Local blood coagulation drives cancer cell arrest and brain metastasis in a mouse model

Brain metastases (BMs) remain a clinical challenge in modern oncology. Up to 50% of patients suffering from metastatic melanoma and 30% to 50% of patients with human epidermal growth factor receptor 2 (HER2)–positive and triple-negative breast cancer develop BMs during the course of their disease.¹  BMs are by nature a multifocal disease in a delicate organ, which explains why treatment options are limited and why median overall survival remains as low as 3 to 9 months.²  Moreover, associated multiple neurologic complications significantly compromise patients’ quality of life. This makes prevention of brain colonization by cancer cells and the subsequent formation of BMs an attractive general concept,^2-5  but the best biological strategy remains obscure, and no prospective clinical trials investigating drugs with respect to their potential for preventing BMs have yet been completed.

Interestingly, a growing body of evidence suggests long-term benefits of antithrombotic drugs on patient survival which, in solid tumors, largely depends on the inhibition of metastatic spread. Antithrombotic therapy prevents cancer-associated adverse effects such as deep vein thrombosis and pulmonary embolism.⁶  Treatment with low molecular weight heparin (LMWH) has primarily been associated with a survival benefit in patients with advanced cancer.^6,7  Although this largely retrospective data must be viewed with caution, they point toward the potential of antithrombotic agents that reaches beyond their current clinical use. The exact mechanisms of action of their potential antimetastatic activity remain elusive, as does the question of whether BMs might also be affected.

The aim of this study was to specifically investigate the impact of cancer cell–associated microthrombus formation within brain capillaries on the distinct steps of the brain metastatic cascade⁵ : intravascular arrest of circulating cancer cells, extravasation, microscopic brain colonization, and formation of a clinically relevant macrometastasis. A newly developed in vivo multiphoton laser-scanning microscopy (MPLSM) method combined with other technologies has allowed unprecedented insights into the interplay of the blood coagulation system with the entire brain metastatic cascade. Anticoagulation strategies and antiplatelet approaches have now been tested for their ability to inhibit the formation of BMs.

For additional information, please see the supplemental Data, available at the Blood Web site.

Informed written consent was obtained from all participants in accordance with the Declaration of Helsinki and the International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use guidelines. The protocol received approval from the ethics committee of the Medical Faculty Mannheim, Heidelberg University (Heidelberg, Germany; 2012-293N-MA). The human breast cancer cell lines Jimt1,⁸  and melanoma cell lines A2058 and H1Dl2 (H1)⁹  were cultured under standard media conditions (Dulbecco’s modified Eagle medium, 10% fetal bovine serum, 1% penicillin and streptomycin). For in vivo imaging,^8,10  cells expressing green fluorescent protein or red fluorescent protein were brain-passaged 3 times as described previously.^5,8 

All animal procedures were approved by the Regierungspraesidium Karlsruhe (Karlsruhe, Germany) (governmental authority for laboratory animal research; G188/12 and G220/16). Male NMRI-nu mice were used for melanoma studies, and female nu/nu mice were used for breast cancer studies. Mice (8 to 12 weeks old) received a chronic cranial window (6 mm in diameter), and a single tumor cell (TC) suspension was injected into the left cardiac ventricle as reported before.^5,8,10 

For in vivo imaging and MPLSM, we used a Zeiss LSM 7MP microscope equipped with a Coherent Chameleon Ultra II laser. Images were taken through a BP500-550/BP575-610 filter with a 20×/1.0 W-Plan-Apochromat objective (Zeiss, Jena, Germany), which allowed for a subcellular resolution of <0.5 µm in a depth of 200 to 500 µm. Rhodamine 6G dye (0.5 mg/mL; Sigma-Aldrich, Schnelldorf, Germany),^11,12  1 mg of tetramethylrhodamine-isothiocyanate (TRITC) dextran (500 g/mol; 52194, Sigma-Aldrich) or fluorescein-isothiocyanate dextran (FITC; 2000 g/mol; FD-2000S, Sigma-Aldrich) was injected intravenously (IV). Multimodal correlative microscopy was performed as described previously^13-15  to retrieve the arrested TCs for electron microscopy (EM) imaging. Cranial magnetic resonance imaging (MRI) was performed in the fourth week after intracardiac TC injection on a 9.4 Tesla small animal nuclear magnetic resonance scanner (BioSpec 94/20 USR, Bruker BioSpin, Ettlingen, Germany).¹⁶ 

