Activation of the Coagulation Mechanism on Tumor Necrosis Factor-stimulated Cultured Endothelial Cells and Their Extracellular Matrix THE ROLE OF FLOW AND FACTOR IX/IXa*

Infusion of tumor necrosis factor (TNF) into tumor- bearing mice led to intravascular clot formation with fibrin deposition in microvessels in the tumor bed in close association with the vessel wall, which could be prevented by active site-blocked factor IXa (IXai). This observation prompted us to examine the role of the intrinsic in activation of the mech- anism on TNF-stimulated human endothelial cell and during to factors or

Infusion of tumor necrosis factor (TNF) into tumorbearing mice led to intravascular clot formation with fibrin deposition in microvessels in the tumor bed in close association with the vessel wall, which could be prevented by active site-blocked factor IXa (IXai). This observation prompted us to examine the role of the intrinsic system in activation of the coagulation mechanism on TNF-stimulated human endothelial cell monolayers and endothelial-derived matrix during exposure to purified coagulation factors or flowing blood. Treatment of endothelial cells in intact monolayers with TNF induced expression of the procoagulant cofactor tissue factor (TF) in a dose-dependent manner, and after removal of the cells, TF was present in the matrix. TNF-treated endothelial cell monolayers exposed to blood anticoagulated with low molecular weight heparin induced activation of coagulation. Addition of IXai blocked the procoagulant response on TNF-treated endothelial cells, and consistent with this, the presence of factor IX/VIIIa enhanced endothelial TF/factor VII(a) factor X activation over a wide range of cytokine concentrations (0-600 p~) .
When TF-dependent factor X activation on endothelial cells was compared with preparations of subendothelium, the extracellular matrix was 10-20 times more effective. IXai blocked TF/factor VII(a) mediated activated coagulation on matrix, but only at lower concentration of TNF (e50 p~) .
Similarly, enhancement of factor Xa formation on matrix by factors IX/VIIIa was most evident at lower TNF concentrations. When anticoagulated whole blood flowing with a shear of 300 s" was exposed to matrices from TNF-treated endothelial cells, but not matrices from control cells, fibrinopeptide A (FPA) generation, fibrin deposition, and platelet aggregate formation were observed. FPA generation could be prevented by a blocking antibody to TF and by active site-blocked factor Xa (Xai) over a wide range of TNF concentrations (0-600 PM), whereas IXai only blocked FPA generation at lower TNF concentrations (e50 PM). Activation of coagulation on matrix from TNF-stimulated endothelial cells was dependent on the presence of platelets, indicating the important role of platelets in propagating the reactions leading to fibrin formation. These observations demonstrate the potential of cytokine-stimulateed endothelium and their matrix to activate coagulation and suggest the importance of the intrinsic system in factor Xa formation on cellular surfaces.
Tumor necrosis factor/cachetin (TNF)' is a central mediator of the host response in diverse conditions ranging from Gram-negative sepsis to ischemia/reperfusion syndrome (1-3) in which abnormalities of coagulation, both localized and generalized, are a common component. TNF has been shown to induce the synthesis and expression of the procoagulant cofactor tissue factor (TF) by endothelium (4, 5) suggesting a mechanism whereby this cytokine can activate cellular clotpromoting mechanisms. Because fibrin formation is often localized to particular vascular beds, we have focussed on the role of endothelial cell TF in the activation of coagulation. TF, the major initiator of coagulation in vivo, binds factor VII/VIIa and promotes activation of factors IX and X (6-8). Although in many in vitro experimental systems employing purified proteins and phospholipids the activation rate of factor X exceeds that of factor IX, suggesting that factor IX could be bypassed, in vivo factor IX appears to have a central role in the hemostatic mechanism (8). This has led us to speculate that on cellular surfaces alternative mechanisms might be operative, providing insights into the possible role of factor IX/IXa in TF-mediated activation of coagulation.
