Multiple Propofol-binding Sites in a γ-Aminobutyric Acid Type A Receptor (GABAAR) Identified Using a Photoreactive Propofol Analog*♦

Background: Propofol binding to GABAAR sites of uncertain location potentiates receptor function and produces anesthesia in vivo. Results: A photoreactive propofol analog identifies propofol-binding sites in α1β3 GABAARs. Conclusion: Propofol binds to each class of intersubunit sites in the GABAAR transmembrane domain. Significance: This study demonstrates that propofol binds to the same sites in a GABAAR as etomidate and barbiturates. Propofol acts as a positive allosteric modulator of γ-aminobutyric acid type A receptors (GABAARs), an interaction necessary for its anesthetic potency in vivo as a general anesthetic. Identifying the location of propofol-binding sites is necessary to understand its mechanism of GABAAR modulation. [3H]2-(3-Methyl-3H-diaziren-3-yl)ethyl 1-(phenylethyl)-1H-imidazole-5-carboxylate (azietomidate) and R-[3H]5-allyl-1-methyl-5-(m-trifluoromethyl-diazirynylphenyl)barbituric acid (mTFD-MPAB), photoreactive analogs of 2-ethyl 1-(phenylethyl)-1H-imidazole-5-carboxylate (etomidate) and mephobarbital, respectively, have identified two homologous but pharmacologically distinct classes of intersubunit-binding sites for general anesthetics in the GABAAR transmembrane domain. Here, we use a photoreactive analog of propofol (2-isopropyl-5-[3-(trifluoromethyl)-3H-diazirin-3-yl]phenol ([3H]AziPm)) to identify propofol-binding sites in heterologously expressed human α1β3 GABAARs. Propofol, AziPm, etomidate, and R-mTFD-MPAB each inhibited [3H]AziPm photoincorporation into GABAAR subunits maximally by ∼50%. When the amino acids photolabeled by [3H]AziPm were identified by protein microsequencing, we found propofol-inhibitable photolabeling of amino acids in the β3-α1 subunit interface (β3Met-286 in β3M3 and α1Met-236 in α1M1), previously photolabeled by [3H]azietomidate, and α1Ile-239, located one helical turn below α1Met-236. There was also propofol-inhibitable [3H]AziPm photolabeling of β3Met-227 in βM1, the amino acid in the α1-β3 subunit interface photolabeled by R-[3H]mTFD-MPAB. The propofol-inhibitable [3H]AziPm photolabeling in the GABAAR β3 subunit in conjunction with the concentration dependence of inhibition of that photolabeling by etomidate or R-mTFD-MPAB also establish that each anesthetic binds to the homologous site at the β3-β3 subunit interface. These results establish that AziPm as well as propofol bind to the homologous intersubunit sites in the GABAAR transmembrane domain that binds etomidate or R-mTFD-MPAB with high affinity.

helix (␤M2-15Ј) and four of the M3 helix (␤M3-4Ј, ␤3Met-286), numbered relative to the conserved Arg and Asp near the N terminus of each subunit's M2 and M3 helices, respectively (7,8). In addition, positions in ␤M4 (9) and in the ␣ subunit cytoplasmic domain (10) have been identified as propofol sensitivity determinants. These propofol sensitivity determinants can be located in models of heteromeric GABA A Rs constructed by homology from the recently solved structure of a homomeric ␤3 GABA A R (11) or the structures of other pentameric ligandgated ion channels, including the Torpedo nicotinic acetylcholine receptor (nAChR) (12), the prokaryotic proton-gated channel GLIC (13), the amine-gated channel ELIC (14), and the invertebrate glutamate-gated channel GluCl (15). In these models, ␤M2-15Ј and ␤M3-4Ј, positions that are also sensitivity determinants for the intravenous anesthetic etomidate, are present in a pocket at the interface between the ␤ and ␣ subunits that contains the transmitter-binding sites in the extracellular domain (referred to as the ␤ ϩ -␣ Ϫ interface) (16,17), although the other sensitivity determinants are not within that intersubunit pocket. That etomidate binds to this intersubunit site was established by the etomidate-inhibitable photoincorporation of reactive etomidate analogs into ␤M3-4Ј and ␣1Met-236 in ␣M1 in a heterogeneous population of GABA A Rs purified from bovine brain (18) and in purified human ␣1␤3 GABA A R (16).
