Molecular modeling and mutagenesis of the ligand-binding pocket of the mGlu3 subtype of metabotropic glutamate receptor

Authors


Address correspondence and reprint requests to David R. Hampson, University of Toronto, 19 Russell St., Toronto, Ontario, Canada M5S 2S2. E-mail: d.hampson@utoronto.ca

Abstract

A homology model of the extracellular domain of the mGlu3 subtype of metabotropic glutamate (mGlu) receptor was generated and tested using site-directed mutagenesis, a radioligand-binding assay using the Group II selective agonist (2S,2′R,3′R)-2-(2′,3′-[3H]dicarboxycyclopropyl) glycine ([3H]DCG-IV), and in a fluorescence-based functional assay in live transiently transfected human embryonic kidney cells. Ten of the 12 mGlu3 mutants (R64A, R68A, Y150A, S151A, T174A, D194A, Y222A, R277A, D301A and K389) showed either no binding or a 90% or greater loss of specific [3H]DCG-IV binding. Several analogous mutations in mGlu2 supported the results obtained with mGlu3. These results demonstrate that the binding of [3H]DCG-IV to mGlu3 is exceptionally sensitive to mutagenesis-induced perturbations. In silico docking of DCG-IV into the agonist binding pocket of mGlu3 facilitated the interpretation the mutagenesis results. Tyrosines 150 and 222, and arginine 277 show close contacts with the third carboxylic acid group in DCG-IV, which is not present in glutamate or (2S,1′S,2′S)-2-(carboxycyclopropyl)glycine (L-CCG-I). Mutation of these three amino acids to alanine resulted in a near complete loss of receptor activation by DCG-IV and retention of near wild-type affinity for L-CCG-I. It is proposed that hydrogen bonding between this carboxylate and tyrosines 150 and 222 and arginine 277 provide a partial explanation for the high affinity and Group II selectivity of DCG-IV. These findings define the essential features of the ligand-binding pocket of mGlu3 and, together with other recent studies on mGlu receptors, provide new opportunities for structure-based drug design.

Abbreviations used
AFBD

2-amino-3-fluorobicyclo[3.1.0]hexane-2,6-dicarboxylic acid

DCG-IV

(2S,2′R,3′R)-2-(2′,3′-dicarboxycyclopropyl)glycine

HEK

human embryonic kidney

L-AP4

l-amino-4-phosphonobutyric acid

L-CCG-I

(2S,1′S,2′S)-2-(carboxycyclopropyl)glycine

mGlu

metabotropic glutamate receptor

The metabotropic glutamate (mGlu) receptors are key modulators of excitatory and inhibitory neurotransmission in the central and peripheral nervous systems. All eight receptor subtypes are expressed in neurons, whereas the mGlu3 and mGlu5 receptors are present in both neurons and astrocytes (Petralia et al. 1996; Bruno et al. 1998; Ferraguti et al. 2001; Tamura et al. 2001). The family of eight mGlu receptors has been subdivided into three groups (Group I, II, and III) based on sequence similarity and pharmacological properties. The mGlu2 and mGlu3 receptor subtypes, which constitute the Group II mGlu receptors, share 66% amino acid sequence identity in the extracellular domain and possess very similar pharmacological profiles (Conn and Pin 1997; Pin et al. 1999; Schoepp et al. 1999; Bräuner-Osborne et al. 2000; Parmentier et al. 2000).

Examples of agonists with high affinity and selectivity for Group II receptors include (2S,2′R,3′R)-2-(2′,3′-dicarboxycyclopropyl)glycine (DCG-IV) (Hayashi et al. 1993; Ishida et al. 1993) and LY354740 (Schoepp et al. 1997; Schaffhauser et al. 1998). DCG-IV and LY354740 are rigid analogs of glutamate with ring structures incorporated into the side chains (Fig. 1); both compounds have substantially higher affinity than l-glutamate for mGlu2 and mGlu3. The newer high affinity Group II selective compounds have been instrumental in establishing the potential of Group II agonists as neuroprotective agents in a variety of disease models. For example, Group II agonists have been shown to protect neurons in models of neurotoxicity (Bruno et al. 1995; Allen et al. 1999) and global ischemia (Bond et al. 2000), and have salutary effects in animal models of anxiety (Helton et al. 1998; Tatarczynska et al. 2001), schizophrenia (Moghaddam and Adams 1998), and Parkinson's disease (Dawson et al. 2000; Breysse et al. 2002).

Figure 1.

Comparison of the structures of glutamate, L-AP4, L-CCG-I and the Group II agonists DCG-IV, LY354740, 2-amino-6-fluorbicyclo[3.1.0]hexane-2,6,-dicarboxylic acid (AFBD), and MGS0028. The compounds are shown in their correct relative and absolute stereochemistries and the C2′ and C3′ carbons atoms are labeled in DCG-IV.

