Down-regulation of Monocyte Tissue Factor Mediated by Tissue Factor Pathway Inhibitor and the Low Density Lipoprotein Receptor-related Protein*

Inflammatory mediators like bacterial lipopolysaccharide induce monocytes to express tissue factor (TF), the cell-surface protein that triggers the blood clotting cascade in hemostasis and thrombotic disease. The physiologic ligand for TF is the serine protease, factor VIIa (FVIIa), and the resulting bimolecular enzyme, TF/FVIIa, can be reversibly inhibited by tissue factor pathway inhibitor (TFPI). Culturing monocytic cells in the presence of both FVIIa and TFPI caused down-regulation of TF expression via reducing its half-life. To exert this effect, FVIIa had to be competent to bind both TF and TFPI, and TFPI had to contain the C-terminal domain required for binding to other cell-surface receptors, including the low density lipoprotein receptor-related protein (LRP). TF down-regulation by FVIIa plus TFPI was abrogated by the 39-kDa receptor-associated protein, which blocks binding of all known ligands to LRP. Furthermore, treatment with FVIIa plus TFPI caused monocyte TF to colocalize with α-adaptin, a component of clathrin-coated pits. Thus, in addition to reversibly inhibiting TF/FVIIa catalytic activity, TFPI also mediates the permanent down-regulation of cell-surface TF in monocytic cells via LRP-dependent internalization and degradation. This represents an unusual mechanism for receptor internalization, requiring ligand-dependent bridging of one cell-surface receptor (TF) to a second cell-surface receptor (LRP), the latter being capable of clathrin-mediated internalization.

Tissue factor (TF) 1 is the integral membrane protein that initiates the blood clotting cascade in hemostasis and many thrombotic disorders (1). TF allosterically enhances the enzyme activity of the plasma serine protease, factor VIIa (FVIIa), and the resulting TF⅐FVIIa complex activates coagulation factors IX and X by limited proteolysis. Recent studies have also shown that TF can act as a signaling receptor upon binding FVIIa (2). Although TF is normally absent from cells in contact with the plasma, monocytes can be induced to express cell-surface TF by several inflammatory mediators, including bacterial lipopolysaccharide (LPS) (3). Expression of TF on circulating monocytes is thought to drive the life-threatening coagulopathy observed in sepsis (4,5) and to cause thrombosis in other disease states such as unstable angina (6). TF is also expressed in monocytes/macrophages resident in atherosclerotic plaques (7,8), where it can trigger thrombus formation following plaque fissure or rupture. It is therefore important to understand how TF expression is controlled in monocytic cells. Several studies have focused on transcriptional control of TF gene expression (reviewed in Ref. 9) and mRNA half-life (10). However, much less is known about post-translational control of TF expression, especially in monocytic cells.
An important physiologic regulator of the enzymatic activity of TF/FVIIa is the protease inhibitor, tissue factor pathway inhibitor (TFPI). TFPI circulates in association with lipoproteins and is present in platelets. In addition, a large pool of TFPI is apparently bound to the vascular endothelium in vivo, which is releasable by heparin injection (11). TFPI has three Kunitz-type inhibitor domains, allowing it to inhibit factor Xa (FXa) and TF/FVIIa simultaneously (12). Whereas the TFPI⅐FXa complex is an especially potent inhibitor of TF/ FVIIa, TFPI can also inhibit TF/FVIIa directly. TFPI binds to two classes of cell-surface proteins, heparan sulfate proteoglycans (HSPGs) and the low density lipoprotein receptor-related protein (LRP) (13). LRP is responsible for mediating clathrindependent endocytosis and subsequent degradation of TFPI by hepatoma cells (14). In addition, TFPI mediates the uptake and degradation of FXa by fibroblasts and hepatoma cells (15). In cultured vascular endothelial and smooth muscle cells, TF is localized, via TFPI-dependent (16) or -independent (17) means, to caveolae, specialized membrane domains whose phospholipid content differs significantly from the remainder of the plasma membrane (18).
In the present study we investigated the role of TFPI in down-regulation of TF expression in human peripheral blood monocytes and the human monocytic leukemia cell line, Mono Mac 6 (MM6) (19). When FVIIa and TFPI bound to TF on the surface of monocytic cells, the half-life of TF was shortened from 3.7 to 1.3 h. Down-regulation of TF expression was dependent upon both FVIIa and TFPI and required that TFPI have an intact C terminus. In contrast to endothelial and smooth muscle cells, no evidence was obtained in monocytes for translocation of TF to caveolae or other membrane domains with altered phospholipid content. However, TF down-regulation by TFPI plus FVIIa in monocytic cells was dependent upon LRP. We propose that when TF/FVIIa on the surface of monocytic cells binds TFPI, the resulting TF⅐FVIIa⅐TFPI complex associates with LRP and is translocated to clathrin-coated pits. TF is thereby internalized and degraded by an unusual mechanism in which one integral membrane protein (TF) is bridged, in a ligand-dependent fashion, to a second integral membrane protein (LRP), the latter promoting clathrin-mediated internalization.
