Molecular docking and 3D-QSAR CoMFA studies on indole inhibitors of GIIA secreted phospholipase A(2).

Abstract

Automated docking allowing a "protein-based" alignment was performed on a set of indole inhibitors of the GIIA secreted phospholipase A(2) (GIIA sPLA(2)). A correlation between the binding scores and the experimental inhibitory activity was observed (r(2) = 0.666, N = 34). All the indole inhibitors were docked in the active site of the GIIA sPLA(2) enzyme, and the best score docking pose of each inhibitor was used for the "protein-based" alignment of the compounds. A three-dimensional quantitative structure-activity relationship (3D-QSAR) model was then established using the comparative molecular field analysis (CoMFA) method. The set of 34 indole inhibitors was divided into two subsets: the training set, composed of 26 compounds, and the test set, consisting of eight compounds. The robustness and the predictive ability of the generated CoMFA model were examined by using the test set. A good correlation (r(2) = 0.997) between predicted and experimental inhibitory activity data allows the validation of the CoMFA model. Finally, the generated CoMFA model was used for the design and evaluation of new compounds. The new designed compounds exert improved predicted inhibitory activity and may be a target for the synthesis of new GIIA sPLA(2) indole inhibitors.

Full text

Page 1

Molecular Docking and 3D-QSAR CoMFA Studies on Indole Inhibitors of GIIA
Secreted Phospholipase A2
Varnavas D. Mouchlis, Thomas M. Mavromoustakos,* and George Kokotos
Laboratory of Organic Chemistry, Department of Chemistry, University of Athens, Panepistimiopolis,
Athens 15771, Greece
Received June 2, 2010
Automated docking allowing a “protein-based” alignment was performed on a set of indole inhibitors of the
GIIA secreted phospholipase A2(GIIA sPLA2). A correlation between the binding scores and the experimental
inhibitory activity was observed (r2) 0.666, N ) 34). All the indole inhibitors were docked in the active
site of the GIIA sPLA2enzyme, and the best score docking pose of each inhibitor was used for the “protein-
based” alignment of the compounds. A three-dimensional quantitative structure-activity relationship (3D-
QSAR) model was then established using the comparative molecular field analysis (CoMFA) method. The
set of 34 indole inhibitors was divided into two subsets: the training set, composed of 26 compounds, and
the test set, consisting of eight compounds. The robustness and the predictive ability of the generated CoMFA
model were examined by using the test set. A good correlation (r2) 0.997) between predicted and
experimental inhibitory activity data allows the validation of the CoMFA model. Finally, the generated
CoMFA model was used for the design and evaluation of new compounds. The new designed compounds
exert improved predicted inhibitory activity and may be a target for the synthesis of new GIIA sPLA2
indole inhibitors.
1. INTRODUCTION
Phospholipase A2corresponds to a superfamily of enzymes
that to date include 15 separate, identifiable groups and
numerous subgroups of PLA2enzymes.1,2These enzymes
catalyze the hydrolysis of the ester bond of membrane
phospholipids at the sn-2 position. The five main categories
