HLA-Cw*03-Restricted CD8⁺ T-Cell Responses Targeting the HIV-1 Gag Major Homology Region Drive Virus Immune Escape and Fitness Constraints Compensated for by Intracodon Variation^{▿ †}

Study subjects.

A total cohort of 778 subjects infected with C-clade HIV was studied. Of these, 578 individuals were recruited in Kwa-Zulu Natal (KZN) Province in South Africa, and 200 C-clade-infected subjects originally from sub-Saharan Africa were recruited in Oxford, United Kingdom, including the mother-child pair SL17. All individuals were naïve to antiretroviral therapy, and clinical data (CD4⁺ count and viral load) were available. However, information regarding the time of infection was not available, as all patients were chronically infected by the time of analysis.

ELISPOT screening.

A total of 778 study subjects were screened for HIV immune responses. Comprehensive epitope screening was achieved by incubating peripheral blood mononuclear cells (PBMCs) with overlapping peptides (OLPs) spanning the entire HIV proteome and measuring the subsequent gamma interferon (IFN-γ) response in an enzyme-linked immunospot (ELISPOT) assay as previously described (1). Of the 778 subjects, 80 were identified as carrying either HLA-Cw*0303 or Cw*0304 by high-resolution HLA typing (5). The HIV gag gene was sequenced from DNA from 640 C-clade-infected individuals. Sequencing was performed with primers published and described previously (24). Of the 80 subjects with HLA-Cw*0303/0304 and ELISPOT data available, 62 also had HIV Gag sequences.

YL9 epitope and variant recognition experiments.

The YL9 epitope had previously been defined and shown to be restricted by the HLA-Cw*0303 and -0304 alleles (22). Peptides representing the amino acid changes T303A/I/V were synthesized (purity, 80%) (MBS 396; Advanced Chemtech, Louisville, KY; MGH Peptide Core Facility automated synthesizer). Recognition of the variant peptides compared to the wild type was determined by measurement of the IFN-γ response to serially diluted peptides for each variant in an ELISPOT assay. This was performed using PBMCs isolated from individuals (n = 7) responding to the wild-type YL9, as previously described (15).

Site-directed mutagenesis.

Mutations in the HIV Gag protein at positions T303A, T303V, and T303I were introduced by site-directed mutagenesis in the gag gene into the p83-2 plasmid (Quikchange I; Stratagene, United Kingdom). The whole plasmid DNA was PCR amplified in a mutagenesis reaction with two complementary primers containing the target mutation. The primers used for the mutagenesis reaction were as follows: T303A, 5′-GAGACTATGTAGACCGATTCTATAAAGCTCTAAGAGCCGAG-3′ and 3′-CTCGGCTCTTAGAGCTTTATAGAATCGGTCTACATAGTCTC-5′; T303I, 5′-GAGACTATGTAGACCGATATAAAATTCTAAGAGCCGAGC-3′ and 3′-GCTCGGCTCTTAGAATTTTATAGAATCGGTCTACATAGTCTC-5′; and T303V, 5′-GAGACTATGTAGACCGATTCTATAAAGTTCTAAGAGCCGAG-3′ and 3′-CTCGGCTCTTAGAACTTTATAGAATCGGTCTACATAGTCTC-5′. The introduction of each mutation was verified by sequencing the whole gag gene in newly generated plasmid clones. The DNA fragment ranging from SapI to ApaI was then subcloned into a new p83-2 vector to avoid potential carryover of additional mutations during the mutagenesis, and the coding region sequence was verified again as previously described (26).

Virus production.

Viral stocks were produced by cotransfection in MT4 cells of p83-2 site-directed mutants and p83-10eGFP as previously described (44). Viral stocks were harvested, and viral RNA was extracted (Qiagen, United Kingdom). The gag and pol genes were PCR amplified and sequenced to confirm the presence of the introduced mutations and the absence of other potential variations. The 50% tissue culture infective dose (TCID₅₀) for each viral stock was determined in MT4 cells using the method of Reed and Muench (33a).

