Alphaxalone Binds in Inner Transmembrane β⁺–α⁻ Interfaces of α1β3γ2 γ-Aminobutyric Acid Type A Receptors

NEUROSTEROIDS (neuroactive steroids), including the general anesthetic alphaxalone (ALX), allopregnanolone, and tetrahydro-deoxycorticosterone, are potent rapid-acting anxiolytics, anticonvulsants, amnestics, and sedative-hypnotics.¹  These effects are linked to enhanced gating of γ-aminobutyric acid type A (GABA_A) receptors, the main inhibitory neurotransmitter receptors in mammalian brain and major molecular targets for the general anesthetics propofol and etomidate.^2,3  Typical synaptic GABA_A receptors consist of 2α, 2β, and 1γ subunits arranged βαβαγ counterclockwise, viewed from the extracellular space.⁴  Each GABA_A subunit contains an N-terminal extracellular domain and a transmembrane domain with four α helices: M1 to M4. Five M2 helices surround a receptor’s central chloride channel, while M1 and M3 helices form an intermediate ring between M2 and M4 helices. Subunit interfaces are designated β⁺–α⁻ (two per receptor), α⁺–β^–, γ⁺–β^–, and α⁺–γ^–, where + corresponds to the M3 face and – is the M1 face.

Data locating neurosteroid sites on GABA_A receptors are sparse and inconsistent (table 1).^5–22  Pharmacokinetic studies indicate that neurosteroids reach GABA_A receptors via membrane lipids.²³  Mutations in α1-M1 at α1M236, α1T237, and α1I239 reduce neurosteroid sensitivity.^5,13  These residues map to outer transmembrane β⁺–α⁻ clefts in homology models based on glutamate-gated chloride (GluCl) channels from Caenorhabditis elegans²⁴  (fig. 1)²⁵  and are identified by photolabeling and substituted cysteine modification-protection (SCAMP) studies as contacts for etomidate and propofol (table 1).²⁶  Ivermectin binds to outer transmembrane intersubunit pockets on GluCl²⁴  and triiodothyronine displaces both ivermectin and allopregnanolone from homologous GABA_A receptor sites, including the etomidate/propofol sites.²⁷  Thus, neurosteroids may act through the outer transmembrane β⁺–α⁻ pockets where etomidate and propofol bind.

Other evidence indicates that neurosteroid sites are separate from etomidate and propofol sites. Neurosteroids synergize with etomidate and its derivatives when coapplied to GABA_A receptors.^28,29  Previous SCAMP experiments find no ALX interactions at several etomidate and propofol contacts in outer transmembrane β⁺–α⁻ clefts or other homologous pockets in α1β3γ2L receptors.^5,21  Other evidence points to inner transmembrane β⁺–α⁻ neurosteroid sites. Mutations in inner α1-M1 at α1Q242 reduce neurosteroid sensitivity.^13,14  The photolabel (3α,5β)-6-azi-pregnanolone (6-AziP) incorporates in inner β3-M3 at β3F301, but this study used β3 homomeric receptors.²²  Finally, β2Y284 mutations also impair neurosteroid effects.¹³  This residue’s location in β3 crystals³⁰  and homology models (fig. 1) suggests neurosteroid sites within β3 intrasubunit helix bundles.

To test whether ALX binds in β⁺–α⁻ transmembrane clefts and to compare ALX sites to those for etomidate and propofol, we used SCAMP to assess drug contacts on β3-M2 and β3-M3 helices in α1β3γ2L receptors. Using the structure of β3 homomeric receptors³⁰  and our GluCl-based structural homology model²⁵  (fig. 1), we selected residues spanning most of the β3-M3 helix, from β3M283 (outer) to β3F301 (inner), most facing the β⁺–α⁻ interface, and several facing the intrasubunit β3 helix pocket. Our results suggest that ALX contacts β3-M3 at β⁺–α⁻ interfacial residues that are adjacent and intracellular to those for propofol and etomidate.

Oocytes were harvested from female Xenopus laevis frogs in accordance with the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health (Bethesda, Maryland). Animal use in this study was approved by the Massachusetts General Hospital Institutional Animal Care and Use Committee (Boston, Massachusetts; protocol No. 2005N000051). Frogs were housed and maintained in a veterinarian-supervised facility and anesthetized in tricaine during oocyte collection. All efforts were made to minimize suffering.

