MHC Class I-restricted Recognition of a Melanoma Antigen by a Human CD4⁺ Tumor Infiltrating Lymphocyte

T cells can mediate the recognition and elimination of virus-infected cells, tissue allografts, cells expressing normal self antigens, and tumor cells in humans and in animal models. Mature T cells develop in the thymus from bone marrow-derived progenitor cells (reviewed in Refs. 1 and 2). T-cell antigens are recognized in association with MHC molecules, resulting in MHC-restricted antigen recognition (3). T cells expressing the CD8 coreceptor on their cell surface recognize 8–10 amino acid peptide fragments bound to MHC class I molecules and thus are MHC class I restricted (4, 5). CD8⁺ T cells are generally CTLs and commonly represent the effector arm of T cell-mediated immunity. T cells expressing the CD4 coreceptor on their cell surface recognize 10–15 amino acid peptide fragments bound to MHC class II molecules and thus are MHC class II restricted (4, 5). CD4⁺ T cells are generally helper T cells that secrete cytokines that initiate and augment the function of CTL and B cells.

Broad cross-reactivity exists among T cells in the periphery and in the thymus, suggesting that epitope mimicry plays an important part in shaping the T-cell repertoire (6, 7). The sequence and relative abundance of individual peptides presented by the thymic epithelium can have dramatic effects on T-cell development (8, 9). Individual peptides are capable of positively selecting a diverse T-cell repertoire in fetal thymic organ cultures (10). In the periphery, individual T-cell clones can cross-react with alloantigens (11) and a broad spectrum of peptides in the context of self MHC (12, 13, 14). These studies suggest that cross-reactivity is an important mechanism for generating and maintaining T-cell immunity.

In the present study, we describe an HLA-A2-restricted, melanoma-reactive CD4⁺ TIL² culture from melanoma patient 1383. Transient transfection experiments identified the antigen that was recognized by TIL 1383 I as tyrosinase. The epitope recognized by TIL 1383 I was a 9 amino acid peptide corresponding to amino acid residues 368–376 from tyrosinase. TIL 1383 I was weakly cytolytic and secreted cytokines in a pattern consistent with T helper 1 (T_h1) cells that produce IFN-γ, IL-2, GM-CSF, and TNF-α but not IL-4 or IL-10. The MHC class I restricted antigen recognition by CD4⁺ T cells supports the notion that epitope mimicry exists in the immune response by T cells.

All melanoma, immortalized fibroblasts (E6E7 transformed) and EBV B-cell lines used in this study were established from surgical specimens (TILs, melanoma, and fibroblasts) or peripheral blood mononuclear cells (EBV B) from melanoma patients undergoing immunotherapy in the Surgery Branch, National Cancer Institute, as described previously (15, 16, 17). Prostate cancer lines 1542 CP3TX (HLA-A2⁻) and 1550 CPTX (HLA-A2⁺) were kindly provided by Dr. Suzanne L. Topalian (Surgery Branch, National Cancer Institute, NIH, Bethesda, MD) and were established from patients undergoing radical prostatectomies at the National Cancer Institute as described previously (17). Renal cell carcinoma lines UOK 181 (HLA-A2⁻) and UOK 131 (HLA-A2⁺) were kindly provided by Dr. W. Marston Linnehan (Urology Branch, National Cancer Institute, NIH, Bethesda, MD) and were established from patients undergoing radical nephrectomy at the National Cancer Institute as described previously (18). Esophageal carcinoma cell lines HCE-4 (HLA-A2⁻) and SKGT-5 (HLA-A2⁺) were kindly provided by Dr. David S. Schrump (Surgery Branch, National Cancer Institute; Ref. 19). Colon carcinoma cell lines LOVO (HLA-A2⁻) and SW480 (HLA-A2⁺) lines, breast carcinoma cell lines MCF7 (HLA-A2⁺) and MDA-MB-231 (HLA-A2⁺), and ovarian carcinoma cell line SKOV3 (HLA-A2⁻) were obtained from American Type Culture Collection (Rockville, MD). 624–28 MEL (HLA-A2⁻) and 624–38 MEL (HLA-A2⁺) were obtained by limiting dilution cloning of 624 MEL (20). 888 A2 MEL, 397 A2 MEL, and SKOV3 A2 were obtained by transfecting 888 MEL, 397 MEL, and SKOV3, respectively, with a cDNA encoding HLA-A2 that was cloned into the pcDNA 3 eukaryotic expression vector (In Vitrogen, Carlsbad, CA). COS-7 cells that were used for transient transfection assays and T2 cells that were used for peptide recognition assays have been described elsewhere (21, 22). All tumor cell lines except the prostate cell lines were maintained in CM that consisted of RPMI 1640 (Biofluids, Rockville, MD) supplemented with 10% heat-inactivated fetal bovine serum (Life Technologies, Inc., Rockville, MD) and 1% penicillin-streptomycin-glutamine (Life Technologies). The prostate cell lines were maintained in Keratinocyte-SFM (Life Technologies) supplemented with 5% heat-inactivated fetal bovine serum, 1% penicillin-streptomycin-glutamine, 10 mm HEPES (Biofluids, Rockville, MD), and 25 mm bovine pituitary extract (Life Technologies).

