The Low Density Lipoprotein Receptor-related Protein Can Function Independently from Heparan Sulfate Proteoglycans in Tissue Factor Pathway Inhibitor Endocytosis*

Tissue factor pathway inhibitor (TFPI) is a plasma serine protease inhibitor that directly inhibits coagulation factor Xa and regulates blood coagulation via inhibition of factor VIIa-tissue factor enzymatic activity. We previously demonstrated that >90% of TFPI bound to a single population of low affinity binding sites on hepatoma cells (2 3 10 sites/cell, Kd 5 30 nM), and, that following binding, the low density lipoprotein receptorrelated protein (LRP) mediated TFPI uptake and degradation. We subsequently reported heparan sulfate proteoglycans (HSPGs) constitute a second receptor system involved in TFPI catabolism. In the present study, mouse embryonic fibroblasts heterozygous and homozygous-negative for disruption of the LRP gene were used to further examine the roles of LRP and HSPGs in TFPI endocytosis. We demonstrate that LRP is absolutely required for degrading I-TFPI. LRP heterozygous and homozygous-negative cells bind I-TFPI similarly, and the 39-kDa protein, an inhibitor of all known ligand interactions with LRP, does not alter I-TFPI binding to these cells. TFPI can be cross-linked to LRP on [S]cysteine-labeled hepatoma and LRP-heterozygous cells but not LRP-negative cells. When HSPGs are blocked with protamine, I-TFPI binds in a 39-kDa protein-inhibitable manner to 41,000 high affinity sites/ hepatoma cell (Kd 5 2.3 nM). Blockade of HSPGs with protamine results in significantly more I-TFPI degradation by LRP-positive cells. TFPI can be cross-linked to LRP in the absence and presence of protamine. However, in the presence of protamine, relative to the total pool of cross-linked proteins, 5-fold more TFPI is crosslinked to LRP. Finally, we show TFPI inhibits I-a2macroglobulin-methylamine binding to hepatoma cells and that carboxyl-terminal residues 115–319 of the 39kDa protein inhibit both I-TFPI degradation and binding when binding conditions contain protamine. Together, our results suggest that while the majority of TFPI binds to cell surface HSPGs, LRP can function independently fromHSPGs in the binding and uptake of TFPI. The low density lipoprotein receptor-related protein (LRP) is a cell-surface glycoprotein composed of two subunits of ;515 kDa and 85 kDa. The ;515-kDa subunit binds ligands and is noncovalently associated with the 85-kDa subunit which contains a single membrane-spanning domain and two NPXY internalization sequences (1). LRP functions as an endocytosis receptor for a rapidly increasing number of diverse ligands that are involved in lipoprotein metabolism, protease/coagulation regulation, and toxin entry (1–13). A 39-kDa protein, also termed receptor-associated protein, copurifies with and binds with high affinity to LRP (1, 14). Numerous studies have found that exogenously added 39-kDa protein inhibits the binding and/or cellular uptake of all ligands by LRP (1). In 1994, an LRP-negative cell line was established from the embryos of LRP-knockout mice (15). LRP-negative cell lines, in addition to the 39-kDa protein and antibodies directed against LRP, are an extremely useful means of establishing whether LRP functions as a receptor for candidate ligands. Using hepatoma cells, we previously reported (10) that LRP mediated the uptake and cellular degradation of tissue factor pathway inhibitor (TFPI), a 42-kDa plasma glycoprotein that inhibits both coagulation factor Xa and tissue factor-initiated blood coagulation (16). We also reported that LRP was not the major cell surface TFPI receptor since the 39-kDa protein did not inhibit I-TFPI binding (10). Because 10% of prebound TFPI was degraded via LRP, we speculated that #10% of TFPI bound directly to LRP which was then internalized and degraded. We also considered the possibility that after TFPI bound to another molecule, TFPI could be transferred to LRP for uptake and degradation since several previously characterized LRP ligands (7, 17) initially bind to other cell surface molecules prior to LRP-mediated uptake and degradation. Heparan sulfate proteoglycans (HSPGs) appear to play an important role in TFPI catabolism (18–21). Evidence for HSPG involvement has come from the following observations. 