Murine TNFα was from Chemicon International (now part of Millipore Corporation, Billerica, MA, USA), human TNFα was from ReliaTech (Wolfenbüttel, Germany), collagen was from Nycomed (now part of Takeda Pharmaceuticals International, Zürich, Switzerland), and anti-human tissue factor-FITC, anti-human P-selectin-RPE, and respective negative controls were from AbD Serotec (Oxford, UK). RNeasy Mini Kit, small interfering RNA (siRNA) against TNFR1 and TNFR2, and Effectene transfection kit were from Qiagen (Hilden, Germany), and predesigned glyceraldehyde-3-phosphate dehydrogenase (GAPDH), TNFR1, TNFR2, plasminogen activator inhibitor 1 (PAI-1), tissue factor, and thrombomudulin TaqMan primers were from Applied Biosystems (Foster City, CA, USA). All other chemicals were from Sigma-Aldrich (Munich, Germany).

Animal experiments were performed in wild-type (WT) or TNFα receptor-deficient C57BL/6 mice. WT mice were purchased from Charles River (Sulzfeld, Germany). TNFR1^−/− and TNFR2^−/− mice were originally obtained from the Jackson Laboratory (Bar Harbor, ME, USA), and TNFR1^−/−R2^−/− mice were subsequently generated by cross-breeding. Surgical procedures were performed under short-term anesthesia induced by a single intraperitoneal injection of midazolam 3 mg/kg (Ratiopharm, Ulm, Germany), fentanyl 0.03 mg/kg (CuraMED Pharma, Baden-Württemberg, Germany), and medetomidinhydrochloride 0.3 mg/kg (Pfizer, Berlin, Germany; produced by Orion Pharma, Espoo, Finland) diluted in 0.9% NaCl. After the experiments, the animals were killed by injection of an overdose (2 g/kg) of sodium pentobarbital (Merial, Hallbergmoos, Germany). All experiments were conducted in accordance with the German animal protection law and approved by the district government of Upper Bavaria (approval reference number AZ 55.2-1-54-2531-162-08). The investigation conforms to Directive 2010/63/EU of the European Parliament.

The dorsal skinfold chamber microcirculatory model was used in mice as described previously 23. Animals with an intact microcirculation underwent carotid artery catheterization for application of drugs or injection of isolated platelets, respectively. Intravital fluorescence microscopy was performed by using a modified microscope (Zeiss Axiotech Vario; Carl Zeiss, Oberkochen, Germany). Images were recorded with a digital camera (AxioCam HSm; Carl Zeiss). For all in vivo experiments, TNFα was administered via carotid artery catheter at a dose of 0.4 µg/kg. This was calculated to match plasma levels of approximately 5 ng/mL, which caused effects in vitro.

Intravital thrombotic vessel occlusion time was assessed in arterioles of WT or TNFα receptor-deficient C57BL/6 mice in the dorsal skinfold chamber model. For induction of intra-arteriolar thrombosis, the ferric chloride superfusion method was used as described previously 2324. Before the experiments, blood vessel flow was digitally recorded and regular blood flow was confirmed for all analyzed arterioles. To visualize vessel lumina before vessel injury, 50 µL of a 5% fluorescein isothiocyanate-labeled dextran solution (FITC-dextran, molecular weight 150,000) was infused via the carotid catheter. Injury to the vascular wall was then performed by application of 30 µL of a ferric chloride solution (25 mmol/L) onto arterioles by using a standardized protocol, and movies were recorded until blood flow ceased.

Whole blood was drawn from anesthetized mice by cardiac puncture. To prevent blood from clotting, syringes contained 10% of sodium citrate. The citrated whole blood was spun at 130g, and the obtained platelet-rich plasma (PRP) was incubated with carboxyfluorescin (carboxyfluorescein diacetate succinimidyl ester 17 µmol/L; Bachem, Bubendorf, Switzerland) in the dark for 30 minutes. Labeled platelets were then spun at 340g and resuspended in a buffered calcium-free physiologic solution (138 mmol/L NaCl, 2.7 mmol/L KCl, 12 mmol/L NaHCO₃, 0.4 mmol/L NaH₂PO₄, 1 mmol/L MgCl₂ × 6 H₂O, 5 mmol/L D-glucose, 5 mmol/L Hepes; pH 7.35). For centrifugation, iloprost (10 ng/mL; Schering, Berlin-Wedding, Germany) was added to prevent platelet activation. The ability of the isolated and stained platelets to aggregate was tested by platelet aggregometry.

