Identification of the Bovine γ-Aminobutyric Acid Type A Receptor α Subunit Residues Photolabeled by the Imidazobenzodiazepine [3H]Ro15-4513*

Ligands binding to the benzodiazepine-binding site in γ-aminobutyric acid type A (GABAA) receptors may allosterically modulate function. Depending upon the ligand, the coupling can either be positive (flunitrazepam), negative (Ro15-4513), or neutral (flumazenil). Specific amino acid determinants of benzodiazepine binding affinity and/or allosteric coupling have been identified within GABAA receptor α and γ subunits that localize the binding site at the subunit interface. Previous photolabeling studies with [3H]flunitrazepam identified a primary site of incorporation at α1His-102, whereas studies with [3H]Ro15-4513 suggested incorporation into the α1 subunit at unidentified amino acids C-terminal to α1His-102. To determine the site(s) of photoincorporation by Ro15-4513, we affinity-purified (∼200-fold) GABAAreceptor from detergent extracts of bovine cortex, photolabeled it with [3H]Ro15-4513, and identified 3H-labeled amino acids by N-terminal sequence analysis of subunit fragments generated by sequential digestions with a panel of proteases. The patterns of 3H release seen after each digestion of the labeled fragments determined the number of amino acids between the cleavage site and labeled residue, and the use of sequential proteolytic fragmentation identified patterns of cleavage sites unique to the different α subunits. Based upon this radiochemical sequence analysis, [3H]Ro15-4513 was found to selectively label the homologous tyrosines α1Tyr-210, α2Tyr-209, and α3Tyr-234, in GABAA receptors containing those subunits. These results are discussed in terms of a homology model of the benzodiazepine-binding site based on the molluscan acetylcholine-binding protein structure.

Interactions of ␥-aminobutyric acid (GABA) with GABA A receptors are modulated by many drugs including benzodiazepines, barbiturates, anesthetics, alcohols, and neurosteroids (8,9). Benzodiazepines are of significant therapeutic interest and are among the most widely prescribed drugs in the world, used to treat anxiety and insomnia, to induce anesthesia, and to reduce seizure activity. The binding of many clinically useful benzodiazepines, such as flunitrazepam, is positively coupled to the agonist-binding sites and results in an allosteric potentiation of apparent GABA binding affinity as measured electrophysiologically. Other drugs, such as the imidazobenzodiazepine Ro15-1788 (flumazenil), can bind to the benzodiazepine site without energetic coupling to the GABA site (zero modulator or benzodiazepine antagonist), whereas for Ro15-4513, an azide analog of Ro15-1788 and the focus of this study, there is negative allosteric coupling with GABA binding (negative modulator or benzodiazepine inverse agonist) for some GABA A receptor subunit compositions (5, 8 -11).
Benzodiazepines bind to a distinct binding site on the GABA A receptor located between an ␣ and the ␥ subunits. The benzodiazepine-binding site is homologous to the GABA-binding sites located at the ␤-␣ interfaces, with many binding site residues conserved (12,13). Mutational analyses have identi-fied amino acids in three discrete regions of the primary structure of an ␣ subunit (14) and three regions of a ␥ subunit (15)(16)(17) that function as determinants of benzodiazepine binding affinity and/or allosteric coupling to the agonist-binding site. Mutational analyses allow the identification of amino acids that are important determinants of the energetics of ligand binding, but these amino acids need not be direct contributors to the ligand-binding site.
Photoaffinity labeling provides an alternative experimental approach to identify amino acids in proximity to a bound ligand (reviewed in Refs. 18 and 19). For the nicotinic ACh receptor, photoreactive agonists and antagonists have provided extensive identification of amino acids contributing to the agonistbinding sites and to the ion channel (see Refs. 20 and 21 and references therein). For the GABA A receptor, irradiation at 254 nm results in the covalent incorporation of the agonist [ 3 H]muscimol, with Phe-65 in the bovine ␣ 1 subunit identified as a labeled amino acid (22). The benzodiazepine [ 3 H]flunitrazepam photoincorporates into several ␣ subunit isoforms (23)(24)(25), and ␣ 1 His-102 has been identified as the primary site of photoincorporation within the ␣ 1 subunit (26,27). [ 3 H]Ro15-4513 can also be photoincorporated into ␣ subunits of the GABA A receptor (10), and its site(s) of incorporation was shown to be C-terminal to ␣ 1 His-102 (28) and possibly within the transmembrane domains of the receptor (29). Ro15-4513 contains a photoreactive azide that upon UV excitation forms a reactive nitrene. Therefore [ 3 H]Ro15-4513 should photolabel amino acids in close proximity to the azide, and identification of those residues will provide a first definition of the orientation of Ro15-4513 within the benzodiazepine-binding site.
In this report we identify the amino acids photolabeled by [ 3 H]Ro15-4513 in GABA A receptors purified by affinity chromatography from detergent extracts of bovine cerebral cortex. Labeled amino acids in the ␣ 1 , ␣ 2 , and ␣ 3 subunits were determined by use of a protein sequencing strategy that depended upon the pattern of 3 H release rather than the direct identification of GABA A receptor subunit amino acids.
Affinity Purification of GABA A Receptor from Detergent Extracts of Bovine Cortex-Fresh whole bovine brain was placed on ice, and the meninges were rapidly removed, and the cortex was dissected away from the remaining structures. The cortex was washed to remove blood and stored in 0.32 M sucrose at Ϫ80°C. In preparation for affinity purification of GABA A receptors, ϳ100 g of bovine cortex were thawed and homogenized at 4°C in 10 volumes of 50 mM Tris (pH 8.0) containing 0.32 M sucrose, using a 55-ml Potter-Elvehjem glass homogenizer and Teflon pestle. The homogenate was centrifuged (550 ϫ g, 10 min) to remove particulate matter, and then the pellet, after centrifugation (50,000 ϫ g, 1 h), was resuspended in 20 volumes of distilled water using a Tekmar Ultra-Turrax (Cincinnati, OH). The suspension was centrifuged (50,000 ϫ g, 1 h), and the pellet was resuspended in 10 volumes ice-cold membrane buffer (50 mM Tris, 50 mM KCl, 1 mM EDTA, 10 g/ml soybean trypsin inhibitor, 0.1 mM phenylmethylsulfonyl fluoride, 0.5 mM dithiothreitol, 2 mM benzamidine hydrochloride, 0.1 mM benzethonium chloride, 100 g/ml bacitracin (pH 8.0)). Following another centrifugation (50,000 ϫ g, 1 h), the membrane pellet was stored at Ϫ20°C overnight.
