsomyndighetens har misslyckats - nu måste politikerna gripa in.” Dagens Nyheter, April 24, 2020. Accessed 12 May 2021. https://www.dn.se/debatt/folkhalsomyndighetens-har-misslyckats-ny-maste-politikerna-gripa-in/
Christensen, J., and S. Hesstvedt. 2019. “Expertisation or Greater Representation? Evidence from Norwegian Advisory Commissions.” European Politics and Society 20 (1): 83–100. doi:10.1080/23745118.2018.1515861.
Christensen, Tom, and Per Laegreid. 2020. “Balancing Governance Capacity and Legitimacy—How the Norwegian Government Handled the COVID-19 Crisis as a High Performer.” Public Administration Review 80 (5): 774–779. doi:10.1111/puar.13241.
Christiansen, P. M., B. Niklasson, and P. Öhberg. 2016. “Does Politics Crowd out Professional Competence? The Organisation of Ministerial Advice in Denmark and Sweden.” West European Politics 39 (6): 1230–1250. doi:10.1080/01402382.2016.1176368.
Cuesta, J. G., P. Aavitsland, H. Englund, O. Gudlaugsson, S. H. Hauge, O. Lyytikäinen, and M. Virtanen, the Nordic Influenza Comparison Group 2016. “Pandemic Vaccination Strategies
and Influenza Severe Outcomes during the Influenza A (H1N1) pdm09 Pandemic and the Post-Pandemic Influenza Season: The Nordic Experience.” *Eurosurveillance* 21 (16): 30208. doi:10.2807/1560-7917.ES.2016.21.16.30208.
Djoef. 2020. “Håndteringen af corona-krisen: Danskere har mere tillid til myndighederne end svenskerne.” Accessed 12 May 2021. https://www.djoef.dk/-/media/documents/djoef/p/para-graf_20_april_2020_2_a4_web.ashx
Elgh, F., Å. Gustafsson, B. Lundbäck 2020., and M. Vaagen. “Fler åtgärder krävs för att bromsa smittspridning.” *Svenska Dagbladet*, March 15, 2020. Accessed 12 May 2021 https://www.svd.se/fler-åtgärder-kravs-att-bromsa-smittspridning
Ellis, R. 1994. *Presidential Lightning Rods the Politics of Blame Avoidance*. Kansas: University Press of Kansas.
Findalen, J., and B. Weichardt. 2020. “Regeringen bremsede positive coronatal.” *Ekstrabladet*, May 27, 2020. Accessed May 12 2021. https://ekstrabladet.dk/nyheder/politik/danskpolitik/regeringen-bremse-positive-coronatal/8139077
Glenn, J., C. Chaumont, and P. V. Dintrans. 2020. “Public Health Leadership in the Times of COVID-19: A Comparative Case Study of Three Countries.” *International Journal of Public Leadership* 17 (1): 81–94. doi:10.1108/IJPL-08-2020-0082.
Goodman, P. 2020. “Sweden Has Become the World’s Cautionary Tale.” *New York Times*, July 7, 2020. Accessed 12 May 2021. https://www.nytimes.com/2020/07/07/business/sweden-economy-coronavirus.html
Helsingen, L. M., E. Refsum, D. K. Gjøstein, M. Løberg, M. Bretthauer, M. Kalager, and L. Emilsson, for the Clinical Effectiveness Research group 2020. “The COVID-19 Pandemic in Norway and Sweden–Threats, Trust, and Impact on Daily Life: A Comparative Survey.” *BMC Public Health* 20 (1): 1–10. doi:10.1186/s12889-020-09615-3.
Hope Dashboard. 2021. “Dashboard.” Accessed 12 May 2021. https://hope-project.dk/dashboard/.
Jensen, H., C. Birk, and M. Lund. 2020. “Interne Dokumenter Afslører Strid om Coronastrategi: »Politikerne Har Villet Udwise Handlekraft for Enhver Pris – Uanset Eksperternes Rådgivning.” *Berlingske*, May 28, 2020. Accessed May 12 2021. https://www.berlingske.dk/samfund/interne-dokumenter-afslorer-strid-om-coronastrategi-politikerne-har
Keldsen, T. 2020. ” Danskeerne har størst tillid til myndighedernes håndtering af coronaviruskrisen i Norden.” Accessed 12 May 2021. https://yougov.dk/news/2020/03/19/danskeerne-har-storst-tillid-til-myndighedernes-han/
Kveimland, S. 2021. “Norges offentlige utredninger 2021: 6 Myndighetenes håndtering av koronapandemien.” *Rapport fra Koronakommisjonen*. Accessed May 12 2021. https://www.koronakommisjonen.no/files/2021/04/Koronakommisjonens_rapport_NOU.pdf.
Lodge, M., and A. Boin. 2020. “Great easing? Leaders Face a Tragic Dilemma, But They Should Not Hide Behind the Backs of Experts.” *LSE blog*. Accessed 12 May 2021. https://blogs.lse.ac.uk/covid19/2020/05/14/leaders-face-a-tragic-dilemma-economic-collapse-or-mass-causalities/
Löningd, C. 2020. ” Sprickan mellan regeringen och Folkhälsoomyndigeten växer.” *Svenska Dagbladet*, November 22, 2020. Accessed 12 May 2021. https://www.svd.se/sprickan-vaxer-sa-blev-tegnell-av-med-sitt-veto
Ludvigsson, J. 2020. ” Sweden: Why the ‘Moral Superpower’ Dissented Over Covid-19.” *Financial Times*, November 17, 2020. Accessed 12 May 2021 https://www.ft.com/content/71c8a636-4d40-4d0e-947f-af8de9b6b82c
Mereckiene, J., S. Cotter, F. D’Ancona, C. Giambi, A. Nicoll, D. Levy-Bruhl and D. O’Flanagan 2010. “Differences in National Influenza Vaccination Policies Across the European Union, Norway and Iceland 2008–2009.” *Eurosurveillance* 15 (44): 1–10. doi:10.2807/ese.15.44.19700-en.
Münchau, W. 2020. “Sweden’s Covid-19 Experiment Holds a Worldwide Warning.” *Financial Times,* September 13, 2020. Accessed 12 May 2021. https://www.ft.com/content/9a1e118f-a4ef-43ab-a981-58b8bd35c0e1
Norwegian government 2020. “Nye nasjonale innstramminger.” *Press release,* November 26, 2020. Accessed 12 May 2021. https://www.regjeringen.no/no/aktuelt/nye-nasjonale-innstrammer/id2776995/
Öberg, S. A., and H. Wockelberg. 2016. “Nordic Administrative Heritages and Contemporary Institutional Design.” In: Carsten Greve, Per Lægreid & Lise H. Rykkja (Eds), *Nordic Administrative Reforms,* 57–78. London: Palgrave Macmillan.
Petridou, E. 2020. Politics and administration in times of crisis: Explaining the Swedish response to the COVID-19 crisis. *European Policy Analysis* 6(2): 147–158. doi:10.1080/10.1002/epa2.1095
Pierre, J. 2020. “Nudges against Pandemics: Sweden’s COVID-19 Containment Strategy in Perspective.” *Policy and Society* 39 (3): 478–493. doi:10.1080/14494035.2020.1783787.
Reuters. 2020. “Swedes confidence in P M’s handling of coronavirus crisis declining, poll shows.” September 2 2020. Accessed 22 June 2021. https://www.reuters.com/article/us-healthcoronavirus-sweden-idUSKBN25T398
Retriever Database. 2021. Accessed 12 May 2021. https://app-retriever-info-com.ezproxy.its.uu.se/login
Rubin, M. 2020. “Coronakrisen er ikke en blankocheck til Mette Frederiksen. Vi skal passe på folkestyret.” *Politiken,* March 26, 2020. Accessed 12 May 2021. https://politiken.dk/debat/ledere/art7727770/Coronakrisen-er-ikke-en-blankocheck-til-Mette-Frederiksen.-Vi-skal-passe-p%C3%A5-folkestyret.
Rubin, O., and D. de Vries. 2020. “Diverging Sensemaking Frames During the Initial Phases of the COVID-19 Outbreak in Denmark.” *Policy Design and Practice* 3 (3): 277–296. doi:10.1080/25741292.2020.1809809.
Rubin, O., N. Errett, R. Upshur, and E. Baekkeskov. 2021. “The Challenges Facing Evidence-Based Decision Making in the Initial Response to COVID-19.” *Scandinavian Journal of Public Health* doi:10.1080/25741292.2020.1809809.
Sayers F. 2020. “The great Covid-19 experiment — did Sweden beat us all?” Evening Standard, 14 September, 2020. Accessed 16 Feb 2021. https://www.standard.co.uk/news/world/sweden-coronavirus-covid19-lockdown-a4546976.html
Serup, S. 2020. ” Er vi gået bemyndigelsesamok?” *B.T.*, April 5, 2020. Accessed 12 May 2021. https://www.bt.dk/samfund/soes-marie-serup-er-vi-gaaet-bemyndigelsesamok.
Siegrist, M., and A. Zingg. 2014. “The Role of Public Trust during Pandemics: Implications for Crisis Communication.” *European Psychologist* 19 (1): 23–32. doi:10.1027/1016-9040/a000169.
Smith, M. 2020. “What is Public Opinion in Sweden Around Coronavirus?” *Yougov.* Accessed 12 May 2021. https://yougov.co.uk/topics/health/articles-reports/2020/10/20/what-public-opinion-sweden-around-coronavirus.
Steinglass, M. 2020. “The Nordic Countries’ Different Approaches to COVID-19 will Continue.” *Economist,* November 17, 2020. Accessed 12 May 2021. https://www.economist.com/the-world-ahead/2020/11/17/the-nordic-countries-different-approaches-to-covid-19-will-continue.
Yarmol-Matusiak, E. A., L. E. Cipriano, and S. A. Stranges. 2021. “A Comparison of COVID-19 Epidemiological Indicators in Sweden, Norway, Denmark, and Finland.” *Scandinavian Journal of Public Health* 49 (1): 69–78. doi:10.1177/1403494820980264. | 2025-03-06T00:00:00 | olmocr | {
"pdf_page_numbers": [
[
0,
2246,
1
],
[
2246,
5667,
2
],
[
5667,
9073,
3
],
[
9073,
12617,
4
],
[
12617,
13094,
5
],
[
13094,
16399,
6
],
[
16399,
19545,
7
],
[
19545,
19940,
8
],
[
19940,
22135,
9
],
[
22135,
24946,
10
],
[
24946,
28355,
11
],
[
28355,
31648,
12
],
[
31648,
35251,
13
],
[
35251,
38197,
14
],
[
38197,
41572,
15
],
[
41572,
45215,
16
]
]
} |
d59fd91d876b35c815b23062ea790468738c5048 | {
"olmocr-version": "0.1.59",
"pdf-total-pages": 108,
"total-fallback-pages": 0,
"total-input-tokens": 251698,
"total-output-tokens": 74188,
"fieldofstudy": [
"Economics"
]
} | Impacts of Carbon Pricing in Reducing the Carbon Intensity of China’s GDP
Jing Cao
Mun Ho
Govinda R. Timilsina
WORLD BANK GROUP
Development Research Group
Environment and Energy Team
June 2016
**Abstract**
In contributing to global climate change mitigation efforts as agreed in Paris in 2015, China has set a target of reducing the carbon dioxide intensity of gross domestic product by 60-65 percent in 2030 compared with 2005 levels. Using a dynamic computable general equilibrium model of China, this study analyzes the economic and greenhouse gas impacts of meeting those targets through carbon pricing. The study finds that the trajectory of carbon prices to achieve the target depends on several factors, including how the carbon price changes over time and how carbon revenue is recycled to the economy. The study finds that carbon pricing that starts at a lower rate and gradually rises until it achieves the intensity target would be more efficient than a carbon price that remains constant over time. Using carbon revenue to cut existing distortionary taxes reduces the impact on the growth of gross domestic product relative to lump-sum redistribution. Recycling carbon revenue through subsidies to renewables and other low-carbon energy sources also can meet the targets, but the impact on the growth of gross domestic product is larger than with the other policies considered.
This paper is a product of the Environment and Energy Team, Development Research Group. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at [email protected].
Impacts of Carbon Pricing in Reducing the Carbon Intensity of China’s GDP
Jing Cao, Mun Ho, and Govinda R. Timilsina
Key Words: Climate change policies, Carbon pricing, China, NDC, CGE modeling
JEL Classification: Q54, D58
1 The authors would like to thank Bert Hofman, Todd Johnson, Garo Batemanian, Karlis Smits, Ying Fan and Mike Toman for their valuable comments and suggestions. The views and interpretations are of authors and should not be attributed to the World Bank Group and the organizations they are affiliated with. We acknowledge World Bank’s Knowledge for Change (KCP) Trust Fund and the Harvard Global Institute for financial support.
2 Jing Cao ([email protected]) is an Associate Professor at the School of Economics and Management, Tsinghua University, Beijing, China; Mun Ho ([email protected]) is a Visiting Scholar at the Resources for Future, Washington, DC and also a member of Harvard China Project; Govinda Timilsina ([email protected]) is a Senior Economist at the Development Research Group, World Bank.
1 Introduction
In its Nationally Determined Contribution (NDC) coming out of the Paris Agreements reached at the 21st meeting of the Conference of Parties to the United Nations Framework Convention on Climate Change (UNFCCC), the Chinese government has set a goal to cut CO₂ emissions per unit of GDP by 60 to 65 percent by 2030, compared to 2005 levels. It also has reiterated previously announced goals that carbon emissions would peak, and that the share of non-fossil fuels would increase to 20% of total energy consumption, by 2030. Earlier in 2016 the government announced the 13th Five-year Plan, with the goals of reducing energy per unit GDP by 18% between 2015 and 2020 and to reduce CO₂ per unit GDP by 20%. The longer term targets imply that China would need additional low carbon infrastructure investments, including further deployment of renewable energy, as well as further improvements in energy efficiency.
This study addresses the following “what-if” question: what would be the economic consequences of China seeking to meet its NDC target with carbon pricing? In practice, China can be expected to use a range of policy instruments to that end. However, an examination of economic consequences using a hypothetical least-cost instrument, a dynamic carbon price, provides a useful context for evaluating more complicated policy portfolios. Moreover, an examination of carbon pricing is consistent with the expressed desire of the government to rely more on market mechanisms.
The analysis is carried out using a multi-sector growth model of China, including a household model component that allows us to discuss the impacts of policies on different demographic groups. While there exist a large number of studies analyzing impacts of a hypothetical carbon tax in China, studies analyzing the potential impacts of carbon pricing to meet China’s NDC do not exist. To the knowledge of authors, this is the first analysis that examines implications of meeting China’s NDC through a carbon pricing mechanism.
We first use the model to develop a base case up to 2030. The base case includes existing plans to expand the renewable energy and nuclear energy included in China’s 12th and 13th Five year plans out to 2020. The base case does not include carbon pricing. The government has announced an intention to extend current experimental carbon trading programs to a national system beginning in 2017. However, no details of the level of cap and coverage of sources have been given, and so the planned national emission trading scheme is not included in the study. In the base case, the carbon emissions are rising continuously during the 2012-2030 period, though
they peak around 2030 through various GHG mitigation measures already in the base case. The base case is then compared to scenarios with carbon pricing as well as a policy for additional investment in renewables for electricity generation. Carbon pricing is implemented in the model through surcharges on fossil fuel prices based on their carbon contents.
The study considers the following policy formulations. First, “lower” carbon pricing and “higher” carbon pricing trajectories are implemented to achieve, respectively, 60% or 65% reduction of carbon intensity of GDP from 2005 levels. For the 65% target, we compare a fixed carbon surcharge on fossil fuels over time to a trajectory where the surcharge starts at a low level but gradually rises.
As has been emphasized by many others in the literature, the method of recycling the revenues generated from carbon pricing is another important policy component, as it has an important impact on the net economic cost of such a pricing policy. We consider two approaches to recycle the carbon revenue to the economy— a lump sum transfer to households, and a cut in the VAT and capital income tax. In addition, the study considers a policy that subsidizes renewable energy with revenue generated through the carbon pricing.
We find that the carbon surcharge on fossil fuels aimed at achieving a 60% reduction in CO₂ intensity of GDP would rise from 1.6 yuan/ton of CO₂ in 2016 to 26 yuan/ton of CO₂ in 2030 (all units measured in 2010 yuan). This carbon price will reduce the absolute level of 2030 emissions by 3.3% with GDP only 0.11% lower than the base case, if carbon revenue is recycled to cut existing taxes. In this scenario, 2030 energy use falls by 2.6% and electricity use by 1.5%. With the carbon pricing policy to achieve a 65% reduction in CO₂ intensity of GDP, the carbon surcharge on fossil fuels rises to 157 yuan/ton of CO₂ in 2030, generating a 13% reduction in energy use, a 16% reduction in CO₂ and a 0.74% lower GDP, all relative to the base case.
If instead of cutting existing taxes, the carbon revenues are recycled by giving them to households as a lump sum rebate, then GDP is 1.2% lower than the base case in 2030, instead of 0.74%. In this case, the same carbon price slows aggregate output growth more, and hence also slows growth energy consumption and emissions more. This reinforces the lesson emphasized by many others – recycling revenues from carbon pricing by reducing existing tax wedges is a useful opportunity for softening adverse economic impacts.
The paper is organized as follows. Section 2 briefly highlights the CGE methodology developed for this study (detailed description of the model is presented in the Appendix A),
followed by results of the model for the base case scenario in Section 3. Section 4 presents results of carbon pricing scenarios to meet China’s NDC, followed by sensitivity analysis in which the baseline is altered in Section 5. Section 6 concludes the paper.
2. The CGE model
The CGE model used for this study is a dynamic recursive growth model where the main agents are the household, producers, government and the rest of the world. Household savings, enterprise retained earnings and government-funded investments are the main sources of investment; unlike most developed economies the government role in China is much larger. Detailed description of the model is presented in Appendix A, here we summarize its key features.
One of the key features of this model that distinguishes it from large number of CGE models available for China for climate policy analysis is that it allows for heterogeneity among households, whereas most existing CGE models have a single representative household. In our model, households are distinguished by size, presence of children, age of head and region. The income elasticity is different for different consumption items and thus projects a different structure of consumption in the future when incomes rise. Such a function allows us to distinguish the impact of policies on different households via the consumption channel.\(^3\) Labor is supplied inelastically by households.
The private household savings rate is set exogenously and total national private savings is made up of household savings and retained earnings of enterprises. These savings, together with allocations from the central plan, finance national investment. They also finance the government deficit and the current account surplus. Investment in period \(t\) increases the stock of capital that is used for production in future periods. The plan part of the capital stock is assumed immobile in any given period, while the market part responds to relative returns. Over time, plan capital is depreciated and the total stock becomes mobile across sectors.
The government imposes taxes on value added, sales and imports. On the expenditure side, it buys commodities, makes transfers to households, pays for plan investment, makes interest payments on the public debt and provides various subsidies. Expenditures on interest and
\(^3\) See Equations A20-A24 for the household demand function in Appendix A.
transfers are exogenous, and the exogenous deficit target is met by making government spending on goods endogenous.
Finally, the rest of the world supplies imports and demands exports. Domestically produced goods are imperfect substitutes for imports. World relative prices and the current account balance are set exogenously in this one-country model, and an endogenous variable for terms of trade clears the current account equation. The world price of commodity \( i \) relative to \( j \) is assumed to be at base year ratios throughout the projection period with the exception of world oil prices, where projections from the U.S. Energy Information Administration 2013 *Annual Energy Outlook* are used (and treated as exogenous).
On the supply side, 33 industries are distinguished, each producing output that is given by a nested series of constant-returns-to-scale CES functions.\(^4\) Primary factors include capital, labor and land. Pure TFP growth and biased technical change are allowed; in particular, energy input per unit output can decline faster than other input-output ratios.
Since the electricity sector is the main contributor of CO\(_2\) emissions in China, in the version of the model used for this study, the electricity sector is disaggregated into one transmission subsector and 9 distinct generation sub-sectors – coal, gas, nuclear, hydro, other, wind, solar, coal-CCS and gas-CCS. The tier structure of electricity production is given in Figure 1. At the top node total output is the aggregate of Generation and Transmission, and Generation is in turn an aggregate of Baseload and Intermittent Renewables.\(^5\) The Baseload aggregate is composed of the output from coal, gas, nuclear, hydro, miscellaneous other minor sources, and the potential technologies, coal with carbon storage and sequestration (CCS) and gas-CCS. The Intermittent Renewables are wind and solar. The elasticities of substitution among these generation sources are presented in Table A2 in Appendix A and also noted in Figure 3; we assume a high degree of substitution among the baseload sources, and we follow other studies in assuming an elasticity of one between Baseload and Renewables.
This structure reflects the reality that average generation costs are different and yet different generation sources co-exist; there are considerations beyond average costs that determine the share of various sources in a highly regulated sector. We should make a technical note regarding this formulation: the imperfect substitution of kWh’s means that the quantity
\(^4\) The production structure is given in Figure A1 and equations A4-A12 in Appendix A.
\(^5\) This aggregation of Baseload and Renewables is similar to that in C-GEM (Qi et al. 2014).
The output from each generation source is also expressed as a nest of CES functions; for nuclear, hydro, wind and solar there is a “Resource” input that represents non-produced inputs such as suitable rivers, and land with wind and sun. In any given period, these resources are given by an upward sloping supply curve to represent the short-run costs of developing such regions. This is similar to the representation in Vennemo et al. (2013) and Sue Wing et al. (2014) as discussed in the Appendix A.
The electricity generation and distribution system in China is dominated by a few large enterprises and tightly regulated by the NDRC. The prices are set by the NDRC after negotiation among the generators, distribution monopolies and major users. The dispatch order (which generation units are used at any moment) is determined for the most part by a “fairness” principle and a loose aim to promote renewables; it is not set according to least-cost dispatch, and it does not reflect bid prices for electricity in a competitive wholesale generation market. For a carbon price to work in the electric power sector, there must thus be reforms that would allow price signals to matter – that the net cost of fossil fuels reflects the carbon price, and that dispatch is sensitive to the prices asked by the generators. Our policy simulation here is predicated on the assumption that such reforms will take place.
There are 33 markets for the commodities; that is, there are 33 endogenously determined prices that equate supply with demand for the domestic commodities identified in the model. The total supply consists of domestically produced goods and imported varieties; these are assumed to be imperfect substitutes. There are three markets for the factors of production – land, capital and labor – and three prices to clear them.
The base case simulation is determined by the projections of the exogenous variables and the initial stocks of debt, capital and labor force. Given the initial stocks, we solve for the three factor prices and the 33 commodity prices that clear the markets in the first period. This gives us all the quantities for the first period, including investment that augments the next period stock.
---
6 The projections of the exogenous projections are described in detail in Appendix A. The population projections come from UN Population Division’s World Population Prospects: 2012 Revision, downloaded from their web site, [http://esa.un.org/unpd/wpp/unpp/panel_population.htm](http://esa.un.org/unpd/wpp/unpp/panel_population.htm).
The details of the construction of the base year data for the power sector and the base case projections of the various generation sources is given in Appendix B. We discuss the impact of using this electricity projection for the base case in section 5.
**Figure 1. Structure of electricity sector**
3. The Base Case Scenario
GDP growth in China between 1978 and 2007, the eve of the Global Financial Crisis, was 9.9% per year. With the stimulus to fight the effects of the Crisis in 2008 and 2009, the growth rate between 2007 and 2011 was maintained at a high 9.6%, however, with the end of the stimulus and the continuing weakness in the world economy, it decelerated to about 7.5% during 2012-14. In the 13th Five-year plan announced in March 2016, the goal was to reduce the energy per unit GDP by 18% between 2015 and 2020, and reduce CO2 per unit GDP by 20%.
---
Cao and Ho (2014) discuss the sources of growth in China for the past 30 years – aggregate productivity and industry productivity performance – and discuss how these might inform projections of future growth. They relate this growth accounting of China to the literature about the middle-income trap and growth slowdowns.
The base case growth path is driven by the exogenous variables including population, labor force quality, capital quality growth, total factor productivity and saving rates. Our assumptions for the drivers are based on our reading of the historical record (Cao and Ho (2014) and are described in Section A.4 in the Model Appendix A. This base case is not designed to replicate precisely any particular projection; it is only intended to provide a point of comparison for the policy cases. The base case growth is summarized in Table 1 and Figure 2. Between 2010 and 2030 the population is projected to rise from 1,360 to 1,470 million, while the working age population falls from 938 to 883 million. With the assumed increase in average hours worked per person (including longer work lives) and labor quality, effective labor supply increase by 8% over these 20 years despite the fall in working population. Our model projects that GDP will grow with an average rate of 6.4% per year during 2015-20 and decelerates to 4.6% during 2020-30. The 6.4% rate for the 2015-2020 period is very close to that assumed in 13th Five-year plan (6.5% for the 2016-2020 period). The consumption share of GDP rises from 35% in 2010 to 54% in 2030 due to falling trends in household saving rates. More precisely, growth decelerates due to lower TFP, lower savings rate, and falling working age population.
Primary energy use grows at 3.9% during 2015-20, implying a fall in energy intensity of 3.4% per year, close to the 13th FYP target of a 18% reduction in intensity. This is to be compared to the historical record given in Figure 2; the intensity index was quite volatile and averaged a decline of 4.1% per year during the 11th Five-year Plan (2006-10). During 2020-30 projected energy use decelerates more, but with the slowing GDP growth, the intensity falls at only 2.1% per year. There is a big change in the composition of energy sources; coal use only grows at 2.2% compared to oil at 4.3% and gas at 7.7% per year during 2015-20.
---
8 GDP projection is endogenous to the model which is calculated based on several variables which are exogenous. They include population growth, share of working age population to total population, saving rates, dividend payout rates, government taxes and deficits, world prices for traded goods, current account deficits, rate of productivity growth, rate of improvement in capital and labor quality. The values of these variables are provided in Table A3 in Appendix A. Note that GDP projections vary from across the sources. For example, IMF (2014) projections were 6.9% for 2012-2020 and 5.3% for 2020-30. World Bank (2013) projections were 7.0% for 2016-2020 and 5.0% for 2026-30. Projections change across the sources and also overtime from the same source due to different assumptions.
Table 1. Base case projection
(All numbers in this table are model outputs except those specified at the bottom of the table)
| Variable | 2010 | 2015 | 2020 | 2030 | 2015-20 growth rate | 2020-30 growth rate |
|---------------------------------|-------|-------|-------|-------|---------------------|---------------------|
| Population (million) * | 1,360 | 1,404 | 1,440 | 1,470 | 0.51% | 0.21% |
| Effective labor supply (bil. 2010 yuan) | 16,687 | 17,529 | 18,100 | 18,098 | 0.64% | 0.00% |
| GDP (billion 2010 yuan) | 40,145 | 58,956 | 80,465 | 126,235 | 6.4% | 4.6% |
| Consumption/GDP* | 0.35 | 0.44 | 0.50 | 0.55 | | |
| Energy use (million tons sce@) | 3,249 | 4,005 | 4,839 | 5,958 | 3.9% | 2.1% |
| Coal use (million tons) | 3,122 | 3,666 | 4,083 | 4,610 | 2.2% | 1.2% |
| Oil use (million tons) | 441 | 537 | 662 | 835 | 4.3% | 2.3% |
| Gas use (million cubic meters) | 107,291 | 166,354 | 241,467 | 381,163 | 7.7% | 4.7% |
| Electricity use (TWh) | 4,206 | 5,554 | 6,886 | 8,951 | 4.4% | 2.7% |
| CO₂ emissions (fossil fuel, million tons) | 7,388 | 8,831 | 10,146 | 11,886 | 2.8% | 1.6% |
| CO₂ emissions (total, million tons) | 8,299 | 9,819 | 11,139 | 12,797 | 2.6% | 1.4% |
| Carbon intensity (kg CO₂/yuan) | 0.184 | 0.150 | 0.126 | 0.094 | -3.4% | -2.9% |
| GDP per capita (2010 yuan) | 29522 | 42004 | 55884 | 85859 | | |
| GDP per capita; PPP US$2005 | 8228 | 11707 | 15575 | 23929 | | |
* Exogenous variables
@ SCE refers to standard coal equivalent
** Model output but calibrated with growth rates from IEA (2014)
Figure 2. Projections of GDP, Energy and Emissions in the Base Case
Table 2. Growth of electricity sector in base case
| | 2010 | 2015 | 2020 | 2025 | 2030 | Annual growth rate |
|--------------------------|------|------|------|------|------|--------------------|
| | | | | | | 2015-20 | 2020-30 |
| Total TWh | 4206 | 5554 | 6886 | 8017 | 8951 | 4.4% | 2.7% |
| Percent share | | | | | | 2.7% | 2.5% |
| Coal | 77.0 | 71.4 | 65.7 | 65.0 | 64.6 | 12.0% | 5.5% |
| Gas | 1.9 | 2.7 | 3.9 | 4.6 | 5.1 | 15.3% | 5.8% |
| Nuclear | 1.8 | 3.4 | 5.6 | 6.7 | 7.6 | 3.7% | 0.4% |
| Hydro | 16.3 | 16.2 | 15.7 | 14.0 | 12.7 | 8.5% | 3.1% |
| Other | 1.9 | 2.4 | 2.9 | 2.9 | 3.0 | 11.5% | 4.2% |
| Wind | 1.2 | 3.4 | 4.8 | 5.2 | 5.6 | 37.7% | 3.2% |
| Solar | 0.0 | 0.3 | 1.4 | 1.4 | 1.5 | 0.4% | 0.0% |
Prices relative to GDP deflator with $P(\text{coal}, 2010) = 1$
| | 2010 | 2015 | 2020 | 2025 | 2030 | 2015-20 | 2020-30 |
|--------------------------|------|------|------|------|------|----------|----------|
| Coal | 1.00 | 0.93 | 0.89 | 0.89 | 0.89 | 1.00 | 0.73 |
| Gas | 1.37 | 1.31 | 1.27 | 1.29 | 1.32 | 1.00 | 0.73 |
| Nuclear | 1.00 | 0.93 | 0.90 | 0.90 | 0.90 | 1.00 | 0.73 |
| Hydro | 0.74 | 0.69 | 0.66 | 0.67 | 0.68 | 1.00 | 0.73 |
| Other | 1.39 | 1.33 | 1.29 | 1.32 | 1.34 | 1.00 | 0.73 |
| Wind | 1.37 | 1.23 | 1.17 | 1.17 | 1.18 | 1.00 | 0.73 |
| Solar | 2.77 | 2.51 | 2.33 | 2.35 | 2.38 | 1.00 | 0.73 |
$P_{\text{GDP}}:P_{\text{Labor}}$
Figure 3. Projections of electricity generation in the base case
Electricity use is projected to grow at 4.4% during 2015-20 and at 3.0% during 2020-30. The change in the structure of generation sources is given in Table 2 and graphed in Figure 3; we see that electricity use grows faster than coal due to the rise of renewables and nuclear energy – coal-fired electricity grows at 2.7% during 2015-20, nuclear at 12%, wind at 11%. Solar starts from a very tiny base but rises rapidly, at 40% per year. In the 2020-30 period, renewable growth decelerates sharply but is still faster than coal, with the exception of hydro growth which falls to 0.4% per year.
As a result of the shift towards cleaner fuels driven by the assumed improvements in energy efficiency and the endogenous changes in prices in the base case, CO₂ emissions, including those from cement manufacture, only grows at 2.6% per year during 2015-20. That is, the CO₂ intensity falls faster than energy intensity; by 2030 it is 58.7% lower than the 2005 level (compared to the NDC target of a 60-65% reduction). CO₂ emissions rise steadily from 8,300 million tons in 2010 to 12,800 million in 2030, thereafter we project a stabilization around 13 billion tons for at least the next 10 years. Coal use rises to 4.1 billion tons in 2020 and then plateaus at about 4.6 billion beginning in 2030.
4. Simulations of Carbon Pricing Scenarios
In our base case the CO₂ intensity falls by 58.7% in 2030 relative to 2015. We consider carbon price trajectories that can reduce CO₂ intensity by 60 or 65% by 2030, as in the NDC. To that end, we consider flat rates that apply for all years, and prices that start low and rise every year so as to achieve the same cumulative emission reductions. We also consider two alternative approaches for recycling the carbon revenue to the economy: (i) recycle the carbon revenue by cutting existing taxes and (ii) rebating it lump-sum to households.
These policy scenarios are defined in Table 3. In the R1CUT scenario, we find through trial-and-error that a carbon price starting at 1.6 yuan/ton of CO₂ in 2015, and rising gradually to 26 yuan in 2030, hits the 60% reduction goal. In this scenario, the carbon revenues collected
---
9 Our energy projection is based on IEA (2014). In the “Current Policies” scenario of IEA, China’s primary energy use is projected to grow at 2.9% during 2012-20 and 1.9% for 2020-30, and project coal use to grow at 2.1% (2012-20) and at 1.3% (2020-30).
10 All yuan units are measured in 2010 yuan, the base year of the model.
11 We should digress here to clarify what is meant by a statement like “a carbon price of ¥10 per ton in each year of the projection.” In an economy with no TFP growth and little capital accumulation, absolute (and thus relative) commodity prices will differ little from the base year. In this case there is little ambiguity about what ¥10 in each year means. With rapid growth in output and incomes per capita, the price of commodities relative to the price of labor is
are offset by cuts in the VAT and capital income tax. In the R2CUT scenario, through trial-and-error we find that a carbon price starting at ¥10/ton of CO₂ in 2015 and rising to ¥157 in 2030, hits the 65% reduction goal. In the F2CUT scenario, we keep the carbon price flat at ¥82 level throughout the study horizon in order to achieve the same cumulative reductions in CO₂ emissions relative to the base case over the 2015-2030 period (-9.8%) as the rising carbon pricing policy R2CUT.12
To examine the effects of different revenue recycling options we have Policy R2lump which has a rising carbon price and the same cumulative CO₂ reductions as R2cut, but here the carbon revenues are returned lump sum to households instead of cutting VATs. The carbon price paths under various scenarios are illustrated in Figure 4. Finally, in the “rising renewable subsidies” (RRsub) scenario we introduce subsidies for renewable power in the electricity sector so that the share of wind, solar and hydro reaches those specified in the IEA (2014). These subsidies are financed by a fossil fuel tax in proportion to carbon content of fuels and it rises over time as the renewable share rises.
Figure 4. Carbon price (Yuan/tCO₂)
falling rapidly, and relative prices are changing due to different capital-labor ratios (and possibly different industry TFP growth). In our accounting system we measure things in terms of the GDP basket, and since relative prices change in our economy, “¥10 per ton” means a different basket of goods in each period. A flat ¥10/ton results in a changing ad valorem rate on the coal price.
12 We also considered an analogous policy F1CUT for comparison with R1CUT, but the results were not particularly useful and so we omit it here.
Table 3. Policy scenarios
| Scenario definition | Scenario name | Target to be achieved under the scenario | Revenue recycling scheme | Carbon surcharge (per ton CO₂, yuan2015) |
|---------------------|---------------|------------------------------------------|--------------------------|----------------------------------------|
| | | | | 2015 | 2025 | 2030 |
| Base case | B0 | 58.7% reduction of CO₂ intensity without carbon pricing | | | | |
| Low carbon pricing with rising rate | R1CUT | 60% reduction in CO₂ intensity by 2030 | Cut VAT and capital tax | 1.6 | 18 | 26 |
| High carbon pricing with rising rate | R2CUT | 65% reduction in CO₂ intensity by 2030 | Cut VAT and capital tax | 9.8 | 108 | 157 |
| | R2LUMP | Same cumulative CO₂ reductions as R2cut | Lump sum rebate to households | 9.3 | 102 | 149 |
| High carbon pricing with flat rate | F2CUT | Same cumulative CO₂ reductions as R2cut | Cut VAT and capital tax | 82 | 82 | 82 |
| Rising renewable subsidies to reach IEA's New Policies scenario | RRSUB | Renewable subsidies financed by carbon revenue | Carbon tax to offset renewables costs | 2.8 | 36 | 58 |
High versus low carbon prices
The NDC announced by the government set a goal for reducing the CO₂:GDP intensity in 2030 by 60-65% compared to 2005 levels. Here we compare the economic impacts of achieving the more stringent target, 65%, with the costs of reducing the intensity by only 60%, recalling that in the base case the reduction was 58.7%. The macroeconomic impacts of R2CUT versus R1CUT are given in Table 4. We present the results for the first year of the carbon pricing, 2015, and the final year. The impact of CO₂ intensity over time is plotted in Figure 5 while the impact on GDP, energy use and CO₂ emissions for the high carbon price case are plotted in Figure 6. The impact on prices and output of each of the 33 industries in 2030 are plotted in Figures 7 and 8.
We first describe the impact of the lower carbon price case (R1). The initial impact on GDP and energy use in 2015 of the small carbon price is correspondingly small, about 0.2% reduction in energy use. By 2030, when the carbon price reached 26¥/ton CO₂, energy use has fallen 2.6%, and CO₂ emissions has fallen 3.3%. The carbon price has discouraged the use of
fossil fuels, and raised the price of electricity, leading to a rise in the prices of energy intensive goods relative to the base case as shown in Figure 7 for 2030. Coal price at the mine-mouth rises by 6.2% while crude oil price rises by 1.2% inclusive of the tax, leading to a 1.4% rise in electricity prices.
Table 4. The effects of achieving higher reductions with higher carbon price case
(R1CUT vs R2CUT; rising carbon prices with offsetting cuts in VAT)
| Variable | 2015 | 2030 |
|-----------------------------------|-----------------------------|-----------------------------|
| | Base case | R1CUT: Low C price | R2CUT: High C price | Base case | R1CUT: Low C price | R2CUT: High C price |
| GDP (billion yuan 2010) | 58,956 | -0.004 | -0.027 | 126,235 | -0.11 | -0.74 |
| Consumption (bil yuan 2010) | 25,290 | -0.003 | -0.017 | 67,264 | -0.11 | -0.71 |
| Investment (bil yuan 2010) | 25,374 | -0.007 | -0.045 | 43,747 | -0.13 | -0.86 |
| Government consumption (bil yuan 2010) | 6,973 | 0.000 | 0.000 | 12,785 | 0.00 | 0.00 |
| Energy use (million tons of see) | 4,005 | -0.21 | -1.21 | 5,958 | -2.6 | -13.1 |
| Coal use (million tons) | 3,666 | -0.29 | -1.71 | 4,610 | -4.1 | -20.0 |
| Oil use (million tons) | 537 | -0.04 | -0.27 | 835 | -0.7 | -4.0 |
| Gas use (billion cubic meters) | 166,354 | -0.09 | -0.53 | 381,163 | -1.8 | -9.4 |
| Electricity (billion kWh) | 5,554 | -0.11 | -0.65 | 8,951 | -1.5 | -7.9 |
| CO2 emissions (inc cement; mil tons) | 9,819 | -0.23 | -1.33 | 12,797 | -3.3 | -16.0 |
Cumulative CO2 (2015-2030) 166,675 -1.9 -9.8
Carbon price yuan/ton CO2 ¥1.6 ¥9.8 ¥26.2 ¥157.1
Carbon price as a share of total revenue 0.11% 0.66% 1.1% 5.8%
| Electricity generation | Base case values | % change | R1CUT: Low C price | % change | R2CUT: High C price | % change |
|------------------------|------------------|----------|-------------------|----------|---------------------|----------|
| Coal | 3967 | -0.18 | -1.04 | | 5782 | -2.7 | -14.2 |
| Gas | 152 | 0.02 | 0.11 | | 459 | -0.3 | -2.2 |
| Nuclear | 190 | 0.06 | 0.35 | | 678 | 1.0 | 5.4 |
| Hydro | 902 | 0.06 | 0.35 | | 1132 | 0.9 | 5.2 |
| Other | 133 | 0.11 | 0.67 | | 271 | 1.6 | 8.5 |
| Wind | 191 | 0.002 | 0.01 | | 498 | 0.01 | 0.12 |
| Solar | 19 | 0.004 | 0.02 | | 130 | 0.04 | 0.25 |
Figure 6. Impacts of carbon pricing to achieve 65% reduction in CO₂ intensity (Scenario R2CUT)
Figure 7. Impacts of carbon pricing on commodity prices in 2030
These changes in relative prices, and the reduction in capital stocks, lead a fall in industry output relative to the base case, as shown in Figure 8. By 2030, the consumption of coal falls in relative terms by 4.1%, oil by 0.7% and electricity by 1.5% (Table 4). The output of energy intensive goods, including primary metals and cement (building materials sector), fall more than the output of consumer goods and services industries. These price distortions reduce GDP, leading to lower investment every year. By 2030 GDP is lower by 0.11%, aggregate consumption is lower by 0.11%, investment by 0.13%, and total energy by 2.6%. As a result of these reductions, CO$_2$ emissions fall by 3.3% in 2030 in Policy R1CUT.
The composition of electricity generation change at the same time that total electricity output falls by 1.5%; coal generation falls by 2.7% in 2030 while nuclear and hydro rise by about 1% (bottom section of Table 4). The price of intermittent renewables (wind and solar) are essentially unchanged and their output remains unchanged leading to a rise in their share contribution.
When we raise taxes much higher in the R2CUT scenario, the effects are amplified. The tax by 2030 is 157\$/ton, and the price of coal rises by 37% instead of 6%. The impact of lower investment and capital stocks due to these price shocks is a 0.23% reduction in 2020 GDP, and a 0.74% reduction in 2030 (compared to the 0.11% cut in GDP under the low tax policy). The tax on CO$_2$ raises revenues equal to 5.8 percent of total revenues and enables a corresponding cut in the VAT and capital taxes. The lower GDP and higher energy prices lead to a 13% reduction in energy use compared to 2.6% under policy R1. Carbon emissions are 16% lower in order to reach the 65% intensity target. Within the electricity group, coal generation falls by 14% while nuclear and hydro rises by more than 5%. This is due to the rise in coal price leading to a rise in the relative price, P(coal electricity)/P(hydro), of 5.6%. Solar power output rises by only 0.3% given that we assumed that the elasticity between intermittent renewables and baseload is only 1, that is, most of the substitution is from coal to nuclear and hydro and not to intermittent sources.
Flat versus rising carbon prices
In this comparison we illustrate the benefits of phasing in a carbon price gradually rather than imposing a flat rate over time. We compare two paths of carbon prices that deliver the same cumulative carbon emission reduction by 2030 – R2CUT versus F2CUT (carbon prices are
plotted in Figure 4). The comparison of impacts is shown in Table 5 for the first and final years. Note that the reduction in cumulative CO$_2$ emissions over 2015-30 is the same, 9.8%, and that the first year the tax is 82¥/ton in F2CUT compared to 9.8¥ in the rising tax case.
With the higher tax rate in the first year of F2CUT, the larger distortion lead to a 0.26% fall in GDP relative to base case, compared to 0.03% in R2CUT. There is a bigger impact on investment that cumulates to a 0.56% reduction in the 2030 capital stock relative to base case, compared to a 0.51% relative reduction in the rising tax scenario.
The other variables such as energy consumption, electricity composition and CO$_2$ emissions follow the same pattern – much bigger changes in year 1 and a more modest change in 2030, but cumulative losses that are bigger. The cumulative reduction in CO$_2$ is the same but the reduction relative to base case in year 2030 is only 9.5% in the flat tax case compared to 16% in the rising tax case. In that year the changes relative to base case in coal generated electricity are -8.2% versus -14.2%, and in hydro they are +2.7% versus +5.2%.
To summarize, the flat tax path that delivers the same cumulative CO$_2$ reduction as the rising tax trajectory imposes greater GDP losses. A richer model that takes explicit account of adjustment costs would generate even bigger differences between a gradually rising tax and a flat tax profile.
*Options for recycling carbon tax revenue*
To give an illustration of the importance of how recycling carbon tax revenues is done, we compare the previous R2cut scenario where we cut existing VAT and capital income tax to a case where the revenues are given back to households as a lump sum transfer. The results for this R2LUMP case are shown in Table 6 for 2030 together with R2cut. These two paths of rising carbon tax rates are set so that the cumulative change in CO$_2$ emissions over 2015-2030 are the same -9.8%. It turns out that the relatively weaker economic performance in the lump-sum case generates lower emissions, and so the same cumulative reduction requires a lower carbon price than in R2CUT. This difference means that the change in carbon intensity in 2030 is not exactly equal to the 65.0% reduction in the R2cut case, but very close.
Table 5. The effects of flat versus rising carbon taxes (F2cut vs. R2cut)
(High tax cases, revenue recycled by cutting VAT and TK)
| Variable | 2015 | | 2030 | |
|------------------------------------|-----------------------|----------------------|-----------------------|----------------------|
| | Base case | F2CUT: Flat C Tax | R2CUT: High C Tax | Base case |
| | values | % change from base | values | % change from base |
| GDP (billion yuan 2010) | 58,956 | -0.26 -0.027 | 126,235 | -0.61 -0.74 |
| Cumulative GDP (not discounted) | | | | |
| Consumption (bil yuan 2010) | 25,290 | -0.17 -0.017 | 67,264 | -0.54 -0.71 |
| Investment (bil yuan 2010) | 25,374 | -0.41 -0.045 | 43,747 | -0.64 -0.86 |
| Capital stock (bil yuan 2010) | 138,851 | 0.00 0.00 | 386,091 | -0.56 -0.51 |
| Energy use (million tons of sce) | 4,005 | -8.9 -1.21 | 5,958 | -7.8 -13.1 |
| Coal use (million tons) | 3,666 | -12.4 -1.71 | 4,610 | -11.9 -20.0 |
| Oil use (million tons) | 537 | -2.2 -0.27 | 835 | -2.4 -4.0 |
| Gas use (billion cubic meters) | 166,354 | -4.3 -0.53 | 381,163 | -5.4 -9.4 |
| Electricity (billion kWh) | 5,554 | -5.0 -0.65 | 8,951 | -4.6 -7.9 |
| CO2 emissions (inc cement; mil tons)| 9,819 | -9.8 -1.33 | 12,797 | -9.5 -16.0 |
| Cumulative CO2 (2015-2030) | | 184,840 | -9.8 -9.8 |
| Carbon tax yuan/ton CO2 | ¥82.5 | ¥9.8 | ¥82.5 | ¥157 |
| Carbon tax as a share of total revenue | 5.0% 0.66% | | 3.3% 5.8% |
| Electricity generation | values | % change from base | values | % change from base |
| Coal | 3967 | -7.96 -1.04 | 5782 | -8.2 -14.2 |
| Gas | 152 | 0.41 0.11 | 459 | -1.1 -2.2 |
| Nuclear | 190 | 2.78 0.35 | 678 | 2.9 5.4 |
| Hydro | 902 | 2.78 0.35 | 1132 | 2.7 5.2 |
| Other | 133 | 5.22 0.67 | 271 | 4.6 8.5 |
| Wind | 191 | 0.09 0.01 | 498 | -0.1 0.12 |
| Solar | 19 | 0.17 0.02 | 130 | 0.1 0.25 |
The carbon tax rate in 2030 is ¥149 in the R2lump case but slightly higher at ¥157 in the tax cut case. The lack of an offsetting cut means that the VAT’s in the lump sum case remain high and generate greater distortions. The net effect for many goods is a higher price in R2lump; for example, the price of primary metals rise by 5.7% versus 4.6% in R2CUT, for food manufacturing it is 2.1% versus 0.9%, and for construction it is 2.9% versus 1.8%. These greater price distortions contribute to a smaller GDP. The cut in capital taxes in R2CUT also allows...
enterprises to retain more earnings and invest more, and thus the capital stock by 2030 falls relative to base case by only 0.51% compared to a 1.3% reduction in R2lump. As a result of this lower capital stock and higher value-added tax rates in the lump sum case, GDP in 2030 is 1.2% below base case, compared to only 0.74% lower in the tax cut case.
With lower aggregate output and lower carbon taxes prices in the lump sum case, the reduction in energy consumption is very similar, 13.2% versus 13.1%. The changes in CO₂ emissions are corresponding similar, -16%, and thus the fall in CO₂:GDP intensity relative to base case is slightly smaller in the R2lump case. Hydro generation is +4.7% in R2lump versus +5.2% in R2CUT. The change in intermittent renewables is different; in the lump-sum case the reduction in total energy demand leads to a 0.22% relative fall in wind power, while it rose by 0.12% in the tax cut case.
In sum, we have similar reductions in emissions at the cost of somewhat greater reduction in GDP growth (relative to base case) with the lump-sum transfer case than the tax-cut case. The slower GDP growth is compounded over time. Note that our gains in efficiency by reducing existing value-added taxes and capital taxes are calculated for a recursive model; in a dynamic model with foresight the impact of cuts in capital taxation would be even bigger, that is, the slowdown in GDP with the tax cuts would be even more modest than we have calculated.
Table 6. The impact of different methods of recycling carbon revenues
(R2CUT vs. R2LUMP, cases with high and rising carbon tax)
| Variable | 2030 Base case values | % change from base |
|---------------------------|-----------------------|--------------------|
| GDP (billion yuan 2010) | 126,235 | -0.74 | -1.17 |
| Consumption (bil yuan 2010)| 67,264 | -0.71 | -0.77 |
| Investment (bil yuan 2010) | 43,747 | -0.86 | -2.13 |
| Capital stock (bil yuan 2010)| 386,091 | -0.51 | -1.28 |
| Energy use (million tons of sce) | 5,958 | -13.1 | -13.2 |
| Coal use (million tons) | 4,610 | -20.0 | -19.9 |
| Oil use (million tons) | 835 | -4.0 | -4.5 |
| Gas use (billion cubic meters) | 381,163 | -9.4 | -9.4 |
| Electricity (billion kWh) | 8,951 | -7.9 | -8.1 |
| CO2 emissions (inc cement; mil tons) | 12,797 | -16.0 | -16.1 |
| Cumulative CO2 (2015-2030) | 184,840 | -9.8 | -9.8 |
| Carbon tax yuan/ton CO2 | ¥157 | ¥149 |
| Carbon tax as a share of total revenue | 5.8% | 5.5% |
| Electricity generation values | % change from base |
|-------------------------------|--------------------|
| Coal | 5782 | -14.2 | -14.3 |
| Gas | 459 | -2.2 | -2.4 |
| Nuclear | 678 | 5.4 | 4.9 |
| Hydro | 1132 | 5.2 | 4.7 |
| Other | 271 | 8.5 | 7.8 |
| Wind | 498 | 0.12 | -0.22 |
| Solar | 130 | 0.25 | 0.00 |
**Subsidies for renewables and other low-carbon sources of electricity**
In the policy RRsub we subsidize the cost of nuclear, hydro, wind, and solar using the revenues from a fossil fuel tax in proportion to their carbon contents. The subsidies are chosen so that the share of each of these sources in generation is equal to the projection by IEA (2014) under their “New Policies Scenario (NPS)”. We focus on the kilowatt-hours generated rather than the capacity of the various sources which are also projected separately in IEA (2014).
Table 7 gives the shares of electricity generation (by kWh) projected in IEA (2014) under their two scenarios – “Current Policies” and “New Policies” for two selected years. We first compute the change in shares for each year of the projection period between the two policies, e.g. under NPS the coal share in 2030 falls by 8.8 percentage points, and the wind share rises by 2.2 points. Although our base case shares not identical with the IEA’s Current Policies projections in each year, they are set to be very close. We set the annual renewable targets in our policy scenarios as our base case shares plus the differences between the IEA scenarios, as illustrated in Table 7 for 2020 and 2030. In the RRsub policy we set subsidies separately for nuclear, hydro, wind, solar and other to hit the higher shares under the New Policies Scenario.
The results of using the carbon tax to subsidize renewables and other low-carbon energy sources are given in Table 8, together with those from R1cut and R2cut for comparison. In RRsub, the subsidies for renewables start at low rates but become quite substantial in the later years as given in the last column of Table 7. The share for wind has to rise from 5.7% in the base case to 7.9%, which requires a subsidy of 48%, while the share of hydro has to rise from 12.8% to 14.7%. As noted in Section 2, these renewables require natural resource inputs that have upward sloping supply curves and thus require higher returns to meet higher resource demands.
Table 7. Renewables target from IEA (2014) “New Policies Scenario” and subsidies in the model.
| | 2020 Current Policies | 2020 New Policies | Difference | 2030 Current Policies | 2030 New Policies | Difference | RRsub subsidy (%) |
|----------|-----------------------|-------------------|------------|-----------------------|-------------------|------------|------------------|
| Coal | 65.4 | 61.9 | -3.6 | 64.1 | 55.3 | -8.8 | 18.9 |
| Nuclear | 5.8 | 6.3 | 0.5 | 7.6 | 9.2 | 1.6 | 13.2 |
| Hydro | 15.6 | 17.1 | 1.4 | 12.8 | 14.7 | 1.9 | 48.3 |
| Wind | 4.8 | 5.6 | 0.8 | 5.7 | 7.9 | 2.2 | 46.7 |
| Solar | 1.4 | 1.8 | 0.4 | 1.5 | 2.5 | 1.0 | |
| Gas, other | 7.0 | 7.4 | 0.5 | 8.3 | 10.4 | 2.1 | |
| Total | 100 | 100 | 0 | 100 | 100 | 0 | |
The value of output of the electricity sector was ¥3240 billion in 2010, with a value added of 810 billion equaling 2.1% of GDP. These high subsidies thus require a sizable carbon tax to finance them. By 2030 the CO₂ tax needed is ¥58/ton, which is in between the ¥26 carbon price
in R1cut and ¥157 in R2cut. These subsidies result in a 2.2% fall in the price of electricity compared to the base case and thus the 1.9% reduction in the consumption of electricity is only slightly bigger than the 1.5% reduction in the R1cut scenario with its much lower carbon price. That is, the carbon prices that raised the price of electricity in R1cut encouraged conservation and generated a 3.3% cut in total CO₂ emissions at a modest cost of reducing GDP by 0.11% in 2030. In this Renewable subsidy scenario, the subsidies discourage electricity saving and mutes the total CO₂ impact of generating less coal electricity.
Total CO₂ emissions in 2030 fall by 7.9% relative to base case, given the 58 yuan carbon price. The cumulative 2015-2030 relative cut in emissions of 2.1% is only slightly bigger than the 1.9% cut in R1cut, with the 26 yuan carbon price. The large carbon price and distortions of the electricity price generate a relative GDP loss of 0.53% in 2030 compared to only 0.11% loss in R1cut. The GDP loss in R2cut is bigger at 0.74% but that achieved a cumulative CO₂ reduction of 9.8% compared to 2.1% for RRsub. That is, the ambitious renewable targets require large distortions in the later years, which generate CO₂ reductions as a byproduct of a greater slowdown in GDP growth.
With the subsidies, coal generation in 2030 falls to 58.4% of total kWh compared to 64.6% in the base case and 63.8% in R1CUT, and hydro rises to a 14.5% share compared to 13.0% in R1CUT. These changes result in a larger reduction coal use in 2030, 10% versus 4.1% in R1cut, and CO₂ emissions fall by 7.9% compared with 3.3% in R1CUT.
5. Sensitivity Analysis on Non-Fossil Fuels Growth Assumptions
In the base case the projection of resource inputs and capital inputs into the electricity generation functions is set according to the projections under the “Current Policies” scenario (CPS) for China in IEA (2014). As shown in Table 2, this scenario assumes a rapid growth of nuclear, wind and solar generation. The rapid expansion of relatively expensive generation resources (i.e., nuclear, wind, solar) implies higher subsidies and ultimately higher costs to the society.
Table 8. Subsidies to renewables financed by carbon taxes (scenario RRsub).
| Variable | 2015 Base case | RRsub | R1CUT | 2030 Base case | RRsub | R1CUT | R2CUT |
|--------------------------------------|----------------|-------|-------|----------------|-------|-------|-------|
| GDP (billion ¥2010) | 58,956 | -0.018| -0.004| 126,235 | -0.53 | -0.11 | -0.74 |
| Consumption (bil ¥2010) | 25,290 | -0.014| -0.003| 67,264 | -0.45 | -0.11 | -0.71 |
| Investment (bil ¥2010) | 25,374 | -0.020| -0.007| 43,747 | -0.61 | -0.13 | -0.86 |
| Government consumption (bil ¥) | 6,973 | -0.023| 0.00 | 12,785 | -0.40 | 0.00 | 0.00 |
| Fossil energy use (mil tons of sce) | 4,005 | -0.41 | -0.21 | 5,958 | -5.8 | -2.6 | -13.1 |
| Coal use (million tons) | 3,666 | -0.63 | -0.29 | 4,610 | -10.1 | -4.1 | -20.0 |
| Oil use (million tons) | 537 | -0.03 | -0.04 | 835 | -1.1 | -0.7 | -4.0 |
| Gas use (billion cubic meters) | 166,354 | -0.29 | -0.09 | 381,163 | -6.0 | -1.8 | -9.4 |
| Electricity (billion kWh) | 5,554 | -0.09 | -0.11 | 8,951 | -1.9 | -1.5 | -7.9 |
| Electricity price | | -0.18 | 0.10 | | -2.2 | 1.4 | 7.9 |
| CO2 emissions (mil tons) | 9,819 | -0.48 | -0.23 | 12,797 | -7.9 | -3.3 | -16.0 |
Cumulative CO2 (2015-2030) 184,840 -2.1 -1.9 -9.8
Carbon tax yuan/ton CO2 ¥2.8 ¥1.6 ¥58 ¥26 ¥157
| Electricity generation | 2015 share of total generation | 2030 share of total generation |
|------------------------|--------------------------------|--------------------------------|
| Coal | 0.714 | 0.646 |
| Gas | 0.027 | 0.051 |
| Nuclear | 0.034 | 0.076 |
| Hydro | 0.162 | 0.127 |
| Other | 0.024 | 0.030 |
| Wind | 0.034 | 0.056 |
| Solar | 0.003 | 0.015 |
In order to get an idea of the economic impacts of this push away from coal and towards non-fossil fuels in the IEA Current Policies Scenario, we conducted an alternative set of simulations with a “no coal reduction” (NCR) scenario and a “CPS targets” scenario. In the NCR case we allow the coal share to fall from the first year 2010 to 2016 as in the base case, but then maintain close to that share out to 2030. The share of hydro in the base case is falling due to the projected difficulty in finding more hydro resources, and we maintain this in NCR. Note that even with the falling share the absolute production of hydro power is rising over time. The
remainder of the electricity demand is met by the other sources (gas, nuclear, wind, solar, others) that rise modestly in their share contribution to offset the falling hydro contribution.
In the CPS case, we do not change the resource and capital supplies exogenously as in the base case, but let them remain at the NCR paths and use a system of taxes and subsidies to hit the higher targets for renewables and nuclear. We impose taxes on coal and subsidies for the rest in order to attract more resources and capital into the non-fossils and discourage coal generation. We require that the tax revenue exactly equal the subsidies so that there is no net (new) transfer to the government.
The comparison of these two scenarios is given in Table 9. In the NCR case, total coal consumption in 2020 is 4,147 million tons compared to 4,083 in the base case. The coal power is cheaper and thus the electricity demand in NCR is slightly higher, 6,949 billion kWh in 2020 compared to 6,886 in the base case. The impact of achieving the CPS targets via taxes and subsidies is substantial even in 2020; coal use is 3.5% lower than in NCR leading to a 2.1% reduction in total energy consumption. The price of average electricity is 1.3% higher leading to a 2.7% reduction in 2020.
By 2030 the impact is magnified by the cumulative GDP losses and reduction in capital stock. Coal use in 2030 is 5.5% lower in the CPS case while total energy consumption in 3.5% lower. Aggregate GDP is 0.29% lower due to these distortions with reductions in both consumption and investment. The generation mix shifts from 70.6% coal and 4.9% nuclear to 65.3% coal and 7.6% nuclear. The wind contribution rises from 4.7% to 5.7%.
While our specifications of the cost functions for the renewables and nuclear are simple, it allows a representation of the costs of finding suitable sites for such generation methods, and also possibly higher costs of transmission. The simulated costs are comparable to the renewable subsidies case shown in Table 8 given that the same cost function elasticities are used. There the cost of reducing CO₂ emissions by 7.9% in 2030 is a 0.53% relative cut in GDP, here the 4.0% reduction in CO₂ reduces GDP relative to base case by 0.29%.
Table 9. Comparing “no coal reduction (NCR)” with IEA “Current Policies” scenarios
| Variable | 2020 NCR values | % change | 2030 Base case values | % change | 2030 CPS vs NCR values | % change |
|-----------------------------------------|-----------------|----------|-----------------------|----------|------------------------|----------|
| GDP (billion ¥2010) | 80,494 | -0.10 | 126,235 | -0.29 |
| Consumption (bil ¥2010) | 38,784 | -0.13 | 67,264 | -0.35 |
| Investment (bil ¥2010) | 31,532 | -0.07 | 43,747 | -0.25 |
| Government consumption (bil ¥2010) | 8,958 | -0.14 | 12,785 | -0.27 |
| Fossil energy use (million tons of sce) | 4,915 | -2.1 | 6,113 | -3.5 |
| Coal use (million tons) | 4,147 | -3.5 | 4,794 | -5.5 |
| Oil use (million tons) | 663 | 0.12 | 836 | -0.16 |
| Gas use (billion cubic meters) | 239,647 | 2.1 | 361,272 | -1.5 |
| Electricity (billion kWh) | 6,949 | -2.7 | 9,102 | -4.3 |
| Electricity price | 1.3 | | | 3.4 |
| CO2 emissions (inc cement; mil tons) | 11,236 | -2.3 | 13,073 | -4.0 |
| Cumulative CO2 (2015-2030) | | | 187,684 | -2.8 |
| Electricity generation share of total generation | 2020 | 2030 |
|--------------------------------------------------|------|------|
| Coal | 0.678| 0.657|
| Gas | 0.037| 0.037|
| Nuclear | 0.047| 0.058|
| Hydro | 0.157| 0.156|
| Other | 0.025| 0.030|
| Wind | 0.046| 0.048|
| Solar | 0.010| 0.014|
6. Conclusion
As part of the Paris Agreement, China has set for itself targets for reducing energy intensity by 60-65% in 2030 compared to 2005 levels. This study examines the economic consequences if the targets were met through a carbon pricing mechanism, where all fossil fuels are taxed in proportion to their carbon contents. The study uses a recursive dynamic computable general equilibrium model of the Chinese economy for the purpose this analysis.
The study finds that the carbon pricing policy would lead to a such a 65% reduction in intensity at a modest cumulative cost to GDP, which would be 0.7% lower than the base case by 2030, when revenues generated from the carbon pricing are recycled to the economy to cut existing value-added taxes. The goal of reducing the intensity by 60% is projected to cut GDP by only 0.1% relative to the base case by 2030. The carbon price not only reduces energy demand but also causes a shift away from more energy intensive industries to less energy intensive ones; it also causes a larger-scale deployment of renewable energy in electricity and heat generation.
A policy focused on promoting renewables in electricity generation using subsidies financed by a fossil fuel tax in proportion to their carbon contents would achieve the carbon reduction at a somewhat greater relative loss in GDP growth than a carbon pricing case. Putting more of the burden of adjustment on the power sector, when there are costs to a rapid ramp-up of hydro, nuclear, wind and solar power, results in a greater distortion of the price of electricity and a more difficult adjustment. A more careful study of the costs of resources going into these renewables is needed, including land, and suitable rivers.
As noted, this study does not attempt to model China’s actual plan to achieve its NDC, which includes both market and non-market measures. Instead, it focuses on the cost of achieving the NDC through a market mechanism, carbon pricing. Comparison of the economic consequences of the carbon pricing mechanism considered here with that of government’s actual plan of implementing the NDC is a natural extension of this study. Moreover, this study focuses only on CO₂ emissions from fossil fuel combustion, and does not include other GHG emissions from fossil fuel combustion and also GHG emissions from industrial processes and land use change. This is a limitation of the study to be addressed in future work.
References
Cao, Jing, Mun S. Ho and Dale Jorgenson. 2013. “The Economics of Environmental Policy in China” in Chris Nielsen and Mun Ho (eds.) Clearer Skies over China: Reconciling Air Quality, Climate and Economic Goals, MIT Press, Cambridge, MA.
Cao Jing and Mun S. Ho. 2014. “China: Economic and GHG emissions Outlook to 2050”. The Chinese version of this is in 重塑未来: 未来20 年的中国与世界, 清华大学能源环境经济研究所 (Reshaping the Future: China and the World in the next 20 years), Tsinghua University, Institute for Energy and Environmental Economics.
Dollar, David. 2015. “China’s rise as a regional and global power” Horizons - Journal of International Relations and Sustainable Development. Summer, Issue 4:162-172.
Financial Times. 2016. Hudson Lockett and Lucy Hornby. “China routes RMB 5tn into transport infrastructure,” May 11 2016.
International Energy Agency (IEA). 2010. Projected Costs of Generating Electricity, IEA and OECD, Paris.
International Energy Agency (IEA). 2014. World Energy Outlook 2014, IEA, Paris.
International Monetary Fund (IMF) (2014), World Economic Outlook, IMF, Washington, DC.
Nielsen, Chris and Mun Ho (eds.). 2013. Clearer Skies over China: Reconciling Air Quality, Climate and Economic Goals, MIT Press, Cambridge, MA.
Qi, Tianyu, Niven Winchester, Da Zhang, Xiliang Zhang and Valerie Karplus. 2014. The China-in-Global Energy Model. MIT JPSPGC Report No 262, May.
Sue Wing, Ian, Kathryn Daenzer, Karen Fisher-Vanden and Katharine Calvin. 2011. Phoenix Model Documentation. PNNL Joint Global Change Research Institute. Available at: http://www.globalchange.umd.edu/models/phoenix/
Vennemo, Haakon, Jianwu He and Shantong Li. 2014. Macroeconomic Impacts of Carbon Capture and Storage in China. Environmental and Resource Economics, 59(3):455-477.
World Bank and Development Research Center (DRC). 2013. China 2030. The World Bank, Washington DC and Development Research Center of the State Council, People’s Republic of China.
Appendix A: Economic-Environmental Model of China
In this appendix we describe the model for China in some detail, beginning with the modeling of each of the main economic agents. Then in section A.2 we describe the data and parameters underlying the model. A previous version of this model of the Chinese economy is used in Nielsen and Ho (2013) and here we describe the updates to it. This is a multi-sector model of economic growth where the main drivers of growth are population, total factor productivity growth and changes in the quality of labor and capital. It has a dynamic recursive structure, i.e. where investment is determined by fixed savings rate as in the Solow model. Consumption demand is driven by a translog household model that distinguishes demand by different demographic groups. The electricity sector is composed of 7 different generation technologies.
A.1 Structure of the Model
We discuss the five main actors in the economy in turn – producers, households, capital owners, government and foreigners. The electricity sector is described in greater detail explaining how the different generation technologies are allocated. For easy reference Table A1 lists variables which are referred to with some frequency. In general, a bar above a symbol indicates that it is a plan parameter or variable while a tilde indicates a market variable. Symbols without markings are total quantities or average prices. To reduce unnecessary notation, we include the time subscript, \( t \), only when necessary to note the time dependence.
A.1.1. Production
Each of the 33 industries is assumed to produce its output using a constant returns to scale technology. Except for electricity, for each sector \( j \) the output at time \( t \), \( Q_{j,t} \), is expressed as:
\[
Q_{j,t} = f(KD_j, LD_j, TD_j, A_{ij}, \ldots, A_{ij}, t),
\]
where \( KD_j \), \( LD_j \), and \( TD_j \), and \( A_{ij} \) are capital, labor, land, and intermediate inputs, respectively. In sectors for which both plan and market allocation exists, output is made up of two components,
\[ Q_I = Q_{I,j} + Q_{I,m}, \]
where \( Q_{I,j} \) and \( Q_{I,m} \) are quantities of plan and market output, respectively.
\[ Q_{I,j} = f(KD_j, LD_j, TD_j, A_{ij}, \ldots, A_{ij}, t), \]
where \( K_D, L_D, T_D, \text{ and } A_j \) are capital, labor, land, and intermediate inputs, respectively. In sectors for which both plan and market allocation exists, output is made up of two components,
\[ Q_{F,j} = Q_{F,j} + Q_{F,m}, \]
where \( Q_{F,j} \) and \( Q_{F,m} \) are quantities of plan and market output, respectively.
Each industry may produce more than one commodity and each commodity may be produced by more than one industry. In the language of the input output tables, we make use of both the Use and Make (or Supply) matrices.
---
13 \( Q_{I,j} \) denotes the quantity of industry \( j \)'s output. This is to distinguish it from, \( Q_{C,j} \), the quantity of commodity \( j \). Each industry may produce more than one commodity and each commodity may be produced by more than one industry. In the language of the input output tables, we make use of both the Use and Make (or Supply) matrices.
the plan quota output \((\overline{Q}_I)_j\) and the output sold on the market \((\overline{Q}_M)_j\). The plan quota output is
sold at the state-set price \((\overline{P}_I)_j\) while the output in excess of the quota is sold at the market price
\((\overline{P}_L)_j\). The \(P_I\) and \(Q_I\) names are chosen to reflect that these are domestic industry variables, as opposed to commodities \((P_C)\) or total supply \((P_S)\), the sum of domestic output and imports.
A more detailed discussion of how this plan-market formulation is different from standard market economy models is given in Garbaccio, Ho, and Jorgenson (1999). In summary, if the constraints are not binding, then the “two-tier plan/market” economy operates at the margin as a market economy with lump sum transfers between agents. The capital stock in each industry consist of two parts – the fixed capital, \(\overline{K}_j\), that is inherited from the initial period, and the market portion, \(\overline{K}_D\), that is rented at the market rate. The before-tax return to the owners of fixed capital in sector \(j\) is:
\[
\text{profit}_j = \overline{P}_I \overline{Q}_I + \overline{P}_L \overline{Q}_M - \overline{P}_K \overline{K}_D - \overline{P}_L \overline{L}_D - \overline{P}_T \overline{T}_D
- \sum_i \overline{P}_S \overline{A}_i - \sum_i \overline{P}_S \overline{A}_i.
\]
For each industry, given the capital stock \(\overline{K}_j\) and prices, the first order conditions from maximizing equation A2, subject to equation A1, determine the market and total input demands.
We represent the production structure with the cost dual, expressing the output price as a
function of input prices and an index of technology. The 3 primary factors and 33 intermediate
inputs for each industry are determined by a nested series of constant elasticity of substitution
(CES) functions taken from the GTAP model (version 7). The nest structure is given in Figure A1
and applies to all industries except electricity which is treated separately below.
At the top tier, output is a function of the primary factor–energy basket (VE) and the non-energy intermediate input basket (M), $Q_{jt} = f(VE_{jt}, M_{jt}, t)$. The VE basket is an aggregate of value added (VA) and the energy basket (E). Value added is a function of the 3 primary factors – capital (K), labor (L) and land (T). The energy aggregate is a CES function of coal, oil mining, gas mining, petroleum refining & coal products, electricity and gas commodities. The materials aggregate (M) is a Cobb-Douglas function of the 27 non-energy commodities.
For the top tier, the value equation and cost function are, respectively:
(A3) \[ PI_{jt}Q_{jt} = P_{jt}^{VE}VE_{jt} + PM_{jt}M_{jt} \]
(A4) \[ PI_{jt} = \frac{K_{jt}}{g_{jt}} \left[ \alpha_{Mjt}^{\sigma_{jt}^{Qjt}}PM_{jt}^{(1-\sigma_{jt}^{Qjt})} + \left(1 - \alpha_{Mjt}\right)^{\sigma_{jt}^{Qjt}}P_{jt}^{VE(1-\sigma_{jt}^{Qjt})} \right] \frac{1}{1-\sigma_{jt}^{Qjt}} \]
where $\alpha_{Mjt}$ is the weight for all non-energy inputs into industry $j$, and $1/\sigma_{jt}^{Qjt}$ is the elasticity of substitution between the two inputs. $g_{jt}$ is the index of the level of technology where a rising value indicates positive TFP growth and falling output prices. We assume that the change follows an
exponential function: \( \dot{g}_j(t) = A_j \exp(-\mu_j t) \). This implies technical change that is rapid initially, but gradually declines toward zero.
The primal function corresponding to the above dual cost is:
\[
QI_{jt} = \frac{g_{jt}}{k_{jt}^Q} \left[ \alpha_{Mji} M_{ji}^\sigma_{ji}^{Qj} + \left(1 - \alpha_{Mji}\right) VE_{ji}^\sigma_{ji}^{Qj} \right] \left(1 - \alpha_{Mji}\right) P_{ji}^{Qj} / P_{ji}^{VE}
\]
The input demands derived from the CES cost function are:
\[
VE_{jt} = \left(\frac{k_{jt}^Q}{g_{jt}}\right)^{1-\sigma_{ji}^{Qj}} \left(1 - \alpha_{Mji}\right)^{\sigma_{ji}^{Qj}} P_{ji}^{Qj} + \alpha_{Mji} PM_{jt}^{\sigma_{ji}^{Qj}} QI_{jt}
\]
\[
M_{jt} = \left(\frac{k_{jt}^Q}{g_{jt}}\right)^{1-\sigma_{ji}^{Qj}} \left[ \alpha_{Mji} PM_{jt}^{\sigma_{ji}^{Qj}} QI_{jt} \right]
\]
The weights for the CES functions are explained in Rutherford (2003) and Klump, McAdam and Willman (2011); these are calibrated using the base year values:
\[
\alpha_{Mj0} = \frac{PM_{j0} M_{j0}^{1/\sigma_{ji}^{Qj}}}{P_{j0}^{VE} VE_{j0}^{1/\sigma_{ji}^{Qj}}} ;
\]
\[
\frac{g_{j0}}{k_{j0}^Q} = \frac{QI_{j0}}{\left[ \alpha_{Mj0} M_{j0}^{\sigma_{ji}^{Qj}} + \left(1 - \alpha_{Mj0}\right) VE_{j0}^\sigma_{ji}^{Qj} \right] \left(1 - \alpha_{Mj0}\right) P_{j0}^{Qj} / P_{j0}^{VE}}
\]
The corresponding value, price and input demand equations for the primary factor-energy basket (VE) and the value-added basket (VA) are:
\[
P_{ji}^{VE} = P_{ji}^{VA} A_{ji} + PE_{ji} E_{jt}
\]
\[
P_{ji}^{VA} = P_{ji}^{KD} KD_{jt} + PL_{ji} LD_{jt} + PT_{ji} TD_{jt}
\]
\[
P_{ji}^{VE} = \frac{1}{\kappa_{ji}^{VE}} \left[ \alpha_{Ej}^{VE} P_{ji} E_{jt}^{1-\sigma_{ji}^{VE}} + \left(1 - \alpha_{Ej}^{VE}\right) P_{ji}^{VA} V_{ji}^{1-\sigma_{ji}^{VE}} \right]^{1-\sigma_{ji}^{VE}}
\]
\[
P_{ji}^{VA} = \frac{1}{\kappa_{ji}^{VA}} \left[ \alpha_{Kj}^{VA} P_{ji}^{KD} K_{ji}^{1-\sigma_{ji}^{VA}} + \alpha_{Lj}^{VA} P_{ji}^{PL} L_{jt}^{1-\sigma_{ji}^{VA}} + \left(1 - \alpha_{Lj}^{VA} - \alpha_{Kj}^{VA}\right) P_{ji}^{PT} T_{jt}^{1-\sigma_{ji}^{VA}} \right]^{1-\sigma_{ji}^{VA}}
The parameters of the value-added-energy and value-added nodes are calibrated to base year values in the following way:
\[
\alpha_{Ej0} = \frac{PE_{j0} E^{1/\sigma_{\mu}^{E}}}{P_{j0}^{VA} V^{1/\sigma_{\mu}^{VA}} + PE_{j0} E^{1/\sigma_{\mu}^{VA}}} ; \quad \kappa_{j0}^{VE} = VE_{j0} / \left[ \alpha_{Ej0} E^{\sigma_{\mu}^{E}} + \left( 1 - \alpha_{Ej0} \right) V^{\sigma_{\mu}^{VA}} \right]^{1/\sigma_{\mu}^{VA}}
\]
\[
\alpha_{Kj0} = \frac{P_{j0}^{KD} K^{1/\sigma_{\mu}^{VA}}}{P_{j0}^{KD} K^{1/\sigma_{\mu}^{VA}} + P_{j0}^{LD} L^{1/\sigma_{\mu}^{VA}} + P_{j0}^{TD} D^{1/\sigma_{\mu}^{VA}}} ; \quad \alpha_{Lj0} = \ldots ; \quad \alpha_{Tj0} = \ldots
\]
\[
\kappa_{j0}^{VA} = VA_{j0} / \left[ \alpha_{Kj0} K^{\sigma_{\mu}^{VA}} + \alpha_{Lj0} L^{\sigma_{\mu}^{VA}} + \left( 1 - \alpha_{Kj0} - \alpha_{Lj0} \right) D^{\sigma_{\mu}^{VA}} \right]^{1/\sigma_{\mu}^{VA}}
\]
Note that the cost functions for the sub-aggregates do not have an index of technology like the top tier; however, the share coefficients – \( \alpha_{Ej} \), \( \alpha_{Kjt} \), etc. – are allowed to change over time to reflect biases in technical change.
The energy basket equations give the demands for the 6 types of energy:
\[
PE_{j} = \sum_{k \in IE} PS_{k} A_{kj}
\]
\[
PE_{j} = \frac{1}{\kappa_{ji}^{E}} \left[ \sum_{k \in IE} \alpha_{kj}^{E} \left( 1 - \alpha_{kj}^{E} \right)^{1/\sigma_{\mu}^{E}} \right]^{1/\sigma_{\mu}^{E}} \quad \text{IE} = \{ \text{coal, oil, gasmine, refine, elect, gas} \}
\]
\[
A_{kj} = \left( \frac{1}{\kappa_{ji}^{E}} \right)^{1/\sigma_{\mu}^{E}} \left[ \alpha_{kj}^{E} \frac{PE_{j} E_{ji}}{PS_{k}} \right]^{\sigma_{\mu}^{E}} E_{j} \quad k \in IE
\]
The non-energy basket is a Cobb-Douglas function and the corresponding equations are:
\[
\alpha_{kjt}^E = \frac{PS_{kjt}^{1/\sigma_j^E}}{\sum_{i \in IE} PS_{ikt}^{1/\sigma_i^E}}; \quad \kappa_{j0}^E = E_{j0} / \left( \sum_{k \in IE} \alpha_{kjt}^E A_{kj0}^{\sigma_j^E - 1} \right)^{\sigma_j^E / \sigma_j^E - 1}
\]
We set the energy share \( \alpha_{Ej} \) to fall gradually over the next 40 years while the labor coefficient, \( \alpha_{Lj} \), rises correspondingly. The composition of the aggregate energy input \( E_j \) (i.e. the \( \alpha_{kj}^E \) coefficients) are also allowed to change over time.
The non-energy basket is a Cobb-Douglas function and the corresponding equations are:
(A12) \( \ln PM_{jt} = \sum_{k \in NE} \alpha_{kjt}^M \ln PS_{kt} \) \quad NE=\{agri, ..., services, admin\}
\[
PM_{jt} M_{jt} = \sum_{k \in NE} PS_{kt} A_{kjt}
\]
\[
A_{kjt} = \alpha_{kjt}^M \frac{PM_{jt} M_{jt}}{PS_{kt}} \quad k \in NE
\]
The price to buyers of industry output includes the indirect tax on output, the externality ad-valorem tax, the externality unit tax:
(A13) \( PI_i^t = (1 + t_i^t + t_i^{ux}) PI_i + t_i^{uu} \)
A carbon tax on coal, e.g., is represented by \( t_{coal}^{uu} \).
Industries versus Commodities
The model distinguishes industries from commodities as in the official Use and Make input-output tables. Each industry may make a few commodities and each commodity may be made by a few industries; e.g. the Refining industry produces Refining commodity and Chemical commodity, and the Chemical commodity comes from Refining, Chemical, Primary Metal and other industries. The quantity of domestic commodity is denoted \( QC \) and its price \( PC \); the sum of column \( i \) in the Make matrix gives the value of commodity \( i \), and the sum of row \( j \) is the industry output value. The relation between commodity and industry output and prices are written as:
(A14) \( VQC_i = PC_i QC_i = \sum_j m_{ji} PI_i Q_i \)
\[
\ln PC_i = \sum_j m_{ji}^r \ln PI_j^i
\]
where \( m_{ji}^r \) is the row share and \( m_{ji}^c \) is the column share.
A.1.2. The electricity sector
In version 17 of the China Model we disaggregate the electricity sector into different generation technologies unlike the previous versions that represent the output of electricity using the common production function (A.4b) above. The production and input structure of this sector is illustrated in Figure A2; this consists of a nested structure of CES functions. At the top tier, electricity output is an aggregate of Transmission & Distribution and Electricity Generation. The price of electricity output (sector 22) is a function of the price of transmission \( P_{j=22,TT}^{i} \) and the price of generation \( P_{j=22,G}^{i} \):
\[
(A.15) \quad P_{jj}^{i} = \frac{\kappa_{ji}^{QI}}{g_{ji}} \left[ \alpha_{Eij}^{\sigma_{i}^{ED}} P_{j=22,TT}^{i} \left( 1-\sigma_{i}^{ED} \right) + \left( 1-\alpha_{Eij}^{\sigma_{i}^{ED}} \right) \sigma_{i}^{ED} P_{j=22,G}^{i} \left( 1-\sigma_{i}^{ED} \right) \right] \frac{1}{1-\sigma_{i}^{ED}}
\]
\( j=22(\text{elec}) \)
The superscript \( tt \) denotes that this is a price inclusive of indirect business taxes that are levied at the level of the generating sectors and transmission. This is explained in (A21b) below. The quantity and value equations are:
\[
(A.16) \quad Q_{ij}^{Ei} = \left( \frac{\kappa_{ji}^{QI}}{g_{ji}} \right)^{1-\sigma_{i}^{ED}} \left[ \alpha_{Eij}^{\sigma_{i}^{ED}} \frac{P_{j=22,TT}^{i}}{P_{j=22,G}^{i}} \right]^{\sigma_{i}^{ED}} Q_{jj}^{i}
\]
\[
Q_{ij}^{TD} = \left( \frac{\kappa_{ji}^{QI}}{g_{ji}} \right)^{1-\sigma_{i}^{ED}} \left[ \left( 1-\alpha_{Eij}^{\sigma_{i}^{ED}} \right) \frac{P_{j=22,TT}^{i}}{P_{j=22,G}^{i}} \right]^{\sigma_{i}^{ED}} Q_{jj}^{i}
\]
\[
P_{j=\text{elec},TT}^{i} Q_{j=\text{elec},TT}^{i} = P_{ij}^{Ei} Q_{ij}^{Ei} + P_{ij}^{TD} Q_{ij}^{TD}
\]
At the Generation node we assume that this consists of Base load sources and Renewables (intermittent) in a way similar to the C-GEM model (Qi et al 2014). The price of Electricity Generation is thus a function of the price of Base load Electricity ($P_{Et}^{BL}$) and price of Renewable Electricity ($P_{Et}^{RE}$):
$$P_{t}^{EG} = \frac{1}{k_{t}^{EG}} \left[ \alpha_{BLt}^{EG} P_{Et}^{BL(1-\sigma_{t}^{EG})} + \left(1 - \alpha_{BLt}^{EG}\right) \sigma_{t}^{EG} P_{Et}^{RE(1-\sigma_{t}^{EG})} \right] \frac{1}{1-\sigma_{t}^{EG}}$$
(A.17)
When $\sigma_{t}^{EG} = 1$, this simplifies to:
$$P_{t}^{EG} = \frac{1}{k_{t}^{EG}} P_{Et}^{BL_{t}^{BL}} P_{Et}^{RE(1-\alpha_{BLt})}$$
Figure A2. Structure of electricity sector
Figure A2. Structure of electricity sector
(b) Transmission and Renewables structure
The quantities of Base load output and Renewable output are:
\[
Q_{Et}^{BL} = \left( \frac{1}{\kappa_i^E} \right)^{1-\sigma_i^E} \left[ \alpha_{BLt} \frac{P_i^E}{P_{Et}^{BL}} \right] Q_i^E
\]
(A.18)
\[
Q_{Et}^{RE} = \left( \frac{1}{\kappa_i^E} \right)^{1-\sigma_i^E} \left[ (1-\alpha_{BLt}) \frac{P_i^E}{P_{Et}^{RE}} \right] Q_i^E
\]
\[
P_t^{RE} Q_t^{RE} = P_{Et}^{BL} Q_{Et}^{BL} + P_{Et}^{RE} Q_{Et}^{RE}
\]
Renewables here consist only of Wind and Solar which are intermittent sources and requires either parallel storage capacities, or conventional backup. We thus assume that such electricity is imperfectly substitutable with base load sources and specify an elasticity of substitution, \( \sigma_i^E \), in a way similar to Qi et al. (2014) for our main parameter value of 1.0.\(^{14}\) All other sources of electricity contribute to the Base load aggregate with a high elasticity of substitution, \( \sigma_i^{BL} = 4.\)\(^{15}\) In the base year, these sources include conventional coal, gas, hydro, nuclear and a minor “other” (oil, biomass, geothermal, etc.) In the future years we allow the options of coal with CCS and gas with CCS. The price of Base load electricity is thus a function of the component prices \( P_{t,EGEN}^{GEN} \),
\[
P_{t,EGEN}^{GEN} = P_{t,GEN}^{coal} + P_{t,GEN}^{gas} + P_{t,GEN}^{nuclear} + P_{t,GEN}^{hydro} + P_{t,GEN}^{other} + P_{t,GEN}^{coal-CCS} + P_{t,GEN}^{gas-CCS}
\]
(A.19)
\[
l=coal, gas, nuclear, hydro, other, coal-ccs, gas-ccs
\]
The price variables have a \( tt \) superscript to denote that they are inclusive of (net) output taxes and subsidies. The value equation and quantities of the various Base load technologies are:
\[
P_{Et}^{BL} Q_{Et}^{BL} = \sum_{l \in BL} P_{lt}^{GEN} Q_{lt}^{GEN}
\]
(A.20)
\(^{14}\) In the Phoenix model (Sue Wing et al 2011), the elasticity of substitution between “peak load” (which includes wind and solar) and “base load” sources is also 1.
\[ Q_{lt}^{GEN} = \left( \frac{1}{\kappa_i^{BL}} \right)^{1-\sigma^{BL}} \left[ \alpha_{lt}^{BL} \frac{P_{lt}^{BL}}{P_{lt}^{rt,GEN}} \right]^{\sigma^{BL}} Q_{Et}^{BL} \]
For the intermittent renewable aggregate we only identify two types in this model: wind and solar (the others are part of the miscellaneous “other” in the base load tier). We assume that wind and solar are close, but not perfect, substitutes with an elasticity \( \sigma^{RE} = 4 \). This is the elasticity chosen in Sue Wing et al. (2011, p 31). The equations for the renewable tier are:
\[
P_{lt}^{RE} = \frac{1}{\kappa_i^{RE}} \left[ \alpha_{wind,t}^{RE} P_{wind,t}^{GEN(1-s^{RE})} \right]^{1-\sigma^{RE}} \left[ \left( 1 - \alpha_{wind,t}^{RE} \right) P_{solar,t}^{GEN(1-\sigma^{RE})} \right]^{\sigma^{RE}} \left( 1 - \sigma^{RE} \right)
\]
(A.21)
\[
Q_{wind,t}^{GEN} = \left( \frac{1}{\kappa_i^{RE}} \right)^{1-\sigma^{RE}} \left[ \alpha_{wind,t}^{RE} \frac{P_{lt}^{RE}}{P_{wind,t}^{RE}} \right]^{\sigma^{RE}} Q_{Et}^{RE}
\]
\[
Q_{solar,t}^{GEN} = \left( \frac{1}{\kappa_i^{RE}} \right)^{1-\sigma^{RE}} \left( 1 - \alpha_{wind,t}^{RE} \right) \frac{P_{lt}^{RE}}{P_{solar,t}^{RE}} \left( \alpha_{wind,t}^{RE} \right) Q_{Et}^{RE}
\]
\[
P_{lt}^{RE} Q_{lt}^{RE} = P_{wind,t}^{lt,GEN} Q_{wind,t}^{GEN} + P_{solar,t}^{lt,GEN} Q_{solar,t}^{GEN}
\]
The price to the purchasers of such electricity is \( P_{lt}^{rt,GEN} \) which includes the net output tax, \( t_{lt,t}^{EL} \) and externality tax, \( t_{lt,t}^{ELxu} \). The prices to the producers are net of this tax:
(A21b) \[ P_{lt}^{rt,GEN} = (1 + t_{lt,t}^{EL}) P_{lt}^{GEN} + t_{lt,t}^{ELxu} \]
The input demand structure for coal generation is given in Figure A2; at the top tier coal power is produced by a low elasticity CES function of value-added-energy (VE) and non-energy intermediates (M).\(^ {16} \) Productivity growth in this sector is represented by the \( g_{ct}^{BL} \) term in the price function (A22). The VE bundle is a CES function of value-added (VA) and energy (E) with \( \sigma^{VE} = 0.5 \), while the VA node has an elasticity \( \sigma^{VA} \) of 1.0 between capital and labor. The Energy aggregate is a function of coal and non-coal energy which is a small item that includes electricity and refined petroleum products (lubricants and vehicle fuels) as described in the data appendix. We set the elasticity between them to a low value (\( \sigma^{E} \)
\(^{15} \) Our specification of base load and renewables follows EPPA-4, which assumes perfect substitution among the base load sources. We have, however, chosen to use an elasticity of 4 as used in the Phoenix model; in a similar setup, Vennemo et al (2014) use an elasticity of 20.
\(^{16} \) In Qi et al. (2014) this is set to be a Leontief function, here we use the low general elasticity between materials and value-added-energy in GTAP of 0.15.
The non-coal quantity is an aggregate of only electricity and refined oil since the other energy inputs (oil mining, gas) are zero.
The input structure for gas generation is similar to the one for coal for the top tiers for output price, price of VE, price of VA and price of M. The equations for the top tiers, in terms of the producer prices, are:
\[
P_{E,ct}^{\text{GEN}} = \frac{k_{ct}^{\text{BL}}}{g_{ct}^{\text{BL}}} \left( \alpha_{M,ct}^{\text{BLc}} \sigma_{BL,ct}^{(1-\sigma_{BL,ct})} P_{M,ct}^{\text{BLL} \left(1-\sigma_{BL,ct}^{M}\right)} \right) + \left( 1 - \alpha_{M,ct}^{\text{BLc}} \right) \sigma_{BL,ct}^{M} P_{BL,ct}^{VE \left(1-\sigma_{BL,ct}^{M}\right)} \left( 1 - \sigma_{BL,ct}^{M} \right) \tag{A.22} \]
\[
P_{E,ct}^{\text{GEN}} Q_{ct}^{\text{GEN}} = P_{M,ct}^{\text{BL}} M_{ct}^{\text{BL}} + P_{ct}^{VE} V_{ct}^{\text{BL}} ; \quad c=\text{coal, coal_ccs, gas, gas_ccs} \]
\[
V_{E,ct}^{BL} = \left( \frac{k_{ct}^{\text{BL}}}{g_{ct}^{\text{BL}}} \right)^{1-\sigma_{BL,ct}^{M}} \left[ \left( 1 - \alpha_{M,ct}^{\text{BLc}} \right) \frac{P_{ct}^{BL}}{P_{ct}^{VE}} \right]^{\sigma_{BL,ct}^{M}} Q_{ct}^{\text{GEN}} \]
\[
M_{ct}^{BL} = \left( \frac{k_{ct}^{\text{BL}}}{g_{ct}^{\text{BL}}} \right)^{1-\sigma_{BL,ct}^{M}} \left[ \alpha_{M,ct}^{\text{BLc}} \frac{P_{ct}^{BL}}{P_{ct}^{VE}} \right]^{\sigma_{BL,ct}^{M}} Q_{ct}^{\text{GEN}} \]
\[
\ln PM_{ct}^{BL} = \sum_{k \in NE} \alpha_{kM,ct}^{BLc} \ln PS_{kt} \tag{A.23} \]
\[
A_{k,ct}^{BLc} = \alpha_{kM,ct}^{BLc} \frac{PM_{ct}^{BL}}{M_{ct}^{BL}} / PS_{kt} \]
\[
P_{BL,ct}^{VE} = \frac{1}{k_{ct}^{VE}} \left[ \alpha_{ct}^{BLc} \sigma_{BL,ct}^{VE} P_{ct}^{BL \left(1-\sigma_{BL,ct}^{VE}\right)} + \left( 1 - \alpha_{ct}^{BLc} \right) \sigma_{BL,ct}^{VE} P_{BL,ct}^{VE \left(1-\sigma_{BL,ct}^{VE}\right)} \right] \left( 1 - \sigma_{BL,ct}^{VE} \right) \tag{A.24} \]
\[
P_{BL,ct}^{VE} V_{BL,ct} = P_{BL,ct}^{VE} V_{BL,ct} E_{BL,ct} + P_{BL,ct}^{VE} E_{BL,ct} \]
\[
V_{A,ct}^{BL = \left( \frac{1}{k_{ct}^{VE}} \right)^{1-\sigma_{BL,ct}^{VE}} \left[ \left( 1 - \alpha_{ct}^{BLc} \right) \frac{P_{BL,ct}^{VE}}{P_{BL,ct}^{VE}} \right]^{\sigma_{BL,ct}^{VE}} V_{BL,ct} } \]
\[
E_{c,t}^{BL} = \left( \frac{1}{\kappa_{c,t}^{VE}} \right)^{1-\sigma_{BL}^{VE}} \left[ \alpha_{Et}^{BL} \frac{P_{c,t}^{VE}}{P_{c,t}^{EL}} \right]^{\sigma_{BL}^{VE}} VE_{c,t}^{BL}
\]
\[
P_{c,t}^{VA} = \frac{1}{\kappa_{c,t}^{VA}} \left[ \alpha_{Kr}^{VA} \frac{P_{c,t}^{VA}}{P_{c,t}^{KD}} P_{c,t}^{KD} \right]^{\sigma_{VA}^{VA}} \left[ \frac{1}{1-\sigma_{BL}^{VA}} \right]
\]
\[
\begin{align*}
K_{c,t}^{BL} &= \left( \frac{1}{\kappa_{c,t}^{VA}} \right)^{1-\sigma_{VA}^{BL}} \left[ \alpha_{Kr}^{VA} \frac{P_{c,t}^{VA}}{P_{c,t}^{KD}} \right]^{\sigma_{VA}^{BL}} V_{c,t}^{BL} \\
LD_{c,t}^{BL} &= \left( \frac{1}{\kappa_{c,t}^{VA}} \right)^{1-\sigma_{VA}^{BL}} \left[ (1-\alpha_{Kr}^{VA}) \frac{P_{c,t}^{VA}}{P_{c,t}^{PL}} \right]^{\sigma_{VA}^{BL}} V_{c,t}^{BL}
\end{align*}
\]
For the energy tier, the equations for coal and noncoal (NC) inputs in sector \( c \) are:
\[
P_{c,t}^{BL} = \frac{1}{\kappa_{c,t}^{Coal}} \left[ \alpha_{Et}^{Coal} \frac{E_{c,t}^{Coal}}{E_{c,t}^{VE}} \right]^{\sigma_{Coal}^{Ecoal}} \left[ \frac{1}{1-\sigma_{BL}^{Coal}} \right]
\]
\[
P_{c,t}^{NC} = \frac{1}{\kappa_{c,t}^{Coal}} \left[ \alpha_{Et}^{NC} \frac{E_{c,t}^{NC}}{E_{c,t}^{VE}} \right]^{\sigma_{NC}^{Ecoal}} \left[ \frac{1}{1-\sigma_{BL}^{NC}} \right]
\]
The elasticities of substitution are summarized in Table A2. The top tier is Leontief between Materials and the VE bundle (\( \sigma_{BL}^{VE} = 0 \)). The substitution between energy and value added (\( \sigma_{BL}^{VA} \)) is set
at 0.5, the value used in the GTAP model.\footnote{The EPPA (Paltsev et al, Table 3) model uses an elasticity of 0.4-0.5; Qi et al. (2014) uses 0.1 and Phoenix (Sue Wing, Fig 2) uses 0.25.} Labor input is a very small share in the electricity sector and we set $\sigma_{VA}^{BL,c} = 1$ following EPPA and Qi et al. (2014). This model explicitly recognizes the small amount of energy used besides the main fuel source unlike the other models mentioned so far; in the case of coal, the non-coal inputs include refined petroleum, electricity and gas. The substitution between the main fuel and the small non-coal energy bundle is set at 0.25, this is similar in spirit to the value of energy-capital substitution used in Phoenix for electricity generation. The substitution among the components of non-coal energy is set at 0.5, the general elasticity for energy inputs in GTAP and EPPA.
The input structure for gas-fired power plants is similar to that for coal, except that in the bottom tier for energy, gas inputs (GS) are aggregated with non-gas (NG; electricity and refined petroleum). The value equations for the gas and gas_ccs nodes are:
\begin{equation}
(P_{EGEN}^{G}_{g,t}, Q_{EGEN}^{G}_{g,t}) = PM_{g,t}^{BL} M_{g,t}^{BL} + P_{VE}^{BL}_{g,t} V_{g,t}^{BL}, \quad g = \text{gas, gas_ccs}
\end{equation}
\begin{equation}
PM_{g,t}^{BL} M_{g,t}^{BL} = \sum_{k \in NE} PS_{k,t} A_{g,k,t}^{BL}
\end{equation}
\begin{equation}
P_{VE}^{BL}_{g,t} V_{g,t}^{BL} = P_{VA}^{BL}_{g,t} V_{g,t}^{BL} + P_{VE}^{BL}_{g,t} E_{g,t}^{BL}
\end{equation}
\begin{equation}
P_{VA}^{BL}_{g,t} V_{g,t}^{BL} = P_{KD}^{BL}_{g,t} K_{g,t}^{BL} + P_{PL}^{BL}_{g,t} L_{g,t}^{BL}
\end{equation}
\begin{equation}
P_{E}^{BL}_{g,t} E_{g,t}^{BL} = P_{GS}^{GS}_{g,t} Q_{g,t}^{GS}_{BL} + P_{NG}^{NG}_{g,t} Q_{g,t}^{NG}_{BL}
\end{equation}
\begin{equation}
P_{GS}^{GS}_{g,t} Q_{g,t}^{GS}_{BL} = \sum_{k \in NG} PS_{k,t} A_{k,t}^{BL}
\end{equation}
The price functions of the energy aggregate, the gas aggregate and the non-gas aggregate are:
\begin{equation}
P_{E}^{BL}_{g,t} = \frac{1}{\kappa_{E,g,t}^{BL}} \left[ \alpha_{E,t}^{BL} \sigma_{E,g,t}^{BL} P_{E,t}^{BL} + (1 - \alpha_{E,t}^{BL}) \sigma_{E,g,t}^{BL} P_{NG,t}^{NG} \right] \frac{1}{1 - \sigma_{E,g,t}^{BL}}
\end{equation}
(A29a)
\begin{equation}
Q_{g,t}^{GS} = \left( \frac{1}{\kappa_{E,g,t}^{BL}} \right)^{1 - \sigma_{E,g,t}^{BL}} \left[ \alpha_{E,t}^{GS} \frac{P_{E,t}^{GS}}{\sigma_{E,g,t}^{GS}} \right] E_{BL}^{g,t}
\end{equation}
(A29b)
\begin{equation}
P_{GS}^{GS}_{g,t} = \frac{1}{\kappa_{GS}^{BL,t}} \left[ \alpha_{GS,g,t}^{GS} \sigma_{GS,g,t}^{GS} P_{GS,t}^{GS} + (1 - \alpha_{GS,g,t}^{GS}) \sigma_{GS,g,t}^{GS} P_{gasprod,t}^{GS} \right] \frac{1}{1 - \sigma_{GS,g,t}^{GS}}
\end{equation}
\footnote{The EPPA (Paltsev et al, Table 3) model uses an elasticity of 0.4-0.5; Qi et al. (2014) uses 0.1 and Phoenix (Sue Wing, Fig 2) uses 0.25.}
The values of the substitution elasticities follow that of coal and are given in Table A2.
We also project fossil fuel technologies that will be cost competitive in the future with a high carbon price: coal integrated gasification with carbon capture (IGCC) and natural gas with carbon capture (NGCC). The IGCC tier structure is given in Figure A2; the NGCC structure is identical. It is the same as the coal tier structure except that the energy node is replaced by a “fuel and sequestration” node. This structure follows Sue Wing et al. (2011, p33) where the fuel and sequestration technology in combined in a Leontief function, and the sequestration input is a fixed factor resource with an upward sloping supply curve.
The equations for $P_{c=coalccs,t}^{BL}$ are the same as (A.22), and (A.23) for $P_{c=coalccs,t}^{M}$, (A.25) for $P_{c=coalccs,t}^{VA}$, (A.26) for $P_{c=coalccs,t}^{BL}$, and (A.27) for $P_{c=coalccs,t}^{NC}$. The price for the fuel-sequestration sub-aggregate is given, in general, by:
$$A_{natgas,g,t}^{BL} = \left( \frac{1}{\kappa_{BL,g,t}^{GS}} \right)^{1-\sigma_{BL,g,t}^{GS}} \left[ \Omega_{natgas,gt}^{GSgas} \left( \frac{P_{BL,gt}^{GS}}{P_{S, natgas, t}^{GS}} \right)^{\sigma_{BL,g,t}^{GS}} \left( \frac{1}{\kappa_{BL,g,t}^{GS}} \right) \right] Q_{BL,g,t}^{GS}$$
$$A_{gasprod,g,t}^{BL} = \left( \frac{1}{\kappa_{BL,g,t}^{GS}} \right)^{1-\sigma_{BL,g,t}^{GS}} \left[ \left( 1 - \Omega_{natgas,gt}^{GSgas} \right) \left( \frac{P_{BL,gt}^{GS}}{P_{S, gasprod, t}^{GS}} \right)^{\sigma_{BL,g,t}^{GS}} \left( \frac{1}{\kappa_{BL,g,t}^{GS}} \right) \right] Q_{BL,g,t}^{GS}$$
$$A_{k,g,t}^{BL} = \left( \frac{1}{\kappa_{BL,g,t}^{NG}} \right)^{1-\sigma_{BL,g,t}^{NG}} \left[ \alpha_{k, BL,g,t}^{NG} \left( \frac{P_{BL,gt}^{NG}}{P_{S, k}^{NG}} \right)^{\sigma_{BL,g,t}^{NG}} \left( \frac{1}{\kappa_{BL,g,t}^{NG}} \right) \right] Q_{BL,g,t}^{NG}$$
$$(A29c) P_{BL,g,t}^{NG} = \frac{1}{\kappa_{BL,g,t}^{NG}} \left[ \sum_{k \in NG} \alpha_{k, BL,g,t}^{NG} \left( \frac{P_{BL,gt}^{NG}}{P_{S, k}^{NG}} \right)^{\sigma_{BL,g,t}^{NG}} \left( \frac{1}{\kappa_{BL,g,t}^{NG}} \right) \right]$$
In the case of a Leontief function with $\sigma_{coalccs}^{seq} = 0$, the equation is simply:
$$(A26c) P_{coalccs,t}^{ESEQ} = \left[ \alpha_{coalccs,t}^{seq} P_{E_{coalccs,t}} + \alpha_{coalccs,t}^{seq} P_{SEQ_{coalccs,t}} \right]$$
$$E_{coalccs,t} = \alpha_{coalccs,t}^{E_{SEQ}} Q_{coalccs,t}^{ESEQ} \quad Q_{coalccs,t}^{SEQ} = \alpha_{coalccs,t}^{SEQ} Q_{coalccs,t}$$
where we may normalize the units of the fuel-sequestration bundle to be the same as the fuel units,
\( \alpha_{coal_{ccs}}^{E, seq} = 1 \) and the sequestration coefficient \( \alpha_{coal_{ccs}}^{seq} \) reflects to addition cost for this technology, per unit fuel.
The price of VE in the ccs sectors is then an aggregate of the \( P_{coal_{ccs}, t}^{VA} \) and \( P_{ccs, t}^{ESEQ} \):
\[
(P_{ccs, t}^{VE}) = \frac{1 - \sigma_{BLc, t}}{1 - \sigma_{BLc}} \left[ \alpha_{BLc, t}^{ESEQ} P_{ccs, t}^{ESEQ} (1 - \sigma_{BLc}) \right] \frac{1}{1 - \sigma_{BLc}}
\]
\( c = \text{coal}_{ccs}, \text{gas}_{ccs} \)
The sequestration resource supply is given by:
\[
Q_{ccs, t}^{ESEQ} = \left( \frac{1}{1 - \sigma_{BLc, t}} \right) \left[ \alpha_{BLc, t}^{ESEQ} P_{ccs, t}^{ESEQ} \right]^{\sigma_{BLc, t}} V_{E, t}^{BLc} \quad c = \text{coal}_{ccs}, \text{gas}_{ccs}
\]
A parallel set of equations hold for NGCC with the price of fuel-sequestration in NGCC given by:
\[
P_{gas_{ccs}, t}^{ESEQ} = \left[ \alpha_{gas_{ccs}, t}^{E, seq} P_{gas_{ccs}, t}^{VE} + \alpha_{gas_{ccs}}^{seq} P_{gas_{ccs}, t}^{ESEQ} \right]
\]
\( E_{gas_{ccs}, t} = \alpha_{gas_{ccs}}^{E, seq} Q_{gas_{ccs}, t}^{ESEQ} \), \( Q_{ESEQ, t} = \alpha_{gas_{ccs}}^{seq} Q_{gas_{ccs}, t}^{ESEQ} \)
The modeling of nuclear power supply is always treated specially in the models cited given its unusual nature. While there is no obvious constraint like the availability of rivers for hydro power, it is recognized that actual construction of nuclear plants has been difficult with substantial opposition by those worried about safety. Vennemo et al. (2014) uses an upward sloping supply function due to “the political suitability of different locations.” Sue Wing et al. (2014) also use a “fixed factor” in their specification of nuclear power, but justify it by saying that a “supply curve is used to parameterize the mining and milling of resources to produce the fuel rods.” EPPA-4 also has a fixed factor in nuclear generation but interprets it as a stock of knowledge that builds over time with cumulative output.
We follow the logic of Vennemo at al., and the Phoenix model, and require a non-reproducible resource input for nuclear power. In the second tier (see Figure A2), the VR bundle is an aggregate of resource \( R^{NUCL} \) and value-added-energy (VE) with an elasticity of \( \sigma_{VR} = 0.4, \) following the EPPA-4.
value (Phoenix uses a top elasticity of 0.5). We follow Vennemo et al. in using a nuclear resource supply elasticity of 2.5.
The specification of the hydro power production function is another complicated matter. Rivers suitable for hydro power is an obvious input, and one might think of the stock of such water resources and the cost of using the marginal river. If the next unused water source requires a more costly structure (or implementation costs including relocation of people) than existing dams, then one may represent this as a production function with a lower TFP factor, or as a paying more for a fixed quantity of effective water resource \( R \). For simplicity we have chosen the latter approach and set up the hydro function like that of nuclear, with a rising supply curve for the water resource input, \( R^{HYDR} \). In the same spirit, our wind and solar output functions follow those of nuclear and hydro with specific Resource inputs. This is illustrated in Figure A2(b). The parameterization of the resource input shares follows that given in EPPA 4 (Paltsev et al. 2005 Table 11); the resource share for wind and solar is set at 0.05 and we assume the resource share for hydro is also 0.05 of total gross output.
The cost functions for the renewables – nuclear, hydro, wind, solar and “other” – are the same. The following equations are for the sub-aggregates – output price \( P^{EGEN}_{b,t} \), intermediate input price \( P^{BL}_{b,t} \), value-added-resource price \( P^{VR}_{BL,b,t} \), value-added-energy price \( P^{VE}_{BL,b,t} \), the energy price \( P^{BL}_{E} \):
\[
P^{EGEN}_{b,t} = \frac{K_{bt}^B}{g_{bt}} \left[ \alpha_{b,t}^{Bb} \sigma_{bt}^{M,b} P^{BL}_{b,t} (1 - \sigma_{bt}^{M,b}) + \left( 1 - \alpha_{b,t}^{Bb} \right) \sigma_{bt}^{M,b} P^{VR}_{b,t} \right]^{1/(1-\sigma_{bt}^{M,b})}
\]
\[
P^{EGEN}_{b,t} (\ln g_{bt}) = P^{BL}_{b,t} (\ln M_{b,t}) + P^{VR}_{b,t} (\ln VR_{b,t})
\]
\[
M^{B}_{b,t} = \left( \frac{K_{bt}^B}{g_{bt}} \right)^{1-\sigma_{bt}^{M,b}} \left[ \alpha_{b,t}^{Bb} \frac{P^{EGEN}_{b,t}}{P^{BL}_{b,t}} \right]^{\sigma_{bt}^{M,b}} Q^{EGEN}_{b,t}
\]
\[
VR^{B}_{b,t} = \left( \frac{K_{bt}^B}{g_{bt}} \right)^{1-\sigma_{bt}^{M,b}} \left[ (1 - \alpha_{b,t}^{Bb}) \frac{P^{EGEN}_{b,t}}{P^{VR}_{b,t}} \right]^{\sigma_{bt}^{M,b}} Q^{EGEN}_{b,t}
\]
\[
\ln P^{BL}_{b,t} = \sum_{k \in NE} \alpha_{b,t}^{kM} \ln PS_{k,t}
\]
\[
P^{BL}_{b,t} M^{B}_{b,t} = \sum_{k \in NE} PS_{k,t} A^{B}_{b,t,k} ; \quad A^{B}_{b,t,k} = \alpha_{b,t}^{CB} P^{BL}_{b,t} M^{B}_{b,t} / PS_{k,t}
\]
\[ P_{Bb,t}^{VR} = \frac{1}{\kappa_{Bb,t}^{VR}} \left[ \alpha_{Rt}^{Bb} \sigma_{Bb}^{VR} P_{Bb,t}^{(1-\sigma_{Bb}^{VR})} + (1 - \alpha_{Rt}^{Bb}) \sigma_{Bb}^{VR} P_{Bb,t}^{(1-\sigma_{Bb}^{VR},E)} \right] \frac{1}{1 - \sigma_{Bb}^{VR}} \]
(A.32)
\[ P_{Bb,t}^{VR} R_{Bb,t}^{VR} = P_{Bb,t}^{VE} E_{Bb,t} + PR_{Bb,t}^{B} \]
\[ R_{Bb}^{B} = \left( \frac{1}{\kappa_{Bb,t}^{B}} \right)^{1-\sigma_{Bb}^{VR}} \left[ \alpha_{Rt}^{Bb} \frac{P_{Bb,t}^{VR}}{PR_{Bb,t}^{B}} \right]^{\sigma_{Bb}^{VR}} VR_{Bb,t}^{B} \]
\[ V_{E_{Bb,t}}^{B} = \left( \frac{1}{\kappa_{Bb,t}^{B}} \right)^{1-\sigma_{Bb}^{VE}} \left[ (1 - \alpha_{Rt}^{Bb}) \frac{P_{Bb,t}^{VR}}{PR_{Bb,t}^{B}} \right]^{\sigma_{Bb}^{VE}} V_{Bb,t}^{B} \]
\[ E_{Bb,t}^{B} = \left( \frac{1}{\kappa_{Bb,t}^{B}} \right)^{1-\sigma_{Bb}^{VE}} \left[ \alpha_{E_t}^{Bb} \frac{P_{Bb,t}^{VR}}{PE_{Bb,t}^{B}} \right]^{\sigma_{Bb}^{VE}} V_{Bb,t}^{B} \]
\[ VA_{Bb,t}^{B} = \left( \frac{1}{\kappa_{Bb,t}^{B}} \right)^{1-\sigma_{Bb}^{VE}} \left[ (1 - \alpha_{Bb}^{Bb}) \frac{P_{Bb,t}^{VE}}{P_{A_{Bb,t}}^{B}} \right]^{\sigma_{Bb}^{VE}} V_{Bb,t}^{B} \]
\[ P_{Bb,t}^{VA}^{A} = \frac{1}{\kappa_{Bb,t}^{VA}} \left[ \alpha_{Kt}^{Bb} \sigma_{Bb}^{VA} P_{A_{Bb,t}}^{KD(1-\sigma_{Bb}^{VA})} + (1 - \alpha_{Kt}^{Bb}) \sigma_{Bb}^{VA} P_{Bb,t}^{(1-\sigma_{Bb}^{VA})} \right] \frac{1}{1 - \sigma_{Bb}^{VA}} \]
(A.34)
\[ P_{Bb,t}^{VA} A_{Bb,t}^{D} = P_{Bb,t}^{KD} K_{Bb,t}^{D} + P_{D_{Bb,t}}^{BL} L_{D_{Bb,t}} \]
\[ K_{Bb,t}^{D} = \left( \frac{1}{\kappa_{Bb,t}^{VA}} \right)^{1-\sigma_{Bb}^{VA}} \left[ \alpha_{Rt}^{Bb} \frac{P_{Bb,t}^{VA}}{P_{D_{Bb,t}}^{KD}} \right]^{\sigma_{Bb}^{VA}} V_{A_{Bb,t}}^{B} \]
\[ L_{D_{Bb,t}} = \left( \frac{1}{\kappa_{Bb,t}^{VA}} \right)^{1-\sigma_{Bb}^{VA}} \left[ \alpha_{P_{Bb,t}^{VA}} \frac{P_{Bb,t}^{VA}}{PL_{Bb,t}^{D}} \right]^{\sigma_{Bb}^{VA}} V_{A_{Bb,t}}^{B} \]
\[ PE_{Bb,t}^{E_{Bb,t}} = \frac{1}{\kappa_{Bb,t}^{E}} \sum_{k \in IE} \alpha_{E_{Bb,t}}^{k} \sigma_{k}^{E_{Bb,t}} P_{S_{k_{t}}}^{(1-\sigma_{Bb}^{E_{Bb,t}})} \frac{1}{1 - \sigma_{Bb}^{E_{Bb,t}}} \] \( IE=\{coal,oil,natgas, refine, elect, gas\} \)
\[ PE_{Bb,t}^{E_{Bb,t}} = \sum_{k \in IE} PS_{k_{t}} A_{k,b,t}^{B} \]
\[ A_{k,b,t}^{B} = \left( \frac{1}{\kappa_{Bb,t}^{E}} \right)^{1-\sigma_{Bb}^{E}} \left[ \alpha_{k_{b,t}}^{E} \frac{PE_{Bb,t}^{E_{Bb,t}}}{PS_{k_{t}}} \right]^{\sigma_{Bb}^{E}} E_{b_{t}}^{B} \]
In Vennemo et al. (2014) the wind resource grows in a logistic manner, while in Phoenix the supply curve is a constant elasticity one. EPPA-4 introduces a “Fixed Factor” for wind and solar and interprets it as knowledge that grows with cumulated output, and sets an elasticity between the Fixed Factor and value-added \( \sigma_{RF_{kl}}^{\text{VR}} \) at 0.6. Paltsev et al. (2005) states that the “Choice of the substitution elasticity creates an implicit supply elasticity of wind in terms of the share of electricity supplied by the technology.”
We interpret of the resource variable for hydro, wind and solar as a limited supply of sites suitable for such technologies, and set the Nuclear, Hydro, Wind and SolarPV resources to grow in the base case at the rates projected in IEA (2014). In any period \( t \), in the policy case, the supply curve for Wind or SolarPV resources is given by:
\[
R^r_t = \bar{R}^r_t P_{R^r_t}^{\text{nucl, hyd, wind, solr}} \quad r=\text{NUCL, HYDR, WIND, SOLR}
\]
where \( \bar{R}^r_t \) is the projected base case resource availability. This means that opening an additional river, or wind farm, or solar farm, will require paying a price for the resource that is higher than the base case price. We follow Vennemo et al. (2014) in using an elasticity of 2.5, which means that this price is only slightly higher than base case resource price. It may be reasonable to not assume an identical function for all possible values of the base case resources availability, that is, when there is a high utilization of these water or wind sites, the marginal cost might increase substantially. That is, imposing a lower elasticity in the future years when there is a high penetration of such renewable sources.
