Suboptimal Activation of CD8⁺ T Cells by Melanoma-derived Altered Peptide Ligands: Role of Melan-A/MART-1 Optimized Analogues¹

Melanoma patients can mount T-cell-mediated immune responses against specific antigens expressed on tumor cells, but such immunological responses are often ineffective in controlling neoplastic growth in vivo. Hence, the ability of tumor cells to develop strategies for avoiding immune recognition and/or destruction, a phenomenon known as “tumor immune escape,” has recently become the object of intensive investigation (1).

In addition to mechanisms involving down-regulation of HLA/tumor antigen expression, release of immunosuppressive cytokines, or expression of proapoptotic molecules (1), suboptimal T-cell triggering could be considered a potential additional cause of tumor immune escape (2, 3). In fact, evidence of suboptimal CD8⁺ T-cell activation has often been reported in tumor-immune interaction, and signs of functional impairment compatible with T-cell tolerance or anergy appear to be a constant feature of T lymphocytes infiltrating tumor lesions (4). In particular, loss of IL-2³ production despite adequate costimulation, as well as development of tolerant phenotype, has been shown to occur in both human (2) and animal tumor models (5, 6).

Hence, suboptimal T-cell triggering or selective blunting of their activation status could be invoked to explain the paradoxical coexistence of T lymphocyte infiltrates with unhampered tumor growth frequently observed in melanoma patients (1). Indeed, the self nature of many of the tumor antigens identified thus far suggests that immunoregulatory mechanisms responsible for tolerance maintenance might also be involved in down-regulating such an antitumor T-cell response in vivo (3, 7).

Among the molecular pathways involved in the maintenance of peripheral T-cell tolerance (2, 8), interaction with peptide analogues with partial agonist or antagonist activity has been shown to play a potential role in mediating the functional impairment of self-reactive T cells (9, 10, 11, 12).

In the present study, we investigated the possibility that the expression by tumor cells of peptide analogues with partial agonist/antagonist activity could contribute in functionally turning-off self-reactive melanoma-specific T cells. This hypothesis stems from our previous observations that the immunodominant Melan-A/MART-1 (referred to hereafter as MART-1_27–35) peptide, a highly immunogenic antigen in HLA-A2⁺ melanoma patients, embodied a general sequence motif frequently occurring within a variety of endogenous peptides and proteins (13), a finding recently confirmed by Dutoit et al. (14). Some of these peptide analogues were found to act as APLs, i.e., as partial agonists or antagonists of MART-1_27–35-reactive T cells by selectively impairing IL-2 production in response to the native peptide (15).

Here we show that suboptimal activation of antigen-specific CD8⁺ T cells, as defined by loss of IL-2 release in the presence of IFN-γ release, can be detected in response to melanoma cells both in vitro and in vivo and that tumor cells express endogenous MART-1_27–35-like peptides exerting partial agonist/antagonist activity on tumor-specific CD8⁺ T cells. Additionally, we report that T-cell responses resistant to such anergy induction can be generated by optimized MART-1 peptide analogues, which were previously shown to produce quantitatively and qualitatively improved antitumor reactivity (16).

A42 is a MART-1_27–35-specific, HLA-A*0201-restricted, cytotoxic CD8⁺ T-cell clone isolated from tumor-infiltrating lymphocytes (17, 18) and maintained in culture with 3000 IU/ml IL-2. Its clonality has been described previously in detail (19). 501mel and 526mel are HLA-A*0201⁺/MART-1⁺ melanoma cell lines (17). 501mel is autologous to A42. The anti-MART-1 T-cell line used for evaluating recognition of different melanoma cell lines was generated by in vitro peptide sensitization from PBMCs of a HLA-A*0201⁺ melanoma patient (patient 7) and extensively characterized for MART-1 specificity in a previous study (20). Generation of peptide-specific T cells was performed by culturing PBMCs from HLA-A*0201⁺ melanoma patients with peptide and exogenous IL-2, as described previously (20). Me 2658, Me 1768, Me 8863, Me 1190, and Me 0966 are HLA-A*0201⁺/MART-1⁺ melanoma cell lines obtained from tumor samples of subjects surgically resected in our Institute. HLA-A2 and MART-1 expression was evaluated by cytofluorometric staining with the anti-HLA-A2 mAb (clone BB7.2) and the anti-MART-1 mAb M27-C10 (21). Data were analyzed by FACSCalibur and CellQuest software (Becton Dickinson, San Jose, CA). 624.28mel is a HLA-A*0201-defective clone derived from HLA-A*0201⁺/MART-1⁺ 624mel. A375 is a MART-1-, tyrosinase-, gp100-defective HLA-A0201⁺ melanoma cell line. T2 is a TAP-deficient lymphoma cell line that effectively presents exogenously supplied peptides in the context of HLA-A*0201.