For intravital studies of platelet interaction with the intact vessel wall, isolated and fluorescent-stained murine platelets from a donor animal were injected via a carotid artery catheter and observed in the dorsal skinfold chamber model. Movie sequences of 30 seconds in four- to six-vessel segments in each animal were recorded and analyzed by using AxioVision Software (Carl Zeiss). Vessels with abnormal flow were excluded from analysis. From the resulting length of the platelet trace in single images, velocities of single platelets were calculated by using the exposure time of each single picture. PVWI was expressed in frequency histograms consisting of all platelet velocities analyzed. Histograms were normalized to the maximum platelet speed within a vessel to exclude biasing influences of altered blood flow velocities between different arterioles. As a consequence, a rightward shift in platelet velocity distribution within a histogram expresses less PVWI, whereas a leftward shift signalizes increased PVWI at the arteriolar wall. Platelets with less than 5% of the velocity of the fastest platelets were defined as rolling platelets.

Platelet aggregation was measured by using the turbidimetric method described by Born 25. Human or murine PRP was obtained by centrifugation (130g) of whole citrated blood drawn from human cubital veins or by cardiac puncture in mice. ADP-, collagen-, or thrombin receptor-activating peptide (TRAP)-induced platelet aggregation was measured photometrically by using a two-channel aggregometer (ChronoLog 490-2D; Chrono-log Corporation, Havertown, PA, USA) under continuous stirring at 1,000 revolutions per minute at 37°C. Written consent was obtained from platelet donors. To assess the effects of endothelial-derived cytokines, 100 µL of supernatant of stimulated human microvascular endothelial cells (HMECs) was added to 300 µL of PRP and incubated for 90 minutes before agonist-induced aggregation was measured.

Human umbilical vein endothelial cells (HUVECs) were isolated and cultured as described previously 26. The procedure was approved by a university ethics review board. HMECs were provided by Ades and colleagues 27 and cultured in M199 media supplemented with 10% fetal calf serum, 10% endothelial growth media (PromoCell, Heidelberg, Germany), and 1% penicillin/streptomycin. The investigation conforms to the principles outlined in the Declaration of Helsinki.

For superoxide measurements, the cytochrome c reduction method was used as previously described 28. The O₂^--dependent part of cytochrome c reduction was calculated from the difference in absorbance (550 nm) between samples incubated with or without superoxide dismutase.

HMECs or HUVECs were grown as described and incubated with sham or TNFα (5 ng/mL) as indicated. Cells were stained by using anti-p-selectin-RPE and anti-tissue factor-FITC or corresponding RPE- or FITC-labeled negative control. For measuring, a FACSCanto II flow cytometer (Becton, Dickinson and Company, Franklin Lakes, NJ, USA) was used. Data were analyzed by using FACSDiva software (Becton, Dickinson and Company).

HMECs were incubated with sham or TNFα for indicated time points. RNA isolation and real-time polymerase chain reaction (PCR) were performed as described previously 29. Commercially available pre-developed TaqMan reagents were used for the human target genes thrombomodulin, PAI-1, tissue factor, TNFR1, and TNFR2, and GAPDH was used as a reference housekeeping gene. All measurements were performed in duplicates.

Cells were stimulated with TNFα as indicated and then lysed with 15 mM n-octyl-D-glycopyranosidase in 0.1 M imidazol buffer; 20 μL of cell lysate and 20 μL of 200 mmol/L CaCl₂ for re-calcification were added to 300 μL of citrated (3.13% sodium citrate) human whole blood from healthy volunteers, and clotting time was measured by thrombelastometry (Rotem; Tem Innovations, Munich, Germany). Stimulation with human recombinant tissue factor was used as a positive control.

To assess the role of TNFα-induced autocrine factors released by endothelial cells, cells were stimulated with TNFα 5 ng/mL for 3 hours. After removal of the TNFα-containing medium and a washing step, fresh medium without TNFα was added to the cells to allow secretion of prothrombotic factors. After 2 hours, this medium was used to stimulate untreated cells in which several parameters then were measured.

Endothelial cells were transfected with 50 nM siRNA against human TNFR1 and TNFR2 by using the magnetofection method in combination with the Effectene kit from Qiagen as previously described 30. Transfected cells were left 48 hours to allow siRNA-mediated knockdown, which was confirmed by real-time PCR and Western blot, which was performed as described elsewhere 30.

