There is increasing evidence that a T-cell response against a variety of defined tumor antigens exists in tumor patients even with advanced disease. In melanoma tumor-reactive T cells have been expanded from tumor tissue and peripheral blood recognizing a variety of antigens including melanosomal antigens, nonmelanoma-specific antigens like the MAGE-family, and tumor-specific mutated antigens (Robbins and Kawakami, 1996).
Little is known, however, about the in vivo representativity of such tumor-reactive T cells. Most investigators have expanded T cells by repeated in vitro stimulation with tumor cells or antigen in the presence of IL-2, which results in quantitative and qualitative changes. An analysis by Faure et al. (1998) using the immunoscope technique has shown that MART-1-specific T-cell clones expanded from tumor-infiltrating lymphocytes were not detectable in situ.
We have performed this study to evaluate whether circulating tumor-reactive T cells are detectable directly ex vivo in unstimulated peripheral blood lymphocytes from patients with metastatic melanoma. We have used the ELISPOT assay to determine T cells by their antigen-induced secretion of cytokines (Herr et al.,1996; Scheibenbogen et al.,1997). The important advantage of this assay is its high sensitivity facilitating the direct ex vivo quantification of antigen-reactive T cells. Furthermore, the direct secretion of IFNγ by T cells in response to antigen provides information about the functional state of the T cells. A close correlation between the number of T cells secreting IFNγ in response to an antigen and the level of specific cytotoxicity as determined with the chromium release assay has been shown by several groups (Di Fabio et al.,1994; Miyahira et al.,1995; Scheibenbogen et al.,1997).
To test the T-cell response against a broad panel of possible melanoma antigens, we used HLA-A2- or HLA-A1-matched melanoma cell lines as target cells in the ELISPOT assay. The feasibility of such an approach is suggested by previous studies showing that most HLA-A2+ melanoma-specific T-cell clones and T-cell lines recognize public antigens present on the majority of HLA-A2+ melanoma cell lines (Anichini et al.,1996; Hom et al.,1993). Using this ex vivo analysis, we were able to detect high frequencies of circulating melanoma-reactive T cells capable of rapid effector function in more than half of the patients with metastatic melanoma.
MATERIAL AND METHODS
Peripheral blood mononuclear cells (PBMC)
Blood samples were obtained from HLA-A1- and HLA-A2-positive healthy subjects and melanoma patients after informed consent. All melanoma patients had metastatic disease (AJCC stage IV, 1 patient stage IIIb). HLA-typing had been performed serologically by the standard NIH microlymphocytotoxicity test. PBMC were isolated from heparinized blood by density gradient centrifugation using Ficoll-Hypaque 1,077 (Biochrom, Berlin, Germany). Cells were washed twice with phosphate-buffered saline (PBS; Biochrom) and cryopreserved at −196° in FCS (Biochrom) containing 10% DMSO (Merck, Berlin, Germany).
For our studies, we used different HLA-A2- and HLA-A1-positive tumor cell lines including the melanoma cell lines SK-MEL-24, SK-MEL-30, and RVH, which were obtained from ATCC (American Type Culture Collection, Rockville, MD), as well as MKR, SB-Mel, HB-Mel (autologous tumor cell line of patient no. 16), Mel 14 (autologous tumor cell line of patient no. 15), Mel 18a (autologous tumor cell line of patient no. 17), Mel 38 (autologous tumor cell line of patient no. 18), and Mel 17 (autologous tumor cell line of patient no. 19), which were all established in our own laboratories. All tumor cell lines were cultured in tissue culture flasks in RPMI 1640 supplemented with 10% FCS, 1 mM L-Glutamine, and penicillin/streptomycin (Biochrom).
Expression of HLA-A1 and HLA-A2 on melanoma cell lines was determined by using mouse monoclonal IgG or IgM antibodies specific for HLA-A2 (BB7.2; ATCC) or HLA-A1 (One Lambda, Canoga Park, CA). Fluorescinated goat anti-mouse IgG or IgM were used as a second antibody. Separated T cells were phenotyped by staining with Simultest CD4-FITC/CD8-PE and Simultest control, as well as by CD45R0-FITC, CD45RA-FITC, and IgG-FITC (all from Immunotech, Hamburg, Germany). Data acquisition was performed on FACScan and FACSCalibur and analysed by using the CellQuest software (Becton Dickinson, Heidelberg, Germany).
