Protein–ligand interaction-guided discovery of novel VEGFR-2 inhibitors

Article abstract of DOI:10.1080/07391102.2019.1635915

As an effective target in abnormal angiogenesis-related tumor treatment, VEGFR-2 has small-molecule inhibitors of various scaffolds being approved for treating diseases such as renal carcinoma, non-small cell lung cancer, etc. However, endogenous and acqu

Full text of DOI:10.1080/07391102.2019.1635915

Journal of Biomolecular Structure and Dynamics
ISSN: 0739-1102 (Print) 1538-0254 (Online) Journal homepage: https://www.tandfonline.com/loi/tbsd20
Protein-Ligand Interaction Guided Discovery of
Novel VEGFR-2 Inhibitors
Yanmin Zhang, Mingliang Zhang, Yuchen Wang, Yuanrong Fan, Xingye
Chen, Yan Yang, Yi Hua, Wuchen Xie, Tao Lu, Weifang Tang, Yadong Chen &
Haichun Liu
To cite this article: Yanmin Zhang, Mingliang Zhang, Yuchen Wang, Yuanrong Fan, Xingye Chen,
Yan Yang, Yi Hua, Wuchen Xie, Tao Lu, Weifang Tang, Yadong Chen & Haichun Liu (2019):
Protein-Ligand Interaction Guided Discovery of Novel VEGFR-2 Inhibitors, Journal of Biomolecular
Structure and Dynamics, DOI: 10.1080/07391102.2019.1635915
To link to this article: https://doi.org/10.1080/07391102.2019.1635915
Accepted author version posted online: 24
Jun 2019.
Submit your article to this journal
View Crossmark data
Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=tbsd20

Protein-Ligand Interaction Guided Discovery of
Novel VEGFR-2 Inhibitors
Yanmin Zhanga,1, Mingliang Zhangb,1, Yuchen Wanga, Yuanrong Fana, Xingye Chena,
Yan Yanga, Yi Huaa, Wuchen Xiea, Tao Lua,b,c, Weifang Tangb, Yadong Chena,* and
Haichun Liua,*
a Laboratory of Molecular Design and Drug Discovery, School of Science, China
Pharmaceutical University, 639 Longmian Avenue, Nanjing 211198, China.
b School of Science, China Pharmaceutical University, 639 Longmian Avenue,
Nanjing, Jiangsu 211198, China
c State Key Laboratory of Natural Medicines, China Pharmaceutical University, 24
Tongjiaxiang, Nanjing 210009, China.
1 The authors contributed equally to this work and should be considered as cofirst
authors.
*Corresponding Authors
For Haichun Liu: Tel.: +86-25-86185163. Fax: +86-25-86185182. E-mail:
tuna888@126.com.
For Yadong Chen: Tel.: +86-25-86185163. Fax: +86-25-86185182. E-mail:
ydchen@cpu.edu.cn.

ABSTRACT
As an effective target in abnormal angiogenesis-related tumor treatment, VEGFR-2
has small-molecule inhibitors of various scaffolds being approved for treating
diseases like renal carcinoma and non-small cell lung cancer etc. However,
endogenous and acquired drug resistance are still considered to be the main
contributors for the failure of VEGFR-2 clinical candidates. Therefore, development
of novel VEGFR-2 inhibitors is still urgently needed in the market but also
challenging. In this work, residues including Asp1046, Ile1025, HIS1026, Cys919 and
Lys868 were identified as the most important residues for Hbonded interaction, while
His1026, Asp1046, Glu885, Ile1025 and Leu840 exhibited critical role for the
nonbonded interactions through a comprehensive analysis of protein-ligand
interactions, which plays critical roles in the binding of compounds and targets.
Guided by the analysis of binding interactions, a total of 10 novel VEGFR-2
inhibitors based on N-methyl-4-oxo-N-propyl-1,4-dihydroquinoline-2-carboxamide
scaffold were discovered through fragment-based drug design and structure-based
virtual screening, which expands the chemical space of current VEGFR-2 inhibitors.
Biological activity evaluation showed that even though the enzymatic activity of these
compounds against VEGFR-2 were inferior to that of the positive controls sorafenib
and motesanib, compound I-10 showed moderate HepG2 cell inhibitory activity with
an IC50 value of 33.65 μM, and eight compounds exhibited moderate or higher
HUVEC inhibitory activity in the range of 19.54-57.98 μM compared to the controls.
Particularly, the HUVEC inhibitory activity of compound I-6 (IC50 = 19.54 μM)

outperformed motesanib and can be used as starting points for further optimization
and development for cancer treatment.
Graphical Abstract
Keywords: VEGFR-2 inhibitors; Protein-ligand interaction; Fragment-based drug
design; Structure-based virtual screening
1. INTRODUCTION
Angiogenesis is the formation of new blood vessels and the division of existing blood
vessels(Kiselyov, Balakin, & Tkachenko, 2007). Normal angiogenesis generally
occurs during ovulation, wound healing and reproduction(Klagsbrun & Moses, 1999).
In the early 1970s, Folkman proposed the hypothesis that tumor growth and
metastasis depend on angiogenesis, and proposed a strategy for treating cancer by
inhibiting angiogenesis(Hanahan & Folkman, 1996). Since then, treating tumors by
inhibiting angiogenesis has been a hot spot in the field of anticancer
research(Kiselyov et al., 2007; Musumeci, Radi, Brullo, & Schenone, 2012). Later,
Ferrara and colleagues found that vascular endothelial growth factor A (VEGF-A),
which stimulates mitosis of endothelial cells and promotes angiogenesis(Holmes,

