Structures of the extracellular regions of the groupII/III metabotropic glutamate receptorsTakanori Muto*, Daisuke Tsuchiya†, Kosuke Morikawa‡, and Hisato Jingami§

Biomolecular Engineering Research Institute, 6-2-3 Furuedai, Suita, Osaka 565-0874, Japan

Communicated by Shigetada Nakanishi, Osaka Bioscience Institute, Osaka, Japan, January 4, 2007 (received for review November 20, 2006)

Metabotropic glutamate receptors play major roles in the activationof excitatory synapses in the central nerve system. We determinedthe crystal structure of the entire extracellular region of the group IIreceptor and that of the ligand-binding region of the group IIIreceptor. A comparison among groups I, II, and III provides thestructural basis that could account for the discrimination of group-specific agonists. Furthermore, the structure of group II includes thecysteine-rich domain, which is tightly linked to the ligand-bindingdomain by a disulfide bridge, suggesting a potential role in transmit-ting a ligand-induced conformational change into the downstreamtransmembrane region. The structure also reveals the lateral interac-tion between the two cysteine-rich domains, which could stimulateclustering of the dimeric receptors on the cell surface. We propose ageneral activation mechanism of the dimeric receptor coupled withboth ligand-binding and interprotomer rearrangements.

crystallization � cysteine-rich region � ligand binding � G protein-coupledreceptor

The metabotropic glutamate receptor (mGluR) is a G protein-coupled receptor (GPCR) that transduces extracellular (EC)

signals into G protein activation through biomembranes. ThemGluR molecule belongs to family C, which includes the GABABreceptor, the calcium-sensing receptor, and some taste and pher-omone receptors (1). The family C GPCR protomer has a large ECregion that recognizes a specific ligand molecule. L-glutamate (Glu)is a principal excitatory neurotransmitter in the central nervesystem, where glutamate receptors play an essential role in regu-lating synaptic activity. The mGluRs are categorized into threegroups, which comprise eight subtypes (2, 3). Each of the groups hasa specific regional distribution in the brain and displays a distinctpharmacological profile. Therefore, structural information aboutthe receptor–ligand interaction is essential for drug development.

An mGluR protomer consists of three distinct regions: the EC,transmembrane (TM), and intracellular regions (Fig. 1A). The ECregion is further divided into two parts: the N-terminal ligand-binding region, which is frequently called the ‘‘Venus flytrapmodule,’’ and the cysteine-rich (CR) domain, which intervenesbetween the ligand-binding and TM regions. Previously, we re-ported the crystal structures of the ligand-binding region of groupI mGluR subtype 1 (mGluR-I1) (4, 5). The structures revealed thatdimeric molecules could undergo quaternary structure changes,depending on the bound ligands. Furthermore, we proposed thatthe structure is in dynamic equilibrium, where the ratio between theactive and resting conformations is modulated by the presence/absence of ligand. In response to our structural analyses, the initialactivation mechanism of mGluR has been investigated by severalbiochemical methods (6–15). However, many issues still remainunresolved. In particular, the absence of structural informationabout the CR domain obscures the concrete and conformationalviews of the coupling mechanism between the ligand-binding andTM regions.

Here we report the crystal structures of the entire EC region ofthe group II mGluR subtype 3 (mGluR-II3), complexed withvarious agonists. This x-ray analysis revealed the structure of the CRdomain, which is unique to the family C GPCRs (16). We alsodescribe the crystal structure of the ligand-binding region of the

group III mGluR subtype 7 (mGluR-III7). A comparison amongthe known mGluR structures, including those of mGluR-I1, pro-vides a structural basis for the discrimination of the group-specificagonists. Furthermore, the crystal structure of the group II receptorreveals interaction between the two CR domains, thereby tying upthe two adjacent dimeric receptors with the R-conformation.Finally, the activation mechanism caused by both ligand-bindingand receptor clustering is proposed, in accordance with the variousagonist/antagonist-bound mGluR structures and previous experi-mental results.

