Identification of a GABAB Receptor Subunit, gb2, Required for Functional GABAB Receptor Activity*

G protein-coupled receptors are commonly thought to bind their cognate ligands and elicit functional responses primarily as monomeric receptors. In studying the recombinant γ-aminobutyric acid, type B (GABAB) receptor (gb1a) and a GABAB-like orphan receptor (gb2), we observed that both receptors are functionally inactive when expressed individually in multiple heterologous systems. Characterization of the tissue distribution of each of the receptors by in situhybridization histochemistry in rat brain revealed co-localization of gb1 and gb2 transcripts in many brain regions, suggesting the hypothesis that gb1 and gb2 may interact in vivo. In three established functional systems (inwardly rectifying K+channel currents in Xenopus oocytes, melanophore pigment aggregation, and direct cAMP measurements in HEK-293 cells), GABA mediated a functional response in cells coexpressing gb1a and gb2 but not in cells expressing either receptor individually. This GABA activity could be blocked with the GABAB receptor antagonist CGP71872. In COS-7 cells coexpressing gb1a and gb2 receptors, co-immunoprecipitation of gb1a and gb2 receptors was demonstrated, indicating that gb1a and gb2 act as subunits in the formation of a functional GABAB receptor.

G protein-coupled receptors are commonly thought to bind their cognate ligands and elicit functional responses primarily as monomeric receptors. In studying the recombinant ␥-aminobutyric acid, type B (GABA B ) receptor (gb1a) and a GABA B -like orphan receptor (gb2), we observed that both receptors are functionally inactive when expressed individually in multiple heterologous systems. Characterization of the tissue distribution of each of the receptors by in situ hybridization histochemistry in rat brain revealed co-localization of gb1 and gb2 transcripts in many brain regions, suggesting the hypothesis that gb1 and gb2 may interact in vivo. In three established functional systems (inwardly rectifying K ؉ channel currents in Xenopus oocytes, melanophore pigment aggregation, and direct cAMP measurements in HEK-293 cells), GABA mediated a functional response in cells coexpressing gb1a and gb2 but not in cells expressing either receptor individually. This GABA activity could be blocked with the GABA B receptor antagonist CGP71872. In COS-7 cells coexpressing gb1a and gb2 receptors, co-immunoprecipitation of gb1a and gb2 receptors was demonstrated, indicating that gb1a and gb2 act as subunits in the formation of a functional GABA B receptor.
Metabotropic GABA B 1 receptors were first distinguished pharmacologically by Hill and Bowery (1). Kaupmann et al. (2) recently cloned two alternatively spliced forms of the GABA B receptor, termed gb1a and gb1b, which belong to the G proteincoupled receptor superfamily and are most closely related to the metabotropic glutamate receptors. Although native GABA B receptors are reported to activate inwardly rectifying K ϩ channels (Kir) (3), recombinant gb1a receptors coexpressed with Kir channels in Xenopus oocytes failed to be functionally active (2,4,12). A recent report has shown that recombinant GABA B receptors fail to couple to effector pathways in a variety of non-neuronal and neuronal cell types, suggesting that additional cellular component(s) are required (4). Failure to show GABA B receptor function coupled with previous reports that GPCRs can undergo dimerization (5)(6)(7)(8), suggested that heterodimerization may be important for GABA B receptor function. Furthermore, receptor heterodimerization appears to rescue function of mutated or chimeric muscarinic and adrenergic receptors (9). We report here that coexpression of gb1 and gb2 receptors are necessary for the formation of functional GABA B receptors and result in heterodimerization.

EXPERIMENTAL PROCEDURES
Expression Constructs for gb1a and gb2 Receptors-The murine gb1a (mgb1a) cDNA was constructed from two expressed sequence tags (IMAGE Consortium clone identification numbers 472408 and 319196), combined by standard PCR methods, and subcloned into the pCINeo (Stratagene) and pcDNA3.1 (Invitrogen) vectors. 2 The rat gb1a receptor was obtained by PCR of rat brain cDNA using oligonucleotide primers based on the published sequence (Ref. 2; GenBank TM accession number Y10369) and subcloned into pcDNA3.1.
Two independently derived cDNAs of human gb2 (GenBank TM accession numbers AF069755 and AF056085, the former having previously been called GPR51 (10)) were used to make expression constructs. A N-terminal FLAG-tagged hgb2/pcDNA3.1 construct encoding a modified influenza hemaglutinin signal sequence (MKTIIALSYIFCLVFA) followed by an antigenic FLAG (DYKDDDDK) epitope was generated by PCR (10). The hgb2 construct used for expression in 293 cells was the coding sequence from GenBank TM accession number AF056085 with the C-terminal splice variant of GenBank TM accession number AF095723 3 inserted into pcDNA3.1 in a manner similar to that for the rat gb1a.
