Elucidation of a Novel Extracellular Calcium-binding Site on Metabotropic Glutamate Receptor 1α (mGluR1α) That Controls Receptor Activation*

Metabotropic glutamate receptor 1α (mGluR1α) exerts important effects on numerous neurological processes. Although mGluR1α is known to respond to extracellular Ca2+ ([Ca2+]o) and the crystal structures of the extracellular domains (ECDs) of several mGluRs have been determined, the calcium-binding site(s) and structural determinants of Ca2+-modulated signaling in the Glu receptor family remain elusive. Here, we identify a novel Ca2+-binding site in the mGluR1α ECD using a recently developed computational algorithm. This predicted site (comprising Asp-318, Glu-325, and Asp-322 and the carboxylate side chain of the receptor agonist, Glu) is situated in the hinge region in the ECD of mGluR1α adjacent to the reported Glu-binding site, with Asp-318 involved in both Glu and calcium binding. Mutagenesis studies indicated that binding of Glu and Ca2+ to their distinct but partially overlapping binding sites synergistically modulated mGluR1α activation of intracellular Ca2+ ([Ca2+]i) signaling. Mutating the Glu-binding site completely abolished Glu signaling while leaving its Ca2+-sensing capability largely intact. Mutating the predicted Ca2+-binding residues abolished or significantly reduced the sensitivity of mGluR1α not only to [Ca2+]o and [Gd3+]o but also, in some cases, to Glu. The dual activation of mGluR1α by [Ca2+]o and Glu has important implications for the activation of other mGluR subtypes and related receptors. It also opens up new avenues for developing allosteric modulators of mGluR function that target specific human diseases.

Metabotropic glutamate receptors (mGluRs) 2 have key functions in a variety of different neurological processes, including memory, learning, pain, synaptic plasticity, and the control of the activity of various circuits throughout the brain (1). The mGluRs belong to family C of the large superfamily of G protein-coupled receptors (GPCRs). Family C GPCRs (also referred to as family 3 GPCRs, the nomenclature that will be utilized here) also include the Ca 2ϩ -sensing receptor (CaSR), GABA B receptors, taste receptors, and putative pheromone receptors (2). All members of the family 3 GPCRs share similar domain architecture, including venus flytrap-like extracellular domains (ECD), heptahelical transmembrane domains, and intracellular C-terminal C-tails. The mGluRs fall into three groups and eight subtypes. Group I comprises mGluR1 and mGluR5 (3). mGluR1 is expressed mainly around a core of ionotropic glutamate receptors in the postsynaptic densities of neurons and functions as a disulfide-linked homodimer (4). Upon activation by its agonists, the intracellular domains of the group I mGluRs associate with the G protein G q/11 to activate phospholipase C, which subsequently converts phosphatidylinositol bisphosphate (PIP 2 ) to diacylglycerol and inositol trisphosphate (IP 3 ), thereby releasing Ca 2ϩ from the endoplasmic reticulum, as well as activating protein kinase C (PKC) and other downstream effectors (5).
The issues of whether mGluRs respond to extracellular calcium ([Ca 2ϩ ] o ) and how calcium binding modulates the family 3 GPCRs have attracted extensive investigation. On the basis of sequence homology to CaSR, mGluRs were postulated to be capable of responding to [Ca 2ϩ ] o . [Ca 2ϩ ] o has been proposed to either activate mGluR1 directly or to act as a positive mGluR1 modulator (6,7). Kubo et al. (6,8) reported that [Ca 2ϩ ] o , as well as Glu, can trigger intracellular responses elicited by mGluR1, mGluR3, and mGluR5. [Ca 2ϩ ] o or Gd 3ϩ further stimulate the activity of mGluR1␣ even after saturation of the Glu response and vice versa (6). In addition, mGluR1␣ responds to 5 mM [Ca 2ϩ ] o in Purkinje cells prepared from global mGluR1␣ knock-out mice in which the receptor has been specifically knocked into Purkinje cells, whereas the Purkinje cells from the mGluR1␣ global knock-out mice themselves cannot sense [Ca 2ϩ ] o (9,10). On the basis of these studies, [Ca 2ϩ ] o is postulated to mediate postsynaptic efficacy through it action on mGluR1 (11). Moreover, Glu triggers [Ca 2ϩ ] i oscillations in a manner that is modulated by [Ca 2ϩ ] o (12), as R, 5-3, 5-Dihy-droxylphenylglycine, an agonist of group I mGluRs, generated inward currents that were enhanced by [Ca 2ϩ ] o (10). In contrast, Nash et al. (13) concluded that mGluR1␣ is not a calciumsensing receptor because its response to the agonist L-quisqualate is not sensitive to [Ca 2ϩ ] o . However, the effect of [Ca 2ϩ ] o on the EC 50 for quisqualate was not examined. Any putative Ca 2ϩbinding sites capable of regulating mGluR signaling remain "invisible" in six crystal structures of the ECD of mGluR1␣ determined to date (14,15), as well as the ECDs and cysteinerich domains of mGluR3 and mGluR7 (15,16). One Gd 3ϩ ion binds to mGluR1 between the helices of lobe 2 (LB2) at the dimer interface of the ECD, far from the Glu-binding site (14,17). Removing the Gd 3ϩ -binding residue, E238Q, eliminated sensitivity to Gd 3ϩ but not sensitivity to [Ca 2ϩ ] o and Glu (17,18). Two Gd 3ϩ ions visible in the crystal structure were ignored by these authors, although one of them is located near the critical hinge region coordinated by Asp-322, Asp-324, and Asp-493 (14). This observation also suggests strongly that a Ca 2ϩ ion could bind to this region of the protein. The invisibility of Ca 2ϩbinding sites in the x-ray structures of the mGluRs represents a major challenge shared among other Ca 2ϩ -modulated proteins functioning at high Ca 2ϩ concentrations, like those in the extracellular fluids, due to their low Ca 2ϩ -binding affinities (K d , ϳ0.1-1.5 mM) and irregular binding geometries (19). Our understanding of the role of Ca 2ϩ as an extracellular signal acting via family 3 GPCRs beyond CaSR is severely hampered by the lack of adequate information about the location and properties of the Ca 2ϩ -binding sites of this class of proteins.