ResultsOverall Architecture of the mGluR-II3 EC Region. The crystal of the ECregion contains two protomers in the asymmetric unit. Because ofthe noncrystallographic two-fold symmetry in the dimer, theirinternal structures are essentially identical. The overall architectureof the dimer complexed with Glu is shown in Fig. 1B. TheN-terminal part (residues 25–508) consists of the ligand-bindingregion. Its general architecture is similar to that of mGluR-I1 (4)and is divided into two domains, LB1 and LB2 (Fig. 1A). Becausethe bound agonist intervenes between the two domains (Fig. 1B),the entire ligand-binding region adopts the closed conformation(4). Although the agonist is bound to both of the protomers, theinterprotomer interface resembles that of the R-conformationobserved in mGluR-I1 complexed with an antagonist (5), ratherthan that of the A-conformation for the agonist-bound form (4).Therefore, the overall conformational state is designated as closed–closed/R, which was not observed in the previous analyses (4, 5).

CR Domain. The present structure includes the CR domains(residues 509–575) at the C terminus, which is followed by theTM and intracellular regions in the full-length receptor (Fig.

Author contributions: T.M., D.T., K.M., and H.J. designed research; T.M. and D.T. performedresearch; T.M. and D.T. analyzed data; and D.T., K.M., and H.J. wrote the paper.

The authors declare no conflict of interest.

Abbreviations: mGluR, metabotropic glutamate receptor; GPCR, G protein-coupled recep-tor; Glu, L-glutamate; EC, extracellular; TM, transmembrane; CR, cysteine-rich; VDW, vander Waals; DCG-IV, (2S,2�R,3�R)-2-(2�,3�-dicarboxycyclopropyl)glycine; 1S,3S-ACPD, (1S,3S)-1-aminocyclopentane-1,3-dicarboxylic acid; 1S,3R-ACPD, (1S,3R)-1-aminocyclopentane-1,3-dicarboxylic acid; 2R,4R-APDC, (2R,4R)-4-aminopyrrolidine-2,4-dicarboxylate.

Data deposition: The atomic coordinates and the structure factors have been deposited inthe Protein Data Bank, www.pdb.org [PDB ID codes 2E4U (mGluR-II3, Glu), 2E4V (mGluR-II3,DCG-IV), 2E4W (mGluR-II3, 1S,3S-ACPD), 2E4X (mGluR-II3, 1S,3R-ACPD), 2E4Y (mGluR-II3,2R,4R-APDC), and 2E4Z (mGluR-III7)].

*Present address: Fuji Gotemba Research Laboratories, Chugai Pharmaceutical Company,1-135 Komakado, Gotemba, Shizuoka 412-8513, Japan.

†Present address: Institute for Advanced Biosciences, Keio University, 403-1 Nipponkoku,Tsuruoka, Yamagata 997-0017, Japan.

‡To whom correspondence may be sent at the present address: Institute for ProteinResearch, Osaka University, Open Laboratories of Advanced Bioscience and Biotechnol-ogy, 6-2-3 Furuedai, Suita, Osaka 565-0874, Japan. E-mail: morikako@protein.osaka-u.ac.jp.

§To whom correspondence may be sent at the present address: Office of Graduate Coursesfor Integrated Research Training, Kyoto University Faculty of Medicine, Yoshida, Sakyo-Ku, Kyoto 606-8501, Japan. E-mail: jingami@mfour.med.kyoto-u.ac.jp.

This article contains supporting information online at www.pnas.org/cgi/content/full/0611577104/DC1.

© 2007 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0611577104 PNAS � March 6, 2007 � vol. 104 � no. 10 � 3759–3764

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Janu

ary

21, 2

021

1A). The structure of the CR domain (Fig. 2A) contains three�-sheets, each composed of two short, antiparallel �-strands.This domain possesses nine cysteine residues, which are allstrictly conserved among the family C GPCRs (Fig. 2D), exceptfor the GABAB receptor. All of these residues are fully oxidized(Fig. 2 A). The four intradomain disulfide bridges contribute tostabilizing the internal architecture of the domain. In contrast,the remaining cysteine residue (C527) is bound to C240 in theLB2 domain (17). Notably, the short connection between the Cterminus of the structure (E567) and the N terminus of thepredicted TM (A577) contains nine residues, which should alsorestrict the relative position between the two regions. Thus, it ismost likely that the CR domain plays a major role in transmittingthe conformational change in the ligand-binding domain to theTM region. Furthermore, this domain prevents a direct inter-action between the ligand-binding and TM regions, because thedistance between the LB2 domain and the C terminus of the CRdomain is too far (Fig. 1B).