The agonist binding site in the mGlu receptors is contained within the large extracellular amino-terminal domain of the protein (Okamoto et al. 1998; Han and Hampson 1999; Peltekova et al. 2000; Selkirk et al. 2002). The ligand-binding domain consists of two lobes that close upon agonist binding to activate the receptor and initiate the propagation of intracellular signals (Parmentier et al. 2002). Recently, the molecular architecture surrounding the binding pocket has been revealed by studies using molecular modeling and mutagenesis, and by the elucidation of the X-ray structure of the rat mGlu1 receptor subtype (Kunishima et al. 2000). Molecular modeling studies carried out in conjunction with mutagenesis experiments have been conducted on the Group I receptor, mGlu1 (Sato et al. 2003), the Group II receptor mGlu2 (Malherbe et al. 2001), and the Group III receptors mGlu4 (Hampson et al. 1999; Rosemond et al. 2002) and mGlu8 (Bessis et al. 2002). With the exception of the earlier study by Hampson et al. (1999), which used as template the leucine–isoleucine–valine binding protein, these studies were based on homology models that utilized the molecular coordinates of the crystal structure of mGlu1, and [3H]quisqualic acid (mGlu1), [3H]LY354740 (mGlu2) and [3H]l-amino-4-phosphonobutyric acid ([3H]L-AP4) (mGlu4) as probes in radioligand-binding experiments. In the present study, we extend the investigation of the molecular pharmacology of mGlu receptor subtypes by probing the ligand-binding pocket of mGlu3 using molecular modeling, site-directed mutagenesis, and the high affinity Group II agonist [3H]DCG-IV.

Materials and methods

Materials

l-Glutamate and (2S,2′R,3′R)-2-(2′,3′-[3H]dicarboxycyclopropyl)glycine ([3H]DCG-IV) (specific activity 18.3 Ci/mmol) were purchased from Tocris Cookson Inc. (Bristol, UK). All culture reagents were purchased from Invitrogen Corp. (Carlsbad, CA, USA).

Expression constructs and site-directed mutagenesis

For the expression of wild-type rat mGlu2 and mGlu3 in human embryonic kidney cells (HEK-293-TSA-201), the EcoRI digested fragment of mGlu2 or mGlu3 in the pBluescript SK phagemid (Tanabe et al. 1992) was subcloned into the EcoRI site of mammalian expression vector pcDNA3.0 (Invitrogen). For the construction of epitope-tagged mGlu3 receptors, a c-myc-tag (EQKLISEEDLDP) was inserted just downstream of the predicted signal peptide using an overlapping PCR strategy. Two separate PCRs were performed using mGlu3-pcDNA3.0 as a template with two pairs of primers (Pair I, m3-BglII-F: 5′-GAAGATCTCCCGATCCCCTATGGTC-3′, m3-cmyc-BclI-R: 5′-TGGGTCCAAGTCTTCTTCGCTGATCAGCTTTTGTTCGTTGTGATCTCCTAAAGAGAGTAAAAATCC-3′; Pair II, m3-cmyc-BclI-F: 5′-GAACAAAAGCTGATCAGCGAAGAAGACTTGGACCCATTTATGAGGAGGGAAATTAAAATAGAAGG-3′, m3-BglII-R: 5′-GAAGATGTCGGCCATGGCTTTGGCC-3′). Primers m3-cmyc-BclI-F and -R contain a BclI restriction enzyme site and the sequence encoding the c-myc epitope. The PCR products were mixed, and a second PCR was performed with primers m3-BglII-F and -R. The amplification products were cut by BglII and ligated into BglII digested mGlu3-pcDNA3.0 for the final expression construct.

To facilitate the subcloning of the mGlu3 mutants, a silent mutation incorporating a BamHI restriction enzyme site was created at serine 510 of the c-myc-tagged mGlu3 using the QuikChange site-directed mutagenesis kit (Stratagene Corp., La Jolla, CA, USA). Briefly, a 1.75-kilobase KpnI–SmaI fragment of the tagged mGlu3 was subcloned into the pBluescript SK vector to make a mGlu3 KpnI/SmaI cassette. Using this cassette as template, a PCR was carried out using pfu Turbo polymerase (Stratagene) and two complementary 32-mer oligonucleotide primers (sense and anti-sense) containing the site of mutation. The PCR was carried out for 16 cycles followed by the addition of 1 µL of DpnI to digest the non-mutated template, and the resultant mutated DNA was transformed into DH5αE. coli-competent cells. Both the mGlu3 KpnI/SmaI cassette and mGlu3 pcDNA3.0 plasmid were digested by KpnI–SmaI, the 1.75-kb KpnI/SmaI fragment from cassette and a 2.65-kb SmaI fragment from mGlu3 pcDNA3.0 were subcloned into the KpnI–SmaI sites of pcDNA3.0 and the resultant construct containing the BamHI site was confirmed by restriction enzyme analysis. Cassettes for mutagenesis were made using the 1.7-kb BamHI fragment of c-myc-tagged mGlu3 and a 2.1-kb KpnI fragment of mGlu2 in pBluescript SK. All mutants of c-myc-tagged mGlu3 and untagged mGlu2 were made in these cassettes using the QuikChange mutagenesis procedure. All of the mutants were sequenced and subcloned back into the pcDNA 3.0 mammalian expression vector.

