Spontaneous CD4+ T Cell Responses against TRAG-3 in Patients with Melanoma and Breast Cancers1

The taxol resistance gene TRAG-3 was initially isolated from cancer cell lines that became resistant to taxol in vitro. TRAG-3 is a cancer germline Ag expressed by tumors of different histological types including the majority of melanoma, breast, and lung cancers. In the present study, we report that patients with stage IV melanoma and breast cancers developed spontaneous IFN-γ-producing CD4+ T cell responses against a single immunodominant and promiscuous peptide epitope from TRAG-3 presented in the context of multiple HLA-DR molecules. The TRAG-3-specific CD4+ T cells and clones were expanded in vitro and recognized not only peptide pulsed APCs but also autologous dendritic cells (DCs) loaded with the TRAG-3 protein. All stage IV melanoma patients with TRAG-3-expressing tumors developed spontaneous CD4+ T cell responses against TRAG-3, demonstrating its strong immunogenicity. None of these patients had detectable IgG Ab responses against TRAG-3. TCRβ gene usage studies of TRAG-3-specific CD4+ T cell clones from a melanoma patient and a normal donor suggested a restricted TCR repertoire in patients with TRAG-3-expressing tumors. Altogether, our data define a novel profile of spontaneous immune responses to cancer germline Ag-expressing tumors, showing that spontaneous TRAG-3-specific CD4+ T cells are directed against a single immunodominant epitope and exist independently of Ab responses. Because of its immunodominance, peptide TRAG-334–48 is of particular interest for the monitoring of spontaneous immune responses in patients with TRAG-3-expressing tumors and for the development of cancer vaccines.

T he CD4 ϩ Th cells play a crucial role in B cell activation for Ab production and in the generation and the maintenance of Ag-specific effector CD8 ϩ T cells (1)(2)(3)(4). Therefore, the identification of Th epitopes is of first importance not only for the study of spontaneous Ag-specific immune responses in patients with cancer but also for the development of cancer vaccines with Th epitopes that may optimally stimulate antitumor CD8 ϩ memory T cells. We and others (5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15) have developed strategies to identify novel MHC class II epitopes capable of stimulating Th1-type tumor Ag-specific CD4 ϩ T cells. These efforts have led to the identification of a number of epitopes that are currently used in phase I/II trials peptide-based vaccines.
We have focused our research efforts on the discovery of epitopes from cancer germline Ags (CGAs), 3 expressed by tumors of different histological types but not by normal tissues, except testis. Because the germ cells in testis do not express MHC class I and class II molecules (16), the CGA-derived epitopes are specifically expressed by tumor cells. CGAs appear to give rise to multiple MHC class II-restricted epitopes recognized by CD4 ϩ T cells (8 -10, 12, 17, 18). In patients with active cancers, CGAs may spontaneously stimulate Abs and circulating CD4 ϩ T cells that are most often detectable only after one round of in vitro stimulation with peptide-pulsed APCs (18,19). Whenever detectable, these CGA-specific CD4 ϩ T cells appeared to be associated with the presence of circulating Ag-specific Abs (18). However, few correlative analyses of Ab and T cell responses have been performed, and more studies are needed to better understand the immunological basis of spontaneous immune responses to tumor Ags in cancer patients and to further optimize cancer vaccines strategies.
Here, we have investigated the spontaneous CD4 ϩ T cell and Ab responses in patients with active solid tumors that commonly expressed the taxol resistance-associated gene-3 (TRAG-3), including melanoma, breast, and lung cancers. TRAG-3 is expressed by a wide range of solid tumors including 78% of melanoma cell lines and all tested melanoma tumor samples (20), 60% of breast cancers (21), and 54% of non-small cell lung carcinoma (NSCLC) (22). The TRAG-3 gene was originally isolated from tumor cell lines that became resistant to taxol in vitro (23) and is induced by demethylation of tumor cell lines (24). Multiple HLA-A2-presented peptides capable of stimulating CD8 ϩ T cells from PBL of normal donors or patients with tumors have been reported (25,26). However, only the epitope TRAG-3 56 -68 successfully stimulated CTLs that recognized and lysed TRAG-3-expressing tumor cells. To date, no MHC class II epitope from TRAG-3 has yet been reported.
In this study, we observed that patients with stage IV melanoma and breast cancer developed IFN-␥-producing CD4 ϩ T cell responses against a single immunodominant and promiscuous epitope from TRAG-3. All patients with stage IV melanoma and TRAG-3-expressing tumors had detectable TRAG-3-specific CD4 ϩ T cells, but no circulating TRAG-3-specific Ab. Our findings add to the diversity of the spontaneous immune responses to CGA-expressing tumors observed in patients with advanced cancers. They also provide a good candidate for the monitoring of spontaneous immune responses in patients with TRAG-3-expressing tumors and for the development of TRAG-3-based cancer vaccines.

