Molecular and Pharmacological Characterization of Native Cortical γ-Aminobutyric AcidA Receptors Containing Both α1 and α3 Subunits

We have investigated the existence, molecular composition, and benzodiazepine binding properties of native cortical α1-α3 γ-aminobutyric acidA (GABAA) receptors using subunit-specific antibodies. The co-existence of α1 and α3 subunits in native GABAA receptors was demonstrated by immunoblot analysis of the anti-α1- or anti-α3-immunopurified receptors and by immunoprecipitation experiments of the [3H]zolpidem binding activity. Furthermore, immunodepletion experiments indicated that the α1-α3 GABAA receptors represented 54.7 ± 5.0 and 23.6 ± 3.3% of the α3 and α1 populations, respectively. Therefore, α1 and α3 subunits are associated in the same native GABAA receptor complex, but, on the other hand, these α1-α3 GABAA receptors from the cortex constitute a large proportion of the total α3 population and a relatively minor component of the α1 population. The pharmacological analysis of the α1- or α3-immunopurified receptors demonstrated the presence of two different benzodiazepine binding sites in each receptor population with high (type I binding sites) and low (type II binding sites) affinities for zolpidem and Cl 218,872. These results indicate the existence of native GABAA receptors possessing both α1 and α3 subunits, with α1 and α3 subunits expressing their characteristic benzodiazepine pharmacology. The molecular characterization of the anti-α1-anti-α3 double-immunopurified receptors demonstrated the presence of stoichiometric amounts of α1 and α3 subunits, associated with β2/3, and γ2 subunits. The pharmacological analysis of α1-α3 GABAA receptors demonstrated that, despite the fact that each α subunit retained its benzodiazepine binding properties, the relative proportion between type I and II binding sites or between 51- and 59-61-kDa [3H]Ro15-4513-photolabeled peptides was 70:30. Therefore, the α1 subunit is pharmacologically predominant over the α3 subunit. These results indicate the existence of active and nonactive α subunits in the native α1-α3 GABAA receptors from rat cortex.

The neuropharmacological effects of benzodiazepines are mediated by the benzodiazepine () binding sites associated with the GABA A 1 receptor complex (for reviews, see Refs. 1 and 2). Based on their affinity for different drugs, two different benzodiazepine binding sites have been identified in the central nervous system. Type I (benzodiazepine receptor 1, 1 ) displays high affinity for CL 218,872 (2), ␤-carboline derivates (3), and the imidazopyridine zolpidem (4,5). Type II (benzodiazepine receptor 2, 2 ) displays low affinity for these compounds. A third benzodiazepine binding site with very low affinity for zolpidem (type II L , 5 ) has also been identified in isolated rat brain membranes (6) and sections (7).
Molecular cloning experiments have demonstrated the existence of five different families of subunits that are components of the GABA A receptor complex. Most of these families comprise several isoforms: ␣ 1 -␣ 6 , ␤ 1 -␤ 3 , ␥ 1 -␥ 3 , ␦, and 1 and 2 (for reviews, see Refs. 8 and 9). A minimum of ␣, ␤, and ␥ subunits should be co-expressed in transfected cells to resemble all the pharmacological properties of native GABA A receptors (10). On the other hand, the presence of different ␣ subunits determines the affinity of the different benzodiazepine binding sites. In this sense, the ␣ 1 -␤ 1 -␤ 3 -␥ 2 combination confers type I pharmacology to the recombinant GABA A receptor (i.e. high affinity for, among others, zolpidem and Cl 218,872) (11). Type II properties are conferred by the presence of ␣ 2 , ␣ 3 , or ␣ 5 subunits (11,12).
Several approaches have been taken to identify which subunits co-exist in the native GABA A receptor complex. However, the subunit composition of the different native GABA A receptor complexes remains unsolved. Immunoprecipitations or immunoaffinity purifications using anti-␣ subunit antibodies (anti-␣ 1 , -␣ 2 , -␣ 3 , -␣ 5 , and -␣ 6 subunits) indicated that a significant proportion of native receptors are made by the association of two different ␣ subunits (such as ␣ 1 ␣ 2 , ␣ 1 ␣ 3 , ␣ 1 ␣ 5 , or ␣ 1 ␣ 6 ) (13)(14)(15)(16)(17) in a single receptor complex. However, other authors have indicated the absence of association between different ␣ subunits (18,19). On the other hand, the pharmacological properties of these GABA A receptors are also unknown.
In the present article we have addressed these questions by determining the molecular and pharmacological properties of the immunopurified receptors using subunit-specific antibodies to the major ␣ subunits expressed in the rat cerebral cortex, the ␣ 1 and ␣ 3 subunits. mmol) were from DuPont NEN. Zolpidem was synthesized in the preclinical research department of Synthélabo Recherche. Cl 218,872 was from Cyanamid. All other benzodiazepines were from Hoffmann-La Roche.
