Reversible Regulation of Tissue Factor–Induced Coagulation by Glycosyl Phosphatidylinositol–Anchored Tissue Factor Pathway Inhibitor
Abstract—Endothelial and tumor cells synthesize tissue factor pathway inhibitor (TFPI-1), which regulates tissue factor (TF) function by TF · VIIa · Xa · TFPI-1 quaternary complex formation (where VIIa and Xa are coagulation factors) and by translocation of these complexes into glycosphingolipid-rich microdomains of the cell membrane. Recombinant TFPI-1 added exogenously to cells is targeted to a degradation pathway. This study analyzes whether quaternary complex formation with endogenous TFPI-1 results also in internalization and degradation. We demonstrate that endogenous TFPI-1 and recombinant TFPI-1 differ in their distribution on the cell surface. Recombinant TFPI-1 is found in phospholipid- and glycosphingolipid-rich membrane domains, whereas endogenous TFPI-1 preferentially localizes to glycosphingolipid-rich microdomains. On quaternary complex formation, endogenous TFPI-1 remains protease sensitive and accessible for antibodies on intact cells, demonstrating that it is not appreciably internalized. Rather, regulation of TF by TFPI-1 is restored within 12 hours, consistent with dissociation of quaternary complexes on the cell surface. Endogenous TFPI-1 can be released from the cell surface by phospholipase treatment, indicating that TFPI-1 either is a glycosyl phosphatidylinositol (GPI)-anchored protein or binds to a GPI-linked receptor. We demonstrate that expression of a recombinant GPI-anchored form of TFPI-1 targets TF · VIIa complexes to glycosphingolipid-rich membrane fractions. Thus, GPI anchoring of TFPI-1 is sufficient for regulation of TF · VIIa complex function by a pathway of reversible inhibition rather than internalization and degradation.
- coagulation cascade
- Kunitz-type inhibitors
- cell surface proteoglycans
- glycosphingolipid-rich microdomains
- Received July 28, 1999.
- Accepted October 7, 1999.
The coagulation pathways are triggered by the cell surface receptor tissue factor (TF) that binds and allosterically activates the serine protease coagulation factor VIIa (VIIa). The cellular initiation of coagulation by the TF · VIIa complex is tightly regulated by the plasma-derived serine protease inhibitor (serpin) antithrombin III (ATIII)1 2 and the cell-associated Kunitz-type inhibitor TF pathway inhibitor (TFPI-1).3 4 Whereas inhibition of TF · VIIa by ATIII results in dissociation of the complex,5 6 binding of Kunitz-type inhibitors to TF · VIIa increases the stability of the complex,5 7 indicating that the latter mechanism of inhibition may provide a more sustained downregulation of the TF pathway. An inhibitor homologous to TFPI-1, TFPI-2, has been identified,8 9 but the role of TFPI-2 in the physiological regulation of TF · VIIa function is unclear. The TFPIs are multidomain inhibitors consisting of 3 Kunitz-type domains and a carboxy terminal highly enriched in basic residues. The carboxy terminal and the preceding third Kunitz-type domain10 are critical for the interaction with heparan-sulfated proteoglycans.
TFPI-1 binds to the surfaces of vascular cells11 12 and can be released into the blood by in vivo injection of heparin,13 suggesting heparan-sulfated proteoglycan cell surface binding sites. Proteoglycan receptors that are known to bind TFPI-1 are the transmembrane-anchored ryudocan/syndecan 414 and the glycosyl phosphatidylinositol (GPI)-anchored glypican 3.15 In a previous study, we found evidence that the primary anchoring of endogenous TFPI-1 on the surface of ECV304 cells is through a GPI linkage,16 and similar conclusions were drawn from studies using primary human umbilical vein endothelial cells in culture.17 The type of membrane anchoring influences the localization of cell surface receptors to specific areas of the cell membrane. For example, GPI-anchored proteins favor glycosphingolipid-rich membrane compartments that can be isolated by density centrifugation as low-density fractions insoluble in cold detergent, whereas transmembrane receptors typically localize to phospholipid-rich membrane areas that are fully detergent soluble and thus nonbuoyant in density gradients.18
In addition to proteoglycan receptors, the transmembrane-anchored LDL receptor–related protein (LRP), which is involved in endocytosis, has also been shown to provide a limited number of cell surface binding sites for TFPI-1. LRP is required for degradation of exogenously added TFPI-1 by cultured cells12 19 and for the accelerated cell surface clearance of TF in the presence of VIIa and TFPI-1 on monocytes.20 In the latter case, TF was shown to remain associated with phospholipid-rich membrane domains. In contrast, we have found that endogenously synthesized GPI-anchored TFPI can direct the TF · VIIa complex to glycosphingolipid-rich microdomains and plasmalemma vesicles/caveolae.16 Glycosphingolipids provide an unfavorable environment for proteolytic activity of the TF · VIIa complex; thus, targeting TF to these membrane structures serves as a mechanism to downregulate TF · VIIa complex function in addition to inhibition by cell-associated TFPI-1.16
In the present study, we determine whether GPI-anchored endogenous TFPI-1 supports the internalization and degradation of the quaternary TF · VIIa · Xa · TFPI-1 complex (where Xa is coagulation factor Xa). Because the majority of studies used recombinant TFPI-1 that was expressed in Escherichia coli, we wished to compare the cellular distribution of this material with the localization of TFPI-1 that is synthesized endogenously by cells. Recombinant TFPI-1 is shown to partition into detergent-insoluble as well as detergent-soluble fractions, a pattern that is distinct from the predominant localization of endogenous TFPI-1 in detergent-insoluble microdomains. Formation of a quaternary complex with endogenous GPI-anchored TFPI-1 does not result in internalization but in slow dissociation of the complex. Thus, the main function of GPI-anchored TFPI-1 is in the transient downregulation of TF · VIIa-dependent initiation of the coagulation pathways on cell surfaces.
