Coexistence of Two β Subunit Isoforms in the Same γ-Aminobutyric Acid Type A Receptor*

Three novel subunit-specific antisera to the β1, β2, and β3 subunits of rat γ-aminobutyric acid type A (GABAA) receptors have been used to study the native receptor in the rat brain. Affinity-purified anti-β1, anti-β2, and anti-β3 antibodies recognized in immunoblots protein bands of 57, 55, and 57 kDa, respectively. Quantitative immunoprecipitation of solubilized GABAA receptors from various rat brain regions showed that the β2 subunit was the most abundant isoform in cerebellum (in 96% of the GABAA receptors) and cerebral cortex (64%) but that it was the least abundant isoform in hippocampus (44%). The β3subunit was found most abundant in hippocampus (64%) followed by cerebral cortex (48%) and cerebellum (33%). The β1subunit was present in a very small proportion of the cerebellar GABAA receptors (3%), but it was present in a high proportion of the GABAA receptors from the hippocampus (49%) and cerebral cortex (32%). Quantitative receptor immunoprecipitation or immunopurification followed by immunoblotting experiments have revealed the existence of colocalization of two different β subunit isoforms in a significant proportion of the brain GABAA receptors. Thus, in the rat cerebral cortex 33% of the GABAA receptors have both β2 and β3 subunits, and 19% of the receptors have both β1 and β3 subunits. The extent of colocalization of β subunit isoforms varied among brain regions, being highest in hippocampus and lowest in cerebellum. These and other results taken together suggest that the number of α, β, and γ subunits (stoichiometry) in the brain GABAA receptor pentamers might not be unique. It might vary depending on receptor type.

The brain GABA A 1 receptors are heteropentameric (1) membrane proteins that form Cl Ϫ channels. To date, six ␣ subunit isoforms (␣ 1 -␣ 6 ), three ␤ (␤ 1 -␤ 3 ), three ␥ (␥ 1 -␥ 3 ), and one ␦ have been cloned from mammalian brain (2). The brain GABA A receptors display high heterogeneity due to the existence of many different pentameric combinations of the 13 subunits (reviewed in Refs. [2][3][4][5][6]. Nonetheless, the complete subunit composition and stoichiometry of most of the native GABA A receptors in various brain regions and cell types is still not known. Studies from heterologous expression systems have provided crucial information about the subunits that are re-quired for the expression of fully functional GABA A receptors and for understanding the relationship between subunit composition and radioligand binding specificity (7)(8)(9)(10). A limitation of studying reconstituted receptors is that it is not known how well they represent the native receptors expressed in the brain. We and others have used subunit-specific antibodies to elucidate the subunit composition of the native GABA A receptors that are expressed in the brain (3,5,6). These studies have shown that (i) the ␣ 1 , ␤ 2/3 , and ␥ 2 subunits coexist in many brain GABA A receptors (11)(12)(13)(14); (ii) two ␣ subunit isoforms can coexist in the same GABA A receptor (15)(16)(17)(18)(19); and (iii) two ␥ subunit isoforms can coexist in the same GABA A receptor (13,20). In contrast to the ␣ and ␥ subunits, little is known about the possible colocalization of two ␤ subunit isoforms in the same brain GABA A receptor due to the difficulty in raising suitable ␤ subunit-specific antibodies. We have previously obtained several antibodies to the ␤ subunits (21-23), but they recognized more than one ␤ subunit isoform. One of them was the mouse monoclonal antibody (mAb) 62-3G1 to affinity-purified GABA A receptors, which recognized both ␤ 2 and ␤ 3 subunits (21). In the present paper, we report the development of novel ␤ subunit isoform-specific antisera (to ␤ 1 , ␤ 2 , or ␤ 3 ) and the use of these antibodies to demonstrate the existence of colocolization of ␤ subunit isoforms in the same brain GABA A receptor. To the best of our knowledge, this is the first evidence suggesting the coexistence of two different ␤ subunit isoforms (␤ 1 and ␤ 3 ; ␤ 2 and ␤ 3 ) in the same brain GABA A receptor. 3  Preparation of Subunit-specific Antibodies-The peptides PLSSREG-FGRGLC (corresponding to amino acid residues 382-393 of rat ␤ 1 ), AGLPRHSFGRNAC (corresponding to residues 382-393 of rat ␤ 2 ), and QSMPKEGHGRYMC (corresponding to residues 380 -391 of rat ␤ 3 ) are located in the large intracellular loop of the corresponding ␤ subunit. This loop is located between the putative transmembrane domains M 3 and M 4 (24). The peptides were synthesized with an additional cysteine at the carboxyl end through which they were coupled to keyhole limpet hemocyanin. New Zealand rabbits were subcutaneously immunized with 100 g of keyhole limpet hemocyanin-coupled peptides in complete Freund's adjuvant. The following immunizations were done in incomplete Freund's adjuvant at 4-week intervals. The antisera used in this study were collected 12 days after the fifth immunization. The antibodies were affinity-purified on the corresponding immobilized peptide as described previously (13).
