Discovery of Novel Vascular Endothelial Growth Factor Receptor 2 Inhibitors: A Virtual Screening Approach

Authors


Corresponding author: Wenfang Xu, xuwenf@sdu.edu.cn

Abstract

A virtual screening approach was performed to develop novel and potent vascular endothelial growth factor receptor 2 inhibitors. The Specs database was filtered by ‘rule of five’, a pharmacophore model, and docking filter. Sixteen molecules were selected for tube formation assay, a naphthalenol group containing molecule, 12, showed good performance in the study. In the following aortic ring assay and 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay, 12 was discovered to efficiently inhibit angiogenesis and tumor cell growth. It is the first time to discover naphthalenol scaffold as potent vascular endothelial growth factor receptor 2 inhibitors. Thus, a molecular dynamic simulation process was applied to discover key features of 12 in binding to vascular endothelial growth factor receptor 2. Hydrophobic interactions were discovered to play significant role in the ligand–receptor binding.

Abbreviations:
HUVECs

human umbilical vein endothelial cells

MTT

3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide

RMSD

root mean square deviation

RMSF

root mean square fluctuation

VEGFR2

vascular endothelial growth factor receptor 2

μL

microliter

μm

micromolar

Vascular endothelial growth factor receptors (VEGFRs) play important roles in tumor angiogenesis and metastasis (1,2). The intrinsic tyrosine kinase activities of VEGFRs were activated by homodimerization or heterodimerization induced by ligands (3). There are three members in the VEGFR family (VEGFR1, 2, 3), among them, only homodimerization of VEGFR2 leads to a strong autophosphorylation of tyrosine residues of VEGFR2 (3,4). While VEGFR1 is weakly involved in transducing the VEGF angiogenic signals and shows poor autophosphorylation ability in response to vascular endothelial growth factor (VEGF) in endothelial cells (5). Vascular endothelial growth factor receptor 3 mainly functions in the establishment and maintenance of lymphatics (3). Thus, increased interests focused on inhibition of VEGFR2 for anti-angiogenesis in cancer therapy.

Numerous VEGFR2 inhibitors have good performances in preclinical and clinical trials (6–9). Sorafenib (10) was approved by U.S. Food and Drug Administration (FDA) in 2005 for treatment of advanced renal cancer. The abundant accomplishments have encouraged considerable experiments focusing on the development of VEGFR2 inhibitors for tumor treatment.

Virtual screening approach has been used to discover potent VEGFR2 inhibitors with new scaffold (11). Herein, the virtual screening process was performed using sybyl8.0 software package (Tripos Inc., St. Louis, MO, USA). The database was firstly screened by ‘rule of five’ filter followed by a pharmacophore model screening. Docking of the filtered molecules into VEGFR2 structure was performed to further reduce the number of candidates. At last, 16 molecules were evaluated by tube formation assay, aortic ring assay and tumor cell inhibition assay. It is encouraging that a naphthalenol group containing molecule showed predominant performances in these tests. For the purpose of designing analogs, the binding mode of molecule with residues in the active site of VEGFR2 was elucidated by a molecular dynamic simulation.

Methods and Materials

Virtual screening

The Specs chemical database (April, 2009; http://www.specs.net) was used for virtual screening. Concord program (Tripos Inc.) was used to convert the chemical structures into 3D conformation. The database was firstly filtered by ‘rule of five’ (10); molecules with >5 H-bond donors, >5 calculated LogP value, >10 H-bond acceptors, and >500 molecular weight were excluded.

The structure of VEGFR2 (PDB entry: 2QU5) (12) downloaded from RCSB protein database was used as the receptor in the docking procedure. Several residues are missing in the crystal structure. Thus, a homology modeling procedure was performed to fill up the incomplete structure by modeler (13) in Discovery Studio2.5. Firstly, the missing residues were added to the sequence of 2QU5. After that, the 3D structure was constructed using 2QU5 as the template.

