The mechanisms by which volatile anesthetics induce vasodilation are unknown. Recent studies of adenosine triphosphate-sensitive potassium channels (KATP channels) in the vascular smooth muscle of the coronary circulation suggest that these channels play a role in the coronary artery dilation produced by hypoxemia, the coronary blood flow (CBF) reactive hyperemic response, and in CBF auto regulation. We therefore conducted this study to determine the role of KATP channels in isoflurane-induced coronary vasodilation. Studies were conducted in six open-chest, anesthetized swine. The left anterior descending coronary artery was cannulated and perfused by blood passed through a membrane oxygenator. This preparation allowed us to administer drugs and volatile anesthetics regionally to the perfused myocardium, minimizing systemic effects. Regional CBF response to 1.5% and 3.0% isoflurane administered via the membrane oxygenator was measured before and after blockade of KATP channels, and was compared to the vasodilation produced by regional administration of several doses of sodium nitroprusside and adenosine. Blockade of KATP channels was achieved by regional intracoronary administration of glibenclamide (1-22 micrograms.kg-1.min-1), a specific blocker of these channels. Administration of 1.5 and 3.0 percent isoflurane increased regional CBF by 29 +/- 29% and by 62 +/- 28%, respectively. Under control conditions, blockade of KATP channels decreased mean CBF by 18%, but did not cause ischemia. KATP channel blockade totally eliminated the vasodilator response to both doses of isoflurane. During KATP channel blockade the response to 3% isoflurane was converted to net vasoconstriction: mean delta CBF = -5% +/- 6%, P = < 0.05 versus control. Negative inotropic effects of isoflurane were not eliminated by glibenclamide. Because KATP channel blockade was so effective in eliminating isoflurane-induced coronary vasodilation, the dose of glibenclamide was decreased in sequential experiments, but total blockade of isoflurane vasodilation was achieved even at the smallest dose of glibenclamide studied (1 microgram.kg-1.min-1). The vasodilator response to nitroprusside was not affected, and the vasodilator response to adenosine was partially inhibited (consistent with their known mechanisms of action). Blockade of KATP channels by glibenclamide completely inhibits isoflurane-induced coronary vasodilation in the regionally perfused swine myocardium. The response to sodium nitroprusside, a drug that induces vasodilation via a different mechanism, was unaffected. The response to adenosine, a drug whose vasodilation is partially mediated via KATP channels, was partially inhibited. These results suggest that in vivo isoflurane-induced coronary artery vasodilation is predominantly mediated by KATP channels.