Metabotropic glutamate receptor 5 is a disulfide-linked dimer.

The sequences of the metabotropic glutamate receptors (mGluRs) show little homology with other members of the G protein-coupled receptor family and exhibit several distinctive features, including a large N-terminal extracellular domain with 17 cysteines in conserved positions. Here we demonstrate that mGluR5, as well as other mGluRs, behave as species approximately twice as large as expected from their sequence, but reducing conditions cause a decrease to the predicted molecular mass. Co-immunoprecipitation experiments using wild type and epitope-tagged receptors demonstrate that this is due to specific, disulfide-dependent dimerization of the receptor. The intermolecular disulfide that mediates dimerization occurs in the extracellular domain, within about 17 kDa from the N terminus.

tors are homologous to the mGluRs (6,7); so together the Ca 2ϩ and glutamate receptors appear to constitute a unique subgroup of this supergene family. Receptor domains responsible for signal transduction also appear different. For example, the third intracellular loop is important for determining the specificity of G protein coupling in most G protein-coupled receptors examined, whereas the C-terminal end of the second intracellular loop is critical in the mGluRs (8 -10). Moreover, all mGluRs have a very large N-terminal extracellular domain (about 65 kDa in mGluR5), constituting about one-half of the protein, whereas most G protein-coupled receptors do not. The glutamate binding domain is believed to lie in this extracellular region (11,12), not within the bundle of membrane-spanning domains, as is typical of the rhodopsin-like receptors.
Another unique structural feature is that there are 21 conserved cysteine residues in all the mGluRs (13). Nineteen of these are in the N-terminal domain and extracellular loops. Nine of the cysteines are at the C-terminal portion of the extracellular domain, and this region has been compared with similar cysteine-rich domains of receptor tyrosine kinases (12). Although the function of these cysteines is unknown, the strict conservation of position implies that the function is a shared and important one for this family of receptors.
Although native iGluRs are thought to function as heteromeric pentamers (14 -16), the mGluRs, by analogy with other G protein-coupled receptors, have been assumed to be monomeric (1,4,17). However, this has not been directly demonstrated. Using biochemical and molecular techniques, here we demonstrate that mGluRs are not monomeric but are instead covalently linked dimers, bound by disulfide bonds between conserved cysteines in the N-terminal extracellular domain.

EXPERIMENTAL PROCEDURES
Antibodies, Western blots, and Immunoprecipitates-Antibodies to wild type (wt) mGluR5 were affinity-purified antipeptide antibodies raised against an immunogen that contained the C-terminal 13 amino acids as described (18). Antibodies to wt mGluR1a were affinity-purified antipeptide antibodies raised against an immunogen that contained the sequence of residues 1116 -1130 (i.e. EFVYEREGN-TEEDEL) of the rat mGluR1a (19). Both monoclonal and polyclonal anti-hemagglutinin (HA) antibodies were obtained from Babco (Berkeley, CA); the polyclonal antibody was used for immunoprecipitation, the monoclonal antibody for Western blots and immunocytochemistry. Antibodies to mGluR2-3 and mGluR4 were the generous gift of Dr. Thomas Knoepfel (CIBA, Basel, Switzerland). Preparation of brain tissue, electrophoresis (on 6 or 7.5% polyacrylamide gels), and transfers onto polyvinylidene difluoride membranes (Immobilon P; Millipore, Waters, MA) were as described (18). For preparation of membranes from transfected cells, cells were washed once in PBS, then subjected to one freeze-thaw cycle. They were scraped into lysis buffer (2 mM HEPES and 2 mM EDTA, pH 7.4, containing protease inhibitors) and homogenized in a glass homogenizer with a motorized Teflon pestle. The nuclear pellet (1000 ϫ g, 5 min) was discarded, and membranes were harvested after pelletting (35,000 ϫ g, 30 min). For immunoprecipitations, membranes were homogenized in PBS containing 0.5 or 1% SDS. The SDS extract was diluted 5-or 10-fold into PBS containing 0.5% dodecyl maltoside to sequester free SDS into mixed micelles, thereby permitting immunoprecipitation. Antireceptor antibody was added, and the mixture was incubated at 4°C overnight. Protein A-Sepharose (Sigma) was added, and the incubation continued for 2 h at room temperature on a rocking table. The protein A pellets were washed three times with PBS before elution with sample buffer and electrophoresis. The sample buffer always contained 2% SDS and, when indicated, 20 mM dithiothreitol (DTT). Samples were heated at 60°C for 3 min before electrophoresis.
