Synthesis and structure-activity relationships of (aryloxy)quinazoline ureas as novel, potent, and selective vascular endothelial growth factor receptor-2 inhibitors

Article abstract of DOI:10.1021/jm2013453

In our continuing search for medicinal agents to treat proliferative diseases, quinazoline derivatives were synthesized and evaluated pharmacologically as epithelial growth factor receptor and vascular endothelial growth factor receptor 2 (VEGFR-2) tyrosine kinase inhibitors. A quantitative structure-activity relationship analysis was conducted to rationalize the structure-activity relationship and to predict how similar the inhibitor-binding profiles of two protein kinases are likely to be on the basis of the docking of lead coumpounds into the ATP-binding site. This model was used to direct the synthesis of new compounds. A series of N-(aromatic)-N′-{4-[(6,7- dimethoxyquinazolin-4-yl)oxy]phenyl}urea were identified as potent and selective inhibitors of the tyrosine kinase activity of VEGFR-2 (fetal liver kinase 1, kinase insert domain-containing receptor). An efficient route was developed that enabled the synthesis of a wide variety of analogues with substitution on several positions of the template. Substitution of diarylurea, competitive with ATP, afforded several analogues with low nanomolar inhibition of enzymatic activity of VEGFR-2. In this paper, we describe the synthesis, structure-activity relationships, and pharmacological characterization of the series.

Full text of DOI:10.1021/jm2013453

Article
pubs.acs.org/jmc
Synthesis and Structure−Activity Relationships of
(Aryloxy)quinazoline Ureas as Novel, Potent, and Selective Vascular
Endothelial Growth Factor Receptor-2 Inhibitors
Antonio Garofalo,†,‡ Amaury Farce,*,†,§ Severine Ravez,†,‡ Amelie Lemoine,†,‡ Perrine Six,†,‡
́
́
Philippe Chavatte,†,§ Laurence Goossens,†,‡ and Patrick Depreux†,‡
†Univ. Lille Nord de France, F-59000 Lille, France
§
‡Institut de Chimie Pharmaceutique Albert Lespagnol (ICPAL) and Laboratoire de Chimie Ther
́
apeutique, UFR de Pharmacie,
́
Univ. Lille 2 Droit et Sante (UDSL), EA4481, F-59006 Lille, France
ABSTRACT: In our continuing search for medicinal agents to treat proliferative diseases,
quinazoline derivatives were synthesized and evaluated pharmacologically as epithelial
growth factor receptor and vascular endothelial growth factor receptor 2 (VEGFR-2)
tyrosine kinase inhibitors. A quantitative structure−activity relationship analysis was
conducted to rationalize the structure−activity relationship and to predict how similar the
inhibitor-binding profiles of two protein kinases are likely to be on the basis of the docking
of lead coumpounds into the ATP-binding site. This model was used to direct the synthesis
of new compounds. A series of N-(aromatic)-N′-{4-[(6,7-dimethoxyquinazolin-4-yl)oxy]-
phenyl}urea were identified as potent and selective inhibitors of the tyrosine kinase activity
of VEGFR-2 (fetal liver kinase 1, kinase insert domain-containing receptor). An efficient
route was developed that enabled the synthesis of a wide variety of analogues with
substitution on several positions of the template. Substitution of diarylurea, competitive
with ATP, afforded several analogues with low nanomolar inhibition of enzymatic activity of VEGFR-2. In this paper, we describe
the synthesis, structure−activity relationships, and pharmacological characterization of the series.
This receptor tyrosine kinase (RTK) has been shown to be an
important mediator of signal transduction in cells.17−21 These
membrane molecules characteristically consist of an extrac-
ellular ligand-binding domain connected through a segment in
the plasma membrane to an intracellular tyrosine kinase
domain.22,23 Binding of the ligand to the receptor results in
receptor dimerization and stimulation of the receptor-
associated tyrosine kinase activity, which leads to phosphor-
ylation of tyrosine residues and initiates a signaling cascade
leading to a variety of cellular responses.24−26
Throughout the past two decades, inhibition of VEGF
activity or VEGFR-2 kinase has been shown to inhibit angio-
genesis, tumor progression, and dissemination in a number of
preclinical and clinical studies.27−34 From examples, the use of the
neutralizing monoclonal antibody against VEGF, bevacizumab,32,33
has demonstrated a prolonged survival in colorectal cancer
patients. In addition, small-molecule tyrosine kinase inhibitors
of KDR, such as sorafenib (a dual Raf−KDR inhibitor)34,35 and
sunitinib (a multitarged kinase),36 have been approved for
treatment of cancers (Figure 1). Up to now, there has been a
lot of research focusing on the development of novel inhibitors
of KDR as vandetanib (ZD6474), orally bioavailable in phase
III, considered to be a dual tyrosine kinase inhibitor targeting
EGFR and VEGFR-2.37−39
INTRODUCTION
■
Protein kinases constitute one of the largest protein families
and are implicated in many diseases by alteration in their
kinase-mediated signaling pathways, such as cancer, diabetes,
and inflammatory disorders.1−3 Protein kinases play important
roles in regulating most of the cellular functions (proliferation,
cell cycle, cell metabolism, survival, apoptosis, DNA damage/
repair, ...) by multiple pathways, which makes them attractive as
drug targets.4−7 In recent years, angiogenesis has become an
attractive therapeutic target.8 Pathological angiogenesis has
been associated with a variety of diseases, including diabetes
retinopathy, psoriasis, rheumatoid arthritis, and cancer.9 This
pathology is an essential series of molecular events that occur
during the formation of new blood vessels from the
endothelium of preexisting vasculature and plays a vital role
in tumor growth. Solid tumors cannot grow beyond a critical
size until they develop new collateral blood vessels to provide
oxygen and nutrients, elimination of waste materials generated
from tumor metabolism, and cell dissemination leading to
metastasis to other organs.10−12
Among the many proangiogenic factors, vascular endothelial
growth factor (VEGF) has been identified as the most common
regulator of tumor angiogenesis, including vascular permeability,
endothelial cell activation, proliferation, and migration.13−16
Crucial steps in angiogenesis are mediated through a specific
VEGF receptor, the kinase insert domain-containing receptor
(KDR; or vascular endothelial growth factor 2 (VEGFR-2)).
Received: October 7, 2011
Published: January 9, 2012
© 2012 American Chemical Society
1189
dx.doi.org/10.1021/jm2013453 | J. Med. Chem. 2012, 55, 1189−1204

Journal of Medicinal Chemistry
Article
Figure 1. Examples of tyrosine kinase inhibitors undergoing clinical trials or marketed.
Often, the fact that the biological targets belong to the same
family of receptors raises the problem of selectivity. Special
efforts to design multiple activating drugs can be successful
using computational methods to support biochemical studies
and to design selective ligands for receptor subtypes.
In the present work, molecular modeling studies of these
novel VEGFR-2 inhibitors were performed using a three-
dimensional quantitative structure−activity relationship (3D
QSAR) and docking approach. 3D QSAR methods, such as
comparative molecular field analyses (CoMFA) and compara-
tive molecular similarity index analyses (CoMSIA), were
applied to these inhibitors to gain insight into how steric,
electrostatic, hydrophobic, and hydrogen-bonding interactions
influence their activities. In our research for selective KDR
kinase inhibitors, we have recently disclosed the structure−
activity relationships (SARs) of quinazoline derivatives (Figure 2)
conformation. Second, 3D QSAR models employing the CoMFA
and CoMSIA methods have been elaborated on the basis of an
alignment of the compounds on the putative bioactive con-
formation of the two reference molecules. The excellent agree-
ment between the docking and the QSAR fields indicates that
the binding mode hypothesis is most probably fairly close to
the biological binding mode, therefore offering a relevant basis
for proposing improvements for the further development of the
series. The second major interest of this study is the predictive
power of the models, more than their insights into the struc-
tural modifications that would enhance the activity of the series,
which is too homogeneous to provide an efficient basis for
suggesting new scaffolds. This predictive power has been validated
on external test sets and further proved on newly designed
selective VEGFR-2 inhibitors. These compounds have been
synthesized and evaluated in our whole cell assay by measuring
the inhibition of VEGFR-2 and EGFR to provide their selec-
tivity and antiproliferative activity toward prostate (PC3), breast
(MCF-7), and colon (HT-29) cancer cells.
CHEMISTRY
■
In our research to discover a new anticancer drug, we designed
and evaluated pharmacologically 200 compounds as potent
kinase inhibitors. Sixty of these compounds were used for this
study. They were prepared via a generalized route outlined in
Schemes 1−3.
Scheme 1. Synthesis of Quinazoline Intermediates
Figure 2. General structures of synthesized carbamic acid esters 1a and
aryloxy derivatives 1b.40−42
as anticancer drugs.41,42 Herein, we describe the SAR and phar-
macology of a novel series of quinazoline carbamic acid esters,
dual EGFR/VEGFR-2 kinase inhibitors,41 and of (aryloxy)-
quinazoline ureas, selective VEGFR-2 kinase inhibitors.42
In an effort to further develop this series, a QSAR model has
been derived from this activity in a two-step process. First, the
docking of the most active compound of the two families of
molecules has been realized to achieve a putative bioactive
1190
dx.doi.org/10.1021/jm2013453 | J. Med. Chem. 2012, 55, 1189−1204

Journal of Medicinal Chemistry
Article
a
Scheme 2
a
Reagents and conditions: (a) Raney nickel, H2, CH2Cl2, ROH (80−90%); (b) Raney nickel, H2, CH2Cl2/THF (1/1), ROH (80−90%); (c) isocyanate,
CHCl3 (70−80%); (d) RCOCl, triethylamine, THF, 0−20 °C (70−85%); (e) Raney nickel, H2, MeOH (80−90%); (i) anilines, 2-propanol, reflux
(60−80%); (j) anilines, NaH, DMF, 50 °C (25−40%).
The synthesis and spectroscopic data of 4-chloroquinazoline
derivatives 7−10 have been reported (Scheme 1).40−42
The synthesis of (aminophenyl)carbamic acid esters 11−27
was performed by a one-pot reduction procedure of the corre-
sponding nitrophenyl isocyanate in the presence of various
alcohols in a CH2Cl2/THF mixture.46 Condensation of 1,4-
phenylenediamine and the corresponding isocyanate in CHCl3
afforded the desired ureas 28−32 in high yields and short reaction
times.47 Amide analogues 33 and 34 were prepared from 2-
methyl-4-nitroaniline according to a two-step procedure: acetyla-
tion from chloride derivatives and reduction under a hydrogen
atmosphere using Raney nickel as the catalyst.41 Target
compounds 35−78 were obtained by nucleophilic substitution
with previously synthesized anilines 11−34 either in 2-propanol
or in DMF in the presence of sodium hydride (40−60%).
The desired anilinoquinazolines 35−78 were isolated as
hydrochloride salts (Scheme 2).
acetamide, a carbamic acid ester, or a urea substituted on an
arylamino ring, have been designed using structure−activity
relationships (Table 1). Suppression of the basic side chain on
the quinazoline core confers an increase of cellular and enzy-
matic inhibitory activity. Our results show that some derivates
of series A and B (6,7-dimethoxyquinazoline) have high activity
on EGFR and VEGFR-2 compared to series C and D. Introduc-
tion of a urea group led to an increase in VEGFR-2 enzymatic
activity but not in EGFR enzymatic activity. Also, replacement
of the urea entity by a carbamic acid methyl ester group as in 42
presents a dual EGFR/VEGFR-2 activity. According to investi-
gations, the substitution by halogens on the middle phenyl
group of carbamic acid ester and urea derivatives (series A and B)
appeared to be a new opportunity to develop a new dual EGFR/
VEGFR-2 inhibitor.
Investigation of the SAR around the lead carbamic acid
methyl ester 42 revealed a range of potent dual EGFR/VEGFR-
2 inhibitors. The substitution by a halogen or methyl on the
middle phenyl group of carbamic acid ester increased enzymatic
activity. The effect of varying the methoxy groups on the
quinazoline core by basic side chain was investigated (series C,
D, and E). Our results led to a better affinity against VEGFR-2
and a loss of activity against EGFR.
In Scheme 3, 4-chloro-6,7-dimethoxyquinazoline (7) was used
to prepare the aryloxy derivatives 79−93.48 Heating at 150 °C in
DMSO of quinazoline 7 with phenol derivatives gave products
79−81 in good yields. Amidification or condensation with aniline
intermediates 94 and 95 afforded carbamic acid esters 82 and 83
and the desired ureas 84−93.42−48
Four series of new quinazolines (series A, B, C, and D),
differentiated by having an ether linker at the 6- or 7-position of
core and incorporating a donor/acceptor group such as an
We have also described the discovery, SAR, and preliminary
biological evaluation of a novel series of VEGFR-2-selective
tyrosine kinase inhibitors characterized by the replacement of
1191
dx.doi.org/10.1021/jm2013453 | J. Med. Chem. 2012, 55, 1189−1204

