The design and discovery of water soluble 4-substituted-2,6-dimethylfuro[2, 3-d]pyrimidines as multitargeted receptor tyrosine kinase inhibitors and microtubule targeting antitumor agents

Article abstract of DOI:10.1016/j.bmc.2014.04.049

The design, synthesis and biological evaluations of fourteen 4-substituted 2,6-dimethylfuro[2,3-d]pyrimidines are reported. Four compounds (11-13, 15) inhibit vascular endothelial growth factor receptor-2 (VEGFR-2), platelet-derived growth factor receptor

Full text of DOI:10.1016/j.bmc.2014.04.049

Accepted Manuscript
The Design and Discovery of Water Soluble 4-Substituted-2,6-dimethyl-
furo[2,3-d]pyrimidines as Multitargeted Receptor Tyrosine Kinase Inhibitors
and Microtubule Targeting Antitumor Agents
Xin Zhang, Sudhir Raghavan, Michael Ihnat, Jessica E. Thorpe, Bryan C. Disch,
Anja Bastian, Lora C. Bailey-Downs, Nicholas F. Dybdal-Hargreaves, Cristina
C. Rohena, Ernest Hamel, Susan L. Mooberry, Aleem Gangjee
PII:
S0968-0896(14)00314-9
DOI:
http://dx.doi.org/10.1016/j.bmc.2014.04.049
BMC 11544
Reference:
To appear in:
Bioorganic & Medicinal Chemistry
Received Date:
Revised Date:
Accepted Date:
28 February 2014
16 April 2014
25 April 2014
Please cite this article as: Zhang, X., Raghavan, S., Ihnat, M., Thorpe, J.E., Disch, B.C., Bastian, A., Bailey-Downs,
L.C., Dybdal-Hargreaves, N.F., Rohena, C.C., Hamel, E., Mooberry, S.L., Gangjee, A., The Design and Discovery
of Water Soluble 4-Substituted-2,6-dimethylfuro[2,3-d]pyrimidines as Multitargeted Receptor Tyrosine Kinase
Inhibitors and Microtubule Targeting Antitumor Agents, Bioorganic & Medicinal Chemistry (2014), doi: http://
dx.doi.org/10.1016/j.bmc.2014.04.049
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

The Design and Discovery of Water Soluble 4-Substituted-2,6-dimethylfuro[2,3-
d]pyrimidines as Multitargeted Receptor Tyrosine Kinase Inhibitors and Microtubule
Targeting Antitumor Agents
†
†
‡
‡
‡
‡
Xin Zhang , Sudhir Raghavan , Michael Ihnat , Jessica E. Thorpe , Bryan C. Disch , Anja Bastian , Lora C.
‡
§
§
£
§,ǁ
Bailey-Downs , Nicholas F. Dybdal-Hargreaves , Cristina C. Rohena , Ernest Hamel , Susan L. Mooberry ,
,
†,ǁ
Aleem Gangjee*
†
Division of Medicinal Chemistry, Graduate School of Pharmaceutical Sciences, Duquesne University, 600
Forbes Avenue, Pittsburgh, PA 15282.
College of Pharmacy, University of Oklahoma Health Science Center, 1110 North Stonewall, Oklahoma City,
OK 73117.
Department of Pharmacology, Cancer Therapy & Research Center, University of Texas Health Science
‡
§
Center at San Antonio, 7703 Floyd Curl Drive, San Antonio, TX 78229.
£
Screening Technologies Branch, Developmental Therapeutics Program, Division of Cancer Treatment and
Diagnosis, Frederick National Laboratory for Cancer Research, National Institutes of Health, 1050 Boyles
Street, Frederick, MD 21702.
ǁ These authors contributed equally to this work
*
To whom correspondence should be addressed. Phone: 412-396-6070. Fax: 412-396-5593.
E-mail: gangjee@duq.edu.
Abbreviations: combretastatin A-4 (CA-4); vascular endothelial growth factor receptor-2
VEGFR-2); Receptor Tyrosine Kinases (RTKs); 2-methoxyestradiol (2ME2); platelet-derived
growth factor receptor β (PDGFR-β); chorioallantoic membrane (CAM)
(
1

Abstract
The design, synthesis and biological evaluations of fourteen 4-substituted 2,6-dimethylfuro[2,3-
d]pyrimidines are reported. Four compounds (11-13, 15) inhibit vascular endothelial growth
factor receptor-2 (VEGFR-2), platelet-derived growth factor receptor β (PDGFR-β), and target
tubulin leading to cytotoxicity. Compound 11 has nanomolar potency, comparable to sunitinib
and semaxinib, against tumor cell lines overexpressing VEGFR-2 and PDGFR-β. Further, 11
binds at the colchicine site on tubulin, depolymerizes cellular microtubules and inhibits purified
tubulin assembly and overcomes both βIII-tubulin and P-glycoprotein-mediated drug resistance,
and initiates mitotic arrest leading to apoptosis. In vivo, its HCl salt, 21, reduced tumor size and
vascularity in xenograft and allograft murine models and was superior to docetaxel and sunitinib,
without overt toxicity. Thus 21 affords potential combination chemotherapy in a single agent.
Introduction
2