Cell culture transfection

For ligand-binding experiments and immunoblots, HEK-293-TSA-201 cells were cultured in modified Eagle's medium with 6% fetal bovine serum and antibiotics. Transient transfections were conducted using a calcium phosphate protocol as described previously (Han and Hampson 1999). In the case of the functional analyses, HEK cells (American Type Culture Collection # CRL 1573) were transiently transfected in 6-well plates using 2 µg of wild-type mGlu3 cDNA or mutant cDNA plus 2 µg of Gα15 cDNA using the Fugene 6 reagent (Roche Biochemicals, Indianapolis, IN, USA) as instructed by the manufacturer. All experiments were conducted on cells or membranes collected 48 h after transfection.

Membrane preparation and radioligand binding

The membrane preparation and the [3H]DCG-IV radioligand-binding assays were performed as described previously for analysis of mGlu2 expressed in CHO cells (Cartmell et al. 1998) with several minor modifications. Forty-eight hours after transfection, HEK cells were washed three times with phosphate-buffered saline, collected by centrifugation (3840 × g, 15 min, 4°C) and frozen at − 70°C. The pellet was suspended in 15 mL of ice-cold lysis buffer I (20 mm HEPES, 10 mm EDTA, pH 7.4) and homogenized with a Polytron. After centrifugation (48 000 × g, 20 min, 4°C), the pellet was washed once by suspension in 20 mL of lysis buffer I and subjected to centrifugation at 48 000 × g for 20 min at 4°C. The membranes were resuspended and homogenized in 10 mL of lysis buffer II (20 mm HEPES, 0.1 mm EDTA, pH 7.4) followed by an additional 10 mL of lysis buffer II; the membranes were recentrifuged at 48 000 × g for 20 min at 4°C and frozen at − 70°C. The thawed membranes were resuspended in 10 mL assay buffer (50 mm Tris-HCl, 2 mm MgCl2 and 2 mm CaCl2, pH 7.0), homogenized, centrifuged (48 000 × g, 20 min, 4°C) and homogenized again in 3 mL of assay buffer. Membrane protein concentrations were determined using the Bio-Rad kit and bovine serum albumin as the standard.

For radioligand binding, the membranes were diluted in assay buffer to a final concentration of 0.25 mg/mL and 50 µg of membrane protein was used in a final assay volume of 250 µL. All binding assays were performed at room temperature using 20 nm[3H]DCG-IV and 500 µm l-glutamate was used to define non-specific binding. After 45 min incubation, membranes were collected by centrifugation (14 000 × g, 4 min), washed with 500 µL of assay buffer and solubilized overnight in 500 µL of 1 m NaOH. The samples were counted on a Packard TriCarb liquid scintillation counter (Meriden, CT, USA). The data were analyzed using Graphpad Prism 3.0 (San Diego, CA, USA).

Functional analysis of mGlu3 mutants

Twenty-four hours after transfection, HEK cells were plated in polyornithine pre-treated 96-well microtiter plates (Costar black microtiter plates) at 100 000 cells/100 µL of Dulbecco's modified Eagle's medium. Forty hours post-transfection, cells were washed once with calcium assay buffer (20 mm HEPES, pH 7.4, 146 mm NaCl, 5 mm KCl, 1 mm MgCl2, 1 mm CaCl2, 1 mg/mL bovine serum albumin and 1 mg/mL glucose) and incubated at 37°C in 100 µL of assay buffer. After a 2 h incubation, the assay buffer was changed and the cells were incubated for another 1 h at 37°C. The assay buffer was removed and 30 µL of assay buffer containing 6 µm of Fluo-4 (Molecular Probes Inc., Eugene, OR, USA) was added into each well; the cells were incubated for 1 h at room temperature in the dark. The cells were washed three times with assay buffer and then incubated in 150 µL of assay buffer for 30 min at room temperature in the dark. DCG-IV and (2S,1′S,2′S)-2-(carboxycyclopropyl)glycine (L-CCG-I) were dissolved in assay buffer and the responses were recorded on a FLEXstation benchtop scanning fluorometer (Molecular Devices Corp., Sunnyvale CA, USA) at room temperature with settings of 485 nm for excitation and 525 nm for emission. The Graphpad Prism 3.0 software was used to plot fluorescence intensities and calculate the EC50 values.

Immunoblotting and immunocytochemistry

The procedures for immunoblotting were carried out as described previously (Hampson et al. 1999). Electrophoresis samples containing 100 mm dithiothreitol were incubated at 37°C for 15 min prior to gel electrophoresis. The immunoblots were labeled with either an antibody raised against the carboxy terminus of mGlu2/3 (Chemicon Corp., Temecula, CA, USA), or an anti-peptide antibody raised against a sequence in the amino-terminal domain of rat mGlu3 (GRINEDRGIQRLEAML) using methodology described previously (Naples and Hampson 2001).