All TILs were established from surgical specimens obtained from patients with metastatic melanoma as described (15). Viable cells (1 × 10⁶; tumor and lymphocytes) from patient 1383 were plated in each of 12 wells of a 24-well plate in a total of 2 ml of TIL medium that consisted of AIM V medium (Life Technologies) supplemented with 5% heat-inactivated pooled AB human serum (Sigma Chemical Co., St. Louis, MO), 1% penicillin-streptomycin-glutamine, and 6000 IU/ml recombinant human IL-2 (Chiron, Emeryville, CA). All TIL cultures were expanded and maintained at densities between 0.5–2.0 × 10⁶ cells/ml. TIL 1235 and TIL 1520 are HLA-A2 restricted CD8⁺ TIL cultures that recognize the MART-1:27–35 and gp100:209–217 peptide epitopes, respectively (22, 23). TIL 888 is an HLA-A24-restricted CD8⁺ TIL culture that recognizes tyrosinase (24). TIL 1558 is an HLA-DR1-restricted CD4⁺ TIL culture that recognizes a mutated triosephosphate isomerase that is expressed only by its autologous 1558 MEL (25).

The cloning of the HLA-A2.1 (22), MART-1 (22), gp100 (26), tyrosinase (24), TRP-2 (27), and NY Eso-1 (28) cDNAs has been described elsewhere. All cDNAs were subcloned into the eukaryotic expression vector pcDNA 3 (In Vitrogen).

Peptides were synthesized by a solid phase method using an AMS 422 multiple peptide synthesizer (Gilson Co., Worthington, OH) as described previously (29). The amino acid sequences of the MART-1, gp100, and tyrosinase peptides used in this study are as follows: MART-1:27–35, AAGIGILTV (22); gp100:209–217, ITDQVPFSV (30); tyrosinase:1–9, MLLAVLYCL (31); tyrosinase:368–376 370 D, YMDGTMSQV (32); and tyrosinase:368–376 370 N, YMNGTMSQV (31). The amino acid sequence of the hepatitis B virus core protein:18–27 23Y peptide (HBV:18–27 23Y) that was used for peptide binding assays was FLPSDYFPSV (33). All peptides were dissolved in DMSO at 2 mg/ml.

The binding of peptides to HLA-A2 was measured using a live cell competitive binding assay as described, except T2 cells were substituted for EBV-transformed B cells (33). Values obtained represented the IC₅₀ for each peptide, which is the concentration of peptide that inhibits 50% of the binding of ¹²⁵I-labeled hepatitis B virus core protein residues 18–27 modified with a tyrosine at P6 (HBV:18–27 23Y) to HLA-A2 molecules on T2 cells. HBV:18–27 23Y was radiolabeled with ¹²⁵I using chloramine T, and labeled peptide was separated from the unincorporated ¹²⁵I using a Sephadex G-10 (Sigma) size exclusion column as described (33). Binding of ¹²⁵I-labeled HBV:18–27 23Y peptide to T2 cells was measured in the presence of 10, 1, 0.1, 0.01, and 0.001 mg/ml unlabeled test peptide. The maximum binding and maximum inhibition values for each assay were determined by incubating T2 cells with ¹²⁵I-labeled HBV:18–27 23Y peptide alone or with an excess (50 mg/ml) of unlabeled HBV:18–27 23Y peptide. Each determination was performed in triplicate, and their average was used to calculate the percentage of inhibition as follows:

The ln(% inhibition) was plotted versus the ln(peptide concentration). Using regression analysis, the IC₅₀ was determined as the peptide concentration that inhibited 50% of the binding of the ¹²⁵I-labeled HBV:18–27 23Y peptide.

Oligonucleotide primers were purchased from Life Technologies. The sequences of the human TCR BV subfamily-specific primers have been described elsewhere (34). Human TCR BV subfamily-specific primers were designed based on the genomic TCR BV sequences reported by Rowen et al. (35). Each primer was designed such that there was no more than a single base mismatch with any member of the subfamily. The official nomenclature proposed by the International Union of Immunological Societies subcommittee on nomenclature has been used to identify all TCR gene segments described (36).

