CLINICAL RESEARCH ARTICLE

ITM2A Expands Evidence for Genetic and Environmental Interaction in Graves Disease Pathogenesis

Xiao-Ping Ye,1,2* Fei-Fei Yuan,1,2* Le-Le Zhang,2* Yu-Ru Ma,2 Man-Man Zhang,2 Wei Liu,2 Feng Sun,2 Jing Wu,2 Meng Lu,2 Li-Qiong Xue,2 Jing-Yi Shi,1 Shuang-Xia Zhao,2 Huai-Dong Song,1,2 Jun Liang3,4 and Cui-Xia Zheng,2 for The

China Consortium for the Genetics of Autoimmune Thyroid Disease Downloaded from https://academic.oup.com/jcem/article/102/2/652/2972067 by guest on 27 September 2021

1State Key Laboratory of Medical Genomics, Shanghai Institute of Endocrinology and Metabolism, Ruijin Hospital affiliated to Shanghai Jiaotong University School of Medicine, Shanghai 200025, China; 2Research Center for Clinical Medicine, Department of Respiration and Endocrinology, The Ninth People’s Hospital Affiliated to Shanghai Jiaotong University School of Medicine, Shanghai 200011, China; 3Department of Endocrinology, The Central Hospital of Xuzhou Affiliated to Xuzhou Medical College, Xuzhou, Jiangsu Province 221109, China; and 4Xuzhou Clinical School of Xuzhou Medical College, The Affiliated XuZhou Hospital of Medical College of Southeast University, Xuzhou, Jiangsu Province 221009, China

Context: Graves disease (GD) is a common autoimmune disease triggered by genetic predisposition and environmental factors. However, the mechanisms of interaction between genetic and envi- ronmental factors contributing to the development of GD remain unknown.

Objective: We aimed to identify GD susceptibility variants and on Xq21.1 locus and interpret the contribution of interaction between genetic predisposition on Xq21.1 and environmental factors to GD.

Design: We performed refining study on Xq21.1 in a 2-stage study and carried out expression quantitative trait locus analysis of the best association signal with GD.

Setting and Participants: A total of 4316 GD patients and 4374 sex-matched controls were collected from the Chinese Han population by cooperation with multiple hospitals.

Results: We identified that rs3827440 or its linkage single nucleotide polymorphisms (SNPs) were probably the causal variant in the Xq21.1 locus, with the most substantial association with GD in our 2 combined cohorts (P = 2.45 3 10 15). The genotypes of rs3827440 were correlated with the ex- pression of ITM2A in monocytes and peripheral blood mononuclear cells (PBMCs) from healthy volunteers. Notably, the expression of ITM2A in monocytes after lipopolysaccharide (LPS) and interferon-g (INF-g) stimulation showed substantial difference among the volunteers that carried different genotypes of rs3827440 (P = 9.40 3 1027 and P = 1.26 3 1025 for 24 hours’ LPS and INF-g stimulation, respectively). Moreover, ITM2A expression was significantly decreased in PBMCs from untreated GD patients than that from controls.

Conclusion: The results suggest that ITM2A might be a susceptibility for GD in the Xq21.1 locus, and environmental factors, such as viral and bacterial infections, probably contribute to GD pathogenesis by interacting with the risk SNP rs3827440 mediating the regulation of ITM2A expression. (J Clin Endocrinol Metab 102: 652–660, 2017)

ISSN Print 0021-972X ISSN Online 1945-7197 *These authors contributed equally to this study.

Printed in USA Abbreviations: CI, confidence interval; CT, cycle threshold; EBV, Epstein-Barr virus; eQTL, Copyright © 2017 by the Endocrine Society expression quantitative trait locus; eSNP, expression SNP; GD, Graves disease; GWAS, Received 9 July 2016. Accepted 24 October 2016. genome-wide association study; HCV, hepatitis C virus; IFN, interferon; LPS, lipopoly- First Published Online 3 November 2016 saccharide; OR, odds ratio; PBMC, peripheral blood mononuclear cell; PCR, polymerase chain reaction; SNP, single nucleotide polymorphism; TSHR, thyroid-stimulating hormone receptor; Th, T-helper; YE, Yersinia enterocolitica.

652 press.endocrine.org/journal/jcem J Clin Endocrinol Metab, February 2017, 102(2):652–660 doi: 10.1210/jc.2016-2625 doi: 10.1210/jc.2016-2625 press.endocrine.org/journal/jcem 653

