Impact of Intrinsic Cooperative Thermodynamics of Peptide-MHC Complexes on Antiviral Activity of HIV-Specific CTL¹

Human CD8⁺ CTLs recognize HIV-infected cells by interaction of their own TCRs with viral peptides bound to HLA class I molecules on the cell surface of the infected cells and eliminate them directly by cytolysis or indirectly through the production of soluble factors such as cytokines and chemokines. Among these activities, the cytotoxic activity of CTLs toward HIV-infected cells is associated with efficient viral containment in vitro and in vivo (1, 2, 3). However, significant differences exist not only in the antiviral activity of HIV-specific CTLs among specificities (4, 5, 6, 7) but also in CTL specificities between early and chronic phases of an HIV infection (8, 9, 10). Changes in CTL specificity could lead to the accumulation of less effective antiviral CTLs in the late chronic phase of an infection (6, 11, 12). There are a number of different possibilities in the literature that potentially explain the heterogeneity in the antiviral activity of CTLs, such as: differences in functional avidity of CTLs toward exogenously pulsed synthetic peptides (7, 13), TCR usage (14, 15), cross-reactive capacity of CTLs toward variant Ags (14, 16), kinetics and amplitude of immunogenic protein expression (9, 17, 18, 19), Ag processing and presentation pathways (20, 21), and binding activity of an antigenic peptide to a given HLA class I molecule (22). However, considering that immunodominant peptides are not always those with the highest density presented at the target cell surface (23, 24) and that immunodominant CTLs are not always correlated with effective antiviral CTL responses (25), an interesting question can be raised as to whether, and if so what, inherent characteristics of target epitope peptides support the efficient recognition by CTLs for the killing of virus-infected cells. As mentioned above, however, the antiviral activity of CTLs stems from multifactorial events, reflecting the consequence of various positive and negative factors that govern viral replication, Ag presentation, and T cell activation (26). Broad comparisons between very different virus strains, peptide Ags, and MHCs provide little information beyond highlighting just the differences. Comparisons between more closely related viral Ags and MHCs could be more revealing.

We previously reported that CD8 T cells specific for an Nef epitope (VY8, VPLRPMTY) were consistently elicited very early in vivo, whereas those specific for another Nef epitope (RY11, RPQVPLRPMTY) were mostly observed in the chronic phase of an HIV infection (10). Remarkably, VY8 is entirely contained within RY11; and both are presented by HLA-B35 with comparable binding activity, as assessed by a cellular HLA stabilization assay (10). As initial preliminary experiments showed that VY8-specific CTLs had more potent cytotoxic activity toward HIV-infected primary CD4⁺ cells than RY11-specific CTLs, in the present study we asked what property of these antigenic peptides is correlated with CTLs having potent antiviral cytotoxic activity. Combining a series of data obtained from T cell lines transduced with the genes for the cognate TCRs, we discovered that the decay of peptide-MHC class I complex (pMHC),³ rather than the functional avidity of TCR-pMHC interactions, substantially influenced the susceptibility of HIV-infected cells to CTL killing. Furthermore, by using a biochemical approach, we found that the peptide intrinsic cooperative thermodynamics of pMHC could be an important factor to support efficient antiviral cytotoxic activity of CTLs.

CTL clones were established as previously described (6, 15) by using PBMC samples taken from HLA-B*3501⁺ individuals (Pt-01, Pt-03, Pt-19, and Pt-33) in the chronic phase of an HIV-1 infection. Briefly, a bulk CTL culture, which had been established by stimulation of PBMC with a synthetic peptide for 1–2 wk, was further seeded at a density of 0.8 or 5 cells/well with a cloning mixture (irradiated allogeneic PBMC and C1R-B3501 cells pulsed with 1 μM peptide in RPMI 1640 with 10% FCS and 100 U/ml recombinant IL-2). Two weeks later, cells showing substantial Ag-specific cytolytic activity were maintained in the medium with peptide stimulation weekly. CTL clone 139 generated from PBMC of Pt-19 was designated as CTL 19-139, and other clones were similarly designated. TCR-encoding genes of CTL clones were obtained by using a SMART PCR cDNA synthesis kit (BD Clontech) and analyzed by the ImMunoGeneTics database (http://imgt.cines.fr) as previously described (27, 28). The study was conducted in accordance with the human experimentation guidelines of Kumamoto University.

