HLA-DPB1-COL11A2 and Three Additional Xmhc Loci Are Independently Associated with RA in a UK Cohort

ORIGINAL ARTICLE HLA-DPB1-COL11A2 and three additional xMHC loci are independently associated with RA in a UK cohort

G Orozco1, A Barton1, S Eyre1, B Ding2, J Worthington1,XKe1,3,4 and W Thomson1,4 1Arthritis Research UK Epidemiology Unit, Manchester Academic Health Science Centre, University of Manchester, Manchester, UK; 2Karolinska Institute, Stockholm, Sweden and 3Institute of Child Health, University College London, London, UK

The aim of this study was to investigate the complex association pattern of the extended major histocompatibility complex (xMHC) region with rheumatoid arthritis (RA) susceptibility to identify effects independent of HLA-DRB1. A total of 1804 RA cases and 1474 controls were included. High-resolution HLA-DRB1 typing was performed. Subjects were genotyped for 1546 single-nucleotide polymorphisms (SNPs) using Affymetrix GeneChip 500 K (Santa Clara, CA, USA) as part of the Wellcome Trust Case Control Consortium Study. Statistical analysis was carried out using PLINK. To avoid confounding by RA- associated HLA-DRB1 alleles, we analyzed xMHC SNPs using a data set with pairwise matching of cases and controls on DRB1 genotypes. A total of 594 case–control pairs with identical DRB1 genotypes were identified. After this adjustment, 104 SNPs remained significantly associated with RA (Po0.05), suggesting that additional RA loci independent of HLA-DRB1 can be found in the xMHC region. Of these, four loci showed the strongest associations with RA (Po0.005): ZNF391, the olfactory receptor (OR) gene cluster, C6orf26-RDBP and HLA-DPB1-COL11A2. An additional locus mapping to the BTN (butyrophilin) cluster showed independent association with RA in anti-cyclic citrullinated peptide-positive patients exclusively. We have validated the previously described independent association of the HLA-DPB1-COL11A2 locus with RA. In addition, association with three novel independent RA loci in the xMHC region (ZNF391, OR2H1 and C6orf26-RDBP) has been detected. Genes and Immunity (2011) 12, 169–175; doi:10.1038/gene.2010.57; published online 3 February 2011