raves disease (GD) is a common and complex au- found many gene expressions and their responses to LPS Gtoimmune disease caused by interaction of genetic or IFN-g stimulation showed substantial difference factors and environmental triggers (1). The prevalence of among subjects carrying different SNP genotypes (cis- GD in the Chinese Han population is approximately expression quantitative trait locus [cis-eQTL] genes) (17). 0.25% to 1.0% with a strong female preponderance, The identification of functional regulatory variants and consistent with that of many other autoimmune diseases associated modulated genes is key to interpreting GWAS (2,3). However, the reason that GD is more prevalent findings and establishing how genes are associated with among women is unknown. A reasonable explanation disease (18). The levels of LPS and IFN-g are increased in is that genes located on the X- may be humans when infected by bacteria or viruses; therefore, contributed to the development of GD. Studies have those findings by Fairfax et al. laid the foundation for reported that the Xp11 region is associated with auto- our investigation into the role of interaction between Downloaded from https://academic.oup.com/jcem/article/102/2/652/2972067 by guest on 27 September 2021 immune thyroid disease in the United Kingdom pop- genetic and environmental factors, such as infections, in ulation and that transcription levels of the forkhead the pathogenesis of GD. box P3 gene (FOXP3) in Xp11 are associated with In the current study, we performed refining study on susceptibility to autoimmune thyroid disease in the Cau- Xq21.1 in a large-scale sample of cohorts including 4316 casians but not in the Japanese and Chinese Han pop- GD patients and 4374 controls and identified rs3827440 215 ulations (4–6). (PCombined =2.453 10 ) was the best SNP associated In our previous genome-wide association study (GWAS), with GD on Xq21.1. Interestingly, the levels of ITM2A we confirmed 4 GD-associated loci previously reported and expression in monocytes after either LPS or IFN-g stimu- identified 2 susceptibility loci in the Chinese Han pop- lation showed substantial difference among the individuals ulation (7). Another extension of this study identified 5 GD who carried distinct genotypes of rs3827440. Moreover, risk loci (8). In the extension study, rs5912838, located the transcription levels of ITM2A in peripheral blood between the immune receptor G –coupled receptor mononuclear cells (PBMCs) were down-regulated in un- 174 gene (GPR174) and integral membrane protein 2A treated GD patients when compared with those of controls. gene (ITM2A), was identified as an important association These findings suggest that environmental triggers, such signal on Xq21.1 in the Chinese Han population. ITM2A, as infection, indicated by LPS or IFN-g stimulation, prob- but not GPR174, escaping X-chromosome inactivation, ably contribute to the development of GD by interacting was arbitrarily considered a susceptibility gene for GD (8). with risk variant rs3827440, mediating the regulation of Moreover, rs3827440, a nonsynonymous single nucleotide ITM2A expression. polymorphism (SNP) in GPR174 highly linked with rs5912838 on Xq21.1, has been identified associated with Materials and Methods GD in 2 independent studies. One was performed in a Caucasian population and the other was in another group Subjects and sample collection of the Chinese Han population (9,10); moreover, rs3827440 All subjects in the current study were from the Chinese Han was also found to be associated with autoimmune Addison population by cooperation with multiple hospitals in China disease in a UK cohort (11). Therefore, the Xq21.1 region (19). This program was approved by the local institutional review board. As previously reported, 1536 patients with GD was considered a susceptibility locus to GD, but the sus- and 1516 sex-matched controls were recruited in the initial ceptibility of SNPs or genes in this region is controversial. GWAS stage (7). Next, an additional 2874 GD patients and Infections by Yersinia enterocolitica (YE), hepatitis 2906 sex-matched controls were collected for the replication C virus (HCV), and Epstein-Barr virus (EBV) have been study. Diagnosis of GD was based on the principles previously frequently cited as major environmental triggers of GD reported (7). Genomic DNA of all subjects was extracted from (12–16). However, many of the individuals infected by YE, peripheral blood leukocytes by using FUJIFILM QuickGene- 610 system (Kurabo Biomedical, Tokyo, Japan). HCV, or EBV were not suffering from GD; this may be explained by their interaction with different genetic pre- Genotyping and quality control dispositions. Although the interaction between genetic GWAS was performed by Illumina Human660-Quad Bead predisposition and environmental factors contributed to Chips in the Chinese National Center the pathogenesis of GD, the evidence supporting the (Shanghai, China). We then obtained the imputed SNPs by hypothesis was scarce. performing imputation analysis on IMPUTE2 based on the Most recently, Fairfax et al. investigated correlation 1000 Genomes Project (June 2011) (20). All of the genotyped and imputed SNPs were filtered by stringent quality control. of SNP genotypes in the whole genome with the levels 2 SNPs with a Hardy-Weinberg equilibrium P . 1 3 10 6 and of gene transcription in monocytes recruited from 432 missing call rates #0.05 and minor allele frequency . 0.01 were healthy Europeans under lipopolysaccharide (LPS), in- included, and the SNPs with sex inconsistencies and cryptic terferon-g (IFN-g) stimulation, or na¨ıve state (17). They relatedness were excluded (8). In the replication stage, SNP 654 Ye et al ITM2A Contributes to Graves Disease Pathogenesis J Clin Endocrinol Metab, February 2017, 102(2):652–660

genotyping was performed by TaqMan SNP Genotyping Assay used as an internal reference. The DCT values for AA and CC of on ABI Vii Real-Time polymerase chain reaction (PCR) system. each individual were calculated by doubling the DCT values obtained for AA or CC. The DCT value for AC was regarded as D D SNP selection and sample size calculation AA or CC, and each CT value was determined by the CT D According to the results in the GWAS stage, we defined the value obtained for AA or CC average plus the AC actual CT chromosomal region by HapMap recombination rates, GWAS value. We used the same primer sequences for ITM2A and data, and linkage disequilibrium information (21). Then we GPR174 as previously reported (8). 3 24 found 272 SNPs with PGWAS # 1 10 (Supplemental Ta- Statistical analysis ble 1). By using Haploview, these 272 SNPs could be captured SNPs in chromosome X in the GWAS stage were analyzed by by 9 tag SNPs with r2 . 0.8; tag SNPs were genotyped in the using logistic regression in PLINK and adding sex as a covariate second replication stage. To detect the genome-wide substantial (22). In the replication and combined cohorts, tag SNPs were difference of the genotyped SNPs at 80% power, the sample size Downloaded from https://academic.oup.com/jcem/article/102/2/652/2972067 by guest on 27 September 2021 also analyzed by logistic regression analysis. Forward and in the replication stage was calculated by QUANTO (version 2-locus logistic regression analyses were performed to determine 1.2.4). In the current study, we therefore included 4316 GD the independent effect and main effect of an individual SNP by patients and 4374 controls in the combined stage. combining the R software package and PLINK (23). Genome- ITM2A wide P value plot was determined by the R package, and the The correlation analysis of eSNPs for and GD- regional plots were generated by SNAP version 2.2 software associated SNPs and modified by the R software package. eQTL analysis was performed in the eQTL gene database established by Fairfax et al. (17). In brief, they detected the Results genotype of 609,704 SNPs and transcriptome data in 432 healthy Europeans. The monocytes were separated from Refining association study in the Xq21.1 region for PBMCs by magnetic-activated cell sorting (Miltenyi). Monocytes GD in the Chinese Han Population separated from 432 individuals were treated by 20 ng/ml INF-g or In our previous study, 1536 GD cases and 1516 sex- 20 ng/ml LPS for 24 hours, respectively. After quality control, the data from 367 samples stimulated by 20 ng/ml INF-g and 322 matched controls were genotyped at 657,366 SNPs using individuals treated with 20 ng/ml LPS were further analyzed in the Illumina Human660-Quad BeadChips at the GWAS study by Fairfax et al. Moreover, monocytes from 261 individuals stage (8,9). After stringent quality control, 486,049 SNPs were stimulated by 20 ng/mL LPS for 2 hours (17). Data across all 4 in 1442 GD cases and 1468 sex-matched controls were conditions were available from 228 individuals, permitting cross- analyzed (8,9) (Table 1). In the initial GWAS stage, treatment comparison (17). By searching the cis-eQTL database, 2 rs5912838 (P = 4.15 3 10 8; odds ratio [OR], 1.37; we investigated the expression SNPs (eSNPs) for P2RY10, GWAS GPR174,andITM2A in the GD susceptibility locus Xq21.1. 95% confidence interval [CI], 1.22 to 1.53) on GPR174- Furthermore, the association of these eSNPs with Graves disease ITM2A at Xq21.1 exhibited the most important differ- was analyzed based on our GWAS data. ence in our genotyped SNPs (8). An approximate 990-kb region on Xq21.1 covering rs5912838 was separated by PBMCs and real-time PCR 2 recombination hotspots (the recombination rate of the Blood samples (100 mL) were donated by 191 healthy right hotspot was .40%; the rate of the left hotspot was Chinese Han volunteers and 56 untreated GD patients for gene about 20%), and this region included 3 annotated genes: expression analysis. The blood samples of 56 GD patients were collected when these patients were first diagnosed with GD and P2RY10, GPR174, and ITM2A [Fig. 1(A)]. Using our did not receive any treatment before. PBMCs were collected by previous GWAS data, we analyzed the genotyped and using Ficoll according to the manufacturer’s instructions. The imputed SNPs on Xq21.1 by logistic regression analysis. rs3827440 and rs5912838 genotypes among the 191 healthy Fourteen genotyped SNPs and 258 imputed SNPs with 2 volunteers were determined by TaqMan SNP Genotyping Assay P #1 3 10 4 were identified in this linkage dis- on an ABI Vii Real-Time PCR system. Expression of the target GWAS genes in PBMCs was determined by quantitative real-time PCR. equilibrium block (Supplemental Table 1). The relative expressions of genes of interest were calculated by Our previous study demonstrated that rs5912838, the comparative cycle threshold (CT) method, and GAPDH was located approximately in the 350-kb region on Xq21.1,

