Contributions of extravascular and intravascular cells to fibrin network formation, structure, and stability

Fibrin network production is essential for hemostasis, and an increasing number of studies implicate abnormal fibrin structure and stability in thrombotic disease, including premature myocardial infarction, and venous thromboembolism.^1,2  Fibrin production is the culmination of procoagulant processes that generate thrombin at an injury site. These processes are initiated by exposure of extravascular tissue factor (TF)–bearing cells after vascular disruption (hemostasis), or triggered by inappropriate expression of intravascular procoagulant activity (thrombosis). Understanding cellular and biochemical mechanisms contributing to fibrin formation, structure, and stability is fundamental for continued development of antithrombotic and thrombolytic therapies.

Previous studies have shown the thrombin concentration catalyzing the conversion of fibrinogen to fibrin influences fibrin structure and stability. Low thrombin concentrations produce coarse fibrin networks that are relatively susceptible to fibrinolysis, whereas high thrombin concentrations produce dense networks that resist fibrinolysis.^3-7  These differences in fibrinolytic susceptibility have significant physiologic implications: clots that are susceptible to fibrinolysis are prone to premature lysis and bleeding, whereas clots that resist fibrinolysis are prone to thrombosis.⁸ 

Although the role of thrombin in fibrin formation has been heavily scrutinized, few studies have examined the contributions of cells—the site of thrombin generation—to fibrin formation, structure, and stability. We previously demonstrated that in situ thrombin generation on the cell surface modulates the 3-dimensional structure and stability of the fibrin network. When thrombin generation is localized to a cell surface, clots develop a significantly denser fiber network within 10 micrometers of the surface compared with distal regions of the clot (40-50 micrometers from the cell surface).⁹  Notably, the proximal, dense fibrin network is resistant to lysis compared with the distal, coarse network, suggesting the clot's structural heterogeneity dictates its fibrinolytic susceptibility.⁹  Together, these findings suggest that when thrombin generation is localized to a cell surface, the fibrin network reflects its spatial location and the procoagulant phenotype of the local cell population.⁹ 

A cell's ability to support thrombin generation is determined by its expression of procoagulant and anticoagulant activities. For example, TF expression initiates thrombin generation, and phosphatidylserine exposure is necessary for robust thrombin generation through prothrombinase activity.¹⁰  Tissue factor pathway inhibitor and thrombomodulin down-regulate thrombin generation by negatively regulating TF and factors V and VIII activity, respectively.^11,12  Because expression of these properties is cell specific, cells from different vascular locations differ in their ability to support thrombin generation.¹³  Extravascular cells (fibroblasts and smooth muscle cells [SMCs]) are highly procoagulant and primed to support hemostasis, whereas intravascular (endothelial) cells suppress thrombin generation and fibrin formation.^14,15  However, exposure of intravascular cells to bacterial lipopolysaccharide,¹⁶  hypoxia,¹⁷  or inflammatory cytokines¹⁸  induces expression of a procoagulant/prothrombotic phenotype. Although prior studies have compared thrombin generation on these cells, the contributions of these cells toward clot structure and stability have not previously been investigated.

In the current study, we compared the abilities of extravascular and intravascular cells to modulate fibrin formation, structure, and stability. Our findings show that procoagulant extravascular cells (fibroblasts, SMCs) supported the formation of dense fibrin networks that resisted fibrinolysis. Network structure depended on procoagulant activity but not interactions between integrins and fibrin. Inflammatory cytokines induced prothrombotic activity on endothelial cells and subsequent production of an abnormally stable fibrin network. Blocking TF reduced fibrin network density and stability of clots produced by cytokine-stimulated intravascular cells. These findings suggest fibrin network formation, structure, and stability are important determinants of hemostasis and vascular cell–specific thrombosis.

Sources of proteins and materials are provided in supplemental Materials (available on the Blood website; see the Supplemental Materials link at the top of the online article). Human blood and vessel segments were obtained through the University of North Carolina at Chapel Hill with institutional review board approval.

Immortalized human fibroblasts (NHF₁-hTert), generously provided by Dr Marila Cordeiro-Stone (University of North Carolina at Chapel Hill), were grown as described.^9,19  Primary human umbilical vein endothelial cells (HUVECs) pooled from 3 donors and aortic smooth muscle cells (SMCs) from 1 donor were from Lonza Walkersville and cultured as directed by Lonza. Cells were used between passages 3 to 10 to reduce phenotypic drift. Cells (7500/well) were seeded in 96-well tissue culture flat-bottom plates, grown to 80% to 95% confluence, and washed with phosphate-buffered saline (PBS) before use.

