CYP19 (aromatase): Exploring the scaffold flexibility for novel selective inhibitors

Article abstract of DOI:10.1016/j.bmc.2008.08.046

Several derivatives out of a series of antifungal agents exhibited a good inhibitory potency against aromatase as well as a fairly good selectivity toward CYP17, even if lacking H-bond accepting substituents. Their common structural feature is a flexible backbone that did not fit into previously reported CYP19 models. Thus, a ligand-based approach was exploited to develop a novel statistically robust, self-consistent and predictive 3D-QSAR model herein proposed as a helpful tool to design new aromatase inhibitors.

Full text of DOI:10.1016/j.bmc.2008.08.046

Bioorganic & Medicinal Chemistry 16 (2008) 8349–8358
Contents lists available at ScienceDirect
Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com/locate/bmc
CYP19 (aromatase): Exploring the scaffold flexibility for novel
selective inhibitors
b
c,
d
a
Sabrina Castellano a, , Giorgio Stefancich , Rino Ragno , Katarzyna Schewe , Marisabella Santoriello ,
*
*
Antonia Caroli c, Rolf W. Hartmann d, Gianluca Sbardella a
a Dipartimento di Scienze Farmaceutiche, Università di Salerno, Via Ponte Don Melillo, I-84084 Fisciano (SA), Italy
b Dipartimento di Scienze Farmaceutiche, Università di Trieste, P.le Europa 1, I-34127 Trieste, Italy
c Dipartimento di Chimica e Tecnologie del Farmaco, Università degli studi di Roma ‘La Sapienza’, P.le Aldo Moro 5, I-00185 Roma, Italy
d Pharmaceutical and Medicinal Chemistry, Saarland University, PO Box 151150, D-66041 Saarbrücken, Germany
a r t i c l e i n f o
a b s t r a c t
Article history:
Several derivatives out of a series of antifungal agents exhibited a good inhibitory potency against aro-
matase as well as a fairly good selectivity toward CYP17, even if lacking H-bond accepting substituents.
Their common structural feature is a flexible backbone that did not fit into previously reported CYP19
models. Thus, a ligand-based approach was exploited to develop a novel statistically robust, self-consis-
tent and predictive 3D-QSAR model herein proposed as a helpful tool to design new aromatase inhibitors.
Ó 2008 Elsevier Ltd. All rights reserved.
Received 20 May 2008
Revised 9 August 2008
Accepted 22 August 2008
Available online 26 August 2008
Keywords:
Aromatase inhibitors
CYP19
3D-QSAR
Ligand-based drug design
Antifungal agents
1. Introduction
tives (Fig. 1), were actually developed in the last two decades.6–25
Among the latter, the recently FDA-approved anastrozole26 and
Estrogen-dependent (ER+) breast cancer accounts for approxi-
mately one-third of all breast cancer patients, and two-thirds of
cases of postmenopausal breast cancer.1,2 This tumor contains
estrogen receptors and requires estrogens for tumor growth and
indeed estrogens are maintained in postmenopausal women breast
tissue nearly at the same level as in premenopausal women.3
Aromatase is a cytochrome P450 (CYP19; EC 1.14.14.1) enzyme
that catalyzes the conversion of androgens androstenedione and
testosterone to the aromatic estrogenic steroids estrone and estra-
diol, respectively, through the aromatization of the A ring of the
substrate.4 This enzyme is present in breast tissue and is an impor-
tant pharmacological target in the anti-cancer therapy, because
intratumoral aromatase is the source of local estrogen production
in breast cancer tissues.5 As a matter of fact, suppression of estro-
gen biosynthesis by aromatase inhibition represents an effective
approach for the treatment of hormone-sensitive breast cancer
and several classes of steroidal and nonsteroidal AR inhibitors
(AIs), such as aminoglutethimide and imidazole or triazole deriva-
letrozole,27 as well as the steroid exemestane,28 are widely used,
even as the first-line drugs in the therapy of breast cancer.29,30
However, the occurrence of important side effects associated
with the prolonged clinical use of AIs31,32 calls for the search of
new, potent, more selective, and less toxic CYP19 inhibitors.
One of the most important feature for strong inhibitor binding
to CYP enzymes is the capability to interact as the sixth ligand with
the iron atom of the heme group, always present in this enzyme
family. This coordination is excellently performed by the lone pair
carried on the sp2 hybridized imidazole nitrogen but other electron
rich heterocycles might be accepted in this position as well. There-
fore we were rather surprised that the introduction of 1-amino-
azoles was little studied in the development of AIs, particularly
if considering the high potency and selectivity of compound
YM-511 (Fig. 1),16,25 the most known among the derivatives con-
taining this moiety.
During our search for novel antifungal agents, we previously
reported the synthesis of a large number of compounds,33–36 many
of which were characterized by the presence of a common scaffold
containing a central 1-amino-azole moiety and two aromatic por-
tions separated from the central one by methylenic linkers (Fig. 2).
As the pursuing of new entities possessing aromatase inhibitory
properties is still active, we decided to firstly explore the potential
* Corresponding authors. Tel.: +39 089 96 9244; fax: +39 089 96 9602 (S.C.); tel.:
+39 06 4991 3937; fax: +39 06 491 491 (R.R.).
E-mail addresses: scastellano@unisa.it (S. Castellano), rino.ragno@uniroma1.it
(R. Ragno).
0968-0896/$ - see front matter Ó 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.bmc.2008.08.046

8350
S. Castellano et al. / Bioorg. Med. Chem. 16 (2008) 8349–8358
N
H
N
N
N
O
O
N
CN
N
H3C
H3C
H3C
NH2
CN
H3C
CN
CH3
( )-aminogluthethimide
anastrozole
( )-fadrozole
N
N
N
N
N
NC
CN
NC
CN
CGS-18320B
letrozole
N
N
N
N
N
N
Br
N
N
N
N
CH3
N
N
OH
N
Cl
NC
NC
F
( )-vorozole
YM-511
MPV-2213ad
Figure 1. Nonsteroidal aromatase inhibitors.
of this template, before decorating it with hydrogen bond accep-
tors at an appropriate distance from the heterocycle, credited to
provide auxiliary interactions that could further improve the bind-
ing affinity and the potency of the inhibitors.21,37–39
Therefore, we tested the in vitro inhibiting activity of deriva-
tives 1–4 (Fig. 2) against aromatase and against 17a-hydroxy-
lase/17,20-lyase (CYP17), another cytochrome P450 involved in
the synthesis of androgens, to evaluate their selectivity toward a
related enzyme. The in vitro assays evidenced a very good inhibi-
tory activity against aromatase for a few tested derivatives. Unex-
pectedly, we were able to obtain (vide infra) only a poor correlation
between the activities of these compounds and their fitting into
the models precedently described by other groups in collaboration
with one of us (R.W.H.).20–24,40–43
2- chloro-5-(chloromethyl)thiophene or with 7-(bromomethyl)-
benzothiophene47 in the presence of tetrabutylammonium hydro-
gen sulfate in dichloromethane/50% sodium hydroxide aqueous
solution to furnish derivatives 1bq, 1eq, 1nq, and 1nr. The alkyl-
ation of the amine 6n with 7-(bromomethyl)benzothiophene was
accomplished only in the presence of potassium tert-butoxide
and 18-crown-6 in diethyl ether to obtain the final compound
1er (Scheme 1).35
Derivatives 2bq and 2eq were in turn prepared by alkylation of
the imidazolylacetophenones 7b48 and 7e49 with 2-chloro-5-(chlo-
romethyl)thiophene in THF in the presence of sodium hydride to
give the intermediate ketones 8bq and 8eq which were reduced
following the Huang–Minlon modification of the Wolff–Kishner
reaction36 to yield the final derivatives 2bq and 2eq (Scheme 2).