Table 1. Sample Set Descriptions

Genotyping Stage Disease State No. Sex Ratio (M/F) Age GWAS GD 1442 335/1107 39 6 14 Control 1468 359/1109 45 6 9 Replication GD 2874 379/2495 40 6 14 Control 2906 555/2351 46 6 15 Combined GD 4316 714/3602 39 6 14 Control 4374 914/3460 45 6 13

Abbreviations: F, female; M, male. doi: 10.1210/jc.2016-2625 press.endocrine.org/journal/jcem 655

OR, 1.31; PForward, 0.0078) was indicated an independent signal associated with GD by forward logistic regression analysis. To further refine the causal SNPs associated with GD on Xq21.1, we selected 9 tagSNPs, which were able to 24 capture the 272 SNPs with PGWAS # 1 3 10 in this locus (r2 . 0.8), for genotyping in the replication stage in a cohort of 2874 GD patients and 2906 controls (Supplemental Table 2). Rs5912838 and rs3827440 were also selected for genotyping because they were previously reported to be significantly associated with GD in the Downloaded from https://academic.oup.com/jcem/article/102/2/652/2972067 by guest on 27 September 2021 Chinese Han population (8,9). Among these 11 selected SNPs, rs3827440 was in high linkage with rs5912838, rs6619800, and rs5912203 (r2 . 0.8) [Fig 1(C)], and rs1736642 was excluded because of a failed probe syn- thesis; thus, 10 SNPs in this locus were further genotyped in the replication stage. After quality control, rs3827440 was the best association signal with GD in the replication 211 stage (PReplication = 4.40 3 10 ; OR, 1.29; 95% CI, 1.20 to 1.39). Meanwhile, in the combined database with 4316 GD cases and 4374 controls, rs3827440 was confirmed as having the most substantial association 215 (PCombined = 2.45 3 10 ; OR, 1.30; 95% CI, 1.22 to 1.39) (Table 2). Additionally, rs1736645 exhibited the

second strongly association signal (PCombined = 5.00 3 2 10 13; OR, 1.27; 95% CI, 1.19 to 1.36) after excluding rs3827440 and its highly linked SNPs, rs5912203 and rs5912838. To further determine the causal variants for GD in this region, we performed forward regression analysis based on the combined database. Consequently, rs3827440

(PForward = 0.001) was found to be an independent GD- associated signal after excluding its highly linked SNPs, rs5912838, rs6619800, and rs5912203. Meanwhile,

rs1736645 (PForward = 0.009) seems to be another in- dependent variant associated with GD on Xq21.1. Moreover, the regression models of all tag SNPs in this region could be improved at the level of P , 0.05 by Figure 1. Regional plots of association results, logistic regression analysis, and linkage disequilibrium block analysis. (A) The GD- rs3827440 using 2-locus conditional logistic regression associated SNPs in the Xq21.1 region in the GWAS stage. The color analysis (Supplemental Fig.). However, all SNPs except of each SNP dot reflects its r2 value, with the top genotyped SNP, rs1736645 (P = 0.012) could not improve the regression rs5912838, represented by the purple dot. (B) Two-locus logistic models at the level of P , 0.05 if rs3827440, as the best regression for rs5912838 on Xq21.1 in the GWAS stage. P values of the other SNPs on Xq21.1 after conditioning on rs5912838 are marker in the Xq21.1 region, was put into these models represented by blue triangles. P values of rs5912838 after (Supplemental Fig.). All of these data indicated that conditioning on the other SNPs are represented by red dots. (C) rs3827440 or its highly linked SNPs were probably the Linkage disequilibrium block analysis for the 10 selected SNPs. Values shown are the r2 values between 2 SNPs. best variants in the Xq21.1 region for the susceptibility to GD. was the signal most strongly associated with GD. In the The effects of rs3827440 on GPR174 and ITM2A current study, rs5912838 was again the signal most expression in monocytes and PBMCs significantly associated with GD among the 272 SNPs Rs3827440 was a nonsynonymous SNP in GPR174 evaluated by 2-locus conditional logistic regression analysis and was located near the 30-untranslated region of 26 [Fig. 1(B)]. In addition, rs1751107 (PGWAS =2.283 10 ; ITM2A (distance, 196 kb); thus, the correlation of 656 Ye et al ITM2A Contributes to Graves Disease Pathogenesis J Clin Endocrinol Metab, February 2017, 102(2):652–660

