Docking of 1,4-Benzodiazepines in the ?1/?2GABAAReceptor
D. Berezhnoy, T. T. Gibbs, and D. H. Farb
Laboratory of Molecular Neurobiology, Department of Pharmacology & Experimental Therapeutics, Boston University School of
Medicine, Boston, Massachusetts
Received January 8, 2009; accepted May 29, 2009
Positive allosteric modulation of the GABAAreceptor (GABAAR)
via the benzodiazepine recognition site is the mechanism
whereby diverse chemical classes of therapeutic agents act to
reduce anxiety, induce and maintain sleep, reduce seizures,
and induce conscious sedation. The binding of such therapeu-
tic agents to this allosteric modulatory site increases the affinity
of GABA for the agonist recognition site. A major unanswered
question, however, relates to how positive allosteric modula-
tors dock in the 1,4-benzodiazepine (BZD) recognition site. In
the present study, the X-ray structure of an acetylcholine bind-
ing protein from the snail Lymnea stagnalis and the results from
site-directed affinity-labeling studies were used as the basis for
modeling of the BZD binding pocket at the ?1/?2subunit inter-
face. A tethered BZD was introduced into the binding pocket,
and molecular simulations were carried out to yield a set of
candidate orientations of the BZD ligand in the binding pocket.
Candidate orientations were refined based on known structure-
activity and stereospecificity characteristics of BZDs and the
impact of the ?1H101R mutation. Results favor a model in
which the BZD molecule is oriented such that the C5-phenyl
substituent extends approximately parallel to the plane of the
membrane rather than parallel to the ion channel. Application of
this computational modeling strategy, which integrates site-
directed affinity labeling with structure-activity knowledge to
create a molecular model of the docking of active ligands in the
binding pocket, may provide a basis for the design of more
selective GABAAR modulators with enhanced therapeutic po-
GABAAreceptors (GABAARs) are pentameric transmem-
brane proteins that belong to the cysteine-loop superfamily of
ligand-gated ion channels and function as GABA-gated Cl?-
selective channels, which mediate most fast inhibitory neu-
rotransmission in the central nervous system (Berezhnoy et
al., 2007). There are 20 related GABAAR subunits in mam-
mals, designated ?1–6, ?1–4, ?1–3, ?, ?, ?, ?, and ?1–3, that can
assemble in multiple combinations to produce different
GABAAR subtypes (Barnard et al., 1998; Bonnert et al.,
1999). The regional and cellular distribution of different
GABAAR subunits is distinct but overlapping, and individual
receptor subtypes exhibit distinct subcellular localizations
(Berezhnoy et al., 2007). Most GABAARs in the adult mam-
malian central nervous system are composed of ?, ?, and ?
subunits, with ?1?2/3?2being the most abundant subtype
(Sieghart and Sperk, 2002).
GABAARs are activated by binding of agonist to recogni-
tion sites located at ?(?)/?(?) subunit interfaces (Berezhnoy
et al., 2007). Agonist-induced receptor activation can be mod-
ulated through allosteric binding sites located at the ?1(?)/
?2(?) subunit interface (the BZD recognition site) (Choh et
al., 1977; Chan and Farb, 1985). Residues implicated in the
formation of the GABA and BZD binding sites are located at
equivalent positions within six loops in the extracellular
N-termini of the ?, ?, and ? subunits (Supplemental Fig. 1)
(Berezhnoy et al., 2007).
Previous attempts have been made to superimpose the
structures of allosteric modulators to construct a pharma-
cophore model for the BZD recognition site (Borea et al.,
1987; Villar et al., 1989; Schove et al., 1994; Zhang et al.,
1995; Huang et al., 1998, 1999; He et al., 2000; Marder et al.,
2001; Verli et al., 2002). However, such models are difficult to
relate to receptor structure. Sigel et al. (1998) determined
affinities for a series of imidazo- and 5-phenyl-1,4-benzodiaz-
This work was supported by the National Institutes of Health National
Institute of Mental Health [Grant R01-MH049469].
Article, publication date, and citation information can be found at
The online version of this article (available at http://molpharm.
aspetjournals.org) contains supplemental material.
ABBREVIATIONS: GABAAR, GABAAreceptor; BZD, benzodiazepine; FNZ, flunitrazepam; DZ, diazepam; RMSD, root-mean-square deviation;
DZ-NCS, diazepam carrying a thiol-reactive –N?C?S group; AChBP, acetylcholine binding protein; Ro 15-4513, 4H-imidazo(1,5-
a)(1,4)benzodiazepine-3-carboxylic acid, 8-azido-5,6-dihydro-5-methyl-6-oxo-, ethyl ester; Ro 15-1788, 8-fluoro-5,6-dihydro-5-methyl-6-oxo-4H-
imidazo[1,5-a][1,4]benzodiazepine-3-carboxylic acid ethyl ester.
Copyright © 2009 The American Society for Pharmacology and Experimental Therapeutics
Mol Pharmacol 76:440–450, 2009
Vol. 76, No. 2
Printed in U.S.A.
epines to wild-type and mutant receptors to delineate the
orientation of these ligands in the recognition site. An extra
hydroxyl group of tyrosine introduced by the ?2F77Y muta-
tion interferes with para-substitutions of the C5-phenyl ring,
suggesting that the phenyl ring is adjacent to ?2Phe77 in the
binding pocket (Sigel et al., 1998). Kucken and colleagues
(2003) used a series of three substituted imidazobenzodiaz-
epines in combination with amino acid mutations of varying
volume at ?2Ala79 to infer the position of compounds similar
to Ro 15-1788 and Ro 15-4513 (Kucken et al., 2003). Photoaf-
finity labeling using [3H]flunitrazepam identified the major
site that incorporates radioactivity as His101 of loop A (Mc-
Kernan et al., 1995; Davies et al., 1996; Duncalfe et al., 1996)
and a second less abundant site as Pro96 (Smith and Olsen,
2000). Likewise, photoaffinity labeling using the imidazoben-
zodiazepine [3H]Ro 15-4513 identified residue Tyr209 of loop
C of the ?1subunit as proximal to the benzodiazepine binding
site (Sawyer et al., 2002). However, the docking position with
respect to specific contact residues cannot be deduced be-
cause of uncertainty of photoaffinity labeling in an environ-
ment containing multiple aromatic residues (Kotzyba-Hibert
et al., 1995). Using a C7-modified diazepam (DZ) carrying a
thiol-reactive –N?C?S group (DZ-NCS), ?1H101C was con-
firmed to be in or near the binding pocket. At the functional
level, the reacted receptor becomes irreversibly locked in a
positively modulated state (Berezhnoy et al., 2004, 2005; Tan
et al., 2007a,b,c).
To further refine the positioning of ligands in the BZD
binding pocket, we use a homology model based on the crystal
structure of AChBP (Brejc et al., 2001). The initial ligand
position was obtained by modeling DZ-NCS covalently linked
to ?1H101C (Fig. 1). This yields two candidate orientations,
one with the C5-phenyl group oriented approximately paral-
lel to the cell membrane, and the other with the C5-phenyl
oriented parallel to the ion channel. We evaluated the con-
sistency of these orientations with respect to four criteria: 1)
the capacity to accommodate a tethered DZ analog ([poly(Me-
BZD)] that was used in the early affinity column purification
of GABAAreceptors (Sigel et al., 1983; Sigel and Barnard,
1984); 2) the effect of the ?1H101R mutation, which abolishes
BZD binding (Wieland et al., 1992; Wieland and Lu ¨ddens,
1994; Benson et al., 1998; Dunn et al., 1999); 3) the two
enantiomers of 3-methyl-substituted FNZ, Ro 11-6896 and
Ro 11-6893 (Niehoff et al., 1982; De Blas et al., 1985); and 4)
the binding affinities of a set of active and inactive BZD
derivatives (Klopman and Contreras, 1985; Zhang et al.,
1994) (Fig. 2). The results show that the docking orientation
with the C5-phenyl parallel to the membrane satisfies all of
these criteria, whereas the orientation with the phenyl par-
allel to the ion channel does not, indicating that the former
orientation in the binding pocket is favored.
