Generation of CTL Recognizing an HLA-A*0201-Restricted Epitope Shared by MAGE-A1, -A2, -A3, -A4, -A6, -A10, and -A12 Tumor Antigens: Implication in a Broad-Spectrum Tumor Immunotherapy¹

The identification of tumor Ags (TA)³ and their recognition by tumor-specific CTL has fuelled the development of immunotherapeutic strategies in cancer. Although numerous TA and their epitopes have been identified, particularly in melanoma (1, 2), the majority of these are quite restricted in expression and their clinical utility remains limited. Therefore, it is imperative to evaluate additional TA/epitopes that are widely expressed in cancer to develop clinically meaningful immune intervention for most cancer patients regardless of tumor type. The p53 and the recently described human telomerase reverse transcriptase (TERT) are classical examples of widely expressed TA (3, 4). However, the use of these TA for tumor immunotherapy raises two major problems related to their expression by normal tissues, including thymus (5). First, their specific CTL repertoire is composed of low-avidity T cells with weak tumor-killing activity that have escaped tolerance, as demonstrated for p53 (6, 7). Second, targeting widely expressed TA can induce autoimmunity that, in some cases, leads to the destruction of essential normal cells and tissues. We have recently obtained evidence indicating that vaccination with dominant TERT epitopes induces autoimmunity that targets TERT-expressing activated B cells (our unpublished observations).

MAGE-A is a multigene family that consists of 12 homologous genes (MAGE-A1 to -A12) located in the q28 region of chromosome X (8). Among the different family members, MAGE-A1, -A2, -A3, -A4, -A6, -A10, and -A12 are abundantly expressed by tumors but not by normal tissues except testis and placenta (9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23). Each one of the seven MAGE-A Ags is detected in primary and metastatic tumors of various histological types including melanoma, lung, bladder, ovarian, and breast carcinomas. Individual MAGE-A expression varies from one tumor type to the other, but overall the large majority of tumors expresses at least one MAGE-A. Targeting epitopes shared by all MAGE-A Ags would be of interest against a broad spectrum of cancers. Moreover, the use of these epitopes will not come up against the tolerance of their CTL repertoire and the risks of autoimmunity. MAGE-A do not belong to the immunological self and the MAGE-A-expressing testis is considered as an immunoprivileged organ and cannot be the target of an autoimmune response.

To find common MAGE-A epitopes susceptible to induce MAGE-A-specific HLA-A*0201-restricted CTL, we first selected peptides likely to share the same antigenicity, even if they lack the HLA-A*0201-specific anchor residues. Sequence modifications were introduced in some of these peptides to improve their binding capacity and their immunogenicity.

In this work we describe one high-affinity heteroclitic peptide (p248V9) corresponding to two HLA-A*0201-restricted, cross-recognized epitopes, derived from MAGE-A2, -A3, -A4, -A6, -A10, -A12 (p248G9) and MAGE-A1 (p248D9). This heteroclitic peptide stimulates CTL that recognize each MAGE-A Ag individually and kill MAGE-A-expressing tumor cells. Therefore, it may be implemented for a broad-spectrum, efficient, but nonetheless safe, vaccination.

The HLA-A*0201 transgenic HHD mice were previously described (24).

Murine RMAS/HHD cells were obtained by transfection of murine RMAS cells with the HHD construct as previously described (24). COS-7 cells and WEHI-164 clone 13 cells were kindly provided by Dr. F. Jotereau (Institut National de la Santé et de la Recherche Médicale, Unité 463, Nantes, France). The HLA-A*0201-expressing human tumor cells were T2 (deficient in TAP1 and TAP2 transporters); M44 and M113 (melanoma; kindly provided by Dr. F. Jotereau), OBR (bladder cancer; kindly provided by Dr. D. Zeliszewsli, Centre National de la Recherche Scientifique UPRES-A8067, Paris, France), MCF-7 (breast cancer), and Caco-2 (colon cancer). Cells were grown in RPMI 1640 medium supplemented with 10% FCS.