For this reason we herein describe the building of a novel robust
and more versatile 3D-QSAR model and the validation of its ability
to predict the aromatase inhibiting activity of a set of newly syn-
thesized derivatives (Fig. 3).
2.2. Biological assays
Compounds 1–4 were tested for inhibition of aromatase using
human placental microsomes incubated with 1b[3H]androstenedi-
one and measuring the tritiated water formed during the aromati-
zation of the substrate, as previously described.44 For the CYP17
inhibition tests, human CYP17 expressed in Escherichia coli, P450
reductase and progesterone as substrate were used.45 The IC50 va-
lue of fadrozole tested in our test system46 is also reported for
comparison (Table 1).
Interestingly, even if lacking H-bond accepting substituents as
CN or NO2,21,37–39 several among the tested derivatives exhibited
a good inhibitory potency against aromatase (compare, e.g., IC50
values of compounds 1bk and 2bk with that of the reference com-
pound fadrozole) as well as a fairly good selectivity toward CYP17.
As a general rule imidazole-containing derivatives resulted more
active than the triazole (both symmetrical 1,3,4 and asymmetrical
1,2,4) counterparts (compare IC50 values of compounds 1bd and
1bn with those of compounds 3bd, 3bn and 4bd and 4bn, respec-
2. Results and discussion
2.1. Chemistry
The novel derivatives 1bq, 1eq, 1nq, 1er, 1nr, 2bq, and 2eq
were prepared (Schemes 1 and 2) following the same synthetical
route previously reported by us.33–36
Briefly, the 1-aminoimidazole derivatives 1 were prepared
starting from the Schiff bases 5b, 5e, and 5n, synthesized by adding
imidazole to a neutral solution of hydroxylamine-O-sulfonic acid in
water and then treating the crude residue obtained after acidificat-
ion and elimination of the water with a solution of the proper aro-
matic aldehyde in ethanol.34 The reduction of the Schiff bases with
sodium borohydride in methanol yielded the corresponding sec-
ondary amines 6b, 6e, and 6n which were finally alkylated with

S. Castellano et al. / Bioorg. Med. Chem. 16 (2008) 8349–8358
8351
X
N
Y
N
Z
R
R1
N
N
N
N
1
a
b
c
d
e
N
2
Me
Cl
F
Me
Cl
F
F
N
N
f
h
k
N
N
g
i
j
3
MeO
OMe Cl
Cl F
N
N
S
N
N
n
o
p
l
m
4
Figure 2. Antifungal compounds 1–4 tested as aromatase inhibitors. In the numeration of the compounds the number (1–4) indicates the azole moiety while the two letters
(a–p) specify the combination of aromatic moieties R and R1. The various alternatives for X, Y, and Z are as shown in structures 1–4.
tively). As far as the aromatic moieties (R and R1, Fig. 2) are con-
cerned, in agreement with what was recently reported by Potter
and co-workers39 the introduction of a biphenyl system as R sub-
stituent (Fig. 2) was beneficial for the activity of the compounds
while the 1-naphthyl-substituted derivatives resulted less active
(compare IC50 values of compounds 1ad and 1an with those of
2bd and 2bn, respectively). Nevertheless, the replacement of the
biphenyl group with a 4-tert-butylphenyl further improved the
efficacy of the derivatives, even if a decrease in selectivity was also
observed (compare IC50 values of compounds 1bn and 1dn). As
regards to the R1 substituent (Fig. 2), the effects of the introduction
of 2- or 4-substituted phenyl rings seemed to correlate with the
electronegativity of the substituents (F > Cl > H), that is, with the
ability of the latter to act as H-bond acceptors, particularly if in
the para-position of the benzene ring. In fact, 4-fluorophenyl deriv-
atives 1bk and 2bk were the most active among all tested com-
pounds, being only 3.7- and 2.2-fold less active than fadrozole,
respectively (Table 1). Moreover, sterical hindrance appeared to
be detrimental for the activity (compare, e.g., IC50 values of com-
pounds 1be, 1bi, and 1bd). Interestingly, the replacement of the
substituted benzene with a thiophene ring resulted in an enhance-
ment of both the inhibitory potency and the selectivity toward
CYP17 (compare IC50 values of compounds 1dn and 1dp).
with the exception of 2-halo-substituted compounds which
resulted threefold more active in the 1-amino-imidazole series 1
(compare IC50 values of compounds 1bg and 1bh with those of
compounds 2bg and 2bh, respectively).
2.3. 3D-QSAR studies
In order to gain further insight, a three-dimensional quantita-
tive structure–activity relationship (3D-QSAR) model was then
developed using a training set with a large structural variety con-
taining derivatives 1–2 and 4ac, 4cn, 4dc, and 4dp50 (31 com-
pounds) plus additional 90 compounds (Supplementary data)
previously reported as aromatase inhibitors.20–22,40–43,51 It is worth
mentioning that, to avoid any lack of homogeneity in the biological
data, all the selected compounds had been tested in our laboratory
and following the same experimental protocol.
The structures were built using the standalone version of the
program PRODRG2.52–54 Due to the lack of any structural informa-
tion we tried to perform a structure-based alignment using a
CYP19 homology model developed by Favia et al. (pdb entry code
1tqa).55 Nevertheless any attempt to build a statistically significant
3D-QSAR model failed, maybe due to the lack of the correct Fe
parameters and also to the fact that in the aforementioned paper
by Favia et al. only the aromatase minimized average structure of
the molecular dynamics studies was available. Therefore we
turned to develop a ligand-based alignment by means of the pro-
gram Surflex developed by Jain.56 For the flexible Surflex-based
The presence of the nitrogen atom directly linked to the N1 of
the azole moiety did not significantly influenced the inhibiting
capability of tested molecules. In fact, the activities of derivatives
1 were comparable or slightly inferior to those of derivatives 2,

8352
S. Castellano et al. / Bioorg. Med. Chem. 16 (2008) 8349–8358
N
N
N
Cl
N
N
N
N
N
N
Cl
Cl
Cl
S
S
S
Cl
1bq
1eq
1nq
N
N
Cl
N
N
N
N
S
Cl
S
1er
1nr
N
N
N
N
Cl
Cl
S
S
2bq
2eq
Figure 3. Newly synthesized compounds used to test the predictive ability of the 3D-QSAR model described in the paper.
R1
R
R
N
N
c
a, b
d or e
N
N
N
NH
N
N
N
N
R
N
H
6b,6e,6n
1bq
1nq
1nr
5b
, R= biphenyl
, R= biphenyl
5e, R= phenyl
5n, R= 2,4-dichlorophenyl
R1= 5-chlorothiophen-2-yl
1eq, R= phenyl
R1= 5-chlorothiophen-2-yl
, R= 2,4-dichlorophenyl
R1= 5-chlorothiophen-2-yl
1er, R= phenyl
R1= benzothiophen-7-yl
, R= 2,4-dichlorophenyl
R1= benzothiophen-7-yl
Scheme 1. Reagents and conditions: (a) NH2OSO3H, water, NaHCO3; (b) biphenyl-4-carbaldehyde or benzaldehyde or 2,4-dichlorobenzaldehyde, ethanol, HCl; (c) NaBH4,
methanol; (d) 2-chloro-5-(chloromethyl)thiophene or 7-(bromomethyl)benzothiophene, 50% aq NaOH, TBAHS, dichloromethane; (e) 7-(bromomethyl) benzothiophene,
tBuOK, 18-C-6, diethyl ether.
