Exploring the dual role of extracellular vesicles in coagulation and immune modulation in glioblastoma
Annabell Wolff, Grit Waitz, Philipp Kaps, Sonja Oehmcke-Hecht, Wendy Bergmann-Ewert, Björn Schneider, Katharina Richter, Charlotte Wagner, Ann-Sophie Becker, Anett Seifert, Daniel Dubinski, Thomas M. Freiman, Thomas Thiele, Sascha Troschke-Meurer, Claudia Maletzki
TL;DR
This study explores how extracellular vesicles from glioblastoma tumors contribute to blood clotting and immune changes, offering new insights into potential biomarkers and treatments.
Contribution
The paper introduces a translational workflow to investigate EV-mediated coagulation and immune modulation in glioblastoma.
Findings
GBM tumors show high TF and PDPN expression with low TFPI, indicating a procoagulant phenotype.
GBM-derived EVs modulate microglial behavior and immune polarization, affecting the tumor microenvironment.
EVs from GBM exhibit procoagulant activity proportional to TF expression and immune-modulating effects.
Abstract
Glioblastoma (GBM) is often complicated by venous thromboembolism (VTE), primarily driven by tissue factor (TF, F3) and podoplanin (PDPN). These factors promote local hypercoagulation and microthrombosis, thereby contributing to tumor progression by enhancing migration, invasion, and inflammation. Both TF and PDPN can be released via extracellular vesicles (EVs), which carry procoagulant and immunomodulatory cargo. We developed a translational workflow combining biobanked tumor samples, clinical data, ex vivo GBM cultures, and coagulation assays to investigate mechanisms of hypercoagulation. Intraoperative blood coagulation was profiled using ClotPro®. Gene expression of coagulation-related markers was analyzed in tumor tissues and cell lines, complemented by RNAseq-based profiling of coagulation–inflammation links. Functional coagulation assays included clotting time, platelet…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6- —Universitätsmedizin Rostock (8980)
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsExtracellular vesicles in disease · Neutrophil, Myeloperoxidase and Oxidative Mechanisms · Protease and Inhibitor Mechanisms
Introduction
Glioblastoma (GBM) is a highly aggressive and heterogeneous malignancy of the central nervous system. Despite decades of research and the discovery of numerous driver mutations, effective targeted therapies capable of eradicating GBM remain elusive. While actionable genetic alterations are rare, there is a well-established connection between tumor progression and cancer-associated thrombosis, as reviewed in^1^. Importantly, thrombotic complications in GBM patients, including venous thromboembolism (VTE) and arterial thromboembolism, are associated with worse clinical outcomes, including reduced overall survival, increased morbidity, and higher mortality^2,3^. The development of local microthrombosis within the tumor capillary network and tumor microenvironment (TME), as well as peripheral venous thromboembolism, has been linked to oncogenic driver mutations such as KRAS,* MET*,* TP53*, and EGFRVIII. This pathological interplay contributes to the regulation of coagulation-related gene expression, collectively referred to as the coagulome, which underlies the prothrombotic condition observed in GBM patients^4^. Among the key mediators, tissue factor (TF) plays a central role^5,6^. TF is a critical receptor for coagulation factor (F) VII, and, particularly in GBM, its expression is frequently linked to pro-tumorigenic effects mediated by other activated clotting factors such as FVIIa, FXa, and thrombin^5^. Additionally, extracellular vesicles (EVs, ø 200 nm) derived from GBM cell-free supernatants have been shown to induce both platelet aggregation and coagulation^7^. Elevated levels of glial-derived and/or TF-bearing EVs even correlate with disease progression^8^. Another important molecular determinant of cancer-associated thrombosis in GBM is podoplanin (PDPN), which also serves as EVs cargo^9,10^. PDPN enhances tumor cell clonogenicity, epithelial-mesenchymal transition, migration, invasion, and inflammation^11^. Acting synergistically with TF, PDPN contributes to the formation of local microthrombosis within the TME.
Preclinical tumor models provide valuable tools for investigating the interplay between biological and genetic factors in cancer-related coagulation disorders. In these models, the expression of coagulation factors is maintained under the regulation of oncogenic signaling pathways. Notably, the clinically more aggressive mesenchymal GBM subtype has been shown to express a distinct set of coagulation-related genes compared to the less aggressive proneural subtype, which is associated with a lower risk of thrombosis^4^. In addition to subtype-specific differences, intratumoral heterogeneity, along with the phenotypic and clonal evolution of GBM cell populations, may significantly influence coagulopathy. Conversely, procoagulant events within the TME can affect the biological behavior and molecular trajectory of tumor cells, highlighting a dynamic and reciprocal relationship between coagulation and tumor progression.
In this study, we developed a workflow combining clinically annotated patient samples, ex vivo tumor cell culture, and analysis of local and systemic coagulation. Given that timely and precise hemostasis monitoring can help to prevent potentially life-threatening complications, we conducted a biomarker-driven evaluation of individualized coagulation profiles. Our data show that GBM cells isolated from different patients may use distinct procoagulant strategies, activating both plasmatic and cellular pathways to drive local hypercoagulability. These findings could contribute to the development of personalized strategies for thrombosis prevention and cancer-specific risk assessment.
Results
Patient cohort and global coagulation profiles
This study included 15 patients with histologically confirmed WHO Grade 4 GBM, with MGMT promoter methylation present in 40% of cases (Table 1). The median age was 64 ± 13.7 years (range 40–81), 42% were female. Intraoperatively obtained blood showed platelet counts were within the normal range, whereas leukocyte counts were slightly elevated. During treatment, 20% of patients experienced thrombotic events, and approximately half of the cohort received anticoagulant therapy (Table 1). The median overall survival was 19.9 months.
Table 1. Clinical patients data, including basic molecular characterization of GBM samples.Patient IDGender/ageOS[month]MGMT statusBlood parametersPlatelets[10^9^/l]^#^leukocytes[10^9^/l]^$^Thromboembolic Events (TE) during treatmentAnticoagulant therapyGBM03m/814.4Unmethylated13715.1Deep vein thrombosisASS, LMWH*GBM06m/713.9Methylated2373.4Not reportedUnknownGBM08f/4713.4Unmethylated27619.2Not reportedUnknownGBM09m/790.9Unmethylated21620.7Deep vein thrombosisLMWHGBM11m/7334.9Methylated1614.9Not reportedUnknownGBM12f/691.9Unmethylated1656.7Not reportedLMWHGBM14m/6327.4Unmethylated2518.3Cerebral venous sinus thrombosisUnknownGBM15m/402.5Unmethylated2029.6Not reportedUnknownGBM17m/4915.9Methylated19916.9Not reportedASSGBM19m/5015.9Unmethylated18212.7Not reportedASSGBM21f/722.3Unmethylated1327.8Not reportedRivaroxabanGBM22m/791.8Unmethylated2007.4Not reportedLMWHGBM23m/731.0Methylated2048.7Not reportedLMWHGBM24m/438.7Unmethylated20015.5Not reportedUnknownGBM26f/728.2Methylated2649.6Not reportedUnknown Σ (% or Median (Range) f/m: 41.9 : 58.1
age: 64.0 ± 13.7
19.9
Methylated: unmethylated
39.1 : 60.9
201 (144)
10 (18)
20
53.3 MGMT O6-methylguanine-DNA methyl-transferase, ASS Acetylsalicylic acid, LMWH low molecular weight heparin; * Enoxaparin; ^#^ Reference range: 150–450 × 10⁹/L; ^$^ Reference range: 4.0–9.0 × 10⁹/L.
