Mechanism of Cl- Selection by a Glutamate-gated Chloride (GluCl) Receptor Revealed through Mutations in the Selectivity Filter*

To learn about the mechanism of ion charge selectivity by invertebrate glutamate-gated chloride (GluCl) channels, we swapped segments between the GluClβ receptor of Caenorhabditis elegans and the vertebrate cationic α7-acetylcholine receptor and monitored anionic/cationic permeability ratios. Complete conversion of the ion charge selectivity in a set of receptor microchimeras indicates that the selectivity filter of the GluClβ receptor is created by a sequence connecting the first with the second transmembrane segments. A single substitution of a negatively charged residue within this sequence converted the selectivity of the GluClβ receptor's pore from anionic to cationic. Unexpectedly, elimination of the charge of each basic residue of the selectivity filter, one at a time or concomitantly, moderately reduced the PCl/PNa ratios, but the GluClβ receptor's mutants retained high capacity to select Cl- over Na+. These results indicate that, unlike the proposed case of anionic Gly- and γ-aminobutyric acid-gated ion channels, positively charged residues do not play the key role in the selection of ionic charge by the GluClβ receptor. Taken together with measurements of the effective open pore diameter and with structural modeling, the study presented here collectively indicates that in the most constricted part of the open GluClβ receptor's channel, Cl- interacts with backbone amides, where it undergoes partial dehydration necessary for traversing the pore.

shared by Cys-loop receptors enabled swapping of pore sequences between cationic and anionic channels so as to assess the involvement of specific amino acids in ion charge selectivity. It was previously shown that concomitant replacement of the residues at positions Ϫ2Ј, Ϫ1Ј, and 13Ј ( Fig. 1, B and C) of cationic receptors by the residues found at the homologous positions of anionic receptors, and vice versa, lead to conversion of ion charge selectivity (10 -13).
Further mutagenesis studies led to the recognition that the different capacities of cationic versus anionic Cys-loop receptors to distinguish between the charge of ions rely on the differences in the amino acid composition at positions Ϫ1Ј and Ϫ2Ј (Fig. 1C) (13)(14)(15)(16). The conserved pore-facing Glu residue at position Ϫ1Ј of cationic Cys-loop receptors was further inferred to form, around the axis of ion conduction, a negatively charged ring that plays the key role in cationic selectivity by interacting with cations and repulsing anions (12,13,15,16). Conversely, a conserved arginine at position 0Ј of anionic Cys-loop receptors was inferred to interact with anions and repulse cations. A basic residue at position 0Ј is also typical of all cationic Cys-loop receptors ( Fig. 1C and the ligand-gated ion channels data base), but it was suggested that local conformational differences in the M1-M2 connecting segment (M1-M2 loop) orient this basic residue to the pore lumen only in anionic Cys-loop receptors (9,15). These local conformational differences have been attributed to a proline residue, which is present exclusively at position Ϫ2Ј of anionic Cys-loop receptors (12, 14 -17).
Unlike all other homomeric anionic Cys-loop receptors studied thus far, the ␤ subunit of the GluCl receptor does not have a proline residue at position Ϫ2Ј, a feature that minimizes the likelihood of causing drastic local conformational changes when mutating its M1-M2 loop. As the GluCl␤ subunit can assemble into a functional anionic homomeric receptor (GluCl␤R) (18), the selectivity filter of the GluCl␤R was readily identified here by microchimerism and then was extensively mutated. Electrophysiological analyses of ionic permeability ratios in a large repertoire of mutants, together with computer-assisted molecular modeling, reveal a novel mechanism of Cl Ϫ selection by a Cys-loop receptor.
