Aromatase, a cytochrome P450, catalyzes three consecutive hydroxylation reactions converting C19 androgens to aromatic C18 estrogenic steroids. On the basis of a recent computer modeling of the active site of aromatase, a hydrophobic pocket, thought to be important for the binding of some aromatase inhibitors, was predicted to extend roughly in the plane of the steroid substrate, from the position that would be occupied by its C4 and C7 atoms. Four mutants, G121A, I125N, F235N, and I474F, were generated to test this model. Although the mutagenesis results have shown that the current model for the active site of aromatase almost certainly contains a number of errors, the results are in general very satisfactory in that they suggest how the model should be altered by local realignments of the aromatase sequence with that of cytochrome P450cam. Among the mutants, I474F is the most interesting one. Its Km value for androstenedione was found to be lower than the wild type enzyme, and the kinetic analysis exhibited a substrate inhibition-like kinetic profile through an "in-cell" assay. In addition, this mutation reduces the binding affinity of an aromatase inhibitor, 4-hydroxyandrostenedione, and increases the binding affinity of two aromatase inhibitors, aminoglutethimide and CGS 16949. This study demonstrates a useful approach, by a combination of computer modeling, site-directed mutagenesis, and inhibitor binding studies, to examine the structure of the active site of aromatase and the binding nature of various aromatase inhibitors.