N
N
N
R2
R2
R2
N
N
N
a
b
Cl
Cl
S
S
O
O
2
7b,7e
R2= H, Ph
8bq,8eq
2bq
2eq
, R = phenyl
2
, R = H
Scheme 2. Reagents and conditions: (a) 2-chloro-5-(chloromethyl)thiophene, NaH, THF, 50 °C; (b) hydrazine hydrate, KOH, diethyleneglycol, 190 °C.
alignment (Fig. 4) the crystal structures of the three highly active
aromatase inhibitors CGS-18320B,57 fadrozole,57 and vorozole58
(Fig. 1) were used as reference molecules (Supplementary data).
Once the structures were aligned, a GRID/GOLPE59 procedure was
applied to buildthe 3D-QSARmodel. To this purpose, thewater, DRY,
OH and C@GRID probes alone and combinations of them (DRY+OH
and DRY+water) were tried and only the preliminary GRID/GOLPE
models showing the higher q2 values were refined by sequential
fractional factorial design (FFD) selections. The DRY+OH combina-
tion gave the best statistical results (Table 2). Furthermore since
many of the molecules included in the training set contained a chiral
center, two models were initially built: one including only com-
pounds with R-configuration (R-model) and the second containing
only derivatives with S-absolute configurations (S-model). Achiral
compounds were equally added to both models. Yet, the statistical
results obtained by the use of different probes for each model were
in favor of the R-model derived with DRY+OH probes combination,
thus we decided to discard the S-model.
The statistical parameters listed in Table 2 show that, in terms
of conventional correlation (r2) and cross-validation (q2 and SDEP)
coefficients, the model resulted statistically robust with values of
0.86, 0.70 and 0.54, respectively (cross-validation 5 random groups
in GOLPE60 program), while the Leave One Out validation method
gave q2 and SDEP values of 0.74 and 0.50, respectively.

S. Castellano et al. / Bioorg. Med. Chem. 16 (2008) 8349–8358
8353
Table 1
Inhibition of CYP19 and CYP17 by derivatives 1–4
X
N
Y
N
R1
R
Z
R1
X
Y
Z
CYP19a IC50
(
lM) or % inhib. (36
lM)
CYP17b % inhib. (2.0
lM)
c
Compound
R
1ad
1an
1bd
1be
1bf
1bg
1bh
1bi
1-Naphthyl
1-Naphthyl
4-tert-Butylphenyl
2,4-Dichlorophenyl
4-tert-Butylphenyl
Phenyl
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
N
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
N
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
N
N
N
N
N
N
N
N
N
4.59
4.00
2.16
1.20
2.96
0.30
0.46
2.2
1.18
0.19
1.51
1.53
3.65
0.35
1.52
0.58
1.00
0.95
1.04
1.09
1.37
1.32
0.11
0.51
1.68
1.24
0.26
(14.8%)
NA
(8.8%)
(5.9%)
(8.8%)
NA
13
11.35
4-40-Biphenyl
4-40-Biphenyl
4-40-Biphenyl
4-40-Biphenyl
4-40-Biphenyl
4-40-Biphenyl
4-40-Biphenyl
4-40-Biphenyl
4-40-Biphenyl
4-40-Biphenyl
4-40-Biphenyl
4-40-Biphenyl
4-tert-Butylphenyl
4-tert-Butylphenyl
4-40-Biphenyl
4-40-Biphenyl
4-40-Biphenyl
4-40-Biphenyl
4-40-Biphenyl
4-40-Biphenyl
4-40-Biphenyl
4-40-Biphenyl
4-40-Biphenyl
4-40-Biphenyl
4-40-Biphenyl
1-Naphthyl
NAd
NA
2-Tolyl
20.57
29.43
NA
29.07
13.57
13.90
42.87
21.67
25.75
NA
43.30
10.70
NA
17.10
25.00
22.55
NA
32.80
15.83
30.20
38.80
18.23
17.70
6.10
22.37
4.43
16.10
15.95
13.05
NA
2-Chlorophenyl
2-Fluorophenyl
4-Tolyl
1bj
1bk
1bl
4-Chlorophenyl
4-Fluorophenyl
2-Methoxyphenyl
4-Methoxyphenyl
2,4-Dichlorophenyl
2,4-Difluorophenyl
2,4-Dichlorophenyl
2-Thienyl
1bm
1bn
1bo
1dn
1dp
2bd
2be
2bf
2bg
2bh
2bi
2bk
2bl
4-tert-Butylphenyl
Phenyl
2-Tolyl
2-Chlorophenyl
2-Fluorophenyl
4-Tolyl
4-Fluorophenyl
2-Methoxyphenyl
4-Methoxyphenyl
2,4-Dichlorophenyl
2,4-Difluorophenyl
Cinnamyl
4-tert-Butylphenyl
2,4-Dichlorophenyl
Cinnamyl
4-tert-Butylphenyl
2,4-Dichlorophenyl
2,4-Dichlorophenyl
Cinnamyl
2,4-Dichlorophenyl
2-Thienyl
Cinnamyl
4-tert-Butylphenyl
2,4-Dichlorophenyl
Cinnamyl
4-tert-Butylphenyl
2,4-Dichlorophenyl
2,4-Dichlorophenyl
Cinnamyl
2bm
2bn
2bo
3ac
3ad
3an
3bc
3bd
3bn
3cn
3dc
3dn
3dp
4ac
4ad
4an
4bc
4bd
4bn
4cn
4dc
4dn
4dp
Fadrozole
1-Naphthyl
1-Naphthyl
4-40-Biphenyl
4-40-Biphenyl
4-40-Biphenyl
Cinnamyl
4-tert-Butylphenyl
4-tert-Butylphenyl
4-tert-Butylphenyl
1-Naphthyl
1-Naphthyl
1-Naphthyl
4-40-Biphenyl
4-40-Biphenyl
4-40-Biphenyl
Cinnamyl
4-tert-Butylphenyl
4-tert-Butylphenyl
4-tert-Butylphenyl
N
N
N
N
N
N
N
N
(10.6%)
(38.3%)
(3.3%)
(30.7%)
13.04
(14.6%)
(24.95%)
(31.4%)
(19.8%)
(27.6%)
17.13
5.06
19.93
4.00
NA
15.47
3.13
NA
12.40
6.03
9.20
21.05
12.40
3.33
N
N
N
N
N
N
N
N
N
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
N
N
N
N
N
N
N
N
N
N
N
2,4-Dichlorophenyl
2-Thienyl
(36%)
9.23
0.05
N
NA
a
Human aromatase, placental microsomes, and substrate [1b,2b-3H]testosterone, 500 nM.
Human CYP17 expressed in E. coli, substrate progesterone, 25 lM.
The given values are mean values of at least three experiments. The deviations were within 5%.
NA, no activity detected.
b
c
d
Besides the statistical evaluation, the assessment of its predic-
tive ability is also an essential requirement for a 3D-QSAR model.
Therefore with the aim to test the potential use of this model as
an aid to design new aromatase inhibitors, an external data set
(test set) of novel derivatives belonging to the series 1 and 2 (deriv-
Table 2
Statistical results of R-model with DRY+water (upper lines values) and DRY+OH
(lower lines values) probes combination
q2
SDEPLOO
q2
r2
CV(5)
SDEPCV(5)
Variablesa
PCsb
LOO
CV(5)
0.66
0.74
0.58
0.50
0.64
0.70
0.82
0.86
0.60
0.54
1958
1475
3
3
a
Number of GRID variables after the FFD selections.