Angiogenesis, the formation of new blood vessels, is essential for tumorigenesis and
1
,2
plays an essential role in tumor growth, invasion and metastasis. Solid tumors depend on the
newly formed vascular network to provide nutrients and to remove metabolic waste in order to
3
grow beyond a few millimeters in diameter.
The activation of Receptor Tyrosine Kinases (RTKs) regulates the transduction of signals
from the extracellular domain of endothelial cells to the nucleus and represents the most
4
important factor that triggers angiogenesis.
Figure 1. Structures of microtubule targeting agents.
Tubulin binding agents (Figure 1) belong to an important class of antitumor agents and
5
-7
are widely used in the clinic for cancer chemotherapy. Most microtubule targeting agents can
be divided into 3 classes based on their interactions within the taxane, vinca, or colchicine site on
6
,7
tubulin.
Drugs that bind within the taxoid site include paclitaxel (Taxol), docetaxel (Taxotere),
and the epothilones. Paclitaxel and other taxoids (and the epothilones) bind to β-tubulin in the
3

8
, 9
interior of the microtubule.
Unlike the other two classes of antimicrotubule agents, the taxoids
6
,10
stimulate tubulin polymerization and are designated as microtubule stabilizers. They are
7
, 10
useful in the treatment of breast, lung, ovarian, head and neck and prostate cancers.
The
second class of microtubule disrupting agents are the vinca alkaloids and these include
vincristine, vinblastine, vindesine and vinorelbine. The vinca alkaloids bind between two αβ-
6
tubulin diemers at a site that is distinct from the taxane site. The vinca alkaloids are important
7
in the treatment of leukemias, lymphomas, non-small cell lung cancer and childhood cancers. A
diverse collection of small molecules, including colchicine and the combretastatins, bind to the
11, 12
colchicine site on β-tubulin at its interface with α-tubulin, a site distinct from the vinca site.
Similar to the vinca alkaloids, colchicine site agents inhibit tubulin polymerization. Colchicine
itself is not used as an antitumor agent but is useful in the treatment of gout and familial
1
3
Mediterranean fever. Although there are no clinically approved anticancer agents that bind
within the colchicine site, several agents including 2-methoxyestradiol, combretastatin A-4 (CA-
4
) phosphorylated prodrug combretastatin A-4 phosphate (CA-4P) (fosbretabulin), the
combretastain CA-1P prodrug (OXi4503), BNC105P, ABT-751 and plinabulin (NPT-2358) have
11, 14, 15
been evaluated in clinical trials.
While CA-4P, CA-1P and others continue to be evaluated
to date no colchicine site agent has advanced to Phase III studies or FDA
approval for anticancer indications, demonstrating the need of developing additional colchicine
11, 16
in clinical trials,
11
site agents for potential clinical evaluations.
Tumor angiogenic mechanisms that are vital for tumor growth and metastasis are
1
7
inhibited by antiangiogenic agents. Antiangiogenic agents that target RTKs, especially the
1
8
VEGF receptor, have found utility in the treatment of multiple types of cancer, but these agents
1
9, 20
are usually not cytotoxic and are mainly cytostatic.
Combination chemotherapy with
4

antiangiogenic and cytotoxic agents has shown significant promise, and several studies with such
1
9, 20
combinations are in progress in the clinic.
Single agents with both antiangiogenic activities
and cytotoxicity could circumvent the pharmacokinetic problems of multiple agents, avoid drug-
drug interactions, be used at lower doses to alleviate toxicity, be devoid of overlapping toxicities
and delay or prevent tumor cell resistance. In this study, we report the design and discovery of
novel bicyclic furo[2,3-d]pyrimidines, some of which possess both RTK and tubulin inhibitory
properties along with potent in vivo antitumor activities.
Rationale
The inhibition of RTKs disrupts angiogenesis and provides an approach for the treatment
1
8, 21
of cancer.
The ATP-binding site of RTKs is an attractive target for small molecule drug
design. Structure determination of ATP-RTK complexes have revealed the regions within or
2
1, 22
close to the binding cleft that are not fully occupied by ATP.
These unoccupied regions show
structural diversity between members of the kinase family. The commonality as well as diversity
among the ATP-binding sites of kinases has allowed the development of general pharmacophore
2
2-25
22
models for rational drug design.
The general model proposed consists of an Adenine
region which is a hydrophobic binding site for the adenine ring of ATP (Figure 2a) as well as for
the heterocyclic scaffold of RTK inhibitors such as quinazolines and pyrimidines. The N1- and
N6-amino nitrogen of the adenine ring of ATP are hydrogen bonded to two amino acid residues
of the Hinge region. The Sugar binding pocket in the ATP binding site accommodates the sugar
moiety of ATP, and the Phosphate binding region binds the triphosphate moiety of ATP. In
addition, a Hydrophobic site I extends in the direction of the lone pair of the N7 of ATP and a
Hydrophobic site II lies below the Adenine region. Both hydrophobic regions are unoccupied by
5

2
2
ATP in the binding site.
Figure 2a. Binding of ATP with RTKs. Representative amino acids from the four binding
pockets (Hinge Region, Hydrophobic Pocket I, Hydrophobic Pocket II and the Phosphate Site) in
the ATP site from VEGFR-2 and PDGFR-β and potential interactions with ATP are shown. The
2
6
potential interaction map for ATP was generated using MOE 2010.10 by superimposition of the
2
7
28
docked poses of ATP in VEGFR-2 (PDB: 1YWN ) and a homology model of PDGFR-β.
6

