Three-dimensional quantitative structure-selectivity relationships analysis guided rational design of a highly selective ligand for the cannabinoid receptor 2

Article abstract of DOI:10.1016/j.ejmech.2010.11.034

This paper describes a three-dimensional quantitative structure-selectivity relationships (3D-QSSR) study for selectivity of a series of ligands for cannabinoid CB1 and CB2 receptors. 3D-QSSR exploration was expected to provide design information for drug

Full text of DOI:10.1016/j.ejmech.2010.11.034

European Journal of Medicinal Chemistry 46 (2011) 547e555
Contents lists available at ScienceDirect
European Journal of Medicinal Chemistry
journal homepage: http://www.elsevier.com/locate/ejmech
Original article
Three-dimensional quantitative structureeselectivity relationships analysis
guided rational design of a highly selective ligand for the cannabinoid receptor 2
a
a
b
b
a
Simone Brogi , Federico Corelli , Vincenzo Di Marzo , Alessia Ligresti , Claudia Mugnaini ,
a
a,
*
Serena Pasquini , Andrea Tafi
Dipartimento Farmaco Chimico Tecnologico, Università degli Studi di Siena, Via Alcide de Gasperi 2, 53100 Siena, Italy
a
b
Endocannabinoid Research Group, Institute of Biomolecular Chemistry, Consiglio Nazionale delle Ricerche, Via dei Campi Flegrei 34, 80078 Pozzuoli (Napoli), Italy
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 6 August 2010
Received in revised form
This paper describes a three-dimensional quantitative structureeselectivity relationships (3D-QSSR)
study for selectivity of a series of ligands for cannabinoid CB1 and CB2 receptors. 3D-QSSR exploration
was expected to provide design information for drugs with high selectivity toward the CB2 receptor. The
proposed 3D computational model was performed by Phase and generated taking into account a number
of structurally diverse compounds characterized by a wide range of selectivity index values. The model
proved to be predictive, with r2 of 0.95 and Q2 of 0.63. In order to get prospective experimental vali-
dation, the selectivity of an external data set of 39 compounds reported in the literature was predicted.
22 October 2010
Accepted 19 November 2010
Available online 27 November 2010
Keywords:
2
The correlation coefficient (r ¼ 0.56) obtained on this unrelated test set provided evidence that the
3D-QSSR
correlation shown by the model was not a chance result. Subsequently, we essayed the ability of our
approach to help the design of new CB2-selective ligands. Accordingly, based on our interest in studying
the cannabinergic properties of quinolones, the N-(adamantan-1-yl)-4-oxo-8-methyl-1-pentyl-1,4-
dihydroquinoline-3-carboxamide (65) was considered as a potential synthetic target. The log(SI) value
predicted by using our model was indicative of high CB2 selectivity for such a compound, thus spurring
us to synthesize it and to evaluate its CB1 and CB2 receptor affinity. Compound 65 was found to be an
extremely selective CB2 ligand as it displayed high CB2 affinity (Ki ¼ 4.9 nM), while being devoid of CB1
affinity (Ki > 10,000 nM). The identification of a new selective CB2 receptor ligand lends support for the
practicability of quantitative ligand-based selectivity models for cannabinoid receptors. These drug
discovery tools might represent a valuable complementary approach to docking studies performed on
homology models of the receptors.
Cannabinoid receptor 2
Selective ligands
Computer-aided drug design
Quinolones
Ó 2010 Elsevier Masson SAS. All rights reserved.
1. Introduction
and inflammatory pain and are intensively investigated as potential
new analgesic and antiinflammatory agents [4]. Unfortunately,
The cannabinoid 1 receptor (CB1 receptor) and the cannabinoid
receptor (CB2 receptor) are members of the G-protein-coupled
CB1/CB2 agonists are not devoid of unwanted side effects, many of
which are thought to be due to activation of central CB1 receptor
rather than peripheral CB1 or CB2 receptors [5].
2
receptor family [1]. While CB1 receptor is abundantly expressed in
the central nervous system (CNS), CB2 receptor is mainly localized
in peripheral nerve terminals and in the tissues of the immune
system [2]. Recent studies have suggested that CB2 receptor is also
expressed in certain subpopulations of the CNS and evidence is
growing that CB1 receptor is also expressed in peripheral tissues
3].
Agonists of both cannabinoid receptor subtypes produce strong
antinociceptive effects in animal models of chronic, neuropathic,
In principle, separating the therapeutic effects of cannabinoid
agonists from their undesired effects could be accomplished by
either preventing the ligands from crossing the bloodebrain barrier
or by increasing the selectivity of the ligands for the CB2 receptor
[6]. Several classes of selective CB2 ligands have demonstrated
efficacy in pre-clinical models of inflammatory pain [7] and have
shown a therapeutic window with regard to CNS side effects [8].
However, none of the CB2-selective agonists that have been
developed to date are completely CB2-specific. Thus, they are all
expected to display CB2 selectivity only within a finite dose range
and to target CB1 receptor as well when administered at a dose that
lies above this range [9]. On the basis of these considerations,
[
*
Corresponding author. Tel.: þ39 0577 234 313; fax: þ39 0577 234 333.
E-mail address: tafi@unisi.it (A. Tafi).
0223-5234/$ e see front matter Ó 2010 Elsevier Masson SAS. All rights reserved.
doi:10.1016/j.ejmech.2010.11.034

548
S. Brogi et al. / European Journal of Medicinal Chemistry 46 (2011) 547e555
interest is growing in developing new structural classes of CNS
penetrant CB2 agonists with high receptor subtype selectivity
suitable for in vivo studies [6,10e12].
might show a functional profile [28] different from that assigned by
structural similarity [13c]. Phase [29], a software package designed
for pharmacophore modeling, structure alignment and activity
prediction has been used for this purpose. Notably, this package
provides the means to align sets of ligands onto a pharmacophore
and to develop 3D-QSAR models able to identify further structural
features that govern molecule activity. In this study, Phase was
firstly applied to develop a common feature CB2 pharmacophore
model to be used as an alignment rule and, then, to carry out a 3D-
QSSR investigation [30].