protein (cloned in a separate plasmid), using the calcium phosphate method. Electrophysiological recordings were made 2-3 days after the transfection, using the whole-cell patch clamp technique. To determine ionic permeability ratios, current-voltage relations were established in various external solutions with and without 100 M ACh as follows. Solution A contained: 140 mM NaCl, 2.8 mM KCl, 2 mM CaCl 2 , 2 mM MgCl 2 , 10 mM glucose, 10 mM HEPES, and 5 mM NaOH, pH 7.35. Solutions B 1,2 . . . are as solution A, but NaCl was diluted with a mannitolcontaining solution to maintain equiosmolar conditions while reducing the external NaCl concentrations to (in mM): B 1 ϭ 70, B 2 ϭ 35, B 3 ϭ 17.5, B 4 ϭ 0. Note that in one exceptional case (chimera 4), the composition of the diluted solutions was slightly different, giving external NaCl concentrations of (in mM): B 1Ј ϭ 100, B 2Ј ϭ 40, B 3Ј ϭ 20, B 4Ј ϭ 10, B 5 ϭ 0. Solutions C 1,2 . . . are as solution A, but NaCl was replaced by sodium isethionate so as to keep the external concentration of Na ϩ ions constant and to replace Cl Ϫ ions by isethionate. The external concentrations of NaCl in these solutions was (in mM): C 1 ϭ 70, C 2 ϭ 35, C 3 ϭ 17.5, C 4 ϭ 0. Solution D is as solution A, but NaCl was replaced by 140 mM sodium acetate. Patch pipettes (1-2 megaohms) were filled with a solution containing 130 mM CsCl, 4 mM MgCl 2 , 4 mM Na 2 -ATP, 1 mM EGTA, 10 mM HEPES, and 10 mM CsOH, pH 7.35. External solutions were delivered by a computer-driven valve manifold system enabling fast exchange of the solutions. Current-voltage relations were deter-mined at 25°C by two methods. (i) Currents evoked by 100 M ACh for 3 s were measured at different holding potentials ranging from Ϫ100 to ϩ50 mV, and (ii) inverted 200-ms long voltage ramps (either from ϩ50 to Ϫ100 mV or from ϩ70 to Ϫ100 mV) were applied 1 and 2 s after the beginning of a 3-s ACh application. The initial holding potential was Ϫ60 mV. Leak currents obtained by the same protocol, in the absence of ACh, were subtracted. Measured reversal potential (E rev ) values were corrected to account for the liquid junction potentials by using the JPCalc software (21) implemented in pClamp version 8.1. Permeability ratios for Na ϩ , Cs ϩ , and Cl Ϫ were calculated using the Goldman-Hodgkin-Katz equation, where R, T, and F are the gas constant, the absolute temperature, and the Faraday's constant, respectively, and the permeability ratios are ␣ ϭ P Na /P Cl and ␤ ϭ P Cs /P Cl . Note that ion activities, instead of ion concentrations, have been used in the analyses of permeability ratios. Ionic activities were calculated on the basis of the Debye-Hückel theory (22) but with the corrections introduced in the Millero-Pitzer method for solutions having an ionic strength greater than 0.1 M (41), as implemented in the Electrolytes program of Aq-Solutions, a software package of programs for the quantitative treatment of equilibria in solution. 3 Model Building-An initial homology model was built by using the atomic coordinates of the AChR structure (23) as a template (Protein Data Bank number 1OED), as recently described (20). After modeling the M1-M2 loop, it was tilted together with the M1 and M2 segments to an intermediate position between the closed and open states previously elaborated by Paas et al. (20) in the case of another Cys-loop chimeric receptor. We readily obtained an intermediate position displaying a distance of ϳ6.1 Å between the van der Waals surfaces of backbone amides located on opposite sides of the pore, at the level of position Ϫ3Ј. This distance is within the range of the effective open pore diameter determined here. Note that the M1-M2 loop of the GluCl␤R (DLH-STAG) is shorter by two amino acids than the loop of the models elaborated in Paas et al. (20) (PPDLHSTAG). As a result, when compared with its position in the chimera modeled in Paas et al. (20), histidine(H)-5Ј of the GluCl␤R's pore (modeled here) moved away from the permeation pathway. Consequently, HϪ5Ј of the GluCl␤R's model is not in contact with the permeating ions. This modeling observation is in line with the incapacity of Zn 2ϩ to block the chimeric ␣7-GluCl␤R (data not shown), unlike the case of a chimera having the transmembrane segments of the 5HT 3A R with the sequence PPDLHSTAG between M1 and M2 (20). The model of chimera 17 was built as above but with a proline, instead of alanine, at position Ϫ2Ј. The root mean square difference between the backbone atoms of the GluCl␤R and chimera 17 is 0.03 Å (throughout the pentameric membrane-embedded domains). The root mean square difference between the backbone atoms of the M1-M2 loops (plus their flanking residues, IϪ10Ј and R0Ј) of these structures is 0.08 Å.