The frozen membrane pellet was thawed on ice and resuspended in 4 volumes of ice-cold solubilization buffer (membrane buffer with 1 M KCl, 0.5% Triton X-100, and 20% glycerol) using an Ultra-Turrax. This suspension was gently stirred for 1 h (4°C) and then centrifuged (150,000 ϫ g, 1 h). The supernatant was applied three times to an Ro7-1986/1 affinity column (4°C) pre-equilibrated with 100 ml of solubilization buffer. After loading, the column was washed with 100 ml of solubilization buffer, 400 ml of wash buffer (membrane buffer with 0.2% Triton X-100 and 10% glycerol), and 200 ml of urea wash buffer (wash buffer with 3 M urea). GABA A receptor was eluted in 5 ml of elution buffer (wash buffer with 3 mM flurazepam and 4 M urea, 1 h of incubation), which was repeated 2 times. Three additional elutions were performed with 30-min incubations (4.5 h total). Following the final incubation, an additional 20 ml of elution buffer was added, and the eluate was pooled (approximately 50 ml) and dialyzed (3 h, 4°C) against 20 mM Tris (pH 8.0), 0.2% Triton X-100, 10% glycerol, 0.1 mM phenylmethylsulfonyl fluoride, and 1 mM ␤-mercaptoethanol.
The dialyzed affinity column eluate was applied to a 5-ml High Q ion exchange resin (prepared according to the manufacturer's instructions) and washed extensively (20 mM Tris, 10% glycerol, 0.2% Triton X-100 (pH 7.4)) to remove flurazepam. The protein eluted from the column with 10 ml of 20 mM Tris, 1 M NaCl, 10% glycerol, 0.2% reduced Triton X-100 (pH 7.4) was collected in 1-ml fractions and assayed for [ 3 H]muscimol binding (see under "Binding Assays"). The one or two fractions containing [ 3 H]muscimol binding were pooled and used for photolabeling.
Photoaffinity Labeling of Affinity-purified GABA A Receptor-Affinity-purified GABA A receptors (ϳ100 -300 pmol in 5 ml of 50 mM Tris, 10% glycerol, 0.2% reduced Triton X-100 (pH 7.4)) were incubated at 4°C with 20 nM [ 3 H]Ro15-4513 for 90 min in a glass Petri dish. In some experiments, 10 M flunitrazepam was included to determine the specificity of [ 3 H]Ro15-4513 photoincorporation. The receptor preparation was irradiated at a distance of 7.5 cm with a 366-nm lamp (UVGL-25 (UVP, Inc., San Gabriel, CA)) for 30 min. After photolabeling, the amount of irreversible binding was estimated by taking 3 aliquots (100 l) of the photolabeled receptor preparation and incubating them with 1 mM flunitrazepam for 1 h at 4°C with gentle shaking. Total protein was precipitated, and the retained 3 H was determined by filtration as described under "Binding Assays." Binding Assays-GABA A receptor binding was assayed at each step of affinity purification by a [ 3 H]muscimol binding assay (32). Briefly, aliquots of membranes or detergent extracts (100 l; 0.2-0.006 mg of protein) taken at different steps during affinity purification were incubated in a final volume of 0.5 ml of 50 mM Tris/HCl (pH 7.4) and 50 mM KCl at 4°C for 30 min with the GABA A receptor agonist [ 3 H]muscimol at a concentration (40 nM) sufficient to occupy all agonist-binding sites. For solubilized GABA A receptor, bovine ␥-globulin (100 l; 30 mg/ml) and 30% polyethylene glycol (300 l) were added after incubation to precipitate total protein, followed by vigorous vortexing. Receptorbound [ 3 H]muscimol was trapped by filtration on glass fiber filters (Whatman GF/B). The filters were rinsed with three 2-ml aliquots of ice-cold buffer (50 mM Tris/HCl (pH 7.4), 50 mM KCl, and 7% polyethylene glycol), and retained 3 H was determined by liquid scintillation counting. All assays were carried out in triplicate, and nonspecific binding was determined by carrying out incubations in the presence of 0.1 mM GABA in parallel tubes. Protein was estimated using the BCA protein assay (Pierce).
SDS-PAGE and Immunoblotting-[ 3 H]Ro15-4513-labeled GABA A receptor was precipitated by methanol and chloroform (33), dissolved in SDS-PAGE sample buffer, and boiled for 5 min, and then the polypeptides were separated by SDS-PAGE (34). Polypeptides were separated on 0.75-or 1.5-mm slab gels consisting of a 10% separating gel and a 3% stacking gel. To visualize separated polypeptides, gels were stained with Coomassie Brilliant Blue. To visualize 3 H distribution by fluorography, the gels were soaked in Amplify (Amersham Biosciences) according to manufacturer's instructions and dried prior to exposure to film (X-Omat AR, Eastman Kodak) for 21-28 days at Ϫ20°C.
Tricine gels (16.5% T, 6% C) were used as described (35). To isolate the labeled peptides by SDS-PAGE, the slab gels were cut into 6-mm strips that were eluted for 3 days in 12 ml of 100 mM NH 4 HCO 3 , pH 8.4, with 0.1% SDS and 2.5 mM dithiothreitol. The eluates were assayed for 3 H, and the samples of interest were filtered, concentrated to ϳ200 l in Ultrafree Biomax 5K concentrators (Millipore), and precipitated with acetone (75%, Ϫ20°C overnight).
Enzymatic Digestions-For digestion with EndoLys-C, samples were resuspended with vortexing in 100 -200 l of 25 mM Tris buffer (pH 8.6) with 0.1% SDS and 500 M EDTA (EKC buffer). Digestions were performed with 0.3 units of EndoLys-C at 25°C for 2 weeks. Subsequent digestions were carried out in the same buffer with the following modifications. For EndoAsp-N (0.2 g, 1 week, 25°C), the sample was diluted 10-fold with H 2 O. For trypsin (10 g, 3 days, 25°C), Genapol C-100 (Calbiochem) was added to the sample at 5 times the SDS concentration. No buffer modifications were necessary for V8 protease (50 g, 3 days, 25°C). For sequential digestions, the activity of the previous enzyme was inhibited irreversibly by addition of 2 mM diisopropyl fluorophosphate (Sigma) at 25°C for Ͼ2 h.