Receptors and Cells-The cDNA fragments containing the full-length mGluR5 or mGluR1a coding sequences were ligated into pcDNA1neo (Invitrogen) downstream of the cytomegalovirus promoter (18). The HA-tagged mGluR5 mutant was constructed using recombinant polymerase chain reaction (20) with two sets of primer pairs. The first set (5Ј-CATGACGACCTTCGCAGAGAT-3Ј, nucleotides 3482 to 3501, GenBank accession number D10891; and 5Ј-ATCCTCTCCCAAATAT-GACACTTATCCATATGATGTTCCAGATTATGCT-3Ј, nucleotides 3720 -3741 followed by the HA epitope in bold) and the second (5Ј-TATCCATATGATGTTCCAGATTATGCTTGAGCCACT-GGAAACTTCCCT-3Ј, representing the HA epitope in bold followed by nucleotides 3781-3801; and 5Ј-CACACACGGTGGAGACATGAGCG-GCCGCTAAA, nucleotides 3897-3918 followed by a NotI restriction site) were used to amplify DNA fragments of 286 or 164 base pairs from wt-mGluR5. Both fragments were used as templates in another round of polymerase chain reactions using the two flanking primers, and the resulting fragments were then used to replace wt sequences in mGluR5. tHA was constructed using HA-tagged mGluR5 as a template together with primer sets fusing sequences within the first intracellular domain in frame with those of the HA-tagged C terminus. The 5Ј-primer set included 5Ј-GAAGTCAGCTGTTGTTGG-3Ј (identical to nucleotides 1843-1860) as well as 5Ј-CGAGTCCACCGAGTCTCTAGACTTGAC-CACCGGAGT-3Ј (complementary to nucleotides 2083-2097 and 3685-3668). The 3Ј-primer set consisted of the complement of the latter primer together with the NotI restriction site-containing primer described above. First and second round polymerase chain reaction was performed as described. The final product was digested with BstEII and NotI and then subcloned into wt-mGluR5 cut with the same enzymes. The resulting plasmid was termed tHA. Both the HA-tagged mGluR5 and tHA were confirmed by sequencing.
HEK cells at 80% confluency were transfected with 15 g of plasmid DNA using LipofectAMINE (Life Technologies, Inc). Forty-eight hours later membranes were prepared and immunoprecipitated with the indicated antibodies as described above. Immunohistochemical analysis of cotransfected cells was done on Lab-Tek chambered glass slides. Primary antibodies included the HA monoclonal antibody and anti-wt-mGluR5, followed by fluorescein isothiocyanate-labeled goat antimouse and CY3-labeled goat anti-rabbit (Jackson ImmunoResearch Labs, Inc.) secondary antibodies, respectively.

Metabotropic Glutamate Receptors Migrate at about Twice Their Predicted Molecular Mass under Nonreducing
Conditions-To examine the possible structural or conformational roles of the conserved cysteines in mGluRs, the electrophoretic mobility of mGluR5 (from rat cortical membranes) was examined under reducing and nonreducing conditions in SDS gels (Fig. 1A). In the presence of the reducing agent DTT or 2-mercaptoethanol (not shown), mGluR5 migrated at an apparent molecular mass of ϳ148 kDa. Deglycosylation with peptide N-glycosidase F reduced this to ϳ130 kDa (not shown), consistent with the size predicted from the primary sequence. However, in the absence of reducing agent, mGluR5 migrated at an apparent molecular mass of ϳ260 kDa (Fig. 1A). Because samples were prepared in the presence of the denaturing detergent, 2% SDS, most noncovalent interactions should have been eliminated.
