Mapping the Agonist-binding Site of GABAB Type 1 Subunit Sheds Light on the Activation Process of GABABReceptors*

The γ-amino-n-butyric acid type B (GABAB) receptor is composed of two subunits, GABAB1 and GABAB2, belonging to the family 3 heptahelix receptors. These proteins possess two domains, a seven transmembrane core and an extracellular domain containing the agonist binding site. This binding domain is likely to fold like bacterial periplasmic binding proteins that are constituted of two lobes that close upon ligand binding. Here, using molecular modeling and site-directed mutagenesis, we have identified residues in the GABAB1 subunit that are critical for agonist binding and activation of the heteromeric receptor. Our data suggest that two residues (Ser246 and Asp471) located within lobe I form H bonds and a salt bridge with carboxylic and amino groups of GABA, respectively, demonstrating the pivotal role of lobe I in agonist binding. Interestingly, our data also suggest that a residue within lobe II (Tyr366) interacts with the agonists in a closed form model of the binding domain, and its mutation into Ala converts the agonist baclofen into an antagonist. These data demonstrate the pivotal role played by the GABAB1 subunit in the activation of the heteromeric GABAB receptor and are consistent with the idea that a closed state of the binding domain of family 3 receptors is required for their activation.

The ␥-amino-n-butyric acid type B (GABA B ) 1 receptor is a member of the family 3 heptahelix receptors (1). This protein family also includes the metabotropic glutamate (mGlu) (2), the Ca 2ϩ -sensing (3), fish olfactory (4), and some mammalian putative taste (5) and pheromone (6,7) receptors. An original feature of these receptors is that their ligand-binding site is located within their large extracellular N-terminal domain structurally independent from the transmembrane heptahelix core region (8 -10). It is currently not known how the agonist binding within this extracellular domain leads to the conformational changes of the heptahelical core region required for G-protein activation (11). Answering this question will give insights into the specific functional properties of these family 3 receptors. As a first step, it is important to get information on the structural features of their ligand binding domain.
It has previously been proposed that the ligand binding domain of family 3 heptahelix receptors displays a three-dimensional structure similar to that of bacterial periplasmic binding proteins (PBP) such as the leucine/isoleucine/valinebinding protein (LIVBP) (12)(13)(14)(15). These PBPs are part of a family of structurally related proteins that also includes AmiC, a cytoplasmic amide-binding protein involved in the control of the amidase operon in Pseudomonas aeruginosa (16,17). These proteins consist of two distinct globular lobes interconnected by three linkers (16,18,19). Numerous structural studies have revealed that these PBPs adopt at least two stabilized conformations as follows: an "open," usually unbound form, and a "closed" form stabilized by the bound ligand (19). In the closed conformation, the ligand is trapped between the two lobes. Accordingly it has been proposed that agonist binding at family 3 receptors stabilizes an active (closed) conformation of their binding domain that in turn activates the heptahelix core region (11).
In agreement with the above described hypothesis, the ligand binding domains of mGlu 1 (8), mGlu 4 (9) and GABA B 1 (10) receptors have been shown to retain their ligand-binding properties in the absence of the heptahelix domain. Moreover, molecular modeling and mutagenesis studies have further confirmed the similarity between the ligand-binding sites of PBPs such as LIVBP and that of family 3 receptors (13,14). Up to now, only residues within one lobe (lobe I) have been identified as being important for agonist binding on family 3 receptors (mGlu 1 , mGlu 4 and Ca 2ϩ -sensing receptors) (13,14,20). Therefore the importance of the second lobe (lobe II) for ligand binding and receptor activation remains to be elucidated.
Another feature of family 3 heptahelix receptors is that they are found mostly as dimers (21,22). The role of this dimerization process in the activation of the receptor is still unclear. Within this receptor family, the GABA B receptor constitutes an excellent model for the study of the activation mechanism of these receptors. Indeed, this receptor is a heterodimer constituted of two subunits, GABA B 1 and GABA B 2, both necessary for a fully functional receptor (23)(24)(25)(26)(27)(28). Accordingly, the role of each subunit in the activation process can be analyzed.
