Tie2-mediated vascular remodeling by ferritin-based protein C nanoparticles confers antitumor and anti-metastatic activities

Bevacizumab was purchased from (InvivoGen, California, USA) and cisplatin (cis-diammineplatinum (II) dichloride, Cis), isopropyl-β-d-1-thiogalactopyranoside (IPTG), imidazole, and dithiothreitol (DTT) were purchased from Sigma-Aldrich (St Louis, MO, USA). Tris-HCl and Triton X-100 were purchased from Amresco (Solon, OH, USA), sodium chloride and urea from Junsei Chemical (Tokyo, Japan), and phenylmethane sulfonyl fluoride (PMSF) and protease inhibitor cocktail tablets from Roche (Basel, Switzerland). Mouse monoclonal antibodies against α-smooth muscle actin (SMA), pimonidazole, and cytokeratin II were purchased from Abcam (Cambridge, UK), HPI (Burlington, MA, USA), and Millipore (Burlington, CA, USA), respectively. Antibodies against CD31 and LYVE-1 were obtained from BD Biosciences (San Jose, CA, USA) and Angiobio (San Diego, CA, USA), respectively. Alexa Fluor 488-, 568-, and 647-conjugated antirabbit IgG and Hoechst 33258 were obtained from Invitrogen (Carlsbad, CA, USA).

The TFG/TFMG vectors were constructed, expressed, and purified as previously reported [11]. Briefly, TFG/TFMG plasmids were transformed into Escherichia coli expression strain BL21 (DE3) and grown in LB medium containing 50 μg/ml kanamycin. Protein expression was induced with 0.1 M IPTG at 37 °C for 5 h, and the cells were lysed (lysis buffer: 50 mM Tris-HCl pH 6.8, 100 mM NaCl, 1% Triton X-100, 1 mM PMSF, 1 mM DTT, protease inhibitor cocktail) and ultrasonicated. Binding was performed (urea binding buffer: 20 mM Tris-HCl pH 6.8, 300 mM NaCl, 10 mM imidazole, 8 M urea) at room temperature for 1 h. The sample was loaded onto nickel ion chelate affinity columns, washed (washing buffer: 20 mM Tris-HCl pH 6.8, 500 mM NaCl, 30 mM imidazole, 8 M urea), and eluted (elution buffer: 20 mM Tris-HCl pH 6.8, 100 mM NaCl, 300 mM imidazole, 8 M urea). Finally, sequential dialysis (dialysis buffer: 20 mM Tris-HCl pH 6.8, 100 mM NaCl, and sequentially diminishing concentration of urea) using a dialysis cassette (MWCO, 10 kDa; Thermo Scientific, Rockford, IL, USA) was conducted to obtain refolded proteins.

SDS-PAGE gel staining assay was performed to evaluate the purity of TFG and TFMG following protein purification. First, the protein samples were separated by gel electrophoresis using a 10% SDS-PAGE gel, followed by washing and visualization of the protein bands by staining with a solution containing Coomassie brilliant blue R250 dye (0.125 g), glacial acetic acid (5 ml), methanol (25 ml), and water (20 ml), followed by destaining. Photographic images of the gels were captured to record the protein bands.

LLC and EA.hy926 (human endothelial somatic cell hybrid) cells were obtained from the American Type Culture Collection (Manassas, VA, USA), and C3H/10T1/2 (clone 8) cells were obtained from the Korean Cell Line Bank (Seoul, South Korea). All of the cells were grown in DMEM (Hyclone, Logan, UT, USA) supplemented with 10% FBS (Hyclone, Logan, UT, USA) and 1% antibiotics (100 units/ml penicillin, 100 mg/ml streptomycin, and 0.25 mg/ml amphotericin B; Invitrogen, Carlsbad, CA, USA) at 37 °C in a humidified atmosphere containing 5% CO₂ and 95% air.

The mice used in this study consisted of 5- to 6-week-old adult male C57BL/6J mice (SLC, Japan) or MMTV-PyMT transgenic mice [20, 21] (Jackson Lab, Sacramento, CA, USA). All mice were housed in a specific pathogen-free facility with individual ventilation systems at Kyungpook National University. The animal handling and experimental procedures were conducted strictly according to the Guidelines for Care and Use of Laboratory Animals issued by the Institutional Ethical Animal Care Committee of Kyungpook National University (Protocol number: KNU 2014-0189 and KNU 2017-0145).

