Modeling and mutagenesis of the active site of human P450c17.

Cytochrome P450c17 is the single enzyme having steroid 17 alpha-hydroxylase and 17,20 lyase activities. We sought to model the active site of this enzyme to identify residues contributing to its catalytic activities, and to test the roles of the identified amino acids by altering them via site-directed mutagenesis. Using the MIDAS-plus program, we modeled P450c17 and the structurally related steroid 21-hydroxylase, P450c21, on the crystallographically determined structure of bacterial P450cam. By positioning the progesterone substrate into each model, we identified five residues that appeared crucial for determining whether progesterone would undergo 17 alpha-hydroxylation (by P450c17) or 21-hydroxylation (by P450c21). Each identified residue in the P450c17 sequence was changed to the corresponding residue in the P450c21 sequence, yielding the four P450c17 mutants L102Y, G111D, G301l, and M369L + l371L. The mutants were transfected into COS-1 cells and their 17 alpha-hydroxylase, 17,20 lyase, and 21-hydroxylase activities were assayed by incubation with [14C]pregnenolone, [3H]17OH-pregnenolone, and [3H]17OH-progesterone and TLC. The L102Y and M369L + l371L mutants retained 50-80% of 17 alpha-hydroxylase and 70-100% of 17,20 lyase activity, while the G111D and G301l mutants lost both activities, but no mutants acquired detectable 21-hydroxylase activity (0.1% of wild type P450c21). Combination of the two mutants that retained partial activity (L102Y and M369L + l371L) yielded a single protein that retained 40% of 17 alpha-hydroxylase and 50% of 17,20 lyase activity, but none of the seven possible vectors expressing two, three, or all four of the mutations in a single enzyme yielded detectable 21-hydroxylase activity. The mutations D298V and D298S were predicted to ablate 17,20 lyase activity while retaining 17 alpha-hydroxylase activity, but were both inactive. These studies indicate that models based on the crystal structure of P450cam correctly predict many gross architectural features of steroidogenic enzymes and that many of the predicted residues are in or near the active site of P450c17. However, because enzymatic activity requires interactions between the enzyme and substrate at distances of less than 1 A, and modeling cannot predict atomic loci to greater than 1.5-2.0 A, it was not possible to design mutants that would confer 21-hydroxylase activity to P450c17. Currently available data cannot predict the structural and amino acid sequence requirements for a specific P450 activity.