Journal of Medicinal Chemistry
Article
Scheme 3. Synthesis of (Aryloxy)quinazoline Derivatives
linkers at the 4-position of the quinazoline skeleton by ether
(series F and G). The diarylurea of (aryloxy)quinazoline
potently inhibited VEGFR-2 at submicromolar concentrations.
Molecular modeling established interactions of these com-
pounds with the binding active site. However, their potent
activity against VEGFR-2 is not reflected by their antiprolifer-
ative activity on cell-based assays due to poor solubility.
According to these results, the substitution by halogens or a
methyl group on the middle phenyl group of urea derivatives
appears to be a new opportunity to develop a new VEGFR-2-
selective inhibitor. Moreover, as a diarylurea moiety is favored
for optimal interaction in the active site, others compounds
were realized to obtain a better affinity.
These compounds were synthesized according to described
procedures in Scheme 4 with 4-chloro-6,7-dimethoxyquinazo-
line (7) as the starting product. Selective reaction of chloride
derivative 7 with 4-amino-3-methylphenol and tetra-n-butyl-
ammonium bromide in a mixture of methyl ethyl ketone and
sodium hydroxide provided the intermediate 96. Condensation
of 94 or 96 with the corresponding isocyanate in chloroform
afforded the desired ureas 97−102 (Scheme 4).
In Table 2, the diarylurea moiety resulted in excellent kinase
inhibition with a nanomolar IC50 value against KDR. There was
a selectivity of (aryloxy)quinazoline for this enzyme, against
EGFR. Introduction of a methyl group on the phenyl ring (99−
102) led to an excellent inhibitory activity against VEGFR-2, in
particular for naphthalene derivative 102 (IC50 = 2 nM against
VEGFR-2). Substitution of the aromatic group by a bulkier
group such as a halide or a naphthalene was tolerated. This
enzymatic inhibition was limited by introduction of an electron
donor group such as methoxy (87 and 97). These quinazolines
were further evaluated in cell-based assays and, unfortunely,
were inactive toward different cancer cell lines.
charges were attributed by the Gasteiger−Huckel method, and
̈
the energy minimization was run with the Tripos force field.50
To produce a meaningful conformation for their alignment,
the best compounds of the two main substitution patterns (91
and 92) were docked into the VEGFR-2 active site (Table 2).
This was carried out by selecting a suitable reference, that is, a
structure of the enzyme cocrystallized with an inhibitor
structurally related to our compounds. The structure of the
protein cocrystallized with a suitable furopyrimidinic inhibitor
(103) was obtained from the Protein Data Bank (http://www.
pdb.org)51 under the entry 1ywn.52 The cocrystallized inhibitor
and water molecules were removed, and hydrogens were added
to the protein, taking care to be as close as possible to the
biological protonation state of the residues. The binding mode
of the compounds was investigated by a two-step process. They
were first docked into the binding site of VEGFR-2 with GOLD
3.2,53 the binding site being defined as a 10 Å sphere around
the cocrystallized inhibitor. The multiple conformations
generated were then ranked by an in-house consensus scoring
based on Goldscore53 and X-Score.54 The consistency of the
best ranked conformation was visually assessed to make sure it
was the most stable binding mode prediction. The molecules
were aligned on the best docking conformation of the closer
reference compound to give a putative bioactive conformation.
The QSAR models were achieved by randomly dividing the
pool of 49 compounds into a series of training sets of 33 mol-
ecules and corresponding test sets of the 16 remaining mol-
ecules (Table 3), following the recommendations by Oprea.55
A later batch of 16 more compounds were predicted before
their testing as an external test set, therefore assessing the
models from the initial 33 compounds on a very large 32-
molecule test set. Each training set was employed to generate
models using the CoMFA56 and CoMSIA57 methodologies
against the activity of the molecules expressed as pIC50, that is,
the opposite of the logarithm of their activity, using the same
set of default parameters and the same box for all the models. If
they passed a q2 threshold of 0.3, their external prediction
capacity was tested by predicting the values of pIC50 for the
MOLECULAR MODELING
■
All the calculations have been carried out under the Sybyl 6.9.1
molecular modeling package49 running on Silicon Graphics
Octane 2 workstations. The ligands were built from the internal
fragment library of Sybyl, and their geometries were optimized
by the Powell method available in the Maximin2 procedure to a
gradient of 0.001 kcal/(mol·Å). The dielectric constant was set
to 7.4 to implicitly represent a biological middle, the atomic
2
corresponding test set compounds, yielding an rpred value.
Another relevant parameter is the number of components N of
the models, showing the relative complexity of the underlying
mathematical model. The resulting fields were depicted as the
1192
dx.doi.org/10.1021/jm2013453 | J. Med. Chem. 2012, 55, 1189−1204

Journal of Medicinal Chemistry
Article
Table 1. Enzymatic (EGFR/VEGR-2) and Cellular (PC3, Prostate Cancer Cell; HT29, Colon Cancer Cell; MCF-7, Breast
Cancer Cell) Results for Quinazoline Derivatives
a,d
compound
proliferative inhibition, % (or IC50, μM)
a
,
b
a,c
no.
series
R
EGFR IC50, μM
VEGFR-2 IC50, μM
PC3
9%
HT29
0%
MCF-7
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
B
B
B
B
B
C
C
C
C
C
C
C
C
D
D
E
H
4.30
3.60
0.38
5.70
>10
>10
>10
6.90
>10
7.05
0.80
4.80
6.15
1.00
1.00
6.40
4.00
0.9
7.00
0.80
5.30
1.65
7.30
4.0
7%
4-Br
37%
6.50 μM
1%
2%
42%
3-Cl-4-F
4-Br-2-F
5.70 μM
0%
4.90 μM
31%
0%
4-(NHCOCH2CH3)
0%
0%
3-CH3-4-(NHCO(CH2)2CH3)
3-CH3-4-(NHCO(CH2)3CH3)
4-(NHCOOCH3)
0%
0%
5%
6.80
5.80
5.60
5.00
6.80
5.05
5.05
0.50
3.30
4.90
0.85
0.65
0.85
5.20
5.25
5.55
7.80
6.50
4.65
5.80
5.10
6.20
4.30
6.90
1.00
6.45
5.20
3.35
6.95
3.70
6.60
7.45
7.25
5.30
7.40
4.80
9.40
5.70
0%
0%
0%
2%
10%
27%
6%
4-(NHCOOCH2CH3)
4-(NHCOO(CH2)2CH3)
4-(NHCOO(CH2)3CH3)
4-(NHCOOCH(CH3)2)
3-Cl-4-(NHCOOCH3)
3-Cl-4-(NHCOOCH2CH3)
3-Cl-4-(NHCOO(CH2)2CH3)
3-Cl-4-(NHCOO(CH2)3CH3)
3-CH3-4-(NHCOOCH3)
3-CH3-4-(NHCOOCH2CH3)
3-CH3-4-(NHCOO(CH2)2CH3)
3-CH3-4-(NHCOO(CH2)3CH3)
3-CH3-4-(NHCOOCH(CH3)2)
4-Cl-3-(NHCOOCH2CH3)
4-CH3-3-(NHCOOCH3)
4-CH3-3-(NHCOOCH2CH3)
−CH2CH3
12%
1%
16%
18%
29%
19%
36%
0%
20%
18%
19%
9.80 μM
0%
32%
24%
22%
30%
0%
29%
20%
5%
0%
41%
6%
10%
0%
0%
0%
0%
0.9
6%
18%
0%
7.60
5.80
0.90
7.20
5.00
>10
>10
>10
>10
>10
9.90
>10
9.45
>10
>10
>10
>10
>10
9.10
>10
>10
>10
9.15
>10
>10
4%
13%
33%
30%
39%
12%
1%
3%
2%
13%
0%
4%
0%
0%
0%
0%
0%
6%
−(CH2)3CH3
phenyl
14%
35%
43%
9.40 μM
36%
16%
36%
0%
16%
26%
1.50 μM
7.15 μM
7.70 μM
26%
10%
9.60 μM
0%
5.80 μM
45%
3-methoxyphenyl
3-chloro-4-fluorophenyl
3-Cl-4-F
6.20 μM
5.90 μM
7.50 μM
2.55 μM
38%
4-Br-2-F
3-Br-4-CH3
3-Cl-4-(NHCOOCH3)
3-Cl-4-(NHCOOCH2CH3)
3-CH3-4-(NHCOOCH3)
3-CH3-4-(NHCOOCH2CH3)
p-(NHCONHCH2CH3)
4-Br-2-F
0%
5.45 μM
1%
0%
0%
0%
0%
11%
0%
0%
0%
0%
0%
35%
0%
3-CH3-4-(NHCOOCH2CH3)
4-Br-2-F
0%
36%
0%
31%
11%
40%
0%
3.60 μM
2.55 μM
2.50 μM
6.15 μM
5.80 μM
7.70 μM
33%
7.10 μM
0%
E
3-Cl-4-(NHCOOCH3)
3-Cl-4-(NHCOOCH2CH3)
3-CH3-4-(NHCOOCH3)
3-CH3-4-(NHCOOCH2CH3)
E
E
E
0%
0%
1193
dx.doi.org/10.1021/jm2013453 | J. Med. Chem. 2012, 55, 1189−1204

