Melanoma-Reactive T Cells in the Bone Marrow of Melanoma Patients: Association with Disease Stage and Disease Duration

Bone marrow has long been established as the major primary lymphoid organ but is emerging now as an important secondary lymphoid organ as well (1). Not only can bone marrow induce primary T-cell-mediated immune responses in situ against local and blood-borne tumor antigens (1, 2) but it can also recruit memory T cells generated at other sites (3) and function as a nest for migrating memory T cells (4). Therefore, bone marrow of tumor patients may represent an interesting organ for analysis and therapeutic exploitation of the patients' repertoire of tumor-reactive T cells. Tumor-reactive memory T cells from the bone marrow of breast cancer patients induced regression of autologous tumors transplanted into nonobese diabetic/severe combined immunodeficient mice (5).

The immunogenicity of malignant melanoma is well established, and tumor-reactive T cells were found in the peripheral blood of tumor patients (6). Clinically, overt metastasis of melanoma to the bone is an infrequent event occurring in advanced stages of the disease (7). Recent studies described melanoma micrometastasis to the bone marrow in stage III to IV patients (8, 9). Although there is some indication that detection of tyrosinase mRNA (9) or immunomagnetic detection of melanoma cells in bone marrow (10) may be associated with a worse prognosis, current data do not support the routine use of bone marrow sampling for staging or stratification purposes. Recently, melanoma-reactive CD8⁺ T cells have been described in the bone marrow of five tumor-free melanoma patients after experimental immunotherapy (11). Using peptide-MHC tetramers and flow cytometry, these cells were at either the same or higher frequency in the bone marrow as in peripheral blood of the same patients. These results are in line with the detection of memory T cells in early-stage breast cancer patients (5). Because bone marrow has been proposed as an alternative source for tumor-specific T cells for the use in adoptive transfer strategies (5), we investigated the occurrence of melanoma-reactive T cells in melanoma patients of different disease stages.

Patients. The study was initiated after approval of the local ethics committee. After obtaining informed consent, peripheral blood was drawn by venipuncture and bone marrow was aspirated from the iliac crest of 40 participating patients. Stage was assigned according to the new American Joint Committee on Cancer (AJCC) criteria (12).

Generation of dendritic cells and T lymphocytes. Dendritic cells were generated as described previously (5). In short, peripheral blood mononuclear cells were obtained by density gradient centrifugation and were subsequently allowed to adhere to plastic. Adherent cells were cultured in X-VIVO 20 medium (Cambrex BioScience, East Rutherford, NJ), recombinant human granulocyte macrophage colony-stimulating factor (50 ng/mL; Essex Pharma, Munich, Germany), and recombinant human interleukin (IL)-4 (1,000 IU/mL; PromoCell, Heidelberg, Germany) for the generation of dendritic cell. After 1 week of culture, dendritic cell was harvested, counted, depleted of contaminating T and B lymphocytes, and used for enzyme-linked immunospot (ELISpot) analysis. Nonadherent cells were maintained in RPMI 1640 (Invitrogen, Karlsruhe, Germany) supplemented with 8% human AB serum (Sigma, Deisenhofen, Germany), recombinant human IL-2 (100 units/mL; PromoCell), and recombinant human IL-4 (60 units/mL). After 1 week, T cells were depleted of contaminating CD15⁺, CD19⁺, and CD56⁺ cells and used for ELISpot analysis.

Cell lines and lysates. U-937 is a histiocytic lymphoma cell line (13) and was obtained from the American Type Culture Collection (Manassas, VA). SK-MEL-23, a kind gift of Dr. L.J. Old (Memorial Sloan-Kettering Cancer Center, New York, NY), is a highly pigmented melanoma cell line. Cells were lysed by five cycles of freezing and thawing. Dendritic cells were pulsed with lysates for 20 hours at a concentration of 200 μg/1 × 10⁶ cells/mL.

