Role of Charged Residues in Coupling Ligand Binding and Channel Activation in the Extracellular Domain of the Glycine Receptor*

The glycine receptor is a member of the ligand-gated ion channel receptor superfamily that mediates fast synaptic transmission in the brainstem and spinal cord. Following ligand binding, the receptor undergoes a conformational change that is conveyed to the transmembrane regions of the receptor resulting in the opening of the channel pore. Using the acetylcholine-binding protein structure as a template, we modeled the extracellular domain of the glycine receptor α1-subunit and identified the location of charged residues within loops 2 and 7 (the conserved Cys-loop). These loops have been postulated to interact with the M2-M3 linker region between the transmembrane domains 2 and 3 as part of the receptor activation mechanism. Charged residues were substituted with cysteine, resulting in a shift in the concentration-response curves to the right in each case. Covalent modification with 2-(trimethylammonium) ethyl methanethiosulfonate was demonstrated only for K143C, which was more accessible in the open state than the closed state, and resulted in a shift in the EC50 toward wild-type values. Charge reversal mutations (E53K, D57K, and D148K) also impaired channel activation, as inferred from increases in EC50 values and the conversion of taurine from an agonist to an antagonist in E53K and D57K. Thus, each of the residues Glu-53, Asp-57, Lys-143, and Asp-148 are implicated in channel gating. However, the double reverse charge mutations E53K:K276E, D57K:K276E, and D148K:K276E did not restore glycine receptor function. These results indicate that loops 2 and 7 in the extracellular domain play an important role in the mechanism of activation of the glycine receptor although not by a direct electrostatic mechanism.

Fast synaptic transmission in the central nervous system is mediated by members of the ligand-gated ion channel (LGIC) 1 receptor superfamily. The glycine receptor (GlyR) is a member of the nicotinic-like LGIC superfamily that includes the nicotinic acetylcholine (nAChR), serotonin type 3 (5-HT 3 R), and ␥-aminobutyric acid (GABA A R) receptors (1,2). Each of these receptors are pentameric complexes arranged around a central ion conducting pore. Individual subunits share a similar membrane topology, with hydropathy analysis predicting a large extracellular domain at the N terminus and four putative transmembrane domains (M1-M4) (1). A key characteristic of these receptors is the integral ion channel that is opened following ligand binding. The extracellular domain contains the ligand binding site, which is spatially separate from the M2 domain that lines the ion channel pore of these receptors (3,4). While there is an accumulated body of data on the structures involved in ligand binding (2) and the ion channel pore (2)(3)(4), relatively little is known about the structures that may link these two spatially separate domains to effect receptor activation.
Inherited mutations located in the regions flanking the M2 domain of the GlyR ␣1-subunit are associated with human startle disease (hyperekplexia) (5) and startle syndromes in other species (6). In humans, the startle mutations R271L/Q and K276E map to the M2-M3 linker domain (5,7). At both locations, the mutations result in the substitution of a positive charged residue. The effect of these mutations is to disrupt signal transduction, as inferred from the increases in EC 50 for glycine and the conversion of taurine from an agonist to an antagonist (8). This was also demonstrated at the single channel level for the K276E mutation (9). An inherited mutation in the M2-M3 region of the nAChR is associated with a form of congenital myasthenic syndrome (10) and a mutation in the GABA A R is associated with a rare form of epilepsy (11), both of which disrupt channel gating. In addition, site-directed mutagenesis and covalent modification studies have demonstrated that the M2-M3 region undergoes a conformational change that is associated with activation of the GlyR (12,13). Similar evidence of conformational change of the M2-M3 domain has been found for the GABA A R (14). We sought to identify structures within the extracellular domain of the human GlyR that are involved in the conformational change that conveys the ligand binding event to opening the channel pore.
