Water access points and hydration pathways in CLC H⁺/Cl⁻ transporters

Significance

CLC transporters are biologically essential proteins that catalyze the transmembrane exchange of chloride for protons. The permeation pathway for chloride through the transporters has been well characterized. Here, we study the more elusive permeation pathway for protons. Through computational modeling, we show that water molecules can permeate deep inside the protein and form continuous wires. To test the hypothesis that these water wires mediate proton transport, we mutated residues predicted to impede water wire formation and experimentally evaluated the effects of the mutations. The results from our concerted computational and experimental approach strongly support the role of water in proton transport by CLCs and provide a valuable framework for investigating their overall transport mechanism.

Abstract

CLC transporters catalyze transmembrane exchange of chloride for protons. Although a putative pathway for Cl⁻ has been established, the pathway of H⁺ translocation remains obscure. Through a highly concerted computational and experimental approach, we characterize microscopic details essential to understanding H⁺-translocation. An extended (0.4 µs) equilibrium molecular dynamics simulation of membrane-embedded, dimeric ClC-ec1, a CLC from Escherichia coli, reveals transient but frequent hydration of the central hydrophobic region by water molecules from the intracellular bulk phase via the interface between the two subunits. We characterize a portal region lined by E202, E203, and A404 as the main gateway for hydration. Supporting this mechanism, site-specific mutagenesis experiments show that ClC-ec1 ion transport rates decrease as the size of the portal residue at position 404 is increased. Beyond the portal, water wires form spontaneously and repeatedly to span the 15-Å hydrophobic region between the two known H⁺ transport sites [E148 (Glu_ex) and E203 (Glu_in)]. Our finding that the formation of these water wires requires the presence of Cl⁻ explains the previously mystifying fact that Cl⁻ occupancy correlates with the ability to transport protons. To further validate the idea that these water wires are central to the H⁺ transport mechanism, we identified I109 as the residue that exhibits the greatest conformational coupling to water wire formation and experimentally tested the effects of mutating this residue. The results, by providing a detailed microscopic view of the dynamics of water wire formation and confirming the involvement of specific protein residues, offer a mechanism for the coupled transport of H⁺ and Cl⁻ ions in CLC transporters.

The chloride channel (CLC) family (1, 2) includes both passive Cl⁻ channels and secondary active H⁺-coupled Cl⁻ transporters (3–8). The latter, also known as H⁺/Cl⁻ exchangers, drive uphill movement of H⁺ by coupling the process to downhill movement of Cl⁻ or vice versa, thereby exchanging the two types of ions across the membrane at fixed stoichiometry (9). ClC-ec1, a CLC from Escherichia coli, has served as the prototype CLC for biophysical studies because of its known crystal structures (10, 11), its tractable biochemical behavior, and its structural and mechanistic similarities to mammalian CLC transporters (3–8, 12–17). Detailed structural and functional studies of ClC-ec1 (9, 11, 18–27) have shed light on some of its key mechanistic aspects. Most prominently, these studies have characterized the Cl⁻ permeation pathway and its lining residues (10, 18, 25) and established the role of E148, also known as Glu_ex, as the extracellular gate for the Cl⁻ pathway (9, 11).

Although much less is known about the H⁺ translocation pathway (and mechanism), experimental studies have provided key information on the involvement of specific residues in H⁺ transport (9, 13, 14, 20, 22, 27, 28). Extensive site-directed mutagenesis studies have zeroed in on two glutamate residues essential for H⁺ transport (Fig. 1A): E148 (Glu_ex), which acts as the main extracellular H⁺ binding site (9, 11, 27), and E203 (Glu_in), which plays a similar role on the cytoplasmic side (20, 22, 28). Neutralization of either glutamate eliminates H⁺ translocation by ClC-ec1 (9, 28). However, the discovery of these H⁺ binding sites also raised a mechanistic puzzle (3, 23): How do protons translocate between the two sites, which are separated by a ∼15-Å-long, largely hydrophobic region within the lumen of the protein?

