Coordination number of zinc ions in the phosphotriesterase active site by molecular dynamics and quantum mechanics

We have run several molecular dynamics (MD) simulations on zinc‐containing phosphotriesterase (PTE) with two bound substrates, sarin and paraoxon, and with the substrate analog diethyl 4‐methylbenzylphosphonate. A standard nonbonded model was employed to treat the zinc ions with the commonly used charge of +2. In all the trajectories, we observed a tightly bound water (TBW) molecule in the active site that was coordinated to the less buried zinc ion. The phosphoryl oxygen of the substrate/inhibitor was found to be coordinated to the same zinc ion so that, considering all ligands, the less buried zinc was hexa‐coordinated. The hexa‐coordination of this zinc ion was not seen in the deposited X‐ray pdb files for PTE. Several additional MD simulations were then performed using different charges (+1, +1.5) on the zinc ions, along with ab initio and density functional theory (DFT) calculations, to evaluate the following possibilities: the crystal diffraction data were not correctly interpreted; the hexa‐coordinated zinc ion in PTE is only present in solution and not in the crystal; and the hexa‐coordinated zinc ion in PTE is an artifact of the force field used. A charge of +1.5 leads to a coordination number (CN) of 5 on both zinc ions, which is consistent with the results from ab initio and DFT calculations and with the latest high resolution X‐ray crystal structure. The commonly used charge of +2 produces a CN of 6 on the less buried zinc. The CN on the more buried zinc ion is 5 when the substrate/inhibitor is present in the simulation, and increases to 6 when the substrate/inhibitor is removed prior to the simulation. The results of both of the MD and quantum mechanical calculations lead to the conclusion that the zinc ions in the PTE active site are both penta‐coordinated, and that the MD simulations performed with the charge of +2 overestimate the CN of the zinc ions in the PTE active site. The overall protein structures in the simulations remain unaffected by the change in zinc charge from +2 to +1.5. The results also suggest that the charge +1.5 is the most appropriate for the molecular dynamics simulations on zinc‐containing PTE when a nonbonded model is used and no global thermodynamic conclusion is sought. We also show that the standard nonbonded model is not able to properly treat the CN and energy at the same time. A preliminary, promising charge‐transfer model is discussed with the use of the zinc charge of +1.5. © 2003 Wiley Periodicals, Inc. J Comput Chem 24: 368–378, 2003