VEGF-A, VEGF-B, and PlGF are high sequence homology ligands of VEGFR1, they bind similarly to VEGFR1 according to the structure data (PDB: 1FLT, 2XAC, and 1RV6). It has been then demonstrated that Loop 1 of VEGF-A binds mainly to D3 of VEGFR1 (Markovic-Mueller et al., 2017) and VEGFR2 (Brozzo et al., 2012). In our previous study, we designed peptides mimicking different binding epitopes of VEGF-A, VEGF-B, and PlGF (Wang et al., 2017). Cyclic peptides derived from Loop 1 showed very interesting inhibition activity to disrupt the interaction of VEGF-A/VEGFR1 in a VEGF-A/VEGFR1 interaction-based ELISA assay (Table 1). Although most of reported inhibitors target VEGFR1 D2, Wiesmann et al had shown by domain deletion analysis that the deletion of D3 from VEGFR1 D1-D2-D3 induce a loss of more than 20-fold VEGF-A binding affinity (Wiesmann et al., 1997). In our reported study, cyclic peptides derived from Loop1: A-cL1, B-cL1, and P-cL1 showed higher activity than those derived from the helix α1 or the loop 2 or the loop 3. We thus focus this study on these three peptides (sequences in Table 2, formulas in Supplementary Scheme S1).

Peptides A-cL1, B-cL1, and P-cL1 were respectively designed from Loop 1 of VEGF-A, VEGF-B, and PlGF, which were supposed to bind mainly to the domain 3 (D3) of VEGFR1. The docking results showed that A-cL1 and P-cL1 had similar interaction mode with VEGFR1, mainly binding to VEGFR1 D3 (Figures 2A,C). However, B-cL1, which had larger cycle, binds differently to VEGFR1. B-cL1 binds to the interface of VEGFR1 D2 and D3 (Figure 2B). The ten highest score peptide structures were superimposed in the binding with VEGFR1 (Supplementary Figures S1–S3). The major interactions between A-cL1, B-cL1, P-cL1, and VEGFR1 were analyzed according to the docking (Table 2).

VEGFR1 and VEGFR2 have very similar structures. The crystal structure of VEGF-A/VEGFR1 D1-D6 (PDB: 5T89) and the crystal structure of VEGF-A/VEGFR2 D2-D3 (3V2A) showed the same binding epitopes on VEGF-A, which suggests that the epitope mimic peptides of VEGF-A are normally able to bind both VEGFR1 and VEGFR2. An in vitro inhibitor selection assay was previously developed in our laboratory based on the interaction of VEGF-A/VEGFR1 (Goncalves et al., 2007b). Inhibitors selected from this assay that are able to inhibit the interaction of VEGF-A/VEGFR1, should probably inhibit as well as the interaction of VEGF-A/VEGFR2 and their downstream signaling. The investigation of A-cL1, B-cL1, and P-cL1 induced in vitro anti-angiogenic effects was performed by cell proliferation, cell migration and tube formation in HUVECs.

A-cL1, B-cL1, and P-cL1 having different VEGFR1 binding affinity showed dose-dependent different inhibitory effects in HUVECs migration and tube formation. In order to verify whether the different activity of A-cL1, B-cL1, and P-cL1 could result in different intensity of downstream signal transduction, we performed western blot analysis of the p-ERK1/2 level in the MAPK/ERK1/2 signaling pathway, which is a proven signal pathway induced by the interaction of VEGF/VEGFRs (Shibuya, 2011). After HUVECs incubation with A-cL1, B-cL1, and P-cL1 at different concentrations (1 μM, 5 and 25 μM) in the presence of VEGF-A (0.2 μg/ml), the p-ERK1/2 level was determined in cell lysate. As shown in Figure 6, bevacizumab (6.5 μM) had the most important inhibition of p-ERK1/2 formation. A-cL1 and B-cL1 significantly decreased the level of p-ERK1/2 in a dose-dependent manner, while P-cL1 showed a limited decrease of p-ERK1/2 formation compared to the control.

