Full text of DOI:10.1110/ps.9.11.2200

Protein Science ~2000!, 9:2200–2209. Cambridge University Press. Printed in the USA.
Copyright © 2000 The Protein Society
Three-dimensional model of the extracellular domain
of the type 4a metabotropic glutamate receptor:
New insights into the activation process
ANNE-SOPHIE BESSIS,1 HUGUES-OLIVIER BERTRAND,2 THIERRY GALVEZ,3
CYRIL DE COLLE,3 JEAN-PHILIPPE PIN,3 and FRANCINE ACHER1
1Laboratoire de Chimie et Biochimie Pharmacologiques et Toxicologiques, UMR8601-CNRS,
Université René Descartes-Paris V, 75270 Paris Cedex 06, France
2Molecular Simulations Inc., Parc Club Orsay Université, 91893 Orsay Cedex, France
3Centre INSERM-CNRS de Pharmacologie-Endocrinologie, UPR 9023-CNRS, 34094 Montpellier Cedex 5, France
~Received June 13, 2000; Final Revision September 8, 2000; Accepted September 18, 2000!
Abstract
Metabotropic glutamate receptors ~mGluRs! belong to the family 3 of G-protein-coupled receptors. On these proteins,
agonist binding on the extracellular domain leads to conformational changes in the 7-transmembrane domains required
for G-protein activation. To elucidate the structural features that might be responsible for such an activation mechanism,
we have generated models of the amino terminal domain ~ATD! of type 4 mGluR ~mGlu4R!. The fold recognition search
allowed the identification of three hits with a low sequence identity, but with high secondary structure conservation:
leucine isoleucine valine-binding protein ~LIVBP! and leucine-binding protein ~LBP! as already known, and acetamide-
binding protein ~AmiC!. These proteins are characterized by a bilobate structure in an open state for LIVBP0LBP and
a closed state for AmiC, with ligand binding in the cleft. Models for both open and closed forms of mGlu4R ATD have
been generated. ACPT-I ~1-aminocyclopentane 1,3,4-tricarboxylic acid!, a selective agonist, has been docked in the two
models. In the open form, ACPT-I is only bound to lobe I through interactions with Lys74, Arg78, Ser159, and Thr182.
In the closed form, ACPT-I is trapped between both lobes with additional binding to Tyr230, Asp312, Ser313, and
Lys317 from lobe II. These results support the hypothesis that mGluR agonists bind a closed form of the ATDs,
suggesting that such a conformation of the binding domain corresponds to the active conformation.
Keywords: ACPT; docking; homology modeling; metabotropic glutamate receptors
Elucidating the activation mechanism of G-protein coupled recep-
tors ~GPCR! is a major issue. In the case of most rhodopsin-like
~family 1, see Bockaert & Pin, 1999! GPCRs, the agonists interact
in a cavity within the 7-transmembrane domain ~TM!, and stabi-
lize the active conformation ~Bockaert & Pin, 1999!. In the case of
family 3 GPCRs, the agonist binding site is located within their
large extracellular domain ~Conn & Pin, 1997; Pin et al., 1999!,
whereas the G-protein activation is still mediated by the intracel-
lular loops of their 7TM domain ~Pin et al., 1994; Gomeza et al.,
1996!. It is actually not known how the agonist binding on the
extracellular domain of these family 3 GPCRs leads to the con-
formational changes required for G-protein activation ~Pin et al.,
1999!.
receptor, several putative pheromones receptors and a taste recep-
tor ~Bockaert & Pin, 1999!. Eight subtypes of metabotropic glu-
tamate receptors have been cloned so far. They are classified in
three groups according to their sequence similarities, transduction
mechanisms, and pharmacological profiles ~Pin et al., 1999; Schoepp
et al., 1999!. Group I is composed of mGlu1R and mGlu5R that
both stimulate PLC hydrolysis. Group II includes mGlu2R and
mGlu3R, which inhibit adenylyl cyclase, as do mGlu4R, mGlu6R,
mGlu7R, and mGlu8R from Group III.
As mentioned above, agonists bind within the large amino ter-
minal extracellular domain ~ATD! of these receptors ~O’Hara et al.,
1993; Takahashi et al., 1993; Okamoto et al., 1998; Hampson et al.,
1999; Han & Hampson, 1999!. Identifying the structure of this
binding domain and the conformational changes resulting from
agonist binding would shed some light on the activation mecha-
nism of these receptors. In 1993, O’Hara et al. proposed that
mGluR ATDs share structural similarity with bacterial periplasmic
amino acid-binding proteins ~PBP!, such as the leucine-binding
protein ~LBP! and the leucine0isoleucine0valine-binding protein
~LIVBP!, the structures of which have been solved by X-ray crys-
This family 3 of GPCRs encompasses the metabotropic gluta-
mate receptors ~mGluR!, the GABAB receptor, a calcium-sensing
Reprint requests to: Francine Acher, Laboratoire de Chimie et Biochimie
Pharmacologiques et Toxicologiques, UMR8601-CNRS, Université René
Descartes-Paris V, 45 rue des Saints-Pères, 75270 Paris Cedex 06, France;
e-mail: acher@biomedicale.univ-paris5.fr.
2200

Model of mGlu4R extracellular domain
2201
tallography ~O’Hara et al., 1993!. This proposal was based on a
low sequence identity ~17% between LIVBP and mGlu1R ATD!
and the mutagenesis of critical binding residues. Since then, the
structures of LIVBP and LBP have been used for the generation of
tridimensional models of mGluR ~Costantino & Pelliciari, 1996;
Costantino et al., 1999; Hampson et al., 1999!, GABAB ~Galvez
et al., 1999!, and calcium-sensing ~Ray et al., 1999! receptor ATDs,
and some of these models have been partially validated by muta-
genesis experiments ~O’Hara et al., 1993; Bräuner-Osborne et al.,
1999; Galvez et al., 1999; Hampson et al., 1999!. Structural sim-
ilarity was also detected between ionotropic glutamate receptor
~iGluR! ligand-binding region and the lysine0arginine0ornithine-
binding protein ~LAOBP!, the histidine-binding protein ~HBP!,
and the glutamine-binding protein ~QBP!, which all belong to a
PBP structural cluster different from the one of LIVBP0LBP ~Tam
& Saier, 1993; Stern-Bach et al., 1994; Lampiden et al., 1998;
Felder et al., 1999!. This observation was later confirmed by the
crystal structure of the ligand-binding core of a GluR subunit
~GluR2! in complex with an agonist ~Armstrong et al., 1998!.