Figure 2b. Proposed horizontal binding mode 1 of 11 with RTKs similar to ATP binding mode.
Figure 2c. Proposed alternate vertical binding mode of 11 with RTKs.
7

Figure 3. Design of antitubulin and RTK inhibitory activities in single agents.
2
9, 30
We had reported previously
that the furo[2,3-d]pyrimidine scaffold containing
compound 1 (Figure 3) had afforded both VEGFR-2 and PDGFR-β inhibitory activity. In
3
1-33
addition, we
had also reported that 6-methyl cyclopenta fused pyrimidines 2 demonstrated
potent tubulin depolymerization activity along with potent in vitro and in vivo antitumor activity.
1
9, 20, 34, 35
The remarkable recent success and clinical trials
of RTK inhibitors in general and
particularly of VEGFR-2 and/or PDGFR-β inhibitors in combination with antitubulin agents
such as paclitaxel prompted our design and development of single agents with both VEGFR2
and/or PDGFR-β inhibition and antitubulin activity.
As an initial study, we elected to structurally engineer tubulin inhibitory activity into our
existing RTK inhibitor 1 (Figure 3). Transposing the 5-substitution of the RTK inhibitor 1 on to
the 4-NH moiety of 1 and placing a 6-CH
3
group on 1 afforded a hybrid furo[2,3-d]pyrimidine
scaffold (Figure 3) capable of antitubulin activity structurally akin to 2 (Figure 3) with the ability
8

to access the hydrophobic region 1 on the general pharmacophore for RTKs in Figure 2.
Figure 4. The structures of furo[2,3-d]pyrimidine RTK inhibitors 3-15.
Thus the hybrid molecules 3-15 with the furo[2,3-d]pyrimidine scaffold (Figure 4) were
designed with potentially both attributes of RTK (VEGFR-2 and/or PDGFR-β) inhibition and
antitubulin activity in single molecules. The 2-substitution of 3-15 was designed as a methyl
group, which is not present in most other 6-6 or 6-5 pyrimidine fused ring system RTK inhibitors
3
6
37
38
39
40
41
42
reported such as gefitinib, erlotinib, vandetanib, lapatinib, afatinib, canertinib, and 6-
[
4-[(4-ethylpiperazin-1-yl)methyl]phenyl]-N-[(1R)-1-phenylethyl]-7H-pyrrolo[2,3-d]pyrimidin-
4
3
43
4
-amine (AEE788) and analogs of linifanib (Figure 5).
9

Figure 5: Structures of kinase inhibitors that lack a 2-methyl group.
3
3
We have shown that for similarly fused cyclopenta[d]pyrimidines, the 2-methyl group
affords the most potent activity against tubulin and tumor cells. On the basis of molecular
modeling (see below), there is sufficient space in the ATP binding site of RTKs to accommodate
a 2-methyl moiety that provides additional hydrophobic binding with the ATP site, compared
with 2-desmethyl analogs. Compounds 3-15 also have a 6-methyl substitution, which has a
small size and can be accommodated in the phosphate or sugar binding site (Figure 2) of multiple
RTKs. Thus compounds 3-15 were designed to fit into the ATP binding site of RTKs exemplified
by 11 in Figures 2b and 2c.
3
7
Similar to several known ATP competitive RTK inhibitors (Figure 5), gefitinib,
3
8
39
40
44
erlotinib, vandetanib, lapatinib, among others, 3-10 (Figure 4) have a 4-anilino
1
0

substitution. Different ring anilino substitutions at the 4-position of the pyrimidine A-ring were
expected to be involved in RTK binding at the Hydrophobic Region I and to influence inhibition
4
5
selectivity as well as antitumor activity. Palmer et al. reported that small lipophilic electron-
withdrawing groups at the 3-position of the anilino moiety in tricyclic ring systems are beneficial.
4
6
Bold et al. also reported that small lipophilic electron-withdrawing groups at the 4-position of
the anilino moiety are beneficial for VEGFR-2 and PDGFR-β inhibition. Thus, anilines with
electron-withdrawing groups at either the 3- or 4-position were incorporated into compounds 6-8.
For comparison, compounds 3 and 10 with a phenyl group and a 4’-methoxy phenyl substitution,
respectively, were also included as 4-substitution on the pyrimidine A ring. Compound 4 with a
naphthalene substitution was designed to determine the bulk tolerance in Hydrophobic region I
3
8
among different RTKs. Similar to erlotinib, a 3-acetylenyl substitution was included in
compound 5. The indole ring rather than a substituted phenyl ring has been reported to provide
2
5a
potent and selective RTK inhibition.
Thus in compound 9, an indole ring was incorporated as
the 4-substitution on the pyrimidine A-ring via a nitrogen linker.
Compared to 3-10, in compounds 11-15, an additional alkyl group with increasing bulk
was designed on the aniline-nitrogen. Based on the general binding mode and molecular
modeling (see below), the ATP site has sufficient space to accommodate a methyl group in 11,
while the larger alkyl groups in 12-15 may not be tolerated due to steric hindrance with the hinge
region in the ATP site. However, these larger alkyl groups could be accommodated in an
alternate vertical binding mode indicated from molecular modeling (see below). The methyl
group on the aniline nitrogen in 11 was introduced to restrict free rotation of the 4-position C-N
bond as well as the 1’-position C-N bond (Figure 2a) and thus restrict the conformation of the
1
1