Antibody Preparation-Peptides NH 3 -␣ 3 (amino acids 1-10, pyroglutamyl-GESRRQEPG) and COO Ϫ -␣ 1 (amino acids 419 -428, PQLKAPT-PHQ) were synthesized and coupled to keyhole limpet hemocyanin, via an extra tyrosine located at the COOH or NH 2 terminus of ␣ 3 or ␣ 1 peptides, by Neosystem SA (Strasbourg, France). For immunizations, rabbits (New Zealand White) were subcutaneously injected with 200 g of coupled peptide emulsified (1:2) in Freund's complete adjuvant followed 20 days later by a booster injection of conjugate with incomplete adjuvant (1:1). Rabbits were then boosted every 2-3 weeks. The animals were bled 10 days after each booster injection. Development of an immune response was followed by immunoprecipitation of the solubilized receptor.
The antibodies were purified through peptide affinity columns. The ␣ 3 and ␣ 1 peptides were coupled to adipic acid dihydrazide-agarose (sigma) or CNBr-activated Sepharose 4B (Pharmacia Biotech), respectively, as recommended by the manufacturer. Two ml of anti-␣ 3 or anti-␣ 1 antisera (diluted 1/5 in PBS) were recirculated, overnight at 4°C, in the corresponding 1-ml peptide column. After washing with 150 ml of PBS, the antibodies were eluted with 50 mM glycine-HCl, pH 2.3, and the fractions (0.5 ml) were neutralized by 1 M Tris, pooled and dialyzed in 1 liter of PBS (overnight at 4°C).
Other antibodies used in this work were the mAb 63-3G1 and anti-␥ 2 and -␥ 3 antibodies. These two polyclonal antibodies were produced using peptides from 2-10 or 1-15 amino acids of the NH 3 ϩ terminus of the ␥ 2 and ␥ 3 subunits, respectively (to be published elsewhere).
For immunoblots, the purified antibodies were labeled with digoxigenin as recommended by the manufacturer (Boehringer Mannheim). The digoxigenin incorporated into anti-␣ 1 or anti-␣ 3 antibodies was determined by enzyme-linked immunosorbent assay or dot blot. Both antibodies displayed a similar activity (not shown).
The GABA A receptor was solubilized at 4 mg of protein/ml at 4°C for 60 min with 0.5% (w/v) sodium deoxycholate, 0.5% (w/v) CHAPS, 140 mM NaCl, and 10 mM Tris-HCl, pH 7.5 (solubilization buffer), containing the same protease inhibitors as above. After centrifugation at 100,000 ϫ g for 60 min at 4°C, the supernatant was collected. The recovery of the benzodiazepine binding activity in the solubilized material represented 80 -90% of the 5 nM [ 3 H]FMZ, 10 nM [ 3 H]FNZ, or 5 nM [ 3 H]zolpidem binding activity found in membranes (also see Ref. 21).
Immunoprecipitation and Immunopurification-For immunoprecipitation experiments, the different antisera were adsorbed to a suspension of protein A-Sepharose (10%, w/v, in solubilization buffer; also see Refs. 20 -22). Ten or 75 l of anti-␣ 1 or anti-␣ 3 antibodies/assay were incubated, for 2 h at 4°C with agitation, with 50 l of 10% protein A-Sepharose in a final volume of 300 l of solubilization buffer. The IgG-protein A-Sepharose complexes were isolated by centrifugation, washed three times with 1.4 ml of solubilization buffer, and used for immunoprecipitation. On the other hand, preimmune sera was also absorbed to protein A-Sepharose and used in control experiments.
The immunoaffinity columns were synthesized as described (15). Briefly, 1-2 mg of each purified antibody were absorbed to 0.5 ml of protein A-Sepharose. The IgG-protein A-Sepharose complex was washed with 40 ml of PBS followed by 5 ml of 0.2 M triethylamine, pH 8.3. The column was then treated with 1.5 ml of 20 mM dimethylpimelimidate in 0.2 M triethylamine, pH 8.3, for 30 min at room temperature. After incubation, the medium was replaced by 1 ml of 0.2 M ethanolamine, pH 8.3, and incubated for 5 min. After coupling, the column was packed and washed, at 10 ml/h, with: 1) 40 ml of PBS; 2) 2 ml of 140 mM NaCl, 1 mM EDTA, 1 mM EGTA, and 50 mM sodium phosphate, pH 11.5; and 3) 20 ml of PBS. The columns were pre-equilibrated with 10 ml of solubilization buffer.