of PLA2enzymes are: the secreted sPLA2s, the cytosolic
cPLA2s, the Ca2+-independent iPLA2s, the PAF acetylhy-
drolases, and the lysosomal PLA2s.
The products of the sn-2 ester bond hydrolysis of phos-
pholipids by the PLA2 enzymes are free fatty acids and
lysophospholipids. The action of the PLA2enzymes on the
phospholipids is of high importance when the esterified fatty
acid at the sn-2 position is arachidonic acid (AA). AA is
converted by different downstream metabolic enzymes (such
as COX-1, COX-2 and 5-LO) to several bioactive lipid
mediators called eicosanoids, including prostaglandins (PGs)
and leukotrienes (LTs).3,4The eicosanoids participate in
many pathological inflammatory conditions, such as athero-
sclerosis5and ischemia diseases.6The lysophospholipids are
precursors of other bioactive mediators, such as the platelet
activating factor (PAF).7
PLA2enzymes are membrane-bound enzymes, and as a
result they have an i-face that has been proposed to make
contact with the substrate interface.8The i-face of the sPLA2
enzymes is a relatively flat surface of 1600 A2, which binds
tightly to the phospholipid bilayers (Kd< 10-13M).9There
are some polar and hydrophobic residues on the flat surface,
through which the sPLA2s bind to the ionic bilayers. The
sPLA2enzymes are interfacial enzymes, and their active site
is localized near the substrate binding i-face.10
There are 10 known members of sPLA2enzymes that have
been identified in mammals, which are numbered and
classified in groups according to the chronological order of
their discovery. The 10 groups of the sPLA2enzymes are:
IB, IIA, IIC-F, III, V, X, and XII.11The characterization
of their molecular structure, the classification, the genome
localization, and the details of their catalytic mechanism
attracted the research interest of many scientists.11-15In the
class of the low molecular weight secreted PLA2enzymes,
the group IIA secreted PLA2(GIIA sPLA2) is of paramount
importance since it is involved in several inflammatory
diseases, such as rheumatoid arthritis, and was cloned in
1989.16,17In vitro studies using recombinant GIIA sPLA2
on phospholipid substrates have provided important informa-
tion about the biochemistry of the enzyme. For instance, the
GIIA sPLA2(known as human nonpancreatic sPLA2) shows
biological activity on ionic phospholipids, such as phos-
phatidylglycerol (PG), phosphatidylserine (PS), and phos-
phatidylethanolamine (PE), but it is inactive on phosphati-
dylcholine (PC).18
GIIA sPLA2 is a disulfide-linked enzyme, with seven
disulfide bonds, which contribute to the folding and stability
of the enzyme structure. In addition, it has a Ca2+-binding
loop and a His/Asp catalytic dyad. The mechanism of
substrate hydrolysis begins by the activation of a water
molecule by the catalytic histidine (His47). Beside this
histidine, there is an aspartate residue (Asp48), which
together with three other residues (Gly29, Gly31, and His27)
construct the conserved Ca2+-binding loop, where the Ca2+
ion is bound.19The hepta-coordinated Ca2+ion provides two
positions for the substrate binding, one axial and one
* Corresponding author. E-mail: tmavrom@chem.uoa.gr. Telephone: +30
2107274293.
J. Chem. Inf. Model. 2010, 50, 1589–1601
1589
10.1021/ci100217k
 2010 American Chemical Society
Published on Web 08/26/2010