Keywords: rheumatoid arthritis; xMHC; susceptibility; HLA-DRB1

Introduction methods revealed that the association of the HLA-DRB1 locus with RA is complex, with a variety of specific Rheumatoid arthritis (RA) is a common disease, affecting alleles conferring different risks. These alleles share a around 1% of the population worldwide. It is an sequence in the third hypervariable region of the gene inflammatory, autoimmune, systemic disease and is char- termed the shared epitope.5 Since then, countless studies acterized by the chronic inflammation and progressive have demonstrated that HLA-DRB1 is consistently linked destruction of the joints that leads to disability.1 RA is a and associated with RA in every population studied.6 complex disease, meaning that multiple genetic variants, HLA-DRB1 lies within the major histocompatibility environmental factors and random events interact to trigger complex (MHC) on chromosome 6p21. The classical pathological pathways. Although many of these etiological extended MHC (MHCx) encompasses a region of factors have not yet been identified, recent groundbreaking B3.6 Mb and it is divided into three subregions (classes advances such as the availability of genome-wide associa- I, II and III), whereas the xMHC encompasses a B7.8 Mb. tion studies have expanded our knowledge about the It is characterized by its great variability and gene genetic factors that contribute to RA, such that 26 loci have density. Many of these genes have immunological now been confirmed (Po5 Â 10À8).2 functions, and variation within the MHC has been found However, the HLA-DRB1 locus identified long before to be associated with the majority of autoimmune the advent of genome-wide association studies remains diseases.7 The region is also characterized by the the major RA susceptibility locus. The association of presence of strong linkage disequilibrium (LD). HLA-DRB1 alleles with RA was described for the first A variety of approaches have been used to determine time in the 1970s.3,4 Subsequent molecular typing whether additional RA susceptibility loci reside within the MHC region, but interpretation of these associations is made difficult by the strong LD and insufficient Correspondence: Dr G Orozco, Arthritis Research UK Epidemiology coverage of the region in many of the studies.8–13 More Unit, Manchester Academic Health Science Centre, University of recently, dense single-nucleotide polymorphism (SNP)- Manchester, Stopford Building, Oxford Road, Manchester M13 9PT, mapping studies have identified several loci conferring UK. RA susceptibility independently of HLA-DRB1 risk E-mail: [email protected] 4These authors share senior authorship. alleles using novel statistical approaches to avoid Received 26 April 2010; revised and accepted 25 June 2010; confounding.14–16 Validation in independent data sets is published online 3 February 2011 particularly important to confirm new RA loci in the Association of xMHC with RA G Orozco et al 170 MHC. The Wellcome Trust Case Control Consortium HLA-G-HLA-A locus, whereas the other four showed (WTCCC) Study provides an ideal data source for this stronger evidence of individual independent effects purpose.17 In the present study, this large and well- (Supplementary Figure 2). Therefore, we have found at characterized data set was examined for the identifica- least four xMHC loci associated with RA independently tion and validation of HLA-DRB1-independent RA of HLA-DRB1: risk loci. Locus 1: close to ZNF391 in the extended class I region; Locus 2: close to OR2H1 in the extended class I region; Locus 3:closetoC6orf26-RDBP in the class III region; and Results Locus 4: HLA-DPB1-COL11A2 in the class II region. In the initial unmatched analysis, we found 745 It has been shown that HLA alleles are differentially SNPs significantly associated with RA (P trendo0.05; associated with anti-cyclic citrullinated peptide (CCP)- Figure 1a). Of these, 192 markers reached genome- positive and -negative disease. Therefore, we next wide significance (Po5 Â 10À7, as defined in the analyzed the unmatched data in anti-CCP-positive and original WTCCC Study17) and 269 were significant anti-CCP-negative patients separately. The anti-CCP þ after Bonferroni correction of the xMHC region (Po3.2 Â subgroup showed a similar pattern of association 10À5). Many of the strongest associations were observed compared with the whole data set, with the strongest in the vicinity of the HLA-DRB1 locus, although a associations around the HLA-DRB1 locus and evidence number of suggestive additional signals closer to the of additional signals (Figure 1b). On the contrary, only telomere were detected. the HLA-DRB1 region surpassed the GW significance To avoid confounding by RA-associated HLA-DRB1 threshold in the anti-CCPÀ subgroup (Figure 1c). alleles, we performed HLA-DRB1 case–control matching Next, we carried out HLA-DRB1-matched analysis at a ratio 1:1, as previously described.14 Given the stratified according to the presence or absence of anti- stringent selection strategy, we expected a decrease in CCP antibodies. A total of 296 and 119 matched case– the statistical power, as the number of individuals control pairs were available for the anti-CCP þ and anti- included in the analysis was reduced significantly (594 CCPÀ groups, respectively. Three loci were associated case–control pairs). A significance threshold was set at with anti-CCP þ RA, independently of HLA-DRB1,at Po0.005, and five loci were identified to be associated Po0.005: the butyrophilin (BTN)-gene cluster locus with RA independently of HLA-DRB1, that is, ZNF391, (particularly in the BTN2A2-BTN3A1 region), HLA-B- OR2H1-OR2H2, HLA-G-HLA-A, C6orf26-RDBP and HCP5 and HLA-DPB1-COL11A2 (Table 2 and Figure 2b). HLA-DPB1-COL11A2 (Table 1 and Figure 2a). Conditional logistic regression analysis provided sup- To further test whether these five loci represent effects port that they represented true independent effects independent of each other and the rest of the region, we (Supplementary Figure 3). In addition, the olfactory performed conditional logistic regression analysis for the receptor (OR) locus could also represent another strongest associated SNP within each locus, conditioning independent effect (P ¼ 0.007) (Figure 2b). With regard on all the SNPs analyzed across the xMHC region. to the anti-CCP-patient subgroup, ZNF391, the OR This analysis also allows the visual display of the in- cluster, and HLA-DPB1-COL11A2 showed some evi- dependence of a locus (a dip and/or a spike against dence of association with RA (P ¼ 0.01), but the weaker the xMHC), as shown in Supplementary Figure 2. A association was primarily due to the smaller sample very complex association pattern was found for the size, in comparison with the anti-CCP þ subgroup

80 80 70 70 60 60 50 50 40 40 -logP -logP 30 30 20 20 10 10 0 0 25700 26700 27700 28700 29700 30700 31700 32700 2570026700 27700 28700 29700 30700 31700 32700 Position on Chr6 (Kb) Position on Chr6 (Kb)

8 7 6 5 4

-logP 3 2 1 0 2570026700 27700 28700 29700 30700 31700 32700 Position on Chr6 (Kb) Figure 1 Unmatched association analysis of 1546 SNPs in the xMHC region using the whole RA cohort (1804 RA patients and 1474controls) (a), anti-CCP þ (1042 RA patients and 1804 controls) (b) and anti-CCPÀ (270 RA patients and 1804 controls) (c) patients only. Red line indicates P-value significance threshold (Po5 Â 10À7).