anilino ring. A methyl group on the aniline-nitrogen was therefore expected to force the phenyl
ring into the Hydrophobic site I (Figures 2b and 2c) and increase RTK inhibition. We were aware
that the replacement of the NH-hydrogen with a methyl group would preclude its hydrogen bond
donor ability in the hinge region and might result in a loss of RTK inhibitory activity. In addition
3
1-33
to conformational restriction, we
recently reported that methylation of the anilino nitrogen in
1
9,
similar 6-5 ring systems afforded potent antitubulin activity. In light of the extensive literature
2
0a,20b, 34, 35
on clinical combination chemotherapy protocols that combine RTK inhibitors
(
bevacizumab, apatinib, gefitinib, sorafenib, erlotanib, lapatinib, pazopanib, inatinib, crizotinib,
sunitinib) with antitubulin agents (paclitaxel, docetaxel, vinorelbine, ixabepilone and others), it
was of paramount importance to determine if N-alkylation of 10, particularly the N-methyl
analog 11, would possess both anti-RTK and antitubulin activity thus affording dual activity and
the potential of combination chemotherapy in a single agent.
Molecular Modeling and Computational Studies
Docking studies were performed for proposed compounds 3 – 15 in the colchicine
4
7
27
binding site of tubulin (PDB: 1SA0 , 3.58Å) and in VEGFR-2 (PDB: 1YWN , 1.71Å) and in a
2
8
26
homology model of PDGFR-β using Molecular Operating Environment (MOE 2010.10)
2
8, 31
(
validated by re-docking the crystal structure ligands) using previously reported methods.
Docking studies of the lead and standard compounds (semaxanib for VEGFR-2 and sunitinib for
28
PDGFR-β) were performed using methods previously reported.
1
2

Docking of compound 11 in the colchicine site of tubulin:
MeO
4
'
N
5
4
N3
6
2
O
7
N
1
11
Figure 6: Superimposition of the docked poses of 11 (pink) and DAMA colchicine (purple) in
the colchicine site of tubulin at the interface of the α-subunit (magenta) and β-subunit (blue) of
4
7
tubulin; Pocket amino acids are depicted in green as sticks. PDB: 1SA0.
Multiple low-energy conformations (within 1 kcal/mol of the best pose) were obtained on
docking 3 – 15 in the colchicine site of tubulin. The docked conformation of 11 (Figure 6, score:
-
5.95 kcal/mol) was selected as a working model for compounds 3 – 15 based on their similarity
to the crystal structure of the bound conformation of DAMA-colchicine (Figure 6, purple) in
tubulin. In addition we elected to use 11 to depict the molecular modeling since it turned out to
be the most potent dual acting agent of the target compounds (see below). The 4’-OMe phenyl
group of 11 overlaps with the triOMe containing A-ring of DAMA-colchicine and interacts with
Leuβ246, Alaβ248, Leuβ253, Alaβ314, Ileβ376 and Valβ316 (Figure 6). The 4’-OMe of 11
1
3

overlaps with the 3’-OMe group in the A ring of DAMA colchicine. The 4’-OMe of 11 forms a
hydrogen bond with Cysβ241 as is observed with the 3’-OMe group of DAMA-colchicine. The
N-Me group of 11 occupies a region in space in proximity to the C5 and C6-positions in the B-
ring of DAMA colchicine and is involved in hydrophobic interactions with Lysβ252, Alaβ248
and Leuβ246. The furo[2,3-d]pyrimidine scaffold of 11 partly overlaps with the C-ring of
DAMA-colchicine and forms hydrophobic interactions with Leuβ253, Asnβ256 and Lysβ250.
These results provide molecular modeling support for the potential tubulin inhibitory activity for
proposed compound 11 and its analogs (3–10 and 12–15) and are in keeping with results from
4
8
recently published molecular modeling studies at the colchicine site of tubulin.
The key binding interactions seen with 11 were also retained for the best docked poses of
analogs 3 – 10 in tubulin. In the docking studies of 12–15 in the colchicine site, homologation at
the aniline nitrogen was expected to increase hydrophobic interactions with the side chain atoms
of Leuβ248 and Metβ259 in the binding pocket and thus maintain potency against tubulin like 11.
1
4

Docking of 11 in the ATP binding site of VEGFR-2 and PDGFR-β:
Figure 7a: Docked pose of 11 (pink) in the vertical binding mode in the ATP binding pocket of
2
7
VEGFR-2 (PDB: 1YWN ). Pocket amino acids are depicted in green as sticks. Amino acids in
the pocket that form strongest interactions with 11 are depicted in green as cylinders.
Figure 7a shows the docked pose of 11 (selected as a working example representative of
2
7
3
– 15) in a vertical binding mode in the ATP binding site of a crystal structure of VEGFR-2.
The furo[2,3-d]pyrimidine scaffold of 11 is placed in the adenine binding pocket of the ATP
binding site of VEGFR-2. The N1 nitrogen of 11 forms a hydrogen bond with the backbone NH
of Cys917 in the Hinge region. Additionally, the scaffold is stabilized by hydrophobic
interactions with Leu838 and Leu1033. The 4-N-methyl anilino moiety interacts with the side
chain carbon atoms of Ile1042 and Cys1043. The 4’-OMe phenyl side chain of 11 is oriented
towards Hydrophobic pocket I and interacts with Leu838, Gly839 and Val846. The docked score
1
5

