Background/Purpose: While genome-wide association studies have identified risk loci for rheumatoid arthritis and other autoimmune diseases, in very few instances have causal variants driving risk been precisely defined.

Methods: We fine-mapped 76 autoimmune disease loci outside of the MHC, combining genotype data from rheumatoid arthritis (RA) and type 1 diabetes (T1D) case-control studies. After sequencing 799 1kb regulatory (H3K4me3) regions within these loci in 568 individuals, we observed accurate imputation for 89% of common variants. We applied Bayesian fine-mapping for these loci in RA (11,475 cases, 15,870 controls), T1D (9,334 cases and 11,111 controls) and combined datasets.

Results: We reduced the number of potential causal variants to ≤5 in 8 RA and 11 T1D loci. We identified causal missense variants in five loci (Table, DNASE1L3, SIRPG, PTPN22, SH2B3 and TYK2) and likely causal non-coding variants in six loci (Table, MEG3, TNFAIP3, CD28/CTLA4, ANKRD55, IL2RA, REL/PUS10). In the DNASE1L3 locus, the previously reported missensse RA lead SNP rs35677470 (posterior=0.81) causes a R206C change in DNASE1L3, which abolishes its nuclease function. In SIRPG, we identified the rs6043409 missense variant, causing a V263A substitution (posterior=0.25). For the loci with non-coding causal candidates, we performed stringent filtering based on allelic function and enhancer activity. Using this approach, in the CD28/CTLA4 locus, we identified independent variants associated with RA and T1D: the rs117701653 SNP near CD28, associated with RA (posterior=0.82), and the rs3087243 SNP associated with both RA and T1D (posterior=0.91). Functional assays including EMSA and luciferase assay showed allele specific, cell type specific activity for rs117701653 but not rs3087243. Additionally, in the TNFAIP3 locus in RA, we identified allele specific and cell type specific enhancer activity for the rs35926684 indel (posterior=0.24), but not the previously reported lead variant rs6920220. Finally, in the MEG3 locus, we identified the rs34552516 indel (posterior=0.37) for T1D with allele specific and cell type specific enhancer activity.

Conclusion: This study demonstrates the potential for dense genotyping and imputation to pinpoint missense and non-coding causal alleles.