Journal of Medicinal Chemistry
Table 1. continued
compound
Article
a,d
proliferative inhibition, % (or IC50, μM)
a
,
b
a,c
no.
series
R
EGFR IC50, μM
VEGFR-2 IC50, μM
PC3
5%
HT29
0%
MCF-7
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
F
3-Cl-4-F
4-Br-2-F
>10
5.90
>10
>10
>10
>10
>10
>10
>10
>10
>10
>10
>10
>10
>10
9.80
0.60
8.90
0.70
0.60
0.40
1.00
0.06
0.07
0.04
0.09
0.05
0.03
0.60
0.60
0%
F
16%
0%
0%
0%
F
3-(NHCOCH3)
4-(NHCOOCH3)
4-(NHCOOCH2CH3)
p-((CH2)3CH3)
0%
5%
F
0%
1%
4%
F
28%
30%
42%
23%
12%
36%
35%
2%
0%
0%
G
G
G
G
G
G
G
G
G
G
31%
8%
39%
0%
p-cyclohexyl
p-phenyl
10%
9%
12%
0%
p-(4-methoxyphenyl)
p-(3-chloro-4-fluorophenyl)
p-(2,4-difluorophenyl)
p-(4-bromophenyl)
p-(2-naphthyl)
26%
10%
13%
0%
30%
6.30 μM
10%
5%
6.80 μM
0%
m-phenyl
19%
19%
7%
m-(4-methoxyphenyl)
2%
0%
a
b
Compounds tested at a concentration of 10 μM. The values are the mean SD of at least three independent experiments (SD < 10%). Inhibition
c
of EGFR (purified from human carcinoma A431 cells) tyrosine kinase activity. Inhibition of VEGFR−2 (recombinant human protein) tyrosine
kinase activity. Cell proliferation was realized by MTS assay at 10 μM from at least three independent determinations. Higher concentrations were
d
not used to avoid precipitation of the compounds in the culture medium.
a
Scheme 4
a
Reagents and conditions: (a) nBu4N+Br−, 2-butanone, 20% NaOH, reflux (84%); (b) isocyanate derivatives, NEt3, CHCl3, room temperature (35−
55%).
zones where the contributions of the fields to the activity were the
largest (positive influence, >80%; negative influence, <20%).
To validate the docking, the cocrystallized ligand 103 was
docked using the same procedure as that used for our com-
pounds. The highest scoring conformation is shown in Figure 3
with the cocrystallized conformation. Its rmsd versus the
crystallographic conformation is a low 0.436 Å. Moreover, all
30 solutions are closely superimposed and form the same
hydrogen bonds with Glu 883 (H on the nitrogens of the urea),
with the backbone H of Asp 1044 (carbonyl of the urea), and
with the backbone H of Cys 917 (N1 of the pyrimidinic cycle).
The only real difference is the position of the terminal methoxy,
lying in the plane of the aromatic group in the crystallographic
structure and perpendicular to it in the docked conformation.
We therefore concluded our docking protocol was sound and
assumed the resultant conformation to be representative of the
bioactive binding mode of our highest activity compounds.
The substitution pattern of the central aromatic tensor of our
molecules is either a para or a meta disubstitution. As easily
understood, the para pattern gives a conformation closer to that
of the cocrystallized compound (Figure 4a), with a good
conservation of the spatial position of the molecule from the
terminal aromatic group up to the central aromatic tensor. The
dimethoxyquinazoline is oriented in the same direction as the
central aromatic−urea chain, rather than being upright as in
103. However, the same H bonds are found, with the distal
nitrogen of the quinazoline linking with Cys 917 rather than the
proximal nitrogen as in the reference compound. The second
leader we have docked conserves the position of the important
urea and has much the same conformation as its para congener.
However, the aromatic tensor plane forms an angle of about
140° with 92. Nonetheless, the distal nitrogen of the dimethoxy-
quinazoline is at 1.1 Å in the 92 counterpart and is therefore also
implied in a H bond with Cys 917 (Figure 4b).
The remaining compounds were aligned on either 91 or 92
following their substitution pattern to yield an alignment. From
it, a number of models were created as outlined previously and
evaluated for their robustness (N, r2) and predictive power (q2,
rpred2). The five best models are summarized in Table 4.
Interestingly, all of them present a number of components N
lower than one-fourth of the compounds included in their
elaboration, thus meaning they are able to handle the activity
values with a fairly simple equation. They are characterized by
high r2 and q2 above the threshold value of 0.3 we have imposed
in the initial selection of the models because this value means a
confidence of 95% that the resulting model is not due to a
chance correlation.58 The major indicative factor of a predictive
2
model is its rpred value. With the exception of the CoMSIA
model generated from training set 5, all are above 0.7, heralding
the capacity of the models to predict the activity of a previously
1194
dx.doi.org/10.1021/jm2013453 | J. Med. Chem. 2012, 55, 1189−1204

Journal of Medicinal Chemistry
Article
Table 2. Enzymatic (EGFR/VEGR-2) and Cellular (PC3, Prostate Cancer Cell; HT29, Colon Cancer Cell; MCF-7, Breast
Cancer Cell) Results for New Quinazoline Urea Derivatives
a
b
Compounds tested at a concentration of 10 μM. The values are the mean SD of at least three independent experiments (SD < 10%). Inhibition
c
of EGFR (purified from human carcinoma A431 cells) tyrosine kinase activity. Inhibition of VEGFR-2 (recombinant human protein) tyrosine
kinase activity. Cell proliferation was realized by MTS assay at 10 μM from at least three independent determinations. Higher concentrations were
d
not used to avoid precipitation of the compounds in the culture medium.
unknown compound with a fairly high accuracy of more than
this aspect. Third, their external predictivity does not appear to
be too strongly overestimated. It should be kept in mind that
the higher the rpred2, the more confident one should be regard-
ing the predictive power of a model. However, evaluation of
the predictive capacity is tricky and can very easily lead to a
false overconfidence over the reality of the prediction. We were
2
70%. Although it has the highest rpred for both CoMFA and
CoMSIA, model 1 is somewhat hampered by its rather low q2,
indicating it has a poorer internal prediction capacity than the
other models. Moreover, it also has a rather high number of
components N for its CoMSIA part, indicating that the
underlying mathematical model is more complex. Model 2 has
the same number of components, but displays improved q2 and
2
rather suspicious of the rpred nearing 0.90. In the meantime,
the goal of the models is to provide biological activities of the
2
only a very slight decrease in rpred for the CoMFA method.
compounds prior to their testing as accurately as possible, that
2
This is correlated to a possible instability in the predictions of
the activity of the training set, as the leave-one-out cross-
validated r2 represents the internal predictive power. In essence,
at least one of the molecules in the training set could not be
predicted from the model derived from the others, resulting in
a large prediction error. While fairly attractive, we have not
chosen it as our final model due to its relatively large N. The
next model fared little better. It is quite correct in terms of
predictive power but in our opinion lacks high enough q2
values, being the second worst model. Moreover, its practical
use may be overburdened by the high number of components
of the CoMFA part. The two remaining models share some
features of interest. First, they display a balanced profile of
stability (r2) and predictivity (q2 and rpred2) that can be seen as a
proof of their inherent robustness. Second, they have a low
number of components, model 4 being better than model 5 in
is, with the highest rpred possible. We therefore set model 4 as
our predictive tool for the last batch of 16 molecules. Interest-
ingly, all the models had a quite close repartition of fields. For
the CoMFA models, the steric field is slightly less important
than the electrostatic field, in a proportion of about 40/60. For
the CoMSIA models, the global trend in field repartition is also
well conserved among the models, although the individual
percentages between the fields are more varied. Roughly, the
steric contribution is about 5%, while the electrostatic fields
contribute about 20%. Acceptor fields are fairly unimportant,
with a 10% contribution, while hydrophobic and hydrogen
bond donor fields are respectively 30% and 35%. This may be
relatively well explained when noting that the loss of donor
hydrogen bonds around the urea has a larger impact on the
binding than the acceptor capacity or size of the molecule, as
long as it is able to fit in the pocket. In our study, the large
1195
dx.doi.org/10.1021/jm2013453 | J. Med. Chem. 2012, 55, 1189−1204

Journal of Medicinal Chemistry
Article
Table 3. CoMFA- and CoMSIA-Predicted Activities
compound
set
test
pIC50
CoMSIA prediction
CoMSIA difference
CoMFA prediction
CoMFA difference
35
36
37
38
39
40
41
42
43
44
45
46
47
49
50
51
55
57
58
59
60
61
62
63
65
66
67
69
71
72
74
75
77
80
81
82
83
84
85
87
89
92
93
97
98
99
100
101
102
5.16
6.12
5.27
5.80
5.14
5.36
5.16
5.24
5.25
5.30
5.16
5.29
5.29
5.48
5.31
6.07
5.28
5.11
5.19
5.33
5.23
5.29
5.21
5.37
6.00
5.19
5.28
5.16
5.18
5.13
5.28
5.13
5.03
6.22
5.05
6.17
6.19
6.42
6.00
7.17
7.03
6.19
5.16
7.40
8.22
7.40
8.40
8.22
8.70
5.44
5.66
5.37
5.81
5.22
5.54
5.31
5.09
5.06
5.14
5.42
5.13
5.25
5.21
5.54
5.49
5.37
4.98
4.82
5.30
5.16
5.65
5.44
5.83
5.83
5.52
5.53
5.43
4.97
5.55
5.53
5.14
5.06
6.66
4.78
6.40
6.14
6.60
6.49
6.96
7.13
4.74
4.86
6.81
7.84
7.53
8.13
7.98
7.66
−0.28
0.46
5.56
5.54
5.57
5.73
5.18
5.65
5.31
5.22
5.01
4.82
5.58
5.04
5.39
5.18
5.75
5.73
5.43
5.32
5.20
5.39
5.31
6.17
5.97
6.09
5.87
5.07
5.41
5.64
5.10
5.58
5.29
4.69
4.78
6.65
4.96
6.15
5.92
6.33
6.44
7.32
7.32
5.25
5.18
7.08
7.47
7.75
7.87
7.72
7.62
−0.40
0.58
training
training
test
−0.10
−0.02
−0.08
−0.18
−0.14
0.14
−0.30
0.06
test
−0.04
−0.29
−0.15
0.01
training
training
training
test
0.19
0.24
training
test
0.16
0.48
−0.25
0.16
−0.42
0.25
training
training
training
test
0.04
−0.10
0.29
0.27
−0.23
0.58
−0.44
0.34
training
training
test
−0.09
0.13
−0.15
−0.22
−0.01
−0.06
−0.08
−0.88
−0.77
−0.72
0.13
test
0.36
test
0.04
test
0.07
test
−0.36
−0.24
−0.46
0.17
training
training
training
training
training
training
training
test
−0.33
−0.25
−0.28
0.21
0.12
−0.13
−0.48
0.08
−0.42
−0.26
−0.01
−0.03
−0.44
0.27
−0.45
−0.01
0.45
training
test
training
training
training
training
training
training
training
training
training
test
0.25
−0.43
0.09
−0.23
0.05
0.02
0.27
−0.18
−0.49
0.21
0.09
−0.44
−0.15
−0.29
0.94
−0.10
1.45
training
test
0.30
−0.03
0.32
0.59
training
training
training
training
test
0.39
0.75
−0.13
0.27
−0.35
0.53
0.24
0.50
1.03
1.08
effect of the hydrophobicity is somewhat of a curse, as a high
hydrophobicity could have a negative impact on the solubility
of the compounds and therefore their cellular effect, while
contributing favorably to the binding at a molecular level.
The repartition of the compounds in the training and test
sets is shown in Figure 5. All three broad classes of activity (good,
from 1 to 2.4; average from 2.5 to 3.4; low above 3.5) are well
represented in both sets, with maybe a slight overrepresentation
of the good activity compounds in the test set (four of the six
compounds of this class). This does not show up prominetly in
the predictions of either the CoMSIA (Figure 6) or the CoMFA
(Figure 7) model, as there are only two compounds predicted
with about 1 log of error by both methods. One of them is
compound 102, which is the most active molecule. Due to the
randomness of the repartition in the training and test sets, it fell
in the test set and could not be predicted accurately from the
compounds in the corresponding training set. By comparison,
the second best compound, 91, was attributed to the training set
and is well predicted, therefore showing that the extrapolation
from a model should be taken with caution. The second
1196
dx.doi.org/10.1021/jm2013453 | J. Med. Chem. 2012, 55, 1189−1204