Peptides were synthesized by Synt:em (Nı̂mes, France). Plasmids containing the cDNA for MAGE-A (pcD-Sra/MAGE-A1, pcD-Sra/MAGE-A2, pcDNAI/MAGE-A3, pcDNAI/MAGE-A4, pcDNAI/MAGE-A6, pcDNAI/MAGE-A10, and pcDNAI/MAGE-A12) were kindly provided by Dr. P. van der Bruggen (Ludwig Institute for Cancer Research, Brussels, Belgium). The plasmid containing the HHD construct was described previously (24).

HHD mice were injected s.c. with 100 μg of peptide emulsified in IFA in the presence of 140 μg of the I-Ab-restricted HBV core-derived Th epitope (128–140: TPPAYRPPNAPIL). After 11 days, 5 × 10⁷ spleen cells were stimulated in vitro with peptide (10 μM) in RPMI 1640 medium supplemented with 10% FCS, 2 μM glutamine, and antibiotics. On day 6 of culture, the bulk responder populations were tested for specific cytotoxicity. Upon response, CTL cultures were restimulated weekly in vitro with 2 × 10⁷ irradiated spleen cells in the presence of 1 to 0.1 μM peptide and 50 U/ml rIL-2 (Proleukin; Chiron, Suresnes, France).

PBMC were collected by leukapheresis from healthy HLA-A*0201 donors. Dendritic cells (DC) were produced from adherent cells cultured for 7 days (2 × 10⁶ cells/ml) in the presence of 500 IU/ml GM-CSF (Leucomax; Schering-Plough, Levallois Perret, France) and 500 IU/ml IL-4 (R&D Systems, Minneapolis, MN) in complete medium (RPMI 1640 supplemented with 10% heat inactivated human AB serum, 2 μM l-glutamine, and antibiotics). On day 7 maturation agents poly(I:C) at 100 ng/ml and anti-CD40 mAb (kindly provided by Dr. J. P. Abastado, Immuno-Designed Molecules, Paris, France) at 2 μg/ml were added in the culture for 24 h. Mature DC were pulsed with 10 μM peptide in the presence of 5 μg/ml β₂-microglobulin for 2 h at 37°C and then irradiated (3500 rad). CD8⁺ cells were purified by positive selection with CD8 MicroBeads (Miltenyi Biotec, Paris, France) according to the manufacturer’s instructions. A total of 2 × 10⁵ CD8⁺ cells were stimulated with 2 × 10⁴ peptide-pulsed DC in complete culture medium supplemented with 1000 IU/ml IL-6 (R&D Systems) and 5 IU/ml IL-12 (R&D Systems) in a final volume of 100 μl/well in 96-well round-bottom plates. From day 7, cultures were weekly restimulated with peptide-loaded DC in the presence of 20 IU/ml IL-2 and 10 ng/ml IL-7 (R&D Systems). After the third in vitro restimulation, CD8 cells were collected and stimulated with allogeneic HLA-A*0201 EBV-transformed B cells in the presence of 10 μM peptide for 16 h. IFN-γ-producing CD8⁺ cells were purified using the IFN-γ secretion assay cell enrichment and detection kit (Miltenyi Biotec) according to the manufacturer’s instructions. Purified CD8⁺ cells were then cultivated in complete medium supplemented with 20 IU/ml IL-2 and 10 ng/ml IL-7 for 1 wk.

Targets were labeled with 100 μCi of ⁵¹Cr for 90 min, washed twice, and plated in 96-well round-bottom plates (3 × 10³ cells/well in 100 μl of RPMI 1640 plus 5% FCS). When RMAS/HHD cells were used as targets they were pulsed with various concentrations of peptides at 37°C for 90 min. A total of 100 μl of effectors were then added to the wells. After a 4-h incubation, 100 μl of supernatant were collected and radioactivity was measured in a gamma counter. The percentage of specific lysis was determined as follows: lysis = (experimental release − spontaneous release)/(maximal release − spontaneous release) × 100.

