Prediction of 5-HT3 receptor agonist-binding residues using homology modeling.

Abstract

5-HT(3) receptors demonstrate significant structural and functional homology to other members of the Cys-loop ligand-gated ion channel superfamily. The extracellular domains of these receptors share similar sequence homology (approximately 20%) with Limnaea acetylcholine binding protein, for which an x-ray crystal structure is available. We used this structure as a template for computer-based homology modeling of the 5-HT(3) receptor extracellular domain. AutoDock software was used to dock 5-HT into the putative 5-HT(3) receptor ligand-binding site, resulting in seven alternative energetically favorable models. Residues located no more than 5 A from the docked 5-HT were identified for each model; of these, 12 were found to be common to all seven models with five others present in only certain models. Some docking models reflected the cation-pi interaction previously demonstrated for W183, and data from these and other studies were used to define our preferred models.

Full text

Page 1

2338 Biophysical JournalVolume 84April 20032338–2344
Prediction of 5-HT3Receptor Agonist-Binding Residues
Using Homology Modeling
David C. Reeves, Muhammed F. R. Sayed, Pak-Lee Chau, Kerry L. Price, and Sarah C. R. Lummis
Department of Biochemistry, University of Cambridge, Cambridge CB2 1AG, United Kingdom
ABSTRACT
ligand-gated ion channel superfamily. The extracellular domains of these receptors share similar sequence homology (;20%)
with Limnaea acetylcholine binding protein, for which an x-ray crystal structure is available. We used this structure as a template
for computer-based homology modeling of the 5-HT3receptor extracellular domain. AutoDock software was used to dock 5-HT
into the putative 5-HT3receptor ligand-binding site, resulting in seven alternative energetically favorable models. Residues
located no more than 5 A˚from the docked 5-HT were identified for each model; of these, 12 were found to be common to all
seven models with five others present in only certain models. Some docking models reflected the cation-p interaction previously
demonstrated for W183, and data from these and other studies were used to define our preferred models.
5-HT3receptors demonstrate significant structural and functional homology to other members of the Cys-loop
INTRODUCTION
The 5-HT3receptor is a member of the Cys-loop family
of LGIC. This family of proteins, which includes nACh,
GABAA, and glycine receptors, are responsible for fast sy-
naptic transmission at chemical synapses and are the target
sites of many neuroactive drugs; however, we do not yet
know the molecular details of their structure. The receptors
are pentamers, and are usually constituted of 2–4 different
subunits; each subunit has a large extracellular N-terminal
region and four putative transmembrane domains (M1–M4).
Two 5-HT3receptor subunits, A and B, have been char-
acterized (see Reeves and Lummis, 2002, for review), and
receptors appear to be able to function as either homo-oli-
gomeric (A only) or hetero-oligomeric (A and B) subunit
complexes (Davies et al., 1999). 5-HT3receptors have been