We report herein the identification of a novel Ca 2ϩ -binding site adjacent to the Glu-binding site in the hinge region of the ECD of mGluR1␣ (14,15), which was found by using our recently developed the MUG (multiple geometries) algorithm (20 -22). MUG is a graphic geometry-based Ca 2ϩ -binding site prediction software. It extracts oxygen clusters from Protein Data Bank (PDB) files and assumes a Ca 2ϩ center for each cluster. The clusters then are verified by setting parameters for geometric filters that define the range of distance between oxygen atoms and Ca 2ϩ . The clusters satisfying the parameter setting were considered candidates for Ca 2ϩ -binding pockets. The putative Ca 2ϩ -binding pockets of lower quality were further modified by allowing rotation of the side chains of predicted liganding residues. To investigate a single Ca 2ϩ -binding site present within a short stretch of amino acids, normally less than 30 residues, we engineered the short loop into a scaffold protein, CD2. Intact CD2 does not bind Ca 2ϩ and is tolerant of site-directed mutagenesis without undergoing changes in its overall structure. Then the metal-binding capabilities of the key predicted Ca 2ϩ -binding residues were further characterized using luminescence energy transfer (LRET) and site-directed mutagenesis (23,24 (14,15,25). Notably, the carboxylate group of the side chain of Glu also contributes to the binding site for Ca 2ϩ . We propose a dual activation mechanism whereby the simultaneous binding of Glu and Ca 2ϩ , at their separate but partially overlapping binding sites, potentiates one another's actions to yield maximal activation of mGluR1␣.

EXPERIMENTAL PROCEDURES
Computational Prediction of Ca 2ϩ -binding Sites in mGluR1␣ and Molecular Modeling-The three-dimensional coordinates of the crystal structures of the ECD of mGluR1␣ were obtained from the PDB (PDB entry codes: 1EWT, 1EWK (15), and 1ISR (14)). Hydrogen atoms were added using the Sybyl7.2 package (Tripos Inc., St. Louis, MO). The identification of putative Ca 2ϩ -binding sites in the ECD of mGluR1␣ was performed using MUG, a graph theory-based algorithm (21) developed by our laboratory. The Ca-O distance in the software was set to 1.6 -3.1 Å with a set average cutoff of 2.4 Å (26,27), and the O-O distance was set to 6.0 Å (21). Side chain atoms were rotated to accommodate Ca 2ϩ -induced local conformational changes (48). Furthermore, electrostatic surface potential maps were constructed using Delphi (28), and GRASP (29) was then used to render and modify the image. The linear, putative Ca 2ϩbinding site was added into the scaffold protein CD2 between Ser-52 and Gly-53 with triple Gly linkers at both ends, and the combined grafting model was generated by Modeler 9v4 (30).
Tb 3ϩ Titration and Ca 2ϩ Competition-In Trp-sensitized Tb 3ϩ -LRET experiments, emission spectra from 500 to 580 nm were recorded with excitation set at 282 nm; slit widths were set at 8 nm for excitation and 12 nm for emission. A glass filter with a cutoff of 320 nm was utilized to circumvent secondary Rayleigh scattering. Tb 3ϩ titration and metal competition assays were performed as described previously (24). 500 mM K ϩ , 10 M La 3ϩ , 10 M Gd 3ϩ , 1 mM Mg 2ϩ , and 1 mM Ca 2ϩ , respectively, was used to selectively compete with Tb 3ϩ . Each experiment was carried out independently in triplicate.