The cDNAs encoding full-length TCRα and TCRβ of interest were separately cloned into a retrovirus vector pMX provided by T. Kitamura (Tokyo University, Tokyo, Japan) and delivered into a TCR-deficient mouse T cell hybridoma cell line TG40 provided by T. Saito (RIKEN Institute, Saitama, Japan) as previously described (27, 28). The human CD8α gene was similarly delivered into the cells as needed (28). Finally the cells showing bright staining with PE-conjugated anti-mouse CD3ε mAb (2C11; BD Pharmingen) were cloned by a limiting dilution method for further functional assays described below.

The HLA-B*3501 tetramers in complex with the VY8 or RY11 peptides were prepared as previously described (28). The CTL clones were stained with PE- and allophycocyanin-labeled HLA-B35 tetramers at 37°C for 15 min followed by anti-CD8-PerCP (BD Biosciences) and anti-CD3-FITC (DakoCytomation) at 4°C for 15 min. For the kinetic analysis of HLA-B35 tetramer dissociation, CTL clones were stained with PE-conjugated tetramer (0.2 μM) for 30 min at 4°C. Then the cells were rapidly washed twice and suspended at 4°C in a buffer (2% BSA in PBS) supplemented with the monomeric type of unconjugated peptide-HLA complex (2 μM) for blocking. A portion of the reaction volume was then removed periodically (0.5, 1, 2, 4, and 8 h), and the cells were subsequently stained with anti-CD8 and anti-CD3 mAbs at 4°C. For the flow cytometric analysis, the CD3⁺CD8⁺ cells were gated and then analyzed for the tetramer binding by flow cytometry with FACSCalibur (BD Biosciences).

Primary CD4⁺ cells were purified from PBMC taken freshly from HIV-negative donors expressing HLA-B*3501 by using a magnetic cell separation system (Miltenyi Biotec) and stimulated with PHA (3 μg/ml; Sigma-Aldrich) for 4 days. After having been labeled with ⁵¹Cr, the activated CD4⁺ cells were pulsed with various concentrations of a synthetic peptide for 1 h at 37°C, washed once with culture medium, and then mixed with CTL clones (4000 cells/well) for 4 h at 37°C. For virus-infected target cells, the activated CD4⁺ cells (4000 cells/well), which had been infected with recombinant HIV-1 or vaccinia virus carrying the nef gene of strain SF2 (10), were mixed with CTL clones at various E:T ratios for 6 h at 37°C after having been labeled with ⁵¹Cr. It should be noted that 30 ± 5% of the cells were p24 Ag-positive, as revealed by intracellular flow cytometric analysis of HIV-infected CD4⁺ cells.

TCR recognition of cognate Ags was measured in terms of IL-2 secretion by TG40 cells transduced with TCR and CD8α as described earlier (27, 28). Unless otherwise specified, C1R-B3501 cells (10⁴ cells/well), TCR-transduced TG40 cells (2 × 10⁴ cells/well), and peptides were mixed and incubated for 48 h at 37°C. The resultant culture supernatant was then collected, and the amount of IL-2 was determined by analyzing the proliferative activity of CTLL-2, an IL-2 indicator cell line. The EC₅₀ value of the peptide was calculated as the concentration of peptide that exhibited a half-maximal activation of TCR-transduced TG40 cells with CD3ε mAb-mediated activation of the cells defined as maximal.

For the kinetic analysis of the peptide dissociation from pMHC, C1R-B3501 cells were first incubated with 100 μM peptide at 37°C for 1 h. Then the cells were rapidly washed twice and suspended at 37°C in culture medium. A portion of the peptide-loaded target cells was then removed periodically (10, 20, 30, 60, 120, 240, 360, 720 min), washed once with culture medium, and subsequently mixed with TCR-transduced TG40 cells. The amount of IL-2 produced by the TG40 cells was then determined as described.