Recombinant TFPI (20), TFPI 13-160 (a C-terminal truncated form of TFPI (21)), and XK1 (FX/TFPI chimera (22)) were generous gifts from Dr. George Broze, Jr. (Washington University, St. Louis). The 39-kDa receptor-associated protein (RAP) and rabbit anti-LRP IgG were prepared as described (23). FX was purified from human plasma and activated to FXa as described (24,25). Human plasma FVII was purified and activated as described (26). FFR-FVIIa was prepared by reaction of FVIIa with FFR-chloromethyl ketone (1:20 molar ratio) for 30 min at room temperature in 100 mM NaCl, 50 mM Tris-HCl, pH 7.5, 0.5% human serum albumin, followed by extensive dialysis to remove unreacted inhibitor. FFR-FVIIa had less than 1% of its original amidolytic activity (27). Recombinant E 2 PD-FVIIa was constructed and purified as described for E 2 PD-FX (28) and was a gift from Dr. Pierre Neuenschwander. TF was purified from human brain by affinity chromatography as described (29) except that antibody TF9 -5B7 was used in place of TF8 -5G9. Recombinant soluble TF was prepared as described (30). Preparation of polyclonal anti-TF IgG and anti-TF mAb were described previously (31). The CHO-B cDNA probe was a gift from Dr. R. Walls of the UCLA School of Medicine (Los Angeles, CA). All proteins except TFPI, XK1, and RAP were supplemented with 0.5% human serum albumin and dialyzed against sterile phosphate-buffered saline before use in cell cultures. Stock solutions of TFPI (8 mg/ml) and TFPI 13-160 (4.7 mg/ml) were in 2 M urea/phosphate buffered saline and were diluted directly into culture medium when used. Controls with similarly diluted urea were performed.
Cell Culture-MM6 cells were grown at 37°C in 5% CO 2 in RPMI-1640, 10% calf serum supplemented with 50 g/ml gentamicin. For experiments, cells were rinsed once with Ca 2ϩ ,Mg 2ϩ -free HBSS and once with AIM-V medium, and then resuspended in AIM-V, 0.01% calf serum unless otherwise indicated. MM6 cells were then cultured in multiwell plates. AIM-V has been used for long term culture of macrophages (32). We found that MM6 cells cultured in AIM-V for 10 days were viable and grew at a rate only slightly lower than cells cultured in RPMI/10% calf serum (data not shown). Platelet-poor human plasma was prepared by centrifugation of heparinized blood for 15 min at 300 ϫ g and recentrifugation of plasma for 15 min at 5,000 ϫ g.
Human peripheral blood mononuclear cells (PBMC) were isolated from heparinized blood by centrifugation using Histopaque, per the manufacturer's instructions, then suspended in AIM-V, 0.01% calf serum, and cultured in multiwell plates or sterile polypropylene culture tubes at 37°C in 5% CO 2. The cells were used in experiments immediately after isolation.
TF ELISA-TF was solubilized from cells and culture media by adding EDTA (7 mM final) and either Triton X-100 (1% final) or octyl glucoside (60 mM final), after which TF levels were measured by a sandwich ELISA (30). EDTA was included to dissociate FVIIa and TFPI from TF, to eliminate competition with anti-TF antibodies. Control experiments in which FVIIa and TFPI were added to lysates of TFexpressing monocytes confirmed that these proteins did not interfere with TF measurement by ELISA in the presence of EDTA. Statistical significance of differences in ELISA values was determined using Student's t test.
For measurement of the half-life of TF, cells were treated for 4 h with LPS and then treated with 10 g/ml cycloheximide; cultures were immediately divided; TFPI and FVIIa were added as indicated, and timed aliquots were removed and processed for ELISA. The effect of staurosporine on TF levels was investigated by incubating LPS-stimulated cells as above for 3.5 h, adding 2 M staurosporine or vehicle (ethanol) and incubating for an additional 30 min. FVIIa and TFPI were then added for 2 h, after which cells were collected for ELISA.