Total cellular RNA was isolated from 5 × 10⁶ TIL 1383 I cells using Trizol (Life Technologies, Inc., Gaithersburg, MD), and cDNA was synthesized using oligo dT_(12–18) and Superscript II reverse transcriptase (Life Technologies, Inc., Gaithersburg, MD) as described (34). Each TCR BV subfamily-specific forward primer was used in conjunction with a BC reverse primer to amplify the TCR BV transcripts present in the TIL culture. PCR reactions consisted of 200 μm deoxynucleotide triphosphate (Oncor, Gaithersburg, MD), 400 nm of both forward and reverse primers, and 1 unit of Taq polymerase (Oncor, Gaithersburg, MD) in a 50-μl reaction. Amplifications were performed in a Perkin-Elmer 9600 DNA thermocycler (Norwalk, CT) under the following conditions: 35 cycles of 92°C denaturation for 30 s, 60°C annealing for 30 s, and 72°C extension for 1 min. PCR products were separated using ethidium bromide-stained agarose gels and electronically imaged using an Eagle Eye II Still Video System (Stratagene, La Jolla, CA).

The DNA sequence analysis of the TCRBV12 band from TIL 1383 I was performed using the BV12a primer and Dye Terminator Cycle Sequencing kits (Perkin-Elmer/ABI, Foster City, CA). Sequences were determined using an ABI Prism 310 Genetic Analyzer (Perkin-Elmer/ABI). The resulting sequences were analyzed using the Genetics Computer Group, Inc. software package (37).

Cell surface expression of TCRBV12, CD3, CD4, and CD8 molecules on TIL 1383 I was measured by immunofluorescence using FITC-conjugated anti-TCR BV12 (Immunotech, Westbrook, ME), PE-conjugated anti-CD3 (Becton Dickinson, San Jose, CA), FITC- or PE-conjugated anti-CD4 (Becton Dickinson), and FITC- or PE-conjugated anti-CD8 mAbs (Becton Dickinson). For analysis, the relative log fluorescence of 10⁴ live cells was measured using a FACScan flow cytometer (Becton Dickinson). For cell sorting, the TCR BV12 mAb was dialyzed against PBS to remove the sodium azide and filter sterilized. Cells (2 × 10⁷) cells from a 54-day-old TIL 1383 I culture were stained with FITC-conjugated anti-TCR BV12 mAb. BV12-expressing cells were collected using a Coulter Elite flow cytometer (Coulter, Miami, FL).

The reactivity of TIL 1383 I was measured in cytokine release assays as described (38). Briefly, TIL 1383 I cells were cocultured overnight with stimulator cells in a 1:1 ratio (the actual number of stimulators and responders used are indicated in the table legends) in RPMI 1640 supplemented with 10% heat-inactivated pooled human AB serum and 1% penicillin-streptomycin-glutamine in a total volume of 200 ml in 96-well U-bottomed plates. Supernatants were harvested, and the amount of cytokine release was measured by ELISA. IFN-γ, TNF-α, IL-4, and IL-10 ELISA kits were purchased from Endogen (Woburn, MA). GM-CSF and IL-2 ELISA kits were purchased from R&D Systems (Minneapolis, MN). The stimulators used in these assays included a panel of HLA-A2⁺ and HLA-A2⁻ tumors, EBV-transformed B cells, fibroblast lines, and peptide-pulsed T2 cells. Peptide-pulsed T2 cells were prepared by incubating 1 × 10⁶ T2 cells/ml with peptide for 2 h at 37°C.

COS-7 cells were transiently transfected using Lipofectamine Plus (Life Technologies, Rockville, MD) as described (39). Briefly, 5 × 10⁴ COS-7 cells/well in 96-well, flat-bottomed plates were incubated in 40 ml of serum-free DMEM medium containing 200 ng of pcDNA 3 containing the cloned tumor-associated antigen cDNA (150 ng) and HLA-A2.1 (50 ng) cDNA at 37°C in a humidified CO₂ incubator (5% CO₂) for 3 h. Then, 160 ml of DMEM medium supplemented with 10% heat-inactivated fetal bovine serum and 1% penicillin-streptomycin-glutamine was added. After 2 days, 100 ml of medium were removed, and 2.5 × 10⁴ T cells in 100 ml of CM were added to each well. The amount of IFN-γ was measured by ELISA as described above.