The extracellular domain of HLA-B*3501 H chain (aa residues 1–276) and β₂-microglobulin (β₂m) were produced in Escherichia coli as insoluble inclusion bodies. These proteins were dissolved in a buffer containing urea and then refolded in the presence of synthetic VY8 or RY11 peptide as previously described (28). In this construct, there was no biotinylated tag sequence at the C terminus of the H chain. Refolded proteins were purified by size-exclusion and anion-exchange chromatography analysis, pooled, dialyzed against PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.8 mM KH₂PO₄ (pH 7.5)), and used both for DSC and CD measurements. The resultant protein solutions were in the concentration range from 0.3 to 0.7 mg/ml, as determined by UV absorption at 280 nm; and the molecular masses of the protein complexes were calculated from the amino acid composition.

For DSC measurements, excessive heat capacity curves were monitored by an ultrasensitive scanning microcalorimeter (VP-DSC; MicroCal) at a heating rate of 1 K/min with a sample cell volume of ∼0.5 ml. The experimental data were baseline-corrected and subjected to deconvolution by using the software package ORIGIN for DSC (MicroCal), based on the assumption that the macromolecule is composed of a number of domains, each of which is involved independently in a “two-state” transition between folded and unfolded states. Each transition is characterized by two parameters, T_m and ΔH_m, in which T_m is the thermal midpoint of a transition and ΔH_m is the calorimetric heat change calculated from the area under the transition peak.

For CD measurements, changes in the ellipticity (as θ) with heating from 4° to 90°C were monitored at 222 nm and other wavelengths by a JASCO J-725 spectropolarimeter with a sample cell volume of ∼0.4 ml in a quartz cell with an optical path length of 2 mm. The T_m value in the CD analysis was calculated by using the standard analysis software provided by the manufacturer (JASCO).

We previously reported that in HIV-infected patients with HLA-B35, Nef protein elicited the most dominant CD8 T cell responses (6), with a short epitope (VY8, VPLRPMTY) being dominant early and a subsequent shift to an N-terminal extended long epitope (RY11, RPQVPLRPMTY) in the chronic phase (10). However, VY8 is entirely contained within RY11 and may therefore be the minimum epitope for CTLs. To clarify this issue, we generated CTL clones by stimulating PBMC of four HIV-infected individuals with either VY8 or RY11 peptide and then analyzed their Ag fine specificity by cytotoxic assays. Three CTL clones (01-127, 19-139, and 33-1) generated with VY8 stimulation showed a potent cytotoxic activity toward primary CD4⁺ cells pulsed with VY8 and an activity of markedly lesser strength toward those pulsed with RY11 peptide (Fig. 1,A), confirming their optimal epitope to be VY8. In contrast, the other three CTL clones (01-113, 01-231, and 03-8) generated with RY11 stimulation showed a potent cytotoxic activity toward primary CD4⁺ cells pulsed with RY11 and no activity toward those pulsed with VY8 (Fig. 1 A), confirming their optimal epitope to be RY11. Ag fine specificity of the CTL clones was also confirmed in terms of the HLA-B35 tetramer binding (see below). These data indicate that VY8 and RY11 are optimal epitopes for HLA-B35 and are recognized by a different set of CTL clones.

We next asked whether CTL antiviral cytotoxic activity is different between specificities. The CTL clones showed a significant difference in functional avidity toward their cognate Ags between the specificities (p = 0.023, two-tailed t test), with EC₅₀ values being 2.9 ± 0.85 × 10⁻¹⁰ and 1.3 ± 0.37 × 10⁻⁸ M for VY8 and RY11, respectively (Fig. 1,A). Next, the same cells were infected with vaccinia virus or HIV-1 expressing Nef_SF2 and analyzed for their susceptibility to killing by the CTL clones. The VY8-specific CTLs showed more potent cytotoxic activity toward virus-infected CD4⁺ cells than the RY11-specific ones regardless of the viruses used (Fig. 1, B and C).

