Female BALB/c mice (aged 7–11 weeks, 140 mice) were purchased from the Medical University of Bialystok (Poland) and housed 5–6 per cage in a temperature-controlled environment (22–25°C), 12-hour light/day cycle, and unlimited access to food (Zoolab, Krakow, Poland) and water throughout the experiment. All experimental procedures involving animals were accomplished in accordance with the approval of the Second Local Ethical Committee on Animal Testing, permit no 91/2018 and 259/2019. Mice welfare was monitored throughout the study once daily. Euthanasia was performed by intraperitoneal (i.p.) injection of ketamine and xylazine, 100 and 10 mg ⋅ kg⁻¹ of body weight, respectively.

In vivo studies utilised the mouse mammary adenocarcinoma 4T1-luc2-tdTomato cell line stably expressing the firefly luciferase gene and tdTomato fluorescent protein (the kind gift of Professor Joanna Wietrzyk from Ludwik Hirszfeld Institute of Immunology and Experimental Therapy, Polish Academy of Sciences (IIET)) at the 5^th passage from resuscitation after their purchase from Caliper Life Sciences Inc., USA; source of the parental line: ATCC, CRL-2539). The cells were cultured as previously described (Smeda et al., 2020a). Prior to i.v. injection, the cells were detached using Accutase solution (Sigma-Aldrich, Poland), centrifuged (300 × g, 4°C, 5 min), stained with Cell Tracker™ Red CMTPX Dye (Invitrogen, C34552) for 30 min at 37°C, rinsed 3 times with Dulbecco’s phosphate-buffered saline (DPBS) (Gibco), DPBS and growth medium 1:1 (Hank’s Balanced Salt Solution (HBSS), IIET, Poland), resuspended in HBSS at the required concentration, and injected into the tail vein of female BALB/c mice (7.5 × 10⁴ cells in 100 μl of HBSS per mouse). For in vitro experiments, the HLMVEC line was used. The cell line was obtained from the European Cell Culture Collection (Cell Applications), cultured in microvascular endothelial cell growth medium (Cell Applications) at 37°C in an atmosphere of 5% CO₂. Both cell cultures were routinely tested for Mycoplasma contamination.

Anaesthetized mice (100 mg ⋅ kg⁻¹ ketamine +10 mg ⋅ kg⁻¹ xylazine, i.p. were injected via the femoral vein with a solution of 2% Evans blue (EB; 60 kDa) dye (Sigma Aldrich) in .9% saline at a dose of 4 mg ⋅ kg⁻¹, as described previously (Smeda et al., 2018). The injected dye solution was left to circulate for 10 min, and the lungs were perfused with PBS for 15 min, isolated, weighted, and homogenised in 200 μl of 50% TCA (dissolved in distilled water). The homogenate was centrifuged (at 12,000 rpm for 12 min at 4°C) and stored at −80°C. Prior to EB concentration measurement (absorbance at 620 nm, plate reader Synergy 4, BioTek), the supernatant was diluted at a 1:3 ratio with 95% ethanol. The results were normalised to tissue weight .

Excised lungs were washed with saline, fixed in formalin, paraffin-embedded, and cut into 5-μm slices. For the 24 h- and 2-day time points, the number of viable cancer cell colonies was counted based on their vivid and stable red fluorescence and normalised to the area of the lung cross-section. For the 7-day time point, quantitative assessment of already well-defined pulmonary metastases was based on classical haematoxylin and eosin (H&E) staining and normalised to the area of the lung cross-section as described previously (Smeda et al., 2020a).

Thirty mice were pre-treated with dabigatran etexilate (Biorbyt, cat. no orb180748) by oral gavage twice daily at a total dose of 100 mg per kg of body weight per day in .05% natrosol, and the same number of mice simultaneously received the vehicle (.05% natrosol) for 3 consecutive days. The dosing regimen effectively reduced pulmonary metastasis in the orthotopic model of breast cancer (DeFeo et al., 2010). All mice were subsequently injected with 7.5 × 10⁴ mouse mammary adenocarcinoma 4T1-luc2-tdTomato (4T1) cells. Dabigatran etexilate or vehicle treatment was continued (twice daily Monday to Friday and once daily over the weekend) until the animal euthanasia at 24 h, 2 days, and 7 days after injection of cancer cells.