Polyclonal FITC-conjugated anti-von Willibrand factor (anti-VWF antibody; GTX28822, GeneTex, Irvine, CA; detailed characterization provided in supplemental Figure 1) and an unspecific, host consistent (sheep immunoglobulin G [IgG]) FITC-conjugated isotype control (20006-1-F, Cedarlane Laboratories, Burlington, Canada) were used. After purification with a column (HiTrap, GE29-0486-84, Sigma-Aldrich), antibodies (4 µg/g) were injected IV 30 minutes before TC injection. Antibody injections (1 µg/g) were given before every imaging session.^17,18 

Treatments with acetylsalicylic acid (ASA [aspirin], 5 µg/g; Bayer, Leverkusen, Germany) and clopidogrel (25 µg/g; AbZ-Pharma, Ulm, Germany) were administered by oral gavage once per day to inhibit clot formation. Tirofiban (0.5 μg/g; #30165, Sigma-Aldrich) was injected IV 30 minutes before TC injection, then every 4 hours for 24 hours, then 3 times per day until day 10 post intracardiac injection (p.i.i.).¹⁹  Injection of polyclonal anti-glycoprotein Ib (GPIb) (4 µg/g; R300, Emfret Analytics, Würzburg, Germany) 2 hours after TC injection resulted in effective platelet depletion for more than 4 days.²⁰  Polyclonal nonimmune IgG (C301, Emfret Analytics) served as an isotype control. Tinzaparin (innohep; Leo Pharma, Neu-Isenburg, Germany) was injected subcutaneously (0.6 IU/g).²¹  Dabigatran (80 mg/kg; Pradaxa, Boehringer Ingelheim, Ingelheim am Rhein, Germany) was given orally via gavage 2 times per day 2 days before intracardiac injection until day 10 p.i.i. Orally administered substances were dissolved in a methylcellulose solution (0.5% methylcellulose; 0.2% polysorbate) in a maximum volume of 0.2 mL per application.

Mice were euthanized on day 3 or day 28 p.i.i. Cryosections (10 µm) of fresh frozen brains without trans-cardiac perfusion were taken 130 µm apart to avoid overlapping of metastases over several slices, and hematoxylin and eosin staining was applied. Images were acquired via slide scanner (Axio Scan.Z1, Zeiss) in bright field configuration with a 20×/0.8 Plan-Apochromat objective (Zeiss) and a Hitachi HV-F202SCL camera.

The following primary antibodies were used: polyclonal sheep anti-human VWF-FITC (1:150; GTX28822, GeneTex), polyclonal rabbit anti DsRed (1:100; 632496, Takara Bio USA, USA), polyclonal rabbit anti-human VWF (1:150, A0082, DAKO, Glostrup, Denmark), and monoclonal mouse anti-human thrombospondin (1:75, MS-421-R7, Thermo Fisher Scientific, Waltham, MA). The following secondary antibodies were used: Alexa 555-conjugated goat anti-rabbit (1:200; Invitrogen, Carlsbad, CA), Alexa 555–conjugated goat anti-mouse (1:200; Invitrogen), and FITC-conjugated goat anti-rabbit (1:200; BD Pharmingen, San Diego, CA). Nuclei were stained with 4,6 diamidino-2-phenylindole (DAPI).

Plasma coagulation time was measured using a mechanical ball coagulometer (MC-4 PLUS, Merlin Medical, Lemgo, Germany). TCs (400 000 per mL) were diluted (1:1, 1:4, 1:8, 1:16) in human plasma and coagulation was initiated by CaCl₂ (0.025M; #ORHO37, Siemens, München, Germany). Thrombin generation of TCs (10 000 per 100 µL) with or without human plasma (centrifuged at 1000g for 15 minutes) was performed according to the manufacturer’s instructions (Technoclone, Wien, Austria). This assay is based on the cleavage of a fluorogenic substrate for the determination of thrombin activity.