After intravenous injection of TRAP (10 nmol/kg) into the tail vein of mice, plasma samples (30 μl) were obtained at 0.5, 1, 1.5, and 2 h. Plasma samples (30 μl) were vigorously mixed with 100 μl acetonitrile and centrifuged at 16,000 g for 10 min. Aliquots (20 μl) of the supernatant were injected into an Agilent 6470 Triple Quadrupole mass spectrometer equipped with an Agilent infinity 1260 high-performance liquid chromatograph (Agilent, Wilmington, DE, USA). TRAP was monitored using multiple reaction modes at m/z 761.6 → 602.5 in the ionization mode with a collision energy of 35 eV. After intravenous injection with DyLight 680 NHS ester-conjugated TFMG (10 nmol/kg) into the tail vein of LLC allograft tumor-bearing mice, plasma samples (30 μl) were obtained at 0.5, 2, 4, 8, 24, and 48 h. The fluorescence of the TFMG in plasma samples was measured at excitation and emission wavelengths of 685 nm and 715 nm, respectively. The calibration standards for TRAP and TFMG in mouse plasma were linear in the range of 0.28–72 nM with a correlation coefficient of over 0.999. The pharmacokinetic parameters were calculated using the non-compartmental model in WinNonlin (version 5.1; Pharsights, Cary, NC, USA).

To generate a LLC allograft tumor model, suspensions of LLC cells (4 × 10⁵ cells in 200 μl of serum-free culture media) were implanted subcutaneously into the dorsal flank regions of 5- to 6-week-old male C57BL/6J mice. At the indicated time points following LLC cells inoculation, the mice were administered TFG (10 nmol/kg) or TFMG (10 nmol/kg), with or without cisplatin (3 mg/kg), or bevacizumab (5 mg/kg) by intravenous injection into the tail vein. Subsequently, tumor volumes were measured based on the formula: V = 0.5 × A × B², where A represents the longest and B the shortest dimension. Eventually, the mice were anesthetized and tissues were harvested for further analyses at the completion of the study period.

Female MMTV-PyMT transgenic mice were employed as another model to observe tumor growth and metastasis. Twelve weeks following their birth, the volumes of every palpable tumor nodule were measured and added to determine the tumor burden for each mouse. Based on tumor burden, the mice were divided into three groups, each receiving TFG (10 nmol/kg), TFMG (10 nmol/kg), or vehicle (control) at the indicated time points. Tumor burdens after each week following the 12th week were measured up to the 15th week for each of the three groups. At the 15th week, the mice were anesthetized and tissues were harvested for further analyses.

For immunohistochemical (IHC) analysis, the harvested tumor or tissue samples were fixed in 3.7% formaldehyde (FA), dehydrated by serially incubating with 10% to 40% sucrose solution, and finally embedded in tissue freezing medium (Leica, Wetzlar, Germany). The frozen blocks were cut into 5 to 10 μm sample sections, which were blocked with 5% BSA in PBST (0.3% Triton X-100 in phosphate-buffered saline (PBS)) and incubated overnight at 4 °C with the appropriate primary antibody: anti-CD31 (BD Biosciences, CA, USA), anti-CD31 (Milipore, MA, USA), anti-α-SMA (Abcam, Cambridge, UK), antipimonidazole (Hypoxyprobe-1, HPI, MA, USA), anticytokeratin II (Millipore, MA, USA), anti-LYVE-1 (Angiobio, CA, USA), anti-Phospho-Tie-2 (R&D systems, MN, USA), anti-PAR1 (Abcam, Cambridge, UK), anti-PAR3 (Santa Cruz, CA, USA), anti-CD68 (Abcam, Cambridge, UK), anti-iNOS (Abcam, Cambridge, UK), anti-arginase1 (BD Biosciences, CA, USA), anti-CD4, PE (Invitrogen, CA, USA), or Alexa Fluor 647 anti-CD8a (BioLegend, CA, USA). After a few washes, the samples were incubated at RT for 1 h with the recommended fluorophore-tagged secondary antibody, their nuclei stained with Antifade Mounting Medium with DAPI (Vector Laboratories, Burlingame, CA) or Hoechst 33258 (Invitrogen, USA), and mounted with fluorescent mounting medium (Sigma, USA). The immunofluorescence images were acquired using a Leica TCS SP5 II confocal microscope (Leica, Wetzlar, Germany) or a Zeiss confocal microscope (Carl Zeiss, Germany). Pimonidazole hydrochloride (60 mg/kg; Hypoxyprobe-1, HPI, USA) was injected intraperitoneally 90 min before sacrifice of mice bearing tumors used for the investigation of hypoxic regions. Serial tissue sections (5 μm) were stained with H&E and observed under an Axio Imager A1 light microscope (Carl Zeiss, Germany).

