Mutations at Beta N265 in γ-Aminobutyric Acid Type A Receptors Alter Both Binding Affinity and Efficacy of Potent Anesthetics

Etomidate and propofol are potent general anesthetics that act via GABAA receptor allosteric co-agonist sites located at transmembrane β+/α− inter-subunit interfaces. Early experiments in heteromeric receptors identified βN265 (M2-15′) on β2 and β3 subunits as an important determinant of sensitivity to these drugs. Mechanistic analyses suggest that substitution with serine, the β1 residue at this position, primarily reduces etomidate efficacy, while mutation to methionine eliminates etomidate sensitivity and might prevent drug binding. However, the βN265 residue has not been photolabeled with analogs of either etomidate or propofol. Furthermore, substituted cysteine modification studies find no propofol protection at this locus, while etomidate protection has not been tested. Thus, evidence of contact between βN265 and potent anesthetics is lacking and it remains uncertain how mutations alter drug sensitivity. In the current study, we first applied heterologous α1β2N265Cγ2L receptor expression in Xenopus oocytes, thiol-specific aqueous probe modification, and voltage-clamp electrophysiology to test whether etomidate inhibits probe reactions at the β-265 sidechain. Using up to 300 µM etomidate, we found both an absence of etomidate effects on α1β2N265Cγ2L receptor activity and no inhibition of thiol modification. To gain further insight into anesthetic insensitive βN265M mutants, we applied indirect structure-function strategies, exploiting second mutations in α1β2/3γ2L GABAA receptors. Using α1M236C as a modifiable and anesthetic-protectable site occupancy reporter in β+/α− interfaces, we found that βN265M reduced apparent anesthetic affinity for receptors in both resting and GABA-activated states. βN265M also impaired the transduction of gating effects associated with α1M236W, a mutation that mimics β+/α− anesthetic site occupancy. Our results show that βN265M mutations dramatically reduce the efficacy/transduction of anesthetics bound in β+/α− sites, and also significantly reduce anesthetic affinity for resting state receptors. These findings are consistent with a role for βN265 in anesthetic binding within the β+/α− transmembrane sites.


