Tissue Factor Pathway Inhibitor Inhibits Endothelial Cell Proliferation via Association with the Very Low Density Lipoprotein Receptor*

Tissue factor pathway inhibitor (TFPI) contains three Kunitz-type proteinase inhibitor domains and is a potent inhibitor of tissue factor-mediated coagulation. Here, we report that TFPI inhibits the proliferation of basic fibroblast growth factor-stimulated endothelial cells. A truncated form of TFPI, containing only the first two Kunitz-type proteinase inhibitor domains, has very little antiproliferative activity, suggesting that the carboxyl-terminal region of TFPI is responsible for this activity. Binding studies revealed that full-length TFPI, but not the truncated TFPI molecule, is recognized by the very low density lipoprotein receptor (VLDL receptor) indicating that this receptor is a novel high affinity endothelial cell receptor for TFPI. The antiproliferative activity of TFPI on endothelial cells is inhibited by the receptor-associated protein, a known antagonist of ligand binding by the VLDL receptor, and by anti-VLDL receptor antibodies. These results confirm that the antiproliferative activity of TFPI is mediated by the VLDL receptor and suggest that this receptor-ligand system may be a useful target for the development of new anti-angiogenic and antitumor agents.

The extrinsic pathway of blood coagulation is initiated when factor VII (fVII) 1 binds to its cellular receptor, tissue factor (TF). The fVIIa⅐TF complex then functions as a potent enzyme, activating factor X which leads to thrombin generation. In addition to initiating coagulation, recent research suggests that fVIIa⅐TF complexes play an important role in angiogenesis (1)(2)(3). Zhang et al. (1) correlated TF expression in tumor cells with the ability of the tumors to secrete vascular endothelial growth factor (VEGF) and, in turn, to induce an angiogenic response when implanted in immunodeficient mice. Although TF is not normally expressed on the surface of vascular endothelial cells, in situ hybridization studies have detected TF mRNA in tumor-associated endothelial cells from patients with invasive breast cancer (2). Finally, in a TF-dependent metastasis model, the binding and proteolytic activity of VIIa was shown to be necessary for the initial steps of tumor metastasis (3).
The tissue factor pathway is regulated by a potent inhibitor termed tissue factor pathway inhibitor (TFPI). TFPI forms a tight complex with both fXa and fVIIa leading to their inhibition. The TFPI molecule contains three Kunitz-type domains and a basic carboxyl-terminal region. By employing site-directed mutagenesis, Girard et al. (4) demonstrated that the second Kunitz domain is required for efficient binding and inhibition of fXa, and both Kunitz domains 1 and 2 are required for the inhibition of fVIIa/TF activity. Mutation of the predicted inhibitory residues of the third Kunitz domain had no significant effect on either function of TFPI. Rather, this portion of the molecule binds to the cell surfaces by interacting with cell-surface glycosaminoglycans (5). TFPI can also bind to the low density lipoprotein receptor-related protein (LRP) (6) and, in doing so, mediates the permanent down-regulation of cellsurface TF in monocytic cells via LRP-dependent internalization and degradation (7).
LRP is a member of the LDL receptor family, which includes the LDL receptor, LRP-1b, gp330/megalin, apoER2, and the very low density lipoprotein receptor (VLDL receptor). LRP is widely expressed in fibroblasts, macrophages, smooth muscle cells, neurons, and activated glial cells (8) but does not appear to be highly expressed in vascular endothelium (9). In contrast, VLDL receptor is expressed in HUVECs, in the endothelium of capillaries and small arterioles (10), and on the endothelial surface of bovine arteries (11). This receptor was originally identified by homology cloning (12) and initially was thought to function in the delivery of triglyceride-rich lipoproteins to peripheral tissues (12,13). However, studies in homozygous knockout mice that lack the VLDL receptor gene revealed that the animals had normal levels of plasma triglyceride and cholesterol indicating that in mice the VLDL receptor is not likely a major receptor involved in lipoprotein catabolism (14).