Even though many efforts have been directed in recent years to
the modeling of CB2 receptor binding, the rational design of novel
CB2-selective ligands by computational methodologies is still
a challenging task [10]. The vast majority of computational studies
on CB receptors consist either of retrospective rationalizations
focused on proteineligand docking simulations using homology
models of both receptors and 3D-QSAR models [13], or in phar-
macophore-based virtual screening protocols [14]. Hence, a lack is
perceived of predictive models for CB2 selectivity, effective to assist
the drug design process. On the other hand, the knowledge of
several CB2-selective classes of compounds might allow pharma-
cophore modeling (PM) to help fill this gap. This technique, in fact,
not only enables fast design of novel structural scaffolds, but also
provides sound alignment rules whereon one could ground
predictive three-dimensional structureeselectivity relationships
2. Results and discussion
A representative set of 64 CB2 ligands was selected (see Fig. 1
and Table S1), taking in no account their functional activity,
among
a
number of 4-quinolone-3-carboxamides recently
synthesized in our laboratory (23e39 and 64) [31] and derivatives
belonging to different structural classes already reported in the
literature (1e13, 17 and WIN55212-2 (58) [28], JWH-015 (14) and
CP-55,940 (55) [32], JWH-181 (15) and JWH-007 (16) [19], AM1241
(18) [33], AM630 (19) [34], 20e22 and 40e42 [7c], 43e45 [20],
L759633 (46) and L759656 (47) [35], 48e51 [21], O-1057 (52) [36],
AMG41 (53) [22], JWH-133 (54) [23], AM855 (56) [24], BAY593074
(57) [37], gp1a (59) [38], 60 [39], SR144528 (61) [40], HU-308 (62)
[41], and GW405833 (63) [42]). The binding affinities of all these
compounds for human recombinant CB1 and CB2 receptors (Ki
values) have been measured according to the same protocol, by
(
3D-QSSR) approaches.
The difficulties inherent in the rational discovery of selective
ligands of CB2 receptor with a clear-cut functional activity profile
agonist/antagonist/inverse agonist) have been recently faced in
(
the case of pharmacophore modeling [14]. Markt and coworkers
demonstrated that CB2 receptor-selective agonists and antagonists/
inverse agonists can be “structurally closely related” so that “the
differences in terms of chemical features are subtle”. Consequently,
these authors have abandoned the idea of generating selective
models for agonists, antagonists and inverse agonists as “discrim-
ination between agonists and antagonists would only be possible
with very restrictive pharmacophore models which would not be
suitable for a virtual screening workflow focused on the discovery
of structurally novel scaffolds” [14]. Actually, the pharmacophore
model developed by these authors, though based on CB2 receptor-
selective agonists only, screened some ligands with moderate
selectivity, different binding behavior and functional activity.
Up to date, only a CoMFA/CoMSIA model of selectivity for indole
ligands of CB1 and CB2 receptor subtypes has been published, in
which the functional activities of the studied set of compounds
3
displacement of the radioligand [ H]-CP-55,940 [7c,19e24,28,
31e42] and are shown in Table 1 (second and third columns).
Our selection was restrained to compounds showing high CB2
affinity (Ki values ꢀ 56.7 nM) and CB1 affinity covering approxi-
mately four orders of magnitude (Ki values ranging from 0.5 to
>10,000 nM). As a consequence, the selectivity index of the
compounds [SI, calculated as Ki(CB1)/Ki(CB2) ratio, fourth column
of Table 1] essentially depends on their affinity at CB1 receptor.
With the aim to develop a bare selectivity model, the log(SI) was
used as the experimental activity variable in Phase. The selectivity
index of some derivatives (“undetermined” compounds hereafter)
was not computable precisely due to their low CB1 affinity [Ki
(CB1) > 10,000 nM]. However, because highly selective compounds
were considered as a source of important structureeselectivity
relationships, two “undetermined” compounds (20 and 24) were
included in the training set, with a 10,000 nM Ki(CB1) value arbi-
trarily assigned, in order to avoid the loss of positive information.
The remainder of the undetermined compounds was included in
the test set to stress the model predictivity. Therefore, the depen-
dent QSSR variable spanned a four logarithmic units range starting
from zero, ensuring statistical significance to the approach.
All the compounds were aligned onto a purpose-built CB2
common feature pharmacophore (details concerning the building
of the pharmacophore and its performance as a retrospective 3D-
QSAR modeling tool are provided in Supplementary material).
Three quantitative selectivity models containing one to three PLS
factors were then generated. Due to the peculiarity of our purpose
and according to Phase manual suggestions, the atom-based
version of the QSAR methodology was preferred to the pharma-
cophore-based one, in order to consider contributions to selectivity
possibly deriving from features other than the pharmacophore [30].
The whole set of 64 molecules was divided into a training set and
a test set represented by 29 and 35 compounds (Table 1), respec-
tively, selected in an unbiased way in an effort to maximize struc-
tural diversity and coverage of experimental activities. Compounds
20, 24 and 59 represented the high boundary of our training set
selectivity range.
(
[
generally proposed as agonists) have not been analyzed in detail
15]. A general strategy for the development of selectivity models,
however, has been recently suggested by Weber and coworkers
through CoMFA/CoMSIA analyses of inhibitors of carbonic anhy-
drase isoforms. These scientists have derived the molecular align-
ment of isozyme selective inhibitors from one enzyme isoform
only, by molecular docking studies of compounds into its binding
site [16]. An analogous approach can be applied in the case of CB1
and CB2 receptors, as the high degree of homology (68%) exhibited
by the transmembrane domains of these targets causes binding
affinities of their respective ligands to be generally correlated. Such
an outcome, in fact, has been even evaluated to be consistent with
the hypothesis that non selective compounds can keep the same
conformation when bound to both subtypes [17]. Moreover, Wiley
and coworkers have accounted for structureeactivity relationships
results suggesting the overlap, albeit incomplete, of the pharma-
cophores for CB1 and CB2 receptors [18].
Based on all the above considerations, we have developed an
inclusive 3D-QSSR model, founded on a CB2 common feature
pharmacophore and able to predict in a semi quantitative manner
the selectivity index (see below) of novel CB2 receptor ligands
belonging to several structural classes. According to the difficulties
discussed above in the prediction of functional activity at CB2
receptor [14], in this study no analysis of ligands functional activ-
ities was performed. On the other hand the functional activity of
several CB2 ligands reported so far in the literature and used in this
study has been not explicitly determined [7c,19e27] so that they

S. Brogi et al. / European Journal of Medicinal Chemistry 46 (2011) 547e555
549
Fig. 1. Chemical scaffolds of the compounds used to generate the 3D-QSSR model (see Table S1 in the Supplementary material for further details).