However, similarities between iGluRs and mGluRs at both sec-
ondary structure and binding site levels are too low to allow the
use of this crystal structure as a template.
Periplasmic binding proteins serve as high-affinity active trans-
port of various nutrients such as sugars, inorganic anions, and
amino acids. Many of these proteins have been crystallized with
and without their ligands and their structures solved by X-ray
~Quiocho & Ledvina, 1996!. They are all constituted of two glob-
ular domains linked by a hinge region and can be found either in
an open unliganded, an open liganded, or a closed liganded form
~Quiocho & Ledvina, 1996!. Accordingly, it has been proposed
that the substrate binds to a first lobe and then stabilizes the closed
form of the protein.
were not considered since they only match on a short part of the
query sequence ~42 amino acids overlap or less! and as a result, do
not allow a global alignment with matching secondary structure
elements ~see below!. Interestingly, a Blosum 50 retrieved only
LIVBP and LBP, whereas a Blosum 75 retrieved 2chs, 1com but
not LIVBP and LBP. Yet, sequence identity with mGlu4R ATD was
low ~19%!, as previously observed ~O’Hara et al., 1993!. Such a
low sequence identity between the query sequence and the remote
homologues falls in the so-called twilight zone, thus we carried out
Psi-Blast ~Altschul et al., 1997! and Seqfold ~Fisher & Eisenberg,
1996! searches to increase the sensitivity and to support the sig-
nificance of the initial results. These methods, which take into
account not only the sequence but respectively sequence profiles
and0or secondary structure predictions, retrieved again LIVBP and
LBP with the lowest E-values. AmiC ~acetamide-binding protein,
PDB code: 1pea! was next retrieved with a lower but acceptable
significance. Other hits were not considered since they displayed a
much lower significance. Interestingly, AmiC belongs to the same
structural cluster as LIVBP and LBP. Indeed, while AmiC shares
only 12% sequence identity with LIVBP and LBP, these three
proteins belong to the same family of the periplasmic binding
protein-like according to the SCOP classification and to the same
cluster within this structural family ~Tam & Saier, 1993; Cham-
berlain et al., 1997!. LIVBP, LBP, and AmiC are bilobate proteins
likely adopting open and closed structures along their activation
mechanism. Although LIVBP and LBP have been shown to un-
dergo closure upon ligand binding ~Olah et al., 1993!, only the
structures of their open forms have been solved ~Sack et al., 1989a,
1989b!. However, a model for a closed state has been proposed,
but is not publicly available ~Olah et al., 1993!. Conversely AmiC
has only been crystallized in a closed state ~Pearl et al., 1994!.
Accordingly, the available 3D structures of LIVBP and LBP can be
used to construct a 3D model of mGlu4R ATD in a putative open
form, whereas the structure of AmiC may constitute a valuable
template for a possible closed form.
To support a similar mechanism for family 3 GPCR ATDs, we
have investigated the possible three-dimensional ~3D!-structures
of mGlu4R ATD taken as a representative member of this receptor
family and using computational methods. Our aim was to generate
a homology-derived model for the open and closed forms of the
ATD, and then to analyze the models in terms of their agreement
with site-directed mutagenesis, pharmacological profiles, and phar-
macophore models ~Bessis et al., 1999; Jullian et al., 1999!. Our
results highlight some new putative residues for each lobe that
could interact with the agonist. Altogether, our data support the
hypothesis that the agonist is bound to a closed form of the ATD,
suggesting this state corresponds to the active state of the receptor.
Alignment
The LIVBP, LBP, AmiC, mGlu1R, mGlu2R, and mGlu4R sequence
alignment, generated as described in Materials and methods, is
displayed in Figure 1. Other mGluR sequences are not shown since
the mGluR family alignment is similar to literature data ~Duvoisin
et al., 1995!. The alignment was further confirmed by the good
correspondence between X-ray and predicted secondary structure
elements of LIVBP and mGluRs ~Fig. 1!. Inspection of the sec-
ondary structure of mGluRs reveals that some strands ~bB, bD, bE,
ba, and bM identified in LIVBP and LBP! were not predicted.
LIVBP aa and ab homologous helices were also not predicted for
mGluRs. However, they were not predicted when LIVBP sequence
was submitted to PHD ~data not shown!. They were also not de-
picted in the initial X-ray structure ~Sack et al., 1989a!. Gaps into
LIVBP or LBP secondary structure elements were avoided, except
for aIX helix to maintain the alignment of conserved residues on
both sides of the gap. This insertion reflects a distortion into the
mGluRs’ aIX helices allowing a proper orientation of hydrophobic
residues, thus improving the structural quality of the model. Yet,
the secondary structure homology between LIVBP and mGlu4R
ATD reaches 58% compared to 19% sequence identity according
to the alignment in Figure 1. This percentage results from the
comparison of the number of residues involved in secondary struc-
ture elements of LIVBP ~b-strands and a-helices, Fig. 1! to the
Results and discussion
Template searching
In 1993, O’Hara et al. proposed that mGluR ATDs share structural
similarity with PBP such as LBP and LIVBP. To identify new
possible structural templates for mGlu4R ATD, we performed new
searches against an updated version of the protein data banks ~PDB!.
We first carried out a classical sequence similarity search using
Fasta3 ~Pearson & Lipman, 1988!. The following results were
obtained when using a Blosum62 matrix. LIVBP ~PDB code: 2liv!
and LBP ~PDB code: 2lbp! were recognized as the best templates
since matching residues were found all along the query sequence
and thus give rise to a subsequent global alignment. Other pro-
posed templates ~isomerase 2chs, chorismate mutase 1com, . . .!

