Tissue Factor-Expressing Tumor-Derived Extracellular Vesicles Activate Quiescent Endothelial Cells via Protease-Activated Receptor-1

Tissue factor (TF)-expressing tumor-derived extracellular vesicles (EVs) can promote metastasis and pre-metastatic niche formation, but the mechanisms by which this occurs remain largely unknown. We hypothesized that generation of activated factor X (FXa) by TF expressed on tumor-derived EV could activate protease-activated receptors (PARs) on non-activated endothelial cells to induce a pro-adhesive and pro-inflammatory phenotype. We obtained EV from TF-expressing breast (MDA-MB-231) and pancreatic (BxPC3 and Capan-1) tumor cell lines. We measured expression of E-selectin and secretion of interleukin-8 (IL-8) in human umbilical vein endothelial cells after exposure to EV and various immunologic and chemical inhibitors of TF, FXa, PAR-1, and PAR-2. After 6 h of exposure to tumor-derived EV (pretreated with factor VIIa and FX) in vitro, endothelial cells upregulated E-selectin expression and secreted IL-8. These changes were decreased with an anti-TF antibody, FXa inhibitors (FPRCK and EGRCK), and PAR-1 antagonist (E5555), demonstrating that FXa generated by TF-expressing tumor-derived EV was signaling through endothelial PAR-1. Due to weak constitutive PAR-2 expression, these endothelial responses were not induced by a PAR-2 agonist peptide (SLIGKV) and were not inhibited by a PAR-2 antagonist (FSLLRY) after exposure to tumor-derived EV. In conclusion, we found that TF-expressing cancer-derived EVs activate quiescent endothelial cells, upregulating E-selectin and inducing IL-8 secretion through generation of FXa and cleavage of PAR-1. Conversion of resting endothelial cells to an activated phenotype by TF-expressing cancer-derived EV could promote cancer metastases.

Tissue factor (TF)-expressing tumor-derived extracellular vesicles (EVs) can promote metastasis and pre-metastatic niche formation, but the mechanisms by which this occurs remain largely unknown. We hypothesized that generation of activated factor X (FXa) by TF expressed on tumor-derived EV could activate protease-activated receptors (PARs) on non-activated endothelial cells to induce a pro-adhesive and pro-inflammatory phenotype. We obtained EV from TF-expressing breast (MDA-MB-231) and pancreatic (BxPC3 and Capan-1) tumor cell lines. We measured expression of E-selectin and secretion of interleukin-8 (IL-8) in human umbilical vein endothelial cells after exposure to EV and various immunologic and chemical inhibitors of TF, FXa, PAR-1, and PAR-2. After 6 h of exposure to tumor-derived EV (pretreated with factor VIIa and FX) in vitro, endothelial cells upregulated E-selectin expression and secreted IL-8. These changes were decreased with an anti-TF antibody, FXa inhibitors (FPRCK and EGRCK), and PAR-1 antagonist (E5555), demonstrating that FXa generated by TF-expressing tumor-derived EV was signaling through endothelial PAR-1. Due to weak constitutive PAR-2 expression, these endothelial responses were not induced by a PAR-2 agonist peptide (SLIGKV) and were not inhibited by a PAR-2 antagonist (FSLLRY) after exposure to tumor-derived EV. In conclusion, we found that TF-expressing cancer-derived EVs activate quiescent endothelial cells, upregulating E-selectin and inducing IL-8 secretion through generation of FXa and cleavage of PAR-1. Conversion of resting endothelial cells to an activated phenotype by TF-expressing cancer-derived EV could promote cancer metastases.
Normal quiescent endothelial cells act as a barrier to cancer metastasis, but pro-inflammatory mediators can activate endothelial cells to induce expression of pro-inflammatory adhesion molecules and secretion of pro-inflammatory cytokines. Such endothelial responses can potentially promote cancer metastasis by facilitating recruitment of tumor cells or bone marrow-derived hematopoietic cells involved in pre-metastatic niche formation (8)(9)(10). Activation of endothelial cells by EV has been observed previously, resulting in increased vascular permeability, angiogenesis, and cancer metastasis (10)(11)(12)(13)(14); however, the mechanisms remain largely unknown.
In this study, we hypothesized that TF-expressing tumorderived EV would activate non-activated endothelial cells, inducing upregulation of adhesion molecules and secretion of cytokines in a PAR-1 and PAR-2 dependent manner. We found that TF-expressing tumor-derived EV from breast and pancreatic cancer cell lines activated endothelial cells within 6 h of exposure, and this activation required TF-mediated generation of FXa. The endothelial responses were mediated through PAR-1 but not PAR-2. Our results indicate that PAR-1 is the main receptor for eliciting TF-expressing tumor-derived EV responses in unstimulated endothelial cells.