For immunocytochemical analyses, HEK cells were washed with phosphate-buffered saline twice at 48 h post-transfection and fixed with phosphate-buffered saline containing 4% paraformaldehyde for 12 min at 25°C. The cells were rinsed twice by phosphate-buffered saline and blocked in phosphate-buffered saline containing 5% goat serum with or without 0.2% Triton X-100 for 30 min at 25°C. The cells were subsequently incubated for 1 h at 25°C with the anti-mGlu2/3 or the mouse anti-c-myc monoclonal antibody at a 1 : 2000 dilution in phosphate-buffered saline containing 5% goat serum with or without 0.25% Triton X-100. After removing the primary antibody, the cells were washed five times for 5 min with phosphate-buffered saline and then incubated for 1 h at 25°C with Alexa Fluor 594 (Molecular Probes Inc.) goat anti-rabbit (mGlu2/3 carboxy terminal antibody) or Alexa Fluor 594 anti-mouse for cells incubated with the c-myc monoclonal antibody at a 1 : 1000 (Upstate Inc., Lake Placid, NY, USA) dilution in phosphate-buffered saline containing 5% goat serum with or without 0.25% Triton X-100. The cells were then washed five times for 5 min each with phosphate-buffered saline, mounted with mounting medium (ThermoShandon, Pittsburgh, PA, USA) and photographed with a Zeiss Axiovert 135 TV microscope equipped with a 590-nm excitation and a 600-nm high pass emission filter at a magnification of 400x using Kodak Tmax 400 film.

Molecular modeling

A homology model of the closed form of the extracellular domain of rat mGlu3 was generated using the X-ray crystal structure of the extracellular domain of rat mGlu1 as a template (Kunishima et al. 2000; PDB coordinates, 1EWK.pdb). The pair-wise sequence alignment between the rat mGlu1 and mGlu3 sequences was produced using the CLUSTALW program (Higgins et al. 1994; version 1.6). The basic pair-wise alignment between mGlu1 and mGlu3 used in the homology modeling is similar to the multiple sequence alignment shown in Fig. 2. All of the critical residues in and around the binding pocket have the same alignment as shown in Fig. 2. However, a few residues present in the loop structures of mGlu3 and outside the binding pocket were manipulated manually to be consistent with the mGlu1 X-ray crystal structure. The MODELER module available in the INSIGHT II molecular modeling software (Accelrys Inc., San Diego, CA, USA) was used to generate the homology model. In the generation of the homology model, the bound glutamate was not considered. The mGlu3 model has no intramolecular steric conflicts and the backbone torsion angles are within the allowed regions in the Ramachandaran phi-psi plot (Ramachandran et al. 1963). The position and orientation of the zwitterionic glutamate was taken from the rat mGlu1 crystal structure. However, some of the side chains were rotated in order to reduce the steric clashes between the glutamate ligand and the side chains in the pocket. The coordinates for the whole complex were energy minimized using the DISCOVER cff91 force field by constraining the backbone C-alpha atoms of all protein residues. After the initial minimization, the coordinates of the entire mGlu3–glutamate complex were energy minimized without any constraints. In all the energy minimization calculations reported in this study, a distance-dependent dielectric constant was used to calculate the electrostatic interactions. The energy minimization was carried out until the maximum derivative was less than 0.09 kcals per angstrom.

Figure 2.

Amino acid sequence alignment of a portion of the amino-terminal domains of mGlu1, 2, 3 and 4. The alignment begins at the first fully conserved glycine residue in the mGlu receptor family (glycine 30 and glycine 37 in mGlu2 and mGlu3, respectively). Amino acids in the binding pocket of the mGlu1 crystal structure are highlighted with a black background. Four mutations in mGlu2 made by Malherbe et al. (2001) are highlighted with a gray background, and the 12 mutations in mGlu3 (present study) are labeled with an asterisk.

A three-dimensional model of the mGlu3 agonist DCG-IV, with correct chiral centers, was generated from a two-dimensional drawing. Conformational analysis was done using the DISCOVER force field on DCG-IV to identify energetically favorable conformations that were within 10 kcals. Each conformation of DCG-IV was docked visually by superimposing amino and carboxyl groups between DCG-IV and the docked l-glutamate. We found one conformation that had good overlap with the bound glutamate. The coordinates of the entire mGlu3–DCG-IV complex were energy minimized without any constraints as described above. The final conformation of DCG-IV (torsion angle of α-amino group, τ1 = 168°) was in good agreement with those calculated for glutamate (τ1 = 177°) and for carboxycyclopropylglycine (τ1 = 153°) (Monn et al. 1997).

Results

Radioligand-binding experiments utilizing a tritiated version of DCG-IV and a microcentrifugation assay procedure showed an extremely strong signal-to-noise ratio in membranes from HEK cells expressing mGlu3. At a concentration of 20 nm[3H]DCG-IV, specific binding represented 94% of the total binding and typically gave about 21 000 specific counts per minute in a liquid scintillation counter (45% counting efficiency). No specific [3H]DCG-IV binding was detected in untransfected HEK cell membranes. The dissociation constant (KD) for [3H]DCG-IV binding to mGlu3 was 0.075 µm ± 0.007 µm. In all experiments, a modified expression plasmid was used in which a c-myc epitope tag was incorporated into the coding region of the rat mGlu3 cDNA after asparagine 27, just downstream from the predicted signal peptide cleavage site after glycine 24; this modified plasmid was used throughout this study. A previous study showed that the addition of a c-myc tag in the same position in mGlu4 did not affect the binding of the Group III agonist [3H]L-AP4 (Han and Hampson 1999). In the present study, the addition of the c-myc tag at the equivalent position in mGlu3 did not appreciably alter [3H]DCG-IV binding compared to the untagged receptor. The IC50 for l-glutamate (using [3H]DCG-IV) with the untagged receptor was 0.40 ± 0.28 µm (mean ± SEM), whereas the IC50 for the c-myc-tagged mGlu3 was 0.35 ± 0.05 µm. The incorporation of the c-myc epitope tag just downstream of the signal peptide was advantageous for immunostaining of fixed cells (see below).