Previous studies of TNF-treated endothelial monolayers have demonstrated the presence of both T F and a binding site for factor IX/IXa, the latter promoting assembly of the intrinsic factor X activation complex on the cell surface (9-11). Therefore, we considered it likely that on endothelium, factor IX/IXa could contribute to activation of coagulation initiated by TF. In the current study, we demonstrate that active site-blocked factor IXa (IXai) does inhibit formation of intravascular thrombi in tumor-bearing animals after the infusion of TNF. In this context, TF/VII-mediated factor Xa ' The abbreviations used are: TNF, tumor necrosis factor; TF, tissue factor; LMWH, low molecular weight heparin; RIA, radioimmunoassay; BSA, bovine serum albumin; HEPES, N-2-hydroxyethylpiperazine-N'-2-ethanesulfonic acid; FPA, fibrinopeptide A. 12067 formation on intact endothelial monolayers incubated with TNF was enhanced by the addition of factor IX/.VIII, whereas IXai blocked the activation of coagulation when anticoagulated blood was incubated with the stimulated endothelial cells. Unexpectedly, there was considerably greater TF activity in the matrix of stimulated endothelial cells than that expressed on the cell surface. Activation of coagulation by matrix-associated TF was only blocked by IXai when the endothelial cells had been incubated with low concentrations of TNF. Matrix from stimulated endothelial cells effectively promoted fibrinopeptide A generation, fibrin deposition, and platelet aggregate formation in a flow system using low molecular weight heparin anticoagulated blood. These results suggest a contribution of the intrinsic system when coagulation is initiated on a intact vessel and point to the importance of TF in the subendothelial matrix as an initiator of coagulation.

EXPERIMENTAL PROCEDURES
Materials-Low molecular weight heparin (LMWH, FragminO) was from KabiVitrum, Stockholm, Sweden. All culture plastics were obtained from Nunc, Roskilde, Denmark, except the ThermanoZ coverslips from Flow Laboratories Inc., Woodcock Hill, United Kingdom. The other tissue culture supplies (media, antibiotics, and trypsin) were from Gibco Biocult, Paisley, Scotland. All other chemicals obtained from commercial sources were of the highest purity grade available.
Preparation of Coagulation Proteins-Human factors IX and X, purified to homogeneity as described previously (12), were generously provided by Dr. W. Kisiel (University of New Mexico, Albuquerque, NM). For preparation of active site-blocked coagulation factors, bovine factors IX and X were used (these were purified as described previously (13,14)). Factor IX was activated with factor XIa (15, w:w) as described (lo), and the product was chromatographed on a Mono Q column using fast protein liquid chromatography equipment (Pharmacia, Uppsala, Sweden) to remove the factor XIa. Based on sodium docecyl sulfate-polyacrylamide gelelectrophoresis, the factor IXa preparations contained <5% residual factor IX. Factor X was activated by incubation with the purified coagulant protein from Russels viper venom (1:100, w:w) (15) generously provided by Dr. Richard Hart (American Diagnostica, Greenwhich, CT). The reaction mixture was subjected to chromatography on QAE-Sephadex to obtain purified factor Xa (16). These preparations of factor Xa contained 4 % residual factor X. Activated coagulation factors were inactivated during incubation with Glu-Gly-Arg-chloromethylketone (30-fold molar excess, Calbiochem) (17), and excess inhibitor was removed by dialysis. The active site-blocked factor IXa (IXai) and Xa (Xai) had no detectable procoagulant activity.
Radioimmunoassays for factor IX were performed by a modification of the method of Suzuki and Thompson (18) as described (19). These assays employed a monospecific rabbit anti-bovine factor IX which did not cross-react with mouse factor IX antigen present in the animal's plasma but recognized bovine factor IX/IXa (detection limit was approximately 1 nM corresponding to 80% binding on the standard curve). The immunoreactivity of bovine factor IX, IXa, and IXai appeared identical in this assay.