For immunopurification, the solubilized GABA A receptor (30 pmol of [ 3 H]FMZ binding activity) was recirculated (10 ml/h), overnight at 4°C through 0.5-ml columns. After absorption, the columns were washed (10 ml/h) with 20 ml of solubilization buffer. The retained material was eluted at 10 ml/h with 3 ml of 140 mM NaCl, 1 mM EDTA, 1 mM EGTA, 0.5% sodium deoxycholate, 0.5% CHAPS, and 50 mM sodium phosphate, pH 11.5. Fractions of 0.5 ml were collected and neutralized with 18 l of 1 M sodium phosphate. The GABA A receptor was identified by determining the binding activity of 5 nM [ 3 H]FMZ. Positive fractions were pooled, and 0.5 mg of BSA plus protease inhibitors was added. Alternatively, the immunoaffinity columns were eluted by treatment with 2% SDS in 10 mM Tris-HCl, pH 6.8, for 30 min at room temperature.
For immunopurification in series, the anti-␣ 1 -immunopurified receptor was recirculating through 0.1-ml anti-␣ 3 immunoaffinity columns. After washing, the immunobeads were either aliquoted for binding assays or eluted with SDS for immunoblot analysis of the retained material (23).
The immunopurification was quantified by determining the binding activity of [ 3 H]FMZ (5 nM) in the solubilized receptor, the column filtrate, and the final eluate.
Other Methods-The affinity purification of the bovine GABA A receptor complex and immunoblots has been described elsewhere (27). Immunoblots were developed with luminol as recommended by the manufacturer (Boehringer Mannheim). Proteins were determined by the method of Lowry et al. (28). SDS-polyacrylamide gel electrophoresis was done according to the method of Laemmli (29). Photoaffinity labeling and fluorography were performed as described (20).
Data Analysis-Scatchard transformations of the saturation curves and the [ 3 H]FNZ displacement curves were adjusted using LIGAND (30), as described in detail elsewhere (6). The significance (p Ͻ 0.05) of the fit of the displacement curves to a one or two binding site model was determined by the F ratio test. The proportions of the different binding sites were calculated from the "maximal density of binding sites" corresponding to each affinity, determined by LIGAND. Densitometric analysis of both the immunoblots and the fluorographs was performed as described (26). 3 and -␣ 1 Antibodies-Specific polyclonal antibodies have been generated against peptides from the NH 2 -terminal domain (amino acids 1-10) of the ␣ 3 subunit or the C-terminal domain (amino acids 419 -428) of ␣ 1 subunit of the GABA A receptor complex. Both polyclonal antibodies immunoprecipitated the native receptors solubilized from rat cortical membranes in a dose-dependent manner (not shown). The maximal immunoprecipitation (64 Ϯ 2 or 22 Ϯ 4% of the [ 3 H]FMZ binding) was achieved with 5 or 25 l of anti-␣ 1 or anti-␣ 3 antibody, respectively. Similar results were obtained with 10 nM [ 3 H]FNZ (not shown). A second round of incubation with saturating amounts of either antibody immunoprecipitated less than 5% of either binding activity (not shown).
2) The immunoprecipitation of the native receptors was specifically inhibited by the peptide used as antigen but not by others corresponding to similar NH 3 ϩ -or COO Ϫ -terminal regions of other ␣ subunits (Fig. 1, B and C). 3) In immunoblots using the affinity-purified GABA A receptor (Fig. 1D) anti-␣ 3 antiserum immunoreacted with a faint band of M r 59,000 (␣ 3 subunit), whereas anti-␣ 1 strongly reacted with a single band of M r 51,000 (␣ 1 subunit). 4) In immunoblots using extracts from cortical membranes (Fig. 1E) anti-␣ 3 immunostained two bands of M r 59,000 and 61,000 (see Ref. 15) and, on the other hand, anti-␣ 1 recognized a M r 51,000 peptide. A nonspecific band of 100 kDa was also observed in some experiments (also see Fig. 2B). The mAb 62-3G1 (specific to ␤ 2 and ␤ 3 , M r 55,000 -57,000 peptides; Refs. 27 and 31) was included as a control. 5) In the three brain regions studied (cortex, hippocampus, and cerebellum), the percentage of immunoprecipitation by these antibodies is consistent with the level of expression of ␣ 3 or ␣ 1 subunits, determined by in situ hybridization or immunoprecipitation (18,19,(32)(33)(34)(35). As expected, the anti-␣ 1 antiserum immunoprecipitated most of the [ 3 H]FMZ binding activity from the cerebellum (85.3 Ϯ 7.5%), followed by the cortex (71.0 Ϯ 5.3%) and hippocampus (52.2 Ϯ 2.0%). Anti-␣ 3 antiserum immunoprecipitated a low proportion of receptors compared with anti-␣ 1 . The maximal immunoprecipitation was obtained in the cortex (25.8 Ϯ 4.7%), followed by the hippocampus (19.1 Ϯ 3.2%) and cerebellum (9.8 Ϯ 3.5%). In conclusion, by all these criteria both antibodies are specific for their corresponding subunits.