TC-induced platelet aggregation was evaluated by light transmission aggregometry (LTA) using a spectrophotometer (PowerWave XS2, BioTek Instruments, Winooski, VT). Platelets were isolated from citrated blood of human donors after approval by the local ethics committee. TCs (10 000 per 100 µL) with or without human plasma were added to the platelets. Tinzaparin (100 IU/mL) was added for 30 minutes, and thrombin (0.5 IU/mL) served as positive control. Light transmission was recorded at 600 nm at 37°C.

Human TCs where stained with phycoerythrin (PE)–conjugated anti-tissue factor (anti-TF; 0.2 mg/mL; #12-1429, Thermo Fisher Scientific) and eFluor780-conjugated fixable viability dye (1:1.000; #65-0865-14, eBioscience, San Diego, CA) for 30 minutes at 4°C. Isotype control PE-conjugated mouse IgG (#12-4714-82, Thermo Fisher Scientific) was used. Data were recorded on a fluorescence-activated cell sorter (FACSCanto II, BD Biosciences, San Jose, CA) and analyzed using FlowJo version 10 software (BD Biosciences).

Zeiss ZEN software and the ImageJ image processing program (National Institutes of Health, Bethesda, MD) were used for image analysis.²²  QuPath version 0.1.2 software (GitHub; Queen’s University Belfast, Belfast, United Kingdom) was used for quantification of cryosections.²³  Power analyses to provide cohort sizes were simulated with R software version 3.4.0 (https://www.r-project.org/) in advance. Unpaired Student t test or Mann-Whitney U test was used for comparing 2 groups, Spearman’s rank correlation coefficient was calculated for correlations, and survival curves were tested with the Kaplan-Meier estimator. Either a χ² test or Fisher’s test was used to compare the frequencies of reaching a next step in the metastatic cascade. SPSS 24.0 software (SPSS, Chicago, IL) was used for statistical analysis, and statistical significance was assumed when P < .05 was reached.

The formation of blood clots at the site of arrested TCs in the brain was first discovered using a correlative light and EM methodology.¹⁵  Here, Jimt1 breast cancer cells were injected intracardially, and intravascular brain arrested TCs were targeted by using multimodal correlative microscopy. EM at 3 to 7 days p.i.i. allowed the visualization of TCs and their microenvironment, including blood components (Figure 1A). Strikingly, an accumulation of aggregated platelets and clots in the vessel lumen was frequently observed at the sites of TC arrest (Figure 1B).

To study the functional impact of blood coagulation factors on the single steps of BM formation in mice, we further developed an in vivo MPLSM methodology through a chronic cranial window, which was originally established in our laboratory.^5,8,10  In general, this technology makes it possible to study the entire sequence of steps that a circulating TC must master to eventually establish a brain macrometastasis: the brain metastatic cascade (Figure 1C). Two different human melanoma cell lines (A2058 and H1) and 1 human breast cancer cell line (Jimt1) were injected intracardially to generate metastases, and MPLSM imaging was initiated within a few minutes after heart injection and repeated over the course of 4 weeks to monitor the early and later steps of brain colonization (Figure 1D). Intravascular cell arrest within the brain occurred exclusively in brain capillaries with a diameter of 5 to 12 µm (Figure 1E) with a cell-vessel diameter ratio of 1, as reported previously.⁵  To label, quantify, and further characterize the occurrence of blood clots in vivo, Rhodamine 6G^11,12  (to label intravascular blood clots via their platelet content) was injected before every imaging session. Furthermore, FITC-labeled polyclonal anti-VWF antibodies were used to visualize VWF fibers.