TF (Xa generation), prothrombinase activity (thrombin generation), and activated protein C generation were measured by chromogenic substrate cleavage on a SpectraMax 340PC plate reader (Molecular Devices), as described.^9,12,20 

Proteins (40 μg) were separated by 4% to 12% sodium dodecyl sulfate–polyacrylamide gel electrophoresis and transferred to NitroBind Pure Nitrocellulose (GE Osmonics). Western blotting for expression of thrombomodulin and integrin subunits αv and β3 was performed as described, using extracellular signal-related kinase 2 as a loading control.²¹  Proteins were visualized with an enhanced chemiluminescence detection kit (Amersham Biosciences).

Blood was collected from 30 healthy subjects (53% female, 40% nonwhite) through a 21-gauge butterfly into a syringe. The first 5 mL was used to prepare platelet-poor plasma and measure partial thromboplastin times for each donor. The following 30 mL was drawn into a separate syringe and immediately transferred to sodium citrate/corn trypsin inhibitor (CTI; 0.105 M/3.2% sodium citrate, pH 6.5, 18.3 μg/mL CTI) to minimize contact activation.²²  Platelet-free plasma (PFP) was prepared by sequential centrifugation (150g for 15 minutes, 13000g for 15 minutes), aliquoted, and snap-frozen in liquid nitrogen within 2 hours of blood collection. Normal pooled PFP was prepared by pooling PFP from persons with normal partial thromboplastin times. PFP was analyzed for activity levels of factors V, VII, VIII, IX, X, and XI, protein C, free protein S, and fibrinogen at University of North Carolina Hospitals McLendon Clinical Laboratory; all were in the normal reference range.

Thrombin generation was measured by calibrated automated thrombography using a Fluoroskan Ascent fluorometer (ThermoLabsystem), as described^23-25  except 80 μL of PFP and 20 μL of HEPES-buffered saline/bovine serum albumin were added to washed cell monolayers to initiate reactions (67.7% plasma, final). Addition of fluorogenic substrate, alone, to cell monolayers did not produce signal, indicating the cell types did not express endogenous substrate cleaving activity (data not shown).

Thrombin generation was calculated using Thrombinoscope software Version 3.0.0.29 (Thrombinoscope BV). The thrombin generation rate was calculated by dividing the peak height by the difference from time to peak to lag time. The endogenous thrombin potential was not reported because thrombin generation curves on unstimulated HUVECs did not return to baseline.

Clotting was initiated by incubating cells with recalcified (10mM, final) normal, pooled PFP (86.7% plasma, final). In some assays, fibrinogen (2 mg/mL), CaCl₂ (5mM, final), and thrombin (2nM, final) were added to cells instead of plasma to examine networks formed in the absence of in situ thrombin generation. Exogenous thrombin addition to unstimulated and stimulated HUVECs caused cells to lose their normal polygonal shape due to actin polymerization.²⁶  To prevent actin polymerization and cell contraction, confluent HUVECs were treated for 30 minutes with the Rho-associated kinase inhibitor Y-27632 (10μM, final) in culture media; this inhibitor does not prevent fibrin from interacting with β3 integrins.²⁷  Fibrinolysis assays were performed by adding tissue plasminogen activator (tPA; 250 ng/mL, final) to PFP at the reaction start. Clot formation and lysis were detected by turbidity at 405 nm in a SpectraMax 340PC plate reader.

Clotting on platelet monolayers was performed by adhering gel-filtered platelets²⁸  (10⁶ total) to collagen-coated (50 μg/mL, final) glass coverslips and activating platelets with SFLLRN (50 μg/mL, final) in the presence of abciximab or control immunoglobulin G (IgG; 0.136μM, final) and MgCl₂ (2mM final). Platelets were allowed to adhere and activate for 30 minutes at 37°C before coverslips were washed with HEPES-buffered saline/bovine serum albumin to remove nonadherent platelets. Recalcified (10mM, final), normal PFP with abciximab or control IgG and lipidated TF (Innovin; 1:120 000, final) was then added to platelets. Clots were allowed to form for 2 hours before imaging.

