The glutamate-receptor ion channels (iGluRs), namely AMPA, kainate
and NMDA receptors, are the major mediators of excitatory synaptic
transmission in the central nervous system1–3. Each subtype has a unique
role. NMDA receptors, which have attracted enormous interest, act as
coincidence detectors. These monitor changes in membrane potential
and the presence of glutamate in the synaptic cleft4. AMPA receptors
have key roles in mediating the rapid excitatory synaptic current, the
kinetics of which is tuned over about a fivefold range in different struc-
tures of the brain5. Kainate receptors have a major modulatory role at
both presynaptic and postsynaptic sites, and kainate-receptor agonists
have long been known to be potent convulsants and environmental
toxins in the food supply6,7. These receptors were identified through
traditional pharmacological approaches, and their naming reflects
this: AMPA, kainate and NMDA are synthetic agonists and naturally
occurring toxins not found in the brain8.
AMPA receptors are encoded by four genes (GluR1–4), kainate recep-
tors by three genes for ‘low-affinity’ subunits (GluR5–7) and two for
‘high-affinity’ subunits (KA1 and KA2), and NMDA receptors by seven
genes (NR1, NR2A–D, NR3A and NR3B). A fourth family encoded by
two δ subunits is much less well understood9. The existence of a large
number of genes, coupled with alternative splicing10 and RNA edit-
ing11,12, makes the functional analysis of native glutamate receptors a
Biophysical studies have revealed many intriguing features for these
channels, such as the unique requirement for coactivation of NMDA
receptors by both glycine and glutamate, which bind to NR1 and NR2
subunits, respectively. But the most exciting recent advance is the solu-
tion of high-resolution crystal structures for the ligand-binding cores
for individual iGluR subunits. This has been achieved for representative
members from each gene family, providing for the first time a detailed
understanding of agonist recognition in a diverse family of ligand-gated
ion channels. From these structures, plausible models have been rigor-
ously tested to identify in broad terms the mechanisms underlying acti-
vation and desensitization, and the action of some allosteric modulators
on the ligand-binding domain.
In this review I discuss the molecular organization of iGluRs; I explain
how crystallographic studies have shaped our thinking about how ago-
nists open the channel; and I summarize what we know about the rapid
desensitization and allosteric modulation of AMPA receptors.
Modular organization of iGluR subunits
The iGluRs have a characteristic modular architecture that seems to
have been formed by fusion of three genetically discrete gene segments
Glutamate receptors at atomic resolution
Mark L. Mayer1
At synapses throughout the brain and spinal cord, the amino-acid glutamate is the major excitatory
neurotransmitter. During evolution, a family of glutamate-receptor ion channels seems to have been
assembled from a kit consisting of discrete ligand-binding, ion-channel, modulatory and cytoplasmic
domains. Crystallographic studies that exploit this unique architecture have greatly aided structural analysis
of the ligand-binding core, but the results also pose a formidable challenge, namely that of resolving the
allosteric mechanisms by which individual domains communicate and function in an intact receptor.
1Building 35, Room 3B1002, Porter Neuroscience Research Center, 35 Lincoln Drive, Bethesda, Maryland 20892, USA.
that in bacteria encode individual proteins (Fig. 1). The amino-terminal
domain (ATD) has homology with leucine-isoleucine-valine-binding
protein (LIVBP), a bacterial periplasmic binding protein, as well as with
the structurally related glutamate-binding domain of the G-protein-
coupled metabotropic receptors. In contrast to its role in G-protein-cou-
pled glutamate receptors, in iGluRs the ATD does not bind glutamate
but instead forms dimers that are major assembly control points limiting
oligomerization to members of the same iGluR subfamily13,14. In NMDA
receptors, the ATD of the NR2A and NR2B subunits bind the negative
allosteric modulators Zn2? and the drug ifenprodil, respectively15–17.
The results of site-directed mutagenesis and molecular modelling stud-
ies, which suggest that Zn2? and ifenprodil promote domain closure
and stabilize a closed cleft conformation of the ATD, raise the question
Figure 1 | Modular organization of a glutamate receptor. a, Diagrammatic
representation of the amino-terminal domain (ATD), the two domain
(D1 and D2) ligand-binding core encoded by the S1 (cyan) S2 (orange)
peptide segments, the ion-channel domain with three membrane-spanning
segments and a pore-loop (P), and the intracellular C terminus. b,
Molecular architecture illustrated by the crystal structures of the mGluR1
ligand-binding domain dimer (green), the GluR5 ligand-binding core
dimer (S1, cyan; S2 orange) and two subunits from the KcsA tetramer
(grey). Note that S1 and S2 contribute to both domains of the ligand-
binding core owing to the crossover of S1 into D2 and S2 in D1. An intact
glutamate receptor is believed to assemble as a dimer of dimers.
γ Glu α
NATURE|Vol 440|23 March 2006|doi:10.1038/nature04709
of whether there are as-yet-unidentified endogenous modulators that
bind to other members of the iGluR family18,19. It is also possible that
the ATD of many iGluRs has lost the ability to bind endogenous ligands
and that its major role now is to control assembly.
The agonist-binding domain of iGluRs is encoded by two polypeptide
segments named S1 and S2, which as shown in Fig. 1 are interrupted by
insertion of the ion-channel pore20. This domain also has sequence and
structural homology with bacterial periplasmic proteins but it differs in
structure from LIVBP and the ATD of iGluRs. With genetic approaches
it is possible to excise the agonist-binding domain from the remainder
of an iGluR subunit and express just this as a soluble domain as a soluble
protein suitable for crystallization21,22. Many agonist and partial agonist
structures have been solved, and this enormous body of new informa-
tion is discussed in depth later in this review.
