Therapeutic Effectiveness of Recombinant Cancer Vaccines Is Associated with a Prevalent T-Cell Receptor α Usage by Melanoma-specific CD8⁺ T Lymphocytes

Tyrosinase-related protein (TRP)-2 is a melanosomal enzyme defined as a melanocyte differentiation antigen because it is expressed in most mammalian melanocytes and melanomas. TRP-2 is the main antigenic target of the immune response elicited in mice by immunization with genetically modified B16 melanoma vaccine (1, 2). Although slight differences have been described under therapeutic versus prophylactic conditions (2), the mouse immune response to TRP-2 is dominated by the generation of CD8⁺ T lymphocytes responsible for both the antitumor response and autoimmunity, i.e., skin depigmentation (vitiligo) due to normal melanocyte destruction (3). Tumor-infiltrating lymphocytes isolated from melanoma patients also recognize TRP-2 peptides in the context of different class I HLA alleles. Mouse and human TRP-2 have about 80% homology at the protein level, and the peptide SVYDFFVWL (TRP-2_180–188) is presented in association with mouse H-2 molecule K^b and human HLA molecule A2 (1, 4, 5).

Peripheral tolerance, which seems to limit the anti–TRP-2 response in the C57BL/6 strain, can be broken by xenoimmunization, or immunization with an altered source of antigen. Immunization with a recombinant adenovirus (rAd) encoding human TRP-2, in fact, elicited an immune response against the respective mouse homologue and completely protected C57BL/6 mice from a lethal melanoma challenge (6). Tolerance could be also broken by linking the mouse self-antigen with a foreign immunogenic protein providing strong CD4 helper sequences, such as the fusion protein between the mouse TRP-2 gene product and the enhanced green fluorescent protein (EGFP) of jellyfish Aequorea victoria (7). Whereas the induction of an effective immune response against TRP-1/gp75 was often accompanied by vitiligo, immunity to TRP-2, delivered by either plasmid DNA or recombinant viruses, was not always associated with widespread vitiligo (8, 9). These interesting findings delineate a window of opportunity between proautoimmune and therapeutic activity that could be exploited to promote the therapeutic effects of recombinant vaccines based on TRP-2.

Based on encouraging preclinical studies, melanoma-specific vaccines were designed and evaluated in clinical trials (10). Unfortunately, partial or complete tumor regression was reported only in a minority of patients (11). This was somewhat expected because most initial trials were conducted in patients with advanced metastatic disease. These trials were designed to identify surrogate end points with respect to vaccine efficacy in the absence of overt tumor regression. The most confounding results of cancer vaccine trials concern the absence of a clear-cut correlation between the clinical responses and the antigen-specific immune response detected in patient-derived T lymphocytes. In fact, vaccination with class I major histocompatibility complex (MHC)-restricted peptides can easily generate tumor-specific CD8⁺ T cells among the circulating lymphocytes of immunized patients, but there is no assay that unambiguously identifies those patients who will respond clinically to immunotherapy (11, 12).

To simultaneously analyze multiple aspects of the TRP-2–specific, CD8⁺ T lymphocyte-dependent response, we designed an experimental protocol to evaluate different immune variables in single mice after prophylactic or therapeutic immunization. In prevention experiments, after immunization with recombinant vaccines encoding TRP-2 antigen, mice underwent splenectomy. After recovering from surgery, they were challenged with a lethal intravenous inoculum of B16 melanoma cells to monitor tumor development. In the therapeutic model, splenectomy was performed in mice that had been previously inoculated with tumor cells and then vaccinated with recombinant vaccines. Fresh splenocytes were used to quantify the number of TRP-2–specific CD8⁺ T cells by cytofluorometry after staining with H-2 class I tetramers and to enumerate the antigen-specific effectors releasing interferon (IFN)-γ in an enzyme-linked immunosorbent spot (ELISPOT) assay. Moreover, by stimulating the antigen-specific T lymphocytes with the K^b-restricted, TRP-2_180–188 peptide for 5 days, peptide-stimulated mixed leukocyte cultures (MLPCs) were designed to expand limited numbers of TRP-2–reactive T lymphocytes. Finally, we studied the T-cell repertoire usage in tetramer-sorted, TRP-2–specific T lymphocytes recovered from peptide-stimulated cultures.