eVs from Tumor cell lines
At 70-90% confluency, tumor cells were serum starved overnight. Tumor-conditioned media were centrifuged at 300 × g for 5 min and 2,500 × g for 20 min to remove cells and debris, then 100,000 × g for 70 min at 4°C in an Optima L-90K ultracentrifuge (Beckman Coulter, Pasadena, CA, USA) to pellet EV. The pellet was resuspended in phosphate-buffered saline (PBS) and used immediately.
nanoparticle Tracking analysis of eVs Samples were diluted with PBS and analyzed using in a Malvern NanoSight NS300 (Westborough, MA, USA) with 3 × 60 s runs at ambient temperature. EV size distribution and concentration were determined by manufacturer's software (V3.0, camera level at 16 and threshold at 10).
atomic Force Microscopy of eVs Extracellular vesicles, prepared as above, were resuspended in water and then centrifuged at 100,000 × g for 2 h at 4°C. The pellet was then resuspended in water. Atomic force microscopy was conducted at the University of Texas Houston AFM Core Facility. A 10 μl sample of EV suspension (at a concentration of 2.5 × 10 11 EV/ml as determined by nanoparticle tracking analysis) was incubated for 10 min on freshly cleaved mica surfaces (V1 mica disks, Ted Pella, Redding, CA, USA). Excess liquid was then removed, and the sample was allowed to dry at room temperature. All samples were freshly prepared and imaged immediately. Atomic force microscopy was performed using RTESP cantilevers (fo = 237-289 kHz, k = 20-80 N/m, Bruker, Santa Barbara, CA, USA) and BioScope II AFM (Bruker). Images (5 μm × 5 μm) were taken using tapping mode operated in air at a scan rate of 0.6 Hz, then analyzed using NanoScope Analysis (V1.40, Bruker).

Procoagulant activity of eVs
Tumor-derived EVs were diluted 1:5 in HEPES buffer (10 mM HEPES, 137 mM sodium chloride, 5 mM calcium chloride, 4 mM potassium chloride, 10 mM glucose, 0.5% bovine serum albumin, and pH 7.4) and incubated with 1 nM FVIIa and 75 nM FX for 15 min at 37°C. The concentration of FX does not limit the reaction because the highest concentration of generated FXa was below 20 nM in our study. A chromogenic substrate (40 µM Spectrozyme, Sekisui Diagnostics, Lexington, MA, USA) was added, and color change was measured at 405 nm every 15 s for 10 min. Factor Xa generation was calculated based on a standard curve of purified FXa. To account for TF-independent FX activation, control values (with FX only) were deducted from reported values.