We next examined TCR-pMHC interactions by analyzing the binding specificity and activity of CTL clones toward HLA-B35 tetramers. CTL 19-139 and 01-231 were exclusively stained by their cognate HLA-B35 tetramers, whereas an Env-specific CTL clone was not stained by any of the HLA-B35 tetramers examined (Fig. 2,A), confirming the specificity of the CTL clones as well as the integrity of our peptide-HLA class I complex preparations. Also, titration of the HLA-B35 tetramers showed comparable binding activity of the CTL clones toward the cognate HLA-B35 tetramers (Fig. 2,B). We then examined the kinetics of the dissociation of cognate HLA-B35 tetramers from CTL clones. There was no substantial difference between CTL 19-139 and CTL 01-231 in dwell time of interaction with the cognate HLA-B35 tetramers (Fig. 2 C), suggesting comparable kinetic interactions between VY8 and RY11-specific TCRs and their cognate pMHC. However, these results appeared to be inconsistent with the data showing the favorable functional avidity of VY8-specific CTLs (as described).

It has been shown that TCR reconstruction on TCR-deficient T cells is advantageous to investigate how the TCR-pMHC interaction influences T cell activation (27, 28, 29, 30) because primary T cells can increase or decrease their sensitivity/avidity for an epitope through changes in their inhibitory receptor expression and membrane organization as well as via a redistribution of signaling molecules in some circumstances (31, 32, 33, 34, 35). To further clarify how the TCR-pMHC interacts, we cloned TCR-encoding genes of CTL 19-139 (VY8 specific) and CTL 01-231 (RY11 specific), and functionally reconstructed their TCRs (designated TCR-139 and TCR-231, respectively) on TCR-deficient T cell line TG40 (27, 28). The resultant TG40-139 and TG40-231 cells showed CD3 expression and HLA-B35 tetramer binding activity at comparable levels (Fig. 3,A), in good agreement with the observations obtained for the parental CTLs (Fig. 2, B and C). After further transduction of the cells with human CD8α, both cells showed IL-2 secretion at a comparable level in response to anti-CD3 mAb stimulation (Fig. 3,B), confirming the integrity of the TCR-mediated signaling machinery in these cells. Then, functional avidity of TG40-139 and TG40-231 cells toward the cognate Ags was determined by coincubation of target cells and peptides. Virtually no difference was observed in their functional avidities (Fig. 3 C), suggesting comparable TCR-pMHC interactions between the specificities. It should be noted that the peptides were always present for the duration of the assay (see below).

During a number of attempts to clarify the reasons for the variable observations among assays, we noticed that the avidity of TCR-transduced cells was much decreased when the peptide-loaded target cells were washed before coincubation with the TCR-transduced cells (Fig. 3,D). Under this washing-off condition, TG40-139 cells showed significantly more potent functional avidity than TG40-231 cells (Fig. 3,D). We then analyzed the rate of peptide-off from pMHC by using the TCR-transduced cells. The target cells, which had been loaded with a peptide followed by washed-off, were taken and subsequently mixed with TG40 cells expressing the cognate TCR. The extent of the TG40 response should be proportional to the actual pMHC dose retained on the target cell surface. The data showed that the decay of the VY8/HLA-B35 complex was much slower than that of the RY11/HLA-B35 one, as the half-life values of pMHC were 3.3 ± 0.83 × 10² and 1.0 ± 0.03 × 10² min for VY8 and RY11, respectively (Fig. 3 E).

To look for variant peptides that could affect pMHC decay and the susceptibility to stimulation of T cells, we examined a series of alanine substitutions in both peptides for their activity to sensitize TCR-transduced T cells under the no-wash condition. VY8 with an alanine substitution at position 5 (designated VY8-5A) showed more potent reactivity with TCR-139 than did VY8, whereas VY8-1A and VY8-3A had moderate reactivity (Fig. 4,A). However, TG40-139 cells did not respond to the other five VY8 variants up to a 100-μM concentration (Fig. 4,A). In TCR-231, RY11-8A showed the most profound response, whereas RY11-3A and RY11-4A had reactivity comparable to that of RY11 (Fig. 4,B). However, RY11-9A showed weak reactivity; and the other seven RY11 variants had no reactivity (Fig. 4 B). We also tested various amino acid variations (57 variant peptides in total) for their reactivity toward TCR-139 and TCR-231 (data not shown), but VY8-5A and RY11-8A showed the most pronounced effects on IL-2 production by TCR-transduced T cells.

