Quantitative Characterization of α6 and α1α6 Subunit-containing Native γ-Aminobutyric AcidAReceptors of Adult Rat Cerebellum Demonstrates Two α Subunits per Receptor Oligomer

-Aminobutyric acidA (GABAA) receptors were purified from adult rat cerebella by anti-α6(1-16 Cys) antibody affinity chromatography. Immunoblots of the α6 subunit-containing receptors showed the copurification of the α1, β2/3, 2, but not α2 and α3 GABAA receptor polypeptides. Further fractionation of this receptor subpopulation by anti-GABAA receptor subunit α6(1-16 Cys) and anti-α1(413-429) antibody affinity columns in series substantiated the coassociation of the α1 and α6 polypeptides. The percentage of coexistence of the two subunits was determined by quantitative immunoblotting, which found that 41 ± 12% of α6 subunit immunoreactivity is associated with the α1 subunit. The ratios of the α1:α6 subunits in the double purified receptor preparations was found to be 1:1, thus determining directly for the first time subunit ratios within native GABAA receptors. The benzodiazepine pharmacology of the α1α6 subunit-containing receptors was shown to be predominantly benzodiazepine-insensitive by quantitative immunoprecipitation assays. These results are the first direct quantitative studies of subunit ratios within a population of native GABAA receptors.

␥-Aminobutyric acid A (GABA A ) receptors were purified from adult rat cerebella by anti-␣6(1-16 Cys) antibody affinity chromatography. Immunoblots of the ␣6 subunit-containing receptors showed the copurification of the ␣1, ␤2/3, ␥2, ␦ but not ␣2 and ␣3 GABA A receptor polypeptides. Further fractionation of this receptor subpopulation by anti-GABA A receptor subunit ␣6(1-16 Cys) and anti-␣1(413-429) antibody affinity columns in series substantiated the coassociation of the ␣1 and ␣6 polypeptides. The percentage of coexistence of the two subunits was determined by quantitative immunoblotting, which found that 41 ؎ 12% of ␣6 subunit immunoreactivity is associated with the ␣1 subunit. The ratios of the ␣1:␣6 subunits in the double purified receptor preparations was found to be 1:1, thus determining directly for the first time subunit ratios within native GABA A receptors. The benzodiazepine pharmacology of the ␣1␣6 subunit-containing receptors was shown to be predominantly benzodiazepine-insensitive by quantitative immunoprecipitation assays. These results are the first direct quantitative studies of subunit ratios within a population of native GABA A receptors.
The GABA A 1 receptors of mammalian brain are fast-acting ligand-gated chloride ion channels. Multiple genes encoding GABA A receptor subunits have been identified by molecular cloning. These are classified on the basis of their respective amino acid sequence similarities into five subunit types. Thus known mammalian GABA A receptor subunits are the ␣1-␣6, ␤1-␤3, ␥1-␥3, ␦, and 1-2, comprising 15 identified to date (for review, see McDonald and Olsen (1994)). Different combinations of these subunits are thought to assemble, probably in pentameric combinations, in vivo to form functional GABA A receptors. The polypeptide complement of any one native GABA A receptor has not been elucidated. However, the biophysical and pharmacological properties of cloned receptors suggest that most probably consist of an ␣␤␥ combination (Olsen and McDonald, 1994;Stephenson, 1995). A subunit stoichiometry (␣) 2 (␤) 1 (␥) 2 was found for a defined expressed GABA A receptor (Backus et al., 1993). For native receptors, a five-fold axis of symmetry was revealed by negative stain elec-tron microscopy thus providing the first evidence for a pentameric quaternary structure in vivo (Nayeem et al., 1994). Furthermore, Khan et al. (1994) recently deduced by immunoprecipitation studies that a GABA A receptor subpopulation in the cerebellum consisted of ␣1␣6␥2 S ␥2 L ␤2/3 subunits where ␥2 S and ␥2 L are splice variants of the ␥2 subunit.