Table 2 shows the statistical parameters derived using Phase
methodology. The model with three PLS factors was preferred and
chosen as the 3D-QSSR model, since it performed better on the
whole than those with fewer factors. The correlation coefficients of
The 3D-QSSR results were visualized using 3D plots of the
crucial volume elements occupied by ligands. Figs. 3e5 show the
3D plot representation of the model as a whole superimposed to
24, 61 (SR144528), and 55 (CP-55,940), respectively. In this
representation blue and red cubes indicate positive and negative
coefficients, respectively, that is volumes in which the occupying
atoms of the ligands cause an increase or a decrease of selectivity.
Cubes having small positive and negative coefficients, which
therefore did not greatly affect selectivity, were filtered out by
setting a ꢁ2.0eꢂ02 coefficient threshold. Notably, compound 24,
showing the greatest CB2 selectivity, mainly occupies blue regions,
while the less CB2-selective derivative 61 occupies some of the red
regions. Finally, derivative 55, a compound showing no CB2
selectivity, mainly occupies red regions. In Fig. 6 only the cubes
occupied by compound 24 are displayed, decomposed into the
contributions to the model by different atom classes. The map
corresponding to electron-withdrawing atoms (including
hydrogen-bond acceptors) is displayed on the left of Fig. 6, while
that corresponding to hydrophobic/non-polar atoms is shown on
the right.
2
2
the model (r ¼ 0.95 and Q ¼ 0.63) were statistically acceptable
when considering both the ratio (0.8) between the number of
compounds in the training and test sets, respectively, and the
number of undetermined compounds included in the test set (11
compounds). Moreover, the high Pearson-R value (Pearson-
R ¼ 0.81) also indicated a close correlation between predicted and
actual selectivity index values. These features, together with the
small number of PLS factors, the large F value and the small variance
ratio (P) supported the robustness of the approach. The linear plot of
calculated/predicted versus actual selectivity index values is dis-
played in Fig. 2 and reported in the fifth and fourth columns of Table
1. Notably, with the exception of 48, the differences between
experimental and calculated values were within one order of
magnitude for all the compounds, demonstrating that the selec-
tivity of compounds in both training and test sets was reasonably
well estimated by the model.

550
S. Brogi et al. / European Journal of Medicinal Chemistry 46 (2011) 547e555
Table 1
Table 2
CB1 and CB2 receptor affinity values [Ki(CB1) and Ki(CB2) columns, nM], experi-
mental (Exp column) and estimated (Calc column) selectivity index values [log(SI)]
for compounds used in the computational study. Estimated log(SI) values were
calculated by application of 3D-QSSR model (see text).
3D-QSSR statistical parameters of the three phase-derived sets of models.
PLS
r2a
SDb
Fc
Pd
RMSEe
Q2f
R-Pearsong
1
2
3
0.54
0.89
0.95
0.82
0.39
0.28
33.6
118.5
173.7
2.34eꢂ06
1.60eꢂ14
9.30eꢂ18
0.61
0.58
0.57
0.56
0.61
0.63
0.75
0.80
0.81
Cmpd
Ki(CB1)
45
Ki(CB2)
Exp log(SI)
Calc log(SI)
a
1
2
3
4
5
6
7
8
0.1
4.4
0.4
0.9
0.2
30.0
25.0
9.0
8.3
4.6
1.3
0.7
3.0
13.3
0.6
2.57
2.30
2.27
1.58
2.30
2.05
1.85
2.05
2.55
2.00
2.30
3.05
0.98
1.47
0.32
0.28
2.05
2.64
2.22
3.42
0.95
2.34
2.01
4.15
3.68
3.38
3.08
3.00
2.79
2.35
2.38
1.38
3.05
3.10
3.14
1.74
2.27
1.81
3.35
2.30
2.28
1.82
0.52
0.15
1.05
2.26
2.21
0.34
0.43
0.19
0.09
0.02
0.07
2.30
0.00
0.60
0.03
1.01
3.95
2.16
2.74
2.64
2.21
2.30
2.49
2.30
2.18
1.80
2.02
2.06
1.86
2.10
1.59
1.97
1.84
2.46
0.93
1.15
0.08
0.67
1.94
2.63
1.72
3.2
a
r2: value of r2 for the regression.
a
845
228
33
b
c
d
e
f
a
SD: standard deviation of the regression.
F: variance ratio.
b
b
P: significance level of variance ratio.
RMSE: root-mean-square error.
28
b
3310
1700
1000
2951
280
234
780
28
2
2
b
Q : value of Q for the predicted activities.
g
a
R: r-Pearson, correlation between the predicted and observed selectivity index
b
values for the test set.
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
6
6
6
6
0b
1
2
b
b
interactions established by the adamantyl group on the amide
moiety and by the C6 substituent (2-furyl group in compound 24).
In order for this ligand to keep these hydrophobic groups in the
right orientation, the trans-conformation of the amide is required
as well as the alignment of the amide carbonyl dipole with the
ketone carbonyl dipole.
The 3D-QSSR model was then subjected to a prospective
experimental validation. First of all, it was used to predict the
selectivity of an external data set of 39 compounds reported in the
literature [6,7c,10e12,19,25e27,43e50] (E1eE39, see Fig. 7 and
a
3
4 (JWH-015)a
5 (JWH-181)a
383
1.3
9.5
3500
5000
5152
b
6 (JWH-007)
2.9
b
7
32.0
11.5
31.2
3.8
a
8 (AM1241)
b
9 (AM630)
a
0
>10,000
42
a
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
4.7
1.20
2.66
2.06
3.28
2.88
3.09
2.20
3.11
1.87
2.95
3.08
2.00
2.81
2.66
2.32
1.91
2.18
1.61
2.50
1.94
2.15
2.11
0.38
0.94
0.95
2.46
2.09
0.99
0.41
1.49
0.13
0.03
0.42
2.19
0.07
0.53
ꢂ0.09
1.88
4.20
2.19
2.83
2.85
2.36
1.25
a
2520
2080
11.6
20.6
0.7
2.3
4.2
b
a
>10,000
>10,000
>10,000
>10,000
>10,000
>10,000
>10,000
>10,000
510
>10,000
>10,000
>10,000
900
Table S2 in the Supplementary material). The correlation coefficient
b
2
(
r ¼ 0.56) of this unrelated set of derivatives was comparable to the
b
b
value obtained for the test set; the linear plot of predicted versus
actual selectivity index values (displayed in Fig. 2 and reported in
the fifth and fourth columns of Table 3) and the differences
between these values (within one order of magnitude for all the
compounds with the exceptions of E25 and E29); provided
evidence that the correlation shown by the model was not a chance
result.