Infusion In case of the IXai infusions, the coagulation factor was diluted in saline and infused via a tail vein such that 5 pg/ animal was given before the TNF and 5 pg/animal was given at the time of the TNF infusion. The level of IXai measured prior to killing the animals, assessed with the RIA ascribed above, was 20 pmol/ml of anticoagulated plasma (levels of factor IX in normal animals are in the range of 60-100 nM). Two hours after the TNF infusion, mice were anesthetized and subjected to whole body beating heart perfusion fixation, as described previously (21). Microvessels from tumor tissue were examined for evidence of thrombi. Cell Cultures-Human umbilical vein endothelial cells were isolated from umbilical veins and cultured according to Jaffe et al. (22) with some minor modifications (23). The cells were identified by their typical characteristics such as the presence of von Willebrand factor.
Routinely, endothelial cells of the second passage were subcultured on gelatin-coated Thermanox@ coverslips. Before seeding the cells, the gelatin on the coverslips was fixed with 0.5% glutardialdehyde. The fixation step with glutardialdehyde did not influence the procoagulant activity on endothelial cells and their matrix upon stimulation. Moreover, virtually identical procoagulant activities on endothelial cells and their matrix upon stimulation were also observed when the endothelial cells were grown on fibronectin (generously provided by Dr. J. van Mourik, Central Laboratory of Bloodtransfusion of the Dutch Red Cross, Amsterdam, The Netherlands) instead of fixed gelatin. Cell monolayers, grown to confluence in 5-7 days, were used (approximately 60,000 cells/cm*). At confluence, the cell culture medium was refreshed 16 h before adding TNF.
Endothelial cells were stimulated for 4 h with TNF. TNF was dissolved in distilled water containing 0.1% bovine serum albumin (BSA) in a concentration of 600 nM and diluted in the cell culture medium to the indicated concentrations. Addition of distilled water, 0.1% BSA alone had no effect on endothelial procoagulant activity. To isolate the extracellular matrix, endothelial cells were exposed to 0.1 M NH,OH for f 1 0 min at room temperature with gentle shaking. The cell layer was completely removed by this procedure leaving the extracellular matrix intact (24-27). Previous studies (28) have demonstrated that a similar procoagulant activity was found on the matrix when the extracellular matrix was isolated with 2 M urea or nitrocellulose-acetate paper stripping. The isolated extracellular matrix was washed three times with phosphate-buffered saline (10 mM phosphate, pH 7.4, and 150 mM NaC1) and was used on the same day.
Factor X Activation with Purified Coagulation Factors-Factor X activation on TNF-stimulated endothelial cell monolayers (1.2 X lo5 cells) and their isolated extracellular matrix was examined as follows. TNF-stimulated endothelial cell monolayers or matrices derived from TNF-stimulated endothelial cells were washed three times with 2 ml of 10 mM HEPES, pH 7.4,137 mM NaCl, 4 mM KC1,3 mM CaC12, 10 mM glucose, and 0.5 mg/ml BSA, and subsequently factor X (200 nM) in 1 ml of this HEPES buffer was incubated on the endothelial cell monolayers or matrices. Where indicated, factor IX (1 nM) and factor VIIIa (10 unit/ml) were present. Factor VI11 was preactivated for 5 min with 0.01 unit/ml thrombin. This concentration of thrombin was sufficient to activate factor VI11 completely but did not influence the absorbance at 405 nm assayed at the end of the experiment. Factor VIIIa activity remained constant during the experimental period examined. The reaction was started by addition of 1 nM factor VIIa. Several aliquots (50 pl) were collected over a 15-min time interval and added to 50 p1 of 50 mM Tris-HC1, pH 7.9,175 mM NaCl, 5 mM EDTA, and 0.5 mg/ml BSA. Factor X activity was assayed with the chromogenic substrate MeO-Co-D-CHG-Gly-Arg-p-nitroanilide (Spectrozyme'" FXa, American Diagnostica Inc., Greenwich, CT). 10 p1 of 2 mM Spectrozyme was added to 100-pl samples in microtiter plates, and the absorbance at 405 nm was monitored with a V,,, reader (Molecular Devices, Menlo Park, CAI. Factor X activation was correlated to a standard curve of purified factor Xa. Blood Collection and Perfusion Studies-LMWH was diluted in saline to a concentration of 200 units/ml. Blood was collected by clean venipuncture in 1:10 (v/v) of this heparin saline. Perfusion studies with steady flow (29) were performed with a rectangular perfusion chamber which was described and characterized extensively elsewhere (30).