Association between ␣ 1 and ␣ 3 Subunits-To determine the presence of ␣ 1 subunits, co-assembled with ␣ 3 subunits in the same receptor complex, we first quantified the [ 3 H]zolpidem binding activity immunoprecipitated by anti-␣ 3 antiserum. [ 3 H]Zolpidem binds with high affinity to ␣ 1 subunit-containing GABA A receptors (type I benzodiazepine binding sites) (11,12,21,22). Therefore, [ 3 H]zolpidem (5 nM) binding activity was used as a marker of the presence of ␣ 1 subunits in the immunoprecipitated receptor (also see Ref. 21). The quantitative immunoprecipitation of [ 3 H]zolpidem binding was tested by two sequential incubations with anti-␣ 1 or anti-␣ 3 antibodies. The second incubation yielded 3.3 Ϯ 2.8 and 3.8 Ϯ 1.7% of immunoprecipitation for anti-␣ 1 and -␣ 3 , respectively, indicating that the immunoprecipitation of the receptor was maximal.
As shown in Fig. 2A, anti-␣ 1 and -␣ 3 antibodies immunoprecipitated 90.0 Ϯ 5.4 and 26.9 Ϯ 3.6% of the [ 3 H]zolpidem binding activity, respectively. These results demonstrated that, in native GABA A receptors, the high affinity [ 3 H]zolpidem (5 nM) binding sites (type I benzodiazepine binding sites) are largely associated with the presence of an ␣ 1 subunit (also see Refs. 21 and 22) and, importantly, that these sites can be immunoprecipitated in association with ␣ 3 subunits.
The co-purification of both ␣ subunits was not due to crossreaction between the antibodies. As shown in Fig. 2C (lane 2), the anti-␣ 1 antibody produces absolutely no immunoreaction products in Western blots of the anti-␣ 3 -immunopurified receptors that have been immunodepleted of the ␣ 1 subunits. On the other hand, the anti-␣ 3 antibody immunoreacted with 59 -61-kDa peptides (Fig. 2C, lane 3), demonstrating the presence of ␣ 3 -containing GABA A receptors. Conversely, no immunoreaction products were produced by the anti-␣ 3 antibody using the ␣ 3 immunodepleted and anti-␣ 1 -immunopurified receptor as an antigen (Fig. 2C, lanes 5 and 6). Furthermore, in membrane preparations of human embryonic kidney cells, transfected with the ␣ 1 -␥ 2 -␤ 2 combination, the anti-␣ 1 antibody immunostained a single 51-kDa peptide, whereas no immunoreaction products were detected using the anti-␣ 3 antibody (not shown). These results clearly demonstrate the absence of cross-reaction between the antibodies and confirm the co-purification of both ␣ subunits in the same receptor complex.
It could be argued that the co-purification of two different ␣ subunits, such as ␣ 1 and ␣ 3 , was due to interactions between individual GABA A receptor complexes through cytoskeletal elements (36) or, on the other hand, to the presence of anomalous receptors (partially assembled receptors) due to the solubilization of intracellular stores in our membrane preparations (37). Additional control experiments were performed to test these possibilities. Treatment of the cortical membranes, prior to solubilization and purification, with 5 g/ml demecolcine (a tubulin-depolymerizing agent) or 10 g/ml cytochalasin D (an actin-depolymerizing agent) did not modify the percentage of  4) immunoaffinity columns. After washing, the immunoaffinity columns were treated with SDS, and the eluted receptor was analyzed by immunoblot using 5 g of purified anti-␣ 1 (lanes 1 and 3) or anti-␣ 3 (lanes 2 and 4) antibodies. C, the ␣ 1 or ␣ 3 subunits were immunodepleted from the solubilized receptor by three sequential incubations with anti-␣ 1 (15 l each) and anti-␣ 3 (75 l each), respectively. The ␣ 1 -or ␣ 3 -immunodepleted receptors were immunopurified by anti-␣ 3 (lanes 2 and 3) and anti-␣ 1 (lanes 5 and 6) immunoaffinity columns. Lanes 1 and 4, anti-␣ 1 and anti-␣ 3 control immunopurifications, respectively. The eluted receptor was analyzed by immunoblot using 5 g of purified anti-␣ 1 (lanes 1, 2, and 6) or anti-␣ 3 (lanes 3-5) antibodies. For all immunoblots, the affinity-purified antibodies were labeled with digoxigenin. Numbers on the left, M r values of the immunostained bands.  Fig. 2A were obtained using purified synaptic membranes as starting material (not shown).
