HLA-Independent Heterogeneity of CD8⁺ T Cell Responses to MAGE-3, Melan-A/MART-1, gp100, Tyrosinase, MC1R, and TRP-2 in Vaccine-Treated Melanoma Patients¹

An important requirement for vaccines against cancer is that they stimulate antitumor CD8⁺ T cells, as these are thought to be the major effector cells in tumor rejection (1, 2). There is an interest in finding peptides that have the ability to stimulate such responses because these peptides would be good candidates for vaccine construction. A number of tumor-derived, HLA-restricted peptides have been identified that are antigenic, i.e., recognized by CD8⁺ T cells in vitro (3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13). However, their immunogenicity remains unknown, as this requires direct evidence that they can stimulate CD8⁺ T cell responses in vivo when used to immunize patients. Recently, we developed a sensitive method to quantify peptide-specific CD8⁺ T cells in peripheral blood so that the small increase in the number of these cells induced by active immunization to Ag can be detected. Using this procedure (14), we directly detected without in vitro sensitization, vaccine-induced CD8⁺ T cell responses to two peptides derived from the melanoma-associated Ags MAGE-3_271–278 (3) and Melan-A/MART-1_27–35 (4).

We have now extended these studies by examining patients’ responses in vivo to a large number of peptides derived from Ags expressed by human melanoma cells. These include MAGE-3 (15), Melan-A/MART-1 (16), gp100 (5), tyrosinase, melanocortin receptor (MC1R)⁴ (17), and dopachrome tautomerase (TRP-2) (18, 19), all of which are known to be antigenic in vitro. The study focused on HLA-A*0201-restricted peptides, the allele most commonly expressed by patients at risk for developing melanoma.

We evaluated 22 sequential, vaccine-treated, HLA-A*02⁺ patients with resected malignant melanoma. Seven (32%) had American Joint Committee on Cancer (AJCC) stage II, nine (41%) had stage III and six (27%) had stage IV disease. HLA typing was performed by complement-mediated cytotoxicity. HLA-A2 subtyping was performed by the PCR-SSP (sequence specific primers) technique using the set containing 5′ and 3′ primers for identifying A*0201 to A*0217 alleles (Dynal, Oslo, Norway) (20).

All patients were immunized to a polyvalent melanoma vaccine prepared, as previously described, from Ags shed into culture medium by a pool of melanoma cell lines (SFM14, SFM20 and SFSKMel28) (21, 22). Briefly, the shed material was collected, concentrated, pooled, treated with 0.5% Nonidet P-40 to break up aggregates, and ultracentrifuged at 100,000 × g for 90 min to deplete alloantigens. The supernatant was filter sterilized, adjusted to a protein concentration of 200 μg/ml, dispensed into sterile glass vials, and frozen at −80°C until used. The vaccine contains MAGE-1, MAGE-3, Melan-A/MART-1, and tyrosinase by Western blotting (14), gp-100 by immune precipitation, and several other melanoma-associated Ags of 45–110 kDa that are expressed by melanoma tissue in vivo and are immunogenic in humans (23). All patients were immunized to vaccine, using alum or liposomes as adjuvant (24). It was administered intradermally split into the four extremities, every 2–3 wk four times. Peripheral blood was collected before immunization and 1 wk following the fourth immunization. Mononuclear cells were separated on Ficoll-hypaque and frozen in liquid N₂ until used.

Peptides were prepared on an Applied Biosystems (Foster City, CA) instrument. Briefly after removal of the α-amino-ter-butylcarbonyl protecting group, the phenylacetamidomethyl resin peptide was coupled with a 4-fold excess of preformed symmetrical anhydride (hydroxybenzyltriazole esters for arginine, histidine, asparagine, and glutamine) for 1 h in dimethylformamide. For arginine, histidine, asparagine, glutamine, and histidine residues, the coupling step was repeated to obtain a high efficiency coupling. Peptides were cleaved by treatment with hydrogen fluoride in the presence of the appropriate scavengers. Synthetic peptides were purified by reverse phase HPLC. The purity and identity of the peptides, which was routinely >95% was determined by amino acid sequence and mass spectrometric analysis respectively.

