Investigation on Quantitative Structure Activity Relationships and Pharmacophore Modeling of a Series of mGluR2 Antagonists.

Abstract

MGluR2 is G protein-coupled receptor that is targeted for diseases like anxiety, depression, Parkinson's disease and schizophrenia. Herein, we report the three-dimensional quantitative structure-activity relationship (3D-QSAR) studies of a series of 1,3-dihydrobenzo[ b][1,4]diazepin-2-one derivatives as mGluR2 antagonists. Two series of models using two different activities of the antagonists against rat mGluR2, which has been shown to be very similar to the human mGluR2, (activity I: inhibition of [(3)H]-LY354740; activity II: mGluR2 (1S,3R)-ACPD inhibition of forskolin stimulated cAMP.) were derived from datasets composed of 137 and 69 molecules respectively. For activity I study, the best predictive model obtained from CoMFA analysis yielded a Q(2) of 0.513, R(2) (ncv) of 0.868, R(2) (pred) = 0.876, while the CoMSIA model yielded a Q(2) of 0.450, R(2) (ncv) = 0.899, R(2) (pred) = 0.735. For activity II study, CoMFA model yielded statistics of Q(2) = 0.5, R(2) (ncv) = 0.715, R(2) (pred) = 0.723. These results prove the high predictability of the models. Furthermore, a combined analysis between the CoMFA, CoMSIA contour maps shows that: (1) Bulky substituents in R(7), R(3) and position A benefit activity I of the antagonists, but decrease it when projected in R(8) and position B; (2) Hydrophilic groups at position A and B increase both antagonistic activity I and II; (3) Electrostatic field plays an essential rule in the variance of activity II. In search for more potent mGluR2 antagonists, two pharmacophore models were developed separately for the two activities. The first model reveals six pharmacophoric features, namely an aromatic center, two hydrophobic centers, an H-donor atom, an H-acceptor atom and an H-donor site. The second model shares all features of the first one and has an additional acceptor site, a positive N and an aromatic center. These models can be used as guidance for the development of new mGluR2 antagonists of high activity and selectivity. This work is the first report on 3D-QSAR modeling of these mGluR2 antagonists. All the conclusions may lead to a better understanding of the mechanism of antagonism and be helpful in the design of new potent mGluR2 antagonists.

Full text

Page 1

Int. J. Mol. Sci. 2011, 12, 5999-6023; doi:10.3390/ijms12095999

International Journal of
Molecular Sciences
ISSN 1422-0067
www.mdpi.com/journal/ijms
Article
Investigation on Quantitative Structure Activity Relationships
and Pharmacophore Modeling of a Series of mGluR2 Antagonists
Meng-Qi Zhang 1, Xiao-Le Zhang 2, Yan Li 1,*, Wen-Jia Fan 1, Yong-Hua Wang 3,4, Ming Hao 1,
Shu-Wei Zhang 1 and Chun-Zhi Ai 3
1 Department of Materials Science and Chemical Engineering, Dalian University of Technology,
Dalian, Liaoning 116024, China; E-Mails: manna1989@mail.dlut.edu.cn (M.-Q.Z.);
rosemarryfan@gmail.com (W.-J.F.); dluthm@yeah.net (M.H.); zswei@dlut.edu.cn (S.-W.Z.)
2 Department of Mathematical Sciences, Dalian University of Technology, Dalian, Liaoning 116024,
China; E-Mail: xlfree@foxmail.com
3 Lab of Pharmaceutical Resource Discovery, Dalian Institute of Chemical Physics,
Graduate School of the Chinese Academy of Sciences, Dalian, Liaoning 116023, China;
E-Mail: aicy@dicp.ac.cn
4 Center of Bioinformatics, Northwest A&F University, Yangling, Shaanxi 712100, China;
E-Mail: yhwang@dlfu.edu.cn
* Author to whom correspondence should be addressed; E-Mail: yanli@dlut.edu.cn;
Tel. +86-411-84986062; Fax: +86-411-84986063.
Received: 2 August 2011; in revised form: 1 September 2011 / Accepted: 9 September 2011 /
Published: 16 September 2011