Cell lines, media, and Abs
Tissues and blood samples utilized for all studies reported in this araticle were obtained under the University of Pittsburgh Cancer Institute (UPCI) Institutional Review Board-approved protocols 96-099 and 99-088. HLA-DR and HLA-DP genotyping of melanoma patients and normal donors was performed using commercial typing panels of PCR primers according to the manufacturer's instructions (Dylan). HLA-DRB1*0401transfected T2 cells, i.e., T2.DR4 cells, and HLA-DR-transfected mouse cells, i.e., L.DR cells, were previously described (8,27). All cell lines were cultured in RPMI 1640 (Invitrogen) supplemented with 10% FCS, L-arginine (116 mg/L), L-asparagine (36 mg/L), and L-glutamine (216 mg/L). The HB55 and HB95 hybridomas, secreting the L243 anti-HLA-DR (class II) mAb and the W6/32 anti-HLA-A, B, C (class I) mAb, respectively, were purchased from American Type Culture Collection. The B7/21 anti-HLA-DP (class II) mAb used in our blocking experiments was previously described (28).

Peptide synthesis
The TRAG-3-derived peptides were synthesized and stored as previously reported (8,9). Synthesis of TRAG-3 peptides was based on the protein sequence encoded by the TRAG-3 cDNA obtained from three melanoma cell lines as detailed below.

Recombinant proteins
The TRAG-3 recombinant protein was expressed in Escherichia coli strain Bl21 (DE3) from a TRAG-3 cDNA cloned in vector pTriEx (Novagen) as described under TRAG-3 cDNA sequencing. The protein was extracted from inclusion bodies in the presence of 8 M urea and purified by ion metal affinity chromatography using the 6-his tag at the N terminus of the protein.
After elution, the protein sample was dialyzed against a buffer containing 2 M urea, 50 mM NaCl, and 20 mM Tris-HCl, pH 8. The TRAG-3 protein was concentrated up to 5 mg/ml. The full length NY-ESO-1 recombinant protein was produced in E. coli and was kindly provided by Drs. Lloyd J. Old and Gerd Ritter (Ludwig Institute for Cancer Research, New York Branch, New York, NY).

TRAG-3 cDNA sequencing
Total RNA was extracted from the three melanoma tumor cell lines UPCI-Mel 285, UPCI-Mel 558, and UPCI-Mel 527.1 using the RNeasy Mini Kit (Qiagen) and following the manufacturer's instructions. Reverse transcription was conducted as previously reported (29). The PCRs were performed in the presence of AmpliTaq DNA polymerase (Applied Biosystems), using the TRAG-3-specific forward primer, HZ30 (5Ј-TGTGGATGGGCCT CATCCAATTAGTT-3Ј) and reverse primer, HZ27 (5Ј-AGCGGCG GTCTTTTATGGAGAGC-3Ј). Both primers were designed according to the published TRAG-3 cDNA sequence (23). The PCR amplification was conducted in 50 l as follows: 5 min at 94°C for 5 min; then 32 cycles (1 min at 94°C, 1 min at 60°C, 1 min at 72°C); and finally 10 min at 72°C. The PCR products were run on 1.5% agarose gel, purified using the QIAquick gel extraction kit (Qiagen), and sequenced using the ABI 3100 automated DNA sequencer.

Construction of pTriEx4-neo-TRAG-3 plasmid
TRAG-3 cDNA was amplified with RT-PCR from the melanoma cell line UPCI-MEL 285 using pfu DNA polymerase (Stratagene Cloning System), and the previously described TRAG-3-specific primers HZ30 and HZ27. The PCR amplification was conducted in 50 l as follows: 5 min at 94°C for 5 min; then 32 cycles (1 min at 94°C, 1 min at 60°C, 1 min at 72°C); and finally 10 min at 72°C. The PCR product was phosphorylated and then run on a 1% agarose gel before purification using QIAquick (Qiagen). pTriEx4-neo eucaryotic expression vector (Novagen) was digested with SmaI restriction enzyme (New England Biolabs), dephosphorylated, and used for ligation with the TRAG-3 cDNA before transformation of Nov-aBlue Singles Competent Cells using a Clonables Ligation/Transformation Kit (Novagen) as recommended by the manufacturer.