Blood was kept at room temperature before use in the perfusions. Perfusions were performed with whole blood or with reconstituted blood. For this latter purpose, washed red blood cells were resuspended in plasma. Washed and packed red blood cells were added to a hematocrit of 0.4 (31). The final platelet count in reconstituted blood was 150,000 platelets/& The coverslips in the perfusion chamber were rinsed before the start of the perfusion with 25 ml of prewarmed (37 "c) 10 mM HEPES-buffered saline. Where indicated, factors IXai and Xai were added to the perfusates in concentrations of 150 nM. For the study of fibrin deposition, peroxidase-labeled fibrinogen' was added to the perfusates (15 ml) which were prewarmed for 5 min a t 37 "C before the start of the perfusion and recirculated then for 5 min. Different wall shear rates were obtained by varying flow rate and chamber width. Since 15 ml of whole blood is exposed to 4.4-cm' cultured cells in this perfusion system and approximately 60,000 cells are present per cm', the cell to volume ratio is k17.500 cells/ml of whole blood. At the end of the perfusion, the chamber was thoroughly rinsed with 30 ml of HEPES-buffered saline. Fibrinopeptide A (FPA) samples were collected from the reservoir at the end of the perfusion. The coverslips were then removed from the chamber and rinsed with 2 ml of HEPES-buffered saline. The perfused part of the coverslip was examined for fibrin deposition as described previously? A previous characterization of this perfusion system has demonstrated that after a lag period of approximately 1 min, the FPA generation in the perfusate increases in a linear fashion up to 5-10 min and stabilizes at longer perfusion times of 10-20 min (28). Hence, the time point of 5 min best reflects steady state formation of FPA in whole blood and was selected for these experiments. It has also been shown (28) that preincubation of endothelial cells or matrix with IgG against human TF, or addition of IgG against human factor VIIa to the perfusate, inhibited the FPA generation in this experimental setting for over SO%, indicating the central role of the TF-factor VIIa complex in FPA generation under the experimental conditions examined.
Afterwards, the coverslips were fixed with 0.5% glutardialdehyde (33). The coverslips were subsequently exposed to osmium tetroxide (2%) as postfixation, dehydrated, and embedded in Epon as described (28,33). For "en face" microphotographs of aggregate formation, the coverslips were fixed after perfusion and stained with May-Grunwald/ Giemsa (30). Microphotographs were made with a Zeiss photomicroscope 111 a t a X 575 magnification.
The Epon with the embedded matrix and adhering platelets were separated from the coverslip by thermoshock. I-pm sections of the Epon-embedded matrices were prepared and stained with methylene blue and basic fuchsin (33) and evaluated for aggregate formation by light microscopy a t X 1000 magnification. The light microscope was interfaced with an image analyzer (AMS 40-1, Analytical Measuring Systems, Saffron Walden, United Kingdom). For each coverslip, a t least 1400 points a t a distance of 1 pm were selected at random and evaluated for aggregate formation.
Activation of Coagulation on Endothelial Cells-Activation of the coagulation mechanism on endothelial cell monolayers was studied under static conditions. For this purpose, confluent endothelial monolayers (1.2 X IO5 cells) were stimulated for 4 h with T N F (0-600 PM). 0.5 ml of 20 units/ml LMWH anticoagulated whole blood was incubated on the cell monolayer a t 37 "C for 5 min. Where indicated, IXai and Xai were present in a concentration of 150 nM. Since 0.5 ml of whole blood is exposed to 1.2 X 10'' cells in this experimental setting, the cell to blood volume ratio in this experimental setting is &240,000 cells/ml of blood. This is approximately 13-fold higher than the surface to volume ratio of the perfusion system. Samples (450 pl) were collected, added to 50 pl of anticoagulant mixture provided in the FPA kit, and assayed for FPA generation.