Journal of Medicinal Chemistry
Article
compounds, such as 98, 100, and 101, were predicted much
more accurately, with only compound 98 being missed by more
than 0.5 log, and only by the CoMFA method. All of them were
used for the training of the model, therefore stressing the
possible effect of the underrepresentation of the highest activity
class in the training set, rendering the model less able to
extrapolate to the best pIC50 values. It should also be noted that
compound 69 was not included in the model due to the un-
determined nature of its IC50. A series of 16 already synthetized
compounds were tested later. We used them to further test the
predictive ability of our model by estimating their activity
before their testing as an external validation set. The results are
presented in Table 5.
Interestingly, the predictions were close enough to the actual
values, with 14 compounds out of 16 predicted within 1 log of
the experimental values and 11 within 0.5 log. Compound 48
was within 1.2 log, and only compound 79 was really mis-
predicted, with about 1.5 log of difference for both the CoMFA
and CoMSIA models. This could be due to its somewhat
shorter length compared to those of most of the other mol-
ecules included in the model, as it is a key synthesis inter-
mediary, lacking the urea moiety. With respect to the reduced
initial training set, these results implied the model fared fairly
well. As a posttesting validation, we tried to create a model
trained on all 49 compounds available in the first stage to more
accurately predict the remaining 16. The CoMFA and CoMSIA
models behaved quite well (Table 6). They are comparable to
model 4, with an improved CoMFA r2, but a poorer CoMFA q2.
However, they were not better in the prediction of the second
series of compounds (Figure 8). In particular, compound 79 is
still mispredicted by about 1.5 log. We therefore have not
evaluated any further the development of this model and have
reverted to model 4.
Figure 3. Docking conformation of furo[2,3-d]pyrimidine 103
synthesized by Y. Miyazaki52 (colored by atom types) compared
with the crystallographic conformation (in green).
The two methods we have employed have the major
advantage of providing a visual 3D depiction of the molecular
fields around the compounds. We have compared them with
the docking of the two lead compounds of our series. The
CoMFA fields (Figure 9) give indications of the steric and
electrostatic zones of interactions. There are two favorable
contours for the steric field. A large contour is next to the
aromatic group of the central tensor, where the pocket is widely
open. A smaller one is at the other end of the urea, showing the
interest of a substitution of an appropriate length on this
moiety. The pocket is relatively tightly packed, so the naphtha-
lene of 91 barely fits in this zone, just as the methoxyphenyl of
93. A totally flexible end group such as the butane of 84 is also
of the appropriate length. However, as none of our compounds
have a longer chain, we can only guess they should not fit as
well in the binding site. The sterically restricted zones comple-
ment this first view with a cluster of medium-sized contours
around the aromatic tensor, on the side of the benzyl that is
opposite that of the favorable zone. This is again in agreement
with the binding site geometry, as they correspond to Val 846
and 914, closing this part of the pocket. Interestingly, a large
hindrance zone and several smaller zones are placed at the
entry of the pocket, indicating that too long groups on the
quinazoline side are not useful for the activity and indeed are
even unfavorable. This may be linked to the larger moieties
replacing the two methoxy groups of the reference molecules,
indicating either a loss of activity due to a lowered solubility or
just an unfavorable entropic effect of fixing these long chains in
a restrictive conformational environment. The electrostatic field
is somewhat less informative, with respectively a small and a
Figure 4. (a) Docking conformation of 91. (b) Docking of 92.
mispredicted compound is 92, which was again classed in the
test set. It fared rather poorly in the CoMSIA model, but slightly
better in the CoMFA model. In the meantime, high-activity
1197
dx.doi.org/10.1021/jm2013453 | J. Med. Chem. 2012, 55, 1189−1204

Journal of Medicinal Chemistry
Article
a
Table 4. Statistical Results for the Top Five Models
CoMSIA
q2
CoMFA
q2
model no.
N
r2
rpred
fields
N
r2
rpred
fields
2
2
1
2
3
4
5
5
5
3
4
5
0.94
0.94
0.89
0.92
0.95
0.62
0.72
0.63
0.74
0.79
0.86
0.75
0.85
0.73
0.66
S, E
S, E
S, E
S, E
S, E
3
3
8
3
3
0.86
0.89
0.98
0.86
0.86
0.49
0.56
0.58
0.60
0.65
0.88
0.81
0.80
0.75
0.73
S, E, H, D, A
S, E, H, D, A
S, E, H, D, A
S, E, H, D, A
S, E, H, D, A
a
Fields used in the models: S (steric), E (electrostatic), H (hydrophobic), D (hydrogen bond donor), A (hydrogen bond acceptor).
Table 5. Prediction for the 16 Compounds of the External
Validation Set
CoMSIA
prediction
CoMSIA
difference
CoMFA
prediction
CoMFA
difference
compound pIC50
48
52
53
54
64
79
80
86
88
90
91
73
76
78
68
70
6.28
6.19
6.07
5.28
5.16
5
5.145
5.212
5.115
5.281
5.478
6.538
6.362
6.903
6.446
7.099
7.145
5.087
5.001
4.996
4.922
5.489
1.135
0.978
5.314
5.522
5.54
0.966
0.668
0.53
0.955
−0.001
−0.318
−1.538
−0.142
0.317
5.164
5.476
6.475
6.209
7.36
0.116
−0.316
−1.475
0.011
−0.14
0.761
−0.169
0.086
−0.135
0.179
0.239
0.636
0.309
6.22
7.22
7.4
0.954
6.639
7.389
7.434
5.275
5.141
5.011
4.834
5.121
7.22
7.52
5.14
5.32
5.25
5.47
5.43
0.121
Figure 5. Repartition of the activities of the compounds in the training
(blue tilted squares) and test (orange triangles) sets of model 4.
0.375
0.053
0.319
0.254
0.548
−0.059
Table 6. Statistical Results for Model 6
CoMSIA
r2
CoMFA
r2
N
q2
N
q2
4
0.894
0.753
3
0.659
0.870
Figure 6. Predictions of CoMSIA model 4 for the training (blue tilted
squares) and test (orange triangles) sets.
Figure 8. Residuals of the second series of compounds by models 4
and 6 for the CoMFA (orange triangles) and CoMSIA (blue tilted
squares) methods.
Figure 7. Predictions of CoMFA model 4 for the training (blue tilted
squares) and test (orange triangles) sets.
that these aromatic groups must bear a slightly decreased elec-
tronic load. The unfavorable contours and the favorable zone at
the methoxy level are nearly useless as they are not related to
the protein.
large favorable zone in the middle of the terminal and central
aromatic groups of the reference compounds, offering the idea
1198
dx.doi.org/10.1021/jm2013453 | J. Med. Chem. 2012, 55, 1189−1204

Journal of Medicinal Chemistry
Article
Figure 9. Steric (green, favorable; yellow, unfavorable) and electro-
static (blue, positive charge favorable; red, negative charge favorable)
CoMFA fields.
The CoMSIA steric fields (Figure 10a) are larger and more
easily interpretable than their CoMFA counterparts, but point
to the same trend. Two large favorable contours surround the
terminal and central aromatics, clearly hinting to the
importance of occupying the right end side of the pocket by
a substitution on the urea. The central zone only includes the
meta-substituted tensor. This may be an artifact related to the
larger number of 1,4-substituted molecules having less impact
on the variations of the fields. On the contrary, the two un-
favorable zones near the methoxy point again to the worsened
activity of compounds with longer chains on the oxygens. As
previously described, the favorable electrostatic contours
(Figure 10a) show the importance of the electronic load of
the central aromatic group. However, the urea is surrounded by
two zones of major interest. A favorable one is placed next to
the carboxy and an unfavorable one next to the distal nitrogen.
They may be related to the hydrogen bonding occurring at
these positions. The same is true for the favorable zone lying
just under the distal nitrogen of the quinazoline. However, the
interest of the larger contours around the methoxy is low, due
to the aforementioned unfavorable steric effect of long chains in
this position and the near to none correlation with residues of
the protein. A large hydrophilic contour nearly completely sur-
rounds the aromatic tensor (Figure 10b). This could herald the
lowered activity of molecules bearing a substitution of this
element other than the straightforward 1,3- or 1,4-disubstitu-
tion. However, the smaller hydrophobic region that includes
most of the urea is less easy to understand. It might come from
the shorter compounds completely missing the urea, for which
a hydrophobic substitution is the most common in our study.
These compounds are in fact synthesis intermediaries that were
tested to assess the effect of the urea. The hydrogen bond
donor fields (Figure 10c) display a large favorable contour in
front of the nitrogen of the urea. As they are involved in a
critical interaction with Glu 883, this contour is in very close
agreement with the docking and the previous knowledge of the
subject. A medium-sized unfavorable contour appears next to
the oxygen of the tensor. Our compounds are built around
either a phenoxyquinazoline or an anilinoquinazoline. There-
fore, we can conclude easily that the hydrogen capacity of the
anilinoquinazoline is not favorable to the activity of the
molecules, bringing a fresh insight into the further development
Figure 10. CoMSIA fields of 91 and 92: (a) steric (green, favorable;
yellow, unfavorable) and electrostatic (blue, positive charge favorable;
red, negative charge favorable), (b) hydrophobic (yellow, hydrophobic
favorable; white, hydrophilic favorable), (c) hydrogen bond donor
(cyan, favorable; purple, unfavorable) and acceptor (magenta,
favorable; red, unfavorable).
of the series. The acceptor fields are also in very good
agreement with the docking. A small favorable contour appears
around the Cys 917 backbone hydrogen involved in an
interaction with the quinazoline. An unfavorable contour is
1199
dx.doi.org/10.1021/jm2013453 | J. Med. Chem. 2012, 55, 1189−1204

Journal of Medicinal Chemistry
Article
Table 7. Prediction for the Four Newly Synthetized Analogues
a
Scheme 5
a
Reagents and conditions: (a) nBu4N+Br−, 2-butanone, 20% NaOH, reflux (40−70%); (b) isocyanate derivatives, NEt3, CHCl3, room temperature
(10−38%).
displayed above the carbonyl of the urea. This group is engaged
in a hydrogen bond with the backbone of Asp 1044, so this
result more closely attracted our attention. Upon a closer in-
vestigation, this zone prohibits the placement of the carbonyl in
an upward position that would hinder its interaction with Asp
1044 rather than clueing toward an unfavorable effect of this
interaction. A few compounds of very feeble activity are point-
ing their carbonyl in this region, with a corresponding loss of
the third hydrogen bond of the urea. Lastly, a large favorable
contour is drawn around the oxygen of the tensor. It shows that
a nitrogen in this position is not the best choice, and reinforces
the unfavorable hydrogen donor contour. Taken together, the
hydrogen bond fields offer an interesting view of the best
scaffold in the series, which should be able to accept a hydrogen
from Cys 917 and Asp 1044, give one or two hydrogens to Glu
883, and bear a tensor devoid of a hydrogen donor. All these
conclusions are in excellent agreement with the docking, further
corroborating the hypothesis of the binding mode we made at
the first step of our study.
were proposed, following these guidelines and the interest of
the substitution of the aromatic linker with a hydrophobic moiety,
among which the bioisostery of methyl to halogen. Four com-
pounds were chosen for biological evaluation as they were pre-
dicted to be the most promising (106−109) (Table 7). These
compounds were synthesized according to procedures
described in Scheme 5 with the 4-chloro-6,7-dimethoxyquina-
zoline (7) as the starting product. Selective reaction of chlo-
ride derivative 7 with 4-amino-3-chlorophenol or 4-amino-2-
chlorophenol and tetra-n-butylammonium bromide in a mixture
of methyl ethyl ketone and sodium hydroxide provided the
intermediates 104 and 105. Condensation of 104 or 105 with
the corresponding isocyanate in chloroform afforded the
desired ureas 106−108 and 109 (Scheme 5).
Most interestingly, none of the compounds were predicted
with an error as high as 0.3 log by the CoMSIA model, while
the CoMFA model fared somewhat worse, with two under-
estimated molecules at more than 0.5 log from their experi-
mental value. Moreover, it appeared that the overall ranking of
the compounds prior to their synthesis presented no match
between the CoMFA ranking and the experimental pIC50. On
the other hand, the best and worst compounds of this new
series were correctly guessed by the CoMSIA model. In more
detail, compound 108, the chlorinated congener of compound
102, was predicted to be close to the latter by the CoMSIA
As they have proved to be quite predictive, these models
were employed to imagine new structures and help rank them
to prioritize the synthesis of the most promising. We have
chosen molecules related to the previous series to keep the
synthetic route close to those already described and to preserve
the cohesion of the whole series. A number of new structures
1200
dx.doi.org/10.1021/jm2013453 | J. Med. Chem. 2012, 55, 1189−1204

