Structural capture of an intermediate transport state of a CLC CI-/H+ antiporter

The CLC family proteins are involved in a variety of cellular processes, where chloride homeostasis needs to be controlled. Two distinct classes of CLC proteins, Cl- channels and Cl-/H+ antiporters, have been functionally and structurally investigated over the last several decades. Recent studies have revealed that the conformational heterogeneity of the critical glutamate residue, Gluex could explain the transport cycle of CLC-type Cl-/H+ antiporters. However, the presence of multiple conformations of the Gluex has been suggested from combined structural snapshots of two different CLC antiporters. Thus, we aimed to investigate the presence of these three intermediate conformations in CLC-ec1, the most deeply studied CLC at both functional and structural levels. By comparing crystal structures of E148D, E148A mutant and wildtype CLC-ec1 with varying anion concentrations, we suggest that the Gluex indeed take at least three distinct conformational states in a single CLC antiporter, CLC-ec1.


INTRODUCTION
The CLC family proteins are evolutionarily well conserved proteins expressed in the biological membranes and are critical for diverse physiological processes that control Clconcentration including the regulation of skeletal muscle and neuronal excitability, acidification of intracellular organelles, and cell volume regulation, among others 1-3 . Genetic mutations of CLC genes are linked to a variety of pathological outcomes, such as myotonia, leukodystrophy, deafness, retinal degeneration, Bartter's syndrome, Dent's disease, and osteopetrosis [4][5][6] . The CLC family proteins are classified functionally into two mechanistically distinct protein families of Clchannels and Cl -/H + antiporters. The CLC-type Clchannels allow Clto move across the membrane according to its electrochemical gradient. In contrast, CLC-type Cl -/H + antiporters catalyze energetically uphill movement of Clwith the compensation of coupled downhill movement of H + , or vice versa [7][8][9] . In addition to conventional Cl -/H + antiporters, the plant system is equipped with CLC-type NO3 -/H + antiporters that transport nitrate into the vacuoles 10,11 and some bacteria use CLC-type F -/H + antiporters to resist the environmental toxicity of fluoride 12,13 .
Regardless of their mechanistic diversity, the CLC family proteins share a highly conserved key glutamate residue, called the external glutamate (Gluex) or the gating glutamate (Glugate), which serves as a gate for the Cl --conducting pathway in both CLC channels and antiporters as well as for H + transit of CLC antiporters during the transport cycle 8,9,14 . Structural studies of various CLC proteins from bacteria to mammals have revealed their molecular architectures at atomic resolution [15][16][17][18][19][20][21][22] , which immediately suggest the molecular mechanisms of how CLC proteins transport ions. In the structures, the Gluex adopts at least three different conformations along the Cltransport pathway: Up-, Middle-, or Down-conformation ( Figure 1). Based on the rotamerically distinct positions of the Gluex in the CLC structures, it was suggested that the glutamate side chain competes for two anion binding sites with permeant chloride ions and accepts a proton after expelling two chloride ions during the transport cycle of Cland H + exchange 19 .
Though three alternative Gluex conformations would explain 2:1 coupled movement of Cland H + in Cl -/H + antiporters, no single CLC protein has shown three conformational states structurally: Up-and Middle-conformations have been observed in the deeply studied CLC-ec1 from E.coli 15,16 ; Down-conformation in a eukaryotic CLC, cmCLC from a thermophilic red alga 17 ; and Up-and Down-conformations in a distinct CLC-type F -/H + antiporter clade, CLC-Eca 22 . Thus, whether the Gluex in a single CLC protein adopts all three rotameric conformations during the transport cycle remains unclear.
Recent studies on the mutation of the external glutamate (Gluex) in CLC-type Cl -/H + antiporters to aspartate, an amino acid with a methylene (-CH2 group) shorter and a similar pKa value, reported extremely slow ion transport in both CLC-ec1 and cmCLC 19 . Moreover, a mutation at the corresponding glutamate (E166D) in the CLC-0 channel reduced singlechannel conductance as well as drastically decreased open probability 23,24 . These previous observations support the idea that the Gluex has three conformational states (Up-, Middle-, and Down-conformations; Fig. 1C), and reflect the penalty of the shorter aspartate side chain to reach the Scen. Recently, it was also suggested that CLC-ec1 could adopt a cmCLC-like state by a conformational locking of the Gluex at the central anion binding site (Scen), which is induced by high external Clconcentration 25 . Computational study also proposed a transport cycle in which the Gluex could occupy the Scen in CLC-ec1 26 .
However, the presence of cmCLC-like "Down" conformation, the Gluex occupying the Scen, has not been observed in CLC-ec1 structure, only in cmCLC. Thus, we aimed to tackle the following: (1) elucidating the structural change in the E148D mutant CLC-ec1, and explain the mechanistical consequences of E148D mutation, if it exists, (2) visualizing the structural transition occurring due to Asp148 protonation of the E148D mutant by comparing its structure with E148N mutant, and (3) obtaining structural insight into the conformational changes of the Gluex, and especially into its presence at the Scen in CLC-ec1 during the transport cycle.
Structural and functional examinations of E148D and E148N CLC-ec1 revealed unexpected rotameric change upon protonation/deprotonation of Asp148, which resulted in limited solution accessibility to the Asp148 residue and could slow ion transport. From the comparison of the E148D mutant and halide-free wild-type structures, we provide evidence for the presence of a new intermediate "Mid-low" state in the transition between "Middle" and "Down" conformations. This "Mid-low" conformation of Gluex is enough to expel a chloride ion from the central binding site similarly observed in the structure of cmCLC.
Additionally, we reconsolidated a previous observation 19 with anomalously detectable short carboxylic acid, bromoacetate, which favors to bind in a carboxylate-down configuration to the ungated E148A mutant in the analogy of "Down" conformation of Gluex. These results suggest that the external glutamate in a single CLC-type Cl -/H + antiporter can undergo a structural twist from Up-to Down-conformation to exchange both Cland H + .