Fibrinopeptide A Assays-A radioimmunoassay kit (Byk-Sangtec, Dietzenbach, Federal Republic of Germany) was used for FPA measurements. Samples of 900 pl were collected before and after the experiments and added to 100 pl of anticoagulant mixture provided in the kit. Instructions of the manufacturer were followed. FPA values were expressed in nanograms/ml plasma. FPA generation was calculated from the increase in FPA level compared with the initial value just before experiments. All samples were assayed in duplicate. Base-line values were less than 2 ng/ml in all experiments. The results presented are the mean of a t least three experiments.

RESULTS
TNF Infusion Studies-In vivo, infusion of TNF leads to disseminated intravascular coagulation at higher dose and localized intravascular thrombosis at lower doses (1). One striking instance of localized clot formation is the intravascular thrombosis observed in the tumor bed after the infusion of TNF (21). Previous studies (4,5) have indicated that TNF induces the expression of procoagulant activity in cultured endothelial cells, suggesting a basis for the initiation of co- agulation. In view of the paucity of TF in the intravascular space, a situation previously shown to favor activation of factor IX by the T F pathway (34, 35), and the presence of binding sites for factor IX/IXa on endothelial cells (9-11), we examined the role of factor IX/IXa in TNF-induced thrombosis.
After the infusion of TNF, intravascular thrombosis was observed localized to the tumor bed (Fig. lA), as described previously (21). In contrast, when mice were infused with TNF and IXai, no fibrin deposition or thrombosis was evident by morphologic criteria (Fig. 1B). Mice infused with TNF and factor IX, the zymogen, developed thrombosis as in animals infused with TNF alone (data not shown). These data suggested that IXai prevented access of native factor IX/IXa to a limited number of cell surface binding sites in the intravascular space which contributed importantly to the development of thrombi. Based on previous reports demonstrating the presence of T F in the blood vessel wall of a renal cell carcinoma (36), as well as in the subendothelium of normal rabbit aortae and umbilical cords (37), we considered it likely that the infused TNF (either alone or in concert with tumorderived mediators) was inducing TF in endothelium and that this T F would be found predominantly in the matrix. In support of this hypothesis, we have recently identified two tumor-derived polypeptides produced by murine meth A cells which induced T F in endothelium and interacted in a synergistic fashion with TNF (38, 39). To further examine mechanisms through which TNF-stimulated endothelial cells could activate coagulation, and the role of factor IX/IXa, studies were carried out with cultured endothelial cells and their extracellular matrices.
Factor XActivation on Endothelial Cells and Their Matrices in a Static System-Incubation of endothelial cells with TNF led to factor VIIa-dependent factor Xa formation due to the expression of TF (Fig. 2). Induction of T F activity occurred in a dose-dependent manner on both intact endothelial cell monolayers and subendothelium, with the rate of factor X activation 10-20 times greater on the matrix preparations than on intact endothelial cell monolayers. Activation of factor Xa occurred in a linear fashion over the experimental period examined. In control experiments where factor VIIa was absent, no factor Xa formation was observed (data not shown).
Addition of factor IX/VIIIa to TNF-treated endothelial cells incubated with factors VIIa and X enhanced factor Xa formation. The effect of factors IX/VIIIa on factor Xa for-  mation after 7 min incubation is shown in Fig. 3. On endothelial cells, factor IX/VIIIa stimulated factor X activation by 2-to 4-fold over a range of TNF concentrations (Fig. 3A). Experiments with matrix from TNF-stimulated endothelial cells showed the most striking effect of factor IX/VIIIa to be at TNF concentrations of <lo p~, with less enhancement at higher cytokine levels (Fig. 3B).