The co-assembling of ␣ 1 and ␣ 3 in the same receptor complex also could be due to redistribution of subunits during solubilization. This possibility was tested by determining the immunoprecipitation by anti-␣ 3 of the diazepam-insensitive [ 3 H]Ro 15-4513 binding sites in solubilized receptors from cerebellar membranes or from the mixture (1:1) of cerebellar plus cortical membranes. The diazepam-insensitive binding sites are associated with the presence of ␣ 6 subunits (38,39), and this subunit is not expressed in the cortex (32,33,39). The immunoprecipitation of diazepam-insensitive [ 3 H]Ro15-4513 binding activity by anti-␣ 3 was very low and similar in both solubilized preparations, pure cerebellar membranes, and a mixture of cerebellar and cortical membranes (0.01 Ϯ 0.01 and 0.01 Ϯ 0.01 pmol, n ϭ 2, respectively), thus indicating that no apparent subunit redistribution takes place due to solubilization procedures.
The association between both ␣ subunits was quantified by immunodepletion experiments. In these experiments, a particular ␣ subunit was depleted by two sequential immunoprecipitations with the specific antiserum. After depletion, the remaining GABA A receptor complex was immunoprecipitated by the other ␣ subunit. As shown in Table II, depletion of ␣ 1 subunits produced a significant decrease in the [ 3 H]FMZ binding activity immunoprecipitated by anti-␣ 3 antiserum (0.21 Ϯ 0.04 versus 0.10 Ϯ 0.01 pmol, respectively). Thus, 54.7 Ϯ 5.0% of the benzodiazepine binding activity immunoprecipitated by anti-␣ 3 was depleted by preincubation with the anti-␣ 1 antiserum. On the other hand, most of the [ 3 H]zolpidem immunoprecipitated by the anti-␣ 3 antiserum was depleted by preincubation with the anti-␣ 1 antibody (89.0 Ϯ 7.8%; 0.05 Ϯ 0.01 versus 0.005 Ϯ 0.004 pmol). These results indicated that most, if not all, of the high affinity binding sites immunoprecipitated by the anti-␣ 3 antiserum were due to the presence of an ␣ 1 subunit.
Reciprocally, depletion of ␣ 3 subunits also affected the immunoprecipitation by the anti-␣ 1 antiserum. As shown in Table  II Thus, 20 -25% of the ␣ 1 population is associated with an ␣ 3 subunit in the same receptor complex.
As mentioned above, the second immunoprecipitation with either antibody or for either binding site yielded a residual 3-4% of immunoprecipitation. Therefore, after two rounds of immunoprecipitation, a particular ␣ subunit should be completely depleted from the solubilized material. However, no attempts were made to detect the depleted subunit remaining in the supernatants. We are aware that some residual amounts of the depleted subunit could persist in the solubilized receptor. Therefore, these results could be, in some extent, underestimated.

Pharmacological Properties of Anti-␣ 1 -or Anti-␣ 3 -immunopurified Receptors from Rat Cortex-The pharmacological properties of the anti-␣ 1 -or anti-␣ 3 -immunopurified receptors were determined by [ 3 H]zolpidem and [ 3 H]FNZ saturation studies and by displacement experiments using aliquots from the immunoaffinity columns.
As shown in H]FNZ were similar in both anti-␣ 3 or anti-␣ 1 -immunopurified receptors and very close to K D values obtained in crude cortical rat membranes for type I or total benzodiazepine binding sites (4,6,24,40).
The presence of two pharmacologically distinct receptors in the anti-␣ 3 and anti-␣ 1 -immunopurified receptors also has been tested by displacement studies of [ 3 H]FNZ by type I-specific ligands (such as zolpidem or Cl 218,872). In both anti-␣ 1and anti-␣ 3 -immunopurified receptors and for both zolpidem and Cl 218,872, the Hill slope of the displacement curves was lower than unity (Table III), indicating the existence of a heterogeneous population of binding sites. Furthermore, in every case, displacement curves were better fitted (based on the extra sum of squares using the program LIGAND, three of three experiments; p Ͻ 0.05) to a two binding site model with high (K i , 5-8 and 40 -70 nM for zolpidem and Cl 218,872, respectively) and low (K i , 390 -430 nM and 2-3.5 M for zolpidem or CL 218,872, respectively) affinity (Table III). In contrast, diazepam displaced the [ 3 H]FNZ binding immunopurified by the anti-␣ 3 antibody, with a Hill coefficient of 1.2 and a single high affinity site (K i , 7.1 nM). It is interesting that all the calculated K i values were very similar to those determined in crude cortical membrane preparations for type I and II benzodiazepine receptors (6,24).
