TG1 cells were purchased from Lucigen (Wisconsin, America) E. coli DH5α, and E. coli BL21 (DE3) were purchased from Weidibio (Shanghai, China). The restriction enzymes, T4 DNA ligase, and PrimeSTAR^® HS DNA Polymerase were purchased from Takara (Tokyo, Japan). HRP-conjugated anti-HA tag monoclonal antibody was purchased from Sino Biological (Beijing, China). Fetal bovine serum (FBS), DMEM, ECM, and penicillin-streptomycin (10,000 U/mL penicillin and 10 mg/mL streptomycin) were obtained from Thermo Fisher Scientific (Waltham, USA). Antibodies for VE-cadherin (VE-Ca), and β-actin were all purchased from Cell Signaling Technology (Danvers, MA). Anti-CD93 antibody was bought from Santa Cruz (Santa Cruz, CA). Anti-His antibody was bought from Proteintech (Chicago, USA). 488-conjugated Goat Anti-Mouse IgG (H + L) antibody was purchased from ABclonal (Wuhan, China). Other reagents unless indicated were obtained from Sigma Chemical Co. (St. Louis, MO).

Human embryonic kidney cells (HEK293) preserved in our lab were cultured in DMEM (high glucose), supplemented with 10% (v/v) fetal bovine serum (FBS), 100 U/mL penicillin, and 100 mg/mL streptomycin. Human umbilical vein endothelial cells (HUVECs) purchased from ScienCell (Carlsbad, CA) were cultured in ECM supplemented with 5% (v/v) FBS and 1% (v/v) ECGS, 100 U/mL penicillin and 100 mg/mL streptomycin. Cells were incubated at 37 °C in a humidified atmosphere (5% CO2).

The phage library was expanded and then infected with M13 helper phage (10¹² CFU/mL) for 1 h. The samples were centrifuged at 8,000 rpm for 5 min, and the bacteria were resuspended in 200 mL of 2×TY medium (containing 100 μg/mL Amp and 50 μg/mL Kana) and incubated at 37°C overnight. The phage display library was recovered from the supernatant by precipitation with PEG/NaCl (20% PEG and 2.5 M NaCl). The ELISA plates were coated with CD93 (200 nM for the first and second rounds, 100 nM for the third and fourth rounds of biopanning) and incubated overnight at 4°C. After washing, 5% BSA solution (0.1% TBST dissolved) was added and incubated for 1 h at room temperature. After washing, prepared phage libraries were added to the wells and incubated for 2 h. The samples were incubated with a glycine-hydrochloric acid solution (20 mM Tris-HCl and 1% BSA, pH = 2.2) for 10 min at room temperature, collected in EP tubes, and immediately neutralized with Tris-HCl buffer to pH = 7.4. After that, infect the eluted phage with TG1. Phage amplification and titer assays were performed to prepare for the next round of panning.

Monoclonal phage ELISA was selected to determine the positive anti-CD93 Nbs clones. The ELISA plates were coated with 100 nM CD93 and BSA incubated overnight at 4°C. After washing, 5% BSA solution (0.1% TBST dissolved) was added and incubated for 1 h at room temperature. After washing, different phage monoclonals were added to the wells and incubated for 2 h. After washing, HRP-conjugated anti-M13 antibody was added and incubated for 1 h, after discarding the secondary antibody and washing, TMB solution was added to each well and incubated at 37°C for 30 min. At this time, the solution turned blue. After incubation, the reaction was terminated by adding 1M HCl, and the 450 nm absorbance value was measured in an enzyme calibrator. The obtained positive clones were cultured at 37°C for 4 h and sent for sequencing (Qing Ke, China). Protein sequences were translated by Expasy (Duvaud et al., 2021).

The amplified gene product (Nde I-HA-Nb-Xho I) was purified, cleaved with Nde I and Xho I enzymes, and ligated into the pET-24a (+) plasmid. Recombinant vectors were transformed into E. coli BL21 (DE3). Cells were induced with 0.75 mM IPTG at 30°C for 10 h. The collected cells were crushed with the high-pressure homogenizer. Purification with Ni-columns, followed by dialysis desalting of the protein solution is essential. Finally, the protein purity was analyzed by SDS-PAGE electrophoresis.