Cells (33 000/well) were seeded in Lab-Tek II Chamber no. 1.5 coverglass (Nalge Nunc International) and grown to 80% to 95% confluence. Clots were formed as for clotting assays with the addition of AlexaFluor-488–labeled fibrinogen (10 μg/150 μL sample, 3.2% of total fibrinogen). Synthetic peptide GRGDSP (500μM, final), which blocks binding to RGD-mediated receptors, or control peptide GRGESP was included as indicated. Clotting proceeded until a constant final turbidity was reached in separate, parallel reactions monitored in a SpectraMax 340PC plate reader.

Clots were imaged at room temperature as described⁹  on a Zeiss LSM5 Pascal laser scanning confocal microscope linked to a Zeiss Axiovert 200M microscope equipped with a Zeiss 63×/1.4 NA oil-immersion plan apo-chromatic lens. The 488-nm line of a medium power multiline argon ion laser was used for excitation and a 505- to 530-nm band-pass filter for emission. A computer equipped with Carl Zeiss software (v1.5) was used to operate the system. Optical sectioning was achieved by closing the pinhole in the front of the detector to 1 airy unit. The zoom factor was 1. Thirty optical sections (1024 × 1024 pixels each) in 3 randomly chosen locations were collected at 0.36-μm intervals in the z-axis at the cell surface (defined as when the cells were in focus using differential interference contrast microscopy) and 40 μm above the cell surface, thus covering a region 10.8 μm thick at and above the cell surface. Image volumes were 146 × 146 × 10 μm. Single images were collected in 15.47 seconds. Optical resolution was approximately 0.14 μm in the xy-plane and approximately 0.5 μm on the z-axis. The sectioning interval in z was smaller than the calculated z-axis optical section resolution to achieve Nyquest sampling in z based on the Zeiss software calculation. No correction was made for refractive index mismatch. Images were deconvolved using 3-dimensional deconvolution algorithms in AutoQuant's Autodeblur (Version x1.4.1; Media Cybernetics Inc). Fibrin network density was analyzed using ImageJ (1.37V; National Institutes of Health) by placing random grids of 2 pixel crosses on individual slices (∼ 121-144 crosses/slice), and counting fibers intersecting the middle of the crosses, divided by the total number of crosses. In slices near the cell surface, crosses placed over cells (indicated by “empty” space) were subtracted from the total number of crosses to correct for the volume occupied by the cell.⁹  Optical resolution (0.14 μm [xy] by 0.5 μm [z]) was near the lower limit of resolution to measure fibrin fiber diameter accurately; therefore, fibrin diameter analysis was not performed.³ 

Saphenous vein segments (2-6 cm) were obtained with informed consent from patients undergoing coronary artery bypass grafting under a protocol approved by the University of North Carolina institutional review board. After removal, vein segments were placed in heparinized saline before immediate transfer to the laboratory where they were placed into PBS and cut longitudinally down the midline, creating 2 vessel halves. Vessel segments were then cut laterally and placed into individual chambers on a permanox slide, lumen side up. Vessels were then washed 5 times with PBS, and recalcified (10mM, final) plasma was added to the vessel lumen. After 2 hours, clots were fixed overnight in 4% paraformaldehyde/1% glutaraldehyde in PBS. Fixed clots were washed and embedded in Epon for thin sectioning. Clots were imaged using a LEO EM910 transmission electron microscope (Carl Zeiss SMT) operating at 80 kV. Digital images were recorded using a Gatan Orius CCD Digital Camera and Digital Micrograph 3.11.0 (Gatan Inc). Fibrin density was determined on 6300× micrographs by placing 1-μm-thick boxes parallel to, and 0.25 μm and 5 μm from, the cell surface and counting fibers inside the box.⁹ 

Continuous variables were expressed as mean plus or minus SD. Pairwise comparisons were performed by Student t test. A 1-way analysis of variance with Dunnett posthoc test using unstimulated HUVECs as the index group was used to control for error of multiple testing. P values less than .05 were considered significant.