Roberts, Thomas, & Cross, 2007), opening the study of targets involving VEGF
signaling pathways. After activation, vascular endothelial growth factor receptor
(VEGFR) plays a major role in the development and progression of cancer(Ivy, Wick,
& Kaufman, 2009).
When VEGF binds to VEGFR, it induces dimerization of VEGFR, causing
modification of the intracellular domain. These conformational changes induce the
exposure of the ATP binding site, followed by receptor dimerization upon ATP
binding, resulting in the autophosphorylation or dephosphorylation of specific
tyrosine residues for downstream signal transduction proteins(Koch, Tugues, Li,
Gualandi, & Claesson-Welsh, 2011). Tyrosine phosphorylation on VEGFRs is
regulated by internalization, degradation and dephosphorylation of different protein
tyrosine phosphatases(Kappert, Peters, Böhmer, & Östman, 2005). This mechanism
leads to the initiation of a typical receptor signaling cascade, which activates several
downstream signal transduction pathways, such as p38-MAPK, Raf/MEK/ERK, and
PI3K/PKB pathways (Figure 1)(Morabito, De Maio, Di Maio, Normanno, & Perrone,
2006). Activation of VEGFR and its downstream pathways will mediate cell
proliferation, increase cell vascular permeability, and accelerate cell migration and
cell survival, which ultimately lead to angiogenesis. Therefore, the development of
inhibitors targeting VEGFR has become an important area in current anti-cancer
research. Particularly, VEGFR-2 acts as a major receptor for angiogenic function,
especially in tumor growth, metastasis and multidrug resistance of tumors(Ivy et al.,
2009). Monoclonal antibodies or small-molecule inhibitors can block the upstream

target VEGF or VEGFR-2, and thereby inhibit the corresponding downstream
pathways, resulting in the inhibition of abnormal angiogenesis (Figure 1)(Ivy et al.,
2009).
Figure 1. VEGFR signal pathways involving angiogenesis(Ivy et al., 2009).
Currently, three monoclonal antibodies including bevacizumab(Ferrara, Hillan,
Gerber, & Novotny, 2004), ramucirumab, aflibercept have been approved for the
treatment of metastatic colorectal cancer. Nine VEGFR-2 small-molecule drugs
(Figure 2) including sorafenib(Folkman & D'Amore, 1996), sunitinib(Risau, 1997),
pazopanib(Klagsbrun & Moses, 1999), vandetanib(Ivy et al., 2009), axitinib(Ferrara,
2004), regorafenib(Pei et al., 2006), nintenanib(Roth et al., 2009), lenvatinib(Li et al.,
2006), apatinib(Roviello et al., 2016) are approved for the treatment of renal
carcinoma, non-small cell lung cancer and thyroid cancer etc. in the past decades,
which fully demonstrates the feasibility of treating tumors by inhibiting VEGFR-2.

Those inhibitors are all ATP competitive inhibitors(Musumeci et al., 2012), and their
binding conformation with VEGFR-2 are shown in Figure 2. Most of those inhibitors
are multi-target inhibitors with nanomolar activity for VEGFR subtypes and those
targets of high sequence homology with VEGFR-2, such as PDFRR, FGFR and c-Kit
etc. In addition, for VEGFR-2, there are a variety of inhibitors in clinical or
preclinical studies at present, and their structures cover a wide chemical space,
including quinazolines, quinolines, quinolines-2-ketones, oxindoles, phthalazines,
furopyrazines,
2-aminonicotinamides,
3-aminothiophene-2-carboxamides,
pyrrolotriazines, oxazoles, pyrrolopyrimidines, imidazolins, oxazolpyrrolocarbazoles
and isothiazoles. Some of those compounds are previously or now in clinical I, II or
III phase studies, alone or in combination for the treatment of various solid or
hematoma tumors(Musumeci et al., 2012).