proposed to be evolutionarily the oldest members of the Cys-
loop LGIC family, and this, combined with the ability of
thesereceptorstofunctionashomo-oligomers,hasmeantthat
5-HT3receptors provide a useful model system for under-
standing critical features of all Cys-loop receptors (Reeves
and Lummis, 2002). Most work to date on this family of
proteins has been performed using nACh receptors, yet
despite many years of study, structural details of the protein-
ligand interactions at the molecular level remained unknown.
Evidence from early biochemical and mutagenesis studies on
nACh receptors indicated that it was the extracellular domain
that contained the ligand-binding site, a hypothesis that was
later confirmed by the construction of a chimaeric protein
consistingoftheN-terminaldomainofthea7neuronalnACh
receptor subunit linked to the C-terminal portion of the
5-HT3Areceptor subunit: the resulting chimera possessed
nACh receptor pharmacological properties and 5-HT3re-
ceptor channel properties (Eisele ´ et al., 1993). The de-
terminationofthestructureoftheAChBP(Brejcetal.,2001),
which is homologous not only to the extracellular domain of
the nACh receptor but also to those of other LGIC including
the 5-HT3receptor, allows definition of the ligand-binding
domain at a greater level of detail than before.
Computer-aided modeling of the extracellular domain of
5-HT3receptor has been attempted previously (Menziani
et al., 2001; Gready et al., 1997), but the structure of AChBP
suggests that these models may not reflect the true binding
site. Several homology-based models of the nACh recep-
tor and GABAAreceptor have been recently constructed
(Le Nove `re et al., 2002; Cromer et al., 2002; Schapira et al.,
2002). In this paper, we present a model of the 5-HT3re-
ceptor extracellular domain constructed by homology with
AChBP, and discuss the accuracy of the alternative docking
models for 5-HT in relation to experimental evidence and se-
quence alignment data.
MATERIALS AND METHODS
Sequence alignment
A sequence alignment of the extracellular domain of the 5-HT3receptor and
AChBP was generated using the program FUGUE (Shi et al., 2001), which
utilizes environment-specific substitutiontablesand structure-dependentgap
penalties.
Modeling
The three-dimensional model of the extracellular region of the 5-HT3
receptor was built using MODELLER (Sali and Blundell, 1993) based on
Submitted October 10, 2002, and accepted for publication December 3,
2002.
Address reprint requests to Sarah C. R. Lummis, Dept. of Biochemistry,
University of Cambridge, Tennis Court Rd., Cambridge CB2 1AG,
UK. Tel.: 144-1223-765950; Fax: 144-1223-333345; E-mail: sl120@
cam.ac.uk.
David C Reeves’ present address is Albert Einstein College of Medicine,
1300 Morris Park Ave., Bronx, NY 10461 USA.
Pak-Lee Chau’s present address is Bioinformatique Structurale, Institut
Pasteur, 75724 Paris, France.
Abbreviations used: 5-HT3, 5-hydroxytryptamine3; LGIC, ligand-gated
ion channel; ACh, acetylcholine; AChBP, acetylcholine binding protein;
nAChR, nicotinic acetylcholine receptor; GABA, gamma-amino butyric
acid; ECD, extracellular domain.
? 2003 by the Biophysical Society
0006-3495/03/04/2338/07$2.00