Abstract: MGluR2 is G protein-coupled receptor that is targeted for diseases like anxiety,
depression, Parkinson’s disease and schizophrenia. Herein, we report the three-dimensional
quantitative structure–activity relationship (3D-QSAR) studies of a series of 1,3-dihydro-
benzo[b][1,4]diazepin-2-one derivatives as mGluR2 antagonists. Two series of models
using two different activities of the antagonists against rat mGluR2, which has been shown
to be very similar to the human mGluR2, (activity I: inhibition of [3H]-LY354740; activity
II: mGluR2 (1S,3R)-ACPD inhibition of forskolin stimulated cAMP.) were derived from
datasets composed of 137 and 69 molecules respectively. For activity I study, the best
predictive model obtained from CoMFA analysis yielded a Q2 of 0.513, R2ncv of 0.868,
R2pred = 0.876, while the CoMSIA model yielded a Q2 of 0.450, R2ncv = 0.899, R2pred = 0.735.
For activity II study, CoMFA model yielded statistics of Q2 = 0.5, R2ncv = 0.715, R2pred = 0.723.
These results prove the high predictability of the models. Furthermore, a combined analysis
OPEN ACCESS

Page 2

Int. J. Mol. Sci. 2011, 12


6000
between the CoMFA, CoMSIA contour maps shows that: (1) Bulky substituents in R7, R3
and position A benefit activity I of the antagonists, but decrease it when projected in R8 and
position B; (2) Hydrophilic groups at position A and B increase both antagonistic activity I
and II; (3) Electrostatic field plays an essential rule in the variance of activity II. In search
for more potent mGluR2 antagonists, two pharmacophore models were developed
separately for the two activities. The first model reveals six pharmacophoric features,
namely an aromatic center, two hydrophobic centers, an H-donor atom, an
H-acceptor atom and an H-donor site. The second model shares all features of the first one
and has an additional acceptor site, a positive N and an aromatic center. These models can
be used as guidance for the development of new mGluR2 antagonists of high activity and
selectivity. This work is the first report on 3D-QSAR modeling of these mGluR2
antagonists. All the conclusions may lead to a better understanding of the mechanism of
antagonism and be helpful in the design of new potent mGluR2 antagonists.
Keywords: 3D-QSAR; mGluR2 antagonist; CoMFA; CoMSIA; pharmacophore modeling

1. Introduction
Glutamate is a useful excitatory neurotransmitter of the nervous system, although its excessive
amount in the brain can lead to cell death through a process called excitotoxicity, which consists of the
over stimulation of glutamate receptors. Excitotoxicity occurs in neurological diseases such as Alzheimer’s
disease, Parkinson’s disease and multiple sclerosis [1]. The major excitatory neurotransmitter substance
in the mammalian central nervous system is L-glutamate, which acts on receptors that are highly
heterogeneous and of two types: ionotropic glutamate receptors (iGluRs) that mediate fast-synaptic
transmission in most neuronal synapses and metabotropic glutamate receptors (mGluRs) which are
G-protein Coupled Receptors linked to multiple second messengers and modulate the ion channel
currents [2,3]. Until now, at least eight mGluRs have been described, subdivided into three groups
based on their primary structure, second-messenger coupling, and pharmacology. Group I receptors
include mGluR 1 and 5, group II mGluR2 and 3, and group III mGluR 4, 6, 7 and 8 [4]. These mGluRs
play essential neuromodulatory roles throughout the brain, as such they are attractive targets for
therapeutic intervention for a number of psychiatric and neurological disorders including anxiety,
depression, Fragile X, Syndrome, Parkinson’s disease and schizophrenia [5].
In the latest decades, pharmacological agents acting at specific mGluR subtypes have been
developed. In particular, mGluR2/3 agonists showing antipsychotic properties and mGluR2/3
antagonists may be useful as antidepressants and cognitive enhancers as demonstrated in different
animal models [6]. These agents include the group II -selective agonist LY354740 and antagonist
LY341495 [7]. These agonists have been reported to have neuroprotective, anxiolytic/anti-panic and
anti-Parkinsonism properties, as well as anti-psychotic potential [8–10]. However, recently studies
showed that LY354740 has been demonstrated to impair the spatial navigation memory using the
water maze and produces a dose-dependent impairment of working memory in the delayed match to