The amino acid sequence of the acetylcholine-binding protein (AChBP) from the snail Lymnaea stagnalis, was found to have a strong (15-24%) homology with the extracellular domains of LGIC subunits (15). The crystal structure of this protein provides a template with which to model these extracellular domains (15). Loop 7 of the AChBP corresponds with the conserved Cys-loop of LGICs, which is essential for a functional receptor but not involved in ligand binding (16,17). The location of loop 7 in the AChBP suggests that the conserved Cysloop of LGIC receptors is in a position to interact with the transmembrane domains of the receptor, and so may be involved in gating (15,18). Loop 2 is also a flexible loop, which in a LGIC receptor subunit is proposed to be in a position to interact with the transmembrane domains. Alignment of the AChBP and GlyR ␣1-subunit sequences shows that loop 2 corresponds to the location of the A52S mutation in spasmodic mice (19,20). The effect of this mutation is to impair the GlyR activation (19,20), resulting in an exaggerated startle phenotype, which suggests that loop 2 might also be involved in gating of the GlyR.
As charged residues and conformational changes in the M2-M3 linker are associated with the activation of the ion channel, we sought to examine the role of charged residues in loops 2 and 7 in the extracellular domain of the GlyR ␣1subunit. Using a combination of cysteine accessibility techniques and mutations that reversed the charge of these residues, we evaluated the role of loops 2 and 7 in the activation of the GlyR. We demonstrate that residues in these loops play an important role in receptor activation but not by a simple electrostatic interaction as has been observed in the GABA A R (21).

EXPERIMENTAL PROCEDURES
Molecular Modeling-Amino acid residues 31-248 of the ␣1-subunit of the human GlyR (Swiss Prot accession P23415) were aligned with residues 1-206 of the AChBP (Protein Data Bank accession 1I9B) according to the alignment of Brejc (15). Pentameric ␣1-subunit models of the GlyR were built in a single run of the program Modeler (22), and the structures were then refined using the Refine_1Ј function of Modeler.
Mutagenesis and Expression of Human GlyR ␣1-Subunit cDNA-The cDNA encoding the wild-type human ␣1-subunit of the GlyR was subcloned into the pCIS expression vector. Note that the ␣1-homopentamer was expressed, not the 3␣:2␤ native heteropentamer, in this study. Site-directed mutagenesis was performed using the oligonucleotidedirected polymerase chain reaction (PCR) technique and confirmed by DNA sequencing of the complete plasmid. To ensure that the sulfhydryl reagents only reacted with the substituted cysteine residues, we used receptor subunits in which the free cysteine residue at position 41 in the extracellular domain had been mutated to an alanine (13). All of the substituted cysteine mutations were introduced on this C41A background. The mutation of acidic and basic residues in the Cys-loop and loop 2 of the extracellular domain, and in the M2-M3 linker domain were introduced on the wild-type GlyR ␣1-subunit.
Electrophysiology-Whole cell patch clamp experiments were performed at room temperature (22.5 Ϯ 0.3°C). The cells were continually superfused with an external bathing solution containing:140 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM HEPES, and 10 mM glucose, adjusted to pH 7.4 with 1 M NaOH. Glycine and taurine concentrations were made up in the external bathing solution and were applied directly to the cells with a modified U-tube. Patch pipettes were fabricated from borosilicate glass capillaries and fire polished to give a tip resistance of 2-6 M⍀. Pipettes were filled with an intracellular solution containing: 120 mM CsCl, 20 mM TEACl, 1 mM CaCl 2 , 2 mM MgCl 2 , 10 mM HEPES, and 11 mM EGTA adjusted to pH 7.2 with CsOH. Whole cell currents were recorded at a holding potential of Ϫ50 mV using an Axopatch-1D amplifier and digitised using pClamp 6.0 software and a Digidata 1200 ADC interface (Axon Instruments, Foster City, CA). At least 80% series resistance compensation was achieved in each experiment. Concentration response curves were constructed from the peak current response to the application of a range of agonist concentrations, with a minimum  of 1 min between successive applications. Inhibition curves were constructed from the peak current response to the co-application of a range of taurine concentrations and a fixed concentration of glycine, with a minimum of 1 min between successive applications.
