The most promising nucleoside analogs that are currently undergoing preclinical and clinical testing for anti-HIV activity belong to the dideoxynucleoside group. We have studied the toxicity of 3'-azido,3'-deoxythymidine (AZT), 2',3'-dideoxycytidine (DDC), and 2',3'-dideoxyinosine (DDI) in canine bone marrow progenitor cells in culture. AZT potently inhibited both canine CFU-GM and CFU-E with IC50 values of 2 and 8 mumol/l respectively, while DDC was relatively non-toxic to either progenitor with IC50 of > 200 mumol/l and 80 mumol/l respectively. DDI was mildly toxic to the bone marrow progenitors, with IC50 values of 62 mumol/l for CFU-GM and 70 mumol/l for CFU-E. Dipyridamole, a nucleoside transport inhibitor, did not influence the toxicity of these dideoxynucleosides in either progenitor at concentrations up to 10 mumol/l. Using uridine as the prototype endogenous nucleoside, we have demonstrated that there is a saturable "zero-trans" nucleoside transport system in canine bone marrow mononuclear cells, which is completely inhibited by 1 mumol/l dipyridamole (Ki = 0.02 mumol/l). None of the dideoxynucleosides appeared to be a substrate for this transport system, and dipyridamole did not alter their influx. Permeation of radiolabeled AZT into bone marrow mononuclear cells was slow and non-saturable, while the permeation of DDI was even slower. DDC did not permeate bone marrow cells well, with very little cell accumulation even after 2 hours of equilibration. Our toxicity data from canine bone marrow progenitor cells paralleled the clinical hematotoxicity profiles of these dideoxynucleosides in AIDS patients and suggest that the myelotoxicity of a nucleoside analog is related to its ability to permeate the progenitor cells in question. Canine bone marrow progenitor cultures may serve well as an in vitro model for drug hematotoxicity studies.