Immunopurification of ␤ Subunit Containing GABA A Receptors-Each anti-␤ subunit antibody was irreversibly coupled through the Fc moiety to Protein A-agarose beads with dimethylpimelimidate, according to the procedure of Schneider et al. (25). The beads were washed extensively and sequentially with (i) phosphate-buffered saline, (ii) elution buffer (100 mM glycine, pH 2.0, 0.05% Triton X-100), and (iii) modified radioimmune precipitation buffer. The GABA A receptors from rat cerebral cortex membranes were solubilized in modified radioimmune precipitation buffer and were bound onto the immobilized anti-␤ subunit antibody (bound to the agarose beads) by overnight incubation at 4°C. The beads were extensively washed with modified radioimmune precipitation buffer followed by 10 mM sodium phosphate, pH 6.8, 0.05% Triton X-100. The GABA A receptors were recovered after incubation with elution buffer followed by immediate neutralization with Tris base. The wash and elution buffers always included the mixture of protease inhibitors.
Immunoblots and Other Procedures-The immunoblots were done according to De Blas and Cherwinski (26). Affinity-purified GABA A / BZD receptors from bovine brain were prepared by BZD affinity chromatography on immobilized Ro 7-1986/1 as described elsewhere (21). The ELISA was performed as described by De Blas et al. (27).

RESULTS
Antisera to ␤ 1 , ␤ 2 , and ␤ 3 Subunits-Novel subunit-specific rabbit antisera to each of the three ␤ subunit isoforms have been prepared. The development of the immune response was monitored by both ELISA with immobilized peptides and immunoprecipitation of [ 3 H]FNZ binding activity of the solubilized GABA A receptors from rat cerebral cortex. Antiserum specificity for the corresponding peptide was shown by ELISA. No cross-reactivity was observed with the other two peptides, and blockage of each of the ELISA reactions (antiserum diluted 1:1,000) was only obtained with the corresponding peptide (i.e. 10, 20, and 50 g/ml for ␤ 1 , ␤ 2 , and ␤ 3 , respectively). Each antiserum immunoprecipitated solubilized native GABA A receptors from rat cerebral cortex (Fig. 1). The anti-␤ 1 , anti-␤ 2 , and anti-␤ 3 antisera immunoprecipitated a maximum of 28, 58, and 34%, respectively, of [ 3 H]FNZ binding from cerebral cortex membranes with a single immunoprecipitation. Maximum immunoprecipitation was obtained with 12, 15, and 20 l of antiserum, respectively. Receptor immunoprecipitation with each antiserum was specifically blocked by the corresponding peptide (i.e. 5, 5, and 25 g/ml for ␤ 1 , ␤ 2 , and ␤ 3 , respectively) but not by the other two peptides (not shown).