Materials and Methods
Homology Modeling. A homology model of the extracellular
domain of the rat GABAAR ?1and ?2subunits was constructed based
on the X-ray structure of the AChBP complexed with nicotine (Pro-
tein Data Bank entry 1uw6) (Celie et al., 2004). The mature protein
sequences of the rat ?1and ?2subunits (accession numbers: ?1,
P62813; ?2, P18508) were aligned with sequences of two adjacent
AChBP subunits (A and B, respectively) using ClustalW (Thompson
et al., 1994) (Supplemental Fig. 1). Because the GABAAR subunits
share only ?18% identity with AChBP, the reliability of the align-
ment was checked by creating a multiple alignment with all ?1–6and
?1–3subunits and ?, ?, ?, and ? subunits of the nicotinic acetylcholine
receptor, taking the secondary structure predictions into account.
Using absolutely conserved residues to “anchor” regions of low ho-
mology, we edited the sequence alignment to align gaps with loops in
the AChBP structure.
Three regions of the GABAAreceptor subunits did not align well:
the N-terminal ?-helix, the region between ?-sheet domains ?4 and
?6, and the region between ?-sheet domains ?8 and ?9 (Supplemen-
tal Fig. 1). In contrast to the alignment of Brejc et al. (2001), we have
aligned the insertion between ?-sheet domains ?4 and ?6 of the
GABAAreceptor with the ?4–?5 extracellular loop of AChBP. This
results in a better alignment with the GABAAR subunits, because
there is more room for the inserted residues compared with the Brejc
et al. alignment, in which the ?5–?6 ?-sheet domain is partially
buried. After alignment, each subunit was modeled independently
using the Build Homology Model module of Discover Studio (Accel-
rys, San Diego, CA). Loops to fill in the gaps between the GABAARs
sequences and the template sequence were built and refined using
the autorotomer feature of the same module. The backbone atoms of
each residue were tethered to the coordinates of corresponding res-
idues in the AChBP template with a force constant of 5 kcal ? Å?1.
This protocol generated 10 receptor dimer models, which were then
subjected to energy minimization to eliminate obvious problems such
as steric clashes, and the model with the lowest occurrence of unfa-
vorable contacts was chosen.
In the resulting dimer model, ?98% of the residues have a back-
bone geometry falling in favorable regions of the Ramachandran
plot. Superimposing the ligand binding domain of the homology
model onto the AChBP yields an average root-mean-square deviation
of 0.7 Å for ?-carbons. When the consensus sites for N-glycosylation
are mapped onto the model, all are found on the solvent-accessible
surface. Residues previously identified as forming the GABA and
BZD binding sites are also on the water-accessible surface, with the
exception of ?2Met57. The available evidence indicates that our
homology model is based on the structure of the AChBP in a confor-
mational state that binds nicotine with high affinity and is thus
presumed to resemble a conformation of the nAChR that binds ACh
with high affinity (i.e., either an open or desensitized state) (Brejc et
al., 2001; Unwin et al., 2002; Unwin, 2003; Celie et al., 2004). A
number of the residues that this alignment predicts to line the BZD
binding pocket, to our knowledge, have not been investigated exper-
imentally. In particular, Lys155, Thr213, and His215 on the ?1
subunit and Asn60 on the ?2subunit are predicted to face the
interior of the binding pocket and are located in close proximity to
residues shown to affect potency of efficacy of BZD site ligands.
Fig. 1. Mechanism of covalent modification by
DZ-NCS: nucleophilic attack of ?1H101C on DZ-
NCS results in an ?1substituent bearing DZ
covalently linked to the drug recognition site. In
the resulting product, the angle formed by –C–
S–C– bond is 120° and allows some degree of
rotation around the –C–S–, –S–C–, and –C–N–
Docking of 1,4-Benzodiazepines with ?1?2?2GABAARs
Modeling of DZ-NCS Tethered to ?1?2BZD Binding Pocket.
After optimization of the receptor model, the ?1H101C mutation was
introduced, covalently linked to DZ-NCS corresponding to the cova-
lent reaction of cysteine with the -NCS reactive group. 1,4-Benzodi-
azepines such as DZ exist in solution as an equimolar mixture of two
chiral conformers due to rapid inversion of the nonplanar seven-
membered ring (Blount et al., 1983). Both conformers of DZ-NCS
were therefore used for docking studies, rather than the single con-
formation of DZ found in the X-ray structure (Camerman and Cam-
erman, 1972) (Fig. 3). During simulation, constraints were applied to
the receptor model such that only the ligand and the residues facing
the interior of the binding pocket were allowed to move: on the ?1
subunit: Phe99, His101, Asn102, Lys155, Tyr159, Thr162, Gly200,
Val202, Ser204, Ser205 Thr206, Val211, Thr213, and His215; on the
?2subunit: Asp56, Tyr58, Asn60, Asp75, Phe77, Ala79, Thr81,
Thr126, Met130, Leu140, Thr142, Arg144, Lys184, Ser186, Val188,
Val190, Thr193, Arg193, and Trp196.
Conformations were searched by rotation of the -CS-NH- bond in
30° increments, followed by a standard dynamics cascade procedure
that included minimization steps, simulated annealing (600 to 50 K),
equilibration, and production steps at 300 K. This resulted in a pool
of DZ-NCS conformations. Ligand orientations in which the C5-
phenyl extended out of the recognition site were discarded, because
the C5-phenyl is essential for high-affinity binding of 1,4-BZDs (Sigel
et al., 1998), and ligands that do not have this moiety are inactive
(i.e., Ro 5-4654; Fig. 2A), indicating that the phenyl group is likely to
be an interaction center.
This procedure yielded two favorable orientations, designated “h”
(horizontal) and “v” (vertical), for each of the two conformers of
DZ-NCS, for a total of four candidate models of bound DZ-NCS,
designated DZ-NCS1h, DZ-NCS1v, DZ-NCS2h, and DZ-NCS2v (Fig.
5). Corresponding models for DZ and FNZ bound to the receptor were
obtained as follows for each of the candidate models: the bond be-
tween the DZ-NCS molecule and the receptor was eliminated; the
native histidine-101 residue of the receptor was restored; and DZ-
NCS was replaced by DZ, FNZ, or other BZD derivatives (Fig. 2).
Subsequently, each model was subjected to the standard dynamics
cascade protocol as described above. Interaction energies were cal-
culated using Calculate Interaction Energy: ligand and receptor
were defined as groups of atoms, dielectric constant was set to 1,
nonbound list radius was 14, and nonbound higher and lower cutoff
distances were set to 12 and 10, respectively.