Immunopurification and Pharmacological Properties of ␣ 3 -␣ 1 GABA A Receptors-The association of ␣ 3 and ␣ 1 subunits and the pharmacological properties of the ␣ 3 -␣ 1 GABA A receptors were further tested by using sequential immunopurification (see Refs. 14 and 16). The GABA A receptor was first immunopurified by anti-␣ 1 immunoaffinity columns, and the eluted

and ␣ 3 subunits by immunodepletion experiments
The immunodepletion of ␣ 1 or ␣ 3 subunits was done by two sequential incubations of the solubilized receptor (1 pmol of [ 3 H]FMZ binding sites) with either anti-␣ 1 (10 ϩ 10 l) and anti-␣ 3 (75 ϩ 75 l) antibodies, respectively. The remaining receptor was immunoprecipitated by incubation with 75 or 10 l of anti-␣ 3 and -␣ 1 antisera, respectively. The binding activity was determined in pellets and supernatants. The results are expressed as pmol of [ 3 H]FMZ (5 nM) or [ 3 H]zolpidem (5 nM) specific binding activity immunoprecipitated in each condition. The percentage of depletion was calculated from the specific binding activity immunoprecipitated before and after depletion. Data are mean Ϯ S.D. of three to six independent experiments. material was absorbed to anti-␣ 3 immunoaffinity columns. As was mentioned before (see Table I), no significant elution from the anti-␣ 3 immunoaffinity columns could be achieved. Therefore, the GABA A receptor retained by the anti-␣ 3 columns was analyzed by binding assays using aliquots of the anti-␣ 3 immunobeads (see Ref. 23) or by immunoblots of the SDS-eluted receptors.
The stoichiometry between both ␣ subunits was estimated by densitometric analysis of semiquantitative immunoblots ( Fig.  4; also see Ref. 16). For these experiments, a fixed amount of receptor was immunoblotted and incubated with increasing concentrations of both antibodies in combination. After 4 h of incubation, the medium was aspirated and replaced by a new batch of antibodies. The immunoreaction products were quantified by densitometry. As shown in Fig. 4, A and B, at saturating concentrations, both antibodies yielded similar amounts of immunoreaction products. Thus, these results indicated a stoichiometry of approximately 1:1 (␣ 3 /␣ 1 ratio, 1.1 Ϯ 0.1, n ϭ 2; Fig. 4B).
Finally, we have tested the pharmacological properties of the

TABLE III Pharmacological characterization of the anti-␣ 1 and anti-␣ 3 immunopurified GABA A receptors from rat cortex
The solubilized receptor was immunoabsorbed to anti-␣ 3 or anti-␣ 1   periments, was 70:30 for high and low affinity, respectively (see Table IV).

DISCUSSION
The molecular composition of native GABA A receptors is unknown. Evidence is accumulating for the existence of different ␣ subunit combinations (such as ␣ 1 ␣ 3 , ␣ 1 ␣ 2 , ␣ 1 ␣ 5 , and ␣ 1 ␣ 6 ) co-assembled in single native GABA A receptor complexes. However, other studies also indicated the absence of co-existence between different ␣ subtypes (18,19). On the other hand, it is currently accepted that the benzodiazepine binding properties of the GABA A receptors are mainly determined by the ␣ subunits (10, 11, 12). Therefore, if two different ␣ subunit subtypes are co-assembled in a single GABA A receptor, two pharmaco-logically different benzodiazepine binding sites could co-exist in a single complex. In the present article we have investigated the possible existence and the pharmacological properties of native ␣ 1 -␣ 3 GABA A receptors from the rat cortex.
The presence of ␣ 1 and ␣ 3 subunits in the same GABA A receptor complex was demonstrated by immunoprecipitation and immunopurification experiments. It has been described that, in transfected GABA A receptors, the high affinity binding sites for zolpidem (type I benzodiazepine binding sites) are determined by the presence of ␣ 1 subunits. Other ␣ subtypes (such as ␣ 2 , ␣ 3 , and ␣ 5 ) confer low affinity for this ligand (type II benzodiazepine receptors) (11,12). Therefore, the association of an ␣ 1 with other ␣ subunits could be estimated by immunoprecipitation of the [ 3 H]zolpidem binding activity. Our immunoprecipitation experiments ( Fig. 2A and Refs. 21 and 22) demonstrate that most, if not all (90.0 Ϯ 5.4%), of the high affinity binding sites for zolpidem are due to the presence of an ␣ 1 subunit in the GABA A receptor. Importantly, 25-30% of these [ 3 H]zolpidem binding sites were immunoprecipitated by anti-␣ 3 antibody, thus suggesting an ␣ 1 -␣ 3 association. Consistently, immunodepletion of the ␣ 1 subunits suppress the immunoprecipitation of [ 3 H]zolpidem binding by the anti-␣ 3 antibody (Table II). Thus, ␣ 1 binding properties could be immunoprecipitated in association with ␣ 3 subunits, suggesting the presence of both ␣ 1 and ␣ 3 subunits in the same receptor complex.