Replication kinetics in Jurkat T cells and primary cells.

Jurkat T cells were infected in triplicate to a multiplicity of infection (MOI) of 0.001 in a total volume of 1 ml with wild-type virus or virus variants and incubated at 37°C for 2 h. The pellets were washed twice with phosphate-buffered saline (PBS) and cultured at 37°C and 5% CO₂, as previously described (31). Viral infectivity was determined daily for 15 days using fluorescence-activated cell sorting (FACS) analysis to measure the percentage of enhanced green fluorescent protein (EGFP)-positive cells, as previously described (32). A similar experiment was performed in peripheral blood mononuclear cells (PBMCs) by purifying these cells from a seronegative donor and stimulating them with phytohemagglutinin (PHA) at 5 μg/ml and 0.5 μg/ml for 72 h. After stimulation, the cells were pulled together, and the cellular pellets were infected with each viral variant at an MOI of 0.02 for 2 h at 37°C. The cellular pellets were washed twice with PBS and cultured at 37°C and 5% CO₂. PBMCs were resuspended in R10 medium supplemented with interleukin 2 (IL-2) (Roche, United Kingdom) at 100 U/ml to a final concentration of 1 × 10⁶ cells/ml. Viral spread and infectivity were measured as the percentage of CD4-p24-positive cells by intracellular staining with anti-CD3 and anti-CD4 antibodies (Becton Dickinson) and HIV Gag-p24 phycoerythrin (PE) clone KC57-RD1 (Coulter Clone, United Kingdom). This analysis was performed using FACS. The FACS data were analyzed with Flowjo v8.6.3.

Clonal and phylogenetic analyses.

Detailed analysis of proviral DNA clones was undertaken for two HLA-Cw*0304-positive individuals within the Oxford cohort: the child, SL17C, of a mother-to-child transmission pair (SL17M/SL17C) and an individual (H004) with the T303V variant as a population sequence. The HIV gag gene was PCR amplified and cloned into the Topo vector (Invitrogen, United Kingdom). Approximately 20 to 30 clones were sequenced per time point and individual. The H004 and SL17M/SL17C sequences were edited and aligned with SeqAlv2.0. jModeltest v0.1.1 (30) was used to infer the best phylogenetic model that was able to implement a discrete gamma distribution (Γ), which models the heterogeneity rate among sites. A maximum-likelihood tree was obtained with PhyML v3.0 (17), with the best evolutionary model found in jModeltest. In order to assess the significance of the nodes, 1,000 bootstrap replicates were performed, and only nodes with bootstrap values higher that 70% were taken to be significant. The phylogenetic trees were edited using FigTree v1.1.2 (http://tree.bio.ed.ac.uk/).

Statistical analysis.

A 2-tailed Fisher's exact test was used to identify the associations between alleles and specific polymorphisms. PRISM software was used for the analysis of viral kinetics, CD4, and viral-load data sets. The infectivity rate of each virus was analyzed by fitting a linear model to the log₁₀-transformed data of EGFP expression and comparing the slopes, as previously described (32). The slopes were considered significantly different if P was <0.05. The Mann-Whitney test was used to compare viral loads between groups, and a 2-tailed unpaired t test was use to compare absolute CD4 counts.

Chronically HIV-infected HLA-Cw*0303/0304-positive individuals make an immunodominant Gag-296-304-specific response.