Automated Ligand Docking. Docking of DZ and FNZ was car-
ried out using the CDOCKER algorithm (Wu et al., 2003) in the
Fig. 2. Structures of BZD derivatives used in docking studies. A, BZD modulators diazepam, flunitrazepam, and flunitrazepam derivatives Ro 11-6896
and Ro 11-6893 carrying an optically active methyl group at the 3-position. B, inactive compounds Ro 05-4864 and Ro 05-4654 do not bind. C, structure
of BZD ligand used by Sigel et al. (1983) to isolate the GABAAR. D and E, structures of active (D) and inactive (E) BZD analogs used to test model
of ligand orientation. High-affinity indicates ligands with IC50?100 nM; low affinity indicates ligands with 100 nM ? IC50? 1 ?M. Compounds
identified as inactive have IC50? 1 ?M (Ro 05-4864, Ro 05-4854, analog 3) or lack activity in behavioral assays (analog 4).
Fig. 3. Diazepam exists in solution as
equimolar mixture of conformers, de-
noted as DZ1 and DZ2, defined as
shown. Simulations identified two
candidate orientations for DZ in the
binding pocket. Each conformer can
potentially bind in either an h- or v-
orientation. These are denoted DZ1h
and DZ1v for conformer DZ1 and
DZ2h and DZ2v for conformer DZ2.
Berezhnoy et al.
Discovery Studio environment. CDOCKER is a grid-based molecular
docking method that uses the CHARMm force field. The receptor is
held rigid while the ligand is allowed to flex during the refinement
process. The binding site cavity for automated docking was assigned
via the protein-ligand interaction menu with a sphere of 8 Å. A set of
20 random ligand conformers was generated from the initial ligand
structure through high-temperature molecular dynamics followed by
random rotations and then refined by grid-based (GRID I)-simulated
annealing and energy minimization. The simulated annealing pro-
cedure consisted of 1000 steps of variable temperature molecular
dynamics. In each cycle, the temperature was scaled from 600 to 50
K over an interval of 10 ps followed by Smart Minimizer energy
minimization to 0.1 kcal ? mol?1? Å?1. The 20 most energy-favorable
ligand conformers were selected for further analysis. Both FNZ and
DZ converged on a set of similar conformations for both manual and
automated docking procedures, consistent with a restricted ste-
reospecific binding site.
Modeling of the ?1?2BZD Binding Pocket. One of the
main challenges of homology modeling is to identify the cor-
rect sequence alignment. The ligand-binding domains of the
GABAAR subunits share only ?18% amino acid identity with
AChBP, which is marginal for effective alignment and ho-
mology modeling. The validity of the GABAAR model is sup-
ported by its consistency with available biochemical data on
the location of critical residues (i.e., glycosylation sites, res-
idues forming GABA, and BZD binding sites). Residues re-
siding at the ?1subunit that have been reported to contribute
to the BZD binding pocket (His101, Tyr159, Thr162, Gly200,
Ser204, Ser205, Thr206, Tyr209, and Val211) are all water-
exposed, as were all such residues on the ?2subunit (Tyr58,
Asp75, Phe77, Ala79, Thr81, Met130, Leu140, Thr142,
Arg144, Lys184, Ser186, Val188, Val190, Thr193, Arg194,
and Trp196), with the single exception of ?2Met57 (Fig. 4).
Modeling of DZ-NCS Linked to the BZD Binding Site.
To determine how BZDs fit into the binding site, the assump-
tion was made that all active BZD-like ligands orient them-
selves similarly in the binding pocket. An important con-
straint is provided by the observation that DZ-NCS retains
modulatory activity when covalently linked to a cysteine
introduced by mutagenesis in place of histidine at position
?1101, a locus that has been identified by mutational analy-
sis as critical for BZD binding. We simulated the covalent
linkage of a DZ-NCS molecule (Fig. 2) in two alternative
conformations to the ?1H101C mutated receptor, because it
has been shown that 1,4-benzodiazepines in solution exist as
mixture of two conformers that have an inversion barrier of
?12 kcal/mol (Fig. 3) (Blount et al., 1983). Modeling indicates
that each conformer can potentially assume two favorable
orientations when bound to ?1H101C: a horizontal (h) orien-
tation in which the phenyl group is approximately parallel to
the plane of the plasma membrane, and a vertical (v) orien-
tation in which the phenyl group extends toward the mem-
brane, approximately parallel to the axis of the ion channel
In the h orientation, the benzodiazepine ring lies in the
Fig. 4. Structure of BZD binding
pocket. The BZD binding site of ???
GABAARs is formed at the ?(?)/?(?)
subunit interface. A, mapping of ac-
cessible volume of the BZD binding
pocket. B, front view of the binding
pocket. Residues on the ? subunit fac-
ing the interior of the binding site are
labeled in black, and those on the ?
subunit are labeled in blue. C and D,
views of the ? and ? subunits from
inside the BZD binding pocket. Resi-
dues that are believed to affect prop-
erties of BZD ligands via direct con-
effects are identified.
Docking of 1,4-Benzodiazepines with ?1?2?2GABAARs
same plane as ?1Phe99 and ?2Phe77, the C5-phenyl group is
directed toward ?1Tyr159, and the carboxyl groups are di-
rected toward ?1Ser204 and ?2Arg194 (Fig. 5, A and B). The
main difference between conformers 1 and 2 is the orienta-
tion of the N1 methyl group: in conformer 1 (DZ-NCS-1h), it
is directed toward ?2Thr193, whereas in conformer 2 (DZ-
NCS-2h), the N1 methyl group is directed out of the binding
pocket (Figs. 5C and 8D).
In the v-orientation, the C5-phenyl group is oriented par-
allel to the intersubunit interface and lies in close proximity
of ?2Trp196 (Fig. 5, E and G). In conformer 1 (DZ-NCS-1v),
N1 methyl group is directed out of the binding pocket (Fig.
5F), whereas in conformer 2 (DZ-NCS-2h), it is directed to-
ward the interior of the binding pocket (Fig. 5H).
Because irreversible binding of DZ-NCS results in persis-
tent GABAAreceptor potentiation, it is likely that the orien-
tation of covalently bound DZ-NCS corresponds to the orien-
tation of other active benzodiazepine site ligands in the
binding pocket. These four orientations were thus used as a
basis for modeling the binding of DZ and FNZ. All models
were subsequently subjected to energy minimization.
Modeling of DZ and FNZ Binding. The classic 1,4-BZDs
DZ and FNZ were then introduced the same position. Energy
minimization runs resulted in orientations that were close to
the original position. The only minor difference between the
two was caused by the presence of a fluorine atom, resulting
in a slight rotation of the C5 phenyl of FNZ molecule. It is
noteworthy that this rotation brings the fluorine atom closer
to hydroxyl group of ?1Tyr209. For both DZ (Fig. 6) and FNZ
(Fig. 7), the C7-substituent of both ligands is located in close
proximity to ?1His101 and ?1Lys155, and this is especially
pronounced with FNZ, which has a strongly electronegative
nitro group that can participate in the hydrogen bonding
with both ?1His101 and ?1Lys155. The carbonyl group of
both ligands faces the ?1Ser204-?2Thr193-?2Arg194 triad,
where it can form hydrogen bonds and coordinate two do-
mains: the tip of loop C, and most of loop F. As with DZ-NCS,
DZ and FNZ can exist in two conformers, which may be
positioned in either the h- or v-orientation. The orientation of
the C5-phenyl is a major difference between h- and v-orien-
tations for both DZ and FNZ. With ligand in the h-orienta-
tion, the C5-phenyl is located in the same plane as ?1Phe99
and ?2Phe77. These residues together form a “floor” to the
binding pocket. The “ceiling” of the binding pocket is formed
by ?1Thr206 and ?1Tyr209 (Figs. 6 and 7).
