GABAB receptors are unique among G-protein-coupled receptors (GPCRs) in their requirement for heterodimerization between two homologous subunits, GABAB1and GABAB2, for functional expression. Whereas GABAB1 is capable of binding receptor agonists and antagonists, the role of each GABAB subunit in receptor signaling is unknown. Here we identified amino acid residues within the second intracellular domain of GABAB2 that are critical for the coupling of GABAB receptor heterodimers to their downstream effector systems. Our results provide strong evidence for a functional role of the GABAB2 subunit in G-protein coupling of the GABAB receptor heterodimer. In addition, they provide evidence for a novel “sequential” GPCR signaling mechanism in which ligand binding to one heterodimer subunit can induce signal transduction through the second partner of a heteromeric complex.
GABABreceptors are G-protein-coupled receptors (GPCRs) that mediate slow synaptic inhibition in the brain and spinal cord (for review, seeBowery, 1993; Kerr and Ong, 1995; Couve et al., 2000). They are unique among type C GPCRs in that they are heterodimers of GABAB1 (Kaupmann et al., 1997) and GABAB2 subunits (Jones et al., 1998; Kaupmann et al., 1998; White et al., 1998; Kuner et al., 1999; Ng et al., 1999), and it is now generally accepted that each subunit is unable to form a functional receptor when expressed in isolation (Kaupmann et al., 1997;Jones et al., 1998; White et al., 1998; Ng et al., 1999). With respect to GABAB1, this is attributable to its retention within the endoplasmic reticulum on homomeric expression (Couve et al., 1998; Calver et al., 2000; Filippov et al., 2000; Margeta-Mitrovic et al., 2000). In this context, GABAB2 has been shown to play a key role in trafficking the GABAB1 subunit to the cell surface (Couve et al., 1998; Filippov et al., 2000). This exposes the N-terminal ligand binding domain of GABAB1, which is not present in GABAB2 (Galvez et al., 2000a,b), enabling it to bind extracellular agonist and subsequently activate downstream signaling pathways. Interestingly, it has been shown recently that C-terminal GABAB1 mutants, which are expressed on the cell surface in the absence of GABAB2, are completely nonfunctional, despite their ability to bind GABA (Calver et al., 2000; Margeta-Mitrovic et al., 2000). GABAB receptor function is restored, however, when these mutants are coexpressed with GABAB2, demonstrating that the GABAB2 subunit may not only be important for correct trafficking of GABAB1 but also for the mediation of agonist-induced G-protein coupling. This hypothesis has been given additional strength by a recent report by Galvez et al. (2001), demonstrating that the extracellular domain of GABAB2 is essential for agonist activation of the heterodimeric receptor.
Mechanisms for G-protein coupling have been investigated widely for members of the rhodopsin–β-adrenergic GPCR family. Growing evidence suggests that the second and third intracellular loops are critical interaction sites important for both G-protein coupling and selectivity. In addition, these intracellular domains have been demonstrated to be important for G-protein coupling of type C GPCRs, such as metabotropic glutamate receptors (mGluRs) (Gomeza et al., 1996;Francesconi and Duvoisin, 1998). At present, the role of the two GABAB subunits in G-protein coupling and downstream signal transduction is unknown. Here, using a site-directed mutagenesis approach, we investigated the importance of residues in the second intracellular loop (il2) of GABAB1 and GABAB2 subunits. We demonstrated that mutations within il2 of GABAB2 can dramatically decrease responses of the heterodimer complex to agonist, to the extent that receptor signaling can be effectively abolished by the introduction of three negatively charged residues into this region. In contrast, reciprocal mutations made in il2 of the GABAB1subunit have no apparent effect on heterodimer signaling. These results are consistent with a model in which agonists bind to the GABAB1 subunit, resulting in signal transduction via G-protein coupling through the GABAB2subunit. This would therefore suggest the existence of a novel “sequential and sideways” signaling cascade that may be of relevance to the growing number of reported GPCR heterodimers (Milligan and Rees, 2000).
