Cross-Clade Detection of HIV-1—Specific Cytotoxic T Lymphocytes Does Not Reflect Cross-Clade Antiviral Activity

Abstract

The genetic divergence of human immunodeficiency virus (HIV)—1 into distinct clades is a serious consideration for cytotoxic T lymphocyte (CTL)—based vaccine development. Demonstrations that CTLs can cross-recognize epitope sequences from different clades has been proposed as offering hope for a single vaccine. Cross-clade CTL data, however, have been generated by assessing recognition of exogenous peptides. The present study compares HIV-1—specific CTL cross-clade epitope recognition of exogenously loaded peptides with suppression of HIV-1—infected cells. Despite apparently broad cross-clade reactivity of CTLs against the former, CTL suppression of HIV-1 strains with corresponding epitope sequences is significantly impaired. The functional avidity of CTLs for nonautologous clade epitope sequences is diminished, suggesting that CTLs can fail to recognize levels of infected endogenously derived cell-surface epitopes despite recognizing supraphysiologic exogenously added epitopes. These data strongly support clade-specific antiviral activity of CTLs and call into question the validity of standard methods for assessing cross-clade CTL activity or CTL antiviral activity in general.

The divergence of HIV-1 into multiple clades worldwide poses a great challenge for vaccine development. The spread of the epidemic has resulted in geographic segregation of genetic sequences into distinct clades that can diverge by 30% [1]. The vast majority of HIV-1 isolates found in the United States and Europe are clade B strains, which are a small minority of the strains in Africa, the origin and epicenter of the HIV-1 epidemic. Although a growing body of work has been performed in populations infected with non—clade B strains, studies of HIV-1 pathogenesis have been disproportionately focused on clade B.

The CD8⁺ T lymphocyte response against HIV-1 appears to be responsible for partially controlling infection during the chronic stable phase of infection and has thus become a focus for many vaccine-development efforts [2, 3]. Cytotoxic T lymphocytes (CTLs) can recognize and kill HIV-1—infected cells, suppressing viral replication by the specific recognition of viral peptides (epitopes) presented by HLA class I molecules [4, 5]. However, even single amino acid changes in epitopes can ablate this function, resulting in the escape of HIV-1 from CTLs [6]. Evolution of HIV-1 sequences to allow escape is believed to play a role in the inability of CTLs to clear or fully contain infection [3, 6]; variability in epitope sequences is therefore clearly a crucial factor in their antiviral activity.

Given the genetic divergence of HIV-1 clades, CTL epitope sequences can contain clade-specific polymorphisms [1]. This raises the issue of whether CTL-based HIV-1 vaccine development using sequences from 1 clade would have utility in regions in which other clades are found. Even assuming that a vaccine might elicit effective CTL responses capable of coping with intraclade genetic diversity, it is unclear whether any vaccine based on a single fixed sequence could yield CTL responses that would include multiple HIV-1 clades.

However, the specificity of CTLs for their targeted epitopes is not absolute but rather shows some promiscuity, and CTLs can usually recognize multiple epitope variants [6–8]. This property of cross-recognition has been extensively evaluated in the context of CTL epitope variation across different HIV-1 clades [9–30], and most investigators have concluded that cross-clade CTL activity is common. A critical implicit underlying assumption common to all of these studies has been that the detection of CTL activity (target cell killing or cytokine release) using exogenous synthetic peptides predicts activity against HIV-1—infected cells endogenously presenting the same peptides. Whether such cross-reactivity defined in routine assays extends to actual HIV-1—infected cells remains poorly understood.

Materials and Methods

HIV-1 permissive target cells. The cell lines Jurkat (expressing HLA B*07) and T1 (expressing HLA A*02) were maintained in RPMI 1640 medium supplemented with l-glutamine, penicillin-streptomycin, and 10% heat-inactivated fetal calf serum, as described elsewhere [4].

