Computational Modeling of Ion Transport in Bulk and through a Nanopore Using the Drude Polarizable Force Field

Abstract

In the past two decades, molecular dynamics simulations have become the method of choice for elucidating the transport mechanisms of ions through various membrane channels. Often, these simulations heavily rely on classical nonpolarizable force fields (FFs), which lack electronic polarizability in the treatment of the electrostatics. The recent advancements in the Drude polarizable FF lead to a complete set of parameters for water, ions, protein, and lipids, allowing for a more realistic modeling of membrane proteins. However, the quality of these Drude FFs remains untested for such systems. Here, we examine the quality of this FF set in two ways, i.e., (i) in simple ionic aqueous solution simulations and (ii) in more complex membrane channel simulations. First, the aqueous solutions of KCl, NaCl, MgCl2, and CaCl2 salts are simulated using the polarizable Drude and the nonpolarizable CHARMM36 FFs. The bulk conductivity has been estimated for both FF sets using applied-field simulations for several concentrations and temperatures in the case of all investigated salts and compared to experimental findings. An excellent improvement in the ability of the Drude FF to reproduce the experimental bulk conductivities for KCl, NaCl, and MgCl2 solutions can be observed but not in the case of CaCl2. Moreover, the outer membrane channel OmpC from the bacterium Escherichia coli has been employed to examine the ability of the polarizable and nonpolarizable FFs to reproduce ion transport-related quantities known from experiment. Unbiased and applied-field simulations have been performed in the presence of KCl using both FF sets. Unlike for the bulk systems of aqueous salt solutions, it has been found that the Drude FF is not accurate in modeling KCl transport properties across the OmpC porin.

Ion Permeation, Selectivity, and Electronic Polarization in Fluoride Channels

“…The experimental D is 1.475 × 10 -5 cm 2 /s in infinitely dilute aqueous solution (81), and (1.4 ± 0.1) × 10 -5 cm 2 /s considering finite concentration (82). Our results therefore indicate that the Drude FF predicts a correct D whereas C36m overestimates D by 50%, which is consistent with the study of ion conductivity by Prajapati et al (83). The permeation rate constant k was calculated from the PMF and D (Eqs.…”

“…Our results therefore indicate that the Drude FF predicts a correct D whereas C36m overestimates D by 50%, which is consistent with the study of ion conductivity by Prajapati et al . (83). The permeation rate constant k was calculated from the PMF and D ( Eqs.…”

Multiscale modeling of ion transport in porous electrodes

“…The CHARMM36‐jun2015 force field is used to describe the carbon and ion interactions 33,34 . A similar force field has been used in many reported studies to describe the nanofluidic system and ion transport 35 . Specifically, in this system, the graphene atoms are charged and their positions are fixed as part of the pore walls, thus only the nonbonded interaction between the graphene atoms and H 2 O, Na + , Cl − should be considered.…”

Benchmarking polarizable and non-polarizable force fields for Ca2+–peptides against a comprehensive QM dataset

“…The Drude-wRMSD model routinely shows unphysical distances <1.5 Å between the cation and 74,76 The unphysically short ion coordination distances observed in Drude-wRMSD lead to an "over-stabilization" phenomenon and presumably an over-binding on the protein surface, as suggested by studies of ion transport in ryanodine receptors 37 and porins. 39 Recent comparative analysis of MD simulations and capillary electrophoresis experiments for dications binding to insulin 17,19 indicates that specific and very tight binding of cations in the physiological pH range leads to over-accumulation of mobile charges on the protein surface modeled with non-polarizable FFs. 17 For example, up to 20 Ca 2+ were reported to bind stably to the full CaM structure, in stark contrast with the anticipated four ions bound to sites present in the EF hands.…”

“…While the parameters were shown to provide excellent performance in various reduced models of binding sites, 11,37 their extension to MD simulations of ion-protein interactions and transport in porin proteins elucidated remarkable issues leading to a hindered ion diffusion in the protein interior as well as apparent over-binding to the protein. 38,39 Recently, Villa et al showed, using the Drude FF, that it is possible to capture the complex interaction surface of Mg 2+ with methyl phosphate in the condensed phase, illustrating the feasibility of developing accurate and transferable polarizable potential functions for metal-ligand interactions. 40 However, success in the final deployment of next-generation polarizable FFs depends critically on assessing (i) the vast chemical space presented by the variety of side chains found in proteins and (ii) the strategies for explicitly including charge transfer terms in the case of strongly interacting cations.…”