Coagulation profiling by the ClotPro^®^ viscoelastic assay and standard laboratory coagulation assays revealed mildly prolonged coagulation times in the EX- and FIB-tests, while IN-test values remained within reference ranges (Fig. 1A). Maximum clot firmness (MCF) also remained within the reference range for the FIB-test but slightly reduced in EX- and IN-tests. Standard laboratory parameters (Quick, INR, aPTT) were unaltered (Fig. 1B). EVs were isolated from patient plasma to evaluate their procoagulant activity. Clotting times (CT) in normal and FXI-deficient plasma showed that GBM-derived EVs significantly shortened CT versus healthy donor EVs (p < 0.0001), with preserved activity in FXI-deficient plasma, indicating independence from the intrinsic pathway (Fig. 1C). Patient-derived GBM cells mirrored this effect, confirming tumor-driven hypercoagulability (p < 0.0001, Fig. 1D). Together, these findings suggest that plasma, EVs, and tumor cells from GBM patients contribute to a hypercoagulable state, likely through activation of the extrinsic pathway activation, underscoring the role of tumor-derived EVs in GBM-associated coagulopathy.
Fig. 1. Coagulation-associated gene expression and peripheral coagulation status in GBMpatient. (A) Peripheral coagulation parameters of GBM patients were assessed by viscoelastic ClotPro^®^ Hemostasis. n = 13–17, mean ± SD. (B) Standard coagulation assays (Quick, INR, and aPTT) were measured. Dotted lines represent the reference range. n = 44, mean ± SD. (C, D) GBM patient plasma-derived EVs clotting time (EV-CT, C) with normal and FXI-deficient plasma and clotting time of GBM cells (GBM-CT, D) with only normal plasma were measured, indicating procoagulant activity. Each dot represents one biological replicate. n = 3–20, mean ± SD, Unpaired t-test, **** p < 0.0001 vs. ctrl. (E) Representative H&E-stained sections of GBM tumors from four patients (GBM03, GBM06, GBM15, GBM24). (F,G) Bubble plots depicting log2 fold changes and significance levels (represented by bubble size) of selected genes related to coagulation and inflammation in primary tumor samples (F) and patient-derived cell lines (G). Red bubbles indicate upregulation, while blue bubbles indicate downregulation. n = 7 (F) and n = 12(G), mean ± SD. (H) Bar graph comparing mean log2 fold changes between primary tumors (red) and patient-derived cell lines (blue) for the indicated genes, showing slight differences in expression patterns across the two sample types. (I) Gene expression of F3,* TFPI1*, and PDPN in primary tumors (top) and patient-derived cell lines (bottom) were analyzed by qRT-PCR n = 13–15 (top), n = 4 (bottom), mean ± SD, One-way ANOVA (Tukey’s multiple comparison test), n.s. = not significant; **p < 0.01, ***p < 0.001. (J) Correlation matrix of F3, TFPI1, and PDPN in GBM samples. Pearson coefficients are shown; color indicates strength and direction: blue = positive, red = negative. n = 6 (K) Representative immunofluorescence images are shown. Scale bars = 200 (10x)/50 (20x)/20 (40x) µm and the data were quantified using ImageJ. n = 6 mean ± SD.
Table 2. Clinical and molecular characteristics of the patient-derived cell lines.Patient IDGender/ageOS [month]MGMT statusMolecular subtype/ IDH-StatusCDKN2ACDK4EZH2MutationsGBM03m/814.4UnmethylatedIDH-wildtype/wtdiploidamplifiedamplified/overexpressedTP53 (R273H (VAF: 48.6 %); R248Q (VAF: 50.3 %); P72R (VAF: 99.6 %); KDR Q472H (VAF: 35.0 %)GBM06m/713.9MethylatedIDH-wildtype/wt diploidhighly amplifiedamplified/overexpressedTP53 G244A (VAF: 99.9 %); P72R (VAF: 100.0 %); PIK3CA I391M (VAF: 30.6 %)GBM14m/6327.4unmethylatedIDH-wildtype/wttriploidamplifiednot amplified/overexpressedTP53 V173L (VAF: 58.6 %); P72R (VAF: 58.3 %); KDR Q472H (VAF: 54.0 %); PTEN D252Y (VAF: 100.0 %)GBM15m/402.5unmethylatedIDH-wildtype/wtpaternal deletionhighly amplifiedamplified/overexpressedTP53 P72R (VAF: 99.9 %)
Expression of coagulation-associated genes is preserved in patient-derived cell lines
The primary tumors displayed the characteristic heterogeneous morphology (Fig. 1E), including microvascular proliferation, necrosis, hemorrhage, and hyperemic intratumoral vessels. Clear microthrombosis could not be detected using conventional light microscopy.
To further assess the local coagulation profile, tumor tissues from primary GBM cases and their matched low-passage (< P10) patient-derived cell lines were analyzed in comparison to normal tissue (Fig. 1F,G). Gene expression profiling revealed modest but consistent differences in coagulation- and inflammation-associated genes between primary tumors and patient-derived cell lines. The comparative analysis of mean log₂ fold changes (Fig. 1H) indicated that expression of coagulation-related genes is generally preserved and, in some cases, enhanced in patient-derived cell lines. Quantitative RT-PCR analysis of the three main coagulation-associated genes (F3,* TFPI1*, and PDPN) in 15 GBM samples confirmed the transcriptomic data. All tumors displayed higher expression levels of F3 and PDPN compared to TFPI1, which was largely undetectable (Fig. 1I, top). Patient-derived GBM cell lines (GBM03, GBM06, GBM14, and GBM15) displayed a similar expression profile, suggesting that F3 and PDPN expression was stably maintained in vitro. Notably, TFPI1 expression was detected in one cell line (GBM14) but was absent in the corresponding primary tumor, potentially reflecting selective clonal expansion or in vitro adaptation (Fig. 1I, bottom). Correlation analysis demonstrated an inverse association between F3 and PDPN expression, suggesting that these factors may be subject to distinct regulatory mechanisms (Fig. 1J). Quantitative analysis of immunofluorescence images of primary GBM specimens revealed heterogeneous and partially overlapping expression of TF and PDPN, with concurrent expression confirmed across independent samples (Fig. 1K; Supplementary Fig. S1). Together, these findings indicate that key coagulation-associated factors are preserved in GBM-derived cell models, supporting their suitability for mechanistic studies. However, the limited number of samples should be considered when interpreting these data.
TF and PDPN abundance are heterogeneous within patient-derived cell lines
TF and PDPN, key regulators of local coagulation, were analyzed in patient-derived GBM cell lines via flow cytometry and immunofluorescence (Fig. 2A–D). Flow cytometry revealed heterogeneity in TF⁺ and PDPN⁺ populations, with GBM14 showing the highest expression (75% TF⁺, 91% PDPN⁺) and the largest TF⁺PDPN⁺ subpopulation (Fig. 2A). Immunofluorescence confirmed these patterns, though PDPN levels in GBM03 appeared lower compared to flow cytometry (Fig. 2B,C). The clinical and molecular characteristics of the GBM cell lines used in this study are summarized in Table 2.
Fig. 2. Heterogeneous expression of TF and PDPN in patient-derived GBM cell lines. (A–C) GBM cells (GBM03, GBM06, GBM14, and GBM15) were stained for TF and PDPN, then analyzed using spectral flow cytometry (A) and immunofluorescence (B,C). Representative immunofluorescence images are shown. Scale bar = 20 μm (B), and the data were evaluated using ImageJ (C). n = 7–9 (A), n = 3 biological replicates with n = 3–4 technical replicates (C), mean ± SD, One-way ANOVA (Tukey’s multiple comparison test), *** p < 0.001, **** p < 0.0001. (D) Clotting times of GBM cells were measured using a coagulometer with normal, FVII-deficient, and FXII-deficient plasma. n = 3, mean ± SD, Two-way ANOVA (Tukey’s multiple comparison test), * p < 0.05, *** p < 0.001, **** p < 0.0001. (E) GBM cells were treated with Clexane, after which TF and PDPN were detected via immunofluorescence. Representative images were shown. The evaluation was performed using ImageJ. n = 3, mean ± SD, Two-way ANOVA (Tukey’s multiple comparison test), ** p < 0.01, *** p < 0.001, **** p < 0.0001. Scale bars = 20 μm (F,G) Platelets and GBM cells were incubated with or without plasma, or with fibrinogen, and then examined by flow cytometry for PS exposure (F) or platelet activation (G). n = 3, mean ± SD, Two-way ANOVA (Tukey’s multiple comparison test), * p < 0.05, *** p < 0.001, **** p < 0.0001 vs. positive ctrl. (H) Platelet aggregation via GBM cells was performed using an aggregometer. Shown are the onset of aggregation (Lag Time) and the strength of platelet aggregation (aggregation in %). n = 3, mean ± SD, One-way ANOVA (Tukey’s multiple comparison test), *p < 0.05, ** p < 0.01.