In a previous study, we provided evidence supporting a similar three-dimensional structure for LIVBP and the binding domain of GABA B 1 subunit, the only subunit able to bind the high affinity GABA B receptor radioligands (24), and we identified Ser 246 as being critical for binding of the antagonist 125 I-CGP64213 (15). In the present study, we have identified residues in the lobe I of the GABA B 1 subunit critical for agonist binding and activation of the heteromeric receptor. Moreover, we show that the mutation of a residue in lobe II of the GABA B 1 subunit converts the specific heteromeric GABA B receptor agonist baclofen into an antagonist. In addition to demonstrating the requirement of GABA binding on GABA B 1 subunit for the activation of the GABA B heteromeric receptor, our data are consistent with the idea that a closed form of the binding domain of family 3 receptors is required for their activation.

EXPERIMENTAL PROCEDURES
Materials-GABA was obtained from Sigma (L'Isle d'Abeau, France). 125 I-CGP64213 was synthesized as described elsewhere (29) and labeled to a specific radioactivity of Ͼ2,000 Ci mmol Ϫ1 (ANAWA AG, Wangen, Switzerland). L-Baclofen was synthesized in the research laboratories of Novartis Pharma in Basel (30). Serum, culture media, and other solutions used for cell culture were from Life Technologies, Inc.
Site-directed Mutagenesis-Single amino acid substitution was carried out by the Quick Change strategy (Stratagene, La Jolla, CA) according to the manufacturer's instructions using pBSB5 as a template (15). For each mutagenesis, two complementary 30-mer oligonucleotides (sense and antisense; Genaxis Biotechnologie, Nimes, France) were designed to contain the desired mutation in their center. To allow a rapid screening of the mutated clones, the primers carried an additional silent mutation introducing (or removing) a restriction site. The presence of each mutation of interest and the absence of undesired ones were confirmed by DNA sequencing. Subsequently, a short fragment surrounding the mutation was subcloned in place of the corresponding wild-type fragment of pRKBR1a (15).
Cell Culture and Expression in HEK 293 Cells-Human embryonic kidney (HEK) 293 cells were cultured in Dulbecco's modified Eagle's medium (Life Technologies Inc.) supplemented with 10% fetal calf serum, penicillin, and streptomycin. Wild-type and mutated expression constructs were transfected in HEK 293 cells by electroporation as described previously (31). Electroporation was carried out in a total volume of 300 l with 10 ϫ 10 6 cells. For membrane preparation, 2 g of plasmid DNA containing either wild-type or the mutated GABA B 1a receptor coding sequences and 8 g of carrier DNA were used. After electroporation, the cells were plated on polyornithine-coated dishes.
Ligand Binding Assay-Ligand competition experiments were performed on membranes of HEK 293 cells prepared as follows. 24 h after transfection, the cells were washed and homogenized in Tris-Krebs buffer (20 mM Tris-Cl, pH 7.4, 118 mM NaCl, 5.6 mM glucose, 1.2 mM KH 2 PO 4 , 1.2 mM MgSO 4 , 4.7 mM KCl, 1.8 mM CaCl 2 ) and centrifuged for 20 min at 40,000 ϫ g. The pellet was resuspended in Tris-Krebs buffer and stored at Ϫ80°C. For ligand competition assays, thawed membranes (10 g of protein) were incubated with 0.1 nM 125 I-CGP64213 in the presence of unlabeled ligands at the indicated concentration. The incubation was terminated by filtration through GF/C Whatman glass fiber filters (Whatman International Ltd., Maidstone, UK). The concentration of 125 I-CGP64213 used in displacement experiments (0.1 nM) is approximately 10 times lower than the affinity of this radioligand on the native or GABA B R1 receptor, such that the IC 50 values measured for all unlabeled compounds are not significantly different from their affinity. Nonspecific binding was determined with 10 mM GABA. Displacement curves were fitted with the Kaleidagraph software (Abelbeck Software) using the equation y ϭ ((y max Ϫ y min )/(1 ϩ (x/IC 50 ) nH )) ϩ y min where IC 50 is the concentration of cold drug necessary to displace half of the specifically bound 125 I-CGP64213, and n H is the Hill number.
