3'-Azido-3'-deoxythymidine (AZT) is the drug most widely used in the treatment of AIDS. Its major drug-related toxicity is bone marrow suppression, which limits the dose of AZT that can be used. It is essential that AZT be phosphorylated for antiviral effect. We have recently demonstrated that thymidine kinase (TK), the initial enzyme in AZT anabolism, follows a circadian pattern in rat bone marrow. The present study was undertaken to determine whether AZT toxicity is related to the time of its administration and whether the variation in toxicity is correlated with the circadian variation in TK activity. Male Sprague-Dawley rats were housed under standardized conditions of light and dark (lights on 0600 to 1800 and lights off 1800 to 0600) for 4 weeks. The animals were randomly divided into seven groups; six groups were administered AZT by intraperitoneal injection at the same dose of 750 mg/kg of body weight at various times (0400, 0800, 1200, 1600, 2000, and 2400), and one group was used as a control. AZT-related toxic effects, including bone marrow toxicity, differed significantly among the treatment groups, depending on the time of AZT administration (by analysis of variance and Cosinor analysis, P < 0.001). The least toxicity was observed in rats receiving AZT at 1600 (10 h after light onset [10 HALO], in late sleep span) and the greatest toxicity was observed in those injected at 0400 (22 HALO, in late activity span). To verify these results, we administered AZT by intraperitoneal injection at an approximately 50% lethal dose (1,500 mg/kg) to two groups of rats, one at 1200 (6 HALO, in the middle of the sleep span) and the other at 2400 (18 HALO, in the middle of the activity span). AZT lethality was significantly higher in rats receiving AZT at 2400 (18 HALO, in the middle of the activity span). Further statistical analysis demonstrated that the variation in AZT toxicity was correlated with the circadian variation in TK activity in bone marrow of the same species (peak activity at 0400 [22 HALO, in late activity span] and trough activity at 1600 [10 HALO, in late sleep span]), suggesting that the circadian variation in TK activity may be the biochemical basis for the observed circadian variation in AZT toxicity. These results may be useful in the design of improved AZT chemotherapeutic regimens.