Cysteine Substitutions Define Etomidate Binding and Gating Linkages in the α-M1 Domain of γ-Aminobutyric Acid Type A (GABAA) Receptors*

Background: Etomidate induces anesthesia via intersubunit transmembrane sites in GABAA receptors. Results: In receptors engineered with α-M1 domain cysteines, GABA accelerates modification. Etomidate inhibits modification at three positions. Conclusion: Etomidate contacts a subdomain of α-M1 linked to channel gating, consistent with in silico model docking. Significance: We identify new structures that both bind anesthetic and modulate channel gating through rearrangement. Etomidate is a potent general anesthetic that acts as an allosteric co-agonist at GABAA receptors. Photoreactive etomidate derivatives labeled αMet-236 in transmembrane domain M1, which structural models locate in the β+/α- subunit interface. Other nearby residues may also contribute to etomidate binding and/or transduction through rearrangement of the site. In human α1β2γ2L GABAA receptors, we applied the substituted cysteine accessibility method to α1-M1 domain residues extending from α1Gln-229 to α1Gln-242. We used electrophysiology to characterize each mutant's sensitivity to GABA and etomidate. We also measured rates of sulfhydryl modification by p-chloromercuribenzenesulfonate (pCMBS) with and without GABA and tested if etomidate blocks modification of pCMBS-accessible cysteines. Cys substitutions in the outer α1-M1 domain impaired GABA activation and variably affected etomidate sensitivity. In seven of eight residues where pCMBS modification was evident, rates of modification were accelerated by GABA co-application, indicating that channel activation increases water and/or pCMBS access. Etomidate reduced the rate of modification for cysteine substitutions at α1Met-236, α1Leu-232 and α1Thr-237. We infer that these residues, predicted to face β2-M3 or M2 domains, contribute to etomidate binding. Thus, etomidate interacts with a short segment of the outer α1-M1 helix within a subdomain that undergoes significant structural rearrangement during channel gating. Our results are consistent with in silico docking calculations in a homology model that orient the long axis of etomidate approximately orthogonal to the transmembrane axis.

Although photolabeling has identified the general region where etomidate binds to GABA A receptors, complementary approaches are needed to more fully explore etomidate binding interactions in intact receptors. Receptor protein used for photolabeling was affinity-purified in detergent and may have undergone structural changes. The photosensitive diazirine moieties of azietomidate and TDBzl-etomidate are appended to one end of the stereoselective parent ligand and may fail to label some side chains in the site.
Given the identification of ␣Met-236 as a likely contact point, etomidate interactions with the ␣-M1 domain are of particular interest. Residues in ␣-M1 have previously been linked to GABA A receptor modulation by barbiturates (23,24) and neurosteroids (25,26) as well as gating transduction between extracellular GABA sites and the transmembrane channel (23,27,28). Cysteine substitution and modification studies (21,22) have provided clues to some side chain orientations and evidence of rearrangements during gating for the pre-M1 and outer ␣-M1 domain from ␣Ile-223 to ␣Met-236.
The aims of this study were to investigate the roles of ␣1-M1 domain residues in receptor gating, etomidate modulation, and etomidate binding interactions in functional ␣1␤2␥2L GABA A receptors in native membranes. We applied the substituted cysteine accessibility method to a series of 13 single-residue cysteine-substituted mutations in the ␣1-M1 domain near ␣1Met-236. In each mutant channel, sensitivity to GABA and etomidate and the effects and rates of pCMBS modification with and without GABA were quantified electrophysiologically. We also tested proximity to bound etomidate at each accessible cysteine by testing if etomidate blocks sulfhydryl modification. Additional biochemical studies were performed on channels harboring the ␣1M236C mutation. Results were compared with in silico etomidate docking calculations in a homology model of the ␤2-M3/␣1-M1 interfacial pocket.

MATERIALS AND METHODS
Animal Use-Female Xenopus laevis were housed in a veterinary-supervised environment in accordance with local and federal guidelines. Frogs were anesthetized by immersion in 0.2% tricaine (Sigma-Aldrich) prior to minilaparotomy to harvest oocytes.
Chemicals-R-(ϩ)-Etomidate was obtained from Bedford Laboratories (Bedford, OH). The clinical preparation in 35% propylene glycol was diluted directly into buffer. Propylene glycol at the resulting concentrations has no effect on GABA A receptor function (10). Picrotoxin (PTX) was purchased from Sigma-Aldrich and dissolved (2 mM) in electrophysiology buffer. Alphaxalone was purchased from MP Biomedical (Solon, OH) and prepared as a stock solution in DMSO. p-Chloromercuribenzenesulfonic acid sodium salt (pCMBS), methyl methanethiosulfonate (MMTS), and 2-aminoethyl methanethiosulfonate hydrobromide (MTSEA) were purchased from Toronto Research Chemicals (North York, Canada). Salts and buffers were purchased from Sigma-Aldrich.