The final element of the electricity sector is transmission and distribution. There is little data on this; even the 500-sector IO table for the U.S. has just one sector for total electric utilities. In the data appendix we describe how the data for Electric Utilities is separated into two sets for Generation and Transmission. The resulting input vector for Transmission includes almost all 33 commodities identified in the model. The tier structure for it is given in Figure A2(b) and is similar to the general structure for all other regular sectors given in Figure A1. The equations for the top tier of the Transmission sector, in terms of the producer price, are:
\[
P_{TD}^t = \frac{K_{TD}^t}{g_{TD}^t} \left[ \alpha_{M,t}^{TD,\sigma_{TD}^M} P_{M_{TD}}^{TD(1-\sigma_{TD}^M)} + \left(1 - \alpha_{M,t}^{TD} \right) P_{TD,j}^{MVE(1-\sigma_{TD}^M)} \right]^{\frac{1}{1-\sigma_{TD}^M}}
\]
\[
P_{TD}^t Q_{TD}^t = P_{TD,j}^{EVE} V_{TD}^t + P_{M_{TD}}^{TD} M_{TD}^t
\]
\[
V_{TD}^t = \left( \frac{K_{TD}^t}{g_{TD}^t} \right)^{1-\sigma_{TD}^M} \left[ \left(1 - \alpha_{M,t}^{TD} \right) P_{TD,j}^{MVE} \right]^{\sigma_{TD}^M} Q_{TD}^t
\]
The equations for the lower tiers are:
\begin{align*}
M^{TD}_t &= \left( \frac{k^{TD}_t}{g^{TD}_t} \right)^{-\sigma^{MTD}_t} \left[ \alpha^{MTD}_t \frac{P^{TD}_t}{PM^{BL}_t} \right]^\sigma^{MTD}_t Q^{TD}_t \\
\end{align*}
\begin{align*}
P^{\mu,TD}_t &= (1 + \eta^{EL}_t) P^{TD}_t
\end{align*}
The equations for the lower tiers are:
\begin{align*}
&\text{(A38)} \quad \ln PM^{TD}_{t} = \sum_{k \in NE} \alpha^{TD}_{kMt} \ln PS_{kt} \\
&P^{VE}_{TD,t} = \frac{1}{\kappa^{VE}_{TD,t}} \left[ (1 - \alpha^{VE}_{YAt}) \sigma^{VE}_{YTD} P^{EY}_{TDJ} + \alpha^{VE}_{YAt} \sigma^{VE}_{YTD} P^{YAE}(1-\sigma^{VE}_{YTD}) \right] \frac{1}{1 - \sigma^{VE}_{YTD}} \\
&P^{VA}_{TD,t} = \frac{1}{\kappa^{VA}_{TD,t}} \left[ \alpha^{VA}_{kTD} \sigma^{VA}_{kTDj} P^{KD}(1-\sigma^{VA}_{kTD}) + (1 - \alpha^{VA}_{kTD}) \sigma^{VA}_{kTD} P^{L}(1-\sigma^{VA}_{kTD}) \right] \frac{1}{1 - \sigma^{VA}_{kTD}} \\
&P_{TD,t}^{VE} = \frac{1}{\kappa^{VE}_{TD,t}} \sum_{k \in IE} \alpha^{EVE}_{k,TDE} \sigma^{EVE}_{kTDj} P^{EY}_{TDJ} \frac{1}{1 - \sigma^{EVE}_{kTD}} \\
&\text{IE} = \{\text{coal, oil, natgas, refine, elect, gas}\} \\
&P^{VE}_{TD,t} = P^{VA}_{TD,t} V_{A_{TD,t}} + PE_{TD,t} E_{TD,t} \\
&PM^{TD}_{t}^{MTD} = \sum_{k \in NE} PS_{k} A_{k,TD,t} \\
&PE_{TD,t}^{MTD} = \sum_{k \in IE} PS_{kt} A_{k,TD,t} \\
&P^{VE}_{TD,t} V_{A_{TD,t}} = P^{VE}_{TD,t} V_{A_{TD,t}} + PE_{TD,t} E_{TD,t} \\
&P^{VA}_{TD,t} V_{A_{TD,t}} = P^{VA}_{TD,t} V_{A_{TD,t}} + PE_{TD,t} E_{TD,t}
\end{align*}
The elasticity of substitution between materials and VE \((\sigma^{QI}_{TD})\) is set at 0.7, and \(\sigma^{VA}_{TD} = 1.0\) following Sue Wing (2011, p 30).
Note on aggregation and units of measurement
Equation (A17) is the cost dual of the quantity function that expresses the index of total generation output \((Q^{GEN}_{E})\) as a CES function of baseload output \((Q^{BL}_{E})\) and intermittent renewables \((Q^{RE}_{E})\). Baseload output is, in turn, a CES aggregate of the output coal, gas, nuclear, wind and others, \(\{Q^{GEN}_{E}\}\), that is, \(Q^{BL}_{E}\) is not a simple linear sum of the \(Q^{GEN}_{E}\)’s. We must thus be careful and distinguish between
the output of kWh of each generation source $l$, and the aggregate indices. $Q_{l,t}^{EGEN}$ is the output measured in billions of Y2010, and the kWh is given by a conversion coefficient:
$$Q_{l,t}^{kWh} = \psi_{l,t}^{EL} Q_{l,t}^{EGEN}$$
The total output in kWh is the simple sum:
$$Q_{TOT,t}^{kWh} = \sum_{l} Q_{l,t}^{kWh}$$
which grows at a (slightly) different rate from the output index $Q_{l,t}^{EG}$.
One may express the output index as a product of a quality index and the total kWh:
$$Q_{l,t}^{EG} = \psi_{l,t}^{EL} Q_{TOT,t}^{kWh}$$
The quality, or composition, index $\psi_{l,t}^{EL}$ represents the impact of changes in shares of components that have different relative prices. For example, the price per kWh of coal is much lower than that of solar and if the share of solar kWh rises, then the quality index rises. This is analogous to the relation between effective labor input, quality of labor and hours worked; the index of labor input is the product of labor quality and total hours, and a rising labor quality indicates that hours from more highly paid workers is rising as a share of total hours worked by all workers. One may think that the term “quality” of electricity may be misleading since kilowatt hours as perfectly homogenous and substitutable, it however, represents an economically meaningful distinction between kWh and value of kWh. The electrons that are identical from the user point of view are distinguished by the method of production – clean versus polluting, steady versus intermittent, near versus far. The $\psi_{l,t}^{EL}$ index represents changes in the costs of production as the composition of methods change; $Q_{l,t}^{EG}$ is an index of the (marketed) economic resources that went into producing $Q_{TOT,t}^{kWh}$ kWh of electricity. The non-marketed resources such as clean air, or clean water, are accounted for separately.
A.1.3. Households
Private consumption in this model is driven by an aggregate demand function that is derived by aggregating over different household types. Each household derives utility from the consumption of commodities, is assumed to supply labor inelastically, and owns a share of the capital stock. It also receives income transfers from the government and foreigners, and receives interest on its holdings of public debt. Aggregate private income is the sum over all households, and this income, after taxes and the payment of various non-tax fees ($FEE$), is written as:
\[(A45) \quad Y^P = \sum_k y^P_k \]
\[Y^P = Y_L + DIV + G_I + G_{\text{transfer}} + R_{\text{transfer}} - FEE - T^{LUMP} \]
\(Y_L\) denotes aggregate labor income from supplying \(LS\) units of effective labor, less income taxes:
\[(A46) \quad Y_L = (1 - t^L)PL LS \quad .\]
The relationship between labor demand and supply is given in equation A63 below. Aggregate supply \(LS\) is a function of the working age population, average annual hours, and an index of labor quality:
\[(A47) \quad LS_t = POP_t^w hr_t q^L_t \quad .\]
\(DIV\) denotes dividends from the households’ share of capital income and is explained below in A75. \(G_I\) and \(G_{\text{transfer}}\) represent interest and transfers from the government, and \(R_{\text{transfer}}\) is transfers from the rest-of-the-world. \(T^{LUMP}\) is the lump sum tax that is used in policy simulations.
Household income is allocated between consumption \((VCC_t)\) and savings. In this model we use a simple Solow growth model formulation with an exogenous savings rate \((s_t)\) to determine private savings \((S^P_t)\):
\[(A48) \quad S^P_t = s_t Y^P_t = Y^P_t - VCC_t \quad .\]
Total consumption expenditures are allocated to the 33 commodities identified in the model. We do this with a demand function estimated over household consumption survey data. This consumption data is at purchaser’s prices and follows the expenditure classification; these have to be linked later to the IO classifications and the factory-gate prices of the IO system. We arrange the demand system in a tier structure shown in Table 1. At the top tier total expenditures is allocated to Food, Consumer Goods, Housing and Services. In the sub-tiers these four bundles are allocated to 27 items.
Table 1. Tier structure of household consumption
| Name | Components in Consumer Expenditures |
|----------|------------------------------------------------------------------------|
| 1 C | Consumption, Food, Consumer goods, Services, Housing |
| | $CC = CC(FD, CG, SV, HS)$ |
| 2 FD | Food, Food & tobacco, Dining out |
| | $C^{FD} = C^{FD}(C1, C2)$ |
| 3 CG | Consumer goods, Clothing, Residential goods, Recreational & misc. goods|
| | $C^{CG} = C^{CG}(CL, RG, RM, C14)$ |
| 4 SV | Services, Communication, Education, Recreational svc, Health, Other services, Imputations, Transportation |
| | $C^{SV} = C^{SV}(C19, C22, C23, C24, C26, C27, TR)$ |
| 5 HS | Housing, Rental & housing services, Utilities-Energy |
| | $C^{HS} = C^{HS}(C5, EN)$ |
| 6 CL | Clothing, Clothes-shoes, Clothing services |
| | $C^{CL} = C^{CL}(C3, C4)$ |
| 7 RG | Residential goods, Furniture, Appliances, Interior Decorations, HH daily-use articles |
| | $C^{RG} = C^{RG}(C10, C11, C12, C13)$ |
| 8 RM | Recreational & Misc. goods, Communications equip, Books, Other goods |
| | $C^{RM} = C^{RM}(C18, C20, C21, C25)$ |
| 9 EN | Energy (dom), Water, Electricity, Coal, Gas |
| | $C^{EN} = C^{EN}(C6, C7, C8, C9)$ |
| 10 TR | Transportation, Gasoline, Vehicle svcs, Transportation fees |
| | $C^{TR} = C^{TR}(C15, C16, C17)$ |
Household $k$’s indirect utility function over the four aggregates in the top tier, $V(p, M_k)$, is of a form that allows for exact aggregation:
\[
\ln V_k = \alpha_0 + \ln(p_k / M_k) \gamma_p + \frac{1}{2} \ln(p_k / M_k) B \ln(p_k / M_k) + \ln(p_k / M_k)^\gamma B_{\mu} A_k ,
\]
where $M_k$ is the expenditures of household $k$, and $p = (p_{FD}^k, p_{CG}^k, p_{HS}^k, p_{SV}^k)$ is the price vector of the 4 bundles. Each household type has its own distinct utility function and $A_k$ is a vector of demographic dummy
variables to indicate the size of the household, the presence of children, the age of the head, and the region. The budget constraint for household $k$ is:
$$M_k = \sum_i p_i^k c_i^k = p_{FD}^k c_{FD}^k + p_{CG}^k c_{CG}^k + p_{HS}^k c_{HS}^k + p_{SV}^k c_{SV}^k$$
Let $w_i^k = p_i^k c_i^k / M_k$ denote the share of expenditure allocated to bundle $i$. Applying Roy’s Identity we get the demand share vector:
$$w^k = \frac{1}{D(p_k)} (\alpha_p + B \ln \frac{p_k}{M_k} + B_{p,t} A_t) = \frac{1}{D(p_k)} (\alpha_p + B \ln p_k - B_i \ln M_k + B_{p,t} A_t)$$
where $D(p_k) = -1 + t B_{p,t} \ln p_k$ and $w^k = (w_{FD}^k, w_{CG}^k, w_{HS}^k, w_{SV}^k)'$.
The aggregate demand is obtained by summing over all household types. Let $n_k$ be the number of households of type $k$; the aggregate share vector is then:
$$w_i = \frac{\sum_k n_k M_k w_i^k}{\sum_k n_k M_k} = \frac{\sum_k n_k M_k w_i^k}{M_i}$$
$$= \frac{1}{D(p_i)} [\alpha_p + B \ln p_i - B_i \sum n_{kt}^i M_{kt} \ln M_{kt} + B_{p,t} \sum n_{kt}^i M_{kt} A_k].$$
The above equations (A52) and (A53) are estimated simultaneously, with (A52) estimated over one year of cross-sectional consumer expenditure data, and (A53) estimated using time series national prices and aggregate consumption expenditures.
To use the estimated equation (A53) in the model that include projections into the future we make some modifications. Firstly, the consumer survey data does not include some items that are in the National Accounts such as imputed rentals for owner-occupied housing and FISIM. We make some adjustments to the $\alpha_p$’s to scale the shares to match the consumption in the Input-Output table for our base year 2010. We project the distribution and demographic terms to account for the aging impact and thus re-write the share demand system as:
$$w_i = \frac{1}{D(p_i)} [\alpha_p + B \ln p_i - B_i (\sum n_{kt}^i M_{kt} \ln M_{kt} + \ln M_k) + B_{p,t} \sum n_{kt}^i M_{kt} A_k]$$
$$w_i = \frac{1}{D(p_i)} [\alpha_p + B \ln p_i - B_i (\xi_{i,t}^{old} + \ln M_k) + B_{p,t} \pi_{i,t}^{old}]$$
Next, the aggregate expenditures on the 4 bundles are allocated to the 27 commodities according to the tier structure in Table 1. This is done with a linear logarithmic function that allows the shares to
change over time. For example, for the Transportation bundle, the value of expenditures \( v_{TR} \), the price index \( p_{TR}^{CE} \) and implied quantity is:
\[
\begin{align*}
(A55) \quad v_{TR} = p_{TR}^{CE} c_{TR} &= p_{15}^{CE} C_{15} + p_{16}^{CE} C_{16} + p_{17}^{CE} C_{17} \\
\ln p_{TR}^{CE} &= \alpha_{18,j} \ln p_{18,j}^{CE} + \alpha_{19,j} \ln p_{19,j}^{CE} + \alpha_{20,j} \ln p_{20,j}^{CE} ; \quad \alpha_{18,j} + \alpha_{19,j} + \alpha_{20,j} = 1 \\
c_{TR,j} &= v_{TR,j} / p_{TR,j}^{CE}
\end{align*}
\]
The demand for gasoline, item 15, is then:
\[
(A56) \quad C_{15,j} = \alpha_{15,j} v_{TR} / p_{15}^{CE}
\]
The consumption items listed in Table 1 are those used in the consumption survey and must be linked to the factory gate values in the Input-Output Accounts. For example, Food & tobacco in the Consumption accounts consist of commodities from Agriculture, Food Manufacturing, Trade (Commerce) and Transportation in the IO categories. The CE superscript denotes that these are prices for the consumption expenditure items. Table 2 gives the bridge that links these two accounts in the benchmark year 2010 for urban consumption. Column \( i \) of the bridge \( H^u \) gives the shares to allocate consumption item \( i \) to the 33 IO commodities. A similar table is constructed for rural consumption. Let be \( V C_{i,t}^{u,CE} \) the vector of consumption values for the urban sector, then the vector of consumption in IO terms is given by:
\[
(A57) \quad V C_{i,t}^{u,IO} = H^u V C_{i,t}^{u,CE}
\]
The prices of the consumption commodities are linked to the prices of the IO commodities via the same share matrix:
\[
(A58) \quad p^{u,CE}_{i,t} = H^u p^{IO}_{i,t} \; ; \quad p^{r,CE}_{i,t} = H^r p^{IO}_{i,t}
\]
The total value of consumption of commodity \( i \) is the sum of the urban and rural components:
\[
(A59) \quad V C_{it}^{IO} = V C_{it}^{u,IO} + V C_{it}^{r,IO} = p^{IO}_{it} C_{it}
\]
The value of national consumption in equation (A48) is the sum over all the commodities:
\[
(A60) \quad V C C_t = \sum_i V C_{it}^{IO} = p_{FD,t} C_{FD,t} + p_{CG,t} C_{CG,t} + C_{HS,t} p_{HS,t} + p_{SV,t} C_{SV,t}
\]
Table 2. Bridge to link Consumption Expenditures (urban) to Input-Output accounts
| | Food, Tobacco | Dining Out | Clothes | Appliance | Health | Care |
|----------------|---------------|------------|---------|-----------|--------|------|
| | 1 | 2 | 3 | ... | 11 | 24 |
| Agri | 1 | 0.141 | 0.280 | 0.000 | 0.000 | 0.000|
| Coal | 2 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000|
| Crude | 3 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000|
| natgas | 4 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000|
| nonenergy | 5 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000|
| Food | 6 | 0.781 | 0.100 | 0.000 | 0.000 | 0.000|
| textile | 7 | 0.000 | 0.000 | 0.030 | 0.000 | 0.000|
| apparel | 8 | 0.000 | 0.000 | 0.776 | 0.000 | 0.000|
| lumber | 9 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000|
| paper | 10 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000|
| Refine | 11 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000|
| Chem | 12 | 0.002 | 0.000 | 0.000 | 0.011 | 0.130|
| Build | 13 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000|
| pmetal | 14 | 0.000 | 0.000 | 0.000 | 0.001 | 0.000|
| metal | 15 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000|
| machin | 16 | 0.000 | 0.000 | 0.000 | 0.020 | 0.000|
| tequip | 17 | 0.000 | 0.000 | 0.000 | 0.030 | 0.000|
| emachin | 18 | 0.000 | 0.000 | 0.000 | 0.638 | 0.088|
| electro | 19 | 0.000 | 0.000 | 0.000 | 0.162 | 0.000|
| Instru | 20 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000|
| Other | 21 | 0.000 | 0.000 | 0.006 | 0.010 | 0.074|
| Elect | 22 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000|
| gasprod | 23 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000|
| constr | 24 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000|
| transp | 25 | 0.006 | 0.000 | 0.026 | 0.023 | 0.000|
| commun | 26 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000|
| commerc | 27 | 0.069 | 0.000 | 0.156 | 0.106 | 0.083|
| Hotel | 28 | 0.000 | 0.620 | 0.000 | 0.000 | 0.000|
| finance | 29 | 0.000 | 0.000 | 0.000 | 0.000 | 0.079|
| realest | 30 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000|
| business | 31 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000|
| service | 32 | 0.000 | 0.000 | 0.006 | 0.000 | 0.546|
| admin | 33 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000|
Sum of share 1.000 1.000 1.000 1.000 1.000
A.1.4. Government and Taxes
In the model, the government has two major roles. First, it sets plan prices and output quotas and allocates investment funds. Second, it imposes taxes, purchases commodities, and redistributes resources. Public revenue comes from direct taxes on capital, labor, value-added taxes, indirect taxes on output, tariffs on imports, the externality tax, and other non-tax receipts:
\[
\text{Rev} = \sum_j t^k (P^k_D K_D - D_j) + t^L P L S + t^T \sum_j (P^T_D K_D + P L LD_j + PT_j T D_j) \\
+ \sum_j t^j P I_j Q I_j + R_{EXT} + \sum_t t^T P M_i M_i + \sum_i t^i (Q I_i - X_i + M_i) + F E E + T^{LUMP}
\]
where \( D_j \) is the depreciation allowance and \( X_i \) and \( M_i \) are the exports and imports of good \( i \).
Externality taxes, such as those on SO2 or CO2, may be on the value of output, or on the quantities. We allow for both, the total revenue from the externality tax on output is:
\[
\text{R}_{EXT} = \sum_j t^j P I_j Q I_j
\]
In one application of the model described in Ho and Nielsen (2007, Chapter 10), the externality tax rate is set proportional to the marginal air pollution damages (\( MD^D \)) from output \( j \):
\[
t^j_{\mu} = \lambda M D^D_{\mu-1}
\]
When we consider a tax on fossil fuels based on the carbon content, the externality tax per unit of fuel \( j \) is:
\[
t^j = t^C c_j,
\]
where \( c_j \) is the carbon content per unit of fuel of type \( j \).
The nontax payments to the government are set as a fixed share of household income:
\[
F E E = \gamma^{N H I} \gamma^p
\]
Total government expenditure is the sum of commodity purchases and other payments:
\[
\text{Expend} = V G G + G_{INV} + \sum P I_j X_j + G_I + G_{IR} + G_{transfer}
\]
Government purchases of specific commodities are allocated as shares of the total value of government expenditures, \( V G G \). For good \( i \):
We construct a price index for government purchases as \( \log PGG = \sum \alpha_i \log PS_i \). The real quantity of government purchases is then:
\[
G = \frac{VGG}{PGG}.
\]
Transfers are set equal to a fixed rate of the population multiplied by the wage rate:
\[
G_{\text{transfer}} = \gamma^{\rho} PL POP_t.
\]
The difference between revenue and expenditure is the deficit, \( \Delta G \), which is covered by increases in the public debt, both domestic (\( B \)) and foreign (\( B^{G^*} \)):
\[
\Delta G_t = \text{Expend}_t - \text{Rev}_t,
\]
\[
B_t + B^{G^*}_t = B_{t-1} + B^{G^*}_{t-1} + \Delta G_t.
\]
The deficit and interest payments are set exogenously and equation A70 is satisfied by making the level of total nominal government expenditure on goods, \( VGG \), endogenous in the base case. In simulating policy cases we would often set the real government expenditures in the policy case equal to those in the base case. In this counterfactual we would use some endogenous tax variable to meet equation A70.
### A.1.5. Capital, Investment, and the Financial System
We model the structure of investment in a fairly simple manner. In the Chinese economy, some state-owned enterprises receive investment funds directly from the state budget and are allocated credit on favorable terms through the state-owned banking system. Non-state enterprises get a negligible share of state investment funds and must borrow at competitive interest rates. There is also a small but growing stock market that provides an alternative channel for private savings. We abstract from these features and define the capital stock in each sector \( j \) as the sum of two parts, which we call plan and market capital:
\[
K_{jt} = \overline{K}_{jt} + \tilde{K}_{jt}.
\]
The plan portion evolves with plan investment and depreciation:
\[
\overline{K}_{jt} = (1-\delta)\overline{K}_{j,t-1} + \psi' T_{jt}, \quad t = 1, 2, \ldots, T.
\]
The rate of depreciation is $\delta$, and $\psi_i$ is an aggregation that converts the investment units to capital stock units. In this formulation, $K_{j0}$ is the capital stock in sector $j$ at the beginning of the simulation. This portion is assumed to be immobile across sectors. Over time, with depreciation and limited government investment, it will decline in importance. Each sector may also rent capital from the total stock of market capital, $\tilde{K}_i$:
$$\tilde{K}_i = \sum_j \tilde{K}_{ji}, \quad \text{where} \quad \tilde{K}_{ji} > 0.$$
The allocation of market capital to individual sectors, $\tilde{K}_{ji}$, is based on sectoral rates of return. As in equation A2, the rental price of market capital by sector is $\tilde{P}_{jD}$. The supply of $\tilde{K}_{ji}$, subject to equation A72, is written as a translog function of all of the market capital rental prices, $\tilde{K}_{j} = K_j(\tilde{P}_{1D}, \ldots, \tilde{P}_{nD})$:
$$\frac{\tilde{K}_{ji}}{\tilde{K}_i} = \alpha_j^{KS} + \sum_i \beta_{ij}^{KS} \ln \tilde{P}_{iD}$$
To simplify the modeling of the capital supply in the electricity sector, we first allocate $\tilde{K}_{elect,t}$ according to (A74b) and then allocate that to the various generation subsectors, using a similar function of the rental rates in the various electricity subsectors:
$$\tilde{K}_{elect,t} = \tilde{K}_{transm,t} + \tilde{K}_{coal,t} + \ldots \tilde{K}_{solar,t}; \quad K_{elect,t} = K_{transm,t} + K_{coal,t} + \ldots K_{solar,t}$$
$$\tilde{P}_{elect,t}K_{elect,t} = \sum_e \tilde{P}_{eD} K_{e,t} \quad \text{e}=\text{transm, coal, … solar}$$
In three sectors, agriculture, crude petroleum and gas mining, “land” is a factor of production. We have assumed that agricultural land and oil fields are supplied inelastically, abstracting from the complex property rights issues regarding land in China. After taxes, income derived from plan capital, market capital, and land is either kept as retained earnings by the enterprises, distributed as dividends, or paid to foreign owners:
---
18 Both K and I are aggregates of many asset types, ranging from computer equipment to structures. The composition of total investment and total capital stock are different and an aggregation coefficient is needed to reconcile the historical series.
\[ \sum_j \text{profits}_j + \sum_j \bar{P}_j^K \bar{K}_j + \sum_j PT_j T_j = \text{tax}(k) + RE + DIV + r(B^*), \]
where \( \text{tax}(k) \) is total tax on capital and value added (the first two terms on the right hand side of equation A2).\(^{19}\)
As discussed below, total investment in the model is determined by savings. This total, \( VII \), is then distributed to the individual investment goods sectors through fixed shares, \( \alpha^j_t \):
\[ PS^t I^t_{it} = \alpha^j_t VII^t. \]
A portion of sectoral investment, \( \bar{I}_t \), is allocated directly by the government, while the remainder, \( \tilde{I}_t \), is allocated through other channels.\(^{20}\) The total, \( I_t \), can be written as:
\[ I_t = \bar{I}_t + \tilde{I}_t = I^d_{it} I^{a1} \ldots I^{dN}_{it}. \]
As in equation A73 for the plan capital stock, the market capital stock, \( \bar{K}_{jt} \), evolves with new market investment:
\[ \bar{K}_{jt} = (1-\delta)\bar{K}_{j,t-1} + \psi^j_t \bar{I}_{jt}. \]
**Non-reproducible assets**
In addition to the capital stock, the households own the non-reproducible assets – land, renewable resources and sequestration resources. The supply of land (or mining resources) is simply assumed fixed for each type (agriculture, coal mining, oil mining):
\[ T_{jt} = T_{j0} \]
The supply curves for nuclear and hydro resources are assumed to be upward sloping curves.
\[ RS_{b,t}^B = \left( \frac{PR_{b,t}^B}{PR_{b0}^B} \right)^\sigma_b^B \]
---
\(^{19}\) In China, a substantial part of the “dividends” are actually income due to agricultural land.
\(^{20}\) It should be noted that the industries in the Chinese accounts include many sectors that would be considered public goods in other countries. Examples include local transit, education, and health.
A.1.6. The Foreign Sector
Trade flows are modeled using the method followed in most single-country models. Imports are considered to be imperfect substitutes for domestic commodities and exports face a downward sloping demand curve. We write the total domestic supply of commodity $i$ as a CES function of the domestic ($DC_i$) and imported good ($M_i$):
\[(A81) \quad DS_i = A_0\left[\alpha^D DC_i^\rho + \alpha^M M_i^\rho\right]^{1/\rho},\]
where $DC$ is the quantity of domestically produced goods that are sold domestically. The elasticity is $\sigma = 1/(1 - \rho)$. The cost dual corresponding to the above primal function is:
\[(A82) \quad PS_i = \frac{1}{A_0}\left[\alpha^\sigma PD_i^{1-\sigma} + \alpha^m M_i^{1-\sigma}\right]^{1/(1-\sigma)}\]
and the value of total domestic supply is:
$$PS_i DS_i = PD_i DC_i + PM_i M_i$$
The purchaser’s price for domestic goods, $PD_i$, is related to the commodity supply price $PC_i$ and is discussed in the export section below. $PS_i$ is the price of the basket of commodity $i$ to domestic purchasers. The price of imports to buyers is the foreign price plus tariffs (less export subsidies), multiplied by a world relative price, $e$:
\[(A83) \quad PM_i = e(1 + t_i)PM_i^* + t_i^{ru} .\]
From (A82) we may derive the demand for imports as:
\[(A.84) \quad \frac{PM_i M_i}{PS_i DS_i} = \frac{\alpha^{m(1-\rho)} M_i^{\rho/(1-\rho)} - 1}{\alpha^{\sigma(1-\rho)} DC_i^{\rho/(1-\rho)} + \alpha^m M_i^{\rho/(1-\rho)}} = \frac{\alpha^\sigma PM_i^{1-\sigma}}{\alpha^\sigma PD_i^{1-\sigma} + \alpha^m PM_i^{1-\sigma}}\]
Domestically produced commodities (QC) are allocated to the domestic market and exports according to a constant elasticity of transformation (CET) function:
\[(A85) \quad QC_{it} = \kappa_i^X \left[\alpha_i^X X_i^{\sigma_i^X - 1} + (1 - \alpha_i^X) DC_i^{\sigma_i^X \sigma_i^D - 1}\right]^{\sigma_i^X / (\sigma_i^X - 1)}\]
The ratio of exports to domestically sold goods depends on the domestic price (PD) relative to world prices adjusted for export subsidies ($s_i^{e}$):
(A85b) \( X_{it} = DC_{it} \left[ 1 - \frac{\alpha^x_{it} PD_{it}}{\alpha^x_{it}} \right]^{\sigma^r_{it}} ; \quad PX_{it} = e_i (1 + s^e_i) PE^e_{it} \)
The value identity is:
(A86) \( PC_{it} QC_{it} = PD_{it} DC_{it} + PX_{it} X_{it} \)
The weights and constant terms are set using base year values:
\[
\alpha^x_{it} = \frac{PD_{i0} X_{i0}^{1/\sigma^r_{i0}}}{PD_{i0}^{1/\sigma^r_{i0}} + PX_{i0} DC_{i0}^{1/\sigma^r_{i0}}} ; \quad \kappa^x_i = QC_{i0} / \left[ \alpha^e_{i0} X_{i0}^{\sigma^e_{i0} - 1} + (1 - \alpha^e_{i0}) DC_{i0}^{\sigma^e_{i0} - 1} \right]^{\sigma^r_{i0} - 1}
\]
The share parameters \( \alpha^x_{it} \) are projected exogenously to take into account the rising role of exports during 1980-2010 and a falling role in the future. The price \( PC \) is given in equation (A14) above, and is also an implicit dual function of (A85), \( PC = f(PX, PD) \).
The current account balance is equal to exports minus imports (valued at world prices before tariffs), less net factor payments, plus transfers:
(A87) \[
CA = \sum_i PX_i X_i (1 + s^r_i) - \sum_i ePM^i_i M_i - r(B^* - G - IR + R_{\text{transfer}}),
\]
\[
= VX - VM - r(B^* - G - IR + R_{\text{transfer}})
\]
Like the government deficits, the current account balances are set exogenously and accumulate into stocks of net foreign debt, both private \( (\bar{B}^p_i) \) and public \( (\bar{B}^{p*}_i) \):
(A88) \[
\bar{B}^p_i + \bar{B}^{p*}_i = \bar{B}^{p*}_{i-1} + B^{p*}_{i-1} - CA_i .
\]
A.1.7. Markets
The economy is in equilibrium in period \( t \) when the market prices clear the markets for the 33 commodities and the three factors. The supply of domestically produced commodity \( i \) must satisfy the total of intermediate and final demands:
(A89) \[
DS_i = \sum_j A_{ij} + C_i + I_i + G_i , \quad i = 1, 2, \ldots, 33.
\]
For the labor market, we assume that labor is perfectly mobile across sectors so there is one average market wage which balances supply and demand. As is standard in models of this
type, we reconcile this wage with the observed spread of sectoral wages using wage distribution coefficients, \( \psi_{jt}^L \). Each industry pays \( PL_{jt} = \psi_{jt}^L PL_t / (1 - t^V_j) \) for a unit of labor. The labor market equilibrium is then given as:
\[
(A90) \quad \sum_j \psi_{jt}^L L_{dt} = LS_t.
\]
For the non-plan portion of the capital market, adjustments in the market price of capital, \( \tilde{P}_j^{KD} \), clears the market in sector \( j \):
\[
(A91) \quad KD_{jt} = \psi_{jt}^K K_{jt},
\]
where \( \psi_{jt}^K \) converts the units of capital stock into the units used in the production function. The rental price \( PT_j \) adjusts to clear the market for “land”:
\[
(A92) \quad TD_j = T_j, \quad \text{where} \quad j = \text{“agriculture”, “crude petroleum”, “gas mining”}.
\]
In this model without foresight, investment equals savings. There is no market where the supply of savings is equated to the demand for investment. The sum of savings by households, businesses (as retained earnings), and the government is equal to the total value of investment plus the budget deficit and net foreign investment:
\[
(A93) \quad S^p + RE + G_{IN} = VII + \Delta G + CA.
\]
The budget deficit and current account balance are fixed exogenously in each period. The world relative price \( (e) \) adjusts to hold the current account balance at its exogenously determined level.
The model is a constant returns-to-scale model and is homogenous in prices, that is, doubling all prices leaves the economy unchanged. We are free to choose a price normalization.
A.1.8 Welfare Other accounting identities
The household welfare function (A50) is chosen to allow aggregation over different households. The aggregation issues are discussed in Jorgenson et al. (2013, Chapter 3); equation (A54) gives the aggregate demand function for the four consumption bundles. Jorgenson et al. expresses social welfare as a function that takes into account the different compositions of
households (different size and number of children), using the concept of household equivalents. The welfare function depends on the average level of consumption as well as inequality of consumption (efficiency and equity). Here we compute only the average levels to give the efficiency measure which is given by:
\[
\ln \bar{V} = \frac{\sum_k m_h(p, A_k) \ln V_k}{\sum_k m_h(p, A_k)}
\]
\(V_k\) is the household utility in (A50), and \(m_h(p, A_k)\) is the household equivalent to the reference household which is aged 18-34, male, elementary school, two members and in the East. The equivalence scale is explained in Jorgenson and Slesnick (1987) and is given by:
\[
\ln m_h(p, A_k) = \frac{1}{D(p)} [\ln p'B_{pA}A_k]
\]
The money measure of welfare is given by a social expenditure function (Jorgenson and Slesnick 1987, eq. 5.15):
\[
\ln M(p, W) = \frac{1}{D(p)} [\ln p'\alpha + \frac{1}{2} \ln p' B \ln p-W] + \ln \sum_k m_h(p, A_k)
\]
The money measure of the change in welfare due to a policy (from \(W^0\) to \(W^1\)) is a function of the policy case measured at base case prices \((P^0)\):
\[
\Delta M = M(p^0, W^1) - M(p^0, W^0)
\]
Gross domestic product in nominal terms is the sum of consumption, investment, government spending, plus net exports:
\[
VGDP = VCC + VII + VGG + VX - VM
\]
To construct real, constant yuan, GDP we need to first define real consumption, investment, etc. These are expressed as the divisia aggregate of the 33 commodities that make up each component, for example, real personal consumption expenditures is:
\[
CC^{div} = \text{divisia}(C; PS^C)
\]
\[
d \Delta \ln \frac{CC^{div}}{CC^{div}_{t-1}} = \sum_i \frac{1}{2}(v^f_{i,t} + v^f_{i,t-1}) \Delta \ln \frac{C_i}{C_{i,t-1}} ; \quad v^f_{i,t} = \frac{PS_i C_i}{VCC_t}
\]
Real GDP is then a divisia index of these components:
A.1.9 Energy, emissions and environmental accounting
To account for atmospheric environmental damages we consider a range of criteria pollutants: particulate matter (PM$_{2.5}$ and PM$_{10}$), sulphur dioxide, nitrogen oxides, VOCs, ammonia. We also account for greenhouse gas emissions, in particular carbon dioxide. The PM concentration is due to primary PM emissions as well as secondary particles such as sulfates and nitrates which are formed from sulfur dioxide and NO$_x$ respectively. The emissions inventory is described in *Clearer Skies*, Chapters 4-6. To illustrate the calculations we describe here a simplified account of energy flows and primary PM, SO$_2$ and NO$_x$ emissions.
We begin by describing the energy variables. Very often a simple indicator of total primary energy production and consumption is produced by summing the energy equivalents of the fossil fuels and primary electricity and heat. This may not be a very useful indicator given that a joule of energy from burning coal is very different in the ease of use from a joule from gasoline or a joule of electricity; a difference that is reflected in the prices per joule. Nevertheless, for comparison with well-known series we compute the standard coal equivalent (sce) of these primary sources of energy.
First, recall that we distinguish between industry output ($Q_I$) and commodity output ($Q_C$). $Q_C$ is the constant yuan quantity of commodity produced (billions of 2010 yuan). $Q_P^t$, the total quantity of coal, crude, and gas produced (whether combusted or not) in year $t$ is given by the commodity output ($Q_C$) multiplied by the fuel conversion coefficient, $\xi^{f}_{mean}$:
$$ (A101) \quad Q_P^t = \xi^{f}_{mean} Q_C^t, \quad f=\text{coal, crude oil, gas mining, electricity} $$
where $\xi^{f}_{mean}$ is the quantity of the commodity output (in million tons, million m$^3$, or billion kWh) per billion yuan of commodity output. For example, the quantity of raw coal produced in million tons is given by $Q_P^{\text{raw coal}} = \xi^{\text{coal}}_{mean} Q_C^{\text{coal},t}$. Since electricity is only a part of the “Electricity, Steam & Hot water” sector, the quantity of electricity produced (in billion kWh) is:
$$ (A101) \quad Q_P^{\text{electric}} = \xi^{\text{electric}}_{mean} \alpha^{\text{el only}}_{elect} Q_C^{\text{electric},t} $$
where $\alpha^{\text{el only}}_{elect}$ is the electricity share of the “Electricity, Steam, & Hot water” sector’s commodity output.
We compute energy consumption in two ways. The first way simply uses the total output of fuels (production based account); the second way sums over the industry consumption of energy that is calibrated to the official estimates in the base year (consumption based account). First, $E^{\text{PROD}}$, the total sce of energy produced domestically, is:
where $e_f$ is the energy content of a unit of fuel $f$ (e.g. tons of sce per ton of oil) and the PRI superscript denotes primary electricity from renewables and nuclear. (In this calculation we ignore the tiny amount of heat from natural sources.) We set $\alpha_{PRIelec}^P$, the share of electricity produced from primary sources, exogenously by considering the projected generation of renewables and nuclear power. Then $Q^{PRIelec}_t$, the quantity of primary electricity produced from renewables and nuclear, is:
\[(A103)\quad Q^{PRIelec}_t = \alpha_{PRIelec}^P E^{Pelec}_t \]
$E^{EXP}$, the total sce of energy exported, on net, is:
\[(A104)\quad E^{EXP}_t = e_{coal}^{\text{mean}} (X_{coal,t} - M_{coal,t}) + e_{oil}^{\text{mean}} (X_{crude,t} - M_{crude,t} + X_{refine,t} - M_{refine,t}) + e_{gas}^{\text{mean}} (X_{natgas,t} - M_{natgas,t})\]
where $X_f$ is the value of exports of fuel $f$ (in billion yuan) and $M_f$ is the value of imports of fuel $f$ (in billion yuan). Exports of electricity are not counted in this measure since it is a secondary energy, the pollution due to the generation of electricity for exports is located in the country and they are not exported.
$E^{CONS}$, the total energy consumed in China (in tons sce) is then given by production less net exports, less changes in inventory ($E^{INV}_t$):
\[(A105)\quad E^{CONS}_t = E^{PROD}_t - E^{EXP}_t - E^{INV}_t\]
\[(A106)\quad E^{CONS}_t = e_{coal}^{\text{mean}} (QC_{coal,t} - X_{coal,t} + M_{coal,t}) + e_{oil}^{\text{mean}} (QC_{crude,t} - X_{crude,t} + M_{crude,t} - X_{refine,t} + M_{refine,t}) + e_{gas}^{\text{mean}} (QC_{natgas,t} - X_{natgas,t} + M_{natgas,t}) + e_{elect} Q^{PRIelec}_t - E^{INV}_t\]
\[(A107)\quad C^{cons}_t = e_{coal}^{\text{mean}} (QC_{coal,t} - X_{coal,t} + M_{coal,t}), \ ext{etc.}\]
The second expression in (A105) substitute in (A102) and (A104) to show that it is the constant yuan output less net exports, multiplied by the fuel conversion coefficient, and multiplied by the energy content
coefficient. In (A106), the variables $CF_{i}^{coal}$, $CF_{i}^{oil}$, $CF_{i}^{gas}$ denote the quantity of fuel consumed in million tons or million $m^{3}$. Here, $CF_{i}$ is calculated as the sum of commodity output ($QC$) and imports ($I$) less exports ($X$), multiplied by the fuel conversion coefficient ($\zeta_{mean}$) to convert the constant yuan of fuel consumed into the quantity of fuel consumed (in million tons or million $m^{3}$).
Second, in consumption-based accounting, we also calculate national energy consumption by adding over each industry, using industry specific information about the consumption of coal, coke, liquid fuels, etc. We first define consumption coefficients ($\zeta_{j}$) by taking the data on fuel actually used (in million tons, million $m^{3}$, or billion kWh) from the China Statistical Year book (CSY 2012, Table 7-9 “Consumption of Energy by Sector”) and dividing by the value of energy purchases given in the Input-Output table.
To disambiguate, the fuel conversion coefficient ($\zeta_{mean}$), presented in equation (A101), is computed using the production data at the aggregate level: the total quantity of the commodity output divided by the total value of the commodity output. In contrast, the consumption coefficient ($\zeta_{j}$), presented here, is computed using the consumption data at the industry level: industry $j$’s consumption of fuel $f$ (in million tons, million $m^{3}$, or billion kWh) divided by the value of industry $j$’s purchases of fuel $f$ (in billion yuan).
Secondary fuels are produced by the Petroleum Refining & Coal Products sector which we group as coke, refined liquids, and other petroleum products. The “other petroleum products,” such as bitumen and lubricants, are assumed to be not combusted (i.e. not contributing to CO2 emissions). Each industry $j$ purchase a different share of coke (coal products) from this sector and we write the value of coke input as a share of the value of Refining & Coal Products in the Use matrix: $\alpha_{ref \_co,j}^{coalpr} U_{refine,j}$ where $\alpha_{s,j}^{f}$ is the share of fuel $f$ in sector $s$ that industry $j$ purchases, and $U_{f,j}$ is the value (in billion yuan) of inputs of fuel $f$ for industry $j$ from the Use matrix.
The value of Refined liquids and Other Petroleum Products consumed are then:
$$\alpha_{refine}^{liquid}(1-\alpha_{ref \_co,j}^{coalpr}) U_{refine,j} ; \quad (1-\alpha_{refine}^{liquid})(1-\alpha_{ref \_co,j}^{coalpr}) U_{refine,j}$$
That is, the value of refined liquids is the product of: 1) the share of liquids in total refined petroleum products; 2) the share of non-coal products in the Refining & Coal Products sector that industry $j$ purchases; and 3) the value of Refining & Coal Products purchased by industry $j$. Similarly, the value of Other Petroleum Products input (on the right) can be interpreted as the product of: 1) the share of non-liquids in total refined petroleum products; 2) the share of non-coal products in the Refining & Coal Products sector that industry $j$ purchases; and 3) the value of Refining & Coal Products purchased by industry $j$.
The energy consumption coefficients for coke and liquid fuels are thus:
where $F_{\text{CSY coke,}j\text{,baseyr}}$ is the quantity of coke consumed by $j$ in million tons in the base year 2010. $F_{\text{CSY liquidfuel,}j\text{,baseyr}}$ is the sum of the quantity of gasoline, kerosene, diesel and fuel oil consumed (given in CSY 2012), and $F_{\text{CSY otherpetroleum,}j\text{,baseyr}}$ is the sum of the quantity of lubricant, bitumen, naphta, etc. consumed (given in the LBL’s China Energy Databook).
(A109) $\alpha_{\text{refine liquid}} = \frac{F_{\text{CSY liquidfuel,}j\text{,baseyr}}}{(F_{\text{CSY liquidfuel,}j\text{,baseyr}} + F_{\text{CSY otherpetroleum,}j\text{,baseyr}})}$
is the quantity share of liquids consumed in consumption of total refined petroleum products for industry $j$.
Our model distinguishes between the Gas Mining sector and the Gas Utilities (or Gas Products) sector; most industries purchase only from Gas Products, while a few purchase from Gas Mining for transformation and combustion – Chemicals, Electricity and Gas Products. For all industries $j$ other than Gas Products the consumption coefficient is the quantity of natural gas purchased by industry $j$ divided by the sum of the values of natural gas and gas products purchased by industry $j$ in the base year:
(A110) $\xi_{j\text{,baseyr}} = \frac{F_{\text{CSY natgas,}j\text{,baseyr}}}{U_{\text{natgas,}j\text{,baseyr}} + U_{\text{gasprod,}j\text{,baseyr}}}$, $j \neq \text{Gas Products}$
In contrast, the energy consumption coefficient for the Gas Products industry is divided by only the value in the Gas Products cell, excluding the Gas Mining cell:
(A111) $\xi_{j\text{,gasprod,baseyr}} = \frac{F_{\text{CSY natgas,}j\text{,baseyr}}}{U_{\text{gasprod,}j\text{,baseyr}}}$
We now move on from calculating energy consumption to calculating energy combustion. The above consumption coefficients refer to the purchases of the different fuels. Some of these oil and gas inputs are not combusted but converted to other products such as fertilizer or bitumen. In the Refining sector part of the crude input is combusted but most are converted to liquid fuels or other petroleum products; the un-combusted portion is represented by the “refining loss” coefficient, $\rho_{j\text{,ref loss}}^{\text{ref,loss}}$, where $(1 - \rho_{j\text{,ref loss}}^{\text{ref,loss}})$ is the fraction of un-combusted crude input. (For industries other than $j=\text{Refining}$, $\rho_{j\text{,ref loss}}^{\text{ref,loss}}$ is simply 1, reflecting that 100% of the crude input is combusted.)
In the Gas Products (Utilities) industry, gas is purchased from the Natural Gas Mining sector and sold to consumers, that is, there is assumed to be no combustion in this industry. In the Chemicals sector, raw gas is purchased from the Gas Mining sector and part of it is converted to plastics and other products. The combusted portion is represented by $\rho_{j=\text{Chemical}}^{\text{gas loss}}$. (For industries other than $j=\text{Chemical}$, $\rho_{j=\text{Chemical}}^{\text{gas loss}}$ is simply
1, reflecting that 100% of the raw gas is combusted). Thus, these loss adjustment coefficients can be thought of as the share of fuel \( f \) that is combusted for industry \( j \).
To disambiguate in advance: \( FT_{ji}^{CSY} \) (used in the previous section) refers to the quantity of fuel \( f \) purchased, while \( FT_{ji}^f \) (used below) refers to the quantity of fuel \( f \) combusted.
The quantity of fuel combusted \( (FT) \) is given by the constant yuan of fuel \( (A_{ij}) \) multiplied by the consumption coefficients \( (\xi^f_j) \) that converts the value of fuel \( f \) to physical quantities (tons of coal, tons of oil, m\(^3\) of gas, kWh of electricity)), and multiplied by these loss adjustments \( (\rho^{f,loss}_j) \). The following equations describe the quantity of fuel combusted for coal, oil, other petroleum products (\( nonliqref \)), and gas in terms of fuels at a finer classification:
\[
\begin{align*}
(\text{A112}) \quad FT_{ji}^{coal} &= Q_{ji}^{rawcoal} + Q_{ji}^{coke} = \xi^\text{coal} \chi_j \rho_j^{\text{coke,loss}} A_{\text{coal},ij,t} + \xi^\text{coalpr} \alpha_j^{\text{coalpr}} A_{\text{refining},jt} \\
&= \xi^\text{coal} \chi_j \rho_j^{\text{coke,loss}} U_{\text{coal},ij} / PS_{\text{coal}} + \xi^\text{coalpr} \alpha_j^{\text{coalpr}} U_{\text{refine},j} / PS_{\text{refine}} \\
FT_{ji}^{oil} &= Q_{ji}^{crude} + Q_{ji}^{refinedoil} = \xi^\text{oil} \chi_j \rho_j^{\text{ref,loss}} A_{\text{oil},ij,t} + \xi^\text{refliq} \alpha_j^{\text{reflq}} U_{\text{refine},j} (1 - \alpha^{\text{coalpr}}_j) A_{\text{refining},jt} \\
&= \xi^\text{oil} \chi_j \rho_j^{\text{ref,loss}} U_{\text{crude},ij} / PS_{\text{crude}} + \xi^\text{refliq} \alpha_j^{\text{reflq}} U_{\text{refine},j} (1 - \alpha^{\text{coalpr}}_j) U_{\text{refine},j} / PS_{\text{refine}} \\
&= \begin{cases}
\xi^\text{gas} \chi_j \rho_j^{\text{gas,loss}} U_{\text{natgas},ij} / PS_{\text{natgas}} + \xi^\text{gasprod} U_{\text{gasprod},ij} / PS_{\text{gasprod}} & j \neq \text{gasprod} \\
\xi^\text{gas} \chi_j \rho_j^{\text{gas,loss}} U_{\text{gasprod},ij} / PS_{\text{gasprod}} & j = \text{gasprod}
\end{cases}
\end{align*}
\]
where that the constant yuan quantity of energy input is given by the value in the Use matrix divided by the fuel price: \( A_j = U_{ij} / PS_{ij} \). \( Q_i \) is the quantity of energy input \( i \) (at the finer classification) combusted, and the quantity of fuel \( f \) combusted is the sum over the \( i \) finer types to give \( FT^f \).
For electricity, \( \alpha^{\text{elec}}_i \) is the share of electricity in Electricity, Steam & Hot Water, and we may similarly define \( FT_{ji}^{\text{elec}} \), the quantity of electricity purchased (in billion kWh) as:
\[
(\text{A113}) \quad FT_{ji}^{\text{elec}} = \xi^{\text{elec}} \chi_j \alpha^{\text{elec}}_i U_{\text{elec},ij} / PS_{\text{elec}}
\]
The above equations (A112) and (A113) are for the industry purchases of energy, a similar set of equations hold for household and investment use of energy:
(A114) \[ FT_{\text{coal}}^{\text{HH,}j} = \frac{\alpha_{\text{coal}}}{\alpha_{\text{coal}}} C_{\text{coal},j} + \frac{\alpha_{\text{coalpr}}}{\alpha_{\text{coalpr}}} C_{\text{refining},j} \]
\[ FT_{\text{oil}}^{\text{HH,}j} = \frac{\alpha_{\text{oil}}}{\alpha_{\text{oil}}} C_{\text{oil},j} + \frac{\alpha_{\text{liquid}}}{\alpha_{\text{liquid}}} \left(1 - \alpha_{\text{coalpr}}\right) C_{\text{refining},j} \]
\[ FT_{\text{gas}}^{\text{HH,}j} = \frac{\alpha_{\text{gas}}}{\alpha_{\text{gas}}} C_{\text{gas},j} + \frac{\alpha_{\text{gasprod}}}{\alpha_{\text{gasprod}}} C_{\text{gasprod},j} \]
\[ FT_{\text{coal}}^{\text{INV}} = \frac{\alpha_{\text{coal}}}{\alpha_{\text{coal}}} I_{\text{coal}}; \quad FT_{\text{oil}}^{\text{INV}} = \frac{\alpha_{\text{oil}}}{\alpha_{\text{oil}}} I_{\text{oil}}; \quad FT_{\text{gas}}^{\text{INV}} = \frac{\alpha_{\text{gas}}}{\alpha_{\text{gas}}} I_{\text{gas}} \]
where \( C_{\text{elect}}, C_{\text{coal}}, C_{\text{refining}}, C_{\text{oil}}, C_{\text{gas}}, \text{ and } C_{\text{gasprod}} \) denote the constant yuan value of Consumption by households of those fuels (in billion yuan). \( I_f \) is the value (in billion yuan) of purchases of fuel \( f \) by the Investor (these are essentially business inventories in the Investment column of the input-output accounts).
The un-combusted portions in this version are the other petroleum products (“nonliqref”) and part of the gas use by the Chemicals industry. We denote the un-combusted fuel use by FU:
\[ (A115) \]
\[ FT_{\text{nonliqref}}^{\text{ref}} = \frac{\alpha_{\text{nonliqref}}}{\alpha_{\text{nonliqref}}} \left(1 - \alpha_{\text{liquid}}\right)\left(1 - \alpha_{\text{coalpr}}\right) U_{\text{refine},j} / PS_{\text{refine}} \quad j=1,\ldots,33 \]
\[ FT_{\text{gas}}^{\text{ref}} = \frac{\alpha_{\text{gas}}}{\alpha_{\text{gas}}} \left(1 - \alpha_{\text{gasloss}}\right) U_{\text{gas},j} / PS_{\text{gas}} \quad j=\text{Chemicals} \]
[In this version we have not separated out the combusted portion of “other petroleum products (nonliqref)”. A more detailed accounting would have refined products divided to “refined liquids”, “other combustible refined products”, and “noncombustible refined products”. The equations would be:
\[ (A115n) \]
\[ FT_{\text{noncomb}}^{\text{ref}} = \frac{\alpha_{\text{noncomb}}}{\alpha_{\text{noncomb}}} \left(1 - \alpha_{\text{liquid}} - \alpha_{\text{comb}}\right)\left(1 - \alpha_{\text{coalpr}}\right) U_{\text{refine},j} / PS_{\text{refine}} \]
The other combustible refined products are LPG, Refinery Gas and other gases and should be counted in the calculation of emissions. In the current version we combine them with the noncombustible refined products (wax, asphalt, etc) to give total “nonliqref”.
The total energy consumed by industry \( j \) or households is the sum of these physical units of primary fossil fuels combusted multiplied by the energy conversion coefficient (\( e_f \), e.g. tons of SCE per ton of coal) plus the electrical energy, plus the un-combusted portions:
\[ (A116) \]
\[ EIND_{\text{HH,}j} = e_{\text{coal}} FT_{\text{coal}}^{\text{HH,}j} + e_{\text{oil}} FT_{\text{oil}}^{\text{HH,}j} + e_{\text{gas}} FT_{\text{gas}}^{\text{HH,}j} + e_{\text{elect}} FT_{\text{elect}}^{\text{HH,}j} + e_{\text{oil}} FT_{\text{nonliqref}}^{\text{HH,}j} \quad j=1,\ldots,33,\text{HH,INV} ; \quad j=\text{elect} \]
When we express energy consumption as above we are counting \( j \)'s use of electricity as energy consumed by \( j \), not as energy consumed by the Electric Utilities when it burns coal to generate electric
power. For a consistent accounting of total national consumption, the net energy consumed by Electric Utilities is only the generation loss plus the Utilities own electricity consumption \( (U_{\text{elect,elect}}) \). The generation loss is given by the energy embodied in the fuels combusted in the power plants less the energy embodied in the delivered thermal electricity (total electricity minus renewables and nuclear \( (Q_{t,\text{Pelec}} - Q_{t,\text{Prielec}}) \)). The net energy consumed by Electric Utilities, \( E\text{IND}_{j=\text{elect}} \), is thus:
\[
E\text{IND}_{j=\text{elect},t} = e_{\text{coal}} F_{t,j}^{\text{coal}} + e_{\text{oil}} F_{t,j}^{\text{oil}} + e_{\text{gas}} F_{t,j}^{\text{gas}} - e_{\text{elect}} (Q_{t,\text{Pelec}} - Q_{t,\text{Prielec}}) + e_{\text{elect}} U_{\text{elect},j} / P_{s,\text{elect}}
\]
(A117)
The national total energy consumption is then the sum over all industries and final demand:
\[
E_{\text{IND}T_{\text{OT}},t} = \sum_{j} E\text{IND}_{j,t} + E\text{IND}_{\text{HH},t}
\]
(A118)
This should be equal to \( E_{t,\text{CONS}} \), the total computed from the production data in equation A107.
Emissions
The national emissions of carbon dioxide may be computed from the production accounts by adding over the emissions from all fossil fuels \( f \). This is given by the quantity of fuel consumed \( (CF_{t,f}) \), multiplied by the energy content coefficient \( (e_f) \), and multiplied by the CO2 intensity, \( (c_f \text{, tons of CO2 per sce of fuel}) \):
\[
EM_{CO2,t}^{\text{fossil}} = c_{\text{coal}} e_{\text{coal}} C_{t}^{\text{coal}} + c_{\text{oil}} e_{\text{oil}} C_{t}^{\text{oil}} + c_{\text{gas}} e_{\text{gas}} C_{t}^{\text{gas}}
\]
(A119)
The quantity of fuel \( f \) consumed, \( CF_{t,f} \), is given in equation A107 above. For non-combustion sources of CO2 we only consider those from cement production processes; this is expressed as an emission factor \( (c_{\text{cement}}) \) multiplied by the cement component of the output of the Building Materials industry:
\[
EM_{\text{CO2},t}^{\text{noncomb}} = c_{\text{cement}} a_{\text{cement}} Q_{\text{Build},t}
\]
(A120)
where \( a_{\text{cement}} \) is cement’s share of the Building Materials industry’s output, and \( Q_{\text{Build}} \) is the value of the output of the Building Materials industry (in billion yuan) which also includes glass and clay products.
Total carbon emissions are then the sum of the fossil emissions and non-combustion ones:
\[
EM_{CO2,t} = EM_{CO2,t}^{\text{fossil}} + EM_{CO2,t}^{\text{noncomb}}
\]
(A121)
Local pollutants
Primary emissions of pollutant $x$ from sector $j$ at period $t$ ($EM_{jxt}$) are produced from fossil fuel combustion and from non-combustion production processes. The combustion emissions are obtained by multiplying the energy input by an emission factor, $\psi_{jxt}$, while the process emissions are output multiplied by the emission factor, $\sigma_{jxt}$. Total emissions from $j$ are thus:
\begin{equation}
EM_{jxt} = \sigma_{jxt}QI_j + \sum_f (\psi_{jxt} FT_{jf})
\end{equation}
\begin{equation}
j=1,\ldots,33
\end{equation}
$x =$ PM$_{10}$, SO$_2$, NO$_X$, $f =$ coal, oil, gas
where $QI_j$ is the output of industry $j$’s (in billion constant yuan2010) and $FT_j$ is the quantity of fuel $f$ combusted by industry $j$. The combustion emission factor ($\psi_{jxt}$) is given in tons of emissions of pollutant $x$ per ton of fuel, while the process emission factor ($\sigma_{jxt}$) is given in tons of primary emissions of pollutant $x$ per billion yuan of industry output.
Households’ use of fuels also generates pollutants:
\begin{equation}
EM_{HHT,xt} = \sum_f (\psi_{HHT,xft} FT_{HHT,f})
\end{equation}
The estimation of emissions in 2005 is reported in *Clearer Skies* (Chapters 4-6) and an updated version for 2010 is used to calibrate these emission factors. The emission factors are projected based on planning documents of the NDRC and other government agencies.
The emissions are then used by the GEOS-Chem atmospheric model to compute the concentration of various criteria pollutants at each grid cell as described in *Clearer Skies* (Chapter 7). We consider the impact of PM and ozone on human health, and concentrate on the main effects – mortality risks, hospital admission due to cardiovascular reasons and due to respiratory reasons, and outpatient visits. The health effect $h$ due to a change in concentration of $x$ ($\Delta C_x$) induced by a policy change is given by (*Clearer Skies* Chap. 8):
\begin{equation}
\Delta HE_{hx} = f_{hx}(\Delta C_x) \times Pop \times BI_h
\end{equation}
where $\Delta HE_h$ denotes the change in the number of cases of health endpoint $h$; $f$ is the C-R function; $Pop$ represents the population exposed to the pollutant; and $BI_h$ represents the baseline incidence of the health endpoint $h$. The total impact, say for mortality, is the sum over all pollutants for $h =$ mortality, $\Delta HE_h = \sum_x \Delta HE_{hx}$.
We also consider the impact of ozone on agriculture output. There is less agreement in the literature about how to model this impact, and as discussed in *Clearer Skies* (Chap. 8) we use three different measures
of ozone exposure (indices \( I_{O3}^i \), \( i=\text{SUM6, AOT40, W126} \)), to compute the impact on the output of maize, rice and wheat. The percentage change in yields is given by:
\[
(A125) \quad \Delta q_{\text{crop}, O3} = \frac{\Delta Q_{\text{crop}}}{Q_{\text{crop}}} = f_{O3}(\Delta I_{O3}^i) \times Q_{\text{crop}} \times B_{\text{crop}}; \quad \text{crop=rice, wheat, maize}
\]
The final step is to calculate the monetary value of these damages. The value is given by the health impact from (A125) multiplied by the willingness to pay value of each type of health effect \( V_{ih} \), and the value of crop damages is the value of the crop \( V_{\text{crop},i} \) multiplied by the percentage change in crop yields:
\[
(A126) \quad V_i = \sum_h \Delta HE_{ih} V_{ih} + \sum_{\text{crop}} V_{\text{crop},i} \Delta q_{\text{crop},i}
\]
A.2 Implementing tax and subsidy policies; environmental policies
In this we describe the implementation of the main policies studied with this model. We first describe carbon tax scenarios, one that is offset by a lump sum rebate to households and one that is offset by cut in existing taxes. We then describe electricity policies, one which subsidize clean energy and one with imposes Renewable Portfolio Standards.
A.2.1 Carbon Tax
We represent a simple carbon tax with revenue neutral recycling as a tax on fossil fuels in proportion to the carbon content of fuel \( j \) (\( c_j \)). We put the tax upstream – on the producers of these (primary) fuels and on the imports of all fuels; the tax per unit output (constant yuan2010) is the carbon content (tons of carbon per yuan) multiplied the carbon tax rate (\( t^w \), yuan per ton of C):
\[
A130 \quad t^w_{jt} = L^w c_j \quad j=\text{coal, crude, natgas}
\]
and the unit tariff on import of fuel commodity \( i \) is:
\[
A131 \quad t^w_{ii} = L^w c_i \quad i=\text{coal, crude, natgas, refine, gasprod}
\]
The unit tax \( t^w_{jt} \) is the term that appears on the right hand side of the purchaser’s price equation (A13), while the unit tariff is on the right side of the import price (A83).
The above simple setup means that all purchasers of fuel \( j \) will pay the carbon tax. A more complicated policy of imposing different rates on different users of fossil fuels will require a tax on intermediate input \( A_{ij} \).
If this tax is to be recycled as a lump sum transfer to households then this is represented by a negative tax, \( T^{LUMP} \), in income equation (A45) above. If this tax is to recycled as a tax cut we introduce a new endogenous variable, \( t^{scale} \), that is applied to all tax rates in period \( t \), e.g. the tax on capital income is this scaling variable multiplied by the base case rate (\( t^k_0 \)):
\[
A132 \quad t^k_t = t^{scale} t^k_0
\]
These offsetting tax variables are chosen to maintain revenue and spending neutrality; we choose the lump sum payment or the tax rate factor such that the real level of aggregate government consumption (GG) and deficit are the same as the base case value for each period:
\[
A134 \quad GG_t = GG^{base}_t; \quad \Delta G_t = \Delta G^{base}_t
\]
A.2.2 Renewables promotion policy
One common policy examined is the use of subsidies to promote renewables in electricity generation. In the base case there is a net output tax \((t_{l,t}^{EL})\) on electricity generated by source \(l\), giving the purchaser price, \(P_{l,t}^{EGEN}\), in equation (A21b). A new subsidy for renewables is represented by \(s_{b,t}^{EL}\).
\[
P_{b,t}^{E,GEN} = (1 + t_{b,t}^{EL} - s_{b,t}^{EL})P_{b,t}^{EGEN} + t_{l,t}^{EL,ext}; \quad b=\text{hydro, wind, solar, ...}
\]
The total payment for this new subsidy is expressed as a negative revenue from the electricity generation sector, \(R_{E,GEN}\):
\[
R_{E,GEN} = -\sum_{b} s_{b,t}^{EL} F_{b,t}^{E,GEN} Q_{b,t}^{E,GEN}
\]
This is added to the total revenue term in equation (A61):
\[
\text{(A61')} \quad \text{Rev} = \sum_{j} t_{j}^{KD} (P_{j}^{KD} D_{j} - D_{j}) + t_{j}^{PL,LS} + t_{j}^{V} \sum_{l} (P_{j}^{KD} D_{j} + P_{j}^{L} D_{j} + P_{j}^{T} D_{j})
\]
\[
+ \sum_{j} t_{j} Q_{j}^{I} + R_{EXT} + \sum_{i} t_{i}^{PM} M_{i} + \sum_{i} t_{i}^{F} (Q_{i} - X_{i} + M_{i}) + FEE + T^{LUMP}
\]
\[
+ R_{E,GEN}
\]
A renewable portfolio standard (RPS) policy requires a certain minimum share of total electricity output to come from renewable sources. In practice, this may be imposed on each company, or imposed on a region and allowing companies to meet the target collectively. We represent the RPS policy in this model as requiring the national share of the kWh from source \(l\) be no less than some target, \(t_{l,t}^{RPS}\):
\[
\frac{Q_{l,t}^{Wh}}{Q_{TOT,t}^{Wh}} \geq t_{l,t}^{RPS}
\]
The RPS may be implemented by subsidies or by allowing producers to pass the higher costs to consumers. A simple way to implement this is a system of taxes and subsidies to the different generators such that the net fiscal burden on the government is zero, and the consumer bears the net impact on the average price of electricity. Let \(t_{l,t}^{RPS}\) be the tax on the output from source \(l\); this is negative if it is a subsidy. The price equation (A21b) is then rewritten as:
\[
P_{l,t}^{E,GEN} = (1 + t_{l,t}^{EL} - t_{l,t}^{RPS})P_{l,t}^{EGEN} + t_{l,t}^{EL,ext}; \quad l=\text{coal, gas, hydro, wind, ...}
\]
Producers will change output given this new vector of prices according to the cost functions given in A17, A19 and A21. This is illustrated in Figure A3.
Revenue neutrality requires:
\[ (A139) \sum t^r_{l,s} P^{GEN}_{l,t} Q^{GEN}_{l,t} = 0 \]
We thus have potentially 8 rates for the \( t^r_{l,s} \), to hit a maximum of 8 independent target shares for the different sources of power; only relative prices matter and the shares must sum to 1.
In this diagram the arc represents the production possibility frontier for coal versus wind. In the base case the producer, and consumer, relative price for coal versus wind is \( \frac{P^0_c}{P^0_w} \) and output and consumption is at point B. The small arcs represent the iso-utility curves of eq. A17. A subsidy for wind of \( s_w \) will move the relative price to \( \frac{P_c}{P_w} \) and production to point S. The price to the consumer becomes \( \frac{(1+s_w)P_c}{P_w} \).
A.3 Parameterizing the electricity sub-model
The parameters of the electricity sub-model are based on three main sources of data – the Input-Output table, the Electric Power Industry Statistics,\(^{21}\) and the International Energy Agency (2010) “Projected Costs of Generating Electricity”. The elasticities are summarized in Table A2
A.4 Parameters, exogenous variables and data sources
The key input into the model is the Social Accounting Matrix (SAM) for 2010. This traces the flow of commodities and payments among the producers, household, government and rest of the world. The SAM is assembled from the 2007 benchmark input output table.\(^{22}\) A summary of this SAM is given in Figure A4, the actual matrix used is disaggregated to the 33 sectors and commodities. From this we derive the labor and capital incomes, the tax revenues for each type of tax, the expenditures on specific commodities by the household, government and foreign sectors, and government payments of all types in equation A76.
These payments are combined with employment and capital input data to give the compensation rates for labor and capital for each sector. The estimates for employment and capital stocks by sector are taken from a productivity study of China (Cao, Ho, Jorgenson, Ren, Sun and Yue 2009) that supplements the official data with labor force surveys. The various tax and subsidy rates are not statutory rates but are implied average rates derived by dividing revenues by the related denominator – value of industry output, capital income, total value added, and imports.
The exogenous variables in the model include total population, working age population, saving rates, dividend payout rates, government taxes and deficits, world prices for traded goods, current account deficits, rate of productivity growth, rate of improvement in capital and labor quality, and work force participation. These variables may, of course, be endogenous (i.e. they interact among each other) but we ignore this and specify them independently. Our assumptions for these exogenous drivers are summarized in Table A3.
The assumption that affects the growth rate the most is the household savings rate, \(s\). Our assumption is to have \(s\) beginning at the observed 41.2% for 2010 and gradually falling to 22% in 2020 and 15% in 2050. National private savings is household savings plus the retained earnings of enterprises.
---
\(^{21}\)电 力 工 业 统 计 资 料 汇 编.
\(^{22}\) The 2010 input-output table is given in NBS (2014). The benchmark IO table for 2007 is derived from detailed enterprise data, the 2010 IO table is extrapolated by the NBS using simpler aggregated data.
The share of retained earnings is assumed to fall, and dividend payouts to rise to reflect the diminishing role of state enterprises in the economy. The dividend rate, i.e. the share not used for retained earnings, was 38.9\% in 2010 and we project it to rise to 58\% by 2020. It should be pointed out that national savings and investment in the Chinese data includes capital such as roads and other public infrastructure, items that are excluded from the “gross fixed private investment” item in most other countries National Accounts.
In the labor supply expression eqn. (A47) we have the product of the working-age population, annual average hours and quality. In Cao and Ho (2014) we discussed various population growth scenarios including the different two-child policies. Projections by age groups are taken from projections made by the Population Division of the Department of Economic and Social Affairs of the United Nations Secretariat\(^{23}\) and in the central population scenario we adjust them to incorporate a looser child policy. The results are plotted in Figure A5.
The composition of the work force changes over time with a bigger portion of educated workers, bigger or smaller portion of more experienced workers, and an older average age. This quality of labor input index, \(q_t^L\), in estimated in Cao et al. (2009) to have grown at 0.9\% per year for the period 1983-2000. Given the expectation of continued higher educational attainment in the future we assume that China's aggregate labor quality continue to rise, but at a diminishing rate. By 2040 the quality index is assumed to grow at only 0.2\% per year. For comparison, the U.S. labor quality growth peaked at 0.5\% during the 1960s, and fell to 0.3\% per year during 1995-2000 (Jorgenson, Ho and Stiroh, 2005, Table 6.5).
Total labor hours depend also on the participation rate and annual working hours. There is no comprehensive data on the number of hours worked and based on comparisons to other countries we project it to rise due to improvements in the functioning of the labor market -- lower underemployment, seasonal unemployment and other labor market frictions. We assume that hours worked per capita rises at 0.2\% per year initially but slowing down over time. The results are plotted in Figure A6.
We allow for improvements in future capital “quality,” or composition, as represented by the \(\psi/\) coefficient in (A73). Cao et al. (2009) note how the composition of the capital stock in China has shifted towards assets with shorter life, i.e. towards a smaller share of structures and a larger share of equipment such as computers. They explain how assets which have shorter useful lives generate higher annual capital services per dollar of capital stock, and hence is of a higher quality in the terminology of Jorgenson, Ho and Stiroh (2005). While the ratio of equipment to structures has moved in different directions over the past 30 years, we believe it will return to a more typical development trend of rising equipment ratios. We project that capital quality rises by 1.5\% per year initially, then gradually decelerating. For land, the supply
\(^{23}\) The demographic projections are from the U.N. Population Division’s World Population Prospects: 2012 Revision, downloaded from their web site, http://esa.un.org/unpd/wpp/unpp/panel_population.htm .
of land for agriculture, oil mining and gas mining is simply set fixed for all periods equal to the base year value.
Tax rates are set equal to those for 2010 derived from the SAM. These are summarized in Table A4. For the government deficit, $\Delta G$, we set it at the base year 1.69% of GDP initially, declining steadily towards zero in the long run. These deficits are cumulated into the stocks of domestic and foreign debt, $B_t$ and $B_t^{Gr}$, assuming a constant division between domestic and foreign borrowing. Data for the stock of debt and interest paid on it comes from the China Statistical Yearbook (NBS 2012, Table 8-13), IMF’s *World Economic Outlook 2012* online database and the 2010 Social Accounting Matrix. Government transfers, $G_{\text{transfer}}$, are set to rise in proportion with population and average wage. The nontax fees paid by enterprises are set to be a fixed share of GDP equal to the base year’s share (Table A4).
The current account balance was in a huge surplus in the mid-2000s but has since declined. There is no consensus about the future evolution of this variable, for simplicity, after setting it as a share of GDP at the observed sample period values, we set it to decline rapidly to zero. This $CA_t$ deficit is also the assumed rate of borrowing from the world. Import prices, $PM_t^*$, are assumed fixed at the base year value for every period with one important exception. World oil price forecasts are taken from the U.S. Energy Information Administration and shown in Figure A7. The model also requires projections of the export share; while this has been rising rapidly in the past, it fell with during the 2008 Global Financial Crisis and never recovered. Given the aim to rebalance the economy away from exports and investment towards consumption, we simply project a constant value for the share parameter.
The base year data for 2010 was constructed in 2013 since then, the macro variables for 2011-13 is now available; these include the GDP, investment and current account surplus. The current account surplus has fallen, and the unusually high share of investment in GDP has risen even more after 2010. We take these into account in setting the savings rate and current account balance as share of GDP for these years.
*Parameters*
The rate of productivity growth is another factor that has a large effect on the base case growth rate of the economy but has little impact on the difference between cases. Total Factor Productivity growth at the industry level in the 1982-2000 period show a very wide range of performance as estimated by Cao et al. (2009), ranging from -10% to 5% per year. The Domar-weighted productivity growth for all industries was 2.7% for 1982-2000. To keep the base case as simple as possible we ignore this wide range of observed
---
24 The projections for crude oil prices are taken from the EIA’s *Annual Energy Outlook 2010*, Table 12, which is available on their web page: [http://www.eia.gov/forecasts/aeo/](http://www.eia.gov/forecasts/aeo/).
TFP growth, and in our projections of sector productivity terms in eqn. (A3) we initially set all the $\mu_j$'s to the same value, 0.018. These are then adjusted to match actual GDP growth rates in the initial years for which we have actual data.
The value share parameters of the production functions ($\alpha_{Kj}$, $\alpha_{Lj}$, etc.) are set to the values in the 2010 IO table in the first year of the simulation. For future periods we change most of these parameters so that they gradually resemble the shares found in the US input output table for 1997. The exceptions to this are the coal inputs for all the sectors, this is set to converge to a value between current Chinese and US1997 shares.\(^{25}\) The rate of reduction in energy use is set at a modest level relative to the rapid improvements in the recent Chinese history. We assume that the share of energy in industry output is reduced gradually to 60% of the 2005 levels in 40 years. This is conservative compared, for example, to the performance in the electric power industry during the 1990-99 period. In that time the thermal output grew 88% whereas coal input only rose 61%, a rate of improvement of some 1.5% per year.\(^{26}\)
The $\alpha_{it}^C$ parameters of the consumption function are set in a similar way. That is, for the first period they are equal to the shares in the 2010 Social Accounting Matrix, and for the future periods they gradually approach US 1997 shares except for coal. This implies a higher projected demand for private vehicles and gasoline than that assumed in most other models of China. The coefficients determining demand for different types of investment goods ($\alpha_{it}^I$), and different types of government purchases ($\alpha_{it}^G$), are projected identically.
The import and export elasticities are set to the values in GTAP v4. The base share of exports and imports are taken from the SAM.
\(^{25}\) We have chosen to use U.S. patterns in our projections of these exogenous parameters because they seem to be a reasonable anchor. While it is unlikely that China’s economy in 40 years time will mirror the U.S. economy of 1997, it is also unlikely to closely resemble any other economy. Other projections, such as those by the World Bank (1994), use the input-output tables of developed countries including the U.S.
\(^{26}\) China Energy Statistics Yearbook 1997-1999, Tables 4-5 and 4-15.
References
Cao, Jing, Mun Ho, Dale Jorgenson, Ruoen Ren, Linlin Sun, Ximing Yue. 2009. “Industrial and Aggregate Measures of Productivity Growth in China, 1982-2000,” *The Review of Income and Wealth*, 55(s1):485-513, July.
Cao Jing and Mun S. Ho. 2014. “China: Economic and GHG emissions Outlook to 2050”. The Chinese version of this is in 重塑未来: 未来20年的中国与世界, 清华大学能源环境经济研究所 (Reshaping the Future: China and the World in the next 20 years), Tsinghua University, Institute for Energy and Environmental Economics.
Jorgenson, Dale, Richard Goettle, Mun Ho and Peter Wilcoxen. 2013. *Double Dividend: Environmental Taxes and Fiscal Reform in the US*, MIT Press, Cambridge, MA.
Jorgenson, Dale, Mun S. Ho and Kevin Stiroh, 2005. *Information Technology and the American Growth Resurgence*, MIT Press, Cambridge, MA.
Ho, Mun S., and Chris P. Nielsen, eds. 2007. *Clearing the Air: The Health and Economic Damages of Air Pollution in China*. Cambridge, MA: MIT Press.
International Energy Agency (IEA). 2010. *Projected Costs of Generating Electricity*, IEA and OECD, Paris.
International Energy Agency (IEA). 2014. *World Energy Outlook 2014*, IEA, Paris.
National Bureau of Statistics (NBS). 2012. *China Statistical Yearbook 2012*, Beijing.
National Bureau of Statistics (NBS). 2014. *China Statistical Yearbook 2014*, Beijing.
Chris Nielsen and Mun Ho (eds.). 2013. *Clearer Skies over China: Reconciling Air Quality, Climate and Economic Goals*, MIT Press, Cambridge, MA.
Paltsev, Sergey, John Reilly, Henry Jacoby, Richard Eckaus, James McFarland, Marcus Sarofim, Malcolm Asadoorian and Mustafa Babiker. 2005. The MIT Emissions Prediction and Policy Analysis (EPPA) Model: Version 4. MIT Joint Program on the Science and Policy of Global Change, Report No. 125, August.
Rutherford, Thomas. 2003. Lecture notes on constant elasticity functions. Mimeo, University of Colorado.
Klump, Rainer, Peter McAdam and Alpo Willman. 2011. The Normalized CES Production Function: Theory and Empirics. European Central Bank, Working Paper 1294.
Sue Wing, Ian, Kathryn Daenzer, Karen Fisher-Vanden and Katharine Calvin. 2011. Phoenix Model Documentation. PNNL Joint Global Change Research Institute. Available at: http://www.globalchange.umd.edu/models/phoenix/
Qi, Tianyu, Niven Winchester, Da Zhang, Xiliang Zhang and Valerie Karplus. 2014. The China-in-Global Energy Model. MIT JPSPGC Report No 262, May.
United States Energy Information Administration (U.S. EIA). 2013. Annual Energy Outlook 2013. Available at http://www.eia.gov/forecasts/archive/aeo13/.
Vennemo, Haakon, Jianwu He and Shantong Li. 2014. Macroeconomic Impacts of Carbon Capture and Storage in China. Environmental and Resource Economics, 59(3):455-477.
Table A1. Selected Parameters and Variables in the Economic Model
| Parameters | Description |
|-----------------------------|-----------------------------------------------------------------------------|
| $S^e_i$ | export subsidy rate on good $i$ |
| $t^c_i$ | carbon tax rate on good $i$ |
| $t^k$ | tax rate on capital income |
| $t^L$ | tax rate on labor income |
| $t^r_i$ | net import tariff rate on good $i$ |
| $t^t_i$ | net indirect tax (output tax less subsidy) rate on good $i$ |
| $t^x$ | unit tax per ton of carbon |
| Endogenous Variables | Description |
|-----------------------------|-----------------------------------------------------------------------------|
| $G_I$ | interest on government bonds paid to households |
| $G_{INV}$ | investment through the government budget |
| $G_{IR}$ | interest on government bonds paid to the rest of the world |
| $G_{\text{transfer}}$ | government transfer payments to households |
| $P^{KD}_i$ | rental price of market capital by sector |
| $PE^*_i$ | export price in foreign currency for good $i$ |
| $PL_i$ | producer price of good $i$ |
| $PL'_i$ | purchaser price of good $i$ including taxes |
| $PL$ | average wage |
| $PL_i$ | wage in sector $i$ |
| $PM^*_i$ | import price in foreign currency for good $i$ |
| $PM'_i$ | import price in domestic currency for good $i$ |
| $PS_i$ | supply price of good $i$ |
| $PT_i$ | rental price of land of type $i$ |
| $QI_i$ | total output for sector $i$ |
| $QS_i$ | total supply for sector $i$ |
| $r(B')$ | payments by enterprises to the rest of the world |
| $R_{\text{transfer}}$ | transfers to households from the rest of the world |
Table A2. Reference values for elasticities of substitution in the Electricity sector
| Description | Value | Sectors/nodes |
|-------------|-------|---------------|
| $\sigma^{ED}$ | Transmission-Generation | 0.7 Electric Utilities |
| $\sigma^{EG}$ | Baseload-Wind/Solar | 1.0 Generation |
| $\sigma^{BL}$ | Among base load sources | 4.0 Baseload Generation |
| $\sigma^{RE}$ | Wind-Solar | 4.0 Wind and Solar Generation |
| $\sigma^{QI}$ | Materials-VE | 0.7 Transmission |
| $\sigma^{VE}$ | Energy-Value added | 0.5 Transmission |
| $\sigma^{VA}$ | Capital-Labor | 1.0 Transmission |
| $\sigma_E^{TD}$ | Among energy inputs | 0.5 Transmission |
| $\sigma_M^{BL,c}$ | Materials-VE | 0.1 Coal, coal-ccs, gas, gas-ccs |
| $\sigma_V^{BL,c}$ | Energy-Value added | 0.5 All generation sources |
| $\sigma_A^{BL,c}$ | Capital-Labor | 0.4 All generation sources |
| $\sigma_E^{BL,c}$ | Coal-Noncoal | 0.25 Coal generation, coal-ccs |
| $\sigma_N^{NC}$ | Among noncoal energy | 0.5 Coal generation, coal-ccs |
| $\sigma_E^{BL,g}$ | Gas-Nongas | 0.25 Gas generation, gas-ccs |
| $\sigma_N^{NG}$ | Among nongas energy | 0.5 Gas generation, gas-ccs |
| $\sigma_{seq}^{coalccs}$ | Energy-Sequestration tech. | 0 Coal-ccs |
| $\sigma_{seq}^{gassccs}$ | Energy-Sequestration tech. | 0 Gas-ccs |
| $\sigma_M^{BL,b}$ | Materials-VR | 0.1 Nuclear, hydro, wind, solar, other |
| $\sigma_V^{BL,b}$ | Resource-VE | 0.4 Nuclear, hydro, other |
| $\sigma_V^{BL,b}$ | Resource-VE | 0.25 wind, solar |
| $\sigma_E^{BL,b}$ | Among energy inputs | 0.5 Nuclear, hydro, wind, solar, other |
| $\epsilon_r^R$ | Resource supply | 2.5 Nuclear |
Table A3. Parameters of base case growth path
| Year | Savings rate | Dividend rate | Population | Work force | Labor input (quality adjusted) | Total Factor Productivity index |
|------|--------------|---------------|------------|------------|--------------------------------|-------------------------------|
| 2010 | 41.2% | 38.9% | 1360 | 938 | 100.0 | 100.0 |
| 2020 | 22.3% | 57.9% | 1440 | 930 | 108.5 | 111.5 |
| 2030 | 17.0% | 63.2% | 1470 | 884 | 108.5 | 122.8 |
| 2040 | 15.3% | 64.9% | 1466 | 838 | 105.9 | 133.9 |
| 2050 | 14.6% | 65.5% | 1434 | 758 | 97.6 | 144.6 |
Table A4. Miscellaneous Tax Rates and Coefficients
| Tax rate on capital income | tk | 0.0805 |
|---------------------------|--------|--------|
| Indirect tax rate on output | tt | 0.0 to 0.033 |
| VAT rate | tv | 0 to 0.199 |
| Import tax rate | tr | 0 to 0.198 |
| Nontax payment share | \(\gamma^{\text{NENT}}\) | 0.0246 |
| Govt transfer rate | \(\gamma^{\text{fr}}\) | 0.2846 |
| Household savings rate (2010) | 0.4125 |
| Dividend payout rate (2010) | 0.3889 |
Figure A4. Summary Social Accounting Matrix for China, 2010 (bil yuan)
| Commodity | Total | Capital account | Total | Capital account |
|--------------|-------|----------------|-------|----------------|
| Commodity | 148659| | 148659| |
| Industry | 137141| | 137141| |
| Labor | 16687 | | 16687 | |
| Capital | 15726 | | 15726 | |
| Land | 1619 | | 1619 | |
| Households | 16687 | | 16687 | |
| Enterprise | 5436 | | 5436 | |
| VAT+BT | 566 | | 566 | |
| Government | 291 | | 291 | |
| Tariff | 1252 | | 1252 | |
| ROW | 10416 | | 10416 | |
| Capital a/c | 9853 | | 9853 | |
| Addendum: GDP=| 40151 | | 40151 | |
Figure A5. Projections of Population (millions)
Figure A6. Projections of work force
Note: Projection taken from US EIA(2013).
Appendix B. Electricity Sector Parameterization and Projection
B.1 Electricity Sector Parameterization and Data Construction
This Appendix describes the construction of the electricity sector data set with the various generation technologies and how that is integrated with the rest of the economic accounts of the China model that is described in the Model Appendix.
The model is based on a Use matrix with 33 commodities and 33 industries and a Make (or Supply) matrix with 33 industries and 33 commodities (Table B5). Sector 22 is the “production and supply of electric power and heat power”, which we label as Electricity & heat. This includes both generation and distribution, including suppliers of steam and hot water, and combined heat-and-power units. The electricity column in the Use matrix gives the values of inputs into the sector including the 33 intermediate inputs, labor and capital inputs. It also gives the value of taxes, net of subsidies, paid by that sector.
The Social Accounting Matrix (SAM) for 2010 was constructed using the official 2007 benchmark input-output matrices (Use and Make) and rebalanced to match the 2010 values for GDP, the final demand components (CIGXM), industry value added, industry gross output and government tax receipts. The key values of this 2010 SAM are given in Figure A4 of the Model Appendix.
Output and prices in electricity sector
The task here is to disaggregate this electricity and heat sector into the various generation technologies and “transmission & distribution”. This is made difficult by the lack of data on prices and yuan values even though there are good data on the kWh output quantities and installed capacity (in GW). It is also made difficult because the measures of output in the National Accounts are not reconciled with in the data from the electric power industry sources.
The first step is to collect the quantity data on the generation capacity and power generated in 2010 and more recent years. The data before 2011 is assembled by LBNL (2013) from various sources of information in China including the NDRC and State Grid companies (see also NBS (2014) Tables 9-6 and 9-15) and we obtain more recent information from China Electricity Council (2014). The capacity and output data are presented in Table B1. The “Others” category includes oil, biomass, geothermal, etc.
27 Electricity Power Industry Statistics, China Electricity Council, Beijing (2014). (电力工业统计资料汇编)
In 2010, of the 966GW of total generating capacity, 68.3% was coal and 22.4% was hydro. The operating hours are quite different for the different technologies and of the 4228 terawatt-hours of power output, 77.1% is from coal and only 16.2% from hydro. Wind power has been rising rapidly with a doubling of output between 2010 and 2012. Solar capacity was negligible in 2010 but reached 28GW in 2014.
Table B1. Electric power sector characteristics
| Capacity (GW) | 2010 | 2011 | 2012 |
|---------------|------|--------|--------|
| Coal | 660.0| 713.3 | 758.8 |
| Gas | 27.0 | 35.0 | 37.4 |
| Nuclear | 10.8 | 12.6 | 12.6 |
| Hydro | 216.1| 233.0 | 249.5 |
| Wind | 29.6 | 46.2 | 61.4 |
| Solar | 0.3 | 2.2 | 3.4 |
| Others | 22.7 | 20.1 | 23.6 |
| Total | 966.4| 1062.4 | 1146.8 |
| Output (TWh, billion kWh) | 2010 | 2011 | 2012 |
|---------------------------|------|--------|--------|
| Coal | 3258.9| 3703.1 | 3719.3 |
| Gas | 78.7 | 111.0 | 109.5 |
| Nuclear | 74.7 | 87.2 | 98.3 |
| Hydro | 686.7| 668.1 | 855.6 |
| Wind | 49.4 | 74.1 | 103.0 |
| Solar | 0.1 | 0.7 | 3.6 |
| Others | 79.1 | 86.3 | 97.2 |
| Total | 4227.7| 4730.5 | 4986.5 |
The next step is to collect prices for each generation technology in order to estimate the output values. There is the feed-in tariff (or on-grid price, 上网电价) and a distribution price (输配电价) which varies by generation technology and location. The average generation price was 0.38 yuan/kWh in 2010 while the distribution price is 0.16 yuan/kWh for the National Grid and 0.20 for the Southern Grid.\(^\text{28}\) The average end-user price was thus 0.58 yuan/kWh (0.38+0.20) in
\(^{28}\) China Times May 15, 2013, given at www.wantchinatimes.com/news-subclass-cnt.aspx?id=20130515000080&cid=1102.
2010. There is no official data on average prices for each generation technology; there are some national benchmark prices set by the NDRC, for example, the price for nuclear power was set at 0.43 yuan/kWh in 2013, and there are three pricing regions for solar power set at 0.9, 0.95 and 1.0 yuan/kWh respectively, in 2013.
Table B2. Costs and prices for electricity in China
a) Feed-in tariffs (yuan/kWh)
| Year | Coal | Gas | Hydro | Nuclear | Wind | Solar |
|------|------|-----|-------|---------|------|-------|
| 2009 | 0.36 | 0.55| 0.28 | 0.40 | 0.53 | 1.08 |
| 2010 | 0.41 | 0.56| 0.30 | 0.41 | 0.56 | 1.08 |
| 2011 | 0.44 | 0.56| 0.31 | 0.41 | 0.58 | 1.08 |
b) Levelized costs (LCOE) from IEA(2010), selected technologies, 5% discounting option
| Technology | Lifetime (years) | Load factor | LCOE US$/MWh | Fuel | O&M | Capital US$/MWh | LCOE yuan/kWh |
|-----------------------------|------------------|-------------|--------------|------|-----|-----------------|---------------|
| Nuclear; CPR-1000 | 60 | 0.85 | 30.0 | 9.33 | 7.18| 13.4 | 0.186 |
| Super critical coal 1119MW | 40 | 0.85 | 29.5 | 23.1 | 1.54| 4.99 | 0.183 |
| Comb. cycle gas 1358MW | 30 | 0.87 | 35.8 | 28.1 | 2.85| 4.90 | 0.222 |
| Hydro 6277MW | 80 | 0.34 | 16.9 | 0 | 2.55| 14.3 | 0.105 |
| Onshore wind 35MW | 25 | 0.22 | 83.2 | 0 | 23.3| 60.0 | 0.516 |
| Solar 10MW | 25 | 0.18 | 186 | 0 | 18.0| 169 | 1.153 |
c) Wind power characteristics
| Capacity | Levelized costs (IEA) | Capacity | Number |
|----------|-----------------------|----------|--------|
| MW | US$/MWh | yuan/kWh | MW | in 2012 |
| 200 | 51.0 | 0.32 | 200+ | 17 |
| 50 | 64.2 | 0.40 | 100-150| 109 |
| 35 | 83.2 | 0.52 | 50 | 1217 |
| 30 | 89.0 | 0.55 | 20-30 | 801 |
| Mean cost| 67.5 | 0.42 | | |
d) Levelized costs with carbon capture from IEA (2010)
| Capture rate (%) | LCOE | Ratio of CCS to Reference plant |
|------------------|------|---------------------------------|
| Coal-Chem absorption 2030 | 85 | 1.49 |
| Coal-Oxy combustion, 2030 | 90 | 1.55 |
The main source of data are the detailed plant-by-plant data reported by the NDRC\textsuperscript{29} and reproduced in LBNL (2013, Tables 6B.12-6B.18) for 2008, 2009 and 2011. We compute the simple average of these plant level on-grid prices separately for coal, gas and hydro plants and these are reported in the top section of Table B2. The prices for 2010 are interpolated between 2009 and 2011 that we describe in greater detail below.
Table B2 also gives the average benchmark prices for nuclear, wind and solar.\textsuperscript{30} Some data for 2013 are deflated back to 2010 using the PPI for electricity. We can see that the prices paid by the state grid companies vary substantially by the type of technology, from 0.30 yuan for hydro to 1 yuan for solar PV. For onshore wind, four different prices are allowed for different regions: 0.51, 0.54, 0.58, 0.61 yuan/kWh respectively for Regions I, II, III and IV. Prices have generally been rising over time except for solar where the benchmark prices have fallen from very generous levels.
To provide a comparison to these estimates of average feed-in tariffs, we also report in Table B2(b) the cost estimates in IEA (2010), \textit{Projected Costs of Generating Electricity}. This provides the levelized costs for various power generation technologies in many countries of the world, including China. We report only a sample of the technologies estimated in IEA (2010), and report the three components of the total levelized costs (LCOE) – fuel, operation & maintenance, and capital costs. The IEA report calculated LCOEs using two different rates of discount, 5% and 10%; we only report the 5% set here. The US$ estimates are converted to yuan at an exchange rate of 6.2 yuan/$.
The cost for the 1GW super critical coal plant in IEA (2010) is substantially lower than the average price allowed for the stock of existing plants in China which includes many smaller and older units, 0.18 versus 0.4. Similarly, the average feed-in tariff of hydro power in China is
\begin{table}[h]
\centering
\begin{tabular}{|c|c|c|}
\hline
Technique & Year & Price \\
\hline
Coal-IGCC-Selexol & 2030 & 85 \\
Gas-CC, Chem absp & 2030 & 85 \\
\hline
\end{tabular}
\end{table}
\textsuperscript{29} \textit{NDRC Price Notices for Electricity}. For example, the Southern Grid Price Notice for 2011 (\textit{调整南方电网电价的通知}，\textit{发改价格}\textsuperscript{[2011]}\textsuperscript{2618号}) is given at:
\url{http://www.ndrc.gov.cn/fzgggz/jggl/zcfg/201112/t20111201_748381.html}
\textsuperscript{30} These controlled price settings are taken from China Climate Change Info-Net at\url{http://www.ccchina.gov.cn/Website/CCChina/UpFile/File443.pdf}, and Xinhua News at\url{http://news.xinhuanet.com/finance/2013-07/09/c_124978789.htm}
much higher than the large-unit cost estimated by IEA. The cost for wind and solar are quite close to the Chinese benchmark prices.
Wind power is expected to be expanded rapidly under the current government plans and we provide more information in Table B2(c). The first two columns give the levelized costs estimated in IEA (2010) for various wind turbine capacities under the 5% discount option. These range from 0.32 to 0.55 yuan/kWh. The last two columns give the size distribution of the wind turbines in 2012.31 The most common type is the 50MW turbine. Using the IEA estimated costs, the average wind cost for that distribution of sizes is 0.42 yuan/kWh. This is also substantially lower than the average feed-in tariffs which depend on the region where the turbine is located. That is, the tariff takes into consideration the wind conditions and not just the cost of operating the turbine.
For technologies that are not currently used but might be considered in the future, especially if there is a carbon emission policy, we report the estimated costs of generation with carbon capture and sequestration (CCS) in part (d) of Table B2. This gives the estimated cost of a plant with CCS relative to a reference plant without CCS. The IEA (2010) projections give estimated costs for 2015, 2020 and 2030 and we report only the 2030 estimates to give readers an idea of the magnitudes involved. For coal with Chemical absorption with a 85% CO2 capture rate, the cost ratio is projected to be 1.49, while the gas combined cycle plant with CCS has a lower cost ratio of 1.23.
With these estimates of the average feed-in tariffs of the various technologies in Table B2, and the output data in Table B1, we estimate the revenues. These output values are computed in order to be reconciled with the values in the Social Accounting Matrix. The values derived from the prices in Table B2 are given in the first column of Table B3. The value of coal power is 78.9% of the total 1,595 billion yuan. Since the price of hydro power is the lowest, the value of hydropower is only 12.4% of total output even though it is 16.2% of the TWh. The other sources of power were less than 3% of the total in 2010.
Turning back to Table B1, we see that the gross output of Electricity and Heat in the input-output table is 3,236 billion. If we take the total of 1595 billion as the generators’ revenues, then the residual for transmission would be 1641 billion yuan (Table B3b), which is larger than the generating sector. As we noted above, in 2010, the average feed-in tariff was 0.38 yuan/kWh.
---
31 This data is given in the 2014 Report on Wind Power Market.
and the average transmission price was 0.20 giving a total end-user price of 0.58 yuan/kWh. That is, the average price for the transmission is quite a bit smaller than the average price paid to the generators.
The national accounts are most likely more comprehensive and take into account aspects of the electric power sector that is not included in these prices. To reconcile these two sets of data, we take a simple approach and scale the feed-in tariffs in Table B2 upwards so that the ratio of generator revenue to total generator plus transmission revenue is 0.38:0.58 as shown in the last two columns of Table B3b. This rescaled total generator revenue of 2121 billion yuan is then applied to the individual technologies and the revised output values are given in the third column of Table B3.
| Table B3. Value of output of power generators in 2010 (billion yuan). |
|---------------------------------------------------------------|
| **Value using prices in Table B2(a) Rescaled** |
| **bil.yuan** | **%** | **bil.yuan** | **%** |
| Coal | 1258.4 | 78.9% | 1673.7 |
| Gas | 41.7 | 2.6% | 55.4 |
| Nuclear | 28.9 | 1.8% | 38.5 |
| Hydro | 197.3 | 12.4% | 262.4 |
| Wind | 26.2 | 1.6% | 34.8 |
| Solar | 0.1 | 0.0% | 0.2 |
| Others | 42.4 | 2.7% | 56.4 |
| **Total** | 1595 | 100% | 2121.3 |
| Table B3b. Gross output and prices of Electricity sector in 2010 |
|---------------------------------------------------------------|
| **Price** | **Value** | **Rescaled value** |
| ¥/kWh | % | bil yuan | % | bil yuan | % |
| Generators | 0.38 | 65.5% | 1595 | 49.3% | 2121 | 65.5% |
| Transmission | 0.20 | 34.5% | 1641 | 50.7% | 1115 | 34.5% |
| **Total Electricity** | 0.58 | 100% | 3236 | 100% | 3236 | 100% |
**Input-output of electricity sector**
With these outputs and prices of each generation technology, we can now disaggregate the Electricity & Heat sector of the Use matrix into the 7 technologies and electric power distribution. The Use column gives the inputs into this sector and is given in the first column of Table B4. We distribute this column to the 7 generation types and distribution in two steps; first divide the Use column into a Generation column and a Distribution column; then divide the Generation column into 7 technologies. We proceeded in the following manner.
First, set the target for total generation to 2,121 billion and for distribution to 1,115 as calculated in Table B3b. Second, allocate the coal mining, gas mining, and oil mining inputs entirely to Generation. Third, to allocate electricity input, we turn to the energy consumption data by industry. The consumption of electrical power by the Electricity & Heat sector is reported to be 568.8 TWh in 2010 and transmission losses are 256.8 TWh compared to the national total consumption of 4,193 TWh.\(^3\) This means that transmission losses were 45.2% of power consumed by the electricity generation and distribution sector. We thus allocate 45.2% of the Use(electricity, electricity) cell to the Distribution column and the remainder to the Generation column. Next, we allocate the total trade and transportation margins used by
\(^3\) China Statistical Yearbook 2011, Tables 7-6 and 7-9.
Table B4. Input-output of electricity sector, 2010 (billion yuan)
| Industry | Total electricity | Generation | Distribution |
|---------------------------|-------------------|-------------|--------------|
| Agriculture | 0.55 | 0.29 | 0.26 |
| Coal mining | 686.55 | 686.55 | 0.00 |
| Oil mining | 24.38 | 24.38 | 0.00 |
| Gas mining | 24.33 | 24.33 | 0.00 |
| Nonenergy mining | 8.79 | 4.66 | 4.13 |
| Food | 22.60 | 11.98 | 10.61 |
| Textile | 0.85 | 0.45 | 0.40 |
| Apparel | 22.04 | 11.69 | 10.35 |
| Lumber | 4.38 | 2.32 | 2.06 |
| Paper | 11.45 | 6.07 | 5.38 |
| Refining & coal prod | 216.11 | 144.79 | 71.32 |
| Chemicals | 21.01 | 11.14 | 9.87 |
| Nonmetallic mineral | 14.45 | 7.66 | 6.79 |
| Primary metals | 18.66 | 9.89 | 8.76 |
| Fabricated metal | 27.18 | 14.41 | 12.77 |
| Machinery | 69.53 | 36.87 | 32.66 |
| Transportation equip | 85.63 | 45.41 | 40.22 |
| Electrical mach. | 331.35 | 175.71 | 155.63 |
| Electronics | 5.07 | 2.69 | 2.38 |
| Instruments | 107.18 | 56.84 | 50.34 |
| Other manufacturing | 7.53 | 3.99 | 3.54 |
| Electricity & Heat | 182.89 | 100.30 | 82.58 |
| Gas utilities | 8.62 | 8.62 | 0.00 |
| Construction | 2.54 | 1.35 | 1.19 |
| Transportation equip | 51.86 | 38.89 | 12.97 |
| Communication | 29.97 | 15.89 | 14.08 |
| Trade | 49.17 | 36.88 | 12.29 |
| Hotel & restaurants | 14.52 | 7.70 | 6.82 |
| Finance | 218.43 | 115.83 | 102.60 |
| Real Estate | 2.94 | 1.56 | 1.38 |
| Business services | 66.07 | 35.04 | 31.04 |
| Services | 89.96 | 47.70 | 42.25 |
| Public admin. | 0.76 | 0.41 | 0.36 |
| Labor | 226.73 | 120.23 | 106.49 |
| Capital | 431.96 | 229.07 | 202.89 |
| Land | 0.00 | 0.00 | 0.00 |
| Taxes | 150.29 | 79.70 | 70.59 |
| **Total gross output** | **3236.30** | **2121.31** | **1114.99** |
electricity to fossil fuels versus non-fuel goods. That is, we assume that these margins only apply to inputs 1 through 21 out of the 33 commodities in the Model.
Up to this point we have allocated the fuel inputs (commodity numbers 2,3,4), electricity (22) and a portion of the trade (27) and transportation (25) margins to Generation and to Distribution. Of the 2,121 billion yuan target total for Generation, 1,080 remains unallocated, and for Distribution 957 remain unallocated. Each of the remaining intermediate inputs (1,5,6,…33), and each of the value added items, is allocated to Generation and to Distribution in proportion to these unallocated amounts. In this way, the sum of all inputs for Generation equals the target 2,121 billion yuan. The results of this exercise are given in the second and third columns in Table B4.
The second step disaggregates the total Generation column to the 7 technologies, where the output targets are given in Table B3 in the “rescaled values” column. First, we allocate the entire coal mining input of Generation to “Coal generation”, oil mining to “Others”, and gas mining and gas utilities to “Gas generation”. See Table B5. Then for each technology we compute the unallocated total as the output target minus these allocated fuels. Each of the non-fuel rows in the Generation column (including the value added rows) is then allocated to the 7 technologies in proportion to these unallocated totals.
The results of this disaggregation are given in Table B5. Coal generation is 79% of total generation output but since fuel input is such a big factor here, its share of capital value-added in total Generation is only 72%. For hydro, the situation is reversed, it has 12.4% of Generation output value but 19.0% of value added.
The above simple disaggregation procedure preserves the Use column of the electricity & heat sector, that is, the sum of generation and distribution output is the electricity & heat output, and for each commodity input, the sum across the 7 generation technologies and distribution is the value in that row of the Use column. This means that the rest of the SAM for the other sectors is undisturbed.
### Table B5. Disaggregation of Generation inputs to the different technologies
| | Generation | Coal | Gas | Nuclear | Hydro | Wind | Solar | Other |
|----------|------------|------|-----|---------|-------|------|-------|-------|
| 1 Agriculture | 0.29 | 0.21 | 0.00 | 0.01 | 0.06 | 0.01 | 0.00 | 0.01 |
| 2 Coal mining | 686.55 | 686.55 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 24.38 |
| 3 Oil mining | 24.38 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 24.38 |
| 4 Gas mining | 24.33 | 0.00 | 24.33 | 0.00 | 0.00 | 0.00 | 0.00 | 24.38 |
| 5 Nonenergy mng | 4.66 | 3.34 | 0.08 | 0.13 | 0.89 | 0.12 | 0.00 | 0.11 |
| 6 Food | 11.98 | 8.59 | 0.20 | 0.33 | 2.28 | 0.30 | 0.00 | 0.28 |
| 7 Textile | 0.45 | 0.32 | 0.01 | 0.01 | 0.09 | 0.01 | 0.00 | 0.01 |
| 8 Apparel | 11.69 | 8.38 | 0.19 | 0.33 | 2.23 | 0.30 | 0.00 | 0.27 |
| 9 Lumber | 2.32 | 1.66 | 0.04 | 0.06 | 0.44 | 0.06 | 0.00 | 0.05 |
| 10 Paper | 6.07 | 4.35 | 0.10 | 0.17 | 1.16 | 0.15 | 0.00 | 0.14 |
| 11 Refining & coal | 144.79 | 103.76 | 2.36 | 4.04 | 27.58 | 3.66 | 0.02 | 3.37 |
| 12 Chemicals | 11.14 | 7.98 | 0.18 | 0.31 | 2.12 | 0.28 | 0.00 | 0.26 |
| 13 Nonmetallic min. | 7.66 | 5.49 | 0.13 | 0.21 | 1.46 | 0.19 | 0.00 | 0.18 |
| 14 Primary metals | 9.89 | 7.09 | 0.16 | 0.28 | 1.88 | 0.25 | 0.00 | 0.23 |
| 15 Fabricated metal | 14.41 | 10.33 | 0.24 | 0.40 | 2.75 | 0.36 | 0.00 | 0.33 |
| 16 Machinery | 36.87 | 26.42 | 0.60 | 1.03 | 7.02 | 0.93 | 0.00 | 0.86 |
| 17 Transportation eq | 45.41 | 32.54 | 0.74 | 1.27 | 8.65 | 1.15 | 0.01 | 1.06 |
| 18 Electrical mach. | 175.71 | 125.92 | 2.87 | 4.91 | 33.47 | 4.44 | 0.02 | 4.08 |
| 19 Electronics | 2.69 | 1.93 | 0.04 | 0.08 | 0.51 | 0.07 | 0.00 | 0.06 |
| 20 Instruments | 56.84 | 40.73 | 0.93 | 1.59 | 10.83 | 1.44 | 0.01 | 1.32 |
| 21 Other manuf. | 3.99 | 2.86 | 0.07 | 0.11 | 0.76 | 0.10 | 0.00 | 0.09 |
| 22 Electricity | 100.30 | 71.88 | 1.64 | 2.80 | 19.11 | 2.54 | 0.01 | 2.33 |
| 23 Gas utilities | 8.62 | 0.00 | 8.62 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 |
| 24 Construction | 1.35 | 0.96 | 0.02 | 0.04 | 0.26 | 0.03 | 0.00 | 0.03 |
| 25 Transportation | 38.89 | 27.87 | 0.64 | 1.09 | 7.41 | 0.98 | 0.00 | 0.90 |
| 26 Communication | 15.89 | 11.39 | 0.26 | 0.44 | 3.03 | 0.40 | 0.00 | 0.37 |
| 27 Trade | 36.88 | 26.43 | 0.60 | 1.03 | 7.02 | 0.93 | 0.00 | 0.86 |
| 28 Hotel & rest. | 7.70 | 5.52 | 0.13 | 0.22 | 1.47 | 0.19 | 0.00 | 0.18 |
| 29 Finance | 115.83 | 83.01 | 1.89 | 3.23 | 22.06 | 2.93 | 0.01 | 2.69 |
| 30 Real Estate | 1.56 | 1.12 | 0.03 | 0.04 | 0.30 | 0.04 | 0.00 | 0.04 |
| 31 Business svcs | 35.04 | 25.11 | 0.57 | 0.98 | 6.67 | 0.89 | 0.00 | 0.81 |
| 32 Services | 47.70 | 34.18 | 0.78 | 1.33 | 9.09 | 1.21 | 0.01 | 1.11 |
| 33 Public admin. | 0.41 | 0.29 | 0.01 | 0.01 | 0.08 | 0.01 | 0.00 | 0.01 |
| Labor | 120.23 | 86.16 | 1.96 | 3.36 | 22.90 | 3.04 | 0.01 | 2.79 |
| Capital | 229.07 | 164.15 | 3.74 | 6.40 | 43.63 | 5.79 | 0.03 | 5.32 |
| Land | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 |
| Taxes | 79.70 | 57.11 | 1.30 | 2.23 | 15.18 | 2.02 | 0.01 | 1.85 |
| Gross output | 2121.3 | 1673.6 | 55.44 | 38.46 | 262.37 | 34.83 | 0.17 | 56.39 |
In order to compare this disaggregation with the other sources of data, we compute the costs shares implied by this input-output data for coal and hydro (the predominant sources of electricity in 2010), and report them in Table B6. The cost shares are given for fuels, other non-fuel intermediate inputs, labor and capital costs in order to compare them with the categories in the IEA (2010) levelized cost calculations. In the IO tables, the finance row contains a large entry which we interpret to include interest margins on loans. When we compute the “capital cost share,” we include both the value added row and the finance row of the Use column in order to be closer to the accounting concepts in IEA (2010). Similarly, the fuel cost share is the sum of the coal row (in factory gate prices) and the trade and transportation margins in order to be consistent with the accounting concepts. The IEA estimated costs for selected technologies were given in Table B2c, and we also report in Table B6 the cost shares averaged over all the coal technologies given in IEA (2010). The average over all the hydro technologies is also given.
Table B6. Input cost structure for generation; input-output table versus IEA (2010).
| | Fuel | Other intermediates | Labor | Capital (inc. Finance) |
|---------------|------|---------------------|-------|------------------------|
| Coal power generation | | | | |
| Table B5 cost shares | 0.446 | 0.353 | 0.053 | 0.148 |
| IEA (2010) costs | 0.768 | 0.047 | 0.007 | 0.177 |
| Hydro power | | | | |
| Table B5 cost shares | 0.000 | 0.657 | 0.093 | 0.250 |
| IEA (2010) costs | 0.000 | 0.206 | 0.032 | 0.762 |
We can see that these cost shares are very different; for coal generation, the IEA (2010) projections allocates 76.8% to fuel and 17.7% to capital compared with the input-output table allocation of 44.6% for fuel, 14.8% to capital and 35.3% to non-fuel intermediates. These differences may be due to:
(a) distinction between average and marginal fuel costs (some coal input is allocated at the controlled price which is much lower than the market price, the IEA calculations are based on the market price);
(b) differences in the cost of capital assumptions (the IEA likely assumes a long run cost of capital in a deregulated capital market, the actual profits in 2010 that is embodied in the National Accounts may be much lower than such a long-run rate of return given that the feed-in tariffs are controlled by the government);
(c) size and vintage of the coal units (the IEA estimates are for new large plants to be constructed now whereas the current stock includes many older and smaller units, the maintenance costs may be quite different for the older units, the old plants may be allowed less generous tariffs);
(d) different coverage of enterprises included in the electricity and heat sector (the National Accounts may include enterprises that are not simple generators or distributors).
Given these distinctions we regard the 2010 SAM values as the correct ones to represent the base year cost structures, and use the IEA (2010) cost shares to represent future cost structures. That is, in our projections of the share parameter in the cost functions (or production functions), we will begin with the estimates in Table B4 and B5 and then gradually trend them towards the cost shares in the IEA rows of Table B6.
B.2 Projections in the base case
The exogenous drivers of economic growth are this model are the demographics (population and working-age population), saving rates, and total factor productivity growth as described in Appendix A. The future structure of the economy is affected by projected consumer preferences, biases in technical change and world commodity prices; production parameters are based on US input-output tables and consumption parameters are estimated using consumer expenditure surveys. This appendix focuses on the projection of the electricity sector which is treated in a distinct manner compared to the other production industries.
As explained in the Appendix A, we regard the investment in the electricity sector as being externally determined by the Plan. There is the detailed 5-year plans and also long-term targets; in the case of electricity generation there are targets for nuclear capacity, wind and solar capacities. IEA (2014) contains projections of China’s energy use, including projections of various electricity generation technologies, based on their reading of these plans and their projections of economic growth.
The IEA (2014, Table 1.1) uses the GDP projections from the IMF and these are given in Table B7. These are very close to our GDP projections (somewhat slower than Cao and Ho 2014), with about 7% in the current decade and slowing to 5% during 2020-30.
IEA (2014) projects three energy scenarios: “Current Policies” scenario is based on “policies that were enacted as of mid-2014”, “New Policies” is based on “the continuation of existing policies and measures as well as the implementation (albeit cautiously) of policy proposals, even if they are yet to be formally adopted”, while the “450 Scenario” is intended to illustrate “what it would take to achieve an energy trajectory consistent with limiting the long-term increase in average global temperature to 2°C.”
We take the “Current Policies” projections of the generation capacities and power output to guide our base case projection of the composition of electricity output. The IEA (2014) report also gives the estimates of actual capacities in 2012. The Chinese data reported in the China Energy Yearbook (and China Statistical Yearbook) that are given in Table B1 are slightly different in that they only include sources that are connected to the grid. We first rebase the 2012 IEA figures to these official data for 2012, and adjust their projections. These rebased projections out to 2040 are plotted in Figure B1 and summarized in Table B7.
Total electricity output is projected to rise from 4,986 TWh in 2012 to 6,930 in 2020, and to 10,333 in 2030. These translate to growth rates of 5.2% per year (2012-20), 3.2% (2020-30) and 1.9% (2030-40). By 2040, with a projected population of 1,466 million, the annual per-capita production will be 8.50 MWh. For comparison, the U.S. net generation in 2010 was 13.3 MWh per capita and Japan’s consumption was 8.34 MWh per capita.
---
33 IEA (2014) page 33.
34 Electricity output is given in Table 7.1 of the Monthly Energy Review published by the US EIA; world consumption is given in the World Bank’s World Development Indicators.
Table B7. Growth and energy projections (% per year).
| | Base Case (IEA "Current Policies") | Policy (IEA "New Policies") |
|---------------------|-----------------------------------|------------------------------|
| | 2012-20 | 2020-30 | 2030-40 | 2012-20 | 2020-30 | 2030-40 |
| GDP (Cao & Ho 2014) | 7.0 | 5.0 | 3.6 | 0.2 | 0.6 | 0.2 |
| GDP (IEA vis IMF) | 6.9 | 5.3 | 3.2 | 0.1 | 0.3 | 0.1 |
| Primary energy (IEA)| 2.9 | 1.9 | 0.89 | 0.1 | 0.3 | 0.1 |
| Electricity (IEA) | 5.2 | 3.2 | 1.9 | 0.1 | 0.3 | 0.1 |
| Elect: Coal | 3.5 | 3.0 | 1.9 | 0.1 | 0.3 | 0.1 |
| Elect: Gas | 14.6 | 6.2 | 3.9 | 0.1 | 0.3 | 0.1 |
| Non-fossil electricity | 9.3 | 3.4 | 1.4 | 0.1 | 0.3 | 0.1 |
| Primary energy (IEA)| 2.4 | 1.4 | 0.41 | 0.1 | 0.3 | 0.1 |
| Electricity (IEA) | 4.6 | 2.6 | 1.4 | 0.1 | 0.3 | 0.1 |
| Elect: Coal | 2.2 | 1.4 | 0.6 | 0.1 | 0.3 | 0.1 |
| Elect: Gas | 14.9 | 7.3 | 4.0 | 0.1 | 0.3 | 0.1 |
| Non-fossil electricity | 10.1 | 3.9 | 2.1 | 0.1 | 0.3 | 0.1 |
In this Current Policies scenario (CPS) the coal share of total output falls from quickly 77.1% in 2010 to 67.3% in 2020 and then decline slowly to 64.1% in 2030, and remain at about 64% thereafter. In terms of growth rates, coal generation grows at 3.5% during 2012-20 compared to 5.2% for total electricity output. Wind and solar together rise from 1.2% of total output in 2010 to 5.5% in 2020, and then continue to rise to 7.3% in 2030.
In the New Policies scenario (NPS) a more aggressive conservation path is assumed, with an even bigger shift to renewables. Figure B2 put the two scenarios for power output in TWh side by side. In the New Policies scenario total output rises only to 9,274 TWh in 2030 compared to 10,333 in the CPS. The annual growth rates of total electricity output are only 4.6% (2012-20), 2.6% (2020-30) and 1.4% (2030-40) compared to 5.2%, 3.2% and 1.9%, respectively, in the CPS. The NPS projects more renewables and the coal share of output here is 61.9% in 2020 versus 67.3% in the CPS, and 55.3% in 2030 versus 64.1%. In terms of absolute output, the CPS has 6620 TWh of coal-generated power in 2030 compared to 5129 TWh in the New Policies scenario.
Figure B2. Projection of TWh in IEA's Current Policies versus New Policies
References
International Energy Agency (IEA). 2010. *Projected Costs of Generating Electricity*, IEA and OECD, Paris.
International Energy Agency (IEA). 2014. *World Energy Outlook 2014*, IEA, Paris.
Lawrence Berkeley National Laboratory (LBNL) 2013. *China Energy Databook Version 8.0*. (eds.) David Fridley, John Romankiewicz and Cecilia Fino-Chen, China Energy Group, LBNL.