Tumor tissues were fixed with 4% paraformaldehyde in PBS at 4 °C for 24 h and frozen sections were prepared. The terminal deoxynucleotidyl transferase-mediated dUTP nick-end green fluorescent labeling (TUNEL) assay was performed with sectioned tissues (5–10 μm thickness) according to the manufacturer’s instructions (TUNEL assay kit, Promega, Madison, WI, USA). Fluorescent images were observed by fluorescence microscopy (Axio Imager A1 microscope Carl Zeiss, Germany).

TFMG was labeled with Dylight 680 NHS ester at a molar ratio of 1:10. Briefly, TFMG (1 mg) was dissolved in PBS (1.5 ml) and Dylight 680 NHS ester (1 mg) was dissolved in DMSO (0.1 ml). TFMG and fluorescent dye were incubated for 2 h at room temperature. The reaction product was passed through a 0.2-μm filter unit, and the unreacted dye was separated on a PD midiTrap™ G-25 (GE Healthcare, UK) column that had been pre-equilibrated with PBS and 2 mM sodium azide. This process yielded 300 μM of TFMG with a ratio of dye per protein greater than 1.5.

EA.hy926 and C3H/10T1/2 cells were transfected with 10 to 200 nM of siRNA for PAR-1, PAR-2, PAR-3, Tie2, G_α13, G_αi, or G_αq (Bioneer, Daejeon, South Korea) using Lipofecatmine RNAiMAX. Silencing of the genes was confirmed by quantitative real-time polymerase chain reaction (qRT-PCR) using the primers shown in Table S2. After total RNA was extracted using a Trizol RNA extraction kit (Invitrogen, California, USA), cDNA was reverse-transcribed using ReverTra Ace® qPCR RT Master Mix (TOYOBO, Osaka, Japan). Quantitative real-time PCR was performed with SsoAdvanced Universal SYBR® Green Supermix (Bio-rad, Hecules, CA, USA) (Bio-rad C1000 Thermocycler). Relative expression to β-actin was calculated by the ΔΔCt method (Livak and Schmittgen 2001).

Endothelial cell/cancer cell (EC) co-cultures, pericyte/cancer cell (PC) co-cultures, and pericyte/endothelial cell/cancer cell (PEC) co-cultures were established using EA.hy926, LLC, and C3H/10T1/2 cells. For the EC co-cultures, LLC cells (2.5 × 10⁵ cells/well) were first seeded in 12-well plates and allowed to adhere for 4 h, after which 12-well Transwell polycarbonate membrane inserts (3 μm; Corning, NY, USA) were placed above each well. EA.hy926 cells (5 × 10⁵ cells/well) were seeded over the Transwell membrane inserts and incubated at 37 °C for 48 h. For PC co-cultures, LLC (2.5 × 10⁵ cells/ well) cells were seeded onto the 12-well plate surface, while C3H/10T1/2 cells (2.5 × 10⁵ cells/well) were seeded above the Transwell membrane inserts. For the PEC co-cultures, LLC (2.5 × 10⁵ cells/ well) cells were seeded onto the 12-well plate surface, and C3H/10T1/2 cells (2.5 × 10⁵ cells/well) were added to the lower surface of the Transwell membrane inserts and was followed by the addition of EA.hy926 cells (5 × 10⁵ cells per well) onto the upper surface of the Transwell membrane inserts. The other procedures remained the same.

EC, PC, and PEC co-cultures were used to conduct permeability assays. The EC, PC, and PEC co-cultures were treated with TFG or TFMG and incubated at 37 °C for 24 h. Transwell membrane inserts were washed with PBS and transferred to fresh 12-well plates containing 500 μl of serum-free DMEM per well. FITC-dextran (150 μl; 50 μg/ml; MW = 3 kDa; Sigma-Aldrich, MO, USA) was added to each of the Transwell membrane inserts, and leakage across the Transwell membrane was determined using a fluorescence plate reader (λ_ex, 485 nm; λ_em, 535 nm).