HIV-1—specific CTL clones. CTL clones were isolated by limiting dilution cloning of peripheral blood mononuclear cells (PBMCs) from HIV-1—infected persons in the Los Angeles area (under an Institutional Review Board—approved informed-consent protocol) and were maintained with periodic restimulations using irradiated allogeneic feeder PBMCs and anti-CD3 antibody, as described elsewhere [4–6]. Clone S42758-RL10— 3.22 recognized the B*07-restricted epitope RPAEPVPLQL in Rev (RL10; HXB2 aa 66–75), and clones S1-SL9—3.23T and S36-SL9—10.18T recognized the A*02-restricted epitope SLYNTVATL in Gag p17 (SL9; aa 77–85) [31]. The latter 2 clones were telomerase transduced, as described elsewhere [32].

Chromium release assays. The CTL clones were screened for killing of Jurkat and T1 cells in standard chromium-release assays, as described elsewhere [4]. Briefly, the target cells were labeled with chromium 51 (⁵¹Cr) in the presence or absence of tested peptides and plated at 1 × 10⁴ cells/well in a 96-well U-bottom plate. CTL clones were added at 5 × 10⁴ cells/well for a 4-h incubation. Supernatant was harvested into LumaPlates (Packard) for quantitation of ⁵¹Cr release with a microscintillation counter (Wallac MicroBeta). Controls included target cells without CTLs (spontaneous) and target cells with 2.5% Triton (maximal). Specific lysis was calculated as follows: specific lysis = (observed chromium release − spontaneous chromium release)/(maximal chromium release − spontaneous chromium release).

HIV-1 mutagenesis and stocks. Virus stocks were generated from plasmid DNA as described elsewhere [6, 33]. The NL4-3 sequence served as the backbone for the HIV-1 mutants produced for this study, with a single amino acid mutation in Nef to ablate HLA class I down-regulatory function and avoid this confounding effect on CTL activity. Consensus sequences were obtained from the Los Alamos National Laboratory HIV Sequence Database (http://www.hiv.lanl.gov/content/hiv-db). Production of the clade B consensus SL9 sequence SLYNTVATL using overlapping polymerase chain reaction mutagenesis was as reported elsewhere [6]. The SLFNTVATL variant was produced using the QuikChange mutagenesis kit (Stratagene) by changing the clade B consensus sequence from TCATTATATAATACAGTAGCAACCCTC to TCATTATTTAATACAGTAGCAACCCTC. The native NL4-3 RL10 sequence RSAEPVPLQL already corresponded to the clade A1 consensus sequence, with the nucleotide sequence CGATCTGCGGAGCCTGTGCCTCTTCAGCTA. The QuikChange kit was used to create the clade B consensus sequence RPAEPVPLQL by mutagenesis to CGACCTGCGGAGCCTGTGCCTCTTCAGCTA; the variant RPTEPVPLQL by mutagenesis to CGACCTACGGAGCCTGTGCCTCTTCAGCTA; and the variant RPEEPVPLQL by mutagenesis to CGACCTGAGGAGCCTGTGCCTCTTCAGCTA.

HIV-1 inhibition assays. The ability of the CTL clones to suppress HIV-1 replication was tested in coculture assays, as described elsewhere [5, 6, 31]. Briefly, Jurkat or T1 cells were infected with the appropriate variant of NL4-3 at an MOI of 0.01 tissue culture infectious dose/cell. These cells were then plated in a 96-well U-bottom plate at 5 × 10⁴ cells/well, for triplicate coculture with and without 1.25 × 10⁴ CTLs. Viral replication was monitored by quantitative p24 antigen ELISA.

Functional avidity assays. The “functional avidity” of CTL clones for epitopes, defined as the concentration of exogenously added epitope required for 50% maximal killing of target cells (HIV-1 uninfected) in the standard chromium-release assays described above, was defined as described elsewhere [4, 34]. In brief, the target cells were incubated with serial dilutions of peptide (Sigma) during the 4-h chromium-release assay. The observed specific lysis was plotted against the peptide concentration to estimate the concentration of peptide yielding 50% of the maximal level of specific lysis (sensitizing dose for 50% maximal killing; see below).