National Bureau of Statistics (NBS). 2014. China Statistical Yearbook 2014. | 2025-03-05T00:00:00 | olmocr | {
"pdf_page_numbers": [
[
0,
195,
1
],
[
195,
1791,
2
],
[
1791,
2839,
3
],
[
2839,
5522,
4
],
[
5522,
8244,
5
],
[
8244,
10669,
6
],
[
10669,
13420,
7
],
[
13420,
15983,
8
],
[
15983,
17180,
9
],
[
17180,
20004,
10
],
[
20004,
22164,
11
],
[
22164,
24194,
12
],
[
24194,
27164,
13
],
[
27164,
28918,
14
],
[
28918,
30249,
15
],
[
30249,
31394,
16
],
[
31394,
35630,
17
],
[
35630,
35725,
18
],
[
35725,
35789,
19
],
[
35789,
38036,
20
],
[
38036,
38344,
21
],
[
38344,
40667,
22
],
[
40667,
44774,
23
],
[
44774,
46255,
24
],
[
46255,
48965,
25
],
[
48965,
52158,
26
],
[
52158,
54344,
27
],
[
54344,
57511,
28
],
[
57511,
59755,
29
],
[
59755,
62485,
30
],
[
62485,
65013,
31
],
[
65013,
66435,
32
],
[
66435,
69630,
33
],
[
69630,
71653,
34
],
[
71653,
72917,
35
],
[
72917,
74979,
36
],
[
74979,
76639,
37
],
[
76639,
78600,
38
],
[
78600,
80498,
39
],
[
80498,
81219,
40
],
[
81219,
81219,
41
],
[
81219,
81305,
42
],
[
81305,
83248,
43
],
[
83248,
86113,
44
],
[
86113,
88253,
45
],
[
88253,
89729,
46
],
[
89729,
92617,
47
],
[
92617,
95101,
48
],
[
95101,
97502,
49
],
[
97502,
100034,
50
],
[
100034,
102379,
51
],
[
102379,
105261,
52
],
[
105261,
107396,
53
],
[
107396,
109862,
54
],
[
109862,
111600,
55
],
[
111600,
114124,
56
],
[
114124,
116387,
57
],
[
116387,
118539,
58
],
[
118539,
121257,
59
],
[
121257,
123125,
60
],
[
123125,
125052,
61
],
[
125052,
127357,
62
],
[
127357,
129147,
63
],
[
129147,
131194,
64
],
[
131194,
133193,
65
],
[
133193,
135188,
66
],
[
135188,
137016,
67
],
[
137016,
139850,
68
],
[
139850,
141871,
69
],
[
141871,
145077,
70
],
[
145077,
148083,
71
],
[
148083,
151096,
72
],
[
151096,
154691,
73
],
[
154691,
157293,
74
],
[
157293,
159903,
75
],
[
159903,
160752,
76
],
[
160752,
163083,
77
],
[
163083,
165249,
78
],
[
165249,
165695,
79
],
[
165695,
166177,
80
],
[
166177,
168842,
81
],
[
168842,
172211,
82
],
[
172211,
175252,
83
],
[
175252,
177663,
84
],
[
177663,
179705,
85
],
[
179705,
180387,
86
],
[
180387,
183747,
87
],
[
183747,
185278,
88
],
[
185278,
186768,
89
],
[
186768,
187848,
90
],
[
187848,
187934,
91
],
[
187934,
187976,
92
],
[
187976,
190436,
93
],
[
190436,
192468,
94
],
[
192468,
195051,
95
],
[
195051,
197818,
96
],
[
197818,
200470,
97
],
[
200470,
202934,
98
],
[
202934,
203826,
99
],
[
203826,
207053,
100
],
[
207053,
209232,
101
],
[
209232,
212349,
102
],
[
212349,
215094,
103
],
[
215094,
217397,
104
],
[
217397,
219187,
105
],
[
219187,
220575,
106
],
[
220575,
221835,
107
],
[
221835,
222291,
108
]
]
} |
7a5e08976b1f9a3a5a393cd4a1d6ce9354c92378 | {
"olmocr-version": "0.1.59",
"pdf-total-pages": 3,
"total-fallback-pages": 0,
"total-input-tokens": 8938,
"total-output-tokens": 3101,
"fieldofstudy": [
"Medicine"
]
} | Targeting adhesion to the vascular niche to improve therapy for acute myeloid leukemia
Myriam L. R. Haltalli & Cristina Lo Celso
Niche hijack by malignant cells is considered to be a prominent cause of disease relapse. Barbier and colleagues uncover (E)-selectin as a novel mediator of malignant cell survival and regeneration which, upon blockade, has the potential to significantly improve therapeutic outcomes.
Not-so-welcome to the neighborhood
Hematopoietic stem cells (HSCs) sit at the top of the hematological hierarchy as multipotent cells with the capacity to self-renew and give rise to various progenitor populations, which maintain the production of short-lived mature lineages (including blood cells) throughout life. They reside within the bone marrow (BM), within specialized niches that facilitate their survival and regulate their function via local interactions with diverse stromal and hematopoietic cells. It is widely accepted that spatially distinct niches exist to support HSCs with a range of differentiation biases and quiescence states. These are broadly split into the central BM niche, located in the deeper BM and rich in sinusoid capillaries, and the endosteal niche, in close proximity to the bone inner surface and rich in arterioles and transitional vessels that feed into the sinusoidal network. In both locations, HSCs reside in perivascular niches where endothelial cells (ECs) and their associated mesenchymal stem cells critically maintain and regulate them.
Hematological malignancies, such as acute myeloid leukemia (AML), tend to arise from the expansion of transformed hematopoietic stem and progenitor cells (HSPCs) with subclonal genotypes, resulting in refractory clones that are able to resist chemotherapy and ultimately lead to the relapse of disease. Leukemia stem cells (LSCs) share many features with healthy HSCs, which has led to the hypothesis that LSCs similarly reside in, and depend on, specific BM “neighborhoods” that support their expansion and survival. There is a growing need for novel therapies to reduce malignancy-associated morbidity and mortality and, due to the fact that many mechanisms of immune evasion and chemoresistance are partly dependant on it, the BM niche represents a unique target to improve treatment outcomes. The recent work by Barbier and colleagues is therefore timely as it reveals, using AML as a model, a new cell-extrinsic niche-based pathway that directly mediates therapy resistance. The authors propose a method with the potential to improve AML treatment efficacy by targeting the vascular niche.
Holding on tight
Endothelial (E)-selectins are cell adhesion molecules expressed in the vascular niche, which have well-characterized roles in leukocyte homing. E-selectin is constitutively expressed on BM
endothelium, aiding the homing and engraftment of circulating HSPCs that possess the necessary counter-receptors. Previous work had demonstrated that adhesion to E-selectin promotes the proliferation of HSCs, directly triggers their activation and induces lineage commitment². Interestingly, based on earlier observations of E-selectin expression in BM areas where leukemia cells home, the hypothesis arose that LSCs may be protected from the effects of chemotherapy through their anchorage in niches enriched in its expression. The work from Winkler et al. uncovered an unexpected additional role for this adhesion molecule as a therapeutic target when its inhibition induced HSC chemoresistance and radioresistance. Overall, their findings raised the question whether specific binding of tumor cells with endothelium expressing E-selectin might regulate tumor growth and proliferation². Barbier et al. address this question and show that, in the context of disease, interaction with E-selectin plays a direct role in promoting malignant cells’ survival in the BM, a textbook example of niche hijack.
Inflammation is a hallmark of cancer and we have previously hypothesized that it is a key player in driving remodeling of the BM niche during leukemia³. E-selectin is expressed by ECs at the leading edge of solid tumors and at pre-metastatic sites, which are also highly inflammatory environments⁴,⁵. Barbier et al. demonstrate that the inflammation generated by AML in the BM directly drives the increased levels of E-selectin on BM ECs of leukemic mice. This resulted in the postulation that the specific upregulation on tumor-associated vasculature may be less involved in the homing of AML cells, but more so as a factor directly contributing to their survival and regeneration by creating a protective niche for LSCs to hold on to, thus promoting therapy resistance.
A novel target for improved therapy
While there is an ever-increasing number of targeted therapies to treat leukemia, patient outcome remains poor, thus there is an urgent need to develop new strategies. Over the past few years, with the advance of increasingly sophisticated technologies to study the BM microenvironment, we have gained fascinating insight into this tissue. We can now appreciate that it is a complex and dynamic entity that actively interacts with healthy hematopoietic cells, as well as their malignant counterparts, playing an important role in directing their fates. Our knowledge of how leukemia cells co-opt and alter BM niches offers many opportunities for therapeutic exploitation, and more targeted interventions of novel pathways to restore healthy hematopoiesis and limit disease relapse.
By elegantly demonstrating that the leukemic blasts most likely to survive chemotherapy were those characterized by higher E-selectin binding potential, Barbier and colleagues suggested that these cells were the main contributors to disease relapse. To test the role of E-selectin in supporting malignant cells, they sought to understand the outcome if this adhesion molecule was absent in the BM (using an E-selectin gene knock-out mouse model) or blocked therapeutically. E-selectin can be inhibited by administering a selective small molecule antagonist called GMI-1271 (also known as uproleselan). It has been shown that this treatment effectively displaces chronic myeloid leukemia (CML) cells from the BM endothelium, promoting cell cycle progression and downregulating the expression of the E-selectin ligand CD44—thus reinforcing the reduced adhesion of CML cells to the BM endothelium⁶. Using an innovative method, the authors investigated whether functional LSCs were amongst those protected by E-selectin-mediated interactions in the vascular niche. They developed a quantitative in vivo LSC chemosensitivity assay, based on the principle of limiting dilution transplants widely used for enumerating HSCs. By comparing the number of persisting LSCs in recipients of BM cells from chemotherapy-treated, E-selectin gene-deleted, or GMI-1271-treated wild-type mice, they demonstrated that the deletion or blockade of E-selectin drastically reduced LSC survival. Therefore, E-selectin mediates significant intracellular pro-survival signaling for LSCs, and this unique characteristic was not replicated by other adhesion molecules and chemokines involved in BM retention.
Up until now, combinations of chemotherapy with approaches that target mainly AML-intrinsic mechanisms have been proposed³, including the use of CXCR4 antagonists⁴ and anti-inflammatory therapies⁵. With this work, the authors bring to light the fact that cell-extrinsic targets could prove hugely valuable in future phases of therapy development. By administering GMI-1271 alongside a standard chemotherapy regime in leukemic mice, Barbier et al. were able to double the duration of mouse survival over chemotherapy alone. Endogenous, healthy, HSPC populations exhibited strikingly different responses when compared to AML blasts, suggesting that this treatment strategy protects these populations in the BM. These intriguing data demonstrate that therapeutic blocking of this vascular niche-mediated survival pathway could complement conventional AML treatment regimens to improve efficacy and extend overall survival.
Future directions and clinical impact
We must harness the progress made so far in creating a detailed map of the BM microenvironment with all of its structures and components at steady state, in order to keep elucidating the underlying mechanisms of disease-mediated niche hijack. The number of clinical trials arising from a deeper understanding of the HSC niche is bound to increase. Based on the work presented by Barbier and colleagues, a phase I/II clinical trial is already in progress to evaluate the use of GMI-1271 alongside standard therapy in relapsed/refractory adult AML (NCT02306291). So far, they have shown greater rates of clinical remission and extended median patient survival, which is consistent with the findings in vivo presented by the authors. The future of these “nicotherapies”⁶ to facilitate chemotherapy looks extremely promising given the rise in demand for new agents with good tolerance in a broad range of AML patients, increased efficacy and, most importantly, the ability to limit disease relapse. The work presented here provides hope that innovative methods of targeting malignant BM niches could be the key to making AML a treatable disease.
Received: 19 June 2020; Accepted: 9 July 2020; Published online: 23 July 2020
References
1. Pinho, S. & Frenette, P. S. Haematopoietic stem cell activity and interactions with the niche. Nat. Rev. Mol. Cell Biol. 20, 303–320, (2019).
2. Méndez-Ferrer, S. et al. Mesenchymal and haematopoietic stem cells form a unique bone marrow niche. Nature 466, 829 (2010).
3. Ishikawa, F. et al. Chemotherapy-resistant human AML stem cells form a unique bone marrow niche. Nature 466, 829 (2010).
4. Lane, S. W., Scadden, D. T. & Gilliland, D. G. The leukemia stem cell niche: current concepts and therapeutic opportunities. Blood 114, 1150–1157 (2009).
5. Duarte, D., Hawkins, E. D. & Celso, C. L. The interplay of leukemia cells and the bone marrow microenvironment. Blood 131, 1507–1511 (2018).
6. Barbier, V. et al. Vascular niche hijack via E-selectin promotes niche-mediated acute myeloid leukemia stem cell chemoresistance and regeneration. Nat. Commun. 11, 2042 (2020).
7. Winkler, I. G. et al. Vascular niche E-selectin regulates hematopoietic stem cell dormancy, self renewal and chemoresistance. Nat. Med. 18, 1651 (2012).
8. Sipkins, D. A. et al. In vivo imaging of specialized bone marrow endothelial microdomains for tumour engraftment. *Nature* **435**, 969–973 (2005).
9. Duarte, D. et al. Inhibition of endostelial vascular niche remodeling rescues hematopoietic stem cell loss in AML. *Cell Stem Cell* **22**, 64–77 (2018).
10. Ye, C. et al. Expression of E-selectin on endothelial cells of small veins in human colorectal cancer. *Int J. Cancer* **61**, 455–460 (1995).
11. Shi, H. et al. Recruited monocytic myeloid-derived suppressor cells promote the arrest of tumor cells in the premetastatic niche through an IL-1β-mediated increase in E-selectin expression. *Int J. Cancer* **140**, 1370–1383 (2017).
12. Godavarthy, P. S. et al. The vascular bone marrow niche influences outcome in chronic myeloid leukemia via the E-selectin - SCL/TAL1 - CD44 axis. *Haematologica* **105**, 136–147 (2019).
13. Pezeshkian, B., Donnelly, C., Tamburo, K., Geddes, T. & Madlambayan, G. J. Leukemia mediated endothelial cell activation modulates leukemia cell susceptibility to chemotherapy through a positive feedback loop mechanism. *Plos ONE* **8**, e60823 (2013).
14. Cho, B.-S., Kim, H.-J. & Konopleva, M. Targeting the CXCL12/CXCR4 axis in acute myeloid leukemia: from bench to bedside. *Korean J. Intern. Med.* **32**, 248–257 (2017).
15. Mead, A. J. et al. Niche-mediated depletion of the normal hematopoietic stem cell reservoir by Flt3-ITD–induced myeloproliferation. *J. Exp. Med.* **214**, 2005–2021 (2017).
16. Levesque, J.-P., Winkler, I. G. & Rasko, J. E. J. Nichotherapy for stem cells: there goes the neighborhood. *Bioessays* **35**, 183–190 (2012).
**Author contributions**
M.L.R.H and C.L.C wrote the article.
**Competing interests**
The authors declare no competing interests.
**Additional information**
Correspondence and requests for materials should be addressed to C.L.C.
Reprints and permission information is available at [http://www.nature.com/reprints](http://www.nature.com/reprints)
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. | 2025-03-04T00:00:00 | olmocr | {
"pdf_page_numbers": [
[
0,
2805,
1
],
[
2805,
10405,
2
],
[
10405,
12531,
3
]
]
} |
e3c1e7affd69b22c106cf05efad75a338b07648b | {
"olmocr-version": "0.1.59",
"pdf-total-pages": 16,
"total-fallback-pages": 0,
"total-input-tokens": 64739,
"total-output-tokens": 16329,
"fieldofstudy": [
"Physics"
]
} | Pulsating Strings in Lunin-Maldacena Backgrounds
Sergio Giardino and Victor O. Rivelles
Instituto de Física, Universidade de São Paulo, C. Postal 66318, 05314-970 São Paulo, SP, Brazil
E-mail: jardino,[email protected]
ABSTRACT: We consider pulsating strings in Lunin-Maldacena backgrounds, specifically in deformed Minkowski spacetime and deformed $AdS_5 \times S^5$. We find the relation between the energy and the oscillation number of the pulsating string when the deformation is small. Since the oscillation number is an adiabatic invariant it can be used to explore the regime of highly excited string states. We then quantize the string and look for such a sector. For the deformed Minkowski background we find a precise match with the classical results if the oscillation number is quantized as an even number. For the deformed $AdS_5 \times S^5$ we find a contribution which depends on the deformation parameter.
KEYWORDS: AdS/CFT Correspondence, Pulsating Strings, Lunin-Maldacena Background.
1. Introduction
Integrability is an important tool for the understanding of several aspects of the AdS/CFT correspondence [1]. The energy of strings in $\text{AdS}_5 \times S^5$ and anomalous dimensions of gauge invariant operators in planar $\mathcal{N} = 4$ super Yang-Mills theory have been successfully compared with the predictions of the thermodynamic Bethe ansatz in several limits [2]. Being functions of the string tension and charges these quantities usually have intricate expressions. Strings at the semiclassical level correspond to operators with large charges in the field theory and a reasonable understanding of this situation has been achieved [3, 4]. On the other side, when the semiclassical parameters are small, corresponding to short operators in the field theory, results for folded strings [5, 6] and pulsating strings [7] are providing new data to support the correspondence.
Despite the huge progress got so far in the $\text{AdS}_5 \times S^5$ case it is also important to understand situations with less supersymmetry. One such a case is the planar gauge theory with a beta deformed superpotential leading to a marginally conformal $\mathcal{N} = 2$ supersymmetric gauge theory [8, 9]. The dual gravitational background was found by Lunin and Maldacena [10] and in the case of $\text{AdS}_5 \times S^5$ it consists of a deformation of $S^5$ through a real parameter $\hat{\gamma}$ which keeps the anti-de Sitter part intact while the dilaton and some RR and NS-NS fields are turned on. Both, string and gauge theories, also show signs of integrability [11, 12, 13, 14, 15] with the corresponding Y-system being discussed recently [16, 17]. Several spinning and rotating string configurations were considered in Lunin-Maldacena backgrounds. However one important class of strings, pulsating strings, have not received much attention so far. They have been studied in $\text{AdS}_5 \times S^5$ [18, 19, 20, 21, 22, 23, 24, 25], $\text{AdS}_4 \times CP^3$ [26, 27] and other backgrounds [28],[29],[30],[31]. In the context of deformed geometry it was found that pulsating and rotating strings in $\text{AdS}_5 \times S^5$ are images of point like strings living in the deformed space [32]. In this paper we will provide some results for pulsating strings in deformed backgrounds when the deformation is small.
Since the classical motion of a pulsating string is periodic we can use its oscillation number\(^1\) $N = \oint p dq/2\pi$ and its energy to characterize the dynamics. The string oscillation
\(^1\)Here $p$ is the canonical momentum to the oscillating coordinate $q$.
---
Contents
1. Introduction 1
2. Lunin-Maldacena Backgrounds 2
3. Pulsating Strings in Deformed Minkowski Spacetime 4
4. Pulsating Strings in the Deformed Sphere of $\text{AdS}_5 \times S^5_{\hat{\gamma}}$ 8
5. Conclusions 12
number is not one of the string charges but it is very useful to describe the behavior of pulsating strings as shown in the $AdS_5 \times S^5$ case [24, 7]. So we consider pulsating strings in deformed Minkowski spacetime and $AdS_5 \times S^5$ to find the relation between the energy and the oscillation number in the regime of small deformation. Since the oscillation number is an adiabatic invariant it can provide some information about the semi-classical regime where $N$ is very large. We then perform the quantization of the string looking for highly excited string states and compute its energy in perturbation theory. When we consider the deformed Minkowski spacetime we find full agreement if the oscillation number is quantized as an even number. In the $AdS_5 \times S^5$ case we consider only strings oscillating in the deformed sphere since $AdS_5$ is not deformed and the solutions in this sector would be the same as in the undeformed case. We find that the oscillation number can be expressed in terms of elliptic functions. We derive the relation between the energy and the oscillation number for short strings in the low energy regime. When the deformation vanishes we recover the results for $AdS_5 \times S^5$ [7]. In the case of large energy we consider the quantization of highly excited strings up to second order in perturbation theory and we find that the energy has a new term proportional to the deformation parameter which is not present in the classical relation between the energy and the oscillation number. We also discuss the contributions coming from the fluctuations of the radial $AdS$ coordinate.
This paper is organized as follows. A short discussion of the Lunin-Maldacena backgrounds is presented in the next section. In Section 3 we consider pulsating strings in the deformed Minkowski spacetime and in Section 4 we consider the case of $AdS_5 \times S^5$. Finally in Section 5 we present some conclusions.
2. Lunin-Maldacena Backgrounds
The beta deformed $\mathcal{N} = 4$ SYM gauge theory has a $U(1) \times U(1)$ global symmetry which is realized geometrically as an isometry of a two torus in the dual string background. It can also be understood as a TsT transformation applied to $AdS_5 \times S^5$ and can be generalized to other backgrounds as well [14, 33]. Keeping with the standard notation the real deformation parameter in the string side is denoted by $\gamma$. Then the Lunin-Maldacena background is given by [10]
$$ ds^2 = R^2 \left[ ds^2_{AdS_5} + \sum_{i=1}^{3} (d\mu_i^2 + G \mu_i^2 d\phi_i) + \gamma^2 \mu_1^2 \mu_2^2 \mu_3^2 G \left( \sum_{i=1}^{3} d\phi_i \right)^2 \right], $$
$$ B_2 = \gamma R^2 G \left( \mu_1^2 \mu_2^2 d\phi_1 \wedge d\phi_2 + \mu_2^2 \mu_3^2 d\phi_2 \wedge d\phi_3 - \mu_1^2 \mu_3^2 d\phi_1 \wedge d\phi_3 \right), $$
$$ e^{2\Phi} = e^{2\Phi_0} G, $$
$$ C_2 = -48\pi N \gamma \omega_1 \wedge d\psi, $$
$$ C_4 = 16\pi N (\omega_1 + G d\omega_1 \wedge d\phi_1 \wedge d\phi_2 \wedge d\phi_3), $$
where $B_2$ is the NS-NS two-form potential, $C_2$ and $C_4$ are the two and four-form RR
potentials respectively, $\Phi$ is the dilaton, and
\[
\begin{align*}
ds_{AdS_5}^2 &= -\cosh^2 \rho \, dt^2 + dp^2 + \sinh^2 \rho \left( d\Psi^2 + \sin^2 \Psi \, d\Phi_1^2 + \cos^2 \Psi \, d\Phi_2^2 \right), \\
G^{-1} &= 1 + \hat{\gamma}^2 \left( \mu_1^2 \mu_2^2 + \mu_2^2 \mu_3^2 + \mu_3^2 \mu_1^2 \right), \\
d\omega_1 &= \sin^3 \alpha \cos \alpha \sin \theta \cos \theta d\alpha \wedge d\theta, \quad \omega_{AdS_n} = d\omega_4, \\
\sum_{i=1}^3 \mu_i^2 &= 1, \quad \hat{\gamma} R^2 = \gamma, \quad R^2 = 4\pi e^{\Psi_0} N, \quad (2.5)
\end{align*}
\]
with $R$ being the AdS radius. Notice that only the $S^5$ sphere has been modified and that when $\hat{\gamma} = 0$ we recover the original $AdS_5 \times S^5$ background.
We can solve the constraint in (2.5) by introducing spherical like coordinates
\[
\begin{align*}
\mu_1 &= \sin \theta \cos \psi, \\
\mu_2 &= \cos \theta, \\
\mu_3 &= \sin \theta \sin \psi,
\end{align*}
\]
so that the metric and the KR field become
\[
\begin{align*}
ds^2 &= R^2 \left\{ ds_{AdS_5}^2 + d\theta^2 + \sin^2 \theta \, d\psi^2 + G \left[ \sin^2 \theta \cos^2 \psi \left( 1 + \hat{\gamma}^2 \sin^2 \theta \cos^2 \theta \sin^2 \psi \right) d\phi_1^2 + \cos^2 \theta \left( 1 + \hat{\gamma}^2 \sin^4 \theta \cos^2 \psi \sin^2 \psi \right) d\phi_2^2 + \sin^2 \theta \sin^2 \psi \left( 1 + \hat{\gamma}^2 \sin^2 \theta \cos^2 \theta \cos^2 \psi \right) d\phi_3^2 \right] + 2G \hat{\gamma}^2 \sin^4 \theta \cos^2 \theta \sin^2 \psi \cos^2 \psi \left( d\phi_1 d\phi_2 + d\phi_2 d\phi_3 + d\phi_1 d\phi_3 \right) \right\}, \\
B_2 &= R^2 \gamma G \sin^2 \theta \left( \cos^2 \theta \cos^2 \psi d\phi_1 \wedge d\phi_2 + \cos^2 \theta \sin^2 \psi d\phi_2 \wedge d\phi_3 - \sin^2 \theta \sin^2 \psi \cos^2 \psi d\phi_1 \wedge d\phi_3 \right), \quad (2.7)
\end{align*}
\]
where now
\[
G^{-1} = 1 + \hat{\gamma}^2 \sin^2 \theta (\cos^2 \theta + \sin^2 \theta \cos^2 \psi \sin^2 \psi). \quad (2.9)
\]
We will use this form to study pulsating strings.
The Lunin-Maldacena deformation technique can also be applied to the ten dimensional Minkowski spacetime [10]. The result is simpler than in the $AdS_5 \times S^5$ case and it will be used in the next section as a background for pulsating strings even though no dual gauge theory is known. The ten dimensional Minkowski spacetime is split into a four dimensional one and the remaining six dimensional space is deformed resulting in
\[
\begin{align*}
ds^2 &= \eta_{\mu\nu} dx^\mu dx^\nu + \sum_{i=1}^3 \left( dr_i^2 + Gr_i^2 d\phi_i^2 \right) + \gamma^2 r_1^2 r_2^2 r_3^2 G \left( \sum_{i=1}^3 d\phi_i \right)^2, \\
B_2 &= \gamma G \left( r_1^2 r_2^2 d\phi_1 \wedge d\phi_2 + r_2^2 r_3^2 d\phi_2 \wedge d\phi_3 - r_1^2 r_3^2 d\phi_1 \wedge d\phi_3 \right), \\
e^{2\Phi} &= G, \quad G^{-1} = 1 + \gamma^2 \left( r_1^2 r_2^2 + r_2^2 r_3^2 + r_1^2 r_3^2 \right).
\end{align*}
\]
No RR fields are present in this case. As expected, the limit $\gamma = 0$ reproduces the ten dimensional Minkowski spacetime.
3. Pulsating Strings in Deformed Minkowski Spacetime
To consider pulsating strings in (2.10) we will use spherical coordinates defined by
\[ r_1 = r \sin \theta \cos \psi, \quad r_2 = r \sin \theta \sin \psi, \quad r_3 = r \cos \theta. \] (3.1)
Firstly we choose a string at the origin of the Minkowski spacetime and with \( \psi = \pi/2 \), \( \phi_1 = \phi_2 = 0 \) and \( \theta \) fixed so that the metric becomes
\[ ds^2 = -dt^2 + dr^2 + Gr^2 \cos^2 \theta d\phi_3^2, \]
\[ G^{-1} = 1 + \gamma^2 r^4 \sin^2 \theta \cos^2 \theta. \] (3.2) (3.3)
With this choice the coupling of the string to the \( B_2 \) field vanishes. We now consider an ansatz for a pulsating string that is wound in \( \phi_3 \) and oscillating in the radial direction
\[ t = \kappa \tau, \quad r = r(\tau), \quad \phi_3 = m \sigma, \quad \theta = \text{constant}, \] (3.4)
with \( m \) being the winding number. The only non trivial equation comes from the Virasoro constraint
\[ r^2 + m^2 r^2 G \cos^2 \theta = \kappa^2. \] (3.5)
At this point we must notice that the fixed angle \( \theta \) appears always in the combination \( m \cos \theta \) and \( \gamma \sin \theta \cos \theta \) which allow us to rescale \( m \) and \( \gamma \) as \( m \cos \theta \rightarrow m \) and \( \gamma \sin \theta \cos \theta \rightarrow \gamma \) respectively, assuming that \( \theta \) is different from zero and \( \pi/2 \).
We can write (3.5) in terms of an effective potential as
\[ \frac{r^2}{m^2} = \frac{\kappa^2}{m^2} - V(r). \] (3.6)
where \( V(r) \) is given by
\[ V(r) = \frac{r^2}{1 + \gamma^2 r^4}, \] (3.7)
The behavior of the effective potential is depicted in Fig. 1. There is an unstable equilibrium point at \( r_M^2 = 1/\gamma \) where the potential reaches its maximum value \( V(r_M) = 1/(2\gamma) \).
The points where the radial velocity vanishes are
\[ r^2_\pm = \frac{1}{2\gamma^2 \kappa^2} \left( 1 \pm \sqrt{1 - 4\gamma^2 \kappa^4 \frac{m^4}{r_M^4}} \right), \] (3.8)
Figure 1: The effective potential in deformed Minkowski spacetime
- 4 -
and the condition for oscillatory motion implies that \( \kappa^2/m^2 \leq 1/(2\gamma) \).
Now (3.5) can be rewritten as
\[
\int dr \sqrt{\frac{1 + \gamma^2 r^4}{(r^2 - r_{\pm}^2)(r^2 - r_-^2)}} = \gamma \kappa d\tau, \tag{3.9}
\]
Let us consider the case of small deformation \( \gamma << 1 \). Then the height of the potential is very large and we can consider a highly excited pulsating string. Its radial motion will be limited to \( r_- \). For small \( \gamma \) the turning points are
\[
r_{\pm}^2 \approx \frac{1}{\gamma^2 \kappa^2} \left( 1 - \gamma^2 \frac{\kappa^4}{m^4} \right), \tag{3.10}
\]
\[
r_-^2 \approx \kappa^2 \frac{m^2}{\gamma^2} \left( 1 + \gamma^2 \frac{\kappa^4}{m^4} \right). \tag{3.11}
\]
In this situation \( \gamma^2 r^4 < 1 \) and the square root in (3.9) can be expanded up to first order in \( \gamma^2 \) resulting in
\[
\int_{0}^{\tau} dr \frac{1 + \gamma^2 r^4/2}{\sqrt{(r^2 - r_{\pm}^2)(r^2 - r_-^2)}} = \gamma \kappa (\tau - \tau_0), \tag{3.12}
\]
where \( \tau_0 \) is an integration constant. The integrals can be written in terms of elliptic functions. Notice that for vanishing \( \gamma^2 \) we recover the results for flat spacetime discussed for instance in [20].
In order to characterize the string dynamics we can compute the relation between the string oscillation number and the energy. The oscillation number associated to the periodic radial motion is \( N = \oint \Pi_r dr/2\pi \), where \( \Pi_r \) is the momentum conjugated to \( r \). To compute \( \Pi_r \) we take the Polyakov action for the ansatz (3.4)
\[
S = -\frac{\sqrt{\lambda \kappa}}{2} \int d\tau \left( 1 - \dot{r}^2 + m^2 r^2 G \right), \tag{3.13}
\]
to find that \( \Pi_r = \sqrt{\lambda \kappa} \dot{r} \). Then
\[
N = \frac{\sqrt{\lambda \kappa}}{2\pi} \oint \dot{r} dr = \frac{\sqrt{\lambda \kappa}}{2\pi} \oint \sqrt{\kappa^2 - m^2 V(r)} dr \tag{3.14}
\]
\[
= 2\frac{\sqrt{\lambda}}{\pi} \gamma \kappa \int_{0}^{r_-} dr \sqrt{(r^2 - r_{\pm}^2)(r^2 - r_-^2)}(1 - \frac{1}{2} \gamma^2 r^4), \tag{3.15}
\]
where in the last line we used the small \( \gamma \) limit. The integrals can be easily evaluated in the limit of small \( \gamma \) and we get
\[
N = \frac{1}{2} \frac{\sqrt{\lambda} \kappa^2}{m} \left( 1 + \frac{5}{16} \gamma^2 \frac{\kappa^4}{m^4} \right). \tag{3.16}
\]
We can then solve for the energy \( E = \sqrt{\lambda \kappa} \) in the limit of small \( \gamma \) to find
\[
E = \sqrt{2\lambda^{1/2}mN} \left( 1 - \frac{5}{8} \frac{\gamma^2 N^2}{\lambda \, m^2} \right). \tag{3.17}
\]
Since adiabatic invariants can be used to probe the highly excited states of the quantum theory the above expression can be regarded an approximation for the energy in the limit of large quantum numbers in the radial direction. In what follows we will check that this is indeed true.
Quantization will be performed by starting with the Nambu-Goto action for the ansatz (3.4)
\[ S = -m \frac{\sqrt{\lambda}}{\kappa} \int dt \, r \sqrt{G} \sqrt{1 - \dot{r}^2}. \]
(3.18)
We find that \( \Pi_r = \sqrt{\lambda m r} \sqrt{G} \dot{r} / \sqrt{k^2 - \dot{r}^2} \) and
\[
\begin{align*}
H^2 &= \Pi_r^2 + \lambda m^2 \frac{r^2}{1 + \gamma^2 r^4} \\
&= \Pi_r^2 + \lambda m^2 r^2 - \gamma^2 \lambda m^2 \frac{r^6}{1 + \gamma^2 r^4},
\end{align*}
\]
(3.19)
(3.20)
where in the last line we split the potential in a term which is independent of \( \gamma \) and another which depends on the deformation. If the deformation vanishes we get a radial harmonic oscillator potential \( \lambda m^2 r^2 \). In order to proceed we assume that the wave function depends only on \( r \) so we have to realize \( \Pi_r^2 \) as the radial component of the Laplacian
\[
\Pi_r^2 = -\frac{1}{\sqrt{-g}} \frac{d}{dr} \left( \sqrt{-g} \frac{d}{dr} \right).
\]
(3.21)
We will first consider the situation where we quantize only the radial motion on the deformed plane (3.2) ignoring the remaining coordinates, that is we take \( \sqrt{-g} = r \cos \theta \sqrt{G} \) in (3.21). In this situation we expect to reproduce (3.17) for higher quantum numbers. The Schrödinger equation to be solved is then
\[
H^2 \Psi = \frac{1}{r} \frac{d}{dr} \left( r \frac{d\Psi}{dr} \right) + \lambda m r^2 \Psi + 2\gamma^2 r^3 \frac{d\Psi}{dr} - \gamma^2 \lambda m^2 \frac{r^6}{1 + \gamma^2 r^4} \Psi = E^2 \Psi,
\]
(3.22)
where we took the limit of small deformation in the third term. To solve (3.22) we will use standard perturbation theory. First we have to choose the unperturbed Hamiltonian. We can either choose the two first terms or the three first terms. In the last case we are considering the quantization in the deformed space while in the first one we are regarding the deformation as a perturbation and quantizing in the undeformed space. Either way we get the same result at the end. So we choose as the unperturbed Hamiltonian the first two terms of (3.22) and find that the normalized wave function is
\[
\Psi_n(r) = \sqrt{2\lambda^{1/2} m} e^{-\frac{1}{2} \sqrt{\lambda} m r^2} L_n(\sqrt{\lambda} m r^2),
\]
(3.23)
where \( L_n \) are Laguerre polynomials. For highly excited states we have
\[
E_{0,n}^2 = 4\sqrt{\lambda} m n,
\]
(3.24)
or \( E_{0,n} = \sqrt{2\lambda^{1/2}m(2n)} \). Since we are quantizing a radial harmonic oscillator we expect that its energy depends only on even integers. Comparison with the lowest order of (3.17) shows complete agreement if the oscillation number is quantized as \( 2n \).
The first order correction to the energy for small deformation is
\[
\delta E_n^2 = 2\gamma^2 \int_0^\infty dr r^4 \frac{d\Psi_n}{dr} \frac{d\Psi_n}{dr} - \gamma^2 \lambda m^2 \int_0^\infty dr \frac{r^7}{1 + \gamma^2 r^4} |\Psi_n|^2
\]
\[
= 2 \frac{\gamma^2}{\sqrt{\lambda m}} n - 20 \frac{\gamma^2}{\sqrt{\lambda m}} n^3.
\]
(3.25)
The first term can be disregard for large \( n \) and the energy is then
\[
E_n = \sqrt{2\lambda^{1/2}m(2n)} \left( 1 - \frac{5}{16} \frac{\gamma^2 (2n)^2}{\lambda m^2} \right),
\]
(3.26)
As expected it agrees with (3.17) if we assume that the oscillation number \( N \) is quantized with eigenvalue \( 2n \).
We can capture some of the quantum effects of the deformed ten dimensional space by taking in \( \Pi^2 \) the contribution of all dimensions by choosing \( \sqrt{-g} = r^5 \sqrt{G} \ldots \) where the dots represent terms which do not depend on \( r \) and cancel out in (3.21). We also assume that the wave function depends only on the radial coordinate. The Schrödinger equation for \( H^2 \) reads now
\[
H^2 \Psi = -\frac{1}{r^5} \frac{d}{dr} \left( r^5 \frac{d\Psi}{dr} \right) + \lambda m^2 r^2 \Psi + 4\gamma^2 r^2 \frac{d\Psi}{dr} - \gamma^2 \lambda m^2 \frac{r^6}{1 + \gamma^2 r^4} \Psi = E^2 \Psi,
\]
(3.27)
where in the third term we already took the limit of small deformation. Again we take the unperturbed potential as that of the radial harmonic oscillator and consider the last two terms as perturbations. The normalized wave function is now
\[
\Psi_n(r) = \sqrt{2\lambda^{3/4} m^{3/2} n} e^{-\sqrt{\lambda m^2} r/2} L_n^{(2)}(\sqrt{\lambda m^2}),
\]
(3.28)
where \( L_n^{(2)} \) are generalized Laguerre polynomials. The energy for high quantum number is still given by the former result (3.24). This happens because only the zero point contribution to the energy changes with the dimension and in the limit of large \( n \) it can be ignored.
Now we have two corrections coming from the deformation. The first one is the \( r^6/(1 + \gamma^2 r^4) \) perturbation in the potential term in (3.27). For large \( n \) it gives the same correction of order \( n^3 \) as in (3.25) even though the wave function and the number of dimensions are different. The second correction in (3.27) has a derivative term and in the limit of large \( n \) we get
\[
\delta^2 E_n^2 = -32 \frac{\gamma^2 n}{\sqrt{\lambda m}}.
\]
(3.29)
It is of order \( n \) while the first term is of order \( n^3 \) so it can be disregarded. This means that in the large \( n \) limit we get the same result (3.26) as in the planar case. This shows that for a highly excited string the classical result (3.17) can be straightforwardly taken to quantum case when \( N \) is quantized as \( 2n \).
4. Pulsating Strings in the Deformed Sphere of $\text{AdS}_5 \times S^5_{\gamma}$
We now consider a pulsating string in the deformed sphere of $\text{AdS}_5 \times S^5_{\gamma}$. We assume that in the $\text{AdS}_5$ sector (2.2) we have $\Psi = \Phi_1 = \Phi_2 = 0$ while in the $S^5_{\gamma}$ sector (2.7) we take $\phi_1 = \phi_2 = 0$ and $\psi = \pi/2$ so that we still have a deformed sphere inside $S^5_{\gamma}$. In this situation the metric reduces to
$$ds^2 = R^2 \left(- \cosh^2 \rho \, dt^2 + d\rho^2 + d\theta^2 + G \sin \theta^2 \, d\phi_3^2\right),$$
(4.1)
while $B_2 = C_2 = 0$, and now $G = 1/(1 + \gamma^2 \sin^2 \theta \cos^2 \theta)$.
We consider the following ansatz for a pulsating string wound $m$ times along $\phi_3$ in the deformed sphere
$$t = \kappa \tau, \quad \rho = \rho(\tau), \quad \theta = \theta(\tau), \quad \phi_3 = m\sigma.$$
(4.2)
The equations of motion are satisfied if $\rho$ is constant, so we take the string at the center of $\text{AdS}_5$. Together with the Virasoro constraints we get the only non-trivial equation
$$\dot{\theta}^2 - \kappa^2 + m^2 G \sin^2 \theta = 0.$$
(4.3)
As before we introduce an effective potential as
$$\frac{\dot{\theta}^2}{m^2} = \frac{\kappa^2}{m^2} - V(\theta)$$
$$V(\theta) = \frac{\sin^2 \theta}{1 + \gamma^2 \sin^2 \theta \cos^2 \theta},$$
(4.4)
which is plotted in Fig.2. The potential has a maximum at $\pi/2$ with value 1 so that there are two situations to be analyzed. If $\kappa^2/m^2 < 1$ then there is a turning point so that $\theta$ is limited to a maximum value $\theta_+$. In this case the energy is small $E^2 < \lambda m^2$ and we have a short string oscillating on the deformed sphere. If $\kappa^2/m^2 > 1$ then $E^2 > \lambda m^2$ and there are no turning points. We now have a string oscillating all the way from the equator to one of the poles of the deformed sphere.
We can rewrite (4.3) as
$$\frac{\dot{\theta}^2}{m^2} = \gamma^2 \frac{\kappa^2}{m^2} \left(\sin^2 \theta_+ - \sin^2 \theta\right)(\sin^2 \theta - \sin^2 \theta_-)}{1 + \gamma^2 \sin^2 \theta \cos^2 \theta},$$
(4.5)
where, in the limit of small deformation, we have
\[ \sin^2 \theta_+ = \frac{\kappa^2}{m^2} \left[ 1 + \hat{\gamma}^2 \frac{\kappa^2}{m^2} \left( 1 - \frac{\kappa^2}{m^2} \right) \right], \quad \sin^2 \theta_- = -\frac{1}{\hat{\gamma}^2} \frac{1}{\sin^2 \theta_+}. \]
(4.6)
For \( \kappa^2/m^2 < 1 \) there is one turning point \( \theta_+ \) while for \( \kappa^2/m^2 > 1 \) there is none.
To find out the oscillating number we need the Polyakov action for the ansatz (4.2)
\[ S = -\frac{\sqrt{\lambda \kappa}}{2} \int d\tau (\cosh^2 \rho - \dot{\rho}^2 - \dot{\theta}^2 + m^2 \sin^2 \theta G), \]
(4.7)
from which it follows that \( \Pi_\theta = \sqrt{\lambda \kappa} \dot{\theta} \). Then the oscillating number is
\[ N = \frac{1}{2\pi} \oint \Pi_\theta d\theta = \frac{1}{2\pi} \sqrt{\lambda \kappa} \oint \dot{\theta} d\theta. \]
(4.8)
Let us consider first the case of short strings when \( \kappa^2/m^2 < 1 \). The oscillation number in this case is given by
\[ N = \frac{2\sqrt{\lambda}}{\pi} \frac{\kappa}{\sin \theta_+} \int_0^{\theta_+} d\theta \sqrt{\sin^2 \theta_+ - \sin^2 \theta} \left[ 1 + \frac{1}{2} \hat{\gamma}^2 (\sin^2 \theta_+ - \cos^2 \theta) \sin^2 \theta \right]. \]
(4.9)
The integrals reduce to elliptic functions which still are \( \hat{\gamma} \) dependent. After further expansion in \( \hat{\gamma} \) we find
\[ N = \frac{2\sqrt{\lambda}}{\pi m} \left[ \mathcal{E} \left( \frac{\kappa^2}{m^2} \right) - \left( 1 - \frac{\kappa^2}{m^2} \right) \mathcal{K} \left( \frac{\kappa^2}{m^2} \right) \right] \]
\[ + \hat{\gamma}^2 \frac{2\sqrt{\lambda}}{\pi m} \left[ -\frac{1}{15} \left( 2 + 3 \frac{\kappa^2}{m^2} - 8 \frac{\kappa^4}{m^4} \right) \mathcal{E} \left( \frac{\kappa^2}{m^2} \right) + \frac{2}{15} \left( 1 - \frac{\kappa^2}{m^2} \right) \left( 1 + 2 \frac{\kappa^2}{m^2} \right) \mathcal{K} \left( \frac{\kappa^2}{m^2} \right) \right], \]
(4.10)
where \( \mathcal{K} \) and \( \mathcal{E} \) are complete elliptic integrals of the first and second kind respectively. The string energy \( E = \sqrt{\lambda \kappa} \) can be used to eliminate \( \kappa \) and allows to find the relation between the energy \( E \) and the oscillation number \( N \). To do that we have to expand the elliptic functions and we have to assume that \( E/(\sqrt{\lambda} m) \) is small. We then obtain
\[ E = \sqrt{2\lambda^{1/2} m N} \left[ 1 - \frac{1}{8} \frac{N}{\sqrt{\lambda} m} - \frac{5}{128} \left( \frac{N}{\sqrt{\lambda} m} \right)^2 + \ldots \right] \]
\[ - \frac{3\sqrt{2}}{8} \hat{\gamma}^2 \frac{N^{3/2}}{(\sqrt{\lambda} m)^{1/2}} \left[ 1 - \frac{41}{24} \frac{N}{24 \sqrt{\lambda} m} + \frac{193}{384} \left( \frac{N}{\sqrt{\lambda} m} \right)^2 + \ldots \right]. \]
(4.11)
As expected there is a natural small expansion parameter \( N/(\sqrt{\lambda} m) \) involving the oscillation number so we are in the regime of low quantum numbers. Also our results reduce to the \( AdS_5 \times S^5 \) results [7] in the undeformed case. Notice we can not perform the quantization for strings with very large energy because we had to assume that \( E/(\sqrt{\lambda} m) \) is small when the elliptic functions were expanded.
When $\kappa^2/m^2 > 1$ the oscillation number is
$$N = \frac{2\sqrt{\lambda}}{\kappa} \frac{\kappa}{\sin \theta_+} \int_0^{\pi/2} d\theta \sqrt{\sin^2 \theta_+ - \sin^2 \theta} \left[ 1 + \frac{1}{2} \gamma^2 (\sin^2 \theta_+ - \cos^2 \theta) \sin^2 \theta \right]. \quad (4.12)$$
After taking the limit of small deformation and expanding the elliptic functions we get
$$N = 2 \frac{\sqrt{\lambda}}{\kappa} E \left( \frac{m^2}{\kappa^2} \right) + \frac{1}{15} \gamma^2 \frac{\sqrt{\lambda}}{\kappa} \left[ -2 \left( 1 + \frac{3 \kappa^2}{2 m^2} - \frac{4 \kappa^4}{m^4} \right) E \left( \frac{m^2}{\kappa^2} \right) + \left( 1 - \frac{\kappa^2}{m^2} \right) \left( 1 + \frac{8 \kappa^2}{m^2} \right) K \left( \frac{m^2}{\kappa^2} \right) \right]. \quad (4.13)$$
The elliptic functions can be expanded in order to express the energy in terms of the oscillation number. In order to do that we have to assume that $E/(\sqrt{\lambda} m)$ is large and we obtain
$$E = N \left[ 1 + \frac{1}{4} \left( \frac{\sqrt{\lambda} m}{N} \right)^2 - \frac{1}{64} \left( \frac{\sqrt{\lambda} m}{N} \right)^4 + \ldots \right]$$
$$- \frac{1}{32} \gamma^2 \frac{\kappa m^2}{N} \left[ 1 - \frac{3}{16} \left( \frac{\sqrt{\lambda} m}{N} \right)^2 + \frac{5}{128} \left( \frac{\sqrt{\lambda} m}{N} \right)^4 + \ldots \right]. \quad (4.14)$$
Now the small expansion parameter is $\sqrt{\lambda} m/N$ so we can consider the regime of large quantum numbers.
To quantize the string we consider the Nambu-Goto action for the ansatz (4.2). Going to the Hamiltonian formalism we get
$$\Pi_\rho = \frac{\sqrt{\lambda} m \sin \theta \sqrt{G}}{\sqrt{\cosh^2 \rho - \hat{\rho}^2 - \hat{\theta}^2}} \dot{\rho},$$
$$\Pi_\theta = \frac{\sqrt{\lambda} m \sin \theta \sqrt{G}}{\sqrt{\cosh^2 \rho - \hat{\rho}^2 - \hat{\theta}^2}} \dot{\theta}, \quad (4.15)$$
and
$$H^2 = \cosh^2 \rho \left( \Pi_\rho^2 + \Pi_\theta^2 + \lambda m^2 G \sin^2 \theta \right),$$
$$= \cosh^2 \rho \left[ \Pi_\rho^2 + \Pi_\theta^2 + \lambda m^2 \sin^2 \theta \left( 1 - \gamma^2 \sin^2 \theta \cos^2 \theta \right) \right], \quad (4.17)$$
where in the last line we took the limit of small deformation. For the moment let us consider the string sitting at the center of AdS. Firstly we will quantize the string in the deformed sphere (4.1). Then $\Pi_\theta^2$ has to be realized like (3.21) (with $r$ replaced by $\theta$) using $\sqrt{-g} = \sqrt{G} \sin \theta$. For small deformation the Schrödinger equation reads
$$H^2 \Psi = - \frac{1}{\sin \theta} \frac{d}{d\theta} \left[ \sin \theta \frac{d\Psi}{d\theta} \right] + \gamma^2 \sin \theta \cos \theta (1 - 2 \sin^2 \theta) \frac{d\Psi}{d\theta}$$
$$+ \lambda m^2 \sin^2 \theta (1 - \gamma^2 \sin^2 \theta \cos^2 \theta) \Psi = E^2 \Psi. \quad (4.18)$$
Taking the first term as the unperturbed Hamiltonian the normalized eigenfunctions are written in terms of Legendre polynomials
\[ \Psi_n(\theta) = \sqrt{2n + 1} P_n(\cos \theta), \tag{4.19} \]
with eigenvalue \( E_{0,n}^2 = n(n + 1) \). For highly excited strings \( E_{0,n} = n \) and it agrees with lowest order term in (4.14) if the oscillation number is quantized as an integer.
The first order correction to the energy is then computed in perturbation theory. For highly excited states we find
\[
\delta_1 E_n^2 = \int_0^{\pi/2} d\theta \sin \theta \Psi_n^* \left[ \dot{\gamma}^2 \sin \theta \cos(1 - 2\sin^2 \theta) \frac{d}{d\theta} + \lambda m^2 \sin^2 \theta (1 - \dot{\gamma}^2 \sin^2 \theta \cos^2 \theta) \right] \Psi_n \\
= \frac{1}{2} \lambda m^2 \left( 1 - \frac{1}{8} \dot{\gamma}^2 \right). \tag{4.20} \]
The second order correction can also be computed in the large \( n \) limit and we get
\[
\delta_2 E_n^2 = \frac{1}{32} \lambda^2 m^4 \frac{1}{n^2} \left( 1 - \frac{1}{8} \dot{\gamma}^2 \right). \tag{4.21} \]
We then find for the energy of highly excited string states
\[
E_n = n \left[ 1 + \frac{1}{4} \frac{\lambda m^2}{n^2} - \frac{1}{64} \left( \frac{\lambda m^2}{n^2} \right)^2 + \ldots \right] - \frac{1}{32} \dot{\gamma}^2 \lambda m^2 \left[ 1 - \frac{3}{16} \frac{\lambda m^2}{n^2} + \ldots \right], \tag{4.22} \]
in perfect agreement with the classical expression (4.14) for the energy if the oscillation number is quantized as \( n \). This is expected because we are quantizing in the deformed sphere. It should be noticed that we could have used as the unperturbed Hamiltonian the first two terms of (4.18) which come together from the Laplacian on the deformed sphere but the result is the same.
We can now include the effects of the full ten dimensional deformed spacetime on \( \theta \) by taking \( \sqrt{-g} = \sqrt{G} \sin^3 \theta \cos \theta \ldots \) in \( \Pi_2^n \), that is,
\[
\Pi_2^n = \frac{1}{G \sin^3 \theta \cos \theta} \frac{d}{d\theta} \left( G \sin^3 \theta \cos \theta \frac{d}{d\theta} \right) \\
= \frac{1}{\sin^3 \theta \cos \theta} \frac{d}{d\theta} \left( \sin^3 \theta \cos \theta \frac{d}{d\theta} \right) - 2 \dot{\gamma}^2 \sin \theta \cos \theta (1 - 2\sin^2 \theta) \frac{d}{d\theta}. \tag{4.23} \]
Taking the first term of (4.23) as the unperturbed Hamiltonian in (4.17) the normalized eigenfunctions are expressed in terms of Jacobi polynomials as
\[
\Psi_n(\theta) = 2\sqrt{n + 1} P_n^{(0,1)}(1 - 2 \cos^2 \theta), \tag{4.24} \]
with eigenvalue \( E_{0,n}^2 = 4n(n + 2) \). Then \( E_{0,n} = 2n \) for large \( n \) and we get agreement with the lowest order term of (4.14) if the oscillation number is quantized as an even integer. Now the wave function is even about \( \theta = \pi/2 \).
The first order correction to the energy in the limit of large \( n \) is
\[
\delta_1 E_n^2 = \int_0^{\pi/2} d\theta \sin^3 \theta \cos \theta \Psi_n^* \left[ 2 \gamma^2 \sin \theta \cos \theta (1 - 2 \sin^2 \theta) \frac{d}{d\theta} \right] \Psi_n \\
+ \lambda m^2 \sin^2 \theta (1 - \gamma^2 \sin^2 \theta \cos^2 \theta) \Psi_n \\
= \frac{1}{2} \lambda m^2 \left( 1 - \frac{1}{8} \gamma^2 \right) - \gamma^2.
\]
(4.25)
This is precisely the result (4.20) with an extra term which came from the one derivative contribution in (4.23). This contribution is at the same order of \( 1/n \) as the other terms. In the case of the deformed Minkowski spacetime (3.25) they were of different orders and the corresponding term was of lower order. To second order we find
\[
\delta_2 E_n^2 = \frac{1}{128} \lambda^2 m^4 \left( 1 - \frac{1}{8} \gamma^2 \right) + \frac{1}{32} \gamma^2 \lambda m^2 \frac{n}{n^2}.
\]
(4.26)
and again we find an extra term which was absent in the deformed Minkowski spacetime case. Notice that the Hamiltonian mixes terms of order \( \gamma^2 \) and \( \lambda m^2 \) so that both corrections are present in the energy. Then the energy of highly excited states is
\[
E = 2n \left[ 1 + \frac{1}{4} \left( \frac{\lambda m^2}{(2n)^2} \right)^2 - \frac{1}{64} \left( \frac{\lambda m^2}{(2n)^2} \right)^2 + \ldots \right] - \frac{1}{32} \gamma^2 \frac{\lambda m^2}{2n} \left( 1 - \frac{3}{16} \frac{\lambda m^2}{(2n)^2} + \ldots \right) - \frac{1}{4} \gamma^2.
\]
(4.27)
The extra contributions coming from the corrections to the energy in (4.25) and (4.26) gave rise to the last term \( \gamma^2/(4n) \). Up to this term we have agreement with (4.14) if the oscillation number is quantized as \( 2n \).
Finally we can incorporate some contribution from the AdS sector along the lines of [20]. We can consider the term \( \Pi_\rho \) in (4.17) and its contribution to the energy. We find that for the unperturbed Hamiltonian there is a shift in the energy \( E \to E + 2(N_\rho + 1) \) where \( N_\rho \) is the quantum number associated to \( \rho \). Since we are considering highly excited oscillation states we can ignore \( N_\rho \) assuming that the wave function is concentrated at the origin of the AdS allowing us to take \( \rho = 0 \) so that the energy is just shifted by 2. The same result holds when we include the deformation because the structure of the equations in AdS remains intact.
5. Conclusions
We have considered pulsating strings in deformed Minkowski spacetime and \( AdS_5 \times S^5 \) for small deformation. At the classical level we have found the relation between the oscillation number and the energy. For \( AdS_5 \times S^5 \) the oscillation number can be expressed in terms of elliptic functions. We have different expansions for low and large energy strings. Since the oscillation number is an adiabatic invariant it can be used to probe the semi-classical regime when \( N \) is very large. We then performed the quantization for highly excited string states. In the deformed Minkowski spacetime the radial quantization lead to wave
functions which are generalized Laguerre polynomials. The energy can then be computed in perturbation theory and corresponds to the expression of the classical energy if the oscillation number is quantized as an even number. For the $AdS_5 \times S^5$ we have found that the oscillation number can be expressed in terms of elliptic functions which have different forms depending on whether we consider small or large energy. The quantization in the low energy regime, corresponding to short strings, has been performed in [7]. We have quantized the oscillatory motion of the string in the large energy case and found that wave functions are Jacobi polynomials. The energy was computed in perturbation theory and we found one extra term which was not predicted by the classical relation between the energy and the oscillation number. The oscillation number has to be quantized as an even number and the extra term depends exclusively on the deformation and not on the string tension or the winding number. It would be interesting to study the stability of this class of strings along the lines of [13].
It is known that the gauge theory operator dual to pulsating strings in $AdS_5 \times S^5$ is composed of non holomorphic products of the complex scalar fields and that there is a precise match between the energy and the anomalous dimension [21, 24, 34, 35]. It would be interesting to extend these results to the deformed case using the results of [17].
As a last remark we note that for a rotating string in $AdS_5 \times S^5$ with angular momentum $J$ and winding number $m$ it is known that its energy can be obtained from the $AdS_5 \times S^5$ energy simply by replacing $|m|$ by $|m + \frac{1}{2} \gamma J|$. This is also true for the fluctuations around the classical solution. For pulsating strings it is easily seen that no such property is present.
Acknowledgments
The work of S.G. was supported by Capes and CNPq. The work of V.O.R. is supported by CNPq grant 304116/2010-6 and FAPESP grant 2008/05343-5.
References
[1] For a review and earlier references see: N. Beisert, C. Ahn, L. F. Alday, Z. Bajnok, J. M. Drummond, L. Freyhult, N. Gromov, R. A. Janik et al., “Review of AdS/CFT Integrability: An Overview,” [arXiv:1012.3982 [hep-th]].
[2] For a review and earlier references see: N. Gromov, “Integrability in AdS/CFT correspondence: Quasi-classical analysis,” J. Phys. A 42, 254004 (2009).
[3] S. Frolov, A. A. Tseytlin, “Semiclassical quantization of rotating superstring in AdS(5) x S**5,” JHEP 0206, 007 (2002). [hep-th/0204226].
[4] N. Beisert, B. Eden, M. Staudacher, “Transcendentality and Crossing,” J. Stat. Mech. 0701, P01021 (2007). [hep-th/0610251].
[5] A. Tirziu and A. A. Tseytlin, “Quantum corrections to energy of short spinning string in AdS5,” Phys. Rev. D 78, 066002 (2008) [arXiv:0806.4758 [hep-th]].
[6] R. Roiban and A. A. Tseytlin, “Quantum strings in AdS5xS5: strong-coupling corrections to dimension of Konishi operator,” JHEP 0911, 013 (2009) [arXiv:0906.4294 [hep-th]].
[7] M. Beccaria, G. V. Dunne, G. Macorini, A. Tirziu and A. A. Tseytlin, “Exact computation of one-loop correction to energy of pulsating strings in AdS_5 x S^5,” J. Phys. A A44, 015404 (2011) arXiv:1009.2318 [hep-th].
[8] R. G. Leigh and M. J. Strassler, “Exactly Marginal Operators And Duality In Four-Dimensional N=1 Supersymmetric Gauge Theory,” Nucl. Phys. B 447, 95 (1995) [arXiv:hep-th/9503121].
[9] A. Mauri, S. Penati, A. Santambrogio and D. Zanon, “Exact results in planar N = 1 superconformal Yang-Mills theory,” JHEP 0511, 024 (2005) [arXiv:hep-th/0507282].
[10] O. Lunin and J. M. Maldacena, “Deforming field theories with U(1) x U(1) global symmetry and their gravity duals,” JHEP 0505, 033 (2005) [arXiv:hep-th/0502086].
[11] R. Roiban, “On spin chains and field theories,” JHEP 0409, 023 (2004) [arXiv:hep-th/0312218].
[12] D. Berenstein and S. A. Cherkis, “Deformations of N = 4 SYM and integrable spin chain models,” Nucl. Phys. B 702, 49 (2004) [arXiv:hep-th/0405215].
[13] S. A. Frolov, R. Roiban and A. A. Tseytlin, “Gauge - string duality for superconformal deformations of N = 4 super Yang-Mills theory,” JHEP 0507, 045 (2005) [arXiv:hep-th/0503192].
[14] S. Frolov, “Lax pair for strings in Lunin-Maldacena background,” JHEP 0505, 069 (2005) [arXiv:hep-th/0503201].
[15] N. Beisert and R. Roiban, “Beauty and the twist: The Bethe ansatz for twisted N = 4 SYM,” JHEP 0508, 039 (2005) [arXiv:hep-th/0505187].
[16] N. Gromov, V. Kazakov and P. Vieira, “Exact Spectrum of Anomalous Dimensions of Planar N=4 Supersymmetric Yang-Mills Theory,” Phys. Rev. Lett. 103, 131601 (2009) [arXiv:0901.3753 [hep-th]].
[17] N. Gromov and F. Levkovich-Maslyuk, “Y-system and beta-deformed N=4 Super-Yang-Mills,” J. Phys. A A44, 015402 (2011) arXiv:1006.5438 [hep-th].
[18] H. J. de Vega, A. L. Larsen, N. G. Sanchez, “Semiclassical quantization of circular strings in de Sitter and anti-de Sitter space-times,” Phys. Rev. D51, 6917-6928 (1995). [hep-th/9410219].
[19] S. S. Gubser, L. R. Klebanov and A. M. Polyakov, “A semi-classical limit of the gauge/string correspondence,” Nucl. Phys. B 636, 99 (2002) [arXiv:hep-th/0204051].
[20] J. A. Minahan, “Circular semiclassical string solutions on AdS(5) x S**5,” Nucl. Phys. B 648, 203 (2002) [arXiv:hep-th/0209047].
[21] J. Engquist, J. A. Minahan and K. Zarembo, “Yang-Mills duals for semiclassical strings on AdS(5) x S**5,” JHEP 0311, 063 (2003) [arXiv:hep-th/0310188].
[22] A. Khan, A. L. Larsen, “Spinning pulsating string solitons in AdS(5) x S**5,” Phys. Rev. D69, 026001 (2004). [hep-th/0310019].
[23] G. Arutyunov, J. Russo, A. A. Tseytlin, “Spinning strings in AdS(5) x S**5: New integrable system relations,” Phys. Rev. D69, 086009 (2004). [hep-th/0311004].
[24] M. Kruczenski and A. A. Tseytlin, “Semiclassical relativistic strings in S**5 and long coherent operators in N = 4 SYM theory,” JHEP 0409, 038 (2004) [arXiv:hep-th/0406189].
[25] M. Smedback, “Pulsating strings on $\text{AdS}(5) \times S^5$,” JHEP 0407, 004 (2004) [arXiv:hep-th/0405102].
[26] B. Chen and J. B. Wu, “Semi-classical strings in $\text{AdS}_4 \times \text{CP}^3$,” JHEP 0809, 096 (2008) [arXiv:0807.0802 [hep-th]].
[27] H. Dimov and R. C. Rashkov, “On the pulsating strings in $\text{AdS}_4 \times \text{CP}^3$,” Adv. High Energy Phys. 2009, 953987 (2009) [arXiv:0908.2218 [hep-th]].
[28] H. Dimov and R. C. Rashkov, “Generalized pulsating strings,” JHEP 0405, 068 (2004) [arXiv:hep-th/0404012].
[29] N. P. Bobev, H. Dimov and R. C. Rashkov, “Pulsating strings in warped $\text{AdS}(6) \times S^4$ geometry,” arXiv:hep-th/0410262.
[30] D. Arnaudov, H. Dimov and R. C. Rashkov, “On the pulsating strings in $\text{AdS}_5 \times T^{1,1}$,” arXiv:1006.1539 [hep-th].
[31] D. Arnaudov, H. Dimov and R. C. Rashkov, “On the pulsating strings in Sasaki-Einstein spaces,” AIP Conf. Proc. 1301, 51-58 (2010) arXiv:1007.3364 [hep-th].
[32] S. A. Frolov, R. Roiban, A. A. Tseytlin, “Gauge-string duality for (non)supersymmetric deformations of $\text{N}=4$ super Yang-Mills theory,” Nucl. Phys. B731, 1-44 (2005). [hep-th/0507021].
[33] L. F. Alday, G. Arutyunov and S. Frolov, “Green-Schwarz strings in TsT-transformed backgrounds,” JHEP 0606, 018 (2006) [arXiv:hep-th/0512253].
[34] N. Beisert, J. A. Minahan, M. Staudacher, K. Zarembo, “Stringing spins and spinning strings,” JHEP 0309, 010 (2003) [hep-th/0306139].
[35] J. A. Minahan, “Higher loops beyond the SU(2) sector,” JHEP 0410, 053 (2004) [hep-th/0405243]. | 2025-03-05T00:00:00 | olmocr | {
"pdf_page_numbers": [
[
0,
1012,
1
],
[
1012,
3850,
2
],
[
3850,
6928,
3
],
[
6928,
9952,
4
],
[
9952,
11996,
5
],
[
11996,
14535,
6
],
[
14535,
17165,
7
],
[
17165,
20183,
8
],
[
20183,
22368,
9
],
[
22368,
25552,
10
],
[
25552,
28315,
11
],
[
28315,
31092,
12
],
[
31092,
34192,
13
],
[
34192,
37219,
14
],
[
37219,
40139,
15
],
[
40139,
41696,
16
]
]
} |
53886ef9c2d9da6d8405d14c272c00645111887f | {
"olmocr-version": "0.1.59",
"pdf-total-pages": 8,
"total-fallback-pages": 0,
"total-input-tokens": 19171,
"total-output-tokens": 5318,
"fieldofstudy": [
"Physics"
]
} | New eccentric eclipsing binary in triple system: SY Phe
P. Zasche
Astronomical Institute, Faculty of Mathematics and Physics, Charles University
Prague, CZ-180 00 Praha 8, V Holešovičkách 2, Czech Republic
Abstract
Analyzing available photometry from Hipparcos, ASAS, Pi of the sky and Super WASP, we found that the system SY Phe is a detached eclipsing binary with similar components and orbital period about 5.27089 day. It has a slightly eccentric orbit, however the apsidal motion is probably very slow. The system undergoes an additional photometric variation on longer time scales superimposed on the eclipsing light curve. It also contains one distant component, hence the third light was also considered.
Key words: stars: binaries: eclipsing, stars: individual: SY Phe, stars: fundamental parameters
PACS: 97.10.-q, 97.80.-d
1 Introduction
The system SY Phe (= HD 9283 = HIP 7024) was discovered as a variable by Hoffmeister (1949). Later, Hoffmeister (1958) noted that the Algol-type curve is of BO Cep-type, and the additional variability is discussed. The Hipparcos satellite (Perryman et al., 1997) reveals two clearly-shaped eclipses and found the period about 5.27140 day. The photometric amplitude is about 0.5 mag. It is therefore remarkable that SY Phe is still classified as a ”Variable Star with rapid variations”, according to Simbad, or GCVS (Samus et al., 2012).
Spectral type of the system was firstly derived as F8 by Spencer & Jackson (1936), while Hoffmeister (1958) gave the type F4. Later Houk (1978) noted
Email address: [email protected] (P. Zasche).
the type F3/F5V. According to the Tycho data (Høg et al., 1997) the photometric index is $B_T - V_T = 0.512$ mag, and van Leeuwen (2007) derived the parallax of the system $\pi = 4.69 \pm 1.49$ mas.
However, the system consists of two visual components separated about 4" on the sky. According to the Washington Double Star Catalog (hereafter WDS1, Mason et al. 2001), the astrometric observations of this double do not show any significant change of the position angle. Therefore, the pair is only weakly gravitationally bounded and its semi-major axis is rather large. The Hipparcos observations indicate that the eclipsing variable is the A component (the brighter one).
2 Period analysis of the system
The system was observed by several automatic and robotized telescopes and surveys. Besides the old Hipparcos data, there exist also some photometry of SY Phe obtained by the ASAS survey (Pojmanski, 2002), ”Pi of the sky” (Burd et al., 2005), and Super WASP (Pollacco et al., 2006). For the photometry see Fig 1. All of these observations were used for deriving the times of minima of SY Phe. Both primary and secondary minima were derived, see Table 1. Some of the minima have relatively large uncertainty, however if we plot these minima times in the $O - C$ diagram (see Fig. 2), there is clearly
seen the eccentric orbit of the system. Primary and secondary minima are well-separated, but the analysis of apsidal motion is still difficult.
The method of apsidal motion analysis was described elsewhere, e.g. Giménez & García-Pelayo (1983), Giménez & Bastero (1995). All of the minima times given in Table 1 were used for the analysis. Unfortunately, the set of our data is still very limited and leads to many different results. From these different solutions, we chose to present here two, which have the lowest rms of the fit. These are given in Table 2. As one can see, the change of argument of periastron is still rather small and one cannot easily derive the correct solution. However, the very fast apsidal motion of about 17 years is less probable due to the typical longer apsidal periods in systems of this type. On the other hand, the Solution II (see Table 2) presents so slow apsidal motion, that the value of \( \dot{\omega} = \frac{d\omega}{dt} \) is even so low that the apsidal period \( U \) grows to many thousand years. The time spread of epochs of photometry is still too short yet, hence new precise observations are therefore needed to confirm this hypothesis with higher conclusiveness.
3 Light curve analysis of the system
For the light curve analysis, there arises several problems with the available data: the photometry from the ”Pi of the sky” survey is unfiltered, and the ASAS data have only poor coverage. Relatively best data coverage is provided by the SWASP data, but these are not obtained in any standard photometric filter. The filter used here is a special broadband filter covering a passband from 400 to 700 nm. Therefore, for the use of these data in our light curve
Table 1
The heliocentric minima times used for the analysis.
| HJD | Error | Type | Filter | Source |
|------------|--------|------|--------|----------------|
| 2400000+ | | | | |
| 48470.71478| 0.00269| Prim | Hp | Hipparcos |
| 48473.34287| 0.00527| Sec | Hp | Hipparcos |
| 52471.31528| 0.00258| Prim | V | ASAS |
| 52473.93542| 0.00685| Sec | V | ASAS |
| 53541.30923| 0.00158| Prim | V | ASAS |
| 53543.92347| 0.02567| Sec | V | ASAS |
| 53709.98250| 0.00564| Prim | - | Pi of the sky |
| 53712.58972| 0.01718| Sec | - | Pi of the sky |
| 53907.61660| 0.00210| Sec | SWASP | SWASP |
| 53936.62590| 0.00017| Prim | SWASP | SWASP |
| 53965.59620| 0.00044| Sec | SWASP | SWASP |
| 53981.40956| 0.00065| Sec | SWASP | SWASP |
| 54002.49111| 0.00225| Sec | SWASP | SWASP |
| 54031.49955| 0.00176| Prim | SWASP | SWASP |
| 54047.31409| 0.00066| Prim | SWASP | SWASP |
| 54297.66860| 0.00205| Sec | SWASP | SWASP |
| 54334.55734| 0.00427| Sec | SWASP | SWASP |
| 54342.48405| 0.00078| Prim | SWASP | SWASP |
| 54363.56689| 0.00044| Prim | SWASP | SWASP |
| 54379.38040| 0.00023| Prim | SWASP | SWASP |
| 54416.27834| 0.00245| Prim | SWASP | SWASP |
| 54437.35735| 0.00290| Prim | SWASP | SWASP |
| 54439.97948| 0.00158| Sec | SWASP | SWASP |
| 54611.29793| 0.00127| Prim | - | Pi of the sky |
| 54613.91990| 0.00132| Sec | - | Pi of the sky |
| 54632.38124| 0.00295| Prim | V | ASAS |
analysis, we used the filter with the most similar transmission curve, which is the $V_T$ filter of Tycho experiment onboard of the Hipparcos satellite. The PHOEBE programme (see e.g. Prša & Zwitter 2005), based on the Wilson-Devinney algorithm (Wilson & Devinney 1971), was used for the analysis.
Due to missing information about the stars, and having only the light curve in one filter, some of the parameters have to be fixed. At first, the ”Detached binary” mode (in Wilson & Devinney mode 2) was assumed for computing. The value of the mass ratio $q$ was set to 1. The limb-darkening coefficients were interpolated from van Hamme’s tables (see van Hamme 1993), the linear cosine law was used. The values of the gravity brightening and bolometric albedo
Table 2
The parameters of the apsidal motion.
| Parameter | Value - Solution I | Value - Solution II |
|-----------|--------------------------|--------------------------|
| HJD0 | 2454163.2646 ± 0.0021 | 2454163.2642 ± 0.0020 |
| P [day] | 5.27088632 ± 0.00000631 | 5.27088780 ± 0.00000636 |
| ε | 0.0072 ± 0.0012 | 0.0168 ± 0.0059 |
| ω₀ [deg] | 217.5 ± 1.3 | 251.6 ± 0.9 |
| ω̇ [deg/cycle] | 0.306 ± 0.013 | < 0.0001 |
| U [yr] | 17.1 ± 0.8 | > 10000 |
| Pₚ [day] | 5.27533454 ± 0.00000632 | 5.27088779 ± 0.00000636 |
Table 3
The light-curve parameters of SY Phe, as derived from our analysis.
| Parameter | Value | Parameter | Value |
|-----------|-------|-----------|-------|
| i [deg] | 87.0 ± 1.3 | F₁ = F₂ | 1 (fixed) |
| Ω₁ | 12.504 ± 0.059 | A₁ = A₂ | 0.5 (fixed) |
| Ω₂ | 12.754 ± 0.045 | G₁ = G₂ | 0.32 (fixed) |
| T₁ [K] | 6890 (fixed) | Derived physical quantities: |
| T₂ [K] | 6351 ± 260 | M₁ [M☉] | 1.36 ± 0.14 |
| L₁ [%] | 52.1 ± 1.3 | M₂ [M☉] | 1.36 ± 0.14 |
| L₂ [%] | 32.9 ± 1.0 | R₁ [R☉] | 1.55 ± 0.09 |
| L₃ [%] | 15.0 ± 2.3 | R₂ [R☉] | 1.52 ± 0.08 |
Coefficients were set at their suggested values for convective atmospheres (see Lucy 1968), i.e. G₁ = G₂ = 0.32, A₁ = A₂ = 0.5. Therefore, the quantities which could be directly calculated from the light curve are the following: the relative luminosities Lᵢ, the temperature of the secondary T₂, the inclination i, and the Kopal’s modified potentials Ω₁ and Ω₂. The synchronicity parameters F₁ and F₂ were also fixed at values of 1.
Because the SWASP data have only limited angular resolution, the third component of the visual double also influences the photometry and the third light has to be considered. For the light curve analysis we used the Solution II of the apsidal motion, hence there is no need of apsidal advance of the ω angle. The ephemerides were also taken from this solution. The temperature of the primary component was fixed at a value of 6890 K (Wright et al., 2003). Here we have to emphasize once again that the temperature was only adopted on the basis of a statistical value for the considered spectral type and not from a good spectroscopic determination. The best fit we were able to reach is presented in Fig. 3. The fit is slightly worse near the primary minimum, which could suggest the primary minimum is more narrow than the secondary minimum. The parameters of the fit are given in Table 3. Regarding the values of M and R in absolute units, see below.
There can be two different explanations of the poor fit to the data. First, the orbital eccentricity of the binary should be higher, which cause the duration
Fig. 3. The light curve fit of the SWASP data from PHOEBE.
of the primary eclipse shorter, which is visible in the fit in Fig. 3. Other explanation is the fact that there are some short-periodic variations outside of the eclipses. Therefore, these variations are also presented in the eclipses and these can shift the data points a bit. The magnitude of the outside-eclipse variations is up to 0.04 mag. If these variations are physical or instrumental is still an open question. Moreover, we found that there is also a season-to-season variation of the light curve. The level of outside of eclipse brightness was changed of about 0.03 mag during the period of SWASP observations (more than 500 days).
We were trying to find some periodicity of these data after removing the light curve fit. This result is shown in Fig. 4, where a steady increase of brightness is superimposed with the sinusoidal variation with the period of about 248.6 days. Having such a long period of variations, we can only hardly identify these modulations as those making the system classified as a system "with rapid variations" as stated in Simbad. All of these findings make any analysis difficult and new more precise standard photometry in different photometric filters would be very helpful.
4 Discussion and conclusions
The first light curve and period analyses of the system SY Phe were performed. Dealing with no spectroscopy of the target, one has to consider some assumptions and many of the physical parameters are still affected by relatively large errors.
Fig. 4. The residuals of the light curve fit with respect to time. The fitted red line represents the fit with period of about 249 days and secular decrease of magnitude. Confidence level of 95% is also shown.
However, we can roughly estimate the internal structure constants for both apsidal motion solutions. Assuming the two eclipsing components have masses of about 1.36 M\(_\odot\) (spectral type F4), then the semimajor axis of the orbit is about 17.8 R\(_\odot\), which yields the values of radii of both components (see Table 3). Using these values, one can calculate the internal structure constants and compare these values with the theoretical ones, e.g. by Claret (2004). The Solution I gives log k\(_2\) value of about 0.686, while Solution II gives log k\(_2\) = -2.198. The latter value agrees well with the theoretical values by Claret (2004), assuming the F4 star is on main sequence and is about \(5 \cdot 10^8\) yr old.
The light curve solution also provides the first rough estimation about the third light from the distant component. Our result of the third light value (15% of the total light) is in excellent agreement with the \(\Delta M\) value provided by the WDS catalogue. The nature of the long-term photometric variations with period about 249 days still remains an open question. Therefore, new more detailed analysis is still needed, especially based on new spectroscopic data together with the photometric observations obtained in various photometric filters.
5 Acknowledgments
We thank the "ASAS", "SWASP" and "Pi of the sky" teams for making all of the observations easily public available. This investigation was supported
by the Czech Science Foundation grant no. P209/10/0715, by the Research Program MSM 0021620860 of the Ministry of Education of Czech Republic, and by the grant UNCE 12 of the Charles University in Prague. This research has made use of the SIMBAD database, operated at CDS, Strasbourg, France, and of NASA’s Astrophysics Data System Bibliographic Services.
References
Burd, A., et al. 2005, NewA, 10, 409
Claret, A. 2004, A&A, 424, 919
Cox, A. N. 2000, Allen’s Astrophysical Quantities
Giménez, A., García-Pelayo, J. M. 1983, Ap&SS, 92, 203
Giménez, A., Bastero, M. 1995, Ap&SS, 226, 99
Hög, E., Bässgen, G., Bastian, U., et al. 1997, A&A, 323, L57
Hoffmeister, C. 1949, AAAN, 12, Nr 1, 23
Hoffmeister, C. 1958, Veroeffentlichungen der Sternwarte Sonneberg, 3, 333
Houk, N. 1978, Ann Arbor : Dept. of Astronomy, University of Michigan: distributed by University Microfilms International
Lucy, L. B. 1968, ApJ, 151, 1123
Mason, B. D., Wycoff, G. L., Hartkopf, W. I., Douglass, G. G., Worley, C. E. 2001, AJ, 122, 3466
Perryman, M. A. C., Lindegren, L., Kovalevsky, J., et al. 1997, A&A, 323, L49
Pojmanski, G. 2002, AcA, 52, 397
Pollacco, D. L., et al. 2006, PASP, 118, 1407
Prša, A., Zwitter, T. 2005, ApJ, 628, 42
Samus N.N., Durlevich O.V., Kazarovets E V., et al. General Catalog of Variable Stars (GCVS database, Version 2012Feb)
Spencer Jones H. Jackson J., 1936, Proper Motions of Stars in the Zone Catalogue -40 to -52 degrees of 20843 Stars for 1900, His Majesty’s Stationery Office, London, 1936
van Hamme, W. 1993, AJ, 106, 2096
van Leeuwen, F. 2007, A&A, 474, 653
Wilson, R. E., Devinney, E. J. 1971, ApJ, 166, 605
Wright, C. O., Egan, M. P., Kraemer, K. E., & Price, S. D. 2003, AJ, 125, 359 | 2025-03-04T00:00:00 | olmocr | {
"pdf_page_numbers": [
[
0,
1606,
1
],
[
1606,
2914,
2
],
[
2914,
4929,
3
],
[
4929,
7404,
4
],
[
7404,
10196,
5
],
[
10196,
11746,
6
],
[
11746,
13409,
7
],
[
13409,
15113,
8
]
]
} |
7781615491f549926620b1c695537a2ed87c209c | {
"olmocr-version": "0.1.59",
"pdf-total-pages": 11,
"total-fallback-pages": 0,
"total-input-tokens": 33618,
"total-output-tokens": 12871,
"fieldofstudy": [
"Computer Science"
]
} | In Search of the Flocks: How to Perform Onomasiological Queries in an Ancient Greek Corpus?
Alek Keersmaekers, Toon Van Hal
University of Leuven
Blijde-Inkomststraat 21, 3000 Leuven, Belgium
{alek.keersmaekers; toon.vanhal}@kuleuven.be
Abstract
This paper explores the possibilities of onomasiologically querying corpus data of Ancient Greek. The significance of the onomasiological approach has been highlighted in recent studies, yet the possibilities of performing ‘word-finding’ investigations into corpus data have not been dealt with in depth. The case study chosen focuses on collective nouns denoting animate groups (such as flocks of people, herds of cattle). By relying on a large automatically annotated corpus of Ancient Greek and on token-based vector information, a longlist of collective nouns was compiled through morpho-syntactic extraction and successive clustering procedures. After reducing this longlist to a shortlist, the results obtained are evaluated. In general, we find that τίθωνος can be considered to be the default collective noun of both humans and animals, becoming especially prominent during the Hellenistic period. In addition, specific tendencies in the use of collective nouns are discerned for specific semantic classes (e.g. gods and insects) and over time. Throughout the paper, special attention is paid to methodological issues related to onomasiologically searching.
Keywords: onomasiology, data querying, collective nouns, Ancient Greek
1. Introduction
This paper explores the possibilities of onomasiologically querying corpus data of Ancient Greek. The significance of the onomasiological approach has been highlighted in recent studies, yet the possibilities of performing ‘word-finding’ investigations into corpus data have not been dealt with in depth. English has a wide range of words denoting groups of animals or people, such as a “pack of dogs”, “a school of fish” and “a gang of bandits”. This paper aims to explore how similar collective nouns can be detected in the Ancient Greek corpus by adopting an onomasiological approach to the data.
The paper is organized as follows. A survey of the state of the field (Section 2) precedes an outline of our strategies adopted to tracing collective nouns in Greek (Section 3). Section 4 analyzes various groups of animate entities in Ancient Greek by means of corpus data and discusses onomasiological change in Ancient Greek. In the concluding part (Section 5), alternative approaches and further avenues are discussed. The case studied in this paper has identifiable morpho-syntactic characteristics (see Section 3.2), but in the future it should be also made possible to find words expressing a certain concept for which the availability of syntactical and morphological annotation is not helpful.
2. State of the field
2.1 Onomasiological searching
Corpus-based research is usually based on a (set of) predefined term(s), of which the meaning is traced. In addition to this semasiological or ‘sense-finding’ approach, it is also conceivable to take a certain meaning (concept or notion) as a starting point, and examine which terms are used to shape this meaning in a corpus. In recent decades, linguists have strongly emphasized the importance and relevance of such an onomasiological or ‘word-finding’ approach (see e.g. Grzega, 2002; Geeraerts, 2009; Fernández-Dominguez, 2019), and more recently there have been increasing advocates of the onomasiological approach among conceptual historians too (see e.g. Müller & Schmieder, 2016; Cananu, 2019). For obvious reasons, querying corpora with a semasiological, word-based approach is much easier than meaning-based onomasiological queries, because unlike a meaning a term is a tangible starting point. In a methodological survey paper published against the backdrop of a corpus-based computational historical semantics project, Bernhard Jussen and Gregor Rohmann mention the onomasiological approach, yet the case-studies they present are semasiological in nature (Jussen & Rohmann, 2015). While there has been some research on querying onomasiological dictionaries (Kipfer, 1986; Sierra, 2008; Moerdijk et al., 2008 on the development of ‘semagrams’), the literature on how to onomasiologically querying corpora is limited (see McGillivray, 2020; see also Kutuzov, 2020). In general, onomasiological search strategies generally boil down to making use of annotations that approximate the concept under investigation as much as possible, the results of which are complemented through bottom-up approaches (see e.g. Goossens, 2013). Hence, this presupposes the presence of an annotated corpus, which is a demanding and time-consuming investment, if such a corpus is not yet available (see Mehl, 2016: 50; 92; Atallah et al., 2018). The type of annotations required depends on the onomasiological task at stake. For certain tasks, part-of-speech tags can be helpful, while for others tasks more detailed morphological, syntactic, semantic and/or pragmatic information is needed.
2.2 Collective nouns
Words as ‘flock’ and ‘herd’ are styled quantifying collectives and collective nouns by Biber et al. (2003: 61-62). The terms have been criticized for being too vague (see the references in Dedé, 2012). Some scholars have treated collective nouns as classifiers or (‘classifier constructions’, cf. Lehrer, 1986). Aikhenvald (2000: 115-116) however argues why such terms do not meet the criteria of genuine
Collective nouns have attracted much scholarly attention. Ancient Greek, even though the ‘oldest grammar of the language’ (West, 1991: 116) and the ‘oldest grammar of the world’ (Vitae, 1896: 103-116) are well-known corpora for Ancient Greek with the level of granularity that is necessary for the research described in this paper has been created. We therefore made use of a bottom-up approach that has become highly popular in recent years to represent semantics computationally, the so-called ‘distributional’ approach to semantics, where the meaning is represented by vectors of real numbers (with semantically similar words or constructions receiving mathematically similar vectors). These vectors are based on the context patterns of words in large text corpora (see Erk, 2012; Lenci, 2018 for more detail). Distributional semantic methods have been applied to Ancient Greek by Boschetti (2010), Rodda, Senaldi & Lenci (2017), Rodda, Probert & McGillivray (2019), Keersmaekers & Van Hal (2021), and Perrone et al. (2021). For this paper we use the implementation of Keersmaekers & Van Hal (2021), which calculates word vectors on the basis of PPMI-scaled syntactic dependency-based co-occurrence counts in the GLAUx corpus, with an SVD-based dimension reduction to 100 latent dimensions.
3. Identifying collective nouns in Ancient Greek
3.1 Starting points of the research
As pointed out in 2.1, onomasiological queries highly benefit from corpus annotations. This is especially true for Ancient Greek, a language with a highly flexible word order and complex inflectional morphology, which reduces the effectiveness of strictly form-based (as opposed to lemma-, morphology- and syntax-based) queries. For Ancient Greek, the most well-known corpora are the Greek treebanks (several annotators, consisting of dependency trees with syntactic, morphological and lemma annotation: see Celano and Keersmaekers et al., 2014), which are manually annotated but not extremely large (1.5M tokens), and the Diorisis corpus (a corpus that is annotated for lemmas and morphology, cf. Vatri and McGillivray, 2018), which is relatively sizable (10.2M tokens) but is automatically annotated and does not contain syntactic information. We therefore made use of the (so far unreleased) GLAUx corpus (Keersmaekers, 2021), a corpus containing literary (8th century BC-3th century AD) and documentary texts (3th century BC-8th century AD) automatically annotated for lemmas, morphology and syntax (28.8M tokens): see Keersmaekers (2021) for an evaluation of the quality of the annotation, which was high enough not to provide any substantial obstacles for the research described below.
Although some steps for semantic annotation of Greek have been taken (see Celano & Crane, 2015 and Keersmaekers, 2020 for semantic role annotation; Bizzoni et al., 2014 and Biagetti et al., 2021 for Ancient Greek WordNet), so far no large-scale semantically annotated corpus resource for Ancient Greek with the level of granularity that is necessary for the research described in this paper has been created. We therefore made use of a bottom-up approach that has become highly popular in recent years to represent semantics computationally, the so-called ‘distributional’ approach to semantics, where the meaning is represented by vectors of real numbers (with semantically similar words or constructions receiving mathematically similar vectors). These vectors are based on the context patterns of words in large text corpora (see Erk, 2012; Lenci, 2018 for more detail). Distributional semantic methods have been applied to Ancient Greek by Boschetti (2010), Rodda, Senaldi & Lenci (2017), Rodda, Probert & McGillivray (2019), Keersmaekers (2020), Keersmaekers & Van Hal (2021), and Perrone et al. (2021). For this paper we use the implementation of Keersmaekers & Van Hal (2021), which calculates word vectors on the basis of PPMI-scaled syntactic dependency-based co-occurrence counts in the GLAUx corpus, with an SVD-based dimension reduction to 100 latent dimensions.
3.2 Morpho-syntactic extraction
In Greek, collective nouns are syntactically well-defined, since they are usually accompanied by a so-called partitive genitive (Benvenuto, 2013). Based on the GLAUx corpus, we could extract all constructions of type ‘noun + animate entity in the genitive plural, having ‘attribute’ as its syntactic feature’. The animacy was determined via supervised machine learning techniques, training a deep learning model1 on data annotated for the animacy class of the lemma as the dependent variable and a 100-dimensional word vector of the lemma (as described in 3.1) as the independent variable(s) (see Keersmaekers, 2020: 103-116). Our training data was an animacy lexicon containing trained with stochastic gradient descent using back-propagation.
1 As implemented in R package h2o (LeDell et al., 2022), using a multi-layer feedforward artificial neural network 74 propagation.
486 animate and 2650 inanimate entities; animate entities yielded precision of 0.941 and recall of 0.914 – an estimation via 10-fold cross validation on the training data. On this basis, 1991 lemmas were labeled as animate, which allowed us to identify possible collective nouns.
Our approach is not infallible: in addition to possible errors in the automatically annotated data, we should note that there are a number of alternative constructions that can express collective nouns and that are not included in the extracted data. For example, the genitive can sometimes be replaced with an adjective (e.g. the LSJ dictionary of Ancient Greek, Jones et al., 1996, cites μελισσαῖον ‘swarm of bees’, with the adjective μελισσάς ‘consisting of bees’). Some collective nouns have no (need for) further attributive specification especially if the animal is already lexicalized in the collective noun itself (e.g. βοοκόλιον “a group of cows”, συβόσιον “a group of pigs”, αἰπόλιον “a group of goats”). Obviously, plural morphology might also be used to indicate a group of animate entities (e.g. simply αίγες ‘goats’ instead of αἰπόλιον αἰγών ‘a flock of goats’). Finally, constructions with a genitive in the singular are conceivable (e.g. ‘a swarm of vermin’ in English). Of the extracted lemmas (5488), only lemmas with a frequency of $\geq 5$ (frequency of the lemma accompanied by an animate genitive plural) were retained (1266 in total). These lemmas thus count as potential collective nouns, out of which we attempted to identify the real collective nouns using several computational techniques.
### 3.3 Visualization and clustering techniques
The query defined in section 3.2 likely has a high recall, since we expect most collective nouns to occur in the construction defined there, even though there are some other ways to express groups of animate entities as discussed above. However, its precision is rather low, since many nouns occurring in the noun + animate genitive plural construction are not collective nouns: this construction admits many more types of nouns such as body parts (e.g. ‘the legs of the horses’), possession relations (e.g. ‘the money of the men’) and so on. To retrieve collective nouns from this large set (1266 nouns), we used a variety of dimension reduction and clustering techniques to find structure in our dataset, as well as lexicographical data (the LSJ dictionary, Jones et al., 1996) and corpus examples from the GLAUs corpus (in case of doubt) to identify collective nouns in these structured datapoints. The dimension reduction and clustering techniques were applied to the cosine distances between the nouns in our dataset, which mathematically represent the ‘semantic distances’ between the nouns (see Erk, 2012: 636-637).
As a first step, we made use of t-SNE (t-distributed stochastic neighbor embedding, Van den Maaten and Hinton, 2008), a dimension reduction technique that allows us to represent high dimensional data (a 1266x1266 matrix representing the cosine distances between nouns) in a low-dimensional (in our case two-dimensional) space, with words that are similar in meaning occurring close to each other on the tsnemap.² This enables us to find structure in the data and identify which words are worth looking at to retrieve collective nouns. For instance, the cluster on the bottom right of Fig. 2 (in dark yellow) contains words that clearly refer to body parts (e.g. μύς ‘muscle’, γαστήρ ‘belly’, ὑρὶς ‘hair’). It is unlikely that a collective noun would occur in such a cluster, so these words can safely be discarded after identifying the thematic coherence of the cluster. Instead, on the bottom/center-left of the plot there are several clusters that clearly contain many collective nouns: military units (in red: e.g. ἔλιος, λόγχος, σώλαμος), words referring to herding (in dark blue, with several words that mean ‘a flock or herd’, such as αἰπόλιον, ἀγέλη, πῶς, but also some non-collective nouns such as γομάς ‘herdsman’) and a small cluster of words referring to groups in general (in yellow: πλῆθος, ὄχλος, πληθύς, ὀξύλος, ὀξυμός, σμῆνος); additionally, a little more doubtful are the clusters in pink (generally containing words related to transport such as ἄραμα ‘wagon’, φορτίον ‘load’ and ἰππος ‘horse’ but also some collective nouns such as σωφορίς ‘pair of horses’ and κτήνος ‘beast’, but also ‘flock’), dark green (mainly poetic words referring to family such as φόλον ‘tribe’, but maybe also ‘swarm’, γάτεβον ‘family’, but also unrelated poetic words such as σμάντορ ‘leader’) and light blue (two words referring to the action of collecting or coming together but maybe also to a collection or group, viz. ἄραμα and σναύραμ). After identifying these clusters, we used dictionaries and corpus data to check whether each word occurring in these clusters is actually a collective noun.
Next, we used two cluster techniques that are prevalent in corpus linguistics to identify additional nouns that we may have missed with the t-SNE analysis, viz. partitioning around medoids (PAM) and hierarchical agglomerative clustering (AGNES).³ The former technique divides the data into a predetermined number ($k$) of clusters. After experimenting with the values for $k$, in the end we settled for a small number of $k=20$ clusters. The latter technique hierarchically clusters all nouns into a tree, with similar words occurring in the same ‘branches’ of the tree – a subpart of the tree, containing many collective nouns, is shown in Fig. 3. As with the t-SNE analysis, we analyzed the thematic coherence of each cluster that was formed (in the case of PAM, simply each of the 20 clusters; in the case of AGNES, branches of the tree occurring roughly at the same height), and looked into more detail at the more ‘promising’ clusters containing many collective nouns. These techniques allowed us to identify some additional collective nouns that we had previously missed: these were especially words in the festive or public domain including θῦσος, χορός and σύλλογος, along with some words thematically related to the words we previously found such as σύστημα (a military unit, or also a group in general) and νύφος (literally ‘cloud’, but also a group of people or animals).
² We made use of the R package Rtse (Krijthe and Van der Maaten, 2018). We used a perplexity of 5, theta of 0.0 and 5 iterations.
³ As implemented in R package cluster (Maechler et al., 2022). We used out-of-the-box settings.
Figure 2: Visualization of the t-SNE embeddings
Figure 3: Subpart of the tree of hierarchical agglomerative clustering (AGNES)
4. Results and discussion
4.1 Discussion of shortlist in general
In total, we have traced 40 collective nouns (see Table 1). The dataset on which this paper is based is available through a csv-file.
5 collective nouns appeared more than 100 times in the data. At the top of the list, without a doubt, is the word θίασος (token frequency 998), which can be regarded as the default collective noun for animate referents. There is no semantic class in which this collective noun does not occur. In some cases, πλῆθος is used too, which is according to most dictionaries merely an Ionian variant (this, however, should be checked against the data). πλῆθος is followed by ἄγγελος (215), φίλον (129), τάξις (125) and χορός (109). Some words are exclusively used as collective nouns for animals, such as πῦ (10), αἰπόλιον (8), θίασος (7) and συβόσιον (5), while words that occur only with human referents are more numerous, viz. λόγος (37); τάγμα (34); σύνταγμα (22); συσσίτιον (12); θίασος (10); συλλογος (10); ἀθροισμα (6).
4.2 Discussion of shortlist in general
In total, we have traced 40 collective nouns (see Table 1). The dataset on which this paper is based is available through a csv-file.
5 collective nouns appeared more than 100 times in the data. At the top of the list, without a doubt, is the word θίασος (token frequency 998), which can be regarded as the default collective noun for animate referents. There is no semantic class in which this collective noun does not occur. In some cases, πλῆθος is used too, which is according to most dictionaries merely an Ionian variant (this, however, should be checked against the data). πλῆθος is followed by ἄγγελος (215), φίλον (129), τάξις (125) and χορός (109). Some words are exclusively used as collective nouns for animals, such as πῦ (10), αἰπόλιον (8), θίασος (7) and συβόσιον (5), while words that occur only with human referents are more numerous, viz. λόγος (37); τάγμα (34); σύνταγμα (22); συσσίτιον (12); θίασος (10); συλλογος (10); ἀθροισμα (6).
Some words designating 'flock' or 'group', such as 'clouds of bees', II. 2.87-89) are not captured in the data. Some words designating 'lock' or 'group', such as βόσκημα, which is a club word (like many other words in this list) referring to a group of judges. Although we included several club words in our shortlist, we excluded those where people assemble for a highly specialized purpose, in this case making a judicial decision. It should be emphasized that by eliminating these words (as well as the δήμος and equivalents mentioned above) we have eliminated some clear examples of collective nouns. However, we felt that the inclusion of these words would give way to considerable noise in the data. Conversely, it should be noted that not all instances included in the shortlist unambiguously refer to a collective noun. This is certainly the case for a word like τάξις, which is very polysemous (e.g. also 'order', 'class', 'rank' etc.). Due to the scale of our undertaking, it was infeasible to inspect the data token-wise. We are however aware that our type-based approach is vulnerable to noise.
4.2 Classes of animals and humans
Next, we divided the lemmas of animals and people in a number of subgroups. These subgroups were semi-automatically created: through hierarchical clustering (AGNES) of the vectors of these lemmas, we first checked which of them were highly semantically related and created subgroups on this basis, but the final groups were created with a high degree of human control (e.g. if the cluster algorithm would cluster a specific fish with a bird together, we would put this fish in the group ‘water animals and fish’ rather than ‘birds’). Fig. 4 shows the frequency of collective nouns with the most frequent groups of animals,
Table 1: Collective nouns denoting animals and humans
| θίασος | 8 0% | 7 1% | 1 0% |
|--------|-----|-----|-----|
| αἰπόλιον | 7 0% | 7 1% | 0 0% |
| ἀθροισμα | 6 0% | 0 0% | 6 0% |
| συβόσιον | 5 0% | 5 1% | 0 0% |
4 The data in Johnston (2019) suggests that collective nouns are rare in Homeric similes.
Figure 4: Variation in the presence of animal collective nouns according to semantic subgroups
Figure 5: Variation in the presence of human collective nouns according to semantic subgroups
vist. animals in general, birds, water animals and fish, insects, livestock and wild mammals. The default word πλήθος is used in all categories, although remarkably less frequently in the case of livestock, insects and, to a lesser extent, birds. It is also notable that ἄγελη (often translated as ‘flock’) is used in every subgroup, and not exclusively for livestock. In the category of insects, the subgroup where ἄγελη is underrepresented, one can notice the use of a number of specific collective nouns that are almost exclusively used for insects, viz. ἵππος, σμῆνος and νέφος, the latter of which is also used for birds (see also 4.3). This specialization seems to be rather atypical: φῦλον, νομή and, to a lesser extent, ζεῦγος are animal collectives that can be used for almost any subgroup. One should also notice the high degree of ‘other’ with livestock: besides the ‘default’ options of ἄγελη and πλήθος, Ancient Greek has a large number of specialized words for livestock (e.g. βοικόλον ‘group of cows’).
The humans can also be divided into a number of subgroups. Fig. 5 makes a distinction between the most frequent subgroups, viz. gods, humans in general, men, professions and soldiers. The data shows that in case of the gods certain collective nouns, viz. χορός and φῦλον, outnumber the ‘default’ use of πλήθος. Among the military category one finds the most specialized collective nouns, such as τάξις, δῆλον, λόχος and φάλαγξ. Strikingly, στρατιά and στρατός (‘army’) are hardly represented in this category. In general, it is noticeable that there are plenty combinatory possibilities.
4.3 Degree of specialization
This leads us to the question of to what extent there are exclusive combinations in Greek, showing one-to-one correspondences between a specific collective noun and a specific animate type, like the English ‘mummer of crows’.
Table 2 shows the top results of a Pointwise Mutual Information (PMI) calculation, a measure of association showing whether two variables co-occur more frequently than expected based on their individual frequencies (see Gries 2010: 275-277 for more detail). We have only included collocational combinations that occur at least five times. The results show that in some cases there is a clear etymological connection between the collective noun and the species at stake (σμῆνος, αἰπόλιον, φυσικόλον), thus logically excluding alternative combinations (such as βουκόλον and αἰπέ). Excluding these words and the Homeric word πῶυ, which is exclusively used for δις ‘sheep’, it seems that especially specific insects (μέλισσα ‘bee’, ἀκρίς ‘grasshopper’, the same goes for less frequent insects such as κηφήν ‘drone’ and σφιξ ‘wasp’) are combined with specific collective nouns (νέφος: ἤπατος; σμῆνος), which are rarely used for animals other than insects (except νέφος which is also often combined with birds). A group of pigeons (τριγών or περιστρέφοι) is mostly referred to as κηφήν, likely indicating a duo.
Table 2: Strongest PMI associations between collective nouns and the genitives occurring with them
Closer inspection reveals that some of the exclusive correspondences in Table 2 might be somewhat deceptive, for example, because all the attestations come from one author. This is the case for the association between δῃς and ὅχλος, which seems to be a personal style characteristic of Origenes.
4.4 Diachronic developments
In the previous sections, we mapped the onomasiology of Ancient Greek collective nouns in a static way. However, this onomasiology is obviously prone to semantic change, i.e. the terms used to express groups of animate entities change over time. This section will consider how computational methods can shed light on this onomasiological change. To this aim, we have divided the data into archaic (8th-6th century BC), classical (5th-4th century BC), Hellenistic (3rd-1st century BC) and Roman eras (1st-4th century AD) (see Fig. 6 and Fig. 7). However, caution is advised here: for instance, we have almost exclusively epic texts from the archaic period, so that developments between the archaic and the classical period may be explained in terms of genre rather than in diachronic terms. In the archaic period, the number of data points is very limited. For the classical period the data for the animals are also rather limited, so that the transition between the Hellenistic and Roman period especially lends itself for a study of diachronic developments.
The main evolution that can be traced with respect to the animal collectives (cf. Fig. 6) is the prominence of πλήθος in the Hellenistic period, which clearly decreases in the Roman period. There is no clear challenger; rather, there seems to be a diversification in general, with, for example, a more frequent use of ἄγελη, φῦλον and ζεῦγος (even though πλήθος and ζεῦγος are likely not simply interchangeable). In a few cases we can also observe a tendency to specialization: it is especially in the Roman period that words for insects are associated with specific collective nouns, namely νέφος: ἤπατος; σμῆνος, whereas in the Hellenistic period πλήθος is still predominating here (see Table 3).
Table 3: Collective nouns used for insects in the Hellenistic and Roman period
| Collective | Hellenistic | Roman |
|------------|-------------|-------|
| πλήθος | 9 | 8 |
| ἄγελη | 0 | 1 |
| φῦλον | 0 | 3 |
| νομή | 0 | 1 |
| σμῆνος | 3 | 15 |
| ἤπατος | 1 | 10 |
| νέφος | 3 | 4 |
| χορός | 0 | 2 |
Figure 6: Evolution in the presence of animal collective nouns over time
Figure 7: Evolution in the presence of human collective nouns over time
For the human collective nouns the diachronic changes are less clear (cf. Fig. 7). A number of archaic collective nouns are used much less frequently in later periods. A clear example is ὄμηλος, which in later periods is mainly taken up by a few authors (especially Philo Judaeus). Another example is φῦλον. Here again we observe the prominence of πλῆθος in the Hellenistic period, but the decline in the Roman period is less pronounced. What is particularly striking is the diachronic increase of θύννος as a collective noun for soldiers in the Roman period (12 instances of θύννος and 58 instances of πλῆθος in the Hellenistic period versus 52 instances of θύννος and 53 instances of πλῆθος in the Roman period). In addition, there is a clear increase of χορός among certain ‘professions’ (2 instances of χορός and 19 instances of πλῆθος in the Hellenistic period versus 23 instances of χορός and 51 instances of πλῆθος in the Roman period). Inspecting the data, this is especially true when the profession has a ‘didactic’ or ‘heralding’ function, e.g. philosophers, teachers and prophets.
5. Conclusions and outlook
The syntactic/morphological-based extraction and clustering techniques have allowed us to detect a large number of collective nouns. Nevertheless, there are some important caveats. The quantitative methods used have enabled us to compile a longlist. A final manual selection, reducing the longlist to a shortlist, nevertheless remained necessary. This step involves a large degree of subjective decisions, many of which can be debated. In addition, we cannot evaluate which relevant words were not found (‘recall’). Furthermore, polysemy causes any clustering technique to be problematic. The multidimensional nature of semantics implies that Ancient Greek equivalents for polysemous and idiosyncratic collocations (such as e.g. English ‘murder of crows’) will be difficult to identify. Some ‘collective nouns’ can also be frequently combined with inanimate entities (e.g. πλῆθος χρημάτων “a group or amount of money”). While these examples were filtered out during the animacy detection described in section 3.2 (i.e. we only included words with a sufficient number of animate genitive attributes in the cluster analysis in section 3.3, and similarly only analyzed words with such attributes during the corpus analysis described in 4), these contexts with inanimate entities were still included in the word vectors, and therefore might distort the results of the cluster analysis. In the future, word vectors modelling the meaning of a word in context rather than the general meaning of a word might allow for a higher degree of precision. The results could also be improved by means of an objective set of criteria whether or not a word can be considered a collective noun. Another difficulty resides in the data scarcity, which makes it very difficult to make statements about the significance of the connection between certain collective nouns and specific animals. By way of example, we see that for θύννος (‘tuna’), attested thrice in the data, three different collective nouns are used: besides the generic πλῆθος, στρατός and θύννος occur. Table 1 shows that both στρατός and θύννος tend to be used as collective nouns of humans (especially in a military context; see 4.4). The question here is whether we are dealing with a fixed, conventional collective noun for tuna or a context-related metaphor. Obviously, close reading of the relevant passages may shed more light on the matter. For this particular case, it seems to be an occasional metaphor twice. However, if there would have been more data, it could be determined with more certainty to what extent the use of στρατός and θύννος is rooted in context or convention. The same applies to many other lemmas, so that it is very difficult to make firm statements about which combinations were idiomatically acceptable in Greek.
There are also alternative methods possible for onomasiological queries, including searching for English translations of the concept in question through lexica (e.g. the English-Greek dictionary by Woodhouse 1987, or reverse-searching the LSJ lexicon by Jones et al. 1996) or through parallel translations, as well as using Ancient Greek WordNets – a first Ancient Greek WordNet was created by Bizzoni et al. (2014), while recently a new attempt has been undertaken by Biagetti et al. (2021). Although we could not systematically compare these approaches to the one adopted in this paper due to time and space constraints, we will briefly address the advantages and disadvantages of both through a quick exploration. Searching the Woodhouse and LSJ lexica for words such as ‘lock’, ‘herd’, ‘crowd’ and ‘group’ returned many words listed in Table 1, but also missed some (e.g. neither lexicon included νέφος under an English lemma referring to a collective noun, for example, and Woodhouse expectedly does not contain Homeric words such as πῶον or post-classical words such as χρημάτων as it is limited to the Classical Attic dialect). On the other hand, they also include words missed by our computational approach, especially low-frequent ones that we filtered out in an initial step (see 3.2), e.g. κόμος (only 3 occurrences with an animate genitive noun). Additionally, they also reveal some alternative constructions to express a group of living beings rather than the noun + genitive construction, e.g. adjective + noun constructions such as μελισσαίος υλόματος (see section 3.2) or δρακονθόμιλος συνοικία “a swarm of dragons” (Woodhouse). However, a big limitation of this approach is that it simply shifts the burden of determining which on words or constructions can express a particular concept from one language (Ancient Greek) to another one (e.g. English). For instance, the word ἄθροισμα is defined, among other definitions, as ‘aggregate’ in the LSJ lexicon. While ‘aggregate’ is certainly a collective noun in English, one must take this English term into account as one of the many possibilities to express collective nouns in order to retrieve ἄθροισμα with a lexical-based method. While the Ancient Greek WordNets seem to be less vulnerable in this respect, as they encode semantic relations between words in the target language – in this case Greek – the WordNet designed by Bizzoni et al. (2014) was in fact based on automatic linking between Greek-English lexica and therefore prone to similar problems, while the Biagetti et al. (2021) WordNet is still in active development. All these human-curated resources are also highly dependent on human judgments and the data they have considered during their developments, while the automatic approach discussed in this paper can easily take the whole Greek corpus into account (although it is fair to say that the quality of the semantic methods is highly dependent on the frequency of specific genres in the input data, see also Perrone et al. 2019).
For this first exploration of onomasiologically searching, we have deliberately chosen a case with identifiable syntactic characteristics. The challenge for future research
consists in choosing less straightforward cases, where syntactic and morphological encoding is significantly less decisive. Without doubt, one of the greatest onomasiological challenges is to trace in the Ancient Greek corpus concepts that may be present but for which lexicalized words are missing (possible examples include modern concepts such as ‘queer’, ‘fashion’, etc.).
6. Acknowledgements
This research was made possible by FWO grant 3H200733: “Language and Ideas: Towards a New Computational and Corpus-Based Approach to Ancient Greek Semantics and the History of Ideas”. We would like to thank three anonymous reviewers for their stimulating criticisms.
7. Bibliographical References
Aikhenvald, A.Y. (2000). Classifiers: a Typology of Noun Categorization Devices. Oxford: Oxford University Press.
Atallah, C., Bras, M., & Vieu, L. (2018). Exploring a Corpus Annotated in Causal Discourse Relations for the Study of Causal Lexical Clues. In Final Action Conference TextLink 2018, Mar 2018, Toulouse, France. Retrieved from https://hal.archives-ouvertes.fr/hal-02982984.
Benvenuto, M.C. (2013). Genitive. In Encyclopedia of Ancient Greek Language and Linguistics. Leiden: Brill. Retrieved from <http://referenceworks.brillonline.com/entries/encyclopedia-of-ancient-greek-language-and-linguistics/genitive-COM_00000140>.
Biagetti, E., Zanchi, C., & Short, W.M. (2021). Toward the Creation of WordNets for Ancient Indo-European Languages. In Proceedings of the 11th Global WordNet Conference. University of South Africa (UNISA): Global Wordnet Association, pp. 258-266.
Biber, D., Conrad, S., & Leech, C. (2003). Longman Student Grammar of Spoken and Written English. Harlow: Longman.
Birkens, M.B., & Sommer, F. (2014). The agreement of collective nouns in the history of Ancient Greek and German. In C. Gianollo, A. Jäger & D. Penka (Eds.), Language Change at the Syntax-Semantics Interface. Berlin: De Gruyter, pp. 183-221.
Bizzoni, Y., Boschetti, F., Diakoff, H., Del Gratta, R., Monachini, M., & Crane, G. (2014). The Making of Ancient Greek WordNet. In Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC’14). Reykjavik: European Language Resources Association (ELRA), pp. 1140-1147.
Boschetti, F. (2010). A Corpus-based Approach to Philological Issues. Unpublished PhD Thesis. University of Trento.
Cananau, I. (2019). Toward a Comparatist Horizon in Conceptual History. History of European Ideas, 45(1), pp. 117-120. https://doi.org/10.1080/01916599.2018.1493307
Celano, G.G. (2019). The Dependency Treebanks for Ancient Greek and Latin. In M. Berti (Ed.), Digital Classical Philology. Berlin & Boston: Walter de Gruyter, pp. 279-298.
Celano, G.G. A., & Crane, G. (2015). Semantic Role Annotation in the Ancient Greek Dependency Treebank. In M. Dickinson, E. Hinrichs, A. Patjuk, & A. Przepiórkowski (Eds.), Proceedings of the Fourteenth International Workshop on Treebanks and Linguistic Theories (TLT14). Warsaw, pp. 26-34.
Dedé, F. (2012). Some Remarks on the Metalinguistic Usage of the Term ‘Collective’. In Proceedings of the First Workshop on the Metalanguage of Linguistics. Models and Applications. University of Udine – Lignano. March 2-3, 2012. Roma: Il calamo, pp. 81-94.
Deroy, L. (1956). La valeur du suffixe préhellénique -nth d’après quelques noms grecs en -νθός. Glotta, 35(3/4), pp. 171-195.
Dieckhoff, A. (2019). Peuples et populisme. Annuaire international de justice constitutionnelle, 34, pp. 691-698. https://doi.org/10.3406/ajic.2019.2719
Erk, K. (2012). Vector Space Models of Word Meaning and Phrase Meaning: A survey. Language and Linguistics Compass, 6(10), pp. 635-653.
Fernández-Dominguez, J. (2019). The Onomasiological Approach. In Oxford Research Encyclopedia of Linguistics. Oxford: Oxford University Press. https://doi.org/10.1093/acrefore/9780199384655.013.5
Geevarghese, D. (2009). Theories of Lexical Semantics. Oxford: Oxford University Press.
Goossens, D. (2013). Assessing Corpus Search Methods in Onomasiological Investigations. In H. Hasselgård, J. Ebeling, S. Oksefjell Ebeling (Eds.), Corpus Perspectives on Patterns of Lexis. Amsterdam & Philadelphia: John Benjamins, pp. 271-292.
Gries, S. Th. 2010. Useful Statistics for Corpus Linguistics. In A Mosaic of Corpus Linguistics: Selected Approaches, In A. Sánchez & M. Almela (Eds.), Studien zur romanischen Sprachwissenschaft und interkulturellen Kommunikation. Frankfurt am Main: Peter Lang, pp. 269-291.
Grzega, J. (2002). Some Aspects of Modern Diachronic Onomasiology. Linguistics, 40(5), pp. 1021-1045. https://doi.org/10.1515/ling.2002.035
Johnston, I. (2019). A List of Homeric (Epic) Similes from the Iliad and Odyssey. Retrieved from http://johnstoniatexts.x10host.com/homer/homersimiles.html.
Jones, H.S., Henry George Liddell, MacKenzie, R., Scott, R., & Thompson, A.A. (1996). A Greek-English Lexicon (New ed. with new supplement). Oxford: Clarendon.
Jooksten, F., Sutter, G.D., Driehe, G., Grondelaers, S., Hartsuiker, R.J., & Speelman, D. (2007). Dutch Collective Nouns and Conceptual Profiling. Linguistics, 45(1), pp. 85-132.
Jussén, B., & Rohmann, G. (2015). Historical Semantics in Medieval Studies: New Means and Approaches. Contributions to the History of Concepts, 10(2), pp. 1-6. https://doi.org/10.3167/choc.2015.100201
Kaczynska, E. (2019). Lakoian βοῶν ‘Band of Boys’ as a collective noun. Graeco-Latina Brunensia, (1), pp. 93-103. https://doi.org/10.5817/GLB2019-1-7
Keersmaekers, A. (2020). A Computational Approach to the Greek Papyri: Developing a Corpus to Study
Variation and Change in the Post-Classical Greek Complementation System. Unpublished PhD Dissertation. KU Leuven. Retrieved from https://lirias.kuleuven.be/retrieve/590983.
Keersmaekers, A. (2021). The GLAUX corpus: methodological issues in designing a long-term, diverse, multi-layered corpus of Ancient Greek. In Proceedings of the 2nd International Workshop on Computational Approaches to Historical Language Change 2021, pp. 39-50.
Keersmaekers, A., Mercelis, W., Swaeleens, C., & Van Hal, T. (2019). Creating, Enriching and Valorizing Treebanks of Ancient Greek. In Proceedings of the 18th International Workshop on Treebanks and Linguistic Theories (TLT, SyntaxFest 2019). Association for Computational Linguistics (ACL), pp. 109-117.
Keersmaekers, A., & Van Hal, T. (2021). A Corpus-Based Approach to Conceptual History of Ancient Greek. In G. Kristiansen, K. Franco, S. De Pascale, L. Rosseel, & W. Zhang (Eds.), Cognitive Sociolinguistics Revisited. Berlin & Boston: Walter de Gruyter, pp. 213-225.
Kipfer, B.A. (1986). Investigating an Onomasiological Approach to Conceptual History of Ancient Greek: A study in World Loanword Lexicography. In Proceedings of the 1st International Workshop on Computational Approaches to Historical Language Change. Florence: Association for Computational Linguistics, pp. 56-66.
Kutuzov, A. (2020). Distributional Word Embeddings in Modeling Diachronic Semantic Change. Unpublished PhD-dissertation. Osaka University.
Kühner, R., & Gerth, B. (1966). Ausführliche Grammatik der griechischen Sprache. München: Hueber.
Kühtje, J., & van der Maaten, L. (2018). Rtsne: T-Distributed Stochastic Neighbor Embedding using a Barnes-Hut Implementation (Version 0.15). Retrieved from https://CRAN.R-project.org/package=Rtsne.
LeDell, E., Gill, N., Aiello, S., Fu, A., Candel, A., Click, C., … H2O.ai. (2022). h2o: R Interface for the ‘H2O’ Scalable Machine Learning Platform (Version 3.36.0.4). Retrieved from https://CRAN.R-project.org/package=h2o.
Lehrer, A. (1986). English Classifier Constructions. Lingua, 68(2-3), pp. 109-148.
Lenci, A. (2018). Distributional models of word meaning. Annual Review of Linguistics, 4, pp. 151-171.
Maechler, M., Rousseuw, P., Struyf, A., Hubert, M., Hornik, K., Studer, M., ... Murphy, K. (2022). cluster: ‘Finding Groups in Data’: Cluster Analysis Extended Rousseeuw et al. (Version 2.1.3). Retrieved from https://CRAN.R-project.org/package=cluster.
McGillivray, B. (2020). Computational methods for semantic analysis of historical texts. In K. Schuster & S. Dunn (Eds.), Routledge International Handbook of Research Methods in Digital Humanities. London: Routledge, pp. 261-274.
Mehl, S. (2016). Corpus Onomasiology: A study in World Englishes. Unpublished PhD-dissertation: UCL London.
Meyer, P., & Tu, N.D.T. (2021). A Word Embedding Approach to Onomasiological Search in Multilingual Loanword Lexicography. In Proceedings of eLex 2021, pp. 78-91.
Moerdijk, F., Tiberius, C., & Niestadt, J. (2008). Accessing the ANW Dictionary. In Coling 2008: Proceedings of the Workshop on Cognitive Aspects of the Lexicon (COGALEX 2008). Manchester: Coling 2008 Organizing Committee, pp. 18-24. Retrieved from https://aclanthology.org/W08-1903.
Müller, E., & Schmieder, F. (2016). Begriffsgeschichte und historische Semantik: ein kritisches Kompendium. Frankfurt am Main: Suhrkamp.
Perrone, V., Palma, M., Hengchen, S., Vatri, A., Smith, J.Q., & McGillivray, B. (2019). GASC: Genre-Aware Semantic Change for Ancient Greek. In Proceedings of the 1st International Workshop on Computational Approaches to Historical Language Change. Florence: Association for Computational Linguistics, pp. 56-66.
Perrone, V., Hengchen, S., Palma, M., Vatri, A., Smith, J.Q., & McGillivray, B. (2021). Lexical Semantic Change for Ancient Greek and Latin. In N. Tahmasebi, L. Borin, A. Jatowt, Y. Xu, & S. Hengchen (Eds.), Computational Approaches to Semantic Change. Berlin: Language Science Press, pp. 287–310.
Rhodes, C. (2014). An Unkindness of Ravens: A Book of Collective Nouns. London: Michael O’Mara Books.
Rodia, M.A., Probert, P., & McGillivray, B. (2019). Vector space models of Ancient Greek word meaning, and a case study on Homer. TAL Traitement Automatique Des Langues, 60(3), pp. 63-87.