Molecular Biology-cDNAs for human GABA A receptor ␣ 1 , ␤ 2 , and ␥ 2L subunits were cloned into pCDNA3.1 vectors (Invitrogen). Mutations in cDNA were created with oligonucleotide-directed mutagenesis using QuikChange kits (Agilent Technologies, Santa Clara, CA). Clones from each mutagenesis reaction were subjected to DNA sequencing through the entire cDNA region to confirm the presence of the mutation and absence of stray mutations.
Oocyte Electrophysiology-Messenger RNA synthesis and Xenopus oocyte expression were performed as described previously (15). Experiments were performed at room temperature (21-23°C) in ND96 buffer (96 mM NaCl, 2 mM KCl, 0.8 mM MgCl 2 , 1.8 mM CaCl 2 , 5 mM HEPES, pH 7.5). Peak current responses to varying GABA concentrations (range from 0.1 M to 30 mM), alone or co-applied with 3.2 M etomidate, were measured in Xenopus oocytes (n Ն 3) using two-microelectrode voltage clamp electrophysiology, as described previously (29). GABA applications were from 5 to 20 s, depending on the time to reach steady-state peak current. Normalizing GABA responses at maximal GABA (1-30 mM) were recorded every second or third sweep. Picrotoxin-sensitive leak was measured using 2 mM PTX, followed by Ͼ5-min washout and a maximal GABA response test. Etomidate (3-10 M) or alphaxalone (2 FIGURE 1. Transmembrane domains and amino acids forming etomidate sites in GABA A receptors. A, a diagram of a ␣1␤2␥2L GABA A receptor crosssectioned in the membrane plane and viewed from the extracellular space illustrates the arrangement of ␣1 (yellow), ␤2 (blue), and ␥2 (green) subunits and the transmembrane domains (M1-M4) within each subunit. The chloride channel is shaded gray. Etomidate sites (red ovals) are located between ␣-M1 and ␤-M3 domains. B, an expanded diagram of one etomidate site (outlined in A), showing a helical wheel projection of the ␣1-M1 sequence from Gln-229 (nearest) through Gln-242. The orientation of the side chains is approximate and based on a homology model built from the structure of G. violaceous pentameric ion channels (33). GABA A receptor residues that are photolabeled by etomidate derivatives (13,14), in both ␣-M1 and ␤-M3, are highlighted in pink. We also illustrate ␤2Asn-265 on ␤-M2, a residue where mutations influence etomidate sensitivity (37,38). M) was used as a gating enhancer together with maximal GABA to estimate GABA efficacy. Oocyte currents were low pass-filtered at 1 kHz (model OC-725B, Warner Instruments, Hamden, CT), digitized at 1-2 kHz (Digidata 1200, Molecular Devices, Sunnyvale, CA), and recorded digitally (pClamp 7, Molecular Devices).
Cysteine Modification with pCMBS and Etomidate Protection in Xenopus Oocytes-Maximally GABA-activated receptors were used for control (reference) pCMBS modification, and etomidate concentrations in protection studies were at least 2 times EC 50 for direct activation (30 -100 M), establishing occupation of Ͼ90% of etomidate sites. The pCMBS concentration used for each mutant channel was chosen so that control modification was complete after 40 -60 s of pCMBS exposure. Oocytes were repetitively stimulated with GABA pulses every 5 min until at least three sequential current responses were constant (Ϯ5%). For modification, oocytes were exposed to pCMBS (alone, with GABA, or with GABA plus etomidate) for 5, 10, or 20 s followed by a 5-min ND96 wash. A GABA response test, when appropriate, was repeated prior to the next pCMBS modification, and up to 15 (but usually fewer than 10) modification test cycles were repeated in each oocyte. For modification rate analysis, peak currents were normalized to a pre-or postmodification control and plotted against cumulative pCMBS exposure in units of s ϫ mM. Normalized data were fitted to single exponential functions to determine the apparent modification rate constant in M Ϫ1 s Ϫ1 .
Electrophysiology in HEK293 Cell Membrane Patches-HEK293 cell maintenance and transfection for functional studies were performed as described previously (15). Current recordings from excised outside-out membrane patches were performed at Ϫ50 mV and room temperature (21-23°C) as described by Scheller and Forman (30). Currents were stimulated using pulses of GABA delivered via a multibarrel superfusion pipette coupled to piezo-electric elements that switched solutions in under 1 ms. Currents were filtered at 5 kHz and digitized at 10 kHz for off-line analysis.