A pharmacophore model based on the ligand (ligand276) in the crystal structure was manually defined for further screening. The model containing four features, two hydrophobic sites, one H-bond donor, and one H-bond acceptor, was generated by UNITY in sybyl7.3 software package (Figure 1A).

Figure 1.

 (A) The pharmacophore model used for virtual screening; (B) pipeline of the virtual screening process.

After the pharmacophore model–based screening, the filtered molecules were docked to the active site of the modeled VEGFR2. Surflex dock program (14,15) in sybyl8.0 software was used; 10 poses were generated for each ligand. The protomol was generated based on the ligand in the crystal structure. Other parameter was set referring the default values. Sorafenib and ligand276 were used as a positive control in this step. Compounds having both high docking scores and suitable binding patterns would be selected for activity assay.

Tube formation assay

Matrigel (Cat. 356234; BD Biosciences) was thawed at 4 °C for overnight. Prechilled 96-well plates were coated with Matrigel (50 μL per well) and incubated at 37 °C for 1 h. 2 × 104 human umbilical vein endothelial cells (HUVECs) in 50 μL culture medium were added to each well. The compounds were dissolved by DMSO to the concentration of 100 mm. To make the final concentration of the compounds to be 100 and 20 μm, 100 μL medium with different concentrations of the compounds was added to these wells. The doses of DMSO were kept not more than 1/1000 of the whole volume. After 8 h of incubation at 37 °C, 5% CO2, endothelial cell tube formation was assessed with OLYMPUS inverted microscope.

Aortic ring assay

At 4 °C, 96-well plates were covered with 70 μL Matrigel that was diluted by 1460 medium (1:1). Then, the coated plates were incubated at 37 °C, 5% CO2 for 30 min. Aortas isolated from mice were cleaned of periadventitial fat and connective tissues and cut into approximately 1-mm-long rings. After being rinsed six times by PBS and two times by 1460 medium, the rings were placed on the covered wells. Another 70 μL 1460 medium with DMSO, molecule 12 and sorafenib was added to each well. Concentrations of the compounds are 1, 0.2, 0.04 μm, respectively. Each concentration per compound was added to three different wells. After 3 days, the medium was changed with the exact composition described above. After 6 days of incubation, the microvessel growth was evaluated with OLYMPUS inverted microscope.

MTT assays

ES-2, HCT116, K562, Hela cells and HUVECs were used in the cell proliferation assays. Five thousand cells were seeded into each well of 96-well plates that were incubated at 37 °C, 5% CO2 for overnight. Then, cells were treated with compound 12 and sorafenib at final concentrations ranging from 6.25 to 100 μm. Cells treated with equal volume of DMSO were used as the control. After 48 h of incubation, 0.5% 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) (Amresco, USA) solution was added to each well. Then, a further 4 h of incubation was performed. After that, the culture medium was removed, and 150 μL DMSO was added to dissolve the formazan. After mixing for 5 min, optical density values were detected at 570 nm on a microplate reader (Thermo, USA).

Molecular dynamic simulation

To discover key features of molecule 12 in binding to VEGFR2, a dynamic simulation process was performed using gromacs4.5 software (http://www.gromacs.org) (16–18). The 12-VEGFR2 complex was derived from the docking process. The simulation was performed using a cubic cell geometry, and the default simple point charge (SPC) water was added to the box. Periodic boundary condition was applied, the distance between the grid box and the protein was set to 1.0 nm. Firstly, a steepest descents minimization for 5000 steps was performed to remove bad van der Waals contacts. After that, a 3000 steps conjugate gradient minimization was applied for further energy optimization. Then, a 1-ns simulated annealing was performed by heating the system from 100 to 300 K. Finally, a 10-ns production simulation was submitted. The particle mesh Ewald (PME) method (19) was used for computing long-range electrostatics. The Berendsen temperature coupling and Berendsen pressure coupling were used to keep the system in a stable environment (300 K temperature and 1 bar pressure).

