Comparative transcriptome analysis reveals the mechanism of cross-protection against Verticillium wilt conferred on sunower by hypovirulent strain Gibellulopsis nigrescens Vn-1

Background: Hypovirulent fungal strain Gibellulopsis nigrescens Vn-1 cross-protects sunower against Verticillium wilt. To discover the mechanism of cross-protection by this hypovirulent strain, we analyzed defense enzyme activities and transcriptomes of root samples infected with virulent and hypovirulent strains. Results: Defense enzyme activities increased after inoculation, with the highest levels observed 24 h post-inoculation. At the same time, defense enzyme gene expressions were upregulated, and H 2 O 2 accumulation decreased. A comparative transcriptome analysis revealed that three specic oxidoreductase-related GO terms were signicantly enriched in the Vn-1 group compared with the control. In addition, 33 resistance genes and 160 susceptibility genes were predicted. Seven transcription factors (TFs), two phytohormone response factors, two E3 ubiquitin-protein ligases, two CCR4-associated factor 1 proteins, and two predicted leucine repeat rich (LRR) receptors were found to contribute to the conferral of resistance on sunower. Conclusions: According to our results, hypovirulent strain G. nigrescens Vn-1 can reduce levels of reactive oxygen species in sunower induced by infection with virulent strains such as V. dahliae V33 by regulating HaCAT expression. Furthermore, multiple resistance-related TFs, phytohormones, and receptors contribute to the formation of specic resistance against virulent strain V. dahliae V33.


Introduction
Many species of Verticillium are globally distributed, soil-borne vascular-wilt fungal pathogens that infect numerous hosts, including sun ower (Helianthus annuus L.) [1], tomato [2], potato [3], lettuce [4][5], cotton [6], and cauli ower, and cause destructive wilt disease [7]. Among members of the genus, V. dahliae has the broadest host range, with the ability to infect over 200 plant species [8][9]. Sun ower is a globally important oilseed crop, and sun ower Verticillium wilt caused by V. dahliae is one of the most destructive diseases in oilseed crop production [10]. In the absence of hosts, Verticillium species are able to survive in soil for up to 14 years by relying on melanized resting structures [11]. Resting structures of Verticillium, which vary according to species, include microsclerotia, chlamydospores, and melanized resting mycelia [12][13][14][15]. In many crops, Verticillium wilt disease is di cult to manage by agricultural and chemical means because these resting structures are present in soil. As a result, the development of resistant cultivars and the enhancement of plant resistance are critical measures to control Verticillium wilt in agricultural production.
Innate and induced resistance (so-called cross-protection) constitute the basal resistance system of plants that protects against biotic and abiotic stresses from the environment [16][17][18]. When plants recognize pathogen-associated molecular patterns or phytopathogen effectors [19][20], they invoke defense responses, including a hypersensitive response, which is localized, and system acquired resistance (SAR), which is both localized and distal from the primary infection [21][22]. SAR is accompanied by the transduction of endogenous signals, including salicylic acid (SA) or jasmonic acid (JA), as well as the expression of pathogenesis-related (PR) proteins and the upregulation of defense enzyme genes [23][24][25]. Cross-protection is typically used to improve plant resistance against phytopathogens in agricultural production [26], with biotic inducers (including endophytes and hypovirulent strains) and abiotic ones (including physical or chemical agents) commonly used [27].
Hypovirulent or avirulent phytopathogens, which are types of biological inducers, can potentially confer host plant resistance against virulent strains [28]. Cross-protection methods using hypovirulent strains have the potential to control plant disease. For example, plants infected with a hypovirulent virus line are usually resistant to superinfection by other strains of the same virus [29][30]. Cross-protection has been used to control plant virus diseases [31], such as tobacco mosaic virus in tomato [32], zucchini yellow mosaic virus [33], and papaya ringspot virus [34][35]. Cross-protection involving phytopathogenic fungi has also been discovered and used to control plant fungal diseases [36][37]. For example, many destructive phytopathogenic fungi have been effectively controlled by cross-protection using hypovirulent strains, including members of the genera Fusarium, Colletotrichum, Verticillium, and Puccinia [38][39]. Hypovirulent Verticillium strains have been previously isolated from cotton and demonstrated as promising agents for biocontrol of cotton Verticillium wilt [40][41]. Zhao et al. recently isolated a hypovirulent strain of Gibellulopsis nigrescens (strain Vn-1) causing minimal wilt in sun ower [42]. This strain was found to confer cross-protection against a virulent strain (V. dahliae V33) on sun ower [42]. Cross-protection functions have also been discovered between different phytopathogens [43]. The exploitation of cross-protection is thus an e cient biocontrol measure for managing plant disease.
Moreover, hypovirulent strains confer broad spectrum antifungal activity to protect plants against multiple phytopathogens.
Although cross-protection has long been known, the underlying mechanisms have been poorly understood until recently [44]. With the development of sequencing technology, transcriptome analysis is now the most e cient method for analyzing interactions between hypovirulent strains and host plants [45][46]. Various analytical approaches in combination with transcriptome sequencing are potentially useful for discovering PR genes and ultimately revealing cross-protection mechanisms. These approaches include 1) genome mapping, 2) differentially expressed gene (DEG) analysis, 3) DEG gene ontology (GO) enrichment analysis, and 4) DEG Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis [47][48][49][50][51].
In this study, we measured expression levels of defense enzyme genes and H 2 O 2 accumulation and analyzed DEGs responsive to hypovirulent strain G. nigrescens Vn-1. Our objectives were as follows: 1) detection of G. nigrescens cross-protection in sun ower; 2) discovery of DEGs involved in formation of the induced resistance; and 3) elucidation of the mechanism of cross-protection against Verticillium wilt conferred by hypovirulent strain G. nigrescens Vn-1.