The ability of anti-MHC, anti-CD4, and anti-CD8 mAbs to inhibit antigen recognition by TIL 1383 I was measured in cytokine release assays as described (38). Briefly, 1 × 10⁴ T cells and 1 × 10⁴ tumor cells were cocultured in RPMI 1640 supplemented with 10% heat-inactivated pooled AB serum and 1% penicillin-streptomycin-glutamine in a total volume of 200 ml in 96-well U-bottomed plates for 24 h. For the MHC blocking assays, the tumor cells were incubated with IVA12 and/or W6/32 mAb (final concentration, 20 mg/ml) for 30 min at room temperature prior to the addition of the T cells. For the CD4/CD8 blocking assays, the T cells were incubated with anti-CD4 (final concentration, 2.5 mg/ml) or anti-CD8 (final concentration, 1.25 mg/ml) mAb for 30 min at room temperature prior to the addition of the tumor cells. The amount of IFN-γ released was measured by ELISA as described above.

The ability of TIL 1383 I to lyse HLA-A2⁺, tyrosinase⁺ target cells was measured in ⁵¹Cr release assays as described (15). Briefly, 10⁶ melanoma or peptide-pulsed T2 cells were labeled for 1 h at 37°C with 200 mCi of ⁵¹Cr (Amersham, Arlington Heights, IL) in CM. Labeled target cells (5 × 10³) were incubated with 4 × 10⁵ (80:1), 1 × 10⁵ (20:1), 2.5 × 10⁴ (5:1), and 6.25 × 10³ (1.25:1) effector cells for 4 h at 37°C in 200 ml of CM. Supernatants were harvested and counted using an Wallac 1470 Wizard automatic gamma counter (Wallac, Gaithersburg, MD). Total and spontaneous ⁵¹Cr release by each target was determined by incubating 5 × 10³ labeled target cells in 2% SDS or CM, respectively, for 4 h at 37°C. Each point represented the average of triplicate wells, and percentage of specific lysis was calculated as follows:

Twelve TIL cultures from patient 1383 were established from an enzymatically digested melanoma lesion by culturing the cells in medium containing 6000 IU/ml of recombinant human IL-2. Each culture was expanded separately and tested on day 17 for its reactivity against the autologous 1383 MEL, an HLA-A2⁺ and an HLA-A2⁻ melanoma cell line. One culture, designated TIL 1383 I, appeared to be HLA-A2 restricted and melanoma reactive in the initial screen (data not shown). TIL 1383 I was expanded, and its reactivity was determined using a more extensive panel of melanoma and nonmelanoma cell lines. As shown in Table 1, TIL 1383 I secreted significant amounts of IFN-γ (at least twice background and >100 pg/ml) when stimulated with 9 of the 11 HLA-A2⁺ melanomas but with none of the five HLA-A2⁻ melanomas (Table 1). TIL 1383 I did not recognize any of the HLA-A2⁺ or HLA-A2⁻ EBV transformed B-cell, fibroblast, prostate carcinoma, ovarian carcinoma, renal cell carcinoma, colon carcinoma, esophageal carcinoma, and breast carcinoma cell lines. These results indicated that TIL 1383 I recognized a shared melanoma antigen in the context of HLA-A2.

TIL 1383 I was then tested for its ability to recognize transfectants expressing a panel of melanoma antigens described previously containing known HLA-A2-restricted epitopes. As shown in Table 2, TIL 1383 I recognized COS cells that had been transiently transfected with the HLA-A2 and tyrosinase cDNAs. COS cells transfected with HLA-A2 and MART-1 were only recognized by TIL 1235, and COS cells transfected with HLA-A2 and gp100 were only recognized by TIL 1520. None of the TILs recognized COS cells transfected with HLA-A2 alone, HLA-A2 and TRP2, or HLA-A2 and Eso I. Recognition of the tumor cell controls in this assay was consistent with each of the TILs being HLA-A2 restricted and melanoma reactive. Therefore, the melanoma-associated antigen recognized by TIL 1383 I was tyrosinase.

