Various nucleoside analogues are being used or are being considered for use as therapeutic drugs to inhibit replication of the HTLV-III/LAV virus in infected human cells. Here, the ability of seven nucleoside analogues, a combination of two analogues, and two other therapeutic compounds to induce genotoxic and cytotoxic damage in vivo was evaluated in the mouse bone marrow micronucleus test. Using a 3-consecutive-day oral treatment protocol, almost all of the test chemicals induced a significant increase in the frequency of micronucleated polychromatic erythrocytes (MN-PCE) in male B6C3F1 mice, ranked in decreasing potency as 6-thioguanine greater than Cytarabine HCl greater than 3'-azido-3'-deoxythymidine (AZT)/2',3'-dideoxycytidine combination = AZT greater than Ribavirin = 2',3'-didehydro-3'-deoxythymidine greater than 2',3'-dideoxyadenosine = 2',3'-dideoxycytidine. The frequency of MN-PCE was not increased significantly by treatment with 2',3-dideoxyinosine (DDI) or pentamidine isethionate (PI). The differential ability of AZT and DDI to induce MN in mouse bone marrow was verified from peripheral blood smears prepared from subchronic (90 day) oral studies. The lack of genotoxic activity by DDI was route-specific since, when tested by intraperitoneal injection, a small but significant increase in MN-PCE was observed. A number of these chemicals induced a significant depression in erythropoiesis. However, there was not a significant correlation between the increase in MN-PCE and the depression in the percentage of PCE. This lack of a correlation suggests that factors other than DNA damage may contribute to the inhibition in the rate of erythropoiesis. The presence of increased levels of micronuclei in bone marrow PCE following treatment with various nucleoside analogues suggests that intrinsic genotoxic activity in mammalian cells should be one factor considered during drug selection for AIDS therapy.