8.3
b
11.0
16.0
44.8
41.9
21.5
8.8
8.0
7.3
16.9
56.6
49.8
4.4
6.3
3.4
14.3
1.7
20.0
1.3
11.8
6.4
3.9
0.2
1.3
9.4
8.0
0.9
3.4
0.6
5.4
45.5
1.3
0.04
13.4
5.4
22.7
12.0
2.4
b
b
b
a
b
b
b
Having gained such a confidence, as a second step, we tested the
ability of our 3D-QSSR model to help the design of new CB2-selec-
tive ligands. Inspection of Fig. 3 clearly shows that the 8-position of
b
b
>10,000
3210
>10,000
1220
b
b
b
a
640
996
5.6
28
14
b
a
b
b
5
6 (L759656)a
4888
1043
a
7 (L759633)
8
9
0
b
8.3
0.5
1.8
11.7
8.4
1.00
677
0.6
22.3
48.3
13.3
a
b
a
1
a
2 (O-1057)
a
3 (AMG41)
4 (JWH-133)a
5 (CP-55,490)
6 (AM855)a
a
7 (BAY593074)a
8 (WIN55212-2)
b
a
9 (gp1a)
0
363
b
1925
2890
10,000
1917
480
a
1 (SR144528)
2 (HU-308)a
a
3 (GW405833)
4
b
a
Compound included in the training set.
Compound included in the test set.
b
This visual representation of the contributions of 24 to the
model highlights the great dominance of hydrophobic/non-polar
cubes with respect to electron-withdrawing ones. In other words,
selectivity appears to be strongly dependent on van der Waals
Fig. 2. Scatter plot for the predicted and observed selectivity index values [log(SI)] as
calculated by the 3D-QSSR model applied to the training set, test set and external test
set compounds.

S. Brogi et al. / European Journal of Medicinal Chemistry 46 (2011) 547e555
551
Fig. 3. Superposition of the 3D-QSSR model and the highly selective compound 24.
Fig. 5. Superposition of the 3D-QSSR model and the non selective compound 55 (CP-
5,940).
5
the quinolone nucleus of compound 24 is surrounded by blue cubes,
which suggests lipophilic substituents in this position might
increase CB2 selectivity. Accordingly, based on our interest in
studying the cannabinergic properties of quinolone derivatives, the
N-(adamantan-1-yl)-4-oxo-8-methyl-1-pentyl-1,4-dihydroquino-
line-3-carboxamide derivative (65) shown in Fig. 8 was considered
as our synthetic test compound. The log(SI) value predicted by using
our model (1.93, see Fig. 8) was indicative of high CB2 selectivity for
such a compound, thus spurring us to synthesize it and to evaluate
its CB1 and CB2 receptor affinity. The synthesis of 65 was carried out
in 5 steps from o-toluidine according to a synthetic protocol
previously utilized for the preparation of other quinolone deriva-
tives [7c]. The synthesis of compound 65 is depicted in Scheme 1.
The binding affinities of compounds 65 for human recombinant CB1
and CB2 receptors were evaluated in parallel with SR144528 [40]
and rimonabant [51] as reference CB2 and CB1 ligands, respec-
tively, as previously described [7c]. The new derivative was found to
be an extremely selective CB2 ligand as it displayed high CB2 affinity
3. Conclusion
The 3D computational model proposed in this study has been
generated taking into account a number of structurally diverse
compounds characterized by different selectivity index values and
might be useful for the discovery of structurally novel selective CB2
ligands. Future studies should provide additional enhancements to
the workflow here employed. Thus, exploiting the repeated
appearance of new selective CB2 scaffolds in the literature, we are
currently enlarging the ligands data set in an attempt to widen its
inclusiveness.
In conclusion, the success of our computational strategy, which
was prospectively tested by an unrelated test set of derivatives
taken from the literature and led to the identification of one new
selective CB2 receptor ligand, lends firm support for the practica-
bility of quantitative ligand-based selectivity models for cannabi-
noid receptors. These drug discovery tools might represent
a valuable complementary approach to docking studies performed
on homology models of the receptors. Phase turned out to be an
appropriate software package to achieve such a goal.
(
Ki ¼ 4.9 nM), while being devoid of CB1 affinity (Ki > 10,000 nM),
with a log(SI) value >3.3.
The comparison between the SI values of compounds 65 and 21
(
8-unsubstituted analogue) demonstrates that the insertion of the
small and lipophilic methyl group into the 8-position enhanced
selectivity by approximately 200 fold, as a result of dramatic
reduction in CB1 affinity.
4
. Methods
4.1. Molecular modeling
Three-dimensional structure building, pharmacophore mapping
and 3D-QSSR studies were carried out on an IBM workstation with
Linux operating system running Maestro 8.0, MacroModel 9.5 and
Phase 2.5 programs (Schrödinger, LLC, New York, NY). Phase,
implemented in the Maestro modeling package, was used to
generate pharmacophore and 3D-QSSR models for cannabinoid
receptor CB2. The 3D structure of all the molecules used in Phase
was built in Maestro. Conformers of each derivative were generated
in MacroModel using the OPLS_2005 force field, GB/SA water and
no cutoff for nonbonded interactions. Molecular energy minimi-
zations were performed using the PRGC method with 5000
maximum iterations and 0.001 gradient convergence threshold.
The conformational searches were carried out by application of the
MCMM torsional sampling method, performing automatic setup
with 20 kJ/mol in the energy window for saving structure and
ꢀ
a 0.5 A cutoff distance for redundant conformers. Pharmacophore
feature sites for the molecules were assigned using a set of features
defined in Phase as hydrogen-bond acceptor (A), hydrogen-bond
donor (D), hydrophobic group (H), negatively charged group (N),
Fig. 4. Superposition of the 3D-QSSR model and the moderately selective compound
61 (SR144528).

552
S. Brogi et al. / European Journal of Medicinal Chemistry 46 (2011) 547e555
Fig. 6. Superpositions of the 3D-QSSR model and compound 24. Only the cubes representing the model that are occupied by the compound are displayed (beyond the ꢁ2.0eꢂ02
threshold), decomposed into contributions by two different atom classes. Left: map corresponding to electron-withdrawing atoms (including hydrogen-bond acceptors). Right: map
corresponding to hydrophobic/non-polar atoms. The blue and red cubes refer to regions in which CB2 selectivity is increased or decreased, respectively, by molecular occupancy.
