Local Conformational Dynamics Regulating Transport Properties of a Cl⁻/H⁺ Antiporter

Abstract

ClC-ec1 is a Cl⁻/H⁺ antiporter that exchanges Cl⁻ and H⁺ ions across the membrane. Experiments have demonstrated that several mutations, including I109F, decrease the Cl⁻ and H⁺ transport rates by an order of magnitude. Using reactive molecular dynamics simulations of explicit proton transport across the central region in the I109F mutant, a two-dimensional free energy profile has been constructed that is consistent with the experimental transport rates. The importance of a phenylalanine gate formed by F109 and F357 and its influence on hydration connectivity through the central proton transport pathway is revealed. This work demonstrates how seemingly subtle changes in local conformational dynamics can dictate hydration changes and thus transport properties.

GRAPHICAL ABSTRACT

ClC-ec1 exchanges Cl⁻ and H⁺ ions across the membrane and regulates various physiological processes. The I109F mutation decreases the ion transport rate by an order of magnitude. In this article, two-dimensional free energy profile of H⁺ transport in the mutant has been constructed. Crucial conformational dynamics of the protein is revealed. The close state of a phenylalanine gate blocks the hydration of the central cavity and hence the H⁺ transport, and vice versa.

Introduction

The chloride channel (ClC) family of transmembrane proteins, composed of Cl⁻ channels and Cl⁻/H⁺ antiporters,^[1] has been discovered in nature across multiple species including many bacteria, some archaea, and almost all eukaryotes (including humans).^[2–4] They are responsible for various physiological processes ranging from epithelial salt homeostasis^[5] to pH regulation of intracellular compartments.^[6] Mutations in human ClC genes have proven to be related to pathologies of bone, muscle, kidney, and eye,^[1] emphasizing their importance in human physiology.

ClC-ec1 (Figure 1A) is a Cl⁻/H⁺ antiporter from Escherichia coli, and also the most structurally, functionally, and mechanistically investigated protein in the ClC family.^[7–12] Structurally, crystallographic data^[7–8] of the protein has revealed its homodimeric structure as well as three Cl^– binding sites (S_int, S_cen, and S_ext) per monomer. The central binding site (S_cen), with distinguishable Cl⁻ occupancy in the crystal structure,^[8] is located in the central cavity and surrounded by several conserved residues including S107, E148, I356, F357, A358, and Y445 (ClC-ec1 numbering). Inferred from the crystal structure, the S_cen anion is stabilized by the backbone atoms of I356 and F357 and the side chains S107 and Y445. The other anion sites are located either above or below S_cen along the membrane normal axis.

Figure 1.

(A) Equilibrated homodimeric structure (monomer A in yellow and monomer B in cyan) of ClC-ec1 I109F mutant with Cl⁻ (green sphere) bound to the S_cen site and some key residues shown in monomer A. Approximate ion transport pathways are indicated as red (H⁺) and green (Cl⁻) dashed arrows. The (B) open and (C) closed state of the phenylalanine gate formed by the mutated residue F109 and F357 in I109F mutant, with residues F109, E148, E203, F357 displayed and labeled. The excess proton and S_cen Cl⁻ are shown as yellow and cyan spheres, respectively.

Functionally, the wild-type (WT) ClC-ec1 exchanges Cl⁻ ions for H⁺ at a stoichiometry of (2.2±0.1):1 under normal conditions,^{[9, 12]} and therefore should be classified as a ClC antiporter. The normal proton transport rate is 1.0 × 10³ s^{– 1},^[12–13] but it can be slowed down or even eliminated by alternative polyatomic anions like NO₃^– and SCN^–.^[14] The anion/proton stoichiometry is also shifted to 7–10:1 or a much higher value by these anions.^[14] The two subunits of the homodimer have also been shown to work independently by experiments employing cross-linking ^[15] and by the monomeric form of ClC-ec1.^[16]

