Abstract
Cancer-related venous thromboembolisms (VTE) are associated with metastasis and reduced survival in patients with urothelial cancer of the bladder. Although previous reports suggest the contribution of tissue factor and podoplanin, the mechanistic linkage between VTE and bladder cancer cell–derived molecules is unknown. Therefore, we compared distinct procoagulant pathways in four different cell lines. In vitro findings were further confirmed by microfluidic experiments mimicking the pathophysiology of tumor blood vessels and in tissue samples of patients with bladder cancer by transcriptome analysis and immunohistology. In vitro and microfluidic experiments identified bladder cancer–derived VEGF-A as highly procoagulant because it promoted the release of von Willebrand factor (VWF) from endothelial cells and thus platelet aggregation. In tissue sections from patients with bladder cancer, we found that VWF-mediated blood vessel occlusions were associated with a poor outcome. Transcriptome data further indicate that elevated expression levels of enzymes modulating VEGF-A availability were significantly connected to a decreased survival in patients with bladder cancer. In comparison with previously postulated molecular players, we identified tumor cell–derived VEGF-A and endothelial VWF as procoagulant mediators in bladder cancer. Therapeutic strategies that prevent the VEGF-A–mediated release of VWF may reduce tumor-associated hypercoagulation and metastasis in patients with bladder cancer.
Implications: We identified the VEGF-A–mediated release of VWF from endothelial cells to be associated with bladder cancer progression.
Introduction
With an estimated 180,000 newly diagnosed cases and 38,000 deaths per year in Europe, cancer of the urinary bladder (UCB) is the second most common genitourinary malignancy in Europe and among the five most common malignancies worldwide (1, 2). Urothelial carcinoma, the most prevalent histologic subtype, accounts for 90% of all UCB, with squamous cell carcinoma and adenocarcinoma being far less frequent (3). While patients with nonmuscle-invasive UCB have an excellent 5-year survival rate of >90%, the survival rate is drastically lower in patients suffering from muscle-invasive bladder cancer or disseminated disease (4, 5). The efficiency of standard chemotherapy with cisplatin, but also of new immunotherapies such as PD-L1 inhibitors, is limited (6). It is well known that patients with cancer harbor an elevated risk of venous thromboembolism (VTE), which is associated with an increased tumor cell dissemination and decreased survival (7–10). A growing body of evidence suggests a correlation between hypercoagulopathy, VTE, and the systemic spread of tumor cells (10). The hypercoagulative phenotype of cancer cells is not restricted to specific entities but appears to be a common feature of several malignancies, including pancreatic and ovarian cancer, melanoma, and glioblastoma (11). Abnormal plasma fibrinogen and D-dimer levels were measured in patients with localized, nonmetastatic renal, prostate, and bladder cancer, suggesting early activation of hemostasis (12). Empirical findings suggest that patients with bladder cancer suffer from an increased risk of developing VTE, indicating a cross-talk between UCB cells and the coagulation system. Various cancer cell–derived factors have been attributed to hypercoagulation in terms of platelet activation, initiation of the plasmatic coagulation, or endothelial interaction (11, 13, 14). In this regard, we recently reviewed the potential relevance of the coagulation system in urological cancer (15). We hypothesized that UCB cells promote hemostasis, VTE, and thereby tumor progression. Now, we studied tumor-associated intravascular clot formation in bladder cancer and we aimed to identify molecular players that promote the cross-talk between UCB cells and distinct coagulatory pathways. We investigated the impact of three previously postulated procoagulant effector molecules on UCB progression: tissue factor (TF), podoplanin (PDPN), and endothelial-derived von Willebrand factor (VWF). TF is a transmembrane receptor physiologically exposed by subendothelial cells such as fibroblasts. In pancreatic cancer, TF expression by tumor cells and increased levels of blood circulating TF+ microvesicles have been attributed to tumor-associated hypercoagulation and VTE (16). Interestingly, high TF serum levels have been previously shown in patients with bladder cancer to coincide with rapid disease progression (13, 17). PDPN is a mucin-type protein found in many tissues and directly interacts with the platelet C-type lectin-like receptor 2 (CLEC-2), which promotes platelet aggregation and activation. Although increased levels of PDPN have been correlated with an increased incidence of VTE in glioblastoma, its relevance for tumor progression is under debate (18). In bladder cancer, high levels of PDPN were detected in serum of patients with invasive, more aggressive, larger, and multifocal tumors (19). VWF is a polymeric glycoprotein stored in Weibel–Palade bodies within the vascular endothelium and is secreted upon cell activation. This endothelial cell activation can be triggered by various compounds including tumor cell–derived VEGF-A (20). Once released into the blood flow, VWF is elongated to a string. In its stretched conformation VWF is able to bind platelets and to induce blood vessel occlusion (21). Increased concentrations of VWF were found in the plasma of patients suffering from bladder cancer (22). Supplementary Fig. S1 summarizes the distinct molecules potentially contributing to cancer-associated hypercoagulation. Prediction of VTE based on clinical parameters is to date difficult (23). Therefore, we aimed to analyze the procoagulant impact of several bladder cancer cell–derived molecules in functional in vitro assays and in alignment with clinical data.