1) Plasma TFPI levels increase severalfold in vivo (18, 20) following intravenous administration of heparin, an effect attributed to the release of TFPI from endothelial or liver cell HSPG/ glycosaminoglycan-binding sites. Further, neutralization of heparin with protamine reduces TFPI levels to preheparin values, presumably by re-exposing TFPI to HSPG/glycosaminoglycan-binding sites (21). 2) Protamine, which competes for HSPG-binding sites, inhibits TFPI binding to hepatoma cells * This work was supported in part by National Institutes of Health Grants HL52040, HL53280, and HL34462. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. § Supported in part by Cardiovascular Training Grant T32-HL07275. i Lucille P. Markey Scholar. Supported by Grant HL20948, the Perot Family Foundation, and the Syntex Scholar Program. ‡‡ To whom correspondence should be addressed: Dept. of Pediatrics, Children’s Hospital, Washington University School of Medicine, One Children’s Place, St. Louis, MO 63110. Tel.: 314-454-6005; Fax: 314-454-0537. 1 The abbreviations used are: LRP, low density lipoprotein receptorrelated protein; TFPI, tissue factor pathway inhibitor; a2M*, a2-macroglobulin-methylamine; HSPGs, heparan sulfate proteoglycans; PAGE, polyacrylamide gel electrophoresis; GST, glutathione S-transferase; DTSSP, 3,39-dithiobis(sulfosuccinimidylpropionate); BSA, bovine serum albumin; PBS, phosphate-buffered saline; PBSc, PBS containing 1 mM CaCl2 and 0.5 mM MgCl2. THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 271, No. 42, Issue of October 18, pp. 25873–25879, 1996 © 1996 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in U.S.A.

The low density lipoprotein receptor-related protein (LRP) 1 is a cell-surface glycoprotein composed of two subunits of ϳ515 kDa and 85 kDa. The ϳ515-kDa subunit binds ligands and is noncovalently associated with the 85-kDa subunit which contains a single membrane-spanning domain and two NPXY internalization sequences (1). LRP functions as an endocytosis receptor for a rapidly increasing number of diverse ligands that are involved in lipoprotein metabolism, protease/coagulation regulation, and toxin entry (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13). A 39-kDa protein, also termed receptor-associated protein, copurifies with and binds with high affinity to LRP (1,14). Numerous studies have found that exogenously added 39-kDa protein inhibits the binding and/or cellular uptake of all ligands by LRP (1). In 1994, an LRP-negative cell line was established from the embryos of LRP-knockout mice (15). LRP-negative cell lines, in addition to the 39-kDa protein and antibodies directed against LRP, are an extremely useful means of establishing whether LRP functions as a receptor for candidate ligands.
Using hepatoma cells, we previously reported (10) that LRP mediated the uptake and cellular degradation of tissue factor pathway inhibitor (TFPI), a 42-kDa plasma glycoprotein that inhibits both coagulation factor Xa and tissue factor-initiated blood coagulation (16). We also reported that LRP was not the major cell surface TFPI receptor since the 39-kDa protein did not inhibit 125 I-TFPI binding (10). Because 10% of prebound TFPI was degraded via LRP, we speculated that Յ10% of TFPI bound directly to LRP which was then internalized and degraded. We also considered the possibility that after TFPI bound to another molecule, TFPI could be transferred to LRP for uptake and degradation since several previously characterized LRP ligands (7,17) initially bind to other cell surface molecules prior to LRP-mediated uptake and degradation. Heparan sulfate proteoglycans (HSPGs) appear to play an important role in TFPI catabolism (18 -21). Evidence for HSPG involvement has come from the following observations. 1) Plasma TFPI levels increase severalfold in vivo (18,20) following intravenous administration of heparin, an effect attributed to the release of TFPI from endothelial or liver cell HSPG/ glycosaminoglycan-binding sites. Further, neutralization of heparin with protamine reduces TFPI levels to preheparin values, presumably by re-exposing TFPI to HSPG/glycosaminoglycan-binding sites (21). 2) Protamine, which competes for HSPG-binding sites, inhibits TFPI binding to hepatoma cells and prolongs the half-life of 125 I-TFPI in mice (19). 3) Under conditions where TFPI is unable to bind to HSPGs, the 39-kDa protein inhibits 125 I-TFPI binding to hepatoma cells (19). 4) In mice overexpressing the 39-kDa protein, the rapid clearance of 125 I-TFPI is virtually eliminated in the presence of protamine (19).