To assess functional relevance, GBM cells were incubated with normal, FVII-deficient, and FXII-deficient plasma. GBM14 exhibited the shortest CT (44 s), consistent with high TF expression. FVII deficiency (extrinsic pathway blockade) markedly prolonged CT (up to 777 s), while FXII deficiency (intrinsic pathway blockade) caused only moderate delays (90–294 s), suggesting a TF-driven, extrinsic pathway–dependent hypercoagulability (Fig. 2D). Other GBM lines exhibited intermediate CTs reflecting their TF/PDPN expression levels. These results suggest that TF abundance is a major determinant of plasma clotting in GBM cells.
To assess the effect of standard anticoagulant therapy, GBM03 and GBM14 cells were treated with low molecular weight heparin (LMWH), which did not significantly alter TF or PDPN expression in GBM03, whereas in GBM14 it significantly reduced TF (p < 0.001) and PDPN (p < 0.01), as shown by representative immunofluorescence images and quantification in Fig. 2E. LMWH-induced reductions in TF and PDPN levels in GBM14 were observed. It remains to be determined whether these effects might involve alterations in membrane organization or lipid raft composition, which could be investigated in future studies.
PDPN+ tumor cells do not significantly trigger platelet activation and aggregation
The presence of PDPN⁺ GBM cells, some co-expressing TF, suggests that GBM may exploit multiple procoagulant mechanisms. PDPN interacts with CLEC-2 to induce platelet activation and aggregation, contributing to cancer-associated thrombosis. To assess this, platelet activation and aggregation were measured using flow cytometry and aggregometry (Fig. 2F–H).
Washed platelets were incubated with GBM cells under three conditions: (1) with 2% plasma, (2) with 2.25 mg/ml fibrinogen, and (3) without plasma. All GBM cell lines induced PS exposure on the platelets under all conditions, with no significant differences (Fig. 2F). The addition of fibrinogen reduced phosphatidylserine (PS) exposure on platelets, whereas plasma had no major effect.
Similarly, the platelet activation marker PAC-1 was upregulated following incubation with the different GBM cell lines (Fig. 2G), although no inter-line variation was observed. CD62P expression remained low overall. Fibrinogen reduced PAC-1 expression, whereas plasma enhanced it, suggesting that plasma factors facilitate platelet responsiveness. Platelet aggregation assays using TF^low^PDPN^high^ GBM03 and TF^high^PDPN^high^ GBM14 with platelet-rich plasma showed that both lines induced aggregation with comparable kinetics and amplitude (Fig. 2H). However, these responses were weaker than those induced by TRAP, suggesting that PDPN-mediated platelet activation contributes modestly to procoagulant activity (Fig. 2H).
Sorting of procoagulant phenotypes reveals high plasticity of patient-derived GBM14 cells
To determine whether TF and PDPN act independently or synergistically in driving hypercoagulability, GBM14 cells, selected for their high TF and PDPN expression, were flow-sorted into TF^+^, PDPN^+^, and double-positive (TF^+^PDPN^+^) subpopulations (Fig. 3A). We analyzed phenotypic plasticity over 21 days, CT values, and ROS involvement (Fig. 3B–E). Longitudinal analysis revealed marked phenotypic plasticity among the sorted populations (Fig. 3B). TF⁺ cells upregulated PDPN by day 7 (Fig. 3B), while PDPN⁺ cells reciprocally upregulated TF⁺, acquiring a double-positive phenotype. The TF⁺PDPN⁺ population initially shifted transiently toward a TF-dominant profile by day 7 but reverted by day 21. Ultimately, all subpopulations converged toward a phenotype resembling the pre-sorting state, underscoring the dynamic plasticity of GBM cells. The TF⁺ fraction exhibited CT values comparable to parental cells (58 s vs. 44 s), whereas all PDPN⁺ fractions showed prolonged CT. Single-positive PDPN⁺ cells demonstrated a doubling of the parental CT, and TF⁺PDPN⁺ cells also displayed significantly shortened CT, reaffirming TF’s procoagulant role (Fig. 3C). Notably, the TF⁻PDPN⁻ population exhibited clotting times comparable to the PDPN⁺ single-positive cells, indicating that the absence of TF does not further reduce coagulation and that PDPN expression alone is sufficient to modulate clotting behavior in GBM cells. To explore potential functional consequences of these shifts, we examined ROS and mitochondrial ROS (MitoSOX) levels by immunofluorescence (Fig. 3D) and fluorescence spectroscopy (Fig. 3E). ROS and mitochondrial ROS (MitoSOX) analyses revealed initially higher ROS in TF⁺ cells (p < 0.05 vs. PDPN⁺), which normalized over time (Fig. 3D,E). Moreover, ROS levels in the TF⁻PDPN⁻ cells were markedly lower (Fig. 3D,E). Collectively, these data highlight GBM cell plasticity and underscore the role of TF in mediating GBM-associated hypercoagulability.
Fig. 3. Sorting of procoagulant GBM phenotypes reveals high plasticity of patient-derived GBM cells. (A) Flow cytometry gating strategy used to isolate viable single GBM cells and distinguish TF⁺, PDPN⁺, and double-positive (TF⁺PDPN⁺) subpopulations based on marker expression. (B) Quantification of TF⁺, PDPN⁺, and TF⁺PDPN⁺ cells over 21 days post-sorting revealed dynamic changes and phenotypic plasticity, with all populations tending to repopulate mixed phenotypes over time n = 1–3, mean ± SD. (C) Clotting times (CT) were measured for basal cells and sorted populations. n = 3, mean ± SD, One-way ANOVA (Tukey’s multiple comparison test), * p < 0.05, *** p < 0.001, **** p < 0.0001. (D) Measurement of reactive oxygen species (ROS) via fluorescence intensity (ROS Brite™) over 21 days. n = 2, mean ± SD, Two-way ANOVA (Tukey’s multiple comparison test), * p < 0.05. (E) Representative fluorescence microscopy images showing mitochondrial ROS production (MitoSOX red) and live cell staining (Calcein AM) in sorted populations. Scale bars = 20 μm.
Tumor-derived EVs retain the procoagulant properties of their cells of origin
EVs inherit functional state from their origin cells and can modulate coagulation both locally and systemically. EVs from GBM cell lines and sorted GBM14 subpopulations were analyzed for size, concentration, and procoagulant activity (Fig. 4A–E). While EV size and yield were comparable across cell lines, their CT values mirrored the procoagulant potential of parent cells: GBM14-derived EVs showed the shortest CT, whereas GBM03-derived EVs had significantly longer CT (Fig. 4A,B; see also Fig. 2D). EVs exerted minimal effects on platelet activation, PS exposure, or aggregation. A modest increase in Annexin V⁺ platelets was observed in the presence of plasma, likely reflecting plasma-derived factors rather than EV-mediated activation (Fig. 4C,D). These findings indicate that under the tested conditions, isolated EVs alone did not directly trigger platelet activation or aggregation. Temporal analysis of EVs from GBM14 subpopulations (days 5–21) revealed stable concentrations but variable CTs (Fig. 4E). EVs from PDPN⁺ cells showed prolonged CT over time (p < 0.01), correlating with decreased TF expression in their cells of origin (Fig. 4F). By day 21, CTs converged toward parental levels, reflecting the reversible and adaptive phenotype of GBM cells and their vesicular output.