The domain architecture could be divided into three modules(Fig. 2D), as reported for the TNF receptor (18). The N-terminalmodule in the CR domain (C509–I532) significantly deviates fromany known structures. Although the structural feature is reminis-

cent of the B module (18), the position of the disulfide bridgebetween C509 and C528 is different from the conserved position inthe authentic B2 module. Therefore, we tentatively named thismodule B2�. The other conserved residues are located near thecharacteristic disulfide bond (Fig. 2A). The hydroxyl group of S510forms a hydrogen bond to the main-chain nitrogen atom of W529,and the large aromatic ring of this residue apparently maintains thelocal structure by van der Waals (VDW) interactions. The con-served architecture is close to both the interdomain boundarybetween Q508 and C509 and the interdomain disulfide bondbetween C240 and C527, implying that the B2� module plays animportant role in fixing the relative orientation between the CR andLB2 domains. Consequently, the CR domain could amplify theconformational change in the ligand-binding domains and therebymore effectively transmit the signal to the TM region. This roleprobably accounts for the uniqueness of the B2� module in thefamily C GPCRs.

In contrast, the structures of the middle (P533–M547) andC-terminal (D548–E567) modules are categorized into the A1module (18). The crystal packing implies their potential function.In the crystal, the A1 module at the C terminus forms intermo-lecular contacts at a noncrystallographic two-fold axis (Fig. 2B).This contact site is composed of T556, D558, G561, and Y563 (Fig.2C). These amino acids, except Y563, exhibit weak sequencesimilarity among mGluRs (Fig. 2D). This interaction tempted us toenvisage a physiological situation in which mGluR may be capableof clustering in a one-dimensional array that cooperatively stabilizesthe closed–closed/R structure. The tandem interactions are indeedpossible on the cell surface, only when the dimer adopts the closed–closed/R structure [supporting information (SI) Fig. 6]. The buriedsurface area within the two CR domains (517 Å2) is smaller thanthe standard value for biological interaction (19). Although the twoCR domains may be unable to dimerize in solution, the A1-module-mediated interaction could play a role in regulating the dynamicequilibrium of the EC region (4, 5) on the cell membrane, wheremolecular diffusion is highly reduced (20). Alternatively, the A1module may be responsible for binding to other proteins, asobserved for TNF receptor 1 (21).

Agonist Binding by the Group II Receptor. We have determined thecrystal structures of mGluR-II3 complexed with five differentagonists: Glu, (2S,2�R,3�R)-2-(2�,3�-dicarboxycyclopropyl)glycine(DCG-IV), (1S,3S)-1-aminocyclopentane-1,3-dicarboxylic acid(1S,3S-ACPD), (1S,3R)-1-aminocyclopentane-1,3-dicarboxylicacid (1S,3R-ACPD), and (2R,4R)-4-aminopyrrolidine-2,4-dicarboxylate (2R,4R-APDC). All of these structures were crys-tallized with essentially identical packing. As a result, their overallstructures are basically the same as that shown in Fig. 1B, except forthe CR domains in the 2R,4R-APDC-bound form, in which onestructure was disordered in the dimer. These agonists were boundin the same interdomain cleft (SI Fig. 7). The relative positions ofthe LB1 and LB2 domains are less variable in the agonist-boundforms (data not shown). These fixed open angles strongly suggestthat the primary effect of the agonist binding is domain closing, asmentioned previously (4, 5).

Fig. 3 illustrates schematic diagrams representing the agonistrecognition by mGluR-II3. These agonists share a common struc-ture of an �-amino acid with one carboxyl group at the �-position.The ligand-binding pocket recognizes these chemical groups in anessentially identical manner. The residues involved in ligand bindingagree with the previous mutational studies (7, 8). Surprisingly,mGluR-II3 copes with these different agonists by rearrangingsolvent molecules, instead of altering the protein conformation.The positions of the W2 and W3 water molecules in the Glu- and1S,3S-ACPD-bound structures are occupied by the additionalcarboxyl group of DCG-IV (Fig. 3 A–C). Consequently, the hy-drogen-bonding network around the ligand is maintained amongthe three structures. Furthermore, DCG-IV binding would cost less

