Solvent effects in active‐site molecular dynamics simulations on the binding of 8‐methyl‐N5‐deazapterin and 8‐methylpterin to dihydrofolate reductase

Abstract

Molecular dynamics (MD) simulation methods have been used to study the active site of the ternary complex formed between avian dihydrofolate reductase (DHFR), NADPH cofactor, and the inhibitor 8-methyl-N5-deazapterin in aqueous solution. Spherical shells of water molecules (initially at the bulk-solvent density) are used to solvate the active site and the surrounding protein surface. Two models for treating the effects of the neglected bulk solvent are then considered. The tethered water (TW) model is characterized by the use of harmonic restraining potentials to tether the water molecules to their initial (bulk solvent) positions; whereas, in the capped water (CW) model, water molecules are prevented from escaping from the solvent shell by the use of half-harmonic potentials, but otherwise their motions within the solvation shell are unrestrained. As measured by overall rms differences between coordinates, the distribution of solvent molecules in the active-site region, and the numbers of hydrogen bonds, the TW model compares favorably with the CW model but requires far fewer water molecules, i.e., relatively small solvent shells. The smaller shells of unrestrained water (CW model) gave rise to a distortion in the orientation of the side chain of the active-site residue Tyr-31, whereas no such distortion was apparent in the TW model or for the larger solvent shells in the CW model. A value for the force constant of 0.005 kcal/mol/Å2 for the tethering potential [Solmajer and Mehler, Int. J. Quant. Chem., 44, 291 (1992)] gave satisfactory results for DHFR, although we found that distance-dependent dielectric functions were unable to reproduce accurately the effects of the explicit water models. The free-energy change for the mutation of 8-methyl-N5-deazapterin to 8-methylpterin was computed using both nonsolvated and solvated models. The solvated models gave free energy differences about 1 kcal/mol lower than for nonsolvated models, but the differences between solvated models was much less than 1 kcal/mol. Overall, the calculated differences in thermodynamic stability of the deazapterin and pterin complexes are in fair agreement with experiment, i.e., a small binding differential is predicted. © 1996 by John Wiley & Sons, Inc.