C57BL/6 (H-2^b) mice (8 weeks old) were purchased from Charles River (Calco, Como, Italy). Animal care and procedures followed institutional guidelines, in compliance with national and international regulations. Mice used for in vivo tumor growth experiments were examined every day and euthanized when the tumor became ulcerated or when one of two perpendicular diameters reached 10 mm and their product was >50 mm². MBL-2 is a leukemia line (H-2^b) derived from a Moloney murine leukemia virus-infected C57BL/6 mouse; B16 is a melanoma line (H-2^b) spontaneously growing in C57BL/6 mice (provided by Dr. I. J. Fidler; M. D. Anderson Cancer Center, Houston, TX). B16LU8, a metastasizing melanoma cell line derived from repeated in vivo passage of B16, was kindly provided by Dr. James C. Yang (Surgery Branch, National Institutes of Health, Bethesda, MD). The 293-K^b cell line is a human embryonic kidney cell line stably transfected with plasmid expressing the H-2 K^b class I molecule. Cell lines were maintained in Dulbecco’s modified Eagle’s medium (Invitrogen, Carlsbad, CA) supplemented with 2 mmol/L l-glutamine, 10 mmol/L HEPES, 20 μmol/L 2-mercaptoethanol, 150 units/ml streptomycin, 200 units/ml penicillin, and 10% heat-inactivated fetal bovine serum (Invitrogen). TRP-2–specific cytotoxic T lymphocyte (CTL) clones 8 and 24 were obtained from a C57BL/6 mouse immunized with pcDNA3-trp-2 plasmid, as described in ref. 8, by several in vitro restimulations with syngeneic splenocytes pulsed with TRP-2_181–188 peptide (10 μmol/L) and after limiting dilution cloning.

Adenoviruses used in this study were constructed through Cre-lox recombination with reagents generously provided by Dr S. Hardy (Somatix, Alameda, CA). Adenoviruses were propagated on 293 cells, purified by cesium chloride density gradient centrifugation, dialyzed according to standard protocols (6, 7), and then stored at −70°C. All vectors used express only the antigen of interest under the cytomegalovirus immediate early promoter. The recombinant vaccinia virus (rVV) encoding the human gp100 epitope as minigene was a kind gift of Dr. Nicholas P. Restifo (National Institutes of Health, Bethesda, MD). Immunization was performed as described previously (13).

Preparation of the VR1055-P15 plasmid encoding the p15E portion of the env gene has been described previously (14). Plasmid amplification was performed with the Endofree plasmid mega kit (Qiagen, Hilden, Germany), following the manufacturer’s instructions. Plasmids were purified with Qiagen columns assuring an endotoxin-free DNA preparation (Qiagen GmbH, Hilden, Germany). For DNA immunization, C57BL/6 mice were anesthetized by avertin inoculation and received intramuscular injection with 100 pmol/L cardiotoxin (Latoxan, Rosans, France). Five days later, mice received intramuscular injection with 100 μg of plasmid DNA in 100 μL of saline.

Mice were depleted of either CD4⁺ or CD8⁺ T cells by four intraperitoneal injections of 200 μg of affinity chromatography-purified GK1.5 (anti-CD4) or 2.43 (anti-CD8) monoclonal antibody (mAb). Depleting mAbs were administered 2 days before and 0, 4, and 8 days after subcutaneous challenge with tumor cells. In therapy experiments, the mAbs were administered 1 day before and 0, 4, and 8 days after vaccination with rAd. Depletion was consistently >98%. As a control, anti-Escherichia coli β-galactosidase (GL117.14; rat IgG2a) was used at the same dose.

Three weeks after plasmid DNA inoculation, spleens were removed, and 2.5 × 10⁷ splenocytes were stimulated in vitro in a MLPC with 1 μmol/L of a nonamer peptide corresponding to amino acids 180 to 188 of TRP-2 protein (SVYDFFVWL). K^b-restricted peptides corresponding to amino acids 604 to 611 of p15E protein (KSPWFTTL, p15E peptide) and control peptide corresponding to the amino acids 96 to 103 of β-galactosidase (β-gal) protein (DAPIYTNV) were synthesized and purified by Technogen (Naples, Italy) and were >95% pure, as indicated by analytical high-performance liquid chromatography. The peptide KVPRNQDWL, spanning amino acids 25 to 33 of human gp100, was a kind gift of Dr. Nicholas P. Restifo. The cultures were set up in 10 mL of Dulbecco’s modified Eagle’s medium-10% fetal bovine serum, maintained in 25-cm² tissue culture flasks (Falcon; Becton Dickinson, Lincoln Park, NJ) for 5 days at 37°C under 5% CO₂, and then tested in either IFN-γ or ⁵¹Cr release assay.