HEK Cell Membrane Preparation for Binding Assays-HEK293T cells at 30 -40% confluence were co-transfected using CaPO 4 precipitation (31) with cDNAs encoding wildtype or mutant ␣1, ␤2, and ␥2L. At 48 h after transfection, cells were harvested and centrifuged at 1500 ϫ g for 10 min. The cell pellet was washed twice in PBS and again pelleted. The pellet was homogenized at 11,000 rpm for 30 s in PBS using an Ultra-Turrax T25 homogenizer (Janke & Kunkel, Staufen, Germany). The homogenate was centrifuged for 30 min at 25,000 ϫ g at 4°C. The membranes were washed and repelleted twice and then resuspended in PBS. samples were suction-filtered on GF/B glass fiber filters. Filters were washed two times with 3 ml of assay buffer and transferred to scintillation vials with 2.5 ml of scintillation fluid (Ecolume, ICN Pharmaceuticals, Aurora, OH) for counting.

MMTS/MTSEA Reaction and Etomidate Protection in
Electrophysiological Data Analysis-Analyses for agonist concentration responses, etomidate-induced left shift, and allosteric co-agonist model fitting followed our approach described elsewhere (15,32). Peak GABA-stimulated currents were normalized to maximal GABA responses, and GABA concentration-response data for individual oocytes in the absence and presence of etomidate were fitted with logistic functions using non-linear least squares (Graphpad Prism version 5), where EC 50 is the half-maximal activating concentration and n H is Hill slope.
EC 50 shift ratios were calculated from the difference in log-(GABA EC 50   Analysis of patch macrocurrents for activation, desensitization, and deactivation kinetics was performed as described previously (32) using non-linear least squares fits to Equation 4 and F-tests at p ϭ 0.99 (Clampfit version 8.0, Molecular Devices) to choose the best number of exponents.
All activation traces were best fit with a single exponent, whereas desensitization and deactivation were consistently fitted with two exponents.
Molecular Modeling-A model of the ␣1␤3␥2 GABA A receptor was constructed using the structure of the pentameric ion channel homolog GLIC from Protein Data Bank entry 3P50 as described (33). ␤2 peptide sequences were substituted into the ␤3 structures, requiring no insertions or deletions and only 26 amino acid replacements, seven of which were in the transmembrane domain (two in M3 and five in M4). No substitutions were near the etomidate binding sites. The model was placed within a simulated membrane force field and minimized with the photoreactive etomidate analog azietomidate placed in the etomidate binding interfaces, consistent with photolabeling results (13,14). Using C-DOCKER in the Discovery Studio modeling software package (Accelrys Inc., San Diego, CA), R-(ϩ)-etomidate (183-Å 3 molecular volume, 213-Å 3 Connolly surface volume) was docked within an 11-Å radius sphere in each binding site, solving for the 100 lowest energy solutions starting with 50 different initial orientations and 50 different molecular dynamics-altered etomidate configurations. Interaction energies for all 200 solutions (100 for each site) were between Ϫ36 and Ϫ39 kcal/mol. As controls, etomidate docking was performed at the other three model intersubunit transmembrane cavities.
Statistical Analysis-Results are reported as mean Ϯ S.D. unless otherwise noted. Wild-type and all mutant group comparisons for log(EC 50 ) and ⌬log(EC 50 ) were performed using one-way ANOVA with Dunnett's post hoc multiple comparisons test in GraphPad Prism. Within the mutant group, statistical comparisons of pCMBS modification rates measured under three conditions were performed using ANOVA with Tukey's post hoc test. Etomidate-dependent [ 3 H]flunitrazepam binding data were analyzed using two-way ANOVA with post hoc multiple comparisons test. Pairwise comparisons were performed using Student's t tests. Statistical significance was inferred at p Ͻ 0.05.