Fig. 1. Architecture of mGluRs. (A) Domain architecture of mGluRs. The redlines in mGluR-II3 and mGluR-III7 indicate the region used for the presentcrystallographic analyses. Similarly, the blue line in mGluR-I1 indicates theregion analyzed in our previous studies (4, 5). (B) Two orthogonal views of themGluR-II3 EC region complexed with Glu. The noncrystallographic two-foldaxis runs on the paper plane in both of the drawings. Each domain is coloredas in A. Cyan and blue stick models represent disulfide linkages and sugarmolecules attached at N209, respectively. Bound glutamate molecules areillustrated by red space-filling models. The disordered segments are repre-sented by broken lines. The two protomers are distinguished by dark/faintcolors to depict the dimeric architecture. (C) Overall architecture of themGluR-III7 ligand-binding region, as illustrated in B. Yellow space-filling mod-els represent the bound 2-(N-morpholino)ethanesulfonic acid molecules.

3760 � www.pnas.org�cgi�doi�10.1073�pnas.0611577104 Muto et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

21, 2

021

in terms of solvent entropy than the binding of glutamate or1S,3S-ACPD to form similar hydrogen bonds. In addition, theVDW contact between Y150 and DCG-IV (Fig. 3B) may alsocontribute to stabilizing the ligand binding. These interactionsagree with the fact that DCG-IV displays higher affinity to mGluR-II3 than does Glu (22, 23). Other VDW contacts with Y222 andG302 are observed in the 1S,3S-ACPD-bound form (Fig. 3C).Considering the lower affinity of 1S,3S-ACPD than that of Glu (22,23), the conformation of the bound agonist might be energeticallyunfavorable. In contrast to the above three agonists, water mole-cules are invisible in the ligand-binding pockets complexed with1S,3R-ACPD and 2R,4R-APDC (Fig. 3 D and E), due to thelimited resolution of the x-ray analyses. Although the �-carboxylgroups of these two compounds have different stereo configura-tions from those of the other three, the hydrogen bonds with R68and K389 are formed but skewed. These configurations may partlyexplain their lower affinity than that of Glu (22, 23). In contrast, theinvisible water molecules partly contribute to the binding of 2R,4R-APDC, because the agonist could bind to the receptors morestrongly than 1S,3R-ACPD (22, 23). The nitrogen atom in thefive-membered ring of 2R,4R-APDC (Fig. 3E) may be hydrogen-bonded with solvent molecules.

The mGluR-III7 Ligand-Binding Region. Fig. 1C illustrates the overallarchitecture of the ligand-binding region of the dimeric mGluR-III7, which lacks the CR domain. The asymmetric unit contains oneprotomer, which forms the dimer related by crystallographic two-

fold symmetry. As in the cases of mGluR-I1 and mGluR-II3, theligand-binding region is divided into two domains (Fig. 1 A and C).The structure of each domain is essentially the same as those of thecorresponding regions in mGluR-I1 (4) and mGluR-II3 (Fig. 1B). Inthe crystal, the protomer adopts the open conformation, probablybecause of the binding of a buffer molecule, 2-(N-morpholino)eth-anesulfonic acid (Fig. 1C; see also SI Fig. 8), in the pocket. The openangle between the LB1 and LB2 domains is comparable with thoseobserved in the mGluR-I1 crystals with an antagonist (5). Theinterprotomer interaction primarily occurs only between the twoLB1 domains, which assume the R-conformation (4). Thus, theoverall architecture is designated as open–open/R.

A structural comparison among mGluR-I1, mGluR-II3, andmGluR-III7 suggests that they recognize Glu in a similar manner.To elucidate the recognition mechanism, the closed protomer ofmGluR-III7 was modeled with reference to the closed protomer ofmGluR-II3 by least-square fitting. Fig. 4A represents the ligand-binding pocket with the conserved residues among the threereceptors. These residues belong to not only the LB1 domain(R78/68/78, S165/151/159, T188/174/182, and K409/389/407; eachresidue number represents that for mGluR-I1/mGluR-II3/mGluR-III7) but also the LB2 domain (Y236/222/230 and D318/301/314).These conserved residues form electrostatic interactions with the�-amino acid and �-carboxyl groups of Glu in both mGluR-I1 (4)and mGluR-II3 (Fig. 3A). The highly conserved structure stronglysuggests that these interactions should also occur in mGluR-III7.The mutational studies of the group III receptor confirmed thatthese residues affect agonist binding (9, 10).