Comprehensive screening using a panel of 410 OLPs spanning the whole HIV proteome showed a high frequency of responses in HLA-Cw*0303/0304-positive subjects against the adjacent Gag peptides OLP-40 (GPKEPFRDYVRFFKTLR) and OLP-41 (YVDRFFKTLRAEQATQDV), with 70 of 80 (87.5%) chronically infected individuals having a detectable response (Fig. 1 A). These OLPs contain the HLA-Cw*0303/0304 optimal restricted epitope YL9, which was previously defined (22). The magnitude of these responses made up >50% of the total Gag responses in almost one-third of Cw*03-positive individuals (26/80) and 100% in 15% of these individuals (Fig. 1B) (this calculation counted responses to OLP-40 or OLP-41, whichever was greater, in order to avoid counting the YL9 response twice). In addition, previous, smaller studies on 24 Cw*0304-positive subjects of the same cohort (12), using the Cw*0304-YL9 tetramer, showed detectable responses to this epitope, with a median response frequency of 2.26% of CD8⁺ T cells, the highest of 8 immunodominant responses that were examined. Furthermore, although it is possible that responses to additional optimal epitopes may have contributed to the responses to the overlapping 18-mer peptides OLP-40/41, in 7 Cw*0304-positive subjects selected on the basis of cell availability for further study, all subjects having responses to OLP-40 or -41 also had a detectable response to the YL9 optimal restricted epitope. Additionally, the magnitude of the YL9 response correlated strongly with that of the response to the OLP and was statistically significant in spite of the small numbers (n = 7) of subjects evaluated (for example, for YL9 and OLP-41, r² was 0.67 and P was 0.02 [data not shown]). No other HLA-Cw*0303- or Cw*0304-restricted HIV-specific responses have been defined to date. Based on these data, the HLA-Cw*0303/04-restricted CTL responses directed to YL9 in Gag are immunodominant, both in terms of the frequency with which the epitope is targeted in Cw*03-positive individuals and also in the magnitude of the YL9-specific response compared to responses to other epitopes in Gag.

HLA-Cw*0303/0304 selects variants in Gag-303 that reduce YL9 immune recognition.

In order to determine whether the HLA-Cw*0303/0304-YL9 response imposes selection pressure on the virus, we analyzed RNA and DNA gag sequences from 640 clade C-infected study subjects. In the region of the YL9 epitope, we identified a strong association between HLA-Cw*0303/0304 and the polymorphisms Val/Ile/Ala/Cys at Gag residue 303 (P = 1.62 × 10⁻¹⁰) (Table 1). Viruses from 45% (28/62) of the subjects carried one of these changes compared to 11% (61/578) of Cw*0303/0304-negative subjects. The most common variants in Cw*03-positive subjects in place of wild-type Thr-303 were valine (39% of variants), alanine (29%), and isoleucine (25%), with cysteine (7%) the least common (T303C was not studied in further detail, since it was very rare in the Cw*0303/0304 population) (Table 2).

To examine the impact of the Cw*0303/04-associated Gag-303 variants on YL9 recognition, synthetic YL9 peptides containing the T303V, T303A, and T303I mutations were synthesized, and their recognition by CD8⁺ T cells was compared to that of the wild-type epitope in an IFN-γ ELISPOT assay of PBMCs from seven Cw*0303/04 individuals based on sample availability (1). The T303A and T303I mutants were less well recognized than the wild-type YL9 epitope (median, 7 study subjects; 50% saturating doses (SD₅₀), 243 nM, 235 nM, and 47 nM, respectively) (Fig. 2 B), but only a marginal effect on the SD₅₀ was found with the individually more common T303V mutation (median SD₅₀, 75 nM) (Fig. 2A). These data suggest that immune-mediated selective pressure on HIV is driven by HLA-Cw*0303/0304 through selection of T303V/A/I/C variants at position 8 of the YL9 epitope. The escape variants T303V/A/I at position 303 reduce, but do not abrogate, YL9 recognition.

In vitro viral replicative capacities of common YL9 T303 variants.