of 11 in the vertical binding mode was -5.42 kcal/mol, similar to that seen with the best docked
pose of the standard compound semaxanib (-5.02 kcal/mol) using identical docking parameters.
Figure 7b: Docked poses of 11 (pink) in the ATP-like horizontal binding mode in the putative
2
7
ATP binding pocket of VEGFR-2 (PDB: 1YWN ). Pocket amino acids are depicted in green as
sticks. Amino acids in the pocket that form strongest interactions with 11 are depicted as green
cylinders.
Figure 7b shows an alternate docked pose of 11 in the horizontal binding mode in the
binding site of VEGFR-2. In this lower scoring (-3.50 kcal/mol) pose, the furo[2,3-d]pyrimidine
4
9
scaffold of 11 is oriented like ATP in the pocket. Compound 11 forms hydrophobic interactions
with Leu838, Val846, Ala864, Phe916 and Cys917 (not labeled). The 4-N-methyl moiety of 11 is
oriented towards the Hinge region and interacts with the side chain carbon atoms of Phe916 and
Glu915 (not labeled). The 4’-OMe phenyl side chain binds in Hydrophobic pocket I and interacts
1
6

with Ile847, Ala864, Val865, Val897, Val912, Val914, Ile1042 and Cys1043. The 6-Me moiety of
1 is oriented towards the Sugar site and forms hydrophobic interactions with the side chain
1
carbon atoms of Leu838. Both these modeling results (Figures 7a and 7b) support the rationale
that the designed analogs should possess VEGFR-2 inhibitory potential. Modeling of the N-H
analogs (3 – 10) indicate that these prefer the horizontal, ATP-like binding mode (Figure 7b)
whereas the N-alkylated analogs (11 – 15) prefer the vertical mode as depicted for 11 in Figure
7
a.
Figure 8a: Docked poses of 11 (pink) in the vertical binding mode in the putative ATP binding
2
8
pocket of PDGFR-β homology model. Pocket amino acids are depicted in green as sticks.
Amino acids in the pocket that form strongest interactions with 11 are depicted as green
cylinders.
Figure 8a shows the best docked pose of 11 in the putative ATP binding pocket of a
2
8
PDGFR-β homology model. The bound pose of the furo[2,3-d]pyrimidine scaffold of 11
17

occupies the adenine binding pocket of the ATP binding site and is in the vertical binding mode
like that in Figure 7a for the docked pose of 11 in VEGFR-2. The N1 nitrogen of 11 forms a
hydrogen bond with the backbone NH of Tyr683 in the Hinge region. Additionally, the scaffold
is stabilized by hydrophobic interactions with Leu838 and Leu1033. The 4-N-methyl aniline
moiety interacts with the side chain carbon atoms of Ile1042 and Cys1043. The 4’-OMe phenyl
side chain of 11 is oriented towards Hydrophobic pocket I and interacts with Leu838, Gly839
and Val846.
The docked score of 11 in the vertical binding mode was -6.66 kcal/mol, similar to the
score of the standard compound sunitinib (-6.38 kcal/mol).
Figure 8b: Docked poses of 11 (pink) in the ATP-like horizontal binding mode in the putative
ATP binding pocket of a PDGFR-β homology model. Pocket amino acids are depicted in green
1
8

as sticks. Amino acids in the pocket that form strongest interactions with 11 are depicted as green
cylinders.
Figure 8b shows the lower scoring (-5.06 kcal/mol) alternate docked pose of 11
2
8
(
horizontal binding mode) in the homology model of PDGFR-β binding site. In this pose, the
furo[2,3-d]pyrimidine scaffold of 11 interacts with Leu604, Val614, Tyr683, Cys684 and Gly687
not labeled). The 4-N-methyl moiety is oriented towards the Hinge region and forms
(
hydrophobic interactions with Tyr683 and Cys684. The 4’-OMe phenyl side chain binds in
Hydrophobic pocket I and interacts with Thr681, Glu682, Tyr683, Leu833, Cys834 and Phe845.
The 6-Me moiety of 11 is oriented towards the Sugar site and forms hydrophobic interactions
with the side chain carbon atoms of Val614 and Ala848.
2
7
Thus, molecular modeling studies in a VEGFR-2 crystal structure (PDB: 1YWN ) and a
2
8
PDGFR-β homology model show that target compounds exemplified by 11 can exist in a
docked form in either a horizontal or a vertical pose. This is similar to the vertical and horizontal
bound conformations observed for other furo[2,3-d]pyrimidines in crystal structures with
5
0
RTKs. The energy difference between the best docked horizontal pose of 11 was within 2
kcal/mol of the lowest energy vertical docked pose in both VEGFR-2 crystal structure and the
PDGFR-β homology model. The horizontal pose conformationally orients the phenyl ring of the
1
aniline group towards the 5-position of the furo[2,3-d]pyrimidine as demonstrated by a H NMR
study (see below). The vertical pose, however, orients the methyl group of the aniline nitrogen
over the 5-position of the furo[2,3-d]pyrimidine of 11.
As expected, the preference for the binding mode(s) adopted by 3-15 was dependent on
the substitution at the aniline nitrogen. Compound 3 (with an N-H) bound predominantly in the
1
9

