Homology modeling in tandem with 3D-QSAR analyses: A computational approach to depict the agonist binding site of the human CB2 receptor

Article abstract of DOI:10.1016/j.ejmech.2011.07.023

CB2 receptor belongs to the large family of G-protein coupled receptors (GPCRs) controlling a wide variety of signal transduction. The recent crystallographic determination of human β2 adrenoreceptor and its high sequence similarity with human CB2 recepto

Full text of DOI:10.1016/j.ejmech.2011.07.023

European Journal of Medicinal Chemistry 46 (2011) 4489e4505
Contents lists available at ScienceDirect
European Journal of Medicinal Chemistry
journal homepage: http://www.elsevier.com/locate/ejmech
Original article
Homology modeling in tandem with 3D-QSAR analyses: A computational
approach to depict the agonist binding site of the human CB2 receptor
,
d
Elena Cichero a, Alessia Ligresti b, Marco Allarà b, Vincenzo di Marzo b, Zelda Lazzati c
,
Pasqualina D’Ursi c, Anna Marabotti c, Luciano Milanesi c, Andrea Spallarossa a, Angelo Ranise a,
,
Paola Fossa a
*
a Dipartimento di Scienze Farmaceutiche, Università di Genova, Viale Benedetto XV n. 3, 16132 Genova, Italy
b Endocannabinoid Research Group, Istituto di Chimica Biomolecolare, CNR, Via Campi Flegrei 34, 80078 Pozzuoli Napoli, Italy
c Institute for Biomedical Technologies e National Research Council (ITB-CNR), Via Fratelli Cervi 93, 20090 Segrate (MI), Italy
d Institute of Genetics and Biophysics “A. Buzzati Traverso” e National Research Council (IGB-CNR), Via P. Castellino 111, 80131 Naples, Italy
a r t i c l e i n f o
a b s t r a c t
Article history:
CB2 receptor belongs to the large family of G-protein coupled receptors (GPCRs) controlling a wide
Received 8 November 2010
Received in revised form
14 July 2011
Accepted 14 July 2011
Available online 23 July 2011
variety of signal transduction. The recent crystallographic determination of human
and its high sequence similarity with human CB2 receptor (hCB2) prompted us to compute a theoretical
model of hCB2 based also on 2 adrenoreceptor coordinates. This model has been employed to perform
b2 adrenoreceptor
b
docking and molecular dynamic simulations on WIN-55,212-2 (CB2 agonist commonly used in binding
experiments), in order to identify the putative CB2 receptor agonist binding site, followed by molecular
docking studies on a series of indol-3-yl-tetramethylcyclopropyl ketone derivatives, a novel class of
potent CB2 agonists. Successively, docking-based Comparative Molecular Fields Analysis (CoMFA) and
Comparative Molecular Similarity Indices Analysis (CoMSIA) studies were also performed. The CoMSIA
model resulted to be the more predictive, showing r2ncv ¼ 0.96, rc2v ¼ 0.713, SEE ¼ 0.193, F ¼ 125.223, and
Keywords:
Cannabinoid receptor
CB2 agonist
Homology modeling
3D-QSAR
CoMSIA
r2
¼ 0.78. The obtained 3D-QSAR models allowed us to derive more complete guidelines for the
pred
design of new analogues with improved potency so as to synthesize new indoles showing high CB2
MD simulations
affinity.
Ó 2011 Elsevier Masson SAS. All rights reserved.
1. Introduction
neuroprotective roles, CB2 agents could also be applied in the
prevention of some neurodegenerative disorders, such as Hun-
Cannabinoid receptors interact with cannabinoid drugs
including the classical cannabinoids, such as
9- tetrahydrocan-
nabinol (
9-THC), their synthetic analogues and the endogenous
tington and Alzheimer’s diseases [13e15]. Other studies have also
highlighted potential roles for CB2 in cancer [16,17], multiple
sclerosis [18] and bone regeneration [19,20]. Since CB2 is expressed
mainly on immune tissues, selective CB2 agonists appear to be
devoid of central effects attributable to CB1 activation.
D
D
cannabinoids [1e4]. Currently, two subtypes of cannabinoid
receptors have been cloned and pharmacologically characterized,
the cannabinoid 1 receptor (CB1) and the cannabinoid 2 receptor
(CB2), even if at present their is some experimental evidence that
supports the existence of additional types of cannabinoid receptors
[5e7]. CB1 is mainly located within the central nervous system
(CNS) at presynaptic nerve terminals, while CB2 is mainly associ-
ated with immune system cells. Thus, CB2 selective compounds
had been described in the literature to be active in different
neuropathic and inflammatory pain models [8e12]. Since CB2
receptors had been found recently in CNS tissues showing some
Both CB1 and CB2 belong to the large family of G-protein
coupled receptors (GPCRs) [21] controlling a wide variety of signal
transduction. Since GPCR are membrane proteins, their expression,
purification, crystallization and structure determination present
major challenges to the discovery of new drugs. In the absence of
experimental data about human cannabinoid receptor 3D struc-
tures, computereaided GPCR-targeted drug design can be per-
formed on the basis of homology modeling techniques in tandem
with ligand-based modeling strategies, such as 3D-QSAR analyses.
Up to now, various computer generated molecular models of
ligand-CB2 complexes have been built and evaluated, some of the
latest being those developed by Salo, Tuccinardi and, more recently
by Durdagi [22]. In the present work, with the aim of gaining a better
* Corresponding author. Tel.: þ39 010 3538238; fax: þ39 010 3538358.
E-mail address: fossap@unige.it (P. Fossa).
0223-5234/$ e see front matter Ó 2011 Elsevier Masson SAS. All rights reserved.
doi:10.1016/j.ejmech.2011.07.023

4490
E. Cichero et al. / European Journal of Medicinal Chemistry 46 (2011) 4489e4505
understanding of the agonist-CB2 receptor interactions and in
particular to outline specific CB2 agonists binding mode, a theoret-
ical model of the human CB2 receptor (hCB2) has been built taking
coordinates file of the GPCR templates were obtained from the
Protein Data Bank [29].
Since CB2 receptor contains many of the conserved motifs
associated with GPCRs, the amino acid sequences of CB2 TM helices
were aligned with the corresponding residues of 2RH1, 3EML and
1F88. Due to the lack of the conserved proline in CB2 trans-
membrane domain 5, the length of this portion has been deter-
mined on the base of hydrophobicity similarity criteria, using the
sequence editor tool in MOE software [27].
into account the human
b2 adrenoreceptor [23], human A2A
adenosine receptor [24] and rhodopsin bovine coordinates [25].
This model has been refined by docking and molecular dynamic
simulations on WIN-55,212-2 (CB2 agonist commonly used in
binding experiments; pKi ¼ 8.89) in order to identify the putative
CB2 receptor agonist binding site.
Successively, in order to assess the reliability of the protein
model built as a tool for virtual screening procedures, we per-
formed docking simulations on a series of indol-3-yl-tetrame-
thylcyclopropyl ketone derivatives, a novel class of potent CB2
agonists [26] (in Fig. 1, compound 17 as representative indol-3-yl-
tetramethylcyclopropyl ketone and WIN-55,212-2 are depicted).
Furthermore, docking-based Comparative Molecular Fields Anal-
ysis (CoMFA) and Comparative Molecular Similarity Indices Anal-
ysis (CoMSIA) studies could provide a complementary tool for drug
design. Thus, on the same set of indol-3-yl-tetramethylcyclopropyl
ketone derivatives, 3D-QSAR studies were also performed. The
obtained CoMFA and CoMSIA maps, offered useful suggestions for
the design of new indol-3-yl-tetramethylcyclopropyl ketones with
improved potency and also allowed us to synthesize four new
indoles showing high CB2 affinity.
The connecting loops were constructed by the loop search
method implemented in MOE. The protein structure was mini-
mized with MOE using the AMBER94 force field [30]. The energy
minimization was carried out by the 1000 steps of steepest descent
followed by conjugate gradient minimization until the rms gradient
of the potential energy was less than 0.1 kcal molꢀ1 Åꢀ1
.
2.3. Molecular docking of WIN-55,212-2
With the aim of outline specific CB2 agonists binding mode,
docking studies had been performed according to the following
protocol. Initially, WIN-55,212-2 (ligand commonly used in binding
experiments) was submitted to docking calculation. The ligand was
docked into the putative ligand binding site using the flexible
docking module implemented in MOE. According to the studies
performed by Tuccinardi and co-workers [31], for the hCB2
receptor agonist activity, formation of H-bond contacts and
pep
2. Materials and methods
interactions with S112 and F197, are necessary. Thus, WIN-55,212-2
was docked into the putative CB2 binding site, including the resi-
dues cited above, as identified by the MOE tool SiteFinder. The
compound best-docked pose, evaluated in terms of "London dG",
has been refined by energy minimization (MMFF94) and rescored
according to "Affinity dG" (kcal/mol of total estimated binding
energy). Following this procedure, on the basis of the final docking
scoring function (S), we identify the most probable WIN-55,212-2
confomer interacting with CB2 (lowest mean S value).
To better refine the derived WIN-55,212-2/receptor complex,
a rotamer exploration of all side chains involved in the agonist
binding was carried out, by rotamer explorer module of MOE.
The derived hCB2/WIN-55,212-2 complex model was mini-
mized by CHARMM27 and submitted to molecular dynamics
simulations.
2.1. Data set
A dataset of forty-six indol-3-yl-tetramethylcyclopropyl ketones
derivatives, screened according to the same pharmacological
protocol, were selected from literature [26]. Compounds 1e46
(Table 1) and WIN-55,212-2 have been built, parameterized (Gas-
teiger-Hückel method) and energy minimized within MOE using
MMFF94 force field (the root mean square gradient has been set to
0.00001) [27].
2.2. Human CB2 receptor homology modeling
Since most of the key residues characteristic of GPCRs are
conserved in CB2 receptor, a hCB2 receptor homology model has
been generated, starting from the X-ray structures of human b2
adrenoreceptor (PDB code: 2RH1; resolution ¼ 2.40 Å), human A2A
adenosine receptor (PDB code: 3EML; resolution ¼ 2.60 Å) and
rhodopsin bovine (PDB: 1F88, resolution ¼ 2.80 Å), as GPCR
templates.
2.4. Molecular dynamics simulations
The human CB2 receptor in complex with WIN-55,212-2
obtained from docking studies has been investigated by means of
MD simulations in lipid bilayer formed by POPC molecules to
simulate the physiological context of the protein. The starting POPC
lipid bilayer system was obtained from Dr. Tieleman’s web page
http://moose.bio.ucalgary.ca/index.php?page¼Structures_and_
Topologies and is formed by 128 lipids and 2460 water, equilibrated
with 1 ns MD at 300 K [32e34]. It was extended by 2 ꢁ 2 ꢁ 1 in xyz
directions, in order to have a bilayer large enough to ensure that the
protein was completely and homogeneously surrounded by lipids.
Molecular dynamic (MD) simulations were performed using
GROMACS package version 4.5.3 [35,36], simulations were per-
formed in parallel (MPI) on a Linux Cluster of 32 nodes, each
SuperMicro equipped of two processors: the 2.50 Ghz INTEL(R)
Quad-core and the 16 GB Xeon(R), for a 256 total processors
interconnected with the Infiniband 4ꢁ network and 512 GB of
RAM. The GROMOS96 43a1 force field [37] modified in order to
include Berger’s parameters for lipids [38] was used throughout
the simulations. The topology for the ligand was created
employing the server PRODRG 2.5 Beta [39], but following
suggestions of literature data [40], charges were checked and
The amino acid sequence of hCB2 was retrieved from the
SWISSPROT database [28] while the three-dimensional structure
Fig. 1. Chemical structure and pKi value of CB2 agonists WIN-55,212-2 and 17.