The purpose of the present study was to demonstrate a direct interaction between TFPI and LRP, to define this affinity, and to further examine the role of HSPGs in TFPI binding and degradation. Domains on the 39-kDa protein required for inhibiting TFPI degradation and cross-competition studies between TFPI and previously characterized LRP-ligands have been examined.
Cell Culture-Rat hepatoma MH 1 C 1 cells (24) and mouse embryonic fibroblasts heterozygous (PEA10 cells) and homozygous-negative (PEA13 cells) for disruption of the LRP gene (15) were grown as described previously.
Cell Binding and Degradation Assays-Assay buffers for 125 I-TFPI and 125 I-␣ 2 M* were Earle's minimum essential medium (with glutamine) containing 3% BSA and Dulbecco's modified Eagle's medium containing 6 mg/ml BSA and 5 mM CaCl 2 , respectively. Cell binding assays were performed as described previously (10). Briefly, washed cells were incubated with assay buffer containing the indicated concentrations of 125 I-ligand in the absence or presence of unlabeled competitor proteins for 2 h at 4°C. The cells were then washed and lysed. Radioactivity of cell lysates was determined in a Packard ␥ counter. Total 125 I-ligand binding was determined in the presence of 125 I-ligand alone. Nonspecific 125 I-ligand binding was determined in the presence of excess unlabeled ligand as specified in the text. Specific 125 I-ligand binding was determined by subtracting nonspecific from total 125 Iligand binding.
Cell degradation assays (10) were performed by adding assay buffer containing 125 I-TFPI in the absence or presence of unlabeled competitor proteins to washed cell monolayers. After incubation at 37°C for 4 h, the overlying media were removed and precipitated with trichloroacetic acid. Degradation of ligand was defined as the appearance of radioactive ligand fragments in the overlying media that were soluble in trichloroacetic acid. Degradation of 125 I-TFPI in parallel dishes that did not contain cells was subtracted from each point.
Metabolic Labeling and Chemical Cross-linking-Cells were grown in 10-cm dishes to 80% confluency. After incubation in medium lacking cysteine, metabolic labeling was initiated by the addition of [ 35 S]cysteine for 5 h as specified in the figure legends. Thereafter, monolayers were washed and incubated at 4°C with buffer containing 10 nM 39-kDa protein or 100 nM TFPI in the absence or presence of 100 g/ml protamine. Cells were then washed with PBS containing 1 mM CaCl 2 and 0.5 mM MgCl 2 (PBSc). Chemical cross-linking (24,26) was performed with 1 mM DTSSP in PBSc for 30 min at 4°C followed by quenching with Tris-buffered saline and solubilization in PBS containing 1% (v/v) Triton X-100 and 1 mM phenylmethylsulfonyl fluoride. After sonication, cell lysates were centrifuged in a Microfuge, and the supernatants were used for immunoprecipitation.
Immunoprecipitation-As described (24,26), cell lysates were mixed with an equal volume of PBS containing 1% (v/v) Triton X-100, 0.5% (w/v) sodium deoxycholate, 1% (w/v) sodium dodecyl sulfate (SDS), and 0.5% (w/v) bovine serum albumin and precleared overnight with normal rabbit serum followed by protein A-agarose. The precleared samples were divided into equal portions and incubated with primary antibody. After incubation with protein A-agarose beads, the immunoprecipitates were washed. In some experiments, the protein A-agarose beads were divided into two equal parts for reducing and nonreducing conditions before the final wash. The immunoprecipitated material was released from the beads by boiling in Laemmli sample buffer (27) with or without 5% (v/v) 2-mercaptoethanol. Samples were analyzed by 6% SDS-polyacrylamide gel electrophoresis (PAGE). Gels were impregnated with Amplify, dried, and exposed to film. Densitometric scanning of the gels was performed on a Molecular Dynamics PhosphorImager. Densitometric scanning of nonreduced samples was quantified by adding the amount of radioactivity retained in the stacking gel, at the top of the separating gel, and at the position LRP. For reduced samples, the amount of radioactivity was determined by scanning the gels at the position of LRP.