Western Blot Analysis-Cells were rinsed with phosphate-buffered saline, resuspended at 1 ϫ 10 7 cells/ml in SDS sample buffer with dithiothreitol, and boiled for 5 min. Proteins were resolved on SDS-PAGE and blotted onto polyvinylidene difluoride membranes, which were then treated with blocking buffer (4% bovine serum albumin and 1% casein in 50 mM Tris-HCl, pH 7.4, 100 mM NaCl, 0.05% Tween 20). Membranes were subsequently treated with 0.5 g/ml biotinylated polyclonal anti-TF IgG in blocking buffer for 1 h at 37°C, rinsed three times with 50 mM Tris-HCl, pH 7.4, 100 mM NaCl, 0.05% Tween 20, and incubated with NeutrAvidin-horseradish peroxidase in blocking buffer for 1 h at room temperature. In some experiments, membranes were co-stained with polyclonal rabbit anti-caveolin antibody, using horseradish peroxidase-conjugated goat anti-rabbit as the secondary antibody. Human endothelial cell lysate as positive control for caveolin staining was provided by the supplier of the anti-caveolin antibody. Membranes were rinsed as above and treated with chemiluminescence substrate (3 ml per 54-cm 2 membrane) for 5 min, after which they were exposed to Kodak BIOMAX x-ray film.
RNA Isolation, Blotting, and Hybridization-RNA was extracted from 1 ϫ 10 6 cells using RNAzol according to the manufacturer's instructions, resolved by electrophoresis on formaldehyde/agarose gels, and blotted onto polyvinylidene difluoride membranes. Membranes were hybridized at 65°C for at least 24 h with radiolabeled TF or CHO-B cDNA probes and quantified using a Molecular Dynamics Phos-phorImager and ImageQuaNT software, with TF signals normalized to CHO-B.
Isolation of Triton-insoluble Membrane Fractions-The caveolar membrane isolation protocol of Lisanti et al. (33) was modified as follows: MM6 cells were rinsed with ice-cold phosphate-buffered saline and resuspended in 1% Triton X-100 in MES-buffered saline (25 mM MES, pH 6.5, 150 mM NaCl) containing 1 mM each NaVO 4 and phenylmethylsulfonyl fluoride. Homogenates were adjusted to 45% sucrose, overlaid with 4 ml each 35 and 5% sucrose (in MES-buffered saline), and centrifuged 16 -20 h at 39,000 rpm in a Beckman SW41 rotor. Triton-insoluble membranes, including caveolae, float to the 5/35% interface (34). Gradient fractions (1 ml) were assayed for light scattering at 600 nm, TF content by ELISA, and protein content (bicinchoninic acid assay). The densest three fractions contained approximately 500 -700 g of protein, and the lighter fractions contained Յ100 g of protein. Proteins were precipitated by 10% trichloroacetic acid and resuspended in SDS sample buffer. The entire fraction or 100 g of protein (whichever was less) was resolved by SDS-PAGE and Westernblotted as above.
Immunofluorescence-MM6 cells were rinsed with ice-cold HBSS plus 0.5% bovine serum albumin and incubated with 10 g/ml biotinylated goat anti-TF IgG (45 min at 4°C). After rinsing, cells were incubated with streptavidin-Oregon Green (45 min at 4°C), rinsed again, and fixed in 3.7% paraformaldehyde (10 min at 4°C). The fixative was rinsed away, and cells were permeabilized with 0.05% Triton X-100 in HBSS (10 min at 4°C). After rinsing, cells were incubated 45 min on ice with HBSS plus 0.2% bovine serum albumin and 10 g/ml anti-␣adaptin mAb, rinsed in HBSS, 0.5% donkey serum, and then treated with donkey anti-mouse IgG conjugated to Cy-3. After final rinsing, cells were collected onto a slide by cytospin centrifugation and mounted in 9:1 glycerol:HBSS, pH 8.5. Dual immunofluorescence detection of TF and ␣-adaptin was performed with a Bio-Rad MRC 1024 confocal microscope (Bio-Rad Life Science Group, Hercules, CA), equipped with a krypton/argon laser. A series of 0.2-m optical sections in the z axis were analyzed with Confocal Assistant software written by T. C. Berlje (University of Minnesota). Images were pseudo-colored and merged using Adobe Photoshop software (Adobe Systems, Inc., Mountain View, CA). Specificity of anti-TF staining was demonstrated by adding a large excess (100 g/ml) of recombinant soluble TF, which reduced the level of staining to that observed in the absence of primary antibody.
