Discovery and Development of a Series of Pyrazolo[3,4- d]pyridazinone Compounds as the Novel Covalent Fibroblast Growth Factor Receptor Inhibitors by the Rational Drug Design

Article abstract of DOI:10.1021/acs.jmedchem.9b00510

Alterations of fibroblast growth factor receptors (FGFRs) play key roles in numerous cancer progression and development, which makes FGFRs attractive targets in the cancer therapy. In the present study, based on a newly devised FGFR target-specific scorin

Full text of DOI:10.1021/acs.jmedchem.9b00510

Subscriber access provided by UNIV OF SOUTHERN INDIANA
Article
The Discovery and Development of a Series of Pyrazolo[3,4-d]pyridazinone
Compounds as Novel Covalent Fibroblast Growth Factor
Receptor (FGFR) Inhibitors by Rational Drug Design
Yulan Wang, Yang Dai, Xiaowei Wu, Fei Li, Bo Liu, Chunpu Li, Qiufeng Liu, Yuanyang
Zhou, Bao Wang, Mingrui Zhu, Rongrong Cui, Xiaoqin Tan, Zhaoping Xiong, Jia Liu, Minjia
Tan, Yechun Xu, Mei-Yu Geng, Hualiang Jiang, Hong Liu, Jing Ai, and Mingyue Zheng
J. Med. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.jmedchem.9b00510 • Publication Date (Web): 23 Jul 2019
Downloaded from pubs.acs.org on July 23, 2019
Just Accepted
“Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted
online prior to technical editing, formatting for publication and author proofing. The American Chemical
Society provides “Just Accepted” as a service to the research community to expedite the dissemination
of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in
full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully
peer reviewed, but should not be considered the official version of record. They are citable by the
Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore,
the “Just Accepted” Web site may not include all articles that will be published in the journal. After
a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web
site and published as an ASAP article. Note that technical editing may introduce minor changes
to the manuscript text and/or graphics which could affect content, and all legal disclaimers and
ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or
consequences arising from the use of information contained in these “Just Accepted” manuscripts.
is published by the American Chemical Society. 1155 Sixteenth Street N.W.,
Washington, DC 20036
Published by American Chemical Society. Copyright © American Chemical Society.
However, no copyright claim is made to original U.S. Government works, or works
produced by employees of any Commonwealth realm Crown government in the course
of their duties.

Page 1 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
of Sciences, State Key Laboratory of Drug Research; ShanghaiTech
University
Liu, Hong; Shanghai Institute of Materia Medica Chinese Academy of
Sciences, Key Laboratory of Receptor Research
Ai, Jing; Shanghai Institute of Materia Medica Chinese Academy of
Sciences, State Key Laboratory of Drug Research; University of the
Chinese Academy of Sciences
Zheng, Mingyue; Shanghai Institute of Materia Medica Chinese Academy
of Sciences, State Key Laboratory of Drug Research
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 2 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
The Discovery and Development of a Series of Pyrazolo[3,4-
d]pyridazinone Compounds as Novel Covalent Fibroblast Growth
Factor Receptor (FGFR) Inhibitors by Rational Drug Design
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Yulan Wang , Yang Dai , Xiaowei Wu2, †, Fei Li1, 3, Bo Liu , Chunpu Li , Qiufeng Liu , Yuanyang Zhou , Bao
1
, †
1, †
1
1, 2
2
1, 4
2
, 5
1
1
1, 4
1, 5
1
1
2
Wang , Mingrui Zhu , Rongrong Cui , Xiaoqin Tan , Zhaoping Xiong , Jia Liu , Minjia Tan , Yechun Xu ,
1
1, 5
2, *
1, 4, *
1, *
Meiyu Geng , Hualiang Jiang , Hong Liu , Jing Ai
, Mingyue Zheng
1
State Key Laboratory of Drug Research, Shanghai Institute of Materia Medica, Chinese Academy of Sciences, 555
Zuchongzhi Road, Shanghai 201203, China
2
Key Laboratory of Receptor Research, Shanghai Institute of Materia Medica, Chinese Academy of Sciences, 555 Zu
Chong Zhi Road, Shanghai, 201203, China
3
School of Chemistry, ShangHai University, 99 ShangDa Road, ShangHai 200444, China
University of Chinese Academy of Sciences, No.19A Yuquan Road, Beijing 100049, China
School of Life Science and Technology, ShanghaiTech University, 393 Middle Huaxia Road, Shanghai 200031,
4
5
China
ABSTRACT
Alterations of fibroblast growth factor receptors (FGFRs) play key roles in numerous
cancer progression and development, which makes FGFRs attractive targets in cancer therapy. In
the present study, based on a newly devised FGFR target-specific scoring function, a novel FGFR
inhibitor hit was identified through virtual screening. Hit-to-lead optimization was then performed
by integrating molecular docking and site-of-metabolism predictions with an array of in vitro
evaluations and X-ray cocrystal structure determination, leading to a covalent FGFR inhibitor 15,
which showed a highly selective and potent FGFR inhibition profile. Pharmacokinetic assessment,
protein kinase profiling and hERG inhibition evaluation were also conducted, and they confirmed
the value of 15 as a lead for further investigation. Overall, this study exemplifies the importance
of the integrative use of computational methods and experimental techniques in drug discovery.
ACS Paragon Plus Environment

Page 3 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
INTRODUCTION
The fibroblast growth factor receptor (FGFR) family is a subfamily of receptor tyrosine
kinases (RTKs), which comprises four highly conserved transmembrane receptor tyrosine kinases,
1
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
FGFR1, FGFR2, FGFR3 and FGFR4. FGFRs play important roles in a variety of cell functions,
such as cell proliferation and differentiation, and biological processes, including development,
2
angiogenesis, homeostasis, and wound repair. Alterations of FGFRs are associated with the
progression and development of several cancers, and they contribute to carcinogenesis in three
main situations, namely, driver mutations, neoangiogenesis and resistance to anticancer agents.
Extensive FGFR alterations in cancers have been observed, such as FGFR1 amplifications in 6%
of small cell lung carcinomas, 20% of squamous non–small cell lung carcinomas,17% of
osteosarcomas, 10%~15% of breast cancers, 5% of ovarian cancers and 9% of esophageal cancers;
FGFR2 translocations in 14% of intrahepatic cholangiocarcinomas; FGFR2 mutations in 12%~14%
of endometrial cancers and 5% of squamous non-small cell lung carcinomas; FGFR2
amplifications in 12%~14% of gastric cancers and 4% of breast cancers; FGFR3 translocations in
3
%~6% of bladder carcinomas, 3% of glioblastomas and 15%~20% of myelomas; FGFR3
mutations in 60%~80% of non-muscle-invasive bladder carcinomas; FGFR4 mutations in 6%~8%
3
of rhabdomyosarcomas; etc. Therefore, targeting FGFRs with small-molecule inhibitors
represents a promising therapeutic strategy for treating the above FGFR-related cancers.4
A large amount of effort has been devoted to developing FGF/FGFR inhibitors as
5
anticancer treatments. Several classes of small-molecule FGFR inhibitors have been reported, and
they can be clustered into two classes based on their inhibition profiles. The first class of inhibitors
are characterized by “multitarget” tyrosine kinase inhibitors (TKIs), which have broad targets,
including FGFR. In addition to FGFR, these “multitarget” TKIs may also target vascular
endothelial growth receptors (VEGFRs), platelet-derived growth factor receptors (PDGFRs) and
other tyrosine kinases, resulting in serious adverse effects in clinical studies, in particularly
VEGFR2-based dose-limiting toxicities.6- 8 The second class is inhibitors with highly selective and
potent inhibitory activities against FGFR. These compounds can specifically target FGFR,
reducing the risks of side effects in clinical usage and making the mechanism of action (MOA)
easier to study. According to their MOA, FGFR-selective inhibitors fall into two categories:
noncovalent inhibitors and covalent inhibitors. To date, several selective inhibitors, including both
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 4 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
9
noncovalent and covalent inhibitors (shown in Figure 1; noncovalent inhibitors: AZD4547 (1),
1
0
11
12
13
14
NVP-BGJ398 (2), E7090 (3), LY2874455 (4), JNJ-42756493 (5), and CH5183284 (6);
covalent inhibitors: TAS 12015, 16 (7)) are in clinical development, and some have demonstrated
clinically meaningful benefits with manageable toxicity profiles in clinical trials. For example, 1,
a selective reversible FGFR inhibitor, has demonstrated a well-tolerable safety profile and modest
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
7
antitumor activity in advanced squamous cell lung cancers in a phase Ib study. In a phase Ib/IIa
study, a combination of 1 with either anastrozole or letrozole showed anti-tumor activity in
1
8
advanced estrogen receptor-positive breast cancers resistant to aromatase inhibitors. In addition,
a selective covalent FGFR inhibitor, 7, has shown a tolerable safety profile and antitumor activity
in patients with cholangiocarcinoma harboring FGFR2 gene fusions and cholangiocarcinoma
(
CCA) patients with FGFR2 gene fusion resistance prior to the FGFR inhibitor 2.15, 16, 19 A phase
II study of 7 with patients with intrahepatic CCA harboring FGFR2 gene fusions has been initiated.
Despite the advances made in FGFR inhibitors, the road to clinical treatment of various diseases
related to alterations of FGFR remains long and challenging. The structural diversification of
highly selective and highly bioactive FGFR inhibitors is still in high demand, and the options for
the clinical treatment of diverse FGFR aberrations in a wide variety of cancers are still very limited,
highlighting the importance of developing FGFR inhibitors of novel chemotypes.
ACS Paragon Plus Environment