Fig. 2. CR domain. (A) Stereo diagram showing the CR domain, colored as in Fig. 1B. Each of the three modular structures is bracketed. (B) Interaction betweenthe two neighboring CR domains along with the crystallographic c axis. The left molecule is drawn in faint colors to visualize the intermolecular interface. Thered arrow indicates the noncrystallographic two-fold axis. (C) Close-up view of the interface. Stick models represent residues on the interface in addition tocysteine residues forming disulfide linkages. (D) Multiple sequence alignment of the CR domain in the family C GPCRs, including human calcium-sensing receptor(hCaSR), murine pheromone receptors EC1-V2R (mPhR1), and EC2-V2R (mPhR2); human mGluR subtypes 1–8 (hmGluR1–8); and rat taste receptors type 1members 2 (rTR2) and 3 (rTR3). The primary and secondary structures of rat mGluR-II3 (rmGluR3) are shown on the top. The bottom line indicates the conservedresidues among these receptors. Capital letters represent the specific amino acid type. Otherwise, a, b, h, and l represent aromatic residue, basic residue,hydrophobic residue, and residue carrying a large hydrophobic moiety, respectively.

Muto et al. PNAS � March 6, 2007 � vol. 104 � no. 10 � 3761

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Janu

ary

21, 2

021

DiscussionGroup Discrimination by mGluR Agonists. Fig. 4A shows the biaseddistribution of the conserved residues that participate in the agonistbinding. The nonconserved residues are located near the entranceof the ligands in the open conformation. Despite harboring thesame ligand (Glu), the ligand-binding region of mGluR-II3 is moreclosed than that of mGluR-I1, because W110 and E292 mutuallyblock further closing (Fig. 4B). As a result, the ligand-bindingpocket is kept open to the bulk solvent through the two channels inthe group I receptors (Fig. 4C). On the other hand, the ligand-binding pocket of the closed mGluR-II3 is inaccessible from theoutside, because of the barrier formed by Y150 and R277 (Fig. 4D),specific for the group II receptors. These residues are responsiblefor creating the distinct shape of the ligand-binding pocket. Al-though a previous homology modeling study proposed that R277might participate in direct hydrogen-bonding with DCG-IV (8),such an interaction is not observed in the crystal structure (Fig. 3B).Instead, the residue most likely plays a role in supporting the rigidligand-binding pocket. A similar effect may be produced by S154and R189 (7). In contrast to the group I/II receptors, the LB1–LB2interface of mGluR-III7 possesses residues with small side chains(Fig. 4E). When the closed protomer is modeled with reference tothe mGluR-II3 structure, the ligand-binding pocket is wider andmore easily accessible to the bulk solvent than those of the othertwo. It is apparently difficult for mGluR-III7 to close the ligand-binding region further, because of the steric hindrance among theresidues surrounding the ligand-binding site. Therefore, the widelyopened ligand-binding pocket would be unique to the group IIIreceptors. This environment partly accounts for the low affinity ofGlu to mGluR-III7 (24).

DCG-IV displays high selectivity and potency for the group IIreceptors (25, 26). The structural differences between mGluR-I1and mGluR-II3 suggest that the side chain of conserved W110 inmGluR-I1 blocks the binding of DCG-IV by steric hindrance at thecontact position of the agonist with Y150 (Fig. 4F). Furthermore,the widely opened binding pocket for the group III receptors (Fig.4E) is unfavorable for binding DCG-IV, because the VDW inter-action with Y150 in mGluR-II3 (Fig. 3B) is absent in the group IIIreceptors.

Implications for Receptor Activation. The mGluR-II3 EC complexedwith various agonists adopts the R-conformation, which appearedin the ligand-free state (4) and the antagonist-bound state (5) formGluR-I1. A comparison of the R-conformations among mGluR-I1, mGluR-II3, and mGluR-III7 indicates that the relative positionbetween the LB1 domains is basically identical (SI Fig. 9). Thehydrophobic residues in the LB1 interface are conserved among thethree groups (4), in addition to the invariant local structures (datanot shown). Thus, all of the mGluRs potentially perform the A-to-Rconversion. Because the clear shift of the C helix in the LB1interface (5) was not observed in the closed protomer of mGluR-II3, the previous hypothesis, which claimed the absence of theopen–open/A state in the dynamic equilibrium (5), appears to belegitimate. The observation of multiple structures not only in theligand-free state (closed–open/A and open–open/R) but also in theagonist-bound state (closed–open/A and closed–closed/R) impliesthat the receptor activation involves a complicated mechanismbased on dynamic equilibrium, although the crystal packing mayhave caused biased trapping of certain conformations.