Table 2. Association Results of Tag SNPs on Xq21.1 in the 2-Stage Population

GWAS (1442 vs 1468) Replication (2874 vs 2906) Combined (4316 vs 4374)

Risk Case Control Case Control Case Control SNP Chr. Position Allele (RAF) (RAF) P OR 95% CI (RAF) (RAF) P OR 95% CI (RAF) (RAF) P OR 95% CI rs5912203 78258106 T 0.63 0.56 3.47 3 1027 1.34 1.20–1.50 0.63 0.57 1.31 3 10210 1.28 1.19–1.39 0.63 0.57 8.50 3 10215 1.30 1.21–1.39 rs12687858 78416403 C 0.72 0.67 1.87 3 1025 1.30 1.15–1.46 0.72 0.67 7.27 3 1028 1.25 1.15–1.35 0.72 0.67 8.21 3 10211 1.26 1.18–1.35 2 2 2 rs3827440 78426988 T 0.63 0.56 1.64 3 10 7 1.35 1.21–1.51 0.63 0.57 4.41 3 10 11 1.29 1.20–1.39 0.63 0.57 2.45 3 10 15 1.30 1.22–1.39 rs5912838 78497118 A 0.64 0.56 4.15 3 1028 1.37 1.22–1.53 0.64 0.58 1.44 3 1029 1.26 1.17–1.36 0.64 0.58 1.69 3 10214 1.29 1.21–1.38 rs5912849 78524009 A 0.66 0.60 8.45 3 1026 1.29 1.15–1.44 0.65 0.62 1.56 3 1024 1.16 1.07–1.25 0.65 0.61 3.08 3 1028 1.20 1.13–1.29 rs6619800 78526120 A 0.61 0.53 1.60 3 1027 1.35 1.21–1.51 0.59 0.54 6.62 3 1027 1.21 1.12–1.30 0.59 0.53 6.94 3 10212 1.25 1.17–1.34 rs5959298 78557332 A 0.71 0.65 6.44 3 1025 1.27 1.13–1.43 0.70 0.66 1.12 3 1025 1.20 1.10–1.29 0.70 0.66 1.35 3 1028 1.22 1.14–1.31 rs6619915 78587587 A 0.50 0.44 2.62 3 1025 1.27 1.14–1.42 0.50 0.45 7.22 3 1027 1.21 1.12–1.30 0.50 0.45 3.91 3 10211 1.24 1.17–1.33 26 29 213 rs1736645 78597195 C 0.60 0.54 4.44 3 10 1.30 1.16–1.46 0.61 0.55 6.93 3 10 1.25 1.16–1.35 0.61 0.55 5.00 3 10 1.27 1.19–1.36 Downloaded from https://academic.oup.com/jcem/article/102/2/652/2972067 by guest on 27 September 2021 rs75810848 78609329 C 0.49 0.43 1.47 3 1025 1.29 1.15–1.44 0.49 0.44 1.87 3 1026 1.20 1.11–1.29 0.49 0.44 4.76 3 10210 1.23 1.15–1.31

Abbreviations: Chr., chromosome; RAF, risk allele frequency. Bold type indicates 2 independent GD-associated SNPs.

GPR174 and ITM2A expression in monocytes or Environmental factors: viral and bacterial infections PBMCs with rs3827440 was further investigated. In- contributed to GD by interacting with the terestingly, through cis-eQTL analysis using tran- susceptibility gene ITM2A scriptome data from the monocytes of 432 Caucasian Cumulating evidence has shown that infections by individuals, we found that the expression level of bacteria and viruses are the major environmental triggers ITM2A in monocytes was correlated with rs3827440 for GD, but little is known about the mechanisms 2 (P =2.483 10 4; Table 3) (17). Rs3827440 showed (12–16). Most recently, the cis-eQTL database developed no eQTL effect on the expressions of P2RY10 and by Fairfax et al. (19) collected the results regarding the GPR174, which were also located in the susceptibility correlation of SNP genotypes with the transcriptome in locus of GD Xq21.1. We then evaluated allele-specific monocytes after LPS and IFN-g stimulation recruited effects of rs3827440 on the expression of ITM2A and from 432 healthy Europeans (19). Monocytes mediate GPR174 in PBMCs from 191 healthy volunteers. The innate immune responses in humans and may be acti- results showed that the rs3827440 genotype was cor- vated by LPS and IFN-g after infection. In the current related with the expression of ITM2A in PBMCs (P = study, we found that the rs3827440 genotypes were 0.008), but not with that of GPR174 (P = 0.114) [Fig. correlated with the expression of ITM2A in both 2(A)]. It has been reported that rs5912838, highly linked monocytes and PBMCs. Thus, it is tempting to assume to rs3827440 (r2 = 0.92), was the best GD-associated that the interaction between infections and risk variant signal on Xq21.1 (8). We also detected the allele- rs3827440 contributed the susceptibility to GD by reg- specific effects of rs5912838 on ITM2A expression ulating expression of ITM2A in monocytes, which is a and found that the genotypes of rs5912838 were also susceptibility gene for GD in the Xq21.1 region. Notably, correlated with the expression of ITM2A in PBMCs by searching the cis-eQTL database, among the 3 an- (P = 0.008) [Fig. 2(B)]. These data suggested that notated genes on Xq21.1, P2RY10, GPR174,and rs3827440 or its linked SNPs probably contributed ITM2A, only ITM2A expression in monocytes was found the susceptibility to GD by regulating the expression to be regulated by SNPs in this region. We identified 63 of ITM2A. SNPs on Xq21.1 correlated with ITM2A expression in