Two HLA-A2-restricted tyrosinase epitopes consisting of amino acid positions 1–9 and 368–376 (designated T9-1 and T9-368, respectively) have been described previously (31). To determine whether one of these two HLA-A2-restricted tyrosinase epitopes were recognized by TIL 1383 I, T2 cells pulsed with either T9-1 or T9-368 were assayed for their ability to stimulate IFN-γ release from TIL 1383 I. The T9-368 peptide undergoes a posttranslational modification that changes the N at position 370, which is encoded by the tyrosinase gene to a D, which is found on the surface of melanoma cells (32). As shown in Table 3, TIL 1383 I recognized T2 cells pulsed with the T9-368 peptide with either N or D at position 370. TIL 1383 I recognized the HLA-A2⁺ melanoma lines but not T9-1, M9-27, or G9-209 pulsed T2 cells, 1383 fibroblasts, or 888 MEL. TIL 1235 recognized M9-27, TIL 1520 recognized G9-209, and TIL 620 recognized both M9-27 and G9-209 as expected. TIL 888, which is known to recognize tyrosinase in the context of HLA-A24, recognized the 888 MEL lines but did not recognize any of the HLA-A2-restricted peptides pulsed on T2 cells. Therefore, the previously described tyrosinase epitope T9-368 was recognized by TIL 1383 I, with both the 370N and the 370D forms of the peptide being recognized efficiently.

The TCR BV genes used by TIL 1383 I was determined by reverse transcription-PCR using TCR BV subfamily-specific primers as described (34). TIL 1383 I was oligoclonal because we could only detect BV12, BV14, and BV24 transcripts, with BV12 being the predominant subfamily used (data not shown). DNA sequence analysis of the BV12 band found only a single productive rearrangement, indicating that the BV12 subfamily was clonally expanded in TIL 1383 I (data not shown). To confirm that BV12⁺ T cells were the predominant T-cell clonotype in this culture, the expression of BV12, CD3, CD4, and CD8 on TIL 1383 I was measured by immunofluorescence. Essentially 100% of the T cells in TIL 1383 I expressed both CD3 and CD4 (Fig. 1,c), with ∼99% of the CD4⁺ T cells expressing BV12 (Fig. 1,d). Less than 1% of the TIL 1383 I cells were CD8⁺ (Fig. 1,e) or CD8⁺ and BV12⁺ (Fig. 1,f). These rare events were attributable to nonspecific staining by the anti-CD8 mAb because there were no CD4⁻ T cells in TIL 1383 I (Fig. 1, c and d), and we were unable to detect CD8a chain transcripts by reverse transcription-PCR (data not shown). We wanted to obtain a pure population of BV12⁺, CD4⁺ T cells from TIL 1383 I to determine whether these were the tyrosinase-reactive cells in the TIL 1383 I culture. Because attempts to clone TIL 1383 I in limiting dilution were unsuccessful, we stained a TIL 1383 I culture with an anti-BV12 mAb and used the cell sorter to further enrich the population of BV12⁺ T cells. After two sorts, we were able to generate a TIL 1383 I culture that was 100% BV12⁺ T cells (Fig. 2,c) from a culture that was 98% BV12⁺ T cells (Fig. 2,b). The sorted cells were expanded and tested for expression of BV12, CD4, and CD8 and for their ability to recognize antigen. After 2 weeks of culture, 100% of sorted cells remained CD4⁺ and BV12⁺ (data not shown). When tested for their antigen recognition, the sorted cells specifically recognized T9-368 pulsed T2 cells, and the amount of IFN-γ released was equivalent to the amount released by the unsorted cells (Table 4). These results indicated that the HLA-A2-restricted recognition of tyrosinase by TIL 1383 I was by a BV12-expressing, CD4⁺ T cell.

Human T cells differ from murine T cells in that activated human T cells express MHC class II molecules and can process and present class II-restricted epitopes to each other. Therefore, it is possible that the tyrosinase peptide could be shed from the tumor cells and presented by the TIL 1383 I cells to each other in the context of MHC class II rather than the MHC class I. To demonstrate that the T9-368 peptide was directly presented to TIL 1383 I cells by HLA-A2 on the tumor cells or T2 cells, we compared cytokine secretion by TIL 1383 I when stimulated with peptide-pulsed T2 cells to T cells stimulated with free peptide. TIL 1383 I secreted significant amounts of IFN-γ when stimulated with T2 cells pulsed with 1.0, 0.1, or 0.01 mg/ml of T9-368 370D peptide (Fig. 3,a) and 1.0 or 0.1 mg/ml of T9-368 370N peptide (Fig. 3,b). In contrast, TIL 1383 I did not secrete IFN-γ when stimulated with the T9-368 370D or 370N peptide alone. The CD8⁺ T cell control TIL 1520 cells also secreted IFN-γ when stimulated with G9-209 pulsed T2 cells (Fig. 3 c). TIL 1520 cells secreted a low but significant amount of IFN-γ when stimulated with free peptide (200 pg/ml with 1.0 mg/ml peptide). When the ability of these peptides to bind to HLA-A2 molecules was measured in a competition binding assay, the IC₅₀ for G9-209, T9-368 370D, and T9-368 370N was measured to be 74, 702, and 249 mm, respectively. One possible explanation for the ability of free G9-209 peptide to stimulate TIL 1520 could be that the binding affinity of G9-209 was high enough to allow a small amount of peptide to be bound to the HLA-A2 molecules on the surface of the T cells. The TIL 1520 cells could then stimulate each other to secrete IFN-γ. In contrast, TIL 1383 I cells were unable to stimulate each other when cultured with free peptide, possibly because the low binding affinity of the T9-368 peptides did not allow efficient binding of peptide to the HLA-A2 molecules on their cell surface. These results indicate that the IFN-γ released by TIL 1383 I could only be attributable to the recognition of T9-368/HLA-A2 complexes on the tumor cells.