Alphaxalone was purchased from Tocris Bioscience (United Kingdom) and propofol (2,6-diisopropylphenol) was purchased from Sigma-Aldrich (USA). Both were stored as 10 mM solutions in dimethyl sulfoxide (DMSO) and diluted in electrophysiology buffer for experiments. R-Etomidate was purchased from Hospira, Inc. (USA) as a 2 mg/ml (~8.2 mM) solution in 35% propylene glycol:water and diluted in electrophysiology buffer for experiments. We have previously shown that DMSO and propylene glycol at the dilutions used during electrophysiology experiments produce no effects on GABA_A receptor function.²⁵  R-mTFD-MPAB (R-allyl-m-trifluoromethyl-mephobarbital)³¹  was stored as a 100 mM solution in DMSO and diluted in electrophysiology buffer for experiments. Para-chloromercuribenzenesulfonic acid sodium salt (pCMBS) was purchased from Toronto Research Chemicals (Canada). Fresh pCMBS stock solutions in electrophysiology buffer were prepared on the day of use and kept on ice until final dilution. γ-Aminobutyric acid (GABA), picrotoxin, salts, and buffers were purchased from Sigma-Aldrich.

Oocytes were prepared for use as previously described.⁵  Complementary DNAs encoding human α1, β3, and γ2L GABA_A receptor subunits in pCDNA3.1 expression vectors (Thermo Fisher Scientific, USA) were used. Cysteine mutations were introduced into β3 by site-directed mutagenesis using QuikChange kits (Agilent Technologies, USA). After sequencing several clones through the entire coding region, one clone for each mutant was chosen for further use. Messenger RNAs were synthesized on linearized DNA templates using mMessage mMachine kits (Thermo Fisher Scientific), purified, and combined at ratios of 1α:1β:5γ (final concentration 1 ng/nl in RNAase-free water). Oocytes were injected with ~50 ng mRNA mix and incubated in ND96 buffer (in mM: 96 NaCl, 2 KCl, 1 CaCl₂, 0.8 MgCl₂, 10 HEPES, pH 7.5) supplemented with ciprofloxacin (2 mg/ml) and amikacin (100 µg/ml) at 17°C for 48 to 72 h before electrophysiologic studies.

Electrophysiologic experiments were performed in ND96 buffer at 21 to 23°C as previously described.⁵  Oocytes were placed in a 30 µl custom flow-cell, impaled with borosilicate glass microelectrodes filled with 3 M KCl (resistance < 1 MΩ), then voltage-clamped at –50 mV (model OC-725C, Warner Instruments, USA). Superfusion solutions in ND96 were controlled by electrical valves (VC-8, Warner Instruments) and delivered at a rate of 2 to 3 ml/min from glass reservoir syringes via polytetrafluoroethylene tubing and a polytetrafluoroethylene micromanifold (MP-8, Warner Instruments). Specialized software and a digital input/output interface (pClamp 8.0 and Digidata 1322, both from Molecular Devices, USA) were used to coordinate delivery of solutions and recordings. Current signals were filtered at 1 kHz, digitized at 100 Hz, and stored on a computer disk for offline analysis.

Each mutant receptor was initially characterized to establish its sensitivity to GABA, maximal GABA efficacy, and whether it was spontaneously active. Voltage-clamped oocytes were exposed to GABA solutions (range: 0.1 µM to 10 mM) for 10 to 20 s, followed by 5 min ND96 wash. Normalization sweeps at the maximum GABA concentration for the specific receptor (greater than 10 × EC₅₀; 1 to 10 mM) were recorded every second or third experiment. At least three oocytes from two different frogs were used for each concentration-response.

Spontaneous activation of GABA_A receptors (in the absence of GABA or anesthetics) was assessed by applying 2 mM picrotoxin to voltage-clamped oocytes. Reversible outward currents during picrotoxin application represent closure of spontaneously active channels. Spontaneous activity was normalized to maximal GABA-elicited current in the same cell (N ≥ 3 cells).

Maximal GABA efficacy for each receptor was estimated by comparing peak currents elicited with maximal GABA (1 to 10 mM) to currents elicited with high GABA supplemented with either 2.5 to 5 µM ALX or 3.2 to 6.4 µM etomidate, depending on the receptor’s drug sensitivity (see GABA EC5 Enhancement section below). Agonist efficacy was calculated by normalizing maximal GABA responses to GABA + anesthetic responses in the same cell, assuming the latter represents 100% activation (N ≥ 3 cells).

