1. Patch clamp recording (whole cell configuration) was used to study the action of ATP on rat phaeochromocytoma (PC12) cells usually held at -70 mV and rapidly superfused with buffered saline. ATP (0.5, 1 or 5 mM), applied from micropipettes by pressure application with brief (< or = 50 ms) pulses, induced inward currents with rapid onset and decay. ADP and alpha, beta-methylene ATP were ineffective. 2. ATP (5 mM) applied with pulses > 200 ms long elicited a complex current response characterized by a rapid peak which faded and was followed by a strong current rebound (lasting several s) as soon as the application was terminated. This type of response was readily replicated as long as ATP applications were spaced at 2-3 min intervals. The amplitude of peak and rebound currents was dependent on the length of pressure pulse and was similarly depressed by bath application of a threshold dose (25 microM) of ATP. Rapid fading and rebound of ATP-induced membrane currents were also observed when the Y-tube method was used for applying this agonist. 3. The reversal potential for peak and rebound currents was the same while the time constant values for peak fading and rebound onset were insensitive to changes in membrane potential between -70 and -40 mV. When ATP was applied to a cell clamped at depolarized potential, no current was observed but rapid return of the membrane potential to -70 mV immediately at the end of ATP application was associated with a large rebound current. 4. Brief (20 ms) application of ATP during the onset of the rebound current strongly and transiently suppressed it. The same application performed during the gradual decay of the rebound wave elicited a transient inward current which was much smaller and shorter than the one observed when the cell was in its resting state. Application of 2 s ATP pulses at 20 s intervals equally reduced the initial peak and rebound currents which recovered at the same rate. 5. The present data are interpreted according to a scheme which suggests two types of ATP receptor desensitization. The first one (D1) would be characterized by fast kinetics and low agonist affinity; rapid recovery from D1 would then be manifested as current rebound presumably due to receptor reactivation. The second desensitized state (D2) has slow kinetics and high affinity for the agonist: it is therefore typically seen with sustained application of a low dose of ATP. It is proposed that desensitization and its recovery can influence the time course of membrane responses mediated by purinoceptors.