HLA-DR peptide-binding assays
HLA-DR and HLA-DP were purified from homozygous EBV cell lines by affinity chromatography using the monomorphic mAb L243 and B7/21, respectively, which were coupled to protein A-Sepharose CL4B gel (Amersham Pharmacia Biotech) as previously described (30,31). The binding to the multiple HLA-DR and HLA-DP4 molecules was performed as previously reported (30 -32). Maximal binding was determined by incubating the biotinylated peptide with the MHC class II molecule in the absence of competitor. Binding specificity for each HLA-DR and HLA-DP4 molecule was ensured by the choice of the biotinylated peptides as described previously (31,32). Data were expressed as the concentration of peptide that prevented binding of 50% of the labeled peptide (IC 50 ).

Induction of CD4 ϩ T cells with peptides
The induction of CD4 ϩ T cells in vitro with dendritic cells (DCs) and the TRAG-3-derived peptides was performed as previously reported (7,9). The CD4 ϩ T cells were restimulated on a weekly basis with irradiated autologous mature DCs pulsed with peptide. After at least three restimulations, the immunoreactivity of the CD4 ϩ T cell cultures was analyzed in IFN-␥ and IL-5 ELISPOT assays. The APCs used in ELISPOT assays were either autologous T blasts or L cells that have been genetically engineered to express HLA-DR1 (L.DR1), HLA-DR4 (L.DR4), HLA-DR7 (L.DR7), HLA-DR 53 (L.DR53), or HLA-DP4 (L.DP4). The CD4 ϩ T cells were cloned by limiting dilution using allogeneic PBL and EBV-B cells as feeders in the presence of IL-2 and PHA and subsequently tested for specificity in IFN-␥ ELISPOT and cytokine release assays. The CD4 ϩ T cell clones were maintained by restimulation every 2 wk, by alternating irradiated allogeneic PBL and EBV-B cells or autologous peptide-pulsed DCs as stimulators.

In vitro sensitization with peptides
In vitro sensitization of CD4 ϩ T cells purified from PBLs of patients and normal donors was performed as previously described (33). Briefly, CD4 ϩ T cells were purified with immunomagnetic beads (Miltenyi Biotec). Two million PBLs previously depleted of CD8 ϩ and CD4 ϩ T cells, were pulsed with 10 g/ml of peptide TRAG-3 34 -48 (4 h, 37°C), irradiated (3000 rads), washed, and used to stimulate 2 ϫ 10 6 autologous CD4 ϩ T cells in Iscove's medium (Mediatech) supplemented with 10% human AB serum (Sigma-Aldrich). After 8 h of incubation, 10 U/ml IL-2 and 5 ng/ml IL-7 were added to the culture wells, and this was repeated every 3-4 days thereafter. The remaining fraction of CD4 ϩ and CD8 ϩ T cells was used to prepare T-APCs. The T-APC were harvested and used as target cells in IFN-␥ and IL-5 ELISPOT assays.

IFN-␥ and IL-5 ELISPOT assays
The recognition of APCs pulsed with peptides or proteins by the TRAG-3-specific CD4 ϩ T cells and clones expanded in vitro was assessed by ELISPOT assays specific for human IFN-␥ and IL-5 as previously reported (8,9). The protein-loaded DCs were prepared as previously reported in the presence of the recombinant TRAG-3 or NY-ESO-1 proteins (30 g/ml). In the blocking experiments, 3 g each of purified anti-HLA mAb were added in each ELISPOT well.
The recognition of T-APCs pulsed with peptides by CD4 ϩ cells isolated from cancer patients and in vitro sensitized as previously in this article was assessed by IFN-␥ and IL-5 ELISPOT assays. Briefly, 10 5 presensitized CD4 ϩ T effector cells and 10 5 targets (T-APCs) pulsed with peptides (10 g/ml) were added to each well and incubated for 48 h in AIM V medium (Invitrogen). Spot numbers and spot sizes were determined with computerassisted video image analysis (Cellular Technologies). For statistical evaluation, a Student t test for unpaired samples was used. p Ͻ 0.05 was considered significant.