The association between both subunits was confirmed by immunopurification experiments. The immunoblot analysis of the anti-␣ 1 -or anti-␣ 3 -immunopurified receptors (Fig. 2B) revealed the presence of ␣ 3 immunoreaction product in the anti-  ␣ 1 -immunopurified receptors and, reciprocally, the presence of ␣ 1 in anti-␣ 3 -immunopurified receptors. Furthermore, the association between both ␣ subunits was not due to interactions with cytoskeletal elements. Taken together, these results demonstrated the existence of ␣ 1 -␣ 3 GABA A receptors from the rat cortex. Immunodepletion experiments indicated that the ␣ 1 -␣ 3 GABA A receptors constituted a relatively minor proportion of the total ␣ 1 -containing GABA A receptors (20 -25% of this population) but 50 -55% of the ␣ 3 containing GABA A receptors. Thus, and in partial agreement with previous reports (13)(14)(15), the association between two different ␣ subunits represents a minor population from the total ␣ 1 -containing receptors but a high proportion of other ␣ subunits, such as ␣ 3 . The presence of different ␣ subtypes, in combination with ␤ 1 -␤ 3 and ␥ 2 subunits, determines the benzodiazepine binding properties of recombinant GABA A receptors (10,11,12). As mentioned above, the ␣ 1 subunit confers type I benzodiazepine binding properties (high affinity for zolpidem and Cl 218,872), whereas the ␣ 3 subunit confers type II binding properties (low affinity for these ligands). Therefore, if two different ␣ subunits, such as ␣ 1 and ␣ 3 , are co-assembled in the same receptor complex, and both ␣ 1 and ␣ 3 subunits are pharmacologically active, two different benzodiazepine binding subtypes should be discriminated in either anti-␣ 1 -and anti-␣ 3 -immunopurified receptors. As shown in Table III, in anti-␣ 1 -and anti-␣ 3 -immunopurified receptors, two different binding sites were identified. The affinities for zolpidem (determined by Scatchard and displacement experiments) or Cl 218,872 (determined by displacement experiments) were similar in both immunopurified receptors and similar to those reported for type I and II benzodiazepine binding sites in cortical membranes (6,7). Furthermore, the affinities for both ligands corresponded to those reported for recombinant receptors containing ␣ 1 subunits (high affinity binding sites) and ␣ 3 subunits (low affinity binding sites) (11,12). In consequence, these results suggest the presence of benzodiazepine binding sites in both ␣ 1 and ␣ 3 subunits co-assembled in a single GABA A receptor complex (also see Ref. 17).
To discern whether both ␣ 1 and ␣ 3 subunits, co-assembled in a single complex, display benzodiazepine binding activity, the GABA A receptor was immunopurified by anti-␣ 1 and anti-␣ 3 affinity columns in series; therefore, the whole population of the isolated GABA A receptors should contain two different ␣ subunits. It is noteworthy that anti-␣ 3 immunoaffinity columns retained 20 -25% of the ␣ 1 -immunopurified GABA A receptors, corroborating the proportion of ␣ 1 to ␣ 3 GABA A receptors calculated by depletion experiments (compare Fig. 3A and Table  II). Immunoblot analysis (Fig. 3B) indicates that ␣ 1 and ␣ 3 subunits are mainly associated with ␤ 2/3 and ␥ 2 in the same receptor complex, consistent with previous experiments (21,22). The ␤ 1 subunits are a relatively minor component of the receptor (41), and, on the other hand, it has been demonstrated that ␥ 1 is not associated with ␥ 2 -containing GABA A receptors (42). Thus, we propose a molecular composition of ␣ 1 , ␣ 3 , ␤ 2/3 and ␥ 2 for these native GABA A receptor complexes from rat cortex.
A relevant question to ascertain the pharmacological activity of the ␣ subunits, co-assembled in a single native GABA A receptor, is the stoichiometry between both subunits in the complex. Thus, we have estimated the stoichiometry between both ␣ subunits by quantifying the immunoreaction products of anti-␣ 1 and anti-␣ 3 antibodies in immunoblots. We are aware that immunoblots are only semiquantitative. However, within the limitations of the technique, the results (Fig. 4) indicated the presence of stoichiometric amounts of each ␣ subunit (ratio 1:1; also see Ref. 16 for discussion). The stoichiometry of ␥ 2 , ␤ 2 , and ␤ 3 subunits was not determined.