Page 5 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 1. FGFR-selective inhibitors in clinical development.
The current study reports the discovery and development of a series of pyrazolo[3,4-
d]pyridazinone derivatives as a novel type of FGFR inhibitors, and it emphasizes the synergetic
use of computer-aided drug design (CADD) with experimental evaluations to address different
issues in our hit identification and hit-to-lead optimization stages. Moreover, a variety of chemical,
biophysical and bio-pharmacological methodologies were used for the full characterization of the
anticancer activity, physicochemical and PK properties, and mechanism of action (MOA) of the
discovered FGFR inhibitor 15, which may serve as a promising new lead for developing potential
anticancer treatments.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 6 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
RESULTS AND DISCUSSION
1
. Hit Discovery by Target Specific Scoring Function and Virtual Screening
The target-specific scoring function has proven effective in enhancing the enrichment of
virtual screening (VS) for novel hit discovery, and for interpreting the subtle commonalities and
differences in the ligands binding to the target binding site of interest. Previously, we built a target-
specific scoring function for methyltransferases, SAMscore, based on the crystal structures of
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
2
0
methyltransferase and their ligands activity. The SAMscore demonstrated improved scoring
ability and has been successfully used to identify novel disruptor of silencing telomeric 1-like
2
0
inhibitors. To date, the crystal structures of many FGFR proteins and FGFR inhibitors have been
reported, providing the basis for the development of FGFR target-specific scoring functions. In
this study, using a workflow similar to that of SAMscore, we developed a target-specific scoring
function for FGFR, RTKscore. First, we selected protein crystal structures for FGFR1, FGFR2 and
10
21
22
FGFR3, and the PDB codes were 3TT0 , 2PVF and 4K33 , respectively. In addition, 1,090
FGFR ligands with known chemical structural information and activity data were collected from
2
3
9
Binding DB. The ligands with reported inhibitory activity data such as > 10 nM were marked
2
4
as inactive. Second, the molecular docking program Glide was used to generate the binding
modes of the FGFR proteins and their ligands. The iterative potential of mean force (iPMF)25
features of the binding mode of each ligand with the corresponding protein were calculated to
characterize the interaction between the ligand and the protein as the input data for modeling
building the model. Third, original features were processed and selected with the machine learning
2
6
27
methods epsilon support vector regression (ε-SVR) and RFE , in which the features were
removed stepwise according to their importance to the model. Here, the importance of each feature
2
was evaluated by calculating the square of its Pearson correlation coefficient (R ) with the ligands’
20
affinity data. Similar to our previous report, iPMF features were calculated to characterize the
protein-ligand interactions, and these features included the pairwise interaction features involving
1
7 types of atoms in the protein, 30 types of atoms in the ligands, and 11 distance ranges. In this
study, a total of 1,125 features were selected for further model building. A five-fold cross
validation process was used to evaluate performance of the model. Finally, we obtained a model
2
called RTKscore with reasonable prediction ability; the R of its 5-fold cross validation was 0.46.
In addition, we compared RTKscore with Glide on a same test set, and RTKscore exhibited a more
ACS Paragon Plus Environment

Page 7 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
2
significant correlation with the pIC50 values of these ligands (Figure 2). The R of RTKscore was
2
0.59, whereas the R of Glide was near zero.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 2. Workflow of RTKscore development and comparation of RTKscore with Glide on a test
set.
As described above, the RTKscore showed better predictive performance for FGFR ligands
than a general-purpose scoring function. We therefore applied the RTKscore in a VS project for
discovering novel FGFR1 inhibitors. The SPECS small-molecule database was used as the ligand
2
8
databases, and compounds containing any PAINS substructures, inorganic atoms, unwanted
functionalities, and reactive groups were filtered out first. There were 190,038 ligands that
remained and were prepared for the virtual screening, and Glide was used to generate the potential
binding modes of each ligand. The RTKscore was used to rescore the poses of each ligand, which
included calculating the iPMF features for each pose, the feature subset extraction and rescoring
with the RTKscore. According to the highest score of each ligand, as scored by RTKscore, we kept
the top 1,000 ligands for further evaluation. To ensure structural diversity among the ligands for
biochemical testing, we clustered ligands into 100 clusters on the basis of their chemical
2
9
similarities as measured by molecular fingerprints ECFP4, and only one or two ligands from
each cluster were selected. Finally, 112 ligands were selected and purchased for FGFR1 inhibitory
activity evaluation. We used ELISA assay to test the inhibitory activities of the ligands against
FGFR1 at 50 µM and found that compound 8 showed high inhibitory activity against FGFR1, with
an IC of 114.5±15.3 nM. Its inhibitory activity against FGFR1-depended KG1 cell proliferation
50
was tested, and the results showed that this compound could inhibit the proliferation of KG1 cell
with IC = 1107.8±264.7 nM.
5
0
2. Hit-to-Lead Optimization by Covalent Inhibitor Design
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 8 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
To better investigate the structure-activity relationships, we used molecular docking to
simulate the binding mode of 8 with FGFR1. As shown in Figure 3B, the putative binding mode
of compound 8 follows the typical pattern of most reported FGFR inhibitors. It may also form
hydrogen bonds with the hinge area of FGFR1, in which the pyridazine moiety of compound 8
forms three hydrogen bonds with the residues Ala564 and Glu562. The benzofuran moiety of
compound 8 was located in the hydrophobic pocket of the protein active site, and the N-phenyl
substituent of the ligand reached the solvent-exposed area of the loop part of the protein. Based on
the putative binding mode of 8 with the FGFR1 protein, we found that the hydrophobic pocket of
FGFR1 has additional space (Figure 3C). To better accommodate the hydrophobic pocket, we
first explored benzofuran substituents. After testing a few kinds of substituents, we found that
compound 9 with a 3,5-dimethyl-2-benzofuran moiety shows more potent inhibitory activities,
with FGFR1 IC = 10.8±1.3 nM and KG1 cell IC = 415.7±8.7 nM.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
5
0
50
ACS Paragon Plus Environment

Page 9 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 3. (A) The structures of compounds 8, 9, 10 and 11; (B) putative binding mode of
compound 8 with FGFR1(PDB code: 3TT0); (C) the active site pocket of FGFR1 depicted as mesh
surface. (D) the putative binding modes of compound 10 (blue) and compound 11 (orange) with
FGFR1 (PDB code: 3TT0) simulated by covalent docking. (E) FGFR1, FGFR2, FGFR3 and
FGFR4 P-loop sequence alignment.
Recently, with the resurgence of covalent inhibitors, more and more covalent small
molecule kinase inhibitors have been approved by FDA.30-32 A large number of kinase targets have
a cysteine located in the vicinity of the ATP pocket that could be targeted by irreversible
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 10 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
inhibitors.33, 34 As for FGFR, covalent targeting the Cys residue in the P-loop of FGFRs is also an
3
5
effective strategy for developing selective inhibitors. As reported by the Gray laboratory, the
covalent inhibitor showed gratifying inhibitory activity against FGFR1 at both the molecular level
3
5
and cellular level. In this study, the simulated binding mode of 8 implied that its N-phenyl
substituent extends out the P-loop of FGFR1, which contains a conserved Cys residue within
FGFR family (shown in Figure 3E marked with a green box, i.e., Cys 488 in FGFR1). The distance
from the phenyl ring of 8 to this Cys residue was 11.1 Å, within the appropriate distance range for
a reactive Michael acceptor (Figure 3B). Based on this observation, two covalent inhibitors 10
and 11 were designed by introducing Michael acceptor substituents with different distance to the
benzene ring of compound 9. We used covalent molecular docking to simulate the binding mode
of 10 and 11 with FGFR1, and the simulated results showed that both compounds could form a
covalent interaction with Cys488 of the FGFR1 P-loop without a significant change in their overall
binding mode (Figure 3D).
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
3
. In vitro Activities of Compounds 10 and 11
Compounds 10 and 11 were synthesized and tested for their inhibitory activity against
FGFRs and a diverse panel of representative human kinases. As shown in Table 1, both 10 and 11
exhibited great inhibitory activity to FGFR1, FGFR2, and FGFR3 in vitro (Table 1). Meanwhile
compound 10 also showed potent inhibitory activity against FGFR4 (Table 1). Compounds 10 and
11 showed excellent selectivity for FGFR were observed across typical angiogenesis-regulating
kinases, including VEGFR1 and VEGFR2, PDGFRα, PDGFRβ, VEGFR1 and VEGFR2, and
other tested human kinases.
Increased FGFR signaling promotes cancer cell proliferation. To elucidate the impact of
1
0 and 11 on FGFR-mediated cancer cell proliferation, five FGFR-driven cell lines harboring
frequently occurring oncogenic forms of FGFRs were chosen: FGFR1-translocated KG1 cells,
FGFR2-amplified H716 cells, FGFR2-amplified SNU16 cells, FGFR3-mutant UMUC14 cells,
and BaF3/FGFR1 cells. As shown in Table 1, compound 10 strongly inhibited cell proliferation
of KG1, H716, SNU16, UMUC14 and BaF3/TEL-FGFR1, and most of its IC values were less
5
0
than 10 nM, making it more potent than AZD4547. The model cell line BaF3/TEL-VEGFR2,
which stably expressed the constitutively active form of TEL-VEGFR2, was used to further assess
ACS Paragon Plus Environment