Since fVIIa⅐TF complexes seem to play a role in the pathological proliferation of endothelial cells, we sought to determine whether TFPI, the natural inhibitor of fVIIa/TF, was capable of regulating this process. Our initial studies focused on determining the effect of TFPI on basic fibroblast growth factor (bFGF)-stimulated endothelial cell growth. We report here that TFPI is a potent inhibitor of the growth of bFGF-stimulated endothelial cells. Furthermore, we show that the antiproliferative activity of TFPI is independent of its fVIIa/TF inhibitory activity but rather results from association with the VLDL receptor.

EXPERIMENTAL PROCEDURES
Materials-Human umbilical vein endothelial cells (HUVECs), human microvascular dermal endothelial cells, media for their proliferation (Endothelial Cell Basal Medium (EBM), and Endothelial Cell Growth Medium (EGM) were purchased from Clonetics (San Diego, CA). Rat C6 glioma cells were obtained from ATCC (Manassas, VA). B16BL6 melanoma cell line is a variant cell line derived from B16F10 melanoma cells obtained from the National Cancer Institute-Central Repository (Frederick, MD) certified free of Mycoplasma and other pathogenic murine viruses. Dulbecco's modified Eagle's medium (DMEM), fetal calf serum (FCS), and L-glutamine were obtained from BioWhittaker (Walkerville, MD). TFPI purified from human plasma or HepG2 cells, recombinant full-length TFPI, recombinant truncated TFPI (1-160 a.a), and a rabbit anti-human TFPI were purchased from American Diagnostica Inc. (Greenwich, CT). bFGF was purchased from R & D Systems (Minneapolis, MN). 2ME 2 , heparin (16 -17 kDa), gelatin, and hematoxylin were from Sigma. Cell proliferation enzymelinked immunosorbent assay BrdUrd kit was from Roche Molecular Biochemicals.
Antibodies-Rabbit polyclonal antibodies against a synthetic peptide corresponding to the human VLDL receptor carboxyl terminus (R2623) has been described (15), and the antibody was affinity-purified on peptide-Sepharose as described. Rabbit polyclonal antibodies prepared against the purified VLDL receptor (R4522) have been described (16) and were purified by chromatography on protein G-Sepharose as described previously (16). Polyclonal antibodies to human LRP (R777) were developed as described (17).
Proliferation Assay of Endothelial Cells-HUVECs were routinely cultured to confluence in EGM. The cells were trypsinized and plated in a 96-well plate at 5,000 cells per well per 100 l of EBM supplemented with 2% FCS and antibiotics. The cells were allowed to adhere to the plate for at least 2 h. Then, bFGF at 10 ng/ml and various concentrations of TFPI or 2ME 2 were added to the wells. The cells were cultured for 48 h at 37°C in a 5% CO 2 atmosphere. Cell proliferation was determined using a uridine incorporation method as described by the manufacturer and confirmed by cell counting (Cell Counter model Z1, Coulter Incorporation, Miami, FL). Inhibition of cell proliferation, as assessed by uridine incorporation, was calculated using Equation 1. Additional experiments were performed using VLDL receptor-specific antibodies to block the antiproliferative activity of TFPI. In these experiments, either fixed or increasing concentrations of the rabbit polyclonal anti-VLDL receptor IgG (R4522) or the control anti-VLDL receptor IgG (R2623) were preincubated with HUVECs for 1 h at 37°C. In other experiments, the VLDL receptor antagonist, RAP, was preincubated with endothelial cells for 24 -36 h. After addition of VLDL receptor antagonists, TFPI and bFGF were added, and HUVEC proliferation was determined as above.
Proliferation of Tumor Cells-Rat C6 glioma and B16BL6 melanoma tumor cells were routinely cultured to confluence in DMEM supplemented with 10% FCS. B16BL6 melanoma cells were used between 6 and 18 passages. For the proliferation assay, tumor cells were plated in 96-well plate at 2,500 cells per well per 100 l of media and were allowed to adhere to the plate for at least 2 h. Various concentrations of TFPI or 2ME 2 were then added to the wells. The cells were cultured for 48 h at 37°C in a 5% CO 2 atmosphere. Cell proliferation was determined using a uridine incorporation method. Inhibition of cell proliferation was defined as described above.