Page 3

Int. J. Mol. Sci. 2011, 12


6001
position (DMTP) task in rats, although in humans there are no reports to date that mGluR2/3 agonists
impair the cognitions [6,11].
Recently, acting as mGluR2 and 3 non-competitive antagonists, a series of in vivo active and
well tolerated 1,3-dihydro-benzo[b][1,4]diazepin-2-one
Woltering T.J. et al [12]. They were reported to partially inhibit the binding of the selective agonist
[3H]-LY354740 to rat mGluR2, fully block the effect of LY354740, (1S,3R)-ACPD and L-glutamate in
cAMP assays. These mGluR2/3 antagonists show pharmacological potentiality by blockade of the
mGluR2/3 agonist LY354740-induced hypoactivity and improvement of a working memory deficit
induced either by LY354740 or scopolamine in the delayed match to position task (DMTP). Also,
combination studies of the antagonists with a cholinesterase inhibitor shows apparent synergistic
effects on working memory impairment induced by scopolamine. In addition, among the series of
mGluR2 antagonists reported, compound 8am was found to exhibit mild antidepressant-like activity in
the mouse, indicating further potential use of mGluR2/3 antagonists as antidepressant drugs [12].
Moreover, mGluR2 antagonists have been considered as major elements of a pharmaceutical
composition to treat and prevent acute or chronic neurological disorders including Alzheimer’s disease
and mild cognitive impairment (United States Patent 7235547). Also, Addex Pharmaceuticals
(a pharmaceutical cooperation) has made plans to move mGluR2 antagonist into clinical trials for
Alzheimer’s disease [13]. The great potential of further biological and pharmaceutical function of
mGluR2 antagonist is still under exploration. Therefore, the antagonism of the series of 1,3-dihydro-
benzo[b][1,4]diazepin-2-one derivatives has great value to investigate.
Quantitative structure activity relationship (QSAR) has been widely used to find out various
interactive fields making impacts on activity, to predict the activities of the inhibitors and thus to help
forecasting and designing of better specific ligands [14–16]. It is a mathematical model of correlation,
statistically validated, between the variation on chemical structure and the activity profile of a series
of compounds [17]. Nowadays, three-dimensional (3D) quantitative structure activity relationship
(3D-QSAR) techniques, especially comparative molecular field analysis (CoMFA) and comparative
molecular similarity analysis (CoMSIA) are routinely used in modern drug design [18]. Furthermore,
pharmacophore model is another method to investigate into the structure-activity relationship of
molecules. Common pharmacophoric features can be obtained from pharmacophore model based on
known active compounds, providing guidance for the rational design of novel selective chemicals. In
addition, molecular docking, which is also attempted in this paper, is utilized more and more in
current drug design process. Here we focused on the study of a series of 8-ethynyl-1,3-dihydro-
benzo[b][1,4]diazepin-2-one derivatives that has been reported as new potent non-competitive
mGluR2/3 antagonist [12]. The aim of the present study is to use the 137 newly fused compounds
mentioned above as a data set to identify their requisite structural features affecting the mGluR2
antagonist effects by a combination of several in silico approaches including CoMFA, CoMSIA and
pharmacophore modeling. Structure-activity relationship concerning mGluR3 is not investigated, for
there are no reported experimental values of activity presently. As far as we know, this study provides
the first 3D-QSAR study and pharmacophore modeling for the series of new mGluR2 antagonists.



derivatives were synthesized by

Page 4

Int. J. Mol. Sci. 2011, 12


6002
2. Results and Discussion
To judge whether a QSAR model is highly qualified, several statistical parameters including
especially the cross-validated correlation coefficient (Q2), non cross-validated correlation coefficient
(R2ncv), standard error of estimate (SEE) and F-statistic values as well as the optimum number of
components (OPN) should be evaluated. Various 3D-QSAR models were generated and the best model
was selected based on the statistically significant parameters obtained. For both 3D-QSAR studies,
good correlations were observed in the obtained CoMFA and CoMSIA models demonstrated by the
high values of Q2, Rncv, Rpred and other statistical results. Table 1 summarizes the statistical results of
the CoMFA and CoMSIA analyses.
Table 1. Summary of comparative molecular field analysis (CoMFA) and comparative
molecular similarity analysis CoMSIA results for activity I/activity II.
PLS Statistics