number of residues involved in analogous predicted secondary
structure elements of mGlu4R ATD ~Fig. 1!. This result supports
the hypothesis of a structural homology between the two classes of
proteins, although a very low sequence identity was displayed. The
final alignment is close to the one previously published by O’Hara
et al. ~1993!, but different from an alternative one proposed by
Costantino et al. ~1999!. It shows four major insertions for mGlu4R
~I1 amino acids 59–67, I2 126–148, I3 353–401, I4 428–439, Fig. 1!
compared to LIVBP and LBP sequences. To avoid the large inser-
tions, Costantino et al. ~1999! suggested a different alignment but
it does not optimally align secondary structure elements and the
resulting model does not agree with recent mutagenesis results ~see
below!. Conversely, the present alignment displays conserved bind-
ing residues Arg, Ser, and Thr at positions 78, 159, and 182.
Concerning AmiC, a pairwise sequence alignment with LIVBP and
LBP was impossible due to the low sequence identity. As a result,
this alignment was structurally deduced as described in Materials
and methods. Interestingly, the conserved Ser and Thr residues
mentioned above are found aligned.
Open form model
On the basis of the sequence alignment with LBP and LIVBP, we
constructed a 3D model for the open form of the mGlu4R ATD.
The X-ray structure of the LIVBP template is shown in Figure 2A.
It reveals a binding site located within a cleft defined by two lobes
~Fig. 2A!. The mGlu4R ATD open model based on LIVBP and
LBP coordinates is displayed in Figure 2C. Using an iterative
approach ~see Materials and methods!, we obtained a model with
a satisfactory Profiles_3D score ~Fig. 3C! that compares favorably
to the LIVBP one ~Fig. 3A!. Only one misfold ~defined as a region
with a negative Profiles_3D score! with no effect on the putative
binding site was observed ~Fig. 2C!. According to MODELER’s
probability density functions ~PDF!, no violations were observed
and only few residues ~Ͻ2%! were located in a disallowed region
of the Ramachandran map ~data not shown!. Insertions I1 to I4 are
colored in blue ~Fig. 1!. Two of the four insertions found in mGlu4R
ATD with respect to LIVBP and LBP, namely I2 ~23 amino acids!
and I3 ~49 amino acids!, had to be respectively truncated from

Model of mGlu4R extracellular domain
2203
Fig. 2. 3D structures of templates from the PDB: ~A! LIVBP and ~B! AmiC. Lobe I, lobe II, and linkers are displayed as a ribbon with
similar colors as in Figure 1: cyan, magenta and yellow, respectively. Models of the mGlu4R ATD: ~C! open form and ~D! closed form.
Insertions I1 to I4 are colored in blue ~trI indicates truncated insertion!, the region ~112–114! of negative Profiles_3D score in orange,
a-helices in red, and b-strands in green as in Figure 1.
amino acid 132 to 144 and 356 to 397. In fact, all models generated
with the large insertions displayed knots between I2 and I3 and in
their structural core. Accordingly, these insertions were reduced to
7 and 10 amino acids as indicated above, to avoid knots while
allowing enough flexibility. Amino acid 124 to 128, at the begin-
ning of I2, were constrained in the conformation of an a-helix as
predicted by PHD ~Fig. 1!. Since all insertions appear out of the
cleft ~Fig. 2C!, the lack of information on these regions should not
affect the topology of the binding site model. Conversely, Arg78,
Ser159, and Thr182, which are conserved in all mGluRs and have
been shown to be critical for agonist binding ~Hampson et al.,
1999!, are found at the surface of lobe I facing the cleft. On the
other hand, Arg106 of mGlu1R, which would be located at the
binding site according to Costantino et al.’s ~1999! alignment, is
not conserved within the mGluRs and aligns with Leu106 of
mGlu4R. In the present model it lies outside the cleft on the surface
of lobe I.
Agonist pharmacophore models have been established for mGlu1,
mGlu2, and mGlu4 receptors, which belong to group I, group II,
and group III mGlu receptors, respectively ~Bessis et al., 1999;
Jullian et al., 1999!. They show that in all three cases glutamate
would be recognized in an extended conformation. Consequently,
the activation by selective Glu-like agonists would result from
specific interactions with the different ligand-binding environ-
ments and not from the selection of a specific glutamate bioactive
conformation. Indeed, each pharmacophore model is characterized
by selective sites such as S4 for mGlu4R ~Fig. 4A!. A few potent
selective agonists are known; they all possess a glutamate embed-
ded in their structure and bind to the protein through the three
glutamate common sites S1 ~amino function!, S2 and S3 ~proximal
and distal acidic functions! ~Fig. 4A! ~Bessis et al., 1999; Jullian
et al., 1999!. The binding of leucine in LIVBP ~Sack et al., 1989a!
shows that the a-amino and a-acidic groups are bound to Thr102
and Ser79, respectively, and that the hydrophobic side chain inter-
acts with Tyr18. By analogy, Thr182 and Ser159 ~mGlu4R!, which
align with Thr102 and Ser79 ~LIVBP!, would bind to S1 and S2,
respectively. The distal site S3 would interact with Lys74 aligning
with Tyr18 of LIVBP ~Fig. 1!. This ligand binding mode is illus-
trated with the selective and highly functionalized agonist 1-amino-
cyclopentane-1,3,4-tricarboxylic acid ~ACPT-I! ~Acher et al., 1997!,
which was manually positioned in the open form of mGlu4R ATD
~Figs. 4A, left and 4B!. Interestingly, Arg78 was found there to
interact with the distal function S3 in agreement with recent mu-
tagenesis experiments ~Hampson et al., 1999!.
We have previously noted that when agonists are superposed in
their bioactive conformations ~S1-S3 superimposed!, all chemical
groups generating selectivity were localized on the same face of
the superposition ~Pin et al., 1999!. With S1-S3 interaction sites
fitted into the open form model, we now observe that most of these
selective chemical features ~including S4! are facing lobe II, while

2204
A.-S. Bessis et al.