To measure pulmonary permeability alongside the progression of cancer, 40 mice were pre-treated with dabigatran etexilate in .05% natrosol twice daily at the final dose of 100 mg ⋅ kg⁻¹ 24 h, while the other 40 received the vehicle (.05% natrosol). Afterwards, sixty mice were injected with 4T1 cells and euthanized at 24-h, 2-day, and 7-day timepoints after the injection, whereas 20 healthy control mice were euthanised after the pre-treatment period to visualise if dabigatran etexilate affected the pulmonary endothelium barrier in the mice not injected with breast cancer cells. Just before euthanasia, all animals were injected intravenously with EB dye solution, perfused with phosphate-buffered saline, and their lungs were excised for subsequent analysis.

Lungs were fixed in formalin, paraffin-embedded, cut into 5-μm slices, and mounted on the slides. For fibrinogen immunohistochemical staining, deparaffinized slide-attached lung cross-sections were stained with polyclonal goat antiserum to mouse fibrin(ogen) 1:250 (Accurate Chemical & Scientific Corporation, YNGMFBGBio), followed by biotin-SP donkey anti-goat secondary antibody 1:600 (Jackson Immuno Research, 112-065-003). Stained lung cross-sections were subsequently scanned with a BX51 microscope equipped with virtual microscopy system dotSlide (Olympus, Japan). Image segmentation was performed in Ilastik (developed by the Ilastik team, with partial financial support by the Heidelberg Collaboratory for Image Processing, HHMI Janelia Farm Research Campus, and CellNetworks Excellence Cluster).

The effect of dabigatran on thrombin activity was measured in murine plasma using a thrombin generation assay according to (Tchaikovski et al., 2007) with major modifications. First, citrated murine plasma was mixed with fluorogenic substrate (Z-Gly-Gly-Arg-AMC) solution and pipetted into the plate wells, and thrombin generation was initiated by the addition of the trigger solution, containing tissue factor (TF), phospholipids (PL), and CaCl₂. As a result, 60 µl of the prepared mixture in a well contained 12 µl of plasma, 9 µl of substrate solution, and 39 µl of trigger solution, with a final concentration of 20% plasma, 1.0 pM TF, 16.25 mM CaCl₂, 4 μM PL, and .43 mM ZGGR-AMC. Each plasma sample was calibrated by replacing the trigger solution with a solution containing α2-macroglobulin-thrombin complex (α2M-T, at a final concentration corresponding with 44 nM thrombin activity). Measurements were performed at 37°C and each sample was tested in duplicate. Fluorescent signals were recorded using Tecan Spark 10M microplate reader (Männedorf, Switzerland) and transformed into thrombin concentration as described previously (Hemker and Kremers, 2013). The effect of dabigatran was evaluated based on the parameter lag time, representing the time required to start the generation of thrombin. All reagents were provided as a gift by Synapse BV, Netherlands.

Lungs were excised, rinsed in saline, dried with tissue paper, weighed, cut into small pieces, and snap-frozen in liquid nitrogen; the samples were stored at − 80°C. Lung tissue was homogenised in T-PER buffer (Thermo Scientific, 78510), total protein concentration was determined with BCA Assay Kit (Thermo Scientific, 23225), and the homogenates were aliquoted and stored at −80°C for further analysis. Concentrations of INFγ and iC3b were determined with commercially available ELISA kits (R&D Systems (MIF00) and MyBioSource (MBS9393088), respectively). For Western blotting, an equal amount of protein was loaded and run on the gels, and then transferred to a nitrocellulose membrane, blocked overnight with 5% dry milk, and incubated overnight with the appropriate primary antibodies in TBS (Tris-buffered saline) directed against the following antigens: Ang-2 (Thermo Scientific, PA5-27297, 1:10 000), E-selectin (Bioss Antibodies, bs-1273R, 1:1000), MMP-9 (Millipore, AB19016, 1:1000). After washing, membranes were incubated with anti-rabbit secondary antibodies conjugated with HRP (Santa Cruz Biotechnology, sc-2357, 1:5000) diluted in TBS buffer. Chemiluminescent signal was developed with Clarity Western ECL Substrate (Bio-Rad, 1705061) using ChemiDoc Imager (Bio-Rad) and densitometric analysis of band intensity was performed in ImageJ. The total protein loaded onto the particular lane after transfer was used as the loading control using stain-free technology from Bio-Rad as previously described (Smeda et al., 2018).

