In the current study we used the therapy of established murine leukemia to identify the lymphocyte subsets responsible for toxicity and for therapeutic efficacy of high-dose IL-2. Initial results confirmed that high-dose IL-2 induces marked proliferation of a variety of host cells, including NK cells, Lyt-2+ T cells, L3T4+ T cells, and B cells. Infusion of antibody to NK-1.1 depleted NK-1.1+ cells in vivo and greatly reduced the toxicity of IL-2, but did not decrease therapeutic efficacy. By marked contrast, depletion of host T cells, either Lyt-2+ or L3T4+, had no effect on toxicity but greatly reduced therapeutic efficacy. The requirement for host T cells for the curative effect of IL-2 gives credence to the possibility that substantial efficacy of high-dose IL-2 against established malignancy may require existent host antitumor immunity. Since the human tumors that have been shown to have the most substantial responses to IL-2 (i.e., malignant melanoma and renal cell carcinoma) are those long considered to be immunogenic in the autochthonous host, the current study predicts that for these, as well as other immunogenic human tumors, it should be possible to decrease the toxicity and thus increase the therapeutic index of IL-2 by selectively depleting NK cells in vivo.