The MHC class I-restricted recognition of tyrosinase by TIL 1383 I was further demonstrated in mAb blocking assays. As shown in Table 5, anti-MHC class I mAb W6/32 inhibited TIL 1383 I recognition of 1383 MEL (35%; 184.8 to 120.4 pg/ml), 1558 MEL (46%; 220.7 to 119.1 pg/ml), and T9-368 pulsed T2 cells (61%; 1146.9 to 451.7 pg/ml). In contrast, anti-MHC class II mAb IVA12 had little effect on TIL 1383 I recognition of melanoma cells (1383 MEL: 9%, 184.8 to 167.3 pg/ml; 1558 MEL: 0%, 234.2 to 220.7 pg/ml) or peptide-pulsed T2 cells (17%; 1146.9 to 955.7 pg/ml). W6/32 inhibited MHC class I restricted IFN-γ release by control of the CD8⁺ T cell lines TIL 1520 or TIL 1235, and IVA12 inhibited class II-restricted IFN-γ release by the control of CD4⁺ T-cell line TIL 1558, as expected (Table 5). The ability of W6/32 to inhibit recognition of HLA-A2⁺ tyrosinase⁺ melanoma cells and peptide-pulsed targets by TIL 1383 I is a further demonstration of its MHC class I-restricted reactivity.

The contribution of the CD4 coreceptor to antigen recognition by TIL 1383 I was determined by blocking cytokine release with anti-CD4 mAb. The amount of IFN-γ released by TIL 1383 I when stimulated with 1558 MEL was reduced by 15% (705.4 to 604.9 pg/ml) and 23% (705.4 to 546.5 pg/ml) in the presence of anti-CD4 mAb and anti-CD8 mAb, respectively (Table 5). Anti-CD4 mAb inhibited MHC class II-restricted recognition of 1558 MEL by CD4⁺ TIL 1558 (44% inhibition; 1651.0 to 930.0 pg/ml) but had no effect on MHC class I-restricted recognition of 1558 MEL by TIL 1520 (4% inhibition; 507.2 to 489.5 pg/ml). Anti-CD8 mAb inhibited MHC class I-restricted recognition of 1558 MEL by CD8⁺ TIL 1520 (46% inhibition; 507.2 to 276.0 pg/ml) but had minimal effects on MHC class II-restricted recognition of 1558 MEL by TIL 1558 (17% inhibition; 1651.0 to 1363.0 pg/ml). These results indicate that the CD4 coreceptor is not required for antigen recognition by TIL 1383 I.

The pattern of cytokine release from TIL 1383 I after antigen stimulation is shown in Table 6. TIL 1383 I secreted IFN-γ, GM-CSF, TNF-α, and IL-2 but not IL-4 or IL-10 when stimulated with either the autologous 1383 MEL, 1558 MEL, or OKT3. By comparison, CD8⁺ TIL 1520 secreted only IFN-γ and GM-CSF when stimulated with 1383 MEL or 1558 MEL and IFN-γ, GM-CSF, and TNF-α when stimulated with OKT3. In this experiment, 1383 fibroblast-stimulated TIL 1383 I secreted <20% of the amount of GM-CSF secreted by TIL 1383 I stimulated with tumor cells. Fibroblasts alone secreted small amounts of TNF-α and GM-CSF. These results indicate that the pattern of cytokine secretion by TIL 1383 I was consistent with a Th₁ phenotype, despite being an MHC class I-restricted CD4⁺ T cell.