Fig. 1.

Cl⁻ and H⁺ permeation pathways in ClC-ec1. (A) View of the ClC-ec1 structure in a lipid bilayer (the simulation system used here), with the identical subunits shown in yellow and orange. The presumed Cl⁻/H⁺ permeation pathways are indicated by green and red lines, respectively. The dashed segment of the red line denotes the pathway investigated in this study. (B) Close-up of the central hydrophobic region, with the residues forming this region shown as orange sticks and labeled. Also shown are key glutamate residues (E202, E203, and E148) as well as the Cl⁻ at the central anion binding site. (C) Hydration of the central hydrophobic region during the 0.4-µs equilibrium simulation, measured as the number of water molecules in this region for each subunit.

Since the report of its first crystal structure, a large number of computational studies have aimed at investigating various molecular details related to the CLC H⁺ transport mechanism (27, 29–34). One model emerging from these studies proposes that water molecules may connect the two H⁺ sites (Glu_ex and Glu_in) and, thereby, facilitate H⁺ transport (29, 30, 34). This idea was initially proposed by Kuang and coworkers (29) on the basis of a hole-searching algorithm applied to static crystal structures of ClC-ec1. In their proposed pathway, water molecules are suggested to form two half-wires that are then connected by the hydroxyl group of Y445 to form a complete path for H⁺ transfer. However, it is known from experiments on the Y445F mutant that this hydroxyl is not required for H⁺ transport (20). Wang and Voth (30) proposed another pathway by combining an improved search algorithm for buried water with short molecular dynamics (MD) simulations, thereby taking into account the dynamic nature of the protein. Their pathway did not rely on Y445 but required reorientation of the side chain of E203 to connect the two H⁺ sites. In another study, these investigators further carried out semiempirical free energy calculations to investigate the Cl⁻/H⁺ coupling mechanism (33).

Although the idea of water-mediated H⁺ transport is intriguing and could be key to understanding H⁺ transport in ClC-ec1, several questions relevant to a water wire mechanism remain unanswered: Can the hydrophobic region between the two H⁺ sites actually be hydrated under equilibrium conditions? What is the access/entry point or points for water from the bulk into the hydrophobic region, which is buried inside the protein, approximately at the midpoint of the membrane? Is it possible to observe the spontaneous formation of water wires through MD simulations? If so, how much do the simulated wire structures differ from the ones proposed by the prior studies based on search algorithms? How could the protein affect the dynamics and/or the thermodynamics of water wires?

In the current study, we have addressed these questions through a combined computational and experimental approach. An extended 0.4-µs MD simulation of a membrane-embedded model of wild-type (WT) ClC-ec1 reveals that the central hydrophobic region can indeed be hydrated by water molecules mainly from the cytoplasmic bulk phase through pathways near the dimer interface via a portal lined by residues E202, E203, and A404. Water wires connecting the two H⁺ sites form spontaneously and repeatedly during the equilibrium simulation. Formation of wires requires a side-chain conformational change of I109 and the occupancy of the central Cl⁻ binding site, S_cen. These simulation results make two strong and testable predictions: that mutations at A404 and I109 will reduce ClC-ec1 activity and that the reduction in activity occurs via effects on the H⁺ branch of the transport mechanism. Our experimental tests and additional simulations performed on one mutant form of the protein fully support these predictions.

Results

Spontaneous Hydration of the Central Hydrophobic Region.

We monitored the number of water molecules within the central hydrophobic region geometrically defined as a polyhedron (SI Appendix, Fig. S1), caged by the centroids of 13 residues surrounding this region (Fig. 1B), selected as described in SI Appendix. Fig. 1C shows the number of water molecules within the central region during an extended 0.4-µs MD simulation of WT ClC-ec1. The simulation started with only one water molecule present in the central region, placed between the Cl⁻ at S_cen (Cl⁻_cen) and Glu_ex, as suggested in previous computational studies (SI Appendix) (30). After the first 50 ns of the simulation, water molecules begin to enter the central region from the bulk. Our water count analysis (Fig. 1C) reveals that the central region can accommodate up to six water molecules and that it experiences frequent emptying and refilling events. Similar water fluctuations were observed in a model system that involves water-confining nonpolar pores, indicating that such systems can undergo thermal fluctuation between filled and empty states when the transfer of free energy of water from bulk phase to the pores approaches zero (35).