Each mutant was also characterized for sensitivity to etomidate, propofol, and ALX. Voltage-clamped oocytes expressing GABA_A receptors were repetitively exposed for 20 s to GABA EC5 (eliciting ~5% of maximal GABA response) separated by 5 min wash until three stable responses (varying by less than 5%) were sequentially recorded. The oocyte was then exposed to anesthetic for 30 s, followed by 20 s exposure to a solution containing GABA EC5, combined with anesthetic at 2 × EC50 for loss-of-righting-reflexes in tadpoles: 2.5 µM alphaxalone,³²  3.2 µM etomidate,³³  or 5 µM propofol.³⁴  For each receptor type and three anesthetics, multiple measurements of current response to GABA EC5 and GABA EC5 + anesthetic were obtained in at least four oocytes from two different frogs. EC5 enhancement (mean ± SEM; N ≥ 4) was calculated from the set of individual oocyte ratios of currents measured with anesthetic present to EC5 GABA alone.

SCAMP studies followed the approach we have described previously.^5,21  In each mutant receptor, functional effects and rates of cysteine modification were assessed electrophysiologically after applications of pCMBS, either alone or together, with maximally activating GABA (1 to 10 mM). Before and after pCMBS exposures, voltage-clamped Xenopus oocytes expressing mutant receptors were exposed to first GABA EC5 (low) and then a maximally activating GABA concentration (high; 1 to 10 mM). After 5 min wash, oocytes were exposed for 10 to 20 s to pCMBS (1 µM to 1 mM), a water-soluble sulfhydryl modifying reagent, either alone or coapplied with maximal GABA (1 to 10 mM). pCMBS exposure was followed by a 3 to 5 min wash in ND96. Electrophysiologic responses to low and high GABA were then retested to assess any irreversible changes in receptor function produced by pCMBS modification (in most cases an increase in the ratio of low vs. high GABA-induced peak currents). By testing a range of pCMBS concentrations this way, we identified conditions resulting in maximal modification effects and those appropriate for studying modification rates.

To measure apparent modification rates, pCMBS exposure conditions (concentration × time) were chosen that produced about 10% of the maximal modification effect per cycle. In nearly all mutants, higher pCMBS concentrations were needed to irreversibly affect receptors when applied alone than when coapplied with GABA. Voltage-clamped oocytes were first repeatedly tested for responses to both low and high GABA, then washed for 5 min in ND96, to confirm that the response ratio was stable (less than 5% variation) before pCMBS exposure. Oocytes were then exposed for 5 to 10 s to pCMBS (with or without GABA), followed by 5 min wash and retesting for low and high GABA responses. At least three cycles of pCMBS exposure/wash/low:high GABA response testing were performed on each oocyte used for rate analysis. The series of modification cycles under the selected conditions typically produced less than 50% of the maximal modification effect. A final modification cycle was performed using 10 × pCMBS concentration for 20 s to fully modify receptors, and subsequent electrophysiologic response was assessed as the maximal modification effect.

Protection experiments were performed in the presence of maximally activating GABA, as previously described,⁵  so control modification conditions were pCMBS + GABA. Oocytes were exposed to anesthetic for 30 s followed by exposure to a solution of pCMBS + GABA + anesthetic. Postmodification wash and response tests were identical to control modification conditions (i.e., usually with no anesthetic present, but see below in this section). Anesthetic concentrations used in initial protection studies were chosen to maximize site occupancy, while enabling washout within 5 min (10 µM etomidate, 20 µM propofol, and 10 µM ALX). In receptors with β3F289, β3F293C, and β3L297C mutations, higher concentrations of anesthetics (50 µM etomidate, 100 µM propofol, or 50 µM ALX) were also used in protection experiments. Under these conditions, anesthetic washout between pCMBS exposure and testing for modification effects was extremely slow. Therefore, we used an alternative approach to low GABA responses, measuring direct activation by anesthetics alone (50 µM etomidate, 100 µM propofol, or 50 µM ALX), normalized to high GABA responses. At least two anesthetics were tested in the same manner, to test for drug-specific interactions. In the case of receptors with β3V290C mutations, we tested for allosteric effects (i.e., whether all anesthetics affect pCMBS modification similarly), by including SCAMP studies with 10 µM mTFD-MPAB, a barbiturate hypnotic that acts through GABA_A receptor sites outside the β⁺–α⁻ interfaces.^8,31  For each cysteine mutant, at least five oocytes were studied in control modification experiments and at least four oocytes were studied in each set of anesthetic protection experiments. Group sample sizes of five per group were based both on prior experience and a power analysis performed as previously described,⁵  using a one-tail Student’s t test with α = 0.017 (adjusted for three drug comparisons to each control).