Rodda, M.A., Senaldi, M.S., & Lenci, A. (2016). Panta Rei: Tracking Semantic Change with Distributional Semantics in Ancient Greek. In CLiC-It/EVALITA.
Sierra, G. (2008). Natural Language Searching in Onomasiological Dictionaries. In Coling 2008: Proceedings of the Workshop on Cognitive Aspects of the Lexicon (COGALEX 2008). Manchester: Coling 2008 Organizing Committee, 32-38. Retrieved from https://aclanthology.org/W08-1905.
Swiggers, P., & Wouters, A. (1998). De Tékhnē grammatikē van Dionysius Thrax: de oudste spraakkwast in het Westen. Leuven: Peeters.
Vatri, A., & McGillivray, B. (2018). The Diorisis Ancient Greek Corpus. Research Data Journal for the Humanities and Social Sciences, 3(1), pp. 55-65.
Viteau, J. (1896). Étude sur le Grec du Nouveau Testament: comparé avec celui des septante. Sujet, complément et attribut. Paris: E. Bouillon.
Woodhouse, S.C. (1987). English-Greek dictionary: a vocabulary of the Attic language (Repr.). London: Routledge and Kegan Paul.
Zwarts, J. (2020). Contiguity and membership and the typology of collective nouns. In Proceedings of Sinn Und Bedeutung, 24(2). Osnabrück: Osnabrück University, pp. 539-554. | 2025-03-04T00:00:00 | olmocr | {
"pdf_page_numbers": [
[
0,
5470,
1
],
[
5470,
10430,
2
],
[
10430,
16909,
3
],
[
16909,
17037,
4
],
[
17037,
21132,
5
],
[
21132,
21322,
6
],
[
21322,
26930,
7
],
[
26930,
27076,
8
],
[
27076,
34181,
9
],
[
34181,
39755,
10
],
[
39755,
45184,
11
]
]
} |
17263ead80c1c02096c7ced217a6469f5edf529d | {
"olmocr-version": "0.1.59",
"pdf-total-pages": 14,
"total-fallback-pages": 0,
"total-input-tokens": 50485,
"total-output-tokens": 19273,
"fieldofstudy": [
"Biology",
"Agricultural And Food Sciences"
]
} | Bioactivity of essential oils for the management of *Tetranychus urticae* Koch and selectivity on its natural enemy *Neoseiulus californicus* (McGregor): A promising combination for agroecological systems
Mauricéa Fidelis de Santana\textsuperscript{a}, Cláudio A. G. Câmara\textsuperscript{a,b}, Vaneska Barbosa Monteiro\textsuperscript{a}, João Paulo Ramos de Melo\textsuperscript{a}, Marcilio Martins de Moraes\textsuperscript{b}
\textsuperscript{a} Laboratório de Investigação Química de Inseticidas Naturais, Departamento de Agronomia, Universidade Federal Rural de Pernambuco, Av. Dom Manoel de Medeiros, s/n, 52171-900 Recife, PE, Brazil.
\textsuperscript{b} Laboratório de Produtos Naturais Bioativos, Departamento de Química, Universidade Federal Rural de Pernambuco. Av Dom Manoel de Medeiros, s/n, 52171-900, Recife, PE, Brazil.
Original research
**ABSTRACT**
The two-spotted spider mite, *Tetranychus urticae*, causes damage to crops grown in northeast Brazil. The adoption of biological control methods and curative methods (plant-based insecticides) is an essential practice for pest management in agroecological systems. Therefore, the aim of the present study was to investigate the chemical properties, toxicity, and ovicidal activity of essential oils (EOs) from *Lippia sidoides*, *Croton rhamnifolioides*, *Croton grewioides*, *Citrus sinensis*, *Citrus limon*, *Citrus aurantiifolia* and *Piper divaricatum* for the control of *T. urticae* and determine the selectivity of these EOs regarding the predator mite *Neoseiulus californicus*. The chemical analysis (gas chromatography–mass spectrometry) of the EOs enabled the identification of 98 compounds. The major constituents were carvacrol (*L. sidoides*), β-caryophyllene (*C. rhamnifolioides*), (E)-anethole (*C. grewioides*), limonene (*Citrus* spp.), safrole and methyl eugenol (*P. divaricatum*). All oils exhibited satisfactory toxicity to the eggs and females of *T. urticae* and were even more toxic than the commercial product Azamax. The *L. sidoides* oil exhibited greater toxicity compared to the other oils, with LC\textsubscript{50} values of 0.05 and 0.09 µL mL\textsuperscript{-1} for females and eggs, respectively. All oils tested were selective to *N. californicus*, with RS values ranging from 3.61 to 23.28 for *C. aurantiifolia* and *C. grewioides*, respectively. Therefore, the use of products based on the EOs studied in combination with the natural enemy *N. californicus* is a viable option in agroecological systems for the management of *T. urticae*.
**Keywords** two-spotted spider mite; plant-based acaricide; *Neoseiulus californicus*; selectivity; agroecological systems
**Introduction**
Brazilian agriculture suffers frequent losses due to the attack of pests. The two-spotted spider mite, *Tetranychus urticae* Koch, causes damage to diverse crops grown in the state of Pernambuco, such as beans, cotton, papaya, grapes and ornamental plants (Ferreira et al. 2015; Monteiro et al. 2015), the latter of which is often grown in protected environments.
The losses caused by agricultural pests are both direct (effects on the crop) and indirect (costs related to the purchasing of pesticides and the consequent environmental contamination and harm to health) (Oliveira et al. 2014). Moreover, the indiscriminate use of these products inevitably leads to resistant pest populations. Indeed, *T. urticae* is the agricultural pest with resistance to the largest number of conventional acaricides (526 cases of resistance to 96 different active ingredients) (APRD 2020).
The main form of controlling the two-spotted spider mite is through conventional acaricides (van Leeuwen et al. 2010; Rincón et al. 2019). However, these products are not permitted in agroecological communities or organic farming activities. Azamax (active ingredient: azadirachtin) is the only plant-based acaricide registered used in agroecological systems in protected environments in the state of Pernambuco. A preventive and curative option for the management of this pest is through biological control and the use of plant-based insecticides (Brzozowski and Mazourek 2018). In Brazil, the predator mite *Neoseiulus californicus* (McGregor) is used for the biological control of *T. urticae*, especially in protected environments (Barbosa et al. 2017).
The use of formulations whose active ingredient is derived from plants, such as essential oils (EOs), has been widely investigated due to the broad action on different types of arthropods as well as biodegradability, low toxicity to mammals and the absence of contamination of the environment (Isman 2020). Moreover, these oils are complex mixtures generally made up of terpenes and phenylpropanoids, which makes the development of resistance in the target pest a much slower process, as demonstrated by Feng and Isman (1995) for the green peach aphid, *Myzus persicae* Sulz., as a mixture of active constituents, including neem, mitigated the development of resistance in comparison to a single active ingredient (azadirachtin). Although there are no reports of the resistance of *T. urticae* to azadirachtin (APRD 2020), the frequent use of this active ingredient in agroecological communities of northeast Brazil could favor the resistance of this pest.
Among EOs with recognized acaricidal properties, species belonging to the genus *Lippia* (*L. sidoides*), *Croton* (*C. rhamnifolioides*) and *Citrus* (*C. aurantiifolia*, *C. limon* and *C. sinensis*) stand out (Júnior et al. 2010; Cavalcanti et al. 2010; Camara et al. 2017; Ribeiro et al. 2019). However, there are few reports on the selectivity of these EOs for the predator mite *N. californicus*.
In the search for plant-based substances for use as active ingredients in acaricidal formulations, the aim of the present study was to determine the chemical composition of EOs from the leaves of *Lippia sidoides*, *Piper divaricatum*, *Citrus sinensis*, *C. limon*, *C. aurantiifolia*, *Croton rhamnifolioides* and *C. greuwioide* and evaluate toxicity to the eggs and adults of *T. urticae*. A further aim was to investigate the effects of these oils on the predator mite *N. californicus*. The results were compared to those achieved with a plant-based acaricide (Azamax) used as the positive control.
**Material and methods**
**Collection of plant material**
The plants collected were identified by Botanist Dra. Maria F.A. Lucena. Voucher of samples were mounted and deposited no Herbário da Universidade Federal de Pernambuco, under number: (46254) *Croton rhamnifolioides* Pax and Hoffm. (Euphorbiaceae), (42193) *Croton greuwioide* Baill (Euphorbiaceae), (48734) *Citrus aurantiifolia* (Christm.) Swingle (Rutaceae), (48736) *Citrus limon* (L.) Burm.f. (Rutaceae) and (48739) *Citrus sinensis* Osbeck var. mimo (Rutaceae). *Lippia sidoides* Cham (Verbenaceae) (genotype LISID4) and *Piper divaricatum* (Piperaceae) (Kato-1063) oils were donated by Prof. Alves, PB from Federal University of Sergipe and Prof. Ramos, CS from Chemistry Departament of UFRPE, respectively.
Chemicals
All monoterpenes (α-pinene, β-pinene, α-phellandrene, limonene, 1,8-cineole, p-cymene, citronellal, camphor, terpinen-4-ol, terpinolene, linalool e α-terpinol), sesquiterpenes (β-caryophyllene, aromadendrene, α-humulene, germacrene D, bicyclogermacrene, spathulenol and caryophyllene oxide) and phenylpropanoid ((Z)-anethole, eugenol, methyl eugenol, safrole) used for chemical constituent identification were purchased from Sigma-Aldrich - Brazil.
Essential oils extraction and GC-FID analysis
The EOs from fresh leaves (100 g) of *C. rhamnifolioides, C. grewioides, C. aurantiifolia, C. limon, C. sinensis* were separately isolated using a modified Clevenger-type apparatus and hydrodistillation for 2h. The oil layers were separated and dried over anhydrous sodium sulfate, stored in hermetically sealed glass containers and kept at low temperature (-5 °C) until analysis. Total oil yields were expressed as percentages (g/100 g of fresh plant material). All experiments were carried out in triplicate. Quantitative GC (500 GC, PerkinElmer Clarus, Shelton, CO, USA) analysis were carried out using a apparatus equipped with a flame ionization detector (FID) and a non-polar DB-5 fused silica capillary column (30 m x 0.25 mm x 0.25 μm) (J & W Scientific). The oven temperature was programmed from 60 to 240 °C at a rate 3 °C min⁻¹. Injector and detector temperatures were 260 °C. Hydrogen was used as the carrier gas at a flow rate of 1 mL min⁻¹ in split mode (1:30). The injection volume was 0.5 µL of diluted solution (1/100) of oil in n-hexane. The amount of each compound was calculated from GC-FID peak areas in the order of DB-5 column elution and expressed as a relative percentage of the total area of the chromatograms. Analyses were carried out in triplicate.
GC-MS analysis
The qualitative Gas Chromatography-Mass Spectrometry (GC-MS) (220-MS IT GC, Varian, Walnut Creek, CA, USA) analysis were carried out using a system with a mass selective detector, mass spectrometer in EI 70 eV with a scan interval of 0.5 s and fragments from 40 to 550 Da. fitted with the same column and temperature program as that for the GC-FID experiments, with the following parameters: carrier gas = helium; flow rate = 1 mL min⁻¹; split mode (1:30); injected volume = 1 µL of diluted solution (1/100) of oil in n-hexane.
Identification of components
Identification of the components was based on GC-MS retention indices with reference to a homologous series of C8-C40 n-alkanes calculated using the Van der Dool and Kratz equation (Van den Dool and Kratz 1963) and by computer matching against the mass spectral library of the GC-MS data system (NIST 11 and WILEY 11th) and co-injection with authentic standards as well as other published mass spectra (Adams 2017). Area percentages were obtained from the GC-FID response without the use of an internal standard or correction factors.
Rearing of *Tetranychus urticae* and *Neoseiulus californicus*
Specimens of *T. urticae* were originally collected in 2008 from grapevine (*Vitis vinifera* L.) in the municipality of Petrolina in the state of Pernambuco, Brazil (09°12’43.9″ S, 40°29’12.7″ W) and then maintained in the laboratory on jack bean (*Canavalia ensiformes* L.) at 25 ± 1 °C, 65 ± 5% relative humidity and a 12-h photoperiod without any exposure to acaricides. The predator mite *N. californicus* was collected from the municipality of Bonito in the state of Pernambuco, Brazil (08°28’13” S, 35°43’43” W) on chrysanthemum (*Dendranthema grandiflora* Tzvelev.) and bred in the laboratory since 2010 with no exposure to acaricides. The breeding method of *T. urticae* and *N. californicus* was according to methodology used by Born *et al.* (2018). The predator mite was reared in plastic arenas (25 cm diameter) maintained in B.O.D. at a mean temperature of 27 °C and a 12-h photoperiod. Jack bean leaf was placed with the margin
surrounded by moistened hydrophilic cotton to avoid the escape of the mites. Cotton fibers were placed on the jack bean leaves to stimulate oviposition. As a food source, *T. urticae* and castor bean pollen (*Ricinus communis* L.) were offered every 2 days.
**Residual contact assay**
The leaf disc painting method described by Araújo *et al.* (2020) was used to test the action of *C. aurantiifolia, C. limon, C. sinensis* var. mimo, *L. sidoides, C. rhamnifolioides, P. divaricatum, C. grewioides* and positive control (Azamax) by contact toxicity. The experiments were performed with open Petri dishes (10 cm diameter). Leaf discs (5 cm diameter) were cut from leaves of greenhouse-grown jack bean (*C. ensiformis*). Test solutions were prepared by diluting the EO in water and DMSO (Dimethylsulfoxide) (0.5%) (negative control). The concentration used in the bioassays ranged from 0.009 to 5.40 μL mL⁻¹ for the EO. The concentration of the botanical and conventional insecticides used as positive control ranged from 0.009 to 10 μL mL⁻¹ for Azamax. Leaf discs (5 cm diameter) were immersed in solutions for 30s. Control mites were held on leaf discs immersed in the water and DMSO. Each leaf disc was infested with 15 adult females of *T. urticae*. Five replicates were used in each bioassay and repeated 2× on different dates using a completely randomized design, totaling 150 mites per concentration. Mortality was determined under a dissecting microscope 24 h after the onset of treatment. Mites were considered dead if the appendages did not move when prodded with a fine paintbrush. The residual contact assays were performed at 25 ± 1 °C, 65 ± 5% RH and a 12-h photoperiod. Fifty adult females of *T. urticae* were placed on leaf discs (8 cm diameter) for 24 hours to effect oviposition. After that period, *T. urticae* were removed. The leaf discs with *T. urticae* eggs were immersed in the concentrations of EO, Azamax and control (water and DMSO) (adapted from Esteves-filho *et al.* 2013). Subsequently were placed to dry for 30 minutes at room temperature. Each leaf disc 300 eggs were left, which served as contaminated food for *N. californicus*. Each leaf disc was infested with 15 adult females of *N. californicus*. Five replicates were used in each bioassay and repeated 2× on different dates using a completely randomized design, totaling 150 mites per concentration. Mortality was determined under a dissecting microscope 48 h after the onset of treatment. Mites were considered dead if the appendages did not move when prodded with a fine paintbrush. The residual contact assays were performed at 25 ± 1 °C, 65 ± 5% RH and a 12-h photoperiod.
**Ovicide assay**
The methodology used in this test was adapted from Esteves-Filho *et al.* (2013). Leaf discs (5 cm diameter) were cut from leaves of greenhouse-grown jack bean (*C. ensiformis*). Leaf discs were infested with 15 adult females of *T. urticae*, which were maintained for 24 hours for oviposition. Then leaf discs with eggs of *T. urticae* were immersed in the concentration of each oil, azamax and control, as bioassays described above. Subsequently were placed to dry for 30 minutes at room temperature. Each leaf disc 50 eggs were left. Each bioassay and repeated 3× on different dates using a completely randomized design, totaling 150 mites per concentration. Evaluation was performed after 96 hours of application of oil, azamax and control, which is recorded the number of emerged larvae.
**Statistical analysis**
For the determination of the lethal concentration necessary for a 50% mortality rate (LC₅₀) and 90% (LC₉₀) of the mite population in the residual contact tests, the mortality data were analyzed using the Probit model implemented in the POLO-Plus 2.0 (LeOra Software 2005) program, with the calculation of 95% confidence levels. Toxicity ratios (TR) and RS (Relative Selectivity) were determined based on the method described by Robertson and Preisler (2017).
Results
Yield and chemical profile of essential oils
The GC-MS of the Croton spp., Lippia sidoides, Piper divaricatum and Citrus spp. oils enabled the identification of 98 compounds (Table 1). The greatest yield of EO was achieved with C. grewioides (2.30 ± 0.18%), followed by C. sinensis (0.78 ± 0.05%), C. limon (0.46 ± 0.06%), C. rhamnifolioides (0.17 ± 0.03%) and C. aurantiifolia (0.17 ± 0.05%).
The C. rhamnifolioides and C. grewioides oils had a predominance of compounds belonging to the classes of sesquiterpenes (66.3 ± 0.6%) and phenylpropanoids (75.7 ± 0.5%), respectively. β-Caryophyllene (33.3 ± 0.6%) was the major component of the C. rhamnifolioides oil and (E)-anethole (55.5 ± 0.4%) was the major component of the C. grewioides oil. The predominance of compounds belonging to the chemical class of phenylpropanoids (84.6 ± 0.5%) was also found in the P. divaricatum oil, the major constituents of which were safrole (49.3 ± 0.5%) and methyl eugenol (31.0 ± 0.2%).
The L. sidoides had a predominance of monoterpenes (92.9 ± 1.0%), with carvacrol (59.5 ± 1.0%) as the major component. An abundance of monoterpenes was found in the Citrus oils, with limonene identified as the major component in the C. limon (68.2 ± 0.5%), C. aurantiifolia (57.7 ± 0.9%) and C. sinensis (90.1 ± 1.1%) oils.
Residual contact and ovicidal assay
The relative toxicities of the oils to adult females and the eggs of the two-spotted spider mite and its natural enemy, N. californicus, are displayed in Table 2. Toxicity varied with the type of oil, development stage of the pest and species (pest and natural enemy).
Adult females were more susceptible to the oils than the eggs. For a better classification of toxicity to the adult females of T. urticae based on the LC₅₀ estimates for the oils, the relative toxicities were divided into three groups ranging from most toxic (Group 1) to least toxic (Group 3). Group 1 comprised only the L. sidoides oil. Group 2 was formed by the P. divaricatum, C. limon, C. rhamnifolioides and C. grewioides oils. Group 3 was formed by the C. aurantiifolia and C. sinensis oils. Regarding relative toxicity to the eggs, two groups were formed: Group 1 comprised only the L. sidoides oil and Group 2 was composed of the C. grewioides, C. rhamnifolioides, C. sinensis, C. limon, C. aurantiifolia and P. divaricatum oils.
Comparing the relative toxicities of the substances tested to the two forms of development of the pest, all oils were more efficient than the positive control (Azamax). The L. sidoides oil stood out in this comparison, which was 9.6-fold and 3.4-fold more toxic to the females and eggs of T. urticae, respectively.
Based on the LC₅₀ estimates and respective confidence intervals, the essential oils were divided into three groups from the most toxic to the least toxic to N. californicus. Group 1 was composed of the L. sidoides and C. aurantiifolia oils. Group 2 was composed of the C. rhamnifolioides oil and Group 3 was composed of the C. grewioides, C. sinensis, C. limon and P. divaricatum oils.
Comparing the toxicity of the oils between species, the oils were more toxic to the pest than the predator, as demonstrated by the LC₅₀ estimates, which were higher for N. californicus. Based on the relative selectivity (RS) calculated for the oils investigated (Table 2), most oils were more selective than the plant-based acaricide (Azamax). The only exception was the C. aurantiifolia oil, which had the same RS as Azamax.
Discussion
Yields and chemical profile of essential oils
The yields of the essential oils from the species analyzed are compatible with those described in previous studies on C. rhamnifolioides (Camara et al. 2017), C. grewioides (Silva et al. 2008),
| Table 1 Chemical composition (% ± DP) of essential oils from leaves of Lippia, Piper and Croton, and peels of Citrus species. |
|---------------------------------------------------------------|
| **Compound** | **IR** | **IR** | **Croton humifusa** | **Croton guayacan** | **Lippia alba** | **Piper cubeba** | **Citrus limon** | **Citrus aurantium** | **Citrus sinensis** | **Method of identification** |
| α-Thujene | 921 | 924 | 1.2±0.0 | - | - | - | 0.4±0.0 | 0.8±0.0 | 2.0±0.0 | RI, MS |
| α-Pinene | 928 | 932 | 1.3±0.0 | 0.4±0.0 | - | 2.9±0.0 | 3.0±0.2 | 1.3±0.1 | RI, MS, CI |
| α-Fenchene | 948 | 945 | 0.2±0.0 | - | - | - | - | - | - | RI, MS |
| Camphene | 949 | 946 | 6.4±0.0 | - | - | - | - | - | - | RI, MS |
| Sabinene | 966 | 969 | - | - | - | 0.4±0.0 | 1.0±0.0 | - | RI, MS |
| β-Pinene | 971 | 974 | 0.8±0.0 | 1.9±0.1 | - | - | - | - | 1.9±0.1 | RI, MS, CI |
| Myrcene | 988 | 988 | - | - | - | - | 4.5±0.1 | 7.8±0.7 | - | RI, MS |
| α-Phellandrene | 1004 | 1002 | 1.5±0.1 | - | - | - | - | - | - | RI, MS, CI |
| β-Phellandrene | 1025 | 1025 | 0.1±0.0 | - | - | - | - | - | 0.1±0.0 | RI, MS |
| Sylvestrene | 1025 | 1025 | - | - | - | 0.4±0.0 | - | - | - | RI, MS |
| 1,8-Cineole | 1030 | 1026 | 10.5±0.6 | 1.1±0.1 | - | - | - | - | 1.1±0.1 | RI, MS, CI |
| (Z)-β-Ocimene | 1031 | 1032 | - | - | - | - | 7.5±0.4 | 15.5±0.3 | 0.3±0.0 | RI, MS |
| (E)-β-Ocimene | 1044 | 1044 | - | - | - | 8.4±0.0 | - | - | - | RI, MS |
| γ-Terpineol | 1055 | 1054 | 1.5±0.1 | 6.1±0.2 | - | 1.0±0.1 | 0.9±0.0 | 0.4±0.0 | RI, MS |
| Dihydro myrcenol | 1072 | 1069 | 3.0±0.2 | - | - | - | - | - | - | RI, MS |
| m-Cymene | 1085 | 1082 | 0.2±0.0 | - | - | - | - | - | - | RI, MS |
| Terpinolene | 1088 | 1086 | - | - | 0.4±0.0 | - | - | - | 0.4±0.0 | RI, MS, CI |
| p-Cymene | 1092 | 1089 | 0.2±0.0 | - | - | - | - | - | - | RI, MS |
| Linalool | 1092 | 1095 | 0.4±0.0 | 0.2±0.0 | - | - | 1.2±0.1 | - | 0.2±0.0 | RI, MS, CI |
| cis-β-Terpinol | 1139 | 1140 | - | - | - | - | - | - | 0.4±0.0 | RI, MS |
| Camphor | 1140 | 1141 | - | 0.8±0.1 | - | - | - | - | - | RI, MS, CI |
| Citronellal | 1145 | 1148 | - | - | - | - | 1.6±0.1 | - | 0.1±0.0 | RI, MS, CI |
| Myrcene | 1141 | 1145 | 0.3±0.0 | - | - | - | - | - | - | RI, MS |
| δ-Terpineol | 1162 | 1162 | 0.1±0.0 | - | - | - | 0.8±0.0 | - | - | RI, MS |
| Borneol | 1170 | 1165 | 0.1±0.0 | - | - | - | - | - | - | RI, MS |
| Terpinen-4-ol | 1174 | 1174 | 1.2±0.0 | - | 1.6±0.1 | - | - | - | - | RI, MS, CI |
| (E)-Isocitral | 1175 | 1177 | - | - | - | - | - | - | 0.2±0.0 | RI, MS |
| α-Terpineol | 1192 | 1186 | 0.3±0.0 | 0.5±0.0 | - | - | - | - | - | RI, MS, CI |
| Methyl chavicol | 1196 | 1195 | - | 1.9±0.1 | - | - | - | - | - | RI, MS |
| γ-Terpineol | 1202 | 1199 | 0.7±0.0 | - | - | - | - | - | - | RI, MS |
| p-Anisaldehyde | 1250 | 1247 | - | 0.5±0.0 | - | - | - | - | - | RI, MS |
| (Z)-Anethole | 1251 | 1249 | - | 4.6±0.1 | - | - | - | - | - | RI, MS, CI |
| (E)-Anethole | 1280 | 1282 | - | 55.5±0.4 | - | - | - | - | - | RI, MS, CI |
| Safrole | 1285 | 1285 | - | - | - | 49.3±0.5 | - | - | - | RI, MS, CI |
| Thymol | 1289 | 1289 | - | - | 11.7±0.4 | - | - | - | - | RI, MS, CI |
| Bornyl acetate | 1290 | 1287 | 0.6±0.0 | - | - | - | - | - | - | RI, MS |
| Carvacrol | 1299 | 1298 | - | - | 59.5±1.0 | - | - | - | - | RI, MS, CI |
| δ-Elemene | 1331 | 1335 | 0.5±0.0 | - | - | - | - | - | - | RI, MS |
| α-Cubeene | 1342 | 1345 | 0.1±0.0 | - | - | - | - | - | - | RI, MS |
| Eugenol | 1356 | 1356 | - | - | 3.1±0.1 | - | - | - | - | RI, MS, CI |
| α-Copaene | 1369 | 1374 | 0.2±0.0 | 2.1±0.1 | - | - | - | - | - | RI, MS |
| β-Cubeene | 1387 | 1387 | 0.8±0.0 | - | 0.6±0.0 | - | - | - | - | RI, MS |
| δ-Elemene | 1389 | 1389 | 0.3±0.0 | 1.0±0.0 | - | - | - | - | - | RI, MS |
| β-Longifolene | 1398 | 1400 | 0.7±0.0 | - | - | - | - | - | - | RI, MS |
| Methyl eugenol | 1401 | 1403 | - | 10.6±0.3 | - | 31.0±0.2 | - | - | - | RI, MS, CI |
| Cyclohexene | 1406 | 1406 | - | - | 0.2±0.0 | - | - | - | - | RI, MS |
| β-Caryophyllene | 1415 | 1417 | 33.3±0.6 | 4.5±0.1 | 2.0±0.1 | 0.4±0.0 | 2.0±0.0 | 1.4±0.0 | 0.1±0.0 | RI, MS, CI |
Table 1 Continued.
| Compound | RIa | RIb | Croton rhamnifolios | Citrus amarantifolios | Citrus limon | Citrus sinensis | Method of identification |
|---------------------------------|-----|-----|---------------------|-----------------------|--------------|-----------------|--------------------------|
| **β-Copaene** | 1433| 1430| 0.1±0.0 | - | - | - | RI, MS |
| *trans*-α-Bergamotene | 1435| 1432| - | 0.3±0.0 | 0.4±0.0 | 1.1±0.1 | RI, MS |
| **β-Humulene** | 1439| 1436| 0.5±0.0 | - | - | - | RI, MS |
| 6,9-Guaiaadiene | 1443| 1442| 0.5±0.0 | - | - | - | RI, MS |
| **cis**-Prenyl-limonene | 1446| 1443| - | - | 0.3±0.0 | - | RI, MS |
| *(Z)*-Methyl isoeugenol | 1451| 1451| - | 2.9±0.1 | - | - | RI, MS |
| **α-Humulene** | 1453| 1452| 0.8±0.0 | - | - | - | RI, MS, CI |
| 9-*epi-**(E)*-Caryophyllene | 1467| 1464| 5.1±0.2 | - | - | - | RI, MS |
| γ-Gurjunene | 1474| 1475| - | - | - | 2.9±0.1 | RI, MS |
| Amorpha-5(7)-dione | 1474| 1479| 0.2±0.0 | - | - | - | RI, MS |
| γ-Murolene | 1480| 1478| 0.1±0.0 | - | - | - | RI, MS |
| γ-Himachalene | 1481| 1481| - | - | 1.2±0.0 | - | RI, MS |
| Germacrene D | 1484| 1484| - | 0.4±0.0 | - | - | RI, MS, CI |
| **β-Selinene** | 1489| 1489| - | - | 1.1±0.0 | - | RI, MS |
| *(E)*-Methyl isoeugenol | 1490| 1491| - | 6.7±0.1 | - | - | RI, MS |
| δ-Selinene | 1495| 1492| 0.5±0.0 | - | 2.0±0.1 | - | RI, MS |
| *trans*-Murola-4(14),5-diene | 1497| 1493| 0.3±0.0 | - | - | - | RI, MS |
| Bicyclogermacrene | 1500| 1502| 0.9±0.0 | - | - | - | RI, MS, CI |
| *(Z)*-α-bisabolene | 1507| 1506| - | - | 0.3±0.0 | - | RI, MS |
| Germacrene A | 1512| 1508| 0.2±0.0 | - | - | - | RI, MS |
| δ-Amorphone | 1514| 1511| 0.1±0.0 | - | - | - | RI, MS |
| γ-Cadinene | 1517| 1513| 0.1±0.0 | - | - | - | RI, MS |
| 7-*epi*-α-Selinene | 1520| 1520| - | - | 0.3±0.0 | - | RI, MS |
| δ-Cadinene | 1521| 1522| - | 1.3±0.0 | - | 7.8±0.1 | RI, MS |
| 10-*epi*-Cubebol | 1535| 1533| - | - | 1.2±0.1 | - | RI, MS |
| **α-Cadinene** | 1540| 1537| 1.5±0.1 | - | - | - | RI, MS |
| **α-Copaen-11-ol** | 1543| 1539| 0.1±0.0 | - | - | - | RI, MS |
| **α-Calacorene** | 1544| 1544| 0.2±0.0 | - | - | - | RI, MS |
| Elemol | 1549| 1548| - | - | - | 2.6±0.0 | RI, MS |
| Germacrene B | 1556| 1559| 1.0±0.1 | - | - | - | RI, MS |
| **β-Calacorene** | 1562| 1564| 0.2±0.0 | - | - | - | RI, MS |
| *(Z)*-Isoeugenol acetate | 1566| 1566| - | - | 1.2±0.0 | - | RI, MS |
| Maalol | 1566| 1566| - | - | 1.4±0.1 | - | RI, MS |
| **α-Cedrene epoxide** | 1569| 1574| 0.1±0.0 | - | - | - | RI, MS |
| Spathulenol | 1572| 1577| 5.9±0.1 | 1.6±0.0 | - | - | RI, MS, CI |
| Caryophyllene oxide | 1580| 1582| 5.8±0.6 | 2.8±0.1 | - | - | RI, MS, CI |
| **cis**-β-Elemeneone | 1594| 1589| 0.4±0.0 | - | - | - | RI, MS |
| Viridiflorol | 1596| 1592| 1.6±0.1 | - | - | - | RI, MS |
| Ledol | 1606| 1602| 0.5±0.0 | - | - | - | RI, MS |
| Humulene epoxide II | 1613| 1608| 1.3±0.0 | - | - | - | RI, MS |
| *epi*-α-Cadinol | 1639| 1638| 0.1±0.0 | - | - | - | RI, MS |
| Himesol | 1643| 1640| 1.3±0.1 | - | - | - | RI, MS |
| **α-Murolol** | 1645| 1644| 0.5±0.0 | - | - | - | RI, MS |
| Cubenol | 1645| 1645| 0.1±0.0 | 0.5±0.0 | - | - | RI, MS |
| **α-Eudesmol** | 1656| 1652| 0.2±0.0 | - | - | - | RI, MS |
| 14-hydroxy-(Z)-caryophyllene | 1668| 1666| 0.9±0.0 | - | - | - | RI, MS |
| **β-Bisabololenal** | 1765| 1768| - | - | - | 1.9±0.0 | RI, MS |
| Total | 79.4±0.8| 99.8±0.5| 98.6±1.1| 98.5±0.5| 97.1±0.6| 98.3±0.9| 97.5±1.1| RI, MS |
| Monoterpenes | 31.2±0.7| 9.6±0.1| 92.9±1.0| 0.8±0.0| 91.2±0.5| 88.6±0.9| 96.4±1.1| RI, MS |
| Sesquiterpenes | 66.3±0.6| 14.5±0.0| 5.8±0.1| 13.3±0.1| 5.9±0.1| 9.7±0.0| 1.1±0.0| RI, MS |
| Phenylpropanoids | 75.7±0.5| - | 84.6±0.5 | - | - | - | - | RI, MS |
C. aurantifolia, C. limon (Ribeiro et al. 2019) and C. sinensis (Júnior et al. 2010) collected in different localities in the state of Pernambuco, Brazil.
The chemical profiles determined for the oils from the species of Croton, Lippia, Piper and Citrus are in agreement with data previously reported for these species and/or their congeners. For example, β-caryophyllene and (E)-anethole, which were respectively the major compounds identified in the C. rhamnifolioides and C. grewioioides oils, were also the main constituents of the oils from these same species collected in Pernambuco (Camara et al., 2017; Silva et al., 2008). Carvacrol (59.5 ± 1.0%), which was the major constituent of the L. sidoides oil in the present study, was also found to be the major component in the leaf oil of this species collected in different localities in Brazil in the states of Minas Gerais, Ceará and Pernambuco (Cavalcanti et al. 2010; Guimarães et al. 2015). The phenylpropanoids safrole and methyl eugenol found to be the major constituents of the P. divaricatum oil were also reported for this species in different localities of Brazil and the world (Barbosa et al. 2012; Souto et al. 2012; de Oliveira et al. 2019; Vilhena et al. 2019). Limonene was the major constituent of the Citrus oils, with proportions ranging from 57.7 ± 0.9% to 90.1 ± 1.1%, which is compatible with data reported in previous studies of these species collected in the state of Alagoas, Brazil (Júnior et al. 2010; Ribeiro et al. 2020).
Residual contact and ovicidal assay
The use of EOs combined with biological control is an ecologically and agronomically compatible practice to control pest populations, leaving the use of synthetic acaricides as the last option (Barzman et al. 2015; Pretty et al. 2018). For pests with a history of resistance to synthetic products, such as T. urticae, the use of EOs is an excellent alternative, as the complex mixture of monoterpenes, sesquiterpenes and phenylpropanoids, which affect different sites in the pest, favors the slower development of resistance (Koul and Walia 2009).
The EOs tested in the present study exhibited greater toxicity to T. urticae than the positive control (Azamax [active ingredient: azadirachtin]). Although there is no evidence of the resistance of T. urticae to azadirachtin, the high LC$_{50}$ of this positive control demonstrates its lower effectiveness regarding the mortality of females and lower ovicidal effect compared to all oils tested. Azadirachtin is the only chemical insecticide/acaricide registered for organic agriculture.
| Treatments | Stage | N* | $\chi^2$ (df) | Slope ± SE | LC$_{50}$ (95% CI) | N* | $\chi^2$ (df) | Slope ± SE | LC$_{50}$ (95% CI) | RS* |
|------------|------|----|-------------|------------|-------------------|----|-------------|------------|-------------------|-----|
| Lippia sidoides | Adults | 1350 | 12.80 (6) | 0.88±0.05 | 0.05 (0.03 – 0.07) | 1500 | 12.11 (8) | 0.94±0.05 | 0.78 (0.65 – 0.93) | 15.60 (10.93 – 21.62) |
| Croton grewioioides | Adults | 1350 | 6.62 (5) | - | - | - | - | - | - | - |
| Croton rhamnifolioides | Adults | 1200 | 2.26 (1.70 – 3.30) | 0.15 (0.13 – 0.18) | 1650 | 2.29 (5) | 0.63±0.07 | 3.80 (2.52 – 6.79) | 13.57 (8.17 – 22.90) |
| Citrus sinensis | Adults | 1200 | 2.31 (10) | 0.96±0.06 | 0.15 (0.12 – 0.18) | - | - | - | - |
| Citrus limon | Adults | 1200 | 1.17±0.06 | 0.15 (0.13 – 0.18) | 1050 | 6.62 (5) | 1.17±0.07 | 1.14 (0.95 – 1.37) | 9.50 (7.31 – 12.13) |
| Citrus aurantifolia | Adults | 1200 | 2.81±0.06 | 0.13 (0.10 – 0.15) | 1350 | 11.76 (7) | 1.37±0.09 | 2.26 (1.70 – 3.30) | 17.38 (13.39 – 22.84) |
| Piper divaricatum | Adults | 1050 | 2.79 (5) | 1.10±0.06 | 0.11 (0.09 – 0.15) | 1200 | 3.94 (6) | 0.91±0.07 | 1.79 (1.40 – 2.44) | 16.27 (11.19 – 22.21) |
| Azamax | Adults | 1650 | 2.08 (8) | 0.99±0.04 | 0.48 (0.38 – 0.63) | 1350 | 3.49 (7) | 0.76±0.05 | 2.03 (1.59 – 2.66) | 4.22 (3.05 – 5.64) |
Fidelis de Santana M. et al. (2021), *Acarologia* 61(3): 564-576; DOI 571
farming in Brazil (Agrofit 2020). However, the product is expensive for small farmers, demonstrating the need for economically viable alternatives for producers.
The EOs tested herein were extracted from cultivated plants as well as some native to the Atlantic Forest and Caatinga biomes of Brazil and are easily found in agricultural niches distributed throughout the northeast region of the country. Among these oils, L. sidoides had the greatest yield (4.80 ± 0.23%) as well as the greatest ovicidal action and toxicity by residual contact to T. urticae females.
The genus Lippia is recognized for its acaricidal properties by both fumigation and residual contact (Santos et al. 2019; Tabari et al. 2020). The residual toxicity for L. sidoides found in the present study is compatible with that described by Cavalcanti et al. (2010) for L. sidoides collected in the state of Sergipe, Brazil. The authors also demonstrated this oil has a fumigant effect. Born et al. (2018) recently reported that the oil from the leaves of Lippia gracilis Schauer collected in the state of Pernambuco, which had the same major component at that identified in the L. sidoides oil (carvacrol), exhibited fumigant and residual contact action (LC50 = 29.70 μL mL−1) against T. urticae. However, the residual toxicity found for the L. sidoides oil analyzed in the present investigation was 594 times greater than that of the L. gracilis oil reported by Born et al. (2018). These results suggest that the major component is not always the active ingredient of the oil and that other factors should be taken into consideration, such as qualitative and quantitative aspects and multiple (synergistic, additive and/or antagonistic) interactions that can be established among the chemical constituents of an essential oil (Moraes et al. 2012; Neves and Camara 2016).
A previous investigation of the biological properties of EOs from species of the genus Piper revealed action against several types of arthropods, including mites of importance to veterinary medicine – Rhipicephalus (Boophilus) microplus (Vinturelle et al. 2017) – and agriculture – Dolichocybe indica Mahunka (Pummuang and Insung 2016) and T. urticae (Ribeiro et al. 2016; Araújo et al. 2020). However, studies addressing the effect of the oil from P. divaricatum on arthropods are restricted to the investigation of the insecticidal potential against stored grain pests – Tribolium castaneum Herbst (Jaramillo-Colorado et al. 2015) – and general pests – Solenopsis saevissima (Smith) (Souto et al. 2012).
Based on the LC50 estimates, the P. divaricatum oil was 53 times more toxic by residual contact than the oil from the leaves of Piper aduncum L. (Araújo et al. 2020) to T. urticae adults. The differences in toxicity may be explained by qualitative and quantitative differences in the chemical composition of these Piper oils.
Ferraz et al. (2010) reported the acaricidal action of oils from the leaves of Piper mikanianum (Kunth) Steud. and Piper xylostaeoides on Rhipicephalus microplus larvae. Comparing these results to those obtained in the present investigation, the P. divaricatum oil was 21 and 56 times more toxic to T. urticae than the Piper oils tested on ticks. Besides differences in the chemical profiles of the oils tested, the greater activity found for the P. divaricatum oil may be attributed to morphological differences among mites/ticks.
Citrus is a widely studied genus due to its toxic (Dutra et al. 2016; Papanastasiou et al. 2017; Farias et al. 2020) and repellent (Camara et al. 2015; Ribeiro et al. 2019) activity against arthropods. The acaricidal activity of Citrus against T. urticae has previously been demonstrated by residual toxicity, fumigation and repellent action (Júnior et al. 2010; Ribeiro et al. 2019). Regarding residual toxicity, the present study reports much lower LC50 values for C. limon (0.13 μL mL−1) and C. aurantiifolia (0.21 μL mL−1) than those reported by Ribeiro et al. (2019), which were 25.18 μL mL−1 and 106.14 μL mL−1, respectively. This divergence may be explained by variations in populations of T. urticae, methodological differences and the percentages of different chemical constituents found in the Citrus oils. For instance, the major component (limonene) was identified in higher proportions in the present study (C. limon: 68.2% and 40.7% in the present investigation and the study by Ribeiro et al. 2019, respectively; C. aurantiifolia: 57.7% and 37.7% in the present investigation and the study by Ribeiro et al. 2019 , respectively).
A previous study on the potential of EOs from species of the genus *Croton* revealed that these oils are promising due to their activities against stored grain pests (Silva et al. 2008; Santos et al. 2019; Ribeiro et al. 2020), pests of interest to human medicine (Carvalho et al. 2016) and synanthropic pests (Brito et al. 2020). Recently, EOs from four *Croton* species (*C. pulegiodorus, C. conduplicatus, C. grevioides* and *C. blanchetianus*) were found to be promising in the control of a tick of interest to veterinary medicine (*Rhipicephalus microplus*) (Castro et al. 2019; Rodrigues et al. 2020). Despite reports that *Croton* oils can cause toxicity to the red spider mite by contact, fumigation and repellence (Neves and Câmara 2011; Camara et al. 2017), to the best of our knowledge, no previous studies have evaluated the acaricidal action of the oil from *C. grevioides* against *T. urticae*.
Comparing the LC$_{50}$ estimates for the *C. grevioides* and *C. rhamnifolioides* oils to those from species of *Croton* reported in the literature regarding toxicity to *T. urticae* by contact, the oils investigated herein were 20 times more toxic than the oil from *C. rhamnifolioides* collected in the municipality of Buique, Pernambuco (Camara et al. 2017).
Investigations of substances derived from plants for the control of *T. urticae* are generally directed at assessing the toxicity of EOs to larvae and/or adults. With the exception of the *C. aurantiifolia* oil, which exhibited ovicidal action by fumigation (Pavela et al. 2016), none of the oils analyzed in the present study has previously been investigated with regards to its ovicidal potential against *T. urticae*.
While no significant differences in the susceptibilities of the eggs and females were found among the *C. grevioides, C. rhamnifolioides, C. aurantiifolia* and *P. divaricatum* oils, the *L. sidoides* and *C. limon* oils were more toxic to the adult females and the *C. sinensis* oil was more toxic to the eggs. These results may be explained by several factors: a) the nature of the EOs (qualitative, quantitative and physicochemical aspects); b) the inherent susceptibility of the forms of development investigated (egg and adult); and c) the method used for the evaluation of the oils (direct contact for the eggs and residual contact for the females).
Although there are no records of ovicidal action by direct contact of the oils tested on *T. urticae*, Lima et al. (2013) reported the toxicity of a commercially acquired *L. sidoides* oil (thymol chemotype) to the eggs of *Aedes aegypti*. The ovicidal action found in the present investigation indicates that the *L. sidoides* oil tested (carvacrol chemotype) was 737 times more toxic than the commercial *L. sidoides* oil. This greater toxicity may be explained by qualitative differences between the oils as well as morphological differences between the eggs of the two target species.
Acaricides that are selective for natural enemies are highly advantageous to integrated pest management programs. Selectivity is defined as the capacity of a product to control the target pest while exerting the lowest possible impact on beneficial organisms, such as predators, parasitoids and pollinizers (Ripper et al. 1951). This selectivity is one of the requirements for natural acaricides to be considered economically viable (Vieira et al. 2007). While few studies have investigated the selectivity of EOs for predator mites, the literature offers promising results for the oils of *P. aduncum, Melaleuca leucadendra* L., *Schinus terebinthifolius* Raddi (Araújo et al. 2020) and *L. gracilis* (Born et al. 2018), which were more selective than the oils tested in the present investigation.
The lower selectivity of the oils in comparison to data reported in the literature may be explained by the method employed in the experiments to assess toxicity to the predator mite. In the present study, we offered leaf disks and eggs of *T. urticae* coated with the oils, whereas Araújo et al. (2020) and Born et al. (2018) only used leaf disks. Thus, there was both a residual effect and toxic effect by ingestion in the present study, causing greater toxicity to the predator. Nonetheless, based on the calculation of relative selectivity (RS), the oils investigated herein can be considered selective for *N. californicus* (Table 2).
The present results show that *L. sidoides* is the most promising among all oils tested for the management of *T. urticae*, as it exhibited the greatest toxicity to the pest and was also selective for *N. californicus*. Due to its abundance and availability, *L. sidoides* can be a viable option for the preparation of a plant-based insecticide for the management the red spider mite.
in agroecological systems in the state of Pernambuco. However, further studies are needed, such as field bioassays, for the cost-benefit analysis of a formulation based on essential oils.
**Acknowledgements**
The authors are grateful to the Conselho Nacional de Desenvolvimento Científico e Tecnológico (PQ-2 302735/2019-4), Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (Nº 88882.436540/2019-01; 88887.474239/2020-00) and Fundação de Amparo à Ciência e Tecnologia do Estado de Pernambuco (FACEPE Nº IBPG-0584-1.06/20; IBPG-0715-1.06/18; APQ-0398-1.06/19; APQ-0476 -1.06/14) for scholarship and funding this study.
**References**
Agrofit 2020. Sistema de agrotóxicos Fitossanitario do Ministério da Agricultura, Pecuária e Abastecimento. [Internet]. [5 october 2020]. http://agrofit.agricultura.gov.br/agrofit_cons/principal_agrofit_cons. APRID: Arthropod Pesticide Resistance Database [Internet]. [02 August 2020]. Available from: http://www.pesticideresistance.org.
Araújo M.J.C., Camara C.A.G., Born F.S., Moraes M.M. 2020. Acaricidal activity of binary blends of essential oils and selected constituents against Tetranychus urticae in laboratory/greenhouse experiments and the impact on Neoseiulus californicus. Exp. Appl. Acarol., 80: 423-440. doi:10.1007/s10493-020-00464-8
Barbosa M.F.C., Demite P.R., Moraes G.J., Poletti M. 2017. Controle Biológico com ácaros predadores e seu papel no manejo integrado de pragas [Internet]. Promip. Available from: http://www.promip.agr.br.
Barbosa Q.P., Câmara C.A.G., Ramos C.S., Nascimento D.C., Lima-Filho J.V., Guimarães E.F. 2012. Chemical composition, circadian rhythm and antibacterial activity of essential oils of Piper divaricatum: a new source of safrrole. Quim. Nova., 35: 1806-1808. doi:10.1590/S0100-404220120009000019
Barzman M., Barberi P., Birch A.N.E., Boonekamp P., Dachbrodt-Saaydeh S., Graf B., Hommel B., Jensen J.E., Kiss J., Kudsk P., Lamichhane J.R., Messean A., Moenon A.C., Ratnadass A., Ricci P., Sarah J.L., Sattin M. 2015. Eight principles of integrated pest management. Agron. Sustain. Dev., 35, 1199-1215. doi:10.1007/s13593-015-0327-9
Born F.S., Camara C.A.G., Melo J.O., Moraes M.M. 2018. Acaricidal property of the essential oil from Lippia graulis against Tetranychus urticae and a natural enemy, Neoseiulus californicus, under greenhouse conditions. Exp. Appl. Acarol., 75: 491-502. doi:10.1007/s10493-018-0296-3
Brito F.A., Bacci L., Santan A.S., Silva J.E., Nizio D.A.C., Nogueira P.C.L., Arrigoni-Blank M.F., Melo C.R., Melo J.O., Blank A.F. 2020. Toxicity and behavioral alterations caused by essential oils of Croton tetradeniuss and their major compounds on Acromyrmex balzani. Crop Prot., 137: 105259. doi:10.1016/j.cropro.2020.105259
Brzozowski L., Mazourek M. 2018. A sustainable agricultural future relies on the transition to organic agroecological pest management. Sustainability, 10: 2023. doi:10.3390/su10102023
Camara C.A., Moraes M.M., Melo J.P., Silva M.M. 2017. Chemical Composition and acaricidal activity of essential oils from Croton rhamnifoliioides Pax and Hofm. in different regions of a caatinga biome in northeastern Brazil. J. Essent. Oil-Bear. Plants, 20: 1434-1449. doi:10.1007/s00436-017-0370-1
Camara C.A.G., Akhtar Y., Isman M.B., Seffrin R.C., Born F.S. 2015. Repellent activity of essential oils and selected compounds against Citrus species from two species of Tetranychus urticae in the laboratory and greenhouse. Crop Prot., 74: 110-115. doi:10.1016/j.cropro.2015.04.014
Carvalho K.S., Silva S.L.C., Souza I.A., Gualberto S.A., Cruz R.C.D. 2016. Toxicological evaluation of essential oil from the leaves of Croton tetradenius (Euphorbiaceae) on Aedes aegypti and Mus musculus. Parasitol. Res., 115: 3441-3448. doi:10.1007/s00436-016-5106-2
Castro K.N.C., Chagas A.C.S., Costa-júnior L.M., Camuto K.M., Brito E.S., Rodrigues T.H.S., Andrade I.M. 2019. Acaricidal potential of volatile oils from Croton species on Rhicephalus microplus. Rev. Bras. Farmacogn., 29: 811-815. doi:10.1590/1984-8285-RBF2019.09.001
Cavalcani S.C.H., Niculae E.U., Blank A.F., Camara C.A.G., Araujo L.N., Alves P.B. 2010. Composition and acaricidal activity of Lippia sidoides essential oil against two-spotted spider mite (Tetranychus urticae Koch). Bioresour. Technol., 101: 829-832. doi:10.1016/j.biortech.2009.08.053
Dutra K.A., Oliveira J.V., Navarro D.M.A.F., Barbosa D.R., Silva J.P.O.S. 2016. Control of Callosobruchus maculatus (FABR.) (Coleoptera: Chrysomelidae: Bruchinae) in Vigna unguiculata (L.) WALP. with essential oils from four species of Croton spp. plants. J. Stored Prod. Res., 68: 25-32. doi:10.1016/j.jspr.2016.04.001
Esteves-Filho A.B., Oliveira J.V., Torres J. B., Matos C.H.C. 2013. Toxicidade de espíromosfeno e acaricidas naturais para Tetranychus urticae Koch e compatibilidade com Phytoseiulus macropilus (Banks). Semina: Ciênc. Agrar., 34: 2675-2686. doi:10.5433/1679-0399.2013v34n4a2675
Farias A.P., Santos M.C., Jumbo L.O.V., Oliveira E.E., Nogueira P.C.L., Filho I.G.S., Teodoro A.V. 2020. Croton essential oils control the cassava green mite, Mononychellus tanajoa, and induce higher predatory responses by the lacewing Ceraeochrysa caligata. Ind. Crop. Prod., 145: 112151. doi:10.1016/j.indcrop.2020.112151
Feng R., Isman M.B. 1995. Selection for resistance to azadirachtin in the green peach aphid, Myzus persicae. Experientia, 51: 831-833. doi:10.1007/BF01922438
Ferraz A.B.F., Balbino J.M., Zini C.A., Ribeiro V.L.S., Bordignon S.A.L., von Poser G. 2010. Acaricidal activity and chemical composition of the essential oil three Piper species. Parasitol. Res., 107: 243-248. doi:10.1007/s00436-010-1878-y
Ferreira C.B.S., Andrade F.H.N., Rodrigues A.R.S., Siqueira H.A.A., Gondim M.G.C. 2015. Resistance in field populations of Tetranychus urticae to acaricides and characterization of the inheritance of abamectin resistance. Crop Prot., 67: 77-83. doi:10.1016/j.cropro.2014.09.022
Guimarães L.G.D.L., da Silva, M.L.M., Reis, P.C.J., Costa, M.T.R., Alves, L.L. 2015. General characteristics, phytochemistry and pharmacognosy of Lippia sidoides. Nat. Prod. Commun., 10: 1861-1867. doi:10.1177/1934578X1501001116
Isman M.B. 2020. Botanical insecticides in the twenty-first century—fulfilling their promise? Annu. Rev. Entomol., 65: 233-249. doi:10.1146/annurev-ento-011019-025010
Jaramillo-Colorado B.E., Duarte-Restrepo E., Pinto-Benitez N. 2015. Evaluación de la actividad replenta de aceites esenciales de plantas Piperaceae del departamento de Chocó, Colombia. Rev. Toxicol., 32: 112-116.
Júnior C.P.A., Camara C.A.G., Neves I.A., Ribeiro N.C., Botelho P.S. Robertson J.L., Preisler H.K. 2017. Pesticide bioassays with arthropods. California: CRC Press., pp. 212.
Ripper W.E., Greenslade R.M., Hartley G.S. 1951. Selective insecticides and biological control. J. Econ. Entomol., 4: 272. doi:10.1093/jen/44.4.448
Robertson J.L., Preiser H.K. 2017. Pesticide bioassays with arthropods. California: CRC Press., pp. 212.
Rodrigues O.G., Falcão B.R.M., Barbosa B.C., Pereira A.V., Aquino V.V.F. 2020. *In vitro* biological activity of the *Croton blanchetianus* (Baill) essential oil against *Rhipicephalus* (Boophilus) *microplus* (Acari: Ixodidae). *J. Appl. Biol. Biotechnol.*, 7: 55-58. doi:10.7324/JABB.2019.70210
Santos M.C., Teodoro A.V., Menezes M.S., Pinto-Zevallos D.M., Arrigoni-Blank M.F., Oliveira E.M.C., Sampaio T.S., Farias A.P., Coelho C.R., Blank A.F. 2019. Bioactivity of essential oil from *Lippia gracilis* Schauer against two major coconut pest mites and toxicity to a non-target predator. *Crop. Prot.*, 125: 104913. doi:10.1016/j.cropro.2019.104913
Santos P.E.M., Silva A.B., Lira C.R.I.M., Matos C.H.C., Oliveira C.R.F. 2019. Contact toxicity of essential oil of *Croton pulegiodorus* Baill on *Sitophilus zeamais* Motschulsky. *Rev. Caatinga*, 32: 329-335. doi:10.1590/1983-21252019v32n205rc
Silva C.G.V., Zago H.B., Júnior H.J.G.S., Camara C.A.G., Oliveira J.V., Barros R., Schwartz M.O.E., Lucena M.F.A., 2008. Composition and Insecticidal Activity of the Essential Oil of *Croton gresuioides* Baill. against Mexican Bean Weevil (*Zabrotes subfasciatus* Boheman). *J. Essent. Oil Res.*, 20: 179-182. doi:10.1080/10412905.2008.9699985
Souto R.N.P., Harada A.Y., Andrade E.H.A., Maia J.G.S. 2012. Insecticidal activity of *Piper* essential oils from the Amazon against the fire ant *Solenopsis saevissima* (Smith) (Hymenoptera: Formicidae) Neotrop. *Entomol.*, 41: 510-517. doi:10.1007/s10377-012-0880-8
Tabari M.A., Rostami A., Khodashenas A., Maggi F., Petrelli R., Giordani C., Taponjdu L.A., Papa F., Zuo Y., Cianfaglione K., Youssefi M.R. 2020. Acaricidal activity, mode of action, and persistent efficacy of selected essential oils on the poultry red mite (*Dermanyssus gallinae*). *Food Chem. Toxicol.*, 138: 111207. doi:10.1016/j.fct.2020.111207
van Leeuwen T., Vontas J., Tsagkarakou A., Dermauw W., Tirry L. 2010. Acaricide resistance mechanisms in the two-spotted spider mite *Tetranychus urticae* and other important *Acaria*: a review. *Insect Biochem. Mol. Biol.*, 40: 563e572. doi:10.1016/j.ibmb.2010.05.008
Vieira P.C., Mafezoli J., Biavatti M.W. 2007. Insecticides of origem vegetal. In: Correa A.G., Vieira P.C. (Eds). *Produtos Naturais no Controle de Insetos*, São Carlos: Edufscar. p. 69-104.
Vilhena C.S., Nascimento L.A.S., Aguiar E.H.A., Silva J.K.D.R., Hamoy M., Torres M.F., Barba L.A.L. 2019. Essential oil of *Piper divaricatum* induces a general anaesthesia-like state and loss of skeletal muscle tonus in juvenile tambaqui, *Colossoma macropomum*. *Aquaculture*, 510: 169-175. doi:10.1016/j.aquaculture.2019.05.057
Vinturelle R., Mattos C., Meloni J., Nogueira J., Nunes M.J., Vaz Jr I.S., Rocha L., Lione V., Castro H.C., Chagas E.F. 2017. *In Vitro* Evaluation of Essential Oils Derived from *Piper nigrum* (Piperaceae) and *Citrus limonum* (Rutaceae) against the Tick *Rhipicephalus* (Boophilus) *microplus* (Acari: Ixodidae). *Biochemistry Research International* Volume 2017, Article ID 5342947, pp. 9. doi:10.1155/2017/5342947 | 2025-03-06T00:00:00 | olmocr | {
"pdf_page_numbers": [
[
0,
0,
1
],
[
0,
3065,
2
],
[
3065,
7081,
3
],
[
7081,
10985,
4
],
[
10985,
14954,
5
],
[
14954,
18696,
6
],
[
18696,
22795,
7
],
[
22795,
30786,
8
],
[
30786,
34797,
9
],
[
34797,
39355,
10
],
[
39355,
44123,
11
],
[
44123,
49551,
12
],
[
49551,
51114,
13
],
[
51114,
54189,
14
]
]
} |
68f1c6c0eb21799bad6c7f886cdd945a62eb2c4d | {
"olmocr-version": "0.1.59",
"pdf-total-pages": 3,
"total-fallback-pages": 0,
"total-input-tokens": 8111,
"total-output-tokens": 2875,
"fieldofstudy": [
"Medicine",
"Political Science"
]
} | E-Cigarette as a Harm Reduction Approach among Tobacco Smoking Khat Chewers: A Promising Bullet of Multiple Gains
Saba Kassim 1,* and Konstantinos E. Farsalinos 2
1 College of Dentistry, Taibah University, Al-Madinah Al-Munawwarah 43353, Saudi Arabia
2 Onassis Cardiac Surgery Center, Sygrou 356, Kallithea 17654, Greece; [email protected]
* Correspondence: [email protected]; Tel.: +966-053-555-8878
Academic Editor: Paul B. Tchounwou
Received: 4 January 2016; Accepted: 15 February 2016; Published: 19 February 2016
Abstract: Khat chewing/use, a green leaf with amphetamine-like effects is socially integrated in the Middle East and Africa. Khat chewing is often associated with tobacco smoking and occurs in closed places, such as a family home setting where the smoke-free laws cannot be implemented. Tobacco cigarette smoking among khat chewers is a significant concern, but there is also second-hand exposure to smoke at home or in places where khat users gather. Evidence suggests that e-cigarettes represent a significantly less harmful form of nicotine intake. Evaluating the effects of e-cigarettes among khat chewers could be important in understanding the impact of e-cigarettes as a harm reduction approach, with the potential to reduce the health risk associated with smoking.
Keywords: khat; tobacco smoking; e-cigarettes; nicotine
1. Introduction
The chewing/use of the amphetamine–like khat leaf from Catha edulis is widespread and socially ingrained in the Middle East and East Africa and their diaspora communities world-wide. Khat chewing has become a national and international public health concern [1] and has recently gained popularity globally [2]. Khat is very often used with tobacco smoking [3]. Khat chewers are either daily tobacco smokers or smoke tobacco when chewing khat [3]. The use of tobacco by khat chewers represents a clear public health concern [4]. There is a significant prevalence of tobacco smoking among adult and school children khat chewers [4] with gender difference in mode of tobacco use: women often smoke waterpipe tobacco whereas men often smoke tobacco cigarettes [5]. An alarming elevation in the level of tobacco dependence associated with higher khat dependence, and very high levels of carbon monoxide among smoking khat chewers have been observed [3]. The estimated of tobacco use among khat users versus smoker none khat users was found higher [4] although the difference in smoking dependence between khat users and non-users has not been established.
The reason that khat users smoke tobacco is mainly to enhance the stimulant effects of khat [3], indicating some interaction between nicotine/tobacco and the chemicals present in khat that awaits identification [3]. The session of khat-use often occurs in closed places, such as a family home setting or meeting places for immigrants [6,7] where the smoke-free laws cannot be implemented. Thus, besides the risks associated with smoking among khat chewers, there is also second and third-hand (environmental) exposure to the toxicants present in cigarette/waterpipe smoke.
Khat chewing is an integral part of different social aspects, and attempts to control its use failed in khat-originating countries [8]. However, efforts should be made to reduce the adverse effects of smoking in khat chewers. Although many khat users reported they had thought about and had
attempted to quit khat and tobacco use [5], there is no evidence on the effects of using approved smoking cessation methods in this population. Moreover, access to and willingness to visit smoking cessation clinics is expected to be limited, considering that most khat users are immigrants or living in low income countries (e.g., Yemen) where implementation of tobacco control and availability of tobacco cessation services are scarce. One of the methods of promoting smoking reduction or cessation would be by switching from inhalation of combustible products to a non-combustible nicotine-delivery product, such as the e-cigarette. The emergence of the e-cigarette as a harm reduction product that could decrease smoking related disease has currently received great attention among regulators and the scientific community. Evidence suggests that, although not absolutely safe [9], it represents a significantly less harmful form of nicotine intake [10,11] and can become an important tool for smoking reduction [12] and cessation [13], even in highly-dependent smokers (based on Fagerström Test for Cigarette Dependence) with very high daily cigarette consumption [14]. Thus, e-cigarettes could be used as a harm-reduction approach, with the main goal of reducing the adverse health effects of smoking in the population of khat users. However, the acceptability and subsequent effects of e-cigarette use has not been tested among tobacco smoking khat chewers. E-cigarettes could be used as an alternative to smoking for this population, due to their ability to deliver nicotine (which is needed to enhance khat effects) as well as substitute the sensory-motor aspects of the smoking behavior. Additionally, their use could potentially reduce second and third-hand exposure to the tobacco cigarette smoke observed in rooms where khat-chewing sessions are performed.
2. Future Research Directions
Future research should consider evaluating the use of e-cigarettes in khat chewing sessions as a specific need within the broader general need for research regarding the safety and efficacy of e-cigarettes [15]. Different research questions related to tobacco use along with khat chewing should be tested, including the following:
1. Can e-cigarettes enhance the effect of khat use similarly to tobacco smoking?
2. Can e-cigarettes effectively reduce the need of khat chewers to smoke, leading to smoking reduction or cessation?
3. Can e-cigarettes have an effect in reducing khat use?
4. Will khat chewers who use waterpipe (predominantly women) accept the use of e-cigarette products that resemble waterpipe smoking (there are electronic shisha products with design similar to waterpipe which use e-cigarette liquid and produce aerosol similar to conventional e-cigarettes)?
5. Will partial or complete switching from tobacco smoking to e-cigarette use result in reduced biomarkers of toxins exposure (nitrosamines, polycyclic aromatic hydrocarbons, etc.) in khat chewers?
3. Conclusions
In khat chewing countries, e-cigarettes could become a promising bullet that might reduce the health risks of smoking khat chewers and the second and third-hand exposure of people living around them. That would be consistent with the Framework Convention on Tobacco Control and Tobacco Treatment (FCTC) goals of treating tobacco dependence and protecting people from second-hand exposure while taking into account national circumstances and priorities [16]. With the current opinion letter we propose the development of a new research agenda, focusing on the efficacy of e-cigarette use on tobacco smoking khat chewers.
Author Contributions: Saba Kassim conceived the idea and wrote the first draft of the manuscript. Both Konstantinos E. Farsalinos and Saba Kassim wrote the manuscript and approved it.
Conflicts of Interest: Saba Kassim declares no conflict of interest. Two of the studies by Konstantinos E. Farsalinos were performed using unrestricted funds provided to the institution by e-cigarette companies in 2013.
References
1. Advisory Council on the Misuse of Drugs (ACMD). Khat: A Review of Its Potential Harms to the Individual and Communities in the UK. Home Office, 2013. Available online: http://www.homeoffice.gov.uk/publications/agencies-public-bodies/acmd1/acmd-khat-report-2013 (accessed on 19 September 2015).
2. Kelly, J.P. Cathinone derivatives: A review of their chemistry, pharmacology and toxicology. Drug Test. Anal. 2011, 3, 439–453. [CrossRef] [PubMed]
3. Kassim, S.; Islam, S.; Croucher, R.E. Correlates of nicotine dependence in UK resident Yemeni khat chewers: A cross-sectional study. Nicotine Tob. Res. 2011, 13, 1240–1249. [CrossRef] [PubMed]
4. Kassim, S.; Jawad, M.; Croucher, R.; Akl, E.A. The epidemiology of tobacco use among khat users: A systematic review. BioMed Res. Int. 2015, 2015. [CrossRef] [PubMed]
5. Nakajima, M.; al’Absi, M.; Dokam, A.; Alsoofi, M.; Khalil, N.S.; Al Habori, M. Gender differences in patterns and correlates of khat and tobacco use. Nicotine Tob. Res. 2013, 15, 1130–1135. [CrossRef] [PubMed]
6. Al-Motarreb, A.; Baker, K.; Broadley, K.J. Khat: Pharmacological and medical aspects and its social use in Yemen. Phytother. Res. 2002, 16, 403–413. [CrossRef] [PubMed]
7. Griffiths, P.; Gossop, M.; Wickenden, S.; Dunworth, J.; Harris, K.; Lloyd, C. A transcultural pattern of drug use: Qat (khat) in the UK. Brit. J. Psychiat. 1997, 170, 281–284. [CrossRef] [PubMed]
8. Luqman, W.; Danowski, T.S. The use of khat (catha edulis) in Yemen. Social and medical observations. Ann. Intern. Med. 1976, 85, 246–249. [CrossRef] [PubMed]
9. Shantakumari, N.; Muttappallymyalil, J.; John, L.J.; Sreedharan, J. Cigarette alternatives: Are they safe? Asian Pac. J. Cancer Prev. 2015, 16, 3587–3590. [CrossRef] [PubMed]
10. Farsalinos, K.E.; Le Houezec, J. Regulation in the face of uncertainty: The evidence on electronic nicotine delivery systems (e-cigarettes). Risk Manag. Healthcare Policy 2015, 8, 157–167. [CrossRef] [PubMed]
11. Nutt, D.J.; Phillips, L.D.; Balfour, D.; Curran, H.V.; Dockrell, M.; Foulds, J.; Fagerstrom, K.; Letlape, K.; Milton, A.; Polosa, R.; et al. Estimating the harms of nicotine-containing products using the mcda approach. Eur. Addict. Res. 2014, 20, 218–225. [CrossRef] [PubMed]
12. Begh, R.; Lindson-Hawley, N.; Aveyard, P. Does reduced smoking if you can’t stop make any difference? BMC Med. 2015, 13. [CrossRef] [PubMed]
13. McRobbie, H.; Bullen, C.; Hartmann-Boyce, J.; Hajek, P. Electronic cigarettes for smoking cessation and reduction. Cochrane Database Syst. Rev. 2014, 12, CD001216. [PubMed]
14. Farsalinos, K.E.; Spyrou, A.; Tsimopoulou, K.; Stefopoulos, C.; Romagna, G.; Voudris, V. Nicotine absorption from electronic cigarette use: Comparison between first and new-generation devices. Sci. Rep. 2014, 4. [CrossRef] [PubMed]
15. WHO. Electronic Nicotine Delivery Systems. Conference of the Parties to the WHO Framework Convention on Tobacco Control: Sixth Session Moscow, Russian Federation. Available online: http://apps.who.int/gb/fctc/pdf/cop6/fctc_cop6_10-en.pdf (accessed on 30 November 2015).
16. WHO. Who Report on the Global Tobacco Epidemic, the Mpower Package. Available online: http://www.who.int/tobacco/mpower/mpower_report_full_2008.Pdf (accessed on 18 October 2015). | 2025-03-04T00:00:00 | olmocr | {
"pdf_page_numbers": [
[
0,
3394,
1
],
[
3394,
7413,
2
],
[
7413,
10684,
3
]
]
} |
c2130276c61f018fdfae4c937e794041361034ed | {
"olmocr-version": "0.1.59",
"pdf-total-pages": 17,
"total-fallback-pages": 0,
"total-input-tokens": 59492,
"total-output-tokens": 14778,
"fieldofstudy": [
"Biology"
]
} | Specialized Proresolving Mediators Protect Against Experimental Autoimmune Myocarditis by Modulating Ca\(^{2+}\) Handling and NRF2 Activation
Almudena Val-Blasco, PhD,a,b,* Patricia Prieto, PhD,c,d,* Rafael Iñigo Jaén, BSc,e,f* Marta Gil-Fernández, BSc,a,e,* Marta Pajares, PhD,a,f,g Nieves Domenech, PhD,d,e,g Verónica Terrón, BSc,a María Tamayo, PhD,e,* Inmaculada Jorge, PhD,ab Jesús Vázquez, PhD,b,c,d Andrea Bueno-Sen, BSc,a María Teresa Vallejo-Cremades, PhD,i Jorge Pombo-Otero, MD,a Sergio Sanchez-García, BSc,a Gema Ruiz-Hurtado, PhD,b Ana María Gómez, PhD,b Carlos Zaragoza, PhD,b,c Carmen Delgado, PhD,d,c,f Lisardo Boscá, PhD,a,d María Fernández-Velasco, PhD,c,d
**VISUAL ABSTRACT**
**HIGHLIGHTS**
- Administration of BML-111, a stable LXA4 analog, protects against cardiac dysfunction by avoiding Ca\(^{2+}\) mishandling induced by autoimmune myocarditis in a mouse model.
- Beneficial effects of the SPMs on intracellular Ca\(^{2+}\) handling are mainly caused by a regulation of SERCA2A by NRF2.
- Cardiac tissue obtained from individuals diagnosed with myocarditis, compared with healthy myocardium tissues, displayed depressed mRNA levels of ATP2A2 (SERCA2A) and NF2L2 (NRF2).
Role of Proresolving Mediators in Myocarditis-induced Ca\(^{2+}\) Mishandling
M yocarditis is an inflammatory disease of the myocardium caused by infectious and noninfectious agents and is considered a precursor of dilated cardiomyopathy.\(^1\) Myocarditis has recently been identified in severe forms of COVID-19 and can worsen the prognosis of patients infected with SARS-CoV-2.\(^2\)
Myocarditis has a heterogeneous clinical course, ranging from asymptomatic forms to acute coronary syndromes, including new-onset heart failure, cardiac arrhythmias, or chronic heart failure.\(^3\) Many of these features have been associated with mishandling of intracellular Ca\(^{2+}\),\(^4\) which is a key factor in the regulation of excitation-contraction (EC)-coupling controlling muscle contraction in the heart. EC-coupling is initiated by an action potential in cardiomyocytes that stimulates a small influx of Ca\(^{2+}\) via sarclemma L-type Ca\(^{2+}\) channels. Entry of Ca\(^{2+}\) into the cytosol triggers a large release of Ca\(^{2+}\) from the sarcoplasmic reticulum (SR) through ryanodine receptors (RyR2), resulting in increased intracellular Ca\(^{2+}\) concentrations that activate myofilaments, prompting cell contraction.\(^5\) Relaxation is achieved by reuptake of Ca\(^{2+}\) into the SR by sarcoplasmic reticulum-adenosine triphosphatase 2A (SERCA2A), and across the plasma membrane by the Na\(^+/Ca^{2+}\) exchanger. SERCA2A activation is highly regulated by the small protein phospholamban. Ca\(^{2+}\) mishandling is closely associated with cardiac dysfunction and arrhythmias, but little is known about EC-coupling in the physiopathology of myocarditis.\(^6\)
The different etiologies of myocarditis together with challenges in diagnosis caused by heterogeneity in clinical presentation limit the options for treatment. In many cases, patients with acute myocarditis are treated for heart failure, but they can also be treated with immunosuppressants, antivirals (including interferon beta), nonsteroidal anti-inflammatory drugs, or colchicine depending on symptomatology.\(^1\) Thus, efforts to develop more specific therapeutic modalities are required. Given that the pathologic phenotype of myocarditis is typically related to cardiac inflammation, specific mediators that promote the resolution...
of inflammation, such as specialized proresolving mediators (SPMs), are relevant candidates for research.
SPMs were first elucidated by Serhan and Savill as eicosanoids that dominate the resolution phase of inflammation and help to re-establish tissue homeostasis. The fatty acid arachidonic acid is the precursor of some of the most studied SPMs, termed lipoxins, that exert their actions principally through engagement with the G protein-coupled receptor FPR2/ALXR. Lipoxins are SPMs enzymatically formed by transcellular biosynthesis via 3 different routes: 2 of which are endogenous pathways for lipoxin A4 (LXA4) and LXB4 biosynthesis, and a third pathway that, in the presence of aspirin, generates epi-derivatives of lipoxins known as 15-epi-lipoxin A4/B4. These “aspirin-triggered lipoxins” retain the main actions of native compounds but with significantly increased potency. Because of their short half-life and rapid inactivation in vivo, several synthetic derivatives of native lipoxins have been developed such as the LXA4 receptor agonist S(S), (6)R, 7-trihydroxyheptanoic acid methyl ester (BML-111) (Supplemental Figure 1), which has improved stability and potency in vivo.
SPMs can prevent the development of chronic inflammation and temper the formation of oxidative stress, and many of their beneficial effects have been attributed to the inhibition of the proinflammatory nuclear factor-kB pathway and to the activation of the antioxidant factor nuclear factor erythroid-derived 2-like 2 (NRF2). NRF2 is a transcription factor that can directly and indirectly activate the expression of anti-inflammatory genes and also induces the transcriptional repression of some proinflammatory genes.
NRF2 may counteract inflammation indirectly by modulating oxidative stress pathways, such as those governed by NQO1 and heme oxygenase 1, leading to an anti-inflammatory response. Although the cardioprotective role of NRF2 has been extensively studied, whether it is involved in the beneficial SPM-mediated cardiac effects through the regulation of intracellular Ca2+ dynamics is unknown.
In the present study, we investigated whether administration of BML-111 in an experimental autoimmune myocarditis (EAM) model in mice prevents cardiac dysfunction and intracellular Ca2+ mishandling via NRF2 activation.
METHODS
ANIMALS. The present study was conducted following the guidelines of the Spanish Animal Care and Use Committee according to the European Union (2010/63/EU) and conformed to the Guide for the Care and Use of Laboratory Animals Published by the U.S. National Institute of Health (NIH Publication No 85-23, revised 1996). The study was approved by the Bioethics Committee of the Community of Madrid (PROEX 144/17).
HUMAN SAMPLES. Endomyocardial biopsies were obtained from patients who developed myocarditis (defined according to the American College of Cardiology and American Heart Association clinical guidelines). All samples were obtained from excess tissue initially extracted for clinical diagnosis. Biopsies were obtained from healthy donors’ hearts from Biobanco A Coruña (Cod.0000796). The 8 patients with myocarditis (6 male and 2 female) had a median age of 31 years. The 5 healthy myocardium donors (4 male and 1 female) had a median age of 56 years. The study was conducted according to Spanish Law for Biomedical Research (Law 14/2007-3 of July) and was compliant with the Declaration of Helsinki. The study and the use of samples were approved by the Research Ethics Committee of Galicia (Ref 2017/541). Written informed consent was obtained from all patients.
STATISTICS. The values shown in graphs correspond to the mean ± SD. Differences between group mean values were estimated with the 2-tailed Student’s t-test for unpaired observations, and more than 2 groups were compared using 1-way analysis of variance followed by Tukey’s post hoc test for multiple comparisons, as indicated. In the proteomics analysis, outliers at the peptide and protein levels were detected at 1% false discovery rate and results of modified cysteine (Cys)-containing peptide abundance changes were tested for significance using the Kolmogorov-Smirnov test. A value of P < 0.05 was considered statistically significant. All analyses were performed using GraphPad Prism Software.
Expanded Methods are provided in the Supplemental Appendix.
RESULTS
STRUCTURAL AND FUNCTIONAL CARDIAC CHARACTERIZATION OF MICE WITH INDUCED EAM AND THE EFFECTS OF BML-111. Infiltrated immune cells are commonly found in the myocardium of patients with myocarditis. Hematoxylin-eosin staining demonstrated that mice treated with α-myosin heavy chain and vehicle (EAM+Veh) had more immune cell infiltration in the myocardium than vehicle-treated control mice (Ctrl+Veh) did (Figure 1B), as expected in EAM model. Cardiac expression of the
FIGURE 1 BML-111 Administration Prevents Cardiac Dysfunction and Systolic Ca\textsuperscript{2+} Dysregulation in EAM-induced Mice
(A) Experimental design. (B) Representative hematoxylin and eosin–stained slides of hearts from control + vehicle, (Ctrl + Veh), control + 5(S), (6)R, 7-trihydroxyheptanoic acid methyl ester (BML-111) (Ctrl + BML), experimental autoimmune myocarditis (EAM) + vehicle (EAM + Veh), and EAM + BML-111 (EAM + BML)-treated mice. Original magnification ×20. (C) Individual and mean values of ejection fraction (EF) in Ctrl + Veh (N = 7), Ctrl + BML (N = 8), EAM + Veh (N = 9), and EAM + BML (N = 10) mice. (D) Representative traces of calcium influx through L-type channels (ICaL) obtained from −60 to +60 mV acquired in a cardiomyocyte from a vehicle-treated mouse. (Right) Mean density values of ICaL recorded at all voltages tested in cardiomyocytes from Ctrl + Veh (N = 12, N = 4), Ctrl + BML (N = 18, N = 3), EAM + Veh (N = 12, N = 5), and EAM + BML (N = 18, N = 4) groups. (E) Representative line-scan confocal images and the corresponding profiles of Ca\textsuperscript{2+} transients from 1 cell in each experimental group. (F) Individual and mean values of peak fluorescence Ca\textsuperscript{2+} transients (left), the decay time constant (Tau) (center), and cell shortening (right) obtained in cardiomyocytes from Ctrl + Veh (N = 21, N = 5), Ctrl + BML (N = 28, N = 4), EAM + Veh (N = 26, N = 4), and EAM + BML (N = 26, N = 4) mice. Data show individual values and mean ± SD. *P < 0.05, **P < 0.01, and ***P < 0.001 versus Ctrl + Veh; and ##P < 0.01 and ###P < 0.001 versus EAM + Veh.
inflammatory cytokines Il1β and Tnfα was also significantly higher in EAM-Veh mice than in Ctrl-Veh mice (Table 1), corroborating the proinflammatory cardiac environment. By contrast, EAM treated mice with BML-111 (EAM+BML) did not exhibit immune cell infiltration or elevated cardiac inflammatory biomarkers (Figure 1B, Table 1), and the mean values of all the parameters were similar to those in control mice treated with BML-111 (Ctrl+BML).
As myocarditis is commonly associated with cardiac dysfunction and hypertrophy, we used trans-thoracic echocardiography to assess whether BML-111 treatment could improve these outcomes in EAM-induced mice. As expected, mean ejection fraction and fractional shortening were both significantly lower without changes to heart rate in EAM-Veh mice than in Ctrl-Veh mice (Figure 1C, Table 1, Supplemental Figure 2), and these functional changes were accompanied by cardiac and cellular hypertrophy, as illustrated by an increase in the heart weight/tibia length ratio, cardiomyocyte surface area, and cardiac Nppa messenger RNA (mRNA) levels in EAM-Veh mice (Table 1). By contrast, treatment with BML-111 prevented cardiac dysfunction (Figure 1C, Supplemental Figure 2) and cardiac and cellular hypertrophy development (Table 1) in EAM-induced mice, with similar mean values of all parameters in Ctrl+BML and EAM+BML groups. Additionally, BML-111 administration prevented both the increase in collagen deposition and the higher expression of Tgfβ1, Col1α1, and Col3α1 induced by EAM (Supplemental Figure 3, Supplemental Table 1).
Furthermore, plasma levels of LXA4 were determined in all experimental groups. Supplemental Figure 4 shows increased values in EAM-Veh and EAM-BML groups compared with in the Ctrl-Veh group. Finally, cardiac expression of key enzymes of the biosynthesis of the LXA4 pathway was determined in all groups. Results indicated no significant changes in mRNA levels of Fpr2/Alix, Alox5, Alox12, and Alox15 between groups (Supplemental Figure 5). Overall, these data demonstrate that BML-111 prevents cardiac inflammation and functional and structural heart remodeling induced by EAM.
**BML-111 COUNTERACTS THE IMPAIRMENT IN SYSTOLIC, DIASTOLIC Ca2+ RELEASE AND SR-Ca2+ LOAD INDUCED BY EAM.** Because cardiac dysfunction is closely related to Ca2+ mishandling,12 we next analyzed EC-coupling in isolated cardiomyocytes. We first questioned whether the Ca2+ influx into cardiomyocytes through L-type Ca2+ channels was altered by EAM, finding that the mean value of the current-voltage density curves for influx into cardiomyocytes through L-type Ca2+ channels was similar among the 4 experimental groups (Figure 1D). We next recorded systolic Ca2+ release by measuring Ca2+ transients elicited in 2 Hz-paced cardiomyocytes. Representative images of the Ca2+ transient recordings and the corresponding profile of each experimental group are shown in Figure 1E. Results showed that BML-111 administration significantly attenuated the decrease in amplitude (Figure 1F, left panel) and the slower time decay of Ca2+ transients (Figure 1F, center panel) induced by EAM, and improved the cell shortening (Figure 1F, right panel). We hypothesized that the slower kinetics of Ca2+ transients in the EAM-Veh group may be caused by an impairment of SERCA2A pump function. Indeed, the rate of Ca2+ uptake was significantly higher in the EAM-Veh group than in the Ctrl-Veh group, whereas values in the EAM+BML and Ctrl+BML groups were similar and lower than the values in the EAM-Veh group (Supplemental Figure 6). We examined whether the functional alterations in SR-Ca2+ uptake were associated with changes in the expression of SERCA2A. Cardiac mRNA levels of SERCA2A (Atp2a2) were significantly lower in the EAM-Veh group than in the Ctrl-Veh group, but they were similar between the EAM+BML and Ctrl-Veh groups (Figure 2A, left), indicating that BML-111 blunted the down-regulation of Atp2a2 expression in EAM-induced mice. Proteomic analysis corroborated the gene expression results (Figure 2A, right). By contrast, no changes between groups were observed in the phosphorylation state of the key regulator of SERCA2A,
phospholamban, neither in Ser 16 nor in Thr 17 sites (Supplemental Figure 7).
Because the observed impairment in systolic Ca\(^{2+}\) release and SERCA2A could also be related to changes in the SR-Ca\(^{2+}\) load, we estimated the total SR-Ca\(^{2+}\) load as caffeine-evoked Ca\(^{2+}\) transients. Representative images of these recordings are shown in the left panel of Figure 2B. Results indicated that...
Figure 3: Comparative Proteomics Analysis Reveals Coordinated Protein Abundance Changes in the Heart Proteome of EAM-Induced Mice After BML-111 Treatment
(A) Functional categories significantly altered (*false discovery rate < 0.05) in the EAM+iVeh, Ctrl+iBML, and EAM+iBML groups with respect to Ctrl+iVeh (control). The Systems Biology Triangle algorithm was used to detect coordinated protein changes by BML-111 treatment. Proteins were functionally annotated using Gene Ontology, Kyoto Encyclopedia of Genes and Genomes, and Reactome terms retrieved from the Uniprot repository. Zc values are log2 ratios of categories expressed in units of SD. Supplemental Table 4 displays Zq values for the protein components of the functional categories that were significantly altered.
(B) Relative quantification of the proteins making up the “response to oxidative stress” category showing a highly interconnected network. Protein-protein associations were taken from the STRING database.33
(C) Redox proteomics analysis using the filter-aided stable isotope labeling of oxidized cysteine (Cys) method detects a significant increase in reversible Cys oxidation levels in the heart proteome in the comparisons EAM+iVeh versus Ctrl+iVeh and EAM+iBML versus Veh+iBML groups taking as reference the global behavior of all the peptides from the proteome, which follows the expected null hypothesis. The increase in oxidation level was significantly lower when comparing the modified Cys-containing peptide abundance changes from untreated animals (EAM+iVeh vs Ctrl+iVeh) with the corresponding BML-treated animals (EAM+iBML vs Ctrl+iBML). The cumulative distribution of Zpq values for the total and the modified Cys-containing peptides is shown. ATP = adenosine triphosphate; MHC = major histocompatibility complex; TCA = tricarboxylic acid; other abbreviations as in Figure 1.
the amplitude of caffeine-evoked Ca$^{2+}$ transients was significantly lower in cardiomyocytes from EAM+Veh mice than from Ctrl+Veh mice (Figure 2B, right). By contrast, the amplitude of caffeine-evoked Ca$^{2+}$ transients was similar between cardiomyocytes from EAM+Veh mice and Ctrl-BML mice and comparable to those from Ctrl+Veh mice.
Changes in SR-Ca$^{2+}$ load are often related to aberrant diastolic Ca$^{2+}$ release. To question this possibility, we analyzed the frequency of small Ca$^{2+}$ leakage
FIGURE 5 Luteolin Blocks the Improvement in Cardiac Function and Ca\textsuperscript{2+} Dynamics Produced by BML-111 in EAM-Induced Mice
(A) Experimental design. (B) Plots illustrate individual and mean values of EF in Ctrl\textsuperscript{+}Veh (N = 10), Ctrl\textsuperscript{+}BML (N = 10), EAM\textsuperscript{+}Veh (N = 7), and EAM\textsuperscript{+}BML (N = 10) mice treated with luteolin. (C-E) Individual and mean values of peak fluorescence Ca\textsuperscript{2+} transients (C), decay time constant (D), and cell shortening (E) in cardiomyocytes from Ctrl\textsuperscript{+}Veh (N = 36, N = 4), Ctrl\textsuperscript{+}BML (N = 25, N = 4), EAM\textsuperscript{+}Veh (N = 26, N = 4), and EAM\textsuperscript{+}BML (N = 37, N = 4) from luteolin-treated mice. Data show individual values and mean \pm SD. *(P < 0.05), **(P < 0.01), and *** (P < 0.001) versus Ctrl\textsuperscript{+}Veh; and $\$ (P < 0.05), $\$$ (P < 0.01), and $$$ (P < 0.001) versus Ctrl\textsuperscript{+}BML. SR = sarcoplasmic reticulum; other abbreviations as in Figures 1 and 2.
in the form of Ca\(^{2+}\) sparks in quiescent cells. We found no changes in the number of these events among the 4 experimental groups (Supplemental Figure 8); however, other proarrhythmogenic forms of spontaneous diastolic Ca\(^{2+}\) release (SCR), such as Ca\(^{2+}\) waves and spontaneous Ca\(^{2+}\) transients, were significantly higher in the EAM+Veh group than in the Ctrl+Veh group (Figure 2C). Notably, the percentage of SCR in EAM+BML mice (4%) was significantly lower than that in EAM+Veh mice (12%; \(P < 0.001\)) (Figure 2C, right), supporting the notion that BML-111 prevents the increased diastolic Ca\(^{2+}\) leak induced by EAM. The effects of BML-111 on diastolic Ca\(^{2+}\) release may contribute to preserve the SR-Ca\(^{2+}\) load at physiological levels, prompting regular systolic Ca\(^{2+}\) release in EAM+BML cotreated mice by maintaining ER\(\text{Ca}^{2+}\)A physiological levels. These beneficial effects can explain the normal cell contractility and thus cardiac function in EAM mice treated with BML-111.
**BML-111 Administration Induces a Protective Effect Against Oxidative Stress Through NRF2 Modulation.** We performed high-throughput multiplexed quantitative proteomics analysis to determine whether the beneficial effects of BML-111 on EAM were related to changes in the cardiac protein profile. Analysis of tissue protein extracts (4 animals per group) allowed the quantification of 2,332 proteins with a false discovery rate of 1% (Supplemental Table 2). Application of the Systems Biology Triangle algorithm revealed significant abundance changes in functional categories (false discovery rate <0.05) in EAM+Veh, EAM+BML, and Ctrl+BML mice compared with in Ctrl+Veh mice (Supplemental Table 3). To improve the biological interpretation, we manually grouped the categories into 6 functional clusters (Figure 3A). Muscle contraction and some mitochondrial processes such as tricarboxylic acid cycle, respiratory electron transport, and amino acid metabolism were significantly lower in the EAM+Veh group than in the Ctrl+Veh group. Contrastingly, these functional categories, as well as other mitochondrial processes such as adenosine triphosphate, glycolylate, pyruvat metabolism, and fatty acid \(\beta\)-oxidation were significantly higher in the Ctrl+BML and EAM+BML groups. In addition, proteins involved in the regulation of muscle contraction through the Ca\(^{2+}\) signaling pathway were more abundant in the Ctrl+BML and EAM+BML groups. Of note, the immune response cluster was significantly elevated in EAM+Veh mice, but it was diminished in EAM+BML mice and was lower still in Ctrl+BML mice. Taken together, these results suggest that BML-111 can recover myocardial contraction by attenuating the mitochondrial dysfunction induced by EAM and by limiting the exacerbated autoimmune response.
Myocarditis is associated with increased oxidative stress. To further explore the protective function of BML-111 against oxidative stress, we focused on the “response to oxidative stress” functional category, which was significantly lower in EAM+Veh mice than in Ctrl+Veh mice (Figure 3A). Proteins implicated in cellular reactive oxygen species removal such as Sod1, Txn1, Cat, and Prdx5 were increased in abundance in both EAM+BML and Ctrl+BML mice (Figure 3B), suggesting that BML-111 stimulates the activity of these detoxifying enzymes. In addition, we analyzed the heart tissue “redoxome” by quantitating the reversible Cys oxidation levels in the 4 experimental groups using the filter-aided stable isotope labeling of oxidized Cys method. We observed a significantly higher abundance of oxidized Cys-containing peptides in EAM+Veh mice than in Ctrl+Veh mice. The levels of reversible Cys oxidation were significantly lower in BML-111-treated animals (Figure 3C), highlighting the protective role of BML-111 against oxidative damage in the heart.
Next we explore in-depth the elevated oxidative stress environment in the EAM model. We determined cardiac oxidative stress by staining for the oxidative DNA damage biomarker 8-hydroxy-2'-deoxyguanosine (8-OHdG), finding that 8-OHdG cardiac staining was significantly greater in EAM+Veh mice than in Ctrl+Veh mice (Figures 4A-4C). As anticipated, 8-OHdG staining in the myocardium was significantly lower in EAM+BML mice than in EAM+Veh mice (Figures 4B and 4C). As many of the beneficial antioxidant actions of lipoxins are mediated through NRF2 activation, we determined its activation by analyzing its nuclear accumulation in heart. We found that nuclear levels of NRF2 were significantly higher in myocardium of BML-111-treated mice (in both Ctrl and EAM-induced mice) than in nontreated mice (Figure 4D), supporting the notion that BML-111 prevents myocarditis-induced oxidative stress by activating NRF2. We also performed a series of experiments in EAM-induced mice coadministered with luteolin, a compound that can induce oxidative stress through inhibition of the NRF2 pathway (schema in Figure 4A).\(^{15}\) Luteolin coadministration blunted the protective effects of BML-111 on oxidative stress, as shown by a significant increase in 8-OHdG staining in the myocardium of Ctrl and EAM-induced mice treated with BML-111 (Figures 4B and 4C). Additionally, luteolin
counteracted the increase in cardiac nuclear NRF2 levels induced by BML-111 administration in both Ctrl and EAM groups (Figure 4D). Overall, these data indicate that luteolin blocks BML-111-induced activation of NRF2 in the myocardium, thus preventing the protective action of BML-111 against cardiac oxidative stress. These results point to luteolin as a useful tool to determine the role of NRF2 in Ca$^{2+}$ mishandling shown in our EAM model.
**LUTEOLIN ADMINISTRATION BLUNTS THE PROTECTIVE EFFECTS OF BML-111 ON CARDIAC FUNCTION AND INTRACELLULAR Ca$^{2+}$ DYNAMICS IN EAM-INDUCED MICE.** We next questioned whether luteolin lessens the protective effects of BML-111 on cardiac function and intracellular Ca$^{2+}$ handling in EAM-induced mice (schema in Figure 5A). The values for ejection fraction and fractional shortening in EAM-induced mice cotreated with BML-111 and luteolin (EAM–Lut+BML) were similar to those found in EAM+Lut mice and were significantly lower than those of Ctrl mice (Figure 5B, Supplemental Figure 9).
At the cellular level, luteolin coadministration dampened the protective effect of BML-111 on systolic Ca$^{2+}$ release and SR-Ca$^{2+}$ load in cardiomyocytes. Indeed, the depressed amplitude of Ca$^{2+}$ transients and cell shortening, as well as the low SR-Ca$^{2+}$ load, in the EAM–Lut group was maintained in the EAM+Lut+BML group (Figures 5C to 5F). Of note, the administration of luteolin sustained the slower Ca$^{2+}$ transient kinetics (Figure 5D), indicating that SERCA2A impairment was maintained in EAM+Lut+BML mice. Supporting these findings, we found higher values for Ca$^{2+}$ uptake rate in cells from EAM-Lut group compared with the Ctrl group, and the coadministration of BML-111 was unable to reverse this increase (Supplemental Figure 10). Finally, the increased SCR in EAM–Lut mice (19%) was also evident in EAM–Lut+BML mice (18%) (Figure 5G).
Overall, these results suggest that NRF2 might be a key player in the protective effects of BML-111 on cardiac dysfunction and intracellular Ca$^{2+}$ mishandling induced by EAM.
**GENETIC DELETION OF NRF2 AVERTS THE IMPROVEMENT IN Ca$^{2+}$ HANDLING INDUCED BY LIPOXINS.** To definitively determine whether NRF2 mediates the effects of SPMs on Ca$^{2+}$ dynamics and to evaluate the direct effect of lipoxins in isolated cardiomyocytes, we conducted a series of experiments using cardiomyocytes isolated from wild-type (Wt) and Nrf2 knockout (Nrf2$^{-/-}$) mice. Cells were incubated for 1 hour with Veh or with 250 nmol/L 15-epi-lipoxin A4 (Epi), an LXA4 derivative widely used in vitro experiments, and systolic Ca$^{2+}$ release and SR-Ca$^{2+}$ load/uptake were determined. Representative Ca$^{2+}$ transients obtained in each condition are shown in Figure 6A. Results showed that acute administration of Epi improved systolic Ca$^{2+}$ release in Wt cardiomyocytes, mainly by increasing the amplitude of Ca$^{2+}$ transients (Figure 6B, left panel) and accelerating their kinetics (Figure 6B, central panel). Also, SR-Ca$^{2+}$ uptake by SERCA2A was improved by Epi treatment (Figure 6B, right panel), which can also contribute to increase SR-Ca$^{2+}$ load in Wt cells (Figure 6C). By contrast, Epi administration to cardiomyocytes from Nrf2$^{-/-}$ mice had no effect on systolic Ca$^{2+}$ release or on SR-Ca$^{2+}$ uptake and load (Figures 6A to 6C). These results support the notion that NRF2 is a key mediator that improves systolic Ca$^{2+}$ release and SR-Ca$^{2+}$ uptake and load in isolated cardiomyocytes and point to SERCA2A as a target of NRF2 action. Interestingly, the effects induced by Epi on Ca$^{2+}$ handling were blunted in cardiomyocytes cotreated with the selective inhibitor of NRF2, ML-385 (Supplemental Figure 11). Consistent with this idea, we found that cardiac tissue from Wt mice previously treated with BML-111 showed elevated protein levels of SERCA2A (Figure 6D). To determine whether the observed effects of Epi on Ca$^{2+}$ handling were mediated by FPR2/ALX receptor, cardiomyocytes were cotreated with Epi and the FPR2/ALX antagonist, WRW4, and both systolic Ca$^{2+}$ release and SR-Ca$^{2+}$ load were analyzed. Supplemental Figure 12 shows that WRW4 administration avoided the effects induced by Epi on Ca$^{2+}$ transient amplitude and kinetics and also prevented the modulation of Epi on SR-Ca$^{2+}$ load, suggesting that Epi induces its acute effects on intracellular Ca$^{2+}$ release through FPR2/ALXR. Overall, these results highlight SERCA2A and NRF2 as key modulators of the beneficial effects of SPMs on Ca$^{2+}$ handling.
**TRANSCRIPTIONAL ACTIVITY OF NRF2 DETERMINES SERCA2A EXPRESSION IN HUMAN VENTRICULAR CARDIOMYOCYTES.** Having shown that BML-111 modulates murine cardiac SERCA2A expression and that NRF2 appears to be the main contributor of its effects on Ca$^{2+}$ dynamics, we next investigated whether the molecular regulation of NFE2L2 (NRF2 coding gene) determines SERCA2A expression. We evaluated SERCA2A expression in a human ventricular cardiomyocyte cell line (AC16) silenced for NFE2L2 expression or overexpressing NFE2L2. Results showed that the down-regulation of NFE2L2 in AC16 cells impaired the expression of its canonical targets (HMOX1 and NQO1) and also decreased the expression of SERCA2A (Figure 7A). Conversely, the
Cardiomyocytes from wild-type (WT) and Nrf2 knockout (Nrf2<sup>−/−</sup>) mice were incubated for 1 hour with Veh or 250 nmol/L 15-epi-lipoxin A4 (Epi). (A) Representative line-scan confocal images and the corresponding profiles of Ca<sup>2+</sup> transients from 1 cell in each experimental group. (B) Individual and mean values of peak fluorescence Ca<sup>2+</sup> transients (left), decay time constant (Tau) (center), and the rate of Ca<sup>2+</sup> uptake (right) in WT+Veh (N = 25, N = 4); WT+Epi (N = 36, N = 4), Nrf2<sup>−/−</sup>+Veh (N = 22, N = 3) and Nrf2<sup>−/−</sup>+Epi (N = 24, N = 3) cardiomyocytes. (C) Individual and mean values of the amplitude of the caffeine-induced Ca<sup>2+</sup> transients (F/F<sub>0</sub>) obtained in WT+Veh (N = 26, N = 4), WT+Epi (N = 36, N = 4), Nrf2<sup>−/−</sup>+Veh (N = 15, N = 3), and Nrf2<sup>−/−</sup>+Epi (N = 15, N = 3) cells. Data show individual values and mean ± SD. **P < 0.01 versus WT+Veh. (D, upper) Representative Western blots of SERCA2A and vinculin from cytosolic heart lysates from WT mice treated with Veh (Ctrl) or BML-111 (BML) for 6 hours. (Lower) Scatter plots summarizing the data in Ctrl (N = 4) and BML-111-treated (N = 5) mice expressed as arbitrary units. Data show individual values and mean ± SD. *P < 0.01 versus Ctrl. Abbreviations as in Figures 1, 2, and 5.
**FIGURE 7** Correlation Between NRF2 and SERCA2A2 Expression in Human Ventricular Cells and in Myocardium From Patients With Myocarditis
(A) AC16 cells were transduced with a lentivirus carrying a short hairpin RNA (shRNA) against a scramble sequence (shCTRL) or against NRF2 (shNRF2). (Left) Representative Western blots of the indicated proteins were determined 7 days post-transduction. Vinculin was used as loading control. (Right) Densitometry quantification of SERCA2A protein levels normalized to vinculin levels. Data are mean ± SD (N = 4). ***P < 0.001 versus shCTRL transduced cells.
(B) AC16 cells were transduced with a lentivirus expressing green fluorescent protein (GFP) or a constitutive active version of NRF2 (NRF2-ΔETGE). (Left) Representative images showing efficient transduction 4 days post-lentivirus delivery. (Center) Representative Western blots of the indicated proteins (NRF2, HMOX1, NQO1, GFP, and SERCA2A) were determined 4 days post-transduction. Vinculin was used as a loading control. (Right) Densitometry quantification of SERCA2A protein levels normalized to vinculin levels. Data are mean ± SD (N = 4). *P < 0.05 and ***P < 0.001 versus GFP transduced cells.
(C) Representative hematoxylin and eosin–stained slides of healthy myocardium (N = 5) and myocardium from patients with myocarditis (N = 8). Original magnification ×10 (right) and ×20 (left).
(D) Individual and mean values of messenger RNA levels of NFE2L2 normalized to human 36B4.
(E) Individual and mean values of ATP2A2 messenger RNA levels in human samples. A linear regression of the data is shown. Data show individual values and mean ± SD. *P < 0.01 and **P < 0.01 versus healthy group. FI = fold induction; other abbreviations as in Figures 1 and 2.
overexpression of NRF2 increased SERCA2A protein levels (Figure 7B). These results strongly suggest that the modulation of NFE2L2 determines SERCA2A expression in human ventricular cardiomyocytes.
**HUMAN MYOCARDIUM FROM PATIENTS WITH MYOCARDITIS HAS LOW NFE2L2 AND ATP2A2 EXPRESSION.** To translate the main results to the human heart, we examined human myocardium from patients with myocarditis. Samples of healthy and myocarditis-positive myocardium were characterized by immunostaining. Representative examples of hematoxylin-eosin staining of healthy cardiac tissue and tissue from patients with a clinical diagnosis of myocarditis are shown in Figure 7C, with evident immune cell infiltration and tissue damage in the latter (Supplemental Figure 13). Analysis of NFE2L2 expression revealed significantly lower myocardial levels in patients with myocarditis than in healthy samples (Figure 7D).
As Ca²⁺ mishandling in the EAM model was closely related to impairment in SERCA2A, we also analyzed ATP2A2 mRNA levels in human myocardium tissue. Results showed that ATP2A2 expression was significantly lower in myocarditis-positive myocardium than in healthy cardiac tissue (Figure 7E).
Finally, correlation analysis revealed a strong and direct association between NFE2L2 and ATP2A2 expression (Figure 7F). These results support our findings in the EAM model, showing a significant impairment of NRF2 and SERCA2A expression and reveal an interconnection between them. By contrast, no changes were found in the expression of FPR2/ALXR or in the key enzymes of the biosynthesis of the LXA4 pathway between groups (Supplemental Figure 14). Overall, our findings identify NRF2 and SERCA2A as novel partners in the functional cardiac remodeling linked to myocarditis.
**DISCUSSION**
Myocarditis is a complex inflammatory disease affecting the myocardium that is triggered by both infectious and noninfectious agents and leads, in many cases, to dilated cardiomyopathy, with a 1-year mortality rate of 15%-20%. The lack of specific treatments for myocarditis adds to the challenges of managing patients with this disease and highlights the need to identify new targets and mechanisms to develop future treatments.
SPMs are potential therapeutic tools for treating cardiovascular diseases with a clear inflammatory component, such as myocarditis, as they can drive the resolution of inflammation to recover a healthy state. Administration of SPMs in various experimental models exerts protective cardiac effects, including halting inflammatory responses, cell death, oxidative stress, and fibrosis development, ultimately improving cardiac function in myocarditis and other cardiovascular diseases.
The potential utility of lipoxins is limited by their short half-life, which has prompted the development of new synthetic drugs such as BML-111, an LXA4 receptor agonist with improved stability and potency. BML-111 has proved effective against different inflammatory diseases, but relatively little is known about its specific cardiac effects. Our study demonstrates that BML-111 administration prevents both structural and functional cardiac remodeling associated with myocarditis in a model of EAM, with functional amelioration closely related to the protection against cardiomyocyte Ca²⁺ mishandling. Specifically, BML-111 treatment prevented the reduced systolic Ca²⁺ release and diminished SR-Ca²⁺ uptake induced by EAM. One of the main protective mechanisms elicited by BML-111 involved the preservation of SERCA2A protein levels, preventing its down-regulation on EAM induction. The maintenance of physiological levels of SERCA2A was pivotal to sustain an adequate SR-Ca²⁺ load, thus contributing to preserve cell contractility and cardiac function in EAM mice treated with BML-111.
Disruptions in EC-coupling can also involve diastolic Ca²⁺ leak through ryanodine receptors in many cardiovascular diseases. Diastolic Ca²⁺ leak in the form of SCR was evident in cardiomyocytes from EAM-induced mice, and these aberrant proarrhythmic events were averted by administered BML-111. The beneficial effects of BML-111 on diastolic Ca²⁺ release together with the maintenance of SERCA2A activity contribute to preserve the SR-Ca²⁺ load at physiological levels, prompting regular systolic Ca²⁺ release and cell contractility in EAM-BML-treated mice. All of these functional results were corroborated by the proteomics analysis, which indicated that the beneficial effects of BML-111 on cardiac function and structure were associated with the prevention of EAM-induced changes in proteins involved in contraction, energy regulation, and Ca²⁺ dynamics. Interestingly, the administration of BML-111 did not modify cardiac expression of FPR2/ALXx receptor or key enzymes of the lipoxin pathway. In contrast, in vitro experiments demonstrated that the treatment of cardiomyocytes with an antagonist of FPR2/ALXR significantly blunted the beneficial effects induced by Epi in systolic Ca²⁺ release and SR-Ca²⁺ load, suggesting that the functional effect of SPMs requires FPR2/ALXR to modulate intracellular Ca²⁺ handling. In this line, an interesting study carried out by Petri...
et al.\(^\text{17}\) pointed to FPR2/ALXR as a main player in the murine atherosclerosis progression.
Our results indicate that SERCA2A is fundamental for BML-111-induced prevention of Ca\(^{2+}\) mishandling, pointing to this SR-adenosine triphosphatase as an excellent candidate to develop new therapies that improve cardiac function in patients with myocarditis. Indeed, SERCA2A gene therapy has been evaluated in clinical trials for heart failure\(^\text{18,19}\) and in preclinical experimental small and large animal models, which show improvements in myocardial contractility.\(^\text{20,21}\) However, discrepancies regarding the beneficial role of targeting SERCA2A were found in the CUPID (Calcium Upregulation by Percutaneous Administration of Gene Therapy in Cardiac Disease) trials\(^\text{19,22}\), which might be explained by the possible low efficiency of gene transduction or possibly by post-translational regulatory factors of SERCA2A in human heart failure. In this regard, our results uncover NRF2 as a key modulator of the expression of SERCA2A in human cardiac cells, supporting an emergent field for clinical research in myocarditis or other cardiovascular diseases such as heart failure that are associated with a clear down-regulation of the SR-adenosine triphosphatase.
NRF2 is a master antioxidant factor that has recently gained much interest in cardiac disease research\(^\text{23-25}\) because of its antioxidant and anti-inflammatory effects. Our results show that the protective effect of BML-111 on cardiac function and Ca\(^{2+}\) handling is related to the promotion of NRF2 activation. Functional Ca\(^{2+}\) analysis revealed that pharmacologic blockade of NRF2 in EAM-induced mice blunted the improvement in the intracellular Ca\(^{2+}\) handling induced by the SPM, with evident harmful effects on cardiac function, and these findings were mirrored in in vitro studies on cardiomyocytes from \(\text{Nrf}2^{-/-}\) mice and cells treated with a selective inhibitor of NRF2. Supporting our data, other studies have reported that genetic deletion or pharmacologic blockade of NRF2 contributes to harmful structural and functional cardiac remodeling by augmenting oxidative stress and the inflammatory response.\(^\text{26-28}\)
Regarding the specific role of NRF2 in the regulation of Ca\(^{2+}\) handling, we found that NRF2 activation by BML-111 in mice or its overexpression in human ventricular cells increased SERCA2A expression, suggesting that the SR-Ca\(^{2+}\) pump is a direct or indirect downstream target of NRF2. Indeed, in preliminary results we have searched the ENCODE database for NRF2-binding enhancer, antioxidant response element (called ARE). Although this database does not report information about NRF2 chromatin immunoprecipitations, other ARE-binding proteins such as MAFF, MAFK, and BACH1 were found to bind the \(\text{ATP}2\text{A}2\) gene at 16 sites located at H3K27Ac and DNAse sensitive sites, consistent with promoter regulatory regions (Supplemental Figure 15).
Supporting our hypothesis was the finding that down-regulation of \(\text{NFE2L}2\) in human cells was associated with a decrease in SERCA2A. BML-111-activated NRF2 triggers the maintenance of SERCA2A levels in a physiological range, improving SR-Ca\(^{2+}\) uptake and promoting systolic Ca\(^{2+}\) release and cardiomyocyte contraction. All of these beneficial effects of BML-111 contributed to prevent the cardiac dysfunction in EAM-induced mice. In line with our results, Erkens et al.\(^\text{26}\) demonstrated that \(\text{Nf}r2^{-/-}\) mice present cardiac dysfunction associated with a down-regulation of SERCA2A. Concerning the inflammatory environment and cardiac NRF2 and SERCA2A regulation, Bai et al.\(^\text{29}\) described that the administration of a strong proinflammatory stimulus (lipopolysaccharide) to mice promoted both cardiac SERCA2A and NRF2 down-regulation, and both processes were attenuated in the presence of the antioxidant resveratrol.
Validating the main findings of our experimental model, we found a down-regulation of both \(\text{NFE2L2F}\) and \(\text{ATP}2\text{A}2\) expression in human myocarditis-positive myocardium, and a significant correlation between the expression of both molecules, pointing to an interconnection between them in human myocarditis. In relation to possible translation of findings to human pathology, it is important to mention that myocarditis has also recently been described as a potential complication of COVID-19,\(^\text{20}\) and cardiac damage and dysfunction induced by myocarditis have been observed in patients both in hospital and over the long term.\(^\text{31}\) Interestingly, a recent lipidomic analysis showed that patients with severe COVID-19 had significantly lower levels of SPMs (resolvins E3) than did peers with moderate disease,\(^\text{32}\) indicating that the loss of SPMs might be associated with the more severe form of COVID-19. Our data may provide a new research direction focusing on NRF2-SERCA2A interventions for the development of specific treatments for myocarditis in general and for cardiac complications in patients with COVID-19.
**CONCLUSIONS**
Our findings point to SPMs as a new therapeutic alternative for treating myocarditis. Our study uncovers that SPMs induce cardioprotection by regulating cardiac function and intracellular Ca\(^{2+}\)
dynamics though the modulation of NRF2 and SERCA2A activation. Our results are particularly relevant for patients with myocarditis in general and also for individuals with COVID-19 and myocarditis, because the employment of specific compounds to promote the resolution phase of the inflammatory response may help to manage the cytokine storm and the consequent tissue damage. Additionally, SPMs can contribute to improve cardiac dysfunction by the amelioration of intracellular Ca²⁺ mishandling, expanding the therapeutic spectrum of these compounds.
ACKNOWLEDGMENTS The technical assistance of Laura Martín-Nunes, Lorena Lizeth Compean, Monica Martin-Belinchón, and Lucía Guerrero-López is gratefully acknowledged. The authors thank Dr Kenneth McCreath for manuscript editing. The authors thank the “Biobanco A Coruña” (Instituto de Investigación Biomédica de A Coruña, Spain) for providing tissue samples.
FUNDING SUPPORT AND AUTHOR DISCLOSURES
This work was supported by the Spanish Ministry of Economy and Competitiveness and the European Regional Development Fund (SAF-2017-84777R), Instituto de Salud Carlos III (ISCIII) (PI17/01093, PI17/01344, and PI19/01482), Sociedad Española de Cardiología, Proyecto Translacional 2019 y Asociación del Ritmo Cardiaco (SEC, España), Proyecto Asociación Insuficiencia Cardiaca (Transplante Cardio) 2020, Fondo Europeo de Desarrollo Regional, Fondo Social Europeo, and CIBERCV, a network funded by ISCIII, Spanish Ministry of Science, Innovation and Universities (PGC2018-097019-B-I00), Ministerio de Economía, Industria y Competitividad/Agencia Estatal de Investigación 10.13039/501100011033 PID2020-112388R-B-100, PID2019-105600RB-I00, the Instituto de Salud Carlos III (Fondo de Investigación Sanitaria grant PR93 [PI17/0019/0003-ISCIII-SGEM/ERDF, ProteoRed]), and “la Caixa” Foundation (project code HHR–00347). The Centro Nacional de Investigaciones Cardiovasculares is supported by the ISCIII, the Ministerio de Ciencia, Innovación y Universidades. Dr Ruiz-Hurtado is Miguel Servet I researcher of ISCIII (CP15/00129 Carlos III Health Institute). Dr Tamayo and R.I. Jaén, and M. Gil-Fernández were or currently are PhD students funded by the Formación de Profesorado Universitario program of the Spanish Ministry of Science, Innovation and Universities (FPUE/17/06135; FPUE/18/08827, FPUE/19/01973). The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
ADDRESS FOR CORRESPONDENCE: Dr María Fernández-Velasco, Instituto de Investigación Hospital la Paz, IDI-PAZ, Paseo de la Castellana 261, 28046 Madrid, Spain. E-mail: [email protected] OR [email protected]. OR Dr Patricia Prieto, Facultad de Farmacia, Universidad Complutense de Madrid, Plaza de Ramón y Cajal s/n, 28040 Madrid, Spain. E-mail: [email protected]. OR Dr Lisardo Bosca, Instituto de Investigaciones Biomédicas Alberto Sols (CSIC-UAM), Arturo Duperier 4, 28029 Madrid, Spain. E-mail: [email protected].
REFERENCES
1. Kindermann I, Barth C, Mahfoud F, et al. Update on myocarditis. J Am Coll Cardiol. 2012;59(9):779-792.
2. Hendren NS, Dzavner MH, Bozkurt B, Cooper LT. Description and proposed management of the acute COVID-19 cardiovascular syndrome. Circulation. 2020;141(23):1903-1914.
3. Biesbroek PS, Beek AM, Germans T, Niessen HWM, Van Rossum AC. Diagnosis of myocarditis: current state and future perspectives. Int J Cardiol. 2015;191:211-219.
4. Ruiz-Hurtado G, Li I, Fernández-Velasco M, et al. Reconiling depresed Ca²⁺ sparks occurrence with enhanced RyR2 activity in failing mice cardiomyocytes. J Gen Physiol. 2015;146(4):295-306.
5. Bers DM. Cardiac excitation-contraction coupling. Nature. 2002;415(6868):198-205.
6. Wahlquist C, Jeong D, Rojas-Muñoz A, et al. Inhibition of miR-25 improves cardiac contractility in the failing heart. Nature. 2014;508(7497):531-536.
7. Serhan CN, Savill J. Resolution of inflammation: the beginning programs the end. Nat Immunol. 2005;6(12):1191-1197.
8. Jaén RI, Fernández-Velasco M, Terrón V, et al. BML-111 treatment prevents cardiac apoptosis and oxidative stress in a mouse model of autoimmune myocarditis. FASEB J. 2020;34(8):10531-10546.
9. Lee TH, Lympay P, Crea AE, Spur BW. Inhibition of leukotriene B₄-induced neutrophil migration by lipoxin A₄: structure-function relationships. Biochem Biophys Res Commun. 1991;180(3):1416-1421.
10. Cuadrado A, Rojo Al, Wells G, et al. Therapeutic targeting of the NRF2 and KEAP1 partnership in chronic diseases. Nat Rev Drug Discov. 2019;18(4):295-317.
11. da Costa RM, Rodrigues D, Pereira CA, et al. Nrf2 as a potential mediator of cardiovascular risk in metabolic diseases. Front Pharmacol. 2019;10:382.
12. Val-Blasco A, Piedras MUJM, Ruiz-Hurtado G, et al. Role of NQO1 in heart failure progression via regulation of Ca²⁺ handling. J Am Coll Cardiol. 2017;69(4):423-433.
13. Suraweera TL, Vasantha Rupasinghe HP, Delaire G, Xu Z. Regulation of Nrf2/ARE pathway by dietary flavonoids: A friend or foe for cancer patients with myocarditis?
Role of Proresolving Mediators in Myocarditis-induced Ca\textsuperscript{2+} Mishandling
JACC: BASIC TO TRANSATIONAL SCIENCE VOL. 7, NO. 6, 2022
JUNE 2022:544–560
Transition of pro-resolving mediators on calcium handling? Antioxidants (Basel). 2020;9(10):973.
14. McCarthy RE, Bohrer JP, Hruban RH, et al. Long-term outcome of fulminant myocarditis as compared with acute (nonfulminant) myocarditis. N Engl J Med. 2000;342(10):690–695.
15. Kain V, Ingle KA, Colas RA, et al. Resolvin D1 activates the inflammation resolving response at splenic and ventricular site following myocardial infarction leading to improved ventricular function. J Mol Cell Cardiol. 2015;82:24–35.
16. Fernández-Velasco M, Rueda A, Rizzi N, et al. Increased Ca\textsuperscript{2+} sensitivity of the rydonine receptor mutant RyR2R4496C underlies catecholaminergic polymorphic ventricular tachycardia. Circ Res. 2009;104(2):201–209. 12 pages following 209.
17. Petri MH, Laguna-Fernández A, González-Diez M, Paulsson-Berne G, Hansson GK, Bäck M. The role of the FPR2/ALX receptor in atherosclerotic disease and plaque stability. Cardiovasc Res. 2015;105(1):65–74.
18. Zeebo K, Yaroshinsky A, Rudy JJ, et al. Long-term effects of AAV1/SERCA2a gene transfer in patients with severe heart failure: analysis of recent cardiovascular events and mortality. Circ Res. 2014;114(1):101–108.
19. Greenberg B, Butler J, Felker GM, et al. Calcium upregulation by percutaneous administration of gene therapy in patients with cardiac disease (CUPID 2): a randomised, multinational, double-blind, placebo-controlled, phase 2b trial. Lancet. 2016;387(10024):1178–1186.
