Laboratoire de Chimie et Biochimie Pharmacologiques et Toxicologiques, Université de Paris UMR 8601, Paris, France.
Laboratoire de Chimie Théorique, Paris, France.
J Comput Chem. 2021 Jan 15;42(2):86-106. doi: 10.1002/jcc.26437. Epub 2020 Nov 10.
Molecular dynamics on the complexes of inhibitors with Zn-metalloproteins are a privileged area of applications of polarizable molecular mechanics potentials. With which accuracy could these reproduce the QC intermolecular interaction energies in the two mono-zinc cores and in the dizinc core, toward full-fledged MD simulations on the entire protein complexes? We considered the complexes of the extended recognition site of a Zn-dependent metallo-β-lactamase, VIM-2, produced by bacteria responsible for nosocomial infections, with five newly synthesized inhibitors sharing an original dizinc binding group, 1,2,4-triazole-3-thione (TZT). We considered the energy-minimized structures of each of the five VIM-2 complexes obtained with the SIBFA potential. Energy decomposition analyses (EDA) at the HF level enabled to compare the QC and the SIBFA ΔE values and their contributions in the zinc cores, with and without TZT, totaling 30 complexes. With one exception, the ΔE(QC) values were reproduced with relative errors <1.5%. We next considered the complex of the entire inhibitors with an extended model of VIM-2 recognition site, totaling up to 280 atoms. ΔE(SIBFA) could closely reproduce ΔE(QC). EDA analyses were resumed on the complexes of each inhibitor arm with its interacting VIM-2 residues. As a last step, EDA results at correlated levels were analyzed for the mono- and dizinc sites enabling comparisons with dispersion-augmented ΔE(SIBFA) and correlated multipoles and polarizabilities. Closely reproducing ΔE(QC) and the contrasting trends of its individual contributions should enable for dependable free energy perturbation studies and comparisons to recent experimental ΔG values, limiting as much as possible the reliance on error compensations.
抑制剂与锌金属蛋白酶复合物的分子动力学是极化分子力学势应用的一个优势领域。这些势在全尺寸 MD 模拟整个蛋白质复合物时,对于重现两个单锌核和双锌核中的 QC 分子间相互作用能的准确度如何?我们考虑了扩展的锌依赖性金属β-内酰胺酶,VIM-2 的识别位点与五个新合成的抑制剂的复合物,这些抑制剂共享一个原始的双锌结合基团,1,2,4-三唑-3-硫酮(TZT)。我们考虑了 SIBFA 势下每个 VIM-2 复合物的能量最小化结构。HF 水平的能量分解分析(EDA)使我们能够比较 QC 和 SIBFA 的ΔE 值及其在锌核中的贡献,包括有和没有 TZT 的情况,共 30 个复合物。除了一个例外,ΔE(QC)值的相对误差<1.5%。接下来,我们考虑了整个抑制剂与 VIM-2 识别位点的扩展模型的复合物,总共达到 280 个原子。ΔE(SIBFA)可以很好地重现ΔE(QC)。我们对每个抑制剂臂与其相互作用的 VIM-2 残基的复合物重新进行了 EDA 分析。最后一步,我们在相关水平上分析了单锌和双锌位点的 EDA 结果,以便与色散增强的ΔE(SIBFA)和相关的多极子和极化率进行比较。准确重现ΔE(QC)及其各个贡献的对比趋势,应该能够进行可靠的自由能扰动研究,并与最近的实验ΔG 值进行比较,尽可能减少对误差补偿的依赖。