Introduction
Etomidate and propofol are potent general anesthetics that act as positive allosteric modulators and agonists at GABA A receptors [1][2][3], the major inhibitory neurotransmitter receptors of the central nervous system and members of the pentameric ligandgated ion channel (pLGIC) superfamily [4,5]. Typical synaptic GABA A receptors consist of a, b, and c subunits, arranged b-a-ba-c counterclockwise viewed from the extracellular space [6]. Structural homology models of abc GABA A receptors, based on crystallography of distantly related pLGICs [5] and b3 homopentameric channels [7], are consistent with a large body of structure-function data. Each homologous subunit has a large extracellular domain, and four a-helices (M1 to M4) forming a transmembrane domain (Fig. 1A, B). The two GABA binding sites are located in the b+/a2 interfaces of the extracellular domain. The transmembrane domains of each subunit form four-helix bundles with all five M2 domains around the central chloride channel (Fig. 1B). Surrounding the M2 domains is a second ring of M1 and M3 a-helices, with M4 domains outermost. Portions of M1, M3 and M4 contact membrane lipids.
In addition to the a-M1 and b-M3 residues identified above, M2 domain residues may contribute to the b+/a2 transmembrane sites for propofol and etomidate. The focus of this study, b-M2 position 265 (M2-159), is of particular and longstanding interest as one of the first sites where mutations were found to affect GABA A receptor sensitivity to alcohols and anesthetics [21,22]. This residue also plays a critical role in determining etomidate and propofol sensitivity in mammalian GABA A Figure 1. Functional and structural models of GABA A receptor interactions with etomidate. Panel A: A side-on ribbon depiction of a structural homology model for a1b3c2L GABA A receptors based on the glutamate-gated chloride channel (GluCl) from Caenorhabditis elegans [20]. Both the extracellular domains and transmembrane domains are shown in relation to membrane lipids. Subunits are color-coded (a1 = blue; b3 = yellow; c2L = green). bN265 and other residues involved in etomidate and propofol binding are depicted as red stick structures in one of two b+/ a2 transmembrane interfacial sites. Panel B: A view of the transmembrane domains from the extracellular space shows the structure of each subunit's four-helix bundle and the arrangement of subunits around the central chloride channel (grey circle). Residues in one interfacial anesthetic site are depicted as red stick structures. Panel C: A close-up view of one b+/a2 transmembrane inter-subunit etomidate binding site in the homology model. Helix backbones are depicted as solid cylinders. The bN265 residue and six anesthetic contact residues identified by photolabeling or cysteine modification/protection are highlighted as labeled space-filling structures. Panel D: A Monod-Wyman-Changeux two-state (inactive = R; active = O) equilibrium co-agonist scheme with two equivalent orthosteric agonist (GABA; G) sites and two equivalent allosteric agonist (etomidate; E) sites is depicted [1]. For simplicity, states with only one occupied agonist site are omitted. The model is defined by five equilibrium parameters (see Eq. 3, methods): L 0 is a basal gating equilibrium (C/O); K G and K E are dissociation constants for respectively, GABA and etomidate binding to inactive receptors; c and d quantify the binding affinity ratios for respectively, GABA and etomidate to active vs. inactive receptors. Maximal agonist efficacies for GABA and etomidate are respectively, (1+L 0 c 2 ) 21 and (1+L 0 d 2 ) 21 . doi:10.1371/journal.pone.0111470.g001 receptors. Interchanging the b-M2-159 asparagine of b2 or b3 and the homologous serine of b1 accounts for the remarkable specificity of etomidate for receptors containing b2/3 subunits [22,23]. Methionine substitution at b2 or b3 N265 eliminates etomidate sensitivity and weakens propofol effects at the receptor level [24,25], and b3N265M transgenic mice show remarkable resistance to both anesthetics [26]. Quantitative mechanistic comparison of etomidate effects in voltage-clamped wild-type a1b2c2L and a1bN265Sc2L receptors expressed in Xenopus oocytes indicated that drug efficacy was reduced and that affinity for receptors might be weakened. Similar studies of a1bN265Mc2L receptors were inconclusive due to the absence of anesthetic effects [27]. Homology models of abc GABA A receptors locate bN265 either near or within the b+/a2 site (Fig. 1C), and in silico docking calculations suggest possible contact with anesthetics [9,12,28,29]. Moreover, contact with M2-159 residues is evident in x-ray diffraction structures demonstrating allosteric modulators bound within inter-subunit sites of crystallized pLGIC homologs: ivermectin contacts M2-159 of GluCl [20], while ethanol and the alkane anesthetic bromoform contact M2-159 sidechains in positively modulated prokaryotic Gloeobacter violaceus ion channel (GLIC) mutants [30]. Nonetheless, no anesthetic photolabel incorporation at GABA A bN265 has been detected. One photoreactive propofol analog incorporates at b3H267 [31], which is predicted to be on the M2 helix face opposite the b+/a2 interface. A SCAM protection study of a1bN265Cc2L receptors reported no propofol protection [13], while another reported protection by n-octanol [32]. No studies have reported whether etomidate protects bN265C from modification. Thus, it remains uncertain whether bN265 mutations impair only the efficacy of etomidate and propofol, a result suggesting indirect interactions, or also the affinity of these drugs for GABA A receptors, consistent with a role in binding.
To discriminate between the effects of bN265 mutations on anesthetic binding versus efficacy, we applied both direct and indirect structure-function approaches. We first performed pharmacological sensitivity and thiol modification-protection studies of a1b2N265Cc2L GABA A receptors expressed in Xenopus oocytes and monitored with voltage-clamp electrophysiology, using etomidate at concentrations up to 300 mM (intending to achieve high site occupancy). These experiments revealed that etomidate neither affects a1b2N265Cc2L receptor function nor inhibits reaction with a water-soluble thiol modifier. We also explored bN265M effects in GABA A receptors containing previously characterized second point mutations: a modifiable and anesthetic-protectable cysteine (a1M236C), and a mutation that mimics bound anesthetic (a1M236W). Our results reveal that bN265M reduces both anesthetic binding affinity at b+/a2 sites and transduction of gating effects associated with anesthetic binding.