(
For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article).
positively charged group (P), and aromatic ring (R). Four highly
active compounds (S1eS11eS26eS41) were selected for generating
the pharmacophore hypotheses for CB2 (Fig. S1 in the
Supplementary material). Common pharmacophore hypotheses
were identified using conformational analysis and a tree-based
partitioning technique. The resulting pharmacophores were then
scored and ranked. The best-generated CB2 pharmacophore model
(CB2PHAM) obtained by Phase consisted of five features: one
hydrogen-bond acceptors (A; represent by red vectors), one
aromatic groups (R; orange rings), three hydrophobic functions (H;
green balls) (see Supplementary material for further details). This
pharmacophore was chosen for further 3D-QSSR analysis. All the
Fig. 7. Chemical scaffolds of the compounds included in the external test set (see Table S2 in the Supplementary material for further details).

S. Brogi et al. / European Journal of Medicinal Chemistry 46 (2011) 547e555
553
Table 3
CB1 and CB2 receptor affinity values [Ki(CB1) and Ki(CB2) columns, nM], experi-
mental (Exp column) and estimated (Calc column) selectivity index values [log(SI)]
for compounds of the external test set. Estimated log(SI) values were calculated by
application of 3D-QSSR model (see text).
Cmpd
Ki(CB1)a
Ki(CB2)a
Exp log(SI)
Calc log(SI)
E1 [43]
E2 [44]
E3 [45]
E4 [46]
E5 [25]
E6 [25]
E7 [47]
E8 [19]
3800
270
>10,000
1700
1000
95
2370
42
2100
>10,000
945
24
0.64
422
16
50
8
35.9
6.5
52.6
25.5
4.6
16
2.19
2.62
1.37
2.02
1.30
1.07
1.81
0.81
1.60
2.59
2.31
1.58
1.07
1.82
2.35
1.5
1.58
1.22
0.95
1.85
1.31
2.16
2.30
2.38
3.41
2.3
3.2
2.74
2.6
1.13
0.24
0.63
2.07
2.37
1.6
2.28
2.6
3.16
1.84
1.78
2.13
1.94
1.98
1.63
0.93
1.77
1.27
1.2
2.04
1.78
1.64
1.54
1.74
1.73
2.09
1.58
1.36
0.92
1.32
1.28
1.63
1.8
1.93
2.23
2.43
2.32
2.05
1.59
1.56
0.8
1.28
1.26
1.55
2.03
1.89
1.94
2.25
1.76
E9 [7c]
E10 [7c]
E11 [11]
E12 [11]
E13 [11]
E14 [11]
E15 [11]
E16 [48]
E17 [48]
E18 [48]
E19 [48]
E20 [48]
E21 [48]
E22 [48]
E23 [6]
616
131
220
710
130
200
390
310
11
3.3
3.1
3.9
5.2
23
34
11
23
23
790
470
3400
1800
650
4152
1995
4887
5444
1422
12.3
1.86
0.94
1300
870
53
9
ꢃ
Scheme 1. Synthesis of compound 65. Reagents and conditions: i. EMME, 120 C, 4 h;
E24 [6]
2.7
1.6
9.8
2.7
9.7
3.5
0.91
1.05
0.22
11
3.7
1.2
5.7
10
1.7
8
ii. diphenyl ether, reflux, 16 h; iii. 2.5 N NaOH, reflux, 4 h, then HCl; iv. 1-amino-
E25 [10]
E26 [10]
E27 [10]
E28 [10]
E29 [10]
E30 [26]
E31 [27]
E32 [27]
E33 [49]
E34 [49]
E35 [50]
E36 [50]
E37 [50]
E38 [12]
E39 [12]
ꢃ
adamantane, HOBt, HBTU, DIPEA, DMF, rt, 20 h; v. 1-iodopentane, K
0 h.
2
CO
3
, DMF, 90 C,
2
molecules used for QSSR studies (Fig. 1) were aligned to the best
pharmacophore hypothesis (CB2PHAM). Atom-based QSSR models
were generated for CB2PHAM hypothesis using the 29-member
ꢀ
training set and a grid spacing of 1.0 A. QSSR models containing one
to three PLS factors were generated.
1100
4100
2500
560
4.2. General chemistry
Reagents were obtained from commercial suppliers and used
a
Binding affinities for human recombinant CB1 and CB2 receptors measured by
without further purifications. IR spectra were recorded on a Per-
kineElmer BX FT-IR system. TLC was carried out using Merck TLC
plates Kieselgel 60 F254. Chromatographic purifications were per-
formed on columns packed with Merck 60 silica gel, 23e400 mesh,
for flash technique. Melting points were taken using a Gallenkamp
3
displacement of radioligand [ H]-CP-55,940.
Fig. 8. Compound 65. Chemical structure, predicted selectivity index and superposition between the 3D-QSSR model and the putative bioactive conformation of this derivative.

554
S. Brogi et al. / European Journal of Medicinal Chemistry 46 (2011) 547e555
melting point apparatus and are uncorrected. H NMR and 13C NMR
spectra were recorded at 200 and 50 MHz, respectively, with
a Bruker AC200F spectrometer, and chemical shifts are reported in
1
Pozzuoli, are very grateful to Mr. Marco Allarà for technical
assistance.
d
values, relative toTMS at
d
0.00 ppm. EI low-resolution MS spectra
Appendix. Supplementary data
were recorded using an Agilent 1100 Series LC/MSD spectrometer
with an electron beam of 70 eV. Elemental analyses (C, H, N) were
performed in-house using a PerkineElmer Elemental Analyzer
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.ejmech.2010.11.034.
2
40C.
References
4
.3. Synthesis of N-(adamantan-1-yl)-8-methyl-4-oxo-1-pentyl-
[
1] (a) L.A. Matsuda, S.J. Lolait, M.J. Brownstein, A.C. Young, T.I. Bonner, Nature
46 (1990) 561e564;
b) S. Munro, K.L. Thomas, M. Abu-Shaar, Nature 365 (1993) 61e65.
[2] V. Di Marzo, M. Bifulco, L. De Petrocellis, Nat. Rev. Drug Discov. 3 (2004)
71e784.