RESULTS
Chemical Cross-linking of TFPI and the 39-kDa Protein to LRP on Hepatoma Cells-To demonstrate a direct interaction between TFPI and LRP, chemical cross-linking was performed using DTSSP, a thio-cleavable, water-soluble, and membraneimpermeant reagent. In Fig. 1, [ 35 S]cysteine-labeled hepatoma cells were subjected to cross-linking following binding with TFPI or the 39-kDa protein. Cell lysates, immunoprecipitated with anti-TFPI, anti-39-kDa protein, anti-LRP IgGs, or normal rabbit serum were analyzed by SDS-PAGE. When cells were cross-linked to TFPI or the 39-kDa protein, a complex with very high apparent molecular mass was seen under nonreducing . Cells were either lysed without cross-linking or were lysed following cross-linking with 1 mM DTSSP. Lysates were immunoprecipitated with 10 g of total anti-TFPI IgG (␣-TFPI), 1 g of affinity purified anti-LRP IgG (␣-LRP), 3 g of affinity-purified anti-39-kDa protein IgG (␣-39K), or 10 l of normal rabbit serum (N.R.). The immunoprecipitates were analyzed by 6% SDS-PAGE under nonreducing and reducing conditions. Gels were exposed to film for 30 h prior to developing. The position of LRP is indicated by an arrowhead. Molecular mass markers in kDa are indicated on the left.
conditions following immunoprecipitation with anti-TFPI ( Fig.  1A) or anti-39-kDa protein (Fig. 1B) IgG. These high molecular mass bands migrated to the identical position as that seen in cell lysates immunoprecipitated with anti-LRP antibody (Fig.  1, A and B), implying TFPI and the 39-kDa protein are binding to LRP. Under reducing conditions and resultant cleavage of the cross-linker, the ϳ515-kDa subunit of LRP was observed using anti-TFPI (Fig. 1A), anti-39-kDa protein (Fig. 1B), and anti-LRP IgGs (Fig. 1, A and B). In the absence of cross-linking, no high molecular mass complex was evident with anti-TFPI ( Fig. 1A) or anti-39-kDa protein IgG (Fig. 1B). In the absence of cross-linking, LRP was not co-immunoprecipitated with anti-39-kDa protein IgG since immunoprecipitation conditions contain 1% SDS. Immunoprecipitation of non-cross-linked lysates with anti-LRP IgG (Fig. 1, A and B) resulted in the appearance of the ϳ515-kDa subunit of LRP under both nonreducing and reducing conditions. No proteins are immunoprecipitated under any condition using normal rabbit serum.
Chemical Cross-linking of TFPI to LRP on Cells Heterozygous for Disruption of the LRP Gene-We next examined whether TFPI could be cross-linked to LRP-heterozygous PEA10 cells. The identical cross-linking experiments were performed on LRP-negative PEA13 cells. Following [ 35 S]cysteine labeling, cells were incubated with unlabeled TFPI and thereafter cross-linked with DTSSP. Fig. 2 demonstrates that LRP is synthesized by PEA10 cells but not by PEA13 cells since the ϳ515-kDa subunit of LRP is only immunoprecipitated with anti-LRP IgG from PEA10 cell lysates. The ϳ515-kDa subunit of LRP is also immunoprecipitated with anti-TFPI IgG from PEA10 cell lysates but not from PEA13 cell lysates, demonstrating TFPI binding directly to LRP.
Effect of the 39-kDa Protein on 125 I-TFPI Degradation and Binding by LRP-negative Cells-To investigate whether LRPnegative cells were capable of mediating the uptake and degradation of 125 I-TFPI, PEA10 and PEA13 cells were incubated with 125 I-TFPI in the absence or presence of the 39-kDa protein at 37°C, and degradation products were assessed. Fig. 3A shows PEA10 cells actively degrade 125 I-TFPI and that the 39-kDa protein inhibits this degradation in a dose-dependent manner with an IC 50 value 2 of ϳ15 nM. PEA13 cells degrade 125 I-TFPI minimally, and the 39-kDa protein has no effect on the amount of 125 I-TFPI degraded. The small amount of 125 I-TFPI degraded by PEA13 cells (ϳ40 fmol/10 6 cells/4 h) may represent an LRP-independent process or more likely is the result of pinocytic internalization since mouse fibroblasts nonspecifically pinocytose ligand at a rate of 0.9 m 3 /cell/min (28). Fig. 3B shows that PEA10 and PEA13 cells bind 125 I-TFPI similarly and that the 39-kDa protein does not alter 125 I-TFPI binding to either cell line.