Animal use
Adult female Xenopus laevis frogs were used as a source of oocytes for voltage-clamp electrophysiological experiments. Frogs were housed in a veterinarian-supervised facility and used in accordance with the NIH Guide for the Care and Use of Laboratory Animals. All animal procedures for this study were approved by the Massachusetts General Hospital Institutional Animal Care and Use Committee (Offices of Laboratory Animal Welfare assurance #A3596-01; MGH protocol 2005N000051). To ameliorate suffering, frogs were anesthetized by immersion in 0.2% tricaine prior to harvesting oocytes (Sigma-Aldrich, St. Louis, MO). Laparotomy wounds were minimized (,0.5 cm), and infiltrated with 0.25% bupivicaine to provide post-procedure analgesia. To reduce the number of frogs used, each frog was subjected to a maximum of six oocyte harvests with at least 8 weeks recovery between procedures.
Chemicals R(+)-Etomidate (2 mg/ml in 35% propylene glycol:water) was from Bedford Laboratories (Bedford, OH). Propofol was purchased from Sigma-Aldrich (St. Louis, MO) and stored as stock solution in DMSO. Alphaxalone was purchased from MP Biomedical (Solon, OH) and prepared as a stock solution in DMSO. Drugs were diluted into electrophysiology solutions on the day of use. Maximal experimental concentrations of propylene glycol (,4%) and DMSO (#0.1%) produced no functional effects on GABA A receptors [1]. Picrotoxin (PTX; from Sigma-Aldrich) was dissolved in electrophysiology buffer (2 mM). p-Chloromercuribenzenesulfonic acid sodium salt (pCMBS) was purchased from Toronto Research Chemicals (North York, Ontario, Canada). Salts and buffers were purchased from Sigma-Aldrich.

Oocyte Electrophysiology
Messenger RNA synthesis and Xenopus oocyte expression were performed as previously described [16]. Electrophysiological experiments were done at 21 to 23uC. All drugs were delivered in ND96 electrophysiology buffer (in mM: 96 NaCl, 2 KCl, 0.8 MgCl 2 , 1.8 CaCl 2 , 5 HEPES, pH 7.5). Peak current responses to varying GABA concentrations (range 0.1 mM to 10 mM) alone or co-applied with anesthetics, were measured in Xenopus oocytes using two microelectrode voltage clamp electrophysiology, as previously described [12]. The duration of GABA application varied depending on the time to reach steady-state peak current. Responses to maximal GABA (1 to 10 mM), were recorded every 2 nd or 3 rd sweep for normalization. Picrotoxin-sensitive spontaneous channel activity was measured by applying 2 mM PTX, followed by .5 minute washout and a maximal GABA response test. Etomidate (10 mM) or alphaxalone (2 mM) were used as gating enhancers with high GABA concentrations to estimate maximal 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).
The pCMBS concentrations used for modification experiments were chosen so that initial modification rate conditions (less than 50% of maximal effect) were maintained at up to 40 s exposure. Oocytes were repetitively stimulated with GABA pulses every five minutes until at least three sequential current responses were constant (65%). For modification and protection experiments, oocytes were exposed to pCMBS (alone, with GABA, with anesthetic, or with GABA+anesthetic) for 5 or 10 s, followed by 5-10 min wash in ND96. Responses to both low GABA (EC10) and high GABA (1 to 3 mM) were tested after each cycle of pCMBS exposure and wash. After up to ten cycles of modification and wash, maximal modification effect was checked using 106 higher pCMBS or 100 s pCMBS exposure. Modification rate analysis was performed on data from individual oocytes. For b2N265C receptor modification, spontaneous current after modification was plotted against [pCMBS] 6time and fitted with single exponential functions. For a1M236C modification, the ratio of low GABA to high GABA responses were calculated, normalized to the premodification control, and plotted against [pCMBS] 6 time (mM6s). Linear least squares fits to the first three to five points were used to determine the initial modification rate in M 21 s 21 . A subset of oocytes were modified first in the absence of etomidate, then in the presence of etomidate. Modification rates in both conditions were independently fitted with linear least squares.

Electrophysiological Data Analysis
Analyses for agonist concentration-responses, etomidate-induced left shift, and allosteric co-agonist model fitting followed our approach described elsewhere [16,27]. Experimental peak 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 (Prism v.5, Graphpad Software): where EC 50 is the half-maximal activating concentration and nH is Hill slope. Etomidate-dependent direct activation of receptors was analyzed similarly. EC 50 shift ratios were calculated from the difference in log(GABA EC 50 ) values (Dlog(EC 50 )) measured in the presence of 3.2 mM etomidate versus control.
PTX-sensitive leak currents (I PTX ) normalized to I max GABA ( IPTX I max GABA ) provided estimates of basal open probability. GABA efficacy was estimated based on enhancement of maximal GABA responses by etomidate or alphaxalone [27].
The estimated fraction of activated receptors, P est open corrected for both basal activity and maximal GABA efficacy, was calculated as previously described [35]: A Monod-Wyman-Changeux (MWC) co-agonist mechanism with two equivalent sites each for GABA and etomidate ( Fig. 1D In Eq. 3, L 0 is a dimensionless basal closed:open gating equilibrium variable, K G and K E are dissociation constants for respectively, GABA and etomidate binding to inactive receptors, and c and d are the respective (dimensionless) ratios of dissociation constants in activated versus inactive receptors. The maximal agonist efficacies of GABA and etomidate are inversely related to, respectively, L 0 c 2 and L 0 d 2 .