Sulfhydryl Reagents and Reactions-We used the charged methane thiosulfonate (MTS) derivatives, 2-sulfonatoethyl methanethiosulfonate (MTSES), and 2-(trimethylammonium)ethyl methanethiosulfonate (MTSET) (Toronto Research Chemicals Inc.). Stock solutions of 100 mM MTSES and MTSET in distilled water were aliquoted into screw cap microcentrifuge tubes and rapidly frozen in an ethanol/dry ice mix before storage at Ϫ20°C. For each application of MTS reagents, a new aliquot was thawed, diluted in bathing solution to the working concentration and used immediately.
The rate of covalent modification of cysteine substituted GlyRs by MTS reagents was measured in the presence and absence of glycine. The peak amplitude of the current in response to an EC 50 concentration of glycine (I 50 ) and a maximum concentration of glycine (I max ) were recorded at least twice prior to MTS reagent addition. For reactions in the open state, 100 M MTSET or 100 M MTSES was co-applied with 2 mM glycine. The change in response to an EC 50 concentration of glycine (relative ⌬I 50 ) was expressed relative to the maximum response according to Equation 1, where I 50 (0) is the current response to an EC 50 concentration of glycine before MTS addition, I 50 (t) is the current response for each cumulative time point, t, of the MTS reaction and I max is the maximum current response. The data expressed in this way were fitted with a single exponential decay (Origin, Microcal Software, Northampton, MA) to obtain an estimate of the first order rate constant. The averaged first order rate constant was determined from at least 3 cells in each case. Data Analysis and Statistics-Concentration response data and inhibition data were plotted on semilogarithmic axes and fitted using a non-linear least squares routine (Origin, Microcal Software, Northampton, MA) with the empirical Hill Equation 2, where I is the peak whole cell current recorded following application of a range of concentrations of the agonist, [A]; I max is the estimated maximum current, EC 50 is the glycine concentration required for a half-maximum response and n H is the Hill co-efficient. Statistics were performed on wild-type and mutant EC 50 values and Hill coefficients using a one-way ANOVA with Fisher's post-hoc test.

RESULTS
Modeling of the GlyR on the AChBP-By threading the extracellular domain of the GlyR sequence on the AChBP crystal structure we were able to identify those residues of the GlyR that are located within loop 2 and loop 7, which may be located close to the M2-M3 linker or the ion channel itself (Fig. 1). Amino acid residues with a charged side chain that were located in these loops were chosen for study, namely Glu-53 and Asp-57 in loop 2, and Lys-143 and Asp-148 in loop 7. The residues known to be involved in ligand binding are located distant (25-40 Å) from this site in the receptor (15).
Effects of Mutations to Cysteine-All of the cysteine mutants were constructed on an ␣1-subunit C41A GlyR background to eliminate the possibility of MTS reagents reacting with the free  cysteine. This receptor showed no alteration in the glycine sensitivity (EC 50 ϭ 0.027 Ϯ 0.004 mM, Table I) or any change after 500 M MTSES or 500 M MTSET were added for 1 min (Fig. 2).
To assess the effect of replacing a charged residue with a cysteine, concentration response curves to glycine were measured for the mutant receptors E53C and D57C in loop 2, and K143C and D148C in loop 7. Each of these receptors showed an increase in EC 50 , with K143C showing the smallest shift (EC 50 ϭ 0.28 Ϯ 0.05 mM, 10-fold increase) and D148C showing the largest shift (EC 50 ϭ 0.97 Ϯ 0.03 mM, 36-fold increase), compared with the C41A receptor (Table I).
