Neurosteroids Allosterically Modulate Binding of the Anesthetic Etomidate to γ-Aminobutyric Acid Type A Receptors*

Photoaffinity labeling of γ-aminobutyric acid type A (GABAA)-receptors (GABAAR) with an etomidate analog and mutational analyses of direct activation of GABAAR by neurosteroids have each led to the proposal that these structurally distinct general anesthetics bind to sites in GABAARs in the transmembrane domain at the interface between the β and α subunits. We tested whether the two ligand binding sites might overlap by examining whether neuroactive steroids inhibited etomidate analog photolabeling. We previously identified (Li, G. D., Chiara, D. C., Sawyer, G. W., Husain, S. S., Olsen, R. W., and Cohen, J. B. (2006) J. Neurosci. 26, 11599–11605) azietomidate photolabeling of GABAAR α1Met-236 and βMet-286 (in αM1 and βM3). Positioning these two photolabeled amino acids in a single type of binding site at the interface of β and α subunits (two copies per pentamer) is consistent with a GABAAR homology model based upon the structure of the nicotinic acetylcholine receptor and with recent αM1 to βM3 cross-linking data. Biologically active neurosteroids enhance rather than inhibit azietomidate photolabeling, as assayed at the level of GABAAR subunits on analytical SDS-PAGE, and protein microsequencing establishes that the GABAAR-modulating neurosteroids do not inhibit photolabeling of GABAAR α1Met-236 or βMet-286 but enhance labeling of α1Met-236. Thus modulatory steroids do not bind at the same site as etomidate, and neither of the amino acids identified as neurosteroid activation determinants (Hosie, A. M., Wilkins, M. E., da Silva, H. M., and Smart, T. G. (2006) Nature 444, 486–489) are located at the subunit interface defined by our etomidate site model.

GABA A 3 receptors (GABA A R) are major mediators of brain inhibitory neurotransmission and participate in most circuits and behavioral pathways relevant to normal and pathological function (1). GABA A R are subject to modulation by endogenous neurosteroids, as well as myriad clinically important central nervous system drugs including general anesthetics, benzodiazepines, and possibly ethanol (1,2). The mechanism of GABA A R modulation by these different classes of drugs is of major interest, including identification of the receptor amino acid residues involved in binding and action of the drugs.
In the absence of high resolution crystal structures of drugreceptor complexes, the locations of anesthetic binding sites in GABA A Rs have been predicted based upon analyses of functional properties of point mutant receptors, which identified residues in the ␣ and ␤ subunit M1-M4 transmembrane helices important for modulation by volatile anesthetics (primarily ␣ subunit) and by intravenous agents, including etomidate and propofol (␤ subunit) (3)(4)(5). Position ␤M2-15, numbered relative to the N terminus of the helix, functions as a major determinant of etomidate and propofol potency as GABA modulators in vitro and in vivo (6 -8). By contrast, this residue is not implicated for modulation by the neurosteroids, potent endogenous modulators of GABA A R (9).
Photoaffinity labeling, which allows the identification of residues in proximity to drug binding sites (10,11), has been used to identify two GABA A R amino acids covalently modified by the etomidate analog [ 3 H]azietomidate (12): ␣1Met-236 within ␣M1 and ␤Met-286 within ␤M3. Photolabeling of these residues was inhibited equally by nonradioactive etomidate and enhanced proportionately by GABA present in the assay, consistent with the presence of these two residues in a common drug binding pocket that would be located at the interface between the ␤ and ␣ subunits in the transmembrane domain (12). Mutational analyses identify these positions as etomidate and propofol sensitivity determinants (13)(14)(15).
