Tissue Factor Pathway Inhibitor in Endothelial Cells Colocalizes With Glycolipid Microdomains/Caveolae
Regulatory Mechanism(s) of the Anticoagulant Properties of the Endothelium
Abstract Tissue factor pathway inhibitor (TFPI), the main downregulator of the procoagulant activity of tissue factor•factor VIIa complex, locates in human endothelial cells (EC) in culture as well-defined clusters uniformly distributed both on the cell surface and intracellularly. We here demonstrate by immunofluorescence that TFPI colocalizes in EC with caveolin, urokinase-type plasminogen activator receptor, and glycosphingolipids. The localization of TFPI in caveolae in resting endothelium is proved by double immunogold electron microscopy for TFPI and caveolin. After ultracentrifugation of rat lung or EC homogenates through density gradients of Nycodenz, TFPI was highly enriched at densities of 1.05 to 1.08 g/mL, together with caveolin and alkaline phosphatase. By ELISA, more than half of the cellular TFPI was detected in Triton X-100-insoluble extracts of EC. TFPI incorporates [1-3H]ethanolamine and is cleaved from the cell surface by phosphatidylinositol–phospholipase C, indicating a specific glycosylphosphatidylinositol-anchorage mechanism for TFPI in the plasma membrane. Clustering of TFPI and its localization in caveolae are dependent on the presence of cholesterol in the membrane. Agonist-induced stimulation of EC caused marked changes of distribution for both TFPI and caveolin at subcellular level, with subsequent increase of the cell surface–associated inhibitory activity toward tissue factor•factor VIIa. Our findings suggest that, beside their function in transcytosis, potocytosis, cell surface proteolysis, and regulation of signal transduction, caveolae also play a direct role in the regulation of EC anticoagulant properties.
- Received March 4, 1997.
- Accepted July 31, 1997.
Endothelial cells (EC) play a central role in the regulation of hemostasis by ensuring the cellular control of both procoagulant and anticoagulant mechanisms. The anticoagulant and profibrinolytic functions predominate in the quiescent state of the endothelium, thus maintaining the blood fluidity (for review, see References 1 and 21 2 ). Blood coagulation is initiated when FVII/VIIa in plasma gains access to TF at sites of blood vessel injury and the resulting TF•FVIIa complex activates factor X to Xa and factor IX to IXa, leading to thrombin generation and the formation of a fibrin clot.3
Although healthy EC do not express TF constitutively, marked expression of this protein occurs both in vitro after perturbation of EC with different agonists4–6 and in vivo, during sepsis7 or within the tumor vasculature.8 Inhibition of TF-induced procoagulant activity occurs primarily via the Kunitz-type TF pathway inhibitor (TFPI),9 which thus contributes to the maintainance of the anticoagulant properties of the endothelium.
TFPI is the most physiologically significant inhibitor of the TF•FVIIa complex, but for exerting its full activity, TFPI needs a small amount of FXa to be generated first.9 Then TFPI uses the tandem Kunitz-type domains in its structure to form a quaternary complex with FXa bound to TF•FVIIa.10
Most studies conducted to date to elucidate the functional and kinetic properties of TFPI make reference either to the TFPI circulating in plasma or to the interaction of recombinant TFPI with TF•FVIIa and FXa in cellular or noncellular systems. These studies have provided important insights into the kinetics of TFPI but also revealed discrepancies when the properties of rTFPI11 were compared to those of cell-expressed TFPI.12
The major pool of TFPI resides in the endothelium, which constitutively expresses the protein in normal conditions.13–15 From circumstantial evidence, it was inferred that TFPI secreted from the cells anchors to sulfated proteoglycans in the glycocalix,16 but to the best of our knowledge, systematic studies concerning the regulation of the protein synthesis, intracellular storage and localization, and secretory events of TFPI in EC have not been undertaken so far.
We showed previously that TFPI in resting EC in culture locates as uniform clusters both on the surface and within an intracellular apical pool and found that, after acute stimulation with thrombin, the anticoagulant potency of the EC toward the TF•FVIIa complex increased significantly, owing to TFPI redistribution and enhanced exposure on the plasma membrane.15 Critical requirements for the assembly of TF•FVIIa complex with cellular TFPI on EC were also described.12 Accordingly, cellular TFPI plays a more important role than the fluid-phase form of the inhibitor in the maintainance of the hemostatic balance.