Effects of MTS Addition to Substituted Cysteine Mutants-The covalent modification of cysteine residues enables the introduction of the positive charged trimethyl ammonium group from MTSET or the negative charged sulfonate group from MTSES. For each of the substituted cysteine mutants, this provides a mechanism for reintroducing a charged side chain similar to that of the original charged amino acid. However, this depends upon the substituted cysteine being available for modification. By reacting the positive charged MTSET to the K143C GlyR and the negative charged MTSES to the E53C, D57C, and D148C GlyRs, we could determine whether these cysteine residues were accessible to MTS reagents, as inferred from any detectable shifts in the concentration response curve. When 500 M MTSES was applied for 1 min to the E53C, D57C, or D148C in the presence or absence of glycine, there was no alteration in the amplitude of the I 50 or I max (Fig. 2). This suggests either the cysteine was not accessible to MTSES for modification, or the sulfonate group of the MTSES had no effect on the response of the mutant receptors.
When 500 M MTSET was applied to the K143C in the presence or absence of glycine, there was no change in the I max but a shift in the I 50 such that it was now ϳ80% of the I max (Fig.  2). A complete concentration-response curve showed that following covalent modification the EC 50 had shifted from 242 M to a value of 70.3 M, which is closer to that of the C41A GlyR. This indicates that the K143C receptor was accessible for modification by MTSET and the addition of a positive charge partially restored the function of the receptor.  Table II for the single mutations  and Table III for the double mutations. I 50 and I max responses, were repeated until the reaction was complete. For the closed state (Fig. 3), the I 50 and I max responses were measured after each 3 s application of 100 M MTSET in the absence of glycine, until the reaction was complete. The change in I 50 currents were expressed relative to the I max (relative ⌬I 50 ) and plotted against the cumulative reaction time (Fig. 3C). First order rate constants were estimated from the fit of a single exponential decay to the data.  Table II). In contrast, the K143E GlyR showed no change in glycine (EC 50 0.028 ϩ 0.002 mM) or taurine (EC 50 0.09 ϩ 0.01 mM) activation when compared with the wild-type GlyR. Application of taurine to the E53K and D148K receptors at concentrations up to 50 mM resulted in no detectable whole cell currents, while taurine applied to D57K elicited maximum currents that were only ϳ10% of those elicited by glycine (Table  II). Co-application of taurine in the presence of 5 mM glycine demonstrated that taurine inhibited the glycine current of the E53K and D57K GlyRs. Taurine inhibition curves were performed for the E53K and D57K GlyRs and the obtained IC 50 values were 1.78 Ϯ 0.12 mM and 12.1 Ϯ 1.5 mM respectively ( Fig. 4; Table II). The whole cell currents elicited from the D148K receptor were too small (I max Ͻ15 pA) to gain a reliable inhibition curve, but a 40 Ϯ 5% reduction in current at 5 mM glycine was seen with co-application of 50 mM taurine (data not shown). Conservative mutations E53D, D57E, and K143R all showed no change from the wild-type receptor (E53D, EC 50 ϭ 0.021 Ϯ 0.003 mM; Table II) or a slight decrease in EC 50 concentration (D57E, EC 50 ϭ 0.008 Ϯ 0.0004 mM; K143R, EC 50 ϭ 0.007 Ϯ 0.002 mM; Table II). From these results, we would infer that channel gating is affected by the charge reversal mutations E53K and D57K in loop2, and D148K in loop 7 of the GlyR.

Rates of Reaction for K143C GlyR-To
Effects of Double Charge Reversal Mutations-Charged residues located in the M2-M3 linker have previously been implicated in channel gating (12,13) and this study has demonstrated the role of loop 2 and loop 7 charged residues in gating. Therefore, we tested for electrostatic interactions between the negatively charged residues of loops 2 and 7, and the positively charged Lys-276 residue of the M2-M3 linker by creating the double reverse charge mutations E53K:K276E, D57K:K276E, and D148K:K276E. None of these doubly mutated GlyRs restored the function of these receptors to that of the wild-type, instead the EC 50 values remained similar to that of the K276E receptor ( Fig. 5; Table III). Thus, there does not appear to be any direct electrostatic interactions that influence channel gating between these residues in loops 2 and 7, and Lys-276 in the M2-M3 linker.
