Application of molecular modeling for prediction of substrate specificity in cytochrome P450 1A2 mutants.

Molecular dynamics (MD) simulations of 7-ethoxy- and 7-methoxyresorufin bound in the active site of cytochrome P450 (P450) 1A2 wild-type and various mutants were used to predict changes in substrate specificity of the mutants. A total of 26 multiple mutants representing all possible combinations of five key amino acid residues, which are different between P450 1A1 and 1A2, were examined. The resorufin substrates were docked in the active site of each enzyme in the productive binding orientation, and MD simulations were performed on the enzyme-substrate complex. Ensembles collected from MD trajectories were then scored on the basis of geometric parameters relating substrate position with respect to the activated oxoheme cofactor. The results showed a high correlation between the previous experimental data on P450 1A2 wild-type and single mutants with respect to the ratio between 7-ethoxyresorufin-O-deethylase (EROD) and 7-methoxyresorufin-O-demethylase (MROD) activities and the equivalent in silico "E/M scores" (the ratio of hits obtained with 7-ethoxyresorufin to those obtained with 7-methoxyresorufin). Moreover, this correlation served to establish linear regression models used to evaluate E/M scores of multiple P450 1A2 mutants. Seven mutants, all of them incorporating the L382V substitution, were predicted to shift specificity to that of P450 1A1. The predictions were then verified experimentally. The appropriate P450 1A2 multiple mutants were constructed by site-directed mutagenesis, expressed in Escherichia coli, and assayed for EROD and MROD activities. Of six mutants, five demonstrated an increased EROD/MROD ratio, confirming modeling predictions.