Working with activated EC, which expressed TF after stimulation with tumor necrosis factor, Sevinsky et al showed that downregulation of TF•FVIIa-dependent activation of FX is mediated by TFPI and occurs by translocation of the quaternary complex into low-density, detergent-insoluble cellular fractions enriched in caveolin.12 Nevertheless, we are unaware of any direct morphologic proof for the localization of TFPI in caveolae in native/unstimulated endothelial cells.
We show here that in resting endothelium, TFPI is exposed on the cell surface via a GPI link, and the inhibitor is specifically targeted by apical vectorial delivery to restricted plasmalemma microdomains of special proteolipidic composition and particular function (caveolae). Our study also analyzes the potential mechanism of TFPI/caveolae-mediated upregulation of EC anticoagulant properties after stimulation with specific agonists.
Antibodies and suppliers are as follows: rabbit anti-full-length rTFPI IgG was developed in our laboratory, immunoaffinity purified, and tested for specificity (50 μg IgG/mL produced 95% inhibition of TFPI activity in normal human plasma and the concentration of IgG that still produced 50% inhibition of TFPI activity was ≈0.6 μg/mL; by immunofluorescence, the experiments of competition, such as staining in the presence of rTFPI, were entirely negative); monoclonal anti-K2TFPI, American Diagnostica (No. 4904, Alpha Labs, Eastleigh, UK); monoclonal anti-caveolin IgG (No. C13620, clone C060, raised against an 11.1-kDa N-terminal fragment of human caveolin), Transduction Laboratories (Affinity Research Products, Mamhead, UK); monoclonal anti-uPAR IgG (gift from Dr Vincent Ellis of our institute); and monoclonal anti-AP IgG (Sigma Chemical Co. Ltd, Poole, UK). [1-3H]-ethanolamine hydrochloride (18 Ci/mmol), Amplify, and Hyperfilm were from Amersham Int (Little Chalfont, UK). Phosphatidylinositol phospholipase C from Bacillus cereus (PI-PLC, 600 U/mg) was from Boehringer Mannheim (Lewes, UK). OGP, aprotinin (>9000 KIU/mL), and Pansorbin were from Calbiochem Novabiochem (Nottingham, UK). Secondary antibodies conjugated to FITC or Texas Red, Vectashield, and AP-substrate kit were from Vector Laboratories, Inc. (Burlingame, California). Human α-thrombin (3000 NIH U/mg) was a gift from Dr J.-M. Freyssinet (Institut d’Hematologie et d’Immunologie, Strasbourg, France). Immobilon P and Ultrafree-CL microcentrifuge filters (NMWL-10,000) were from Nihon Millipore (Kogyo KK, Japan). Nycodenz was from Nycomed AS (Oslo, Norway). Secondary antibodies conjugated to 5- or 10-nm colloidal gold were from BioCell Research Labs (Cardiff, UK), and all the other reagents used for electron microscopy were from TAAB Laboratory Equipment Ltd (Aldermaston, UK). CT-B conjugated to FITC, as well as secondary antibodies conjugated with biotin, cell culture media and supplements, Triton X-100, filipin, nystatin, mannosamine (2-amino-2-deoxy D-mannose), heparinase III from Flavobacterium heparinum (heparitinase, 200 to 600 U/mg), calcium ionophore A23187, sodium chlorate, PMSF, 2-[N-morpholino]ethanesulfonic acid (MES), N-[2-hydroxyethylpiperazine]-N′-2-ethanesulfonic acid (HEPES), Tris, BSA, sodium orthovanadate, triethanolamine (TEA), and all other reagents were purchased from Sigma.
We used the immortalized human endothelial cell line EA.hy926 described by Edgell et al17 and HUVEC obtained from fresh umbilical cords.18 The EC split at low passage were grown on Petri dishes (60-mm diameter), T-75 flasks, or glass coverslips precoated with gelatin, either in Dulbecco’s modified Eagle medium or in M199, containing 4 mmol/L l-glutamine, 100 U/mL penicillin, 0.1 mg/mL streptomycin, 15 mmol/L HEPES, and 1:5 diluted heat-inactivated fetal bovine serum. All cells were used at ≈80% morphological confluence, and the assays were carried out in serum-free medium.