The structure of the CR domain and its spatial relationships with

Table 1. Distance between the two CR domains in the mGluR-II3dimer

Quaternary structure Distance, Å

Open–open/R 104Closed–open/R 109Closed–closed/R 126Open–open/A 90Closed–open/A 57Closed–closed/A 46

Each structural model was generated from the atomic coordinates formGluR-II3 complexed with Glu (closed–closed/R). The open protomers wereproduced by superimposition onto the mGluR-I1 structure complexed with(S)-(�)-methyl-4-carboxyphenylglycine (open–open/R) (5). To generate theA-conformations, the closed–open/A structure of mGluR-I1 complexed withGlu (4) was considered. The values shown are for distances between the twoC� atoms of the E567 residues, which are the visible C-terminal ends of thecrystal structure.

Fig. 3. Agonist recognition by mGluR-II3. Schematic drawings for the binding of Glu (A), DCG-IV (B), 1S,3S-ACPD (C), 1S,3R-ACPD (D), and 2R,4R-APDC (E).Hydrogen atoms attached at the C� atom of the ligands are modeled with the corresponding ideal geometries. Only the residues/water molecules that directlyinteract with one of the agonists are drawn. Red and blue lines indicate hydrogen-bonding and VDW contact, respectively. Either of the two carboxyl oxygenatoms connected by the green line in B is likely to be protonated, as suggested from the short distance between them (2.4 Å).

3762 � www.pnas.org�cgi�doi�10.1073�pnas.0611577104 Muto et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

21, 2

021

the ligand-binding region allow us to discuss the connection be-tween the EC and TM region less ambiguously than before (4). Weconstructed the six structural models of the EC region (closed–closed/A, closed–open/A, open–open/A, closed–closed/R, closed–open/R, and open–open/R) and evaluated the distances betweenthe two CR domains in the dimer (Table 1). Given these distancesand the size of the TM region (�35 Å), estimated from the structureof rhodopsin (27), these structural models could be classified intotwo groups. The two TM regions in the dimeric receptor couldassociate with each other only in the closed–open/A and closed–closed/A conformations. On the other hand, the spacing in theR-conformations is too large for them to associate directly witheach other. The two TM regions within an mGluR dimer may notalways contact each other, whereas the dimer itself can be retainedthrough the interprotomer disulfide bridge in the EC region (SI Fig.10) (28).

Although we speculated that the A- and R-conformations mightrepresent the ‘‘active’’ and ‘‘resting’’ states, respectively (4), theiractual involvement in receptor activation is still controversial (29).As a simple interpretation, the closing of the EC region may besufficient to induce a conformational change in the adjacent TMregion, which could be activated in a separated state. However, thisidea neglects the A-conformations, despite the functionality of theGd3�-binding site (5, 12). To allocate the roles for both the A- andR-conformations, the original hypothesis proposed in our previousreport (4) should be reevaluated (Fig. 5). The former induces thedirect association of the two TM regions, which could activate Gprotein, whereas they hardly associate in the latter. The structure ofthe dimeric receptor is likely to be in a dynamic equilibriumcomposed of the five components (5). The A-to-R transition could

Fig. 4. The ligand-binding pocket. (A) Conserved amino acid residues in-volved in ligand binding. Red, green, and blue stick models represent thestructures for mGluR-I1, mGluR-II3, and mGluR-III7, respectively. The threeclosed protomers were superimposed by least-square fitting. The model of theclosed mGluR-III7 was constructed as described in the text. (B) Difference in theopen angle of the closed protomer between mGluR-I1 (purple) and mGluR-II3(green). The yellow stick model represents the bound Glu. The black arrowindicates the view direction in A. (C–E) The ligand-binding pockets of mGluR-I1(C), mGluR-II3 (D), and mGluR-III7 (E) as viewed in A. Each molecular model iscolored as in Fig. 1B. Red lines indicate the conserved surface shown in A. (F)Diagram representing the DCG-IV binding of mGluR-II3, viewed from direc-tions similar to that in A. The structure of the closed protomers of mGluR-I1(green) (4) is superimposed onto the mGluR-II3 structures (yellow).