Formation of Connected Water Wires.

The presence of water molecules in the central region is necessary but not sufficient for H⁺ transport; formation of connected water wires must also occur. Thus, we monitored water wires by looking for the presence of any continuous hydrogen-bonded chain between Glu_ex and Glu_in during the simulation (SI Appendix). The analysis shows that although the central region is often occupied with water molecules (Fig. 1C), continuous water wires in this region can only form transiently (Fig. 2A). The wires formed repeatedly in both ClC-ec1 subunits during the 0.4-µs simulation. The probability of finding a water wire averaged over the two subunits, and excluding the first 50 ns of the simulation (initial hydration phase), was ∼1.5%, suggesting that even though thermodynamically not highly favorable, water wires do have a nonnegligible chance to arise. The lifetime of the water wires was 0.6 ± 0.2 ns (SI Appendix, Fig. S2).

Fig. 2.

Water wires in the central hydrophobic region of ClC-ec1. (A) Water wire formation between Glu_in and Glu_ex during the simulation. Formation events occurring in each subunit are shown in black and red, respectively. The segment marked by the blue bar was used for comparison with the A404L simulation. (B, Upper) Top 10 sites (purple spheres) most frequently occupied by water molecules forming the water wires. The sites are ranked in reverse order according to their distance to Glu_in. (Lower) representative structures of the two most common types of water wires observed in the simulation. The probability of each type is indicated. (Left) Five of the water molecules close to Glu_ex are in the hydrophobic region. (Right) Four of the water molecules close to Glu_ex are in the hydrophobic region. (C) Average interaction energies between water on individual water sites, as indicated in B, and Glu_in, Glu_ex, and Cl_cen⁻. The contributions of other individual amino acids are only between −1 and 1 kcal/mol. (D) Average probability of the water wires obtained from short (1 ns) simulations of ClC-ec1 in the presence (red) or absence (black) of Cl⁻ at S_cen, starting from the most representative wire structure. For each case, 16 simulations were performed with different starting atomic velocities. In the presence of Cl⁻, the drop in the probability of water wires is fit to p_wire(t) = e^−t/τ (dashed line), yielding τ = 1ns with a fitting error (χ²) of 0.02.

We characterized the configuration of the water wires by first identifying distinct spatial sites between Glu_ex and Glu_in with a high probability of water occupancy (Fig. 2B) and then clustering wire structures according to the pattern of the occupancy of these sites (SI Appendix). The water wires are structurally more variable around the Glu_in end but share a similar structure in the vicinity of Glu_ex (SI Appendix, Fig. S3A). At a coarser level (Fig. 2B), the majority of the water wires (∼75%) are long, exhibiting a kink in the middle to avoid steric clashes with Y445, whereas the second most populated configuration is composed of a straighter, shorter wire, the occurrence of which coincides with a slight rotation of the Y445 side chain (∆χ₁ ∼−10° to −20°).

Stabilization of Water Wires by Central Cl⁻.