EC were rinsed with HEPES buffer (10 mmol/L HEPES, 0.137 mol/L NaCl, 2.7 mmol/L KCl, 5 mmol/L glucose, 5 mmol/L CaCl2) pH 7.4, supplemented with 2 g/L BSA, and incubated either with 1 NIH U thrombin/mL for 30 minutes at 37°C15 or with 5 μmol/L calcium ionophore A23187 for 10 minutes at 37°C. Supernatants were assayed for TFPI antigen and activity, and monolayers were fixed with 3% (w/v) paraformaldehyde (PFA) in PBS for 1 hour at room temperature and processed for immunostaining or used immediately for assaying the TFPI anticoagulant potency on the cell surface.
Sodium Chlorate or Heparitinase
To remove surface heparan sulfate proteoglycans, cells were incubated for 2 hours at 37°C with 1 U/mL heparitinase in serum-free medium. To inhibit sulfation of glycosaminoglycans, EC were cultivated for 24 hours in the presence of 30 mmol/L sodium chlorate (modified from Gleizes et al19). Supernatants and monolayers were harvested and processed as described above.
Phosphatidylinositol-Specific Phospholipase C (PI-PLC)
To identify a putative GPI-anchor, EC were incubated with 0.1 U PI-PLC/mL in 70 mmol/L TEA buffer pH 7.5, supplemented with 2 g/L BSA, for up to 2 hours at 37°C.20 Supernatants were harvested and cell monolayers processed as above.
EC were incubated for up to 4 hours in glucose-free medium (RPMI 1640) in the presence or absence of 10 mmol/L mannosamine, which blocks the incorporation of GPI glycans without affecting the synthesis of the protein itself.21 Supernatants and cells were used for quantitative estimations and immunostaining.
Filipin or Nystatin
EC were rinsed with ice-cold PBS, incubated for 15 minutes at 4°C with one of the cholesterol complexant agents filipin (5 mg/mL) or nystatin (100 mg/mL), both dissolved in dimethylsulfoxide,22 then fixed with PFA in PBS for 30 minutes at room temperature, rinsed with PBS, and processed for quantitative estimations or immunofluorescence.
We used the indirect immunofluorescence procedure previously described.15 For surface staining, cells were incubated with the primary antibodies (anti-TFPI IgG, anti-uPAR IgG) or CT-B-FITC (10 μg/mL), either live at 6°C to 8°C before fixation or after fixation at room temperature. For the intracellular detection of TFPI, caveolin, uPAR, and GSL, fixed cells were permeabilized with 0.2 g/L saponin before immunostaining.
In some experiments, native EC were incubated on ice with anti-TFPI IgG, then treated with 1% (w/v) Triton X-100 for 30 minutes on ice and fixed with PFA.23 Alternatively, after the incubation with anti-TFPI IgG at 6°C to 8°C, some EC were warmed for 1 hour at 37°C, fixed, permeabilized, and incubated with anti-caveolin IgG (modified from Parton et al24).
For all the conditions described, samples were rinsed, incubated with the appropriate mixtures of secondary antibodies conjugated to Texas Red or FITC, and mounted with Vectashield on glass slides. Fluorescence microscopy and digital image collection were performed with a Bio-Rad MRC 600 confocal laser unit attached to a Nikon Diaphot inverted microscope (Bio-Rad Microscience Ltd).
Detergent Extraction of the Cells
Sequential extraction of EC25,26 was performed as follows: EC monolayers were incubated for 30 minutes on ice with 1% Triton X-100 in MES-buffered saline (25 mmol/L MES, 0.15 mol/L NaCl, pH 6.5) containing 1 mmol/L PMSF, 10 mg/mL aprotinin, 1 mmol/L sodium orthovanadate, and 10 mmol/L EDTA. The supernatant representing Triton-soluble proteins was removed, microfuged for 2 minutes, and frozen. The remaining cells were lysed for 30 minutes on ice on a rocking platform with 60 mmol/L OGP in Tris-buffered saline (TBS, 0.1 mol/L Tris-HCl, 0.15 mol/L NaCl, pH 7.8) plus the same inhibitors cocktail. The OGP lysate was collected from the plates and cleared by microfugation, and supernatant, assumed to represent Triton-insoluble material, was frozen down.