Immunoblots with Rat Brain Cerebral Cortex Membranes and Affinity-purified Receptors from Bovine Brain-The antibodies were affinity-purified on the corresponding immobilized peptide and used for immunoblotting. Fig. 2A shows that anti-␤ 2 and anti-␤ 3 recognized 55-and 57-kDa peptides, respectively, in crude membranes from rat cerebral cortex. These results are consistent with the immunoblot results obtained with our mAb 62-3G1 to ␤ 2/3 ( Fig. 2A, lane 5). The immunoreactivity was specifically blocked by the corresponding peptide ( Fig. 2A, lanes 2 and 4), but not by the other two peptides (not shown). Protein bands of similar mobilities were also revealed by anti-␤ 2 and ␤ 3 antibodies as well as with mAb 62-3G1 in immunoblots of affinity-purified GABA A receptors from bovine cerebral cortex (Fig. 2B). The protein bands obtained with immunoblots of purified receptors were wider and more diffuse than the ones obtained with membrane immunoblots. This feature has also been observed in previous studies with other antibodies (12, 13). We do not know whether the result reflects the partial degradation of the GABA A receptors during receptor purification despite using a mixture of protease inhibitors. The anti-␤ 1 antibody revealed in immunoblots a faint 57-kDa and a fainter 53-kDa protein band, which are difficult to show by photography.
Immunoprecipitation of GABA A /BZD Receptors from Various Brain Regions-We and others have previously shown that two or three sequential immunoprecipitations with the same antibody are required for reaching the maximum (quantitative) immunoprecipitation of detergent-solubilized GABA A receptors from crude membrane preparations (12,13,18,19,22,23,28). Quantitative immunoprecipitations of solubilized GABA A /BZD receptors, as measured by [ 3 H]muscimol and [ 3 H]FNZ bindings, revealed that the relative abundance of ␤ 1 , ␤ 2 , and ␤ 3 subunits in the GABA A receptors varies among different brain regions (Table I) Nevertheless, the results were not identical, probably because of the well known observation that not all GABA A receptors have BZD (FNZ) binding sites. The immunoprecipitation of affinity-purified GABA A /BZD receptors from bovine cerebral cortex (data not shown) also indicated that the relative abundance of the three ␤ subunits was similar to that of the solubilized receptors from rat cerebral cortex (␤ 2 Ͼ ␤ 3 Ͼ ␤ 1 ).
Immunoprecipitations of the solubilized receptors were also done with a mixture of the three anti-␤ subunit antisera. The immunoprecipitation values for both [ 3 H]FNZ and [ 3 H]muscimol binding were 91-98% (not shown) in the three regions of the rat brain, suggesting that most of the brain GABA A recep-tors have at least one ␤ subunit.
Subunit Colocalization-In each of the studied brain regions, the additive value of the immunoprecipitations obtained with each of the three anti-␤ isoform antibodies was larger than 100% (Table I). Thus, in cerebral cortex the values were 144 and 149% for [ 3 H]muscimol and [ 3 H]FNZ, respectively; in hippocampus the corresponding values were 157 and 162%, and in the cerebellum they were 132 and 109%, respectively. These results suggest that some GABA A receptors have two different ␤ subunit isoforms colocalizing in the same receptor. The colocalization of ␤ subunit pairs was further quantified in the rat cerebral cortex with the procedure described elsewhere for studying the colocalization of ␣ 1 and ␣ 6 (13). Table II shows that when mixing two different anti-isoform antisera (␤ 1 ϩ ␤ 2 or ␤ 1 ϩ ␤ 3 or ␤ 2 ϩ ␤ 3 ), the immunoprecipitation value obtained with the antibody mixture (A column) was smaller than the value obtained by adding the two immunoprecipitation values obtained with the two antisera separately (B column). The difference (B Ϫ A) indicates the extent of subunit colocalization. Thus, in the rat cerebral cortex 33% of the GABA A receptors contain both ␤ 2 and ␤ 3 subunit isoforms, whereas 19% contain ␤ 1 and ␤ 3 . In contrast, there is little colocalization of ␤ 1 and ␤ 2 (8%) in the cerebral cortex GABA A receptors.