Quantitatively Determined Membrane Expression of the mGluR1␣ Mutants Using Flow Cytometry-PcDNA-mGluR1␣ (donated by Dr. Randy Hall's laboratory) contained a FLAG tag at the N terminus of the receptor, and mCherry was genetically fused to the C terminus with a linker, GGNSGG. After 2 days of transient expression of mGluR1␣ and its mutants (D318I, D322I, E325I, and N335I) in HEK293 cells grown on polylysinecoated dishes, cells were incubated in 1ϫ phosphate-buffered saline (PBS) supplemented with 1/1000 anti-FLAG and 1/100 fetal bovine serum (FBS) at 4°C. The cells were then washed three times with 1ϫ Tris-buffered saline (TBS) and fixed using 4% formaldehyde at room temperature for 15 min. After being washed three times with 1ϫ TBS, the receptors on the cell sur-face were then labeled with Alexa Fluor 488 goat anti-mouse IgG (Invitrogen) for 30 min at 37°C. The cells were then collected in 1ϫ PBS, and the intensity of green and red fluorescence was measured using LSRFortessa (BD Biosciences). The ratios of green and red fluorescence from mGluR1␣ and the mutated receptors were normalized to the amount of receptors expressing on the cell surface relative to total receptors (total cellular expression of receptor). Data were collected from three dishes.
Measurement of [Ca 2ϩ ] i Responses of mGluR1␣ and Its Mutants with or without [Ca 2ϩ ] o or Glu-Measurement of [Ca 2ϩ ] i was performed as described (24). In brief, wild type mGluR1␣ and its mutants (D318I, D322I, D324I, E325I, and E328I) were transiently transfected into HEK293 cells and cultured for 2 additional days. The cells on the coverslips were subsequently loaded using 4 M Fura-2 AM in 2 ml of physiological saline buffer (10 mM HEPES, 140 mM NaCl, 5 mM KCl, 0.55 mM MgCl 2 , and 1 mM CaCl 2 , pH 7.4). The coverslips were mounted in a bathing chamber on the stage of a fluorescence microscope. Fura-2 emission signals from single cells excited at 340 or 380 nm were collected utilizing a Leica DM6000 fluorescence microscope in real time as the concentration of extracellular Ca 2ϩ was increased in a stepwise manner. The ratio of emitted fluorescence at 510 nm resulting from excitation at 340 or 380 nm was further analyzed to obtain the intracellular Ca 2ϩ response as a function of changes in [Ca 2ϩ ] o . Then, the sensitivity of mGluR1␣ and its mutants (D322I, D324I, E325I, and E328I) to extracellular glutamate was measured by increasing the extracellular glutamate concentration in the presence of 1.8 mM Ca 2ϩ . The glutamate concentrations at which the intracellular Ca 2ϩ responses of mGluR1␣ and its mutants were first observed and, subsequently, saturated were determined. Moreover, to further characterize the influence of [Ca 2ϩ ] o to Glu-induced [Ca 2ϩ ] i release through wild type mGluR1␣, an additional 5 or 10 mM Ca 2ϩ was added to the perfusate. [Ca 2ϩ ] i was measured as described above during changes in [Ca 2ϩ ] o and/or Glu.
Measurement of Intracellular Ca 2ϩ Release Mediated by mGluR1␣ and Its Mutants in the Presence of Extracellular Gd 3ϩ -Changes in [Ca 2ϩ ] i in response to the addition of Gd 3ϩ were determined as just described. Specifically, cells were incubated in incubation buffer (140 mM NaCl, 4 mM KOH, 10 mM HEPES, 1.5 mM CaCl 2 , 1 mM MgCl 2 , 10 mM glucose, pH 7.4) for up to 1.5 h, and Gd 3ϩ (made up in 140 mM NaCl, 4 mM KOH, 10 mM HEPES, and 0.3 mM MgCl 2 , pH 7.4) was added at the concentrations described under "Results." The [Ca 2ϩ ] i responses of mGluR1␣ after the introduction of mutations in the Glu-binding site were measured similarly.
Data Analysis and Curve Fitting-At each agonist concentration, all of the transfected cells in the microscopic field from three independent experiments were selected for analysis, and at least 60% of the cells displaying normal responses were analyzed. The cells that did not respond to the agonists or displayed a sigmoidal curve with a stable plateau after treatment with high [Ca 2ϩ ] o were excluded. These latter cells maintained a constant, high plateau of the intracellular Ca 2ϩ concentration, perhaps because the plasma membrane was excessively permeable to Ca 2ϩ . To normalize the concentration response curves for the responses to [Ca 2ϩ ] o , the maximal response of wild type mGluR1␣ to extracellular Glu was set at 100% so that the maximal responses of mutant receptors to [Ca 2ϩ ] o or Glu were transformed into percentages relative to the response of WT mGluR1␣ to Glu. Data were fitted using the Hill equation as described previously (23).

