The Plant Journal (2003) 35, 800±810 doi: 10.1046/j.1365-313X.2003.01849.

A role for glycine in the gating of plant NMDA-like receptors

Christian Dubos1,y, David Huggins2,y, Guy H. Grant2, Marc R. Knight1 and Malcolm M. Campbell1,
1
Department of Plant Sciences, University of Oxford, South Parks Road, Oxford OX1 3RB, UK, and
2
Department of Chemistry, Physical and Theoretical Chemistry Laboratory, University of Oxford, South Parks Road,
Oxford OX1 3QZ, UK

Received 2 May 2003; revised 25 June 2003; accepted 4 July 2003.


For correspondence (fax ‡44 1865 274074; e-mail malcolm.campbell@plants.ox.ac.uk).
y
These authors contributed equally to this work.

Summary
The amino acid glycine has a well-established role in signalling in the mammalian central nervous system.
For example, glycine acts synergistically with the major excitatory neurotransmitter, glutamate, to regulate
the in¯ux of ions such as calcium, through N-methyl-D-aspartate (NMDA) receptors. Plants possess NMDA-
like receptors, generically referred to as glutamate receptors (GLRs), named on the basis of their presumed
ligand, glutamate. Previously, glycine has not been implicated in plant GLR activity or any other aspect of
plant signalling. Using transgenic Arabidopsis seedlings expressing aequorin to monitor ligand-mediated
changes in the cytosolic concentration of Ca2‡ ([Ca2‡]cyt), the data presented herein show that glutamate
and glycine act synergistically to control ligand-mediated gating of calcium in plants. Glutamate and gly-
cine synergism also regulates hypocotyl elongation. Transient increases in [Ca2‡]cyt mediated by glutamate
and glycine, as well as hypocotyl elongation, were inhibited by 6,7-dinitroquinoxaline-2,3 dione (DNQX), a
competitive inhibitor of animal GLRs. Using a multiscale docking algorithm in combination with a mole-
cular model of the ligand-binding domain of plant GLRs, evidence is provided indicating that glycine, and
not glutamate, is likely to be the natural ligand for most plant GLR subunits. These ®ndings uncover a
hitherto unconsidered role for glycine signalling in plants, and suggest that the synergistic action of glu-
tamate and glycine at NMDA-like receptors predates the divergence of plants and animals.

Keywords: glycine, calcium, ionotropic glutamate receptors, NMDA.

Introduction
Mammalian ionotropic glutamate receptors (GLRs) are
multimeric, ligand-gated channels, comprised of four or
®ve subunits that function to perceive glutamate during
neurotransmission (Armstrong et al., 1998). Binding of
glutamate to the ligand-binding domain of these receptors
results in gating of the transmembrane channel, allowing
cations to enter the cell. There are three different groups of
mammalian ionotropic GLRs (AMPA, N-methyl-D-aspartate
(NMDA) and Kainate receptors), which are based on their
pharmacological properties and structural similarities
(Armstrong et al., 1998).
Twenty putative GLR subunits are encoded in the Arabi-
dopsis thaliana genome, and these show the greatest
sequence similarity to the mammalian NMDA receptor
subunits (Chiu et al., 1999, 2002; Davenport, 2002; Lacombe
et al., 2001; Lam et al., 1998). NMDA receptors are com-
posed of assemblies of three different classes of subunits,
NR1, NR2 and NR3 (Das et al., 1998). Each subunit has three
800
trans-membrane regions, one membrane-embedded
domain and two domains that form the ligand-binding site
(Figure 1). This structure is also predicted for the plant
receptors (Chiu et al., 1999). Based on sequence similarity,
it has been assumed that plant GLRs may function in a
manner analogous to mammalian GLRs.
Mammalian ionotropic GLRs, such as NMDA receptors,
function as ligand-gated channels for the transport of ions
such as Ca2‡ or Na‡ (Scatton, 1993). Similarly, plant GLRs
have been predicted to function as Ca2‡ channels that are
gated by glutamate (Davenport, 2002; White et al., 2002).
Several lines of evidence suggested that genes encoding
GLR subunits might be involved in regulating cellular and
developmental phenomena in plants. Treatment of
Arabidopsis seedlings with the GLR antagonist 6,7-dinitro-
quinoxaline-2,3 dione (DNQX) provided preliminary, sug-
gestive evidence for the involvement of GLRs in the control
of hypocotyl elongation (Lam et al., 1998). Treatment of
ß 2003 Blackwell Publishing Ltd