Fig. 3. Profiles_3D plots: ~A! LIVBP, ~B! AmiC, ~C! mGlu4R ATD open form, and ~D! closed form.
residues interacting with S1-S3 sites ~Thr182, Ser159, Lys74, Arg78!
are situated on lobe I, as shown for ACPT-I ~Fig. 4B!. The gluta-
mate entity of ACPT-I, which is mimicked by carbons 1 to 3, lines
lobe I while carbon 4 ~bearing the selective S4 function! and car-
bon 5 are facing lobe II. These observations suggest that lobe II
plays an important role in the selectivity of agonist binding in these
receptors. In agreement with this proposal, Takahashi et al. ~1993!
reported that residues 225 to 355 of mGlu2R were critical in de-
fining the characteristic pharmacological properties of this recep-
tor compared to those of mGlu1R. According to our alignment and
3D models for the ATD of mGluR, these residues constitute the
main part of lobe II ~Fig. 1!. Yet, in our open model, the side
chains of the residues of lobe II facing the putative glutamate
binding site in lobe I are too distant from the ligand to contact it
~Fig. 4B!. Therefore, they cannot in the present form be involved
in ligand binding selectivity. These observations lead us to gener-
ate a model of the closed form of the mGlu4R ATD to get an insight
into additional residues that might bind glutamate and other selec-
tive agonists.
displayed in Figure 2D, was examined using MODELER’s PDF,
Profiles_3D score, and Ramachandran maps. Few residues ~Ͻ2%!
were found in the disallowed regions of the Ramachandran map
~data not shown!, and the Profiles_3D plot ~Fig. 3D! is comparable
to the LIVBP ~Fig. 3A! and AmiC ~Fig. 3B! ones. A similar mGlu4R
ATD model was obtained by Hampson et al. ~1999!. However,
these authors did not point out any critical residues belonging to
lobe II and involved in agonist binding. To detect such residues,
ACPT-I was docked in our closed model and used as a molecular
probe, taking advantage of its four functional groups.
Docking of ACPT-I in the closed form model
ACPT-I includes a glutamate structure; therefore, it is likely to
bind to the receptor through the three glutamate binding sites S1-S3
~see above!. It also holds a fourth specific hydrophilic function
~S4, Fig. 4A! ~Bessis et al., 1999!. ACPT-I was chosen in its
putative bioactive conformation as determined in the pharmaco-
phore model. It was manually docked into the binding site of the
mGlu4R ATD closed form model similarly to leucine into LIVBP
~Sack et al., 1989a! and L-serin-O-phosphate ~L-SOP! into the
mGlu4R ATD model ~Hampson et al., 1999!: 1-amino group ~S1!
in interaction with Thr182, 1-carboxylate ~S2! with Ser159, and
3-carboxylate ~S3! with Lys74 and Arg78. This initial complex was
submitted to molecular mechanics and dynamics as described in
Materials and methods.
The final model ~Figs. 4A, right and 4C! shows that ACPT-I was
kept in the same conformation as in the pharmacophore model
~Bessis et al., 1999! since the 5-membered ring remained in the
same pucker all along the molecular dynamics. Therefore, this
result provides a validation of the previous pharmacophore. The
present model suggests an important network of interactions in-
volving amino acids of both lobes and anchoring the ligand by all
its heteroatoms ~Figs. 4C, 5!. This model allows us to propose how
each of the S1 to S4 sites are tied to the protein. The three protons
of the amino group ~S1! are respectively bound to the oxygen atom
Closed form model
Atomic coordinates of LIVBP and LBP closed form are not yet
publicly available ~Olah et al., 1993; Trakhanov & Quiocho, 1995!.
However, the closing of the cleft between the two lobes is a general
movement for the whole family of PBP ~Quiocho & Ledvina,
1996; O’Hara et al., 1999!. A hinge-bending motion occurs while
the structure of each lobe remains similar. Consequently, only the
linker 3D structure of LIVBP or LBP closed form is required to
construct the whole mGlu4R ATD model using structures of the
open form lobes. Coordinates of this hinge region were obtained
using the structure of the three interdomain segments of AmiC
~Fig. 2B! after performing a structure-based alignment with LIVBP.
This last alignment was deduced from the superposition of lobe I
of LIVBP and AmiC, followed by superposition of lobe II as
reported ~Pearl et al., 1994!. The resulting closed model, which is

Model of mGlu4R extracellular domain
2205
Fig. 4. A: Docking of ACPT-I ~carbon atoms in green, nitrogens in blue, oxygens in red! in the mGlu4R ATD models ~Ca trace shown,
lobes and hinge colored as in Fig. 2A!: open form on the left ~manual docking!, closed form on the right ~computational docking!.
Molecular structure of ACPT-I is shown with S1 to S4 binding sites as defined in the pharmacophore model ~Bessis et al., 1999! and
cyclic carbon atoms numbering. Expanded view of the binding of ACPT-I to the open form model ~B! and the closed form model ~C!.
Side chains of binding residues ~experimentally defined ~Hampson et al., 1999! or proposed in this study! are displayed in cyan ~lobe
I! or magenta ~lobe II!.

2206
A.-S. Bessis et al.
1998!. The two oxygen atoms of the a-acidic group ~S2! are bound
to the hydroxyl side chain of Ser159, one of them to the backbone
NH of Ser159, the other one to the backbone NH of Thr182. One
of the oxygen atoms of the second acidic group ~S3! is bound to the
three ammonium functions of Arg78 and Lys317, the other one to
Lys74 ammonium function and to the proton of the hydroxyl group
of Ser157. The flexible side chain of Arg78 appears to be well
orientated in the binding pocket through cation-P interactions with
three aromatic residues Phe55, Tyr179, and Phe408. The two ox-
ygen atoms of the third acidic group ~S4! are bound to the ammo-
nium function of Lys74; furthermore, one of these atoms is bound
to the ammonium function of Lys317 and the other one to the
proton of Ser313 hydroxyl ~Fig. 5!. A stabilizing electrostatic in-
teraction occurs between S3 and Arg78, Lys317, Lys74 side-chain
functions as well as between S4 and Lys74, Lys317. Carbon 2 of
ACPT-I and Ala180 methyl group are in lipophilic interaction, as
well as carbon 5 and Tyr230 phenol ring.
