A Superagonist Variant of Peptide MART1/Melan A_27–35 Elicits Anti-Melanoma CD8⁺ T Cells with Enhanced Functional Characteristics: Implication for More Effective Immunotherapy¹

It is now well established that CD8⁺ tumor infiltrating lymphocytes recovered from melanoma patients recognize a variety of unmodified self peptides that are derived from lineage-specific proteins (1, 2), and are bound to MHC class I molecules. Among such epitopes are those derived from the melanosomal proteins tyrosinase, gp100/Pmel17, and MART1. Although robust CTL activity may be observed ex vivo, often in the presence of excess cytokines and antigen, a clear understanding of the functional status of these cells in vivo, and their impact on tumor growth, has been elusive. In contrast, several mechanisms have been described that potentially contribute to tumor cell evasion of the immune response (3), suggesting that any antitumor efficacy achieved by immune effectors may be offset by countermeasures that result ultimately in tumor progression.

However, the prevailing dynamic in vivo between immune effector and tumor cells may not reflect a “fair fight.” Little evidence is available to indicate that CTLs directed to tumor-expressed self epitopes are indeed engaged in a full-fledged autoaggressive response, stemming from a true breach of tolerance. In recent years, studies focused on the influences of costimulation and cytokines on immune responses in general, and autoimmunity in particular, have led to a better appreciation of the factors that contribute to a sustained and effective response. Several such studies also indicate that qualities of the peptide ligand itself can significantly impact the quality of initiated or existing responses (4, 5, 6, 7), by providing either optimal or suboptimal stimuli that ultimately affect T cell functions. These studies also extend the relevance of peptides acting in the capacity of APLs³ (8), suggesting that the activities described for APLs in vitro (ranging from superagonists to antagonists) might be harnessed to gain control over immune responses in vivo. Although the majority of such studies has involved CD4⁺ T cells, recent studies indicate that parallels exist for CD8⁺ cells as well (5, 9). These latter reports in particular highlight the critical differences that may exist among CTL responses obtained with different, but related, immunogenic agonists.

Such studies suggest the potential for finding highly optimized peptides that might positively modulate CTL autoreactivity as it occurs in melanoma patients, thereby better exploiting these responses for therapeutic purposes. In a previous study, we described a singly substituted analogue of the MART1_27–35 peptide (AAGIGILTV, native peptide; LAGIGILTV, variant “1L”) that showed heteroclitic activity toward the CTL raised to the native MART1_27–35 peptide (10). We view heteroclitic activity, as do others (5), as applying to peptides of which the dose-response curves in functional assays are dramatically left-shifted relative to those of the native peptide, without a commensurate gain in binding affinity for HLA class I. Here, we report that the 1L superagonist can generate CTLs with quantitatively and qualitatively enhanced recognition of melanoma cells expressing the native peptide. These results suggest that a superagonist peptide might be used clinically to exploit a latent capacity of the T cell repertoire to respond more aggressively to the MART1_27–35 self epitope.

Anti-MART1 and 1L T cells were generated from peripheral blood of HLA-A2.1⁺ melanoma patients, as previously described in detail (11). Briefly, lymphocytes were stimulated weekly with autologous irradiated PBMCs pulsed with 1 μm peptide. Exogenous IL-2 (30–60 IU/ml) was added 1 day after restimulation. Before being used as effectors in functional assays, bulk cultures were depleted of CD4⁺ cells by incubation with Mab-coated magnetic beads (Dynal Dynabeads; Oxoid, Oslo, Norway). A42 is a MART1_27–35-reactive CTL clone isolated from tumor infiltrating lymphocytes, as described previously (12); it was maintained in culture with 120 IU/ml IL-2. The HLA-A2.1⁺ melanoma line 501 mel is autologous to clone A42. T2 is a peptide transport-deficient cell line that effectively presents exogenously supplied peptides in the context of HLA-A2.1.