Ion transport and anion binding in the E148D and E148N mutants
The E148D mutant mediates both Cland H + movements across reconstituted proteoliposomes with ~10-fold smaller transport rates (~180 Cl -/sec at pH 4.5) than that of wild-type CLC-ec1 19,27,28 (Fig. 2A Clwith a turnover rate of ~110 Cl -/sec at pH 4.5, but fails to transport H + , as similarly observed in the E148Q mutant, a protonated Gluex mimic of wild-type CLC-ec1 29 ( Fig. 2A Interestingly, the stoichiometry of Cland H + transport in the E148D mutant is 3.8 ± 0.2 (Fig. 2C, n=3) compared to ~2 to 1 in the wild-type 7,28 . Altered Cl -/H + coupling ratios have been previously observed in mutants showing reduced Clbinding at the central binding site including Y445F and Y445I mutants 7,27 , and mutants limiting intracellular proton access such as E203K and E202L mutants 28,32 . Since altered Cl -/H + coupling of these mutants showed increased Clslippage owing to reduced H + transport, E148D mutation could also affect H + transport and change the stoichiometry of coupling.
In order to test whether markedly reduced ion transport observed in the E148D mutant is caused by changed anion binding affinity, we measured the equilibrium anion binding affinity of the E148D mutant by using isothermal titration calorimetry(ITC) at pH 7.5 28,29 .
Although the Clbinding affinity is lowered in the E148D mutant (~4-fold), the binding affinity change cannot solely account for the slowed ion transport (~10-fold). On the other hand, the similar Clbinding affinity and marginal difference between Cltransport rates in E148N and E148Q mutants (~130 Cl -/sec and ~300 Cl -/sec, respectively; Supplementary Table 1, 2) were observed. These results strongly suggest that the conformational changes of Asp148 at the Gluex position and anion binding configuration in E148D mutants should differ from those of wild-type during the transport cycle. If this is the case, we should be able to see the structural changes induced by E148D mutation.