Figure 2. Binding mode of FDA approved nine VEGFR-2 inhibitors
However, whether it is bound to the active conformation of the kinase (type І) or to
the inactive conformation of the kinase (type ІІ), it is basically an ATP competitive
inhibitor. Although some of these compounds are derived from minor chemical
modifications to drugs that have been approved or clinically tested, most new
inhibitors may be more active than their parent drugs. It is worth noting that, in
addition to drug-target interactions, small changes in chemical structure can lead to
different solubility or permeability, leading to strikingly different biological behaviors,
especially when the target compound is a non-polar molecule, such as tyrosine kinase
inhibitor. Additionally, endogenous and acquired resistance are still considered to be
the main factors leading to short-term clinical effects and failure of anti-angiogenic
drugs(Barouch-Bentov & Sauer, 2011; Davis et al., 2011). As various type of

VEGFR-2 inhibitors have already discovered, it is more challenging to obtain a novel
hinge region scaffold. Therefore, discovering novel VEGFR-2 inhibitors with
anti-angiogenic activity are still urgently needed in the market despite the challenge.
Fortunately, efficient computer-aided drug design strategies offers opportunity and
advantage to find promising VEGFR-2 inhibitors(Y. Zhang et al., 2017), especially
the protein-ligand interaction analysis and fragment-based drug design(Bian & Xie,
2018).
In this paper, the binding cavity of VEGFR-2 and its key residues were described
and summarized by analyzing the protein-ligand interaction of VEGFR-2 crystal
complexes first. Then, a fragment replacement method was used to construct an
in-house compound database to expand the diversity of chemical space for small
molecules in virtual screening study. On the basis of the protein-ligand interaction
analysis,
a
series
of
novel
VEGFR-2
inhibitors
with
N-methyl-4-oxo-N-propyl-1,4-dihydroquinoline-2-carboxamide scaffold was obtained
by structure-based screening of an in-house database and then submitted to chemical
synthesis. The VEGFR-2 kinase activity and cell viability of the selected compounds
were further evaluated, and several VEGFR-2 inhibitors with the novel scaffold in the
hinge region were obtained, which demonstrates the application domain of
protein-ligand guided discovery of kinase inhibitors.
2. MATERIALS AND METHODS
2.1 Molecular Modeling.
2.1.1 FTMAP

As a binding site detection method, FTMap samples billions of probe positions on
dense translational and rotational grids based on the extremely efficient fast Fourier
transform correlation approach and utilizes the sum of correlation functions for
scoring.(Brenke et al., 2009) Sixteen small organic probe molecules of different
hydrogen bonding capability and hydrophobicity are utilized to scan the entire surface
of the protein to find the potential binding cavity. Those 16 probes consists of benzene,
acetamide, ethane, acetonitrile, urea, methylamine, phenol, benzaldehyde, ethanol,
isopropanol, isobutanol, acetone, acetaldehyde, dimethyl ether, cyclohexane and
N,N-dimethylformamide.(Ngan et al., 2012) Any cluster with at least six probes of the
lowest mean interaction energies are retained and regarded as consensus sites or hot
spots. The hot spots are ranked according to the number of probes captured, with
consensus site of the most probes representing the most critical sites. In addition,
hydrogen bond (Hbonded) interactions and nonbonded interactions between each
residue and the probes are determined. Those interactions include hydrophobic
interactions, electrostatic interaction, cation π interaction, van der Waals forces, ionic
interaction and etc. The frequency of probes are employed to evaluate the interactions
occurred in contacting with certain residues. The online platform
(http://ftmap.bu.edu)(Ngan et al., 2012) was employed for FTMap analysis.
2.1.2 Molecular Fragment Replacement
Unlike other fragment-based drug design methods, the purpose of molecular
scaffold replacement is to replace part of the original molecule in an effort to maintain
some of the critical binding interaction.(Bergmann, Linusson, & Zamora, 2007) The

in-house compound database was generated by MOE’s scaffold replacement
module(Y. Zhang et al., 2012). Crystal structure of VEGFR-2 in complexed with
sorafenib (PDB 4ASD) was prepared by Protonate3D(Labute, 2009) tool to add
hydrogen atoms and assign ionization states and designated as the receptor to ensure
that the generated compounds can better match the binding pocket of the receptor. The
lead compound was modified by deleting its hinge region part. As molecular
properties are important for the discovery of druglike molecules; thus, molecular
weight less than 600 D, slogP (-4 to 8), topological polar surface area (TPSA, 40 to
140) were set in the descriptor filter for finding druglike molecules (Ertl, Rohde, &
Selzer, 2000; Wildman & Crippen, 1999). Through the protein-ligand interaction
analysis, we learned that Cys919 is important for the binding; therefore, a one-feature
pharmacophore was generated by the Pharmacophore Query tool and set as another
constraint during the fragment searching phase. Finally, the Add Group to ligand
module was adopted which extend a compound at one or more site to find suitable
fragments for the construction of potential hits. To enrich the chemical diversity of the
generated compounds, the RECAP module was used to generate more hits based on
the former compound database.
2.1.3 Molecular docking
Due to its excellent performace in the virtual screening of VEGFR-2 inhibitors
from our pervious study(Yanmin Zhang et al., 2013), crystal structure (PDB 3EWH)
of VEGFR-2 was downloaded from the Protein Data Bank (PDB) and prepared using
the Protein Preparation Wizard workflow in the Schrꢀdinger suite(Ding et al., 2017;