Peptides were synthesized as described previously (13). MART1_27–35, also referred to in the text as native peptide, has the sequence AAGIGILTV and derives from the melanosomal protein MART1. Mono-substituted variants of MART1_27–35 used in this study carry Leu or Glu at position 1 (P1) of the peptide (designated 1L and 1E, respectively), Val at P6 (6V), and Ser at P8 (8S). The sequences and origins (shown in parentheses) of additional human protein-derived peptides used are as follows: (a) LIGLGVISI (renal sodium/phosphate transporter); (b) LAGLGLLVI (signal sequence receptor, a-subunit); (c) LLVAGVLVL (CD8b); (d) LLGIILLVL (IFN-induced monokine); (e) VMGLGVLLL (angiogenin); (f) AAGLSLLTL (E-selectin); (g) SLGLGLLPV (G protein coupled receptor); (h) AVGIGIAVV (CD9); (i) IGGIGTPVP (translation elongation factor); (j) LVVLGLLAV (glutamyl transferase); (k) ALGLGLLPV (G protein-coupled receptor); (l) AIVIGILIA (glucose transporter); (m) AVVIGIIIV (glutamate transporter); (n) LLIGIGLVL (bone marrow stromal antigen-2). Also used as a negative control was the Flu A matrix protein M1 peptide (designated “M1Flu” in figures; residues 58–66, GILGFVFTL).

Cytolytic activity was measured in a standard 4-h ⁵¹Cr release assay using target cells treated as described in each figure. Specific lysis was calculated from [(experimental cpm-nonspecific cpm)/(total cpm-nonspecific cpm)] × 100%, where total cpm was determined from detergent-releasable counts, and nonspecific cpm were the counts released in the absence of added CTLs. HLA-A2 blocking of T-cell lysis was performed by preincubating target cells with the anti-HLA-A2 Mab CR11.351. For cytokine release immunoassay, 1 × 10⁵ target cells were incubated in the absence or presence of different peptides in round-bottomed microtiter plates for 2 h at 37°C. Lymphocytes (1 × 10⁵) were added in 150 μl/well (total volume, 250 μl/well) and incubated overnight (20 h). Plates were then centrifuged, and 200-μl aliquots of supernatant were removed from each well and stored at −80°C until assayed; for cytokine detection, 50 μl (for IFNγ) or 100 μl (for IL-2) were used. Immunoassays were done using ELISA kits for IFNγ (MABTECH, Nacka, Sweden) and IL-2 (Quantikine; R&D Systems, Minneapolis, MN). In the antagonist assays, T2 cells were incubated for 2 h at 37°C with suboptimal concentrations of the MART1_27–35 peptide (i.e., 30 ng/ml for T cells raised with MART1_27–35 peptide and 30 pg/ml for T cells generated with 1L peptide), using for both lymphocyte cultures an E:T ratio of 2.5:1. T2 cells were then washed and plated into microwells containing different concentrations of the peptides being tested for antagonism. After incubation for 1 h at 37°C, CTLs were added, and the standard cytolysis assay was carried out as described above. Baseline cytolysis under these conditions typically resulted in a 30–40% specific ⁵¹Cr release. Data are reported as percentages of lysis inhibition, calculated as follows: [100 - (Lysis in the absence of antagonist peptide): (% lysis in the presence of the antagonist peptide) × 100].

Total RNA from T lymphocytes was extracted by using the RNAzolB (Cinna/Biotecs, Friend-Sword, TX) Single-stranded cDNA synthesis was carried out on 2 μg of total RNA with oligo-dT and Moloney murine leukemia virus-derived reverse transcriptase without RNase H activity (MMLV RT RNaseH-, Superscript; Life Technologies, Gaithersburg, MD), according to the manufacturer’s instructions. cDNA was suspended in 40 μl of water. Quantitative assessment of the TCRBV gene usage among peptide-specific CTL lines was carried out with PCR, as previously described (13). To verify that under our PCR conditions the amplified product was proportional to the amount of the target template in the original sample, each cDNA was serially diluted (2-fold dilution up to 1:128), and each dilution was amplified by PCR with TCRBC-specific primers. The amplification was performed in 25 μl of reaction mixture in the presence of 1 μl of the diluted cDNA, 200 μm of each dNTPs, 1 μm of each primer, and 0.625 units of Taq polymerase (Ampli Taq). By a DNA thermal cycler (Perkin-Elmer Corp., Emeryville, CA), 25 cycles of amplification were performed under the following conditions: 95°C, 30-s denaturation; 60°C, 30-s annealing; and 72°C, 1-min extension. Negative controls were included with no cDNA in the mixture. PCR reactions (10 μl) were analyzed by Southern blot using a C β oligonucleotide probe. Autoradiographs were scanned by a computer-assisted imaging system (Molecular Dynamics, Sunnyvale, CA), and the existing linearity between PCR-amplified products and cDNA dilution was verified for each sample used in this study. The same amount of β-specific cDNA template was then used for all of the samples used in the present study. PCR analysis of the TCRBV repertoire was conducted using a panel of described oligonucleotide primers (14). The level of TCRBV-specific amplification was measured by a densitometric scanning (Eagle-eye; Stratagene) and computer-assisted analysis (Eagle Sight; Stratagene) of ethidium bromide-stained gel loaded with 10 μl of each amplification product. Each TCRBV was expressed as a percentage of the sum of all TCRBV signals detected. Repertoire determination was performed at least three times for each sample.