Page 11 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
the cellular selectivity of compound 10 for FGFR against VEGFR2. As shown in Table 1,
compared with its high potency against cell proliferation of BaF3/TEL-FGFR1 (IC < 0.3 nM),
5
0
compound 10 had no significant inhibitory effect on VEGFR2-mediated cell proliferation (IC =
50
6
119.8±1040.6 nM), confirming the high selectivity of compound 10 for FGFRs against VEGFR2
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
at the cellular level.
Table 1. Molecular and cellular inhibitory activity evaluation of compounds 10 and 11.
IC (nM)
IC (nM)
5
0
50
Cell
Kinase
1
0
11
1
10
11
1
FGFR1
FGFR2
FGFR3
4.8 ± 1.6
2.3 ± 0.1
15.0 ± 1.1
16.3 ± 1.1
6.6 ± 1.6
12.5 ± 1.2
0.6 ± 0.1
0.4 ± 0.0
3.4 ± 0.4
KG1
<0.3
<0.3
0.6±0.0
1.3±0.1
36.4±5.3
3.7±0.6
4.9±0.6
10.3±0.6
SNU16
UMUC14
BaF3/TEL-
FGFR1
6.4±2.1
FGFR4
23.1 ± 3.6
>200
49.8 ± 7.6
<0.3
9.8±2.8
>10000
0.5±0.0
BaF3/TEL-
VEGFR2
6119.8±104
0.6
VEGFR2 >1000
>1000
57.9 ± 22.1
416.9±10.0
VEGFR-1 >1000
67.6%@1000
58.8%@1000
\
\
\
\
\
\
\
\
\
\
\
PDGFR-β >1000
Ret
78.4%@1000 >1000
c-Src
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
>1000
c-Met
ALK
EGFR
ErbB2
ErbB4
EPH-A2
ABL
To further evaluate the cellular activity of compound 10 for targeting FGFR kinase, we
analyzed its effects on the phosphorylation of FGFR and its major downstream signaling molecule,
Erk. Two representative human cancer cell lines with FGFR aberrations were used, namely, KG1
and SNU16. We found that 10 showed significant inhibition of the phosphorylation of FGFR1 and
FGFR2 in the individual cancer cell lines. The phosphorylation of Erk was also inhibited (Figure
4
). Thus, at the cellular level, 10 potently inhibits FGFR signaling.
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 12 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 4. The effects of compound 10 on the phosphorylation of FGFR and the downstream
effector Erk in the KG1 and SNU16 cell lines.
4
. Investigation of the Covalent Binding Mechanism Investigation
To elucidate the interaction mechanism of this series of compounds, we solved the
cocrystal structures of compound 9 in complex with FGFR1 (PDB id: 6ITJ), compound 10 with
FGFR4 (PDB id: 6IUP), and compound 11 with FGFR4 (PDB id: 6IUO). Overall, these cocrystal
structures provided a solid structural basis for understanding the binding mode of this series of
compounds, and also confirmed our molecular design concepts. Figure 5A shows the crystal
structure overlaid with the predicted structure, where the binding mode of 9 determined in the
crystal structure is highly consistent with the binding mode simulated by molecular docking.
Figure 5B compares the crystal structures of complexed 10 and 11, which differ from each other
by only a -CH - group in the acrylamide moiety. As expected, compounds 10 and 11 have binding
2
modes highly similar to that of compound 9, and the acrylamide moiety of these two compounds
reached the targeted Cys residue in the P-loop. The P-loop is highly flexible, and the minor
structural differences between the compounds significantly influenced the spatial arrangement of
the loop. For compound 11, a covalent bond with the Cys477 residue of FGFR4 can be observed.
For compound 10, the distance from the carbon atom of the Michael acceptor to the sulfur atom
of Cys477 is 2.6 Å, which is slightly longer than the C-S bond length. Because the crystallographic
structure is static and represents the spatial average over the whole crystal, we cannot rule out a
potential covalent bonding interaction between 10 and FGFR4 given such a short C-S distance and
the highly flexibility of the binding site.
ACS Paragon Plus Environment

Page 13 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 5. (A) The cocrystal structure (shown in green) of compound 9 and the FGFR1 complex
(PDB id: 6ITJ) and the putative binding mode (shown in pink) of compound 9 and FGFR1
simulated by molecular docking. (B) The crystal structures of the complex of compound 10 (shown
in magenta) with FGFR4 (shown in salmon) (PDB id: 6IUP) and the complex of compound 11
(shown in lemon) with FGFR4 (shown in cyan) (PDB id: 6IUO).
To investigate the potential covalent bond between 10 and FGFR4, we used protein mass
spectrometry to analyze the amino acid residues modified by compound 10 in the FGFR4 kinase
domain. The FGFR4 kinase domain protein was incubated with compound 10 and digested with
trypsin. Then, the tryptic peptides were subjected to LC-MS/MS analysis. The MASCOT program
was used to identify protein modification by searching the Uniport database. The target peptide,
LVLGKPLGEGCFGQVVR (the peptide contains the Cys residue that may be covalently bound),
has been identified (Figure S1). Further analysis showed that the Cys residue of the peptide,
LVLGKPLGEGCFGQVVR, was modified by 440.46 Da, which exactly matches the molecular
weight of a compound 10 (Figure S2). These results indicated that compound 10 could bind to the
FGFR4 protein and form specifically a covalent bond with the residue of Cys477.
5
. Optimization of PK Properties
The pharmacokinetics (PK) properties of compound 10 in rat were determined, and it
showed low oral bioavailability in rats (Table 4). To improve the PK profile of 10, we first used
site-of-metabolism (SOM) prediction programs to identify the reactive sites of 10. Three SOM
prediction programs were used: SOME-UGT36, 37, SMARTCyp and the metabolism prediction
38
3
9
module of Maestro (Schrödinger LLC, New York, NY, 2015). The 5-methyl group on the
benzofuran moiety was predicted to be a reactive site by all three programs (Figure 6). In addition,
the Michael acceptor moiety that is prone to nucleophilic reactions. Overall, we considered that
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 14 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
the Michael acceptor of 10 and the 5-methyl group on the benzofuran are high priorities for site
modification.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 6. SOM prediction of compound 10 by three programs. SMARTCyp (orange circles,
metabolism probability ordered by size of circle), SOME (blue triangles) and SOME-UGT (grey
diamonds), and Schrödinger (green pentagons, metabolism probability ordered by size).
The optimization of the covalent reactive groups is a key challenge in covalent drug design.
An ideal reactive group should have sufficient reaction activity to form the desired bond with the
target protein, but indiscriminate and high reactivity may lead to insurmountable PK problems and
toxicity in the late stage of drug development.40, 41 For compound 10, we aim to slightly weaken
the reactivity of the Michael acceptor, to reduce its binding to off-target nucleophiles, for example,
glutathione (GSH) and proteins with exposed nucleophilic centers such as serum albumin. These
unspecific interactions may account for the fast clearance of 10. Several studies have reported that
a useful optimization strategy is to add an N,N-dimethylaminomethyl substituent or an analogues
to the end of the Michael acceptor,42, 43 which would not only reduce the electrophilicity but also
enhance the steric hindrance of the Michael acceptor. Moreover, a basic functional group onto the
Michael acceptor of the ligand would improve the reactivity between the ligand and the target
protein by catalyzing the intramolecular Michael addition and/or having an inductive effect from
4
4
the protonated basic group. Thus, we substituted the Michael acceptor moiety of compound 10
with several different amino methyl substituents.
Regarding the 5-methyl group on the benzofuran moiety of 10, the methyl group can be
replaced with a halogen to improve metabolic stability in a straight forward manner. In this study,
we replaced the 5-methyl group with two different halogen atoms: chlorine and bromine. As shown
in Table 2, five derivatives were synthesized for biochemical and metabolic evaluation. First, we
tested the inhibitory activities of these derivatives against FGFR1 and FGFR2, and their inhibitory
ACS Paragon Plus Environment

Page 15 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
activities on the proliferation of KG1 cells and SUN16 cells. The results suggested that these
structural modifications retained the in vitro activities of these compounds. All of these derivatives
potently inhibit FGFR1 and FGFR2 enzymatic activity and effectively inhibit the proliferation of
FGFR1-depended cancer cells KG1. Second, we compared the solubility of these compounds
using a kinetic solubility determination assay. Among them, compounds 10 and 12 showed poor
solubility (20 µM), and compounds 13, 14 and 16 were more soluble, with solubilities comparable
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
4
5
46
to that of ibrutinib (17) (an anticancer drug with low solubility ), and compound 15 showed the
highest solubility (200 µM, Table 2). The metabolic stability of compounds 10, 12, 13 and 15 in
mouse liver microsomes were evaluated. As shown in Table 2, in vitro T1/2 values for the four
compounds ranged from 29 min for compound 12 and 31 min and 63 min for compound 10 and
compound 13, respectively, up to 103 min for compound 15 which showed the best stability in
mouse liver microsomes because of the longest half-life.
Table 2. Derivatives of compound 10 and their biochemical and metabolic evaluation.
1
1
0: R = CH , R = H
14: R = CH , R =
1 3 2
1
3
2
2: R = Cl, R = H
1
2
N
N
1
1
5: R = Cl, R =
1 2
N
13: R = CH , R =
1
3
2
6: R = Br, R =
1
2
Solubility
at pH 7.4
(µM)
20
Microsomal stability
FGFR1
nM)
FGFR2
(nM)
KG1 cell
(nM)
Compound ID
in vitro T1/2
(
(min)
31
29
63
--
1
1
1
1
1
1
0
2
3
4
5
6
1.2±0.3
0.8±0.3
2.4±0.1
30.9±7.3
20.4±0.9
0.4±0.1
0.6 ± 0.1
--
1.5±0.1
1.1±0.0
1.9±0.7
8.0±0.7
7.2±3.6
2.0±0.0
0.4±0.0
--
<0.3
<0.3
20
<0.3
50
2.7±1.4
3.7±2.5
0.4±0.1
8.3 ± 2.1
--
50
200
50
103
--
1
--
--
1
1
7
8
50
--
--
--
--
200
--
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 16 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
To cost-effectively evaluate the reactivity of the designed covalent inhibitors, their
undesirable reactions with GSH protein were evaluated in vitro. GSH protein was incubated
separately with compounds 10, 13, 15, and 16 separately, and then their GSH binding metabolites
4
7
were determined with LC-MS/MS, and covalent drugs 17 and afatinib (18) were chosen as
controls. As shown in Figure 7A, the GSH adduct of 10 was the only quickly generated metabolite,
and other compounds demonstrated comparable stabilities toward off-target thiols of GSH. Thus,
we supposed that the poor PK profile of compound 10 is mainly caused by its high intrinsic
chemical reactivity profile, and the modified acrylamide functional groups in compounds 13, 15
and 16 have acceptable reactivity profiles compared with the covalent drugs on the market. This
argument is further supported by the following rat whole blood stability assay,48, 49 in which we
incubated compounds 13 and 15 in whole blood for 5 h and detected the remaining parent
compounds at different time points. As shown in Figure 7B, the prototype of compound 10
decreased quickly, and compound 15 decreased relatively slowly. Moreover, after 5 h, with 10, the
parent compound tended to be depleted, while 20% of compound 15 remained. In general,
compound 15 showed good solubility, a weaker binding affinity for GSH, and improved stability
in whole blood. Therefore, we chose 15 as the lead compound for further characterization.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 7. (A) GSH affinity evaluation of compounds 10, 13, 15, and 16; (B) blood stability
evaluation of compounds 10 and 15.
6
. Characterization of Lead Compound 15
Biochemical Kinetic profiling. We further did the kinetic assessment of compounds 10 and
1
5. The FGFR1 was used as the representative target kinase. At first, a classic diluting assay was
conducted to determine whether compounds 10 and 15 inhibited FGFR via irreversible binding.
As shown in Figure 8, in contrast to the known reversible FGFR inhibitor 1 that dissociate quickly
ACS Paragon Plus Environment