E. Cichero et al. / European Journal of Medicinal Chemistry 46 (2011) 4489e4505
4491
Table 1
Molecular structure of CB2 ligands 1e46.
O
R4
R5
R2
N
R6
R1
R7
Comp
R1
R2
H
R4
F
R5
F
R6
F
R7
F
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
1
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
2
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
OH
H
H
H
F
H
H
H
H
H
H
H
H
H
H
H
H
OH
H
H
3
CI
Br
H
H
H
H
H
H
OH
H
H
H
H
4
H
5
CI
6
Br
7
CH3
CF3
SO2CH3
H
8
9
10
11
12
13
14
15
H
OH
H
CH3
H
O
CH3
H
H
O
(continued on next page)

4492
E. Cichero et al. / European Journal of Medicinal Chemistry 46 (2011) 4489e4505
Table 1 (continued )
Comp
R1
R2
R4
R5
R6
R7
H
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
-CH2
CH3
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
O
CH3
H
H
H
H
H
O
OCH2Ph
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
OCH2Ph
H
OCH2Ph
H
H
OCH2Ph
NH2
H
H
H
H
H
H
H
H
H
H
H
H
H
O(CH2)4OH
O(CH2)4Br
CN
H
H
H
CH2OH
H
CH3
-CH2
O
H
CO2CH3
H
H
H
H
H
H
CN
CN2NH2
CO2CH3

E. Cichero et al. / European Journal of Medicinal Chemistry 46 (2011) 4489e4505
4493
Table 1 (continued )
Comp
R1
R2
R4
R5
R6
R7
-CH2
CH3
CH3
34
35
36
37
38
39
40
41
42
43
44
45
H
H
OCH2Ph
OH
H
H
O
O
O
O
O
-CH2
H
H
O
-CH2
H
H
OH
OH
H
-CH2
-CH2
O
O
H
H
O
-(CH2)2
-(CH2)2
-(CH2)2
-(CH2)2
-(CH2)2
-(CH2)2
-(CH2)2
-(CH2)2
N
H
H
H
H
H
H
H
H
H
H
H
H
O
O
O
O
O
O
O
O
N
N
N
N
N
N
N
H
H
OH
H
H
H
H
OH
H
CH3
H
H
O
CH3
H
H
H
H
H
H
O
H
H
OCH2Ph
CH3
H
H
H
H
NO2
-(CH2)2
N
46
H
NH2
H
H
H
O
reassigned using Antechamber program [41] and applying the
semiempirical AM1 e bond charge correction (AM1 e BCC)
method [42,43]. The CB2 receptoreligand complex was mildly
minimized in vacuo in order to reduce steric hindrance, using
a Steepest Descent algorithm. The orientation of CB2 recep-
toreligand complex regarding to the lipid membrane was checked
based on available information in literature [44], and inserted into
POPC bilayer using the molecular graphic program VMD [45]. After
then, further 19701 water molecules were added to the system, to
cover completely the protein by the solvent, and 12 chloride ions
were added for neutralization. The coordinates of this system were
used as a starting point to insert the protein into the lipid bilayer,
with the aid of the newly developed g_membed computational tool
[46]. G_membed first decreases the width of the protein in the xey
plane and remove all molecules (lipids and water) that overlap
with the narrowed protein. In our case, only 22 POPC molecules
were removed. After that, 1000 MD steps were applied to re-
growth the protein to its original size. In order to perform this
process, the position of protein þ ligand group in the membrane
was blocked and the internal protein þ ligand interactions were
excluded. A cut-off of 1.4 nm was applied to van der Waals
interactions and the electrostatic interactions were treated using
PME [47] with a real space cut-off of 1.0 nm. All bonds were
constraint using the LINCS algorithm [48] allowing a time step of
2 fs. The temperature and pressure were kept constant at 310 K
and 1 bar using velocity rescaling [49] and Berendsen semi-
isotropic pressure coupling [50]. Periodic boundary conditions
were used to exclude surface effects. All these parameters for
g_membed protocol were kindly provided by Dr. M.G. Wolf
(personal communication). After that, the protein
þ
ligand
complex embedded in the membrane was submitted to 1 ns MD
simulation to equilibrate the system. The settings of these MD

4494
E. Cichero et al. / European Journal of Medicinal Chemistry 46 (2011) 4489e4505
simulations were the same as previously, with the exception that
the protein was allowed to move and that the internal
protein þ ligand interactions were restored. To check for the
correct equilibration of the system, we followed the stabilization
of the energy of the system, and of the box vectors to ensure that
the membrane has a stable lateral area. Finally, the production MD
simulation was carried out with the same settings as the equili-
bration MD, for further 20 ns.
Next, several analyses were conducted using programs built
within the GROMACS package, and results were visualized and
elaborated with the aid of the freeware program Grace (http://
plasmagate.weizmann.ac.il/Grace). The energy components,
temperature and pressure, and the box vectors were analyzed to
confirm the stabilization of the system. The equilibration of the
system was reached after 5 ns of simulation, therefore all analyses
were made starting from 5 ns. Visualization and analysis of model
features was carried out using VMD [45] and Pymol [51] programs.
To obtain a representative structure for the trajectory, we per-
formed a cluster analysis using GROMACS utilities, with a cut-off of
0.1 nm for RMSD calculation after partial least-square fitting. A
single cluster was obtained for the whole trajectory. An average
structure was then calculated and minimized using the same
protocol described above. The resulting minimized structure of the
complex between protein and WIN-55,212-2 was used as repre-
sentative structure for the following docking study.
2.6.2. CoMFA and CoMSIA interaction energies
CoMFA method [52] is a widely used 3D-QSAR technique to relate
the biological activity of a series of molecules to their steric and
electrostatic fields, which are calculated placing the aligned mole-
cules, one by one, into a 3D cubic lattice with a 2 Å grid spacing. The
van der Waals potential and Coulombic terms, which represent steric
and electrostatic fields, respectively, were calculated using the stan-
dard Tripos force field method. The column-filtering threshold value
was set to 2.0 kcal/mol to improve the signal-noise ratio. A methyl
probe with a þ1 charge was used to calculate the CoMFA steric and
electrostatic fields. A 30 kcal/mol energy cut-off was applied to avoid
infinity of energy values inside the molecule. The CoMSIA method
[53] calculates five descriptors, namely steric, electrostatic and
hydrophobic parameters, and the H-bond donor and H-bond
acceptor properties. The similarity index descriptors were calculated
using the same lattice box employed for the CoMFA calculations and
a sp3 carbon as probe atom with a þ1 charge, þ1 hydrophobicity
and þ1 H-bond donor and þ1 H-bond acceptor properties.
2.6.3. Partial least-square (PLS) analysis and models validation
The partial least-squares (PLS) approach, an extension of the
multiple regression analysis, was used to derive the 3D-QSAR
models. CoMFA and CoMSIA descriptors were used as independent
variables and pEC50 values were used as dependent variables. Prior
to the PLS analysis, CoMFA and CoMSIA columns with a variance of
less than 2.0 kcal molꢀ1 were filtered by using column-filtering to
improve the signal-to-noise ratio.
The leave one out (LOO) cross-validation method was used to
check the predictivity of the derived model and to identify the
optimal number of components (ONC) leading to the highest cross-
validated r2 (r2cv). In the LOO methodology, one molecule is omitted
from the dataset and a model is derived involving the rest of the
compounds. Employing this model, the activity of the omitted
molecule is then predicted.
The ONC obtained from cross-validation methodology was used
in the subsequent regression model. Final CoMFA and CoMSIA
models were generated using non-cross-validated PLS analysis. To
further assess the statistical confidence and robustness of the
derived models, a 100-cycle bootstrap analysis was performed. This
is a procedure in which n random selections out of the original set
of n objects are performed several times (100-times were required
to obtain a good statistical information). In each run, some objects
may not be included in the PLS analysis, whereas some others
might be included more then once. The mean correlation coeffi-
cient is represented as bootstrap r2(r2boot).
The stereochemistry of the model has been validated through
the analysis of Ramachandran plot, by means of Procheck. In order
to verify the reliability of the derived receptor model, it has been
superimposed and compared to the 2RH1 coordinates (being the
template with the higher sequence similarity).
2.5. Molecular docking of compounds 1e46
The final hCB2/WIN-55,212-2 model (obtained by the MD
simulations) has been used as starting point for the automatic
docking analysis of compounds 1e46. Thus, each compound was
docked into the putative CB2 binding site using the flexible docking
module implemented in MOE, and refined by minimization with
CHARMM27, following the same procedure used for WIN-55,212-2
docking analysis, as described above. Since both the protein model
built and the docking protocol applied proved to be reliable (by
properly explaining why compounds 1e46 are more or less active
as hCB2agonists), we performed the following 3D-QSAR analyses
on the basis of the compound selected docking poses, which
resulted to be aligned into the hCB2 putative binding site.
2.6. 3D-QSAR analyses
2.6.4. Predictive correlation coefficient (r2
)
pred
To further validate the CoMFA and CoMSIA derived model, the
In tandem with structure-based drug design, the docking-based
3D-QSAR approach of CoMFA [52] and CoMSIA analyses [53] by
SybyX 1.0 software [54], could provide a complementary tool for
drug design. Thus, the compound selected docking poses, aligned
into the hCB2 putative binding site, were submitted to 3D-QSAR
studies.
predictive ability for the test set of compounds (expressed as r2
was determined by using the following equation:
)
pred
rp2red ¼ ðSD ꢀ PRESSÞ=SD
SD is the sum of the squared deviations between the biological
activities of the test set molecules and the mean activity of the
training set compounds and PRESS is the sum of the squared
deviation between the observed and the predicted activities of the
test set compounds.
2.6.1. Training set and test set
All the compounds were grouped into a training set, for model
generation, and a test set, for model validation, containing 37 and 9
compounds respectively. The molecules of the test set represent the
24% (estimated as a good percentage to validate a molecular model)
of the training set. Both the training and the test set were divided
manually according to a representative range of biological activities
and structural variations. For QSAR analysis, Ki values have been
transformed into pKi values and then used as response variables.
Compounds receptor affinity covered 4 log orders.
All calculations were carried out using a PC running the
Windows XP operating system and an SGI O2 Silicon Graphics.
2.7. Chemistry
The newly synthesized compounds 48aed (Table 2) were
obtained starting from indoles 47a-b by alkylation, employing KOH