We previously identified a synthetic MART1_27–35 analogue, containing a Leu at P1 (LAGIGILTV; “1L”), that shows heteroclitic behavior with a variety of anti-MART1_27–35 effectors (10). To confirm this previous observation, we tested recognition (assessed as cytotoxic activity) by the MART1_27–35-specific clone A42 (12) of different singly substituted MART1_27–35 analogues. As shown in Fig. 1 A, 1L peptide is the most efficiently recognized among several variant peptides of which the dose-response curves are left-shifted relative to that of the native peptide. The increased responsiveness to the 1L peptide cannot be attributed to enhanced class I binding, because it binds to HLA-A2.1 with an affinity similar to that of the native peptide (10).

Given the strong recognition of the 1L analogue by anti-MART1 CTL, we evaluated whether this heteroclitic epitope could be more efficient in the generation of specific T cells able to cross-react with the native peptide MART1_27–35. Peripheral blood lymphocytes from HLA-A2⁺ melanoma patients were stimulated in vitro either with the 1L analogue or with the native peptide, according to a protocol previously described (11). After 2 weeks of culture, T cells were tested for cytotoxic activity against peptides endogenously presented by HLA-A2⁺ melanoma cells (501mel) or exogenously pulsed onto T2 cells. As reported in Fig. 1,B, 1L resulted in the generation of cytotoxic T cells more effective in mediating lysis of MART1_27–35-expressing targets (either peptide-pulsed T2 or 501mel) than the T cells raised with the native peptide. In two out of five patients (1 and 4), specific cytotoxic T cells were generated only with 1L, and no significant activity was obtained using the native epitope. Background cytolysis of T2 pulsed with the irrelevant peptide from Flu matrix M1Flu was, in all cases, tested below 5%, supporting the fine antigen specificity of 1L-generated T cells. Lysis by the anti-MART1/Melan A clone A42 is also reported in Fig. 1 B as a positive control. Cytolytic activity of anti-1L T cells on 501mel was mediated by a specific recognition of HLA-A2/MART1 complex because it could be blocked by an anti-HLA-A2 Mab and no significant killing of the 501 lymphoblastoid cell line could ever be detected (data not shown). No lysis of autologous lymphoblastoid cell lines was observed either, thus confirming the absence of detectable autoreactivity toward antigens other than MART1 (data not shown).

Peptide titrations were done to assess how efficiently anti-1L CTLs recognized the native peptide. As shown in Fig. 2,A, 1L-specific T cells responded to significantly lower concentrations of the MART1_27–35 peptide than those detected by anti-native T cells. Anti-1L T cells were also able to mediate effective melanoma lysis at lower E:T ratios compared with CTLs raised to cognate peptide (Fig. 2 B).

Specific CTLs exert several immunological functions in addition to cytotoxic activity against antigen-bearing targets. For a more extensive characterization of anti-1L T cell functions, IFNγ release in response to the parental MART1 peptide (exogenously presented by T2 cells or endogenously expressed by the melanoma 501mel) was evaluated at different weeks of in vitro culture. As reported in Fig. 2, C and D, when effectors were stimulated by the T2 plus native peptide and 501mel, respectively, all of the anti-1L T cells tested showed a higher and earlier production of IFNγ than T cells raised with the cognate peptide. Significant recognition of the HLA-A2⁺/MART1⁺ melanoma cells (501mel) could also be detected in all cases tested.

IL-2 production has been shown recently to be a feature not only of CD4⁺ helper T cells, but also of fully activated CD8⁺ cells (15). Impaired IL-2 secretion also has been associated with decreased functional efficacy of CD8⁺ T cells (5, 16). In our system, only anti-1L T cells were able to secrete IL-2 in response to the MART1_27–35 cognate peptide, either exogenously presented on T2 cells or endogenously expressed by HLA-A2⁺ melanoma cells (501mel; Fig. 3 A). No significant IL-2 secretion in response to melanoma cells could be observed with anti-native T cells, which were found in one case (patient 5) to release barely detectable levels of IL-2, only in response to peptide-pulsed T2 cells.