In Situ Hybridization Histochemistry-Adjacent coronal rat brain sections were hybridized with labeled antisense and sense riboprobes directed against rgb2 (GenBank TM accession number AF058795) or rgb1 as described previously (11). 4 Probes were generated by amplification of rgb2 with JC216 (T3 promoter followed by bases 1172-1191) and JC217 (T7 promoter and the complement of bases 1609 -1626) or with JC218 (T3 promoter and bases 2386 -2405) and JC219 (T7 promoter and the complement bases 2776 -2793) or by amplification of rgb1a with JC160 (T3 promoter and bases 631-648) and JC161 (T7 promoter and the complement of bases 1024 -1041). For colocalization experiments, probes were either labeled with digoxigenin (rgb1a) or 35 S (rgb2). Detection of the radiolabeled rgb2 probe was performed using emulsion after detection of the digoxigenin-labeled rgb1 probe on the same brain slices.
Melanophore Functional Assay-Growth of Xenopus laevis melanophores and fibroblasts and DNA transfections by electroporation were * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
b These authors contributed equally to this work. c To whom correspondence should be addressed. Tel.: 514-428-3589; Fax: 514-428-4900; E-mail: gordon_ng@merck.com. e Supported by a Young Investigator Award from the National Alliance for Research on Schizophrenia and Depression.
performed as described previously (13). To monitor the efficiency of transfection an internal control GPCR was used (pcDNA1-cannabinoid 2). For G i -coupled responses (pigment aggregation), cells were preincubated in the presence of 100 l/well of 70% L-15 medium containing 2.5% fibroblast-conditioned growth medium, 2 mM glutamine, 100 g/ml streptomycin, 100 units/ml penicillin, and 15 mM HEPES, pH 7.3, for 30 min to induce pigment dispersion. Absorbance readings at 600 nm were measured using a Bio-Tek Elx800 Microplate reader (ESBE Scientific) before and after 1.5 h incubation with the ligands.
Stable and Transient Transfections and Determination of Cellular cAMP Response in HEK-293 Cells-Hgb2 and rgb1a cDNAs in pcDNA3.1 were used to transfect HEK-293 cells. Stably expressing cells were identified after selection in geneticin (0.375 mg/ml) by dot blot analysis. For coexpression experiments the stable cell lines hgb2-42 and rgb1a-50 were transiently transfected with hgb2 or rgb1a, and cells were assayed for cAMP responses. Wild-type HEK-293 cells or HEK-293 cells stably and transiently expressing the hgb2 and rgb1a receptors were lifted in 1ϫ phosphate-buffered saline, 2.5 mM EDTA, counted, pelleted, and resuspended at 1.5 ϫ 10 5 cells/100 l in Krebs-Ringer-HEPES medium (14), 100 M Ro 20-1724 (RBI) and assayed for agonist-mediated inhibition of forskolin-stimulated cAMP synthesis using a modified version of an assay previously described by Kaupmann et al. (2). cAMP determinations were made using a solid phase modification (15) of the cAMP radioimmunoassay described by Brooker et al. (16) and previously reported (17).
Kir Channel Activity-Xenopus oocytes were isolated as described (18). cDNA constructs for human Kir 3.1, Kir 3.2 channel isoforms (gifts from Dr. Hubert Van Tol, University of Toronto), G i ␣1 (a generous gift of Dr. Maureen Linder, Washington University), and cDNAs of mgb1a or FLAG-hgb2 (subcloned into pT7TS, a gift of Dr. Paul Krieg, University of Texas) were linearized and transcribed using T7 RNA polymerase and the mMessage mMachine (Ambion). Individual oocytes were injected with 5-10 ng (in 25-50 nl) of Kir3.1 and Kir3.2 constructs with mRNAs for mgb1a or FLAG-hgb2 receptors with and without G i ␣1. Recordings were made after 48 h as described (18). Standard recording solution was KD-98, 98 mM KCl, 1 mM MgCl 2 , 5 mM K-HEPES, pH 7.5, unless otherwise stated. Data collection and analysis were performed using pCLAMP v6.0 (Axon Instruments) and Origin v4.0 (MicroCal) software. For subtraction of endogenous and leak currents, records were obtained in ND-96, 96 mM NaCl, 2 mM KCl, 1 mM MgCl 2 , 5 mM Na-HEPES, and these were subtracted from recordings in KD-98 before further analysis.