If both ␣ subunits display benzodiazepine binding sites, the double-immunopurified receptors should display type I and II binding properties in similar proportions, and two peptides should be photoaffinity labeled by [ 3 H]Ro15-4513 to a similar extent. Indeed, the pharmacological analysis of the ␣ 1 -␣ 3 GABA A receptors indicated the presence of two different benzodiazepine binding sites. Both Cl 218,872 and zolpidem discriminated between two different binding sites with high (type I) and low affinities (type II). The calculated K i values for either ligand were similar to those of immunopurified ␣ 1 or ␣ 3 receptors (compare Tables III and IV) and to cerebral membranes (6,7). However, the proportion between both binding sites (70:30 for high and low affinity, respectively) demonstrates that the ␣ 1 subunits are predominantly active over the ␣ 3 subunits. It could be argued that the different proportions between both binding sites, determined by displacement experiments, is due to differences in the K D values of ␣ 1 and ␣ 3 subunits for the

H-labeled ligand ([ H]FNZ or [ H]FMZ
). However, these results were confirmed by [ 3 H]Ro15-4513 photoaffinity-labeling experiments at three different degrees of saturation. As expected, in the double-immunopurified receptors, two photolabeled peptides of 51 kDa (corresponding to ␣ 1 subunits) and 59 -61 kDa (␣ 3 subunits) were identified. However, despite the fact that both ␣ subunits are assembled in stoichiometric amounts in the same receptor complex, the proportion between both photolabeled peptides (at all three concentrations) was 70:30 for 51 and 59 -61 kDa, respectively (Fig. 5). Thus, ␣ 1 subunits are pharmacologically predominant over the ␣ 3 subunits. It should be noted that [ 3 H]Ro15-4513 photolabeled most, if not all, the benzodiazepine binding sites from cerebral membranes (90% in this work; also see Ref. 43).
Our data could be explained by the existence of at least two pharmacologically different populations of ␣ 1 -␣ 3 GABA A receptors (see Fig. 6 for a model). As shown in Fig. 6A, 70% of the ␣ 1 -␣ 3 GABA A receptors may be assembled by a functional ␣ 1 subunit associated with an inactive ␣ 3 subunit. The remaining 30% of the population may be constituted by a functional ␣ 3 subunit associated with inactive ␣ 1 subunits. Nevertheless, we cannot completely exclude the existence of ␣ 1 -␣ 3 GABA A receptors containing two benzodiazepine binding sites (Fig. 6B). In such a model, in which two functional ␣ 1 and ␣ 3 subunits are co-localized in the same receptor complex, 60% of the benzodiazepine binding sites should be conferred by GABA A receptors containing two functional ␣ subunits and 30% by functional ␣ 1 subunits associated with inactive ␣ 3 subunits. Our results do not allow discrimination between these two models.
The presence or absence of active benzodiazepine binding sites could be determined by the distribution of the ␣ and ␥ 2 subunits in the pentameric GABA A receptor complex (10,44). It has been proposed that both ␣ and ␥ 2 subunits are implicated in the benzodiazepine binding sites (44,45), and, on the other hand, the GABA A receptors may contain two ␣ subunits, two ␤ subunits, and a single ␥ 2 subunit (46,47). Thus, the predominance of ␣ 1 pharmacology (type I benzodiazepine binding sites or the 51-kDa photolabeled peptides) could be interpreted by the presence of a single ␥ 2 subunit properly associated with the ␣ 1 subunit in the ␣ 1 -␣ 3 GABA A receptor complex (Fig. 6A). In these receptors, the ␣ 3 subunits should lack the benzodiazepine binding sites (see Fig. 6). On the other hand, two different ␥ 2 subunits could also co-exist in the same receptor complex (46,48). If this is the case, both ␣ subunits could display benzodiazepine binding properties (Fig. 6B).
The physiological significance of GABA A receptors containing two different ␣ subtypes, such as an ␣ 1 -␣ 3 combination, is unknown. The ␣ 1 subunit is highly and uniformly expressed in all cortical layers, whereas the expression of the ␣ 3 subunit is localized in layers V and VI (49). Therefore, the ␣ 1 -and ␣ 3containing GABA A receptors should be restricted to these cortical layers. Co-localization of ␣ 1 and ␣ 3 subunits has been also observed in other discrete brain regions (such as mitral cells of the olfactory bulb and the medial septum; Ref. 49). On the other hand, in recombinant GABA A receptors, the co-expression of ␣ 1 , ␣ 3 , ␤ 2 , and ␥ 2 subunits confers unique functional properties, distinct from GABA A receptors containing a single ␣ subtype (50,51). Therefore, the presence and pharmacological activity of two different ␣ subunit subtypes in native receptor complexes, localized in discrete brain areas and/or cellular regions, could influence the functional and pharmacological properties of the GABA A receptor. The existence and pharmacological properties of ␣ 1 ␣ 3 -containing receptors increase the heterogeneity of the native GABA A receptor complex in the central nervous system.
In summary, our results demonstrate the existence of cortical GABA A receptors containing both ␣ 1 and ␣ 3 subunits in stoichiometric amounts. Furthermore, both ␣ subunits retained their benzodiazepine binding properties. However, the ␣ 1 subunit is pharmacologically predominant over ␣ 3 subunits, indicating the existence of active and nonactive benzodiazepine binding sites associated with these ␣ subunits.