Determination of Inositol Phosphate (IP) Accumulation-HEK 293 cells were transfected as described above with either wild-type or mutated GABA B 1a receptor expressing plasmids (pRKBR1a, 2 g), GABA B 2 receptor (pCI-Neo-BR2, 2 g) (25), G qi9 (2 g) (32), and carrier DNA (4 g). Determination of IP accumulation in transfected cells was performed 15 h after transfection and after overnight metabolic labeling of transfected cells with [ 3 H]myoinositol (23.4 Ci/mol, PerkinElmer Life Sciences). After three washes with Krebs buffer (NaCl 146 mM, KCl 4.2 mM, MgCl 2 0.5 mM, glucose 0.1%, Hepes 20 mM, pH 7.4), the stimulation was conducted for 30 min in Krebs buffer containing 10 mM LiCl and the indicated concentration of agonist. The stimulation was stopped by replacement of the incubation medium with perchloric acid (5%) on ice, and the IPs were purified on Dowex AG 1-X8 (Bio-Rad) columns. Results are expressed as the amount of IP produced over the radioactivity present in the membranes. Concentration-response curves were fitted as described for the binding assay using the equation y ϭ ((y max Ϫ y min )/(1 ϩ (x/EC 50 ) nH ))ϩ y min where EC 50 is the agonist concentration necessary to produce half of the maximal response.
Biotinylation of Membrane Proteins and Western Blotting-Cell surface proteins of transfected HEK 293 cells were labeled with the membrane impermeant reagent Sulfo-NHS-Biotin (Pierce). Briefly, adherent cells were washed 3 times on ice with borate buffer (10 mM H 3 BO 3 , pH 8.8, 140 mM NaCl) and incubated with 0.2 mg/ml sulfo-NHS-biotin in borate buffer for 30 min. The reaction was stopped by washing followed by incubation in borate buffer with 100 mM NH 4 Cl for 10 min on ice. The cells were scraped in lysis buffer (Hepes 20 mM, pH 7.4, NaCl 100 mM, EDTA 5 mM), and the membranes were pelleted and solubilized in lysis buffer containing 1% Nonidet P-0, 0.5% sodium deoxycholate, and 0.1% SDS at a final concentration of 5 mg/ml. The lysate was centrifuged for 1 h at 100,000 ϫ g. The soluble fraction was incubated with streptavidin-coated agarose beads, overnight at 4°C. Bound proteins were analyzed using Tricine-SDS-polyacrylamide gel electrophoresis and immunoblotted using the polyclonal anti-GABA B 1 receptor antibody (33) and an ECL chemiluminescence system (Amersham Pharmacia Biotech). Analysis of wild-type or mutant receptor expression was conducted according to the same protocol.
Homology Modeling of Open and Closed Conformations and Docking of Baclofen-To generate the open form model of the GABA B 1 receptor extracellular domain, we modified the multiple alignment proposed in previous work (15) in order to take into account the secondary structure prediction of the GABA B 1 receptor extracellular domain, as determined using the PHD program, and the sequence of the recently cloned GABA B 2 receptor. Compared with our previous alignment, this resulted in a change from residue 467 to 486 containing helix ␣IX of LBP/LIVBP (18). The alignment of AmiC with LIVBP and LBP was deduced from the structural superposition of each lobe of the closed form of AmiC on the corresponding lobe of the open form of LIVBP (16).
The open form model of the GABA B 1 receptor extracellular domain was generated by the automated homology modeling tool MODELER 5.00 (InsightII version 980, MSI) (34) as described previously (15). The Only the latter has been used for the modeling experiments shown in this paper. The three-dimensional/one-dimensional compatibility scores for these models (35), as determined using the Profiles three-dimensional algorithm using a sequence window of 21 amino acids (InsightII version 980, MSI), is always positive and similar to that previously published (15). The global score is in the range of those determined for the refined three-dimensional structures of LIVBP, LBP, or AmiC determined from x-ray (data not shown). Most residues of our models were in allowed regions of the Ramachandran's map.