Crystal structures of the external gate mutants E148D and E148N
To examine whether E148D mutation induces any structural change, we solved crystal structures of E148D and E148N mutants with FAB antibody fragments 16 at resolutions of 2.95Å and 2.7Å, respectively (Table 1). The overall structures of both E148D and E148N mutants are essentially identical to that of the wild-type (Ca r.m.s. deviation 0.26 and 0.19 Å, respectively; Supplementary Fig 2). Crystal structures were obtained with Brto take advantage of anomalous diffraction in identifying anion binding sites with confidence ( Fig. 3, Supplementary Table 3). Since the binding affinity differences between Cland Brin both E148D and E148N mutants are minimal (1.5~3-fold), Clbinding in these two mutants can be inferred from anomalous Brsignals.
Intriguingly, anomalous Brsignals at the Scen, the central binding site at which Clalways localizes in wild-type CLC-ec1, vanished in the E148D mutant while the side-chain of Asp148 residue sat near the external binding site. Instead, a strong anomalous Brsignal remained at the Sint, the internal binding site, which has been considered a low-affinity anion binding site 29 (Fig. 3A). In the E148N mutant structure, three anomalous Brsignals were observed along the Cltransport pathway, just as in E148Q mutant 16 (Fig. 3B). However, the rotameric position of the Asn148 residue is very different from that of the Gln148 in E148Q mutant. The side chain of Asn148 takes a rotamer horizontally away from the position of Asp148 inducing ~3.2Å movement of the carboxyl carbon from the center of Cltransport pathway, whereas the Gln148 side chain vertically rotates ~56 o with ~3.6Å translation of the carboxyl carbon from the Sext (Fig. 3C, D). These distinct side chain positions of Asp148 and Asn148 could reflect a conformational change in the E148D mutant upon protonation of Asp148, as occurs between the wild-type and E148Q mutant. Comparison of the solvent accessibilities of Asn148 and Gln148 residues suggests that the E148D mutant experiences difficulty in delivering a proton to the outside of the protein, and vice versa ( Supplementary   Fig. 3). This could retard H + transport, followed Clslippage, and result in an altered Cl -/H + stoichiometry in E148D mutant (Fig. 2C).

The E148D mutant mimics the conformation of a new intermediate state of the transport cycle
What is the driving force to expel Clfrom the central binding site of E148D mutant while Asp148 is present near the external binding site (Fig. 3A)? To obtain a possible explanation, we measured atomic distances between residues involved in Clcoordination at the central site in the E148D mutant and wild-type CLC-ec1 structures in the presence and absence of halide ions ( Fig. 4 and Supplementary Table 4). In the E148D mutant structure, the carboxyl group of Asp148 slightly moved down to the central Clbinding site compared to the position of Glu148 in the Middle position in the wild-type structure, and atomic distances between carboxyl (Asp148) and hydroxyl oxygens (Ser107 or Tyr445) decreased by 1.6~1.9Å (Fig. 4A, B; Supplementary Fig. 4A). Thus, it is natural to imagine that anions are not able to sit at the central binding site in the E148D mutant in considering the ionic radii of Cland Br -, ~1.8Å and ~2Å, respectively 34 . Interestingly, the position of the carboxyl group of Asp148 is almost identical to that of Glu148 carboxylate in the halide-free wild-type CLC-ec1 structures previously solved by two independent groups 33 In the wild-type CLC-ec1 structure, the backbone amides from helix F and helix N offer an "anion hungry" binding site, Sext, which Cland the deprotonated Glu148 side chain compete with one another to occupy (Fig. 1C) 15,16 . It is tempting to ask whether the carboxylates of both Asp148 in the E148D mutant and Glu148 in the wild-type in the absence of halide ions have enough coordinating groups to sit in between the external and central binding sites, as Glu148 carboxylate is stabilized by nine backbone amides within a 4Å distance in the wild-type structure while a chloride ion occupies the central binding site (Fig.   4D). The answer is yes: nine to ten backbone amides generate a snug coordination geometry for both cases (Fig. 4E, F). Thus, it can be inferred from the analogy between the structures of halide-free wild-type CLC-ec1 and the E148D mutant that the halide-free conformation is enough to expel Clfrom the Scen, and that it could be a new intermediate, "Mid-low" conformation of Gluex in the transport cycle.