Materials and Methods
Cell culture
The human UCB cell lines RT4, RT112, and T24/83 were obtained from the European Collection of Authenticated Cell Cultures. The simian virus 40 large T antigen immortalized UROtsa cell line served as a model for the benign urothelium and were provided by Philipp Erben (University of Heidelberg, Mannheim, Germany). All cell lines were tested negative for Mycoplasma using the VenorGeM Classic Kit (Minerva Biolabs GmbH). Cell line authentication through the identification of cell-specific SNPs was done as reported previously (ref. 24; Multiplexion GmbH). Cells were cultured in 75 cm2 flasks containing RPMI1640 medium, enriched with 1% nonessential amino acids, 1% glutamine, 1% antibiotics (penicillin and streptomycin), and 10% FBS (Boehringer Mannheim). The incubation was performed at 37°C in humid atmosphere of 5% CO2. Twice a week the media were changed with subcultivation at 70%–80% of confluency. Primary human umbilical vein endothelial cells (HUVEC) were freshly isolated from umbilical cord veins and grown in culture medium composed of two-thirds M199 supplemented with 10% heat-inactivated FCS, 1% antibiotics (penicillin and streptomycin), and one-third EGM-2 (Lonza; ref. 25). HUVECs were maintained at 37°C, 5% CO2. Urothelial cell lines were used a maximum of 85 passages. HUVECs were used a maximum of four passages.
Generation of UCB cell–derived supernatants
For standardized generation of supernatants, UCB cells were grown to confluency in a 75 cm2 flask. After removal of the culture medium, the cells were thoroughly washed with HEPES-buffered Ringer solution (HBRS) consisting of 140 mmol/L NaCl, 5 mmol/L KCl, 1 mmol/L MgCl2, 1 mmol/L CaCl2, 5 mmol/L glucose, and 10 mmol/L HEPES. After incubation in 10 mL RPMI1640 without additives for 24 hours, the conditioned medium was harvested from the intact cell layer and centrifuged at 1,800 × g for 10 minutes to eliminate cell debris. Finally, the supernatants were frozen at −80°C until experiments were performed.
Stimulation of HUVECs
HUVECs were grown on gelatin-coated coverslips in standard culture 12-well plates. At confluency, culture medium was removed and cells were rinsed twice with prewarmed HBRS. Subsequently, different stimuli were applied: UCB cell supernatants; recombinant human VEGF165 (10 ng/mL, R&D Systems), recombinant human IL8 (2 ng/mL, R&D Systems), or human thrombin 0.5 U/mL (Sigma-Aldrich). Where indicated, the VEGF-A–neutralizing antibody bevacizumab (0.65 μg/mL, Roche Diagnostics GmbH) or anti-IL8 (0.72 μg/mL, R&D Systems) were added to the cells. Prewarmed serum-free medium served as control. The supernatants of endothelial cells were collected after 15 minutes of incubation at 37°C, centrifuged at 300 × g for 5 minutes to remove cell detritus, and kept at −20°C for later VWF and P-selectin analysis. The cells on coverslips were fixed and subjected to immunofluorescence staining.
Proteome profiler array
The proteome profiler Human Angiogenesis Antibody Array (C Series, RayBiotech) was used to measure the relative levels of UCB cell–released proteins in the conditioned medium. The assay was performed according to the manufacturer's instructions. An equal volume of each supernatant was used for the assay. The dot blot density was analyzed using ImageJ software.
Immunofluorescence staining of HUVECs
After stimulation, HUVECs on coverslips were fixed in ice-cold methanol for 30 minutes, washed with HEPES, and blocked in 2% BSA dissolved in incubation buffer (0.1% BSA and 0.3% Triton X-100 in HBRS) for 1 hour at room temperature. For detection, the antibodies rabbit anti-human VWF (Dako) at a dilution of 1:200 and FITC-conjugated goat anti-rabbit (BD Pharmingen) at a dilution of 1:400 were used in incubation buffer for 1 hour at room temperature. Costaining for P-selectin was performed using a mouse anti-human CD62p antibody (LSBio) at a dilution of 1:200 and an Alexa555–conjugated goat anti-mouse antibody (Invitrogen) at a dilution of 1:400 in incubation buffer for 1 hour at room temperature. Nuclei were stained with 4,6 diamidino-2-phenylindole (DAPI) diluted 1:1,000 in PBS for 10 minutes. Finally, coverslips were embedded with Mowiol–glycol solution with freshly added 50 mg/mL of 1,4-Diazabicyclo[2.2.2]octane (DABCO, Sigma-Aldrich). Microscopy was performed using an Observer.Z1 (Zeiss) and the data were processed with Zeiss Zen software (version 1.1.2.0) and ImageJ (version 1.51j8).