Rabbit polyclonal antibodies to TFPI and TF and monoclonal antibody TF9–10H10 to TF have previously been described.16 A polyclonal antibody to TFPI-2 was raised in the rabbit by injections of the TFPI-2 cDNA in a mammalian expression plasmid. Purified plasmid DNA (50 to 200 μg) in 10 mmol/L citrate (pH 6.5) and 0.25% bupivacaine was injected intramuscularly monthly until 1:1000 diluted serum showed reactivity with an extracellular matrix preparation from the TFPI-2–expressing T24 bladder carcinoma cell line. After SDS-PAGE on 8% to 16% gradient gels, the antiserum reacted with 3 bands with apparent molecular masses of 30 to 40 kDa. A control antibody raised against an E coli–produced fusion protein of TFPI-2 showed identical reactivity. Additional antibodies were kindly provided by the following investigators: chicken antibody to the TFPI · Xa complex (Dr John Bognacki, American Diagnostica, Greenwich, Conn), monoclonal antibody Ryu-2–30-27 to ryudocan/syndecan 4 (Dr Tetsuhito Kojima, Nagoya University, Nagoya, Japan),14 monoclonal antibody S1 against glypican 1 (Dr Guido David, University of Leuven, Leuven, Belgium),21 antibody to protein C inhibitor (PCI; Dr Koji Suzuki, Mie University, Japan),22 and antibody R777 to LRP (Drs Maria Kounnas and Dudley Strickland, American Red Cross, Rockville, Md).23 Antibodies were purchased from the following sources: polyclonal antibody to ATIII (Dako), monoclonal anti-LRP (American Diagnostica), monoclonal antibody to caveolin (anti—22-kDa Rous sarcoma virus Src substrate, ICN Biomedical Inc), TRITC-labeled rabbit anti-chicken IgG (Sigma Chemical Co), and horseradish peroxidase–conjugated secondary antibodies and reagents for chemiluminescence detection (Amersham Life Sciences). Recombinant VIIa and plasma-derived factors X and Xa were prepared as described16 or purchased from Hematologic Technologies Inc. PCI was purified as described previously.24 Recombinant human TFPI-1 produced in E coli25 was generously provided by Dr Abla Creasey (Chiron Corp, Emeryville, Calif). ATIII was purchased from Calbiochem. Phosphatidylinositol-specific phospholipase C from Bacillus cereus was obtained from Boehringer-Mannheim Biochemica. Mounting solution (Slowfade Antifade) was from Molecular Probes.
The human cell line ECV304 (CR-1998) and Chinese hamster ovary (CHO-K1, CRL 9618) cells were obtained from American Tissue Culture Collection. Before the experiments, ECV304 monolayers were stimulated for 5 hours with 10 ng/mL tumor necrosis factor-α to induce maximal TF expression. All incubations and washes were carried out in cell buffer (mmol/L: HEPES 21, NaCl 137, KCl 5, Na2HPO4 0.75, glucose 5.5, and CaCl2 2, pH 7.4).
Triton X-114 Fractionation to Isolate Membrane-Associated TFPI-1
Confluent cells were detached with a cell dissociation buffer (Life Technologies), washed, and adjusted in cell buffer to 6×106 cells per milliliter. To release GPI-anchored proteins, the cell suspension was incubated with phospholipase C (1 U/mL) for 60 minutes at 37°C. Cells were washed once, and the cell pellet was lysed in 1% Triton X-114 in 0.1 mol/L Tris, 10 mmol/L EDTA, 2000 U/mL aprotinin, and 100 μmol/L phenylmethylsulfonyl fluoride (PMSF) with repeated mixing on ice for 15 minutes. Cell debris was removed by centrifugation at 14 000g at 4°C. Phase separation was induced by incubation of the cleared detergent lysate for 5 minutes at 37°C, followed by brief centrifugation at 14 000g to separate the detergent and aqueous phase. The detergent pellicle was extracted with acetone, and the membrane proteins were resuspended in nonreducing SDS sample buffer. An equivalent of 5×105 cells was loaded per lane for separation by SDS-PAGE and detection by Western blotting by use of polyclonal anti-TFPI antibody.16
Isolation of Detergent-Insoluble Membrane Fractions by Sucrose Gradient Centrifugation
Cell lysates were fractionated after cold Triton X-100 solubilization, as previously described.16 In brief, experimental treatment of cell monolayers typically involved TF · VIIa complex formation by incubating cells with 5 nmol/L VIIa in cell buffer for 1 hour at 4°C, followed by addition of 50 to 200 nmol/L factor X or 10 nmol/L factor Xa for an additional 30 minutes at 37°C. In experiments that analyzed the binding of exogenously added recombinant TFPI-1, the inhibitor was included at the indicated concentrations during an incubation at 4°C for 1 hour, followed by an incubation as described above for 30 minutes at 37°C. After the incubations, cell monolayers were washed twice with cell buffer with reduced (0.5 mmol/L) CaCl2, and cells were harvested by scraping in the same ice-cold buffer. The cell pellet was resuspended in 2 mL ice-cold 1% Triton X-100 containing MES-buffered saline (mmol/L: MES 25, NaCl 150, NaVO4 1, PMSF 1, and CaCl2 0.1, pH 6.5) and homogenized with a loose-fitting Dounce homogenizer, followed by the addition of sucrose to 40%. This lysate was overlaid with 10 mL of a 30% to 5% sucrose gradient in MES-buffered saline without Triton X-100 and centrifuged at 39 000 rpm in an SW40Ti rotor (Beckman Instruments) for 16 to 20 hours. Fractions of 1.5 mL were collected, precipitated with 