Adoptive immunotherapy with lymphokine-activated killer cells and recombinant interleukin 2 (IL 2) can produce significant reduction of visceral metastases in tumor-bearing mice and, as shown recently, in humans with disseminated cancer. Because further dose escalations of IL 2 have been prevented by the development of a vascular leak syndrome (VLS) in both mice and humans, we investigated this VLS in mice undergoing the systemic administration of high-dose IL 2. A model for quantitating capillary permeability was used in which 125I-bovine serum albumin was injected i.v., and 2 hr later, tissues were counted in a gamma analyzer. A permeability index (PI) was calculated by dividing the mean counts per minute (cpm) of tissues from IL 2-treated mice by those from control animals. The injection of IL 2 produced increases in vascular permeability that were most pronounced in the thymus, spleen, lungs, liver, and kidneys (PI = 18.0, 10.0, 9.7, 6.7, and 6.3, respectively, on day 6). The development of the VLS was highly dependent on the number of days of IL 2 treatment (for example, the lungs contained 638, 1382, 3350, and 6187 cpm after 0, 1, 3, and 6 days of IL 2, respectively). Moreover, the degree of the VLS was directly related to the dose of IL 2 administered. Measurement of the wet and dry weights of lungs from IL 2-treated mice demonstrated that IL 2 produced a dramatic increase in their water weight (from 0.10 g at base line to 0.22 g after 200,000 U of IL 2 for 6 days). The injection of the IL 2 excipient failed to induce capillary leakage in tissues. Immunosuppression of mice by pretreatment irradiation (500 rad) or by injection of cyclophosphamide or by concurrent use of cortisone acetate markedly reduced or eliminated the development of the VLS. Similarly, the VLS was not observed in nude mice receiving IL 2. Thus, the administration of IL 2 produces a dose-limiting VLS that may be mediated, directly or indirectly, by host lymphoid elements.