ELISA
Recombinant proteins were diluted in coating buffer (15 mM Na 2 CO 3 , 30 mM NaHCO 3 (pH 9.6), with 0.02% NaN 3 ) to the concentration of 1 g/ml and adsorbed to polypropylene flat-bottom 96-well plates (Corning-Costar) overnight at 4°C. Plates were washed with PBS and saturated overnight at 4°C by addition of 200 l/well PBS containing 2% BSA. After the plates were washed, serum was diluted in PBS-2% BSA buffer and added to the plates (50 l/well). After 2 h of incubation at room temperature, plates were washed, the secondary Ab (goat anti-human IgG-AP; Caltag Laboratories) was added (50 l/well), and the plates were further incubated for 1 h at room temperature. Following additional washes, 50 l/well substrate solution (phosphatase substrate system; Kirkegaard & Perry Laboratories) were added, and plates were incubated for 25 min at room temperature and read immediately (MRX Microplate Reader; Dynatech). Sera were tested over a range of 4-fold dilutions as previously described (34).

TCR␤ gene usage
Total RNA was isolated from 1 ϫ 10 6 CD4 ϩ T cells using the RNeasy Mini Kit (Qiagen) and reverse transcription was conducted as previously reported. A panel of 24 specific primers complementary to TCR ␤-chain variable and a primer for the constant region were used to determine the diversity of the TCR ␤-chain variable repertoire (35). The PCR amplification was conducted in 50 l as follows: 5 min at 94°C for 5 min, then 40 cycles (1 min at 94°C, 1 min at 60°C, 1 min at 72°C) and finally 10 min at 72°C. The PCR products were run on 1.5% agarose gel and purified using the QIAquick gel extraction kit (Qiagen, Valencia, CA). The PCR product was cloned into pCR4-TOPO vector (Invitrogen, Carlsbad, CA) and sequenced using the ABI 3100 automated DNA sequencer.

