Evidence of viral adaptation to HLA class I-restricted immune pressure in chronic hepatitis C virus infection

Abstract

Cellular immune responses are an important correlate of hepatitis C virus (HCV) infection outcome. These responses are governed by the host's human leukocyte antigen (HLA) type, and HLA-restricted viral escape mutants are a critical aspect of this host-virus interaction. We examined the driving forces of HCV evolution by characterizing the in vivo selective pressure(s) exerted on single amino acid residues within nonstructural protein 3 (NS3) by the HLA types present in two host populations. Associations between polymorphisms within NS3 and HLA class I alleles were assessed in 118 individuals from Western Australia and Switzerland with chronic hepatitis C infection, of whom 82 (69%) were coinfected with human immunodeficiency virus. The levels and locations of amino acid polymorphisms exhibited within NS3 were remarkably similar between the two cohorts and revealed regions under functional constraint and selective pressures. We identified specific HCV mutations within and flanking published epitopes with the correct HLA restriction and predicted escaped amino acid. Additional HLA-restricted mutations were identified that mark putative epitopes targeted by cell-mediated immune responses. This analysis of host-virus interaction reveals evidence of HCV adaptation to HLA class I-restricted immune pressure and identifies in vivo targets of cellular immune responses at the population level.

FIG. 1.

Similar HLA-A, -B, and -C allele distributions within the Swiss and WA cohorts. Bars indicate the percentage of individuals with the respective HLA alleles, with dark gray bars indicating the HLA allele distributions within the Swiss cohort and light gray bars indicating the WA cohort. The bar corresponding to “others” groups the alleles occurring at low frequencies. The extremely polymorphic nature of HLA-B is reflected in a larger number of alleles at low frequency compared to HLA-A or -C. The HLA allele distributions for the three loci did not differ significantly between the two cohorts (P = 0.5, 0.1, and 0.6 for HLA-A, -B, and -C alleles, respectively, by Fisher's exact test).

FIG. 2.

Levels of amino acid polymorphism exhibited within NS3 are remarkably similar between the Swiss and WA cohorts. (A) Amino acid polymorphism profiles of the two cohorts within the NS3 region. Known conserved motifs of the NS3 helicase region are indicated by gray boxes and named according to the respective publications (13, 39). The inset provides a detailed analysis of the polymorphism profile of amino acids 1057 to 1127, with arrows indicating active sites within NS3 protease (16). (B) Sites within NS3 where the polymorphism rates between the two cohorts differ by more than 10%. Higher polymorphism within the WA cohort is indicated by bars above the line, and when the amount of polymorphism is greater in the Swiss cohort, bars are below the line. Diamonds mark differences in consensus amino acid between the cohorts.

FIG. 3.

Phylogenetic analyses of the NS3 region in HCV reveals clustering of neutral or “near-neutral” changes that appear to disperse when examining sites are under selective pressure. Nucleotide sequences greater than 1,000 bp in the NS3 region were aligned using the program ClustalW, and the resultant alignment was then used in the phylogenetic package Mega 3.1 (15). Phylogenetic analysis of the NS3 region was performed using the neighbor-joining method based on the modified Nei-Gojobori model for synonymous and nonsynonymous sites with pairwise deletion. Bootstrap replications (n = 1,000) were performed for both analyses. Sequences with less than 1,000 bp were not included; thus, not all individuals contributed to this analysis. (A) Analysis of synonymous changes reveals distinct clusters. Within the genotype 1a cluster, sequences form additional clusters which largely disappear when analyzing only nonsynonymous changes (B). Viral sequences that fall into clusters in panel A are marked with identical colors; sequences that do not fall into clusters are not colored. Triangles mark sequences from Swiss individuals, and circles mark sequences from WA individuals. For both trees, numbers at nodes are bootstrap values, indicating the percentage of 1,000 replications. Only bootstrap values above 60% are shown. The bars at the bottom of the figure indicate the number of substitutions per site.

FIG. 4.

Significant (P < 0.05) associations of HLA alleles with viral polymorphisms within NS3 mark relevant immunological sites within the virus. (A) HLA associations with viral polymorphisms (P < 0.05) are shown as asterisks; those within or flanking (5 amino acids) published epitopes are boxed. The HLA allele and position of the association are shown. Five of the seven positions shown had an association with an allele-wise false discovery rate of less than 10%, while two positions (NS3-1094 and NS3-1635) had false discovery rates above this cutoff. (B) The top panel shows published HLA-restricted epitopes as listed in the HCV database at http://www.hcv.lanl.gov. Specific HLA associations with viral mutations found within the NS3 region are shown as either positive associations (variation from consensus amino acid is overrepresented in the HLA-positive group) or negative associations (maintenance of consensus amino acid is overrepresented in the HLA-positive group) for HLA-A, -B, and -C alleles. Red indicates associations that fall within published epitopes, and blue indicates an association flanking a published epitope. Vertical bars indicate the amount of amino acid variation within the combined cohort for each site, and the indicated amino acid is consensus. Red vertical bars indicate nonconservative amino acid changes, and blue bars indicate conservative amino acid changes (22). Odds ratios are indicated below the HLA allele. Hashed boxes indicate likely HLA haplotype structures. The conserved helicase motifs within this region are indicated and span highly conserved residues in the combined cohort.

