Cryptic dimer interface and domain organization of the extracellular region of metabotropic glutamate receptor subtype 1.

Previously, we produced the whole extracellular region of metabotropic glutamate receptor subtype 1 (mGluR1) in a soluble form. The soluble receptor retained a ligand affinity comparable with that of the full-length membrane-bound receptor and formed a disulfide-linked dimer. Here, we have identified a cysteine residue responsible for the intermolecular disulfide bond and determined domain organization of the extracellular region of mGluR1. A mutant, C140A, was a monomer under nonreduced conditions by SDS-polyacrylamide gel electrophoresis; however, C140A was eluted at the position similar to that of mGluR113, the wild type soluble receptor, by size exclusion column chromatography. Furthermore, C140A bound a ligand, [(3)H]quisqualate, with an affinity similar to that obtained by mGluR113. Oocytes injected with RNA for full-length mGluR1 containing C140A mutation showed responses to ligands at magnitudes similar to those with wild type full-length RNA. Thus, elimination of the disulfide linkage did not perturb the dimer formation and ligand signaling, suggesting that cryptic dimer interface(s) possibly exist in mGluR1. Limited proteolysis of the whole extracellular fragment (residue 33-592) revealed two trypsin-sensitive sites, after the residues Arg(139) and Arg(521). A 15-kDa NH(2)-terminal proteolytic fragment (residue 33-139) was associated with the downstream part after the digestion. Arg(521) was located before a cysteine-rich stretch preceding the transmembrane region. A new shorter soluble receptor (residue 33-522) lacking the cysteine-rich region was designed based on the protease-sensitive boundary. The purified receptor protein gave a K(d) value of 58.1 +/- 0.84 nm, which is compatible to a reported value of the full-length receptor. The B(max) value was 7.06 +/- 0. 82 nmol/mg of protein. These results indicated that the ligand-binding specificity of mGluR1 is confined to the NH(2)-terminal 490-amino acid region of the mature protein.


Introduction
L-glutamate is a major neurotransmitter of excitatory synapses in mammalian brains.
Metabotropic glutamate receptors (mGluRs) 1 are thought to modulate synaptic neurotransmission (1). Eight subtypes of mGluRs and the calcium sensing receptor (CaR) (2) share amino acid sequence homology. mGluRs possess seven transmembrane segments and are considered to be G-protein coupled receptors (GPCRs). mGluRs contain a large extracellular region (~600 amino acids) in contrast to ordinary GPCRs. The NH 2 -terminal extracellular regions have amino acid sequence similarity (3) to that of Leucine-, Isoleucine-, Valine-binding protein (LIVBP) (4), one of the bacterial periplasmic binding proteins. The LIVBP-like region is followed by a cysteine-rich region that precedes the first transmembrane segment. LIVBP-like regions are also found in the GABA B receptor (5), the putative pheromone receptor (6,7). The cysteine-rich region is not shared by the bacterial binding proteins and the GABA B receptor. These receptor proteins which contain a large extracellular region similar to that of mGluR have been designated as family 3 GPCR (8). Thus, structural resemblance among the family 3 receptors has been predicted.
Previously we were able to express the whole extracellular region of mGluR1 in a soluble form (9). The receptor protein secreted into the culture medium retained the ligand binding affinity and selectivity comparable with those of the full-length membrane bound receptor.
Interestingly the soluble receptor of mGluR1 was a cysteine-linked dimer. Dimer or oligomer forms of other subtypes of mGluRs, mGluR5 (10) and mGluR4 (11), or of CaR (12,13) have been also reported. However, the precise mechanism of ligand binding to the mGluR is unknown. It is still unclear whether glutamate binds to mGluRs in an analogous way to the Venus' flytrap model proposed in the bacterial binding proteins (14), whereby the ligand bound to one lobe is trapped to another lobe by the bending motion of the hinge region.
5 Role(s) of dimerization of mGluR1 in ligand binding and signal transmission remain to be elucidated.
The significance of dimerization in signal transduction of single-span transmembrane receptors for growth factors or cytokines has been studied extensively (15). The main focus is on whether structural details of extracellular regions have an influence on intracellular signaling (16,17). For GPCRs, hetero-or homo-dimer formation has been reported in a few cases such as some subtypes of dopamine receptors (18), β2-adrenergic receptor (19) and muscarinic acetylcholine receptors (20). Domain swapping among transmembrane segments has been proposed to be a causative factor in self-association of adrenergic receptors (21).
Technical difficulties have been accompanied with those studies because of seven hydrophobic transmembrane segments. However, more convincing evidence has recently emerged of dimer formation in GPCRs. A heterodimeric opioid receptor dimer with an affinity and ligand selectivity different from those of the homodimer has been reported (22). Dimerization in the intracellular region is also reported to be essential for functional expression of the GABA B heterodimer receptor (23)(24)(25). These results suggest a possibility that dimerization of GPCR is not a rare phenomenon but rather a common occurrence.
In this study, with the advantage of purified dimerized material, we have determined a cysteine residue responsible for the intermolecular disulfide bonding by amino acid point mutation. Effects of disruption of the disulfide bond on ligand binding and signaling were examined. The existence of a cryptic dimer interface in mGluR is discussed. In order to delineate ligand binding core of mGluR1, we furthermore defined the domain organization of the extracellular region by proteolysis experiments.  (9). Briefly, after being infected with baculoviruses for soluble receptors, High Five cells were incubated for four or five days.

