A series of thiol androgens were synthesized and investigated to characterize structural features important for the inhibition of aromatase. Analogues of androstenedione with thiol groups in either the 2 alpha-, 10 beta-, or 19-positions caused time-dependent inhibition of human placental aromatase. When their KI and kcat values were compared with those of 4-hydroxyandrost-4-ene-3,17-dione (4-OHa) and 10 beta-propargylestr-4-ene-3,17-dione (PED), the thiol androgen 10 beta-mercaptoestr-4-ene-3,17-dione (10 beta-SHnorA) proved to be the most potent suicide substrate. However, 19-mercaptoandrost-4-ene-3,17-dione (19-SHA) was the best all-around inhibitor. All compounds except 19-SHA exhibited normal type I P-450 difference spectra with partially purified/solubilized, human placental aromatase. The Ks values for the series of compounds compared qualitatively to the KI values determined from the time and concentration-dependent inhibition experiments. 19-SHA induced split Soret peaks at 380 and 474 nm, which suggested binding of the 19-thiolate directly to the ferric iron of aromatase. This binding could be displaced by aminoglutethimide but not by androstenedione. The inhibitory activity of 19-SHA may be explained by two independent mechanisms: (1) suicide inactivation of aromatase in the ferrous state; and (2) a direct "hyper-type II" binding to the remaining portion of the cytochrome in the ferric state. A free thiol group was necessary for the suicide inhibitory activity of 19-SHA; time-dependent inactivation of aromatase by 19-(acetylthio)androst-4-ene-3,17-dione (19-SAcA) and 19-xanthogenylandrost-4-ene-3,17-dione (19-XanA) could be prevented if the microsomes were preincubated with a carboxyesterase inhibitor. Aromatase previously inactivated by either thiol androgens,4-OHA, or PED could not be reactivated after incubation with the disulfide reducing agent dithiothreitol, which suggests that a disulfide bond may not be involved in aromatase inactivation by these inhibitors.