While binding to Arg78, Ser159, and Thr182 was expected, the
model suggests new bindings to Tyr230 and Asp312, which are
conserved in all mGluRs, and to Lys74, Ser157, Ser313, and Lys317,
which are subtype or group specific ~Table 1!. The model also
predicts a cluster of aromatic residues ~Phe55, Tyr179, Phe408!
located around Arg78, which are also conserved in all mGluRs
~Table 1!. Asp312 and Tyr230 probably play a critical role in
anchoring the amino functional group of all mGluRs agonist. Since
they are situated on lobe II, they do not contribute to the binding
of agonists in the open form, but might play an important role in
the activation process. Interestingly, when the glutamyl residue of
parathyroid Ca2ϩ-sensing receptor, which aligns with Asp312 of
mGlu4R, is mutated, altered calcium homeostasis has been de-
scribed ~Brown & Hebert, 1997!. This observation also supports
the putative critical role of this residue.
Fig. 5. Synthetic scheme of all interactions observed between ACPT-I and
residues from the two lobes ~Table 1! of the closed form model of mGlu4R
ATD. Hydrogen bonds are displayed with dashed lines. When interaction
occurs between charged functional groups, electrostatic interaction may
also take place. A P symbol indicates the cation P interaction between the
amino group of ACPT-I and Tyr230. The cation P interactions between
Arg78 and the aromatic cluster formed by Phe55, Tyr179, and Phe408 are
not materialized for clarity. Residues that are conserved in all mGluRs are
in red, those that are either group specific or one receptor type specific
are in blue.
of Thr182 side chain, to the oxygen atom of Ala180 backbone, and
to the two oxygen atoms of Asp312. Simultaneously, an electro-
static interaction takes place between the positive charge of the
ACPT-I amino group and the negative charge of Asp312. Tyr230
contributes to the positioning of the agonist ammonium moiety by
means of a cation-P interaction ~Dougherty, 1996; Pullman et al.,
Lys74, Arg78, Lys317, Ser157, and Ser313 form an important
basic and hydrophilic cluster that can strongly bind the two distal
Table 1. Residuesa involved in the binding of ACPT-I to mGlu4R ATD and homologous residuesb from other
mGluRs according to the alignment of Figure 1 and known mGluR alignment (Duvoisin et al., 1995)
Group III
mGlu6
Group II
mGlu3
Group I
Lobe
I
mGlu4
mGlu7
mGlu8
mGlu2
DmGluA
mGlu1
mGlu5
Ⅲ
Phe55
—
Gln
—
—
—
—
—
—
—
—
Asn
—
—
—
—
—
—
—
—
—
—
Ala
—
—
—
—
—
—
Arg
—
—
—
—
—
—
—
—
Arg
—
—
—
—
—
—
—
—
Arg
—
—
—
Pro
—
—
—
Tyr
—
Gly
—
—
Ser
—
—
Tyr
—
Gly
—
—
Ser
—
Lys74
Arg78
Ser157
Ser159
Tyr179
Ala180
Thr182
Ⅲ
Ⅲ
Phe408
—
—
—
II
Tyr230
Asp312
Ser313▫
Lys317▫
—
—
—
—
—
—
—
—
—
—
—
—
—
—
Gly
Leu
—
—
Gly
Gln
—
—
Gly
Gln
—
—
Gly
Arg
—
—
Gly
Arg
Ⅲ
aIn bold are residues identified as directly contacting the carboxylic or amino group of glutamate, with ~ ! are residues identified
as likely stabilizing the side chain of Arg78 and those labeled with a ~▫! are residues contacting ACPT-I and which are group III
selective.
bResidues identical to those of mGlu4R are indicated with —.

Model of mGlu4R extracellular domain
2207
acidic functions of ACPT-I ~S3 and S4 sites! and may also explain
the specific binding of phosphonic ligands as L-AP4 ~2-amino-4-
phosphono butyric acid! on group-III mGluRs. Among these res-
idues, which are shared between the two lobes ~Table 1; Fig. 5!,
only Arg78 is common to all mGluRs. Consequently, only Lys74,
Ser157, Ser313, and Lys317 could eventually explain ACPT-I se-
lectivity. It can be noted that Ser313 and Lys317, which belong to
lobe II, are conserved in group III, as it is not the case for Lys74
and Ser157 from lobe I ~Table 1!. No clear data on the involvement
of Lys74 in the activation process are yet available, but two types
of results indicate that Ser157 does not play an essential role in
agonist binding nor in selectivity. While an alanine residue is found
in mGlu8R in place of Ser157 in mGlu4R, ACPT-I affords similar
potency on both receptors ~De Colle et al., 2000!. Furthermore,
when this residue ~Ser157! was mutated to alanine, hardly any
effect was noted on agonist binding ~Hampson et al., 1999!. We
suggest that the role of Ser157 in agonist binding is reduced be-
cause of the strong ionic interactions taking place between the
distal acidic function S3 and the three basic residues Lys74, Arg78,
and Lys317. According to Hampson’s model, the Ser157 side-
chain hydroxyl would bind to Arg78, providing a stabilization of
this important residue. However, Arg78 would be already posi-
tioned by an aromatic cluster ~see above! and binding to Ser157
would not be essential. Future studies using new probes or mutants
will clarify the role of the binding site residues.
sion number P31423! was used as a query sequence to search
structural data banks ~e.g., Brookaven Protein Data Bank, SCOP
Library!. This was done according to various methods: Fasta3,
Psi-Blast, Seqfold, based on different techniques: sequence, se-
quence profile similarities, secondary structure prediction. The PDB
was first searched using the Fasta3 method ~Pearson & Lipman,
1988! available at the EBI server ~http:00www.ebi.ac.uk.0fasta3!
using the default parameters and different blosum scoring matri-
ces. We then performed a Psi-Blast run ~Altschul et al., 1997!
using a standalone version. The sequence profile was built by
searching the SwissProt Data Bank with a Blosum62 matrix. The
converged profile was subsequently utilized to search the PDB. We
performed as well a Seqfold search ~Fisher & Eisenberg, 1996!