Mouse citrated blood (blood/3.8% citrate at 10:1 (v/v)) was diluted with saline and washed with Tyrode’s buffer. Washed blood samples obtained from dabigatran-treated or untreated cancer cell-injected mice were directly subjected to thrombin (.025 U ml⁻¹) activation and antibody staining described below, whereas washed blood samples derived from healthy mice were pre-incubated with 1, 10, 30, and 100 ng ml⁻¹ of dabigatran or dabigatran solvent alone (DMSO; the final DMSO concentration was <1%) to assess the effects of the given dabigatran concentrations on platelet thrombin-induced (.025 U ml⁻¹ or .1 U ml⁻¹) reactivity. Each blood sample was double-stained with an antibody panel that included a platelet-specific antigen, FITC-conjugated GpIIb/IIIa (CD41/61) (Rat IgG2a, clone Leo.F2, Emfret Analytics, #M025-1, 1:5) for platelet identification as well as a platelet activation marker the PE-conjugated active form of GPIIb/IIIa (Rat IgG2b, clone: JON/A, Emfret Analytics, M023-2, 1:5). Platelets were identified based on their forward- and side-scatter characteristics and gated based on the expression of platelet-specific antigen CD41/61 as previously described (Smeda et al., 2018). Unstained platelets were used to establish the level of autofluorescence that was set to fall within the first log order of brightness for each fluorescence channel. FITC-conjugated isotype control antibody (Rat IgG (negative control), Emfret Analytics, P190-1, 1:5) was used to assess non-specific binding for each individual sample. Basal and thrombin-induced (.025 U ml⁻¹ or .1 U ml⁻¹) reactivity of circulating platelets was assessed based on the measured expression/binding level of an active form of GPIIb/IIIa expressed as the percentage of all platelets above isotype control fluorescent signal and/or the median fluorescence intensity (MFI). Flow cytometric analyses of platelet activation was performed using BD LSR II instrument (BD Biosciences) and analysed in BD FACSDiva 6.0 software (Becton Dickinson, Oxford, United Kingdom). Measurements were made on a logarithmic scale and at least 10,000 events were collected for each sample. Appropriate colour compensation was determined in samples singly stained with FITC-conjugated anti-CD41/61. FITC and PE dyes were excited by a 488 nm laser.

The concentration of dabigatran (administered to mice as a pro-drug dabigatran etexilate) was assessed in plasma samples using the UHPLC-MS/MS technique. The plasma volume of 50 µl was spiked with 5 µl of internal standard (IS, dabigatran-¹³C₆) at the concentration of 1 µg ml⁻¹. After gently shaking, 150 µl of .1 M HCl in methanol was added, mixed for 10 min and chilled at 4°C for 10 min. The supernatant collected after sample centrifugation (15,000 rpm, 4°C, 15 min) was directly injected onto an UltiMate 3000 UHPLC system (Thermo Fisher Scientific, Waltham, MA, United States) combined with a TSQ Quantum Ultra triple quadrupole mass spectrometer (Thermo Fisher Scientific, Waltham, MA, United States). The chromatographic analysis was conducted using an ACQUITY UHPLC BEH C18 (3.0 × 100 mm, 1.7 μm, Waters, Milford, MD, United States) analytical column and applying .1% FA in ACN (A) and .1% FA in H₂O (B) as mobile phases delivered in the following gradient elution program: 95% B for 1 min, 95%–5% B for 3 min, 5%–95% B for .5 min, and 95% B for 2.5 min for column equilibration. The mass spectrometric detection was conducted in an electrospray positive ionization mode, and selected ion transitions were used for quantification: 472.4→172.0 (CE = 39 V) and 478.3→172.1 (CE = 39 V) for dabigatran and dabigatran-¹³C₆, respectively. The plasma concentration of dabigatran was calculated based on the regression equation determined for the calibration curve plotted for dabigatran and expressed as the relationship between the peak area ratios of analyte/IS to the nominal concentration of the analyte.