Table 3. Stimulated-eQTL Results of ITM2A and GWAS

SNP SNP Pos. P GWAS P LPS2 P LPS24 P IFN24 P Na¨ıve Value Min. Dataset rs5912724 78305516 4.39 3 1027 NA 2.11 3 1024 1.42 3 1023 2.70 3 1023 LPS24 rs5912792 78425636 1.61 3 1027 8.94 3 1023 9.40 3 1027 1.26 3 1025 2.48 3 1024 LPS24 rs3810712 78426471 1.85 3 1027 8.69 3 1023 1.19 3 1026 2.06 3 1025 3.32 3 1024 LPS24 rs3810711 78426488 1.61 3 1027 8.94 3 1023 9.40 3 1027 1.26 3 1025 2.48 3 1024 LPS24 rs3827440 78426988 1.64 3 1027 8.94 3 1023 9.40 3 1027 1.26 3 1025 2.48 3 1024 LPS24 rs5912838 78497118 4.15 3 1028 6.87 3 1023 2.04 3 1026 3.20 3 1025 2.64 3 1024 LPS24 rs5959298 78557332 6.44 3 1025 1.43 3 1023 1.49 3 1024 3.08 3 1023 NA LPS24 2 2 2 2 rs1751107 78601102 2.28 3 10 6 4.90 3 10 5 NA 2.74 3 10 4 2.59 3 10 4 LPS2

Abbreviations: Na¨ıve, monocytes without LPS or IFN treatment; P IFN-g 24 value, P value of monocytes treated with IFN-g for 24 hours; P LPS2 value, P value of each SNP-regulated ITM2A gene expression in monocytes treated with LPS for 2 hours; P LPS24 value, P value of monocytes treated with LPS for 24 hours; Pos., position. Bold type indicates 2 independent GD-associated SNPs. doi: 10.1210/jc.2016-2625 press.endocrine.org/journal/jcem 657

Figure 2. Association of the transcription levels of ITM2A and GPR174 and rs3827440 and rs5912838. (A) The transcription level of ITM2A but

not GPR174 was regulated by susceptibility SNP rs3827440 in PBMCs from healthy individuals (TT, n = 51; TC, n = 60; CC, n = 78). (B) The Downloaded from https://academic.oup.com/jcem/article/102/2/652/2972067 by guest on 27 September 2021 transcription level of ITM2A but not GPR174 was regulated by susceptibility SNP rs5912838 in PBMCs from healthy individuals (AA, n = 74; AC, n = 59; CC, n = 55). (C) Transcription levels of ITM2A and GPR174 in PBMCs from 56 untreated GD patients were significantly decreased compared with that in 191 healthy individuals PBMCs. Data are presented as the mean 6 standard error of the mean. **P # 0.01; ***P # 1 3 1026. Statistical analysis was performed by 1-way analysis of variance. mRNA, messenger RNA. monocytes after LPS and IFN-g stimulation with different The possible causal SNP rs3827440 on Xq21.1 was a genotypes. SNPs that affect gene expression were referred nonsynonymous SNP in GPR174, and its genotypes were to as eSNPs. Among these 63 eSNPs, 50 SNPs were in- reported to be associated with GPR174 messenger RNA cluded in our GWAS data after quality control, but only 8 levels in PBMCs (9). However, in the current study, the

SNPs were associated with GD at the level of PGWAS # transcription level of ITM2A but not GPR174 in PBMCs 2 1 3 10 4 (Fig. 3; Table 3; Supplemental Table 3). from volunteers without thyroid disease was regulated by Moreover, compared with the correlations among these 8 the different alleles of rs3827440 and its linked SNP eSNPs and ITM2A expression in the monocytes without rs5912838. Moreover, by searching the cis-eQTL data- LPS or IFN-g stimulation, the correlations between these base, ITM2A expression in monocytes was found to be eSNPs with ITM2A expression in monocytes stimulated regulated by rs3827440 as well as its linkage SNPs such by LPS or IFN-g for 24 hours were dramatically as rs5912838, rs5912724, rs5912792, rs3810712, and strengthened indicated by the smaller P value (Table 3). rs3810711. All the data suggested that rs3827440 and its Interestingly, rs3827440, as well as its highly linked SNPs high linkage SNPs probably contributed to the suscep- such as rs5912838, rs5912724, rs5912792, rs3810712, tibility to GD on Xq21.1 by regulating the expression of and rs3810711 (r2 . 0.9; Supplemental Table 4), showed ITM2A, but not GPR174, in monocytes. More recently, the most substantial correlation with ITM2A expression GWAS have had provided more success in identifying in monocytes stimulated by LPS and IFN-g for 24 hours genetic variants associated with GD. Before now, more (Table 3). Conversely, only rs1751107 was most sig- than 20 loci were well-known as having a creditable nificantly associated with ITM2A expression in mono- susceptibility to GD (7,8,19,24). However, among the cytes after LPS stimulation for 2 hours. Among these 8 GD-associated genes in the non-MHC region, ITM2A eSNPs associated with GD, rs5959298, conditioned by showed a strong association with GD (OR, 1.30), similar rs3827440 in the combined cohort of our results, did not to that of TSHR (OR, 1.35) and CTLA4 (OR, 1.30) in the affect ITM2A expression under na¨ıve state (Table 3). Chinese Han population (7). Interestingly, we also found that ITM2A and GPR174 Viral and bacterial infections, especially YE, function expression was dramatically decreased in PBMCs from as triggers in the pathogenesis of GD (12). YE outer untreated GD patients compared with PBMCs from membrane porin F protein shares cross-immunogenicity 2 control subjects (P = 4.73 3 10 7 for ITM2A and P = with a leucine-rich domain of thyroid-stimulating hor- 2 4.36 3 10 10 for GPR174, respectively) [Fig. 2(C)]. mone receptor (TSHR) and is recognized by an antihu- man TSHR antibody. Immunization of mice with YE Discussion outer membrane porin F protein can induce the secretion of TSHR autoantibodies (12,13). Moreover, HCV and Previous studies have provided evidence that Xq21.1 is a EBV are the most important candidate viruses for auto- susceptibility locus for GD in the Chinese Han population immune thyroid disease (25). Clinically, IFN therapy against (8). Here, by using our GWAS data and tag SNP strategy, HCV infection leads to autoimmune thyroiditis (26); we performed a refining study to identify GD suscepti- however, the relationship between the previously mentioned bility SNPs and genes in this region; we found that mechanisms and the development of GD remains merely an rs3827440 was the best SNP in the Xq21.1 region with the attractive hypothesis. The connection, if any, between en- 2 most important association with GD (P =2.453 10 15). vironmental triggers and genetic factors remains elusive. 658 Ye et al ITM2A Contributes to Graves Disease Pathogenesis J Clin Endocrinol Metab, February 2017, 102(2):652–660