Immunoprecipitation and Immunoblotting of GABA B Receptors-COS-7 cells (ATCC) were cultured and transiently transfected with FLAG-hgb2 and mgb1a receptor DNAs alone and in combination using LipofectAMINE reagent (Life Technologies, Inc.) according to the manufacturer's instructions. Membranes prepared from these cells were digitonin-solubilized, and mgb1a/FLAG-gb2 heterodimers were immunoprecipitated with either a mouse anti-FLAG M2 antibody (Kodak IBI) targeting the FLAG-gb2 receptor or anti-gb1 receptor rabbit polyclonal antibodies 1713.1-1713.2 2 using previously described conditions (6). The immunoprecipitates were then washed and subjected to SDSpolyacrylamide gel electrophoresis and immunoblotted with either the anti-FLAG or anti-gb1 antibodies using previously reported conditions (6) and as described below.

RESULTS AND DISCUSSION
Tissue Distribution of gb1 and gb2 Receptor mRNAs-Northern blot analysis revealed that many brain regions, including cortex, show overlapping expression patterns for gb2  and gb1 receptor mRNA (10). In situ hybridization in adjacent coronal sections of rat parietal cortex using 35 S-labeled antisense riboprobes indicates that mRNAs for both gb1 and gb2 receptors are coexpressed in this brain region (Fig. 1, A and B). No hybridization signal was detected with the control 35 Slabeled sense gb1 and gb2 riboprobes (data not shown). To examine expression at the cellular level, digoxigenin-labeled gb1 and radiolabeled gb2 riboprobes were hybridized to the same sections (Fig. 1, C and D). Overlay of the gb1 and gb2 hybridization signals revealed that messages for both receptors are generally expressed in the same neurons (Fig. 1E). In other major brain regions, including the hippocampus, thalamus, and cerebellum, 3 gb2 and gb1 receptor mRNAs are colocalized with Ն95% of gb2 expressing cells also expressing gb1. This suggests that their coexpression may be required for activity in these neurons.
Recombinant gb1a Receptors Require Coexpression of gb2 for Functional Activity-In melanophores transiently cotransfected with the mgb1a and hgb2 receptors, GABA mediated a dose-dependent pigment aggregation response with an IC 50 value of 3-7 M (n ϭ 3), which is absent in mock-transfected cells and cells transfected with the mgb1a or hgb2 alone (Fig.  2). The GABA-mediated inhibitory activity represented 42-56% (n ϭ 3) of a control G i -coupled CB2 cannabinoid receptor response (Fig. 2, inset). GABA activity could be inhibited by the CGP71872 antagonist (n ϭ 3), indicating that this was GABA B receptor-specific (Fig. 2).
To confirm the observations that functional GABA B receptor results from the coexpression of gb1a and gb2 receptors and that it is G i -coupled, we directly examined modulation of cAMP levels in HEK-293 cells. In cell lines stably expressing the individual receptors we observed small and inconsistent responses in assays to examine agonist-mediated modulation of cAMP synthesis (Fig. 3). However, transient hgb2 transfection of HEK-293 cells stably expressing rgb1a (rgb1a-50) and transient rgb1a transfection of HEK-293 cells stably expressing hgb2 (hgb2-42) significantly enhanced the ability of baclofen and GABA to inhibit forskolin-stimulated cAMP synthesis. Hgb2-42/rgb1a and rgb1a-50/hgb2 cells exhibited 28 and 34% reductions in forskolin-stimulated cAMP synthesis with 30 M baclofen and 40 and 43% decreases with 30 M GABA, respectively (Fig. 3). Although inhibition of cAMP synthesis was Currents were evoked by 500-ms voltage commands from a holding potential of Ϫ10 mV, delivered in 20 mV increments from Ϫ140 to 60 mV. B, in a protocol designed to measure the effects of various receptors on Kir currents, oocytes were held at Ϫ80 mV (a potential where significant inward current is measured). Expression of mgb1a or FLAG-gb2 receptors alone with or without G i ␣1 resulted in no modulation of current after GABA treatment. Co-expression of mgb1a and FLAG-hgb2 receptors followed by treatment with 100 M GABA resulted in stimulation of Kir 3.1/3.2. Shown are representative traces from at least three independent experiments under each condition.  5 and 7). gb1a-gb2 heterodimers are observed only in mgb1a/FLAG-gb2 coexpressing cells due to the fact that the gb1 receptor was co-immunoprecipitated with the FLAG-gb2 receptor using the FLAG antibody and detected with gb1 antibodies (lane 12). gb1 receptors are not detected in mock-transfected cells and cells expressing gb1a or FLAG-gb2 (lanes 9 -11). gb1-gb2 heterodimers are observed only in mgb1a/FLAG-gb2 expressing cells due to the fact that the FLAG-gb2 receptor was coimmunoprecipitated using the gb1 receptor antibodies and detected with FLAG antibody (lane 16). No FLAG-gb2 receptors are detected in mock-transfected cells or cells expressing mgb1a or FLAG-gb2 (lanes [13][14][15]. The immunoblots shown are from one to three independent experiments. sometimes observed with rgb1a-50/rgb1a and hgb2-42/hgb2, these effects were relatively small (0 -20% inhibition; Fig. 3) when compared with those observed with the coexpressing cells. Neither baclofen nor GABA in the absence of forskolin had any effect on cAMP synthesis (Fig. 3). In addition, wildtype HEK-293 cells did not exhibit baclofen-or GABA-mediated inhibition of forskolin-stimulated cAMP synthesis (Fig. 3). These data demonstrate that the functional GABA B receptor requires both the gb1 and gb2 receptors for signaling via adenylyl cyclase.