Replacement of ?1His101 with arginine abolishes the bind-
ing of classic 1,4-benzodiazepines, so we examined the impact
of this mutation on the interaction of DZ and FNZ with the
binding pocket. In the h-orientation, severe steric interfer-
ence is evident for both conformers of DZ and FNZ between
the arginine residue and the aromatic moiety adjacent to the
benzodiazepine ring and the C5-phenyl group, resulting in
considerably unfavorable energies of interaction (Table 1). In
the v-orientation, binding of DZ and FNZ is somewhat desta-
bilized but remains energetically favorable, with the benzo-
diazepine ring and nitro group fitting between the ?1H101R
and ?1Lys155 residues (Fig. 6 and 7). The profound impact of
the ?1H101R mutation on binding is thus more consistent
with DZ and FNZ being bound in the h-orientation.
The ?2F77Y mutation has been shown to decrease binding
affinities of both DZ and FNZ by ?226- and ?170-fold in
radioligand binding experiments but does not affect DZ po-
tency in electrophysiological experiments (Buhr et al., 1997).
As modeled, the ?2F77Y residue faces the BZD ligands in the
Fig. 5. Positioning of covalently bound DZ-NCS. Four energetically favorable orientations of DZ-NCS resulted from a search for lowest energy
conformations via systematic rotation of the -CS-NH- bond. DZ-NCS’s seven-member benzodiazepine ring is yellow and C5-phenyl is painted blue. For
each conformer, low-energy binding orientations include a horizontal orientation in which the C5-phenyl group of the ligand is approximately parallel
to the plasma membrane (DZ-NCS1h, A and B; DZ-NCS2h, C and D) and a vertical orientation in which the C5-phenyl extends toward the plasma
membrane (DZ-NCS1v, E and F; DZ-NCS2v, G and H).
Berezhnoy et al.
binding pocket, but we did not detect any unfavorable inter-
action between this residue and DZ or FNZ in either the h- or
v-orientation (Table 1).
To test the structural model of the BZD recognition site,
the impact of replacing FNZ with other BZD derivatives was
examined. The 1,4-benzodiazepine used for initial isolation of
GABAAR [poly(Me-BZD), Fig. 2C] was attached to the aga-
rose column via a polymethyl linker attached at the N1
position. High-affinity binding of the ligand was retained
despite the presence of the linker, arguing that the linker
must be able to extend out of the binding pocket with mini-
mal perturbation of receptor structure. Docking studies sug-
gest that the linker could exit the binding pocket from either
the top or the side of the receptor (Fig. 8), but if the linker
exits from the top, its length is probably insufficient to avoid
steric interference between the receptor and the agarose
resin bead (Fig. 8, C and D). In contrast, the length of the
linker is adequate to avoid steric interference if the ligand is
bound with the N1 substituent facing toward the side of the
receptor, such that the linker can exit from the side of the
receptor as depicted in Fig. 8, A and B. This requires the
bound ligand to be in the h-orientation if it is in conformer 2
and in the v-orientation if it is in conformer 1.
To further evaluate the model and to assess whether active
BZDs are bound in the h-orientation or the v-orientation,
FNZ was replaced with a number of active and inactive BZD
derivatives. Ro 5-4864 (Fig. 2B), which does not bind to
?1?2?2GABAAreceptors (Sigel et al., 1998), bears a para-
substituent on the C5 phenyl that points directly toward
?1Tyr159 when in the h-orientation. This is likely to cause
steric hindrance, which could explain its lack of activity; this
steric clash is not present in the v-orientation. Ro 5-4654,
which lacks the C5 phenyl group, is inactive (Sigel et al.,
1998), arguing that interactions with this group are required
for high-affinity binding of classic 1,4-benzodiazepines.
The orientation of benzodiazepines in the binding pocket
was further evaluated by replacing FNZ with the active (Fig.
2D) and inactive (Fig. 2E) BZD derivatives previously stud-
ied experimentally by Zhang et al. (1994) (analogs 1–3) and
by Klopman and Contreras (1985) (analog 4). Interaction
energies for these analogs in each orientation were calculated
(Table 2). Analogs 1 and 2, which are active in displacing
[3H]FNZ binding from rat brain membranes (Zhang et al.,
1994), were well accommodated in the h-orientation in either
conformer 1 (Fig. 9A) or conformer 2 (Fig. 9B). In contrast,
there was steric hindrance for both conformers in the v-
orientation (Fig. 9, C and D). Analog 3, which has little or no
affinity for the BZD recognition site (Zhang et al., 1994),
yielded unfavorable interaction energies in all orientations
because of steric clashes with the residues and backbone of
the ?1subunit, as did analog 4, which has been reported to
have very low anticonvulsant potency in vivo (Klopman and
Contreras, 1985) (Fig. 10).
To evaluate whether the model reproduces the stereospec-
ificity of BZD binding, two optically active FNZ derivatives
were docked into the binding pocket. The dextrorotary Ro
11-6896, with the C3-methyl pointing up, exhibits more than
100-fold greater affinity in binding studies than its levorota-
tory enantiomer Ro 11-6893 (Fig. 2), which has the methyl
group pointing down (Niehoff et al., 1982). In the h-orienta-
Fig. 6. Orientations of DZ were modeled after orientations of the DZ-NCS. The methyl substituent at the N1 atom of DZ is much smaller in size than
the polymethyl linker used for affinity purification (Fig. 9) and is not able to restrict the ability of the benzodiazepine ring to undergo inversions,
making it more difficult to deduce which of the two conformers is likely to be prevalent in the binding pocket. DZ’s seven-member benzodiazepine ring
is yellow, and the C5-phenyl is blue. A and B, DZ1h; C and D, DZ2h; E and F, DZ1v; G and H, DZ2v orientation. In all of these models, the C7-chloro
group is directed toward the ?1His101 and ?1Lys155 residues, and the C2 carbonyl group is located in close proximity to ?1Ser204, ?1Ser205, ?2Thr193,
and ?2Arg194, where it is able to make hydrogen bonds. Pairs of images depict the BZD binding pocket viewed from outside of the receptor (A, C, E,
and G) and from within the binding pocket looking toward the ? subunit (B, D, F, and H).
Docking of 1,4-Benzodiazepines with ?1?2?2GABAARs
tion, interaction energies for these two compounds repro-
duced the observed stereospecificity of binding, with both
conformers of Ro 11-6896 exhibiting more favorable binding
energies than Ro 11-6893. For the v-orientation, results were
mixed, with Ro 11-6896 being favored over Ro 11-6893 in
conformer 1 but Ro 11-6893 being favored in conformer 2.
The h-orientation thus best reproduces the observed ste-
reospecific binding of these ligands (Table 3).
Automated Docking of DZ and FNZ. To assess the
validity of results obtained using manual docking, automated
docking was carried out using the CDocker algorithm (Wu et
al., 2003), and the 20 most energetically favorable conformers
were selected for further analysis. This algorithm yielded a
number of models resembling the v- and h-orientations ob-
tained by manual docking, as well as orientations that were
distinct. This result can be explained by the way the docking
algorithm operates: random conformations are generated
and seeded within the binding pocket, and subsequent mo-
lecular dynamics and energy minimization finds a local en-
ergy minima without regard for known structure-function
data. Overall, interaction energies from automated docking
(Table 3) were somewhat less favorable than for manual
docking, most likely because the random starting position
of the ligand resulted in a less efficient optimization than in
the manual search, in which the starting position for optimi-
zation was based on information derived from DZ-NCS
Automated docking yielded 3 orientations for DZ, desig-
nated DZ dock1–3, each constituting approximately one third
of the total pool (Supplemental Fig. 2; interaction energies in
Table 3). Two of these orientations superimpose well with
manual docking orientations: DZ dock1 with DZ2h (with
RMSD of 1.8 Å) and DZ dock2 with DZ2v (RMSD of 1.36 Å);
DZ dock3, although favorable in energetic terms, was differ-
ent from the orientations obtained by manual docking. In
this orientation, the C5-phenyl group points outside the bind-
ing pocket; however, it is known that the presence of a chlo-
rine atom at the para-position of this group (Ro 5-4864)
eliminates activity. This is most likely due to steric hin-
drance, because chlorine in the ortho-position is tolerated
(Sigel et al., 1998), indicating that the phenyl group probably
does not point out of the pocket.