Generation of a Soluble VLDL Receptor Domain-A fragment of the human VLDL receptor was generated as a glutathione S-transferase fusion protein. The polymerase chain reaction primers used are sense, AGTTGGATCCGGGAGAAAAGCCAAATGTGAA, and antisense, AGT-TGAATTCTTAGTAGTAACACTCTTTCAGGGGCTCATCACT. These primers generated a 989-bp fragment corresponding to the 8 ligand binding repeats of the VLDL receptor. Two tyrosine residues were added after the final cysteine of repeat 8 so the fragment could be radioiodinated. The polymerase chain reaction fragment was digested with BamHI and EcoRI, and the resulting fragment was ligated into the pGEX-2T vector. The entire insert of this final pGEX-VLDLR1-8 construct was sequenced.
Escherichia coli BL21 cells were transformed with pGEX-2T-VLDLr1-8 and grown in a 10-liter fermenter. When cells were in mid-log phase isopropyl-1-thio-␤-D-galactopyranoside was added to induce fusion protein production. After 3 more hours, cells were centrifuged, and the pellets were resuspended in 10 mM Tris, 1 mM EDTA, pH 7.4, and lysed by treatment with 1 mg/ml lysozyme and sonication. After centrifugation, the supernatant was incubated with 50 M MgCl 2 , 25 g/ml DNase I for 1 h at 4°C. The digest was diluted 2-fold into GSH wash buffer (20 mM HEPES, 0.1 M KCl, 1 mM dithiothreitol, 0.5 mM EDTA, pH 7.5) and applied to the GSH-Sepharose column. Following washing, the column was eluted with wash buffer containing 25 mM GSH.
The GST-VLDLr1-8 fusion protein was digested with 20 nM thrombin for 1.5 h at 25°C, and the reaction was stopped with 200 nM D-Phe-Pro-Arg-CH 2 Cl. The digestion reaction was re-applied to the GSH-Sepharose, and the flow-through containing free VLDLr1-8 was recovered. This was dialyzed against 5% acetic acid overnight and lyophilized. The fragment was then resuspended in nitrogenated 6 M guanidinium HCl, 50 mM Tris, 1 mM dithiothreitol, pH 8.5, and refolded according to the method of Dolmer et al. (18) by diluting 10-fold into nitrogenated 50 mM Tris, 10 mM CaCl 2 , 1 mM GSH, 0.5 mM GSSG, pH 8.5, and then dialyzing into 200ϫ volume of this same buffer overnight at 25°C. The refolded VLDLr1-8 was subjected to RAP affinity chromatography, and bound VLDLr1-8 was eluted from RAP-Sepharose with 20 mM EDTA.
For the solid phase binding assays, purified VLDLr1-8 was immobilized on microtiter wells at a concentration of 4 g/ml. The microtiter wells were then blocked with 3% bovine serum albumin. RAP, uPA, TFPI, and truncated TFPI were radioiodinated using IODO-GEN-coated tubes, whereas VLDL was iodinated by the iodine monochloride method as described (16). To form the uPA⅐PAI-1 complex, a 3-fold molar excess of PAI-1 was incubated with 125 I-labeled uPA. To determine binding affinities, radiolabeled ligand at the indicated concentration was added to the initial well, and a 3-fold serial dilution was performed, and ligand binding was allowed to continue 1 h at 37°C. Specific binding was determined using 100-fold molar excess of RAP. After binding, wells were washed 3 times and counted in a ␥ counter.