Fig. 4EVs retain the procoagulant properties of their cells of origin. (A) Nanoparticle tracking analysis (NTA) showing average size and concentration of EVs isolated from four patient-derived GBM cell lines (GBM03, GBM06, GBM14, GBM15). n = 3, mean ± SD. (B) Clotting time assay reveals significantly reduced clotting times for EVs derived from GBM03, GBM06, and GBM14 compared to GBM15, indicating differential procoagulant activity. n = 3, mean ± SD, One-way ANOVA (Tukey’s multiple comparison test), * p < 0.05, ** p < 0.01 (C) Flow cytometric analysis of platelet activation after EVs exposure. EVs were tested in the presence of plasma, fibrinogen, or in their absence. n = 3, mean ± SD, Two-way ANOVA (Tukey’s multiple comparison test), **** p < 0.0001 vs. positive ctrl. (D) Impact of procoagulant patient-derived GBM cell line-derived EVs on platelet aggregation. Lag time defines the onset of aggregation and aggregation strength. Lag times exceeding 20 min were represented as 1200 s. n = 2–3, mean ± SD. (E) Longitudinal analysis of EV particle counts and clotting times after sorting of TF⁺, PDPN⁺, and TF⁺PDPN⁺ subpopulations over 21 days. n = 3, mean ± SD, Two-way ANOVA (Tukey’s multiple comparison test), * p < 0.05, ** p < 0.01 vs. parental (F) ELISA-based quantification of TF levels in EVs. n = 2, mean ± SD.
Tumor-derived EVs reduce migration, induce senescence, and induce a M1-like phenotype in microglia
To investigate the interaction between GBM-derived EVs and brain-resident microglia, we first assessed EV uptake and its impact on microglial migration (Fig. 5A–C). GBM NIR-derived EVs were efficiently internalized by microglia, with the highest uptake observed for EVs from the TF^high^PDPN^high^ GBM14 cell line (6 h: p < 0.0001 vs. GBM06 NIR-derived EVs; p < 0.01 vs. GBM15 NIR-derived EVs, Fig. 5A,B). Notably, exposure to these EVs significantly reduces microglial migration (Fig. 5C), suggesting EV-mediated microglial retention within the TME.
Fig. 5. Tumor-derived EVs reduce migration, induce senescence, and a M1-like phenotype in microglia. (A) Representative fluorescence microscopy images showing time-dependent uptake of EVs derived from different GBM cell lines by microglia. Scale bar = 20 μm (B) Quantitative analysis of EVs uptake confirms time- and cell line–dependent differences in internalization. n = 3, mean ± SD, Two-way ANOVA (Tukey’s multiple comparison test), ** p < 0.01, **** p < 0.0001. (C) Transwell migration assay reveals significantly reduced migratory capacity in microglia following EVs exposure. n = 3, mean ± SD, One-way ANOVA (Tukey’s multiple comparison test), ** p < 0.01. (D,E) Immunofluorescence staining and quantification show increased expression of senescence-associated markers p21, p16, and p53 in EVs-exposed microglia compared to untreated controls. Scale bar = 20 μm (F,G) Immunofluorescence and corresponding quantification indicate no significant changes in connexin43 or the proliferation marker Ki67, suggesting that EVs-induced senescence is not accompanied by altered proliferation or gap junctional communication. Scale bar = 20 μm (H,I) Immunofluorescence and corresponding quantification for iNOS demonstrate increased expression in microglia treated with GBM-derived EVs, indicating induction of an M1-like pro-inflammatory phenotype. LPS-treated microglia serve as a positive control. n = 3, mean ± SD, One-way ANOVA (Tukey’s multiple comparison test), * p < 0.05, *** p < 0.001. Scale bar = 20 μm.
Further analyses assessed microglial proliferation, senescence, and polarization using immunofluorescence (Fig. 5D–I). EVs from TF^low^PDPN^high^ GBM03 and TF^high^PDPN^high^ GBM14 cells modestly increased Ki67⁺ microglia (~ 30%; Fig. 5D,E), without altering connexin 43 (Fig. 5F–G).
TF^low^PDPN^high^ GBM03- and TF^high^PDPN^high^ GBM14-derived EVs induced a modest M1-like phenotype in microglia, marked by elevated iNOS expression (LPS vs. GBM14_EVs, p < 0.05; GBM03/14_EVs vs. GBM06_EVs, p < 0.05 and p < 0.001; Fig. 5H,I). In contrast, TF^low^PDPN^low^ GBM06- and TF^low^PDPN^low^ GBM15-derived EVs, as well as EVs from the double-negative (TF⁻PDPN⁻) GBM14 population, induced senescence markers and comparable or reduced iNOS expression. Collectively, these results may suggest that GBM-derived EVs may modulate microglial activation and polarization in a manner influenced by TF and PDPN expression, indicating that these factors likely contribute to proinflammatory M1 responses and TME remodeling.
EVs from TF+ GBM cells influence leukocyte activation, TF expression, ROS, and NETosis, while EVs from PDPN+ GBM cells promote senescence
Given their presence in circulation, the immunomodulatory effects of GBM-derived EVs were assessed, focusing on leukocyte and neutrophil activation - focusing on TF expression, ROS production, NET formation (MPO, citH3), and senescence induction (Ki-67, p16, p21, p53). EVs from flow-sorted GBM14 subpopulations (TF⁺, PDPN⁺, TF⁺PDPN⁺, and TF^−^PDPN^−^) were co-cultured with leukocytes or neutrophils from healthy donors (Fig. 6).
Fig. 6. Tumor-derived EVs carrying TF and PDPN differentially modulate leukocyte and neutrophil functions. (A) Representative immunofluorescence images show pronounced leukocyte activation following exposure to tumor-derived EVs. Scale bar = 20 μm (B) Quantification of MPO, TF, and ROS expression in leukocytes exposed to TF⁺ and TF^+^PDPN^+^ EVs. n = 4, mean ± SD, One-way ANOVA (Tukey’s multiple comparison test), * p < 0.05. (C) Representative immunofluorescence images of neutrophil activation through tumor-derived EVs. Scale bar = 20 μm (D) EVs promote neutrophil extracellular trap (NET) formation, demonstrated by immunofluorescence staining of citH3. Scale bar = 20 μm (E) Quantification of Neutrophil activation marker MPO and citH3, as well as TF and ROS. n = 3, mean ± SD. (F,G) PDPN⁺ EVs induce neutrophil senescence, as shown by elevated expression of senescence-associated markers in immunofluorescence images (F) and quantitative analysis (G) n = 3, mean ± SD. Scale bar = 20 μm (H,I) Immunofluorescence (H) and corresponding quantification (I) of Connexin43 and the proliferation marker Ki67. n = 3, mean ± SD. Scale bar = 20 μm.
EVs from TF^+^ GBM cells modulate MPO release in leukocytes, a response not observed with PDPN⁺ EVs. Both EVs from TF^+^ and TF⁺PDPN⁺ GBM cells also elevated leukocyte TF expression and ROS levels (Fig. 6A,B). In neutrophils, TF expression remained largely unchanged, although EVs from PDPN^+^ GBM cells slightly increased TF and ROS, whereas EVs from TF⁻PDPN⁻ GBM cells induced no response in leukocytes or neutrophils. (Fig. 6C,E). NET formation, marked by citH3, was strongest after exposure to EVs from TF^+^ GBM cells, supporting a TF-dependent mechanism (Fig. 6D,E). In terms of senescence, EVs from PDPN⁺, TF⁺PDPN⁺, and even TF⁻PDPN⁻ promoted a senescence-like phenotype, characterized by elevated senescence markers and reduced Ki67, without affecting connexin 43 (Fig. 6F–I). This indicates that while TF and PDPN drive specific proinflammatory and NETosis responses, other EV-mediated signals, present even in the double-negative EVs, can contribute to senescence induction.