Genes and Immunity Association of xMHC with RA G Orozco et al 171 Table 1 xMHC loci showing independent association with RA (Po0.005) after adjustment for HLA-DRB1 alleles by pairwise matching on HLA-DRB1 genotype (n ¼ 594 matched case–control pairs)

SNP BP Nearest gene Major allele Minor allele MAF RA MAF controls P-value OR (95% CI) rs1883404 27458360 ZNF391 T A 0.21 0.26 0.0039 0.75 (0.62–0.91) rs9468111 27463062 ZNF391 A T 0.21 0.26 0.0038 0.75 (0.62–0.91) rs9468112 27465670 ZNF391 A G 0.21 0.27 0.0008 0.72 (0.59–0.87) rs2206251 27466371 ZNF391 T C 0.21 0.26 0.0034 0.75 (0.62–0.91) rs2206252 27466441 ZNF391 C A 0.21 0.26 0.0049 0.76 (0.62–0.92) rs6936539 27466896 ZNF391 G C 0.21 0.26 0.0043 0.75 (0.62–0.92) rs7754411 27467563 ZNF391 A T 0.21 0.26 0.005 0.76 (0.62–0.92) rs4455661 27481109 ZNF391 C T 0.21 0.26 0.005 0.76 (0.62–0.92) rs2107192 29541921 OR2H1-OR2H2 G A 0.22 0.27 0.0046 0.76 (0.62–0.92) rs2746149 29543334 OR2H1-OR2H2 T C 0.18 0.13 0.0013 1.46 (1.16–1.85) rs1233495 29552100 OR2H1-OR2H2 G A 0.32 0.27 0.0033 1.31 (1.09–1.57) rs1233487 29576788 OR2H1-OR2H2 A G 0.32 0.27 0.0037 1.31 (1.09–1.57) rs389600 30004978 HLA-G-HLA-A C T 0.23 0.17 0.0014 1.41 (1.14–1.74) rs707937 31838993 C6orf26-RDBP C G 0.17 0.21 0.0043 0.74 (0.59–0.91) rs550513 32028666 C6orf26-RDBP C T 0.06 0.09 0.0029 0.61 (0.45–0.85) rs3129249 33217534 HLA-DPB1-COL11A2 C T 0.19 0.24 0.0028 0.74 (0.61–0.90) rs3129248 33217636 HLA-DPB1-COL11A2 A G 0.35 0.49 0.0011 0.75 (0.64–0.89)

Abbreviations: BP, base pair; CI, confidence interval; MAF, minor allele frequency; OR, odds ratio; RA, rheumatoid arthritis; SNP, single- nucleotide polymorphism; xMHC, extended major histocompatibility complex.

3.5 3.5 ZNF391 HLA-G- BTN cluster HLA-DPB1 OR cluster HLA-A HLA-DPB1 3 C6orf26- 3 HLA-B-HCP5 RDBP 2.5 2.5 OR cluster 2 2 ZNF391 1.5

-logP 1.5 -logP 1 1 0.5 0.5 0 0 25700 26700 27700 28700 29700 3070031700 32700 25700 26700 27700 28700 29700 30700 31700 32700 Position on Chr6 (Kb) Position on Chr6 (Kb)

3.5 3 2.5 ZNF391 OR cluster 2 HLA-DPB1

-logP 1.5 1 0.5 0 25700 26700 27700 28700 29700 30700 31700 32700 Position on chr6 (Kb) Figure 2 Plots showing significantly associated SNPs in the xMHC region after pairwise matching on HLA-DRB1 genotype in the whole RA cohort (a), in CCP þ patients (b) and in CCPÀ patients (c). Red line indicates P-value significance threshold (Po0.005).

(Figure 2c), as none of the top SNPs in each of these (P ¼ 0.005), but the effect size was not increased loci showed significant heterogeneity by Breslow–Day compared with that found for rs9468112 or its proxies test between anti-CCP þ and anti-CCPÀ groups after alone (odds ratio 0.76; 95% confidence interval 0.62–0.93 HLA-DRB1 matching (data not shown). for the haplotype and 0.72–0.76 for the SNPs) (Table 1 Next, we investigated in more detail the LD present in and Supplementary Table 1). the ZNF391, OR2H1 and HLA-DPB1-COL11A2 loci, because The next most strongly associated SNP was rs3129248. they were the most strongly associated regions and This SNP is located in an intergenic region B55 Kb from contained multiple SNPs associated at the Po0.005 thresh- the HLA-DPB1 gene (Figure 3b). A number of SNPs old in the HLA-DRB1-adjusted analysis (Figure 2a). showed significant P-values in this locus. These asso- The SNP showing the strongest association with ciated SNPs seemed to be located in two differentiated RA after pairwise matching on HLA-DRB1genotype, LD blocks. The first contains the top hit within the rs9468112, is located in the ZNF391 gene. A total of 13 region, rs3129248, and it is located B21 Kb upstream SNPs, mapping to the same LD block and exhibiting very of the COL11A2 gene. The second block is closer to the tight LD, were associated with RA (Po0.05) in this HLA-DPB1 gene (B12 Kb downstream), and overlaps region (Figure 3a). We found one haplotype significantly with the HLA-DRB1-independent RA risk locus pre- increased in controls compared with RA patients viously identified by Ding et al.14