MATERIALS AND METHODS
Site-directed mutagenesis. The GABAB1b and GABAB2 subunits were tagged with either an N-terminal c-myc (wtGABAB1) or hemaglutinin (HA) (wtGABAB2) epitope tag (Calver et al., 2001). PCR primers were designed to include single, double, triple, or quintuple amino acid changes in both the wtGABAB1 and wtGABAB2, resulting in 12 mutant constructs: GABAB1 E579K, GABAB1 E583K, GABAB1 E579K/E580K, GABAB1 E579K/E580K/E583K(referred to as GABAB1 tripleK), GABAB2 K586E, GABAB2 K590E, GABAB2 K586E/M587E,GABAB2 K586E/M587E/K590E(referred to as GABAB2 tripleE), GABAB1 K577/578A, GABAB1 K581/582A, GABAB1 K577/578/581/582/586A(referred to as GABAB1 KQA), and GABAB1 K577/578/581/582/586E(referred to as GABAB1 KQE). Mutagenesis experiments were performed using the Quikchange site-directed mutagenesis kit (Stratagene, La Jolla, CA) according to the instructions of the manufacturer, and all amino acid changes were confirmed by full double-stranded sequencing.
C-terminal truncates (from amino acid 747 of GABAB1) of the GABAB1mutant constructs (glutamate to lysine only) were made by PCR amplification using forward primer 5′-CCCGAATTCATGGG GCCCGGGGCC-3′ and reverse primer 5′-CCCAAGCTTCCTCG GGTGATCAGCCTG-3′. A GABAB1N/2 chimera was generated by ligation of two PCR products (GABAB1b 1–470 and GABAB2 479–954). PCR amplification was performed using forward primer 5′-CCCGA ATTCATGGGGCCCGGGGCC-3′ and reverse primer 5′-CCCAAGC TTCTGGTCAGCTGGGGGGGAC-3′ to generate fragment GABAB1b 1–470. PCR amplification with forward primer 5′-CCCAAGCTT ATCATCCTGG AGCAGCTGCGGAAG-3′ and reverse primer 5′-CCCCTCGAGTTACAGGCCCGAGACCA TGACTC-3′ was used to generate fragment GABAB2479–954. All constructs were cloned into the eukaryotic expression construct pcDNA3.1 (Invitrogen, Paisley, UK). Schematic representations of these constructs are shown in Figure1 c.
Transfection. Human embryonic kidney 293 (HEK293) cells were transfected with either wild-type (wt) or mutant GABAB1- or GABAB2-containing plasmids using Lipofectamine Plus (Life Technologies, Paisley, UK) according to the instructions of the manufacturer. Cells were maintained in Minimal Essential Medium supplemented with 10% fetal bovine serum and 1% nonessential amino acids (all from Life Technologies). After transfection, cells were left for 24 hr before being subcultured for immunocytochemistry or the Ca2+ mobilization assay.
The GABAB1b/GABAB2 cell line was generated in Chinese hamster ovary cells as described previously (Hirst et al., 2000). The GABAB2 cell line was generated in HEK293 cells by transient transfection of the GABAB2 cDNA as described above and selection of positive cells in 800 μg/ml G418.
Immunocytochemistry. HEK293 cells transfected with either wild-type or mutant constructs were subcultured onto glass coverslips, fixed with 4% paraformaldehyde, and then either permeabilized with 0.1% Triton X-100 for 10 min or washed with PBS. Cells were incubated with primary antibody [anti-c-myc, 1:5000; anti-HA, 1:5000 (Boehringer Mannheim, Brüssel, Belgium); and anti-GABAB1b, 1:1000] for 60 min and washed, and then secondary antibody [goat anti-mouse IgG FITC for c-myc; goat anti-rat IgG FITC for HA (Sigma, St. Louis, MO); and goat ant-rabbit IgG FITC for GABAB1b] was used at 1:100 dilution for 60 min. Cells were finally washed, mounted onto glass slides using Citifluor (Citifluor Ltd., London, UK), and viewed using a Leica (Nussloch, Germany) confocal microscope.
Ca 2+ mobilization assay. Transfected cells were seeded into 96-well plates at a density of 50,000 cells per well and incubated at 37°C in 5% CO2 for 24 hr before use. The intracellular Ca2+ mobilization in response to agonist was then measured as described previously (Calver et al., 2000). The data were iteratively curve fitted using a four parameter logistic model (Bowen and Jerman, 1995) as mean ± SEM (with each data point being determined in triplicate) of one representative experiment. Significance of data were assessed using a one-way ANOVA, followed by post hoc t test (least square difference). Data were considered significant ifp < 0.05.
