HIV-1 variants are archived throughout infection and persist in the reservoir

Abstract

The HIV-1 reservoir consists of latently infected cells that persist despite antiretroviral therapy (ART). Elucidating the proviral genetic composition of the reservoir, particularly in the context of pre-therapy viral diversity, is therefore important to understanding reservoir formation and the persistence of latently infected cells. Here we investigate reservoir proviral variants from 13 Zambian acutely-infected individuals with additional pre-therapy sampling for a unique comparison to the ART-naïve quasispecies. We identified complete transmitted/founder (TF) viruses from seroconversion plasma samples, and additionally amplified and sequenced HIV-1 from plasma obtained one year post-infection and just prior to ART initiation. While the majority of proviral variants in the reservoir were most closely related to viral variants from the latest pre-therapy time point, we also identified reservoir proviral variants dating to or near the time of infection, and to intermediate time points between infection and treatment initiation. Reservoir proviral variants differing by five or fewer nucleotide changes from the TF virus persisted during treatment in five individuals, including proviral variants that exactly matched the TF in two individuals, one of whom had remained ART-naïve for more than six years. Proviral variants during treatment were significantly less divergent from the TF virus than plasma variants present at the last ART-naïve time point. These findings indicate that reservoir proviral variants are archived throughout infection, recapitulating much of the viral diversity that arises throughout untreated HIV-1 infection, and strategies to target and reduce the reservoir must therefore permit for the clearance of proviruses encompassing this extensive diversity.

Fig 1. Sampling strategy for study participants.

Viral RNA was reverse transcribed, amplified, and sequenced from pre-therapy plasma samples, while proviral DNA was amplified from cells collected during treatment. For six individuals (IDs Z1094F, Z1123M, Z1788F, Z1047M, Z1165M, and Z1658F), we additionally amplified and sequenced proviral DNA from cells at the last pre-therapy time point (red) to assess divergence from the transmitted/founder virus of contemporaneous sequences from plasma vs. cells. EDI = Estimated Date of Infection.

Fig 2. Bayesian root date estimation.

Bayesian inference was used to estimate the root date, or time of infection, from pre-therapy plasma env sequences. The 95% highest posterior density intervals surround the mean estimated root dates of the Bayesian inference (circles). The clinically estimated infection dates (EDIs) are depicted as stars.

Fig 3. Maximum-likelihood (ML) trees for viral and proviral variants in four individuals.

Representative ML phylogenetic trees for participants Z1123M (A), Z1124F (B), Z326M (C), and Z1047M (D) rooted on the respective transmitted/founder (TF) virus (grey) identified from the seroconversion sample and depicting all viral variants from one year post-infection (blue), the last ART-naïve sample (red), and during treatment (purple diamonds). Variants from cells collected at the last ART-naïve time point are shown in open red diamonds, while all plasma variants are in filled circles.

Fig 4. Divergence from transmitted/founder (TF) virus increases during ART-naïve infection and decreases on treatment.

(A) Patristic distances from the TF virus, or root-to-tip values, from the maximum-likelihood (ML) trees are shown for all 13 participants, with the single instance of multivariant infection in participant Z1658F shown at the far right. Where two samples were assessed during treatment, only the first is shown here. Mean values are indicated by horizontal black bars. (B) Summary of the mean intrapatient patristic distances, where the mean distance is significantly different between each time point assessed (Wilcoxon matched-pairs signed rank tests). (C) Proportion of proviruses seeded into the reservoir, by era, as estimated from the placement of reservoir sequence in the phylogeny. The mean values of proportions for all 13 volunteers are shown, while proportions for each participant are given in S4 Fig. A majority of variants are most closely related to variants present at the last ART-naïve time point. The percentage of variants demonstrating APOBEC hypermutation is also indicated, though these sequences were excluded from analysis.

Fig 5. Regression-based inference of time of provirus integration.

Representative figures for participant Z1165M (A-C) and participant Z634F (D-F). Maximum-likelihood trees of the env gene for pre-therapy variants (circles), including individual seroconversion variants (grey), one year sequences (blue), last pre-therapy sequences (red circles, plasma and open red diamonds, cells), and proviral variants (filled diamonds) in A and D. Two samples during treatment were assessed for both participants, with the first in purple, and subsequent in black. Trees were rooted to optimize the correlation between root-to-tip distance and sampling time for all pre-therapy plasma variants. The linear model relating root-to-tip distances to sampling time is shown in the dashed lines of figures B and E, with the pre-therapy variants denoted as colored dots, and the phylogenetic relationships between them denoted as faint grey lines. Proviral variants from samples collected during treatment are shown in filled diamonds in the same manner. The estimated integration dates of the proviral variants and 95% confidence intervals are shown in the plots C and F. Figures for additional individuals are in S5 Fig.

Fig 6. Maximum-likelihood (ML) trees of all variants for participants with two samples during treatment.

ML phylogenetic trees for all four participants: Z1788F (A), Z634F (B), N133M (C), and Z1165M (D) rooted on the respective transmitted/founder virus (grey) and depicting all viral variants from one year post-infection (blue), the last ART-naïve sample (red), and during treatment (purple and black diamonds, with second sample in black). Sequences from cells collected at the last ART-naïve time point are in open red diamonds, while plasma sequences are depicted in circles.

Fig 7. Distance from transmitted/founder (TF) virus decreases with subsequent sampling during treatment.

Patristic distances from the TF virus, or root, for reservoir variants as a proportion of the greatest intrapatient patristic distance (tree height) from the maximum-likelihood phylogenetic tree; means shown in horizontal black bars. Distances are lower for variants sampled at the second time point during treatment compared to the first (unpaired t test).