Estimation of etomidate binding affinity from cysteine protection
Because a1M236C is fully protectable by etomidate [12] or propofol, anesthetic-dependent inhibition of the a1M236C modification rate reflects anesthetic site occupancy, as we reported for b2M286C [14]. We therefore plotted a1M236C modification rate results as a function of etomidate concentration and fitted logistic functions (Eq. 1) to the data using non-linear least squares with a Hill slope of 1.0. The 50% protection concentration (PC 50 ) is assumed to be the anesthetic dissociation constant (K E ).

Structural Homology Modeling
A model of the human a1b3c2 GABA A receptor was constructed using the Prime module in Schrödinger Maestro 9.7 (release 2014-1, run on the SBGrid Consortium cluster housed at Harvard Medical School, Boston, MA). The structure was based on the GluCl structure template PDB 3RHW [20] after removal of co-crystallized ivermectin and antibody fragments. Separate amino acid sequence alignments of each GABA A subunit to GluCl were performed using the ClustalW algorithm. Alignments were gap-free in all M1 through M3 domains. The intracellular ends of GABA A subunit M3 and M4 were established by the strong alignment of homologous sequences in the middle of these domains and the length of the helices in the GluCl crystal structure. Intervening M3-M4 loops of GABA A subunit sequences were then truncated and replaced with that of the modified GluCl used for crystallography. Alignment introduced a modest number of gaps into extracellular domains and M4 sequences. Some of these were edited out or moved to preserve local GluCl secondary structure. The final sequence alignments ( Figure S1) are similar to those of Bertaccini et al [28]. Subunit monomer models were built using the ''knowledge-based methods'' option in Prime, and the heteropentamer model was assembled from five subunit monomer models in the established pentameric arrangement [6]. Model refinement used VSGB solvation [36] and a simulated membrane forcefield (dielectric = 80) bracketing the transmembrane helices.
Stepwise structural energy minimization proceeded first with nontemplate loops, then non-conserved sidechain rotamers, and finally full all-atom minimization. Molecular graphics images were produced using UCSF Chimera software version 1.8.1.

Statistical Analysis
Results are reported as mean 6 standard deviation unless otherwise noted. Statistical comparisons of results for three or more groups were performed using ANOVA with Tukey's post-hoc test. Pairwise comparisons were performed using Students t-tests. Statistical significance was inferred at p,0.05.

Structural Homology Modeling
A homology model for a1b3c2L GABA A receptors based on GluCl is depicted in Fig. 1, panels A through C. This model is similar to other GABA A receptor homology models based on both GLIC and GluCl [9,12,28,37]. The transmembrane structure of b3 subunits in our model closely matches that of crystallized b3 homomeric GABA A (PDB 4COF) [7]. Residues thought to contribute to etomidate or propofol binding (aL232, aM236, aT237, aI239, bM286, and bV290) all appear in or near the b+/ a2 transmembrane interfacial pockets of the model structure. In addition, bN265 (M2-159) is also adjacent to the b+/a2 interface of the model.