VLDL Receptor-mediated Degradation of TFPI-Mouse PEA13 cells, an LRP-negative cell line, were infected with one of two adenovirus constructs essentially as described (16): Ad-VLDLr, which encodes for the VLDL receptor, and Ad-LacZ, a control virus which encodes for the ␤-galactosidase gene. Briefly, either Ad-LacZ or Ad-VLDLr was added to PEA-13 cells at a concentration of 50,000 particles/cell and incubated overnight. After incubation, cells were rinsed and split into 12-well dishes. TFPI binding, internalization, and degradation studies were performed as described (16). 25 nM 125 I-TFPI in binding buffer (DMEM, 2% bovine serum albumin, 10 mM HEPES, Neutridoma®-NS media supplement (Roche Molecular Biochemicals)) was incubated with infected PEA13 cells for various lengths of time. Control wells included excess cold RAP, which inhibits ligand binding to the VLDL receptor, or 1 g/ml chloroquine, which inhibits lysosomal degradation of internalized ligands. At various time points the media were removed and mixed with 50% trichloroacetic (1:5 v/v). The pellet was collected by centrifugation and contained intact ligand, whereas the supernatant contained degraded ligand. Adherent cells were then treated with trypsin, EDTA, 0.5 mg/ml proteinase K (Sigma) for 10 min at 4°C. Cells are gently rinsed off the dish and centrifuged. After centrifugation, surface-bound TFPI was localized in the supernatant, whereas the cell pellet contained internalized TFPI. The radioactivity associated with each fraction was determined in a ␥ counter.

TFPI Inhibits Endothelial Cell
Proliferation-To determine the effect that TFPI has on endothelial cell growth, bFGF-stimulated endothelial cells were grown in the absence or presence of increasing concentrations of TFPI. As shown in Fig. 1A, TFPI had a dramatic effect on bFGF-stimulated endothelial cell growth, inhibiting proliferation of endothelial cells in a dose-dependent fashion with an IC 50 value of 350 nM. In several additional experiments, the IC 50 value for TFPI ranged from 150 to 400 nM. In contrast to its effect on endothelial cells, TFPI had no effect on the growth of B16BL6 melanoma cells (Fig. 1B), Lewis lung carcinoma (LLC), or rat C6 glioma cells (data not shown). As a control in each experiment, 2ME 2 , a naturally occurring antiproliferative agent (19), was used and inhibited the proliferation of both endothelial cells (Fig. 1A) and B16BL6 cells (Fig. 1B). Inhibition of cell proliferation by TFPI was reversible and was restored by removal of TFPI and addition of fresh media containing bFGF. TFPI derived from nonrecombinant sources exhibited a similar antiproliferative activity. Full-length protein purified from human plasma or HepG2 cells inhibited bFGFpromoted proliferation of HUVECs in a dose-dependent manner with a similar IC 50 value (data not shown).
The Antiproliferative Activity of TFPI Is Lost When the Kunitz 3 Domain of TFPI Is Deleted-TFPI is composed of three Kunitz-type proteinase inhibitor domains. Domains 1 and 2 are required for factor VIIa and Xa binding (4), whereas domain 3, along with the carboxyl terminus of TFPI, can bind to heparin (20). To identify the region of TFPI responsible for this antiproliferative activity, proliferation studies were performed with both full-length TFPI and a truncated TFPI containing only domains 1 and 2 (1-160 a.a.). The results of this experiment (Fig. 2) demonstrated that truncated TFPI exerts a diminished inhibitory action, suggesting that Kunitz-3 and the carboxyl terminus of TFPI play an important role in the antiproliferative action of this protein.