IFN-γ ELISpot. ELISpot analysis was done as described previously (5). In short, 96-well nitrocellulose filtration plates (Millipore, Schwalbach, Germany) were coated with the monoclonal antibody 1-D1K (Mabtech, Nacka Strand, Sweden). Dendritic cell pulsed with different lysates was coincubated with autologous T cells (1:5, dendritic cell to T cell ratio) for 40 hours. All samples were analyzed at least in triplicates. Plates were developed using a biotinylated secondary antibody (clone 7-B6-1, Mabtech), streptavidin-alkaline phosphatase conjugate, and the chromogenic substrate nitroblue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate. Spots were automatically counted with the use of a computer-assisted video image analyzer (KS Elispot, Zeiss, Goettingen, Germany). Individuals were considered as responders if the number of spots formed in the presence of dendritic cell loaded with tumor lysate was significantly higher than in the presence of control lysate-pulsed dendritic cell.

Statistical analysis. Numbers of IFN-γ-secreting cells in ELISpot in different experimental conditions were compared using two-sided t test. Contingency tables were analyzed using Fisher's exact test. Continuous variables from two groups were compared using Mann-Whitney test. Stage dependence of tumor-reactive bone marrow T cells was analyzed using exact Cochran-Armitage test. Differences between melanoma reactivity in bone marrow and peripheral blood were analyzed using McNemar test. A significance level of ≤0.05 was considered significant.

Forty patients participated in the study between August 2002 and August 2005. Clinical information for each patient is shown in Table 1.

Melanoma-specific reactivity was evaluated as the number of T cells producing IFN-γ after stimulation by autologous dendritic cell pulsed with lysate of the well-characterized melanoma cell line SK-MEL-23 as a source of tumor antigens compared with dendritic cell loaded with U-937 lysate as a control. One representative ELISpot result is shown in Fig. 1. We decided to use lysate of a melanoma cell line as test antigen because autologous tumor cells were not available from most patients especially in early stages, although the use of HLA-restricted peptides from defined tumor antigens has the disadvantage of a restriction to only a few CD8 T-cell epitopes in HLA-selected patients. Our methodology with the use of lysate-pulsed dendritic cell as antigen-presenting cell may underestimate the frequency of tumor-reactive CD8⁺ cells because it is based on the cross-presentation pathway of antigen presentation, which is less efficient (14). Nevertheless, our method has the important advantage of additionally assaying for a CD4⁺ response, which is an integral component of an efficient and lasting Th1-biased cellular immune response (15).

Significantly increased numbers of melanoma-reactive T cells were found in the bone marrow of seven patients (“bone marrow responders”) and in peripheral blood of four patients (“peripheral blood responders”; Table 1). Background spot numbers were similar in bone marrow and peripheral blood samples from responders and nonresponders (Fig. 2). In three of five HLA-A2⁺ patients (16, 23, and 27), who were treated with peptide vaccination, sufficient T cells were obtained to do additional IFN-γ ELISpot assays using dendritic cells pulsed with the respective peptides from the tumor antigens gp100, Melan A, MAGE3, and/or tyrosinase that were used for vaccination. Interestingly, none of the patients showed a peptide-specific immune response (data not shown) or a response against melanoma cell lysate, suggesting that the vaccination did not induce long-term memory responses.

The frequencies of melanoma-reactive T cells in responder patients varied between 4 and 251 among 10⁵ total T cells and were comparable in bone marrow and peripheral blood (Table 1). Although melanoma-reactive T cells were found in peripheral blood of patients with all tumor stages, the presence of tumor-reactive T cells in bone marrow was significantly associated with advanced tumor stages. Table 2 summarizes the clinical characteristics of bone marrow responders and nonresponders. Six of seven positive bone marrow samples were derived from stage IV melanoma patients but none from patients with stage I to II melanoma (Table 3). In stage IV patients, there was a trend that tumor-reactive T cells were more often in bone marrow than in peripheral blood (P = 0.06; Table 3). The simultaneous presence of melanoma-reactive T cells in peripheral blood and bone marrow was detected in only 1 of 10 responder patients. This observation points to the possibility that priming of melanoma-reactive T cells can occur in different lymphoid organs at different stages of the disease (e.g., in tumor-draining lymph nodes at earlier stages of the disease and in bone marrow at later stages). We did not find a significant correlation between preceding immunotherapy and tumor immune reactivity. Nevertheless, we also cannot exclude such an association because five of seven patients with tumor-reactive T cells in their bone marrow had received IFN-α pretreatment (Table 1).

