HIV-1 evolutionary dynamics under non-suppressive antiretroviral therapy

Background: Viral population dynamics in long term viraemic antiretroviral therapy (ART) treated individuals have not been well characterised. Prolonged virologic failure on 2nd-line protease inhibitor (PI) based ART without emergence of major protease mutations is well recognised, providing an opportunity to study within-host evolution. Methods: Using next-generation Illumina short read sequencing and in silico haplotype reconstruction we analysed whole genome sequences from longitudinal plasma samples of eight chronically infected HIV-1 individuals failing 2nd-line regimens from the ANRS 12249 TasP trial, in the absence of high frequency major PI resistance mutations. Plasma drug levels were measured by HPLC. Three participants were selective for in-depth variant and haplotype analyses, each with five or more timepoints spanning at least 16 months. Results: During PI failure synonymous mutations were around twice as frequent as non-synonymous mutations across participants. Prior to or during exposure to PI, we observed several polymorphic amino acids in gag (e.g. T81A, T375N) which are have also been previously associated with exposure to protease inhibitor exposure. Although overall SNP frequency at abundance above 2% appeared stable across time in each individual, divergence from the consensus baseline sequence did increase over time. Non-synonymous changes were enriched in known polymorphic regions such as env whereas synonymous changes were more often observed to fluctuate in the conserved pol gene. Phylogenetic analyses of whole genome viral haplotypes demonstrated two common features: Firstly, evidence for selective sweeps following therapy switches or large changes in plasma drug concentrations, with hitchhiking of synonymous and non-synonymous mutations. Secondly, we observed competition between multiple viral haplotypes that intermingled phylogenetically alongside soft selective sweeps. The diversity of viral populations was maintained between successive timepoints with ongoing viremia, particularly in env. Changes in haplotype dominance were often distinct from the dynamics of drug resistance mutations in reverse transcriptase (RT), indicating the presence of softer selective sweeps and/or recombination. Conclusions: Large fluctuations in variant frequencies with diversification occur during apparently "stable" viremia on non-suppressive ART. Reconstructed haplotypes provided further evidence for sweeps during periods of partial adherence, and competition between haplotypes during periods of low drug exposure. Drug resistance mutations in RT can be used as markers of viral populations in the reservoir and we found evidence for loss of linkage disequilibrium for drug resistance mutations, indicative of recombination. These data imply that even years of exposure to PIs, within the context of large stable populations displaying ongoing selective competition, may not precipitate emergence of major PI resistance mutations, indicating significant fitness costs for such mutations. Ongoing viral diversification within reservoirs may compromise the goal of sustained viral suppression.