1,4-dihydroquinoline-3-carboxamide (65)
3
(
A mixture of 2-methylaniline (1.07 g, 10 mmol) and diethyl
ethoxymethylenemalonate (EMME) (2.16 g, 10 mmol) was heated
at 120 C for 4 h and cooled to room temperature. Diphenyl ether
7
[
3] (a) J. Fernández-Ruiz, J. Romero, G. Velasco, R.M. Tolón, J.A. Ramos,
M. Guzmán, Trends Pharmacol. Sci. 28 (2007) 39e45;
ꢃ
(
15 mL) was added and the reaction mixture was heated at reflux
(b) G. Kunos, D. Osei-Hyiaman, S. Bátkai, K.A. Sharkey, A. Makriyannis, Trends
Pharmacol. Sci. 30 (2009) 1e7.
4] (a) V. Di Marzo, Nat. Rev. Drug Discov. 7 (2008) 438e455;
temperature for 16 h. After cooling to room temperature, 2.5 N
NaOH (20 mL, 50 mmol) was added and the reaction mixture was
refluxed for 4 h. After cooling, 12 N HCl was added to the reaction
mixture allowing the precipitation of the acid derivative, which
was collected by filtration, washed with water, then petroleum
ether, and recrystallized from ethanol to give 8-methyl-1,4-dihy-
droquinoline-4-one-3-carboxylic acid (66) as a beige solid (0.6 g,
[
(b) P. Pacher, S. Batkai, G. Kunos, Pharmacol. Rev. 58 (2006) 389e462.
[5] R.G. Pertwee, A. Thomas, Therapeutic applications for agents that act at CB1
and CB2 receptors. in: P.H. Reggio (Ed.), The Cannabinoid Receptors. Humana
Press, Totowa, 2009, pp. 361e392.
6] I. Sellitto, B.L. Le Bourdonnec, K. Worm, A. Goodman, M.A. Savolainen,
G.H. Chu, C.W. Ajello, C.T. Saeui, L.K. Leister, J.A. Casse, R.N. DeHaven,
C.J. LaBuda, M. Koblish, P.J. Little, B.L. Brogdon, S.A. Smith, R.E. Dolle, Bioorg.
Med. Chem. Lett. 20 (2010) 387e391.
7] (a) R.J. Gleave, P.J. Beswick, A.J. Brown, G.M.P. Giblin, P. Goldsmith,
C.P. Haslam, W.L. Mitchell, N.H. Nicholson, L.W. Page, S. Patel, S. Roomans,
B.P. Slingsby, M.E. Swarbrick, Bioorg. Med. Chem. Lett. 20 (2010) 465e468;
(b) M.G. Cascio, D. Bolognini, R.G. Pertwee, E. Palazzo, F. Corelli, S. Pasquini,
V. Di Marzo, S. Maione, Pharmacol. Res. 61 (2010) 349e354;
(c) S. Pasquini, L. Botta, T. Semeraro, C. Mugnaini, A. Ligresti, E. Palazzo,
S. Maione, V. Di Marzo, F. Corelli, J. Med. Chem. 51 (2008) 5075e5084.
8] G.M.P. 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, R. Green, J. Med.
Chem. 50 (2007) 2597e2600.
[
ꢃ
1
3
0% overall yield): R
f
¼ 0.45 (CH
2 2
Cl /MeOH 95:5); mp: 254 C; H
[
NMR (200 MHz, CDCl
3
):
d
¼ 15.37 (s, 1H), 8.61 (s, 1H), 8.12 (d,
J ¼ 8.0 Hz, 1H), 7.70e7.65 (m, 1H), 7.46e7.40 (m, 2H),
ꢂ1
2
7
.50e2.45 ppm (m, 3H); IR (CHCl
0 eV) m/z: 204 [M þ H] ; Anal. calcd for C11
3
):
n
¼ 1625, 1709 cm ; MS (ESI,
þ
H
9
NO
3
: C 65.02, H 4.46,
N 6.89, found: C 65.32, H 4.36, N 6.69.
[
[
The acid derivative 66 (408 mg, 2.0 mmol) was dissolved in dry
DMF (10 mL) and HOBt (260 mg, 2.0 mmol), HBTU (1.52 g,
4
(
room temperature for 30 min, more DIPEA (0.4 mL, 3.0 mmol) was
added and the reaction mixture was stirred at room temperature
for 20 h. K
.0 mmol), DIPEA (0.4 mL, 3.0 mmol) and 1-aminoadamantane
360 mg, 2.4 mmol) were added to the solution. After stirring at
9] R.G. Pertwee, Br. J. Pharmacol. 156 (2009) 397e411.
[
[
10] J.H.M. Lange, M.A.W. van der Neut, H.C. Wals, G.D. Kuil, A.J.M. Borst, A. Mulder,
A.P. den Hartog, H. Zilaout, W. Goutier, H.H. van Stuivenberg, B.J. van Vliet,
Bioorg. Med. Chem. Lett. 20 (2010) 1084e1089.
11] J.M. Frost, M.J. Dart, K.R. Tietje, T.R. Garrison, J.K. Grayson, A.V. Daza, O.F. El-
Kouhen, B.B. Yao, G.C. Hsieh, M. Pai, C.Z. Zhu, P. Chandran, M.D. Meyers,
J. Med. Chem. 53 (2010) 295e315.
[12] E.J. Gilbert, G. Zhou, M.K.C. Wong, L. Tong, B.B. Shankar, C. Huang, J. Kelly,
B.J. Lavey, S.W. McCombie, L. Chen, R. Rizvi, Y. Dong, Y. Shu, J.A. Kozlowski,
N.Y. Shih, W. Hipkin, W. Gonsiorek, A. Malikzay, C.A. Lunn, L. Favreau,
D.J. Lundell, Bioorg. Med. Chem. Lett. 20 (2010) 608e611.