Platelet preparation
Platelets were prepared as reported previously (26). Briefly, blood was drawn into sodium citrate–coated tubes (Sarstedt) and centrifuged at 120 × g for 15 minutes at room temperature to obtain platelet-rich plasma (PRP). To inhibit artificial platelet activation, PRP was supplemented with apyrase (75 μU/mL; Sigma-Aldrich Chemie GmbH) and prostaglandin E1 (100 nmol/L; Sigma-Aldrich) and centrifuged at 1,200 × g for 15 minutes. The platelet pellet was resuspended in wash buffer containing Calcein red-orange (5 μmol/L; Invitrogen) at a final dilution of 1:1,000 and incubated for 30 minutes in darkness at room temperature. Subsequently, platelets were centrifuged for 12 minutes at 1,200 × g and washed twice with wash buffer (36 mmol/L citric acid, 5 mmol/L glucose, 5 mmol/L KCl, 1 mmol/L MgCl2, 103 mmol/L NaCl, 2 mmol/L CaCl2, 75 μU/mL apyrase, 100 nmol/L PGE2, 3.5 mg/mL BSA, pH 6.5). Finally, the pellet was resuspended in wash buffer.
Light transmission aggregometry
The ability of each cancer cell line mediating platelet aggregation was evaluated by light transmission aggregometry (LTA) using a Chrono-log Model 700 (Chrono-Log Corporation). PRP was obtained by centrifugation of citrated blood tubes (S-Monovette, Sarstedt) at 180 × g for 12 minutes. To abolish the chelating effect of Na-Citrate, CaCl2 was added into the PRP to a final concentration of 2 mmol/L. In each experiment 5 × 105 tumor cells were added to the PRP and aggregation of platelets was monitored for 30 minutes at constant stirring of 1,200 rpm at 37°C.
Microfluidics
A laminar flow was induced in microfluidic devices employing an air pressure–based Pump System (IBIDI GmbH) as described previously (27). Briefly, 0.5 × 106 HUVECs were cultured on gelatin-coated μ-slides 0.2 Luer (IBIDI GmbH) for 48 hours under slight flow (1 dyn/cm²). Consecutively, each slide was connected to the pump system filled with 2 mL HRBS and 2 mL tumor supernatant as indicated in the results section. The buffer contained 40% washed erythrocytes and fluorescence-labeled platelets (4 × 106 cells/mL). Perfusion was performed with a shear stress of 6 dyn/cm2 for 15 minutes. The endothelial layer was stained with CellTracker Blue (Thermo Fisher Scientific). Calcein red-orange (Invitrogen) labeled platelets bound to VWF strings were detected in real time using fluorescence microscopy equipped with a 20× objective (Observer.Z1, Zeiss).
Flow cytometry
The surface expression of TF and thrombomodulin (TM) on UCB cells was quantified by flow cytometry. Cells were harvested with Accutase (Sigma-Aldrich) and incubated with 1 μg of a primary mouse anti-human TF or mouse anti human TM (American Diagnostica) or the isotype-specific antibodies in HBRS + 1% BSA for 1 hour on ice. The cells were washed and incubated with a phosphatidylethanolamine-conjugated secondary antibody (1:50) for 30 minutes on ice in the dark. Consecutively, 10,000 cells were analyzed using FACS Canto II and FACS Diva Software (Becton Dickinson).
Thrombin generation assay
Thrombin activity was analyzed as described previously (20). Briefly, 2 × 106 UCB cells were suspended in lysis buffer (RayBiotech) and equal protein loads were ensured using a Bradford protein assay. Platelet-poor normal plasma (PPP) pooled from healthy volunteers was prepared by a two-step centrifugation process: 3,000 × g at 4°C for 15 minutes followed by 11,000 × g for 5 minutes. One well of a 96-well plate was filled with 50 μL of UCB cell lysate and 50 μL of 1:15 PBS-diluted PPP. After incubation for 2 minutes at 37°C, the coagulation reaction was started by adding 5 μL of 250 mmol/L CaCl2. The reaction was performed at 37°C and stopped after 15 minutes by adding 5 μL EDTA (saturated solution). Ultimately, hydrolysis of the thrombin-specific chromogenic substrate S-2238 (Chromogenix-Instrumentation Laboratory Spa) at a concentration of 0.2 mmol/L showed the generation and enzymatic activity of thrombin. Absorbance was measured once per minute at 405 nm, 37°C, for 60 minutes with a Microplate Reader (BioTek Instruments). The results were expressed as maximal rate of hydrolysis (Vmax). To control the specificity of the enzyme reaction, a goat anti-human TF polyclonal antibody (10 μg/mL; American Diagnostica) or Hirudin 10 U/mL (Pentapharm) was added.