Results in text and figures are mean ± SEM unless otherwise indicated.

Digitized GABA concentration-response data were corrected for baseline leak currents and digitally filtered (10 Hz low-pass, Bessel function) using Clampfit 9.0 software (Molecular Devices). Peak currents were normalized to control (maximal currents), and combined GABA data from multiple cells (N ≥ 3) was fitted with logistic equations using Prism 5.02 (GraphPad Software Inc., USA):

where EC₅₀ is the half-maximal activating GABA concentration, and nH is the Hill slope. Mean GABA EC₅₀ and 95% CI are reported. To assess whether mutations altered GABA EC₅₀ relative to wild-type, we performed sum-of-squares F-tests in GraphPad Prism 5.02, using P < 0.0045 as a statistical significance threshold (the Bonferroni correction for P < 0.05 with 11 comparisons).

To test whether mutations altered spontaneous activity and/or GABA efficacy from wild-type values, we used one-way ANOVA with post hoc Dunnett’s tests (in GraphPad Prism 5.02). To test whether mutations affected receptor sensitivities to etomidate, propofol, or ALX, EC5 enhancement data for the three equipotent anesthetic concentrations in wild-type and all functional cysteine mutants was tabulated and analyzed with two-way ANOVA and Bonferroni posttests for wild-type versus mutation for each anesthetic (GraphPad Prism 5.02).

Inferences regarding contact between receptor-bound anesthetics and substituted cysteine sidechains were made when an anesthetic inhibited pCMBS modification selectively, with at least one other anesthetic failing to inhibit modification. Apparent pCMBS modification rates were calculated from data for individual oocytes expressing cysteine mutants. Either normalized maximal GABA responses (for α1β3T262Cγ2L) or normalized low:high GABA response ratios (all other mutants) were plotted against cumulative pCMBS exposure (M × s) and fitted by linear least squares with y-axis intercepts fixed at 1.0. The linear slope, under conditions of partial modification, is presumed to be proportional to the bimolecular reaction rate between pCMBS and the substituted cysteine sulfhydryl.

For α1β3T262Cγ2L data, apparent modification rates were calculated as the absolute values of the negative fitted slopes. Absolute slopes less than 10 M^–1s^–1 (the lower limit of detection) were assigned a rate of 10 M^–1s^–1 for statistical analysis. To identify anesthetics that either accelerated or inhibited modification of each substituted cysteine, apparent rates from control and anesthetic protection studies for that mutant were log transformed, tabulated, and compared using one-way ANOVA (GraphPad Prism 5.02) with P < 0.05 as a significance threshold.

Based on both crystallographic data for β3 homomeric GABA_A receptors (Protein Data Bank 4COF)³⁰  and our α1β3γ2L structural homology model based on GluCl bound to ivermectin (Protein Data Bank 3RHW; fig. 1),^25,26  we identified nine β3-M2 and M3 helix residues facing the β⁺–α⁻ cleft: T262, T266, M283, M286, F289, V290, F293, L297, and F301. We created mutant β3 cDNAs encoding cysteine substitutions at these positions, as well as at Y284, G287, and E298, which are predicted to instead face the intrasubunit β3 helix-bundle pocket. Wild-type and mutant β3 subunits were coexpressed with wild-type α1 and γ2L subunits in Xenopus oocytes, and functionally characterized using two-microelectrode voltage-clamp electrophysiology. No GABA-activated currents were detected when β3 subunits with Y284C mutations were coexpressed with α1 and γ2L, which was consistent with prior reports.³⁵  All other mutations produced GABA-sensitive ion channels with sufficient oocyte currents elicited by 1 to 10 mM GABA (greater than or equal to 0.5 µA at –50 mV) for further experiments. Table 2 summarizes GABA EC₅₀, spontaneous activation, apparent maximal GABA efficacy, and the effect of pCMBS application in these mutant receptors, in comparison to wild-type α1β3γ2L. Six mutations (β3T266C, β3M286C, β3G287C, β3F293C, β3L297C, and β3E298C) significantly increased GABA EC₅₀, and one (β3F289C) reduced GABA EC₅₀ approximately fivefold. Four mutant receptors characterized by increased GABA EC₅₀ also exhibited significantly reduced GABA efficacy (β3M286C, β3F293C, β3L297C, and β3E298C). Like other mutations that sensitize receptors to GABA,^7,36  β3F289C was associated with both high GABA efficacy and measurable spontaneous activation. Our observations were also consistent with previous studies of β2M286C, β2G287C, and β2F289C mutations.^{20,21,35,37,38} 