Page 2

the crystal structure of AChBP determined to 2.7 A˚. The pentamer was
generated by superimposing the model onto each protomer of AChBP;
special care was taken not to alter the coordinate axes of reference. The
generated pentamer model was then energy minimized in SYBYL using the
AMBER force field (Weiner et al., 1984) by moving side chains alone, to
relieve short contacts at the interprotomer interfaces. Electrostatic terms
were also included in these minimization cycles.
Ligand docking
The structure of 5-HT was constructed with idealized geometry using
MacroModel (Mohamadi et al., 1990), with hydrogen atoms added
assuming a pH of 7. The resultant structure, together with the modeled 5-
HT3R, were used as input for AutoDock 3.05 (Goodsell and Olson, 1990;
Morris et al., 1998). The number of points used in each Cartesian direction
was 60, thus giving a spacing of 0.375 A˚and a total of 216,000 points. The
grid center was manually set to an arbitrary point within the binding site
space. Ten genetic algorithm runs were performed on each docking exercise,
with a population size of 50, and the maximum number of generations was
set to 27,000. Seven docked structures were obtained. These structures were
then used as input for a program written by one of us (P.-L.C.) to extract all
amino acids that possess at least one atom within 5 A˚of the 5-HT ligand.
Potential hydrogen-bonding interactions were identified using SwissPDB-
Viewer(Guex andPeitsch,1997).Hydrogenbondsbetween 1.20and 2.76A˚
were allowed with a minimum angle of 1208. Figures were rendered using
POV-Ray 3.1 for Macintosh.
RESULTS
The sequence identity between the 5-HT3receptor ECD and
AChBP, as shown in Fig. 1, is relatively low (19% for the a-
subunit), although this value is similar to that between
AChBP and ACh receptor subunit ECDs (an average of
22%). However, the predicted secondary structure of
a typical nACh subunit ECD has 61.2% identity with
AChBP, and it has been calculated that the secondary
structure similarity between AChBP and the a7 nACh
receptor subunits is around 80% (Le Nove `re et al., 2002).
Thus the secondary structure of the closely related 5-HT3
receptor ECD, and indeed all Cys-loop LGIC ECDs, is likely
to be well represented by the AChBP structure. Our model of
the 5-HT3receptor extracellular domain therefore adopts an
immunoglobulin-like fold similar to that found for AChBP
(Grutter and Changeux, 2001).
The actual location of the agonist-binding site in AChBP
has not been finally determined, as the structure has not
yet been crystallized with ACh bound. However, a HEPES
molecule was identified in the putative binding pocket. As
this molecule has a charged amine like ACh, and the residues
associated with the binding site were those previously
described from labeling and mutagenesis studies, it seems
highly likely that this will prove to be the binding pocket
of the protein. This region, which is located at the subunit
interfaces, is indicated in Fig. 2 A. The relative orientations
of loops A–E, all of which contribute to the binding site in
our models, are shown in Fig. 2 B.
The model of the binding site of 5-HT can be described as
a cleft with a predominance of aromatic residues. Of these,
F226 and W90 are toward the membrane or lower side of
the cleft, whereas W183, Y234, and Y143 are toward the top
of the cleft and therefore would be closer to the presynaptic
cell. The docking procedure resulted in seven energetically
favorable models for positioning 5-HT in this binding site.
In models 1–3 (Fig. 3), 5-HT is oriented such that the pri-
mary amine is in the lower part of the cleft, with the rings of
5-HT sandwiched between the aromatic rings of Y234 and
W183, and there is the potential to form a hydrogen bond
with the main chain carboxyl of S182. The hydroxyl of 5-HT
appears to be in a hydrophilic pocket formed by R92, D229,
and Q151. The models differ in having subtle variations in
the locations of the primary amine and the rings in the cleft,
and this is reflected in the different lengths of the potential
hydrogen bond with S182: for model 1, 2.4 A˚; for model 2,
2.6 A˚; and for model 3, 2.2 A˚.
In models 4 and 5 (Fig. 3), 5-HT is oriented such that the
rings are at the lower part of the cleft and the primary amine
is between W183 and Y234. The hydroxyl group is located
in a hydrophilic pocket formed by T179, E236, and N128,
with the potential to form hydrogen bonds with the latter
two. There is also the potential to form a third hydrogen bond
with the main chain carboxyl of S182 (of 2.0 A˚, model 4) or
W183 (of 2.3 A˚, model 5). Models 4 and 5 differ by subtle
changes in the location of the primary amine—in model 4,
the nitrogen is 4.1 A˚from the nearest carbon of the Y234
ring and 4.8 A˚from the nearest carbon of the W183 benzene
ring, whereas in model 5 the distances are 3.9 A˚and 5.2 A˚,
respectively. There is also a difference in the twist of the
rings, which would result in different length hydrogen bonds
FIGURE 1
domain of the 5-HT3Areceptor subunit with that of AChBP. The alignment
is annotated using the program JOY (Mizuguchi et al., 1998). Key to
alignment: a-helix, red x; b-strand, blue x; 310helix, maroon X; solvent
accessible, lower case X; solvent inaccessible, upper case X; hydrogen bond
to main-chain amide, boldface X; hydrogen bond to main-chain carbonyl,
underlined X; disulfide bond, cedilla c ¸; positive f-torsion angle, italic X.
Alignment of the amino acid sequence of the extracellular
Homology Model of the 5-Hydroxytryptamine3Receptor 2339
Biophysical Journal 84(4) 2338–2344