A further hypothesis is that residues in loops 2 and 7 may interact with each other. We tested for interactions between negatively charged residues in loop 2 and Lys-143 of loop 7 by creating the double reverse charge mutations E53K:K143E and D57K:K143E. The EC 50 of the E53K:K143E mutant was similar to the E53K receptor, and the D57K:K143E receptor had shifted slightly to the right of the D57K receptor ( Fig. 6; Table  III). Therefore, electrostatic interactions between loops 2 and 7 of the extracellular domain do not appear to influence gating.

DISCUSSION
Modeling of the GlyR-By threading the amino acid sequence of the GlyR ␣1-subunit extracellular domain over the AChBP structure, we were able to identify the location of residues that may be involved in various functions of the receptor. In this model, the spasmodic loop and the conserved Cys-loop of the GlyR correspond to loop 2 and loop 7, respectively, of the AChBP structure. These are located in regions that are postulated to be close to the ion channel (15,18) and to the M2-M3 linker (Fig. 1). We propose that conformational changes occur in the spasmodic loop and Cys-loop of the GlyR upon ligand binding that are an important part of the process linking ligand binding to channel gating. This is consistent with the previous demonstration that the spasmodic mutation results in impaired receptor gating but does not affect ligand binding (19,20), and the M2-M3 linker is involved in channel gating (12)(13)(14).
Conformational Changes That Mediate Channel Activation following Ligand Binding-By using MTS modification of cysteine residues, it is possible to identify residues that are either altered in their own conformation, or else the environment around them is altered by the conformational changes of the receptor. The cysteine-substituted mutant K143C was able to be modified by MTSET, resulting in a decrease in the EC 50 value. By showing that the rate of this reaction was faster in the open state than the closed state, we demonstrate that this residue is more accessible to modification in the open state. As accessibility is dependent upon the environment surrounding the sulfur atom of the cysteine, these results suggests that the introduced cysteine is in a more hydrophilic environment when the receptor is in the open state compared with the conformation adopted in the closed state (23).
For each of the E53C, D57C, and D148C GlyR mutations, the receptor did not show any change in glycine activation following the addition of MTS reagents. This could be the result of one of several possibilities. The structure of the receptor may have been altered by the cysteine residue such that the introduced cysteine is not in the same position as the original residues and is now in a hydrophobic environment; the residues may be buried in a pocket that the MTS reagent cannot access, despite actually being surrounded by a hydrophilic en- vironment; or the MTS reagents may react but not change the function of the receptor. We consider the latter possibility unlikely, considering that removal of charge on these residues changes so drastically the function of the receptor. Finally, the residues may shift from a hydrophobic to a hydrophilic area during the signal transduction process, but the length of time that the residues are accessible for modification is so short, because of the altered response of the receptor to glycine, that the MTS reactions are not detectable.
Role of Charged Residues in Loops 2 and 7 of the Extracellular Domain-Both the E53K and D57K mutations resulted in an increased EC 50 and the agonist taurine was converted to an antagonist. We infer from these results that these two loop 2 charged residues (Glu-53 and Asp-57) are involved in channel gating or the signal transduction process leading to channel gating. The conversion of taurine from an agonist to a partial agonist or an antagonist has previously been demonstrated by mutations within the M2-M3 linker, establishing this region as being involved in channel gating (8). The two charged residues in loop 7 (Lys-143 and Asp-148) were also shown to be involved in channel gating. The D148K mutation, similar to E53K, resulted in an increased EC 50 and the conversion of the agonist taurine to an antagonist. While the K143E receptor was not different to the wild-type receptor, the K143C mutation did increase the EC 50 . As MTSET reacted faster in the open state, we can infer this residue is not involved in ligand binding, as the MTSET would otherwise be competing with glycine to react at the introduced cysteine and thus would be expected to react faster in the closed state. This is the first study to show the role of charged residues in both loop 2 and loop 7 of the extracellular domain of the GlyR in channel gating.