horizontal ATP-like binding mode in both VEGFR-2 and PDGFR-β, while both binding modes
were seen for the bound poses of compound 11 which has a 4-N-methyl group with a preference
for the vertical mode. Further homologation of the 4-N alkyl substitution in 12-15 led to
predominant docked binding in the vertical binding mode. This can be explained by two reasons:
(
a) in the horizontal binding mode, where the 4-N substitution is oriented towards the Hinge
region, increase in the size of the substitution leads to steric clashes with the amino acids; and (b)
in the vertical binding modes where the N-alkyl substitution is oriented towards Sugar site
(
Figure 7a, VEGFR-2) or towards the Hydrophobic pocket I (Figure 8a, PDGFR-β), increasing
the size of the alkyl substitution was predicted to promote hydrophobic interactions with amino
acids in the binding pocket (for example, with Ile1042 and Cys1043 in VEGFR-2 and Val614 in
PDGFR-β).
Thus, molecular modeling studies of compounds 3 – 15 (exemplified by 11) in tubulin,
VEGFR-2 and PDGFR-β suggested that these compounds have the potential for both kinase and
tubulin inhibitory activity.
Chemistry
Scheme 1. The synthesis of 4-chloro-2,6-dimethylfuro[2,3-d]pyrimidine 20.
2
0

The synthesis of the key intermediate, compound 20, is shown in Scheme 1. A three step
reaction, starting from diethyl propargyl malonate 16 was successfully employed in the synthesis
of 4-chloro-2,6-dimethylfuro[2,3-d]pyrimidine 20 (Scheme 1). The condensation of diethyl
propargyl malonate 16 and acetamidine hydrochloride 17 was attempted as a route to pyrimidine
1
8. The use of anhydrous MeOH and sodium metal were essential for the cyclization of the
pyrimidine ring. Intramolecular cyclization of 18 to the furo[2,3-d]pyrimidine 19 was carried out
under H SO (conc.) at r.t, and provides high yields (87%) of 19. This procedure is also used to
2
4
scale up the reaction for the synthesis of gram quantities of 19, as an intermediate, for the bulk
synthesis of 11 and 21 required for preclinical and in vivo studies. During scale up, conversion of
1
8 to 19 required cooling the reaction to r.t. in an ice water bath.
Scheme 2. The synthesis of N-aryl-2,6-dimethylfuro[2,3-d]pyrimidin-4-amines 3-15.
2
1

Chlorination of 19 with POCl
3
afforded the 4-chloro-2,6-dimethylfuro[2,3-d]pyrimidine
2
0. This key intermediate was reacted with the appropriate anilines (Scheme 2), in iPrOH or
nBuOH at reflux in the presence of a catalytic amount of HCl, to give target compounds 3-11.
Compounds 12-15 were synthesized via alkylation of 10 with the appropriate iodoalkane in the
presence of NaH (Scheme 2).
Scheme 3. The synthesis of hydrochloric acid salt 21.
Treatment of 11 with HCl (g) in anhydrous ether (Scheme 3) afforded the HCl salt 21 in
9
6% yield. Several different conditions were attempted to obtain 21. In the initial attempt, 100
mg of 11 was dissolved in anhydrous ether to obtain a clear solution. After HCl (g) was bubbled
through the solution, a white precipitate formed but instantly re-dissolved. A similar phenomenon
was observed when the amount of 11 was increased up to 300 mg. An alternate method to obtain
2
1 involved iPrOH as the solvent and conc. HCl as the acid source. However, no precipitate was
5
1
observed under these conditions. A literature search revealed that HCl (g) reacts with diethyl
ether to form an oxonium salt, which can dissolve a small amount but not large amounts of
organic salts. Thus, the amount of 11 was increased to 2 g. Using diethyl ether and HCl (g), 21
1
was obtained in 96% yield. Although the H NMR spectrum did not show a distinguishable peak
+
o
for the newly formed 1-N H proton, 11 and 21 have distinct melting points (108 C for 11 and
2
2

o
2
73 C for 21). In addition, elemental analysis confirmed the stoichiometry of the salt as exactly
one molecule of HCl per molecule of 21 with an additional 0.3 molecules of H O. Salt 21 has
2
excellent water solubility (>1 g/mL) and was prepared as an aqueous solution for in vivo
evaluations. .
Biological evaluations and discussion
RTK inhibitory activity of compounds 3-15 and 21 was evaluated using human tumor
cells known to express high levels of EGFR, VEGFR-1, VEGFR-2 or PDFGR-β using a
phosphotyrosine ELISA cytoblot and are listed in Table 1. Compounds known to inhibit a
particular RTK were used as positive controls for these assays. The effect of compounds on cell
proliferation was measured in A431 cancer cells known to overexpress EGFR. Finally, the effects
of selected compounds on blood vessel formation was assessed using the chicken embryo
chorioallantoic membrane (CAM) assay, a standard test for angiogenesis and the results are
presented in Table 1. In the CAM assay, purified angiogenic growth factors were placed locally
on a vascularized membrane of a developing chicken embryo together with compounds of
interest. Digitized images of the vasculature were taken 48 h after growth factor administration
and the number of vessels per unit area evaluated as a measure of vascular density.
Table 1: RTK inhibitory activities of compounds 3 – 15 and 21.
PDGFR-β
VEGFR-2 kinase kinase
CAM
A431
Cytotoxicity
EGFR kinase
inhibition (nM)
angiogenesis
inhibition
(µM)
Compound
inhibition (nM)
inhibition
nM)
(nM)
(
3
4
5
6
7
8
113.2 ± 30.2
145.2 ± 20.8
30.0 ± 4.9
50.3 ± 10.2
32.2 ± 5.3
76.3 ± 10.1
66.2 ± 8.1
58.2 ± 7.1
70.2 ± 8.0
n.d.
48.2 ± 5.9
49.1 ± 8.0
22.6 ± 3.0
55.5 ± 9.0
18.0 ± 3.0
60.0 ± 10.1
n.d.
n.d.
n.d.
16.2 ± 2.5
66.1 ± 7.2
n.d.
18.2 ± 1.8
n.d.
122.1 ± 3.4
30.2 ± 4.3
n.d.
102.3 ± 19.1
n.d.
n.d.
2
3