Results

Cross-recognition of epitope variants from multiple clades by HIV-1—specific CTL clones. To test HIV-1—specific CTLs for cross-clade epitope recognition (table 1) by conventional methods, 3 HIV-1—specific CTL clones derived from individuals infected in the United States were screened by chromium-release assay using exogenously peptide-labeled target cells (10 μg/mL). Clone S42758-RL10—3.22 targeting a B*07-restricted epitope in Rev (RL10; aa 66–75) was tested for recognition of target cells loaded with epitopes corresponding to the Los Alamos HIV database consensus sequences for clades B/C/G (RPAEPVPLQL), A1 (RSAEPVPLQL), A2/F2 (RPTEPVPLQL), and F1 (RPEEPVPLQL), and clones S1-SL9—3.23T and S36-SL9—10.18T targeting an A*02-restricted epitope in Gag (SL9; aa 77–85) were tested against epitopes corresponding to consensus sequences for clades B/A1/C/D (SLYNTVATL) and G/K (SLFNTVATL). By this assay, all of the clones showed high levels of cross-clade reactivity (figure 1). The RL10-specific clone recognized all variants similarly (figure 1A), suggesting cross-clade reactivity of the CTLs derived from a clade B—infected person against clades A1, A2, C, F1, F2, and G. The 2 SL9-specific clones also recognized the screened epitope sequences similarly (figure 1B), again suggesting cross-clade reactivity, with clade B—elicited CTLs recognizing epitope sequences from clades A1, C, D, G, and K. These data agreed with past findings of cross-clade recognition using exogenous peptides, as determined by chromium-release or cytokine-detection assays.

Figure 1

Cross-clade cross-reactivity against exogenous peptides exhibited by HIV-1—specific cytotoxic T lymphocyte (CTL) clones. Uninfected cells were loaded with exogenously added epitopes (10 μg/mL) as target cells for CTL clones. A, Jurkat cells loaded with the B*07-restricted Rev RPAEPVPLQL epitope or the indicated variants for recognition by the CTL clone S42758-RL10—3.22. The results are representative of 3 independent experiments. B, T1 cells loaded with the A*02-restricted Gag SLYNTVATL epitope or the indicated variant for recognition by the CTL clone S1-SL9—3.23T. The results are representative of 4 independent experiments. C, T1 cells loaded with the A*02-restricted Gag SLYNTVATL epitope or the indicated variant for recognition by the CTL clone S36-SL9—10.18T. The results are representative of 3 independent experiments.

Epitope variation between clades and the ability of CTLs to suppress HIV-1 replication. To evaluate whether detection of cross-clade reactivity with exogenously added epitopes predicts efficient CTL antiviral activity against HIV-1 strains with the corresponding epitope sequences, the CTL clones were tested for their ability to inhibit the replication of HIV-1 containing the same epitope variants. The HIV-1 strain NL4-3 was point mutated in RL10 or SL9 to create epitope variants corresponding to consensus sequences for other clades (table 1), and these viruses were tested for susceptibility to the CTL clones. Clone S42758-RL10-3.22 potently suppressed HIV-1 with the RL10 sequence corresponding to the consensus for clade B and clades C/G (figure 2A; ∼1000-fold). By comparison, the clone had relatively poor antiviral activity against HIV-1 with the 3 other epitope sequences representing consensus sequences for clades A1, A2/F2, and F1 (figure 2B–2D). Clones S1-SL9-3.23T and S36-SL9-10.18T also suppressed HIV-1 containing the SL9 epitope sequence corresponding to the consensus for clade B and clades A1/C/D (figures 3A). The antiviral activities of these clones were also somewhat reduced by altering the epitope sequence to the consensus sequence for clades G/K (figure 3B and 3D). The ability of the CTLs to kill HIV-1—infected cells in conventional chromium-release assays correlated highly with virus suppression, as reported earlier [5]. Overall, these results showed that variation of epitope sequences across clades can markedly reduce the ability of HIV-1—specific CTLs to recognize infected cells and suppress viral replication.