The ion-channel pore of iGluRs has striking sequence homology with
that of bacterial potassium channels, suggesting that iGluRs also con-
tain a pore-loop motif first identified in crystal structures of bacterial
potassium channels23,24 (see also the review in this issue by Ashcroft,
p. 440). This view was reinforced by the discovery of GluR0, a bacte-
rial glutamate-gated potassium channel25. However, in all iGluRs the
channel orientation in the membrane is inverted compared with that
found in classical potassium channels. Another important difference
is that for all potassium channel structures solved so far, the subunits
are arranged with perfect four-fold symmetry. In contrast, the agonist-
binding domains of iGluRs are arranged as dimers of dimers. There is
indirect evidence that for eukaryotic iGluRs, which do not discriminate
between Na? and K?, the extracellular pore entrance itself lacks four-
The greatest diversity among iGluR subtypes is found in their cyto-
plasmic domain, which varies in size from less than 20 to around 500
amino acids. This is a key site of allosteric modulation by calmodulin,
protein kinases and phosphatases27–29. There is a large array of proteins
that coassemble in the postsynaptic density, and many of these interact
with glutamate receptors, either directly or through intermediate part-
ners30. Unfortunately very little is known about the molecular organi-
zation of this region. Algorithms for secondary-structure prediction
indicate only short stretches of ?-helix and ?-sheet, suggesting that
this region acts as more of a handle for cytoskeletal and cytoplasmic
proteins, where the recognition sites are encoded not by classical three-
dimensional structured domains, but rather by relatively short primary
signature sequences. At both the structural and functional level this is
a vast area to explore.
Ligand binding at atomic resolution
The general principle that glutamate, or glycine in the case of the
NMDA receptor NR1 subunit, binds in the cleft of a clam-shell-like
motif formed from the S1–S2 segments of a single iGluR subunit was
anticipated from earlier structural analysis of bacterial periplasmic
ligand-binding proteins. These served as the initial models for the lig-
and-binding domains of iGluRs31,32. However, the chemistry of the lig-
and-binding pocket in periplasmic proteins, and even in the prokaryotic
iGluR homologue GluR0 (ref. 33), differs appreciably from that for ver-
tebrate iGluRs. What has emerged from crystallographic studies of the
ligand-binding cores is atomic-resolution knowledge of the mechanisms
that give each receptor its characteristic ligand-binding properties. This
is an extraordinary advance that required the solution of multiple recep-
tor–ligand complexes, the large majority at resolutions better than 2 Å,
and at present the protein databank contains structures for more than
60 iGluR structures. Although the majority are for agonist and partial
agonist complexes with GluR2, structures have also been solved for the
kainate receptor GluR5 and GluR6 subunits, the NMDA receptor NR1
and NR2A subunits, and for several competitive antagonists.
In all of these structures an arginine side chain on helix D forms the
major binding site for the ligand ?-carboxyl group (Fig. 2). This residue
is essential for the high-affinity binding of glutamate or glycine, and
its mutation, even to lysine, greatly lowers agonist affinity. In all nine
AMPA and kainate receptor genes, the ligand ?-amino group is bound
by a conserved glutamate side chain. In NMDA receptors, this gluta-
mate is replaced by an aspartate, and in the NR2A complex the direct
hydrogen bond to the agonist ?-amino group is replaced by a solvent-
mediated contact. Surprisingly, the ?-carboxyl group is not bound by a
counter charge from a lysine or arginine side chain, but instead makes
hydrogen-bond contacts with the main-chain peptide bond and the
hydroxyl group of a conserved threonine side chain. The conformation
adopted by glutamate is similar in AMPA, kainate and NMDA recep-
tor NR2 subunits. As a result the ?-carboxyl group projects towards
helix F in all iGluR structures. This finding was unexpected because
structure–function studies with conformationally restricted ligands sug-
gested that glutamate would adopt an extended conformation when
bound to NMDA receptors but a folded conformation when bound to
Five key themes emerge from comparison of these structures. First, as
shown in Fig. 3, the ligand-binding core is much larger than is necessary
to accommodate glutamate. The exception is NR1, which binds glycine.
NR1 has a binding pocket just about the right size for this smaller resi-
due, and therefore excludes glutamate by steric effects35. The cavity is
completely closed off from the extracellular solution, except for NR2A,
for which a solvent-filled tunnel connects the glutamate ligand in the
interior of the protein to the extracellular environment. Second, sub-
type selectivity arises from amino-acid differences in domain 2, which
shape the size and chemistry of the ligand-binding pocket; steric effects
have key roles in permitting the binding of subtype-selective agonists
I T L
I T Y
I T A
V I D F S
I D F S
V I D F S
VD F S
G S T K E F F
G S T M T F F
G S S M T F F
G T V N G F
L E S T M N E Y I
E S T S I E Y I T
E S T M N E
Y V T
Figure 2 | Conserved structural elements in the agonist-binding site of
iGluRs. a, The sequence alignment and secondary-structure assignment
for GluR2 (?-helix in pink; ?-sheet in yellow) indicate conserved features in
the binding sites of AMPA, kainate and NMDA receptors. Key side chains
that bind glutamate are highlighted in blue, red or cyan and marked with
an asterisk; black and grey boxes indicate identical and conserved amino
acids. b, The crystal structure for the GluR6 glutamate complex shows,
as ball-and-stick models, the Arg (R492), Glu (E707) and Thr (T659) side
chains that bind glutamate. Dashed lines indicate hydrogen bonds with
main-chain and side-chain atoms. Not shown are four additional solvent-
mediated hydrogen bonds that connect the agonist molecule with the
NATURE|Vol 440|23 March 2006
and excluding other ligands36. In contrast, domain 1 shares a common
shape and chemistry in all iGluR subtypes, consistent with its role in
binding the ?-amino and ?-carboxyl groups, which are common to all
iGluR agonists. Third, the extent of agonist-triggered domain closure
varies between receptor subtypes, and also for individual agonists, par-
tial agonists and antagonists36-39. Fourth, competitive antagonists work
by a foot-in-the-door mechanism, trapping the ligand-binding domain
in a wide open conformation35,37,40. Finally, the ligand-binding cavity
contains trapped water molecules that have key roles in the agonist-
Partial agonists are ligands that when applied at a saturating concen-
tration to a receptor produce less than the maximal response. Two
complementary techniques have revealed how these ligands act.