Chimeric Design and Current Amplitudes of the Various Mutants-
As a first step, we generated a chimeric subunit where the extracellular segment of the AChR ␣7 subunit was fused to the GluCl␤ subunit segment that folds in the membrane and cytoplasm (see "Experimental Procedures" and supplemental Fig. S1). The ␣7-GluCl␤ chimeric subunit assembles into a homopentameric receptor (␣7-GluCl␤R), like the Two of five subunits were removed to expose the pore lumen formed by the membrane-embedded regions. All receptor subunits of this superfamily share similar amino acid sequences, the same transmembrane topology and a completely conserved disulfide bridge (indicated by boldfaced C) in their extracellular ligand-binding domain (represented in red), which accommodates the neurotransmitter (NT ). B, the second transmembrane segment (M2) lines the channel pore. The numbers punctuated with a prime relate to the location relative to the position commonly considered as the first residue of M2 (K0Ј or R0Ј). The selectivity filter is considered to be composed of amino acids belonging to the M1-M2 loop, where position Ϫ1Ј is playing a crucial role in the selection of cations over anions by cationic Cys-loop receptors. C, amino acid sequence alignment of five subunits of Cys-loop receptors. Shown are: part of M1, the entire M2, and the M1-M2 loop that displays a highly conserved negatively charged residue (black box). M2 is shown with a black horizontal line above the sequence. The multiple sequence alignment takes into account the common view that a proline residue occupies position Ϫ2Ј in most anionic Cys-loop receptor subunits, whereas none of the cationic Cys-loop receptor subunits has a proline at this position. Note that gaps are also numbered. AChR-␣7, chick ␣7 subunit of an acetylcholine-activated cationic channel (Swiss-Prot accession number P22770); 5HT 3A R, 5-hydroxytryptamine (serotonin)-activated cationic channel from mouse (Swiss-Prot accession number P23979); GlyR-␣1, human ␣1 subunit of a glycine-activated anionic channel (Swiss-Prot accession number P23415); GABA A R-1, human 1 subunit of an anionic channel activated by ␥-aminobutyric acid (Swiss-Prot accession number P24046); GluClR-␤, ␤ subunit of a glutamate-activated chloride channel from Caenorhabditis elegans (Swiss-Prot accession number Q17328). Note that each of these individual subunits is capable of assembling into a functional homopentamer.
case of all the mutants studied here. As such, the resulting ␣7-GluCl␤R responds to ACh but not to glutamate, allowing us to express our chimeric mutants in HEK-293 cells grown in serum-containing medium with no need of glutamate depletion. That is, this chimeric receptor carries the entire pore of the weakly desensitizing GluCl␤R (18) without being constitutively open (by glutamate present in the medium) and cytotoxic to the cells. Fig. 2A shows typical current traces of the chimeric ␣7-GluCl␤R measured at different membrane voltages. Then, based on previous studies carried out on a few vertebrate Cys-loop receptors, mutations were introduced in regions suspected to contribute to ion charge selectivity. This approach resulted in chimeric mutants that provided robust responses (Table 1).
Ion Charge Selectivity of the ␣7-GluCl␤R and Its Various Mutants-The I-V relations plotted for the ␣7-GluCl␤R show that omission of almost all external Cl Ϫ ions shifts the reversal potential (E rev ) to a positive voltage (e.g. Fig. 2B), by 54.5 Ϯ 1.2 mV (mean Ϯ S.E. from 12 cells, after correcting with liquid junction potentials). The latter value is close to the maximal theoretical shift calculated based on the Cl Ϫ equilibrium (Nernst) potential (61.4 mV; calculated using ionic activities). The extent of the shift depended on the external concentrations of Cl Ϫ (Fig.  3A) and showed that the ␣7-GluCl␤R chimera is highly selective to Cl Ϫ relative to Na ϩ and Cs ϩ (P Cl /P Na ϭ ϳ45 and P Cl /P Cs ϭ 26; Table 1), as expected from a chimera harboring the pore of the GluCl␤R (18). Previous studies suggested that a positively charged residue substituted at the extracellular mouth of Cys-loop receptor mutants (position 19Ј) partially contributes to ionic selectivity by attracting Cl Ϫ ions (24) or repulsing divalent cations (15). It was also shown that an Arg 19Ј 3 Glu mutation in an anionic-to-cationic converted glycine receptor mutant contributes to cationic conductance (25). We therefore replaced the Measurements that were performed with a normal solution (140 mM external NaCl; solution A under "Experimental Procedures"), sodium isethionate (instead of the external NaCl), or mannitol (instead of the external NaCl) are shown in black, blue, and red, respectively. Note that these are examples for the maximal shifts in reversal potentials (E rev values) obtained when the external concentration of NaCl is zero (in blue and red curves). Also note that the E rev values, which are extracted from I-V relations such as those shown in panels B-F, should further be corrected to account for the liquid junction potentials (see "Experimental Procedures").

