Naturally occurring mutations in the polymerase gene of human immunodeficiency virus type 1 (HIV-1) have important implications for therapy and the outcome of clinical studies. Using 42 virus isolates obtained from the UNAIDS sample collection, we analyzed the protease (99 amino acids [aa]) and the first 297 aa of reverse transcriptase (RT) coding regions. Based on the V3 sequence analysis, the collection includes subtype A (n = 5), subtype B (n = 12), subtype C (n = 1), subtype D (n = 11), and subtype E (n = 13) viruses. Of the 42 protease genes, 37 contained naturally occurring mutations at positions in the gene that contribute to resistance to protease inhibitors (indinavir, saquinavir, ritonavir, and nelfinavir) in clade B isolates. The phenotypic effect of these substitutions in non-B isolates is unclear. The The 5'half RT coding region of the 42 isolates was found to be less variable, although 19 of the 42 RT sequences contained amino acid substitutions known to contribute to nucleoside and/or nonnucleoside drug resistance. Since the virus isolates were obtained in 1992, it is unlikely that the infected subjects received protease inhibitors, but we found evidence that one subject acquired a zidovudine (AZT)-resistant HIV-1 strain from a contact who had received AZT. Phylogenetic analysis identified five subtype pol clusters: A, B, C, D, and A'. Comparison of env and pol sequences of the same viruses showed no more recombination events than were already identified on the basis of gag/env comparison (M. Cornelissen, G. Kampinga, F. Zorgdrager, J. Goudsmit, and the UNAIDS Network for HIV Isolation and Characterization, J. Virol. 70:8209-8212, 1996). In one of the known recombinants, a crossover site between subtypes A and C could be identified, and in another, a crossover site could not be identified due to lack of a reference subtype F pol sequence. We analyzed the ds/da ratio of gag, pol, and env sequences of 35 isolates, excluding the recombinants. Our analysis showed that gag and pol are subjected to purifying selection with an average ds/da ratio above 1, independent of the subtype and in contrast with V3 (ds/da approximately 1). Based on the low ds/da ratio of the intergroup analysis of A/E and B/D gag and pol sequences, we analyzed the evolutionary relation between subtypes B and D in more detail by constructing separate phylogenetic trees for synonymous and nonsynonymous substitutions. Our analysis suggests a common ancestry for subtypes B and D that is distinct from that of subtypes A and E.