Tumor cell suspensions were obtained by mechanical processing or enzymatic digestion of HLA-A0201⁺ melanoma lesions as described elsewhere (22) and stored in liquid nitrogen. Routine histological diagnosis was performed for tumor samples used in this study. All lesions had a T-cell infiltrate ranging from 10% to 50% of total cell number.

In the present study, the following peptides were used: MART-1_27–35 (AAGIGILTV; Ref. 13); MART-1_31–39 (GILTVILGV; Ref. 18); MART-1_27–35 1L (LAGIGILTV; Ref. 16); MART-1_26–35 2L (ELAGIGILTV; Ref. 23); Flu A matrix protein M1 (GILGFVFTL; Ref. 13); Escherichia coli methionine synthase (AAGIGIIQI; Ref. 13); bone marrow stromal antigen (Bst-2; LLGIGILVL; Ref. 13); and MART-1_32–40 (ILTVILGVL; Ref. 24). Peptides were synthesized as described previously (13).

Acid elution and reverse phase-HPLC resolution of melanoma peptides were performed as reported previously from two different melanoma lines, 624mel and 526mel (24, 25). Individual HPLC fractions were lyophilized and reconstituted in 200 μl of HBSS (Life Technologies, Inc.) and stored at −20°C before use in functional assays.

Cytokine release in supernatants was detected by ELISA kits for IFN-γ (Mabtech, Nacka, Sweden), IL-2 (Quantikine; R&D, Minneapolis, MN), and IL-10 (Bender MedSystems, Vienna, Austria), according to each manufacturer’s instructions. A total of 10⁵ effectors/well and an E:T ratio of 1:1 were used. IL-2 release was also confirmed by proliferation assay with the IL-2-dependent CTLL line. ELISPOT assay was performed as described previously (22), using the IFN-γ ELISPOT kit (Mabtech) and IL-2 ELISPOT kit (BD Biosciences, San Diego, CA). PBMCs (1.67 × 10⁵ cells/well) or cultured T cells (5 × 10⁴ cells/well) were incubated with 1.67 × 10⁴ cells/well targets. For the evaluation of different peptide recognition, target cells were preincubated with 2 μm of each specific peptide. In case of fresh suspensions, cells from fresh tumor suspensions (10⁵ cells/well) were used in either the presence or absence of the anti-class I HLA w6.32 mAb (source). Plates were counted by automatic ELISPOT reader system (AID Bioline, Torino, Italy). Results are presented as number of spots/10⁶ PBMCs. For statistical analysis, Student’s t test for unpaired samples was used. Values of P < 0.05 were considered significant.

Biotinylated MART-1_27–35/HLA-A*0201 monomers (a gift from Dr. Claudia Giachino; Maugeri Foundation Institute, Pavia, Italy) had been synthesized as described previously (26). Tetramer binding was evaluated as double staining with anti-CD8 mAb (FITC antihuman CD8 IgG1; BD PharMingen) according to the manufacturer’s instructions. Cells were then analyzed using FACSCalibur and CellQuest software (Becton Dickinson).

This assay was performed as described previously (15). Briefly, T cells were incubated overnight with different adherent melanoma cell lines or with A375 cells pulsed with melanoma-derived peptide fractions, with MART-1 derived or control peptides. T cells were then harvested and evaluated for cytokine release in response to 501mel. For statistical analysis, Student’s t test for unpaired samples was used. Values of P < 0.05 were considered significant.