Recently, photoaffinity labeling studies with R-[ 3 H]mTFD-MPAB, a photoreactive barbiturate, identified a second class of general anesthetic-binding sites in human ␣1␤3␥2 GABA A Rs at the ␤ Ϫ -␣ ϩ and ␤ Ϫ -␥ ϩ subunit interfaces (19). Although etomidate bound selectively to the ␤ ϩ interface sites and certain barbiturates bound selectively at the ␤ Ϫ interface sites, propofol inhibited photolabeling at both classes of sites, but only at concentrations (IC 50 ϳ40 M) that were ϳ10-fold higher than the concentrations necessary to potentiate GABA responses. This discrepancy suggests that propofol may bind with higher affinity to other, as yet unidentified, sites in the GABA A R. A reactive propofol analog (o-propofol diazirine (o-PD)) was recently shown to photoincorporate in expressed ␣1␤3 GABA A R into ␤3His-267 (␤M2-17Ј), an amino acid in the ␤ subunit M2 helix in proximity to the R-mTFD-MPAB site, but projecting into the lumen of the ion channel near the interface between the extracellular and transmembrane domains (20).
In this report, we identify propofol-binding sites in a purified human ␣1␤3 GABA A R using 2-isopropyl-5-[3-(trifluoromethyl)-3H-diazirin-3-yl]phenol (AziPm), a photoreactive propofol analog that potentiates GABA responses and acts as a general anesthetic ( Fig. 1) (21). Propofol and AziPm are nAChR inhibitors, and photoaffinity labeling of the Torpedo nAChR established propofol-inhibitable photoincorporation of [ 3 H]AziPm into two sites in the TMD as follows: an intrasubunit site in the ␦ subunit helix bundle, and a site in the ion channel (22). Propofol and AziPm are also inhibitors of GLIC, and in GLIC crystals propofol binds in the TMD in the intrasubunit pocket formed by the four transmembrane helices (23). In purified GLIC in detergent solution, propofol inhibited [ 3 H]AziPm photolabeling of amino acids in that binding pocket (24). Based upon the identification of the GABA A R amino acids photolabeled by [ 3 H]AziPm and the effects of propofol, AziPm, and o-PD on GABA A R pho-tolabeling by [ 3 H]azietomidate and R-[ 3 H]mTFD-MPAB, we found in this study that propofol, AziPm, and o-PD bind in the ␣1␤3 GABA A R to the same intersubunit sites as etomidate and R-mTFD-MPAB, i.e. the homologous sites at the ␤ ϩ -␣ Ϫ , ␣ ϩ -␤ Ϫ , and ␤ ϩ -␤ Ϫ subunit interfaces. We found no evidence of [ 3 H]AziPm photolabeling of GABA A R amino acids that would be located in intrasubunit binding pockets or in the ion channel.
Purification of Expressed ␣1␤3 GABA A Rs-␣1␤3 GABA A Rs containing the FLAG epitope at the N terminus of the ␣1 subunit were purified from a tetracycline-inducible, stably transfected HEK293S cell line (27). Briefly, membrane fractions containing 6 -10 nmol of [ 3 H]muscimol-binding sites, collected from cells growing on 40 -60 tissue culture dishes (15 cm), were resuspended at 1 mg of protein/ml and solubilized overnight in 300 ml of a purification buffer (50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 2 mM CaCl 2 , 5 mM KCl, 5 mM MgCl 2 , 4 mM EDTA, 20% glycerol, pepstatin, chymostatin, and leupeptin (10 g/ml each), 2 g/ml aprotinin, and 1 mM phenylmethanesulfonyl fluoride) supplemented with 2.5 mM n-dodecyl-␤-D-maltopyranoside. Solubilized FLAG-␣1␤3 GABA A Rs were purified by elution from an anti-FLAG M2 affinity resin in elution buffer (purification buffer supplemented with 11.5 mM cholate, 0.86 mM asolectin, and 1.5 mM FLAG peptide). For competition photolabeling studies, GABA A R was also purified by solubilization for 2.5 h in purification buffer supplemented with 30 mM n-dodecyl-␤-D-maltopyranoside followed by elution from the anti-FLAG affinity resin after washing in purification buffer supplemented with 5 mM CHAPS and 0.2 mM asolectin, as described for the ␣1␤3␥2 GABA A R (19). For both protocols, typical purification yields were ϳ1.5 nmol of purified receptor (50 -60 nM binding sites) in 15-25 ml of elution buffer. For the GABA A R purified in 0.86 mM asolectin, 11 Photoaffinity Labeling-Aliquots of FLAG-␣1␤3 GABA A Rs in elution buffer were used for analytical and preparative scale photolabeling (40 -80 l and 1 ml of ␣1␤3 GABA A R, per condition, respectively). Appropriate volumes of radiolabeled, photoreactive anesthetic solutions in methanol were transferred to glass tubes, and solvent was evaporated under an argon stream. Freshly thawed GABA A R in elution buffer was added to the tube, and radioligand was resuspended with gentle vortexing during 30 min on ice to a final [ 3 H]AziPm concentration of ϳ5 M (2.5 Ci per condition) for analytical or ϳ10 M (90 Ci per condition) for preparative