Figure 1. Schematic diagram of GLR structure.
The regions of the ligand-binding domain that form the focus of these
studies are highlighted in grey.
plants with a putative GLR agonist also suggested an
involvement of GLRs in hypocotyl elongation (Brenner
et al., 2000). Unrelated studies predicted that glutamate-
gated channels may be involved in regulating changes in
the cytosolic concentration of Ca2‡, [Ca2‡]cyt (Dennison and
Spalding, 2000). Overexpression of one member of the GLR
family in Arabidopsis implicated at least one GLR subunit in
the control of calcium homeostasis and related develop-
mental phenomena (Kim et al., 2001).
To date, studies on plant GLRs have focused on the
interaction of glutamate with these receptors (Chiu et al.,
1999; Davenport, 2002; Dennison and Spalding, 2000; Kim
et al., 2001; Lacombe et al., 2001; Lam et al., 1998), and
have not addressed the hypothesis that glutamate may not
be the sole natural ligand for these receptors in plants. This
is surprising, as mammalian NMDA receptor subunits,
which share the highest degree of sequence similarity with
the plant GLR subunits, are well known to bind to glycine. In
fact, NMDA receptors must bind not only glutamate, but
also the co-agonist, glycine, in order to function (Ivanovic
et al., 1998). Furthermore, the NR3 family of NMDA receptor
subunits are not responsive to glutamate, and can be gated
by glycine alone (Chatterton et al., 2002). In animals, the
ratio of glutamate-binding subunits to glycine-binding sub-
units is thought to determine the relative responsiveness of
the multisubunit NMDA receptor to both glutamate and
glycine (Chatterton et al., 2002).
Glycine has recently been shown to be a key modulator of
cell-to-cell signalling in the mammalian central nervous
system. In addition to the role that glycine plays in stimu-
lating NMDA receptors to initiate calcium signalling, gly-
cine stimulation of NMDA receptors also primes the
receptors for clathrin-mediated endocytosis (Nong et al.,
2003). Thus, glycine functions to modulate not only the
immediate response of neurons to neurotransmitters, but it
also conditions future responsiveness by altering the occur-
rence of receptors on the cell surface. Recognition of the
roles played by glycine has profoundly changed the under-
standing of synaptic signalling in the central nervous sys-
tem (Nong et al., 2003).
ß Blackwell Publishing Ltd, The Plant Journal, (2003), 35, 800±810
Unsuspected glycine signalling in plants 801
Data from the middle of the last century suggested that
glycine might function as a signalling molecule in plants. In
1939, White found that glycine favoured the growth of
excised tomato roots in the presence of appropriate com-
binations of carbohydrates, thiamine and inorganic salts
(White, 1939). Further work based on White's early obser-
vations suggested that cotyledon-derived glycine modu-
lated root development in concert with other amino acids in
a synergistic fashion (Fries, 1953; Skinner and Street, 1953).
In all these instances, glycine was thought to function
purely as a nutrient, and the potential involvement of
glycine as a signal in plants was not considered. The recent
discovery of NMDA-like GLRs in plants and the role that
glycine plays in gating these receptors in animals suggest
that it may be worthwhile to consider the potential involve-
ment of glycine signalling in plants.
In order to test the hypothesis that glycine may function
to gate plant GLRs, we have examined glycine-mediated
changes in [Ca2‡]cyt in transgenic Arabidopsis seedlings
expressing aequorin, a Ca2‡-sensitive luminescent protein
(Knight et al., 1991, 1996, 1999). In addition, the effects of
glutamate and glycine on hypocotyl elongation were inves-
tigated. In order to determine if any glutamate- or glycine-
mediated changes in [Ca2‡]cyt or in hypocotyl elongation
were dependent on GLRs, the plant material was also
analysed after treatment with DNQX. Finally, the interaction
of glutamate and glycine, as well as DNQX, with plant GLRs
was examined using a molecular modelling approach. The
latter approach was based on a multiscale docking algo-
rithm that has been used successfully to examine ligand
binding to proteins such as the anthrax toxin (Glick et al.,
2002a,b,c). Taken together, the ®ndings of these analyses
open the door to the possibility that glycine may function as
a signalling molecule in plants.
Results and discussion
Glycine mediates changes in [Ca2‡]cyt
Previous studies have shown that glutamate can induce
changes in the cytosolic concentration of [Ca2‡]cyt, and
it was suggested that this occurred through GLRs
(Dennison and Spalding, 2000). If plant GLRs function like
their mammalian counterparts, glycine would also be
predicted to gate GLRs and thereby change [Ca2‡]cyt. This
hypothesis was tested by measuring [Ca2‡]cyt in trans-
genic Arabidopsis seedlings expressing aequorin (Knight
et al., 1991, 1996, 1999). Consistent with the previous
®ndings (Dennison and Spalding, 2000), glutamate induced
an increase in [Ca2‡]cyt in a concentration-dependent
fashion (Figure 2a). However, glycine also induced an
increase in [Ca2‡]cyt in a concentration-dependent manner
(Figure 2a). Amino acids with similar structures, aspartate
802 Christian Dubos et al.

Figure 2. [Ca2‡]cyt changes in plants in response to application of glutamate (Glu), glycine (Gly), aspartate (Asp), alanine (Ala), DNQX or combinations thereof.
The dotted horizontal lines indicate the standard deviation around the control treatment, which was an injection of MS medium, with no effector, as a control for
the touch response.
(a) Arabidopsis seedlings were treated with the effectors at the concentrations indicated, and changes in [Ca2‡]cyt were determined as described previously by
Knight et al. (1996). An analysis of variance was used to compare the mean of each effector treatments relative to the control, which was MS medium alone; $:
signi®cant effect relative to control (injection of MS alone); P < 0.0005. Error bars show the SE of the mean.
(b) As per above, but a cold treatment was used to induce the change in [Ca2‡]cyt as described previously by Knight et al. (1996).

