HLA footprints on human immunodeficiency virus type 1 are associated with interclade polymorphisms and intraclade phylogenetic clustering

Abstract

The selection of escape mutations has a major impact on immune control of infections with viruses such as human immunodeficiency virus (HIV). Viral evasion of CD8(+) T-cell responses leaves predictable combinations of escape mutations, termed HLA "footprints." The most clearly defined footprints are those associated with HLA alleles that are linked with successful control of HIV, such as HLA-B*57. Here we investigated the extent to which HLA footprint sites in HIV type 1 (HIV-1) are associated with viral evolution among and within clades. First, we examined the extent to which amino acid differences between HIV-1 clades share identity with sites of HLA-mediated selection pressure and observed a strong association, in particular with respect to sites of HLA-B selection (P < 10(-6)). Similarly, the sites of amino acid variability within a clade were found to overlap with sites of HLA-selected mutation. Second, we studied the impact of HLA selection on interclade phylogeny. Removing the sites of amino acid variability did not significantly affect clade-specific clustering, reflecting the central role of founder effects in establishing distinct clades. However, HLA footprints may underpin founder strains, and we show that amino acid substitutions between clades alter phylogeny, underlining a potentially substantial role for HLA in driving ongoing viral evolution. Finally, we investigated the impact of HLA selection on within-clade phylogeny and demonstrate that even a single HLA allele footprint can result in significant phylogenetic clustering of sequences. In conclusion, these data highlight the fact that HLA can be a strong selection force for both intra- and interclade HIV evolution at a population level.

FIG. 1.

Sites of interclade variability and HIV-associated polymorphisms in p24 Gag. Consensus sequences (as of 2004) for HIV clades A1, A2, AE, B, and C are shown. Sites of interclade variability are marked with gray bars, and × indicates the sites of HLA-B-mediated selection pressure identified by analysis of subjects with C-clade infections in Durban (30). The four HLA-B*57/5801 epitopes in p24 Gag are enclosed in open boxes, with arrows indicating sites of HLA-B*57/5801-associated escape mutation that have been described in previous studies of B- and C-clade infections (6, 29, 30). There is a correlation between sites of interclade variability and sites of HLA-B-mediated selection; P < 10⁻⁶ (Fisher's exact test).

FIG. 2.

Proportion of amino acid residues at which HLA-B-associated polymorphisms are detected. All amino acids in each protein are represented, divided into “variable residues” at which there are interclade differences in amino acids and “conserved residues” that are identical among consensus sequences for clades A1, A2, AE, B, and C. The proportion of each of these sites at which HLA-B selection has been previously identified (30) is shown in gray. (a) p24 Gag. (b) RT. P values were calculated with Fisher's exact test.

FIG. 3.

Phylogenetic trees illustrating the preservation of clade phylogeny in the presence and absence of sites of clade-specific difference. Sequences from clades A1, A2, AE, B, and C were selected at random from www.hiv.lanl.gov, and ML phylogenetic trees were constructed (midpoint rooted, bootstrap values based on 100 replicates). Each clade is enclosed within a box to show clustering in the presence and absence of sites that vary between clades. The 2004 consensus sequence for each clade is marked by ×. (A) Tree constructed from complete p24 sequences (693 nucleotides). (B) Tree constructed from p24 sequences in the absence of 27 codons at which there is amino acid difference between clade consensus sequences (612 nucleotides).

FIG. 4.

Phylogenetic trees illustrating altered distribution of taxa when codons determining clade-specific differences are swapped between clades. Twenty sequences were selected at random from the clades indicated within dashed boxes. ML trees were constructed from nucleotides (midpoint rooted, bootstrap values based on 100 replicates). (A) Codons defining amino acids characteristic of clade A1 (selected from the clade A1 consensus) were superimposed on 20 sequences from clade B. The clade B sequences are shown twice, once without alteration (marked B) and once with A1-clade amino acids superimposed (marked B+A1). (B) Codons defining amino acids characteristic of clade C (selected from the clade C consensus) were superimposed on the same 20 sequences from clade B. As before, the clade B sequences are shown twice, unchanged (marked B) and bearing the characteristic C-clade codons (marked B+C).

FIG. 5.

Phylogenetic clustering in sequences from individuals with HLA-B*5703 with clustering analyzed in the presence and absence of the footprint sites. Phylogenetic clustering mediated by an HLA-B*5703 footprint was assessed in 100 NJ trees for Gag (with 20% of the sequences from subjects with HLA-B*5703) and 50 for Nef (with 16% of the sequences from subjects with HLA-B*5703). The mean difference in parsimony score is plotted, with error bars showing 95% confidence intervals. A significantly greater degree of phylogenetic clustering was observed in the presence of HLA-B*5703 footprint sites than in the absence of these sites for both proteins (P < 0.0001; paired t test).

FIG. 6.

Phylogenetic clustering as a consequence of artificial imposition of HLA-B*5703 mutations on HLA-B*5703-negative Gag sequences. ML phylogenetic trees constructed from 100 Gag sequences (selected at random from a pool of HLA-B*5703-negative individuals). The same 100 sequences are represented in each tree, and the same 20 are marked with arrows. (A) Twenty taxa were selected at random (marked with arrows). In this panel, no footprint mutations have been added. (B) The same sequences after the addition of three HLA-B*5703 footprint mutations to the arrowed sequences. (C) The same sequences after the addition of five footprint mutations. Progressive phylogenetic clustering among sequences bearing the HLA-B*5703 footprint is evident.

FIG. 7.

Model to show difference in phylogenetic clustering as five HLA-B*5703 footprint mutations were superimposed on Gag sequences. Each panel shows the mean difference in parsimony score between trees with no mutations and trees built from the same taxa with sequential HLA-B*5703 footprint mutations superimposed. Addition of mutations increases the difference in parsimony scores, reflecting progressive phylogenetic clustering (error bars show 95% confidence intervals; r² from linear regression with the y intercept set to go through the origin). (A) One hundred Gag sequences were selected at random from HLA-B*5703-negative individuals and used to construct an NJ phylogenetic tree. Five characteristic HLA-B*5703 footprint mutations were added, one at a time, to 20 sequences in each tree. Twenty repetitions are shown. (B) All 526 taxa from HLA-B*5703-negative subjects were used. The footprint mutation was added to a varying proportion of sequences (individual sequences selected at random). Ten repetitions are shown.