Apixaban Suppresses the Release of TF-Positive Microvesicles and Restrains Cancer Cell Proliferation through Directly Inhibiting TF-fVIIa Activity

Abstract The activation of protease-activated receptor (PAR)-2 by factor Xa (fXa) promotes the release of tissue factor-positive microvesicles (TF+MV), and contributes to proliferation in cancer cells. This study examined the ability of direct oral anticoagulants (DOACs), apixaban and rivaroxaban, to inhibit the release of TF+MV from two cell lines (MDA-MB-231 and AsPC-1) as well as cell proliferation. Activation of the cells with fXa (10 nM) enhanced the release of TF+MV but was suppressed in the presence of either DOAC. These MVs were found to contain fVIIa, but not fXa. Incubation of cell lines with apixaban (1.8 µM) but not rivaroxaban (1.8 µM), in the absence of fXa decreased the release of TF+MV below that of resting cells, in a PAR2-dependent manner. Furthermore, incubation with apixaban reduced the proliferation rate in both cells lines. Incubation of purified fVIIa with apixaban but not rivaroxaban resulted in complete inhibition of fVIIa proteolytic activity as measured using two fVIIa chromogenic substrates. Pre-incubation of the cells with an inhibitory anti-fVIIa antibody, with apixaban or the blocking of PAR2 suppressed the release of TF+MV to a comparable level, and reduced cell proliferation but the effect was not cumulative. This study has established that the activation of PAR2 by TF–fVIIa complex is the principal mediator in augmenting the release of TF+MV as well as cancer cell proliferation. Importantly, for the first time we have shown that apixaban selectively inhibits the proteolytic activity of fVIIa as well as the signalling arising from the TF–fVIIa complex.


Introduction
The bidirectional association between coagulation and cancer has been established. 1 Advanced malignancies are known to cause a wide range of thrombotic diseases including venous thrombosis, pulmonary embolism and disseminated intravascular coagulation, 2 but patient survival may be improved by the administration of anticoagulants. 3,4 Tissue factor (TF) has recently been shown to be an interlinking molecule between the coagulation and cancer and the correlation between levels of TF expression and a poor overall prognosis has been demonstrated in a number of different cancer types. [5][6][7] Furthermore, cancer cells are capable of releasing large quantities of TF-positive microvesicles (TF þ MV) following activation. 8,9 These TF þ MV circulate within the bloodstream and are suggested to participate in widespread thrombosis [10][11][12][13] as well as causing damage to the vascular endothelium, 14 although a direct elevation of TF þ MV does not explain the cancer-associated thrombosis across different tumour types. A possible trigger for the activation of cancer cells is likely to occur when the cells come into contact with blood components. Coagulation proteases are known to cleave a family of receptors called protease-activated receptors (PARs) on the surface of cells. Among these receptors, PAR2 is known to be cleaved by factor Xa (fXa) alone, and also by the TF-fVIIa-fXa ternary complex. 15,16 PAR2 has also been shown to be a target for the proteolytic activity of TF-fVIIa. 15 The up-regulation of PAR2 protein in cancer cells has also been linked to aggressive cancer phenotypes. [17][18][19][20] The activation of PAR2 on the surface of various cancer cell types leads to the incorporation of TF within MV 21 and also can promote increased cellular proliferation, invasion and migration, as well as the expression of interleukin (IL)-8 and vascular endothelial growth factor. 16,17,22,23 More recently, a number of studies have demonstrated the induction of MV release through TFmediated PAR2 activation and have studied the underlying mechanisms. [24][25][26][27][28] Direct oral anticoagulants (DOACs) are a recent group of anticoagulants with a fast onset of action and have the advantage of not needing regular monitoring by blood testing. 29 Apixaban and rivaroxaban are two such compounds which function by directly inhibiting the proteolytic activity of fXa, acting on both the circulating and clot-bound fXa. 30,31 In this study, we report the contribution of TF-fVIIa as a major activator of PAR2 in cancer cells, resulting in the release of TF þ MV and enhancement of cancer cell proliferation. In addition, we have identified a novel and surprising property of apixaban in suppressing this mechanism by directly inhibiting the proteolytic activity of fVIIa.

