Two different enzyme assays, both based on the interaction of native reverse transcriptase (RT) and 3'-azido-2',3'-deoxythymidine triphosphate (AZT-TP), were used to characterize the enzymes from 18 HIV-1 isolates with decreased sensitivity to AZT in cell culture. The first assay, which measures the balance between incorporation and excision of AZT monophosphate in the presence of dNTP substrate (in terms of IC(50)), gave an approx. 9-fold variation in sensitivity to AZT-TP. There was a correlation between the IC(50) values and the sensitivity of the corresponding virus to AZT in cell culture (r=0.60, P<0.01). The second assay, which was designed specifically for measurement of chain termination in the absence of dNTP substrate (as the concentration of AZT-TP giving 50% residual primer function, or CT(50)), revealed a more than 600-fold difference between the different isolate RTs. For the majority of enzymes there was a strict correlation between the results from the two assays; however, four isolates exhibited significantly higher CT(50)/IC(50) ratios than the other isolates. These differences were not related to sensitivity of the corresponding viruses to AZT but to the occurrence of certain mutations in their pol gene. The four deviating isolates contained either a minimum of four AZT-specific substitutions, including Thr-215-->Tyr (isolates 134 and 143), or some of the known specific substitutions combined with Thr-39-->Ala (isolates 80 and 157). The Thr-39-->Ala substitution has previously been recorded in connection with AZT/Foscarnet combination therapy.