Construction of transfer vectors for expression of mGluRs in insect cells
Both transfer vectors, pVLmGluR113 and pVLmGluR114, were constructed from pmGluR108 (9). pVLmGluR113 was made as follows. A set of complementary oligonucleotides, HJ110 (5'-CACAGGCTGTGAGCCCATTCCTGTCCGTTATCTTGAGTGGAGTGACATAGAATAG TGAT-3') and HJ111 (5'-CTAGATCACTATTCTATGTCACTCCACTCAAGATAACGGACAGGAATGGGCTCAC AGCCTGTGAGCT-3'), were annealed and cloned into the SacI/XbaI-digested fragment of pmGluR108. pVLmGluR114 was made as follows. Polymerase chain reaction (PCR) was done with primers TO1 (5'-CATCAATGCCATCTATGCCATGGC-3') and HJ112 (5'-TCTAGATTACTAAGATCGTACCATTCCGCTTTTGTTC-3') using pmGluR108 as a template. TO1 contains an NcoI site. HJ112 contains an XbaI site. The PCR product was digested with NcoI and XbaI and was cloned into the NcoI/XbaI-digested fragment of pmGluR108. Transfer verctors for mutant receptors were made as follows. Mutants in the wild type background are denoted with the number of the residue and amino acid substituted.

ATCAGGCAGGGCTCGGTTCAGC-3'). Following the protocol of Mutan-Super Express
Km Kit (TAKARA), the plasmids pKFC67A, pKFC109A and pKF140A, which contain the 9 mutated sequences, were obtained. 391 bp NotI-EcoRI fragments of pKFC67A and pKFC109A and a 390 bp EcoRI-AatI fragment were ligated into the NotI/EcoRI-digested pVLmGluR113 and EcoRI/AatI-digested pVLmGluR113, yielding pVLmGluR113C67A, pVLmGluR113C109A and pVLmGluR113C140A. All the PCR products were fully sequenced. These transfer vectors were used for transfection to insect cells with BaculoGold Baculovirus DNA.
Color development was done by a commercial detection kit (Promega).

Native PAGE
Samples were analyzed by 2-15% or 4-20% linear gradient polyacrylamide gels not containing SDS in the gel matrix. SDS was not included in either the running buffer or the sample buffer. Gels were siver stained. Molecular weight standards were from Daiichi Pure Chemicals.

Purification of the soluble receptors
A conditioned medium of receptor-producing High Five cells was concentrated to 10-50 fold in the presence of protease inhibitor mixtures (1 mM of PMSF, 5 µg/ml of leupeptin, 10 µg/ml of benzamidine, 10 µg/ml of trypsin inhibitor and 1 µg/ml of aprotinin) and washed by immunoaffinity column which was conjugated with MAb mG1Na-1 as described previously (9). The column was washed by 30 ml of 10 mM Hepes, pH 7.5, containing 200 mM NaCl, and the bound material was eluted with 100 mM CAPS, pH 11, containing 200 mM NaCl.
Eluate was neutralized with 2 M Hepes pH 7.5. Aliquots of the eluate were analyzed by

Ligand binding
Ligand binding was performed with the polyethylene glycol (PEG) precipitation method as described previously (28). Briefly, 20 nM or 40 nM of [ 3 H]quisqualate (323 GBq/mmol) (a gift from BANYU Tsukuba Research Institute) and soluble receptor samples were mixed in was analyzed by the software of Prism II (Graphpad Software, San Diego, CA). Saturation binding curves were fitted to a one-site binding model and K d and B max values were calculated.