To evaluate whether melanoma cells could trigger suboptimal T-cell activation, HLA-A*0201-restricted MART-1_27–35-specific T cells (20) were analyzed for the ability to release IFN-γ and IL-2 in response to MART-1, provided as exogenous peptide or endogenously processed by melanoma cells. These T cells, which were fully triggered (i.e., released both IFN-γ and IL-2 when exposed to T2 cells pulsed with MART-1_27–35 exogenous peptide; Fig. 1, A and B), were tested for optimal recognition of a panel of melanoma cell lines expressing comparable levels of HLA-A*0201 and MART-1 (Table 1). As reported in Fig. 1, similar amounts of IFN-γ were released by the anti-MART-1 T cells in response to different melanoma lines, whereas significant levels of IL-2 release could be detected in only two of seven cell lines tested.

These data suggested that melanoma cells could suboptimally trigger antigen-specific T-cell reactivity inducing selected (i.e., IFN-γ release) lymphocyte functions, without mediating full T-cell activation (i.e., IL-2 release).

To see whether melanoma cells could themselves down-modulate T-cell recognition through the expression of peptide analogues with APL activity, we analyzed IL-2 and IFN-γ release by MART-1_27–35-specific T cells in response to peptides eluted from the suboptimally recognized melanoma cell lines 624mel and 526mel. HLA-associated peptides were stripped from the tumor cell surface, fractionated by HPLC, and used to reconstitute the epitope recognized by the anti-MART-1_27–35 T-cell clone A42 (17, 18, 19).

Distinct peptide fractions (i.e., fractions 43–44, 48–49, and 51–52 for 624mel and fractions 43–44 and 49–50 for 526mel) were reproducibly able to induce IFN-γ production by the anti-MART-1 T-cell clone (Fig. 2, A and B). Comparable data were observed when tumor necrosis factor α release or lytic activity was analyzed (data not shown). However, only fraction 43–44 from both melanomas could elicit significant IL-2 production by the same effector cells (Fig. 2, C and D). To address whether this pattern of epitope reconstitution could be due to the presentation of MART-1_27–35 APLs by 624mel and 526mel cells, we tested HLA-A*0201⁺ melanoma cell-derived peptides for the ability to induce T-cell anergy (i.e., selective impairment of antigen-mediated IL-2 production) in MART-1-specific CD8⁺ T cells. To this purpose, we used a previously described specific antagonism/anergy assay (15). The anti-MART-1 T-cell clone A42 was exposed overnight to melanoma-eluted peptide fractions 48–49 and 51–52 from 624mel and fraction 49–50 from 526mel, pulsed on HLA-A*0201⁺ MART-1⁻ A375 melanoma cells. Subsequently, lymphocytes were collected and tested for IFN-γ and IL-2 release in response to MART-1⁺ melanoma cells (501mel), known to trigger optimal recognition (i.e., induction of both IFN-γ and IL-2 release) by A42 cells. Fraction 43–44 from 526mel (reconstituting optimal T-cell activation) and an irrelevant peptide fraction (40–41) from 624mel were also used as negative controls. Synthetic peptides (Bst-2 and E. coli) were included as control epitopes with and without antagonist activity, respectively (13).

No change in IFN-γ release in response to 501mel was observed after exposure to melanoma-eluted peptide fractions (Fig. 3, A and C), while a significant impairment of IL-2 release was detected after T cell incubation with fractions 48–49 and 51–52 from 624mel (Fig. 3,B) and fraction 49–50 from 526mel (Fig. 3 D). As expected, the synthetic antagonist peptide Bst-2 mediated a comparable effect. On the contrary, no down-regulation of IL-2 production was induced by fraction 43–44 (mediating both IFN-γ and IL-2 in epitope reconstitution assay), by the irrelevant peptide fraction 40–41, or by the null peptide from E. coli. Notably, an anti-Flu GILGFVFTL epitope-specific T-cell line was not inhibited by preexposure to fraction 48–49 and 51–52 (data not shown), indicating that the observed phenomenon was not due to toxic effects of fractions themselves. This phenomenon was observed only when antigen-presenting cells (i.e., A375) were used for exposing A42 to peptide fractions, and no self-presentation of exogenous peptides was obtained by A42 T cells (data not shown).

