The Glycine N-Methyltransferase Case Study: Another Challenge for QM-Cluster Models?

Abstract

The methyl transfer reaction between SAM and glycine catalyzed by glycine N-methyltransferase (GNMT) was examined using QM-cluster models generated by Residue Interaction Network ResidUe Selector (RINRUS). RINRUS is a Python-based tool that can build QM-cluster models with rules-based processing of the active site residue interaction network. This way of enzyme model-building allows quantitative analysis of residue and fragment contributions to kinetic and thermodynamic properties of the enzyme. Many residue fragments are important for the GNMT catalytic reaction, such as Gly137, Asn138, and Arg175, which interact with the glycine substrate, and Trp30, Asp85, and Tyr242, which interact with the SAM cofactor. Our study shows that active site fragments that interact with the glycine substrate and the SAM cofactor must both be included in the QM-cluster models. Even though the proposed mechanism is a simple one-step reaction, GNMT may be a rather challenging case study for QM-cluster models because convergence in energetics requires models with >350 atoms. "Maximal" QM-cluster models built with either qualitative contact count ranking or quantitative interaction energies from functional group symmetry adapted perturbation theory provide acceptable results. Hence, important residue fragments that contribute to the energetics of the methyl-transfer reaction in GNMT are correctly identified in the RIN. Observations from this work suggest new directions to better establish an effective approach for constructing atomic-level enzyme models.

Figure 1:

The active site of the X-ray crystal structure 1NBH chain A and the 3D structure shown in stick representation of the maximal model (Res27). The seed that includes the SAM and Gly is highlighted in magenta, and all other residues in the RIN are shown in green.

Figure 2:

2D structure of the maximal Res27 QM-cluster model which includes all residues that interact with SAM and glycine in PDB 1NBH.

Figure 3.

Free energy diagram of different sized Himo-based models (in blue, computed using B3LYP/6–31G(d’)/6–31G, built the same way as Himo) along with the Himo-Seed model (SAM plus Gly) and Himo-largest model plus His142 (Himo-L+H) built based on the models in ref 35. The free energies of the original models reported in ref 35 are shown in red. Symbols on top (triangles), middle (circles) and bottom (crosses) show the free energies of activation, reaction and charges for each model, respectively. The dashed line in magenta represents the experimental rate constant value converted to a free energy of activation of 17.5 kcal/mol.

Figure 4.

Free energy diagram of the 23 different sized RINRUS models along with the largest model in ref built and computed in this study (B3LYP/6–31G(d’)/6–31G). Figure on top, middle and bottom show the free energies of activation for N-demethylation, reaction free energies, and charges for each model, respectively. The energies of Himo-L1 model (98 atoms) are shown in blue empty symbols. The dashed line in magenta represents the experimental value converted free energy of activation of 17.5 kcal/mol. The red dotted lines show the tight energy convergence range (within ±1 kcal/mol of the ΔG^‡ and ΔG_rxn of the maximal model Res27), the green dotted lines show the loose energy convergence range (within ±3 kcal/mol of the ΔG^‡ and ΔG_rxn of the maximal model Res27).

Figure 5:

Free energies of activation and reaction of QM-cluster model built based on probe ranking of residue fragments interactions with Gly substrate only as the seed. The dashed line in magenta represents the experimental value converted free energy of activation of 17.5 kcal/mol. The red dotted lines show the tight energy convergence range (within ±1 kcal/mol of the ΔG^‡ and ΔG_rxn of the maximal model Res27), the green dotted lines show the loose energy convergence range (within ±3 kcal/mol of the ΔG^‡ and ΔG_rxn of the maximal model Res27).

Figure 6:

Free energy diagram of the 21 different sized models built based on residue ranking using F-SAPT interaction energy with seed Gly only. Figure on top, middle and bottom show the free energies of activation for N-demethylation, reaction free energies, and charges for each model, respectively. The dashed line in magenta represents the experimental value converted free energy of activation of 17.5 kcal/mol. The red dotted lines show the tight energy convergence range (within ±1 kcal/mol of the ΔG^‡ and ΔG_rxn of the maximal model Res27), the green dotted lines show the loose energy convergence range (within ±3 kcal/mol of the ΔG^‡ and ΔG_rxn of the maximal model Res27).

Figure 7:

Free energy diagram of the 24 different sized models built based on residue ranking using F-SAPT interaction energy with seed Gly plus SAM. Figure on top, middle and bottom show the free energies of activation for N-demethylation, reaction free energies, and charges for each model, respectively. The dashed line in magenta represents the experimental value converted free energy of activation of 17.5 kcal/mol. The red dotted lines show the tight energy convergence range (within ±1 kcal/mol of the ΔG^‡ and ΔG_rxn of the maximal model Res27), the green dotted lines show the loose energy convergence range (within ±3 kcal/mol of the ΔG^‡ and ΔG_rxn of the maximal model Res27).

Scheme 1:

Reaction mechanism of the methyl transfer by Glycine N-methyltransferase (GNMT), which catalyzes the transfer of a methyl group from S-adenosylmethionine (SAM) to glycine, resulting in the formation of S-adenosylhomocysteine (SAH) and sarcosine (N-methylglycine). In this one step reaction, methyl group is located in-between the S atom of SAM and N atom of glycine.