A recent mutagenesis study (16) identified two other residues in GABA A R ␣M1 and ␤M3 as critical for direct activation by neurosteroids, ␣Thr-236 (rat numbering, corresponding to ␣1Thr-237, bovine numbering used here and by Li et al. (12)) 4 and ␤Tyr-284. These residues were also proposed to contribute to a neurosteroid binding pocket in the transmembrane domain at the interface between ␤ and ␣ subunits, based upon their location in an alternative GABA A R structural model that positioned those amino acids, and not ␣1Met-236 or ␤Met-286, at the subunit interface. For GABA A Rs and other members of the Cys-loop superfamily of neurotransmitter-gated ion channels, the transmembrane domain of each subunit is made up of a loose bundle of four ␣ helices (M1-M4), with M2 from each subunit contributing to the lumen of the ion channel and M4 positioned peripherally in greatest contact with lipid, as seen in the structures of the Torpedo nicotinic acetylcholine receptor (nAChR) (17) and in distantly related prokaryote homologs (18). However, uncertainties in the alignment of GABA A R subunit sequences relative to those of the nAChR result in alternative GABA A R homology models (12,19,20) that differ in the location of amino acids in the M3 and M4 membrane-spanning helices and in the M1 helix in some models (16,21).
If etomidate and neurosteroids both bind at the same ␤/␣ interface in the GABA A R transmembrane domain, the limited space available for ligand binding suggests that their binding pockets might overlap and that ligand binding would be mutually exclusive. To address this question, we photolabeled purified bovine brain GABA A R with [ 3 H]azietomidate in the presence of different neuroactive steroids and determined by protein microsequencing whether active neurosteroids inhibited labeling of ␣1Met-236 and ␤Met-286, as expected for mutually exclusive binding, or resulted in [ 3 H]azietomidate photolabeling of other amino acids, a possible consequence of allosteric interactions. Active steroids failed to inhibit labeling and enhanced labeling of ␣1Met-236, clearly indicating that the neurosteroid and the etomidate sites are distinct. Our GABA A R homology model that positions ␣1Met-236 and ␤Met-286 at the ␤/␣ interface, but not that of Hosie et al. (16), is also consistent with cysteine substitution cross-linking studies (20,22), which define the proximity relations between amino acids in the ␣M1, ␣M2, ␣M3, and ␤M3 helices, and these results support the interpretation that the two residues photolabeled by [ 3 H]azietomidate are part of a single subunit interface binding pocket, whereas the steroid sensitivity determinants identified by mutagenesis neither are at the ␤/␣ subunit interface nor are contributors to a common binding pocket.

EXPERIMENTAL PROCEDURES
Solubilization and Purification of Bovine Brain GABA A Rs-GABA A R was purified as reported previously (12) on a benzodiazepine Ro7/1986-1 affinity column. Detailed conditions for solubilization and purification are described there.
Photoaffinity Labeling of Purified GABA A R-The peak [ 3 H]muscimol binding fraction (5 ml) from each affinity column elution was used for photolabeling without further dialysis or concentration. KCl was added to a final concentration of 0.1 M. An aliquot of GABA A R (ϳ40 nM [ 3 H]muscimol binding sites, 2.5 ml for preparative labeling or 0.14 ml for analytical) was equilibrated with [ 3 H]azietomidate (final concentration, 2 M) Ϯ additional drugs, incubated for 1 h in the dark (4°C), and irradiated (30 min, 365 nm). After photolabeling, the total protein was precipitated with methanol/chloroform, solubilized in sample loading buffer, and fractionated by SDS-PAGE. The resulting gel lanes were cut into 3-mm slices, and 3 H incorporation was determined either directly by liquid scintillation counting (analytical labeling) or after elution from each slice into 1 ml of elution buffer (12). Aliquots of the eluted bands were assayed for 3 H and pooled for proteolytic digestion.