Genes and Immunity Association of xMHC with RA G Orozco et al 172 Table 2 Independently associated xMHC loci (Po0.005) after been implicated in susceptibility to skeletal and cartilage adjustment for HLA-DRB1 alleles by pairwise matching on disorders, and hearing loss.27–32 HLA-DPB1 is an attrac- HLA-DRB1 genotype in CCP+ RA patients tive candidate gene for RA within this region, as it has an important role in the immune system by presenting SNP BP Nearest gene P-value OR (95% CI) peptides to CD4 þ T cells. Furthermore, the other two genes encoding classical HLA class II molecules, rs9461245 26462551 BTN2A2-BTN3A1 0.0024 0.55 (0.38–0.81) HLA-DRB1 and HLA-DQA1, are established RA risk rs12203176 26473574 BTN2A2-BTN3A1 0.0015 0.54 (0.37–0.79) genes. Interestingly, this locus has also been shown to rs7355 26502138 BTN2A2-BTN3A1 0.0006 0.55 (0.39–0.78) rs1804837 26502299 BTN2A2-BTN3A1 0.0004 0.54 (0.39–0.76) be associated with type 1 diabetes independently of 33,34 rs4712986 26504182 BTN2A2-BTN3A1 0.0009 0.56 (0.40–0.79) HLA-DRB1/HLA-DQB1. To formally assign this rs12208788 26506534 BTN2A2-BTN3A1 0.0007 0.56 (0.40–0.78) effect to HLA-DPB1, however, genotyping of HLA-DP is rs10807008 26513243 BTN2A2-BTN3A1 0.0012 0.54 (0.37–0.78) desirable in the relevant cohorts. rs17610161 26513969 BTN2A2-BTN3A1 0.0022 0.56 (0.38–0.81) Additionally, we found evidence of new independent rs10946823 26524595 BTN2A2-BTN3A1 0.0044 0.63 (0.46–0.87) RA susceptibility loci within the MHC extended class I rs4712996 26549527 BTN2A2-BTN3A1 0.0050 0.63 (0.45–0.87) rs6929846 26566244 BTN2A2-BTN3A1 0.0016 0.60 (0.44–0.83) region. The top hit from our study maps to an intronic rs2893857 26582040 BTN2A2-BTN3A1 0.0023 0.54 (0.37–0.80) haplotype in the zinc finger protein 391 (ZNF391) gene. rs10484441 26582721 BTN2A2-BTN3A1 0.0014 0.53 (0.36–0.78) Zinc finger proteins have diverse functions, and can rs4713001 26595289 BTN2A2-BTN3A1 0.0016 0.53 (0.36–0.79) act as enzymes, storage proteins, replication proteins rs2596460 31525489 HLA-B-HCP5 0.0029 0.46 (0.28–0.77) and transcription factors.35 The exact role of ZNF391 rs3129249 33217534 HLA-DPB1-COL11A2 0.0049 0.67 (0.51–0.89) is currently unknown, but the similarity between its rs3129248 33217636 HLA-DPB1-COL11A2 0.0005 0.65 (0.51–0.82) sequence and sequences of known proteins suggests that it may be involved in transcriptional regulation Abbreviations: BP, base pair; CCP, cyclic citrullinated peptide; CI, (UniProtKB/Swiss-Prot accession number Q9UJN7; confidence interval; OR, odds ratio; RA, rheumatoid arthritis; SNP, http://www.uniprot.org). single-nucleotide polymorphism; xMHC, extended major histocom- The closest gene to the second RA-associated locus in patibility complex. the extended class I region is OR2H1 (olfactory receptor, family 2, subfamily H, member 1). Suggestive evidence for association with RA has been previously found for With regard to the OR2H1 locus, there were a number other members of the OR supercluster.15 The role that of SNPs significantly associated with RA in the same LD these highly polymorphic genes might be having in RA block: the top hit rs2746149 (and its proxies rs2523443 etiological pathways is not evident, as they are involved and rs1233493), rs1233495 (and its proxy rs1233487) and in the sensory perception