Functional Characterization of Wild-type and ␣1M236C␤2␥2L GABA A Receptors
GABA-and Etomidate-dependent Channel Gating-Initial studies focused on the ␣1Met-236 residue identified in photolabeling studies (13,14). GABA-and etomidate-dependent gating, spontaneous activation, and macrocurrent rapid kinetics were characterized in ␣1M236C␤2␥2L channels expressed in Xenopus oocytes or HEK293 cells and compared with ␣1␤2␥2L (wild type). Results of oocyte studies are summarized in Table 1. Functional characteristics of ␣1␤2␥2L receptors in oocytes were similar to those reported in earlier studies (10,15,32). Wild-type GABA EC 50 value averaged 37 M, and the addition of 3.2 M etomidate reduced GABA EC 50 18-fold to about 2 M. Etomidate directly activated wild-type channels with an average EC 50 of 30 M, and maximum etomidate activation averaged 42% of the maximal GABA response. In oocytes expressing ␣1M236C␤2␥2L GABA A receptors, GABA EC 50 averaged about 10-fold higher (390 M) than that in wild-type ( Fig. 2A (solid symbols) and Table 1). In the presence of etomidate, ␣1M236C␤2␥2L receptor currents elicited with maximal GABA (10 -30 mM) increased about 4-fold, indicating a low agonist efficacy for GABA. GABA EC 50 was reduced about 5-fold ( Fig. 2A, open symbols). Etomidate alone was a potent and efficacious agonist of ␣1M236C␤2␥2L receptors, inducing maximal currents that were 4-fold larger than those elicited by 10 -30 mM GABA (Fig. 2B). Rapid application of GABA (10 mM) to voltage-clamped outside-out patches from HEK293  EC 50 values and confidence intervals were calculated from the average and S.D. of fitted log(EC 50 ) values from individual oocytes (n Ͼ 3). GABA efficacy was calculated from the ratio of maximal GABA response to maximal GABA plus an enhancing drug (etomidate or alphaxalone). Etomidate efficacy is expressed as the ratio of maximal etomidate current to the maximal GABA current. Comparisons with wild type were performed using one-way ANOVA with Dunnett's post hoc test. CI, confidence interval.

Receptor
43 ( (34). Also, GABA efficacy in this mutant may differ in HEK293 cells and Xenopus oocytes. Voltage-clamped oocytes (n ϭ 3) expressing ␣1M236C␤2␥2L receptors displayed small picrotoxin-sensitive leak currents in the absence of GABA (Fig. 2D) that were 1-2% of that elicited with 10 mM GABA (0.3-0.5% of maximal currents elicited with 100 M etomidate). Oocytes expressing wild-type receptors displayed no picrotoxin-sensitive leak currents. GABA and etomidate efficacies (maximal fraction of receptors activated) were assessed using single-sweep experiments, where receptors were initially activated with one of these agonists, followed by the addition of a co-agonist. Maximal ␣1M236C␤2␥2L currents elicited by 10 mM GABA increased about 4-fold (300% over base line) with the addition of 2 M alphaxalone (Fig. 2E), whereas currents elicited by 100 M etomidate were enhanced only 4 -5% with the addition of 3 mM GABA (Fig. 2F). Assuming that  Table 2. D, spontaneous channel gating current in an oocyte. The small outward current during PTX application is due to inhibition of active channels. Current elicited with 10 mM GABA in the same cell is also displayed. Average spontaneous activity is 1.8 Ϯ 0.28% of maximal GABA response. E, estimation of maximal GABA efficacy in an oocyte. GABA (10 mM; white bar) alone elicits a current that is enhanced severalfold with co-application of alphaxalone (2 M; black bar). Average results for GABA efficacy are reported in Table 1. F, estimation of etomidate agonist efficacy in an oocyte. Etomidate (100 M; black bar) elicits a maximal current that is only modestly enhanced with co-application of GABA (3 mM; white bar). Average results for etomidate efficacy are reported in Table 1. G, allosteric co-agonist modeling of GABA and etomidate activation. Estimated P o was calculated using Equation 2 from data in A and B (same symbols used for each data set). Equation 3 was fitted to estimated P o using nonlinear least squares with both [GABA] and [etomidate] as input variables. Lines represent the fitted model. Fitted model parameters are reported in Table 3.

TABLE 2 Wild-type and mutant GABA A receptor rates of activation, desensitization, and deactivation
Rates for mutant receptors did not significantly differ from those for wild type. amp, amplitude.

Maximal activation rate
Fast desensitization rate (s ؊1 ) (amp (%)) both of these combinations of co-agonists open all activatable receptors, these results indicate that the efficacies of GABA and etomidate are ϳ0.24 and 0.96, respectively. Monod-Wyman-Changeux Model Analysis-A two-state MWC allosteric co-agonist model accounts for the gating effects of both GABA and etomidate in oocytes (10) and provides a mechanistic framework for interpretation of the functional effects of mutations (15,16,32). A non-linear least squares fit to this model (Equation 3) was performed using estimates of [GABA] and [etomidate]-dependent P o derived from ␣1M236C␤2␥2L oocyte data corrected for spontaneous channel activity and rescaled for GABA efficacy (Equation 2).