Single-channel recording can measure, on an absolute scale, the fraction
of time that a channel spends in the open state41. In contrast, crystal-
lographic structures reveal a snapshot of proteins caught in one of the
many different conformations they sample during their functional life
cycle. For AMPA receptors, these methods showed that agonist efficacy
is directly coupled to the extent of domain closure38. This conclusion
was based on analysis of a series of 5-position halogen-substituted wil-
lardiines, partial agonists for which the extent of domain closure varied
owing to steric effects. Confirming this picture are additional crystal
structures for individual structurally diverse partial agonists for both
AMPA and kainate receptors including (S)-2-amino-3-(4-bromo-3-
hydroxy-isoxazol-5-yl)propionic acid and kainic acid36,39. The appeal
of this partial agonist mechanism is that it also explains the stepwise
increase in single-channel conductance observed during dissociation
of competitive antagonists from an AMPA receptor in the presence of
a saturating concentration of glutamate42.
Mutation of Leu 650 to Tyr in the AMPA receptor reduced the steric
hindrance that prevented kainate from acting as a full agonist. But
this experiment also revealed a second mechanism closer in spirit to
the widely accepted Monod–Wyman–Changeux (MWC) model for
allosteric gating43. The MWC-like mechanism, in which the affinity
of ligands for a single active conformation determines their efficacy,
was suggested by the observation that the L650T mutation reduced the
efficacy but did not reduce the extent of domain closure for the com-
plex with AMPA. This result hints that that the reduced efficacy occurs
because of a decreased stability of the fully activated AMPA-bound state.
However, for kainate the L650T mutation increased both the efficacy
and the extent of domain closure, indicating that ligands can activate
receptors by more than one mechanism.
Subsequent studies on NMDA receptors have reinforced the view that
different mechanisms can underlie partial agonist activity. The extent of
domain closure for the NR1 subunit partial agonists ACPC and ACBC,
which produce only 80% and 40% of the maximum response, is the
same as that for full agonists44. The high-resolution crystal structures
solved for the NR1 subunit partial agonists revealed local conforma-
tional changes in the ligand-binding site, with the consequence that
despite identical domain closure, the structures of the NR1 agonist and
partial agonist complexes are indeed different. But the difference is much
more subtle than observed for AMPA receptor partial agonists. Single-
channel analysis of the partial agonist action of homoquinolinic acid
on the glutamate-binding site of the NR2A subunit revealed a reduced
rate constant for channel activation, which was linked to different bind-
ing mechanisms for glutamate and homoquinolinic acid, predicted by
molecular dynamics (MD) simulations45. Although the homology model
developed for the NR2A glutamate complex differs in detail from the
subsequently published NR2A crystal structure46,47, and assumes that
the extent of domain closure is identical for homoquinolate and gluta-
mate, the approach of combining high-resolution single-channel record-
ing with MD simulations based on experimentally determined crystal
structures has great potential. MD simulations on the AMPA-receptor
glutamate complex, which measured the stability of hydrogen bonds
and hydrophobic contacts, give a hint of what is possible for studies on
the nanosecond timescale48. More recent work, also on AMPA recep-
tors, revealed for the first time transitions from the open to closed cleft
conformation during 20-ns simulations49. This work also revealed that
as a result of differences in domain closure, water exchange rates within
the ligand-binding pocket were much faster for the partial agonist kain-
ate than for glutamate.
Coupling of ion-channel activity to agonist binding
A dimer is present in the vast majority of iGluR crystal structures,
including the recently solved NMDA receptor NR1/NR2A hetero-
dimer47. As discussed below, such dimers have a key role in coupling
the binding of glutamate to the activation of ion-channel gating. The
dissociation constant, Kd, for dimer formation by the isolated ligand-
binding cores is in the millimolar range for wild-type iGluRs, except
for GluR0, for which the Kd is 0.8 µM. Most probably, dimer formation
occurs readily in vitro owing to the high protein concentrations present
in crystals. This would be mimicked in vivo by the close proximity of
the ligand-binding cores resulting from dimer formation by the ATDs
that coassemble at much lower protein concentrations. In all subtypes
of iGluR, the ligand-binding core dimers are ‘glued’ together by hydro-
phobic contacts made by the surface of helices D and J in domain 1,
together with hydrogen bonds and salt bridges, the details of which
vary according to receptor subtype. Notably, in the dimer assembly of
AMPA and kainate receptors, domain 2 is free to move in the transi-
tion from the apo to agonist-bound conformation. Although the same
is true in NMDA receptors, additional intersubunit contacts between
Figure 3 | Solvent-accessible surfaces of the ligand-binding cavity in iGluR
agonist crystal structures. a, Superposition of the cavity interior in GluR2
(blue), GluR5 (green) and GluR6 (red); the bound ligand is glutamate plus
seven water molecules shown as red spheres in the GluR5 complex.
b, Cavities for NR1 (yellow) and NR2A (blue); the bound ligands are glycine
for NR1 and glutamate plus four water molecules (green) in the NR2A
complex; water molecules in the tunnel are shown as red spheres. Note that
the glutamate ligand is much larger than can be accommodated by the NR1
NATURE|Vol 440|23 March 2006
symmetry53,56. Cyclothiazide slows desensitization by greatly stabilizing
the dimer interface, whereas aniracetam slows deactivation by stabiliz-
ing the ligand-binding domain in the agonist-bound, closed-cleft, active
conformation. The above studies are striking examples of how the com-
bination of crystallographic and functional studies can dissect complex
allosteric mechanisms. Indeed, iGluRs are the only ligand-gated ion
channels for which there is any structural basis for the mechanism of
action of allosteric modulators.
The process of gating and desensitization is of course more complex
than can be explained by the simple two-state models shown in Fig. 4.