HPLC-HPLC fractionation was performed on an Agilent 1100 binary HPLC system with an in-line degasser, column heater, and a diode array absorbance detector. Separations were carried out at 40°C on a 100 ϫ 2.1-mm Aquapore BU-300 7-m column (PerkinElmer Life Sciences). Solvent A was 0.08% trifluoroacetic acid in water, and solvent B was 60% acetonitrile, 40% isopropyl alcohol, 0.05% trifluoroacetic acid. The gradient (% solvent B) is indicated in the figures. To neutralize the acid, HPLC fractions were collected in 1.5-ml tubes containing 20 l of 250 mM Tris (pH 8.1). For enzymatic digestion after HPLC, fractions of interest were rotary-evaporated to near dryness and resuspended as above.
Sequence Analysis-N-terminal sequence analysis was performed on an Applied Biosystems Procise 492 protein sequencer on Biobrenetreated micro-trifluoroacetic acid filters (ABI 401111). Depending upon the application, two different injection loops were used on the amino acid analyzer. For high sensitivity detection of 3 H release, five-sixths of each cycle of Edman degradation were collected for scintillation counting, whereas one-sixth was utilized for residue identification. Although no GABA A receptor sequences were identified, the other peptide sequences detected were quantified to ensure proper sequencer function. For high sensitivity detection of phenylthiohydantoin-derivatives, twothirds of each cycle of Edman degradation were used for amino acid analysis, and only one-third was collected for 3 H determination. HPLC fractions of interest were drip-loaded directly onto the filters (45°C) using a syringe pump.
Data Base Searches-Edman degradation was used to determine the 3 H release profile for 3 H-labeled subunit fragments that were subjected to sequential proteolytic fragmentation with a panel of side chainspecific proteases. The observed 3 H release patterns identified the distribution of Lys, Arg, Asp, and Glu on the N-terminal side of the 3 H-labeled amino acid. The experimentally determined distributions of amino acids were used to search protein sequence data bases using the program Pattern Search at pir.georgetown.edu/pirwww/search/patmatch.html. Pattern Search searches for exact matches to proposed sequences in which multiple amino acids can be designated for each position in the sequence. Blast was used to identify the parent proteins of the amino acids sequences identified by Edman degradation (www.ncbi.nlm.nih.gov/BLAST/).
Molecular Modeling-A model of the extracellular region of the bovine GABA A receptor (2␣ 1  , 2␤ 1(10 -218) ␥ 2(25-233) ) was constructed from the snail acetylcholine-binding protein (AChBP) structure (36) using the Homology module of Insight II on a Silicon Graphic work station. The sequences for the GABA A receptor subunits were from the National Center for Biotechnology Information (NCB accession numbers P08219, P08220, and P22300), and the coordinates for the AChBP structure (Protein Data Bank code 1I9B) were from the Research Collaboratory for Structural Bioinformatics. Insertions (␣ 1 residues 82, 103, 122-123, 151, and 174 -176; ␤ 1 residues 78, 102, 119 -120, 144, and 171; and ␥ 2 residues 93, 115, 134 -135, 159, and 186) and deletions (in ␣ 1 two residues between 205 and 206; in ␤ 1 two residues between 200 and 201; and in ␥ 2 two residues between 215 and 216) required to facilitate the alignment between the AChBP and the bovine GABA A receptor subunits were placed at turns to preserve the overall ␤-sheet structure. The ␥ 2 subunit was placed next to an ␣ 1 subunit such that the residues identified previously at the benzodiazepine site were in proximity. Once the pentamer was constructed, the structure was minimized using the Discovery module.
The structures of flunitrazepam and Ro15-4513 were constructed and minimized using the Builder module. Although we were able to build and minimize Ro15-4513, the force fields required for the Docking module are not programmed to accept an azide moiety. This moiety was altered to CϭCϭO to allow for docking. Structures were placed within the predicted benzodiazepine-binding site located at the interface between the ␣ 1 and ␥ 2 subunits, and the Docking module was used to identify the best ligand orientation. The binding site was defined as ␣ 1 residues 102, 103, 156, 160, 161, 203, 205, 206, 207, 210, and 212 and ␥ 2 residues 54, 58, 77, 79, 81, 98, 130, 132, 138, 140, 142, 144, 184, 185, and 194. For the Ro15-4513 ligand, Ͼ200 orientations were sampled, and the minimized best fit (illustrated in Fig. 10) was obtained 117 times. Other orientations of the planar ligand were also sampled, although their interaction energies were not as favorable as the best fit structure. A similar docking was performed with the flunitrazepam model resulting in a prominent single preferred orientation. Table I, GABA A receptor was purified ϳ100-fold with a 90% recovery from a Triton X-100 extract of membranes from 100 g of bovine cortex on an affinity column of the benzodiazepine Ro7-1986/1 and then concentrated to a small volume by use of an anion exchange column. Ro7-1986/1 is an analog that binds to diazepam-sensitive GABA A receptors (37). Recombinant GABA A receptors having a high affinity for diazepam contain the ␥ 2 subunit in combination with ␤ subunits and either ␣ 1 , ␣ 2 , ␣ 3 , or ␣ 5 subunits (for review see Ref. 8). Recombinant GABA A receptors containing only ␣ 4 or ␣ 6 subunits do not bind diazepam and should not be retained on the Ro7-1986/1 column.

Purification of GABA A Receptors-As summarized in
The initial binding capacity (B max ) of the cortical membranes, as assayed by the binding of a saturating concentration of [ 3 H]muscimol, was ϳ0.6 pmol/mg protein (ϳ1 nmol total). Approximately 80% of the GABA A receptor was recovered in the Triton X-100 extract, and ϳ90% of the solubilized GABA A receptor was retained on the affinity column after washing with 3 M urea. After elution of the column with 3 mM flurazepam in 4 M urea, ϳ90% of the solubilized GABA A receptor (730 pmol) was recovered in 50 ml with a [ 3 H]muscimol binding capacity 200-fold higher than the initial cortical membranes. The eluate was dialyzed to remove urea and to reduce the flurazepam concentration, and anion exchange chromatography was then used to further remove fluazepam and to concentrate the receptor. This step did effectively concentrate the GABA A receptor into a volume of 1-2 ml, although recovery was poor (20 -50%), and there was no further purification. Subunit Composition of Affinity-purified GABA A Receptor-In previous investigations, GABA A receptors affinity-purified with an Ro7-1986/1 affinity column from detergent extracts of bovine or rat cortex were found to contain ␣ 1 , ␣ 2 , and ␣ 3 subunits (31,38). To identify the GABA A receptor ␣ subunits present in the affinity column eluate after High Q ion exchange chromatography, immunoblot analysis was performed using antibodies specific for the ␣ 1 , ␣ 2 , ␣ 3 , or ␣ 5 subunit. An antibody specific for the ␣ 1 subunit recognized a 51-kDa band, the anti-␣ 2 subunit antibody recognized a 50-kDa band, and the anti-␣ 3 subunit antibody recognized a 56-kDa band as well as a 29-kDa band, presumably an ␣ 3 subunit proteolytic fragment (Fig. 1). Anti-␣ 5 subunit antibodies did not recognize any polypeptides in the eluate from the Ro7-1986/1 column (data not shown).