Functional characteristics of a1b2N265Cc2L GABA A receptors
Our initial approach to testing steric interactions between bN265 and etomidate followed the substituted cysteine modification and protection strategy that we and others have used to test anesthetic interactions with GABA A receptor transmembrane domains [12][13][14]32]. We electrophysiologically characterized a1b2N265Cc2L GABA A receptors for spontaneous activity, GABA EC 50 and efficacy, and sensitivity to both etomidate modulation and direct activation (Fig. 2). Oocytes expressing a1b2N265Cc2L receptors produced GABA-activated chloride currents with GABA EC 50 similar to a1b2c2L receptors ( Fig. 2A, solid symbols, Table 1). However, etomidate at a concentration that produces general anesthesia in tadpoles (3.2 mM) did not significantly enhance GABA-activated a1b2N265Cc2L currents and did not alter GABA EC 50 ( Fig. 2A, open symbols). High concentrations of etomidate (.30 mM) elicited very small currents with maximal amplitudes around 1 to 2% of maximal GABA responses (Fig. 2B), consistent with a prior report by McCracken et al [32]. Picrotoxin-sensitive spontaneous chloride leak in a1b2N265Cc2L receptors was not detectable (Fig. 2C) and, thus less than experimental noise (,0.2% of maximal GABA-elicited currents), indicating a very low spontaneous opening probability. In contrast to etomidate, alphaxalone (2 mM) strongly modulated low GABA responses in a1b2N265Cc2L receptors, and enhanced maximal GABA currents by 5 to 10% (Fig. 2D). This indicates that over 90% of a1b2N265Cc2L receptors are activated at GABA concentrations of 1 to 10 mM, which is slightly higher than GABA efficacy estimates for a1b2c2L (Table 1). Thus, a1b2N265Cc2L receptors, like a1b2N265Mc2L [27], display a functional phenotype with basal activity and GABA sensitivity similar to wild-type receptors, but extremely low sensitivity to etomidate. These functional data do not indicate whether insensitivity to etomidate is due to low drug affinity, low drug efficacy, or both.

Etomidate does not protect b2N265C from pCMBS modification
Exposing a1b2N265Cc2L GABA A receptors to a water soluble thiol-reactive probe, para-chloromercuribenzene sulfonate (pCMBS), produced currents that did not fully reverse with pCMBS washout (Fig. 3A). These currents were blocked by picrotoxin (not shown), indicating that covalent modification of b2N265C with pCMBS irreversibly activated a1b2N265Cc2L channels. Plotting the post-wash current against cumulative pCMBS exposure resulted in an average apparent modification rate of 9506155 M 21 s 21 (n = 6). Addition of GABA increased the apparent rate of modification to 38006600 M 21 s 21 , (n = 3). In contrast, control experiments in wild-type a1b2c2L receptors revealed no functional effects of 2 mM pCMBS exposures up to 60 s (120 mM6s; n = 5; data not shown).
Absent any data to guide estimation of etomidate affinity in a1b2N265Cc2L receptors, we tested protection using a range of concentrations up to 300 mM, which fully protected b2M286C from modification in both the absence and presence of GABA [14]. However, in a1b2N265Cc2L receptors, co-administration of pCMBS with etomidate at up to 300 mM, with or without GABA, did not alter the rate or extent of the pCMBS-receptor reaction (Fig. 3, B-D).
There are several hypotheses for this lack of etomidate protection at bN265C: 1) etomidate may bind normally, but not near the bN265 residue, implying that the bN265C mutation eliminates drug transduction/efficacy; 2) the bN265C mutation may reduce affinity for etomidate so that site occupancy remains low at concentrations up to 300 mM; and 3) etomidate may bind near bN265, but without obstructing pCMBS access. To address some of these possibilities, we assessed the effects of combining bN265M with previously characterized second mutations in the etomidate site. In order to reduce potential receptor assembly variations caused by subunit-subunit interfacial mutations, we also used concatenated dimer and trimer subunit assemblies for these double-mutant experiments.