9
1
90.2 ± 10.2
42.2 ± 7.1
68. 2 ± 6.2
70.7 ± 14.3
73.3 ± 9.2
340.1 ± 28.3
84.4 ± 10.3
82.3 ± 10.2
n.d.
102.1 ± 17.2
43.1 ± 5.0
19.1 ± 3.0
10.6 ± 2.3
8.5 ± 0.9
34.4 ± 7.1
15.6 ± 3.0
20.2 ± 3.8
12.9
28.3 ± 4.0
63.5 ± 22.8
22.8 ± 4.9
27.1 ± 4.0
29.9 ± 4.6
56.5 ± 6.2
23.1 ± 3.7
21.9 ± 3.0
n.d.
50.0 ± 6.7
n.d.
22.6 ± 3.2
n.d.
0
11
20.8 ± 4.2
50.1 ± 8.1
42.3 ± 6.3
120.7 ± 20.1
50.2 ± 13.1
55. 8 ± 8.0
n.d.
3.9 ± 0.4
1
1
1
1
2
2
3
4
5
1
19.2 ± 3.0
16.7 ± 2.1
32.6 ± 1.9
18.9 ± 2.3
3.7 ± 0.5
semaxanib
DOX
60.0 ± 10.1
0.04 ± 0.0031
18.2 ± 2.1
1.3 ± 0.07
29.1 ± 1.9
n.d.
n.d.
n.d.
1.3 5± 0.03
10.6
cisplatin
sunitinib
erlotinib
n.d.
n.d.
n.d.
172.1 ± 19.4
1.2 ± 0.2
18.9 ± 2.7
124.7 ± 18.2
83.1 ± 10.1
n.d.
n.d.
n.d.
IC50 values of compounds in cell lines with high expression levels of the indicated kinase, A431
cytotoxicity and inhibition of the CAM formation
In the cellular assays using cells with high expression of EGFR, all the compounds
showed lower potency than the standard erlotinib. Compounds 5 and 7 were the most potent and
were about 25-fold less potent than erlotinib. An iPr substitution on the N4-aniline moiety in 14
affords the least potent analog. Against VEGFR-2 expressing cells, all the target compounds
showed good to excellent potency comparable to the standard compounds, sunitinib and
semaxanib. Among them 11 and 21 were the most active compounds. An N4-alkyl substitution
affords the most potent compounds (11 – 15). In the PDGFR-β expressing cells, all the tested
compounds showed better potency than the standard sunitinib. Compound 5 was the most potent
compound in the series and had an IC50 of 16.2 nM. Compounds 11 and 21 were equipotent and
were the second most potent compounds in the series against PDGFR-β expressing cells. They
were about 4-fold more potent than sunitinib. In the A431 cytotoxicity assay, all of the target
compounds were about 10- to 50-fold less potent than cisplatin.
In the CAM angiogenesis inhibitory assay, 11 and 21 were again the most potent
2
4

compounds of the series showing single-digit µM inhibition of angiogenesis, which is about 2-
fold less potent than sunitinib and 9-fold more potent than erlotinib.
Although the anilino nitrogen in the N-methyl compounds 11 and 21 cannot form
hydrogen bonds as hydrogen bond donors with RTK at the hinge region, both compounds
showed activities in cells expressing various RTKs, comparable to and even slightly better than
the other NH analogs. These results validate our design rationale for RTK inhibitory activity of
the proposed hybrid compounds.
A
B
36
2
5

1
Figure 9. The conformations and H NMRs of compound 10 (A) and 11 (B).
1
A H NMR study, in DMSO-d
6
, was carried out to explore the conformation of 10 and 11
-N and N-C ) connecting the phenyl ring and the
(
Figure 9). In compound 10, the σ bonds (C
1
4
pyrimidine ring are both freely rotatable, while these bonds were restricted in 11, where the
1
additional methyl group was introduced on the N4-position. According to the H NMR spectrum,
the furo[2,3-d]pyrimidine 5-H proton in 11 is more shielded than in 10, which suggested a
nearby diamagnetic anisotropic cone. Due to the bulk of the N-methyl group, the conformation
of 11 is restricted and the phenyl ring in 11 has to position itself on top of the 5-H proton (Figure
9
), which accounts for the observed shielding effect.
1
H NMR studies of 11 in a DMSO-d solution indicates that the compound exists in a
6
conformation similar to that seen in the horizontal docked poses (Figures 6, 7b and 8b) and not in
the higher scoring vertical docked pose predicted for VEGFR-2 and PDGFR-β. Clearly, steric
hindrance of the 4-N-substitution with the Hinge region in the docked pose forces the alternate
vertical orientation which adopts the higher energy conformation in order for the OMe-phenyl
ring to interact with Hydrophobic pocket I. In the docked pose in the tubulin colchicine site
1
(
Figure 7a), 11 approximates the conformation suggested by the H NMR and predicted by the
2
6
1
energy minimized conformation of 11 by molecular modeling (MOE 2010.10 ). The H NMR
conformation for 10 is also predicted by the lowest energy conformation for 10 obtained by
2
6
molecular modeling (MOE 2010.10 ).
2
6