Radioligand binding. Membranes from transfected cells were prepared, and radioligand binding to [3H]CGP54626 was performed as described previously (Calver et al., 2000). The concentration of drug-inhibiting-specific radioligand binding by 50% (IC50) was determined by iterative curve fitting (Wood et al., 2000). pK i values (the negative log10 of the molarK i) for receptor binding were then calculated from the IC50 values as described byCheng and Prusoff (1973) using K Dvalues determined previously in saturation binding studies (3 nm). B max values were calculated from the specific bound accounting for receptor occupancy at this radioligand concentration using the Hill–Langmuir adsorption isotherm. Data are expressed as the mean ± SEM of at least three separate experiments from two independent sets of transfections.
[35S]GTPγS binding assays.Cells were homogenized in 20 mm HEPES and 10 mm EDTA, pH 7.4 (4°C), and centrifuged at 48,000 × g for 20 min at 4°C. Membrane pellets were rehomogenized in 20 mm HEPES and 0.1 mm EDTA, pH 7.4, (4°C). The pellet was recentrifuged as described above, the supernatant was discarded, and the pellet was resuspended in 20 mm HEPES and 0.1 mm EDTA and stored at −80°C until required.
[35S]GTPγS binding assays were performed in a 20 mm HEPES buffer, pH 7.4, containing 3 mm MgCl2 and 100 mm NaCl. Cell membranes were preincubated with 10 μm GDP, and increasing concentrations of GABA or baclofen for 30 min at 30°C and then 0.1 nm [35S]GTPγS was added to each well and incubated for an additional 30 min. Nonspecific binding was determined in the presence of 20 μm GTPγS. The incubation was terminated by rapid filtration through J. Whatman glass fiber filters (GF/B) (Semat International, St. Albans, UK) and washed with 3 ml of ice-cold HEPES buffer, pH 7.4, containing 3 mm MgCl2. Radioactivity was determined by liquid scintillation spectrometry using a Packard (Meridian, CT) TopCount.
cAMP accumulation assays. cAMP levels in cells were determined by radioimmunoasay (SMP004; NEN, Boston, MA) according to the instructions of the manufacturer. In brief, cells were washed once with Ca2+-free PBS, scrapped up in the same buffer, and pelleted by a 5 min centrifugation at 400 ×g. The pellet resuspended in manufacturers stimulation buffer, and ∼50,000 cells were added to the appropriate wells of the NEN flash plates together with 10 μm forskolin, plus increasing concentration of agonist. Plates were incubated for 15 min at 37°C, before the addition of the manufacturers detection mixture containing [125I]cAMP tracer (0.16 μCi/ml) to the wells. Plates were covered and left for 24 hr before counting on a Packard TopCount.
Neuron preparation and cDNA injection. Neuron isolation and injection procedures have been described previously (Caulfield et al., 1994; Couve et al., 1998; Filippov et al., 2000). Briefly, single superior cervical ganglion (SCG) neurons were dissociated from 15-to 19-d-old rats and plated on laminin-coated glass coverslips. Five hours after plating, neurons were microinjected into the nucleus with plasmids carrying cDNAs for the GABAB receptor subunits, in addition to a plasmid encoding green fluorescent protein to subsequently identify successfully injected cells. Electrophysiological recordings were routinely made 16–20 hr after injection at room temperature (20°C).
Ca 2+ channel current recording. Currents through voltage-gated Ca2+ channels were recorded using the conventional whole-cell patch-clamp method as described previously (Caulfield et al., 1994; Couve et al., 1998; Filippov et al., 2000). Ca2+ channel currents were routinely evoked every 20 sec with 100 msec depolarizing rectangular test pulse to 0 mV from a holding potential of −90 mV. Ca2+ channel current amplitudes were measured isochronally 10 msec from the onset of the rectangular test pulse, i.e., near to the peak of the control current. As reported previously (Filippov et al., 2000), currents were primarily N-type with negligible contribution by L-type channels. A racemic mixture of (+)- and (−)-baclofen was used in all experiments. Data are presented as means ± SEM as appropriate. Student's test (unpaired) was applied to determine statistical significance. Difference were considered significant if p < 0.05.
