The channel-forming polyperforins P1 and P2 are thought to be formed from the contents of dense core vesicles of cytolytic effector cells. To test this hypothesis, granules from various cytotoxic effector cells were assayed for cytolytic activity on nucleated or unnucleated targets. The results show that in general, granules from cytolytic effector cells are cytolytic, whereas granules from noncytotoxic cells are not. Cytotoxicity of granules is not specific, but there appears to be a preference in that nucleated targets are lysed better than are erythrocytes by granules from T killer or natural killer cells. Granules from CTLL-2, however, preferentially lyse erythrocyte targets. This cell line has been in culture for a long period of time and has lost its cytotoxicity. We tested whether granules from CTLL-2 caused formation of transmembrane pores in erythrocyte target membranes. We found that granule- and complement-induced lesions have similar pore sizes. They are big enough to allow the total release of alpha-bungarotoxin, an 8000 Mr polypeptide with dimensions of 4 X 2.5 nm. Larger molecules are released partially or not at all. Under acidic conditions (pH 5.4) granules do not permeabilize target membranes. This may suggest a pH-dependent control mechanism in the formation, insertion, or function of polyperforin channels, in addition to a previously recognized Ca2+-dependent mechanism. Permeabilization of lipid vesicles by granules was studied to explore what the molecular requirements for channel insertion into membranes may be. Release of alpha-bungarotoxin induced by granules was observed in liposomes made of soybean lipid with or without cholesterol, suggesting that no membrane component other than lipid is required for the insertion of polyperforins, and that the action of polyperforins does not require other mechanisms in the target cell. When pure lecithin from soybean and egg, or synthetic phosphatidylcholines were used, slower release or no release of macromolecules was observed. We suggest that some kind of lipid specificity is required for perforin action. This may be related to the hydrophobic region of the lipid bilayer rather than to the polar portion, because different lecithins with varying fatty acid composition gave similar results.