The ability of TIL 1383 I to lyse HLA-A2⁺ melanoma cells and T9-368 pulsed T2 cells was tested in 4-h ⁵¹Cr release assays. As shown in Fig. 4,a, TIL 1383 I weakly lysed 888 A2 MEL and T9–368 370D pulsed T2 cells but failed to lyse autologous 1383 MEL, 1383 fibroblasts, and M9–27 pulsed T2 cells. TIL 888, which recognizes tyrosinase in the context of HLA-A24, efficiently lysed only the HLA-A24⁺ melanoma lines 888 MEL and 888 A2 MEL (Fig. 4,b). TIL 1235 efficiently lysed 888 A2 MEL and M9-27-pulsed T2 cells but not 1383 MEL, 1383 fibroblasts, and T9-368 pulsed T2 cells (Fig. 4 c). The failure of 1383 MEL to be lysed by TIL 1235 suggested that 1383 MEL may be resistant to cell-mediated lysis. Therefore, despite being a CD4⁺ T cell, TIL 1383 I could weakly lyse some of the HLA-A2⁺, tyrosinase⁺ targets.

Classically, CD4⁺ T cells recognize antigens presented by MHC class II molecules, whereas CD8⁺ T cells recognize antigens presented by MHC class I molecules (3, 4, 5). Recently, several examples of MHC class I-restricted CD4⁺ T cells have been reported (40, 41, 42, 43, 44, 45, 46). However, the existence of MHC class I-restricted antigen recognition by CD4⁺ T cells has remained controversial because no single study has been able to exclude the presence of contaminating CD8⁺ T cells in the culture, determine the MHC class I restriction element, and identify the nominal antigen recognized by the T cells. In this study, we described an MHC class I-restricted CD4⁺ T cell isolated from the infiltrating lymphocytes of a human melanoma lesion. Immunofluorescence analysis indicated that 100% of the cells in the bulk TIL 1383 I culture and cells sorted for high expression of BV12 were CD3⁺CD4⁺ T cells with no evidence of contaminating CD8⁺ T cells (Figs. 1 and 2). The reactivity of TIL 1383 I with a broad panel of tumor targets indicated that the antigen recognized by TIL 1383 I was melanoma associated and HLA-A2 restricted (Table 1). The antigen recognized by TIL 1383 I was tyrosinase, with the 368–376 peptide being the antigenic peptide (Tables 2 and 3). Identification of the antigenic peptide enabled us to conclusively demonstrate that a ligand for the TCR expressed by the CD3⁺CD4⁺BV12⁺ T cell in TIL 1383 I was the T9-368 peptide derived from tyrosinase bound to HLA-A2 (Fig. 3).

Current models of MHC restriction, antigen specificity, and T-cell development require that T cells are positively selected by peptide/MHC molecules expressed on the host thymic epithelial cells (47, 48). From these models, there are several potential ways an MHC class I-restricted CD4⁺ could have developed. One possibility is that TIL 1383 I was positively selected by T9–368/HLA-A2 complexes expressed on the thymic epithelium. Peptides presented by MHC class I molecules are generally derived from proteins encoded endogenously by the antigen-presenting cell, implying that all of the peptides involved in the positive selection of CD8⁺ T cells must be derived from proteins encoded by the thymus. The expression of tyrosinase is limited to cells of the melanocyte/melanoma lineage (49). Unless the T9-368 peptide was shed by melanocytes and exchanged with peptides on MHC class I molecules in the thymus, T9-368/HLA-A2 complexes should not exist in the thymus and would therefore be unavailable to positively select TIL 1383 I. However, we cannot exclude the formal possibility that TIL 1383 I was positively selected by some cross-reactive, thymus-derived peptide presented by HLA-A2. It is also possible that TIL 1383 I could be functionally a CD4⁺ CD8⁺ T cell that carries a mutant CD8 chain that fails to bind our anti-CD8 mAb, making TIL 1383 I appear to be a CD4⁺CD8⁻ T cell. However, the model that is most consistent with the current theories of positive selection is that TIL 1383 I was positively selected by some undefined peptide in association with an MHC class II molecule in this patient, and the HLA-A2-restricted recognition of the T9-368 peptide is merely an unusual cross-reaction.

Cross-reactivity is commonly observed among mAbs but has also been attributed to T cells (6, 7, 12, 13, 14). In the immune response to human melanoma, epitope mimicry was directly investigated by determining the ability of MART-1-reactive CTLs to recognize a panel of naturally occurring peptide mimics derived from human self antigens and pathogens (12). Many of these peptides were recognized by the MART-1-reactive CTL clones, and their pattern of peptide recognition (which peptides were recognized and the magnitude of the response) was unique to each T-cell clone. It is, therefore, possible that epitope mimics of the T9-368 peptide exist that can be recognized by TIL 1383 I or that the T9-368 peptide is itself a mimic of some other unknown peptide.