The response of the barrier formed by human lung microvascular endothelial cells (HLMVECs) to an inflammatory agent (IL-1β, 10 ng ml⁻¹) was assessed in real-time in a fully standardized manner by continuously recording changes in electrical resistance with an ECIS system (Bernas et al., 2010), using 8W10E + or 96W10E + electrode chamber arrays and an ECIS Z-Theta system (Applied Biophysics), along with the associated software v.1.2.126 PC. Immediately after cell seeding, resistance measurements (in Ohms) were initiated, and the increase in resistance with respect to time indicated that cells were forming contacts between each other. The steady state was achieved when maximum resistance was reached, which indicated the formation of a tight monolayer.

On that day, platelets were isolated from the human citrate blood samples obtained from the Blood Donation Centre. Obtained platelet-rich plasma (PRP) (blood was centrifuged at 260 × g with slow deceleration) was supplemented with prostacyclin to a final concentration of 100 ng ml⁻¹ and centrifuged to obtain platelet pellets (960 × g for 10 min). Platelet pellets were resuspended in PBS with albumin (1 mg ml⁻¹) and glucose (1 mg ml⁻¹), subsequently centrifuged in presence of prostacyclin (100 ng ml⁻¹, 810 × g for 10 min), and then the supernatant was discarded. Washed platelets were resuspended in Tyrode’s buffer to a final number of 1 × 10⁶ PLT and stimulated with bovine thrombin (.1 U ml⁻¹), collagen (5 μg ml⁻¹), or ADP (20 μM) or remained unstimulated for 15 min, including their centrifugation time (10 min, 900 × g). Some platelet samples were pre-treated with dabigatran (30 ng ml⁻¹) and subsequently stimulated with bovine thrombin (.1 U ml⁻¹) or remained unstimulated. After centrifugation to discard platelets, platelet releasates were added to HLMVEC cultures with or without IL-1β, and Tyrode’s buffer was added to control wells. Changes in cell resistance were measured in real-time for the consecutive 75 min.

Measurement of the plasma concentration of selected endothelial dysfunction markers was performed using the microLC/MS-MRM method as described previously (Walczak et al., 2015; Suraj et al., 2018; Suraj et al., 2019a; Suraj et al., 2019b). The panel used in this study included biomarkers of endothelium permeability (Ang-1, Ang-2, sFLT-1 (soluble form of fms-like tyrosine kinase 1), and sTie-2 receptor), biomarker of glycocalyx disruption (SDC-1), biomarkers of endothelium inflammation and haemostasis (sE-sel, sICAM-1, vWF, t-PA, and PAI-1), and biomarkers of platelet activation (sP-sel and THBS-1). To assess the plasma concentration of the selected biomarkers, the UHPLC Nexera system (Shimadzu, Kyoto, Japan) connected with a highly sensitive mass spectrometer, QTrap 5500 (Sciex, Framingham, MA, United States), were used. Sample preparation included proteolytic digestion with porcine trypsin to achieve the unique and reproducible peptide sequences that were applied as the surrogates of the proteins suitable for LC-MS/MS analyses. A detailed description of the targeted analysis of the selected panel of proteins was presented elsewhere (Suraj et al., 2018; Suraj et al., 2019a; Suraj et al., 2019b).

In brief, total protein concentration in plasma was assessed using a NanoDrop 8000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, United States). The samples were diluted to a concentration of 7 mg ml⁻¹ using 25 mM ammonium bicarbonate (NH₄HCO₃) (Sigma-Aldrich) and 210 μg of total protein was denatured with 10% solution of sodium deoxycholate (Sigma-Aldrich) in 25 mM NH₄HCO₃, then diluted with 25 mM NH₄HCO₃. Samples were reduced with 50 mM tris(2-carboxyethyl)phosphine (TCEP) (Sigma-Aldrich) in 25 mM NH₄HCO₃ for 30 min at 60°C and alkylated in the dark with 100 mM iodoacetamide (IAM) (Sigma-Aldrich) in 25 mM NH₄HCO₃ for 30 min at 37°C. Next, the excess of IAM was quenched by the addition of 100 mM DL-dithiothreitol (DTT) (Sigma-Aldrich) in 25 mM NH₄HCO₃. The incubation of the samples lasted 30 min at 37°C. Proteolytic digestion was carried out for 16 h at 37°C with sequencing grade modified trypsin (400 µg ml⁻¹) in a 50:1 ratio (substrate:enzyme) (Promega, Madison, USA). Prepared internal standard peptide solutions obtained from dissolved powders of internal standard peptides, synthesized and quality-controlled by Innovagen (Lund, Sweden) (Suraj et al., 2018; Suraj et al., 2019b); specific for each target peptide sequence were implemented just before ceasing the digestion. Finally, this process was stopped by the addition of formic acid (FA) (Sigma-Aldrich) at a final concentration of .5% v/v. Next, sodium deoxycholate was pelleted. The sample supernatant obtained after centrifugation (3,000 × g for 10 min at 23°C) was desalted and concentrated by the micro-solid phase extraction (μSPE) procedure using the Oasis HLB elution plate with 2 mg of sorbent mass per well (Waters, Milford, USA). Finally prepared samples were lyophilized and, directly before LC-MS/MS analyses, resuspended in 50 μl 20% ACN in H₂O.