Cell Culture and Determination of Cell Numbers
Cell lines were selected on the basis of TF expression and MV release, and not on the basis of tissue of origin. 8 MDA-MB-231 breast cancer cell line (ATCC, Teddington, United Kingdom) were cultured in Dulbecco's modified Eagle medium and AsPC-1 pancreatic cancer cell line (ATCC) were cultured in RPMI-1640. All media were supplemented with 10% (v/v) heat-inactivated foetal calf serum (FCS) to ensure the lack of any functional enzymes. Human dermal blood primary endothelial cells (HDBEC), devoid of endogenous TF were cultured in MV media containing 5% (v/v) FCS and growth supplements (PromoCell, Heidelberg, Germany). Unless otherwise stated, the cells were adapted to serum-free medium prior to use in the experiments. Cell numbers were determined using the crystal violet staining procedure and interpreted from a standard curve as previously described. 32,33

Preparation of Test Reagents
Apixaban and rivaroxaban were obtained as pure compounds from Bistrol Myers Squibb (New York, United States) and Bayer (Leverkusen, Germany), respectively, and used at the approximate therapeutic range (0.18-1.8 µM). The compounds were dissolved in dimethyl sulfoxide (DMSO) and diluted in phosphate-buffered saline (PBS) pH 7.4, to 4 mg/ mL working stock solutions. Appropriate controls using similarly diluted DMSO were used alongside, as controls. PAR2-agonist peptide (PAR2-AP; SLIGKV) was synthesised (Severn Biotech Ltd, Kidderminster, United Kingdom) and used at a final concentration of 20 µM. Coagulation fXa and fVIIa (Enzyme Research Laboratories, Swansea, United Kingdom) were diluted stepwise to the required concentrations. Blocking antibodies against PAR2 (SAM11) and PAR1 (ATAP2) were obtained from Santa Cruz Biotechnology (Heidelberg, Germany) and were incubated with the cells at 20 µg/mL to block PAR2 and PAR1, respectively. 34 An inhibitory polyclonal rabbit anti-human fVIIa antibody was obtained from Abcam (Cambridge, United Kingdom) and incubated with the cells at 20 µg/mL, which was optimised beforehand.

Preparation of the TF-Containing Microvesicles
To generate the TF-containing MVs, MDA-MB-231 and AsPC-1 cells were propagated in 25 cm 2 flasks. The cells were washed with PBS and adapted to respective serum-free medium, for 2 hours. The released MVs were then prepared from the conditioned media by ultracentrifugation according to described procedures. 35 The MVs were then washed with PBS and collected again. In other experiments, the cells were activated by incubation with fXa (10 nM) or PAR2-activating peptide (SLIGKV; 20 µM) for 30 minutes prior to collection. In variations of the experiments, the cells were supplemented with combinations of PAR2-blocking antibody (SAM11; 20 µg/mL) or PAR1-blocking (ATAP2; 20 µg/mL) or an immunoglobulin G (IgG) isotype (20 µg/mL), together with apixaban (0-1.8 µM), rivaroxaban (0-1.8 µM) or the DMSO vehicle control.

Analysis of the TF-Containing Microvesicles
The functional density of the released MVs was determined using the Zymuphen MP assay kit (Hyphen BioMed/Quadratech, Epsom, United Kingdom) and the MV density determined from the standards provided by the kit. The Thrombosis and Haemostasis Vol. 119 No. 9/2019 released MV-associated TF antigen was measured using the Quantikine TF-enzyme-linked immunosorbent assay (ELISA) kit (R&D Systems, Abingdon, United Kingdom) according to the manufacturer's instructions. The procoagulant activity of the purified MVs was measured using calibrated automated thrombogram (CAT) assay. The CAT analysis was performed in normal plasma and repeated in fVII-deficient plasma (Hyphen BioMed/Quadratech). In some experiments, the MVs were pre-incubated with an inhibitory mouse anti-human TF antibody (HTF-1; 20 µg/mL) or a control mouse isotype IgG (20 µg/mL; New England Biolabs, Hitchin, United Kingdom). For comparison, plasma was activated with Innovin thromboplastin reagent (Dade Behring, Deerfield, Illinois, United States) at a range of 0 to 10 U/mL (1 U/mL ¼ 1.3 ng/mL), and coagulation was also confirmed using a thrombin calibrator (Stago, Reading, United Kingdom). The 'lag time' to the onset of thrombin generation for each sample was compared with those of the Innovin standard to determine the relative TF activity within the MV samples.