10% trichloroacetic acid, and resuspended in nonreducing SDS-PAGE sample buffer at a protein concentration of 1 mg/mL, based on protein concentration in the fractions determined by bicinchoninic acid assay (Pierce Chemical Co). A constant amount of protein was separated on 8% to 16% gradient gels, followed by semidry transblot transfer onto Immobilon P (Millipore Corp) and Western blotting with polyclonal antibodies and chemiluminescence detection. To detect glypican 1 in the detergent-insoluble fractions, confluent ECV304 cells in ten 150-cm2 flasks were separated by sucrose gradient. The detergent-insoluble fractions were pooled and precipitated with acetone. After resuspension, the pool was treated with 25 mU heparinase (Calbiochem) in 50 mmol/L Tris, 100 mol/L NaCl, 1 mmol/L PMSF, and 20 μg/mL leupeptin, pH 7.0, for 2 hours at 37°C,26 followed by SDS-PAGE for Western blotting with monoclonal antibody S1.27
Expression of GPI-Anchored TFPI-1 in CHO-K1 Cells
The plasmid pRc/RSV-hCD14DAF encoding CD14 and the GPI attachment site from decay-accelerating factor (DAF) was kindly provided by Dr Vladimir Kravchenko (The Scripps Research Institute, La Jolla, Calif).28 The HindIII-ClaI fragment encoding CD14 was replaced by the coding sequence for TFPI obtained by polymerase chain reaction with a flanking 5′ HindIII site 5′ from the initiation of the translation site and a 3′ NarI site that created an in-frame fusion with the 37 carboxy-terminal residues of DAF. An HindIII-ApaI fragment of this construct was further subcloned into pcDNA3.1Zeo to provide a similar vector backbone as the TFPI-1 construct in pcDNA3 that was used as the control in the transfection experiments. CHO-K1 cells were cotransfected with the DAF-TFPI or the wild-type TFPI construct and a TF expression plasmid by use of lipofectamine (Life Technologies). Membrane association of TFPI-1 was analyzed by Western blotting after isolation of membrane-associated proteins by Triton X-114 fractionation, as described above. To analyze the translocation of TF, transfected cells were incubated with or without VIIa (5 nmol/L) and factor X (50 nmol/L) for 30 minutes, followed by separation of Triton X-100–insoluble complexes by sucrose gradient centrifugation, as described above.
ECV304 cells were grown on coverslips to confluence. After incubation for the indicated times in the presence of 5 nmol/L VIIa and 200 nmol/L factor X or 10 nmol/L factor Xa alone, the coverslips were rinsed in ice-cold cell buffer, incubated for 1 hour with chicken anti-TFPI · Xa complex (1:50) or rabbit anti–TFPI-1 antibodies in cell buffer with 5% BSA at 4°C, fixed for 2 minutes with acetone at 4°C, and incubated for 1 hour with TRITC-labeled secondary antibody. After they were washed in the cell buffer, the coverslips were mounted in Slowfade and analyzed by confocal laser scanning microscopy.
Assay for Cell Surface Proteolytic Activity of TF · VIIa and Protease Sensitivity of TFPI-1
In 24-well tissue culture plates, cells were grown to confluence. After washes, cells were preincubated with 50 μg/mL cycloheximide (Sigma) in cell buffer for 30 minutes, and cycloheximide was included at the same concentration in all subsequent incubation steps to prevent de novo synthesis of TF and TFPI-1. Cells were incubated with VIIa (5 nmol/L) and factor X (500 nmol/L) for 30 minutes at 37°C to form quaternary complexes. After washes, serum-free medium containing cycloheximide (50 μg/mL) was added. To determine TF activity, cells were assayed 0.5, 4.5, 12.5, and 24.5 hours after the initial treatment with factors VIIa and X. Cells were incubated with or without 100 μg/mL polyclonal anti-TFPI IgG at 37°C for 15 minutes, followed by addition of VIIa (5 nmol/L) and factor X (500 nmol/L). Aliquots of 50 μL were taken during a 30-minute incubation period and quenched in 150 μL of 0.1 mol/L EDTA for a chromogenic assay to measure Xa generation. The rate of Xa generation was constant in the presence of the anti-TFPI antibody, and this activity was equated to the total TF activity (TFtotal) on these cells. In the absence of anti-TFPI antibodies, the rate initially declined, reflecting inhibition by TFPI. The subsequent stable rate of factor X activation gave the residual TF activity (TFres). TFtotal−TFres equals the amount of TF that is regulated by TFPI and is thus a measure for the inhibitory pool of TFPI-1 on the cell surface. Data for regulated TF are given as a percentage value relative to the total cellular TF, defined by the following formula: 100×(TFtotal−TFres)/(TFtotal). Statistical analysis was performed by t test for paired samples (Statview).
To analyze protease sensitivity of surface-bound TFPI-1, ECV304 cells were plated in 6-well plates and subjected to the above-described protocol in the presence of cycloheximide. After 1 and 12 hours, cells were harvested with cell dissociation buffer and incubated with or without trypsin (0.5 mg/mL) for 60 minutes at 4°C. Soybean trypsin inhibitor (1 mg/mL) was added, and cells were extensively washed before Triton X-114 membrane fractionation for Western blotting. All experiments were repeated at least 3 times, and representative data are shown.
