The rationale of treating melanoma patients by infusion with tumor-infiltrating leukocytes (TIL) is to perform an adoptive therapy through injection of tumor-specific T cells. Nonetheless, methods currently used for ex vivo TIL expansion have not been evaluated for their efficacy to expand TAA-specific T cells. We have addressed this question here, using a culture method in which high TIL growth was induced by a polyclonal T cell stimulus. Intracellular cytokine assays were performed to measure the proportion of T cells responding to autologous tumor cells among the lymphocytes from lymph node biopsies (TIL) of 26 patients with stage III melanoma. The data show that TIL from 18 of these patients contained detectable amounts of tumor-specific T cells before expansion. Although they decreased somewhat in percent abundance during expansion, they were still present afterwards, ranging from 0.3 to 13.8%. Since a median number of 1.7 x 10(10) TIL was obtained from these patients (starting from 3.6 x 10(6) TIL), a total amount of tumor-reactive cytokine-secreting TIL of between 2.8 x 10(6) and 1.12 x 10(9) was obtained in each case from 18 patients. The TIL populations from 8 patients did not contain tumor-reactive T cells: neither before expansion, nor after expansion. Lack of tumor-reactive TIL only occurs for patients bearing several tumor-invaded lymph nodes (40%), but not for those having a single invaded lymph node. Therefore, high numbers of tumor-reactive T cells can be produced, through a polyclonal TIL stimulation, from most early stage III melanoma patients but from only about half of the patients with a more disseminated disease. For this last group, the possibility of getting tumor-reactive TIL can be predicted by checking the presence of these cells before expansion.