Contents lists available at ScienceDirect Computational and Theoretical Chemistry journal homepage: www.elsevier.com/locate/comptc Can a quantum mechanical cluster model explain the special stereospecificity of glyoxalase I? Samaneh Parvaneh, Hadi Parsa, Mehdi Irani ⁎ Department of Chemistry, University of Kurdistan, P.O. Box 66175-416, Sanandaj, Iran ARTICLE INFO Keywords: DFT Enantioselectivity Enzyme Glyoxalase I Mechanism QM-cluster ABSTRACT Quantum mechanical (QM) cluster approach is a useful and convenient tool to study mechanisms of enzymatic reactions. Our investigation is to propose a proper QM-cluster that reproduces the special specificity of glyox- alase I (GlxI). GlxI converts S and R enantiomers of hemithioacetal to only S-D enantiomer of its product. It shows this special specificity with two glutamate residues that are symmetrically coordinated to a Zn ion. However, molecular dynamics simulations showed that one of the glutamates, Glu-172 is more flexible than the other one, Glu-99. Results indicate that a QM-cluster with a more flexible model of Glu-172 can reproduce proposed mechanisms for the S-substrate. However, the same cluster can reach to the R-L enantiomer of the product from the R substrate. We propose the necessity of using much more expensive, hybrid QM/MM methods or a very big QM-cluster to model and study reaction mechanisms of stereospecific enzymes like GlxI. 1. Introduction An essential step in studying a chemical reaction is to clarify its mechanism. This requires identifying intermediates and transition states along the reaction path, and energetics of each reaction step, as well. However, most intermediates and all transition states are short- lived chemical species and are hard to be characterized. Special ex- perimental methods may be applied to trap and characterize the short- lived species. However, these require expensive apparatus and ha- zardous chemicals. Computational methods are excellent alternatives to the experimental ones for studying reaction mechanisms. They do not have the limitation of trapping the short-lived chemical species and are much more environmentally friendly. Furthermore, computational methods consume only electrical energy but not chemicals. These methods are divided into two major classes, i.e., quantum mechanical (QM) and classical molecular mechanics (MM) methods. Computation time is an important aspect of the computational methods. Indeed, the MM methods are much faster than the QM ones, but are silent to study breaking and forming of chemical bonds. In other words, the classical methods are inaccurate for studying chemical reactions. On the other hand, QM methods and a variant of them (density func- tional theory; DFT) are promising to study reaction mechanisms. However, a DFT computation time costs N 4 , where N is the size of the QM system [1]. This imposes limitations in the number of atoms that can be included in the DFT calculations (around 200 atoms). The DFT methods had been used to study mechanisms of a few organic reactions [2–7]. However, intact QM methods cannot be used to study reactions, catalyzed by enzymes, biocatalysts with thousands of atoms. Studying such big systems by QM methods seems impossible, from the hardware point of view. However, there are two solutions for this problem; i) using hybrid QM/MM methods [8–16] and ii) modeling enzymes with rather small clusters (the QM-cluster approach) [17]. The former is much expensive and needs powerful hardware. However, the latter is less expensive and can be performed on hardware accessible to most theoretical chemists. In the QM-cluster method, a few residues of enzymes active sites are separated from a crystal structure as a representative of the whole biocatalysts. The separated atoms, which make a cluster, are used to model enzymatic reactions and to investigate their mechanisms [17]. To compensate the removed surroundings of the active site, two things must be done; i) the QM-cluster is immersed in a continuum solvent model with a dielectric constant of 4 [17], and ii) one or more atoms of each amino acid in the cluster are fixed [18]. The latter is done because the backbone of enzymes hardly moves during chemical reactions, but geometry optimizations move the atoms in the cluster and this makes strange structures and energies [18]. The QM-cluster method has ex- tensively been used to study enzymatic reactions [19–25]. In this study, we test four different QM-clusters (cf. section 2.2) of human glyoxalase I (GlxI; lactoylglutathione lyase; EC 4.4.1.5); to find the best cluster that can represent the special stereospecificity of the https://doi.org/10.1016/j.comptc.2020.112944 Received 5 May 2020; Received in revised form 27 June 2020; Accepted 13 July 2020 ⁎ Corresponding author. E-mail address: m.irani@uok.ac.ir (M. Irani). Computational and Theoretical Chemistry 1188 (2020) 112944 Available online 21 July 2020 2210-271X/ © 2020 Elsevier B.V. All rights reserved. T