A homology model of rat mGlu3 was generated based on the molecular coordinates of the crystal structure of rat mGlu1 (Kunishima et al. 2000); this model was used in conjunction with multiple sequence alignments as a guide for site-directed mutagenesis of the glutamate binding pocket. Several residues in the ligand-binding pockets of the mGlu receptors are conserved in all eight members of the mGlu receptor family (Fig. 2). The conserved amino acids mutated in the binding pocket of mGlu3 are R68, S151, T174, D194, Y222, D301, and K389. In each of these residues, mutation to alanine essentially eliminated [3H]DCG-IV binding (Figs 3a and b).

Figure 3.

Protein expression and relative specific binding of [3H]DCG-IV to mGlu2 and mGlu3 mutants. Single point binding assays using 20 nm[3H]DCG-IV were conducted in triplicate on membranes from HEK cells expressing wild-type or mutant receptors. Specific binding of the mutants was expressed as a percentage of wild-type receptor binding (mean ± SEM). Representative immunoblots using an antibody raised against a peptide sequence in the binding pocket of mGlu3 for the mGlu3 mutants, or a carboxy terminal anti-mGlu2/3 antibody for the mGlu2 mutants are shown above the corresponding column. (a) mGlu3 mutants on lobe I; (b) mGlu3 mutants on lobe II; (c) mGlu2 mutants on lobes I and II.

The remaining amino acids mutated in mGlu3 are not conserved in all eight mGlu receptors. These include R64, Y150, R277, S278, and Q306. Mutation of R64 to alanine eliminated binding, mutation of Y150, R277, and Q306 to alanine produced substantial decreases in binding, whereas mutation of S278 to alanine had little effect on [3H]DCG-IV binding to mGlu3 (Figs 3a and b). Thus, with the exception of S278A and Q306A, the amount of residual specific [3H]DCG-IV binding was too low to conduct further affinity or inhibition analyses. With the S278A mutant, 95% of the wild-type activity was retained (Table 1), whereas in the case of Q306A, sufficient binding remained (13% of wild-type mGlu3, Fig. 3b) to carry out competition experiments. Using unlabeled l-glutamate as the competing compound, the IC50 of Q301A (1.77 ± 0.45 µm) was decreased fivefold compared to the unmutated receptor (IC50 = 0.35 ± 0.05 µm, Fig. 4). It should be noted that, based on the molecular model of mGlu3, the absence of effects with the S278A mutation, and the relatively small effect of the Q306A mutation is likely explained by the fact that these two residues are positioned slightly outside the ligand-binding pocket.

Table 1.  mGlu2 and mGlu3 mutants generated and characterized
 Percentage normalized
binding to wild type
Equivalent residue
in mGlu1
Lobe
  1. Summary of binding results for wild and mutant receptors. Twenty nanomolar [3H]DCG-IV was used in all experiments. Specific binding was expressed as a percentage of wildtype receptor binding (mean ± SEM, n = 3).

mGlu3 mutation
mGlu3-cmyc-N100  
R64A0 ± 3.9Y74I
R68A1.8 ± 3.9R78I
Y150A6.5 ± 4.3S164I
S151A0.3 ± 0.4S165I
T174A0.1 ± 0.3T188I
K389A0 ± 0.3K409I
D194A0.2 ± 0.4D208II
Y222A4.0 ± 0.5Y236II
R277A7.4 ± 0.8E292II
S278A95.5 ± 4.6G293II
D301A0.2 ± 0.1D318II
Q306A13.2 ± 2.3R323II
mGlu2 mutation
mGlu2100  
K377A0.85 ± 1.2K409I
D188A0.4 ± 1.2D208II
R271A8.7 ± 1.1E292II
L300A33 ± 2.8R323II
Figure 4.

Inhibition of [3H]DCG-IV binding by l-glutamate. Binding curves for wild-type mGlu3 (□) and the mGlu3 Q306A mutant (▪) are shown. The IC50 values were 0.38 ± 0.16 µm and 1.77 ± 0.45 µm, respectively (n = 3).