For "docking" experiments, baclofen was designed with a deprotonated carboxylic group and a protonated amino group. Initially, baclofen (R-enantiomer) was manually docked in the closed form model of GABA B 1 receptor, the carboxylate moiety closed to the side chain of Ser 246 and its ␥-amino group facing the side chain of Asp 471 . In this position, the chlorophenyl group is pointing toward lobe II and is in the proximity of Tyr 366 . In order to suppress the steric hindrances and geometric inconsistencies, the ligand-protein complex was submitted to molecular mechanics calculations using the Discover 3.00 calculation engine with the CFF force field (Insight II version 980, MSI). The non-bond cut-off method and the dielectric constant were respectively set up as cell multipole and distance-dependent (⑀ ϭ r). Initially, energy minimization were performed using a Steepest Descent algorithm (until the maximum derivative was less than 2 kcal/mol/Å) followed by a conjugated gradient algorithm (until the maximum derivative was less than 0.01 kcal/mol/Å), whereas the C␣ trace was tethered with a quadratic potential. Then molecular dynamics was applied to the minimized system at constant volume and temperature (298 K). The integration time step was set up to 1 fs. During the dynamics, the force constant value of the quadratic potential was reduced from 100 to 60, 30, 20, 10, and 0 every 40 ps. A snapshot of the system was saved every 400 fs. Once the system was equilibrated, the coordinates of 20 snapshots were averaged and resubmitted to the previously described minimization protocol with no C␣ restraints. Contacts between the ligand and the protein were subsequently analyzed using the web interface of the WHATIF program.

Identification of Putative Structural Regions of the GABA B 1
Receptor Involved in Ligand Binding-Several splice variants have been identified for the GABA B 1 receptor which all possess an identical ligand binding domain but differ in their extreme N-terminal sequence (24) or in their heptahelix regions (36,37). All our studies have been performed with the rat GABA B 1a receptor (the first Met being residue number 1). A three-dimensional model of the GABA B 1 binding domain has been constructed based on the structure of the open forms of LBP and LIVBP (Fig. 1a). According to this three-dimensional model and the known structural elements involved in leucine binding in LIVBP, and in amide binding in AmiC (see Fig. 1), the GABA-binding site would be formed by the following: (a) the loop between strand ␤C and helix ␣III (loop ␤C-␣III according to the nomenclature of the secondary structural elements of LIVBP proposed by Sack et al. (18), residues 243-250), (b) the N-terminal portion of the loop ␤D-␤E (residues 266 -272), and (c) the N termini of helices ␣I (residues 184 -192) and ␣IX (residues 463-473) (Fig. 1). We therefore undertook a systematic analysis, by site-directed mutagenesis, of all residues that can interact with ligands within these regions (see Fig. 1b). Each mutant was analyzed for its ability to bind 125 I-CGP64213 and to be activated by the agonists GABA and baclofen (Fig. 2).
Occupancy of the GABA B 1 Subunit by GABA or Baclofen Plays a Pivotal Role in the Activation of the Heteromeric Receptor-We have previously reported that several mutations of the GABA B 1 subunit either increase or decrease the potency of GABA and baclofen in displacing 125 I-CGP64213 bound to the GABA B 1 subunit expressed alone (see Table I). The effect of these mutations, as well as additional ones generated during this study, on the potency of GABA and baclofen at the heteromeric GABA B receptor was analyzed after co-expression of the different GABA B 1 mutants with the wild-type GABA B 2 subunit in HEK 293 cells. The activation of the GABA B heteromeric receptor was analyzed by measuring the stimulation of phospholipase C via the chimeric G-protein G qi9 (32,38). The potencies (EC 50 values) of both GABA and baclofen were measured on the different heteromers and compared with the binding affinity (IC 50 values which, under our experimental conditions, are similar to K i ) of both GABA and baclofen as determined by displacement of bound 125 I-CGP64213 (see Table I) on GABA B 1 mutants expressed alone. A good correlation was found between GABA (Fig. 3) and baclofen (data not shown) binding affinity values (IC 50 ) on most GABA B 1 mutants, and the EC 50 values were measured with our functional assay on the mutant heteromeric receptors. These data show that agonist binding at the GABA B 1 receptor is critical for the activation of the heteromeric GABA B receptor and validates the use of such a functional assay to further analyze the molecular determinants of agonist binding at the GABA B 1 receptor for which no 125 I-CGP64213 binding could be measured.