Transcriptome analysis
To investigate whether UCB affects expression levels of coagulation-related proteins at the site of the tumor, we analyzed mRNA data from more than 400 patients with bladder cancer in two publicly available databases (28, 29). Gene transcription levels were correlated with overall survival status of the patients or T-status, depending on the data source.
VWF staining in patients with bladder cancer
Paraffin blocks were cut (3 μm sections) and sections were deparaffinized. Antigen retrieval was achieved using citric acid buffer (pH 6.0, 95°C for 40 minutes). Slides were treated with 1% hydrogen peroxide in methanol to suppress endogenous peroxidase activity. After blocking in 1% BSA, the slides were incubated for 30 minutes at 37°C with a polyclonal rabbit anti-human VWF antibody (Agilent-DAKO, dilution 1/1,000 in 1% BSA) followed by incubation with 2 μg/mL of biotinylated anti-rabbit IgG secondary antibody (R&D Systems) for 30 minutes at room temperature. Subsequently, the sections were stained using Standard Ultra-Sensitive ABC Peroxidase Staining Kit (Pierce/Thermo Fisher Scientific) and 3,3′-diaminobenzidine (DAB; Vector Laboratories), counterstained by hematoxylin, dehydrated, and mounted with a cover slip. To analyze vessel activation, vessel contours were marked separately for vessel lumen and vessel wall and VWF staining intensity per area was measured using ImageJ software. A ratio based on the staining intensity per area was calculated using the following equation: [(lumen − wall)/(lumen + wall)]. Values between 0 and 1.0 were considered as activated vessel showing more VWF in the lumen than in the vessel wall, values from 0 to −1.0 represented a quiescent vessel with the majority of VWF stored in the endothelium.
Statistical analysis
Results were expressed as mean ± SD of at least three independent experiments. Statistical analysis was performed with GraphPad Prism software and significance was proven with an unpaired Student t test or a Wilcoxon–Mann–Whitney Test, as appropriate (*, P ≤ 0.05; **, P ≤ 0.01).
Results
Procoagulatory profile of UCB cells
We performed qRT-PCR in four different urothelial cell lines (Fig. 1A; Supplementary Fig. S2A) to analyze the expression of genes potentially involved in the tumor–endothelium or tumor–platelet interaction (30–32). UROtsa cells are supposed to resemble the benign urothelium, whereas RT112, RT4, and T24 cells have been generated from transitional bladder carcinoma (33, 34).
Interplay between bladder cancer cells, plasmatic coagulation, and platelets. A, Expression of PDPN and TF. The three bladder cancer cell lines, RT4, T24, and RT112 have been compared with the benign UROtsa cells. Data were normalized to β-Actin and relative mRNA expression is shown on a logarithmic scale. Data are presented as mean ± SD, n = 3. *, P < 0.05; Student t test. B, The potential of the cell lines to induce platelet aggregation in platelet rich plasma was assessed by LTA measurements. Gray lines represent the SD. Data are shown as mean ± SD, n = 3. C, Direct physical interaction between 5-chloromethylfluorescein diacetate–labeled T24 and UROtsa cells (green) with washed and Calcein red-orange–labeled platelets (red). Scale bar, 20 μm.
Among the tested genes, we found only a markedly differential expression of TF and PDPN (Fig. 1A). High levels of TF were detected in T24 but also in the benign UROtsa cells. Moreover, UROtsa cell were characterized by the highest expression of PDPN. To prove the ability of the UCB cell lines to induce the plasmatic coagulation or to interact with platelets, we applied LTA (Fig. 1B). In-line with the high expression of PDPN, only UROtsa cells were able to induce platelet aggregation (Fig. 1B). LTA results were further confirmed by fluorescence microscopic analysis of mixed fluorescence-labeled platelets and tumor cells, showing that UROtsa cells offers many binding sites for platelets (Fig. 1C). High levels of PDPN have previously been found at high levels in healthy urothelium suggesting that PDPN expression may not contribute to the progression of UCB but to the physiologic hemostasis of the healthy bladder (35). Unexpectedly, also high levels of TF expression in T24 cells were not able to trigger the plasmatic coagulation and thus platelet aggregation in our LTA measurements. Although we verified functional TF in T24 by a chromogenic activity assay (Supplementary Fig. S2B), we detected high levels of the thrombin antagonist TM at the surface of T24 cells (Supplementary Fig. S2C). Because neither PDPN nor TF appeared to be clearly associated with UCB-mediated hypercoagulation we continued to analyze the impact of the UCB cell secretome on its ability to promote coagulation through the activation of endothelial cells.