Saturation Binding of 125 I-TFPI to MH 1 C 1 Cells in the Absence and Presence of Protamine-We previously reported that the majority of 125 I-TFPI binding was not to LRP (10), and that protamine, a competitor for HSPG-binding sites, inhibited 125 I-TFPI binding to MH 1 C 1 cells (19). We also reported that when protamine was included in the binding medium, the 39-kDa protein inhibited 125 I-TFPI binding (19). Thus, under conditions where TFPI was unable to bind to HSPGs, more TFPI was available for binding to LRP. Therefore, to examine the affinity of TFPI binding to LRP, saturation binding experiments were performed with 125 I-TFPI in the absence or presence of protamine. As seen in Fig. 4A, in the absence of protamine, 125 I-TFPI bound specifically to MH 1 C 1 cells over the concentration range of 0.6 -12 nM. Nonspecific binding increased linearly and accounted for 10% of total 125 I-TFPI binding. Saturation of specific binding was not reached at a 125 I-TFPI concentration of 12 nM. Scatchard analysis (29) yielded ϳ5 ϫ 10 6 sites/cell with an apparent K d value of 20 nM (Fig. 4A, inset), similar to our previous study (10). In Fig. 4B, saturation binding of 125 I-TFPI was performed in the presence of 100 g/ml protamine, a concentration previously shown (19) to inhibit 90% of 125 I-TFPI binding to MH 1 C 1 cells. In the presence of protamine, 125 I-TFPI bound specifically to MH 1 C 1 cells, and saturation of specific binding was observed at a 125 I-TFPI concentration of ϳ7 nM. Protamine reduced 125 I-TFPI binding by Ͼ90%. Nonspecific binding increased linearly and accounted for ϳ40% of total binding. Scatchard analysis (29) yielded 41,000 sites/cell with an apparent K d value of 2.3 nM (Fig. 4B, inset).  shows that in the presence of 100 g/ml protamine and 500 nM 39-kDa protein, 125 I-TFPI binding is inhibited and is identical to nonspecific 125 I-TFPI binding.
Effect of Protamine on the Cellular Degradation of 125 I-TFPI in the Absence and Presence of the 39-kDa Protein-Since 39-kDa protein-inhibitable 125 I-TFPI binding was observed in the presence of protamine ( Fig. 4B and Ref. 19), we hypothesized that by blocking TFPI from binding to cell surface HSPGs, more TFPI could bind to LRP. Thus, the effect of protamine on 125 I-TFPI degradation was examined. In the absence of protamine, the 39-kDa protein inhibited 125 I-TFPI degradation in a dose-dependent manner in MH 1 C 1 (Fig. 5A) and PEA10 (Fig.  5B) cells. 1 M 39-kDa protein inhibited 70 -80% of 125 I-TFPI degradation by these cell lines. PEA13 cells (Fig. 5B) degrade 125 I-TFPI minimally. In the presence of protamine, the amount of 125 I-TFPI degraded by MH 1 C 1 (Fig. 5A) and PEA10 (Fig. 5B) cells was approximately doubled and the 39-kDa protein inhibited Ͼ95% of this degradation. In contrast to the LRP-positive MH 1 C 1 and PEA10 cells, protamine inhibited 125 I-TFPI degradation by LRP-negative PEA13 cells (Fig. 5B), suggesting a minor fraction of 125 I-TFPI degradation is mediated by a protamine-sensitive LRP-independent process.
Chemical Cross-linking of TFPI to LRP in the Absence and Presence of Protamine-The ability of protamine to alter the amount of 125 I-TFPI cross-linked to LRP was next examined. In Fig. 6, [ 35 S]cysteine-labeled MH 1 C 1 cells were subjected to cross-linking following binding with TFPI in the absence or presence of protamine. Under nonreducing conditions, a very high apparent molecular mass complex remaining largely in the stacking gel was immunoprecipitated with anti-TFPI antibody when TFPI was cross-linked to cells following binding in the absence (lane 1) and presence (lane 7) of protamine. Densitometric scanning revealed that ϳ4-fold more radioactivity  5, 11, and 17). In the absence of cross-linking, LRP was immunoprecipitated with anti-LRP antibody (lanes 14 and 17) and not with anti-TFPI antibody (lanes 13 and 16). No proteins were immunoprecipitated using normal rabbit serum (lanes 3, 6, 9, 12, 15, and 18).