Next, we examined the decay of a series of alanine variant peptides from HLA-B35. The pMHC decay of VY8-5A and VY8-3A was substantially delayed with a half-life of >4 × 10³ min, whereas the half-life of VY8-1A was slightly more rapid with one of 2.0 ± 0.80 × 10² min (Fig. 4,C). None of the other VY8 variants showed any reactivity (Fig. 4,C), consistent with the data obtained under the no-wash condition (Fig. 4,A). Among the RY11 variants, the decay of RY11-8A was substantially delayed with a half-life of 2.7 ± 0.44 × 10³ min, whereas that of RY11-3A, RY11-4A, and RY11-9A was slightly delayed with half lives of 2.5 ± 0.44 × 10², 2.0 ± 0.40 × 10², and 2.2 ± 0.40 × 10² min, respectively (Fig. 4,D). None of the other RY11 variants showed any reactivity (Fig. 4,D), consistent with the data from the no-wash condition (Fig. 4 B). Taken together, VY8-5A and RY11-8A provided the most profound effects on pMHC decay and T cell stimulation among the wild-type and variant peptides examined.

To ask whether the variant Ags can improve the susceptibility to killing by CTLs, we took advantage of the fact that the alanine substitution at Pro⁸² of Nef_SF2 resulted in the generation of both VY8-5A and RY11-8A variant Ags. Primary CD4⁺ cells were transfected with the wild-type or the Ala⁸² variant nef genes. The cells with the Ala⁸² variant showed substantially increased susceptibility to killing by both CTL 19-139 and CTL 01-231 (Fig. 5), suggesting that the variant antigenic peptides with slower pMHC decay rendered HIV-infected cells more susceptible to CTL-mediated viral containment.

To characterize the biochemical differences between VY8/HLA-B35 and RY11/HLA-B35 complexes, we analyzed the thermostability of free β₂m and that of these heterotrimers (composed of β₂m, H chain, and peptide) by DSC. The heat capacity curve of β₂m, a protein composed of a stable domain containing exclusively β strands, showed a single two-state transition at T_m of 64.18°C (Fig. 6,A and Table I), in good agreement with a previous report (36). In contrast, the heat capacity curve of RY11/HLA-B35 was characterized by two partly overlapping peaks with the melting temperature of the first peak (T_m¹) substantially below that of β₂m (Fig. 6, A and B and Table I). The other single transition at the high temperature peak appeared to result from the melting of β₂m, as the T_m² value of RY11/HLA-B35 was comparable to that of free β₂m (Table I). These results suggest that the melting of RY11/HLA-B35 started with unfolding of the H chain and concomitant release of folded β₂m, which subsequently melted at the higher temperature.

In contrast, the heat capacity curve of VY8/HLA-B35 appeared to be quite different from that of the RY11 counterpart, as the deconvolution of the VY8/HLA-B35 experimental heat capacity profile showed three peaks, which heavily overlapped each other and were less well separated (Fig. 6, A and C and Table I). The transition of the third peak, which was separated by a shoulder at the high temperature side of the second peak (Fig. 6,C), could be correlated with the melting of β₂m, as both melting temperatures were comparable, although interestingly, the ΔH_m³ value of VY8/HLA-B35 was much lower than that of free β₂m (Table I). The melting of the H chain of VY8/HLA-B35 could not be annotated on a single transition, either T_m¹ or T_m². Rather, the results suggested that the melting of the entire VY8/HLA-B35 complex occurred simultaneously and cooperatively with the H chain and β₂m.

