Acetylcholine is believed to dilate normal blood vessels by promoting the release of a vasorelaxant substance from the endothelium (endothelium-derived relaxing factor). By contrast, if the endothelium is removed experimentally, acetylcholine constricts blood vessels. We tested the hypothesis that muscarinic cholinergic vasodilation is impaired in coronary atherosclerosis. Graded concentrations of acetylcholine and, for comparison, the nonendothelial-dependent vasodilator nitroglycerin were infused into the left anterior descending artery of eight patients with advanced coronary stenoses (greater than 50 percent narrowing), four subjects with angiographically normal coronary arteries, and six patients with mild coronary atherosclerosis (less than 20 percent narrowing). Vascular responses were evaluated by quantitative angiography. In several segments each of four normal coronary arteries, acetylcholine caused a dose-dependent dilation from a control diameter of 1.94 +/- 0.16 mm to 2.16 +/- 0.15 mm with the maximal acetylcholine dose (P less than 0.01). In contrast, all eight of the arteries with advanced stenoses showed dose-dependent constriction, from 1.05 +/- 0.05 to 0.32 +/- 0.16 mm at the highest concentration of acetylcholine (P less than 0.01), with temporary occlusion in five. Five of six vessels with minimal disease also constricted in response to acetylcholine. All vessels dilated in response to nitroglycerin, however. We conclude that paradoxical vasoconstriction induced by acetylcholine occurs early as well as late in the course of coronary atherosclerosis. Our preliminary findings suggest that the abnormal vascular response to acetylcholine may represent a defect in endothelial vasodilator function, and may be important in the pathogenesis of coronary vasospasm.