The degree of colocalization of TF with ␣-adaptin was quantified using an image analysis protocol described elsewhere (35). Briefly, the original confocal images were converted into binary images with Adobe Photoshop using a thresholding tool, such that no additional pixels were introduced into the signal after binarization. For these experiments, the resolution was approximately 0.2 m/pixel. The individual binarized TF images were then multiplied with binarized ␣-adaptin signals at a 50% scale, so that the TF pixels remained white when they colocalized with ␣-adaptin pixels, but were gray when they did not colocalize. Colocalization was calculated as the percentage of white pixels relative to total pixels (gray plus white). For these analyses, a total of 6 -11 cells and 5609 -10,883 pixels were examined. (36) and in monocytic cell lines such as THP-1 (10) and MM6 (37). In many published studies, monocytes and monocytic cell lines have been cultured in media supplemented with serum. In this study, when MM6 cells were cultured in the serum-free medium AIM-V, we observed that adding very low concentrations of calf serum (0.01% and below) enhanced TF expression in response to LPS, whereas higher concentrations of serum (0.05-10%) reduced TF expression relative to that seen with 0.01% serum (Fig. 1). Low levels of serum may enhance the responsiveness of MM6 cells to LPS by providing a source of LPS-binding protein (38), but it was unclear why higher concentrations of serum caused MM6 cells to express lower levels of TF following LPS stimulation.

Effect of TFPI and FVIIa on TF Expression-LPS induces TF expression in cultured peripheral blood monocytes
Serum contains FVIIa, the primary ligand for TF, and also TFPI, a reversible protease inhibitor that binds to the TF⅐FVIIa complex. Accordingly, we hypothesized that these molecules may bind to TF on monocytic cells and promote the downregulation of TF expression. MM6 cells or PBMC were cultured in AIM-V medium supplemented with 0.01% calf serum to support optimal TF expression, typically measured 4 -6 h after stimulating the cells with 10 ng/ml LPS. (0.01% serum contributes negligible levels of FVIIa or TFPI.) Cells incubated with TFPI after LPS stimulation expressed TF at levels comparable to cells treated with LPS alone (Fig. 2). Cells treated with FVIIa had very slightly reduced levels of TF. However, MM6 cells and PBMC treated with both TFPI and FVIIa had TF levels that were typically 50 -60% of cells treated with LPS alone (Fig. 2). TFPI can react both with FXa and with FVIIa bound to TF. In some settings, the TFPI⅐FXa complex has a higher affinity for TF/FVIIa than does TFPI alone (39). However, addition of FXa to TFPI and FVIIa resulted in levels of TF expression by MM6 cells and PBMC that were similar to those of cells treated with TFPI and FVIIa without FXa. To simplify interpretation of results, most of the rest of the experiments in this study were therefore performed using TFPI and FVIIa without added FXa.
The time course of TF expression in MM6 cells in response to LPS stimulation was altered when TFPI and FVIIa were included in the culture medium. Thus, MM6 cells cultured in the absence of these two proteins expressed more TF, and did so over a longer period of time, than did cells treated with a combination of TFPI and FVIIa (Fig. 3). However, Northern blot analysis of RNA isolated 1 h after LPS stimulation revealed that TFPI and FVIIa had no effect on TF mRNA levels in MM6 cells (data not shown), indicating that these ligands affected a later stage of TF biosynthesis and/or degradation. Accordingly, the effect of TFPI and FVIIa on the half-life of TF protein in MM6 cells was examined by inducing TF expression with LPS, after which cycloheximide was added to block further protein synthesis. TF antigen levels were then quantified by ELISA (Fig. 4). Treating cells with TFPI and FVIIa reduced the half-life of TF protein 3-fold compared with control cells (mean half-lives of 3.7 h versus 1.3 h; n ϭ 3). In subsequent experiments, the effect of TFPI and FVIIa on TF expression was evaluated in cells incubated for 6 h, since this time point consistently showed the largest difference in TF levels with and without treatment.
Effect of TFPI and FVIIa on Detergent Solubility of TF-Previous studies showed that TF on the surfaces of endothelial cells (16) and smooth muscle cells (17) is associated, at least in part, with caveolae. In endothelial cells, translocation of TF from Triton-soluble to Triton-insoluble membrane fractions (associated with caveolae (18)) was induced by the addition of FVIIa and FX and required endogenous TFPI (16). It is not clear that caveolae per se exist in monocytes. However, if TFPI and FVIIa can induce the translocation of TF to Triton-insoluble membrane domains in monocytes and MM6 cells, it might render this protein undetectable by our ELISA.