Distinct UCB cell supernatants activate endothelial cells
We and others have previously shown that tumor cells secrete various molecules (e.g., VEGF-A) which could promote endothelial cell activation. Because endothelial cell activation is characterized by the release of VWF and soluble P-selectin (sP-selectin; ref. 36), we used ELISA to measure both compounds in the supernatants of HUVEC after exposure to UCB cell–conditioned medium. Treatment with RT4 supernatants induced the strongest endothelial cell activation (Fig. 2A and B). Immunofluorescence analysis further confirmed the RT4-mediated release of VWF as indicated by the formation of ultra-large VWF fibers and the elevated exposure of P-selectin at the cellular border (Fig. 2C). These results were further validated by microfluidic experiments mimicking the pathophysiologic condition of tumor blood vessels (37).
Activation of endothelial cells with tumor cell supernatants. HUVECs were treated for 15 minutes with the supernatants of the indicated UCB cell lines. Thrombin (0.5 U/mL) was used as a positive control, nonconditioned culture medium was used to measure the basal level of VWF release (Control). Endothelial cell activation was quantified in the supernatants of HUVECs by ELISA detecting the release of VWF (A) and sP-selectin (B). Data are presented as mean ± SD, n = 8. *, P < 0.05; **, P < 0.01; Student t test. C, Immunofluorescence detection of VWF (green) and P-selectin (red) in nonstimulated HUVECs (control) and HUVECs treated with supernatants of RT4 cells. Nuclei were stained in blue. Scale bars, 20 μm.
To this end, HUVECs were perfused with tumor cell–conditioned medium supplemented with physiologic concentrations of human platelets. Time-dependent trapping of platelets by VWF fibers was recorded by video microscopy and representative snap shots of the experiments are shown in Fig. 3A. Representative video files are presented in the Supplementary Data (Supplementary Video S1 and S2). The quiescent endothelium recruited negligible amounts of platelets (Fig. 3A, control), however stimulation of endothelial cells with RT4-derived supernatants triggered VWF secretion and consecutive typical platelet binding in a “beads-on-a-string-like” configuration (Fig. 3A, RT4). No relevant endothelial cell activation has been detected after stimulation with RT112, T24, or UROtsa cell–derived supernatants. To confirm the direct interaction between platelets and VWF fibers, we stained the flow chamber after the experiments by immunofluorescence. As indicated in Fig. 3B, VWF strings clearly colocalize with platelets in the RT4-treated group.
Microfluidic experiments mimicking tumor blood vessels. A, Representative snapshots taken during flow experiments. HUVECs were stimulated with nonconditioned medium (control) or supernatants of RT4 cells (RT4) and perfused with Calcein red-orange–labeled platelets at a shear rate of 6 dyn/cm². Platelet binding (white circular spot) was recorded for 10 minutes. Scale bar, 100 μm. Corresponding video files of the flow experiments are provided in the Supplementary Data. The VWF-mediated platelet binding has been quantified (number of VWF–platelet strings) to compare the procoagulant impact of supernatants harvested from RT4, T24, RT112, and UROtsa cells. Data are shown as mean ± SD, n = 3. *, P < 0.05. B, Binding of Calcein red-orange–labeled platelets (red) to VWF fibers has been confirmed by immunofluorescence staining of the microfluidic chambers after the flow experiment. Scale bar, 100 μm.
Endothelial VWF has previously been attributed to angiogenesis (38). We therefore further investigated whether the UCB cell–derived supernatants differ in their ability to induce tube formation in HUVECs. In agreement with their ability to induce VWF release, RT4 cells showed the strongest proangiogenic potential. T24 cells exhibited a trend toward increased tube formation, whereas supernatants from UROtsa and RT112 cells had no impact (Supplementary Fig. S3A). Immunofluorescence analysis further revealed that the tube forming endothelial cells are VWF depleted, suggesting a mechanistic linkage between VWF release and new blood vessel formation (Supplementary Fig. S3B).
Secretome profiling array
To determine which UCB cell–derived soluble factors account for the acute endothelial cell activation, the conditioned medium of RT4, T24, RT112, and UROtsa cells was analyzed by a proteome profiling microarray. The selected array allows the relative quantification of 46 proteins involved in angiogenesis, tissue inflammation, and vascular remodeling (Fig. 4A). The calculated heatmap indicates that the cell lines RT4 and T24 released the highest levels of proangiogenic and proinflammatory factors in contrast to the RT112 and UROtsa cells. To identify possible candidates responsible for the procoagulant activation of the endothelium, we correlated the data of the VWF ELISA (Fig. 2A) with the microarray results. The corresponding correlation coefficients (r2) are listed in Fig. 4B. Highest correlations were found with VEGF-A, IL8, and tissue inhibitor of metalloproteinases 1 (TIMP-1), suggesting their potential contribution to the endothelial cell activation. We confirmed the secretome data of IL8 and VEGF-A in the conditioned cell media by ELISA (Fig. 4C and D).