Page 17 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
allowing recovery of enzyme activity, compounds 10 and 15 markedly prevent recovery of enzyme
activity, strongly supporting the irreversible binding mode of compounds 10 and 15 to FGFR1.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Figure 8. Compound 10 (A) and 15 (B) irreversibly binds to FGFR1. Enzyme activity of FGFR1
was assayed by a Caliper EZ Reader under three different conditions: without the enzyme
(background), without the compound (non-pre-incubation control), and pre-incubated with the
compound. “Conversion” here represents the enzyme activity and means “the percent of
conversion of the substrate peptide”.
ꢀ
Then, we conducted traditional kinetics analysis measuring Kinact/K which defines the
i
5
0
second order rate constant for covalent binding to the target protein. The efficient covalent target
engagement is apparent for compound 10 and 15 to FGFR1 with a calculated Kinact/K ratio of 0.38,
i
−
1
−1
0.070 (μM s ), respectively (Table 3). This K /K value was comparable to some EGFR
inact
i
covalent inhibitors or FGFR covalent inhibitor.51, 52 In addition, compared with compound 10 (Ki
1.9 nM), weaker affinity for compound 15 to FGFR1 (K = 10 nM) was observed, which is
=
i
consistent with the biochemical and cellular IC50 data (Table 2). The kinetics studies also
suggested covalent irreversible binding of the compound 10 and 15 to FGFR.
Table 3. Kinetics values of compounds 10 and 15 binding to FGFR1.
Compound
ID
K
(s )
Kinact/Ki
(μM s )
inact
K (nM)
i
−
1
−1 −1
1
1
0
5
7.2 × 10-4
7.1 × 10-4
1.9
10
0.38
0.070
Protein kinase profiling. To better understand the general kinase target selectivity trends
and potential toxicity poisonousness of compound 15, the commercial KinaseProfiler Service
Eurofins Scientific, Inc.) was utilized to evaluate the selectivity of 15 against a panel of 405
(
kinases at three fixed concentrations (10 nM, 100 nM and 1000 nM). The KinaseProfiler assay
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 18 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
protocols measure the percent inhibition of phosphorylation of a peptide substrate in the presence
of a fixed concentration of ATP (10 µM). Remarkably, compound 15 showed high specificity for
FGFR kinases over all other kinases tested, and no appreciable inhibition of the other kinases was
observed in the presence of 10 nM, with most maintaining >70% of their control activity. At a
concentration of 100 nM, compound 15 inhibited a small number of kinases other than FGFRs,
including c-RAF, SIK, DDR1, Arg, Yes, Flt1, and DNA-PK. The location of the profiled kinases
and active interactions depicted in a kinase dendrogram tree is shown in Figure 9. Details of
profiles including all the data from the Eurofins Kinaseprofiler (405 targets) are available in
Supporting Information. This experiment further supported that compound 15 has high selectivity
for the FGFR family.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
ACS Paragon Plus Environment

Page 19 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
53
Figure 9. Kinase profile of compound 15 drawn with KinMap. Kinase inhibition ratios above 50%
at corresponding concentrations are marked as red circles.
In vivo PK evaluation. The in vivo PK evaluation of compound 15 was performed in rats
via IV (10 mg/kg), PO (20 mg/kg) and IP (20 mg/kg). Compared with compound 10, the PK profile
of compound 15 shows a significant improvement. As summarized in Table 4, when dosing via
IV, the plasma exposure of compound 15 was AUC = 4412.8 ng·h/mL that was four times of
0
-t
compound 10, and the half-life period of compound 15 was 8.55 h. When dosing via IP, the plasma
exposure was AUC = 16790.8 ng·h/mL and the half-life period was 4.63 h. Unfortunately, when
0-t
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 20 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
dosing via PO, the bioavailability of 15 was almost zero. The results revealed that 15 has an
acceptable PK profile only when dosing via IV and IP.
Table 4. Pharmacokinetic properties of compounds 10 and 15 in SD rats.
Dose
(mg/kg)
Tmax
(h)
Cmax
AUC0-t
AUC0-∞
MRT
(h)
t1/2
(h)
CL
Vss
F
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
(ng/mL) (ng·h/mL) (ng·h/mL)
(L/h/kg) (L/kg) (%)
IV 10 0.083 1878.0
1093.9
6.4
1272.4
6.4
5.04 6.25
2.17 1.77
8.60
--
--
--
--
1
1
0
5
PO 20 0.861
12.9
0.29
IV 10
PO 20
IP 20
--
--
4308.7
32.6
4412.8
51.5
1.64 8.55
7.33 21.64
2.27
--
28.0
--
--
~0
--
0.5
0.5
7.3
5042.5 16790.8
16984.3 3.30 4.63
--
--
Species difference study. To understand the metabolic species difference of 15, we used an
in vitro model of hepatic subcellular fractions to incubate compounds and analyzed the metabolites
with UPLC-UV/Q-TOF MS. The metabolites were ranked in ascending order by their mass-to-
charge ratio or their LC retention time when the metabolites had the same mass-to-charge ratio.
The results of the metabolism of 15 by hepatic subcellular fractions (Table S4) show that 15 was
comparatively stable that after 180 min of incubation, and only 20% of the parent compounds was
metabolized. The main metabolic pathways of 15 are amino oxidation, GSH binding and amino
demethylation. 15 is more likely to bind GSH in human and monkey hepatic subcellular fractions
than it is in dog, rat and mouse. The proposed metabolic pathways of 15 in the hepatics subcellular
fractions are shown in Figure S3. Species differences suggested that there are few variations
among hepatic metabolic pathways of 15 in dogs, rats, mice, monkeys and humans. In addition,
the metabolism of 15 in human hepatic subcellular fractions is simpler than those of other species.
hERG inhibitory activity. A blockade of the hERG channel may cause drug-induced
prolongation of the QT interval, which has become a major concern in drug discovery and
development. To evaluate the cardiotoxicities of the candidates, we used a patch-clamp experiment
5
4
to determine the hERG inhibitory activities of compounds 9, 10, 13 and 15, with cisapride (19)
as the positive control. As shown in Table 5, the hERG IC values of most of the compounds are
5
0
highger than 10 µM, and compounds 10 and 15 exhibited hERG IC values higher than 40 µM.
5
0
ACS Paragon Plus Environment