Figure 2

Considerable variation in the ability of cytotoxic T lymphocytes (CTLs) to suppress HIV-1 with RL10 epitope sequences from different clades. Jurkat cells were acutely infected with HIV-1 containing the indicated variants of the RL10 epitope and screened for replication in the absence and presence of the CTL clone S42758-RL10—3.22. The plotted data indicate means of triplicates (error bars indicate 1 SD) for 1 experiment, and the results are representative of 3 independent experiments.

Figure 3

Variation in the ability of cytotoxic T lymphocytes (CTLs) to suppress HIV-1 with SL9 epitope sequences from different clades. T1 cells were acutely infected with HIV-1 containing the indicated variants of the SL9 epitope and were screened for replication in the absence and presence of the CTL clones S1-SL9—3.23T (A and B) or S36-SL9—10.18T (C and D). The plotted data indicate means of triplicates (error bars indicate 1 SD), and the results are representative of 2 independent experiments for each clone.

Table 1

RL10 and SL9 epitope sequence variation between HIV-1 clades.

Lack of correlation between target cell recognition using exogenous peptides and antiviral activity. To compare the ability of the CTL clones to kill exogenously peptide-loaded target cells with the ability to suppress HIV-1 replication, the results of the chromium-release killing assays (figure 1) and virus-suppression assays (figures 2 and 3) were each normalized to activity against the clade B consensus sequence. This comparison for clone S42758-RL10—3.22 (figures 4A) revealed that while killing of exogenously peptide-loaded target cells was equivalent across clades (ratio near 1.0), the inhibitory activity of viruses with epitope sequences from the other clades was poor compared with clade B (ratio <0.2). For the clones S1-SL9—3.23T and S36-SL9—10.18T (figure 4B and 4C), the killing of peptide-loaded target cells also was similar, but the reduction in killing with cross-clade epitope variation was more modest. Overall, this analysis highlighted the lack of correlation between CTL detection using exogenous peptides and CTL antiviral capacity.

Figure 4

Equivalent antiviral activity not necessarily indicated by equivalent target cell recognition using exogenous peptides. Specific lysis data (as shown in figure 1) were normalized as a ratio of killing versus the clade B consensus sequence, and viral inhibition data (as shown in figures 2 and 3) were normalized as a ratio of log10 units of virus suppression versus the clade B consensus sequence. A, Means and SDs from 3 independent lysis experiments and 3 independent inhibition experiments, plotted for cytotoxic T lymphocyte (CTL) clone S42758-RL10—3.22. B, Means and SDs from 4 independent lysis experiments and 2 independent inhibition experiments, plotted for CTL clone S1-SL9—3.23T. C, Means and SDs from 3 independent lysis experiments and 2 independent inhibition experiments, plotted for CTL clone S36-SL9—10.18T.

Diminished CTL functional avidity for epitope variants from other clades. To assess whether reduced CTL functional avidity for epitope variants could play a role in the phenomenon of reduced antiviral activity, the CTL clones were screened using peptide titration assays against the epitope variants for the different clade consensus sequences (figure 5). Clone S42758-RL10—3.22 had the highest functional avidity for the RL10 sequence corresponding to the consensus for clade B and clades C/G, but it had reduced avidity (by ∼1–2 log₁₀ U) for the other 3 epitope variants corresponding to the consensus sequences for clades A1, A2/F2, and F1. Similarly, the clones S1-SL9—3.23T and S36-SL9—10.18T recognized the clade B and clades A1/C/D consensus sequence epitope with greater avidity than did the clade G/K epitope sequence (by ∼0.5–1 log₁₀ U). These findings suggested that CTLs recognize their epitope sequences from the infecting clade more avidly than they recognize versions from other clades and that the reduced avidity for other clades can result in variably reduced antiviral activity for HIV-1, in agreement with prior observations [5].