Ser 246 of GABA B 1a Receptor Is Critical for Agonist Binding and Activity-The mutation of a residue interacting with GABA in the GABA B 1 subunit is expected 1) to either largely decrease the potency of GABA in displacing 125 I-CGP64213 binding or to suppress 125 I-CGP64213 binding and 2) to largely decrease the potency of agonists in activating the receptor. Within the loops ␤C-␣III and ␤D-␤E, the mutations of Ser 246 , Ser 265 , and Tyr 266 were found to largely decrease 125 I-CGP64213 binding (Fig. 4), but only the mutation of Ser 246 affected the potency of GABA and baclofen at the GABA B receptor (Fig. 1b and Table I). As shown in Fig. 4, 125 I-CGP64213 did not bind to the S246A mutant although correctly expressed in HEK 293 cells. Both GABA and baclofen were able to activate the heteromeric receptor containing the S246A mutated GABA B 1 subunit, indicating a correct expression, folding, and dimerization of the receptor. However, the EC 50 values of both agonists were increased by a factor of 1000 compared with those determined with the wild-type receptor (Fig. 5). Interestingly, Ser 246 aligns with Ser 79 of LIVBP which interacts with ligands ( Fig. 1b) (18) thus suggesting that the hydroxyl group of Ser 246 directly interacts with the agonists GABA and baclofen. In agreement with this proposal, mutation of this residue into the more bulky Asn residue, or into Pro which may affect the topology of this loop, suppressed both 125 I-CGP64213 binding and agonist activation of the receptor (Table I and Fig.  5). Furthermore, even the replacement of Ser 246 by a Thr residue resulted in a large decrease in agonist potency (Table I and Fig. 5), suggesting that the orientation of the hydroxyl group is critical for a high agonist potency. All these latter mutants were found to be correctly expressed and targeted to the plasma membrane when co-expressed with the wild-type GABA B 2 subunit (data not shown).
Exploring Helix I and IX, Asp 471 Is a Critical Residue for Agonist Action-We applied the same strategy to identify residues possibly interacting with agonists within helices I and IX. The mutations of Phe 463 , Tyr 470 , and Asp 471 were found to largely decrease 125 I-CGP64213 binding (Fig. 4), but only the mutation of Asp 471 affected the potency of GABA and baclofen to activate the GABA B receptor (Fig. 1b and Table I). As shown in Fig. 4, no 125 I-CGP64213 binding could be measured on membranes expressing the Asp 471 mutant receptor. Moreover, co-expression of the mutant D471A with the wild-type GABA B 2 receptor did not lead to a functional response upon application of GABA or baclofen (Fig. 6, a and b). Western blots clearly show that the lack of function of this mutant receptor was not due to a lack of expression (Fig. 3b). Moreover, the mutant D471A was correctly inserted in the plasma membrane when co-expressed with the GABA B 2 subunit (Fig. 6c), as shown by the large amount of biotinylated subunits after exposure of the intact cells to the non-permeant protein reagent, sulfo-NHSbiotin. Taken together, these data reveal that Asp 471 is a critical residue for agonist binding in GABA B 1 receptors. In agreement with this proposal, the conservative mutation of this Asp into Glu was sufficient to dramatically reduce 125 I-CGP64213 binding. However, a significant increase in IP formation could be detected upon application of either GABA or baclofen to cells co-expressing D471E and the wild-type GABA B 2 receptor, showing a very partial recovery of the agonist action at this mutant receptor (Fig. 6). This partial recovery is likely to have been underestimated since the expression level of the D471E mutant was lower than that of the wild-type or D471A mutant (Fig. 4b).