Enzymatic Digestion, Reversed-phase HPLC, and Protein Microsequencing-Digestion with EndoLys-C (Princeton Separations) was carried out using 2.5 g of protease for aliquots of GABA A R subunit in 50 l of digestion buffer for 3 days at 25°C. Reversed-phase HPLC and peptide microsequencing were performed as described (12). 3 H release in each cycle of Edman degradation was determined by liquid scintillation counting of fractions for 6 ϫ 10 min. Standard error bars are included in the plots (see Figs. 1 and 2). To determine the amount of peptide sequenced, the individual residues were fit to the equation I x ϭ I 0 ϫ R x , where I x is the pmol detected in cycle x, I 0 is the starting amount of the peptide, and R is the average repetitive yield. The range of specific residue photolabeling ((2ϫ(cpm n Ϫ cpm nϪ1 )/I 0 ϫ R x ) was calculated using the I 0 ϫ R x values calculated with 95% confidence. Treatment with o-phthalaldehyde (OPA) during sequencing was as described (12,23). OPA reacts preferentially with primary rather than secondary amines (i.e. proline) and can be used at any cycle of Edman degradation to block sequencing of peptides not containing an N-terminal proline.

RESULTS AND DISCUSSION
Neuroactive Steroids Potentiate GABA A R Photolabeling-When purified bovine brain GABA A Rs photolabeled with [ 3 H]azietomidate were fractionated by SDS-PAGE, 3 H incorporation was detected in GABA A R subunit polypeptides of ϳ50 -55 kDa that was inhibitable by Ͼ90% in the presence of nonradioactive etomidate and was shown to result from labeling of ␣1Met-236 (or the homologous methionine in ␣2, -3, -5) and ␤Met-286 (12). To test for interactions between azietomidate and neurosteroids, we photolabeled GABA A R with [ 3 H]azietomidate in the presence of 1 mM GABA (which enhances binding) and included (individually) in the labeling mixture three different neuroactive steroids: 5␣pregnan-3␣-ol-20-one (allopregnanolone), tetrahydro-deoxycorticosterone (THDOC), or alphaxolone, at various concentrations (0, 0.1, 0.3, 1, 10 M) in parallel. As shown in Fig. 1, all three steroids enhanced GABA A R subunit labeling in a concentrationdependent manner. The concentration dependence of enhancement could be fit to logistic curves, yielding half-effect concentrations of 0.6, 0.9, and 0.8 M for allopregnanolone, THDOC, and alphaxalone, respectively, under the experimental conditions employed. An inactive isomer of allopregnanolone, isopregnanolone (5␣-pregnan-3␤-ol-20-one), produced no modulation of labeling (Fig. 1A), demonstrating the pharmacological specificity of the observed enhancement of labeling.
Neurosteroid Modulation of Photolabeling of ␣1Met-236/ ␤Met-286-Protein microsequencing established that the GABA A R subunit photolabeling seen in the absence or presence of GABA resulted from photolabeling of ␣1Met-236 and ␤Met-286 5 (12). As neurosteroids actually increased [ 3 H]azietomidate photolabeling of GABA A R subunits, we wanted to determine whether the presence of neurosteroids increased the 5 As discussed in Li et al. (12), because of the high degree of sequence conservation for ␣ subunit subtypes in the M1 region and ␤ subunit subtypes in the M3 region, including protease cleavage sites, the labeled amino acids cannot be assigned to specific subunit subtypes, and therefore they are referred to as ␣Met-236 (using the ␣1 subunit numbering) and ␤Met-286.