Results
Gibellulopsis nigrescens Vn-1 induced the defense response of sun ower We previously identi ed the hypovirulent strain G. nigrescens Vn-1 as a promising candidate for conferral of cross-protection against sun ower Verticillium wilt [42]. To investigate the mechanism of induced resistance achieved with this hypovirulent strain in the present study, we measured defense enzyme activities in sun ower at different time points after inoculation. We found that the activities of four defense enzymes, namely, peroxidase (POD), superoxide dismutase (SOD), catalase (CAT), and phenylalanine ammonia lyase (PAL), were induced after inoculation. Three of these enzymes (POD, SOD, and CAT) had their highest levels of catalytic activity at 24 h post-inoculation (hpi). The highest PAL activity occurred at 12 hpi, with no signi cant difference observed between 12 hpi and 24 hpi (Fig. 1A).
Defense enzyme activities after inoculation with G. nigrescens Vn-1 were higher than those following infection with V. dahliae V33, a virulent strain causing more serious symptoms of sun ower Verticillium wilt. A quantity Real Time Polymerase Chain Reaction (qRT-PCR) analysis revealed that the expressions of HaPOD, HaSOD, HaCAT, and HaPAL were upregulated at 24 hpi (Fig. 1B). Interestingly, sun ower seedlings inoculated with G. nigrescens Vn-1 or V. dahliae V33 exhibited reduced H 2 O 2 accumulation compared with a control (CK) sample at 24 hpi (Fig. 2). These results indicate that hypovirulent strain G. nigrescens Vn-1 was able to induce the defense response of sun ower and that 24 hpi was the critical time point, as expressions of the resistance genes were upregulated. The hypovirulent and virulent strains were both able to clear H 2 O 2 accumulated in sun ower seedlings during infection.

Overview of Illumina RNA-sequencing data
To investigate the mechanism of resistance induced by hypovirulent strain G. nigrescens Vn-1, we performed transcriptome sequencing of sun ower root samples inoculated with G. nigrescens Vn-1 or V. dahliae V33, as well as a control, at 24 hpi. The sequencing results were used for comparative transcriptome analysis. More than 45 million clean reads (> 6.5 Gb) were generated per root sample ( Table 1). The transcriptome assemblies obtained for the three groups of sun ower roots were pooled and used to assemble complete transcripts based on the sun ower genome. After elimination of incomplete transcripts, 86.17% to 87.30% of clean reads were mapped to the reference genome; in total, 83.17% to 84.34 % of clean reads were uniquely mapped (Fig. 3A, Table S1). More than 95% of the reads were located within exons (Fig. 3B, Table S1). In this study, a gene was considered to be expressed in a sample if a transcript was detected in the cDNA library for three replicates. Relatively few genes were highly expressed (Fig. 3C). In each group of sun ower root samples, the data from three replicates were highly correlated (Pearson's r > 0.95), thus indicating that the transcriptome pro les were highly reproducible (Fig. 3D).