Results and Discussion

Virtual screening

To save computational time, Lipinski’s rule of five was used to remove unsuitable molecules in Specs database. The number of molecules was reduced from 167 475 to 157 822 in this step (Figure 1B). Then, the refined database was filtered by the generated pharmacophore model. Molecules matching both the shape and features of the model were chosen for further study. Seven thousand and fifty-five molecules were selected for docking into the active site of VEGFR2 using ligand276 and sorafenib as positive controls.

Before docking-based screening, ligand276 in the crystal structure was firstly docked to VEGFR2 to examine whether the method was credible. The docking conformation sharing low root mean square deviation (RMSD) value (2.89 Å) with the crystallized structure suggests that the docking method is reliable (Figure 2). Sorafenib has a total score of 8.07, and the score of ligand276 is 7.23. One hundred and twenty-two molecules have higher total score than sorafenib, and 374 molecules have higher total score than ligand276. The top 200 molecules were qualified by molecular scaffold, binding pattern, and physicochemical property evaluation. At last, 16 molecules with novel scaffolds were selected for biological activity assay (Figure 3).

Figure 2.

 Poses of the crystal structure (carbon atoms were colored white) and the docking structure (carbon atoms were colored gray) of ligand276 in the active site of vascular endothelial growth factor receptor 2 (VEGFR2). Structures of ligand276 were rendered in ball and sticks, and structures of residues were in sticks. The picture was created by UCSF Chimera (20).

Figure 3.

 The screened molecules used for biological activity assay and the positive control.

Binding pattern analysis

A v-shaped pocket is formed in the active site of VEGFR2. There are polar residues (such as Lys53, Glu70, Val99, and Thr101) around the bottom of the pocket, and residues (such as Leu74, Phe103, Leu170, and Phe182) in the two opening of the channel are hydrophobic. The pose of ligand276 matches the shape of the channel. The benzimidazole ring locates at the bottom of the pocket, and the side chains stretch to these two openings (Figure 2). The binding conformations of molecules 07 (Figure 4A), 12 (Figure 4B), 16 (Figure 4C), and sorafenib (Figure 4D) also matched the shape of the pocket revealed by docking study. The naphthalenol rings of molecules 07 and 12, the pyrazole ring of molecule 16, and the benzene ring of sorafenib bind to the bottom of the channel. There are strong hydrophobic interactions between these ligands and the protein. H-bond interactions also make contribution to the binding of these ligands to VEGFR2. Molecule 07 has H-bond interactions with Cys180, molecule 12 has H-bond interactions with Asp181 and Lys53. Asp181 also has H-bond interactions with molecule 16 and sorafenib. Cys104 also binds to sorafenib with hydrogen bond.

Figure 4.

 Binding patterns of molecule 07 (A), 12 (B), 16 (C), sorafenib (D) in vascular endothelial growth factor receptor 2 (VEGFR2).

The screened molecules inhibit tube formation

The inhibitory effect on angiogenesis of the screened molecules was firstly evaluated by in vitro assays. To investigate the effects of the derived molecules on endothelial cell tube formation, HUVECs (2 × 104) were added with different concentrations of the screened molecules onto Matrigel layers. After 8 h of incubation, the ability of endothelial cells forming tube-like structures was detected by inverted microscope. At concentration of 100 μm, three molecules, 07, 12, and 16, completely inhibited the formation of tube-like structures compared with the controls (Figure 5). While at concentration of 20 μm, only molecule 12 completely inhibited the tube growth. Therefore, molecule 12 was investigated by further studies.

Figure 5.

 Effects of screened molecules on tube formation, (A) human umbilical vein endothelial cells (HUVECs) treated with 100 μm molecule 07; (B) HUVECs treated with 100 μm molecule 12; (C) HUVECs treated with 100 μm molecule 16; (D) HUVECs treated with 100 μm sorafenib; (E) HUVECs treated with 20 μm molecule 07; (F) HUVECs treated with 20 μm molecule 12; (G) HUVECs treated with 20 μm molecule 16; (H) HUVECs treated with 20 μm sorafenib; (I) HUVECs treated with culture medium (negative control).