We have adopted the approach of the determination of native GABA A receptor subunit complements by the purification of subsets of receptors by immunoaffinity chromatography using subunit-specific antibodies (e.g. Duggan et al. (1991) and Pollard et al. (1993)). We have focused primarily on the ␣ subunit complements of native receptors. We found, in agreement with several other groups, that the majority of native GABA A receptors contain a single ␣ subunit variant (e.g. Duggan et al. (1991), McKernan et al. (1991, and Benke et al. (1991)). However, we identified, by the use of different specificity antibody affinity columns in series, minor receptor populations that were heterogeneous with respect to their ␣ subunit complement e.g. ␣1␣2, ␣1␣3, and ␣2␣3 subunit-containing receptors Pollard et al., 1993). Thus, we proposed the assembly of at least two ␣ subunits per receptor oligomer. Several other groups have since reported the copurification or coimmunoprecipitation of other ␣ subunit variants (e.g. Mertens et al. (1993) and Kern and Sieghart (1994); summarized by Stephenson (1995)).
We have recently concentrated efforts on the GABA A receptor ␣6 subunit-containing receptors. This is because this subunit is uniquely expressed in adult rat brain in a single cell type, the cerebellar granule cell Thompson et al., 1992). Furthermore, cloned ␣6␤x␥x (and also ␣4␤x␥x) receptors have a distinct benzodiazepine pharmacology in that they have high affinity for the partial inverse agonist Ro 15-4513 but very low affinity for the classical allosteric benzodiazepine regulators such as diazepam . This pharmacological profile corresponds to the previously described diazepam-insensitive site (abbreviated in this paper to the benzodiazepine-insensitive site, BZ-IS) (Sieghart et al., 1991). We previously reported the purification of calf cerebellar ␣6 subunit-containing GABA A receptors, but low yields in the isolation procedure precluded their detailed characterization (Pollard et al., 1993). The isolation efficiency has been improved by using adult rat cerebellum as starting material, thus permitting further analysis, including for the first time quantitative measurements of immunoreactivity. We report these results in this paper.

Methods
Production of GABA A Receptor ␦ Subunit Antibodies-The peptides ␦(320 -337 Cys), DYRKKRKAKVKVTKPRAEC, and ␦(2-12), PHHGA-RAMNDIC, were coupled via the C-terminal cysteine to keyhole limpet hemeocyanin. Polyclonal antibodies were raised in rabbits and were affinity-purified by the respective peptide affinity resins all as described previously . Both specificity anti-␦ subunit antibodies recognized a M r , 57,000 Ϯ 500 immunoreactive band in addition to a protein with M r 64,000. The higher molecular weight species has been observed with other cysteine-coupled peptides with completely different amino acid sequences; it is a nonspecific, non-␦ subunit protein.
Preparation of Membrane-bound, Detergent-solubilized, and Detergent-treated Membrane Fractions from Rat and Calf Brain-Membranes and Na ϩ deoxycholate extracts of rat and calf cerebellum were prepared as described previously (Duggan and Stephenson, 1990). The pellet obtained after Na ϩ deoxycholate solubilization was rehomogenized with 10 mM HEPES, pH 7.4, containing 1 mM EDTA, 1 mM benzamidine, centrifuged at 20,000 ϫ g for 30 min at 4°C, and the pellet retained. This was then resuspended as above and termed the detergent-treated membranes.
Immunoaffinity Purification of GABA A Receptors-Immunoaffinity purification of GABA A receptors from Na ϩ deoxycholate extracts of adult rat cerebellum was carried out using a rabbit anti-␣6(1-16 Cys) Fab antibody fragment affinity column, a sheep anti-␣1(413-429) whole antibody affinity column, or the anti-␣6(1-16 Cys) Fab and anti-␣1(413-429) antibody affinity columns in series all exactly as described previously, except that in the double immunoaffinity column purification experiments, O.05% (w/v) bovine serum albumin was added to the single GABA A receptor ␣6 subunit-purified material Pollard et al., 1993). Purified receptor subpopulations were analyzed by both immunoblotting and radioligand binding assays.
Polyacrylamide Gel Electrophoresis and Immunoblotting-Immunoblotting was carried out as before using sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) under reducing conditions in 10% polyacrylamide slab gels, the chloroform/methanol method for protein precipitation, and the ECL Western blotting system for detection (Duggan et al., 1992). Immunoblots were quantified by densitometry using the Molecular Dynamics personal densitometer.