EA.hy926 cells (1 × 10⁶) were seeded in 100 mm cell culture plates, incubated overnight at 37 °C, and treated with 100 nM TFG or TFMG for 24 h. The cells were then washed with PBS, detached, and lysed with RIPA buffer solution. An appropriate amount of protein was mixed with anti-PAR-3 antibody and incubated overnight at 4 °C under constant rotation. Magnetic beads were added and the mixture was rotated at 4 °C for 4 h to permit binding. After washing and eluting, immunoblots were recorded to visualize the co-immunoprecipitated proteins.

EA.hy926 cells were seeded into 6-well plates, incubated overnight at 37 °C, and treated with 100 nM TFG or TFMG for a suitable duration. The cells were then washed with PBS, detached, and lysed with RIPA buffer solution. Equal amounts of proteins were separated using SDS-PAGE gels and transferred to PVDF membranes. After blocking, the membranes were incubated with antibodies against p-Tie2 (1:1000), Tie2 (1:200), p-Akt (1:1000), Akt (1:1000), FoxO1 (1:1000), p-FoxO3a (1:1000), FoxO3a (1:1000), ROCK-1 (1:1000), ZO-1 (1:1000), or β-actin (1:5000) at 4 °C for 12 h. Membranes were washed three times for 10 min and incubated with a 1:3000 dilution of anti-mouse, anti-goat, or antirabbit antibodies for 2 h. The Western blots were visualized by chemiluminescent detection.

EA.hy926 cells (2 × 10⁵ cells/well) were seeded onto glass coverslips in 4-well plates, incubated overnight at 37 °C, and treated with 100 nM TFMG for 24 h. The cells were then washed with PBS, fixed with 3.7% formaldehyde (Sigma-Aldrich, St Louis, MO, USA), and permeabilized with 0.1% Triton X-100 (Amresco, Solon, OH, USA). After 1 h of blocking with 3% BSA, the cells were incubated overnight at 4 °C with the required primary antibody solution in 0.1% BSA. After washing with PBS the following day, the cells were incubated at room temperature for 1 h with the appropriate fluorescence-tagged secondary antibody solution in 0.1% BSA. The coverslips were mounted onto glass slides using mounting gel and the images were visualized and recorded using a Leica TCS SP5 II confocal microscope (Leica, Wetzlar, Germany).

Statistical analysis was carried out using Graph Pad Prism 7.0. Significant differences for individual pairs of means were determined using Student’s t test and for lager datasets comparing more than 3 groups, one-way ANOVA was used followed by post hoc Dunnett’s multiple comparison test where appropriate to analyze the level of significance, with p values < 0.05 considered statistically significant. Two-way ANOVA was used when there is more than one independent variable and multiple observation for the mean differences between groups. All the data were expressed as means ± standard deviation (SD) of at least 3 replicates. For the analysis of synergism of combination therapy, the Bliss independence model was utilized [22]. The expected response tumor volume (V_Exp) for the combination groups was defined by the formula: V_Exp = (V₁ × V₂)/V_C, where V₁ and V₂ are the mean tumor volumes in the single-treatment group and V_C is the mean volume in the control group. The tumor volume change from the baseline of control group was defined as ΔV = V_C − V₀ (initial volume). Then, the range around the V_Exp was calculated, via upper (U) and lower (L) limits, as follows: V_ExpU = V_Exp + 0.15 × ΔV and V_ExpL = V_Exp − 0.15 × ΔV. If V_Obs is less than V_ExpL, it indicates a synergistic effect, and if V_Obs is in-between V_ExpL and V_ExpU, it denotes an additive effect. V_Obs: the observed mean tumor volume in the combination groups.