Prediction of a Novel Ca 2ϩ -binding Site Adjacent to the Glu-
binding Site in the ECD of mGluR1␣-We recently developed the computational algorithm MUG, which predicts Ca 2ϩ -binding sites using graph theory by identifying all possible liganding oxygen clusters and finding maximal cliques. The positions of Ca 2ϩ and its liganding groups in 144 calcium-binding proteins can be predicted with 0.22-0.49 Å accuracy by geometric filters established on the basis of an extensive survey of known Ca 2ϩbinding sites in the Protein Data Bank (19). To accommodate Ca 2ϩ -induced conformational changes, the side chains of putative Ca 2ϩ -binding ligand residues were subjected to rotation using a rotamer library (MUG SR ) (48). Fig. 1 shows one predicted Ca 2ϩ -binding site identified here in the crystal structure of the mGluR1␣ ECD (PDB entry code: 1EWK) using the MUG algorithm. Two other predicted sites not included in this report were also revealed by MUG, one of them (site 2) residing in the Mg 2ϩ -binding pocket (Leu-86 -Gly-102) inferred from the crystal structure and the other one located within a long loop (Asp-125-Lys-153) that was invisible in the crystal structure because of its high flexibility and was repaired using Modeler (30). The third predicted Ca 2ϩ -binding site (site 3), encompassing Ser-129 to Gly-144, was present within this missing loop. The predicted Ca 2ϩ -binding site studied in detail here comprises the carboxyl side chains of Asp-318 and Glu-325, the main chain carbonyl Asp-322 in a flexible loop of mGluR1␣, and the carboxyl side chain of Glu-701 (a ligand for glutamate). This predicted Ca 2ϩ -binding site is located at the hinge region in the ECD adjacent to the reported Glu-binding site (Arg-74, Ser-165, Thr-188, Asp-208, Tyr-236, Asp-318, and Lys-409) (15,25), with Asp-318 predicted to be involved in both Glu and Ca 2ϩ binding. Thus, Ca 2ϩ and Glu, when bound to the receptor, both bind to Asp-318. Asp-318 and Asp-322 can be identified in the Glu-free form (PDB entry code: 1EWT), whereas the direct binding of Ca 2ϩ to the carboxyl side chain of the agonist Glu-701 is visualized only in the Glu-loaded form (PDB entry code: 1EWK). Thus, the agonist Glu provides an additional ligand for Ca 2ϩ when the former is bound to the receptor, which is very different from intracellular Ca 2ϩ -binding trigger proteins such as calmodulin that lack any additional chelating groups from molecules other than the residues within the Ca 2ϩ -binding protein itself, except for water. Fig. 1B shows that the predicted Ca 2ϩ -binding pocket has a highly negatively charged surface as revealed by Delphi in the structures of three solved forms of the ECD within mGluR1␣ (Fig. 1B).
Obtaining Site-specific Ca 2ϩ /Ln 3ϩ -binding Affinities by a Grafting Approach-To probe the Ca 2ϩ -binding capability of the predicted Ca 2ϩ -binding site in mGluR1␣, we utilized our grafting approach by inserting the protein sequence encompassing the putative mGluR1␣ Ca 2ϩ -binding site into the host protein, CD2.D1 (denoted as CD2-mGluR1-1). The inserted sequence contains all predicted Ca 2ϩ -binding residues except Glu-701. This approach had previously enabled us to obtain site-specific Ca 2ϩ binding affinities of the EF-hand motifs from calmodulin and linear Ca 2ϩ binding sequences, free from the limitations of working with membrane proteins (32,33). The putative mGluR1␣ Ca 2ϩ -binding site was flanked by flexible triple-Gly linkers and inserted between Ser-52 and Gly-53 of CD2.D1 ( Fig. 2A) to ensure a native-like conformation and close proximity (Ͻ15ϳ20 Å) to Trp-32 in order to enhance the Tb 3ϩ -LRET signal. Indeed, grafting the putative Ca 2ϩ -binding loop from mGluR1␣ did not significantly change the secondary and tertiary structures of CD2, as revealed by circular  (8)). B, Tb 3ϩ titration and Ca 2ϩ competition (shown as inset) of CD2.D1. The engineered protein bound Tb 3ϩ and Ca 2ϩ with dissociation constants of 49 Ϯ 9 M and 1.8 Ϯ 0.1 mM, respectively. Substitution of putative metal-binding ligand residues with Ile decreases Tb 3ϩ binding affinity. Tb 3ϩ binding curves of a series of CD2-mGluR1 double mutants, E331I/E333I, D324I/E325I, D318I/D322I, and E328I/N335I. All measurements were carried out in a buffer containing 20 mM PIPES and 10 mM KCl, pH 6.8. C, Tb 3ϩ binding curves of a series of CD2-mGluR1 double mutants: CD2.D1-1 (E331I/E333I), CD2.D1-2 (D324I/E325I), CD2.D1-3 (D318I/D322I), and CD2.D1-4 (E328I/ N335I). All of the measurements were carried out in a buffer containing 20 mM PIPES and 10 mM KCl, pH 6.8. D318I/D322I obviously decreased Tb 3ϩ binding affinity, whereas D324I/E325I displays two phases. The Tb 3ϩ binding affinity of the engineered protein was clearly impaired by these two pair of mutations (n ϭ 3). D, La 3ϩ binding to the engineered protein CD2.D1 monitored by 1 D 1 H NMR. 1 or 2 mM La 3ϩ results in the peak split at the aromatic group region. E, metal selectivity of CD2.D1. The addition of 500 mM K ϩ , 1 mM Ca 2ϩ , 1 mM Mg 2ϩ , 100 M Gd 3ϩ , or 100 M La 3ϩ to the pre-equilibrated Tb 3ϩ (30 M) and protein (3 M) solution was carried out independently. The resultant changes in the Tb 3ϩ luminescence signal were monitored at 545 nm (*, p Ͻ 0.05; **, p Ͻ 0.01).