Statistical significance of in vivo data was assessed in GraphPad Prism 5.03 with a parametric two-way ANOVA or non-parametric Kruskal-Wallis test followed by the appropriate post hoc test (Bonferroni or Dunns Multiple Comparison Tests) based on the normality of the data distribution (tested with a Shapiro-Wilk normality test), homogeneity of variances (tested with an F test), and the variable scale. The data were presented as the median and the interquartile range (IQR) (from lower (25%) to upper (75%) quartile). Only p values ≤.05 were considered significant. We analysed the differences between untreated mice at given timepoints (black symbols on the graphs), the differences between dabigatran-treated mice at given timepoints (grey symbols on the graphs), and the differences between untreated and dabigatran-treated mice at the given timepoint, and significant differences (p < .05) were indicated in the graphs. The significant outliers identified by Grubbs test were excluded from the statistical analysis. The in vitro data illustrating changes of the endothelial barrier in real time were subjected to the analysis of covariance (ANCOVA) based on their parallel course confirmed with the model of equal slopes analysis in Statistica 13.

Intravenous injection of 4T1 breast cancer cells into female BALB/c mice resulted in robust accumulation of cancer cells in their lungs. The number of metastatic cancer cells was high 24 h after 4T1 cancer cell injection (Figures 1A,B) and then progressively declined on the 2^nd and 7^th day after cancer cell injection (Figures 1A,D,F–H). These findings could be explained by the progressive apoptosis of intravascular cancer cells that did not form metastatic colonies, as shown previously by (Qiu et al., 2003). In the lungs of dabigatran-treated mice, the number of accumulated cancer cells was not different than that in the lungs of untreated mice 24 h after the injection (Figures 1A–C); however, on the 2^nd and 7^th day after 4T1 cancer cell i.v. administration, the number of early cancer cell colonies and well-established pulmonary metastases, respectively, were substantially higher in dabigatran-treated mice than in the untreated mice (Figures 1A,D–J).

To verify whether the unfavourable effects of dabigatran on pulmonary metastasis could be related to inhibition of fibrin deposition or innate immune response, we quantitively analysed fibrin levels in the murine lungs (Figure 2A) in relation to thrombin generation in plasma (Figure 2B), interferon γ (IFNγ) concentration in lung homogenates (Figure 2C), and complement activation (Figure 2D). Dabigatran treatment of 4T1 breast cancer cell-injected mice did not affect fibrin deposition in the lungs (Figure 2A). At early timepoints (24 h and 2 days after injection), dabigatran treatment also did not affect the lag time of thrombin generation in the plasma (Figure 2B) (though it tended to be increased 24 h after injection for dabigatran-treated mice), did not alter the concentration of IFNγ reflecting activation of NK cells (Takeda et al., 2011) (Figure 2C), or did not modify the activation of complement (Figure 2D). The changes in the above parameters observed on the 7^th day after i.v. injection were rather secondary and might have resulted from the difference in the metastatic count at that time between untreated and dabigatran-treated mice (Figure 1F).