SigmaStat Software (SYSTAT, Chicago, IL, USA) was used to calculate statistical differences. Data were analyzed by using Student t test for normally distributed variables or the Mann-Whitney rank sum test, when normal distribution was not given. Data are expressed as mean ± standard error of the mean. Differences were considered significant when the error probability level was a P value of less than 0.05.

To analyze the effect of TNFα on thrombus formation in vivo, we assessed the time to thrombotic arteriolar vessel occlusion following injury by ferric chloride superfusion to the vascular wall. The time to complete thrombotic vessel occlusion upon injury was significantly accelerated from 260 ± 31 seconds in control animals to 158 ± 31 seconds in animals treated with TNFα over a period of 4 hours prior to the experiment (n = 8, P <0.05 versus control animals). Vessel occlusion time was not significantly accelerated compared with control animals when TNFα was given only half an hour prior to the experiment (241 ± 54 seconds; n = 5; Figure 1a).

To investigate the contribution of signaling via the TNFα receptor subtypes, we measured time to thrombotic vessel occlusion in TNFR1^−/−, TNFR2^−/−, and TNFR1-/R2^−/− mice. The ability of TNFα to accelerate arteriolar thrombus formation in vivo observed in WT mice (Figure 1a) was further enhanced in TNFR1^−/− mice, in which the mean vessel occlusion time after treatment with TNFα was reduced to 62 ± 13 seconds (n = 5, P <0.05 versus sham-treated TNFR1^−/−; Figure 2a), corresponding to a TNFα-dependent relative acceleration of thrombotic vessel occlusion time of 37% ± 12% in WT mice and of 71% ± 6% in TNFR1^−/− mice (n = 5 to 8; P <0.05 versus WT treated with TNFα; Figure 2b). In mice lacking TNFR2 or both TNFR1 and TNFR2, TNFα did not significantly accelerate thrombus formation in vivo. Without TNFα treatment, time to thrombotic vessel occlusion was not significantly affected in either of the TNFα receptor-deficient animals (Figure 2a).

Next, we assessed transient interaction of platelets to the vessel wall in vivo by intravital microscopy of vessels in the dorsal skinfold chamber. Following systemic TNFα treatment, interaction of injected labeled platelets with the endothelium was slightly enhanced as indicated by a leftward shift in platelet flow velocities and a decrease in the median platelet velocity from 4.0 mm/second in control animals (n = 3,294 platelets, 5 different animals) to 3.7 mm/second in TNFα-treated animals (n = 3,239 platelets, 5 different animals, P <0.05 versus control animals). The fraction of rolling platelets (platelets with less than 5% of the blood flow velocity) was only 0.1% ± 0.1% of all analyzed injected platelets in control animals and not significantly higher in TNFα-treated animals, in which it was 0.4% ± 0.6% (n = 5; data not shown).

In contrast, the leftward shift in platelet velocity pattern seen in WT mice after treatment with TNFα was much more prominent in TNFR1^−/− mice, indicating more interaction of platelets with the endothelium (Figure 3a). Moreover, TNFα increased the fraction of rolling platelets from 0.4% ± 0.3% in WT mice to 4.1% ± 1.0% in TNFR1^−/− mice (n = 5, P <0.01 versus WT; Figure 3b). In TNFα-treated TNFR2^−/− animals, the fraction of rolling platelets was not significantly different compared with WT mice (1.1% ± 0.4%). The maximum platelet velocities in the analyzed vessels as an approximate measure of flow velocity were not significantly different before and after TNFα treatment or between the genotypes.

To test for direct effects of TNFα on platelets, we assessed platelet aggregation in vitro in human PRP by using light transmission aggregometry (Born's method). Incubation of PRP with TNFα 5 ng/mL for different time intervals did not affect the ability of platelets to form aggregates upon stimulation with several commonly used platelet agonists - ADP (5 µmol/L), collagen (5 to 10 µg/mL), or TRAP (5 µmol/L) - either after a short time of incubation or after 4 hours (n = 8; results for ADP in Figure 4a). TNFα itself did not cause platelet aggregation either. Also, the addition of supernatant of TNFα-treated endothelial cells (with or without siRNA knockdown of TNFR1 or TNFR2) to PRP had no influence on platelet aggregation (n = 4; Figure 4b).

To check whether direct activation of platelets is responsible for prothrombotic TNFα effects in TNFR1^−/− mice, we performed platelet aggregation studies ex vivo in PRP from TNFR1^−/− mice. Like in human PRP, platelets from these animals showed no differences in the ability to aggregate with or without prior TNFα incubation (5 ng/mL) to different stimuli, including ADP and collagen, at different stimulation times (n = 5; Figure 4c).