E. Cichero et al. / European Journal of Medicinal Chemistry 46 (2011) 4489e4505
4495
Table 2
2 ꢁ NeCH2), 2.63 (t, J ¼ 6.9 Hz, 2H, NeCH2), 4.45 (t, J ¼ 6.9 Hz, 2H,
NeCH2), 7.30e7.42 (m, 2H, AreH), 7.67e7.84 (m, 4H, AreH),
8.03e8.15 (m, 3H, AreH), 8.57 (s, 1H, indole H-2), 8.67 (s, 1H, CH]
C). Anal. Calcd for C24H25N3O2S: C, 68.71; H, 6.01; N, 10.02; S, 7.64.
Found: C, 68.92; H, 5.97; N, 9.97; S, 7.74.
Newly synthesized indoles 48ae48d.
Y
X
R
N
2.7.4. (E)-ethyl 2-(phenylcarbonyl)-3-(1-(2-(piperidin-1-yl)ethyl)-
1H-indol-3-yl)prop-2-enoate (48b)
R1
Yield: 57%; m.p.: 135e138 ꢂC (acetone); IR (KBr): 2943e2762
(CH2, CH3), 1705 (COOEt), 1667 (COPh); 1605 (C]C) cmꢀ1. 1H NMR
X
Y
R
R1
Yield
55
(DMSO-d6)
d
: 1.12 (t,J ¼ 7.3 Hz, 3H, CH3), 1.20e1.34 (m, 6H, 3 ꢁ CH2),
2.10e2.29 (m, 4H, 2 ꢁ NeCH2), 2.41 (t, J ¼ 7.0 Hz, 2H, NeCH2), 4.15 (q,
J ¼ 7.3 Hz, 2H, CH2), 4.35 (t, J ¼ 7.0 Hz, 2H, NeCH2), 7.18e7.32 (m, 3H,
AreH), 7.42e7.60 (m, 4H, AreH), 7.82e7.90 (m,1H, AreH) 7.91e7.99
(m, 2H, AreH), 8.19 (s, 1H, CH]C), Anal. Calcd for C27H30N2O3: C,
75.32; H, 7.02; N, 6.51. Found: C, 75.64; H, 7.11; N, 6.11.
48a
48b
CN
SO2C6H5
H
N
N
N
N
COPh
COOEt
H
57
2.7.5. (E)-3-(1-(2-morpholinoethyl)-1H-indol-3-yl)-2-
(phenylsulfonyl)prop-2-enenitrile (48c)
O
O
48c
48d
CN
SO2C6H5
COOEt
H
H
40
40
Yield: 40%; m.p.: 143e145 ꢂC (ethyl acetate); IR (KBr):
2970e2800 (CH2,CH3), 2199 (CN), 1561 (C]C), 1332 (SO2), 1145
(SO2) cmꢀ1. 1H NMR (DMSO-d6)
d
: 2.37e2.47 (m, 4H, 2 ꢁ NeCH2),
2.68 (t, J ¼ 6.9 Hz, 2H, NeCH2), 3.50e3.60 (m, 4H, 2 ꢁ OeCH2), 4.47
(t, J ¼ 6.9 Hz, 2H, NeCH2), 7.31e7.42 (m, 2H, AreH), 7.68e7.83 (m,
4H, AreH), 8.00e8.11 (m, 3H, AreH), 8.57 (s, 1H, indole H-2), 8.70
(s, 1H, CH]C). Anal. Calcd for C23H23N3O3S: C, 65.54; H, 5.50; N,
9.97; S, 7.61. Found: C, 65.36; H, 5.70; N, 10.06; S, 7.32.
COPh
as base, Aliquat as phase transfer catalyst and chloroalkylamines as
alkylating agents (Scheme 1). The synthesis of compounds 47a-
b was performed starting from indole via iminium salt formation
and condensation according to the procedure already described
[55].
2.7.6. (E)-ethyl 3-(1-(2-morpholinoethyl)-1H-indol-3-yl)-2-(phen-
ylcarbonyl)prop-2-enoate (48d)
Yield: 40%; m.p.: 145e147 ꢂC (ethyl acetate); IR (KBr):
2970e2795 (CH2, CH3), 1704 (COOEt), 1667 (COPh); 1616 (C]C)
cmꢀ1
.
1H NMR (DMSO-d6)
d
: 1.12 (t,J ¼ 7.0 Hz, 3H, CH3), 2.46 (t,
2.7.1. (2E)-3-(1H-indol-3-yl)-2-(phenylsulfonyl) acrylonitrile (47a)
Yield: 71%; m.p.: 214e216 ꢂC (ethyl acetate). IR (KBr): 3352 (NH),
2210 (CN),1567 (C]C) 1305 (SO2),1140 (SO2) cmꢀ1. 1H NMR (CDCl3)
J ¼ 7.1 Hz, 2H, NeCH2), 3.26e3.35 (m, 4H, 2 ꢁ OeCH2), 4.14 (q,
J ¼ 7.0 Hz, 2H, CH2), 4.25 (t, J ¼ 7.1 Hz, 2H, NeCH2), 7.17e7.30 (m, 2H,
AreH), 7.40e7.69 (m, 5H, AreH), 7.76e7.83 (m,1H, AreH) 7.88e7.95
(m, 2H, AreH), 8.19 (s, 1H, CH]C), Anal. Calcd for C26H28N2O4: C,
72.20; H, 6.53; N, 6.48. Found: C, 71.85; H, 6.90; N, 6.35.
d
: 7.26e7.30 (m, 2H, AreH), 7.40e7.60 (m, 4H, AreH), 7.78e7.83 (m,
1H, AreH) 7.94e7.99 (m, 2H, AreH), 8.41 (d, J ¼ 3.0 Hz, 1H, indole
H-2), 8.47 (s, 1H, CH]C), 9.67 (br s, 1H, NH). Anal. Calcd for
C17H12N2O2S: C, 66.22; H, 3.92; N, 9.08; S, 10.40. Found: C, 65.95; H,
3.67; N, 9.07; S, 10.02.
2.8. Compound 48aed biological activity evaluation
Biological activity of aminoalkylindoles was evaluated in terms
of percentage inhibition and Ki. In both cases, competitive binding
assays were performed using membranes from HEK-293 cells
transfected with the human recombinant CB1 receptor
(Bmax ¼ 2.5 pmol/mg protein) and human recombinant CB2
receptor (Bmax ¼ 4.7 pmol/mg protein). Receptors were incubated
with [3H]-CP-55,940 (0.14 nM/kd ¼ 0.18 nM and 0.084 nM/
kd ¼ 0.31 nM respectively for CB1 and CB2 receptor) as the high
2.7.2. Ethyl (2E)-2-benzoyl-3-(1H-indol-3-yl)acrylate (47b)
Yield: 38%; m.p.: 160e163 ꢂC (diethyl ether); IR (CHCl3): 3465
(NH), 2939 (CH2CH3), 1705 (COOEt), 1667 (COPh); 1609 (C]C)
cmꢀ1
.
1H NMR (CDCl3)
d: 1.09 (t,J ¼ 7.1 Hz, 3H, CH3), 4.15 (q,
J ¼ 7.1 Hz, 2H, CH2), 7.15e7.38 (m, 6H, AreH), 7.42e7.52 (m, 1H,
AreH), 7.75e7.80 (m, 1H, AreH) 7.89e7.94 (m, 2H, AeH), 8.26 (s,
1H, CH]C), 8.71 (br s, 1H, NH). Anal. Calcd for C20H17NO3: C, 75.22;
H, 5.37; N, 4.39. Found: C, 75.07; H, 5.69; N, 4.35.
affinity ligand and displaced with 10
mM WIN 55212-2 as the
heterologous competitor for non specific binding (Ki values 9.2 nM
and 2.1 nM respectively for CB1 and CB2 receptor). All compounds
were tested following the procedure described by the manufacturer
(Perkin Elmer, Italia). Displacement curves were generated by
incubating drugs with [3H]-CP-55,940 for 90 min at 30 ꢂC. Ki values
were calculated by applying the ChengePrusoff equation to the IC50
values (obtained by GraphPad) for the displacement of the bound
radioligand by increasing concentrations of the test compound.
2.7.3. (E)-2-(phenylsulfonyl)-3-(1-(2-(piperidin-1-yl)ethyl)-1H-
indol-3-yl)prop-2-enenitrile(48a)
Yield: 55%; m.p.: 155e156 ꢂC (ethanol); IR (KBr): 2932e2770
(CH2,CH3), 2202 (CN), 1566 (C]C), 1340 (SO2), 1153 (SO2) cmꢀ1. 1H
NMR (DMSO-d6)
d: 1.28e1.55 (m, 6H, 3 ꢁ CH2), 2.30e2.48 (m, 4H,
Y
Y
X
X
KOHaq
R1-Cl Aliquat
DCM
3. Results and discussion
R2
R2
N
H
N
R1
3.1. Human CB2 receptor homology modeling
47a
47b
48a,b
48c,d
Since most of the key residues characteristic of GPCRs are
Scheme 1. Synthesis of the compounds 48aed.
conserved in CB2 receptor, a hCB2 receptor homology model has