We investigated whether the decrease in measured TF levels was due to insolubility in Triton X-100 in three ways. First, we compared octyl glucoside versus Triton X-100 lysis of MM6 cells, since octyl glucoside has been shown to solubilize caveolar proteins (40). However, TF levels in MM6 cells treated with TFPI and FVIIa and lysed with octyl glucoside were indistinguishable from cells lysed with Triton X-100 (data not shown). Second, MM6 cells were treated or not with TFPI and FVIIa, and cells were lysed by boiling in SDS sample buffer followed by SDS-PAGE and Western blotting. Consistent with the ELISA results, TF levels were much lower in lysates from cells treated with TFPI and FVIIa compared with control cells (data not shown). Third, association of TF with Triton-insoluble membrane fractions was examined by lysing MM6 cells with Triton X-100 followed by fractionation on sucrose density gradients. This method has been used in other cell types to separate caveolar membranes and their resident proteins from the remainder of the plasma membrane; the low density, light scattering fractions contain Triton-insoluble membranes and caveolin (33). When MM6 lysates were fractionated by density gradient centrifugation and TF antigen levels were quantified by ELISA or Western blot, essentially all of the TF was associated with the high density fractions (fractions 6 -9) that con-tain Triton-solubilized membrane proteins (Fig. 5). Negligible amounts of TF antigen (less than 2%) colocalized with the light scattering, low density fractions under any of the assay conditions. In these experiments, cells were lysed 15 min after addition of TFPI, FVIIa, and FXa, since this was sufficient to cause translocation of TF from high density to low density membrane fractions in endothelial cells (16). However, MM6 cells lysed 6 h after treatment with TFPI and FVIIa (with or without FXa) gave equivalent results (data not shown). Additionally, caveolin could not be detected in MM6 cell extracts by Western blotting.
Structural Requirements of TFPI and FVIIa for Down-regulation of TF-The studies above indicated that TF levels were down-regulated in monocytes and MM6 cells in response to FVIIa and TFPI, resulting from a decreased half-life of TF protein and not from translocation of the protein to caveolae or other Triton-insoluble membrane fractions. We hypothesized that FVIIa and TFPI both bound to TF on the cell surface and that this binding event altered the half-life of TF. Alternatively, however, FVIIa and/or TFPI might interact with other receptors or binding sites on the cell surface which might indirectly cause the observed reduction in TF half-life.
In order to elucidate the structural requirements for the interaction between TFPI, FVIIa, and TF responsible for the reduction in TF half-life, we employed altered forms of FVIIa with modified binding characteristics. These included activesite blocked FVIIa (FFR-FVIIa), which binds TF with high affinity but which cannot bind TFPI (41), and a recombinant, truncated form of FVIIa (E 2 PD-FVIIa), which lacks the 4-carboxyglutamate-rich domain and the first EGF-like domain. E 2 PD-FVIIa is unable to bind to either the plasma membrane or to TF, as demonstrated by functional assays and dot blots. 2 However, because the active site of E 2 PD-FVIIa is intact, it can still bind to, and be inhibited by, TFPI (data not shown). Unlike cells treated with TFPI plus FVIIa, cells treated with a combination of TFPI and FFR-FVIIa, or with TFPI and E 2 PD-FVIIa, expressed TF at levels that were indistinguishable from those of control cells (Fig. 6). This indicates that FVIIa must be competent to bind TFPI and that TFPI⅐FVIIa complexes must be competent to bind to TF, in order to cause the observed reduction in TF half-life.
To examine further the structural requirements for the interaction between TFPI, FVIIa, and TF responsible for the reduction in TF half-life, we employed TFPI 13-160 , a recombinant, truncated form of TFPI lacking both the third Kunitz domain and the C-terminal region of the protein. TFPI  inhibits the TF⅐FVIIa complex as readily as full-length TFPI (42) but is unable to bind specifically to receptors on hepatoma cells (21). In addition, we also employed a hybrid protein, XK1, that consists of the light chain of FXa and the first Kunitz domain of TFPI (22). The first Kunitz domain of TFPI mediates reversible binding to the TF⅐FVIIa complex (43), whereas the light chain of FXa mediates binding to membranes containing exposed negatively charged phospholipids. Because of these properties, XK1 is an especially efficient inhibitor of the TF⅐FVIIa complex (22). However, like TFPI 13-160 , XK1 lacks the C-terminal region of TFPI. Both TFPI  and XK1 inhibited the enzymatic activity of purified TF⅐FVIIa complexes in vitro (data not shown).
Unlike cells treated with FVIIa plus TFPI, when LPS-stimulated MM6 or PBMC were treated with a combination of FVIIa and either XK1 or TFPI 13-160 , they expressed TF at levels that were essentially the same as cells treated with LPS alone (Fig. 7). This indicates that inhibition of TF/FVIIa enzymatic activity by TFPI is not sufficient to down-regulate TF levels and that the intact C terminus of TFPI is required.