GABA_A receptor mutations may alter anesthetic modulation, which can in turn affect the conditions appropriate for SCAMP tests for drug contacts. We therefore characterized each mutant receptor’s sensitivity to etomidate, propofol, and ALX by measuring anesthetic enhancement of activation by EC5 GABA. Results are summarized in f igure 2. Drug solutions of 3.2 µM etomidate, 5 µM propofol, and 2.5 µM ALX are all twice the EC50 for tadpole loss-of-righting reflexes, and also similarly enhance the gating of wild-type α1β3γ2L GABA_A receptors activated with EC5 GABA⁵  (fig. 2). Compared to wild-type, two mutations, β3M286C and β3F289C, reduced EC5 enhancement by 3.2 µM etomidate, while β3F293C, β3L297C, and β3E298C increased EC5 enhancement by etomidate. EC5 enhancement by 5 µM propofol was also reduced by β3M286C and β3F289C, as well as by β3F293C. EC5 enhancement by 2.5 µM ALX was reduced by β3F289C, β3F293C, and β3L297C.

To establish conditions for SCAMP experiments, we examined the effects of pCMBS exposure, both alone and coapplied with GABA, in each of the cysteine mutants. Wild-type α1β3γ2L receptors were unaffected by pCMBS exposure at 1 mM for 60 s (N = 4). In all but one (β3M283C) of the functional cysteine-substituted mutant receptors we studied, exposure to pCMBS alone or with maximally-activating GABA concentrations induced consistent irreversible functional changes that significantly differed from repeated baseline GABA responses before pCMBS exposure (fig. 3A–I; table 2). In α1β3T262Cγ2L receptors, pCMBS exposure similarly reduced activation by both low and high GABA (fig. 3A). In the other mutant receptors, pCMBS exposure enhanced GABA sensitivity, increasing low:high response ratios in the range of twofold to 13-fold (table 2). With the exception of β3G287C, modification in the presence of GABA required lower pCMBS concentrations than without GABA at all substituted cysteines, resulting in faster apparent modification rates (fig. 3J). Results in α1β3M286Cγ2L receptors (currents not shown in fig. 3) were consistent with earlier studies of α1β2M286Cγ2L.^20,21 

We previously have shown that SCAMP reliably identifies anesthetic contacts when drugs significantly and selectively inhibit pCMBS modification.⁵  Thus, apparent initial rates of cysteine modification in control conditions (pCMBS + GABA) were compared to rates in the presence of added ALX, etomidate, or propofol in each of the modifiable mutant receptors. We chose control pCMBS modification conditions in the presence of maximally activating GABA because: (1) GABA enhances anesthetic binding and thus site occupancy; (2) GABA accelerates pCMBS modification (fig. 3J); and (3) GABA helps to establish similar mixtures of functional receptor states in both control modification and protection experiments.^6,21  Initial protection conditions included 10 µM etomidate, 20 µM propofol, or 10 µM ALX along with GABA and pCMBS. In some mutant receptors that displayed low apparent affinity for anesthetics, we also used fivefold higher protecting anesthetic concentrations. In these cases, we used equivalent high concentrations of at least one other anesthetic to test for drug-specific protection.

Normalized modification data and rate analyses for nine mutations are shown in figure 4 and summarized in figure 4J.²¹  The apparent rate of modification of α1β3T262Cγ2L receptors (fig. 4A) was unaffected by etomidate (red symbols and lines), but accelerated by propofol (green symbols and lines). Modification of α1β3T266Cγ2L receptors (fig. 4B) was accelerated by all three anesthetics, suggesting an allosteric effect. β3M286C protection was fully consistent with previous SCAMP studies of α1β2M286Cγ2L receptors, showing that both etomidate and propofol block modification, while ALX weakly accelerates pCMBS modification (summarized in fig. 4J).^20,21  Modification of α1β3G287Cγ2L receptors (fig. 4C) was unaffected by the three anesthetics. Modification of α1β3F289Cγ2L receptors was weakly blocked by 10 µM etomidate, unaffected by 20 µM propofol, and accelerated by 10 µM ALX (data not shown). Because this mutant was insensitive to anesthetics (fig. 2), we also tested 50 µM etomidate, which inhibited the apparent rate of β3F289C modification over tenfold, while neither 100 µM propofol nor 50 µM ALX inhibited modification (fig. 4D). Modification of α1β3V290Cγ2L receptors (fig. 4E) was strongly blocked by 10 µM etomidate, 20 µM propofol, and 10 µM ALX. To test whether β3V290C modification was allosterically inhibited by anesthetics that do not bind in β⁺–α⁻ sites, we also tested the effect of 10 µM mTFD-MPAB, a potent barbiturate that selectively binds to GABA_A receptor α⁺–β^– and γ⁺–β^– transmembrane interfaces.⁸  Modification of receptors with β3V290C mutations was unaffected by 8 µM mTFD-MPAB (fig. 4J), indicating that inhibition of modification by etomidate, propofol, and ALX was likely steric rather than allosteric.