Receptor-associated Protein (RAP) Inhibits the Antiproliferative Activity of TFPI-We next sought to determine the mechanism of the antiproliferative activity of TFPI. TFPI is known to bind and inhibit the activity of fVIIa⅐TF complexes. However, we were unable to detect tissue factor antigen on proliferating endothelial cells using immunoblotting experiments, suggesting that another receptor is likely responsible for the antiproliferative activity of TFPI. Previous studies have shown that TFPI binds to the LRP (6), and thus we employed RAP to determine whether TFPI is mediating its effect via an LDL receptor family member. RAP was initially identified when it was noted to copurify with LRP during ligand affinity chromatography (21). This molecule binds with high affinity to LRP, VLDL receptor, and gp330/megalin and antagonizes the binding of all ligands to these receptors (22). In the current study, a fixed amount of RAP was first preincubated with cells, and 24 -36 h later increasing amounts of TFPI were added, such that the molar excess of RAP over TFPI varied from 2-to 8-fold. The results (Fig. 3) demonstrate that RAP inhibits the TFPI antiproliferative effect by shifting the IC 50 value of TFPI from 300 to 700 nM. Webb et al. (23) demonstrated that incubation of cells with RAP for up to 5 days had no effect on the cell surface levels of the VLDL receptor nor on its biological activity. Thus, we conclude that RAP acts as a competitive inhibitor of TFPI, preventing it from binding to a member of the LDL receptor family. RAP alone had no effect on bFGF stimulation of endothelial cell growth and does not affect the binding of 125 I-  1-276 a.a.) and truncated (1-160 a.a.) TFPI molecules. Full-length TFPI (closed circles) and truncated TFPI (open circles) were added to subconfluent cultures of HUVECs. After incubation at 37°C for 48 h, cell proliferation was assessed by uridine incorporation and expressed as percent inhibition as described under "Experimental Procedures." Each point represents the average of three wells. For each point the error is less than 15%. labeled TFPI to heparin sulfate proteoglycans (HSPG) (data not shown), confirming the results of Warshawsky et al. (6).
VLDL Receptor-specific Antibodies Inhibit the Antiproliferative Activity of TFPI-Whereas LRP has been previously reported to bind TFPI, this receptor could not be detected in HUVEC cell extracts by immunoblotting or RAP ligand blotting (9). In contrast, the structurally related VLDL receptor is expressed in HUVECs in the endothelium of capillaries and small arterioles and on the endothelial surface of bovine arteries, suggesting that the VLDL receptor may be responsible for the effect of TFPI. To test this, we employed specific antibodies to LRP and the VLDL receptor. Although previous studies have determined that the anti-VLDL receptor IgG blocks ligand uptake by this receptor (16), the specificity of this antibody on endothelial cells has not been demonstrated. To investigate the specificity of the anti-VLDL receptor antibody, immunoblotting experiments on membrane preparations from endothelial cells were performed. The results of this experiment (Fig. 4) confirm that the anti-VLDL receptor antibody is highly specific and only reacts with two polypeptides with estimated molecular weights of 130,000 and 95,000. Multiple forms of the VLDL receptor exist (13) resulting from alternate splicing of the pre-mRNA; type I receptors contain the O-linked sugar domains and have an apparent molecular weight of ϳ130,000 under nonreducing conditions, whereas type II receptors lack the O-linked sugar domain and have an apparent molecular weight of 95,000 -100,000 under nonreducing conditions. VLDL receptor mRNA that lacks exon 4 (encoding for L3) has also been detected in the brain (24).
LRP antibodies had no effect on TFPI inhibition of endothelial cell proliferation (data not shown). In contrast, anti-VLDL receptor IgG did partially neutralize the effect of TFPI on HUVEC proliferation (Fig. 5A), whereas a control IgG prepared against the VLDL receptor cytoplasmic domain had no effect in this assay (Fig. 5A). In the absence of TFPI, anti-VLDL receptor IgG had no effect on bFGF-stimulated HUVEC proliferation. Fig. 5B demonstrates the effect of increasing antibody concentration on HUVEC proliferation while maintaining a constant concentration of TFPI. Anti-VLDL receptor IgG significantly reduced the effect of TFPI on endothelial cell growth, although it did not completely restore the growth to levels seen in the absence of TFPI. These experiments confirm that the VLDL receptor is involved in this effect. Fig. 5C shows an experiment in which fixed levels of anti-VLDL receptor IgG were incubated in the presence of increasing amounts of TFPI, demonstrating that high concentrations of TFPI can overcome the blocking effects of anti-VLDL receptor IgG. This apparent competition between TFPI and VLDL receptor antibody further supports our hypothesis that the VLDL receptor plays an important role in the TFPI-mediated inhibition of endothelial cell growth.