Recombinant TFPI-1 Has Detergent-Soluble and Low-Density Binding Sites on ECV304 Cells
Because the internalization and degradation of TFPI has been extensively studied with recombinant protein that was exogenously added to cells, we wished to determine whether the cellular distribution of the previously used E coli–derived TFPI was identical to TFPI-1 that was endogenously synthesized by ECV304 cells, which served as our model to study the regulation of TF function by cell-associated TFPI-1.16 E coli–derived recombinant full-length TFPI-1 was added to ECV304 monolayers. Cells were incubated with TFPI-1 for 30 minutes at 37°C, washed to remove nonspecifically bound material, and subsequently lysed with cold detergent for sucrose gradient fractionation and characterization of the distribution of cell-bound TFPI-1 by Western blotting. Addition of 100 to 300 nmol/L of TFPI-1 to ECV304 monolayers was required to reproducibly observe binding with the typical distribution shown in Figure 1A⇓. This concentration range is somewhat higher than previously reported affinities of recombinant TFPI-1 for cell surfaces. Because our experiments were performed in the absence of carrier proteins that had been routinely included in previous binding studies,12 20 we attribute the requirement for higher concentrations in our experiments to losses due to nonspecific protein adsorbance of TFPI-1 to the tissue culture plastic. This notion is supported by control experiments in which TFPI-1 was added to a concentrated cell suspension rather than to a cell monolayer, followed by cell fractionation on a sucrose gradient. Under these conditions, binding of recombinant TFPI-1 was observed at 10 nmol/L, and the distribution was comparable to the monolayer experiments.
The sucrose density fractionation separates the cell lysate in detergent-soluble material, which partitions in the high-density fractions 7 and 8, and detergent-insoluble material, which floats in the sucrose gradient because of the high content of glycosphingolipids. Under our conditions, the glycosphingolipid-rich material is typically visible as an opalescent band in fractions 3 to 5. These fractions also contain caveolin, the marker protein for caveolae, which represent a structurally defined subcompartment of the glycosphingolipid-rich microdomains of cell membranes.18 In Figure 1A⇑, the distribution of TFPI-1 in sucrose density separations of lysates from control cells (top) and cells incubated with 300 nmol/L recombinant TFPI-1 (second panel) or 300 nmol/L TFPI-1 in the presence of 5 U/mL heparin (third panel) are shown. The protein loading and exposure times of the Western blots for TFPI-1 in Figure 1A⇑ were identical, and the intensities of the bands thus reflected the relative amounts of the respective species bound to the cells. Endogenous TFPI-1 migrated at the 50-kDa marker, whereas the exogenously added TFPI-1, which lacked carbohydrate moieties resulting from recombinant expression in E coli, migrated close to the 36-kDa marker protein. Endogenous TFPI-1 was present in only the detergent-insoluble fractions, whereas exogenous TFPI-1 was found in detergent-insoluble and detergent-soluble fractions. Endogenous TFPI-1 was not significantly displaced by exogenous recombinant TFPI-1. In all experiments performed, there appeared to be less available binding sites for exogenously added TFPI-1 in fraction 3, whereas a much larger number of apparently unsaturated sites were present in fractions 4 to 8. Cell surface binding of recombinant TFPI-1 was drastically reduced in the presence of heparin (Figure 1A⇑), indicating that heparin either competes with heparan-sulfated proteoglycan binding sites or influences the TFPI-1 conformation to reduce the association with putative cellular receptors. Displacement of the endogenous TFPI-1 by 5 U/mL heparin was not apparent under the experimental conditions, suggesting that dissociation induced by heparin is slow, that higher affinity heparin-independent binding sites exist in these fractions, or that a portion of the endogenous TFPI-1 is directly GPI anchored, as previously suggested.17
Expression of Known TFPI-1 Receptors by ECV304 Cells
The binding of exogenous TFPI-1 throughout the membrane fractions suggested that multiple binding sites for recombinant TFPI-1 existed on ECV304 cells. The expression of known binding proteins for TFPI-1 was analyzed at the mRNA or protein level. By reverse-transcriptase PCR, ECV304 cells expressed mRNA (data not shown) for 2 of the known GPI-anchored heparan-sulfated proteoglycan receptors, glypican 1 and 3, the latter of which binds TFPI-1.15 In pooled low-density fractions from large quantities of cells (from ten 150-cm2 flasks) that were treated with heparinase, the glypican 1–reactive antibody S1 detected a 68-kDa band (data not shown) that had a molecular mass corresponding to the published size of the glypican 1 core protein.27 These data are consistent with localization of GPI-anchored proteoglycan receptors in the glycosphingolipid-rich microdomains. Attempts to study in detail the cellular partitioning of glypican 1 and 3 protein between detergent-soluble and -insoluble fractions failed because of low sensitivity and considerable background staining of available antibodies. In the detergent-soluble fractions, Western blotting detected the 2 previously identified TFPI-1–binding transmembrane receptors, syndecan 4 and LRP, which were absent from the low-density fractions (Figure 1B⇑). Syndecan 4 cell surface expression was also detected on intact cells by immunofluorescence, providing further support that this heparan-sulfated proteoglycan can serve as a cell surface binding site for TFPI-1. Thus, the partitioning of known binding receptors for TFPI-1 into specific membrane domains of ECV304 cells likely accounts for the distribution of recombinant exogenously added TFPI-1 to detergent-soluble and detergent-insoluble membrane domains.
Serpins Do Not Localize to Detergent-Insoluble Microdomains
To analyze whether binding to detergent-soluble and -insoluble binding sites is a common characteristic of heparin-binding protease inhibitors, we analyzed binding of 2 serpins, ATIII and PCI, to EVC304 cells (Figure 1C⇑). These plasma inhibitors were added at near physiological concentrations to ECV304 monolayers, and cells were treated as described above for TFPI-1. ATIII (500 nmol/L) and PCI (50 nmol/L) partitioned predominantly in the detergent-soluble fractions, demonstrating that the TFPI binding sites in the low-density fractions are not shared between exogenously added heparin-binding serine protease inhibitors and TFPI-1. TFPI-2, the highly homologous Kunitz-type inhibitor that is also synthesized by ECV304 cells, showed a distribution remarkably similar to the exogenously added TFPI-1 (Figure 1A⇑), indicating that at least some of the cellular binding sites are shared between these 2 inhibitors. TFPI-2 is known to be heterogeneous, with apparent molecular masses ranging from 27 to 33 kDa under nonreducing conditions that are due to differences in glycosylation.29 According to analysis by SDS-PAGE on gradient gels, the apparent molecular mass of the TFPI-2 isoforms in extracellular matrix preparations ranged from 30 to 40 kDa (not shown). In the cell lysate fractionation (Figure 1A⇑), no 30-kDa form of TFPI-2 was detectable, and the higher molecular weight glycosylation isoforms of TFPI-2 showed preferential localization to low-density fractions. Taken together, these data demonstrate a unique association of the homologous heparin-binding Kunitz-type inhibitors TFPI-1 and -2 with detergent-insoluble microdomains.