TABLE 1
Permeability ratios (P X /P Y ) determined for the chimeric ␣7-GluCl␤R and its mutants Note that: (i) data are means Ϯ S.E.; (ii) all values were rounded to the closest decimal number; (iii) pair sequence alignment between the ␣7 and the GluCl␤ receptors dictates the movement of the ␣7 GϪ3Ј (alignment in Fig. 1C) by one position downstream, i.e., it is GϪ2Ј in Table I; (iv) gaps in the sequence are also numbered; (v) residues shown in blue do not appear in the native sequence of either the ␣7-AChR or the GluCl␤R; (vi) the entire sequence of the ␣7-GluCl␤ chimeric subunit is provided in the Supplemental Data (Fig. S1).
a Measured in response to 100 M acetylcholine at Ϫ60 mV. b Number of cells. c Ion charge selectivity: A, anionic; C, cationic. d The P Cl /P Cs ratio of chimeras 2 or 12 does not statistically differ from that of the ␣7-GluCl␤R (P ϭ 0.059 and 0.150, respectively; two-tailed, unpaired t test). e The P Cl /P Cs ratio of chimeras 13, 14 or 15 significantly differs from that of the ␣7-GluCl␤R (P ϭ 0.0015, 0.0025, and 0.0009, respectively; two-tailed, unpaired t test). f NF, not functional. segment 19Ј NAKL 22Ј of the GluCl␤R pore by the homologous segment of ␣7 ( 19Ј AEIM 22Ј ), and the resulting chimera retained Cl Ϫ selectivity, albeit with a lower P Cl /P Na ratio, but with no change in the P Cl /P Cs ratio (chimera 1, Table 1 and Supplemental Fig. S2A). Cl Ϫ selectivity was also observed when the positive charge of the first amino acid of M2 (position 0Ј) was neutralized (chimera 2, Table 1 and Figs. 2C and 3A).
Position Ϫ1Ј is located in a constriction that extends from position 2Ј toward the bottom of the pore (20, 26 -30). Consistently, it was previously shown that a substitution of Glu at position 2Ј of the homomeric Gly␣1R converted its selectivity (31). In addition, Glu substituted at positions Ϫ3Ј or Ϫ4Ј of the ␤ subunit of a heteromeric GABA A ␣ 2 ␤ 3 ␥ 2 receptor imparted permeability to cations along with anions (32). Here, concomitant replacement of the M1-M2 loop and the residues at positions 0Ј and 2Ј of the GluCl␤R's pore by those of ␣7 resulted in a fully cationic channel (chimera 3, Table 1 and supplemental Fig. S2B). Replacing only the M1-M2 loop also produced a fully cationic channel, indicating that a hydroxyl group at position 2Ј does not play a role in ion charge selectivity (chimera 4, Table 1 and Figs. 2E and 3B). The latter conclusion well agrees with the observations that following non-polar substitutions at position 2Ј, the muscle AChR retains permeability to monovalent cations that are considered to interact with this position while traversing the pore (28). Cationic selectivity was also observed when chimera 4 was further modified so as either to carry the M1-M2 sequence of the cationic 5HT 3A R (chimera 5, Table 1 and supplemental Fig. S2C) or to carry a neutral residue at position Ϫ7Ј (chimera 6, Table  1, Fig. 3B, and supplemental Fig. S2D). The latter modification indicates that the so-called cytoplasmic ring (DϪ7Ј) does not play a role in ion charge selectivity. As long as a Glu residue occupied position Ϫ1Ј, further gradual changes in the sequence of the M1-M2 loop toward the sequence of the GluCl␤R (chimeras 7 and 8) did not convert the mutants back to anionic receptors, but they retained cationic selectivity, even when the GluCl␤R's pore was carrying a single GϪ1ЈE substitution (chimera 9) ( Table 1, Figs. 2F and 3B, and supplemental Fig. S2E and S2F). It should, however, be noted that in three cases, a slight decrease in the relative permeability to Na ϩ (i.e. increase in P Cl /P Na ) was observed (chimeras 6, 7, and 9, Table 1).