Treatment with dabigatran did not alter lung permeability of mice not injected with 4T1 breast cancer cells (Figure 3A). Surprisingly, 4T1 cancer cell injection resulted in increased pulmonary vascular permeability in dabigatran-treated mice 24 h after cancer cells injection (Figure 3A). The higher lung permeability 24 h after i.v. injection in dabigatran-treated and 4T1 cancer cell-injected mice was followed by increased breast cancer cell extravasation and settlement in the lungs 2 days after injection (Figure 1A) that was associated with an increased inflammatory response in the lungs of dabigatran-treated mice (Angiopoietin-2 (Ang-2), Figure 3B; E-selectin, Figure 3C; and matrix metalloproteinase 9 (MMP-9), Figure 3D).

The Figure 4 shows selected plasma parameters that might indicate systemic endothelial dysfunction. To assess the functional status of the endothelium at the systemic level in the absence or presence of dabigatran treatment in 4T1 breast cancer cell-injected mice, selected protein biomarkers of endothelial dysfunction were measured in the mouse plasma (Figure 4) as previously described in the orthotopic 4T1 breast cancer model (Suraj et al., 2019a). The panel used in this study included selected biomarkers of endothelial permeability (Angiopoietin 1 (Ang-1), Ang-2, soluble form of Tie-2 (sTie-2), and soluble form of VEGF receptor 1 (sFLT-1); Figures 4A–D, respectively), glycocalyx damage (syndecan 1 (SDC-1); Figure 4E), endothelial inflammation and haemostasis (soluble form of E-selectin (sE-sel), soluble form of ICAM-1 (sICAM-1), von Willebrand factor (vWF), plasminogen activator inhibitor-1 (PAI-1), t-PA (tissue plasminogen activator); Figures 4F–J, respectively), and platelet activation (soluble form of P-selectin (sP-sel) and thrombospondin 1 (THBS-1); Figures 4K,L, respectively).

There were no significant changes in the plasma concentration of Ang-1 (Figure 4A), while plasma Ang-2 concentration and soluble form of Tie-2 receptor for angiopoetins were both increased by dabigatran treatment 2 days after the injection of breast cancer cells compared to untreated mice (Figure 4B). However, concentration of soluble form of sFLt-1 receptor for VEGF was lower in the plasma of dabigatran-treated mice 24 h after the injection compared to untreated mice (Figure 4D). Plasma concentration of SDC-1 was increased by dabigatran treatment 24 h after the injection of cancer cells, but this difference disappeared 2 days after cancer cells injection i.v. (Figure 4E). No changes were detected in the plasma sE-sel (Figure 4F), sICAM-1 (Figure 4G), vWF (Figure 4H), and PAI-1 (Figure 4I), while tPA-1 concentration in the plasma was decreased by dabigatran treatment 24 h after i.v. injection of cancer cells (Figure 4J). Concentrations of sP-sel (Figure 4K) and THBS-1 (Figure 4L) were lower in dabigatran-treated mice 2 days after injection of 4T1 breast cancer cells.

The most important changes between untreated and dabigatran-treated mice were detected 24 h and 2 days after i.v. injection of cancer cells. They corresponded to increased vascular permeability in the lungs 24 h after i.v. injection of cancer cells (enabling more efficient extravasation of cancer cells) what, consequently, induced more pronounced inflammatory response in the lungs of dabigatran-treated mice 2 days after i.v. injection of cancer cells. We claim that higher pulmonary endothelium permeability 24 h after i.v. and, consequently, higher levels of Ang-2 (Figure 3B), E-selectin (Figure 3C) and MMP-9 (Figure 3D) 2 days after i.v., provided more favourable microenvironemnt for the increased settlement of 4T1 breast cancer cells in the lungs of dabigatran-treated mice. The earliest plasma marker indicating more pronounced endothelial damage in dabigatran-treated mice was increased syndecan-1 (SDC-1), the proteoglycan of endothelial glycocalyx and determinant of blood vessel permeability (Fernández-Sarmiento et al., 2020) In fact, plasma SDC-1 was higher 24 h after i.v., indicating its increased shedding from endothelial surface into the circulation in dabigatran-treated mice at that time (Figure 4E) what was compatible with the alterations in lung endothelial permeability at that time. Interestingly, 2 days after i.v. injection of cancer cells, upregulation of Ang-2 (Figure 4B) and the soluble form of its receptor Tie-2 (Figure 4C) were detectable in the plasma, while the concentration of sFlt-1 in the plasma was not increased (Figure 4D), indicating that, in contrast to Ang-2-dependent pathway, VEGF-dependent signalling might be not involved in the negative effects of dabigatran in 4T1 breast cancer cell-injected mice. Finally, in dabigatran-treated mice 2 days after i.v., injection of cancer cells concentration of biomarkers of platelet activation was lower as judged by a lower THBS-1 (Figure 4L) and sP-sel (Figure 4K), underscoring platelet inhibition, what could have resulted in inefficient platelet-dependent protection of pulmonary endothelium integrity in inflammatory conditions (Middleton et al., 2016). Interesting early changes were also observed in the plasma concentrations of plasminogen activator (t-PA) that, surprisingly, was lower in dabigatran-treated mice 24 h after i.v. injection of cancer cells (Figure 4J) what might have facilitated pro-inflammatory activation of macrophages as shown by (Miles and Parmer, 2017).