In Vitro Thrombosis Model-To more closely simulate the situation in uiuo, an in vitro thrombosis model was employed using anticoagulated whole blood (20 units/ml of LMWH) flowing with a shear of 300 s-'. The blood was anticoagulated with LMWH, since this agent effectively inhibits the formation of thrombin in solution but has only a small effect on thrombin generated at and bound to surfaces (28, 40). In the presence of LMWH, FPA generation was observed on matrix derived from TNF-stimulated endothelial cells during the 5min perfusion period but not on matrix from quiescent endothelial cell cultures (Fig. 4). In view of the limited factor X activation seen with intact cells, it was not surprising that FPA generation was difficult to accurately detect in the flow system where dilution of coagulation products in a larger volume is inherent in the experimental system. Because of these results, further studies with intact endothelial cell monolayers employed a static system with anticoagulated blood (see below), and subsequent perfusion studies utilized matrix preparations. PERFUSION STUDIES OVER MATRIX Generation of FPA, fibrin deposition, and platelet aggregate formation on the matrix depended on the concentration of TNF to which the endothelium originally had been exposed (Fig. 5, A-C). The latter indices of activation of coagulation steadily increased in a dose-dependent manner, being halfmaximal at 10-20 pM, and maximal by 50-100 PM.
To investigate the pathway contributing to activation of coagulation, we examined the role of factor IXa and factor Xa. In a previous study (28) we have demonstrated that initiation of coagulation in this setting was initiated by the procoagulant cofactor TF initially synthesized by the endothelial cells in response to TNF. The contribution of factors IX/IXa and X/Xa were assessed using active site-blocked forms that have been shown to recognize cellular and phosholipid surfaces similar to the native enzymes (10,16), although they are devoid of enzymatic activity. Propagation of the procoagulant response involved formation and function of factor IXa at the lower concentrations of TNF (5-50 PM), as indicated by inhibition of platelet aggregate, FPA, and fibrin formation in the presence of factor IXai (150 nM, Fig. 5 , A-C). The inhibitory effect of factor IXai was dependent on the amount of active site-blocked enzyme added and was halfmaximal at 75 nM IXai. In contrast, Xai inhibited coagulation more effectively at all concentrations of TNF (Fig. 5 , A-C) pointing to the importance of effective factor Xa assembly into the prothrombinase complex for activation of coagulation. The inhibitory effect of Xai was half-maximal at 15 nM using this experimental system. Separate experiments, in which the effect of IXai and Xai was tested after 3 min of perfusion, instead of 5 min, gave similar results (data not shown).
Morphological studies were also performed, allowing direct visualization of fibrin fibrils and platelet aggregates (Fig. 6). Micrographs from matrix derived from TNF-stimulated endothelial cells showed platelet aggregates with connecting fibrin fibrils covering much of the coverslip in the presence of TNF alone (Fig. 6A). In the presence of IXai, there was a reduction in the size and number of platelet aggregates and evidence of fibrin deposition was almost completely absent, except for the presence of small fibrin fibrils near platelet aggregates (Fig.  6B). Addition of Xai completely blocked the formation of platelet aggregates and fibrin deposition on matrix completely (Fig. 6C). In the latter case only platelets adhered to the matrix were visualized.