Journal of Medicinal Chemistry
Article
model, while it was predicted to be noticeably worse by the
CoMFA model, which proved to be very precise. This could be
due to a less accurate account of the charge repartition on the
aromatic center bearing the chlorine by the CoMSIA model.
On the basis of methylated compounds 100−102, we have
tried the chlorinated aromatic tensor analogues 106−109.
Relocating the chlorine on this benzene from position 3 to
position 2 had the effect of improving the CoMSIA-predicted
values, as we expected in part due to the fact that compound
109 is the sole molecule with this substitution pattern. How-
ever, it had the reverse effect for CoMFA, most probably due to
an overevaluation of the electrostatic effect of the halogen on
the charge repartition of the central aromatic group. Experi-
mentally, compound 109 was slightly overpredicted by the
CoMSIA model, while compound 106 was largely under-
estimated by the CoMFA model. Comparing these compounds
with their methylated congener, compound 100, the CoMSIA
model accurately predicted the new molecules to be somewhat
less good, but rather close to compound 101, of only slightly
lower activity. The effect of this different pattern on the CoMFA
prediction is intriguing. Although this model consistently gave not
so good predictions for the best compounds of the series, such a
high underestimation for compound 106 appears to be related
again to the higher electrostatic field contribution in this model.
Compound 107, the chlorinated analogue of compound 101,
suffered from the same drawback, with a serious underesti-
mation from the CoMFA model, while it has been reasonably
well predicted by the CoMSIA model. In hindsight, these com-
pounds showed the validity of the models in interpolating to
more fully complete the chemical cover of the series but also
illustrated the greater robustness of the CoMSIA predictions
and the rather low interest of the CoMFA model in predicting
better than average activity compounds, maybe due to an over-
emphasis on the electrostatic field around the central aromatic
group.
contour, up to the largest group still able to fit in the binding site,
with a combined docking−QSAR−in vitro assay methodology.
EXPERIMENTAL SECTION
■
Chemistry. Melting points were determined with a Buchi 535
̈
capillary melting point apparatus and are uncorrected. Kieselgel 60 F-
254 commercial plates were used for analytical TLC as well as UV light
and/or with iodine to follow the course of the reaction. Flash chro-
matography (FC) was performed with silica gel Kieselgel Si 60, 0.063−
0.200 mm (Merck). The structure of each compound was confirmed
1
by IR (Bruker VECTOR 22 instrument) and H NMR (300 MHz,
Bruker AC300P spectrometer). Chemicals shifts (δ) are reported in
parts per million downfield from TMS, J values are in hertz, and the
splitting patterns are abbreviated as follows: s, singlet; d, doublet; t,
triplet; q, quartet; m, multiplet. The purity of the compounds was
tested by HPLC separation followed by APCI+ (atmospheric pressure
chemical ionization) mass spectral detection on an LC−MS system,
Thermo Electon Surveyor MSQ, and was >95%.
4-Chloro-6,7-dimethoxyquinazoline (7).43 To a solution of
methyl-2-amino-4,5-dimethoxybenzoate (4.0 g, 19.0 mmol) in DMF
(40 mL) and methanol (10 mL) were added formamide (76.0 mmol)
and sodium methoxide (54.0 mmol). The resulting mixture was
refluxed for 16 h. After quenching by water (100 mL), the mixture was
neutralized by 1 M HCl solution. The precipitate was collected by
filtration, washed with H2O (30 mL) and Et2O (30 mL), and dried in
vacuo to give quinazolinone 2 as a white solid (82%) which was used
directly in the next step. A mixture of 2 (3.0 g, 15.0 mmol) and phos-
phorus oxychloride (30 mL) was refluxed for 2 h. After removal of the
solvent, the residue was dissolved in ice−water (50 mL), and the
mixture was neutralized by ammonium hydroxide. The precipate was
collected by filtration and dissolved in CH2Cl2 (100 mL). The organic
layer was washed with a 1 M solution of K2CO3 (3 × 40 mL) and
brine (1 × 40 mL) and dried over CaCl2, and the solvent was removed
under reduced pressure. The spectroscopic data for compound 3 are in
agreement with those reported in the literature.
4-[(6,7-Dimethoxy-4-quinazolinyl)oxy]aniline (94).43 A solution
of 7 (2.00 g, 8.90 mmol), 4-aminophenol (1.17 g, 10.70 mmol), and
tetra-n-butylammonium bromide (1.43 g, 4.45 mmol) in methyl ethyl
ketone (20 mL) and a 20% solution of NaOH (10 mL) was refluxed
for 30 min. After dilution by CHCl3 (100 mL) and H2O (20 mL), the
organic layer was washed with water and brine and dried over MgSO4.
The solvent was removed under reduced pressure, and then the
precipitated solid was collected by filtration and washed with MeOH
(15 mL) to give 94 as a white solid (89%). The spectroscopic data for
compound 94 are in agreement with those reported in the literature.
3-[(6,7-Dimethoxy-4-quinazolinyl)oxy]aniline (95).48 A solution
of 7 (2.00 g, 8.90 mmol), 3-aminophenol (1.17 g, 10.70 mmol), and
tetra-n-butylammonium bromide (1.43 g, 4.45 mmol) in methyl ethyl
ketone (20 mL) and a 20% solution of NaOH (10 mL) was refluxed
for 30 min. After dilution by CHCl3 (100 mL) and H2O (20 mL), the
organic layer was washed with water and brine and dried over MgSO4.
The solvent was removed under reduced pressure, and then the
precipitated solid was collected by filtration and washed with MeOH
(15 mL) to give 95 as a white solid (93%). The spectroscopic data for
compound 95 are in agreement with those reported in the literature.
4-[(6,7-Dimethoxy-4-quinazolinyl)oxy]-3-methylaniline (96). A
solution of 7 (2.00 g, 8.90 mmol), 4-amino-3-methylphenol (1.12 g,
10.70 mmol), and tetra-n-butylammonium bromide (1.43 g, 4.45 mmol)
in methyl ethyl ketone (20 mL) and a 20% solution of NaOH (10 mL)
was refluxed for 30 min. After dilution by CHCl3 (100 mL) and H2O
(20 mL), the organic layer was washed with water and brine and dried
over MgSO4. The solvent was removed under reduced pressure, and
then the precipitated solid was collected by filtration and washed with
MeOH (15 mL) to give 96 as a white solid (84%). Mp: 221−223 °C.
IR (cm−1): 3384 (NH2), 1205 (CCO), 1075 (C−O−C methoxy).
1H NMR (DMSO-d6): δ (ppm) 8.51 (s, 1H, ArH), 7.52 (s, 1H, ArH),
7.33 (s, 1H, ArH), 6.57 (d, 1H, J = 2.10 Hz, ArH), 6.43 (dd, 1H, J = 2.10
and 8.00 Hz, ArH), 6.36 (d, 1H, J = 8.01 Hz, ArH), 5.22 (s, 2H, NH2),
CONCLUSION
■
A series of 60 new VEGFR-2 inhibitors have been synthesized
and tested on the enzyme. Their activity was evaluated in vitro.
In an effort to further develop this series, a QSAR model has
been derived from this activity in a two-step process. First, the
docking of the most active compound of the two families of
molecules has been realized to achieve a putative bioactive con-
formation. Second, 3D QSAR models employing the CoMFA
and CoMSIA methods have been elaborated on the basis of an
alignment of the compounds on the putative bioactive con-
formation of the two reference molecules. The excellent agree-
ment between the docking and the QSAR fields indicates that
the binding mode hypothesis is most probably fairly close to
the biological binding mode, therefore offering a relevant basis
for proposing improvements for the further development of the
series. The second major interest of this study is the predictive
power of the models, more than their insights into the struc-
tural modifications that would enhance the activity of the series,
which is too homogeneous to provide an efficient basis for
suggesting new scaffolds. This predictive power has been vali-
dated on external test sets and further proved on newly synthe-
sized compounds before their biological evaluation. It was
employed to propose some new compounds, the most promising
of which have been synthesized and tested, with fairly promising
results. We plan to further investigate the effect of the modifica-
tion of the tail of the compounds, lying in a large favorable steric
1201
dx.doi.org/10.1021/jm2013453 | J. Med. Chem. 2012, 55, 1189−1204

