Though data from cell lines are abundant, the reason for the development of resistance to 1-beta-D arabinofuranosylcytosine (ara-C) in vivo remains unresolved. A broad interpatient variation of metabolic parameters has further complicated interpretation of the results. The present study compares ara-C metabolism in leukemic blasts of two patients with newly diagnosed disease, before and after repeated treatment with ara-C containing chemotherapy regimens in vivo. Membrane transport of ara-C was unchanged after treatment. In addition, cell-free extracts of blasts obtained after treatment failure showed an unchanged cytidine deaminase activity. Though deoxycytidine kinase activity in cell extracts was unaltered or increased after treatment failure, the activity in situ, measured as the rate of 1-beta-D-arabinofuranosylcytosine triphosphate (ara-CTP) formation, was decreased. This could be shown to be due to an expansion of the deoxycytidine triphosphate (dCTP) pool. The severalfold increase in dCTP pool was accompanied by a decrease in thymidine triphosphate (dTTP) pool and correlated with a decrease in deoxycytidylate deaminase (dCMP-deaminase) activity in cell free extracts. Low dCMP-deaminase activity had been shown to confer an ara-C resistant phenotype to cell lines in vitro. Data presented in this paper show that a selection for leukemic blasts with low dCMP-deaminase activity can also be favored by ara-C containing treatment regimens in vivo. Our data suggest that this mechanism might contribute to treatment failure.