Sequence analysis of the second and third intracellular loops of GABAB1 and GABAB2
Using sequence alignments, we compared residues in the second and third intracellular loops of the GABABsubunits and the related type C mGluRs (Fig. 1 a). The basic residue Arg775 in the third intracellular loop (il3) of mGluR1 has been shown to be important for coupling to G-proteins (Francesconi and Duvoisin, 1998) and is conserved throughout the mGluR family and also in GABAB2 (Fig.1 a). In contrast, the corresponding residue in GABAB1 is a lysine. In il2 of all of the mGluRs, there is a conserved lysine at position 690 (numbered with respect to mGluR1), except for mGluR2 in which the equivalent residue is an arginine, and this basic residue has been shown to be crucial for the interaction of the mGluRs with their respective G-protein (Francesconi and Duvoisin, 1998). Similarly, we noted that GABAB2 also has a basic lysine at this key position. This is in contrast to the GABAB1subunit, in which the acidic amino acid glutamate is substituted for this lysine, resulting in a reversal of charge at this site. These findings may be of functional relevance, because it has been proposed previously that electrostatic interactions between positive charges in the intracellular domains of GPCRs may be important for coupling to a negatively charged face of the G-protein (Fanelli et al., 1998).
Amino acids in the second intracellular loop of GABAB2are critical for GABAB receptor signaling through the chimeric G-protein Gqi5
Based on our sequence analysis of the GABAB1 and GABAB2intracellular loops, we targeted three amino acid residues (K586, M587, and K590) within il2 of GABAB2 for mutagenesis studies (Fig. 1 a,b). These residues were mutated to glutamates either individually or in combination to give a series of single, double, or triple mutants: GABAB2 K586E, GABAB2 K590E, GABAB2 K586E/M587E, and GABAB2 tripleE (Fig.2 a). When transfected into HEK293 cells in isolation, both wild-type and mutant GABAB2 subunits could be detected on the cell surface in the absence of membrane permeabilization (Fig.2 b). In addition, coexpression of each GABAB2 mutant with wtGABAB1demonstrated that each mutant retained its ability to traffic the GABAB1 subunit to the cell surface (Fig.2 b). Having demonstrated that these mutants behaved normally in terms of trafficking to the cell surface, we next tested the ability of the mutated GABAB2 subunits to functionally couple to G-proteins when coexpressed with the wild-type GABAB1 subunit. Although GABAB receptors are known to inhibit adenylyl cyclase activity by preferentially coupling to Go/i, they can also be made to signal via the phospholipase C pathway when coexpressed with the chimeric G-protein Gqi5 (Franek et al., 1999; Wood et al., 2000). Activation of this pathway results in a mobilization of intracellular Ca2+ stores, which can then be measured on a fluorimetric imaging plate reader (FLIPR). Coexpression of wtGABAB1 and wtGABAB2subunits with Gqi5 in HEK293 cells resulted in a robust Ca2+ mobilization in response to GABA (pEC50, 7.19 ± 0.06; n= 36). In the same system, expression of wtGABAB1with either GABAB2 K586Eor GABAB2 K590E resulted in significantly reduced pEC50 values when compared with coexpression with wtGABAB2(6.35 ± 0.07, n = 11, p < 0.001; and 6.07 ± 0.04, n = 12, p < 0.001, respectively) (Fig. 2 c). Expression of wtGABAB1 with the double-mutant GABAB2 K586E/M587Eresulted in additional reduction of the observed pEC50 in response to GABA (5.40 ± 0.11;n = 10; p < 0.001). In contrast to the other GABAB2 mutants studied, expression of GABAB2 tripleE in combination with wtGABAB1 resulted in no detectable responses in this functional assay (n = 12) (Fig. 2 c).
Based on these results, we next prepared mutations in the second intracellular loop of GABAB2 that would help us investigate the relative importance of losing or gaining positive charge within this stretch of five amino acids. To do this, we examined the effects of changing residues K586, M587, and K590 to neutral alanines instead of negatively charged glutamates. As with the previous set of glutamate mutants, cell surface expression of each of the alanine mutants in isolation appeared no different to that of the wtGABAB2 subunit (data not shown). Expression of wtGABAB1 and Gqi5 in combination with GABAB2 K586A or GABAB2 K590A resulted in significantly reduced pEC50 values when compared with coexpression with wtGABAB2 (6.30 ± 0.13, n = 8, p < 0.001; and 5.75 ± 0.08, n = 10, p < 0.001, respectively) (Fig. 2 d). Coexpression with the double-mutant GABAB2 K586A/M587Aresulted in similar significant reduction in pEC50 in response to GABA (6.16 ± 0.07;n = 8; p < 0.001). In contrast to the other GABAB2 mutants studied, and similar to the GABAB2 tripleE mutant, expression of GABAB2 tripleA in combination with wtGABAB1 resulted in no detectable responses in this functional assay (Fig. 2 d).