and alanine, did not induce any change in [Ca2‡]cyt

Glycine and glutamate modulation of [Ca2‡]cyt is
(Figure 2a).
DNQX-sensitive

In order to test the hypothesis that glutamate and glycine

Glycine and glutamate act synergistically to modulate
[Ca2‡]cyt
Seedlings were treated with a mixture of glutamate and
glycine to determine if the compounds acted synergisti-
cally. The combination of glutamate and glycine was effec-
tive at inducing an increase in [Ca2‡]cyt at concentrations
that were two orders of magnitude less than those required
to give an equivalent increase when each of the compounds
was added separately (Figure 2a). Addition of a combina-
tion of aspartate and alanine had no effect, even at con-
centrations that were 10-fold greater than the optimum for
glutamate and glycine (Figure 2a).
The increase in [Ca2‡]cyt in response to the combination
of glutamate and glycine did not appear to be dose-depen-
dent at concentrations above 0.01 mM. At even these low
concentrations, the increase in [Ca2‡]cyt was as great as that
observed when each amino acid was used on its own at the
maximum concentration of 10 mM. This was in contrast to
the increase in [Ca2‡]cyt observed when each amino acid
was added on its own, which was dose-dependent up to
1 mM. This suggests that the gating of Ca2‡ by the combi-
nation of glutamate and glycine was a saturable response,
and that saturation of the channel occurred at very low
concentrations of the amino acids. Given that 10 mM con-
centrations of each individual amino acid also did not
induce a statistically signi®cant increase in [Ca2‡]cyt relative
to that observed at 1 mM (Figure 2a) suggests that this
response is also saturable.
ß Blackwell Publishing Ltd, The Plant Journal, (2003), 35, 800±810
were eliciting their gating effect at GLRs, plants were pre-
treated with a compound that inhibits mammalian iono-
tropic GLRs, DNQX (Armstrong and Gouaux, 2000), prior to
treatment with the agonists. DNQX pre-treatment comple-
tely abolished the increase in [Ca2‡]cyt normally induced
by the addition of glutamate, or glycine, or both in com-
bination (Figure 2a). DNQX pre-treatment did not abolish
an increase in [Ca2‡]cyt brought about by cold treatment
(Figure 2b), showing that the effect of DNQX was neither
toxic nor generic, and that it was related to the perception of
glutamate and glycine. The ability of DNQX to block the
effect of these ligands suggests that they bind to the same
receptor as DNQX, an ionotropic GLR.
DNQX-sensitive glycine and glutamate modulation of
[Ca2‡]cyt occurs in cotyledons, but not in roots
Ligand-induced increases in [Ca2‡]cyt in whole plants were
monitored by visualisation of [Ca2‡]-dependent changes in
aequorin luminescence using a photon-counting camera
(Knight et al., 1999). This revealed that the glutamate- and
glycine-gated changes in [Ca2‡]cyt completely overlapped
in both time (Figure 3a) and location (Figure 3b). The
increase in [Ca2‡]cyt took place throughout the plant body,
but was strongest in the cotyledons and the root, regardless
of whether the ligand used was glutamate, or glycine, or a
combination of both (Figure 3b). DNQX completely inhi-
bited ligand-gated changes in [Ca2‡]cyt in aerial tissue, but