Native functional GABA B receptors have been reported to couple to Kirs (3). Co-expression of the mgb1a and hgb2 with Kir 3.1/3.2 resulted in a significant stimulation of Kir current in response to GABA (301 Ϯ 20.6% (n ϭ 3) increase over control current) measured at Ϫ80 mV, which could subsequently be washed out with control solution (Fig. 4) or blocked with the CGP71872 antagonist (data not shown). Modulation of Kir 3.1/3.2 was not seen in oocytes expressing mgb1a or hgb2 individually even in the presence of G i ␣1 (Fig. 4). The dependence of functional GABA B receptor activity on the coexpression of gb1a and gb2 suggests that these receptors may undergo heterodimerization.
gb1a and gb2 Form a Heterodimer-Immunoblot analysis revealed selective expression of mgb1a monomers and homodimers in mgb1a and mgb1a/FLAG-gb2 expressing cells and expression of FLAG-gb2 receptors in FLAG-gb2 and mgb1a/ FLAG-gb2 expressing cells (Fig. 5, lanes 1-8). To demonstrate the existence of gb1a-gb2 heterodimers, we utilized a differential co-immunoprecipitation and immunoblotting strategy. Anti-gb1 receptor antibodies were used to blot receptors immunoprecipitated with anti-FLAG antibodies (Fig. 5, lanes 9 -12). No mgb1a immunoreactivity was detected in samples prepared from mock vector transfected cells, FLAG-gb2 expressing cells, and mgb1a receptor expressing cells as expected because these species could not be immunoprecipitated with the anti-FLAG antibody and detected with the anti-gb1 antibody (Fig. 5, lanes  9 -11). Immunoreactive ϳ250-(representing the mgb1a-gb2 heterodimer) and ϳ130-kDa species (representing the mgb1a monomer) were detected only in cells coexpressing the mgb1a and FLAG-gb2 receptors, demonstrating that gb1a and gb2 can only be co-immunoprecipitated as part of a complex (Fig. 5, lane  12). Similar demonstration of gb1a-gb2 heterodimerization was obtained when coexpressed receptors were immunoprecipitated first with anti-gb1 antibodies followed by immunoblotting with the anti-FLAG antibody (Fig. 5, lane 16). The ϳ250-kDa species represents the heterodimer, whereas the ϳ130-kDa species represents the FLAG-gb2 monomer. No FLAG-gb2 immunoreactivity was detected in samples prepared from mock vector transfected cells, FLAG-gb2 expressing cells and mgb1a receptor expressing cells as expected because these species could not be immunoprecipitated with the anti-gb1 antibody and detected with the anti-FLAG antibody (Fig. 5, lanes 13-15). The gb1a-gb2 heterodimer, which is stable in SDS, might result from SDS-resistant intermolecular transmembrane interactions as reported for the formation of ␤2-adrenergic and dopamine D2 receptor homodimers (5,6). The monomer presumably results from partial disruption of protein-protein binding domains (Sushi Repeats) or C-terminal ␣-helical domains in mgb1a. 2 Disulfide bonds may also contribute to dimer formation as has been reported for the structurally related metabotropic glutamate receptor (7).
We have shown in three disparate expression systems that coexpression of the gb1 and gb2 receptors is required for functional GABA B receptor activity. That the two receptors are co-immunoprecipitated is consistent with the formation of heterodimers. We cannot rule out the possibility that either receptor may function as a monomeric or homodimeric receptor in the appropriate native cells. Indeed there are regions of the brain, such as caudate/putamen, where gb1 mRNA is abundant and gb2 mRNA is undetectable (Fig. 1, A and B). In such regions gb1 would either require a different subunit or function as a monomer or homodimer. However, our in situ hybridization analysis indicates that in brain regions expressing gb2, gb1 is expressed in a majority of the same neurons, suggesting that native receptors are heterodimers in these cells. GPCRs are commonly thought of as monomeric receptors. However, in light of our functional data and in situ hybridization findings, it seems appropriate to consider the gb1 and gb2 receptor proteins to be subunits of a functional GABA B receptor.
Addendum-During the editorial review of this manuscript, similar work from four independent groups was published reporting the formation of functional GABA B receptor heterodimers (19 -22).