Mechanistically, the anion and proton transport (PT) pathways have been identified (see Figure 1A) based on crystal structures^[7–8] as well as numerous mutagenesis experiments.^{[13, 17]} These pathways intersect in the central cavity around the S_cen site, which is the most stable binding site confirmed by isothermal titration calorimetry (ITC) experiments^[18] and computational calculations of the free energy profile.^[19] Three anion binding sites, S_int, S_cen, and S_ext, were identified in the anion transport pathway, while E148 and E203 were recognized as important proton binding sites.^{[7–8, 13, 17]} Surprisingly, there are no crystallographic waters to date, nor other titratable residues, between these two residues,^[7–8] complicating our understanding of the PT mechanism. Such proton carriers are crucial for the Grotthuss shuttling of PT, since the excess proton diffuses via the dynamic formation and concomitant cleavage of covalent O–H bonds among proton-carrying molecules, unlike other hydrated ions.

Based on several molecular dynamics (MD) simulations of the WT protein, it was proposed that a transient water network could be formed to facilitate PT.^[20–23] Our reactive MD simulations further revealed an important dynamical coupling between PT and an increased hydration between E203 and E148.^{[20, 23]} These simulations also demonstrated how Cl⁻ at S_cen is necessary for PT through the central cavity in one direction, and influential in proton release from E148 in the other.^[23–25] Our recent work further demonstrated how multiple competing pathways with different intermediates and rate limiting steps cumulatively contribute to the total anion/proton flux.^[25] This notion, that multiple sequences of transitions between microstates contribute to the total flux,^[25] is quite different from the commonly proposed single-pathway mechanisms,^{[12, 26]} and explains for the first time the origin of the 2.2:1 stoichiometry.

Despite being extensively studied, several aspects of ClC-ec1 remain unanswered or under debate. For example, it is still unclear how the ClC family of proteins are partitioned into channels and antiporters despite sharing very similar structures.^[27] Similar to other channels, but unlike other active secondary transporters such as LacY and NapA,^[28] ClC-ec1 does not seem to use large-scale conformational changes to enable antiporting.^[29] However, it is thought that conformational dynamics of residues along the transport pathway, including I109,^[22] E148,^{[23, 30]} and F357,^[30] are important for transport properties. The side chain rotation of I109 was proposed to be important for hydration in the central cavity.^[22] Similarly, the concerted movement of F357 and E148 side chains was suggested to be an essential component of the transport cycle based on MD simulations of WT ClC-ec1, combined with uptake assay and ITC measurements on the F357A mutant.^[30]

Several experimental^{[12, 31–34]} and computational^[34] studies have also suggested that conformational changes in helical structures further away from the transport pathways could contribute to the ion transport cycle. However, no crystal structure of ClC-ec1 or other homologs has shown such global conformational changes, and the magnitude as well as the impact of these structural rearrangements remains unclear.

In the absence of large conformational changes, it is curious how the functionality of ClC proteins are altered and regulated. Experiments have demonstrated that alternative anions modulate the transport properties of ClC-ec1.^[14] We were able to explain this effect of alternative anions in S_cen by their ability to disrupt the water network in the central cavity and emphasized the importance of a complete water network for PT.^[35]

Several ClC mutants including the I109F mutant have also been shown to disrupt ion transport.^[22] For the I109F mutant, the PT rate is significantly reduced by an order of magnitude to ~1.0 × 10² s⁻¹, and the Cl⁻ rate is decreased from ~2.3 × 10³ s⁻¹ to ~2.6 × 10² s⁻¹. In non-reactive MD simulations of the WT protein, side-chain rotation of I109 was observed, and the residue was identified as a gating residue lying within the central cavity.^[22] However, the simulation did not directly simulate PT, and the I109 rotation was not characterized statistically or energetically. Interestingly, PT from E203 to E148 is not rate-limiting in one orientation based on a more recent reactive MD study,^[23] which contradicts the proposed gating effect of I109 between E203 and E148. Nevertheless, the rate-limiting step is not necessarily transferrable to the I109F mutant, and no simulation of PT in the I109F mutant has been performed yet. Thus, further computational investigations of the I109F mutant are necessary to understand the mechanism by which this mutation regulates function.