Analysis of the fVIIa and fXa Antigen and Activity
The presence of fX and fVII antigen was detected by Western blot analysis. The samples were separated by 12% (w/v) sodium dodecyl sulphate-polyacrylamide gel electrophoresis, transferred onto nitrocellulose membranes, blocked with Tris-buffered saline with Tween 20 (TBST) (10 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.05% Tween-20). The membranes were then probed with a mouse monoclonal anti-fX antibody (156106), or a rabbit polyclonal anti-fVII antibody (R&D Systems), both diluted 1:2,000 (v/v) in TBST. The membranes were developed with alkaline phosphatase-conjugated goat anti-mouse or goat anti-rabbit antibodies (Santa Cruz), respectively, diluted 1:4,000 (v/v) and bands were visualised using the Western Blue stabilised alkaline phosphatase substrate (Promega, Southampton, United Kingdom). fVII/fVIIa antigen levels were determined using the Assaymax FVII-ELISA kit (Assaypro/Universal Biologicals Ltd., Cambridge, United Kingdom) according to the manufacturer's instructions.
MV and cell surface TF-fVIIa activity was measured by modification of previously described procedures. 36,37 MVs were used directly while cells (5 Â 10 4 ) were washed with PBS prior to assaying. The samples were incubated with fVIIa (10 nM; Enzyme Research Labs) in HEPES-buffered saline pH 7.4, containing 1% (w/v) bovine serum albumin and 5 mM CaCl 2 , together with fX (100 nM) and the fXa substrate (0.2 mM; Hyphen) diluted in the same buffer (200 µL). The samples were incubated for 60 minutes to develop the colour. Aliquots (150 µL) were then transferred to a 96-well plate containing 2% (v/v) acetic acid (50 µL) and the absorptions measured immediately at 410 nm. The amount of fXa generated was determined using a standard curve prepared using fXa (0-20 nM; Enzyme Research Labs). To detect MV-associated fVIIa activity, in some experiments fVIIa was omitted from the above procedure. The absorption measurements were compared to a set of controls prepared using a range of fVIIa (0-10 nM) which were supplemented with recombinant TF (10 U/mL) and analysed for fXa-generation potential as above. The activity of fVIIa was also measured directly using Pefachrome FVIIa substrate (LOXO, Dossenheim, Germany), as well as a novel and specific chromogenic substrate for fVIIa (NH2-Asn-Leu-Thr-Arg-pNA) both of which were used at a range of 0.2 to 1.6 mM final concentration. MV-associated fVIIa activity and the ability of apixaban (1.8 µM) and rivaroxaban (1.8 µM) to inhibit the purified fVIIa (50 nM) and MV-associated fVIIa was examined.

Analysis of the Expression of fVII and IL-8 Messenger Ribonucleic Acid
The expression of fVII messenger ribonucleic acid (mRNA) and IL-8 mRNA was measured by reverse transcription polymerase chain reaction (RT-PCR) as previously described. 31 Total RNA was extracted using the Ribozol solution (VWR, Lutterworth, United Kingdom). The expression of fVII mRNA or IL-8 was measured by GoTaq 1-Step RT-qPCR System (Promega) using QuantiTect primers for fVII, IL-8 and β-actin (Qiagene, Manchester, United Kingdom) and relative amounts determined using a reference.

Suppression of the Expression of fVII and PAR2 by siRNA Knock Down
In order to suppress the expression of fVII and PAR2, cells were transfected with a specific set of Silencer small interfering RNA (siRNA) (5 pmol; Life Technologies, Paisley, United Kingdom) to suppress the expression of fVII and PAR2, respectively, or transfected with a comparable set of control siRNA (5 pmol; Life Technologies) prior to activation. The concentration of each siRNA was optimised by examining the expression of each protein separately beforehand.