~InsightII version 2000, Molecular Simulations Inc., San Diego,
California!. Seqfold gives the ability to search against a fold li-
brary ~e.g., SCOP, http:00www.scop.mrc-lmb.cam.ac.uk0scop! using
either a DSC secondary structure prediction ~King & Sternberg,
1996! alone or in combination with the Psi-Blast profile.
Sequence alignment
To get an optimal sequence alignment between mGlu4R ATD and
LIVBP and LBP, we proceeded in three steps. A ClustalW ~version
1.6! ~Thompson et al., 1994! multiple sequence alignment was first
performed using the sequences of metabotropic glutamate receptor
Altogether, ACPT-I, which is trapped in the domain interface,
binds to amino acids coming from both lobes and, as such, may
contribute to the stabilization of a closed form of mGlu4R ATD, as
described for acetamide bound to AmiC ~Pearl et al., 1994!.
Assuming a common activation mechanism for mGluRs, LIVBP,
and AmiC, the selected templates determined the magnitude of the
mGlu4R ATD hinge motion, which is ;358 ~Chamberlain et al.,
1997!. However, the amplitude of the rotation of the two domains
appears to be quite variable among the family members. While
PBPs exhibit a large domain movement, with the amide sensor
protein AmiC a small relative hinge motion was detected ~Cham-
berlain et al., 1997; O’Hara et al., 1999!. Thus, while the inter-
subdomain hinge angle seems to be optimal for ACPT-I bound to
the mGlu4R ATD in a closed form model, we have no experimental
data to attest to the accuracy of the angle of the open liganded or
unliganded form. In fact, this conformation may not be stable since
we suppose that an equilibrium occurs between the open and closed
conformations with and without a bound ligand, and that activation
would be induced by the closed-liganded form ~Costantino et al.,
1999; Galvez et al., 1999!, in agreement with recent data obtained
on the GABA-B1 receptor ~Galvez et al., 2000!. To obtain the best
model of this latter form, we had to start with the generation of a
bound open form, because of the choice of the LIVBP template.
Obviously this first model could not account for the specific in-
teractions of several agonists such as ACPT-I, while it was the case
with the closed form allowing interaction of the agonist functional
groups with both lobes.
ATD ~rat mGlu1-8R plus the Drosophila receptor DmGluAR!, Ca2ϩ
-
sensing receptor ATD, LIVBP, and LBP. A Blosum30 was used as
a scoring matrix; gap penalties and extension gap penalties were
set to 30 and 0.05, respectively. The alignment was then manually
modified to avoid gaps into secondary structure elements. For
LIVBP and LBP, secondary structures were considered according
to various methods: crystallographic data ~PDB Sum, http:00www.
biochem.ucl.ac.uk0bsm0pdbsum!, Kabsch and Sander algorithm
~InsightII version 980, MSI!, and secondary structure predictions
~Rost & Sander, 1993! using PHD ~http:00www.maple.bioc.
columbia.edu0predictprotein0!. For mGlu4R ATD, only the PHD
prediction could be used. The resulting alignment was then used to
generate 3D models of the mGlu4R ATD open form that were
evaluated as described below. The alignment was further itera-
tively modified to increase the structural quality of our model. The
alignment of AmiC with LIVBP and LBP was deduced from the
structural superposition of each lobe of the closed form of AmiC
on the corresponding lobe of the open form of LIVBP ~Pearl et al.,
1994!.
Open and closed models of mGlu4R ATD (AA47–492)
Both mGlu4R ATD models ~open and closed forms! were gener-
ated by the automated homology modeling tool MODELER 5.00
~InsightII version 980, MSI! ~Sali & Blundell, 1990!. The 3D open
form model was generated using the coordinates of LIVBP and
LBP open forms from Escherichia coli deposited at the PDB. The
sequence alignment used was obtained as described above. Then
the quality of each model generated by MODELER was evaluated
with different independent criteria. Invalid models with important
structural violations, especially knots, were eliminated thanks to
the PDF of MODELER. The remaining models were then submit-
ted to the Profiles_3D algorithm using a sequence window of 21
amino acids ~Profiles_3D, InsightII version 980, MSI!~Luthy et al.,
1992!. This algorithm calculates the compatibility between a given
sequence of amino acids, and its associated modeled structures and
Ultimately, the results of this study support the initial hypothesis
that mGluRATD would close upon ligand binding similarly to PBP.
Materials and methods
Template searching
To identify remote structural homologues, mGlu4R ATD ~amino
acids 1 to 492 from Rattus norvegicus mGlu4R, SwissProt acces-

2208
A.-S. Bessis et al.
helps to identify misfolded regions. This statistical tool allowed us
to locally modify the sequence alignment and to generate itera-
tively new models until we reached an optimal score equivalent to
96% of theoretical score. The best resulting open model as well as
its Profiles_3D score are displayed in Figures 2C and 3C. The
corresponding final alignment ~Fig. 1! was retained to generate a
3D closed model of mGlu4R ATD with truncated I2 an I3 inser-
tions, as described above. This model was constructed using the
coordinates of the two lobes of LIVBP ~Lobe I: amino acids 1–118,
253–325; Lobe II: amino acids 125–247, 332–344! and LBP ~Lobe
I: amino acids 1–118, 253–327; Lobe II: amino acids 125–247,
334–346! crystalline open forms, the coordinates of the two lobes
of the open form model of mGlu4R ATD ~Lobe I: amino acids
47–199, 341–470; Lobe II: amino acids 206–334, 480–492!, and
the coordinates of the three linkers ~amino acids 121–129, 252–
264, 335–344! of the hinge region of AmiC crystalline closed
form. The best resulting model as well as its Profiles_3D are
displayed in Figures 2D and 3D.
References
Acher F, Tellier F, Brabet I, Fagni L, Azerad R, Pin J-P. 1997. Synthesis and
pharmacological characterization of aminocyclopentane tricarboxylic acids
~ACPT!: New tools to discriminate between metabotropic glutamate recep-
tor subtypes. J Med Chem 40:3119–3129.
Altschul SF, Madden TL, Schaffer AA, Zhang J, Zhang Z, Miller W, Lipman DJ.