Photolabeling Affinity-purified GABA A Receptor with [ 3 H]Ro15-4513-Affinity-purified GABA A receptor (40 nM muscimol sites) was equilibrated with 20 nM [ 3 H]Ro15-4513 and then photolabeled. After photolabeling, ϳ50% of the [ 3 H]Ro15-4513 appeared irreversibly incorporated, as judged by the amount remaining bound 1 h after the addition of 1 mM flunitrazepam as competitor. When GABA A receptors were photolabeled in the absence or presence of 10 M flunitrazepam and the 3 H-labeled polypeptides determined by SDS-PAGE, an ϳ50-kDa band was the primary site of flunitrazepam-inhibitable labeling, along with additional specific 3 H incorporation in several bands between 46 and 57 kDa (Fig. 2). The mobilities of the labeled bands were generally consistent with the protein bands recognized by antibodies against bovine GABA A receptor ␣ 3 , ␣ 1 , and ␣ 2 subunits (Fig. 1).
Radiosequence Analyses Using Sequential Proteolytic Fragmentation-Based upon the binding of [ 3 H]muscimol to the affinity-purified GABA A receptor preparation, the receptors were purified ϳ200-fold from the crude membrane preparation, and active GABA A receptor comprised ϳ2% of the protein in the preparation. In addition, the receptor preparation was heterogeneous, with some receptors containing ␣ 1 subunits and others with ␣ 2 or ␣ 3 subunits. We reasoned that for this heterogeneous, partially purified preparation, it might be difficult to purify by HPLC and/or SDS-PAGE a labeled subunit fragment to a sufficient extent to allow direct identification of the labeled peptide by the released phenylthiohydantoin-derivatives (lower detection limit of ϳ0.5 pmol). However, the high radiochemical specific activity of [ 3 H]Ro15-4513 (23 Ci/mmol) enabled us to develop a radiosequence strategy to identify the labeled amino acid in the presence of non-labeled contaminating fragments. Whereas 1 pmol of [ 3 H]Ro15-4513-labeled protein would contain ϳ18,000 cpm, sequence analysis of samples containing as little as 2,000 -3,000 cpm could lead to a clear identification of the cycle of Edman degradation containing the labeled amino acid, even though the level of labeled peptide would be too low for amino acid detection.
To identify the sites of photoincorporation by [ 3 H]Ro15-4513 in GABA A receptor subunits, we used a panel of proteases with defined side chain specificities to cleave the labeled subunit fragments and Edman sequencing with 3 H detection in each cycle (radiosequence analysis) to determine the number of amino acids between the site of cleavage and the labeled residue(s). For example, after digestion with EndoLys-C, which cleaves at the C-terminal side of lysines, the release of 3 H after n cycles of Edman degradation would indicate that 3 H was incorporated in a peptide containing a lysine n amino acids before the labeled residue. For this analysis, we utilized En-doLys-C, trypsin (cleavage C-terminal of lysines and argin-  ines), V8 protease (cleavage C-terminal of glutamates), and EndoAsp-N (cleavage N-terminal of aspartates). By use of sequential digestions with as many a four enzymes, we could identify the distribution of these amino acids (Lys, Arg, Asp, and Glu) on the N-terminal side of the site(s) of labeling, and we hoped that this empirically determined distribution would produce a unique identification of the site(s) of [ 3 H]Ro15-4513 incorporation.
Isolation of Material for Radiosequence Analysis-Affinitypurified GABA A receptor (200 pmol) was photolabeled with [ 3 H]Ro15-4513 and separated by preparative slab gel SDS-PAGE (10% acrylamide). The resulting gel was cut into 6-mm strips, and each strip was eluted and counted for 3 H as shown in Fig. 3A. A broad peak of 3 H was detected spanning three bands (6 -8) centered at 56 kDa, and ϳ80% of the 3 H in these bands was recovered after concentration, precipitation, and resuspension. This region corresponded to the gel region con-taining the ␣ 1 , ␣ 2 , and ␣ 3 subunits identified by immunoblot (Fig. 1). Because the ␣ 3 -immunoreactive material had lower mobility (higher M r ) than the ␣ 1 or ␣ 2 bands, we analyzed the sites of 3 H incorporation separately for Bands 6 and 8 with the hope that each would contain different GABA A receptor ␣ subunits.

Identification of [ 3 H]Ro15
-4513 Incorporation at ␣ 3 Tyr-234 -The sequential digestion strategy that we used to characterize the site(s) of [ 3 H]Ro15-4513 photoincorporation within the GABA A receptor subunit(s) isolated from Band 6 (centered at 66 kDa) is summarized in Fig. 4A. When an aliquot of Band 6 (2,000 cpm) was sequenced, no release of 3 H (Ͻ2 cpm above background) was detected in 25 cycles of Edman degradation (data not shown). Material from Band 6 was digested with EndoLys-C, and when the digest was fractionated by reversedphase HPLC (Fig. 3B), there was a peak of 3 H in fractions 22-23 (ϳ43% organic) which was pooled for further analysis. Radio-sequence analysis of an aliquot of this pool (5,500 cpm, Fig. 5A) revealed no prominent peak of 3 H release, although there was a gradual increase in the background 3 H release during the 30 cycles of Edman degradation that would be consistent with gradual wash off of the labeled peptide from the filter. The remainder of the pool was digested with EndoAsp-N. When an aliquot of that digest was sequenced (5,500 cpm, Fig.  5B), there were peaks of 3 H release in cycles 19 (76 cpm) and 27 (34 cpm). When the remaining material was digested with V8 protease and an aliquot sequenced (5,500 cpm, Fig. 5C), there was a peak of 3 H release in cycle 9 (324 cpm) with no evidence of the 3 H release seen previously in cycles 19 and 27. The remainder of the material was digested with trypsin. When this digest was sequenced (5,500 cpm, Fig. 5D), there was 3 H release in cycle 6 (476 cpm) with no 3 H release remaining in cycle 9.