We constructed short ferritin (sFn) by deleting the short helix E and loop from the human ferritin light chain, and produced the TRAP-ferritin-PC-Gla (TFG) protein by inserting the EPCR ligand (PC-Gla domain) at the C-terminus and PAR-1 activator (TRAP peptide) at the N-terminus to the genetically engineered sFn (Fig. S1A-S1B) [11]. A matrix metalloproteinase (MMP)-2 cleavage site in-between the sFn and PC-Gla domains was alternatively inserted to enable the release of PC-Gmla from the nanocages at MMP-2-activating sites, called TFMG (TRAP-sFn-MMP-2-PC-Gla). Thus, the potential interference of PC-Gla with remaining TRAP-sFn during simultaneous double binding to respective receptors was avoided. We reproducibly generated highly pure PCNs showing strong TFG and TFMG bands, corresponding to their molecular weights of approximately 27 kDa and 27.5 kDa, respectively, whereas no secondary bands were observed throughout the gel other than the specific TFG and TFMG bands, indicating a high purity (Figure S1C). We used the same constructs produced in our previous study in which we already determined particle size by dynamic light scattering analyses as well as TEM (Fig. S1D) [11]. The mean outer diameters of TFG and TFMG were 17.0 and 18.5 nm, respectively (Fig. S1E), which were increased compared with the size of wild-type ferritin (6.9 nm) [23].

We investigated the plasma stability of the TRAP domain and the pharmacokinetic properties of TFMG in LLC tumor-bearing mice (Fig. 1a). The TRAP peptide alone was unstable in the mouse plasma with a half-life of 9.5 min (Fig. 1b). These results suggest that when administered in vivo, TRAP itself could not stably initiate pharmacological activity. However, TFMG showed a much longer half-life (13.83 ± 1.08 h) and plasma TFMG concentration was maintained over 0.6 nM for 48 h following intravenous injection of TFMG (10 nmol/kg) (Fig. 1c). The pharmacokinetic parameters calculated from the plasma concentration profile of TFMG are shown in Table S1. Collectively, the pharmacokinetic results and plasma stability for TFMG suggest an advantage of TFMG nanoparticles as anticancer therapeutics compared with the use of TRAP peptide alone.

To investigate the effects of PCNs on tumor growth, we used an LLC allograft tumor model, in which TFG (10 nmol/kg), TFMG (10 nmol/kg), or vehicle (control) was administered at days 7, 10, and 13 following LLC cell inoculation (Fig. 1d). We found that both TFG and TFMG significantly suppressed tumor growth (Fig. 1e) and prolonged survival (Fig. 1f) compared with the controls. In addition, the PCNs exhibited good delivery efficacy to the tumor mass and various organs (Fig. S2A), whereas they did not induce hemorrhage in the internal organs or the tumor masses (Fig. S2B and S2C). Thus, we examined the effects of TFG/TFMG on apoptosis in the major organs by performing TUNEL assays (Fig. S2D). Apoptotic cells at the kidney and heart were not visible in both vehicle control and treatment group, but reduced apoptotic cells were observed at the liver, lung, and spleen at TFG/TFMG-treated group compared to the control group. To determine the maximal effects of the PCNs, we examined various administration routes including intravenous, subcutaneous, and intramuscular routes in the LLC allograft tumor model (Fig. S2E). We found that TFMG administration by an intravenous route exhibited higher antitumor effects compared with other administration routes without causing reduced body weight or tissue damage (Fig. S2F-S2H). Based on these results, we administrated TFMG by an intravenous route in subsequent mouse model studies. These results suggest that TFG and TFMG have antitumor effects with good delivery efficacy in vivo.

We also examined the effects of the PCNs on tumor metastasis in an LLC allograft tumor model. The PCNs were administered every third day from day 7 to 28 (8 times) following LLC cell inoculation. The tumors were dissected from the mice on day 14 after inoculation and examined for lung metastasis on day 28 (Fig. 2a). Interestingly, we found that TFG and TFMG significantly reduced tumor metastasis and decreased the number of metastatic nodules in the lungs (Fig. 2b). Additionally, we examined tumor metastasis using cytokeratin (epithelial cells marker) and LYVE-1 (lymphatic vessels marker) in cytokeratin+ tumor cells and found that TFG and TFMG significantly reduced metastasis to inguinal lymph nodes (Fig. 2c).

To further elucidate the anti-metastatic potential of the PCNs, we employed another tumor model, the MMTV-PyMT spontaneous breast cancer model, which spontaneously develops metastatic breast cancer. MMTV-PyMT mice were treated with TFG (10 nmol/kg), TFMG (10 nmol/kg), or vehicle every 3 days from week 12 following their birth up to week 14. Tumor size and the number of palpable tumor nodules were assessed at week 15 (Fig. 2d). We found that TFG and TFMG significantly reduced tumor burden (Fig. 2e) and decreased the number of tumor nodules (Fig. 2f) compared with control mice, respectively. In addition, we also found that TFG and TFMG significantly decreased metastatic nodules in the lungs (Fig. 2g). These results indicate that TFG and TFMG exhibit anti-metastatic effects in tumor implants and conventional transgenic tumor mouse models.