For FNZ, automated docking resulted in five orientations,
FNZ dock1–5, respectively constituting 30, 20, 25, 5, and 20%
of the total model pool. FNZ dock1 and dock2 are very similar
to FNZ2v and can be superimposed with RMSDs of 0.81 and
3.04 Å (Supplemental Fig. 3, A and B), whereas FNZ dock4
and dock5 closely resemble the orientation of FNZ2h (RMSDs
of 1.06 and 1.46Å, respectively) (Supplemental Fig. 3, G–J).
FNZ dock3 (Supplemental Fig. 3 E and F) somewhat resem-
bles FNZ2h but differs from it by a larger RMSD of 4.2 Å that
is caused by “sinking” of the nitro-phenyl part of DZ in the
Modeling the molecular interactions of ligands with recep-
tors provides a means of refining structure-activity relation-
ships to aid drug discovery. Crystallization of glutamate re-
ceptor ligand-binding domains has helped to visualize
Fig. 7. Orientations of FNZ in the binding pocket were modeled after orientations of DZ-NCS. FNZ shows potency and efficacy in binding and
electrophysiological assays very similar to DZ, but it differs in that it contains a nitro group, which is a strong hydrogen bond acceptor; additionally,
it has a fluoro group in the ortho-position of the C5 phenyl. FNZ’s seven-member benzodiazepine ring is yellow, and the C5-phenyl is blue. Binding
models were generated corresponding to the h-orientation (A and B, FNZ1h; C and D, FNZ2h), and v-orientation (E and F, FNZ1v; G and H, FNZ2v).
All of these models share two common features: the C7-nitro group is directed toward the ?1His101 and ?1Lys155 residues, and the C2 carbonyl group
is located in close proximity to ?1Ser204, ?2Thr193, and ?2Arg194, where it is able to make hydrogen bonds. Pairs of images depict the BZD binding
pocket viewed from outside of the receptor (A, C, E, and G) and from within the binding pocket looking toward the ? subunit (B, D, F, and H).
Berezhnoy et al.
binding pockets for agonists and allosteric modulators (Jin et
al., 2005). Currently, the only structural data available for
GABAARs are enhanced electron-microscopy images of nico-
tinic acetylcholine receptors and X-ray structures of acetyl-
choline binding proteins from Aplysia californica (Ulens et
al., 2006) and Lymnea stagnalis (Brejc et al., 2001; Celie et
al., 2004), which share ?18% sequence identity with the
GABAAR extracellular domain.
The 1,4-benzodiazepine FNZ photoaffinity labels residue
?1His101 (McKernan et al., 1995; Duncalfe et al., 1996;
Smith and Olsen, 2000), indicating that the FNZ nitro-group
is located near this residue, but the lack of detailed informa-
tion about the structure of the reaction product and the
functional consequences of modification precludes precise po-
sitioning of FNZ in the binding pocket. Photoaffinity binding
of FNZ blocks potentiation by chlordiazepoxide, but because
only ?25% of receptors are irreversibly bound, it was not
possible to determine whether FNZ photoaffinity binding
results in persistent potentiation (Gibbs et al., 1985).
Exposure of ?1H101C receptors to DZ-NCS results in irre-
versible reduction of [3H]Ro 15-1788 binding, indicating co-
valent binding of DZ-NCS within the binding pocket (Berezh-
noy et al., 2004). A caveat is that affinity-labeling could
“capture” a minor orientation that does not contribute appre-
ciably to the action of reversibly bound BZDs. DZ-NCS also
modifies ?1N102C and ?2A79C, albeit with lower efficiency
(Tan et al., 2007a), and an NCS analog of Ro 15-4513, which
lacks the pendant phenyl and may have greater freedom to
orient in the binding pocket, reacts with ?1residues 101, 157,
202, and 211. Confidence that DZ-NCS covalently linked to
?1 residue 101 occupies the binding pocket similarly to re-
versibly bound DZ is increased because ?1His101 is a known
contact residue that is critical for pharmacological activity of
BZDs (McKernan et al., 1995; Davies et al., 1996; Duncalfe et
al., 1996; Smith and Olsen, 2000) and because covalent link-
age of DZ-NCS results in irreversible potentiation compara-
ble with that produced by DZ (Berezhnoy et al., 2004). We
therefore used the position of DZ-NCS within the binding site
as a basis for modeling how BZDs occupy the binding pocket.
The structure of AChBP complexed with nicotine (Celie et
al., 2004), which probably reflects a high-affinity configura-
tion of the binding pocket similar to that associated with the
open or desensitized receptor (Brejc et al., 2001; Unwin et al.,
2002; Celie et al., 2004), was chosen as a basis for homology
modeling of the GABAAR based on the hypothesis that con-
formational changes associated with binding of allosteric
modulators to the BZD recognition site resemble those that
accompany binding of nicotine to AChBP. The structural
similarity of the BZD recognition site to the GABA binding
site suggests that positive modulation by BZDs probably
involves conformational changes similar to the activation by
GABA, resulting in downstream conformational changes that
stabilize the active state(s) of the receptor (Downing et al.,
2005). The hypothesis that BZDs interact with their recogni-
tion site in an agonist-like manner is supported by the ob-
servation that DZ, FNZ, and zolpidem directly activate
GABAAreceptors containing the ?1L263S (Downing et al.,
2005; Ru ¨sch and Forman, 2005) or ?2L245S mutations (Bi-
anchi and Macdonald, 2001) in the absence of GABA.
The plausibility of this model is supported by the observa-
tion that, with one exception, all glycosylation sites and all
residues implicated in GABA and BZD binding are exposed to
water. Only one residue reported as important for FNZ bind-
ing, ?2Met57, is buried; however, the neighboring residue,
?2Tyr58, which also has been implicated in maintaining
high-affinity binding of FNZ, is exposed, suggesting that
effects of mutating ?2Met57 may be allosteric (Kucken et al.,
Modeling of the binding of DZ and FNZ yielded results
similar to DZ-NCS, in which each conformer could be bound
in either the h- or v-orientation. Introduction of the ?1H101R
mutation, which results in 500- to 800-fold reduction in af-
finity of classic benzodiazepines (Wieland et al., 1992; Wie-
land and Lu ¨ddens, 1994; Benson et al., 1998; Dunn et al.,
1999), resulted in steric clashes of arginine residues with
both conformers of DZ and FNZ in the h-orientation. In
contrast, this mutation was accommodated by both DZ and
FNZ in the v-orientation. The h-orientation is thus more
consistent with the large impact of this mutation on DZ and
FNZ binding affinity.
In summary, the impact of the ?1H101R mutation, the lack
of activity of Ro 5-4864, the activity of analogs 1 and 2, and
the higher affinity of Ro 11-6896 compared with Ro 11-6893
are consistent with the h-orientation but not the v-orienta-
Docking energies of DZ (DZ1h, DZ1v, DZ2h, and DZ2v) and FNZ
(FNZ1h, FNZ1v, FNZ2h, FNZ2v) in ?1H101R and ?2F77Y mutant
Potential energies were calculated using the Calculate Interaction Energy protocol
as described under Materials and Methods.