1997. Gapped BLAST and PSI-BLAST: A new generation of protein data-
base search programs. Nucleic Acids Res 25:3389–3402.
Armstrong N, Sun Y, Chen G-Q, Gouaux E. 1998. Structure of a glutamate-
receptor ligand-binding core in complex with kainate. Nature 395:913–917.
Bessis A-S, Jullian N, Coudert E, Pin J-P, Acher F. 1999. Extended glutamate
activates metabotropic receptor types 1, 2 and 4: Selective features at mGluR4
binding site. Neuropharmacology 38:1543–1551.
Bockaert J, Pin J-P. 1999. Molecular tinkering of G-protein coupled receptors:
An evolutionary success. EMBO J 18:1723–1729.
Bräuner-Osborne H, Jensen AA, Sheppard PO, O’Hara P, Krogsgaard-Larsen P.
1999. The agonist-binding domain of the calcium-sensing receptor is lo-
cated at the amino-terminal domain. J Biol Chem 274:18382–18386.
Brown EM, Hebert SC. 1997. Calcium-receptor-regulated parathyroid and renal
function. Bone 20:303–309.
Chamberlain D, O’Hara BP, Wilson SA, Pearl LH, Perkins SJ. 1997. Oligomer-
ization of the amide sensor protein AmiC by X-ray and neutron scattering
and molecular modeling. Biochemistry 36:8020–8029.
Conn PJ, Pin J-P. 1997. Pharmacology and functions of metabotropic glutamate
receptors. Ann Rev Pharmacol Toxicol 37:205–237.
Costantino G, Macchiarulo A, Pellicciari R. 1999. Modeling of amino-terminal
domains of group I metabotropic glutamate receptors: Structural motifs
affecting ligand selectivity. J Med Chem 42:5390–5401.
Docking of ACPT-I in the putative binding site
of mGlu4R ATD closed form
Costantino G, Pelliciari R. 1996. Homology modeling of metabotropic gluta-
mate receptors ~mGluRs!. Structural motifs affecting binding modes and
pharmacological profile of mGluR1 agonists and competitive antagonists.
J Med Chem 39:3998–4006.
De Colle C, Bessis A-S, Bockaert J, Acher F, Pin J-P. 2000. Pharmacological
characterization of the rat metabotropic glutamate receptor type 8a revealed
strong similarities and slight differences with the type 4a receptor. Eur J
Pharmacol 394:17–26.
ACPT-I ~Fig. 3A! was chosen with protonated amino group and
deprotonated carboxylic groups, while a neutral pH ionization state
was set for the residue side chains. ACPT-I, in the conformation of
the agonist mGlu4R pharmacophore model, was initially manually
docked in mGlu4R ATD closed form, in agreement with the posi-
tion of l-leucine in the LIVBP active site. The ammonium and the
a-carboxylate functions were facing Thr182 and Ser159, respec-
tively, while one of the two distal acidic functions was pointing
toward Lys74 to let the second one point to the lobe II. The ligand-
protein complex was submitted to energy minimization using Steep-
est Descent ~until derivative less than 2 kcal0mol0Å! and Conjugated
Gradient ~until derivative less than 0.01 kcal0mol0Å! while Ca_trace
was tethered using a quadratic potential. This was performed using
the Discover 3.00 calculation engine with the CFF force field
~InsightII version 980, MSI!. The nonbond cut-off method and the
dielectric constant were set up to cell multipole and distance-
dependent, respectively ~E ϭ r!. Discover 3.00 and CFF force field
were further used to perform a molecular dynamics at 298 K on the
previously minimized system. The tethering force on Ca trace was
successively decreased by reducing the force constant value of the
quadratic potential from 100 to 60, 30, 20, 10, and 0 every 40 ps.
The molecular dynamics was then continued during 240 ps. The
integration time step was set up to 1 fs, and the calculations were
performed at constant volume and temperature. A snapshot of the
system was saved every 400 fs. Once the system was equilibrated,
the coordinates of 20 snapshots were averaged and submitted again
to the previously described minimization protocol with no Ca
restraints. Contacts between the ligand and the protein were sub-
sequently analyzed using the web interface of the WHATIF pro-
gram ~http:00www.swift.embl-heidelberg.de0servers2!.
Dougherty DA. 1996. Cation-P interaction in chemistry and biology: A new
view of benzene, Phe, Tyr, and Trp. Science 271:163–167.
Duvoisin RM, Zhang CX, Ramonell K. 1995. A novel metabotropic glutamate
receptor expressed in the retina and olfactory bulb. J Neurosci 15:3075–3083.
Felder CB, Graul RC, Lee AY, Merkle H-P, Sadée W. 1999. The venus flytrap
of periplasmic binding proteins: An ancient protein module present in mul-
tiple drug receptors. AAPS Pharmsci. www.pharmsci.org
Fisher D, Eisenberg D. 1996. Protein fold recognition using sequence-derived
predictions. Protein Sci 5:947–955.
Galvez T, Parmentier M-L, Joly C, Malitschek B, Kaupmann K, Kuhn R,
Bittiger H, Froestl W, Bettler B, Pin J-P. 1999. Mutagenesis and modeling
of the GABAb receptor extracellular domain support a venus flytrap mech-
anism for ligand binding. J Biol Chem 274:13362–13369.
Galvez T, Prézeau L, Milioti G, Franek M, Joly C, Froestl W, Bettler B, Bertrand
H-O, Blahos J, Pin J-P. 2000. Mapping the agonist binding site of GABAB
type 1 subunit sheds light on the activation process of GABAB receptors.
J Biol Chem. Forthcoming.
Gomeza J, Joly C, Kuhn R, Knöpfel T, Bockaert J, Pin J-P. 1996. The second
intracellular loop of metabotropic glutamate receptor 1 cooperates with the
other intracellular domains to control coupling to G-proteins. J Biol Chem
271:2199–2205.
Hampson DR, Huang X-P, Pekhletski R, Peltekova V, Hornby G, Thomsen C,
Thøgersen H. 1999. Probing the ligand-binding domain of the mGluR4
subtype of metabotropic glutamate receptor. J Biol Chem 274:33488–33495.
