The proliferation and/or survival of a variety of cells is dependent on cellular hydrogen peroxide (H(2)O(2)) production. We tested whether this was true of leukemic cells, using cell lines from leukemic patients (CEM, 697, Mn-60, and Tanoue). We found that addition of catalase inhibited proliferation of all cell lines and induced death in two. However, this turned out to be due to arginase contamination of the catalase. Pure arginase inhibited cell proliferation and survival, which was reversible by adding L-arginine, demonstrating the L-arginine dependency of these cells. The glutathione peroxidase mimetic ebselen killed the cells by a novel, rapid form of death, preceded by cell blebbing and prevented by N-acetylcysteine, suggesting toxicity is not due to ebselen's antioxidant activity. Addition of N-acetylcysteine to remove endogenous H(2)O(2) stimulated survival and proliferation, suggesting that basal levels of H(2)O(2) promoted cell death. Consistent with this, leukemic cell death was induced by adding as little as 5 microM H(2)O(2). Ascorbic acid, even at 100 microM, induced death through H(2)O(2) production. Thus H(2)O(2) does not promote proliferation and survival, rather the opposite, and previous literature may have misinterpreted the effects of antioxidants. Arginase, H(2)O(2), ascorbic acid, and ebselen might be useful in the treatment of leukemia.