It is known that HIV immune escape is limited by structural constraints (16). In addition, particular CTL escape mutations in HIV Gag-p24 have been associated with a reduction in the viral RC (4, 11, 26, 32). Structurally, the YL9 Gag responses target the MHR of the Gag p24 protein, a region of the HIV proteome that is highly conserved between retroviruses (25). The MHR is located within helix 8 of the C-terminal domain of p24 (see Fig. S1 in the supplemental material) and has been implicated in virus particle assembly (8). To examine the effect on virus RC of HIV escape variants selected by an HLA-C-restricted response, mutations at residues T303V/I/A in YL9 were introduced into the p83-2 vector (NL43 backbone), and the virus RC was measured in Jurkat T-cell and primary cultures, as previously described (32). Viruses with the YL9 escape mutations T303A and T303I were demonstrated to have substantially decreased infectivity rates and impaired kinetics compared to wild-type virus (Fig. 3 A and B). Furthermore, introduction of T303V was found to increase the virus RC to levels similar to that of wild-type virus. Comparable results were obtained in replication kinetics experiments in PBMCs measured by both GFP⁺ and intracellular p24 staining (Fig. 3C and D). In addition, all viruses were sequenced at the end of the replication kinetics experiments. No reversion to wild type or additional compensatory changes were observed in the Gag coding region for any of the T303 variants. Overall, these results indeed demonstrate that Cw*03-restricted YL9 responses induce changes in HIV RC through viral immune escape. However, the most commonly observed variant, T303V, resulted in minimal loss of CD8⁺ T-cell recognition, with no detectable reduction in the viral RC.

Balance between YL9 immune escape and viral replicative capacity drives two-step nucleotide substitution concomitant with a decrease in absolute CD4 counts and an increase in the viral load.

T303V is the most common variant in the population. However, T303V requires 2 nucleotide substitutions from the T303 wild type compared to the single substitution needed for T303A or T303I selection (Fig. 4 A). Therefore, selection of T303V involves the previous selection of the immune escape variant T303A or T303I. To examine in more detail the selection of these distinct YL9 escape mutants, the gag gene was PCR amplified from a Cw*0304-positive infected individual for whom T303V had previously been determined to be the predominant (population) sequence, and proviral DNA clonal analyses were carried out. Phylogenetic analysis showed the coexistence of viral populations encoding the T303A/V viral variants, together with the wild type, in the same subject (Fig. 4B). This result suggests the presence of the intermediate form, T303A, en route to T303V, but not the outgrowth of T303A in the viral population.

These data suggest that, for most subjects with Cw*0303/04, a balance exists between the selection of the T303I/A variants poorly recognized by CTL but significantly reducing viral replicative capacity and the selection of T303V, where viral escape is modest but the reduction in viral replicative capacity is also minimal. Comparison of absolute CD4 counts and viral loads in Cw*03-positive individuals (Fig. 4C and D) suggested that selection of T303V brings about the optimal result for the virus: Meanwhile the T303I/A mutations overall do not have substantial impacts on the viral load or CD4 counts (median viral load, 19,900 RNA copies/ml; median CD4 count, 373 cells/μl) compared with those individuals with wild-type virus (median viral load, 25,900 copies/ml; median CD4 count, 433 cells/μl); Cw*03-positive subjects carrying the T303V mutants had a median viral load of 96,550 HIV RNA copies/ml, almost 5-fold higher than those individuals who had virus with the wild-type or T303I/A mutants (P = 0.096; Mann-Whitney U test) and half of the absolute CD4 numbers (203 cells/μl; P = 0.023; 2-tailed unpaired t test). There was also a statistically significantly higher absolute CD4 count in the subjects whose autologous virus expressed the T303I/A variants than in those with T303V (absolute CD4 counts, 373/μl versus 203/μl; P = 0.023; 2-tailed unpaired t test) (Fig. 4C). These data suggest either that Gag-303-V gives the virus the optimal result, increasing in vivo replication with unaffected virus RC and adequately reduced CTL recognition, or that the Gag-303-V variant may be selected in the setting of high viremia and a weak immune response against the virus at a time when selection pressure is not strong enough to select for escape mutants that reduce viral replicative capacity.

These findings are supported by data from an HIV-infected mother-child pair in which the expression of HLA-Cw*0304 in the infant selected the Val at position 303 in an 11-month study period (Fig. 5 A). A maximum-likelihood phylogenetic tree for the mother-child pair demonstrated in vivo virus evolution toward T303V in the absence of minority variants, as opposed to the presence of T303A or T303I selected by the continuous and stronger immune pressure present in adults (Fig. 5B).