Fig. 8. Positioning of BZD ligand with
polymethyl linker in binding pocket. A
tethered BZD ligand was used in the af-
finity column for initial isolation of the
and size of this ligand suggests that the
polymethyl linker can exit the binding
pocket either from the side of the binding
pocket (A and B) or from the top (C and
D). The latter is less likely, because the
length of the linker (highlighted in red in
C and D) attached to the affinity column
would not be expected to permit the li-
gand to reach the binding pocket.
Docking of 1,4-Benzodiazepines with ?1?2?2GABAARs
tion of 1,4-BZDs in the binding pocket. In addition, the model
is consistent with the success of a tethered affinity ligand in
the initial purification of the GABAAreceptor, and the lack of
activity of analogs 3 and 4.
Although the model suggests that DZ and FNZ should be
able to bind in either the h-orientation or the v-orientation,
evidence suggests that this does not occur. The profound
impact of the ?1H101R mutation on binding of DZ and FNZ
is inconsistent with the modest effect of this mutation on the
binding energies of these two ligands in the v-orientation,
arguing that little if any binding occurs in this orientation. It
is unclear why the v-orientation is not realized in practice. In
addition to the uncertainties inherent in a homology model
that is derived from the crystal structure of a different pro-
tein binding a different ligand, a crystal structure represents
a static “snapshot” of binding, and does not reproduce the
conformational changes that probably occur in the initial
interaction between ligand and receptor.
In addition, the model was unable to explain the effects of
the ?2F77Y mutation, which reduces DZ and FNZ binding
affinities by 230- and 170-fold, respectively, but did not result
in steric clashes between DZ or FNZ and ?2F77Y for any of
conformers/orientations tested. This may indicate the exis-
tence of an additional favorable orientation in which the BZ
directly contacts this residue, as proposed by Sancar et al.
(2007); however, the introduction of nonaromatic ?2F77L,
Interaction energies of analogs 1 to 4
Interaction energies (potential, van der Waals, and electrostatic energies) of receptor
and affinity probes were calculated for each basic orientation as described under
Materials and Methods using the Calculate Interaction Energy protocol.
Conformers and Analog Nos.Energy
Conformer 1 h
Conformer 2 h
?7 ? 107
?1 ? 108
?1 ? 1012
?1 ? 109
?2 ? 109
?1 ? 107
Fig. 9. Docking of active BZD analogs. Models of FNZ binding obtained in
manual docking runs were tested using analogs 1 (red-yellow-blue) and 2
(orange-yellow-blue) (Fig. 2), which bind with moderate affinities (260
and 55 nM, respectively; Zhang et al., 1994). Each analog was docked as
conformer 1 (A and C) or conformer 2 (B and D) in the h-orientation (A
and B) or v-orientation (C and D). Both analogs interacted favorably in
conformer 1 or 2 in the h-orientation (A and B). In the v-orientation (C
and D), both analogs exhibited steric interference with residues ?1His101
and ?1Lys155 (red arrows), resulting in highly unfavorable interaction
energies for both conformers (Table 1).
Fig. 10. Docking of inactive BZD analogs. Models of FNZ binding ob-
tained in manual docking runs were tested using analogs 3 and 4 depicted
in Fig. 2. Analogs 3 (green-yellow-blue) (Zhang et al., 1994) and 4 (violet-
yellow-blue) (Klopman and Contreras, 1985) are inactive. Both analogs
exhibited steric clashes (red arrows) with residues ?1His101 and
?1Lys155 when docked as conformer 1 (A and C) or conformer 2 (B and D)
in either the h-orientation (A and B) or the v-orientation (C and D).
Docking energies of Ro 11-6896 abbreviated Me(?) in different
orientations ?Me1(?) h, Me1(?) v, Me2(?) h, and Me2(?) v? and Ro
11-6893 abbreviated Me(?) in different orientations ?Me1(?) h, Me1(?)
v, Me2(?) h, and Me2(?) v?
Potential energies were calculated using the Calculate Interaction Energy protocol
as described under Materials and Methods.
?78 Me1(?) h
Berezhnoy et al.
?2F77I, or bulky ?2F77W residues at this position produces
only modest effects on DZ and FNZ affinity (Buhr et al.,
1997), whereas introduction of ?2F77C completely abolishes
FNZ binding (Teisse ´re and Czajkowski, 2001). The lack of
correlation with residue volume suggests that the effect of
this mutation may relate to conformational changes associ-
ated with receptor activation, rather than binding, which
may not be reflected by our model.
In a recent study, Sancar et al. (2007) reported automated
docking of FNZ and zolpidem, which resulted in a FNZ posi-
tion that differs significantly from our results. In this model,
the N1-methyl substituent is directed toward the membrane
and is buried in the binding site, the carboxyl group is in
close proximity to ?2Arg144 and ?1Thr206, and a fluorine
atom is located next to ?1Tyr209, with ?2Thr193 and
?2Arg194 located close to C5-phenyl moiety. Sancar et al.
(2007) reported that the ?2R194D mutation produced no
change in [3H]FNZ binding affinity, whereas Padgett and
Lummis (2008) found that ?2R194N and ?2R194K mutations
reduced maximum DZ potentiation of ?1?2?2receptors by
2.5- and 4-fold, respectively, whereas mutation of the neigh-
boring residue ?2Ser195 to threonine reduced DZ potentia-
tion by 5-fold. These results, which indicate that mutations of
loop F influence BZD efficacy rather than potency, suggest
that conformational changes within loop F are coupled to
By inspection of the proximity of residues facing the ligand
in the present model (Supplementary Table 2), we were not
able to identify specific bonds or interactions that were dom-
inant. Rather, the key observation of this study is one of
ligand orientation. However, some predictions from this
model may be informative. First, loop F is located near loop C
such that ?2Arg194 is near ?1Ser204/?1Ser205, whereas
?2Thr193 is close to ?2Trp196/?2Arg197. This arrangement
would permit the formation of hydrogen bonds within these
residue triads, possibly coordinating with the carboxyl group
of DZ or FNZ. This orientation is supported by the recent
findings of Tan et al. (2007b), who discovered that a DZ-NCS
analog with a reactive group in the 3-position of the benzo-
diazepine ring covalently labels ?1Ser205 and ?1Thr206. It is
possible then that this may reflect an activated configuration
of the binding pocket. Second, hydrogen bonds may also form
between ?1Lys155 and the FNZ nitro group oxygen atoms.
Finally, ?–? interactions could occur between the ?1Tyr159
and ?1Tyr209 and the pendant phenyl moiety.
The present study focuses on the orientation of classic
BZDs in the binding pocket, and it is unclear whether non-
BZD ligands orient similarly; however, mutagenesis and
docking studies of the non-BZD ligands zolpidem and eszo-
piclone indicate that interactions with ?1His101, ?1Ser204,
and ?2Arg194 contribute to orienting these ligands in the
binding pocket (Hanson et al., 2008).
In this study, we attempted to integrate the available
structure-activity data on the interaction of the most studied
class of BZD binding site ligands with the structure-function
data for the most studied GABAAR isoform. Docking to a
molecular model for the BZD recognition site indicates that
the key structural elements of classic 1,4-benzodiazepines,
the 1,4-benzodiazepine ring and the pharmacologically cru-
cial C5-phenyl group, most likely are oriented in the binding
pocket in parallel to the plasma membrane and perpendicu-
lar to the Cl?channel. Application of this computational
modeling strategy, which integrates site-directed affinity la-
beling with structure-activity knowledge to create a molecu-
lar model of the docking of active ligands in the binding
pocket, may provide a basis for the design of novel GABAAR
modulators with enhanced therapeutic potential.