Since previous studies have suggested that tumor vascular normalization is an alternative therapeutic strategy for malignant tumors by restoring the balance between pro- and antiangiogenic signaling [24], we examined the effects of the PCNs on remodeling the tumor vasculature in the mouse models. To determine PCN-mediated structural and functional vascular changes in the tumor core, we evaluated the pericyte coverage and hypoxic regions of the allograft tumor from the LLC tumor model. We found a significant increase of α-SMA⁺ pericyte coverage both in the central and peripheral tumoral regions (Fig. 3a). There was also a significant decrease of hypoxic regions (Fig. 3b, PIMO) following TFG or TFMG treatment compared with the controls. In addition, we found that TFG and TFMG significantly reduced the hypoxic regions in the tumor core while enhancing the pericyte coverage of blood vessels compared with controls in the MMTV-PyMT mouse model (Fig. 3c). These results suggest that the PCNs increase the normalization of abnormal tumor vasculature and decrease hypoxic areas within tumors, leading to inhibition of tumor growth.

Abnormal vasculature creates a hypoxic microenvironment that polarizes inflammatory cells toward immune suppression [24]. Therefore, to elucidate the underlying tumor microenvironmental changes through vascular normalization by PCNs, we examined whether PCNs induced changes in immune cell activity within tumor tissues. We found that the positive area stained with CD8 was significantly upregulated with minimal changes of the positive area stained with CD4 by TFG or TFMG treatment. The ratio of CD4⁺/CD8⁺ was significantly decreased in TFG or TFMG-treated groups relative to the control in allograft LLC tumor model (Fig. 4a) as well as in the MMTV-PyMT spontaneous breast cancer model (Fig. 4b). These suggested that TFG or TFMG treatment induced significantly upregulation of cytotoxic CD8⁺ T lymphocytes with minimal alteration of CD4⁺ T lymphocytes in the animal models (Fig. 4c). To check for functional cytotoxic T cell activity, we analyzed apoptosis by the TUNEL assay in tumor tissues. We found that TFMG treatment significantly increased apoptosis both in the central and peripheral regions of the tumor tissues compared with controls (Fig. 4d). Moreover, the population of CD68⁺/Arg-1⁺ M2-like macrophages was significantly decreased in LLC tumor tissues following TFG and TFMG treatment compared with controls (Fig. 4e(a)). In contrast, the population of CD68⁺iNOS⁺ M1-like macrophages was increased by TFG and TFMG in tumor tissues (Fig. 4e(b)). Similar results were observed in the MMTV-PyMT spontaneous breast cancer model (Fig. 4f(a and b)). Decreased M2-like macrophages were observed, whereas increased M1-like macrophages and CD8⁺ T lymphocytes were evident in tumor tissue. M1-like macrophages tend to normalize tumor vasculature, whereas M2-like macrophages take part in the formation of abnormal blood vessels [25, 26]. In addition, M1-like macrophages stimulate naïve T cells to elicit a Th1/cytotoxic response, while M2-like macrophages induce T cell responses without antitumor activity [27, 28]. Therefore, M1-like cells could suppress tumor growth, whereas M2-like cells promote it [27, 29‐31]. PCN increased M1-like macrophages and diminished M2-like macrophages, as well as CD8⁺ T lymphocytes in tumor tissue, suggesting PCN might induce normalization of tumor vasculature through enhancing the reprogramming of macrophages.