Metabolic Radiolabeling and Immunoprecipitation
EC were incubated with [1-3H]ethanolamine (20 mCi/mL)21 for 16 hours, with and without 5 mmol/L mannosamine. The mannosamine-treated cells and half of the native ones were washed and lysed with 60 mmol/L OGP in TBS to obtain whole cell lysates. The other half of resting cells were washed and incubated either for 4 hours at 37°C in TEA buffer only or for 2 hours with 0.1 U/mL PI-PLC as described above. Supernatants were collected and microfuged for removal of cellular debris, and protease inhibitors were added. TFPI in cellular lysates and supernatants was immunoprecipitated by using a combination of techniques:25,27 Samples diluted in 0.1 mol/L Tris-HCl, pH 8.1, containing 1 g/L Tween 80 and 1 g/L SDS, were precleared with nonimmune rabbit IgG and Pansorbin, then incubated with 5 μg/mL of anti-TFPI IgG for 2 hours at 37°C and adsorbed with Pansorbin. The immunosorbent was then washed three times by centrifugation with the same detergent-containing buffer supplemented with 1 mol/L NaCl, then boiled in Laemmli sample buffer, including reducing agent, and subjected to SDS-PAGE. Radiolabeled proteins were detected by fluorography using Amplify and exposure to Hyperfilm, according to the manufacturer’s instructions.
Cell supernatants collected after different treatments were concentrated in Ultrafree. CL filters, and protease inhibitors were added. Total protein in concentrated supernatants and cell lysates was precipitated with 72% (w/v) TCA,28 and precipitates were resuspended in Laemmli sample buffer, boiled, and subjected to SDS-PAGE. Proteins on gels were electrotransferred onto Immobilon P29, and TFPI was detected with the anti-TFPI IgG followed by biotin-conjugated secondary antibody, streptavidin-gold (AuroProbe BL Plus kit from Amersham), and silver enhancement.
We performed both preembedding and postembedding immunogold localization of TFPI. For preembedding, cell monolayers were fixed for 30 minutes at room temperature with a mixture of 2% PFA and 0.05% glutaraldehyde in PBS and washed with PBS, quenched with 0.1 mol/L glycine in PBS, blocked for 30 minutes with Dako antibody diluent supplemented with 1:100 diluted nonimmune goat serum, and incubated for 1 hour at room temperature with anti-TFPI IgG (50 μg/mL), followed by goat anti-rabbit IgG coupled to 10-nm gold particles (diluted 1:100). After washing, monolayers were fixed for 30 minutes in 2.5% glutaraldehyde in 0.1 mol/L Na cacodylate-HCl buffer, postfixed for 1 hour in 1% OsO4 in the same buffer, scraped off the dishes, washed, dehydrated as pellet in graded ethanol, embedded in Spurr resin, and polymerized for 2 hours at 70°C.
For postembedding, endothelial cells were fixed with 3% PFA in PBS for 90 minutes at room temperature, while mouse lung tissue was fixed by perfusion with 3% PFA in PBS for 5 minutes followed by immersion fixation at room temperature for a further 1 hour. Both EC and mouse lung were dehydrated in ascending series of ethanol while the temperature was progressively lowered, following the protocol referred to as PLT (progressive lowering of temperature).30 The samples were embedded overnight in 100% Lowicryl K4M at −35°C, and polymerization was carried out by irradiation in gelatin capsules using indirect UV light in an RMC automatic low-temperature embedding tissue processor for 24 hours at −35°C. The process was finalized by hardening under UV light at room temperature for 1 to 2 days. Thin sectioning was carried out on a Reichert Ultracut microtome (Reichert-Jung Optische Werke, Vienna, Austria), and sections placed on Formvar-coated 200-mesh nickel grids were immunogold labeled essentially as described.31 In brief, sections were quenched, blocked, and incubated for 2 hours at 37°C with each of the following mixtures: for EC, rabbit anti-TFPI IgG and mouse anti-caveolin IgG (both at 50 μg/mL), then goat anti-rabbit IgG-Au10 plus goat anti-mouse IgG-Au5 (1:50 dilution for both); for mouse lung, anti-TFPI IgG followed by goat anti-rabbit IgG-Au10, at the same dilutions as above. After washing, grids were counterstained by sequential incubation in 1% OsO4 (40 minutes), 2.5% uranyl acetate (1 minute), and lead citrate (15 seconds). All the samples were examined with a Philips 201 electron microscope (Eindhoven, The Netherlands). Controls included omission of the first antibodies or replacement with nonimmune rabbit and mouse IgGs.
