Abstract Body

Background: Viral load setpoint is a major correlate of HIV disease progression. Genome-wide association studies have identified common human polymorphisms that together explain no more than 15% of its phenotypic variance. Concurrently, studies of the impact of HIV genetic diversity on viral load have produced highly variable estimates. Here we present a joint assessment of the respective contributions of human and viral variation to HIV viral load at setpoint.

Methods: Human genotype data across the Major Histocompatibility Complex (MHC) region, full-length consensus HIV sequences and setpoint viral load results were available for 1034 treatment naïve individuals of European ancestry, infected with HIV-1 subtype B. Heritability (h2) estimation was carried out with GCTA using three kernel matrices representing: 1) the human Genetic Relatedness Matrix (GRM) derived from the MHC, 2) the viral GRM derived from the full-length sequences, and 3) the sample-specific noise. The human GRM was estimated from 27 common polymorphisms shown to strongly influence viral load selected by LASSO. Bootstrapped phylogenetic trees were inferred from the viral sequences using RAxML. The viral GRM was derived from the phylogenetic trees by taking the branch length of the shared ancestry (i.e. the distance from the root to the most recent common ancestor). The estimates were repeated on 30 bootstrap trees and 15 bootstrap replicates of the samples. Clinical site was included as a covariates in the analyses.

Results: Estimating the host heritability of HIV viral load using the host GRM alone yielded a median estimate of h2=8% with an interquartile range (IQR) of 1% across 15 bootstrap replicates of the samples. The estimates of the viral heritability drawn from 30 bootstrapped viral trees had a median of 29% (IQR=10%). Combining both the host and viral relatedness matrices showed a comparable viral heritability of 26% (IQR=9%) but a decreased host contribution of 4% (IQR=0%).

Conclusions: This is the first estimate of the combined and respective contributions of the host and the viral genomes to the observed variability of HIV viral load. We showed that both the pathogen and host genomes have detectable impacts on the clinical outcome of infection, which are however not independent. These results suggest that the main predictor of clinical outcome is the viral genotype and that host determination of viral load is largely dependent on its ability to select viral variants.