Tyr 366 in Lobe II Decreased GABA and Baclofen Affinity and Converts Baclofen into an Antagonist-In PBPs, or related proteins, which have been crystallized in a closed form complexed with a ligand, residues from both lobes contact the ligand (16, 19, 39 -41). We therefore examined the role of residues within lobe II in ligand binding and agonist activity. We focused our efforts on residues within the loops ␤F-␣V and ␤H-␣VII because they both face the putative GABA-binding site, and a residue interacting with amide in AmiC is located within one of these loops (Fig. 1a and Fig. 8, a and c). None of the alanine substitutions generated in these loops prevent 125 I-CGP64213 binding (Fig. 1b, Table I, and Fig. 3a). The IC 50 values of both GABA and baclofen for displacement of 125 I-CGP64213 binding were similar to those of the wild-type receptor for all these mutants except Y366A. For this mutant, the IC 50 for values of these two agonists were increased by a factor 50 -100 (Table I and Fig. 7a), whereas displacement by cold CGP64213 was not affected (1.1 Ϯ 0.1 nM for the wild-type and 1.0 Ϯ 0.4 nM for Y366A, n ϭ 3). Although GABA and baclofen displayed a similar affinity for this mutant (Fig. 7a), the Y366A mutated subunit formed a functional receptor activated by GABA but not by baclofen (Fig. 7b). In cells co-expressing this receptor and GABA B 2, GABA stimulated IP formation by a factor 5-10, as observed with the wild-type receptor, indicating a correct expression, targeting to the plasma membrane, and association with the wild-type GABA B 2 subunit. However, no significant formation of IP could be measured on this mutant receptor with baclofen (Fig. 7b). Accordingly, if GABA and baclofen bind to the same site in GABA B 1, baclofen is expected to act as an antagonist at this mutant receptor (or as a very partial agonist). Indeed, as shown in Fig. 7c, 5 mM baclofen was found to inhibit the action of GABA at this mutant receptor in a competitive manner. Thus, the mutation Y366A converts baclofen into a competitive antagonist.
Molecular Modeling of the GABA B 1 Extracellular Domain in a Closed Form and Docking of Baclofen-In order to rationalize the effect of the mutations described above, the positions of the mutated residues were visualized in the three-dimensional model of the extracellular domain of the GABA B 1 receptor described above (Fig. 1a). As shown in Fig. 8, a and c, Ser 246 and Asp 471 are correctly located to interact with GABA in lobe I. The distance between the hydroxyl and carboxylic groups of these two residues (10 Å) is compatible with this possibility; however, the side chain of Tyr 366 in loop ␤H-␣VII of lobe II, although facing the putative GABA-binding pocket in lobe I, is located more than 15 Å from these residues. As such, this open form model could not easily explain the binding and functional properties of the Y366A mutant.
We therefore undertook the construction of a three-dimensional model for a closed form of the GABA B 1 binding domain using the coordinates of a three-dimensional model of the closed form of LIVBP (42) kindly provided by F. Quiocho, and the coordinates of the closed liganded form of AmiC (16). In the closed form model, Tyr 366 comes in close proximity to Ser 246 and Asp 471 . Baclofen, which corresponds to a chlorophenyl derivative of GABA (see Fig. 2), has been manually docked in this three-dimensional model, with its carboxyl group facing Ser 246 and its amino group facing Asp 471 . The model has then been subjected to simulation protocols as described under "Experimental Procedures." In this model, the carboxyl group of baclofen forms H bonds with the hydroxyl group of Ser 246 (from lobe I) and those of Tyr 366 (from lobe II) (distance O-O Ͻ3 Å), and its amino group forms a salt bridge with the carboxylate of Asp 471 (distance N-O Ͻ3 Å) (Fig. 8d). This latter interaction is stabilized by the aromatic ring of Tyr 266 and an H bond with the backbone (carbonyl of Pro 467 ). The chlorophenyl group of baclofen was found to fit into a pocket lined by Tyr 366 and Trp 394 from lobe II and Leu 468 from lobe I. DISCUSSION Our mutagenesis and modeling study has identified a set of residues in the GABA B 1a subunit that are critical for agonist binding and activation of the heteromeric GABA B receptor. Our data suggest that two residues in lobe I, Ser 246 and Asp 471 , and one residue in lobe II interact with GABA and baclofen. Interestingly, the mutation of Tyr 366 into Ala not only decreases GABA and baclofen affinity but also converts baclofen into an antagonist.