To examine the factors that facilitate the formation of the water wires in the central hydrophobic region, we calculated interaction energies between the water wires and various elements in their environment, including both every residue of the protein and Cl⁻_cen. Glu_ex, Glu_in, and Cl⁻_cen are the strongest stabilizing elements for the water wires among all of the elements investigated. Stabilization of the water molecules close to Glu_ex mostly originates from Cl⁻_cen (∼4−12 kcal/mol; Fig. 2C). Moreover, calculation of electrostatic potential reveals that the Cl⁻_cen ion creates a negative electrostatic potential in the central hydrophobic region (SI Appendix, Fig. S3B). Under the electrostatic influence of Cl⁻_cen, the water molecules occupying the middle part of the water wires usually align their dipoles toward Cl⁻_cen (SI Appendix, Fig. S3 C and D). Thus, Cl⁻_cen could play an important electrostatic role not only in promoting the water wires but also in determining their structures. The effect of Cl⁻_cen on the stability of the water wires was also evaluated in a 200-ns simulation performed in the absence of Cl⁻_cen. During this simulation, no water wire formed in one of the subunits, and the chance of observing a water wire in the other subunit was found to be only ∼0.2%, a probability far lower than the value obtained in the presence of Cl⁻_cen (∼1.5%). The effect of Cl⁻_cen on the lifetime of the water wires was further assessed in two additional sets of MD simulations (16 independent simulations in each set), both initiated from the most representative water wire structure, either in the presence of Cl⁻_cen or with a vacant S_cen (SI Appendix). In the presence of Cl⁻_cen, the average lifetime of the water wires is ∼1 ns, consistent with the value (0.6 ± 0.2 ns) determined in the extended equilibrium simulation. In contrast, on removal of Cl⁻, the wires become highly unstable, consistently breaking within the first 0.2 ns of the simulations (Fig. 2D). These results indicate that the occupancy of S_cen by Cl⁻ likely plays a critical electrostatic role in the formation and stabilization of water wires.

Accessibility of Water to the Central Hydrophobic Region.

To investigate how water molecules reach the central region from the bulk, all trajectories for individual water molecules visiting this buried region at any time during the simulation were collected (SI Appendix). Each trajectory records amino acids that have been visited by a water molecule approaching the central region. A total of 87 water trajectories were identified in this manner. In all cases, water reaches the central region from the cytoplasmic bulk. Although a significant majority of water molecules (82 of 87) penetrate the protein via a pathway at the dimer interface (SI Appendix, Fig. S4A), in five cases, water molecules followed the pathway primarily associated with Cl⁻ permeation (10, 11, 17). Through a similar analysis, we also find that the majority (n = 78) of water molecules leaving the central region use the pathway located at the dimer interface, and only a small fraction (n = 6) leave via the Cl⁻ pathway. The low probability found for water to enter or leave through the Cl⁻ pathway in principle could be caused by the presence of a Cl⁻ ion in an intracellular Cl⁻ binding site, known as S_int (10), which appears to block the access of cytoplasmic water to the central region. However, we can safely rule out this possibility, as S_int exhibits an ion occupancy of only ∼10% and frequently exchanges its Cl⁻ ion with cytoplasmic solution.

The frequency at which an amino acid is visited by entering water molecules can be used to estimate the involvement of the amino acid in the water pathway. Our analysis, as detailed in the SI Appendix, identified A404, E202, and Glu_in (Fig. 3) as the three most visited amino acids in all 87 water trajectories, with probabilities of 77%, 65%, and 53%, respectively. Structural examination of the protein reveals that these three residues form a portal that separates the central region from cytoplasmic solution (Fig. 3). Most notably, the most visited amino acid by water is A404, a residue with a small side chain. One can therefore hypothesize that a bulkier substitution at position 404 would interfere with efficient water permeation into the central hydrophobic region, and thereby retard H⁺ transport.

Fig. 3.

Entryways of water into the central hydrophobic region. Stereo view of close-up of the dimer interface on the cytoplasmic side. Side chains are shown if they have a probability > 0.2 of being visited by water molecules entering the central region, with the color indicating their probability of being visited (scale bar on the right).

Experimental Verification of the Portal Hypothesis.