Identification of the VLDL Receptor as a Novel TFPI Receptor-The ability of RAP and anti-VLDL receptor IgG to inhibit the antiproliferative activity of TFPI on endothelial cells suggests that the VLDL receptor is involved in this response. We next sought to determine whether TFPI can directly bind to the VLDL receptor. To measure this, a soluble form of the VLDL receptor (sVLDLr1-8) containing the ligand binding portion of the VLDL receptor (12,25) was expressed in E. coli. Following refolding and affinity purification, the sVLDLr1-8 was analyzed by SDS-PAGE which revealed that the purified receptor migrated as a single band under reducing (Fig. 6A, lane 1) and nonreducing conditions (Fig. 6A, lane 2). The faster electrophoretic mobility for sVLDLr1-8 under nonreducing conditions is consistent with the notion of a compactly folded structure arising from the formation of intra-chain disulfide bonds. sVLDLr1-8 was recognized by the anti-VLDLR IgG (Fig. 6B,  lanes 3 and 4) confirming that this antibody is capable of binding to the VLDL receptor ligand binding region.
The functional properties of the recombinant sVLDLr1-8 were examined by solid phase binding assays. In these experiments, sVLDLr1-8 was coated on microtiter wells, and the ability of 125 I-labeled RAP, VLDL, and uPA⅐PAI-1 complexes to bind to this molecule was measured. The results (Fig. 7) indicate high affinity binding of RAP (K D ϭ 1.4 nM) and uPA⅐PAI-1 complexes (K D ϭ 7.3 nM) to the immobilized receptor. The K D values measured in these experiments are virtually identical to the K D of 0.7 and 15 nM measured for the binding of RAP and uPA⅐PAI-1 complexes to the intact VLDLr, respectively (9,15). Immobilized sVLDLr1-8 also recognized 125 I-labeled VLDL particles, and its binding to immobilized sVLDLr1-8 was inhibited by excess RAP (Fig. 7C). These data confirm that the sVLDLr1-8 is correctly folded and capable of binding known VLDLr ligands with appropriate affinity.
We next employed the solid phase assay to measure the ability of 125 I-TFPI and 125 I-TFPI-(1-160) to bind to immobilized sVLDLr1-8. Specific binding was determined by subtracting background binding observed in the presence of a large molar excess of RAP. Full-length TFPI bound in a specific and saturable manner to the sVLDLr1-8 with an apparent K D of 80 nM determined by nonlinear regression analysis (Fig. 8). A 100-fold molar excess of RAP blocked more than 95% of TFPI binding to VLDL receptor. In contrast to the ability of fulllength TFPI to bind sVLDLr1-8, no specific binding of the truncated TFPI to the sVLDLr1-8 was detected under these conditions. Taken together, these data indicate that TFPI binds to the VLDL receptor and that this binding requires the third Kunitz-type domain and/or the flanking sequences. It is interesting to note that the truncated form of TFPI, which does not bind to the VLDL receptor, is also much less effective in endothelial proliferation studies.
LDL receptor family members were initially characterized as endocytic receptors that target ligands to lysosomes for proteolytic degradation. By having demonstrated specific binding of TFPI to the VLDL receptor in vitro, we next sought to determine whether the VLDL receptor was also able to mediate the cellular internalization and degradation of TFPI. These experiments were performed in mouse embryonic fibroblasts (PEA-13) known to be genetically deficient in LRP, since LRP is known to bind, internalize, and degrade TFPI. These fibroblasts also express very little VLDL receptor, thus the VLDL receptor gene was introduced into these cells using an adenovirus-VLDL receptor construct. This experimental system has been previously used to identify and characterize VLDL receptor ligands (16). In this way, the contribution of the VLDL receptor to TFPI internalization and degradation can be readily measured. Cells expressing the VLDL receptor are able to internalize and degrade TFPI (Fig. 9) in a RAP-sensitive fashion, confirming the role of the VLDL receptor in this process. In contrast, parental cells, or cells infected with an adenovirus construct expressing the lacZ gene, are unable to internalize 125 I-labeled TFPI. DISCUSSION An association between components of the coagulation pathway and tumor growth has been recognized for more than a century (26). Multiple studies have implicated fVIIa⅐TF complexes in the progression of tumor growth and metastases by promoting fibrin deposition and angiogenesis. Additional coagulation proteins, including thrombin (27) and protein C (28,29), are also able to regulate endothelial cell growth. In light of these observations, we sought to investigate whether TFPI could regulate the growth of endothelial cells. Our studies focused on TFPI for two reasons. First, the ability of TFPI to block fVIIa/TF activity suggested that it might regulate the signaling activity of this complex, and second, previous reports suggested that high concentrations of TPFI might have antiproliferative or apoptotic effects on smooth muscle cells and endothelial cells (30,31).