Expression of mGluRs in oocytes and electrophysiology
A 777 bp SacII-AatI fragment of pKFC140A and a 3.2 kbp AatI-SacII fragment of pmGR1 (29) and a 3.7 kbp SacII-SacII fragment of pmGR1 were ligated. The resulting plasmid, pmGR1C140A, and pmGR1 were used as templates for in vitro transcription to yield complementary RNA (cRNAs). The plasmid DNA was linealized by NotI and capped cRNA was synthesized with MEGAscript T7 kit (Ambion, Austin, TX). Xenopus laevis oocytes were prepared according to the standard procedure. The follicular cell layer was removed by treatment with Ca 2+ -free ND solution (96 mM NaCl, 2 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 and 5 mM Hepes, pH 7.6) containing 100 µg/100 ml collagenase-B (Yakuruto, Tokyo, Japan) for 1.5 h at 18 ˚C. The oocytes were then injected with 10 ng of cRNAs. The injected oocytes were incubated in ND solution for 1-2 d. A two-electrode voltage clamp system was used to measure glutamate and quisqualate induced chloride ion currents. The membrane potential was held at -60 mV in all measurements. All currents were filtered at 1 kHz, and data were stored and anlyzed using MacLab system (Bio-Research-Center, Nagoya, Japan). All measurements were conducted at room temperature.

Cross-linking procedure
The cross-linker, EGS, was dissolved at 20 mM in dimethylsulfoxide and diluted into the protein solution to give the final concentration of 1 mM. The reaction proceeding at 25 °C for

Low angle rotary shadowing
The purified mGluRs in 50 mM Hepes, pH 7.5, was equilibrated with glycerol (up to 50% (v/v) ). Final concentration of the protein was 50 µg/ml each. 50 µl of the sample was sprayed onto mica surface cleaved freshly by using a painter's airbrush (Olympus Model SP-B, f 0.18 mm). Then, the sample on the mica was rapidly brought into a freeze-etching device equipped with a large turbo pump (FR 7000, Hitachi, Mito, Japan), dried for 10 min (room temperature) in vacuum (1x10 -6 Pa) and then cooled down to -100 ˚C. Subsequently, specimens were rotary shadowed with platinum by an electron gun positioned at an angle of 2.5˚ to the mica surface and followed by carbon evaporation. Shadowed films were removed from the mica by slowly soaking the mica into water, mounted on copper grids and examined under a JEOL 100CX electron microscope (JEOL Co., Ltd., Tokyo). Figure 1 shows a schematic view of the transfer vectors, pVLmGluR113, pVLmGluR113C67A, pVLmGluR113C109A, pVLmGluR113C140A and pVLmGluR114, used for making the recombinant viruses for mGluR113, mGluR113C67A, mGluR113C109A, mGluR113C140A and mGluR114. pVLmGluR113 encodes complementary DNA (cDNA) corresponding to the region Met 1 -Glu 592 of the primary amino acid sequence of mGluR1. pVLmGluR114 encodes cDNA corresponding to the region Met 1 -Ser 522 . pmGluR108 that was used in our previous study (9) encoded the amino acid region identical with that of pVLmGluR113 and six histidine codons at its COOH-terminus.

Results
In this study, a His tag-sequence of pmGluR108 was omitted in pVLmGluR113.