scale experiments. GABA A R was equilibrated with [ 3 H]azietomidate or R-[ 3 H]mTFD-MPAB at final concentrations of 0.9 or 1.5 M, respectively. Receptors were then equilibrated for 10 min with 1 mM GABA before addition of appropriate concentrations of nonradioactive anesthetic. After further incubation on ice for 30 min, the aliquots were transferred to 96-well plastic plates (Corning catalog number 2797) or 3.5-cm diameter plastic Petri dishes (Corning catalog number 3001) for analytical or preparative photolabelings, respectively, and irradiated on ice for 30 min at a distance of 0.5 to 1 cm with a 365 nm lamp (Spectroline Model EN-16, Spectronics Corp, Westbury, NJ). Stock solutions of nonradioactive AziPm (200 mM), propofol (1 M), R-mTFD-MPAB (60 mM), and etomidate (60 mM) were prepared in methanol, and all samples were photolabeled at a methanol concentration of 0.5% (v/v).
SDS-PAGE and Subunit Fragmentation-Following irradiation, samples were mixed with an equal volume of electrophoresis sample buffer (16), incubated for 30 -60 min at room temperature, and then fractionated by SDS-PAGE on a 6% Trisglycine gel. For analytical scale labeling, samples were loaded onto wells 2 cm deep, 0.8 cm wide, and 0.15 cm thick (sample volume, 150 l). Preparative scale labeling samples were loaded onto wells 2 cm deep, 11.3 cm wide, and 0.15-cm thick (sample volume, 1.5 ml). Subunits resolved by SDS-PAGE were visualized by Coomassie Brilliant Blue stain and excised to measure incorporated 3 H (for analytical scale experiments) or eluted and digested to generate peptide fragments for sequence analysis. For analytical scale experiments, the excised subunits were incubated overnight with 200 l of deionized water and 500 l of TS-2 tissue solubilizer (Research Products), and then 3 H incorporation was determined by liquid scintillation counting after adding 5 ml of Ecoscint A (National Diagnostics).
After photolabeling on a preparative scale, GABA A R subunits were recovered from the excised gel bands as described (16) and resuspended in 200 l of digestion buffer (15 mM Tris, 500 M EDTA, and 0.1% SDS (pH 8.5)). Aliquots (ϳ90 l) from gel bands enriched in ␣1 or ␤3 subunits were digested at room temperature with 0.5 units of EndoLys-C for 14 days or 2.5 g of EndoGlu-C for 2-4 days, following which the digests were fractionated by HPLC or directly subjected to protein microsequence analysis. For chemical cleavage at the C terminus of methionines, samples immobilized on PVDF sequencing filters were treated with cyanogen bromide as described (28,29). For chemical cleavage at the C terminus of tryptophans, samples on PVDF filters were treated with BNPS-skatole as described (30), except that after precipitation of excess BNPS-skatole, the digestion solution was loaded onto a second PVDF filter, and material on the two filters was sequenced simultaneously (16). ␣1␤3 GABA A R amino acids photolabeled by R-[ 3 H]mTFD-MPAB were identified as described for the ␣1␤3␥2 GABA A R (19).
Quantification of Anesthetic Inhibition of GABA A R Photolabeling-The concentration dependence of inhibition of 3 H incorporation into GABA A R subunits was fit by nonlinear least squares using SigmaPlot to a single site model, Equation 1, where f(x) is the 3 H counts/min (cpm) incorporated into a subunit at the inhibitor total concentration x; f 0 is the subunit 3 H in the absence of inhibitor; IC 50 is the total inhibitor concentration reducing photolabeling by 50%, and f ns is the nonspecific subunit photolabeling. IC 50  Reversed-phase HPLC and Sequence Analysis-Subunit fragments generated by enzymatic digestion were fractionated by reversed phase HPLC (rpHPLC) on an Agilent 1100 binary HPLC system, using a Brownlee C4-Aquapore column (100 ϫ 2.1 mm, 7 particle size) at 40°C. Solvent A was 0.08% trifluoroacetic acid in water, and solvent B was 0.05% trifluoroacetic acid in 60% acetonitrile and 40% 2-propanol. A nonlinear elution gradient increasing from 5 to 100% solvent B in 80 min was used at a flow rate of 200 l/min, with 0.5-ml fractions collected. 3 H distribution was determined by counting aliquots (10%) of each fraction, and peptide elution was monitored by absorbance at 215 nm. The rpHPLC fractions containing peaks of 3 H were pooled and drop-loaded at 45°C onto Micro TFA glass fiber filters (Applied Biosystems). Digests of intact GABA A R subunits and selected rpHPLC fractions were loaded directly onto PVDF filters using Prosorb (Applied Biosystems) sample preparation cartridges. All filters were treated after loading with Biobrene (Applied Biosystems) before sequencing.