We first compared the abilities of extravascular cells (fibroblasts, SMCs) and intravascular cells (HUVECs) to support thrombin generation and fibrin formation in recalcified, contact-inhibited PFP. As in previous studies,^15,29  thrombin generation on fibroblasts and SMCs demonstrated a significantly shorter lag time, higher rate, shorter time to peak, and increased peak height than on unstimulated HUVECs (Figure 1A, Table 1). Although the final peak turbidities of all clots were similar, fibroblasts and SMCs supported a significantly faster onset and rate of clot formation than unstimulated HUVECs (Figure 1B, Table 1).

To determine how inflammatory cytokines influence fibrin formation and stability, we first treated HUVECs with 0 to 1nM tumor necrosis factor-α (TNFα) for 4 hours. Consistent with previous findings,^18,30  TF activity (factor Xa generation) was dose dependent with respect to TNFα concentration (Figure 1A inset). Subsequent assays were performed with HUVECs treated with 1nM TNFα for 4 hours. We first examined thrombin generation and fibrin formation, structure, and stability. Compared with unstimulated HUVECs, TNFα-stimulated HUVECs exhibited a significantly shorter lag time and time to peak, and increased rate and peak height of thrombin generation (Figure 1A, Table 1). Compared with unstimulated HUVECs, TNFα stimulation also significantly shortened the onset and increased the rate of clot formation (Figure 1B, Table 1). We observed similar findings with interleukin-1β (IL-1β)–stimulated cells (data not shown). In short, cytokine stimulation increased HUVEC procoagulant activity to a level similar to that of extravascular cells (SMCs).

We next incubated fibroblast, SMC, and unstimulated and TNFα-stimulated HUVEC monolayers with recalcified, normal pooled PFP and used laser scanning confocal microscopy to examine the fibrin structure of clots produced by these cells. Consistent with thrombin generation and fibrin formation measurements, clots produced by fibroblasts and SMCs were composed of densely and moderately packed networks, respectively, whereas clots produced by unstimulated HUVECs were composed of an open, coarse network (Figure 2A-C,E-G, Table 2). TNFα-stimulated (Figure 2D,H) and IL-1β–stimulated (data not shown) HUVECs produced significantly denser networks than unstimulated HUVECs, more like those produced by SMCs. Together, these findings show extravascular cells (fibroblasts, SMCs) produce denser fibrin networks than unstimulated intravascular (endothelial) cells, but stimulating endothelial cells with proinflammatory cytokines induces their adoption of an extravascular cell–like prothrombotic phenotype that produces an abnormally dense fibrin network (Table 2).

In addition to individual effects of cells on network density, our analysis also showed clots produced by cultured fibroblasts, SMCs, and HUVECs demonstrated increased fibrin density near (0-10 μm) versus distal (> 10 μm) to the cell surface (Figure 2 compare A, B, C, and D with E, F, G, and H, respectively, 2M, Table 2, and Campbell et al⁹ ). We confirmed these findings in clots produced by freshly excised human blood vessels, by adding recalcified, normal PFP to the lumen of excised, washed, saphenous veins isolated from patients undergoing coronary bypass surgery. Clots formed in vessels for at least 2 hours before fixation, thin sectioning, and imaging in cross-section by transmission electron microscopy. Similar to our in vitro findings with cultured cells (Figure 2A-M and Campbell et al⁹ ), clots produced in vessels ex vivo showed significantly (P = .007) increased fibrin density near versus distal to the vessel wall (1.23 ± 0.08 vs 0.57 ± 0.20 fibers/μm², respectively; Figure 2N). These findings confirm that the observed fibrin density gradient is not an artifact of cultured cells, but is also likely to arise from procoagulant activity expressed by cells within the vasculature.

We previously showed fibroblasts produce this fibrin density gradient via thrombin diffusion from the site of generation on the cell surface.⁹  However, other studies have shown cellular integrins, specifically αIIbβ3 and αvβ3, increase fibrin network density and organization on platelets³¹  and HUVECs,³²  respectively. Those studies were performed by adding exogenous thrombin to plasma³¹  or fibrinogen³²  in the presence of cells. Thus, although both SMCs and HUVECs express αv and β3 subunits^32-34  (data not shown), the relative roles of in situ thrombin generation and integrin interactions in dictating fibrin structure in plasma clots have not been established for SMCs or HUVECs.