This observed pattern of 3 H releases after sequential proteolysis indicated the following. 1) One or more residues labeled

FIG. 3. Purification by HPLC of [ 3 H]Ro15-4513-labeled GABA A receptor subunit fragments. An affinity-purified preparation of bovine GABA A receptor (ϳ250 pmol in 5 ml) was equilibrated with [ 3 H]Ro15
-4513 (20 nM) and irradiated as described under "Experimental Procedures." The labeled material was separated by preparative SDS-PAGE (10% acrylamide), and the gel was then cut into 6-mm strips. A, total 3 H distribution recovered from the eluted gel bands as determined by liquid scintillation counting of 1% aliquots. A broad peak of 3 H was centered at Band 7 (ϳ56 kDa), and an additional peak was seen at the dye front. When the eluates of the bands were concentrated, precipitated, and resuspended in preparation for proteolysis, Ͼ77% of the 3 H was recovered in Bands 6 -8, whereas Ͻ7% of the 3 H in the dye front was recovered. The eluates from Bands 6 -8 were digested individually with EndoLys-C, and the resulting fragments were separated by reversed-phase HPLC as described under "Experimental Procedures." The 3 H distributions (q) are shown as well as the absorption at 214 nm (O) and the % solvent B (---) for Band 6 (B), Band 7 (C), and Band 8 (D). For Band 6 (ϳ66 kDa), 80,000 cpm were injected; total recovery was 80%, and ϳ50% of 3 H recovered a single peak (fractions 22 and 23). For Band 7 (ϳ56 kDa), 72,000 cpm were injected, total recovery was 70%, with ϳ40% in fractions 22 and 23. For Band 8 (ϳ48 kDa), 90% of the 114,000 cpm injected was recovered, with ϳ40% in fractions 22 and 23.

FIG. 4. Digestion strategy for analysis of [ 3 H]Ro15-4513-labeled GABA A receptor subunit fragments.
by [ 3 H]Ro15-4513 occurred 18 and/or 26 amino acids after an Asp residue. 2) This same labeled residue(s) occurs 9 amino acids after a Glu (they are the same site(s) because the 3 H release seen after EndoAsp-N treatment was no longer present after digestion with V8 protease). 3) The same labeled residue(s) occurs 6 amino acids after an Arg (not a Lys, because EndoLys-C did not initially cut there). A search of the bovine GABA A receptor subunit sequences determined that this pattern of amino acids (DX 7 DX 8 EX 2 R) occurred only in the ␣ 3 subunit, with sites of cleavage at Asp-208, Asp-216, Glu-225, and Arg-228 identifying ␣ 3 Tyr-234 as the amino acid labeled by [ 3 H]Ro15-4513 (Fig. 5E). The 3 H release seen in cycles 19 and 27 after EndoAsp-N digestion would be accounted for by [ 3 H]Ro15-4513 incorporation at ␣ 3 Tyr-234 and partial cleavage at either Asp-216 or Asp-208. The only inconsistency with this conclusion is the fact that V8 protease produced no cleavage at ␣ 3 Glu-233, the amino acid preceding ␣ 3 Tyr-234. However, a plausible explanation for the lack of cleavage at ␣ 3 Glu-233 is that the modification at ␣ 3 Tyr-234 prevents V8 protease action at the peptide bond between ␣ 3 Glu-233 and the modified tyrosine.
To determine the uniqueness of this pattern of amino acid residues, we used the program Pattern Search to search the PIR data base for the pattern 3(X K )-D-7(X KD )-D-8(X KD )-E-2(X KDE )-R-4(X KDER )-2X where X K ϭ not K; X KD ϭ not K or D; X KDE ϭ not K, D, or E; and X KDER ϭ not K, D, E, or R. From this search we identified 43 sequences that matched this distribution, only six of which were mammalian. Four of the mammalian sequences were the ␣ 3 subunit of the GABA A receptor (from different species); the other 2 sequences were variants of human intestinal mucin 2. The only bovine sequence identified was the GABA A receptor ␣ 3 subunit.

Identification of [ 3 H]Ro15-4513 Incorporation at ␣ 1 Tyr-210 -We next characterized the site(s) of [ 3 H]Ro15
-4513 photoincorporation within the GABA A receptor subunit(s) isolated from Band 8 (centered at 48 kDa, Fig. 3A). A summary of the digestion strategy is presented in Fig. 4B. Sequence analysis of an aliquot (2,000 cpm) of Band 8 revealed no 3 H release (Ͻ2 cpm above background) in 25 cycles of Edman degradation (data not shown). When the Band 8 material was digested with EndoLys-C and fractionated by reversed-phase HPLC, the 3 H elution profile (Fig. 3D) was similar to that seen for Band 6, with a peak of 3 H in fractions 22 and 23 (ϳ45% organic). When these fractions were pooled and an aliquot was sequenced (8,500 cpm, Fig. 6A), there was no release of 3 H in 30 cycles of Edman degradation other than a gradual increase in background radioactivity. When the remainder of the pool was digested with trypsin and an aliquot was sequenced (8,500 cpm, Fig. 6B), there was a peak of 3 H release in cycle 23 (148 cpm). When the rest of the material was digested with En-doAsp-N and an aliquot was sequenced (8,500 cpm, Fig. 6C), there was a prominent peak of 3 H in cycle 12 (397 cpm) and no evidence of the 3 H release previously seen in cycle 23. This pattern of 3 H releases after sequential proteolysis indicated the following. 1) A residue labeled by [ 3 H]Ro15-4513 occurred 23 amino acids after an Arg residue (not a Lys, because EndoLys-C did not cleave at that position). 2) This same labeled residue occurs 11 amino acids after an Asp. 3) There were no lysines within 30 amino acids of the labeled amino acid. A search of the amino acid sequences for the bovine GABA A receptor subunits found that this pattern of specific residues occurs only in the ␣ 1 subunit. Cleavage at Arg-187 and Asp-199 would identify ␣ 1 Tyr-210 as the amino acid labeled by [ 3 H]Ro15-4513 (Fig. 6D). Significantly, ␣ 1 Tyr-210 is homologous to ␣ 3 Tyr-234.
We also analyzed the occurrence of this pattern with the PIR Pattern Search program (6(X K )-R-11(X KR )-D-9(X KRD )-2X where X K ϭ not K; X KR ϭ not K or R; and X KRD ϭ not K, R, or D). This pattern was far from unique; 10,823 protein sequences were identified of which 41 were of published bovine proteins. However, the ␣ 1 subunit was the only GABA A receptor subunit sequence to match this pattern.