Radioligand Binding and Immunoprecipitation Assays-Radioligand binding and immunoprecipitation assays were both carried out as described previously using the polyethyleneimine assay for the measurement of specific ligand binding activities (Stephenson and Duggan, 1990). For the measurement of [ 3 H]Ro 15-4513 binding activity, total binding was determined using 10 M Ro 15-4513 to define nonspecific binding. For the measurement of benzodiazepine-sensitive (BZ-S) [ 3 H]Ro 15-4513 binding activity, 10 M flunitrazepam was used in the binding assay. BZ-IS [ 3 H]Ro 15-4513 binding activity was thus calculated as the difference between total and the benzodiazepine-sensitive (BZ-S) binding activity. RESULTS We have previously reported the purification of ␣6 subunitcontaining GABA A receptors from calf cerebellum by anti-␣6(1-16 Cys) Fab antibody affinity chromatography (Pollard et al., 1993). In that study we reported the copurification of the ␣6 and ␣1 subunit immunoreactivities. However, because of the low yield in the isolation procedure, we were unable to characterize in detail the coassociation of ␣6 and ␣1 subunit immunoreactivities by the use of different specificity antibody columns in series, as has been employed for other GABA A receptor subpopulations (cf. Duggan et al., 1991;Pollard et al., 1993). Furthermore, there were no detectable radioligand binding activities in the purified receptor preparations, thus precluding the characterization of the pharmacological properties of native GABA A receptor ␣6 subunit-containing receptors. An investigation of the efficiency of the purification procedure showed that for the standard conditions of receptor solubilization using 0.5% (w/v) Na ϩ deoxycholate and 150 mM KCl, the majority of ␣6 subunit immunoreactivity remained in the calf cerebellar detergent-treated membranes (Fig. 1). This was in contrast to the efficiency of solubilization under the same conditions from adult rat cerebellum, where at least 50% of the ␣6 subunit immunoreactivity was found in the solubilized preparation ( Fig. 1). The comparative results for the extraction of BZ-S and BZ-IS [ 3 H]Ro 15-4513 binding sites from rat cerebellum are shown in Table I. There was no appreciable difference in the efficiency of detergent solubilization between the two pharmacological classes of receptor. Furthermore, the percentage efficiency of solubilization of binding sites agreed with that found for ␣6 subunit immunoreactivity (Table I and Fig. 1). This is in contrast to Uusi-Oukari (1992), who showed that BZ-S sites were preferentially solubilized from pig cerebellum. Although this differential sensitivity to detergent extraction appears to be species-dependent, it is important also to make the point that it may reflect differences between receptor subtypes in their subcellular compartmentalization and/or association with cytoskeletal elements particularly at less accessible sites such as synapses. For the results herein and indeed for all other papers addressing native GABA A receptor subunit complements by biochemical approaches, the assumption is made that the solubilized preparation is representative of the entire functional, ␣6 subunit-containing GABA A receptor population.
Purification and Characterization of ␣6 Subunit-containing GABA A Receptors from Adult Rat Cerebellum-The alternative use of rat cerebellum as the source of ␣6 subunit-containing GABA A receptors with the increased efficiency of solubilization permitted their detailed biochemical characterization.  1. Solubilization of GABA A receptor ␣6 immunoreactivity from adult rat and calf cerebellum. Membranes, Na ϩ deoxycholate extract, and Na ϩ deoxycholate-treated membranes were all prepared from adult rat and calf cerebellum. Samples (25 g of protein/gel lane) were precipitated and analyzed by immunoblotting using affinity-purified anti-␣6(1-16 Cys) antibodies (2.5 g, final concentration) all as described under "Experimental Procedures." Lane 1, benzodiazepine affinity-purified GABA A receptor from calf cortex; lanes 2-4, calf cerebellum; lanes 5-7, rat cerebellum. Lanes 2 and 5, detergent-treated membranes; lanes 3 and 6, Na ϩ deoxycholate-solubilized membranes; lanes 4 and 7, membranes. The positions of prestained protein standards (kDa ϫ 10 Ϫ3 ) are on the left; ␣1 and ␣6 refer to the antibodies used for immunoblotting. activity, it can be seen that 18% of the sites applied are retained by the anti-␣6(1-16 Cys) Fab antibody column. There is a recovery of 4% of the total [ 3 H]Ro 15-4513 sites following pH 11.5 elution. This efficiency is the same order of magnitude found using different specificity GABA A receptor antibody columns, e.g. for ␣1 and ␣2 . When the pharmacology of the total [ 3 H]Ro 15-4513 binding activity bound to the anti-␣6(1-16 Cys) Fab antibody column was subfractionated into the BZ-S and BZ-IS sites, typical values found were 75% of the BZ-IS sites compared to 7% of the BZ-S sites (Table II). Mean values obtained for the retention of BZ-IS and BZ-S were 74 Ϯ 3% and 2 Ϯ 5%, respectively (n ϭ 4). From Table III, it can be seen that the percentage of ␣6 subunit immunoreactivity (79 Ϯ 7%) and BZ-IS sites retained by the anti-␣6(1-16 Cys) Fab antibody affinity column were the same.