One interpretation of this result is that mGluR5 is covalently attached by intermolecular disulfide bonds to another component of the membrane, but other possibilities must be considered. A different electrophoretic mobility in the presence of reducing agents may reflect an altered conformation due to cleavage of intramolecular disulfide bonds. However, such bonds usually promote compact structures that on reduction increase the Stokes radius and, hence, increase the apparent molecular mass of the protein. Another possibility is that spu-rious disulfide bonds may form between free sulfhydryls during SDS denaturation of the protein, leading to artifactual covalent association of mGluR5 with another protein. To test this, membranes were treated with iodoacetate to alkylate free sulfhydryls and then solubilized with SDS ( Fig. 1A, right lane). The receptor still behaved as a high molecular mass species, indicating that the disulfide bonds responsible for holding this species together were present prior to solubilization, as part of the native structure (Fig. 1A).
To verify this reduction-dependent alteration in molecular mass by an independent technique, Sephacryl S-400 gel filtration chromatography in the presence of SDS was used. For this experiment cortical membranes (2 mg of protein) were dissolved in PBS containing 1% SDS (Ϯ10 mM DTT), which also served as column buffer. Reduction led to a large decrease in apparent molecular mass of the receptor (Fig. 1B). Taken together, these results indicate that mGluR5 is covalently attached via disulfide bonds to another component(s) of the membrane.
This molecular mass shift was not unique to mGluR5. Under nonreducing conditions mGluR1a (another group 1 mGluR), mGluR2-3 (a Group 2 mGluR) and mGluR4 (a Group 3 mGluR) all migrated as species about twice as large as expected, with reduction causing a shift to the appropriate molecular mass (data not shown).
Metabotropic Glutamate Receptors Are Dimers under Nonreducing Conditions-What is the nature of the molecule to which mGluR5 is attached? The receptor may be bound by disulfide bridges to a distinct molecule; therefore, the high molecular mass species would be heteromeric, or alternatively, the receptor may be a homodimer.
When mGluR5 was expressed in HEK cells, it migrated as the high molecular mass form in nonreducing gels (not shown); this indicated that either the receptor forms homodimers, or that HEK cells endogenously express the mGluR5-associated protein.
To determine which of these hypotheses is correct, FIG. 1. mGluRs have a high apparent molecular mass under nonreducing conditions. A, Western blot analysis of rat cortical membranes (25 g of protein/lane) using an antibody directed to the C terminus of wild type mGluR5. DTT caused the receptor to shift to a lower apparent molecular mass. Pretreatment of the membranes with iodoacetate (IAA) did not alter the apparent molecular mass. B, Sephacryl S-400 column chromatography of SDS-solubilized rat brain membranes in the presence or absence of DTT indicate that mGluR5 migrates as a lower molecular mass species in the presence of DTT. Protein was measured spectroscopically (OD280), and mGluR5 was detected by assaying every other fraction by Western blotting. Column dimensions, 75 ϫ 1.5 cm.
cross-immunoprecipitation experiments were performed. A plasmid encoding an mGluR5 epitope tagged at the C terminus was constructed. Because the antibody we have used to recognize wt mGluR5 is directed toward the C terminus (18), the nucleotides coding for the wt C terminus were removed and replaced by the sequence encoding the HA epitope ( Fig. 2A; Ref.  21). After expression in HEK cells, HA-mGluR5 also behaved as a high molecular mass species in nonreducing gels, indicating that alteration of the C terminus did not disrupt formation of the disulfide-bound complex (not shown). If the receptors form covalent homodimers, some of the wt-mGluR5 and HA-mGluR5 may be expected to be found in the same dimer when both receptors are expressed in the same cell. If, however, each receptor is bound to an unidentified, distinct molecule to form the high molecular mass form, no wt-mGluR5-HA-mGluR5 heterodimers should be found.