Mutations outside the dimer interface, in the linker from S2 to the sec-
ond transmembrane segment, attenuate desensitization in both AMPA
and kainate receptors. But in the absence of structural data, it is not
possible to explain how this occurs57. Mutations in the ligand-binding
pocket can also profoundly affect desensitization: notably, replacement
by alanine of a tyrosine side chain that caps the binding site in AMPA
receptors converts kainate from a very weak partial agonist into a potent
and strongly desensitizing ligand, most probably because of relief of
steric hindrance and increased domain closure58. Here the mechanism
is easier to conceptualize, although the lack of structural data for the
alanine mutant means that a definitive interpretation is not possible. The
same mutation also greatly reduces the potency for glutamate, possibly
as a result of disruption of hydrogen-bond contacts and solvent structure
in the ligand-binding site.
20 Å41 Å
Figure 4 | Activation of iGluR gating, resulting from agonist-induced
expansion of the ligand-binding core dimer. a, Diagrammatic
representations of the resting, activated and desensitized states to illustrate
how domain closure is linked to separation of the ligand-binding core
dimer domain 2 segments. Lower panel, crystal structures of the GluR5
ligand-binding core dimer in the resting (antagonist bound) and active
glutamate-bound conformations40. Currently no structure for the
desensitized state has been reported. b, Diagrammatic representations
indicating a hypothetical model of how binding of an allosteric modulator
to the ATD could pull the agonist-binding core dimer apart and trigger
channel closure. Small red circles represent glutamate.
domains 1 and 2 suggest mechanisms for allosteric coupling between
the glutamate- and glycine-binding sites.
The structure of the dimer assembly in iGluRs immediately suggested
a mechanism for ion-channel gating. In this model, the movement of
domain 2 towards domain 1 during the transition from the open cleft
of the apo state to the closed cleft of the glutamate-bound complex pro-
duces a scissors-like outward motion of the linkers connecting the ion-
channel transmembrane (TM) segments to the ligand-binding core (Fig.
4). Competitive antagonists jam the ligand-binding cores in their resting
conformation by a foot-in-the-door mechanism. Although many details
remain to be resolved, it seems reasonable to assume that gating of the
ion channel in iGluRs results from movement of the TM segments that
are pulled open by contraction of the ligand-binding cores in a dimer
assembly. Because each subunit contains a complete agonist-binding
site, which is coupled to TM segments in the same subunit, it is easy
to imagine how the activation of individual subunits could generate
partial opening of the ion channel. This would lead to the generation
of multiple subconductance states characteristic of AMPA and kain-
ate subtype iGluRs. This mechanism is similar to that subsequently
proposed for the Ca2?-activated K channel MthK33,37,50. Although this
principle is probably the same for NMDA receptors the details are much
less well understood. Binding of either glutamate or glycine alone does
not trigger ion-channel gating, and when the channel does open, it
visits only one subconductance state, suggesting a cooperative gating
scheme involving simultaneous, or tightly coupled, transitions by pairs
of channel gates51,52.
Desensitization and allosteric modulation
The gating model shown in Fig. 4 can explain how the channel opens
because dimer assembly occurs mainly through the domain 1 surface,
leaving domain 2 free to move. This suggests a model in which desensi-
tization occurs when the dimer undergoes a conformational rearrange-
ment that breaks contacts at the domain 1 surface. This uncoupling
allows the pair of subunits in a dimer to rotate around the central axis,
relieving strain on the ion-channel linkers and allowing the ion chan-
nel to close, even though the agonist remains bound with high affinity.
This model was tested by constructing mutants that either increased
or decreased the stability of dimer formation. The effects on dimer
stability measured for isolated S1–S2 constructs correlated with the
extent of desensitization measured in intact receptors bearing the same
mutants53. A subsequent systematic analysis of dimer contacts observed
in the crystal structure was performed by site-directed mutagenesis, and
changes in the kinetics of desensitization were measured54. These results
confirmed that in AMPA receptors the dimer interface must undergo
a rearrangement in order for desensitization to proceed. It also verified
that the stability of the dimer assembly in the resting and active state is
generated by a network of intermolecular contacts that are distributed
over an area of about 900–1400 Å2. A major challenge for the future
will be to identify the conformation of the ligand-binding cores in the
desensitized state of a dimer assembly. An intriguing extension of this
model, which could explain negative allosteric modulation by ligands
such as Zn2? and ifenprodil, is possible if one considers that domain
closure of the ATD will pull apart the protomers in a ligand-binding
core dimer assembly (Fig. 4).
The profound desensitization of AMPA receptors is strongly attenu-
ated by small-molecule, positive allosteric modulators that seem to
act through two somewhat related mechanisms. At the macroscopic
level this can be modelled as resulting from either block of entry into
the desensitized state (as occurs for cyclothiazide) or stabilization of
the open state (as occurs for aniracetam55). Biochemical and crystal-
lographic experiments have established that both classes of compound
bind in a solvent-filled hydrophobic crevice at the base of the ligand-
binding core dimer, but at different sites. In sedimentation equilibrium
experiments using purified GluR2 ligand-binding cores, both cyclo-
thiazide and CX614 greatly stabilized the dimer. Cyclothiazide binds
to a pair of sites at the periphery of the dimer, whereas aniracetam and
CX614 bind to a single site located on the molecular two-fold axis of
NATURE|Vol 440|23 March 2006
NMDA-receptor dimer assemblies
The recent crystallization of a heterodimer of the ligand-binding cores
of the NR1–NR2A complex, and the biochemical demonstration using
disulphide-linked receptors that this assembly is the native form, as
opposed to separate NR1 and NR2 homodimers, is a major achievement
that explains several unique functional properties of NMDA recep-
tors47. Although functional experiments have yet to begin in earnest,
two results are of particular note. First, the slow deactivation of NMDA
receptors, which is key to their biological function at synapses59, occurs
in part as a result of an amino-acid difference in the NR1 subunit in
which a tyrosine residue at the base of the dimer interface occupies the
site to which aniracetam binds in AMPA receptors. The tyrosine thus
acts as an endogenous allosteric modulator that slows deactivation47.