Model analysis of pooled ␣1M236C␤2␥2L receptor data was constrained by an L 0 value (200) derived from the spontaneous gating measurement. The resulting fit and parameters are shown in Fig. 2G and compared with model parameters fitted to wild-type data ( Table 3). The four free fitted parameters were well constrained by the data sets. MWC model analysis indicates that the M236C mutation dramatically reduced GABA efficacy (Table 3; c for ␣1M236C␤2␥2L is 70-fold larger than for wild-type) while only slightly reducing etomidate efficacy (d for ␣1M236C␤2␥2L does not significantly differ from the wildtype value). The relatively small GABA EC 50 shift with etomidate is mostly due to the low efficacy of GABA, whereas spontaneous receptor activity accounts for efficacious etomidate agonism. A similar sensitivity to etomidate agonism was observed previously in ␣1M236W␤2␥2L receptors, which also display spontaneous gating (15).
pCMBS Modification and Etomidate Protection-Application of pCMBS (up to 2 mM for 60 s, with or without 1 mM GABA) to oocytes expressing wild-type GABA A receptors produced no detectable changes in electrophysiological properties, including spontaneous activation and responses to low and high GABA (data not shown). Applying pCMBS to oocytes expressing ␣1M236C␤2␥2L receptors irreversibly increased leak currents and reduced 10 mM GABA current responses, suggesting that modification activates and enhances desensitization. In the absence of GABA, pCMBS concentrations over 1 mM were needed to produce functional changes within 60 s, and the apparent rate of modification at room temperature was low (16 M Ϫ1 s Ϫ1 ; Table 4). Co-application of 10 mM GABA with pCMBS increased the rate of modification over 10-fold to 210 M Ϫ1 s Ϫ1 (Fig. 3A). Because ␣1M236C␤2␥2L channels display spontaneous activity and were readily activated by etomidate, we used a maximally GABA-activated receptor condition (10 mM GABA) for control studies of pCMBS modification to compare with modification with GABA plus etomidate. For protection, we used 30 M etomidate, based on the fitted allosteric model (Table 3), which indicated an etomidate dissociation constant for GABA-activated ␣1M236C␤2␥2L receptors (K E ϫ d) near 1 M. Co-application of 30 M etomidate and 10 mM GABA reduced the apparent pCMBS modification rate about 10-fold compared with GABA alone (Fig. 3, B and C, and Table 4).
Etomidate protection of ␣1M236C from sulfhydryl modification was also assessed using a biochemical assay in HEK293 Results are from nonlinear least squares fits to Equation 3, which describes an allosteric two-state equilibrium mechanism with two classes of agonist sites (one for GABA and one for etomidate), each with two equivalent sites (10). L 0 is a dimensionless basal equilibrium gating variable, inversely related to spontaneous activity. For wild-type receptors, spontaneous activity was undetectable, and we constrained L 0 to a previous estimate (25,000) for wild type (32). K G and K E are equilibrium dissociation constants for GABA and etomidate binding to inactive states, and c and d are dimensionless parameters representing the respective ratios of binding constants in active versus inactive states. The agonist efficacies of GABA and etomidate are inversely related to c 2 and d 2 , respectively.

Cysteine Substitution Scan of ␣1-M1 Residues
GABA-and Etomidate-dependent Gating-The molecular length of etomidate in an extended configuration is ϳ12 Å, about the distance of two full turns along an ␣-helical axis. We therefore characterized single residue cysteine mutants along the ␣1-M1 domain from ␣1Q229 to ␣1Q242, spanning four helical turns with the photolabeled ␣1Met-236 residue in the middle of this region. All cysteine-substituted receptors were characterized using a two-microelectrode voltage clamp in Xenopus oocytes to assess GABA EC 50 , GABA response modulation by 3.2 M etomidate, and etomidate direct activation. Results are summarized in Table 1 and Fig. 4.

Etomidate and GABA A Receptor ␣-M1 Domain Cysteines
In the presence of 3.2 M etomidate, five mutant channels exhibited GABA EC 50 shifts that significantly differed (one-way ANOVA, p Ͻ 0.05) from the wild-type value (Table 1 and Fig.  4B). Receptors with ␣1Q229C and ␣1M236C mutations showed etomidate shifts smaller than that in wild type, and ␣1T230C, ␣1Y231C, and ␣1S241C were associated with larger shifts. All ␣1-M1 domain cysteine-substituted mutant receptors were directly activated by etomidate (Table 1). Etomidate direct activation EC 50 values varied from 3 M (␣1I235C␤2␥2L) to 40 M (␣1Q229C␤2␥2L), and the maximal relative efficacy of etomidate activation, normalized to maximal GABA efficacy, varied from about 0.24 in ␣1Q229C␤2␥2L channels up to about 4.0 in both ␣1P233C␤2␥2L and ␣1M236C␤2␥2L receptors. Whereas maximum etomidate agonist efficacy relative to GABA was significantly increased by P233C and M236C mutations, the estimated normalized etomidate efficacy for the mutants (calculated 95% confidence intervals for GABA efficacy ϫ etomidate efficacy, from Table 1) did not significantly differ from wild type. In other words, high relative etomidate efficacy simply reflected low GABA efficacy with P233C and M236C mutants. Overall, six mutations, all but one extracellular to Thr-237, altered etomidate sensitivity as assessed by direct activation EC 50 or EC 50 shifts (Table 1).