At day 1 p.i.i., 99% of Jimt1 cells, 96% of A2058 cells, and 97% of H1 cells showed any level of clot formation at the very site of their arrest in brain microvessels. The high resolution of in vivo MPLSM allowed us to further specify clot aggregation patterns and determine different clot volumes (Figure 2A-B). An extensive clot was defined by a size that was 50% to 100% of the size of the intravascular arrested TC and was detectable in 37% of Jimt1 breast cancer cells and 8% of A2058 and 23% of H1 melanoma cells. In contrast, a minor clot was defined by a size that was <50% of the size of the intravascular arrested TC, whereas the size of an excessive clot exceeded the size of the arrested TC (>100%; Figure 2B). Clot formation was visible immediately after the injection of TCs and usually disappeared after 3 to 7 days when cells extravasated. Remarkably, in vivo tracking of the long-term fate of arrested TCs (Figure 2C) revealed that successful extravasation was in general indisputably associated with clot formation.

Arrested cancer cells with excessive clot formation did not successfully extravasate, possibly reflecting the presence of nonvital cancer cells in this situation, but cancer cells with extensive (ie, relevant) clot formation on day 1 extravasated more efficiently than cells with no extensive clot formation, which reached significance for the Jimt1 (P = .013) and the A2058 (P = .021) but not for the H1 model (P = .557) (Figure 2D). Importantly, only those TCs with any clot association at day 1 p.i.i. were able to grow out to a macrometastasis thereafter (Figure 2E).

Formation of a relevant (extensive) clot was detected only at TC arrest sites but nowhere else in the investigated brains and not in control animals intracardially injected with phosphate-buffered saline, suggesting that relevant clots are not primarily triggered by the heart injection method itself but are exclusively induced by the TCs at their arrest sites. These data indicate that clot formation is an indispensable prerequisite for TC extravasation in the brain.

Several receptors and their ligands have been proposed to induce TC-mediated platelet aggregation.²⁴  To investigate the ability of TCs to induce platelet aggregation, we used LTA. However, when washed platelets were incubated with Jimt1, A2058, and H1 cells, no aggregation of platelets was detected (Figure 3A). In line with that observation, FACS analysis for receptors of direct TC-platelet interactions showed no expression of GPIb or GPIIb/GPIIIa (supplemental Figure 2A). These data suggest that tumor-initiated clotting was not mediated by direct TC-platelet interactions, but instead may involve plasmatic components. TF is the primary initiator of the coagulation cascade. As a transmembrane glycoprotein, TF is the receptor for the coagulation protease factor VII (FVII), and the formation of the TF-FVII complex triggers thrombin generation and clot formation.²⁵  TF is constitutively expressed in the subendothelial spaces and adventitia and is highly expressed in many types of cancer.²⁶  FACS analysis revealed that Jimt1 cells express TF, but the melanoma A2058 and H1 cells do not (Figure 3B). Reduced plasma coagulation time demonstrated the thrombotic activity of all TCs, in which the prothrombotic activity of the breast cancer cell line Jimt1 was highest (Figure 3C). In addition, the presence of coagulation factors by coincubation with human plasma resulted in a marked increase in thrombin generation in all TC lines (Figure 3D), suggesting a thrombin-dependent mechanism for TC-triggered blood clotting. Indeed, LTA confirmed that TC-derived thrombin promoted platelet aggregation (Figure 3E; supplemental Figure 2B). Incubation of TCs with tinzaparin (Figure 3E; supplemental Figure 2C) or hirudin (supplemental Figure 2D) abolished this effect. Therefore, we conclude that TC-generated thrombin is the main trigger for the TC-mediated clot formation.

For in vivo proof of our findings, we treated mice with antiplatelet therapy (APT: ASA plus clopidogrel [ASAC]) along with LMWH tinzaparin and VWF blocking by using a polyclonal anti-VWF antibody (Figure 4A). The efficacy of platelet inhibition induced by ASAC was confirmed by an increased tail bleeding time (supplemental Figure 2E). In line with this observation, therapeutic thrombin inhibition by the LMWH was demonstrated via decreased thrombin generation (supplemental Figure 2F).

MPLSM after 24 hours confirmed a significantly reduced clot formation for all 3 treatment groups (Figure 4B). As expected, untreated control mice frequently exhibited a prolonged TC arrest in brain microvessels for at least 24 hours (Figure 4C). As mentioned above, a prolonged retention of TCs within brain capillaries is a prerequisite for TC extravasation.⁵  All anticoagulatory treatments provoke a significant reduction of Jimt1 TCs, but not A2058 and H1 TCs, within brain microvessels (Figure 4D).