Evidence of [ 3 H]Ro15
-4513 Incorporation at ␣ 2 Tyr-209 -Material from Band 7 (ϳ56 kDa) was also digested with En-doLys-C and fractionated by reversed-phase HPLC (Fig. 3C). When aliquots from fractions 20 -22 were sequenced individually, similar 3 H release patterns were seen for fractions 20 and 21 that differed from the release seen for fraction 22 (Fig. 7).  (Fig. 3B) were pooled, rotary-evaporated, and resuspended in 100 l of EKC buffer (22,000 cpm recovered). A, an aliquot from the EndoLys-C digest was sequenced. B, the remaining material was digested with EndoAsp-N from which an aliquot was sequenced. C, the remaining material was then digested with V8 protease, and an aliquot of that digest was sequenced. D, the remainder was digested with trypsin and then sequenced. The sequential digests were performed as described under "Experimental Procedures," and for each sample equal amounts of 3 H (5500 cpm) were loaded on the filter and sequenced for 30 cycles of Edman degradation, after which 1160 (A), 780 (B), 410 (C), and 300 (D) cpm remained on the filter. After the initial EndoLys-C digest (A), no release of 3 H above background was detected other than a 50 cpm anomaly in cycle 23 which was not reproduced in two other sequencing studies of similarly prepared samples. Upon digestion with EndoAsp-N, release of 3 H was seen in cycles 19 and 27 (B). 3 H release in those cycles was eliminated by digestion with V8 protease, which resulted in 3 H release in cycle 9 (C). After digestion with trypsin, 3  For fractions 20 and 21, when samples containing 800 and 1550 cpm were sequenced, there were peaks of 3 H release in cycle 6 of 95 and 176 cpm, respectively. In contrast, radio-sequence analysis of a sample of fraction 22 containing 1730 cpm had only 17 cpm released in cycle 6. Sequence analysis of fractions 20 -21 from the HPLC of Bands 6 and 8 also revealed 100 -200 cpm 3 H release in cycle 6 (data not shown), which had not been seen in the sequencing of the peak 3 H fractions (Figs. 5A and  6A). An examination of the amino acid sequence of the bovine GABA A receptor ␣ 2 subunit in the region homologous to the identified sites of [ 3 H]Ro15-4513 photolabeling in the ␣ 1 and ␣ 3 subunits indicated that 3 H release in cycle 6 was consistent with labeling of the ␣ 2 subunit at Tyr-209, which is homologous to ␣ 1 Tyr-210 and ␣ 3 Tyr-234. EndoLys-C digestion of the ␣ 2 subunit should produce a 16-amino acid fragment containing Tyr-209 in cycle 6 (Fig. 7B). This peptide is shorter than the labeled peptides produced by EndoLys-C digestion of the ␣ 1 or ␣ 3 subunits (64 and 43 amino acids, respectively), and the shorter peptide would be expected to elute at a lower % organic solvent. Because the first digestion was with EndoLys-C, there were no other proteases that would cleave between ␣ 2 Lys-203 and ␣ 2 Tyr-209. Comparison of the ␣ 2 and ␣ 3 subunit sequences in this region reveals that the only difference in protease sites is an Arg-Lys substitution 6 amino acids before the labeled tyrosines. Thus, without the initial digestion with EndoLys-C, the 3 H release pattern seen for labeled ␣ 2 subunit material after digestion with EndoAsp-N, V8 protease, or trypsin would be the same as we observed for the labeled ␣ 3 subunit (Fig. 5).
Bovine Mitochondrial F 1 -ATPase Peptides Co-purify with 3 H-Labeled GABA A Receptor Fragments-Based upon the identified sites of [ 3 H]Ro15-4513 incorporation in GABA A receptor ␣ subunits, EndoLys-C digestion of labeled ␣ 1 , ␣ 2 , and ␣ 3 subunits would generate labeled fragments of 64, 16, and 43 amino acids. In preliminary experiments on an analytical scale, we established that digestion with EndoLys-C produced an ϳ7-kDa fragment that by radiosequence analysis after redigestion with EndoAsp-N had a 3 H release profile consistent with the presence of the ␣ 1 subunit fragment labeled at ␣ 1 Tyr-210. We attempted to purify the 3 H-labeled ␣ 1 subunit fragment from an EndoLys-C digest by a combination of Tricine SDS-PAGE and reversed-phase HPLC. When the labeled subunits were separated by SDS-PAGE, the peak of 3 H migrated at ϳ56 kDa (Fig. 8A, band 10), and this incorporation was reduced by Ͼ85% when photolabeling was carried out in the presence of 10 M flunitrazepam. When material eluted from Band 10 was digested with EndoLys-C and the digest fractionated by Tricine SDS-PAGE, there were specifically labeled bands of 16, 8, and 4 kDa (Fig. 8B) (and similar 3 H distributions for the EndoLys-C digests of Bands 9 and 11). When the 8-kDa band was excised, eluted, and purified by reversed-phase HPLC (Fig. 8C), the majority of the 3 H was recovered in a single peak at 47% organic (fraction 23). When this fraction was sequenced, no bovine GABA A receptor subunit sequences were detected, but there was a clear primary sequence identified in 30 cycles of Edman degradation that was a fragment of the bovine mitochondrial F 1 -ATPase A chain (1E1QA) beginning at Thr-3 (initial yield, 3 pmol). There were also three or four other peptides present at initial levels of 1-2 pmol, although no clear sequence assignment was possible. Based upon the radiochemical spe-  (Fig. 3D) were pooled, rotary-evaporated, and resuspended in 200 l of EKC buffer (38,000 cpm recovered). A, an aliquot from the EndoLys-C digest was sequenced. B, the remaining material was digested with trypsin from which an aliquot was sequenced. C, the remaining material was then digested with EndoAsp-N, and an aliquot of that digest was sequenced. The sequential digests were performed as described under "Experimental Procedures," and for each sample equal amounts of 3 H (8600 cpm) were loaded on the filter and sequenced for 30 cycles of Edman degradation, after which 1550 (A), 620 (B), and 450 (C) cpm remained on the filter. No release of 3 H above background was seen when the initial EndoLys-C digest was sequenced (A). After digestion with trypsin, prominent release of 3 H was seen in cycle 23 (B). 3  cific activity of [ 3 H]Ro15-4513, the 6,500 cpm sequenced was incorporated in ϳ0.4 pmol of labeled GABA A receptor subunit fragment (and ϳ1 pmol of unlabeled subunit fragment if the labeled and unlabeled peptides remained together throughout the purification.) This level of GABA A receptor subunit fragment would not have not been detectable in the presence of a complex mixture of peptides at the ϳ2-3 pmol level. Fractions 24 and 25 were also sequenced and were found to contain fragments from the F 1 -ATPase chain D (1E79D) beginning at Leu-406 and Ala-9, respectively, at initial yields of ϳ0.5 pmol and identified in 20 cycles of Edman degradation. Once again, no GABA A receptor subunit sequences were detected. DISCUSSION The goal of this study was to use photoaffinity labeling to identify the amino acid(s) in the benzodiazepine-binding site that is photolabeled by [ 3 H]Ro15-4513, an imidazobenzodiazepine that contains an azide substituent that upon UV illumination will produce a reactive nitrene. Because of the defined structure of the photoreactive intermediate, identification of the amino acids labeled by [ 3 H]Ro15-4513 will define the orientation of the ligand within the benzodiazepine-binding site. Previous studies have shown that for receptors containing the ␣ 1 , ␣ 2 , ␣ 3 , or ␣ 5 subunits, the binding of Ro15-4513 was negatively coupled to the binding of GABA, in contrast to the positive allosteric coupling seen for flunitrazepam at those receptor subtypes or for Ro15-4513 binding to receptors containing ␣ 4 or ␣ 6 subunits (11).