of smell. Nevertheless, it has rs2107192 (and its proxies rs1233490 and rs2107193) been proposed recently that olfactory abnormalities may (Figure 3c). One haplotype, tagged by the minor G be associated with autoimmune conditions involving allele of rs1233490, was strongly associated with RA the central nervous system.36 RA patients demonstrate (P ¼ 6 Â 10À4; odds ratio 0.72; 95% confidence interval a higher prevalence of psychiatric disorders, such as 0.59–0.87), but the effect size was lower than that found depression, than the normal population,37 and several for rs1233490 alone (odds ratio 0.77; 95% confidence studies show a link between depression and olfactory interval 0.64–0.94; Table 1 and Supplementary Table 2). impairment.36 Other neighboring genes include the The haplotype tagged by the top hit in the region, ubiquitin D pseudogene 1 (UBDP1) and MAS1 onco- rs2746149, was also associated with RA (P ¼ 0.03; odds gene-like (MAS1L). Further validation, fine mapping and ratio 1.28; 95% confidence interval 1.01À1.62), but again functional studies will be necessary to identify the the effect of the haplotype was more modest than etiological variants within these new potential RA loci the SNP. and to elucidate their role in disease pathways. With regard to the stratified analysis according to RA patients’ anti-CCP status, the pattern of association was Discussion in general similar to that found in the non-stratified analysis (Figures 2a, b and c) in contrast with previous In this study, we have confirmed that multiple indepen- studies.14 Although the BTN gene cluster appeared to be dent loci are associated with RA in the xMHC complex. a risk locus exclusively for anti-CCP þ patients, a formal The most convincing signals after HLA-DRB1 mapped to test of heterogeneity of all the top SNPs in the associated the MHC class II (HLA-DPB1-COL11A2) and extended loci did not reveal significant heterogeneity between class I (ZNF391 and OR2H1) regions. anti-CCP þ and anti-CCPÀ groups after HLA-DRB1 HLA-DPB1 alleles have been studied in the past as matching with controls. Interestingly, the BTN gene potential RA susceptibility markers, with conflicting cluster has been also associated with type 1 diabetes.38 results, possibly reflecting an underpowered study de- The BTN cluster encodes proteins of as yet unknown sign and/or clinical heterogeneity.18–26 Our result in a UK functions, but they are thought to have an immunologi- cohort together with previous studies in Swedish and cal role, given that they show strong similarity with other North American populations14,15 provides strong evi- xMHC genes and with CD80 and CD86 costimulatory dence that HLA-DPB1-COL11A2 is a genuine RA risk receptors.39,40 Regarding the association found in the locus independent of HLA-DRB1 and the rest of the HLA-B-HCP5 locus for anti-CCP þ patients, it is difficult xMHC region. Several SNPs were found to be associated to ascertain whether it represents a different effect with RA in this intergenic region, distributed in two from that found in Corf26-RDBP in the non-stratified different LD blocks. This pattern was similar to that analysis (Figures 2a and b). The lack of statistically found by Lee et al.15 Polymorphisms in COL11A2 have significant associations mapping to the xMHC in the