To examine the contribution of peptides on the thermostability profile of pMHC, we also analyzed RY11-8A and VY8-5A in complex with HLA-B35 by DSC. Both single mutations showed substantial effects on the heat capacity profiles of overall pMHC complexes and increased total enthalpy values compared with those of their respective wild-type counterparts (Fig. 6, A, D, and E and Table I). Notably, transitions at high temperature peaks in the variant peptide complexes, corresponding to T_m² for RY11-8A/HLA-B35 and T_m³ or T_m⁴ for VY8-5A/HLA-B35, appeared to rely on a contribution from β₂m, although these enthalpy costs were considerably larger than the enthalpy change of free β₂m (Table I), suggesting a substantial contribution from the H chain to these transitions.

To further examine the contribution of peptides on the thermostability profile of pMHC, we obtained CD profiles of these pMHC heterotrimers to see the thermally induced changes in their secondary structures. We observed substantial differences in CD profiles between β₂m alone and all pMHC heterotrimers examined (Fig. 6,C). β₂m alone melted with T_m in the CD analysis of 66.3°C (Fig. 6,D), in good agreement with the T_m¹ value in the DSC analysis (Table I) as well as with a previous report (36). In contrast, all pMHC heterotrimers had much larger negative molar ellipticity at 222 nm at a low temperature range than β₂m alone (Fig. 6,C), most likely reflecting the presence of α helices in their H chain subunits. In addition, each pMHC complex showed a different reduction in their negative molar ellipticity with increasing temperatures (Fig. 6,C), highlighting the contribution of peptides on the thermostability of the secondary structure in their H chain subunits. Specifically, RY11/HLA-B35 melted with a T_m (CD) of 52.1°C (Fig. 6,D), a value consistent with the T_m¹ in the DSC analysis, confirming the observation made by DSC that melting of RY11/HLA-B35 started with unfolding of the H chain subunit. In addition, VY8/HLA-B35 melted at a much higher temperature (Fig. 6,D) than RY11/HLA-B35, confirming the potent thermostability of the VY8/HLA-B35 complex as observed in the DSC analysis. Finally, the HLA-B35 complexes with the variant peptides melted at higher temperatures than their wild-type counterparts (Fig. 6, C and D), confirming again the contribution of variant peptides on the profound thermostability in these pMHC complexes as observed in DSC analysis.

In the present study, using TCR-reconstructed cells we clearly showed that the difference in antiviral cytotoxic activity between mature CTLs specific for two different but closely related antigenic Nef peptides (VY8 and RY11) was not caused by the difference in functional avidity of TCR-pMHC interactions. Rather, our data demonstrated that the antiviral activity of these effector CTLs was much influenced by peptide intrinsic factors including peptide-off rate and cooperative thermodynamics of the cognate pMHC. The data obtained by introduction of a mutation in the Nef protein that resulted in the alteration of both epitopes to VY8-5A and RY11-8A further confirmed the association between these peptide intrinsic factors and the susceptibility of Nef-expressing cells to killing by the cognate CTLs. Our results are partly in line with those of previous studies demonstrating that the peptide-off rate of pMHC is an important factor for generating immunodominance hierarchy in class I (34, 37, 38, 39) and class II (40, 41) MHC-restricted T cell responses, i.e., the slower the peptide-off rate, the greater the abundance of a given pMHC on the surface of APCs, which leads to the generation of immunodominant T cell responses (26). However, immunodominant peptides are not always those with the highest density presented at the target cell surface (23, 24), and immunodominant CTLs do not always play a dominant role in containment of HIV replication (25). The results shown here extend these previous findings that interdependent and cooperative thermostability profiles of pMHC induced by antigenic peptides can be associated with efficient recognition by CTLs for killing virus-infected target cells.