Figure 3. [Ca2‡]cyt changes in plants in response to application of glutamate
(Glu), glycine (Gly), DNQX or combinations thereof.
Arabidopsis seedlings were treated with the effectors at the concentrations
indicated, and changes in [Ca2‡]cyt were determined as described previously
by Knight et al. (1996). Ligand-based changes in [Ca2‡]cyt were (a) examined
luminometrically or (b) visualised using a photon-counting camera as
described previously by Knight et al. (1999).
only slightly changed the activity in the root. These results
reveal that GLRs found in aerial tissues of Arabidopsis
behave in a manner analogous to their mammalian coun-
terparts in that they are DNQX-sensitive. These results also
suggest that either DNQX does not penetrate the root to
inhibit GLR activity there, or, that at least two types of
receptors function to perceive glutamate and glycine in
Arabidopsis: a DNQX-sensitive form in aerial tissues and
another that is DNQX-insensitive found in the roots.
DNQX-mediated hypocotyl elongation is competitively
inhibited by glycine and glutamate
Plant GLRs have been implicated in the control of an
important component of light-mediated development ±
hypocotyl elongation (Brenner et al., 2000; Lam et al.,
1998). As has been shown previously (Lam et al., 1998),
plants grown in the presence of DNQX have elongated
hypocotyls (Figure 4a). If DNQX elicits this response by
antagonising plant GLRs, then simultaneous application
of GLR agonists should reverse this effect. Both glutamate
and glycine reversed DNQX-induced hypocotyl elongation
in a concentration-dependent fashion (Figure 4). Glycine
was at least as effective as glutamate at reversing the DNQX
ß Blackwell Publishing Ltd, The Plant Journal, (2003), 35, 800±810
Unsuspected glycine signalling in plants 803
effect. Glycine and glutamate functioned synergistically
when applied together with DNQX to seedlings, function-
ing at signi®cantly lower concentrations than when they
were used alone (Figure 4b).
While DNQX pre-treatment of seedlings completely
inhibited ligand-gated changes in [Ca2‡]cyt in aerial tissues,
as monitored by aequorin (Figure 3b), it is clear that growth
of seedlings in medium containing a mixture of DNQX and
amino acids allows for competition to occur between the
agonists and the antagonist, so that the effect is not abso-
lute but concentration-dependent. It is important to note
that the experiments with aequorin document the ability of
DNQX pre-treatment to block the immediate change in
[Ca2‡]cyt, and do not indicate the long-term effect on
[Ca2‡]cyt, gating by the three ligands acting together. The
growth of plants in a combination of DNQX with the amino
acids would almost certainly result in a dynamic competi-
tion for the binding site by the inhibitor and the two
agonists. Furthermore, long-term exposure of plants to
DNQX may alter the abundance of GLRs, and thereby alter
plant responsiveness to the agonists. In animals, the abun-
dance of GLRs is affected by the extent to which a neuron is
stimulated by neurotransmitters. The greater the stimula-
tion, the lower the abundance of GLRs and vice versa
(Platenik et al., 2000; Scannevin and Huganir, 2000; Sheng
and Kim, 2002). Similarly, continual exposure of plants to
DNQX and concomitant decreases in GLR stimulation may
lead to an increase in the abundance of GLRs, which would,
in turn, enhance plant responsiveness to any glutamate and
glycine that is present. This may account for the ability of
glutamate and glycine to reverse the effects of DNQX when
plants are grown in a mixture of these ligands. Finally, it
may also be that the combined action of glutamate and
glycine reverses the inhibitory effects of DNQX through an
alternative signalling mechanism, which may be calcium-
independent and which could account for the apparent
differences between the calcium in¯ux data and the hypo-
cotyl elongation data.
Molecular modelling shows binding of plant GLRs to
glutamate, glycine and DNQX
As was the case in previous studies, the results described
herein provide circumstantial evidence for the interaction
between plant GLRs and their putative ligand, glutamate.
Beyond this, however, the current results provide more
comprehensive evidence in support of the hypothesis that
plant GLRs function like their mammalian counterparts in
that the gating of [Ca2‡]cyt is modulated by both glutamate
and glycine, and that this is inhibited by DNQX. Further-
more, the results presented herein show that this signalling
mechanism may be involved in the control of hypocotyl
elongation. Nevertheless, it remains to be determined if any
of the compounds in question actually interact with plant
804 Christian Dubos et al.
GLRs. As an important ®rst step in determining if gluta-
mate, glycine and DNQX are able to bind to plant GLRs, a
molecular modelling approach was employed.
The structure of the extracellular ligand-binding region of
an A. thaliana GLR subunit (AtGLR) was modelled on the
crystal structure of the analogous region of a Rattus nor-
vegicus GLR (PDB ID 1FTJ; Armstrong and Gouaux, 2000).
The structure of the rat GLR and its interaction with its
ligands was empirically determined by crystallography as
well as detailed functional analyses of GLR-ligand interac-
tions (Armstrong and Gouaux, 2000; Armstrong et al., 1998;
Yoneda and Ogita, 1991). Using a multiscale docking algo-
rithm that provides accurate predictions of protein±ligand
interactions (Glick et al., 2002a,b,c), the nature of the inter-
action of the plant GLRs with glutamate and glycine was
explored. The A. thaliana GLR, AtGLR2.9, was chosen for
the modelling study as it exhibited the greatest sequence
similarity to the R. norvegicus GLR, particularly in those
regions implicated in ligand binding (Figures 5 and 6).
The multiscale docking algorithm runs a series of pro-
gressive iterations to explore the three-dimensional space
of a protein molecule for the most likely binding sites of a
chosen ligand. In early iterations, many possible binding
sites are found, as all surfaces of the protein are available
for ligand binding. As the algorithm proceeds, preferred
binding sites are identi®ed as they are `discovered' iteration
upon iteration. For those proteins where there is a good ®t
ß Blackwell Publishing Ltd, The Plant Journal, (2003), 35, 800±810
Figure 4. Variation in hypocotyl elongation in
response to application of glutamate (Glu), gly-
cine (Gly), DNQX or combinations thereof.
(a) Arabidopsis seedlings were grown in med-
ium containing the effectors at the concentra-
tions indicated, and changes in hypocotyl
elongation were visualised.
(b) The quantitative response to the effectors
was determined. Values represent the mean of a
minimum of 15 seedlings. Pairwise t-tests were
used to compare the mean of each treatment
against the mean for the seedlings grown in MS
plus DNQX ($: signi®cant difference; P <
0.0001) or MS alone (*: no signi®cant differ-
ence; P < 0.0001). Error bars show the SE of the
mean.
between the protein and the ligand, the iterations reveal
that only one or several overlapping protein sites are opti-
mal for that ligand.
Using this approach, the docking algorithm con®rmed
that glutamate bound to the R. norvegicus GLR at the site
predicted by the crystal structure (Figures 7a±c and 8a,b). In
contrast, attempts to dock glutamate to the AtGLR2.9
model failed. The ®nal iteration revealed that there was a
high likelihood that glutamate would only associate with
the surface of the protein, and not in the predicted ligand-
binding site (Figure 7d±f). Furthermore, attempts to `force'
glutamate into the predicted ligand-binding site in silico
showed that this could not occur naturally because of
signi®cant steric hindrance (Figure 8c,d). This is attributa-
ble to the replacement of one of the key residues that are
required for binding of the side chain carboxylate group of
glutamate (Figure 8d). The residue in question, Thr655 of
the R. norvegicus GLR, is conserved amongst all known
mammalian GLRs (Figure 5). In AtGLR2.9, Thr655 is
replaced by a bulky, hydrophobic Phe residue, which blocks
the ligand-binding pocket where glutamate would dock
(Figures 6 and 8d). In fact, Thr655 is replaced either by
Phe or the bulky, hydrophobic branch-chain amino acids,
Leu or Ile, in 18 of the 20 AtGLR subunits, and is missing
altogether from one other (Figure 5). This suggests that
glutamate is not the natural ligand for the majority of AtGLR
subunits.
 Unsuspected glycine signalling in plants 805

Figure 5. Multiple sequence alignment of GLR

sequences.
The peptide sequences (and their correspond-
ing accession numbers) that were used in the
alignment included: R. norvegicus GLRs, Rat-
GlurB (CAA38465), RatKain1 (AAA02873), NR2a
(AAC03565),NR2b(AAA41714),NR2c(AAA41713),
NR2d (AAC37647), NR1a (AAA16366) and
NR3b (AAL69893); Homo sapiens GLR, NR3a
(AAL40734) and A. thaliana GLRs, AtGLR1.1
(AAF26802.1), AtGLR1.2 (BAA96960.1), AtGLR1.3
(BAA96961.2), AtGLR1.4 (AAF02156.1), AtGLR2.1
(AAB61068.1), AtGLR2.2 (AAD26895.1), AtGLR2.3
(AAD26894.1), AtGLR2.4 (CAA19752.1), AtGLR2.5
(CAB96656.1), AtGLR2.6 (CAB96653.1), AtGLR2.7
(AAC33239.1),AtGLR2.8(AAC33237.1),AtGLR2.9
(AAC33236.1), AtGLR3.1 (AAF63223.1), AtGLR3.2
(CAA18740.1), AtGLR3.3 (AAG51316.1), AtGLR3.4
(AAB71458.1),AtGLR3.5(AAC69939.1),AtGLR3.6
(CAB63012.1) and AtGLR3.7 (AAC69938.1). Resi-
duesimportantfor ligandbinding arehighlighted
in grey in domain one (a) and domain two (b).