Taken together, these data suggest that two of three peptide fractions recognized by fully functional MART-1_27–35-specific effectors may act as partial agonist or APLs of MART-1 peptides. Additionally, these results rule out the possibility that the partial agonist activity of melanoma-eluted fractions (Fig. 2) could be related to HPLC artifacts (i.e., immunodominant epitope smearing along the column). To confirm these data, we performed the anergy assay by preexposing T cells to 624mel cells, from which peptide fractions were derived. 501mel (HLA-A*0201⁺/MART-1⁺), A375 (HLA-A*0201⁺/MART-1⁻), or 624.28mel (HLA-A*0201⁻/MART-1⁺) cells were additionally tested as negative controls. Preexposure to 624mel induced the same effect observed with eluted peptide fractions, i.e., conserved IFN-γ production (Fig. 3,E) and impairment of IL-2 release (Fig. 3,F). However, no T-cell anergy was observed after preincubation with melanoma cells negative for either the antigen (A375) or the HLA restriction allele (624.28mel; Fig. 3, E and F), thus suggesting that the phenomenon was strictly dependent on MART-1 and HLA-A*0201 expression. Notably, exposure to 501mel, mediating both IFN-γ and IL-2 release and thus optimal T-cell recognition (Fig. 1), did not display any anergistic activity on A42 (Fig. 3, E and F).

We and others had previously characterized and sequenced one naturally processed and presented melanoma peptide (MART-1_32–40) eluting in HPLC fraction 48–49 and recognized by anti-MART-1_27–35 T-cell clones and TILs (24). We thus evaluated whether this peptide could be responsible for the observed anergistic activity in fraction 48–49. As reported in Fig. 4,A, MART-1_32–40 acted as a partial agonist, mediating IFN-γ but not IL-2 release by the anti-MART-1 T-cell clone A42. Using the above-described anergy assay, selective loss of IL-2 release (Fig. 4,B), associated with retainment of IFN-γ releasing activity against 501mel (Fig. 4 C), was observed after preexposure of T cells to the 32–40 peptide but not to the native MART-1_27–35 epitope. These results suggest a role for MART-1_32–40 in the APL activity of fraction 48–49.

Taken together, our data indicate that partial agonist/antagonist peptides endogenously processed and presented by MART-1⁺/HLA-A*0201⁺ melanoma cells are involved in impairing IL-2 secretion in response to melanoma cells by a mechanism based on TCR-specific engagement. To evaluate whether this phenomenon could also occur in vivo, HLA-class I-restricted IFN-γ and IL-2 release by freshly isolated CD8⁺ T cells infiltrating melanoma lesions was analyzed by ELISPOT assay. Indeed, in four of seven different HLA-A*0201⁺ melanoma patients evaluated, HLA-class I-restricted secretion of IFN-γ could be observed in fresh cellular suspensions obtained from melanoma nodules infiltrated with T cells (Fig. 5). On the contrary, class I-restricted IL-2 secretion was not detected in any of the investigated lesions (Fig. 5). Although several mechanisms could be responsible for T-cell anergy in the neoplastic environment, the lack of T-cell-mediated IL-2 production at the tumor site could be interpreted as a potential sign of APL expression by tumor cells.

LAGIGILTV (27–35/1L) and ELAGIGILTV (26–35/2L) are two MART-1 optimized peptide analogues known to generate more potent anti-MART-1 T-cell responses (16, 23). Although they bear the same amino acid substitution (Leu) at position 1 and 2, respectively, these two epitopes have been shown to exert their enhanced immunogenicity through different mechanisms, i.e., by mediating superagonist activity (27–35/1L; Ref. 16) related to a more efficient interaction with the TCR and by binding more efficiently to HLA-A*0201 molecules (26–35/2L), as compared with the native peptide (23).

To evaluate whether T cells generated with these two optimized MART-1 analogues could be differentially susceptible to APL-mediated anergy, we characterized the functional features of T cells generated with MART-1_27–35/1L and MART-1_26–35/2L. MART-1-specific T cells were thus generated by in vitro peptide sensitization of PBMCs from HLA-A*0201 melanoma patients. After 2-week cultures, the frequency of T cells cross-recognizing MART-1_27–35 native peptide and the MART-1⁺ melanoma line 501mel, as detected by IFN-γ ELISPOT assay, was reproducibly higher in T cells raised with the optimized variants as compared with native peptide counterpart MART-1_27–35 (Fig. 6,A). Staining with HLA-A*0201 tetramers containing the native MART-1_27–35 peptide confirmed that an increased number of T cells cross-recognizing native epitope could be raised with the two optimized variants in comparison with natural peptide (Fig. 6 B). The 2L peptide variant appeared in all cases associated with the highest levels of specific T-cell generation, whereas 1L peptide-elicited responses appeared more case sensitive.