Modification of α1β3F293Cγ2L receptors was accelerated by etomidate and propofol, but unaffected by 10 µM ALX. Increasing ALX to 20 µM (fig. 4F, dashed purple lines) or 50 µM (fig. 4F, solid purple lines) resulted in significantly reduced rates of β3F293C modification in comparison to 50 µM etomidate and 100 µM propofol (fig. 4F). Modification of α1β3L297Cγ2L receptors was unaffected by low concentrations of etomidate, propofol, or ALX (not shown). Because α1β3L297Cγ2L is relatively insensitive to ALX (fig. 2), we performed additional SCAMP experiments with 50 µM ALX versus 50 µM etomidate in this mutant, revealing inhibition by ALX, but not etomidate (fig. 4G). Modification of α1β3E298Cγ2L receptors (fig. 4H) was unaffected by any of the anesthetics. Modification of α1β3F301Cγ2L receptors (fig. 4I) was weakly, but significantly, blocked by 10 to 20 µM ALX and unaffected by 10 to 20 µM etomidate.

On the opposite face of the transmembrane β⁺–α⁻ cleft, Hosie et al.¹³  identified mutant effects on neurosteroid sensitivity at three residues in α1-M1: α1T237, α1I239, and α1Q242 (table 1). We previously reported that receptors with both α1I239C and α1Q242C mutations are unaffected by pCMBS, precluding SCAMP studies.⁶  To supplement our studies of β3-M2 and β3-M3 residues, we used SCAMP to test whether ALX protects the cysteine substitution at α1T237. No inhibition of pCMBS modification rates in α1T237Cβ3γ2L receptors by 10 µM ALX was observed (data not shown), whereas 10 µM etomidate inhibited modification, in agreement with previous results.⁶ 

Our aims in this study were to assess hypothesized ALX contacts with β3 sidechains that face transmembrane β⁺–α⁻ clefts in α1β3γ2L GABA_A receptors, and to compare these with etomidate and propofol contacts. Using electrophysiology, we studied ten mutant receptors with single cysteine-substitutions in β3-M2 or β3-M3 helices, in which the sulfhydryl modifier pCMBS produced irreversible functional changes. Based on drug-specific inhibition of pCMBS modification, we infer a number of anesthetic contact residues: etomidate binds near β3M286, β3F289, and β3V290 (fig. 5A); propofol binds near β3M286 and β3V290 (fig. 5B); and ALX binds near β3V290, β3F293, β3L297, and β3F301 (fig. 5C). Mapping these residues onto our α1β3γ2L structural model (figs. 5D–I) suggests that all three anesthetics bind in transmembrane β⁺–α⁻ intersubunit clefts, with overlapping etomidate and propofol sites extending from the middle of β3-M3 (near β3V290) extracellularly (figs. 5, D and E), and the ALX site extending from β3V290 intracellularly (fig. 5F).

Single-point mutations that affect neurosteroid sensitivity in heteromeric mammalian GABA_A receptors (table 1) are found throughout the transmembrane β⁺–α⁻ cleft. Our SCAMP results for ALX provide evidence of contact with four inner β3-M3 residues facing the β⁺–α⁻ interface. The strongest prior evidence for an inner transmembrane β⁺–α⁻ neurosteroid site is β3F301 photolabeling with 6-AziP,²²  but the use of homomeric β3 receptors and failure to test if neurosteroids block 6-AziP labeling make it far weaker than studies in heteromeric receptors using photolabeling derivatives of etomidate and propofol.²⁶  Mutations at both α1I239 and α1Q242, located opposite β3F293 in our structural model (fig. 1), impair receptor sensitivity to neurosteroids^13,14,28  and α1Q242C confers insensitivity to ALX, but not to etomidate (unpublished data). The lack of pCMBS-induced effects in receptors with α1I239C and α1Q242C mutations⁶  precludes SCAMP tests and contrasts with our current findings in inner β3-M3 mutants. Other indirect support for inner transmembrane neurosteroid sites include evidence that a membrane-impermeant steroid positively modulates GABA_A receptors only when applied intracellularly.²³  Docking calculations using the β3 homomeric GABA_A receptor structure³⁰  also locate pregnanolone and allopregnanolone sites near both β3F301 and β3L297.³⁹ 

