B y studying changes in the clonal composition of HIV-1 populations during the first weeks of zidovudine (ZDV) treatment before the development of ZDV resistance-conferring mutations, we demonstrated previously a selective inhibition of nonsyncytium-inducing (NSI) HIV-1, even when present as coexisting population in individuals also harboring syncytium-inducing (SI) HIV-1. In this study, we observed the opposite in individuals receiving didanosine (ddI) treatment. In these individuals (n = 7) a median -0.98 log change (range -1.55-0.08) in infectious cellular SI load was observed, whereas the coexisting NSI load was only minimally affected (median -0.15 log, range -1.27-0.50; P = 0.03). The virus phenotype-dependent treatment responses were independent of the clonal composition of HIV-1 populations at baseline. Individuals treated with a combination of ZDV and ddI revealed an equal decline of both NSI and SI infectious cellular load (n = 4; NSI: median -1.55 log, range -2.19 to -1.45; SI: median -1.47 log, range -1.81 to -0.86; P = 0.56). To test the hypothesis that the previously reported optimal activation of ZDV and ddI in activated and resting T cells, respectively, in combination with the differential T cell tropism of NSI and SI HIV-1 is the basis for the observed virus phenotype specific efficacy of nucleoside analogs, we studied the effect of treatment with a protease inhibitor that does not require intracellular activation. Individuals receiving ritonavir (n = 4) indeed showed equal declines in NSI and SI infectious cellular load (NSI: median -2.37 log, range -2.59 to -2.16; SI: median -2.82 log, range -3.14 to -2.50; P = 0.25). Our data suggest HIV-1 phenotype as an additional parameter in the design of optimal treatment regimens.