ß Blackwell Publishing Ltd, The Plant Journal, (2003), 35, 800±810

806 Christian Dubos et al.

There are two types of mammalian NMDA receptor sub-

Figure 6. A multiple sequence alignment of the R. norvegicus GLR
(AAA41240) with the A. thaliana GLRs, AtGLR2.9 (AAC33236.1) and
AtGLR1.1 (AAF26802.1) shows locations of residues that have been impli-
cated in ligand binding highlighted in grey.
The two regions that bind glutamate are shown. In domain one, relatively
highly conserved residues at Pro478, Thr480 and Arg485 bind the backbone
of glutamate ($). In domain two, residues corresponding to Ser654, Thr655
and Glu705 from the R. norvegicus GLR also bind to glutamate (^).
units: glutamate and glycine receptors (Ivanovic et al.,
1998). Glutamate is bound by NR2 subunits, whereas gly-
cine is bound by NR1 and NR3 subunits (Ivanovic et al.,
1998). All NR2 subunits possess a conserved glutamate-
binding GST motif, corresponding to Gly653, Ser654 and
Thr655 in the glutamate-binding site of the R. norvegicus
GLR (Figure 5). In contrast, most of the AtGLR sequences
are similar to the glycine-binding subunits NR1 and NR3,
where Thr655 is replaced by amino acids with large, hydro-
phobic side chains (Figure 5).
To determine if glycine could function as a ligand for
plant GLR subunits, glycine was docked to the AtGLR2.9
model. Glycine preferentially docks with the proposed glu-
tamate-binding site (Figures 7g±i and 8e,f). The data
strongly suggest that glycine is the natural agonist for
the majority of AtGLRs. The one exception to this is

Figure 7. `Snapshots' of the cumulative iterations from the multiscale docking algorithm, for the interactions of GLR subunits and putative ligands.
The full ligand-binding domain, showing the amino-acid residues important in binding in the ligand-binding pocket (in yellow), is shown with the potential
binding sites (green dots).
(a) An early iteration of the R. norvegicus GLR docking with glutamate.
(b) A middle iteration of the R. norvegicus GLR docking with glutamate.
(c) The last cumulative iteration of the R. norvegicus GLR docking with glutamate. Note that all potential binding sites cluster in the ligand-binding pocket.
(d) An early iteration of AtGLR2.9 docking with glutamate.
(e) A middle iteration of AtGLR2.9 docking with glutamate.
(f) The last cumulative iteration of AtGLR2.9 docking with glutamate. Arrows denote the fact that no speci®c binding site has been identi®ed.
(g) An early iteration of AtGLR2.9 docking with glycine.
(h) A middle iteration of AtGLR2.9 docking with glycine.
(i) The last cumulative iteration of AtGLR2.9 docking with glycine. Note that all potential binding sites coalesce to unify in a single site in the ligand-binding
pocket.

ß Blackwell Publishing Ltd, The Plant Journal, (2003), 35, 800±810

 Unsuspected glycine signalling in plants 807

Figure 8. Molecular models of ligands binding

to GLR subunits.
The full ligand-binding domains showing the
amino-acid residues important in binding (in
green) are shown with the ligands (in red) in
panels (a,c,e,g,i), whereas the detailed ligand-
binding pockets with the amino acids impli-
cated in binding are shown with the ligands
in panels (b,d,f,h,j).
(a) Model of the ligand-binding site of the
R. norvegicus GLR docked with glutamate.
(b) Detailed model of the ligand-binding pocket
of the R. norvegicus GLR docked with gluta-
mate.
(c) Model of the ligand-binding site of AtGLR2.9
docked with glutamate, showing the regions of
steric hindrance as coloured spheres.
(d) Detailed model of the ligand-binding pocket
of AtGLR2.9 docked with glutamate, showing
the regions of steric hindrance as coloured
spheres.
(e) Model of the ligand-binding site of AtGLR2.9
binding to glycine.
(f) Detailed model of the ligand-binding pocket
of AtGLR2.9 binding to glycine.
(g) Model of the ligand-binding site of AtGLR1.1
binding to glutamate.
(h) Detailed model of the ligand-binding pocket
of AtGLR1.1 binding to glutamate.
(i) Model of the ligand-binding site of AtGLR2.9
binding to DNQX.
(j) Detailed model of the ligand-binding pocket
of AtGLR2.9 binding to DNQX.