In the pH 11.5 eluted fractions, again, BZ-S and BZ-IS [ 3 H]Ro 15-4513 binding activity were both assayed. The results were difficult to quantify because of the low level of total [ 3 H]Ro 15-4513 binding activity, but it was observed that flunitrazepam displaced a significant proportion of the total binding activity (see below). There was no significant retention of specific [ 3 H]flunitrazepam binding sites by the anti-␣6(1-16 Cys) antibody column. In agreement, [ 3 H]flunitrazepam binding to the purified receptors was not detectable (n ϭ 2, results not shown).
To determine the other GABA A receptor polypeptides that copurified with the ␣6 subunit-containing GABA A receptors, the pH 11.5 fractions 2-4 were pooled and analyzed by immunoblotting. Fig. 2 shows the results, where it can be seen that ␣1, ␤2/3, ␥2, and ␦ subunit immunoreactivities were found. As for the bovine preparations, there was no detectable ␣2 or ␣3 subunit immunoreactivities. Note that we were unable to use the anti-␤3(379 -393) antibody ; this was raised to the bovine ␤3(379 -393) sequence. The homologous rat sequence has three amino acid differences, and the rat ␤3 subunit is not recognized by the bovine antibody. The ␤ subunits were thus identified by the monoclonal antibody bd-17, which recognizes both rat ␤2 and ␤3 subunits.
Quantification of the Percentage Coassociation of the ␣1 and ␣6 GABA A Receptor Subunits in Adult Rat Cerebellum-The coassociation of the ␣1 and ␣6 subunits was investigated further and quantitatively by using different immunoaffinity columns in series. Thus GABA A receptors were first purified by anti-␣6(1-16 Cys) Fab antibody affinity chromatography. The receptor-containing fractions were pooled and applied to a sheep anti-␣1(413-429) whole antibody affinity column. The affinity column filtrate and the pH 11.5 eluted fractions were analyzed for radioligand binding activity and immunoreactivity. No [ 3 H]Ro 15-4513 binding was detected in the eluted fractions, but ␣6 and ␣1 GABA A receptor subunit immunoreactivities were found and these were quantified by densitometry. Fig. 3 shows representative results for these experiments. On application of the ␣6 subunit-containing receptors to the anti-␣1(413-429) antibody column, it can be seen that all the ␣1 subunit immunoreactivity is retained in contrast to the M r 57,000 ␣6 subunit, which was detectable in the anti-␣1(413-429) antibody affinity column filtrate. Both immunoreactivities were found in the pH 11.5 eluted fractions. Fig. 3 (A and B) shows the elution profile from a single experiment.     (3) summarizes the results obtained from three separate purifications. It demonstrates the coelution of the ␣1 and ␣6 subunit immunoreactivities. The results for the quantification of the immunoblots are summarized in Table III. For these experiments, equal volumes were applied for each sample thus permitting direct comparisons following densitometry. It was found that 79 Ϯ 7% of the applied ␣6 subunit immunoreactivity was retained initially by the anti-␣6 subunit antibody column. On application to the second different specificity antibody column, 41 Ϯ 12% ␣6 subunit immunoreactivity was retarded (n ϭ 2).