Location of Labeled Residues in GABA A R Homology Models-
The proposal that etomidate binds in the GABA A R transmembrane domain at the interface of ␤ and ␣ subunits followed from the locations of ␣1Met-236 and ␤M3-286 in a GABA A R homology model (12) derived from the structure of the Torpedo nAChR (Fig. 3A), whereas the proposal that neurosteroids bind at that interface was consistent with the locations of ␣Thr-237 and ␤Tyr-284, positions determining sensitivity to direct activation by neurosteroids, in an alternative structural model (16) (Fig. 3B) that positions neither ␣1Met-236 nor ␤Met-286 at the interface between the subunits. The location of ␣1Met-236 at the ␤-␣ interface was consistent with other models based upon the nAChR structure (19,20), and as shown in Fig. 3, the recent identification (22) of the positions in ␣M1 that can be crosslinked to ␤M286C or ␤F289C is consistent with the homology model of Li et al. (12) but not that of Hosie et al. (16). Further support for this alignment comes from Jansen and Akabas (20), who used Cys mutagenesis and sulfhydryl cross-linking to orient the ␣M3 segment relative to ␣M2, which suggests proximity between ␤M2-Thr-262 and ␤M3 residues Leu-296, Tyr-299, and Ala-300, as in Fig. 3A. The proposal that each GABA A R contains two equivalent etomidate binding sites, one at each ␤-␣ interface (12), is consistent with kinetic data obtained by electrophysiology, suggesting that two equivalent sites can mediate both agonist enhancement and direct channel gating by etomidate (24).  Photoaffinity labeling with [ 3 H]azietomidate led to the identification of ␣1Met-236 and ␤Met-286 in an unbiased search for photolabeled amino acids in GABA A R subunit digests (12). Although photoaffinity labeling is a powerful tool to identify amino acids contributing directly to drug binding sites in proteins, results can be biased because the photoreactive intermediate may react preferentially with only certain amino acid side chains, and there is also the possibility that highly reactive side chains distant from the binding site may be labeled (10,11).
The aliphatic diazirine group in [ 3 H]azietomidate results in a reactive aliphatic carbocation intermediate following irradia-tion, which reacts most efficiently with Asp, Glu, and Tyr and also with Gln, Ser, Cys, and His (23). The labeling of two relatively less reactive methionines with the etomidate analog (12) indicates that they might be the most reactive side chains within the binding site, but they are unlikely to be labeled as distant, highly reactive side chains. In addition, aliphatic carbocation intermediates are known to be highly reactive with water (t1 ⁄ 2 Ͻ1 ns (25)), further limiting the possibility of diffusional encounters with reactive side chains distant from the binding site. That the photolabeling of ␣1Met-236 and ␤Met-286 was potentiated in the presence of GABA, fully inhibitable by etomidate (12), and inhibitable by propofol 6 provides further evidence that [ 3 H]azietomidate is acting as a specific affinity label and that the photolabeling results provide a direct identification of two amino acids contributing to the etomidate binding site.
In contrast, Li et al. (12) were not able to determine whether or not the M2 residue (␤Asn-265) identified as a determinant of etomidate and propofol sensitivity (6 -8, 26) is labeled by [ 3 H]azietomidate, although propofol does not protect ␤N265C from reaction with sulfhydryl reagents, whereas it protects ␤M286C (14). In our homology model, this position (Fig. 3A) may be accessible both from the interior of the ␤ subunit helix bundle and from the etomidate binding pocket at the ␤/␣ interface, and further studies are needed to determine whether it contributes directly to the etomidate binding site and/or is crucial for the coupling between anesthetic binding and GABA A R gating.
Our results and complementary cross-linking studies (22) indicate that the neurosteroid activation determinants in ␣M1 and ␤M3 are not contributing to a common binding pocket at the interface between ␤ and ␣ subunits, although Akk et al. (27) support the data of Hosie et al. (16), implicating residues in ␣M1 for steroid effects. Further studies with photoreactive neurosteroids (28) may be necessary to identify neurosteroid binding sites.  (12), with the residues in ␣M1 and ␤M3 photolabeled by [ 3 H]azietomidate (circled residues in green) contributing to a common binding pocket at the ␤-␣ subunit interface. Also included in pink is the position in ␤M2 that functions as a determinant of etomidate/azietomidate anesthetic potency in vivo (7,29), the residues in ␣M1 and ␤M3 identified as sensitivity determinants for direct activation by neurosteroids (boxed residues in yellow) (16), and the positions in ␣M1 and ␤M3 that when mutated to Cys can form intersubunit cross-links (red and orange) (22). B illustrates the locations of those residues in an alternative GABA A R model (16).