The DSC and CD measurements showed significantly different thermostability profiles among HLA-B35 in complex with VY8, RY11, and their variant peptides. In comparison of the thermostability of HLA-B35 complexes between wild-type peptides and their variants, we found significant effects of the mutations on thermal stabilization of the entire pMHC, as the total enthalpy values required for unfolding of HLA-B35 in complex with the variant peptides were substantially increased compared with those for their respective wild-type counterpart. This thermal stabilization by the variant peptide was corroborated by the DSC and CD analyses and is most likely associated with slow dissociation of these peptides from pMHC, as observed in the cell-based assays. In contrast, the calorimetric unfolding enthalpy of RY11/HLA-B35 obtained by deconvolution of the experimental heat capacity curve was higher than that of its VY8 counterpart, although the cell-based assays showed rapid dissociation of RY11 from pMHC. In this regard, it is possible that RY11 and HLA-B35 may bind with multiple different conformations; because relatively long peptides can be accommodated on the peptide binding groove of HLA class I molecules with their central region bulged (42, 43, 44, 45) or either end extended away (46). This result is less likely in this study, however, because TCR-231-transduced cells responded to target cells pulsed with RY11 but not to those with truncated peptides such as VY8 and RM9 (RPQVPLRPM, data not shown). Also, the HLA-B35 tetramers prepared with RY11 showed binding exclusively with TCR-231 but not with other TCRs including VY8-specific TCR-139. It is also conceivable that the peptide-off rate from the membrane-bound and glycosylated form of pMHC (i.e., present on the cell surface) is not correlated with the thermostability of the soluble form of pMHC (i.e., using bacterially produced extracellular domain). Alternatively, peptides that are endogenously loaded onto the empty MHC class I with the assistance of a specialized multimeric unit called the peptide-loading complex (47, 48) in the endoplasmic reticulum could have conformational characteristics different from those of molecules refolded in vitro. More importantly, the RY11/HLA-B35 complex showed two relatively simple two-state transitions in thermal unfolding, in which a high-temperature transition corresponds to free β₂m. Such an unfolding pattern has been reported for various self-peptides in association with HLA-B27 (36, 49). In contrast, the VY8/HLA-B35 complex and two other variant complexes showed significantly interdependent and cooperative unfolding processes among heterotrimer subunits and structural domains, suggesting the critical contribution of β₂m in maintaining antigenicity of the peptide in association with the H chain. However, whether such a conformational characteristic in a given pMHC can be directly attributable to efficient docking by cognate TCRs, to preferential loading of peptides in an intracellular peptide selection process, or to both events needs to be examined by further intense experiments.

In HIV-infected cells, antigenic peptides are generated through the endogenous MHC class I Ag processing and presentation pathway for CTL recognition. Peptides generated in the cytosol mainly by the proteasome are translocated into the endoplasmic reticulum by mediation of the TAP, and then loaded onto the empty MHC class I (47, 50). It has been reported that sequence specificities by TAP can influence the efficiency of epitope presentation by cell surface MHC class I molecules (47, 50, 51). It is therefore conceivable that the increased susceptibility of Nef-expressing cells to CTL recognition by the Ala⁸² mutation observed in our study might be attributable to the enhancement of the TAP efficiency, in addition to peptide intrinsic factors including peptide-off rate and thermodynamics of the cognate pMHC. However, this scenario is less likely because it is well known that the amino acid substitutions in the middle of epitopes, such as VY8-5A and RY11-8A in our study, have only a limited effect on TAP efficiency (52, 53, 54). Considering that a number of human viruses including HIV-1 can somehow abrogate the TAP function (50, 55), how the TAP efficiency influences the susceptibility of cells infected with various variant viruses to CTL killing is an important question to be addressed in future studies.

Both VY8 and RY11 share anchor residues, proline at position 2 and tyrosine at the C terminus, which are optimal for binding with HLA-B*3501 (56, 57); and both Ags are dominantly recognized in HLA-B*3501⁺ individuals with an HIV-1 infection (6, 10, 58). Even in such a case, the improved thermostability of pMHC by an amino acid substitution within the epitopes, even other than a primary anchor residue, can substantially enhance the susceptibility to recognition by CTLs for killing target cells, suggesting that the altered peptide ligand strategy is capable of enhancing CTL-mediated immune responses against HIV-1 infection similar to that used for anti-cancer vaccines targeting self-Ags (34, 39). Our data thus highlight the importance of incorporating thermostability data in the process of rational optimization of Ags that support profound antiviral activity by HIV-specific CTLs.

We thank S. Dohki, Y. Idegami, and T. Akahoshi for excellent technical help.

The authors have no financial conflict of interest.