Downregulation of TF · VIIa Function by Endogenous TFPI-1 Is Reversible
Our previous study16 indicated that GPI-anchored TFPI-1 plays an important role in translocating TF to the nonprocoagulant environment of glycosphingolipid-rich microdomains. The fate of the translocated complex and the reversibility of the downregulation had not yet been investigated. ECV304 cells were stimulated with tumor necrosis factor-α for 5 hours, followed by addition of cycloheximide to block de novo protein synthesis during the subsequent experimental manipulation. The chosen dose of cycloheximide blocked protein synthesis by >95%, as judged from intrinsic labeling experiments, without affecting cell attachment and viability during the duration of the experiment. Performing the experiments in the presence of cycloheximide ensured that changes in cell surface protein expression did not result from de novo protein synthesis.
We had previously established that TF is in excess of TFPI-1 in tumor necrosis factor-α–stimulated ECV304 cells16 and found that TF activity is rapidly downregulated by complex formation with TFPI-1. On quaternary complex formation, TF activity and TFPI-1 inhibitory function on the cell surface are thus concordantly reduced. To form a quaternary complex with cell-associated TFPI-1, cells were exposed to VIIa and substrate X for 30 minutes, followed by removal of the unbound protein and continuation of cell culture in serum-free growth medium At various times after return to culture medium, cells were analyzed for TF activity on the cell surface. By comparing TF activity in the presence and absence of function-blocking anti–TFPI-1 antibodies, we determined the amount of TF that is regulated by TFPI-1, which is given in Figure 2⇓ as a percent value based on the total TF activity. On untreated cells, up to 50% of the TF activity was regulated by TFPI-1. Cells that were preincubated with VIIa and substrate X for 30 minutes to form the quaternary complex with TFPI-1 had reduced levels of total TF functional activity, which provided evidence that a fraction of TF is already in an inhibited complex with TFPI-1. Only a minor fraction of the remaining TF (<15%) was regulated by TFPI-1 immediately and 4 hours after the pretreatment with VIIa and substrate X (Figure 2⇓). There was no significant difference between the treated and nontreated cells in the absolute TF activity that was not regulated by TFPI-1, demonstrating that the manipulations did not influence the activity of TF activity per se. Prolonged incubation in the absence of VIIa and substrate X restored TFPI-1–dependent regulation of TF function in the treated cells, and the percentage of TF regulated by TFPI-1 on treated and control cells was indistinguishable after 12 and 24 hours. Thus, formation of the quaternary complex did not induce enhanced degradation of TF or TFPI-1. Notably, to detect TF activity after prolonged incubation, it was necessary to add VIIa, indicating that the TF · VIIa complex also dissociated after disruption of the quaternary complex.
TFPI-1 Is Not Internalized After Quaternary Complex Formation With TF
The suggested slow dissociation of the quaternary complex may involve internalization and a time-dependent reappearance of TF and TFPI-1 on the cell surface. TFPI-1 is highly protease sensitive, and treatment of intact cells at low temperatures to restrict internalization results in complete proteolytic degradation of the cell-associated inhibitor.16 Internalization of TFPI-1 should result in protection from extracellular proteases. Intact cells pretreated with VIIa and substrate X along with controls were exposed to trypsin at various times after formation of the quaternary complex. Figure 3⇓ demonstrates that at 1 and 12 hours after complex formation, the amount of cell-associated TFPI-1 was similar in treated and untreated cells. All cellular TFPI-1 was susceptible to extracellular protease treatment, demonstrating that no significant pool of the cell-associated inhibitor was in an inaccessible internal compartment of the cell. Identical results were obtained when the cells were immediately processed after quaternary complex formation for 30 minutes (data not shown), excluding the rapid reappearance of the complex after an initial internalization event or a closure of caveolae.
We further used antibody staining of intact cells to demonstrate that TFPI-1 remains accessible for a non–cell-permeable antibody. The intensity of staining for TFPI-1 did not change over time after the addition of VIIa and substrate X (Figure 4A⇓). We further used an antibody to the TFPI · Xa complex to detect complex formation of TFPI-1 on the cell surface. This antibody does not react with free TFPI-1 on unperturbed ECV304 cells (Figure 4B⇓). Addition of Xa or incubation of the cells with VIIa and the substrate X resulted in an indistinguishable staining intensity, consistent with a similar level of cell surface expression of the Xa · TFPI-1 and the quaternary complex (Figure 4B⇓). Staining intensities were similar 2 hours after complex formation, demonstrating that neither the GPI-anchored Xa · TFPI-1 complex nor the quaternary complex is substantially internalized under our experimental conditions. These results are different from previous studies with exogenously added recombinant TFPI-1, which mediates efficient internalization of Xa on hepatoma cells or embryonic fibroblasts,30 31 providing evidence for a distinct function of the GPI-anchored pools of TFPI-1.