[ 3 H]Flunitrazepam has been shown to photoincorporate into ␣ 1 His-102 (26,27), which had been identified by mutational analyses as a major affinity determinant of benzodiazepine binding (39). ␣ 1 His-102 is also an important determinant of Ro15-4513 efficacy, as substitutions at that position regulate equilibrium binding affinity (40) and determine whether it acts as a positive or negative modulator of GABA responses (41 (43), possibly within amino acids 247-289 that span the end of the first transmembrane segment to the beginning of the third transmembrane segment (29).
In this report we have used a radiochemical sequencing strategy to identify the amino acids labeled by [ 3 H]Ro15-4513 in a heterogeneous preparation of GABA A receptors purified from bovine cortex on a benzodiazepine affinity column. The purification procedure resulted in an ϳ200-fold purification of a population of receptors containing, based upon immunoblot analysis, ␣ 1 , ␣ 2 , or ␣ 3 subunits. Based upon the level of specific [ 3 H]muscimol binding, active GABA receptors comprise ϳ1% of the protein in the preparation, and, not surprisingly, GABA A receptor subunits or subunit fragments could not be directly detected by Edman degradation when labeled subunits were FIG. 9. Sequence alignments of GABA A receptor subunit segments containing benzodiazepine-binding site affinity determinants. Shown are alignments of segments of the bovine ␣ 1 , ␣ 2 , and ␣ 3 subunits along with segments of the ␥ 2 subunit with residues (*) identified by mutational analyses as benzodiazepine affinity determinants (7,16). Predicted cleavage sites for the proteases used in this study are denoted in boldface type. The homologous tyrosines photolabeled by After photolabeling, polypeptides were fractionated by preparative SDS-PAGE (10% acrylamide), and the gels were cut into 6-mm slices and eluted. The total 3 H recovered from each of the gel slices is shown along with the mobilities of the molecular weight standards. B, material recovered from Band 10 (ϳ56 kDa (128,000 cpm Ϫflunitrazepam, 16,000 cpm ϩflunitrazepam)) was digested with EndoLys-C, and the digest was fractionated by Tricine SDS-PAGE with the distribution of 3 H determined after elution from the gel slices. C, the material recovered from Band J (14,700 cpm Ϫflunitrazepam, 400 cpm ϩflunitrazepam) was purified by reversed-phase HPLC. The 3 H distribution (q, Ϫflunitrazepam, 10% of each fraction counted) is shown as well as the absorption at 214 nm (O) and the % solvent B (---). No bovine GABA A receptor subunit peptide sequences were detected when the peak of 3 H or adjacent fractions were sequenced, although other bovine sequences were identified (see text). isolated by SDS-PAGE or when subunit fragments were further purified by SDS-PAGE and reversed-phase HPLC. However, by use of N-terminal sequence analysis to determine the patterns of 3 H release as labeled subunit fragments were digested sequentially with a panel of four proteases, we established that [ 3 H]Ro15-4513 is photoincorporated into homologous positions in the three subunits: ␣ 1 Tyr-210, ␣ 2 Tyr-209, and ␣ 3 Tyr-234. Our results are most compelling for the ␣ 3 subunit, for which positive identification of the 3 H release profile was made after digestion with 4 proteases, whereas the identification of the labeled amino acids in the ␣ 1 and ␣ 2 subunits utilized only 3 and 1 proteases, respectively. Although the subunit heterogeneity of the brain GABA A receptor preparation initially complicated the identification of the sites by making it appear that multiple sites of labeling might exist within a single subunit, the fact that our data are consistent with photoincorporation into homologous Tyr in all three subunits actually strengthens the conclusion about each individual identification.
The position in the rat GABA A receptor (␣ 1 Tyr-209) occupied by the tyrosine residues photolabeled by [ 3 H]Ro15-4513, as well as ␣ 1 Thr-206, have been shown by site-directed mutagenesis to affect the affinity of ligands whose binding is coupled either positively or negatively to the GABA site (14,44). In addition, ␣ 1 Val-211 and the corresponding ␣ 5 Ile-215 have been identified as imidazobenzodiazepine affinity determinants (45). In a pharmacophore model of the benzodiazepine-binding site developed based upon the results of mutational analyses (46), ␣ 1 Tyr-209 was proposed to interact with all ligands. However, in the model it was not positioned in proximity to the azide of Ro15-4513 (or to ␣ 1 His-101).
The radiosequence strategy that we have used permits a positive identification of [ 3 H]Ro15-4513-labeled amino acids in the ␣ 1 , ␣ 2 , and ␣ 3 subunits, but in the absence of direct identification of subunit masses by Edman degradation, our results provide no information about the relative efficiency of incorporation of [ 3 H]Ro15-4513 into the different subunits or about the relative abundance of each subunit in our receptor preparation. It is most likely that the labeled ␣ 1 , ␣ 2 , and ␣ 3 subunits are each contained within distinct populations of monomeric GABA A receptors. However, further immunoaffinity purification by use of subunit-specific antibodies would allow a direct determination of whether labeled ␣ 1 and ␣ 3 subunits, for example, coassemble in a single receptor. For rat cortical membranes, both ␣ 1 and ␣ 3 subunits were photolabeled with [ 3 H]Ro15-4513, and based upon immunoaffinity purification with subunit specific antisera, a minor proportion of labeled ␣ 1 subunits copurifies with ␣ 3 subunits (47).