Three weeks after immunization, mice were challenged subcutaneously with a lethal dose of B16 melanoma cells (2 × 10⁴) and then monitored for 100 days after tumor injection. Tumor growth was monitored every 3 days by caliper measurement. Alternatively, C57BL/6 mice were inoculated intravenously with 10⁵ B16LU8 cells. Lungs were removed 14 days after challenge, and pulmonary metastases were counted in a blind fashion. In therapy experiments, the same dose of tumor cells was used, but tumor was injected 3 days before immunization. All the in vivo experiments were conducted in mice randomized before tumor injection or before treatment for the therapeutic model. Mice were bred in filtered cages placed inside biohazard closets. For the adoptive transfer, B6 mice were inoculated with 10⁵ B16LU8 tumor cells via tail vein injection on day 0. On day 3, 5 × 10⁶ TRP-2–specific CTLs were administered intravenously, and 30,000 IU of recombinant interleukin (IL)-2 were given intraperitoneally twice a day for 3 days.

Soluble H-2–peptide tetramers were produced using a previously described method (14). Soluble purified complexes were biotinylated using BirA enzyme (Avidity, Denver, CO). Phycoerythrin (PE)-labeled tetramers were produced by mixing the biotinylated complexes with Extravidin-PE (Sigma, St. Louis, CO) and validated by staining CTL clones with the appropriate specificity. Each tetramer batch was titrated and used at the optimum concentration (5 μg/mL) of K^b heavy chain.

Fresh or in vitro stimulated splenocytes (10⁶ per sample) were resuspended in 50 μL of fluorescence-activated cell sorting (FACS) buffer (0.9% NaCl solution containing 2% bovine serum albumin and 0.02% NaN₃; both from Sigma) with antimouse Fc-γ receptor 2.4G2 mAb (ATCC HB-197) for 10 minutes at room temperature to reduce the nonspecific staining. Cells were further labeled with either PE-conjugated TRP-2 tetramer-PE (TRP-2-TET, 5 μg/mL) or β-gal tetramer-PE (β-gal-TET) for 20 minutes at room temperature. Each sample was then stained at 4°C with rat antimouse CD8-Tricolor (0.1 μg per 10⁶ cells; clone CTCD8α; Caltag, Burlingame, CA) and with hamster antimouse CD3-fluorescein isothiocyanate (1 μg per 10⁶ cells; clone 145-2C11; Caltag). Before analysis, cells were washed twice, resuspended in FACS buffer, and analyzed with a FACSCalibur flow cytometer (Becton Dickinson, San Jose, CA). Data were analyzed using Cell Quest software (Becton Dickinson). The triple-positive cells (CD3⁺/CD8⁺/TRP-2-TET⁺) were sorted to obtain highly purified populations using a FACSVantage DiVa (Becton Dickinson) equipped with a 488-nm argon laser (I 305C-BD; Innova Coherent, Santa Clara, CA), at a rate of 8,000 to 12,000 cells per second. The sorted populations used in each experiment were >97% pure. Adequate controls regarding the cell viability were performed using light-scattered parameters and propidium iodide/annexin V staining. The triple-positive dead cells were excluded from the sorting procedure using light-scattered parameter in the back gate approach.

The ⁵¹Cr-labeled target cells (2,000 cells per well) were incubated with effector cells at various effector to target ratios in 96-well microplates (Falcon; Becton Dickinson). After a 6-hour incubation at 37°C, supernatants were harvested, and radioactivity was counted in a microplate scintillation counter (Top-Count; Packard Instruments Co., Meriden, CT). For peptide pulsing, ⁵¹Cr-labeled target cells (10⁶ per mL) were incubated with relevant peptides (1 μmol/L, final concentration) for 30 minutes at 37°C and then washed twice before use. In cold target inhibition assays, 5 × 10⁶ splenocytes from mice immunized with rAd were stimulated in vitro with 10⁵ γ-irradiated syngeneic MCA38 cells, which were previously infected for 18 hours with Ad-human TRP (hTRP)-2 at a multiplicity of infection = 100, in the presence of 10 IU of recombinant IL-2. For the assay, unlabeled inhibitor B16, YAC-1, β-gal–loaded 293K^b cells, or TRP-2–loaded 293K^b cells were seeded together with labeled B16 cells at different cold to hot target ratios. CTLs from the cultures were then added at an effector to hot target ratio of 100. The percentage of inhibition by cold targets was calculated as follows: inhibition (%) = 100 × [1 − lysis of CTLs with cold targets (%)/lysis of CTLs in the absence of cold targets (%)].

Splenocytes (10⁵ cells) from MLPCs were restimulated for 24 hours in triplicate wells with an equal amount of target cells; the supernatants were harvested and tested for the IFN-γ released in a sandwich enzyme-linked immunosorbent assay (ELISA) assay (Endogen, Boston, MA).