GPI Anchoring of TFPI-1 Is Sufficient for Localization to Detergent-Insoluble Microdomains
Our previous study demonstrated that the majority of the endogenously synthesized TFPI-1 in ECV304 cells is released by phospholipase C treatment,16 indicating that GPI anchoring is responsible for the targeting of TFPI-1 to the low-density microdomains and the resulting regulation of TF function. To directly test this hypothesis, we attempted to reconstitute the TF translocation pathway by a TFPI-1 fusion protein that included at its carboxy terminal the GPI attachment sequence of DAF. Expression of wild-type soluble TFPI-1 in CHO cells did not result in cell surface association of TFPI-1 (Figure 5A⇓), suggesting that TFPI-1 binding sites with characteristics similar to those in ECV304 cells were not expressed by CHO cells. Transfection with the cDNA that encodes for the fusion protein of TFPI-1 and the GPI attachment sequence of DAF (DAF-TFPI) resulted in recovery of TFPI-1 in the membrane fraction of a Triton X-114 phase separation (Figure 5A⇓). DAF-TFPI was also detected in the supernatant, likely attributable to shedding or incomplete utilization of the GPI attachment sequence, as described for other overexpression experiments.28 Treatment of cells that were transfected with DAF-TFPI with phosphoinositol-specific phospholipase C (1 U/106 cells) resulted in a loss of recovery of TFPI-1 in the membrane fraction (Figure 5B⇓), providing evidence for GPI anchoring of the fusion protein.
CHO cells were transiently cotransfected with DAF-TFPI and TF. Cells transfected with the DAF-TFPI fusion protein showed enhanced downregulation of TF · VIIa proteolytic function compared with cells that were transfected with soluble TFPI-1, which showed no stable membrane anchoring. In transient transfection experiments, the percentage of TF activity regulated by TFPI-1 was 24±13% on the DAF-TFPI–transfected cells versus 9±16% on the TFPI-1–transfected cells (n=5). Cells cotransfected with DAF-TFPI and TF were further analyzed by Western blotting for the localization of the proteins after sucrose density gradient fractionation of cells that were lysed with cold detergent. In the absence of VIIa and substrate X, the DAF–TFPI-1 fusion protein localized selectively to the low-density fractions, whereas TF was exclusively found in the detergent-soluble fractions (Figure 5C⇑). When cells were exposed to VIIa and substrate X to promote quaternary complex formation of TF · VIIa · Xa with the fusion protein, TF translocated to the detergent-insoluble fractions. The stronger intensity of the signal for TFPI-1 in fraction 5 in the right panel of Figure 5C⇑ was not a consistent finding, and we conclude that the localization of the fusion protein did not change appreciably on complex formation. GPI anchoring of TFPI-1 is thus sufficient to regulate TF · VIIa function by translocation to glycosphingolipid-rich microdomains.
In our previous study,16 we demonstrated a pathway for regulation of the proteolytic function of TF · VIIa by cell-associated GPI-anchored TFPI-1. The TF · VIIa complex has high proteolytic activity in the detergent-soluble fractions, but TF · VIIa function is negligible in glycosphingolipid-rich detergent-insoluble microdomains of the cells. TF is translocated to these microdomains on association with GPI-anchored TFPI-1. This translocation requires both VIIa and the product of the TF · VIIa complex, Xa, consistent with a typical quaternary complex formation of TF · VIIa · Xa · TFPI that has been documented biochemically as the inhibitory mechanism of TFPI-1.3 The presented data show that direct fusion of a GPI anchor to TFPI-1 is sufficient for the specific localization of TFPI-1 to detergent-insoluble microdomains and for the reconstitution of the translocation pathway in transfected CHO cells. Direct GPI linkage of TFPI-1, previously suggested for endothelial cells,17 would thus provide a regulatory pathway for TF function in vivo.
However, exogenously added TFPI-1 also associates with detergent-insoluble microdomains through apparently unsaturated binding sites on ECV304 cells. Thus, a direct GPI anchoring of TFPI-1 is not the only mechanism for targeting of TFPI-1 to low-density glycosphingolipid-rich microdomains. ECV304 cells express members of the GPI-anchored glypican family of heparan-sulfated proteoglycans that likely associate with the detergent-insoluble membrane domains as a result of their specific membrane attachment. A previous study has demonstrated that the glypican 3 core protein binds specifically to immobilized TFPI-1,15 and we find no significant displacement of the endogenous inhibitor from ECV304 cells on incubation with heparin. Clusters of negatively charged residues in glypican 1 and 3 are highly conserved32 and may contribute to protein-protein interactions with basic regions of TFPI, resulting in high-affinity binding that is only inefficiently inhibited by heparin. Protein-protein interactions may be essential for cell surface binding of secreted TFPI in light of the low concentration of secretion by endothelial cells that accumulate <0.5 nmol/L TFPI-1 per 106 cells per 24 hours in the culture supernatant.33 34
Because heparin infusion releases TFPI-1 into the circulation,13 heparan-sulfated binding sites for TFPI-1 must also exist in vivo. We find that exogenously added TFPI-1 was bound to detergent-insoluble and -soluble binding sites on ECV304 cells and that this binding was drastically reduced in the presence of heparin. The transmembrane-anchored heparan-sulfated proteoglycan syndecan 4, which binds TFPI-1,14 was found to partition into the detergent-soluble fractions, indicating that this class of cell-associated proteoglycans constitutes part of the heparin-sensitive binding sites for exogenously added TFPI-1. Notably, other heparin-binding proteins, such as the serpins ATIII and PCI in the present study, also bound to the detergent-soluble fractions but failed to localize appreciably to the detergent-insoluble fractions. In contrast, TFPI-2 that is endogenously synthesized by ECV304 cells showed a membrane partitioning similar to exogenously added recombinant TFPI-1. This suggests that serpins and TFPIs share some proteoglycan binding sites in the detergent-soluble fractions but that the association with detergent-insoluble microdomains is a specific property of TFPIs. Because glypican 1 has been found to bind to another Kunitz-type inhibitor, the amyloid precursor protein inhibitor,35 one may speculate that association with glycosphingolipid-rich microdomains is a specific property of Kunitz-type protease inhibitors.