Approximation of Binding of Apixaban to fVIIa
The crystal structures of fVIIa (4ylq), fXa (2P16) and apixaban (GG2) were obtained on Brookhaven format (Protein Data Bank) and used to estimate the location and efficiency of binding using Autodock 4v2.6. The Autodock graphical interface AutoDockTools 1.5.6 was used, the polar hydrogens were retained and partial charges added to the proteins using the Gasteiger charges. The search space was limited to an area of 20 Â 20 Â 20 Å, centred around the hydroxyl group of Ser195 in the enzymatic site of fVIIa and fXa. For each enzyme, 25 Â ligand orientations (poses) were examined and ranked according to the scoring function.

Statistical Analysis
All data represent the calculated mean values from the number of experiments stated in each figure legend AE the calculated standard error of the mean. Statistical analysis was carried out using the Statistical Package for the Social Sciences (SPSS Inc. Chicago, Illinois, United States). Significance was determined using one-way analysis of variance and Tukey's honesty significance test or, where appropriate, by paired t-test.

Cancer Cell-Derived Microvesicles Contain Functional TF and fVIIa
Throughout the study, MVs were purified from the serumfree conditioned media, from resting MDA-MB-231 and AsPC-1 cell lines due to high expression of TF and MV release, and was irrespective of tissue of origin. Measurement of procoagulant activity by CAT analysis showed that the MVs contained functional TF which was inhibited on pre-incubation with HTF-1 antibody (►Fig. 1A-C). Moreover, the qualitative analysis of the MVs showed that these possessed endogenous fXa-generation potential without the need for additional fVIIa (►Fig. 1D), and were also functional when examined in fVII-deficient plasma (►Fig. 1E-G) but not in fXdeficient plasma (not shown). Examination of the fVIIa by Western blot indicated the presence of active fVIIa in both cell lines, exhibiting multiple bands (►Fig. 1H) which have previously been attributed to the presence of glycosylation variants. 38,39 The extrahepatic expression of fVII has been demonstrated in a number of cell lines from various tissues. 40

Apixaban Directly Inhibits the Proteolytic Activity of fVIIa
Incubation of purified fXa (10 nM) with either apixaban (1.8 µM) or rivaroxaban (1.8 µM) resulted in maximal inhibition of fXa activity as measured using the fXa-chromogenic substrate (►Fig. 2A). In addition, following supplementation of the cells with apixaban or rivaroxaban, the purified MVs were washed with PBS and purified again by ultracentrifugation. To ensure the removal of residual DOAC, the resultant MVs were incubated with fXa (10 nM) and the activity measured using a fXa-chromogenic substrate. The washed MVs did not appear to possess any anti-fXa activity against purified fXa, precluding any apixaban or rivaroxaban carry over by the MVs (►Fig. 2B). Incubation of Perfachrome substrate with either fVIIa or fXa resulted in the colour production (►Fig. 2C). In contrast, the designed chromogenic substrate (NH2-Asn-Leu-Thr-Arg-pNA; 0.4 mM) was activated by purified fVIIa but not purified fXa. Importantly, apixaban (1.8 µM) but not rivaroxaban (1.8 µM) was capable of inhibiting purified fVIIa (5 nM) (►Fig. 2D). This level of inhibition was comparable to that achieved following preincubation of the purified fVIIa with an inhibitory polyclonal anti-fVII antibody. Moreover, the inclusion of TF (1 U/mL) did not alter the inhibitory influence of apixaban towards fVIIa (►Fig. 2E). The rate of reaction for the purified fVIIa (50 nM) was examined using the Perfachrome substrate (0.4-1.6 mM), by measuring the rate of change in absorption (410 nm) over 60 minutes. These analyses were carried out in the presence and absence of apixaban (►Fig. 2F) which indicated competitive inhibition of fVIIa by apixaban (►Fig. 2G), with a calculated Ki of 4.4 nM. Finally, incubation of MVs purified from the conditioned media of MDA-MB-231 cells with apixaban but not rivaroxaban eliminated the MV-associated fVIIa activity which was measured using the chromogenic substrate (NH2-Asn-Leu-Thr-Arg-pNA; 0.4 mM) (►Fig. 2H).
Apixaban and Rivaroxaban Prevent the Release of TF þ MV in Response to fXa Incubation of MDA-MB-231 and AsPC-1 cells with fXa (10 nM) resulted in the rapid release of MVs from these cell lines (►Fig. 3A and B) which also contained amplified levels of TF antigen (►Fig. 3C and D) and TF activity (►Fig. 3E and F). Moreover, inclusion of higher concentrations of apixaban (1.8 µM) or rivaroxaban (1.8 µM) were effective in reducing the release of MVs and the associated TF antigen and activity to comparable levels, with the exception of TF antigen released from MDA-MB-231 cells. This could arise as a consequence of very high incorporation of TF into these MVs and is in agreement with our previous findings. 8 Additionally, PAR2 activation can induce both MV release and incorporation of TF into the MVs. 21,44 Therefore, the lower thrombin generation may be attributed to reductions in both of these processes.