Transient transfection of HEK cells with both plasmids led to uptake and expression of both receptors in the same individual cells (Fig. 2B). Membranes prepared from the cotransfected cells were solubilized in 0.5% SDS to disrupt noncovalent interactions between proteins. As expected, anti-wt did not immunoprecipitate any HA-tagged proteins from cells expressing only HA-mGluR5 (Fig. 2C), nor did anti-HA bring down wt-mGluR5 (data not shown). However, HA-mGluR5 was immunoprecipitated from extracts of cotransfected cells by either anti-HA or anti-wt antibody (Fig. 2C). Moreover, wt-mGluR5 was also immunoprecipitated from these extracts when either antibody was used (not shown). Thus, these data indicate that mGluR5 polypeptides form dimers.
To determine the specificity of dimer formation, we performed an analogous experiment in which wild type mGluR1a was cotransfected with HA-mGluR5. If the assembly of mGluR dimers is specific, mGluR1a and HA-mGluR5 should not form heterodimers. As shown in Fig. 3, antibody selective for mGluR1a did not immunoprecipitate any HA-containing bands from cells transfected with mGluR1a, mGluR1a and HA-mGluR5, or HA-mGluR5 (Fig. 3, top gel, left three lanes), but it did immunoprecipitate mGluR1 from cells transfected with mGluR1a or mGluR1a and HA-mGluR5 (Fig. 3, bottom gel, left two lanes). Similarly, the antibody selective for HA did not immunoprecipitate any mGluR1a from cells transfected with mGluR1a, mGluR1a and HA-mGluR5, or HA-mGluR5, but it did immunoprecipitate HA-mGluR5 from cells transfected with HA-mGluR5 or HA-mGluR5 and mGluR1a. Thus, despite the 60% amino acid identity between mGluR1a and mGluR5, (22), they do not heterodimerize. These data indicate that there is great specificity in the assembly of the metabotropic receptor dimers.
Metabotropic Glutamate Receptors Are Linked via Their Nterminal Extracellular Domains-To localize which part of mGluR5 was involved in dimer formation, two types of experiments were performed.
In the first set of experiments, a mutant receptor, truncated after the first transmembrane domain and tagged with the HA epitope at the C terminus, was constructed (tHA; Fig. 2A). When this mutant receptor was expressed in HEK cells and immunoprecipitated with anti-HA, it migrated during electrophoresis as a dimer (160 kDa) under nonreducing conditions and as a monomer (doublet of 80 -90 kDa) under reducing conditions (Fig. 4A). Therefore, the locus for disulfide-mediated dimerization is in the N-terminal half of the receptor, most of which is extracellular.
When the tHA receptor was co-expressed with wt-mGluR5 and then immunoprecipitated with antibody to HA, an additional HA-containing band (220 kDa) was observed on the Precipitated products were reduced using 20 mM DTT and resolved on a 6% SDS-polyacrylamide gel. Separated products were transferred to a polyvinylidene difluoride membrane and probed with anti-HA monoclonal antibody followed by enhanced chemiluminescence. Molecular mass markers are shown on the right. Antibody to the wt receptor immunoprecipitated HA receptor from cotransfected cells, indicating that heterodimerization occurred. FIG. 3. mGluR1a and mGluR5 do not form heterodimers. Membranes from cells expressing wt-mGluR1a (1a), HA-mGluR5 (HA), or both (1a & HA), were treated with SDS and immunoprecipitated with anti-mGluR1a (1a) or anti-HA (HA) polyclonal antibodies. Precipitated products were reduced using 20 mM DTT and resolved on a 6% SDSpolyacrylamide gel. Separated products were transferred to a polyvinylidene difluoride membrane, and the blots probed with anti-HA monoclonal (top blot) or anti-mGluR1a polyclonal (bottom blot) antibodies followed by enhanced chemiluminescence. Molecular mass markers are shown on the left (201 and 120 kDa). Antibody to the mGluR1a precipitated mGluR1a, but not HA-mGluR5, and anti-HA immunoprecipitated HA-mGluR5, but not mGluR1a. Aliquots of each cell extract (no immunoprecipitation) are in the right three lanes on each gel. ippt AB, immunoprecipitation antibody. nonreduced gels, suggesting the formation of heterodimers between truncated-HA and wt receptors (Fig. 4A). In agreement with this interpretation, antibody to the wt receptor immunoprecipitated the 220-kDa heterodimer, but not the 160-kDa tHA homodimer, from cotransfected cells. The wt antibody did not precipitate any HA-containing species from cells transfected with only the wt or tHA mutant. Taken together, these results indicate that the truncated receptor forms both homodimers and heterodimers.