Page 2

equatorial.20The high-resolution crystal structures of the
GIIA sPLA2enzyme have defined an enclosed active site
with a hydrophobic region which is located near the
N-terminal helix.21,22This hydrophobic region contributes
to the binding of a phospholipid molecule and to the
interfacial binding of the enzyme to the phospholipid
bilayers.
The GIIA sPLA2enzyme is an attractive target for the
development of new inhibitors, which might lead to thera-
peutic drugs for diseases where the enzyme is involved. It
has also been crystallized with or without different ligands,
and this fact renders the enzyme a suitable target for drug
design using computational methods.23-28Many different
classes of synthetic and natural inhibitors are known to
date.29-33However, in order to enhance the design of new
GIIA sPLA2inhibitors, the requirements are: (i) the knowl-
edge of the exact location of the active site and the
understanding of the inhibitor-enzyme interactions; and (ii)
the establishment of drug design computational protocols to
predict the activity of new designed molecules.
The main computational methods used in the rational drug
design can be divided into two groups: (i) protein-based
studies, which include molecular docking studies23,24and
molecular dynamics simulations25,26(MD) (wherein the
receptor-ligand interactions model is simulated); and (ii)
quantitative structure-activity relationship (QSAR) ana-
lysis,27,28which does not require a priori hypothesis about
the receptor structure. The QSAR analysis is based on the
pharmacophore alignment of the ligands and implies a
common biochemical mechanism.
Molecular docking is an extensively used computational
method in rational drug design,34and the main principals
that govern this technique have been described in recent
review articles.35-37For any docked pose, the binding score
is calculated generally as the sum of the electrostatic, van
der Waals and hydrophobic interactions, and hydrogen
bonding. Some scoring functions include also metal-binding
and solvation terms. Methodologies have been developed to
pick up the pose that simulates best the biological molecular
interactions.38
The 3D-QSAR comparative molecular fields analysis
(CoMFA)39requires the alignment of all the studied mol-
ecules in a three-dimensional space. In conventional CoMFA
studies, the molecules are fitted to a reference molecule,
which is the most rigid or constrained molecular structure
among the most active compounds. The hypothesis in this
case is that the conformation of the reference compound is
supposed to correspond to the “biologically active” confor-
mation. In the case of a known X-ray crystal structure with
high resolution, a structure-based design protocol reduces
the uncertainty about the determination of the “bioactive
conformation” of the reference compound. A protocol of
automated molecular docking of all the available compounds
in the receptor active site creates an indubitable “receptor-
based” alignment for the CoMFA analysis.
The present study is consisted of molecular docking
calculations and the generation of a CoMFA model on a set
of GIIA sPLA2indole inhibitors reported in the literature.40
All the GIIA sPLA2indole inhibitors were docked in the
enzyme active site using GLIDE 5.5.41-43The extra-
precision44(XP) mode of GLIDE was used for the docking
calculations. The best score docking pose of each indole
inhibitor was used in a “protein-based” alignment for the
CoMFA procedure. The quality of the CoMFA analysis
depends greatly on the alignment of the studied compounds.
“Protein-based” alignment via molecular docking allows the
development of a high-quality CoMFA model.
2. COMPUTATIONAL METHODS
2.1. Preparation of the GIIA sPLA2Enzyme File. Four
crystal structures of the GIIA sPLA2 enzyme which are
deposited in the RCSB protein data bank were downloaded
(PDB IDs: 1DB4 holo form 2.20 Å X-ray resolution,451DB5
holo form 2.80 Å X-ray resolution,451KVO holo form 2.00
Å X-ray resolution,46and 1J1A holo form 2.20 Å X-ray
resolution47). The objective was to judge which one is
sufficient for docking the indole inhibitors. The procedure
for this determination is consisted by the following steps:
(i) using “superposition panel” of Maestro 9.048all the crystal
structures were superimposed based on all the backbone
atoms including beta carbons. The crystal structure with PDB
ID: 1DB4 was chosen as the reference structure for the
superposition (rmsd between: 1DB4-1DB5: 0.147 Å,
1DB4-1J1A: 0.746 Å, 1DB4-1KVO: 0.487 Å). No sig-
nificant structural differences were observed (see Figure 1
in Supporting Information); (ii) the active site region was
examined to determine if the superimposed ligands can fit
into the reference site without steric clashes. No significant
steric clashes were observed; (iii) the active site region of
all the crystal structures, in turn, was examined in order to
determine whether any residues in the superimposed protein
differ appreciably in position or conformation from those in
the reference site. No significant differences were observed.
Thus, the 1DB4.pdb file has been chosen for the molecular
docking calculations. This file contains a single unit of the
GIIA sPLA2 enzyme cocrystallized with a native indole
inhibitor, which is structurally similar with the indole
inhibitors used in this study. The 1DB4.pdb crystal structure
was prepared using the “Protein Preparation Wizard” panel49
of Schro ¨dinger 2009 molecular modeling package. In par-
ticular, using the “preprocess and analyze structure” tool,
the bond orders were assigned, all the hydrogen atoms were
added, the calcium ion was treated in order to have the
correct geometry and formal charge (+2), the disulfide bonds
were assigned, and all the water molecules in a distance
greater than 5 Å from any heterogroup were deleted. Using
Epik 2.0,50,51a prediction of the heterogroups ionization and
tautomeric states was performed. An optimization of the
hydrogen-bonding network was performed using the “H-bond
assignment” tool. Finally, using the “impref utility”, the
positions of the hydrogen atoms were optimized by keeping
all the heavy atoms in place.
2.2. Preparation of Ligands Files. All the indole ligands
were built and adjusted using the Maestro 9.0 molecular
builder. All the hydrogen atoms were added, and the ligands
were submitted in full structure optimization, using the
minimization procedure of MacroModel 9.7.52For the
minimization, a standard molecular mechanics energy func-
tion (OPLS_200553force field) and the Polak-Ribiere
conjugated gradient method (5000 iterations with gradient
0.01 kJ/mol·Å)54were used. Solvent effects were modeled
with the generalized Born/surface area (GB/SA) implicit
solvent model55using water as solvent and normal non-
1590
J. Chem. Inf. Model., Vol. 50, No. 9, 2010
MOUCHLIS ET AL.