To investigate the role of integrins in determining fibrin structure on these cells, we formed clots via in situ thrombin generation in the presence of 500μM GRGDSP peptide or GRGESP control peptide. This concentration of GRGDSP fully inhibits RGD-mediated interactions^32,35  and blocks platelet aggregation (data not shown), but does not alter the rates of thrombin generation or fibrin polymerization in plasma (data not shown). Fibrin density produced by fibroblasts, SMCs, or HUVECs was not altered by GRGDSP peptide (Figure 2 compare E, F, G, and H with I, J, K, and L, respectively, Table 2, supplemental Figure 1). Clinically relevant concentrations of anti-β3 integrin abciximab (0.136μM)³¹  similarly failed to reduce network density (Figure 3), indicating the lack of involvement of integrins in determining network structure on these cells.

To evaluate the role of in situ thrombin generation in fibrin network production, we triggered clotting on fibroblasts, SMCs, and HUVECs via exogenous thrombin addition to purified fibrinogen. These reactions bypass in situ thrombin generation, and fibrin formation occurs within a uniform thrombin concentration. Under these conditions, fibroblasts, SMCs, and HUVECs did not produce a fibrin density gradient like that seen during in situ thrombin generation, and GRGDSP peptide did not decrease fibrin density in these clots (supplemental Figure 2, supplemental Table 3). These results confirm the central role of cellular procoagulant activity in determining the 3-dimensional network structure on fibroblasts, SMCs, and HUVECs.

Because previous studies have shown a significant influence of β3 integrins on platelets,³¹  we compared findings with fibroblasts, SMCs, and HUVECs with those with platelets. Clots were formed on platelets via in situ thrombin generation (plasma plus soluble TF) or exogenous thrombin addition (fibrinogen plus thrombin) in the absence or presence of abciximab. Under these conditions, the thrombin generation rate did not change in the absence or presence of abciximab and was similar to HUVECs (data not shown). Figure 4 shows patches of increased fibrin density in clots produced by in situ thrombin generation (Figure 4A,C), likely reflecting contributions of platelet procoagulant surface in these reactions. Reactions with exogenous thrombin addition show a more uniform network (Figure 4B,D). We also observed focal density (arrows in Figure 4A-B) where individual platelets concentrated the surrounding fibrin network density in clots produced during in situ thrombin generation or exogenous thrombin addition. In contrast to findings on fibroblasts, SMCs, and HUVECs, abciximab significantly reduced fibrin density (larger patches and focal density) in clots produced by platelets during in situ thrombin generation, as well as during exogenous thrombin addition (Figure 4E). The effect of abciximab demonstrates the significant influence of platelet β3 integrin interactions on network density, likely reflecting the high concentration of β3 integrin subunits on platelets.³¹  Together, these findings suggest the mechanism(s) dictating fibrin network structure on fibroblasts, SMCs, and HUVECs differs from those on platelets. Specifically, αIIbβ3 dictates network structure in platelet-rich clots, whereas integrin interactions play a minor to no role in clots formed by fibroblasts, SMCs, and HUVECs.

Cellular procoagulant activity and fibrin network structure dictate clot fibrinolytic stability.^4,5,9  To compare lysis of clots produced by extravascular and intravascular cells, we added tPA (250 ng/mL, final) to recalcified normal PFP, incubated the tPA-treated PFP with cells, and monitored fibrin formation and lysis by a turbidity increase and subsequent decrease, respectively, at 405 nm.⁴  Traditional measures of fibrinolysis (“lysis time”—the time between the midpoints of the increase and decrease in turbidity¹  and “1/2 lysis time”—the time the turbidity drops to one-half the maximum turbidity^36,37 ) are performed using a constant type and concentration of trigger to initiate clotting (thrombin or TF). However, because our experiments were initiated by cells expressing different TF concentrations, these lysis times were affected by the initiation of thrombin generation and subsequent effects on the lag times and rates of clot formation, limiting the utility of the traditional lysis measures. Therefore, we used the time to peak turbidity and the peak turbidity as measures of the time to maximum fibrin formation and the peak incorporation of fibrin into clot, respectively.⁴  The ability of cellular procoagulant activity to increase fibrin stability shortens the time to peak and increases the peak turbidity. In these assays, fibroblasts and SMCs supported rapid formation of clots with high peak turbidity, whereas HUVECs produced the longest time to peak and lowest peak turbidity (Figure 5). Clots produced by TNFα-stimulated HUVECs exhibited a faster time to peak and higher peak turbidity than unstimulated HUVECs (Figure 5). In contrast, no differences in time to peak or peak turbidity were observed in plasma clots produced by exogenous thrombin addition (10nM) to these cells (data not shown), indicating the observed differences were due to cellular procoagulant activity and not cellular fibrinolytic or antifibrinolytic activities. Together, these findings suggest extravascular cells and TNFα-stimulated HUVECs produce stable networks that resist lysis, whereas unstimulated HUVECs produce less fibrin accumulation.