Samples were sequenced using a Procise 492 protein sequencer (Applied Biosystems) programmed to use 2 ⁄ 3 of the material from each cycle of Edman degradation for PTH-derivative quantification and 1 ⁄ 3 to measure the 3 H release by scintillation counting. For some samples, sequencing was interrupted at designated cycles, and the sample filter was treated with OPA to chemically isolate for further sequencing only those fragments containing a proline in the designated cycle. OPA reacts with primary amines, but not secondary amines, and treatment with OPA blocks further sequencing of any fragment not containing a proline at that cycle (31,32). The amount of PTH-derivative released (in picomoles) for a given residue was quantified using their peak height in the chromatogram, background-corrected, compared with a standard peak, and fit by nonlinear least squares to Equation 2, where I x is the mass of the peptide residue in cycle x (in picomoles); I 0 is the initial amount of peptide (in picomoles), and R is the average repetitive yield. Amino acid derivatives whose amounts could not be accurately estimated (His, Trp, Ser, Arg, and Cys) were omitted from the fit. The efficiency of photolabeling (in cpm/pmol) for a given amino acid residue was calculated by Equation 3, where cpm x is the 3 H released in cycle x.
Molecular Modeling-The Discovery studio 2.5.5 molecular modeling package (Accelrys, Inc.) was used as described (19) to dock propofol, AziPm, and o-PD in potential anesthetic binding pockets in two homology models of a human ␣1␤3 GABA A R with a ␤3-␣1-␤3-␣1-␤3 subunit order as follows: (i) a model described previously (16) derived from the crystal structure of GLIC (Protein Data Bank code 3P50); and (ii) a model based upon the recently published structure of a human homopentameric ␤3 GABA A R ((Protein Data Bank code 4COF (11)). This new model was created by replacing the ␤3 sequences of the subunits designated A and C with the human ␣1 sequence, an alignment requiring two single residue insertions in the structure at ␣1Thr-172 and ␣1Gly-185 and the removal of the cytoplasmic loop between M3 and M4 (␣1Arg-313 to ␣1Lys-383) and N-and C-terminal truncations. This model was placed within a membrane force field and partially minimized to eliminate high energy interactions induced by the ␣1 sequence replacements (two cycles of minimization, final system energy ϭ Ϫ211,878 kcal/mol). Although propofol, AziPm, or o-PD docked readily at the intersubunit anesthetic-binding site at the ␣ ϩ -␤ Ϫ interface, we were unable to dock to the ␤ ϩ -␣ Ϫ or ␤ ϩ -␤ Ϫ interfaces without modifying the side chain orientations of ␤3Asn-265 (␤M2-15Ј) and/or ␣1Met-236 (in ␣M1). After these side chains were rotated out of the pockets, propofol was placed into the ␤3-␣1 or ␤3-␤3 pocket, and the system was minimized for two cycles, followed by 10 additional cycles of minimization after removal of propofol. The CHARMm-based molecular dynamics simulated-annealing program CDOCKER was used to dock propofol, AziPm, and o-PD within the pockets in the transmembrane domain at the ␤ ϩ -␣ Ϫ , ␣ ϩ -␤ Ϫ , and ␤ ϩ -␤ Ϫ subunit interfaces and in the intersubunit pockets accessible from the ion channel in proximity to ␤3His-267 (␤3 GABA A R-based model only; no pocket in the GLIC-based model). We also docked the anesthetics at the top and bottom of the ion channel and in the intrasubunit pockets. Although no intrasubunit pockets were present in the ␤3 GABA A R-based model, in the GLIC-based model ␤3His-267 contributed to the ␤3 subunit pocket. For the intersubunit sites, randomly oriented and randomly distributed molecules of propofol (9 -30), AziPm (6 -30), and o-PD (6) were seeded within binding site spheres (12 Å radii) centered on the proposed anesthetic-binding sites defined between M2-15Ј, the conserved proline in M1 (␣1Pro-233/␤3Pro-228), and the conserved aromatic residue in M3 (␣1Tyr-294/␤3Phe-289). For each binding site, CDOCKER was set up to first generate 10 -40 random conformations for each replica using high temperature molecular dynamics, and 10 -40 random orientations of each molecule were generated within the binding