To test our hypothesis that the portal is a bottleneck for water penetration into the central region, we introduced bulkier side chains into the portal via a series of mutants at position 404 and tested their transport activity. H⁺ transport can be quantified by reconstituting ClC-ec1 into proteoliposomes and monitoring changes in extravesicular [H⁺] (36). Transport is initiated by removing extravesicular Cl⁻ (replacing with the impermeant isethionate), thus generating a gradient for Cl⁻ to move out of the liposomes, driving H⁺ in through ClC-ec1’s 2:1 Cl⁻/H⁺ exchange mechanism. Because this exchange transports three net charges across the membrane per cycle, buildup of an electrical gradient halts any measurable H⁺ transport until the addition of valinomycin dissipates the electrical gradient by shuttling K⁺ ions. As shown in Fig. 4 A and C, the rate of H⁺ transport through ClC-ec1 decreases in parallel to the volume of the side chain introduced at position 404. Mutation to the smaller side chain glycine has no measurable effect, whereas mutation to valine, leucine, or tryptophan progressively cripples H⁺ transport. Because H⁺ transport is tightly coupled to Cl⁻ transport, the rates of Cl⁻ transport decrease roughly in parallel to those of the H⁺ (Fig. 4 B and C). For the two largest side chains, H⁺ transport is slowed so drastically that Cl⁻ has the opportunity to slip through occasionally in an uncoupled manner, as can be seen in the comparison of the ratio of these rates (SI Appendix, Fig. S5A).

Fig. 4.

Experimental and computational tests of portal hypothesis. (A) H⁺ transport catalyzed by ClC-ec1 containing different substitutions at residue 404. H⁺ uptake was initiated at t = 10 s (indicated with arrow) by the addition of valinomycin. (B) Cl⁻ transport catalyzed by WT and the 404 mutants. The arrow indicates the point of valinomycin addition. (C) Summary of Cl⁻ and H⁺ unitary turnover rates determined for each of the mutants. Asterisk denotes the WT residue. (D) Effects of A404L mutation in WT background (Left) versus the uncoupling E148A background (Right). Cl⁻ turnover rates were normalized to their respective background. (E) Comparison between portal structures in WT and A404L. The snapshot shown for WT portal is the one at t = 178 ns in the WT simulation, which was also used to construct the mutant simulation. The last frame of the A404L simulation is shown to represent the mutant portal.

If, as predicted, substitutions at position 404 inhibit transport predominantly by thwarting formation of the water wires, then effects should not be felt at the Cl⁻ pathway. To test this prediction, we examined the effect of A404L mutation specifically on Cl⁻ transport (36). Whereas the A404L mutation decreases Cl⁻ transport through WT ClC-ec1 13-fold, it has only a relatively minor (threefold) effect on Cl⁻ transport by E148A ClC-ec1 (Fig. 4D). Together, these experimental results support the hypothesis that A404 forms part of the portal that acts a bottleneck for water entry, and thus H⁺ transport.

To further complement these experiments, we also carried out a 110-ns simulation of the A404L mutant and compared the water entrance rate through the portal with the WT system. An initial model of the mutant was constructed, as described in the SI Appendix, on the basis of a selected frame from the WT simulation. A 110-ns segment of the WT trajectory right after this frame, as denoted by a blue bar in Fig. 2A, was used for comparison of the water entrance rates. Note, however, that the permeation rate for WT is essentially constant throughout the simulation (SI Appendix, Fig. S4B). Within the compared trajectories (both 110 ns long), far fewer water molecules permeated through the portal of the A404L mutant than those in WT (6 vs. 34) (SI Appendix, Fig. S4C). Inspection of the simulated structures (Fig. 4E) reveals that C_δ of the introduced leucine occupies the space between the E202 and Glu_in side chains, leaving little room for water molecules to cross the portal. Because of the resulting diminished rate of water penetration from bulk solvent, formation of water wires was not at all observed in the mutant simulation, even though the central region was hydrated at the beginning of the simulation. In contrast, water wires formed repeatedly in the same period for the WT simulation (Fig. 2A). The observed retardation of water wire formation agrees with the ∼35-fold reduction in H⁺ flux observed in our experiments (Fig. 4C).

Conformational Changes Required by Water Wires.