variants (18). Interestingly, we found that ITM2A ex- pression in monocytes was regulated by 63 specific SNPs located on the Xq21.1 locus. Among these, 8 SNPs, in- cluding rs3827440 and rs1751107, were significantly associated with GD. Meanwhile, rs3827440, as well as its highly linked SNPs, showed the most substantial corre- lation with ITM2A expression in monocytes subjected to LPS and IFN-g stimulation for 24 hours. Based on these data, we propose that the contribution of rs3827440 to GD susceptibility is most likely mediated by the response of ITM2A expression in monocytes subjected to LPS and Downloaded from https://academic.oup.com/jcem/article/102/2/652/2972067 by guest on 27 September 2021 IFN-g stimulation. Therefore, we hypothesize that the interaction of infections, such as bacterial and viral in- fections, with genetic variants, such as rs3827440, affects the expression of susceptibility gene ITM2A in monocytes, contributing to the pathogenesis of GD. ITM2A was identified as a target gene of GATA-3, which is critical for the development of CD4+ Thelper (Th) cells, Th2 cell differentiation, regulatory T-cell func- tion (27), and CD8+ cytolytic T-cell (Tc cell) activation (11), and homeostasis (28–31). ITM2A is induced during major histocompatibility complex–mediated positive selection of double-positive thymocytes (31,32). Peripheral na¨ıve Th and Tc cells express a low level of ITM2A, which could be transiently induced after T-cell activation. Meanwhile, ITM2A expression is significantly higher in effector Th cells than in regulatory T cells (33,34). ITM2A overexpression in transgenic mice causes downregulation of CD8 in CD4+CD8+ double-positive thymocytes. Furthermore, ITM2A deficiency in mice results in an attenuated Th- cell–dependent humoral immune response and a partial Figure 3. Regional plots of association results of GWAS and defect in the transition from CD4+CD8dull to CD8 single- stimulus-specific eQTLs of ITM2A. (A) GD-associated SNPs in the Xq21.1 region in the GWAS stage; rs3827440, rs5912838, positive stage (28). Considering that GD is an autoimmune rs1736645, and rs1751107 are shown as indicated by the arrows. disease and that CD8+ T cells are diminished in both the (B) Total eQTL results of ITM2A on Xq21.1. The P values of 63 peripheral and thyroid tissues of patients with GD, leading ITM2A eSNPs regulating expression in monocytes with or without + + LPS or IFN-g stimulation. (C) The effect of 8 GD-associated SNPs to an increase in the ratio of CD4 /CD8 (35, 36), ITM2A is 24 with PGWAS # 1 3 10 on the expression of ITM2A in monocytes probably involved in the immunologic pathogenesis of GD. P with or without LPS or IFN-g treatment. GWAS of each SNP is In conclusion, the current study uncovered that shown as a black reverse triangle. LPS2, monocytes stimulated with 20 ng/mL LPS for 2 hours; LPS24, monocytes stimulated with 20 ng/mL rs3827440 was the best association signal with GD by LPS for 24 hours; IFN-g, monocytes stimulated with 20 ng/mL IFN-g conducting a refining study on Xq21.1 based on our for 24 hours; na¨ıve, monocytes without any treatment. previous GWAS data. Meanwhile, the levels of ITM2A expression in monocytes after LPS and IFN-g stimulation As mentioned, Fairfax et al. established the cis-eQTL showed substantial difference among healthy Europeans database that catalogues genetic variants in the whole carrying different genotypes of rs3827440. These data genome that regulate messenger RNA expression in suggest that ITM2A was probably a susceptibility gene to monocytes after LPS and IFN-g stimulation. LPS is a GD on Xq21.1 and also that environmental factors, such well-known bacterial product, and virus infections in- as viral and bacterial infections, likely contribute to GD duce an increase of cytokines, such as IFN in vivo.Both pathogenesis by interaction with susceptibility SNP IFN and LPS activate the innate , trig- rs3827440 or its high linkage SNPs to regulate ITM2A gering activity in monocytes; thus, the cis-eQTL data- expression in monocytes. This study reports that how base provided an opportunity to investigate how infections, genetic and environmental factors interact to contribute as environmental factors, trigger development of autoim- to the development of GD and provide evidence to in- mune diseases such as GD by interaction with genetic terpret GWAS data. doi: 10.1210/jc.2016-2625 press.endocrine.org/journal/jcem 659