pCMBS Modification and Etomidate Protection-In the resting closed state (no GABA), exposure to pCMBS up to 1 mM for 60 s produced no functional changes in receptors containing ␣1Y231C, ␣1P233C, ␣1I235C, ␣1V238C, ␣I239C, and ␣Q242C mutations. For the other seven cysteine-substituted mutant channels, evidence of pCMBS modification in the absence of GABA was observed with apparent rates ranging from 10 to 3000 M Ϫ1 s Ϫ1 (Table 4). Along the transmembrane axis, closed state aqueous accessibility extended from Gln-229 to Ser-241.
Modification of cysteine substitutions at the most extracellular residues we studied, Q229C and T230C, was much faster than at more internal residues, with L240C slowest. A helical wheel projection based on the homology model (Fig. 5A) locates all closed state accessible residues within one hemi-face of ␣-M1, projecting toward ␤2-M3 and the intersubunit space or toward ␣-M2. None are predicted to face lipid in the model.
With the addition of GABA at Ͼ10 ϫ EC 50 , pCMBS modification was observed in eight mutant receptors: the seven that were modified in the resting state plus ␣1I235C␤2␥2L. Rates of pCMBS modification in the presence of GABA varied from 20 to 520,000 M Ϫ1 s Ϫ1 (Table 4). GABA significantly increased pCMBS modification rates at all modifiable ␣1-M1 cysteinesubstituted sites, except ␣1Ser-241. The ratio of modification rates in the presence versus absence of GABA varied from 2-fold at ␣1L232C and ␣1L240C up to 170-fold at ␣1T230C (Fig. 5B). A rate ratio could not be calculated for ␣1I235C receptors, which displayed no modification in the absence of GABA.

DISCUSSION
Our main findings are that cysteine substitution at most residues in the outer GABA A receptor ␣-M1 domain impair GABA agonism, whereas effects on etomidate modulation vary. GABA accelerates pCMBS modification at most accessible positions. Etomidate protection from modification indicates proximity to three (two newly identified) residues in this region.
The Outer ␣1-M1 Domain Is Linked to GABA Agonism and the Channel State-Cysteine substitutions in ␣1-M1 from Gln-229 through Met-236 (excluding Cys-234) consistently impair GABA agonism, increasing GABA EC 50 , reducing GABA efficacy, or both ( Table 1). Five of these seven cysteine substitu-  Fig. 1B shows pCMBS-modified residues as colored circles; green, modification enhanced by etomidate; magenta, modification reduced by etomidate. The heavy outline around Ile-235 indicates modification only in the presence of GABA. B, bars represent log ratios (mean Ϯ S.D. (error bars)) of pCMBS modification rates measured in the presence of maximally activating GABA (Ͼ10 ϫ EC 50 ) to the rate without GABA (n Ն 3 for each condition). *, rates in the presence versus absence of GABA differ significantly with p Ͻ 0.05. †, a ratio for I235C could not be calculated because no modification was evident without GABA. C, bars represent log ratios (mean Ϯ S.D.) of pCMBS modification rates measured in the presence of maximally activating GABA plus etomidate (30 -100 M) to rates in the presence of GABA alone (n Ն 3 for each condition). Black bars, reduced rates; white bars, increased rates. *, rates in the presence versus absence of etomidate differ significantly with p Ͻ 0.05. OCTOBER 18, 2013 • VOLUME 288 • NUMBER 42 tions are modifiable with pCMBS, and modification accelerates with application of GABA (Table 4). In contrast, ␣1-M1 residues intracellular to Met-236 do not significantly affect GABA sensitivity and are relatively inaccessible to pCMBS. Modifica-tion by pCMBS signifies accessibility via an aqueous pathway (39). Thus, water-accessible ␣1-M1 residues are largely extracellular to Thr-237, consistent with structural homology models in which the space between transmembrane domains nar- rows intracellularly. Accelerated pCMBS modification by GABA suggests increased mobility or aqueous access. Bali and Akabas (21) have also examined substituted cysteine accessibility in ␣1-M1 from Gly-224 to Met-236. At residues examined in both studies (Gln-229 to Met-236), the pattern of GABA-dependent pCMBS modification at Q229C, T230C, L232C, and M236C was similar. Contrasting with our results, Bali and Akabas reported lower GABA EC 50 values and higher pCMBS modification rates and reported that Y231C mutations abolish channel function. These differences may be due to their rat "Cys-light" channel background, lacking native transmembrane cysteines, including ␣1Cys-234. Importantly, our main conclusions are unaffected by these differences.