4496
E. Cichero et al. / European Journal of Medicinal Chemistry 46 (2011) 4489e4505
As shown in Fig. 3, the following CB2 residues are conserved
among the three GPCRs: (i) L42, N51 and L64 residues in TM1, (ii)
Y70, L76, A79, D80, L82 in TM2, (iii) L124, A128, the (E/D)RY motif
(130e133) in TM3, (iv) the A150, W158 and W172 residues in TM4,
(v) L213 in TM5, (vi) the CWXP motif (257-260 amino acids) in
TM6, (vii) the NPXIY sequence (295e299 residues) and R307 in
TM7. In addition, between the target protein and the b2 adrenor-
eceptor (which is the most similar template to the hCB2 receptor
sequence) the following residues are also conserved : (i) the L45,
V52, V54 and I58 residues in TM1, (ii) F72, I73, S75 and A77 in
TM2, (iii) F106, V113, T118, A119 and S120 in TM3, (iv) V152, S161,
L163, S165 and L167 in TM4, (v) S193, F197 and L201 in TM5, (vi)
K245, T246, L247 and G220 in TM6, (vii) N291, S292 and R302 in
TM7.
Fig. 2. Pairwise percentage residues identity values calculated evaluated between the
four GPCRs.
been generated, starting from the X-ray structures of human b2
adrenoreceptor (PDB code: 2RH1), human A2A adenosine receptor
(PDB code: 3EML) and rhodopsin bovine (PDB: 1F88), as GPCR
templates.
The alignment of the CB2 (P34972) fasta sequence on the X-ray
coordinates of human b2 adrenoreceptor [23], human A2A adeno-
sine receptor [24] and rhodopsin bovine [25] was performed on the
basis of the Blosum62 matrix (MOE software). The reliability of the
four GPCRs alignment can be verified by high values of the pairwise
percentage residue identity (PPRI) reported in Fig. 2.
As concern the alignment of the hCB2 (P34972) amino acidic
sequence on the X-ray coordinates of 2RH1, 3EML and 1F88, the
higher PPRI values were the one towards 2RH1 (24.6) and towards
1F88 (20.1), in comparison with that on 3EML (16.3). Thus, the hCB2
model was derived by means of the X-ray coordinates of 2RH1
(used as template for hCB2 sequence residues 29e227 and
237e316) and 1F88 (used as template for hCB2 sequence residues
228e236), aligned together on the basis of the GPCRs conserved
residues. The hCB2 fragments corresponding to the sequence resi-
dues 1e28 and 317e360 were not taken into account, since they are
not involved in the agonistereceptor interactions.
3.2. Molecular docking of WIN-55,212-2
The obtained CB2 receptor model has been employed to
perform a docking procedure on WIN-55,212-2 (CB2 agonist
commonly used in binding experiments) in order to analyse the
putative CB2 receptor agonist binding site. In agreement with
site-directed mutagenesis data and with the computational
results available in literature [31], the agonist selected docking
pose occupies a lipophilic pocket delimited by TM3, TM5 and
TM6.
In detail (Fig. 4), the oxygen atom of the morpholinic group
displays a H-bond interaction with the S112 side chain while the
oxygen at the position 7 of the indole moiety shows a H-bond with
the N188 backbone. The 1-naphthyl substituent establishes Van der
Waals interactions and
and L262.
pep stacking with W194, L196, F197, I198
Fig. 3. Sequence alignment of the human CB2 receptor on the basis of the human b2 adrenoreceptor (2RH1) and the bovine rhodopsin (1F88) coordinates.

E. Cichero et al. / European Journal of Medicinal Chemistry 46 (2011) 4489e4505
4497
Fig. 4. Selected docking pose of WIN-55,212-2 into the putative CB2 (obtained by site-
directed docking procedure, followed by complex refinement) agonist binding site. H-
bond interactions are depicted in green. Residues involved in hydrophobic contacts are
labelled. (For interpretation of the references to colour in this figure legend, the reader
is referred to the web version of this article.)
Fig. 5. Graphic of RMS fluctuation of residues in hCB2 during the MD simulation
(excluding the first 5 ns of equilibration). Residues involved in interactions with the
ligand WIN-55,212-2 are indicated and labelled.
The derived hCB2/agonist complex model has been refined by
molecular dynamic simulations.
distances between S112 and atom OBE and between S285 and NAB
are stable around 3 Å from 6 to 20 ns. Instead, from MD data, the H-
bond between N188 residue with the oxygen of the naphthyl
methanone moiety (atom NAB) seems to be less stable since the
distance between the donor and the acceptor atom is variable
between 2.5 and 11.5 A during the simulation.
3.3. Molecular dynamic simulations
In order to analyse how the motion of the protein (especially of
the side chains in the binding pocket) can influence the binding of
the ligand, molecular dynamic (MD) simulations in explicit POPC
membrane at physiological temperature (310 K) were performed
on the complex between hCB2 and agonist WIN-55,212-2 starting
from the conformation obtained by docking simulations. The final
trajectory of the complex was energetically stable, with average
temperature and pressure close to values of 310 K and 1 bar. The
RMSD analysis confirms that the complex reaches an equilibrium
after about 5 ns of simulation (see Supplementary Material).
The analysis of root mean square fluctuations (RMSF) can be
used as a reference to evaluate motility of the residues of the
protein. Here we found that the main fluctuations (>0.2 nm) are
present on residues belonging to mobile loops of the protein and on
the N-and C-terminal part of the protein, far from the ligand
binding pocket. Notably, residues that were found to interact with
the agonist in docking simulations, are among the most stable
residues of the protein (RMSF < 0.15 nm) (Fig. 5). Also, the ligand
shows only limited fluctuations (<0.1 nm) during the simulation.
This is crucial, because several important interactions between the
protein and the ligand are made via H-bonds, and the strength of
this kind of bond is determined by the geometry of the interaction,
which is strongly affected by mobility of the interactors. In this case,
the fluctuations are very low, and it is possible to hypothesize that
the interactions between protein and ligand are very stable during
the time.
In order to analyse the variability among different frames of the
simulation, we carried out a cluster analysis on the trajectory. The
calculation using a cut-off of 0.1 nm for the RMSD after partial least-
square fit, calculated on the structure of the protein, produced only
a single cluster for the whole trajectory. The same result was
obtained using as a reference the structure of the ligand.
These results indicate the high similarity between the structures
in the simulation and suggest that no important differences are
present among the several structures in terms of ligand motility, as
it was also suggested by the low RMSF value.
Finally, we compared the representative structure of the
complex obtained from the trajectory with the one derived from
docking simulation.
The comparison between the docking and molecular dynamics
studies shows a different binding mode of the agonist in the
putative pocket. The hydrogen bonded pattern is conserved for
residues S112 and N188 in both simulations; molecular dynamics
shows an additional hydrogen bond involving the S285 (Fig. 7).
Therefore, the molecular portrait obtained by MD simulation is
that of a protein in which the ligand is bound with a high confor-
mational stability, and in which H-bonds and hydrophobic inter-
actions keep the molecule strictly associated to the protein,
confirming the hypotheses made on the basis of docking
experiments.
The final model backbone conformation was inspected by
Ramachandran plot, showing absence of outliers among the resi-
dues belonging to the putative binding site (see Fig. S1 in the
supplementary data). As depicted in Fig. 8, the derived hCB2 model
was superimposed to the coordinates of the human b2 adrenor-
eceptor, which is the template showing the higher similarity to the
target protein [compare the hCB2 PPRI value towards 2RH1 (24.6)
with those on 1F88 (20.1)], displaying a quite positive root mean
square deviation value (RMSD ¼ 1.113 Å, calculated on the carbon
atom alignment).
To support this hypothesis, we analyzed the presence and
persistence of H-bonds between WIN- 55,212-2 and hCB2 using
Gromacs tools. We detected the presence of a maximum of 3 H-
bonds with an average number of H-bonds per frame equal to 1.28.
The residues mainly involved in these bonds are S112, which
interacts with the oxygen of the morpholine ring moiety (atom
OBE) and S285, bound to the methyl morpholine ring moiety (atom
NAB). We also measured the variation of the distances between
these atoms during the simulation (Fig. 6), and found that the

4498
E. Cichero et al. / European Journal of Medicinal Chemistry 46 (2011) 4489e4505
Fig. 6. Variation of distances of heavy atoms (donor and acceptor) involved in H-bonds between hCB2 and WIN-55,212-2 (excluding the first 5 ns of equilibration). Distance
between side chain oxygen of S112 and morpholine ring moiety (atom OBE) (a); distance between side chain oxygen of S285 and the methyl morpholine ring moiety (atom NAB)
(b); distance between amide nitrogen of N188 and oxygen of the naphthyl methanone moiety (atom NAB) (c).
As shown in Fig. 9, the agonist WIN-55,212 is surrounded by the
following conserved residues: (i) V113 in TM3, (ii) S193 and F197 in
TM5, (iii) W258 in TM6.
Notably, the F197 and W258 side chain are the only conserved
residues which resulted to be rotated, in comparison with those of
the b2 adrenoreceptor, being more in proximity and slightly far
from the agonist, respectively. Consequently, according to our
studies, the geometry of the derived hCB2 model binding pocket
seems to be particularly affected by the establishment of hydro-
phobic contacts, such as those with F197.
be a more or less active hCB2agonist), we performed docking
simulations on a series of known potent hCB2 agonists.
Thus, the derived WIN-55,212-2/receptor complex has been
used as starting point for the automatic docking simulation on
a series of indol-3-yl-tetramethylcyclopropyl ketone derivatives.
According to our calculations all the indol-3-yl-tetramethylcy-
clopropyl ketone derivatives display a H-bond interaction with S112
side chain by the oxygen atom of the carbonyl group and share the
following Van der Waals contacts between: (i) the R1 substituent
and L108, V113, L169, (ii) the tetramethylcyclopropyl ketone
portion and the F87, F117 and C288 side chain, (iii) the agonist
indole moiety and L108, and M265.
More in details, as shown in Fig. 10, 17 (the most active
compound of the series) is engaged in two additional H-bond
contacts between the methoxy group at the indole position 7 and
the N188 side chain, and between the pyrane oxygen atom and the
T114 nitrogen atom.
3.4. Molecular docking of compounds 1e46
In order to verify if the protein model built (and the docking
protocol previously applied) could be employed as a reliable tool for
virtual screening procedures (by explaining why a compound could
On the contrary, compound 13, bearing a hydroxyl group at the
indole position 7, displays H-bond contact between the R7
substituent and the P184 carbonyl oxygen atom, while any inter-
actions concerning the pyrane oxygen atom are missing.
Similarly, compounds 14, 15, 16, bearing the methoxy group at
the indole position 4, 5 and 6, respectively, are also unable to
display H-bonds with T114. On the other hand, the compound 14, 15
methoxy groups, and the one of compound 16, are involved in H-
bond interactions with the L182 and the L185 nitrogen atom,
respectively.
Accordingly, compound 13 (pIC50 ¼ 8.50) and compound 14e16
(pIC50 ¼ 8.34e9.29) show lower pIC50 values in comparison with
that of compound 17 (pIC50 ¼ 9.90).
Compounds 10 (pIC50
¼
8.41), 11 (pIC50
¼
8.72), 12
(pIC50 ¼ 8.07), bearing the hydroxyl moiety at the indole position 4,
5 and 6, miss any contacts between the pyrane ring and T114, but
compound 10 and 11 seem to be engaged in H-bonds with the S285
side chain, being more potent than compound 12.
Compound 41 (indole 15 analogue) and compound 42 (indole 16
analogue), bearing an ethyl-morpholine substituent at the indole
position 1 (instead of a methyl-pyrane one), are not able to properly
interact with the L182 and L185 nitrogen atom, respectively.
Fig. 7. The WIN-55,212-2 poses (green) and (cyan) obtained by docking and molecular
dynamics simulations respectively. Only residues involved in the hydrogen bond-
interactions are depicted in stick and labelled. (For interpretation of the references
to colour in this figure legend, the reader is referred to the web version of this article).