Number of principal components.
b
Figure 4. The 124 compounds of the training set as aligned by Surflex.

8354
S. Castellano et al. / Bioorg. Med. Chem. 16 (2008) 8349–8358
Table 3
Inhibition of CYP19 and CYP17 by derivatives 1bq, 1eq, 1nq, 1er, 1nr, 2bq, and 2eq
N
N
Z
R1
R
R1
Z
CYP19a IC50
(
lM)
CYP17b % inhib. (2.0
lM)
c
Compound
R
1bq
1eq
1nq
1er
1nr
2bq
2eq
4-40-Biphenyl
Phenyl
2,4-Dichlorophenyl
Phenyl
2,4-Dichlorophenyl
4-40-Biphenyl
Phenyl
5-Chloro-2-thienyl
5-Chloro-2-thienyl
5-Chloro-2-thienyl
7-Benzothienyl
7-Benzothienyl
5-Chloro-2-thienyl
5-Chloro-2-thienyl
N
N
N
N
N
CH
CH
1.04
1.06
0.32
9.96
2.49
1.07
0.87
23.60
7.30
56.20
5.25
4.07
25.70
7.97
a
Human aromatase, placental microsomes, and substrate [1b,2b-3H]testosterone, 500 nM.
b
c
Human CYP17 expressed in E. coli, substrate progesterone, 25
lM.
The given values are mean values of at least three experiments. The deviations were within 5%.
atives 1bq, 1eq, 1nq, 1er, 1nr, 2bq, and 2eq, Fig. 3) was subjected
to the 3D-QSAR model and then tested against CYP19 (and CYP17
as well) following the same protocols reported above. The results
of the biological assays are reported in Table 3.
Indeed, although the pIC50 values of the newly synthesized mol-
ecules did not span over a wide range, their prediction was fairly
good. The correlations between experimental and calculated/pre-
dicted pIC50 values (SDEP = 0.99) are shown in Figure 5.
Analysis of the GOLPE60 PLS coefficients plots revealed that
three areas (A, B, and C in Fig. 6) surrounding the ligand-based
aligned molecules are important for the activity. In all the mole-
cules of the training set the A region (green circle) is always occu-
pied by substituents, thus highlighting the importance of this area
in modulating the inhibitory potency. In fact, the large cyan-col-
ored polyhedrons in this area are related to significant electrostatic
and hydrogen bond interactions.
On the other side, the poor presence of large polyhedrons in the
C region (blue circle) may suggest that this region is not significa-
tive for the activity. Nevertheless, most of the training set mole-
cules are highly superimposed in this area, leading to low
standard deviation on the grid points surrounding it. As a matter
of fact, the PLS algorithm considers the regions corresponding to
Figure 6. GRID plots of the PLS coefficients for the 3D-QSAR model. The cyan
contours represent negative coefficients under ꢀ0.052 energy value while the
yellow contours represent the positive coefficients over +0.0030. For a better
visualization two active compounds (compound 15 of the BMC1998 pool and
compound 11R of the BMCL2006 pool; Supplementary data) are reported. For the
sake of clarity hydrogen atoms are not displayed.
the missing polyhedrons as poorly correlating with the training
set activities.
As regards to the B region (red circle), this is occupied by the
azole moieties and is mainly described by positive PLS coefficients
(yellow polyhedrons), in perfect agreement with the recognized
importance of the coordination capability of such groups with
the heme iron of the CYP enzyme.
The activity contribution plot, different for every molecule
within the training set, gives the possibility to display spatial re-
gions that are individually important for the selected molecules.
In general, in the activity contribution plots cyan polyhedrons
(negative values) show areas where a negative contribution to
the inhibitory activity is associated, while yellow polyhedrons (po-
sitive values) indicate areas of positive contribution to the aroma-
tase inhibitory activity.
This is more clearly explained in Figure 7, reporting the activity
contribution plots relative to derivatives 2bk, 2bh, and 2bm,
respectively, the most and two of the least active among the com-
pounds of the series 2 (Table 1). The decrease of activity from 2bk
(4-fluorophenyl-substituted) to 2bm (4-methoxy-phenyl-substi-
tuted) could be ascribed to the higher sterical hindrance exerted
by the methoxy group as showed by the cyan polyhedron close
to the methyl group (central panel of Fig. 7). On the other hand,
the lower activity of 2bh if compared to 2bk could be attributed
Figure 5. Experimental versus recalculated/predicted pIC50 plot. The molecules of
the training set are indicated with blue squares while the compounds of the test set
are indicated with orange circles. CGS-18320B, fadrozole, and vorozole were used as
reference compounds and are indicated with deep red triangles.

S. Castellano et al. / Bioorg. Med. Chem. 16 (2008) 8349–8358
8355
Figure 7. Activity contribution plots of the PLS coefficients for derivatives 2bk (left), 2bm (center), and 2bh (right). The conformations of the three compounds are displayed
as atom type colors. The cyan contours represent negative activity contribution areas while the yellow contours represent the positive activity contribution areas. For the sake
of clarity hydrogen atoms are not displayed.
Table 4
training set, we took advantage of the Surflex program and of the
GRID/GOLPE procedure to develop a novel 3D-QSAR model, charac-
Mulliken charges calculated with the program GAMESS
terized by high robustness and self-consistency as substantiated by
the high statistical coefficients. The analysis of the PLS coefficients
plots revealed three regions crucial for the interaction of the inhib-
itors with the CYP enzyme and this fact, together with the capabil-
ity to predict the activity of an external data set, propose this
model as a helpful tool to design new aromatase inhibitors.
Compound
Substituent in position 4
Mulliken charge
1bi
Cl
F
OMe
F
OMe
0.038931
ꢀ0.409229
ꢀ0.715180
ꢀ0.411690
ꢀ0.820991
1bk
1bm
2bk
2bm
to the establishment by the 2-F substituent of unfavorable interac-
tions (cyan polyhedrons in right panel of Fig. 7), that instead are
missing in the case of the 4-F group (left panel of Fig. 7).
4. Experimental
4.1. Chemistry
The difference in activity induced by the substituents (fluorine,
chlorine, or methoxy) in the para-position could be explained by a
combination of both sterical and electronic factors. In fact, ab initio
calculations performed for the derivatives 1bj, 1bk, 1bm, 2bk, and
2bh at the MP2/3-21G* level using the quantum chemistry pro-
gram GAMESS61 revealed that the negative charge on the fluorine
atoms (compounds 1bk and 2bk, Table 4) is sufficient to account
for a charge–charge bond or even for accepting a weak hydrogen
bond. On the contrary, this interaction is not allowed for both
the chlorine atom, positively charged and much bigger than fluo-
rine (compare activity values of 1bj and 1bk in Table 1), and the
methoxy substituent, negatively charged but whose free rotation
could hinder the formation of the favorable interactions (compare
activity values of 1bm and 2bm in Table 1 with those of 1bk and
2bk, respectively). Such scenario is fully in agreement with the
activity contribution plots reported in Figure 7.
4.1.1. General
All chemicals were purchased from Aldrich Chimica (Milan, Italy)
or from Lancaster Synthesis GmbH (Milan, Italy) and were of the
highest purity. All solvents were reagent grade and, when necessary,
were purified and dried by standard methods. Reactions were rou-
tinely monitored by TLC performed on aluminum-backed silica gel
plates (Merck DC, Alufolien Kieselgel 60 F254) with spots visualized
by UV light (k = 254 and 365 nm) or using a KMnO4 alkaline solution.