Together, these findings indicate that GBM-derived EVs exert distinct and complementary effects: TF⁺ EVs promote leukocyte activation and NETosis, while PDPN⁺ EVs induce senescence-like changes. These results support a model in which TF and PDPN in EVs are key contributors, while additional EV cargo molecules likely modulate senescence and other immune responses. Further studies are needed to determine whether these mechanisms operate in vivo and how they impact disease progression and patient outcomes.
Discussion
This study provides further mechanistic insight into the prothrombotic nature of GBM tumor cells. We engaged a translational workflow combining biobanked plasma and tumor samples, ex vivo GBM cultures, and coagulation assays, and found that GBM cell lines variably express TF and PDPN. This directly correlates with their clotting activation potential because GBM-cells directly facilitate TF-mediated thrombin generation and PDPN-driven platelet activation. Their EVs further modulate microglial behavior and trigger immune polarization with an impact on TME remodeling. Overexpression of TF and PDPN not only drives coagulation, tumor growth, epidermal-mesenchymal transition (EMT), and metastasis, but also contributes to immune modulation and evasion^12^.
TF and PDPN are highly upregulated across GBM patients, though with substantial intertumoral heterogeneity. In 2002, Guan et al. reported elevated F3 (TF) mRNA levels in GBM^6^, a finding later confirmed by TCGA data showing F3 expression in 93% of GBM cases, with 65% exhibiting moderate to high expression levels associated with poor prognosis^13^. PDPN has similarly emerged as an independent prognostic marker^14,15^, with evidence supporting its role in EV-mediated contribution to systemic hypercoagulability^9^. In our cohort, standard coagulation parameters (Quick, INR, aPTT) remained within normal ranges, suggesting no systemic hypercoagulability in these patients before surgery. In contrast, EVs extracted from the plasma of GBM patients significantly shortened CT, highlighting their potential role in cancer-associated thrombosis. In preclinical GBM models, we observed reduced cellular clotting times, confirming tumor-intrinsic procoagulant activity. While the extrinsic coagulation pathway, mediated by TF, was the primary driver, platelet aggregation assays revealed that cellular coagulation further exacerbates the prothrombotic state.
The roles of TF and PDPN in GBM-driven hypercoagulation are well-established; direct comparisons remain limited. However, the contribution of each part has not been assessed yet in a direct comparison. We addressed this by profiling both factors across four patient-derived GBM cell lines, which revealed a substantial heterogeneity of TF and PDPN expression. GBM14 exhibited the highest expression of both TF and PDPN, whereas GBM03 showed high PDPN but low TF expression. In contrast, GBM06 and GBM15 displayed low levels of both factors. Interestingly, MGMT promoter methylation was detected exclusively in GBM06, which also had low PDPN expression, consistent with prior reports linking MGMT methylation to PDPN suppression^16^. Beyond MGMT methylation, PTEN status may also play an important role^17^. Loss-of-function mutations in this tumor suppressor activate the PI3K/Akt signaling pathway, which not only promotes tumor growth but also upregulates procoagulant markers. Additional profiling of coagulation and inflammation-associated genes exhibited significant upregulation of PDPN, PLAU, THBS1, VEGFA, ANGPT2, and MMP9 in primary GBM cases, reflecting an active thrombo-inflammatory TME. These procoagulant and inflammatory gene signatures were retained in patient-derived GBM cell lines. Correlation analyses revealed a negative association between PDPN and both the anticoagulant gene TFPI1 and the procoagulant gene, suggesting complex regulation of coagulation factors.
An additional novel finding of our study is the co-expression of TF and PDPN in GBM14 cells. Notably, the patient from whom the GBM14 cell line was derived developed cerebral sinus venous thrombosis during the disease course. This case highlights the potential translational relevance of our in vitro findings, linking elevated tumor-intrinsic expression of TF and PDPN with clinically evident hypercoagulability. Although in previous research this dual expression has been studied, functional insights remain limited. Our phenotypic sorting of GBM14 subpopulations revealed dynamic plasticity, with a progressive enrichment of TF⁺ cells across all sorted groups. This suggests an intrinsic bias toward a TF-dominant phenotype, which may also manifest during tumor progression or recurrence in vivo. Given the influence of oxidative stress on tumor plasticity, we investigated ROS production in GBM subpopulations. Elevated ROS levels were observed predominantly in TF⁺ cells, whereas the TF⁻PDPN⁻ double-negative population exhibited lower ROS levels, indicating a potential link between TF expression and oxidative stress. Interestingly, PDPN⁺ and TF⁺PDPN⁺ cells also generated ROS, but at lower levels. Possibly, these levels were sufficient to trigger TF induction but insufficient to maintain co-expression. These findings align with previous reports linking TF to oxidative stress^18,19^, and collectively highlight ROS as a potential driver of the procoagulant phenotype on GBM. The dynamic regulation of TF and PDPN underlines the adaptive nature of GBM cells and their role in shaping the tumor’s thrombo-inflammatory environment.
EVs are nanoscale particles secreted by viable cells, including tumor cells, into the surrounding environment and circulation, where they act as bioactive cargoes capable of modulating coagulation, immune responses, and tumor progression^20^. By carrying signatures of their origin cells—such as the prothrombotic factors TF and PDPN, these vesicles can directly influence systemic and local thrombo-inflammatory processes^9^. Supporting this concept, a previous study in a xenograft model co-expressing both factors reported increased intravascular fibrin deposition and vessel-occluding microthrombi, implicating EVs as key mediators^9^. In our study, TF⁺ EVs likely initiate the coagulation cascade, whereas PDPN⁺ EVs exerted minimal to no effects on platelet activation and aggregation under the tested conditions, indicating a limited contribution to cellular procoagulant activity, which, taken together, could potentially support a modest prothrombotic feedback loop. Additionally, PDPN⁺ EVs contributed to senescence and microglial modulation, and EVs from TF⁻PDPN⁻ populations exhibited limited procoagulant activity but could still induce moderate iNOS expression in microglia, suggesting additional multifactorial EV effects.
Beyond these procoagulant effects, we also observed that tumor-derived EVs were rapidly internalized by microglia. This uptake may reflect both immune surveillance and potential reprogramming of microglia to support tumor-associated inflammation, growth, and coagulation. However, uptake kinetics varied between tumor-derived EVs of different GBM cell origins. EVs from the TF^high^PDPN^high^ GBM14 cells exhibited the highest uptake and were associated with reduced microglial migration, suggesting that EV cargo modulates microglial behavior. Our data also imply that EVs of TF^high^PDPN^high^ origin cells take part in regulating microglial senescence and polarization. These findings are consistent with a recent report linking senescence with hypercoagulability following irradiation, where TF promoted clonal expansion of irradiated SA-βGal⁺ GBM cells, activating intrinsic oncogenic signaling and extrinsic coagulation cascades alike^21^. In contrast, PDPN counteracts senescence by inhibiting the p16^Ink4a^/Rb axis^22^, which aligns with our observation of increased microglial proliferation upon exposure to EVs from TF^low^PDPN^high^ GBM03 and TF^high^PDPN^high^ GBM14 cells. In GBM14, TF-up and PDPN-downregulation may create a compensatory balance between proliferation and senescence. However, this hypothesis remains speculative, as only a minority of cells exhibited proliferative activity, suggesting the involvement of additional regulatory pathways.