We believe that the variation in the way each T-cell clone recognizes antigen is reflected by the variation in the TCR V genes used by the T-cell clone. Epitope mimicry should then lead to TCR diversity in the immune response to a single antigenic peptide. Despite reports of restricted TCR V gene usage by melanoma reactive T cells (50), we have found TCR β chain usage by melanoma-reactive T cells to be quite diverse (34, 51, 52). In one patient, as many as 10 distinct TCR clonotypes using seven different TCR BV genes were used to recognize the same gp100:209–217/HLA-A2 complex.³ These observations are consistent with the hypothesis that the vast TCR diversity we observe among melanoma-reactive T cells may be attributable to the broad cross-reactivity and redundancy of the immune system.

If epitope mimicry can lead to CD4⁺ T cells that recognize antigen in the context of MHC class I molecules, then T cells should be more common than they appear. There are probably two main reasons why there are very few reported examples of class I-restricted CD4⁺ T cells: (a) most class I-restricted T cells are isolated based on their ability to lyse ⁵¹Cr-labeled targets. Because most CD4⁺ T cells are not cytolytic, class I-restricted CD4⁺ T cells would be generally overlooked in lysis assays. In fact, had we used lysis assays as our initial screen, TIL 1383 I would have been classified as nonreactive and discarded because it failed to lyse the autologous melanoma (Fig. 4); and (b) most T cells require the CD4 or CD8 coreceptors to respond to antigen stimulation (53, 54). CD4 binds to the b2 domain of MHC class II molecules (55), and CD8 binds to the a3 domain of MHC class I molecules (56). The function of these coreceptors is to stabilize the TCR/peptide/MHC complex (57, 58) and enhance T-cell signaling by augmenting CD3z and ZAP70 phosphorylation (59). However, there is T-cell to T-cell variation in the requirement for CD8 for T-cell function (60). Those T cells whose function is CD8 dependent have low- to intermediate-affinity TCRs, and those T cells whose function is CD8 independent have high-affinity TCRs (61, 62). CD4 and CD8 independent clones are relatively rare, suggesting that T cells bearing high-affinity TCRs are relatively rare. Our inability to block T9-368 peptide or tumor cell recognition with anti-CD4 mAb, together with incomplete blocking of IFN-γ secretion by anti-MHC class I mAb, strongly suggests that the BV12 TCR expressed by TIL 1383 I must have sufficient affinity for the T9-368/HLA-A2 complex that it can activate the T cell in the absence of the CD4 coreceptor. Despite having a high-affinity TCR, TIL 1383 I has relatively low avidity compared with other class I-restricted T cells (recognized 10 nm peptide versus ≤ 1 nm peptide; Fig. 3).

TIL 1383 I may prove to be a useful reagent for addressing basic questions of T-cell development and immunological tolerance. We and others have shown that the transfer of low-affinity, MHC class I-restricted TCRs with apparent low affinity into cells that lack the CD8 coreceptor result in cells with low avidity (63, 64, 65). Coexpression of CD8 increased the avidity of the T cells, resulting in enhanced antigen recognition (64, 65). It is interesting to note that the reactivity of TIL 1383 I is similar to the reactivity of T9-368-reactive CD8⁺ T cells. In both cases, the T cells had higher avidity for the T9-368 370D peptide, which is expressed on the surface of melanoma cells than the T9-368 370N peptide, which is encoded by the tyrosinase gene (32). Therefore, the only apparent difference between T9-368-reactive CD8⁺ T cells and TIL 1383 I is the high-affinity TCR expressed by TIL 1383 I. Because high avidity T cells appear to be more effective in treating virus-infected mice than low avidity T cells (66), expressing this high-affinity TCR in normal CD8⁺ T cells should result in high avidity T cells that may be more effective in eliminating metastatic disease. Another possible benefit of this high-affinity TCR would be to transfer this MHC class I reactivity to normal CD4⁺ T cells. Given that TIL 1383 I recognized tumor cells without any contribution from the CD4 coreceptor, we would predict peripheral blood lymphocyte-derived CD4⁺ T cells bearing this TCR would also recognize HLA-A2⁺, tyrosinase⁺ melanoma cells. The ability of tumor-reactive CD4⁺ T cells to mediate tumor regression could be evaluated in a greater number of patients because HLA-A2 is expressed by ∼50% of all melanoma patients.

We thank Drs. John R. Wunderlich, Suzanne L. Topalian, Timothy Clay, and Alfred Singer for thoughtful suggestions and critical reading of the manuscript. We also thank Pat Koen and Arnold Mixon for valuable assistance with flow cytometry.