Role of Platelets in Activation of Coagulation on Matrix
Derived from TNF-treated Endothelial Cells-Although activation of coagulation is initiated by TF associated with the matrix, propagation of the procoagulant response could involve the multiple cell types present in blood, This led us to compare FPA generation in plasma with that observed in whole blood after perfusion over matrices derived from TNFtreated endothelium. On the matrix, FPA generation was reduced in plasma at all TNF concentrations, with the most marked inhibition at lower TNF concentrations (Fig. 7A). To more specifically establish the contribution of blood cells to FPA generation, perfusion studies with reconstituted blood were performed over matrices derived from endothelial cells stimulated with 50 PM TNF (Fig. 7B). Reconstitution of plasma with erythrocytes to a hematocrit of 0.4 did not enhance FPA generation, but when platelets were added the procoa lant response was restored to that observed in blood Matrix Derived from TNF-stimulated Endothelial Cells-In view of the range of shear rates present in viuo, where the procoagulant response must occur, it was important to compare activation of coagulation observed at 300 s-l with that observed at a higher wall shear rate (1300 s-') and in the absence of shear (0 s-'), In both the static and the high shear system, exposure of blood to matrix derived from stimulated endothelial cells led to FPA generation (Fig. 8, A-B). Activation of coagulation was dependent on the TNF concentration to which the cells originally had been exposed and was blocked at low concentrations of TNF by IXai and at all TNF concentrations by Xai.
When studies of platelet aggregate formation were performed at 1300 s-', the results were also comparable with those observed at 300 s-'. Activation of coagulation was dependent on the TNF concentration up to 50 PM and was blocked by IXai at low TNF concentrations and by Xai at all TNF concentrations (data not shown). Platelet aggregate formation in the static system was not measurable since in the absence of flow, formation of platelet aggregates is negligible.
Activation of Coagulation on TNF-stimulated Endothelial Cell Monolayers-Although endothelial cells expressed only low amounts of T F (Fig. 2), making it difficult to study in the flow system, they were capable of initiating activation of coagulation. When endothelial cell monolayers (1.2 X IO5 cells) were treated with TNF (0-600 PM) and then exposed to anticoagulated blood, generation of FPA occurred in a dosedependent manner. FPA generation could be prevented by factor Xai (150 nM) over a wide range of TNF concentrations. In contrast to previous studies with matrix, IXai (150 nM) also blocked FPA generation at all concentrations of TNF examined (0-600 pM) (Fig. 9).

DISCUSSION
In the response of tumor vasculature to TNF, intravascular fibrin deposition, progressing to occlusive thrombosis, is initiated when the integrity of the endothelial cell monolayer is still intact (21). We therefore considered that TNF-mediated induction of endothelial cell T F (4, 5) could initiate coagulation in undamaged vessels of the tumor bed. In view of the presence on the vessel surface of specific binding sites for factor IX/IXa, which can promote assembly of the intrinsic factor X activation complex (9-ll), the possiblity was considered that factors IX/VIII contribute to the factor Xa formation that leads to fibrin deposition in this setting. In support Nl o r of this hypothesis, infusion of IXai prevented thrombus fromation in the tumor vasculature of mice infused with TNF. Furthermore, the effect of IXai was observed at a concentration less than that of the zymogen, factor IX, present in mouse plasma (20 nM uersw 60-100 nM). This suggested that IXai was blocking participation of the relatively small amounts of endogenously formed factor IXa in the intrinsic factor X activation complex, presumably, by competing for a Confluent endothelial monolayers were exposed to increasing TNF Concentrations (0-600 PM TNF) for 4 h, and the isolated matrix was either exposed to 0.5 ml 20 units/ml LMWH blood for 5 min (wall shear rate = 0 s") ( A ) or perfused with LMWH blood for 5 min a t a wall shear rate of 1300s" ( B ) . Control perfusates (+) were compared with perfusates containing IXai (0)   limited number of high affinity sites on cell surfaces such as the endothelial cell (9-11).
These results led us to focus initially on the role of the intrinsic system in TF-mediated activation of coagulation on the surface of stimulated endothelial cells. Since changes in permeability of the endothelial cell monolayer accompany thrombin generation (34), it was important to compare the TF activity present on the surface of an intact monolayer with that of a matrix preparation from the same TNF-treated endothelial cells. The extracellular matrix of cultured endothelial cells closely resembles the vascular basement membrane in composition and structural array of organization (24-27) and is used as a model system to study the interaction of blood cells and coagulation factors with the subendothelium (40, 42-44). In our experiments, only small amounts of T F activity were observed on intact endothelial cell monolayers, compared with that found in the matrix following removal of the endothelial cells. These observations led us to compare characteristics of the activation of coagulation on TNFtreated endothelium with that observed on the matrix.