Nycodenz Density Gradient Centrifugation
The fractionation procedure was performed as described.32 Briefly, rat lung tissue or HUVEC monolayers were homogenized by using a tissue homogenizer, with 5 mmol/L Tris-HCl buffer, pH 7.4, containing 0.22 mol/L sucrose and 0.1 g/L Tween 80, and centrifuged at 800g for 5 minutes. Supernatants representing total homogenates were fractionated through a Nycodenz stepwise gradient (5 to 35 mg/mL) by ultracentrifugation for 2.5 hours at 202 000g, after which 1-mL fractions were collected (densities ranged from 1.03 to 1.17 g/mL). We assayed each fraction by ELISA for TFPI antigen15 and dot-blot for AP activity and caveolin concentration. For detection of AP,28 samples from each fraction were dotted on PVDF membrane and assayed with substrate from a Vector AP-substrate kit. For caveolin and TFPI immunodetection, gradient fractions were incubated for 30 minutes with OGP (final concentration: 60 mmol/L), then applied on the PVDF membrane and blotted with the monoclonals anti-caveolin IgG or anti-K2 TFPI, each followed by biotin-conjugated anti-mouse IgG and the AuroProbe BL Plus kit as described for the Western blotting. AP activity, caveolin and TFPI concentrations, respectively, are approximately proportional to the color intensity of the dots on the membrane.
Measurement of TFPI Antigen and Anticoagulant Activity
TFPI antigen and activity were determined in EC supernatants, cellular lysates, fractions from the Nycodenz gradients, and on the cell surface, essentially as previously described:15 The antigen was measured by both direct ELISA (on cell monolayers) and indirect competitive ELISA (supernatants); the activity was determined with a two-stage chromogenic assay, in which the residual capacity of TF-FVIIa to activate FX after incubation with TFPI-containing samples and small amounts of FXa was estimated. Protein estimation in the samples was done by the Bradford assay.33
Effect of Heparitinase and Sodium Chlorate on TFPI on the EC Surface
Treatment of both HUVEC and EA.hy926 cells neither induced enhancement of TFPI antigen or activity in the medium (Fig 1a⇓, HEP) nor decreased anticoagulant activity on the cell surface (Fig 1b⇓, HEP and NaClO3). Immunofluorescence staining on heparitinase-treated cells (not shown) confirmed the quantitative data, indicating that cellular TFPI is not exposed on the EC surface through interactions with glycosaminoglycans only.
TFPI Is Anchored in the Membrane Through a GPI Moiety
Effect of PI-PLC
The enzyme cleaves the hydrophobic part of the glycolipid anchor, resulting in the release of the otherwise intact protein into the medium.27 When EC were treated with PI-PLC, ≈40% increase of TFPI antigen and activity was detected in the supernatants of both HUVEC and EA.hy926 cells (Fig 1a⇑, PI-PLC), and the anticoagulant activity on the cell surface decreased equivalently (Fig 1b⇑, PI-PLC). Immunofluorescence microscopy revealed considerable reduced immunostaining for TFPI on the cell surface (Fig 2⇓c; compare to Fig 2a⇓). The Western blot in Fig 3⇓a shows the presence of two additional bands reactive with the anti-TFPI IgG in the cell medium after incubation of HUVEC with PI-PLC (lane B), as compared to constitutively secreted TFPI (lane A).
Incorporation of [1-3H]Ethanolamine
The fluorographies in Fig 3b⇑ show that TFPI in EA.hy926 cells incorporates this radioactive main precursor of the GPI-anchor (lane A, whole cell lysate). TFPI that was constitutively secreted from radiolabeled cells exhibited a very weak radioactive signal, at Mr of ≈30 kDa (lane C). Significantly more radioactive protein was detected in the cell medium after PI-PLC treatment (lane D), with strong signals at Mr of ∼45 kDa and 60 to 70 kDa, comparable with the ones observed for whole cell lysates (lane A).