To evaluate whether qualitative differences in cytokine profiles could characterize T cells generated with MART-1 analogues as compared with lymphocytes generated with native peptide, we then analyzed IL-2 and IL-10 release in response to specific antigenic stimuli. As shown in Fig. 7, the 1L analogue reproducibly generated T cells with the highest capacity to release IL-2 in response to native MART-1 in all of the three melanoma patients evaluated, although improved release as compared with the native peptide was mediated by the 2L analogue as well. Fig. 7 also reports IL-10 levels detected in the same supernatants tested for IL-2 release. In contrast to what was observed for IL-2, variant T cells generated with 2L peptide produced higher amounts of IL-10 as compared with lymphocytes originated with 1L or native peptides. This finding suggests that optimized analogues may indeed lead to the generation of qualitatively different MART-1-specific T cells, possibly mediating a differential TCR triggering or the expansion of cross-reacting lymphocytes with different functional features.

We then investigated whether anti-MART-1 T cells generated with either 1L or 2L, due to their improved functional features, could show different sensitivity to T-cell anergy mediated by melanoma-eluted peptide fractions. T cells raised with different MART-1 peptides were exposed to HLA-A*0201⁺/MART-1⁻ A375 mel cells loaded with 624mel-eluted peptide fractions 43–44, 48–49, and 51–52, irrelevant fraction 40–41, E. coli peptide, and the known natural antagonist peptide Bst-2 (15).

As for the anti-MART-1 T-cell clone A42, selective impairment of IL-2 release to 501mel, in the presence of conserved IFN-γ secretion, was reproducibly observed when T cells generated with native 27–35 peptide were pretreated with fraction 48–49, fraction 51–52, or Bst-2 (Fig. 8, A and B, top panels), but not with the irrelevant fraction 40–41, fraction 43–44, or the E. coli peptide. On the contrary, IL-2 release by 1L-raised T cells was independent of preexposure to either melanoma-eluted fraction 40–41 or 51–52 or peptide antagonists, such as Bst-2. Thus these lymphocytes resulted insensitive to anergy induction by melanoma-derived peptides (Fig. 8, A and B, middle panels). Surprisingly, 2L-specific T cells were sensitive to melanoma anergizing fractions, which inhibit IL-2 release in response to 501mel without affecting IFN-γ response. Similar data could be obtained by preexposing T cells to 624mel, from which peptide fractions were eluted (data not shown). These results indicate that T cells generated with the 1L analogue display unique features enabling them to resist APL-mediated T-cell anergy. Their potentiated ability to release IL-2 in response to the native epitope could play a pivotal role in the phenomenon.

In this study, we show that melanoma cells can be suboptimally recognized by anti-MART-1 T cells because of endogenous processing and presentation of APLs with partial agonist/antagonist activity on antigen-specific CD8⁺ T cells. This form of suboptimal T-cell activation is characterized by a reduction of IL-2 release upon recognition of HLA-peptide complex (IFN-γ, tumor necrosis factor α release, or lysis), and it may occur in vivo as well, in T cells infiltrating MART-1⁺/HLA-A2⁺ melanoma lesions. Moreover, we found that this form of T-cell anergy can be overcome when T cells are generated in the presence of optimized peptide analogues of the same epitope against which T-cell response was down-modulated by APLs expressed by melanoma. Resistance to APL-induced T-cell anergy correlates with the ability to release IL-2 after antigenic stimulation and appears to be a feature of T cells generated with superagonist analogues.

Self-specific T cells could play a major role in immune surveillance against tumor cells (7). Low-avidity autoreactive T cells escaping thymic selection can be found in periphery of healthy individuals (27). These cells respond to self-peptide/MHC complexes with low-affinity interactions resulting in a pronounced defect of IL-2-related expansion in the presence of IFN-γ production and cytolytic activity (3).