AtGLR1.1, which contains the glutamate-binding GST motif
in the correct region of domain 2 (Figures 5 and 6). When
the ligand-binding domain of AtGLR1.1 was modelled on
the R. norvegicus receptor, glutamate bound to the same
residues as in the mammalian GLRs (Figure 8g,h). Thus,
while the majority of AtGLR subunits should bind glycine,
we predict that plant GLR channels may function with both
glycine and glutamate. This prediction is entirely consistent
with the observations that we made with respect to both the
gating of [Ca2‡]cyt and the changes in hypocotyl elongation.
ß Blackwell Publishing Ltd, The Plant Journal, (2003), 35, 800±810
Analysis of the modelling results highlights residues
implicated in glutamate binding in mammalian GLRs that
also appear important for glycine binding in the AtGLRs. For
example, the Glu and Arg residues in the binding site appear
to be important for ligand binding in both the R. norvegicus
GLR (Glu705, Arg485) and the AtGLR (Glu724, Arg529).
Both are perfectly conserved over all receptors investigated
(Figure 5). Similarly, other amino acids implicated in estab-
lishing the three-dimensional structure of the ligand-bind-
ing region of domain 1, such as Thr480 and Ile481, are
808 Christian Dubos et al.
conserved between species (Figure 5). In contrast, the bind-
ing-site Pro residue at position 478 in the R. norvegicus GLR
is replaced by Asp522 in AtGLR2.9, and appears to bind to
the glycine via a side chain interaction. The replacement of
Pro with Asp would anchor glycine more strongly within
the binding site. In domain 2, Ser654 and Thr655 of the
R. norvegicus GLR are replaced by non-polar Ala and Phe
residues, respectively, in the AtGLR. The Ala681 residue in
the AtGLR is predicted to bind glycine via its backbone
nitrogen (much as Ser 654 binds glutamate in mammalian
GLRs). Crucially, the Phe682 residue in AtGLR is not pre-
dicted to interact with glycine; rather, the presence of this
amino acid hinders binding of other amino acids, such as
glutamate, because of the presence of their side chains.
DNQX appears to have the capacity to bind to all plant
GLR subunits. This ®nding supports the hypothesis that
DNQX interacts with plant GLRs (Lam et al., 1998). Model-
ling with AtGLR2.9 showed that DNQX bound in the pre-
dicted ligand-binding pocket (Figure 8i,j). Thus, DNQX
would be predicted to compete for the plant GLR ligand-
binding pocket with other compounds, such as GLR ago-
nists. This is entirely consistent with the predicted interac-
tion of DNQX with animal GLRs. Importantly, these ®ndings
are also consistent with the observation that glutamate and
glycine compete with DNQX in the gating of [Ca2‡]cyt
(Figure 2) and in the regulation of hypocotyl elongation
(Figure 4), and substantiate the contention that glutamate
and glycine mediate their effect at plant GLRs.
Future studies should aim to obtain empirical evidence
for ligand binding to plant GLR subunits. The ®ndings
presented here suggest that this may not be a trivial task.
Given the fact that the animal receptors function as hetero-
meric channels (Armstrong et al., 1998) and given the pre-
ponderance of plant genes encoding the different subunits,
it may prove dif®cult to devise experimental systems that
allow the re-constitution of the plant channels to provide an
accurate indication of the in vivo ligands. Indeed, animal
GLRs are unable to bind to their ligands when expressed as
homomeric channels. However, the modelling work pre-
sented here should provide guidance for future studies
aimed at determining the ligands for these receptors, at
least in vitro. Where glycine might not have previously been
considered as a ligand for these receptors, or where dif®-
culties may have been encountered when using glutamate
as a ligand, new alternatives can be considered. Further-
more, the contention that these receptors may not function
as amino acid receptors, based on lack of glutamate bind-
ing (Davenport, 2002), may have to be re-assessed.
Conclusion
These studies provide evidence that glycine signalling may
be widespread, to include the plant kingdom. The results
show that the activation of GLRs by the synergistic action of
ß Blackwell Publishing Ltd, The Plant Journal, (2003), 35, 800±810
glutamate and glycine is an ancient mechanism, which
predates the split between plants and animals some
1000 million years ago. Furthermore, the preponderance
of plant GLRs that can bind glycine, relative to those that
bind glutamate, suggests that glycine receptor subunits
may have diversi®ed to function in different developmental
contexts. As is the case in animals, where timing and locali-
sation of the expression of different NMDA receptor sub-
units can alter cell responsiveness to agonists (Chatterton
et al., 2002), plants may also alter their responsiveness to
glycine by altering the expression of different GLR subunits.
Indeed, the subunit encoded by AtGLR1.1, which is the only
subunit that the modelling results predict to bind glutamate,
may be amongst those subunits that is the most important
in altering GLR speci®city. Recent evidence suggests that
AtGLR1.1 does play a role in perceiving C:N balance in
Arabidopsis (Kang and Turano, 2003), but it remains to be
determined if this relates to the perception of the glycine.
While earlier studies suggested that glycine might func-
tion as a signal in plants (Fries, 1953; Skinner and Street,
1953; White, 1939), this possibility was not explored. It is
tempting to speculate that glycine may function as a sig-
nalling molecule in plants. Diurnal changes in glycine levels
have been documented in tobacco. In fact, in tobacco, the
levels of glycine in the leaves increased by a factor of almost
sixfold from the end of the night until the plants were
exposed to 9 h of daylight (Geiger et al., 1998). This was
in contrast to the concentration of glutamate in the leaves,
where the levels remained almost constant over the same
time period (Geiger et al., 1998). Furthermore, growth of
tobacco plants in elevated carbon dioxide had very little
effect on the normal diurnal changes in glutamate levels, but
dampened the extent of the ¯uctuations in glycine concen-
tration in the leaves (Geiger et al., 1998). These changes in
glycine levels and the responsiveness of the levels to altered
carbon dioxide suggest that glycine could function as a sig-
nalling molecule, at least in tobacco. It is notable that glu-
tamate remained virtually unchanged over the time period
that was monitored, while glycine levels increased during
the day. This observation is consistent with glycine func-
tioning as the signalling molecule, as opposed to glutamate.
These observations, coupled with the ®ndings presented
herein, open the door to future studies aimed at investigating
the hitherto undiscovered role of glycine in plant signalling
and to explore the role of glycine signalling in organism
development that extends beyond the nervous system.
Experimental procedures
Growth of Arabidopsis plants
Seeds from A. thaliana wild-type plants (Col-0 ecotype) were
sterilised by sequential treatment with 70% ethanol, satura-
ted calcium hypochlorite and sterile water, and then germinated