Functional enrichment analysis of DEGs
To investigate sun ower genes potentially involved in resistance to Verticillium wilt, we compared the three transcriptomes. We identi ed a total of 1,790 DEGs between Vn-1 and WT groups, including 404 and 1,386 DEGs with up-and downregulated expressions, respectively. We also identi ed 3,469 DEGs between V33 and WT groups, including 1,204 and 2,445 DEGs with up-and downregulated expressions, respectively. Moreover, 744 DEGs were discovered between Vn-1 and V33 groups, of which 501 and 243 were up-and downregulated, respectively (Fig. 4A). Many more genes were differentially expressed between V33 and CK groups than between Vn-1 and CK groups. Venn diagram analysis of DEGs revealed that 1,167 DEGs were included in either Vn-1 vs. CK or V33 vs. CK comparisons, whereas only 58 DEGs were included in all three comparison groups (Vn-1 vs. CK, V33 vs. CK, and Vn-1 vs. V33) ( Fig 4B).

Validation of DEGs
To con rm the reliability of the DEGs identi ed in the comparative transcriptome analysis, we randomly selected 16 DEGs that were differentially expressed among the three comparison groups for qRT-PCR veri cation. The qRT-PCR results for most DEGs were highly correlated with the RNA-seq data (Pearson's r ³ 0.80); the exception was HannXRQ_Chr01g0021411 (Pearson's r = 0.79) (Fig. 5). This strong correlation indicates that the RNA-seq data were valid and reliable.

Gene Ontology (GO) enrichment analysis of DEGs
To functionally characterize DEGs, we performed a GO enrichment analysis. The number of signi cantly enriched GO terms in the V33 vs. CK comparison group was much higher than in Vn-1 vs. CK and Vn-1 vs. V33 comparison groups. The nine most signi cantly enriched GO terms in the Vn-1 vs. CK group were as follows: oxidation-reduction process, response to oxidative stress, single-organism metabolic process, lipid metabolic process, siroheme biosynthetic process, siroheme metabolic process, heme biosynthetic process, cellular glucan metabolic process, and glucan metabolic process (Fig. 6A, Table S2). Oxidationreduction, lipid metabolic, carbohydrate metabolic, and signaling processes were the major subcategories of processes within the biological process category. These four subcategories were also enriched in the V33 vs. CK comparison group (Fig. 6B, Table S2). In both Vn-1 vs. CK and V33 vs. CK comparison groups, the most abundant GO term in the cellular component category was extracellular region. In the molecular function category, oxidoreductase activity was the most signi cantly enriched subcategory in all three comparison groups (Fig. 6, Table S2). According to this GO enrichment analysis of DEGs, enriched resistance-related GO terms in Vn-1 vs. CK and V33 vs. CK comparison groups were highly similar, which suggests that both hypovirulent and virulent strains could induce defense responses at early stage of infection.
To identify GO terms associated with Verticillium wilt resistance in sun ower, we compared the terms signi cantly enriched in Vn-1 vs. CK and V33 vs. CK groups. We found 34 enriched GO terms common to both groups. In addition, 23 and 57 signi cantly enriched terms were unique to Vn-1 vs. CK and V33 vs. CK groups, respectively (Fig. 7). We speculated that these GO terms were related to resistance and susceptibility to Verticillium wilt.
KEGG pathway enrichment analysis of DEGs DEG functions were also examined by KEGG pathway enrichment analysis. DEGs identi ed in the three comparison groups were found to be related to 17 KEGG pathway (Table S3). According to the results of the analysis, infection with either hypovirulent or virulent strains in uenced the expressions of genes related to the following pathways: biosynthesis of secondary metabolites, biosynthesis of unsaturated fatty acids, fatty acid elongation, fatty acid metabolism, phenylalanine metabolism, phenylpropanoid biosynthesis, and plant-pathogen interaction. Majority of DEGs contributing to plant-pathogen interaction pathway, were down-regulated during the Verticillium species interacting with sun ower ( Fig  S1). Infection with hypovirulent strain Vn-1 additionally affected the expressions of genes in alanine, aspartate, and glutamate metabolism; cutin, suberin and wax biosynthesis; and ribosome pathways ( Table S3, Fig. S2), whereas virulent strain V33 was also able to trigger gene expressions in carbon xation in photosynthetic organisms, carbon metabolism, citrate cycle (TCA cycle), fatty acid degradation, glycolysis/gluconeogenesis, pentose phosphate, and proteasome pathways (Table S3, Fig.  S3). Moreover, only three KEGG pathways were signi cantly enriched in DEGs in the Vn-1 vs. V33 comparison: biosynthesis of unsaturated acids, fatty acid metabolism, and plant-pathogen interaction (Table S3). These results suggest that genes in many KEGG pathways, including alanine, aspartate, and glutamate metabolism; cutin, suberin and wax biosynthesis; and ribosome pathways, were involved in the speci c resistance induced by the hypovirulent strain.