The characteristic myeloid cell infiltration in GBM, particularly by tumor-associated macrophages, provides an explanation for our findings of microglial polarization by EVs from TF^high^PDPN^high^ GBM cells. In our study, EVs from TF^low^PDPN^high^ GBM03 and TF^high^PDPN^high^ GBM14 induced iNOS, indicating M1-like microglial polarization, while EVs from TF^low^PDPN^low^ GBM06 reduced iNOS, suggesting an M2-like shift, and EVs from the TF⁻PDPN⁻ GBM14 population did not significantly alter iNOS expression, suggesting that TF and PDPN are key contributors to M1 polarization. This suggests that EVs from TF^high^PDPN^high^ cells preferentially promote M1 polarization, while prior studies have shown that TF can support both M1 (via PAR-2 activation) and M2 (via FXa-AKT/ERK signaling) phenotypes^23,24^. In GBM, PDPN is a key regulator of the TME, associated with M2 polarization, neutrophil degranulation, and increased CD68^+^CD163^+^ M2-like macrophages^25^. Although this association is well documented, our findings do not fully align. In our study, EVs from TF^high^PDPN^high^ GBM14 cells induced M1-like microglial polarization, and interestingly, EVs from GBM03, which are TF low but PDPN high, also showed a trend towards M1-like polarization. This suggests that PDPN’s effect on microglial polarization may be context-dependent, influenced by the combination of EV cargo, the activation status of the recipient cells, or other microenvironmental factors. One possible explanation for the M1-like polarization observed with TF^high^PDPN^high^ EVs is that PDPN, while associated with M2 polarization in other contexts, may synergize with TF in the same vesicles to activate pro-inflammatory pathways, overriding the M2-promoting effects of PDPN alone. Beyond TF and PDPN, additional EV components, such as miRNAs or other signaling molecules, may further modulate macrophage polarization in a tumor-specific manner^26^. While our data suggest differential effects on M1 polarization, further studies are needed to clarify the specific contributions of these EVs to tumor–immune cell interactions within the TME. Specifically, expression levels of connexin 43 remained unchanged, which either indicates that this gap junction protein is not directly involved in the immediate cellular response to EVs uptake or that its expression is regulated in a way that does not fluctuate significantly in this context. This warrants further investigation into whether connexin 43 plays a subtler role in the broader context of EV-mediated cellular reprogramming.
EVs from TF^high^PDPN^high^ cells also activate leukocytes and induce NETosis, reflecting a TF-dependent inflammatory response, supported by increased MPO, a hallmark of neutrophil activation. TF⁺ EVs promoted leukocyte activation, ROS generation, and NET formation, whereas PDPN⁺ EVs induced senescence without strong ROS/NET effects, illustrating complementary roles of TF and PDPN in immune modulation. In contrast, EVs from TF⁻PDPN⁻ cells did not induce significant leukocyte or neutrophil activation, ROS production, or NET formation, while PDPN⁺ EVs contributed primarily to senescence induction, indicating that the proinflammatory and NETosis effects are largely TF-dependent. This TF-dependent immune activation aligns with our finding of M1-like polarization of microglia in response to TF^high^PDPN^high^ GBM14-derived EVs. Co-incubation with GBM-derived EVs further upregulated TF and ROS in leukocytes and neutrophils, suggesting a role for immune thrombosis in tumor progression and immune evasion^27^. We showed that EVs from TF^high^ cells promote neutrophil activation and induce NET formation, as evidenced by elevated MPO and citH3. As NETs are negatively charged and can initiate the intrinsic coagulation cascade via FXII, their induction further links GBM with an increased thrombotic risk – consistent with findings in breast cancer^28^. However, several limitations must be acknowledged. The relatively small and heterogeneous patient cohort, together with the variability between cell lines, may have limited the statistical robustness and generalizability of the findings. Although patient-derived GBM cultures retain essential transcriptional and phenotypic features, in vitro systems cannot fully capture the complexity of the (in vivo) tumor microenvironment. Likewise, the use of isolated extracellular vesicle preparations may not entirely reflect the composition and biological activity of circulating EVs in patients. While the results strongly support a causal contribution of TF and PDPN to GBM-associated thromboinflammation, additional validation is required to delineate their specific roles in disease progression. The observed interplay between oxidative stress, TF expression, and EV biogenesis also remains incompletely understood and merits further mechanistic exploration. To strengthen and extend these findings, future studies should include larger patient cohorts and employ advanced in vitro 3D models, such as patient-derived organoids, which more faithfully preserve tumor–immune–stromal interactions within the native microenvironment and thus offer a more physiologically relevant platform for translational research.
Taken together, our results demonstrate that TF and PDPN are critical immune modulators in GBM, with TF confirmed as the dominant driver of thromboinflammation. GBM-derived EVs are internalized by microglia, reducing their motility, promoting a senescence-like state, and altering polarization. Overall, this study underscores that procoagulant TF⁺ EVs orchestrate immune and prothrombotic responses via multiple mechanisms, and that EVs lacking both TF and PDPN do not elicit these responses, highlighting the complementary and distinct roles of these factors in GBM-mediated immune modulation. These observations warrant evaluation as potential biomarkers for thrombotic risk stratification in GBM patients.
Conclusion
This study underscores the critical role of extracellular vesicles and tumor-intrinsic mechanisms in driving tumor-associated hypercoagulation in GBM. By integrating clinical data, ex vivo GBM models, and tumor-derived extracellular vesicles, we identified key molecular and functional pathways linking vesicle-mediated prothrombotic activity with tumor-driven inflammation. These findings show how both extracellular vesicles and tumor-specific mechanisms shape a procoagulant tumor microenvironment and provide a foundation for future research exploring therapeutic strategies to modulate coagulation in GBM.
Materials and methods
Study aim
To investigate tumor-associated hypercoagulation in glioblastoma, with particular emphasis on the dual role of extracellular vesicles in coagulation and immune modulation, using an integrated translational approach combining clinical data, biobanked tumor tissue, ex vivo GBM cell models, and tumor-derived extracellular vesicles.
Materials
The study was conducted at the Rostock University Medical Center (Ethics Registration ID: A2018-0167). Tumor tissue and intraoperative blood samples (collected in citrate tubes) were obtained from patients diagnosed with IDH-wildtype GBM during surgical resection, following written informed consent. Patient-derived GBM cell lines were established from freshly resected tumor specimens as described before^29^. Continually growing cell cultures were serially passaged and regularly stocked in low passages. For this study, GBM03, GBM06, GBM14, and GBM15 cells were included. In selected experiments, GBM cells transduced via a lentiviral vector carrying the gene for the near-infrared (NIR) fluorescent protein (iRFP680), which emits red fluorescence, in viable cells, were employed^30,31^. In addition, hMC3 microglia were used. Cell lines were cultured in 2D using Dulbecco’s Modified Eagle Medium/Nutrient Mixture F-12 (DMEM/F12) supplemented with 10% fetal calf serum, 6 mmol/l L-glutamine, and 1% penicillin/streptomycin (all from PAN-Biotech, Aidenbach, Germany) for GBM Cell lines and Eagle’s Minimum Essential Medium (EMEM, ATCC, Manassas, Virginia, US) supplemented with 10% fetal calf serum (PAN-Biotech) for hMC3, incubated at 37 °C in a humidified atmosphere with 5% CO_2_.
Patient- or tumor-derived EVs were isolated via differential centrifugation from either patient plasma or cell culture supernatants. Leukocytes and platelets from healthy donors were used with the informed consent of the individuals.
ClotPro analysis and coagulation assay
Citrate blood from GBM patients was immediately subjected to ClotPro^®^ viscoelastometry for performing EX-test, IN-test, and FIB-test. The EX-test (extrinsic pathway) and FIB-test (fibrinogen function) are performed with tissue factor activation, while the IN-test assesses the intrinsic pathway of coagulation. The remaining blood was centrifuged (2000 x g, 10 min), the plasma was collected and stored at −80 °C for coagulation assays.
Tumor histology and immunohistochemistry
Histopathological examination of the primaries was done by a trained pathologist, and additional staging information was compiled from patients’ clinical charts. H&E sections were obtained from paraffin-embedded tumors.