20. Miyamoto MI, Del Monte F, Schmidt U, et al. Adenosinergic gene transfer of SERCA2a improves left-ventricular function in aortic-banded rats in transition to heart failure. Proc Natl Acad Sci U S A. 2000;97(2):793–798.
21. Byrne MJ, Power JM, Prevolos A, Mariani JA, Hajjar RJ, Kaye DM. Recirculating cardiac delivery of AAV2/SERCA2a improves myocardial function in an experimental model of heart failure in large animals. Gene Ther. 2008;15(23):1550–1557.
22. Hayward C, Banner NR, Morley-Smith A, Lyon AR, Harding SE. The current and future landscape of SERCA gene therapy for heart failure: a clinical perspective. Hum Gene Ther. 2015;26(5):293–304.
23. Vashi R, Patel BM. NRF2 in cardiovascular diseases: a ray of hope. J Cardiovasc Transl Res. 2021;14(3):573–586.
24. Kang GJ, Kim EJ, Lee CH. Therapeutic effects of specialized pro-resolving lipids mediators on cardiac fibrosis via NRF2 activation. Antioxidants (Basel). 2020;9(12):1259.
25. Milanesi E, Manda G, Dobre M, et al. Distinctive under-expression profile of inflammatory and redox genes in the blood of elderly patients with cardiovascular disease. J Inflamm Res. 2021;14:429–442.
26. Eriksen R, Kramer CM, Lüktstädt W, et al. Left ventricular diastolic dysfunction in Nrfl2 knockout mice is associated with cardiac hypertrophy, decreased expression of SERCA2a, and preserved endothelial function. Free Radic Biol Med. 2015;89:906–917.
27. Li J, Ichikawa T, Villacorta L, et al. Nrfl2 protects against maladaptive cardiac responses to hemodynamic stress. Arterioscler Thromb Vasc Biol. 2009;29(11):1843–1850.
28. Ge C, Hu L, Lou D, et al. Nrfl2 deficiency aggravates PM2.5-induced cardiomyopathy by enhancing oxidative stress, fibrosis and inflammation via RIPK3-regulated mitochondrial disorder. Aging. 2020;12(6):4836–4865.
29. Bai T, Hu X, Zheng Y, Wang S, Kong J, Cai L. Resveratrol protects against lipopolysaccharide-induced cardiac dysfunction by enhancing SERCA2a activity through promoting the phospholamban oligomerization. Am J Physiol Heart Circ Physiol. 2016;311(4):H1051–H1062.
30. Kawakami R, Sakamoto A, Kawai K, et al. Pathological evidence for SARS-CoV-2 as a cause of myocarditis: JACC Review Topic of the Week. J Am Coll Cardiol. 2021;77(3):314–325.
31. Fried JA, Ramasubbu K, Bhatt R, et al. The variety of cardiovascular presentations of COVID-19. Circulation. 2020;141(23):1930–1936.
32. Schwarz B, Sharma L, Roberts L, et al. Severe SARS-CoV-2 infection in humans is defined by a shift in the serum lipidome resulting in dysregulation of eicosanoid immune mediators. J Immunol. 2021;206(2):329–334.
33. Szklarczyk D, Gable AL, Lyon D, et al. STRING v11: Protein-protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets. Nucleic Acids Res. 2019;47(D1):D607–D613.
KEY WORDS calcium handling, myocarditis, NRF2, pro-resolving mediators, SERCA2a
APPENDIX For supplemental methods, results, figures, and tables, please see the online version of this paper. | 2025-03-05T00:00:00 | olmocr | {
"pdf_page_numbers": [
[
0,
1197,
1
],
[
1197,
3520,
2
],
[
3520,
8367,
3
],
[
8367,
9991,
4
],
[
9991,
14166,
5
],
[
14166,
14578,
6
],
[
14578,
16448,
7
],
[
16448,
16960,
8
],
[
16960,
18017,
9
],
[
18017,
23292,
10
],
[
23292,
28603,
11
],
[
28603,
29946,
12
],
[
29946,
31706,
13
],
[
31706,
36882,
14
],
[
36882,
42276,
15
],
[
42276,
47303,
16
],
[
47303,
51966,
17
]
]
} |
818a9563fbe96085e53d3431c571ba94cd749c44 | {
"olmocr-version": "0.1.59",
"pdf-total-pages": 13,
"total-fallback-pages": 0,
"total-input-tokens": 28396,
"total-output-tokens": 10101,
"fieldofstudy": [
"Political Science"
]
} | IT Governance Maturity for Uganda’s Higher Institutions of Learning
Lillian Ndagire, Kyambogo University, Uganda*
Gilbert Maiga, Makerere University, Kampala, Uganda
Benedict Oyo, Gulu University, Uganda
ABSTRACT
The daily application of information technology (IT) in the public sector organizations and its positioning as a critical driver for economic growth require a focus on implementing IT governance. Despite the increasing application of IT, there is limited academic systematic research regarding the maturity of IT governance and process improvement in higher institutions of learning (HILs). IT in HILs is highly complex, and managing complex systems of processes and platforms in HILs necessitates tools to assess and provide guidelines for integrating organizational processes. This research, therefore, evaluated the IT governance processes in HILs in Uganda and established the maturity level to ensure continuous improvement and organizational maturity. Eight HILs in Uganda were measured using 15 IT processes of the COBIT framework and rated with the generic maturity model. Data were analyzed using MS Excel. Results indicated IT governance maturity level of HILs in Uganda was at Level 2 (repeatable).
KEYWORDS
COBIT, EIT Governance Evaluation, Generic Maturity Model, Higher Institutions of Learning, Information Technology, IT Governance, IT Governance Maturity, IT Processes
INTRODUCTION
The day-to-day use of IT in organizations has led to increased investment in IT systems (Adaba & Rusu, 2014). NITA-U (2018) notes that IT usage in public sector organizations enhances effective and efficient public service delivery. For HILs, IT enables automated access to educational services through IT platforms for academic and management functions (Montenegro & Flores, 2015). The IT systems in HILs are complex and diverse, consisting of a diverse set of technologies involving various applications, platforms, educational systems, and cloud applications to support their teaching, learning, research, and administrative processes (Bianchi et al., 2017). Hence, managing IT systems in HILs requires a focus on IT governance (Laita & Belaisaoui, 2017; Nyeko et al., 2018).
IT governance is a central portion of an organization’s governance consisting of leadership, organizational structures, and processes to enable the IT in organizations to sustain and extend the strategies and objectives of the organization (ITGI, 2003). According to Weill and Ross (2004), IT
governance involves applying IT processes that enable IT activities to be aligned with the organization’s mission, strategy, and objectives. Numerous standards on IT governance exist. The prominent ones include Information Technology Infrastructure Library that describes practices for managing IT services (Zhang et al., 2013); Control Objectives for Information and related Technology (COBIT) that describes policies and practices for control of IT and security (ISACA, 2004); and ISO 38500 for IT corporate governance (OGC, 2008). However, the COBIT framework is adopted in this study due to its wide acceptance in IT governance practice (ITGI & PwC, 2008).
RELATED LITERATURE
IT in Higher Institutions of Learning
Eckel and King (2004) note that higher institutions consist of post-secondary education, research guidance, and training conducted at institutions like universities licensed by state authorities as educational institutions. Yudatama et al. (2017) further state that HILs are supposed to be non-profit entities encompassing academic and administration sections. The administration section supports the academic area for the smooth running of the institution considering education as the primary business (Yudatama et al., 2017). According to Forest and Kinser (2002), HILs carry out teaching, exacting applied work research, and social services. IT facilitates dissemination of knowledge, supports and improves academic activities, and enables sharing educational content (Yasemin et al., 2008; Lockyer et al., 2001). Additional, through the proper use of IT systems, disadvantaged groups of people can be reached (Toro & Joshi, 2012).
IT Governance Maturity
IT governance maturity is a measure that entails a collection of IT capabilities that concerns what the IT department provides for the business (Axios, 2018) by indicating the progress of an organization over time (IGI Global, 2021). Microsoft IF&SRT (2009) further states that a maturity model as a measure designates and evaluates the practices and processes that let an organization achieve dependable and supportable results. In addition, Axios (2018) asserts that a mature IT organization is pertinent to business goals, competently functions, and can rapidly change as business requirements change. In contrast, a low maturity IT organization may incline to provide wrong things at an unacceptable cost and fail to change as business requirements change.
For HILs, a maturity model works as a point of reference to understand the reality of the situation institutions must follow to realize excellence by defining the path to undertake and providing quality mechanisms at each maturity level (Carvalho et al., 2018). Studies on IT governance maturity exist, such as a comparative study of maturity models of different subareas of education was identified and categorized in HILs (Carvalho et al., 2018). Tocto-Cano et al. (2020) present a method that detects gaps in existing maturity models for universities since they do not indicate their whole dimensions. Duarte and Martins (2013) provide an extension to a business process maturity model for HILs. Đurek and Redep (2016) proposed a method to prioritize elements in the digital maturity framework for HILs and assessing the digital maturity level of HILs in Croatia.
Challenges of IT Governance Maturity in HILs
HILs are unique institutions whose technological infrastructure comprises heterogeneous technologies such as diversity of applications, educational systems, cloud applications, different platforms (Duarte & Martins, 2013). In addition, HILs have a large spectrum of platforms like student relationship management, learning management systems, survey tools, and business intelligence (Carvalho et al., 2018; Duarte & Martins, 2013). Hence, HILs must have a cohesive strategy that is capable of supporting their transverse processes.
Besides, HILs lack standard academic management processes (Carvalho et al., 2018). Each HILs follows its internal procedures, which become an obstacle to adopt standard software packages. However, some commercial and open-source products have been developed, such as Moodle (Carvalho et al., 2018), and initiatives for interoperability of processes in HILs (Ribeiro et al., 2016) make it evident to alter this situation.
Managing such complex processes and platforms in HILs necessitates tools to assess and provide guidelines for integrating organizational processes and information systems. Thus, an IT governance maturity model using 15 processes of COBIT (Guldentops et al., 2002; ISACA, 2003) and scored using the generic maturity model (ITGI, 2003: 2007) was used to assess the maturity level of IT governance in HILs in Uganda.
Theoretical Framework
Following an earlier paragraph, the COBIT framework was created by IT Governance Institute, Information Systems Audit and Control Association (ITGI, 2008). It has three parts. Namely: Criteria for information, IT resources, and processes for IT (COBIT, 2007). IT processes have four domains (COBIT, 2007) with 34 cases of IT (Devos & Van de Ginste, 2015). The four domains involve; acquire and implement, deliver and support, plan and organize, and monitor and evaluate (COBIT, 2007). Furthermore, COBIT consists of management guidelines for assessing, implementing, and improving IT management consistent with organizational business goals (Van Grembergen & De Haes, 2008).
One of the evaluation tools is the generic maturity model. The maturity model provides guidelines undertaken by management in organizations to do self-evaluation through measuring the level of management processes for 34 IT processes of COBIT (ITGI, 2003). According to Nfuka and Rusu (2010), maturity level ranges from 0 to 5, showing the state for every IT process and what should be in place to realize a higher level. Also, ITGI (2008) notes that maturity level shows IT governance in an organization and compares it with other geographical locations to develop strategies for improvement. The generic maturity model, as highlighted above, is given in Table 1.
Several evaluations on IT governance maturity have been conducted. For example, the IT governance maturity level evaluates Swedish electric utilities’ support systems and administrative processes (Simonsson et al., 2007). Establishing a governance maturity reference benchmark for the public and not-for-profit organizations (Guldentops et al., 2002; Liu & Ridley, 2005). The study on maturity level and its effects on IT governance for public sector organizations in Australia (Liu & Ridley, 2005). Conversely, Yanosky and Caruso (2008) present a poorly moderated maturity level of IT governance for HILs. While these studies underscore the importance of IT governance in ensuring the successful adoption of IT systems, many public sector organizations in developing countries like Uganda are yet to streamline IT governance.
Table 1. Generic maturity model (Source: ITGI (2003; 2007)
| Level | Process Description |
|-------|---------------------|
| 0 Non-existent | There are no identifiable processes, and the organization has not identified problems |
| 1 Initial | The organization identifies problems, but there is a lack of standardized procedures. Instead, there are ad-hoc approaches used. |
| 2 Repeatable | Processes apply comparable procedures used by various people responsible for a similar job; however, official communication of ordinary practices is limited, which may result in errors |
| 3 Defined Process | Organizations follow formalized and documented procedures, but deviations are likely to occur, and methods may not be sophisticated |
| 4 Managed | Monitoring and measuring of compliance of procedures are followed by management and acts when defaulted. Processes steadily progress, and system automation is limited |
| 5 Optimized | Processes are of good practice evaluated against results of constant development with other organizations. IT is incorporated to automate workflows and enhance quality |
Ugandan Context
IT governance maturity is still low in public sector organizations in Uganda. For example, a survey on “IT performance for public sector organizations” indicated a low maturity level of IT governance (NITA-U, 2018). The survey shows that the ICT technical committee is at 29.9%, the ICT steering committee is at 28.6%, and 53.9% were positioned at the unit level and 33.9% at the department level. Also, knowledge, IT resources, and culture limitations for developing countries (Ndou, 2004; Bakari, 2007) express a need to evaluate IT governance maturity. Evaluation of IT governance maturity level shows the condition of IT processes and what is required to sustain and increase IT governance in such changing environment (Amanat, 2018; Nfuka, 2012).
Therefore, this paper evaluates the IT governance maturity level in eight HILs in Uganda. The study analyses 15 COBIT processes scored with the generic maturity model (ITGI, 2003; 2007) (see table 2) as in earlier related studies (Amanat, 2018; Nfuka & Rusu, 2010). Furthermore, a comparison of IT governance maturity level in HILs in Uganda was done with selected public sector organizations in Pakistan (Amanat, 2018) and Australia (Liu & Ridley, 2005) as well as with an international range of nations (ISACA, 2003).
RESEARCH METHODOLOGY
Empirical Source
Eight public degree-awarding HILs in Uganda were selected to evaluate IT governance maturity. HILs were chosen because of their high dependency on IT for teaching, learning, research, administrative processes, and community outreach. Given that the selected HILs are examples of public sector organizations of a developing country (Uganda), the attained state of maturity on how they plan, implement, support, and monitor IT signifies IT processes’ relative maturity in organizations with similar set-up and environment. HILs were: Gulu University (GU) which was established after appointing a technical task force in 2001 to set up the institution (Gulu University, 2018). Busitema University (BUS) was established to improve equitable access to university education in the eastern part of the country (Busitema University, 2019). Makerere University (MUK), the largest and oldest HIL, was first established as a technical school in 1922 (Roach, 2011). Lira University (LU) was the first public institution teaching hospital in Uganda established to train students in health sciences.
Table 2. COBIT IT processes (Guldentops et al., 2002; ISACA, 2003)
| COBIT domain | IT processes |
|--------------------|------------------------------------------------------------------------------|
| Plan and Organize (PO) | PO1 - Define a strategic IT plan |
| | PO3 - Determine the technological direction |
| | PO4 - Define the IT processes, organization, and relationships |
| | PO5 - Manage the IT investment |
| | PO6 - Communicate management aims and direction |
| | PO9 - Assess and manage IT risks |
| | PO10 - Manage projects |
| Deliver and Support (DS) | DS1 - Define and manage service levels |
| | DS4 - Ensure continuous service |
| | DS5 - Ensure systems security |
| | DS11 - Manage data |
| Acquire and Implement (AI) | A11 - Identify automated solutions |
| | A12 - Acquire and maintain application software |
| | A16 - Manage changes |
| Monitor and Evaluate (ME) | ME1 - Monitor and evaluate IT performance |
Soroti University (SU) was the newest degree-awarding institution established by the Government of Uganda after lobbying by stakeholders from the Teso sub-region in the Soroti district (Beinomugisha, 2015). Kyambogo University (KYU) was established in 2003 after a merger of Uganda Polytechnic Kyambogo, Institute of Teacher Education Kyambogo, and Uganda National Institute of Special Education (Cula, 2005). Kabale University (KAB) donates to the development of Kigezi and Africa at large through service delivery, research, and training (Kushaba, 2012). Mbarara University of Science and Technology (MUST) was established in the former Nursing and Midwifery School to cover the gap of health professionals in the country (MUST, 2019).
An exploratory survey was conducted between September and October 2020. A questionnaire consisting of 6 Likert scales ranging from 0 to 5 was administered to 51 participants to ascertain their agreement on established, formalized, and documented IT processes. The development of the questionnaire was based on a maturity measurement tool by ITGI (2007) that was customized to fit the studied environmental context and the 15 IT processes of COBIT.
A total of 51 persons participated in the survey involving IT and business representatives, specifically IT and Business Directors/Managers as in earlier related studies (Amanat, 2018; Nfuka & Rusu, 2010; ISACA, 2003). The selection of participants was based on their role in IT leadership and decision-making.
**Research Process**
The researcher informed participants of the purpose of the survey. Documents were gathered by Chief Information Officers (CIOs) in HILs, which were availed to the researcher. Hence, documents were used to verify data, thus adding credibility (Yin, 2003) to the rated maturity level. Such documents included IT policies, procedures, strategies, structures, plans, performance reports, and meeting minutes. Whenever a higher maturity level was quoted and not supported by an accompanying document, the researcher noted the document, clarified it, and discussed the document with the participants. Participants were requested to indicate the actual score depending on formalized IT processes. Most participants understood the IT processes, although some confused them at the beginning. The data collected was analyzed using MS Excel software and represented graphically using charts (Manikandan, 2011).
**RESULTS**
This section presents results of IT governance maturity in eight public degree-awarding HILs in Uganda. The comparative maturity level study (Carvalho et al., 2018; Amanat, 2018) for IT processes, domain level, IT processes at the institutional level, and country (economic) status is as follows.
**IT Governance Maturity Level for HILs in Uganda**
The average maturity level across the eight HILs in Uganda was 2.72 (see table 4) with a range between 0.79 and 3.91 and 80% (12 over 15) IT processes were greater than 2.00 compared to 20%
| HILs Respondents | GU | BUS | MAK | LU | SU | KYU | KAB | MUST | Total |
|------------------|----|-----|-----|----|----|-----|-----|------|-------|
| Directors/ managers of IT | 2 | 4 | 3 | 3 | 2 | 2 | 3 | 2 | 21 |
| Directors/ managers of business | 4 | 3 | 5 | 3 | 4 | 4 | 3 | 4 | 30 |
| Total | 6 | 7 | 8 | 6 | 6 | 6 | 6 | 6 | 51 |
IT processes were lower than 2.00. This showed that the lower end was at an initial stage (level 1), whereas the higher level was at the defined process stage (level 3). On average, processes apply comparable procedures used by various people responsible for a similar job; however, official communication of standard practices is limited, resulting in errors. Results showed that issues need to be addressed to increase the IT governance maturity level.
**Comparison of IT Processes**
IT processes in HILs in Uganda were compared, as shown in Figure 1. It was observed that some IT processes performed reasonably well, and others performed poorly.
IT processes that reasonably performed include the following:
1. **DS1**: Define and manage service levels with a maturity level of 3.91. A possible reason could be that all HILs needed to follow procurement and disposal procedures as stated in Public Procurement and Disposal of Public Assets Act 1 of 2003 (PPDA, n.d.). No HIL receives a service without a formally signed contract between the institution and the service provider.
2. **PO1**: Define a strategic IT plan with a maturity level of 3.81. This score is that all HILs in Uganda had an IT strategic plan which was included in the master plan of the institution. Secondly, e-learning is an emphasized mode of remote teaching and learning in institutions due to the COVID-19 pandemic. NCHE set guidelines for e-learning in HILs across the country (Daily Mornitor, 2020); hence, IT cannot be avoided since it’s an enabler for this activity and its direction, as indicated in the strategic plan.
3. **DS11**: Manage data with a maturity level of 3.23. This might be due to the massive use of IT systems (such as human resource management systems, Integrated Financial Management, and System Academic Information Management System) developed externally and internally to store and reserve institutional academic and administrative data.
These findings were consistent with the previous study of a developing country (Amanat, 2018; Nfuka & Rusu, 2010).
In contrast, the IT processes that performed poorly include:
1. **PO9:** Assess IT risks with a maturity level of 0.79. The low performance could be associated with the fact that institutions use ad hoc approaches on IT risks that hinge on a system-to-system basis. In addition, institutions did not hold a budget for IT risk management due to the limited funds since IT risk management was not considered a priority. These findings are consistent with previous authors of developing countries (Amanat, 2018; Nfuka & Rusu, 2010).
2. **DS4:** Ensure continuous service with a maturity level of 0.85. The reason for the low score could be that institutions had sustainability and maintenance challenges of IT systems. Some IT systems were donor-funded, and their management was challenged when donor funds ended. For example, the Blackboard learning management system was a project for learning and teaching at Makerere University (Abigail, 2018), whose management was terminated because of high license costs (Ssekakubo et al., 2011).
**Domain-Level Comparison**
A comparison of domains of 15 IT processes in eight HILs in Uganda was made. Results in figure 2 show variations in scores for the different domains.
According to Figure 2, it is observed that domains scored different average maturity levels (figure 5.2). Domain Plan and Organise scored highest at 2.87. This is consistent with previous scholars (Nfuka & Rusu, 2010; Guldentops et al., 2002) who had a high average maturity level for plan and organize domain. The high average maturity level could be that before IT systems are introduced in HILs, they should be desired, and plans should be placed before implementation. For instance, due to the COVID-19 pandemic, all HILs must have e-learning facilities before conducting online lessons.
On the other hand, domain Deliver and Support had the lowest average maturity level at 2.44. The low score could be due to the high failure rate of IT systems that do not perform to the users’ expectations. Also, the IT systems functions may not be aligned with the business goals. But, again, this is inconsistent with previous studies (Nfuka & Rusu, 2010; Guldentops et al., 2002), indicating higher maturity in the delivery and support domain due to policies and regulations.
Comparison of Maturity Level of IT Process at Institutional Level
A comparison for the maturity level for each of the 8 HIL in Uganda was carried out (see Figures 3 and 4). MUK scored the highest of 3.4. The possible reason could be that MUK was the oldest degree-awarding HIL in Uganda, having reasonably implemented IT governance mechanisms. Process DS5 - Ensure systems security scored highest of 4.9. This meant that MUK is sensitive to its information and applies reasonable security procedures to protect its data. Followed by AI2 - Acquire and Maintain Application Software with an average maturity level of 4.8. The possible reason for this score may be developing the various IT systems for both academic and administrative activities. The lowest IT process was AI1 - Identify automated solutions of 0.7 followed by AI6 - Manage changes of 1.4.
The second HIL with a high maturity level was KYU of 3.1. The possible reason for the high maturity level may be that KYU is at its best at operationalizing most of its IT functions and affiliated tertiary institutions. IT processes PO1 - Define a strategic IT plan and PO3 - Determine technological direction had the highest IT maturity level of 4.4. IT process PO9 - Assess IT risks and DS4 - Ensure continuous service had the lowest average maturity level of 0.7.
SU was the lowest of 2.3. The possible explanation for the low score could be that SU was the newest institution functional for two years where IT systems were being set up and at initial stages. The highest IT process was AI1 - Identify automated solutions of 4.5. Followed by DS1 - Define and manage service levels of 4.3. The lowest IT processes were: DS4 - Ensure continuous service of 0.4 and PO9 - Assess IT risks of 0.4.
Comparison With Developing Country, Developed Country, and Internationally
The scored average maturity level was compared with similar studies of selected public sector organizations in Pakistan (developing country), Australia (developed country), and international with various nations. The average maturity level for 15 IT processes was already indicated in previous research.
For Pakistan in Ali (2018), for Australia in Liu and Ridley (2005), and internationally across a range of nations in Guldentops et al. (2002). The data for studied Australian public sector organizations was published in 2005 and the international public sector benchmark in 2002; their numerical comparison could not be found (Amanat2018); hence, they could be compared in the chart. However, the comparison between the studied HILs and public sector organizations in Pakistan is in figure 5.
Analysis showed that the average maturity level for HILs in Uganda was higher than Pakistan but lower than Australia and intentionally from various nations. The maturity level of IT processes of selected public sector organizations in Pakistan ranged from 1.6 to 3.1, with the majority above the maturity level of 2.0 (60% or 9 out of 15), falling between 2.1 and 2.7. The maturity level of IT processes for the public sector in Australia ranged from 2.5 to 3.5, with the majority above the
maturity level of 3 (60% or 9 out of 15) falling between 3 and 3.5. The maturity level of international public sector organizations ranged from 2 to 3, with the majority above 2.5 (87% or 13 out of 15) ranging between 2.5 and 3.0. Based on the generic maturity model, this indicated that IT processes for developed countries (such as Australia) and international are relatively well defined with standardized and documented measures (ITGI, 2000). This is comparable with the public sector organizations of developing countries such as Uganda that prompted to learn from them and improve.
Concerning the public sector organizations of developing countries such as Uganda prompted to learn from them and improve. IT process DS11-Manage Data performed well, like the studied public sector organizations in Australia. This meant that studied HILs in Uganda were equally well in managing data and related activities. Also, the studied HILs in Uganda should learn from the studied public sector organizations in Australia and improve on IT processes that performed low. Besides, most IT processes in the studied HILs in Uganda performed lower than the studied international public sector benchmark. This showed that studied HILs in Uganda were still lower than the studied international public sector benchmark. This could be caused by the time range between the two studies. Hence, the need to improve lower-performed IT processes in studied HILs in Uganda. Moreover, most of the IT processes in the studied HILs in Uganda performed better than Pakistan’s selected public sector organizations. This could be caused by the different contexts in which the organizations are working.
CONCLUSION AND FURTHER RESEARCH
The study sought to determine the IT governance maturity level for HILs in Uganda. This was attained by assessing the maturity of IT processes for eight HILs in Uganda as a developing country and compared with developing country (Pakistan), developed country (Australia), and internationally using various nations to benchmark for learning and embracing best practices. Analysis showed IT governance maturity level for eight HILs in Uganda was 2.72 (level 2: repeatable). Indicating that processes apply comparable procedures used by various people responsible for a similar job; however, official communication of standard practices is limited, resulting in errors. The studied HILs in Uganda performed better than 6 studied public sector organizations in Pakistan but lower than those in Australia and internationally. Most importantly, the study provided a reference benchmark of the maturity level of IT processes in HILs in Uganda previously unexplored.
More research on evaluating the maturity of IT processes in other public sector and private sector organizations in Uganda whose maturity level is unknown is still required. This will enable identifying the IT governance gaps and setting plans towards achieving the desired level of strategic alignment, IT governance maturity, and IT governance.
ACKNOWLEDGMENT
This study was funded in part by the Swedish International Development Cooperation Agency and Makerere University.
FUNDING AGENCY
Publisher has waived the Open Access publishing fee.
REFERENCES
Abigail, I. (2018). A Report on the Status of Blackboard Courses from 2002 to November 2008. https://www.oerafrica.org/resource/report-status-blackboard-courses-2002-november-2008
Adaba, G. B., & Rusu, L. (2014). IT Governance practices in a public organization in Ghana. International Journal of Innovation in the Digital Economy, 5(2), 14–23. doi:10.4018/ijide.2014040102
Ali, A. (2018). IT Governance Implementation Framework for Public Sector Organizations of Pakistan. University of Engineering & Technology, Taxila. http://prr.hec.gov.pk/jspui/handle/123456789/11309
Axios, S. (2018). IT Maturity Matters Because IT Matters. https://www.thinkhdi.com/library/supportworld/2018/it-maturity-matters-because-it-matters.aspx
Bakari, J. B. (2007). A Holistic Approach for Managing ICT Security in Non-Commercial Organizations: A case in a developing country. Stockholm University.
Beinomugisha, G. (2015). Soroti University key for the 2016 general election in Teso. https://www.sun.ac.ug/about-us/university-background/
Bianchi, I., Sousa, R., Pereira, R., & Hillegersberg, J. (2017). Baseline mechanisms for IT governance at universities. Proceedings of the 25th European Conference on Information Systems (ECIS), 1551–1567.
Busitema University. (2019). Historical Background. https://busitema.ac.ug/about/our-history/
Carvalho, J. V., Pereira, R. H., & Rocha, Á. (2018). A comparative study on maturity models for information systems in higher education institutions. The 2018 International Conference on Digital Science, 150–158.
COBIT. (2007). Information Systems Audit and Control Association, Control Objectives for Information and Related Technology, 4 (1st ed.). IT Governance Institute.
Cula, A. (2005). Kyambogo University: Establishment of the University. The Uganda Higher Education Review, 2(2), 23–26.
Devos, J., & Van de Ginste, K. (2015). Towards a Theoretical Foundation of IT Governance – The COBIT 5 case. The Electronic Journal Information Systems Evaluation, 18(2), 95–103.
Duarte, D., & Martins, P. V. (2013). A maturity model for higher education institutions. Journal of Spatial and Organizational Dynamics, 1(1), 25–44.
Đurek, V., & Redep, N. B. (2016). Review on e-readiness assessment tools. Cent. Eur. Conf. Inf. Intell. Syst., 161–169.
Eckel, P. D., & King, J. E. (2004). An overview of higher education in the United States: Diversity, access and the role of the marketplace. Academic Press.
Forest, J., & Kinser, K. (2002). Higher Education in the United States: An Encyclopedia. ABC-CLIO.
Guldentops, E., Van Grembergen, W., & De Haes, S. (2002). Control and governance maturity survey: Establishing a reference benchmark and a self-assessment tool. Information Systems Control Journal, 6, 32–35.
Gulu University. (2018). Historical Background of Gulu University. https://gu.ac.ug/our-history/
IGI Global. (2021). What is Maturity Model. https://www.igi-global.com/dictionary/maturity-metrics-health-organizations-information/18047
ISACA. (2003). Further material on IT governance maturity. ISACA.
ISACA. (2004). COBIT Student Book. IT Governance Institute.
ITGI. (2000). Control Objectives for Information and Related Technologies. ITGI.
ITGI. (2003). Board briefing on IT governance (2nd ed.). ITGI.
ITGI. (2007). IT Governance Institute. COBIT 4.1.
ITGI. (2008). IT Governance global status report. IT Governance Institute.
ITGI & PwC. (2008). *IT Governance Global Status Report*. www.itgi.org
Kushaba, A. (2012). Museveni Endorses Take over of Kabale University. Uganda Radio Network. https://www.kab.ac.ug/about-kab/history-facts/background/
Mornitor, D. (2020). NCHE issues guidelines for university e-learning. https://www.monitor.co.ug/uganda/news/national/nche-issues-guidelines-for-university-e-learning-1898632
Laita, A., & Belaissaoui, M. (2017). Information technology governance in public sector organizations. *Europe and MENA Cooperation Advances in Information and Communication Technologies*, 331–340.
Lira University. (2020). About Lira University. https://lirauni.ac.ug/about-lira-university/
Liu, Q., & Ridley, G. (2005). IT Control in the Australian public sector: an international comparison. *Proceedings of the Thirteenth European Conference on Information Systems*.
Lockyer, L., Patterson, J., & Harper, B. (2001). ICT in higher education: Evaluating outcomes for health education. *Journal of Computer Assisted Learning*, 17(3), 275–283. doi:10.1046/j.0266-4909.2001.00182.x
Manikandan, S. (2011). Measures of central tendency: Median and mode. *Journal of Pharmacology & Pharmacotherapeutics*, 2(3), 214–215. doi:10.4103/0976-500X.83300 PMID:21897729
Microsoft IF&SRT. (2009). *Microsoft Innovation framework & self-reflection tool*. http://www.istoolkit.com/self_reflection.html
Montenegro, C. W., & Flores, D. A. (2015). An integrated model for ICT governance and management applied to the council for evaluation, accreditation and quality assurance of higher education institutions in Ecuador (CEAACES). *2015 International Conference on Computing, Communication and Security (ICCCS)*, 1–9. doi:10.1109/CSSCS.2015.7374158
MUST. (2019). About MUST. https://www.must.ac.ug/about-must/our-story/
Ndou, V. (2004). E – Government for Developing Countries: Opportunities and Challenges. *The Electronic Journal on Information Systems in Developing Countries*, 18(1), 1–24. doi:10.1002/j.1681-4835.2004.tb00117.x
Nfuka, E. N. (2012). *IT Governance in Tanzanian public sector organizations*. Stockholm University.
Nfuka, E. N., & Rusu, L. (2010). IT governance maturity in the public sector organizations in a developing country: The case of Tanzania. *Proceedings of the Sixteenth Americas Conference on Information Systems*, 1–12.
NITA-U. (2018). *NITA-U Strategic Plan 2018/19 – 2022/23*. Author.
Nyeko, S., Niwe, M., Moya, M., & Kituyi, G. (2018). *IT Governance Steering Committee, IT Competence and IT Governance Performance in Ugandan Public Universities*. In *2018 IST-Africa Week Conference (IST-Africa)*. IEEE.
OGC. (2008). *Best Management Practice in IT Service Management (ITIL) and Project Management (PRINCE2)*. www.bestmanagement-practice.com
PPDA. (n.d.). *Public Procurement and Disposal of Public Assets [Act 1 of 2003]*. Retrieved November 30, 2020, from https://www.ppda.go.ug/download/ppda_act/ppda_act/PPDA_ACT-2003.pdf
Ribeiro, L. M., Pereira, R. H., Pacheco, O., Bernardes, M., & Martins, R. T. (2016). Interoperability between information systems of Portuguese higher education institutions. *EUNIS 22nd Annual Congress Book of Proceedings*, 203–214.
Roach, P., Setter, J., & Esling, J. (2011). *Cambridge English Pronouncing Dictionary*. Cambridge University Press.
Simonsson, M., Johnson, P., & Wijkström, H. (2007). Model-based IT governance maturity assessments with COBIT. *European Conference on Information Systems*.
Ssekakubo, G., Suleman, H., & Marsden, G. (2011). Issues of adoption: have e-learning management systems fulfilled their potential in developing countries? *Proceedings of the South African Institute of Computer Scientists and Information Tech*. doi:10.1145/2072221.2072248
Tocto-Cano, E., Paz Collado, S., López-Gonzales, J. L., & Turpo-Chaparro, J. E. (2020). A Systematic Review of the Application of Maturity Models in Universities. *Information (Basel), 11*(10), 466. doi:10.3390/info11100466
Toro, U., & Joshi, M. (2012). ICT in higher education: Review of literature from the period 2004-2011. *International Journal of Innovation, Management and Technology, 3*, 2023.
Van Grembergen, W., & De Haes, S. (2008). Strategies and Models for IT Governance. Implementing Information Technology Governance: Models, Practices and Cases.
Weill, P., & Ross, J. (2004). *IT Governance: How Top Performers Manage IT Decision Rights for Superior Results*. Harvard Business Press.
Yanosky, R., & Caruso, J. B. (2008). *Process and politics: IT governance in higher education*. ECAR Key Findings.
Yasemin, , Petek, Kar, & Turgay, (2008). A Structural Equation Model for ICT Usage in Higher Education. *Journal of Educational Technology & Society, 11*, 262–273.
Yin, R. K. (2003). *Case Study Research. Design and Methods. Third Edition. Applied social research method series, Volume 5*. Sage Publications.
Yudatama, U., Nazief, B. A. A., Hidayanto, A. N., & Muhammad, M. (2017). Factors Affecting Awareness and Attitude of IT Governance Implementation in The Higher Education Institution: A Literature Review. *2017 3rd International Conference on Science in Information Technology (ICSIITech)*.
Zhang, Y., Zhang, J., & Chen, J. (2013). Critical success factors in IT service management implementation: People, process, and technology perspectives. *2013 International Conference on Service Sciences (ICSS)*, 64–68.
Ndagire Lillian is an Assistant Lecturer in the Department of Networks, School of Computing and Library Science, Kyambogo University. She is pursuing a PhD in Information Systems in the College of Computing and Information Sciences at Makerere University. The research area is IT governance for public sector organizations of developing countries. She holds a Masters of Information Technology from Makerere University and a Bachelor of Science (Computer Science & Economics). Areas of specialization are database management, operating systems implementation, network configuration, systems analysis, and design. Lillian participates in several activities such as curriculum development for several IT-related programs and supervises research students, both undergraduates and postgraduates.
Gilbert Maiga (PhD) is the current Dean of the School of Computing and informatics Technology at Makerere University. He is an Associate Professor in the Department of Information Technology where he has mentored several Graduate Students researches to completion in the broad field of Information systems. He is also a former chair of the department of Information Technology at Makerere University. He has more than 15 years of teaching and research experience. His main areas of interest for research over are in: Information Systems evaluation, e-services (e-health and e-Governance) systems development and adoption. He has also published work on the use of Ontologies for biomedical data Integration.
Benedict Oyo holds a PhD in Information Systems. He is the current Dean, Faculty of Science at Gulu University. Benedict’s current research focuses on the role of ICT innovations in development. This is reflected by three main areas: Implementation of MOOCs in low bandwidth environments; Application of crowdsourcing in open courseware development; Application of System Dynamics in modelling agricultural, health and livelihood systems; Benedict has led several systems development at Gulu University, including: Graduate Tracking System, Publication System, Covid registration and attendance tracking system, Human Resource Management System, and an offline eLearning system. Benedict was in 2017 appointed to the 10-member board of the National ICT Initiatives Support Programme (NIISP) under the Ministry of ICT&NG. Benedict serves on nine journal committees as an external reviewer, and has published over 15 peer reviewed articles since his PhD graduation in 2012. | 2025-03-05T00:00:00 | olmocr | {
"pdf_page_numbers": [
[
0,
2489,
1
],
[
2489,
6378,
2
],
[
6378,
10513,
3
],
[
10513,
14715,
4
],
[
14715,
18092,
5
],
[
18092,
20041,
6
],
[
20041,
22446,
7
],
[
22446,
24579,
8
],
[
24579,
25564,
9
],
[
25564,
28784,
10
],
[
28784,
32180,
11
],
[
32180,
35923,
12
],
[
35923,
40038,
13
]
]
} |
71e9ab41fd2456825004640b11b8b852936a20d9 | {
"olmocr-version": "0.1.59",
"pdf-total-pages": 13,
"total-fallback-pages": 0,
"total-input-tokens": 37698,
"total-output-tokens": 12799,
"fieldofstudy": [
"Medicine"
]
} | Tick-Tock Chimes the Kidney Clock – from Biology of Renal Ageing to Clinical Applications
Joshua Rowlanda Artur Akbarov a Akhlaq Maana James Ealesa John Dormerb Maciej Tomaszewskia,c
a Division of Cardiovascular Sciences, Faculty of Medicine, Biology and Health, University of Manchester, Manchester, b University Hospitals of Leicester NHS Trust, Leicester, c Division of Medicine, Central Manchester NHS Foundation Trust, Manchester Academic Health Science Centre, Manchester, UK
Key Words
Age • Kidney • Estimated glomerular filtration rate • Nephrosclerosis • Senescence • Chronic kidney disease
Abstract
Ageing of the kidney is a multi-dimensional process that occurs simultaneously at the molecular, cellular, histological, anatomical and physiological level. Nephron number and renal cortical volume decline, renal tubules become atrophic and glomeruli become sclerotic with age. These structural changes are accompanied by a decline in glomerular filtration rate, decreased sodium reabsorption and potassium excretion, reduced urinary concentrating capacity and alterations in the endocrine activity of the kidney. However, the pace of progression of these changes is not identical in everyone - individuals of the same age and seemingly similar clinical profile often exhibit stark differences in the age-related decline in renal health. Thus, chronological age poorly reflects the time-dependent changes that occur in the kidney. An ideal measure of renal vitality is biological kidney age – a measure of the age-related changes in physiological function. Replacing chronological age with biological age could provide numerous clinical benefits including improved prognostic accuracy in renal transplantation, better stratification of risk and identification of those who are on a fast trajectory to an age-related drop in kidney health.
Introduction
The intrinsic measure of biological age has captivated the imagination of the general public for years. The appeal may stem from a desire to quantify one’s remaining years, the pursuit of novel treatments to extend lifespan, or as a justification for poor health. For the
clinician, interest in calculating a person's biological age dates back to at least the 1960s, when researchers monitoring the health of survivors from the Hiroshima and Nagasaki bombings sought to quantify their biological age via the amalgamation of predictive biological markers [1].
Biological age can be defined as an intrinsic measure of the age-related changes in physiological reserve; that is the capacity for organs to carry out normal physiological function. In the last 50 years, numerous attempts have been made to develop a reliable algorithm to measure biological age [2, 3]. The mechanisms that underlie ageing are more complex than initially expected. Consequently, biological ageing algorithms have grown increasingly complex. Recent evidence suggests that each organ has a unique ageing pattern, indicating that the biological age of each organ should be calculated individually [4].
Herein, we provide an overview of structural and functional changes that occur as the kidney ages (see Fig. 1). We summarise research progress on inter and intra-individual differences in biological age and expound the clinical importance of accurate calculation of an individual's biological kidney age.
Fig. 1. Molecular, structural (microscopic and macroscopic), and functional dimensions of kidney ageing. Cellular dysfunction due to oxidative stress and the resultant inflammation, in combination with telomere shortening, lead to cellular senescence and apoptosis. Reduced reparative capacity and extracellular matrix dysregulation is associated with microscopic damage known as nephrosclerosis. Sclerotic glomeruli shrivel, leading to reduction in cortical volume. At a macro-anatomical level ageing is associated with cortical scarring and parenchymal calcification. These anatomical changes are accompanied by reductions in glomerular filtration rate, tubular dysfunction and aberrant endocrine activity. Abbreviations: GFR: Glomerular filtration rate, RAAS: Renin-angiotensin-aldosterone system.
filtration rate (eGFR) are commonly used as a measure of age-related decline in kidney health. Our earlier family-based studies documented the heritable nature of eGFR and that the proportion of its variance explained by the heritable additive component is actually higher than that of blood pressure [5]. The estimated heritability for age-related drop in eGFR (h² = 0.33) is generally less significant than that for eGFR (h² = 0.38-0.75) [5–9], although it appears that monozygotic twins exhibit higher correlation for age-related changes in eGFR than dizygotic twins [10].
Over 60 single nucleotide polymorphisms (SNPs) were associated with eGFR in genome-wide association studies [11, 12]. Similar to other complex polygenic traits, the extent to which these genetic variants explain the proportion of inter-individual variance in the decline in eGFR is minimal [11]. Only 3 SNPs have been associated with eGFR decline so far [8]. Further evidence for the contribution of genes to the development of age-related changes in the kidney come from gene expression studies – hundreds of genes are up- or down-regulated in the human kidney in response to ageing [13]. MicroRNAs (miRNA) - small noncoding RNAs that regulate gene expression post-transcriptionally are also associated with renal ageing. Indeed, at least 18 miRNAs were significantly upregulated and 10 miRNAs were downregulated with ageing in the kidney of the rat [14]. Surprisingly few studies directly addressed the influence of environmental factors on the ageing of the kidney. Dietary factors such as caloric restriction reduce the rate of age-associated autophagy and oxidative stress in the kidney, a process mediated via SIRT1, AMPK and mTOR [15]. Methionine consumption (found in red meat, cheese and nuts) appears to have the opposite effect as documented in experimental models [16].
**Key molecular mechanisms of renal ageing**
**Oxidative stress and inflammation**
The free radical theory of ageing proposes that oxidative stress damages cellular constituents, leading to age-related decline. The kidney deploys an arsenal of mechanisms to prevent reactive oxygen species from wreaking havoc. For example, superoxide dismutase 1 and 2 (SOD1 and SOD2) soak up free radicals, preventing organelle and DNA damage. Unfortunately, like many other antioxidants, SOD expression declines with age [17]. As exhibited in knockout (KO) mice, absence of SOD1 leads to glomerulonephritis, nephrocalcinosis and lymphocyte infiltration. Consequently, KO mice have a reduced lifespan [18]. As the kidney ages, damage driven by oxidative stress leads to accumulation of macrophages and lymphocytes in the renal tissue [19]. Infiltrating macrophages release IFNγ, IL-6 and TNFα, which activate key master transcription factors including STAT1, STAT3 and NFκB [20]. Of these transcription factors, NFκB has been studied in extensive detail, with the hope that if able to target it, you could halt the age-associated inflammatory cascade.
**Cellular senescence and telomere shortening**
Central to the ageing process lies cellular senescence, the irreversible growth arrest that constricts renal regenerative capacity and propagates a pro-inflammatory state termed the senescence-associated secretory phenotype (SASP). In acute senescence e.g. post acute kidney injury (AKI), the SASP coordinates the removal of senescent cells through immune surveillance. However, in ageing, immune system dysfunction prevents effective clearance of senescent cells, leading to persistent SASP factor expression which causes inflammatory and fibrotic damage to surrounding cells [21]. Recently, Baker et al. demonstrated therapeutic clearance of senescent cells using a drug-inducible transgene to initiate apoptosis improved renal function, reduced glomerulosclerosis and ultimately led to increases in mouse lifespan [22]. One of the key instigators of cellular senescence is telomeric shortening. The telomere theory of ageing proposes that lifespan is predetermined by a finite capacity for cellular replication, called the Hayflick limit [23]. There is a documented inverse correlation between
the length of telomeres in the kidney and chronological age [24]. Telomeric attrition was also associated with increased susceptibility to AKI and decreased graft survival post-transplant [25, 26]. However, telomere shortening is not an ideal biomarker of ageing. For one, telomere dysfunction can cause cellular senescence independent of telomere shortening [27]. Moreover, telomere length stops being a useful predictor of age-related morbidity and mortality in those older than 85 [28]. Furthermore, a cell’s telomere length is indicative of its’ replicative history, and does not necessarily correlate with its’ biological age.
**Structural and functional changes in renal ageing**
**Histology**
At birth, the human kidney contains approximately 900,000-1 million nephrons [29]. No new nephrons are formed after 36 weeks gestation [29]. Ageing is associated with the depletion of approximately 4,500 nephrons per year [30, 31]. This equates to loss of almost half one’s nephrons between early adulthood (18-19) and old age (70-75). Nephron number is proportional to eGFR throughout one’s life, except in the elderly age group (70-75) where nephron number drastically drops with minimal consequence to eGFR. One possibility is that compensatory hypertrophy of residual nephrons maintains eGFR in some elderly populations [31]. This could explain (at least to some extent) the inter-individual differences seen in eGFR in elderly populations [32–34].
The drop in nephron numbers with age is accompanied by the changes in renal histology. The glomerular basement membrane thickens, its capillaries shrivel and are replaced by fibrotic tissue (glomerular sclerosis), the renal tubules collapse (tubular atrophy), extracellular matrix components accumulate, expanding the interstitial space (interstitial fibrosis) and arterial walls thicken and lose their elasticity (arteriosclerosis) [35]. This tetrad of abnormalities is termed nephrosclerosis [36]. Pairwise comparison indicates these age-related histological abnormalities are highly correlated with one another [36]. However cumulatively, as measured by nephrosclerosis score; they are not associated with age-related eGFR decline, perhaps due to the involution and eradication of sclerotic glomeruli distorting findings [36].
Nephrosclerosis is the most common pathway for kidney injury in ageing [37]. It has its roots in early life, prior to the development of chronic kidney disease (CKD) [36], but is clinically silent throughout a major part of its natural history; unlike many other renal conditions it does not usually manifest with proteinuria [38]. However, the gradual progression of nephrosclerotic changes leads inevitably to a loss in functional nephron reserve and atrophy increasing the susceptibility of affected patients not only to severe presentations of progressive CKD but also other renal disorders such as acute kidney injury [39]. Nephrosclerosis was also reported to cluster with high mortality and high risk of end-stage renal disease [40]. As a facet of ageing, histological appearance is unique in that it represents the wounds of time, the afflictions faced by an organ.
**Macroscopic changes**
From the 4-5th decade onwards, renal volume declines [41]. Diminution is largely confined to the renal cortex, as renal medulla volume remains relatively stable perhaps due to tubular hypertrophy or unaccounted increases in renal sinus fat. Loss of renal mass, in combination with atherosclerotic plaque formation and an increase in renal sympathetic tone, results in declining renal blood flow, at a rate of 10% per decade, from the 4-5th decade onwards [39, 42]. Inadequate renal perfusion likely contributes to the age-associated decline in GFR. Furthermore, ageing is associated with increased prevalence of renal cysts, parenchymal calcifications and cortical scars [39].
### Functional changes
Multiple studies report variable age-associated decline in eGFR from 0.4-2.6 ml/min/year [33, 34, 36, 43, 44]. The rate of decline increases with age. In 20-30-year olds eGFR decreased by 0.82 ml/min/1.73m^2/year in comparison to those over 50 where it decreased by 1.15 ml/min/1.73m^2/year [34]. Ageing is associated with tubular dysfunction including decreased sodium reabsorption, potassium excretion and reduced urinary concentrating capacity [45]. These changes in part account for the increased risk of dehydration and AKI observed in elderly individuals [46]. Ageing also affects the endocrine function of the kidney. Despite decreased renin expression, angiotensin II activity increases with age [47, 48]. This may be due to increases in angiotensin II receptor sensitivity or differential regulation of systemic and intrarenal renin-angiotensin systems [49, 50]. Renal conversion of 25-hydroxyvitamin D into 1,25-dihydroxyvitamin D declines with age, contributing to the vitamin D deficiency commonly seen in the elderly [51]. Commonly used to distinguish age-associated renal dysfunction from chronic kidney disease, elderly individuals exhibit raised levels of erythropoietin, perhaps due to subclinical blood loss, increased erythrocyte turnover or increased insensitivity to the effects of erythropoietin [52].
### Intra and inter-individual variation in ageing
A person’s chronological age can drastically differ from their underlying biological age. Belsky et al. used an array of biomarkers, including creatinine clearance and blood urea nitrogen, to measure the biological ages of a large cohort of 38-year-olds [2]. Despite being the same chronological age, individual biological ages ranged from 28-61. Some participants aged 3 biological years for every calendar year, whereas others exhibited almost no physiological age-related change in a year.
The pace of renal ageing varies between individuals. Findings by Rule et al. suggest substantial variation in nephrosclerosis scores between healthy individuals within the same age bracket [36]. For example, 10% of individuals aged 60-69 had a nephrosclerosis score of 0, whilst another 10% had a sclerosis score of 4. Our data from the TRANScriptome of renAL humAn TissuE (TRANSLATE) study [53, 54] illustrates how two individuals of the same chronological age and seemingly similar clinical profile exhibit stark contrast in renal histological appearance (see Fig. 2). Longitudinal studies by Linderman, Jiang and Cohen demonstrate 36%, 43% and 15.4% of healthy individuals exhibit no age-related decline in eGFR respectively [32–34]. These findings indicate that age-associated functional decline in healthy kidneys exhibits marked inter-individual variation.
It is not clear at present what offers a protection against age-related decline in the structural integrity and function of the kidney. The differences in the observed phenotypic changes of the kidney can be explained (at least to some extent) by the differences in gene expression profiles between those with faster and slower renal ageing. Indeed, Rodwell et al. evaluated the age-associated changes in gene expression in 74 kidneys [13]. They identified 447 age-regulated genes that form a molecular profile of ageing. In doing so, they noted that certain individuals despite being chronologically younger, had a gene expression profile suggestive of someone more senior (fast agers), and vice versa (slow agers).
Apart from these apparent differences in pace of renal ageing between individuals, variation exists in ageing of different organs from the same individual. For example, comparison of age-associated transcriptomic changes in kidney and muscle, demonstrated minor overlap in age-related gene expression, suggesting discrete molecular ageing mechanisms [13, 55]. Cadaveric studies measuring telomere length in 12 different tissues, including the kidney, demonstrates high variability in the extent of telomeric attrition between organs, suggesting divergent pace of ageing [56]. In addition, different histological
components of the same organ from the same individual respond differently to ageing. Within the kidney, Rodwell et al. demonstrate little overlap in expression of ageing-associated genes between the cortex and the medulla [13]. Likewise, Melk et al. observed a faster rate of telomeric attrition in the renal cortex, than the renal medulla [24].
Quantifying the biological age of the kidney
There are a number of potential surrogates of biological renal age (Table 1) [3, 27-28, 32-33, 36, 57-58]. Telomere length and eGFR have previously been utilised to predict biological age [2], however, they have their own limitations as discussed earlier. Molecular markers such as Klotho may have a potential role in renal ageing. Indeed, overexpression of Klotho has been shown to extend the lifespan of mice by 20-30% [59]. In the kidney, Klotho is involved in the prevention of cellular senescence, regulation of interstitial fibrosis, and suppression of inflammation [59-61]. Urinary and serum Klotho levels have already been utilised as prognostic markers of CKD [62, 63]. Nephrosclerosis is a potential proxy, having been closely correlated with age [36], age-associated kidney gene expression [13] and renal transplant outcome [64]. However, obtaining histological samples via renal biopsy is not feasible in patients without clear clinical indication i.e. evidence of overt nephropathy. The
Fig. 2. Inter-individual variation in nephrosclerosis. Histology images from two TRANScriptome of renal humAn TissuE (TRANSLATE) study patients (1 and 2) with a seemingly comparable clinical profile; both 62-year-old men with hypertension and obesity. Despite similar clinical histories the two patients exhibit dramatically different nephrosclerotic changes. 1. Normal interlobular artery. 2. Global glomerulosclerosis. 3. Glomerular collapse. 4. Tubular atrophy and interstitial fibrosis with inflammation. 5. Interlobular artery with fibrointimal thickening. A and C = Periodic acid–Schiff, x40. B and D = Masson’s trichrome, x100.
1- No nephrosclerosis
2- Severe nephrosclerosis
A
B
C
D
1
2
3
4
5
successful identification of robust signatures of kidney-specific biological age will require exploiting new molecular strategies such as transcriptomics, epigenomics, proteomics and metabonomics. Each of them offers an unbiased systematic insight into thousands of genes and molecules many of which are the key determinants of the individual trajectories of kidney ageing. A particularly promising strategy is transcriptomic profiling of cells harvested from urine samples [58]. Additional strategies, including the measurement of circulating levels of cell-free DNA have shown promise in the prediction of outcomes in renal transplantation [65, 66], and this could be potentially further exploited in studies on renal ageing. Next-generation RNA-sequencing-based profiling of epigenetic master regulators, such as small (miRNA, small interfering RNA, piwi-interacting RNA) and long non-coding RNAs may shed an insight into these ageing signatures [67, 68]. Horvath’s DNA methylation-based measure of biological age has already shown the potential of exploiting human epigenome in search of signatures of tissue ageing [3]. Ideally, the use of “omics” should be integrated with the objective and direct measures of age-related kidney damage such as histologically confirmed nephrosclerosis. The availability of resources where both histologically-confirmed measures of age-related kidney damage together with biological materials suitable for omics-type profiling (i.e. TRANSLATE Study) [53, 54] brings us closer to finding the multi-marker signatures of biological kidney age. We believe that the future of a kidney-specific ageing signature lies in systems biology – combining genetics, transcriptomics, epigenetics, proteomics, clinical and histological data. Such efforts are currently underway as part of our TRANSLATE Study [53, 54] and in the NEPTUNE Study [69].
**Clinical prediction of biological kidney age**
The attractiveness of determining the biological age of the kidney lies in its diagnostic and predictive potential. The shortage of donor kidneys has led to increased use of suboptimal organs, often from elderly donors. In Great Britain, the UK kidney donor risk-index scoring system is commonly used to screen the allograft quality. It incorporates donor age, which is known to predict poor outcome after transplant [70]. Biological kidney age could supersede chronological age, as a more accurate predictor of prognosis. Thus, kidney donors rejected due to their old age, may now be able to donate if found to be younger than their expected biological age, increasing the pool of available kidneys. Conversely, organs of apparent optimal suitability for renal transplantation based on chronological age, may require more careful monitoring upon transplantation should their biological age be much older than their chronological age. The value of incorporating markers of ageing into prognostic algorithms in renal transplantation is receiving increasing attention due to its potential to improve outcome in recipients [71]. For example, two non-coding RNAs, miR-217 and miR125b were shown to predict delayed graft function with a 61% sensitivity and 91% specificity.
Given that chronological age is the major risk factor for AKI, it is tempting to speculate that the prediction of risk and monitoring of clinical outcomes in patients with AKI can be
### Table 1. Surrogates of renal biological age – past, present and future
| Signature | Source | Limitations | References |
|-----------|--------|-------------|------------|
| eGFR | Blood | Significant inter-individual differences. Indirect measure based on serum levels of creatinine. Poor correlation with histological measures of age-related kidney damage. | Linderman et al [32], Jiang et al [33] |
| Telomere length | Blood/Tissue sample | Mostly measured in leukocytes from peripheral blood. Poor correlation with biochemical measures of kidney function. Cell replication ≠ biological cell age. | Martin-Ruiz et al [28] |
| Klotho expression | Urine/Blood | Reliability of urinary/serum measurement questionable. | Akimoto et al [57], Rule et al [36] |
| Nephrosclerosis | Tissue sample | Measurement requires invasive procedure. | Rule et al [36] |
| Transcriptomic signatures | Urine/Blood/Tissue sample | Limited evidence in relation to kidney ageing. | Suthanthiran et al [58] |
| Epigenetic signatures | Urine/Blood/Tissue sample | Limited evidence in relation to kidney ageing. | Horvath et al [3] |
improved by knowledge of their biological kidney age. This is particularly relevant to AKI due to drug nephrotoxicity (approximately 20% of AKI) [72] as clinical guidelines recommend careful monitoring and increased caution when prescribing high-risk medications to the elderly (based on their chronological age) [73]. Information on biological kidney age could more accurately inform the decision-making process. For example, the Mehran contrast-induced AKI risk score advocated by the Kidney disease: Improving global outcomes (KDIGO) guidelines advises that those chronologically older than 75 are at increased risk of contrast nephropathy [74]. If an individual chronologically younger than 75 was found to have a biological age greater than 75, it would be sensible to assume increased risk and if appropriate, modify their management plan in accordance with best practice guidelines.
Perceived biological age currently plays a role in surgical decision making, often conveyed in the notes as "remarkably fit for 88" or "a rather old 71-year-old". With respect to the kidney, eGFR is typically the only measure of renal vitality used in pre-operative assessments. Reduced eGFR prior to surgery is associated with increased mortality, independent of AKI [75]. However, eGFR is a suboptimal measure of overall renal health. As previously noted, eGFR only measures one aspect of renal function, it does not reflect the age-related kidney change in a significant cohort of the population [32–34], and it poorly correlates with other markers of renal biological age e.g. nephrosclerosis score [36]. Thus, pre-operative appraisal of biological kidney age in addition to eGFR could provide more accurate prognostic information.
Advanced chronological age is independently associated with poor prognosis in IgA nephropathy [76] and renal malignancy [77]. Conversely, increased chronological age at disease onset in autosomal recessive polycystic kidney disease is associated with improved prognosis [78]. In these conditions, biological age could replace chronological age as a prognostic indicator. Other kidney diseases with a less well-defined role of age on outcome include diabetic nephropathy [79], focal segmental glomerulosclerosis [80] and membranous nephropathy [81].
Ageing is associated with increased prevalence of chronic kidney disease (CKD), indeed, almost half of elderly individuals fulfil the current diagnostic criteria for CKD [82, 83]. Elderly patients with CKD have a greater adjusted risk of death [83] and exhibit lower functional kidney reserve when they present with CKD [84]. Current diagnostic criterion relies on a fixed threshold to identify CKD, with little consideration for age-related decline in eGFR, with some suggesting this leads to over-diagnosis of otherwise healthy individuals [85, 86]. One possible solution is incorporation of biological kidney age into diagnostic criteria in this group of patients.
For the healthy individual, whose baseline eGFR is otherwise normal, biological kidney age could be used to identify those with accelerated renal ageing. For example, Rule et al. demonstrate in a population of healthy living kidney donors, 15% of those aged 30-39 exhibit substantial age-related nephrosclerotic change (indicating accelerated renal ageing), yet all these individuals have a measured GFR >75 ml/min/1.73m² [36]. It is reasonable to assume these fast renal agers are more susceptible to age-related renal disease (e.g. chronic kidney disease). Thus, an increased rate of kidney ageing could warrant prophylactic measures i.e. prescription of prospective anti-ageing therapies [87].
**Conclusion**
Renal ageing is associated with a progressive decline in eGFR and structural disfigurement at a microscopic and macroscopic level. At a molecular level, this is accompanied by cellular senescence, telomere shortening, apoptosis and fibrosis. There are numerous gaps in knowledge on how molecular changes influence renal histology and how histological changes translate into a decline in renal function. Furthermore, little is known about the genetic and environmental factors that influence individual rates of age-related decline. Intra-individual
differences in organ ageing advocate the development of a kidney-specific ageing signature. This successful identification of such signatures relies on combining “omics”, clinical and histological data. This will have several potential benefits for clinical nephrology including improved prognostics and identification of high-risk patients.
**Disclosure statement**
The authors of this review have no conflicts of interest.
**Acknowledgements**
This work was supported by a British Heart Foundation grant: PG/17/35/33001.
**References**
1. Hollingsworth JW, Hashizume A, Jablon S: Correlations between tests of aging in Hiroshima subjects--an attempt to define "physiologic age". Yale J Biol Med 1965;38:11–26.
2. Belsky DW, Caspi A, Houts R, Cohen HJ, Corcoran DL, Danese A, Harrington H, Israel S, Levine ME, Schaefer JD, Sugden K, Williams B, Yashin AI, Poulton R, Moffitt TE: Quantification of biological aging in young adults. Proc Natl Acad Sci 2015;112:E4104–E4110.
3. Horvath S: DNA methylation age of human tissues and cell types. Genome Biol 2013;14:R115.
4. Ori A, Toyama BH, Harris MS, Bock T, Iskar M, Bork P, Ingolia NT, Hetzer MW, Beck M: Integrated Transcriptome and Proteome Analyses Reveal Organ-Specific Proteome Deterioration in Old Rats. Cell Syst 2015;1:224–237.
5. Debiec R, Christofidou P, Denniff M, Bloomer LD, Bogdanski P, Wojnar L, Musialik K, Charchar FJ, Thompson JR, Waterworth D, Song K, Vollenweider P, Waeger G, Zukowska-Szczechowska E, Samani NJ, Lambert D, Tomaszewski M: Urotensin-II System in Genetic Control of Blood Pressure and Renal Function. Plos One 2013;8:e83137.
6. Raggi P, Su S, Karohl C, Vekedr E, Rojas-Campos E, Vaccarino V: Heritability of Renal Function and Inflammatory Markers in Adult Male Twins. Am J Nephrol 2010;32:317–323.
7. Fox CS, Yang Q, Cupples LA, Guo CY, Larson MG, Leip EP, Wilson PWF, Levy D: Genomewide Linkage Analysis to Serum Creatinine, GFR, and Creatinine Clearance in a Community-Based Population: The Framingham Heart Study. J Am Soc Nephrol 2004;15:2457–2461.
8. Gorski M, Tin A, Garnaas M, McMahon GM, Chu AY, Tayo BO, Pattaro C, Teumer A, Chasman DI, Chalmers J, Hamet P, Tremblay J, Woodward M, Aspelund T, Eiriksdottir G, Gudnason V, Harris TB, Launer LJ, Smith AV, Mitchell BD, et al.: Genome-wide association study of kidney function decline in individuals of European descent. Kidney Int 2015;87:1017–1029.
9. Pattaro C, Aulchenko YS, Isaacs A, Vitart V, Healy G, Polasek O, Kotschy A, Biloglav Z, Campbell S, Hastie N, Lauc G, Meitinger T, Oostra BA, Gyllensten U, Wilson JF, Pichler I, Hicks AA, Campbell H, Wright A, et al.: Genome-wide linkage analysis of serum creatinine in three isolated European populations. Kidney Int 2009;76:297–306.
10. Moayyeri A, Hart DJ, Snieder H, Hammond CJ, Spector TD, Steves CJ: Aging Trajectories in Different Body Systems Share Common Environmental Etiology: The Healthy Aging Twin Study (HATS). Twin Res Hum Genet 2016;19:27–34.
11. Pattaro C, Teumer A, Gorski M, Chu AY, Li M, Mijatovic V, Garnaas M, Tin A, Sorice R, Li Y, Taliun D, Olden M, Foster M, Yang Q, Chen MH, Pers TH, Johnson AD, Ko YA, Fuchsberger C, Tayo B, et al.: Genetic associations at 53 loci highlight cell types and biological pathways relevant for kidney function. Nat Commun 2016;7:10023.
12 Gorski M, van der Most PJ, Teumer A, Chu AY, Li M, Mijatovic V, Nolte IM, Cozza C, Maltiun D, Gomez F, Li Y, Tayo B, Tin A, Feitosa ME, Aspelund T, Attia J, Biffar R, Bochud M, Boerwinkle E, Borecld I, et al.: 1000 Genomes-based meta-analysis identifies 10 novel loci for kidney function. Sci Rep 2017;7:45040.
13 Rodwell GEJ, Sonu R, Zahn JM, Lund J, Wilhelmy J, Wang L, Xiao W, Mindrinos M, Crane E, Segal E, Myers BD, Brooks JD, Davis RW, Higgins J, Owen AB, Kim SK: A transcriptional profile of aging in the human kidney. PLoS Biol 2004;2:e427.
14 Bai XY, Cai GY, Zhuo L, Gao JJ, Dong D, Cui S, Feng Z, Shi SZ, Bai XY, Sun XF, Chen XM: Short-term calorie restriction protects against renal senescence of aged rats by increasing autophagic activity and reducing oxidative damage. Mech Ageing Dev 2013;134:570–579.
15 Ning YC, Cai GY, Zhuo L, Gao JJ, Dong D, Cui S, Feng Z, Shi SZ, Bai XY, Sun XF, Chen XM: Short-term calorie restriction protects against renal senescence of aged rats by increasing autophagic activity and reducing oxidative damage. Mech Ageing Dev 2013;134:570–579.
16 Grant L, Lees EK, Forney LA, Mody N, Gettys T, Brown PAJ, Wilson HM, Delibegovic M: Methionine restriction improves renal insulin signalling in aged kidneys. Mech Ageing Dev 2016;157:35–43.
17 Lim JH, Kim EN, Kim MY, Chung S, Shin SJ, Kim HW, Yang CW, Kim YS, Chang YS, Park CW, Choi BS: Age-associated molecular changes in the kidney in aged mice. Oxid Med Cell Longev 2012;2012:171383.
18 Zhang Y, Unnikrishnan A, Deepa SS, Liu Y, Li Y, Ikono Y, Sosnowska D, van Remmen H, Richardson A: A new role for oxidative stress in aging: The accelerated aging phenotype in Sod1(-/-) mice is correlated to increased cellular senescence. Redox Biol 2016;11:30–37.
19 Xi Y, Shao F, Bai X-Y, Cai G, Lu Y, Chen X: Changes in the expression of the Toll-like receptor system in the aging rat kidneys. PloS One 2014;9:e96351.
20 O’Brown ZK, Van Nostrand EL, Higgins JP, Kim SK: The Inflammatory Transcription Factors NFκB, STAT1 and STAT3 Drive Age-Associated Transcriptional Changes in the Human Kidney. PLoS Genet 2015;11:e1005734.
21 Sturmlechner I, Durik M, Sieben CJ, Baker DJ, van Deursen JM: Cellular senescence in renal ageing and disease. Nat Rev Nephrol DOI: 10.1038/nrneph.2016.183.
22 Baker DJ, Childs BG, Durik M, Wijers ME, Sieben CJ, Zhong J, Sakness RA, Jeganathan KB, Verzosa GC, Pezeshki A, Khazaie K, Miller JD, van Deursen JM: Naturally occurring p16(Ink4a)-positive cells shorten healthy lifespan. Nature 2016;530:184–189.
23 Jin K: Modern Biological Theories of Aging. Aging Dis 2010;1:72–74.
24 Melk A, Ramassar V, Helms LMH, Moore R, Rayner D, Solez K, Halloran PF: Telomere Shortening in Kidneys with Age. J Am Soc Nephrol 2000;11:444–453.
25 Westhoff JH, Schildhorn C, Jacobi C, Hönme M, Hartner A, Braun H, Kryzer C, Wang C, von Zglinicki T, Kränzlin B, Greitz N, Melk A: Telomere shortening reduces regenerative capacity after acute kidney injury. J Am Soc Nephrol 2010;21:327–336.
26 Koppelstaetter C, Schratzberger G, Perco P, Hofer J, Mark W, Ollinger R, Oberbauer R, Schwarz C, Mitterbauer C, Kainz A, Karkoszka H, Wieczek A, Mayer B, Mayer G: Markers of cellular senescence in zero hour biopsies predict outcome in renal transplantation. Aging Cell 2008;7:491–497.
27 Arai Y, Martin-Ruiz CM, Takayama M, Abe Y, Takebayashi T, Koyasu S, Suematsu M, Hirose N, von Zglinicki T: Inflammation, But Not Telomere Length, Predicts Successful Ageing at Extreme Old Age: A Longitudinal Study of Semi-supercentenarians. EBioMedicine 2015;2:1549–1558.
28 Martin-Ruiz CM, Gussekloo J, van Heemst D, von Zglinicki T, Westendorp RGJ: Telomere length in white blood cells is not associated with morbidity or mortality in the oldest old: a population-based study. Aging Cell 2005;4:287–290.
29 Bertram JF, Douglas-Denton RN, Diouf B, Hughson MD, Hoy WE: Human nephron number: implications for health and disease. Pediatr Nephrol 2011;26:1529.
30 Hoy WE, Hughson MD, Bertram JF, Douglas-Denton R, Amann K: Nephron number, hypertension, renal disease, and renal failure. J Am Soc Nephrol 2005;16:2557–2564.
31 Denic A, Lieske JC, Chakkera HA, Poggio ED, Alexander MP, Singh P, Kremers WK, Lerman LO, Rule AD: The Substantial Loss of Nephrons in Healthy Human Kidneys with Aging. J Am Soc Nephrol 2017;28:313–320.
32 Lindeman RD, Tobin J, Shock NW: Longitudinal studies on the rate of decline in renal function with age. J Am Geriatr Soc 1985;33:278–285.
33 Jiang S, Sun X, Gu H, Chen Y, Xi C, Qiao X, Chen X: Age-related change in kidney function, its influencing factors, and association with asymptomatic carotid atherosclerosis in healthy individuals--a 5-year follow-up study. Maturitas 2012;73:230–238.
34 Cohen E, Nardi Y, Krause I, Goldberg E, Milo G, Garty M, Krause I: A longitudinal assessment of the natural rate of decline in renal function with age. J Nephrol 2014;27:635–641.
35 Zhou XJ, Rakheja D, Yu X, Saxena R, Vaziri ND, Silva FG: The aging kidney. Kidney Int 2008;74:710–720.
36 Rule AD, Amer H, Cornell LD, Taler SJ, Cosio FG, Kremers WK, Teector SC, Stegall MD: The association between age and nephrosclerosis on renal biopsy among healthy adults. Ann Intern Med 2010;152:561–567.
37 Meyrier A: Nephrosclerosis: update on a centenarian. Nephrol Dial Transplant 2015;30:1833–1841.
38 Meyrier A: Nephrosclerosis: a term in quest of a disease. Nephron 2015;129:276–282.
39 Denic A, Glassock RJ, Rule AD: Structural and functional changes with the aging kidney. Adv Chronic Kidney Dis 2016;23:19–28.
40 Vikse BE, Aasarød K, Bostad L, Iversen BM: Clinical prognostic factors in biopsy-proven benign nephrosclerosis. Nephrol Dial Transplant 2003;18:517–523.
41 Wang X, Vrtiska TJ, Avula RT, Walters LR, Chakkera HA, Kremers WK, Lerman LO, Rule AD: Age, kidney function, and risk factors associate differently with cortical and medullary volumes of the kidney. Kidney Int 2014;85:677–685.
42 Hollenberg NK, Adams DF, Solomon HS, Rashid A, Abrams HL, Merrill JP: Senescence and the Renal Vasculature in Normal Man. Circ Res 1974;34:309–316.
43 Poggio ED, Rule AD, Tanchanco R, Arrigain S, Butler RS, Srinivas T, Stephany BR, Meyer KH, Nurko S, Fatica RA, Shoskes DA, Krishnamurthi V, Goldfrarb DA, Gill I, Schreiber MJ: Demographic and clinical characteristics associated with glomerular filtration rates in living kidney donors. Kidney Int 2009;75:1079–1087.
44 Lauretani F, Semba RD, Bandinelli S, Miller ER, Ruggiero C, Cherubini A, Guralnik JM, Ferrucci L: Plasma polyunsaturated fatty acids and the decline of renal function. Clin Chem 2008;54:475–481.
45 Martin JE, Sheaff MT: Renal ageing. J Pathol 2007;211:198–205.
46 Abdel-Kader K, Palevsky P: Acute Kidney Injury in the Elderly. Clin Geriatr Med 2009;25:331–358.
47 Weidmann P, De Myttenaere-Bursztein S, Maxwell MH, de Lima J: Effect on aging on plasma renin and aldosterone in normal man. Kidney Int 1975;8:325–333.
48 Sangaralingham SJ, Wang BH, Huang L, Kumfu S, Ichiki T, Krum H, Burnett JC: Cardiorenal fibrosis and dysfunction in aging: Imbalance in mediators and regulators of collagen. Peptides 2016;76:108–114.
49 Thompson MM, Oyama TT, Kelly FJ, Kennefick TM, Anderson S: Activity and responsiveness of the renin-angiotensin system in the aging rat. Am J Physiol Regul Integr Comp Physiol 2000;279:R1787-1794.
50 Wray DW, Nishiya SK, Harris RA, Richardson RS: Angiotensin II in the Elderly. Hypertension 2008;51:1611–1616.
51 Veldhuij J, Wei R, Oz L, Dhawan P, Jeon YH, Christakos S: Vitamin D, calcium homeostasis and aging. Bone Res 2016;4:1604.
52 Ralser WR, Sheng S, McKeel J, Artz AS, Dengduluri N, Tecson J, Taub DD, Brant LJ, Ferrucci L, Longo DL: Serum erythropoietin and aging: a longitudinal analysis. J Am Geriatr Soc 2005;53:11360–11365.
53 Marques FZ, Romaine SP, Denniff M, Eales J, Dormer J, Garrel D, Wojnar L, Musialik L, Duda-Raszewska B, Kiszka B, Duda M, Morris BJ, Samani NJ, Danser AJ, Bogdanski P, Zulowska-Szczechowska E, Charchar FJ, Tomaszewski M: Signatures of miRNA-181a on the Renal Transcriptome and Blood Pressure. Mol Med 2015;21:739–748.
54 Tomaszewski M, Eales J, Denniff M, Myers S, Chew GS, Nelson CP, Christofidou P, Desai A, Büss C, Wojnar L, Musialik K, Zojwiak J, Debrec E, Dominiczak AF, Navis G, van Gilst WH, van der Harst P, Samani NJ, Harrap S, Bogdanski P, et al.: Renal Mechanisms of Association between Fibroblast Growth Factor 1 and Blood Pressure. J Am Soc Nephrol 2015;26:3151–3160.
55 Zahn JM, Sonu R, Vogel H, Crane E, Mazan-Manczczar K, Rabkin R, Davis RW, Becker KG, Owen AB, Kim SK: Transcriptional profiling of aging in human skin reveals a common aging signature. PLoS Genet 2006;2:e115.
56 Dlouha D, Maluskova J, Kravolova Lesna I, Lanskia V, Hubacek JA: Comparison of the relative telomere length measured in leukocytes and eleven different human tissues. Physiol Res 2014;63 Suppl 3:S343-350.
57 Akimoto T, Yoshizawa H, Watanabe Y, Numata A, Yamazaki T, Takeshima E, Iwazu K, Komada T, Otani N, Morishita Y, Ito C, Shizuki K, Ando Y, Muto S, Kuro-o M, Kusano E: Characteristics of urinary and serum soluble Klotho protein in patients with different degrees of chronic kidney disease. BMC Nephrol 2012;13:155.
58 Suthanthiran M, Schwartz JE, Ding R, Abecasis M, Dadhania D, Samstein B, Knechtle SJ, Friedewald J, Becker VT, Sharma VK, Williams NM, Chang CS, Hoang C, Muthukumar T, August P, Kessler KS, Fairchild RL, Hricik DE, Heeger PS, Han L, Liu J, Riggs M, Ilde DN, Bridges ND, Shaked A: Urinary-Cell mRNA Profile and Acute Cellular Rejection in Kidney Allografts. N Engl J Med 2013;369:20–31.
59 Kurosu H, Yamamoto M, Clark JD, Pastor JV, Nandi A, Gurnani P, McGuinness OP, Chikuda H, Yamaguchi M, Kawaguchi H, Shimomura I, Takayama Y, Herz J, Kahn CR, Rosenblatt KP, Kuro-o M: Suppression of aging in mice by the hormone Klotho. Science 2005;309:1829–1833.
60 Xu Y, Sun Z: Molecular basis of Klotho: from gene to function in aging. Endocr Rev 2015;36:174–193.
61 Kalaitzidis RG, Duni A, Siamopoulos KC: Klotho, the Holy Grail of the kidney: from salt sensitivity to chronic kidney disease. Int Urol Nephrol 2016;48:1657–1666.
62 Hu MC, Shi M, Zhang J, Quiñones H, Griffith C, Kuro-o M, Moe OW: Klotho deficiency causes vascular calcification in chronic kidney disease. J Am Soc Nephrol 2011;22:124–136.
63 Rotondi S, Pasquali M, Tartaglione L, Mucci ML, Mandanici G, Leonangeli C, Sales S, Farcomeni A, Mazzaferro S: Soluble a-Klotho Serum Levels in Chronic Kidney Disease. Int J Endocrinol DOI: 10.1155/2015/872193.
64 Remuzzi G, Cravedi P, Perna A, Dimitrov BD, Turturro M, Locatelli G, Rigotti P, Baldan N, Beatini M, Valente U, Scalamogna M, Ruggenenti P: Cell-Free DNA: An Upcoming Biomarker in Transplantation. Am J Transplant 2015;15:2541–2551.
65 Kim J, Kim KM, Noh JH, Yoon JH, Abdelmohsen K, Gorospe M: Long noncoding RNAs in Diseases of Aging. Biochim Biophys Acta 2016;1859:209–221.
66 Harries LW: MicroRNAs as Mediators of the Ageing Process. Genes 2014;5:656–670.
67 McGuinness D, Leierer J, Shapter Q, Mohammed S, Ginell-Littlejohn M, Kingsmore DB, Little A-M, Kerschbaum J, Schneeberger S, Maglione M, Nadalin S, Wagner S, Königsrainer A, Atiken E, Whalen H, Clancy M, McConnochie A, Koppelstaetter C, Stevenson KS, Shiels PG: Identification of Molecular Markers of Delayed Graft Function Based on the Regulation of Biological Ageing. Plos One 2016;11:e0146378.
68 Wang X, Bonventre JV, Parrish AR: The aging kidney: increased susceptibility to nephrotoxicity. Int J Mol Sci 2014;15:15358–15376.
69 Mehraan R, Aymong ED, Nikolsky E, Lasic Z, Lakovou I, Fahy M, Mintz GS, Lansky AJ, Moses JW, Stone GW, Leon MB, Dangas G: A simple risk score for prediction of contrast-induced nephropathy after percutaneous coronary intervention. J Am Coll Cardiol 2004;44:1393–1399.
70 Duan ZY, Cai GY, Chen YZ, Liang S, Liu SW, Wu J, Qiu Q, Lin SP, Zhang XG, Chen XM: Aging promotes progression of IgA nephropathy: a systematic review and meta-analysis. Am J Nephrol 2013;38:241–252.
77 Hupe MC, Merseburger AS, Lokesher VB, Eggers H, Rott H, Wegener G, Abbas M, Kuczyk MA, Herrmann TR: Age-an independent prognostic factor of clinical outcome in renal malignancies: results of a large study over two decades. World J Urol 2014;32:115–121.
78 Halvorson CR, Bremmer MS, Jacobs SC: Polycystic kidney disease: inheritance, pathophysiology, prognosis, and treatment. Int J Nephrol Renov Dis 2010;3:69–83.
79 Retnakaran R, Cull CA, Thorne R, Adler AI, Holman RR: Risk Factors for Renal Dysfunction in Type 2 Diabetes: U.K. Prospective Diabetes Study 74. Diabetes 2006;55:1832–1839.
80 Hodgin JB, Bitzer M, Wickman L, Afshinnia F, Wang SQ, O’Connor C, Yang Y, Meadowbrooke C, Chowdhury M, Kikuchi M, Wiggins JE, Wiggins RC: Glomerular Aging and Focal Global Glomerulosclerosis: A Podometric Perspective. J Am Soc Nephrol 2015;26:3162–3178.
81 Yamaguchi M, Ando M, Yamamoto R, Akisya S, Kato S, Katsuno T, Kosugi T, Sato W, Tsuboi N, Yasuda Y, Mizuno M, Ito Y, Matsu S, Maruyama S: Patient Age and the Prognosis of Idiopathic Membranous Nephropathy. PLoS One DOI:10.1371/journal.pone.0110376.
82 Coresh J, Selvin E, Stevens LA, Manz J, Kusek JW, Eggers P, Lente FV, Levey AS: Prevalence of Chronic Kidney Disease in the United States. JAMA 2007;298:2038–2047.
83 O’Hare AM, Choi AI, Bertenthal D, Bacchetti P, Garg AX, Kaufman JS, Walter LC, Mehta KM, Steinman MA, Allon M, McClellan WM, Landefeld CS: Age affects outcomes in chronic kidney disease. J Am Soc Nephrol 2007;18:2758–2765.
84 Barai S, Gambhir S, Prasad N, Sharma RK, Ora M: Functional renal reserve capacity in different stages of chronic kidney disease. Nephrology 2010;15:350–353.
85 Hallan SI, Orth SR: The Conundrum of Chronic Kidney Disease Classification and End-Stage Renal Risk Prediction in the Elderly – What Is the Right Approach? Nephron Clin Pract 2010;116:c307–c316.
86 Glassock RJ, Winearls C: Ageing and the Glomerular Filtration Rate: Truths and Consequences. Trans Am Clin Climatol Assoc 2009;120:419–428.
87 ClinicalTrials.gov. Metformin in Longevity Study (MILES). Available from: https://clinicaltrials.gov/ct2/show/NCT02432287 [Accessed 2017 Jan 18]. | 2025-03-06T00:00:00 | olmocr | {
"pdf_page_numbers": [
[
0,
2139,
1
],
[
2139,
4151,
2
],
[
4151,
8297,
3
],
[
8297,
12162,
4
],
[
12162,
16251,
5
],
[
16251,
18357,
6
],
[
18357,
22854,
7
],
[
22854,
27064,
8
],
[
27064,
30358,
9
],
[
30358,
34823,
10
],
[
34823,
39232,
11
],
[
39232,
42343,
12
],
[
42343,
44497,
13
]
]
} |
d7039110a35c7f9028367ff1a04230c8fc1c8110 | {
"olmocr-version": "0.1.59",
"pdf-total-pages": 7,
"total-fallback-pages": 0,
"total-input-tokens": 26065,
"total-output-tokens": 8527,
"fieldofstudy": [
"Biology"
]
} | Transcription factor Ikaros Represses Protein Phosphatase 2A (PP2A) Expression through an Intronic Binding Site*
Kamalpreet Nagpal 1, Katsue Sunahori Watanabe 1, Betty P. Tsao 5, and George C. Tsokos 1,2
Received for publication, February 13, 2014, and in revised form, March 26, 2014 Published, JBC Papers in Press, April 1, 2014, DOI 10.1074/jbc.M114.558197
Background: PP2A is a serine/threonine phosphatase playing a central role in the pathology of the autoimmune disease SLE.
Results: Ikaros binds to an intronic site in PP2A and modulates its expression.
Conclusion: Ikaros represses PP2A expression by recruiting histone deacetylase HDAC1.
Significance: This study proposes a novel pathway for regulation of PP2A, a critical molecule in SLE pathogenesis.
Protein phosphatase 2A (PP2A) is a highly conserved and ubiquitous serine/threonine phosphatase. We have shown previously that PP2A expression is increased in T cells of systemic lupus erythematosus patients and that this increased expression and activity of PP2A plays a central role in the molecular pathogenesis of systemic lupus erythematosus. Although the control of PP2A expression has been the focus of many studies, many aspects of its regulation still remain poorly understood. In this study, we describe a novel mechanism of PP2A regulation. We propose that the transcription factor Ikaros binds to a variant site in the first intron of PP2A and modulates its expression. Exogenous expression of Ikaros leads to reduced levels of PP2Ac message as well as protein. Conversely, siRNA-enabled silencing of Ikaros enhances the expression of PP2A, suggesting that Ikaros acts as a suppressor of PP2A expression. A ChIP analysis further proved that Ikaros is recruited to this site in T cells. We also attempted to delineate the mechanism of Ikaros-mediated PP2Ac gene suppression. We show that Ikaros-mediated suppression of PP2A expression is at least partially dependent on the recruitment of the histone deacetylase HDAC1 to this intronic site. We conclude that the transcription factor Ikaros can regulate the expression of PP2A by binding to a site in the first intron and modulating chromatin modifications at this site via recruitment of HDAC1.
Systemic lupus erythematosus (SLE)3 is a multifactorial autoimmune disease that primarily affects women in their reproductive years (1). It affects many different organs, like the skin, joints, brain, heart, and kidney. In addition to genetic, environmental, and hormonal factors, immune system irregularities, especially in T cells, is one of the contributing factors toward the pathology of this disease (2). In T cells isolated from SLE patients, aberrant signaling leads to atypical characteristics, such as enhanced tyrosine phosphorylation as well as increased calcium influx (3). One of the signaling defects is the reduced expression of the T cell receptor (TCR)-associated CD3ζ chain. In SLE T cells, the FcRγ structurally and functionally replaces the CD3ζ chain (4). FcRγ couples to spleen tyrosine kinase (syk) instead of ζ-associated protein 70 (ZAP70) kinase, thus resulting in defective signaling (5). Another hallmark of SLE T cells is a reduction in their ability to produce IL-2, an essential cytokine for T cell proliferation and effector functions (6).
One of the key components contributing to the signaling defects in SLE pathogenesis is the serine/threonine protein phosphatase 2A (PP2A). PP2A is a ubiquitously expressed, highly conserved serine/threonine phosphatase that plays a key role in a number of cellular processes like cell division, motility, cytoskeletal dynamics, etc. (7). PP2A has a tripartite structure consisting of the scaffold subunit A and the catalytic subunit C forming the core enzyme and one of the many regulatory subunits binding to the core enzyme to form a functional holoenzyme (8). We have shown previously that PP2A protein and mRNA levels as well as the enzymatic activity of the catalytic subunit are increased in T cells from SLE patients compared with healthy individuals (9). This enhanced expression of PP2A is one of the factors contributing to the molecular defects seen in SLE (10).
We have demonstrated previously that a cAMP response element in the PP2A promoter is hypomethylated in SLE T cells and that the binding of the transcription factors cAMP response element-binding protein and SP1 contribute to the expression levels of PP2A (11). However, additional mechanisms that affect the expression of PP2A are still not clear and warrant an extensive study. In 2011, a genome wide association study (GWAS) study identified an SNP in the first intron of PPP2CA that was associated with SLE (12). The risk allele of this SNP was associated with renal disease, anti-double-stranded DNA, and anti-ribonucleoprotein antibodies. Moreover, PP2A expression was higher in patients carrying this allele, suggesting that this SNP may play a role in regulating PP2A expression.
We hypothesized that a transcription factor binding at this site might be controlling PP2A expression. Here, we report that
Ikaros Represses PP2A Expression
the transcription factor Ikaros binds to this variant site in the first intron of PP2A. Ikaros acts as a repressor, and its binding reduces the expression of PP2A. We further show that Ikaros exerts its repressive effect by the recruitment of the histone deacetylase HDAC1, which maintains the chromatin in a closed conformation, precluding the binding of the transcription machinery. Thus, we demonstrate, for the first time, that Ikaros acts as a repressor for the serine/threonine phosphatase PP2A and, thus, identify a novel means of control of PP2A, a critical molecule involved in lupus pathogenesis.
EXPERIMENTAL PROCEDURES
Purification of T cells—Peripheral blood was collected by venipuncture, and CD3⁺ T cells were purified using a rosette T cell purification kit (Stem Cell Technologies) as described above (11). T cells were maintained in RPMI 1640 medium supplemented with 10% fetal calf serum and 1% penicillin-streptomycin (Sigma-Aldrich) and maintained in a humidified incubator (37 °C, 5% CO₂). All studies were approved by the institutional review board (Committee on Clinical Investigations) at the Beth Israel Deaconess Medical Center, and informed consent was obtained from all participants.
Oligonucleotide Pulldown Assay—Custom-synthesized, biotin-labeled WT as well as control oligos were purchased from Integrated DNA Technologies. Sequences of the oligos were as follows: WT oligonucleotide, 5’-TTCCAGCCTCCTCCCTC-CCAAAGCACCAGGATTG-3’; control oligonucleotide, 5’-ACCTGACGCCTAAGACGCTTACTCGCCTCGCGC-3’. The pulldown assay was carried out using streptavidin magnetic beads (PureBiotech LLC, Middlesex, NJ) as described before (13). The eluates were run on a 4–20% BisTris NuPAGE precast gel (Invitrogen), transferred to a PVDF membrane, and blotted with anti-Ikaros antibody (Santa Cruz Biotechnology, Inc.).
Luciferase Assay—PP2A core promoters followed by 384 bases spanning the Ikaros binding site in the first intron were cloned upstream of the luciferase gene in the pGL3 reporter vector. Two million 293T cells were transfected with either the reporter plasmid by itself (250 ng) or in combination with different amounts of the Ikaros expression vector (50, 100, or 200 ng) using Lipofectamine 2000 (Invitrogen) according to the instructions of the manufacturer. Twenty-four hours after transfection, cells were collected and lysed using passive lysis buffer (Promega), and luciferase activity was quantified by using the Dual-Luciferase assay system (Promega) according to the instructions of the manufacturer.
Site-directed Mutagenesis of the Reporter Plasmid—The luciferase reporter plasmid described above was mutated using Pfu Turbo polymerase (Stratagene). The mutagenesis primers were as follows: Del-1, 5’-CCTACAAATCTCTGGCTGGAGAGTCTCAGAACGCTC-3’; Del-2, 5’-CTTACATACTCCATTTGCTGTTCTCGCCGAGCCTGAGCTC-3’; mutant, 5’-CTTACATACTCCATTTGCTGTTCTCGCCGAGCCTGAGCTC-3’. The reverse primer in each case was the respective reverse complement of the forward primer.
Reporter ChIP and ChIP Assays—293T cells were transfected with either only the wild-type or the mutant reporter (described above) or the reporter in combination with the Ikaros expression plasmid using Lipofectamine 2000. Twenty-four hours after transfection, cells were collected, and the ChIP assay was performed using the MAGnify ChIP kit (Invitrogen) according to the instructions of the manufacturer. Briefly, cells were fixed for 10 min with 1% formaldehyde to cross-link DNA-protein and protein-protein complexes. The cross-linking reaction was stopped using 1.25 m glycine for 5 min. The cells were lysed, sonicated to shear DNA, and sedimented, and then the diluted supernatants were incubated overnight with the respective antibodies. 10% of the diluted supernatants were saved as “inputs” for normalization. Several washing steps were followed by protein digestion using proteinase K. Reverse cross-linking was carried out at 65 °C. DNA was subsequently purified and amplified by quantitative PCR on a LightCycler 480 real-time PCR system (Roche) using specific primers flanking the intronic site and the luciferase gene (reverse primer). Threshold cycle (i.e. Ct) values were used to calculate relative mRNA expression by the ΔΔCt relative quantification method.
In the case of regular ChIP assays, 5 million freshly isolated primary T cells were used for each antibody/sample. There was no transfection, and endogenous proteins were used for immunoprecipitation of the chromatin. The ChIP assay was carried out as described above using the MAGnify ChIP kit. The primers used were as follows: 5’-CAGAATTTGATGATACAGAACATTGA-3’ (forward) and 5’TAGAACAAGGTGGTCTCCTACATCAT-3’ (reverse). The antibodies used for the ChIP assays were anti-Ikaros (Santa Cruz Biotechnology, Inc.) and anti-HDAC1 (Millipore).
Transfections in T Cells—Plasmid or siRNA transfections in primary T cells were carried out using the Nucleofector system (Lonza). Five million freshly isolated T cells were resuspended in 100 µL of Nucleofector solution, and the respective amount of plasmid or siRNA was added. For most studies, 0.5, 1, or 2 µg of the plasmid/1 million cells was used. In the case of siRNA, either 5 or 10 nM final concentration was used. Cells were transfected using the U-014 program and rescued immediately in prewarmed RPMI medium supplemented with 10% fetal calf serum and 1% penicillin-streptomycin. For plasmid transfections, cells were harvested 36 h post-transfection for real-time and protein analysis. For silencing studies, 72 h after siRNA transfection, cells were harvested for further analysis.
Real-time PCR Analysis—After harvesting the cells, total RNA was isolated using the RNeasy Plus kit (Qiagen). 300 ng of the total RNA was reverse-transcribed into cDNA using RNA-to-cDNA premix (Clontech). Real-time PCR amplification was carried out with SYBR Green I using LightCycler 480 (Roche). Threshold cycles (Ct values) were used to calculate relative mRNA expression by the relative quantification method.
Coimmunoprecipitation Assay—293T cells were transfected with the various combinations of plasmids (0.5 µg each) using Lipofectamine 2000. Twenty-four hours after transfection, the cells were lysed in 1 ml of radioimmune precipitation assay buffer (Boston BioProducts) supplemented with EDTA-free complete protease inhibitor mixture (Roche). After spinning down the cells, the supernatants were incubated with 1 µg of Ikaros antibody (catalog no. H-100X, Santa Cruz Biotechnology, Inc.) for 2 h at 4 °C. 10% of the supernatant was saved as
RESULTS
IKZF1 Binds to a Specific Site in the First Intron of PP2Ac—Elevated levels of PP2A have been observed in T cells isolated from lupus patients, and this increased expression is a contributing factor in a number of molecular abnormalities. In addition to the already known factors that affect the expression of PP2A, a GWAS study by Tan et al. (12) revealed a genetic variant in the first intron of PPP2CA that was not only positively associated with SLE but also affected the expression levels of PP2A in patients (Fig. 1A). Thus, we hypothesized that a transcription factor binding at this site might be responsible for the observed changes in PP2A levels. Using an in silico transcription factor binding search, we identified Ikaros (IKZF1) as a putative factor binding at this specific intronic site. To confirm the binding of Ikaros to this site, we used a custom-synthesized, biotin-conjugated DNA oligonucleotide spanning the particular site. A random oligonucleotide of similar length served as a control. The oligos were incubated with Jurkat cell nuclear extracts, and streptavidin magnetic beads were used to pull down the interacting proteins. The eluates were run on a gel, and the membrane was probed for Ikaros. As shown in Fig. 1B, Ikaros was pulled down specifically with the oligo corresponding to the PP2A intronic site. The control oligo was unable to bring down Ikaros protein, suggesting that the binding of Ikaros to this site is specific.
Ikaros Represses PP2A Expression—To determine whether the binding of Ikaros to this site has any functional consequences, we employed a luciferase reporter system. The PP2A core promoter, followed by a 384-base pair region surrounding the intronic site, was cloned upstream of a gene encoding firefly luciferase (Fig. 2A). The expression of the luciferase gene is thus under direct control of the PP2A promoter as well as the intronic region. As compared with the empty vector, the PP2A reporter vector had significant activity (Fig. 2B). However, the exogenous expression of Ikaros, achieved by cotransfecting the expression plasmid along with the reporter vector, reduced the reporter activity to about half. Moreover, Ikaros had a dose-dependent effect, with increasing amounts of the expression plasmid leading to an even greater reduction in reporter activity.
To further prove that it is indeed the binding of Ikaros to the intronic region of PP2A that is responsible for the reduction in reporter activity, we generated a mutant version of the reporter plasmid, with the Ikaros binding site deleted. As compared with the wild-type reporter, the mutant reporter had enhanced reporter activity. Furthermore, in contrast to the wild-type construct, the mutant reporter activity was unaffected by the
Ikaros Represses PP2A Expression
expression of Ikaros (Fig. 2C). This suggests that the binding of Ikaros to the specific site in the first intron of PP2A results in reduction of PP2A reporter activity.
Ikaros Represses the Expression of PP2Ac—The luciferase reporter assays with the transient expression of Ikaros suggest that Ikaros acts as a repressor of PP2A induction. Thus, we asked whether it can regulate the expression of PP2A. To this end, we overexpressed Ikaros in CD3+ T cells isolated from healthy individuals. After 36 h in culture, cells were collected and processed for immunoblotting. Forced expression of Ikaros led to a significant decrease in the protein levels of PP2A (Fig. 3A, right panel). Moreover, this decrease was dose-dependent. Increasing the amount of Ikaros transfected into the cells led to an even greater decrease in PP2A expression. To determine whether Ikaros represses PP2A expression at a transcriptional level, we assessed the mRNA levels of PP2A in cells expressing Ikaros. PP2A mRNA levels were similarly reduced in response to Ikaros overexpression (Fig. 3A, left panel). We next investigated the effect of Ikaros silencing on PP2A expression. siRNA-enabled silencing of Ikaros led to increased levels of PP2A message as well as protein (Fig. 3B, left and right panels, respectively). Thus, Ikaros acts as a repressor for PP2A expression.
Ikaros Is Recruited to the First Intron in PP2A—We employed ChIP assays to confirm the recruitment of Ikaros to this particular site in a cellular setting. Reporter ChIP assays were performed in 293T cells. The cells were transfected with the PP2A promoter-intron reporter construct (described above) in combination with either the empty vector (pCMV) or an Ikaros expression vector. Twenty-four hours post-transfection, DNA-protein complexes were cross-linked and cells were lysed, sonicated, and immunoprecipitated with Ikaros antibody or a control IgG. Immunoprecipitated DNA was purified, and the presence of the specific intronic DNA was quantified using real-time PCR analysis. We observed a significantly increased recruitment of Ikaros to the intron of PP2A in the Ikaros-transfected samples compared with the pCMV-transfected cells, suggesting the recruitment of Ikaros to this site in the first intron of PP2A (Fig. 4B, left panel). We also mutated the wild-type reporter to generate three mutants (depicted in Fig. 4A) to test the specificity of this site. Two of these mutants were deletion mutants where either only the core Ikaros binding site (Del-2) or a few bases in addition to the core site were deleted (Del-1), and the third mutant had some of the bases changed, leaving the Ikaros binding site partially intact. We observed that both deletion mutants were unable to recruit Ikaros, suggesting that the absence of this site precludes the binding of Ikaros (Fig. 4B, right panel). With the mutant where some of the bases were changed, leaving the Ikaros binding site partially intact, we could see a reduction in recruitment compared with the wild-type reporter but not a complete abrogation. This suggests that an intact Ikaros binding site is necessary for the recruitment of Ikaros to the PP2A intron. We also wanted to verify the recruitment of Ikaros to the endogenous PPP2CA gene in T cells by ChIP assays in primary T cells. In this case, we immunoprecipitated the endogenous Ikaros protein using an Ikaros-specific antibody and analyzed its recruitment to the PP2A intron in a more physiologically relevant setting. We observed enhanced recruitment to this site in the case of Ikaros-specific antibody in comparison with control IgG, thus confirming that Ikaros is recruited to the intron of the endogenous PPP2CA gene.
Ikaros-mediated Repression of PP2A Is Dependent on the Recruitment of HDAC1—Ikaros is a zinc finger transcription factor that also harbors a dimerization domain at its carboxyl terminal through which it can recruit many different proteins and transcriptional regulators, including chromatin-modifying complexes like Sin3A, Sin3B, etc. (14). One of the proteins with which it is known to interact is histone deacetylase 1 or
FIGURE 4. Ikaros is recruited to the specific site in the intron of PP2A. A schematic of the mutant (Mut) reporters used in these experiments. B, reporter ChIP in 293T cells. 293T cells were transfected with either only the wild type (left panel) or the mutant reporter (right panel) or the reporter in combination with the Ikaros expression plasmid using Lipofectamine 2000. Twenty-four hours after transfection, cells were collected, and a ChIP assay was performed using the MAGnify ChIP kit. The region spanning the specific intron site was amplified by quantitative PCR and normalized to the values obtained from the input DNA. The graph shows mean ± S.D. of three observations.
HDAC1. Thus, we asked whether the observed repressor activity of Ikaros in the case of PP2A could be due to the recruitment of HDAC1 to this particular intronic site. We first tested the interaction between Ikaros and HDAC1. Transient transfections in HEK 293T cells, followed by immunoprecipitation assays were carried out to address this question. As shown in Fig. 5A, HDAC1 could be coimmunoprecipitated with Ikaros, indicating that Ikaros and HDAC1 do indeed physically interact. Using biotinylated oligos, we also show that HDAC1 binds to the particular site in the first intron of PP2A (Fig. 5B). In both reporter ChIP as well as endogenous ChIP in primary T cells, we could see increased recruitment of HDAC1 to the intronic site compared with control IgG (Fig. 5C, left and right panels, respectively), thus confirming that HDAC1 does bind to this site in PP2A intron 1.
Having established that HDAC1 interacts with Ikaros and is recruited to this site, we wanted to prove that the repression of
FIGURE 5. Ikaros-mediated repression of PP2A expression is dependent on HDAC1. A, a communoprecipitation (IP) assay showing the binding of Ikaros with HDAC1. 293T cells were transfected with the various combinations of plasmids using Lipofectamine 2000. Twenty-four hours after transfection, the cells were lysed, and the supernatants were incubated with Ikaros antibody for 2 h at 4 °C. After the preincubation with the antibody, agarose A/G beads were added to each sample and incubated overnight at 4 °C. The immunoprecipitates were subsequently run on a gel, transferred to a PVDF membrane, and blotted for the indicated proteins. The saved inputs were also run on the gel to confirm equal expression of proteins in all samples. IB, immunoblot. B, a biotin-conjugated, intron-specific oligo (sp. oligo) or a random control oligo were used in pulldown assays with Jurkat cell nuclear extracts. The eluates were run on a gel and probed with anti-HDAC1 antibody. C, reporter ChIP (left panel) and ChIP (right panel) in 293T and primary T cells, respectively. 293T cells (left panel) were transfected with the reporter in combination with the Ikaros and HDAC1 expression plasmids using Lipofectamine 2000. Twenty-four hours after transfection, cells were collected, and a ChIP assay was performed using the MAGnify ChIP kit. For ChIP with endogenous protein in primary T cells (right panel), 5 million freshly isolated primary T cells were used for each antibody/sample. The ChIP assay was carried out using the MAGnify ChIP kit. The region spanning the specific intron site was amplified by quantitative PCR and normalized to the values obtained from the input DNA. The graph shows mean ± S.D. of three observations.
Ikaros Represses PP2A Expression
PP2A expression that we see is due to HDAC1. To this effect, we used two different approaches. Using siRNA, we silenced HDAC1 in cells overexpressing Ikaros and assessed the mRNA levels of PP2Ac. As expected, the overexpression of Ikaros led to a reduction in PP2A mRNA levels. However, siRNA-mediated silencing of HDAC1 in these cells restored the PP2A levels to about 50% of the original, suggesting that, in the absence of HDAC1, Ikaros-mediated PP2A repression is not as competent (Fig. 5D). Secondly, we used an HDAC1 inhibitor, trichostatin A to confirm this. T cells were transfected with Ikaros, followed by a 12-hour treatment with trichostatin A. The treatment of cells with trichostatin A led to the complete restoration of PP2A expression (Fig. 5E). Thus, the repression of PP2A expression seen upon Ikaros overexpression is at least partly due to the recruitment of HDAC1 to the same site. HDAC1 recruitment to this site might influence the accessibility of transcription factors to the promoter, thus affecting gene expression of PP2A.
DISCUSSION
In this study we propose a novel mechanism for regulating the expression of PP2A, a serine/threonine phosphatase that is central to the molecular defects seen in the pathogenesis of SLE. We show, for the first time, that the expression of PP2A can be controlled by the binding of transcription factor Ikaros to a specific site in the first intron. Ikaros acts as a repressor for PP2A expression, and this can be explained by the ability of Ikaros to bind and recruit the histone deacetylase HDAC1 to this particular site. Thus, we provide evidence for a previously unknown regulatory mechanism of one of the most important factors involved in SLE.
T cells of SLE patients have higher levels of PP2A. This enhanced expression as well as activity translates to defects in the T cell signaling machinery, such as reduced expression of CD3ζ as well as decreased IL-2 production. We have also demonstrated recently that higher levels of PP2A might be one of the factors contributing to hypomethylation of DNA in patients, another hallmark of SLE pathogenesis (15). Although PP2A expression is clearly a significant aspect controlling the outcome of T cell signaling in SLE, not many studies have focused on the mechanisms affecting the expression of PP2A. We not only bring to the forefront the transcriptional control of PP2A, but also delineate a novel pathway for its regulation.
Conventionally, transcriptional regulation involves various factors binding to the promoter or the initiation site of the protein coding sequence. Nevertheless, there are countless studies exemplifying the instances where transcription factor binding to the introns can influence the expression of the gene, and intron-mediated enhancement of genes is a well studied branch of gene expression (16). A recent study by Hoffmann et al. (17) describes an intronic enhancer that promotes the expression of the UCP3 gene by the binding of SP1/SP3 transcription factors. Another report demonstrates the regulation of fibroblast growth factor receptor 3 by an intronic element (18). The intron-mediated control of genes is not limited to transcriptional enhancement. In fact, Ikaros itself has been shown to bind to an intronic site in the opioid receptor gene and cause its repression (19). How exactly an intronic element affects gene expression is not very clear. Perhaps the looping and folding of DNA brings these distal enhancer/repressor elements closer to the regulatory sequences of the promoter of the respective genes.
Ikaros (IKZF1) belongs to a family of zinc finger transcription factors (20). It binds to the regulatory elements of its target genes in a sequence-dependent manner. Ikaros can act as an activator or repressor of transcription, depending on which proteins it binds and recruits via its C terminus (21). Its repressive activity can be classified into three main categories: chromatin modification, corepressor recruitment, and competition with transcriptional activators (22). In the case of PP2A, we find that Ikaros recruits HDAC1 to this specific site in the first intron of PP2A. The repressive activity is dependent on the presence of HDAC1 because its silencing or the chemical inhibition of its activity restores the expression of PP2A. However, we believe that HDAC1 is not the sole factor responsible for Ikaros-mediated repression of PP2A because we only see a partial restoration of PP2A levels. Additional proteins/factors that might be involved in the regulation of PP2A levels remain to be investigated.
Because our study demonstrates that the binding of Ikaros to an intronic element reduces the levels of PP2A, it is conceivable that, in SLE patients, binding of Ikaros is compromised, thus giving rise to enhanced PP2A levels. However, the question remains why and how Ikaros differentiates between patients and healthy individuals. Is it the expression of Ikaros itself that controls its binding and, hence, the differential expression of PP2A? This would suggest that, in SLE patients, the expression of Ikaros would be decreased compared with normal controls. Indeed, a report by Hu et al. (24) does imply that in peripheral blood mononuclear cells from SLE patients, Ikaros mRNA levels are reduced (23). Whether this difference in expression of Ikaros translates to the regulation of PP2A remains to be seen.
In conclusion, we have shown that Ikaros is a modulator of PP2A expression. It binds to a site in the first intron of PPP2CA and reduces the expression levels of PP2A. The repressive ability of Ikaros depends, at least in part, on the recruitment of the histone deacetylase HDAC1, thus affecting the accessibility to the chromatin. Thus, we define a novel mechanism for the regulation of PP2A, a key component of SLE pathogenesis.
REFERENCES
1. Tsokos, G. C. (2011) Systemic lupus erythematosus. N. Engl. J. Med. 365, 2110–2121
2. Grammatikos, A. P., and Tsokos, G. C. (2012) Immune deficiency and autoimmunity: lessons from systemic lupus erythematosus. Trends Mol. Med. 18, 101–108
3. Moulton, V. R., and Tsokos, G. C. (2011) Abnormalities of T cell signaling in systemic lupus erythematosus. Arthritis Res. Ther. 13, 207
4. Krishnan, S., Warke, V. G., Nambiar, M. P., Tsokos, G. C., and Farber, D. L. (2003) The FcRγ subunit and Syk kinase replace the CD3ζ chain and ZAP-70 kinase in the TCR signaling complex of human effector CD4 T cells. J. Immunol. 170, 4189–4195
5. Krishnan, S., Farber, D. L., and Tsokos, G. C. (2003) T cell rewiring in differentiation and disease. J. Immunol. 171, 3325–3331
6. Solomou, E. E., Juang, Y. T., Gourley, M. F., Kammer, G. M., and Tsokos, G. C. (2001) Molecular basis of deficient IL-2 production in T cells from patients with systemic lupus erythematosus. J. Immunol. 166, 4216–4222
7. Cohen, P. T., Brewis, N. D., Hughes, V., and Mann, D. J. (1990) Protein serine/threonine phosphatases: an expanding family. FEBS Lett. 268, 13756 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 289 • NUMBER 20 • MAY 16, 2014
8. Mumby, M. C., and Walter, G. (1993) Protein serine/threonine phosphatases: structure, regulation, and functions in cell growth. Physiol. Rev. 73, 673–699
9. Katsiari, C. G., Kyttaris, V. C., Juang, Y. T., and Tsokos, G. C. (2005) Protein phosphatase 2A is a negative regulator of IL-2 production in patients with systemic lupus erythematosus. J. Clin. Invest. 115, 3193–3204
10. Juang, Y. T., Wang, Y., Jiang, G., Peng, H. B., Ergin, S., Finnell, M., Magilavy, A., Kyttaris, V. C., and Tsokos, G. C. (2008) PP2A dephosphorylates Elf-1 and determines the expression of CD3ε and FcγRI in human systemic lupus erythematosus T cells. J. Immunol. 181, 3658–3664
11. Sunahori, K., Juang, Y. T., and Tsokos, G. C. (2009) Methylation status of CpG islands flanking a cAMP response element motif on the protein phosphatase 2Ac promoter determines CREB binding and activity. J. Immunol. 182, 1500–1508
12. Tan, W., Sunahori, K., Zhao, J., Deng, Y., Kaufman, K. M., Kelly, J. A., Langefeld, C. D., Williams, A. H., Comeau, M. E., Ziegler, J. T., Marion, M. C., Bae, S. C., Lee, J. H., Lee, J. S., Chang, D. M., Song, Y. W., Yu, C. Y., Kimberly, R. P., Edberg, J. C., Brown, E. E., Petri, M. A., Ramsey-Goldman, R., Vilá, L. M., Reveille, J. D., Alarcón-Riquelme, M. E., Harley, I. B., Boackle, S. A., Stevens, A. M., Scofield, R. H., Merrill, J. T., Freedman, B. L., Anaya, J. M., Criswell, L. A., Jacob, C. O., Vyse, T. J., Niewold, T. B., Gaffney, P. M., Moser, K. L., Kastner, P., Chan, S. (2011) Ikaros in B cell development and function. World J. Biol. Chem. 2, 132–139
13. He, C. F., Liu, Y. S., Cheng, Y. L., Gao, J. P., Pan, T. M., Han, J. W., Quan, C., Sun, L. D., Zheng, H. F., Zuo, X. B., Xu, S. X., Sheng, Y. J., Yao, S., Hu, W. L., Li, Y., Yu, Z. Y., Yin, X. Y., Zhang, X. J., Cui, Y., and Yang, S. (2010) TNIP1, SLC15A4, ETS1 and IKZF1 are associated with clinical features of systemic lupus erythematosus in a Chinese Han population. Lupus 19, 1181–1186
14. Hu, W., Sun, L., Gao, J., Li, Y., Wang, P., Cheng, Y., Pan, T., Han, J., Liu, Y., Lu, W., Zuo, X., Sheng, Y., Yao, S., He, C., Yu, Z., Yin, X., Cui, Y., Yang, S., and Zhang, X. (2011) Down-regulated expression of IKZF1 mRNA in peripheral blood mononuclear cells from patients with systemic lupus erythematosus. Rheumatol. Int. 31, 819 – 822
Ikaros Represses PP2A Expression | 2025-03-05T00:00:00 | olmocr | {
"pdf_page_numbers": [
[
0,
5089,
1
],
[
5089,
11682,
2
],
[
11682,
14459,
3
],
[
14459,
18618,
4
],
[
18618,
22026,
5
],
[
22026,
29102,
6
],
[
29102,
31447,
7
]
]
} |
a10984cb409c26726107b11f4a1eb36e5e8b92c8 | {
"olmocr-version": "0.1.59",
"pdf-total-pages": 4,
"total-fallback-pages": 0,
"total-input-tokens": 11913,
"total-output-tokens": 3247,
"fieldofstudy": [
"Physics"
]
} | UNDERSTANDING THE PHYSICS OF THE X-FACTOR
S. C. O. Glover\textsuperscript{1} and M.-M. Mac Low\textsuperscript{2}
Abstract. We study the relationship between the H\textsubscript{2} and CO abundances in simulated molecular clouds using a fully dynamical model of magnetized turbulence coupled to a detailed chemical network. We find that the CO-to-H\textsubscript{2} conversion factor for a given molecular cloud, the so-called X-factor, is determined primarily by the mean extinction of the cloud, rather than by its metallicity. Our results explain the discrepancy observed in low metallicity systems between cloud masses derived from CO observations and other techniques such as infrared emission, and predict that CO-bright clouds in low metallicity systems should be systematically larger and/or denser than Milky Way clouds.
1 Introduction
Observed star formation takes place within giant molecular clouds (GMCs), so understanding how these clouds form and evolve is a key step towards understanding star formation. The main chemical constituent of any GMC is molecular hydrogen (H\textsubscript{2}), but this is very difficult to observe \textit{in situ}. For this reason it is common to use emission from carbon monoxide (CO) as a proxy for H\textsubscript{2}.
In order to do this, however, it is necessary to understand the relationship between the distributions of the H\textsubscript{2} and the CO. They have rather different formation mechanisms: H\textsubscript{2} forms on the surface of dust grains, while CO forms in the gas-phase as a product of ion-neutral chemistry. Both are readily photodissociated by ultraviolet (UV) radiation, but H\textsubscript{2} can protect itself from this radiation via self-shielding even in relatively low column density gas (Draine & Bertoldi \textsuperscript{1996}).
We therefore expect to find large variations in the CO/H\textsubscript{2} ratio within any given GMC, with the H\textsubscript{2} filling a significantly larger volume of the cloud than the CO.
Despite this, there is good evidence that CO \textit{emission} is a good tracer of H\textsubscript{2} mass within the Milky Way (see e.g. Solomon \textit{et al.} \textsuperscript{1987}, Dame \textit{et al.} \textsuperscript{2001}). A
number of independent studies have shown that GMCs in the Galactic disk show a good correlation between the integrated intensity of the $J = 1 \rightarrow 0$ rotational transition line of $^{12}\text{CO}$ and the $\text{H}_2$ column density. This correlation is typically described in terms of a conversion factor $X_{\text{CO}}$ (the ‘X-factor’), given by
$$X_{\text{CO}} = \frac{N_{\text{H}_2}}{W_{\text{CO}}} \simeq 2 \times 10^{20} \text{cm}^{-2} \text{K}^{-1} \text{km}^{-1} \text{s}, \quad (1.1)$$
where $W_{\text{CO}}$ is the velocity-integrated intensity of the CO $J = 1 \rightarrow 0$ emission line, averaged over the projected area of the GMC, and $N_{\text{H}_2}$ is the mean $\text{H}_2$ column density of the GMC, averaged over the same area.
However, the issue of the environmental dependence of $X_{\text{CO}}$ remains highly contentious. Extragalactic measurements of $X_{\text{CO}}$ that use a virial analysis to determine cloud masses find values for $X_{\text{CO}}$ that are similar to those obtained in the Milky Way, with at most a weak metallicity dependence (e.g. Rosolowsky et al. 2003, Bolatto et al. 2008). On the other hand, measurements that constrain GMC masses using techniques that do not depend on CO emission consistently find values for $X_{\text{CO}}$ that are much larger than the Galactic value and that are suggestive of a strong metallicity dependence (e.g. Israel 1997, Leroy et al. 2009).
Numerical simulations provide us with one way to address this observational dichotomy. If we can understand the distribution of CO and $\text{H}_2$ in realistic models of GMCs, then we may begin to understand why the different types of observation give such different results. In this contribution, we summarize the results from some of our recent numerical simulations that self-consistently model both the chemistry and the turbulent dynamics of the gas within GMCs, and discuss what they can tell us about the behaviour of $X_{\text{CO}}$.
2 Method
We have performed a large number of simulations of the chemical and thermal evolution of the turbulent, dense interstellar medium using a modified version of the ZEUS-MP magnetohydrodynamical code. Our modifications include the addition of a simplified treatment of hydrogen, carbon and oxygen chemistry, a detailed atomic and molecular cooling function, and a treatment of the effects of ultraviolet radiation using a six-ray approximation. Full details of these modifications can be found in Glover et al. (2010).
Our simulations begin with initially uniform atomic gas, threaded by a uniform magnetic field with strength $B_0 = 5.85 \mu\text{G}$. The initial velocity field is turbulent, with power concentrated on large scales, and with an initial rms velocity of $5 \text{km} \text{s}^{-1}$. We drive the turbulence so as to maintain approximately the same rms velocity throughout the simulation. We adopt periodic boundary conditions for the gas and in most cases use a cubical simulation volume with a side length $L = 20 \text{pc}$. In a few simulations, we adopt a smaller-sized box, with $L = 5 \text{pc}$. We have run simulations with a variety of mean densities and metallicities, in order to span a range of different physical conditions. Full details of these simulations can be found in Glover & Mac Low (2010).