In the second experiment addressing location of the disulfide bond(s), intact HEK cells expressing the wt receptor were incubated with trypsin for various periods. We reasoned that proteolytic removal of all or part of the extracellular domain would generate a receptor fragment that does not dimerize, and from the size of this fragment we could infer the approximate location of the relevant cysteine(s). Since the wt antibody is directed toward the C terminus, all the proteolysis products observed will necessarily have intact C termini and loss of some length of the N terminus. Treatment with trypsin removed only a small fragment from the N terminus of the receptor, decreasing the apparent molecular mass by about 17 kDa (Fig. 4B,  ϩDTT). This very limited digestion suggests that access of the protease to potential cleavage sites was restricted by steric factors due to the secondary structure of the extracellular domain. However, even very short periods of proteolysis removed the site of dimerization, since the proteolyzed receptor migrated at the monomer molecular mass under nonreducing conditions (Fig. 4B, ϪDTT). These results indicate that the cysteine(s) responsible for disulfide bond formation are in the N-terminal 17 kDa of mGluR5.

DISCUSSION
The experiments described here clearly show that mGluR5 normally exists as a dimer on the plasma membrane. Dimerization is mediated via a cysteine or cysteines located within 17 kDa from the N terminus in the extracellular domain. Since heterodimers between mGluR5 and mGluR1a do not form, but the truncated mGluR5 containing only the extracellular region and one transmembrane domain does dimerize, the information providing the specificity of mGluR5 dimerization also resides in the N-terminal region of the molecule. Because mGluR1a, 2-3 and 4 also migrate as dimeric species, we propose that dimerization may be a general property of the mGluR family.
Several authors who have examined mGluRs using Western blot analysis have noted the presence of high molecular mass aggregated forms of the receptors (23, 24), even in the presence of reducing agents. In the most extreme cases these aggregates are so large that the receptor polypeptides do not enter the gel. This aggregation may reflect strong interactions among the denatured, highly hydrophobic, multiple transmembrane domains in mGluRs. In our hands, aggregation is avoided by heating samples minimally before electrophoresis (60°C, 3 min) and by using only SDS or dodecyl maltoside as solubilizing detergents. We believe the dimer we describe is the native form of the receptor and not an artifact because: 1) it is a discrete band with a characteristic and appropriate molecular mass and not a smear, as aggregates usually are; 2) it can be converted to the monomer by reducing agents, whereas mGluR aggregates cannot; 3) even under conditions that minimize aggregation, as described above, we never observed any monomer except when samples were reduced (or proteolyzed as in Fig. 3B), suggesting that all the receptor is initially present in the membrane as dimer; and 4) the truncated mutant (tHA), which contains only a single transmembrane domain and, hence, should exhibit little tendency to aggregate, also migrated as a dimer (Fig. 3A).
To date, there is no evidence that other G protein-coupled receptors exist as covalent dimers. The ␣1-adrenergic receptor of the rat ventricle behaves as a 77-kDa species in the presence or absence of DTT (25). Similarly, the electrophoretic mobility of the neuromedin B receptor is unaffected by DTT (26). A cholecystokinin receptor and an opioid receptor apparently increase in molecular mass when treated with DTT (27,28), suggesting that disulfide bonds are maintaining compact conformations of these polypeptides. It is unlikely, therefore, that covalent, disulfide-dependent dimerization is a universal feature of G protein-coupled receptors. This structural feature may be unique to the mGluRs and perhaps the related Ca 2ϩsensing receptors (6,7). It is worth noting that the Ca 2ϩsensing receptors do have the conserved cysteines characteristic of this subgroup.