RAP Abrogates TF Down-regulation-In hepatoma cells, TFPI has been shown to bind to at least two classes of cell membrane proteins, HSPGs and LRP (13). In those cells, TFPI binds to LRP with a K d of 2.3 nM (14) and to HSPGs with a K d of about 30 nM, which outnumber LRP by about 10-fold (44). It is thought that both receptor systems may function to clear TFPI from the plasma in vivo (13), but LRP is required for the cellular uptake and degradation of TFPI by hepatoma cells in vitro (14). Previous studies have shown that TFPI must have an intact C-terminal region in order to bind to HSPGs or LRP (13,21), matching the requirements found above for mediating TF down-regulation. We hypothesized, therefore, that the TF⅐FVIIa⅐TFPI complex interacts with HSPGs or LRP via the C-terminal region of TFPI and that this binding is essential in down-regulating TF expression. Degradation of TFPI is inhibited by RAP (14), a protein that inhibits binding and uptake by all known ligands that bind to LRP (45,46), whereas protamine sulfate inhibits the binding of TFPI to HSPGs (13). Therefore, in order to determine whether binding of TFPI to HSPGs or LRP is involved in the down-regulation of TF in monocytic cells, we conducted experiments in the presence of protamine sulfate or RAP (Fig. 8). Protamine sulfate failed to block the ability of TFPI plus FVIIa to down-regulate TF expression. In fact, MM6 cells treated with protamine sulfate (with TFPI plus FVIIa) exhibited about 11% lower TF levels relative to cells treated with TFPI and FVIIa alone (p Ͻ 0.005). In hepatoma cells, protamine sulfate increases by 5-fold the amount of TFPI that can be cross-linked to LRP and enhances LRP-mediated degradation of TFPI (14). On the other hand, RAP completely abrogated the effect of TFPI plus FVIIa on TF levels in both MM6 cells and monocytes (Fig. 8). This indicated that the TF⅐FVIIa⅐TFPI complex interacts with LRP in order to effect the down-regulation of TF levels, possibly via endocytosis with subsequent degradation of TF.
TFPI and FVIIa Promote the Colocalization of TF with ␣-Adaptin-LRP has been localized to clathrin-coated pits in both F9 (47) and glioblastoma cells (48), so this would be a plausible pathway for internalization and degradation of TF. The protein kinase C inhibitor, staurosporine, inhibits internalization of various cell-surface receptors, including LRP (49). When added to cells 30 min prior to TFPI and FVIIa, 2 M staurosporine completely abrogated the down-regulation of TF induced by TFPI and FVIIa (data not shown). This effect of staurosporine suggested that TFPI-mediated loss of TF might involve internalization, possibly by directing TF to clathrincoated pits.
Since staurosporine can affect multiple cellular processes, we used dual-label immunofluorescence to examine more directly whether TFPI plus FVIIa could promote the colocalization of TF with ␣-adaptin, a component of the adaptor protein-2 complex which associates with clathrin-coated pits of the plasma membrane (50). MM6 cells were stimulated with LPS for 3.5 h, after which FVIIa and either TFPI or TFPI 13-160 were added. The cells were incubated for a further 30 min before being processed for immunofluorescence labeling, as described under "Experimental Procedures." In control cells (not treated with FVIIa or TFPI), TF was distributed in a granular pattern that became more punctate when the cells were treated with a combination of FVIIa and either TFPI or TFPI   (Fig. 9, top row; green). When the staining pattern of anti-␣-adaptin (Fig.  9, middle row; red) was merged with the pattern of anti-TF mAb (green), significant colocalization was observed in cells treated with a combination of FVIIa and TFPI, as evidenced by the appearance of yellow color (Fig. 9, bottom row). In contrast, little or no yellow staining was observed in control cells or in cells treated with a combination of FVIIa and TFPI   (Fig.  9, bottom row). This was confirmed by quantifying the degree of colocalization of green and red pixels in the images as described under "Experimental Procedures." Control cells had only 3.2 Ϯ 0.3% (mean Ϯ S.E.) colocalization of green pixels (TF) with red pixels (␣-adaptin), and cells treated with FVIIa and TFPI 13-160 had only 9.6 Ϯ 2.0% colocalization. In contrast, cells treated with FVIIa and TFPI had 26.6 Ϯ 4.4% colocalization of TF and ␣-adaptin signals. This indicates that, as with TF down-regulation, the intact C terminus of TFPI is required to promote colocalization of TF with ␣-adaptin, a component of clathrincoated pits. No fluorescent signal was observed with monoclonal anti-caveolin staining of these cells (not shown), consistent with the results above from Western blot analysis of MM6 cell lysates.

DISCUSSION
In vitro, endothelial cells and monocytes express TF in response to a variety of inflammatory mediators, including LPS (reviewed in Refs. 9 and 51). In vivo, TF expression by endothelial cells is very rare, even in models of lethal septic shock and endotoxemia (52)(53)(54). On the other hand, TF expression on circulating monocytes has been well documented in animal models of bacteremia and LPS treatment and in humans with sepsis (52,55,56). In addition, it is clear that the intravascular coagulation associated with sepsis and endotoxemia is TF-dependent (4,5). Since these studies strongly argue that induced expression of TF on circulating monocytes drives the coagu-lopathies observed in sepsis, it is clearly important to understand how the LPS-induced expression of TF in monocytes is regulated.