Page 21 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
55, 56
In general, a safe drug candidate is expected to have an hERG IC value greater than 10 µM,
5
0
so compound 15 has a relatively low risk of hERG toxicity.
Table 5. hERG inhibitory activity evaluation.
Compound
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
9
10
13
15
19
ID
hERG IC50
1
9.53
> 40 32.85 > 40
0.03
(µM)
CONCLUSION
CADD has the potential to improve the efficiency of drug research and discovery to assist
or accelerate decision making throughout the entire development process. Although an increasing
number of successful drug discovery cases involving CADD methods are being reported,57, 58 the
shortcomings of CADD methods are also obvious. For instance, due to inaccurate or inapplicable
5
9
computational models, the false positive ratio in virtual screenings remains high, and it is still of
challenge to reliably prioritize synthetic efforts to focus on the optimization of compounds with
favorable activities as well as absorption, distribution, metabolism, excretion, and toxicity
properties. The rational application of CADD methods requires experimental researchers to better
understand the limitations of computational models and choose the appropriate models by
comprehensively considering their computation costs and the expected accuracy level. In this study,
we first devised an FGFR-targeted scoring function, the RTKscore, to address the inaccuracy
associated with activity rank-ordering of virtual screenings. Based on the RTKscore, we
successfully identified compound 8 from a commercial chemical library, and this compound has a
novel structure and exhibits sub-nanomolar inhibitory activity (FGFR1 IC = 114.5±15.3 nM)
5
0
against FGFR. Structure-guided optimization then led to compound 9 with improved molecular
and whole-cell activities against FGFR (FGFR1 IC = 10.8±1.3 nM, KG1 IC = 554.4±447.2
50
50
nM), and covalent compound 10 with subnanomolar inhibitory activity toward FGFRs and FGFR-
dependent cancer cell lines. Molecular biology experiments, X-ray crystallography and protein
mass spectrometry studies showed that 10 could selectively target FGFR and form a covalent bond
with the Cys residue in the loop of the FGFRs. Despite its excellent in vitro activity, compound 10
showed a poor PK profile due to its low solubility in water and metabolic instability issues. To
address these problems, we further used site-of-metabolism prediction to identify the metabolically
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 22 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
labile sites on compound 10 and successfully obtained a more soluble and stable compound, 15.
In vivo PK experiments indeed showed that 15 has high plasma exposure and an acceptable t1/2
when dosing via IP and IV, despite that there was no improvement in oral exposure. Moreover,
the hERG inhibitory activity evaluation experiment indicated that 15 has a relatively low risk of
hERG toxicity. These results indicated that compound 15 could be a promising FGFR lead
compound. Taken together, this study describes our efforts in rational design of novel FGFR
inhibitors. Accordingly, further medicinal chemistry exploration around compound 15 is currently
ongoing with the aim to provide validated candidates for the novel anticancer treatments.
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Experimental Section
RTKscore development
The protocol for the development of the RTKscore is similar to that of SAMscore development.
First, we collected the crystal structures of FGFR1 (PDB id: 3TT0), FGFR2 (PDB id: 2PVF), and FGFR3
(PDB id: 4K33) from PDB, and the ligand information for these proteins from the BindingDB, and these
data point included the ligand structure and activity information with the corresponding target protein. A
total of 1,090 ligands were collected. The activity data of these ligands were converted to pIC50 (푝퐼퐶50
=
9 ― 푙표푔 (퐼퐶 )), and the ligands without explicit data value were marked as inactive with pIC50 = 0.
10 50
We divided all the ligands into two groups, the training set and the test data in a ratio of 5:1, in which the
training data set had 908 ligands and the test data set had 182 ligands.
The molecular docking program Glide was used to generate the binding modes of each ligand with
the corresponding target protein. Before molecular docking, the ligands were prepared by the LigPrep
module in Schrödinger software (LigPrep, version 3.4; Schrödinger, LLC: New York, NY, 2015), and the
protein structures were optimized by the module of Protein Preparation Wizard module in the Maestro
program (Maestro, version 10; Schrödinger, LLC: New York, NY, 2015). Based on the optimized protein
structures, protein Grid files were generated with the center of the ATP binding site by the Receptor Grid
Generation module in the Glide program (Schrödinger, LLC: New York, NY, 2015). The prepared ligand
conformations were docked to the corresponding target protein Grid files by Glide with the SP precision
mode. All parameters for the above processes were the default parameters. The conformation of the lowest
Glide Emodel score for each ligand was selected as the pose of the ligand for further study.
Based on the binding mode of the ligands, the iPMF features for each ligand were calculated with
in-house Python scripts for each ligand. Seventeen types of atoms in the proteins, 30 types of atoms in the
ligand, and 11 distance ranges (0.0-2.0 Å, 0.0 -3.0 Å, …, and 0.0 – 12.0 Å) were considered. Overall, there
were 5610 iPMF features for each ligand. Then, epsilon support vector regression (ε-SVR) and recursive
feature elimination (RFE) methods were used to perform feature selection, and the unimportant features
were eliminated step by step until the optimal feature sub set of features was identified. The importance of
each feature was evaluated by the square of the Pearson correlation coefficient. Fifty unimportant features
were eliminated every time, and the data set with sub features were trained with the ε-SVR method and
evaluated by 5-fold cross validation (CV). Finally, the model with 1125 features, the RTKscore, was
obtained. The RTKscore was compared with Glide on the test data set, and the squares of the correlation
efficient of the ligands’ pIC values with the score predicted RTKscore or Glide was calculated.
5
0
Virtual screening
ACS Paragon Plus Environment

Page 23 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
The SPECS database, containing 207, 163 compounds, was used as the ligand database for virtual
screening. First, we removed compounds containing inorganic atoms, PAINS substructures, low
draggability or reactive substructures. After filtering, the remaining 190,038 compounds were prepared
using the same protocol used for RTKscore development for virtual screening. The prepared protein
structure of FGFR1 (PDB id: 3TT0) from RTKscore development was chosen as the receptor structure for
virtual screening. The prepared ligands were docked to the receptor protein structure with Glide, and 5
conformations of each ligand were outputted. Next, we used the RTKscore to evaluate each ligand pose,
which included iPMF features calculation, feature selection, and score prediction. According to ligands’
scores predicted by the RTKscore, the top 1000 ligands were kept for further analysis. We clustered the
ligands into 100 clusters based on fingerprint ECFP4 with software Pipeline Pilot v7.5 software. To ensure
selected ligands were structurally diverse, we selected one or two compounds from each cluster. Finally,
there were 112 compounds were selected and purchased for biochemical evaluations.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Molecular modeling
For noncovalent inhibitors, the binding mode of the ligand and protein was simulated with the
molecular docking program Glide. The protocol of the molecular docking was similar to that mentioned
above, and we outputted all conformations of the ligand and checked the conformations manually.
60
For covalent inhibitors, we simulated their binding mode with CovDock, a covalent docking program
from Schrödinger (Schrödinger, LLC: New York, NY, 2015). As the residue Cys488 of the FGFR1 crystal
structure (PDB id: 3TT0) was mutated to an alanine, we first mutated the residue 488 back to cystine, and
then the protein structure and ligands were prepared with the same protocol as described above. After
choosing Cys488 of FGFR1 as the reaction residues of the target protein, and the reaction type was set as a
Michael addition reaction. Other parameters were left at the default settings.
General chemistry information
As shown in schemes 1, we developed an efficient route to obtain derivatives 8-16. Herein, we use
2
’-hydroxyacetophenone derivatives 17a-d as the starting materials (Scheme 1) and treated them with ethyl
bromoacetate and K CO in DMF to obtain the intermediates 19a-d. Subsequent treatment with 21a-c to
2
3
afforded the pyrazole intermediates 22a-f. Compounds 8, 9, and 23a-d were synthesized by the reaction of
intermediates 22a-f and hydrazine hydrate in EtOH under one-pot conditions. The reductions of compounds
23a-c with Fe powder afforded the aniline derivatives, which were ideally poised to undergo acylation with
the corresponding acyl chlorides to afford compounds 10 and 12-16. The Cbz group of 23d was deprotected
by Pd(OH) /H in MeOH to generate the amine intermediate, which could then undergo acylation with
2
2
acryloyl chloride to afford compound 11.
a
Scheme 1
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 24 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
O
R1
R1
R1
a
O
O
CN
b
R1
OH
O
O
O
1
7a-d
19a,
19b,
1
R = H
1
R = H, Me, Cl, Br
18a-d
1
O
R = Me
1
1
9c, R = Cl
d
NC
1
1
9d, R = Br
R3
N
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
O
N
NH2
H
N
Cl
O
c
R3
N
1
3
O
22a, R = H, R = H
1
3
O
22b, R = Me, R = H
3
Et
1
3
R3
21a,
22c, R = Me, R = NO
2
R = H
3
1
3
21b, R = NO
22d, R = Cl, R = NO2
2
2
0a-c
1
3
2
2e, R = Br, R = NO
2
1c, R3
NHCbz
1
3
2
=
22f, R = Me, R =
NHCbz
R1
R1
O
NH2
R2
O
f, g
N
N
NH
e
O
HN
N
NH2
O
10, 12~16
N
N
R3
HN
N
g, h
O
1
3
8
9
2
2
, R = H, R = H
NH2
O
1
3
O
, R = Me, R = H
1
3
HN
3a, R = Me, R = NO
N
2
N
1
3
3b, R = Cl, R = NO
HN
2
N
1
3
23c, R = Br, R = NO
2
O
1
3
11
23d, R = Me, R =
NHCbz
a
o
o
Reagents and conditions: (a) Ethyl bromoacetate, K CO , DMF, 100 C; (b) CH CN, NaH, THF, 50 C;
c) Ethyl 2-chloroacetoacetate, NaNO , NaOAc, EtOH, H O/HCl (3:1, v/v), 0 C-rt; (d) Et N, DCM, rt; (e)
N H H O, HCl (conc.), EtOH, 100 C, MW; (f) For compunds 23a-c, Fe, HCl (conc.), MeOH/H O (6:1,
v/v), 80 C, 1 h; (g) Acryloyl chloride or acryloyl chloride amino derivatives, K CO , THF, rt; (h) For
2
3
3
o
(
2
2
3
.
o
2
4
2
2
o
2
3
o
compund 23d, Pd(OH) /C, HCl (conc.), MeOH, 40 C, 8 h.
2
General methods.
The reagents (chemicals) were purchased and used without further purification. Nuclear magnetic
resonance (NMR) spectroscopy was performed on a Bruker AMX-400 and AMX-300 NMR (TMS as the
IS). Chemical shifts are reported in parts per million (ppm, δ) downfield from TMS. Proton coupling
patterns are described as singlet (s), doublet (d), triplet (t), quartet (q), and multiplet (m). Low- and high-
resolution mass spectra (LRMS and HRMS) were acquired with electron impact, electrospray, and matrix-
assisted laser desorption ionization (EI, ESI, and MALDI) analyses on a Finnigan MAT-95, LCQ-DECA
spectrometer and an IonSpec 4.7 T instrument. HPLC analyses of all final compounds for biological testing
were carried out on an Agilent 1260 Series HPLC with an Agilent Extend-C18 column (150×4.6 mm, 5
µm) (Table S5). All final compounds achieved a minimum of 95% purity. The known acryloyl chloride
6
1
amino derivatives were already prepared.
4
-Amino-3-(3-methylbenzofuran-2-yl)-2-phenyl-2,6-dihydro-7H-pyrazolo[3,4-d]pyridazin-7-one (8).
To a solution of 22a (500 mg, 1.35 mmol) in EtOH (5 mL) was added hydrazine hydrate (472 mg,
.42 mmol). The mixture was stirred at 60 °C overnight. The reaction mixture was diluted with water (20
9
ACS Paragon Plus Environment