E. Cichero et al. / European Journal of Medicinal Chemistry 46 (2011) 4489e4505
4499
Fig. 8. The superimposition of the final hCB2 model on the 2RH1 coordinates is depicted as side-view (a) and as top-view (b). The WIN-55,212-2 pose is depicted by space filling
curve.
In fact, compound 41 misses any H-bond interaction with L182
while, as shown in Fig. 11, 42 displays an H-bond contact with the
N188 backbone nitrogen atom.
(ONC ¼ 6) to give a non-cross validated r2 (r2ncv) ¼ 0.94, a test set r2
(r2pred) ¼ 0.71, Standard Error of Estimate (SEE) ¼ 0.245, steric
contribution ¼ 0.461 and electrostatic contribution ¼ 0.539. The
model reliability thus generated was supported by bootstrapping
results. All statistical parameters supporting CoMFA model are
reported in Table 3.
Accordingly, compounds 41 (pIC50 ¼ 6.63) and 42 (pIC50 ¼ 8.80)
show lower pIC50 values in comparison with those of compound 15
(pIC50 ¼ 8.34) and compound 16 (pIC50 ¼ 9.29).
For the indole selected docking poses, scoring function values
have been reported in Table S2 (supplementary data).
The final CoMSIA model consisting of steric, electrostatic,
hydrophobic, H-bond acceptor and H-bond donor fields with
a r2 ¼ 0.96, r2
¼ 0.78, SEE ¼ 0.193, steric contribution ¼ 0.110,
ncv
pred
electrostatic contribution ¼ 0.231, hydrophobic contribution ¼ 0.159,
3.5. CoMFA and CoMSIA analyses
H-bond acceptor contribution
¼
0.241 and H-bond donor
contribution ¼ 0.259 was derived. All statistical parameters sup-
The compound selected docking poses, aligned into the hCB2
putative binding site, were submitted to 3D-QSAR studies.
porting CoMSIA model are reported in Table 3.
Experimental and predicted binding affinities values for the
training set and test set are reported in Table 4, while distribution of
experimental and predicted pKi values for training set according to
the final CoMFA and CoMSIA models are represented in Fig. S3
(supplementary data).
On the basis that CoMFA and CoMSIA field effects on the target
properties can be viewed as 3D coefficient contour plots, identi-
fying important regions where any change in these fields may affect
the compound affinity, they could be helpful to optimize indol-3-
yl-tetramethylcyclopropyl ketones as CB2 agonists. The 3D-QSAR
CoMFA and CoMSIA analyses were performed dividing
compounds 1e46 into a training set (2e14, 16e23, 25e31, 33, 34,
36, 38, 40, 42, 43, 45, 46) for model generation and into a test set (1,
15, 24, 32, 35, 37, 39, 41, 44) for model validation. CoMFA and
CoMSIA studies were developed using, respectively, CoMFA steric
and electrostatic fields and CoMSIA steric, electrostatic, hydro-
phobic, H-bond acceptor and H-bond donor properties, as inde-
pendent variables, and the ligand pKi as dependent variable.
The final CoMFA model was generated employing non-cross-
validated PLS analysis with the optimum number of components
Fig. 9. The superimposition of the final hCB2 model on the 2RH1 coordinates is depicted: (a) the two proteins are shown as side-view and the WIN-55,212-2 pose is depicted by
space filling curve. (b) the two proteins are shown as top-view and the WIN-55,212-2 pose is depicted by ball and stick. The conserved residues located 5 Å from the agonist are
shown and labelled.

4500
E. Cichero et al. / European Journal of Medicinal Chemistry 46 (2011) 4489e4505
Table 3
Summary of CoMFA and CoMSIA results.
CoMFA Model
CoMSIA Model
No. compounds
37
6
37
6
Optimal number of components
(ONC)
Leave one out r2 (r2
)
0.562
0.678
0.245
0.94
76.178
0.461
0.539
e
0.602
0.713
0.193
0.96
125.223
0.110
0.231
0.159
0.241
0.259
0.98
loo
Cross-validated r2 (r2
)
cv
Std. error of estimate (SEE)
Non cross-validated r2 (r2
)
ncv
F value
Steric contribution
Electrostatic contribution
Hydrophobic contribution
H-bond acceptor contribution
H-bond donor contribution
e
e
Bootstrap r2 (r2
)
0.97
0.169
boot
Standard error of estimate r2
0.149
boot
(SEE r2
)
boot
Test set r2 (r2
)
0.71
0.78
pred
Fig. 10. The compound 17 selected docking into the hCB2 putative binding site is
depicted. The agonist is reported in ball and stick, C atom in dark green. Residues
located 5 Å from the agonist are shown and labelled. H-bond are coloured in cyano.
(For interpretation of the references to colour in this figure legend, the reader is
referred to the web version of this article.)
Table 4
Experimental and predicted pKi values of compounds 1e46.
Compound
CoMFA Model
Pred. pKi
CoMSIA Model
Exp. pKi
Pred. pKi
Pred. pKi
Residual
analysis maps are described and discussed in the following
sections.
1a
2
3
4
5
6
7
8
8.31
8.80
8.52
8.18
8.60
9.19
9.83
8.46
8.53
8.41
8.72
8.07
8.50
8.48
8.34
9.29
9.90
8.03
8.90
9.05
8.52
8.03
6.58
7.14
7.52
6.95
7.59
6.62
7.23
8.30
7.21
8.86
9.39
8.75
9.16
7.67
8.99
8.36
7.04
6.96
6.63
8.80
8.53
7.57
6.53
7.49
8.41
8.61
8.46
8.32
8.52
8.79
9.28
8.89
8.59
8.91
8.46
7.81
8.43
8.24
8.04
9.67
9.86
8.08
8.69
9.27
8.44
8.16
6.62
7.37
7.73
7.04
7.55
6.56
7.14
8.45
7.53
7.81
9.32
8.81
8.77
7.94
8.36
8.53
7.29
6.85
7.09
8.72
8.51
7.72
6.49
7.46
ꢀ0.10
0.19
0.06
7.93
8.57
8.46
8.60
8.90
8.97
9.33
8.69
8.73
8.68
8.87
7.97
8.50
8.75
8.63
9.31
9.51
8.09
8.62
9.16
8.99
8.05
6.38
7.59
7.46
7.02
7.68
6.47
7.39
8.52
6.93
8.21
9.41
8.84
8.87
7.56
8.67
8.40
7.50
7.20
7.15
8.54
8.39
7.87
6.74
7.33
0.38
0.23
0.06
ꢀ0.42
ꢀ0.30
0.22
3.5.1. CoMFA steric and electrostatic regions
ꢀ0.14
As shown in Fig. 12a (for simplicity, only the structure of
compound 17, showing the highest pKi value in the dataset, is
depicted and used as a model) the steric contour map predicts
favourable interaction polyhedra (green) for positions 4, 6 and 7 of
the indole moiety, around the tetramethylcyclopropyl group and in
proximity of the pyrane ring. The introduction of bulky group
around position 5 of the indole scaffold is disfavoured (yellow). The
reliability of the steric map calculations is verified by the higher pKi
value of 17 (R7 ¼ OCH3, pKi ¼ 9.90) in comparison with that of 13
(R7 ¼ OH, pKi ¼ 8.50) and by the higher affinity of 6 (R6 ¼ bromo,
pKi ¼ 9.19), 16 (R6 ¼ OCH3, pKi ¼ 9.29), 35 (R5 ¼ OH, R6 ¼ OCH3,
0.08
0.40
0.55
ꢀ0.43
ꢀ0.06
ꢀ0.49
0.26
0.26
0.07
0.24
0.30
0.50
ꢀ0.23
ꢀ0.20
ꢀ0.27
ꢀ0.15
0.10
9
10
11
12
13
14
15a
16
17
18
19
20
21
22
23
24a
25
26
27
28
29
30
31
32a
33
34
35a
36
37a
38
39a
40
41a
42
43
44a
45
46
0.00
ꢀ0.27
ꢀ0.29
ꢀ0.02
0.39
ꢀ0.38
0.04
ꢀ0.04
0.21
ꢀ0.06
0.28
ꢀ0.22
ꢀ0.11
ꢀ0.47
ꢀ0.02
0.21
0.08
ꢀ0.13
ꢀ0.04
ꢀ0.23
ꢀ0.21
ꢀ0.09
0.04
0.06
0.09
ꢀ0.15
ꢀ0.32
1.05
ꢀ0.45
0.06
ꢀ0.07
ꢀ0.09
0.15
ꢀ0.16
ꢀ0.22
0.28
0.65
0.07
ꢀ0.02
ꢀ0.09
0.30
0.11
0.33
ꢀ0.04
ꢀ0.46
ꢀ0.24
ꢀ0.52
0.26
ꢀ0.06
0.39
ꢀ0.27
0.64
ꢀ0.17
ꢀ0.25
0.11
ꢀ0.46
0.08
0.02
0.14
Fig. 11. The compound 41 selected docking into the hCB2 putative binding site is
depicted. The agonist is reported in ball and stick, C atom in orange. Residues located
5 Å from the agonist are shown and labelled. H-bond are coloured in red. (For inter-
pretation of the references to colour in this figure legend, the reader is referred to the
web version of this article.)
ꢀ0.15
ꢀ0.30
ꢀ0.21
0.16
0.04
0.03
a
Test set compounds.