Schrodinger, 2011). Water molecules beyond 5 Å of the cognate ligand were deleted
and only hydrogens were selected for the restrained minimization with the force field
of OPLS_2005. A grid box of similar size with the crystallized ligand was generated
as the binding cavity for potential VEGFR-2 inhibitors using the Receptor Grid
Generation module. The in-house database compounds were minimized by the
LigPrep package(Lu et al., 2018). Because its excellent performance in a
cross-docking analysis(Yanmin Zhang et al., 2013), Glide standard precision (SP)
mode was chosen for molecular docking(Yanmin Zhang et al., 2013). The top 10
poses of each ligand were minimized by a post-docking procedure that saved the best
pose for further analysis.
2.2 Chemistry Experiments.
Commercially available chemicals were used as purchased without further
purification. Solvents were purified and stored according to standard procedures. All
reactions were monitored by thin-layer chromatography (TLC), and silica gel plates
with fluorescence F-254 were used and visualized with UV light. Column
chromatography was carried out on silica gel (300 −400 mesh). The melting points
were determined on Digital melting point instrument (Shanghai Physical Optical
Instrument Factory WRS-1A) and X-4 digital display micro melting point measuring
instrument (Beijing Tektronix Instrument Co., Ltd.). The NMR spectra were measured
on a Bruker Avance III-300 and/or Bruker Avance III-500 instrumentment, TMS is
determined by internal standard. Chemical shifts are expressed as δ units using
tetramethylsilane as the external standard (in NMR description, s, singlet; d, doublet; t,

triplet; q, quartet; m, multiplet; and br, broad peak). All coupling constants (J) are
reported in Hertz. MS analysis was performed on Agilent 1100 LC/MSD and 2010
GC/MS mass spectrometer (Agilent, USA). IR is SHIMADZU FTIR-8400S, KBr
tablet. TLC plate is made of silica gel GF254 (produced by Qingdao Ocean Chemical
Plant), ans column chromatography uses 300-400 mesh silica gel (produced by
Qingdao Ocean Chemical Plant). The details of the synthesis process and structure
spectra data for the target compounds as well as the intermediates can be found in the
Supporting Information.
2.3 Biological Assay.
2.3.1 In vitro Kinase Assay.
One single concentration (20 μM) inhibition rate determination was entrusted to
China Pharmaceutical University New Drug Screening Center. The inhibitory
property on the enzyme activity was assessed by Cisbio’s HTRF® KinEASE™ Kit, in
which a unique substrate with a single phosphorylation site would be recognized by
the phospho-tyrosine antibody labeled with the cryptate (Eu(K)). The detection
reagent captures the phosphorylated substrate and the obtained homogeneous
time-resolved fluorescence (HTRF) signal is in the proportion of the amount of
phosphorylation.(Tardieu, 2007) Based on HTRF, the KinEASE assay consists of two
steps: the first step is the kinase reaction phase and then is the detection step. 2 μL of
VEGFR-2 kinase solution, 2 μL of biotin substrate and 4 μL of compound isolated in
SEB-supplemented kinase buffer were added to each well for incubation in the
enzymatic reaction step. For the initiation of reaction, 2 μL of LATP was then added

at room temperature (18-22°C). The reaction was carried out for 1 hour. In the second
step, the detection reagents, 5 μL of tyrosine kinase antibody-Eu (K) and 5 μL of
streptavidin-XL665 (SA-XL665) in EDTA, were added to each well and incubated at
room temperature for 1 hour. The signal was detected using the Beckman Coulter
platform HTRF detection module.
2.3.2 Anti-proliferative assay
The cell activity assessment were entrusted to Crown Bioscience Inc. Fluorescein,
ATP and oxygen were treated as substrates for the luciferase in the CTG reagent to
produce oxidized fluorescein and release energy in the form of light. The produced
light amount is proportional to the total amount of ATP, which can reflect the total
number of living cells (human umbilical vein endothelial cells, HUVEC and liver
hepatocellular carcinoma, HepG2), and the fluorescence intensity is used to calculate
the anti-cell proliferation rate. The method consists of several steps. The first step was
cell plating, in which exponential-phase cells were collected and viable cells were
counted using a Vi-Cell XR cell counting instrument. On the basis of the density in
the cell culture medium, 90 μL of cell suspension was added to each well of a 96-well
cell culture plate. The final cell concentration was about 2000 to 4000 cells per well
(the specific cell density was regulated based on cell growth). The next step was
compound partitioning, where 10 mM stock solutions in DMSO were used to dissolve
target compound and these solutions were diluted 10-fold with the medium solution.
10 μL of 10-fold compound dilution in total was added to each cell line per well,
resulting in a final drug concentration of 10 μM and a final DMSO concentration of