To examine how protein conformations correlate with the hydration of the central region and the formation of the water wires, all of the simulated frames were grouped according to the number of water molecules present in the central region. All of the frames with complete water wires were placed into a separate group. The global conformational changes for each of the ensembles were analyzed through the principal component analysis, as detailed in SI Appendix, revealing no apparent correlation with the hydration of the central region or formation of the water wires (SI Appendix, Fig. S6). The local conformations were characterized by side-chain conformational distributions of residues within the central region (SI Appendix, Table S1), with the distributions across ensembles compared by evaluating their cosine similarity scores, a quantity used to describe similarity between distributions (SI Appendix). The results, as detailed in the SI Appendix, identify I109 as the residue most conformationally sensitive to the formation of the water wires (SI Appendix, Table S1). I109 can adopt two side-chain conformers (SI Appendix, Fig. S7), each of which essentially appears either in the absence (χ₂ = ∼−180°) or the presence (χ₂ = ∼−60°) of the water wires. Structural comparison (Fig. 5 A and B) shows that in the absence of water wires, I109, F199, F357, and Y445 are within close contact, occupying the space to be taken by the water wire; the space becomes available once C_δ of I109 has moved away from this region through side-chain rotation. These results suggest that the formation of the water wires requires a specific conformational change of the I109 side chain.

Fig. 5.

Coupling of residue I109 to water wires. (A) Side chain conformation of I109 in the ClC-ec1 crystal structure. The conformational change that occurs on formation of water wires is indicated by a curved arrow. (B) Rotation of I109 as observed in the simulated structure with the most representative water wire. (C) Summary of Cl⁻ and H⁺ unitary turnover rates determined for each of the I109 mutants. Asterisk denotes the WT residue. (D) Effects of the I109F mutation in the WT background (Left) versus in the uncoupled E148A background (Right). Cl⁻ turnover rates are normalized to their respective background.

To experimentally test that I109 is important in the formation of water wires (and thus H⁺ transport), we made a series of mutations at this position. All five substitutions had significant effects on both H⁺ and Cl⁻ transport rates (Fig. 5C). An effect on Cl⁻/H⁺ stoichiometry was also observed in I109W, which was most severely crippled in its ability to transport H⁺ (SI Appendix, Fig. S5B). Given the extreme sequence conservation at I109 and surrounding residues (SI Appendix, Fig. S8), the observed deleterious effect caused by the mutations does not distinguish between a mechanism by which the mutation disrupts the H⁺ permeation pathway specifically, as predicted computationally, versus a nonspecific mechanism. To address this point, we further evaluated I109F in the E148A background, where H⁺ transport is eliminated (as was done for A404L in the experiments shown in Fig. 4D). Strikingly, the I109F mutation, which decreases Cl⁻ transport through WT ClC-ec1 by more than 80%, has absolutely no effect on Cl⁻ transport by E148A ClC-ec1 (Fig. 5D).

Discussion

An unexplained detail of the CLC transport mechanism is how protons are translocated across a 15-Å central hydrophobic region between the established proton sites, Glu_ex and Glu_in (3, 9, 22, 28). To resolve this key mechanistic detail, we performed extended equilibrium MD simulations and complementary experimental analysis to evaluate whether water wires might mediate proton transport between the two sites and to characterize factors that support water wire formation.

To connect Glu_ex and Glu_in through any water wire, the central hydrophobic region first needs to be sufficiently hydrated. Wang and Voth (30), using a modified version of DOWSER (37) for water placement and short MD simulations, reported that 3–4 water molecules could be placed in this region. It remained, however, unclear whether this pattern of hydration was thermodynamically possible, especially when the protein is in equilibrium with bulk water. We demonstrate here that under equilibrium conditions, this region can indeed accommodate as many as six water molecules (Fig. 1C). Emptying and refilling events occur frequently in this region, as manifested by the fluctuation of its water content, indicating a constant exchange of water between the central region and bulk solvent.