Etomidate and GABA A Receptor ␣-M1 Domain Cysteines
A variety of data support the hypothesis that GABA activation expands the space around ␣1-M1, increasing both water and side chain mobility to a depth near ␣1Leu-240. Cysteine substitutions could impair GABA-induced gating by reducing side chain volume and increasing water in transmembrane pockets, increasing entropic energy costs. Cysteine, being smaller than most of the residues we mutated, does not occlude etomidate binding, which displaces water and reduces entropy, stabilizing open channel states. Tryptophan substitution at ␣1Met-236 or ␤2Met-286 enhances gating in the same manner while occluding the etomidate site (15). A similar mechanism was proposed by Jenkins et al. (40) for volatile anesthetic actions at intrasubunit pockets. Bali et al. (22) observed GABA-dependent disulfide formation between ␤2M286C and ␣1Q229C and suggested that these residues approach each other during gating. However, this intersubunit disulfide bond stabilizes a non-conducting receptor state. GABA-dependent cross-linking could also result from increased side chain or backbone mobility. Delineating transmembrane domain motions during channel gating and desensitization remains an important experimental challenge.
Outer ␣-M1 Domain Linkage to Etomidate Modulation-Several outer ␣-M1 domain mutations also affected etomidate sensitivity with a pattern distinct from that for GABA sensitivity (Fig. 4). Two cysteine substitutions, Q229C and M236C, reduce etomidate-induced left-shift, whereas T230C, Y231C, and S241C increased modulation. MWC allosteric analysis indicates that the ␣1M236C mutation insignificantly affects etomidate allosteric efficacy (Table 3, d). Thus, the small GABA EC 50 shift in ␣1M236C␤2␥2L receptors reflects only a portion of etomidate modulation; the remainder is evident as increased GABA efficacy (Fig. 1A). The divergence of cysteine mutant effects on orthosteric and allosteric agonists rules out isolated effects on channel gating (L 0 in MWC models). Considering evidence that etomidate contacts ␣-M1 domain residues (discussed below), we speculate that its movements differ when gating is triggered by GABA alone versus with etomidate bound, altering open state stability and channel gating modes.
We used 30 -100 M etomidate, concentrations associated with direct receptor agonism, to achieve high etomidate site occupation in protection experiments. Importantly, etomidate agonism is mediated by the same sites that produce modulation by low (general anesthetic) concentrations (10). Despite high predicted site occupancy, etomidate protection at L232C and T237C was less complete than at M236C, suggesting less steric proximity. An alternative protection mechanism whereby etomidate blocks the aqueous pCMBS access pathway is unlikely, because access to L240C and S241C, furthest from the membrane surface, was preserved with etomidate bound. The possibility that etomidate indirectly induces movement of protected residues to become pCMBS-inaccessible also exists. However, a steric protection mechanism is also supported by comparison with Q229C, T230C, and I235C. In receptors with these mutations, etomidate increased pCMBS modification rates beyond those with GABA alone (Table 4 and Fig. 5), paralleling electrophysiological evidence that etomidate increased channel activation. Notably, etomidate also enhanced channel gating in the three protected Cys-mutant receptors (Table 1).
Our results suggest that etomidate interactions are limited to 1.5 helical turns of ␣1-M1, about 8 Å along the transmembrane axis. The long leg of etomidate fits into a right cylinder of diameter 7.4 Å and length 11.9 Å, sufficiently long to contact residues on two or three sequential helical turns. Thus, our results are most consistent with etomidate binding with its long axis approximately perpendicular (90 Ϯ 45°) to the transmembrane axis.
Comparison with Structural Homology Models and Crystallized Receptors-In silico docking of etomidate to the ␣-M1/ ␤-M3 cavity in a structural homology model (Fig. 7) placed etomidate adjacent to all three protected residues (Leu-232, Met-236, and Thr-237) with its long axis oriented approximately orthogonal to the transmembrane axis. Thus, cysteine protection results agree with the predictions of this and previous models of the etomidate site (13,14). In the model, etomidate also contacts Ile-228, Pro-233, and Leu-240. We did not include Ile-228 in our study, and ␣1P233C␤2␥2L receptors were unaffected by exposure to pCMBS, precluding protection tests. The P233C mutation also did not alter etomidate sensitivity (Table 1 and Fig. 4B). Thus, ␣1P233 could be an etomidate contact point with a "silent" cysteine mutation. The L240C mutation produced no change in etomidate sensitivity and was modified by pCMBS but not protected. The model places Leu-240 farther from etomidate than the protected residues, and a cysteine mutation would further increase this separation. Assessing other Pro-233 and Leu-240 mutations might provide additional information about etomidate interactions.