By analyzing protease sensitivity and immunoreactivity of TFPI-1 on intact cells, we demonstrate that quaternary complex formation with GPI-anchored TFPI-1 on ECV304 cells does not result in appreciable internalization of the inhibitor. This clearly distinguishes endogenous TFPI-1 from exogenously added recombinant TFPI-1 that undergoes internalization, on the basis of several studies.12 19 20 31 The different biological properties of the endogenous and recombinant inhibitor may result from lack of glycosylation of the recombinant material, which is an important topic for future studies. Because exogenously added TFPI-1 partitions throughout detergent-insoluble and -soluble fractions, we suggest that the internalization pathway of exogenously added inhibitor involves binding of TFPI-1 to the detergent-soluble proteoglycans. This would facilitate the cooperation20 36 with endocytotic receptors, such as LRP, which was detected in the same membrane environment. As an additional difference from the degradative pathway described for exogenous TFPI-1, we find that the inhibitory function of endogenous TFPI-1 is restored after prolonged incubation, which allowed for the dissociation of the quaternary complex. The high-affinity association of endogenously synthesized TFPI-1 with glycosphingolipid-rich microdomains thus constitutes a reversible pathway to regulate cell surface proteolysis.
TFPI-1 plays an essential role in the in vivo regulation of the cellular initiation of coagulation by TF.37 38 Because heparin infusion releases TFPI-1 presumably from the vascular endothelium into the circulation,13 one might be faced with decreased anticoagulant potential of the vessel wall because of depletion of cell-associated TFPI-1 during the course of anticoagulant therapy with heparin. Our data indicate that the endogenously bound TFPI-1 on cultured cells is not significantly displaced by heparin and that heparin is rather more effective in preventing the binding of exogenously added TFPI-1 to possibly lower affinity binding sites in the detergent-soluble fractions. Heparin may compete with binding of ATIII, the other potential physiological inhibitor of the TF pathway, to these proteoglycan binding sites, resulting in lack of inhibitors for the TF pathway in the phospholipid-rich membrane areas of the cell. As a consequence, inhibition of TF in the presence of heparin would be directed toward the translocation-dependent pathway that positions TF in an environment that does not support the proteolytic activity of TF · VIIa.16 This may provide the most efficient and prolonged control of TF-initiated coagulation and thus cooperate with the fluid-phase anticoagulant properties of heparin in plasma.
This study was supported by National Institutes of Health grants PO1 HL-16411 (W.R.) and RO1 HL-58869 (L.V.M.R.) and from the American Heart Association (I.O., Y.M.). W.R. is an Established Investigator of the American Heart Association. Antibodies were generously provided by Drs John Bognacki, Tetsuhito Kojima, Koji Suzuki, Dudley Strickland, Maria Kounnas, and Guido David. We thank Dr Vladimir Kravchenko for plasmids, Dr Abla Creasey for recombinant TFPI-1, Dr Francisco España for PCI purification, and Pablito Tejada, Richard Savary, Cindi Biazak, and Jennifer Royce for excellent technical assistance.
Rao LVM, Rapaport SI, Hoang AD. Binding of factor VIIa to tissue factor permits rapid antithrombin III/heparin inhibition of factor VIIa. Blood. 1993;81:2600–2607.
Lawson JH, Butenas S, Ribarik N, Mann KG. Complex-dependent inhibition of factor VIIa by antithrombin III and heparin. J Biol Chem. 1993;268:767–770.
Rao LV, Nordfang O, Hoang A, Pendurthi U. Mechanism of antithrombin III inhibition of factor VIIa/tissue factor activity on cell surfaces: comparison with tissue factor pathway inhibitor/factor Xa-induced inhibition of factor VIIa/tissue factor activity. Blood. 1995;85:121–129.
Dickinson CD, Ruf W. Active site modification of factor VIIa affects interactions of the protease domain with tissue factor. J Biol Chem. 1997;272:19875–19879.
Sprecher CA, Kisiel W, Mathewes S, Foster DC. Molecular cloning, expression, and partial characterization of a second human tissue-factor-pathway inhibitor. Proc Natl Acad Sci U S A. 1994;91:3353–3357.
Miyagi Y, Koshikawa N, Yasumitsu H, Miyagi E, Hirahara F, Aoki I, Misugi K, Umeda M, Miyazaki K. cDNA cloning and mRNA expression of a serine proteinase inhibitor secreted by cancer cells: identification as placental protein 5 and tissue factor pathway inhibitor-2. J Biochem (Tokyo). 1994;116:939–942.
Callander NS, Rao LVM, Nordfang O, Sandset PM, Warn-Cramer B, Rapaport SI. Mechanisms of binding of recombinant extrinsic pathway inhibitor (rEPI) to cultured cell surfaces: evidence that rEPI can bind to and inhibit factor VIIa-tissue factor complexes in the absence of factor Xa. J Biol Chem. 1992;267:876–882.
Warshawsky I, Herz J, Broze GJ Jr, Schwartz AL. The low density lipoprotein receptor-related protein can function independently from heparan sulfate proteoglycans in tissue factor pathway inhibitor endocytosis. J Biol Chem. 1996;271:25873–25879.
Kojima T, Katsumi A, Yamazaki T, Muramatsu T, Nagasaka T, Ohsumi K, Saito H. Human ryudocan from endothelium-like cells binds basic fibroblast growth factor, midkine, and tissue factor pathway inhibitor. J Biol Chem. 1996;271:5914–5920.
Mast AE, Higuchi DA, Huang ZF, Warshawsky I, Schwartz AL, Broze GJ Jr. Glypican-3 is a binding protein on the HepG2 cell surface for tissue factor pathway inhibitor. Biochem J. 1997;327:577–583.