Peptide binding to purified HLA-A*0201 molecules was measured as previously described (25). Briefly, the assay is based on the inhibition of binding to detergent-solubilized HLA molecules of a radiolabeled standard peptide with strong binding affinity for HLA-A*0201. The standard peptide, FLPSDYFPSV, was radioiodinated by the chloramine T method using ¹²⁵I (ICN, Irvine, CA). HLA-A*0201 concentration yielding ∼15% of bound peptide (approximately in the 10 nM range) was used in the inhibition assays. Various doses of the test peptides (1 nM to 10 μM) were incubated with 5 nM radiolabeled standard peptide and HLA-A*0201 molecules for 2 days at room temperature in the presence of a mixture of protease inhibitors and 1 μM β₂-microglobulin (Scripps Laboratories, San Diego, CA). At the end of the incubation period, the percent HLA bound radioactivity was determined by gel filtration. The peptides were selected based on their capacity to bind to HLA-A*0201 with an affinity greater than an IC₅₀ of 500 nM, which is known to be in the immunogenic range for cytotoxic T lymphocyte epitopes (26).

We determined the number of peptide-specific CD8⁺ T cells in peripheral blood by filter spot assay as previously described (14). Briefly, 96-well polyvinylidene dilfuoride filter plates (Millipore, Bedford, MA) were precoated with mAbs to human IFN-γ, (Biosource, Camarillo, CA) and seeded with ∼20,000 HLA-A*0201⁺ target cells/well (SFM20 · A2) (14). The target cells were pulsed with 20 nM of test peptide before addition of effectors. The HLA-A*0201-restricted influenza peptide FluM1_58–66 (GILGFVFTL) to which most individuals have a CD8⁺ T cell response was added to some wells as a positive control (27). mAb IVA12 (anti-HLA-DR, DP, and DQ (HB145 from the American Type Culture Collection, Manassas, VA) was added to all wells to prevent presentation of class II Ags by targets or residual monocytes. Effector PBMC were thawed, resuspended in RPMI 1640 with 10% FBS and depleted of monocytes on plastic. The cells were then added to the wells, incubated 48 h at 37°C and 5% CO₂, washed with PBS-Tween 20, incubated overnight with goat anti-IFN-γ (R&D, Minneapolis, MN), followed by biotinylated rabbit anti-goat Ig (Sigma, St. Louis, MO) and extravidin alkaline phosphatase (Sigma). Spots representing individual T cells that had been stimulated by peptides and released IFN-γ were visualized with BCIP/NBT (Kirkegaard & Perry Laboratories, Gaithersburg, MD) and counted with a dissecting microscope.

The number of CD8⁺ T cells was calculated as the number of IFN-γ-secreting cells in wells without Abs minus those in wells with anti-CD8 Abs (OKT8 from the American Type Culture Collection). The number of peptide-specific CD8⁺ T cells was determined by subtracting the number of CD8⁺ cells reacting to targets pulsed with an A*0201-restricted control peptide (YLKKIKNSL from falciparum malaria) (28) from the number of CD8⁺ T cells reacting to melanoma peptide-pulsed targets. All assays were conducted twice, once in duplicate and once in triplicate, and the average value used. The mean SD was ±0.95. The test was considered positive if there were ≥5 peptide-specific CD8⁺ T cells per 500,000 PBL, which is twice the maximum SD. The number of vaccine-induced peptide-specific CD8⁺ T cells was calculated by subtracting the number of peptide-specific CD8⁺ T cells present before treatment from that after four vaccine immunizations in the same patient. The vaccine was considered to have stimulated a CD8⁺ T cell response to that peptide if this number was ≥5.

The peptides used in this study were selected for their potential ability to stimulate CD8⁺ T cells in vivo based on being derived from Ags known to be associated with melanoma, i.e., MAGE-3, Melan-A/MART-1, gp100, tyrosinase, MC1R, or TRP-2; having 9–10 amino acid residues; containing an HLA-A*0201 binding motif; and by having high binding affinity for HLA A*0201 in vitro. A total of 45 melanoma peptides that satisfied these criteria is provided in Table I.

We examined the ability of the 45 HLA-A*0201-restricted peptides selected as described above, an HLA-A*0201-restricted malaria peptide as a negative control (28), and an HLA-A*0201-restricted influenza peptide as a positive control (27) to be recognized by CD8⁺ T cells from melanoma patients. We measured the number of CD8⁺ T cells directed to each peptide before and after immunization to a polyvalent melanoma vaccine containing the Ags from which the melanoma peptides were derived. The study was conducted on 22 sequential HLA-A*02⁺ patients who were immunized to the polyvalent vaccine. All but one patient were A*0201⁺. Patient 484 was A*0205+.