RNA sequencing and gene expression analysis of coagulation-associated genes
Total RNA from primary tumors and cell lines was isolated using the RNeasy Mini Kit (Qiagen, Hilden, Germany) with on-column DNase treatment, following the manufacturer’s instructions. RNA quality was assessed using a Nanodrop spectrophotometer, and samples with OD260/280 ≥ 2.0 and OD260/230 ≥ 2.0 were selected for sequencing. RNA sequencing was performed with poly(A) enrichment on the Illumina PE150 platform, generating approximately 30 million paired-end reads per sample. Bubble plots illustrating coagulation-associated gene expression were generated using R Studio (version 4.4.2).
For gene expression analysis of coagulation-associated genes, RNA isolation from primary tumors and cell lines was again performed using the RNeasy^®^ kit (Qiagen, Hilden, Germany) according to the manufacturer’s protocol. 1 µg of total RNA was reverse transcribed into cDNA using Random Hexamer Primers (50 ng/20 µL reaction volume). The reverse transcription reaction was completed by adding RT buffer, dNTP mix, and Reverase™ enzyme (Bioron, Römerberg, Germany).
The 25 ng of cDNA were mixed with 0.65 µL of predesigned Taqman™ gene expression assays 6-FAM-3ʹNFQ-MGB F3, 6-FAM-3ʹNFQ-MGB TFPI1, or 6-FAM-3ʹNFQ-MGB PDPN (Applied Biosystems, Darmstadt, Germany), 0.65 µL of in-house 5-VIC-3ʹNFQ-MGB GAPDH for normalization, 6.5 µL of master mix, and 3.2 µL of water. The reaction was performed on a LightCycler^®^ ViiA7 system (Applied Biosystems) as previously described^32^. Gene expression levels were quantified using the 2^−ΔCT^ method, where ΔCT = CT_target − CT_housekeeping.
Full-spectrum flow cytometry
Furthermore, a modified staining approach was employed to study TF (CD142) and PDPN. In brief, viability and antibody (Ab) staining were conducted on 250,000 cells. Cell viability was investigated using the Zombie NIR™ Fixable Viability Kit (BioLegend, San Diego, California, United States, 1:5000). All further procedures were performed using staining buffer (PBS, 2 mM EDTA, 2% BSA). Therefore, PE-conjugated anti-human CD142 Ab (clone: NY-2, stock: 50 µg/ml, BioLegend, 1:50) and Alexa Fluor^®^ 488-conjugated anti-human Podoplanin Ab (clone: NC-08, stock: 200 µg/ml, BioLegend, 1:100) were used. Extracellular staining was done for 30 min at RT, followed by two washing steps (350 x g, 5 min). Cells were resuspended in 0.25 ml staining buffer. All flow cytometry measurements were performed on the Cytek™ Aurora three laser spectral flow cytometer. Data were analyzed using SpectroFlo™ Version 3.3 and FlowJo™ Version 10.6.1.
Immunofluorescence staining of TF and PDPN with and without low-weight molecular heparin (LWMH) on primary tumor sections and tumor cells
Immunofluorescence staining of TF and PDPN was performed on primary tumor cryosections and tumor cells. Sections were washed three times with PBS and blocked with 2% BSA in 1x PBS for 1 h at room temperature in a humidified chamber. Ab staining was done at 4 °C overnight using PE anti-human CD142 Ab (clone: NY-2, stock: 50 µg/ml, BioLegend, 1:20) and Alexa Fluor^®^ 488 anti-human Podoplanin Ab (clone: NC-08, stock: 200 µg/ml, BioLegend, 1:20).
For tumor cells a total of 3,000 cells/well were seeded in µ-slides 18 wells (ibidi, Gräfelfing, Germany), followed by anticoagulant treatment with LWMH (Clexane, 16 mg/l), every 24 h for three days. Control cells were left untreated. Blocking was performed with 0.5% Triton^®^ X-100, 2% BSA for 60 min and specific Ab staining was done at RT for 2 h using PE anti-human CD142 Ab (clone: NY-2, stock: 50 µg/ml, BioLegend, 1:50) and Alexa Fluor^®^ 488 anti-human Podoplanin Ab (clone: NC-08, stock: 200 µg/ml, BioLegend, 1:100).
For both, nuclear staining was performed using DAPI (stock: 1.75 mg/mL; Thermo Fisher Scientific, Waltham, Massachusetts, USA, 1:1,750), followed by microscopic examination (AxiovertA.1, Zeiss, Jena, Germany). Data were analyzed using ImageJ.
PS exposure and platelet activation
In general, activated platelets can be characterized by CD62P (P-selectin) and CD41/CD61 expression. Only a subpopulation of activated platelets exposes PS on their surface, exhibiting a procoagulant phenotype. Blood from healthy AB0 blood group 0 donors was collected in tubes containing acid citrate dextrose solution A (ACD-A). Platelet-rich plasma (PRP) and platelet-poor plasma (PPP) were prepared by centrifugation. Platelets were isolated from PRP as described^33^. Briefly, PRP was washed twice with Tyrode buffer containing 0.35% bovine serum albumin (BSA), 0.1% glucose, 2.5 U/mL apyrase, 1 U/mL hirudin, pH 6.3. Finally, the platelet pellet was resuspended in a bicarbonate-based suspension buffer containing 0.35% BSA, 0.1% glucose, 0.2 M MgCl_2_, 0.2 M CaCl_2_, pH 7.2. Platelets were adjusted to 300,000 platelets/ml. For PS exposure analyses, 15 µL of platelets and 15,000 tumor cells or 1 × 10^9^ particle/ml EVs in 15 µL were added to 96-well plates containing two metal spheres/well and stirred magnetically for 20 min. Staining with APC Annexin V (BioLegend) and FITC anti-human CD62P (clone: AK4, stock: 200 µg/ml, BioLegend, 1:2) was performed in binding buffer (BioLegend) for 20 min. For activation analyses, platelets were co-incubated with tumor cells or EVs for 20 min without shear stress, followed by a 20 min staining with APC anti-human CD62P (clone: AK4, stock: 200 µg/ml, BioLegend, 1:10)and FITC anti-human PAC-1 (clone: PAC-1, BioLegend, 1:10). Reactions were stopped with PBS. In both experiments, PBS served as a negative control; a mixture of 20 µM thrombin receptor activator peptide 6 (TRAP-6) and 100 ng/mL convulxin served as a positive control. Measurements were conducted with or without 5 µL PPP and 2.25 mg/mL fibrinogen, using a FACSCalibur flow cytometer and CellQuestPro 6.0 software.
Platelet aggregometry
To assess tumor cell- and EVs-induced platelet aggregation, platelet-rich plasma (PRP) was pre-warmed at 37 °C for 5 min in a light-transmission aggregometer (Chrono-Log, Haverton PA, U.S.). A total of 225 µL PRP was aliquoted into four glass cuvettes, with PPP in a fifth cuvette serving as reference. After pre-incubation, TRAP (20 µM, positive control), NaCl (negative control), or tumor cells/EVs were added to the PRP cuvettes to a final volume of 250 µL, and aggregation was recorded over 12–20 min.
Cell sorting approach
GBM14 cells were sorted into three populations: TF^+^, PDPN^+^, TF^+^PDPN^+^, and TF^−^PDPN^−^. In the approach, 250,000 cells were stained in 100 µl FACS staining buffer (PBS, 2 mM EDTA, 2% BSA) with PE anti-human CD142 Ab (clone: NY-2, stock: 50 µg/ml, BioLegend, 1:50) and Alexa Fluor^®^ 488 anti-human Podoplanin Ab (clone: NC-08, stock: 200 µg/ml, BioLegend, 1:50). The cells were washed and resuspended in FACS staining buffer, and sorted for live (30 µM DAPI, BioLegend) cells on a BD FACSAria™ IIIu. The cells were cultivated and used for clotting or functional assays.