Second, in contrast to the situation for AMPA and kainate receptors, in
which the ligand-binding core dimer interface is generated exclusively
by domain 1 contacts, in NMDA receptors both domain 1 and domain
2 contribute to the dimer interface. This suggests a mechanism for allo-
steric coupling between NR2 and NR1, which underlies the large varia-
tion in glycine affinity observed on coassembly with the NR2A–NR2D
subunits. Experimental tests to define exactly how this occurs have yet
to be reported60,61.
The ion channel
A variety of approaches subsequently revealed the transmembrane
topology of iGluRs. From a structural perspective the most fruitful
have been cysteine-scanning mutagenesis experiments in which the
rate of covalent modification of cysteine side chains was measured62,63.
This information was used to infer the structure of the channel. These
experiments have reached a high degree of sophistication such that the
voltage dependence of modification has been measured to locate the
position of side chains in the membrane electric field64. In conceptually
related experiments, scanning mutagenesis with glutamate, alanine and
tryptophan residues has given clues about the surface topology at the
cytoplasmic entrance to the pore and the location of the binding site
for polyamines, which act as endogenous pore blockers for AMPA and
kainate receptors with a glutamine at the Q/R editing site23.
The picture that has emerged from these studies (Fig. 5) suggests a
three-helix bundle supporting a pore-loop motif that forms the closest
point of contact with ions near the cytoplasmic entrance to the channel65.
The ion-channel pore is lined mainly by residues from M2, the second
transmembrane segment, with Ca2? permeability determined both by
the amino acid at the tip of the pore loop (the Q/R site) and by the
DRPEER sequence in the linker between M2 and S2 in NMDA recep-
tors66. Until recently little attention was given to the M3 segment, which
is present in all eukaryotic iGluRs but not their prokaryotic homologues.
In addition, its location relative to M1 and M2 is uncertain.But M3 is
vital for channel activity in eukaryotes, and its removal, together with
the associated cytoplasmic C terminus, produces inactive receptors.
Suprisingly, coexpression of M3 and truncated receptors restores activ-
ity, showing that there is no need for a covalent link between M3 and
S2 during receptor activation67. Also significant is the observation that
the GYKI family of noncompetitive AMPA receptor antagonists seem
to interact with the S2–M3 linker68.
The gating model for iGluRs, based on movements of the linker
regions in the ligand-binding core crystal dimers, suggests that the top
of the channel, near the extracellular surface, must undergo large confor-
mational movements. Unfortunately the details are fuzzy and even the
location of the barrier to ion permeation in the closed and desensitized
states remains to be unambiguously determined. The dimer of dimers
model implied by the crystal structures of the ligand-binding cores has
not been confirmed at a structural level, and no tetrameric assembly has
yet been visualized at a resolution sufficient to resolve how the individual
subunits are positioned with respect to each other. Despite this, evidence
for a dimer of dimers is now overwhelming. This potentially forces the
issue of a mismatch between the expected four-fold symmetry of the
ion-channel pore, and the two-fold symmetry required by the dimer
of dimers arrangement, but only if the ion-channel pore has four-fold
symmetry. State-dependent crosslinking experiments on AMPA recep-
tors suggest that the top of the helix-bundle crossing implied by a KcsA-
based structural model is better modelled as a dimer of dimers rather
than a tetramer, indicating that if a symmetry mismatch does occurs its
location is deeper in the channel26. However, for NMDA receptors in
which the NR1 and NR2 subunits form obligate heteromeric assemblies,
the symmetry mismatch goes as far as the tip of the pore loop. This is not
surprising given the different amino-acid sequence in the NR1 and NR2
subunits in this region. In retrospect there is no compelling reason to
expect four-fold symmetry within the ion-channel pore for any eukaryo-
tic glutamate receptor. However, for GluR0, which is highly potassium-
selective, the situation is less clear . All of the K-channel structures solved
so far exhibit four-fold symmetry, but coordination of K+ ions does not
require this per se, so perhaps the structure of even the GluR0 pore is
Glutamate receptors and disease
The availability of selective and potent NMDA receptor antagonists,
especially the channel blocker MK801, triggered an enormous research
effort in the 1980s when it was discovered that excessive calcium influx
through NMDA receptors triggered neurodegeneration in animal mod-
els of ischaemia. The initial surge of excitement was rapidly tempered
by studies that showed that NMDA receptor antagonists had severe
behavioural side effects. This, coupled with increasingly sophisti-
cated research on stroke mechanisms that revealed a pathophysiology
much more complicated than could be explained by hyperactivation
of NMDA receptors alone, greatly diminished enthusiasm for further
work with glutamate-receptor antagonists as neuroprotective agents.
However, the issue is not dead, and there is reasonable hope that, by
fine-tuning receptor-subtype selectivity and affinity, clinically useful
drugs will eventually emerge. Impetus for this is a growing body of evi-
dence that glutamate receptors have key roles in a much broader variety
of neurodegenerative and psychiatric diseases than first anticipated and
To ligand-binding domain
Figure 5 | The pore region of the KcsA potassium channel used as a
template to map the results of site-directed mutagenesis on iGluR channel
block by polyamines. At the tip of the pore loop (P), but nowhere else in the
channel, introduction of tryptophan produces a pM affinity-binding site
for polyamines23. At the cytoplasmic entrance to the channel, red spheres
indicate amino-acid ?-carbon atom positions for which introduction of
glutamate side chains supports polyamine block. In the linker from M2 to
the ligand-binding domain, the DRPEER sequence probably contributes to
a Ca2+-binding site near the entrance to the channel66.
NATURE|Vol 440|23 March 2006
the approval of the NMDA-receptor channel blocker Memantine for
treatment of dementia in the United States and the European Union69.