Inhibition of 125 I-TFPI Degradation by 39-kDa Protein Constructs-To define regions of the 39-kDa protein required for the inhibition of 125 I-TFPI interaction with LRP, the ability of a series of GST-fusion proteins encoding distinct regions of the 39-kDa protein to inhibit 125 I-TFPI degradation by MH 1 C 1 cells was examined. Fig. 7A shows the GST-fusion protein encoding carboxyl-terminal residues 115-319 of the 39-kDa protein (GST/115-319) inhibits 125 I-TFPI degradation identically to the full-length GST-39-kDa protein (GST/1-319). Both GST/1-319 and GST/115-319 inhibited 80% of 125 I-TFPI degradation at concentrations of 1 M. GST/115-319 (500 nM) also inhibited 85% of specific 125 I-TFPI binding at 4°C when protamine (100 g/ml) was included in the binding medium but did not alter 125 I-TFPI binding when protamine was omitted (data not shown). Cross-competition Binding Experiments between TFPI and LRP-Ligands-Competition binding experiments on MH 1 C 1 cells were performed to determine whether TFPI could alter the binding of previously characterized LRP-ligands. As seen in Table I, the 39-kDa protein and TFPI inhibit 94% and 85% of specific 125 I-␣ 2 M* binding, respectively. Each inhibits 125 I-␣ 2 M* binding with an IC 50 value of 20 nM (data not shown).

DISCUSSION
Using hepatoma cells, we previously demonstrated that LRP mediates the cellular degradation of 125 I-TFPI since degradation was inhibited by antibodies directed against LRP and by the 39-kDa protein (10). In the present study, we demonstrate that LRP is required for the degradation of 125 I-TFPI since LRP-heterozygous cells, in contrast to LRP-negative cells, actively degrade 125 I-TFPI in a 39-kDa protein-inhibitable manner. Fragments of the 39-kDa protein, which differentially regulate ␣ 2 M* and tissue-type plasminogen activator binding to LRP (23), also differentially inhibit 125 I-TFPI degradation and HSPG-independent 125 I-TFPI binding. The observation that protamine enhances 125 I-TFPI degradation by LRP-positive MH 1 C 1 and PEA10 cells is potentially important. While prevention of TFPI from binding to HSPGs may provide additional TFPI for uptake and degradation by LRP, this is unlikely since the absolute concentration of free 125 I-TFPI is essentially unchanged in the absence and presence of protamine. An alternative explanation may be that the "effective" 125 I-TFPI concentration is increased by protamine: protamine may increase the affinity and/or efficiency of TFPI endocytosis by LRP. The finding that 1 M 39-kDa protein inhibits Ͼ95% of 125 I-TFPI degradation by LRP-positive cells in the presence of protamine whereas 70 -80% is inhibited in the absence of protamine supports this possibility. Further studies will be necessary to specifically define this mechanism. Nonetheless, the use of protamine provides a strategy to dissect the role of LRP in TFPI endocytosis, and together our results suggest that the direct binding of TFPI to LRP, rather than the transfer of TFPI from HSPGs to LRP, is the predominant mechanism by which TFPI endocytosis occurs, i.e. if transfer was an efficient mechanism, protamine should not enhance, and should, in fact, inhibit 125 I-TFPI degradation since protamine would prevent 125 I-TFPI binding to HSPGs and subsequent transfer to LRP.  In addition, if transfer occurred, the 39-kDa protein should inhibit 125 I-TFPI degradation to a similar extent in the absence and presence of protamine, unless transfer was a very slow process. Additional data supporting a direct interaction between TFPI and LRP and this mechanism as responsible for TFPI uptake and degradation are seen in our cross-linking experiments. After cross-linking unlabeled TFPI to [ 35 S]cysteine-labeled MH 1 C 1 and LRP-heterozygous cells, LRP is immunoprecipitated using antibodies directed against TFPI. No specific cross-linking of TFPI to any species is observed on LRP-negative cells. Consistent with the hypothesis that more TFPI binds to LRP when HSPGs are blocked with protamine is the observation that relative to the total cohort of cross-linked proteins, 5-fold more TFPI is cross-linked to LRP in the presence of protamine. In other words, in the presence of protamine, substantially less TFPI binds to the cell surface so fewer cell surface proteins would be cross-linked to TFPI. Indeed, under nonreducing conditions and in the presence of protamine, ϳ4fold less total radioactivity is immunoprecipitated with anti-TFPI antibody. Under reducing conditions, protamine does not alter the amount of LRP immunoprecipitated with anti-TFPI antibody since saturating concentrations of TFPI (ϳ5 times the K d value) would saturate LRP in the presence of protamine and would saturate both LRP and the major TFPI binding species in the absence of protamine.
