Background: Genome-wide association studies of HIV outcome consistently identify the MHC region as the major genetic influence on disease progression. Through establishment of the International Collaboration for the Genomics of HIV we sought to bring together all existing GWAS data in HIV patients to maximize power to uncover further association signals.
Methods: Genome-wide SNP data were collected from 25 clinical centers. Plasma viral load measurements obtained during the chronic phase of untreated infection were available for 6,315 individuals of European ancestry. Missing genotypes were imputed using the 1,000 Genomes Project reference. Single marker association testing was performed per study using linear regression and combined across studies by meta-analysis. Classical HLA alleles and amino acid variants were imputed using the SNP2HLA pipeline. Heritability estimates were calculated using GCTA.
Results: Consistent with previous studies, the top association signal was amino acid position 97 in HLA-B (p=4e-143), with independent associations at positions 67 (p=4e-112) and 45 (p=8e-49) in HLA-B and positions 77 (p=2e-12) and 95 (p=4e-5) in HLA-A. All 5 amino acid positions are located in the peptide-binding groove. Controlling for these positions fully accounted for classical class I HLA allele associations. Independent of the MHC, a second peak of association in the CCR5 region (top SNP rs4317138 p=8e-19) was observed. Conditioning out the known effect of CCR5Δ32, 57 SNPs remained associated at genome-wide significance (p<5e-8) suggesting multiple additional causal variants in this region. Considering all SNPs and assuming an additive genetic model, we obtained a narrow-sense heritability (h2) estimate of 26% (p<2e-16) for viral load. Removing the HLA and CCR5 regions, we observed a reduced, yet still significant h2 estimate of 16% (p=5e-8).
Conclusions: By combining available GWAS data in HIV infected individuals we refined the known association signals in HLA class I and CCR5. Controlling for these main effects, we uncovered additional, independent associations in known regions. Heritability analysis suggested that variation outside known regions, captured through common SNPs, also contribute to HIV control.