Q2
R2ncv
SEE
F
R2pre
SEP
OPN
Activity I Activity II
CoMFA
0.513
0.868
0.296
96.022
0.876
0.288
7
CoMSIA
0.450
0.899
0.273
101.353
0.735
0.420
8
Contribution
0.112
0.277
0.184
0.135
0.256
0.035
CoMFA
0.503
0.715
0.265
20.907
0.723
0.241
6
CoMSIA
0.367
0.657
0.285
24.927
0.667
0.265
4
Steric 0.488
0.461



0.051

Electrostatic
Hydrophobic
H-donor
H-acceptor
Clogp
Q2, cross-validated correlation coefficient after the leave-one-out procedure; R2
correlation coefficient; SEE, standard error of estimate; F, ratio of R2
= R2
standard error of prediction; OPN, optimal number of principal components.
0.917



0.083
0.455
0.409


0.136
ncv, non-cross-validated
ncv explained to unexplained
ncv/(1 − R2
ncv); R2
pre, predicted correlation coefficient for the test set of compounds; SEP,
2.1. Results for Activity I
2.1.1. 3D-QSAR Statistical Results
During the molecular modeling process, 110 compounds out of the total 137 mGluR2 antagonists
were used as training set and the remaining 27 compounds were used as test set (shown in Tables S1–S4
in supporting information). The best results were obtained at a column filtering of 1 kcal/mol for both
steric and electrostatic fields for CoMFA analysis. PLS analysis showed a Q2 value of 0.513 with 7
components for CoMFA, indicating a good internal predictive capacity of the model. A high
correlation coefficient of 0.868 for the non-cross-validated model shows its self-consistency.

Page 5

Int. J. Mol. Sci. 2011, 12


6003
In addition, other statistical results including a SEE of 0.296 and an F-test value of 96.022 are
reported (Table 1). Steric field descriptors explain 0.488 of the variance, while the electrostatic
descriptors contribute 0.461, suggesting a balanced percentage of the influence exerted by the two
fields on the activity of the antagonist. In addition, ClogP contributes a minor percentage of 5.1,
indicating that the hydrophobicity of compound does affect its antagonistic activity to some extent.
For the CoMSIA analysis, all combinations of the five different field descriptors (the steric,
electrostatic, hydrophobic, H-bond-donor and H-bond acceptor) were used, with attempt to seek for the
best CoMSIA model while avoiding the risk of possible omitting of some optimal ones. As a result, out
of all CoMSIA models established using the same training set as used in the CoMFA analysis, the
optimal CoMSIA model was made use of all five field parameters, and reveals a validated Q2 of 0.450
for 8 components, a Rncv2 of 0.899, F value of 101.353 and SEE of 0.273. The electrostatic field was
demonstrated contributing a lot to the model with a sum of 27.7 percentages. In addition, the hydrogen
bond acceptor field descriptor fulfills its role by correlating with the mGluR2 antagonist activity with a
fraction of 0.256. Steric field contributes 11.2 percent of the variances. H-bond donor field and
acceptor field contribute 0.184 and 0.135 respectively.
All the statistical parameters of CoMFA and CoMSIA obtained show that the models generated are
reasonable. Furthermore, the accuracy of these models was elucidated using an external test set of 27
compounds. With a high predictive coefficient R2pred of 0.876 (R2ncv = 0.868) for CoMFA and 0.735
(R2ncv = 0.899) for CoMSIA achieved, the test sets potently validate the efficacy of the CoMFA and
CoMSIA models. Figure 1A,B illustrate the correlation plots of the experimental versus the predicted
pIC50 values of the training (filled black square) and test sets (filled blue circle) for the optimal CoMFA
and CoMSIA models, respectively. Clearly, a good correlation was observed from this figure since the
predicted values are almost as accurate as the experimental activities for the whole dataset, and all
points are rather uniformly distributed around the regression line. This good agreement between the
predicted and experimental activity data proves the satisfactory predictive ability of both the CoMFA
and CoMSIA models.
Figure 1. The ligand-based correlation plots of the predicted versus the actual pIC50 values
using the training (filled black square) and the test (filled blue circle) set compounds based
on (A) CoMFA for activity I (Dataset A: 137 compounds); (B) CoMSIA models for activity I
(Dataset A: 137 compounds); (C) CoMFA for activity II (Dataset B: 69 compounds).