This work focuses on the mechanistic details of how a mutation influences PT properties. Through comprehensive simulations with the correct description of delocalization of an excess charge and Grotthuss shuttling, we have determined the free energy profile for PT through the central cavity in I109F mutant. We find that an important phenylalanine gate is formed between F109 and F357 that disrupts the hydration network in the central cavity and thus slows down PT by an order of magnitude. We further demonstrate how water connectivity is crucial for PT, and propose a new approach for regulating hydration to alter transport properties. This alternative approach of controlling hydration through local conformational dynamics is expected to influence the PT mechanism of other similar proteins, such as cytochrome c oxidase (CcO).^[36]

Methods

System Setup

The simulation system consisted of the ClC-ec1 homodimer (PDB ID: 1OTS),^[8] 163 POPE lipids, 17 Cl^– anions, and ~11,000 TIP3P water molecules in a 92 Å × 92 Å × 79 Å box with periodic boundary conditions. The equilibrated wild-type structure^[23] was mutated to I109F, followed by energy minimization. It was then equilibrated with the GROMACS 5.1 software^[37] for ~0.5 μs. After the equilibration, the proton on the E203 residue in monomer A was treated as the excess proton. All following discussions and simulations were done for monomer A, since two subunits have been shown to work independently.^[15–16]

To simulate PT explicitly, multiscale reactive molecular dynamics (MS-RMD)^[38] was employed. To correctly describe charge delocalization and Grotthuss shuttling of proton transport, MS-RMD performs an on-the-fly diagonalization of a Hamiltonian matrix and calculates the most probable mixture of states, each with a different bonding topology.

Simulations were performed with the RAPTOR package^[38] embedded in the LAMMPS software,^[39] with umbrella potentials implemented in the PLUMED 2 package.^[40] Additional simulation details, and the parametrization procedure of MS-RMD models are explained in the Supplementary Material.

Definition of Collective Variables

To find the most relevant collective variable (CV) coupled to the water dynamics, test simulations were performed. They used the same settings as production simulations, except that water density CV^[41] was adopted as an initial guess of the second CV, and that the MS-RMD parameters for E148 and E203 were not reparametrized (i.e., these parameters were retrieved from our previous publication,^[42] see more details in Supplementary Material). The absolute value of covariance between the candidate CV and the water-wire gap ratio (a quantity reflecting the quality of water network, described in the next section) was calculated to determine the CV most correlated with hydration level.

In production simulations, two-dimensional replica-exchange umbrella sampling (REUS)^[43] was employed to calculate potentials of mean force (PMFs) with an accelerated convergence rate. The first CV, ξ ₁, was the ratio-based center of excess charge (CEC) position defined in our previous work.^[35] This ratio-based CV describes the distance between donor and acceptor and the progress of proton transport simultaneously, but does not introduce artificial bias.^[35] The second CV, ξ ₂, was set to be the tip distance between the outmost carbon atoms of the phenyl rings in the F109 and F357 side chains, as it has the largest absolute covariance with the gap ratio, and was shown to be correlated to the hydration level in the central cavity (see Figure 2, S1). Other details of generating PMFs are described in the Supplementary Material for the sake of simplicity.

Figure 2.

Scatter-point plot of (A) gap ratio or (B) KSP length vs tip distance of Phe gate calculated from REUS simulations, and (C, D) the corresponding density plots. Data points were clustered into (A) 3 or (B) 2 components according to Gaussian mixture model,^[45] with each cluster colored red, green, and gray, respectively. Black lines partition the plot into four quadrants. These lines were placed along the dividing line between red and green clusters, such that minimal points are in quadrants I and III. The density plot was normalized such that the integral over the area is unity.

Characterization of Water Connectivity

The water-wire gap ratio and K-shortest path (KSP) length were used to quantify water connectivity. They were plotted against the tip distance CV (ξ₂) to identify correlation. Neither of these measures is continuously differentiable, so they cannot be used as a CV for REUS.