The mouse IFN-γ development Module System Kit (R&D System Inc., Minneapolis, MN) was used according to the manufacturer’s instructions. Development was performed with the ELiSpot Blue color module (R&D System Inc.), and the number of spots was counted blindly by two operators under a dissecting microscope and expressed as the mean number of spots ± SE of triplicate determinations. Each mean value was subtracted from the one derived from effector cells cultured with target cells pulsed with an irrelevant peptide.

RNA was extracted using guanidium hydrochloride-containing Trizol reagent (Life Technologies, Inc., Gaithersburg, MD) according to the manufacturer’s instructions. First-strand cDNA synthesis was performed using oligo(dT) as a primer for reverse transcription of 1 μg of total RNA using Moloney murine leukemia virus reverse transcriptase (Life Technologies, Inc.) as described previously (15).

Polymerase chain reaction (PCR) amplification was performed as reported in detail elsewhere (15). Briefly, after titration of the different templates, cDNAs were amplified for 30 cycles under nonsaturating PCR conditions. The T-cell receptor (TCR) repertoire was analyzed using a panel of TCR BV and AV family-specific primers by PCRs as described previously (15, 16). In each PCR reaction, the common TCR-α or -β primer was labeled at the 5′ end with 5′ 6-carboxyfluorescein. In addition to the TCR BV or AV family-specific and constant primers, the single reaction also contained a β-actin–specific primer pair amplifying a 6-carboxyfluorescein–labeled 320-bp product as an internal PCR control for verification of cDNA integrity and the fidelity of the single PCR reactions (17). The TCR AV5 repertoire was also screened by a PCR reaction with a specific AJ primer in place of the TCR Constant α.

After the amplification, the TCR-CDR3 length analysis was evaluated by the method known as spectratyping or immunoscope (15, 16, 17, 18). According to this technique, the composition of each BV TCR family is visualized by running the PCR products for the different TCR families on a sequencing gel (15, 16, 17, 18). Normally, each TCR family is resolved by this technique as a series of bands having a Gaussian distribution. Each alteration, in either the distribution or intensity of single bands, represents a perturbation in the given BV TCR family (15, 16, 17, 18). In brief, an equivalent volume of PCR-labeled product was mixed with formamide dye-loading buffer and 0.2 μL of Rox-labeled size marker (Applied Biosystems, Foster City, CA), heated at 94°C for 2 minutes, and run on a sequencing gel in a fluorescence-based DNA sequencer (377 abi; Applied Biosystems). The data were analyzed by means of Genescan software (Applied Biosystems), which allows for the visualization of gel bands as chromatograms and assigns size and peak areas to different PCR products Sequencing of single TCR AV5 spectratype bands was performed using an internal primer and automated sequencing according to an established protocol (15).

The clonotypic analysis and frequency analysis of the given clonotype among the tetramer-enriched lymphocytes was analyzed, as described recently (19, 20), using a primer based on the AV5 sequence found in mouse T-cell clone 24 recognizing B16 melanoma. Briefly, real-time PCR was performed in a 5700 Thermal Cycler (Applied Byosistems), using the 5′ nuclease assay. For the clonotypic analysis, we designed a primer specific for the CDR3 region of the whole TCR AV plus a primer specific for TCR AV5. All of the primers were designed using Primer Express software (Applied Byosistems). To assess the frequency of the clonotypic transcripts among the transcripts for the entire TCR AV repertoire, two series of quantitative PCRs were carried out in parallel, using the AV5 primer and the TaqMan probe, and either the clonotypic primer or the Constant AV primer. The corresponding threshold cycles (CT) for the clonotype (CTc) and the whole AV (Cta) were measured. Clonotype frequency was determined as the ratio f (clonotype) = Corr × 2 Cta − CTc, where Corr is a correcting factor accounting for the difference in amplification efficiency of different primer pairs (19).

The Wilcoxon Mann-Whitney U test was used to examine the null hypothesis of rank identity between two sets of data. Kaplan-Maier plots and the Mantel-Haenszel test were used to compare survival of mice belonging to different treatment groups. Spearman rank correlation, a distribution-free analog of correlation analysis, was applied to compare two independent discrete random variables. All P values presented are two-sided.