E. Cichero et al. / European Journal of Medicinal Chemistry 46 (2011) 4489e4505
4501
Fig. 12. Contour maps of CoMFA steric regions (green, favoured; yellow, disfavoured) are displayed around compound 17, depicted in stick mode and coloured by atom type (a).
Contour maps of CoMFA electrostatic regions are shown around compound 17. Blue regions are favourable for more positively charged groups; red regions are favourable for less
positively charged groups (b). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
pKi ¼ 9.16) and 42 (R6 ¼ OCH3, pKi ¼ 8.80) compared to those of 5
(R6 ¼ chloro, pKi ¼ 8.60), 12 (R6 ¼ OH, pKi ¼ 8.07), 36 (R5 ¼ OH,
R6 ¼ OH, pKi ¼ 7.67) and 40 (R6 ¼ OH, pKi ¼ 6.96), respectively.
Moreover, the results are in agreement with the high binding
affinity of 33 (R6 ¼ phenyl, pKi ¼ 9.39), with the higher pKi of 10
(R4 ¼ OH, pKi ¼ 8.41) and 14 (R4 ¼ OCH3, pKi ¼ 8.48) compared to
that of 18 (R4 ¼ OCH2Ph, pKi ¼ 8.03), and with the following affinity
33 (R6 ¼ phenyl) (pKi ¼ 8.30e9.39) in comparison with that of 31
(R6 ¼ CH2NH2, pKi ¼ 7.21) and are also supported by the evidence
of the higher affinity levels of 17 (R7 ¼ OCH3, pKi ¼ 9.90), 42
(R6 ¼ OCH3, pKi ¼ 8.80) and 35 (R5 ¼ OH, R6 ¼ OCH3, pKi ¼ 9.16) in
comparison to those of 13 (R7 ¼ OH, pKi ¼ 8.50), 40 (R6 ¼ OH,
pKi ¼ 6.96) and 36 (R5 ¼ OH, R6 ¼ OH, pKi ¼ 7.67), respectively. The
results are also in accordance with the higher pKi values of 11
(R5 ¼ OH, pKi ¼ 8.72) and 39 (R5 ¼ OH, pKi ¼ 7.04) compared to
those of 15 (R5 ¼ OCH3, pKi ¼ 8.34) and 41 (R5 ¼ OCH3, pKi ¼ 6.63),
respectively, and with the following affinity trend: 2 (R5 ¼ fluoro,
pKi ¼ 8.80) > 3 (R5 ¼ chloro, pKi ¼ 8.52) > 4 (R5 ¼ bromo,
pKi ¼ 8.18). Finally, compounds 11 (R5 ¼ OH, pKi ¼ 8.72) and 39
(R5 ¼ OH, pKi ¼ 7.04) display higher binding affinity than 15
(R5 ¼ OCH3, pKi ¼ 8.34) and 41 (R5 ¼ OCH3, pKi ¼ 6.63),
respectively.
trends:
2
(R5
¼
fluoro, pKi
¼
8.80)
>
3
(R5
¼
chlorine,
pKi ¼ 8.52) > 4 (R5 ¼ bromo, pKi ¼ 8.18); 11 (R5 ¼ OH,
pKi ¼ 8.72) > 15 (R5 ¼ OCH3, pKi ¼ 8.34) > 27 (R6 ¼ CH2OCH3,
pKi ¼ 7.59); 38 (R5 ¼ H, pKi ¼ 8.36) > 39 (R5 ¼ OH, pKi ¼ 7.04) > 41
(R5 ¼ OCH3, pKi ¼ 6.63). Accordingly, compound 35 (R5 ¼ OH,
R6 ¼ OCH3, pKi ¼ 9.16) shows higher binding affinity value than 34
(R5 ¼ OCH2Ph, R6 ¼ OCH3, pKi ¼ 8.75).
According to the electrostatic field contour maps of the CoMFA
analysis plotted in Fig.12b, less positive moieties are predicted to be
favoured (red areas) around the pyrane oxygen atom and the indole
positions 6 and 7. On the other side, more electropositive substit-
uents are predicted to be beneficial (blue polyhedra) in the vicinity
of the indole positions 4 and 5. These results are in agreement with
the higher pKi values of 5 (R6 ¼ chlorine), 6 (R6 ¼ bromo), 9
(R6 ¼ SO2CH3), 16 (R6 ¼ OCH3), 30 (R6 ¼ CN), 32 (R6 ¼ CO2CH3) and
The CoMSIA steric and electrostatic regions are in agreement
with the CoMFA steric and electrostatic areas.
3.5.2. CoMSIA hydrophobic, H-bond acceptor and H-bond donor
regions
The calculated CoMSIA hydrophobic contours (Fig. 13) predict
favourable hydrophobic substituents (yellow areas) around the
tetramethylcyclopropil group, the positions 4 and 5 of the indole
moiety and in proximity of positions 5 and 6 of the pyrane ring. On
the contrary, lipophilic substituents around the 6 and 7 positions of
indole moiety result to be disfavoured (white polyhedra). The
reliability of the hydrophobic map calculations is verified by higher
affinity of 19 (R5 ¼ OCH2Ph, pKi ¼ 8.90), 24 (R5 ¼ O(CH2)4Br,
pKi ¼ 7.14), and 27 (R5 ¼ CH2OCH3, pKi ¼ 7.59) compared to those of
15 (R5 ¼ OCH3, pKi ¼ 8.34), 23 (R5 ¼ O(CH2)4OH, pKi ¼ 6.58), and 26
(R5 ¼ CH2OH, pKi ¼ 6.95), respectively. These results are in
agreement with the difference in pKi levels between: (i) 7
(R6 ¼ CH3, pKi ¼ 9.83) and 8 (R6 ¼ CF3, pKi ¼ 8.46); (ii) 16
(R6 ¼ OCH3, pKi ¼ 9.29) and 20 (R6 ¼ OCH2Ph, pKi ¼ 9.05); (iii) 32
(R6 ¼ CO2CH3, pKi ¼ 8.86) and 9 (R6 ¼ SO2CH3, pKi ¼ 8.53); (iv) 42
(R6 ¼ OCH3, pKi ¼ 8.80) and 43 (R6 ¼ OCH2Ph, pKi ¼ 8.53); (v) 17
(R7 ¼ OCH3, pKi ¼ 9.90) and 21 (R7 ¼ OCH2Ph, pKi ¼ 8.52).
To take into account the role of H-bond acceptor and H-bond
donor groups for the affinity towards the human CB2 receptor, the
corresponding CoMSIA contours were calculated. As shown in
Fig. 14a, H-bond acceptor groups are predicted to be favoured
(magenta regions) around the oxygen atom of pyrane ring, the
Fig. 13. Contour maps of CoMSIA hydrophobic regions (yellow, favoured; white, dis-
favoured) are shown around compounds 17, shown in stick mode and coloured by
atom type. (For interpretation of the references to colour in this figure legend, the
reader is referred to the web version of this article.)

4502
E. Cichero et al. / European Journal of Medicinal Chemistry 46 (2011) 4489e4505
Fig. 14. CoMSIA hydrogen bond acceptor polyhedra are shown around compounds 17, depicted in stick mode and coloured by atom type. H-bond acceptor groups: magenta,
favoured; green, disfavoured (a). CoMSIA hydrogen bond donor polyhedra are shown around compounds 17, depicted in stick mode and coloured by atom type. H-bond donor
groups: purple, disfavoured; cyan, favoured (b). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
oxygen atom of carbonyl function and in proximity of the indole
position 7. Besides, H-bond donors (Fig. 14b) are predicted to be
beneficial (cyan areas) in the vicinity of the indole position 4, while
the introduction of a H-donor group seems to be disfavoured
(purple) in a small area close to the R7 methoxy group in compound
17. The reliability of these calculations is verified by the higher
affinity of 46 (R4 ¼ NH2, pKi ¼ 7.49) in comparison to that of 45
(R4 ¼ NO2, pKi ¼ 6.53).
the CoMFA electrostatic map highlights the importance of elec-
tronegative groups at the indole position 1 and 7, which could be
involved in H-bond interactions with the key residues T114 and
N188. The CoMSIA hydrophobic (Fig. S6) map points out the
beneficial presence of hydrophobic substituents at the following
positions: (i) at the indole position 5, establishing hydrophobic
contacts with M265, C288, (ii) in proximity of the tetramethylcy-
clopropyl group, enhancing interactions with F87, L182 and L289,
(iii) and around the methylene located between the indole position
1 and the pyrane ring, being probably involved in contacts with the
residues L108 and L192. The CoMSIA H-bond donor and H-bond
acceptor maps (Fig. S7) confirm the importance of the formation of
the hydrogen bond between the indole derivative carbonyl group
and S112, and between the pyrane ring oxygen atom and N188.
Furthermore, the introduction of H-donor groups onto the indole
position 4 can allow the establishment of H-bond contacts with
S285.
The information obtained by the 3D-QSAR contour maps and
the molecular docking studies provide useful suggestions for the
synthesis of new indol-3-yl-tetramethylcyclopropyl ketone deriv-
atives with possible improved potency. In detail, a 2- methyl, 6-
methyl or 2,6-dimethylpyrane ring would seem particularly
favourable as substituent R1, detecting hydrophobic contacts with
L108, V113, T114, and L169. Regarding the indole moiety, an alkyl
amine group at position 4 could enhance the ligand potency by H-
bond interactions with S285 (when R7 is a methoxy group, being H-
bonded to N188). On the other hand (when R4 is a hydroxyl group,
3.5.3. A comparison between the CoMFA and CoMSIA analyses and
the hCB2/17 docking model
In order to verify the reliability of the 3D-QSAR model, CoMFA
and CoMSIA maps have been compared with the docking analysis
results. For simplicity, in Figs. S4eS7 (supplementary data) we have
reported only the CoMFA steric, electrostatic and the CoMSIA
hydrophobic and the H-bond acceptor maps, superimposed to the
docking model of compound 17 into the putative receptor binding
site.
The CoMFA steric model proves to match with the hCB2 putative
binding site 3D topology, suggesting bulky substitution in prox-
imity of three deep receptor pockets (Fig. S4). The first cavity (P1),
including F87, S112, F117, L182 and C288, surrounds the tetrame-
thylcyclopropyl group, while the second pocket (P2, delimited by
L108, V113, T114, S165 and L169) is occupied by the substituent at
the indole position 1. Finally, the substituents located at the indole
position 6 and 7 are oriented towards a third receptor pocket (P3),
including residues N188, L192, L196 and F197. As shown in Fig. S5,
Fig. 15. Chemical structure of selective CB2 agonists I and II (showing the highest pKi values in the 1,2,3,4-tetrahydropyrrolo[3,4-b]indole series and in the benzimidazole series,
respectively). The common 1,5-disubstituted-indole or ebenzimidazole scaffold is depicted in blue. (For interpretation of the references to colour in this figure legend, the reader is
referred to the web version of this article.)