In contrast, integrating the residues of the M1-M2 loop of ␣7 within the sequence of the GluCl␤R's loop while keeping a Gly at position Ϫ1Ј (i.e. the native amino acid of the GluCl␤R) provided an anionic channel (chimera 10, Table 1 and supplemental Fig. S2G). Deleting the GluCl␤R residues at positions Ϫ6Ј and Ϫ5Ј (LH) slightly reduced the P Cl /P Na and P Cl /P Cs ratios, but these mutants still displayed considerable capacity to select Cl Ϫ over Na ϩ and Cs ϩ (chimeras 11 and 12, Table 1, Fig. 3A, and supplemental Fig. S2H and S2I). Further deletion of the Thr that precedes position Ϫ2Ј in the GluCl␤R, alone (chimera 13) or together with neutralization of the charge either at position Ϫ7Ј (chimera 14) or at position 0Ј (chimera 15), resulted in channels that retained high Cl Ϫ over Na ϩ selectivity but became slightly permeable to Cs ϩ (Table 1, Figs. 2D and 3A, and supplemental Fig. S2J and S2K). Shortening the M1-M2 loop to four amino acids rendered the chimeric receptor nonfunctional (chimera 16, Table 1). Proline at position Ϫ2 was shown to play a role in determining the (small) pore diameter of a glycine receptor, indicating that pore size also contributes to ion charge selectivity (Ref. 17, reviewed by Keramidas et al. (9)). Interestingly, an ␣7-GluCl␤R having a proline at position Ϫ2Ј (chimera 17) displays P Cl /P Na and P Cl / P Cs ratios closely similar to those of the ␣7-GluCl␤R (Table 1 and supplemental Fig. S3).
Relative Permeability of Chloride-selective Chimeras to Isethionate and Acetate-To assess the dimensions of the open pore of the ␣7-GluCl␤R and anionic mutants, we examined the permeability to the organic anions isethionate and acetate relatively to the permeability for Cl Ϫ . Table 2 shows that the ␣7-GluCl␤R is not permeable to isethionate but allows acetate to permeate slightly. Elimination of the side chain at position 0Ј (chimera 2) or deleting residues belonging to the M1-M2 loop of the GluCl␤R (chimeras 12-15) turned these mutants permeable to isethionate and increased the relative permeability to acetate ( Table  2). The increase in the relative permeability to isethionate linearly correlated with the increase in the relative permeability to acetate (Fig. 4), indicating that these mutations have effectively widened the open pore. Notably, the ␣7-GluCl␤R-AϪ2ЈP mutant (chimera 17) was found to be as permeable to acetate as the ␣7-GluCl␤R (P Acet /P Cl ϭ 0.1 Ϯ 0.013, mean Ϯ S.E. from 5 cells; e.g. supplemental Fig. S3A).

DISCUSSION
The Location of the Selectivity Filter in the GluCl␤R's Channel-Previous studies on the homomeric ACh-␣7, Gly␣1, and 5HT 3 receptors had shown that concomitant substitutions at positions Ϫ2Ј, Ϫ1Ј, and  Table 1 under column N. Note that in all cases (panels A and B and all the other chimeras), replacement of external NaCl was performed by mannitol solutions containing decreasing concentrations of NaCl (see "Experimental Procedures").

