ATP-inhibited potassium channels (K(ATP)) were studied in excised, inside-out patches from cultured adult mouse pancreatic beta-cells and HIT cells. In the absence of ATP, ADP opened K(ATP) channels at concentrations as low as 10 microM and as high as 500 microM, with maximal activation between 10 and 100 microM ADP in mouse beta-cell membrane patches. At concentrations greater than 500 microM, ADP inhibited K(ATP) channels while 10 mM virtually abolished channel activity. HIT cell channels had a similar biphasic response to ADP except that more than 1 mM ADP was required for inhibition. The channel opening effect of ADP required magnesium while channel inhibition did not. Using creatine/creatine phosphate solutions with creatine phosphokinase to fix ATP and ADP concentrations, we found substantially different K(ATP)-channel activity with solutions having the same ATP/ADP ratio but different absolute total nucleotide levels. To account for ATP-ADP competition, we propose a new model of channel-nucleotide interactions with two kinds of ADP binding sites regulating the channel. One site specifically binds MgADP and increases channel opening. The other, the previously described ATP site, binds either ATP or ADP and decreases channel opening. This model very closely fits the ADP concentration-response curve and, when incorporated into a model of beta-cell membrane potential, increasing ADP in the 10 and 100 microM range is predicted to compete very effectively with millimolar levels of ATP to hyperpolarize beta-cells. The results suggest that (i) K(ATP)-channel activity is not well predicted by the "ATP/ADP ratio," and (ii) ADP is a plausible regulator of K(ATP) channels even if its free cytoplasmic concentration is in the 10-100 microM range as suggested by biochemical studies.