In addition to the 12 mGlu3 mutants, several mutations were made in the highly homologous Group II receptor, mGlu2. These mutants were produced to verify results in mGlu3 and/or because they were not examined in a previous mutagenesis study of mGlu2 (Malherbe et al. 2001). The mGlu2 mutants D188A and K377A, corresponding to D194 and K389, respectively, in mGlu3, showed no DCG-IV binding activity, whereas the R271A mutation, corresponding to R277 in mGlu3, displayed 9% of the specific binding activity measured in the wild-type mGlu2 receptor (Fig. 3c). Thus, the results obtained with this set of mGlu2 mutants were very similar to those observed with the analogous mutants in mGlu3. The fourth mGlu2 mutant analyzed was L300A. In contrast to the equivalent mutation in mGluR3, Q306A, which showed 13% of wild-type mGlu3 binding, L300A in mGlu2 showed substantially more [3H]DCG-IV binding (33% of wild-type mGlu2) (Fig. 3c).

The wild-type mGlu3 and selected mGlu3 mutants were also examined in a fluorescence-based functional assay in live HEK cells. The cells were co-transfected with the mGlu3 cDNA and a cDNA coding for the promiscuous G-protein subunit Gα15 to switch the signal transduction pathway from the inhibition of cAMP production to the stimulation of phosphoinositide turnover and release of intracellular calcium which was detected using the calcium-sensitive dye Fluo-4. In this assay, wild-type mGlu3 yielded EC50 values for DCG-IV and L-CGG-I of 0.08 ± 0.03 and 0.09 ± 0.04 µm, respectively (Table 2); the EC50 value for DCG-IV was very close to the KD value obtained in the ligand-binding experiments noted above (0.075 µm). All of the mutants analyzed in this assay caused a 688-fold or greater increase in the EC50 value for DCG-IV compared to the wild-type receptor. With several mutants, including R68A, S151A, T174A, D194A, and D301A, EC50 values could not be calculated because responses were only obtained at two or three of the highest drug concentrations tested and full dose–response curves could not be generated because of drug insolubility and/or pH changes in the assay buffer at higher drug concentrations. With the S150A, Y222A, and R277A mutants a clear dichotomy was observed between the responses obtained with DCG-IV and L-CCG-I. All three mutations caused a dramatic reduction in the affinity for DCG-IV but showed relatively little effect on responses to L-CCG-I (Table 2).

Table 2.  EC50 values for wild-type and mutant mGlu3 receptors
ReceptorEC50m)
DCG-IVL-CCG-I
  1. EC50 values (in micromolar concentrations) for mGlu3 and mGlu3 mutants from the fluorescence-based functional assay conducted on live transiently transfected HEK cells. Each value is the mean ± SEM of 2–4 determinations.

Wild-type0.08 ± 0.030.09 ± 0.04
R64A417 ± 32066 ± 21
R68A> 1000> 1000
Y150A306 ± 1190.06 ± 0.05
S151A> 1000> 1000
T174A> 1000> 1000
D194A> 100> 100
Y222A> 10000.15 ± 0.06
R277A> 500.24 ± 0.18
D301A> 100> 100
K389A55 ± 29> 100

The reduction or elimination of DCG-IV binding seen with the various mutants might be caused by global protein misfolding induced by perturbation of the tertiary structure of the protein, rather than a selective elimination of an important determinant of ligand binding. In some cases, global protein misfolding causes retention of the misfolded protein in the endoplasmic reticulum of the cell. To assess this possibility, immunocytochemical analyses were conducted on fixed HEK cells expressing wild-type or mutants receptors. Cells expressing the c-myc-tagged mGlu3 and immunostained with an anti-c-myc monoclonal antibody showed prominent cell-surface labeling in non-permeabilized cells consistent with the extracellular location of the c-myc epitope tag inserted adjacent to the predicted signal peptide sequence (Fig. 5a). Mock-transfected HEK cells showed no staining with the c-myc antibody under the same conditions (not shown). Systematic analysis of all of the mGlu3 mutants, several examples of which are illustrated in Figs 5(b)–(f), showed cell-surface labeling that was similar to, or indistinguishable from the surface labeling of the wild-type receptor.

Figure 5.

Immunocytochemical analysis of selected mGlu3 mutants. (a) Wild-type mGlu3; (b) Y150A; (c) D194A; (d) R277A; (e) Q306A; (f) K389A. Non-permeabilized HEK cells were immunostained with a monoclonal anti-c-myc primary antibody and an Alexa 594-conjugated anti-mouse secondary antibody.

To examine potential cell permeabilization by the fixation procedure alone, we treated HEK cells transfected with the wild-type mGlu3 with or without Triton X-100 and labeled them with a polyclonal antibody that recognizes the intracellular carboxy terminus of mGlu2 and mGlu3 (Chemicon Corp.). Intense labeling along the perimeter of the cell (plasma membrane) was observed only in cells treated with Triton X-100. In cells not treated with detergent, very faint diffuse intracellular staining was observed, indicating a low level of antibody permeabilization, and the intense labeling along the plasma membrane seen in Triton-treated cells was not visible (data not shown). Therefore the plasma membrane staining seen with the anti-c-myc antibody to the extracellular epitope does appear to give an accurate indication of cell-surface labeling of mGlu3 and the mGlu3 mutants. Cell-surface expression was also substantiated by the fact that functional responses were obtained with the mutants. Taken together, the immunocytochemical results and functional analyses indicate that the mutant receptors are trafficked to the cell surface like the wild-type receptor.