[13] (a) S. Durdagi, M.G. Papadopoulos, P.G. Zoumpoulakis, C. Koukoulitsa,
T.A. Mavromoustakos, Mol. Divers. 14 (2010) 257e276;
2 3
CO (1.39 g, 10 mmol) and n-pentyl iodide (1.43 mL,
1
0 mmol) were added to the reaction mixture, which was heated at
ꢃ
9
0 C for 20 h, then poured into ice and extracted with AcOEt. The
organic layers were washed with brine, dried over anhydrous
Na SO and evaporated to dryness. The crude residue was purified
by flash column chromatography using CH Cl /MeOH 97:3 as the
eluent to afford the title compound 65 as a white solid (128 mg, 17%
overall yield), which was recrystallized from ethanol: R
¼ 0.78
/MeOH 97:3); mp: 140 C; H NMR (200 MHz, CDCl ):
¼ 9.87 (s, 1H), 8.66 (s, 1H), 8.49e8.40 (m, 1H), 7.45e7.40 (m, 1H),
2
4
2
2
(b) E.K. Tiburu, S. Tyukhtenko, L. Deshmukh, O. Vinogradova, D.R. Janero,
f
A. Makriyannis, Biochem. Biophys. Res. Commun. 384 (2009) 243e248;
(c) E. Cichero, S. Cesarini, L. Mosti, P. Fossa, J. Mol. Model. 16 (2010) 677e691;
(d) J.Z. Chen, X.W. Han, A. Makriyannis, J. Wang, X.Q. Xie, J. Med. Chem. 49
ꢃ
1
(
CH
2
Cl
2
3
d
(
(
2006) 625e636;
7
9
3
3
.34e7.30 (m, 1H), 4.36 (t, J ¼ 7.7 Hz, 2H), 2.73 (s, 3H), 2.12e2.07 (m,
e) A.M. Ferreira, M. Krishnamurthy, B.M. Moore II, D. Finkelstein, D. Bashford,
H), 1.71e1.65 (m, 8H), 1.24e1.20 (m, 4H), 0.83 ppm (t, J ¼ 6.3 Hz,
Bioorg. Med. Chem. 17 (2008) 2598e2606;
1
3
H); C NMR (50 MHz, CDCl
3
):
d
¼ 13.8, 22.2, 23.8, 28.3, 29.6, 30.7,
(f) K. Worm, R.E. Dolle, Curr. Pharm. Des. 15 (2009) 3345e3366.
14] P. Markt, C. Feldmann, J.M. Rollinger, S. Raduner, D. Schuster, J. Kirchmair,
S. Distinto, G.M. Spitzer, G. Wolber, C. Laggner, K.H. Altmann, T. Langer,
J. Gertsch, J. Med. Chem. 52 (2009) 369e378.
[
6.6, 41.8, 51.6, 57.5, 112.7, 124.9, 126.0, 126.2, 129.9, 137.4, 139.5,
50.1, 163.6, 176.8 ppm; IR (CHCl
ꢂ1
1
3
):
n
¼ 1657 cm ; MS (ESI, 70 eV)
þ
m/z: 407 [M þ H] ; Anal. calcd for C26
H
34
N
2
O
2
: C 76.81, H 8.43, N
[15] G.B.L. De Freitas, L.L. da Silva, N.C. Romeiro, C.A.M. Fraga, Eur. J. Med. Chem. 44
2009) 2482e2896.
(
6
.89, found: C 76.51, H 8.58, N 7.09.
[
[
[
[
16] A. Weber, M. Böhm, C.T. Supuran, A. Scozzafava, C.A. Sotriffer, G. Klebe,
J. Chem. Inf. Model. 46 (2006) 2737e2760.
17] M. Fichera, G. Cruciani, A. Bianchi, G. Musumarra, J. Med. Chem. 43 (2000)
4
.4. Biology
2300e2309.
18] J.L. Wiley, I.D. Beletskaya, E.W. Ng, Z. Dai, P.J. Crocker, A. Mahadevan,
R.K. Radzan, B.R. Martin, J. Pharmacol. Exp. Ther. 301 (2002) 679e689.
19] J.W. Huffman, G. Zengin, M.J. Wu, G. Hynd, K. Bushell, A.L. Thompson,
S. Bushell, C. Tartal, D.P. Hurst, P.H. Reggio, D.E. Shelley, M.P. Cassidy,
J.L. Wiley, B.R. Martin, Bioorg. Med. Chem. 13 (2005) 89e112.
CB1 and CB2 binding assays: receptor binding assays were
performed exactly as described previously [7c], using membranes
of cells over-expressing the human recombinant CB1 or CB2
receptors.
[20] R. Silvestri, M.G. Cascio, G. La Regina, F. Piscitelli, A. Lavecchia, A. Brizzi,
S. Pasquini, M. Botta, E. Novellino, V. Di Marzo, F. Corelli, J. Med. Chem. 51
(2008) 1560e1576.
Acknowledgements
[21] S. Durdagi, A. Kapou, A. Kourouli, T. Andreou, S.P. Nikas, V.R. Nahmias,
D.P. Papahatjis, M.G. Papadopoulos, T. Mavromoustakos, J. Med. Chem. 50
(
2007) 2875e2885.
Authors from the University of Siena gratefully acknowledge
[
22] D.P. Papahatjis, S.P. Nikas, T. Andreu, A. Makriyannis, Bioorg. Med. Chem. Lett.
financial support from Siena Biotech S. p. A. Authors from CNR,
12 (2002) 3583e3586.

S. Brogi et al. / European Journal of Medicinal Chemistry 46 (2011) 547e555
555
[
[
[
[
[
[
23] J.W. Huffman, J. Liddle, S. Yu, M.M. Aung, M.E. Abood, J.L. Wiley, B.R. Martin,
Bioorg. Med. Chem. 7 (1999) 2905e2914.
24] A.D. Khanolkar, D. Lu, P. Fan, X. Tian, A. Makriyannis, Bioorg. Med. Chem. Lett.
[40] M. Rinaldi-Carmona, F. Barth, J. Millan, J.M. Derocq, P. Casellas, C. Congy,
D. Oustric, M. Sarran, M. Bouaboula, B. Calandra, M. Portier, D. Shire,
J.C. Breliere, G.L. Le Fur, J. Pharmacol. Exp. Ther. 284 (1998) 644e650.
[41] L. Hanus, A. Breuer, S. Tchilibon, S. Shiloah, D. Goldenberg, M. Horowitz,
R.G. Pertwee, R.A. Ross, R. Mechoulam, E. Fride, Proc. Natl. Acad. Sci. U.S.A. 96
(1999) 14228e14233.
[42] K.J. Valenzano, L. Tafesse, G. Lee, J.E. Harrison, J.M. Boulet, S.L. Gottshall,
L. Mark, M.S. Pearson, W. Miller, S. Shan, L. Rabadi, Y. Rotshteyn, S.M. Chaffer,
P.I. Turchin, Y. Elsemore, M. Toth, L. Koetzner, G.T. Whiteside, Neurophar-
macology 48 (2005) 658e672.
