Detachment of HCO3– from the Active Site of Carbonic Anhydrase: Molecular Dynamics Simulation and Machine Learning

Gong Chen,Diannan Lu,Jianzhong Wu,Zheng Liu

DOI: https://doi.org/10.1021/acs.jpcc.8b05298

2018-01-01

Abstract:While tremendous efforts have been devoted, for many years both experimentally and theoretically, to explore the potential of carbonic anhydrase (CA) as an industrial catalyst for CO(2)sequestration, little is known about the kinetics of HCO3- detachment from the active site of the enzyme, an essential step affecting the performance of the CO2 hydration process. Using molecular dynamics simulation and the Markov-state model, we established a global kinetic profile for HCO3- detachment from CA. Using human carbonic anhydrase II (hCA-II) as the model enzyme, we calculated the commitors of all Markov states for HCO3- detachment through the analysis of transition paths and found that HCO3- prefers to diffuse toward the bulk solution once it dissociates with the binding sites. We also established the kinetic map for HCO3- detachment using molecular simulation data with the Markov-state model and identified six distinct pathways for HCO3- diffusion toward the bulk solution. These pathways were determined according to the electrostatic interaction gradient and the solvation gradient of HCO3- produced by amino-acid residues of hCA-II and the surrounding water molecules, respectively. To explore the relationship of HCO3- residence time at different binding sites as a function of steric hindrance, local charge density, and hydrophobicity, we proposed an artificial neural network (ANN) algorithm to correlate HCO3- occupancy probability, which reflects binding strength in different binding sites, with the molecular properties of hCA-II. The predicted HCO3- occupancy probability agrees perfectly with the simulation results, suggesting the great potential of an ANN for the re-engineering of carbonic anhydrase as well as other enzymes toward tailored applications.