Prediction of genes related to Verticillium wilt resistance in sun ower
To detect Verticillium wilt resistance genes in sun ower, we analyzed the DEGs uncovered by comparative transcriptome methods. DEGs upregulated in Vn-1 vs. CK and Vn-1 vs. V33 comparison groups but downregulated or not signi cantly changed in V33 vs. CK were predicted to be resistance genes. As a result, 33 genes were predicted to participate in resistance against V. dahliae (Fig. 8). As detailed in Table  S4, these predicted resistance genes encode seven transcription factors, two phytohormone response factors, two E3 ubiquitin-protein ligases, and two CCR4-associated factor 1 proteins. In addition, two proteins (with conserved leucine-rich repeat (LRR) domains) encoded by HannXRQ_Chr01g0025331 and HannXRQ_Chr14g0444761 were predicted as resistance-related receptors (Fig. S4).
In contrast to resistance genes, genes related to susceptibility were expected to be upregulated in V33 vs. CK and V33 vs. Vn-1 comparison groups and downregulated or unchanged in the Vn-1 vs. CK group. As a result, 160 genes were predicted to be related to susceptibility to Verticillium wilt in sun ower (Fig. 8, Table S5). Further research to verify these potential resistance and susceptibility genes is required.

Discussion
In a previous study, we found that a hypovirulent Verticillium strain provided cross-protection against virulent Verticillium strains in sun ower. In the present study, we monitored the defense response of sun ower at different time points after inoculation with hypovirulent Verticillium strain Vn-1 and performed a transcriptome analysis of samples at 24 hpi. According to our results, 24 hpi was the critical time point for inducing the defense response of sun ower. We also analyzed transcriptome data obtained at 24 hpi. The strong correlation among data sets from the three biological replicates of each sample are indicative of the high reproducibility and reliability of the transcriptome pro ling performed in this study, while the correlation between qRT-PCR and RNA-seq results suggests that the DEG analysis was accurate. As revealed by the transcriptome analysis, three speci c oxidoreductase-related GO terms were signi cantly enriched in the Vn-1 group compared with the control. Many speci c genes were only signi cantly differentially expressed in the Vn-1 vs. CK comparison group. This result suggests that hypovirulent strain Vn-1 induces sun ower speci c resistance to the virulent strain by regulating expression levels of those DEGs.
In this study, we discovered that the 24-hpi time point was the critical stage of infection, with the highest activities of defense enzymes. During this time, HaPAL, HaPOD, HaCAT, and HaSOD were also upregulated. These results suggest that V. dahliae Vn-1 and G. nigrescens induced the defense response of sun ower at 24 hpi. At this time point, a small number of germinated conidia attach to the root tip surface, and germ tube starts to penetrate into the adjacent epidermal root [52][53]. These results suggest that a small number of conidia of Verticillium species can induce a defense response in sun ower by secreting pathogenetic factors.
Reactive oxygen species (ROS) are important biotic signaling molecules involved in multiple physiological processes, such as seed germination, biotic or abiotic stress response, root development, and stomatal closure [54][55][56][57][58][59]. In addition, ROS burst, especially that of H 2 O 2 , can regulate plant innate immunity and contribute to hypersensitive response (HR) [60]. Under normal conditions, ROS are rapidly metabolized by antioxidant enzymes or compounds such as vitamins and glutathione [61][62][63][64]. When plants are infected by phytopathogens, however, excessive ROS are generated, which leads to resistance formation [65]. Either directly or indirectly, phytopathogens help eliminate excessive ROS by secreting effectors for further host infection [66]. In this study, V. dahliae V33 and G. nigrescens were able to remove redundant H 2 O 2 in sun ower seedlings by regulating the activities of HaCAT and HaPOD. These ndings suggest that V. dahliae V33 and G. nigrescens secrete effectors to reduce H 2 O 2 levels to facilitate infection of sun ower.