Separation of PBMC was done by positive enrichment using CD8− and CD4− MACSMicroBeads and MiniMACS positive selection column type MS+ according to the producer's instructions (Miltenyi Biotec, Bergisch-Gladbach, Germany). Separation of CD8+/CD45RA+ and CD8+/CD45RO+ T-cell subsets was done by the MACS CD8+ Multi-Sort Kit (Miltenyi Biotec). After separation, cells were washed and resuspended in Iscoves modified DMEM (Biochrom) supplemented with 10% AB-serum (PAA, Linz, Austria), 1 mM L-Glutamin, and 1% penicillin/streptomycin (Biochrom) and kept overnight at 37°C and 5% CO2, before analysed in the ELISPOT assay. FACS analyses showed purity of the separated fractions of ≥ 90% in all experiments.
Ninety-six-well nitrocellulose plates (Millititer; Millipore, Bredford, MA) were coated with 50 μl/well of 8 μg/ml anti-human IFNγ MAb (code no. 1598-00; Genzyme, Rüsselsheim, Germany) overnight. Then wells were washed and blocked with Iscove's modified DMEM (Biochrom) + 10% AB-serum (PAA) for 2 hr at 37°C. PBMC or T cells were incubated in a concentration of 1.5 × 105/well, 1.5 × 104/well, and 1.5 × 103/well, admixed with tumor cells in a concentration of 1 × 104/well. PBMC or T cells alone or incubated with pokeweed mitogen served as negative or positive controls, respectively. After 24 hr of incubation in the antibody-coated plates at 37°C and 5% CO2, the plates were washed 6 times with PBS + 0.05% Tween 20. Wells were incubated for 24 hr at 4°C with 50 μl/well of biotinylated mouse anti-human IFNγ MAb (clone 4S.B3; Pharmingen, Hamburg, Germany) at a concentration of 2.5 μg/ml. After washing 4 times with PBS, 100 μl/well streptavidine-alkaline phosphatase (BioRad, Munich, Germany) diluted 1/1,000 was added for 2 hr at room temperature. After another washing step with PBS, 100 μl/well of BCIP/NBT substrate (BioRad) was added for 20–30 min. Color development was stopped by washing under running tap water. After drying at room temperature, spots were counted using an automized image analysis system (AID, Straßberg, Germany).
Antibodies used for blocking studies
The purified monoclonal mouse antibody BB7.2 directed against HLA-A2 (clone obtained from ATCC) and the monoclonal mouse antibody AIMHC II specific for HLA-DR (Immunotech) were used for blocking studies.
Wilcoxon signed rank test was calculated to test whether mean numbers of melanoma-reactive T cells between melanoma patients and healthy subjects showed significant differences.
Frequencies of T cells reactive with HLA-matched tumor cell lines in peripheral blood of healthy subjects
PBMC from healthy subjects served as negative control. We analysed the T-cell reactivity from 9 HLA-A2+ and 7 HLA-A1+ healthy subjects against 4 HLA-A2+ or 2 HLA-A1+ melanoma cell lines, respectively. No or few T cells secreted IFNγ in response to each melanoma cell line in the healthy individuals (0.006% ± 0.02% of PBMC, mean ± 3 SD, n = 9, in response to 4 HLA-A2+ melanoma cell lines; 0.009% ± 0.03% of PBMC, mean ± 3 SD, n = 7, in response to 2 HLA-A1+ melanoma cell lines; Tables I and II).
Table I. T-CELL RESPONSE AGAINST A PANEL OF HLA-A2+ MELANOMA CELL LINES IN HLA-A2+ HEALTHY SUBJECTS AND MELANOMA PATIENTS
Table II. T-CELL RESPONSE AGAINST 2 HLA-A1+ MELANOMA CELL LINES IN HLA-A1+ HEALTHY SUBJECTS AND MELANOMA PATIENTS
Number of IFNγ-secreting T cells/106 PBMC
Mean +3 SD
Frequencies of T cells reactive with HLA-matched tumor cell lines in peripheral blood of melanoma patients
We next tested the T-cell response of 11 HLA-A2+ and 3 HLA-A1+ patients with metastatic melanoma against the same HLA-A2− or HLA-A1-matched tumor cell lines. Patient characteristics are shown in Table III. Nine out of 14 patients had high numbers of tumor-reactive T cells with up to 0.81%, 0.78%, 0.53%, 0.10%, 0.09%, 0.07%, 0.06%, 0.04%, and 0.04% of PBMC secreting IFNγ against at least one of the allogeneic HLA-matched melanoma cell lines (patient nos. 1–7, 12, 13; Tables I and II). These frequencies exceeded the mean and 3-fold SD of the T-cell responses observed in healthy individuals (Tables I and II). The other 5 melanoma patients (patient nos. 8–11 and no. 14) had frequencies of tumor-reactive T cells in a similar range as found in several healthy individuals.