Our findings indicated that coagulant activity, and not integrins, dictated fibrin structure and stability. To probe the coagulant mechanism by which these cells dictate fibrin structure and stability, we characterized their specific procoagulant and anticoagulant activities. Fibroblasts, SMCs, and unstimulated and TNFα-stimulated HUVECs had similar prothrombinase activities, suggesting procoagulant phospholipid exposure was similar on each cell type (Figure 6A, Table 3). Consistent with previous studies,^38,39  SMCs, unstimulated HUVECs, and TNFα-stimulated HUVECs expressed thrombomodulin antigen (data not shown) and supported similar rates of protein C generation, whereas fibroblasts did not express thrombomodulin and supported significantly lower protein C generation (Table 3). Because fibroblasts, SMCs, and HUVECs produced clots with significantly different fibrin densities, these findings suggested neither prothrombinase activity nor activated protein C generation specifically dictated fibrin formation, structure, and/or stability on these cell types.

We then measured TF activity on these cells. Consistent with previous studies,^15,29  fibroblasts and SMCs supported 131-fold and 8.7-fold higher rates of factor Xa generation, respectively, than HUVECs (Table 3). Stimulating HUVECs for 4 hours with 1nM TNFα increased factor Xa generation 19.7-fold (Table 3). Anti-TF antibody (10 μg/mL, final) significantly reduced TF activity by 78%, 65%, and 85% on fibroblasts, SMCs, and TNFα-stimulated HUVECs, respectively (Table 3). Because the anti-TF antibody and recombinant factor VIIa (rFVIIa) were added to reactions simultaneously, competition between rFVIIa and anti-TF antibody for binding to TF prevented complete inhibition of TF activity. Together, these data showed a positive relationship between TF activity (Xa generation) and fibrin network density on these cells (Figure 6B), suggesting TF activity was the dominant mechanism differentiating procoagulant activity and subsequent fibrin network density of clots produced by these cells.

Previous studies suggested TF inhibitors, including antibodies against TF⁴⁰  and active-site inhibited FVIIa,⁴¹  reduce thrombus propagation and thrombus weight in animal models of thrombosis. Given the positive correlation between TF activity and fibrin network density, we hypothesized that TF and network structure are important targets for antithrombotic agents.

To determine whether TF inhibition alters fibrin structure and stability, we treated unstimulated and TNFα-stimulated HUVECs with anti-TF⁴²  or control IgG (10 μg/mL, final) and measuredthrombin generation and fibrin formation and structure. On unstimulated HUVECs, addition of anti-TF antibody slightly and nonsignificantly prolonged the lag time of thrombin generation (Figure 7A) and onset of clot formation (Figure 7B), suggesting these unstimulated, cultured cells express a low level of TF activity, likely due to their extraction from their native environment under shear. As in Figure 1B, the final turbidities of clots produced by these cells did not differ (Figure 7B) despite the substantial differences in cellular procoagulant activity (Figure 7A). Addition of anti-TF to unstimulated HUVECs did not affect fibrin network structure (Figure 7C-D). However, addition of anti-TF antibody to TNFα-stimulated HUVECs significantly (P < .05) delayed the onset and decreased the peak height and time to peak thrombin generation (Figure 7A), prolonged the onset of clot formation (Figure 7B), and reduced the fibrin network density to levels found in clots produced by unstimulated HUVECs (Figure 7C-D). In the presence of tPA, anti-TF antibody prolonged the time to peak turbidity in clots produced by TNFα-stimulated HUVECs (Figure 7E). Importantly, the peak turbidity (fibrin incorporation) produced by TNFα-stimulated HUVECs in the presence of anti-TF was statistically indistinguishable from that produced by unstimulated HUVECs (Figure 7F). These findings suggest inhibiting thrombin generation, either through TF, Xa, or direct thrombin blockade, would significantly reduce fibrin formation and stability.