Concentration of solutions after reactions and extractions involved
the use of a rotary evaporator operating at a reduced pressure of
ꢁ10 Torr. Organic solutions were dried over anhydrous sodium sul-
fate. Chromatographic separations were performed on silica gel (Sil-
ica gel 60, 0.063–0.200 mm; Merck DC) or on alumina (aluminium
oxide 90, active, neutral, 0.063–0.200 mm; Merck DC) columns.
Melting points were determined on a Gallenkamp melting point
apparatus in open capillary tubes and are uncorrected. Infrared
(IR) spectra (KBr) were recorded on a Shimadzu FTIR-8000 instru-
ment. 1H NMR spectra were recorded at 300 MHz on a Bruker
Avance 300 spectrometer; chemical shifts are reported in d (ppm)
units relative to the internal reference tetramethylsilane (Me4Si).
Mass spectra were recorded on a Finnigan LCQ DECA TermoQuest
(San Jose, CA) mass spectrometer using an electrospray ion source
(ESI-MS). The elemental compositions of the compounds agreed to
within 0.4% of the calculated value. When the elemental analysis
is not included, crude compounds were used in the next step with-
out further purification. As a rule, samples prepared for physical and
biological studies were dried in high vacuum over P2O5 for 20 h at
temperatures ranging from 25 to 110 °C, depending on the sample
melting point. The Schiff bases 5b, 5e, and 5n,34 the corresponding
secondary amines 6b, 6e, and 6n,34 the imidazolylacetophenones
7b48 and 7e,49 as well as 7-(bromomethyl)benzothiophene47 were
prepared according to procedures described in the literature.
3. Conclusion
With the aim to investigate the potential of their common flex-
ible backbone as a template for developing novel and selective aro-
matase inhibitors, we have tested the inhibitory activity against
CYP19 and CYP17 of a library of antifungal agents previously re-
ported by us. Indeed, even if lacking H-bond accepting substituents
as CN or NO2, several among the tested derivatives exhibited a
good inhibitory potency against aromatase as well as a fairly good
selectivity toward CYP17. However when we decided to build a
3D-QSAR model to rationalize and better understand the struc-
ture–activity relationships of such scaffold we realized the inade-
quacy to this purpose of
a structure-based approach using
previously described models. Therefore we turned to a ligand-
based approach and after pooling the structures of 124 previously
reported aromatase inhibitors into
a structural-diversity-rich

8356
S. Castellano et al. / Bioorg. Med. Chem. 16 (2008) 8349–8358
4.1.2. Synthesis of 3-(5-chlorothiophen-2-yl)-2-(1H-imidazol-1-
yl)-1-phenylpropan-1-one (8eq)
Anal. Calcd for C22H19ClN2SꢃC2H2O4: C, 61.47; H, 4.51; N, 5.97.
Found: C, 61.60; H, 4.52; N, 5.96.
To a stirred suspension of NaH (0.576 g, 24 mmol) in dry THF
(20 ml) a solution of 2-(1-imidazolyl)acetophenone 7e49 (3.72 g,
20 mmol) in dry THF was added. After stirring at room temper-
ature for 1 h, a solution of 2-chloro-5-(chloromethyl)thiophene
(3.34 g, 20 mmol) in dry THF (20 mL) was added dropwise and
the reaction was kept stirring at 50 °C for 3 h. The reaction mix-
ture was then cooled to room temperature, quenched with
methanol (2.0 mL) and concentrated under vacuum. The crude
product was partitioned between diethyl ether (100 mL) and
water (100 mL) and the aqueous layer was separated and ex-
tracted with diethyl ether (3ꢂ 50 mL). The combined organic
layers were washed with brine, then dried over Na2SO4 and con-
centrated. The resulting crude material was purified by column
chromatography on silica gel eluting with AcOEt to yield pure
8eq (2.28 g, 36%) as a colorless glass-like solid which was crys-
tallized as an oxalate salt from ethanol/diethyl ether (mp 176–
178 °C). 1H NMR (CDCl3): d 7.93–7.82 (m, 2H), 7.65–7.40 (m,
4H), 7.04 (s, 1H), 6.99 (s, 1H), 6.64 (d, 1H, J = 3.8 Hz), 6.36 (d,
1H, J = 3.8 Hz), 5.70 (dd, 1H, J = 8.8, 5.1 Hz), 3.55 (dd, 1H,
J = 15.4, 5.1 Hz), 3.35 (dd, 1H, J = 15.4, 8.8 Hz). ESI-MS m/z: 317
(M+H). IR (KBr) 1697 (C@O) cmꢀ1. Anal. Calcd for C16H13ClN2OSꢃ-
C2H2O4: C, 53.14; H, 3.72; N, 6.89. Found: C, 53.25; H, 3.72; N,
6.88.
4.1.6. Synthesis of N-benzyl-N-(5-chlorothiophen-2-yl-methyl)-
1H-imidazol-1-amine (1eq)
A
solution of N-benzyl-1H-imidazol-1-amine 6e34 (1.73 g,
10.0 mmol) and tetrabutylammonium hydrogen sulfate (3.39 g,
10.0 mmol) in DCM (50 mL) was treated with 2-chloro-5-(chloro-
methyl)thiophene (2.51 g, 15.0 mmol) and 50% sodium hydroxide
aqueous solution (30 mL). After stirring at room temperature for
15 h, DCM (60 mL) was added and the mixture was poured into
ice-water (60 mL). The aqueous layer was separated and extracted
with DCM (3ꢂ 50 mL). Combined organic layers were washed with
brine, dried over Na2SO4 and concentrated. The resulting crude
material was purified by column chromatography on silica gel
eluting with AcOEt to yield pure 1eq (1.00 g, 33%) as a colorless
glass-like solid which was crystallized as an oxalate salt from eth-
anol/diethyl ether (mp 147 °C). 1H NMR (CDCl3): d 7.30–7.17 (m,
6H), 7.10–7.03 (m, 1H), 6.95–6.88 (m, 1H), 6.65 (d, 1H,
J = 1.8 Hz), 6.53 (d, 1H, J = 1.8 Hz), 4.24 (s, 2H), 4.15 (s, 2H). ESI-
MS m/z: 304 (M+H). Anal. Calcd for
C
15H14ClN3S.C2H2O4: C,
51.84; H, 4.09; N, 10.67. Found: C, 51.96; H, 4.10; N, 10.64.
4.1.7. Synthesis of N-benzyl-N-(benzo[b]thiophen-7-ylmethyl)-
1H-imidazol-1-amine (1er)
A
mixture of N-benzyl-1H-imidazol-1-amine 6e34 (1.73 g,
4.1.3. Synthesis of 1-[1,10-biphenyl]-4-yl-3-(5-chlorothio-phen-
2-yl)-2-(1H-imidazol-1-yl)propan-1-one (8bq)
10.0 mmol), potassium tert-butoxide (1.12 g, 10.0 mmol) and 18-
crown-6 (0.27 g, 1.00 mmol) in diethylether (20 mL) was stirred
at room temperature under nitrogen for 15 min. The reaction mix-
ture was cooled to 0 °C and a solution of 7-(bromomethyl)benzo-
thiophene47 (1.99 g, 10.0 mmol) in diethylether (10 mL) was
added dropwise. The mixture was stirred at room temperature for
4 h and then poured into water (60 mL). The aqueous layer was sep-
arated and extracted with AcOEt (3ꢂ 50 mL). The combined organic
phases were washed with brine, dried over Na2SO4 and concen-
trated. The resulting crude material was purified by column chro-
matography on silica gel eluting with AcOEt to yield pure 1er
(1.15 g, 36%) as a colorless glass-like solid which was crystallized
from benzene/ligroin (mp 110–112 °C). 1H NMR (CDCl3): d 7.78–
7.69 (m, 1H), 7.52–7.08 (m, 11H), 6.91–6.86 (m, 1H), 4.45 (s, 2H),
4.17 (s, 2H). ESI-MS m/z: 320 (M+H). Anal. Calcd for C19H17N3S: C,
71.44; H, 5.36; N, 13.15. Found: C, 71.60; H, 5.37; N, 13.13.