The presence of extra anion binding site above the external binding site
A recent study showed that high external Clconcentration (above 300 mM) significantly slowed both Cland H + transports in CLC-ec1 by limiting conformational changes of the Gluex from Down-to Up-conformations 25 . Also, a computational study on CLC-ec1 suggested an extra Cl-binding site above the Sext 36 . On the contrary, it was reported that high Brconcentration up to 400 mM increased anomalous Brdensities at both Scen and Sint instead of revealing any Brbinding at Sext or above it in the wild-type CLC-ec1 structure 33 . These data propelled us to identify an anion binding site near Sext in E148D mutant since Asp148 side chain moves slightly down as a Mid-low conformation, which might give a chance to reveal the external anion binding at Sext or above it. Impressively, we were able to see an extra anomalous Brsignal above Sext by increasing Brconcentration: anomalous Brsignal does not appear at the extra binding site (the extra external site, Sxet) in 20 mM NaBr condition, but begins to appear in 50 mM NaBr and becomes prominent in 200 mM NaBr ( Fig. 5A; Supplementary Table 3). The bromide ion at Sxet is coordinated by the backbone amide and side chain guanidinium group of Arg147 and is located ~6 Å above Sext along the anion transport pathway (Fig. 5B). Interestingly a crystallographic water molecule was reported at the position of Sxet in the wild-type CLC-ec1 structures 16,28 . These results suggest that Sxet is a genuine low-affinity anion binding site in CLC-ec1, which can be occupied by either a water molecule or halide ion depending on the anion concentration. Thus, it is possible to imagine that the conformational change of Gluex from Middle to Up-conformations might be hampered by extra Clbinding at Sxet, which induces the slowed ion transport in high [Cl -] condition as observed previously 25 .

Bromoacetate can sit at the Scen in the E148A mutant
Observations of the E148D and wild-type structures indicated that the halide-free Previously, it was shown that carboxylic acids such as glutamate or gluconate in solution can restore H + transport in the ungated mutant E148A in which the Gluex is substituted with alanine, and also that glutamate might occupy the Scen site in E148A CLC-ec1 19 . These observations imply that the Gluex might go further down in CLC-ec1; however, the presence of glutamate, specifically its g-carboxylate, at the Scen is not conclusive, since it was deduced from the omit (Fo -Fc) map at a 3.05 Å resolution. Thus, we decided to re-visit a carboxylate occupancy at the Scen with an anomalously detectable short carboxylic acid, bromoacetic acid (BAA) instead of glutamate (Fig. 6A). Before determining crystal structure of E148A and BAA complex, we first tested whether BAA can bind to the Clbinding site in E148A CLC-ec1 by measuring the thermodynamic Clbinding affinities in the presence of a varied concentration of BAA. The binding affinity of E148A CLC-ec1 to Clwas decreased by ~10-fold (20 µM to 200 µM) in the presence of 10 mM BAA and it was completed abolished by 50 mM BAA (Fig. 6B~D).
Interestingly, two anomalous bromine signals were found along the Cltransport pathway in the E148A mutant structure: a slightly stronger signal near the Sext and a weaker signal near the Scen (Fig. 6E; Supplementary Table 3). We speculate that the anomalous signals reflect alternatively positioned bromoacetates in the anion transport pathway, since two anomalous densities are placed in its proximity and Fo -Fc density is not significantly extruded by the two anomalous signals (Fig. 6E). Based on the anomalous intensities, it can be inferred that both the carboxylate-down and carboxylate-up configurations are favorable in the Cltransport pathway. Especially, the bromine group is found at the Sext and the carboxylate group is placed at the Scen in the carboxylate-down configuration (Fig. 6E~G).
These results strongly suggest that the carboxylate of the Gluex can move down to the Scen confirming the presence of a cmCLC-like conformation in CLC-ec1. However, this may not necessarily be the case when the side chain is tethered to the protein, since the favorability might depend more on the internal energies of Glu148 rotamers connected to the backbone attached to the protein. Nevertheless, we examined at the chemical properties of carboxylate groups dissociated from the protein in the Cltransport pathway.