It seems that, in melanoma, the fraction of tumor patients with tumor-reactive bone marrow T cells (7 of 40, 17.5%) is lower than in primary breast cancer patients (19 of 29, 66%; ref. 5) and in patients with pancreatic cancer (41 of 41, 100%; ref. 16). Whereas memory T cells were found predominantly in early stages of primary breast cancer, melanoma-reactive bone marrow T cells were associated with advanced disease. It was shown before in other tumor entities that frequencies and functional capacities of tumor antigen-specific T cells decreased with disease progression most likely due to tumor-induced immune suppression (4). It is remarkable that the opposite trend seems to hold true for bone marrow T-cell responses of melanoma patients.

Our findings lead to a modification of the interpretation of the findings reported by Letsch et al. (11) who described melanoma-reactive CD8⁺ T cells in bone marrow of all five tested patients.

Four of the five patients had stage IV disease (International Union Against Cancer), and four of five of the patients had received tumor peptide vaccination before the analysis. Given the higher number of patients investigated in our study and the inclusion of patients with early-stage disease, we believe that the study of Letsch et al. (11) had a bias toward advanced disease. In agreement with our interpretation of an association of bone marrow immune responsiveness with late-stage disease, four of the seven bone marrow responders in our study had measurable tumor load at the time of the bone marrow sampling. Furthermore, variables clinically associated with stage [longer disease duration and lactate dehydrogenase (LDH) activity] were also associated with the responder status in bone marrow of patients (Table 2), although such correlation was not found for peripheral blood T cells (Table 1). Interestingly, median duration of disease, as determined by the interval between the initial diagnosis and the bone marrow sampling, was longer in stage IV patients with a tumor-reactive T-cell memory in bone marrow than in those without (median, 12 months versus 43 months). Nevertheless, we did not find a significant survival advantage in patients with positive bone marrow response as analyzed by Kaplan-Meier analysis (data not shown). This discrepancy may be related to the observation that priming in the bone marrow occurred preferentially in late phases of advancing disease.

The immunobiology of melanoma may be different from breast cancer. Although for breast cancer micrometastatic disease is a negative predictor of survival (17, 18), the situation in melanoma is being debated (9, 10).

At present, the mechanism behind the occurrence of tumor-reactive effector and memory T cells in the bone marrow is not entirely clear. Although it has been shown in the mouse that the bone marrow can be a priming site for T-cell responses (1, 2), there is also evidence that T cells primed in lymph nodes or spleen can migrate to the bone marrow (3, 19). It seems plausible that tumor-reactive T cells found in the bone marrow were primed there. This would only occur in situations in which tumor antigens reach the bone marrow environment in sufficient concentration. Although formal proof is lacking, we propose from the presently available data that this happens only in advanced stages of melanoma. In cancers, which gain access to the bone marrow early, such as breast cancer, tumor-reactive T cells may be found in bone marrow also early.

In conclusion, we found a stage dependency of the occurrence of melanoma-reactive bone marrow T cells. Only a subset of advanced stage melanoma patients showed tumor-reactive T cells in their bone marrow. The frequency of these cells in the bone marrow is comparable with the situation described for breast cancer patients (5) and was not significantly related to any therapeutic intervention or treatment outcome, contrasting earlier results (11). Although not directly addressed in our study, these cells are most likely memory T cells because we detected antigen-dependent IFN-γ secretion during the 40-hour stimulation period only in the CD45RA T-cell fraction (data not shown). They may be an attractive source of effector T cells from advanced stage patients for future immunotherapies because they are less exposed to the negative effect of the tumor and its microenvironment and, if primed in the bone marrow, may preferentially recognize metastatic tumor cells.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.