The alkylation of the imidazolylacetophenone 7b48 according to
the same procedure described above yielded pure 8bq (34%) as a
white solid (mp 142–143 °C). 1H NMR (CDCl3): d 7.96 (d, 2H,
J = 7.9 Hz), 7.70–7.30 (m, 8H), 7.08 (s, 1H), 7.03 (s, 1H), 6.68 (d,
1H, J = 3.9 Hz), 6.40 (d, 1H, J = 3.9 Hz), 5.73 (dd, 1H, J = 8.9,
5.3 Hz), 3.55 (dd, 1H, J = 15.2, 5.3 Hz), 3.35 (dd, 1H, J = 15.2,
8.9 Hz). ESI-MS m/z: 393 (M+H). IR (KBr) 1681 (C@O) cmꢀ1. Anal.
Calcd for C22H17ClN2OS: C, 67.25; H, 4.36; N, 7.13. Found: C,
67.38; H, 4.35; N, 7.12.
4.1.4. Synthesis of 1-(5-chlorothiophen-2-yl)-2-(1H-imidazol-
1-yl)-3-phenylpropane (2eq)
A mixture of 3-(5-chlorothiophen-2-yl)-2-(1H-imidazol-1-yl)-
1-phenylpropan-1-one 8eq (1.78 g, 5.64 mmol), hydrazine mono-
hydrate (1.37 mL, 28.2 mmol), ethylene glycol (50 mL) and potas-
sium hydroxide (2.59 g, 46.2 mmol) was heated at 185–190 °C
for 3 h. The reaction mixture was then cooled to room tempera-
ture, poured into 500 mL of water, acidified to pH ꢄ 1.0 with con-
centrated hydrochloric acid and extracted with chloroform (3ꢂ
50 mL). The organic extract was washed with brine, dried over
Na2SO4 and concentrated. The crude product was purified by col-
umn chromatography on silica gel eluting with AcOEt to yield pure
2eq (0.46 g, 27%) as a colorless glass-like solid which was crystal-
lized as an oxalate salt from ethanol/diethyl ether (mp 182–
184 °C). 1H NMR (CDCl3): d 7.36–6.80 (m, 8H), 6.63 (d, 1H,
J = 3.1 Hz), 6.34 (d, 1H, J = 3.1 Hz), 4.40–4.19 (m, 1H), 3.34–2.95
(m, 4H). ESI-MS m/z: 303 (M+H). Anal. Calcd for
4.1.8. Synthesis of N-(1,10-biphenyl-4-ylmethyl)-N-(5-
chlorothiophen-2-ylmethyl)-1H-imidazol-1-amine (1bq)
The alkylation of N-[(1,10-biphenyl)-4-ylmethyl]-1H-imidazol-
1-amine 6b34 with 2-chloro-5-(chloromethyl)thio-phene following
the same procedure described for 1eq, yielded pure 1bq (81%) as a
colorless glass-like solid which was crystallized as a hydrochloride
salt from ethanol/diethyl ether (mp 169–171 °C). 1H NMR (CDCl3):
d
7.60–7.25 (m, 10H), 7.13 (s, 1H), 6.98 (s, 1H), 6.69 (d, 1H,
J = 3.7 Hz), 6.57 (d, 1H, J = 3.7 Hz), 4.29 (s, 2H), 4.27 (s, 2H). ESI-
MS m/z: 380 (M+H). Anal. Calcd for C19H18ClN3SꢃHCl: C, 60.58; H,
4.60; N, 10.09. Found: C, 60.69; H, 4.60; N, 10.07.
C
16H15ClN2SꢃC2H2O4: C, 55.03; H, 4.36; N, 7.13. Found: C, 55.15;
4.1.9. Synthesis of N-(5-chlorothiophen-2-ylmethyl)-N-(2,4-
dichlorobenzyl)-1H-imidazol-1-amine (1nq)
H, 4.36; N, 7.11.
The alkylation of N-(2,4-dichlorobenzyl)-1H-imidazol-1-amine
6n34 with 2-chloro-5-(chloromethyl)thiophene following the same
procedure described for 1eq, yielded pure 1nq (32%) as a white
glass-like solid which was crystallized from benzene/ligroin (mp
106 °C). 1H NMR (CDCl3): d 7.38–7.33 (m, 1H), 7.25–7.22 (m, 1H),
7.14–7.08 (m, 3H), 6.98–6.91 (m, 1H), 6.65 (d, 1H, J = 3.8 Hz),
6.55 (d, 1H, J = 3.8 Hz),4.29 (s, 2H), 4.27 (s, 2H). ESI-MS m/z: 374
(M+H). Anal. Calcd for C15H12Cl3N3S: C, 48.34; H, 3.25; N, 11.27.
Found: C, 48.44; H, 3.26; N, 11.24.
4.1.5. Synthesis of 1-(1,10-biphenyl-4-yl)-3-(5-chloro-thiophen-
2-yl)-2-(1H-imidazol-1-yl)propane (2bq)
Following the same procedure described above, title derivative
2bq was obtained starting from 8bq as a pure colorless glass-like
solid (yield: 34%) which was crystallized as an oxalate salt from
ethanol/diethyl ether (mp 143–144 °C). 1H NMR (CDCl3): d 7.40–
6.78 (m, 12H), 6.65 (d, 1H, J = 3.2 Hz), 6.40 (d, 1H, J = 3.2 Hz),
4.42–4.18 (m, 1H), 3.36–2.98 (m, 4H). ESI-MS m/z: 379 (M+H).

S. Castellano et al. / Bioorg. Med. Chem. 16 (2008) 8349–8358
8357
4.1.10. Synthesis of N-(benzo[b]thiophen-7-ylmethyl)-N-(2,4-
dichlorobenzyl)-1H-imidazol-1-amine (1nr)
Italy) and for having provided the GRID program. This work was
partially supported by grants from Regione Campania 2003, LR 5/
02, Ministero dell’Università e della Ricerca Scientifica e Tecnolog-
ica (PRIN 2005), and Università di Salerno, Italy (G.S.). Thanks are
due to Martina Jankowski for the help performing the enzyme
assays.
The alkylation of N-(2,4-dichlorobenzyl)-1H-imidazol-1-amine
6n34 with 7-(bromomethyl)benzothiophene47 following the pro-
cedure described for 1eq, yielded pure 1nr (26%) as a white
glass-like solid which was crystallized as an oxalate salt from
ethanol/diethyl ether (mp 145–147 °C). 1H NMR (CDCl3):
d
7.77–7.68 (m, 1H), 7.51–7.08 (m, 9H), 6.93–6.84 (m, 1H), 4.49
(s, 2H), 4.31 (s, 2H). ESI-MS m/z: 388 (M+H). Anal. Calcd for
Supplementary data
C
19H15Cl2N3SꢃC2H2O4: C, 52.73; H, 3.58; N, 8.78. Found: C,
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.bmc.2008.08.046.