To study the long-term effects of APT and anticoagulation with LMWH or dabigatran on the brain metastatic cascade, we continued the corresponding treatments until day 10 p.i.i., and tracked the fate of those cancer cells that managed to perform an intravascular arrest until day 28 (Figure 5A). Here, no difference in the subsequent steps of the brain metastatic cascade was observed when compared with controls (Figure 5B). To investigate the overall impact of different antithrombotic strategies, brain macrometastasis formation was quantified 4 weeks after heart injection in the breast cancer model. After this time point, control mice would start to become symptomatic. Hematoxylin and eosin staining was used to identify metastases on slides of brain tissue (Figure 5C; false color delineations for quantification). Here, anticoagulation with tinzaparin resulted in a reduction in both metastasis count and metastasis size (Figure 5D). Next, we aimed to get a comprehensive picture of the anti-brain metastatic activities of additional antiplatelet and anticoagulant strategies. Because ASAC does not affect platelet GPIIb/GBIIIa-fibrinogen interactions, we further induced thrombocytopenia by treatment with anti-GPIb serum and blocked the interaction of platelets with fibrinogen and VWF with the clinically approved anti-GPIIb/GPIIIa antagonist tirofiban. Cranial MRI scans demonstrated that platelet inhibition had no impact on total BM load (Figure 5E-G). In contrast, the clinically approved thrombin inhibitor dabigatran strongly reduced overall BM formation (Figure 5E-G), confirming the pivotal role of TC-derived thrombin and the plasmatic coagulation system for brain colonization of cancer cells.

Because thrombin mediates the secretion of endothelial VWF (a multimeric molecule known to promote thrombosis and metastasis^21,27 ), in vivo VWF imaging was performed using a specific fluorescence-labeled anti-VWF antibody. With this technique, we were able to observe hemostatically active VWF fibers in the lumen of cerebral microvessels (Figure 6A-B). In contrast to platelet accumulation, the VWF signal was not as abundant, but it was more refined to specific microvascular areas adjacent to arrested TCs (supplemental Figure 3A). In line with that observation, immunohistochemical analyses confirmed VWF and fibrin positivity at the site of arrested TCs (Figure 6C; supplemental Figure 3B). Next, we revisited our data in breast cancer patients suffering from BMs. To this end, we analyzed tumor-free brain tissue adjacent to BMs. In human brain tissue, we could confirm VWF release and platelet binding next to human breast cancer brain metastases (Figure 6D). Because luminal VWF fibers mediate the recruitment and extravasation of neutrophils promoting inflammation and metastasis,^28,29  we next analyzed the contribution of monocytes and neutrophils in early steps of the metastatic cascade (supplemental Figure 3C). In contrast to lung metastasis, we observed no monocytes or neutrophils in BMs.

Administration of the anti-VWF fluorescent antibody in vivo demonstrated VWF positivity in 79% of all intravascularly arrested Jimt1 cells and 50% of A2058 and 31% of H1 melanoma cells (Figure 7A), which is consistent with the differential clot formation rates in these models. Compared with untreated controls, the subsequent metastatic outgrowth in the perivascular niche into a micrometastasis from day 3 to day 14 was significantly impaired in the Jimt1 breast cancer model (Figure 7B-C). Similar trends were detectable for the melanoma models, but statistical significance was not reached (Figure 7C; supplemental Figure 4A). To investigate whether anti-VWF treatment remains effective when it is started after TC extravasation, administration of anti-VWF antibody was initiated as late as day 10 after intracardial TC injection. Here, anti-VWF treatment no longer showed any significant effects on the fate of TCs in the perivascular niche (Figure 7C). To analyze leakage for VWF during initial steps of brain colonization by TCs, sodium fluorescein (molecular weight, 376.27 g/mol)⁸  and IgG extravasation were measured. We could not detect any increase in vascular permeability within the first 5 days after TC injection (supplemental Figure 4B-E). Together these data indicate that the therapeutic anti-VWF antibody solely displayed its activity within brain microvessels and had no access to extravasated TCs.