Mutations in the second intracellular loop of GABAB2have no effect on agonist or antagonist binding
Having shown a marked effect on GABABreceptor signaling after mutation of selected residues within the second intracellular loop of GABAB2, we wanted to see whether the agonist and antagonist binding properties of the mutant receptor heterodimer were consistent with that of the wild-type receptor. Analysis of membrane homogenates expressing each GABAB2 mutant in combination with wtGABAB1 had no effect on the ability of GABA to displace the antagonist CGP54626 in competition binding assays. The potency of GABA observed in homogenates prepared from cells expressing GABAB2 K586E, GABAB2 K590E, GABAB2 K586E/M587E, or GABAB2 tripleE with wtGABAB1 subunit was not significantly different from that observed when the wild-type GABAB2subunit was coexpressed with the GABAB1 subunit (pK i values of 4.57 ± 0.17, 4.22 ± 0.05, 4.48 ± 0.22, 4.84 ± 0.54, and 4.37 ± 0.12, respectively; n = 3) (Fig. 2 e).B max values of [3H]CGP54626 specifically bound did not differ significantly between wild-type and mutants and ranged from 5.6 ± 3.3 to 8.3 ± 3.7 pmol/mg protein.
Negatively charged amino acids in the second intracellular loop of GABAB1 are not critical for GABAB receptor signaling through Gqi5
The demonstration that mutation of specific residues within the second intracellular loop of GABAB2 can have pronounced effects on GABAB signaling through Gqi5 led us to examine the functional importance of similarly positioned residues within the GABAB1 subunit. We therefore studied the consequences of mutating the stretch of negative amino acids found in the second intracellular loop of GABAB1 and in a similar position to those residues mutated previously in GABAB2 (Fig.3 a). Each of the mutants GABAB1 E579K, GABAB1 E583K, GABAB1 E579K/E580K, and GABAB1 tripleK were transfected into HEK293 cells with and without wtGABAB2. In the absence of GABAB2, neither wt or mutant GABAB1 variants were expressed at the cell surface (Fig. 3 b and data not shown). However, when coexpressed with the GABAB2 subunit, cell surface expression of all GABAB1 mutants was observed as expected (Fig. 3 b and data not shown). These experiments therefore suggest that mutation of these acidic residues in the second intracellular loop of GABAB1 do not affect trafficking of the subunit to the surface by GABAB2. We then tested the ability of these constructs to activate Gqi5 in response to GABA. Coexpression of GABAB2 with GABAB1 E579K, GABAB1 E579K/E580K, GABAB1 E583K, or GABAB1 tripleK produced no significant changes in functional response when compared with coexpression with the wtGABAB1 subunit (pEC50 values for wtGABAB1, 7.19 ± 0.06, n = 36; GABAB1 E579K, 7.03 ± 0.13, n = 8; GABAB1 E579K/E580K, 7.18 ± 0.16, n = 8; GABAB1 E583K, 7.02 ± 0.13, n = 7; and GABAB1 tripleK, 7.18 ± 0.16, n = 7) (Fig. 3 c). This suggests that these acidic residues are not critical in the signaling of the GABAB1 receptor dimer through Gqi5.
Addition of positively charged residues to the second intracellular loop of a cell surface-expressed GABAB1 is not sufficient for receptor function in the absence of GABAB2
Although mutation of acidic residues in the second intracellular loop of GABAB1 appeared to have no functional effect with respect to heterodimer signaling, we also wanted to test the function of GABAB1 mutants when expressed on the cell surface in the absence of GABAB2. We and others have shown previously that removal of the C-terminal domain of the GABAB1 subunit results in its cell surface expression in the absence of GABAB2. However, despite this cell surface expression, the GABAB1subunit remains ineffective in coupling to downstream effector systems in the absence of GABAB2 (Calver et al., 2000;Margeta-Mitrovic et al., 2000). In this set of experiments, we therefore produced four C-terminal truncates: GABAB1C- E579K, GABAB1C- E579K/E580K, GABAB1C- E583K, or GABAB1C- tripleK. Expression of these mutants showed that they were all able to reach the cell surface in the absence of GABAB2 (Fig.3 b and data not shown). We then compared the ability of each GABAB1 C-terminal truncate mutant to couple to Gqi5 after stimulation by GABA. All of the GABAB1 mutants tested were nonfunctional when expressed alone (n = 4) (Fig. 3 d and data not shown). In contrast, expression of each truncated GABAB1 mutant gave a robust signal when coexpressed with GABAB2, no different to that seen with wtGABAB1 (data not shown). This suggests that removal of negatively charged glutamic acid residues within the second intracellular loop of GABAB1 is not sufficient to allow signaling and strengthens the argument that the GABAB2 subunit is absolutely necessary for G-protein signaling.