[43] G.H. Chu, C.T. Saeui, K. Worm, D.G. Weaver, A.J. Goodman, R.L. Broadrup,
J.A. Cassel, R.N. DeHaven, C.J. LaBuda, M. Koblish, B. Brogdon, S. Smith, B. Le
Bourdonnec, R.E. Dolle, Bioorg. Med. Chem. Lett. 20 (2009) 5931e5935.
[44] B.B. Yao, G. Hsieh, A.V. Daza, Y. Fan, G.K. Grayson, T.R. Garrison, O. El
Kouhen, B.A. Hooker, M. Pai, E.J. Wensink, A.K. Salyers, P. Chandran,
C.Z. Zhu, C. Zhong, K. Ryther, M.E. Gallagher, C.L. Chin, A.E. Tovcimak,
V.P. Hradil, G.B. Fox, M.J. Dart, P. Honore, M.D. Meyer, J. Pharmacol. Exp.
Ther. 1 (2009) 141e151.
[45] M. Naguib, P. Diaz, J.J. Xu, F. Astruc-Diaz, S. Craig, P. Vivas-Mejia, D.L. Brown,
Br. J. Pharmacol. 7 (2008) 1104e1116.
[46] H. Ohta, T. Ishizaka, M. Yoshinaga, A. Morita, Y. Tomishima, Y. Toda, S. Saito,
Bioorg. Med. Chem. Lett. 17 (2007) 5133e5135.
[47] H. Iwamura, H. Suzuki, Y. Ueda, T. Kaya, T. Inaba, J. Pharmacol. Exp. Ther. 296
(2001) 420e425.
[48] A.J. Goodman, C.V. Ajello, K. Worm, B. Le Bourdonnec, M.A. Savolainen,
H. O’Hare, J.A. Cassel, G.J. Stabley, R.N. DeHaven, C.J. LaBuda, M. Koblish,
P.J. Little, B.L. Brogdon, S.A. Smith, R.E. Dolle, Bioorg. Med. Chem. Lett. 19
(2009) 309e313.
9
(1999) 2119e2124.
25] P.L. Ferrarini, V. Calderone, T. Cavallini, C. Manera, G. Saccomanni, L. Pani,
S. Ruiu, G.L. Gessa, Bioorg. Med. Chem. 12 (2004) 1921e1933.
26] M. Krishnamurthy, S. Gurley, B.M. Moore II, Bioorg. Med. Chem. 16 (2008)
6489e6500.
27] A.K. Nadipuram, M. Krishnamurthy, A.M. Ferreira, W. Li, B.M. Moore II, Bioorg.
Med. Chem. 11 (2003) 3121e3132.
28] 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, J. Med. Chem. 51 (2008) 1904e1912.
29] S.L. Dixon, A.M. Smondyrev, E.H. Knoll, S.N. Rao, D.E. Shaw, R.A. Friesner,
J. Comput. Aided Mol. Des. 20 (2006) 647e671.
[
[
[
30] Phase, Version 2.5, User Manual. Schrödinger Press, LLC, New York, 2005.
31] S. Pasquini, A. Ligresti, C. Mugnaini, T. Semeraro, L. Cicione, M. De Rosa, F. Guida,
L. Luongo, M. De Chiaro, M.G. Cascio, D. Bolognini, P. Marini, R. Pertwee,
S. Maione, V. Di Marzo, F. Corelli, J. Med. Chem. 53 (2010) 5915e5928.
32] R.G. Pertwee, Curr. Med. Chem. 6 (1999) 635e664.
[
[
33] M.M. Ibrahim, H. Deng, A. Zvonok, D.A. Cockayne, J. Kwan, H.P. Mata,
T.W. Vanderah, J. Lai, F. Porreca, A. Makriyannis, T.P. Malan Jr., Proc. Natl. Acad.
Sci. U.S.A. 100 (2003) 10529e10533.
[
34] R.A. Ross, H.C. Brockie, L.A. Stevenson, V.L. Murphy, F. Templeton,
A. Makriyannis, R.G. Pertwee, Br. J. Pharmacol. 126 (1999) 665e672.
35] K. Worm, D.G. Weaver, R.C. Green, C.T. Saeui, D.M.S. Dulay, W.M. Barker,
J.A. Cassel, G.J. Stabley, R.N. DeHaven, C.J. LaBuda, M. Koblish, B.L. Brogdon,
S.A. Smith, R.E. Dolle, Bioorg. Med. Chem. Lett. 19 (2009) 5004e5008.
36] R.G. Pertwee, T.M. Gibson, L.A. Stevenson, R.A. Ross, W.K. Banner, B. Saha,
R.K. Razdan, B.R. Martin, Br. J. Pharmacol. 129 (2000) 1577e1584.
37] J. De Vry, D. Denzer, E. Reissmueller, M. Eijckenboom, M. Heil, H. Meier,
F. Mauler, J. Pharmacol. Exp. Ther. 310 (2004) 620e632.
[
[
[
[
[49] H. Ohta, T. Ishizaka, M. Tatsuzuki, M. Yoshinaga, I. Iida, T. Yamaguchi,
Y. Tomishima, N. Futaki, Y. Toda, S. Saito, Bioorg. Med. Chem. 16 (2008)
1111e1124.
[50] H. Ohta, T. Ishizaka, M. Tatsuzuki, M. Yoshinaga, I. Iida, Y. Tomishima, Y. Toda,
S. Saito, Bioorg. Med. Chem. Lett. 17 (2007) 6299e6304.
[51] (a) M. Rinaldi-Carmona, F. Barth, M. Héaulme, D. Shire, B. Calandra, C. Congy,
S. Martinez, J. Maruani, G. Néliat, D. Caput, P. Ferrara, P. Soubrié, J.C. Brelière,
G. Le Fur, FEBS Lett. 350 (1994) 240e244;
38] G. Murineddu, P. Lazzari, S. Ruiu, A. Sanna, G. Loriga, I. Manca, M. Falzoi,
C. Dessì, M.M. Curzu, G. Chelucci, L. Pani, G.A. Pinna, J. Med. Chem. 49 (2006)
7502e7512.
[39] E. Stern, G.G. Muccioli, R. Millet, J.F. Goossens, A. Farce, P. Chavatte, J.H. Poupaert,
D.M. Lambert, P. Depreux, J.P. Hénichart, J. Med. Chem. 49 (2006) 70e79.
(b) L.A. Sorbera, J. Castaner, J.S. Silvestre, Drugs Future 30 (2005) 128e137.