Among the three peptides, B-cL1 showed the most important in vitro anti-angiogenic activity. In order to further study the in vivo anti-angiogenic activity of B-cL1, we performed an evaluation in a chick embryo chorioallantoic membrane (CAM) model, which is a common physiological and pathological angiogenesis study model in vivo (Rezzola et al., 2020). As shown in Figure 7A, on day 9, capillaries are well formed on the surface of the chick embryo chorionic epithelium. After treatment with different concentrations of B-cL1 (1 μM, 5 and 25 μM) for 48 h, formation of new capillaries was inhibited compared with control treated by PBS (Figure 7A). At 1 μM of B-cL1, short capillaries formation was still observed. However, the increase in thickness of blood vessels was inhibited. At 5 μM of B-cL1, both the formation of new capillaries and the development of blood vessels thickness were inhibited. At 25 μM of B-cL1, destruction of pre-existing capillaries was observed. AngioTool (Zudaire et al., 2011), a software for quantitative analysis of angiogenesis, which allows to quantifier the percentage of capillary area in total area (Klotz et al., 2015; Sawaguchi et al., 2017), was used to analyze the increasing or decreasing vessels area after treatment. The results showed that B-cL1 was able to inhibit new capillaries formation in a dose-dependent manner (Figure 7B).

Gastric cancer is one of the most common malignancy that threaten the health of human beings, and the human gastric cancer SGC-7901 cell line is widely used in anti-tumoral studies in cellular level and on xenografted mouse model (Lv et al., 2019). This kind of tumor is largely explored in clinical trial by anti-angiogenic approach, including anti-VEGF, which have shown some benefits in second line treatment (Park et al., 2015). Consequently, anti-tumoral activity of B-cL1 was evaluated by VEGF-induced (50 ng/ml) human gastric cancer SGC-7901 cell proliferation and on BALB/c mice subcutaneous xenograft of human gastric cancer SGC-7901 cells.

B-cL1 showed a dose-dependent inhibition of human gastric cancer SGC-7901 cell proliferation at five concentrations (0.2, 1, 5, 25 and 125 μM), no cytotoxicity was observed event at 125 μM (Figure 8A). It showed similar anti-proliferation activity at 25 μM as bevacizumab at 6.5 μM. Meanwhile, B-cL1 (5 mg/kg/day and 10 mg/kg/day) and bevacizumab (5 mg/kg, only once) were intravenously administrated for 2 weeks on BALB/c nude mice subcutaneous xenograft of SGC-7901 cells, starting when the tumor volume arrived at 150–300 mm³. Tumor volume and mice body weight were measured every 2 days. After 14 days of treatment, the mice were euthanized using CO₂ followed by cervical dislocation to ensure death and the tumors were separated (Figure 8B). The results showed that B-cL1 (5 mg/kg/day) reduced 60% of tumor weight, 53% of tumor volume, compared to control group; B-cL1 (10 mg/kg/day) showed 62% reduction of tumor weight, 51% reduction of tumor volume, compared to control group. As a positive control, bevacizumab (5 mg/kg) reduced 57% of tumor weight and 57% of tumor volume with only one administration (Figures 8C,D). It is worth to indicate that no mortality of mice was observed during the 2 weeks of treatment, and the body weights of mice was increased reasonably (Figure 8E), which suggest weak toxicity of B-cL1 for an administration up to 10 mg/kg/day.

The design and synthesis of peptides derived from different epitopes of VEGFR1 binding sites on VEGF-A, VEGF-B, and PlGF have been reported in our previous work (Wang et al., 2017). Briefly, to mimic the loop structure, the original peptide sequences have been modified and then cyclized through disulfide bonds. To facilitate the reading, we rename these peptides (peptide 14, 18, and 19 in previous work) as described in the Table 1.

Protein Preparation Wizard Workflow program (Gayatri et al., 2016) provided in Maestro 9.0 was used to build the structures of peptides A-cL1, B-cL1, and P-cL1 based on the original crystal structures of VEGF-A, VEGF-B, and PlGF with VEGFR1-D2 (PDB code: 1FLT, 2XAC, 1RV6). Peptides were built by mutating underlined residues to cysteine residues and formed disulfide bond on the side chains (Table 1). They were then aligned and docked near the binding pocket of Loop1 on VEGFR1 (using complex structure of VEGF-A with VEGFR1 D1-D6, PDB: 5T89) in Pymol to generate the initial PDB complex file of A-cL1/5T89, B-cL1/5T89 and P-cL1/5T89. Afterwards, the peptide-protein complexes were adapted to optimize the side chains, in order to eliminate internal collision that are not related to intermolecular interactions in Rosetta FlexPepDock program (Raveh et al., 2010). At last, the docking was performed with optimized structures of the three peptides on Rosetta online server (http://flexpepdock.furmanlab.cs.huji.ac.il/index.php) (Raveh et al., 2010; London et al., 2011). Ten structure models of each peptide with highest score were obtained among 200 generated structures (Supplementary Figures S1–S3) and were output for binding analysis.