A b+/a2 site occupancy reporter reveals that bN265M mutations reduce etomidate affinity We previously reported that GABA A receptors with cysteine substitutions at a1M236 are modified by pCMBS and that bound etomidate blocks a1M236C modification [12]. Thus, as we showed for bM286C [14], anesthetic-dependent reduction in the a1M236C modification rate reflects b+/a2 site occupation. For the experiments described here, we preferred a1M236C as a site occupancy reporter because, unlike bM286C, it maintains etomidate sensitivity [12]. We first demonstrated that, in GABA A receptors assembled from concatenated subunit dimers and trimers, a1M236C is both pCMBS modifiable and protected by both etomidate and propofol (Fig. 4). We then examined anesthetic protection at a1M236C in receptors assembled from concatenated dimers and trimers with bN265M mutations, for comparison (Fig. 5).
Voltage-clamped oocytes expressing b3-a1M236C dimers and b3-a1M236C-c2L trimers produced GABA-activatable receptorchannels that were both modulated and directly activated by etomidate (Fig. 4A, B; Table 1). Etomidate efficacy as an agonist was similar to that of GABA (Fig. 4B, which was 6364.4% (n = 3) based on enhancement of maximal currents at high GABA (Fig. 4A, C). These pharmacological responses were quantitatively consistent with an MWC allosteric co-agonist model (Fig. 4D). Exposing b3-a1M236C/b3-a1M236C-c2L receptors to pCMBS   irreversibly increased their sensitivity to low GABA (20 mM < EC10-20) relative to high GABA, indicating enhanced channel gating (Fig. 4E). The apparent rate of a1M236C modification increased about 3-fold when pCMBS was co-applied with GABA, and was further accelerated by addition of alphaxalone, a positive allosteric modulator that does not interact with etomidate or propofol sites (Fig. 4F, G). Because etomidate directly activates b3-a1M236C/b3-a1M236C-c2L receptors, we did not test etomidate protection in the absence of GABA. Etomidate at 3 mM did not significantly alter the apparent rate of a1M236C modification in GABA-activated receptors (Fig. 4G). We hypothesized that this result reflected two opposing effects of etomidate, as we have previously reported [12,14]: etomidate both increased the fraction of activated channels relative to GABA alone (see Figs. 4A, C), while also blocking pCMBS access. Therefore, we used the GABA+alphaxalone modification rate as a control for analysis of etomidate-dependent protection results, because the fraction of activated receptors was maximized when GABA was co-applied with either alphaxalone or etomidate. Relative to the rate observed with GABA plus alphaxalone, 3.2 mM etomidate slowed modification about three-fold, and 32 mM etomidate reduced the modification rate over twenty-fold (Fig. 4G). Logistic analysis of these modification rate data indicated 50% protection at 1.07 mM etomidate (PC 50 ; 95% CI = 0.24 to 4.7 mM).
A b+/a2 site mutation that mimics bound anesthetic reveals that bN265M impairs transduction To study the effect of bN265M mutations on b+/a2 site transduction, without the need to establish drug site occupancy, we used another mutation. The azi-etomidate photolabeled a1M236 sidechain abuts the b+/a2 cleft where etomidate binds, and a1M236W mutation mimics the positive allosteric effects of anesthetic binding while reducing etomidate modulation [16]. Thus, the tryptophan substituted sidechain at a1M236 likely occupies the same space as bound etomidate, and shares transduction mechanisms that enhance channel gating.