To better define the relative contribution of the endothelium to TNFα-dependent prothrombotic effects, we analyzed pro- and anti-coagulant factors in primary human endothelial cells (HUVECs) and HMECs. We first measured the release of endothelial superoxide (O₂^-) upon TNFα stimulation in HUVECs by using the cytochrome c assay. The photometrically recorded superoxide dismutase-sensitive fraction of cytochrome c reduction, expressed as the difference in the absorbance, depicts the production of O₂^- and was 2.2 ± 0.6-fold increased after stimulation with TNFα for half an hour in a dose of 5 ng/mL (n = 15, P <0.05 versus control; Figure 5a).

Next, we used flow cytometry to measure the adhesion molecule p-selectin, which is described to mediate platelet rolling on the endothelium 31. Whereas p-selectin was barely detectable on the surface membrane of HUVECs either non-stimulated or stimulated with TNFα (5 ng/mL) for only 30 minutes, it was significantly upregulated after 4 hours, with an increase in the mean relative fluorescence of 50.4% ± 13.7% (P <0.05, n = 5; Figure 5b).

Thrombomodulin expression was found to be suppressed by 79.8% ± 2.4% upon treatment with TNFα 5 ng/mL for 4 hours (P <0.05, n = 8; Figure 5c). Furthermore, PAI-1, one of the most potent inhibitors of fibrinolysis, and tissue factor (CD142), which plays a crucial role in the activation of the extrinsic pathway of the coagulation cascade, were measured by real-time PCR in HMECs upon stimulation with TNFα for 30 minutes or 4 hours at 5 ng/mL. After 4 hours, mRNA levels of PAI-1 were more than fivefold higher compared with baseline (fold increase 5.3 ± 0.5; P <0.05, n = 16; Figure 5d), and both tissue factor expression and tissue factor protein on the surface membrane of HMECs were significantly upregulated after 4 hours of treatment with TNFα 5 ng/mL (11.6 ± 2.0-fold and 3.4 ± 0.3-fold increase versus baseline, respectively, P <0.05, n = 8 to 10; Figure 5e). Ultimately, in an endothelial-dependent clotting assay in whole blood, clotting time upon induction with HMEC lysates after stimulation with TNFα (5 ng/mL for 4 hours) was accelerated from 12.5 ± 0.3 minutes to 5.5 ± 0.4 minutes (P <0.05, n = 4; Figure 5f).

To test whether endothelial cells stimulated with TNFα secrete mediators which contribute to TNFα-dependent prothrombotic effects, HUVECs or HMECs were incubated for 30 minutes (for p-selectin) and 4 hours (for thrombomodulin, PAI-1, and tissue factor expression) with supernatants of cells previously stimulated with TNFα at 5 ng/mL for 3 hours. Whereas p-selectin was slightly but not significantly increased (n = 6; Figure 5g), expression levels of thrombodulin were reduced by 44% ± 12% (P <0.05, n = 4; Figure 5h), and mRNA levels of PAI-1 and tissue factor were significantly increased (3.3 ± 03-fold and 9.9 ± 0.8-fold increase versus baseline, respectively, P <0.05, n = 4; Figure 5i,j).

To define the relative contribution of the TNFR subtypes to induction of single genes relevant in thrombosis, knockdowns of TNFR1 and TNFR2 were performed in HMECs and HUVECs by using siRNA. As tested by real-time PCR and Western blot after 24 and 48 hours, respectively, for both receptors, a knockdown of about 55% could be achieved (Figure 6j).

Whereas upregulation of p-selectin was inhibited only by knockdown of TNFR2 (Figure 6a), TNFα-dependent changes in thrombomodulin and PAI-1 expression were significantly reduced by TNFR1 knockdown (Figure 6b,c). TNFα-induced upregulation of tissue factor was significantly reduced by knockdown of both TNFR subtypes (Figure 6d).

Thrombomodulin suppression induced by the supernatant of TNFα-stimulated HMECs was abolished after TNFR2 knockdown (n = 4; Figure 6f), and p-selectin, PAI-1, and tissue factor upregulations induced by soluble mediators released upon TNFα stimulation were not changed after knockdown of either of the TNF receptor subtypes (n = 4; Figure 6e,g,h). Knockdown of both TNFR1 and TNFR2 significantly inhibited endothelial-dependent TNFα-induced acceleration of blood clotting (P <0.05, n = 4; Figure 6i).