During our extended equilibrium simulation, we not only observed spontaneous hydration of the central region but also saw spontaneous formation of continuous water wires connecting Glu_ex and Glu_in. The wires form frequently and last for up to 2.4 ns, which is long enough for H⁺ to be exchanged between Glu_ex and Glu_in through the Grotthuss mechanism (38). The structure of the water wire observed in the current study (Fig. 2) is substantially different from that reported by Kuang and colleagues (29), in which a connected wire can only be realized after the inclusion of the hydroxyl group of Y445. The extracellular part of our wire model is similar to another model proposed by Wang and Voth (30). However, their wires end in the middle of the central region, whereas our wires extend all of the way to Glu_in.

A key aspect of the process of hydration, which is highly relevant to the frequency of the formation of water wires, is how water in bulk solvent gains access to the central region. We demonstrate that water molecules hydrating the central region come from bulk solvent on the cytoplasmic side, not the extracellular side, and that they enter the central region via a portal at the dimer interface lined by E202, Glu_in, and A404 (Fig. 3). In particular, both the simulations and our experiments clearly show that the small size of A404 is a critical factor for efficient access of water to the central region. It is noteworthy that mutation of this small hydrophobic residue into other hydrophobic residues produces such unambiguous effects on H⁺ transport. In contrast, aggressive mutagenesis of the polar residues leading to and surrounding Glu_in yielded largely insignificant effects on transport, with the sole exception of E202 (36). Interestingly, according to our results, E202 is one of the identified portal-lining residues and the second most important residue for water permeation (Fig. 3). All of these observations highlight the functional importance of the characterized portal in H⁺ translocation by ClC-ec1.

Previous studies have suggested that ion transport in the CLC family may involve protein conformational changes (24, 30, 31, 34), ranging from a global breathing motion of the CLC dimer (31) to a flipping motion of the Glu_in side chain (30). Our analysis shows that a change involving only the movement of C_δ methyl group of I109 is sufficient for the formation of water wires. The importance of such movement is supported by our experimental results. When I109 is replaced by larger side chains (Phe or Trp), considerable reduction of the H⁺ transport activity occurs (Fig. 5C), as expected because of steric constriction. When I109 is replaced by smaller side chains (Ala or Val), reduction of transport also occurs, although to a lesser degree (Fig. 5C). In this case, subtle collapse of the surrounding hydrophobic residues may slightly disfavor water wires. Specificity for the H⁺ transport pathway is demonstrated by the complete absence of effect on Cl⁻ transport in I109F/E148A (Fig. 5D).

Interaction analysis provides additional mechanistic insight and a physical explanation for the long-known phenomenon that the presence of Cl⁻ at S_cen correlates with Cl⁻/H⁺ transport stoichiometry. Early studies showed that the crystallographic density for anions at S_cen decreases with mutations at Y445 (18) or when Cl⁻ is replaced by polyatomic anions (19), and that these decreases correlate strikingly well with Cl⁻/H⁺ coupling stoichiometry. To better understand this phenomenon, Picollo and coworkers (27) combined functional studies and free energy simulations to demonstrate that Cl⁻ and H⁺ bind simultaneously to ClC-ec1 and that binding of two Cl⁻ ions is required to lower the free energy of H⁺ binding to Glu_ex, providing a convincing case that the effect is a key feature of the CLC antiport mechanism. In agreement with these results, using semiempirical free energy calculations, Zhang and Voth suggested that the presence of both central and internal Cl⁻ ions can facilitate H⁺ transport by lowering the kinetic barrier of H⁺ transfer from Glu_in to Glu_ex (33), given the presence of preformed water wires (30). Our results show that the presence of the Cl⁻_cen is also important for the stability of the water wire itself, adding a crucial layer of detail to our understanding of the Cl⁻/H⁺ coupling mechanism.