Genes and Immunity Association of xMHC with RA G Orozco et al 173 rs9468112 region rs3129248 region Recombination rate (cM/Mb) Recombination rate (cM/Mb) 8 80 8 80 recombination rate 0.8 recombination rate 0.8 6 0.5 60 6 0.5 60 2 r2 r 4 40 rs9468112 4 rs3129248 40 P=8e-04 P=0.001 2 20 2 20

Observed (-logP) 0 0

Observed (-logP) 0 ZNF391 ZNF184 0 HLA-DPA1 COL11A2 HLA-DPB1 RXRB

27500 27600 33200 Chromosome 6 position (hg18) (kb) Chromosome 6 position (hg18) (kb)

rs2746149 region Recombination rate (cM/Mb) 8 80 recombination rate 0.8 6 0.5 60 r2 4 rs2746149 40 P=0.001328 2 20

Observed (-logP) 0 0

29200 29400 29600 Chromosome 6 position (hg18) (kb)

Figure 3 Association and LD plots for the ZNF391 (a), HLA-DPB1 (b) and OR2H1 (c) loci. Top hit from each region is marked with a red diamond.

anti-CCP-negative group of patients is in accordance The matching of cases and controls 1:1 according to with previous studies.14 However, this seems to be a HLA-DRB1 genotypes has been proposed as the optimal reflection of the modest sample size of this cohort (270 way to adjust for confounding by the effect of DRB1 RA patients). Finally, stratified analysis results should be genotype, and has proved successful in the identification considered with caution, given the impossibility to of new independent RA susceptibility loci in the xMHC recruit a sufficiently large number of RA patients with region.14,15 The advantage of this approach over adjust- high-resolution HLA-DRB1 typing and anti-CCP data ing for the carriage of shared epitope alleles is that it from the WTCCC cohort. avoids the residual confounding that might exist in the A previous work investigating SNPs in the MHC latter, given the complexity of the DRB1 locus and the region conducted in a UK population found two fact that different shared epitope alleles confer different different loci associated with RA, mapping to the disease risks. However, the DRB1 matching has the VARS2L gene and upstream of HLA-B.16 As mentioned implicit drawback of a significant reduction of the above, we found a signal downstream HLA-B in anti- sample size and, consequently, of the statistical power, CCP þ patients, but it was almost 100 Kb away from that making meta-analysis of multiple studies desirable. found by Vignal et al. (rs2442728).16 No SNPs in LD with In conclusion, the HLA-DPB1-COL11A2 locus has been rs2442728 were tested in our data set; the closest proxy confirmed as an independent RA risk locus in our study. (rs887464, r2 ¼ 0.47) was not significantly associated with Furthermore, our results identified other additional RA. However, our results are difficult to compare with effects in the xMHC region, supporting the long- Vignal et al.’s,16 because the methods to control con- standing hypothesis that additional non-DRB1 effects founding by HLA-DRB1 were different in both studies. contribute to RA disease susceptibility. With the avail- Whereas the previous UK study adjusted the results for ability of multiple dense SNP and HLA genotype data in the number of shared epitope copies, we used the DRB1 large well-characterized case–control cohorts, a more matching approach. detailed picture of these effects can be established.

