6868
V. Corradi et al. / Bioorg. Med. Chem. Lett. 21 (2011) 6867–6871
recently discovered by our research group using computer-aided
methods.9
This compound (BV02; Fig. 1) inhibits the interaction between
14-3-3
r
and c-Abl in vitro with a lethal dose 50 (LD50) in the
M). BV02 promoted the nucle-
low micromolar range (LD50 = 1.04
l
ar localization of c-Abl and stimulated the pro-apoptotic signals in
Ba/F3 cells expressing the wt and the T315I Bcr-Abl construct.10,11
Due to its peculiar mechanism of action, BV02 was considered as a
candidate for the treatment of Chronic Myeloid Leukemia (CML),
and particularly to overcome the drug resistance associated with
the disease progression. Unfortunately, later NMR studies revealed
that BV02 underwent to a spontaneous chemical cyclization at
room temperature, being the ‘open’ 2-carbamoyl benzoic moiety
in equilibrium with its ftalimidic ‘closed’ form (see Supplementary
data). This equilibrium is pH dependent and was observed either
by solubilising the open derivative purchased from Asinex vendor
(purity >95%) or the one obtained by chemical synthesis.11 From a
medicinal chemistry point of view, the simultaneous existence of
two closely related equilibrium structures of the same hit com-
pound could represents an obstacle for undertaking further ‘hit
to lead’ optimization steps, since the identity of the chemical moi-
ety responsible for the observed biological activity could not be
easily determined. Accordingly, BV02 was not further investigated,
but the identification of closely related analogues that may not un-
dergo to such rearrangement is currently under study.
In the present work we wish to report how the application of
molecular modeling coupled with biophysical and biochemical
techniques could efficiently lead to the identification of protein–
protein interaction inhibitors. Considering the chemical stability
problem experienced with BV02, computational results have been
carefully revised in order to discard those compounds that may
suffer from possible spontaneous chemical rearrangements at
room temperature.
The Virtual Screening (VS) protocol based on the sub-sequential
application of pharmacophore modeling and ligand docking, which
has been previously described,9 has been used in this work to
screen the whole Asinex database in search for potential 14-3-
3r/c-Abl protein–protein interaction inhibitors. In addition, to
overcome the steric restrictions intrinsic in structure-based phar-
macophoric modeling, high throughput docking (HTD) has been
also applied alongside the VS. In the previously described approach
the virtual screening was performed against the Asinex Gold col-
lection, which consists of about 200,000 compounds. Here, we used
the whole Asinex library, which contains more than 600,000 small
molecules. Therefore, it should not be surprising if different results
were obtained with respect to those previously discussed.9 In addi-
tion, only compounds endowed with better scoring values than
BV02 were selected from the database. As a result, a small set of
the best ranking compounds identified by both the VS and the
HTD protocols were purchased for a preliminary biological evalua-
tion. Among them, compounds BV01 and BV101 (Fig. 1) showed
considerable cytotoxicity against Bcr-Abl-expressing Ba/F3 cells
and were further investigated.
Fig. 2. Docking-based binding mode of BV02 (the reference hit compound), BV01
and BV101 within the amphipathic groove of 14-3-3 (coordinates of the protein
were taken from the crystallographic structure coded by PDB 1YWT). The surface of
14-3-3 is coloured according to its electrostatic potential. Blue = positive charge.
Red = negative charge. Colour intensity is proportional to the charge density. The
negatively charged ‘heads’ of BV02 and BV01 interact in a positively charged region
of the protein (basic residues are labelled).
r
r
molecular fragments,13 was herein used to compare BV01 and
BV101 against the reference BV02. The calculation of the Tanimoto
coefficient was out carried with Discovery Studio 3.014 by using
both the FCFP_6 and ECFP_6 sets of fingerprints, which gave for
BV01 a value of 0.1489 and 0.1351, respectively. This means that
that this compound is considerably different from BV02. BV101 is
the 4-nitro-2-[(4-phenyl-6-p-tolyl-pyrimidin-2-yl)-hydrazonom-
ethyl]-phenol and, similarly to BV02, shows very few degrees of
freedom. However, the chemical diversity with respect to BV02 is
high as in the case of BV01. The Tanimoto similarity index calcu-
lated by using both the FCFP_6 and ECFP_6 sets of fingerprints is
0.094 and 0.1065, with respect to BV02. Due to the lack of the
carboxylic group, BV101 shows a less pronounced negative charge
on the scaffold ‘head’ with respect to the other compounds. Its
BV01 is the 4-(4-hexyloxycarbonyl-phenylamino)-3,5-dinitro-
benzoic acid and presents a different scaffold than the reference
compound BV02. The only similarity is represented by the negative
charge distributed on the ‘head’ of the inhibitor, which, according
to docking studies, should interact in the 14-3-3r amphipathic
groove in close proximity of the cluster of basic residues consisting
of K49, R56, R60 and R129 (Fig. 2). This is also in agreement with
the orientation of the phosphate group of the bound phospho-pep-
tide in the crystal structure (PDB code: 1YWT).12
The ‘tail’ of BV01 is represented by an aliphatic and very flexible
lipophilic chain, whereas BV02 shows a conjugated aromatic sys-
tem. The Tanimoto coefficient, which describes the similarity be-
tween two compounds based on the presence or absence of