Human umbilical vein endothelial cells (HUVECs), of passage 3 to 6, were purchased from Shanghai Fuheng Biotechnology Co. LTD. and cultured in endothelial cell culture medium (ECM: Gibco, Life Technologies, United States), with 10% fetal bovine serum (FBS: Gibco, Life Technologies, United States), 100 U/mL penicillin and 100 mg/ml streptomycin in an incubator at 37°C and containing 5% carbon dioxide (CO₂). Human gastric cancer SGC-7901 cells, from Cell Bank of the Chinese Academy of Sciences (Shanghai, China), were cultured in Roswell Park Memorial Institute (RPMI) 1,640 medium (Gibco, Life Technologies, United States), with 10% fetal bovine serum (FBS: Gibco, Life Technologies, United States), 100 U/mL penicillin and 100 mg/ml streptomycin in an incubator at 37°C and containing 5% carbon dioxide (CO₂).

HUVECs (3 × 10³) were seeded on the 96-well plate per well in ECM containing 5% FBS, and incubated overnight at 37°C, with 5% CO₂. Then the medium was removed. New serum-free medium containing 0.2 μg/ml VEGF-A (R&D Systems, United Kingdom) with different concentrations of peptides (0.2, 1, 5, 25 and 125 μM), bevacizumab (Roche, Switzerland) at 6.5 μM (1 mg/ml) or control group (serum-free medium) were added and incubated for another 48 h (6 wells/concentration/group). The effects of proliferation were quantified by Cell Counting Kit-8 (CCK-8, Sigma, United States) assays. The absorbance was measured at 450 nm with an enzyme-linked immunoassay reader (AMR-100, ALLSHENG, Hangzhou, China). The experiment was repeated three times.

The six-well plate was coated with HUVECs and incubated with ECM containing 10% FBS until the confluence of monolayer cells reached 100%. A scratching wound was created with 10 μl pipette tip. Then the medium was removed. New medium containing only 2% serum and 0.2 μg/ml (5 nM) VEGF-A (R&D Systems, United Kingdom) with different concentrations of peptides (1 μM, 5 and 25 μM), bevacizumab (1 mg/ml, 6.5 μM) or 2% serum medium (control group) were added and incubated at 37°C, with 5% CO₂ for another 12 h. Cell migration was observed and photographed at 6 and 12 h under a microscope (TE-2000U, NIKON, Japan). The results were calculated from three independent experiments with four replicates. Wound area (the open area of the scratch) was quantified using ImageJ, and the percentage of wound closure were calculated: [1 - (wound area at 6 h or 12 h/wound area at 0 h)] × 100%.

Growth factor-reduced basal membrane extract (BD Biocoat^TM, 356,230, United States) was melted overnight in refrigerator at 4°C, and precoated on the 96-well plate. The 96-well plate was then incubated at 37°C for 1 h until Matrigel solidified. 5 × 10⁴ HUVECs/well, pre-treated with different concentrations of peptides (1 μM, 5, and 25 μM), bevacizumab (6.5 μM) or 2% serum medium (control group) were added to the plate and incubated at 37°C for 6 h. Resulting tube networks were observed and photographed by using the microscope (TE-2000U, NIKON, Japan). The numbers of cell branches were quantified in a blind manner. The results were calculated from three independent experiments with four replicates.