Apixaban but not Rivaroxaban Suppress the Release of TF þ MV from Non-Activated Cells
In addition to suppressing the cellular responses to fXa, the influence of DOAC on the release of TF þ MV from MDA-MB-231 and AsPC-1 cells under resting conditions was examined. Interestingly, apixaban (1.8 µM) but not rivaroxaban (1.8 µM) was capable of reducing the release of MVs from both cells lines to levels below that observed in resting cells (►Fig. 4A and B). This reduction was also reflected in the levels of MVassociated TF antigen (►Fig. 4C and D).

Apixaban Reduces the rate of Cell Proliferation in Non-Activated Cells
In order to assess the influence of DOAC on cell proliferation, MDA-MB-231 cells, AsPC-1 cells and HDBEC (2 Â 10 4 ), were placed in respective complete media, and were then supplemented with apixaban (1.8 µM), rivaroxaban (1.8 µM) or the DMSO vehicle with a further addition on day 2. Cell numbers were then determined after 4 days using the crystal violet staining procedure. Incubation of MDA-MB-231 and AsPC-1 cells with 1.8 µM apixaban reduced cell proliferation by 28 and 19%, respectively (►Fig. 5A and B), but had no detectable influence on primary endothelial cells (►Fig. 5C). Our data are in line with supplementation experiments carried out using apixaban, used at 5-to 50-fold higher than the therapeutic levels. 45 To further assess the involvement of endogenous fVIIa and PAR2, the expression of these proteins was   To determine the optimal concentrations for DOAC, fXa (10 nM) was pre-incubated for 30 minutes with a range of apixaban (0-1.8 µM) and rivaroxaban (0-1.8 µM). fXa activity was then determined using a chromogenic substrate (A) (n ¼ 3; Ã p < 0.05 vs. fXa alone). To ensure that the released microvesicles were free of any DOAC carry over, the microvesicles were incubated with fXa (10 nM) and fXa-chromogenic substrate (0.2 mM) and any reduction in fXa activity was recorded (B). A synthetic fVIIa-chromogenic substrate (NH2-Asn-Leu-Thr-Arg-pNA; 0.4 mM), or Perfachrome (0.4 mM), was incubated with purified fVIIa (5 nM) or purified fXa (10 nM) at 37°C for 1 hour and the absorptions at 410 nm determined (C) (n ¼ 4; Ã p < 0.05 vs. untreated sample). Samples of purified fVIIa (5 nM) were then pre-incubated with apixaban (1.8 µM), rivaroxaban (1.8 µM), an inhibitory polyclonal anti-fVIIa antibody or a control isotype, prior to addition of the fVIIa-chromogenic substrate and incubated at 37°C for 1 hour and measuring the absorption at 410 nm (D) (n ¼ 4; Ã p < 0.05 vs. untreated sample, #p < 0.05 vs. fVIIa alone). Combinations of tissue factor (TF) (1 U/mL), fVIIa (5 nM) and apixaban (1.8 µM) were incubated with Perfachrome substrate for 1 hour at 37°C and the absorptions measured at 410 nm (E) (n ¼ 5; Ã p < 0.05 vs. untreated sample). Samples of purified fVIIa (50 nM) were incubated with a range of concentrations of Perfachrome substrate (0.4-1.6 mM) in the presence or absence of apixaban (1.8 µM) and the rates of reaction measured (F) (n ¼ 4). By constructing Lineweaver-Burk plots, the Ki for the competitive inhibition of fVIIa by apixaban was determined to be 4.