Peptide TRAG-3 34 -48 is a novel promiscuous epitope that binds to multiple HLA-DR molecules
Based on the published TRAG-3 cDNA sequence, we have designed specific oligonucleotide primers and used proof-reading polymerase to clone the TRAG-3 cDNA from three different melanoma cell lines as previously described in Materials and Methods. We have found that all three cDNA sequences were identical and slightly different from the previously reported TRAG-3 cDNA: G instead of A at position 251; T instead of A at position 263; A instead of T at position 264; and A instead of G at position 340 (23). These nucleotide differences result in a 3-aa difference at the protein level: R instead of H at position 33; L instead of H at position 37; and T instead of A at position 63. On the basis of the new TRAG-3 sequence, we have chosen 15 overlapping 15-mer peptides. All the peptides include an aliphatic or aromatic residue in positions 1-5 because their presence at these positions is a requisite for the binding to HLA-DR and HLA-DP4 molecules. Indeed, aliphatic and aromatic residues are accommodated by the P1 pocket of these molecules and constitute the main anchor residue (32,36). We have then synthesized the 15 overlapping peptides and evaluated their binding capacities to 10 different HLA-DR and 2 HLA-DP4 molecules including the 7 molecules encoded by the HLA-DRB1 genes (i.e., HLA-DRB1*0101, HLA-DRB1*0301, HLA-DRB1*0401, HLA-DRB1*0701, HLA-DRB1*1101, HLA-DRB1*1301, and HLA-DRB1*1501), three molecules encoded by the HLA-DRB3, HLA-DRB4, and HLA-DRB5 genes and the HLADPB1*0401 and -0402 molecules. The data are presented in Table I.
We next performed a series of additional experiments to investigate the immunogenicity of these peptide sequences.
Only peptide TRAG-3 34 -48 was capable of stimulating IFN-␥producing CD4 ϩ T cells derived from PBLs of the two melanoma patients and the two normal donors (Fig. 1).
Bulk CD4 ϩ T cells isolated from patient 1 (UPCI-MEL 285) and stimulated with peptide TRAG-3 34 -48 produced both IFN-␥ and IL-5 spots in the presence of L.DR1 cells and L.DR4 cells pulsed with peptide TRAG-3 34 -48 but not with an irrelevant peptide (NY-ESO-1 119 -143 ) (Fig. 2, A and B). These CD4 ϩ T cells also displayed reactivity against autologous DCs previously loaded with the TRAG-3 protein but not with an irrelevant protein (NY-ESO-1). This reactivity was partially inhibited by the addition of the anti-HLA-DR mAb (L243) but not the anti-HLA-A, B, C mAb (w6/32).
Bulk CD4 ϩ T cells isolated from normal donor 1 and stimulated with peptide TRAG-3 34 -48 peptide specifically recognized L.DR7 cells but not L.DR11 cells or L.DP4 cells pulsed with peptide TRAG-3 34 -48 in IFN-␥ and IL-5 ELISPOT assays (Fig. 2, C and  D). The CD4 ϩ T cells failed to produce significant amount of IFN-␥ or IL-5 in the presence of APCs pulsed with peptide NY-ESO-1 119 -143 , used as an irrelevant peptide. These CD4 ϩ T cells displayed reactivity against autologous DCs previously loaded with the TRAG-3 protein that was inhibited by addition of the anti-HLA-DR mAb (L243) but not the anti-HLA-A, B, C mAb (w6/32). These CD4 ϩ T cells did not recognize autologous DCs loaded with the irrelevant protein NY-ESO-1 protein.
Bulk CD4 ϩ T cells isolated from normal donor 2 and stimulated with peptide TRAG-3 34 -48 peptide produced IFN-␥and IL-5 in the presence of L.DR4 cells but not L.DP4 cells pulsed with peptide TRAG-3 34 -48 (Fig. 2, D and E). The CD4 ϩ T cells failed to produce significant amount of IFN-␥ or IL-5 in the presence of APCs pulsed with peptide NY-ESO-1 119 -143 , used as an irrelevant peptide. These CD4 ϩ T cells displayed reactivity against autologous DCs previously loaded with the TRAG-3 protein but not with an irrelevant protein (NY-ESO-1). This reactivity was inhibited by addition of the anti-HLA-DR mAb (L243) but not the anti-HLA-A, B, C mAb (w6/32).
Peptide TRAG-3 34 -48 stimulated HLA-DRB1*0101 and -DRB1*0701-restricted CD4 ϩ T cell clones from a melanoma patient and a normal donor that recognized autologous protein-pulsed DCs Several clones were obtained by limiting dilution from the CD4 ϩ bulk T cells of patient 1 and normal donor 1 that recognized the TRAG-3 34 -48 peptide.
Clone 61/58 derived from PBLs of patient 1 specifically produced IFN-␥ but not IL-5 in the presence of L.DR1 cells pulsed with peptide TRAG-3 34 -48 and autologous DCs loaded with the TRAG-3 protein in IFN-␥ ELISPOT assays (Fig. 3A). The reactivity of this Th1-type CD4 ϩ T cell clone against autologous TRAG-3-loaded DCs was partially inhibited by the addition of the anti-HLA-DR mAb (L243) but not the anti-HLA-A, B, C mAb (w6/32). No significant amount of IFN-␥ production was observed in the presence of peptide TRAG-3 34 -48 alone, other L.DR cells pulsed with peptides, or autologous DCs fed with an irrelevant protein (NY-ESO-1).
Clone 62/4 derived from PBLs of normal donor 1 produced not only IFN-␥ but also IL-5 in ELISPOT assays in the presence of L.DR7 pulsed with peptide TRAG-3 34 -48 and autologous DCs fed with the TRAG-3 protein (Fig. 3, C and D). No significant amount of IFN-␥ production was observed in the presence of peptide TRAG-3 34 -48 alone, other L.DR cells pulsed with peptides, or autologous DCs fed with an irrelevant protein (NY-ESO-1). The ability of this Th0-type CD4 T cell clone to produce IFN-␥ in the presence L.DR7 cells, preincubated with various concentrations of peptide TRAG-3 34 -48 , was evaluated. Half-maximal stimulation of clone 62/4 required peptide loading concentrations of ϳ600 nM (Fig. 3E).
We also observed that the peptide sequence TRAG-3 34 -48 37H (with H instead of L at position 37) encoded by the previous reported TRAG-3 cDNA sequence (23) was not recognized by our CD4 ϩ T cell clones (Fig. 3, A, C, and D).  Table I. Ten thousand of the responder CD4 ϩ T cells were incubated in 48-h IFN-␥ ELISPOT assays in the presence of T blasts pulsed with either the relevant TRAG-3 peptide (p) or peptide NY-ESO-1 119 -143 as control (10 g/ml). Spots were developed and counted by computer-assisted video image analysis. Each symbol represents the mean spot number of triplicates with 10 4 CD4 ϩ T cells initially seeded per well.