A growing body of work has implicated fibrin structure and stability as important determinants of hemostasis and thrombosis. Most studies have focused on plasma contributions to clotting, showing that plasma clots from patients with bleeding disorders are composed of an abnormally coarse, unstable fibrin network, whereas plasma clots from patients with a personal or family history of thrombosis have an abnormally dense network that resists fibrinolysis.^1,2,43  Few studies have examined cellular contributions to clot structure and stability. In the current study, we showed that extravascular and intravascular cells support fibrin formation, structure, and stability. Our findings show extravascular cells supported the formation of a dense fibrin network that resisted fibrinolysis. Cytokine-stimulated intravascular cells similarly produced an abnormally dense, stable fibrin network that was inhibited by TF blockade. These findings suggest specific mechanisms by which different vascular cells contribute to hemostatic and thrombotic clot formation.

All of the examined cell types produced a denser fibrin network proximal to the cell surface compared with distal regions of the clot, consistent with our previous findings that thrombin generation on a fibroblast surface dictates the 3-dimensional network structure.⁹  The current study extends the generality of this observation to 3 additional cell types (SMCs, HUVECs, and platelets) as well as cells lining freshly excised human saphenous veins. Experiments with excised saphenous veins in particular (Figure 2N) confirm the observation is not restricted to cultured cells, but is likely to arise from procoagulant activity of intravascular cells in vivo. Although a previous study showed the αv and β3 integrins significantly increase fibrin organization in clots produced on HUVECs,³²  we were unable to confirm this finding. Rather, we found in situ thrombin generation was both necessary and sufficient to increase fibrin density in clots produced by HUVECs, without further effect of cellular receptors. Differences between our findings and the previous work³²  may reflect differences in the source of HUVECs used in the study; the previous study used HUVECs from individual donors, whereas we used HUVECs pooled from 3 donors. We note, however, that although integrins did not significantly modulate fibrin structure of clots formed by TF-bearing cells, these findings do not preclude a role for cellular integrins in anchoring the fibrin network to the vessel wall. Further investigations of how clots anchor to surrounding cells are necessary to elucidate mechanisms that predispose some thrombi to embolism.

It is interesting that differences in fibrin structure seen by microscopy were not reflected in the final turbidity measurements of these clots (Figures 1B, 7B). In prior studies, turbidity and fibrin structure were studied using homogeneous fibrin networks. Our data suggest the fibrin structural gradient produced during in situ thrombin generation uncouples these measurements. These findings suggest studies performed in the presence of cells should include a microscopic method to directly visualize the fibrin network and validate turbidity measurements.

Fibroblasts and SMCs, present in the outmost and medial layers of the vessel wall, respectively, supported robust thrombin generation and clot formation, and produced dense, stable fibrin networks. After vascular breach, exposure of these cells to blood would initiate a rapid procoagulant response. In contrast, intravascular (endothelial) cells are specifically designed to prevent thrombin generation and fibrin production. The unstimulated HUVECs used in our study supported low thrombin generation and only coarse network formation. This basal activity of cultured endothelial cells probably reflects phenotypic changes upon removal of these cells from their native environment under flow.⁴⁴  Regardless, the observed differences between fibrin networks formed by endothelial cells and extravascular cells are profound, and suggest exposure of extravascular cells to blood would promote a stable fibrin network after vascular injury.

In atherosclerotic disease, exposure of blood to collagen and subendothelial cells (eg, SMCs) upon plaque rupture would be expected to produce a thrombus composed primarily of platelets and reinforced by a dense, overly stable fibrin network. Fibrin-rich venous clots, conversely, are thought to arise on intact endothelium.^45,46  In this case, exposure of intravascular cells to inflammatory mediators, including bacterial lipopolysaccharide,¹⁶  hypoxia,¹⁷  or cytokines,¹⁸  may induce inappropriate intravascular TF expression. In our study, stimulating endothelial cells with TNFα increased TF expression and thrombin generation and caused the formation of dense fibrin networks, similar to those produced by extravascular cells (SMCs). We observed similar effects after treatment with IL-1β (data not shown), indicative of a generalized mechanism by which inflammatory mediators trigger a procoagulant/prothrombotic phenotype and subsequent intravascular fibrin formation.