Recombinant viruses for the receptors were prepared as described in "Experimental
Procedures". mGluR113 encodes the whole extracellular region of mGluR1. Amino acid mutation was introduced at the first three consecutive cysteine residues (Cys67, Cys109 and Cys140). The mutant receptor proteins were designated C67A, C109A and C140A. Concentrated culture medium of insect cells infected with the recombinant viruses for cysteine mutants as well as for mGluR113 was subjected to SDS-PAGE and analyzed by immunoblotting (Figure 2A).
In the presence of 20 mM dithiothreitol (DTT) (left pannel), the three mutants were detected as approximate 65 kDa bands. Under nonreduced conditions (right pannel), C67A and C109A was electrophoresed at around 130 kDa bands; however, C140A was electrophoresed much faster with a position corresponding to the monomer band. The sizes of C67A and C109A appeared significantly different from that of mGluR113, suggesting a conformational change. These results clearly showed that alanine mutation at Cys 140 eliminated the intermolecular disulfide bonding.
14 Next we compared ligand binding capacities between C140A and mGluR113. As concentrated culture medium of insect cells infected with the viruses for C140A showed ligand binding capacity comparable to that of mGluR113 (data not shown), we next purified C140A by an immunoaffinity column as described in "Experimental Procedures". Although the purified C140A sample showed a band equal to that of mGluR113 under reduced conditions ( Figure 2B left pannel), C140A migrated much faster than mGluR113 under nonreduced conditions (center pannel). Surprisingly, this mutated receptor migrated as a dimer on native-polyacrylamide gels as did mGluR113 (right pannel). We also observed a small amount of larger oligomers of C140A. These data indicated that mutation at Cys 140 eliminated disulfide bonding in the extracellular portion of mGluR1, but the extracellular portion of the C140A receptor folded in a manner that maintained noncovalent dimerization of the receptor molecule. To reinforce this result, C140A was subjected to gel filtration column chromatography in parallel with mGluR113 ( Figure 3). Calibration with molecular size markers showed the eluting position of mGluR113 to be around 185 kDa and that of C140A to be around 150 kDa. Both receptors were definitely larger than a monomer and seemed to be a dimer or a larger oligomer.
The ligand binding ability of C140A was quantitively examined by comparing the inhibition of [ 3 H]quisqualate binding to the purified C140A and mGluR113 with unlabeled glutamate and quisqualate ( Figure 4A). Both receptors did not make a great difference in dose-response curves of unlabeled glutamate and quisqualate. These data indicated that there was no great difference between mGluR113 and C140A in affinities to glutamate and quisqualate. Then we next examined the importance of the disulfide linkage in signal transmission of mGluR1 using the oocyte expression system ( Figure 4B). A messenger RNA (mRNA) which encodes a full-length of mGluR1 whose Cys 140 was mutated to alanine, was created. Oocytes injected with the mutated mGluR1mRNA as well as the wild type Next in order to understand the domain organization of the mGluR1 extracellular region, we first determined the NH 2 -terminal amino acid sequence of soluble receptor protein, mGluR113, that was generated by removal of a signal peptide and found that mGluR113 began with a residue Ser 3 3 . Proteinase digestion was then performed using the purified mGluR113 ( Figures 5A and 5B).  Figure 5B) and proceeded to the next experiments.
We examined the ligand binding capacity of trypsinized mGluR113. Surprisingly, trypsin digestion, at up to a ratio of 1/10 which is larger than that used in Figure 5, did not abolish ligand binding capacity of mGluR113, as shown in Figure 6A. Trypsinized mGluR113 did show [ 3 H]quisqualate binding comparable to that of undigested mGluR113. In order to know whether or not the 15 kDa NH 2 -terminal fragment (residue 33-139) is required for ligand binding, the trypsinized sample was loaded on a native polyacrylamide gel that did not contain SDS in the gel matrix ( Figure 6B). The trypsinized mGluR113 was electrophoresed at a very close position to that of the undigested mGluR113. To prove that the NH 2 -terminal fragment is attached to the remaining fragment, we performed a cross-linking experiment ( Figure 6C).  result is in contrast to our previous study (9), in which recombinant virus derived from pmGluR103 that encoded cDNA corresponding to the region Met 1 -Glu 492 of mGluR1, did not efficiently produce the soluble receptor protein, mGluR103. The COOH-teminal 30 amino acids of mGluR114 may be involved in a secondary structure and contribute to protein stability. Next to purify mGluR114, the 40-fold concentrated medium was applied on an immunoaffinity column as described in "Experimental Procedures". Then we further purified the material by Resource Q column chromatography. Figure 4B shows Coomasie staining of the purified material of mGluR114 and mGluR113. We also confirmed that the NH 2 -terminal amino acid sequence of mGluR114 is identical with that of mGluR113 (data not shown).
Using purified mGluR114 as well as mGluR113, a ligand binding study was performed.
mGluR114 and mGluR113 showed saturable binding as in Figures 8A and 8B . The Next we performed rotary shadowing of the soluble receptor proteins, mGluR114 and mGluR113 ( Figure 9). Both were globular proteins consisting of two similar components facing each other, consistent with the above finding that they were electrophoresed as dimers in the native gels. Remarkable difference in the shadowing image was not observed between mGluR114 and 113, implying that the LIVBP-like region can fold regardless of the presence or absence of the cysteine-rich region.