Our results provide a framework for refining models of the CLC transport mechanism. The proposed models vary significantly in detail, but all invoke movement of Glu_ex coupled to substrate binding and unbinding, and all rely on a step in which Glu_ex is protonated from the intracellular side. One way in which the models differ from one another is in the location of the Glu_ex side chain during this protonation from the intracellular side: in a “down,” “intermediate,” or “middle” position (SI Appendix, Fig. S9). [The “up” position is occupied when Glu_ex is protonated/deprotonated from the extracellular side (11).] In the transport cycle model proposed by Feng and colleagues, the Glu_ex side chain is protonated from the intracellular side when in a down position, occupying S_cen, a state observed in the CLC structure from Cyanidioschyzon merolae (cmCLC) (39). In comparison, the model proposed by Miller and Nguitragool involves movement of Glu_ex to an “intermediate” position between S_ext and S_cen, which remains occupied by Cl⁻ (23), a conjecture supported by the recent structure of a putative ClC-ec1 transport cycle intermediate (40). Finally, the studies of Picollo and coworkers (discussed earlier) indicate that Glu_ex is more favorably protonated when in the middle position (occupying S_ext) than when positioned down at S_cen. Our study supports this latter possibility by revealing water wires that connect Glu_in to Glu_ex in the middle position. Nevertheless, because our study has focused solely on WT ClC-ec1, we cannot rule out that water wires could also connect to Glu_ex when it is located in the alternatively proposed positions. Interestingly, Feng and colleagues found that free glutamate can support H⁺ transport in the absence of Cl⁻ by occupying the Cl⁻-permeation pathway; however, the precise position and orientation of the glutamate could not be determined unequivocally from the 3-Å structure in that study (41). Future studies to evaluate the effects of glutamate positioning on water wire formation in both ClC-ec1 and cmCLC will help resolving the details of the CLC transport mechanism as well as any potential differences among CLC homologs.

The mutants specifically designed in the present study to examine the importance of the portal and the conformational coupling of side chains to water wires (A404 and I109, respectively) provide unique guidance for investigating the coupling between Cl⁻ and H⁺ transport. The portal mutant A404L predominantly affects H⁺ transport but has a secondary effect on Cl⁻ transport, as indicated by the threefold inhibition observed in the E148A background (Fig. 4D). A comparable minor effect on Cl⁻ transport was observed with mutations at E202 (36), another portal residue identified in the current study. In contrast, I109F at the center of the protein exclusively inhibits the H⁺ branch of the CLC transport mechanism (Fig. 5D). The distinction between effects of mutations at 404 and 109 is subtle but remarkable. Residue 109 closely abuts the Cl⁻-permeation pathway and is part of the highly conserved CLC signature sequence GSGIPE (found in both Cl⁻/H⁺ transporters and Cl⁻ channels) (SI Appendix, Fig. S8) (12). Residue S107 in this sequence appears to be an intracellular gate to the Cl⁻-permeation pathway (3, 10). Mutations at 107 unvaryingly disrupt anion transport, with mixed effects on H⁺ transport (21, 25, 42–44). The surgical precision with which I109F retards H⁺ transport demonstrates that different regions of the GSGIPE signature sequence can have separable effects on H⁺ and Cl⁻ transport, despite the close juxtaposition of the H⁺/Cl⁻ pathways at this region of the protein. In contrast, the portal residues are 10–15 Å removed from the Cl⁻-permeation pathway, yet mutations here exert a small but significant effect on Cl⁻ transport. Thus, investigating conformational coupling between the Cl⁻ permeation pathways and the portal may provide additional clues to the CLC transport mechanism.

Materials and Methods

WT and mutant ClC-ec1 proteins were overexpressed, purified, reconstituted, and assayed essentially as described (20, 36), except that Cl⁻ and H⁺ transport were measured simultaneously. The ClC-ec1 structure (Protein Data Bank ID code 1OTS) (10), embedded into a 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine (POPE) bilayer and solvated in 150 mM NaCl solution, was modeled with CHARMM and simulated with NAMD. Solutions and conditions used for reconstitution and transport assay and the details of simulations and analysis are described in SI Appendix.