Examination of the performance of semiempirical methods in QM/MM studies of the SN2-like reaction of an adenylyl group transfer catalysed by ANT4′

Abstract

Quantum mechanical (QM) semiempirical methods (SMs), combined with molecular mechanics (MM) force fields, are extensively used in theoretical studies of enzymatic reactions. Despite being several orders of magnitude faster than ab initio methods, their correctness is essential to be used in calculations requiring statistical simulations. Herein, a wide range of SMs are examined, from those based on s and p orbitals, sp-SMs (MNDO, AM1, PM3 and RM1), to those including d orbitals, spd-SMs, either based on approximations to the Hartree–Fock theory (MNDO/d, PM6 and AM1/d-PhoT) or derived from density functional theory (DFTB3). These QM Hamiltonians are used within a multiscale QM/MM additive scheme, to clarify their usefulness in mechanistic studies of phosphoryl-transfer reactions. The SN2-like reaction of the adenylyl group transfer catalysed by 4′-O-Nucleotidyltransferase (ANT4′) was selected as a benchmark. Geometrical characteristics of stationary structures, the shape of potential energy surfaces together with the barrier heights and kinetic isotope effects (KIEs), obtained with the different SMs/MM methods were compared with results obtained at higher M06-2X/MM level of theory. Critical limitations of the sp-SMs in the present mechanistic study were detected. The spd-SMs describe the reaction as a concerted process, same as the reference method M06-2X, but none of them is free of limitations. PM6 reproduces the biased trend of previous sp-SMs stabilizing structures of phosphorous atoms with certain pentavalent character, while AM1/d-PhoT and DFTB3 describe TSs more dissociative than M06-2X, which determines the lower quality of the computed primary and secondary 16O/18O KIEs. Efforts to improve the SMs can be guided by the exposure of their limitations, which were supported by the results of a second studied phosphoryl-transfer reactions; the hydrolysis of phosphodiester bond at the 3′-end of the viral DNA (vDNA). Thus, for instance, further increases in SMs accuracy can be achieved by improving the training and survey reference data sets, a more complete set of parameters for describing intermolecular interactions or further developments of spd-SMs.