site spheres. The lowest 25-100 energyminimized docking solutions, generated using simulated annealing and full potential minimization, were collected and ranked according to CDOCKER interaction energies. In the ␤3 GABA A R model, all three anesthetics were predicted to bind stably at the following: 1) each intersubunit anesthetic-binding site; 2) at each ␤3His-267-associated pocket near the ion channel; and 3) at the bottom of the ion channel at the levels of M2-2Ј-M2-6Ј. Less favorable binding was predicted in the ion channel at the level of M2-13Ј. Communication between the ␤-␤ intersubunit anesthetic sites and the ␤3His-267-associated pocket at the ␤-␤ interface near the ion channel was blocked by ␤3Pro-228 from M1 on the ␤ Ϫ side and from ␤3Thr-262 (M2-12Ј) and ␤3-Thr-266 (M2-16Ј) from M2 on the ␤ ϩ side in the crystal structure. In the GLIC-based model using CDOCKER interaction energies, AziPm and propofol were each predicted to bind with highest affinity at the ␤ ϩ -␣ Ϫ and ␤ ϩ -␤ Ϫ interface, and with lower affinity at the ␣ ϩ -␤ Ϫ interface and in the ion channel. No stable binding was predicted in the intrasubunit pockets.
Connolly surface representations defined by a 1.4-Å diameter probe of the ensemble of the 25 lowest CDOCKER interaction energy docking solutions for both propofol and AziPm are shown, along with the AziPm molecule docked with the lowest CDOCKER interaction energy, for the ␤ ϩ -␣ Ϫ , ␣ ϩ -␤ Ϫ , and ␤ ϩ -␤ Ϫ intersubunit sites. Also shown in Connolly surface representation is the ␤-␤ intersubunit pocket accessible from the ion channel in proximity to ␤3His-267.

Amino acid Control
Localization of GABA A R Structural Domains Containing Amino Acids Photolabeled by [ 3 H]AziPm-To provide an initial characterization of the GABA A R subunit regions containing photolabeled amino acids, GABA A Rs were photolabeled on a preparative scale with [ 3 H]AziPm in the absence and presence of propofol, and rpHPLC was used to fractionate EndoLys-C and EndoGlu-C digests of material eluted from the gel bands enriched in ␤3 subunits (Fig. 4) and ␣1 subunits (data not shown). For both digests, all 3 H was eluted in two broad peaks in a region of the gradient (50 -70% solvent B) reported previously to contain fragments beginning near the N terminus of the M1 and M3 helices (16), and no 3 H was recovered in the hydrophilic fractions containing fragments from extracellular domains of GABA A R subunits.
Aliquots of the EndoLys-C digests of labeled ␣1 and ␤3 subunits were sequenced to identify the cycles of Edman degradation with peaks of 3 H release indicative of the presence of a photolabeled amino acid (Fig. 5). In the ␣1 subunit digest, there were peaks of 3 H release in cycles 17 and 19. In the ␤3 subunit digest, there was a peak of 3 H release in cycle 12. Because the digests contained all subunit fragments, peaks of 3 H release could not be directly associated with specific subunit fragments. However, based upon the rpHPLC fractionation of the subunit digests, [ 3 H]AziPm was likely to be incorporated into one or more of the subunit transmembrane helices. For each subunit, digestion with EndoLys-C produced fragments beginning before the M1, M3, and M4 helices (Fig. 5). Therefore, for the ␣1 subunit digest, the peaks of 3 H release in cycles 17 and 19 would be consistent with photolabeling in ␣M1 of ␣1Ile-239 in the fragments beginning at ␣1Arg-223 or ␣1Ile-221, respectively. For the ␤3 subunit digest, the peak of 3 H release in cycle 12 was consistent with photolabeling in ␤3M1 of ␤3Met-227, the amino acid photolabeled by R-[ 3 H]mTFD-MPAB (  (Fig. 6). First, an EndoLys-C digest of the ␣1 subunit was sequenced with OPA treatment in cycle 11 (at ␣1Pro-233) to prevent further sequencing of other fragments not containing a proline at

R anesthetic-binding sites
IC 50 values, the total anesthetic concentrations resulting in 50% inhibition of photolabeling of ␣1␤3 GABA A R purified in 0.2 mM asolectin, 5 mM CHAPS, were determined as described under "Experimental Procedures" (mean Ϯ range, two independent experiments); EC 50 value for anesthesia, tadpole loss of righting reflex.   Table 2.