Previous functional, SCAMP, and photolabeling evidence (table 1) all locate etomidate and propofol sites in outer transmembrane β⁺–α⁻ clefts. In comparing ALX contacts in β3-M3 with those for etomidate and propofol, we found that, with the exception of β3V290C, ALX contacts were mutually exclusive with propofol or etomidate contacts. We also recently reported that etomidate contacts α1L232, and that both etomidate and propofol contact α1M236, while ALX contacts neither.⁵  Altogether, our current results indicate that ALX binds in inner transmembrane β⁺–α⁻ cleft sites abutting outer transmembrane etomidate/propofol sites, with possible contact of outer and inner sites near β3V290.

Neurosteroids enhance GABA_A receptor photolabeling by etomidate derivatives²⁸  and neurosteroid-etomidate combinations synergize in both enhancing GABA_A receptor gating and anesthetizing animals.²⁹  An allosteric mechanism for this synergy through mutual coupling of sites to channel gating is suggested by our observations that both etomidate and propofol accelerate pCMBS modification of β3F293C in the ALX sites, while ALX accelerates pCMBS modification at β3M286C in the etomidate/propofol sites. Direct contact between neurosteroids and etomidate in abutting sites could also mutually enhance drug binding, contributing to functional synergy.

Our current results extend the map of propofol and etomidate contacts on the β+ aspect of the outer β⁺–α⁻ sites (table 1; fig. 5). Functional and SCAMP results with β3M286C echoed previous studies of β2M286C.^20,21  We identified two additional etomidate contact residues, β3F289 and β3V290, while propofol protects β3V290C but not β3F289C. Thus, the β⁺–α⁻ sites for propofol and etomidate overlap, agreeing with previous SCAMP and photolabel competition results (table 1).^5,8,12  Interestingly, despite evidence that propofol and etomidate might contact β2/3N265 on the M2 helix (table 1), we found no evidence of contact at β3T262 or β3T266 that also abut β⁺–α⁻ interfaces in structural models (fig. 1).

The functional effects of both cysteine-substitution and pCMBS modification provide insight into allosteric linkages and aqueous accessibility at the residues we studied. Spanning from M286 to E298, most β3-M3 cysteine mutations altered GABA EC₅₀ and/or GABA efficacy (table 2), indicating that this region is coupled to ion channel gating. Similar observations were made in a series of α1-M1 cysteine substitutions.⁶  Cysteine mutants throughout β3-M3 were also accessible to pCMBS, indicating an aqueous pathway extending intracellularly to at least β3F301, and echoing similar findings on the β1-M2 helix.⁴⁰ 

Mutant functional analyses underlie many of the hypotheses we have tested (table 1) and it is tempting to infer drug contacts from the altered anesthetic sensitivities of cysteine mutants (fig. 2). However, we recently compared SCAMP with tryptophan mutant drug sensitivity for two photolabeled residues and four anesthetics, finding perfect agreement between SCAMP and photolabeling, but poor concordance with mutant drug sensitivities.⁵  There are multiple other examples of SCAMP identifying anesthetic contacts in GABA_A receptors that weren’t photolabeled,^5,6,26,41  but only one published report of SCAMP disagreeing with photolabeling.²⁵ 

Our SCAMP approach requires functional heterologous receptor expression, quantifiably consistent cysteine modification effects, and drug occupation of a large fraction of sites.²⁶  Even under these conditions, we cannot formally rule out allosteric effects in SCAMP experiments. However, allosteric mechanisms should strongly link the functional effects of different anesthetics to inhibition of modification in relevant mutants. Comparing figures 2 and 4J, such correlations are absent at many positions where modification was inhibited: F289, V290, F293, and F301. Moreover, drug specificity was demonstrable at every protected cysteine (fig. 4J). Thus, our SCAMP results are more compatible with a steric mechanism rather than an allosteric mechanism for inhibiting pCMBS modification. Inferences of steric interactions between receptor-bound drugs and substituted cysteines are strengthened when protection is concentration-dependent and profound. ALX protection at β3F293C, β3L297C, and β3F301C was relatively weak compared to results for etomidate, propofol, and mTFD-MPAB at some of their outer transmembrane contacts.^5,6,21  For β3F293C and β3L297C, this is attributable to low ALX affinity (see next paragraph). The β3F301C sidechain may be located at the periphery of the steroid site, limiting ALX protection at this position.

