17-(Allylamino)-17-demethoxygeldanamycin (17AAG), a compound that is proposed for clinical development, shares the ability of geldanamycin to bind to heat shock protein 90 and GRP94, thereby depleting cells of p185erbB2, mutant p53, and Raf-1. Urine and plasma from mice treated i.v. with 17AAG contained six materials with absorption spectra similar to that of 17AAG. Therefore, in vitro metabolism of 17AAG by mouse and human hepatic preparations was studied to characterize: (a) the enzymes responsible for 17AAG metabolism; and (b) the structures of the metabolites produced. These materials had retention times on high-performance liquid chromatography of approximately 2, 4, 5, 6, 7, and 9 min. When incubated in an aerobic environment with 17AAG, murine hepatic supernatant (9000 x g) produced each of these compounds; the 4-min metabolite was the major product. This metabolism required an electron donor, and NADPH was favored over NADH. Metabolic activity resided predominantly in the microsomal fraction. Metabolism was decreased by approximately 80% in anaerobic conditions and was essentially ablated by CO. Microsomes prepared from human livers produced essentially the same metabolites as produced by murine hepatic microsomes, but the 2-min metabolite was the major product, and the 4-min metabolite was next largest. There was no metabolism of 17AAG by human liver cytosol. Metabolism of 17AAG by human liver microsomes also required an electron donor, with NADPH being preferred over NADH, was inhibited by approximately 80% under anaerobic conditions, and was essentially ablated by CO. Liquid chromatography/mass spectrometry analysis of human and mouse in vitro reaction mixtures indicated the presence of materials with molecular weights of 545, 601, and 619, compatible with 17-(amino)-17-demethoxygeldanamycin (17AG), an epoxide, and a diol, respectively. The metabolite with retention time of 4 min was identified as 17AG by cochromatography and mass spectral concordance with authentic standard. Human microsomal metabolism of 17AAG was inhibited by ketoconazole, implying 3A4 as the responsible cytochrome P450 isoform. Incubation of 17AAG with cloned CYP3A4 produced metabolites 4 and 6. Incubation of 17AAG with cloned CYP3A4 and cloned microsomal epoxide hydrolase produced metabolites 2 and 4, with greatly decreased amounts of metabolite 6. Incubation of 17AAG with human hepatic microsomes and cyclohexene oxide, a known inhibitor of microsomal epoxide hydrolase, did not affect the production of metabolite 4 but decreased the production of metabolite 2 while increasing the production of metabolite 6. These data imply that metabolite 2 is a diol and metabolite 6 is an epoxide. Mass spectral fragmentation patterns and the fact that 17AG is not metabolized argue for the epoxide and diol being formed on the 17-allylamino portion of 17AAG and not on its ansamycin ring. These data have implications with regard to preclinical toxicology and activity testing of 17AAG as well as its proposed clinical development because: (a) production of 17AG requires concomitant production of acrolein from the cleaved allyl moiety; and (b) 17AG, which was not metabolized by microsomes, has been described as being as active as 17AAG in decreasing cellular p185erbB2.