1. Transmembrane current through ATP-regulated K+ channels (I-K(ATP)) was measured in ventricular heart cells of the guinea-pig in the whole-cell and cell-attached patch configurations under conditions of metabolic poisoning with the mitochondrial uncoupler 2,4-dinitrophenol (DNP). 2. Maintained exposure of the cells to DNP resulted in a transient appearance of whole-cell I-K(ATP). When I-K(ATP) had reached several nanoamps, blocking the forward-running Na+-K+ pump with 0.5 mM strophanthidin decreased I-K(ATP) after a delay. The time course of this decrease could be described by a single exponential function, which yielded a time constant (tau) of 4.51 +/- 1.89 s (n = 8). 3. Hyperpolarization from 0 mV to -100 or -150 mV for 2 s caused I-K(ATP) (measured at 0 mV) to decrease by 34.2 +/- 14.1% (n = 8) and 37.6 +/- 9.4% (n = 8), respectively. After the hyperpolarizing pulse, I-K(ATP) returned to its higher initial level within a couple of seconds. 4. Driving the pump backwards by removing the extracellular Kf ions caused the permanent disappearance of DNP-induced I-K(ATP). 5. Application of 0.5 mM strophanthidin in the absence of external Kf ions induced a transient increase in I-K(ATP) as did washing out the glycoside (n = 5). 6. When pump action was inhibited by using Na+, K+-free Tyrode solution (see Methods) in the bath, strophanthidin did not have a comparable direct effect on I-K(ATP). 7. In cell-attached patches, strophanthidin applied via the bath caused a reduction in I-K(ATP) with a similar time course to that in whole-cell experiments. This suggests that the interaction between the pump molecules and the K-ATP channels is not restricted to closely neighbouring molecules. 8. The data support the hypothesis that [ATP] at the cytosolic face of the membrane may drop to practically zero, thereby passing an 'ATP window' in which the channels first open and then close, and that the submembrane [ATP] is readily controlled by the cytosolic [ATP].