[ 3 H]AziPm Photolabeling in Other Transmembrane Helices-
Photolabeling within ␤M2, if it occurred, was at Ͻ15% the efficiency of ␤3Met-227, based upon the levels of 3 H released during sequencing of a sample containing the fragment beginning at ␤3Ile-242 (11 pmol), produced by BNPS-skatole cleavage at ␤3Trp-241. Photolabeling within ␣M2, if it occurred, was at Ͻ10% the efficiency of ␣1Ile-239, based upon sequence analysis of the fragment beginning before ␣M2 at ␣1Ser-251 (6 pmol) produced by subunit digestion with EndoGlu-C and then chemical isolation of that fragment during sequence analysis by treatment with OPA at cycle 3. Photoincorporation within ␣M3 was characterized when the fragment beginning at ␣1Asp-287 was sequenced along with the fragment beginning at ␣1Thr-237 (Fig. 6B). Because no peak of 3 H release was detected other than the peak in cycle 3 expected for the photolabeling of ␣1Ile-239, any photolabeling in ␣M3 was at Ͻ15% the efficiency of ␣1Ile-239.

DISCUSSION
In this study, we characterize the interactions of propofol with the ␣1␤3 GABA A R by directly identifying the GABA A R amino acids photolabeled by [ 3 H]AziPm, a photoreactive propofol analog, and by comparing propofol, AziPm, and o-PD interactions with the intersubunit anesthetic-binding sites identified by [ 3 H]azietomidate and R-[ 3 H]mTFD-MPAB (16,19). When photolabeling was analyzed at the level of GABA A R subunits, we found that propofol, AziPm, and o-PD each inhibited [ 3 H]azietomidate or R-[ 3 H]mTFD-MPAB photolabeling in a manner consistent with competitive inhibition, and con-3 Sequence analyses of ␤3 subunit fragments isolated by rpHPLC were characterized by substantial, decreasing "wash-off" of 3 H in the first 3-4 cycles of Edman degradation for samples enriched in ␤M1-␤M2 (Fig. 7) or ␤M3/ ␤M4 (Fig. 8) that was not seen for ␣1 subunit samples. The source is unknown for this incorporation that is unstable under the acid and/or base treatments necessary for Edman degradation, but there was no evidence of unstable 3 H incorporation when enzymatic digests of ␤ subunits were fractionated by rpHPLC in 0.1% trifluoroacetic acid (Fig. 4).  versely, propofol, etomidate, and R-mTFD-MPAB each inhibited [ 3 H]AziPm photoincorporation to a similar extent. 4 It was surprising that etomidate or R-mTFD-MPAB each inhibited [ 3 H]AziPm ␤ subunit photolabeling to the same extent and with IC 50 values of 1 M because they bind with high affinity and selectivity at the ␤ ϩ -␣ Ϫ and ␣ ϩ -␤ Ϫ subunit interfaces, respectively, and they do not inhibit each other's photolabeling at those sites with high affinity. However, in the ␣1␤3 GABA A R, etomidate and R-mTFD-MPAB were both potent inhibitors of R-[ 3 H]mTFD-MPAB photolabeling of ␤3Met-286 and ␤3Phe-289, but only R-mTFD-MPAB was a potent inhibitor of photolabeling of ␤3Met-227, the amino acid accounting for ϳ95% of 3 H incorporation in the ␤ subunit (Table 1). This indicates that in the ␣1␤3 GABA A R, both R-mTFD-MPAB and etomidate (16) bind with high affinity to the homologous site at the ␤ ϩ -␤ Ϫ subunit interface and that [ 3 H]AziPm may be preferentially photolabeling amino acids at that interface. Protein microsequencing identified propofol-inhibitable photolabeling by [ 3 H]AziPm of amino acids in the etomidate site at the GABA A R ␤ ϩ -␣ Ϫ subunit interface (␣1Met-236/ ␣1Ile-239 in ␣M1 and ␤3Met-286 in ␤M3) and in the R-mTFD-MPAB site at the ␣ ϩ -␤ Ϫ subunit interface (␤3Met-227 in ␤M1). ␤3Met-286 and ␤3Met-227 are also present at the ␤ ϩ -␤ Ϫ interface of an ␣1␤3 GABA A R that does not occur in an ␣1␤3␥2 GABA A R. However, AziPm clearly binds in proximity to ␤3Met-227 in ␤M1 at the ␣ ϩ -␤ Ϫ interface because AziPm inhibits R-[ 3 H]mTFD-MPAB ␤ subunit photolabeling but etomidate does not. As discussed above, the concentration dependence of R-mTFD-MPAB and etomidate inhibition of [ 3 H]AziPm ␤ subunit photolabeling indicates that [ 3 H]Az-iPm also photolabels amino acids at the ␤ ϩ -␤ Ϫ intersubunit site.