Journal of Medicinal Chemistry
Article
4.02 (s, 3H, OCH3), 3.98 (s, 3H, OCH3), 2.15 (s, 3H, CH3). LC−MS
(APCI+): m/z calcd for C17H17N3O3 312 [(M + H)+].
(s, 3H, OCH3), 2.28 (s, 3H, CH3). LC−MS (APCI+): m/z calcd for
C28H24N4O4 481 [(M + H)+].
4-[(6,7-Dimethoxy-4-quinazolinyl)oxy]-3-chloroaniline (104). A
solution of 7 (2.00 g, 8.90 mmol), 4-amino-3-chlorophenol hydro-
chloride (1.93 g, 10.70 mmol), and tetra-n-butylammonium bromide
(1.43 g, 4.45 mmol) in methyl ethyl ketone (20 mL) and a 20%
solution of NaOH (10 mL) was refluxed for 30 min. After dilution by
CHCl3 (100 mL) and H2O (20 mL), the organic layer was washed
with water and brine and dried over MgSO4. The solvent was removed
under reduced pressure, and then the precipitated solid was collected
by filtration and washed with MeOH (15 mL) to give 104 as a brown
solid (70%). Mp: 208−210 °C. IR (cm−1): 3385 (NH2), 1201 (C
CO), 1070 (C−O−C methoxy), 1048 (C−Cl). 1H NMR (DMSO-
d6): δ (ppm) 8.53 (s, 1H, ArH), 7.51 (s, 1H, ArH), 7.36 (s, 1H, ArH),
7.22 (d, 1H, J = 2.67 Hz, ArH), 6.98 (dd, 1H, J = 2.79 and 8.73 Hz,
ArH), 6.84 (d, 1H, J = 8.76 Hz, ArH), 5.34 (s, 2H, NH2), 3.97 (s, 3H,
OCH3), 3.95 (s, 3H, OCH3). LC−MS (APCI+): m/z calcd for
C16H14ClN3O3 331 [(M + H)+ for 35Cl] and 333 [(M + H)+ for 37Cl].
4-[(6,7-Dimethoxy-4-quinazolinyl)oxy]-2-chloroaniline (105). A
solution of 7 (2.00 g, 8.90 mmol), 4-amino-2-chlorophenol (1.54 g,
10.70 mmol), and tetra-n-butylammonium bromide (1.43 g, 4.45 mmol)
in methyl ethyl ketone (20 mL) and a 20% solution of NaOH (10 mL)
was refluxed for 30 min. After dilution by CHCl3 (100 mL) and H2O
(20 mL), the organic layer was washed with water and brine and dried
over MgSO4. The solvent was removed under reduced pressure, and
then the precipitated solid was collected by filtration and washed with
MeOH (15 mL) to give 108 as a brown solid (40%). Mp: >250 °C. IR
(cm−1): 3385 (NH2), 1203 (CCO), 1068 (C−O−C methoxy),
1051 (C−Cl). 1H NMR (DMSO-d6): δ (ppm) 8.50 (s, 1H, ArH), 7.52
(s, 1H, ArH), 7.38 (s, 1H, ArH), 7.05 (d, 1H, J = 8.81 Hz, ArH), 6.70
(d, 1H, J = 2.62 Hz, ArH), 6.60 (dd, 1H, J = 2.32 and 8.80 Hz, ArH),
5.35 (s, 2H, NH2), 4.00 (s, 3H, OCH3), 3.95 (s, 3H, OCH3). LC−MS
(APCI+): m/z calcd for C16H14ClN3O3 331 [(M + H)+ for 35Cl] and
333 [(M + H)+ for 37Cl].
General Procedure for Urea Derivatives 97 and 98. To a stirred
solution of 94 (0.20 g, 0.67 mmol) and NEt3 (0.17 g, 1.68 mmol) in
10 mL of CHCl3 were added phenyl isocyanate derivatives (0.80 mmol).
After 16 h, the residue was filtered off, washed by chloroform (5 mL) and
petrolum ether (10 mL), and recrystallized.
N-{4-[(6,7-Dimethoxyquinazolin-4-yl)oxy]phenyl}-N′-(3-
methoxyphenyl)urea (97). Crystallization from EtOH/H20 (95/5)
gave pure 97 as a white solid (81%). Mp: 237−239 °C. IR (cm−1):
3214 (NH), 1695 (CO), 1208 (CCO), 1079 (C−O−C
methoxy). 1H NMR (DMSO-d6): δ (ppm) 8.80 (s, 1H, NH), 8.70 (s,
1H, NH), 8.52 (s, 1H, ArH), 7.50−7.60 (m, 3H, ArH), 7.36 (s, 1H,
ArH), 7.15−7.25 (m, 4H, ArH), 6.94 (d, J = 6.80 Hz, 1H, ArH), 6.55
(d, J = 6.80 Hz, 1H, ArH), 4.01 (s, 3H, OCH3), 3.96 (s, 3H, OCH3),
3.72 (s, 3H, OCH3). LC−MS (APCI+): m/z calcd for C24H22N4O5
447 [(M + H)+].
N-(3-Bromophenyl)-N′-{4-[(6,7-dimethoxyquinazolin-4-yl)oxy]-
phenyl}urea (98). Crystallization from acetonitrile gave pure 98 as a
white solid (69%). Mp: >250 °C. IR (cm−1): 3214 (NH), 1697 (C
1
O), 1206 (CCO), 1080 (C−O−C methoxy), 1056 (C−Br). H
NMR (DMSO-d6): δ (ppm) 8.75 (s, 1H, NH), 8.70 (s, 1H, NH), 8.51
(s, 1H, ArH), 7.50−7.70 (m, 3H, ArH), 7.34 (s, 1H, ArH), 7.15−7.25
(m, 4H, ArH), 6.98 (d, J = 7.40 Hz, 1H, ArH), 6.70 (d, J = 7.40 Hz,
1H, ArH), 4.01 (s, 3H, OCH3), 3.97 (s, 3H, OCH3). LC−MS
(APCI+): m/z calcd for C23H19BrN4O4 495 [(M + H)+ for 79Br] and
497 [(M + H)+ for 81Br].
General Procedure for Urea Derivatives 99−102. To a stirred
solution of 96 (0.20 g, 0.64 mmol) and NEt3 (0.19 g, 1.60 mmol) in
10 mL of CHCl3 were added phenyl isocyanate derivatives (0.83 mmol).
After 16 h, the residue was filtered off, washed by chloroform (5 mL) and
petrolum ether (10 mL), and recrystallized.
N-(4-Bromophenyl)-N′-{4-[(6,7-dimethoxyquinazolin-4-yl)oxy]-3-
methylphenyl}urea (99). Crystallization from EtOH/H20 (95/5) gave
pure 99 as a white solid (65%). Mp: >250 °C. IR (cm−1): 3213 (NH),
1695 (CO), 1206 (CCO), 1079 (C−O−C methoxy), 1057
(C−Br). 1H NMR (DMSO-d6): δ (ppm) 9.02 (s, 1H, NH), 8.52
(s, 1H, NH), 8.01 (s, 1H, ArH), 7.90 (d, J = 9.10 Hz, 1H, ArH), 7.57
(s, 1H, ArH), 7.51 (d, J = 8.70 Hz, 2H, ArH), 7.39 (s, 1H, ArH), 7.28
(m, 2H, ArH), 7.05−7.15 (m, 3H, ArH), 4.01 (s, 3H, OCH3), 3.97 (s,
3H, OCH3), 2.28 (s, 3H, CH3). LC−MS (APCI+): m/z calcd for
C24H21BrN4O4 509 [(M + H)+ for 79Br] and 511 [(M + H)+ for 81Br].
N-(3-Bromophenyl)-N′-{4-[(6,7-dimethoxyquinazolin-4-yl)oxy]-3-
methylphenyl}urea (100). Crystallization from EtOH/H20 (95/5)
gave pure 100 as a white solid (60%). Mp: >250 °C. IR (cm−1): 3214
(NH), 1695 (CO), 1206 (CCO), 1079 (C−O−C methoxy),
1052 (C−Br). 1H NMR (DMSO-d6): δ (ppm) 8.85 (s, 1H, NH), 8.75
(s, 1H, NH), 8.10 (s, 1H, ArH), 7.50−7.65 (m, 3H, ArH), 7.35 (s, 1H,
ArH), 7.15−7.25 (m, 4H, ArH), 6.95 (m, 1H, ArH), 4.01 (s, 3H, OCH3),
3.97 (s, 3H, OCH3), 2.27 (s, 3H, CH3). LC−MS (APCI+): m/z calcd for
C24H21BrN4O4 509 [(M + H)+ for 79Br] and 511 [(M + H)+ for 81Br].
N-(3-Chloro-4−fluoroanilino)-N′-{4-[(6,7-dimethoxyquinazolin-
4-yl)oxy]-3-methylphenyl}urea (101). Crystallization from acetoni-
trile gave pure 101 as a white solid (75%). Mp: >250 °C. IR (cm−1)
3213 (NH), 1695 (CO), 1206 (CCO), 1079 (C−O−C
methoxy), 1057 (C−Cl). 1H NMR (DMSO-d6): δ (ppm) 9.21 (s, 1H,
NH), 8.52 (s, 1H, NH), 8.08 (s, 1H, ArH), 7.80−7.90 (m, 2H, ArH),
7.52 (s, 1H, ArH), 7.30−7.40 (m, 2H, ArH), 7.11 (s, 1H, ArH), 7.08
(d, J = 2.30 Hz, 1H, ArH), 4.01 (s, 3H, OCH3), 3.97 (s, 3H, OCH3),
2.28 (s, 3H, CH3). LC−MS (APCI+): m/z calcd for C24H20ClFN4O4
483 [(M + H)+ for 35Cl] and 485 [(M + H)+ for 37Cl].
General Procedure for Urea Derivatives (106−108). To a stirred
solution of 104 (0.20 g, 0.60 mmol) and NEt3 (0.15 g, 1.50 mmol) in
10 mL of CHCl3 were added phenyl isocyanate derivatives (0.72
mmol). After 16 h at room temperature, the solvent was removed
under reduced pressure. The residue was dissolved in AcOEt, washed
with a solution of HCl (1 N, 3 × 15 mL), a 10% solution of K2CO3
(3 × 15 mL), and brine, and dried over MgSO4. The solvent was
removed under reduced pressure. The residue was purified by FC
(CH2Cl2/MeOH, 95/5).
N-(3-Bromophenyl)-N′-{4-[(6,7-dimethoxyquinazolin-4-yl)oxy]-3-
chlorophenyl}urea (106). Crystallization from EtOH/H20 (95/5)
gave pure 106 as a white solid (10%). Mp: >250 °C. IR (cm−1): 3210
(NH), 1694 (CO), 1219 (C−F), 1205 (CCO), 1081 (C−O−
1
C methoxy), 1055 (C−Br), 1050 (C−Cl). H NMR (DMSO-d6): δ
(ppm) 9.61 (s, 1H, NH), 8.57 (s, 1H, ArH), 8.45 (s, 1H, NH), 8.15
(d, 1H, J = 9.60 Hz, ArH), 7.89 (s, 1H, ArH), 7.55−7.58 (m, 2H,
ArH), 7.39 (s, 1H, ArH), 7.23−7.32 (m, 3H, ArH), 7.16 (m, 1H,
ArH), 3.99 (s, 3H, OCH3), 3.97 (s, 3H, OCH3). LC−MS (APCI+):
m/z calcd for C23H18BrClN4O4 529 [(M + H)+ for 79Br/35Cl], 531
[(M + H)+ for 79Br/37Cl], 531 [(M + H)+ for 81Br/35Cl], and 533
[(M + H)+ for 81Br/37Cl].
N-(3-Chloro-4-fluoroanilino)-N′-{4-[(6,7-dimethoxyquinazolin-4-
yl)oxy]-3-methylphenyl}urea (107). Crystallization from EtOH/H20
(95/5) gave pure 107 as a white solid (10%). Mp: >250 °C. IR (cm−1):
3211 (NH), 1699 (CO), 1205 (CCO), 1079 (C−O−C
1
methoxy), 1057 (C−Cl). H NMR (DMSO-d6): δ (ppm) 9.89 (s, 1H,
NH), 8.68 (s, 1H, ArH), 8.54 (s, 1H, NH), 8.16 (d, 1H, J = 9.06 Hz,
ArH), 7.83 (dd, J = 7.01 and 4.47 Hz, 1H, ArH), 7.59 (s, 1H, ArH), 7.56
(d, 1H, J = 2.16 Hz, ArH), 7.42 (s, 1H, ArH), 7.30−7.39 (m, 3H, ArH),
4.01 (s, 3H, OCH3), 3.98 (s, 3H, OCH3). LC−MS (APCI+): m/z calcd
for C23H17Cl2FN4O4 503 [(M + H)+ for 35Cl /35Cl], 505 [(M + H)+ for
37Cl /35Cl], and 507 [(M + H)+ for 37Cl /37Cl].
N-{4-[(6,7-Dimethoxyquinazolin-4-yl)oxy]-3-methylphenyl}-N′-
(2-naphthyl)urea (102). Crystallization from EtOH/H20 (95/5) gave
pure 102 as a white solid (55%). Mp: >250 °C. IR (cm−1): 3213
(NH), 1697 (CO), 1206 (CCO), 1079 (C−O−C methoxy).
1H NMR (DMSO-d6): δ (ppm) 9.17 (s, 1H, NH), 8.81 (s, 1H, NH),
8.52 (s, 1H, ArH), 8.15−8.25 (m, 2H, ArH), 7.60−7.80 (m, 2H, ArH),
7.30−7.50 (m, 5H, ArH), 7.13 (s, 1H, ArH), 7.07 (d, J = 2.10 Hz, 1H,
ArH), 6.95 (d, J = 2.40 Hz, 1H, ArH), 4.01 (s, 3H, OCH3), 3.97
N-{4-[(6,7-Dimethoxyquinazolin-4-yl)oxy]-3-chlorophenyl}-N′-(2-
naphthyl)urea (108). Crystallization from EtOH/H20 (95/5) gave
pure 108 as a white solid (10%). Mp: >250 °C. IR (cm−1): 3213
(NH), 1697 (CO), 1206 (CCO), 1079 (C−O−C methoxy),
1202
dx.doi.org/10.1021/jm2013453 | J. Med. Chem. 2012, 55, 1189−1204