The water-wire gap ratio was calculated following our previous procedure.^[35] The breadth-first algorithm was used to search for any possible water network connected through hydrogen bonds, which were defined by distance criterion and angle criterion: (1) the distance between the donor and acceptor heavy atoms is shorter than 3.5 Å; and (2) the angle formed by the donor atom, the central hydrogen, and the acceptor atom is larger than 150 degrees. With the maximum depth set to 20, the searching algorithm was performed twice, once from E203 to E148 and once in the opposite direction, so as to search for all paths between the donor and the acceptor. The gap ratio was the gap distance divided by channel length, where “gap distance” was defined as the smallest distance between two disconnected groups of heavy atoms in the water networks starting from either E203 or E148, and “channel length” was given by the smallest distance between two groups of side-chain oxygen atoms of either E203 or E148.

To calculate the K-shortest path length (K = 1), the “distance matrix” of water molecules in the transport channel as well as oxygen atoms of side chains of E148 and E203 was calculated first. The 𝑖𝑗-th element 𝑑_ij was given by

$$d_{ij}=1+{\left(r_{ij}/r_0\right)}^6,$$ (1)

where r_ij represents the distance between the i-th atom and the j-th atom, and 𝑟₀ is a scaling factor and set as 3.0 Å. On the undirected graph given by the distance matrix, the length of the shortest path starting from carboxyl oxygen atoms of E203 to those of E148 was calculated using Yen’s algorithm.^[44] The natural logarithm of the resulting quantity was taken for a smaller range and better description of water connectivity. The calculation was implemented in a revised version of PLUMED 2^[40] and Boost Graph Library.

Clustering of data points in Figure 2 and S1 were done using Gaussian mixture model,^[45] calculated with scikit-learn package^[46] implemented in Python 3. Other clustering algorithms (k-means++,^[47] spectral clustering, etc.) gave similar results. The black lines separating four quadrants were placed along the dividing line between two clusters, such that minimal points are in quadrants I and III. The density plot was normalized such that the integral over the area is unity.

Results and Discussion

Tip Distance of the Phenylalanine Gate Representing Water Connectivity

To study PT in the I109F mutant of ClC-ec1, we employed REUS for enhanced sampling. As with all rare event processes, finding the CVs that define the slowest motions limiting the process of interest is of paramount importance. As shown in our previous work,^[35] the water density CV,^[41] which was used to simulate PT for ClC-ec1 with alternative anions in S_cen, was not perfect in all situations. At high water densities, it favored water molecules clustering below the S_cen anion instead of a continuous water-wire along the PT pathway. As a result, the minimum free energy path was uncoupled (i.e., it had a Pi-shape, with two ~90-degree kinks) for the nitrate system, meaning the PT process did not show a clear dependence on the water density CV, especially through the transition barrier region. Although we recovered their correlation through post-processing, we want to capture this directly through our enhanced sampling simulations. We developed several quantities to characterize water connectivity directly, but they are not continuously differentiable and cannot be used as a CV in REUS. Therefore, we characterized alternative degrees of freedom (DOFs) in the system to find the most relevant CV coupled to the water dynamics and to get an idea of factors influencing water structure.

From the trajectories of the test simulations, we observed some conformational changes of F109 and F357 (Figure 1B, 1C). Given that these two residues were proposed to affect transport properties,^{[22, 30]} we supposed that these conformational changes could couple with the amount of water around the S_cen anion. We proposed three candidate CVs: (1) the minimum distance between the carbon atoms in two Phe residues, (2) the distance between the center of the phenyl carbons, (3) the tip distance between the outmost carbon atoms of two Phe residues. After analyzing the test simulations, we discovered that the tip distance has the largest absolute covariance with the water network gap ratio. We believe that the tip distance CV is flexible enough to allow rotation along the principal axis of the phenyl group without being changed, but it is not overly flexible in that it forbids free rotation along an arbitrary axis.