Positively charged amino acids in the second intracellular loop of GABAB1 are not critical for GABAB receptor signaling through Gqi5
We also noted when aligning the GABAB1 and GABAB2 il2 sequences that there were five positively charged lysine residues in this region of GABAB1, although they do not directly align with the positive residues within il2 of GABAB2 that we implicated in G-protein coupling. We therefore wanted to exclude the possibility that these amino acids might be involved in receptor signaling. We studied the effects of mutating the five lysine residues in il2 of GABAB1 to neutral alanines and/or negatively charged glutamates, both in pairs and all five together (Fig. 4 a). Each of the mutants GABAB1 K577/578A, GABAB1 K581/582A, GABAB1 KQA, and GABAB1 KQE were transfected into HEK293 cells together with GABAB2. This resulted in cell surface expression of all of the GABAB1 mutants as expected (Fig.4 b). We then tested the ability of these constructs to activate Gqi5 in response to GABA. Coexpression of GABAB2 with GABAB1 K577/578A, GABAB1 K581/582A, GABAB1 KQA, or GABAB1 KQE resulted in the expression of functional GABAB receptors. Neither of the double mutations (GABAB1 K577/578A or GABAB1 K581/582A) produced any significant changes in the potency of GABA when compared with the wild-type receptor, and although the potency of GABA at the quintuple mutant subunits (GABAB1 KQA and GABAB1 KQE) was slightly lower than the wild type, this was a very minor effect when compared with the mutations in GABAB2(pEC50 values for wtGABAB1, 7.19 ± 0.06, n = 36; GABAB1 K577/578A, 7.28 ± 0.04, n = 9; GABAB1 K581/582A, 7.17 ± 0.08, n = 9; GABAB1 KQA, 6.74 ± 0.08, n = 12; and GABAB1 KQE, 6.76 ± 0.03, n = 9) (Fig. 4 c). These data suggest that the basic residues in il2 of GABAB1 are also not critical for the signaling of the GABABreceptor dimer through Gqi5.
Exchanging charged residues between the second intracellular loops of GABAB1 and GABAB2 abolishes GABAB receptor functional coupling to G-proteins
Because GABAB2 tripleE when coexpressed with the wild-type GABAB1 produced a nonfunctional receptor, we coexpressed GABAB2 tripleE with GABAB1 tripleK in HEK293 cells. Despite both subunits being expressed at the cell surface (Fig.5 a), no functional response was detected in the Ca2+ mobilization assay (n = 8) (Fig. 5 b). This suggests that GABAB2 is essential for G-protein coupling in the GABAB receptor.
Amino acids in the second intracellular loop of GABAB2are critical for GABAB receptor signaling through endogenous G-proteins in superior cervical ganglion cells
It was important to determine that the functional effects observed using our GABAB2 mutant constructs in conjunction with Gqi5 could be reproduced in an effector system coupled to endogenously expressed G-proteins. We demonstrated previously that SCG neurons express only the GABAB1 subunit. However microinjection of GABAB2 constructs into these neurons results in the expression of functional GABAB receptors that inhibit Ca2+ channels (Filippov et al., 2000). Microinjecton of GABAB2 K590E and GABAB2 tripleE resulted in cell surface expression of GABAB2 (Fig.6 a) and GABAB1 subunits (data not shown). Expression of the wtGABAB2 subunit and stimulation by baclofen resulted in an ∼53 ± 2.02% inhibition of voltage-activated Ca2+ currents (Fig.6 b,c), as measured 10 msec after the voltage pulse. Expression of GABAB2 K590E resulted in significantly reduced inhibition of Ca2+currents when compared with wtGABAB2 (42 ± 1.92%) inhibition (Fig. 6 b–d). In contrast to the other mutants studied, injection of the GABAB2 tripleE mutant resulted in no baclofen-stimulated inhibition of Ca2+ channel opening above that observed in mock-injected neurons. This demonstrates that the GABAB2 tripleE mutant, in conjunction with endogenous GABAB1, forms a completely nonfunctional GABAB receptor (Fig.6 b,c).
