The penetration of Bordetella pertussis adenylate cyclase into various mammalian cells exhibits similar kinetics; the accumulation of both intracellular cyclase activity and cyclic AMP is rapid, reaching constant levels after 15-60 min of incubation. The kinetics of enzyme penetration into turkey erythrocytes is different; cyclase activity and cyclic AMP accumulate linearly and do not reach constant levels even after 6 h of incubation. In the preceding paper [Friedman, Farfel & Hanski (1987) Biochem. J. 243, 145-151] we have suggested that the constant level of intracellular cyclase activity reflects a steady state formed by continuous penetration and intracellular inactivation of the enzyme. In contrast with other mammalian cells, no inactivation of cyclase is observed in turkey erythrocytes. These results further support the notion that there is continuous penetration and deactivation of the invasive enzyme in mammalian cells. A 5-6-fold increase in specific activity of the invasive cyclase is detected in a pellet fraction of human lymphocytes in which a similar increase in specific activity of the plasma-membrane marker 5'-nucleotidase is observed. A similar increase in the invasive-cyclase specific activity is detected in a membrane fraction of human erythrocytes. Cyclase activity in a membrane-enriched fraction of human lymphocytes reached a constant level after 20 min of cell exposure to the enzyme. Similar time courses were observed for accumulation of cyclase activity and cyclic AMP in whole lymphocytes [Friedman, Farfel & Hanski (1987) Biochem, J. 243, 145-151]. We suggest therefore that cyclic AMP generation by the invasive enzyme as well as the intracellular inactivation process occur while it is associated with a membrane fraction identical, or closely associated, with the plasma membrane.