Characterization of the UCB cell supernatants. A, The secretome of UCB cells was analyzed by a microarray probing 46 proteins. T24 cells released considerable amounts of different proangiogenic and proinflammatory cytokines. RT4 cells secrete less proinflammatory cytokines but highest levels of VEGF-A. RT112 and UROtsa did not show a pronounced secretory activity apart from MMP-1 and TIMP-1/-2. High protein levels are shown in red, low protein levels are shown in blue. B, Levels of the secreted proteins were aligned with the ability of the UCB cell supernatants to induce the release of procoagulant VWF from HUVECs (Fig. 2). The correlation coefficient, r², suggests IL8, TIMP-1, and VEGF-A as potential procoagulant endothelial cell activators. Verification of IL8 (C) and VEGF-A (D) levels by ELISA in UCB cell supernatants.
VEGF-A inhibition blocks VWF-mediated platelet binding
Through the correlation of our proteome data with the ability of the UCB cells to induce the release of VWF, we identified IL8, TIMP-1, and VEGF-A as possible inductors of the procoagulant endothelial cell activation. Because TIMP-1 tunes the activity of matrix metalloproteinases (MMP) and thus matrix remodeling rather than a direct endothelial cell activation, we focused in the next set of experiments on the procoagulant activity of IL8 and VEGF-A. Recombinant human IL8 failed to elicit VWF secretion from HUVECs. Moreover, the antibody-based neutralization of IL8 in the RT4 supernatant was ineffective to prevent endothelial cell activation (Fig. 5A). In contrast, recombinant human VEGF-A–induced VWF release and blockage of RT4-derived VEGF-A by the VEGF-A–neutralizing antibody bevacizumab completely prevented VWF exocytosis. In a comparable manner, also the exposure and the release of sP-selectin after VEGF-A modulation was affected (Fig. 5B). To prove the impact of VEGF-A on the extent of platelet adhesion to the endothelium, we repeated our microfluidic experiments in the presence of bevacizumab. In-line with the ELISA experiments, RT4-induced VWF-mediated trapping of platelets to the endothelium was completely abolished upon VEGF-A blockage (Fig. 5C).
VEGF-A inhibition impedes UCB cell–induced procoagulant activation of endothelial cells. A, In comparison with thrombin treatment (0.5 U/mL), HUVECs were not stimulated by recombinant IL8 (2 ng/mL). Blockage of IL8 in the supernatants of RT4 by neutralizing antibodies (0.72 μg/mL) or isotype control (+iso) failed to prevent the release of VWF. Nonconditioned medium was used as a negative control (Control). Thrombin (0.5 U/mL) was used as positive control. Data are shown as mean ± SD, n = 3. B, Stimulation of HUVECs with recombinant VEGF-A (10 ng/mL) induced a strong secretion of VWF and sP-selectin. Neutralization of VEGF-A in the supernatants of RT4 by bevacizumab (Beva, 0.65 μg/mL) blocked the procoagulant activation of HUVECs. Nonconditioned medium was used as a negative control (Control). Thrombin (0.5 U/mL) was used as positive control. Data are shown as mean ± SD, n = 3. **, P < 0.01; Student t test. C, Representative images of snapshots taken during flow experiments. HUVECs were perfused (shear rate 6 dyn/cm²) with RT4-conditioned medium containing physiologic concentrations of Calcein red-orange–labeled platelets (white circular spots). Quantification of the flow experiments indicates a reduced platelet binding (rel. platelet adhesion) in the presence of bevacizumab. Data are shown as mean ± SD, n = 3. *, P < 0.05; Student t test. Scale bar, 100 μm.
Blood vessel occlusion in patients with UCB and transcriptome analysis
Our in vitro experiments suggest that hypercoagulation in patients with bladder cancer is related to the release of VWF from the activated endothelium. To confirm this finding, we analyzed the secretion of endothelial VWF into the lumen of blood vessels in human UCB tissues (Fig. 6A and B; Supplementary Fig. S4). Overall, 200 vessels from high grade (n = 5 patients), 200 vessels from low grade (n = 5 patients), and 160 vessels from healthy control tissue (n = 5 patients) were analyzed. To quantify the magnitude of the blood vessel occlusion in relation to a potential activation of the endothelium, a ratio between the luminal VWF and the vessel wall–retained VWF (nonreleased VWF) was calculated. The quantification revealed a significant higher proportion of occluded blood vessels in high grade tumors compared with low grade tumors and healthy controls. As expected, healthy controls had the smallest proportion of occluded vessels (Fig. 6C). In our previous research, we have found a comparable relationship between luminal and nonreleased endothelial VWF in tissue sections of murine malignant melanoma (39).