Thus, loss of IL-2 release in the presence of IFN-γ response against tumor cell lines expressing both HLA and the antigen possibly involves the same mechanisms controlling self-tolerance (28). Among them, APL expression could play a role, particularly in the MART-1 model (13, 15, 29). In fact, MART-1 model is a paradigm of how thymic selection can maintain in periphery a large T-cell pool recognizing a self-peptide that has an immunodominant role in the T-cell response to melanoma (27, 30, 31). The MART-1-specific T-cell repertoire has been shown to cross-react with different natural occurring APLs of the immunodominant MART-1 epitopes (13, 14). The data here presented suggest that peptide analogues with APL activity are possibly expressed by melanoma cells themselves and may contribute in down-modulating T-cell response, thus representing a new form of tumor immune escape. Although no direct sequencing of the fractions was performed in the present study, our results are compatible with those published previously and indicating that the immunodominant MART-1_27–35 migrates in the first fraction, whereas MART-1_32–40 is present in the 48–49 peak and may contribute to the antagonistic activity described here (24, 32).

Our data on the role of APLs expressed on tumor cells in switching off T-cell-specific tumor recognition are in line with those described in other systems. In fact, APL activity is strictly related to the great versatility of TCR-dependent signaling complex that can interact with a broad range of ligands (11). Hierarchical organization of TCR signaling thresholds for cytokine responses may explain the loss of IL-2 production in the presence IFN-γ release upon weak stimuli (12). Most of the self-antigen/tumor antigen-derived epitopes are indeed weak stimulators of the TCR (33). Because T-cell response depends on the efficiency of specific TCR-ligand interaction, the coexpression of different peptides eliciting different immune functions in the same pool of antigen-specific TCRs may represent a possible form of tumor immune escape by down-modulating antigen-specific T-cell activity in vivo. This mechanism of shaping specific T-cell response could indeed contribute to explanation of the coexistence of tumor-specific T cells and tumor cells bearing both antigens and HLA molecules.

Several data support the existence of a partially activated T cell in human melanoma lesions (31). Here we report that T cells freshly isolated from melanoma lesions can release IFN-γ in the absence of IL-2 production. This observation, which is in agreement with previous data obtained by immunohistochemistry and mRNA analysis, strongly supports the suboptimal activation of CD8⁺ T cells at tumor site (34, 35, 36), and it is compatible with results from clinical trials that indicate that high-dose IL-2 administration contributes to revert anergy status of antitumor T-cell populations in humans (37).

Given the limited efficacy achieved thus far by peptide-based vaccines in melanoma patients (38), raising T cells that optimally interact with tumor cells in vivo is one of the primary objectives in designing new vaccination strategies. Indeed, suboptimal T-cell activation by APLs has been shown in the murine system to be overcome by T cells generated in the presence of optimized peptide ligands (5). In the MART-1_27–35 system, two modified variants of the same epitope with different HLA affinity but similar efficacy in qualitatively (mostly by increasing IL-2 release) and quantitatively improving anti-native peptide CD8⁺-specific responses appear to induce MART-1-specific T cells that differ in terms of sensitivity to APL-mediated functional down-modulation. Indeed, suboptimal activation of antigen-specific T cells could be by-passed by lymphocytes generated with the superagonist peptide 1L only. This effect appeared to be related to the improved ability of 1L-generated T cells to release substantial amounts of IL-2 in response to the native MART-1 peptide. This ability could contribute to overcome anergizing stimuli mediated by tumor-expressed APLs or by other immunosuppressive mechanisms present at the tumor site, thus creating a more favorable milieu for T-cell activity. Therefore, among strategies to obtain immune-mediated tumor regressions in cancer patients, the usage of optimized peptide analogues with superagonist activity should be considered a potential tool to revert tumor-mediated functional impairment of antigen-specific T cells.

We thank Drs. Claudia Giachino and Stefan Stevanovic (Department of Immunology, Institute for Cell Biology, University of Tubingen, Tubingen, Germany) for providing reagents and useful suggestions. We are also indebted to Grazia Barp for editorial assistance.