Detection of TRAG-3-specific IFN-␥ producing CD4 ϩ T cells in the peripheral blood of melanoma and breast cancer but not lung cancer patients
We next wanted to address whether TRAG-3 could stimulate spontaneous CD4 ϩ T cell and Ab responses from the blood of patients with TRAG-3-expressing tumors. Therefore, we investigated the blood of 12 patients with melanoma, 13 patients with breast cancers, 14 patients with lung cancers, and 9 normal donors for the presence of CD4 ϩ T cells capable of recognizing peptide TRAG-3 34 -48 and peptide HA 60 -73 as control (Table II). CD4 ϩ T cells were first isolated from PBLs and stimulated with peptide-pulsed APCs for one round of in vitro stimulation before IFN-␥, IL-5, and IL-10 ELISPOT assays, as described in Materials and Methods.
We observed TRAG-3-specific CD4 ϩ T cells from 9 of 12 patients with stage IV melanoma (Fig. 4A). All of the nine patients with TRAG-3-specific CD4 ϩ T cells had TRAG-3-expressing melanoma as assessed by RT-PCR and shown in Table II. These TRAG-3-specific CD4 ϩ T cells were Th1-type CD4 ϩ T cells for six of nine patients and Th0-type CD4 ϩ T cells for the remaining three patients. We also detected HA-specific CD4 ϩ T cells in 11 of the 12 melanoma patients (Fig. 4D). Four of thirteen patients with breast cancers had circulating IFN-␥-producing (i.e., Th-1 type) TRAG-3-specific CD4 ϩ T cells (Fig. 4B). Two of the patients with TRAG-3-specific CD4 ϩ T cells had stage IV disease, and two had stage III disease. HA-specific CD4 ϩ T cell responses were evaluated in 9 of the 13 breast cancer patients and were detectable in 3 patients (Fig. 4E). None of the 14 patients with stage IV NSLCC had detectable TRAG-3-specific CD4 ϩ T cells (Fig. 4C). HA-specific CD4 ϩ T cell responses were analyzed from nine patients and were detectable in only two patients (Fig. 4F). Of note and as controls, none of the nine normal donors had detectable TRAG-3-specific CD4 ϩ T cells (Fig. 4, A-C). HA-specific CD4 ϩ T cells were detectable in seven of nine normal donors (Fig. 4, D-F).
We next investigated whether our immunological findings were correlated with taxane therapy (Table II). Eight of the thirteen breast cancer patients received taxanes, but only two patients previously treated with taxanes developed TRAG-3-specific CD4 ϩ T cells. Twelve of fourteen lung cancer patients were previously treated with taxanes; none of them developed TRAG-3-specific CD4 ϩ T cells. Most importantly, none of the nine melanoma patients with TRAG-3-expressing tumors and TRAG-3-specific CD4 ϩ T cells had received taxanes.
Altogether, our data demonstrate the existence of spontaneous TRAG-3-specific CD4 ϩ T cells in all stage IV melanoma patients with TRAG-3-expressing melanoma and in a limited number of patients with stage III/IV breast cancers but not in any of the patients with stage IV lung cancers. Furthermore, no circulating TRAG-3-specific Abs have been detected from the blood of these patients, although some of these patients had NY-ESO-1-specific CD4 ϩ T cell responses that coexist with the presence of circulating anti-NY-ESO-1 Abs.