Table III. T-CELL RESPONSE AGAINST AUTOLOGOUS TUMOR CELL LINES AND AGAINST A PANEL OF ALLOGENEIC MELANOMA CELL LINES IN MELANOMA PATIENTS
To determine whether the release of IFNγ occurred already after a brief antigen contact in vivo, we tested the T-cell response after 6 hr of coincubation in the ELISPOT assay in 1 patient. A similar number of IFNγ-secreting T cells were observed in patient no. 1 after 6 hr and 24 hr (0.26% and 0.23% of PBMC secreting IFNγ in response to the melanoma cell line SBMel and 0.07% and 0.08% of PBMC secreting IFNγ in response to the melanoma cell line SK-MEL-30). The spots were, however, less intense and smaller after 6 hr as compared with 24-hr assay time. Therefore, we used 24 hr for all subsequent experiments.
T-cell subset analyses
In 2 HLA-A2+ patients (nos. 1 and 2) with high frequencies of tumor-reactive T cells, we isolated CD8+ and CD4+ T cells from PBMC. The T cells reactive with the HLA-A2-matched tumor cells were only found in the CD8+ compartment (Fig. 1). The reactivity of CD8+ T cells against the melanoma cell line RVH could be inhibited in these patients with the HLA-A2-specific BB.7.2 antibody more than 90% and 75%, respectively (Fig. 2), while the anti-HLA-DR antibody had no effect.
To further define the phenotypic characteristics of the CD8+ tumor-reactive T cells observed in our study, we separated CD8+/CD45RA+ and CD8+/CD45RO+ T cells in patient no. 1. A T-cell response was found in both subsets (Fig. 3). The phenotype CD45RO+/IFNγ+ is characteristic of memory T cells, whereas the CD45RA+/IFNγ+ T cells were shown to have the functional characteristics of effector-type CD8+ T cells (Hamann et al.,1997).
Frequencies of T cells reactive with autologous tumor cell lines in comparison with allogeneic tumor cell lines
To determine whether the T-cell responses detected against allogeneic melanoma cell lines are representative of the T-cell responses against autologous melanoma cells, we analysed 5 additional patients from whom autologous melanoma cell lines were available. Four out of 5 patients showed a T-cell response against their autologous tumor with 0.06% (no. 15), 0.03% (no. 19), 0.01% (no. 16), and 0.005% (no. 17) of PBMC secreting IFNγ in response to the tumor cells (Table III). In 2 of these 4 patients (nos. 15 and 16), similar frequencies of tumor-reactive T cells were observed against the autologous tumor and against the allogeneic HLA-A2-matched tumor cell lines SK-MEL-24, SK-MEL-30, and RVH in patient no. 15, and against the SK-MEL-24 in patient no. 16, respectively. Patient no. 19 had less than 0.01% of reactive T cells with the HLA-A1+ allogeneic cell lines. In contrast, patient no. 17 had 10–20-fold higher numbers of T cells reactive with the 4 allogeneic cell lines, suggesting that his autologous cell line may have lost the target antigen(s) or the ability to present them. In patient no. 18, no T-cell response against the autologous and less than 0.01% of T cells reactive with the allogeneic melanoma cell lines were observed.
Clinical significance of melanoma-reactive T melanoma-reactive cells
Three patients with high frequencies of tumor-reactive T cells (no. 4, 6, and 12) had achieved a complete or partial tumor regression following previous immunotherapy with high-dose IL-2 and/or IFNα containing regimens, and 1 patient (no. 17) had SD following vaccination with IL-12-transfected autologous melanoma cells (Table III). Another patient (no. 3) had presented with a lymph node metastasis of melanoma origin without a detectable primary melanoma, suggesting that the primary tumor lesion had been immunologically destroy. Six out of 11 patients, in whom tumor-reactive T cells could be demonstrated (Nos. 1, 2, 5, 7, 13, and 15) had, however, a growing metastatic tumor, and 3 out of these 6 patients had received previous immunotherapy without clinical response, suggesting the presence of escape mechanisms.
In 11 out of 19 melanoma patients analysed in this study, we could demonstrate increased numbers of circulating T cells directly secreting IFNγ in response to public target antigens present on HLA-A1- or HLA-A2-matched melanoma cell lines. As expected from our methodological approach, the T-cell response against the allogeneic melanoma cell lines in melanoma patients was mediated by CD8+ T cells and was HLA-A2-restricted as demonstrated in 2 HLA-A2-positive patients. Similar frequencies of T cells in response to HLA-A1- or HLA-A2-matched allogeneic and autologous tumor cells were observed in 3 out of 5 patients, in whom autologous cell lines were available. In 1 patient, only a T-cell response against the autologous tumor was detectable, possibly reflecting the recognition of a private antigen. In contrast, 1 patient had a strong T-cell response against all 4 allogeneic cell lines but not against his autologous cells, suggesting that his autologous cell line may have lost the target antigen(s) or the ability to present them.