0.1%. The plates were placed in a 37 ° C incubator containing 5% CO 2 for 72 hours,
followed by the plate detection step. Based on the manufacturer's instructions, 50 μL
of CTG solution which was thawed previously and equilibrated to room temperature
was added to each well after 72 h of drug treatment. The solution was mixed using a
microplate shaker for 2 minutes. After incubation for 10 minutes at room temperature,
the fluorescence signal value was measured by an Envision 2104 plate reader. Those
with more than 30% cell inhibition at a concentration of 10 μM were further tested for
10 concentrations (from 10 nM to 1000 μM). Data processing: inhibition
ratio=1-Vsample/Vvehicle control *100%. Vsample is for drug treatment group while Vvehicle
control is for solvent control group. Non-linear regression models were constructed and
S type dose survival curves were plotted by GraphPad Prism 5.0 software
(https://www.graphpad.com/scientific-software/prism/) and the IC50 values were
calculated. Each test was duplicated three times.
3. RESULTS AND DISCUSSION
3.1 Binding Site Analysis
The hot spots in the VEGFR-2 binding pocket were analyzed by FTMap based on a
total of 38 VEGFR-2 crystal structures. Table S1 in the Supporting Information
shows the distribution of the hot spots derived from the 16 different probe types. It
has been proved that the main hot spots binds at least 16 probe clusters in a druggable
target and in combination with its nearby hot spots, and they constitute the binding
site which potentially incorporating drug-size ligands(Dima et al., 2011; Landon,
Lancia-Dr, Thiel, & Vajda, 2007; Y. Zhang et al., 2016). From Table S1, FTMap

could recognize at least one druggable site in all the VEGFR-2 crystal structures
except for PDB 2XIR, 3CPC and 4AGC. However, they had at least four hot spots
constituted by at least 10 probe clusters and their first hot spot all contained 15 probe
clusters. For example, the top six hot spots for the FDA approved VEGFR-2 inhibitors
crystal structures were shown in Figure 3, from which we could see that most of the
hot spots were situated in the binding pocket of VEGFR-2 crystallized ligands. The
hot spot distribution of the available VEGFR-2 crystal structures with FDA approved
VEGFR-2 inhibitors consisted of 4ASD (sorafenib), 4AGD (sunitinib), 3CJG
(pazopanib), 4AGC (axitinib), 3C7Q (nintedanib) and 3WZD (lenvatinib). The hot
spot distributions of the other 36 VEGFR-2 complexes demonstrated similar
distributions as crystal structures complexed with the FDA approved VEGFR-2
inhibitors.
Figure 3. The top six consensus sites (CS) detected by FTMap for crystal complexes
of six marketed VEGFR-2 inhibitors. CSs are denoted as lines with CS1 (cyan), CS2
(magenta), CS3 (yellow), CS4 (salmon), CS5 (gray), and CS6 (slate blue)

successively in descending number of probe clusters. The number in the parentheses
is the number of probe clusters detected in that hot spot.
As shown in Figure 3, the binding cavity of VEGFR-2 can be generally divided
into three major parts consisting of the hinge region, DFG-motif and hydrophobic
back pocket. The hinge region and the hydrophobic back pocket obtained at least one
consensus sites (CS), individually. According to Table S1, sites at PDB 3C7Q and
3WZD hinge region were all CS1, with 20 and 18 probe clusters captured,
respectively. In PDB 4AGD and 3CJQ, CS1s of 22 and 28 probe clusters were located
in the DFG-motif region while for PDB 4AGC, CS1 of 15 probe clusters was found in
the hydrophobic back pocket. These results indicated the importance of this region,
and further confirmed that the type II kinase inhibitors generally have better activity
and selectivity and the DFG-motif region is an essential part for the binding of type II
inhibitors for most kinase targets. Still, some consensus sites such as CS5 of 8 probe
clusters in PDB 4AGD, CS3 of 13 probe clusters, CS2 of 15 probe clusters and CS6
of 6 probe clusters in PDB 4AGC, CS3 of 12 probe clusters in PDB 3WZD were
situated in the wide range of the binding pocket, demonstrating the potential of sites
to expand and optimize the current VEGFR-2 inhibitors. In summary, the FTMap
binding site detection can accurately probe the binding pocket of VEGFR-2 and can
be used as a confirmation for potential binding sites.
3.2 Protein-Ligand Interaction Analysis
Hitherto, a total of 38 VEGFR-2 crystal structures (PDB) were resolved with
various types of VEGFR-2 inhibitors(Yanmin Zhang et al., 2013). A comprehensive

analysis of the protein-ligand binding interaction was conducted by detecting all the
VEGFR-2 crystal structures. Herein, the frequency of both Hbonded and nonbonded
interactions (Figure 4) for each residue in the 38 PDB complexes were calculated.
Figure 5 and Figure 6 show the hot map of the Hbonded and nonbonded interactions,
respectively. The corresponding values of the Hbonded and nonbonded interactions
frequency can be found in the Supporting Information Table S2 and S3. From Figure
4 and Figure 5, 13 residues indicated Hbonded interaction with the cognate ligand in
more than 30 PDB complexes. They were Cys919 (38/38), Leu840 (33/38) in the
hinge region; Cys 1045 (36/38), Asp1046 (38/38), Glu885 (37/38), Lys868 (37/38)
around the DGF-motif region; Ile1025 (36/38), His1026 (34/38), ARG1027 (33/38),
ASP1028 (35/38) in the hydrophobic back pocket; and Asn923 (38/38), Ser925
(31/38), Arg1032 (38/38) within the solvent accessible region.
VEGFR-2 inhibitors are generally classified as type I (DGF-in) and type II
(DGF-out) class. From the crystal complexes analysis, it can be seen that the type I
inhibitors such as axitinib (PDB 4AGD) and sunitinib (PDB 4AGC) generally bind to
the above key residues in the hinge region and the solvent accessible region. The type
II inhibitors such as sorafenib (PDB 4ASD), lenvatinib (PDB 3WZD), and apatinib
(PDB 3EFL) mainly bind to key residues in the hinge region and the hydrophobic
back pocket. In addition, at least 20 PDB complexes obtained Hbonded binding
interactions with residues Arg842, Ala881, Val899, Glu917, Arg929, Cys1024,
Asn1033, Ile1044, Phe1047, Gly1048, Leu1049, Ala1050, Arg1066, Tyr1082,
Glu1097, Ala1103 and Ser1104 within the binding pocket of VEGFR-2. Among all