The GABAB2 subunit alone is unable to functionally respond to GABA
It is known that the GABAB1 subunit contains a binding site for GABA in its N-terminal domain (Malitschek et al., 1999), but it is still unclear from published work as to whether GABAB2 is capable of binding and responding to GABA on its own. We therefore generated a GABAB2stable cell line in HEK293 cells, in which we confirmed GABAB2 expression on the cell surface by both immunocytochemistry and Western blotting (data not shown). We were unable to observe any [35S]GTPγS binding during stimulation with GABA up to a concentration of 1 mm (Fig. 7 a). Similarly, we were unable to demonstrate any inhibition of forskolin-stimulated adenylate cyclase activity in these cells in response to GABA at a concentration of up to 100 μm (Fig. 7 b). In cell lines expressing both GABAB1 and GABAB2 on the other hand, we were readily able to demonstrate functional coupling using both of these techniques with the pharmacology expected from recombinant GABABreceptors (Fig. 7 a,b).
The N-terminal domain of GABAB2 is also necessary for normal GABAB receptor signaling
To investigate whether the N-terminal domain of GABAB2 is important in normal GABAB receptor function, we made a chimeric GABAB2 construct by replacing the N-terminal binding domain of GABAB2 with that of GABAB1. Expression of this GABAB1N/2 subunit was clearly observed on the cell surface, and furthermore it retained the ability to traffic wtGABAB1 to the cell surface (Fig.7 c). However, functional analysis of cells expressing both GABAB1N/2 and wtGABAB1exhibited a complete lack of response to GABA (n = 4) (Fig. 7 d). This suggests that the N-terminal domain of GABAB2 may also be important in the mediation of receptor responses to GABA, despite the fact that this subunit does not actually bind GABA.
GABAB receptors are the only members of the type C family of GPCRs that have been shown to function as heterodimers. They are also distinct from other reported GPCR heterodimers in which both members of the heterodimer complex are able to form functional receptors when expressed as monomers (Jordan and Devi, 1999; AbdAlla et al., 2000; Gines et al., 2000; Gomes et al., 2000; Rocheville et al., 2000). In this study, we wanted to further analyze the role of each GABAB subunit with respect to G-protein coupling and signal transduction, focusing on residues within the second and third intracellular loops. Residues within these domains have been demonstrated previously to be critical for G-protein interactions in a number of receptors. More specifically, recent studies on the functional roles of the cytoplasmic domains of rhodopsin have suggested that the third intracellular loop contains sites important in determining G-protein specificity, whereas the second intracellular loop contains regions essential for G-protein activation (Yamashita et al., 2000). Sequence comparison of residues in il2 and il3 for both the GABAB subunits and the related mGluRs highlighted clear differences between GABAB1 and GABAB2 of potential relevance for G-protein coupling (Fig. 1 a), suggesting that it may be GABAB2 rather than GABAB1 that is involved in the coupling of the GABAB receptor to G-proteins and downstream signaling. Furthermore, comparison of 46 il2 sequences from various species demonstrated that all of the mGluRs and GABAB2 contained fewer than two negatively charged residues within this loop. This is in sharp contrast to GABAB1, which contains four negatively charged residues. This may be of particular relevance in the context of modeling studies by Fannelli et al. (1998), who investigated the electrostatic complimentarity of receptor-G-protein complexes. These studies favor a model whereby “opening” of the cytosolic domains of the receptor allows for an interaction between the electrostatically positive surface of the domains of the receptor and the negatively charged surface of the Gαsubunit (Higgs and Reynolds, 2001). In this respect, it is of interest that the electrostatic potential of Go, reported to be the predominant G-protein interacting with GABAB receptors (Leaney and Tinker, 2000), is proposed to be one of the most negative of the Gα subunits (C. Reynolds, personal communication). Based on these observations, we can speculate that the negatively charged residues of the GABAB1 subunit would make it a less attractive candidate for G-protein coupling than GABAB2.
To determine the functional role that il2 residues play in GABAB receptor function, we made a series of amino acid substitution mutants at K586, M587, and K590 in GABAB2.Single or double substitutions of these residues to glutamate or alanine did not affect the ability of GABAB2 to reach the cell surface itself or to traffic GABAB1 to the cell surface but led to a significant reduction in agonist potency when compared with the wild-type receptor heterodimer. In addition, when all three residues were substituted by glutamates or alanines, it resulted in a receptor that was completely unresponsive to agonists. These changes in agonist efficacy were not attributable to aberrant folding of the receptor subunits or changes in binding affinity because both agonist and antagonist binding was unaltered when compared with wild-type receptor. Moreover, our binding data were consistent with immunocytochemical data and suggested that there was no significant effect on receptor expression levels. Importantly, these mutants also interfered with the physiological coupling of GABAB receptors to Ca2+ channels in SCG neurons, confirming that mutation of these residues in il2 of GABAB2was sufficient to abolish GABAB receptor signaling. These data therefore identify key residues important in G-protein coupling and activation of the receptor heterodimer and is in agreement with previous proposals suggesting that the heptahelical domains of GABAB2, and not GABAB1, contain these molecular determinants (Calver et al., 2001; Galvez et al., 2001).