Detection of reactive oxygen species using ROS Brite™ 670 and mitochondrial superoxide with MitoSOX™ Red
ROS levels were measured using ROS Brite™ 670 (stock: 10 mM, AAT Bioquest, Pleasanton, California, United States, 1:100), a cell-permeable dye that emits red light upon oxidation. On days 0, 7, 14, and 21, 5,000 sorted GBM14 cells/well were seeded in 96-well plates, incubated overnight (37 °C), followed by ROS Brite staining (final concentration: 100 µM, 15 min, 37 °C) and fluorescence was measured (ex/em = 640/680 nm) using a Tecan reader Infinite^®^ M Plex (Tecan, Männedorf, Switzerland). To detect mitochondrial superoxide, 1,000 GBM14 cells were seeded in µ-slides 18 wells (ibidi) and incubated overnight. Cells were stained with MitoSOX™ Red (Stock: 5 mM, Invitrogen, Waltham, Massachusetts, United States, 1:1250) and Calcein-AM (4 µM, 30 min, 37 °C). After two PBS washes, samples were imaged using a Zeiss Axiovert A.1 microscope.
EVs isolation and nanoparticle tracking analysis (NTA)
EVs were isolated from 1 ml of supernatant of sorted GBM14 cells. After centrifugation at 400 x g for 10 min to remove debris, supernatant was centrifuged three times with PBS at 14,000 x g for 30 min to isolate EVs. Finally, EVs were resuspended in the remaining PBS and stored at − 80 °C.
For quantification, EVs were diluted in PBS and analyzed using a NanoSight^®^ LM10 (Malvern Panalytical, Malvern, United Kingdom, Software: NTA 3.3). Five 30 s video recordings were acquired per sample, with a thermometer attached to monitor temperature. Mean size and concentration were calculated, and PBS background was subtracted.
Clotting of GBM cells and EVs with normal, FXII-deficient, and FVII-deficient plasma & TF ELISA
The clotting time of GBM cells and EVs was quantified using a coagulometer, preheated to 37 °C. Cells were adjusted to a concentration of 10,000 cells/ml. EVs were diluted with PBS to 1 × 10^9^ particles/ml. For measurement, the following reagents were added sequentially: normal, FXII-deficient, or FVII-deficient plasma, cell/EVs suspension, and 25 mM CaCl_2_. EVs were clotted with normal and FXI-deficient plasma only. The clotting time was analyzed. The TF-ELISA was performed according to the manufacturer’s instructions (Elabscience^®^).
NIR tumor cell-derived EVs uptake via microglia
Tumor-derived EVs uptake by microglia (hMC3) was studied using EVs from the supernatants of GBM NIR cells as outlined above. Therefore, 10,000 hMC3 were seeded in a 24-well plate, and 1 × 10^9^ particles/ml of the respective NIR-GBM EVs or 1x PBS (solvent) were added as control. The analysis was conducted at four distinct time points (6, 24, 48, and 72 h) using a microscope (Axiovert A.1, Zeiss, Jena, Germany).
Migration assay
Microglial migration in response to tumor-derived EVs was assessed using a transwell assay. Microglia were first cultured in serum-free medium for 24 h. Subsequently, serum-containing medium was added to the lower chambers of 24-well transwell plates (8 μm pore size), and 500,000 microglia in serum-free medium were seeded into the upper chambers. Tumor-derived EVs (1 × 10⁹ particles) were added to the microglia for 24 h. After incubation, non-migrated cells on the upper membrane surface were gently removed with a swab. Migrated microglia on the lower membrane surface were stained with Calcein AM (4 µM, 30 min, 37 °C), and migration was quantified by fluorescence measurement using Tecan reader Infinite^®^ M Plex.
Immunofluorescence staining for EVs-induced proliferation, senescence, and myeloid polarization
To distinguish between proliferative and senescent phenotypes following EVs exposure, 5,000 microglia/well were seeded in µ-slide 18-well (ibidi). Cells were fixed and then simultaneously blocked and permeabilized using 0.5% Triton^®^ X-100 in 2% BSA (60 min, room temperature).
Proliferation was assessed using Alexa Fluor™ 647 anti-human Ki-67 Ab (clone: Ki-67, stock: 0.08 mg/ml, Invitrogen, 1:50) and Alexa Fluor™ 488 anti-human connexin 43 Ab (clone: CX-1B1, stock: 0.5 mg/mL Thermo Fisher Scientific, 1:50). Senescence was evaluated using Alexa Fluor^®^ 594 p53 Ab (clone: DO-1, stock: 0.5 mg/ml, BioLegend, 1:250), Alexa Fluor^®^ 488 anti-human p21^Waf1/Cip1^ Ab (clone: 12D1stock: 50 µg/ml, Cell Signaling Technology, 1:300), and Alexa Fluor^®^ 546 anti-human p16 Ab (clone: F-12, stock: 200 µg/ml, Santa Cruz Biotechnology, Dallas, Texas, United States, 1:50).
M1 polarization was detected via Alexa Fluor^®^ 546 anti-human iNOS Ab (clone: C-11, stock: 200 µg/ml, Santa Cruz Biotechnology, 1:50). All antibodies were incubated for 2 h at room temperature. Nuclei were stained with DAPI (stock: 1.75 mg/mL; Thermo Fisher Scientific, 1:1750). Imaging and quantification were performed using fluorescence microscopy (AxiovertA.1, Zeiss) and ImageJ.
EVs-induced leukocyte and neutrophil activation
Peripheral blood was centrifuged (2,000 x g, 10 min), plasma removed, and erythrocytes lysed in NH₄Cl buffer to isolate leukocytes. Neutrophils were further purified using the MojoSort™ Neutrophil Isolation Kit (BioLegend) with a 10 min magnetic incubation, followed by erythrocyte lysis. A total of 250,000 leukocytes or neutrophils in 100 µL were placed within circles pre-drawn on slides using a Dako Pen to confine the cells. Cells were then stimulated with 1 × 10⁹ particles/ml sorted GBM14-derived EVs for 4 h at 37 °C, followed by fixation and blocking. For detection, cells were stained with CoraLite^®^ Plus 488 anti-human MPO Ab (clone: 4C11F6, stock: 1 mg/ml, Proteintech, Chicago, Illinois, United States, 1:200), PE anti-human CD142 Ab (clone: NY-2 stock: 50 µg/ml, BioLegend, 1:50), and ROS Brite™ 670 (stock: 10 mM, AAT Bioquest, 1:100). Nuclei were counterstained with DAPI (stock: 1.75 mg/mL; Thermo Fisher Scientific, 1:1,750).
To detect NETosis, neutrophils were incubated overnight at 4 °C with rabbit anti-citH3 Ab (stock: 0.5 mg/ml, Thermo Fisher Scientific, 1:50) followed by a secondary Ab: anti-rabbit IgG (H + L) F(ab′)₂–Alexa Fluor^®^ anti-human 647 Ab (stock: 0.5 mg/ml, BioLegend, 1:250), and DAPI (stock: 1.75 mg/mL; Thermo Fisher Scientific, 1:1,750). All samples were analyzed via fluorescence microscopy (Axiovert A.1, Zeiss) and quantified using ImageJ.
Statistics
All values are given as mean ± SD. Statistical evaluation was performed using GraphPad PRISM software, version 10.5.0 (GraphPad Software, San Diego, CA, USA). The criterion for significance was set to p < 0.05. All experiments were done in biological triplicate. After proving the assumption of normality (Shapiro–Wilk test), One-way ANOVA (Dunnett’s multiple comparison), Two-way ANOVA (Tukey’s multiple comparison), or t-test was performed. If the normality test failed, the Kruskal–Wallis or Mann–Whitney U-Test was performed. Survival distributions were compared using the log-rank test.
Supplementary Information
Below is the link to the electronic supplementary material.
Supplementary Material 1