In three mutant receptors, α1β3F289Cγ2L, α1β3F293Cγ2L, and α1β3L297Cγ2L, high anesthetic concentrations demonstrated concentration-dependent block of pCMBS modification. In these mutants, weak EC5 enhancement (fig. 2) indicated weak drug binding based on the Monod-Wyman-Changeux allosteric principle that positive gating modulation reflects the relative affinity of ligands for active (open) versus inactive (closed) receptors. Thus, weak EC5 enhancement relative to wild-type implies reduced drug affinity for GABA-activated receptors and a need for high drug concentrations to occupy most binding sites. In addition, α1β3F293Cγ2L receptors were characterized by low GABA efficacy, with maximal GABA activating only about 16% of these receptors (table 2) under control modification conditions. With addition of etomidate or propofol, the fraction of activated and desensitized receptors increased, allosterically accelerating β3F293C modification (fig. 4J). Adding ALX to high GABA likely produced two opposing effects on α1β3F293Cγ2L modification: increased activation/desensitization that accelerates modification, and steric protection that inhibits modification. In initial experiments, 10 µM ALX produced approximate balance in these opposing effects, while higher ALX concentrations resulted in overall slowing of modification.

Crystallographic studies of pentameric ligand-gated ion channels reveal that small anesthetics and alcohols can occupy both intersubunit and intrasubunit transmembrane pockets.^42–44  In this study, we examined two mutations (β3G287C and β3E298C) that are predicted to face the β3 intrasubunit helix bundle pocket, in both outer and inner regions of β3-M3 (fig. 1). While we observed altered GABA sensitivity as evidence of pCMBS access and modification in these mutants, no anesthetic protection was observed (figs. 4 and 5). These results are evidence against the presence of positively modulating anesthetic sites in β3 intrasubunit pockets.

Endogenous and synthetic neurosteroids are potent neuromodulators with broad therapeutic potential. Our current SCAMP studies locate positively modulating ALX sites on α1β3γ2L GABA_A receptors in inner transmembrane β⁺–α⁻ intersubunit clefts. These neurosteroid sites are adjacent to outer transmembrane β⁺–α⁻ sites where etomidate and propofol act, suggesting both direct and indirect mechanisms for cooperativity between neurosteroids and etomidate.^28,29  Two other outer transmembrane intersubunit sites, in α⁺–β^– and γ⁺–β^– clefts, bind propofol and barbiturates.^5,8  No ligands have yet been identified for the transmembrane α⁺–γ^– cleft and the inner transmembrane portions of α⁺–β^– and γ⁺–β^– interfaces, but membrane lipids probably modulate ion channel activity by interacting with transmembrane intersubunit clefts.⁴⁵  In summary, large portions of the five transmembrane intersubunit clefts in α1β3γ2L GABA_A receptors are allosterically coupled to ion channel gating. Subregions of these clefts form sites for hydrophobic modulators that in several cases, including that of neurosteroids, display remarkable drug selectivity. Structural variations in these intersubunit interfaces also contribute to subtype-selective GABA_A receptor pharmacology.

The authors thank Youssef Jounaidi, Ph.D. (Instructor, Department of Anesthesia, Critical Care, and Pain Medicine, Massachusetts General Hospital, Boston Massachusetts) for his help with molecular biology. Karol Bruzik, Ph.D. (Department of Medicinal Chemistry and Pharmacognosy, University of Illinois, Chicago, Illinois) provided mTFD-MPAB. Keith W. Miller, D.Phil. (Department of Anesthesia, Critical Care, and Pain Medicine, Massachusetts General Hospital) and Jonathan B. Cohen, Ph.D. (Department of Neurobiology, Harvard Medical School, Boston, Massachusetts) provided helpful comments on the manuscript.

This work was supported by grant Nos. GM089745 and GM058448 from the National Institutes of Health, Bethesda, Maryland.

The authors declare no competing interests.