Rheumatoid arthritis association at 6q23

The WTCCC genome-wide association screen (GWA) of 1,860 rheumatoid arthritis cases and 2,938 healthy controls confirmed association with SNPs within the HLA region and the PTPN22 gene (P < 1 × 10^-7; ref. 1). Nine other loci showed strong evidence for association (P = 1 × 10^-5 - 5 × 10^-7). SNPs at these loci were genotyped in an independent cohort of 5,063 rheumatoid arthritis cases and 3,849 healthy controls (Supplementary Methods and Supplementary Table 1 online), using the Sequenom iPlex platform (http://www.sequenom.com), to establish whether they are genuinely associated with the disease. In this cohort, we had 80% power to detect most of the effect sizes reported in the initial study at P < 0.05 (Supplementary Table 2 online). We selected ten SNPs for genotyping; these included the known rheumatoid arthritis susceptibility variant rs2476601, mapping to the PTPN22 gene, and nine previously unknown SNPs identified by the WTCCC study1. A Bonferroni correction of 9 was applied to account for the previously unknown loci investigated, resulting in a P value threshold of P < 0.006 for claims of significance in this validation study. We regarded SNPs validated at P values between 0.05 and 0.006 as suggestive evidence.

We detected strong association with the rheumatoid arthritis-causing SNP in the PTPN22 gene (rs2476601), as expected (odds ratio (OR) = 1.53, 95% CI = 1.39-1.68, trend P = 2.0 × 10^-18). This was not a completely independent replication, as association of rheumatoid arthritis with this locus has been reported in previous studies using some of the same samples included in the current study2-4. However, it confirmed the suitability of this cohort for validation studies. Of the nine newly identified SNPs tested, rs6920220 (G > A) showed association with rheumatoid arthritis in this cohort (OR for minor allele = 1.23, 95% CI = 1.15-1.33, trend P = 1.1 × 10^-8) (Table 1). For this SNP, the allele frequencies were similar across control groups tested in the WTCCC study and the healthy controls tested here (minor allele frequency (MAF) 0.22 and 0.21, respectively). We therefore undertook a combined analysis of the WTCCC data and the validation data, and we obtained strong statistical evidence for association between this SNP and rheumatoid arthritis (OR = 1.22, 95% CI = 1.15-1.29, trend P = 3.6 × 10^-12) (Table 1).

The validation samples came from six different centers in the UK, raising the possibility that the results were affected by population substructure and heterogeneity. The WTCCC study paid particular attention to the potential effect of population structure on disease association studies and found only 13 loci exhibiting notable geographical variation, none of which overlapped with loci in the current study. In addition, a stratified analysis by center revealed that the association of rs6920220 with rheumatoid arthritis was independently observed in 4 of the 5 centers tested (one center had no controls, and therefore association could not be statistically tested for this center; Supplementary Table 3 online). No heterogeneity was detected among the samples from the different centers or among the samples in the WTCCC study, and combined evidence from the different centers (using a Cochran-Mantel-Haenszel test) attained a significance level virtually the same as that from the combined samples (Supplementary Table 3). This study, therefore, provided convincing evidence for the association of rs6920220 with rheumatoid arthritis.

We did not detect evidence for interaction with HLA-DRB1 or PTPN22 susceptibility markers, suggesting that the rs6920220 variant confers an additional effect independent of the two known susceptibility genes. Together, HLA-DRB1 (shared epitope), PTPN22 (rs2476601) and rs6920220 account for 8.5% of the total variance in rheumatoid arthritis susceptibility.

Rheumatoid arthritis is a phenotypically heterogeneous disease with a female gender bias and variation in a number of features, such as the presence of autoantibodies (in particular, rheumatoid factor and antibodies to cyclic citrullinated peptides (CCP)) and joint erosions. Analysis of the WTCCC study indicated a significant association for SNP rs6920220 in anti-CCP-positive (OR = 1.38, 95% CI = 1.22-1.55, trend P = 1.7 × 10^-7) but not in anti-CCP-negative individuals (OR = 1.03, 95% CI = 0.82-1.29). The same pattern was seen in rheumatoid factor-positive and negative subgroups (Table 2). However, it should be noted that the anti-CCP-negative subgroup was small. In the combined WTCCC and validation cohorts, reasonable numbers in both subgroups were attained, and stronger associations were again observed in the antibody-positive subjects (Table 2). Indeed, the difference between the anti-CCP-positive and negative subgroups was statistically significant (trend P = 0.003). This lends support to the mounting evidence that the clinical phenotype of rheumatoid arthritis may actually comprise at least two genetically distinct subsets, one of which is characterized by the presence of antibody to CCP and is associated with exposure to cigarette smoking5,6.

The rs6920220 SNP maps to an intergenic region of 6q23. Four SNPs genotyped by the HapMap Consortium (rs6933404, rs2327832, rs6927172 and rs17264332) in the same region have a pairwise r² = 1 with rs6920220, and all five SNPs map to a single linkage-disequilibrium block spanning 60 kb (Supplementary Fig. 1 online). The block contains no known genes or transcripts but does encompass part of the PTPN11 pseudogene. The latter, notably, is a pseudogene for a locus that was highlighted in a combined analysis of autoimmune diseases (rheumatoid arthritis, type 1 diabetes and inflammatory bowel disease) in the WTCCC study. A recent report has suggested that up to 20% of pseudogenes may be transcribed, but none have been reported to be functional in humans to date7. The block lies between the genes OLIG3 and TNFAIP3 (also known as A20). OLIG3 appears to be important in development and differentiation of neuronal cells8, and therefore it is not an obvious candidate for association with rheumatoid arthritis susceptibility. The TNFAIP3 gene acts as a negative regulator of the transcription factor NF-κB in response to tumor necrosis factor-α and toll-like receptor-induced signals but not in response to IL-1β-induced activation9,10. Mice deficient for the TNFAIP3 protein developed multiorgan inflammation, including inflammation of joints11.

Of the remaining eight SNPs, we found nominal statistical evidence for association of rs743777 mapping to the IL2RB gene and observed the most significant effect under a recessive model (P = 0.005), in line with the findings from the WTCCC study1 (Supplementary Table 4 online). In the combined cohort analysis, this SNP showed modest evidence for association (OR = 1.13, 95% CI = 1.07-1.19; trend P = 5.3 × 10^-6), suggesting the effect, if any, is weak. Larger studies or meta-analysis of additional cohorts may be required to confirm this association. The WTCCC study observed that rs11761231, located close to PODXL, had a sex-differentiated effect in rheumatoid arthritis1. The present study provided only borderline evidence (trend P = 0.042) to support this observation (Supplementary Table 4). Although a stronger association was observed in females (trend P = 0.025) than in the combined male and female samples (and no significant association with rheumatoid arthritis was observed in males alone), a two-degree-of-freedom sex-differentiated test that combined trend tests in males and females did not reach a significance level of P < 0.05. For the other six SNPS, one (rs3816587) failed genotyping quality control measures, and no evidence for association was detected for any of the others in the validation dataset (Supplementary Table 4). Notably, rs2104286, which maps to IL2RA, a gene already associated with a number of other autoimmune diseases (type 1 diabetes, multiple sclerosis and Graves’ disease) and known to play a role in T cell regulation, was not associated with rheumatoid arthritis in this dataset12,13.

In conclusion, a previously unknown locus has emerged for rheumatoid arthritis, and the challenge will now be to characterize the etiological variant and determine its role in susceptibility.