To determine the ratios of the ␣1:␣6 subunits in the double purified receptor preparations, two sets of experiments were carried out. First, the primary ␣1 and ␣6 affinity-purified antibody dose dependences were determined, for a fixed antigen concentration, in immunoblots of double immunoaffinity-purified receptors (Fig. 4A). Second, using the antibody concentration at saturation (Fig. 4A), the ␣6␣1 subunit-containing antigen was varied and the resultant immunoreactive bands quantified (Fig. 4B). The ␣1:␣6 subunit ratio was 0.95 Ϯ 0.1, which was the mean value for each antigen concentration and for n ϭ 2 preparations. In immunoprecipitation assays (see below), it was found that the anti-␣6(1-16 Cys) antibody did not pellet all the ␣6 subunit immunoreactivity even when used at saturation. Although immunoblotting and immunoprecipitation are two different experimental paradigms, further investigation was required to ensure that the ␣6 subunit was not being underestimated by using an antibody with low avidity. Thus, two immunoblots were carried out in parallel. In the first a single incubation with a saturating concentration of primary anti-␣6(1-16 Cys) antibody was used. For the second immunoblot, this was processed as for the first except that the initial primary antibody was aspirated and the immunoblot then incubated with a fresh primary anti-␣6(1-16 Cys) antibody at the same saturating concentration (Fig. 4A). Both immunoblots were then quantified by densitometry but no differences in the amount of ␣6 subunit immunoreactivity were found (results not shown).
Immunoblots of ␣6␣1 double immunoaffinity-purified receptors showed the coassociation of ␤2/3 and ␦ subunits (n ϭ 1; results not shown), but attempts to show the localization of the ␥2 subunit have so far been negative.
The same immunoprecipitation assays were carried out on GABA A receptors purified from adult rat cerebellum by anti-␣1(413-429) antibody affinity chromatography. Here, the ␣1 antibody immunoprecipitated close to 100% of total, BZ-S, and BZ-IS [ 3 H]Ro 15-4513 sites as should be the case for an ␣1 subunit-purified preparation. However, the ␣6 subunit antibody immunoprecipitated a maximum of 30 Ϯ 5% of total [ 3 H]Ro 15-4513 sites (i.e. ␣1␣6 subunit containing). When these binding sites were subfractionated into the BZ-IS and BZ-S sites, the values were not significant for the BZ-S but 47 Ϯ 7% compared to a predicted 100% for the BZ-IS sites. Increasing the antibody concentration did not effect the percentage of [ 3 H]Ro 15-4513 binding sites precipitated, but it was found that at these high antibody concentrations, ␣6 subunit immunoreactivity was still present in the supernatant. Thus, the inability to immunoprecipitate all the BZ-IS [ 3 H]Ro 15-4513 sites may be attributed to the low avidity for the antibody. This has been encountered before for immunoprecipitations with the anti-␥2(1-15 Cys) antibody where the problem was overcome by sequential immunoprecipitations with fresh batches of antibody (Duggan et al., 1992). This was not possible here because of the low levels of binding activity. Significantly, when the pharmacology of the immunoprecipitated pellet was determined directly instead of as a percentage of the total starting activity, 91 Ϯ 10% (n ϭ 2) of the [ 3 H]Ro 15-4513 binding was BZ-IS. DISCUSSION In this paper, we have described the purification of ␣6 subunit-containing GABA A receptors with the retention of their [ 3 H]Ro 15-4513 radioligand binding activities. We have sub- stantiated the coexistence of the ␣1 and ␣6 subunit in single receptor oligomers and, in addition, we have quantified their percentage coassociation. In double immunoaffinity-purified receptors, the ␣1:␣6 subunit ratio was 1:1 and the benzodiazepine pharmacology of this subset of receptors was BZ-IS. Thus the use of the ␣6 and ␣1 immunoaffinity columns in series not only proved the coexistence of these two gene products in one receptor (41% of the ␣6 subunit receptor population, Table III) but also showed that in the rat cerebellum, either single variant ␣1 or ␣6 subunit-containing receptors exist. These results are in agreement with the emerging pattern from several groups. That is, that at least for the GABA A receptor ␣ subunits, different isoforms do partially coexist within the same receptor molecule (summarized by Stephenson, 1995). With specific reference to the ␣1 and ␣6 subunits, the findings herein are in agreement with the localization of ␣1 and ␣6 GABA A receptor subunit-like immunoreactivities at the electron microscopic level where synapses in the cerebellum were found containing either ␣1 or ␣6 or both ␣1 and ␣6 polypeptides (Nusser et al., 1995). Moreover, Mathews et al. (1994) coexpressed ␣1, ␣6, ␤2, and ␥2 polypeptides in mammalian cells and showed that the resultant pharmacological properties were distinct from both ␣1␤1␥2 and ␣6␤1␥2 receptors and best explained by an ␣1␣6␤2␥2 hybrid receptor. Quirk et al. (1994), however, found no evidence for coassociation of ␣1 and ␣6 subunits but this may be explained by low avidity antibodies. Similarly, Korpi and Luddens (1993) found no evidence for the coassociation of all four subunits following transient expression in mammalian cells.