Page 3

with the keto group of N128 (2.8 A˚and 2.5 A˚for models
4 and 5, respectively).
In model 6, the indole N atom of 5-HT is toward the top of
the cleft, and is similarly located in model 7, although here
5-HT has rotated ;1808. The primary amines are therefore at
quite distinct locations at either upper right of the cleft with
the N atom 3.2 A˚from the nearest atom of Y143 (model 6),
or to the lower left at a distance of 3.7 A˚from the nearest
atom of F226 (model 7). The hydroxyl groups, however, are
both located close to Y153: in model 6 there is the potential
to form a hydrogen bond of length 2.3 A˚, whereas in model 7
it is 3.3 A˚from the nearest carbon atom. The former is the
only potential hydrogen bond we identified in model 6, and
in model 7 there is also only one potential hydrogen bond,
between the indole N atom and the backbone carbonyl of
S182.
DISCUSSION
We have used information available from the x-ray crystal
structure of AChBP to model the extracellular domain of
the 5-HT3receptor. This region contains the binding sites for
5-HT, which lie at the subunit interfaces. Simulated docking
of 5-HT with our model of the binding site using the pro-
gram AutoDock revealed seven energetically favorable posi-
tions of 5-HT. Examination of the residues within 5 A˚of
5-HT showed that many (12 out of 17) were common to
all positions (see Table 1), and most of these residues, as
expected, fall in the binding loop regions that have been
proposed to constitute the binding site (Fig. 2 B). Our data
also identify residues that have the potential to form
hydrogen bonds with the agonist, and therefore may be
involved not only in its correct localization but also in
transducing agonist binding into channel opening. The
results show that homology modeling can be valuable in
locating residues involved in agonist binding, and in pro-
viding experimentally testable hypotheses to characterize
agonist-receptor interactions.
AChBP is homologous to the extracellular domain of the
Cys-loop family of LGIC and therefore, given that there are
as yet no atomic resolution structures of the latter proteins,
homology modeling with AChBP provides a route to iden-
tifying important features of the extracellular domain. It
should be remembered, however, that AChBP lacks many
features of the complete receptor and therefore the lim-
itations of such data must be borne in mind. For example,
combining structural information from this protein with the
highest available resolution images from the nACh receptor
has revealed that in the absence of ACh, the extracellular
domains of the a-(ligand-binding) subunits differ by rota-
tions of their inner pore-facing parts when compared to the
AChBP subunit structures (Unwin et al., 2002). As the struc-
ture of AChBP does, however, appear to provide a reason-
able fit with the nACh receptor data in the presence of ACh,
the use of the model we have derived to examine the docking
of 5-HT should yield a sufficiently accurate estimation of
the agonists’ location to provide useful experimentally test-
able hypotheses. Indeed, the residues that we identified as
\5 A˚ from 5-HT are all located in the binding loops
A–E that were originally defined from photo-affinity label-
ing and mutagenesis studies of the related nACh receptor.
These studies showed that residues contributing to the bind-
ing site were located in three regions of the a-subunit (loops
A–C) and three regions on the adjacent b- or g-subunit
(loops D–F), although the latter appeared to play a less sig-
nificant role. Consistent with these data, examination of
the putative ligand-binding pocket of the5-HT3receptorsug-
gests that residues from the loop B and C regions play the
major role, with fewer residues from loops A, D, and E being
involved. Loop F has not yet been fully defined in the
structure of AChBP, and thus we currently have no loop F
residues in our model.
Toidentify whichofthedockingpositionsismostlikelyto
represent the correct orientation of 5-HT in the binding site,
we used both experimental data and evidence from sequence
conservation.ThelatterisindicatedinTable1andshowsthat
76%oftheresiduesweidentifiedtobewithin5A˚of5-HTare
conserved in all agonist-binding 5-HT3receptor subunits
identified to date. The only region that contains residues that
are not conserved is loop C. As we would anticipate that
critical residues would be conserved, models 4 and 5 are
favored; here only two nonconserved residues are present in
the binding site, as compared to four in the other models.
Conversely, comparing those residues that are conserved but
are not present in our defined binding site region, the most
likely models are 4, 5, and 7 (one residue not present),
whereas in the other models two residues are not present.
More definitive evidence, however, can be obtained from
FIGURE 2
extracellular domain. (A) Model of 5-HT3receptor based
on AChBP showing the location of the putative ligand-
binding pocket. The white arrow points perpendicular to
the membrane in the plane of the ion channel away from
the membrane. (B) The five ligand-binding loops A–E for
5-HT3receptor as defined for the nACh receptor. Key:
loop A (yellow); loop B (blue); loop C (green); loop D
(red); loop E (gray).
Homology model of the 5-HT3 receptor
2340Reeves et al.
Biophysical Journal 84(4) 2338–2344

Page 4

Homology Model of the 5-Hydroxytryptamine3Receptor 2341
Biophysical Journal 84(4) 2338–2344