Barnard EA, Skolnick P, Olsen RW, Mohler H, Sieghart W, Biggio G, Braestrup C,
Bateson AN, and Langer SZ (1998) International Union of Pharmacology. XV.
Subtypes of gamma-aminobutyric acidA receptors: classification on the basis of
subunit structure and receptor function. Pharmacol Rev 50:291–313.
Benson JA, Lo ¨w K, Keist R, Mohler H, and Rudolph U (1998) Pharmacology of
recombinant ?-aminobutyric acid A receptors rendered diazepam-insensitive by
point-mutated receptors. FEBS Lett 431:400–404.
Berezhnoy D, Gravielle MC, and Farb DH (2007) Pharmacology of the GABAA
receptor, in Handbook of Contemporary Neuropharmacology (Sibley D, Hanin I,
Kuhar M, and Skolnick P eds) vol 1, pp 465–568, John Wiley & Sons/Wiley-
Interscience, Hoboken, NJ.
Berezhnoy D, Baur R, Gonthier A, Foucaud B, Goeldner M, and Sigel E (2005)
Conformational changes at benzodiazepine binding sites of GABAAreceptors de-
tected with a novel technique. J Neurochem 92:859–866.
Berezhnoy D, Nyfeler Y, Gonthier A, Schwob H, Goeldner M, and Sigel E (2004) On
the benzodiazepine binding pocket in GABAAreceptors. J Biol Chem 279:3160–
Bianchi MT and Macdonald RL (2001) Agonist trapping by GABAAreceptor chan-
nels. J Neurosci 21:9083–9091.
Blount JF, Fryer RI, Gilman NW, and Todaro LJ (1983) Quinazolines and 1,4-
benzodiazepines. 92. Conformational recognition of the receptor by 1,4-
benzodiazepines. Mol Pharmacol 24:425–428.
Bonnert TP, McKernan RM, Farrar S, le Bourdelle `s B, Heavens RP, Smith DW,
Hewson L, Rigby MR, Sirinathsinghji DJ, Brown N, et al. (1999) Theta, a novel
?-aminobutyric acid type A receptor subunit. Proc Natl Acad Sci U S A 96:9891–
Borea PA, Gilli G, Bertolasi V, and Ferretti V (1987) Stereochemical features con-
trolling binding and intrinsic activity properties of benzodiazepine-receptor li-
gands. Mol Pharmacol 31:334–344.
Brejc K, van Dijk WJ, Klaassen RV, Schuurmans M, van Der Oost J, Smit AB, and
Sixma TK (2001) Crystal structure of an ACh-binding protein reveals the ligand-
binding domain of nicotinic receptors. Nature 411:269–276.
Buhr A, Baur R, and Sigel E (1997) Subtle changes in residue 77 of the ? subunit of
?1?2?2GABAAreceptors drastically alter the affinity for ligands of the benzodi-
azepine binding site. J Biol Chem 272:11799–11804.
Camerman A and Camerman N (1972) Stereochemical basis of anticonvulsant drug
action. II. Molecular structure of diazepam. J Am Chem Soc 94:268–272.
Celie PH, van Rossum-Fikkert SE, van Dijk WJ, Brejc K, Smit AB, and Sixma TK
(2004) Nicotine and carbamylcholine binding to nicotinic acetylcholine receptors as
studied in AChBP crystal structures. Neuron 41:907–914.
Chan CY and Farb DH (1985) Modulation of neurotransmitter action: control of the
gamma-aminobutyric acid response through the benzodiazepine receptor. J Neu-
Choh DW, Farb DH, and Fischbach GD (1977) Chlordiazepoxide selectively aug-
ments GABA action in spinal cord cell cultures. Nature 269:342–344.
Davies M, Martin IL, Bateson AN, Hadingham KL, Whiting PJ, and Dunn SM (1996)
Identification of domains in human recombinant GABAAreceptors that are pho-
toaffinity labelled by [3H]flunitrazepam and [3H]Ro 15-4513. Neuropharmacology
De Blas AL, Sangameswaran L, Haney SA, Park D, Abraham CJ Jr, and Rayner CA
(1985) Monoclonal antibodies to benzodiazepines. J Neurochem 45:1748–1753.
Downing SS, Lee YT, Farb DH, and Gibbs TT (2005) Benzodiazepine modulation of
partial agonist efficacy and spontaneously active GABAAreceptors supports an
allosteric model of modulation. Br J Pharmacol 145:894–906.
Duncalfe LL, Carpenter MR, Smillie LB, Martin IL, and Dunn SM (1996) The major
site of photoaffinity labeling of the ?-aminobutyric acid type A receptor by
[3H]flunitrazepam is histidine 102 of the ? subunit. J Biol Chem 271:9209–9214.
Dunn SM, Davies M, Muntoni AL, and Lambert JJ (1999) Mutagenesis of the rat ?1
subunit of the ?-aminobutyric acidAreceptor reveals the importance of residue 101
in determining the allosteric effects of benzodiazepine site ligands. Mol Pharmacol
Gibbs TT, Chan CY, Czajkowski CM, and Farb DH (1985) Benzodiazepine receptor
photoaffinity labeling: correlation of function with binding. Eur J Pharmacol
Hanson SM, Morlock EV, Satyshur KA, and Czajkowski C (2008) Structural require-
ments for eszopiclone and zolpidem binding to the gamma-aminobutyric acid
type-A (GABAA) receptor are different. J Med Chem 51:7243–7252.
He X, Huang Q, Ma C, Yu S, McKernan R, and Cook JM (2000) Pharmacophore/
receptor models for GABAA/BzR alpha2beta3gamma2, alpha3beta3gamma2 and
alpha4beta3gamma2 recombinant subtypes. Included volume analysis and com-
parison to alpha1beta3gamma2, alpha5beta3gamma2, and alpha6beta3gamma2
subtypes. Drug Des Discov 17:131–171.
Huang Q, Cox ED, Gan T, Ma C, Bennett DW, McKernan RM, and Cook JM (1999)
Studies of molecular pharmacophore/receptor models for GABAA/benzodiazepine
receptor subtypes: binding affinities of substituted beta-carbolines at recombinant
alpha x beta 3 gamma 2 subtypes and quantitative structure-activity relationship
studies via a comparative molecular field analysis. Drug Des Discov 16:55–76.
Huang Q, Liu R, Zhang P, He X, McKernan R, Gan T, Bennett DW, and Cook JM
(1998) Predictive models for GABAA/benzodiazepine receptor subtypes: studies of
quantitative structure-activity relationships for imidazobenzodiazepines at five
Docking of 1,4-Benzodiazepines with ?1?2?2GABAARs
recombinant GABAA/benzodiazepine receptor subtypes [alphaxbeta3gamma2 (x ?
1–3, 5, and 6)] via comparative molecular field analysis. J Med Chem 41:4130–
Jin R, Clark S, Weeks AM, Dudman JT, Gouaux E, and Partin KM (2005) Mecha-
nism of positive allosteric modulators acting on AMPA receptors. J Neurosci
Klopman G and Contreras R (1985) Use of artificial intelligence in the structure-
activity correlation of anticonvulsant drugs. Mol Pharmacol 27:86–93.
Kotzyba-Hibert F, Kapfer I, and Goeldner M (1995) Recent trends in photoaffinity
labeling. Angew Chem Int Ed Engl 34:1296–1312.