Acknowledgments Autoimmune Thyroid Disease. A genome-wide association study identifies two new risk loci for Graves’ disease. Nat Genet. 2011;43: Address all correspondence and requests for reprints to: Huai- 897–901. Dong Song, MD, PhD, Department of Endocrinology and 8. Zhao SX, Xue LQ, Liu W, Gu ZH, Pan CM, Yang SY, Zhan M, Wang HN, Liang J, Gao GQ, Zhang XM, Yuan GY, Li CG, Du Molecular Biology, State Key Laboratory of Medical Genomics, WH, Liu BL, Liu LB, Chen G, Su Q, Peng YD, Zhao JJ, Ning G, Shanghai Institute of Endocrinology and Metabolism, Ruijin Huang W, Liang L, Qi L, Chen SJ, Chen Z, Chen JL, Song HD. Hospital affiliated to Shanghai Jiaotong University School of China Consortium for the Genetics of Autoimmune Thyroid Dis- Medicine, Shanghai 200025, China. Email: huaidong_s1966@ ease. Robust evidence for five new Graves’ disease risk loci from a 163.com. staged genome-wide association analysis. Hum Mol Genet. 2013; 22:3347–3362. We are grateful to all volunteers who participated in this 9. Chu X, Shen M, Xie F, Miao XJ, Shou WH, Liu L, Yang PP, Bai study. YN, Zhang KY, Yang L, Hua Q, Liu WD, Dong Y, Wang HF, Shi This work was supported in part by the National Natural JX, Wang Y, Song HD, Chen SJ, Chen Z, Huang W. An X Downloaded from https://academic.oup.com/jcem/article/102/2/652/2972067 by guest on 27 September 2021 Science Foundation of China (Grants 81270863, 81430019, chromosome-wide association analysis identifies variants in ’ 50 81370888, 31571296, 81472177, 81270624, 81400776, GPR174 as a risk factor for Graves disease. J Med Genet. 2013; : 479–485. and 31501015), Shanghai Municipal Commission of Health 10. Szyma´nski K, Miskiewicz P, Pirko K, Jurecka-Lubieniecka B, Kula D, and Family Planning Outstanding Young Scholars (Grant Hasse-Lazar K, Krajewski P, Bednarczuk T, Ploski R. rs3827440, a XYQ2013086), Shanghai Science and Technology Committee nonsynonymous single nucleotide polymorphism within GPR174 Youth Sailing Program (Grant 14YF1411800), Natural Science gene in , is associated with Graves’ disease in Polish Foundation of Jiangsu Province (Grant BK2012575), and Caucasian population. Tissue Antigens. 2014;83:41–44. Jiangsu Provincial Special Program of Medical Science (Grant 11. Napier C, Mitchell AL, Gan E, Wilson I, Pearce SH. Role of the X-linked gene GPR174 in autoimmune Addison’s disease. J Clin BL2014027). Endocrinol Metab. 2015;100:E187–E190. Author contributions: H.D.S., C.X.Z., and J.L. conceived and 12. Wenzel BE, Heesemann J, Wenzel KW, Scriba PC. Antibodies to designed the project. X.P.Y., C.X.Z., and J.L. contributed to the plasmid-encoded of enteropathogenic Yersinia in patients project management. X.P.Y., F.F.Y., J.W., M.L., Y.R.M., F.S., with autoimmune thyroid disease. Lancet. 1988;1:56. and M.M.Z. contributed to the replication genotyping. F.F.Y., 13. Wang Z, Zhang Q, Lu J, Jiang F, Zhang H, Gao L, Zhao J. Identification of outer membrane porin f protein of Yersinia L.L.Z., and Y.R.M. took part in the statistical analysis and enterocolitica recognized by antithyrotopin receptor antibodies in expression quantitative trait locus analysis. W.L., L.Q.X., J.Y.S, Graves’ disease and determination of its epitope using mass spec- S.X.Z., and J.L. took part in the collection of clinical sam- trometry and bioinformatics tools. J Clin Endocrinol Metab. 2010; ples, extracted DNA, and undertook sample quality control. 95:4012–4020. X.P.Y., H.D.S., and C.X.Z. wrote the manuscript. 14. Trikudanathan GV, Ahmad I, Israel JL. Concurrent autoimmune ’ Disclosure Summary: The authors have nothing to disclose. hepatitis and grave s disease in hepatitis C during pegylated in- terferon alpha-2a and ribavirin therapy. Saudi J Gastroenterol. 2011;17:348–352. References 15. Brix TH, Hansen PS, Hegedus L, Wenzel BE. Too early to dismiss Yersinia enterocolitica infection in the aetiology of Graves’ disease: 1. Menconi F, Marcocci C, Marino M. Diagnosis and classification evidence from a twin case-control study. Clin Endocrinol (Oxf). 69 – of Graves’ disease. Autoimmun Rev. 2014;13:398–402. 2008; :491 496. 2. Burch HB, Cooper DS. Management of Graves disease: a review. 16. Luo G, Seetharamaiah GS, Niesel DW, Zhang H, Peterson JW, JAMA. 2015;314:2544–2554. Prabhakar BS, Klimpel GR. Purification and characterization of 3. Tunbridge WM, Evered DC, Hall R, Appleton D, Brewis M, Clark Yersinia enterocolitica envelope proteins which induce antibodies 152 F, Evans JG, Young E, Bird T, Smith PA. The spectrum of thyroid that react with human thyrotropin receptor. J Immunol. 1994; : – disease in a community: the Whickham survey. Clin Endocrinol 2555 2561. (Oxf). 1977;7:481–493. 17. Fairfax BP, Humburg P, Makino S, Naranbhai V, Wong D, Lau E, 4. Imrie H, Vaidya B, Perros P, Kelly WF, Toft AD, Young ET, Jostins L, Plant K, Andrews R, McGee C, Knight JC. Innate immune Kendall-Taylor P, Pearce SH. Evidence for a Graves’ disease sus- activity conditions the effect of regulatory variants upon monocyte 343 ceptibility locus at chromosome Xp11 in a United Kingdom pop- gene expression. Science. 2014; :1246949. ulation. J Clin Endocrinol Metab. 2001;86:626–630. 18. Gregersen PK. Genetics. A genomic road map for complex human 5. Ban Y, Tozaki T, Tobe T, Ban Y, Jacobson EM, Concepcion ES, disease. Science. 2014;343:1087–1088. Tomer Y. The regulatory T cell gene FOXP3 and genetic suscep- 19. Song HD, Liang J, Shi JY, Zhao SX, Liu Z, Zhao JJ, Peng YD, Gao tibility to thyroid autoimmunity: an association analysis in Cau- GQ, Tao J, Pan CM, Shao L, Cheng F, Wang Y, Yuan GY, Xu C, casian and Japanese cohorts. J Autoimmun. 2007;28:201–207. Han B, Huang W, Chu X, Chen Y, Sheng Y, Li RY, Su Q, Gao L, Jia 6. Inoue N, Watanabe M, Morita M, Tomizawa R, Akamizu T, WP, Jin L, Chen MD, Chen SJ, Chen Z, Chen JL. Functional SNPs Tatsumi K, Hidaka Y, Iwatani Y. Association of functional poly- in the SCGB3A2 promoter are associated with susceptibility to morphisms related to the transcriptional level of FOXP3 with Graves’ disease. Hum Mol Genet. 2009;18:1156–1170. prognosis of autoimmune thyroid diseases. Clin Exp Immunol. 20. Howie BN, Donnelly P, Marchini J. A flexible and accurate ge- 2010;162:402–406. notype imputation method for the next generation of genome-wide 7. Chu X, Pan CM, Zhao SX, Liang J, Gao GQ, Zhang XM, Yuan association studies. PLoS Genet. 2009;5:e1000529. GY, Li CG, Xue LQ, Shen M, Liu W, Xie F, Yang SY, Wang HF, Shi 21. Johnson AD, Handsaker RE, Pulit SL, Nizzari MM, O’Donnell CJ, JY, Sun WW, Du WH, Zuo CL, Shi JX, Liu BL, Guo CC, Zhan M, de Bakker PI. SNAP: a web-based tool for identification and an- Gu ZH, Zhang XN, Sun F, Wang ZQ, Song ZY, Zou CY, Sun WH, notation of proxy SNPs using HapMap. Bioinformatics. 2008;24: Guo T, Cao HM, Ma JH, Han B, Li P, Jiang H, Huang QH, Liang L, 2938–2939. Liu LB, Chen G, Su Q, Peng YD, Zhao JJ, Ning G, Chen Z, Chen JL, 22. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender Chen SJ, Huang W, Song HD. China Consortium for Genetics of D, Maller J, Sklar P, de Bakker PI, Daly MJ, Sham PC. PLINK: 660 Ye et al ITM2A Contributes to Graves Disease Pathogenesis J Clin Endocrinol Metab, February 2017, 102(2):652–660