Page 3

bonded cutoffs. The constant dielectric electrostatic treatment
was selected with a dielectric constant of 1. The minimized
structures of the indole ligands were subsequently prepared
using LigPrep 2.3.56In particular, using LigPrep 2.3 all the
structures were checked to be correct, ionization states were
generated for all the indole inhibitors by taking into account
the metal mode of Epik 2.0 ionizer for the metalloproteins,
and the OPLS_2005 force field was used for the structure
optimization of all the indole ligands.
2.3. Receptor Grid Generation for the GIIA sPLA2
Enzyme. Glide is a grid-based ligand docking with energetics
approach and searches for favorable interactions between
ligands and receptors. The shape and properties of the
receptor are represented on a grid by different sets of fields
that provide progressively more accurate scoring of the ligand
pose (by “pose” we mean a complete specification of the
ligand: position and orientation relative to the receptor, core
conformation, and rotamer-group conformations). These
fields are generated as preprocessing steps in the calculation
and hence need to be computed only once for each receptor.
For the grid generation of the GIIA sPLA2 enzyme, the
binding site was defined using the native indole ligand
cocrystallized with the enzyme. The default value (1.00) for
the van der Waals radii scaling factor was chosen, which
means no scaling for the nonpolar atoms was performed (no
flexibility was simulated for the receptor). Glide uses two
boxes to organize the calculation: (i) the grids are calculated
within the space defined by the enclosed or outer box. This
is also the box in which all the ligand atoms must be
contained; (ii) acceptable positions for the ligand center
during the site point search lie within the ligand diameter
midpoint box or the inner box. This box gives a more precise
measure of the effective size of the search space. However,
ligands can move outside this box during grid minimization.
The native indole inhibitor has a very similar size with the
indole ligands of this study, and for defining the size of the
inner box, the option “docking ligands similar in size to
the workspace ligand” was chosen. The Cartesian coordinates
of the inner box, X, Y, and Z length were set to 12 Å.
2.4. Protocol of Grid-Based Ligand Docking with
Energetics (Glide). Glide uses a series of hierarchical filters
to search for possible locations of the ligand in the active
site region of the receptor. For the grid-based ligand docking
with energetics, the receptor grid generation file was used
(see Section 2.3). The next step is to define the conforma-
tional space of the ligand. Glide uses a sophisticated fashion
described by Friesner and co-workers to choose six lowest-
energy poses.42Finally, the three to six lowest-energy poses
obtained in this fashion are subjected to a Monte Carlo
procedure that examines nearby torsional minima. This
procedure is needed in some cases to properly orient
peripheral groups and occasionally alters internal torsion
angles. To soften the potential for nonpolar parts of the
ligands the “scaling of van der Waals radii” was used with
“scaling factor” of 0.8 and “partial charge cutoff” of 0.15.
Glide uses the GlideScore in order to rank poses. GlideScore
is a modified and expanded version of ChemScore57scoring
function that is used for predicting binding affinity and rank-
ordering ligands in database screens. By using GlideScore
as the scoring function, a composite Emodel score is then
used to rank the poses of each ligand and to select the poses