Kucken AM, Teisse ´re JA, Seffinga-Clark J, Wagner DA, and Czajkowski C (2003)
Structural requirements for imidazobenzodiazepine binding to GABAAreceptors.
Mol Pharmacol 63:289–296.
Kucken AM, Wagner DA, Ward PR, Teisse ´re JA, Boileau AJ, and Czajkowski C
(2000) Identification of benzodiazepine binding site residues in the ?2 subunit of
the ?-aminobutyric acid A receptor. Mol Pharmacol 57:932–939.
Marder M, Estiu ´ G, Blanch LB, Viola H, Wasowski C, Medina JH, and Paladini AC
(2001) Molecular modeling and QSAR analysis of the interaction of flavone deriv-
atives with the benzodiazepine binding site of the GABAAreceptor complex. Bioorg
Med Chem 9:323–335.
McKernan RM, Wafford K, Quirk K, Hadingham KL, Harley EA, Ragan CI, and
Whiting PJ (1995) The pharmacology of the benzodiazepine site of the GABA-A
receptor is dependent upon the type of ?-subunit present. J Recept Signal Trans-
duct Res 15:173–183.
Niehoff DL, Mashal RD, Horst WD, O’Brien RA, Palacios JM, and Kuhar MJ (1982)
Binding of a radiolabeled triazolopyridazine to a subtype of benzodiazepine recep-
tor in the rat cerebellum. J Pharmacol Exp Ther 221:670–675.
Padgett CL and Lummis SC (2008) The F-loop of the GABAAreceptor ?2 subunit
contributes to benzodiazepine modulation. J Biol Chem 283:2702–2708.
Ru ¨sch D and Forman SA (2005) Classic benzodiazepines modulate the open-close
equilibrium in alpha1beta2gamma2L gamma-aminobutyric acid type A receptors.
Sancar F, Ericksen SS, Kucken AM, Teisse ´re JA, and Czajkowski C (2007) Struc-
tural determinants for high-affinity zolpidem binding to GABA-A receptors. Mol
Sawyer GW, Chiara DC, Olsen RW, and Cohen JB (2002) Identification of the bovine
gamma-aminobutyric acid residues photolabeled by the imidazobenzodiazepine
[3H]Ro15-4513. J Biol Chem 277:50036–50045.
Schove LT, Perez JJ, and Loew GH (1994) Molecular determinants of recognition and
activation at the cerebellar benzodiazepine receptor site. Bioorg Med Chem
Sieghart W and Sperk G (2002) Subunit composition, distribution and function of
GABAAreceptor subtypes. Curr Top Med Chem 2:795–816.
Sigel E and Barnard EA (1984) A ?-aminobutyric acid/benzodiapzepine receptor
complex from bovine cerebral cortex: improved purification with preservation of
regulatory sites and their regulations. J Biol Chem 259:7129–7223.
Sigel E, Schaerer MT, Buhr A, and Baur R (1998) The benzodiazepine binding pocket
of recombinant ?1?2?2 ?-aminobutyric acid A receptors: relative orientation of
ligands and amino acid side chains. Mol Pharmacol 54:1097–1105.
Sigel E, Stephenson FA, Mamalaki C, and Barnard EA (1983) A ?-aminobutyric
acid/benzodiazepine receptor complex from bovine cerebral cortex. J Biol Chem
Smith GB and Olsen RW (1994) Identification of a [3H]muscimol photoaffinity
substrate in the bovine ?-aminobutyric acid A receptor ? subunit. J Biol Chem
Smith GB and Olsen RW (2000) Deduction of amino acid residues in the GABAA
receptor alpha subunits photoaffinity labeled with the benzodiazepine flunitraz-
epam. Neuropharmacology 39:55–64.
Tan KR, Baur R, Gonthier A, Goeldner M, and Sigel E (2007a) Two neighboring
residues of loop A of the alpha1 subunit point towards the benzodiazepine binding
site of GABAAreceptors. FEBS Lett 581:4718–4722.
Tan KR, Baur R, Gonthier A, Goeldner M, and Sigel E (2007b) The binding site for
benzodiazepines on GABAAreceptors. Soc Neurosci Abstr 33:141.11.
Tan KR, Gonthier A, Baur R, Ernst M, Goeldner M, and Sigel E (2007c) Proximity-
accelerated chemical coupling reaction in the benzodiazepine-binding site of ?-ami-
nobutyric acid type A receptors: superposition of different allosteric modulators.
J Biol Chem 282:26316–26325.
Teisse ´re JA and Czajkowski C (2001) ?-Strand in the ?2 subunit lines the benzodi-
azepine binding site of the GABAAreceptor: structural rearrangements detected
during channel gating. J Neurosci 21:4977–4986.
Thompson JD, Higgins DG, and Gibson TJ (1994) CLUSTAL W: improving the
sensitivity of progressive multiple sequence alignment through sequence weight-
ing, position-specific gap penalties and weight matrix choice. Nucleic Acids Res
Ulens C, Hogg RC, Celie PH, Bertrand D, Tsetlin V, Smit AB, and Sixma TK (2006)
Structural determinants of selective ?-conotoxin binding to a nicotinic acetylcho-
line receptor homolog AChBP. Proc Natl Acad Sci U S A 103:3615–3620.
Unwin N (2003) Structure and action of the nicotinic acetylcholine receptor explored
by electron microscopy. FEBS Lett 555:91–95.
Unwin N, Miyazawa A, Li J, and Fujiyoshi Y (2002) Activation of the nicotinic
acetylcholine receptor involves a switch in conformation of the alpha subunits. J
Mol Biol 319:1165–1176.
Verli H, Albuquerque MG, Bicca de Alencastro R, and Barreiro EJ (2002) Local
intersection volume: a new 3D descriptor applied to develop a 3D-QSAR pharma-
cophore model for benzodiazepine receptor ligands. Eur J Med Chem 37:219–229.
Villar HO, Uyeno ET, Toll L, Polgar W, Davies MF, and Loew GH (1989) Molecular
determinants of benzodiazepine receptor affinities and anticonvulsant activities.
Mol Pharmacol 36:589–600.
Wieland HA, Lu ¨ddens H, and Seeburg PH (1992) A single histidine in GABAA
receptors is essential for benzodiazepine agonist binding. J Biol Chem 267:1426–
Wieland HA and Lu ¨ddens H (1994) Four amino acid exchanges convert a diazepam-
insensitive, inverse agonist-preferring GABAAreceptor into a diazepam preferring
GABAAreceptor. J Med Chem 37:4576–4580.
Wu G, Robertson DH, Brooks CL 3rd, and Vieth M (2003) Detailed analysis of
grid-based molecular docking: a case study of CDOCKER—A CHARMm-based MD
docking algorithm. J Comput Chem 24:1549–1562.
Zhang W, Koehler KF, Harris B, Skolnick P, and Cook JM (1994) Synthesis of
benzo-fused benzodiazepines employed as probes of the agonist pharmacophore of
benzodiazepine receptors. J Med Chem 37:745–757.
Zhang W, Koehler KF, Zhang P, and Cook JM (1995) Development of a comprehen-
sive pharmacophore model for the benzodiazepine receptor. Drug Des Discov
Address correspondence to: Dr. David H. Farb, Laboratory of Molecular
Neurobiology, Department of Pharmacology and Experimental Therapeutics,
Boston University School of Medicine, 72 East Concord Street, L-603, Boston,
MA 02118. E-mail: firstname.lastname@example.org
Berezhnoy et al.