To further investigate the role of il2 residues in the heterodimer, we substituted the corresponding negatively charged residues in GABAB1, E579, E580, and E583, to lysines. These GABAB1mutants were all expressed on the cell surface after coexpression with GABAB2, and all responded to agonist as effectively as the wild-type receptor heterodimer. Similarly, GABAB1 mutants in which all five positively charged lysine residues in this region were changed to either neutral or negatively charged amino acids are completely functional within heterodimers with wild-type GABAB2. Furthermore, a number of other GABAB1 mutants we expressed all showed a complete lack of responsiveness to GABA in the absence of GABAB2. Although this data does not completely discount the possibility that G-proteins may be interacting with il2 of the GABAB1 subunit, it does suggest that it is less likely that there are electrostatic interactions between GABAB1 and G-proteins. A recent study however has reported that replacement of the entire seven transmembrane and intracellular domains of GABAB1 with that of GABAB2 (termed GB1/2) reduces the efficacy of G-protein coupling to a “GB1/2 GB2” receptor (Galvez et al., 2001). Although this may be indicative of a modulatory role for GABAB1 in G-protein coupling, we would suggest that it is more likely as a result of less efficient signaling between GABAB1 and its “downstream” partner, GABAB2.
Molecular modeling studies have identified residues in the GABAB1 subunit that are critical for binding of GABAB agonists and antagonists, and these residues are absent from the GABAB2 subunit (Galvez et al., 2000a,b). In this study, we demonstrated that GABAB2 expressed in the absence of GABAB1 in HEK293 cells is unable to respond functionally to GABA, by either inhibition of adenylate cyclase activity or binding of [35S]GTPγS. However, the absence of a ligand binding site does not necessarily mean that GABAB2 is not important in signal transduction. Indeed, the recent solving of the crystal structures for the N-terminal binding domain of mGluR1 demonstrate that movements between all four globular lobes of the N-terminal domains of the dimer are likely to cause shifts in transmembrane and intracellular domains, resulting in receptor activation (Kunishima et al., 2000). These findings may help to explain results from this and other studies in which deletion or substitution of substantial parts of the GABAB2 N-terminal domain results in a complete loss of GABAB receptor function (Jones et al., 2000; Galvez et al., 2001). This therefore suggests that the N-terminal domain of the GABAB2 subunit, although unable to bind GABA itself, may instead be involved in the conformational changes induced after ligand binding to the GABAB1subunit.
In summary, our study has determined three residues within il2 of the GABAB2 subunit critical for G-protein signaling of the GABAB receptor heterodimer, demonstrating that this signaling absolutely requires the presence of the GABAB2 protein. Although our data do not rule out a possible involvement of GABAB1 in the signaling process, these studies, together with current knowledge of this receptor heterodimer, strongly suggest a novel mechanism of “sideways” signal transduction, whereby agonist binds to the GABAB1 subunit, resulting in a conformational change that is in some way passed on to the GABAB2 subunit, perhaps through its N-terminal and/or transmembrane domains. This in turn results in the receptor forming an activated state suitable for G-protein recruitment, via a direct interaction with GABAB2, and thus downstream signal transduction (Fig. 8). Such novel sideways signaling may allow increased complexity of receptor signal transduction, which may also be applicable to other GPCR heterodimers.
A.C. and S.J.M. are supported by the Wellcome Trust and the Medical Research Council. A.K.F. and D.A.B. are supported by the Wellcome Trust. We thank Prof. Derek Middlemiss and Prof. Gary Price for helpful discussion and comments on this manuscript.
M.J.R and A.R.C. contributed equally to this work.
Correspondence should be addressed to Menelas N. Pangalos, Department of Neurology CEDD, GlaxoSmithKline, New Frontiers Science Park, Third Avenue, Harlow, Essex CM19 5AW, UK. E-mail:.