Fig. 1. Estimate of the CO-to-H$_2$ conversion factor $X_{\text{CO,est}}$, plotted as a function of the mean visual extinction of the gas, $\langle A_V \rangle$. At $\langle A_V \rangle > 3$, the values we find are consistent with the value of $X_{\text{CO}} = 2 \times 10^{20}$ cm$^{-2}$ K$^{-1}$ km$^{-1}$ s determined observationally for the Milky Way by Dame et al. (2001), indicated in the plot by the horizontal dashed line. At $\langle A_V \rangle < 3$, we find evidence for a strong dependence of $X_{\text{CO,est}}$ on $\langle A_V \rangle$. The empirical fit given by Equation 3.1 is indicated as the dotted line in the Figure, and demonstrates that at low $\langle A_V \rangle$, the CO-to-H$_2$ conversion factor increases roughly as $X_{\text{CO,est}} \propto A_V^{-1.5}$.
3 Results
Because the CO in many of our simulations is optically thick, an accurate determination of $W_{\text{CO}}$ would require a full non-LTE radiative transfer calculation, a complex undertaking outside the scope of our present study. Instead, we make use of a simpler procedure to determine an estimate for the CO-to-H$_2$ conversion factor, denoted as $X_{\text{CO,est}}$. We first select a set of independent sightlines through our simulation, one per resolution element. We next compute H$_2$ and CO column densities along each of these sightlines. We convert each of the CO column densities into an estimate of the optical depth of the gas in the CO J = 1 $\rightarrow$ 0 transition, under the assumptions that (a) the CO level populations are in LTE, (b) the gas is isothermal, with a temperature equal to the CO-weighted mean temperature found in the actual simulation, and (c) the CO linewidth is uniform, and is given by $\Delta v = 3$ km s$^{-1}$. Given an estimate for the CO optical depth, we then compute an estimate for $W_{\text{CO}}$ using the same technique as in Pineda et al. (2008). Finally, we average over all the sightlines to compute a mean intensity $\langle W_{\text{CO}} \rangle$ for the simulation, and do the same for the H$_2$ to arrive at a mean H$_2$ column density. $X_{\text{CO,est}}$ is then simply the ratio of these two values.
Using this procedure, we have computed $X_{\text{CO,est}}$ for each of our simulations. Figure 1 shows how the values we obtain depend on the mean visual extinction of the gas, $\langle A_V \rangle$. We see from the figure that there is a clear change in the behaviour of $X_{\text{CO,est}}$ at $\langle A_V \rangle \sim 3$. In clouds with larger mean extinctions, $X_{\text{CO,est}}$ is roughly
constant and has a value consistent with the observationally-determined value for the Milky Way. On the other hand, for smaller mean extinctions, $X_{\text{CO, est}}$ increases sharply with decreasing $\langle A_V \rangle$. This behaviour is caused by a rapid fall-off in the CO abundance with decreasing mean extinction, which leads to a corresponding sharp drop in $W_{\text{CO}}$. Because it does not self-shield efficiently, CO molecules are protected from photodissociation primarily by dust extinction, and so as this decreases, the CO abundance decreases much more rapidly than the $H_2$ abundance, resulting in the rapid increase we find for $X_{\text{CO, est}}$. The dependence of $X_{\text{CO, est}}$ on $\langle A_V \rangle$ can be described by the empirical fitting function
$$X_{\text{CO, est}} \approx \begin{cases} 2.0 \times 10^{20} & A_V > 3.5 \\ 2.0 \times 10^{20} (A_V/3.5)^{-3.5} & A_V < 3.5 \end{cases}$$
illustrated in Figure 1 by the dotted line.
The sharp fall-off in $W_{\text{CO}}$ with decreasing mean extinction has an important consequence. In order to detect CO emission from GMCs in Local Group galaxies, the integrated intensity of the emission must be $\sim 1 \text{ K km s}^{-1}$ or higher, and we find in our simulations that only the clouds with $\langle A_V \rangle > 1$ have integrated intensities above this value. Therefore, GMCs detectable in CO will always sit on the right-hand side of Figure 1 in the regime where $X_{\text{CO, est}}$ is roughly constant, providing a simple explanation for why values of $X_{\text{CO}}$ in extragalactic systems derived using CO observations are always roughly the same as the Galactic value. On the other hand, GMCs with mass determinations that do not rely on CO are not constrained to fall in the regime where $W_{\text{CO}}$ is large, and so may occur anywhere in the plot, thereby explaining why they are often found to have X-factors that are much larger than the Galactic value. A further consequence of the behaviour of $X_{\text{CO}}$ and $W_{\text{CO}}$ is the prediction that CO-bright clouds in low metallicity systems must be larger and/or denser than their Milky Way counterparts, since at lower metallicity a larger surface density of gas is required to produce the necessary mean extinction.
References
Bolatto, A. D., Leroy, A. K., Rosolowsky, E., Walter, F., & Blitz, L. 2008, ApJ, 686, 948
Dame, T. M., Hartmann, D., and Thaddeus, P. 2001, ApJ, 547, 792
Draine, B. T., and Bertoldi, F. 1996, ApJ, 468, 269
Glover, S. C. O., and Mac Low, M.-M. 2010, MNRAS, in press; arXiv:1003.1340.
Glover, S., Federrath, C., Mac Low, M.-M., and Klessen, R. S. 2010, MNRAS, 404, 2
Israel, F. P. 1997, A&A, 328, 471
Leroy, A. K., Bolatto, A., Bot, C., Engelbracht, C. W., Gordon, K., Israel, F. P., Rubio, M., Sandstrom, K., and Stanimirović, S. 2009, ApJ, 702, 352
Pineda, J. E., Caselli, P., and Goodman, A. A. 2008, ApJ, 679, 481
Rosolowsky, E., Engargiola, G., Plambeck, R., and Blitz, L. 2003, ApJ, 599, 258
Solomon, P. M., Rivolo, A. R., Barrett, J., and Yahil, A. 1987, ApJ, 319, 730 | 2025-03-06T00:00:00 | olmocr | {
"pdf_page_numbers": [
[
0,
2252,
1
],
[
2252,
5569,
2
],
[
5569,
8123,
3
],
[
8123,
11198,
4
]
]
} |
14f272dd016995b8386222f223c939d71cfc7b3c | {
"olmocr-version": "0.1.59",
"pdf-total-pages": 28,
"total-fallback-pages": 0,
"total-input-tokens": 69140,
"total-output-tokens": 17866,
"fieldofstudy": [
"Computer Science"
]
} | Systems, Actors and Agents:
Operation in a multicomponent environment
Mark Burgin
University of California, Los Angeles
405 Hilgard Ave.
Los Angeles, CA 90095
Abstract. In this paper, we further develop multi-agent approach by creating new types of system models. The problem is that conventional models of multi-agent and multicomponent systems implicitly or explicitly assume existence of the absolute time or even do not include time in the set of defining parameters. However, it is rationalized theoretically and validated experimentally that there are different times and time scales in a variety of real systems – physical, chemical, biological, social, etc. Thus, the goal of this paper is construction of multi-agent and multicomponent system models with concurrency of processes and diversity of actions. To achieve this goal, a mathematically based system actor model is elaborated and its properties are studied.
Keywords: time, system, actor, agent, action, process, interaction, environment
1. Introduction
Multi-agent approach is becoming more and more popular in the area of computing, networking, artificial intelligence, robotics, distributed control, resource management, collaborative decision support systems, data mining, etc. (Buşoniu, et al, 2010; Shoham and Leyton-Brown, 2008; Vlassis, 2007; Weiss, 1999). Usually, it is assumed that a multi-agent system is a group of autonomous, interacting entities sharing a common environment, which they perceive with sensors and upon which they act with actuators.
Time is a critically important characteristic of any real-life system. However, not all features of time are adequately presented in the conventional multi-agent models.
If we analyze existing approaches and directions in the multi-agent approach, we can see that in all dynamic models of multi-agent systems, either time is implicitly induced by actions of agents and system states or it is explicitly assumed that unique time exists for the whole system. An archetypal
example of this situation is the absolute Newtonian time in the physical universe, which is innate for the entire classical physics.
However, relativity theory and various experiments disproved this assumption bringing forth the concept of local time (cf., for example, (Einstein, et al, 1923)). The system theory of time extends this principle much further (Burgin, 1992; 2002). Other researchers also advocated existence of different times or different time scales in their theories (cf., for example, (Prigogine, 1980; Barwise and Seligman, 1997)). Besides, as Norbert Wiener (1961) writes, one of the most famous philosophers of the 20th century Bergson lays special emphasis on the distinction between the reversible time of physics, in which nothing new happens, and the irreversible time of evolution and biology, in which there is always something new (Bergson, 1910). In spite of this, time in general systems theory is similar to time in classical physics, namely, either all models of systems in general systems theory are still based on the principle of absolute (global) time or time is not explicitly expressed in these models.
At the same time, there are many systems, in which it is unfeasible to introduce and preserve global time. For instance, it is proved that clock synchronization becomes impossible under definite conditions (Lamport, 1984; Dolev, et al, 1986; Fischer, et al, 1985; Attiya and Ellen, 2014). This precludes introduction of global time. As a result, in some systems, only local time (local time scale) can be treated in a consistent way. In addition, there are systems, in which global time (global time scale) coexists with a variety of local times (local time scales). Moreover, often these different times and time scales cannot be synchronized. All systems with these properties, which we call concurrent systems, cannot be portrayed by conventional models in general systems theory.
The goal of this paper is to construct more advanced than utilized now models of multi-agent distributed systems using the concept of local time, which exists and can be different in distinct components and parts of real systems according to the system theory of time (Burgin, 1992; 2002). These models provide descriptions and tools for exploration not only of classical systems with one global time but also of relativistic and concurrent systems, which can multiplicities of time.
It is interesting that absence of global time results in nonexistence of global states in a multicomponent system due to the concurrent functioning of the components and parts. As a result, time becomes multidimensional and demands specific unconventional mathematical structures for its representation.
In addition, exploring and modeling systems with a variety of independent and incoherent local times (local time scales), we come to the concepts of an observer, observation, synchronization and coordination of times and actions, which have not been studied in general systems theory. For instance, there are systems, in which it is possible to synchronize local time in different components and establish in such a way, global (absolute) time and a global time scale. However, conventional models from general systems theory give only a partial picture of these systems and do not allow exploration of synchronization. Note that synchronization plays a pivotal role in diversity of systems from organisms of people (cf., for example, (Winfree, 1987)) to computer networks (cf., for example, (Mills, 1991; Burgin, et al, 2016)) to distributed databases (cf., for example, (Lindsay, et al, 1979)) to physics and metrology (cf., for example, (Boixo, et al, 2006)) to human-computer interaction (cf., for example, (Burgin, et al, 2001)) to service systems (cf., for example, (Marzullo, 1983)). As Birman (2005) writes, clock synchronization is a necessary and critical part in most distributed systems.
To develop our model employing different structures of local time and local time scales, we use methods and approaches developed for concurrent processes in computer science and information theory. On the first stage of our research, we construct a kinetic system model by fundamentally advancing and further developing the Actor Model originally constructed for distributed computations (Hewitt, et al, 1973). Here we expand the scope of this model from computational systems and processes to general systems making it applicable for any system comprised of interacting subsystems, e.g., for organization, society, group of people or a computer network.
We call the basic component of the System Actor Model (SAM) constructed in this paper an actor although the conventional research typically uses the term agent. The reason for this is that according to the common usage, an agent is a system (an actor) who/that acts on behalf of another system (actor). Besides, in political science and economics, an agent is a person or entity able to make decisions and take actions on behalf of, or that impact, another person or entity called the principal (Rees, 1985; Eisenhardt, 1989).
This methodology allows treating agents as specific actors, who/which act on behalf other actors - principals. As a result, it is possible to represent any multi-agent multicomponent system by the System Actor Model but not every actor system can be represented by a multi-agent multicomponent system. There are many situations, especially, in society, when this difference between agents and free actors is very important. Taking into account that actors can be software systems, we see that
software agents are a very special but essentially important case of actors. It is possible to compare
relation between actors and agents with the relation between a function and a computable, e.g.,
recursive, function.
The System Actor Model is more flexible than agent models. For instance, agents are usually
treated as autonomous systems perceiving with sensors and acting with actuators (Bușoniu, et al,
2010; Vlassis, 2007; Weiss, 1999). At the same time, actors can be directed or controlled by other
actors. Some of actors can be without sensors and/or actuators. For instance, in problems of resource
management, identifying each resource with an actor can make available a helpful, detailed
perspective on the system while each of them might not have sensors and/or actuators and could be
controlled and managed by a central authority.
This paper is structured in the following way. In Section 2, which goes after Introduction, we
describe and explore the computational actor model. In Section 3, we construct and explore the
system actor model, for which the computational actor model becomes a very special case. Besides,
we go much further in comparison with the computational actor model by elaborating mathematical
models of actors and environments where these actors function. This allows us to obtain many
properties of actions, events, actors and their systems by rigorous mathematical techniques. One of
the main targets of this work is to construct mathematical tools for exploration of collaborative and
multi-agent systems, social network analysis and developing computational qualitative methods for
data mining and digital humanities.
2. Actor model in computer science
The Computational Actor Model (CAM) and its methodology were developed in the theory of
computation to provide constructive and theoretical tool for modeling, analyzing and organizing
concurrent digital computations (Hewitt, et al, 1973; Hewitt, 2012). In CAM, actors are interpreted
as computing devices or computational processes. We will call them computational actors.
To make the model uniform, the concept of a messenger, which is also a computational actor, is
used instead of the concept of a message. An arbitrary event in the model is the receipt of a
messenger, which impersonates a message, by the target (recipient) computational actor.
In CAM, computational actors perform computations based on information about other
computational actors and asynchronously communicate using their addresses for sending and
receiving messages. Additionally, computational actors can make decisions about their actions and
behavior, create other computational actors, and determine how to react to the received messages. It is possible to treat all these actions as events in the space of computational actors although this not done in the original Computational Actor Model described in (Hewitt, et al, 1973).
Computational actors are described by two groups of axioms - structural axioms and operational axioms.
**Structural axioms** determine that the local storage of a computational actor may contain addresses of other computational actors such that satisfy one of the following conditions:
1. The addresses were provided when the computational actor was created.
2. The addresses have been received in messages.
3. The addresses were installed in computational actors created by the given computational actor.
**Operational axioms** determine what a computational actor can do. Namely, a computational actor can:
1. Create more computational actors
2. Send messages to other computational actors
3. Receive messages from other computational actors
Hewitt explains that CAM is rooted in physics while other theoretical models of computation are based on mathematics and/or logic (Hewitt, 2007). As a result, CAM has many properties similar to properties of physical models, especially, in quantum physics and relativistic physics. For instance, detailed observation of the arrival order of the messages for a computational actor affects the results of actor’s behavior and can even increase indeterminacy. According to CAM, the performance of a computational actor is exactly defined when it receives a message while at other times, it may be indeterminate. Note that in reality, existence of nondeterministic models of computation, such as nondeterministic Turing machines (cf., for example, (Burgin, 2005)), shows that in some cases, the performance of a computing system or process cannot be exactly defined.
An important feature of CAM is that it can model systems that cannot be represented by the deterministic Turing's model while the latter is a special case of CAM. As Milner wrote (Milner, 1993), Hewitt had explained that a value, an operator on values, and a process could all be computational actors. Taking into account that computational actors can be interpreted as software systems, we can see that software agents are a very special but essentially important case of computational actors. The relation between computational actors and software agents is similar to
the relation between the concept of a number and the concept of a rational number. As we know, there are numbers that are not rational.
Being very useful for concurrent computations, CAM has very limited applications beyond computer science. That is why, taking the concept of an actor in all its generality and building a mathematical representation of a system actor, for which a computational actor is a very particular case, we extend CAM far beyond the area of computers, computer networks and computations.
3. Actor model in systems theory
The basic concept in the System Actor Model (SAM) is the concept of an actor or more exactly, of a system actor, which, in particular, can be a computational actor. In what follows, we mostly call system actors simply by the name actor when it does not cause ambiguity.
Informally, a system actor is a system that functions in some environment interacting with other systems. It means that System Actor Model developed in this paper is inherently dynamic.
This notion of an actor is more formally described in the following way.
**Definition 3.1.** Taking a system $E$ of interacting systems $\{R_k ; k \in K\}$, which have the lower rank than $E$, we call the systems $R_k$ actors and treat them as actors in $E$, while $E$ is called the environment of each of the actors $R_k$.
Note that in contrast to the Computational Actor Model where computational actors are processes or operators, a system actor can be (or can be interpreted as) an arbitrary system or an element/component of an arbitrary system, e.g., people, social networks, living beings, cells of living beings, molecules, artificial systems, such as computers or computer networks, processes and/or imaginary systems, such as heroes of novels or movies. Besides, computational actors can perform only three types of actions – create new actors, send messages and receive messages (Hewitt, 2007). In comparison with these limited abilities, system actors, in general, can perform any actions. Possible actions are described by the axioms that determine the environment of system actors.
Although some authors call such systems by the name agent (cf., for example, (Doyle, 1983; Minsky, 1986)), it is more reasonable to call them actors because according to the common usage, an agent is a system (an actor) who/that acts on behalf of another system (actor). In addition, in political science and economics, an agent is a person or entity able to make decisions and take actions on behalf of another person or entity called the principal (Rees, 1985; Eisenhardt, 1989).
To build a mathematical model of an environment with actors, we construct a mathematical model (description) of an actor and an environment. Note that there is no similar mathematical model (description) in the computational Actor Model.
A **formal actor** *(system actor representation)* $A$ is described by a name and five structural components - three sets called *set components* of the actor $A$ and two functions (or relations) called *functional components* of the actor $A$. Namely, we have the following structure
$$A = (\text{Rel}_A, \text{Act}_A, \text{Trn}_A; \text{React}_A, \text{Proact}_A)$$
$A$ is a name of the actor
Three sets (set components) are:
- $\text{Rel}_A$ is the set of properties and relations of the actor $A$ (usually only relations in the environment $E$ are considered)
- $\text{Act}_A$ is the set of possible actions of the actor $A$
- $\text{Trn}_A$ is the set of possible actions aimed at the actor $A$
Two functions (functional components), which are multivalued in the general case, are:
The *reaction function* *(reaction relation)* shows responses of the actor $A$ to actions on $A$
$$\text{React}_A: \text{Trn}_A \rightarrow \text{Act}_A$$
The *proaction function* *(proaction relation)* shows actions of the actor $A$ instigated by properties and relations of $A$
$$\text{Proact}_A: \text{Rel}_A \rightarrow \text{Act}_A$$
Reactions and proactions determine behavior of the actor.
It is possible to consider the following example of a tentative proaction.
**Example 3.1.** If an actor $B$ is a friend of an actor $A$, then $A$ is doing something good for $B$.
The next example shows a prescribed proactions.
**Example 3.2.** If an actor $B$ is a friend of an actor $A$, then $A$ always accepts messengers (massages) sent by $B$.
As an example of reactions, we can consider the following situation.
**Example 3.3.** The *action* aimed at an actor $A$ is an e-mail from an actor $B$.
The *reaction* of $A$ is the response to this e-mail.
Relations between an actor and data structures or knowledge structures, which may also be represented as actors, can represent the memory of the actor. Then self-actions can change this
memory performing computation, making decisions and deliberating subsequent actions. Note that it is possible to represent relations by properties and properties by relations (Burgin, 1985; 1990).
Parts and elements of actor’s components have their modalities described below.
First, in this description of an actor $A$, it is useful to make a distinction between actualized parts (elements) and tentative parts (elements) of actor’s components. For instance, some relations of $A$ exist while others are only possible. Then the former relations are actualized while the latter are tentative. In a similar way, some actions have been performed or/and are performed while others are only possible. Then the former actions are actualized while the latter are tentative.
Second, if an actor has a knowledge system, then it is useful to make a distinction between acknowledged parts (elements) and implicit parts (elements) of actor’s components. For instance, an actor $A$ can know about some of its relations and do not know about others. Then the former relations are acknowledged while the latter are implicit.
Usually the components of an actor satisfy some restrictions. For instance, if an actor $A$ is an automaton that does not give any output, e.g., if $A$ is an accepting finite automaton, then all action of $A$ are self-actions. In a formal setting, restrictions are described by axioms.
Properties, relations and actions have various properties including temporal properties. For instance, a singular action is performed at one moment of time, while performance of a regular action always demands some interval of time. In the theory of computational automata, all actions are singular (Burgin, 2005).
An important relation in this model is acquaintance. Namely, each actor $A$ has a list of names (addresses) of forward acquaintances $\text{FAcq}(A)$ and a list of names (addresses) of backward acquaintances $\text{BAcq}(A)$. These lists regulate communication of the actor $A$. Namely, the actor $A$ can send messages (messengers) only to forward acquaintances from $\text{FAcq}(A)$ and can receive messages (messengers) from only backward acquaintances from $\text{BAcq}(A)$. In particular, an actor (a system) can get feedback only from its backward acquaintances and can send feedback only to its forward acquaintances.
This assumption is formalized by the following axioms.
Let $\text{SMes}(A, B)$ denotes the action of sending a messenger (a message) by an actor $A$ to an actor $B$, $\Rightarrow$ denotes implication, $\Diamond$ denotes modal value “possible” and $\neg\Diamond$ denotes modal value “impossible”. For instance, $\Diamond \text{SMes}(A, C)$ means that the actor $A$ can send messages to the actor $C$.
**Axiom SM.** a) $\forall A, C \ (C \in \text{FAcq}(A) \Rightarrow \Diamond \text{SMes}(A, C))$.
b) \( \forall A, C (C \notin \text{FAcq}(A) \Rightarrow \neg \Diamond \text{SMes}(A, C)) \).
Informally, Axiom SMa means that the actor \( A \) can send messages (messengers) to any of its forward acquaintances. Axiom SMb means that the actor \( A \) cannot send messages (messengers) to any actor that (who) is not its forward acquaintance.
**Proposition 3.1.** If Axiom SM is true, then
\[ \forall A, C (C \in \text{FAcq}(A) \iff \Diamond \text{SMes}(A, C)) \]
**Proof.** By Axiom SMa, we have
\[ \forall A, C (C \in \text{FAcq}(A) \Rightarrow \Diamond \text{SMes}(A, C)) \]
Thus, we have to prove only
\[ \forall A, C (\Diamond \text{SMes}(A, C) \Rightarrow C \in \text{FAcq}(A)) \]
Let us assume that the actor \( A \) can send messages to some actor \( C \), i.e., \( \Diamond \text{SMes}(A, C) \), but \( C \) does not belong to the forward acquaintances of \( A \), i.e., \( C \notin \text{FAcq}(A) \). However, by Axiom SMb, we have \( \neg \Diamond \text{SMes}(A, C) \) and by principle of the Excluded Middle, our assumption is incorrect. Thus, we have
\[ \forall A, C (\Diamond \text{SMes}(A, C) \Rightarrow C \in \text{FAcq}(A)) \]
Proposition is proved.
Let \( \text{RMes}(C, A) \) denotes the action of receiving a messenger (a message) by an actor \( A \) from an actor \( C \).
**Axiom RM.**
a) \( \forall A, C (C \in \text{BAcq}(A) \Rightarrow \Diamond \text{RMes}(C, A)) \).
b) \( \forall A, C (C \notin \text{BAcq}(A) \Rightarrow \neg \Diamond \text{RMes}(C, A)) \).
Informally, Axiom RMA means that the actor \( A \) can receive messages (messengers) from any of its backward acquaintances. Axiom RMb means that the actor \( A \) cannot receive messages (messengers) from any actor that (who) is not its backward acquaintance.
**Proposition 3.2.** If Axiom RM is true, then
\[ \forall A, C (C \in \text{BAcq}(A) \iff \Diamond \text{Mes}(C, A)) \]
**Proof** is similar to the proof of Proposition 3.1.
Note that \( C \in \text{FAcq}(A) \) does not always mean that \( A \in \text{BAcq}(C) \). Indeed, it is possible that an actor \( A \) can send messages to an actor \( C \) but \( C \) cannot receive messages from \( A \).
The following axiom for the environment \( E \) remedies this situation.
**Connectivity Axiom CA.** \( \forall A, C \in E (C \in \text{FAcq}(A) \iff A \in \text{BAcq}(C)) \).
Informally, it means that an actor $A$ can receive messages (messengers) from an actor $B$ if and only if $B$ can send messages (messengers) to $A$.
Acquaintances that belong to both lists $\text{FAcq}(A)$ and $\text{BAcq}(A)$ are called friends. We denote this set by
$$\text{Fr}(A) = \text{FAcq}(A) \cap \text{BAcq}(A).$$
In many cases (but not always), lists $\text{FAcq}(A)$ and $\text{BAcq}(A)$ coincide. In this case, they also coincide with the list $\text{Fr}(A)$.
Let us assume that Axioms CA, SM and RM are true.
**Proposition 3.3.** $\forall A, B \ (B \in \text{Fr}(A) \Rightarrow A \in \text{Fr}(B))$
*Proof.* The formula $B \in \text{Fr}(A)$ means that $B \in \text{FAcq}(A)$ and $B \in \text{BAcq}(A)$. By Axiom CA, we have
$$A \in \text{BAcq}(A) \text{ and } A \in \text{FAcq}(A)$$
Thus, $A \in \text{Fr}(B)$.
Proposition is proved.
Proposition 3.3 allows proving the following result.
**Proposition 3.4.** If in the environment $E$, all acquaintances are friends, then $E$ satisfies Axiom CA.
In the process of actor functioning, the lists of acquaintances and friends can change.
There are five basic types of actor relations:
- **Inner relations** are relations between parts and elements of the actor $A$. For instance, if an actor $A$ is an organization, then relations between members of this organization are inner relations of $A$.
- **Internal** relations are relations between the actor $A$ and its parts and elements. For instance, if an actor $A$ is an organization, then the relation “a member $H$ of $A$ receives salary from $A$” is an internal relation of $A$.
- **Outer relations** are relations of the actor $A$ to other actors, their parts, elements and the environment. For instance, if actors $A$ and $B$ are organizations, then cooperation between $A$ and $B$ is an outer relation of $A$.
- **Intermediate** relations are relations of parts and elements of the actor $A$ to other actors, their parts, elements and the environment. For instance, if an actor $A$ is an organization,
then any relation between a member $H$ of $A$ and an actor $K$ who is not a member of $A$ is an intermediate relation of $A$.
- **External relations** are relations of other actors, in which the actor $A$ is included. For instance, if actors are companies, then “to be a supplier” is an external relation of $A$ when $A$ is a supplier for another company.
Note that it is possible to consider actions, reactions and proactions as relations. However, it is more efficient to treat these structures separately making emphasis on the functionality and dynamics.
According to the theory of autopoiesis developed by Maturana and Varela (1973), relations and properties play a crucial role for autopoietic systems, which can be described briefly as self-producing devices, or a self-generating systems with the ability to reproduce themselves recursively. Relations and properties of a system determine the structure of this system (Burgin, 2012). Indeed, autopoietic systems are structure-determined systems according to the principle of structural determinism, which states that the potential behavior of the system depends on its structure (Maturana, 1997). It means that all actions of actors representing autopoietic systems are functions of relations and properties.
Observing actions in the real world, we see that there are different types, classes, groups and kinds of actions. Let us consider some of them.
Temporal characteristics of actions determines three groups of reactions and proactions:
- **Sharp immediate reaction (proaction)** of $A$ starts immediately after the beginning of the corresponding action on $A$ (immediately after the property or relation becomes overt).
- **Reserved immediate reaction (proaction)** of $A$ starts when the corresponding action on $A$ ends (when the corresponding property or relation becomes comprehensible).
- **Delayed reaction (proaction)** of $A$ is performed when some time passes after the corresponding action on $A$ (when some time passes after the corresponding property or relation becomes comprehensible).
Definitions imply the following results.
**Proposition 3.5.** If an action $a$ is not immediate, then $a$ and any sharp immediate reaction to $a$ are parallel in time.
**Proposition 3.6.** An action and a reserved immediate reaction to it are strictly sequential in time.
**Proposition 3.7.** An action and a delayed immediate reaction to it are sequential in time.
There are other temporal relations between separate actions and events.
**Definition 3.2.** a) *Temporal independence* of events (actions) $E_1$ and $E_2$ means autonomy of their occurrence, i.e., either $E_1$ can take place before $E_2$ or $E_2$ can take place before $E_1$ or they can take place at the same time.
b) Two events (actions) are *temporally dependent* if they are not are temporally independent.
For instance, events in two disconnected computing systems are temporally independent. Note that disconnectedness means that these computers are not connected not only to one another but also to another system, for example, to the Internet. However, temporal independence does not prohibit simultaneous occurrence or coincidence of actions and events.
**Proposition 3.8.** Temporal dependence is a transitive relation.
Another important concept is temporal incomparability.
**Definition 3.3.** a) *Temporal incomparability* of events (actions) $E_1$ and $E_2$ means that it is not known whether they happen at the same time or which of them happens before the other.
b) Two events (actions) are *temporally comparable* if they are not are temporally incomparable.
For instance, events in two disconnected computers, which are not observed by the same observer, are temporally incomparable.
**Proposition 3.9.** Temporal comparability is a transitive relation.
Temporal independence and incomparability are related to concurrency.
**Definition 3.4.** Concurrency of two or more events or actions means their temporal independence and/or temporal incomparability, or in other words, that time of their happening is independent and sometimes incomparable.
As temporal independence allows simultaneous occurrence or coincidence, the introduced concept of concurrency comprises other interpretations of this term.
Concurrency is intrinsically related to such properties of events and actions as parallelism and sequentiality.
**Definition 3.5.** Two or more events or actions are parallel if their time intervals intersect (moments of their occurring coincide when they have zero duration, i.e., they are momentary).
For instance, when people read and understand some text, these actions are usually parallel but not always strictly parallel.
Note that independence of events allows them to be parallel. It implies that some parallel events can also be concurrent.
Proposition 3.10. If a momentary event (action) $E_1$ is parallel to a momentary event (action) $E_2$ and the event (action) $E_2$ is parallel to a momentary event (action) $E_3$, then all three events (actions) are parallel.
If the events are not momentary, then this result is not always true. For instance, let us consider events $E_1$, $E_2$ and $E_3$ such that $E_1$ starts at time 0 and ends at time 3, $E_2$ starts at time 2 and ends at time 5, and $E_3$ starts at time 4 and ends at time 7. Then the event $E_1$ is parallel to the event $E_2$ and the event $E_2$ is parallel to the event $E_3$, but the event $E_1$ is not parallel to the event $E_3$.
However, for interval events (actions), i.e., events (actions) with interval duration, it is possible to prove a result similar to Proposition 3.9.
Proposition 3.11. If an interval event (action) $E_1$ is parallel to an interval event (action) $E_2$, the event (action) $E_2$ is parallel to an interval event (action) $E_3$ and the event (action) $E_1$ is parallel to the event $E_3$, then all three events (actions) are parallel.
However, if the events are neither interval nor momentary, then this result is not always true. For instance, let us consider events $E_1$, $E_2$ and $E_3$ such that $E_1$ starts at time 0 and ends at time 3, $E_2$ starts at time 2 and ends at time 5, and $E_3$ starts at time 0 and continues to time 1, then restarts at time 4 and ends at time 7. Then the event $E_1$ is parallel to the event $E_2$ and the event $E_2$ is parallel to the event $E_3$, the event $E_1$ is parallel to the event $E_3$ but all three events are not parallel.
Definition 3.6. Two or more events or actions are strictly parallel if their beginning and end coincide and they go (take place) in the same time.
For instance, when the user switches her computer on (the first event), the computer starts working (the second event, which is strictly parallel to the first event).
Proposition 3.12. If an event (action) $E_1$ is strictly parallel to an event (action) $E_2$ and the event (action) $E_2$ is strictly parallel to an event (action) $E_3$, then the event (action) $E_1$ is strictly parallel to the event (action) $E_3$.
Remark 3.1. For parallel events (actions), this result is not always true.
Definition 3.7. a) Two events or actions are sequential if one of them, say $E_2$, starts after the other, say $E_1$, ends.
b) In this case, the event (action) $E_2$ is called subsequent to the event (action) $E_1$.
For instance, reception of information is subsequent to sending this information but usually it is not strictly subsequent.
Proposition 3.13. The relation between events and actions to be sequential is transitive.
Another important relation between events and actions is to be strictly sequential.
**Definition 3.8.** a) Two events or actions are strictly sequential if one of them, say \( E_2 \), starts exactly at the moment the other, say \( E_1 \), ends.
b) In this case, the event (action) \( E_2 \) is called strictly subsequent to the event (action) \( E_1 \).
In the theory of finite automata, it is assumed that starting from the second transition, each transition of the automaton is strictly subsequent to the previous transition (Burgin, 2005).
**Proposition 3.14.** If an event (action) \( E_1 \) is strictly subsequent to an event (action) \( E_2 \) and the event (action) \( E_2 \) has positive duration and is strictly subsequent to an event (action) \( E_3 \), then event (action) \( E_1 \) is not strictly subsequent to the event (action) \( E_3 \).
There are also structural characteristics of actions. One of them is direction.
Direction of actions determines three groups of actions:
- An external action of an actor is directed at other actors (cf. Figure 2).
- An internal action or a self-action of an actor is directed at the same actor and usually results in self-transformation (cf. Figure 1).
- A combined action of an actor is directed both at other actors and at the same actor (cf. Figure 3).
**Figure 1.** A self-action is an action of an agent on itself.
**Figure 2.** An external action is directed at other actors
**Figure 3.** A combined action goes inside and outside.
Example 3.4. Reception of information is an example of a self-action.
Example 3.5. Computation performed by a system actor and any computational operation are examples of a self-action.
Example 3.6. Decision-making of a system actor is an example of a self-action.
Example 3.7. Sending information is an example of an external action.
Example 3.8. Working an inductive Turing machine transforms the content of its working register and from time to time, sends this content to the output register (Burgin, 2005). The action of the machine when it is doing both operations at the same time is a combined action.
Another structural characteristic of actions is modality, which determines the status of actions in the environment. There are three modalities of actions – positive, negative and neutral – and each of them contains four classes.
Positive modalities of actions:
- Possible actions
- Tolerable actions
- Permitted actions
- Performed actions
Negative modalities of actions:
- Impossible actions
- Intolerable actions
- Prohibited actions
- Not performed (but possible/permitted) actions
Neutral modalities of actions:
- Unknown actions
- Unidentified actions
- Unspecified actions
- Indefinite actions
There are definite relations between modalities of actions.
Proposition 3.15. a) Any unknown action is unidentified.
b) Any unidentified action is unspecified.
c) Any performed action is possible.
d) Any unknown possible and permitted action is not performed.
Structural characteristics of actions show that there are *simple actions* and *compound actions*, which are compositions of other actions. Compositions of actions are constructed using operations with actions. For instance, performing one action after another gives us the sequential composition of these actions.
If an action \( a \) is a composition of actions \( a_1, a_2, a_3, \ldots, a_n \), for example, \( a = \omega(a_1, a_2, a_3, \ldots, a_n) \) where \( \omega \) is an \( n \)-ary operation with actions, then any action \( a_i \) \((i = 1, 2, 3, \ldots, n)\) is included in or is a *part* of the action \( a \). It is denoted by \( a_i \subseteq a \).
Informally, the relation \( b \subseteq a \) means that performance of the action \( a \) includes performance of the action \( b \).
**Proposition 3.16.** For any actions \( a, b \) and \( c \), relations \( a \subseteq b \) and \( b \subseteq c \) imply the relation \( a \subseteq c \).
Indeed, as a composition of compositions of actions is a composition of actions, relations \( a \subseteq b \) and \( b \subseteq c \) imply the relation \( a \subseteq c \).
It means that the relation “to be a part” or “to be included” is transitive.
Composition preserves direction of actions.
**Proposition 3.17.** A composition of internal (external or combined) actions of the same actor is an internal (external or combined) action.
Organization of actions determines three groups of actions:
- *Direct actions* does not include additional operations (actions)
- *Mediated actions* include additional operations (actions or processes), for example, such as computation, meditation, contemplation or actions of other actors
- *Void actions or inactions*
Not to perform an action is also an action. It is a *void action*. All other actions are *proper actions*.
It is possible to build the system Actor Model (SAM) with one void action or with different void actions. It is possible to give a more precise description of actor’s behavior when SAM allows different void actions. In this case, we have the following definition.
**Definition 3.9.** Not to perform an action \( a \) is the inaction \( \neg a \).
For instance, when a person \( A \) is standing near the river and doing nothing seeing a person \( B \) is drowning, this is a negative void action. When the Allies did nothing to prevent Hitler from seizing Austria and a part of Czechoslovakia, it was also a negative void action.
At the same time, there are positive void actions. For instance, when a person does not steal, it is a positive void action.
The concept of inaction or non-action plays an important role in Taoism because one of its central principles is the Principle of non-action (Wu wei in Chinese). Wu wei from the Tao Te Ching literally means non-action or non-doing and is connected to the paradox weiwuwei: "action without action" (Kirkland, 2004; Klaus, 2009).
Let us consider some properties of void actions.
**Proposition 3.18.** \( \neg \neg a = a \).
Informally, it means that when non-doing of action \( a \) is not performed, then action \( a \) is performed. In essence, this is a version of the Principle of Excluded Middle because the proof of Proposition 3.18 uses this Principle and it is possible to consider systems of actors for which this assertion is not true.
Common sense tells us that independently in what way you compose non-doing, it will always be non-doing. We formalize this impression in the following axiom.
**Emptiness Axiom EA.** If \( a_1, a_2, a_3, \ldots, a_n \) are actions and \( \omega \) is an \( n \)-ary operation with actions, then
\[
\omega(\neg a_1, \neg a_2, \neg a_3, \ldots, \neg a_n) = \neg \omega(a_1, a_2, a_3, \ldots, a_n)
\]
Axiom EA implies the following result.
**Proposition 3.19.** A composition of inactions is an inaction.
However, in general, Axiom EA is not always valid and a composition of inactions can be a proper action. For instance, let us consider the binary composition \( L(x, y) \), which combines two actions inferring the third action when only three actions can be performed. To provide an example of this situation, we can take the situation when a person can only either run (action \( a \)) or walk (action \( b \)) or stand (action \( c \)). Then combining two inactions \( \neg a \) (not running) and \( \neg b \) (not walking), we have \( L(a, b) = c \), which is a proper action.
**Proposition 3.20.** If \( a \subseteq b \), then \( \neg b \subseteq \neg a \).
Indeed, if an action \( b \) includes an action \( a \), then the absence of \( a \) implies and thus, includes, the absence of \( b \).
It is useful to consider the total inaction \( T_{IA} \) when simply nothing is done.
**Proposition 3.21.** For any action \( a \), we have \( \neg a \subseteq T_{IA} \).
Definitions imply the following result.
**Proposition 3.22.** A composition of non-void (proper) actions is a mediated action.
There other important types of actions.
A *primitive action* is a direct action that depends only on the input actions of other actors in the case of reactions or only properties and relations in the case of proactions.
An *automatic action* is a direct action that depends both on actions of other actors and on properties/relations.
Note that an inaction also can be primitive or automatic.
**Proposition 3.23.** When an action $a$ is primitive (automatic), the inaction $¬a$ is also primitive (automatic).
Automatic actions allow unification of reactions and proactions in one (multivalued in a general case) function of combined actions
$$\text{Combact}_A: \text{Trn}_A \times \text{Rel}_A \to \text{Act}_A$$
In this context, the function $\text{React}_A$ is a restriction of the function $\text{Combact}_A$ when the action on $A$ is void and the function $\text{Proact}_A$ is a restriction of the function $\text{Combact}_A$ when the property/relation is void. This gives us the following result.
**Proposition 3.24.** Any primitive action is an automatic action.
Different types of actions spawn different types of actors.
**Definition 3.10.** A *behaviorally primitive actor* $A$ has only primitive actions.
For instance, finite automata with one state are behaviorally primitive actors because their actions depend only on the input.
**Definition 3.11.** A *behaviorally automatic actor* $A$ has only automatic actions.
For instance, finite automata are behaviorally primitive actors because their actions depend on both the input and inner state.
Proposition 3.24 implies the following result.
**Corollary 3.1.** Any behaviorally primitive actor is a behaviorally automatic actor.
There are various relations between actors.
**Definition 3.12.** Two actors are *identical* if they have the same structural components.
For instance, in contemporary industry, identical copies of many devices, such as vehicles, planes, computers and cell phones, are produced. In the system Actor Model, all these copies are represented by identical actors.
**Lemma 3.1.** Identity is an equivalence relation in sets of actors.
It is possible to find identical actors in many areas. One of them is theory and technology of information processing. Thus, there are models of computational systems, which contain many (sometimes, infinite) identical computing elements. Examples are cellular automata, artificial neural networks and iterative arrays.
For instance, a cellular automaton is a system of identical finite automata called cells, which form a net and interact with one another. A cellular automaton is determined by the following parameters (Burgin, 2005):
1. The space organization of the cells. In the majority of cellular automata, cells organized in a simple rectangular grid (mostly it is a one-dimensional string of cells and a two- or three-dimensional grid of cells), but in some cases, other arrangements, such as a honeycomb or Fibonacci trees.
a. The topology of the cellular automaton is determined by the type of the cell neighborhood, which consists of other cells that interact with this cell. In a grid, these are normally the cells physically closest to the cell in question. For instance, if each cell has only two neighbors (right and left), it defines linear topology. Such cellular automata are called linear or one-dimensional. It is possible to consider linear automata with the neighborhood of some radius \( r > 1 \). When there are four cells (upper, below, right, and left), the CA has two-dimensional rectangular topology. Such cellular automata are called planar or two-dimensional.
2. The dynamics of a cellular automaton, which determines by what rules cells exchange information with each other.
Traditionally, only rectangular organization of the cells and their neighborhoods has been considered for cellular automata. Recently, researchers have begun studies of cellular automata in the hyperbolic plane or on a Fibonacci tree (Margenstern, 2002). It is proved that such automata are more efficient than traditional cellular automata in the Euclidean plane. This higher efficiency is a result of a better topology in cellular automata in the hyperbolic plane.
According to the system Actor Model, each element of a cellular automaton is an actor and its actions consist of computing and communicating operations.
Looking at computer technology, we see that from the perspective of a manufacturer, products, e.g., computers, of the same type are identical.
Another important relation between actors is dynamic equivalence.
**Definition 3.13.** Two actors are *dynamically equivalent* if they have the same action components.
When it is necessary to solve the same problem for different input data, it is possible to use equivalent actors to this in a parallel or concurrent mode. This is often done in multiprocessor computers where identical processors perform necessary computations.
**Lemma 3.2.** Dynamic equivalence is an equivalence relation in sets of actors.
Identity of actors is a stronger relation than dynamic equivalence.
**Lemma 3.3.** Identical actors are dynamically equivalent.
Dynamic equivalence determines similarities in actor’s behavior.
**Proposition 3.25.** An actor without actions is dynamically equivalent to an actor that has only void actions.
**Proposition 3.26.** An actor $A$ dynamically equivalent to a behaviorally primitive actor $B$ is behaviorally primitive.
**Proposition 3.27.** An actor $A$ dynamically equivalent to a behaviorally automatic actor $B$ is behaviorally automatic.
Another important relation between actors is homology.
**Definition 3.14.** Two actors $A$ and $B$ are *homological* if all their corresponding structural components are isomorphic.
For instance, for homological actors $A$ and $B$, there are isomorphisms between $\text{Rel}_A$ and $\text{Rel}_B$, between $\text{React}_A$ and $\text{React}_B$, and between $\text{Proact}_A$ and $\text{Proact}_B$.
**Example 3.9.** Let us consider two deterministic finite automata $A$ and $B$. They have the same set of states and the same set of start and final states. The first has the alphabet $\{0, 1\}$ and the second the alphabet $\{a, b\}$. Besides, all transitions of $A$ produced by input 0 are the same as all transitions of $B$ produced by input $a$ and all transitions of $A$ produced by input 1 are the same as all transitions of $B$ produced by input $b$. Then these automata are homological actors.
**Lemma 3.4.** Homology is an equivalence relation in sets of actors.
Identity of actors is a stronger relation than homology.
**Lemma 3.5.** Identical actors are homological.
Let us assume that isomorphisms between $\text{React}_A$ and $\text{React}_B$ and between $\text{Proact}_A$ and $\text{Proact}_B$ preserves primitive actions. Then we have the following result.
**Proposition 3.28.** An actor $A$ homological to a behaviorally primitive actor $B$ is behaviorally primitive.
Let us assume that isomorphisms between React$_A$ and React$_B$ and between Proact$_A$ and Proact$_B$. preserves automatic actions. Then we have the following result.
**Proposition 3.29.** An actor $A$ homological to a behaviorally automatic actor $B$ is behaviorally automatic.
A weaker type of relations is dynamic homology
**Definition 3.15.** Two actors $A$ and $B$ are dynamically homological if all their corresponding action components are isomorphic.
**Lemma 3.6.** Dynamic homology is an equivalence relation in sets of actors.
Dynamic equivalence of actors is a stronger relation than dynamic homology.
**Lemma 3.7.** Dynamically equivalent actors are dynamically homological.
Let us assume that isomorphisms between React$_A$ and React$_B$ and between Proact$_A$ and Proact$_B$. preserves primitive actions. Then we have the following result.
**Proposition 3.30.** An actor $A$ dynamically homological to a behaviorally primitive actor $B$ is behaviorally primitive.
Let us assume that isomorphisms between React$_A$ and React$_B$ and between Proact$_A$ and Proact$_B$. preserves automatic actions. Then we have the following result.
**Proposition 3.31.** An actor $A$ dynamically homological to a behaviorally automatic actor $B$ is behaviorally automatic.
According to their structure, we discern four classes of actors:
- A **structurally prime actor** $A$ does not have components or parts.
- A **structurally primitive actor** $A$ does not have components or parts, which are also actors.
- A **structurally composite actor** $A$ has parts and/or components.
- A **structurally compound actor** $A$ has parts and/or components, which are also actors.
In the actor’s structure elements are also treated as parts.
The scale of observation defines what actors are prime. Thus, to be a prime actor depends on the scale of observation/treatment. For instance, in the observation scale of society, people are primitive actors. At the same time, in the observation scale of biology, people are composite actors.
The scale of modeling defines what actors are primitive. Thus, to be a primitive actor depends on the scale of modeling/representation. For instance, in the modeling scale of society, it is natural to
represent people as primitive actors. At the same time, in the modeling scale of biology, it is natural to represent people as compound actors.
It is possible to develop a scale (ranging) of actors and deal with parts and components of a actor in this scale. Namely, an actor \( A \) that is a part/component of another actor \( B \) has lower range than \( B \).
The system (environment) \( E \) can be a model of a real system \( R \), which can be physical, mental or structural. The system \( R \) is called a *modeled domain of \( E \)*. In general, one environment \( E \) can model different domains.
Let us consider a modeled domain \( R \) of an environment \( E \).
**Proposition 3.32.** If \( R \) is the modeled domain of environment \( E \) and a subdomain \( P \) of \( R \) is a modeled domain of \( D \), then there is an injection of the set of all actors from \( D \) into the set of all actors from \( E \).
It is possible to introduce the following axiom
**Modeling Axiom MA.** Any object in the modeled domain \( R \) is modeled by an actor in \( E \).
If Pythagoras asserted “Everything is a number,” the Modeling Axiom states “Everything and everybody is an actor.”
The computational Actor Model that satisfies the Modeling Axiom is called the universe of CAM (Agha, 1986).
Let us consider an actor \( A \) with the inner structure \( Q \).
**Proposition 3.33.** If the Modeling Axiom is valid for an environment \( E \) and its modeled domain \( R \), then:
(a) Any structurally primitive actor is structurally prime.
(b) Any structurally composite actor is structurally compound.
**Corollary 3.2.** If the Modeling Axiom is valid for an environment \( E \) and its modeled domain \( R \), then there are only structurally primitive and structurally compound actors in \( E \).
**Definition 3.16.** A *primary actor* \( A \) is not a part or component of other actors.
According to their communication, we discern five classes of actors – closed, inactive, generative, undemanding and open actors.
**Definition 3.17.** A *closed actor* \( A \) does not send and receive messengers (messages).
The concept of a closed actor allows treating almost anything, for example, tables, chairs, mountains, rivers, words, sounds, etc. as actors.
**Definition 3.18.** An *inactive actor* \( A \) does not send messengers (messages).
For instance, a sleeping woman does not send messengers (messages). Another example of an inactive actor is a receptor such as an automaton, which accepts input but gives no output (Burgin, 2005).
Definitions imply the following result.
**Lemma 3.8.** Any closed actor $A$ is inactive.
The dual concept to an inactive actor is a non-receptive actor.
**Definition 3.19.** A *non-receptive actor* $A$ does not receive messengers (messages).
An example of a non-receptive actor is a generator, i.e., such as an automaton, which does not accept input but gives output (Burgin, 2005). Another example of a non-receptive actor is a black hole (Thorne, 1994; Davies, 1995).
Definitions imply the following results.
**Lemma 3.9.** Any closed actor $A$ is non-receptive.
It means that the property “to be closed” is stronger than the property “to be non-receptive.”
**Lemma 3.10.** A non-receptive and inactive actor $A$ is closed.
Opposite to closed actor are open actors.
**Definition 3.20.** An *open actor* $A$ sends and receives messengers (messages).
Definitions imply the following results.
**Lemma 3.11.** Any open actor $A$ is active.
It means that the property “to be open” is stronger than the property “to be active.”
**Lemma 3.12.** A receptive and active actor $A$ is open.
It is possible to distinguish actor by messages they send.
**Definition 3.21.** An *undemanding actor* $A$ does not send requesting messengers (requests).
Definitions imply the following results.
**Lemma 3.13.** Any inactive actor $A$ is undemanding.
Lemmas 3.9 and 3.13 imply the following result.
**Corollary 3.3.** Any closed actor $A$ is undemanding.
It is possible to develop a scale (ranging) of actors and deal with parts and components of a primary actor in this scale.
Because an actor is functioning in some environment, it is also practical to use an extended actor representation, which includes relevant characteristics of the environment.
An extended actor representation consists of two names, three sets and four functions (or relations)
\[(A, E) = (\text{Rel}_A, \text{Act}_A, \text{Trn}_A; \text{React}_A, \text{Proact}_A, \text{VReact}_A, \text{VProact}_A)\]
\(A\) is a name of the actor.
\(C\) is a name of the actor’s environment.
Three sets are:
- \(\text{Rel}_A\) is the set of properties of \(A\) and relations of \(A\) to other actors and the environment
- \(\text{Act}_A\) is the set of possible actions of \(A\)
- \(\text{Trn}_A\) is the set of possible actions on \(A\)
Four functions (multivalued in the general case) are:
The reaction function shows responses of \(A\) to actions on \(A\)
\(\text{React}_A: \text{Trn}_A \rightarrow \text{Act}_A\)
Proactions show actions on \(A\) instigated by properties and relations of \(A\)
\(\text{Proact}_A: \text{Rel}_A \rightarrow \text{Act}_A\)
For instance, if \(B\) is a friend of \(A\), then \(A\) is doing something good for \(B\).
The virtual reaction function shows responses of \(A\) to all possible actions
\(\text{VReact}_A: \text{Act}_E \rightarrow \text{Act}_A\)
Here \(\text{Act}_E\) is the set of all possible actions in \(E\).
The virtual proaction function shows actions on \(A\) instigated by all properties and relations, which exist in \(E\)
\(\text{VProact}_A: \text{Rel}_C \rightarrow \text{Act}_A\)
Here \(\text{Rel}_C\) is the set of all possible properties and relations in \(E\).
Definitions imply the following results.
**Lemma 3.14.** \(\text{React}_A\) is a restriction of \(\text{VReact}_A\).
**Lemma 3.15.** \(\text{Proact}_A\) is a restriction of \(\text{VProact}_A\).
In the System Actor Model, we also have a mathematical model of an environment.
An environment representation is described by a name, two sets and two functions (or relations)
\[E = (\text{Rel}_E, \text{Act}_E, \text{Trn}_E; E\text{React}_E, E\text{Proact}_E)\]
\(A\) is a name of the actor
Two sets are:
- \(\text{Rel}_E\) is the set of all possible properties and relations in \(E\).
• \( \text{Act}_E \) is the set of all possible actions in \( E \).
Two functions (multivalued in the general case) are:
\( \text{EReactions} \) show all possible responses to actions in \( E \)
\[ \text{EReact}_E: \text{Trn}_E \rightarrow \text{Act}_E \]
\( \text{EProactions} \) show all possible actions instigated by properties and relations in \( E \)
\[ \text{EProact}_E: \text{Rel}_E \rightarrow \text{Act}_E \]
Note that the systems \( R_k \) in the environment \( E \) can have different ranks. For instance, in society, actors include separate individuals, organizations, countries, and so on.
**Definition 3.22.** a) If an actor \( A \) is a proper subsystem of an actor \( B \), then the *rank* of \( A \) is lower than the rank of \( B \).
b) If actors \( A \) and \( B \) consist of elements of the same rank, then the *rank* of \( A \) is equal to the rank of \( B \).
By definition, the environment \( E \) has the highest rank in the system Actor Model.
**Proposition 3.34.** Elements, parts and components of an actor \( A \) have lower rank than \( A \).
4. Conclusion
We have built a mathematical model of multicomponent interactive systems, which is called the System Actor Model and based on the formal structure of actors functioning in a multifarious convoluted environment. Different properties of such systems represented by an environment with actors have been obtained. Actions and events are analyzed in this context and different classes of events and actions are explicated and studied. Actors are also classified according to their traits. In addition, we elaborated a mathematical model of the environment. One of the main targets of this work is to construct mathematical tools for exploration of social systems. To conclude, we formulate some open problems for the System Actor Model.
The first cluster of problems is related to actions.
**Problem 1.** Formalize and study results of actions.
**Problem 2.** Formalize and study consequences of actions.
**Problem 3.** Formalize and study causes of actions.
**Problem 4.** Formalize and study in more detail structural, temporal and spatial characteristics of actions.
The second cluster of problems is related to actors.
**Problem 5.** Formalize and study tasks of actors.
**Problem 6.** Formalize and study obligations of actors.
**Problem 7.** Formalize and study norms of actors.
**Problem 8.** Formalize and study values of actors.
The third cluster of problems is related to concepts of agents and oracles, which are connected to the concept of actors.
**Problem 9.** Formalize and study relations between agents and actors.
**Problem 10.** Formalize and study relations between oracles and actors.
**References**
1. Agha, G. A. (1986) *ACTORS: A Model of Concurrent Computation in Distributed Systems*, The MIT Press Series in Artificial Intelligence, The MIT Press, Cambridge, Massachusetts
2. Attiya H. and Ellen, F. Impossibility Results for Distributed Computing, Morgan & Claypool Publishers, May 1, 2014
3. Barwise, J. and Seligman, J. (1997) *Information Flow: The Logic of Distributed Systems*, Cambridge Tracts in Theoretical Computer Science 44, Cambridge University Press
4. Bergson, H. *Time and Free Will: An Essay on the Immediate Data of Consciousness*, George Allen and Unwin, London, 1910
5. Birman, K.P. (2005) Clock Synchronization and Synchronous Systems, in *Reliable Distributed Systems*, Springer New York, pp. 493-508
6. Boixo, S., Caves, C.M., Datta, A. and Shaji, A. (2006) On decoherence in quantum clock synchronization, *Laser Physics*, v. 16, issue 11, pp. 1525-1532
7. Burgin, M. (1985) Abstract theory of properties, in *Non-classical Logics*, Institute of Philosophy, Moscow, pp. 109-118 (in Russian)
8. Burgin, M. (1990) Abstract Theory of Properties and Sociological Scaling, in *Expert Evaluation in Sociological Studies*, Kiev, pp. 243-264 (in Russian)
9. Burgin, M. A System Approach to the Concept of Time, *Philosophical and Sociological Thought*. - 1992. – No. 8 (in Russian and Ukrainian)
10. Burgin M. Time as a Factor of Science Development, *Science and Science of Science*, 1997, No. 1/2, pp. 45-59
11. Burgin, M. *Elements of the System Theory of Time*, LANL, Preprint in Physics 0207055, 2002, 21 p. (electronic edition: http://arXiv.org)
12. Burgin, M. *Super-recursive Algorithms*, Springer, New York/Heidelberg/Berlin, 2005
13. Burgin, M. (2012) *Structural Reality*, Nova Science Publishers, New York
14. Burgin, M., Karplus, W. and Liu, D. The Problem of Time Scales in Computer Visualization, in “Computational Science”, Lecture Notes in Computer Science, v. 2074, part II, 2001, pp.728-737
15. Burgin, M., Mikkilineni, R. and Morana, G. Intelligent organisation of semantic networks, DIME network architecture and grid automata, International Journal of Embedded Systems, v. 8, No. 4, 2016, pp. 352-366
16. Buşoniu, L., Babuška, R. and De Schutter, B. Multi-agent reinforcement learning: An overview, in Innovations in Multi-Agent Systems and Applications (D. Srinivasan and L.C. Jain, eds.), v. 310 of Studies in Computational Intelligence, Berlin, Germany: Springer, pp. 183–221, 2010.
17. Davies, P. About Time, Simon & Schuster, New York/London/Tokyo, 1995
18. Dolev, D., Halpern, J.Y. and Strong, H. R. On the Possibility and Impossibility of Achieving Clock Synchronization, Journal of Computer and System Sciences, v. 32, 230-250 (1986)
19. Einstein, A., Lorentz, H.A., Weil, H., and Minkowski, H. The Principle of Relativity, Dover, 1923
20. Fischer, M.J., Lynch, N.A. and Paterson, M.S. Impossibility of Distributed Consensus with One Faulty Process, Journal of the Association for Computing Machinery, v. 32, No. 2, April 1985, pp. 374-382
21. Hewitt, C. (2007) What is Commitment? Physical, Organizational, and Social, Lecture Notes in Artificial Intelligence (Vázquez-Salceda, J. and Pablo Noriega, P. EdsS), Springer
22. Hewitt, C. (2012) What is computation? Actor Model versus Turing’s Model, in A Computable Universe, Understanding Computation & Exploring Nature as Computation (H. Zenil, Ed.) World Scientific Publishing Company/Imperial College Press
23. Hewitt, C., Bishop, P. and Steiger, R. (1973) A Universal Modular Actor Formalism for Artificial Intelligence, IJCAI'73 Proceedings of the 3rd international joint conference on Artificial intelligence, Morgan Kaufmann Publishers Inc., San Francisco, CA, pp. 235-245
24. Kirkland, R. Taoism: The Enduring Tradition, Routledge, London/New York, 2004
25. Klaus, H. The Tao of Wisdom. Laozi – Daodejing. Chinese-English-German. Hochschulverlag, Aachen, 2009
26. Lamport, L. Unsolved problems, solved problems, and non-problems in concurrency , in Proceedings of the 3rd ACM Conference on Principles of Distributed Computing, 1984, pp. I-11
27. Lindsay, B. G., Selinger, P. G., Galtieri, C., Gray, J. N., Lorie, R. A., Price, T. G., Putzolu, F., Traiger, I. L., and Wade, B. W. Notes on distributed databases. IBM Res. Rep. RJ2571, IBM Research Division, San Jose, Calif., 1979.
28. Marzullo, K. Loosely-Coupled Distributed Services: A Distributed Time System, Ph.D. dissertation, Stanford University, 1983
29. Maturana, H. (1997) Metadesign, in Articulos y Conferences “Diez Años de Post-Racionalismo en Chile”, Instituto de Terapia Cognitiva Web, Santiago
http://www.inteco.cl/articulos/006/doc_Ing1.htm last visited 9/6/2017
30. Maturana, H. and Varela, F. (1998) *The Tree of Knowledge: The Biological Roots of Human Understanding*, Shambhala, Boston
31. Mills, D.L. (1991) Internet time synchronization: the Network Time Protocol. *IEEE Trans. Communications COM-39*, v. 10, pp. 1482-1493
32. Milner, R. (1993) Elements of Interaction, *Communications of the ACM*, v. 36 No. 1, pp. 78-89
33. Prigogine, I. *From being to Becoming: Time and Complexity in Physical Systems*, San Francisco, 1980
34. Shoham, Y. and Leyton-Brown, K. *Multiagent Systems: Algorithmic, Game Theoretic and Logical Foundations*, Cambridge University Press, Cambridge, 2008
35. Thorne, K.S. *Black Holes and Time Warps*, Norton, New York, 1994
36. Vlassis, N. *A Concise Introduction to Multiagent Systems and Distributed Artificial Intelligence*, Synthesis Lectures in Artificial Intelligence and Machine Learning, Morgan & Claypool Publishers, 2007
37. Weiss, G. (ed.) *Multiagent Systems: A Modern Approach to Distributed Artificial Intelligence*, MIT Press, New York/London, 1999
38. Wiener, N. *Cybernetics, or Control and Communication in the Animal and the Machine*, MIT Press and Wiley, New York/London, 1961
39. Winfree, A.T. *The Timing of Biological Clocks*, Scientific American Library, New York, 1987 | 2025-03-06T00:00:00 | olmocr | {
"pdf_page_numbers": [
[
0,
2007,
1
],
[
2007,
4725,
2
],
[
4725,
7603,
3
],
[
7603,
10224,
4
],
[
10224,
12697,
5
],
[
12697,
15285,
6
],
[
15285,
17468,
7
],
[
17468,
20311,
8
],
[
20311,
22652,
9
],
[
22652,
24686,
10
],
[
24686,
27126,
11
],
[
27126,
29513,
12
],
[
29513,
32223,
13
],
[
32223,
33800,
14
],
[
33800,
35220,
15
],
[
35220,
37748,
16
],
[
37748,
40143,
17
],
[
40143,
42369,
18
],
[
42369,
44829,
19
],
[
44829,
47110,
20
],
[
47110,
49460,
21
],
[
49460,
51823,
22
],
[
51823,
53778,
23
],
[
53778,
55805,
24
],
[
55805,
57975,
25
],
[
57975,
60287,
26
],
[
60287,
63200,
27
],
[
63200,
64473,
28
]
]
} |
07cafe64907aa83edbaf485ef24807a3788d645c | {
"olmocr-version": "0.1.59",
"pdf-total-pages": 11,
"total-fallback-pages": 0,
"total-input-tokens": 30543,
"total-output-tokens": 10297,
"fieldofstudy": [
"Medicine",
"Psychology"
]
} | Effect of patient-initiated versus fixed-interval telePRO-based outpatient follow-up: study protocol for a pragmatic randomised controlled study
Liv Marit Valen Schougaard1*, Caroline Trillingsgaard Mejdahti2,6, Klaus Hvam Petersen1, Anne Jessen1, Annette de Thurah2,3,4, Per Sidenius5, Kirsten Lomborg2,4,6 and Niels Henrik Hjollund1,7
Abstract
Background: The traditional system of routine outpatient follow-up of chronic disease in secondary care may involve a waste of resources if patients are well. The use of patient-reported outcomes (PRO) could support more flexible, cost-saving follow-up activities. AmbuFlex is a PRO system used in outpatient follow-up in the Central Denmark Region. PRO questionnaires are sent to patients at fixed intervals. The clinicians use the PRO data to decide whether a patient needs a visit or not (standard telePRO). PRO may make patients become more involved in their own care pathway, which may improve their self-management. Better self-management may also be achieved by letting patients initiate contact. The aim of this study is to obtain data on the effects of patient-initiated follow-up (open access telePRO) on resource utilisation, quality of care, and the patient perspective.
Methods: The study is a pragmatic, randomised, controlled trial in outpatients with epilepsy. Participants are randomly assigned to one of two follow-up activities: a) standard telePRO or b) open access telePRO. Inclusion criteria are age ≥ 15 years and previous referral to standard telePRO follow-up at Aarhus University Hospital, Denmark. Furthermore, patients must have answered the last questionnaire via the Internet. The number of contacts will be used as the primary outcome measure. Secondary outcome measures include well-being (WHO-5 Well-Being Index), general health, number of seizures, treatment side effects, mortality, health literacy (Health Literacy Questionnaire), self-efficacy (General Self-Efficacy scale), patient activation, confidence, safety, and satisfaction. In addition, the patient perspective will be explored by qualitative methods. Data will be collected at baseline and 18 month after randomisation. Inclusion of patients in the study started in January 2016. Statistical analysis will be performed on an intention-to-treat and per-protocol basis. For qualitative data, the interpretive description strategy will be used.
Discussion: The benefits and possible drawbacks of the PRO-based open access approach will be evaluated. The present study will provide important knowledge to guide future telePRO interventions in relation to effect on resource utilisation, quality of care, and the patient perspective.
Trial registration: ClinicalTrials.gov: NCT02673580 (Registration date January 28, 2016)
Keywords: Patient-reported outcomes, TelePRO, Clinical practice, Outpatient clinic, Outpatient follow-up, Open access, Randomised controlled trial
* Correspondence: [email protected]
1AmbuFlex, Regional Hospital West Jutland, Herning, Denmark
Full list of author information is available at the end of the article
© The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
Background
The Danish health care system is changing from inpatient towards a greater outpatient activity. From 2002 to 2009, there was a 50% increase in outpatient activity in Denmark, primarily related to the number of contacts per patient [1]. At the same time, there appears to be a growing need of health care services especially for the growing group of patients with chronic diseases and an increased focus on patient involvement. The challenge is to manage this without compromise on quality of care and patient outcomes. Follow-up visits for patients with chronic diseases in secondary care are traditionally based on regular pre-booked visits, which may be arranged when the patient is well. Thus patients as well as clinicians may find such visits unnecessary. The volume of appointments leads to capacity issues in outpatient clinics that struggle to respond rapidly to patients’ requests for help [2].
One way of handling this challenge may be to let patients report essential information on health status and symptoms from home before or instead of visiting the outpatient clinic. Patients’ own reports on health condition are termed patient-reported outcomes (PRO). The American Food and Drug Agency definition of PRO, “A measurement based on a report that comes directly from the patient about the status of a patient’s health condition without interpretation of the patient’s response by a clinician or anyone else” [3], focuses on the source of information and points out the importance of the patient perspective. The use of PRO in clinical practice is becoming increasingly common, and studies have reported improved patient-clinician communication, more effective self-management, and better utilisation of resources when PROs are used, whereas findings related to effects on patient outcomes are less consistent [4–7].
PRO may facilitate patient involvement because the problems reported as important by the patient are taken into consideration in the decision-making process [8–10]. However, patient involvement is not a goal in itself but rather a means to increase the patient’s self-management. Self-management refers to the individual’s ability to manage symptoms, treatment, physical and psychosocial consequences, and life style changes inherent in life with a chronic disease [11]. In practice, PRO is supposed to promotes a patient-centred dialogue between the patient and the clinicians in which the patient’s view and opinion on his health are included. Thus, implementing PRO into clinical practice allows patients to actively participate in their own care, by which their self-management may improve [8].
AmbuFlex is a generic clinical PRO system which is not limited to specific patient groups, organisations or medical record systems [12]. As of December 2015, AmbuFlex had been implemented in nine patient groups at 15 outpatient clinics in Denmark [13]. An analysis initiated by the Danish government based on experiences with AmbuFlex has demonstrated a positive national business case and considerable quality gain [14]. The Danish government and Danish regions, who run the public hospitals, have decided on an agreement for nationwide implementation of PRO in three diagnostic group, including epilepsy, before 2020. AmbuFlex was implemented for epilepsy outpatients at Aarhus University Hospital in March 2012 and is now used at three neurological departments in the Central Denmark Region. As of August 2016, 4,513 epilepsy outpatients have been referred to AmbuFlex, which are about two-thirds of all epilepsy outpatients in the region. The PRO questionnaire used contains information on specific aspects of daily life with epilepsy and has been developed in close cooperation with clinicians and patients. Face validity is fundamental and has been ensured during the development of the questionnaire [12, 13]. A graphical PRO overview is presented to the clinicians, who use the PRO data for clinical decisions together with other available clinical data in the record to decide whether the patient needs a visit or not. If a PRO questionnaire is used to evaluate the patient’s need for a hospital visit, the PRO data must be obtained outside the hospital. This is called tele-patient-reported outcome (telePRO) [12]. Experiences from epilepsy outpatient clinics have shown that of 8,256 PRO-based contacts, 48% were handled without additional contact to the patient other than the PRO questionnaire [13]. A preliminary interview study has indicated that patients experience greater flexibility in care, the saving of time, improved communication with the clinicians, and increased knowledge about their own disease [13, 15].
The AmbuFlex method used at the three neurological departments is called standard telePRO. In standard telePRO, regular scheduled visits are replaced with fixed questionnaires at intervals similar to those of the former pre-booked visits. A patient-initiated approach “open access” telePRO has been developed in which patients have access to their own PRO data and are able to initiate contact with the clinic by filling in a PRO questionnaire. A review by Whear et al. investigated the effectiveness of patient-initiated clinics in chronic conditions in secondary care and included seven randomised trials. The review found that the risk of harm from using the patient-initiated clinic model is low in patients with breast cancer, inflammatory bowel disease, and rheumatoid arthritis. The included studies found few significant differences in clinical outcomes between traditional appointment scheduling and the patient-initiated follow-up method. In four of the studies, the patient-initiated model was associated with savings in clinician time and resource use [2]. A review by Taneja et al, that included five of the same randomised studies reached the same
conclusion [16], while another review showed no significant differences in psychological and health-related quality of life outcomes between consultant-led and patient-initiated clinics. Patients have reported better satisfaction in patient-initiated clinics compared to usual care [17]. The patient-initiated method used was broadly the same in the studies included in the three reviews. Patients could request clinical advice by calling the clinic and, if necessary, arranging an appointment to see a clinician. However, none of the included studies used PRO as the main access point in the open access intervention, and all studies contain methodological limitations [2, 16, 17].
Objectives
The aim of this study is to provide insight into the effects of patient-initiated telePRO follow-up. The specific aims are to compare resource utilisation, quality of care, and the patient perspective of two outpatient follow-up activities: a) standard telePRO (fixed-interval telePRO follow-up) and b) open access telePRO (patient-initiated telePRO follow-up). We hypothesise that 1. Number of contacts is less in open access telePRO, 2. Quality of care in open access telePRO is at least as good as in standard telePRO, and 3. Patient self-management and experiences in open access telePRO are better than standard telePRO.
Methods
The study follows the (Additional file 1: SPIRIT checklist): Standard protocol items for clinical trials [18].
Design
This study is a pragmatic two-arm randomised controlled trial. Participants are randomly assigned to one of two follow-up activities: (a) standard telePRO or (b) open access teleRPO.
Study population
Participants are epilepsy outpatients recruited from the epilepsy clinic at Aarhus University Hospital in Central Denmark Region, Denmark.
Inclusion criteria
a) Age ≥ 15 years
b) Diagnosis or suspicion of epilepsy (IC-D 10 codes: G40, Z033a, DR568 and DR568E)
c) Already referred to standard telePRO by a clinician
d) Able to answer the questionnaire via the Internet, indicated by having answered the last questionnaire via the Internet
Exclusion criterion
a) Referred to telePRO follow-up with proxy questionnaire. Patients can be referred to a proxy questionnaire if they have cognitive problems and need help from a relative or health professionals.
Intervention
Reference group – standard telePRO
AmbuFlex (standard telePRO) is used in three epilepsy outpatient clinics in Central Denmark Region. In standard telePRO, outpatient follow-up activity is determined by a clinician and patients receive a questionnaire at fixed intervals (3, 6, or 12 months). The questionnaire includes information about aspect of daily life with epilepsy such as seizures, symptoms, medication adherence, and social aspects. Responses are automatically processed according to a specific algorithm and given a “green”, “yellow”, or “red” status. A red status indicates that the patient needs or wishes contact with the clinic, a green status that the patient has no current need of attention, while a yellow status indicates that the patient may need to be seen in the clinic, but a clinician has to decide whether further contact is needed. The patient can always overrule a decision by requesting contact. They can choose two different contact forms in the questionnaire: telephone consultation or a face-to-face consultation at the clinic. Non-responders get three reminders and are contacted if do not respond. Clinicians keep track of incoming yellow and red responses, and non-responders, and this information is presented on a PRO alert list. The PRO overview (Fig. 1) is presented graphically to the clinician within the electronic health record system, and used as decision aid together with other available health record information to decide whether the patient needs a visit or not [13].
Intervention group – open access telePRO
In open access, contact to the outpatient clinic is initiated by the patient by filling in a PRO questionnaire. The same questionnaire is used as in standard telePRO, but the patient decides when to respond. The patients can access a PRO overview, “My Epilepsy”, customised for patient use via a secure login at the Danish national health website “Sundhed.dk”. The clinicians handle questionnaires in the same way as in standard telePRO.
The open access website “My Epilepsy”: design and features
A prototype website, “My Epilepsy”, was developed to collect PRO in patient-initiated outpatient follow-up. The website was linked with the Danish National Health Website ‘Sundhed.dk’. The website, “My Epilepsy”, was customised for patient use and designed to allow patients to: a) answer a PRO questionnaire to get in contact with the clinic, b) view their personal PRO data (previously questionnaire responses), c) view information about the epilepsy questionnaire and specific questions, and d) have access to contact information to
the epilepsy outpatient clinic. A research team that included, outpatients with epilepsy and experts in telePRO, patient involvement, software technology, clinical epilepsy, provided inputs to design the prototype website. The research team developed the initial website specifications, constructed the website, and elicited feedback from epilepsy outpatients (n = 6), using cognitive interviewing techniques to study the manner in which the patients understood and responded to the website. The interface is shown in Fig. 2. Patients emphasised the importance of a user-friendly interface with clear and concise information. Patients were interested in tracking change over time and in using the website because it gave them the potential to communicate with their clinicians at a time decided by themselves. They found it conceivable that access to their previous questionnaires could give them a better understanding of their chronic disease. Finally, they pointed out the need for a telephone number if they required immediate contact. Patients had few problems assessing and using the site.
The website consists of four core elements:
a) Answer questionnaire: Here, patients can answer the epilepsy questionnaire when they need to get in contact with the clinic. The questionnaire is the same as in standard telePRO. When the patient has completed the questionnaire, the response is automatically sent as a red request to the PRO alert list at the epilepsy clinic. The clinician assesses the response and contacts the patient as soon as possible. The clinic has reserved appointments in their booking system to ensure that patients get a quick appointment. As a “safety net”, patients have to answer the questionnaire before twice the fixed interval has elapsed. For example, if the patient is referred with a 12-month interval, the patient has to respond within two years. If not, the patient is automatically sent a questionnaire, given a red status, and is contacted by a clinician.
b) Previous answers: In this element, all of the patient’s previous questionnaire responses are available. Patients have access to a PRO overview interface and specific and detailed questionnaire responses in the same manner as the clinicians. The overview interface is shown in Fig. 3. It is customised to monitor selected PRO data and to illustrate changes in health status over time. Colour codes indicate the severity of the symptoms reported by the patient. A red or orange bar indicates a self-reported problem, a yellow bar some problem, and a green bar indicates no problems. Note: Labels were translated from Danish.
Fig. 2 The open access telePRO website "My Epilepsy". Note: Labels were translated from Danish.
Fig. 3 PRO response overview customised to outpatients with epilepsy. A red or orange bar indicates a self reported problem, a yellow bar some problem, and green bar indicates no problems. Note: Labels were translated from Danish.
Randomisation
Pre-randomisation designs prevent change in behaviour in the control group because of disappointment about the allocation [19]. Eligible standard telePRO participants will be pre-randomised to standard telePRO follow-up (no change) or open access telePRO follow-up. Control as well as intervention participants receive the baseline questionnaire together with the fixed PRO questionnaire. The clinicians respond to the fixed PROs as usual and will not have access to the baseline questionnaire. Control participants will continue with fixed interval questionnaires and no change will be undertaken. Intervention participants will receive detailed information about the open access approach two weeks after a clinician’s response to the fixed PRO questionnaire. The study coordinator will forward written information to the included intervention participants. Individuals who do not agree to participate will continue with standard telePRO follow-up. Due to the nature of the intervention neither patients nor clinicians can be blinded to allocation. The randomisation is performed with an algorithm developed as part of the WestChronic software [12]. The allocation ratio open access/standard is 0.55/0.45. This ratio was selected to account for an expected number of patients in the open access arm who do not wish to participate.
Study timeline
Inclusion and randomisation with baseline assessments will take place from January 2016. Follow-up assessment will take place 18 months after randomisation. Baseline and follow-up assessments are shown in Table 1. Figure 4 presents the inclusion of patients and the stages in the study.
Outcomes
The effects of patient-initiated follow-up (open access telePRO) will be evaluated with regard to three different aspects: resource utilisation, quality of care, and the patient perspective. Resource utilisation will constitute the primary outcome, measured by number of contacts. Quality of care and the patient perspective constitute the secondary outcomes. Quality of care includes pivotal clinical quality measures (mortality, seizure, and treatment side effects) as well as more general patient-oriented quality measures (well-being and general health). The patient perspective includes measures related to self-management, such as health literacy, self-efficacy, and patient activation. Measures of confidence, safety, and satisfaction will be used to describe patient experiences. The patient perspective is primarily
| Outcomes | Data sources | Measurement/month |
|---------------------------|-------------------------------------------------------------|-------------------|
| Resource utilisation | | |
| 1. Number of contacts | The Hospital Business Intelligence Register, Central Denmark Region | 0–18 |
| Quality of care | | |
| 2. Well-being | WHO-Five Well-being Index (WHO-5) | 0, 18 |
| 3. General health | Item from The Short Form Health Survey (SF-36) | 0, 18 |
| 4. Mortality | The Hospital Business Intelligence Register, Central Denmark Region | 0–18 |
| 5. Number of seizures | Item from the epilepsy questionnaire, Central Denmark Region | 0, 18 |
| 6. Treatment side effects | Item from the epilepsy questionnaire, Central Denmark Region | 0, 18 |
| Patient perspective a | | |
| 7. Health literacy | The Health Literacy Questionnaire (HLQ) sub scale 4, 6 and 9 | 0, 18 |
| 8. Self-efficacy | General Self-Efficacy scale (GSE) | 0, 18 |
| 9. Patient activation | Items from Patient Activation Measure (PAM) | 0, 18 |
| 10. Confidence, safety, and satisfaction | Items from a PREM questionnaire, Danish Cancer Society | 0, 18 |
* The patient perspective is primarily explored by qualitative methods in a complementary PhD study
explored by qualitative methods in a complementary PhD study. An overview of primary and secondary outcomes, data sources, and measurement timeline is shown in Table 1.
**Primary outcome**
**Resource utilisation**
Number of contacts includes all contacts with the outpatient clinic in the study follow-up period, including face-to-face consultations with a physician, face-to-face consultations with a nurse, and telephone consultations. In addition, other health care contacts will be gathered, e.g., epilepsy-related emergency room visits and hospitalisations as well as hospitalisation related to co-morbidity. Data will be gathered from the Hospital Business Intelligence Register in Central Denmark Region.
**Secondary outcomes**
**Quality of care**
Patients’ well-being will be measured by using the Danish version of WHO-Five Well-being Index (WHO-5). WHO-5 was developed by the World Health Organisation for the assessment of well-being among patients with diabetes [20]. WHO-5 consists of five positively worded items reflecting present mental well-being within the previous two weeks. Items are rated on a 6-point scale ranging from 5 “all of the time” to 0 “at no time”. The instrument has demonstrated sufficient psychometric properties in a wide range of chronic conditions [20, 21]. Patients’ general health will be measured by using one item from the Danish version of The Short Form Health Survey (SF-36); “In general, would you say your health is: excellent, very good, good, fair, or poor” [22, 23]. The validity and reliability of this item are well documented [24]. Data on mortality will be gathered from the Hospital Business Intelligence Register in Central Denmark Region. Finally, number of seizures and treatment side effects will be collected from ad hoc items in the epilepsy questionnaire used at epilepsy clinics in Central Denmark Region. The validity and reliability of these items have not yet been documented.
**Patient perspective**
Health literacy will be measured by using the Danish version of Health Literacy Questionnaire (HLQ) [25, 26]. HLQ was developed to measure a wide range of health literacy needs of people in the community. The HLQ includes nine conceptually subscales with a total of 44
items containing five scales with agree/disagree response options and four scales with difficulties in perform tasks response options. The HLQ has well-documented psychometric properties [26]. In this study, the HLQ sub-scales 4, 6, and 9 will be used; 4. Social support for health, 6. Ability to actively engage with healthcare providers, 9. Understand health information well enough to know what to do. Self-efficacy will be measured by using the Danish version of General Self-Efficacy Scale (GSE) [27, 28]. GSE was designed to assess optimistic self-belief to cope with difficult demands in life [27, 28]. GSE includes ten items with a response range from 1 “not at all true” to 4 “exactly true”. The GSE scale has been used in a range of research projects in different countries and populations, where it typically yielded sufficient psychometric properties [29]. Patient activation will be measured by two ad hoc items developed with inspiration from the Danish version of the Patient Activation Measure (PAM) [30]. Confidence, safety, and satisfaction will be measured by using ad hoc items developed with inspiration from a Danish PREM (patient-reported experience measure) questionnaire from the Danish Cancer Society.
In addition, the patient perspective will be explored in a complementary qualitative PhD study. The primary aim of this study is to explore the mechanisms of actions related to standard telePRO and open access telePRO. Interpretive description (ID) will be used as the research approach [31]. Patients’ experiences with telePRO will be explored in individual interviews and participant observations in outpatient clinics. The target group for participation is patients with epilepsy, referred to standard telePRO or open access telePRO follow-up in the three neurological departments in Central Denmark Region.
Other measurements
Demographic information such as sex, age, education, marital status, and duration of epilepsy diagnosis will be obtained from baseline questionnaires.
Sample size
Statistical power was calculated for the primary outcome number of contacts. Based on literature review [32] the number of consultations (n) and standard deviation (SD) was; $n = 4.64$, $(SD = 2.38)$ in conventional follow-up and $n = 4.12$, $(SD = 3.41)$ in open access follow-up. We expect at least a difference of one contact between the groups. Given a statistical power of 90%, $p$-value 0.05, and allocation ratio 0.8, we will need a sample size of 172 patients in the standard telePRO group and 214 patients in the open access telePRO group. To account for attrition and loss to follow-up, we will recruit a total of approximately 500 participants. For qualitative data, a purposeful sample of at least five participants from each group will be interviewed.
Analyses
Descriptive statistics will be used to describe differences in the baseline characteristics of participating patients in the two arms of the trial. Statistical analysis will be intention to treat, whereby all randomised participants will be included in the analysis according to their randomised allocation. The primary outcome, total number of contacts in the two arms, will be analysed using a sample t-test. If the distribution of data is skewed, we will use medians and nonparametric tests. For secondary outcomes, a chi-square test or logistic regression will be used for dichotomous outcome data and sample t-test or multiple linear regression analysis will be used for continuous outcome data. Non-parametric tests will be used if continuous data are not normally distributed. Demographics covariates (sex, age, education, marital status, and epilepsy diagnosis duration) will be included in the per-protocol analysis.
For qualitative data, ID will be employed as the overriding research approach. ID is an inductive research strategy in which constant comparative method with concurrent data collection and analysis is utilised to gain a deeper insight and understanding of human experiences within their natural context. The result is a comprehensive interpretation, potentially a model of explanation of the phenomenon under study, which can provide clinical practice with a research-based choice of action [31]. ID is considered appropriate in the present study because the approach is suited for exploration of specific clinical issues, in this case how patients with epilepsy experience standard and open access telePRO follow-up.
Ethics
The risks to participants are considered to be minimal as all eligible participants are referred to standard telePRO follow-up by clinicians at the epilepsy clinic. As a “safety net” to ensure that no patients are lost in the open access arm, the patients have to answer the epilepsy questionnaire before twice the fixed interval has elapsed. If lack of response the patient is reallocated into standard telePRO with a red status and a clinician will contact the patient. Furthermore, all patients are informed to call the clinic in pressing need of attention.
The Danish Data Protection Agency has accepted the study. In addition, the Danish research ethics committee in Central Denmark Region was contacted and has stated that approval from the committee is not necessary for this present study. Therefore, written informed consent was not obtained from the participants. Prior to study participation patients in the intervention group receive written information about the study. Study participation is entirely voluntary and participants are informed they can withdraw from the study at any time without affecting future care. In the qualitative complementary PhD study,
the participants gave written informed consent prior to enrolment, and the study was approved by the Danish Data Protection Agency.
Data security
All data activities in the study are documented and stored in the WestChronic web-system [12]. The system is physically located in Central Denmark Regions Server Park behind the firewall and Threat Management Gateway. Regular backup is performed weekly. All data transactions fulfill conditions established by the Danish Data Protection Agency.
Discussion
During the last decade, the use of PRO in clinical practice has become increasingly common, and to our knowledge, AmbuFlex is the first generic PRO system that uses PRO as the basis for outpatient follow-up [13]. The focus of this trial will be to evaluate the effect of a patient-initiated open access telePRO intervention compared to standard telePRO with respect to resource utilisation, quality of care, and the patient perspective. Ideally, we would have preferred to compare the two arms (standard and open access) with conventional follow-up with pre-booked outpatient visits to the clinic. However, this was not possible because the epilepsy clinics in Central Denmark Region have used standard telePRO follow-up since 2012. Thus, we will compare two rather similar outpatient follow-up activities, which will probably result in only small differences in effect between the groups. Evaluation of the effect must be done using reliable, valid, and clinically meaningful measures. This study includes outcome measures based on recommendations from clinical experts, researchers, and the literature [33].
Loss to follow-up is one of the main concerns in randomised controlled studies [34]. Loss to follow-up in this study is related to the open access group of patients since study participation in the open access arm is entirely voluntary, and participants can choose to continue with standard telePRO follow-up. Loss of statistical efficiency can be overcome by increased the number of participants in the study [19]. We have taken this into consideration and will include 10% more patients in the open access telePRO arm. In addition, we will recruit a larger number of participants than the minimum sample size calculation indicated.
Only web-responders will be included in the open access arm, and the results may therefore be generalizable only to this subgroup of epilepsy patients. These patients may differ with respect to education, age, and use of new technologies compared to the entire group of epilepsy patients. In another study in progress, the aim is to examine determinants for referral to telePRO follow-up. Data from this study can be used to compare the study population with the entire group of patients with epilepsy.
Another potential challenge may be how individuals in the intervention group will use the “My Epilepsy” website. Some patients are better able to decide themselves when they need to contact the clinic, while others are more reserved and afraid to be a nuisance. Several patients that have used standard telePRO have pointed out the benefit of getting a fixed questionnaire once a year. They do not believe they would remember to answer if they had to do it on their own. This could signify that even though they may not feel the need for a clinical appointment, but do feel a form of security in answering the fixed interval questionnaire. This will be taken into consideration in the study, since all patients in the intervention group will receive a questionnaire if they do not respond within two times the referred interval, for example, within 24 months if they are assigned a 12-month questionnaire interval. Another concern could be that patients in the open access group could choose to make a call instead of answering the questionnaire when they need to get in contact with the clinic. If they behave in this way, the benefit of using PRO in clinical practice will be reduced.
Standard telePRO has been well integrated into clinical practice in three epilepsy clinics in Central Denmark Region since 2012. A new patient-initiated approach has been developed that may result in potential benefits in terms of the patient perspective and resource utilisation. The potential benefits as well as possible drawbacks need to be evaluated. We have decided to combine qualitative and quantitative research methods in two parallel PhD studies. The intention of the complementary qualitative PhD study is to further explain the findings from the randomised study by providing a description of the various ways in which telePRO is manifested and an interpretation of the underlying mechanisms of action. The two studies will complement each other and contribute with important research-based knowledge to guide future telePRO interventions in relation to effect on resource utilisation, quality of care, and the patient perspective.
Trial status
On going.
Additional file
Additional file 1: SPIRIT 2013 Checklist: Recommended items to address in a clinical trial protocol and related documents*. (DOC 122 kb)
Abbreviations
GSE: General Self-Efficacy scale; HLQ: Health literacy questionnaire; ID: Interpretive description; PAAM: Patient activation measure; PREM: Patient-reported experience measure; PRO: Patient-reported outcome; SD: Standard deviation; SF-36: Short form health survey; WHO-5: WHO-Five well-being index
Acknowledgements
This study is funded by Aarhus University, the Health Research Fund of Central Denmark Region, and the Danish foundation TrygFonden.
Funding
Aarhus University, the Health Research Fund of Central Denmark Region, and the Danish foundation TrygFonden have funded this PhD study. The funding sources had no role in design of this study and will not have any role during its execution, analysis, interpretation of the data, or decision to submit results.
Availability of data and materials
Not applicable.
Authors’ contributions
NHH and LMVS conceived the study in collaboration with PS, AdT, and KL. LMVS, KHP, CTM, and NHH participated in the development of the open access intervention. KHP and NHH developed the open access website and the randomisation algorithm as part of the WestChronic software. LMVS participated in recruitment of participants, data collection and registration. LMVS drafted the manuscript. All authors have read and approved the final manuscript.
Competing interests
The authors declare that they have no competing interests.
Consent for publication
Not applicable.
Ethics approval and consent to participate
The Danish research ethics committee in Central Denmark Region was contacted. According to Danish law, the committee has stated that approval from the committee is not necessary for this present study. Therefore, written informed consent was not obtained from the participants.
Data monitoring and dissemination policy
A data monitoring committee was not needed due to minimal risk in the intervention group. The results of the study will only be published in peer reviewed journals.
Protocol version
Issue date 6 Jan 2017, version number: 01.
Author details
1AmbuFlex, Regional Hospital West Jutland, Herning, Denmark. 2The Research Programme in Patient involvement, Aarhus University Hospital, Aarhus, Denmark. 3Department of Rheumatology, Aarhus University Hospital, Aarhus, Denmark. 4Department of Clinical Medicine, Aarhus University, Aarhus, Denmark. 5Department of Neurology, Aarhus University Hospital, Aarhus, Denmark. 6Department of Public Health, Aarhus University, Aarhus, Denmark. 7Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark.
Received: 15 November 2016 Accepted: 13 January 2017
Published online: 26 January 2017
References
1. Inderigs- og Sundhedsministeriet. Analyse af aktiviteten i sygehusvæsenet. 2010. http://www.sum.dk/-/media/Filer%20-%20Publikationer/Analyse-af-syg...Accessed 6 June 2016.
2. Whear R, Abdul-Rahman AK, Boddy K, Thompson-Coon J, Perry M, Stein K. The clinical effectiveness of patient initiated clinics for patients with chronic or recurrent conditions managed in secondary care: a systematic review. PLoS One. 2013;8(10):e74774.
3. US Department of Health and Human Services Food and Drug Administration. Guidance for industry: patient-reported outcome measures: use in medical product development to support labeling claims. 2009. http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/UCM193282.pdf. Accessed 6 Jun 2016.
4. Marshall S, Haywood K, Fitzpatrick R. Impact of patient-reported outcome measures on routine practice: a structured review. J Eval Clin Pract. 2006;12(5):559–68.
5. Chen J, Ou L, Hollis SJ. A systematic review of the impact of routine collection of patient reported outcome measures on patients, providers and health organisations in an oncologic setting. BMC Health Serv Res. 2013;13:211–6963. 13-211.
6. Johansen MA, Bernsten GK, Schuster T, Henriksen E, Horsch A. Electronic symptom reporting between patient and provider for improved health care service quality: a systematic review of randomized controlled trials, part 2: methodology, quality and effects. J Med Internet Res. 2012;14(5):e126.
7. Valderas JM, Kotzeva A, Espallargues M, Guyatt G, Ferrans CE, Halyard MY, Revicki DA, Symonds T, Parada A, Alonso J. The impact of measuring patient-reported outcomes in clinical practice: a systematic review of the literature. Qual Life Res. 2008;17(2):179–93.
8. Bennett AV, Jensen RE, Basch E. Electronic patient-reported outcome systems in oncology clinical practice. CA Cancer J Clin. 2012;62(5):336–47.
9. Boyce MB, Browne JP, Greenhalgh J. The experiences of professionals with using information from patient-reported outcome measures to improve the quality of healthcare: a systematic review of qualitative research. BMJ Qual Saf. 2014;23(6):508–18.
10. Greenhalgh J. The applications of PROs in clinical practice: what are they, do they work, and why? Qual Life Res. 2009;18(1):115–23.
11. Barlow J, Wright C, Sheaby J, Turner A, Hainsworth J. Self-management approaches for people with chronic conditions: a review. Patient Educ Couns. 2002;48(2):177–87.
12. Hjollund NH, Larsen LP, Beiring K, Johnsen SP, Riskjaer E, Schougaard LM. Use of Patient-Reported Outcome (PRO) measures at group and patient levels: experiences from the general integrated PRO system, WestChronic Interact. J Med Res. 2014;3(1):e5.
13. Schougaard LM, Larsen LP, Jessen A, Sidenius P, Dorflinger L, de Thurah A, Hjollund NH, AmbuFlex tele-patient-reported outcomes (telePRO) as the basis for follow-up in chronic and malignant diseases. Qual Life Res. 2016;25(3):525–34.
14. Danish Agency for Digitalisation and PA Consulting Group. Analyse af Patient Rapporterede Oplysninger (PRO). 2015. http://www.digst.dk/digital/-velfaerd/Strategi-for-digital-velfaerd/Rapporter-og-analyser/Sundhed/-Analyse af-patientrapporterede-oplysninger. Accessed 6 Jun 2016.
15. Center for Telemedicin, Region Midjylland. AmbuFlex – Behovsstyrede ambulante patientbeford med spørgeskemaer baseret klinik selvmontering. Bilag Patientperspektiv – interview med 13 epilepsipatienter. 2013 (In Danish).
16. Taneja A, Su’a B, Hill AG. Efficacy of patient-initiated follow-up clinics in secondary care: a systematic review. Intern Med J. 2014;44(12a):1156–60.
17. Whear R, Abdul-Rahman AK, Thompson-Coon J, Boddy K, Perry MG, Stein K. Patient initiated clinics for patients with chronic or recurrent conditions managed in secondary care: a systematic review of patient reported outcomes and patient and clinician satisfaction. BMC Health Serv Res. 2013;13:501. 6963-13-501.
18. Chan AW, Tetzlaff JM, Altman DG, Laupacis A, Gotzsche PC, Krleza-Jeric K, Hrobjartsson A, Mann H, Dickersin K, Berlin JA, Dore CJ, Parulekar WR, Summerskill WS, Groves T, Schulz CF, Sox HC, Rockhold FW, Rennie D, Moher D. SPIRIT 2013 statement: defining standard protocol items for clinical trials. Ann Intern Med. 2013;158(3):200–7.
19. Zelen M. A new design for randomized clinical trials. N Engl J Med. 1979;300(22):1242–5.
20. Bech P, Gudex C, Johansen KS. The WHO (Ten) well-being index: validation in diabetes. Psychother Psychosom. 1996;65(4):183–90.
21. Topp CW, Ostergaard SD, Søndergaard S, Bech P. The WHO-5 well-being index: a systematic review of the literature. Psychother Psychosom. 2015;84(3):167–76.
22. Bjørner JB, Damsgaard MT, Watt T, Groenvold M. Tests of data quality, scaling assumptions, and reliability of the Danish SF-36. J Clin Epidemiol. 1998;51(11):1001–11.
23. Bjørner JB, Thunedborg K, Kristensen TS, Modvig J, Bech P. The Danish SF-36 Health Survey: translation and preliminary validity studies. J Clin Epidemiol. 1998;51(11):1199–9.
24. Gandek B, Ware Jr JE, Aaronsen NK, Alonso J, Apolone G, Bjørner J, Brazier J, Bullingner M, Fukuhara S, Kaasa S, Leplege A, Sullivan M. Tests of data quality, scaling assumptions, and reliability of the SF-36 in eleven countries: results from the IQOLA Project. International Quality of Life Assessment. J Clin Epidemiol. 1998;51(11):1149–58.
25. Maindal HT, Kayser L, Norgaard O, Bo A, Elsworth GR, Osborne RH. Cultural adaptation and validation of the Health Literacy Questionnaire (HLQ): robust nine-dimension Danish language confirmatory factor model. Springerplus. 2016;5(1):1232. 016-2887-9 . eCollection 2016.
26. Osborne RH, Batterham RW, Elsworth GR, Hawkins M, Buchbinder R. The grounded psychometric development and initial validation of the Health Literacy Questionnaire (HLQ). BMC Public Health. 2013;13:658. 2458-13-658.
27. Luszczynska A, Scholz U, Schwarzer R. The general self-efficacy scale: multicultural validation studies. J Psychol. 2005;139(5):439–57.
28. Schwarzer R, Jerusalem M. Generalized self-efficacy scale. In: Weinman J, Wright S, Johnston M, editors. Measures in health psychology: a user’s portfolio. causal and control beliefs. Windsor: NFER-NELSON; 1995. p. 35–7.
29. Scholz U, Gutiérrez-Dona B, Sud S, Schwarzer R. Is general self-efficacy a universal construct? Psychometric findings from 25 countries. Eur J Psychol Assess. 2002;18(3):242–51.
30. Maindal HT, Sokolowski I, Vedsted P. Translation, adaptation and validation of the American short form patient activation measure (PAM13) in a Danish version. BMC Public Health. 2009;9:209. 2458-9-209.
31. Thorne S. Interpretive Description. Walnut Creek: Left Coast Press, Inc; 2008.
32. Williams JG, Cheung WY, Russell IT, Cohen DR, Longo M, Leny B. Open access follow up for inflammatory bowel disease: pragmatic randomised trial and cost effectiveness study. BMJ. 2000;320(7234):544–8.
33. Fountain NB, Van Ness PC, Swain-Eng R, Tonn S, Bever Jr CT, American Academy of Neurology Epilepsy Measure Development Panel and the American Medical Association-Convened Physician Consortium for Performance Improvement Independent Measure Development Process. Quality improvement in neurology: AAN epilepsy quality measures: report of the quality measurement and reporting subcommittee of the American academy of neurology. Neurology. 2011;76(1):94–9.
34. Sorensen HT, Lash TL, Rothman KJ. Beyond randomized controlled trials: a critical comparison of trials with nonrandomized studies. Hepatology. 2006; 44(5):1075–82. | 2025-03-04T00:00:00 | olmocr | {
"pdf_page_numbers": [
[
0,
3690,
1
],
[
3690,
9530,
2
],
[
9530,
14444,
3
],
[
14444,
17064,
4
],
[
17064,
17392,
5
],
[
17392,
21679,
6
],
[
21679,
23924,
7
],
[
23924,
29552,
8
],
[
29552,
34931,
9
],
[
34931,
42577,
10
],
[
42577,
44734,
11
]
]
} |
357d7891478e9a80f7bb22633967ec22cfd4a640 | {
"olmocr-version": "0.1.59",
"pdf-total-pages": 9,
"total-fallback-pages": 0,
"total-input-tokens": 32543,
"total-output-tokens": 12451,
"fieldofstudy": [
"Physics"
]
} | Conditional quantum operation of two exchange-coupled single-donor spin qubits in a MOS-compatible silicon device
Mateusz T. Mądzik, Arne Laucht, Fay E. Hudson, Alexander M. Jakob, Brett C. Johnson, David N. Jamieson, Kohei M. Itoh, Andrew S. Dzurak & Andrea Morello
Silicon nanoelectronic devices can host single-qubit quantum logic operations with fidelity better than 99.9%. For the spins of an electron bound to a single-donor atom, introduced in the silicon by ion implantation, the quantum information can be stored for nearly 1 second. However, manufacturing a scalable quantum processor with this method is considered challenging, because of the exponential sensitivity of the exchange interaction that mediates the coupling between the qubits. Here we demonstrate the conditional, coherent control of an electron spin qubit in an exchange-coupled pair of $^{31}$P donors implanted in silicon. The coupling strength, $J = 32.06 \pm 0.06$ MHz, is measured spectroscopically with high precision. Since the coupling is weaker than the electron-nuclear hyperfine coupling $A \approx 90$ MHz which detunes the two electrons, a native two-qubit controlled-rotation gate can be obtained via a simple electron spin resonance pulse. This scheme is insensitive to the precise value of $J$, which makes it suitable for the scale-up of donor-based quantum computers in silicon that exploit the metal-oxide-semiconductor fabrication protocols commonly used in the classical electronics industry.
Building useful quantum computers is a challenge on many fronts, from the development of quantum algorithms\textsuperscript{3} to the manufacturing of scalable hardware devices\textsuperscript{2}. For the latter, adapting the fabrication processes already in use in the classical electronics industry—silicon-based metal-oxide-semiconductor (MOS) processing\textsuperscript{13–16} and ion implantation\textsuperscript{17–19}—to the construction of quantum hardware would represent a great technological head start. This was the insight that triggered the first proposal of encoding quantum information in the spin state of donor atoms in silicon\textsuperscript{10}. Qubits defined by individual donor-bound electron spins have demonstrated quantum gate fidelities beyond 99.9\% (ref. 11), and coherence lifetimes approaching 1 s (ref. 12). The next challenge is the demonstration of robust two-qubit logic operations, necessary for universal quantum computation.
In their simplest form, two-qubit logic gates can be executed using three distinct strategies. The first requires the two qubits to have approximately the same energy splitting, $\epsilon_1 \approx \epsilon_2$, and turning on the qubit–qubit interaction $J$ for a finite amount of time\textsuperscript{14}, yielding a native SWAP gate\textsuperscript{15}. The second strategy implements a controlled-Z gate by dynamical control of $J$. The coupling is switched on for a calibrated time period, whereby the target qubit acquires a phase shift proportional to the change in precession frequency determined by the state of the control qubit\textsuperscript{16,17}. The third strategy implements a native controlled-rotation (CROT) gate via resonant excitation of the target qubit, whose transition frequency can be made to depend on the state of the control qubit.
The CROT gate is related to the controlled-NOT operation that appears in most quantum algorithms, but imparts an additional phase of $\pi/2$ to the target qubit. This gate requires the individual qubits’ energy splittings to differ by an amount $\delta \epsilon = |\epsilon_1 - \epsilon_2|$ much larger than their coupling $J$. It was used in early nuclear magnetic resonance (NMR) experiments\textsuperscript{18}, superconducting qubits\textsuperscript{19} and, more recently, was adapted to electron spin qubits in semiconductors, where the energy detuning $\delta \epsilon$ can be provided by a difference in Landé $g$-factors between the two electron spins\textsuperscript{16,20} or by a magnetic field gradient\textsuperscript{21}. For electron spin qubits, the coupling $J$ originates from the Heisenberg exchange interaction. The main advantage of this type of gate is that it can be performed while keeping $J$ constant—an essential feature when locally tuning $J$ is either impossible or impractical. Moreover, the precise value of $J$ is unimportant, as long as it is smaller than $\delta \epsilon$, and larger than the resonance linewidths.
Fig. 1 Two-qubit metal-oxide-semiconductor device. a Scanning electron micrograph of a device similar to the one used in the experiment, with labels describing the function of the aluminum gates on the surface. b Schematic cross section of the device, depicting a pair of donors $\approx$10 nm beneath a thin SiO$_2$ dielectric, inside an isotopically enriched $^{28}$Si epilayer.
in the control donor, respectively; their average is \( A = \frac{(A_1 + A_2)}{2} \) and their difference \( \Delta A = (A_1 - A_2) \).
To simplify the problem, we draw the energy levels diagrams shown in Fig. 2a, where we assume that both donors have the same hyperfine coupling \( A = 100 \text{ MHz} \). A more general and extensive discussion of the two-donor spin Hamiltonian is given in the Supplementary Note 1.
When the nuclei are in a parallel configuration (\( |\uparrow\uparrow\rangle \) or \( |\downarrow\downarrow\rangle \)), the uncoupled electron spins have the same energy splitting. Upon introducing an exchange coupling \( J \), the electronic eigenstates become the singlet \( |S\rangle = (|\uparrow\downarrow\rangle - |\downarrow\uparrow\rangle)/\sqrt{2} \) and triplet \( |T_+\rangle = |\uparrow\downarrow\rangle + |\downarrow\uparrow\rangle)/\sqrt{2} \); \( |T_0\rangle = |\uparrow\downarrow\rangle \); \( |T_-\rangle = |\downarrow\uparrow\rangle \) states. An oscillating magnetic field can induce electron spin resonance (ESR) transitions between the triplets, corresponding to the ESR lines \( \ell_2 \) and \( \ell_5 \) in Fig. 2b. The singlet state has a total spin of zero, and cannot be accessed by ESR. Since the energy splittings \( |T_+\rangle \leftrightarrow |T_0\rangle \) and \( |T_0\rangle \leftrightarrow |T_-\rangle \) are identical, an ESR transition can occur irrespective of the state of the control qubit. These unconditional resonances do not constitute two-qubit logic operations.
If instead, we prepare the nuclear spins in opposite orientations (\( |\uparrow\downarrow\rangle \) or \( |\downarrow\uparrow\rangle \)), the hyperfine interaction detunes the uncoupled electrons by \( \Delta \epsilon \equiv A \). Introducing a weak exchange coupling \( J \ll A \) results in electronic eigenstates of the form \( |\uparrow\downarrow\rangle \); \( |\downarrow\uparrow\rangle \); \( |\uparrow\downarrow\rangle \); \( |\downarrow\downarrow\rangle \), where \( |\uparrow\downarrow\rangle = \cos \theta |\uparrow\downarrow\rangle + \sin \theta |\downarrow\uparrow\rangle \); \( |\downarrow\uparrow\rangle = \cos \theta |\downarrow\uparrow\rangle - \sin \theta |\uparrow\downarrow\rangle \); and \( \tan(2\theta) = J/A \). In this case, for each antiparallel nuclear orientation there exist two distinct frequencies (\( \ell_1 \) and \( \ell_3 \) for \( |\uparrow\downarrow\rangle \); \( \ell_4 \) and \( \ell_6 \) for \( |\downarrow\uparrow\rangle \)), separated by \( J \), at which the target qubit would respond, depending on the state of the control. Therefore, a \( \pi \)-pulse on any of these resonance lines embodies a form of two-qubit CROT gate. Defining \( |\downarrow\rangle \) as the computational \( |1\rangle \) state, \( \ell_1 \) and \( \ell_4 \) yield CROT gates, i.e., rotations of the target qubit conditional on the control being in the \( |1\rangle \) state, while \( \ell_3 \) and \( \ell_6 \) yield zero-CROT gates (Fig. 2b).
Importantly, the ability to perform a CROT gate depends only on the ability to apply a selective \( \pi \)-pulse on one of the conditional resonances. The precise value of \( J \) is unimportant, as long as it exceeds the resonance linewidth (~10 kHz in our devices) and is smaller than \( A = 100 \text{ MHz} \). The value of \( J \) sets an approximate limit to the speed of the CROT operation\(^{13} \), since the spectral width of the CROT pulse (approximately equal to the Rabi frequency) must be lower than the frequency spacing \( = J \) between resonances conditioned on opposite control qubit states\(^{20} \). This limitation can be circumvented to some extent by replacing the simple resonant \( \pi \)-pulse with more sophisticated control schemes\(^{28,29} \), as demonstrated, e.g., in quantum dot systems\(^{20} \). Overall, this scheme affords a wide tolerance in the physical placement of the donors.
**Ion implantation strategies.** We fabricated two batches of devices designed to exhibit exchange interaction between donor pairs. In addition to the implanted \( ^{31}\text{P} \) donors, the devices include a single-electron transistor (SET) to detect the donor charge state, four electrostatic gates to control the donor potential, and a microwave antenna to deliver oscillating magnetic fields (see Fig. 1a).
The ion implantation step was executed using two different strategies. We first implanted a batch of devices with a low fluence of \( \text{P}^{+2} \) molecular ions, accelerated with a 20 keV voltage (corresponding to 10 keV/atom). When a \( \text{P}^{+2} \) molecule hits the surface of the chip, the two P atoms break apart and come to rest at an average distance that depends on the implantation energy. We chose the energy and the fluence \( (5 \times 10^{10} \text{ donors/cm}^2) \) to obtain well-isolated pairs; that is, we used the choice of acceleration energy to determine the most likely distance between donors resulting from an individual \( \text{P}^{+2} \) molecule (see Fig. 3c), and adapted the fluence to obtain a low probability of donor pairs overlapping with each other. A representative charge stability diagram of this type of devices, taken by sweeping the SET top gate voltage, stepping the donor gate voltage, and monitoring the
transistor current, is shown in Fig. 3a. A small number of isolated donor charge transitions—identifiable as near-vertical breaks in the regular patterns of SET current peaks—reveals well-separated individual donors, but too low a chance that two donors may be found in close proximity.
We thus fabricated another batch of devices, where we implanted a high fluence (1.25 × 10^{12} donors/cm^{2}) of single P\textsuperscript{+} ions at 10 keV energy. This yields a 25-fold increase in the donor density (see Fig. 3d, f and Supplementary Note 2), reflected in the much larger number of observed charge transitions in a typical stability diagram (Fig. 3b).
In a device with high-fluence P\textsuperscript{+} implanted donors, we identified a pair of charge transitions that, under suitable gate tuning, cross each other (Fig. 4a). As expected from the electrostatics of double quantum dots, this results in a “honeycomb diagram”, where the crossing between the charge transitions is laterally displaced by the mutual charging energy of the two donors\textsuperscript{30}. Note that this in itself does not provide any indication of the existence of a quantum-mechanical exchange coupling. Spin exchange would appear as a curvature in the sides of the honeycomb diagram\textsuperscript{32}, but its value would need to be >>1 GHz to be discernible in this type of experiment.
### Spectroscopic measurement of exchange interaction
The experimental methods for control and readout of the 31P donors follow well-established protocols. We perform single-shot electron spin readout via spin-dependent tunneling into a cold charge reservoir\textsuperscript{32,33}, and coherent control of the electron\textsuperscript{34} and nuclear\textsuperscript{35} spins via magnetic resonance, where an oscillating magnetic field is provided by an on-chip broadband microwave antenna\textsuperscript{36}.
Controlling the two pulsing gates above the donor implantation area allows us to selectively and independently control the charge state of each donor, which can be set to either the neutral D\textsuperscript{0} (electron number N = 1) or the ionized D\textsuperscript{+} (N = 0) state. In particular, we can freely choose the electrochemical potential of the donors with respect to each other, i.e., which of the donors ionizes first, while the other remains neutral (see Supplementary Movie).
On the stability diagram in Fig. 4a, we identify the four regions corresponding to the neutral (N = 1) and ionized (N = 0) charge states of each donor. For example, the boundary between the (0\textsubscript{c},0\textsubscript{t}) and (0\textsubscript{c},1\textsubscript{t}) regions is where the second donor (target) can be read out via spin-dependent tunneling to the SET island\textsuperscript{32,33}, while the first (control) remains ionized. This is because, when transitioning from, e.g., (0\textsubscript{c},1\textsubscript{t}) to (0\textsubscript{c},0\textsubscript{t}), the lost charge is absorbed by the island of the SET, which is tunnel coupled to the donors\textsuperscript{32}. At low electron temperatures (T\textsubscript{el} ≈ 100 mK) and in the presence of a large magnetic field (B\textsubscript{0} = 1.4 T), the tunneling of charge from donor to SET island becomes spin dependent, since only the |↑⟩ state has sufficient energy to escape from the donor. This mechanism provides the basis for the single-shot qubit readout\textsuperscript{33}. Therefore, the boundary (0\textsubscript{c},0\textsubscript{t}) ↔ (0\textsubscript{c},1\textsubscript{t}) is where we can observe the spin target donor, while it behaves as an isolated system, since the control donor is ionized at all times.
This expectation is confirmed by the ESR spectrum shown in Fig. 4b, which exhibits the two ESR peaks consistent with the two possible nuclear spin orientations of a single 31P donor\textsuperscript{35}. Since we are measuring a single atom, each trace normally contains only one peak, but occasionally the nuclear spin flips direction during the scan, so a single trace can also exhibit both peaks. Since the intrinsic ESR linewidth is very narrow (a few kilohertz in isotopically enriched 28Si (ref. 12)), finding the resonances is a time-consuming process. To speed this up, we used adiabatic spin inversion\textsuperscript{37} with a 6 MHz frequency chirp, resulting in a large electron spin-up fraction whenever a resonance falls within the frequency sweep range. The 6 MHz width of the frequency sweeps is the cause of the artificial width and shape of the resonances shown in Fig. 4b, c.
In the next step, we operate near the boundary (1\textsubscript{c},0\textsubscript{t}) ↔ (1\textsubscript{c},1\textsubscript{t}) where the target donor is read out, but the control donor is in the neutral D\textsuperscript{0} charge state, with an electron bound to it. Repeatedly measuring the ESR spectrum of the target donor, now reveals four possible ESR peaks. We interpret this as evidence for the presence of...
Despite the 6 MHz width of the ESR lines caused by the adiabatic inversion, it is clear by comparing Fig. 4b, c that the addition of a second electron introduces a significant Stark shift of both the hyperfine coupling $A_t$ and the $g$-factor $g_t$ of the target donor. While Stark shifts of donor hyperfine couplings and $g$-factors as a function of applied electric fields have been observed before, including on single donors, the observation of such shifts from the addition of a single charge in close proximity is novel. We anticipate that a systematic analysis of $A$ and $g$ Stark shifts under controlled conditions may help elucidating the precise nature of the electron wavefunctions in exchange-coupled donors, and benchmarking the accuracy of microscopic models.
Once the approximate frequencies of the ESRs are found by adiabatic inversion with chirped pulses, we switch to short constant-frequency pulses in order to measure linewidths limited solely by the pulse excitation spectrum. Here, unlike the experiments in Fig. 4, the four different nuclear spin configurations $|0\downarrow,0\downarrow\rangle, |0\uparrow,0\uparrow\rangle, |0\downarrow,0\uparrow\rangle, |0\uparrow,0\downarrow\rangle$ are deliberately set by projective nuclear readout followed, if needed, by coherent manipulation of the individual nuclear spins with NMR pulses. To address a specific nuclear spin, we keep the target donor ionized while the control donor is in the neutral state, with its electron spin in $|\uparrow\rangle$. This renders the NMR frequencies of each nucleus radically different, with $\nu_{\text{lin}} = g_t g_e B_0 = 24.173$ MHz and $\nu_{\text{nc}} = y_z B_0 + A_t/2 = 67.92$ MHz (see Supplementary Note 5 for details on the nuclear spin initialization).
The full ESR spectrum is presented in Fig. 5b along with insets that display the individual power-broadened resonance peaks. The experimental ESR spectrum shown in Fig. 5b can be compared to the numerical simulations of the full Hamiltonian (Eq. (1)) for the specific parameters of this donor pair. In addition to the exchange coupling $J$, the Hamiltonian contains five unknown parameters: the contact hyperfine couplings $A_t$ and $A_c$, the electron $g$-factors $g_t$ and $g_c$, and the static magnetic field $B_0$. Although $B_0$ is imposed externally, its precise value at the donor sites can have a slight uncertainty, e.g., due to trapped flux in the superconducting solenoid, or positioning the device slightly off the nominal center of the field. $B_0$ can be combined with the average of $g_t$ and $g_c$ to yield an average of the Zeeman energy $E_Z/h = (g_t + g_c)\muBB_0/2h$ of the donor electrons, which would rigidly shift the manifold of ESR frequencies. If, in addition, we assume that $g_t = g_c$, we are left with four free fitting parameters, $J/A_t, A_c, E_Z/h$ which can be extracted from the knowledge of the four ESR frequencies.
In the numerical simulations, we vary the hyperfine coupling of target and control donors, $A_t$ and $A_c$, to find a combination of values that allows matching all four ESR frequencies at the same magnitude of the exchange interaction $J$. Figure 5a shows the result of the simulation that best matches the ESR spectrum of Fig. 5b, using $A_t = 97.75 \pm 0.07$ MHz, $A_c = 87.57 \pm 0.16$ MHz, and $J = 32.06 \pm 0.06$ MHz. Errors indicate the 95% confidence levels. With these values, all ESR frequencies were matched with a maximum error $\Delta E = \max(|E_{\text{sim}} - E_{\text{exp}}|: E_{\text{sim}} - E_{\text{exp}}): |E_{\text{lin}} - E_{\text{exp}}|: |E_{\text{lin}} - E_{\text{exp}}|) = 4.74$ kHz, only slightly larger than the 30 kHz resolution of the measurement itself. This spectroscopic method constitutes the most accurate measurement of exchange interaction between phosphorus donor pairs obtained to date.
The extracted values of $A$ are far from the bulk value $A_{\text{bulk}} = 117.53$ MHz and rather different between the two donors. This could be due to local variations in lattice strain and electric fields within the device, which can be substantial even on a scale $\approx 10$ nm. Strain, in particular, varies dramatically near the tips of the control gates, and is well-known to cause changes in hyperfine coupling. The possible influence of strain on the spin relaxation time $T_1$ is discussed in ref. 43.
The observation of exchange coupling in the appropriate range for CROT operations was not unique to this particular device. A similar value of $J = 30$ MHz was measured on a second device, fabricated in the same batch and with the same high-dose P+ implantation strategy (see Supplementary Note 6).
The high-fluence devices contain many donors, as visible in the charge stability diagram (Fig. 3b). One may thus expect to encounter more complex clusters of interacting donors, instead of isolated pairs. However, the high-resolution ESR spectrum in Fig. 5b, which does not contain extra resonances beyond those expected from a pure two-donor system, rules out other coupled configurations are deliberately initialized by nuclear magnetic resonance. All ESR peaks match the simulation by choosing the parameters in the ESR spectrum as a function of exchange coupling $J$, using the system Hamiltonian (Eq. (1)) with parameters matching the experimental results.
