1. Intracellular mechanism(s) for controlling the opening of muscarinic K+ channels in the absence of an applied muscarinic agonist were studied in rabbit atrium by applying the patch clamp technique to isolate single myocytes. 2. In the cell-attached patch configuration, currents due to the activity of both the muscarinic K+ channel and the inward rectifying K+ channel were recorded. However, while the inward rectifying K+ channel currents were observed in only ten patches of 211 examined, spontaneous opening (i.e. in the absence of a muscarinic agonist) of the muscarinic K+ channel currents was observed in all patches examined in these atrial cells. 3. The single-channel currents due to spontaneous opening of muscarinic K+ channels were identified on the basis of their very similar conductance and gating properties to the unitary events which have been recorded when 0.5-mu-M-acetylcholine is included in the pipette and 10-mu-M-GTP is present in the internal side of the patch membrane. 4. Although the spontaneous opening of the muscarine K+ channels disappeared soon after excision of the patch membrane, this type of channel activity reappeared following application of ATP and MgCl2 to the internal side of the torn-off patch, as expected from previous publications. 5. The K+ channel activity induced by the ATP and Mg2+ (measured as the product of the number of channels, N, times the probability of opening, P(o)) was strongly dependent upon concentration of free Mg2+; it was half-maximal at 2.2 x 10(-4) M [Mg2+]i. However, after the muscarinic K+ channels had been activated by 100-mu-M-guanosine 5'-O-3-thiotriphosphate (GTP-gamma-S) together with ATP and Mg2+, an increase in the Mg2+ concentration from 5.5 x 10(-5) to 2 x 10(-3) M failed to enhance this channel activity. 6. Pertussis toxin, which is known to uncouple muscarinic receptors from associated G-proteins (G(i) or G(o)), failed to inhibit the ATP- and Mg2+ -induced activation of this K+ channel in the absence agonists. 7. In experiments made to test whether the Mg2+-ATP requirements results from an obligatory phosphorylation reaction, ATP was replaced with adenylyl-imidodiphosphate (AMP-PNP), an analogue of ATP which is resistant to hydrolysis. This K+ channel activity was not present when ATP was replaced with AMP-PNP. 8. This Mg2+ -ATP-dependent phosphorylation does not appear to involve activation of protein kinase C since 12-O-tetradecanoylphorbol-13-acetate (TPA, 10 nM), an activator of protein kinase C, did not increase the spontaneous opening of this muscarinic K+ channel in the cell-attached patch configuration; and this channel activity was not inhibited by 1-(5-isoquinolinylsulphonyl)-2-methylpiperazine (H-7, 20-mu-M). 9. To assess the possibility that the phosphorylation reaction required for channel activation may involve a membrane-bound nucleotide diphosphate kinase (NDPK) we applied nucleotide triphosphates which are known to be able to substitute for ATP as substrates for the NDPK. Consistent with this possibility, uridine 5'-triphosphate (UTP) (1 mM), thymidine 5'-triphosphate (TTP) (1 mM) and guanosine 5'-triphosphate (GTP) (1 mM) were able to activate these K+ channels in the presence of 2 mM-MgCl2; however GTP was much less effective than ATP. 10. Additional evidence for the involvement of NDPK in the spontaneous activation of these K+ channels was obtained by applying nucleotides known to be non-specific inhibitors of this enzyme. Adenosine 5'-diphosphate (ADP, 500-mu-M), guanosine 5'-O-(2-thiodiphosphate) (GDP-beta-S, 100-mu-M) and guanosine 5'-diphosphate (GDP, 25-mu-M) each strongly inhibited the spontaneous K+ channel activity induced by ATP and Mg2+. 11. These results demonstrate that the muscarinic K+ channel is regulated by intracellular nucleotides and Mg2+, even when the muscarinic receptors are not occupied by agonist; and they suggest that a transphosphorylation reaction may be essential for this regulation, possibly involving a membrane-bound nucleotide diphosphate kinase.