Fig. 5. Hypothetical model for mGluR activation. Three structural modelsrepresent the active (bottom, left), resting (bottom, right), and insulated (top)states, respectively. Each model, viewed in parallel to the membrane, wasconstructed by using the atomic coordinates of mGluR-II3 EC (present work)and rhodopsin (26). For clarity, the closed–open/A and open–open/R confor-mations are highlighted in the active and resting states, respectively, withdark colors, whereas the other possible conformations are drawn with faintcolors. In all of these models, the C termini of the EC region (black arrowheads)are anchored to the same position at the corresponding TM. The position isseparated by �12 Å from the N terminus of the first TM helix (white arrow-heads). The distance is long enough for the nine residues to connect the tworegions in mGluR (black broken lines). Yellow and green spheres indicate theC� atoms of the residues representing loops I (K67; connecting the first andsecond TM helices) and II (R147; connecting the third and fourth TM helices),respectively. The interatomic distances for I–I (yellow lines) (66 Å in A; 54 Å inR), II–II (green lines) (42 Å in A; 55 Å in R), and I–II (purple lines) (53 Å in A; 53Å in R) are consistent with the experimental data (13). The A-to-R transition inthe dynamic equilibrium could induce lateral translation of the blue TM, asindicated by the gray arrows.

Muto et al. PNAS � March 6, 2007 � vol. 104 � no. 10 � 3763

BIO

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Janu

ary

21, 2

021

induce the rearrangement of the TM regions, as suggested byTateyama et al. (13) (Fig. 5). It was also reported that the dimericreceptor could be activated by a single protomer closing (14, 15).These data are consistent with our mechanism, because closing ofone protomer is sufficient to induce the A-conformation (4).Furthermore, the intermolecular interactions within the crystallattice (Fig. 2B) suggest that the closed–closed/R structure repre-sents another structural state that insulates the receptors fromequilibrium. Both the A- and R-conformations coexist in equilib-rium, even if two protomers in the dimer are fully closed. Excessamounts of the closed–closed/R molecules on the cell surfaceshould cause aggregation of the receptors, which could stabilize theR-conformation. This structural state might be related to thedesensitized one, where the receptor responsiveness to the agonistsis attenuated. The intermolecular interactions by the CR domainssuggest another possibility that a cluster of proteins may coopera-tively function for the receptor activation. Interactions with otherproteins, including G protein and scaffold proteins, might induceunexpected conformational changes in mGluR. To test thesepossibilities, further studies of mGluRs are required.

Materials and MethodsThe recombinant proteins for the crystallographic analyses wereproduced with the baculovirus expression system and were purifiedto homogeneity. The EC region of mGluR-II3, whose two glyco-sylation sites (N414 and N439) were mutated with glutamine, wascrystallized in a buffer containing 0.2 M NH4H2PO4 (pH 4.9–5.1)and 2.5–5.0% (wt/vol) PEG 6000. Crystals of the ligand-bindingdomain of mGluR-III7 were obtained in a mixture of the B2 and C3buffers in Grid Screen ammonium sulfate (Hampton Research,Aliso Viejo, CA) in a 1:4 ratio. Both of the crystal structures weresolved by the molecular replacement method using the atomiccoordinates of mGluR-I1 (Protein Data Bank ID code 1EWK) andwere refined to provide the final structures with reasonable crys-tallographic statistics. Detailed experimental procedures and sta-tistics are described in SI Table 2 and SI Materials and Methods.

We thank Drs. T. Oyama and N. Shimizu for help with x-ray data collection;Ms. R. Sata, Ms. H. Kishi, and Mr. T. Tomura for technical assistance; andDr. S. Nakanishi for valuable comments. This work was partly supported bythe New Energy and Industrial Technology Development Organization.