Surprisingly, in humans there are no diseases that have been unam-
biguously linked to mutations in glutamate-receptor genes, although
several animal diseases have been discovered, which have provided us
with useful mouse models. Two of the most interesting are lurcher and
stargazer. Lurcher is produced by a point mutation in the second TM
segment of the δ2 subunit. This mutation causes spontaneous ion-chan-
nel activity, possibly as a result of attenuated desensitization, which leads
to excessive cation flux and neurodegeneration of Purkinje cells, which
express δ2 at high levels70–72. The stargazer phenotype results from a
loss of AMPA-receptor trafficking to synapses on cerebellar granule
cells, which has been traced to a mutation in the γ2 member of a family
of auxiliary membrane proteins, which modulate both AMPA receptor
trafficking and gating kinetics73–75. The finding that AMPA receptors in
vivo coassemble with an auxiliary subunit, and that this modifies their
affinity for agonists and their kinetics of deactivation and desensitiza-
tion, was entirely unexpected.
The extraordinary advances in understanding the ligand-binding mech-
anisms of iGluRs arose largely from crystallographic efforts. Although
many important details will be learned in the immediate future from
crystallization of the ligand-binding cores of additional iGluR subtypes
and complexes, the major challenge is to extend this approach to other
receptor domains, perhaps to an intact receptor, and to figure out how
the domains communicate with each other. As a consequence of our
limited ability to overexpress and purify eukaryotic membrane proteins,
it is far from clear when or if this will be possible. As a result, details of
the ion-permeation process, the binding sites for channel blockers, and
the site of interactions with recently identified accessory membrane
proteins remain to be determined. In this sense, our understanding
of Cys-loop receptors, for which structures are available for the intact
protein, albeit at lower resolution, is more advanced (see the review by
Sine and Engel, p. 448 in this issue). Even with (or without) this infor-
mation, studies using a variety of experimental approaches including
calorimetry, fluorescense resonance energy transfer (FRET), electron
microscopy, NMR, and of course electrophysiological techniques, will
be required to bring life to the static iGluR structures revealed by crys-
tallography. Further investment in computational studies is another
approach with great potential, but major advances will require the
development of techniques to describe protein motion on the micro-
second-to-millisecond time scale. The past five years have been incred-
ibly productive and I hope progress will continue at this rate as more
complex structural problems are tackled.
1. Watkins, J. C. & Evans, R. H. Excitatory amino acid transmitters. Annu. Rev. Pharmacol.
Toxicol. 21, 165–204 (1981).
Mayer, M. L. & Westbrook, G. L. The physiology of excitatory amino acids in the vertebrate
central nervous system. Prog. Neurobiol. 28, 197–276 (1987).
Collingridge, G. L. & Lester, R. A. J. Excitatory amino acid receptors in the vertebrate central
nervous system. Pharmacol. Rev. 41, 143–210 (1989).
Collingridge, G. L. & Bliss, T. V. Memories of NMDA receptors and LTP. Trends Neurosci. 18,
Geiger, J. R. P. et al. Relative abundance of subunit mRNAs determines gating and Ca2+
permeability of AMPA receptors in principle neurons and interneurons in rat CNS. Neuron
15, 193–204 (1995).
Lerma, J. Roles and rules of kainate receptors in synaptic transmission. Nature Rev.
Neurosci. 4, 481–495 (2003).
Perl, T. M. et al. An outbreak of toxic encephalopathy caused by eating mussels
contaminated with domoic acid. N. Engl. J. Med. 322, 1775–1780 (1990).
Watkins, J. C. & Jane, D. E. The glutamate story. Br. J. Pharmacol. 147 Suppl 1, S100–S108
Hollmann, M. in Ionotropic Glutamate Receptors in the CNS (eds Jonas, P. & Monyer, H.)
3–98 (Springer, Berlin, 1999).
10. Sommer, B. et al. Flip and flop: a cell-specific functional switch in glutamate-operated
channels of the CNS. Science 249, 1580–1585 (1990).
11. Sommer, B., Köhler, M., Sprengel, R. & Seeburg, P. H. RNA editing in brain controls a
determinant of ion flow in glutamate-gated channels. Cell 67, 11–19 (1991).
12. Puchalski, R. B. et al. Selective RNA editing and subunit assembly of native glutamate
receptors. Neuron 13, 131–147 (1994).
13. Ayalon, G. & Stern-Bach, Y. Functional assembly of AMPA and kainate receptors is
mediated by several discrete protein-protein interactions. Neuron 31, 103–113 (2001).
14. Ayalon, G., Segev, E., Elgavish, S. & Stern-Bach, Y. Two regions in the N-terminal domain of
ionotropic glutamate receptor 3 form the subunit oligomerization interfaces that control
subtype-specific receptor assembly. J. Biol. Chem. 280, 15053–15060 (2005).
15. Paoletti, P., Ascher, P. & Neyton, J. High-affinity zinc inhibition of NMDA NR1-NR2A
receptors. J. Neurosci. 17, 5711–5725 (1997).
16. Perin-Dureau, F., Rachline, J., Neyton, J. & Paoletti, P. Mapping the binding site of the
neuroprotectant ifenprodil on NMDA receptors. J. Neurosci. 22, 5955–5965 (2002).
17. Masuko, T. et al. A regulatory domain (R1-R2) in the amino terminus of the N-methyl-d-
aspartate receptor: effects of spermine, protons, and ifenprodil, and structural similarity to
bacterial leucine/isoleucine/valine binding protein. Mol. Pharmacol. 55, 957–969 (1999).
18. Paoletti, P. et al. Molecular organization of a zinc binding N-terminal modulatory domain in
a NMDA receptor subunit. Neuron 28, 911–925 (2000).
19. Pasternack, A. et al. Alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)
receptor channels lacking the N-terminal domain. J. Biol. Chem. 277, 49662–49667
20. Stern-Bach, Y. et al. Agonist-selectivity of glutamate receptors is specified by two domains
structurally related to bacterial amino acid binding proteins. Neuron 13, 1345–1357 (1994).
21. Kuusinen, A., Arvola, M. & Keinanen, K. Molecular dissection of the agonist binding site of
an AMPA receptor. EMBO J. 14, 6327–6332 (1995).