Major findings
We used both direct and indirect structure-function strategies to investigate the role of the GABA A receptor bN265 (M2-159) residue in the binding and efficacy of etomidate and propofol within their established b+/a2 interfacial sites. Cysteine substitution at bN265 produced receptors that were profoundly insensitive to etomidate (Fig. 2). Moreover, etomidate at concentrations up to 300 mM did not protect bN265C from covalent thiol modification (Fig. 3). The absence of etomidate effects obscures whether bN265C alters anesthetic binding vs. efficacy. However, the impact of bN265M mutations was revealed in studies of receptors containing second ''reporter'' mutations within the b+/ a2 anesthetic sites. These indirect experiments show that bN265M mutations both weaken anesthetic binding and impair b+/a2 site transduction (efficacy).

Anesthetic binding versus efficacy in GABA A receptors
The challenge of distinguishing whether ligand-gated ion channel mutations affect agonist binding, efficacy, or both, is well-known [40]. Both etomidate and propofol are allosteric agonists at GABA A receptors [1,2]. Within the MWC mechanistic framework, mutations can affect apparent sensitivity to anesthetics in three ways [41]: 1) altering the basal inactive-active equilibrium (L 0 in Eq. 3), 2) altering anesthetic binding to inactive receptors (K E ), and 3) altering anesthetic efficacy (d), which is equivalent to selectively altering binding to activated receptor states (dK E ). Both current and previous [25,27,32] functional studies reveal that bN265C and bN265M mutations eliminate etomidate sensitivity, providing no information regarding changes in drug binding or efficacy. To investigate drug occupancy (binding) of b+/a2 anesthetic sites independent of drug effects, we used a modifiable and protectable reporter cysteine, a1M236C. Concatenated b3-a1M236C/b3-a1M236C-c2L receptors, like free subunit a1M236Cb2c2L receptors [12], retained sensitivity to both etomidate modulation and direct activation. MWC model analysis indicated that the respective dissociation constants for etomidate in inactive (K E ) and active (dK E ) receptors are approximately 50 mM and 0.5 mM (Figs. 4D, 7A). Thus, etomidate allosteric efficacy (d) in this receptor is approximately 0.01 (Figs. 4D, 7A), comparable to estimates for both wild-type receptors [1] and free subunit a1M236Cb2c2L receptors [12]. Etomidate occupation of its two sites per b3-a1M236C/b3-a1M236C-c2L receptor is thus predicted to shift the closed:open equilibrium by d 22 < 10,000-fold toward open states (Fig. 7A). This accounts for the robust etomidate agonism observed in these channels (Fig. 4B). The MWC estimate for etomidate affinity in activated b3-a1M236C/ b3-a1M236C-c2L receptors (dK E < 0.5 mM) is also in good agreement with that for free-subunit a1M236Cb2c2L receptors(dK E < 1 mM) [12] and our current etomidate protection results (PC 50 < 1 mM; Fig. 5).
In comparison, protection experiments in double-mutant concatenated b3N265M-a1M236C/b3N265M-a1M236C-c2L receptors indicated only ,50% occupancy at 300 mM etomidate, implying extremely low-affinity binding to both inactive (K E < 300 mM) and GABA-activated (dK E < 300 mM) receptors (Fig. 7B). Etomidate's similar affinity for both inactive and active receptors in the double mutant receptors implies an allosteric efficacy near 1.0 (Fig. 7B). This means that etomidate binding to receptors with bN265M mutations doesn't shift the closed:open state distribution, consistent with the observed absence of drug effects on channel activity. This estimated effect of bN265M on etomidate efficacy (10,000-fold reduction for two sites) is much larger than the efficacy effect (25-fold reduction for two sites) that was previously estimated for b2N265S mutations [27]. Comparing results for a1M236Cb3c2L and a1M236Cb3N265Mc2L receptors (Fig. 7) suggests that methionine substitution reduces etomidate affinity for inactive receptors about 6-fold (K E < 50 mM vs. 300 mM), while efficacy (d) is weakened about 100-fold per site. Together, reductions of both drug affinity and efficacy account for the approximately 600-fold reduction in apparent etomidate binding affinity for GABA-activated receptors (dK E < 0.5 mM vs. 300 mM).
Our evidence demonstrating reduced etomidate affinity in both inactive and GABA-activated receptors is consistent with a direct role for bN265 in drug binding. This interaction is also suggested by in silico docking calculations [12]. Additional indirect support for this hypothesis comes from studies of M2-159 residues on other pLGICs that contact alcohols and anesthetics [30,39,42]. In addition, other outer M2 domain residues may also contribute to anesthetic sites in GABA A receptors [31] and related pLGICs [30,43,44]. However, the indirect nature of our evidence cannot rule out the possibility that bN265 mutations indirectly influence etomidate binding.
The propofol sites in b+/a2 interfaces overlap with the etomidate sites Recently, both aM236 and bM286 in a1b3 receptors were identified as incorporation sites for a propofol photo-label analog [10]. Our protection studies at a1M236C confirm this locus as a propofol contact residue. Propofol also protects bM286C from thiol modification [13]. Azi-etomidate photolabeling [8] and SCAM protection [12] identify both a1M236 and bM286 as etomidate contact points, and propofol inhibits azi-etomidate photolabeling at these residues [45]. Thus, both aM236 and bM286 are common contact points for both etomidate and propofol. Our results (Figs. 4, 5) also show that bN265M reduces b+/a2 site affinity for propofol. Bali & Akabas [13] noted that propofol modulates a1b2N265Cc2 receptors, and we also observed propofol effects in receptors with bN265M mutations. This likely reflects propofol modulation via at least two other transmembrane interfacial sites: b-/a+ and b-/c+ [45], that also bind a potent barbiturate photolabel [46] but not etomidate. The presence of at least four propofol sites per abc GABA A receptor present challenges for interpretation of its effects in mutant channels.
The bN265M mutation impairs transduction between b+/ a2 anesthetic sites and channel gating We further assessed the effect of bN265M on b+/a2 anesthetic site transduction by testing the gating effects of a1M236W, a mutation that mimics bound anesthetic [16,38]. The b2N265M mutation reduced the gating effects of a1M236W mutations about ten-fold, but did not eliminate them (Fig. 6). Extending this conclusion to bound etomidate suggests that reduced efficacy/ transduction alone may not fully account for the profound etomidate-insensitivity of receptors containing bN265M mutations. This is consistent with our result from a1M236C protection studies that also show profound effects on anesthetic efficacy combined with more modest, but significant reductions in etomidate affinity for resting state receptors.
The impact of bN265 mutations also appears to be highly selective for b+/a2 anesthetic site ligands, further supporting its involvement through local steric interactions. Both GABA EC 50 and basal channel gating are weakly altered by b2N265S or bN265M mutations [27], and our current results indicate that the same is true for bN265C mutations. Based on our data and others [32], receptors with bN265 mutations also retain modulation by alphaxalone, pentobarbital, and benzodiazepines. Evidence indicates that all of these drugs act primarily via allosteric sites that do not overlap with the b+/a2 interfacial sites where both etomidate and propofol bind.