No Evidence of [ 3 H]Ro15-4513
Labeling of ␣ 1 His-102 or Amino Acids within ␣ 1 97-117-Previous photolabeling and mutational studies of the GABA A receptor indicate that at least 2 additional regions of an ␣ subunit primary structure may come in contact with Ro15-4513 at the benzodiazepine-binding site, specifically residues near ␣ 1 His-102 and ␣ 1 Tyr-160 (7,13). An examination of an alignment of the bovine ␣ 1-3 sequences for these regions (Fig. 9) indicated that if they were labeled, we would have seen patterns of 3 H release after proteolytic fragmentation significantly different from those that we observed in our experiments. For the regions preceding ␣ 1 His-102 and ␣ 1 Tyr-160, all three subunits would produce the same patterns of protease cleavages, EX 19 KX 4 D and EX 4 DX 6 K, respectively. Although we have no direct proof that cleavages at these sites occurred, the efficiency of cleavages that occurred before ␣ 3 Tyr-234 and the corresponding regions of the ␣ 1 and ␣ 2 subunits makes it likely that these cleavages also occur. The patterns of 3 H release we observed lead us to conclude that there is no significant [ 3 H]Ro15-4513 photoincorporation in ␣ 1 His-102, in the corresponding positions in the ␣ 2 or ␣ 3 subunits, or in nearby amino acids. The 3 H release patterns seen during sequence analysis of the major peak of 3 H in the HPLC fractionation of EndoLys-C digests of GABA A receptor subunits (Figs. 5 and 6) are also not consistent with labeling at or near ␣ 1 Tyr-160, but the 3 H release in the sixth cycle of Edman degradation of fractions 20 and 21 of the EndoLys-C digest (Fig. 7), which we attribute to labeling of ␣ 2 Tyr-209, is also consistent with labeling of ␣ 1 Tyr-162 or the corresponding position in the other subunits. However, in other experiments when the materials eluted from gel Bands 6 -8 were digested with EndoAsp-N and then sequenced, 3 H release was limited to cycles 12 and 19 with no release in cycle 14 as would be expected for labeling of ␣ 1 Tyr-162 (data not shown).
A Model of the Benzodiazepine-binding Site-To gain further appreciation of the selective labeling of ␣ 1 Tyr-210 by [ 3 H]Ro15-4513, we constructed an homology model of an extracellular domain of GABA A receptor containing 2 ␣ 1 , 2 ␤ 1 , and a ␥ 2 subunits, based upon the crystal structure of the snail AChBP, a soluble, secreted protein homologous to the extracellular domain of the nicotinic ACh receptor (36,48). Shown in Fig. 10 is a stereo representation of the benzodiazepine-binding site within our model with Ro15-4513 docked in an energetically favored orientation. In this orientation the photoreactive nitrogen of the azide is ϳ5 Å from ␣ 1 Tyr-210 and 4 Å from ␣ 1 His- FIG. 10. Stereo representation of Ro15-4513 bound within a GABA A receptor benzodiazepine-binding site. An homology model of the extracellular region of the bovine GABA A receptor (␣ 1 ␤ 1 ␥ 2 ) was constructed from the crystal structure of the molluscan AChBP (36), and Ro15-4513 was docked in the benzodiazepine-binding site located at the ␣ 1 -␥ 2 subunit interface. The ␤-sheets from the AChBP structure are denoted (␤1, ␤2, . . . ). Ro15-4513 is in stick representation with atom types denoted: carbon, black; nitrogen, blue; and oxygen, red (compare with Fig. 2B). Hydrogens were excluded for simplification. GABA A receptor amino acids previously identified by mutational analyses as potential contributors to the benzodiazepine-binding pocket are denoted in ball and stick representation. These residues are contained within 3 primary structure segments from both the ␣ 1 and ␥ 2 subunits that are each indicated by different colors. The ␣ 1 residues localized to the benzodiazepine site are ␣ 1 His-102 (yellow); Tyr-160, Tyr-162, and Thr-163 (red); and Gly-201, Ser-205, Thr-207, Tyr-210, and Val-212 (blue); the ␥ 2 residues are Met-57 and Tyr-58 (brown); Phe-77, Gly-79, and Thr-81 (green); and Met-130 and Thr-142 (magenta). In this orientation the photoreactive nitrogen of Ro15-4513 (*) was ϳ5 Å from ␣ 1 Tyr-210 (cyan) and 4 Å from ␣ 1 His-102.
102, the residue that is labeled by [ 3 H] flunitrazepam, and it is also within 4 -6 Å of ␣ 1 Tyr-160, ␣ 1 Ser-205, and ␥ 2 Phe-77. In a docking simulation with flunitrazepam (not shown), we found that it was oriented with its nitro group within 4 Å of ␣ 1 His-102, consistent with one proposed mechanism of photoincorporation via a nucleophilic attack on a benzene carbon ortho to the nitrogen (49), and within 5 Å of ␣ 1 Tyr-210.
In terms of the model, the labeling of ␣ 1 Tyr-210 by [ 3 H]Ro15-4513 and the lack of labeling of ␣ 1 His-102 is very striking. This might result from the preferential reaction of the photoreactive intermediate with Tyr rather than His, as photolabeling of tyrosines by azide photoaffinity labeling has been frequently observed (19), whereas labeling of His has not been reported to our knowledge. However, the labeling of ␣ 1 Tyr-210 also occurs in the absence of labeling of ␣ 1 Tyr-160. Alternatively, the lack of labeling of ␣ 1 His-102 or ␣ 1 Tyr-160 may, in fact, occur because the structure of the benzodiazepine-binding site occupied by Ro15-4513 may differ significantly from that in the homology model. Because the binding of flunitrazepam is coupled positively to the binding of GABA, whereas the binding of Ro15-4513 is coupled negatively, flunitrazepam will stabilize the same conformation of the GABA A receptor that is favored by the binding of GABA, whereas Ro15-4513 will stabilize a conformation with low affinity for GABA, presumably the resting state of the receptor. Changes of the accessibility of substituted cysteines for chemical modification have provided evidence that the binding of GABA results in a change in the structure of the benzodiazepine-binding site (50).
The homology model is based upon the structure of the AChBP in a single conformational state (36), one that binds ACh with high affinity and is presumed to be closest in structure to a nicotinic ACh receptor conformation that binds ACh with high affinity, i.e. either an open channel or desensitized state. A recent comparison of the structure of the AChBP with that of the extracellular domain of the Torpedo nicotinic ACh receptor in the resting state identifies significant differences in structure within the agonist-binding site (51). In future studies it will be important to determine whether [ 3 H]Ro15-4513 photolabels amino acids other than ␣ 1 Tyr-210 when photolabeling is carried out in the presence of GABA or another agonist.