Page 5

experimental studies. The most useful of these are probably
by the Lester lab, who have shown that the 18 amine of 5-HT
forms a cation-p interaction with W183 in the binding site
(Beene et al., 2001). Mutagenesis studies have previously
shown that this tryptophan is a critical ligand-binding residue
(Spier and Lummis, 2000), and it is homologous to W149
in the nACh receptor, which also is involved in a cation-p
interaction with the agonist (Zhong et al., 1998). In other
LGIC this residue also appears important; aromatic residues
are found in the homologous position in GABAAand glycine
receptors, where again they have been shown to play a role in
ligand binding (Schmieden et al., 1993; Vandenberg et al.,
1992; Amin and Weiss, 1993). These data favor models 4
and 5 and indeed allow us to dismiss models 6 and 7, where
the aromatic center of tryptophan is not sufficiently close to
the 18 amine of 5-HT to form such an interaction (Gallivan
and Dougherty, 1999).
W183 is located in the loop B region. Our models suggest
that up to four other residues from this region are also in the
binding site: T181, S182, and L184 were observed in all
models, and T179 in models 4 and 5. T179 is conserved not
only in all agonist binding (A) subunits, but also in B and in
the recently identified but as yet uncharacterized C subunits
(Reeves and Lummis, 2002), but not in the nACh receptor
or in AChBP. This suggests it may play an important role,
further supporting models 4 and 5. Potential hydrogen bonds
with some of these loop B residues also provide a useful
experimentally testable point of distinction between these
two models. In model 4, the backbone carboxyl group of
S182 is an appropriate distance and orientation to form
a hydrogen bond with primary amine of 5-HT, whereas in
model 5 a hydrogen bond could form with the backbone
carboxyl group of W183 (Fig. 3).
In loop A, the only residue within 5 A˚of 5-HT is N128.
This residue has not been previously identified as a ligand-
binding residue, but interestingly the adjacent glutamate,
E129, and also F130 have been proposed from mutagenesis
studies to be involved in both agonist and antagonist binding
(Steward et al., 2000). It is possible that changing these
residues alters the location or orientation of N128. It is also
possible, however, that our alignment may not be accurate in
this region. Further mutational analysis will be required to
distinguish between these possibilities.
Experimental evidence has also shown that loop C resi-
dues play a critical role in forming the binding site. Simulta-
neous mutagenesis of I228, D229, and I230 significantly
changed the EC50for 5-HT, although the single mutations
have yet to be investigated (Hope et al., 1999). D229 is
one of two acidic residues in the loop C region, and there
is evidence that at least one of these is in close proxi-
mity to the ligand. The fluorescence profile of the 5-HT3re-
ceptor antagonist, GR-flu, is pH dependent, and when bound
suggests an acidic character for the binding site (Tairi et al.,
1998). Interestingly, D229, which is in the binding site in
models 1, 2, 3, 6, and 7, is not conserved in all binding sub-
units. This is in contrast to E236, which is present in models
4, 5, and 7; thus, sequence data support the latter models.
Y234 has also been shown in mutagenesis experiments
to play a role at the binding site (Price and Lummis, 2001),
perhaps in stabilizing the primary amine of 5-HT. In models
4 and 5 it is located at the opposite side of this amine group
to W183 and may form another cation-p interaction here.
Other mutagenesis studies of the loop C region also sug-
gest that a number of these residues are involved in ligand
binding; this region was found to be critical in controlling
the potency of both the 5-HT3receptor agonist meta-chloro-
phenylbiguanide (Mochizuki et al., 1999) and the antagonist
d-tubocurarine (Yan et al., 1999).
In loop D, two residues appear in all our models, W90
and R92, and there are experimental data from mutagenesis
studies to support their role in the binding site (Mochizuki
et al., 1999; Yan et al., 1999). These residues correspond to
ligand-binding residues W53 and Q55 in the AChBP, which,
as predicted by Yan et al. (1999) for the 5-HT3receptor, are
part of a b-sheet structure in our model.
Two residues from loop E are also apparent in all our
FIGURE 3
Amino acids shown in each case project within a 5 A˚radius of 5-HT. 5-HT is shown for clarity in a space-filling view, but these radii are not to scale. The
protein is displayed in the same 3D orientation in each panel. Potential hydrogen bonds are shown as dotted green lines (lengths given in text). All other atoms
are colored according to the CPK system (blue, nitrogen; red, oxygen; white, carbon).
Docking models for the binding of 5-HT to the 5-HT3receptor (1–7). Models 1–7, respectively, of 5-HT docked in the receptor binding site.
TABLE 1
site within 5 A˚of 5-HT
Residues of the putative 5-HT3receptor binding
5-HT3A
Model
a1 nAChR
ResidueLoopResidue1234567CON
A
B
N128
T179
T181
S182
W183
L184
F226
I228
D229
I230
Y234
E236
W90
R92
Y143
Q151
Y153
1
?
1
1
1
1
1
1
1
1
1
?
1
1
1
1
1
1
?
1
1
1
1
1
1
1
1
1
?
1
1
1
1
1
1
?
1
1
1
1
1
1
1
1
1
?
1
1
1
1
1
1
1
1
1
1
1
1
1
?
?
1
1
1
1
1
?
1
1
1
1
1
1
1
1
1
?
?
1
1
1
1
1
?
1
1
?
1
1
1
1
1
1
1
1
1
?
1
1
1
1
1
1
?
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
y
y
y
y
y
y
y
n
n
n
n*
y
y
y
y
y
y
Y93
K145
G147
T148
W149
T150
Y190
C192
C193
P194
Y198
D200
R55
K57
L79
T117
T119
C
D
E
CON, conservation of this residue in all known 5-HT3A(agonist-binding)
subunits to date. Bold indicates experimental evidence is available to sup-
port the role of this residue in the ligand-binding site. Equivalent residues
in a nACh receptor subunit are shown for comparison.
*, conserved as aromatic.
2342 Reeves et al.
Biophysical Journal 84(4) 2338–2344