The IC 50 values for inhibition of photolabeling in the ␣ subunit by [ 3 H]azietomidate and in the ␤ subunit by R-[ 3 H]mTFD-MPAB define the apparent affinities of anesthetics for the sites at the two ␤ ϩ -␣ Ϫ and two ␣ ϩ -␤ Ϫ interface sites per ␣1␤3 GABA A R, respectively. Inhibition of [ 3 H]AziPm photolabeling in the ␤ subunit defines anesthetic affinity for the site at the ␤ ϩ -␤ Ϫ interface (Table 2) The locations of these three classes of homologous intersubunit anesthetic-binding sites in the GABA A R transmembrane domain are shown in Fig. 9 in a homology model of a ␤3␤3␣1␤3␣1 GABA A R based upon the structure of the homomeric ␤3 GABA A R (11), along with the location at the ␤ ϩ -␤ Ϫ interface of a pocket accessible from the ion channel and in contact with ␤3His-267, the residue photolabeled by o-PD (20). Also included in the figure is an alignment of ␣1 and ␤3 subunit sequences in the M1-M3 transmembrane domain with the photolabeled amino acids highlighted.
Docking and the Intersubunit Binding Sites-As in the GABA A R homology models based upon GLIC or GluCl (16,19), in the model based upon the ␤3 GABA A R, azietomidate (volume, 240 Å 3 ) and R-mTFD-MPAB (275 Å 3 ) are predicted to bind stably within each of the three classes of intersubunit pockets. Propofol and AziPm, which have the same molecular volumes (182 Å 3 ), as well as o-PD (volume, 150 Å 3 ), are also predicted to bind stably and with similar energies at each of the intersubunit pockets and also in the channel accessible pockets near ␤3His-267. It should be noted that the GABA A R we photolabeled was purified in the absence of cholesterol, although previous reconstitution studies indicate that cholesterol is essential for function (37,38). Furthermore, it is probable that cholesterol actually binds the GABA A R, and candidate sites would certainly include the intersubunit transmembrane cavities identified as anesthetic sites here (39,40). However, bound cholesterol was not localized in the ␤3 GABA A R crystal structure, although the receptor was purified in the presence of 1 M cholesterol (11), and the dimensions of the intersubunit pockets differ only subtly from those in GLIC or GluCl, purified in the absence of cholesterol (13,15).
AziPm Photolabeling of Nonintersubunit-binding Sites-Although we did not identify photolabeled amino acids in regions other than the intersubunit-binding sites, the observed pharmacological specificity of [ 3 H]AziPm photolabeling at the level of the intact ␤ subunit suggests that other photolabeled amino acids may remain to be identified. Although the efficiency of photolabeling of ␣1Ile-239 (240 cpm/pmol) was similar to the level of propofol-inhibitable ␣1 subunit photolabeling (ϳ280 cpm/pmol), the levels of photolabeling of ␤3Met-227 (ϳ20 cpm/pmol) or ␤3Met-286 (3 cpm/pmol) were much lower than the level of propofol-inhibitable ␤3 subunit photolabeling (ϳ1,000 cpm/pmol, Fig. 3). Furthermore, the high levels of 3 H released in the first cycles of Edman degradation of samples enriched in fragments containing ␤M1-␤M2 (Fig. 7) or ␤M3 (Fig. 8) (19,34), which contain the same photoreactive trifluoromethylphenyl diazirine, no evidence was seen for similar apparent instability of photolabeled residues. Further studies using mass spectroscopic sequencing techniques may identify additional photolabeled amino acids in those fragments, which are likely to be located within the intersubunitbinding sites because R-mTFD-MPAB and etomidate are potent inhibitors of that photolabeling. This pharmacological specificity of the unidentified photolabeling in the ␤3 subunit makes it highly unlikely that it results from [ 3 H]AziPm photolabeling of ␤3His-267, the amino acid photolabeled by o-PD (20). Consistent with this, photolabeling of a nAChR histidine by [ 3 H]azioctanol, an aliphatic diazirine, was readily detected by Edman degradation (42).