52.84; H, 3.59; N, 8.76.
4.2. Biology
References and notes
1. American Cancer Society web site. http://www.cancer.org and http://
www.imaginis.com.
2. Chen, S.; Zhou, D.; Okubo, T.; Kao, Y. C.; Eng, E. T.; Grube, B.; Kwon, A.; Yang, C.;
Yu, B. Ann. N. Y. Acad. Sci. 2002, 963, 229.
3. Geisler, H. J. Steroid Biochem. Mol. Biol. 2003, 86, 245.
4. Bruno, R. D.; Njar, V. C. O. Bioorg. Med. Chem. 2007, 15, 5047.
5. Brueggemeier, R. W.; Hackett, J. C.; Diaz-Cruz, E. S. Endocr. Rev. 2005, 26, 331.
6. Brodie, A.; Brodie, H. B.; Callard, G.; Robinson, C.; Roselli, C.; Santen, R. J. Steroid
Biochem. Mol. Biol. 1993, 44, 321.
4.2.1. CYP19 preparation and assay
The microsomal fraction of freshly delivered human term
placenta provided the source of the aromatase enzyme. The
compounds were tested for aromatase inhibitory activity62 using
the 3H2O-method introduced by Thompson and Siiteri.44 [1b,
2b-3H]testosterone was used as substrate.
7. Foster, A. B.; Jarman, M.; Leung, C. S.; Rowlands, M. G.; Taylor, G. N.; Plevey, R.
G.; Sampson, P. J. Med. Chem. 1985, 28, 200.
8. Daly, M. J.; Jones, G. W.; Nicholls, P. J.; Smith, H. J.; Rowlands, M. G.; Bunnett, M.
A. J. Med. Chem. 1986, 29, 520.
9. Hartmann, R. W.; Batzl, C.; Pongratz, T. M.; Mannschreck, A. J. Med. Chem. 1992,
35, 2210.
10. Bossche, H. V. J. Steroid Biochem. Mol. Biol. 1992, 43, 1003.
11. Karjalainen, A.; Kalapudas, A.; Södervall, M.; Pelkonen, O.; Lammintausta, R.
Eur. J. Pharm. Sci. 2000, 11, 109.
4.2.2. CYP17 preparation and assay
Human CYP17 was obtained from E. coli coexpressing human
CYP17 and NADPH-P450 reductase63 by homogenization and frac-
tional centrifugation.45 The assay was performed using the 50,000g
sediment and progesterone as substrate as described.45
4.3. Molecular modeling
12. Attar, E.; Bulun, S. E. Fertil. Steril. 2006, 85, 1307.
13. Geisler, J.; Lonning, E. Clin. Cancer Res. 2005, 11, 2809.
14. Michaud, L. B. Am. J. Health-Syst. Pharm. 2005, 62, 266.
15. Miller, W. R. Semin. Oncol. 2003, 30, 3.
16. Recanatini, M.; Cavalli, A.; Valenti, P. Med. Res. Rev. 2002, 22, 282. and
references therein reported.
17. Kelloff, G. J.; Lubet, R. A.; Lieberman, R.; Eisenhauer, K.; Steele, V. E.; Crowell, J.
A.; Hawk, E. T.; Boone, C. W.; Sigman, C. C. Cancer Epidemiol. Biomarkers Prev.
1998, 7, 65.
18. Saberi, M. R.; Vinh, T. K.; Yee, S. W.; Griffiths, B. J. N.; Evans, P. J.; Simons, C. J.
Med. Chem. 2006, 49, 1016.
All the tested molecules were built, starting from ASCII text,
using the standalone version of PRODRG51–53 in conjunction with
GROMACS suite.64 The alignment of the compounds was achieved
by means of the Surflex program,55 using the crystal structures of
CGS-18320B,56 fadrozole,56 and vorozole57 as reference molecules,
previously rigidly aligned with the program Surflex. The Chimera
1.2176 program65 was used to produce the images on a 3GHz
AMD CPU equipped IBM compatible workstation with the SUSE
9.0 version of the Linux operating system.
The training set of 124 molecules was obtained by pooling 31
compounds belonging to the 1–2 and 4 series,50 the three reference
compounds (vorozole, anastrozole, and fadrozole), and 90 previ-
ously reported aromatase inhibitors (Fig. 4).20–22,40–43,51
19. Su, B.; Diaz-Cruz, E. S.; Landini, S.; Brueggemeier, R. W. J. Med. Chem. 2006, 49,
1413.
20. Lézé, M. P.; Le Borgne, M.; Pinson, P.; Palusczak, A.; Duflos, M.; Le Baut, G.;
Hartmann, R. W. Bioorg. Med. Chem. Lett. 2006, 16, 1134.
21. Recanatini, M.; Bisi, A.; Cavalli, A.; Belluti, F.; Gobbi, S.; Rampa, A.; Valenti, P.;
Palzer, M.; Palusczak, A.; Hartmann, R. W. J. Med. Chem. 2001, 44, 672.
22. Cavalli, A.; Bisi, A.; Bertucci, C.; Rosini, C.; Paluszcak, A.; Gobbi, S.; Giorgio, E.;
Rampa, A.; Belluti, F.; Piazzi, L.; Valenti, P.; Hartmann, R. W.; Recanatini, M. J.
Med. Chem. 2005, 48, 7282.
23. Gobbi, S.; Cavalli, A.; Rampa, A.; Belluti, F.; Piazzi, L.; Palusczak, A.; Hartmann,
R. W.; Recanatini, M.; Bisi, A. J. Med. Chem. 2006, 49, 4777.
24. Gobbi, S.; Cavalli, A.; Negri, M.; Schewe, K. E.; Belluti, F.; Piazzi, L.; Hartmann, R.
W.; Recanatini, M.; Bisi, A. J. Med. Chem. 2007, 50, 3420.
25. Okada, M.; Yoden, T.; Kawaminami, E.; Shimada, Y.; Kudoh, M.; Isomura, Y.;
Shikama, H.; Fujikura, T. Chem. Pharm. Bull. 1996, 44, 1871.
26. Howell, A.; Robertson, J. F. R.; Vergote, I. A. Breast Cancer Res. 2003, 82, 215.
27. Scott, J. L.; Keam, J. S. Drugs 2006, 66, 353.
4.3.1. GRID calculations
The interaction energies were calculated by using the program
GRID66 (version 22) with a grid spacing of 1 Å and the grid dimen-
sions (Å): Xmin/Xmax, ꢀ8.0/20.0; Ymin/Ymax, ꢀ13.0/13.0; and Zmin
/
Zmax, ꢀ13.0/13.0.
28. Lonning, P. E.; Bajetta, E.; Murray, R.; Tubiana-Hulin, M.; Eisenberg, P. D.;
Mickiewicz, E.; Celio, L.; Pitt, P.; Mita, M.; Aaronson, N. K.; Fowst, C.; Arkhipov,
A.; di Salle, E.; Polli, A.; Massimini, G. J. Clin. Oncol. 2000, 18, 2234.
29. Thurlimann, B.; Keshaviah, A.; Coates, A. S.; Mouridsen, H.; Mauriac, L.; Forbes,
J. F.; Paridaens, M.; Castiglione-Gertsch, M.; Gelber, R. D.; Rabaglio, M.; Smith,
I.; Wardley, A.; Price, K. N.; Goldhirsch, A. N. Engl. J. Med. 2005, 353, 2747.