Xenoimmunization with rAd-hTRP-2 prevented the development of lung metastases after challenge with B16 melanoma in all vaccinated mice (Fig. 1,A). This effect was entirely mediated by CD8⁺ T lymphocytes because depletion of CD8⁺ T cells, but not CD4⁺ T cells, abrogated the preventive activity (Fig. 1,A). The vaccination protocol generates an antigen-specific CTL response against the immunodominant K^b-restricted TRP-2_180–188 peptide. In fact, splenocytes from mice immunized with rAd-hTRP-2 and stimulated in vitro with syngeneic cells infected with the same virus showed a strong cytotoxic activity against B16 cells and 293K^b pulsed with the TRP-2_180–188 peptide, but not toward control cells pulsed with the irrelevant β-gal_96–103 peptide (Fig. 1,B). As expected, cytotoxicity against B16 cells was abrogated in a dose-dependent fashion by the addition of an excess of cold B16. The same competition curve was obtained with TRP-2–pulsed 293K^b cells, but not with either 293K^b cells loaded with the irrelevant β-gal peptide or the natural killer target YAC-1 cells (Fig. 1 B), suggesting that the TRP-2–specific CTLs were responsible for all or most of the lytic activity of the cultures stimulated with the whole rAd-hTRP-2.

Even after tumor rejection, the immune response in mice vaccinated with rAd-hTRP-2 did not spread significantly toward other antigens expressed by the melanoma cell line B16. Splenocytes of mice that had rejected B16 pulmonary metastases after preventive vaccination with rAd-hTRP-2 were stimulated in MLPCs with the immunodominant peptides of the mouse melanoma antigens TRP-2, gp100, or p15E (13, 14). These peptide-stimulated splenocytes were tested for the ability to recognize B16 cells by releasing IFN-γ, an assay useful to detect T lymphocytes with high avidity toward the tumor antigen (see below). Only the MLPCs stimulated with TRP-2 peptide contained effectors that recognized B16 cells, provided that the mice had been immunized with rAd-hTRP-2 but not with the control rAd-EGFP. This suggests that tumor growth in the host did not generate per se an immune reactivity to different melanoma antigens including TRP-2 (Fig. 1 C). No reactivity against either gp100 or p15E was detected in mice that rejected the melanoma after immunization with rAd-hTRP-2, whereas antimelanoma reactivity directed to gp100 and p15E antigens could be easily induced by immunizing mice with either a plasmid DNA encoding p15E or a rVV encoding the altered peptide of gp100 (13, 14). These results are in agreement with previously published data. In fact, immune response elicited by irradiated B16 melanoma cells engineered to produce granulocyte macrophage colony-stimulating factor and given in conjunction with anti–CTLA-4 was directed exclusively to TRP-2 antigen, with no evidence of reactivity to melanocyte antigens gp100, tyrosinase, Melan-A/MART-1, or TRP-1 (2).

To evaluate the efficacy of different immunogens, the rAds encoding either green fluorescent protein fused to mouse TRP-2 [mTRP (rAd-EGFP-mTRP-2)] or murine TRP-2 (rAd-mTRP-2) were compared with rAd-hTRP-2. Mice underwent splenectomy 14 days after vaccination with different rAds. Splenocytes were used for staining with H-2 K^b tetramers assembled with the TRP-2_180–188 peptide (TRP-2-TET) as effectors in the ELISPOT assay and for MLPC. After 5 days, we performed tetramer staining and IFN-γ release from the MLPC. On day 21, these mice were challenged with an intravenous lethal inoculum of B16 melanoma cells. Fourteen days later, lungs were removed, and pulmonary metastases were counted in a blind fashion (Supplemental Fig. S1A).

As expected, immunization with rAd-mTRP-2 did not prevent tumor take (Fig. 2,A). After immunization with rAd-hTRP-2 but not with rAd-mTRP-2, all mice showed significant changes in the immunologic variables examined (ELISPOT assay, TRP-2-TET⁺ cells ex vivo and in MLPC, and IFN-γ released by MLPC-derived lymphocytes, Fig. 2,B–D). The fusion product, EGFP-mTRP-2, delivered by rAds also conferred protection from challenge. However, whereas MLPC-derived T lymphocytes released IFN-γ when cocultured with peptide-pulsed K^b target cells, there was no significant recognition of the B16 melanoma cells, suggesting the generation of low avidity effectors (Fig. 2,C). To further examine their relationship, we compared the number of pulmonary metastases and the percentage of TET⁺ lymphocytes in MLPC. The experimental data fit a two-parameter exponential decay curve (metastasis fraction = 0.48e^{−2.53(%Tet+cells}); Rsqr = 0.996) strongly suggesting that every increment in the number of TRP-2-TET⁺ cells affects the tumor progression at a fixed rate: Small increments are highly effective until a plateau is reached in the antitumor effect (Fig. 2 E). These results indicate that predictive markers of the efficacy of genetic vaccination can be found in a prevention model. In particular, the number of TRP-2-TET⁺ lymphocytes expanded in vitro after a single cycle of stimulation with the peptide had the highest predictive value.