Anesthetics and bN265C protection
In contrast to negative bN265C protection results with etomidate (this study) and propofol [13], McCracken et al [32] reported that n-octanol blocks thiol modification at b2N265C. If all of these anesthetics bind in the b+/a2 sites, why don't they all protect bN265C? One explanation consistent with our results is that bN265C reduces anesthetic affinity more than bN265M; thus b+/a2 site occupancy may have been very low during our etomidate protection experiments. Secondly, the possibility that 300 mM etomidate occupies a1b2N265Cc2L sites without obstructing pCMBS modification cannot be excluded. bN265C forms disulfide bonds with other cysteine substitutions on a-M1, b-M3, and b-M1 [18], revealing potential bN265 interactions with both b+/a2 inter-subunit anesthetic sites and b intra-subunit helix bundles. Structural homology models (e.g. Fig. 1) also show bN265 oriented toward b-M3, lying between intra-and intersubunit pockets. Other pLGIC subunit helix bundles are known to form sites that anesthetics occupy [47] and the GABA A b helix bundles could be access pathways for pCMBS to reach the thiol of bN265C without interference from propofol or etomidate. If octanol, a small flexible molecule, binds in both b helix bundles and b+/a2 sites, this could explain its unique protection at bN265C. Thirdly, the small size and flexibility of n-octanol relative to propofol and etomidate may allow it to bind closer to bN265 within the b+/a2 sites. Intriguingly, high-resolution structures of GLIC mutants suggest that the relative size and occupancy of inter-subunit and intra-subunit transmembrane pockets influences whether small alcohols and anesthetics act as gating enhancers versus inhibitors [30]. Moreover, the positive modulating anesthetic sites on GLIC are formed between M2 domains on adjacent subunits, rather than between M1 and M3 domains. Thus, examination of anesthetic interactions with both intra-subunit helix bundle pockets and various parts of the intersubunit pockets of GABA A receptor may reveal distinct sites where small versus large drugs bind and act.
bN265 and etomidate photolabels If bN265 sidechains contact anesthetics, how also to explain the absence of anesthetic photolabeling at this residue? The large effects of bN265 mutations on etomidate sensitivity are fully consistent with complementary structure-function data on etomidate derivatives, if we assume limitations on their binding orientation. Modification at the ester leaving group of etomidate, including appending large photoreactive groups, preserves anesthetic activity and GABA A receptor modulating efficacy [48,49]. Photolabels of this type label residues on a-M1 and b-M3, near the lipid-protein interface. In contrast, based on inhibition of azietomidate labeling, p-trifluorodiaziryl (TFD) substitution on the phenyl ring of etomidate reduces affinity for etomidate sites ,50fold relative to a similar p-TFD-benzyl substitution at the ester position [45]. Other bulky additions to the phenylethyl group or reorientation around the chiral carbon adjacent to the phenyl group of etomidate also dramatically reduce anesthetic activity in animals and/or modulatory efficacy in GABA A receptors [49][50][51][52][53]. Together, these results suggest that the phenylethyl end of bound etomidate is oriented toward the transmembrane pore and M2 domains (i.e. bN265), in a sterically constrained posture that likely contributes to etomidate stereoselectivity.
Finally, GABA A receptors form other transmembrane anesthetic sites, including other inter-subunit pockets, intra-subunit helical bundles, and the ion channel [54,55]. Different anesthetics display distinctive specificities for these various sites [10,45]. Defining the pLGIC structural elements involved in site-selective anesthetic binding will be essential to understanding the mechanisms of these important drugs, while also informing development of more receptor sub-type selective modulators.

Supporting Information
Figure S1 Alignment of GABA A receptor subunit amino acid sequences with GluCl. Each subunit sequence was independently aligned as described in methods. Predicted secondary structure domains are indicated by blue bars for beta sheet and red bars for alpha-helix. (TIF)