residues, Lys868 (16/38), Glu885 (13/38), Cys919 (16/38), Ile1025 (29/38), His1026
(23/38), Arg1027 (14/38), Cys1045 (13/38), and Asp1046 (35/38) appeared in at least
10 PDB crystal structures with the percent of Hbonded interactions frequency higher
than 5%. The results of nonbonded interaction showed similar trend with the Hbonded
interaction analysis. As shown in Figure 4 and Figure 6, the same 13 residues
showed nonbonded interactions with their cognate ligands in at least 30 PDB
complexes as compared to the Hbonded interaciton. Except Cys1045 (36/38) and
ASP1046 (38/38) were slightly different with the Hbonded interactions, the other 11
residues exerted nonbonded interactions in the same number of PDB complexes.
Similarly, Leu840 (10/38), Glu885 (16/38), Cys919 (4/38), Ile1025 (11/38), His1026
(27/38), Asp1046 (18/38) showed nonbonded interactions with cognate ligands in at
least 10 PDB complexes. In summary, for the contributon of Hbonded interaction,
Asp1046, Ile1025, HIS1026, Cys919 and Lys868 were the most important residues
while His1026, Asp1046, Glu885, Ile1025 and Leu840 exhibited the most evident
nonbonded interactions.
Figure 4. key residues detected from FTMap Hbonded and nonbonded analysis (PDB
3WZD) and the number of residues with a certain percent of frequency.

Figure 5. FTMap Hbonded interaction analysis of VEGFR-2 crystal structures.

Figure 6. FTMap nonbonded interaction analysis of VEGFR-2 crystal structures.
The specific Hbonded and nonbonded interactions between the cognate ligands
with representative VEGFR-2 crystal structures were shown in Figure 7. Figure 7
showed the available PDBs crystallized with six FDA approved VEGFR-2 inhibitors

including sorafenib (PDB 4ASD), sunitinib (PDB 4AGD), pazopanib (PDB 3CJG),
axitinib (PDB 4AGC), nintedanib(PDB 3C7Q) and 3WZD (lenvatinib). Among the
six listed inhibitors, sorafenib (PDB 4ASD) and lenvatinib (PDB 3WZD) are belong
to the type II kinase inhibitors as they both occupy the back hydrophobic pocket once
the DFG-motif is in the out conformation. They shared similar distributions in the
Hbonded and nonbonded interactions. Specifically, in terms of the Hbonded
interaction frequency, for PDB 4ASD, residues with the top five highest Hbonded
interaction frequency were Arg1051 (14.2%), Asp1046 (9.6%), Tyr1082 (8.7%),
Arg1027 (8.4%) and Ile1025 (6.9%). For PDB 4AGD, the corresponding residues
were Arg1051 (29.2%), Tyr1059 (8.0%), Ser803 (5.6%), Arg842 (4.2%) and Asn923
(3.8%). While for PDB 3CJG, the top five highest Hbonded interaction frequency
were Asp1046 (33.0%), Phe1047 (11.2%), Leu889 (8.2%), Asn1033 (6.4%) and
Arg1032 (5.1%). For PDB 4AGC, Arg1051 (37.1%), Asn923 (8.2%), Arg1027 (7.1%),
Arg842 (4.8%) and Leu802 (3.4) ranked the top five for the Hbonded interaction
frequency. In terms of PDB 3C7Q, the corresponding top five residues were Cys919
(28.3%), Lys868 (9.6%), Arg1032 (7.9%), Asp1046 (7.3%) and Asn1033 (6.2%)
while for PDB 3WZD, residues with the top five highest Hbonded interaction
frequency consisted of Asp1046 (17.9%), Cys919 (116.3%), Asn1033 (9.0%),
Arg1032 (8.8%) and Phe 1047 (8.0%). Generally, residues with higher Hbonded
interaction frequency also obtained relatively higher nonbonded interactions
frequency. However, there was one exception. Even through Leu840 did not achieve
high Hbonded interaction frequency as no PDBs appeared in the top five, its