The gating effect of the phenylalanine gate formed by F109 and F357 (Figure 1B, 1C) was further quantified by the clustering in the correlation plot between this tip distance and two other quantities reflecting water connectivity: the water network gap ratio and KSP length (Figure 2, S1). Although we used test simulations to identify the correlation, we further confirmed this by running additional analyses on REUS simulations. The latter (see Figure 2A) is consistent with results from test simulations (Figure S1). Therefore, the discussion below is based on newer analyses. The data in Figure 2A clusters in either the upper-left region, where the tip distance is larger than 7 Å and the gap-ratio lower than ~0.4, or the lower-right region, where the tip distance is shorter and the gap-ratio is larger. Thus, a well-connected water network (gap-ratio less than ~0.3) is strongly correlated with an open Phe gate. Note that the line at the zero-gap region represents a completely connected water network, based on the hydrogen bond definition with distance criterion and angle criterion. These completely connected states are dominated by the larger tip distances (gap ratio above 6 Å) and do not exist for tip distances below 5 Å. As expected, there is an abrupt jump in values from 0 to ~0.2, because our hydrogen bond definition is discontinuous (i.e., 0/1-based). A small shift in the position of a water oxygen/hydrogen atom in a continuous water chain might disconnect its hydrogen bond to a neighboring water molecule along the chain if this bond is just at the definition boundary based on the discontinuous criteria. Such an abrupt change will suddenly increase the gap distance from 0 to the new O–O distance, which is ~3 Å if the angle criterion is broken and ~3.5 Å if the distance criterion is broken. Given the approximate Glu–Glu distance at ~15 Å, it is reasonable to have a gap in the gap ratio between 0 and ~0.2.

In the correlation plot between tip distance and KSP length (Figure 2B), for which a longer length means worse water connectivity, the data shows a similar pattern, with dividing lines at ~7 Å tip distance and at 4 KSP length. Again, the formation of high-quality water networks (KSP length < 3) requires an open Phe gate (tip distance > ~6 Å), and the best-quality water networks (KSP length < 1.8) only accompany a fully open Phe gate (tip distance > ~8 Å).

These plots (Figure 2) demonstrate that the conformational change of F109 and F357 is important to water structure. Despite a few exceptions, the Phe–Phe tip distance is correlated with the quality of water network within the transport channel. Based on these results, we decided to directly sample the tip distance to control the hydration of the PT pathway. Unlike the gap ratio and KSP length, the tip distance has the advantage of being continuously differentiable in sampling regions. Thus, the tip distance was used in REUS for enhanced sampling of the channel hydration.

Phenylalanine Gate Regulating Proton Transport

With two CVs defined, we employed REUS to sample PT between E203 and E148. Since both orientations of the protein in the membrane are possible in the in vitro experiments,^[48] we consider the rate-limiting step to be either from E203 to E148, or vice-versa, depending on which has a lower barrier. The orientation with the higher barrier should have lower transport rate and thus contribute less to the experimentally measured flux. The 2D PMF (Figure 3A) shows a 14.3 ± 0.6 kcal mol⁻¹ free energy barrier for PT across the transport channel (E148 to E203). The increased transport barrier compared to WT (by ~6 kcal mol⁻¹)^[23] shifts the rate-limiting step to the transport within the central region. Based on transition state theory (TST), the rate constant for this PT step is estimated to be (1.6 ± 1.1) × 10² s⁻¹, which is consistent with the experimental value of ~1.0 × 10² s⁻¹.^[22] Note that the protonated E203 state is ~1.5 kcal mol^–1 more stabilized than the protonated E148 state in the I109F mutant of ClC-ec1. In contrast, our previous studies^[23] showed energetically downhill PT from E203 to E148 in WT ClC-ec1, using the same MS-RMD methods and sampling protocols. This should be attributed to the decrease in pK_a of E148, as was suggested by its correlation with decreased Cl^– transport rate.^{[25, 49]}

Figure 3.

(A) 2D PMF of I109F ClC-ec1, with Cl⁻_cen present in S_cen. Irrelevant high-energy areas (red) are not sampled to reduce the cost of computation. The black line traces the minimum free energy path for PT from E203 to E148. (B) The extracted 1D PMF along the minimum free energy path. The error of the PMF is ~0.6 kcal mol⁻¹, estimated from splitting the trajectories and block-averaging. Five key points along the path are labeled and described in the Results and Discussion section.