It was shown previously that prophylaxis and therapy of cancer after vaccination with cell-based vaccines are distinct experimental situations because they can generate different cellular mechanisms leading to tumor rejection (2). We thus investigated whether predictive markers for recombinant vaccine efficacy could be unveiled in a therapeutic model. Induction of therapeutic tumor immunity was assessed by intravenous challenge with B16 melanoma cells, followed 3 days later by a single inoculation of recombinant rAd-EGFP, rAd-mTRP-2, rAd-EGFP-mTRP-2, or rAd-hTRP-2. Whereas immunization with rAd-mTRP-2 and rAd-EGFP-mTRP-2 (data not shown) was not sufficient to affect the overall survival, melanoma-bearing mice receiving the rAd-hTRP-2 showed a significant prolongation of survival, and about 50% of the mice were completely cured and remained disease-free during the observation period (Fig. 3,A). These mice will be referred to as regressors. Depletion of CD8⁺ T cells before vaccination with rAd-hTRP-2 adversely affected the therapeutic activity of the vaccine because tumor developed at the same speed as the negative control, confirming a prevalent role of CD8⁺ T lymphocytes in tumor eradication (Fig. 3,B). Interestingly, CD4⁺ T cells appear to be required in the later phase of tumor rejection because mice depleted of this lymphocyte population developed tumor, but with slower kinetics (Fig. 3 B).

Analysis of the immune response in single mice (Supplemental Fig. S1B) did not reveal any characteristic that predicted this long-term therapeutic effect. Complete regressors did not have the highest TRP-2-TET⁺ cell expansion or the more functionally active effectors, ex vivo or after in vitro stimulation with the TRP-2_180–188 peptide (Table 1), a situation similar to that observed in melanoma patients. The overall analysis indicated that the mice rejecting the tumor in the therapy model presented intermediate levels of TRP-2-TET⁺ cell expansion, as compared with the group that did control tumor growth for some time but did not reject the tumor completely (Table 1).

To further characterize the immune response in this therapeutic model, we analyzed the TCR repertoire usage of the mice combining tetramer sorting with T-cell repertoire examination by means of CDR3 length analysis, a technique known as spectratyping or immunoscope. The nucleotidic sequence of each CDR3 rearrangement defines a T-cell clonotype. Splenocytes from mice treated with rAd-hTRP-2 were cultured with TRP-2_180–188 peptide and the triple-positive lymphocytes (TRP-2-TET⁺/CD3⁺/CD8⁺) separated by high-speed cell sorting. Tetramer-guided analysis of TCR α- and β-chains revealed a severe restriction in TCR BV and TCR AV gene segment usage and a difference in repertoire between regressor and progressor mice. In Fig. 4,A, we show a synopsis of the complete TCR BV and TCR AV repertoire in two animals, one regressor and one progressor, after therapy with rAd-hTRP-2 (for actual spectratypes, see Supplemental Fig. S2). The mouse that was cured by therapeutic immunization showed a severely skewed repertoire; in particular, the TCR AV compartment presented a more marked skewing in comparison with TCR BV usage. Similar results were found in all animals studied (data not shown). Even more strikingly, we observed the expression of high message levels for a given α family (TCR AV5) in all of the survivors. By spectratype analysis using a combination of AV and AJ primers, this family was resolved as a single peak (Fig. 4,B). These peaks were directly sequenced using an internal primer (data not shown) and found to contain a major CDR3 sequence originally described in CTL clone 24 derived from a TRP-2 immune mouse (see below). To estimate the frequency of this clonotype in the two animal groups, we performed clonotype analysis as detailed in Materials and Methods. The analysis showed that this particular rearrangement of TCR AV5 was not detectable or was detectable at very low levels in progressor mice; in particular, the average of that particular TCR AV5 sequence in nonsurvivor and survivor animals was 3.5 and 27.2 of the total TCR AV repertoire, respectively (P = 0.037; Fig. 4 C). In some of the survivors, this clonotype represented about 50% of the total TCR-α present in the tetramer-sorted population. The same analysis did not show a significant association between TCR BV usage and either survival or death of the animals (data not shown). Thus, these two groups could not be distinguished by quantitative assessment of the TRP-2–specific immune response, but rather by analysis of the TCR composition of the expanded TRP-2–specific lymphocyte population.

Mice that rejected the tumor after challenge were monitored for more than 1 year. None developed vitiligo during this time (data not shown), a finding corroborated in other experimental melanoma models (8, 9).