The N-terminal extracellular domain of the mGluRs is related to the bacterial periplasmic binding proteins (PBPs) (12), as are extracellular domains of some iGluRs (12,29). PBPs constitute a family of proteins involved in high affinity transport of amino acids, sugars, and other nutrients into bacteria. Three-dimensional crystal structures reported for several PBPs have indicated that these proteins are composed of two distinct globular domains with a ligand binding cleft between them (30, 31). Recently, O'Hara et al. (12) proposed a threedimensional model of the structure of mGluR1a based on PBP structural information. Their alignment permitted them to FIG. 4. The disulfide bond responsible for dimerization is in the N-terminal region of mGluR5. A, Western blot analysis of immunoprecipitates (ippt Ab) from SDS-solubilized membranes prepared from cells expressing wt, tHA, or both, using monoclonal anti-HA to visualize the immunoreactive bands. Antibody to the wt receptor did not immunoprecipitate any HA-positive bands from cells expressing only wt or tHA receptors but did precipitate the 220-kDa band present in co-transfected cells. Antibody to HA immunoprecipitated the 160-kDa band from tHA-expressing cells and both 220-and 160-kDa proteins from co-transfected cells. On reducing gels, all immunoprecipitated bands behaved as monomers. B, intact HEK cells expressing wt mGluR5 were treated with trypsin (standard 0.05% trypsin and 0.53 mM EDTA in Hank's balanced salt solution from Life Technologies, Inc.) for the times indicated, and then membranes were prepared and electrophoresed under reduced and nonreduced conditions. Trypsin treatment led to a ϳ17-kDa decrease in the apparent molecular mass of the monomer (reduced, top), and a loss of dimerization (nonreduced, bottom). make several successful predictions concerning the glutamate binding site of the receptor, lending credence to the structural comparison. Our results indicate that the cysteine responsible for dimerization of mGluR5 is in the N-terminal 17 kDa. Within this region, there are four cysteines, all but one of which is conserved among the mGluRs. Based on the alignment of mGluRs and PBPs suggested by O'Hara et al. (12), it is conceivable that two of these cysteines are involved in intramolecular disulfides, leaving the remaining cysteine(s) available for intermolecular interactions.
Receptors with intracellular tyrosine kinase domains dimerize on ligand binding, and this is critical for signal transduction (32)(33)(34). Our evidence suggests that the cysteines responsible for mGluR dimerization are not those present in the cysteinerich tyrosine kinase receptor-like domain (12). Dimerization of tyrosine kinase receptors brings the monomeric receptors into close proximity, allowing each member of the pair to phosphorylate the other, thereby providing the binding sites necessary for initiating assembly of the signal-transducing apparatus on the intracellular face of the membrane. These active tyrosine kinase receptor dimers are linked noncovalently, whereas mGluR dimers are covalently bound.
Earlier studies of adrenergic receptors emphasized the role of two disulfide-linked, conserved cysteine residues in the first and second extracellular loops. For example, in the mammalian ␤ 2 -adrenergic receptor maintenance of this extracellular disulfide bridge is important for high affinity agonist binding and function (35,36). In contrast, DTT had no effect on binding of ligands to muscarinic receptors (37) or prostaglandin E 2 receptors (38) but potentiated binding to and functioning of H1 histamine receptors (39 -42). Therefore, one cannot predict a priori what effect reducing conditions will have on receptor functioning. We have preliminary results indicating that maintenance of extracellular disulfides of mGluR5 are critically important for maintenance of signal transduction through this receptor. 2 Consistent with this result, Vignes et al. (43) showed that inositol phosphate production stimulated by glutamate in synaptoneurosomes was blocked by DTT, whereas that stimulated by carbachol was not. It is interesting that N-methyl-Daspartate receptors are influenced oppositely by redox state: extracellular reduction leads to potentiation of receptor function, not inhibition (44 -46). Perhaps ambient redox conditions lead to a complementary and coordinate regulation of iGluRs and mGluRs.