From the 2D PMF, the minimum free energy path and extracted 1D free energy profile (Figure 3B) show an inverse U-shaped path compared to the Pi-shape in our previous work on alternative anions bound to ClC-ec1,^[35] suggesting a possibly similar stepwise mechanism of PT but more coupling between the Phe gate (and thus water connectivity) and PT. Thus, the similarity of paths is consistent with the above discussion, in that the conformation of the Phe gate indeed influences the water dynamics within the channel and thereby PT. With mostly hydrophobic residues in the transport cavity, the system starts in a dehydrated state with closed phenylalanine gate and an excess proton loaded at E203 (Figure 1C, S2A). After the phenylalanine gate is opened (shown as increased tip distance in the 2D PMF), the hydration of the central cavity increases (Figure S2B). This is consistent with the 6–7 Å tip distance region in Figure 2. With a connected water-wire, the excess proton is transported to the center (Figure S2C) and through the phenylalanine gate (Figure 1B, S2D). The gate then closes due to interactions between the two Phe residues and the channel recovers its original dry state attributable to hydrophobic effect (Figure S2E). The increased transport barrier in the I109F mutant compared to WT^[23] indicates that the interaction between F109 and F357 is sufficiently stabilized, compared to the interaction between I109 and F357 in the WT protein, to hinder the formation of a connected water-wire, but not strong enough to completely block water penetration and PT.

The observed mechanism of PT re-emphasizes the importance of a connected water network around the excess charge, which is also the reason why alternative anions slow or block PT in WT protein. It also demonstrates that in addition to the chemical nature of alternative anions,^[35] the conformational dynamics of the phenylalanine gate, formed by F109 (the mutated residue) and F357, can also regulate water connectivity and hence PT through the central cavity. A similar gating mechanism was reported in CcO,^[36] where the side-chain rotation of N139 controls the asparagine gate formed by N121 and N139, and thus regulates the hydration level of D-channel. Noticeably, these local conformational changes are coupled with the proton movement in the protein. In the ClC mutant, the closed state of the Phe gate is ~2.5 kcal mol⁻¹ more stable than the open state when proton is bound to E203, but the open state becomes dominant when proton moves to the center (region C in Figure 3A). A similar coupling effect between PT release from D132 and conformational changes of the asparagine gate in WT CcO were also reported.^[36]

Conclusions

Multiscale reactive MD simulations with enhanced free energy sampling were performed to investigate the mechanism of I109F mutation slowing down PT in ClC-ec1. With both the mutation and PT treated explicitly, we constructed a 2D PMF for the PT process across the central region past the central Cl^– anion binding site (S_cen). Based on transition state theory, the rate constant estimated from the PMF is consistent with the experimental measurements,^[22] where the PT rate is decreased by an order of magnitude to ~1.0 × 10² s⁻¹. Our result is also consistent with the decreased Cl⁻ transport rate, since protonated E148, which is essential for Cl^– transport, is destabilized in our simulations.

By comparing our results with previous work on the nitrate/thiocyanate-bound ClC-ec1,^[35] we herein confirm a similar hydration–PT–dehydration stepwise mechanism, but in this case the controlling factor is the conformation of the phenylalanine gate formed by F109 and F357. The correlation plot shows the relationship between the tip distance of the gate and the water connectivity. With these results combined, we conclude that it is the formation and stabilization of the Phe gate that disrupts water connection and hence the proton transport across the central region. This is an alternative approach to regulate PT, compared to the previously reported regulation by the chemical nature of the transported anion. Thus, this work demonstrates how subtle local conformational dynamics can significantly influence the ease of forming hydration networks and thus the rates of charge transport. Such subtle conformational dynamics could be an important factor distinguishing a ClC antiporter from a ClC channel, and are certainly influential regulators of ClC functionality.

Despite quantitively matching experimental results of PT rate, our study does not fully explain the effect of I109 mutation on Cl⁻ transport. Our future research will include a more complete investigation of the whole Cl⁻/H⁺ transport cycle of the mutant using multiscale kinetic modeling (MKM).^[25]