E. Cichero et al. / European Journal of Medicinal Chemistry 46 (2011) 4489e4505
4503
Table 5
representative compounds I and II are depicted) have shown high
binding affinity toward CB2 receptor and good selectivity over CB1
receptor.
Thus, in order to identify the key structural features impacting
the binding affinity, a computational study of 3D-QSAR was per-
formed. Initially, the two classes of selective CB2 agonists were
studied separately (one CoMFA and one CoMSIA model for each
series was derived), subsequently, all compounds together were
Binding affinities of compounds 48ae48d towards hCB1 and hCB2 receptors.
Compound
CoMFA Model mmmmodelModel
Pred. pKi
CoMSIA Model
Pred. pKi
48a
48b
48c
48d
8.68
7.01
8.32
6.83
8.43
6.34
8.27
6.28
Table 6
Binding affinities of compounds 48ae48d towards hCB1 and hCB2 receptors.
Compd
Displacement% on CB1a
IC50 on CB1
mM
Ki on CB1
m
M
Displacement% on CB2b
IC50 on CB2
m
M
Ki on CB2
mM
Exp. pKi on CB2
48a
48b
48c
48d
at 10
at 10
m
m
M (83.00)
M (25.08)
163.4
>10
0.062
>10
at 10
at 10
m
m
M (101.00)
M (70.75)
0.0075
1.39
0.0102
1.58
8.12
5.86
7.99
5.80
>10
0.155
>10
5.48
0.0401
6.22
at 1
m
M (85.01)
at 1
m
M (95.93)
M (66.04)
at 10
mM (43.22)
at 10 m
a
Data reported as percent of displacement of [3H]-CP 55,940 (0.5 nM) from hCB1 receptor, at an inhibitor concentration of 10 or 1
mM.
b
Data reported as percent of displacement of [3H]-WIN 55,212-2 (0.8 nM) from hCB2 receptor, at an inhibitor concentration of 10 or 1
m
M.
being H-bonded to S285), an aminomethyl- or aminomethyl group
(NH2CH2- or NH2CH2CH2-) could be introduced at the indole
position 7, being engaged in H-bond with N188. In addition, the
indole position 6 could be exploited to establish hydrophobic
contacts with L196 and F197 by introducing a small alkyl group.
aligned (on the basis of the common scaffold depicted in blue in
Fig. 15), and the final CoMFA and CoMSIA models were calculated.
As concern the indole positions 1 and 3 discussed above, 3D-
QSAR studies performed on 1,2,3,4-tetrahydropyrrolo[3,4-b]indole
and benzimidazole derivatives point out that the following key
features could enhance the ligand hCB2 agonist activity: (i) bulky
substituents bearing electronegative (and H-bond acceptor) groups
at the indole position 1 (see Fig. 15), (ii) electronegative (and H-
bond acceptor) groups located between the indole position 3 and
a bulky hydrophobic substituent (see Fig. 15).
Accordingly, four aminoalkylindoles have been designed and
synthesized, bearing an aminoalkyl group and a vinyl moiety
(disubstituted by polar groups) at the indole position 1 and 3,
respectively. For these compounds, the predicted pKi values (by
means of the CoMSIA model calculated on indoles 1e47), have been
listed in Table 5.
3.6. Newly synthesized compounds 48aed
On the basis of the 3D-QSAR analyses previously discussed,
several favourable structural modifications on the indole moiety
have been identified. Among them, the introduction of H-bond
acceptor groups around the indole position 1 and 3 (see Fig. 14a)
emerged to be particularly interesting. In order to verify the reli-
ability of these results, we compared the CoMSIA maps obtained on
compounds 1e47 with those derived from CoMFA and CoMSIA
studies, already performed by us, on other two series of CB2
agonists, structurally related to indole derivatives [56]. Briefly,
novel classes of CB2 agonists based on 1,2,3,4-tetrahydropyrrolo
[3,4-b]indole and benzimidazole scaffolds [57,58] (in Fig. 15,
As reported in Table 6, compounds 48a and 48c show high Ki
values towards the human CB2 receptor, within a nanomolar range
while compounds 48b and 48d show binding affinities in the
micromolar range. Notably, compound 48a proves to be a CB2
selective ligand.
The MD CB2/agonist complex previously discussed has been
employed to perform molecular docking studies on indoles
48ae48d, in order to rationalize the biological data and to inves-
tigate their binding mode within the molecular target. The most
interesting compound, 48a (Fig. 16), displays H-bond contacts
between the sulphonyl and the N188 side chain, and between the
pyperidine protonated nitrogen atom and the key residue S112. The
indole moiety is engaged in pep with F117 and F197, and in Van der
Waals contacts with V261. The Y cyano and phenyl group are
properly oriented towards residues T114 and T118, and towards
L182 and L185, respectively.
4. Conclusions
In this work, with the aim of gaining a better understanding of
the agonist-CB2 receptor interactions, a theoretical model of the
human CB2 receptor (hCB2) developed by homology modeling and
molecular dynamic techniques has been discussed. The model and
docking protocol reliability has been verified by docking studies on
a series of potent CB2 agonists (indol-3-yl-tetramethylcyclopropyl
ketone derivatives). Successively, the docking-based CoMFA and
CoMSIA analyses, performed on the same series of compounds,
allowed us to derive more complete guidelines for the design of
Fig. 16. Compound 48a (green) selected docking pose into the MD hCB2/WIN-55,212-2
complex. Only residues involved in ligand/receptor interactions are depicted in stick
and labelled. H-bond are depicted in cyano. (For interpretation of the references to
colour in this figure legend, the reader is referred to the web version of this article.)