We previously have shown that several natural analogues (peptide sequences contained within endogenous human proteins) of MART1_27–35 act as partial agonists or antagonists for anti-MART1_27–35 T cells (10, 13). Under conditions typically used to assay antagonist activity, we noted that several of these analogues effectively inhibited CTL lysis of target cells pulsed with a suboptimal concentration of native MART1/Melan A peptide. However, we also observed that these same antagonist peptides, tested over a broad concentration range, could not inhibit lysis of target cells when suboptimally pulsed with the 1L variant peptide (10). To evaluate whether CTLs raised with 1L analogue could show a comparable insensitivity to antagonist peptides, anti-1L T cells were tested in an antagonist assay with target cells suboptimally pulsed with the MART1/Melan A_27–35 peptide, using a panel of natural analogues, which included peptides previously observed to be antagonists with other anti-MART1/Melan A_27–35 CTLs (10). As shown in Fig. 3, B and C, 6 of 14 tested peptides inhibited lysis of MART1/Melan A _27–35-pulsed targets by CTLs raised against the native peptide. The anti-native peptide effector populations derived from patients 2 and 3 were each antagonized by different, though partially overlapping, subsets of peptides, most likely reflecting differences in the composition of their responding repertoires, as further suggested by the TCRVβ analysis presented below. On the contrary, no significant antagonism was observed when anti-1L T cells were used (Fig. 3, B and C), thus confirming the improved immune functions of T cells generated with a superagonist variant.

To evaluate whether the different functional behavior of T cells generated with the 1L analogue or with the native peptide could be correlated to an in vitro selection of a different TCR repertoire, the TCRBV families expressed by the two specific T cell populations were analyzed by a semiquantitative reverse transcription-PCR. Each repertoire was compared with the TCRBV expressed by unstimulated CD8⁺ fresh lymphocytes. Fig. 4 reports data obtained for patients 2 and 5 for whom unstimulated PBMCs were available. Comparing the TCRBV repertoire of fresh CD8⁺ T cells with the corresponding cell line obtained after specific in vitro stimulation with 1L or native peptide, a different distribution of TCRBV was observed, suggesting that an antigen-driven selection occurred during the in vitro culture. In addition, inside each patient, the TCRBV repertoire was differently affected by stimulation with 1L or with the native peptide, indicating that the final effector T cells greatly differs for their general TCR composition. Comparable data were obtained in patient 3 (data not shown). Additionally, in both patients reported here (Fig. 4, A and B), a preferential expansion of TCRBV13 and 14 could be observed in 1L-stimulated T lymphocytes, whereas no common usage was detectable for T cells generated with the cognate peptide.

T lymphocytes have been shown to have no single ligand specificity, but to recognize a large array of peptides, termed APLs (8). These APLs can mediate a number of different outcomes in interacting T cells, ranging from inducing selective immunological functions (partial agonist) to completely turning off their functional capacity (antagonist). In addition to suboptimal ligands, TCR can also be triggered by superagonists, which are analogues that enhance T-cell stimulation by inducing immunological functions not detected with the cognate ligand (4, 5, 6, 7).

We recently showed that a heteroclitic variant of the melanoma-melanocyte-derived epitope MART1/Melan A_27–35, containing a Leu instead of an Ala at P1 (LAGIGILTV; 1L), could hyperstimulate MART1/Melan A-specific T cells (10). Here, we reported that this analogue, which binds to MHC with affinity comparable with the native peptide, displays superagonist activity by promoting the generation, in HLA-A2⁺ melanoma patients, of MART1/Melan A-specific T cells with enhanced immunological functions as compared with cells raised with the native epitope. The superagonist activity of 1L can be detected as an increase in the magnitude of T cell responses to the original epitope (either exogenously pulsed on T2 cells or endogenously expressed by melanoma cells); 1L-raised T cells indeed lyse with higher efficiency MART1/Melan A_27–35-expressing targets and release in their presence cytokines such as IFNγ earlier and to a higher extent than T cells generated with the original epitope. Furthermore, these T cells exert half-maximal effector responses at concentrations of the MART1/Melan A_27–35 peptide significantly lower than that required to sensitize lymphocytes raised with the cognate epitope.

In addition to these quantitative modifications, 1L superagonist can induce qualitatively new responses, driving T cells to further secrete IL-2 in response to tumor cells (not detected in T cells generated with the native peptide), and to undergo a faster and different clonal expansion. TCR repertoire analysis shows, in fact, that 1L expands a different and more restricted panel of Vβ families than that obtained when the native peptide is used.