We have previously reported that Ser 246 is critical for 125 I-CGP64213 binding at GABA B 1a receptors (15). By using a functional assay, we show here that this residue plays a critical role in both GABA and baclofen action, their apparent affinity being decreased by a factor 1000 when Ser 246 is mutated into Ala or even Thr. In accordance with the possible importance of this residue, a search within the Drosophila and Caenorhabdi-  Table I. A similar correlation was also observed with baclofen (data not shown). binding (IC 50 in M) and receptor activation (EC 50 in M) Mutants in boldface are those we proposed to be critical for agonist binding and action. NB, specific 125 I-CGP64213 is less than 10% that measured on membranes expressing the wild-type (WT) receptor. NE, no significant stimulation of IP formation with 1 mM of the agonist. tis elegans genomes for GABA B 1-like receptors shows that this residue is conserved through evolution (43). Moreover, this residue aligns with Ser 79 of LBP and LIVBP, the hydroxyl group of which forms H bonds with the ␣-carboxylic group of leucine (18) (Fig. 9). Ser 246 also aligns with Ser 85 of AmiC which interacts with acetamide ( Fig. 1) (16). This Ser residue is also conserved in all mGlu receptors and has been proposed to form H bonds with the ␣-carboxylic group of glutamate ( Fig. 9) (13, 14). Accordingly we propose that Ser 246 of GABA B 1a also contacts via an H bond the carboxylic group of GABA and baclofen. This proposal is consistent with both our modeling studies and the identification of Asp 471 as possibly forming an ionic interaction with the amino group of these GABA B agonists (see below). In addition to this Ser residue, the hydroxyl group of a Thr 102 HEK 293 cells (b). a, 10 g of membrane proteins were used, and the specific binding of 125 I-CGP64213 was measured using 1 nM of this radioligand. b, 10 g of membrane proteins were used, and the GABA B 1 receptor protein was detected using an antibody directed against its C terminus. Results are from one experiment representative of at least two others. WT, wild type. in LIVBP (Fig. 1b) forms H bonds with the ␣-amino group of leucine (18). A conserved Thr has been proposed to play the same role in mGlu receptors (13,14) (Fig. 9). In GABA B 1a, this Thr residue aligns with Ser 269 , which does not play a critical role in GABA and baclofen binding. Interestingly, neither GABA nor baclofen possesses an ␣-amino acid moiety (Fig. 9).

FIG. 4. Effect of some mutations on the specific binding of 125 I-CGP64213 (a) and on the expression level of the protein after transient expression in
Indeed we previously reported that Ser 269 was involved in the effect of Ca 2ϩ on this receptor subtype (43).
Among all other mutated residues from lobe I, only Asp 471 was found to play a critical role in the action of GABA and baclofen. Mutation of this residue into Ala suppresses antagonist binding as well as GABA-and baclofen-induced activation of the receptor. This total loss of function did not result from a lack of expression of this mutant protein. Moreover, the D471A mutant was found to be inserted correctly into the plasma membrane when co-expressed with the GABA B 2 receptor. Since GABA B 2 is required for the plasma membrane localization of the GABA B 1 receptors (28,44), these data also indicate that the mutation of Asp 471 does not affect the formation of the heterodimer. In agreement with such an important role of Asp 471 in agonist binding, this residue is conserved in Drosophila and C. elegans GABA B 1-like receptors (data not shown). Moreover, our modeling studies revealed that the position of this residue allows its side chain to interact correctly with the amino group of GABA. Indeed, this residue is located at the bottom of the binding pocket and points toward a hydrophobic environment that lacks any polar residue in LIVBP and constitutes the binding pocket for the side chain of leucine (18). In this region in mGlu receptors, the side chain of an Arg (Arg 78 ) points toward the binding pocket ( Fig. 9) and has been proposed to interact with the ␥-carboxylic group of glutamate in mGlu 4 (14) and mGlu 1 (45) receptors.
Taken together, our data show that leucine, glutamate, and GABA interact in a similar binding pocket in lobe I of LIVBP, mGlu, and GABA B 1 receptors, respectively (Fig. 9). They also highlight how the same binding pocket in lobe I has evolved within these LIVBP-like proteins to specifically recognize different but related molecules (Fig. 9).