in liquid Murashige and Skoog (MS) medium (Gibco-BRL,
Gaithersberg, MD, USA) containing 0.1% MES [2-(-N-morpholi-
no)ethanesulfonic acid]. Following sterilisation, seeds were incu-
bated in MS for 2 days at 48C to ensure uniform germination.
Plants were at 228C under long-day conditions (16-h light/8-h dark)
with an average light intensity of 60.58 mmol m 2 sec 1. All
compounds used for treatments were purchased from Sigma
(Sigma±Aldrich, St Louis, MO, USA).
Measurement of changes in [Ca2‡]cyt as detected
by aequorin
Arabidopsis seedlings were grown on MS liquid medium as
described above. Re-constitution of aequorin was performed
in vivo, essentially as described previously by Knight et al.
(1991). Changes in calcium following application of effectors
was determined as previously described by Knight et al. (1996).
Calcium-dependent aequorin luminescence was imaged as
described previously by Knight et al. (1999), using an intensi®ed
charge-coupled device (CCD) camera (model EDC-02), with a cam-
era control unit (HRPCS-2) and image acquisition and processing
software (IFS216), all from Photek (St. Leonards-on-Sea, UK).
Measurement of hypocotyl length
After the growing period, MS was removed and replaced by 100%
ethanol. Hypocotyls were scanned and measured using NIH-
IMAGE software (version 1.62; National Institute of Health, USA).
Multiple sequence alignments and molecular modelling
The crystal structure of the R. norvegicus glutamate receptor
construct in complex with glutamate was obtained from the Pro-
tein Data Bank (PDB ID 1FTJ). The construct was based on the
receptor GlurB (accession number AAA41240) and contained the
two glutamate-binding domains (S1 and S2) joined by a linker
region (Armstrong and Gouaux, 2000; Armstrong et al., 1998).
The rat glutamate receptor sequence and the plant AtGLR sequen-
ces were initially aligned using both Insight Homology (Accelrys,
San Diego, CA, USA) and PSIPRED (McGuf®n et al., 2000) algo-
rithms. The two sequences from AtGLR2.9 and AtGLR1.1, corres-
ponding to the two glutamate-binding domains, were removed
from the plant sequences and joined together by the linker
sequence. The homology package of Insight II was then used to
assign co-ordinates to the amino acid sequence of AtGLR2.9 and
AtGLR1.1. The models obtained were then re®ned by CHARMM to
optimise the three-dimensional structure (Brooks et al., 1983).
The ligands were then docked to this structure using the multi-
scale docking algorithm (Glick et al., 2002a). In each case, the
ligand±protein complex was re®ned by CHARMM to optimise
the structure after docking was complete (Brooks et al., 1983).
Acknowledgements
We are most grateful to Drs Mark Fricker and Heather Knight for
their constructive criticism and encouragement. We would also
like to thank three anonymous reviewers for very thoughtful and
useful comments on the manuscript. D.H. was supported by a
studentship from the Engineering and Physical Sciences Research
Council, UK. This work was supported by grants from the Biotech-
nology and Biological Sciences Research Council of the UK to
M.M.C.
ß Blackwell Publishing Ltd, The Plant Journal, (2003), 35, 800±810
Unsuspected glycine signalling in plants 809
References
Armstrong, N. and Gouaux, E. (2000) Mechanisms for activation
and antagonism of an AMPA-sensitive glutamate receptor:
crystal structures of the GluR2 ligand binding core. Neuron.
28, 165±181.
Armstrong, N., Sun, Y., Chen, G.Q. and Gouaux, E. (1998) Structure
of a glutamate-receptor ligand-binding core in complex with
kainate. Nature, 395, 913±917.
Brenner, E.D., Martinez-Barboza, N., Clark, A.P., Liang, Q.S.,
Stevenson, D.W. and Coruzzi, G.M. (2000) Arabidopsis mutants
resistant to S(‡)-beta-methyl-alpha, beta-diaminopropionic
acid, a cycad-derived glutamate receptor agonist. Plant Physiol.
124, 1615±1624.
Brooks, B.R., Bruccoleri, R.E., Olafson, B.D., States, D.J.,
Swaminathan, S. and Karplus, M. (1983) CHARMM: a program
for macromolecular energy, minimization, and dynamics calcu-
lations. J. Comput. Chem. 4, 187±217.
Chatterton, J.E., Awobuluyi, M., Premkumar, L.S. et al. (2002)
Excitatory glycine receptors containing the NR3 family of NMDA
receptor subunits. Nature, 415, 793±798.
Chiu, J., DeSalle, R., Lam, H.M., Meisel, L. and Coruzzi, G. (1999)
Molecular evolution of glutamate receptors: a primitive signal-
ing mechanism that existed before plants and animals diverged.
Mol. Biol. Evol. 16, 826±838.
Chiu, J.C., Brenner, E.D., DeSalle, R., Nitabach, M.N., Holmes, T.C.