In the present study, we found that expression of TF in monocytes and MM6 cells in response to LPS was modulated by culturing the cells in the presence of the plasma proteins, FVIIa and TFPI. Neither FVIIa nor TFPI by themselves was able to substantially alter the expression of TF in monocytic cells, but together they caused a 3-fold reduction in TF half-life. FVIIa is the physiologic ligand for TF, whereas TFPI is a reversible inhibitor of the enzymatic activity of the TF⅐FVIIa complex. Incubating monocytes expressing TF with a combination of FVIIa and TFPI therefore results in the formation of a trimolecular complex (TF⅐FVIIa⅐TFPI) on the cell surface. The studies presented here indicate that, in addition to inhibiting the enzymatic activity of the TF⅐FVIIa complex, TFPI promotes the permanent removal of cell-surface TF in monocytic cells via internalization and subsequent degradation.
Previous studies with endothelial and smooth muscle cells have shown that TF can localize to caveolae and that translocation of TF to caveolae is promoted by TFPI (16). Localization of TF to caveolae or similar Triton-insoluble membrane microdomains has not previously been reported in monocytes. In the present study we were able to demonstrate the existence of Triton-insoluble membrane fractions in MM6 cells, although these cells failed to stain with antibodies to caveolin. It is presently unclear whether or not monocytic cells have caveolae, although our observations are consistent with previous reports that peripheral blood monocytes do not express caveolin (57) and that THP-1 cells (a human monocyte-macrophage cell line) do not appear to have membrane domains with the ultrastructural configuration of caveolae (58). In any case, we found no evidence for association of TF in MM6 cells with Triton-insoluble microdomains, either in the presence or absence of FVIIa and TFPI.
TFPI-mediated down-regulation of TF was only observed when cells were treated with forms of FVIIa that are competent to bind to both TF and TFPI. TFPI contains multiple Kunitz domains and can inhibit both FVIIa and FXa simultaneously. Thus, TFPI can form trimolecular complexes with TF (TF⅐FVIIa⅐TFPI) as well as tetramolecular complexes (TF⅐FVIIa⅐TFPI⅐FXa; see Fig. 10 for a schematic diagram). We found that TFPI-mediated down-regulation of TF was independent of FXa. FX/FXa binds to multiple cell-surface proteins including protease-nexin 1 (59), EPR-1 (60), and Mac-1 (61), and is mitogenic for some cell types (62). On hepatoma cells and fibroblasts, FXa also directs the internalization and degradation of the complex of FXa and TFPI by a mechanism different from that of free TFPI (15). Since FXa was not essential for observing TF-mediated down-regulation, and since it has potentially other confounding effects on cells, we chose to focus most of our studies on the effect of FVIIa and TFPI on TF expression. However, our studies clearly demonstrate that TFPI plus FVIIa can mediate TF down-regulation in the presence of FXa as well.
Down-regulation of TF by TFPI required the C-terminal domain of TFPI, since truncated forms of this protein (TFPI  and XK1) were ineffective despite the fact that they are potent inhibitors of TF/FVIIa enzymatic activity. The C-terminal domain of TFPI mediates binding to both HSPGs and LRP, which were therefore candidates for mediating its effect on TF half-life. Protamine sulfate, which competes with TFPI for binding to HSPGs (13), was unable to inhibit the down-regulation of TF by TFPI. Thus, the down-regulation of TF by TFPI does not appear to be dependent upon TFPI interaction with HSPGs. RAP is a 39-kDa protein that blocks binding of all known ligands, including TFPI, to LRP (45,46). As LRP mediates the internalization and degradation of TFPI in hepatoma cells (13), we hypothesized that binding of TFPI to the TF⅐FVIIa complex could result in the loss of TF through LRP-mediated internalization and subsequent degradation. Many cell types express LRP, including peripheral blood monocytes and monocytic cell lines (45,63,64). MM6 cells also express LRP, as determined by flow cytometric analysis of cells stained with rabbit anti-LRP IgG (data not shown). Our studies showed that RAP abrogated the effect of TFPI on TF expression levels, consistent with an essential role for LRP in the process of TF downregulation mediated by TFPI and FVIIa.
