Binding of flexible ligands to proteins: Valproic acid and its interaction with cytochrome P450cam



A combined theoretical and experimental study of the binding and interaction of valproic acid (VPA) with the bacterial cytochrome P450cam enzyme and the determination of regio- and stereoselective hydroxylation product distribution was performed. From the experiments, C4[BOND];OH VPA was found to be the predominant hydroxylation product with a small amount of C5[BOND]OH VPA formed. The experimental stereoselectivity of hydroxylation was 2R4S > ∼ 2S4R > 2R4R > ∼ 2S4S and 2S5 > ∼ 2R5. The overall goals of the theoretical study were twofold: (1) to characterize as completely as possible, using energy optimization and molecular dynamics simulations, the interactions of flexible ligands with their target proteins, and (2) to determine the extent to which these results could be used to develop criteria to predict or explain the experimentally observed regio- and stereoselectivity of hydroxylation of the flexible ligands. Among the useful insights into the behavior of flexible ligands upon binding to their traget proteins obtained are (1) a large change in conformation occurs for many conformers of VPA upon binding to P450cam, (2) low- energy conformers of VPA do not necessarily lead to optimum interactions with the target protein, and (3) the most favorable mode of interaction of this flexible ligand with the protein binding site has been identified and found to be a result of strong electrostatic interactions between VPA and both Tyr96 and Asp297. For the study of the hydroxylated VPA products, the challenging aspect of this problem was to determine criteria for weighing the contribution of each of the possible protein–ligand complexes. To this end, various possibilities were examined and compared with the experimental results. No single complex was found to reproduce the observed experimental regio- and stereoselectivity. This result indicates that more than one bioactive form of VPA contributes to its oxidation. Results most consistent with experiment are obtained by using the interaction energy of the protein–ligand complex as a criterion for including its contribution to product formation. Although there are remaining disparities between predicted and observed product distributions, the combined theoretical and experimental effort has led to insights into the modes of interaction of this flexible ligand that lead to its observed product specificity. © 1993 John Wiley & Sons, Inc.