T cells were stimulated by T2 cells loaded with peptide (10 μM) and by tumor cells in the presence of 20 μg/ml brefeldin A (Sigma, Oakville, Canada). In blocking experiments tumor cells were preincubated with an anti-HLA-A*0201 (BB7.2) and an anti-HLA-B/C (B1.23.2) mAb for 1 h. Six hours later T cells were washed, stained with PE-conjugated anti-CD8 mAb (Caltag Laboratories, Burlingame, CA) in PBS for 25 min at 4°C, washed, and fixed with 4% paraformaldehyde. Then, cells were permeabilized with PBS/0.2% saponin/0.5% BSA (Sigma) and stained with allophycocyanin-conjugated anti-IFN-γ mAb (BD PharMingen, Mississauga, Canada). Cells were analyzed on FACSCalibur (BD Biosciences, Mountain View, CA).

A total of 2.2 × 10⁴ simian COS-7 cells were plated in flat-bottom 96-well plates in DMEM 10% FCS, in triplicate per condition. Eighteen hours later, the medium was discarded and 100 ng of each DNA plasmid was put in contact with COS cells in DMEM, 10% Nuserum (Collaborative Biochemical Products, Belford, MA), 10 mM chloroquine, and 10 mg/ml DEAE dextran. After a 4-h incubation at 37°C, transfection medium was discarded and 50 μl PBS 10% DMSO was added for 2 min. Transfected COS-7 cells were incubated in DMEM 10% FCS during 40 h and then used with 5 × 10⁴ murine CTL in a TNF-α secretion assay.

Transfected COS-7 cells at day 4 and human tumor cells were used as stimulating cells. When necessary, they were incubated with 10 μM peptide for 2 h. A total of 5 × 10⁴ T cells were then added in 50 μl RPMI and 10% FCS and incubated for 6 h. Each condition was tested in triplicate. A total of 50 μl of the supernatant was collected. Standard dilutions were prepared in 50 μl with final doses of TNF-α ranging from 10⁴ to 0 pg/ml. On both the supernatants and the standard dilutions, 3 × 10⁴ TNF-α sensitive WEHI-164c13 cells in 50 μl were added. They were incubated for 16 h at 37°C. Inhibition of cell proliferation was evaluated by the MTT colorimetric method (25).

Isolation of total RNA from tumor cells was performed using TRIzol (Invitrogen, Cergy Pontoise, France) reagent. For cDNA synthesis, RNA (2 μg) was diluted in water, 4 μl of 5× reverse transcriptase buffer (Invitrogen), 1 μl of 10 mM dNTP, 2 μl of a 20 μM solution of oligo(dT), 20 U of RNasin (Promega, Madison, WI), 2 μl of 0.1 M DTT, and 200 U of Moloney murine leukemia virus reverse transcriptase (Invitrogen) in a 20-μl reaction volume, and incubated at 42°C for 60 min. PCR amplification was performed with DNA polymerase TaKaRa Taq (Takara Biomedicals, Shiga, Japan) using pairs of oligonucleotide primers that are highly specific for each MAGE-A gene (8, 26, 27, 28). To ensure that the RNA is not degraded, a PCR assay with primers specific for β-actin was conducted. Samples were scored positive when a band of appropriate size was visible on an agarose gel in the presence of ethidium bromide.

The protocol used has been described previously (29). Briefly, T2 cells were incubated with various concentrations of peptides ranging from 100 to 0.1 μM at 37°C for 16 h, and then stained with the BB7.2 mAb to quantify the expression of HLA-A*0201. For each peptide concentration, the HLA-A*0201-specific staining was calculated as the percentage of staining obtained with 100 μM of the reference peptide HIVpol 589 (IVGAETFYV). The relative affinity (RA) was determined as follows: RA = (concentration of each peptide that induces 20% of HLA-A*0201 expression/concentration of the reference peptide that induces 20% of HLA-A*0201 expression).