As mentioned above, the CDR3 sequence that revealed the T-cell clonotype discriminating tumor regressors from progressors was originally found in CTL clone 24. Interestingly, the CTL clones 8 and 24, obtained by the same cloning procedure from the splenocytes of a TRP-2–vaccinated C57BL/6 mouse, shared a common AV gene (GenBank accession numbers AY089787 and AY089788, respectively) but had different BV rearrangements (GenBank accession numbers AY089784 and AY089785, respectively). This allowed us to investigate the properties of TRP-2–specific clones that differed only in the TCR β-chain. Both clones 8 and 24 were able to lyse a syngeneic cell line pulsed with TRP-2_180–188 (Fig. 5,A). However, a titration assay performed to evaluate the functional TCR avidity for the TRP-2_180–188-K^b antigenic complexes clearly indicated that CTL clone 24 possessed a somewhat higher avidity (Fig. 5,A). In fact, clone 24 exhibited an EC₅₀ (2 × 10⁻¹¹ mol/L) that was about 10 times lower than the EC₅₀ of clone 8 (2.5 × 10⁻¹⁰ mol/L). In agreement with these results, both clones released similar levels of INF-γ when stimulated with peptide-pulsed syngeneic cells, but only clone 24 was able to secrete INF-γ against B16 parental melanoma and the variant B16LU8 (Fig. 5 B). Taken together, these data suggest that TCR-α usage is predominantly shaped by the recognition of TRP-2. However, the β-chain contributes considerably in determining the function and avidity of specific T cells.

We speculated that the higher avidity clone 24 might better recognize the few antigen complexes exposed on B16 tumor cells also in vivo. C57BL/6 mice inoculated intravenously with B16LU8 cells received an inoculum of 5 × 10⁶ cells of the different CTL clones 3 days later. IL-2 was given intraperitoneally for 3 days after adoptive transfer. A group of mice treated only with IL-2 was used as a control. The IL-2 treatment alone reduced the number of metastases in comparison with untreated C57BL/6 mice (P = 0.007). The adoptive transfer of clone 8 marginally increased the antitumor effects exerted by IL-2 treatment alone (P = 0.11). On the other hand, the adoptive transfer of clone 24 significantly reduced the total number of pulmonary metastases (clone 24 versus IL-2, P = 0.006). These results clearly reveal a strict association of the ability of CTLs to recognize melanoma in vitro with the TCR avidity for the peptide-MHC complexes and with the antitumor effects in adoptive transfer experiments.

In the present study, a multivariable analysis of the CD8⁺ T-cell response in mice vaccinated with different recombinant vaccines encoding the melanoma antigen TRP-2 delineated three main situations. The simplest finding involved the activity of the powerful rAd-hTRP-2 vaccine in prophylaxis of tumor challenge. All of the mice mounted a vigorous immune response against the mouse K^b-restricted epitope, as indicated by the significant changes in the number and functions of TRP-2–specific T lymphocytes. The immune response was so strong that correlation with the number of metastases was not feasible because the majority of the mice did not present countable metastases after challenge with B16 (7 of 10 mice had no metastases). T lymphocytes stimulated with the TRP-2_180–188 peptide in vitro also recognized the wild-type melanoma. Nevertheless, it must be noted that prophylactic vaccination with the xenogeneic form of the antigen, although valuable for several experimental mouse tumor antigens, would be difficult to implement and interpret in a clinical setting.

The second finding involved formulation of a clinically applicable immunogen (chimeric protein between an immunogenic protein and mouse TRP-2 expressed in rAd) that was shown to have some protective effect on tumor challenge, but with a thwarted immune response. TRP-2–specific T lymphocytes, which proliferated after in vitro stimulation, efficiently recognized the immunodominant peptide, but not the B16 melanoma, which displays a low number of K^b-peptide complexes on the surface due to down-regulation of class I H-2 molecules (21). This finding suggests that T lymphocytes with low avidity TCR were the main effectors elicited by these vaccines. The percentage [and the absolute number (data not shown)] of TRP-2-TET⁺ lymphocytes in peptide-stimulated cultures correlated directly with the antimetastatic activity of the vaccine. Moreover, rAd-EGFP-mTRP-2 vaccination was successful in eliciting effector lymphocytes detected by an ELISPOT assay for IFN-γ, and the number of effector lymphocytes also correlated with the prophylactic efficacy (data not shown).