cell-Based elisa for adhesion Molecule expression
Endothelial cells were exposed to EV or agonists for 6 h at 37°C, then fixed with 2% paraformaldehyde for 15 min. Wells were blocked with 5% normal goat serum (Jackson ImmunoResearch, West Grove, PA, USA) for 30 min, then incubated with 1 µg/ml antibody against E-selectin (1.2B6, Abcam) overnight at 4°C. Wells were incubated with horseradish peroxidase-conjugated secondary antibody (1:150, Thermo Scientific, Waltham, MA, USA) for 1 h, then with 3,3′,5,5′-tetramethylbenzidine substrate (Thermo Scientific) for 30 min. Washing with PBS was performed between incubation steps. Sulfuric acid (2 N) was added, and absorbance was read at 450 nm (Molecular Devices SpectroMax M3, Sunnyvale, CA, USA). Data were normalized to untreated endothelial cells due to baseline fluctuations and presented as fold change.
elisa for cytokine secretion Cultured media from endothelial cells was collected and stored at −80°C. Samples were centrifuged at 15,000 × g for 10 min at 4°C to remove debris and aggregates. The interleukin-8 (IL-8) concentration (linear range 31.2-2,000 pg/ml) was measured using commercial kits according to the manufacturer's instructions (R&D Systems ELISA DuoSets, Minneapolis, MN, USA) and was not normalized to untreated endothelial cells.

immunofluorescent Microscopy for endothelial e-selectin and Par expression
Endothelial cells were grown on fibronectin-coated coverslips, then exposed to EV and fixed as above. Cells were stained with an antibody against E-selectin, PAR-1, or PAR-2 (10 µg/ml) for 30 min, followed by an Alexa-488-conjugated secondary antibody for 30 min. Coverslips were sealed using mounting medium with DAPI (VECTASHIELD, Vector Laboratories, Burling Game, CA, USA) and imaged (Nikon Eclipse TE2000-U, Melville, NY, USA; Photometrics CoolSNAP HQ2 camera, Tucson, AZ, USA).
cell Viability assay Endothelial viability was determined using a commercial kit based on ATP quantification (CellTitre Glo, Promega, Madison, WI, USA). Data were normalized to untreated endothelial cells and presented as a percentage.

statistical analysis
Data were parametric and represented as mean ± SD. Means of different treatments were compared with a Student's t-test and α = 0.05 using Prism 6.0 (GraphPad, La Jolla, CA, USA).

ethics approval statement
No ethics approval statement was required for the study.

Tumor-Derived eVs express TF antigen and Procoagulant activity
Tissue factor antigen was expressed on the surface of MDA-MB-231, BxPC3, and Capan-1 cells, as shown by flow cytometric analysis (Figure 1A). Similarly, EV derived from these tumor cell lines expressed TF antigen on immunoblotting, with flotillin and CD9 expression confirming successful isolation of EV ( Figure 1B; Figure S1 in Supplementary Material). EV from   all three cell lines generated similar amounts of FXa ( Figure 1C). The size distribution of EV derived from MDA-MB-231 and BxPC3 cells was similar, consisting mostly of small vesicles (approximately 80% of population less than 200 nm), with a small amount of large vesicles, likely membrane-derived microvesicles (>200 nm) ( Figure 1D). However, Capan-1 cells generated more large microvesicles (close to 50%). The size distribution was verified using atomic force microscopy ( Figure 1E), which showed vesicles with a range of size (~50-300 nm). We observed more large individual vesicles from Capan-1 cells, supporting the nanoparticle tracking analysis. The size of released EV has been shown to vary among different cell lines derived from different tissues and is dependent on genes involved in vesicle trafficking (31).

endothelial cells exhibit high constitutive expression of Par-1 and Weak expression of Par-2
We evaluated PAR-1 and PAR-2 expression on non-activated E4+ HUVEC with flow cytometry and immunofluorescent microscopy. We found PAR-1 was highly expressed, whereas weak PAR-2 expression was only seen in a small proportion of E4+ HUVEC, as shown by the small peak to the right of the isotype control on flow cytometric analysis and few positive staining cells on immunofluorescent microscopy (Figure 2). To verify this phenotype was not a consequence of the ORF4 gene, we also tested HUVEC and found a similar pattern of expression for both PAR-1 and PAR-2 ( Figure S2A in Supplementary Material), as reported previously (23). Strong surface expression of PAR-1 and PAR-2 was observed on MDA-MB-231 breast cancer cells with flow cytometry (Figure 2), which were used as a positive control (32).