For the non-␣ subunits coassociated with the ␣6 polypeptide, the results reported here are in agreement with previous reports, where the coassociation of ␣6 with ␥2 (Khan et al., 1994;Quirk et al., 1994), ␦ (Quirk et al., 1994), and ␤2/3 (Khan et al., 1994) was found. But Quirk et al. (1994) identified ␣6␥2 and ␣6␦ as two distinct populations, where the latter did not bind [ 3 H]Ro 15-4513. In the ␣1␣6 double immunoaffinity-purified receptors, we were unable to detect the GABA A receptor ␥2 subunit by immunoblotting. Negative results here are difficult to interpret definitively because they may be explained by both the low levels of purified receptor and antibody avidity, a particular problem with the anti-␥2 subunit antibody used (cf. Stephenson et al., 1990;Duggan et al., 1992). But, it may also be that the ␥2 subunit is associated with single ␣6 variant receptors. Consequently, the (␣1␣6␤2/3␦) receptor identified here may be similar to (␣6␦) receptors described by Quirk et al. (1994), which do not bind [ 3 H]Ro 15-4513. Further analysis of the anti-␣1(413-429) post-column filtrate should clarify this.
The direct determination of the number of ␣ subunits (a 1:1 ratio for ␣1:␣6, therefore predicting two per receptor) is the first for native GABA A receptors. It is in agreement with the 1:1 ratio predictions for native receptors where the coexistence of two but not three different ␣ subunits were detectable , the inferred subunit complement of native cerebellar receptors, ␣1␣6␤2/3␥2 L ␥2 S (Khan et al., 1994), and the subunit complement of an (␣1) 2 (␤1) 1 (␥2) 2 cloned receptor (Backus et al., 1993). The quantification described in this paper is not ideal because antibody molecules are bivalent and the antibodies used are polyclonal albeit to a restricted epitope. Thus it is a possibility that the number of antibodies bound per subunit may not be stoichiometric. However, this is unlikely because 1) steric hindrance would reduce the probability of two antibody molecules binding to different epitopes within the restricted 16-amino acid peptide sequence, and 2) the binding of one antibody molecule to two subunits would have an equal probability for the ␣1 and ␣6 subunits following reduction and denaturation in SDS-PAGE.
The study of cloned, single ␣ GABA A receptors showed that the benzodiazepine subpharmacology was dependent on the type of ␣ subunit (Luddens and Wisden, 1991). The ␣6 subunitcontaining cloned receptors show BZ-IS pharmacology in contrast to ␣1 receptors, which are BZ-S . A single point mutation in the ␣6 subunit, ␣6R100H, results in a mutant receptor with high affinity for the classical benzodiazepines (Wieland et al., 1992). From the results reported herein, for (␣1␣6) receptors, the ␣6 subunit dominates the pharmacology yielding BZ-IS pharmacology. This would be in agreement with Mathews et al. (1994).
Conclusions-In conclusion, we have demonstrated a direct correlation between the BZ-IS, [ 3 H]Ro 15-4513 binding site pharmacology and native, ␣6 subunit-containing GABA A receptors. Furthermore, quantitative results showed that 41% of all the ␣6 subunit-containing receptors coexist with an ␣1 subunit and that in these receptors, the ␣6:␣1 ratio was 1:1 and they displayed BZ-IS pharmacology. The functional significance of the extensive GABA A receptor heterogeneity remains to be solved. However, the fact that we find all combinations of association of these two polypeptides in percentages that preclude the random association of actively transcribed genes suggests that there may be an active sorting/assembly mechanism to form functional receptor subtypes.