Figure 5

Variation in HIV-1—specific cytotoxic T lymphocyte (CTL) functional avidity for epitope sequences from different clades. Each CTL clone was assessed for specific lysis of uninfected target cells over serial 4-fold dilutions of exogenously added epitope for the indicated epitope variants. The results are representative of 3 or more independent measurements. The sensitizing dose for 50% maximal killing (SD₅₀) values of clone S42758-RL10—3.22 for the indicated variants (RPAEPVPLQL, -S--------, --T-------, and --E-------) were 4.41 ± 0.16, 6.66 ± 0.33, 5.10 ± 0.31, and 5.19 ± 0.32 log₁₀ pg/mL, respectively (means ± SDs) (A). The SD₅₀ values of clone S1-SL9—3.23T for the indicated variants (SLYNTVATL and --F------) were 3.09 ± 0.59 and 3.56 ± 0.33 log₁₀ pg/mL, respectively (B). The SD₅₀ values of clone S36-SL9—10.18T for the indicated variants (SLYNTVATL and --F------) were 3.32 ± 0.48 and 3.53 ± 0.35 log₁₀ pg/mL, respectively (C).

Discussion

The importance of tailoring potential HIV-1 vaccines to match the diversity of the virus remains unknown. Some successful vaccines—for example, the Streptococcus pneumoniae vaccine [35]—have dealt with the issue of pathogen divergence by having a multivalent design that represents the majority of circulating subtypes of the pathogen. This is possible because these pathogens have evolved into relatively fixed subtypes. In the case of the more variable influenza virus, the combination of evolution and intraspecies genetic recombination presents diversity too great for this approach; thus, the strategy has been to create custom vaccines each year to match the prevalent circulating strain during the annually periodic influenza season. However, the genetic diversity of HIV-1 dwarfs that of influenza virus [36], and numerous HIV-1 clades continue to spread and evolve simultaneously worldwide.

An effort to address diversity has been to assess the degree to which CTLs might cross-react against epitopes across different clades of HIV-1 [9–30]. Most of these studies have used detection of CTL cytokine release or killing by exogenous peptide loading as a presumptive surrogate for cross-reactivity against viruses from other clades. Generally, most investigators have concluded from these assays that the degree of cross-clade activity is high, although many of these studies have demonstrated the highest CTL activity against autologous clades. Such data have formed the basis for proposing testing of vaccines with HIV-1 sequences from a fixed clade in geographic regions where other clades predominate, for example, testing of a clade B—based recombinant canarypox vaccine in Uganda, where almost all HIV-1 strains are clades A and D [37].

Our data provide a cautionary perspective on interpreting data on the cross-reactivity of CTLs across different HIV-1 clades. Using a chromium-release assay for cytolysis of exogenously peptide-loaded cells (which is less sensitive than cytokine-detection methods, such as the interferon-γ enzyme-linked immunospot [ELISpot] assay), we obtained results that agreed with findings from other studies concerning high crossclade cross-reactivity of HIV-1—specific CTL responses. CTL clones from persons infected in a clade B—endemic area were generally capable of recognizing epitope variants representing the consensus sequences of other clades. Strikingly, when the same CTL clones were tested for their ability to suppress HIV-1 replication of strains containing the same epitope variants, the results were different; epitope variation variably ablated the antiviral capability of the CTLs. Some epitope variants that were well recognized as exogenously loaded peptides were virtually unrecognizable in the context of HIV-1 suppression. Although the experiments are limited to 3 HIV-1—specific CTL clones, the trends are strongly suggestive, and these data indicate that CTLs elicited by an epitope in one clade may lose antiviral activity against HIV-1 from another clade with epitope variation, despite detection of “cross-clade responses” by standard assays using exogenous peptide loading.

The likely mechanism of this phenomenon relates to the functional avidity of CTLs for their target epitope. CTLs raised against a particular epitope sequence from clade B (presumably in patients from a clade B—endemic area) can have reduced avidity for the corresponding epitope sequence from another clade. Although excess loading of target cells with exogenous epitope can overcome this reduction in avidity and allow detection of CTL activity, the reduction is not sufficient to allow recognition of the amount of epitope presented on an HIV-1—infected cell. Recent work [38] has demonstrated the thresholds of avidity required for recognizing cells via the epitopes in the present study .