Table 2. Tumor cell inhibitory activity (GI50,nM) of 11 (NCI 60 cell line panel).
Panel/Cell
line
GI50
Panel/ Cell line GI50
(nM)
Panel/ Cell line GI50
(nM)
Panel/ Cell line
GI50
(nM)
(nM)
NSCLC
Renal Cancer
Ovarian cancer
Prostate Cancer
DU-145
A549/ATCC 37.6
786 – 0
43.6
19.5
55.9
16.3
56.7
75.2
OVCAR-3
OVCAR-4
OVCAR-5
OVCAR-8
SK-OV-3
29.1
60.4
55.8
36.6
25.2
26.2
42.2
EKVX
64.8
32.2
84.1
40.7
33.1
<1
A498
Breast Cancer
MCF7
HOP-62
NCI-H226
NCI-H23
NCI-H460
NCI-H522
CNS Cancer
SF-268
ACHN
CAKI-1
SN 12C
UO-31
MDA-MB-231/ATCC 44.3
HS 578T
BT-549
15.3
48.0
34.9
Melanoma
LOX IMVI
MALME-3M
M14
Colon Cancer
COLO 205
HCC-2998
HCT-116
HCT-15
HT29
54.9
42.3
23.3
33.9
37.7
22.9
14.6
MDA-MB-468
Leukemia
CCRF-CEM
HL-60(TB)
K-562
20
33.6
13.8
20.0
33.3
53.2
30.8
28.5
34
25.3
17.3
13.2
67.9
42.8
32.0
SF-295
SK-MEL-2
SK-MEL-28
SK-MEL-5
UACC-62
SF-539
25.3
32.5
21.1
29.8
SNB-19
SNB-75
U251
MOLT-4
KM12
RPMI-8226
SR
SW-620
Compound 11 was initially evaluated in the NCI 60 cell line panel and showed potent
5
2
GI50s against most of the NCI 60 cancer cell lines (Table 2). A COMPARE analysis (Table 3)
of the NCI 60 cell line data showed tubulin inhibitor vincristine sulfate to have the closest
Pearsons correlation coefficient with 11. Other compounds, such as vinblastine sulfate and
maytansine, also tubulin binding agents, were ranked as the next closest correlations. This clearly
suggested antitubulin activity and warranted the further evaluation of 11 and the other analogs as
2
7

potential tubulin binding agents.
Table 3. NCI COMPARE analysis results for 11.
Rank
Vector
Correlation
0.600
Cell line
49
1
2
vincristine sulfate S67574 -3M TGI 2 days AVGDATA
maytansine S 153858 -4M TGI 2 days AVGDATA
vinblastine sulfate S49842 -5.6M TGI 2 days AVGDATA
0.494
49
3
0.458
49
Based on the results of the COMPARE correlation, the effects of 11 and 21 on cellular
microtubules were evaluated. A-10 cells were treated with a range of concentrations of 11, 21
and related compounds. Both 11 and 21 caused dose-dependent loss of cellular microtubules,
similar to the effects of CA-4 (Figure 5). Using these cells an EC50 (concentration required to
cause 50% loss of cellular microtubules) (Table 4) was calculated for 11 and other analogs. The
3
3
EC50 value for 11 was 103 nM, whereas CA-4 had an EC50 of 7 nM, confirming the
COMPARE analysis results and validating our initial design rationale with respect to antitubulin
inhibitory activity for the proposed hybrid compounds (11 – 13, 15 and 21). To our knowledge,
11 and 21 are the first furo[2,3-d]pyrimidines reported to have potent dual RTK inhibition and
microtubule disrupting activity. Structure–activity analyses of the other members of this series
identified that the size of the N-alkyl group determines the microtubule depolymerization
potency (Table 4). Compound 11 (EC50 = 103 nM) with an N-methyl group is the most potent
tubulin depolymerizer in the series. The ethyl substituted 12 (EC50 = 1010 nM) is about 10-fold
less potent than 11. Larger N-substitutions, including propyl, isopropyl and n-butyl groups,
resulted in a loss of microtubule depolymerizing activity (Table 4). In addition, an N-H, without
an alkyl group (10) is also inactive.
2
8

Table 4. Microtubule depolymerizing activities of 10 – 15.
EC50 for
Compound
CA-4
microtubule
depolymerization
7 nM
1
1
1
1
1
1
2
0
1
2
3
4
5
1
> 40 µM
103 nM
1.1 µM
13.5 µM
> 40 µM
>40 µM
570 nM
Figure 10. Compounds 11and 21 inhibit the polymerization of purified porcine brain tubulin.
Compounds were incubated at 10 µM with 2.2 mg/mL purified porcine brain tubulin in the
presence of 10% glycerol, and GTP. CA-4 was used as a positive control.
2
9