Sevinsky JR, Rao LVM, Ruf W. Ligand-induced protease receptor translocation into caveolae: a mechanism for regulating cell surface proteolysis of the tissue factor-dependent coagulation pathway. J Cell Biol. 1996;133:293–304.
Lupu C, Goodwin CA, Westmuckett AD, Emeis JJ, Scully MF, Kakkar VV, Lupu F. Tissue factor pathway inhibitor in endothelial cells colocalizes with glycolipid microdomains/caveolae: regulatory mechanism(s) of the anticoagulant properties of the endothelium. Arterioscler Thromb Vasc Biol. 1997;17:2964–2974.
Warshawsky I, Broze GJ, Schwartz AL. The low density lipoprotein receptor-related protein mediates the cellular degradation of tissue factor pathway inhibitor. Proc Natl Acad Sci U S A. 1994;91:6664–6668.
Hamik A, Setiadi H, Bu G, McEver RP, Morrissey JH. Down-regulation of monocyte tissue factor mediated by tissue factor pathway inhibitor and the low density lipoprotein receptor-related protein. J Biol Chem. 1999;274:4962–4969.
David G, Lories V, Decock B, Marynen P, Cassiman J-J, van den Berghe H. Molecular cloning of a phosphatidylinositol-anchored membrane heparan sulfate proteoglycan from human lung fibroblasts. J Cell Biol. 1990;111:3165–3176.
Suzuki K, Nishioka J, Kusumoto H, Hashimoto S. Mechanism of inhibition of activated protein C by protein C inhibitor. J Biochem. 1984;95:187–195.
Mikhailenko I, Kounnas MZ, Strickland DK. Low density lipoprotein receptor-related protein/α2-macroglobulin receptor mediates the cellular internalization and degradation of thrombospondin: a process facilitated by cell-surface proteoglycans. J Biol Chem. 1995;16:9543–9549.
Diaz-Collier JA, Palmier MO, Kretzmer KK, Bishop BF, Combs RG, Obukowicz MG, Frazier RB, Bild GS, Joy WD, Hill SR, Duffin KL, Gustafson ME, Junger KD, Grabner RW, Galluppi GR, Wun T-C. Refold and characterization of recombinant tissue factor pathway inhibitor expressed in Escherichia coli. Thromb Haemost. 1994;71:339–346.
Lories V, De Boeck H, David G, Cassiman J-J, van den Berghe H. Heparan sulfate proteoglycans of human lung fibroblasts. J Biol Chem. 1998;262:854–859.
De Boeck H, Lories V, David G, Cassiman J-J, van den Berghe H. Identification of a 64 kDa heparan sulphate proteoglycan core protein from human lung fibroblast plasma membranes with a monoclonal antibody. Biochem J. 1987;247:765–771.
Lee J-D, Kravchenko V, Kirkland TN, Han J, Mackman N, Moriarty A, Leturcq D, Tobias PS, Ulevitch RJ. Glycosyl-phosphatidylinositol-anchored or integral membrane forms of CD14 mediate identical cellular responses to endotoxin. Proc Natl Acad Sci U S A. 1993;90:9930–9934.
Rao CN, Reddy P, Liu YY, O’Toole E, Reeder D, Foster DC, Kisiel W, Woodley DT. Extracellular matrix-associated serine protease inhibitors (Mr 33,000, 31,000, and 27,000) are single-gene products with differential glycosylation: cDNA cloning of the 33-kDa inhibitor reveals its identity to tissue factor pathway inhibitor-2. Arch Biochem Biophys. 1996;335:82–92.
Ho G, Toomey JR, Broze GJ, Schwartz AL. Receptor-mediated endocytosis of coagulation factor Xa requires cell surface-bound tissue factor pathway inhibitor. J Biol Chem. 1996;271:9497–9502.
Ho GY, Broze GJ Jr, Schwartz AL. Role of heparan sulfate proteoglycans in the uptake and degradation of tissue factor pathway inhibitor-coagulation factor Xa complexes. J Biol Chem. 1997;272:16838–16844.
Bajaj MS, Kuppuswamy MN, Saito H, Spitzer SG, Bajaj SP. Cultured normal human hepatocytes do not synthesize lipoprotein-associated coagulation inhibitor: evidence that endothelium is the principal site of its synthesis. Proc Natl Acad Sci U S A. 1990;87:8869–8873.
Ameri A, Kuppuswamy MN, Basu S, Bajaj SP. Expression of tissue factor pathway inhibitor by cultured endothelial cells in response to inflammatory mediators. Blood. 1992;79:3219–3226.
Williamson TG, Mok SS, Henry A, Cappai R, Lander AD, Nurcombe V, Beyreuther K, Masters CL, Small DH. Secreted glypican binds to the amyloid precursor protein of Alzheimer’s disease (APP) and inhibits APP-induced neurite outgrowth. J Biol Chem. 1996;271:31215–31221.
Narita M, Bu G, Olins GM, Higuchi DA, Herz J, Broze GJ Jr, Schwartz AL. Two receptor systems are involved in the plasma clearance of tissue factor pathway inhibitor in vivo. J Biol Chem. 1995;270:24800–24804.
Sandset PM, Warn-Cramer BJ, Rao LVM, Maki SL, Rapaport SI. Depletion of extrinsic pathway inhibitor (EPI) sensitizes rabbits to disseminated intravascular coagulation induced with tissue factor: evidence supporting a physiologic role for EPI as a natural anticoagulant. Proc Natl Acad Sci U S A. 1991;88:708–712.
Huang Z-F, Higuchi D, Lasky N, Broze GJ Jr. Tissue factor pathway inhibitor gene disruption produces intrauterine lethality in mice. Blood. 1997;90:944–951.