HUVECs or HEK293 cells (4 × 10³) were seeded into 96-well plates and incubated for 24 h. Then, different concentrations of Nbs (0, 0.0001, 0.01, 0.1, 1, 10, 100, 1,000, and 10,000 nM) were added. After incubating at 37°C for 24 and 48 h, MTT solution (final concentration 0.5 mg/mL) was added and left for 4 h to form blue-purple methanate crystals. Then a triplet (10% SDS-5% isobutanol-0.01 M HCl) was added and left overnight to fully dissolve the crystals. The absorbance values at 570 nm and 630 nm (OD) were read. The percentage of absolute absorbance values between the drug administration group and the control group was determined as the cell viability.

The affinity of anti-CD93 Nb was determined by indirect-ELISA (Zhang et al., 2016). The ELISA plates were coated with 5 μg/mL CD93 and BSA and incubated overnight at 4°C. After washing, 5% skimmed milk solution (0.1% TBST dissolved) was added and incubated for 1 h at room temperature. After washing, different concentrations of Nb were added to the wells and incubated for 2 h. After washing, HRP-conjugated anti-HA antibody was added and incubated for 1 h, After discarding the secondary antibody and washing, TMB solution was added to each well and incubated at 37°C for 30 min. At this time, the solution turns blue. After incubation, the reaction was terminated by adding 1M HCl, and the 450 nm absorbance value was measured in an enzyme calibrator.

HUVECs or HEK293 (4 × 10⁵) cells per sample were suspended in PBS containing 5% BSA and then incubated with/without 10 μM (10 μM, which had reached the plateau phase, was selected for cell binding according to the ELISA assay.) NC81/NC89 at 37°C for 1 h. Resuspend the cells collected by centrifugation. The cells were washed and incubated with anti-His antibody for 1 h. After washing, the cells were incubated with a fluorescent secondary antibody for 1 h. Finally, the cells were washed and the binding assay was performed with a FACS flow cytometer.

Ibidi inserts were fixed in a 24-well plate, and 5 × 10³ HUVECs were placed into a 24-well plate and then starved of serum-free ECM for 12 h. When the cells were grown in the chamber to 70% confluence, the insert was removed vertically and washed 1-2 times with DPBS. Three random photographs were then taken immediately under the microscope noted as 0 h. Complete medium containing different concentrations of Nbs (0, 0.1, 1, 5, and 10 μM) were added to each well. 12 h later, three random photographs of the scratch location of each well were taken with an OlYMPUS inverted microscope. The wound migrated distances were measured with ImageJ.

Change the complete medium to serum-free medium 1 day in advance and incubate for 12 h. Matrigel (BD Biosciences) was thawed on ice for 24 h, then 50 μL Matrigel per well was coated into a 96-well tissue culture plate and incubated at 37 °C for 0.5 h to solidify. 1 × 10⁴ HUVECs were seeded into the 96-well plate and added with 200 μL ECM supplemented with 1% (v/v) FBS, 1% (v/v) ECGS and different concentrations of Nbs (0, 1 and 10 μM). Incubated at 37°C for 3–12 h, and the formation of endothelial tubes was photographed with an inverted OlYMPUS microscope at ×100 magnification. The endothelial tubes were counted with ImageJ.

Samples were separated by SDS-PAGE gel electrophoresis and electrophoretically transferred onto PVDF membranes, and then the membranes were probed with the appropriate combination of β-actin (1:1,000), CD93 (1:200), VE-Ca (1:1,000) and horseradish peroxidase-conjugated secondary antibodies. Proteins in the membranes were visualized by enhanced chemiluminescence kits. The protein bands were quantified by the average ratios of integral optic density following normalization to the expression of internal control β-actin, CD93, or VE-Ca, and the results were further normalized to control.

The protein sequences were submitted to the I-TASSER webserver for homology modeling using default settings. According to the evaluation results, the highest-score model was adopted for molecular docking. HADDOCK software was used to perform molecular docking for NC81/NC89 and CD93 (van Zundert et al., 2016; Honorato et al., 2021). The solvent-exposed residues of CD93 were analyzed by WHATIF tools and the CDR residues of NC81/NC89 were used. Based on the evaluation of the feedback docking clusters, the docking cluster with the smallest Z-score was the best, and the amino acid residues of CD93 interacting with NC81/NC89 were further analyzed by PDBePISA.