To investigate the molecular mechanism of PCN-induced vascular normalization, we established an in vitro EC co-culture system with LLC cells using Transwell plates (Fig. 5a). The EC co-cultures with TFG and TFMG significantly reduced vascular permeability in vitro (Fig. 5b) in both a time- and dose-dependent manner (Fig. 5c, d), confirming the in vivo animal results. PAR-1 signaling in endothelial cells has been known to involve activation of other protease-activated receptors, such as PAR-2 and PAR-3, by receptor heterodimerization [32‐35]. Hence, we examined the involvement of PAR-2 and PAR-3 in the reduction of endothelial permeability induced by the PCNs. We performed an siRNA-mediated knockdown of PAR-1, PAR-2, and PAR-3, followed by measurement of the effects on TFMG-induced permeability abatement. The siPAR-1, siPAR-2, and siPAR-3 siRNAs exhibited more than 90% efficiency in suppressing gene expression in the ECs at 48 h post-transfection (Fig. S3). The endothelial permeability was significantly enhanced by PAR-1 and PAR-3 knockdown in ECs. However, PAR-2 knockdown had no significant effect on the reduction of endothelial permeability induced by TFMG (Fig. 5e), suggesting that only PAR-1 and PAR-3 are involved in the anti-permeable effect of PCNs in ECs.

Notably, PAR-3 was shown to potentiate the PAR-1-mediated response by PAR-1/PAR-3 heterodimerization [32]. Therefore, to test the role of the PAR-1 interaction with PAR-3 in the PCN-mediated effect, we examined whether TFG and TFMG treatment increased the PAR-1/PAR-3 interaction. We performed a co-immunoprecipitation analysis of PAR-1 and PAR-3 following TFG or TFMG treatment and found that the PAR-1/PAR-3 interaction was significantly enhanced (Fig. 5f). We also observed increased binding of PAR-1/PAR-3 by TFG and TFMG treatment (Fig. 5g) compared with the control. Together, these results suggest that both TFG and TFMG induce vascular normalization through heterodimerization of PAR-1 and PAR-3, leading to PAR-1-mediated PAR-3 activation.

The tyrosine-protein kinase receptor, Tie2, expressed primarily in endothelial cells, has recently been identified as a crucial target for vascular normalization in cancer [21, 36]. Tie2 activation has been implicated in noncanonical PAR-3 activation-mediated stabilization of vascular integrity [37]. To investigate whether Tie2 has a role in TFG/TFMG-induced vascular normalization, we examined the effects of siRNA-mediated endothelial Tie2 silencing and antibody-mediated blockage of endothelial Tie2 receptors on in vitro vascular permeability using EC co-cultures. As measured by qRT-PCR, Tie2 siRNA exhibited more than 90% efficacy in suppressing gene expression at 48 h post-transfection (Fig. S4A). TFMG-induced abatement of in vitro permeability was significantly rescued by endothelial Tie2 knockdown and by blockage of the endothelial Tie2 receptor using anti-Tie2 antibody (Fig. 6a). This demonstrates a role for endothelial Tie2 in TFMG-induced vascular normalization.

In addition, immunofluorescence analysis revealed that TFMG treatment significantly enhanced ZO-1 levels, particularly at the cell-cell junctions of endothelial cells (Fig. 6b), confirming that the cellular tight junctions were increased when permeability was inhibited by the PCNs. To identify the Tie2 signaling pathways that increase cell junctional proteins, we performed a Western blot analysis. The results revealed that TFMG treatment of EC monocultures caused significant increases in Tie2 phosphorylation and its downstream target, Akt (Fig. 6c). This suggests that Akt activation results from Tie2 activation following TFMG-induced PAR-1/PAR-3 heterodimerization. We found that phosphorylated FoxO3a at S253 was increased by TFMG treatment, whereas FoxO3a levels were decreased, indicating that it is targeted for degradation by phosphorylation (Fig. 6c). We also observed a significant increase in the level of Rho kinase (ROCK-1), a downstream target of Rho-GTPase, and an increase in the level of the tight-junction protein zonula occludens-1 (ZO-1) following TFMG treatment of endothelial cells compared with untreated cells (Fig. 6c). To confirm the hypothesis that TFMG induces vascular normalization by increasing PAR-1/PAR-3 interaction and Tie2 activation in vivo, we performed an IHC analysis of tumor tissues. We found that phosphorylated Tie2 was upregulated by TFMG where endothelial PAR-3 was expressed (Fig. 6d). In addition, the interaction between PAR-1/PAR-3 in CD31-expressed endothelial cells was increased by TFMG treatment compared with the control in tumor tissues (Fig. 6e), suggesting that the PAR-1/PAR-3 interaction results in Tie2 activation and is involved in the vascular normalization by the PCNs.