to be reported to the user. Emodel combines GlideScore, the
nonbonded interaction energy, and for flexible docking, the
excess internal energy of the generated ligand conformation.
In the present study the extra-precision44(XP) mode of
GlideScore scoring function was used. XP mode can be used
when the active site of the receptor contains a metal and
often works well. Glide assigns a special stability to ligands
in which anions coordinate to the metal center. To benefit
from this assignment, groups such as carboxylates, hydrox-
amates, and thiolates must be anionic. The protein residues
that line the approach to the metal center need to be
protonated in a manner compatible with the coordination of
an anionic ligand, such as a carboxylate or hydroxamates.
With respect to the GIIA sPLA2enzyme, the calcium ion
has charge of +2, but the effective charge which GLIDE
uses to dock the ligands is +1 because the ionic carboxylate
group of Asp48, that bears a charge -1, binds to the calcium
ion. The GlideScore XP scoring function also includes terms
which assign penalties to structures where statistical results
indicate that one or more groups is inadequately solvated.
A large database of cocrystallized structures has been used
by the creators of the scoring function, to optimize the
parameters associated with the penalty terms. No penalties
on the docked structures were observed. The force field used
for the docking was the OPLS_2001.
2.5. CoMFA. 2.5.1. Data Set (Training and Test Sets).
The 34 indole inhibitors and the corresponding biological
data used in this study have been selected from the
literature.40The molecular structures and the GIIA sPLA2
inhibitory activity data for these 34 indole inhibitors are
summarized in Table 1. All the collected biological data (%
inhibition) were measured in vitro under the same experi-
mental conditions.40
These 34 compounds were divided into two subsets: 26
of them were used as a training set, and 8 were used as a
test set. The compounds were separated to training and test
sets in order for the two sets to have a good molecular
diversity. Both, training and test sets included active and
inactive compounds to a scale from 0 to 100% inhibition
against the GIIA sPLA2enzyme.
2.5.2. Data Set Alignment. In the 3D-QSAR CoMFA
analysis, biological conformation and alignment rule selection
are two important factors to construct a reliable model. In
the present study, the poses from the automated molecular
docking calculations by GLIDE were used for the alignment
method. Compound 1 was selected as a template molecule
for the alignment. Considering the structural similarities of
the studied indole inhibitors, 17 common atoms of com-
pounds 1-34 were selected for the alignment. The alignment
of the molecules was performed using the module “database
alignment” in SYBYL 8.0 molecular modeling package. The
alignment results are shown in Figure 2 in the Supporting
Information with the common atoms highlighted in magenta
color in compound 1.
2.5.3. CoMFA Analysis. In the present study, a “protein-
based” alignment was used (the conformations of the studied
compounds used for the alignment were chosen by docking
these compounds in the GIIA sPLA2enzyme active site).
Atomic charges for the aligned molecules were calculated
using the Gasteiger-Hu ¨ckel method, which is a combination
of two other charge computational methods: the Gasteiger-
Marsili58method to calculate the σ component of the atomic
charge and the Hu ¨ckel59method to calculate the π compo-
COMFA STUDIES ON INDOLE INHIBITORS
J. Chem. Inf. Model., Vol. 50, No. 9, 2010 1591