Because mGluR1␣ is modulated by various polyvalent cations, including Ca 2ϩ , Gd 3ϩ , Tb 3ϩ , La 3ϩ , Mn 2ϩ , and Mg 2ϩ (6), we tested the metal binding selectivity of CD2.D1 by applying K ϩ , Mg 2ϩ , La 3ϩ , or Gd 3ϩ to compete with prebound Tb 3ϩ . Fig.  2C shows that the luminescence intensity of Tb 3ϩ decreased significantly upon adding trivalent La 3ϩ or Gd 3ϩ , indicating that Tb 3ϩ bound to the pocket was replaced by these metal ions. Gd 3ϩ had the strongest capacity to displace Tb 3ϩ . Similarly, adding La 3ϩ to CD2.D1 produced a split in the resonance of CD2.D1 at 10 ppm (Fig. 2D). Ca 2ϩ competed more effectively than Mg 2ϩ , whereas K ϩ failed to compete with Tb 3ϩ .
Next, we utilized mutagenesis studies to examine the contribution of proposed ligand-binding residues to metal-binding capability. Double substitutions of negatively charged residues by Ile to delete negative charges but preserve bulky side chains in the proposed binding pocket, as seen in the mutants D324I/ E325I, D318I/D322I, and E328I/N335I, produced 2.3-, 6.1-, or 98-fold increases in the respective dissociation constant values (Fig. 2, B and E). However, removing the negative charges from the non-Ca 2ϩ -binding residues, Glu-331 and Glu-333 (E331I/ E333I) (Fig. 2, B and E) produced a less than 2-fold change in the K d , with a modest alteration in Tb 3ϩ binding to the predicted binding pocket.
Membrane Expression of mGluR1␣ and Its Mutants-WT mGluR1␣ and its mutant forms (D318I, D322I, and E325I) were expressed heterogeneously in HEK293 cells. The receptors expressed on the cell membrane were visualized by confocal microscopy (supplemental Fig. S2) and flow cytometry (Fig. 3). Supplemental Fig. S2 shows that the mutations did not affect the distribution of the receptors on the membrane. We calculated the ratio of intensities of green and red fluorescence measured using flow cytometry (LSRFortessa, BD Biosciences); the mutant receptors displayed expression levels on the cell membrane comparable with that of the wild type receptor, although E325I displayed a somewhat lower membrane expression level ( Fig. 3; n ϭ 3). Thus, the mutations involving the predicted Ca 2ϩ -binding site had little effect on the surface expression of the respective mutant receptors.
Extracellular Ca 2ϩ Triggers mGluR1␣-mediated Intracellular Responses-We next examined the mGluR-mediated intracellular Ca 2ϩ responses in HEK293 cells transfected with mGluR1␣-mCherry. The fluorescent protein mCherry was fused to mGluR1␣ to correlate cellular responses with the expression of mGluR1␣. We chose HEK293 cells as a model because this cell line lacks endogenous mGluR1␣ (34). mGluR1␣-mCherry was well expressed and correctly targeted to the cell membrane (Figs. 3 and 4A and supplemental Fig. S2), and Fura-2 was efficiently loaded (Fig. 4B). Single cell, real time imaging was performed using fluorescence microscopy. To minimize receptor desensitization by agonists, the responses to each concentration of added Ca 2ϩ or Glu were examined using separate coverslips.