The antiproliferative activity of TFPI appears specific for endothelial cells as no inhibition was observed in any tumor cell line tested. Our initial hypothesis was that the antiproliferative activity of TFPI was mediated through fVIIa⅐TF complexes. However, we were unable to detect any expression of tissue factor on HUVECs. Thus, we speculated that a different receptor was responsible for the observed activity. It is unlikely that TFPI binding to HSPGs was responsible for the antiproliferative activity of TFPI, since RAP was able to inhibit the TFPI antiproliferative effect. As RAP does not affect the binding of TFPI to HSPGs, we concluded the mechanism involved association of TFPI with a member of the LDL receptor family.  2 and 4) conditions. Following electrophoresis, the bands were visualized with Coomassie Blue (A), or samples were transferred to nitrocellulose and immunoblotted with 1 g/ml anti-VLDL receptor IgG (R4522) (B).
Since the VLDL receptor is expressed on endothelial cells, we examined the potential involvement of the VLDL receptor in this activity. Several lines of evidence substantiate that TFPI binding to the VLDL receptor is a necessary component of the antiproliferative activity of TFPI. First, TFPI binds to the VLDL receptor with an apparent K D of 80 nM, which is close to the IC 50 value for the inhibition of endothelial cell proliferation (150 -400 nM). Second, RAP inhibits both the binding of TFPI to the VLDL receptor as well as its antiproliferative activity. Third, a specific polyclonal antibody to the VLDL receptor, which is known to block ligand recognition by this receptor (16), also blocked the antiproliferative effect of TFPI on endothelial cells. Finally, a deletion mutant of TFPI that failed to bind to the VLDL receptor in vitro was also much less active in inhibiting endothelial cell proliferation. All of these data corroborate the role of the VLDL receptor in this process and indicate that TFPI and the VLDL receptor are capable of modulating endothelial cell growth.
It should be noted that a severalfold molar excess of VLDL receptor antagonist was unable to block completely the TFPI antiproliferative activity, suggesting that the VLDL receptor may not be completely responsible for this effect. There are several possible explanations for this result. First, the concentrations of inhibitor used may not have been high enough to block completely the VLDL receptor-mediated TFPI effect. Second, the binding of TFPI to HSPGs may enhance the ability of TFPI to associate with the VLDL receptor and inhibit proliferation. A number of other VLDL receptor ligands have high affinity for HSPGs, including lipoprotein lipase and thrombospondin. The binding of these ligands to LRP is promoted when prior HSPG binding occurs (32,33). Finally, we cannot exclude the possibility that another mechanism also contributes to the TFPI activity that is insensitive to VLDL receptorbinding antagonists.  1-160 a.a.) to immobilized sVLDLr1-8. Soluble VLDLr1-8 (4 g/ml) was immobilized on microtiter plates (100 l/well). Indicated concentrations of 125 I-labeled TFPI (filled circles) or truncated TFPI (filled squares) was added to the wells. Following incubation, the wells were washed, and bound counts/min were determined. Specific binding was determined by 100-fold molar excess of RAP.
FIG. 9. VLDL receptor-dependent degradation of TFPI. PEA13 cells infected with either adenovirus-VLDL receptor (circles) or adenovirus-LacZ (inverted triangles) were used to measure degradation of TFPI. 35 nM 125 I-labeled TFPI was incubated with the cells for the indicated times, and the extent of cellular mediated degradation was measured as described under "Experimental Procedures." TFPI degradation is only seen with Ad-VLDL receptor-infected cells (filled circles). No degradation is seen in LacZ-infected cells (filled triangles), whereas excess RAP (open circles) dramatically reduces VLDL receptor-mediated TFPI degradation.