Page 25 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
mL). The reaction mixture was then extracted with DCM (3 × 30 mL), and the combined organic layers
were dried over Na SO and concentrated under reduced pressure. The residue was dissolved in EtOH (5
mL) and added few drops HCl (con.). The mixture was stirred at 100 °C for 2 h under microwave irradiation.
The reaction mixture was then concentrated under reduced pressure and purified by column
2
4
chromatography using silica with 10% MeOH in DCM as eluent to give 8 (210 mg, 32% yield), as a yellow
1
solid. H NMR (500 MHz, DMSO-d ) δ 11.48 (s, 1H), 7.66 (d, J = 7.5 Hz, 1H), 7.59 (d, J = 8.3 Hz, 1H),
6
1
3
7
1
1
3
.50 – 7.39 (m, 6H), 7.36 – 7.30 (m, 1H), 5.20 (s, 2H), 1.92 (s, 3H). C NMR (126 MHz, DMSO) δ 155.00,
54.47, 143.19, 143.11, 138.92, 137.37, 129.44, 128.61, 127.63, 126.19, 124.65, 123.22, 120.64, 119.85,
14.82, 111.54, 8.12. LRMS (ESI, m/z): 358.0 [M+H] ; HRMS (ESI) m/z: cacld for C H N O ([M+H] ):
58.1299; found: 358.1290.
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
+
+
2
0
15
5
2
4-Amino-3-(3,5-dimethylbenzofuran-2-yl)-2-phenyl-2,6-dihydro-7H-pyrazolo[3,4-d]pyridazin-7-one
(9).
1
Compound 9 was prepared in a similar manner as described for compound 8. Yield: 31%. H NMR
(500 MHz, DMSO-d ) δ 11.49 (s, 1H), 7.52 – 7.36 (m, 7H), 7.26 – 7.17 (m, 1H), 5.19 (s, 2H), 3.07 (s, 3H),
6
1
3
1.87 (s, 3H). C NMR (126 MHz, DMSO-d ) δ 155.02, 152.94, 143.21, 143.10, 138.92, 137.43, 132.44,
6
1
29.46, 129.41, 128.67, 127.74, 127.39, 124.60, 120.25, 119.59, 114.76, 111.12, 20.86, 8.11. LRMS (ESI,
+
+
m/z): 372.0 [M+H] ; HRMS (ESI) m/z: cacld for C H N O ([M+Na] ): 394.1274; found: 394.1272.
2
1
17
5
2
N-(4-(4-Amino-3-(3,5-dimethylbenzofuran-2-yl)-7-oxo-6,7-dihydro-2H-pyrazolo[3,4-d]pyridazin-2-
yl)phenyl)acrylamide (10).
To a solution of 23a (500 mg, 1.2 mmol) in MeOH/H O (6:1, 14 mL) was added HCl (0.5 mL) and
2
Fe powder (268 mg, 4.8 mmol). The mixture was stirred at 80 °C for 6 h. The reaction mixture was filtered
and the filtrate was concentrated under reduced pressure. The residue was dissolved in THF (5 mL) and
added K CO (248.8 mg, 1.8 mmol). Then a solution of acryloyl chloride (217.2 mg, 2.4 mmol) in THF (5
2
3
mL) was added to the reaction mixture dropwise at rt. Until TLC showed the reaction completed,
the reaction mixture was then concentrated under reduced pressure and purified by column chromatography
1
using silica with 10% MeOH in DCM as eluent to give 10 (120 mg, 23% yield), as a yellow solid. H NMR
(
400 MHz, DMSO-d ) δ 11.48 (s, 1H), 10.95 (s, 1H), 7.82 (d, J = 8.9 Hz, 2H), 7.55 – 7.32 (m, 4H), 7.22
6
(
d, J = 8.9 Hz, 1H), 6.59 (dd, J = 17.0, 10.2 Hz, 1H), 6.25 (d, J = 17.0 Hz, 1H), 5.75 (d, J = 10.2 Hz, 1H),
1
3
5.19 (s, 2H), 2.41 (s, 3H), 1.89 (s, 3H). C NMR (126 MHz, DMSO-d ) δ 163.55, 155.00, 152.93, 143.17,
6
142.91, 140.04, 137.48, 133.82, 132.40, 131.78, 128.67, 127.58, 127.34, 127.15, 125.05, 120.25, 119.47,
+
114.61, 111.11, 20.86, 8.14. LRMS (ESI, m/z): 441.9 [M+H] ; HRMS (ESI) m/z: cacld for C H N O
([M+H] ): 441.1670; found: 441.1681.
2
4
21
6
3
+
N-(4-(4-Amino-3-(3,5-dimethylbenzofuran-2-yl)-7-oxo-6,7-dihydro-2H-pyrazolo[3,4-d]pyridazin-2-
yl)benzyl)acrylamide (11).
To a solution of 23d (540 mg, 1.0 mmol) in MeOH (5 mL) was added HCl (1 mL) and Pd(OH) /C
2
(
20%, 35 mg, 0.05 mmol). The mixture was stirred at 40 °C overnight under H atmosphere. The reaction
2
mixture was filtered and the filtrate was concentrated under reduced pressure. The residue was dissolved in
THF (5 mL) and added K CO (207.3 mg, 1.5 mmol). Then a solution of acryloyl chloride (181 mg, 2
2
3
mmol) in THF (5 mL) was added to the reaction mixture dropwise at rt. Until TLC showed the reaction
completed, the reaction mixture was then concentrated under reduced pressure and purified by column
chromatography using silica with 10% MeOH in DCM as eluent to give 11 (99 mg, 21% yield), as a yellow
1
solid. H NMR (400 MHz, DMSO-d ) δ 11.51 (s, 1H), 8.82 (t, J = 5.9 Hz, 1H), 7.48 (d, J = 8.4 Hz, 1H),
6
7
6
.45 – 7.38 (m, 3H), 7.32 (d, J = 8.5 Hz, 2H), 7.22 (d, J = 8.6 Hz, 1H), 6.30 (dd, J = 17.1, 10.2 Hz, 1H),
.12 (dd, J = 17.1, 2.0 Hz, 1H), 5.61 (dd, J = 10.2, 2.0 Hz, 1H), 5.19 (s, 2H), 4.37 (d, J = 6.0 Hz, 2H), 2.40
1
3
(s, 3H), 1.89 (s, 3H). C NMR (126 MHz, DMSO-d ) δ 164.88, 155.09, 153.01, 143.26, 143.06, 140.81,
6
137.66, 137.42, 132.52, 131.52, 128.70, 128.06, 127.70, 127.46, 125.78, 124.56, 120.32, 119.69, 114.86,
+
111.22, 41.54, 20.91, 8.19. LRMS (ESI, m/z): 456.0 [M+H] ; HRMS (ESI) m/z: cacld for C H N O
([M+H] ): 455.1826; found: 455.1837.
2
5
23
6
3
+
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 26 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
N-(4-(4-Amino-3-(5-chloro-3-methylbenzofuran-2-yl)-7-oxo-6,7-dihydro-2H-pyrazolo[3,4-
d]pyridazin-2-yl)phenyl)acrylamide (12)
1
Compound 12 was prepared in a similar manner as described for compound 10. Yield: 20%. H
NMR (400 MHz, DMSO-d ) δ 11.50 (s, 1H), 10.43 (s, 1H), 7.82 – 7.69 (m, 3H), 7.64 (d, J = 8.8 Hz, 1H),
6
7
5
1
1
.45 – 7.38 (m, 3H), 6.43 (dd, J = 17.0, 10.1 Hz, 1H), 6.26 (d, J = 17.0 Hz, 1H), 5.78 (d, J = 11.2 Hz, 1H),
1
3
.23 (s, 2H), 1.95 (s, 3H). C NMR (126 MHz, DMSO-d ) δ 163.45, 154.95, 153.03, 143.17, 142.99,
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
39.85, 139.00, 133.88, 131.53, 130.38, 127.71, 127.56, 126.98, 125.98, 125.23, 120.29, 119.79, 119.57,
+
14.86, 113.23, 8.17. LRMS (ESI, m/z): 461.0 [M+H] ; HRMS (ESI) m/z: cacld for C H N O Cl
2
3
17
6
3
+
([M+Na] ): 483.0943; found: 483.0952.
(E)-N-(4-(4-Amino-3-(3,5-dimethylbenzofuran-2-yl)-7-oxo-6,7-dihydro-2H-pyrazolo[3,4-
d]pyridazin-2-yl)phenyl)-4-(dimethylamino)but-2-enamide (13).
1
Compound 13 was prepared in a similar manner as described for compound 10. Yield: 19%. H
NMR (400 MHz, DMSO-d ) δ 11.48 (s, 1H), 10.61 (s, 1H), 7.80 – 7.71 (m, 2H), 7.49 – 7.37 (m, 4H), 7.23
6
(dd, J = 8.5, 1.6 Hz, 1H), 6.76 (d, J = 15.4 Hz, 1H), 6.38 (d, J = 15.4 Hz, 1H), 5.18 (s, 2H), 2.41 (s, 3H),
13
2
1
4
.36 (s, 6H), 1.89 (s, 3H). C NMR (126 MHz, DMSO-d ) δ 163.21, 155.08, 152.99, 143.25, 142.97,
6
39.95, 137.51, 133.94, 132.49, 128.72, 127.67, 127.41, 125.21, 120.29, 119.55, 114.67, 111.17, 58.66,
4.04, 20.91, 8.20. HRMS (ESI) m/z: cacld for C H N O ([M-H] ): 496.2103; found: 496.2109.
-
2
7
26
7
3
(E)-N-(4-(4-Amino-3-(3,5-dimethylbenzofuran-2-yl)-7-oxo-6,7-dihydro-2H-pyrazolo[3,4-
d]pyridazin-2-yl)phenyl)-4-morpholinobut-2-enamide (14).
1
Compound 14 was prepared in a similar manner as described for compound 10. Yield: 9%. H
NMR (400 MHz, DMSO-d ) δ 11.48 (s, 1H), 10.41 (s, 1H), 7.74 (d, J = 9.0 Hz, 2H), 7.49 (d, J = 8.4 Hz,
6
1
H), 7.44 (s, 1H), 7.41 – 7.36 (m, 2H), 7.23 (dd, J = 8.5, 1.6 Hz, 1H), 6.79 – 6.70 (m, 1H), 6.33 (d, J = 15.1
1
3
Hz, 1H), 5.19 (s, 2H), 3.72 – 3.57 (m, 4H), 3.29 – 3.11 (m, 2H), 2.49 – 2.30 (m, 7H), 1.90 (s, 3H). C NMR
126 MHz, DMSO-d ) δ 162.32, 155.00, 152.93, 143.16, 142.93, 139.88, 137.48, 133.87, 132.40, 128.69,
(
6
1
27.59, 127.33, 125.18, 120.23, 119.51, 119.46, 114.62, 111.12, 66.00, 52.99, 45.66, 20.86, 8.15. HRMS
-
(ESI) m/z: cacld for C H N O ([M-H] ): 538.2208; found: 538.2211.
29 28 7 4
(E)-N-(4-(4-Amino-3-(5-chloro-3-methylbenzofuran-2-yl)-7-oxo-6,7-dihydro-2H-pyrazolo[3,4-
d]pyridazin-2-yl)phenyl)-4-(dimethylamino)but-2-enamide (15).
1
Compound 15 was prepared in a similar manner as described for compound 10. Yield: 25%. H
NMR (400 MHz, DMSO-d ) δ 11.49 (s, 1H), 10.70 (s, 1H), 7.80 – 7.73 (m, 3H), 7.64 (d, J = 8.8 Hz, 1H),
6
7
2
1
5
.49 – 7.31 (m, 3H), 6.83 – 6.73 (m, 1H), 6.47 (d, J = 15.4 Hz, 1H), 5.23 (s, 2H), 3.92 (d, J = 6.8 Hz, 2H),
.75 (s, 6H), 1.95 (s, 3H). C NMR (126 MHz, DMSO) δ 162.32, 154.92, 153.03, 143.14, 142.99, 139.67,
13
38.97, 134.05, 131.54, 130.38, 127.69, 126.98, 125.97, 125.27, 120.28, 119.79, 119.68, 114.85, 113.22,
+
6.64, 41.92, 8.16. HRMS (ESI) m/z: cacld for C H ClN O ([M+H] ): 518.1702; found: 518.1707.
2
6
25
7
3
(E)-N-(4-(4-Amino-3-(5-bromo-3-methylbenzofuran-2-yl)-7-oxo-6,7-dihydro-2H-pyrazolo[3,4-
d]pyridazin-2-yl)phenyl)-4-(dimethylamino)but-2-enamide (16).
Compound 16 was prepared in a similar manner as described for compound 10. Yield: 14%. H NMR (400
1
MHz, DMSO-d ) δ 11.50 (s, 1H), 11.02 (s, 1H), 7.92 (d, J = 5.1 Hz, 1H), 7.82 (d, J = 9.0 Hz, 2H), 7.58 –
6
7
=
.52 (m, 2H), 7.42 – 7.34 (m, 2H), 6.86 – 6.77 (m, 1H), 6.55 (d, J = 15.4 Hz, 1H), 5.23 (s, 3H), 3.73 (d, J
1
3
5.7 Hz, 2H), 2.61 (s, 6H), 1.94 (s, 3H). C NMR (126 MHz, DMSO-d ) δ 162.77, 155.01, 153.42, 143.23,
6
1
1
43.00, 140.00, 138.80, 133.90, 130.99, 128.71, 127.01, 125.20, 124.80, 123.33, 121.59, 119.72, 119.64,
+
15.59, 114.88, 113.72, 57.27, 42.51, 8.19. HRMS (ESI) m/z: cacld for C H N O Br ([M+H] ): 562.1197;
26 25 7 3
found: 562.1210.
Ethyl 3-methylbenzofuran-2-carboxylate (18a).
ACS Paragon Plus Environment