Genes and Immunity Association of xMHC with RA G Orozco et al 174 Patients and methods All statistical analyses were performed using PLINK.43 Haplotypes were estimated using Haploview.44 Patients RA cases and controls cohorts were part of the WTCCC Study and have been described elsewhere.17 Briefly, Conflict of interest a total of 1804 RA patients satisfying the 1987 American College of Rheumatology Criteria for RA modified The authors declare no conflict of interest. for genetic studies 41,42 and 1474 controls from the 1958 Birth Cohort (http://www.b58cgene.sgul.ac.uk/) were included in the study. All subjects were Caucasian Acknowledgements from the United Kingdom and were recruited after providing informed consent. In all, 75% of the patients This work was supported by the Arthritis Research UK were female, the age of onset was 46.3±14.6 years and (Grant reference number 17552). We are grateful to the 84% were positive for rheumatoid factor. Anti-CCP data NIHR Manchester Biomedical Research Centre and the were available for 1313 RA patients, of which 1042 (79%) WTCCC. GO is funded by the European Union (Marie were anti-CCP-positive and 271 (21%) were negative. Curie IEF Fellowship PIEF-GA-2009-235662). This cohort is therefore representative of a hospital-based series of patients with RA. None of the samples was included in any previous study on HLA associations References with RA. Anti-CCP antibody titer was measured at one point in 1 Klareskog L, Catrina AI, Paget S. Rheumatoid arthritis. time using a commercially available kit (Diastat Anti- Lancet 2009; 373: 659–672. CCP Kit; Axis-Shield Diagnostics Limited, Dundee, UK). 2 Raychaudhuri S. Recent advances in the genetics of rheuma- Patients with titers X5U mlÀ1 were defined as positive toid arthritis. Curr Opin Rheumatol 2010; 22: 109–118. for anti-CCP antibodies. 3 Stastny P. Mixed lymphocyte cultures in rheumatoid arthritis. J Clin Invest 1976; 57: 1148–1157. 4 Stastny P. Association of the B-cell alloantigen DRw4 with Genotyping rheumatoid arthritis. N Engl J Med 1978; 298: 869–871. High-resolution HLA-DRB1 typing was performed 5 Gregersen PK, Silver J, Winchester RJ. The shared epitope using Dynal RELI SSO kits (DYNAL Biotech, Bromborough, hypothesis. An approach to understanding the molecular genetics of susceptibility to rheumatoid arthritis. Arthritis UK), and was available for 1536 patients and 1126 Rheum 1987; 30: 1205–1213. controls. Subjects were genotyped using the Affymetrix 6 Newton JL, Harney SMJ, Wordsworth BP, Brown MA. A 17 GeneChip 500 K, as part of the WTCCC Study. After review of the MHC genetics of rheumatoid arthritis. Genes data quality control,17 1580 SNPs spanning a region of Immun 2004; 5: 151–157. B7.8 Mb within chromosome 6p, encompassing the 7 Fernando MMA, Stevens CR, Walsh EC, De Jager PL, Goyette xMHC (Chr6:25 751 771–33 537 930, NCBI build 35) were P, Plenge RM et al. Defining the role of the MHC in selected (Supplementary Figure 1). To guard against autoimmunity: a review and pooled analysis. PLoS Genet spurious associations, a stringent Hardy–Weinberg equi- 2008; 4: e1000024. librium threshold (Po0.001) was further applied, leaving 8 Jawaheer D, Li W, Graham RR, Chen W, Damle A, Xiao X et al. a total of 1546 SNPs for the subsequent analyses. Dissecting the genetic complexity of the association between human leukocyte antigens and rheumatoid arthritis. Am J Hum Genet 2002; 71: 585–594. Statistical analysis 9 Mulcahy B, Waldron-Lynch F, McDermott MF, Adams C, Amos CI, Zhu DK et al. Genetic variability in the tumor Association of HLA-DRB1 alleles with RA was inter- necrosis factor-lymphotoxin region influences susceptibility to rogated by logistic regression. Statistical analysis of the rheumatoid arthritis. Am J Hum Genet 1996; 59: 676–683. SNPs mapping to the xMHC region was carried out in 10 Newton JL, Harney SM, Timms AE, Sims AM, Rockett K, three stages: Darke C et al. Dissection of class III major histocompatibility complex haplotypes associated with rheumatoid arthritis. 1. Unmatched association analysis: Genotype counts in Arthritis Rheum 2004; 50: 2122–2129. 2 cases and controls were analyzed using the w -test for 11 Ota M, Katsuyama Y, Kimura A, Tsuchiya K, Kondo M, trend. Naruse T et al. A second susceptibility gene for developing 2. Adjustment of the association results for HLA-DRB1 rheumatoid arthritis in the human MHC is localized within a alleles: High-resolution HLA-DRB1 typing was used 70-kb interval telomeric of the TNF genes in the HLA class III to perform case–control matching at a ratio 1:1, as region. Genomics 2001; 71: 263–270. previously described.14,15 Using this approach, 594 12 Kilding R, Iles MM, Timms JM, Worthington J, Wilson AG. case–control pairs with identical DRB1 genotypes Additional genetic susceptibility for rheumatoid arthritis telomeric of the DRB1 locus. Arthritis Rheum 2004; 50: 763–769. were identified. Logistic regression was used to 13 Brintnell W, Zeggini E, Barton A, Thomson W, Eyre S, Hinks A perform the association analysis in this data set. et al. Evidence for a novel rheumatoid arthritis susceptibility 3. Conditional logistic regression: The top associated SNPs locus on chromosome 6p. Arthritis Rheum 2004; 50: 3823–3830. from stage 2 (Po0.005) were further tested for 14 Ding B, Padyukov L, Lundstrom E, Seielstad M, Plenge RM, independence by logistic regression analysis condi- Oksenberg JR et al. Different patterns of associations with anti- tioning on all the SNPs analyzed across the xMHC citrullinated protein antibody-positive and anti-citrullinated region. protein antibody-negative rheumatoid arthritis in the extended major histocompatibility complex region. Arthritis Stratified analyses were then carried out according to Rheum 2009; 60: 30–38. anti-CCP status, and statistical analyses as listed above 15 Lee HS, Lee AT, Criswell LA, Seldin MF, Amos CI, Carulli JP (1–3) were repeated for both subgroups. et al. Several regions in the major histocompatibility complex