Before vaccine treatment, elevated numbers of circulating peptide-specific CD8⁺ T cells were found in 8 of the 22 patients. These patients reacted to 14 of the 45 melanoma peptides (see Table I). One or more of these peptides was derived from each of the 7 Ags studied. However, elevated CD8⁺ T cells to any one peptide were present in no more than 1 patient (5%).

Following four vaccine immunizations, a peptide-specific CD8⁺ T cell response was induced or augmented to 22 (47.8%) of the peptides, as indicated by a minimum increase of 5 peptide-specific CD8⁺ T cells over baseline measurement in the same patient (see Table II). Of the induced responses, at least one of the inducing peptides was derived from each of the proteins studied. The most frequently positive peptides were gp100_585–613 IMPGQEAGL) to which responses were induced in 3 (14%) of the 22 patients (see Table II). The highest vaccine-induced responses were directed to MAGE-3_271–278 (FLWGPRALV), Melan-A/MART-1_56–64 (ALMDKSLHV) and gp100_13–21 (AVIGALLAV) which all were >40 peptide-specific CD8⁺ T cells/500,000 PBL over baseline measurements. Twenty-one of the 22 patients were also tested for their ability to recognize the HLA-A*0201-restricted influenza peptide FluMI_58–66 (27) as a positive control. This peptide was recognized by 76% of the patients. There was no response to the control malaria peptide in any patient either before or after vaccine immunization. Overall, vaccine immunization induced a peptide specific CD8⁺ T cell response to at least one peptide in 59% of the 22 patients. Responses were induced in 7 of 14 (50%) of patients without a pre-existing response, and further induced in 4 of 8 (50%) patients who had a pre-existing response before vaccine treatment.

The relative responses to each Ag were evaluated based on the frequency of vaccine-induced, peptide-specific, CD8⁺ T cell responses to at least one peptide derived from that Ag. The results are summarized in Table III. All of the melanoma Ags stimulated a CD8⁺ T cell response to at least one peptide on that Ag. The most frequently recognized Ag was gp100, which induced CD8⁺ T cell responses in 27% of patients. Overall, 59% of the patients had a vaccine-induced CD8⁺ T cell response to at least one of the Ags studied.

There was striking heterogeneity in the ability of individual patients to develop vaccine-induced CD8⁺ T cell responses to individual peptides on the same Ag. An example of this heterogeneity in responses to peptides of gp100 is illustrated in Table IV. Individual patients immunized to the same amount of gp100-containing vaccine developed CD8⁺ T cell responses to one gp100 peptide but not to others, whereas the reverse was true of other patients. Overall, 27% patients developed a CD8⁺ T cell response to at least one of the gp100 peptides, but no more than three patients (14%) reacted to the same peptide. There was a similar degree of heterogeneity in vaccine-induced responses to peptides derived from the other Ags (data not shown). Nine to 18% of the patients reacted to at least one peptide from each Ag, but no more than three patients (14%) reacted to the same peptide that Ag.

There was also marked heterogeneity in the responses of individual patients to individual Ags. As illustrated in Table III, some patients developed vaccine-induced CD8⁺ T cell responses to peptides of one Ag but not to those of another Ag, whereas the reverse was true for other patients. Though 59% of patients developed CD8⁺ T cell responses to at least one of seven Ags studied, only 9% to 27% of patients reacted to the same Ag.

The clinical status of the patients in the study is shown in Table V. All survived beyond 1 year, and over 75% are still surviving. Of those patients who had a vaccine-induced CD8⁺ T cell response to at least one of the Ags tested, a higher proportion (though not statistically significant) were progression-free after 1 year as compared with those who had no response (54% vs 22%). This positive correlation was unrelated to the stage of disease or the type of adjuvant administered with the vaccine as these two variables were equally distributed in both groups. Most likely a larger study will be needed to show statistically significant differences.