The discrepancy between assays is likely an artifact of methodology. Excess concentrations of exogenous peptides are used to ensure detection of CTLs in standard assays such as the ELISpot assay [2], because the functional avidity of CTLs can vary over several orders of magnitude [34]. Thus, this approach is slanted to include detection of low-avidity CTLs. However, clade variation in the recognized epitope can result in loss of avidity that is not detected under conditions of excess peptide. Reduced avidity may render the CTLs incapable of recognizing the amount of epitope endogenously produced by an infected cell, with loss of antiviral activity under physiological conditions of epitope presentation.

Our experimental approach to examining virus suppression may still underestimate the loss of antiviral activity due to clade variation. Non—clade B consensus epitope sequences were evaluated in the context of a clade B molecular clone virus, allowing controlled comparisons of the impact of epitope variation on T cell receptor recognition by minimizing other differences. However, clade-specific polymorphisms in sequences flanking the epitope could interfere with epitope processing and further reduce recognition. Again, bypassing endogenous epitope production by using exogenous peptides results in failure to detect such differences. Another factor that we omitted with our experimental approach is the impact of Nef-mediated down-regulation of the HLA class I molecule, which might be expected to accentuate further the differences in recognition of different epitope variants by limiting the availability of presentation. We chose to perform our experiments with Nef that does not down-regulate HLA class I to avoid this factor that could exaggerate our findings. Furthermore, we have noted that this function of Nef is highly variable between different patient sequences (data not shown).

Overall, our data suggest that CTL responses can be highly clade specific. Most of the epitope variants we examined did not differ from the clade B epitope at major anchor residues for HLA class I binding. Thus, these epitopes conceivably could generate CTL responses that show the analogous phenomenon of clade specificity for other non-B clades that we observe here for clade B, but further work will be required to explore that possibility. Although our data argue against implementing clade B—based vaccination in areas in which other clades are endemic, a demonstration of CTL specificity for other clades would similarly argue against vaccination for other clades in areas of clade B endemicity.

Finally, our data suggest other worrisome implications beyond the interpretation of cross-clade immunity for vaccine studies. The same principle applies broadly to any case of epitope variation, such as mutational escape during natural infection or mismatch between vaccine and actual infecting HIV-1 sequences. The detection of a CTL “response” to a particular epitope sequence by exogenous-peptide assays, such as the ELISpot assay or intracellular cytokine staining, does not ensure that the detected CTLs can recognize the epitope in the context of an HIV-1—infected cell. Furthermore, this principle can apply to vaccines even in the absence of epitope-sequence variation. A vaccine could generate CTL responses that can recognize epitopes in exogenous-peptide assays but lack sufficient avidity to recognize the same epitopes presented by infected cells (at lower concentrations). Maximizing HIV-1 protein expression by vaccine vectors (through promoter selection and codon optimization) has been the norm in vaccine design, but this could theoretically set the conditions for preferential expansion of lowavidity CTLs that cannot recognize physiological levels of epitopes. Notably, the ELISpot assay has been the only validated test of immunogenicity for ongoing phase 2 and 3 human HIV-1 vaccine trials, and it would not detect this shortcoming in vaccine-induced CTLs.

In conclusion, these data underscore the importance of having a functional understanding of standard CTL measurements and their relationships to antiviral activity. The ability to trigger CTL activity against a particular epitope sequence in such assays does not equate to efficient CTL recognition of cells infected with HIV-1 containing that sequence. It is likely that prior studies of cross-clade CTL reactivity have markedly underestimated the impact of epitope variation between clades on the ability of CTLs to recognize and clear HIV-1, and our findings have potentially significant implications for the evaluation of antiviral immunity in pathogenesis and vaccine studies.

Potential conflicts of interest: none reported.

Presented in part: First Annual Ugandan AIDS Conference, Kampala, Uganda, 7 December 2006.

Financial support: Public Health Service (grant AI043203).