Molecule 12 inhibits angiogenesis in vitro

Aortic ring assays were performed to further examine the inhibitory activity of molecule 12 on angiogenesis. The isolated aortas from mice were used, which was cut into approximately 1-mm-long aortic rings. The rings were put on diluted Matrigel and covered by 1460 medium with DMSO, molecule 12, and sorafenib. After 6 days of incubation, the microvessel growth of the rings was compared by taking photographs with inverted microscope. Molecule 12 and sorafenib inhibited the microvessel growth in a dose-dependent manner. At concentration of 1 μm, the growth of microvessels was completely inhibited (Figure 6). While reduced the concentration to 0.2 μm, a few microvessels have grown up. Their inhibitory effects at 0.04 μm of concentration were not obvious. The result suggests that molecule 12 has dose-dependent effect on inhibition of angiogenesis in vitro, and the effect is similar with sorafenib in the aortic ring assay.

Figure 6.

 Effects of molecular on angiogenesis in the aortic ring assays, (A) mice aortic ring treated with 1 μm molecule 12; (B) mice aortic ring treated with 0.2 μm molecule 12; (C) mice aortic ring treated with 0.04 μm molecule 12; (D) mice aortic ring treated with 1 μm sorafenib; (E) mice aortic ring treated with 0.2 μm sorafenib; (F) mice aortic ring treated with 0.04 μm sorafenib; (G) mice aortic ring treated with 1/1000 DMSO (negative control).

Molecule 12 inhibit tumor cell and HUVEC growth in vitro

To examine the inhibitory effect of the screened molecules on tumor cell and HUVEC growth, ES-2, HCT116, K562, Hela cells, and HUVECs were used for MTT assays. After incubation for overnight on 96-well plates, these cells were treated with compound 12 and sorafenib at concentrations ranging from 6.25 to 100 μm. DMSO was used for negative control. After another 48 h of incubation, adding of MTT, removing of medium and dissolving of formazan, optical density values were detected at last. Molecule 12 had good performance in the MTT assays compared with the positive control in these cell lines (Table 1). The result suggests that molecule 12 that can effectively inhibit HUVEC growth is also a very potent tumor cell growth inhibitor in vitro.

Table 1.   Inhibitory activity (IC50, μm) of molecule 12 and sorafenib on tumor cell and HUVEC growth
CompoundCell
ES-2HCT116K562HelaHUVEC
  1. HUVEC, human umbilical vein endothelial cells.

12 17.48.9012.314.22.7
Sorafenib50.914.420.521.06.4

Molecular dynamic simulation analysis

To probe the binding mode of molecule 12 in the active site of VEGFR2, a dynamic simulation was performed using the gromacs software. The simulation should provide a reliable protein conformation for further analysis. The time-dependent RMSD (Figure 7A) and radius of gyration plots (Figure 7B) in the 10-ns simulation have been convergent at the beginning of the 10-ns production simulation. The ligand also shares very small movements after 850 ps (Figure 7C). The result indicates that the structure of the ligand–protein complex is credible for further structural analysis.

Figure 7.

 Time-dependent RMS (A) and radius of gyration (B) plot and of vascular endothelial growth factor receptor 2 (VEGFR2) and time-dependent RMS (C) plot of molecule 12 in the 10-ns molecular dynamic simulation.

The root mean square fluctuation (RMSF) plot shows that residues in the C-terminal are most flexible, and the loops also share large magnitude of movements (Figure 8). Residues 43, 190, 192, 193–197, 259, 304, and 305 have fluctuation values more than 0.3 nm. All the residues around molecule 12 (residues 51, 53, 70, 71, 73, 74, 84, 86, 99, 101, 161, 163, 168, 179, 181–185, and 201) have very small magnitude of movements (mean RMSF value is 0.09 nm). It suggests that interactions between molecule 12 and surrounding residues are stable.

Figure 8.