The mechanism by which TFPI binding to the VLDL receptor blocks bFGF-stimulated endothelial cell proliferation remains to be determined at this time and may involve signaling events. The VLDL receptor can participate in signal transduction pathways as shown in studies demonstrating that knockout mice lacking both the VLDL receptor and the closely related apoE receptor 2 have a defect in neuronal migration (34). Mice lacking the neuronal ECM protein Reelin, the disabled gene product (Dab1), or VLDL receptor/ApoER2 have identical phenotypes. Dab1 is an adaptor protein that interacts with the cytoplasmic domains of LDL receptor family members including the VLDL receptor (35,36) and appears to function in a tyrosine kinase signaling pathway. These findings suggest that the VLDL receptor and apoER2 participate in transmitting the extracellular Reelin signal through intracellular signaling processes mediated by mDab1. The hypothesis that the VLDL receptor signals in response to Reelin binding was recently supported by studies (37) showing direct binding of Reelin to the VLDL receptor and changes in tyrosine phosphorylation in response to Reelin-VLDL receptor association.
We propose that similar mechanisms occur in endothelial cells in response to TFPI binding to the VLDL receptor. TFPI binding could initiate signal transduction by cytoplasmic adaptor proteins (possibly a member of the disabled family of genes) in association with the VLDL receptor cytoplasmic domain, which would ultimately result in the inhibition of HUVEC proliferation. Alternatively, it is possible that the VLDL receptor does not signal directly but that it is involved in the formation of a multiprotein signaling complex induced by TFPI binding. Currently, the function of the VLDL receptor is not known. This molecule is most abundant in skeletal muscle, heart, adipose tissue, and brain (13,38). Subsequent studies confirmed that this receptor is expressed in the endothelium (10,11) and in macrophages present in human atherosclerotic lesions (11,16). The role of the VLDL receptor in neuronal development (34) and recent findings by Webb et al. (23) indicating a major role for the VLDL receptor in breast cancer cell migration suggest an important role for this receptor in cellular signaling events.
Several ligands for the VLDL receptor have been identified in addition to TFPI. These include apolipoprotein E (apoE), thrombospondin-1 (TSP-1), urokinase complexed to its inhibitor, plasminogen inhibitor type I (PAI-1), and Lp(a) (9,16). It is interesting to note that apoE (39), TSP-1 (40), and PAI-1 (41) have been reported to inhibit endothelial cell proliferation and angiogenesis, whereas urokinase (42) and Lp(a) (43) have been reported to enhance angiogenesis. It is unknown whether these molecules require binding to the VLDL receptor to exert effects on endothelial cells. However, it is compelling to speculate that the endocytic activity of the VLDL receptor could regulate the local concentrations of these various molecules, thus controlling the local balance of pro-and anti-angiogenic molecules. Therefore, the VLDL receptor may play a pivotal role in regulating endothelial cell growth and angiogenesis, directly through TFPI binding and indirectly through its effect on the turnover of pro-and anti-angiogenic molecules.
Based upon these in vitro results, it will be important to determine whether TFPI has anti-angiogenic and antitumor activity in vivo, and at present it is unknown whether TFPI/ VLDL receptor interaction has an antiproliferative effect on endothelial cells in vivo. Studies are currently underway with VLDL receptor knockout mice to determine if the VLDL receptor and its interaction with TFPI plays a role in pathological angiogenesis.
In conclusion, our data identify TFPI as an inhibitor of endothelial cell proliferation. We show that the VLDL receptor is a novel TFPI receptor on endothelial cells and, remarkably, that the VLDL receptor mediates the antiproliferative effect of TPFI. Together these data demonstrate the ability of the VLDL receptor to regulate the normal and pathological growth of endothelial cells and to position these families of molecules as potential targets for the development of new therapies of angiogenesis-based diseases.