As previously described (29), T2 cells were incubated overnight with 100 μM of each peptide at 37°C in serum-free medium. Cells were washed, incubated with brefeldin A (10 μg/ml) for 1 h, washed again, and incubated at 37°C during 0, 2, 4, or 6 h in the presence of brefeldin A at 0.5 μg/ml. Cells were stained with the BB7.2 mAb. Dissociation complex (DC₅₀) was defined as the time required for the loss of 50% of the HLA-A*0201/peptide complexes stabilized at t = 0 h.

The MAGE-A1, -A2, -A3, -A4, -A6, -A10, and -A12 are expressed in a large variety of tumors with a frequency ranging from 5 to 76% (Table I). However, a very high percentage of tumors express at least one MAGE-A.

The aim of this work was to identify epitopes belonging to different MAGE-A and sharing the same antigenicity. The antigenicity of HLA-A*0201-bound peptides is defined by their sequence that interacts with the TCR and is located between the P3 and P8/9. We aligned the sequences of MAGE-A1, -A2, -A3, -A4, -A6, -A10, and -A12 to find nonamers and decamers sharing the P3–P8 and P3–P9 regions, respectively. Four groups of peptides were identified, and their sequences and the MAGE-A they belong to are shown in Table II. Peptides were designed by the position of the first residue in the MAGE-A1 sequence. p262 had a complete sequence homology in all MAGE-A, whereas p174 was identical in all MAGE-A Ags except MAGE-A10. The homology group p248 was composed by two peptides, p248G9 (MAGE-A2, -A3, -A4, -A6, -A10, and -A12) and p248D9 (MAGE-A1), differing at the C-terminal position. The fourth group (p264) was excluded from this study because it corresponds to the already known HLA-A*0201-restricted MAGE-A3 epitope (FLWGPRALV) that is reported to be not efficiently processed by tumor cells (30).

We then evaluated the capacity of selected peptides to bind (RA) and to stabilize (DC₅₀) the HLA-A*0201 molecule (Table II). As we reported previously, these parameters allow for distinguishing high-affinity peptides (RA < 5 and DC₅₀ > 2 h) that are immunogenic in both HHD mice and humans from low-affinity peptides (RA > 5 and DC₅₀ < 2 h) that are nonimmunogenic (29). According to these criteria, peptides p174, p262, p248G9, and p248D9 had a low affinity for HLA-A*0201. To enhance affinity of p174, p262, p248G9, and p248D9 and thus increase their immunogenicity, we modified their sequences at HLA-A*0201-specific primary (P2 and P9/10) and/or secondary anchor positions. p262, p248G9, and p248D9 lacked the anchor residues at P2 or C-terminal position. The heteroclitic variants p262L2 and p248V9 exhibited high binding affinity (RA = 0.2 and 1.8, respectively) and stabilization capacity (DC₅₀ = 6 and 4 h, respectively). For the p174 we produced the p174Y1V10 variant by replacing the residue at P1 with a tyrosine as previously described (29) and the residue at P10 with the strongest anchor valine. P174Y1V10 was a strong binder (RA = 2.5) but weakly stabilized the HLA-A*0201 molecule (DC₅₀ < 2 h).

To study whether heteroclitic peptides can induce CTL able to recognize the corresponding native peptides, we vaccinated HHD mice with heteroclitic peptides, stimulated their spleen cells with 10 μM peptide in vitro, and tested cultures for killing RMAS/HHD cells loaded with the heteroclitic or the native peptides (Fig. 1,A). p174Y1V10 was unable to elicit a CTL response, probably because of its weak stabilization capacity. CTL generated in p262L2-vaccinated mice recognized the heteroclitic variant but not the native p262. In contrast, p248V9-vaccinated mice developed a CTL response directed against the heteroclitic but also against the native p248G9 and p248D9 peptides. We established a CTL line (CTL248) from HHD spleen cells primed in vivo with p248V9 and in vitro repetitively restimulated with decreasing concentrations of p248G9 (10 to 1 μM). CTL248 was maintained in culture with 1 μM p248G9 and recognized the p248G9 and p248D9 with a relatively high avidity (Fig. 1 B).