4504
E. Cichero et al. / European Journal of Medicinal Chemistry 46 (2011) 4489e4505
[20] A.I. Idris, R.J. van’t Hof, I.R. Greig, S.A. Ridge, D. Baker, R.A. Ross, S.H. Ralston,
Regulation of bone mass, bone loss and osteoclast activity by cannabinoid
receptors, Nat. Med. 11 (2005) 774e779.
new analogues with improved potency so as to synthesize new
indoles showing high CB2 affinity.
[21] Z.L. Lu, J.W. Saldanha, E.C. Hulme, Seven-transmembrane receptors: crystals
clarify, Trends Pharmacol. 23 (2002) 140e146.
[22] O.M. Salo, K.H. Raitio, J.R. Savinainen, T. Nevalainen, M. Lahtela-Kakkonen,
J.T. Laitinen, T. Järvinen, A. Poso, Virtual screening of novel CB2 ligands using
a comparative model of the human cannabinoid CB2 receptor, J. Med. Chem.
48 (2005) 7166e7171;
Acknowledgements
This work was financially supported by the University of Gen-
ova, MIUR FIRB ITALBIONET (RBPR05ZK2Z and RBIN064YAT_003),
flagship InterOmics. Prof. Stefano Moro, Dr. Marco Fanton and Dr.
Giuseppe Marson are gratefully acknowledged. E.C. was financially
supported by a post-doc fellowship, Area Chimica, University of
Genova.
T. Tuccinardi, P.L. Ferrarini, C. Manera, G. Ortore, G. Saccomanni, A. Martinelli,
Cannabinoid CB2/CB1 selectivity. Receptor modeling and automated docking
analysis, J. Med. Chem 49 (2006) 984e994;
S. Durdagi, M.G. Papadopoulos, P.G. Zoumpoulakis, C. Koukoulitsa,
T. Mavromoustakos, A computational study on cannabinoid receptors and
potent bioactive cannabinoid ligands: homology modeling, docking, de novo
drug design and molecular dynamics analysis, Mol. Divers 14 (2010)
257e276.
Appendix. Supplementary material
[23] V. Cherezov, D.M. Rosenbaum, M.A. Hanson, S.G. Rasmussen, F.S. Thian,
T.S. Kobilka, H.J. Choi, P. Kuhn, W.I. Weis, B.K. Kobilka, R.C. Stevens, High-
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.ejmech.2011.07.023.
resolution crystal structure of an engineered human beta2-adrenergic
G
protein-coupled receptor, Science 318 (2007) 1258e1265.
[24] V.P. Jaakola, M.T. Griffith, M.A. Hanson, V. Cherezov, E.Y. Chien, J.R. Lane,
A.P. Ijzerman, R.C. Stevens, The 2.6 angstrom crystal structure of a human A2A
adenosine receptor bound to an antagonist, Science 322 (2008) 1211e1217.
[25] K. Palczewski, T. Kumasaka, T. Hori, C.A. Behnke, H. Motoshima, B.A. Fox, I. Le
Trong, D.C. Teller, T. Okada, R.E. Stenkamp, M. Yamamoto, M. Miyano, Crystal
structure of rhodopsin: a G protein-coupled receptor, Science 289 (2000)
739e745.
[26] J.M. Frost, M.J. Dart, K.R. Tietje, T.R. Garrison, G.K. Grayson, A.V. Daza, O.F. El-
Kouhen, L.N. Miller, L. Li, B.B. Yao, G.C. Hsieh, M. Pai, C.Z. Zhu, P. Chandran,
M.D. Meyer, Indol-3-yl-tetramethylcyclopropyl ketones: effects of indole ring
substitution on CB2 cannabinoid receptor activity, J. Med. Chem. 51 (2008)
1904e1912.
References
[1] S.L. Palmer, G.A. Thakur, A. Makriyannis, Cannabinergic ligands, Chem. Phys.
Lipids 121 (2002) 3e19.
[2] A.C. Howlett, F. Barth, T.I. Bonner, G. Cabral, P. Casellas, W.A. Devane,
C.C. Felder, M. Herkenham, K. Mackie, B.R. Martin, R. Mechoulam,
R.G. Pertwee, International Union of Pharmacology. XXVII. Classification of
cannabinoid receptors, Pharmacol. Rev. 54 (2002) 161e202.
[3] R.G. Pertwee, Cannabinoid receptors and pain, Prog. Neurobiol. 63 (2001)
569e611.
[4] P. Goya, N. Jagerovic, L. Hernandez-Folgado, M.I. Martin, Cannabinoids and
neuropathic pain, Mini Rev. Med. Chem. 3 (2003) 765e772.
[27] MOE: Chemical Computing Group Inc. Montreal. H3A 2R7 Canada. http://
www.chemcomp.com.
[28] A. Bairoch, R. Apweiler, The SWISS-PROT protein sequence database and its
supplement TrEMBL in 2000, Nucleic Acids Res. 28 (2000) 45e48.
[29] H.M. Berman, J. Westbrook, Z. Feng, G. Gilliland, T.N. Beth, H. Weissig,
I.N. Shindyalov, P.E. Bourne, The protein data bank, Nucleic Acids Res. 28
(2000) 235e242.
[30] W.D.C.P. Cornell, C.I. Bayly, I.R. Gould, K.M. Merz, D.M. Ferguson,
D.C. Spellmeyer, T. Fox, J.W. Caldwell, P.A.J. Kollman, A second generation
force field for the simulation of proteins, nucleic acids and organic molecules,
J. Am. Chem. Soc. 117 (1995) 5179e5196.
[5] A. Calignano, G. La Rana, D. Piomelli, Antinociceptive activity of the endoge-
nous fatty acid amide, palmitylethanolamide, Eur. J. Pharmacol. 419 (2001)
191e198.
[6] V. Di Marzo, C.S. Breivogel, Q. Tao, D.T. Bridgen, R.K. Razdan, A.M. Zimmer,
B.R. Martin, Levels, metabolism, and pharmacological activity of anandamide
in CB(1) cannabinoid receptor knockout mice: evidence for non-CB(1), non-
CB(2) receptor-mediated actions of anandamide in mouse brain, Neuro-
chemistry 75 (2000) 2434e2444.http://www.ncbi.nlm.nih.gov/pubmed/
11080195.
[31] T. Tuccinardi, P.L. Ferrarini, C. Manera, G. Ortore, G. Saccomanni, A. Martinelli,
Cannabinoid CB2/CB1 selectivity. Receptor modeling and automated docking
analysis, J. Med. Chem. 49 (2006) 984e994.
[32] D.I. Chan, D.P. Tieleman, H.J. Vogel, Molecular dynamics simulations of beta-
ketoacyl-, beta-hydroxyacyl-, and trans-2-enoyl-acylcarrier proteins of
Escherichia coli, Biochemistry 49 (2010) 2860e2868.
[33] S.J. Marrink, X. Periole, D.P. Tieleman, A.H. de Vries, Comment on ‘On using
a too large integration time step in molecular dynamics simulations of coarse-
grained molecular models, Phys. Chem. Chem. Phys. 12 (2010) 2254e2256.
[34] S. Baoukina, S.J. Marrink, D.P. Tieleman, Lateral pressure profiles in lipid
monolayers, Faraday Discuss. 144 (2010) 393e409.
[7] N. Hájos, C. Ledent, T.F. Freund, Cannabinoids inhibit hippocampal GABAergic
transmission and network oscillations, Neuroscience 12 (2000) 3239e3249.
[8] G.T. Whiteside, G.P. Lee, K.J. Valenzano, The role of the cannabinoid CB2
receptor in pain transmission and therapeutic potential of small molecule CB2
receptor agonists, Curr. Med. Chem. 14 (2007) 917e936.
[9] J.C. Ashton, Cannabinoids for the treatment of inflammation, Curr. Opin.
Investig. Drugs 8 (2007) 373e384.
[10] G.M. Giblin, C.T. O’Shaughnessy, A. Naylor, W.L. Mitchell, A.J. Eatherton,
B.P. Slingsby, D.A. Rawlings, P. Goldsmith, A.J. Brown, C.P. Haslam,
N.M. Clayton, A.W. Wilson, I.P. Chessell, A.R. Wittington, Green, R Discovery of
2-[(2,4-dichlorophenyl)amino]-N-[(tetrahydro-
2H-pyran-4-yl)methyl]-4-
[35] C. Kutzner, D. Van der Spoel, M. Fechner, E. Lindahl, U.W. Schmitt, B.L. De
Groot, H.J. Grubmüller, Speeding up parallel GROMACS on high-latency
networks, J. Comput. Chem. 28 (2007) 2075e2084.
[36] B. Hess, C. Kutzner, D. Van der Spoel, E.J. Lindahl, GROMACS 4: algorithms for
highly efficient, load-balanced, and scalable molecular simulations, Chem.
Theory Comput. 4 (2008) 435e447.
[37] W.F. Van Gunsteren, S.R. Billeter, A.A. Eising, P.H. Hunenberger, P. Kruger,
A.E. Mark, W.R.P. Scott, I.G. Tironi, Biomolecular simulation, in: The GRO-
MOS96 manual and user guide. Vdf Hochschulverlag AG an der ETH Zurich,
1996, pp. 1e1042.
[38] O. Berger, O. Edholm, F. Jähnig, Molecular dynamics simulations of a fluid
bilayer of dipalmitoylphosphatidylcholine at full hydration, constant pressure,
and constant temperature, Biophys. J. 72 (1997) 2002e2013.
[39] A.W. Schuettelkopf, D.M.F. Van Aalten, PRODRG e a tool for high-throughput
crystallography of protein-ligand complexes, Acta Crystallogr. D 60 (2004)
1355e1363.
[40] J.A. Lemkul, W.J. Allen, D.R. Bevan, Practical Considerations for Building
GROMOS-Compatible small molecule Topologies, J. Chem. Inf. Model. 50
(2010) 2221e2235.
[41] J. Wang, W. Wang, P.A. Kollman, D.A. Case, Automatic atom type and bond
type perception in molecular mechanical calculations, J. Mol. Graph. Model. 25
(2006) 247e260.
[42] A. Jakalian, B.L. Bush, D.B. Jack, C.I. Bayly, Fast efficient generation of high-
quality atomic charges, AM1-BCC Model: Method J. Comput. Chem. 21
(2000) 132e146.
[43] A. Jakalian, D.B. Jack, C.I. Bayly, Fast, efficient generation of high-quality
atomic charges. AM1-BCC model: II. Parameterization and validation,
J. Comput. Chem. 23 (2002) 1623e1641.
(trifluoromethyl)- 5-pyrimidinecarboxamide, a selective CB2 receptor agonist
for the treatment of inflammatory pain, J. Med.Chem. 50 (2007) 2597e2600.
[11] J. Guindon, A.G. Hohmann, Cannabinoid CB2 receptors: a therapeutic target
for the treatment of inflammatory and neuropathic pain, Br. J. Pharmacol. 153
(2008) 319e334.
[12] A.D. Khanolkar, D. Lu, M. Ibrahim, R.I. Duclos, G.A. Thakur, T.P. Malan,
F. Porreca, V. Veerappan, X. Tian, C. George, D.A. Parrish, D.P. Papahatjis,
A. Makriyannis, Cannabilactones: a novel class of CB2 selective agonists with
peripheral analgesic activity, J. Med. Chem. 50 (2007) 6493e6500.
[13] J. Fernandez-Ruiz, J. Romero, G. Velasco, R.M. Tolon, J.A. Ramos, M. Guzman,
Cannabinoid CB2 receptor: a new target for controlling neural cell survival?
Trends Pharmacol. Sci. 28 (2007) 39e45.
[14] M. Maccarrone, N. Battista, D. Centonze, The endocannabinoid pathway in
Huntington’s disease: a comparison with other neurodegenerative diseases,
Prog. Neurobiol. 81 (2007) 349e379.
[15] D. Centonze, A. Finazzi-Agro, G. Bernardi, M. Maccarrone, The endocannabi-
noid system in targeting inflammatory neurodegenerative diseases, Trends
Pharmacol. Sci. 28 (2007) 180e187.
[16] R.J. McKallip, C. Lombard, M. Fisher, B.R. Martin, S. Ryu, S. Grant,
P.S. Nagarkatti, M. Nagarkatti, Targeting CB2 cannabinoid receptors as a novel
therapy to treat malignant lymphoblastic disease, Blood 100 (2002) 627e634.
[17] G. Velasco, I. Galve-Roperh, C. Sánchez, C. Blázquez, M. Guzmán, Hypothesis:
cannabinoid therapy for the treatment of gliomas? Neuropharmacology 47
(2004) 315e323.
[18] R.G. Pertwee, Cannabinoids and multiple sclerosis, Pharmocol. Therapeut 95
(2002) 165e174.
[19] O. Ofek, M. Karsak, N. Leclerc, M. Fogel, B. Frenkel, K. Wright, J. Tam, M. Attar-
Namdar, V. Kram, E. Shohami, R. Mechoulam, A. Zimmer, I. Bab, Peripheral
cannabinoid receptor, CB2, regulates bone mass, Proc. Natl. Acad. Sci. U.S.A.
103 (2006) 696e701.
[44] Javier Ramos, Víctor L. Cruz, Javier Martínez-Salazar, Nuria E. Campillo, Juan
A. Páez, Dissimilar interaction of CB1/CB2 with lipid bilayers as revealed by

E. Cichero et al. / European Journal of Medicinal Chemistry 46 (2011) 4489e4505
4505
molecular dynamics simulationw, Phys. Chem. Chem. Phys. 13 (2011)
3660e3668.
[45] W. Humphrey, A. Dalke, K. Schulten, VMD: visual molecular dynamics, J. Mol.
Graph 14 (1996) 33e38 27-8.
[46] M.G. Wolf, M. Hoefling, C. Aponte-Santamaría, H. Grubmüller, G. Groenhof,
g_membed: efficient insertion of a membrane protein into an equilibrated
lipid bilayer with minimal perturbation, J. Comput. Chem. 31 (2010)
2169e2174.
[47] U. Essmann, L. Perera, M.L. Berkowitz, T. Darden, H. Lee, L.G. Pedersen,
A smooth Particle Mesh Ewald method, J. Chem. Phys. 103 (1995) 8577e8593.
[48] B.J. Hess, P-LINCS: a parallel Linear constraint Solver for molecular simulation,
Chem. Theory Comput. 4 (2008) 116e122.
[49] G. Bussi, D. Donadio, M. Parrinello, Canonical sampling through velocity
scaling, J. Chem. Phys. 126 (2007) 014101.
[50] H.J.C. Berendsen, J.P.M. Postma, W.F. Van Gunsteren, A. Di Nola, J.R. Haak,
Molecular dynamics with coupling to an external bath, J. Chem. Phys. 81
(1984) 3684e3690.
[51] W.L. DeLano, The PyMOL molecular graphics system DeLano Scientific. San
Carlos, CA, USA, 2002.http://www.pymol.org.
[52] R.D. Cramer III, D.E. Patterson, J.D. Bunce, Prog. Clin. Biol. Res. 291 (1989)
1e165.
[53] G. Klebe, U. Abraham, T. Mietzner, Molecular similarity indices in a compar-
ative analysis (CoMSIA) of drug molecules to correlate and predict their bio-
logical activity, J. Med. Chem. 37 (1994) 4130e4146.
[54] Sybyl X 1.0 Tripos Inc 1699 South Hanley Road. St Louis. Missouri. 63144. USA
25.
[55] A. Ranise, S. Cesarini, A. Spallarossa, F. Sancassan, F. Bondavalli, O. Bruno,
S. Schenone, G. Menozzi, P. Fossa, L. Mosti, Unprecedented one-pot stereo-
selective synthesis of Knoevenagel-type derivatives via in situ condensation
of N-methyleniminium salts of ethylenethiourea and ethyleneurea with
active methylene reagents, Synthesis-Stuttgart 16 (2007) 2495e2502.
[56] E. Cichero, S. Cesarini, L. Mosti, P. Fossa, CoMFA and CoMSIA analyses on
1,2,3,4-tetrahydropyrrolo[3,4-b]indole and benzimidazole derivatives as
selective CB2 receptor agonists, J. Mol. Model. 16 (2010) 1481e1498.
[57] D. Pagé, H. Yang, W. Brown, C. Walpole, M. Fleurent, M. Fyfe, F. Gaudreault, St-
Onge S New 1,2,3,4-tetrahydropyrrolo[3,4-b]indole derivatives as selective
CB2 receptor agonists, Bioorg. Med. Chem. Lett. 22 (2007) 6183e6187.
[58] D. Pagé, E. Balaux, L. Boisvert, Z. Liu, C. Milburn, M. Tremblay, Z. Wei, S. Woo,
X. Luo, Y.X. Cheng, H. Yang, S. Srivastava, F. Fhou, W. Brown, M. Tomaszewski,
C. Walpole, L. Hodzic, S. St-Onge, C. Godbout, D. Salois, K. Payza, Novel
benzimidazole derivatives as selective CB2 agonists, Bioorg. Med. Chem. Lett.
13 (2008) 3695e3700.