As previously reported, PAR-1 interaction with PAR-3 alters the PAR-1/G_α13, but not PAR-1/G_αq, binding conformation, and thereby favors a distinct G_α13-activated downstream pathway [32]. To further elucidate which G-protein isoform is involved in Tie2 activation by TFMG, we examined Tie2 activation by knocking down several G-proteins with siRNA in ECs. G_α13, G_αi, and G_αq siRNA exhibited more than 60% efficacy in suppressing gene expression at 48 h post-transfection (Fig. S4B). Activation of Tie2 and Akt was induced by TFMG and reduced in the absence of the G_α13 G-protein (Fig. 6f), suggesting that Gα₁₃ mediates Tie2 activation by TFMG. Taken together, these results indicate that PCN-induced vessel normalization depends on endothelial Tie2-induced tightening of endothelial junctions which results in reduced vascular permeability (Fig. 6g).

While pericyte-expressed Tie2 has been suggested to play a vital role in controlling tumor-associated angiogenesis and tumor growth [38], modest expression of PARs has also been associated with pericytes [39‐41]. Therefore, we examined whether pericyte Tie2, in addition to endothelial Tie2, has any direct effect on TFMG-induced vascular normalization. To investigate the involvement of pericyte Tie2 in TFMG-induced reduction of vascular permeability, we established a PC and a pericyte-endothelial cancer cell (PEC) co-culture system (Fig. S5A and S5B). TFMG treatment did not inhibit in vitro permeability in the PC co-culture system (Fig. S5C), whereas TFMG treatment induced significant inhibition of in vitro permeability in the PEC co-culture system (Fig. S5D). In addition, siRNA-mediated silencing of pericyte PAR-1, PAR-2, PAR-3, or Tie2 resulted in no significant changes in TFMG-induced permeability reduction (Fig. S5D). These results indicate that there are no direct effects of pericyte PARs or Tie2 in TFMG-induced vascular normalization. In addition, TFMG treatment of pericyte monocultures did not result in the activation of pericyte Tie2, nor the activation of the downstream target Akt (Fig. S5E). This further supports the finding that TFMG-induced vascular normalization is not associated with pericyte Tie2.

Tumor vessel normalization is expected to not only reduce metastasis but also to improve delivery of chemotherapeutic drugs by enhancing perfusion [42, 43]. Based on the effects of the PCNs in reversing the functional impairment of the disorganized tumor vessels, we investigated whether PCNs enhanced the delivery and antitumor effects of chemotherapeutic agents. To examine the synergistic effects of PCNs with chemotherapeutic agents, TFMG was administered individually or co-administered with varying doses of cisplatin or doxorubicin to LLC tumor-bearing mice, and their respective antitumor effects were evaluated against vehicle-administered control mice. We found that a combined treatment of TFG or TFMG with cisplatin or doxorubicin enhanced the additive antitumor effects of either cisplatin (Figs. 7a, b; S6A; and S6B) or doxorubicin (Fig. S6C and S6D), leading to significant inhibition of tumor progression, not only compared with the control but also compared with cisplatin or doxorubicin monotherapy. Cisplatin combined with TFG or TFMG markedly enhanced the survival relative to the control or cisplatin monotherapy (Fig. 7c). In addition, cisplatin coadministration with TFG or TFMG significantly reduced the hypoxic area of tumors (Fig. 7d, PIMO), while significantly enhancing the pericyte coverage of the tumor blood vessels (Fig. 7e).

Next, to confirm the role of PCNs in normalizing the tumor vasculature, we assessed vascular perfusion using DyLight 488-conjugated lectin [44]. TFMG as well as cotreatment with cisplatin increased perfused lectin to the tissues (Fig. 7f). Additionally, the leakage of FITC-dextran from the vessels to the tissues [45] was significantly decreased by treatment with TFMG or the cisplatin combination (Fig. 7g). To compare histological alterations induced by cisplatin or antiangiogenic agents, we performed H&E staining of the tumor tissues following treatment with cisplatin or bevacizumab. We found a well-organized morphology in TFMG- or bevacizumab-treated tissue, but not in the cisplatin-treated group. Many dead cells were found in the tissues of the combined cisplatin plus TFMG group compared with the control group (Fig. S6E), suggesting that the cytotoxic activity of cisplatin was increased through restored perfusion by TFMG. These results suggest that PCNs increase the chemotherapeutic effects of cisplatin through normalization of the tumor vessels.