22. Chen, G. Q., Sun, Y., Jin, R. & Gouaux, E. Probing the ligand binding domain of the GluR2
receptor by proteolysis and deletion mutagenesis defines domain boundaries and yields a
crystallizable construct. Protein Sci. 7, 2623–2630 (1998).
23. Panchenko, V. A., Glasser, C. R. & Mayer, M. L. Structural similarities between glutamate
receptor channels and K+ channels examined by scanning mutagenesis. J. Gen. Physiol. 117,
24. Kuner, T., Seeburg, P. H. & Guy, H. R. A common architecture for K+ channels and ionotropic
glutamate receptors? Trends Neurosci. 26, 27–32 (2003).
25. Chen, G. Q., Cui, C., Mayer, M. L. & Gouaux, E. Functional characterization of a potassium-
selective prokaryotic glutamate receptor. Nature 402, 817–821 (1999).
26. Sobolevsky, A. I., Yelshansky, M. V. & Wollmuth, L. P. The outer pore of the glutamate
receptor channel has 2-fold rotational symmetry. Neuron 41, 367–378 (2004).
27. Soderling, T. R. & Derkach, V. A. Postsynaptic protein phosphorylation and LTP. Trends
Neurosci. 23, 75–80 (2000).
28. Tavalin, S. J. et al. Regulation of GluR1 by the A-kinase anchoring protein 79 (AKAP79)
signaling complex shares properties with long-term depression. J. Neurosci. 22, 3044–
29. Vissel, B., Krupp, J. J., Heinemann, S. F. & Westbrook, G. L. Intracellular domains of NR2
alter calcium-dependent inactivation of N-methyl-D-aspartate receptors. Mol. Pharmacol.
61, 595–605 (2002).
30. Peng, J. et al. Semiquantitative proteomic analysis of rat forebrain postsynaptic density
fractions by mass spectrometry. J. Biol. Chem. 279, 21003–21011 (2004).
31. Mano, I., Lamed, Y. & Teichberg, V. I. A venus flytrap mechanism for activation and
desensitization of alpha-amino-3-hydroxy-5-methyl-4-isoxazole propionic acid receptors.
J. Biol. Chem. 271, 15299–15302 (1996).
32. Quiocho, F. A. & Ledvina, P. S. Atomic structure and specificity of bacterial periplasmic
receptors for active transport and chemotaxis: variation of common themes. Mol.
Microbiol. 20, 17–25 (1996).
33. Mayer, M. L., Olson, R. & Gouaux, E. Mechanisms for ligand binding to GluR0 ion channels:
crystal structures of the glutamate and serine complexes and a closed apo state. J. Mol.
Biol. 311, 815–836 (2001).
34. Watkins, J. C., Krogsgaard-Larsen, P. & Honoré, T. Structure-activity relationships in the
development of excitatory amino acid receptor agonists and competitive antagonists.
Trends Pharmacol. Sci. 11, 25–33 (1990).
35. Furukawa, H. & Gouaux, E. Mechanisms of activation, inhibition and specificity: crystal
structures of NR1 ligand-binding core. EMBO J. 22, 1–13 (2003).
36. Mayer, M. L. Crystal structures of the GluR5 and GluR6 ligand binding cores: molecular
mechanisms underlying kainate receptor selectivity. Neuron 45, 539–552 (2005).
37. Armstrong, N. & Gouaux, E. Mechanisms for activation and antagonism of an AMPA-
sensitive glutamate receptor: Crystal structures of the GluR2 ligand binding core. Neuron
28, 165–181 (2000).
38. Jin, R., Banke, T. G., Mayer, M. L., Traynelis, S. F. & Gouaux, E. Structural basis for partial
agonist action at ionotropic glutamate receptors. Nature Neurosci. 6, 803–810 (2003).
39. Hogner, A. et al. Structural basis for AMPA receptor activation and ligand selectivity:
Crystal structures of five agonist complexes with the GluR2 ligand-binding core. J. Mol.
Biol. 322, 93–109 (2002).
40. Mayer, M. L., Ghosal, A. K., Dolman, N. P. & Jane, D. E. Crystal structures of the kainate
receptor GluR5 ligand binding core dimer with novel GluR5 selective antagonists.
J. Neurosci. 26, 2850–2859 (2006).
41. Colquhoun, D. & Ogden, D. C. Activation of ion channels in the frog end-plate by high
concentrations of acetylcholine. J. Physiol. (Lond. ) 395, 31–159 (1988).
42. Rosenmund, C., Stern-Bach, Y. & Stevens, C. F. The tetrameric structure of a glutamate
receptor channel. Science 280, 1596–1599 (1998).
43. Armstrong, N., Mayer, M. & Gouaux, E. Tuning activation of the AMPA-sensitive GluR2
ion channel by genetic adjustment of agonist-induced conformational changes. Proc. Natl
Acad. Sci. USA 100, 5736–5741 (2003).
44. Inanobe, A., Furukawa, H. & Gouaux, E. Mechanism of partial agonist action at the NR1
subunit of NMDA receptors. Neuron 47, 71–84 (2005).
45. Erreger, K. et al. Mechanism of partial agonism at NMDA receptors for a conformationally
restricted glutamate analog. J. Neurosci. 25, 7858–7866 (2005).
46. Chen, P. E. et al. Structural features of the glutamate binding site in recombinant NR1/
NR2A N-methyl-d-aspartate receptors determined by site-directed mutagenesis and
molecular modeling. Mol. Pharmacol. 67, 1470–1484 (2005).
47. Furukawa, H., Singh, S. K., Mancusso, R. & Gouaux, E. Subunit arrangement and function in
NMDA receptors. Nature 438, 185–192 (2005).
48. Mamonova, T., Hespenheide, B., Straub, R., Thorpe, M. F. & Kurnikova, M. Protein flexibility
using constraints from molecular dynamics simulations. Phys. Biol. 2, S137–S147 (2005).