VWF-mediated blood vessel occlusion in patients with bladder cancer. A, Peritumoral blood vessel (high grade UCB) occluded by large intraluminal VWF fiber networks (arrow). Scale bar, 100 μm. B, In healthy control tissue, VWF was mostly stored in endothelial cells (arrowhead) indicating nonactivated endothelial cells. Scale bar, 100 μm. C, Quantitative analysis of vessels in healthy tissue (n = 160), in low grade tumors (n = 200), and in high grade tumors (n = 200) revealed significantly more activated blood vessels in the peritumoral area than in healthy tissue, with high grade tumors showing more endothelial cell activation than low grade tumors. *, P < 0.05; Wilcoxon–Mann–Whitney test. D, MMP9 and HPSE have been shown to control the bioavailability of VEGF-A within the microenvironment of tumors. Both enzymes were significantly upregulated in deceased bladder tumor patients suggesting also increased levels of VEGF-A. In contrast, no or only a weak connection between TF or PDPN expression and the patients' outcome was found. mRNA expression is shown on a logarithmic scale. Further transcriptome data are depicted in the Supplementary Data.
To further validate the histologic findings, we analyzed public transcriptome data of more than 400 patients with bladder cancer (28, 29). In total, we correlated 17 pro- and anticoagulatory genes with the patients' overall survival (Supplementary Table S1). In agreement with our in vitro results, neither PDPN nor TF expression appeared to be strongly related to the patients' survival (Fig. 6D). Decreased survival was also not connected to elevated VEGF-A levels (Supplementary Table S1). However, high levels of the matrix remodeling enzymes MMP-9 and HPSE, which were both known to control the availability of VEGF-A in tissues (40, 41), were significantly associated with a reduced overall survival (Fig. 6D).
Discussion
In this study, we compared the three malignant urothelial cell lines RT4, T24, and RT112 and the benign UROtsa cell line. The descriptive and functional analysis focused on procoagulatory factors that have previously been attributed to tumor-associated hypercoagulation and metastasis. Our findings were further validated by public transcriptome data and biopsies of human patients with UCB using IHC.
In the first set of experiments, we have analyzed the potential impact of cell surface–exposed TF and PDPN on their ability to induce coagulation. Both membrane receptors have been shown to either activate the plasmatic coagulation or to interact with platelets (18, 42–44). Interestingly, none of the malignant cancer cells were able to induce blood clotting, although significant amounts of TF were expressed by T24 and RT112 cells. LTA measurements indicate that only the benign UROtsa cell line had a strong ability to promote platelet aggregation, which correlates with a high PDPN expression. In-line with our results, PDPN is found at high levels in nonmalignant tissues including prostate and urinary bladder tissues, as well as stromal cells underneath the vascular endothelium (45). Lack of PDPN expression favors bleedings, suggesting a physiologic role of PDPN in hemostasis (46). However, overexpression of PDPN in bladder cancer cells was shown to promote pulmonary metastasis in mice (35). The prometastatic properties of PDPN have been explained by the formation of tumor cell platelet clusters upon intravenous injection of the cancer cells, which increases the probability of the tumor cells to be trapped in the microcirculation of the lung. To clarify the impact of PDPN on the progression of bladder cancer cells, further studies using animal models with spontaneously metastasizing bladder cancer are required.
Interestingly, also high levels of TF in T24 or RT112 cells were not sufficient to induce plasmatic coagulation. This might be at least partially explained by comparable high levels of TM at the surface of the UCB cells. The membrane protein TM strongly binds thrombin. While TM-bound thrombin has a reduced affinity to fibrinogen it can efficiently activate the anticoagulant protein C (47, 48). Our results indicate that the expression of TF in bladder cancer cells is not sufficient to mediate coagulation. Next to tumor cells, also activated monocytes or endothelial cells could produce procoagulant TF, which may contribute to hypercoagulation in patients with tumor (49, 50). Therefore, we suggest that the pathophysiologic impact of monocyte and endothelial cell–derived TF and tumor cell–derived proinflammatory molecules that trigger the expression of TF should to be considered in future research. Indeed, our secretome analysis indicate that especially T24 cells express high levels of proinflammatory cytokines such as GM-CSF, IL6, or CCL-5, which may promote procoagulatory activation of immune and endothelial cells (51, 52). Moreover, CCL-5 triggers the production of platelets in the bone marrow (53). The related thrombocytosis may further increase to probability of VTE in patients with cancer independent of the tumor cell–derived TF. Hypercoagulation in patients is not compulsorily connected to single cancer cell–derived factors but the result of a complex and systemic interplay between various procoagulant molecules and a large range of different cells. Therefore, the investigation of isolated molecular aspects in vitro may provide only limited insights into the relationship of malignancy and tumor cell–related procoagulant molecules such as TF or PDPN. However, in-line with our in vitro findings which suggest a limited impact of both molecules, also the transcriptome analysis of UCB tissues indicated an only weak correlation between patients' survival and expression of TF or PDPN.