To gain deeper insights into the effects of VWF inhibition on total BM load, MRI and overall survival experiments were performed. Histologic analyses and computer-assisted quantifications demonstrated that VWF inhibition correlates with a significant reduction of both the total number and total size of macrometastases (Figure 7D-E; supplemental Figure 5A). Statistical significance was not reached for survival experiments, but 2 mice lived for more than 60 days, which has never been observed with control mice in this robust model, in which death from brain and systemic metastases always occurs around day 30 (supplemental Figure 5B-C). Therefore, we conclude that VWF inhibition is a promising approach for preventing the formation of BMs.

A growing body of evidence suggests that platelets and the coagulation system can be involved in the process of organ metastasis.^30,31  However, the mechanisms of the interplay between TCs and the cellular and plasmatic coagulation system is far from being clear, which includes the process of BM formation. Our newly developed intravital microscopy methodology allowed us to dissect which steps of the multistep process of brain metastasis are actually affected by clot formation and antithrombotic therapy in the living mouse: first and foremost, it is the TC arrest in brain capillaries. To the best of our knowledge, this methodology was used for the first time to convincingly demonstrate the importance of specific microvascular clot formation at arrested TCs for successful organ metastasis.

Here we provide evidence for a key role of thrombin for clot formation in the brain microvasculature as the starting point of brain metastasis. TF on the cell surface interacts with the plasmatic coagulation and promotes the generation of thrombin.³²  A high TF expression in the tumor tissue of cancer patients has been correlated with tumor progression, a worse prognosis, and thrombotic complications.³³  In line with previous studies,^34,35  we show a high TF expression of Jimt1 breast cancer cells. However, the absence of TF in the A2058 and H1 melanoma cell lines did not abolish thrombin generation, which suggests a direct activation of thrombin by these cells. Indeed, previous publications identified a tumor-derived cysteine protease, which promotes thrombin generation via activation of coagulation FX and is referred to as cancer procoagulant.³⁶  Thrombin then exerts its procoagulant activity via enzymatic conversion of fibrinogen to fibrin polymers,³⁷  platelet aggregation, and the secretion of endothelial VWF multimers.^34,35,38  By using in vivo MPLSM, we found that the initial TC arrest was associated with platelet recruitment and the deposition of fibrinogen and VWF. Furthermore, clot formation represented an essential step in TC extravasation and therefore in BM. We also demonstrated that antithrombin strategies with dabigatran or the LMWH tinzaparin can reduce TC arrest and brain metastatic load of specific tumors. Clinical studies using LMWH in cancer patients have variable results with regard to prolonged overall survival.^6,39-41  Reasons for this controversial outcome might be explained by inclusion of different tumor types, disease stages, and the applied LMWH. Indeed, differences in length and chemical composition define antimetastatic properties of heparins.^21,27  Tinzaparin exerts its activity via inhibition of thrombin and FX, and the binding of VEGF-A, both required for TC-mediated VWF secretion.²¹  In addition, genetic inhibition of VWF multimer degradation in a mouse with depletion of ADAMTS13 (an enzyme that cleaves and inactivates VWF) resulted in increased lung metastases after intravenous injection of TCs.²⁷  Here, tinzaparin treatment and VWF inhibition caused less TC arrest in the breast cancer model and the H1 melanoma model. Our results further suggest that VWF is also relevant for the process of micrometastasis growth in the typical vascular co-option mode along brain capillaries: anti-VWF treatment resulted in a reduced BM count and a smaller mean macrometastasis size in both melanoma and breast cancer models. Moreover, TC survival in the perivascular niche was reduced in the breast cancer model.

