HIV-1 Evolutionary Dynamics under Nonsuppressive Antiretroviral Therapy

Abstract

Prolonged virologic failure on 2nd-line protease inhibitor (PI)-based antiretroviral therapy (ART) without emergence of major protease mutations is well recognized and provides an opportunity to study within-host evolution in long-term viremic individuals. Using next-generation sequencing and in silico haplotype reconstruction, we analyzed whole-genome sequences from longitudinal plasma samples of eight chronically infected HIV-1-positive individuals failing 2nd-line regimens from the French National Agency for AIDS and Viral Hepatitis Research (ANRS) 12249 Treatment as Prevention (TasP) trial. On nonsuppressive ART, there were large fluctuations in synonymous and nonsynonymous variant frequencies despite stable viremia. Reconstructed haplotypes provided evidence for selective sweeps during periods of partial adherence, and viral haplotype competition, during periods of low drug exposure. Drug resistance mutations in reverse transcriptase (RT) were used as markers of viral haplotypes in the reservoir, and their distribution over time indicated recombination. We independently observed linkage disequilibrium decay, indicative of recombination. These data highlight dramatic changes in virus population structure that occur during stable viremia under nonsuppressive ART. IMPORTANCE HIV-1 infections are most commonly initiated with a single founder virus and are characterized by extensive inter- and intraparticipant genetic diversity. However, existing literature on HIV-1 intrahost population dynamics is largely limited to untreated infections, predominantly in subtype B-infected individuals. The manuscript characterizes viral population dynamics in long-term viremic treatment-experienced individuals, which has not been previously characterized. These data are particularly relevant for understanding HIV dynamics but can also be applied to other RNA viruses. With this unique data set we propose that the virus is highly unstable, and we have found compelling evidence of HIV-1 within-host viral diversification, recombination, and haplotype competition during nonsuppressive ART.

Keywords: antiretroviral resistance; clinical failure; drug resistance evolution; human immunodeficiency virus.

FIG 1

Sequence divergence for eight patients under nonsuppressive ART. These data were for SNPs detected by Illumina NGS at <2% abundance. Sites had coverage of at least 10 reads. (a and b) In both (a) synonymous and (b) nonsynonymous mutations, there was an idiosyncratic change in the number of SNPs relative to the reference strain over time. (c) Mixed-effects linear model of divergence from the baseline time point and (d) consensus C subtype. The trend and 95% confidence interval (CI) (gray shadows) suggest that there is no strong positive or negative linear relationship between divergence and time.

FIG 2

(a) Multidimensional scaling showing clustering of HIV whole genomes from consensus sequences with high intrapatient diversity. Multidimensional scaling (MDS) was created by determining all pairwise distance comparisons under a TN93 substitution model, color-coded by patient. Axes are MDS-1 and MDS-2. (b) Maximum likelihood phylogeny of constructed viral haplotypes for all patients. The phylogeny was rooted on the clade C reference genome (GenBank accession number AF411967). Reconstructed haplotypes were genetically diverse and did not typically cluster by time point.

FIG 3

(a) Pairwise linkage disequilibrium decays rapidly with increasing distance between SNPs. Lines represent patterns of LD for each patient examined in-depth. There was a constant decrease in linkage disequilibrium over the first 800 bp. (b) Putative recombination breakpoints and drug resistance-associated mutations of all longitudinal consensus sequences belonging to three patients, 15664, 16207, and 22763. All sequences were colored uniquely; perceived recombination events supported by 4 or more methods implemented in RDP5 are highlighted with a red border and italic text to show the major parent and recombinant portion of the sequence. Drug resistance-associated mutations are indicated with a red arrow, relative to the key at the bottom of the image. For ease of distinguishment, the K65R mutations are indicated with a blue arrow.

FIG 4

Drug regimen, adherence, and viral dynamics within patient 15664. (a) Viral load and drug levels. At successive time points the drug regimen was noted and the plasma drug concentration measured by HPLC (nmol/L). The patient was characterized by multiple partial suppression (<750 copies/mL, 16 months; <250 copies/mL, 22 months) and rebound events (red dotted line) and poor adherence to the drug regimen. (b) Drug resistance- and non-drug resistance-associated nonsynonymous mutation frequencies determined by Illumina NGS. The Patient had large population shifts between time points 1 and 2, consistent with a hard selective sweep, coincident with the shift from the 1st-line regimen to 2nd-line. (c) Synonymous mutation frequencies. All mutations with a frequency of <10% or >90% at two or more time points were tracked over successive time points. Most changes were restricted to gag and pol regions and had limited shifts in frequency, i.e., between 20 and 60%. (d) Maximum likelihood phylogeny of reconstructed haplotypes. Haplotypes largely segregated into three major clades (labeled A to C). Majority and minority haplotypes, some carrying lamivudine resistance mutation M184V. Clades referred to in the text body are shown to the right of the heatmap.

FIG 5

Drug regimen, adherence, and viral dynamics of patient 22763. (a) Viral load and regimen adherence. At successive time points the regimen was noted and plasma drug concentration was measured by HPLC (nmol/L). The patient had therapeutic levels of drug at several time points (3, 5, and 8), indicating variable adherence to the prescribed drug regimen. (b) Drug resistance- and non-drug resistance-associated nonsynonymous mutation frequencies. The patient had numerous drug resistance mutations in dynamic flux. Between time points 4 and 7, there was a complete population shift, indicated by reciprocal competition between the RT mutation K65R and the TAMs K67N and K70R. (c) Synonymous mutations frequencies. All mutations with a frequency of <10% or >90% at two or more time points were followed over successive time points. Several env mutations mimicked the nonsynonymous shifts observed between time points 2 and 4, suggestive of linkage. (d) Maximum likelihood phylogeny of reconstructed haplotypes. Time points 1 to 4 were found in distinct lineages. In later time points, from 5 to 8, haplotypes became more intermingled while maintaining antagonism between K65R- and K67N-bearing viruses.

FIG 6

Drug regimen, adherence, and viral dynamics within patient 16207. (a) Viral load and drug levels. At successive time points the regimen was noted and the plasma drug concentration measured by HPLC (nmol/L). The patient displayed ongoing viremia and poor adherence to the prescribed drug regimen. (b) Drug resistance- and non-drug resistance-associated nonsynonymous mutation frequencies. The patient had only one major RT mutation, K103N, for the duration of the treatment period. Several antagonistic nonsynonymous switches predominantly in env were observed between time points 1 and 4. (c) Synonymous mutation frequencies. All mutations with a frequency of <10% or >90% at two or more time points were followed over successive time points. In contrast to nonsynonymous mutations, most synonymous changes were in pol, indicative of linkage to the env coding changes. (d) Maximum likelihood phylogeny of reconstructed haplotypes. Haplotypes were again clearly divided intro three distinct clades; each clade contained haplotypes from all time points, suggesting a lack of hard selective sweeps and intermingling of viral haplotypes with softer sweeps. Most viral competition occurred outside drug pressure.