The third finding involved the model most similar to the clinical setting. Therapy of established tumors required a strong vaccine formulation, and yet only about half of the treated mice completely rejected the tumor. Even with this small tumor burden, therapy appeared to be a stochastic event, as in human melanoma patients. All mice, in fact, developed an easily detectable immune response against TRP-2, but the breadth of the response did not allow us to distinguish tumor progressors from regressors. Qualitative rather than quantitative differences were thus suspected. Individual differences in the orientation of the CTL response to a tumor antigen were previously explained by the stochastic timing of recruitment of different epitope-specific T cells, a sort of “first come, first served” hypothesis (22). Indeed, it has been elegantly proven that T cells that encounter the antigen at early time points can account for a significant part of the specific response, even though they are not the most frequent in the preimmune repertoire (23). Our results might reflect a scenario in which only those mice that present an in vivo expansion of a selected population of T cells bearing particular TCR AV chains are able to reject the preexisting tumor. The properties and dynamics of these “fittest” T cells are currently not known, although we can speculate that lymphocytes possessing the AV5 chain (together with a few other AV clonotypes) might include high avidity TRP-2–specific T cells, whose prototype is represented by clone 24.

Combining the sorting of TRP-2–specific CD8⁺ T cells and quantitative PCR-based T-cell repertoire analysis greatly improves the accuracy in detecting T-cell clones that could not otherwise be detected by using functional assays, semiquantitative TCR analysis, or tetramer sorting alone (19). Functional assays based on the estimation of the overall immune reactivity against the melanoma antigen could not be sufficiently predictive in therapeutic vaccination because they reflect an oligoclonal expansion that might or might not include the therapeutic CTL clones. In a prophylactic setting, on the other hand, this oligoclonal response might be sufficient to control the growth of the limited number of B16 melanoma cells inoculated in mice regardless of the clonal composition of the TRP-2–responsive population.

Preferential AV usage in tumor regressors was not entirely surprising because antigen recognition by CTLs was shown to require a specific α-chain pairing with a variety of TCR β-chains (24), suggesting a biased TCR AV usage in peptide recognition that has been confirmed by analysis of TCR-MHC-peptide complex crystals (25, 26). Moreover, a restricted TCR-α repertoire has been described in CTLs recognizing the Melan-A/MART-1 melanoma antigen in HLA A2 context (27, 28, 29), although in these studies, different CDR3 rearrangements were found in the presence of the same AV gene usage. In one report, Melan-A–specific T cells isolated from melanoma patients were found to have a frequent usage of the AV 2.1 chain but a large BV chain repertoire (29). This preferential usage is not related to an antigen-driven narrowing of the TCR affinities or peripheral homeostatic expansion of selected clones in tumor-bearing hosts but rather reflects a constraint already present in the preimmune repertoire. Our data are apparently discrepant; however, some important differences need to be highlighted. We analyzed the TCR repertoire on a T-cell population able to bind TRP-2 tetramers. AV preferential usage was found in animals that regressed the tumor, thus it is possible that we preferentially selected the lymphocytes that recognize TRP-2 with higher efficiency. Moreover, the Melan-A/MART-1 antigenic system is unique in that a sizeable pool of naïve Melan-A/MART-1–specific CD8 T cells is generated during thymic selection (30, 31).

Although studies with melanoma patients failed to reveal a correlation between AV 2.1 usage and avidity of antigen recognition, some CDR3 public or homologous sequences within the AV 2.1-AJ 35 rearrangements were more frequently found in CTLs derived from different donors that recognized the tumor with high avidity (29). Recurrent or homologous AV sequences appeared also to pair preferentially with BV 14, suggesting that additional factors such as the specific CDR3 loop or the pairing with some BV chains could influence the overall avidity. In this regard, the mouse AV5 chain with a conserved CDR3 region described in this article was found in at least four clones isolated from the same bulk culture of mice immunized with pcDNA3-TRP-2: paired with BV7 in clones CTL24 and CTL20 and paired with BV8.2 in clones CTL7 and CTL8. As shown in Fig. 5, these CTL clones possessed different avidities toward TRP-2-K^b complexes, thus confirming the relevance of the α-chain to guide antigen recognition and the requirement for α- and β-chain pairs to shape the strength of interaction with the antigen-MHC complex.

Our data strongly support the concept that the presence of specific T-cell clonotypes is a requirement in breaking peripheral tolerance and mounting a therapeutic immune response in tumor-bearing hosts toward a tumor-associated antigen such as TRP-2. Additional studies are needed to investigate whether an efficient expansion of the protective T-cell clonotype requires a preimmune bias in the T-cell repertoire of single animals, as proposed in human studies (28, 29, 31), or is due to a stochastic usage of unselected repertoire. However, the possibility of identifying a close correlation between a particular TCR usage or a given T-cell clonotype and the efficacy of the immune response against a tumor antigen may open new scenarios in the immunotherapy of tumors.

The authors would like to thank Susanna Mandruzzato and Jack Gorski for helpful discussions and critical reading of the manuscript, Pierantonio Gallo for assistance with graphics, Vito Barbieri for technical assistance in mouse studies, and Lisa Smith for the editing.