and Coruzzi, G.M. (2002) Phylogenetic and expression analysis
of the glutamate-receptor-like gene family in Arabidopsis thali-
ana. Mol. Biol. Evol. 19, 1066±1082.
Das, S., Sasaki, Y.F., Rothe, T. et al. (1998) Increased NMDA current
and spine density in mice lacking the NMDA receptor subunit
NR3A. Nature, 393, 377±381.
Davenport, R. (2002) Glutamate receptors in plants. Ann. Bot.
(Lond.), 90, 549±557.
Dennison, K.L. and Spalding, E.P. (2000) Glutamate-gated calcium
¯uxes in Arabidopsis. Plant Physiol. 124, 1511±1514.
Fries, N. (1953) Limiting factors in the growth of the pea seedling
root. Physiol. Plant. 6, 292±300.
Geiger, M., Walch-Liu, P., Engels, C., Harnecker, J., Schulze, E.-D.,
Ludewig, F., Sonnewald, U., Scheible, W.-R. and Stitt, M. (1998)
Enhanced carbon dioxide leads to a modi®ed diurnal rhythm of
nitrate reductase activity in older plants, and a large stimulation
of nitrate reductase activity and higher levels of amino acids in
young tobacco plants. Plant Cell Environ. 21, 253±268.
Glick, M., Grant, G.H. and Richards, W.G. (2002a) Docking of
¯exible molecules using multiscale ligand representations. J.
Med. Chem. 45, 4639±4646.
Glick, M., Grant, G.H. and Richards, W.G. (2002b) Pinpointing
anthrax-toxin inhibitors. Nat. Biotechnol. 20, 118±119.
Glick, M., Robinson, D.D., Grant, G.H. and Richards, W.G. (2002c)
Identi®cation of ligand binding sites on proteins using a multi-
scale approach. J. Am. Chem. Soc. 124, 2337±2344.
Ivanovic, A., Reilander, H., Laube, B. and Kuhse, J. (1998) Expres-
sion and initial characterization of a soluble glycine binding
domain of the N-methyl-D-aspartate receptor NR1 subunit. J.
Biol. Chem. 273, 19933±19937.
Kang, J. and Turano, F.J. (2003) The putative glutamate receptor
1.1 (AtGLR1.1) functions as a regulator of carbon and nitrogen
metabolism in Arabidopsis thaliana. PNAS, 100, 6872±6877.
Kim, S.A., Kwak, J.M., Jae, S.K., Wang, M.H. and Nam, H.G. (2001)
Overexpression of the AtGluR2 gene encoding an Arabidopsis
homolog of mammalian glutamate receptors impairs calcium
utilization and sensitivity to ionic stress in transgenic plants.
Plant Cell Physiol. 42, 74±84.
810 Christian Dubos et al.
Knight, M.R., Campbell, A.K., Smith, S.M. and Trewavas, A.J.
(1991) Transgenic plant aequorin reports the effects of touch
and cold-shock and elicitors on cytoplasmic calcium. Nature,
352, 524±526.
Knight, H., Trewavas, A.J. and Knight, M.R. (1996) Cold calcium
signaling in Arabidopsis involves two cellular pools and a change
in calcium signature after acclimation. Plant Cell, 8, 489±503.
Knight, H., Veale, E., Warren, G.J. and Knight, M.R. (1999) The sfr6
mutation in Arabidopsis suppresses low-temperature induction
of genes dependent on the CRT/DRE sequence motif. Plant Cell,
11, 875±886.
Lacombe, B., Becker, D., Hedrich, R. et al. (2001) The identity of
plant glutamate receptors. Science, 292, 1486±1487.
Lam, H.M., Chiu, J., Hsieh, M.H., Meisel, L., Oliveira, I.C., Shin, M.
and Coruzzi, G. (1998) Glutamate-receptor genes in plants.
Nature, 396, 125±126.
McGuf®n, L.J., Bryson, K. and Jones, D.T. (2000) The PSIPRED
protein structure prediction server. Bioinformatics, 16, 404±405.
Nong, Y., Huang, Y.Q., Ju, W., Kalia, L.V., Ahmadian, G., Wang,
Y.T. and Salter, M.W. (2003) Glycine binding primes NMDA
receptor internalization. Nature, 422, 302±307.
ß Blackwell Publishing Ltd, The Plant Journal, (2003), 35, 800±810
Platenik, J., Kuramoto, N. and Yoneda, Y. (2000) Molecular
mechanisms associated with long-term consolidation of the
NMDA signals. Life Sci. 67, 335±364.
Scannevin, R.H. and Huganir, R.L. (2000) Postsynaptic organization
and regulation of excitatory synapses. Nat. Rev. Neurosci. 1,
133±141.
Scatton, B. (1993) The NMDA receptor complex. Fundam. Clin.
Pharmacol. 7, 389±400.
Sheng, M. and Kim, M.J. (2002) Postsynaptic signaling and plas-
ticity mechanisms. Science, 298, 776±780.
Skinner, J.C. and Street, H.E. (1953) Studies on the growth of
excised roots. II. Observations on the growth of excised ground-
sel roots. New Phytologist, 53, 44±67.
White, P.R. (1939) Glycine in the nutrition of excised tomato roots.
Plant Physiol. 14, 527±538.
White, P.J., Bowen, H.C., Demidchik, V., Nichols, C. and Davies,
J.M. (2002) Genes for calcium-permeable channels in the
plasma membrane of plant root cells. Biochim. Biophys. Acta,
1564, 299±309.
Yoneda, Y. and Ogita, K. (1991) Neurochemical aspects of the
N-methyl-D-aspartate receptor complex. Neurosci. Res. 10, 1±33.