Apixaban Reduces TF þ MV Release through Preventing PAR2 Activation
In order to determine if the release of TF þ MV and the increased rate of proliferation was in response to the activa-tion of PAR2 or PAR1, the cells were pre-incubated with the blocking antibodies against PAR2 (SAM11) or PAR1 (ATAP2) in the presence and absence of the DOAC. Any alterations in the release of TF þ MV were then analysed as above. Inhibition of PAR2 using SAM11 antibody reduced the release of TF þ MV from both cell lines in response to fXa (10 nM). This reduction was comparable in magnitude to those achieved on supplementation with apixaban or rivaroxaban (►Fig . 6A and B). In contrast, incubation with apixaban or rivaroxaban did not prevent the enhancement of the TF þ MV release following In the absence of fXa, only apixaban reduced the release of TF þ MV to levels attainable by blocking PAR2, but combinations of apixaban together with SAM11 antibody presented no further inhibitory potential, when tested in MDA-MB-231 cells (►Fig. 6F). To further examine the contribution of PAR2 to pro-metastatic cell function, sets of MDA-MB-231 cells were incubated with apixaban (1.8 µM), PAR2-blocking antibody (SAM11; 20 µg/mL) or used untreated. Total RNA was the extracted from the cells and the expression of IL-8 mRNA was measured using β-actin as a reference. The incubation of cells with apixaban resulted in 40% reduction in IL-8 expression compared to the untreated cells, while the inhibition of PAR2 reduced IL-8 expression by 56% (►Fig. 6G). Additionally, inhibition of PAR2 in these cells resulted in 47% decrease in cell proliferation (►Fig. 6H) and in agreement with our previous findings. 43

Apixaban Functions through Preventing the Activation of PAR2 by TF-fVIIa Complex
Since fXa appears to be absent from either of the cell lines or the MVs derived from them, the hypothesis that the auto-activation of PAR2 in cell lines was mainly induced by TF-fVIIa complex rather than fXa was investigated. Furthermore, the possibility that apixaban may be influencing TF-fVIIa directly was also examined. Supplementation of cells with fVIIa (5 nM) or fXa (10 nM) induced the release of MVs to a similar level (►Fig. 7A and B). However, the release of TF within the MVs was significantly amplified following incubation of cells with fVIIa (5 nM) (►Fig. 7C and D). Furthermore, the inclusion of apixaban, but not by rivaroxaban, abolished the release of MVs from the cell lines, in response to fVIIa (5 nM) (►Fig. 7E and F), as well as preventing the incorporation of TF into the released MVs (►Fig. 7G and H). Finally, pre-incubation of the MDA-MB-231 and AsPC-1 cells with an inhibitory anti-fVIIa antibody suppressed the rate of cell proliferation by 30 and 20%, respectively.

Apixaban is Estimated to Bind to fVIIa and fXa with Similar Affinity and Orientation
The examination of binding of apixaban to fVIIa showed a suitable binding orientation in which the catalytic site may be blocked (►Fig. 8). Furthermore, the calculated binding energy for the interaction of apixaban with fVIIa and fXa were calculated to be comparable (-9.78 and -11.66 kcal/mol, respectively). In addition, the ligand binding efficiencies were also determined to be of the same order of magnitude (-0.29 and -0.34, respectively).

