Authors:
U. Özbek (New York, NY, US)
A. Kennedy (Bethesda, MD, US)
S. Singh (Miami, FL, US)
M. Barbieri (Miami, FL, US)
I. Konidari (Miami, FL, US)
J. McCauley (Miami, FL, US)
M. Dorak (Liverpool, GB)
Single-nucleotide polymorphisms (SNPs) across the HLA region are often significant in genome-wide association studies (GWAS), but most of these associated SNPs do not correspond to classical HLA alleles, even in immune-mediated disorders. Another level of functional HLA region variation (epitopes/supertypes) is relevant in disease associations, but has not attracted much attention in GWAS. To initiate a systematic survey of their involvement with GWAS findings, we examined correlations between SNPs and common HLA epitopes (Bw4/Bw6; C1/C2) / genetic supertypes (DR52/DR53) in a panel of 95 HLA-typed cell lines using ImmunoChip genotyping. We examined univariate correlations between 8,045 SNPs that passed quality control procedures and HLA epitopes/supertypes. We examined the reported GWAS associations of SNPs that were most significantly correlated with HLA epitopes/supertypes on GRASP and PhenoScanner. We found proxy SNPs for each epitope/supertype with r>0.50 (P<0.0001). The HLA-C region SNP rs9264942 correlated with Bw4. This SNP is the top GWAS hit for HIV-1 control (P=3E-35) and implicated in HLA-C expression levels. Our result offers an alternative explanation that rs9264942 is a proxy for Bw4 (a known marker for HIV-1 control). rs9264942 is also the top GWAS marker for Crohns disease in the HLA class I region. The strongest proxy for C1/C2, rs12211087, is the top genome-wide risk marker for psoriasis via its proxy (r2=1) rs4406273 (P=5E-723). Four SNPs showed strong correlations (r>0.97) with DR53 (rs2133035, rs9271574, rs9271488, rs3104389) and were among the top GWAS hits for rheumatoid arthritis, ulcerative colitis and sarcoidosis, and strongest eQTLs for HLA-DQA2, -DQA1, -DRB6 and -DRB1. We identified three ImmunoChip SNPs (rs2097432, rs9271613, rs9271850) as DR52 proxies, and their GWAS associations included the top genome-wide risk marker for systemic sclerosis (via proxy rs3129763; P=9E-187). The proxies for DR52 were the strongest eQTLs for HLA class II genes (P=3E-189 for HLA-DQA1). Using two SNPs together increased the strength of the correlation with DR52 (for rs3129887-A - rs2097432-G, r=0.94, P=1E-38). Our results suggest that the interpretation of HLA region SNP associations can be improved by taking into account additional levels of functional variation within the HLA region.