In further experiments, we analyzed the ability of tumor cell–derived supernatants to induce an acute and procoagulatory endothelial cell activation associated with the exocytosis of Weibel–Palade bodies and thus the release of the procoagulant VWF and the exposure of P-selectin. We discovered that RT4 cells have the strongest ability to promote endothelial cell activation. Microfluidic experiments further indicated that only RT4 cells were able to induce VWF-related platelet aggregation at the endothelial interface. By correlating the ability to induce VWF release with the secretome profile of each cell line, we identified IL8, TIMP-1, and VEGF-A to potentially trigger procoagulant activation of endothelial cells. Because TIMP-1 might not directly contribute to endothelial cell activation, we focused on IL8 and VEGF-A. Antibody-based neutralization of either IL8 or VEGF-A or stimulation with recombinant proteins suggests that VEGF-A is responsible for Weibel–Palade body exocytosis. We confirmed this finding in microfluidic experiments showing that the VEGF-A–blocking antibody bevacizumab prevented VWF-mediated platelet aggregation. In clinical studies it could be demonstrated that high serum VEGF-A levels were a significant predictor of overall and cancer-specific survival in patients with bladder cancer (54). In a mouse model with xenografted UCB cells, it was shown that tumor-infiltrating CD4+ cells induce VEGF-A release through an IL1, androgen receptor, HIF1α axis leading to enhanced metastasis (55). In our in vitro study we investigated the potential impact of tumor cell–derived VEGF-A on the procoagulant activation of endothelial cells. However, in tumor tissues, VEGF-A is not only produced by the tumor cells but also at high levels by tumor-associated macrophages and neutrophils (56). In addition, it has been shown that tumor-infiltrating monocytes, tumor-associated macrophages, and especially tumor-associated neutrophils secret considerable amounts of MMP-9 into the tumor microenvironment, which further promote the liberation of VEGF-A from the extracellular matrix (57–61). Moreover, we and others have previously reported that in melanoma and myeloma MMP-9 together with HPSE contribute to the release of VEGF-A through the hydrolysis of proteoglycans and the cutting of associated heparan sulfate chains (40, 62). Indeed, transcriptome analysis showed that the VEGF-A–related hydrolytic enzymes MMP-9 and HPSE were significantly connected to a decreased patients' survival. Of note, the mRNA expression of MMP9 or HPSE may not directly correlate with the effective hydrolytic activity of the enzymes within the tumor. Therefore, further data on their tissue distribution, quantity, and activity may give relevant insights into the procoagulant connection between MMPs, HPSE, and VEGF-A in UCB.
Procoagulant activation in bladder cancer tissue was indicated in our study by the release of VWF into the lumen of the tumor blood vessels and the associated blood vessel occlusion, which correlated with the tumor grade of the patients with UCB. In accordance with our results, higher VWF antigen and P-selectin levels have been measured in the blood plasma of patients with UCB with invasive or lymph node–positive disease in comparison to superficial or localized bladder cancer (22, 63). Comparable findings have been made in colorectal carcinoma or malignant melanoma underlining that VWF is essential for cancer cell dissemination (64).
In conclusion, our data demonstrate that UCB cell–derived VEGF-A is able to induce a prothrombotic intravascular micro-milieu by exerting endothelial cell activation. The release of VWF from these procoagulant endothelial cells promoted platelet binding and was associated with blood vessel occlusions in UCB tumor tissues. Future studies are required to prove whether inhibition of VEGF-A prevent VTE in patients with bladder cancer and whether this correlates with a reduced tumor progression.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: C. Bolenz, C. Gorzelanny
Development of methodology: J.R. Robador, C. Gorzelanny
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. John, J.R. Robador, S. Vidal-y-Sy, P. Houdek, E. Wladykowski, C. Günes
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. John, S.W. Schneider, A.T. Bauer, C. Gorzelanny
Writing, review, and/or revision of the manuscript: A. John, C. Günes, C. Bolenz, S.W. Schneider, C. Gorzelanny
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Vidal-y-Sy, P. Houdek, E. Wladykowski, C. Bolenz, S.W. Schneider
Study supervision: C. Bolenz, C. Gorzelanny
Acknowledgments
We thank Annette Steidler for excellent technical assistance and Dr. Cleo-Aron Weis for supervision of the histologic analysis. We also thank Birgit Schneider for the schematic drawing. This research was funded by the German Society of Urology (DGU; Ferdinand Eisenberger grant), grant number JoA1/FE-16.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Footnotes
Note: Supplementary data for this article are available at Molecular Cancer Research Online (http://mcr.aacrjournals.org/).
Mol Cancer Res 2020;18:1099–109
- Received October 21, 2019.
- Revision received January 15, 2020.
- Accepted March 26, 2020.
- Published first March 31, 2020.
- ©2020 American Association for Cancer Research.
References
Volume 18, Issue 7