Han G, Hampson DR. 1999. Ligand binding to the amino-terminal domain of
the mGluR4 subtype of metabotropic glutamate receptor. J Biol Chem
274:10008–10013.
Jullian N, Brabet I, Pin J-P, Acher F. 1999. Agonist selectivity of mGluR1 and
mGluR2 metabotropic receptors: A different environment but similar rec-
ognition of an extended glutamate conformation. J Med Chem 42:1546–1555.
King RD, Sternberg MJE. 1996. Identification and application of the concepts
important for accurate and reliable protein secondary structure prediction.
Protein Sci 5:2298–2310.
Lampiden M, Pentikäinen O, Johnson MS, Keinänen K. 1998. AMPA receptors
and bacterial periplasmic amino acid-binding proteins share the ionic mech-
anism of ligand recognition. EMBO J 17:4704–4711.
Luthy R, Bowie JU, Eisenberg D. 1992. Assessment of protein models with
three-dimensional profiles. Nature 356:83–85.
O’Hara BP, Norman RA, Wan PT, Roe SM, Barrett TE, Drew RE, Pearl LH.
1999. Crystal structure and induction mechanism of AmiC-AmiR: A ligand-
regulated transcription antitermination complex. EMBO J 18:5175–5186.
Acknowledgments
This work was supported by grants from the CNRS ~PCV97-115! and
RETINA France. The authors wish to thank Parke-Davis Pharmaceutical
Research ~Ann Arbor, Michigan! for allowing a Fulbright scholarship to
A.-S. Bessis, and G. Milioti, N. Jullian, K. Olswezski, B. Hartmann for
suggestions and critical reading of the manuscript.

Model of mGlu4R extracellular domain
2209
O’Hara PJ, Sheppard PO, Thogersen H, Venezia D, Haldeman BA, Mc Grane V,
Houamed KH, Thomsen C, Gilbert TL, Mulvihill ER. 1993. The ligand-
binding domain in metabotropic glutamate receptors is related to bacterial
periplasmic binding proteins. Neuron 11:41–52.
Okamoto T, Sekiyama N, Otsu M, Shimada Y, Sato A, Nakanishi S, Jingami H.
1998. Expression and purification of the extracellular ligand binding region
of metabotropic glutamate receptor subtype 1. J Biol Chem 273:13089–13096.
Olah GA, Trakhanov S, Trewhella J, Quiocho FA. 1993. Leucine0isoleucine0
valine-binding protein contracts upon binding of ligand. J Biol Chem 268:
16241–16247.
Pearl L, O’Hara B, Drew R, Wilson S. 1994. Crystal structure of AmiC: The
controller of transcription antitermination in the amidase operon of Pseudo-
monas aeruginosa. EMBO J 13:5810–5817.
Pearson WR, Lipman DJ. 1988. Improved tools for biological sequence analysis.
Proc Natl Acad Sci USA 85:2444–2448.
Rost B, Sander C. 1993. Prediction of protein secondary structure at better than
70% accuracy. J Mol Biol 232:584–599.
Sack JS, Saper MA, Quiocho FA. 1989a. Periplasmic binding protein structure
and function. Refined X-ray structures of the leucine0isoleucine0valine-
binding protein and its complex with leucine. J Mol Biol 206:171–191.
Sack JS, Trakhanov SD, Tsigannik IH, Quiocho FA. 1989b. Structure of the
l-leucine-binding protein refined at 2.4 Å resolution and comparison with
the Leu0Ile0Val-binding protein structure. J Mol Biol 206:193–207.
Sali A, Blundell TL. 1990. Definition of general topological equivalence in
protein structures: A procedure involving comparison of properties and
relationships through simulated annealing and dynamic programming. J Mol
Biol 212:403–428.
Schoepp DD, Jane DE, Monn JA. 1999. Pharmacological agents acting at sub-
types of metabotropic glutamate receptors. Neuropharmacology 38:1431–
1476.
Pin J-P, De Colle C, Bessis A-S, Acher F. 1999. New perspective in the devel-
opment of selective metabotropic glutamate receptor ligands. Eur J Phar-
macol 375:277–294.
Pin J-P, Joly C, Heinemann SF, Bockaert J. 1994. Domains involved in the
specificity of G protein activation in phospholipase C coupled metabotropic
glutamate receptor. EMBO J 13:342–348.
Pullman A, Berthier G, Savinelli R. 1998. Interaction of the tetrametylammo-
nium ion with the cycles of aromatic amino acids beyond the SCF ab initio
level. J Am Chem Soc 120:8553–8554.
Quiocho FA, Ledvina PS. 1996. Atomic structure and specificity of bacterial
periplasmic receptors for active transport and chemotaxis: Variation of com-
mon themes. Mol Microbiol 20:17–25.
Stern-Bach Y, Bettler B, Hartley M, Sheppard PO, O’Hara P, Heinemann SF.
1994. Agonist selectivity of glutamate receptors is specified by two domains
structurally related to bacterial amino acid-binding proteins. Neuron 13:
1345–1357.
Takahashi K, Tsuchida K, Tanabe Y, Masu M, Nakanishi S. 1993. Role of the
large extracellular domain of metabotropic glutamate receptors in agonist
selectivity determination. J Biol Chem 268:19341–19345.
Tam R, Saier MHJ. 1993. Structural, functional, and evolutionary relationships
among extracellular solute-binding receptors of bacteria. Microbiol Rev
57:320–346.
Thompson JD, Higgins DG, Gibson TJ. 1994. CLUSTAL W: Improving the
sensitivity of progressive multiple sequence alignment through sequence
weighting, positions-specific gap penalties and weight matrix choice. Nu-
cleic Acids Res 22:4673–4680.
Trakhanov S, Quiocho FA. 1995. Influence of divalent cations in protein crys-
tallization. Protein Sci 4:1914–1919.
Ray K, Hauschild BC, Steinbach PJ, Goldsmith PK, Spiegel AM. 1999. Iden-
tification of the cysteine residues in the amino-terminal extracellular domain
of the human Ca2ϩ receptor critical for dimerisation. J Biol Chem 274:
27642–27650.