Ion Charge Selectivity in an Anionic Cys-loop Receptor
13Ј converted their ion charge selectivity (10 -14). In further experiments, an AϪ1ЈE substitution introduced concomitantly with a deletion of PϪ2Ј in the homomeric anionic Gly␣1R, GABA A 1R, and MOD-1 (an invertebrate Cl Ϫ -selective 5HT 3 R) led to predominant cationic permeability accompanied with some permeability to Cl Ϫ (P Cl /P Na ϭ 0.13, ϳ0.37, and small shifts in E rev values, respectively) (13,15,16). Similar changes in ion charge selectivity were observed also when an AϪ1ЈE substitution was introduced together with a deletion of AϪ2Ј in the ␤ subunit of an ␣ 2 ␤ 3 ␥ 2 heteromeric GABA A R (P Cl /P Na ϭ 0.4 and P Cl /P K ϭ ϳ0.26) (32,33). A single AϪ1ЈE substitution in the Gly␣1R, GABA A 1R, and MOD-1 imparted a large component of permeability to Na ϩ , but the mutants still allowed Cl Ϫ to permeate (P Cl /P Na ϭ 0.34, ϳ1.4, and small shifts in E rev values, respectively) (13,15,16).
Here, a single GϪ1ЈE substitution in the GluCl␤R's pore (chimera 9) converted the ion charge selectivity almost completely (P Cl /P Na ϭ 0.08 and P Cl /P Cs ϭ 0.06). The finding that the permeability to Cl Ϫ did not exceed 9% of the permeability to Na ϩ (in chimera 9) indicates that mutations that convert the ion charge selectivity unlikely induce drastic structural changes in the selectivity filter of the GluCl␤R's pore. Since additional mutations toward the cationic receptors' sequence further reduced the P Cl /P Na and P Cl /P Cs ratios down to 0.02 and 0.01, respectively (chimeras 8, 5, and 4), we infer as follows. In wild-type cationic Cys-loop receptors, the M1-M2 loop shapes the selectivity filter so as to optimally orient the carboxyl moieties of position(s) Ϫ1Ј in a mode that leads to (i) effective interactions with Na ϩ ions, (ii) repulsion of Cl Ϫ ions, and (iii) generation of a negative electrostatic potential close to the bottom of the M2 segments (34), so as to counterbalance the positive dipoles of the M2 helices.
Since replacement of the M1-M2 loop of the GluCl␤R by that of the ␣7-AChR or the 5HT 3A R is sufficient to completely convert the selectivity from highly anionic to highly cationic selectivity (chimeras 4 and 5), we conclude that the apparatus acting as the selectivity filter of the GluCl␤R is formed by a sequence belonging to the M1-M2 loop. Based on the mild effect seen with chimera 1, we, however, do not exclude some contribution of charged residues in the outer wide vestibule to ionic permeation by affecting ionic movements and the local concentration of ions.
Positively Charged Side Chains Are Not Fundamental for Cl Ϫ Selection by the GluCl␤R-The capacity of cationic Cys-loop receptors (wild types and mutants) to select cations over anions on the basis of counter charges in the selectivity filter raises the question of whether (or not) an inverse mechanism of ion charge selectivity takes place in anionic Cysloop receptors. It was previously suggested that the residues composing the M1-M2 linker orient differently when the ion charge selectivity of the receptor is changed following mutations in this segment. It was further proposed that, unlike the case of cationic Cys-loop receptors, the conserved basic side chain of position 0Ј of the anionic Gly␣1 receptor points to the pore lumen to form a positively charged ring that interacts with the passing Cl Ϫ ion and, on the other hand, repulses cations (9,15). Here, a single R0ЈG mutation in the GluCl␤R's pore reduced the P Cl /P Na ratio only in a moderate extent, whereas the mutant retained high capability to select Cl Ϫ over Na ϩ and Cs ϩ (chimera 2). This observation indicates that positive charge at position 0Ј does not play a key role in Cl Ϫ selectivity by the GluCl␤R. Although one may reasonably argue that the mild reduction in the P Cl /P Na ratio (chimera 2) does indicate some positive-charge contribution by R0Ј, we assume that R0Ј of the GluCl␤R has a role in stabilizing the shape of the selectivity filter. Computer-assisted molecular modeling indicates that the long side chain of R0Ј forms multiple van der Waals interactions with HϪ5Ј, LϪ6Ј, and IϪ10Ј, on the other side of the M1-M2 loop (Fig. 5C). Elimination of this bond network might therefore destabilize the architecture of the M1-M2 loop and cause local structural changes in the main chain, also around the C␣ atom of R0Ј.