To further corroborate the existence of colocalization of two ␤ subunit isoforms in the same GABA A receptor, solubilized receptors from rat cerebral cortex were immunopurified with each of the anti-␤ subunit antibodies followed by immunoblotting with the other two ␤ subunit antibodies. Fig. 3A shows the presence of both the 57-kDa ␤ 1 (lane 1) and the 57-kDa ␤ 3 (lane 5) subunits in the GABA A receptors that have been immunopurified on immobilized anti-␤ 1 . However, no ␤ 2 subunit (lane 3) was detected in these receptors. This finding agrees with the results of Table II obtained with a different assay, which also show little colocalization (8%) of ␤ 1 and ␤ 2 subunits in cerebral cortex (Table II).  from various regions of the rat brain Data are cumulative values after three sequential immunoprecipitations with each antibody. The receptors were solubilized from the various regions and incubated with 15 l of anti-␤ 1 , 20 l of anti-␤ 2 , or 15 l of anti-␤ 3 . Then the supernatant of the first immunoprecipitation was incubated with the same volume of the same antibody and immunoprecipitated under the same conditions. The procedure was repeated a third time with the supernatant of the second immunoprecipitation. The results are the percentage of binding activity in the pellets with respect to all of the solubilized binding activity (supernatant ϩ pellet). Data  sera (not shown). The 57-kDa ␤ 3 peptide band, but not the 55-kDa ␤ 2 peptide, was revealed with mAb 62-3G1, which recognizes both ␤ 2 and ␤ 3 subunits (lane 7), also supporting the colocalization of ␤ 3 (but not ␤ 2 ) with ␤ 1 . Fig. 3B shows the presence of both the 55-kDa ␤ 2 (lane 3) and the 57-kDa ␤ 3 subunit (lane 5) in the receptors that have been immunopurified on immobilized anti-␤ 2 . No protein band was revealed with anti-␤ 1 antibody (lane 1), indicating once more the low extent of colocalization of ␤ 1 and ␤ 2 . The immunoreactivities for anti-␤ 2 and anti-␤ 3 could be specifically blocked by the corresponding synthetic peptide (lanes 4 and 6, respectively). Two bands corresponding to 55 kDa (␤ 2 ) and 57 kDa (␤ 3 ) were revealed with mAb 62-3G1 (lane 7), as expected with this mAb that recognizes both ␤ 2 and ␤ 3 . Fig. 3C shows immunoblots of GABA A receptors immunopurified on immobilized anti-␤ 3 . These receptors had the 57-kDa ␤ 1 (lane 1), the 55-kDa ␤ 2 (lane 3), and the 57-kDa ␤ 3 (lane 5), as revealed by the corresponding anti-␤ 1 , anti-␤ 2 , and anti-␤ 3 antibodies. The presence of ␤ 3 was also revealed with the mAb 62-3G1 (lane 7). No ␤ 2 was revealed with this mAb because (i) the mAb 62-3G1 reacts less with ␤ 2 than ␤ 3 in immunoblots and (ii) there is considerably less ␤ 2 than ␤ 3 in the GABA A receptors immunopurified on anti-␤ 3 . The immunoreactivity of each antibody with the respective subunit was specifically blocked by the corresponding synthetic peptide (Fig. 3C, lanes 2, 4, and 6) but not by the other subunit peptides (not shown). Therefore, the results shown in Fig. 3 and Table II strongly argue in favor of the coexistence of either ␤ 1 and ␤ 3 or ␤ 2 and ␤ 3 subunits (but not ␤ 1 and ␤ 2 ) in the same GABA A receptor pentamers.

DISCUSSION
Specificity of the Anti-␤ 1 , Anti-␤ 2 , and Anti-␤ 3 Antisera-These ␤ subunit isoform-specific antibodies immunoprecipitated native solubilized receptors, and in immunoblots they reacted with the corresponding GABA A receptor subunit. The specificity of these antibodies has been demonstrated by several lines of evidence. (i) The selected peptide sequences are highly specific for the corresponding ␤ 1 , ␤ 2 , or ␤ 3 subunit as computer analysis indicates. In addition, no homologous peptide sequences were found in other GABA A receptor subunits.
(ii) No cross-reactivity between each antibody and the peptides of the other subunits was found in ELISA. In addition, each antibody-peptide reaction was dose-dependently blocked by the corresponding peptide but not by the others. (iii) Receptor immunoprecipitation was specifically inhibited by the corresponding peptide but not by the others. (iv) Immunoprecipitations of GABA A receptors from various brain regions showed that the relative abundance of the ␤ subunit peptides correlated with the relative abundance of the corresponding mRNAs (29 -32). (v) In immunoblots, each isoform-specific antibody recognized a protein band (or two) of the expected size for that particular ␤ subunit (14,(21)(22)(23)(33)(34)(35). In addition, the immunoblot reaction could be specifically blocked by the corresponding peptide but not by the others.