Discussion
Previously we have reported dimer formation of soluble forms of mGluR1 expressed in insect cells (9), in which the dimer was disulfide linkaged. In this investigation we have made cysteine to alanine mutants and identified the cysteine residue responsible for the  and CaR might not be exactly alike. Cysteine residues responsible for intermolecular disulfide dimer or oligomer formation of m3 muscarinic receptor, which is one of the classical GPCRs, have recently been reported (32). Interestingly, a mutant in which the two responsible cysteines located at the 2nd and 3rd extracellular loops were mutated to alanines, lost the intermolecular disulfide linkage but retained the capacity to form non-covalent receptor dimer or multimer.
In this study, we have also succeeded in outlining the domain structure of the extracellular region of mGluR1. A number of Arg and Lys residues are distributed in the extracellular region, nevertheless, trypsin digested mGluR113, which consists of the whole extracellular region of mGluR1, at the site of Arg 521 . Thus the extracellular region of mGluR1 was subdivided into two domains: an NH 2 -terminal LIVBP-like region and a cysteine-rich region preceding the first transmembrane segment. We call the former part the ligand binding domain (LBD) and the latter the cysteine-rich domain (CRD) hereafter. Furthermore five major fragments generated by partial tryptic digestion was assigned according to the two trypsin sites.
An extracellular fragment devoid of CRD, mGluR114, was expressed well and showed a K d value of 58.1 ± 0.84 nM for its ligand, [ 3 H]quisqualate, which is close to that of mGluR113. Both of the soluble receptors, mGluR113 and mGluR 114, showed dimer under nonreduced conditions and monomer under reduced conditions. Thus cysteine 140 responsible for the intermolecular disulfide bond resides within LBD. Because the trypsindigested mGluR113 electrophoresed at a position similar to that of the undigested one under a native condition, nick introduced by trypsin treatment seems not to interfere with the dimer interface. Interestingly, trypsin-digested mGluR113 as well as mGluR114 (data not shown) 20 retains ligand binding activity, suggesting that the digestion site, Arg 139 , is exposed to the surface and resides in a flexible region. The NH 2 -terminal 15 kDa fragment seems to be associated with the remaining part after trypsin digestion. Therefore we did not make an expression construct that lacked the corresponding cDNA sequence. Thus whether the 15 kDa fragment is required for ligand binding could not be determined. The 15 kDa region might have a role for folding LBD. We speculate that the ligand binding site forms a rather rigid structure.
Provided that two molecules of the ligand, quisqualate, bind to a dimer form of the soluble receptor, the stoichiometry of binding can be calculated to be approximately 39-44% based on the B max values of mGluR113 and mGluR114 for the ligand. Only half of the material might retain binding capacity. Or we may have been unable to measure maximal binding capacity because our PEG precipitation assay is not expected to resolve low affinity site, if any, with a K d in the micromolar range. However, if we assume that cooperative binding is operated in the dimer form of the mGluR1 or the two binding sites possess different affinities, this value of stoichiometry would be meaningful. Interestingly, it has been reported that a purified 42 kDa ligand-binding fragment of GluR-D, an α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor, one of glutamate-gated ion channels, showed a B max value of 6-12 nmol/mg, indicating 50% stoichiometry of the theoretical maximum (33).
Model buildings of part of the extracellular region of monomer mGluR1 were proposed on the basis of an atomic structure of a bacterial binding protein that resides in the periplasmic space (34,3). If the dimerized form is the principal form of active mGluR1, a novel mechanism may be at work in ligand binding and signal transmission in mGluR. Although the cysteine residue(s) responsible for the intermolecular disulfide bonding has been determined, the precise mechanism of the dimer formation has not been explored, and the character of the dimer interface still remains to be elucidated. Whether or not dimer formation is critical for ligand binding and signal transmission in mGluRs is a topic that should be examined further. Because the cryptic dimer interface which is different from that formed by the intermolecular disulfide bond in mGluR1 is sensitive to SDS detergent, electrostatic or hydrophobic interaction can be suspected to be a driving force for the dimer formation of mGluR1. Intriguingly, a stretch of 24 consecutive uncharged amino acids in the extracellular region of mGluR1 (residues 155 to 178) has been pointed out when the primary amino acid sequence was determined (29).
Whether or not CRD interacts with LBD or the transmembrane region is still unknown.
The absence of large differences between the rotary shadowing imagings of mGluR113 and mGluR114 suggests that CRD appears not to perform major roles in dimer formation.