HUVECs treated with different concentrations of peptides (1 μM, 5, and 25 μM), bevacizumab (6.5 μM) or serum-free medium (control group) were lysed with RIPA lysis buffer (Beyotime, Shanghai, China) containing protease and phosphatase inhibitors (Sigma, United States). After centrifugation at 4°C, protein concentration was determined by bicinchoninic assay (BCA) (Beyotime, Shanghai, China). Polyacrylamide gel electrophoresis was performed with the same amount of protein of different groups. The protein on the gel was then transferred to the PVDF membrane (Millipore, United States), and the membrane was blocked with skimmed milk powder in Tris-buffered saline (TBS, 5%) for 2 h. Later, the membrane was washed with TBST (1 × Tris Buffered saline, 0.1%Tween20), and incubated with the primary antibody (anti-phospho ERK1/2, anti-ERK1/2, and anti-GAPDH as control, Cell Signaling Technology, United States) and TBST overnight at 4°C. The membrane was washed 3 times with TBST, and then incubated with horseradish peroxidase-conjugated goat anti-mouse IgG secondary antibody and goat anti-rabbit IgG secondary antibody for 1 h at room temperature. At last, proteins were analyzed by ECL reagent (WesternBright ECL, United States) following the manufacturer’s protocol, after 3 times washing with TBST. The experiment was repeated three times. Protein expression was quantified by ImageJ.

A chick chorioallantoic membrane (CAM) assay was carried out to determine the in vivo anti-angiogenic activity of peptide B-cL1. Two-days old fertilized specific pathogen free (SPF) eggs (Ningbo Chunpai Agricultural Science and Technology Co. LTD., Ningbo, China) were incubated at 37.5°C, 60–70% relative humidity for 1 week. Blood vessels in SPF eggs were observed within lighting. Then a window of about 2 cm² was opened on the eggs and a sterilized silica gel ring (1.5 cm inner diameter) was placed onto the CAM. 20 μL of different concentration of B-cL1 (1 μM, 5, and 25 μM in PBS) or PBS (control group) were added inside the silica gel ring (10 eggs per group). The window was recovered with sterilized tape, and the eggs were incubated for another 48 h. The sterilized tape was removed, and the area of capillary blood vessels photographed under a stereomicroscope (Soptop, Shanghai, China). The experiment was repeated three times, and the percentage of capillary area in total analyzed area was calculated using AngioTool (Zudaire et al., 2011). The percentage of increasing/decreasing capillary area was then calculated: percentage of capillary area at 48 h - percentage of capillary area at 0 h.

Human gastric cancer SGC-7901 cells (2 × 10³) were seeded on the 96-well plate per well in RPMI 1640 medium containing 2% FBS, and incubated overnight at 37°C, with 5% CO₂. Then the medium was removed. New serum-free medium containing 50 ng/ml VEGF-A (R&D Systems, United Kingdom) with different concentrations of peptides (0.2, 1, 5, 25, and 125 μM), bevacizumab (Roche, Switzerland) at 6.5 μM (1 mg/ml) or control group (serum-free medium) were added and incubated for another 48 h (6 wells/concentration/group). The effects of proliferation were quantified by Cell Counting Kit-8 (CCK-8, Sigma, United States) assays. The absorbance was measured at 450 nm with an enzyme-linked immunoassay reader (AMR-100, ALLSHENG, Hangzhou, China). The experiment was repeated three times.

The evaluation of antitumoral activity of B-cL1 was performed in Laboratory of Experimental Animal Science, Hangzhou Normal University (Hangzhou, China) maintained under standardized environmental conditions, with approved protocols by the Institutional Animal Care and Use Committee (IACUC) of Hangzhou Normal University. Human gastric cancer SGC-7901 cells (1 × 10⁶ cells/500 μL) from Cell Bank of the Chinese Academy of Sciences (Shanghai, China) were injected to the right flanks of 8 weeks old female BALB/c mice. When tumor size grown to 100–300 mm³, BALB/c mice were randomized to groups (n = 6). B-cL1 (5 mg/kg/day and 10 mg/kg/day) and PBS (control group) were intravenously administrated for 2 weeks, bevacizumab (5 mg/kg) was intravenously administrated only once during the 2 weeks as positive control. The tumor volume was measured every 2 days by a digital Vernier caliper, using the following formula: v = a² × b × 0.52 (where a is the shortest diameter of tumor and b is the longest diameter of tumor). After 14 days of treatment, the mice were euthanized using CO₂ followed by cervical dislocation to ensure death. Then, the tumors were separated, photographed and weighed on balance (OHAUS Adventure, United States).

Data are expressed as the arithmetic mean ± SEM of at least three different experiments using the GraphPad Prism software version 8.00 (San Diego, United States). The statistical significance of results was evaluated by one-way ANOVA, with probability values *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, being considered as significant, “ns” meaning non-significant.