According to our qRT-PCR analysis, HaCAT and HaPOD expression levels in samples inoculated with G. nigrescens were signi cantly higher than those in seedlings infected with V. dahliae. Three oxidoreductase-related GO terms (GO:0016651, GO:0016705, and GO:0016627) were only enriched in the Vn-1 vs. CK comparison group. These results suggest that oxidoreductases regulate the concentration of ROS and the speci c resistance of sun ower to V. dahliae V33 when induced by G. nigrescens Vn-1.
Through comparative transcriptome analysis, we discovered 33 resistance genes speci cally upregulated in the Vn-1 vs. CK comparison group. Seven (32.32%) of these resistance genes encode members of WRKY, MYB, and Zinc-nger CCCH TF families. Previous studies of durum wheat, Arabidopsis, rice, and cotton have revealed the involvement of the Zinc-nger CCCH protein group in biotic and abiotic stress responses [67][68][69]. In addition, WRKY-type TFs confer differential tolerance to biotic and abiotic stresses in many plant species [70][71][72][73]. Further research has revealed that AtWRKY70 regulates SA-and JA-dependent defense signal pathways [74] by regulating the PR gene AtNPR1 [75], with similar results reported in rice [76]. In addition, MYB-type TFs are involved in many resistance process in hosts during phytopathogen infection, such as promoting cuticular wax and glucosinolate biosynthesis [77][78], enhancing the expression of defense-related genes [79], and even triggering programmed plant cell death [80]. These results indicate that hypovirulent strain G. nigrescens promotes the speci c resistance of sun ower to V. dahliae via many types of TF genes. The functions of the resistance genes predicted in this study need further functional veri cation.
Finally, two predicted resistance-related proteins (HannXRQ_Chr01g0025331 and HannXRQ_Chr14g0444761) contain LRR domains and may serve as receptors during pathogenesis. Our analysis of conserved domains also revealed that HannXRQ_Chr01g0025331 additionally contains Rx and NB-ARC domains, thus suggesting that this protein is a CC-NB-LRR receptor. Previous studies have indicated that the Rx (CC-NB-LRR) protein confers resistance to Potato Virus X (PVX), with the PVX coat protein acting as the AVR determinant [81][82]. Furthermore, Pb1 encodes a CC-NB-LRR protein and confers durable, broad-spectrum resistance to Magnaporthe oryzae in rice [83]. Taking into consideration the gene-for-gene principle, we speculate that HannXRQ_Chr01g0025331 acts as a resistance gene in rice to confer speci c resistance against virulent strain V. dahliae. In addition, an avirulent gene should be present in hypovirulent strain Vn-1 genome but not in virulent strain V. dahliae V33.
In this study, we predicted 33 resistance and 160 susceptibility genes. The much larger proportion of susceptibility genes explains why sun ower cultivar LD5009 is susceptible to Verticillium wilt. This conclusion is consistent with the results of a resistance gene identi cation study performed by Cao et al [84]. Further functional veri cation of the resistance and susceptibility genes predicted in this study is needed.

Methods
Strain culture conditions Virulent (V. dahliae V33) and hypovirulent (G. nigrescens Vn-1) strains were both isolated from sun ower and grown on potato dextrose agar at 26°C in darkness.

Sun ower seedling cultivation and inoculation
The susceptible sun ower variety LD5009 (Beijing Kafry Technology Co., Beijing, China) was used for inoculation. Three sun ower seeds were sown per pot in plastic pots (height × diameter, 10 cm × 12 cm). The pots were maintained under 16-h light/8-h dark conditions for 15 days. The relative humidity and temperature (day/night) was kept at 60%/65% and 28°C/26°C, respectively.
To prepare inoculum, ve mycelial plugs (5 mm in diameter) of virulent and hypovirulent strains were excised from 15-day-old colonies, placed in conical asks with 200 mL Czapek liquid medium, and cultured in a shaking incubator at 100 rpm for 10 days at 26°C [40]. Conidia were then collected from the medium by centrifugation for 20 min at 5,000 g and diluted with sterilized distilled water to a concentration of approximately 1.0 × 10 7 conidia mL -1 . Seedling inoculation was accomplished by incubating the plastic pots on plates containing 50 mL of conidial suspension for 40 min until the suspension was totally absorbed by the seedling substrate (2:1 [v/v] vermiculite:sand). Each replicate consisted of 10 pots, with three replicates used for each strain. Sterilized water distilled water was used as a control.