Page 4

nent of the atomic charge. The total charge is the sum of
the charges calculated by the two methods. The CoMFA
fields are generated by creating a grid around the molecule
and calculating the steric and electrostatic potentials at each
point on the grid using a charged probe atom. The CoMFA
calculations were performed using Tripos Advance CoMFA60
module in SYBYL 8.0. The steric and electrostatic field
energies were calculated using the Lennard-Jones and
coulomb potentials, respectively with 1/r2distance-dependent
dielectric constant in all intersections of regularly spaced (0.2
nm) grid. The sp3carbon atom with radius of 1.53 Å and
charge +1.0 was used as a probe to calculate the steric and
electrostatic energies between the probe and the molecules
using the standard Tripos force field.61The truncation for
both the steric and electrostatic energies was set to 30 kcal
mol-1. This indicates that any steric or electrostatic field
value that exceeds this value will be replaced with 30 kcal
mol-1, thus makes a plateau of the fields close to the center
of any atom.
2.5.4. Partial
Validations. The initial PLS analysis is used to correlate the
experimental inhibitory activity of the indole inhibitors
against the GIIA sPLA2enzyme with the CoMFA values
containing magnitude of steric and electrostatic potentials.
CoMFA standard scaling was applied to all the CoMFA
analysis. The full PLS analysis was run with a column
filtering of 2.0 kcal mol-1to reduce the noise and to speed
up the calculation. In CoMFA analysis, descriptors were
treated as independent variables, whereas the % inhibition
against the GIIA sPLA2 enzyme values were treated as
dependent variables in the PLS regression analyses to derive
the 3D-QSAR model. The model was assessed by their cross-
validated r2(rcv
The final model (non-cross-validated conventional analysis)
was developed from the model with the highest cross-
validated r2(rcv
sessed by the r2conventional correlation coefficient, the s
standard error of prediction, and the F value (Fisher test).
To obtain confidence limits and test the stability of
obtained PLS model, for the conventional CoMFA run,
bootstrapping64was performed (100 runs, column filtering:
2.00 kcal mol-1). The idea is to simulate a statistical sampling
procedure by assuming that the original data set is the true
population and generating many new data sets from it. These
new data sets (called bootstrap samplings) are of the same
size as the original data set and are obtained by randomly
choosing samples (rows) from the original data, with repeated
selection of the same row being allowed. The statistical
calculation is performed on each of these bootstrap sam-
plings, with new values being calculated for each of the
parameters to be estimated. The difference between the
parameters calculated from the original data set and
the average of the parameters calculated from the many
bootstrap samplings is a measure of the bias of the original
calculation.60
Least-Squares(PLS)Analysisand
2) using leave-one-out (LOO) procedure.62,63
2). The non-cross-validated model was as-
3. RESULTS AND DISCUSSION
3.1. Molecular Docking Results. In this section, the
results derived from the docking of the indole inhibitors in
the GIIA sPLA2enzyme active site will be discussed. The
attention has been focused on the most characteristic
receptor-ligand interactions for a few key inhibitor structures
and especially for compound 1, the most active one in the
data set.
The established automated molecular docking was first
applied to two indole inhibitors already cocrystallized with
the GIIA sPLA2enzyme and described by Schevitz et al.45
These two compounds (see Table 1 in the Supporting
Information) are structurally very similar to the indole
inhibitors in this study. By docking these compounds, it was
possible to examine if GLIDE is able to reproduce crystal-
lographic experimental data for structural similar compounds.
Afterward, the 34 indole inhibitors in this study were docked
in the GIIA sPLA2enzyme active site, and the inhibitory
activity was compared with the XP binding scores calculated
by GLIDE. The purpose of this docking was to examine if
there is a correlation between the experimental % inhibitory
activity and the theoretical XP binding scores.
The crystallographic data reveal the following interactions
of indole845and indole645with the GIIA sPLA2enzyme
Table 1. In Vitro % Inhibition of the Indole Inhibitors40Against
the GIIA sPLA2Enzyme and the XP Binding Scores Calculated by
GLIDE
compoundR group
% inhibition
against GIIA
sPLA2
XP
GScore
Training Set
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
2,6-Cl2-C6H3CH2
3-I-C6H4CH2
3-(OCF3)-C6H4CH2
3-Cl-C6H4CH2
3-Br-C6H4CH2
2-(CF3)-C6H4CH2
2,6-F2-C6H3CH2
CH3(CH2)6
4-(CH3)-C6H4CH2
3-CN-C6H4CH2
2,4-F2-C6H3CH2
4-(CF3)-C6H4CH2
C6H5CH2
4-F-C6H4CH2
CH3(CH2)5
2,5-F2-C6H3CH2
(CH3)2CHCH2
C6H5COCH2
3-(NO2)-C6H4CH2
CH3(CH2)2
CH3CH2
4-CN-C6H4CH2
1-(8-CH2Br)-naphthaleneCH2
4-(7-CH3O)-coumarinylCH2
2-(OCH3)-5-(NO2)-C6H3CH2
phthalimidoCH2
100
98
97
96
95
93
91
89
88
86
84
83
82
80
79
76
64
61
57
51
39
29
24
-11.21
-10.45
-10.45
-10.61
-10.29
-10.68
-10.73
-9.70
-10.41
-10.47
-10.69
-9.71
-10.72
-10.59
-9.96
-10.63
-9.76
-9.41
-9.78
-9.32
-9.17
-9.74
-9.18
-9.97
-8.85
-6.48
9
7
0
Test Set
27
28
29
30
31
32
33
34
2-napthaleneCH2
2-(CF3)-4-F-C6H3CH2
3-(CH2Br)-C6H4CH2
CH3(CH2)3
CH3(CH2)4
C6F5CH2
(S)-(+)-CH3CH2CH(CH3)CH2
phthalimido(CH2)4
97
93
86
69
57
33
22
-10.82
-10.46
-10.64
-9.90
-9.48
-9.38
-8.97
-8.632
1592
J. Chem. Inf. Model., Vol. 50, No. 9, 2010
MOUCHLIS ET AL.