nonbonded interactions frequency were relatively greater than other residues. The
corresponding nonbonded interactions frequency for Leu840 were 4ASD (1.7%),
4AGD (2.4 %), 3CJG (8.9%), 4AGC (5.9%), 3C7Q (1.2%) and 3WZD (9.4%),
respectively.
Overall, in the top five Hbonded interaction frequency, Cys919 in the hinge region
only appeared in PDB 3C7Q (28.3%) and 3WZD (16.3%) whereas Asp1046 in the
DFG-motif appeared in four PDBs including PDB 3C7Q (7.3%), 3CJG (33.0%),
3WZD (17.9%) and 4ASD (9.6%). Despite that Asp1046 in PDBs 4AGC and 4AGD
did not got any top five Hbonded interaction frequency, their cognate ligands have a
strong Hbonded interaction with Arg1051 of Hbonded interaction frequency of 37.1%
and 29.2%, respectively. In addition, Arg1032, Asn1033 and Phe1047 were also quite
critical as they appeared in at least two PDBs in the ranking of top five Hbonded
interaction frequency. Moreover, the other 32 PDBs (Figure 5 and Figure 6) showed
similar trend with the six PDBs complexed with FDA approved drugs in terms of both
Hbonded and nonbonded interaction frequencies (Figure 7). In total, Asp1046,
Cys919, Arg1051, Phe1047, Arg1032, Asn1033 were the most important residues
forming interactions with the approved VEGFR-2 inhibitors, which was consistent
with previous results on the detected interactions for all VEGFR-2 crystal structures.

Figure 7. Hydrogen bond frequency maps in the six proteins crystal complexes of six
marketed VEGFR-2 inhibitors. The red bar and blue bar represent the Hbonded
interactions (Hbonded %) and the nonbonded interactions (Nonbonded %),
respectively.
3.3 Structure-Based Virtual Screening
Fragment-based drug design generally includes fragment growing, fragment linking
and de novo drug design(Bian & Xie, 2018). Molecular scaffold replacement belongs
to fragment growing as it retains part of the known inhibitors either in the hinge
region or the other pocket such as the DFG motif or the hydrophobic back pocket in

the kinase targets. Due to its relatively conserved strategy, the successful rate is higher
than fragment linking and de novo drug design. The design and optimization of the
lead compound is shown in Figure 8A. In the first step of the workflow, hinge region
scaffold 4-hydroxy-N-methylpicolinamide of sorafenib was replaced by retaining the
diphenylurea part which occupys the DFG motif and the hydrophobic back pocket. A
total of 15595 compounds were generated by this module. After duplication, 14777
compounds were kept for later evaluation. Then, the generated compounds were
submitted for molecular docking, which can explore the best binding mode between
the target (here is VEGFR-2) and the target compounds and give a comprehensive
docking score. PDB 3EWH was used as the receptor as it was selected as the most
foavorable crystal structure for VEGFR-2 docking-based virtual screening by a
comprehensive self-docking and cross-docking analysis(Yanmin Zhang et al., 2013).
Third, through the protein-ligand interaction analysis, we learned that Asp1046,
Ile1025, HIS1026, Cys919 and Lys868 and Leu840 were key residues responsible for
Hbonded or nonbonded interactions. Thus, only the docked compounds having either
Hbonded or nonbonded interactions with Asp1046, Ile1025, HIS1026, Cys919,
Lys868, Leu840 and His1026 were kept for the last manually selection step. Among
them,
a
novel
scaffold
in
the
hinge
region
of
N-methyl-4-oxo-N-propyl-1,4-dihydroquinoline-2-carboxamide
was obtained.
Furthermore, as our former study(Yang et al., 2015) showed that the R1, R2 and R3
sbustituents can be replaced by halogen (F, Cl, CH3 or CF3) to adjust the
physiochemical properties, a total of 10 compounds were finally designed for further

evaluation (Table 1). Additionally, rather than specifically affecting one desired target,
pan assay interference (PAINS) compounds are often regarded as false positives as
they tend to react with numerous biological targets simultaneously.(Baell & Holloway,
2010) Thus, the designed 10 compounds were submitted to PAINS remover
(http://cbligand.org/PAINS/)(Baell & Holloway, 2010) to avoid any false positives.
Fortunately, all 10 compounds passed the PAINS filtration process and were adopted
for chemical synthesis.
Figure 8. (A) Lead compound design and optimization workflow. (B) The docking
conformation of designed compounds (PDB 3EWH); (C) The docking binding mode
of the lead compound I-1 (PDB 3EWH). (D) Comparison of the docking binding
mode of I-1 (PDB 3EWH) and sorafenib crystalized with VEGFR-2 (PDB 4ASD); (E)
Comparison of the docking binding mode of I-1 (PDB 3EWH) and motesanib
crystalized with VEGFR-2 (PDB 3EFL).
As shown in Figure 8B, the binding mode of the 10 target compounds were similar
to each other as the all occupied they three major pockets including the hinge region,
DFG-motif and hydrophobic back pocket and can be regarded as type II inhibitors