Genes and Immunity Association of xMHC with RA G Orozco et al 175 confer risk for anti-CCP-antibody positive rheumatoid arthri- 30 McGuirt WT, Prasad SD, Griffith AJ, Kunst HP, Green GE, tis, independent of the DRB1 locus. Mol Med 2008; 14: 293–300. Shpargel KB et al. Mutations in COL11A2 cause non- 16 Vignal C, Bansal AT, Balding DJ, Binks MH, Dickson MC, syndromic hearing loss (DFNA13). Nat Genet 1999; 23: Montgomery DS et al. Genetic association of the major 413–419. histocompatibility complex with rheumatoid arthritis impli- 31 Videman T, Saarela J, Kaprio J, Nakki A, Levalahti E, Gill K cates two non-DRB1 loci. Arthritis Rheum 2009; 60: 53–62. et al. Associations of 25 structural, degradative, and 17 Wellcome Trust Case Control Consortium. Genome-wide inflammatory candidate genes with lumbar disc desiccation, association study of 14 000 cases of seven common diseases bulging, and height narrowing. Arthritis Rheum 2009; 60: and 3000 shared controls. Nature 2007; 447: 661–678. 470–481. 18 Carthy D, MacGregor A, Awomoi A, Rigby AS, Thomson W, 32 Vikkula M, Mariman EC, Lui VC, Zhidkova NI, Tiller GE, Donn R et al. HLA-DPB1*0201 is associated with particular Goldring MB et al. Autosomal dominant and recessive clinical features of rheumatoid arthritis. Rev Rhum Engl Ed osteochondrodysplasias associated with the COL11A2 locus. 1995; 62: 163–168. Cell 1995; 80: 431–437. 19 Gao X, Fernandez-Vina M, Olsen NJ, Pincus T, Stastny P. 33 Howson JM, Walker NM, Clayton D, Todd JA. Confirmation HLA-DPB1*0301 is a major risk factor for rheumatoid factor- of HLA class II independent type 1 diabetes associations in the negative adult rheumatoid arthritis. Arthritis Rheum 1991; 34: major histocompatibility complex including HLA-B and 1310–1312. HLA-A. Diabetes Obes Metab 2009; 11(Suppl 1): 31–45. 20 Perdriger A, Semana G, Quillivic F, Chales G, Chardevel F, 34 Nejentsev S, Howson JM, Walker NM, Szeszko J, Field SF, Legrand E et al. DPB1 polymorphism in rheumatoid arthritis: Stevens HE et al. Localization of type 1 diabetes susceptibility evidence of an association with allele DPB1 0401. Tissue to the MHC class I genes HLA-B and HLA-A. Nature 2007; Antigens 1992; 39: 14–18. 450: 887–892. 21 Perdriger A, Guggenbuhl P, Chales G, Le DP, Yaouanq J, 35 Coleman JE. Zinc proteins: enzymes, storage proteins, Genetet B et al. The role of HLA-DR-DR and HLA-DR-DP transcription factors, and replication proteins. Annu Rev interactions in genetic susceptibility to rheumatoid arthritis. Biochem 1992; 61: 897–946. Hum Immunol 1996; 46: 42–48. 36 Ortega-Hernandez OD, Kivity S, Shoenfeld Y. Olfaction, 22 Sage DA, Evans PR, Cawley MI, Smith JL, Howell WM. HLA psychiatric disorders and autoimmunity: is there a common DPB1 alleles and susceptibility to rheumatoid arthritis. genetic association? Autoimmunity 2009; 42: 80–88. Eur J Immunogenet 1991; 18: 259–263. 37 Bruce TO. Comorbid depression in rheumatoid arthritis: 23 Seidl C, Koch U, Brunnler G, Buhleier T, Frank R, Moller B pathophysiology and clinical implications. Curr Psychiatry et al. HLA-DR/DQ/DP interactions in rheumatoid arthritis. Rep 2008; 10: 258–264. Eur J Immunogenet 1997; 24: 365–376. 38 Viken MK, Blomhoff A, Olsson M, Akselsen HE, Pociot F, 24 Singal DP, Sastry A, Buchanan WW. HLA-DPB1 alleles in Nerup J et al. Reproducible association with type 1 diabetes in patients with rheumatoid arthritis. Clin Exp Rheumatol 1994; the extended class I region of the major histocompatibility 12: 183–186. complex. Genes Immun 2009; 10: 323–333. 25 Yelamos J, Garcia-Lozano JR, Moreno I, Romero M, Garcia A, 39 Linsley PS, Peach R, Gladstone P, Bajorath J. Extending the B7 Sanchez B. Frequency of HLA-DPB1 alleles in a Spanish (CD80) gene family. Protein Sci 1994; 3: 1341–1343. population: their contribution to rheumatoid arthritis suscept- 40 Rhodes DA, Stammers M, Malcherek G, Beck S, Trowsdale J. ibility. Eur J Immunogenet 1994; 21: 91–98. The cluster of BTN genes in the extended major histocompat- 26 Yen JH, Chen JR, Tsai WJ, Tsai JJ, Liu HW. HLA-DPB1 ibility complex. Genomics 2001; 71: 351–362. polymorphism in patients with rheumatoid arthritis in 41 Arnett FC, Edworthy SM, Bloch DA, McShane DJ, Fries JF, Taiwan. J Rheumatol 1995; 22: 2034–2037. Cooper NS et al. The American Rheumatism Association 1987 27 Brunner HG, van Beersum SE, Warman ML, Olsen BR, revised criteria for the classification of rheumatoid arthritis. Ropers HH, Mariman EC. A Stickler syndrome gene is linked Arthritis Rheum 1988; 31: 315–324. to chromosome 6 near the COL11A2 gene. Hum Mol Genet 42 MacGregor AJ, Bamder S, Silman AJ. A comparison of the 1994; 3: 1561–1564. performance of different methods of disease classification for 28 Jakkula E, Melkoniemi M, Kiviranta I, Lohiniva J, Raina SS, rheumatoid arthritis. Results from an analysis from a nation- Perala M et al. The role of sequence variations within the genes wide twin study. J Rheumatol 1994; 21: 1420–1426. encoding collagen II, IX and XI in non-syndromic, early-onset 43 Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, osteoarthritis. Osteoarthritis Cartilage 2005; 13: 497–507. Bender D et al. PLINK: a tool set for whole-genome association 29 Maeda S, Ishidou Y, Koga H, Taketomi E, Ikari K, Komiya S and population-based linkage analyses. Am J Hum Genet 2007; et al. Functional impact of human collagen alpha2(XI) gene 81: 559–575. polymorphism in pathogenesis of ossification of the posterior 44 Barrett JC, Fry B, Maller J, Daly MJ. Haploview: analysis and longitudinal ligament of the spine. J Bone Miner Res 2001; 16: visualization of LD and haplotype maps. Bioinformatics 2005; 948–957. 21: 263–265.

Supplementary Information accompanies the paper on Genes and Immunity website (http://www.nature.com/gene)

Genes and Immunity