To test whether p248V9-induced CTL recognize endogenously processed MAGE-A1, -A2, -A3, -A4, -A6, -A10, and -A12, we first stimulated the CTL248 line with COS cells cotransfected with cDNA encoding the HHD construct (α₁ and α₂ domains of HLA-A*0201 and the α₃ and intracellular domains of D^b) and each one of the seven MAGE-A. CTL activation was evaluated by measure of TNF-α secretion. Results show that the CTL248 line responded to COS cells coexpressing HHD and any of the MAGE-A tested but not to COS cells expressing HHD or MAGE-A separately (Fig. 2). This demonstrates that CTL248 recognize all the endogenous MAGE-A presented by HLA-A*0201.

In a second set of experiments, we tested whether CTL248 recognize HLA-A*0201 and MAGE-A-expressing human tumor cells. We stimulated CTL248 with the M44 (MAGE-A1, -A2, -A3, and -A4), M113 (MAGE-A2 and -A3), OBR (MAGE-A6 and -A10), and the MAGE-A-negative MCF-7 and Caco-2 tumor cell lines. All these lines were HLA-A*0201 positive and the MAGE-A expression was assessed by RT-PCR. CTL activation was evaluated by measure of secreted TNF-α. CTL248 responded to all the MAGE-A⁺ cells but not to MCF-7 and Caco-2 cells (Fig. 3). These results were further confirmed by measuring the number of IFN-γ-producing CTL248 cells upon stimulation (data not shown). To demonstrate that tumor cell recognition was HLA-A*0201 restricted, we stimulated CTL248 with M44 cells in the presence of an anti-HLA-A*0201 (BB7.2) and an anti-HLA-B/C (B1.23.2) mAb. CTL248 response was inhibited by the BB7.2 but not by the B1.23.2 mAb (Fig. 3 B).

Taken together these data show that p248V9-induced CTL recognize endogenously processed MAGE-A1, -A2, -A3, -A4, -A6, -A10, and -A12.

We then assessed the capacity of p248V9 to generate CTL in vitro from HLA-A*0201 healthy donors’ PBMC. CTL were induced by three weekly in vitro stimulations with the p248V9. IFN-γ-producing cells upon peptide stimulation were isolated and cultured for 7–10 days before testing. CTL were generated in seven of nine healthy donors tested. Results from one donor are presented in Fig. 4. CTL responded to T2 cells loaded with p248V9, p248G9, and p248D9 but not to T2 cells loaded with an irrelevant peptide (HIVgag₇₆) (Fig. 4,A). They responded also to HLA-A*0201⁺MAGE-A⁺ M44 and M113 but not to the HLA-A*0201⁺MAGE-A⁻ Caco-2 cells (Fig. 4,B). We verified the HLA-A*0201 restriction of M44 cell recognition in an anti-HLA-A*0201 Ab blocking experiment. The response of CTL to M44 cells was blocked by the HLA-A*0201-specific BB7.2 Ab but not by the HLA-B/C-specific B1.23.2 Ab (Fig. 4 C).

These results demonstrate that the heteroclitic p248V9 stimulates human CTL that recognize MAGE-A-expressing tumor cells.

In this paper we describe a heteroclitic peptide (p248V9) that stimulates CTL able to recognize endogenous MAGE-A1, -A2, -A3, -A4, -A6, -A10, and -A12 Ags presented by HLA-A*0201. The p248V9 peptide corresponds to two new, low HLA-A*0201 affinity, cross-recognized cryptic epitopes derived from MAGE-A2, -A3, -A4, -A6, -A10, and -A12 (p248G9) and MAGE-A1 (p248D9). This heteroclitic epitope may have a great interest for tumor immunotherapy for several reasons. First, it may generate CTL directed against a very large variety of tumors. In fact, although MAGE-A Ags are expressed by different tumors with a frequency ranging between 5 and 75% (Table I), >80% of all tumors express at least one MAGE-A Ag and could, therefore, be targeted by the p248V9-induced CTL. Second, the use of p248V9 peptide for tumor immunotherapy will lower the risk of tumor escape due to loss of nominal Ag (31, 32). Considering that very often tumors simultaneously express several MAGE-A Ags (33), to escape p248V9-specific CTL response these tumors will have to lose all the MAGE-A they express, which is less likely. Third, because MAGE-A is not expressed in normal tissues except testis, the p248G9- and p248D9-specific CTL repertoire recruited by the heteroclitic p248V9 epitope should not have been negatively selected and the high-avidity CTL providing the strongest antitumor effect (34) should not have been eliminated. Fourth, the absence of MAGE-A in normal tissues limits the risk of autoimmunity that could be a great problem for other universal TA, like the human TERT, which is expressed in activated T and B cells and in the CD34⁺ hematopoietic progenitors (5). The three latter points underline the advantage of targeting the p248G9 and p248D9 MAGE-A epitope compared with targeting epitopes from other universal TA like human TERT or p53.

