For the BANK1
SNPs, data were available from a previous GWAS in a UK population including 2000 RA cases and 3000 controls.23
In addition, part of the cohort used in our study had been previously genotyped for UBE2L3
Therefore, non-overlapping samples from these studies were added to the current cohort for the analysis of the three markers. Control allele frequencies for all SNPs tested conformed to Hardy–Weinberg expectations (). Four SNPs, rs6568431 mapping to the ATG5
locus, rs2736340 close to BLK
, rs4963128 mapping to the KIAA1542
locus and rs5754217 in UBE2L3
showed nominal evidence for association (p<0.05). After applying a Bonferroni correction for the number of SNPs tested, only the SNPs mapping to BLK
retained statistically significant evidence for association. It should be noted that control allele frequencies were similar and the direction of association was the same at all four loci with that observed in the SLE studies, although effect sizes were smaller. Therefore, although non-significant after applying a Bonferroni correction, ATG5
remain interesting candidates for further investigation. Repeating the association analysis in the anti-CCP positive and negative subgroups of RA cases did not substantially alter the observed effect sizes (data not shown). A test of heterogeneity of odds ratios (ORs) showed that the effect sizes were similar among all RA, anti-CCP positive RA and anti-CCP negative RA subgroups for all analysed SNPs (p>0.05).
Association results for SLE risk variants genotyped in UK RA cases and controls
A meta-analysis including our UK cohort and the US cohort in which association of BLK
with RA was described for the first time showed a strong association between the rs2736340 SNP in the BLK
locus and RA (p=5.6×10−11
, OR=1.14 95% CI 1.10 to 1.19). We also performed a pooled analysis for UBE2L3
, expanding the validation cohort from the meta-analysis of Stahl et al24
with the UK non-overlapping samples genotyped in our study, and including data from GWAS. In this expanded analysis, the rs5754217 at the UBE2L3
locus now reaches genome-wide significance (p=2.3×10−10
, OR=1.14 95% CI 1.09 to 1.19).
We next performed a pathway analysis including the loci that have shown evidence of association with both SLE and RA (HLA-DRB1, PTPN22, STAT4, TNFAIP3, FCGR2A, PRDM1, IRF5, PXK, BLK and UBE2L2). All the over-represented pathways are involved in the immune response, such as dendritic cell maturation, T helper (Th) cell differentiation or CTLA4 signalling in cytotoxic T lymphocytes (). As is the case in the overlapping loci, genes exclusively associated with RA (AFF3, ANKRD55/IL6ST, C5orf13, CCL21, CCR6, CD2/CD58, CD28, CD40, CTLA4, IL2/IL21, IL2RA, IL2RB, KIF5A, PRKCQ, PTPRC, RBPJ, REL, SPRED2, TAGAP, TRAF1/C5 and TRAF6) are part of immune response pathways, with pathways related to Th cell activation being most over-represented. However, when analysing genes associated only with SLE (ATG5, BANK1, ITGAM, JAZF1, PHRF1, PTTG1, TNFSF4, TNIP1 and UHRF1BP1), a different pattern emerged. Four significantly over-represented pathways were identified but each included only one gene and only one of them was immune response related ().
Over-represented (p<0.05) canonical pathways in which RA and SLE overlapping loci, RA only loci and SLE only loci are involved
The number of SLE-associated SNPs showing at least nominal evidence for association with RA was higher than would be expected by chance alone. We explored, therefore, the extent of the total burden of SLE susceptibility alleles in RA. For this analysis we included previously generated genotype data for the markers: rs7574865 in STAT4, rs2476601 in PTPN22 and rs5029937 in TNFAIP3; and from this study: rs10489265 in TNFSF4, rs10516487 in BANK1, rs10036748 in TNIP1, rs2431697 in PTTG1, rs11755393 in UHRF1BP1, rs6568431 in ATG5, rs864745 in JAZF1, rs2736340 in BLK, rs4963128 in KIAA1542, rs9888739 in ITGAM and rs5754217 in UBEL2L3. We found that the mean number of SLE risk alleles carried by patients with RA was significantly higher than that found in controls (2.9 vs 2.5, p=9.1× 10−7).
Finally, we calculated the sibling recurrence risk ratio (λs) for the confirmed RA and SLE overlapping loci (PTPN22
, STAT4, TNFAIP3, FCGR2A, PRDM1, IRF5, PXK, BLK
). We estimate that, after excluding HLA-DRB1
alleles, these explain 5.8% of the genetic susceptibility to RA as a whole, while all the non-HLA confirmed RA loci identified to date24
are able to explain 10.7%.