TCR␤ CDR3 studies of TRAG-3-specific CD4 ϩ T cell clones
To study the diversity of the T cell repertoire of the TRAG-3specific CD4 ϩ T cells, we have cloned and sequenced the TCR␤ CDR3 gene region of 10 different CD4 ϩ T cell clones isolated either from patient 1 or normal donor 1 as previously described in Materials and Methods. The data are presented in Table III. We observed that the 10 clones isolated from patient 1 exhibited 1 unique TCR␤ CDR3 region, whereas the TRAG-3-specific CD4 ϩ T cell clones derived from normal donor 1 expressed 3 distinct TCR␤ CDR3 regions. Altogether, our data suggest a limited TCR repertoire of TRAG-3-specific CD4 ϩ T cells in patients with TRAG-3-expressing tumors.
Importantly, TRAG-3-specific CD4 ϩ T cells recognized autologous DCs previously loaded with the TRAG-3 protein but not with an irrelevant protein, thus indicating that peptide TRAG-3 34 -48 encompasses one naturally processed and presented epitope. Our findings contrast with previous studies showing that CGAs usually give rise to multiple MHC class II epitopes recognized by CD4 ϩ T cells (8 -10, 12, 17, 18), reporting the first example of tumor Ag-specific CD4 ϩ T cell responses directed against a single immunodominant epitope. The selectivity of the T FIGURE 4. TRAG-3-specific Th1type vs Th2-type immunoreactivity of CD4 ϩ T cells in patients with melanoma, breast, and lung cancers. CD4 ϩ T cells were isolated from the peripheral blood of patients with melanoma, breast, and lung cancers and normal donors and then stimulated with autologous APCs pulsed with peptide TRAG-3 34 -48 , peptide HA 60 -73 , peptide NY-ESO-1 87-111 , or peptide NY-ESO-1 157-170 . CD4 ϩ T cells were then analyzed in IFN-␥ and IL-5 ELISPOT assays for their reactivity to autologous T-APCs pulsed with: peptide TRAG-3 34 -48 (A-C); peptide HA 60 -73 (D-F), for melanoma, breast cancer, and lung cancer patients, respectively. Each symbol represents the combined IFN-␥and IL-5 ELISPOT data for each individual patient with melanoma, breast cancer, or lung cancer (F) and each normal donor (E). ELISA for NY-ESO-1-specific Abs were performed as previously reported in the section Materials and Methods (all data Ͻ1/400 are considered negative). We have also presented the combined data of anti-NY-ESO-1 Ab and IFN-␥ ELISPOT assays from melanoma patients and normal donors for: peptide NY-ESO-1 87-111 (F) and peptide NY-ESO-1 157-170 (G). Each symbol represents the combined IFN-␥ELISPOT and NY-ESO-1-specific Abs data for each individual melanoma patient (F) or normal donor (E). cells to respond to a very limited number of peptides within complex protein Ags is called immunodominance (37). A number of mechanisms appear to contribute to immunodominance including events related to Ag processing and presentation as well as T cell recognition (38). Our binding data on TRAG-3, showing that a single amino acid region is capable of binding to multiple HLA-DR molecules, strongly support the pivotal role of HLAbinding properties for the determinant selection of the immunodominant peptide, TRAG-3 34 -48 (37,39). Furthermore, our data support the hypothesis that immunodominance is correlated with binding to multiple MHC class II molecules. The correlation between highly degenerate HLA-DR binding and immunodominance has been previously established for a number of epitopes derived from foreign and self-Ags (40 -45). Such a correlation is also suggested by a recent study in cancer patients showing that the spontaneous NY-ESO-1-specific CD4 ϩ T cell responses are directed toward two immunodominant and promiscuous HLA-DR-restricted peptides, NY-ESO-1 81-100 and NY-ESO-1 119 -143 , previously identified by us and others (8,9,18,46).
There is now ample evidence that some CGAs may spontaneously give rise to T cell and Ab responses in patients with CGAexpressing tumors (18,19,(47)(48)(49). Whenever detectable, spontaneous immune responses to CGAs were found only in a fraction of the patients with CGA-expressing tumor. In contrast, our findings show that all patients with TRAG-3-expressing stage IV melanoma had spontaneous TRAG-3-specific CD4 ϩ T cells, supporting the strong immunogenicity of TRAG-3. This observation may not apply to all types of CGA-expressing tumors as spontaneous CD4 ϩ T cell responses against TRAG-3 were found in only four tested stage III/IV breast cancer patients and in none of the tested patients with stage IV NSCLC. We have not investigated TRAG-3 expression in the primary tumor or metastases of the breast cancer and NSCLC patients and cannot rule out that few breast tumors and none of the 14 lung cancer patients' tumors expressed TRAG-3. However, this seems very unlikely given that previous works have demonstrated that ϳ60 and 54% of breast cancers and NSLCC express TRAG-3, respectively (21,22). Furthermore, as only three of nine and two of nine patients with breast cancers and stage IV NSCLC, respectively, developed HA-specific CD4 ϩ T cells, we also cannot exclude that the capability of the Ag-specific CD4 ϩ T cells to produce cytokines may have been impaired because of their clinical status or ongoing chemotherapy.