TF-expressing Tumor-Derived eV Upregulate endothelial e-selectin expression and induce il-8 secretion in a TF-, FXa-, and Par-1-Dependent, but Par-2-independent, Manner
Endothelial cells were treated with EV derived from MDA-MB-231, BxPC3, and Capan-1 cell lines for 6 h. We observed an increase in endothelial E-selectin surface expression and IL-8 secretion only when EVs were pretreated with FVIIa and FX (Figures 3A,B).
We saw similar upregulation in non-activated HUVEC with tumor-derived EV pretreated with FVIIa and FX ( Figures S2B,C in Supplementary Material), indicating that these responses were not due to E4+ ORF gene insertion. Subsequent experiments were performed with MDA-MB-231-derived EV as our model tumor EV. The EV-induced changes in endothelial cells required both FVIIa and FX, suggesting that generation of FXa by TF-FVIIa on EV is required (Figures 3C,D). In support of this, we found that FXa alone induced a mild increase in E-selectin expression and IL-8 secretion, although the response was far weaker than that induced by EV pretreated with FVIIa and FX ( Figure S3A in Supplementary Material). This suggests that FXa generated on the surface of cell membranes is more effective than free FXa for inducing pro-inflammatory responses in endothelial cells.
An antibody against TF (HTF-1) and FXa inhibitors (FPRCK and EGRCK) reduced and abolished upregulation of E-selectin and secretion of IL-8 in E4+ HUVEC, respectively (Figures 4A-D), supporting that these responses were mediated by TF-mediated generation of FXa. The FXa inhibitor, FPRCK and EGRCK, also inhibited endothelial responses elicited by purified FXa ( Figures  S3A,B  microscopy, E-selectin expression was observed on a subpopulation of E4+ HUVEC exposed to tumor-derived EV and was abolished with FPRCK ( Figure 4E). Although we performed our experiments in serum-free media, traces of thrombin may be present. Thrombin has been shown to induce E-selectin expression and IL-8 secretion via cleavage of PAR-1 (33). Inclusion of 10 U/ml hirudin, a specific thrombin antagonist, abolished endothelial responses to thrombin ( Figure S3C in Supplementary Material), but did not significantly reduce the responses to MDA-MB-231-derived EV pretreated with FVIIa and FX (Figures 4C,D). Since FXa has been shown to activate both PAR-1 and PAR-2 (19-21), we sought to determine which of these PARs was involved in the observed endothelial responses, using an agonist and inhibitor approach. We observed an increase in E-selectin expression and IL-8 secretion in E4+ HUVEC with the PAR-1 agonist peptide, TFLLR (Figures 5A,B) but not the scrambled peptide, RLLFT. The canonical PAR-1 agonist, thrombin, also stimulated these endothelial responses ( Figure  S3C in Supplementary Material). Pretreatment of endothelial cells with a PAR-1 antagonist, E5555, abolished the responses to both the PAR-1 agonist peptide and MDA-MB-231-derived EV (Figures 5A,B). By contrast, a PAR-2 agonist (SLIGKV) or scrambled peptide (VKGILS) had no effect on the endothelial cells; the PAR-2 antagonist (FSLLRY) did not reduce endothelial responses to MDA-MB-231-derived EV (Figures 5C,D). We attribute the lack of PAR-2-mediated response to the weak PAR-2 expression in a subset of the non-activated endothelial cells (Figure 2). To verify that the PAR-2 agonist was functional, we treated MDA-MB-231 (which strongly expresses PAR-1 and PAR-2 as shown in Figure 2) and E4+ HUVEC with the PAR-1 and PAR-2 agonist for 30 min, and evaluated downstream PAR-mediated signaling by immunoblotting for phosphorylated ERK (34) (Figure S4 in Supplementary Material). Phosphorylation of ERK was observed in MDA-MB-231 tumor cells with both PAR-1 and PAR-2 agonists. However, in E4+ HUVEC, ERK phosphorylation was observed with the PAR-1, but not the PAR-2, agonist. Taken together, our data support that the tumor-derived EVs are stimulating the endothelial cells primarily via PAR-1. Neither tumor-derived EV nor inhibitors