The ␤ Subunit Isoforms in the GABA A Receptors from Different Brain Regions-The specificity of these antibodies and their ability to recognize the native receptor enabled us to quantify by immunoprecipitation the prevalence of the ␤ 1 , ␤ 2 , and ␤ 3 subunit isoforms among the GABA A receptors from various brain regions. The ␤ 2 subunit was the most abundant of the isoforms in cerebellar GABA A receptors (84 -96%) and in cerebral cortex (64 -66%), but it was the least abundant isoform in hippocampus (41-44%). The ␤ 3 subunit was found most abundant in hippocampal receptors (62-64%), followed by cerebral cortex (48 -52%) and cerebellum (23-33%). The ␤ 1 subunit was present in a very small proportion of the cerebellar GABA A receptors (2-3%) but was present in a high proportion of hippocampal receptors (49 -59%) and to a lesser extent in cerebral cortex (31-32%). Based upon these findings, it seems that the relative protein expression of the ␤ subunit isoforms and assembly into GABA A receptors varies among different brain regions. The functional significance for this distinct regional pattern of expression remains unknown. It has been proposed that the main site(s) for muscimol binding is on the ␤ subunit, although the ␣ subunit also contributes to it (36,37). Recombinant studies have shown that the ␤ subunit may affect the affinity of the GABA A receptors for GABA and other agonists (38), although it does not significantly affect the affinity or efficacy of other ligands acting at the BZD, barbiturate, or steroid binding sites (8,39). However, it has been reported that the anticonvulsant drug loreclezole was highly selective for receptors containing ␤ 2 or ␤ 3 but not ␤ 1 (40).
Colocalization of Two ␤ Subunit Isoforms in the Same    H]muscimol binding immunoprecipitated with anti-␤ 1 , anti-␤ 2 , and anti-␤ 3 antisera were higher than 100%, particularly in hippocampus and cerebral cortex, as shown in Table I. These results suggest that a substantial proportion of GABA A receptor in the brain contains at least two ␤ subunit isoforms and that the extent of ␤ subunit colocalization varies among brain regions, being highest in hippocampus (57-62%) and lowest in cerebellum (9 -32%), whereas in cerebral cortex it is 44 -49%. Moreover, when mixing two antisera such as anti-␤ 1 and anti-␤ 3 , or anti-␤ 2 and anti-␤ 3 , the immunoprecipitation value was significantly smaller than the additive value obtained after immunoprecipitating with each antiserum. We have calculated that in the cerebral cortex ␤ 2 and ␤ 3 colocalized in 33% of the GABA A receptors, whereas ␤ 1 and ␤ 3 colocalized in 19% of the receptors. The colocalization of ␤ 1 and ␤ 2 was less common (8%). Finally, immunopurification with a subunit-specific antibody followed by immunoblotting with the other anti-␤ subunit antibodies (Fig. 3) confirmed the colocalization of ␤ 1 and ␤ 3 isoforms on one hand and ␤ 2 and ␤ 3 on the other in some GABA A receptors. The immunoblots also confirmed the low extent of colocalization of ␤ 1 and ␤ 2 in cerebral cortex. Table I also shows that in cerebellum there is little or no colocalization of ␤ 1 with either ␤ 2 or ␤ 3 .