To investigate a correlation between the presence of TRAG-3specific CD4 ϩ T cells and TRAG-3-specific IgG Ab, we studied the occurrence of spontaneous humoral responses to TRAG-3 and NY-ESO-1 (as a control) with ELISA. Whereas the stage IV melanoma patients in this study with naturally occurring NY-ESO-1specific CD4 ϩ T cells had circulating Abs against NY-ESO-1, none of the patients with spontaneous TRAG-3-specific CD4 ϩ T cells had detectable Ab responses against TRAG-3. Our observation of IFN-␥-producing CD4 ϩ T cells in the absence of circulating TRAG-3-specific Abs clearly indicates that spontaneous CGAspecific CD4 ϩ T cells may exist independently of the presence of CGA-specific circulating Abs, revealing a novel profile of spon-taneously acquired immune response against CGAs. Interestingly, four of the melanoma patients with TRAG-3-specific CD4 ϩ T cells also had circulating NY-ESO-1-specific CD4 ϩ T cells and NY-ESO-1-specific Abs, suggesting that CGA-specific CD4 ϩ T cells within the same individual may differ in their capability to activate B cells and promote CGA-specific IgG Abs.
Two previous experimental studies in animals have suggested that the diversity of the CD4 ϩ T cell repertoire plays a critical role in the generation of isotype-switched IgG autoantibodies (50,51). Another study in humans has reported the existence of Th2-type CD4 ϩ T cells directed against a single immunodominant epitope from mugwort pollen in allergic patients, which coexisted with Ag-specific IgE Abs and exhibited a large TCR repertoire (52). These observations led us to examine the TCR repertoire of the TRAG-3-specific CD4 ϩ T cells derived from PBL of either a melanoma patient with TRAG-3-expressing tumor or a normal donor. We observed that all 10 clones derived from the melanoma patient exhibited the same TCR␤ CDR3 region, whereas 3 distinct TCR␤ CDR3 regions were found among the 10 CD4 ϩ T cell clones derived from the normal donor. Although we cannot exclude a selective in vitro expansion of one specific CD4 ϩ T cell clone that occurred only for the melanoma patient, our data suggest an in vivo selection of one specific T cell clonotype in the melanoma patient with TRAG-3-expressing tumor. It is thus possible that the help provided by the TRAG-3-specific CD4 ϩ T cells is of limited specificity and less likely to promote a sustained effector B cell response and Ab production in patients with TRAG-3-expressing tumors. Further TCR studies of TRAG-3-specific CD4 ϩ T cells from a larger number of patients with TRAG-3-expressing tumors are needed to confirm our findings in one melanoma patient.
Whenever detectable, the TRAG-3-specific CD4 ϩ T cells obtained from patients with melanoma and breast cancers were either Th1-type or Th0-type CD4 ϩ T cells because they produced IFN-␥, or IFN-␥ and IL-5, respectively. We have not identified TRAG-3-specific CD4 ϩ T cells or clones that produced only Th-2-type cytokines, suggesting no Th2 skewing of the spontaneous TRAG-3-specific CD4 ϩ T cell responses as previously reported for MAGE-6 (53). These findings are in line with our previous studies of NY-ESO-1-specific CD4 ϩ T cells in patients with active NY-ESO-1-expressing melanoma (12). They illustrate that Th2 skewing is not a phenomenon that is generalized for all CGA-specific CD4 ϩ T cell responses in cancer patients.
The majority of the breast cancer and lung cancer patients in this study have been treated with taxane-based chemotherapy, which might be expected to induce TRAG-3 expression by the patients' tumor. However, and as indicated in Table II, we have not found that taxane-based treatment induced spontaneous immune responses against TRAG-3 in vivo. Furthermore, none of the patients with stage IV melanoma in this study with TRAG-3-expressing tumors and TRAG-3-specific CD4 ϩ T cells had previously received taxanes. These findings strongly suggest that taxanes play no major role in the expression of TRAG-3 or in the induction of TRAG-3-specific CD4 ϩ T cells. In summary, this study reports a novel immunological pattern of CGA-specific immune responses and adds to the diversity of spontaneous immune responses in patients with CGA-expressing tumors. Because of the common expression of TRAG-3 by melanoma and its strong spontaneous immunogenicity in patients with stage IV melanoma, the TRAG-3 34 -48 epitope represents a good candidate for the monitoring and the dissection of tumor Ag-specific spontaneous immune responses in patients with advanced melanoma.