The bifunctional Bordetella adenylate cyclase toxin-hemolysin (ACT) penetrates target cell membranes, forms cation-selective channels and subverts cellular signaling by catalyzing uncontrolled conversion of ATP to cAMP. While primarily targeting phagocytes expressing the alphaMbeta2 integrin (CD11b/CD18), the toxin can also penetrate mammalian erythrocytes lacking the receptor and membrane endocytosis. We sought here to analyze the membrane interactions of ACT in a liposome model. Insertion of ACT into liposome membranes required calcium and caused leakage of entrapped fluorescent probes due to liposome disruption, as indicated by similar release kinetics for the approximately 398 Da FITC probe and its approximately 4400 Da dextran conjugate. However, the non-acylated proACT, which does not penetrate cellular membranes, exhibited higher capacity to bind and lyze liposomes than the mature toxin, showing that the fatty-acyl modification was not required for penetration of ACT into the lipid bilayer. Individual deletions within the channel-forming, acylation and repeat domains of ACT abolished its capacity to disrupt both liposomes and erythrocytes. In contrast to erythrocyte binding, however, the liposome binding was only lost upon a simultaneous deletion of both the channel-forming and acylation domains, suggesting that the acylation domain was also involved in liposome penetration of ACT. Moreover, substitutions of glutamates 509 and 516 by lysines, which strongly enhanced the channel-forming and hemolytic activity of ACT, did not affect its capacity to disrupt liposomes. This shows that the mechanism of ACT action in cellular membranes is not fully reproduced in liposome membranes.