Discussion

In the homology model constructed for the ligand-binding domain of mGlu3 there are nine amino acids (R64, R68, Y150, S151, T174, Y222, R277, D301, and K389) within the binding pocket that make direct contacts with one or more functional groups in the docked DCG-IV molecule (Fig. 6). Two residues interact with the α-amino group of DCG-IV (threonine 174 and aspartate 301), at least one interacts with the α-carboxylic acid (serine 151), and three amino acids interact with the 2′ carboxylic acid (γ-carboxylic acid) group of DCG-IV (arginines 64 and 68, and lysine 389). In addition, tyrosine 150 along with tyrosine 222 and arginine 277 establish bonding interactions with the third carboxylic acid (3′ carboxylic acid) group in DCG-IV.

Figure 6.

Depiction of the binding pocket of rat mGlu3 based on the molecular model. For clarity, only the residues making direct contacts with the DCG-IV molecule are shown and selected hydrogen bonds that are within 2.7–2.9 angstroms are shown as thin yellow lines. The atoms and bonds are color coded according to atom type; red for oxygen, blue for nitrogen and green for carbon. The carbon backbone of the DCG-IV molecule is colored cyan.

A general conclusion that can be drawn from the experimental findings presented here is that [3H]DCG-IV binding to mGlu3 is exquisitely sensitive to mutagenesis-induced perturbation. Ten of the 12 amino acids in mGlu3 and three of the four amino acids in mGlu2 that were mutated to alanine produced a 90% or greater decrease in [3H]DCG-IV binding; half of the mutants generated completely eliminated binding. The exceptional mutagenesis-induced sensitivity of DCG-IV binding to mGlu3 is likely due to the fact that DCG-IV is both a more rigid and a more complex molecule than l-glutamate and the Group III agonist L-AP4. This rigidity is due in part to the constraints on the C1′–C2′ bond in DCG-IV.

Threonine 174 and aspartate 301 in mGlu3, which interact with the α-amino group of the ligand, and serine 151, which interacts with the α-carboxylic acid group of the ligand, are conserved in all eight mGlu receptor subtypes. Conversion of each of these residues to alanine induced a large decrease or elimination of [3H]DCG-IV binding. Similar results with the equivalent mutations were seen with [3H]quisqualic acid binding to mGlu1 (Sato et al. 2003). Mutations in the analogous amino acids in mGlu2 caused a loss of the binding of the Group II radiolabeled agonist [3H]LY354740, except for threonine 168 (equivalent to threonine 174 in mGlu3), which showed a 21-fold reduction in affinity compared to wild-type mGlu2 (Malherbe et al. 2001).

Mutation of these three conserved amino acids also caused a loss or large reduction in binding or activation of the Group III mGlu receptors mGlu4 and mGlu8 (Hampson et al. 1999; Bessis et al. 2002; Rosemond et al. 2002). Of considerable interest is the observation that mutation of tyrosine 227 or aspartate 309 to alanine in mGlu8 (equivalent to tyrosine 222 and aspartate 301 in mGlu3) converted the Group III antagonists ACPT-II and MAP4 into agonists (Bessis et al. 2002). Based on a homology model of the extracellular domain of mGlu8, it was suggested that these compounds are antagonists because steric hindrance prevents the closure of the Venus flytrap domain in the wild-type receptors, and that removal of these side chains alleviates the hindrance thereby allowing closure of the flytrap module.

Our results with mGlu3 indicate that the conserved residues at serine 151, threonine 175, and aspartate 301 in mGlu3 appear to be crucial anchor points for the binding of a variety of mGlu receptor ligands with diverse structures. This is consistent with the various molecular models and the crystal structure of mGlu1 all indicating that these amino acids establish binding interactions with the α-amino and α-carboxylic acid functional groups which are common features of mGlu ligands.

Three additional amino acids establish bonds with the 2′ carboxylic acid group of DCG-IV (arginines 64 and 68, and lysine 389); mutation of any of the three eliminated detectable [3H]DCG-IV binding. Of these, arginine 68 and lysine 389 are conserved in all mGlu receptors. Similar to that observed with [3H]DCG-IV binding to mGlu3, both of these mutations eliminated [3H]L-AP4 binding to mGlu4 (Hampson et al. 1999; Rosemond et al. 2002). Surprisingly, although mutation of arginine 78 in mGlu1 (equivalent to R68 in mGlu3) eliminated [3H]quisqualic acid binding to mGlu1 (Jensen et al. 2000; Sato et al. 2003), mutation of lysine 409 (equivalent to R389 in mGlu3) had no effect on [3H]quisqualic acid binding (Sato et al. 2003). Thus, lysine 389 in mGlu3 is an example of a conserved amino acid in the binding pocket of mGlu receptors that, upon mutation, produces differential effects on ligand binding amongst the various mGlu ligands and receptor subtypes.

Arginine 64 is not conserved in the mGlu receptor family. In mGlu2 this position is occupied by arginine 57. As observed in the present study on [3H]DCG-IV binding to mGlu3, this residue was also found to be critical for the binding of [3H]LY354740 to mGlu2 (Malherbe et al. 2001). In contrast, mutation of the equivalent amino acid in mGlu4 (lysine 74) had no effect on the binding of [3H]L-AP4. However, a double mutation of lysine 74 together with lysine 317 drastically reduced [3H]L-AP4 (Rosemond et al. 2002). Thus the amino acids that occupy these positions in the binding pockets of the mGlu receptors may contribute to the subtype selectivity of the various mGlu receptor ligands.