a tool set for whole-genome association and population-based Tiret L. Influence of sex and genetic variability on expression of linkage analyses. Am J Hum Genet. 2007;81:559–575. X-linked genes in human monocytes. Genomics. 2011;98:320–326. 23. Diabetes Genetics Initiative of Broad Institute of Harvard and MIT, 28. Tai TS, Pai SY, Ho IC. Itm2a, a target gene of GATA-3, plays a Lund University, and Novartis Institutes of BioMedical Research, minimal role in regulating the development and function of T cells. Saxena R, Voight BF, Lyssenko V, Burtt NP, de Bakker PI, Chen H, PLoS One. 2014;9:e96535. Roix JJ, Kathiresan S, Hirschhorn JN, Daly MJ, Hughes TE, Groop 29. Wang Y, Misumi I, Gu AD, Curtis TA, Su L, Whitmire JK, Wan YY. L, Altshuler D, Almgren P, Florez JC, Meyer J, Ardlie K, Bengtsson GATA-3 controls the maintenance and proliferation of T cells Bostrom K, Isomaa B, Lettre G, Lindblad U, Lyon HN, Melander downstream of TCR and cytokine signaling. Nat Immunol. 2013; O, Newton-Cheh C, Nilsson P, Orho-Melander M, Rastam L, 14:714–722. Speliotes EK, Taskinen MR, Tuomi T, Guiducci C, Berglund A, 30. Pai SY, Truitt ML, Ho IC. GATA-3 deficiency abrogates the de- Carlson J, Gianniny L, Hackett R, Hall L, Holmkvist J, Laurila E, velopment and maintenance of T helper type 2 cells. Proc Natl Acad Sjogren M, Sterner M, Surti A, Svensson M, Svensson M, Tewhey Sci USA. 2004;101:1993–1998. R, Blumenstiel B, Parkin M, Defelice M, Barry R, Brodeur W, 31. Tai TS, Pai SY, Ho IC. GATA-3 regulates the homeostasis and Camarata J, Chia N, Fava M, Gibbons J, Handsaker B, Healy C, activation of CD8+ T cells. J Immunol. 2013;190:428–437. Downloaded from https://academic.oup.com/jcem/article/102/2/652/2972067 by guest on 27 September 2021 Nguyen K, Gates C, Sougnez C, Gage D, Nizzari M, Gabriel SB, 32. Kirchner J, Bevan MJ. ITM2A is induced during thymocyte se- Chirn GW, Ma Q, Parikh H, Richardson D, Ricke D, Purcell S. lection and T cell activation and causes downregulation of CD8 Genome-wide association analysis identifies loci for type 2 diabetes when overexpressed in CD4(+)CD8(+) double positive thymocytes. and triglyceride levels. Science. 2007;316:1331–1336. J Exp Med. 1999;190:217–228. 24. Cooper JD, Simmonds MJ, Walker NM, Burren O, Brand OJ, Guo 33. Marson A, Kretschmer K, Frampton GM, Jacobsen ES, Polansky H, Wallace C, Stevens H, Coleman G, Wellcome Trust Case JK, MacIsaac KD, Levine SS, Fraenkel E, von Boehmer H, Young Control Consortium, Franklyn JA, Todd JA, Gough SC. Seven RA. Foxp3 occupancy and regulation of key target genes during newly identified loci for autoimmune thyroid disease. Hum Mol T-cell stimulation. Nature. 2007;445:931–935. Genet. 2012;21:5202–5208. 34. McHugh RS, Whitters MJ, Piccirillo CA, Young DA, Shevach EM, 25. Fallahi P, Ferrari SM, Vita R, Benvenga S, Antonelli A. The role of Collins M, Byrne MC. CD4(+)CD25(+) immunoregulatory T cells: human parvovirus B19 and hepatitis C virus in the development gene expression analysis reveals a functional role for the glucocorticoid- of thyroid disorders. Rev Endocr Metab Disord. 2016. induced TNF receptor. Immunity. 2002;16:311–323. 26. Tomer Y, Blackard JT, Akeno N. Interferon alpha treatment and 35. Morshed SA, Latif R, Davies TF. Delineating the autoimmune thyroid dysfunction. Endocrinol Metab Clin North Am. 2007;36: mechanisms in Graves’ disease. Immunol Res. 2012;54:191–203. 1051–1066 (x-xi). 36. Saitoh O, Nagayama Y. Regulation of Graves’ hyperthyroidism 27. Castagne R, Zeller T, Rotival M, Szymczak S, Truong V, Schillert with naturally occurring CD4+CD25+ regulatory T cells in a mouse A, Tregouet DA, Munzel T, Ziegler A, Cambien F, Blankenberg S, model. Endocrinology. 2006;147:2417–2422.