DiscUssiOn
In this study, we demonstrate that TF-expressing EVs induce a pro-adhesive and pro-inflammatory phenotype in quiescent endothelial cells. This effect was mediated by TF-dependent generation of FXa and PAR-1 activation. Our results suggest that PAR-1 is the dominant receptor on unstimulated endothelial cells for the TF-FVIIa-FX complex on tumor-derived EV. Endothelial responses induced by FXa have largely been ascribed to PAR-2 (21,22,26,27); however, this was not the case in our study, likely due to the weak PAR-2 expression on the endothelial cells. Our study focused on the 6-h acute response of unstimulated ECs upon exposure to tumor-derived EV. PAR-1 can trans-activate PAR-2 (35), and PAR-2 expression can be upregulated in endothelial cells after several hours of exposure to FXa, inflammatory cytokines, or hypoxic conditions (22). Thus, it is possible that after the initial stimulation by PAR-1, PAR-2 may also play a role in endothelial responses to TF-expressing tumor-derived EV.
Using immunofluorescent microscopy, we observed that the upregulation of E-selectin expression occurred in a distinct subset of cells (~15-40% dependent on agonist). These data suggest that there is heterogeneity within a given endothelial population and  that tumor-derived EV may only activate "receptive" endothelial cells by virtue of surface receptor expression or triggering of permissive downstream signaling responses. The mechanisms by which tumor-derived EV bind to or fuse with endothelial cells are not fully elucidated. The conversion of quiescent endothelial cells to a pro-adhesive and pro-inflammatory phenotype by tumor-derived EV could impact metastasis. Dysfunctional, but not normal, endothelial cells promote tumor inflammation, invasion, and metastasis (36,37). Endothelial activation could potentially be one of the first steps in which pre-metastatic niches are formed. Upregulation of endothelial adhesion molecules, such as E-selectin, could promote rolling and arrest of bone marrow-derived cells and tumor cells for metastasis (38)(39)(40). Increased secretion of IL-8 could induce inflammation, further benefiting metastasis by recruiting proinflammatory cells and promoting growth factor and chemokine secretion (41). We investigated E-selectin expression and IL-8 secretion in endothelial cells in response to tumor-derived EV, but other proteins and chemokines that may impact metastasis, such as intracellular adhesion molecule-1 (42) and monocyte chemoattractant protein-1 (43), may be upregulated by tumorderived EV as well.
Large quantities of high TF-expressing tumor-derived EV in the circulation are likely to generate thrombin, which will activate PAR-1 more strongly than FXa and override any observed FXa-mediated responses (20). Thrombin generation may also trigger paraneoplastic thrombosis (albeit, controversial) (5) or promote thrombosis through impairing anti-tumor immune defenses (44). However, it is more likely that endothelial cells at sites of metastasis are exposed to low concentrations of continuously shed EV over a long period of time versus single large boluses. We speculate that low concentrations of TF-expressing EV may carry a "coat" of FXa, which does not generate thrombin systemically; with continuous exposure and binding of EV to endothelial cells, a threshold concentration required to activate endothelial cells via PAR-1 cleavage may be reached. Initial PAR-1-mediated signaling could then "prime" quiescent endothelial cells into a pro-inflammatory phenotype, facilitating pre-metastatic niche formation and metastasis of the primary tumor. Our data suggest that FXa inhibitors (for example, rivaroxaban) could be useful as preventative adjunct therapy in cancer patients to not only reduce cancer-associated thrombosis but also minimize tumor-induced EV endothelial activation (45), which may help reduce metastasis.
aUThOr cOnTriBUTiOns SC contributed to experimental design, acquisition and analysis of data, and draft of manuscript. JP contributed to acquisition