The greater activation of coagulation observed on matrix allowed for studies using a perfusion system with whole blood anticoagulated with LMWH. This system, which has been described in detail elsewhere (28), employs the LMWH to slow down fibrin formation by promoting inactivation of thrombin in the fluid phase. Matrices derived from TNFtreated endothelial cells, but not those from untreated control cells, promoted formation of FPA, platelet aggregates, and fibrin over a range of shear rates. On the matrix, factor Xai, an inhibitor of factor Xa assembly into the prothrombinase complex (16), blocked activation of coagulation at all concentrations of TNF examined. In contrast, IXai only blocked coagulation on matrix from endothelial cells stimulated with lower concentrations of TNF (<50 pM). Consistent with these results, enhancement of factor X activation, when factors IX/ VIIIa were added to matrix preparations incubated with factors VIIa/X, was maximal only in those derived from endothelial cells treated with lower TNF concentrations. The mechanism(s) responsible for the dependence of activation of coagulation on matrix derived from endothelial cells stimulated with lower concentrations of TNF on factors IX/IXa is unclear, at present, although several possibilities are evident. A previous study has indicated that at lower TF concentrations factor IX activation is favored compared with factor X, and thus factor IXa/VIII-mediated factor Xa formation may assume a greater importance (34). In addition, matrix preparations derived from endothelial cell monolayers may contain residual endothelial cell factor IX/IXa binding sites or a new class of sites which could potentially facilitate assembly of the factor X activation complex. When greater amounts of T F were present in the matrix ( X 0 PM), the role of factor IXa becomes less important, probably due to the greater extent of direct factor X activation by the TF-factor VIIa complex.
Although coagulation on matrix was inititated by T F associated with the matrix, effective propagation of the procoagulant response required the presence of platelets. This most likely reflects the presence of assembly sites for both factor X activation and prothrombinase complexes on the surface of the stimulated platelet (45-48). Compared with quiescent endothelium and certain peripheral blood cells, it has been found that the platelet is a very effective surface for the interaction leading to assembly of the prothrombinase complex (32, 49). Thus, there is a close link between TF/factor VIIa-mediated activation of factors IX and X on the matrix and their subsequent participation in propagation of the procoagulant response on the platelet surface.
The limited amount of TF expressed on endothelial cell monolayers made it more difficult to analyze their properties in the flow system. It was not surprising that FPA generation was hard to detect in the flow system where dilution of coagulation products is inherent in the experimental system. In the static system, TF-initiated activation of coagulation, leading to FPA generation was observed with intact endothelial cell monolayers. Factor IX/IXa appeared to have a central role in the reactions leading to factor Xa and thrombin formation in this setting, as indicated by inhibition of FPA generation in the presence of IXai at all TNF concentrations. Consistent with this observation, the presence of factors IX/ VIIIa in reaction mixtures containing stimulated endothelial cells, factors VIIa and X, enhanced factor Xa formation over a wide range of TNF concentrations. This could be explained by the presence of a limited number of high affinity endothelial cell surface binding sites which can facilitate assembly of the factor X activation complex. The importance of factor IX/IXa in promotion of coagulation on intact TNF-treated endothelial cell monolayers might also result from the low concentrations of T F on the endothelial cell surface thereby favoring factor IX activation (34).
These observations indicate that cellular binding sites for coagulation factors, especially those on the endothelium and the platelet, can significantly influence activation of coagulation on stimulated endothelial cells and their matrices. Furthermore, the matrix derived from stimulated endothelial cells appears to promote coagulation much more effectively than intact monolayers. This finding emphasizes a link between the pathogenesis of thrombosis and perturbations in vessel wall barrier function which lead to increased endothelial cell monolayer permeability, thereby allowing access of plasma coagulation proteins to the TF containing matrix.