The response of wild type mGluR1␣ to Glu was investigated in physiological saline buffer with 1.8 mM Ca 2ϩ . Wild type mGluR1␣ only responded at Ն0.5 M Glu, and this response was saturated at 30 M Glu. Fig. 6A shows that at 1.8 mM Ca 2ϩ , the addition of Ͼ30 M Glu evoked large [Ca 2ϩ ] i responses. . Surface expression of WT mGluR1␣ and its mutants. mGluR1␣ carries FLAG tag at its N terminus and mCherry at its C terminus. mGluR1␣ and its mutants were transiently expressed in HEK293 cells seeded on 50-mm dishes coated with polylysine. After incubation with anti-FLAG, the receptors on the membrane could be visualized using a secondary antibody, Alexa 488-anti-mouse IgG (Invitrogen). Emissions at 520 and 610 nm were collected by flow cytometry (LSRFortessa, BD Biosciences); these represent receptors present on the cell membrane and overall, respectively. Ratios of fluorescence at 520 -610 nm indicate the membrane expression levels of WT mGluR1␣ and its mutants. Emission at 520 nm (green signal) reflects the membrane expressed receptors, whereas the red signal at 610 nm from mCherry is a measure of total expression of the receptor. NT indicates non-transfected cells, which display no fluorescence. Although E325I displays a relatively lower surface expression, the other mutants have membrane expression level comparable with that of WT mGluR1␣ (n ϭ 3).  Fig. 5 and Table 2), although all mutants were expressed at levels comparable with the WT receptor, as assessed by immunofluorescence and flow cytometry ( Fig. 3 and supplemental Fig. S2). These results suggest that the predicted Ca 2ϩbinding residues, Asp-318, Glu-325, and Asp-322 (especially the first two), are important for the sensitivity of mGluR1␣ to modulation by [Ca 2ϩ ] o . However, S166A maintains Ca 2ϩ sensitivity with a lower maximal response (Fig. 4C), although Ser-166 was previously reported to be a potential Ca 2ϩ -binding residue (6).  Table 2) but retains Glumediated [Ca 2ϩ ] i responses (Fig. 6A). However, its EC 50 value for Glu-mediated responses is increased by ϳ18-fold ( Fig. 6B and Table 2). These results further confirm that Glu-325 contributes to Ca 2ϩ binding without directly liganding Glu (as shown by earlier studies of the binding site for Glu, which did not identify Glu-325 as a Glu ligand). However, the proximity of the Ca 2ϩ -and Glu-binding sites may produce indirect, conformational effects of mutating residue 325 on Glu binding. Furthermore, D322I exhibited a reduction in EC 50 for [Ca 2ϩ ] o by only 33%, consistent with it making a relatively minor contribution as a ligand for Ca 2ϩ binding. In contrast to the marked impact of D318I and E325I on the EC 50 for [Ca 2ϩ ] o , removal of other charged residues, such as D324I and E328I, did not alter either the EC 50 (3 and 8% changes, respectively) or the magnitude of the response to [Ca 2ϩ ] o significantly in the absence of Glu (104 Ϯ 10 and 102 Ϯ 5, respectively, of the control level) (Fig. 5 and Table 2).

Effect of Mutating Glu-binding Site on [Ca 2ϩ ] i Responses to Glu and Ca 2ϩ
-To further explore the synergistic interaction between the predicted Ca 2ϩ -and Glu-binding sites, four mutations at Glu ligand residues (S165A, T188A, D208I, and Y236F) were generated. Consistent with studies reported previously (25), T188A and D208I entirely eliminated Glu sensitivity, whereas S165A and Y236F could be activated only by high concentrations (100 M) of Glu (Fig. 8A). Interestingly, all receptors with mutated Glu-binding ligand residues (exception for Asp-318) retained a sensitivity to [Ca 2ϩ ] o (Fig. 8B and Table 3), although their EC 50 values were increased compared with that of wild type mGluR1␣ (Table 3), again perhaps owing to local conformational effects of mutating the Glu-binding site on Ca 2ϩ binding. S165A and D208I increased the EC 50 of the wild type receptor for [Ca 2ϩ ] o from 3.0 to 8.1 and 4.6 mM, respectively, although their maximal responses were comparable to that of the wild type receptor (Fig. 8B and Table 3). Conversely, T188A and D208I exhibited much reduced maximal responses (26 and 66%, respectively), whereas their EC 50 values were comparable with that of the wild type receptor (Table 3). Taken together, these data show that it is possible to generate mGluR1␣ variants responding to either Glu or to Ca 2ϩ alone. Thus mGluR1␣ can function as a true [Ca 2ϩ ] o -sensing receptor, as certain mutants, such as S165A and D208I, do not respond to Glu but maintain their Ca 2ϩ -sensing capability with only a modest increase in the EC 50 for [Ca 2ϩ ] o .

Effects of Mutations in Predicted Ca 2ϩ -binding Site on Gd 3ϩ -induced [Ca 2ϩ ] i Responses-Gd
3ϩ is also revealed at the hinge region in the Fourier map, where it shares residues Asp-322 and Asp-324 from the loop that contributes to Ca 2ϩ binding (14). Because of the low resolution of this crystal structure (4 Å), the highly flexible loop that binds Gd 3ϩ in the crystal structure, and the similarity of the binding geometries of Gd 3ϩ and Ca 2ϩ , these two cations probably share, at least in part, the same residues. To address this possibility, the responses to [Gd 3ϩ ] o of D318I and E325I were compared with that of the wild type receptor. Consistent with results reported by Abe et al. (17,36), the dose-response profiles of wild type mGluR1␣ display a bell-shaped curve. However, the introduction of the mutation D318I or E325I completely eliminated the receptor sensitivity to [Gd 3ϩ ] o (Fig. 9).

DISCUSSION
We utilized the computational algorithm MUG (22) to predict a novel Ca 2ϩ -binding site in mGluR1␣ adjacent to the Glu-binding site shown in Fig. 1. This predicted Ca 2ϩ -binding site (comprising Asp-318, Glu-325, Asp-322, and the carboxylate side chain of Glu-701) does not completely overlap the Glu-binding site (15,25). However, both sites include Asp-318, which our data suggest is involved in both Glu and Ca 2ϩ binding. The metal-binding capability of the predicted Ca 2ϩ -binding residues in mGluR1␣ was verified by a grafting approach. Like wild type mGluR1␣, the predicted Ca 2ϩ -binding site grafted in a scaffold protein (CD2) exhibited metal selectivity for Ca 2ϩ and its trivalent analogs, Gd 3ϩ , Tb 3ϩ , and La 3ϩ , in contrast to the physiological monovalent cation, K ϩ , which exhibited no measurable affinity for the predicted site. The metal-binding capability of the predicted metal-binding ligand residues in mGluR1␣ was further verified by replacing negatively charged residues with Ile. The Ca 2ϩ binding affinity of mGluR1␣ determined by the grafting approach (ϳ1.8 mM) is within the physiological concentration of [Ca 2ϩ ] o in the nervous system (0.8 -1.5 mM) (37), although this Ca 2ϩ binding constant may be changed slightly in vivo by the local microenvironment and/or the presence of Glu released by nearby cells.