Journal of Medicinal Chemistry
Article
1057 (C−Cl). 1H NMR (DMSO-d6): δ (ppm) 9.62 (s, 1H, NH), 9.33
(s, 1H, ArH), 8.53 (s, 1H, NH), 8.09 (m, 2H, ArH), 7.76−7.84 (m,
4H, ArH), 7.50 (m, 1H, ArH), 7.32−7.48 (m, 3H, ArH), 7.21 (s, 1H,
ArH), 6.74 (d, J = 8.13 Hz, 1H, ArH), 4.01 (s, 3H, OCH3), 3.97 (s,
3H, OCH3), 2.28 (s, 3H, CH3). LC−MS (APCI+): m/z calcd for
C27H21ClN4O4 501 [(M + H)+ for 35Cl /35Cl], 503 [(M + H)+ for 37Cl
/35Cl], and 505 [(M + H)+ for 37Cl /37Cl].
ACKNOWLEDGMENTS
■
Kinase assays were realized on the Binding-Platform of ICPAL,
Universite Lille 2, supported by PRIM (Pole de Recherche
Interdisciplinaire pour le Medicament). We thank the
́
̂
́
LaRMNS, RMN Laboratory, Faculty of Pharmacy.
N-(3-Bromophenyl)-N′-{4-[(6,7-dimethoxyquinazolin-4-yl)oxy]-2-
chlorophenyl}urea (109). To a stirred solution of 105 (0.20 g, 0.60
mmol) and NEt3 (0.15 g, 1.50 mmol) in 10 mL of CHCl3 was added
3-bromophenyl isocyanate (0.14 g, 0.72 mmol). After 48 h at room
temperature, the solvent was removed under reduced pressure. The
residue was dissolved in AcOEt, washed with a solution of HCl (1 N,
3 × 15 mL), a 10% solution of K2CO3 (3 × 15 mL), and brine, and
dried over MgSO4. The solvent was removed under reduced pressure.
The residue was washed by EtOH. Crystallization from EtOH/H2O
(95/5) gave pure 109 as a white solid (38%). Mp: >250 °C. IR
(cm−1): 3200 (NH), 1618 (CO), 1236 (C−F), 1208 (CCO),
ABBREVIATIONS USED
■
ATP, adenosine 5′-triphosphate; DMF, N,N-dimethylformamide;
DMSO, dimethyl sulfoxide; EtOAc, ethyl acetate; EtOH, ethanol;
FC, flash chromatography; MeOH, methanol; rt, room
temperature
REFERENCES
■
(1) Barton, J.; Blackledge, G.; Wakeling, A. Growth factor and their
receptors: new targets for prostate cancer therapy. Urology 2001, 58,
114−122.
1
1137 (C−O−C methoxy), 1048 (C−Br), 1031 (C−Cl). H NMR
(DMSO-d6): δ (ppm) 9.05 (s, 1H, NH), 9.01 (s, 1H, NH), 8.54 (s,
1H, ArH), 7.58 (d, 2H, ArH), 7.57 (s, 1H, ArH), 7.40 (m, 3H, ArH),
7.16−7.35 (m, 4H, ArH), 3.99 (s, 3H, OCH3), 3.98 (s, 3H, OCH3).
LC−MS (APCI+): m/z calcd for C23H18BrClN4O4 529 [(M + H)+ for
79Br/35Cl], 531 [(M + H)+ for 79Br/37Cl], 531 [(M + H)+ for
81Br/35Cl], and 533 [(M + H)+ for 81Br/37Cl].
Cell Culture and Cell Proliferation Assay. Human prostate
cancer cells (PC3) and breast cancer cell line MCF7 were grown at
37 °C in a humidified atmosphere containing 5% CO2 in RPMI-1640
medium (Sigma), minimum essential medium (MEM; Sigma), and
Dulbecco's modified Eagle's medium (DMEM; Gibco) supplemented
with 10% fetal bovine serum, glutamine (2 mM), penicillin (100 IU/mL),
and streptomycin (100 μg/mL). Colon cancer cells (HT29) were grown
at 37 °C in a humidified atmosphere containing 5% CO2 in DMEM +
Glutamax-I (Gibco) supplemented with 10% fetal bovine serum, penicillin
(100 IU/mL), and streptomycin (100 μg/mL).
In the cell proliferation assay, cells were plated in triplicate on 96-
well plates (3000 cells/well) and incubated for 24 h. The cells were
then incubated in culture medium that contained various concen-
trations of tested compounds, each dissolved in less than 0.1% DMSO.
After 72 h, cell growth was estimated by the colorimetric MTS test.
In Vitro Kinase Assays. Kinase assays were performed in 96-well
plates (Multiscreen Durapore, Millipore) using [γ-32P]ATP (Amersham
Biosciences) and the synthetic polymer poly(Glu4/Tyr) (Sigma
Chemicals) as a phosphoacceptor substrate. Tested compounds
were dissolved in DMSO. The final concentration of DMSO in the
assay solutions was 0.1%, which was shown to have no effect on the
kinase activity.
EGFR Tyrosine Kinase Activity. A 20 ng sample of EGFR (purified
from human carcinoma A431 cells, Sigma Chemicals) was incubated
for 1 h at 28 °C using various concentrations of tested compounds in
kinase buffer (HEPES, 50 mM, pH 7.5, BSA, 0.1 mg/mL, MnCl2, 10 mM,
MgCl2, 5 mM, Na3VO4, 100 μM, DTT, 0.5 mM, poly(Glu4/Tyr),
250 μg/mL, ATP, 5 μM, [γ-32P]ATP, 0.5 μCi).
(2) Cohen, P. The role of protein phosphorylation in human health
and disease. Eur. J. Biochem. 2001, 268, 5001−5010.
(3) Levitzki, A. Protein kinase inhibitors as a therapeutic modality.
Acc. Chem. Res. 2003, 36, 462−469.
(4) Cohen, P. Protein kinasesthe major drug targets of the twenty-
first century? Nat. Rev. Drug Discovery 2002, 1, 309−315.
(5) Zwick, E.; Bange, J.; Ullrich, A. Receptor tyrosine kinase as
targets for anticancer drugs. Trends Mol. Med. 2002, 8, 17−22.
(6) Traxler, P.; Bold, G.; Buchdunger, E.; Caravatti, G.; Furet, P.;
Manley, P.; O’Reilly, T.; Wood, J.; Zimmermann, J. Tyrosine kinase
inhibitors: from rational design to clinical trials. Med. Res. Rev. 2001,
21, 499−512.
(7) Traxler, P. Tyrosine kinases as targets in cancer therapy
successes and failures. Exp. Opin. Ther. Targets 2001, 21, 499−512.
(8) Carmeliet, P. Angiogenesis in health and disease. Nat. Med. 2003,
9, 653−660.
(9) Otrock, Z. K.; Makarem, J. A.; Shamseddine, A. I. Vascular
endothelial growth factor family of ligands and receptors: reviews.
Blood Cells, Mol. Dis. 2007, 38, 258−268.
(10) Bishop-Bailey, D. Tumour vascularisation: a druggable target.
Curr. Opin. Pharmacol. 2009, 9, 96−101.
(11) Hida, K.; Hida, Y.; Shindoh, M. Understanding tumor
endothelial cell abnormalities to develop ideal anti-angiogenic
therapies. Cancer Sci. 2008, 99, 459−466.
(12) Otrock, Z. K.; Mahfouz, R. A. R.; Makarem, J. A.; Shamseddine,
A. I. Understanding the biology of angiogenesis: review of the most
important molecular mechanisms. Blood Cells, Mol. Dis. 2007, 39,
212−220.
(13) Senger, D. R.; Van de Water, L.; Brown, L. F.; Nagy, J. A.; Yeo,
K. T.; Yeo, T. K.; Berse, B.; Jackman, R. W.; Dvorak, A. M.; Dvorak, H.
F. Vascular permeability factor (VPF, VEGF) in tumor biology. Cancer
Metastasis Rev. 1993, 12, 303−324.
(14) Roy, H.; Bhardwaj, S.; Yla-Herttuala, S. Biology of vascular
̈
endothelial growth factors. FEBS Lett. 2006, 580, 2879−2887.
(15) Huang, Z.; Bao, S. D. Roles of main pro- and anti-angiogenic
factors in tumor angiogenesis. World J. Gastroenterol. 2004, 4, 463−470.
(16) Gleadle, J. M.; Ratcliffe, P. J. Hypoxia and the regulation of gene
expression. Mol. Med. Today 1998, 4, 122−129.
VEGFR−2 Tyrosine kinase Activity. A 10 ng sample of VEGFR-2
(recombinant human protein, Invitrogen) was incubated for 1 h at
28 °C using various concentrations of tested compounds in kinase
buffer (Tris, 50 mM, pH 7.5, BSA, 25 μg/mL, MnCl2, 1.5 mM, MgCl2,
10 mM, DTT, 2.5 mM, Na3VO4, 100 μM, β-glycerophosphate, 5 mM,
poly(Glu4/Tyr), 250 μg/mL, ATP, 5 μM, [γ-32P]ATP, 0.5 μCi).
The reaction was stopped by adding 20 μL of trichloroacetic acid,
100%. The wells were screened and washed 10 times with trichloro-
acetic acid, 10%. The plates were counted in a Top Count for 1 min/well.
(17) Ferrara, N.; Davis-Smyth, T. The biology of vascular endothelial
growth factor. Endocr. Rev. 1997, 18, 4−25.
(18) Roskoski, R. Jr. Vascular endothelial growth factor (VEGF) signal-
ing in tumor progression. Crit. Rev. Oncol. Hematol. 2007, 62, 179−213.
(19) Roskoski, R Jr. VEGF receptor protein-tyrosine kinases: structure
and regulation. Biochem. Biophys. Res. Commun. 2008, 375, 287−291.
(20) Shibuya, M.; Claesson-Welsh, L. Signal transduction by VEGF
receptors in regulation of angiogenesis and lymphangiogenesis. Exp.
Cell Res. 2006, 312, 549−560.
AUTHOR INFORMATION
■
Corresponding Author
(21) Rahimi, N. Vascular endothelial growth factor receptors:
molecular mechanisms of activation and therapeutic potentials. Exp.
Eye Res. 2006, 83, 1005−1016.
*Phone: +33 (0)3 20 96 40 40. Fax: +33 (0)3 20 96 49 06.
E-mail: amaury.farce-2@univ-lille2.fr.
1203
dx.doi.org/10.1021/jm2013453 | J. Med. Chem. 2012, 55, 1189−1204