Following treatment of cells with a combination of TFPI and FVIIa, TF colocalized with ␣-adaptin, a component of clathrincoated pits of the plasma membrane. We propose that binding of the C-terminal domain of TFPI to LRP directs the TF⅐FVIIa⅐TFPI complex to clathrin-coated pits, where the com-plex can be internalized (see Fig. 10). This represents an unusual mechanism for receptor internalization and degradation, in which one cell-surface receptor protein (TF) is bridged via a divalent ligand to a second cell-surface receptor protein (LRP), the latter being capable of undergoing clathrin-mediated internalization. A situation reminiscent of this exists for the LRPmediated internalization of the complex of urokinase, plasminogen activator inhibitor type-1, and the urokinase plasminogen activator receptor (uPAR), which also employs a ligand-dependent bridging mechanism (63). However, whereas the protease⅐inhibitor complex bound to uPAR is apparently degraded within lysosomes, the glycosylphosphatidylinositollinked receptor (uPAR) is recycled back to the cell surface within 2 h (65). TF, however, is apparently not recycled. Although not investigated here, it is likely that FVIIa and TFPI are also not recycled and that LRP-mediated internalization of TF results in destruction of the entire receptor-protease-inhibitor (TF⅐FVIIa⅐TFPI) complex by directing it to lysosomes. FIG. 9. TFPI and FVIIa promote colocalization of TF with ␣-adaptin. MM6 cells were stimulated with 10 ng/ml LPS for 3.5 h, after which a combination of 50 nM FVIIa and either 75 nM TFPI or 75 nM TFPI 13-160 were added, as indicated. Control cells received neither FVIIa nor TFPI. After 30 min, cells were collected and incubated with biotinylated polyclonal antibody to TF, followed by streptavidin-Oregon Green. Cells were then fixed, permeabilized, and incubated with mAb to ␣-adaptin, followed by donkey anti-mouse IgG conjugated to Cy-3. Confocal microscopy, employing optical sections near the basal portion of MM6 cells, revealed staining for TF (green, top row) or ␣-adaptin (red, middle row). When the images were merged (bottom row), colocalization of staining for TF and ␣-adaptin (yellow) was readily observable in cells treated with TFPI and FVIIa but not in control cells or cells treated with a combination of TFPI  and FVIIa. The bar represents 20 m.
FIG. 10. Schematic diagram of complexes formed with LRP. A, the active site of FVIIa (in the TF⅐FVIIa complex) can interact with the first Kunitz domain of TFPI (K 1 ) to form the TF⅐FVIIa⅐TFPI complex. This trimolecular complex then associates with LRP via the C-terminal region of TFPI (indicated with the letter C). B, alternatively, TFPI can interact with the active site of FXa via its second Kunitz domain (K 2 ), and the resulting TFPI⅐FXa complex can interact with the TF⅐FVIIa complex via the first Kunitz domain of TFPI as in A. The tetramolecular complex consisting of TF⅐FVIIa⅐TFPI⅐FXa then associates with LRP via the C-terminal domain of TFPI as in A. We propose that when either of these receptor-protease-inhibitor complexes binds to LRP, the entire complex is translocated to clathrin-coated pits for internalization and degradation of TF. C, multiple copies of RAP can bind to LRP, blocking binding of all known ligands to this receptor, including TFPI (reviewed in Ref. 46).
The plasma concentration of TFPI is about 2.5 nM, which is lower than the concentrations employed in this study. However, most of the TFPI circulating in plasma is truncated at the C terminus, is bound to lipoproteins, and has very low specific activity (11). In vivo, a major pool of highly active, full-length TFPI is apparently bound to the endothelium and is releasable by administration of heparin (11). Moreover, there is evidence that TF and TFPI may be expressed simultaneously by LPSinduced or adherent monocytes (54,66). Additionally, fulllength TFPI is released by platelets upon stimulation with thrombin (67). Therefore, local concentrations of TFPI at sites of wounds or infection may be much higher than in plasma. Therapeutic levels of TFPI shown to be effective in reducing mortality in animal models of sepsis and other diseases range from 25 to 90 nM (68 -70), consistent with the TFPI concentrations employed here. In vivo, TFPI regulates TF/FVIIa function by inhibiting its enzymatic activity. However, TFPI is a reversible protease inhibitor, and it is conceivable that TFPI and/or TFPI⅐FVIIa complexes may dissociate from cell-surface TF. In this case, the newly exposed TF molecules would become available to trigger the clotting cascade again. In the present study, we have described an additional mechanism for down-regulation of TF function by TFPI. Thus, TFPI (or the TFPI⅐FXa complex) binds to the TF⅐FVIIa complex on cell surfaces and, via interaction with LRP, causes the entire complex to be translocated to clathrin-coated pits. TF complexes are then internalized and degraded, resulting in permanent down-regulation of coagulant activity in monocytes and monocytic cells.