HIV evolution in early infection: selection pressures, patterns of insertion and deletion, and the impact of APOBEC

Abstract

The pattern of viral diversification in newly infected individuals provides information about the host environment and immune responses typically experienced by the newly transmitted virus. For example, sites that tend to evolve rapidly across multiple early-infection patients could be involved in enabling escape from common early immune responses, could represent adaptation for rapid growth in a newly infected host, or could represent reversion from less fit forms of the virus that were selected for immune escape in previous hosts. Here we investigated the diversification of HIV-1 env coding sequences in 81 very early B subtype infections previously shown to have resulted from transmission or expansion of single viruses (n = 78) or two closely related viruses (n = 3). In these cases, the sequence of the infecting virus can be estimated accurately, enabling inference of both the direction of substitutions as well as distinction between insertion and deletion events. By integrating information across multiple acutely infected hosts, we find evidence of adaptive evolution of HIV-1 env and identify a subset of codon sites that diversified more rapidly than can be explained by a model of neutral evolution. Of 24 such rapidly diversifying sites, 14 were either i) clustered and embedded in CTL epitopes that were verified experimentally or predicted based on the individual's HLA or ii) in a nucleotide context indicative of APOBEC-mediated G-to-A substitutions, despite having excluded heavily hypermutated sequences prior to the analysis. In several cases, a rapidly evolving site was embedded both in an APOBEC motif and in a CTL epitope, suggesting that APOBEC may facilitate early immune escape. Ten rapidly diversifying sites could not be explained by CTL escape or APOBEC hypermutation, including the most frequently mutated site, in the fusion peptide of gp41. We also examined the distribution, extent, and sequence context of insertions and deletions, and we provide evidence that the length variation seen in hypervariable loop regions of the envelope glycoprotein is a consequence of selection and not of mutational hotspots. Our results provide a detailed view of the process of diversification of HIV-1 following transmission, highlighting the role of CTL escape and hypermutation in shaping viral evolution during the establishment of new infections.

Figure 1. Posterior probabilities of belonging to the selection class (ω>1) for all sites in gp160.

The dashed line indicates the 0.5 posterior probability, which we used as a threshold to assign sites to the selection site class. Flat parts of the graph correspond to sequence regions that were masked either because they were poorly aligned or coding in more than one frame.

Figure 2. Three-dimensional structure context of the selected sites identified in gp120.

(A) 13 sites from the dataset excluding sequences with evidence of APOBEC-mediated hypermutation, and (B) 4 additional sites identified analyzing the complete dataset. The sites depicted in blue represent those sites that are embedded in a known or potential CTL epitope and circled site numbers are potentially affected by APOBEC hypermutation. Sites marked with an asterisk occur in a region for which there is no available structure, therefore the positions are shown in proximity to their actual locations.

Figure 3. Selected sites that are embedded in potential CTL epitopes.

The patient consensus sequence is at the top of each alignment, with Fiebig stage, patient ID, and CON for consensus indicated. The proposed epitope is shown beneath the patient consensus, followed by the HLA. Previously reported epitopes are provided in full, predicted epitopes are written with uppercase letters representing the anchor motif embedded in a string of x's. (A–E) are based on i) evidence for selection from the model implemented in HyPhy, ii) evidence of selection from locally concentrated regional evidence for selection relative to extremely homogeneous early infection alignments, and iii) HLA appropriate known epitopes or anchor motifs found using the Los Alamos HIV immunology database epitope location finder, ELF. (F) This potential escape mutation was found by the phylogenetic test for changes enriched in later Fiebig stages, and was also supported by (ii) and (iii) above (Table 2). Mutations in (A–C) also were favored in later Fiebig stages. (G) This set of mutations includes the only other sites in the entire data set of 81 subjects where mutations tended to cluster over a small region in one person. No site in this region had statistical support for positive selection, but this cluster of mutations was embedded in a potential epitope.

Figure 4. Maximum likelihood tree including all sequences from all 81 homogeneous patients illustrating the pattern of change in one site.

The tree is based on all nucleotide sequences and all positions in Env, and illustrates the mutational patterns in the context of the full tree at amino acid position 83. Position 83 is highly conserved, and is His in most sequences in most subjects (indicated by the red Xs). In the 4 subjects that are featured, it changes from His to Tyr in at least some of the sequences in each subject, indicated by the blue 0 s. Although the trees were generated for each patient separately for the HyPhy analysis, this kind of repeated mutation pattern in a single site is an illustration of the kind of change captured in the dN/dS ratio of the HyPhy analysis (Table 2). This site, 83, was also identified as interesting in the full tree analysis as it tended to change from H to Y in patients at later Fiebig stages (Table 2).

Figure 5. Cumulative distribution of insertions and deletions in acutely infected patients.

(A) The frameshifting single base indels observed in early infection were distributed throughout the Env gene, with a few concentrated regions localized in particular strings of like bases (red and black lines in the figure). As these substitutions are lethal, they must have arisen in the newly infected individual. In contrast, most of the single base indels in the 21 heterogeneous infection cases were compensatory, associated with nearby mutations that resolved the reading frame, and most of the compensated indels are localized in the hypervariable domains. (B) Similarly, multiple base frameshifting indels are distributed throughout Env, while in-frame indels are concentrated in the hypervariable regions.