It is worth noting that dabigatran treatment did not affect the concentration of abovementioned markers of endothelial dysfunction in plasma in healthy mice with the exception of vWF, the concentration of which was lowered (Supplementary Figure S1).

At the time of blood collection, the plasma concentration of dabigatran was about 10 ng ⋅ ml⁻¹ in all experimental groups (Figure 5A). To test whether such a concentration of dabigatran affected thrombin-induced platelet activation, platelets were stimulated ex vivo by low (.025 U ml⁻¹) or high (.1 U ml⁻¹) concentrations of thrombin in the absence or presence of dabigatran at a concentration range of 1–100 ng ⋅ ml⁻¹, and the expression of active GpIIb/IIIa receptor was quantified by flow cytometry (Figures 5B,C). As shown in Figures 5B,C, even a 1 ng ⋅ ml⁻¹ concentration of dabigatran inhibited platelet activation induced by .025 U ml⁻¹ of thrombin. In turn, 30 ng ⋅ ml⁻¹ dabigatran completely inhibited platelet activation induced by a higher concentration of thrombin (.1 U ml⁻¹). Furthermore, the reactivity of platelets obtained from dabigatran-treated mice was diminished both at 24 h as well as on the 7^th day after i.v. injection of 4T1 breast cancer cells, as evidenced by diminished expression of the active form of GPIIb/IIIa in dabigatran-treated mice compared with non-treated mice in response to thrombin (.025 U ml⁻¹) (Figures 5D–F), directly supporting the inhibitory effects of dabigatran treatment in vivo on thrombin-dependent platelet activation.

To verify whether platelet releasate from thrombin-activated platelets could protect pulmonary endothelial barrier integrity in inflammatory conditions, HLMVECs monolayers were challenged with the inflammatory mediator interleukin (IL)-1β. This cytokine was chosen as it is released in vivo by activated monocytes/macrophages, increasing pulmonary permeability in the model of experimental metastasis (Häuselmann et al., 2016). IL-1β impaired endothelial barrier function (control vs. IL-1β, p < .001) (Figures 6A,B,D). However, in the presence of the releasate from washed non-stimulated (quiescent) human platelets R [PLT_q] (Figure 6A) as well as in the presence of releasate of washed thrombin-stimulated human platelets R [PLT_thr] (Figure 6B), IL-1β-induced impairment of endothelial barrier function was attenuated (p < .05 and p < .001, respectively), clearly showing platelet-dependent protection of the endothelial barrier in the lungs. The protective effect of platelet releasates afforded by thrombin-stimulated human platelets (p < .001) (Figure 6B) was more pronounced compared with non-stimulated platelets (p < .05) (Figure 6A). Most importantly, dabigatran substantially inhibited the pulmonary endothelium barrier-supportive effect of thrombin-stimulated platelet releasate (R [PLT_thr + D]) in the absence of IL-1β (Figure 6C) and in the presence of IL-1β (Figure 6D) (p < .001 and p < .001, respectively).

Furthermore, the barrier-protective effect of platelet releasates in the presence of IL-1β was to some degree agonist-specific (Figure 6; Supplementary Figure S2). Although thrombin- and collagen-stimulated PLT releasates protected the endothelial barrier integrity in the absence and in the presence of IL-1β, ADP-stimulated PLT releasate had the negative effect on the endothelial barrier in the absence (p < .001) as well as in the presence (p < .001) of IL-1β compared to control.