The homology between the different MAGE-A varies from 48% (MAGE-A10 vs MAGE-A12) to 95% (MAGE-A3 vs MAGE-A6). As a result, the alignment of all the MAGE-A sequences reveals only one completely identical nonameric peptide (p262). However, the complete homology is not necessary for peptides to be cross-recognized by CTL. The homology can be limited to the central segment of peptides (P3–P8/9) that contains residues contacting the TCR (35). Moreover, the presence of the HLA-A*0201-specific primary anchor motifs, which is a prerequisite for a high affinity, and immunogenicity were not necessary either. The p248G9 and p248D9 epitopes did not have an anchor motif at P9 in any of the MAGE-A sequences and exhibited a low affinity for HLA-A*0201. The substitution of the P9 residue with the anchor motif V gave an heteroclitic peptide with high affinity, strong immunogenicity, and the same antigenicity as the native peptides. The preservation of the native peptide antigenicity has been reported for several heteroclitic peptides, having substituted the anchor residues (36, 37, 38, 39). The rare exceptions of this rule must concern very nonconservative substitutions, as is the case for the p262L2, where an E is substituted by an L at P2. In conclusion, the search of cross-recognized HLA-A*0201-restricted peptides from sequences with relatively high homology can be exclusively based on their homology in the P3/C-terminal P segment. This strategy can effectively be applied to other multigenic families of TA like HER, BAGE, and GAGE to enrich the pool of universal tumor epitopes. The search for a common HER epitope is currently under way in our laboratory.

The p248G9/D9 is the unique common MAGE-A epitope presented by HLA-A*0201. The p262, which is identical to all MAGE-A sequences, is unlikely to be naturally processed because of the unfavorable residue (E) at the anchor P2. Moreover, the previously described MAGE-A3 epitope FLWGPRALV, which differs in the seven MAGE-A only at the C-terminal position, has recently been shown to be ineffectively processed by tumor cells (30). Concerning the numerous other HLA-A*0201-restricted MAGE-A epitopes, their sequences are either unique to the MAGE-A they belong to or they are shared by very few MAGE-A (40, 41, 42). One of the rare exceptions is the HLA-B*3701-restricted promiscuous epitope described by Tanzarella et al. (23). This epitope is shared by MAGE-A1, -A2, -A3, and -A6 but not by MAGE-A4, -A10, and -A12.

In conclusion, in this paper we describe an MAGE-A-derived HLA-A*0201-restricted heteroclitic peptide (p248V9) that induces CTL able to recognize each MAGE-A Ag individually and subsequently to kill MAGE-A-expressing tumor cells. This heteroclitic peptide can be used for a broad-spectrum tumor immunotherapy.

We thank Dr. P. van Der Bruggen for providing the MAGE-A encoding plasmids, Drs. F. Jotereau and D. Zeliszewsli for providing tumor cells, Dr. J. P. Abastado for providing the anti-CD40 mAb, Dr. H. Echchakir for providing pairs of oligonucleotide primers specific for each MAGE-A gene, and Dr. I. Miconnet for critical reading of the manuscript. We also thank Y. Lecluse for technical assistance in cytometry.