It is interesting to note the difference in mechanisms dictating fibrin structure in clots produced by TF-bearing cells (procoagulant activity) and platelets (β3 integrin–fibrin interactions). Smyth et al⁴⁷  previously showed variable protection of β3-deficient mice from thrombosis initiated by different mechanisms. Whereas β3-deficient mice showed reduced thrombus formation in a model of arterial (platelet-rich) thrombosis, the β3-deficient mice were only minimally protected from TF-initiated microvascular thrombosis.⁴⁷  Our findings, consistent with these in vivo thrombosis models, provide a mechanistic basis for the effectiveness of β3-blocking strategies in preventing platelet-rich arterial thrombi⁴⁸  versus anticoagulants to prevent fibrin-rich venous thrombi.^49,50 

Our findings show a crucial role for TF activity in fibrin network formation, structure, and stability. Compared with unstimulated HUVECs, TNFα-stimulated HUVECs expressed significantly higher TF activity, but similar lipid and thrombomodulin expression. Thus, increased TF activity alone was sufficient to increase thrombin generation and fibrin density to levels seen in clots produced by extravascular SMCs. Blocking TF activity on TNFα-stimulated endothelial cells reduced procoagulant activity and fibrin density and stability to that of unstimulated cells, lending support to a mechanism by which VIIa/TF inhibition reduces venous thrombosis.⁴⁰ 

Our current experimental design omits the effects of shear on fibrin formation, structure, and stability. Previous studies have demonstrated that endothelial cells are less susceptible to TNFα-induced TF expression in the presence of shear.⁵¹  However, once stimulated, endothelial cells support robust thrombin and fibrin formation under low shear, whereas high shear suppresses fibrin deposition.⁵²  Interestingly, higher TF concentrations increase the shear threshold that attenuates fibrin production.⁵³  Together, these studies suggest that reduced shear (stasis) promotes endothelial cell susceptibility to cytokine stimulation, and that the combination of activated endothelial cells and low shear promotes local thrombus formation. Although the previous studies examined the role of shear and TF concentration on fibrin deposition,^53,54  the resulting fibrin structures were not evaluated. Based on our current study, evaluating fibrin structure under different shear forces may offer insight into the role of stasis in thrombus formation.

Our results provide an important foundation for future studies of cell types from different vascular beds. Our system allows for the examination of how cellular properties such as TF⁵⁵  and thrombin¹³  expression on endothelial cells from the arterial, venous, and microvascular circulation, or binding of TF microparticles,⁵⁶  contribute to clot structure and stability. Such studies are critical for understanding the role of the endothelium in vascular bed–specific thrombosis. Furthermore, our system enables the examination of how interactions between soluble clotting factors and vascular cells promote clot formation. Effects of hypercoagulability and endothelial dysfunction may be additive or even synergistic with regard to thrombin generation and fibrin production. Studies on the combined effects of these cellular and plasma activities remain interesting next steps in these investigations.

The authors thank Laura Gray, Brittany Larson, Colleen McGary, and Kellie Machlus for their excellent technical assistance.

This study was supported by research funding from the National Institutes of Health (K01AR051021 [A.S.W.] and T32ES007017 [R.A.C.]), the Gustavus and Louise Pfeiffer Research Foundation (A.S.W.), the National Hemophilia Foundation (A.S.W.), and Novo Nordisk (A.S.W.).

National Institutes of Health

Contribution: R.A.C. designed and performed the experiments, analyzed and interpreted the data, and wrote the manuscript; K.A.O. designed and performed the experiments, and analyzed and interpreted the data; C.H.S. and B.C.S. contributed vital reagents; and A.S.W. designed the research, analyzed and interpreted data, and wrote the manuscript.

Conflict-of-interest disclosure: A.S.W. received research funding from Novo Nordisk. The remaining authors declare no competing financial interests.

R.A.C.'s current affiliation is Department of Human Molecular Biology and Genetics, University of Utah, Salt Lake City. K.A.O.'s current affiliation is Department of Molecular and Integrative Physiology, University of Michigan, Ann Arbor. C.H.S.'s current affiliation is Division of Cardiothoracic Surgery, Department of Surgery, University of Utah, Salt Lake City.

Correspondence: Alisa S. Wolberg, Department of Pathology and Laboratory Medicine, University of North Carolina at Chapel Hill, 815 Brinkhous-Bullitt Bldg, CB #7525, Chapel Hill, NC 27599-7525; e-mail: alisa_wolberg@med.unc.edu.