DISCUSSION
In the present study, we examined a series of Gluex mutants to gain understanding of the ion transport mechanism of CLC Cl -/H + antiporter. From the combined approaches of ion transport measurements, equilibrium binding studies and x-ray crystallography on CLC-ec1, we found the following: (1) aspartate mutation of the Gluex significantly slows ion transport rate and slightly reduces anion binding; (2) slowed ion transport in the E148D mutant could be attributable to the limited solution accessibility to the Asp148 residue, which makes protonation and deprotonation of the Asp148 side chain difficult, as seen in the structure of Our data provide evidence of intermediate states in the transport cycle of CLC Cl -/H + antiporters, and specifically elucidate how movement of Gluex can be stabilized in the transition between the "Middle" and the "Down" conformations. The uniquely suggested "Mid-low" conformation of the Gluex is stabilized by backbone amides from helix F and helix N, which are thought to be coordinate the "Middle" conformation of Gluex. Once the side chain of Gluex adopts the "Mid-low" conformation, Gluex can move further to the Scen, since bromoacetate favors the carboxylate-down configuration to the ungated E148A mutant.
The structure of CLC-1 Clchannel from Homo sapiens was recently determined by cryo-electron microscopy 21 . Interestingly, the Glugate locates to the side of the Cltransport pathway in CLC-1 while its protonation state is unclear. By coincidence, a notably different rotameric position of Glugate in CLC-1 is quite similar to that of Asn148 in the E148N mutant Recently, structural studies on a CLC-type F -/H + antiporter from Enterococcus casseliflavus revealed two conformations, in which the equivalent Gluex was exposed to either the extracellular face in the Up position or to the intracellular face in the Down position 22 . Thus, it can be considered a general rotameric twist in CLC-type antiporters that the external glutamate or the gating glutamate switches its positions to Up, Middle, and Down during the transport cycle regardless of differences in transport mechanism of anion/H + exchanges.
Unlike other alternatively accessing transporters, CLC Cl -/H + antiporters have been considered rather static without large conformational changes 37,38 . As a coupled transporter, the Gluex up-conformation, which potentially allows access from both intracellular and extracellular sides, should be avoided to maintain stoichiometrically coupled movements of Cl -/H + . It has been suggested that both tyrosine (Tyrcen) and serine (Sercen) residues near the Scen act as an inner gate, whose opening occurs only when Gluex caps the external Clentrance 39

Protein purification and crystallography
Analytical-grade chemicals and reagents were purchased from Sigma-Aldrich or otherwise specified, and lipids were from Avanti Polar Lipids. Expression and purification of CLC-ec1 was performed as described 7,28 . The protein was run on a size-exclusion column equilibrated with 100 mM NaCl, 10 mM Tris-HCl, pH 7.5, 5mM Decyl Maltoside (Anatrace) for ion flux experiments.
Protein was concentrated to 10~20 mg/mL and mixed with an equal volume of crystallization solution in a hanging-drop vapor-diffusion chamber. Crystals, appearing in 20-28% (w/v) PEG 400, 0-20 mM Na/K tartrate, pH 7.5-9.0 in 3-10 days at 22 °C, were cryo-protected by slowly increasing PEG concentration in the mother liquor to ~35%, followed by freezing in liquid N2.
Data sets were collected at beamline BL-5C or BL-11C Pohang Accelerator Laboratory (PAL; Pohang, Korea) at X-ray wavelength of 1 Å, or at 0.919 Å for Branomalous diffraction. Data were integrated and scaled using HKL2000 or imosflm, and initial models were obtained by molecular replacement against 4ENE or 1OTS using Phaser in the CCP4 software suite.
Models were rigid-body refined in REFMAC5 and further refined in Phenix. Protein Data Bank accession codes for each crystal are reported in Table 1.

Cland H + flux assay
Formation of liposomes reconstituted with wild-type or mutant CLC-ec1 (1-5 µg protein/mg lipid) and ion flux measurements have been described in detail 28   Critical residues(Gluex, Sercen, and Tyrcen) for Cltransport in the structures of CLC antiporters are shown as stick models (yellow, carbon in wild-type CLC-ec1; magenta, carbon in E148Q CLC-ec1; cyan, carbon in cmCLC for carbon; blue, nitrogen; and red, oxygen). Chloride ions are shown as green spheres. The PDB accession numbers for wild-type CLC-ec1, E148Q CLC-ec1, and cmCLC are #1OTS, #1OTU, and #3ORG, respectively. C. Diagram of conformational diversity of Gluex in CLC antiporters. The internal, central and external anion binding sites are marked as Sext, Scen, and Sint, respectively. Note that helix F (hF) and helix N (hN) offer helical dipoles (d + ) to stabilize both Cland Gluex binding at Sext 15 .     . Red dashed lines indicate distances between Cland coordinating residues. Numbers next to the dashed lines indicate distances in Å. Each distance was measured in chain A from the PDB structures of wild-type and mutant CLC-ec1 (#1OTS for wild-type with Cl -(A), #4KJP for wild-type without halide ion (B), and #6AD7 for E148D (C)) using the PyMOL software. Coordination of neighboring amino acids to the carboxylates of Glu148 and Asp148 residues in wild-type and mutant CLC-ec1 (D~F). Red dashed lines indicate distances less than 4 Å between nearby amino acid residues and g-carboxylates of the Gluex of wild-type CLC-ec1 in the presence (D) or absence (F) of Cl -, and b-carboxylates of Asp148 in the E148D mutant (E).