Page 5

active site: (i) two interactions between two oxygen atoms
of the ligands and the calcium ion and (ii) one hydrogen
bond with His47. In the case of indole8, one oxygen atom
of the phosphonate group participates in a hydrogen bond
with Lys62 through a water molecule placed near the enzyme
active site. The compounds indole8 and indole6 were docked
successfully by GLIDE. The interactions aforementioned
were reproduced, and the conformation of each ligand was
almost identical to the crystallographic one. In addition, three
other hydrogen bonds were observed between one hydrogen
atom of the amide group and Asp48, the oxygen atom of
the amide group and Gly29 and one of the oxygen atoms of
the carboxylate or phosphonate group with Gly31 (see
Figure 3 and Table 1 in the Supporting Information).
The docking calculations were then applied to the 34
indole inhibitors in this study. The computed XP binding
scores of the indole inhibitors from the docking calculations
with GLIDE were compared with their experimental inhibi-
tory activity against the GIIA sPLA2enzyme (Table 1). The
linearity of the plot (r2) 0.666, N ) 34) presents a good
correlation between the XP binding scores and the inhibitory
activity of the indole inhibitors (Figure 1).
In Figure 2 the binding of the most active compound 1 is
presented and reveals the main interactions with the enzyme
active site. The observed interactions are: (i) two interactions
of two oxygen atoms (the oxygen atom of the amide group
and one of the oxygen atoms of the carboxylate group) with
the calcium ion (COO-···Ca2+2.30 and NHCO···Ca2+2.70
Å); (ii) one hydrogen bond of one of the hydrogen atoms of
the amide group with the nitrogen atom of His47 (H· · ·N,
1.90 and N· · ·N 2.90 Å); (iii) one hydrogen bond of one of
the hydrogen atoms of the amide group with the oxygen atom
of Asp48 (H···O 1.80 and O···O 2.70 Å); (iv) one hydrogen
bond of the oxygen atom of the amide group with the
hydrogen atom of the amine group of Gly29 (O.. .H-N 2.10
and O· · ·N 2.80 Å); and (v) one hydrogen bond of one of
the oxygen atoms of the carboxylate group with the hydrogen
atom of the amine group of Lys62 through a water molecule
placed near the active site of the enzyme (O···H-O···H-N
2.50, 1.90 and O· · ·O· · ·N 3.00, 2.90 Å). The phenyl ring
of compound 1 participates in aromatic (π-π) stacking
interactions with the residues Phe5 (Rcen ) 6.49 Å, θ )
9.40°) and His6 (Rcen ) 5.76 Å, θ ) 70.77°), and the indole
ring participates also in aromatic (π-π) stacking interactions
with Phe5 (Rcen ) 5.88 Å, θ ) 84.86°) and His47 (Rcen )
6.64 Å, θ ) 70.20°). Aromatic (π-π) stacking interactions
are one of the forces governing molecular recognition. Burley
and Petsko have reported that aromatic (π-π) stacking
interactions in proteins operate at distances (Rcen) of 4.5-7.0
Å between the center mass of the rings, the rings’ centroids.65
The angle (θ) between normal vectors of interacting aromatic
rings is typically between 30° and 90°, producing a “tilted-
T” or “edge-to-face” arrangement of interacting rings. Hunter
and co-workers66have reported that aromatic (π-π) parallel
stacking interactions (θ < 30°) between phenylalanine
residues in proteins are also favorable, if the rings are offset
from each other. The two chlorine atoms participate in van
der Waals contacts with the carbon atoms of the site chains
of Phe5, Ile9, Ala17, Ala18, and Tyr21.
In the 34 indole inhibitors described in this study, the main
structural changes are in the R group. This position is
governed mainly by steric effects in the active site of the
GIIA sPLA2enzyme. In Figure 3 the binding of the inactive
compound 25 is presented. The main interactions of com-
pound 25 in the active site of the enzyme are: (i) two
interactions of the two oxygen atoms (the oxygen atom of
the amide group and one of the oxygen atoms of the
carboxylate group) with the calcium ion (COO-···Ca2+2.30
and NHCO· · ·Ca2+2.60 Å); (ii) one hydrogen bond of one
of the hydrogen atoms of the amide group with the nitrogen
Figure 1. Plot of the experimental % inhibition against the GIIA sPLA2enzyme versus the XP binding scores calculated by GLIDE. The
linear regression statistics for the data set are also presented.
COMFA STUDIES ON INDOLE INHIBITORS
J. Chem. Inf. Model., Vol. 50, No. 9, 2010 1593