We further demonstrated that mGluR1␣ could be activated by either Glu or [ We have also shown that mutating predicted Ca 2ϩ -binding residues abolishes or significantly not only reduces [Ca 2ϩ ] o sensitivity but also, in some cases, affects Glu-mediated responses. For example, E325I completely abolished [Ca 2ϩ ] o -mediated [Ca 2ϩ ] i responses in the absence of Glu (Fig. 3C and Table 2). At the same time, this mutant retained sensitivity to Glu, albeit with an ϳ18-fold reduction in EC 50 (without any decrease in maximal response) (Fig. 5 and Table 2). Thus, despite its not being a Glu-binding residue, the presence of an intact Ca 2ϩbinding ligand at Glu-325 considerably enhanced the affinity of mGluR1␣ for Glu. How could this take place? These results, in fact, are consistent with separate but overlapping Ca 2ϩ -and Glu-binding sites (Figs. 1 and 10). As noted, positively charged Arg-323 is also very close to Ca 2ϩ in the model, but it might not FIGURE 6. Intracellular Ca 2؉ responses to extracellular Glu in HEK293 cells transfected with WT mGluR1␣ or its mutants. Three negatively charged residues in the predicted Ca 2ϩ -binding pocket (Asp-318, Asp-322, and Glu-325) were mutated into Ile. Along with WT mGluR1␣, the mutants were transiently expressed in HEK293 cells. In the presence of 1.8 mM Ca 2ϩ , extracellular Glu-induced intracellular Ca 2ϩ release was measured by recording emission intensities at 510 nm excited at 340 and 380 nm, respectively. A, responses to Glu of mutations on Ca 2ϩ -binding site. Except for mutant D318I, two other mutants, D322I, E325I, and WT mGluR1␣ display responsiveness to Glu. B, maximal response of WT mGluR1␣ and its mutants to Glu at a saturating concentration. Each single data point was performed in an individual dish, and the cells expressed mGluR1␣ showing responses to Glu were selected for analysis (n ϭ 3).  (Fig. 9). Although Glu-238, located at the interface of the two protomers, has been reported as a functional Gd 3ϩ site, I120A, another mutation located at the interface of the receptor loses sensitivity to [Ca 2ϩ ] o (25). This indicates that the mutations at the interface of the two monomers cause markedly reduced activation of the receptor by either [Gd 3ϩ ] o or [Ca 2ϩ ] o and that the site at the hinge region is a true Gd 3ϩbinding site. This also highly suggests that our predicted Ca 2ϩbinding site can likewise bind Gd 3ϩ .
In view of these findings, we propose a working model of dual activation of mGluR1␣ by the two physiological activators, [Ca 2ϩ ] o and Glu, via their overlapping and interacting binding pockets at the hinge region of the ECD (Fig. 10). Increased concentrations of either Glu or [Ca 2ϩ ] o partially activate mGluR1␣. However, full activation of mGluR1␣ with maximal sensitivity and maximal amplitude of the response to Glu requires simultaneous binding of both Glu and Ca 2ϩ , with Asp-318 playing a key role in the synergy between the two agonists. In this sense, mGluR1␣ can be viewed as a "coincidence detector," requiring the binding of both ligands for maximal intracellular signaling.