**Fig. 5 Conditional and unconditional coherent control of the target qubit in the presence of an exchange-coupled control qubit.** a Simulated evolution of the ESR spectrum with parameters matching the experimental results. b Measured ESR spectrum of the target electron in the $(1c,1t)$ charge region. The control electron is kept in the $|\downarrow\rangle$ state, while the four nuclear spin configurations are deliberately initialized by nuclear magnetic resonance. All ESR peaks match the simulation by choosing the parameters $J = 32.06 \pm 0.06$ MHz, $A_t = 97.75 \pm 0.07$ MHz, $A_c = 87.57 \pm 0.16$ MHz, with maximum error $\Delta \varepsilon = 47.4$ kHz. c–f Target qubit Rabi oscillations measured on each of the resonances $\ell_1$ (c), $\ell_2$ (d), $\ell_4$ (e), and $\ell_5$ (f). A $\pi$-pulse on $\ell_1$ or $\ell_4$ transitions constitutes a CROT two-qubit logic gate (Fig. 2b). The same microwave source output power (8 dBm) has been used to drive all Rabi oscillations. The frequency $\Omega$ of the observed Rabi oscillations exhibits variations of up to a factor 4 between resonances, possibly due to a non-monotonic frequency response of the microwave transmission line. The visibility of the Rabi oscillations is systematically lower in the conditional resonances $\ell_1$ and $\ell_4$, as compared to the unconditional ones $\ell_2$ and $\ell_5$.
For the “trivial” resonances, where the nuclear spins are either $|\uparrow_c\downarrow_t\rangle$ (E1, red line) or $|\uparrow_t\downarrow_c\rangle$ (E5, pink line), the Rabi oscillations have a visibility $V_{\text{Rabi}} = P_{\uparrow\downarrow} - P_{\uparrow\uparrow} = 0.75$. In contrast, the nontrivial, conditional resonances $\ell_1$ and $\ell_4$ have a significantly lower visibility $V_{\text{Rabi}} = 0.5$. We considered whether this could be explained by the fact that $\ell_1$ and $\ell_4$ represent transitions to the $|\uparrow\downarrow\rangle$ state rather than $|\uparrow\uparrow\rangle$. Given the measured $J = 32.06$ MHz and $A = 92.66$ MHz, the final state for resonances $\ell_1$ and $\ell_4$ is $|\downarrow_c\uparrow_t\rangle = 0.986|\uparrow_c\downarrow_t\rangle + 0.166|\uparrow_t\downarrow_c\rangle$. This would account for only a 2.7% loss in visibility when measuring the transition through the target qubit.
Another possible contribution to the loss of Rabi visibility can arise because, in a coupled qubit system, measuring one qubit can affect the state of both. Here, the single-shot measurement of the target electron can result in the $|\downarrow\downarrow\rangle$ state being projected to $|\downarrow_c\downarrow_t\rangle$ or $|\uparrow_t\downarrow_c\rangle$. If the system is projected to $|\downarrow_c\downarrow_t\rangle$ and the control electron is not reinitialized in $|\downarrow_c\downarrow_t\rangle$ for the next single-shot measurement, the ESR resonances $\ell_1$ or $\ell_4$ become inactive. Resetting the control electron to the $|\downarrow_c\downarrow_t\rangle$ state requires waiting a relaxation time $T_1$, during which no excitation of the target spin would be achieved on $\ell_1$ or $\ell_4$. In this device, we measured $T_1 = 3.4 \pm 1.3$ ms on the target electron spin (Supplementary Note 3). Therefore, even though the chance of projection to $|\downarrow_c\downarrow_t\rangle$ is low (2.7%), this effect could propagate over several measurement records. This hypothesis can be verified by inspecting the single-shot readout traces (Supplementary Note 4). After a $\pi$-pulse on $\ell_1$ or $\ell_4$, we observe instances where a few successive readout traces show a $|\downarrow\uparrow\rangle$ outcome. However, such instances of missing target excitation do not last for more than $\approx 20$ ms—two orders of magnitude less than the measured $T_1$ of the target
**Resonant CROT gate.** Coherent control of one of the two-electron spins is demonstrated in Fig. 5c–f. ESR control of the electron spin is performed in the $(1c,1t)$ region, with the control electron in the $|\downarrow\rangle$ state. We observe Rabi oscillations for all four nuclear spin configurations. Electron spin rotations driven on $\ell_1$ ($|\uparrow_c\downarrow_t\downarrow_t\rangle \leftrightarrow |\uparrow_c\downarrow_t\uparrow_t\rangle$, yellow line) and $\ell_4$ ($|\downarrow_c\uparrow_t\downarrow_t\rangle \leftrightarrow |\downarrow_c\uparrow_t\uparrow_t\rangle$, green line) are conditional upon the control electron being in the $|\downarrow\rangle$ state. Therefore, a $\pi$-pulse on one of these ESR resonances constitutes a CROT two-qubit gate.
The high-fluence devices contain many donors, as visible in the charge stability diagram (Fig. 3b). One may thus expect to encounter more complex clusters of interacting donors, instead of isolated pairs. However, the high-resolution ESR spectrum in Fig. 5b, which does not contain extra resonances beyond those expected from a pure two-donor system, rules out other coupled donors. This is probably because the vast majority of other donors, presumably located behind (with respect to the SET) the pair being measured, are in the ionized charge state. To completely eliminate concerns around spurious donors, we will in the future adopt a counted single-ion implantation method, which allows to introduce individual donors with a confidence close to 99.9% (ref. 44).
The high-fluence devices contain many donors, as visible in the charge stability diagram (Fig. 3b). One may thus expect to encounter more complex clusters of interacting donors, instead of isolated pairs. However, the high-resolution ESR spectrum in Fig. 5b, which does not contain extra resonances beyond those expected from a pure two-donor system, rules out other coupled donors. This is probably because the vast majority of other donors, presumably located behind (with respect to the SET) the pair being measured, are in the ionized charge state. To completely eliminate concerns around spurious donors, we will in the future adopt a counted single-ion implantation method, which allows to introduce individual donors with a confidence close to 99.9% (ref. 44).
The high-fluence devices contain many donors, as visible in the charge stability diagram (Fig. 3b). One may thus expect to encounter more complex clusters of interacting donors, instead of isolated pairs. However, the high-resolution ESR spectrum in Fig. 5b, which does not contain extra resonances beyond those expected from a pure two-donor system, rules out other coupled donors. This is probably because the vast majority of other donors, presumably located behind (with respect to the SET) the pair being measured, are in the ionized charge state. To completely eliminate concerns around spurious donors, we will in the future adopt a counted single-ion implantation method, which allows to introduce individual donors with a confidence close to 99.9% (ref. 44).
The high-fluence devices contain many donors, as visible in the charge stability diagram (Fig. 3b). One may thus expect to encounter more complex clusters of interacting donors, instead of isolated pairs. However, the high-resolution ESR spectrum in Fig. 5b, which does not contain extra resonances beyond those expected from a pure two-donor system, rules out other coupled donors. This is probably because the vast majority of other donors, presumably located behind (with respect to the SET) the pair being measured, are in the ionized charge state. To completely eliminate concerns around spurious donors, we will in the future adopt a counted single-ion implantation method, which allows to introduce individual donors with a confidence close to 99.9% (ref. 44).
The high-fluence devices contain many donors, as visible in the charge stability diagram (Fig. 3b). One may thus expect to encounter more complex clusters of interacting donors, instead of isolated pairs. However, the high-resolution ESR spectrum in Fig. 5b, which does not contain extra resonances beyond those expected from a pure two-donor system, rules out other coupled donors. This is probably because the vast majority of other donors, presumably located behind (with respect to the SET) the pair being measured, are in the ionized charge state. To completely eliminate concerns around spurious donors, we will in the future adopt a counted single-ion implantation method, which allows to introduce individual donors with a confidence close to 99.9% (ref. 44).
The high-fluence devices contain many donors, as visible in the charge stability diagram (Fig. 3b). One may thus expect to encounter more complex clusters of interacting donors, instead of isolated pairs. However, the high-resolution ESR spectrum in Fig. 5b, which does not contain extra resonances beyond those expected from a pure two-donor system, rules out other coupled donors. This is probably because the vast majority of other donors, presumably located behind (with respect to the SET) the pair being measured, are in the ionized charge state. To completely eliminate concerns around spurious donors, we will in the future adopt a counted single-ion implantation method, which allows to introduce individual donors with a confidence close to 99.9% (ref. 44).
The high-fluence devices contain many donors, as visible in the charge stability diagram (Fig. 3b). One may thus expect to encounter more complex clusters of interacting donors, instead of isolated pairs. However, the high-resolution ESR spectrum in Fig. 5b, which does not contain extra resonances beyond those expected from a pure two-donor system, rules out other coupled donors. This is probably because the vast majority of other donors, presumably located behind (with respect to the SET) the pair being measured, are in the ionized charge state. To completely eliminate concerns around spurious donors, we will in the future adopt a counted single-ion implantation method, which allows to introduce individual donors with a confidence close to 99.9% (ref. 44).
electron spin. Therefore, also this explanation appears improbable. Overall, we conclude that even performing a simple Rabi oscillation on a conditional resonance in exchange-coupled donors unveils unexpected details that warrant further investigation.
The presence within the device of strong electric fields, which can affect the value of $J$ and thereby the frequency of the conditional resonances, appears not to introduce spurious spin dephasing. We have measured the dephasing time of the target electron $T_2^\ast$ on both $\Xi_5$ (unaffected by $J$) and $\Xi_1$ (dependent on $J$), and found similar values $T_2^\ast \approx 9 \mu s$, within the error margins (Supplementary Note 3).
The complete benchmarking of a two-qubit logic gate requires the coherent control and individual readout of both qubits, and the operation of the target qubit conditional on an arbitrary state of the control qubit. The present device, comprising a very thin ESR antenna, was damaged by an electrostatic discharge before we could complete the benchmarking of the full two-qubit logic gate. Future devices will be equipped with thicker antenna to prevent this issue.
For the readout, it is often but not always possible to read two (or more) spins sequentially using the same charge sensor. This depends simply on whether all donors electrons have a tunnel time to the reservoir that falls within a usable range (typically 10-100 ms). We are currently developing new device designs, inspired by the flip-flop qubit proposal34 that afford a greater degree of control of all tunnel couplings. Even if only one donor (e.g., the target) happens to be readable, the control donor spin states can be read out via a quantum non-demolition (QND) method by using the target electron as ancilla qubit, as already demonstrated in exchange-coupled double quantum dot systems17,46,47. This process requires a long relaxation time $T_1$ of the electron spins in presence of weak exchange coupling. The target electron $T_1 = 3.4 \pm 1.3 \mu s$ measured here is close to that of single, uncoupled donor electrons spins43, and indicates that an ancilla-based QND readout will be an available option for future experiments.
Methods
Sample fabrication. Silicon MOS processes are employed for the donor spin qubit device fabrication. A silicon wafer is overgrown with a 0.9 μm thick eplayer of the isotopically purified $^{28}$Si with $^{28}$Si residual concentration of 730 p.p.m. (ref. 39). Heavily doped n+ regions for Ohmic contacts and lightly doped p regions for leakage prevention are defined by phosphorus and boron thermal diffusion. A field oxide (200 nm thick SiO2) is grown using a wet thermal oxidation process. The central active area is covered with a high-temperature oxide (8 nm thick SiO2) grown in dry conditions. Subsequently, an aperture of 90 mm × 100 mm is defined in a PMMA mask using electron-beam lithography (EBL). Through this aperture, the samples are implanted with atomic (P) or molecular (P2) phosphorus ions at an acceleration voltage of 10 keV per ion. During implantation, the samples were tilted by 7° to minimize the possibility of channeling implantation. The final P atom position in the device is determined using full cascade Monte Carlo SRIM simulations. The projected range of the implant is 10 nm beyond the SiO2/Si interface. The size of the PMMA aperture is taken into account when determining the P–P donor spacing. Post implantation, a rapid thermal anneal (3 s at 1000 °C) is performed for donor activation and implantation damage repair. A nanoelectronic device is defined around the implantation region through two EBL steps, each followed by thermal deposition of aluminum (20 nm thickness for layer 1; 40 nm for layer 2). Between each aluminum layer, the Al2O3 is formed by immediate, post-deposition sample exposure to a pure, low pressure (100 mTorr) oxygen atmosphere. The final step is a forming gas anneal (400 °C, 15 min, 95% N2/5% H2) aimed at passivating the interface traps.
Experimental setup. The device was placed in a copper enclosure and wire-bonded to a gold-plated printed circuit board using thin aluminum wires. The sample was mounted in a Bluefors LD400 cryogen-free dilution refrigerator with base temperature of 14 mK, and placed in the center of the magnetic field produced by the superconducting solenoid in persistent mode ($\approx 1.4$ T). The magnetic field was oriented perpendicular to the short-circuit termination of the on-chip microwave antenna and parallel to the sample surface.
DC bias voltages, sourced from Stanford Instruments SIM928 isolated voltage sources, were delivered to the SET top gate, the barrier gates and the DC donor gates through 20 Hz low-pass filters. Microwave pulses for ESR were generated by an Agilent E8257D 50 GHz analog source; RF pulses for NMR were produced by a Agilent N5182B 6 GHz vector source. RF and microwave signals to be delivered to the microwave antenna were combined at room temperature and delivered through a semi-rigid coaxial cable fitted with a 10 dB attenuator mounted at the 4 K plate and a 3 dB attenuator at the 14 mK stage. The SET current was measured by a Femto DLPCA-200 transimpedance amplifier at room temperature ($10^7$ V/A gain, 50 kHz bandwidth), followed by a Stanford Instruments SIM910 IFET post-amplifier ($10^6$ V/V gain). Stanford Instruments SIM965 analog filter (50 kHz cutoff, low-pass Bessel filter), and acquired via an AlazarTech ATS9440 PCI digitizer card. The instruments were synchronized by a SpinCore Pulseblaster-ESR TTL generator.
Data availability
The experimental and simulation data that support the findings of this study are available in Figshare with the identifier https://doi.org/10.6084/m9. figshare.13291913.
Received: 28 July 2020; Accepted: 2 December 2020; Published online: 08 January 2021
References
1. Montanaro, A. Quantum algorithms: an overview. npj Quantum Inf. 2, 1–8 (2016).
2. Ladd, T. D. et al. Quantum computers. Nature 464, 45–53 (2010).
3. Veldhorst, M., Enkink, H., Yang, C. & Dzurak, A. Silicon CMOS architecture for a spin-based quantum computer. Nat. Commun. 8, 1–8 (2017).
4. Pillarsetty, R. et al. Qubit device integration using advanced semiconductor manufacturing process technology. In 2018 IEEE International Electron Devices Meeting (IEDM). 6.3 (IEEE, San Francisco, CA, 2018).
5. Vinet, M. et al. Towards scalable silicon quantum computing. In 2018 IEEE International Electron Devices Meeting (IEDM). 6.5 (IEEE, San Francisco, CA, 2018).
6. Govoreanu, B. et al. Moving spins from lab to fab: A silicon-based platform for quantum computing device technologies. In 2019 Silicon Nanoelectronics Workshop (SNW). 1-2 (IEEE, Kyoto, Japan, 2019).
7. Jamieson, D. N. et al. Controlled shallow single-ion implantation in silicon using an active substrate for sub-20-kev ions. Appl. Phys. Lett. 86, 202101 (2005).
Discussion
We have presented the experimental observation of weak exchange coupling between the electron spins of a pair of 31P donors implanted in $^{28}$Si. The exchange interaction $J = 32.06 \pm 0.06$ MHz was determined by ESR spectroscopy, and falls within the range $J < A$ where a native CROT two-qubit logic gate can be performed by applying a rf-pulse to the target electron after setting the two donor nuclear spins in opposite states. These results represent the first demonstration of hyperfine-controlled CROT gate for donor electrons33, a scheme that is intrinsically robust to uncertainties in the donor location, since it only requires $J$ to be smaller than $A = 100$ MHz, and larger than the inhomogeneous ESR linewidth $\approx 10$ kHz.
The present work already unveiled peculiar effects, such as the Stark shift of hyperfine coupling and $g$-factors in the presence of an exchange-coupled electron, and unexplained features in the visibility of conditional qubit rotations. These effects call for detailed theoretical models of donor exchange under strain and electric fields, significantly expanding the existing theories of exchange in bulk silicon33.
Future experiments will focus on benchmarking the fidelity of a complete one- and two-qubit gate set, and studying the noise channels affecting the operations. The suitability of this exchange-based logic gate for large-scale quantum computing will be assessed by integrating deterministic, counted single-ion implantation within the fabrication process8,44, and studying the device yield and gate performance while subjected to realistic fabrication tolerances. The fine spectral resolution afforded by our resonant control methods will provide precious insights and experimental validation to a wide suite of theoretical models of donor physics and quantum device fabrication.
8. Van Donkelaar, J. et al. Single atom devices by ion implantation. J. Phys. Condens. Matter 27, 154204 (2015).
9. Singh, M. et al. Electrostatically defined silicon quantum dots with counted antimony donor implants. Appl. Phys. Lett. 108, 062101 (2016).
10. Kane, B. E. A silicon-based nuclear spin quantum computer. Nature 393, 133–137 (1998).
11. Dehollain, J. P. et al. Optimization of a solid-state electron spin qubit using gate set tomography. N. J. Phys. 18, 103018 (2016).
12. Muhonen, J. T. et al. Storing quantum information for 30 seconds in a nanoelectronic device. Nat. Nanotechnol. 9, 986 (2014).
13. Kalra, R., Laucht, A., Hill, C. D. & Morello, A. Robust two-qubit gates for donors in silicon controlled by hyperfine interactions. Phys. Rev. X 4, 021044 (2014).
14. Loss, D. & DiVincenzo, D. P. Quantum computation with quantum dots. Phys. Rev. A 57, 120 (1998).
15. Petta, J. R. et al. Coherent manipulation of coupled electron spins in semiconductor quantum dots. Science 309, 2180–2184 (2005).
16. Veldhorst, M. et al. A two-qubit logic gate in silicon. Nature 526, 410 (2015).
17. Watson, T. et al. A programmable two-qubit quantum processor in silicon. Nature 555, 633–637 (2018).
18. Cory, D. G., Price, M. D. & Havel, T. F. Nuclear magnetic resonance spectroscopy: an experimentally accessible paradigm for quantum computing. Phys. D Nonlinear Phenom. 120, 82–102 (1998).
19. Plantenberg, J., De Groot, P., Harmans, C. & Mooij, J. Demonstration of controlled-NOT quantum gates on a pair of superconducting quantum bits. Nature 447, 836 (2007).
20. Huang, W. et al. Fidelity benchmarks for two-qubit gates in silicon. Nature 569, 532 (2019).
21. Zajac, D. M. et al. Resonantly driven CNOT gate for electron spins. Science 359, 439–442 (2018).
22. Kohl, W. & Luttinger, J. Theory of donor states in silicon. Phys. Rev. 98, 915 (1955).
23. Koiller, B., Hu, X. & Sarma, S. D. Exchange in silicon-based quantum computer architecture. Phys. Rev. Lett. 88, 027903 (2001).
24. Dehollain, J. P. et al. Single-shot readout and relaxation of singlet and triplet states in exchange-coupled p 31 electron spins in silicon. Phys. Rev. Lett. 112, 236801 (2014).
25. González-Zalba, M. F. et al. An exchange-coupled donor molecule in silicon. Nano Lett. 14, 5672–5676 (2014).
26. Broome, M. et al. Two-electron spin correlations in precision placed donors in silicon. Nat. Commun. 9, 980 (2018).
27. He, Y. et al. A two-qubit gate between phosphorus donor electrons in silicon. Nature 571, 371–375 (2019).
28. Economou, S. E. & Barnes, E. Analytical approach to swift nonleaky entangling gates in superconducting qubits. Phys. Rev. B 91, 161405 (2015).
29. Russ, M. et al. High-fidelity quantum gates in si/si double quantum dots. Phys. Rev. B 97, 085421 (2018).
30. van der Wiel, W. G. et al. Electron transport through double quantum dots. Rev. Mod. Phys. 75, 1 (2002).
31. Yang, S., Wang, X. & Sarma, S. D. Generic Hubbard model description of semiconductor quantum-dot spin qubits. Phys. Rev. B 83, 161301 (2011).
32. Morello, A. et al. Architecture for high-sensitivity single-shot readout and control of the electron spin of individual donors in silicon. Phys. Rev. B 80, 081307 (2009).
33. Morello, A. et al. Single-shot readout of an electron spin in silicon. Nature 467, 687 (2010).
34. Pla, J. et al. Single atom electron spin qubit in silicon. Nature 489, 541 (2012).
35. Pla, J. J. et al. High-fidelity readout and control of a nuclear spin qubit in silicon. Nature 496, 334 (2013).
36. Dehollain, J. et al. Nanoscale broadband transmission lines for spin qubit control. NanoTechnology 24, 015202 (2012).
37. Laucht, A. et al. High-fidelity adiabatic inversion of a 31P electron spin qubit in natural silicon. Appl. Phys. Lett. 104, 092115 (2014).
38. Bradbury, F. R. et al. Stark tuning of donor electron spins in silicon. Phys. Rev. Lett. 97, 176404 (2006).
39. Laucht, A. et al. Electrically controlling single-spin qubits in a continuous microwave field. Sci. Adv. 1, e1500022 (2015).
40. Asaad, S. et al. Coherent electrical control of a single high-spin nucleus in silicon. Nature 579, 205–209 (2020).
41. Dreher, L. et al. Electrostatic hyperfine tuning of phosphorus donors in silicon. Phys. Rev. Lett. 106, 037601 (2011).
42. Mansir, J. et al. Linear hyperfine tuning of donor spins in silicon using hydrostatic strain. Phys. Rev. Lett. 120, 167701 (2018).
43. Tenberg, S. B. et al. Electron spin relaxation of single phosphorus donors in metal-oxide-semiconductor nanoscale devices. Phys. Rev. B 99, 205306 (2019).
44. Jakob, A. M. et al. Deterministic single ion implantation with 99.8% confidence for scalable donor-qubit arrays in silicon. Preprint at https://arxiv.org/abs/2009.02892 (2020).
45. Tosi, G. et al. Silicon quantum processor with robust long-distance qubit couplings. Nat. Commun. 8, 1–11 (2017).
46. Nakajima, T. et al. Quantum non-demolition measurement of an electron spin qubit. Nat. Nanotechnol. 14, 555–560 (2019).
47. Xue, X. et al. Repeative quantum nondemolition measurement and soft decoding of a silicon spin qubit. Phys. Rev. X 10, 021006 (2020).
48. Itoh, K. M. & Watanabe, H. Isotope engineering of silicon and diamond for quantum computing and sensing applications. MRS Commun. 4, 143–157 (2014).
Author/s:
Madzik, MT; Laucht, A; Hudson, FE; Jakob, AM; Johnson, BC; Jamieson, DN; Itoh, KM; Dzurak, AS; Morello, A
Title:
Conditional quantum operation of two exchange-coupled single-donor spin qubits in a MOS-compatible silicon device
Date:
2021-01-08
Citation:
Madzik, M. T., Laucht, A., Hudson, F. E., Jakob, A. M., Johnson, B. C., Jamieson, D. N., Itoh, K. M., Dzurak, A. S. & Morello, A. (2021). Conditional quantum operation of two exchange-coupled single-donor spin qubits in a MOS-compatible silicon device. NATURE COMMUNICATIONS, 12 (1), https://doi.org/10.1038/s41467-020-20424-5.
Persistent Link:
http://hdl.handle.net/11343/272609
File Description:
Published version
License:
CC BY | 2025-03-05T00:00:00 | olmocr | {
"pdf_page_numbers": [
[
0,
1493,
1
],
[
1493,
4855,
2
],
[
4855,
10136,
3
],
[
10136,
15142,
4
],
[
15142,
19509,
5
],
[
19509,
30453,
6
],
[
30453,
39204,
7
],
[
39204,
44464,
8
],
[
44464,
45178,
9
]
]
} |
3f1f67aa02139216cd06c7813b0aa0e6edc93498 | {
"olmocr-version": "0.1.59",
"pdf-total-pages": 18,
"total-fallback-pages": 0,
"total-input-tokens": 80363,
"total-output-tokens": 18833,
"fieldofstudy": [
"Mathematics"
]
} | COMBINATORIAL CONSTRUCTIONS OF
THREE-DIMENSIONAL SMALL COVERS
YASUZO NISHIMURA
Abstract. In this paper we study two operations on 3-dimensional small
covers called a connected sum and a surgery. These operations correspond
to combinatorial operations on \((\mathbb{Z}_2)^3\)-colored simple convex polytopes. Then we
show that each 3-dimensional small cover can be constructed from \(T^3\), \(\mathbb{R}P^3\)
and \(S^1 \times \mathbb{R}P^2\) with two different \((\mathbb{Z}_2)^3\)-actions by using these operations. This
result is a generalization or an improvement of results in [3], [5], [8] and [12].
1. Introduction
A small cover was introduced by Davis and Januszkiewicz [1] as an
\(n\)-dimensional closed manifold \(M^n\) with a locally standard \((\mathbb{Z}_2)^n\)-action such that its orbit space
is a simple convex polytope \(P\) where \(\mathbb{Z}_2\) is the quotient additive group \(\mathbb{Z}/2\mathbb{Z}\). They
showed that there exists a one-to-one correspondence between small covers and
\((\mathbb{Z}_2)^n\)-colored polytopes (cf. [1, Proposition 1.8]). Here a pair \((P, \lambda)\)
is called a \((\mathbb{Z}_2)^n\)-colored polytope when \(P\) is an \(n\)-dimensional simple convex polytope with
the set of facets \(\mathcal{F}\) and a function \(\lambda: \mathcal{F} \to (\mathbb{Z}_2)^n\) satisfying the following condition:
\((\\ast)\) if \(F_1 \cap \cdots \cap F_n \neq \emptyset\) then \(\{\lambda(F_1), \cdots, \lambda(F_n)\}\) is linearly independent.
We say that two \((\mathbb{Z}_2)^n\)-colored polytopes \((P_i, \lambda_i) (i = 1, 2)\) are equivalent when
there exists a combinatorial equivalence of polytopes \(\phi: P_1 \to P_2\) such that \(\lambda_2 \phi = \theta \lambda_1\)
for some \(\theta \in \text{Aut}(\mathbb{Z}_2)^n\). The \(n\)-dimensional torus \(T^n\) and the real projective
space \(\mathbb{R}P^n\) with standard \((\mathbb{Z}_2)^n\)-actions are examples of small covers over the \(n\)-cube
\(I^n\) and the \(n\)-simplex \(\Delta^n\) respectively.
In this paper we are interested in constructions of 3-dimensional small covers \(M^3\)
from basic small covers by using some operations. In [3] Izmestiev studied a class
of 3-dimensional small covers which are called linear models and are correspondent
to 3-colored polytopes. He introduced two operations on linear models called a
connected sum \(\natural\) and a surgery \(\circ\) and proved the following theorem (cf. [3, Theorem
3]).
Theorem 1.1 (Izmestiev). Each linear model \(M^3\) can be constructed from \(T^3\) by
using three operations \(\natural, \circ\) and \(\circ^{-1}\) where \(\circ^{-1}\) is the inverse of \(\circ\).
In [12] we generalized Theorem 1.1 to orientable small covers \(M^3\) which are correspond-
to 4-colored polytopes. We introduced a new operation called the Dehn
surgery \(\circ^D\), and showed that each orientable small cover \(M^3\) can be constructed from
\(T^3\) and \(\mathbb{R}P^3\) by using four operations \(\natural, \circ, \circ^{-1}\) and \(\circ^D\) (cf. [12, Theorem 1.10]).
Later Lü and Yu [8] considered a construction of general small covers $M^3$. They introduced new operations $\sharp$, $\sharp^\text{ve}$, $\sharp^\Delta$ and $\sharp_i^\otimes$ ($i \geq 3$) and showed the following theorem (cf. [8] Theorem 1.2).
**Theorem 1.2** (Lü and Yu). Each small cover $M^3$ can be constructed from $\mathbb{R}P^3$ and $S^1 \times \mathbb{R}P^2$ with a certain $(\mathbb{Z}_2)^3$-action by using seven operations $\sharp$, $\sharp^{-1}$, $\sharp^\text{ve}$, $\sharp^\Delta$, $\sharp_i^\otimes$ and $\sharp_i^\otimes$.
Operations appeared in Theorem 1.2 are all “non-decreasing” i.e. they do not decrease the number of faces of an orbit polytope, and therefore the use of the surgery $\sharp$ is prohibited unlike Theorem 1.1. In [5] Kuroki pointed out that the operations $\sharp^D$, $\sharp^e$ and $\sharp^\text{ve}$ can be obtained as compositions of $\sharp$ and $\sharp$ such as
$$
\sharp^D = \sharp \circ \sharp^p \mathbb{R}P^3, \quad \sharp^e = \sharp \circ \sharp^p \mathbb{R}P^3 \quad \text{and} \quad \sharp^\text{ve} = \sharp^2 \circ \sharp, \quad \text{respectively (cf. [5] Theorem 4.1)}.
$$
Therefore our result in [12] can be improved as follows: Each orientable small cover $M^3$ can be constructed from $\mathbb{R}P^3$ and $T^3$ by using three operations $\sharp$, $\sharp$ and $\sharp^{-1}$. (cf. [5] Corollary 4.4). Moreover Lü-Yu’s result can be rewritten by using $\sharp$ instead of $\sharp^e$ and $\sharp^\text{ve}$ as follows (cf. [5] Corollary 4.8): Each small cover $M^3$ can be constructed from $\mathbb{R}P^3$ and $S^1 \times \mathbb{R}P^2$ with a certain $(\mathbb{Z}_2)^3$-action by using six operations $\sharp$, $\sharp$, $\sharp^{-1}$, $\sharp^\Delta$, $\sharp_i^\otimes$ and $\sharp_i^\otimes$. Then a problem arises (cf. [5] Problem 5.2).
**Problem 1.3**. What are basic small covers from which we can construct all 3-dimensional small covers using the three operations $\sharp$, $\sharp$ and $\sharp^{-1}$?
We give a solution to this problem. The following is our main result.
**Theorem 1.4**. Each small cover $M^3$ can be constructed from $T^3$, $\mathbb{R}P^3$ and $S^1 \times \mathbb{R}P^2$ with two different $(\mathbb{Z}_2)^3$-actions by using two operations $\sharp$ and $\sharp$.
In the above theorem we do not use the inverse surgery $\sharp^{-1}$. As a corollary we obtain improvements of Theorem 1.1 and our previous result in [12].
**Corollary 1.5**. (1) Each linear model $M^3$ can be constructed from $T^3$ by using two operations $\sharp$ and $\sharp$.
(2) Each orientable small cover $M^3$ can be constructed from $T^3$ and $\mathbb{R}P^3$ by using two operations $\sharp$ and $\sharp$.
These results are equivariant analogues of a well-known result (cf. [4]): “Each closed 3-manifold can be constructed from the 3-sphere $S^3$ by using the Dehn surgeries”.
This paper is organized as follows. In section 2 we recall the definition and the basic facts about small covers briefly, and we introduce some basic 3-dimensional small covers. In section 3 we establish several operations on $(\mathbb{Z}_2)^3$-colored polytopes. In section 4 we discuss the constructions of $(\mathbb{Z}_2)^3$-colored polytopes, and prove Theorem 1.4. In section 5 we follow the standpoint of Lü and Yu, and discuss a non-decreasing construction of small covers by using the inverse surgery $\sharp^{-1}$ instead of the decreasing surgery $\sharp$. We shall point out that there is a gap in the proof of Theorem 1.2 in [8] (Remark 5.5) and improve their result as follows.
**Theorem 1.6**. (1) Each linear model $M^3$ can be constructed from $T^3$ by using three operations $\sharp$, $\sharp^e$ and $\sharp^{-1}$.
(2) Each orientable small cover $M^3$ can be constructed from $T^3$ and $\mathbb{R}P^3$ by using three operations $\sharp$, $\sharp^e$ and $\sharp^{-1}$.
(3) Each small cover $M^3$ can be constructed from $\mathbb{R}P^3$ and $S^1 \times \mathbb{R}P^2$ with two different $(\mathbb{Z}_2)^3$-actions by using four operations $\sharp$, $\sharp^e$, $\sharp^{-1}$ and $\sharp_i^\otimes$.
In section 6 we shall make a remark on a 2-torus manifold which is an object of a little wider class than small covers. If the object is expanded to this class, the argument becomes easier. We prove the following theorem.
**Theorem 1.7.** (1) Each linear model of a locally standard 2-torus manifold over $D^3$ can be constructed from $S^3$ by using inverse surgery $\natural^{-1}$.
(2) Each orientable locally standard 2-torus manifold over $D^3$ can be constructed from $S^3$ by using two surgeries $\natural^{-1}$, $\natural^D$ and the blow up $\mathbb{R}P^3$.
(3) Each locally standard 2-torus manifold over $D^3$ can be constructed from $S^3$ by using the inverse surgery $\natural^{-1}$ and connecting $\mathbb{R}P^3$, $S^1 \times_{\mathbb{Z}_2} S^2$, $S^1 \times \mathbb{R}P^2$ with certain $(\mathbb{Z}_2)^3$-actions by operations $\natural$ and $\natural^e$.
2. Basics of small covers
In this section we recall the definitions and basic facts on small covers (see [1] for detail). Let $P$ be an $n$-dimensional simple convex polytope with facets (i.e., codimension-one faces) $\mathcal{F} = \{F_1, \cdots, F_m\}$. A small cover $M$ over $P$ is an $n$-dimensional closed manifold with a locally standard $(\mathbb{Z}_2)^n$-action such that its orbit space is $P$. For a facet $F$ of $P$, we define $\lambda(F)$ to be the generator of the isotropy subgroup at $x \in \pi^{-1}(\text{int} F)$ where $\pi : M \to P$ is the orbit projection. Then a function $\lambda : \mathcal{F} \to (\mathbb{Z}_2)^n$ is called a characteristic function of $M$ which satisfies the following condition.
\[
(*) \text{ if } F_1 \cap \cdots \cap F_n \neq \emptyset \text{ then } \{\lambda(F_1), \cdots, \lambda(F_n)\} \text{ is linearly independent.}
\]
Therefore $\lambda$ is a kind of face-coloring of $P$. Then we call a function satisfying $(*)$ a $(\mathbb{Z}_2)^n$-coloring of $P$. We say that two $(\mathbb{Z}_2)^n$-colored polytopes $(P_1, \lambda_1)$ and $(P_2, \lambda_2)$ are equivalent when there exists a combinatorial equivalence of polytopes $\phi : P_1 \to P_2$ such that $\lambda_2 \phi = \theta \lambda_1$ for some $\theta \in \text{Aut}(\mathbb{Z}_2)^n$. Conversely, given a simple convex polytope $P$ and a $(\mathbb{Z}_2)^n$-coloring $\lambda : \mathcal{F} \to (\mathbb{Z}_2)^n$ satisfying $(*)$, we can construct a small cover $M$ such that its characteristic function is the given $\lambda$ as follows:
\[
M(P, \lambda) := P \times (\mathbb{Z}_2)^n / \sim,
\]
where $(x, t) \sim (y, s)$ is defined as $x = y \in P$ and $s - t$ is contained in the subgroup generated by $\lambda(F_1), \cdots, \lambda(F_k)$ such that $x \in \text{int}(F_1 \cap \cdots \cap F_k)$. We say that two small covers $M_i$ over $P_i$ ($i = 1, 2$) are $\text{GL}(n, \mathbb{Z}_2)$-equivalent on a combinatorial equivalence of polytopes $\phi : P_1 \to P_2$ when there exists a $\theta$-equivariant homeomorphism $f : M_1 \to M_2$ such that $\pi_2 \circ f = \phi \circ \pi_1$ i.e. $f(g \cdot x) = \theta(g) \cdot f(x) \ (g \in (\mathbb{Z}_2)^n, \ x \in M_1)$ for some $\theta \in \text{Aut}(\mathbb{Z}_2)^n$. Moreover we say that two small covers are equivalent when they are $\text{GL}(n, \mathbb{Z}_2)$-equivalent on some equivalence $\phi : P_1 \to P_2$. In [7] this equivalence and a $\text{GL}(n, \mathbb{Z}_2)$-equivalence on the identity are called a weakly equivariant homeomorphism and a $D$-$J$ equivalence, respectively. Davis and Januszkiewicz proved that a small cover $M$ over $P$ with a characteristic function $\lambda$ is $\text{GL}(n, \mathbb{Z}_2)$-equivalent on the identity to $M(P, \lambda)$ when we fix a polytope $P$ (cf. [1] Proposition 1.8). Therefore we can identify the equivalence class of a small cover $M(P, \lambda)$ with the equivalence class of a $(\mathbb{Z}_2)^n$-colored polytope $(P, \lambda)$.
**Example 2.1.** The real projective space $\mathbb{R}P^n$ and the $n$-dimensional torus $T^n$ with standard $(\mathbb{Z}_2)^n$-actions are examples of small covers over the $n$-simplex $\Delta^n$ and the $n$-cube $I^n$ respectively. Figure [1] shows their characteristic functions on the polytopes in the case $n = 3$, where $\{\alpha, \beta, \gamma\}$ is a basis of $(\mathbb{Z}_2)^3$. We notice that a
\((\mathbb{Z}_2)^n\)-coloring on \(\Delta^n\) is unique up to equivalence. Therefore we denote the colored simplex by \(\Delta^n\) by omitting coloring.
**Figure 1.** Characteristic functions of \(\mathbb{R}P^3\) and \(T^3\).
A small cover over \(P\) with \(n\)-coloring (i.e. \(\lambda(F)\) is a basis of \((\mathbb{Z}_2)^n\)) is called a linear model. An example of a linear model is the torus \(T^n\) shown in Example\[2.1\]
In this case the \(n\)-coloring of \(P\) (i.e. the linear model) is unique up to equivalence. In case \(n = 3\), it is well-known that a simple convex polytope is 3-colorable if and only if each face contains an even number of edges.
In [11, Theorem 1.7], we gave a criterion of when a small cover is orientable. We recall the criterion in the case \(n = 3\).
**Theorem 2.2.** A 3-dimensional small cover \(M(P, \lambda)\) is orientable if and only if \(\lambda(F)\) is contained in \(\{\alpha, \beta, \gamma, \alpha + \beta + \gamma\}\) for a suitable basis \(\{\alpha, \beta, \gamma\}\) of \((\mathbb{Z}_2)^3\).
From the above theorem the small covers \(\mathbb{R}P^3\) and \(T^3\) given in Figure 1 are both orientable. We call a \((\mathbb{Z}_2)^3\)-coloring satisfying the orientability condition in the above theorem an orientable coloring of \(P\). Since each triple of \(\{\alpha, \beta, \gamma, \alpha + \beta + \gamma\}\) is linearly independent, the orientable coloring is just an ordinary 4-coloring.
**Example 2.3.** We consider small covers on the 3-sided prism \(P^3(3) = I \times \Delta^2\). There exist three types of \((\mathbb{Z}_2)^3\)-coloring on \(P^3(3)\) shown in Figure\[2\] up to equivalence. The first example \(M(P^3(3), \lambda_1)\) is non-equivariantly homeomorphic to \(S^1 \times \mathbb{R}P^2\). The second example \(M(P^3(3), \lambda_2)\) is not equivariantly homeomorphic to \(M(P^3(3), \lambda_1)\) but non-equivariantly homeomorphic to \(S^1 \times \mathbb{R}P^2\) (cf. [8, Lemmas 4.2 and 4.3]). The last example \(M(P^3(3), \lambda_3)\) is orientable and homeomorphic to \(\mathbb{R}P^3 \sharp \mathbb{R}P^3\) where \(\sharp\) is the connected sum (see the following section).
**Figure 2.** Basic three types of \((\mathbb{Z}_2)^3\)-coloring on 3-sided prism \(P^3(3) = I \times \Delta^2\); \(\lambda_1, \lambda_2\) and \(\lambda_3\) respectively.
**Example 2.4.** It is easily verified that there exist four types of \((\mathbb{Z}_2)^3\)-coloring on the 3-cube \(I^3 = P^3(4)\). One of them is the 3-colored cube which is already seen in Figure\[1\] and is denoted by \((I^3, \lambda_0)\). The other three types are shown in Figure\[3\] The associated small covers are homeomorphic to \(S^1 \times K\), a twisted \(K\)-bundle over \(S^1\) and a twisted \(T^2\)-bundle over \(S^1\) according to \(\lambda_1, \lambda_2\) and \(\lambda_3\) respectively, where \(K = \mathbb{R}P^2 \sharp \mathbb{R}P^2\) is the Klein’s bottle (more precisely see [8, Lemmas 5.3 and 5.4]).
COMBINATORIAL CONSTRUCTIONS OF THREE-DIMENSIONAL SMALL COVERS
Figure 3. Three types of \((\mathbb{Z}_2)^3\)-coloring on 3-cube \(I^3\): \(\lambda_1\), \(\lambda_2\) and \(\lambda_3\) respectively (except the 3-colored cube in Figure 1).
Remark 2.5. In [8] the \(GL(3,\mathbb{Z}_2)\)-equivalence on the identity (D-J equivalence) is adopted as an equivalence relation of \((\mathbb{Z}_2)^3\)-colored polytopes, i.e. \((P,\lambda) \sim (P,\theta\lambda)\) for \(\theta \in \text{Aut}(\mathbb{Z}_2)^3\). Therefore it is written that there exist five (resp. seven) types of \((\mathbb{Z}_2)^3\)-coloring on \(P^3(3)\) (resp. \(I^3\)) in [8]. Discussing D-J equivalence classes only when the orbit polytope \(P\) is fixed has the meaning. However, the orbit polytopes will be not fixed in the following sections. Then we adopt our equivalence (the weakly equivariantly homeomorphism) instead of the D-J equivalence. In this paper we shall rewrite results in [8] to our standpoint by our equivalence. The difference between the D-J equivalence and our equivalence is not essential in the discussion of the following sections.
3. OPERATIONS ON SMALL COVERS
Henceforth we assume that \(n = 3\) and \((P,\lambda)\) is a pair of a 3-dimensional simple convex polytope \(P\) with a \((\mathbb{Z}_2)^3\)-coloring \(\lambda\), and \(\{\alpha, \beta, \gamma\}\) is a basis of \((\mathbb{Z}_2)^3\). We call a 3-dimensional simple convex polytope a 3-polytope for simplicity. From the Steinitz’s theorem (see [2] etc.) combinatorially equivalence classes bijectively correspond to 3-connected 3-valent simple planar graphs i.e. 1-skeleton of \(P\). Here a graph \(\Gamma\) is called \(k\)-connected, \(l\)-valent and simple if \(\Gamma\) is connected after cutting any \((k - 1)\) edges, the degree of each vertex is \(l\), and there is no loop and no multi-edge, respectively. In this section we recall some operations on \((\mathbb{Z}_2)^3\)-colored polytopes (or small covers), which were introduced in [3], [8] and [12].
Definition 3.1 (the connected sum \(\#\)). The operation \(\#\) in Figure 4 (from left to right) is called the connected sum (at vertices) and its inverse (from right to left) is denoted by \(\#^{-1}\). These operations also can be defined for non-colored polytopes. Remark that \(P_1\# P_2\) is also a 3-polytope for any 3-polytopes \(P_i\) \((i = 1, 2)\) from the Steinitz’s theorem. The operation \(\#\) corresponds to the connected sum \(M(P_1, \lambda_1)\sharp M(P_2, \lambda_2)\) around fixed points of them (cf. [11] 1.11 or [8] Definition 3). We say that \((P,\lambda)\) is decomposable (as a \((\mathbb{Z}_2)^3\)-colored polytope) when there exist two \((\mathbb{Z}_2)^3\)-colored polytopes \((P_i, \lambda_i)\) \((i = 1, 2)\) such that \((P,\lambda) = (P_1,\lambda_1)\sharp(P_2,\lambda_2)\). Similarly we say that \(P\) is decomposable as a non-colored polytope when \(P = P_1\sharp P_2\) as non-colored polytopes for some \(P_i\) \((i = 1, 2)\).
Specifically the connected sum with \(\Delta^3\) on polytopes, denoted by \(\sharp\Delta^3\) (and often called a cutting vertex or bistellar 0-move), corresponds to the operation called a blow up on small covers (Figure 5). Its inverse \(\#^{-1}\Delta^3\) (often called a bistellar 2-move) is called a blow down.
Definition 3.2 (the surgery \(\natural\)). The operation \(\natural\) in Figure 6 (from left to right) is called the surgery along an edge \(e\) and its inverse \(\natural^{-1}\) (from right to left) is called the inverse surgery along a pair of edges \(e_1\) and \(e_3\). The operations \(\natural\) and \(\natural^{-1}\) both
Figure 4. The connected sum $\#$ and its inverse $\#^{-1}$.
Figure 5. The blow up $\# \Delta^3$ and the blow down $\#^{-1} \Delta^3$.
correspond to the ordinaly surgeries on small covers (cf. [3]). In the previous papers [3], [5], [8] and [12], surgeries $\#$ and $\#^{-1}$ were not distinguished and they both were denoted by the same symbol $\#$.
Figure 6. The surgery $\#$ and its inverse $\#^{-1}$.
We do not allow the surgeries $\#$ and $\#^{-1}$ when the 3-connectedness of the 1-skeleton of $P$ is destroyed after doing it, i.e. the following cases respectively:
**in case $\#$:** if and only if $F_2$ and $F_4$ are adjacent to a same face except $F_1$ and $F_3$ (involve the case when $F_1$ or $F_3$ is a quadrilateral),
**in case $\#^{-1}$:** if and only if $F'_1$ is adjacent to $F'_3$.
**Definition 3.3** (the connected sum along edges $\#^e$). The operation $\#^e$ in Figure 7 (from left to right) is called the connected sum along edges and its inverse is denoted by $(\#^e)^{-1}$. We notice that the operation $\#^e$ is obtained as the composition $\#^e = \# \circ \#$ as shown in the same figure (cf. [5] Theorem 4.1(2)). The operation $\#^e$ corresponds to the connected sum along the circle $\pi^{-1}(e)$ on a small cover $M$ where $\pi : M \to P$ is the projection (cf. [5]).
Specifically the operations $\#^e P^3(3)$ (along a vertical edge in Figure 2) and $\#^e \Delta^3$ are often called the cutting edge and the bistellar 1-move, respectively (Figure 8). The former (left diagram) corresponds to a blow up along the circle $\pi^{-1}(e)$ on a small cover. In this diagram we can choose not only $\beta + \gamma$ but also $\alpha + \beta + \gamma$ as a color of the center square when $* = 0$. The latter operation $\#^e \Delta^3 = \# \circ \# \Delta^3$ corresponds to the Dehn surgery of type $\Delta^3$ on a small cover (cf. [12] or [5] 3.5). This operation is denoted by $\#^D$ and is called the Dehn surgery. This operation can
Figure 7. The connected sum along the edges \( \#^e \) and its inverse \((\#^e)^{-1}\). The figure also shows that \( \#^e = \natural \circ \sharp \).
be done along an edge \( e \) which satisfies the following condition:
\[
\sum_{e} \lambda(F) := \sum_{\{F \in \mathcal{F} | e \cap F \neq \emptyset\}} \lambda(F) = 0.
\]
We call such an edge 0-sum edge (or 4-colored edge in orientable case). We notice that the Dehn surgery \( \natural^D \) does not change the number of faces, and is invertible because \((\natural^D)^{-1} = \natural^D\).
Figure 8. The cutting edge \( \#^e P^3(3) \) and the Dehn surgery \( \natural^D = \#^e \Delta^3 \).
From the Steinitz's theorem, a 3-polytope \( P \) is decomposable as a non-colored polytope if and only if there exist three edges such that they are not adjacent to each other and the 1-skeleton of \( P \) becomes disconnected after cutting them. Obviously if an orientable (4-)colored polytope \( P \) is decomposable as a non-colored polytope then \((P, \lambda)\) is also decomposable (as a \((\mathbb{Z}_2)^3\)-colored polytope). However we need a little attention for non-orientable colored polytopes. We say that \((P, \lambda)\) is quasi-decomposable when there exist two \((\mathbb{Z}_2)^3\)-colored polytopes \((P_1, \lambda_1)\) and \((P_2, \lambda_2)\) such that either \((P, \lambda) = (P_1, \lambda_1)\#^e(P_2, \lambda_1)\) or \((P, \lambda) = (P_1, \lambda_1)\#^e(P_2, \lambda_2)\), except \( P = P_1\#^e \Delta^3(= \#^D P_1)\).
Remark 3.4. Notice that if a 1-skeleton of \( P \) becomes disconnected after cutting three edges \( \{e', e'', e'''\} \) then these three edges are not adjacent to each other or meet at a vertex. In fact if a pair \( \{e', e''\} \) of these three edges is adjacent to each
other and the other edge $e''$ is not adjacent to $e' \cap e''$ then the 1-skeleton of $P$ becomes disconnected after cutting the edge $e''$ and the edge which is adjacent to $e' \cap e''$ and different from $e'$ and $e''$. This contradicts the 3-connectedness of the 1-skeleton of $P$.
**Proposition 3.5.** Let $(P, \lambda)$ be a $(\mathbb{Z}_2)^3$-colored polytope, but not $P^3(3)$. If $P$ is decomposable as a non-colored polytope then $(P, \lambda)$ is quasi-decomposable.
**Proof.** It is sufficient to treat the case that $P$ is indecomposable as a $(\mathbb{Z}_2)^3$-colored polytope. Since $P$ is decomposable as a non-colored polytope, there exist three non-adjacent edges such that $P$ becomes disconnected after cutting them out, and colors of the three faces adjacent to these edges are not linearly independent as shown in Figure 9.
**Figure 9.** The decomposition of a polytope along a 3-cycle of 2-independent faces.
Since $P \neq P^3(3)$, $P$ has at least six faces so we may assume that there are at least two faces under the pillar ($F_i$'s) in the first diagram. We first assume that $F'_3 = F'_2$ (equivalently $e_{21} = e_{31}$ because if it is not so, the 1-skeleton of $P$ becomes disconnected after cutting these two edges). Then the 1-skeleton of $P$ becomes disconnected after cutting three edges $e_1$, $e_{23}$, and $e_{32}$. Since $P$ is indecomposable, these three edges actually meet at a vertex $F_1' \cap F_2' \cap F_3'$ (see Remark 3.4). It should be $F_1' = F_2' = F_3'$ and it is a triangle. This contradicts the assumption that there are at least two faces under the pillar. Therefore the assumption $F'_2 = F'_3$ is denied, and by a similar discussion we can reach the conclusion that $F'_i (i = 1, 2, 3)$ are different faces each other. We notice that if $F_3' \cap F'_2 = \emptyset$ then it is clear that $F_1' \cap F'_2 = F_2' \cap F'_1 = \emptyset$. Therefore we can assume that $F'_3 \cap F'_2 = \emptyset$ by changing the role of $F_i$'s if necessary.
Now we can do the surgery $\natural^{-1}$ for edges $e_1$ and $e_{32}$, and decompose $P$ into two $(\mathbb{Z}_2)^3$-colored polytopes $P_1$ and $P_2$ by cutting three non-adjacent edges $e_{1}'$, $e_2$ and $e_{31}$ (second and third diagrams). Then we have $\natural^{-1}P = P_1 \natural P_2$ or equivalently $P = P_1 \natural P_2$. □
Notice that the surgery $\natural$ and the Dehn surgery $\natural^D$ are not allowed along an edge of a quadrilateral and a triangle, respectively, and the inverse surgery $\natural^{-1}$ is not allowed along a pair of adjacent edges. The following is a key lemma to relate the surgery to the connected sum.
**Lemma 3.6.** Let $(P, \lambda)$ be a $(\mathbb{Z}_2)^3$-colored polytope. Suppose that the 3-connectedness of the 1-skeleton of $P$ is destroyed after doing surgeries $\natural^{-1}$ or $\natural^D$, but not the above trivial prohibited cases. Then $(P, \lambda)$ is quasi-decomposable. In particular when $(P, \lambda)$ is (orientable) 4-colored, $(P, \lambda)$ is decomposable as a $(\mathbb{Z}_2)^3$-colored polytope.
Proof. In consequence of Proposition 3.6 it is sufficient to prove that \((P, \lambda)\) is decomposable as a non-colored polytope.
(1) **in case** \(\tau^{-1}\): When the inverse surgery \(\tau^{-1}\) is not allowed in the right diagram of Figure 6, \(F'_2\) is adjacent to \(F'_3\). Then cutting the three non-adjacent edges \(e_1, e_3\) and \(F'_1 \cap F'_3\) makes the 1-skeleton of \(P\) disconnected. That is \(P\) is decomposable as a non-colored polytope.
(2) **in case** \(\tau^D\): Since \(\tau^D = (\tau^{-1}\Delta^3) \circ \tau^{-1}\) and there is no obstacle for the blow down \(\tau^{-1}\Delta^3\), the allowance of \(\tau^D\) depends only on that of \(\tau^{-1}\).
\(\square\)
4. Constructions of Small Covers
In this section we discuss constructions of \((\mathbb{Z}_2)^3\)-colored polytopes (i.e. small covers) by using two operations \(\sharp\) and \(\star\). Henceforth polytopes are considered as \((\mathbb{Z}_2)^3\)-colored polytopes. In [3], Izmestiev proved the following theorem which is a combinatorial translation of Theorem 1.1.
**Theorem 4.1** (Izmestiev). Each 3-colored polytope \((P^3, \lambda)\) can be constructed from \((F^3, \lambda_0)\) by using three operations \(\sharp, \star\) and \(\tau^{-1}\).
We start from linear models and consider constructions of orientable small covers (i.e. 4-colored polytopes). Let \(P\) be an \(l\)-gonal face of \(P\). We say that \(F\) is \(j\)-independent \((j = 2, 3)\) when the rank of \(\{\lambda(F_1), \ldots, \lambda(F_1)\}\) is \(j\) where \(F_1, \ldots, F_l\) are faces adjacent to \(F\). In the case of orientable small covers, a \(j\)-independent face is a face such that the number of colors of adjacent faces is \(j\) \((j = 2, 3)\). Similarly we say that an edge is \(j\)-colored \((j = 3\) or \(4)\) when the number of the four faces adjacent to the edge is \(j\).
**Proposition 4.2.** Each 4-colored polytope \((P^3, \lambda)\) can be constructed from 3-colored polytopes and \(\Delta^3\) by using two operations \(\sharp\) and \(\tau^D\).
**Proof.** By induction on the number of faces of \(P\), it is sufficient to prove the following
\((*)\) Each 4-colored polytope \(P \neq \Delta^3\) can be decomposed into two polytopes after doing the Dehn surgery \(\tau^D \circ (\tau^D)^{-1}\) finitely many times.
Assume that \(P\) is 4-colored and not \(\Delta^3\). Then there exists a 3-independent face. Let \(F\) be a 3-independent face such that the number of its edges is minimum among 3-independent faces of \(P\), and \(k\) be this number. We prove the above \((*)\) by induction on \(k\). If \(k = 3\) (i.e., \(F\) is a triangle) then we get a colored decomposition \(P = P^0\sharp \Delta^3\) immediately. We assume \(k \geq 4\). Since \(F\) is a 3-independent face, there exists a 4-colored edge \(e\) of \(F\) (see Figure 10).
We notice that there exist no triangular face of \(P\) because \(k \geq 4\). If the Dehn surgery \(\tau^D\) is not allowed along an edge then \(P\) decomposes into two polytopes from Lemma 3.6. Therefore we may assume that the Dehn surgery \(\tau^D\) is allowed along every 4-colored edge of \(F\). If the 3-independence of \(F\) is preserved under the Dehn surgery \(\tau^D\) along some edge, then we can reduce \(P\) to \(\tau^D P\) which has a \((k - 1)\)-gonal 3-independent face, and the proof ends by induction on \(k\). Therefore it is sufficient to show the existence of such an edge.
In Figure 10 we assume that \(F\) becomes 2-independent after doing \(\tau^D\) along the edge \(e\). Then an adjacent face of \(F\) which is painted as \(\beta\) must be unique, and the other faces are painted by \(\alpha\) and \(\gamma\) alternatively such as \(* = \gamma, \ldots, * = \alpha\). In particular when \(k = 4\) (or even), the contradiction arises because \(* = *\).
Figure 10. A 4-colored edge $e$ of a 3-independent face $F$.
$k \geq 5$ and this situation arises, we can do the Dehn surgery $\natural D$ along the edge $e'$ (or $e''$) preserving the 3-independence of $F$.
**Remark 4.3.** In the proof of Proposition 4.2 when we ignore the coloring of $P$, the Dehn surgery $\natural D$ can be continued until a triangle appears for all faces, and then leads to a well-known fact that “Each 3-polytope is bistellarly equivalent to each other” or equivalently “the PL-homeomorphism class of $S^2$ is unique” (cf. [10]).
Combining the above proposition and Theorem 4.1 and noting the relation $\natural D = \natural \circ (\sharp \Delta^3)$, we have the following corollary immediately (cf. [12, Theorem 1.10] and [5, Corollary 4.4]).
**Corollary 4.4.** Each 4-colored polytope $(P^3, \lambda)$ can be constructed from $(I^3, \lambda_0)$ and $\Delta^3$ by using three operations $\sharp$, $\natural$ and $\natural^{-1}$.
Next we consider a construction of all $(\mathbb{Z}_2)^3$-colored polytope. We recall the basic fact that each 3-polytope has a face which has edges less than six (cf. [2] etc.). Such a face is called a small face. If each small face can be compressed so that the number of faces of $P$ decreases then we can reduce all $(\mathbb{Z}_2)^3$-colored polytopes to some basic polytopes by induction on the number of faces. At first we compress 3-independent small faces.
**Proposition 4.5.** Let $P$ be a $(\mathbb{Z}_2)^3$-colored polytope except $\Delta^3$ and $P^3(3)$ as a non-colored polytope. If there exists a 3-independent small face of $P$, then either $P$ or $\natural D P$ is quasi-decomposable.
**Proof.** If there exists a triangular face of $P$ except $\Delta^3$ and $P^3(3)$ then $P$ is decomposable as a non-colored polytope and so $(P, \lambda)$ is quasi-decomposable from Proposition 3.5. Therefore we can assume that $P$ has no triangular face. Let $F$ be a 3-independent small face of $P$.
(1) When $F$ is a quadrilateral, the situation around $F$ is shown as left of Figure 11 where $a_i, b_j \in \mathbb{Z}_2$ with $b_2a_3 = 0$ and at least one of $a_1$ and $b_1$ is nonzero. By a symmetry we may assume that $a_1 = 1$. Since an adjacent triangle does not exist, and we can always blow down $(\sharp^e)^{-1} P^3(3)$ for $F$ along the horizontal edges (if $a_3b_1 = 0$) or the vertical edges (if $b_1 = 1, b_2 = 0$), as shown in Figure 8. That is $P = P^e \sharp^e P^3(3)$.
(2) When $F$ is a pentagon, the situation around $F$ is shown as right of Figure 11 where $a_i, b_j, c_k \in \mathbb{Z}_2$ with $a_2b_3 + b_2 + b_1c_3 + b_3 = 1$ and at least one of $a_1$, $b_1$ and $c_1$ is nonzero. We prove that there exists a 0-sum edge $e$ of $F$ such that $F$ is transformed by $\natural D P$ into a 3-independent quadrilateral. Then $\natural D P$ is quasi-decomposable from the case (1). Here if the Dehn surgery $\natural D$ is not allowed then $P$ is quasi-decomposable from Lemma 3.6.
i) The case $a_1 = 1$ (the case $c_1 = 1$ can be treated similarly).
a) When \(a_2 = 1\), \(e_2\) is a 0-sum edge. If \(c_1 = 0\) or \(b_1 + b_2 = 1\) then the Dehn surgery \(\hat{z}^D\) along the edge \(e_2\) preserves the 3-independence of \(F\) because the rank of \(\{\lambda(F_1), \lambda(F_3), \lambda(F_4)\}\) is three. If \(c_1 = 1\) and \(b_1 = b_2 = 0\) then we have \(b_3 = 1\) and \(e_3\) is a 0-sum edge and \(\{\lambda(F_1), \lambda(F_2), \lambda(F_3)\}\) is linearly independent. If \(c_1 = 1\) and \(b_1 = b_2 = 1\) then we have \(c_3 = 1\) and \(e_1\) is a 0-sum edge and \(\{\lambda(F_2), \lambda(F_3), \lambda(F_4)\}\) is linearly independent. In all cases the Dehn surgery \(\hat{z}^D\) along a certain 0-sum edge preserves the 3-independence of \(F\).
b) If \(a_2 = 0\) then we have \(b_2 = 1\) and \(b_3 + c_3 = 1\). Therefore we obtain \(\sum_{e_2} \lambda(F) = (b_1 + c_1)\alpha\) and \(\sum_{e_3} \lambda(F) = (b_1 + c_1 + 1)\alpha\), so either \(e_4\) or \(e_5\) is a 0-sum edge. Since \(\{\lambda(F_1), \lambda(F_2), \lambda(F_3)\}\) is linearly independent, the Dehn surgery \(\hat{z}^D\) along \(e_4\) or \(e_5\) preserves the 3-independence of \(F\).
ii) The case \(a_1 = c_1 = 0\) and \(b_1 = 1\). We have \(a_2 b_3 + b_2 = 1, b_2 c_3 + b_3 = 1\) and \(\{\lambda(F_1), \lambda(F_2), \lambda(F_4)\}\) is linearly independent. In this case since \(\sum_{e_3} \lambda(F) = (a_2 + b_2 + 1)\beta + (b_3 + 1)\gamma = a_2(1 + b_3)\beta + b_2 c_3 \gamma\) and \(\sum_{e_5} \lambda(F) = (b_2 + 1)\beta + (b_3 + c_3 + 1)\gamma = a_2 b_3 \beta + c_3(1 + b_2)\gamma\), either \(e_3\) or \(e_5\) is a 0-sum edge (if \(a_2 = c_3 = 1\) then \(b_2 + b_3 = 1\)). Then the Dehn surgery \(\hat{z}^D\) preserves the 3-independence of \(F\).
**Remark 4.6.** In the above proposition, except an irregular quasi-decomposition because of the prohibition of \(\hat{z}^D\), we have the fact that each 3-independent small face \(F\) is compressible: such as \(P = P^\rho \Delta^3\) when \(F\) is a triangle, \(P = P^\rho P^3(3)\) or \(P^\rho P^3(3)\) when \(F\) is a quadrilateral and \(P = \hat{z}^D(P^\rho P^3(3))\) or \(\hat{z}^D(P^\rho P^3(3))\) when \(F\) is a pentagon respectively. In all cases the number of faces decreases by this decomposition.
**Proposition 4.7.** Let \(P\) be a \((\mathbb{Z}_2)^3\)-colored polytope except \(\Delta^3, P^3(3)\) and \(I^3\) as a non-colored polytope. If there exists a 2-independent small face of \(P\) then either \(P\) or \(\hat{z}^{-1} P\) is quasi-decomposable.
**Proof.** If there exists a triangular face of \(P\) except \(\Delta^3\) and \(P^3(3)\) then \(P\) is decomposable as a non-colored polytope and so \((P, \lambda)\) is quasi-decomposable from Proposition 3.6. Therefore we can assume that \(P\) has no triangular face. Let \(F\) be a 2-independent small face of \(P\). We notice that the inverse surgery \(\hat{z}^{-1}\) is allowed in the category of \((\mathbb{Z}_2)^3\)-colored polytopes when \((P, \lambda)\) is not quasi-decomposable by Lemma 3.6.
1) When \(F\) is a quadrilateral, the number of quadrilaterals adjacent to \(F\) is at most two because \(P \neq I^3\) and the situation around a 2-independent quadrilateral \(F\) is shown as Figure 12 where \(\ast = \beta\) or 0. If \(F_1\) and \(F_2\) are quadrilateral (the third
Figure 12. The compression of a 2-independent quadrilateral.
Figure 13. The compression of a 2-independent pentagon.
Remark 4.8. When $F$ is a pentagon in the proof of the above proposition, although the compression of the triangle of $\sharp^{-1}P$ does not change the number of faces compared with the beginning, $F$ is transformed into a quadrilateral by this step (see the third diagram). Then we apply the argument (2) in the proof of Proposition 4.7 to the quadrilateral so that the number of faces in the resulting polytope is one less than the number of faces in $P$.
In consequence of Propositions 4.5 and 4.7 we can reduce any $(\mathbb{Z}_2)^3$-colored polytope to $\Delta^3$, $I^3$ and $P^3(3)$ with a certain coloring by using the surgeries $\sharp^{-1}$, $\sharp^D = (\sharp^{-1}\Delta^3) \circ \sharp^{-1}$ (without $\sharp$) and the inverses of connected sums $\sharp^\tau$, $\sharp^{\tau'} = \sharp \circ \sharp$. From Examples 2.3 and 2.4 the possible colorings on $P^3(3)$ (resp. $I^3$) are only three...
Corollary 4.10. We restrict the above theorem to 3- (resp. 4-) colored polytopes, and along vertical edges and \((\#)\) by using two operations \(\Delta^3\) and \(\Delta^3\). Since the surgeries \(\zeta\) and \(\zeta^{-1}\) preserves the number of colors of faces, and the connected sum \(\#\) increases the number of faces, it is clear that these four polytopes can not be constructed from others by using only \(\zeta\), \(\zeta\) and \(\zeta^{-1}\). Therefore we have,
**Theorem 4.9.** Each \((\mathbb{Z}_2)^3\)-colored polytope \((P^3, \lambda)\) can be constructed from \(\Delta^3\), \((I^3, \lambda_0)\), \((P^3(3), \lambda_1)\) and \((P^3, \lambda_2)\) by using two operations \(\zeta\) and \(\zeta\).
The topological translation of the above theorem is Theorem 4.3 shown in the introduction. We restrict the above theorem to 3- (resp. 4-) colored polytopes, and obtain improvements of Theorem 4.4 and Corollary 4.5 as follows.
**Corollary 4.10.** (1) Each 3-colored polytope \((P^3, \lambda)\) can be constructed from \((I^3, \lambda_0)\) by using two operations \(\zeta\) and \(\zeta\).
(2) Each 4-colored polytope \((P^3, \lambda)\) can be constructed from \(\Delta^3\) and \((I^3, \lambda_0)\) by using two operations \(\zeta\) and \(\zeta\).
5. **Non-decreasing constructions of small covers**
Since the operations \(\zeta\) and its inverse \(\zeta^{-1}\) both correspond to surgeries on small covers, we followed Izmestiev’s standpoint in [8] and used the surgery \(\zeta\) in the previous section. However in [8] Lü and Yu considered a “non-decreasing” construction by only operations that number of faces is not decreased, and therefore the use of \(\zeta\) is prohibited. To cancel some obstacles they produced new operations \(\zeta^{\text{even}}, \zeta^\Delta\) and \(\zeta^\ominus\), and showed the following theorem (cf. [8] Theorem 1.1).
**Theorem 5.1** (Lü and Yu). Each \((\mathbb{Z}_2)^3\)-colored polytope \((P^3, \lambda)\) can be constructed from \(\Delta^3\) and \((P^3(3), \lambda_2)\) by using seven operations \(\zeta, \zeta^\circ, \zeta^{\text{even}}, \zeta^{-1}, \zeta^\Delta, \zeta^\bigcirc\) and \(\zeta^\bigcirc\).
However there is a gap in the proof of their paper (we shall point it out later). In this section we also consider a non-decreasing construction of small covers in their standpoint. At first we start with 3-colored polytopes (i.e. linear models). In [8] Izmestiev claimed that each 3-colored polytope can be constructed from 3-colored prisms \(P^3(2I)\) by using \(\zeta\) and \(\zeta^{-1}\) in the proof of Theorem 5.1. From the relation \(P^3(2I) = P^3\zeta^e \cdots \zeta^e P^3\), we can obtain a construction of 3-colored polytopes as follows.
**Proposition 5.2.** Each 3-colored polytope \((P^3, \lambda)\) can be constructed from \((I^3, \lambda_0)\) by using three operations \(\zeta, \zeta^e\) and \(\zeta^{-1}\).
From the above examination we use the operation \(\zeta^e\) and \(\zeta^{-1}\) instead of \(\zeta\) below. Then we can also use the Dehn surgery \(\zeta^D\) and its inverse because of the relations \(\zeta^D = \zeta^e \Delta^3\) and \(\zeta^{D^{-1}} = \zeta^D\). Applying Proposition 5.2 to the above proposition, we have,
**Proposition 5.3.** Each 4-colored polytope \((P^3, \lambda)\) can be constructed from \(\Delta^3\) and \((I^3, \lambda_0)\) by using three operations \(\zeta, \zeta^e\) and \(\zeta^{-1}\).
On the other hand there exist some obstacles for the construction of general \((\mathbb{Z}_2)^3\)-colored polytopes. At first we must prove that Lemma 3.6 also holds for the surgery \(\zeta\).
Lemma 5.4. Let \((P, \lambda)\) be a \((\mathbb{Z}_2)^3\)-colored polytope and \(e\) be an edge of \(P\) but not an edge of a quadrilateral. Suppose that the 3-connectedness of the 1-skeleton of \(P\) is destroyed after doing surgery \(\natural\) along the edge \(e\). Then \((P, \lambda)\) is quasi-decomposable.
Proof. In the Figure\[1\] we assume that the surgery \(\natural\) destroys the 3-connectedness of the 1-skeleton of \(P\). Then there exists a face \(F\) such that \(F \cap F_2 \neq \emptyset\) and \(F \cap F_4 \neq \emptyset\) (see Figure\[1\]). Since neither \(F_1\) nor \(F_3\) is a quadrilateral, \(P \neq P^3(3)\) and we can assume that \(e_1\) is not adjacent to \(e_2\) (i.e., \(R \neq Q\)). When \(e'_1\) is adjacent to \(e_4\) (i.e., \(R' = Q'\)), the 1-skeleton of \(P\) becomes disconnected after cutting the three non-adjacent edges \(e_1, e_2, e'\). Therefore \(P\) is decomposable as a non-colored polytope, and so \(P\) is quasi-decomposable from Proposition\[1\]. We assume that \(e'_1\) is not adjacent to \(e_4\) (i.e., \(P' \neq Q'\)). We do the inverse surgery \(\natural^{-1}\) along the pair of edges \(\{e'_1, e_4\}\) where \(i = 3\) when \(\lambda(F)\) is either \(\alpha\) or \(\alpha + \beta\), and \(i = 1\) when it is not so. If the inverse surgery \(\natural^{-1}\) is not allowed then \((P, \lambda)\) is quasi-decomposable from Lemma\[3\]. Then the graph of \(\natural^{-1}P\) becomes disconnected after cutting the three non-adjacent edges \(e_2, e_i\) and the edge to which \(e'_1\) and \(e_4\) were glued by \(\natural^{-1}\), and \(\{\lambda(F), \lambda(F_2), \lambda(F_4)\}\) is linearly independent. Therefore \(\natural^{-1}P\) is decomposable as a \((\mathbb{Z}_2)^3\)-colored polytope such as \(\natural^{-1}P = P_1 \sharp P_2\), or equivalently \(P = P_1 \sharp e P_2\) i.e. \(P\) is quasi-decomposable. \(\square\)
![Figure 14. The obstacle of the surgery \(\natural\).
Remark 5.5. In [3] Izmestiev used the above lemma only when \(F_4\) in Figure\[1\] is a quadrilateral. In this case \(P\) is always decomposable as a non-colored polytope. In [8] Lü and Yu claimed that this argument can be generalized to every case under the hypothesis of Lemma\[5\] without a proof (cf. [8, Proposition 2.5]), and proved Theorem\[7\] using this claim when \(F_4\) is a pentagon, too. However their claim is incorrect (see Figure\[1\]). Although there is a gap in their proof of Theorem\[7\] the proof is complemented by using Lemma\[5\] instead of their key lemma [8, Proposition 2.5]. Furthermore the theorem is improved by replacing \(\natural\) with \(\natural^{\square}\) as follows: Each \((\mathbb{Z}_2)^3\)-colored polytope \((P^3, \lambda)\) can be constructed from \(\Delta^3\) and \((P^3(3), \lambda_2)\) by using seven operations \(\natural^3, \natural^e, \natural^{ee}, \natural^{-1}, \natural^c\) \((i = 3, 4, 5)\).
From the discussion of the previous section, we can reduce each \((\mathbb{Z}_2)^3\)-colored polytope \(P\) to polytopes which have less faces than \(P\) by using the inverses of \(\natural\) and \(\natural^c\) when \(P\) has a 3-independent small face, or a 2-independent triangle, or a pair of 2-independent quadrilaterals adjacent to each other. Moreover we point out that each 2-independent pentagon can be compressed by using the surgery \(\natural\) as shown in Figure\[1\].
Figure 15. A counter example of [8, Proposition 2.5].
Figure 16. Another compression of a 2-independent pentagon. In the first diagram we may assume that $F_3$ is not quadrilateral by replacing it by $F_2$ if necessary. Then we can do the surgery $\natural$ for the edge $e_3$ and transform $F$ into a triangle (second diagram). Here when the surgery $\natural$ is not allowed, $P$ is quasi-decomposable from Lemma 5.4. Then the triangle can be compressed by $(\sharp e)^{-1}P^3(3)$ and we have $P = \natural^{-1}(P'\sharp e^2P^3(3))$ (third diagram).
In general when colors of two faces on ends of an edge of big faces coincide, we can do the surgery $\natural$ along this edge and decrease the number of faces. Then we can reduce $P$ to $\tilde{P}$ which satisfies the following conditions:
1. $\tilde{P}$ is not quasi-decomposable,
2. each small face of $\tilde{P}$ is an isolated 2-independent quadrilateral,
3. two colors of faces on ends of every edge which is adjacent to big faces do not coincide.
Figure 17. Example of an irreducible polytope: truncated octahedron with a $(\mathbb{Z}_2)^3$-coloring (cf. [8, Example 2.1]).
There are many polytopes satisfying the above condition (see Figure 17). Obviously such a polytope is irreducible by using the inverses of only operations $\triangledown$, $\triangle$ and $\ast^{-1}$. Then we need a coloring change operation $\triangledown^c_i$ in [8].
**Definition 5.6** (The coloring change $\triangledown^c_i$). The operation in Figure 18 is called the coloring change $\triangledown^c_i$ for a 2-independent $i$-gon. This operation is defined as the connected sum along faces to the $i$-gonal prism $P^3(i)$ in particular $\triangledown^c_i = \triangledown^\Delta(P^3(3), \lambda_2)$ (see [8]). It is clear that $\triangledown^c_i$ is invertible because $(\triangledown^c_i)^{-1} = \triangledown^c_i$.
Figure 18. The coloring change $\triangledown^c_i$ for 2-independent $i$-gon.
By using the operation $\triangledown^c_i$, we can change a color of each 2-independent quadrilateral, and compress it by the surgery $\ast$. Moreover the 3-colored cube $(P^3, \lambda_0)$ is obtained by this operation from other basic polytopes such as $\triangledown^c_i(P^3, \lambda_i)$ ($i = 1$ or 3). Therefore we have an improvement of Theorem 5.1 as follows.
**Theorem 5.7.** Each $(\mathbb{Z}_2)^3$-colored polytope $(P^3, \lambda)$ can be constructed from $\Delta^3$, $(P^3(3), \lambda_1)$ and $(P^3(3), \lambda_2)$ by using four operations $\triangledown$, $\triangledown^c$, $\ast^{-1}$ and $\triangledown^c_i$.
The topological translations of Propositions 5.2, 5.3 and Theorem 5.7 are stated in Theorem 1.6.
6. **Locally standard 2-torus manifolds over $D^3$**
In this section we shall give a remark for 2-torus manifolds. A 2-torus manifold $M^n$ is an $n$-dimensional closed smooth manifold with an effective action of $(\mathbb{Z}_2)^n$ (see [6], [7] for detail). If the action is locally standard then the orbit space $Q$ is a nice manifold with corners. When $Q$ is a simple convex polytope, $M$ is a small cover.
We consider the case that $Q$ is a 3-dimensional disc $D^3$ with a simple cell decomposition of the boundary $\partial D^3$, i.e. a locally standard 2-torus manifold over $D^3$. This class is a little wider than 3-dimensional small covers. In fact the 1-skeleton of $Q$ is a 2-connected 3-valent planner graph. This graph is simple and 3-connected if and only if $Q$ is a simple convex polytope. In this category there is little obstacle of surgeries. Therefore it becomes easy to discuss in previous sections.
**Example 6.1.** In Figure 19 we show the characteristic functions of $S^3$ with a standard $(\mathbb{Z}_2)^3$-action and three different $(\mathbb{Z}_2)^3$-colorings of the 2-sided prism $P^3(2)$, respectively. Then the associated 2-torus manifolds $M(P^3(2), \lambda_i)$ are non-equivariantly homeomorphic to $S^1 \times S^2$, $S^2$-bundle over $S^1$ characterized by the conjugation $z \mapsto \bar{z}$ on $S^2 = CP^1$ and $S^1 \times S^2$ according to $i = 0, 1, 2$. We denote $M(P^3(2), \lambda_1)$ by $S^1 \times_{\mathbb{Z}_2} S^2$ where a $\mathbb{Z}_2$-action on $S^1 \times S^2$ is given as follows: $t \cdot (s, z) = (-s, \bar{z})$.
Remark 6.2. We can easily verify the following relations:
1. $\# \circ$ is trivial and $\#^e \circ = \sharp$.
2. $\# P^3(2)$ (or $\#^e P^3(2)$ along the horizontal edge) is a blow up shown in Figure 20 and $\#^e P^3(2)$ (along the vertical edge) is trivial.
3. $\sharp I^3(1, \lambda_0) = (P^3(2), \lambda_0)$ and $\sharp (P^3(2), \lambda_0) = \circ$.
4. $\sharp D^3 = (P^3(2), \lambda_2)$.
5. $\sharp P^3(3, \lambda_1) = (P^3(2), \lambda_1)$.
We notice that if $Q$ is not 3-connected then $Q$ is decomposable as a $(\mathbb{Z}_2)^3$-colored cell decomposition of $D^3$. Therefore applying the above remark (3), (4) and (5) to Theorem 4.9 we obtain the following corollary immediately.
Corollary 6.3. Each $(\mathbb{Z}_2)^3$-colored cell decomposition of $D^3$ can be constructed from $\circ$, $(P^3(2), \lambda_0)$, $(P^3(2), \lambda_1)$ and $(P^3(2), \lambda_2)$ by using two operations $\sharp$ and $\sharp^e$.
In the category of 2-torus manifolds, there is little obstacle for surgeries and blow downs. Therefore we need not consider the case that surgeries are not allowed (e.g. Lemmas 3.6 and 5.4), and obtain the following theorem.
Theorem 6.4. (1) Each 3-colored cell decomposition of $D^3$ can be constructed from $\circ$ by using the inverse surgery $\sharp^{-1}$.
(2) Each 4-colored cell decomposition of $D^3$ can be constructed from $\circ$ by using the inverse surgery $\sharp^{-1}$, the Dehn surgery $\sharp D^3 (= \sharp^e D^3)$ and the blow up $\sharp^e D^3$.
(3) Each $(\mathbb{Z}_2)^3$-colored cell decomposition of $D^3$ can be constructed from $\circ$ by using the inverse surgery $\sharp^{-1}$ and connecting $D^3$, $(P^3(2), \lambda_1)$ and $(P^3(3), \lambda_2)$ by the operations $\sharp$ and $\sharp^e$.
Proof. Let $(Q, \lambda)$ be a $(\mathbb{Z}_2)^3$-colored cell decomposition of $D^3$ but not $\circ$. If a 2-gonal face appears in the following discussion then $(P^3(2), \lambda_1)$ is separated from $Q$ or we do the surgery $\sharp$ and a 2-gon is compressed immediately.
(1) Each 3-colored cell decomposition except $\varnothing$ can be done the surgery $\natural$ and decrease the number of faces.
(2) In the proof of Proposition 4.2 the Dehn surgery $\natural D$ can be continued until a triangle appears because there is no obstacle of $\natural D$. Therefore each 4-colored cell decomposition of $D^3$ can be reduced to a 3-colored cell decomposition by using $\natural D$ and the blow down $\natural - 1 \Delta_3$.
(3) In the proofs of Propositions 4.5 and 4.7 we need not consider the quasi-decomposition by prohibition of surgeries. When $Q$ has a 3-independent small face, $Q$ can be reduced by the blow downs $\natural - 1 P^3(3)$, $\natural - 1 P^3(2)$, $\natural$ and $\natural D$. When $Q$ has a 2-independent triangle, $Q$ can be reduced by the blow downs $\natural - 1 P^3(3)$ and $\natural - 1 P^3(2)$ (along the horizontal edge). Since each 2-independent quadrilateral (or pentagon) has a 3-colored edge, we can do the surgery $\natural$ along this edge in this category and decrease the number of faces. Therefore we can reduce $Q$ to the basic polytopes $\Delta_3$, $P^3(3)$ and $P^3(2)$ by using $\natural$ and inverses of $\natural$ and $\natural D$. From the relations (3), (4) and (5) in Remark 6.2, $\natural(P^3(3), \lambda_2) = \natural D \circ \natural \Delta_3$, $\natural(P^3(3), \lambda_1) = \natural - 1 \circ \natural(P^3(2), \lambda_1)$ and so on. Then the proof is complete.
The topological translation of the above theorem is stated in Theorem 1.7.
Acknowledgments. Finally the author would like to thank Professor M. Masuda for his advice and stimulating discussions.
References
1. M. Davis and T. Januszkiewicz, Convex polytopes, Coxeter orbifolds and torus actions, Duke Math. J. 61 (1991), 417–451.
2. B. Grünbaum, Convex Polytopes, Graduate Text in Mathematics 221, 2nd ed., Springer 2003.
3. I. V. Izmestiev, Three-dimensional manifolds defined by coloring a simple polytope, Math. Note 69 (2001), 340–346.
4. R. Kirby, A calculus for framed links in $S^3$, Invent. Math. 45 (1978), 35–56.
5. S. Kuroki, Operations on 3-dimensional small covers, Chinese Annals Math. Ser. B, 31B(3) (2010), 393–410.
6. Z. Lü, 2-torus manifolds, cobordism and small covers, Pacific J. Math., 241 (2009), 285–308.
7. Z. Lü and M. Masuda, Equivariant classification of 2-torus manifolds, Colloq. Math., 115 (2009), 171–188.
8. Z. Lü and L. Yu, Topological types of 3-dimensional small covers, to appear in Forum Mathematicum, arXiv:0710.4196
9. Z. Lü and L. Yu, On 3-manifolds with locally standard $(\mathbb{Z}_2)^3$-actions, preprint, arXiv:0807.3062v2.
10. E. E. Moise, Geometric Topology in Dimensions 2 and 3, Graduate Texts in Math. 47, Springer-Verlag, 1977.
11. H. Nakayama and Y. Nishimura, The orientability of small covers and coloring simple polytopes, Osaka J. Math. 42-1 (2005), 243–256.
12. Y. Nishimura, Equivariant surgeries of small covers (Japanese), RIMS Kokyuroku 1393 (2004), 44–47. | 2025-03-05T00:00:00 | olmocr | {
"pdf_page_numbers": [
[
0,
3026,
1
],
[
3026,
7089,
2
],
[
7089,
11316,
3
],
[
11316,
14260,
4
],
[
14260,
17874,
5
],
[
17874,
19834,
6
],
[
19834,
21600,
7
],
[
21600,
24693,
8
],
[
24693,
28498,
9
],
[
28498,
31533,
10
],
[
31533,
34770,
11
],
[
34770,
35797,
12
],
[
35797,
39389,
13
],
[
39389,
42747,
14
],
[
42747,
43884,
15
],
[
43884,
46981,
16
],
[
46981,
48993,
17
],
[
48993,
51952,
18
]
]
} |
f31a14e0351c1b228220254e236aff384e981274 | {
"olmocr-version": "0.1.59",
"pdf-total-pages": 7,
"total-fallback-pages": 0,
"total-input-tokens": 26438,
"total-output-tokens": 9278,
"fieldofstudy": [
"Medicine"
]
} | Relationship Between Platelet Count and In-hospital Mortality in Adult Patients With COVID-19: A Retrospective Cohort Study
Qilin Yang†‡, Jun Gao‡, Xiaomei Zeng¹, Junyu Chen* and Deliang Wen*¹
¹Department of Critical Care, The Second Affiliated Hospital of Guangzhou Medical University, Guangzhou, China,
²Department of Nephrology, Peking University International Hospital, Beijing, China
Background: The coronavirus disease 2019 (COVID-19) has become a global pandemic. Systemic inflammation in COVID-19 patients has been associated with poor clinical outcome. This study aims to determine the relationship between platelet count and in-hospital mortality.
Methods: The original data of this study were from article development and validation of a predictive model of in-hospital mortality in COVID-19 patients. In this secondary analysis, we adopted multi-variable logistic regression analyses and smooth curve fitting to assess the independent association between platelet count and in-hospital mortality. We further applied a two-piecewise linear regression model to examine the nonlinear association between platelet count and in-hospital mortality.
Results: Of the 2006 patients, the average age of the participants was 65.9 ± 16.5 years and 42.6% were women. We observed a U-shaped relationship between platelet count and in-hospital mortality. We found two different slopes, the correlations between platelet count and in-hospital mortality of COVID-19 patients were totally different below and above the inflection point which was around $370 \times 10^9/L$. On the left side of the inflection point, the OR was 0.996 (OR: 0.996, 95%CI: 0.994–0.998, $p < 0.001$). On the right side of the inflection point, the OR was 1.011 (OR: 1.011, 95%CI: 1.001–1.021, $p = 0.029$).
Conclusions: A U-shaped association between platelet count and in-hospital mortality was found in the patients with COVID-19. The optimal of platelet count associated with the lowest risk of in-hospital mortality was around $370 \times 10^9/L$.
Keywords: platelet count, in-hospital mortality, ferritin, coronavirus disease 2019, systemic inflammation
BACKGROUND
The coronavirus disease 2019 (COVID-19), an infectious disease caused by a novel strain of human coronavirus, has become the focus of attention worldwide (1). Systemic inflammation in COVID-19 patients has been associated with poor clinical outcome (2–4). Platelets, nucleate megakaryocyte fragments circulating in the blood, play a crucial role in inflammatory diseases (5). There is growing recognition of the critical role of platelets in inflammation and immune responses (6). Previous studies have shown that platelet count is correlated with COVID-19 mortality (7–9).
In the general (10), COPD (11), venous thromboembolism (12), and elderly (13) populations, a U-shaped association was recognized between platelet count and mortality, though its role in COVID-19 remains unclear. Using data from the study of development and validation of a predictive model of in-hospital mortality in COVID-19 patients (14), a respective cohort study enriched for the presence of comorbidity and containing adjudicated events, we investigated post-hoc the association of platelet count measured at basement with in-hospital mortality.
PARTICIPANTS AND METHODS
Data Source
The original data of this study were from the development and validation of a predictive model of in-hospital mortality in COVID-19 patients study (14). Since Diego et al. have relinquished the ownership of the original dataset to PLoS ONE (https://journals.plos.org/plosone/s/data-availability), we can use this dataset to perform secondary analysis based on different scientific hypotheses. The original study was granted an exempt status and the requirement for obtaining informed consent was waived by the Ethics Committee for Clinical Research of the Hospital Universitario Fundación Jiménez Díaz (14). We followed the Strengthening the Reporting of Observational Studies in Epidemiology guidelines to report this study (15).
Study Population
The original study retrospectively evaluated consecutive hospitalized patients with confirmed moderate or severe COVID-19 from four hospitals [Hospital General de Villalba (Collado Villalba, Madrid), Hospital Infanta Elena (Valdemoro, Madrid), Hospital Universitario Rey Juan Carlos (Móstoles, Madrid), and Hospital Universitario Fundación Jiménez Díaz in Madrid] from 27 February to 17 April 2020. The diagnosis of COVID-19 was based on World Health Organization interim guidance and confirmed by RNA detection of 2019-nCoV in the clinical laboratory of Hospital Universitario Fundación Jiménez Díaz. Diego et al. extracted de-identified data from the Huawei (Huawei Technologies Co., Ltd., Shenzhen, China) platform and the collaboration of Indizen-Scalian (Madrid, Spain). Four patients younger than 18 years old and 60 patients missing platelet count data were excluded in further analysis.
Variable Extraction
Baseline Platelet Count
Baseline platelet count was first collected either in the emergency department or within 3 days from admission to a ward from electronic medical records (14).
Covariates
We included the following variables based on published literature and clinical experience: demographic characteristics and chronic comorbidities (arterial hypertension, diabetes mellitus, smoking habit, cardiovascular disease, and pulmonary disease). Laboratory values included body mass index (BMI), lactate dehydrogenase (LDH), ferritin, D-dimer, absolute lymphocyte count, estimated glomerular filtration rate (eGFR), activated partial thromboplastin time (APTT), and fibrinogen.
Outcome
The outcome was in-hospital mortality which was monitored up to 17 April 2020 (14).
Statistical Analysis
Descriptive analysis was performed for all patients. Categorical variables were expressed as numbers and percentages. Continuous variables were expressed as mean and standard deviation (SD) for normal distributions or median and interquartile range for skewed distributions. We used the chi-square test, one-way ANOVA, and Kruskal–Wallis test for the comparison of categorical, normally distributed, and non-normally distributed continuous variables, respectively. We used dummy variables to indicate missing covariate values (16).
We adopted multi-variable logistic regression analyses and smooth curve fitting to assess the independent association between platelet count and in-hospital mortality. We further applied a two-piecewise linear regression model using a smoothing curve to examine the nonlinear association between platelet count and in-hospital mortality. A likelihood ratio test was conducted to compare the one-line linear regression model with the two-piecewise linear model. All the analyses were performed with the statistical software packages R 3.3.2 (http://www.R-project.org, The R Foundation) and Free Statistics.
TABLE 1 | Multivariable logistic regression models evaluating the association between platelet count and in-hospital mortality.
| Variable | Total n | In-hospital mortality n (%) | Model I OR (95%CI) | P-value | Model II OR (95%CI) | P-value | Model III OR (95%CI) | P-value |
|---------------------------|---------|-----------------------------|--------------------|---------|---------------------|---------|---------------------|---------|
| Platelet <100 (×10^9/L) | 81 | 28 (34.6) | 3.72 (2.03–6.81) | <0.001 | 3.03 (1.54–5.98) | 0.001 | 3.65 (1.74–7.66) | 0.001 |
| Platelet 100–300 (×10^9/L)| 1561 | 300 (19.2) | 1.67 (1.10–2.54) | 0.015 | 1.58 (1.00–2.48) | 0.048 | 1.93 (1.18–3.16) | 0.009 |
| Platelet 300–400 (×10^9/L)| 225 | 28 (12.4) | 1.00 (Ref) | 1.00 (Ref) | 1.00 (Ref) | 1.00 (Ref) | 1.00 (Ref) | 1.00 (Ref) |
| Platelet ≥400 (×10^9/L) | 139 | 20 (14.4) | 1.18 (0.64–2.19) | 0.595 | 1.19 (0.61–2.33) | 0.613 | 1.50 (0.73–3.09) | 0.270 |
Model I, No adjustment.
Model II, Adjusted for age and sex.
Model III, Adjusted for all covariates in Table 2.
TABLE 2 | Baseline characteristics of platelet count analysis.
| Variables | Total (n = 2,006) | <100 (n = 81) | 100–300 (n = 1,561) | 300–400 (n = 225) | ≥400 (n = 139) | P-value |
|---------------------------|-------------------|--------------|---------------------|------------------|---------------|---------|
| Age (years) | 65.9 ± 16.5 | 69.3 ± 17.0 | 66.0 ± 16.5 | 64.5 ± 16.8 | 64.4 ± 16.6 | 0.108 |
| Sex (Female), n (%) | 854 (42.6) | 27 (33.3) | 655 (42.0) | 113 (50.2) | 59 (42.4) | 0.038 |
| Smoking, n (%) | 78 (3.9) | 3.7 (3.7) | 62 (4.0) | 9 (4.0) | 4 (2.9) | 0.964 |
| Cardiovascular disease, n (%) | 312 (15.6) | 20 (24.7) | 234 (15) | 39 (17.3) | 19 (13.7) | 0.097 |
| Pulmonary disease, n (%) | 332 (16.6) | 17 (21) | 271 (17.4) | 32 (14.2) | 12 (8.6) | 0.027 |
| Diabetes, n (%) | 400 (19.9) | 17 (21) | 298 (19.2) | 50 (22.2) | 35 (25.2) | 0.287 |
| Hypertension, n (%) | 895 (44.6) | 40 (49.4) | 695 (44.5) | 98 (43.6) | 62 (44.6) | 0.840 |
| BMI (Kg/m²) | 28.2 ± 6.0 | 28.5 ± 5.2 | 28.3 ± 6.0 | 28.6 ± 6.8 | 26.6 ± 4.7 | 0.157 |
| D-dimer (µg/l) | 623 (336, 1,106) | 807 (424, 1,442) | 567.0 (318, 1,016) | 730 (391, 1,330) | 1002 (593, 2,110) | <0.001 |
| Ferritin (ng/ml) | 775 (394, 1,484) | 879 (425, 1,803) | 775.0 (376, 1,476) | 756 (388, 1,428) | 783 (532, 1,430) | 0.597 |
| EGFR (mL/min/L찾기울드) | 89.7 (76.4, 101.2) | 89.7 (76.4, 100.3) | 88.7 (75.4, 99.9) | 94.4 (80.4, 103.6) | 94.7 (78.9, 105.3) | 0.002 |
| Lymphocyte (×10^9/L) | 1.0 (0.7, 1.3) | 0.7 (0.5, 1.1) | 0.9 (0.7, 1.3) | 1.1 (0.8, 1.4) | 1.1 (0.7, 1.5) | <0.001 |
| LDH (µL/L) | 298 (233, 384) | 278 (209, 352) | 293 (232, 380) | 320 (257, 409) | 326 (248, 399) | <0.001 |
| Fibrinogen (mg/dl) | 703.5 ± 206.0 | 615.1 ± 167.4 | 690.3 ± 191.7 | 775.6 ± 253.7 | 786.6 ± 241.6 | <0.001 |
| APTT (seconds) | 31.1 ± 7.9 | 32.5 ± 5.9 | 31.2 ± 8.6 | 30.3 ± 4.4 | 30.7 ± 4.7 | 0.158 |
BMI, body mass index; eGFR, estimated glomerular filtration rate; LDH, lactate dehydrogenase; APTT, activated partial thromboplastin time; SD, standard deviation.
software version 1.3 (17). A two-tailed test was performed and \( p < 0.05 \) was considered statistically significant.
RESULTS
Baseline Characteristics of Participants
From the original cohort, after excluding 4 patients younger than 18 years old and 60 patients missing platelet count data on admission, 2,006 patients were included in our study. Among all these patients, the average age of the participants was 65.9 ± 16.5 years and 42.6% were women. A total of 3.7% of patients were smokers. Compared with the first platelet count group, the fourth platelet count group contained more women, cardiovascular disease, higher D-dimer, EGFR, lymphocyte, and LDH levels, and less pulmonary disease.
Outcome
The overall in-hospital mortality was 18.7%. Figure 1 shows the in-hospital mortality in different platelet count groups. The in-hospital mortality in groups 1–4 was 34.6, 19.2, 12.4, and 14.4%, respectively. The results of the univariate and multivariate logistic regression models are shown in Table 1 and Table 4. In the fully adjusted model (adjusted for all covariates in Table 2), categorized platelet count in the multivariate logistic regression model seemed to confirm a non-linear relationship between platelet count and in-hospital mortality. The 300–400(×10^9/L) platelet count group had the lowest in-hospital mortality.
We tried to look at different thresholds to identify patients at risk and used <100 × 10^9/L vs. 100–550 × 10^9/L vs. >550 × 10^9/L for sensitivity analysis. Compared with 100–550 × 10^9/L groups, the ORs of <100 × 10^9/L and >550 × 10^9/L were 2.34 (1.35–4.07) and 1.69 (0.51–5.6) after adjusting for all covariates in Table 2.
The Nonlinearity Relationship Between Platelet Count and In-hospital Mortality
Through the multivariate logistic regression model and smooth curve fitting, we observed that the relationship between platelet count and in-hospital mortality was non-linear (Figure 2). Data were fit to a piecewise multivariate logistic regression model.
and found two different slopes. In our study, the \( P \)-value for the non-linear test was 0.037 (Table 3), we thus used a two-piecewise model to fit the link between platelet count and in-hospital mortality. We found an inflection point at about 370 \( \times 10^9/L \) (Figure 2). On the left side of the inflection point, the OR was 0.996 (OR: 0.996, 95% CI: 0.994–0.998, \( p < 0.001 \)). On the right side of the inflection point, the OR was 1.011 (OR: 1.011, 95% CI: 1.001–1.021, \( p = 0.029 \)). It suggests that the risk of in-hospital mortality started to decrease by 0.4% per 1 \( \times 10^9/L \) platelet change until a platelet count of \( \sim 370 \times 10^9/L \). Then the risk of in-hospital mortality started to increase by 1.1% per 1 \( \times 10^9/L \) platelet change (\( P \)-value for non-linear test was 0.037).
**Other Risk Factors for In-hospital Mortality in Patients With COVID-19**
Univariate logistic and multivariable logistic regression analysis of risk factors for in-hospital mortality in patients with COVID-19 is reported in Table 4. We found age, male, history of pulmonary disease, history of diabetes, D-dimer, and LDH were independent risk factors for in-hospital mortality in this cohort (all \( P < 0.05 \)).
We also detected the association between platelet count and ferritin in order to understand the relationship between platelet count and inflammation. Based on Figure 3, platelet count was negatively associated with ferritin below 300 \( \times 10^9/L \).
**DISCUSSION**
In this observational retrospective cohort study, we tried to examine the optimal platelet count associated with in-hospital mortality in patients with COVID-19. A U-shaped association between platelet count and in-hospital mortality was found in the cohort. The correlations between platelet count and in-hospital mortality of COVID-19 patients were totally different below and above the inflection point which was around 370 \( \times 10^9/L \). Platelet count, as assessed at baseline, was negatively associated with in-hospital mortality of COVID-19 patients below 370 \( \times 10^9/L \), and it was positively associated above 370 \( \times 10^9/L \). The optimal platelet count associated with the lowest risk of in-hospital mortality was around 370 \( \times 10^9/L \).
Low platelet count is a common laboratory finding in patients with severe COVID-19 (18). Our study found that 4% (81) of patients had a platelet count <100 \( \times 10^9/L \). In a previous study, thrombocytopenia was associated with poor outcome in patients with COVID-19 (18–21). Their results are akin to part of our findings where platelet counts <100 \( \times 10^9/L \) correlated with the highest in-hospital mortality. However, most of the studies converted platelet count into dichotomous variables and only compared the outcome of platelet counts <100 \( \times 10^9/L \) (22) or 125 \( \times 10^9/L \) (23) or 150 \( \times 10^9/L \) (24) with other COVID-19 patients. It was hard in these studies to find either a non-linearity relationship between platelet count and in-hospital mortality or an optimal platelet count associated with the lowest risk of in-hospital mortality.
We used smooth curve fitting (25) and a two-piecewise linear regression model (25) to determine that a high platelet count may also lead to increased mortality in patients with COVID-19. Whereas, in multivariable logistic regression analysis, we only found an increasing trend of mortality in group 4 (platelet \( \geq 400 \times 10^9/L \)) (OR: 1.5, 95% CI: 0.73–3.09, \( P = 0.27 \)). This instability may be the result of the highly sensitive smooth curve fitting in analyzing change trends.
In fact, a U-shaped association between platelet count and in-hospital mortality is universal in different diseases. Several studies have found a non-linear relationship between platelet count and outcomes. Fawzy et al. (11) demonstrated a U-shaped association with platelet count and risk of 3-year all-cause mortality in stable COPD. Van et al. found low and high platelet counts were associated with non-cardiovascular mortality in the elderly, including cancer mortality. Di Micco et al. (12) found a U-shaped relationship between platelet count and the 3-month rate of major bleeding and fatal bleeding in patients with VTE.
The association between platelet count and clinical outcome may not reveal any causality as many parameters could both be the cause and/or the consequence of the changing platelet count.
**TABLE 3** The non-linearity relationship between platelet count and in-hospital mortality.
| Threshold of driving pressure | OR | 95% CI | \( P \)-value |
|-----------------------------|--------|-------------|---------------|
| <370 \( \times 10^9/L \) | 0.996 | 0.994–0.998 | <0.001 |
| \( \geq 370 \times 10^9/L \) | 1.011 | 1.001–1.021 | 0.029 |
| Non-linear test | 0.037 | | |
Adjusted for all covariates in Table 2.
For example, thrombocytopenia could be due to an inflammatory response. The pathophysiology of thrombocytopenia in COVID-19 is hypothetically caused by the alteration of platelet production and consumption (and/or destruction) (26). It affects platelet production by either directly or indirectly affecting hematopoietic stem cells (HSCs), reducing thrombopoietin production, and megakaryocyte maturation due to an increase of specific inflammatory cytokines (27). In this study we also explored the relationship between platelet count and ferritin which is a bio-marker of inflammation and found that a lower platelet count represented higher inflammation in this cohort. Similar to other studies, we found age (21), male gender (28), and underlying chronically illness (29) were risk factors of in-hospital mortality of patients with COVID-19. Our study also found that D-dimer, representing thrombotic risk (30), and LDH, representing systemic inflammation (31), were also associated with prognosis in multivariable logistic regression analysis. These results suggest that age, sex, underlying chronically illness, D-dimer, and LDH deserve further study.
Our research has the following limitations that need attention. First, residual confounders potentially exist, as with all retrospective analyses. By maximizing the sample size, we adjusted for all possible confounders we could. Second, our data are only from Spain and cannot cover other populations. Third, the association between platelet count and clinical outcome may not reveal any causality. However, we tried various techniques such as the non-linearity relationship test and used different thresholds group analyses to confirm this relationship which is worthy of further investigation.
## CONCLUSION
A U-shaped association between platelet count and in-hospital mortality was found in patients with COVID-19. The optimal platelet count associated with the lowest risk of in-hospital mortality was around $370 \times 10^9/L$.
---
### TABLE 4 | Univariate logistic and multivariable logistic regression models evaluating the association between platelet count and in-hospital mortality.
| Variable | Univariate logistic analysis | Multivariable logistic analysis |
|----------------------------|------------------------------|---------------------------------|
| | OR (95%CI) | P-value | OR (95%CI) | P-value |
| Platelet <100 ($\times 10^9/L$) | 3.72 (2.03–6.81) | $<$0.001 | 3.65 (1.74–7.66) | 0.001 |
| Platelet 100–300 ($\times 10^9/L$) | 1.67 (1.10–2.54) | 0.015 | 1.93 (1.18–3.16) | 0.009 |
| Platelet $\geq$400 ($\times 10^9/L$) | 1.00 (Ref) | | 1.00 (Ref) | |
| Age (years) | 1.08 (1.07–1.09) | $<$0.001 | 1.09 (1.07–1.101) | $<$0.001|
| Sex (Female), n (%) | 0.74 (0.59–0.92) | 0.008 | 0.53 (0.39–0.70) | $<$0.001|
| Smoking, n (%) | 1.07 (0.62–1.84) | 0.8054 | 1.16 (0.58–2.31) | 0.682 |
| Cardiovascular disease, n (%) | 2.30 (1.76–3.00) | $<$0.001 | 0.87 (0.62–1.22) | 0.407 |
| Pulmonary disease, n (%) | 1.36 (1.03–1.80) | 0.0311 | 1.43 (1.01–2.01) | 0.042 |
| Diabetes, n (%) | 2.08 (1.62–2.67) | $<$0.001 | 1.39 (1.02–1.90) | 0.040 |
| Hypertension, n (%) | 2.83 (2.25–3.56) | $<$0.001 | 0.97 (0.71–1.32) | 0.830 |
| BMI (Kg/m$^2$) | 0.99 (0.98–1.00) | 0.022 | 0.997 (0.96–1.03) | 0.845 |
| D-dimer (µg/l) | 1.002 (1.001–1.003) | $<$0.001 | 1.003 (1.002–1.004) | $<$0.001|
| Ferritin (ng/ml) | 1.00 (0.9999–1.0001) | 0.705 | 1.000 (0.9999–1.0002) | 0.459 |
| EGFR (ml/min/l.73m$^2$) | 0.98 (0.978–0.983) | $<$0.001 | 0.998 (0.989–1.01) | 0.721 |
| Lymphocyte ($\times 10^9/L$) | 1.04 (0.98–1.10) | 0.234 | 1.06 (0.98–1.15) | 0.181 |
| LDH (U/L) | 1.002 (1.001–1.003) | $<$0.001 | 1.004 (1.003–1.004) | $<$0.001|
| Fibrinogen (mg/dl) | 1.00 (0.999–1.000) | 0.044 | 1.00 (0.998–1.001) | 0.600 |
| APTT (seconds) | 0.997 (0.988–1.007) | 0.613 | 1.009 (0.995–1.024) | 0.193 |
---
**FIGURE 3** | Relationship between platelet count and ferritin. Adjusted for all covariates in Table 2 except ferritin.
DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.
ETHICS STATEMENT
Ethical review and approval was not required for the study on human participants in accordance with the local legislation and institutional requirements. Written informed consent from the patients/participants or their legal guardian/next of kin was not required to participate in this study in accordance with the national legislation and the institutional requirements.
REFERENCES
1. Sohrabi C, Alsafi Z, O’Neill N, Khan M, Kerwan A, Al-Jabir A, et al. World health organization declares global emergency: a review of the 2019 novel coronavirus (COVID-19). Int J Surg. (2020) 76:71–6. doi: 10.1016/j.ijsu.2020.02.034
2. Izcovitch A, Ragusa MA, Tortosa F, Lavena Marzio MA, Agnoletti C, Bengolea A, et al. Prognostic factors for severity and mortality in patients infected with COVID-19: a systematic review. PLoS ONE. (2020) 15:e241955. doi: 10.1371/journal.pone.0241955
3. Tjendra Y, Al Mana AF, Espejo AP, Akgun Y, Millan NC, Gomez-Fernandez C, et al. Predicting disease severity and outcome in COVID-19 patients: a review of multiple biomarkers. Arch Pathol Lab Med. (2020) 144:1465–74. doi: 10.5858/arpa.2020-0471-SA
4. Zhang J, Cao YY, Tan G, Dong X, Wang BC, Lin J, et al. Clinical, radiological, and laboratory characteristics and risk factors for severity and mortality of 289 hospitalized COVID-19 patients. Allergy. (2021) 76:533–50. doi: 10.1111/all.14496
5. Assinger A, Schrottmaier WC, Salzmann M, Rayes J. Platelets in sepsis: an update on experimental models and clinical data. Front Immunol. (2019) 10:1687. doi: 10.3389/fimmu.2019.01687
6. Thomas MR, Storey RF. The role of platelets in inflammation. Thromb Haemost. (2015) 114:449–58. doi: 10.1160/TH14-12-1067
7. Xu J, Yang X, Yang L, Zhou X, Wang Y, Wu Y, et al. Clinical course and predictors of 60-day mortality in 239 critically ill patients with covid-19: a multicenter retrospective study from Wuhan, China. Crit Care. (2020) 24:394. doi: 10.1186/s13054-020-03098-9
8. Wei Y, Zeng W, Huang X, Li J, Qiu X, Li H, et al. Clinical characteristics of 276 hospitalized patients with coronavirus disease 2019 in Zengdu district, Hubei province: a single-center descriptive study. BMC Infect Dis. (2020) 20:549. doi: 10.1186/s12879-020-05252-8
9. Ferguson J, Rosser JJ, Quintero O, Scott J, Subramanian A, Gummia M, et al. Characteristics and outcomes of coronavirus disease patients under nonsurge conditions, Northern California, USA, March-April 2020. Emerg Infect Dis. (2020) 26:1679–85. doi: 10.3201/eid2608.201776
10. Kabat GC, Kim MY, Verma AK, Manson JE, Lin J, Lessin L, et al. Platelet count and total and cause-specific mortality in the women’s health initiative. Ann Epidemiol. (2017) 27:274–80. doi: 10.1016/j.annepidem.2017.02.001
11. Fawzy A, Anderson JA, Cowans NJ, Crim C, Wise R, Yates JC, et al. Association of platelet count with all-cause mortality and risk of cardiovascular and respiratory morbidity in stable copd. Respir Res. (2019) 20:86. doi: 10.1186/s12891-019-1059-1
12. Di Micco P, Ruiz-Giménez N, Nieto JA, Auesky D, del Molino F, Valde R, et al. Platelet count and outcome in patients with acute venous thromboembolism. Thromb Haemostasis. (2013) 110:1025–34. doi: 10.1160/TH13-04-0352
13. Van Der Bom JG, Heckbert SR, Lumley T, Holmes CE, Cushman M, Folsom AR, et al. Platelet count and the risk for thrombosis and death in the elderly. J Thromb Haemost. (2009) 7:399–405. doi: 10.1111/j.1538-7836.2008.03267.x
AUTHOR CONTRIBUTIONS
QY designed the study and wrote the manuscript. JC conducted data analysis and wrote the manuscript. XZ drew the figures. JC conducted data analysis. DW conducted data interpretation and modified the manuscript. All authors contributed to the article and approved the submitted version.
ACKNOWLEDGMENTS
We thank Dr. Diego Velasco-Rodriguez and Dr. Juan-Manuel Alonso-Dominguez (Department of Hematology, Hospital Universitario Fundacion Jimenez Diaz, IIS-FJD, Madrid, Spain) for sharing data on PLOS ONE.
Conflict of Interest: 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.
Publisher's Note: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.
Copyright © 2022 Yang, Gao, Zeng, Chen and Wen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms. | 2025-03-05T00:00:00 | olmocr | {
"pdf_page_numbers": [
[
0,
2726,
1
],
[
2726,
6914,
2
],
[
6914,
12649,
3
],
[
12649,
17657,
4
],
[
17657,
22739,
5
],
[
22739,
26912,
6
],
[
26912,
27970,
7
]
]
} |
3aff39054b8ba4dbcec9f71a1a285c14b8fe04c2 | {
"olmocr-version": "0.1.59",
"pdf-total-pages": 14,
"total-fallback-pages": 0,
"total-input-tokens": 40251,
"total-output-tokens": 15009,
"fieldofstudy": [
"Medicine",
"Biology"
]
} | INTRODUCTION
Acute lymphoblastic leukemia (ALL) is a blood cancer in which the immature lymphoid progenitors are neoplastic and show deregulated proliferation. This malignancy is the most common type of leukemia in children under 15 years of age.\textsuperscript{1} Despite impressive improvements in the management of ALL, treatment failure still occurs in nearly 20%-30% of patients.\textsuperscript{2,3}
Iron protects childhood acute lymphoblastic leukemia cells from methotrexate cytotoxicity
Marjan Abedi | Soheila Rahgozar | Abolghasem Esmaeili
Department of Cell and Molecular biology & Microbiology, Faculty of Biological Science and Technology, University of Isfahan, Isfahan, Iran
Correspondence
Soheila Rahgozar, Department of Cell and Molecular Biology & Microbiology, Faculty of Biological Science and Technology, University of Isfahan, Hezar Jarib Street, 81746-73441 Isfahan, Iran. Email: [email protected]
Funding information
This work was supported by a PhD research grant to MA from The University of Isfahan.
Abstract
Drug resistance is a fundamental clinical concern in pediatric acute lymphoblastic leukemia (pALL), and methotrexate (MTX) is an essential chemotherapy drug administered for the treatment. In the current study, the effect of iron in response to methotrexate and its underlying mechanisms were investigated in pALL cells. CCRF-CEM and Nalm6 cell lines were selected as T and B-ALL subtypes. Cells were pretreated with ferric ammonium citrate, exposed to the IC50 concentration of MTX and cell viability was assessed using MTT, colony formation, and flow cytometry assays. Iron-loaded cells were strongly resistant to MTX cytotoxicity. The inhibitory effect of N-acetyl cysteine to reverse the acquired MTX resistance was greater than that of the iron chelator, deferasirox, highlighting the importance of iron-mediated ROS in MTX resistance. Subsequently, the upregulation of BCL2, SOD2, NRF2, and MRP1 was confirmed using quantitative RT-PCR. Moreover, a positive correlation was demonstrated between the MRP1 expression levels and bone marrow iron storage in pALL patients. Further supporting our findings were the hematoxylin and eosin-stained histological sections showing that iron-treated nude mice xenografts demonstrated significantly more liver damage than those unexposed to iron. Overall, iron is introduced as a player with a novel role contributing to methotrexate resistance in pALL. Our findings suggest that the patients' bone marrow iron stores are necessary to be assessed during the chemotherapy, and transfusions should be carefully administrated.
KEYWORDS
acute lymphoblastic leukemia, drug resistance, iron, methotrexate
1
INTRODUCTION
Acute lymphoblastic leukemia (ALL) is a blood cancer in which the immature lymphoid progenitors are neoplastic and show deregulated proliferation. This malignancy is the most common type of leukemia in children under 15 years of age.\textsuperscript{1} Despite impressive improvements in the management of ALL, treatment failure still occurs in nearly 20%-30% of patients.\textsuperscript{2,3}
Moreover, the overall survival of patients with relapsed ALL has remained between 25% and 40% over the years. Although many genetic and epigenetic factors may play role(s) in relapse, multidrug resistance (MDR) is considered as the major element. Several molecular mechanisms are proposed for the generation of MDR, including the upregulation of ABC transporters contributing to a decrease in the cellular accumulation of drugs, mutations in the key genes controlling cell death, and the (hyper) activation of DNA repair, antiapoptotic molecules and survival-related signaling pathways.
Iron metabolism is shown to be altered in tumors. This modification can be partially explained by the iron critical impact on cell biology. The biosynthesis of heme, for example, and the production of Fe-S clusters may regulate the activity of the myriad enzymes modulating cellular metabolism and proliferation. On the other hand, iron overload is implicated as a carcinogen in some studies. The possible mechanisms through which iron may exert its carcinogenic effects include the generation of reactive oxygen species (ROS), activation of the oxidative responsive transcription factors such as AP-1 and NF-κB, influencing JNK and MAPK pathways, and induction of several alterations in the immune system response and cell cycle growth. Moreover, iron may play a role in drug resistance. An association between tumor chemoresistance and H-ferritin subunit overexpression of iron was previously reported. Furthermore, it has been revealed that there is a close relationship between iron deprivation and reduced MDR1 expression in the K562 cell line. Additionally, Li and colleagues demonstrated that iron strengthens the paracrine loop of IL-6 and confers resistance to breast cancer cells against chemotherapy. However, the exact molecular mechanism by which this tiny but vital element can affect the response to therapy remains to be fully established.
A positive correlation between the ALL patients bone marrow iron stores and poor response to treatment was previously shown. Consequently, we hypothesized that iron may participate in drug resistance, and iron overload can be considered a risk factor for ALL relapse. In an attempt to explicit the role of iron in the leukemia cells survival and response to therapy, ferric ammonium citrate was applied on leukemia cell lines and response to methotrexate was primarily investigated. Supportive findings were then obtained through additional ex vivo experiments and generation of ALL xenograft models.
2 | MATERIAL AND METHODS
2.1 | Cell lines
The authenticated leukemic cell lines CCRF-CEM and Nalm6 were obtained from the cell bank of Pasteur Institute of Iran. Cells at passages 10-25 were used for experiments. The culture medium of these cell lines was RPMI1640 (Gibco) supplemented with 10% heat-inactivated fetal bovine serum, FBS (Gibco). Cells were incubated in a humidified atmosphere containing 5% CO2.
2.2 | Patients and sampling
Eighteen children referred to the Seyed-ol-Shohada Hospital (Isfahan, Iran) and newly diagnosed with Philadelphia negative ALL (5 females, 13 males; mean range, 5.09 ± 0.84 years; range, 0.8-14 years) entered the study in addition to 15 non-cancer age matched controls. The project was conducted in accordance with the Declaration of Helsinki, and permitted by the Ethics Committee of the University of Isfahan (agreement number 89/72763). Bone marrow samples were collected with full written informed consent of parents. Extraction of the mononuclear cells and RNA isolation was performed as previously described. Primary data obtained in ALL patients analyses are illustrated in Table S1. Response to treatment in the ALL patients was assessed using the presence of minimal residual disease (mrd), a year after treatment. Gene rearrangements of immunoglobulin heavy chain (IgH) and T-cell receptor gamma (TcRγ) were assessed using PCR-SSCP (polymerase chain reaction coupled single-strand conformation polymorphism).
2.3 | Chemicals and reagents
Ferric ammonium citrate (FAC) was acquired from Merck. Human holo-transferrin, 2,7'-Dichloro-fluorescin diacetate (DCFH-DA), N-acetyl-l-cysteine (NAC), potassium ferrocyanide and nitric acid 65% were purchased from Sigma. Methotrexate (MTX) was bought from Santa Cruz Biotechnology, Inc. Deferasirox (Exjade) (DFX) was obtained from Novartis.