49. Arinaminpathy, Y., Sansom, M. S. & Biggin, P. C. Binding site flexibility: Molecular simulation
of partial and full agonists with a glutamate receptor. Mol. Pharmacol. 69, 5–12 (2006).
NATURE|Vol 440|23 March 2006
50. Jiang, Y. et al. The open pore conformation of potassium channels. Nature 417, 523–526
51. Banke, T. G. & Traynelis, S. F. Activation of NR1/NR2B NMDA receptors. Nature Neurosci. 6,
52. Schorge, S., Elenes, S. & Colquhoun, D. Maximum likelihood fitting of single channel
NMDA activity with a mechanism composed of independent dimers of subunits. J. Physiol.
569, 395–418 (2005).
53. Sun, Y. et al. Mechanism of glutamate receptor desensitization. Nature 417, 245–253
54. Horning, M. S. & Mayer, M. L. Regulation of AMPA receptor gating by ligand binding core
dimers. Neuron 41, 379–388 (2004).
55. Partin, K. M., Fleck, M. W. & Mayer, M. L. AMPA receptor flip/flop mutants affecting
deactivation, desensitization, and modulation by cyclothiazide, aniracetam, and
thiocyanate. J. Neurosci. 16, 6634–6647 (1996).
56. Jin, R. et al. Mechanism of positive allosteric modulators acting on AMPA receptors. J.
Neurosci. 25, 9027–9036 (2005).
57. Yelshansky, M. V., Sobolevsky, A. I., Jatzke, C. & Wollmuth, L. P. Block of AMPA receptor
desensitization by a point mutation outside the ligand-binding domain. J. Neurosci. 24,
58. Holm, M. M. et al. A binding site tyrosine shapes desensitization kinetics and agonist
potency at GluR2. A mutagenic, kinetic, and crystallographic study. J. Biol. Chem. 280,
59. Lester, R. A. J., Clements, J. D., Westbrook, G. L. & Jahr, C. E. Channel kinetics determine the
time course of NMDA receptor-mediated synaptic currents. Nature 346, 565–567 (1990).
60. Monyer, H. et al. Heteromeric NMDA receptors: molecular and functional distinction of
subtypes. Science 256, 1217–1221 (1992).
61. Laurie, D. J. & Seeburg, P. H. Ligand affinities at recombinant N-methyl-d-aspartate
receptors depend on subunit composition. Eur. J. Pharmacol. 268, 335–345 (1994).
62. Kuner, T., Wollmuth, L. P., Karlin, A., Seeburg, P. H. & Sakmann, B. Structure of the NMDA
receptor channel M2 segment inferred from the accessibility of substituted cysteines.
Neuron 17, 343–352 (1996).
63. Beck, C., Wollmuth, L. P., Seeburg, P. H., Sakmann, B. & Kuner, T. NMDAR channel
segments forming the extracellular vestibule inferred from the accessibility of substituted
cysteines. Neuron 22, 559–570 (1999).
64. Sobolevsky, A. I., Yelshansky, M. V. & Wollmuth, L. P. State-dependent changes in the
electrostatic potential in the pore of a GluR channel. Biophys. J. 88, 235–242 (2005).
65. Wollmuth, L. P. & Sobolevsky, A. I. Structure and gating of the glutamate receptor ion
channel. Trends Neurosci. 27, 321–328 (2004).
66. Watanabe, J., Beck, C., Kuner, T., Premkumar, L. S. & Wollmuth, L. P. DRPEER: a motif in
the extracellular vestibule conferring high Ca2+ flux rates in NMDA receptor channels. J.
Neurosci. 22, 10209–10216 (2002).
67. Schorge, S. & Colquhoun, D. Studies of NMDA receptor function and stoichiometry with
truncated and tandem subunits. J. Neurosci. 23, 1151–1158 (2003).
68. Balannik, V., Menniti, F. S., Paternain, A. V., Lerma, J. & Stern-Bach, Y. Molecular
mechanism of AMPA receptor noncompetitive antagonism. Neuron 48, 279–288 (2005).
69. Lipton, S. A. Paradigm shift in neuroprotection by NMDA receptor blockade: Memantine
and beyond. Nature Rev. Drug Discov. 5, 160–170(2006).
70. Zuo, J. et al. Neurodegeneration in Lurcher mice caused by mutation in delta2 glutamate
receptor gene. Nature 388, 769–773 (1997).
71. Kohda, K., Wang, Y. & Yuzaki, M. Mutation of a glutamate receptor motif reveals its role in
gating and delta2 receptor channel properties. Nature Neurosci. 3, 315–322 (2000).
72. Klein, R. M. & Howe, J. R. Effects of the lurcher mutation on GluR1 desensitization and
activation kinetics. J. Neurosci. 24, 4941–4951 (2004).
73. Tomita, S. et al. Stargazin modulates AMPA receptor gating and trafficking by distinct
domains. Nature 435, 1052-1058 (2005).
74. Turetsky, D., Garringer, E. & Patneau, D. K. Stargazin modulates native AMPA receptor
functional properties by two distinct mechanisms. J. Neurosci. 25, 7438–7448 (2005).
75. Priel, A. et al. Stargazin reduces desensitization and slows deactivation of the AMPA-type
glutamate receptors. J. Neurosci. 25, 2682–2686 (2005).
Acknowledgements Work in the author’s laboratory is supported by the intramural
research programme of NICHD, NIH, DHHS. Synchotron diffraction data were
collected at Southeast Regional Collaborative Access Team (SER-CAT) 22-ID
beamline at the Advanced Photon Source, Argonne National Laboratory. Use of the
Advanced Photon Source was supported by the US Department of Energy, Office of
Science, Office of Basic Energy Sciences.
Author Information Reprints and permissions information is available at
npg.nature.com/reprints and permissions. The author declares no competing
financial interests. Correspondence and requests for materials should be addressed
to the author (email@example.com).
NATURE|Vol 440|23 March 2006