Surprisingly, a His residue at position Ϫ5Ј of the GluCl␤R also cannot play a key role in Cl Ϫ selectivity since chimeras 11-15, which lack this histidine, remained anionic. These observations raise the question of whether one basic side chain could compensate for the absence of the FIGURE 4. The relative permeability to isethionate plotted as a function of the relative permeability to acetate. Each data point represents an individual chimera as follows: black triangle, ␣7-GluCl␤R; inverted blue triangle, chimera 2; magenta square, chimera 12; green circle, chimera 13; gray square, chimera 14; red diamond, chimera 15. Data points were fitted with a linear regression to the equation: P Acet /P Cl ϭ a ϫ P Ise /P Cl ϩ b, where a is the slope, and b is the intercept with the y axis. Error bars correspond to the S.E. values provided in Table 2. loop (and its flanking residues, R0Ј and IϪ10Ј) organized in a 5-fold symmetry around the axis of ion conduction. Chloride is shown in magenta, with an equatorially bound water molecule. Note that the side chain of HϪ5Ј cannot be in contact with the permeating Cl Ϫ ions (see a notion under "Experimental Procedures"). C, space-filling model that corresponds to a side view of two facing segments; each includes the GluCl␤R's M1-M2 loop and its flanking residues, R0Ј and IϪ10Ј. Chloride is shown as in panel A. A stick model of a third subunit is shown (without hydrogen atoms) at the back. The distances between the van der Waals surfaces of two facing backbone amides at the level of AϪ2Ј and TϪ3Ј are 6.52 and 6.07 Å, respectively. D, a model as in panel C but of chimera 17, which has a proline at position Ϫ2Ј. Chloride, as in panel A. The two long black arrows indicate the spheres of hydrogen atoms bonded to the C␥ atoms of two facing prolines. The distance between the van der Waals surfaces of the closest facing hydrogen atoms of two facing prolines is 7.32 Å. The distance between the backbone amides at the level of TϪ3Ј is 6.1 Å. In panels B-D: (i) the green, blue, red, and grayish spheres correspond to carbon, nitrogen, oxygen, and hydrogen atoms, respectively; (ii) the black arrows show side chains, and the gray arrows show backbone amides; (iii) the Cl⅐⅐⅐H-O(water) hydrogen bond used here is 2.89 Å (taken from Protein Data Bank number 102L), giving rise to a 6.1-Å long Cl Ϫ -water complex. Note that additional water molecules can potentially be in contact with Cl Ϫ , particularly above and below the ion along the axis of ion conduction. Also note that the hydrogen bond between a Cl Ϫ ion and water can be as short as 2.46 Å (PDB number 1C4D) so, taken together with a possible N-H⅐⅐⅐Cl hydrogen bonding, Cl Ϫ can be accommodated with an equatorially bound water molecule in anionic receptors having smaller open pore diameter.
other basic residue owing to reorganization of the M1-M2 loop, which might take place when mutating the loop. In the current case, the concomitant elimination of both basic side chains (positions Ϫ5Ј and 0Ј, chimera 15) did not considerably change the capacity of the receptor to select Cl Ϫ over Na ϩ , albeit with moderately reduced P Cl /P Na ratio and more profound decrease in the P Cl /P Cs ratio. These observations indicate that positively charged side chains do not play a pivotal role in Cl Ϫ selectivity by the GluCl␤R. It further implies that the imidazole moiety of HϪ5Ј in the GluCl␤R's pore either is uncharged or, as predicted by the structural model (Fig. 5C), cannot be in contact with the permeating Cl Ϫ ions.
How Does the GluCl␤R Pore Select Cl Ϫ over Na ϩ ?-The permeability of a channel to an ion reflects the ease of an ion to enter and pass through the selectivity filter. Permeability ratios provide an estimate of the difference between the hydration energy in water and the solvation energy provided by the selectivity filter. If direct interactions of Cl Ϫ with the selectivity filter of the GluCl␤R do not involve positively charged residues (as discussed above), then what is the mechanism of Cl Ϫ selectivity by the GluCl␤R? Potential interactions between a Cl Ϫ ion and the hydroxyls at positions Ϫ3Ј are also excluded as ThrϪ3Ј is deleted in chimeras 13-15, which remain selective to Cl Ϫ . The side chain of SerϪ4Ј is likely not involved since it is common to cationic Cys-loop receptors as well.