Discussion
The activation of PAR1 and PAR2 is known to induce the release of MVs from cells. However, the incorporation of TF within the MVs is mediated by PAR2 but not PAR1. 21 Proteolytic activation of PAR2 during coagulation, is attained through the action of fXa and TF-fVIIa complex. 15 However, the expression and activation of fVII and fX by the cancer cell lines needs clarification. Furthermore, the contributions of each of these enzymes to the auto-activation of PAR2 and the release of TF þ MV have not been examined previously. Examination of these properties in two cell lines indicated that the auto-activation of PAR2 in these cell lines was mediated through TF-fVIIa complex, although supplementation with fXa further enhanced this process. In contrast, since no fXa was detected in the cells, it may be assumed that fXa does not play a major role in the maintenance of accelerated proliferation in these cancer cells. This finding also suggests that the expression and activation of fVII/fVIIa by cancer cells may be an additional mediator in the release of procoagulant MVs which needs to be taken into consideration as a risk factor, in conjunction to the expression of TF. [17][18][19][20]22,23 In fact, the released MVs from these cells appeared to contain both TF and fVIIa, and were procoagulant even if fVIIa was omitted from the assay, or when using fVII-deficient plasma. More-over, the cell-derived MVs showed proteolytic activity towards a fVIIa-specific chromogenic substrate which was inhibited in the presence of an inhibitory antibody towards fVIIa. In circulation, TF-fVII þ MV may interact with the surface of various cells 36 where they can potentiate PAR2 activation and the further release of MVs from the activated cells. Consequently, the TF-fVIIa complex associated with MVs amplifies the exposure and release of further procoagulant material into the bloodstream. Ultimately, changes in vascular structure through inflammatory responses and cellular apoptosis may constitute an indirect mechanism through which the TF-fVII þ MV may promote thrombosis associated with cancer.
The comparison of the effectiveness of the DOAC identified a remarkable property of apixaban as an inhibitor of the proteolytic activity of fVIIa, as well as the PAR2 signalling by TF-fVIIa complex. This ability was specifically dependent on fVIIa, was comparable to pre-incubation of the cells with fVIIa-inhibitory antibody and was distinct to that of inhibiting fXa in plasma. [46][47][48][49] Furthermore, the inhibitory potential of apixaban was not replicated on inclusion of rivaroxaban despite the higher sensitivity of coagulation mechanism to the latter. 50 In fact, the estimated Ki value for apixaban towards human fVIIa (4.4 nM) was in line but higher than the published values against human fXa; 0.08 nM 51  8 µM). Cell numbers were then measured after 2 day (E). Sets of cells were adapted to serum-free medium and pre-incubated with blocking antibodies to PAR2 (SAM11; 20 µg/ mL) or PAR1 (ATAP2; 20 µg/mL) in the presence of apixaban (1.8 µM), rivaroxaban (1.8 µM) or DMSO control. The density of microvesicles released from resting MDA-MB-231 (F) was determined using the Zymuphen MP assay kit (n ¼ 4; Ã p < 0.05 vs. DMSO alone). Sets of cells were incubated with apixaban (1.8 µM) or SAM11 antibody (20 µg/mL) and compared to respective vehicle-treated cells. Total ribonucleic acid (RNA) was extracted using the Ribozol solution. The expression of fVII messenger RNA (mRNA) or interleukin (IL)-8 was measured by GoTaq 1-Step RT-qPCR System using QuantiTect primers for fVII, IL-8 and β-actin, and relative amounts determined using a reference (G) (n ¼ 4; Ã p < 0.05 vs. untreated sample). Sets of MDA-MB-231 cells (2 Â 10 4 ) were also pre-incubated with antibody (20 µg/mL) and cell numbers determined cell proliferation compared to the untreated cells (H) (n ¼ 5; Ã p < 0.05 vs. untreated sample). Fig. 7 Apixaban prevents cell activation by inhibiting tissue factor-factor VIIa (TF-fVIIa) activity. Samples of fXa (10 nM) or fVIIa (5 nM) were incubated with an inhibitory anti-fVIIa polyclonal antibody (20 µg/mL) or an equivalent isotope control antibody, for 1 hour. Sets of MDA-MB-231 and AsPC-1 (2 Â 10 5 ) were adapted to serum-free media and incubated with the untreated enzymes, or