2.4 | Viability/proliferation assays
Cellular proliferation capacity and viability were evaluated by MTT assay (Atocel). CCRF-CEM and Nalm6 cells were seeded into 6-well plates (1 × 10⁶ cells/well) with exposure to different concentrations of FAC, from 0 to 6400 µmol/L, for 24 hours. Afterwards, cells were washed twice with PBS, then seeded into 96-well plates (10 × 10⁴ cells/150 µL culture medium) and incubated with the approximate IC50 concentration of MTX (0.5 µmol/L for CCRF-CEM and 1 µmol/L for Nalm6) for 72 and 96 hours, respectively (in relation to their cell doubling time). Subsequently, 10 µL of MTT solution (0.5 mg/mL) was added into the wells. The formazan crystals were dissolved with 100 µL DMSO after 3 hours, and absorbance was measured using a stat fax 2000 microplate reader (Awareness
Technology, Inc) at 492 nm wavelength. All the experiments were carried out at least with three different subcultures of cells and performed in triplicate in each run. To investigate whether there was a direct association between iron and MTX resistance, cells were pretreated with 20 µmol/L DFX for 15 minutes, and 1000 µmol/L NAC for 3 hours, followed by exposure to 400 µmol/L FAC for 24 hours. Cells were then treated with the IC50 concentration of MTX for 72 hours and percent viability was assessed and compared with cells untreated with either DFX or NAC, using MTT assay.
2.5 **Intracellular iron measurement**
The cellular iron uptake was validated by the quantification of the total iron content of cells using atomic absorption flame emission spectrophotometry (AAS). Cells were treated with 400 µmol/L FAC for 24 hours in the 6-well plates (10 x 10^5 cells/2 mL culture medium). Cells were then washed twice with PBS, harvested and counted. 30 x 10^5 cells were suspended in 65% HNO_3 (1 mL) for 24 hours at room temperature. Finally, lysed cells were analyzed by AA-6200 Atomic Absorption Spectrophotometer (Shimadzu Scientific Instruments, Inc). The concentration of iron was calculated per 10^6 cells.
2.6 **Colony formation assay**
To investigate the proliferation capacity of the cells, their ability to form colonies in a semisolid medium was evaluated. The suspended cells (10 x 10^3 cells/1.5 mL/well) were plated into 6-well plates over an agar underlay (0.3% agar), in a medium containing 10% FBS, 0.1% penicillin-streptomycin and 0.5% agar. Cells were incubated in a 5% CO2 atmosphere at 37°C for 21 days. Subsequently, 40 µL of MTT solution (5 mg/mL) was added to each well and incubated for 96 hours. Eventually, the number of colonies was counted using a light microscope by three researchers in a blind manner according to Table 1. Finally, the percentage of colony formation efficiency, CFE, was calculated as below.
\[
\text{CFE\%} = \frac{\text{Number of colonies in the treated well}}{\text{Number of colonies in the untreated well}} \times 100.
\]
2.7 **Apoptosis detection by flow cytometry**
Cells apoptosis rate was detected by Annexin A5 Apoptosis Detection Kit (Biolegend) according to the manufacturer’s guidelines. Briefly, cells (10 x 10^6 cells/mL) were suspended in Annexin V binding buffer. The tubes containing 100 µL of cell suspension were supplemented by 5 µL FITC Annexin V and 10 µL Propidium Iodide solution. They were then incubated at room temperature for 15 minutes in the dark. 400 µL Annexin V binding buffer was added to each tube and analyzed by flow cytometry.
2.8 **Detection of the ROS**
To detect ROS generation, cells (5 x 10^5/well) were incubated in 500 µL FBS-free medium containing 40 µmol/L DCF for 15 minutes in dark. Samples were run in duplicates. Cells were then diluted with 500 µL ice-cold PBS and the fluorescence intensity was measured, immediately, using a Partec CyFlow ML Flow Cytometer supported by FloMax software in the FL-1 channel. Histograms were plotted using FlowJo software version 7.6.1. H_2O_2 was used as a positive control.
2.9 **RNA isolation and real-time PCR**
RNA isolation, DNase treatment and cDNA synthesis were carried out using TRIzol reagent (Ambion), DNase I and cDNA synthesis kit (Thermo Scientific) according to the manufacturer’s instructions, respectively. Real-time PCR was performed using SYBR Premix Ex Taq II kit (Takara) according to the previously described method. All PCR reactions were done in duplicates during two separate experiments using Chromo 4 Real Time PCR Detection System (Bio-Rad Laboratories, Inc). The comparative 2^-ddCT method was used for data analysis and calculation of relative quantification of gene expression. Two separate experiments were performed for each gene. A list of primers is given in Table S2.
2.10 SOD activity assay
Total superoxide dismutase activity was assessed according to the standard method described by Beauchamp and Fridovich. Briefly, 2 × 10⁶ cells were suspended in 1.5 mL extraction buffer, containing 0.01 M phosphate buffer and 0.2% polyvinylpyrrolidone (PVP) (Sigma), and centrifuged at 10 774 g for 20 minutes at 4°C. Then, 50 µL of supernatant was added to a 1000 µL reaction solution containing 50 mmol/L potassium phosphate buffer (pH 7.8), 13 mmol/L methionine, 2 mmol/L riboflavin, 75 mmol/L nitroblue tetrazolium (NBT) (Sigma), and 0.1 mmol/L EDTA. Test tubes were incubated at room temperature at 5000 lux light intensity for 15 minutes. Finally, the rate of NBT reduction was spectrophotometrically measured at 560 nm. One unit of superoxide dismutase was assigned as the amount required for 50% inhibition of the NBT reduction. The enzyme activity was calculated as units per µg protein.
2.11 Animal model and treatments
Athymic nude mice (C57BL/6 Nude, 4- to 6-week-old female) were obtained from Pasteur Institute of Iran. Animal experiments were performed in the Department of Biology Specific-Pathogen-Free Animal Laboratory, University of Isfahan, and approved by the university's Ethics Committee on Animals Handling (Ethic number: IR.UI.REC.1396.056). Cages, beddings, water, and foods were sterilized and changed twice a week. After a 2-week delay for giving the animals enough time for adaptation, mice received intraperitoneal administration of 300 mg/kg cyclophosphamide. Seventy-two hours later, 15 × 10⁶ CCRF-CEM cells in 100 µL FBS were injected subcutaneously into the mice upper midback. Iron treatment began 3 days after transplantation by injecting 50 mg/kg iron sucrose (Venofer) intraperitoneally twice a week. In order to verify the engraftment, heart blood was analyzed by flow cytometry using antibodies against human CD7, CD3, CD5, and CD4. After 20 days, transplantation was confirmed and chemotherapy was started the day after treatment. 5 mg/kg MTX was inoculated intraperitoneally into each mouse, once a week, for 25 days. Mice were sacrificed by deep anesthesia using chloroform in an appropriate chamber, 2 months posttransplantation. Spleen, liver, brain, and bone marrow were harvested after death and were stained with H&E, according to the conventional technique.
2.12 Prussian blue staining
For the detection of iron storage levels in patients samples, Perl's Prussian blue staining was carried out. The air-dried bone marrow specimen was fixed in methanol for 30 s. After alcohol evaporation, staining with the ferrocyanide potassium solution (2 g per 40 mL 3.6% hydrochloric acid) was performed at room temperature for 30 minutes. Slides were counterstained with neutral red (100 mg dissolved in 100 mL ethanol 50%) for 10 minutes and rinsed in tap water. The mounted slides were inspected under a light microscope. Iron grading was performed using the conventional Gale's method. Murine spleen and liver tissues were fixed in 10% formalin, and then embedded in paraffin. Subsequently, the tissue sections were deparaffinized, hydrated and Prussian blue staining was carried out. Finally, the sections were dehydrated, cleared in xylene and mounted in synthetic resin.
2.13 Statistical analysis
All data were presented as the mean ± SEM and the probability value of P < .05 was considered to be significant. Differences between the groups were evaluated by using unpaired and two-tailed t tests. The correlation between the patients’ bone marrow iron stores and MRP1 expression levels was determined by chi-square test. Results were statistically analyzed using Graph Pad Prism 7.0.
3 RESULTS
3.1 Protective effect of iron in response to MTX
A positive correlation was previously demonstrated between the bone marrow iron stores of ALL patients and poor response to treatment. Consequently, we hypothesized that iron participates in drug resistance, and iron overload can be considered a risk factor for relapse. To elucidate the in vitro role of iron in response to therapy, MTX was considered as a key chemotherapy agent in leukemia treatment, and the impact of FAC on MTX-treated CCRF-CEM and Nalm6 cells was assessed. Iron-loaded cells were established by 24 hours pretreatment with FAC. Cells were exposed to the IC50 concentrations of MTX (0.5 and 1 µmol/L) and incubated for 72 and 96 hours, respectively. The difference between the incubation times and the IC50 concentration of MTX for the two cell lines was due to their diverse doubling times and different sensitivities to MTX (data not shown). 400 and 1600 µmol/L FAC showed maximum protection from MTX for CCRF-CEM and Nalm6 cell lines, respectively (Figure 1A-1,B-1).
Statistical analyses were performed for evaluating the protecting impact of 400 and 1600 µmol/L FAC on CCRF-CEM and Nalm6 cells, respectively, in the presence of MTX (Figure 1A-2,B-2). Results showed considerable resistance...
FIGURE 1 The pretreatment of cells with FAC for 24 h protected cells from MTX cytotoxicity. (A-1, 2) The CCRF-CEM cell line was treated with increasing concentrations of FAC (0-6400 µmol/L) for 24 h. Cells were washed twice with PBS, then seeded and incubated with the approximate IC50 concentration of MTX (0.5 µmol/L) for 72 h. Cell viability was then assessed using MTT assay. Statistical analysis for 400 µmol/L FAC showed maximum protection from MTX. (B-1, 2) The Nalm6 cell line was treated with increasing concentrations of FAC (0-6400 µmol/L) for 24 h. Cells were washed twice with PBS, then seeded and incubated with the IC50 concentration of MTX (1 µmol/L) for 96 h. Cell viability was then assessed using MTT assay. Statistical analysis for 1600 µmol/L FAC showed maximum protection from MTX. (C-1) CCRF-CEM and (C-2) Nalm6 cells were treated with FAC for 24 h and lysed by 65% HNO3. The intracellular iron content was then measured per 10^6 cells using atomic absorption flame emission spectrophotometry (AAS). Results showed a significant increase in intracellular iron upon cells exposure to FAC. Values are mean ± SEM of five separate experiments in triplicates, **P < .01 and ****P < .0001
to MTX in iron-loaded cells compared with control cells without iron treatment (67.34 ± 5.83% vs 44.62 ± 4.25% [mean ± SEM; n = 5], \( P < .01 \) for CCRF-CEM cells and 102.20 ± 3.49% vs 40.96 ± 7.41% [mean ± SEM; n = 5], \( P < .0001 \) for Nalm6 cells).
The cellular iron uptake was confirmed by atomic absorption flame emission spectrophotometry prior to MTT assays (Figure 1C).
Further experiments were conducted on CCRF-CEM cell line since the adaptive resistance to MTX in these cells was generated by lower concentrations of FAC than that of Nalm6.
### 3.2 Colony formation and apoptosis assays
The colony formation ability of 24 hours iron-loaded CCRF-CEM cells was studied, while cells were treated with MTX, using soft agar assay. Results showed an increased colony formation efficacy (CFE) of 24.09% for iron-loaded cells compared with those treated with MTX alone (mean ± SEM; \( n = 2 \), \( P = .016 \)) (Figure 2A).
Apoptosis assay was conducted on 24 hours FAC pre-treated CCRF-CEM cells followed by 72 hours exposure to MTX. It was demonstrated that the rate of apoptosis was decreased in iron-loaded cells compared with the FAC-ununtreated samples (33.03% vs 41.88%, respectively). In other words, the rate of viability was increased by 14.10% in those cells (mean ± SEM; \( n = 2 \), \( P = .015 \)) (Figure 2B1,2). Data were shown as percent viability of the untreated control cells.
### 3.3 Assessment of the ROS levels in iron-loaded cells
It is widely known that iron produces ROS via Fenton reaction.\(^{21}\) The generated ROS levels were particularly higher in iron-loaded CCRF-CEM cells than their relative controls (155.27 ± 6.36% vs 100% [mean ± SEM; \( n = 2 \), \( P < .001 \)) (Figure 3A,B). This effect was reversed by the addition of deferasirox, an iron chelator, or N-acetyl cysteine as a ROS scavenger (Figure 3C). Since the impact of N-acetyl cysteine to reverse the defensive effect of iron against MTX was greater than that of deferasirox, it was suggested that iron might exert its protective effect indirectly, by the generation of intracellular ROS.
### 3.4 Evaluation of the gene expression profiles involved in iron-induced ROS using real-time PCR
To unravel the mechanism through which iron may cause resistance to MTX, the impact of FAC on the expression profile of two categories of genes, antioxidants and survival/proliferation-related genes was assessed using real-time PCR in CCRF-CEM cells. It was shown that among the aforementioned genes, the expression levels of SOD2, NRF2, CTNNB1 (\( \beta \)-catenin), IL6, and BCL2 were increased in the iron-loaded cells compared with the FAC-ununtreated controls (7.32 ± 0.77, 1.41 ± 0.03, 14.79 ± 2.63, 5.02 ± 0.79, and 0.85 ± 0.03 fold change, respectively) (Figure 4A). Moreover, the expression levels of NRF2 and BCL2 remained elevated when cells were incubated with MTX for 72 hours, highlighting the role of these genes in iron-induced resistance to MTX (2.25 ± 0.56, 3.78 ± 0.19) (Figure 4B). Furthermore, it was demonstrated that the expression level of MRPI, the NRF2-target gene, was increased followed by treatment with MTX by 5.95 ± 2.65 fold (mean ± SEM, \( P < .05 \)). These data may confirm the activation of the transcription factor NRF2 followed by its overexpression. Furthermore, results demonstrated that the activity of superoxide dismutase was increased by 7.02-fold while treating cells with 400 µmol/L FAC for 24 hours, then 0.5 µmol/L MTX for 72 hours (Figure 4C). These data showed both the transcriptional and, indirectly, translational overexpression of the SOD gene upon 24 hours exposure to FAC.
### 3.5 Ex vivo analyses
To validate the in vitro findings, the expression levels of MRPI were quantified in the mononuclear cells harvested from patients with ALL using real-time PCR. The rationale for selecting MRPI among other genes, besides its overexpression upon exposure to MTX, was the functional effect of this gene as a drug efflux pump and its correlation with NRF2, as its transcription factor. NRF2 was overexpressed while the cells were overloaded with iron in 24 hours and could have induced the overexpression of MRPI while cells were subsequently exposed to MTX for 72 hours. The cutoff point of the MRPI expression levels was defined as two. Interestingly, the mRNA expression levels of MRPI showed a positive association with the patients' bone marrow iron stores (Figure 5). Results showed that from the three samples with high expression levels of MRPI (>2) and iron grade > 3 (Table S1), all patients were low responders to chemotherapy (Table S1). Response to treatment was evaluated by assessing the MRD status of the ALL patients, 1 year after chemotherapy.
### 3.6 In vivo experiments
To further validate our findings, in vivo studies were conducted. After a 2-week delay for giving the animals enough time for adaptation, treatments were started as described in...
Iron treatment was initiated at day 3, followed by CCRF-CEM transplantation, during which 50 mg/kg of Venofer (iron sucrose) was injected twice a week. Iron storage was identified in the mice liver and spleen after three injections, on day 10, at the total iron dose of 150 mg/kg. However, to maintain an iron overload condition in mice, Venofer was persistently administrated up to 850 mg/kg. Perl’s Prussian blue staining confirmed the establishment of iron overload mice models (Figure 6C). To identify the in vivo effect of iron on MTX resistance, treatment was
Figure 6A. Iron treatment was initiated at day 3, followed by CCRF-CEM transplantation, during which 50 mg/kg of Venofer (iron sucrose) was injected twice a week. Iron storage was identified in the mice liver and spleen after three injections, on day 10, at the total iron dose of 150 mg/kg.
initiated the day after the leukemia engraftment was authenticated by flow cytometry (Figure 6B). MTX was injected intraperitoneally, once per week. Mice were sacrificed 2 mo/62 d after transplantation. Iron treatment decreased the survival rate of the ALL mice models by 8.40%. Results were supportive of our in vitro findings, especially the colony formation assay. Effect of iron on the MTX-treated leukemic mice showed that the CCRF-CEM transplanted iron-loaded mice treated with MTX (AIM) group were considerably less responsive to chemotherapy than those unexposed to iron (AM). Altered liver histopathology and decline in body weight were observed in the AIM group by 26.11% and 11%, respectively (Figure 6D,E-2). These findings strengthened the in vitro results regarding iron involvement in MTX resistance.
4 | DISCUSSION
Iron plays critical roles in cellular proliferation and metabolic activities.\(^9\) Several studies have reported its protective role, especially, against cancer.\(^{22-24}\) In contrast, it has been demonstrated that iron overload may be carcinogenic,\(^{10,25}\) contributing to altered cellular metabolism shown in tumors.\(^{8,12}\) Moreover, iron is suggested to play a role in drug resistance.\(^{14-16}\) Given the inconsistency and lack of adequate information regarding the role of iron in ALL, the impetus of this work was to evaluate the impact of this compound on the response of lymphoblastic, malignant cells to MTX and to unveil the molecular mechanism underlying this effect.
Iron was previously introduced as one of the responsible mediators for apoptosis, necrosis, and ferroptosis (iron-mediated programed cell death).\(^{26-30}\) Furthermore, an association between tumor chemo-resistance and altered intracellular iron content was formerly described.\(^{14,15}\) Iron chelation was shown to have antiproliferative effects in several cancers\(^ {31-33}\) and it was demonstrated that iron-chelating agents can reverse the resistance of cancer cells to chemotherapeutic agents.\(^ {34,35}\) Despite emerging interest in using iron chelators in combinational chemotherapy, it has been revealed that some of these reagents, irrespective of their harmful effects on normal cells,\(^ {36}\) exert their antitumor activity regardless of their chelating function.\(^ {37,38}\) Ultimately, the mechanism by which iron confers drug resistance to cancer cells has remained elusive. Growing efforts were made to unravel these mechanisms, particularly in breast cancer.\(^ {8,16,39}\) Considering the effect of iron in ALL, we previously reported a compelling positive correlation between the bone marrow iron stores (BMIS) of ALL patients and poor response to treatment.\(^ {17}\) The current work has introduced iron, for the first time, as a defender of lymphoblastic malignant cells against MTX cytotoxicity (Figures 1 and 2). The maximum relative adaptive
FIGURE 3 The protective effect of iron against MTX might be, indirectly, through the generation of intracellular ROS. A, ROS detection using 2′,7′-Dichloro-fluorescin diacetate (DCFH-DA). The CCRF-CEM cells (5 × 10^5 cells/500 µL per well) were treated with 400 µmol/L FAC, then incubated in FBS-free medium containing 40 µmol/L DCFH-DA for 15 minutes in dark followed by dilution with 500 µL ice-cold PBS. The fluorescence intensity, which was correlated with the amount of the generated oxidized form of the indicator (DCF) and the level of the intracellular ROS, was then measured using flow cytometry. H_2O_2 was used as a positive control. The iron-loaded cells showed a significant shift in fluorescence intensity to the right. B, FAC promoted the intracellular ROS levels up to 55.27 ± 6.36%, compared with the FAC-untreated cells. C, 15 minutes or 3 h pretreated cells with 20 µmol/L of deferasirox (DFX), an iron chelator, or 1000 µmol/L N-acetyl cysteine (NAC), an ROS scavenger, respectively, were exposed to 400 µmol/L FAC for 24 h. Then, iron-loaded cells were exposed to the IC50 concentration of MTX for 72 h. Cellular viability was evaluated by MTT assay. Both DFX and NAC countered the iron-induced resistance to MTX by 80.06 ± 0.99% and 109.01 ± 0.96%, respectively. Values are given as the mean ± SEM of two to three separate experiments in triplicates, *P < .05, **P < .01 and ***P < .001
resistance to MTX was induced by the application of 400 and 1600 µmol/L FAC to CCRF-CEM and Nalm6 cells, respectively. According to the review literature, different concentrations of FAC used for the generation of iron-loaded cell lines were 360 M for rat heart blood cells,40 30 µmol/L for Hep3B,41 and 200 µmol/L for MCF7 cells.42 Furthermore, the intracellular iron status of CCRF-CEM and Nalm6 was different (41.20 ppb vs 106.18 ppb per 1 × 10^6 cells, respectively) (Figure 1C). It is assumed that this can explain the contrasting sensitivity of these two cell lines to MTX (the IC50 concentrations of 0.5 and 1 µmol/L for CCRF-CEM and Nalm6, respectively). The intracellular availability of higher concentrations of iron in Nalm6 might be a determinant for its lower sensitivity to MTX. The contribution of different
**FIGURE 4** Iron-induced alterations in the mRNA expression profiles of some iron and ROS related genes. A. The expression pattern of the antioxidant and survival/proliferation-related genes was measured followed by 24 h treatment of CCRF-CEM cell line with 400 µmol/L FAC, and compared with those in the untreated cell line. B. The gene transcripts expression profile was assessed in the FAC-treated cells which were exposed to MTX for 72 h, and compared with that in the iron-loaded cells without MTX posttreatment. GAPDH1 was applied as the reference gene. The comparative 2^−ΔΔCT^ method was used for data analysis and the related groups were compared using t test. C. The activity of superoxide dismutase was increased by 7.02-fold in the iron-loaded cells exposed to 0.5 µmol/L MTX (FM) for 72 h when compared with MTX-treated cells (M). These data showed both the transcriptional and, indirectly, translational overexpression of the SOD gene upon 24 h exposure to FAC. All values are results of two different experiments in duplicates shown as mean ± SEM, *P < .05 and **P < .01
**FIGURE 5** The positive correlation between MRP1 expression and iron storage in bone marrow. The MRP1 mRNA expression level of the mononuclear cells, obtained from 18 ALL patients, was associated with their bone marrow iron stores. Real-time PCR was performed in two different experiments in duplicates to assess the patients cellular MRP1 expression levels. GAPDH1 was applied as the reference gene and the comparative 2^−ΔΔCT^ method was used for data analysis. Patients iron stores was graded by Perl's Prussian blue staining, as described in materials and methods. The comparative analysis between the MRP1 expression levels and iron storage was performed using the two-tailed chi-square test. Black bars display iron grade > 3, gray bars represent iron grade ≤ 3, BM, bone marrow, *P = .016
intracellular iron contents in dissimilar responses to H2O2 was reported in mouse lymphoma cell lines43; however, no publication concerning leukemia cells is available. Since the adaptive resistance to MTX in CCRF-CEM cells was generated by lower concentrations of FAC than that of Nalm6, the remaining experiments were conducted on this cell line.
In an attempt to unravel the underlying mechanism(s) contributing to the iron-mediated resistance against MTX, the present study demonstrated that BCL2 and IL-6 were overexpressed in the iron-loaded leukemic cells (Figure 4). The developed resistance to MTX and decreased apoptosis in the iron-loaded ALL cells, in line with the prolonged overexpression of BCL2, elicits a binary function for iron as an antianti-poptotic and antioxidant molecule in these cells. Moreover, regarding the partial effect of iron on apoptosis (Figure 2B), it is assumed that the antioxidant capacity of this element is more important. The overexpression of IL-6 in iron-loaded cells is consistent with Li and colleagues’ investigations into breast cancer.16 It is widely known that IL-6 can activate the STAT3 signaling pathway. STAT3 promotes cell survival and proliferation via upregulation of its target genes, BCL2 and c-myc.44 Although in the iron-loaded leukemic cells, STAT3 activator, IL-6, and its downstream target, BCL2, were both overexpressed, a slight increase in the mRNA expression level of STAT3 was detected. Western blot analysis is required to validate the implication of the STAT3 pathway in the resistance of ALL cells to MTX. Moreover, it is interesting to note that β-catenin mRNA levels were also increased in the iron-loaded ALL cell line. The positive regulatory role of iron for β-catenin is previously described in the colorectal cancer cell lines.45 Likewise, it is discovered that the STAT3 promoter possesses a functional element for TCF binding,46 through which β-Catenin may upregulate the expression of STAT3 in esophageal squamous cell carcinoma. A putative crosstalk between the aforementioned genes is illustrated in Figure 7.
The current work introduces iron, for the first time, as an ROS inducer in ALL cells. Data showed that the antioxidant ability of N-acetyl cysteine to reverse the defensive effect of iron against MTX was surprisingly greater than that of the iron chelator, deferasirox (Figure 3); implying that iron mainly confers MTX resistance through the induction of intracellular ROS. Support for this hypothesis is our results demonstrating that ROS levels were particularly higher in iron-loaded cells than their relative control cells. It is interesting to note that iron may promote the intracellular levels of ROS at a critical concentration. It means that the above-optimal concentrations of FAC do not increase ROS levels relatively (Figure S1). It is assumed that the extra amount of iron may be converted to the antioxidant ferritin inside the cell, thereby inhibiting the intracellular free radicals cytotoxicity. Published data showed that iron may indirectly induce the generation of ROS through the hydrogen peroxide participating in the Fenton reaction 21 (Figure 7). Subsequently, the current study illustrated that the pretreatment of CCRF-CEM cells with FAC and MTX develops upregulation of SOD2 and NRF2 genes (Figure 4). It is proposed that the iron-mediated ROS triggers resetting of the redox state through the overexpression of its two aforementioned key regulators. Although the noteworthy alteration in SOD levels has been described in malignant cells,37 its involvement in tumor progression and chemo-resistance is still controversial.48-52 Results demonstrated that the mRNA expression levels of SOD2 and its
 The in vivo experiments. A, Schematic illustration of the in vivo experiments. The timeline of treatments and alleged groups are displayed. C57BL/6 Nude mice received 300 mg/kg intraperitoneal cyclophosphamide, on day 0. 15 × 10⁶ CCRF-CEM cells were injected subcutaneously in the mice upper midback on day 3. Iron treatments were begun 3 d after transplantation by injecting 50 mg/kg Venofer (iron sucrose) intraperitoneally twice a week. The mice of indicated groups, AIM and AM, received 5 mg/kg intraperitoneal MTX, weekly, the day after the engraftment. Mice were sacrificed 2 mo after transplantation. The control group with no transplantation, is not shown. B, Verification of the leukemia engraftment in mice. A representative flow cytometry-based detection of the transplanted CCRF-CEM in the mouse heart blood, using antibodies against human CCRF-CEM surface markers. The identified human CD3, CD4, and CD7 markers (71.50%, 51.34%, and 46.21%, respectively) confirmed leukemia engraftment. C, The confirmed accumulation of iron in mouse tissues. Representative macroscopic views of spleen, liver, and bone marrow aspirate of the CCRF-CEM transplant mice, stained by Perl’s Prussian blue (×400). Blue dots were significantly increased in the iron sucrose injected tissues compared with those in the control mouse. D, A comparison between the body weights of MTX-treated iron-loaded (AIM) and control leukemic mice (AM) from the beginning of the chemotherapy treatment until the last day of experiment. E-1, Representative microscopic views of liver and bone sections of the abovementioned mice groups, stained with Hematoxylin and Eosin. The iron-loaded leukemic mice group treated with MTX (AIM) showed significant liver, but not brain damage compared with the control groups of leukemic mice unexposed to iron, but treated with MTX (AM). Significant histological modifications in brain and liver tissues are demonstrated in untreated leukemic mice (A) compared with the un-transplanted nonleukemic mice with no treatment (UT). E-2, For quantitative morphometry of liver sections, images were photographed from each stained section using a magnification of 40× objective lens. The images were transformed into RGB stack format via Image J software and the area percent with red staining equal to or greater than a defined threshold was computed with this software. By subtracting the total red-colored area from 100, the percentage of the damage in each tissue section was calculated. White spaces of the lumens of blood vessels and artifacts were excluded beforehand. Interestingly, results obtained from the blind interpretation of data were consistent with the scoring analysis performed by the Image J program. Three sections per mouse were evaluated. Data were reported as mean ± SEM; *P = .04. A, CCRF-CEM transplanted leukemic mice; AM, MTX-treated leukemic mice; AI, iron-loaded leukemic mice; AIM, iron-loaded leukemic mice, treated with MTX, UT, Un-transplanted mice with no treatment.
enzymatic activity (Figure 4) were increased in the iron-loaded cells when exposed to MTX. On the other hand, although several attempts have been made to identify the role of \( \text{NRF2} \) in drug resistance,\(^{53-55} \) our findings were the first data to illustrate the overexpression of this gene in the iron-loaded ALL cells. NRF2 is a key regulator for both the antioxidant protection and detoxification of cells. Considering the detoxification role of NRF2, it was shown that \( \text{MRP1} \), the NRF2 target gene and drug efflux pump, is overexpressed in the MTX-resistant iron-loaded cell lines as well as the ALL patients' primary cells with high storage of bone marrow iron (Figures 4 and 5). More interestingly, these patients were all poor responders to chemotherapy. Investigating larger populations of ALL patients in prospective cohort studies may help intensify the validity of results provided in this study.
To examine the in vivo relevance of our findings regarding the effect of iron on response to MTX, iron-loaded leukemic mice were established (Figure 6C). Results showed that despite the acceptable efficacy of MTX treatment in leukemic mice, iron-loaded animals had notably altered liver histopathology and decline in body weight compared with the leukemic mice unexposed to iron (Figure 6). These data supported the contribution of iron to the poor response to MTX.
Taken together, we introduced iron, for the first time, as a responsible compound with critical implications in the resistance of ALL malignant cells to MTX. ROS acts as a secondary messenger for the iron-mediated resistance to MTX; and \( \text{BCL2}, \text{SOD2}, \) and \( \text{NRF2} \) genes are the effector mediators in this scenario. The comprehensive functions of iron in ALL drug resistance have yet to be understood. Additional experiments of RNA interference to knock
---
**Figure 7** Iron a tiny molecule with huge effects. The schematic diagram demonstrating at least two pathways by which iron may confer resistance to leukemic cells. First, iron induces moderate levels of ROS, shown in gray, through the Fenton reaction mediated by hydrogen peroxide \((\text{Fe}^{2+} + \text{H}_2\text{O}_2 \rightarrow \text{Fe}^{3+} + \text{OH}^- + \cdot \text{OH})\). ROS, then, may trigger the antioxidant defense system via upregulation of \( \text{BCL2} \) and \( \text{NRF2} \). Thereby, the malignant cell can survive against the toxic levels of MTX-induced ROS (demonstrated in red). At the same time, the antiapoptotic activity of \( \text{BCL2} \) may inhibit the MTX cytotoxic effects. Upregulation of \( \text{BCL2} \) may also be mediated by iron itself through the overexpression of \( \text{STAT3} \), possibly via IL-6 and \( \beta \)-catenin. Moreover, iron enhances \( \text{MRP1} \) expression indirectly via the activation of its transcription factor, NRF2. The MTX efflux through this ABC transporter protects the cell from MTX cytotoxicity. \( \rightarrow \), activation; \( \rightarrow \), inhibition/repression; \( \rightarrow \), possible/indirect activation; \( \leftrightarrow \), transport; \( \text{STAT3} \) activates transcription of the gene (shown by yellow rectangles); \( \text{MRP1} \), multidrug resistance-associated protein 1 (also known as \( \text{ABCC1} \)); \( \text{TCF} \), T-cell factor/lymphoid enhancing factors family of DNA-binding factors; \( \text{AREs} \), antioxidant response elements; RR, ribonucleotide reductase; MTX, methotrexate; ROS, reactive oxygen species.
down the selected mRNAs and Western blot analyses for protein studies must be performed, and further preclinical investigations are required in this burgeoning area of research. However, in an attempt to minimize the probability of drug resistance and improve the treatment outcomes in children with ALL, our findings suggest that the patients’ bone marrow iron stores must be assessed during the chemotherapy, and the number of blood transfusions should be carefully monitored.
ACKNOWLEDGMENTS
The authors thank Dr Jamal Moshtaghian for his generous help with animal experiments, Dr Ardesher Talebi for his insight and expertise with the analysis of the pathological data, Dr Alireza Moafi for kindly providing patients bone marrow samples, and all patients and their parents who participated in the present study.
CONFLICT OF INTEREST
We have no competing interests to declare.
AUTHOR CONTRIBUTIONS
Dr Rahgozar had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis and was involved in study design. Abedi was involved in acquisition of data and statistical analysis. Abedi, Rahgozar, and Esmaeili were involved in analysis and interpretation of data. Abedi and Rahgozar were involved in manuscript preparation.
DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.
ORCID
Soheila Rahgozar https://orcid.org/0000-0003-1376-255X
REFERENCES
1. Siegel R, Naishadham D, Jemal A. Cancer statistics, 2013. CA Cancer J Clin. 2013;63(1):11-30.
2. Mitchell C, Richards S, Harrison CJ, Eden T. Long-term follow-up of the United Kingdom medical research council protocols for childhood acute lymphoblastic leukaemia, 1980–2001. Leukemia. 2010;24(2):406.
3. Belson M, Kingsley B, Holmes A. Risk factors for acute leukemia in children: a review. Environ Health Perspect. 2006;115(1):138-145.
4. Raetz EA, Bhatla T. Where do we stand in the treatment of relapsed acute lymphoblastic leukemia? Hematology. 2012;12(1):129-136.
5. Leonard GD, Fojo T, Bates SE. The role of ABC transporters in clinical practice. Oncologist. 2003;8(5):411-424.
6. Bakos E, Homolya L. Portrait of multifaceted transporter, the multidrug resistance-associated protein 1 (MRP1/ABCC1). Pflug Arch Eur J Physiol. 2007;453(3):621-641.
7. Hoffmann EK, Lambert IH. Ion channels and transporters in the development of drug resistance in cancer cells. Philos Trans R Soc B Biol Sci. 2013;369(1638):20130109.
8. Miller LD, Coffman LG, Chou JW, et al. An iron regulatory gene signature predicts outcome in breast cancer. Cancer Res. 2011;71(21):6728-6737.
9. Rouault TA, Tong WH. Iron–sulfur cluster biogenesis and human disease. Trends Gener. 2008;24(8):398-407.
10. Toyokuni S. Role of iron in carcinogenesis: cancer as a ferrotoxic disease. Cancer Sci. 2009;100(1):9-16.
11. Norat T, Lukanova A, Ferrari P, Riboli E. Meat consumption and colorectal cancer risk: dose response meta-analysis of epidemiological studies. Int J Cancer. 2002;98(2):241-256.
12. Yang WC. Iron metabolism and leukemia. Adv Tech Biol Med. 2015;3:122.
13. Steegmann-Olmedillas JL. The role of iron in tumour cell proliferation. Clin Transl Oncol. 2011;13(2):71-76.
14. Epstein S, Glickstein H, Picard V, et al. H-ferritin subunit overexpression in erythroid cells reduces the oxidative stress response and induces multidrug resistance properties. Blood. 1999;94(10):3593-3603.
15. Fang D, Bao Y, Li X, et al. Effects of iron deprivation on multidrug resistance of leukemic K562 cells. Chemotherapy. 2010;56(1):9-16.
16. Li J, He K, Liu P, Xu LX. Iron participated in breast cancer chemoresistance by reinforcing IL-6 paracrine loop. Biochem Biophys Res Commun. 2016;475(2):154-160.
17. Moafi A, Ziaie M, Abedi M, et al. The relationship between iron bone marrow stores and response to treatment in pediatric acute lymphoblastic leukemia. Medicine. 2017;96(44):e8511.
18. Rahgozar S, Moafi A, Abedi M, et al. mRNA expression profile of multidrug-resistant genes in acute lymphoblastic leukemia of children, a prognostic value for ABCA3 and ABCA2. Cancer Biol Ther. 2014;15(15):35-41.
19. Beauchamp C, Fridovich I. Superoxide dismutase: improved assays and an assay applicable to acrylamide gels. Anal Biochem. 1971;44(1):276-287.
20. Gale E, Torrance J, Bothwell T. The quantitative estimation of total iron stores in human bone marrow. J Clin Invest. 1963;42(7):1076-1082.
21. Toyokuni S. Iron and thiols as two major players in carcinogenesis: friends or foes? Front Pharmacol. 2014;5:200.
22. Alarifi S, Ali D, Alkahtani S, Alhader M. Iron oxide nanoparticles induce oxidative stress, DNA damage, and caspase activation in the human breast cancer cell line. Cell Death Differ. 2014;15(1):416-424.
23. Kiessling MK, Klemke CD, Kamiński MM, Galani IE, Kramer PH, Gliwok K. Inhibition of constitutively activated nuclear factor-κB induces reactive oxygen species-and iron-dependent cell death in cutaneous T-cell lymphoma. Cancer Res. 2009;69(6):2365-2374.
24. Ferle-Vidović A, Poljak-Blaži M, Rapić V, Škare D. Ferrocenes. 2010;24(2):406.
25. Alarifi S, Ali D, Alkahtani S, Alhader M. Iron oxide nanoparticles induce oxidative stress, DNA damage, and caspase activation in the human breast cancer cell line. Cell Death Differ. 2014;15(1):416-424.
26. Kiersling MK, Klemke CD, Kamiński MM, Galani IE, Kramer PH, Gliwok K. Inhibition of constitutively activated nuclear factor-κB induces reactive oxygen species-and iron-dependent cell death in cutaneous T-cell lymphoma. Cancer Res. 2009;69(6):2365-2374.
27. Ferle-Vidović A, Poljak-Blaži M, Rapić V, Škare D. Ferrocenes. 2010;24(2):406.
28. Kiersling MK, Klemke CD, Kamiński MM, Galani IE, Kramer PH, Gliwok K. Inhibition of constitutively activated nuclear factor-κB induces reactive oxygen species-and iron-dependent cell death in cutaneous T-cell lymphoma. Cancer Res. 2009;69(6):2365-2374.
29. Ferle-Vidović A, Poljak-Blaži M, Rapić V, Škare D. Ferrocenes. 2010;24(2):406.
30. Kiersling MK, Klemke CD, Kamiński MM, Galani IE, Kramer PH, Gliwok K. Inhibition of constitutively activated nuclear factor-κB induces reactive oxygen species-and iron-dependent cell death in cutaneous T-cell lymphoma. Cancer Res. 2009;69(6):2365-2374.
31. Ferle-Vidović A, Poljak-Blaži M, Rapić V, Škare D. Ferrocenes. 2010;24(2):406.
32. Kiersling MK, Klemke CD, Kamiński MM, Galani IE, Kramer PH, Gliwok K. Inhibition of constitutively activated nuclear factor-κB induces reactive oxygen species-and iron-dependent cell death in cutaneous T-cell lymphoma. Cancer Res. 2009;69(6):2365-2374.
33. Ferle-Vidović A, Poljak-Blaži M, Rapić V, Škare D. Ferrocenes. 2010;24(2):406.
28. Munoz JP, Chiong M, Garcia L, et al. Iron induces protection and necrosis in cultured cardiomyocytes: Role of reactive oxygen species and nitric oxide. *Free Radic Biol Med*. 2010;48(4):526-534.
29. Lu B, Chen XB, Ying MD, He QJ, Cao J, Yang B. The role of ferroptosis in cancer development and treatment response. *Front Pharmacol*. 2018;9:892.
30. Salvador GA, Uranga RM, Giusto NM. Iron and mechanisms of neurotoxicity. *Int J Alzheimers Dis*. 2011;2011:1-9.
31. Foa P, Maiolo A, Lombard L, Villa L, Polli E. Inhibition of proliferation of human leukemic cell populations by deferoxamine. *Scand J Haematol*. 1986;36(1):107-110.
32. Blatt J, Stitely S. Antineuroblastoma activity of deferoxamine in leukemia and myeloma cells. *Blood*. 2009;114(14):3064-3073.
33. Callens C, Coulon S, Naudin J, et al. Targeting iron homeostasis induces cellular differentiation and synergizes with differentiating agents in acute myeloid leukemia. *J Exp Med*. 2010;207(4):731-750.
34. Eberhard Y, McDermott SP, Wang X, et al. Chelation of intracellular iron with the antifungal agent ciclopirox olamine induces cell death in leukemia and myeloma cells. *Blood*. 2009;114(14):3064-3073.
35. Whitnall M, Howard J, Ponka P, Richardson DR. A class of iron chelators with a wide spectrum of potent antitumor activity that overcomes resistance to chemotherapeutics. *Proc Natl Acad Sci USA*. 2006;103(40):14901-14906.
36. Benadiba J, Rosilio C, Nebout M, et al. Iron chelation: an adjuvant therapy to target metabolism, growth and survival of murine PTEN-deficient T lymphoma and human T lymphoblastic leukemia/lymphoma. *Leuk Lymphoma*. 2017;58(6):1433-1445.
37. Lee D-H, Jang PS, Chung NG, Cho B, Jeong DC, Kim HK. Deferasirox shows in vitro and in vivo antileukemic effects on murine leukemic cell lines regardless of iron status. *Exp Hematol*. 2013;41(6):539-546.
38. Messa E, Carturan S, Maffè C, et al. Deferasirox is a powerful NF-xB inhibitor in myelodysplastic cells and in leukemia cell lines acting independently from cell iron deprivation by chelation and reactive oxygen species scavenging. *Haematologica*. 2010;95(8):1308-1316.
39. Chekhun VF, Lukyanova N, Burlaka AP, et al. Iron metabolism disturbances in the MCF-7 human breast cancer cells with acquired resistance to doxorubicin and cisplatin. *Int J Oncol*. 2013;43(5):1481-1486.
40. Link G, Saada A, Pinson A, Konijn AM, Hershko C. Mitochondrial respiratory enzymes are a major target of iron toxicity in rat heart cells. *J Lab Clin Med*. 1998;131(5):466-474.
41. Pollak Y, Mechlovich D, Amit T, et al. Effects of novel neuroprotective and neurorestorative multifunctional drugs on iron chelation and glucose metabolism. *J Neural Transm*. 2013;120(1):37-48.
42. Lemler DJ, Lynch ML, Tesfay L, et al. DCYTB is a predictor of outcome in breast cancer that functions via iron-independent mechanisms. *Breast Cancer Res*. 2017;19(1):25.
43. Lipinski P, Drapier J-C, Oliveira L, Retmanska H, Sochanowicz B, Kruszewski M. Intracellular iron status as a hallmark of mammalian cell susceptibility to oxidative stress: a study of L5178Y mouse lymphoma cell lines differentially sensitive to H2O2. *Blood*. 2000;95(9):2960-2966.
44. Lui GY, Kovacevic Z, Richardson V, Merlot AM, Kalinowski DS, Richardson DR. Targeting cancer by binding iron: dissecting cellular signaling pathways. *Oncotarget*. 2015;6(22):18748.
45. Brookes MJ, Boult J, Roberts K, et al. A role for iron in Wnt signaling. *Oncogene*. 2008;27(7):966.
46. Yan S, Zhou C, Zhang W, et al. β-Catenin/TCF pathway upregulates STAT3 expression in human esophageal squamous cell carcinoma. *Cancer Lett*. 2008;271(1):85-97.
47. Oberley TD, Oberley LW. Antioxidant enzyme levels in cancer. *Histol Histopathol*. 1997;12(2):525-535.
48. Sentürker S, Karahalil B, Inal M, et al. Oxidative DNA base damage and antioxidant enzyme levels in childhood acute lymphoblastic leukemia. *FEBS Lett*. 1997;416(3):286-290.
49. Battisti V, Maders LDK, Bagatini MD, et al. Measurement of oxidative stress and antioxidant status in acute lymphoblastic leukemia patients. *Clin Biochem*. 2008;41(7-8):511-518.
50. Chang B, Yang H, Jiao Y, et al. SOD2 deregulation enhances migration, invasion and has poor prognosis in salivary adenoid cystic carcinoma. *Sci Rep*. 2016;6:25918.
51. Yeung BHY, Wong KY, Lin MC, et al. Chemosensitisation by manganese superoxide dismutase inhibition is caspase-9 dependent and involves extracellular signal-regulated kinase 1/2. *Br J Cancer*. 2008;99(2):283.
52. Wilkes J, Alexander M, Cullen J. Superoxide dismutases in pancreatic cancer. *Antioxidants*. 2017;6(3):66.
53. Rushworth SA, MacEwan DJ. The role of nrf2 and cytoprotection in regulating chemotherapy resistance of human leukemia cells. *Cancers*. 2011;3(2):1605-1621.
54. Wu RP, Hayashi T, Cottam HB, et al. Nrf2 responses and the therapeutic selectivity of electrophilic compounds in chronic lymphocytic leukemia. *Proc Natl Acad Sci USA*. 2010;107(16):7479-7484.
55. Khamari R, Trinh A, Gabert PE, et al. Glucose metabolism and NRF2 coordinate the antioxidant response in melanoma resistant to MAPK inhibitors. *Cell Death Dis*. 2018;9(3):325.
SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section.
How to cite this article: Abedi M, Rahgozar S, Esmaeil A. Iron protects childhood acute lymphoblastic leukemia cells from methotrexate cytotoxicity. *Cancer Med*. 2020;9:3537–3550. [https://doi.org/10.1002/cam4.2982] | 2025-03-04T00:00:00 | olmocr | {
"pdf_page_numbers": [
[
0,
3105,
1
],
[
3105,
8302,
2
],
[
8302,
12157,
3
],
[
12157,
17099,
4
],
[
17099,
18305,
5
],
[
18305,
23232,
6
],
[
23232,
24091,
7
],
[
24091,
28404,
8
],
[
28404,
31114,
9
],
[
31114,
31114,
10
],
[
31114,
37864,
11
],
[
37864,
41388,
12
],
[
41388,
48173,
13
],
[
48173,
53643,
14
]
]
} |
7cf50a96a7801106bcffdec99c403fab3503844c | {
"olmocr-version": "0.1.59",
"pdf-total-pages": 10,
"total-fallback-pages": 0,
"total-input-tokens": 27318,
"total-output-tokens": 6162,
"fieldofstudy": [
"Mathematics"
]
} | Semicontinuity of the Automorphism Groups of Domains with Rough Boundary
Steven G. Krantz
Abstract: Based on some ideas of Greene and Krantz, we study the semicontinuity of automorphism groups of domains in one and several complex variables. We show that semicontinuity fails for domains in $\mathbb{C}^n$, $n > 1$, with Lipschitz boundary, but it holds for domains in $\mathbb{C}^1$ with Lipschitz boundary. Using the same ideas, we develop some other concepts related to mappings of Lipschitz domains. These include Bergman curvature, stability properties for the Bergman kernel, and also some ideas about equivariant embeddings.
1 Introduction
A domain in $\mathbb{C}^n$ is a connected open set. If $\Omega$ is a domain, then we let Aut ($\Omega$) denote the group (under the binary operation of composition of mappings) of biholomorphic self-maps of $\Omega$. When $\Omega$ is a bounded domain, Aut ($\Omega$) is a real (never a complex) Lie group.
A notable theorem of Greene/Krantz [GRK2] says the following:
Theorem 1.1. Let $\Omega_0$ be a smoothly bounded, strongly pseudoconvex domain with defining function $\rho_0$ (see [KRA1] for the concept of defining function). There is an $\epsilon > 0$ so that, if $\rho$ is a defining function for a smoothly bounded, strongly pseudoconvex domain $\Omega$ with $\|\rho_0 - \rho\|_{C^k} < \epsilon$ (some large $k$) then the automorphism group of $\Omega$ is a subgroup of the automorphism group of $\Omega_0$. Furthermore, there is a diffeomorphism $\Phi : \Omega \rightarrow \Omega_0$ such that the mapping
$$\text{Aut} (\Omega) \ni \varphi \longmapsto \Phi \circ \varphi \circ \Phi^{-1}$$
is an injective group homomorphism of Aut ($\Omega$) into Aut ($\Omega_0$).
---
1Subject Classification Numbers: 32M05, 32M17, 32M25, 30F10.
2Key Words: several complex variables, one complex variable, automorphism group, Lipschitz boundary.
In what follows we shall refer to this result as the “semicontinuity theorem.”
It should be noted that, although this theorem was originally proved for strongly pseudoconvex domains in $\mathbb{C}^n$, the very same proof shows that the result is true in $\mathbb{C}^1$ for any smoothly bounded domain $\Omega_0$. In fact the proof, while parallel to the original proof in [GRK2], is considerably simpler in the one-dimensional context.
The original proof of this result, which was rather complicated, used stability results for the Bergman kernel and metric established in [GRK1] and also the idea of Bergman representative coordinates. An alternative approach, using normal families, was developed in [KIM]. The paper [GRK3] produced a method for deriving a semicontinuity theorem when the domain boundaries are only $C^2$. The more recent work [GKKS] gives a new and more powerful approach to this matter of reduced boundary smoothness. The paper [KRA2] gives yet another approach to the matter, and proves a result for finite type domains.
It is geometrically natural to wonder whether there is a semicontinuity theorem when the boundary has smoothness of degree less than 2. On the one hand, experience in geometric analysis suggests that $C^2$ is a natural cutoff for many positive results (see [KRP]). On the other hand, Lipschitz boundary is very natural from the point of view of dilation and other geometric operations.
The purpose of this paper is to show that the semicontinuity theorem fails for domains in $\mathbb{C}^n$, $n > 1$, with Lipschitz boundary. But it holds for domains in $\mathbb{C}^1$ with Lipschitz boundary. The reason for this difference is connected, at least implicitly, with the failure of the Riemann mapping theorem in several complex variables. We shall explain this point in more detail as the presentation develops.
2 The Several-Complex-Variable Situation
The main result of this section is the following:
**Theorem 2.1.** Let $n > 1$ and consider domains in $\mathbb{C}^n$. There is a sequence $\Omega_j$ of strongly pseudoconvex domains with Lipschitz boundary and another domain $\Omega$ with Lipschitz boundary so that $\Omega_j \rightarrow \Omega$ in the Lipschitz topology on defining functions and so that
For each \( j \), \( \text{Aut}(\Omega_j) = \mathbb{Z} \);
\( \text{(b)} \) \( \text{Aut}(\Omega) = \{ \text{id} \} \).
See [HEL] for a consideration of strongly pseudoconvex domains with less than \( C^2 \) boundary. This result shows that the semicontinuity theorem fails for domains with Lipschitz boundary.
It should be understood that all the domains considered in this paper have finite connectivity. In particular, the complement of the domain only has finitely many components. And each component of the complement has Lipschitz boundary. We do not allow boundary components that are a single point. Each boundary component is the closure of an open set.
We shall use some ideas in [LER] in constructing the example enunciated in the theorem. We shall make our construction in \( \mathbb{C}^2 \). But it is easy to produce analogous examples in any \( \mathbb{C}^n \).
**Proof of the Theorem:** Let \( \psi \in C_c^\infty(\mathbb{C}^n) \) be such that
\( \text{(i)} \) \( \text{supp} \psi \subseteq B(0, 1) \);
\( \text{(ii)} \) \( \psi \geq 0 \);
\( \text{(iii)} \) \( \psi(0) = 1 \).
We will build our domains by modifying the unit ball \( B \) in \( \mathbb{C}^2 \). We will make particular use of these automorphisms of the unit ball, for \( a \) a complex number of modulus less than 1:
\[
\Psi_a(z_1, z_2) = \left( \frac{z_1 - a}{1 - \overline{a}z_1}, \frac{\sqrt{1 - |a|^2}z_2}{1 - \overline{a}z_1} \right).
\]
See [RUD].
We define
\[
\eta_1(z_1, z_2) = -1 + |z_1|^2 + |z_2|^2 - \left( \frac{1}{10} \psi \left( 10\left( (z_1, z_2) - (\sqrt{3/4}, 1/2) \right) \right) \right).
\]
Set
\[ U_1 = \{ (z_1, z_2) \in \mathbb{C}^2 : \eta_1(z_1, z_2) < 0 \} \].
Clearly \( U_1 \) is a domain with smooth boundary. It is a ball with a “bump” attached at the point \( (\sqrt{3/4}, 1/2) \).
Now define
\[ \Omega_1 = \bigcup_{j=-\infty}^{\infty} \Psi_1^{2j}(U_1) . \]
We see that \( \Omega_1 \) has infinitely many bumps which accumulate at the points \((1, 0)\) and \((-1, 0)\). It is because of those accumulation points that the boundary of \( \Omega_1 \) is only Lipschitz.
In general we let, for \( k \geq 2 \),
\[ \eta_k(z_1, z_2) = -1 + |z_1|^2 + |z_2|^2 - (1/10^k)\psi\left(10^k((z_1, z_2) - (\sqrt{(1/2)^k - 1/2}, 1 - (1/2)^k))\right) . \]
Set
\[ U_k = \{(z_1, z_2) \in \mathbb{C}^2 : \eta_k(z_1, z_2) < 0\} . \]
Clearly \( U_k \) is a domain with smooth boundary. It is a ball with a “bump” attached at the point \((\sqrt{(1/2)^k - 1/2}, 1 - (1/2)^k)\).
Now define, for \( k \geq 2 \),
\[ \Omega_k = \Omega_{k-1} \cup \bigcup_{j=-\infty}^{\infty} \Psi_1^{2^j+k-1}(U_k) . \]
We see that \( \Omega_k \) has infinitely many bumps which accumulate at the points \((1, 0)\) and \((-1, 0)\). It is because of those accumulation points that the boundary of \( \Omega_1 \) is only Lipschitz.
Finally we let
\[ \Omega = \bigcup_{k=1}^{\infty} \Omega_k . \]
Now it is clear that \( \Omega_k \to \Omega \) in the Lipschitz topology on defining functions. Furthermore, the ideas in [LER] show that the automorphism group of \( \Omega_k \) consists precisely of the mappings \( \Psi_1^{2^j+k-1}, j \in \mathbb{Z} \). So the automorphism group of \( \Omega_k \) is canonically isomorphic to \( \mathbb{Z} \). But it is also clear that the automorphism group of \( \Omega \) consists of the identity alone.
That completes the construction described in the theorem.
3 The One-Variable Situation
The one-variable result is this:
**Theorem 3.1.** Consider domains in $\mathbb{C}^1$. Let $\Omega_0 \subseteq \mathbb{C}^1$ be a bounded domain with Lipschitz boundary and defining function $\rho_0$. If $\epsilon > 0$ is sufficiently small then, whenever $\Omega$ is a bounded domain with Lipschitz boundary and defining function $\rho$ satisfying $\|\rho_0 - \rho\|_{\text{Lip}} < \epsilon$ then the automorphism group of $\Omega$ is a subgroup of the automorphism group of $\Omega_0$. Moreover, there is a diffeomorphism $\Phi : \Omega \to \Omega_0$ so that the mapping
$$\text{Aut}(\Omega) \ni \varphi \mapsto \Phi \circ \varphi \circ \Phi^{-1} \in \text{Aut}(\Omega_0)$$
is an injective group homomorphism.
We see here that the situation is in marked contrast to that for several complex variables. Our proof of this result will rely on uniformization for planar domains, a result which has no analogue in several complex variables.
**Proof of the Theorem:** Fix the domain $\Omega_0$ and let $\Omega$ be of distance $\epsilon$ from $\Omega_0$ in the Lipschitz topology.
It is a standard result of classical function theory that a finitely connected domain in the plane, with no component of the complement equal to a point, is conformally equivalent to the plane with finitely many nontrivial closed discs excised—see [AHL] or [KRA3]. Call this conformal mapping the “normalization” of the domain. What is particularly nice about this result is that the proof is constructive and it is straightforward to see that the normalization of $\Omega$ is close to the normalization of $\Omega_0$ just because $\Omega$ is close to $\Omega_0$. Indeed the normalization of $\Omega$ will be close to that of $\Omega_0$ in the $C^2$ topology. Just because once it is close in the Lipschitz topology then it is automatically close in a smoother topology (because the boundary consists of finitely many nontrivial circles).
Thus we may apply the one-dimensional version of the semicontinuity theorem for $C^2$ boundary to see that the automorphism group of the normalization of $\Omega$ is a subgroup of the automorphism group of $\Omega_0$. And the diffeomorphism $\Phi$ exists as usual. Now we may use the normalizing conformal mapping to transfer this result back to the original domains $\Omega_0$ and $\Omega$.
That completes the proof. \qed
We note that another approach to constructing the normalization map is by way of Green’s functions. This method is also quite explicit and constructive. Stability results for elliptic boundary value problems are well known. So this again leads to a proof of the semicontinuity theorem by transference to the normalized domain.
4 Related Results in One Complex Dimension
Key to the work of Greene-Krantz in [GRK1] and [GKR2] is a stability result for the Bergman kernel. In that theorem, the authors consider a base domain Ω₀ and a “nearby” domain Ω. As usual, we define “nearby” in terms of closeness of the defining functions in a suitable topology. But it is useful to note that, in this circumstance, there is a diffeomorphism Π : Ω → Ω₀ which is close to the identity in a suitable $C^k$ topology. With this thought in mind, Greene and Krantz proved the following:
Theorem 4.1. Let Ω₀ be a fixed, smoothly bounded, strongly pseudoconvex domain. Let Ω be a domain which is “ε-close” to Ω₀ in a $C^k$ topology. Let Π be the mapping described in the preceding paragraph. If ε is small enough, then the Bergman kernel $K_Ω$ for Ω is close to $K_{Ω₀} \circ Π$ in the $C^m$ topology for some $0 < m < k$.
This result also holds in one complex dimension, and the proof in that context is actually much easier.
Our remark now is that this theorem is actually true in the Lipschitz topology. We use the argument of the last section. Namely, if Ω is close to Ω₀ in the Lipschitz topology, then the normalization of Ω is close to the normalization of Ω₀ in a smooth topology. This the one-dimensional version of Theorem 4.1 applies to the normalized domains. The result follows.
5 Equivariant Embeddings
A lovely result of Maskit [MAS] is the following:
Theorem 5.1. Let $\Omega \subseteq \mathbb{C}$ be any planar domain. Then there is a univalent, holomorphic embedding $\Phi : \Omega \to \mathbb{C}$ so that the automorphism group of the image domain $\Phi(\Omega)$ consists only of linear fractional transformations.
An elegant corollary of Maskit’s result is that if $\varphi$ is any automorphism of a planar domain that fixes three points then $\varphi$ is the identity mapping. This follows because it is clear that any linear fractional transformation that fixes three points is the identity.
We would like to remark here that the ideas in this paper give a “poor man’s version” of this theorem. For let $\Omega$ be any domain with Lipschitz boundary as we have been discussing. So each component of the complement is the closure of a region having Lipschitz boundary. Now the normalizing map sends this domain $\Omega$ to a planar domain bounded by finitely many disjoint circles. It is easy to see, using Schwarz reflection and Schwarz’s lemma, that any conformal self-map of such a domain must be linear fractional. So any such map that fixes three points must be the identity.
6 The Bun Wong/Rosay Theorem
A classical result in several complex variables is this (see [ROS], [WON]):
Theorem 6.1. Let $\Omega \subseteq \mathbb{C}^n$ be a bounded domain. Let $P \in \partial \Omega$ and assume that $\partial \Omega$ is strongly pseudconvex in a neighborhood of $P$. Suppose that there are a point $X \in \Omega$ and automorphisms $\varphi_j$ of $\Omega$ such that $\varphi_j(X) \to P$ as $j \to \infty$. Then $\Omega$ is biholomorphic to the unit ball.
In a similar spirit, Krantz [KRA4] proved the following result:
Theorem 6.2. Let $\Omega \subseteq \mathbb{C}$ be a bounded domain and let $P \in \partial \Omega$ have the property that $\partial \Omega$ near $P$ is a $C^1$ curve. Suppose that there are a point $X \in \Omega$ and automorphisms $\varphi_j$ of $\Omega$ such that $\varphi_j(X) \to P$ as $j \to \infty$. Then $\Omega$ is conformally equivalent to the unit disc.
In this section we will re-examine Theorem 6.2 in the context of this paper, that is in relation to finitely connected domains with Lipschitz boundary. As noted, such a domain is conformally equivalent to a domain $\hat{\Omega}$ whose boundary consists of finitely many circles. Now we have the following possibilities:
(a) If $\partial \hat{\Omega}$ consists of just one circle, then $\hat{\Omega}$ is the disc, and there is nothing to prove.
(b) If $\partial \hat{\Omega}$ consists of two circles, one inside the other, then $\hat{\Omega}$ is (conformally equivalent to) an annulus. Then the automorphism group of such a domain is two copies of the unit circle. In particular, it is compact. So the hypotheses of Theorem 6.2 cannot obtain.
(c) If $\partial \hat{\Omega}$ consists of two circles, neither of which is inside the other, then the domain is unbounded. The automorphism group of such a domain is compact, and the hypotheses of Theorem 6.2 do not apply.
(d) If $\partial \hat{\Omega}$ consists of at least three circles, with all the circles but one lying inside the other one, then it is well known (see [JUL] or [HEI2]) that the automorphism group of $\hat{\Omega}$ is finite. Then the hypotheses of Theorem 6.2 cannot obtain.
Thus we see by inspection that Theorem 6.2 is true in the context of the domains that we have been discussing in this paper.
7 Curvature of the Bergman Metric
It is a matter of considerable interest to know the curvature properties of the Bergman metric on a planar domain. In particular, negativity of the curvature near the boundary is a useful analytic tool (see [GRK1]). If $\Omega$ is a planar domain with Lipschitz boundary, then its normalized domain is bounded by finitely many circles. The asymptotic boundary behavior of the Bergman kernel on such a domain is very well understood—see [APF]. In particular, the kernel near a boundary point $P$ is asymptotically very much like the kernel for the disc. Thus a straightforward calculation confirms that the curvature of the Bergman metric near the boundary is negative. Of course this statement pulls back to the original domain in a natural way.
8 Closing Remarks
It is natural to want to consider the results presented here in either the $C^1$ topology or even the $C^{2-\epsilon}$ topology. At this time the techniques are not available to attack those questions.
In several complex variables, one would also like to prove semicontinuity theorems for broad classes of domains. This will be the subject for future papers.
REFERENCES
[APF] L. Apfel, Localization Properties and Boundary Behavior of the Bergman Kernel, thesis, Washington University in St. Louis, 2003.
[GKKS] R. E. Greene, K.-T. Kim, S. G. Krantz, and A.-R. Seo, Semi-continuity of automorphism groups of strongly pseudoconvex domains: the low differentiability case, preprint.
[GRK1] R. E. Greene and S. G. Krantz, Deformations of complex structure, estimates for the $\bar{\partial}$-equation, and stability of the Bergman kernel, Advances in Math. 43(1982), 1–86.
[GRK2] R. E. Greene and S. G. Krantz, The automorphism groups of strongly pseudoconvex domains, Math. Annalen 261(1982), 425-446.
[GRK3] R. E. Greene and S. G. Krantz, Normal families and the semicontinuity of isometry and automorphism groups, Math. Z. 190(1985), 455–467.
[HEI1] M. Heins, A note on a theorem of Radó concerning the $(1,m)$ conformal maps of a multiply connected region into itself, Bulletin of the AMS 47(1941), 128–130.
[HEI2] M. Heins, On the number of 1-1 directly conformal maps which a multiply-connected plane regions of finite connectivity $p(> 2)$ admits onto itself, Bulletin of the AMS 52(1946), 454–457.
[HEL] G. M. Henkin and J. Leiterer, Theory of Functions on Strictly Pseudoconvex Sets with Nonsmooth Boundary, with German and Russian summaries, Report MATH 1981, 2. Akademie der Wissenschaften der DDR, Institut für Mathematik, Berlin, 1981.
[HES] Z.-X. He and O. Schramm, Fixed points, Koebe uniformization and circle packings, Ann. of Math. 137(1993), 369–406.
[JUL] G. Julia, Leçons sur la représentation conforme des aires multiplement connexes, Paris, 1934.
[KIM] Y. W. Kim, Semicontinuity of compact group actions on compact differentiable manifolds, Arch. Math. 49(1987), 450–455.
[KRA1] S. G. Krantz, *Function Theory of Several Complex Variables*, 2nd ed., American Mathematical Society, Providence, RI, 2001.
[KRA2] S. G. Krantz, Convergence of automorphisms and semicontinuity of automorphism groups, *Real Analysis Exchange*, to appear.
[KRA3] S. G. Krantz, *Cornerstones of Geometric Function Theory: Explorations in Complex Analysis*, Birkhäuser Publishing, Boston, 2006.
[KRA4] S. G. Krantz, Characterizations of smooth domains in $\mathbb{C}$ by their biholomorphic self maps, *Am. Math. Monthly* 90(1983), 555–557.
[KRP] S. G. Krantz and H. R. Parks, *The Geometry of Domains in Space*, Birkhäuser Publishing, Boston, 1996.
[LER] L. Lempert and L. Rubel, An independence result in several complex variables, *Proc. Amer. Math. Soc.* 113(1991), 1055–1065.
[MAS] B. Maskit, The conformal group of a plane domain, *Amer. J. Math.*, 90 (1968), 718–722.
[ROS] J.-P. Rosay, Sur une characterization de la boule parmi les domaines de $\mathbb{C}^n$ par son groupe d’automorphismes, *Ann. Inst. Four. Grenoble* XXIX(1979), 91–97.
[RUD] W. Rudin, *Function Theory in the Unit Ball of $\mathbb{C}^n$*, Springer-Verlag, New York, 1980.
[WON] B. Wong, Characterizations of the ball in $\mathbb{C}^n$ by its automorphism group, *Invent. Math.* 41(1977), 253–257.
Steven G. Krantz
Department of Mathematics
Washington University in St. Louis
St. Louis, Missouri 63130
[email protected] | 2025-03-05T00:00:00 | olmocr | {
"pdf_page_numbers": [
[
0,
1896,
1
],
[
1896,
4155,
2
],
[
4155,
5965,
3
],
[
5965,
7541,
4
],
[
7541,
9916,
5
],
[
9916,
11670,
6
],
[
11670,
14159,
7
],
[
14159,
16246,
8
],
[
16246,
17990,
9
],
[
17990,
19400,
10
]
]
} |
ff84b25dc4d390f7e811e2fc42c09a8c72ab7550 | {
"olmocr-version": "0.1.59",
"pdf-total-pages": 19,
"total-fallback-pages": 0,
"total-input-tokens": 50551,
"total-output-tokens": 15800,
"fieldofstudy": [
"Agricultural And Food Sciences"
]
} | RESEARCH ARTICLE
Cannabis lighting: Decreasing blue photon fraction increases yield but efficacy is more important for cost effective production of cannabinoids
F. Mitchell Westmoreland*, Paul Kusuma, Bruce Bugbee
Crop Physiology Laboratory, Utah State University, Logan, Utah, United States of America
* [email protected]
Abstract
LED technology facilitates a range of spectral quality, which can be used to optimize photosynthesis, plant shape and secondary metabolism. We conducted three studies to investigate the effect of blue photon fraction on yield and quality of medical hemp. Conditions were varied among studies to evaluate potential interactions with environment, but all environmental conditions other than the blue photon fraction were maintained constant among the five-chambers in each study. The photosynthetic photon flux density (PPFD, 400 to 700 nm) was rigorously maintained at the set point among treatments in each study by raising the fixtures. The lowest fraction of blue photons was 4% from HPS, and increased to 9.8, 10.4, 16, and 20% from LEDs. There was a consistent, linear, 12% decrease in yield in each study as the fraction of blue photons increased from 4 to 20%. Dry flower yield ranged from 500 to 750 g m\(^{-2}\). This resulted in a photon conversion efficacy of 0.22 to 0.36 grams dry flower mass yield per mole of photons. Yield was higher at a PPFD of 900 than at 750 \(\mu\)mol m\(^{-2}\) s\(^{-1}\).
There was no effect of spectral quality on CBD or THC concentration. CBD and THC were 8% and 0.3% at harvest in trials one and two, and 12% and 0.5% in trial three. The CBD/THC ratio was about 25 to 1 in all treatments and studies. The efficacy of the fixtures ranged from 1.7 (HPS) to 2.5 \(\mu\)mol per joule (white+red LED). Yield under the white+red LED fixture (10.4% blue) was 4.6% lower than the HPS on a per unit area basis, but was 27% higher on a per dollar of electricity basis. These findings suggest that fixture efficacy and initial cost of the fixture are more important for return on investment than spectral distribution at high photon flux.
Introduction
Cannabis is a high value crop that can be profitably grown in controlled environments under sole-source electric lights [1], but the cost of electricity is a high fraction of overall production costs [2]. High-pressure sodium (HPS) lights are commonly used in Cannabis cultivation because they have a low upfront cost and high photon output. However, advances in light-
emitting diode (LED) technology [3, 4] has led to more efficient fixtures (watts of output per watt of input), but these fixtures vary in their efficacy (micromoles of photons per joule of energy input; \( \mu \text{mol J}^{-1} \)), depending on the choice of LED and drive current [5]. These differences in efficacy have a significant impact on energy use in controlled environment plant production. LED fixtures can be made with unique spectra that have the potential to increase flower yield and quality (cannabinoid profile) [6–8].
Spectral effects on photosynthesis have been studied for over 70 years (Hoover 1937). McCree [9, 10] and Inada [11] found that the quantum yield (moles of carbon fixed per mole of absorbed photon) of red (601 to 700 nm) photons was about 25% greater than blue (401 to 500 nm) photons and about 5% greater than green photons (501 to 600 nm). Yield photon flux (YPF) considers this photosynthetic response and weights photons from 300 to 800 nm according to their relative quantum yield [12, 13]. Although this spectral effect on quantum yield is well known, YPF has not been widely used to define photosynthetic photons. Instead, the widely used standard is the photosynthetic photon flux density (PPFD), defined as the number of photons within the waveband of photosynthetically active radiation (PAR; 400 to 700 nm) per square meter and second. PPFD assumes all photons in this range equally drive photosynthesis. Although this is not strictly correct, McCree [10] applied the equal weighting standard to the commercially available lighting technologies at the time (before LEDs were available) and concluded that this simpler PPFD definition adequately predicted quantum yield. At the time, spectroradiometers were slow, heavy, non-portable, and expensive, and the equal weighting definition could be used to make a lower cost quantum sensor.
The traditional definition of PAR includes photons between 400 and 700 nm, but recent studies now indicate that far-red photons (701 to 750 nm) are equally efficient at driving photosynthesis when coupled with shorter wavelengths [14, 15]. Far-red photons preferentially excite photosystem I in the photosynthetic electron transport chain, effectively releasing a bottleneck on the pool of reduced plastoquinone and photosystem II, thus allowing for a more rapid re-oxidation and more efficient photosynthesis [16].
As demonstrated by studies under monochromatic spectra or low PPFD—blue, green and red photons can alter plant morphology. Numerous studies have shown no effect of substituting green and red photons under a constant fraction of blue photons. For example, Son and Oh [17] and Kang et al. [18] demonstrated that substituting 10% red photons with green photons had no effect on leaf area or plant diameter in lettuce under 10, 20 or 30% blue. Snowden et al. [19] saw little to no effect on morphology in multiple species when increasing the fraction of green from 2 to 41% at about 11% blue. By contrast, multiple studies (including the studies that found no effect of substituting green and red photons) have shown a reduction in leaf area and dry mass gain with an increasing blue photon fraction from 5 to 30% [17–25].
Most spectral studies have investigated responses to blue photons at a PPFD of 500 \( \mu \text{mol m}^{-2} \text{s}^{-1} \) or less, but blue photon responses can interact with intensity [19, 21], so it is difficult to extrapolate the findings on blue photon response at low PPFD to intensities greater than 500 \( \mu \text{mol m}^{-2} \text{s}^{-1} \). This is particularly important with Cannabis, which has an increasing rate of photosynthesis up to a PPFD of 1500 \( \mu \text{mol m}^{-2} \text{s}^{-1} \) [26] and is increasingly being grown under high PPFD in commercial cultivation.
Far-red photons can also alter plant morphology by increasing stem elongation and leaf expansion, which typically increases radiation capture and thus yield [25, 27, 28]. The mechanism underlying these effects is well studied. Activation of the photoreceptor phytochrome (Pfr) inhibits the activity of a group of transcription factors called PHYTOCHROME INTER-ACTING FACTORS (PIFs). PIFs promote the expression of genes related to elongation, including auxin synthesis. Increased far-red fraction inactivates the phytochrome and releases the inhibition of PIFs, which promotes auxins and thus increases cell elongation [29].
Cannabinoids are a unique class of secondary metabolites that are synthesized by the genus *Cannabis*. Since Gaoni and Mechoulam [30, 31] first described the structure and synthesized tetrahydrocannabinol (THC) and cannabidiol (CBD) *in vitro*, more than 150 cannabinoids have been described [32, 33]. Cannabinoid synthesis occurs primarily in capitate-stalked glandular trichomes that are highly concentrated on the bracts of pistillate flowers [34]. The biosynthetic pathway *in vivo* has been extensively studied and is primarily under genetic control [32, 35–37], although environmental factors can influence the final cannabinoid concentrations [38]. Environmental effects on cannabinoid concentration are of significant interest as the distinction between marijuana (greater than 0.3% THC) and hemp (less than 0.3% THC) has become a legal concern and the demand increases for medical grade *Cannabis* with predictable and consistent cannabinoid profiles [39–41]. The beneficial effects of spectral quality on secondary metabolism has been well studied [42, 43], so the use of unique spectra to regulate cannabinoid biosynthesis in controlled environment *Cannabis* production warrants exploration.
Early work on *Cannabis* suggested that PPFD [44, 45], spectral quality [46], and photoperiod [47] can alter cannabinoid biosynthesis. Mahlberg and Hemphill [46] used Plexiglas filters to provide monochromatic light to greenhouse grown hemp and found that leaves of plants grown under a red filter contained 3 times higher cannabinoid concentration than plants grown under blue and green filters. They also reported a shift in the ratio of cannabinoids, in particular cannabichromene (CBC) and THC, under monochromatic filters compared to sunlight. However, there were differences in non-photosynthetic radiation among treatments, which make the results difficult to interpret. Nonetheless, this early study provided evidence that radiation changes can influence cannabinoid concentration.
LED technology has facilitated a more rigorous investigation into the effect of spectral quality on cannabinoid production [7]. Magagnini et al. [6] reported significant increases in flower yield among plants of a high THC *Cannabis* variety grown under mogul-base HPS (8% blue) compared to two LED fixtures with 14% and 24% blue, but plants grown under HPS had a lower cannabinoid concentration than the two LED treatments. Notably, the total amount of cannabinoids (cannabinoid yield) was not significantly different among the treatments. Namdar et al. [8] found significant differences in morphology and cannabinoid concentration among plants grown under LEDs, T5 fluorescent tubes, and HPS at different growth stages, but the PPFD ranged from 90 to 500 μmol m⁻² s⁻¹ among the three spectral treatments, making these results difficult to interpret. A PPFD of 90 μmol m⁻² s⁻¹ is only slightly above the light compensation point [26]. The results of Namdar [8] are likely caused by differences in photon flux rather than spectral quality, as intensity typically has a large impact on growth [1, 48].
Our objectives were twofold: 1) investigate the effect of spectral quality at a constant high PPFD on yield and quality of *Cannabis* in a controlled environment and 2) evaluate the effect of fixture efficacy on economic yield. We hypothesized that yield would increase as the fraction of blue photons decrease, but fixture efficacy would have a larger impact on the economics of indoor *Cannabis* cultivation.
**Materials and methods**
**Environmental conditions**
It is critical to maintain constant environmental conditions other than the variable being studied. To achieve constant conditions among treatments within a study, plants were grown in a walk-in growth room that had five 1 m² photon-independent sections and common atmospheric conditions (Fig 1).
Each independent section had white reflective walls to precisely define the growth area and simulate the presence of additional plants around the perimeter. Went [49] pioneered this approach at the CalTech Phytotron. Reflective side-walls are commonly used to minimize guard row effects and facilitate extrapolation to larger areas. They have been used in multiple studies of canopy photosynthesis [15, 50, 51]. In all studies, the canopy filled the entire chamber by about week 3.
Overhead fans provided continuous, ample air exchange of 0.8 meters per second at the top of the canopy, measured with a hot-wire anemometer (TSI, Inc. model 8330). The room was uniformly enriched to 1000 ± 50 ppm CO$_2$, continuously measured with an infra-red gas analyzer (LI-COR, Inc., model 850). Temperature was measured with a shielded, fan-aspirated, precision thermistor (Apogee Instruments Inc., model ST-100) mounted at canopy height in each chamber. Humidity was measured with a temperature and relative humidity sensor (Campbell Scientific Inc., model HMP45A).
Measurements were made every 10 seconds and five minute averages of all environmental data were recorded with a data logger (Campbell Scientific Inc., model CR1000). Fixtures were randomized among chambers in each trial. The environmental conditions of each trial are listed in Table 1.
### Spectral treatments
Spectral treatments included a double-ended high-pressure sodium fixture (DE-HPS), a warm white LED fixture (3000 K), a cool white LED fixture (5000 K) and two white+red combination LED fixtures (white+red 1 & white+red 2). These combinations were selected to achieve a relatively constant fraction of green and red photons while providing a five-fold difference in
blue photons (Table 2). These fixtures represent an industry standard fixture (HPS) and a low to high range of efficacies for LED fixtures [52]. The efficacy for DE-HPS was the average of the flat plane integration method and integrating sphere method described in Nelson and Bugbee [3]. The remaining fixtures were independently tested by TÜV SÜD America.
It is critical to maintain constant PPFD among spectral treatments within a study. PPFD at canopy height was kept constant throughout the study by raising the light fixtures as plants grew. PPFD measurements and fixture adjustments were made twice per week for the first four weeks of the flowering period, then once per week for the final four weeks. PPFD was measured using a full-spectrum quantum meter (Apogee Instruments Inc., model MQ-500). The highly reflective chamber walls and the fixture distance from the canopy minimized non-uniformity of lighting within a chamber. The PPFD in the corners of all chambers was within 15% of the average PPFD. Spectral traces of the five fixtures are shown in Fig 2. Measurements of HPS without glass mounted below the fixture. White+red 2 LED was switched with HPS without glass in trial three to quantify the effect of increased longwave radiation.
Yield photon flux (YPF) was normalized to HPS. Clear-glass was mounted below the HPS to reduce longwave radiation.
| Table 2. Efficacy, spectral distribution, and incoming radiation of the five spectral treatments. |
|---------------------------------------------------------------|-----------------|-----------------|-----------------|-----------------|
| Efficacy ($\mu$mol J$^{-1}$) | DE-HPS | 3000 K | White + Red 1 | White + Red 2 | 5000 K |
| % Blue (400–500 nm) | 1.72 | 2.13 | 2.51 | 2.40 | 2.43 |
| % Green (501–600 nm) | 4 | 10 | 10 | 16 | 20 |
| % Red (601–700 nm) | 43 | 39 | 41 | 40 | 49 |
| % Far Red (701–750 nm) | 53 | 51 | 49 | 44 | 31 |
| PPFD ($\mu$mol m$^{-2}$ s$^{-1}$) | 900 | 900 | 900 | 900 | 900 |
| YPF ($\mu$mol m$^{-2}$ s$^{-1}$) | 846 | 801 | 810 | 792 | 783 |
| Normalized YPF | 1 | 0.95 | 0.96 | 0.94 | 0.93 |
| Shortwave Radiation (280–4,000 nm; W m$^{-2}$) | 327 | 222 | 230 | 235 | 253 |
| Longwave Radiation (4,000–100,000 nm; W m$^{-2}$) | 474 | 461 | 454 | 455 | 453 |
*Yield photon flux (YPF) was normalized to HPS. Clear-glass was mounted below the HPS to reduce longwave radiation.
https://doi.org/10.1371/journal.pone.0248988.t002
Fig 2. Spectral traces of a 1000 W DE-HPS and four LED fixtures. Measurements were made at a PPFD of 900 $\mu$mol m$^{-2}$ s$^{-1}$.
https://doi.org/10.1371/journal.pone.0248988.g002
were made with a spectroradiometer (Apogee Instruments Inc., model PS-300). YPF was calculated using the weighting factors in Sager et al. (1988). The YPF (normalized to HPS) increased from 0.93 to 1.0 as the blue fraction decreased from 20% to 4% (Table 2).
HPS bulbs emit more longwave (thermal) radiation, causing higher leaf and flower bud temperature than LEDs at equal air temperature and PPFD [53]. In all studies, glass was mounted below an HPS fixture to make the longwave radiation similar to the LEDs (Table 2). Radiation measurements were made using a four-way net radiometer (Apogee Instruments Inc., model SN-500-SS). Canopy temperature was measured using a forward-looking infrared camera (FLIR Systems, model E6). In trial three, the white+red 2 LED fixture was replaced with an HPS without glass to quantify the effect of increased longwave radiation on canopy temperature and yield.
**Plant material**
Rooted cuttings of the medical hemp cultivar ‘Trump’ (T1) were transplanted into #2 plastic pots (6.3 L) filled with a soilless mix of 3:1 peat/vermiculite that was amended with 1.6 g per liter of dolomitic lime to increase pH to 5.5. Gypsum (CaSO$_4\cdot$2H$_2$O) was added at 0.8 g per liter to provide additional calcium and sulfur. The cultivar ‘Trump’ was selected because it has a compact growth habit and high cannabinoid concentrations. Plants were grown for 7 to 14 days (18 light:6 dark) and selected for uniformity before being switched to an inductive photoperiod (12 light:12 dark). Plant density was six plants per m$^2$ in trials one and three and two plants per m$^2$ in trial two. This facilitated testing a potential interaction of spectral quality with plant density and thus the rate of increase in photon capture.
Plants were irrigated daily to a 10% excess with a complete liquid fertilizer (Peter’s Peat-lite professional 20-10-20 [20N-4.4P-16.6K]) at a rate of 120 mg per liter N (26 mg/L P, 100 mg/L K, 1 mg/L Mg, 0.15 mg/L B, 0.15 mg/L Cu, 0.6 mg/L Fe, 0.3 mg/L Mn, 0.06 mg/L Mo, 0.3 mg/L Zn). Greencare micronutrients (Greencare Fertilizers, Inc.) were added at a rate of 7 mg per liter (0.12 mg/L B, 0.12 mg/L Cu, 0.49 mg/L Fe, 0.25 mg/L Mn, 0.05 mg/L Mo, 0.25 mg/L Zn). AgSil 16H (PQ Corporation) was added using a second proportioner for the liquid fertilizer at a rate of 8.4 mg/L Si (18 mg/L SiO$_2$; 34 mg/L H$_4$SiO$_4$; 0.3 mmol Si per liter). Electrical conductivity (EC) and pH of the nutrient solution were 1.2 ± 0.1 mS cm$^{-1}$ and 6.8 ± 0.1. Each pot had two 1 gallon-per-hour pressure compensating drip emitters (DIG, model B221B). Drip emitters were tested at the start and end of each trial to ensure uniformity.
**Plant measurements**
Root zone status was monitored throughout the study by measuring EC and pH of leachate. Changes to irrigation duration or frequency were made to maintain a leachate EC of about 1.0 mS cm$^{-1}$.
At approximately two weeks after the start of the inductive photoperiod and each week thereafter, flower samples were harvested from a subset of each spectral treatment to characterize the time course of cannabinoid accumulation. Flowers were sampled at a uniform height. Samples were air-dried for five days on well-ventilated shelves (25°C, 35% RH) to a moisture level of 10 ± 2% by weight.
At harvest, leaves, flowers, and stems were manually separated, weighed and dried. All fan leaves were separated from the flowers; the small (sugar) leaves that subtend the inflorescence were not separated from the flowers. After five days, dry mass was measured and a sample of the flowers from each plant was analyzed for final cannabinoid concentration. Trial two was terminated after seven weeks; trials one and three were terminated after eight weeks. The flower buds appeared physiologically mature at harvest in all trials.
Cannabinoid extraction
Dried flower samples were ground to a coarse powder using a stainless-steel grinder (KitchenAid, model BCG111OB). A 0.6-gram subsample of the dried flower material was transferred to an aluminum extraction cell (Q-Cup) containing a 40 μm cellulose filter base (C9) topped with a 0.25 μm membrane filter (M2). Flower samples within the cells were extracted with reagent-grade methanol (10 mL top volume, 5 mL bottom volume, 5 mL top rinse; Fisher Scientific) using an Energized Dispersive Guided Extraction (EDGE®) automated solvent extraction system (CEM Co.) [54]. The temperature in the Q-Cup was maintained at 100˚C with a hold time of 3 minutes. Samples were extracted twice to maximize recoveries. Blank cells were extracted between every sample.
Cannabinoid analysis
Sample extracts were analyzed using an Agilent 1100 series (Agilent Technologies, Santa Clara, CA) reverse phase high-performance liquid chromatography (HPLC) equipped with diode array detector and a Kinetex 2.6 μm C18 polar 100 Å, 100 x 4.6 mm column (Phenomenex Inc.). The mobile phase was a 70/30 ratio of 0.1% formic acid-acetonitrile and 0.1% formic acid-distilled water and the flow rate was 1 mL per min. A reference standard containing the cannabinoid compounds of interest (Cayman Chemical Inc., Phytocannabinoid Qualitative Mixture 10) was used to prepare the external calibration standards. CBDA was analyzed at 210 nm and had a retention time of 2.2 min; CBD was analyzed at 230 nm and had a retention time of 2.9 min; THC was analyzed at 285 nm and had a retention time of 4.0 min; and THCA was analyzed at 275 nm and had a retention time of 4.6 min. Minor cannabinoids such as CBG were excluded from the analysis due to low concentrations. Peaks were auto-integrated using ChemStation (Agilent Technologies). CBD and THC equivalents (CBD$_{eq}$ and THC$_{eq}$) are calculated using Eqs 1 and 2. These account for the conversion of the naturally produced acidic cannabinoids to the neutral form (decarboxylation). The 0.877 multiplier is the molecular weight ratio of the neutral form to the acidic form [55].
\[
CBD_{eq} = CBD + (CBDA \times 0.877)
\]
\[
THC_{eq} = THC + (THCA \times 0.877)
\]
Statistical analysis
The study was a randomized block design with the three studies in time as blocks. Environmental conditions and plant density were changed among studies to determine potential interactions of the blue photon treatment with other factors. This approach is similar to three replicate studies in the field over three growing seasons. To eliminate a potential chamber effect, treatments were randomly assigned to a chamber between trials. Each study included five fixed blue photon fraction levels. The community of plants in each chamber (bordered by the reflective walls) was treated as an experimental unit, rather than treating individual plants within a chamber as replicates. Yield was calculated as the total mass of dry flower per unit area.
The data were analyzed with a linear mixed model using the LME4 package in RStudio (R statistical software, version 1.2). Blue photon fraction was treated as a fixed effect and trial was treated as a random effect. One spectral treatment (white+red2) was exchanged in trial three to provide an HPS without glass treatment. This meant that HPS with glass and three LED treatments had three replicates in time, the white+red2 had two replicates in time, and the
HPS without glass was not replicated. The estimated trial variance on slope (blue photon fraction) was zero, which indicates a common slope for the three trials. Therefore, a random intercept regression model was used for further analysis in which intercept varied due to trials with a fixed blue photon fraction slope for all trials [56].
Results
Yield
As percent blue increased from 4 to 20%, flower yield decreased by 12.3%. This means that flower yield increased by 0.77% per 1% decrease in blue photons (Fig 3; p = 0.04). It is important to note that there was no significant interaction of blue photon fraction with trial (indicated by parallel slopes among the three trials). This indicates a consistent effect of percent blue across a range of yields and environments (Fig 3).
Harvest index (HI) is the ratio of usable biomass to total biomass, here defined as the ratio of dried flower to total dry above-ground mass (flowers, leaves and stems). Typical values for HI in crop plants range from 30 to 50% [57]. HI ranged from 55 to 65% among spectral treatments in the three trials, but there was no effect of blue photons on HI (S1 Fig; p = 0.91). There was no significant interaction with trial on harvest index.
The effect of blue photon fraction on height was not statistically significant (data not shown; p = 0.13).
Fig 3. The effect of blue photons on flower yield in three trials. Fixture names of each blue photon fraction are shown with arrows. The black line represents the linear mixed model with blue photon fraction as a fixed effect and trial as a random effect. The slopes of the lines for the three studies were parallel, indicating no interaction among studies. Yield increased by 0.77% per 1% decrease in blue photons (p = 0.04).
https://doi.org/10.1371/journal.pone.0248988.g003
Cannabinoid concentration
Blue photon fraction had no effect on final cannabinoid concentration (Fig 4A and 4B; CBD<sub>eq</sub> p = 0.32; THC<sub>eq</sub> p = 0.51).
In trial one, the ratio of CBD<sub>eq</sub> to THC<sub>eq</sub> rose to 34 to 1, but dropped to 25 to 1 at harvest (Fig 5G). In trials two and three the ratio was approximately 25 to 1 at each sampling point.
Fig 4. The effect of blue photons on (A) CBD<sub>eq</sub> and (B) THC<sub>eq</sub> concentration at harvest. The black line represents the linear mixed model with percent blue photons as a fixed effect and trial as a random effect. There was no significant effect of blue photon fraction on CBD<sub>eq</sub> (p = 0.32) or THC<sub>eq</sub> (p = 0.51) concentration at harvest.
https://doi.org/10.1371/journal.pone.0248988.g004
The final ratio of CBD<sub>eq</sub> to THC<sub>eq</sub> was 24 to 1 in all treatments and trials. In trial one, the THC<sub>eq</sub> concentration was 0.31 ± 0.03% at harvest while in trial three, the THC<sub>eq</sub> concentration increased to 0.52 ± 0.06%. In trial two, the concentration of THC<sub>eq</sub> rose to 0.46 ± 0.07% at about week five, but then dropped to 0.32 ± 0.03% at harvest (Fig 5D–5F).
In trial one, the concentration of CBD<sub>eq</sub> was about 7.6 ± 0.7% at harvest, while in trial three the concentration of CBD<sub>eq</sub> increased to 11.8 ± 1%. In trial two, the concentration of CBD<sub>eq</sub> rose to 11 ± 1.7% at week 5, but then dropped to 7.9 ± 0.7% at harvest (Fig 5A–5C).
**Efficacy**
An advantage of LED fixtures is their high efficacy and thus low electrical (operating) costs compared to HPS. At an energy cost of $0.10 USD per kWh, using the efficacies listed in Table 2, it costs 1.6 cents per mol of photons from an HPS fixture and 1.1 to 1.3 cents per mol for the LEDs (S2 Fig).
Photon conversion efficacy (PCE) is calculated by dividing dry flower yield (g m<sup>-2</sup> d<sup>-1</sup>) by the daily light integral (DLI; mol m<sup>-2</sup> d<sup>-1</sup>). PCE ranged from 0.22 to 0.36 g per mol, based on photon integration from transplanting to harvest (Fig 6). It is important to include the photons during the 7 to 14-day vegetative growth phase in this analysis. In these studies, the PCE would have been about 10% higher with closer pot spacing during the vegetative growth phase. The
canopy closed (filled the entire chamber) in all trials by about week 3. PCE is calculated per unit ground area, not per plant area. A lower plant density would thus typically reduce photon capture and PCE.
PCE values facilitate calculation of flower yield per dollar of electrical energy for the light fixtures. Yield per dollar of electricity increased as efficacy increased in all trials (Fig 7; \( p < 0.01 \)). The white+red 1 fixture (10% blue) had the highest efficacy (2.51 \( \mu \text{mol J}^{-1} \)) and produced 27% more flower per dollar of electricity on average than HPS and 16% more flower per dollar than the other LEDs with lower efficacy.
**Discussion**
**Yield**
*Potential underlying physiological basis for the observed yield reduction.* Analyzing the results by YPF indicates that a decrease in quantum yield with an increasing blue photon fraction would account for 7% of the 12% decrease in yield (S6 Fig). Although leaf area was not measured, photon capture may have also contributed to the yield reduction. Far-red photons likely had a small contribution to the 12% decrease in yield. Thus, four physiological responses could have contributed to the 12% decrease in yield: 1) blue fraction effect on quantum yield, 2) blue fraction effect on leaf expansion and photon capture, 3) far-red fraction effect on photosynthesis, and 4) far-red fraction effect on leaf expansion and photon capture.
1. Blue photons have a lower quantum yield due to photon absorbance by non-photosynthetic pigments within leaves [9].
2. Increasing the fraction of blue photons is typically associated with decreased leaf expansion and thus reduced photon capture [17, 19].
3. Far-red photons from 701 to 750 nm are photosynthetically active [14, 15].
4. Far-red photons can increase yield by modifying morphology and increasing photon capture [25].
There is the potential that changes in the fraction of other wavelengths could have contributed to the results. Percent blue ranged from 4 to 20% (a 5-fold increase), percent green ranged from 39 to 49%, percent red ranged from 31 to 53%, and percent far-red ranged from 2 to 6%. Snowden et al. [19] found that changes in green photon fraction from 2 to 41% had a minimal effect of growth of seven species, so the 6% range in these studies likely had a minimal effect on yield or yield parameters. The shift from blue to red photons likely increased
For the above reasons, it is unlikely that the changes in green or red photon fraction were the primary factors responsible for the effects on yield. This is consistent with previous studies on other species [17–19].
These results are similar to those of Magagnini et al. [6] who reported a decrease in flower yield in Cannabis (in one of two replicate studies) when the fraction of blue increased from 8 to 14%, but there was no further change from increasing the fraction of blue from 14 to 24%. We found a linear decrease in flower yield up to 20% blue photons. The PPFD in this study was double the intensity of Magagnini et al. [6], which may have affected the response.
**Effects of other environmental factors.** There were differences in average daily temperature and day night differential among studies, but the effect of blue photons was consistent in all studies. The lower temperature and lower PPFD in trial one likely contributed to the lower yield compared to trials two and three.
The canopy temperature in the unshielded HPS treatment averaged 0.3˚C higher than the shielded HPS, and 0.9˚C higher than the LED treatments (S3 Fig). This is consistent with the leaf and canopy temperature model of Nelson and Bugbee [53]. Although the unshielded HPS treatment was only replicated in one study, yield was within 3% of the yield of the other three replicate HPS treatments.
The photosynthetic response of plants to PPFD is non-linear, so the effect of intensity becomes less significant at high PPFD [26]. The photosynthetic response of plants to temperature, especially between 20˚ and 30˚C, is generally more significant and is expected to have a larger effect on photosynthesis [58]. The optimum temperature for photosynthesis is variable among species and is typically related to the region of origin of a species [59]. Bazzaz et al. [60] determined that Cannabis acclimated to warmer temperatures (32˚C/23˚C, day/night) had a higher photosynthetic rate than plants acclimated to cooler temperatures (23˚C/16˚C, day/night). The difference in yield among trials is therefore likely due to differences in temperature, but further research is needed to understand the effect of temperature on flower yield of Cannabis. In any case, the effect of spectra on yield was consistent among trials.
It is possible that Cannabis cultivars with unique morphologies and days to flower would respond differently to blue photon fraction. The effects of blue photon fraction vary among species [19]. Although growth parameters (e.g. height, fresh weight) vary among cultivars within a species, the effects of spectra on these parameters are expected to be small [23]. About 7% of the 12% reduction in yield can be accounted for by lower YPF, a response that is consistent across species [9]. The cultivar “Trump” was selected from among 12 diverse cultivars because it represents an average time to flower. Future research is needed to evaluate potential interactions of spectra with Cannabis cultivars.
**Cannabinoid concentration**
**Spectral effects.** Spectral quality can trigger synthesis of secondary metabolites, which include photo-protective pigments (e.g. anthocyanins), but no theoretical mechanism that links spectral quality to cannabinoid biosynthesis has been elucidated. In this study, there was no statistical difference in final cannabinoid concentration among spectral treatments. Magagnini et al. [6] reported an increase in CBD and THC concentration under 14 and 24% blue (from LEDs) compared to 8% blue under a mogul-base HPS fixture. Additionally, they saw an increase in CBG concentration, a precursor to both CBD and THC, with an increasing fraction of blue photons. They hypothesized that the first enzyme in the cannabinoid pathway is
responsive to blue photons. Photoreceptors are likely under-saturated at lower light intensities allowing for an increased sensitivity to spectral quality. This could explain why an effect on cannabinoid concentration was observed at the lower PPFD of Magagnini et al. [6] and not at the higher PPFD in this study.
**Time-course of cannabinoid accumulation.** Cannabinoid biosynthesis is relatively well understood, although the effects of environment are not well characterized. Cannabinoid degradation, especially in vivo, is far less studied but the implications could be significant. In trial two, both CBD and THC peaked around week five followed by a decrease in the last two weeks of flowering. This decrease has been reported previously [61, 62]. It is not necessarily caused by temperature. The underlying effect of environment on cannabinoid accumulation is not yet clear.
Growth dilution could explain a decline in the final weeks of flowering. This occurs when the plant accumulates biomass faster than cannabinoids. Alternatively, cannabinoids could be degrading. Mahlberg and Kim [34] indicate that trichome maturity can be assessed by color, which is related to the relative cannabinoid content. They found that mature glands are translucent and contain the highest cannabinoid concentration, aged glands are yellow and contain lower cannabinoids and senescent glands are black or brown and contain the lowest cannabinoids. They propose this could be caused by polymerization, transportation, or volatilization of cannabinoids but the mechanism remains to be elucidated. Regardless of the mechanism, the fact that THC concentration can decline in the final weeks of flowering has implications for hemp growers where regulations dictate a maximum THC concentration. Further research is needed to determine the stability and longevity of cannabinoids in vivo, and whether environmental factors can be altered to reduce THC.
**Fixture design and efficacy**
LEDs have facilitated rapid progress in understanding spectral effects on plant growth and secondary metabolism, but these data indicate that efficacy has a larger effect than spectra on the economics of indoor Cannabis cultivation. Decreasing the blue fraction from 20% to 4% resulted in a 12.3% increase in dry mass yield, while increasing the efficacy from 1.72 to 2.51 μmol J⁻¹ resulted in a 27% increase in yield per dollar of electricity. Similar results were reported for 13 lettuce cultivars grown in greenhouses supplemented with HPS or LED lights [23].
Red LEDs have a higher efficacy than blue (and by proxy white) LEDs because red photons have less energy than green and blue. This indicates that LED fixture manufacturers and growers should consider white+red fixtures that have a high portion of red [5]. The white+red 1 (10% blue) treatment had the highest yield per dollar of electricity. The highest efficacy fixtures listed by the Design Lights Consortium are white+red or blue+red combinations [52].
One drawback of high fraction red LED fixtures is low color rendering index (CRI; Rₙ). This metric indicates how well a light source reveals the colors of an object relative to a reference, usually an incandescent light or sunlight. A newer metric, color fidelity index (CFI; Rᵥ), which comes from Technical Memo 30 (TM-30), is mathematically more complex and is designed to provide an improved metric to assess the unique mixtures of colors in LED fixtures [63]. Both CRI and CFI range from 0 to 100. These metrics are routinely used for human lighting applications but are rarely considered in controlled environment agriculture. Broad-spectrum light sources (usually with white LEDs) have a higher CRI and CFI than narrow-spectrum sources [64]. Unfortunately, high-efficacy green LEDs are not yet available, but green photon fraction increases the CRI and CFI. This facilitates identification of pests, pathogens, or nutrient disorders. HPS has a CRI of 43 and a CFI of 55. White LEDs typically have a CRI above 70 and a CFI above 85 [65].
A photograph of plants under a DE-HPS and a 3000 K LED is shown in S4 Fig. There is a tradeoff between CRI/CFI and efficacy (S5 Fig). The importance of establishing a baseline CFI that optimizes identification of abnormalities while still maximizing efficacy is an overlooked parameter in horticultural lighting.
Optimized lighting is critical to the economics of indoor Cannabis cultivation. These data indicate that efficacy is more important than spectra, and that fixture manufacturers should consider reducing the blue fraction to improve economic yield of Cannabis.
Supporting information
S1 Fig. Effect of blue photons on harvest index (HI) in three studies. HI is the ratio of usable biomass to total above ground biomass, here defined as the ratio of flowers to flowers, leaves and stems. There was no significant effect of blue photons on HI (p = 0.91). HI was highest in trial one, which had the highest average temperature of the three trials.
S2 Fig. Example calculation to determine grams per dollar of electricity. PPFD, efficacy, photon conversion efficiency (g per mol), and cost of electricity are inputs to the calculation. Assuming $0.10 per kWh and 0.3 g per mol, it costs $1.00 to produce 3 g of flower.
S3 Fig. Thermal images of the canopy under five spectral treatments from trial three. Darker shades indicate cooler temperatures. Note that the flower buds (white color) are about 2 C warmer than the leaves. HPS without glass was about 1 C warmer than the LEDs and about 0.3 C warmer than the HPS with glass. Thermal images can be used to detect water stress or disease before visual symptoms become apparent.
S4 Fig. Photograph of two adjacent spectral treatments separated by the chamber wall. The color rendering index (CRI) and color fidelity index (CFI) is low under HPS lights. White LEDs allow for easier identification of pests, pathogens, or nutrient disorders.
S5 Fig. Effect of blue photon fraction on efficacy and CRI (top graph) and CFI (bottom graph) of an HPS fixture and four LED fixtures. CRI and CFI describe the ability of a light source to distinguish true colors of an object relative to a reference. Fixtures with a higher CRI or CFI typically have lower efficacy. HPS has low CRI, CFI and efficacy compared to LEDs.
S6 Fig. Effect of blue photon fraction on canopy photosynthesis. Canopy photosynthesis measurements were made under blue + red LEDs. The system used to make measurements has been described previously [14, 15], and references cited therein. Decreasing the blue photon fraction significantly increased canopy photosynthesis (p = 0.01).
S1 Data.
Acknowledgments
We thank Alec Hay for technical assistance, Dr. Xin Dai for assistance with statistical design and analysis, and Dr. Bill Doucette, Emily Angell and Jeff Wight at the Utah Water Research Laboratory, and Brandon Forsyth at the Utah Department of Agriculture and Food, for cannabinoid
analysis. We also thank the Utah Department of Agriculture and Food for providing a certificate to research industrial hemp and Dr. Royal Heins for his helpful review comments.
Author Contributions
**Conceptualization:** F. Mitchell Westmoreland, Paul Kusuma, Bruce Bugbee.
**Data curation:** F. Mitchell Westmoreland.
**Formal analysis:** F. Mitchell Westmoreland.
**Funding acquisition:** Bruce Bugbee.
**Methodology:** Paul Kusuma.
**Supervision:** Bruce Bugbee.
**Writing – original draft:** F. Mitchell Westmoreland.
**Writing – review & editing:** F. Mitchell Westmoreland, Paul Kusuma, Bruce Bugbee.
References
1. Backer R, Schwinghamer T, Rosenbaum P, McCarty V, Eichhorn Bilodeau S, Lyu D, et al. Closing the yield gap for cannabis: a meta-analysis of factors determining cannabis yield. Front Plant Sci. 2019; 10: 495. https://doi.org/10.3389/fpls.2019.00495 PMID: 31068957
2. Mills E. The carbon footprint of indoor Cannabis production. Energy Policy. 2012; 46: 58–67.
3. Nelson JA, Bugbee B. Economic analysis of greenhouse lighting: light emitting diodes vs. high intensity discharge fixtures. PLoS One. 2014; 9: e99010. https://doi.org/10.1371/journal.pone.0099010 PMID: 24905835
4. Wallace C, Both AJ. Evaluating operating characteristics of light sources for horticultural applications. Acta Hortic. 2016; 1134: 435–444.
5. Kusuma P, Pattison PM, Bugbee B. From physics to fixtures to food: current and potential LED efficacy. Hortic Res. 2020; 7: 1–9.
6. Magagnini G, Grassi G, Kotiranta S. The Effect of Light Spectrum on the Morphology and Cannabinoid Content of Cannabis sativa L. Med Cannabis Cannabinoids. 2018; 1: 19–27.
7. Eichhorn Bilodeau S, Wu B-S, Rufyikiri A-S, MacPherson S, Lefsrud M. An update on plant photobiology and implications for cannabis production. Front Plant Sci. 2019; 10: 296. https://doi.org/10.3389/fpls.2019.00296 PMID: 31001288
8. Namdar D, Charuvdi M, Ajampura V, Mazuz M, Ion A, Kamara I, et al. LED lighting affects the composition and biological activity of Cannabis sativa secondary metabolites. Ind Crops Prod. 2019; 132: 177–185.
9. McCree KJ. The action spectrum, absorptance and quantum yield of photosynthesis in crop plants. Agric For Meteorol. 1971; 9: 191–216.
10. McCree KJ. Test of current definitions of photosynthetically active radiation against leaf photosynthesis data. Agric For Meteorol. 1972; 10: 443–453.
11. Inada K. Action spectra for photosynthesis in higher plants. Plant Cell Physiol. 1976; 17: 355–365.
12. Sager JC, Smith WO, Edwards JL, Cyr KL. Photosynthetic efficiency and phytochrome photoequilibria determination using spectral data. Trans ASAE. 1988; 31: 1882–1889.
13. Barnes C, Tibbits T, Sager J, Deitzer G, Bubenheim D, Koerner G, et al. Accuracy of quantum sensors measuring yield photon flux and photosynthetic photon flux. HortScience. 1993; 28: 1197–1200. PMID: 11537894
14. Zhen S, Bugbee B. Far-red photons have equivalent efficiency to traditional photosynthetic photons: Implications for redefining photosynthetically active radiation. Plant Cell Environ. 2020; 43: 1259–1272. https://doi.org/10.1111/pce.13730 PMID: 31990071
15. Zhen S, Bugbee B. Substituting far-red for traditionally defined photosynthetic photons results in equal canopy quantum yield for CO2 fixation and increased photon capture during long-term studies: Implications for re-defining PAR. Front Plant Sci. 2020; 11: 1433. https://doi.org/10.3389/fpls.2020.581156 PMID: 33014004
16. Zhen S, van Iersel MW. Far-red light is needed for efficient photochemistry and photosynthesis. J Plant Physiol. 2017; 209: 115–122. https://doi.org/10.1016/j.jplph.2016.12.004 PMID: 28039776
17. Son KH, Oh MM. Growth, photosynthetic and antioxidant parameters of two lettuce cultivars as affected by red, green, and blue light-emitting diodes. Hortic Environ Biotechnol. 2015; 56: 639–653.
18. Kang WH, Park JS, Park KS, Son JE. Leaf photosynthetic rate, growth, and morphology of lettuce under different fractions of red, blue, and green light from light-emitting diodes (LEDs). Hortic Environ Biotechnol. 2016; 57: 573–579.
19. Snowden MC, Cope KR, Bugbee B. Sensitivity of Seven Diverse Species to Blue and Green Light: Interactions with Photon Flux. PLoS ONE. 2016; 11: e0163121. https://doi.org/10.1371/journal.pone.0163121 PMID: 27706176
20. Dougher TA, Bugbee B. Differences in the Response of Wheat, Soybean and Lettuce to Reduced Blue Radiation. Photochem Photobiol. 2001; 73: 199–207. PMID: 11272735
21. Cope KR, Bugbee B. Spectral Effects of Three Types of White Light-emitting Diodes on Plant Growth and Development: Absolute versus Relative Amounts of Blue Light. HortScience. 2013; 48: 504–509.
22. Cope KR, Snowden MC, Bugbee B. Photobiological interactions of blue light and photosynthetic photon flux: Effects of monochromatic and broad-spectrum light sources. Photochem Photobiol. 2014; 90: 574–584. https://doi.org/10.1111/php.12233 PMID: 24372324
23. Hernandez E, Timmons MB, Mattson NS. Quality, Yield, and Biomass Efficacy of Several Hydroponic Lettuce (Lactuca sativa L.) Cultivars in Response to High Pressure Sodium Lights or Light Emitting Diodes for Greenhouse Supplemental Lighting. Horticulture. 2020; 6: 7.
24. Wang J, Lu W, Tong Y, Yang Q. Leaf morphology, photosynthetic performance, chlorophyll fluorescence, stomatal development of lettuce (Lactuca sativa L.) exposed to different ratios of red light to blue light. Front Plant Sci. 2016; 7: 250. https://doi.org/10.3389/fpls.2016.00250 PMID: 27014285
25. Meng Q, Runkle ES. Far-red radiation interacts with relative and absolute blue and red photon flux densities to regulate growth, morphology, and pigmentation of lettuce and basil seedlings. Sci Hortic. 2019; 255: 269–280.
26. Chandra S, Lata H, Khan IA, Elsohly MA. Photosynthetic response of Cannabis sativa L. to variations in photosynthetic photon flux densities, temperature and CO2 conditions. Physiol Mol Biol Plants. 2008; 14: 299–306. https://doi.org/10.1007/s12298-008-0027-x PMID: 23572895
27. Park Y, Runkle ES. Far-red radiation promotes growth of seedlings by increasing leaf expansion and whole-plant net assimilation. Environ Exp Bot. 2017; 136: 41–49.
28. Park Y, Runkle ES. Far-red radiation and photosynthetic photon flux density independently regulate seedling growth but interactively regulate flowering. Environ Exp Bot. 2018; 155: 206–216.
29. Legris M, Ince YÇ, Fankhauser C. Molecular mechanisms underlying phytochrome-controlled morphogenesis in plants. Nat Commun. 2018; 10: 5219. https://doi.org/10.1038/s41467-019-13045-0 PMID: 31745087
30. Gaoni Y, Mechoulam R. Isolation, structure, and partial synthesis of an active constituent of hashish. J Am Chem Soc. 1964; 86: 1646–1647.
31. Mechoulam R, Gaoni Y. A total synthesis of dl-Δ1-tetrahydrocannabinol, the active constituent of hashish1. J Am Chem Soc. 1965; 87: 3273–3275. https://doi.org/10.1021/ja01092a065 PMID: 14324315
32. Andre CM, Hausman J-F, Guerriero G. Cannabis sativa: The Plant of the Thousand and One Molecules. Front Plant Sci. 2016; 7: 19. https://doi.org/10.3389/fpls.2016.00019 PMID: 26870049
33. Hanusˇ LO, Meyer SM, Muñoz E, Taglialatela-Scafati O, Appendino G. Phytocannabinoids: a unified critical inventory. Nat Prod Rep. 2016; 33: 1357–1392. https://doi.org/10.1039/c6np00074f PMID: 27722705
34. Mahlberg PG, Kim ES. Accumulation of Cannabinoids in Glandular Trichomes of Cannabis (Cannabaceae). Journal of Industrial Hemp. 2004; 9: 15–36.
35. Shoyama Y, Yagi M, Nishioka I, Yamauchi T. Biosynthesis of cannabinoid acids. Phytochemistry. 1975; 14: 2189–2192.
36. Carvalho A, Hansen EH, Kayser O, Carlsen S, Stehle F. Designing microorganisms for heterologous biosynthesis of cannabinoids. FEMS Yeast Res. 2017; 17: 4. https://doi.org/10.1093/femsyr/fox037 PMID: 28582498
37. Zager JJ, Lange I, Srividya N, Smith A, Lange BM. Gene Networks Underlying Cannabinoid and Terpenoid Accumulation in Cannabis. Plant Physiol. 2019; 180: 1877–1897. https://doi.org/10.1104/pp.18.01506 PMID: 31138625
38. Campbell BJ, Berrada AF, Hudalla C, Amaducci S, McKay JK. Genotype × environment interactions of industrial hemp cultivars highlight diverse responses to environmental factors. Agroecosystems, Geosciences & Environment. 2019; 2: 1–11.
39. Potter DJ. A review of the cultivation and processing of cannabis (Cannabis sativa L.) for production of prescription medicines in the UK. Drug Test Anal. 2014; 6: 31–38. https://doi.org/10.1002/dta.1531 PMID: 24115748
40. Chandra S, Lata H, ElSohly MA, Walker LA, Potter D. Cannabis cultivation: methodological issues for obtaining medical-grade product. Epilepsy Behav. 2017; 70: 302–312. https://doi.org/10.1016/j.yebeh.2016.11.029 PMID: 28820406
41. Schluttenhofer C, Yuan L. Challenges towards Revitalizing Hemp: A Multifaceted Crop. Trends Plant Sci. 2017; 22: 917–929. https://doi.org/10.1016/j.tplants.2017.08.004 PMID: 28886910
42. Ouzounis T, Rosenqvist E, Ottoen CO. Spectral effects of artificial light on plant physiology and secondary metabolism: a review. HortScience. 2015; 50: 1128–1135.
43. Thoma F, Somborn-Schulz A, Schlehuber D, Keuter V, Deerberg G. Effects of Light on Secondary Metabolites in Selected Leafy Greens: A Review. Front Plant Sci. 2020; 11: 497. https://doi.org/10.3389/fpls.2020.00497 PMID: 32391040
44. Fairbairn JW, Liebmann JA. The cannabinoid content of Cannabis sativa L grown in England. J Pharm Pharmacol. 1974; 26: 413–419. https://doi.org/10.1111/j.2042-7158.1974.tb09306.x PMID: 4154985
45. Mary W, Crombie L. The influence of photosynthesis and SKF inhibitors on cannabinoid production in Cannabis sativa. Phytochemistry. 1977; 16: 1369–1371.
46. Mahlberg PG, Hemphil JK. Effect of light quality on cannabinoid content of Cannabis sativa L. (Cannabaceae). Bot Gaz. 1983; 144: 43–48.
47. Valle JR, Vieira JE, Aucelio JG, Valio IF. Influence of photoperiodism on cannabinoid content of Cannabis sativa L. Bull Narc. 1978; 30: 67–68. PMID: 248289
48. Vanhove W, Van Damme P, Meert N. Factors determining yield and quality of illicit indoor cannabis (Cannabis spp.) production, Forensic Sci Int. 2011; 212: 158–163. https://doi.org/10.1016/j.forsciint.2011.06.006 PMID: 21737218
49. Went FW. The experimental control of plant growth. Volume 17. Chronica Botanica Co; 1957.
50. Monje O, Bugbee B. Adaptation to high CO2 concentration in an optimal environment: radiation capture, canopy quantum yield and carbon use efficiency. Plant Cell Environ. 1998; 21: 315–324. https://doi.org/10.1046/j.1365-3040.1998.00284.x PMID: 11543216
51. Van Iersel MW, Bugbee B. A multiple chamber, semicontinuous, crop carbon dioxide exchange system: design, calibration, and data interpretation. J Am Soc Hortic Sci. 2000; 125: 86–92. PMID: 11762389
52. Horticultural lighting Design Lights Consortium. [cited 21 Jan 2021]. Database: Horticultural Lighting [Internet]. https://www.designlights.org/horticultural-lighting/search/?search=
53. Nelson JA, Bugbee B. Analysis of environmental effects on leaf temperature under sunlight, high pressure sodium and light emitting diodes. PLoS One. 2015; 10: e0138930. https://doi.org/10.1371/journal.pone.0138930 PMID: 26448613
54. Kinross AD, Hageman KJ, Doucette WJ, Foster AL. Comparison of Accelerated Solvent Extraction (ASE) and Energized Dispersive Guided Extraction (EDGE) for the analysis of pesticides in leaves. J Chromatogr. 2020; 1627: 461414.
55. Establishment of a Domestic Hemp Production Program. In: Federal Register [Internet]. 31 Oct 2019 [cited 27 Apr 2020]. https://www.federalregister.gov/documents/2019/10/31/2019-23749/establishment-of-a-domestic-hemp-production-program
56. Tabachnick BG, Fidell LS, Ullman JB. Using multivariate statistics. 5th ed. Pearson Boston, MA; 2007.
57. Unkovich M, Baldock J, Forbes M. Variability in harvest index of grain crops and potential significance for carbon accounting: examples from Australian agriculture. Advances in agronomy. Elsevier; 2010. pp. 173–219.
58. Chandra S, Lata H, Khan IA, ElSohly MA. Temperature response of photosynthesis in different drug and fiber varieties of Cannabis sativa L. Physiol Mol Biol Plants. 2011; 17: 297. https://doi.org/10.1007/s12298-011-0068-4 PMID: 23573022
59. Berry J, Bjorkman O. Photosynthetic response and adaptation to temperature in higher plants. Annu Rev Plant Physiol. 1980; 31: 491–543.
60. Bazzaz FA, Dusek D, Seigler DS, Haney AW. Photosynthesis and cannabinoid content of temperate and tropical populations of Cannabis sativa. Biochem Syst Ecol. 1975; 3: 15–18.
61. Pacifico D, Miselli F, Carboni A, Moschella A, Mandolino G. Time course of cannabinoid accumulation and chemotype development during the growth of Cannabis sativa L. Euphytica. 2008; 160: 231–240.
62. Aizpurua-Olaizola O, Soydaner U, Öztürk E, Schibano D, Simsir Y, Navarro P, et al. Evolution of the Cannabinoid and Terpene Content during the Growth of Cannabis sativa Plants from Different Chemotypes. J Nat Prod. 2016; 79: 324–331. https://doi.org/10.1021/acs.jnatprod.5b00949 PMID: 26836472
63. Wei M, Royer M, Huang HP. Perceived colour fidelity under LEDs with similar Rf but different Ra. Light Res Technol. 2019; 51: 858–869.
64. Park Y, Runkle ES. Spectral effects of light-emitting diodes on plant growth, visual color quality, and photosynthetic photon efficacy: White versus blue plus red radiation. PLoS One. 2018; 13: e0202386. https://doi.org/10.1371/journal.pone.0202386 PMID: 30114282
65. Technical memorandum: IES method for evaluating light source color rendition. Illuminating Engineering Society. 2020. [Cited 2020 August]. https://store.ies.org/product/tm-30-20-ies-method-for-evaluating-light-source-color-rendition/ | 2025-03-05T00:00:00 | olmocr | {
"pdf_page_numbers": [
[
0,
2507,
1
],
[
2507,
6945,
2
],
[
6945,
10809,
3
],
[
10809,
12728,
4
],
[
12728,
15343,
5
],
[
15343,
19170,
6
],
[
19170,
22609,
7
],
[
22609,
24423,
8
],
[
24423,
25316,
9
],
[
25316,
26861,
10
],
[
26861,
28284,
11
],
[
28284,
29271,
12
],
[
29271,
33043,
13
],
[
33043,
37080,
14
],
[
37080,
39998,
15
],
[
39998,
43472,
16
],
[
43472,
48274,
17
],
[
48274,
53048,
18
],
[
53048,
53694,
19
]
]
} |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.