Thirty-nine melanoma patients were treated with cyclophosphamide (350 mg/m2) followed 3 days later by 5 daily doses of interleukin 2 (3.6 million units/m2 i.v.) weekly for 2 weeks. This cycle was repeated at least twice with a 1-week interval between cycles. Natural killer and lymphokine-activated killer (LAK) cell activity in peripheral blood mononuclear cells were measured before treatment and on the last day of each cycle by chromium release assays. Development of LAK activity of greater than 10 lytic units was correlated with a clinical response. There was no correlation between natural killer activity and clinical response. Antibody-dependent cell-mediated cytotoxicity of in vivo-induced LAK cells after the addition of mouse monoclonal antibodies (MAbs) in vitro was measured in 30 cases on the last day of each interleukin 2 cycle. Anti-GD3 MAbs MB3.6, 11C64, 6H4, and R24 increased LAK cell cytotoxicity against GD3-positive GD2 melanoma cells while anti-GD2 MAb 14.18 increased LAK cell cytotoxicity against GD3-negative GD2-positive melanoma cells. MAb 9.2.27 (IgG2a) directed against a chondroitin sulfate proteoglycan and its core protein with a molecular weight of 250,000 (p250) on human melanoma cells did not mediate antibody-dependent cell-mediated cytotoxicity. The effector cells in these antibody-dependent cell-mediated cytotoxicity. The effector cells in these antibody-dependent cell-mediated cytotoxicity reactions were Leu-19 positive. In preincubation experiments the MAbs showed superior binding to the melanoma target cells than to effector cells. Our results show that low dose interleukin 2 preceded by low dose cyclophosphamide effectively induces LAK cells in vivo. The cytotoxicity of these in vivo-activated LAK cells can be augmented in vitro by mouse MAbs against glycolipid antigens on the tumor.