the samples pre-incubated with the inhibitory or the isotype antibodies, for 2 hours. The density of microvesicles released from MDA-MB-231 (A) and AsPC-1 cells (B) was determined using the Zymuphen MP assay kit (n ¼ 5; Ã p < 0.05 vs. respective enzyme without any antibody). The TF antigen content of the microvesicles from MDA-MB-231 (C) and AsPC-1 cells (D) was also determined using the Quantikine TF-enzyme-linked immunosorbent assay (ELISA) assay (n ¼ 5; Ã p < 0.05 vs. respective enzyme without any antibody). Sets of cells were supplemented with fVIIa (5 nM) in the presence of apixaban (1.8 µM), rivaroxaban (1.8 µM) or the dimethyl sulfoxide (DMSO) vehicle. The density of microvesicles released from MDA-MB-231 (E) and AsPC-1 cells (F) was determined (n ¼ 5; Ã p < 0.05 vs. DMSO control). The TF antigen content of the microvesicles from MDA-MB-231 (G) and AsPC-1 cells (H) was also determined using the Quantikine TF-ELISA assay (n ¼ 5; Ã p < 0.05 vs. DMSO control). and 0.74 nM. 52 Moreover, the estimated Ki value is close to that reported for the synthetic fVIIa inhibitor, BMS-593214 (5 nM). 53 However, due to dissimilar analytical procedures used and also sensitivity to TF preparations, 47,52 we expect these differences to be much greater and therefore emphasise the need for verification, and urge caution at this point. In contrast, the lower circulating concentration of fVIIa compared to fXa suggests that apixaban may be effective in blocking the action of TF-fVIIa complex in vivo directly, as well as by preventing downstream mechanisms. 54 In either case, the rate of TF þ MV release was not further suppressed following blocking of PAR2 with SAM11 antibody. Therefore, the underlying mechanism appears to involve PAR2 signalling and requires the presence of TF-fVIIa complex. PARs are G protein-coupled receptors which act as sensors for the presence of active proteases, in particular the coagulation mechanism. 55,56 The discussions on the multiple roles of PAR2 activation, together with the multiple second messengers, the pathways involved and the cellular outcomes, has been thoroughly delineated previously 57,58 and is therefore beyond the scope of this study.
As stated above, the inhibitory function of apixaban towards fVIIa was not detectable with rivaroxaban. The differential action of various DOACs when examined with different cell lines has been demonstrated previously. 59 We have not examined this feature in any of the other fXablocking DOACs which include edoxaban, darexaban, otamixaban, betrixaban, letaxaban and eribaxaban. Also, despite the inhibition studies, at this point it is not possible to confirm any direct molecular interaction between apixaban and fVIIa. However, the dissimilar specificity of apixaban and rivaroxaban can prove to be a useful tool in deciphering the mechanisms of TF-fVIIa signalling and unscrambling these from the cellular outcomes of fXa action.
In conclusion, this study has shown that the up-regulation of the release of TF þ MV from cancer cells is mediated through auto-activation of PAR2 by the TF-fVIIa complex, on the surface of these cells. This in turn enhances the rate of proliferation in cancer cells. Importantly, we have for the first time shown that apixaban selectively inhibits the proteolytic activity of fVIIa and, therefore, may prove to be a useful agent in controlling cancer cell proliferation as well as the associated risk of thrombosis.
What is known about this topic?
What does this paper add?
• Apixaban and rivaroxaban reduce the release of TF þ MV through the activation of PAR2 by fXa. • Cellular TF-fVIIa complex also induces auto-activation of PAR2 on cancer cells, resulting in TF þ MV release and cell proliferation. • Apixaban competitively inhibits the proteolytic activity of fVIIa with similar kinetics to that of fXa. • Apixaban reduces cancer cell proliferation by inhibiting TF-fVIIa activity.

Authors' Contributions
This study was designed by S.F., A.M. and C.E., and the experimental work carried out by S.F. and Y.M. The data were evaluated by S.F., Y.M., A.M. and C.E. and the manuscript was prepared by S.F. and C.E.

Funding
This study was partly sponsored by Bristol-Myers Squibb (CV185-484) and pure substances were provided by Bristol Meyer Squibb (apixaban) and Bayer UK (rivaroxaban).