Tamoxifen increases the risk of human endometrial cancer and is a potent carcinogen in rat liver, in which it produces DNA adducts and cytogenetic damage. Nevertheless its prophylactic use against breast cancer in healthy women is under investigation in several large trials. To investigate whether rat hepatocarcinogenicity predicts human hepatocarcinogenicity we used genetically engineered bacterial and mammalian target cells to investigate how alpha-hydroxy-tamoxifen, a major phase I metabolite of tamoxifen, is further metabolised by rat and human phase II enzymes, sulfotransferases, to mutagenic and DNA-adduct-forming species. We expressed rat hydroxysteroid sulfotransferase a, a liver-specific enzyme, and corresponding human sulfotransferase in bacteria (Salmonella typhimurium) and in a mammalian cell line (Chinese hamster V79 cells) and tested alpha-hydroxytamoxifen for DNA adduct formation and mutagenicity in these systems, using unmodified cells as controls. In cells that expressed rat hydroxysteroid sulfotransferase, alpha-hydroxytamoxifen was mutagenic and formed the same pattern of DNA adducts as that found in the liver of tamoxifen-treated rats. Alpha-hydroxytamoxifen was not activated, or was at least 20 times less active in cells expressing human hydroxysteroid sulfotransferase. All the other six known human xenobiotic-metabolising sulfotransferases were also expressed in S. typhimurium. None activated alpha-hydroxytamoxifen to a mutagen. These results suggest that the risk of DNA adduct formation, and cancer, in the human liver is low and explain why tamoxifen is a powerful carcinogen to the rat liver, and why standard short-term tests fail to detect its mutagenicity.