Our main conclusions showing colocalization of ␤ 1 and ␤ 3 or ␤ 2 and ␤ 3 differ from those of Benke et al. (14). They reported that the additive value of GABA A receptor immunoprecipitations obtained with their anti-␤ 1 , anti-␤ 2 , and anti-␤ 3 antibodies was close to 100%. They concluded that GABA A receptors largely contain only a single type of ␤ subunit isoform. Our anti-␤ 2 and their anti-␤ 2 antibodies precipitated a similar percentage of [ 3 H]FNZ and [ 3 H]flumazenil binding activity, respectively, from solubilized rat brain membranes (66 versus 57%, respectively). However, we got significantly higher immunoprecipitation values with our anti-␤ 1 (31 versus 8%) and anti-␤ 3 (52 versus 14%) antibodies. Several factors may account for the difference between the two groups: (i) different brain regions were used (our cerebral cortex versus their whole brain); (ii) we did three sequential immunoprecipitations with the same antibody, but they did a single one. We and others have shown that one immunoprecipitation does not precipitate 100% of the GABA A receptors (12,13,18,19,22,23,28). The affinity of their ␤ 1 and ␤ 3 antibodies for the native solubilized receptors may be lower than ours. It is worth to point out that the immunoprecipitation of [ 3 H]FNZ binding in cerebral cortex obtained with our anti-␤ 3 antibody (52%, Table I) was similar to the values (49 and 50%) obtained by two other groups (Refs. 33 and 41, respectively) using their respective anti-␤ 3 antisera to peptides different than ours. One of these groups found that their anti-␤ 3 immunoprecipitated 17% of [ 3 H]FNZ binding in cerebellum (42), which is also close to the 23% value that we are reporting for cerebellum in Table I. The absence of cross-reactivity of our anti-␤ 1 in the immunoprecipitation assays was confirmed not only by the specific blockage with the ␤ 1 peptide, as indicated above, but also by the little immunoprecipitation of [ 3 H]FNZ binding (2%) obtained in cerebellum. However, the same anti-␤ 1 antibody precipitated 31 and 59% of the GABA A receptors in cerebral cortex and hippocampus, respectively. Moreover, anti-␤ 2 and anti-␤ 3 immunoprecipitated 84 and 23% of [ 3 H]FNZ binding in cerebellum. Benke et al. (14) did not test for colocalization of the ␤ 1 -␤ 3 , or ␤ 2 -␤ 3 pairs by immunopurification and immunoblotting as we did. They only tested for the presence of ␤ 1 in the receptors immunopurified on anti-␤ 2 , and they could not detect colocalization of ␤ 1 and ␤ 2 . In agreement with their results, we could find no evidence for significant colocalization of ␤ 1 and ␤ 2 in our immunoblotting experiments. However, we found colocalization of ␤ 1 -␤ 3 and ␤ 2 -␤ 3 .
Our conclusions on the existence of colocalization of two different ␤ subunit isoforms in native brain receptor are supported by the studies of Chang et al. (43) on recombinant receptors in oocytes injected with ␣ 1 , ␤ 2 , and ␥ 2 subunit mRNAs. It seems that these subunits form pentameric assemblies composed mostly of two ␣ subunits, two ␤ subunits, and one ␥ subunit (43). In addition, when tandem constructs of ␣ 6 and ␤ 2 were cotransfected with ␥ 2 (44) the pentameric assemblies had two ␣ subunits, two ␤ subunits, and one ␥ subunit. However, this is not the only stoichiometry observed in all recombinant receptors or native receptors. Thus, it has been shown (45) that host cells cotransfected with ␣ 3 , ␤ 2 , and ␥ 2 mainly form pentameric receptors composed of two ␣, one ␤, and two ␥. Functional pentameric combinations of two ␣, two ␤, and one ␥ as well as one ␣, two ␤, and two ␥ subunits were also formed although in smaller quantities. Moreover, we have also reported that in the cerebellum some native GABA A receptors have two ␣ subunits, one ␤ subunit, and two ␥ subunits with the following pentameric composition: ␣ 1 ␣ 6 ␥ 2S ␥ 2L ␤ 2/3 (18). Thus, all of the results taken together suggest that the number of ␣, ␤, and ␥ subunits forming various pentamers varies depending on the subunit composition and/or the brain region or cell type that express them, which contributes to the heterogeneity of brain GABA A receptors.
In summary, novel and specific antibodies to ␤ 1 , ␤ 2 , and ␤ 3 subunits and quantitative GABA A receptor immunoprecipitation assays have allowed us to reveal the relative contribution of ␤ 1 , ␤ 2 , and ␤ 3 subunit isoforms to the GABA A receptors of various brain regions. Moreover, these results together with experiments on receptor immunopurification and immunoblotting with the ␤ subunit antibodies have revealed for the first time the existence in the brain of colocalization of two different ␤ subunit isoforms in a substantial proportion of the brain GABA A .