Analysis of defense enzyme activities and detection of H 2 O 2 deposition
To analyze whether the hypovirulent strain could induce the defense response of sun ower, H 2 O 2 staining was performed at 24 hpi, and defense enzyme activities were monitored at 12, 24, 48, 72, and 96 hpi using an enzyme activity detection kit (Grace Biotechnology, Suzhou, China) according to the manufacturer's instructions.

RNA extraction and qRT-PCR
To analyze the expressions of defense enzyme genes and validate the RNA-seq results, we collected and rapidly froze root samples in liquid nitrogen at 24 hpi. Total RNA was extracted from the samples using a MiniBEST Plant Extraction kit (Takara, Dalian, China) and reverse transcribed using oligo (dT)18 primer and M-MLV reverse transcriptase (Takara). qRT-PCR ampli cations were performed on a Roche LightCycler 96 system (Roche, USA). The 18S gene was used as an internal control. Relative expression levels were calculated by the 2 -ΔΔCt method [85]. Three replicates were performed per biological sample. Primer pairs used in this study are listed in Table S6.

Library preparation for transcriptome sequencing
To analyze sun ower DEGs induced by hypovirulent strain G. nigrescens Vn-1, sterilized water and the virulent strain V. dahliae V33 were used as controls. Total RNA was extracted from sun ower samples at 24 hpi. A cDNA library was prepared after puri cation, terminal repair, A-tailing, sequencing adapter ligation, size selection, and PCR enrichment using a NEBNext Ultra RNA Library Prep kit. Libraries were sequenced on an Illumina HiSeq X Ten RNA-seq platform (Illumina, San Diego, CA, USA).

Data analysis
The sequencing raw data were recorded in a FASTQ le containing sequence information (reads) and corresponding sequencing quality details. Each sample was treated as an independent biological replicate. After removal of adapter-and poly-N-containing reads and reads of low quality from the raw data, Q20 and Q30 values, GC contents, and sequence duplication levels of the clean data were calculated. Transcriptome assembly was accomplished based on the method described by Grabherr et al [86].
All remaining clean reads were mapped to the H. annuus reference genome (accession PRJNA396063 in the NCBI/Resource/Genome database) using TopHat2 software [87]. Sequencing depth and gene fragment lengths were examined based on the expected number of fragments per kilobase (FPKM) [88], with an FPKM value of 1 set as the threshold for determining whether a gene was expressed [89]. A violin plot of FPKM distributions was generated to compare gene expression levels under different conditions. The nal FPKM for each strain was the mean value of three biological replicates.

Differential expression analysis
Differential expression of genes between groups was analyzed using DESeq (1.10.1), an R package that provides statistical routines for determining differential expression in digital gene expression data using a model based on the negative binomial distribution. The resulting P-values were adjusted using Benjamini and Hochberg's approach for controlling the false discovery rate. Genes with an adjusted P-value < 0.05 according to DESeq were considered to be differentially expressed. Declarations XZ and BZD conceived the project, designed the experiment and wrote the manuscript. LJ, DW, YJZ and HYZ revised manuscript. All authors reviewed the manuscript.

Competing interests
The authors declare that they have no competing interests.

Consent for publication
Not applicable          Heat map of predicted Verticillium wilt resistance and susceptibility genes in sun ower Heat maps of 33 predicted resistance genes (B) and 160 predicted susceptibility genes (B) identi ed by comparative transcriptome analysis. Genes upregulated in Vn-1 vs. CK and Vn-1 vs. V33 comparison groups but downregulated or not signi cantly changed in the V33 vs. CK group were predicted to be resistance genes. Genes upregulated in V33 vs. CK and V33 vs. Vn-1 groups but downregulated or unchanged in the Vn-1 vs. CK group were predicted to be susceptibility genes.

Page 26/26
This is a list of supplementary les associated with this preprint. Click to download.