The reverse charge E53K mutation shifts the concentrationresponse curve (46-fold increase) to a similar degree as the E53C mutation (33-fold increase). This is not the case for the Asp-57 residue, where the D57K GlyR has a smaller shift in EC 50 (9-fold increase) compared with the D57C GlyR (42-fold increase). The Lys-143 residue is similarly intriguing, with virtually no change in the EC 50 for K143E GlyR, but a 10-fold shift in the EC 50 for the K143C GlyR, compared with wild type. When the loop 7 sequence is aligned with that of other LGICs (Fig. 7), the Lys-143 residue aligns with a positively charged arginine residue in the nAChR ␣7-subunit, however it is a negatively charged glutamic acid residue in the GABA A R ␣1subunit. Therefore, at position 143 of the GlyR the presence of a charge rather than the valency of the charge may be important for the function of the receptor. In comparison, the negatively charged glutamic acid residue at position 53 appears to be conserved as an acidic residue across ␣-subunits, as is the aspartic acid residue at position 148 (Fig. 7).
Charged Residues in Loops 2 and 7 of the GABA A R-During the course of this study, it was shown that charged residues in loops 2 and 7 of the GABA A R ␣1-subunit interact with residues in the M2-M3 loop to activate channel gating (21). Kash et al. (21) show that three negatively charged residues (Asp-57, Glu-59, and Asp-149) in these loops cause an increase in the EC 50  Table II for the single mutations and Table III for the double mutations. FIG. 7. Sequence alignment of loops 2 and 7 of the extracellular domain of selected subunits of LGIC receptors. The human sequences for the LGIC subunits shown were aligned to that of the human GlyR ␣1-subunit. Those residues that were investigated in this study (Glu-53, Asp-57, Lys-143, and Asp-148) are shown in circles. The location of the spasmodic mutation (A52S) in loop 2 is shown in bold. The charged residues in the GABA A R ␣1-subunit that were investigated by Kash et al. (21) are boxed. Conserved residues are marked with an asterisk, including the cysteines that form a disulfide bond in the Cys-loop (loop 7). for GABA when the charge is reversed. A sequence alignment of the GlyR ␣1and GABA A R ␣1-subunits indicates that the residues in the GABA A R ␣1-subunit involved in gating interactions align with several of the residues chosen for study in the GlyR (Fig. 7). Using mutant cycle analysis of double reverse charge mutations, Kash et al. (21) were able to demonstrate direct electrostatic interactions between both residues D57 (loop 2) and Asp-149 (loop 7) and the Lys-279 residue in the M2-M3 linker of the GABA A R ␣1-subunit. However, double reverse charge mutations of the Glu-53, Asp-57, and Asp-148 residues and the Lys-143 and Lys-276 residues of the GlyR failed to show any evidence for direct electrostatic interactions. This suggests that while charged residues are clearly involved in the gating process they are not acting through direct electrostatic interactions between these particular pairs of residues in the GlyR.
Implications for Members of the LGIC Receptor Superfamily-This study has confirmed the importance of charged residues in loops 2 and 7 of the extracellular domain of the GlyR in the signal transduction process that links the ligand binding event to channel opening. Unlike the GABA A R (21), there is at present no evidence for direct electrostatic interactions between charged residues in loops 2 or 7 and residue Lys-276 in the M2-M3 loop of the GlyR ␣1-subunit. This indicates that while the mechanism of signal transduction may be similar between the various receptors of the LGIC superfamily, it is not identical. Studies on the nAChR and 5-HT 3 receptor, in addition to further studies on the GABA A R and GlyR, will be necessary to determine the breadth of difference in this mechanism between LGIC receptors. The process of forming theoretical models, mutation and covalent modification analysis and further refining of molecular models will continue to be an invaluable tool in the structural and functional definition of the processes of receptor activation in the LGIC superfamily.