The magnitude of IL-2 production has been shown to be critical for determining the overall extent of the immune responses in both CD4⁺ and CD8⁺ T cell subpopulations (4, 15), whereas its lack is considered a key factor in mediating T cell anergy (4, 15, 16, 17). The evidence that T cells generated with the 1L superagonist produce IL-2 in response to the MART1/Melan A_27–35 peptide expressed on tumor cells, a feature undetectable with lymphocytes raised in the presence of the native epitope, represents in our opinion an important observation that may significantly change the therapeutic potential of the MART1_27–35 peptide. Anti-melanoma T lymphocytes have been, in fact, described as defective in IL-2 production when interacting with tumor cells, and this defect associated with the lack of efficiency to control tumor growth in vivo (18). Moreover, data from our laboratory have recently shown that IL-2 production is one of the first functions that T cells lose when interacting with partial agonists or antagonists (10, 13). The possibility of stimulating, by a superagonist variant, a T cell population potentially able to substain its own growth, once interacting with relevant antigen, and contribute in activating bystander cells should create a more favorable milieu for T-cell activity.

A further factor supporting the ability of T cells generated with 1L to better interact with their HLA-A2/MART1_27–35 ligand is their refractoriness, at least in vitro, to the antagonist effect mediated by MART1/Melan A_27–35 natural analogues. These analogues have been shown on the contrary to induce a strong inhibitory effect on T cells raised with the native peptide (10), therefore, supporting the hypothesis that anti-MART1/Melan A-reactive T cells may be negatively regulated in vivo by encountering antagonist epitopes (3).

Given the lack of evidence that tumor-specific T cells can significantly impact tumor growth in vivo and the limited results obtained with the first clinical protocols (19), new strategies for increasing the immunogenicity of tumor-derived peptides and to potentially improve their therapeutic efficacy have to be found. Thus far, the most commonly attempted approach has been to increase HLA binding affinity of native peptides by introducing conservative substitutions at HLA-binding anchor positions, given the correlation observed with viral antigens between immunogenicity and MHC peptide-binding affinity (20). Melanoma-derived modified peptides, the HLA binding affinity/stability of which is higher than that displayed by the native epitopes, were found to be significantly more immunogenic i.e., more efficient than the cognate peptide in inducing cytotoxic T cells to gp100 (21) and MART1-derived HLA-A2.1 binding epitopes (22). The strategy of modifying natural tumor epitope to improve their immunogenicity has been also recently tested in clinics with promising results. In fact, Rosenberg et al. (23) have shown that vaccination with a modified analogue of an HLA-A2-binding peptide of the melanoma antigen gp100 can mediate tumor regression in 42% of patients with metastatic melanoma, a success rate not achieved in the previous studies with natural gp100 epitopes (23). Of note is the fact that clinical responses could only be observed when IL-2 was administered along with the vaccine. According to the authors, this observation suggests that exogenous IL-2 may help overcoming the state of anergy possibly induced in T cells by the interaction with the tumor antigen in the absence of costimulation, evidence that has been also proved in murine models (24). These latter in vivo data confirm the central role of IL-2 in T cell-mediated antitumor responses and further highlight the clinical potential of a vaccine based on the use of superagonist peptides, which could mediate in situ IL-2 production by specific T cells once interacting with the relevant tumor antigen.

Tumor immunotherapy is, perhaps, the one clinical setting in which the induction of an autoaggressive cellular response seems to be desirable. The superagonist variant of MART1_27–35 that we describe extends the concept of using modified self peptide immunogens to achieve this goal. Although modifications that improve peptide binding and complex stability offer demonstrable advantages, a key issue is the overall quality of the CTL response induced. The data presented here suggest that the 1L variant is capable of optimally inducing CTL in vitro that subsequently are capable of responding to low levels MART1_27–35 with enhanced production of IFN-γ and IL-2, in addition to cytolysis, and reduced susceptibility to negative regulatory mechanisms. The strategy of identifying self-peptide analogues with bona fide superagonist activity offers the potential of improving in vivo, peptide-based tumor immunotherapy.

We are grateful to Drs. Yongsheng Song and Ettore Appella (National Cancer Institute/NIH, Bethesda, MD) for peptide sythesis. We also thank Drs. Claudia Lombardo and Flavio Arienti of our Institute for HLA typing and buffy coats of melanoma patients.