 (A) Root mean square fluctuation (RMSF) plot of vascular endothelial growth factor receptor 2 (VEGFR2) in the 10-ns simulation, residues in the active site (residues 51, 53, 70, 71, 73, 74, 84, 86, 99, 101, 161, 163, 168, 179, 181–185, and 201) is represented as black points. (B) The 3D structure of VEGFR2, which is colored by RMSF value. The color is changed from red to white, and to blue with RMSF value increasing.

Hydrophobic interactions make significant contributions to the ligand–receptor binding (Figure 9). The naphthyl group can form strong hydrophobic interaction with surrounding residues such as His161, Asp163, Asp181, Leu184, and Ala185. The naphthalenol group has hydrophobic interactions with His79, Val83, and Glu70. The triazole group is responsible for H-bond interactions. It can form H-bond interaction with NH3 group of Lys53, the mean distance between NAD of molecule 12 and NZ of Lys53 is 0.56 nm in the 10 nm simulation. The naphthalenol group also contributes to the polar interactions with the receptor.

Figure 9.

 Binding mode of molecule 12 in vascular endothelial growth factor receptor 2 (VEGFR2) revealed by molecular dynamic simulation. (A) 3D representation of molecule 12 in the active site; (B) 2D representation of ligand–receptor interaction by Ligplot (21).

In the molecular dynamic simulation process, the water molecules penetrate into the protein structure, and inevitably, they will influence the ligand–receptor interactions. Thus, the influence of the solution was investigated by energy analysis. In our study, the mean value of Lennard-Jones interaction energy between molecule 12 and VEGFR2 is −143.9 KJ/mol during the 10-ns simulation (Figure 10A). The value between molecule 12 and water molecules is 2.7 KJ/mol (Figure 10B). The result suggests that the Lennard-Jones interaction is very strong between molecule 12 and VEGFR2, and the solution has very small influence.

Figure 10.

 Energy plot between molecule 12 and vascular endothelial growth factor receptor 2 (VEGFR2), and the solution. (A) Lennard-Jones interaction energy between molecule 12 and VEGFR2; (B) Lennard-Jones interaction energy between molecule 12 and the solution.

Kim and co-workers (11) have published their results of virtual screening VEGRF2 inhibitors previously. They discovered three potent VEGFR2 inhibitors from testing 100 compounds. In the present study, three potent inhibitors were discovered from testing 16 compounds. And the screened compounds were analyzed by comprehensive investigations such as tube formation assays, aortic ring assays, and MTT assays. Compound 12 has good performance in all these studies. The binding patterns were not only studied by docking method but also investigated by molecular dynamic simulation. The limitation of our results is the lack of enzyme inhibition assays. The activity of compound 12 in inhibition of VEGFR2 and a series of other kinases will be tested in our further work.

Conclusion

To develop novel VEGFR2 inhibitors, a virtual screening approach was performed in discovering lead structures. After ‘rule of five’, pharmacophore-based and docking-based screening, 16 molecules were selected for biological activity assays. The filtered molecules were firstly evaluated by tube formation assays, and a naphthalenol group containing molecule, 12, caught our attention in this study. Thus, aortic ring assays and MTT assays were performed on molecule 12, and it shows good performances in both studies. Then, molecular dynamic simulation was performed to find core features of molecule 12 in binding to VEGFR2 and to provide clues for further structural modification. Hydrophobic interaction formed by the naphthyl and naphthalenol group was approved to make significant contributions to the ligand–receptor binding. H-bond interaction formed by the triazole group also contributes to the binding. Our work emphasizes the prospect and necessity of development potent VEGFR2 inhibitors by structural modifications of molecule 12.

Acknowledgment

This work was supported by National Nature Science Foundation of China (Grant No. 21172134), Ph.D. Program Foundation of Ministry of Education of China (Grant No. 20110131110037), National High Technology Research and Development Program of China (Grant No. 2011ZX09401-015), and Graduate Innovation Foundation of Shandong University (GIFSDU) (No. yyx10020). We also acknowledge Accelrys for providing the Discovery Studio 2.5 package.

Ancillary