Page 27 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
To a solution of 2’-hydroxyacetophenone 17a (2.0 g, 14.4 mmol) in dry DMF (20 mL) was added
K CO (2.98 g, 21.6 mmol) and ethyl bromoacetate (2.88 g, 17.3 mmol), respectively. The mixture was
2
3
o
stirred at 100 C overnight under Ar. The reaction mixture was then cooled to rt. The suspension was filtered
through a Celite cartridge, and the cartridge rinsed with EA. The filtrate was diluted with EA and water.
The reaction mixture was then extracted with EA (3 × 40 mL), and the combined organic layers were dried
over Na SO and concentrated under reduced pressure. The residue was purified by column
2
4
chromatography on silica gel with using 5% EA in hexane as eluent to give 18a (1.8 g, 61% yield), as a
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
white solid. H NMR (400 MHz, CDCl ) δ 7.64-7.62 (m, 1H), 7.56-7.52 (m, 1H), 7.46-7.42 (m, 1H), 7.32-
3
7
.26 (m, 1H), 4.45 (q, J = 7.1 Hz, 2H), 2.59 (s, 3H), 1.44 (t, J = 7.1 Hz, 3H).
Ethyl 3,5-dimethylbenzofuran-2-carboxylate (18b).
1
Compound 12b was prepared in a similar manner as described for compound 11a. Yield: 41%. H
NMR (400 MHz, CDCl ) δ 7.45 – 7.34 (m, 2H), 7.27 – 7.21 (m, 1H), 4.44 (q, J = 7.1 Hz, 2H), 2.56 (s, 3H),
3
2.46 (s, 3H), 1.44 (t, J = 7.1 Hz, 3H).
Ethyl 5-chloro-3-methylbenzofuran-2-carboxylate (18c).
1
Compound 12c was prepared in a similar manner as described for compound 11a. Yield: 43%. H
NMR (400 MHz, CDCl ) δ 7.60 (d, J = 2.1 Hz, 1H), 7.47 (d, J = 8.8 Hz, 1H), 7.39 (dd, J = 8.8, 2.1 Hz,
3
1
H), 4.45 (q, J = 7.1 Hz, 2H), 2.56 (s, 3H), 1.44 (t, J = 7.1 Hz, 3H).
Ethyl 5-bromo-3-methylbenzofuran-2-carboxylate (18d).
1
Compound 12d was prepared in a similar manner as described for compound 11a. Yield: 45%. H
NMR (400 MHz, CDCl ) δ 7.76 (d, J = 1.8 Hz, 1H), 7.52 (dd, J = 8.8, 1.8 Hz, 1H), 7.42 (d, J = 8.8 Hz,
3
1
H), 4.45 (q, J = 7.1 Hz, 2H), 2.55 (s, 3H), 1.44 (t, J = 7.1 Hz, 3H).
3
-(3-Methylbenzofuran-2-yl)-3-oxopropanenitrile (19a).
To a solution of ethyl 3-methyl-2-benzofurancarboxylate 11a (2 g, 9.79 mmol) and acetonitrile (1.0
mL, 19.6 mmol) in dry THF (25 mL) was added sodium hydride (60%, 0.51 g, 12.7 mmol) portionwise.
o
The resulting mixture was stirred at 50 C for 8 h under Ar, and then allowed to cool to room temperature.
The mixture was quenched with saturated ammonium chloride solution (50 mL) and then was acidified
with 1N HCl solution until pH 5-6. The resulting precipitate was filtered, washed with water, and dried to
afford 12a (1.3 g) as a pale-yellow solid. The compound was used without further purification. LRMS (ESI,
-
m/z): 198.1 [M-H] .
3
-(3,5-Dimethylbenzofuran-2-yl)-3-oxopropanenitrile (19b)
Compound 19b was prepared in a similar manner as described for compound 19a. Yield: 73%.
-
LRMS (ESI, m/z): 212.0 [M-H] .
3
-(5-Chloro-3-methylbenzofuran-2-yl)-3-oxopropanenitrile (19c)
Compound 19c was prepared in a similar manner as described for compound 19a. Yield: 45%.
-
LRMS (ESI, m/z): 233.0 [M-H] .
3
-(5-Bromo-3-methylbenzofuran-2-yl)-3-oxopropanenitrile (19d)
Compound 19d was prepared in a similar manner as described for compound 19a. Yield: 38%.
-
LRMS (ESI, m/z): 277.1 [M-H] .
(
Z)-Ethyl 2-chloro-2-(2-phenylhydrazono)acetate (21a).
To a solution of aniline (5.0 g, 52.08 mmol) in 3M HCl (40 mL) at 0 °C was added a solution of
sodium nitrite (3.95 g, 57.3 mmol) in water (10 mL) dropwise. The mixture was stirred at 0°C for 1 h, then
a solution of NaOAc (4.7 g, 57.3 mmol) and ethyl 2-chloroacetoacetate (9.43 g, 57.3 mmol) in EtOH (30
mL) was added and the mixture was stirred at rt for 4 h. The reaction was quenched with water (50 mL),
ACS Paragon Plus Environment

