We previously showed that exercise induces myocardial preconditioning in dogs and that early preconditioning is mediated through mitochondrial adenosine triphosphate-sensitive potassium channels. We decided to study if late preconditioning by exercise is also mediated through these channels. Forty-eight dogs, surgically instrumented and trained to run daily, were randomly assigned to 4 groups: (1) Nonpreconditioned dogs: under anesthesia, the coronary artery was occluded during 1 hour and then reperfused during 4.5 hours. (2) Late preconditioned dogs: similar to group 1, but the dogs run on the treadmill for 5 periods of 5 minutes each, 24 hours before the coronary occlusion. (3) Late preconditioned dogs plus 5-hydroxydecanoate (5HD): similar to group 2, but 5HD was administered before the coronary occlusion. (4) Nonpreconditioned dogs plus 5HD: similar to group 1, but 5HD was administered before the coronary occlusion. Infarct size (percent of the risk region) decreased by effect of exercise by 56% (P < 0.05), and this effect was abolished with 5HD. 5HD by itself did not modify infarct size. Exercise did not induce myocardial ischemia, and the hemodynamics during ischemia-reperfusion period did not differ among groups. These effects were independent of changes in collateral flow to the ischemic region. We concluded that late cardiac preconditioning by exercise is mediated through mitochondrial adenosine triphosphate-sensitive potassium channels.