Background With evidence initially from The Wellcome Trust Case Control Consortium genome wide association study (GWAS), a region on 12q13 has been convincingly and reproducibly associated with rheumatoid arthritis (RA) and several other autoimmune diseases (AIDs) including type 1 diabetes (T1D) and multiple sclerosis (MS). The strongest association for RA was with a SNP, rs1678542, within intron 15 of KIF5A; however, the locus is a gene rich region, with no clear candidate gene implicated.
Objectives The aim of this study was to refine the association at chromosome 12q13, identify the true causal variant(s) and the gene on which they may act.
Methods Fine mapping of the region was carried out in 3,870 UK cases and 8,430 UK controls. Genotyping was performed on a custom made Illumina Infinium genotyping array, “Immunochip”, to capture all known variation in the region. After stringent sample (>99%) and SNP (>99%) quality control (QC), association testing and conditional logistic regression, to identify independent effects in the region, were carried out in PLINK. SNPs with p<0.0001 after conditioning, were classed as independent.
The expression of four candidate genes (KIF5A, PIP4K2C, METTL21B and CYP27B1) in the region was measured in whole blood from 52 UK RA patients using quantitative real time PCR (qPCR) and correlated with genotypes at all SNPs in the region using linear regression carried out in PLINK, to identify possible expression quantitative trait loci (eQTL). Expression levels were normalised against an average of two reference genes. This experimentally derived data was compared to expression of all genes in the region correlated with genotype at all HapMap SNPs, using publically available expression data from multiple tissues.
Results After QC 762 SNPs were available for analysis. The Immunochip results showed that the peak of association has shifted from the associated SNP within KIF5A identified through GWAS, to a SNP rs2069506 located within an intron of CDK4 (p=3.89×10-6 OR 0.87 95% CI 0.82-0.92). The newly identified association peak localises to a region which has shown previous association with other AID, including T1D and MS. Indeed the most associated SNP in the region, rs2069506, is highly correlated (r20.96) with the MS associated SNP rs703842. Conditional logistic regression identified one effect in the region suggesting the association is due to rs2069506 or a highly correlated SNP.
Quantitative real time PCR of candidate genes in the region and analysis of publically available expression data identified a significant eQTL between SNPs in the 12q13-14 region and METTL21B expression in multiple cell types, the strongest being between rs10431552 and METTL21B in lymphoblastoid cells from all 270 HapMap individuals p=8.27×10-35. This SNP also showed association with RA p=6.19×10-6 (OR 0.87) and is correlated with rs2069506 (r2 0.96).
The peak of eQTL overlaps that of association, suggesting that SNPs in the region associated with RA also regulate METTL21B expression.
Conclusions This study has refined the association with RA in the 12q13-14 region, and has localised the association to a region previously associated with MS and T1D. The identification of a significant eQTL in the region also implicates METTL21B as a strong candidate causal gene.
Disclosure of Interest None Declared