30. Jakesz, R.; Jonet, W.; Gnant, M.; Mittleboeck, M.; Greil, R.; Tausch, C.; Hilfrich,
J.; Kwasny, W.; Menzel, C.; Samonigg, H. Lancet 2005, 366, 455.
31. Osborne, C.; Tripathy, D. Annu. Rev. Med. 2005, 56, 103.
32. Jackson, J.; Miller, W. R.; Dixon, J. M. Expert Opin. Drug Saf. 2003, 2, 73.
33. Castellano, S.; La Colla, P.; Musiu, C.; Stefancich, G. Arch. Pharm. (Weinheim,
Ger.) 2000, 333, 162.
34. Castellano, S.; Stefancich, G.; Musiu, C.; La Colla, P. Arch. Pharm. (Weinheim,
Ger.) 2000, 333, 299.
35. Setzu, M. G.; Stefancich, G.; La Colla, P.; Castellano, S. Farmaco 2002, 57, 1015.
36. Castellano, S.; Stefancich, G.; Chillotti, A.; Poni, G. Farmaco 2003, 58, 563.
37. Furet, P.; Batzl, C.; Bhatnagar, A.; Francotte, E.; Rihs, G.; Lang, M. J. Med. Chem.
1993, 36, 1393.
4.3.2. GOLPE analyses
PLS models were calculated with GOLPE 4.5.1267 running on a
SGI O2 R10000 equipped with the IRIX operating system 6.5.11.
To measure the goodness of the model the statistical indices r2,
q2 and SDEP were employed.
4.3.3. Ab initio calculations
Partial atomic charges on the substituents in the para-position
of compounds 1bj, 1bk, 1bm, 2bk, and 2bh of the training set were
calculated by using the GAMESS package61 and the MP2/3-21G* ba-
sis set.
Acknowledgments
38. Koymans, L. M. H.; Moereels, H.; Bossche, H. V. J. Steroid Biochem. Mol. Biol.
1995, 53, 191.
39. Jackson, T.; Lawrence Woo, L. W.; Trusselle, M. N.; Purohit, A.; Reed, M. J.;
Potter, B. V. L. ChemMedChem 2008, 3, 603.
Many thanks are due to Prof. Gabriele Cruciani and Prof. Sergio
Clementi (Molecular Discovery and MIA srl) for the use the GOLPE
program in their chemometric laboratory (University of Perugia,

8358
S. Castellano et al. / Bioorg. Med. Chem. 16 (2008) 8349–8358
40. Le Borgne, M.; Marchand, P.; Duflos, M.; Delevoye-Seiller, B.; Piessard-Robert,
S.; Le Baut, G.; Hartmann, R. W.; Palzer, M. Arch. Pharm. (Weinheim, Ger.) 1997,
330, 141.
41. Le Borgne, M.; Marchand, P.; Delevoye-Seiller, B.; Robert, J. M.; Le
Baut, G.; Hartmann, R. W.; Palzer, M. Bioorg. Med. Chem. Lett. 1999, 9,
333.
53. Schuttelkopf, A. W.; van Aalten, D. M. Acta Crystallogr., Sect. D Biol. Crystallogr.
2004, 60, 1355.
54. van Aalten, D. M.; Bywater, R.; Findlay, J. B.; Hendlich, M.; Hooft, R. W.; Vriend,
G. J. Comput. Aided Mol. Des. 1996, 10, 255.
55. Favia, A. D.; Cavalli, A.; Masetti, M.; Carotti, A.; Recanatini, M. Proteins Struct.
Funct., Bioinf. 2006, 62, 1074.
42. Marchand, P.; Le Borgne, M.; Palzer, M.; Le Baut, G.; Hartmann, R. W. Bioorg.
Med. Chem. Lett. 2003, 13, 1553.
43. Hartmann, R. W.; Palusczak, A.; Lacan, F.; Ricci, G.; Ruzziconi, R. J. Enzyme Inhib.
Med. Chem. 2004, 19, 145.
44. Thompson, E. A., Jr.; Siiteri, P. K. J. Biol. Chem. 1974, 249, 5364.
45. Hutschenreuter, T. U.; Ehmer, P. B.; Hartmann, R. W. J. Enzyme Inhib. Med. Chem.
2004, 19, 17.
56. Jain, A. N. J. Med. Chem. 2003, 46, 499.
57. Van Roey, P.; Bullion, K. A.; Osawa, Y.; Browne, L. J.; Bowman, R. M.; Braun, D.
G. J. Enzyme Inhib. Med. Chem. 1991, 5, 119.
58. Peeters, O. M.; Schuerman, G. S.; Blaton, N. M.; De Ranter, C. J. Acta Crystallogr.,
Sect. C Cryst. Struct. Commun. 1958, 1993, 49.
59. Ragno, R.; Simeoni, S.; Valente, S.; Massa, S.; Mai, A. J. Chem. Inf. Model. 2006,
46, 1420.
46. Hartmann, R. W.; Bayer, H.; Gruen, G.; Sergejew, T.; Bartz, U.; Mitrenga, M. J.
Med. Chem. 1995, 38, 2103.
60. Baroni, M.; Costantino, G.; Cruciani, G.; Riganelli, D.; Valigi, R.; Clementi, S.
Quant. Struct.-Act. Relat. 1993, 12, 9.
47. Nussbaumer, P.; Petranyi, G.; Stuetz, A. J. Med. Chem. 1991, 34, 65.
48. Nardi, D.; Tajana, A.; Leonardi, A.; Pennini, R.; Portioli, F.; Magistretti, M. J.;
Subissi, A. J. Med. Chem. 1981, 24, 727.
49. Baji, H.; Flammang, M.; Kimny, T.; Gasquez, F.; Compagnon, P. L.; Delcourt, A.
Eur. J. Med. Chem. 1995, 30, 617.
50. The structures of derivatives of the series 3 and 4 (with the exception of
compounds 4ac, 4cn, 4dc, and 4dp), for which just the percentages of
inhibition and not IC50 values were reported in Table 1, were not included in
the training set.
51. Recanatini, M.; Cavalli, A. Bioorg. Med. Chem. 1998, 6, 377.
52. For academic research purposes PRODRG2 is freely available as a WWW
service at http://davapc1.bioch.dundee.ac.uk/prodrg/index.html.
61. Schmidt, M. W.; Baldridge, K. K.; Boatz, J. A.; Elbert, S. T.; Gordon, M. S.; Jensen,
J. H.; Koseki, S.; Matsunaga, N.; Nguyen, K. A.; Su, S.; Windus, T. L.; Dupuis, M.;
John, A.; Montgomery, J. Comput. Chem. 1993, 14, 1347.
62. Hartmann, R. W.; Batzl, C. J. Med. Chem. 1986, 29, 1362.
63. Ehmer, P. B.; Jose, J.; Hartmann, R. W. J. Steroid Biochem. Mol. Biol. 2000, 75, 57.
64. Van Der Spoel, D.; Lindahl, E.; Hess, B.; Groenhof, G.; Mark, A. E.; Berendsen, H.
J. C. J. Comput. Chem. 2005, 26, 1701.
65. Pettersen, E. F.; Goddard, T. D.; Huang, C. C.; Couch, G. S.; Greenblatt, D. M.;
Meng, E. C.; Ferrin, T. E. J. Comput. Chem. 2004, 25, 1605.
66. Goodford, P. J. J. Med. Chem. 1985, 28, 849.
67. GOLPE 4.5; Multivariate Infometric Analysis Srl.: Viale dei Castagni 16, Perugia,
Italy, 1999.