Our proposed working model of mGluR1␣ is supported by previous studies from a number of groups working on both cultured cells and native brain tissue (6). Francesconi and Duvoisin (38) reported that mGluR1␣ in transfected cells is activated by [Ca 2ϩ ] o in the absence of Glu, indicating that mGluR1␣ is a Ca 2ϩ -sensing receptor. Kubo et al. (6) showed that [Ca 2ϩ ] o , as well as Glu, triggers intracellular responses in cultured cells and oocytes expressing mGluR1, mGluR3, and mGluR5. In terms of studies on endogenous mGluRs in native neurons, Tabata et al. (10)  sponses (6) and lowers sensitivity to Glu (39). Our data show that S166A maintains sensitivity to [Ca 2ϩ ] o albeit with reduced responsiveness (Fig. 4C). This residue is located at the ECD hinge joint but away from our predicted site in the flexible hinge region (Fig. 1B). The observed decrease in mGluR1␣ sensitivity observed with S166D could be a result of tuning the Ca 2ϩ binding affinity and Ca 2ϩ -induced conformational change by altering electrostatic interactions around the predicted Ca 2ϩ -binding site (40). As shown in Fig. 1C, the predicted Ca 2ϩ -binding residue Asp-322 is conserved in group I mGluRs. Glu-325 is highly conserved in group I (mGluR1 and mGluR5) and group II (mGluR2 and mGluR3) mGluRs (Fig. 1C). Interestingly, mGluR5 in group I and mGluR3 in group II sense [Ca 2ϩ ] o at physiological levels, whereas mGluR2 is activated only when [Ca 2ϩ ] o is more than 10 mM (6). On the basis of our observation that E325I abolished Ca 2ϩ -induced responses for mGluR1␣ but retained responsiveness to Glu, we concluded that Glu-325 might be very important for Ca 2ϩ binding in the mGluR family generally. Analysis by the Contacts of Structural Units server indicates that Glu-325 interacts electrostatically with Arg-297 and Arg-323 in the Glu-bound structures of mGluR1␣ (41), suggesting that Glu-325 stabilizes the local structure through an electrostatic interaction. As revealed by the grafting approach ( Fig. 2A), E325I significantly reduces the Ca 2ϩ -binding ability in mGluR1␣, possibly by disturbing the favorable local charge environment. Fig. 1C shows that Asp-318 in mGluR1␣, located at the hinge region, is conserved in all members of the three GPCRs, corresponding to Asp-295 of mGluR2, Asp-301 of mGluR3, Asp-309 of mGluR8, and Glu-297 of CaSR (23,24,42). Figs. 4 and 6 clearly demonstrate that Asp-318 contributes not only to Ca 2ϩbut also Glu-triggered [Ca 2ϩ ] i responses. This residue seems to play an essential role in the activation of mGluRs. Consistent with this finding, a D318A mutation was shown previously to reduce receptor expression on the membrane and abolish Glutriggered [Ca 2ϩ ] i and inositol trisphosphate responses (25).
Our findings also appear to be applicable to other members of the three GPCRs, especially CaSR. The mutation E297I in CaSR, equivalent to D318I in mGluR1␣, impairs receptor activation (23,24,42). Glu-297 is an important residue in our reported Ca 2ϩ -binding site in the CaSR hinge region (23,24) (42) have shown that the missense mutations E297K and Y218S significantly reduce the maximal response of the CaSR. Although E297K was considered a key factor in impairing protein folding, thus leading to lower expression on the cell surface and impaired responsiveness to [Ca 2ϩ ] o , our unpublished data 3 show that E297I has a membrane expression level comparable with that of the wild type CaSR. Therefore, the low expression level of E297K could be, at least in part, the result of the substitution of an unfavorable positive charge, which modifies the local charge balance, leading to reduced folding efficiency. Furthermore, our assessment of the surface expression of D318I by flow cytometry showed that it was at the same level as wild type; this echoes the impact of mutating the equivalent residue in CaSR (e.g. Glu-297). It has been postulated that residues Ser-170, Asp-190, Gln-193, Ser-296, and Glu-297 are critical for Ca 2ϩ binding to CaSR and functionality of the receptor (42), which is in excellent agreement with our prediction. In addition, CaSR functions primarily as a [Ca 2ϩ ] o -sensing receptor but can also integrate information about protein metabolism (i.e. amino acids) with that of divalent cations (e.g. calcium) (46). CaSR displays sensitivity to amino acids, especially phenylalanine and other aromatic amino acids, likely via three serine residues (Ser-169 -Ser-171) at a site corresponding to the Glu-binding pocket in mGluR1␣. The double mutation T145A/S170T specifically abolishes CaSR responsiveness to amino acids while leaving [Ca 2ϩ ] o sensing intact (47). Our unpublished data 5 show that Ca 2ϩ and phe-3 C. Zhang and J. Yang, unpublished work.  nylalanine also synergistically modulate the signaling functions of CaSR.
In summary, we have predicted and confirmed experimentally a calcium-binding site in the extracellular domain of mGluR1␣. We have also shed new light on the co-activation of mGluR1␣ by Glu and [Ca 2ϩ ] o . These findings provide novel perspectives on mGluR1␣, which may be viewed as capable of integrating information from two very different types of ligands (an amino acid neurotransmitter and a divalent cation). The levels of [Ca 2ϩ ] o in the brain are highly dynamic (37), and the affinity constants that we have determined in our studies on calcium binding to mGluR1␣ are well within the dynamic, physiological range of [Ca 2ϩ ] o in the brain. For family 3 GPCRs other than CaSR, the physiological importance of [Ca 2ϩ ] o binding has been uncertain; but the findings reported here may be useful in resolving this mystery by allowing for the development of knock-in mutations to mGluR1␣, and resultant mouse models, that disrupt the ability of the receptor to bind [Ca 2ϩ ] o while leaving Glu binding intact. Moreover, because many of the key calcium-binding residues defined in our studies are conserved for other family 3 GPCRs, our findings may have relevance for a host of other receptors beyond just mGluR1␣ (6, 7). Family 3 GPCRs have tremendous potential as therapeutic targets, and therefore the advances described here may facilitate the development of novel family 3 GPCR-targeted drugs for use in the treatment of many different diseases.