Our study also identified 4 residues within lobe I, Ser 265 , Tyr 266 , Phe 463 , and Tyr 470 , the mutation of which dramatically decreases 125 I-CGP64213 binding but does not change the properties and apparent affinity of both GABA and baclofen. Although the lack of significant binding to the S265A, Y266F, and Y470F mutants may result from their low expression level, this cannot be the case for the F463A and Y470A mutants (data not shown and Fig. 4). One may therefore propose that these two residues are required for the binding of this large GABA B antagonist that possesses two aromatic moieties (Fig. 2). However, additional experiments are required to characterize better the role of these residues in GABA B antagonists binding and action.
The data discussed above demonstrate the pivotal role of lobe I in ligand binding to the GABA B 1 receptor and in the agonist activation of the heteromeric receptor. However, our data also revealed that the mutation of Tyr 366 from lobe II largely decreases the affinity of both GABA and baclofen. This mutant subunit is correctly expressed and targeted to the plasma membrane when associated with the wild-type GABA B 2 subunit, as shown by its activation by GABA which is to an extent similar to that obtained with the wild-type receptor. This large decrease in affinity therefore suggests that Tyr 366 has either direct or indirect contact with these agonists. Such an interaction does not appear to be possible in our three-dimensional model for an open form of the GABA B 1 binding domain. However, in a closed form model, the hydroxyphenyl of Tyr 366 points toward the GABA-binding site in lobe I, and its oxygen forms an H bond with the carboxylic group of baclofen. Further experiments are required to validate this possibility. However, these data already show that residues in lobe II affect the affinity of agonists, in agreement with the observation that residues from both lobes contact the ligands in many PBP-like proteins (16, 19, 39 -41). Interestingly, this mutation also prevents baclofen from activating the heteromeric receptor and converts it into a competitive antagonist. In our three-dimensional model, the chlorophenyl group of baclofen points toward a pocket lined by the hydroxyphenyl group of Tyr 366 . It is therefore possible that the replacement of this Tyr by an Ala changes the position of surrounding residues such that the chlorophenyl group in no longer accepted in a closed/active conformation of the binding domain, therefore converting baclofen into an antagonist. Although additional work is required to demonstrate this hypothesis, it is interesting to note that a mutation in lobe II, lining the binding pocket of AmiC, converts the bulky butyramide from an antagonist to an agonist (46).
The GABA B receptor requires the presence of two subunits, GABA B 1 and GABA B 2, for its full activity. Although one role of GABA B 2 is to allow the correct insertion of GABA B 1 in the plasma membrane (28,44), other roles in receptor functioning have not yet been characterized. Our data show a correlation between the affinity of both GABA and baclofen measured in binding studies on wild-type or mutant GABA B 1 receptors, and their EC 50 values were measured in a functional assay after co-expression with GABA B 2 receptor. Moreover, the mutation of Asp 471 , a residue that is likely to play a critical role in GABA binding according to our modeling studies, suppresses the function of the heteromer. Finally, a single mutation in the GABA B 1 subunit is sufficient to convert the selective heteromeric GABA B receptor agonist baclofen into an antagonist. Taken together, and in agreement with recent additional data (47), these results add further strength to the proposal that GABA binding on the GABA B 1 subunit is required and is the main determinant for the activation of the heteromer. Comparison of the putative binding pocket of GABA B 1 and GABA B 2 reveals that although Asp 471 of GABA B 1 is conserved in GABA B 2, Tyr 366 and Ser 246 are replaced by Asp and Pro in GABA B 2, respectively (see Fig. 1b). The mutation of Ser 246 into Pro in GABA B 1 is sufficient to suppress the action of GABA on the heteromer. However, GABA and baclofen have been shown to activate occasionally the GABA B 2 receptor expressed alone (25,48), suggesting that the GABA B 2 receptor is able to bind these two agonists. Our data indicate that if GABA and baclofen bind to the GABA B 2 subunit, they have to do it in a different way from GABA B 1.
In conclusion, our data strongly support the importance of lobe I for binding properties of agonists in family 3 heptahelix receptors. They also reveal that residues in lobe II can be critical for the agonist property of family 3 receptor ligands. This is consistent with the proposal that the closure of the two lobes (the so-called Venus fly-trap mechanism of action (19)) constitutes a key step in family 3 receptor activation.