DCG-IV possesses a third carboxylic acid functional group at C3′, absent in L-CCG-I, and a cyclopropyl ring, both of which are absent in l-glutamate and L-AP4 (Fig. 1). This third carboxylate group projects in the opposite direction of the C2′ carboxylic acid. In the homology model, this acidic group makes contacts with tyrosine 150, tyrosine 222, and arginine 277 (Fig. 6). We suggest that the positively charged side chain of arginine 277 interacts with the third carboxylate group on DCG-IV. This suggestion is consistent with our results, showing a dramatic decrease in [3H]DCG-IV binding to the equivalent mutation in mGlu2 (R271A, Table 1), and the lack of effect of the same mutation on [3H]LY354740 binding to mGlu2 (Malherbe et al. 2001). LY354740 lacks the third carboxylic acid group (Fig. 1) and thus, consistent with the experimental observations, would not be predicted to be affected by this mutation. Further evidence for interaction of the third carboxylate of DCG-IV with tyrosines 150 and 222 and arginine 277 was obtained in the functional analysis of these mutants, where very large decreases in affinity for DCG-IV were seen, whereas, in contrast, all three mutants displayed either wild-type affinity or relatively small decreases in affinity for L-CCG-I (Table 2). Together these findings provide strong confirmation for the correct orientation of DCG-IV docking into the binding pocket in the mGlu3 homology model (Fig. 6).

In the binding of ligands to mGlu receptors, both hydrogen bonding and electrostatic interactions are essential (Macchiarulo et al. 2003). In the present study, the 50-fold higher affinity of DCG-IV for mGlu3 (EC50 = 0.09 µm) compared to the putative endogenous ligand l-glutamate (EC50 = 4.1 µm; Brabet et al. 1998; Schoepp et al. 1999) corresponds to only about 1.6 kcal/mol of additional binding energy. This could result from a small entropic contribution, a weak hydrogen bond, or only a weak charge–charge interaction. It is possible that within the microenvironment of the ligand/receptor complex, the third carboxyl of bound DCG-IV is in the unionized form, and functioning as a hydrogen bond donor/acceptor in association with arginine 277 in mGlu3, and with arginine 271 in mGlu2. The hydrogen bond acidity inline image of an unionized carboxyl group is only about twice that of a primary hydroxyl group (0.54 vs. 0.33, respectively), and its hydrogen bond basicity inline image is essentially the same as those for ketone carbonyl and primary hydroxyl groups (0.42 vs. 0.49 and 0.45, respectively; Abraham and Platts 2001). Thus, the hydrogen-bonding properties of an unionized carboxyl group could account for the observed affinity difference between DCG-IV and glutamate, and suggests that other hydrogen bonding groups might behave similarly.

Indeed, an increase in binding affinity was observed upon introduction of a ketone carbonyl group (equivalent to the third carboxyl carbonyl group of DCG-IV) at position 4 of the 2-amino-3-fluorobicyclo[3.1.0]hexane-2,6-dicarboxylic acid (AFBD) structure to form MGS0028 (Fig. 1; Nakazato et al. 2000). Both AFBD and MGS0028 are Group II agonists. Using tritiated AFBD as the radioligand, MGS0028 showed approximately sevenfold and 12-fold higher affinities for mGlu2 and mGlu3, respectively, than the non-ketonic structure of AFBD (Nakazato et al. 2000). The binding energy of these affinity enhancements are on the order of 1 kcal/mol, the energy expected for a hydrogen bond. The lack of conservation of these arginines and thus the inability to form the proposed hydrogen bond in the other receptor subtypes may help to explain the selectivity of DCG-IV and related compounds for Group II mGlu receptors.

In conclusion, the results presented here define multiple specific interactions between DCG-IV and mGlu3. Each of these interactions appears to be required for maintaining high affinity binding, since the individual removal (i.e. conversion to alanine) of any of the nine key residues in the pocket appears to destabilize the overall receptor/ligand complex to the point of complete or nearly complete loss of [3H]DCG-IV binding. Thus molecular modeling and mutagenesis approaches, additional crystal structures of mGlu receptors complexed with different drugs (Tsuchiya et al. 2002), and pharmacophore strategies (Jullian et al. 1999; Bertrand et al. 2002), including a proposed hydrogen bonding site on mGlu3, now provide a powerful platform for progression to structure-based design of new Group II mGlu receptor ligands.

Acknowledgments

We thank Dr S. Nakanishi for the rat mGlu2 and mGlu3 cDNAs, Dr M. Simon for the Gα15 cDNA, A. Jain and E. Rosemond for cell transfections, and Drs J. R. Damewood and E. C. Johnson for comments on the manuscript. This work was supported by the Canadian Institutes of Heath Research, AstraZeneca Pharmaceuticals LP, and NPS Pharmaceuticals Inc.

Ancillary