Data are expressed as the mean ± standard error of the mean (SEM). Differences between groups were assessed by non-parametric one-way analysis of variance (ANOVA) followed by the least significant difference (LSD) post hoc test when ANOVA found a significance value of F and no variance inhomogeneity. Otherwise, Mann-Whitney U non-parametric ANOVA was performed. p < 0.05 is considered to be statistically significant.

CD93-specific Nbs were screened by solid-phase screening (Zhang et al., 2016), and phage titer assays showed an approximately 100-fold increase in antigen-bound clones in the fourth round compared to the first round (Table 1), indicating a large enrichment of clones bound to CD93 in the fourth round (Figure 1A). From the fourth enrichment, 288 monoclonal phage Nbs were selected and better monoclonal phages of anti-CD93 were selected by monoclonal phage ELISA (Figures 1B–G). As T/N ratios >2.0 were considered positive clones (Yan et al., 2014), a total of nine clones had ratios higher than 2 (Figures 1B–G). All positive clones were sent to the company for sequencing, and the only difference between the 2 Nbs (NC81 and NC89) was the different CDR3 sequences and lengths (Figure 1H).

The Nb gene was correctly inserted into the PET24a (+) vector by polymerase chain reaction (PCR) (Figure 2A). The recombinant protein was expressed in E. coli. BL21 (DE3) after induction by 1 mM IPTG at 30°C for 10 h. After purification by Ni-NTA affinity chromatography, the protein purity was greater than 95% by SDS-PAGE analysis (Figure 2B). In addition, the protein concentration of Nbs was determined, and the yield of Nbs ranged from 2.35 to 8.6 mg/L. The sequence of NC81/NC89 was simulated with I-TASSER and the corresponding tertiary structure of NC81/NC89 was obtained. (Yang and Zhang, 2015). The C-score in the simulation structure is usually between −5 and 2, with the highest C-score value indicating a high confidence level. In addition, a TM-score >0.5 usually indicates the model with the correct topology. As shown in Figures 2C, D, the optimal NC81 model had a C score of −0.81, RMSD of 6.0 ± 3.7 Å, and estimated TM of 0.61 ± 0.14, and the best model of NC89 had a C-score of −0.39, RMSD of 5.1 ± 3.3 Å and estimated TM of 0.66 ± 0.13. These evaluations showed that the obtained NC81/NC89 structural model was highly credible and that the CDR3 regions of both NC81/NC89 were exposed, favoring the binding of CD93 (Figures 2C, D).

To evaluate the affinity of anti-CD93 Nbs, indirect ELISA was performed to calculate the value of the equilibrium association constant (Ka) through four-parameter logistic curve fitting (Figures 3A, B). ELISA results indicated that the Ka values of NC81 and NC89 were 3.54 × 10⁻⁶ M and 2.8 × 10⁻⁶ M, respectively. The diversity of CDR1 and CDR2 of Nbs has a certain effect on antigen-binding function (Muyldermans, 2013). However, CDR1 and CDR2 of Nbs were customized in this designed library and only randomized CDR3 to improve solubility expression, which may have an impact on antibody affinity. Thus, molecular docking was used to analyze whether the fixed CDR1 and CDR2 of Nbs exerted binding function (Lugano et al., 2018). As shown in Figure 3C, the results of FC showed that CD93 on HUVECs was more than twice that on 293T cells. In addition, the binding ability of NC81 and NC89 to HUVECs was examined with His antibody, and the results showed that NC81 and NC89 were able to bind specifically to HUVECs (Figure 3D).

The binding ability of Nbs treated at different temperatures to the antigen was examined by ELISA. As shown in Figures 3E, F, there was no significant change in NC89 and NC81 activity from 30°C to 80°C, and even at 80°C good binding ability was maintained, indicating that the Nbs have good thermal stability.

The Nbs frameworks in this study were derived from shark V-NAR, and normally shark-derived antibodies have high thermal stability. The structural changes of Nbs after different temperature treatments were investigated by circular dichroism (CD). The CD spectra of NC81 (Figure 3G) and NC89 (Figure 3H) treated at different temperatures had positive peaks at 192 nm and negative peaks at 203 nm. The positive peak at 192 nm represents the α-helix and the negative peak at 203 nm represents the β-fold. The results showed that the positive peak at 192 nm and the negative peak at 203 nm did not change significantly under 30°C–70°C treatment, indicating that Nbs still maintained their normal structure at 30°C–70°C. Meanwhile, NC81 showed a slight shift of the positive peak at 192 nm and the negative peak at 203 nm at 80°C, while NC89 maintained a similar structure. This suggests that NC89 may have higher thermal stability than NC81.