Following reduction and SDS-PAGE, the fate of the remaining radioactivity initially retained in the stacking gel under nonreducing conditions in the absence of protamine could be attributed to the fact that when unlabeled TFPI is cross-linked to cell surface proteins, it is cross-linked not only to LRP but also to numerous other species of various molecular masses. Under nonreducing conditions, these cross-linked complexes are retained in the stacking gel, whereas under reducing conditions, the radioactivity from these complexes is widely distributed throughout the gel lane.
As stated, 4-fold more radioactivity was retained in the stacking gel under nonreducing conditions in the absence of protamine than in the presence of protamine. Yet, under reducing conditions, only LRP was evident in the absence (and presence) of protamine. In the absence of protamine, it is not clear why the major TFPI binding species was not also apparent following reduction of the cross-linker. One possibility is that the majority of TFPI binding is to a cell surface carbohydrate moiety and will not be cross-linked to TFPI using DTSSP. Another possibility is that free amine groups on the major TFPI binding species are not accessible to cross-linking with DTSSP. It is also conceivable that there are no cysteines available for metabolic labeling within this TFPI binding species. Finally, this molecule may not be solubilized with 1% Triton X-100. Support for this latter explanation comes from a recent study (30) which reported that endogenous TFPI is bound to the endothelial cell surface by a glycosylphosphatidylinositollinked receptor that is insoluble in Triton X-100.
Another line of evidence supporting a direct interaction between TFPI and LRP comes from our saturation binding experiments on hepatoma cells performed in the absence and presence of protamine. In the absence of protamine, TFPI binds to 2-5 ϫ 10 6 low affinity sites/cell (K d ϭ 20 -30 nM) (10) whereas in the presence of protamine, TFPI binds to 41,000 high affinity sites/cell (K d ϭ 2.3 nM). Previous reports on ligand binding affinities and numbers of binding sites/cell of LRPligands are similar to that of TFPI binding in the presence of protamine. For example, the K d value for tissue-type plasminogen activator and 39-kDa protein binding to LRP is 3-6 nM (24, 31) whereas higher affinities, 40 pM and 2 nM, have been reported for ␣ 2 M* (32). The number of tissue-type plasminogen activator binding sites on hepatoma cells is 60,000 -100,000 (24,31). Thus, the similarity in affinity and number of binding sites/cell between TFPI in the presence of protamine and previously characterized LRP-ligands is consistent with direct interaction between TFPI and LRP.
Finally, the findings that TFPI inhibits 125 I-␣ 2 M* binding and that ␣ 2 M* partially inhibits 125 I-TFPI binding when HSPGs are blocked with protamine also suggest that at least a fraction of cell surface TFPI binding is directly to LRP since ␣ 2 M* is known to bind only to LRP. The inability of unlabeled ␣ 2 M* to alter 125 I-TFPI binding supports our hypothesis that only a small fraction of cell surface TFPI binding is to LRP. Our results further suggest TFPI and ␣ 2 M* may bind to the same site on LRP. Differences in on/off rates for TFPI and ␣ 2 M* could account for the varying extents of inhibition of binding observed. Alternatively, TFPI and ␣ 2 M* may function as noncompetitive inhibitors for binding and therefore bind to distinct sites on LRP. Future competition binding experiments between TFPI and ␣ 2 M* are required to resolve these issues.
In summary, we have shown that TFPI binds directly to LRP, that LRP is required for mediating the cellular degradation of TFPI, and that under conditions where TFPI is unable to bind to HSPGs, TFPI binds in a 39-kDa protein-inhibitable manner specifically and with high affinity to the hepatoma cell surface. Thus, binding, uptake, and degradation of TFPI can occur independent of HSPGs. The two receptor systems involved in TFPI catabolism appear to have different functions: LRP serves to remove TFPI from the circulation, whereas HSPGs may anchor TFPI to the cell surface so that TFPI remains in contact with plasma proteins to function as an anticoagulant.