like sorafenib (Figure 8D) and motesanib (Figure 8E) etc.. Similar to that of
sorafenibi and motesanib, the binding mode of lead compound I-1 (Figure 8C)
showed that compounds I-1 formed hydrogen bond interactions with Asp1046 and
Glu885 as well as Cys919, which were recognized as some of the most critical
residues for the binding of small molecules to VEGFR-2. Moreover, those target
compounds also involved in the nonbonded interactions with residues like Asp1046,
Ile1025, HIS1026 and Leu840 etc., which was consistent with the former
protein-ligand interaction analysis. Moreover, the binding mode of these target
compounds overlapped well with the cystallized conformation of known inhibitors
such as sorafenib (Figure 8D), and motesanib (Figure 8E) etc., further demonstrating
the posibility as potential VEGFR-2 inhibitors.
3.4 Chemistry
3.4.1 Scheme 1. Synthesis of Compounds I-1~I-8
Under
sulfuric
acid
conditions,
the
cyano
group
of
2-methyl-4-nitrophenylacetonitrile alcoholysis was used to obtain ethyl phenylacetate
1(Wojciech Dmowski, 1987). Compound 1 was reduced by NaBH4 to give compound
2(Ferri, Costa, Biancardo, Argazzi, & Bignozzi, 2007). Bromination of compound 2
provided compound 3(Wee et al., 2009). Compound 3 was in a saturated aqueous
solution of methylamine to obtain compound 4. Compound 5 was obtained by
addition reaction of aniline and dimethyl acetylenedicarboxylate(Mazzoni et al., 2010;
Pandey et al., 2013). The compound 5 was cyclized in Dowtherm to give compound 6
(Mazzoni et al., 2010; Pandey et al., 2013). Under basic conditions compound 6 was

hydrolyzed to give compound 7. Then compound 7 and compound 4 were condensed
to give amide 8(Akihiro, 2009) . Hydrogenation reduction of nitro group followed by
compound 8 was to give compound 9(Frost et al., 2010).Then compound 9 was
reacted with various substituted anilines to obtain the target compounds I-1~I-8
(Scheme 1).
Scheme 1:
Reagents and conditions:
a) EtOH, 98%H2SO4, 120℃,4h; b) 4eq NaBH4, EtOH, reflux, overnight; c) 1.4eq
CBr4 and 1.4 eq PPh3, dry CH2Cl2, rt, 30~60min; d) 40% methylamine solution,
overnight, rt; f)Dowtherm, 240℃, 30min or PPA, 130℃, 1h; g) NaOH, MeOH, rt, 2h;
h) EDCI, HOBt, Et3N, dry DMF, 5℃~rt, overnight; i) 5%Pd/C, H2, MeOH, rt, 30min;
j) aniline derivatives, CDI, dry DCM, rt.
3.4.2 Synthesis of Compound I-9
The synthetic process started from 1-(2-bromoethyl)-4-nitrobenzene and

compounds 10-12 were prepared by a method similar to that for compound 4, 8 and 9.
Compound 12 was reacted with 4-chloro-3-trifluoromethylbenzene isocyanate to
provide target compound I-9 (Scheme 2).
Scheme 2:
Reagents and conditions:
a) 40%methylamine water solution, rt; b) 56, EDCI, HOBt, dry DMF; c) 5%Pd/C
MeOH; d) 4-chloro-3-isocyanato-2-(trifluoromethyl)benzene, dry CH2Cl2, rt.
3.4.3 Synthesis of Compound I-10
Under basic conditions, 3, 4-Difluoronitrobenzene and ethyl cyanoacetate were
reacted to provide compound 13(Nicle, 1984). Then compound 13 was hydrolyzed
and decarboxylated under acidic conditions to obtain compound 14(Hans-Georg,
2005). Sodium borohydride reduced the ester to give compound 15(Ferri et al., 2007).
Bromination of compound 15 gave compound 16. Compound 16 was placed in
aqueous methylamine to obtain compound 17. Compounds 18, 19 were prepared by a
method similar to that of compound 8, 9, which were reacted with
4-chloro-3-trifluoromethylbenzene isocyanate in dichloromethane to obtain the target
compound I-10 (Scheme 3).

Scheme 3:
Reagents and conditions:
a) KOH, DMSO, rt, overnight; b) 98%H2SO4, 50%CH3COOH, reflux, 10h; c) 4eq
NaBH4, EtOH, reflux, overnight; d) CBr4/PPh3, dry THF, rt; e) 40% methylamine
water solution/THF, rt, overnight; f) 56, EDCI, HOBt, Et3N, dry DMF, 5℃~rt,
overnight;
4-chloro-3-isocyanato-2-(trifluoromethyl)benzene, dry DCM, rt.
In summary, total of 10 hit
g)
5%Pd/C,
H2,
rt,
30min;
h)
a
compounds
with
N-methyl-4-oxo-N-propyl-1,4-dihydroquinoline-2-carboxamide scaffold which have
not been reported in the literature were synthesized (Table 1). At the same time, 19
synthesized intermediate compounds not reported in the literature were also obtained.
All the target compounds were confirmed by IR, 1H-NMR and MS, and some of the
unreported intermediates were confirmed by 1H-NMR or MS.