Although AziPm and o-PD are both trifluoromethylphenyl diazirines, they may photoincorporate by different reactive intermediates. Their UV absorption spectra differ significantly. AziPm possesses a well resolved diazirine absorption band centered at 370 nm with an extinction coefficient (⑀ 370 ϭ 670 M Ϫ1 cm Ϫ1 ) (21), similar to most trifluoromethylphenyl diazirines (43). o-PD has only a tailing absorption above 300 nm with ⑀ 350 ϭ 70 M Ϫ1 cm Ϫ1 (20). The fact that AziPm photolabels aliphatic and nucleophilic side chains is consistent with the reactivity expected for a carbene intermediate. Photoactivation of o-PD will lead to the formation of a 2-hydroxyphenylcarbene, which is predicted (44) to be 45 kcal/mol less stable than the o-quinone methide that can be formed by intramolecular rearrangement, a highly reactive electrophile that will react preferentially with nucleophilic amino acid side chains such as histidine (45).
AziPm as a Probe for Propofol-binding Sites in Pentameric Ligand-gated Ion Channels-Although the incorporation of a photoreactive diazirine in anesthetics as large as etomidate or MPAB can be accomplished with only minor perturbation of their core structures, the development of photoreactive analogs of anesthetics as small as propofol requires a more dramatic perturbation of structure. AziPm acts as a GABA A R modulator and tadpole anesthetic at similar concentrations as propofol (21). However, AziPm was of much lower efficacy than propofol as a modulator, and it inhibited GABA responses at concentrations above 3 M, although propofol potentiated even at 30 M (21). AziPm photolabeled amino acids in the propofol-binding sites identified by x-ray crystallography in apoferritin (21), a soluble model protein, and in GLIC, a proton-gated prokaryotic pentameric ligand-gated ion channel (23,24).
Propofol and AziPm both inhibit the Torpedo nAChR. However, propofol binds preferentially to the nAChR in the desensitized state (ϩ agonist), and AziPm binds preferentially in the resting, closed channel state in the absence of agonist (22). [ 3 H]AziPm photolabeling identified three binding sites in the nAChR TMD as follows: (i) an intrasubunit site within the ␦ subunit helix bundle; (ii) a site in the ion channel, and (iii) a site at the ␥-␣ interface (22). Photolabeling of the intrasubunit site, which is equivalent to the propofol and AziPm site in GLIC, occurred in the desensitized state, but not the resting state, and propofol inhibited photolabeling competitively. Propofol also inhibited [ 3 H]AziPm photolabeling in the ion channel, but this inhibition was likely to be allosteric because photolabeling was also inhibited by agonist stabilization of the nAChR in the desensitized state. Propofol clearly did not bind to the site at the ␥-␣ interface, because it potentiated rather than inhibited photolabeling. The photolabeling studies with GLIC and nAChR established that AziPm photoincorporates into a wide range of amino acid side chains, including aliphatic, aromatic, and nucleophilic, which demonstrates that it has the photoreactivity necessary to incorporate into binding sites of varying structure. The results also demonstrate, however, that in the nAChR propofol binds to some, but not all, of the sites binding AziPm.
Conclusions-Based upon [ 3 H]AziPm photolabeling of ␣1␤3 GABA A Rs, AziPm and propofol each bind to the ␤ ϩ -␣ Ϫ , ␣ ϩ -␤ Ϫ , and ␤ ϩ -␤ Ϫ intersubunit sites. Etomidate and R-mTFD-MPAB, anesthetics of complex structure, bind with Ͼ50-fold selectivity to the different classes of GABA A R intersubunit sites. In contrast, the modest differences in propofol affinity for the ␤ ϩ -␤ Ϫ , ␤ ϩ -␣ Ϫ , and ␣ ϩ -␤ Ϫ sites in the ␣1␤3 GABA A R establish that an anesthetic of such simple structure binds with little selectivity at each intersubunit site. Characterization of [ 3 H]AziPm photolabeling of ␣1␤3␥2 GABA A R (19) will provide further definition of the selectivities of propofol and other anesthetics for intersubunit sites in an ␣␤␥ GABA A R subtype representative of the heterotrimeric GABA A Rs expressed most abundantly in the brain.