One important and yet unresolved question is whether such tumor-reactive T cells can destroy the tumor in vivo. We have only indirect evidence from functional and phenotypical analyses that the T cells identified in our study may indeed be able to destroy melanoma cells in vivo. In our study, tumor-reactive T cells were characterized by the ability to directly secrete IFNγ in response to melanoma cells, which has been shown to be a property of memory and effector T cells in contrast to unprimed T cells (Hamann et al.,1997). Phenotypic analysis of melanoma-reactive T cells in 1 patient demonstrated IFNγ-secreting T cells in the CD8+/CD45RO+ as well as in the CD8+/CD45RA+ subpopulation, suggesting the presence of memory as well as effector T cells reactive with melanoma in this patient. CD45RA+/IFNγ-secreting T cells were shown to be able to directly release granzyme B and perforin and exert cytotoxic effector function after antigenic contact without further in vitro stimulation (Hamann et al.,1997). In a recent study, Lee et al. (1999) demonstrated circulating T cells specific for the melanoma-associated peptides MART-1 and tyrosinase in melanoma patients using the tetramer technology identifying T cells by their T-cell receptor-peptide binding affinity. Using this methodological approach, specific T cells were demonstrated in 1 patient, which were not able to secrete IFNγ in response to the specific peptide and were unable to lyse peptide-sensitized target cells (Lee et al.,1999). This study shows that functionally unresponsive T cells do exist in patients with metastastic melanoma as well. These T cells would, however, not have been identified in our analysis characterizing specific T cells by the ability to secrete IFNγ.
Indirect evidence for the clinical significance of tumor-reactive T cells present in our patients is provided by the clinical course. There were 4 out of 11 patients with tumor-reactive T cells with a spontaneous or immunotherapy-induced tumor regression, indicating a possible immunologically mediated control of tumor growth. Despite the demonstration of tumor-reactive T cells, 6 out of 11 patients had, however, a growing metastatic tumor, suggesting the presence of escape mechanisms. Several mechanisms of tumor-cell escape have been demonstrated, including the loss of tumor antigens, MHC antigens, or defects in the antigen processing pathway (Marincola, 1997).
One possibility is that the T cell responses we observed had been induced by previous systemic therapies, especially with T-cell-stimulating cytokines such as IL-2 and IFNα or with tumor vaccines. There were, however, several patients with tumor-reactive T cells who had not received any systemic treatment prior to T-cell analysis, suggesting a spontaneous development of tumor-directed T-cell responses in certain patients.
Although a variety of melanoma antigens recognized by cytotoxic T cells have been characterized so far, little is known about the antigen repertoire in human melanoma in vivo. Previous studies have shown that the majority of melanoma-reactive T-cell clones grown from peripheral blood do not recognize melanosomal antigens (Anichini et al.,1996; Mazzochi et al.,1996). Anichini et al. (1996) had analysed the reactivity of more than 100 T-cell clones, and most of the T-cell clones did recognize HLA-A2-positive allogeneic melanoma cell lines but not melanocytes. Similar results were obtained by Mazzochi et al. (1996) with HLA-A3-restricted T-cell clones recognizing HLA-A3 epitopes shared by melanomas but not by melanocytes. This is in accordance with our finding that the majority of patients had T cells reactive with the melanoma cell line SK-MEL-24, which has lost all melanosomal antigens (Slingluff et al.,1996).
In summary, our study shows the presence of an unexpected high number of CD8+ T cells reactive with melanoma cells in many patients with advanced disease. Our data also support the hypothesis that the antigen repertoire in melanoma is larger than is currently known. The identification of the target structures of this melanoma-directed T-cell response and the reasons for the “tumor escape” observed in several patients will provide important insights for the further development of antigen-specific treatment approaches.
Table IV. PATIENT CHARACTERISTICS: TUMOR STATUS AT TIME OF ANALYSIS AND MELANOMA-DIRECTED T-CELL RESPONSE
Status at time of analysis
Maximum percentage of PBMC reactive with allogeneic melanoma cell lines
Prior systemic therapy
Response to therapy
Location of metastases
Ln, lymph node.
This patient had recurrent skin metastases, which had all been resected.–n.a. = not applicable; NED = no evidence of disease.