The determination of relative permeabilities to isethionate and acetate reveals that, although the GluCl␤R's pore is not permeable to isethionate, it is slightly permeable to acetate (ϳ10% of the Cl Ϫ permeability level). Taking into account the second largest dimension of these two organic anions (Table 2, footnotes), it is reasonable to conclude that the effective open pore diameter of the GluCl␤R is larger than 5.18 Å and smaller than 6.2 Å. These dimensions fall within the range of the effective open pore diameter determined previously for other anionic Cys-loop receptors such as the Gly-and GABA-activated channels (5.2-6.1 Å) (35)(36)(37)(38). Given that the Cl⅐⅐⅐H-O(water) hydrogen bond can be as short as ϳ2.5 Å, Cl Ϫ can snugly fit in the selectivity filter of the GluCl␤R with one equatorially bound water molecule, as predicted by the structural model (Fig. 5). As such, ion-dipole interactions can potentially take place between a Cl Ϫ ion and the backbone amides of positions Ϫ2Ј (but see below) and Ϫ3Ј. In contrast, a Na ϩ ion would stay outside the selectivity filter where its energy is lower because: (i) even with an equatorially bound water molecule, it is much smaller than the narrowest part of GluCl␤R's open pore, and (ii) it would not interact with the positive dipoles of the backbone amides. Mutations that widen the pore (chimeras 13-15) probably introduce slight conformational change in the M1-M2 loop, thereby moving the backbone amides of positions Ϫ2Ј and Ϫ3Ј sideways. As such, the progressive widening of the pore results in increasing leak of Cs ϩ through the pore, whereas the permeability to Na ϩ remains effectively zero (chimeras 2 and 12 versus chimeras 13-15).
Our observations that a proline substitution at position Ϫ2Ј (chimera 17) does not essentially change the ion charge selectivity or permeability to acetate give rise to one of the following possibilities. (i) Proline does not change the conformation of the main chain in this mutant, as in the textbook case of a T4-lysozyme-A82P mutant (39), or (ii) proline induces a change in the backbone conformation topologically above position Ϫ2Ј. In any event, proline at position Ϫ2Ј prevents the backbone nitrogen of this position to serve as a hydrogen bond donor or to contribute to Cl Ϫ -dipole interactions. Conclusively, the backbone amides of position(s) Ϫ3Ј of the GluCl␤R actually serve as the predominant sites for partial dehydration of a permeating Cl Ϫ ion. In line with the experimental results, structural modeling reveals that an AϪ2ЈP mutation in the GluCl␤R's pore does not essentially change the backbone conformation and does not impede the interaction of Cl Ϫ with backbone amides at level Ϫ3Ј (Fig. 5D).
It should be noted that the dipole moment of the M2 helices gives partial positive charge at the helices' amino ends. Taken together, the mechanism underlying Cl Ϫ selectivity in the GluCl␤R involves a narrow open pore diameter, attraction of Cl Ϫ by the partial positive charge of the M2 N termini, and the interactions of Cl Ϫ with backbone amides, where the Cl Ϫ ion undergoes partial dehydration. This mechanism might apply to other Cl Ϫ -selective ion channels. Indeed, x-ray crystallography studies have recently shown that dehydrated Cl Ϫ ions in the pore of a bacterial ClC channel are stabilized by electrostatic interactions with helix dipoles and by partial charges from backbone amide groups and side-chain hydroxyl groups (40).
The Selectivity Filter Overlaps the Activation Gate-Previous studies by Karlin and colleagues (30,34) and by Paas et al. (20) have located the activation gate of two different Cys-loop receptors in a constricted part close to the bottom of the pore. Paas et al. (20) further probed gating motions in a Cys-loop chimeric receptor by monitoring the state-dependent accessibility of Zn 2ϩ ions to histidines introduced along the channel pore and alignment of the experimental results with computerassisted molecular models. Consequently, it has been concluded that channel gating predominantly involves rigid tilting motions of the M2 segments, which widen (open) or narrow (close) the bottom pore constriction (20), where ionic selectivity takes place as well (current and the aforementioned other studies). Such a structural overlap implies that components of the activation gate act as the selectivity filter upon channel opening.