As shown in Figures 4A–D, the MTT results showed that anti-CD93 Nbs NC81/NC89 significantly reduced the proliferation of HUVECs in a dose-dependent manner (p < 0.05). However, in 293T cells, there was no significant change in the results of the control and test groups (Figures 4A–D). The results showed that NC81 and NC89 could effectively and specifically inhibit the proliferation of HUVECs.

The effect of anti-CD93 Nbs NC81/NC89 on HUVEC migration was observed by scratch assay. As shown in Figures 4E,F, the gap width of Nbs treated group was more significantly reduced than that of the control group from 0 to 12 h (p < 0.05, p < 0.01, p < 0.001), indicating that NC81 and NC89 could dose-dependently inhibit the migration of endothelial cells. Furthermore, NC89 was able to significantly inhibit HUVEC migration at 1 μM.

Although angiogenesis is a complex process in several kinds of cells, the maturation of endothelial cells into a capillary tube is a critical early step (Carmeliet, 2003; Lutsenko et al., 2003; Ferrara, 2010; Saharinen et al., 2011). To investigate whether the anti-CD93 Nbs NC81/NC89 could inhibit tube formation, HUVECs were cultured on Matrigel in the presence of different concentrations of NC81/NC89 (1 μM and 10 μM) and 30 ng/mL VEGF (Figures 5A–C). As shown in Figures 5A–C, there was obvious tube formation after the addition of VEGF, and capillaries were found to gradually disappear with increasing added NC81/NC89 concentration. Statistical results on tube lengths and numbers of nodes revealed that NC81 and NC89 significantly inhibited angiogenesis in HUVECs in a concentration-dependent manner (p < 0.001).

It has been shown that CD93 increases vascular permeability by inhibiting VE-Ca proteins and thereby increasing vascular permeability (Lugano et al., 2023) (Figure 6A). As shown in Figures 6B–D, the addition of 10 μM NC81 and NC89 to HUVECs significantly reduced the protein expression of CD93 and VE-Ca compared to the control, whereas no significant changes were observed in 293T cells.

In order to study the interaction between NC81/NC89 and CD93, the NC81-CD93 and NC89-CD93 complexes were simulated with HADDOCK. HADDOCK, a semiflexible docking method, is suitable for docking construction of antigens and Nbs (Skottrup, 2017). The extracellular domain of CD93 was selected to dock with solvent-exposed residues of NC81/NC89, and the CD93-NC81/CD93-NC89 complex was modeled by HADDOCK. According to HADDOCK, its Z-score indicates how many standard deviations this structure is from the mean in terms of scores (the more negative the better). The evaluations retrieved from HADDOCK were −2.0 for the Z-score of CD93-NC81 with an RMSD of 0.5 ± 0.3 concerning the overall lowest energy structure, and −2.0 for the Z-score of CD93-NC89 with an RMSD of 0.7 ± 0.4 concerning the overall lowest energy structure, (Figures 7A,B). These results indicate that the complex structure model population obtained has high confidence. The optimal model cluster was assessed by PDBePISA to explore the interaction residues of NC81/NC89 and CD93 (Daniele et al., 2023). The paratopes on IC81 were made up of the following residues: ASP4, TYR8, GLU55, ARG106, TYR3, ARG91, VAL103, ARG106, CYS108 and LYS48. The epitopes on CD93 that bind to NC81 are ARG81, GLY197, GLU277, ASN268, ARG82, ASN265, GLU213, LEU191 and ASN267. The paratopes on NC89 were made up of the following residues: ALA10, ASP107, ARG104, CYS117, and MET115. The epitopes on CD93 that bind to NC89 are AYR262, ARG189, PRO190, ASN265 and ASP296. The interactions between CD93 and NC81/NC89 could be classified into hydrogen bonds and salt bridges. Interestingly, only ASN265 was bound to both NC81 and NC89 among the binding sites of CD93, and in the Nbs not only the CDR3 region but also some amino acids in the framework were involved in binding. (Figures 7C,D).