Correlates of protective cellular immunity revealed by analysis of population-level immune escape pathways in HIV-1

Abstract

HLA class I-associated polymorphisms identified at the population level mark viral sites under immune pressure by individual HLA alleles. As such, analysis of their distribution, frequency, location, statistical strength, sequence conservation, and other properties offers a unique perspective from which to identify correlates of protective cellular immunity. We analyzed HLA-associated HIV-1 subtype B polymorphisms in 1,888 treatment-naïve, chronically infected individuals using phylogenetically informed methods and identified characteristics of HLA-associated immune pressures that differentiate protective and nonprotective alleles. Over 2,100 HLA-associated HIV-1 polymorphisms were identified, approximately one-third of which occurred inside or within 3 residues of an optimally defined cytotoxic T-lymphocyte (CTL) epitope. Differential CTL escape patterns between closely related HLA alleles were common and increased with greater evolutionary distance between allele group members. Among 9-mer epitopes, mutations at HLA-specific anchor residues represented the most frequently detected escape type: these occurred nearly 2-fold more frequently than expected by chance and were computationally predicted to reduce peptide-HLA binding nearly 10-fold on average. Characteristics associated with protective HLA alleles (defined using hazard ratios for progression to AIDS from natural history cohorts) included the potential to mount broad immune selection pressures across all HIV-1 proteins except Nef, the tendency to drive multisite and/or anchor residue escape mutations within known CTL epitopes, and the ability to strongly select mutations in conserved regions within HIV's structural and functional proteins. Thus, the factors defining protective cellular immune responses may be more complex than simply targeting conserved viral regions. The results provide new information to guide vaccine design and immunogenicity studies.

Fig 1

The proportion of sequence variation attributable to HLA-associated selection pressures differs markedly by HIV-1 protein. The proportion of codons exhibiting ≥99.5% amino acid conservation at the population level in our cohort and the proportion of variable codons (those exhibiting <99.5% amino acid conservation) harboring no known HLA-associated polymorphisms are indicated, along with the proportion of codons harboring at least one HLA-associated polymorphism. Numbers in parentheses in the x-axis labels indicate the total length in amino acids of each protein (including the stop codon, if present). **, P < 0.001; ***, P < 0.0001 (Fisher's exact test compared to all other proteins combined).

Fig 2

Widespread differential escape between HLA allele group members correlates with evolutionary distance between HLA genes. Individual HLA-A, -B, and -C types are indicated. The proportion of differential escape (defined as the proportion of HLA-associated polymorphisms for which the odds of escape differed significantly between at least one HLA subtype and the rest of the group) differs markedly for various common types of the HLA-A, -B, and -C loci (A). HLA types exhibiting the highest level of differential escape were those exhibiting the greatest intratype genetic diversity (B). HLA-B*07 and B*15 (exhibiting the highest and lowest proportion of differential escape, respectively) and B*40 are labeled for special interest.

Fig 3

The distribution and natural log odds ratio of escape within known epitopes is nonuniform and biased toward anchor residues. The distribution and natural log odds ratio of escape mutations within known CD8⁺ 9-mer epitopes (A, B, and C) and 8-/10-/11-mer epitopes (D, E, and F) are shown. Frequencies of escape by relative position within or occurring at the ±3 amino acids flanking the epitope N and C termini are depicted as histograms in panels A and D. In panel D, notation of the type C−3 on the x axis indicates the number of positions upstream of the C terminus (in this case, 3). These same data are depicted as fold differences from expected frequency of escape occurring at HLA-specific anchor, nonanchor, and epitope flanking positions (based on the null hypothesis of equal escape probability across all positions) and are shown in panels B and E. *, P < 0.05; **, P <0.001; ***, P < 0.0001 (binomial test). The natural log of the odds ratio of escape at HLA-specific anchor, nonanchor, and epitope flanking positions are shown as scatter plots (C and F).

Fig 4

Fold changes in predicted peptide-HLA binding affinities following escape at anchor residues. Fold changes in predicted peptide-HLA binding affinities as a consequence of escape at HLA-specific anchor (versus nonanchor) residues within 9-mer (A) and 8-, 10-, and 11-mer (B) epitopes are shown as Tukey box plots.

Fig 5

Population-level immune escape pathways in HIV-1 reveal correlates of protective immunity: selected results. Individual HLA-A, -B, and -C alleles are as indicated. Significant inverse correlations were observed between the total number of sites under immune selection by a given HLA across the entire HIV-1 proteome (A), the median odds ratio of escape in Gag (B), and the percent escape at HLA anchor residues (C) and published hazard ratios of AIDS progression. HLA-B*57:01 and B*27:05 are labeled for special interest.