Journal of Medicinal Chemistry
Page 28 of 47
1
2
3
4
5
6
7
8
9
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
then extracted with EA (3 × 40 mL), and the combined organic layers were dried over Na SO and
2
4
concentrated under reduced pressure. The residue was purified by column chromatography using silica with
1
1
0% EtOAc in hexane as eluent to give 21a (9 g, 76% yield), as a yellow solid. H NMR (400 MHz, CDCl )
3
δ 8.33 (s, 1H), 7.37 – 7.31 (m, 2H), 7.25 – 7.21 (m, 2H), 7.08 – 7.01 (m, 1H), 4.39 (q, J = 7.1 Hz, 2H), 1.41
(t, J = 7.1 Hz, 3H).
(Z)-Ethyl 2-chloro-2-(2-(4-nitrophenyl)hydrazono)acetate (21b)
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
Compound 21b was prepared in a similar manner as described for compound 21a. Yield: 82%。
1
H NMR (400 MHz, CDCl ) δ 8.62 (s, 1H), 8.30 – 8.17 (m, 2H), 7.35 – 7.28 (m, 2H), 4.41 (q, J = 7.1 Hz,
3
2H), 1.42 (t, J = 7.1 Hz, 3H).
(Z)-Ethyl 2-(2-(4-((((benzyloxy)carbonyl)amino)methyl)phenyl)hydrazono)-2-chloroacetate (21c).
Compound 21c was prepared in a similar manner as described for compound 21a. Yield: 60%。
+
LRMS (ESI, m/z): 413.0 [M+Na] .
Ethyl 4-cyano-5-(3-methylbenzofuran-2-yl)-1-phenyl-1H-pyrazole-3-carboxylate (22a).
To a solution of 19a (1.0 g, 4.6 mmol) in DCM (10 mL) was added NEt (1.86 g, 18.4 mmol) and
3
2
1a (1.25 g, 5.5 mmol), respectively. The mixture was stirred at rt for 5 h. The reaction mixture was diluted
with water (30 mL). The reaction mixture was then extracted with EA (3 × 30 mL), and the combined
organic layers were dried over Na SO and concentrated under reduced pressure. The residue was purified
2
4
by column chromatography using silica with 15% EA in hexane as eluent to give 22a (1.0 g, 56% yield),
1
as a yellow solid. H NMR (400 MHz, CDCl ) δ 7.51 (dd, J = 8.5, 5.3 Hz, 1H), 7.39 (s, 5H), 7.11 – 7.03
3
(m, 2H), 4.54 (q, J = 7.1 Hz, 2H), 2.26 (s, 3H), 1.48 (t, J = 7.1 Hz, 3H).
Ethyl 4-cyano-5-(3,5-dimethylbenzofuran-2-yl)-1-phenyl-1H-pyrazole-3-carboxylate (22b).
Compound 22b was prepared in a similar manner as described for compound 22a. Yield: 43%。
1
H NMR (400 MHz, CDCl ) δ 7.41 – 7.35 (m, 5H), 7.24 – 7.15 (m, 3H), 4.54 (q, J = 7.1 Hz, 2H), 2.46 (s,
3
3
H), 2.23 (s, 3H), 1.48 (t, J = 7.1 Hz, 3H).
Ethyl 4-cyano-5-(3,5-dimethylbenzofuran-2-yl)-1-(4-nitrophenyl)-1H-pyrazole-3-carboxylate (22c).
Compound 22c was prepared in a similar manner as described for compound 22a. Yield: 41%。
1
H NMR (400 MHz, DMSO-d ) δ 8.39 – 8.30 (m, 2H), 7.79 – 7.69 (m, 2H), 7.57 – 7.52 (m, 1H), 7.38 (d,
6
J = 8.5 Hz, 1H), 7.25 (dd, J = 8.5, 1.7 Hz, 1H), 4.45 (q, J = 7.1 Hz, 2H), 2.42 (s, 3H), 2.21 (s, 3H), 1.37 (t,
J = 7.1 Hz, 3H).
Ethyl 5-(5-chloro-3-methylbenzofuran-2-yl)-4-cyano-1-(4-nitrophenyl)-1H-pyrazole-3-carboxylate
(22d).
Compound 22d was prepared in a similar manner as described for compound 22a. Yield: 46%。
-
LRMS (ESI, m/z): 450.0 [M-H] .
Ethyl 5-(5-bromo-3-methylbenzofuran-2-yl)-4-cyano-1-(4-nitrophenyl)-1H-pyrazole-3-carboxylate
(
22e).
1
Compound 22e was prepared in a similar manner as described for compound 22a. Yield: 36%。 H
NMR (400 MHz, CDCl ) δ 7.52 – 7.47 (m, 2H), 7.40 – 7.37 (m, 1H), 7.30 – 7.26 (m, 2H), 7.23 (s, 1H),
3
7
.22 – 7.18 (m, 1H), 4.54 (q, J = 7.1 Hz, 2H), 2.47 (s, 3H), 2.29 (s, 3H), 1.48 (t, J = 7.1 Hz, 3H).
Ethyl 1-(4-((((benzyloxy)carbonyl)amino)methyl)phenyl)-4-cyano-5-(3,5-dimethylbenzofuran-2-yl)-
H-pyrazole-3-carboxylate (22f).
1
ACS Paragon Plus Environment

Page 29 of 47
Journal of Medicinal Chemistry
1
2
3
4
5
6
7
8
9
1
Compound 22f was prepared in a similar manner as described for compound 22a. Yield: 55%。 H
NMR (400 MHz, CDCl ) δ 7.41 – 7.27 (m, 11H), 7.24 – 7.16 (m, 2H), 5.12 (s, 2H), 4.54 (q, J = 7.1 Hz,
3
2
H), 4.40 (d, J = 6.0 Hz, 2H), 2.47 (s, 3H), 2.24 (s, 3H), 1.48 (t, 3H).
4
-Amino-3-(3,5-dimethylbenzofuran-2-yl)-2-(4-nitrophenyl)-2,6-dihydro-7H-pyrazolo[3,4-
d]pyridazin-7-one (23a).
Compound 23a was prepared in a similar manner as described for compound 8. Yield: 32%. H
NMR (400 MHz, DMSO-d ) δ 11.55 (s, 1H), 8.35 – 8.28 (m, 2H), 7.76 – 7.69 (m, 2H), 7.52 – 7.44 (m,
1
1
1
1
1
1
1
1
1
1
1
2
2
2
2
2
2
2
2
2
2
3
3
3
3
3
3
3
3
3
3
4
4
4
4
4
4
4
4
4
4
5
5
5
5
5
5
5
5
5
5
6
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
1
2
3
4
5
6
7
8
9
0
6
2
H), 7.25 (dd, J = 8.5, 1.3 Hz, 1H), 5.27 (s, 2H), 2.42 (s, 3H), 1.94 (s, 3H).
4-Amino-3-(5-chloro-3-methylbenzofuran-2-yl)-2-(4-nitrophenyl)-2,6-dihydro-7H-pyrazolo[3,4-
d]pyridazin-7-one (23b).
Compound 23b was prepared in a similar manner as described for compound 8. Yield: 44%. LRMS
+
(
ESI, m/z): 458.8 [M+Na] .
4
-Amino-3-(5-bromo-3-methylbenzofuran-2-yl)-2-(4-nitrophenyl)-2,6-dihydro-7H-pyrazolo[3,4-
d]pyridazin-7-one (23c).
Compound 23c was prepared in a similar manner as described for compound 8. Yield: 26%. LRMS
+
(
ESI, m/z): 480.8, 482.7 [M, M+2] .
Benzyl
(4-(4-amino-3-(3,5-dimethylbenzofuran-2-yl)-7-oxo-6,7-dihydro-2H-pyrazolo[3,4-
d]pyridazin-2-yl)benzyl)carbamate (23d).
1
Compound 23d was prepared in a similar manner as described for compound 8. Yield: 28%. H
NMR (400 MHz, DMSO-d ) δ 11.49 (s, 1H), 7.87 (t, J = 6.1 Hz, 1H), 7.49 (d, J = 8.4 Hz, 1H), 7.45 – 7.37
m, 3H), 7.35 – 7.28 (m, 6H), 7.23 (d, J = 8.5 Hz, 1H), 5.17 (s, 2H), 5.03 (s, 2H), 4.23 (d, J = 6.1 Hz, 2H),
6
(
2
.41 (s, 3H), 1.88 (s, 3H).
Kinase inhibition assay
FGFR1, FGFR2, FGFR3, FGFR4, VEGFR2, VEGFR1, PDGFRβ, Ret, c-Src and c-Met active
proteins were purchased from Eurofins. The kinase activities were assessed using ELISA assay. Briefly,
6-well plates were precoated with 20 μg/mL poly (Glu, Tyr) 4:1 (Sigma, St Louis, MO, USA) as a substrate.
A 50-μL aliquot of 10 μmol/L ATP solution diluted in kinase reaction buffer (50 mmol/L HEPES [pH 7.4],
9
5
1
0 mmol/L MgCl , 0.5 mmol/L MnCl , 0.2 mmol/L Na VO , and 1 mmol/L DTT) was added to each well;
μL of the indicated compound diluted in 1% DMSO (v/v) (Sigma, St. Louis, MO, USA) was then added
2 2 3 4
to each reaction well. DMSO (1%, v/v) was used as the negative control. The kinase reaction was initiated
by the addition of purified tyrosine kinase proteins diluted in 49 μL of kinase reaction buffer. After
incubation for 60 min at 37 °C, the plate was washed three times with phosphate-buffered saline (PBS)
containing 0.1% Tween 20 (T-PBS). Antiphosphotyrosine (PY99) antibody (100 μL; 1:500, diluted in 5
mg/mL BSA T-PBS) was then added. After a 30-min incubation at 37 °C, the plate was washed three times,
and 100 μL of horseradish peroxidase-conjugated goat anti-mouse IgG (1:1000, diluted in 5 mg/mL BSA
T-PBS) was added. The plate was then incubated at 37 °C for 30 min and washed 3 times. A 100-μL aliquot
of a solution containing 0.03% H O and 2 mg/mL o-phenylenediamine in 0.1 mol/L citrate buffer (pH 5.5)
was then added. The reaction was terminated by the addition of 50 μL of 2 mol/L H SO , causing a color,
and the plate was then analyzed using a multiwell spectrophotometer (SpectraMAX190, from Molecular
Devices, Palo Alto, CA, USA) at 490 nm. The inhibition rate (%) was calculated using the following
equation: [1-(A490/A490 control)] ×100%. The IC values were calculated from the inhibition curves in
2
2
2
4
5
0
two separate experiments.
Kinase kinetic inhibition assessment
ACS Paragon Plus Environment