Journal of Medicinal Chemistry
Article
(22) Schlessinger, J. Cell signaling by receptor tyrosine kinases. Cell
2000, 103, 211−225.
(23) Hunter, T. Tyrosine phosphorylation: thirty years and counting.
Curr. Opin. Cell Biol. 2009, 21, 140−146.
N. A.; Scarborough, R. M. Identification of orally active, potent, and
selective 4-piperazinylquinazolines as antagonists of the platelet-
derived growth factor receptor tyrosine kinase family. J. Med. Chem.
2002, 45, 3772−3793.
(41) Garofalo, A.; Goossens, L.; Lemoine, A.; Ravez, P.; Six, P.;
Howsam, M.; Farce, A.; Depreux, P. (4-(6,7-Disubstituted quinazolin-
4-ylamino)phenyl) carbamic acid ester: a novel series of dual EGFR/
VEGFR-2 receptor tyrosine kinase inhibitors. MedChemComm 2011,
2, 65−72.
(42) Garofalo, A.; Goossens, L.; Farce, A.; Lemoine, A.; Ravez, P.;
Six, P.; Depreux, P. Impact of aryloxy-linked quinazolines: a novel
series of selective VEGFR-2 receptor tyrosine kinase inhibitors. Bioorg.
Med. Chem. Lett. 2011, 21, 2106−2112.
(24) Gibson, K. H.; Grundy, W.; Godfrey, A.; Woodburn, J. R.;
Ashton, S. E.; Curry, B. J.; Scarlett, L.; Barker, A. J.; Brown, D. S.
Epidermal growth factor receptor tyrosine kinase: structure-activity
relationships and antitumour activity of novel quinazolines. Bioorg.
Med. Chem. Lett. 1997, 7, 2723−2728.
(25) Kohno, M.; Pouyssegur, J. Targeting the Erk signalling pathway
in cancer therapy. Ann. Med. 2006, 38, 200−211.
(26) Djordjevic, S.; Driscoll, P. C. Structural insight into substrate
specificity and regulatory mechanisms of phosphoinositide 3-kinases.
Trends Biochem. Sci. 2002, 27, 426−432.
(27) Gourley, M.; Williamson, J. S. Angiogenesis: new targets for the
development of anticancer chemotherapies. Curr. Pharm. Des. 2000, 6,
417−439.
(28) Brunella, C.; Costanza, E. Possible novel targets for therapeutic
angiogenesis. Curr. Opin. Pharmacol. 2009, 9, 102−108.
(29) Raffetto, J. D.; Khalil, R. A. Matrix metalloproteinases and their
inhibitors in vascular remodeling and vascular disease. Biochem.
Pharmacol. 2008, 75, 346−359.
(30) Hida, K.; Hida, Y.; Shindoh, M. Understanding tumor endothelial
cell abnormalities to develop ideal anti-angiogenic therapies. Cancer Sci.
2008, 99, 459−466.
(31) Masood, R.; Kundra, A.; Zhu, S.; Xia, G.; Scalia, P.; Smith, D. L.;
Gill, P. S. Malignant mesothelioma growth inhibition by agents that
target the VEGF and VEGF-C autocrine loops. Int. J. Cancer 2003,
104, 603−610.
(32) Booy, E. P.; Johar, D.; Maddika, S.; Pirzada, H.; Sahib, M. M.;
Gehrke, I.; Loewen, S.; Louis, S. F.; Kadkhoda, K.; Mowat, M.; Los, M.
Monoclonal and bispecific antibodies as novel therapeutics. Arch.
Immunol. Ther. Exp. 2006, 54, 85−101.
(43) Furuta, T.; Sakai, T.; Senga, T.; Osawa, T.; Kubo, K.; Shimizu,
T.; Suzuki, R.; Yoshino, T.; Endo, M.; Miwa, A. Identification of
potent and selective inhibitors of PDGF Receptor autophosphor-
ylation. J. Med. Chem. 2006, 49, 2186−2192.
́
(44) Desroses, M.; Laconde, G.; Depreux, P.; Henichart, J. P.
Synthesis of unsymmetrical dialkoxy quinazolines. Org. Prep. Proced.
Int. 2004, 36, 445−452.
(45) Telliez, A.; Desroses, M.; Pommery, N.; Briand, O.; Farce, A.;
́
Laconde, G.; Lemoine, A.; Depreux, P.; Henichart, J. P. Derivatives of
Iressa, a specific epidermal growth factor receptor inhibitor, are
powerful apoptosis inducers in PC3 prostatic cancer cells.
ChemMedChem 2007, 2, 318−332.
(46) Garofalo, A.; Goossens, L.; Lebegue, N.; Depreux, P. A novel
and efficient one-pot synthesis of (aminophenyl)carbamic acid ester.
Synth. Commun. 2011, 41, 2007−2016.
(47) Elliott, R. D.; Thomas, H. J.; Shaddix, S. C.; Adamson, D. J.;
Brockman, R. W.; Riordan, J. M.; Montgomery, J. A. Nitrosoureido
nucleosides as potential inhibitors of nucleotide biosynthesis. J. Med.
Chem. 1988, 31, 250−254.
(48) Garofalo, A.; Goossens, L.; Baldyrou, B.; Lemoine, A.; Ravez, S.;
Six, P.; David-Cordonnier, M. H.; Bonte, J. P.; Depreux, P.; Lansiaux, A.;
Goossens, J. F. Design, synthesis and DNA-binding of N-alkyl(anilino)-
quinazoline derivatives. J. Med. Chem. 2010, 53, 8089−8103.
(49) SYBYL 6.9.1; Tripos Inc.: St. Louis, MO, 2003.
(33) Ferrara, N.; Hillan, K. J.; Novotny, W. Bevacizumab (Avastin), a
humanized anti-VEGF monoclonal antibody for cancer therapy.
Biochem. Biophys. Res. Commun. 2005, 333, 328−335.
(34) Scanga, A.; Kowdley, K. Sorafenib: a glimmer of hope for un-
resectable hepatocellular carcinoma? Hepathology 2009, 49, 332−334.
(35) Escudier, B.; Eisen, T.; Stadler, W. M.; Szczylik, C.; Oudard, S.;
Siebels, M.; Negrier, S.; Chevreau, C.; Solska, E.; Desai, A. A.; Rolland,
F.; Demkow, T.; Hutson, T. E.; Gore, M.; Freeman, S.; Schwartz, B.;
Shan, M.; Simantov, R.; Bukowski, R. M. Sorafenib in advanced clear-
cell renal-cell carcinoma. N. Eng. J. Med. 2009, 49, 332−334.
(36) Motzer, R. J.; Michaelson, M. D.; Redman, B. G.; Hudes, G. R.;
Wilding, G.; Figlin, R. A.; Ginsberg, M. S.; Kim, S. T.; Baum, C. M.;
DePrimo, S. E.; Li, J. Z.; Bello, C. L.; Theuer, C. P.; George, D. J.; Rini,
B. I. Activity of SU11248, a multitargeted inhibitor of vascular
endothelial growth factor receptor and platelet-derived growth factor
receptor, in patients with metastatic renal cell carcinoma. J. Clin. Oncol.
2006, 24, 16−24.
(37) Ryan, A. J.; Wedge, S. R. ZD6474a novel inhibitor of VEGFR
and EGFR tyrosine kinase. Br. J. Cancer 2005, 92, S6−S13.
(38) Ciardiello, F.; Bianco, R.; Caputo, R.; Caputo, R.; Damiano, V.;
Troiani, T.; Melisi, D.; De Vita, F.; De Placido, S.; Bianco, A. R.;
Tortora, G. Antitumor activity of ZD6474, a vascular endothelial
growth factor receptor tyrosine kinase inhibitor, in human cancer cells
with acquired resistance to antiepidermal growth factor receptor
therapy. Clin. Cancer Res. 2004, 10, 784−793.
(39) Morabito, A.; Piccirillo, M. C.; Falasconi, F.; De Feo, G.; Del
Giudice, A.; Bryce, J.; Di Maio, M.; De Maio, E.; Normanno, N.;
Perrone, F. Vandetanib (ZD6474), a dual inhibitor of vascular
endothelial growth factor receptor (VEGFR) and epidermal growth
factor receptor (EGFR) tyrosine kinases: current status and future
directions. Oncologist 2009, 14, 378−390.
(40) Garofalo, A.; Goossens, L.; Lemoine, A.; Farce, A.; Arlot, Y.;
Depreux, P. Quinazoline-urea, new protein kinase inhibitors in
treatment of prostate cancer. J. Enzyme Inhib. Med. Chem. 2010, 25,
158−171. Pandey, A.; Volkots, D. L.; Seroogy, J. M.; Rose, J. W.; Yu, J.
C.; Lambing, J. L.; Hutchaleelaha, A.; Hollenbach, S. J.; Abe, K.; Giese,
(50) Clark, M.; Cramer, R. D. III; van Opdenbosch, N. Validation of
the general purpose Tripos 5.2 force field. J. Comput. Chem. 1989, 10,
982−1012.
(51) Berman, H. M.; Westbrook, J.; Feng, Z.; Gary, G.; Bhat, T. N.;
Weissig, H.; Shindyalov, I. N.; Bourne, P. E. The Protein Data Bank.
Nucleic Acids Res. 2000, 28, 235−242.
(52) Miyazaki, Y.; Matsunaga, S.; Tang, J.; Maeda, Y.; Nakano, M.;
Philippe, R. J.; Shibahara, M.; Liu, W.; Sato, H.; Wang, L.; Nolte, R. T.
Novel 4-amino-furo[2,3-d]pyrimidines as Tie-2 and VEGFR2 dual
inhibitors. Bioorg. Med. Chem. Lett. 2005, 15, 2203−2207.
(53) Jones, G.; Willett, P.; Glen, R. C.; Leach, A. R.; Taylor, R.
Development and validation of a genetic algorithm for flexible docking.
J. Mol. Biol. 1997, 267, 727−748.
(54) Wang, R.; Lai, L.; Wang, S. J. Further development and
validation of empirical scoring functions for structure-based binding
affinity prediction. J. Comput.-Aided Mol. Des. 2002, 16, 11−26.
(55) Oprea, T. I.; Waller, C. L.; Marshall, G. R. Three-dimensional
quantitative structure-activity relationship of human immunodeficiency
virus (I) protease inhibitors. 2. Predictive power using limited exploration
of alternate binding modes. J. Med. Chem. 1994, 37, 2206−2215.
(56) Cramer, R. D. III; Patterson, D. E.; Bunce, J. D. Comparative
molecular field analysis (CoMFA). 1. Effect of shape on binding of
steroids to carrier proteins. J. Am. Chem. Soc. 1988, 110, 5959−5967.
(57) Klebe, G.; Abraham, U.; Mietzner, T. Molecular similarity
indices in a comparative analysis (CoMSIA) of drug molecules to
correlate and predict their biological activity. J. Med. Chem. 1994, 37,
4130−4146.
(58) Clark, M.; Cramer, R. D. III. The probability of chance
correlation using partial least squares (PLS). Quant. Struct.-Act. Relat.
1993, 12, 137−145.
1204
dx.doi.org/10.1021/jm2013453 | J. Med. Chem. 2012, 55, 1189−1204