1. Pin JP, Galvez T, Prezeau L (2003) Pharmacol Ther 98:325–354.2. Nakanishi S, Masu M (1994) Annu Rev Biophys Biomol Struct 23:319–348.3. Hollmann M, Heinemann S (1994) Annu Rev Neurosci 17:31–108.4. Kunishima N, Shimada Y, Tsuji Y, Sato T, Yamamoto M, Kumasaka T,

Nakanishi S, Jingami H, Morikawa K (2000) Nature 407:971–977.5. Tsuchiya D, Kunishima N, Kamiya N, Jingami H, Morikawa K (2002) Proc Natl

Acad Sci USA 99:2660–2665.6. Sato T, Shimada Y, Nagasawa N, Nakanishi S, Jingami H (2003) J Biol Chem

278:4314–4321.7. Malherbe P, Knoflach F, Broger C, Ohresser S, Kratzeisen C, Adam G, Stadler

H, Kemp JA, Mutel V (2001) Mol Pharmacol 60:944–954.8. Yao Y, Pattabiraman N, Michne WF, Huang XP, Hampson DR (2003)

J Neurochem 86:947–957.9. Rosemond E, Peltekova V, Naples M, Thogersen H, Hampson DR (2002) J Biol

Chem 277:7333–7340.10. Rosemond E, Wang M, Yao Y, Storjohann L, Stormann T, Johnson EC,

Hampson DR (2004) Mol Pharmacol 66:834–842.11. Suzuki Y, Moriyoshi E, Tsuchiya D, Jingami H (2004) J Biol Chem 279:35526–

35534.12. Abe H, Tateyama M, Kubo Y (2003) FEBS Lett 545:233–238.13. Tateyama M, Abe H, Nakata H, Saito O, Kubo Y (2004) Nat Struct Mol Biol

11:637–642.14. Kniazeff J, Bessis AS, Maurel D, Ansanay H, Prezeau L, Pin JP (2004) Nat

Struct Mol Biol 11:706–713.15. Hlavackova V, Goudet C, Kniazeff J, Zikova A, Maurel D, Vol C, Trojanova

J, Prezeau L, Pin JP, Blahos J (2005) EMBO J 24:499–509.

16. Liu X, He Q, Studholme DJ, Wu Q, Liang S, Yu L (2004) Trends Biochem Sci29:458–461.

17. Rondard P, Liu J, Huang S, Malhaire F, Vol C, Pinault A, Labesse G, Pin JP(2006) J Biol Chem 281:24653–24661.

18. Naismith JH, Sprang SR (1998) Trends Biochem Sci 23:74–79.19. Zhu H, Domingues FS, Sommer I, Lengauer T (2006) BMC Bioinformatics 7:27.20. Kusumi A, Nakada C, Ritchie K, Murase K, Suzuki K, Murakoshi H, Kasai RS,

Kondo J, Fujiwara T (2005) Annu Rev Biophys Biomol Struct 34:351–378.21. Banner DW, D’Arcy A, Janes W, Gentz R, Schoenfeld HJ, Broger C, Loetscher

H, Lesslauer W (1993) Cell 73:431–445.22. Wright RA, Arnold MB, Wheeler WJ, Ornstein PL, Schoepp DD (2001)

J Pharmacol Exp Ther 298:453–460.23. Pin JP, Acher F (2002) Curr Drug Targets CNS Neurol Disord 1:297–317.24. Schoepp DD, Jane DE, Monn JA (1999) Neuropharmacology 38:1431–1476.25. Hayashi Y, Momiyama A, Takahashi T, Ohishi H, Ogawa-Meguro R, Shige-

moto R, Mizuno N, Nakanishi S (1993) Nature 366:687–690.26. Cartmell J, Adam G, Chaboz S, Henningsen R, Kemp JA, Klingelschmidt A,

Metzler V, Monsma F, Schaffhauser H, Wichmann J, Mutel V (1998) Br JPharmacol 123:497–504.

27. Palczewski K, Kumasaka T, Hori T, Behnke CA, Motoshima H, Fox BA, LeTrong I, Teller DC, Okada T, Stenkamp RE, et al. (2000) Science 289:739–745.

28. Tsuji Y, Shimada Y, Takeshita T, Kajimura N, Nomura S, Sekiyama N,Otomo J, Usukura J, Nakanishi S, Jingami H (2000) J Biol Chem 275:28144–28151.

29. Parnot C, Kobilka B (2004) Nat Struct Mol Biol 11:691–692.

3764 � www.pnas.org�cgi�doi�10.1073�pnas.0611577104 Muto et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

21, 2

021