The major findings in this study are the identification of multiple melanoma-associated peptides that can stimulate CD8⁺ T cell responses in vivo in humans, and the demonstration of marked HLA-independent heterogeneity in the ability of these peptides to induce CD8⁺ T cell responses in different individuals. A multipronged strategy was used to identify these peptides. 1) We immunized patients to a polyvalent vaccine that contains multiple melanoma-associated Ags. This permitted the direct and concurrent identification of multiple immunogenic peptides by measuring the presence and level of peptide-specific CD8⁺ T cells in peripheral blood before and after vaccine treatment. The vaccine contains MAGE-3, Melan-A/MART-1, gp100, and tyrosinase (14). It also contains MC1R and TRP-2, as shown in this study by its ability to stimulate CD8⁺ T cell responses to these Ags. 2) We used a screening strategy to select peptides that had a high potential to stimulate CD8⁺ T cells. It was based on their having 9–10 amino acid residues, an anchor motif, and high binding affinity to, HLA-A*0201 in vitro. This made it feasible to focus studies on those peptides most likely to be immunogenic. 3) We used a sensitive filter spot assay for peptide-specific CD8⁺ T cells to evaluate the response to these peptides in vivo. The assay could detect fewer than 5 peptide-specific CD8⁺ T cells per 500,000 PBL (14) and was sensitive enough to quantify the small increase in peptide-specific CD8⁺ T cells which appears in peripheral blood following vaccine immunization. Our method was reproducible within ±2.0 IFN-γ-producing cells/500,000 PBL and required only a small amount of the blood with minimum manipulation compared with assays using purified CD8⁺ T cells. Using this strategy, we identified multiple peptides presented by HLA-A*0201 and derived from MAGE-3, Melan-A/MART-1, gp100, tyrosinase, MC1R, and TRP-2 that have immunogenic potential as evidenced by their ability to stimulate CD8⁺ T cell responses in vivo. Several of these peptides were previously known to be recognized by CD8⁺ T cells in vitro, including FLWGPRALV from MAGE-3 (3), AAGIGILTV and GILTVILGV from Melan-A/MART-1 (4, 29), MLGTHTMEV from gp100 (30) and FLPWHRLFL from tyrosinase (31).

All of the Ags we studied including MAGE-3, Melan-A/MART-1, gp100, tyrosinase, MC1R, and TRP-2 were able to stimulate CD8⁺ T cell responses in vivo because patients immunized to a vaccine containing these Ags had vaccine-induced responses to peptide epitopes derived from them. There were differences in the frequency with which individual Ags stimulated CD8⁺ T cell responses, ranging from gp100 which induced CD8⁺ T cell responses in 27% of the patients to MC1R which induced responses in 9%. However, we could not determine whether this was due to differences in intrinsic immunogenicity in vivo, differences in the relative amount of each Ag present in the vaccine, or a selection bias in the number or type of peptides derived from each Ag selected for testing.

Surprisingly, we did not find any of the peptides to be were clearly immunodominant, as none induced CD8⁺ T cell responses to a much greater extent than the others. This is in contrast to well-described immunodominance previously reported for one peptide we studied, Melan-A/MART-1_27–35. The inability to detect frequent immune responses to this particular peptide did not appear to result from lack of sensitivity as we detected responses to the influenza peptide positive control in the majority of patients. Because of its much lower binding affinity than all of the other peptides, it is possible that we would have seen greater recognition had we used more peptide in this specific case or used the decapeptide that has been reported by Romero et al. (32) to have much greater recognition by T cells from different individuals than the nonapeptide.

Our most important finding was the marked HLA-independent heterogeneity in the ability of different patients to develop CD8⁺ T cell responses to the same peptide or to the same Ag. This was not due to differences in the amount of peptide or Ag used to immunize the patients, since all received the same dose of the same vaccine. Nor was it due to differences in the immunological competence, since all patients were immunologically competent as evidenced by their ability to react to recall Ags and the majority of patients reacted to the control influenza peptide FluM1_58–66. Nor was it due to differences in the HLA subclass restriction of the peptides or of the patients, since 21 of the 22 patients were A*0201⁺ and the peptides were all HLA-A*0201-restricted. The one patient who was of the A*0205 subtype did not respond to any peptide. Nor was it due to differences in the solubility or affinity of the peptides used in the assays, since all patients cells were exposed to all of the peptides under the same conditions.

In summary, we have directly demonstrated the ability of seven melanoma-associated Ags to stimulate a CD8⁺ T cell response in humans and identified multiple CD8⁺ T cell peptide epitopes on each of these Ags that can be recognized by CD8⁺ T cells in vivo. These peptides and Ags are attractive candidates for vaccine construction because they can stimulate CD8⁺ T cell responses. There was marked HLA-independent heterogeneity in the ability of the same peptide or Ag to stimulate CD8⁺ T cell responses in different patients, and no single peptide or Ag was clearly immunodominant. Regardless of the reasons for this heterogeneity, the implication for cancer vaccine design is that the stimulation of CD8⁺ T cell responses to cancer will be maximized by constructing vaccines from multiple peptides or multiple Ags.

We thank Dunlu Chen and Michelene Rivas for vaccine preparation, Ann Byrne and Elise Kelman for performing assays, and Arvind K. Menon and Peter S. Masiakos for HLA typing.