VWF has been involved in many biological processes, including cell proliferation, coagulation, and inflammation,⁴²  which might explain the effects of anti-VWF treatment observed here. One hypothesis was that VWF expression in cells of the vascular wall is important for supporting perivascular co-optive metastatic growth, possibly through integrins. Interaction of VWF with integrins is known,⁴³  and integrin beta-1 inhibition has been shown to reduce the capacity of co-optive TC growth in the perivascular niche in the brain.⁴⁴  Another explanation could be that VWF-induced clots prolong the survival of arrested TCs and promote the extravasation process, since it was demonstrated that luminal VWF fibers mediate the recruitment and extravasation of neutrophils and TCs.^27,28 

TC-platelet interactions fostering BMs could be excluded by the findings of this study. Simultaneous inhibition of platelet cyclooxygenase 1 (COX-1) and adenosine 5′-diphosphate–mediated platelet activation with ASAC^45,46  did not affect BMs. Previous studies demonstrated that this dual therapy with clopidogrel and aspirin even promotes mortality in a mouse model for breast cancer.⁴⁷  Because it was shown that integrin-mediated TC-platelet interactions promote metastasis,^48,49  platelet-depleted plasma and the GPIIb/GPIIIa antagonist tirofiban were applied. Again, we could not detect any effect on the formation of BMs upon those treatments. Our in vitro examinations showing that TCs are not able to mediate platelet aggregation via cell-cell contact provide a sound explanation for the observed lack of therapeutic efficacy. So far, no specific antiplatelet therapy has been included in standard tumor treatment.⁴⁸ 

Limitations of this study should be considered. One limitation is that although the in vivo MPLSM methodology allowed deep insights into the brain metastatic cascade, it did not allow investigations of extracranial metastases. Furthermore, the glass used to cover the cranial window allows the evaluation of only relatively small brain volumes up to a depth of 500 μm for intravital imaging studies.

All in all, we show here that microclot formation at the exact location of TC arrest in brain microvessels drives BM formation. From there, blocking the plasmatic coagulation system by using direct (dabigatran, LMWH) and indirect (LMWH) thrombin inhibitors attenuates BM formation in our mouse model (supplemental Table 1). No bleeding complications were detectable, which mirrors clinical observations that an anticoagulatory therapy in patients with BMs did not increase the risk of intracerebral bleedings.^50,51  Moreover, inhibition of VWF by using an anti-VWF antibody also attenuated BM formation. These results point to a pivotal role of thrombin and VWF for brain colonization by TCs. In contrast, various solely antiplatelet treatments were not effective for reducing overall BM formation in this study. However, these results do not exclude a relevant role of platelets to act in concert with the plasmatic coagulation system to push tumor spreading. The single-agent benefits of antithrombin treatments (by using direct inhibitors of FII, FX, and LMWH) and anti-VWF treatments shown here, together with their good tolerability, make these drugs prime candidates for future clinical BM prevention trials.

The authors thank Sabine Vidal-y-Si for excellent technical help.

This work was supported by grants from the German Research Council (RTG2099) (M.J.F., J.R.R., C.T., B.K., A.T.B., S.W.S., and F.W.), from the Deutsche Krebshilfe (German Cancer Aid), Priority Program “Translational Oncology,” #70112507, “Preventive strategies against brain metastases” (F.W., W.W., and A.S.B.), and from the Erich and Gertrud Roggenbuck-Stiftung (A.T.B. and J.R.R.).

Contribution: M.J.F., S.W.S., A.T.B., and F.W. wrote the manuscript, which was reviewed and edited by all authors; F.W., W.W., and S.W.S. designed the study and supervised the work; M.J.F., A.S.B., V.V., G.S., C.T., and J.K.G. performed in vivo experiments; M.J.F., J.R.R., C.T., M.A.K., K.G., B.K., F.T.M., M.O.B., M.B., and A.T.B. performed ex vivo experiments; M.F. and M.O.B. provided magnetic resonance imaging; and M.A.K and Y.S. performed electron microscopy experiments.

Conflict-of-interest disclosure: F.W. has received research funding from Roche, Genentech, Boehringer Ingelheim, and Divide & Conquer. G.S. is an employee of Carl Zeiss Microscopy. The remaining authors declare no competing financial interests.

Correspondence: Frank Winkler, Neurology Clinic and National Center for Tumor Diseases, University Hospital Heidelberg, Im Neuenheimer Feld 400, 69120 Heidelberg, Germany; e-mail: frank.winkler@med.uni-heidelberg.de.