FIG. 5.

Location of known epitopes and HLA associations within NS3. The model of NS3 was taken from the NCBI PubMed Molecular Graphics Program Cn3D. The positions of the epitopes shown in Fig. 4, 6, and 8 are marked (yellow). Black arrows mark position P1 of the respective epitopes; red arrows mark significant HLA associations.

FIG. 6.

HLA-restricted mutations in individuals with chronic HCV infection within published HLA class I-restricted epitopes. The first line indicates the published epitopes, and the second line indicates the consensus amino acid from both cohorts, with significant associations found in this study shown by white letters on a black background and with an arrow. Identity with consensus is indicated by dots, and deletions are indicated by white dashes on a black background; the sequences are sorted by the presence of the respective HLA alleles. Mutations within epitopes are shown as white letters on a black background, outside epitopes are shown in dark gray, and positions that appear to contain HCV genotype 1b-specific polymorphisms are boxed. A lowercase letter indicates the presence of two major viral quasispecies including consensus. Sequences from individuals infected with HCV genotype 1b are marked by a closed circle. Individuals with incomplete sequence for the respective epitopes were excluded. HLA-B*0801- and HLA-A*0101-positive individuals are marked with circles in the respective phylogenetic trees. Trees include all sites and were generated as described in the legend to Fig. 3. Individual sequences with mutations at positions with significant associations are marked with closed circles; those with consensus amino acids are marked with open circles. (A) Demonstration of a positive HLA association with viral polymorphism in two overlapping epitopes (B8-1395 and B8-1402). The K1397R and L1410X substitutions were significantly (P = 0.007 and P = 0.006 in multivariate analyses, respectively, see Materials and Methods) more frequent in HLA-B*0801-positive than in HLA-B*0801-negative individuals. One individual has a large deletion spanning two neighboring HLA-B*0801 epitopes (marked with a closed triangle) and could not be included in the phylogenetic analyses. (B) Demonstration of a negative HLA association with viral polymorphism (F1444Y) within the known HLA-A1-1436 epitope. The consensus amino acid F appears to be maintained within the HLA-A*0101-positive individuals compared to the increased incidence of mutation from consensus (Y) in the HLA-A*0101-negative individuals. The phylogenetic analyses do not reveal clustering of individual sequences that share the same HLA-restricted polymorphism.

FIG. 7.

Statistical significance of HLA associations with viral polymorphism within published epitopes for HLA-B8 (A) and HLA-A2 (B). We assessed the likelihood of finding true associations between HCV polymorphism and HLA alleles HLA-A2 and -B8 by comparing the distributions of the Fisher's exact P values for amino acid residues lying within and outside of published HLA-A2 and -B8 epitopes. For example, the 631 univariate P values for the HLA-B8 associations with polymorphism in NS3 were classified as to whether the position was within or outside of the published HLA-B8 epitopes (www.hcv.lanl.gov). Using the P values as the “time” base, Kaplan-Meier product limit estimators compare the distributions of the P values for the groups; observations were censored for P values of >0.50. The curves compare the P values for associations of the HLA alleles lying outside of (solid line) and within (broken line) the published epitopes for the allele noted. The differences in the curves were evaluated using proportional hazards models that included covariates for position within a published epitope (dichotomous indicator) and the estimated power to detect an OR of >2.0 or <0.5. For HLA-B8, the early and overall separation of the curves indicates that the P values for residues within the epitopes are smaller overall than those outside epitopes, and this difference is significant after adjustment for the power of the tests for association (proportional hazards model adjusted P value = 0.009). In contrast, no statistically significant difference was observed for the P values for HLA-A2 associations when stratified by residue location within or outside of A2 epitopes (P = 0.602). Finally, there were no differences between the P values comparing (i) HLA-B8 associations within and outside of HLA-A2 epitopes and (ii) HLA-A2 associations within and outside of HLA-B8 epitopes (data not shown).

FIG. 8.

Lack of significant HLA associations within known immunodominant HLA-A2 epitopes. (A) A2-1073. (B) A2-1406. Examination of the two HLA-A2 epitopes finds no significant accumulation by multivariate analysis (see Materials and Methods) of mutations at single residues within the HLA-A*0201-carrying individuals. All three patients with the Y1082F mutation affecting proteasomal processing were infected with HCV genotype 1b, as were the individuals in the experimental setting described by Seifert and colleagues (26). The first line indicates the published epitopes, and the second line indicates the consensus amino acid from both cohorts. Identity with consensus is indicated by dots, and deletions are indicated by white dashes on a black background; the sequences are sorted by the presence of the respective HLA alleles. Mutations within epitopes are shown by white letters on a black background, outside epitopes are shown in dark gray, and mutations that appear specific to HCV genotype 1b are boxed. A lowercase letter indicates the presence of two major viral quasispecies, including the consensus. Sequences from individuals infected with HCV genotype 1b are marked by a closed circle; one individual with a large deletion is marked by a closed triangle. Individuals with incomplete sequence for the respective epitopes were excluded.