High resolution structural evidence locating allosteric agonists in transmembrane intersubunit sites is emerging from several pentameric ligand-gated ion channel models. Co-crystals of C. elegans GluCl channels locate ivermectin (19) in sites that are structural homologs of etomidate sites in GABA A receptors. Ivermectin is much larger than etomidate and makes extensive contacts with GluCl M1 domains. Co-crystallization of propofol and desflurane with GLIC (46) identified intrasubunit transmembrane sites, although cysteine accessibility studies by Ghosh et al. (47) suggest that propofol inhibition is mediated by intersubunit sites. Recent crystal data from positively modulated GLIC-F238A mutants also locate anesthetics and alcohols in intersubunit cavities (48). Co-crystallization of Erwinia chrysanthemi (ELIC) channels with bromoform also identified an intersubunit site deep in the membrane (near residues corresponding to GABA A ␣1Gln-242 and ␣1Trp-246), along with other sites (49). Thus, both prokaryotic and eukaryotic pentameric ion channels harbor multiple types of anesthetic binding sites, linked to both enhancing and inhibitory anesthetic effects (50).
In our current study, cysteine mutations at residues predicted to project into the intrasubunit helical bundle pocket displayed weaker effects on GABA and etomidate sensitivity than at residues facing the intersubunit site. This suggests that intersubunit transmembrane pockets in GABA A receptors influence channel gating more than neighboring intrasubunit transmembrane pockets. Accordingly, although we and others (21) find evidence for variable ␣-M1 side chain movements with gating, suggesting local secondary structure changes, our data also point to important quaternary rearrangements (at subunit interfaces) linked to gating. Despite agreement between model docking and protection results, the limitations of in silico docking to current homology models are evident from control computations showing that etomidate docks with similar stability to all five transmembrane subunit interfaces. This is probably because the model was built on a homomeric template. Nonetheless, recent photolabeling studies demonstrate that R-(ϩ)-etomidate binds to ␤-M3/␣-M1 interfaces with greater than 100-fold higher affinity than to ␣-M3/␤-M1 or ␥-M3/␤-M1 interfaces that are selectively photolabeled by a barbiturate derivative (33). More broadly, links between crystallized receptor structures and functional states remain tenuous, and cocrystal structures reveal no significant rearrangements in anesthetic binding cavities (46,48), highlighting the value of complementary approaches.
Can Mutant Function Predict Anesthetic Contact?-We thoroughly characterized cysteine mutant channel functions yet found no phenotypic "fingerprint" for residues contacting etomidate. Considering the three protected ␣1-M1 mutants we identified and ␤2Met-286 (16), two (␤2Met-286 and ␣1Met-236) reduce etomidate left shift, and three (␤2Met-286, ␣1Met-236, and ␣1Leu-232) reduce GABA sensitivity. However, in ␣1Q229C␤2␥2L, a mutant with functional characteristics similar to ␣1M236C␤2␥2L and ␤2M286C␤2␥2L, etomidate enhances rather than inhibits pCMBS modification. MWC allosteric analysis, which derives independent parameters for etomidate binding and efficacy, also fails to unambiguously identify etomidate binding site mutations. MWC analysis of ␣1M236C indicates minimal changes from wild type in both etomidate affinity and efficacy, whereas cysteine protection and photolabeling clearly place this residue in the etomidate binding site. Functional analysis of multiple mutations at candidate residues might improve inferences regarding ligand interactions. MWC analyses of ␣1M236W, ␤2M286W, and ␤2M286C mutations (15,16) identified dominant effects on the etomidate efficacy parameter (d), which reflects binding to the high affinity activated state. Our current protection results were obtained in GABAbound receptors and thus reflect interactions with open and/or desensitized states. Although ␤2M286C is protected by etomidate in resting (closed) state receptors (16), the low affinity of this state makes its study more challenging.
In summary, we find that cysteine substitutions at outer ␣1-M1 residues (Gln-229 through ␣Met-236) reduce GABA sensitivity and variably affect etomidate sensitivity. Accessibility to pCMBS was evident in ␣1-M1 residues predicted to face the intersubunit space or the ␣1-M2 domain and few others. GABA enhanced modification at most of these positions, indicating gating-coupled rearrangements at the sub-domain level. Etomidate inhibited modification of L232C, M236C, and T237C, indicating contact with a limited portion of ␣-M1 at the subunit-subunit interface. These results agree with in silico model docking calculations, suggesting that the long axis of etomidate lies perpendicular to the ␣1-M1 domain.