Effect of Mannosamine
TFPI in the cell medium increased 40% to 70% after treatment with mannosamine (Fig 1a⇑, MAN), and the activity of TFPI on the cell surface was equivalently diminished, for both HUVEC and EA.hy926 cells (Fig 1b⇑, MAN). The quantitative data are supported by immunofluorescence, which showed decreased intensity of staining for TFPI on the HUVEC surface after mannosamine treatment (Fig 2d⇑). Intracellularly, accumulation of the protein in tubuloreticular structures confined to perinuclear areas was frequently observed (Fig 2e⇑, compare to resting cells in Fig 2b⇑). The fluorography showed that mannosamine also inhibited the incorporation of [1-3H]ethanolamine into cellular TFPI, resulting in weaker radioactive signals, for all the protein bands (Fig 3b⇑, lane B).
Colocalization of TFPI With Caveolin and uPAR
TFPI does not colocalize with von Willebrand factor in Weibel-Palade bodies in EC in culture.15 We investigated by double immunostaining whether TFPI associates with other proteins resident of different granules/vesicles: tPA, caveolin, or GPI-anchored proteins, such as uPAR. TFPI and tPA showed a completely distinct distribution by immunofluorescence32 (not shown). By contrast, TFPI partially colocalized with uPAR over the HUVEC surface (Fig 4a⇓), and treatment with PI-PLC, although seemingly more efficient for uPAR, decreased the intensity of staining for both antigens (Fig 4b⇓).
We found by preembedding and postembedding immunoelectron microscopy techniques carried out on prefixed EC and mouse lung endothelium that gold-immunolabeled TFPI appeared as clusters located both on the plasmalemma proper (Fig 5⇓, a-mouse lung endothelium, e-EA.hy926 cells, and f-HUVEC; arrows), as well as within or nearby small, noncoated invaginations of the plasma membrane (Fig 5b⇓, EA.hy926 cells, Fig 5d⇓, HUVEC; black arrowheads). Morphologically, the invaginations/vesicles were recognized as caveolae by their characteristic 60- to 80-nm size and flask shape. Coated pits were consistently devoid of staining (Fig 5f⇓, cp), while vesicles of dimension similar to caveolae within the apical cytoplasm exhibited labeling for TFPI (Figs 5a⇓ and 5d⇓). Frequently, these vesicles/invaginations bearing TFPI were positively identified as caveolae by double immunogold labeling for TFPI and caveolin (Fig 5c⇓, tangential section at the plasmalemma level, Fig 5d⇓, vertical section; white arrowheads). Both types of control experiments performed were negative.
By immunofluorescence and confocal microscopy examination, a strong superposition of TFPI-clusters and caveolin-containing vesicles was revealed underneath the plasma membrane (Fig 4c⇑). At Golgi level, both colocalization of the two antigens, best seen on the cell periphery, and distinct immunostaining for TFPI and caveolin could be observed (Fig 4d⇑). When live cells were incubated with anti-TFPI IgG at 6 to 8°C to optimize labeling of caveolae24 and then were fixed, permeabilized, and stained for caveolin, the TFPI-specific fluorescence appeared in smaller but more uniformly distributed dots, which significantly colocalized with caveolin (Fig 4e⇑). When live cells were cold-incubated with anti-TFPI IgG, then warmed at 37°C to allow for internalization of the bound antibody and then permeabilized and immunostained for caveolin, superposition between TFPI and caveolin was observed intracellularly for almost every cluster of TFPI (Fig 4f⇑). This suggests that the anti-TFPI IgG bound to TFPI during the cold-incubation was partially internalized via caveolae while cells warmed up.
Enrichment of TFPI Within Triton-Insoluble Complexes
After sequential extractions of EC with 1% (w/v) Triton X-100 and 60 mmol/L OGP, more than half of the cellular TFPI was associated with the OGP fraction, which contained approximately 8% of the total cellular protein. TFPI content normalized for total protein was ≈13 times higher in the OGP fraction than in the Triton-soluble one and diminished by ≈35% after treatment with PI-PLC (Fig 6a⇓). This figure correlates well with the decrease of TFPI antigen and activity on the EC surface, as determined by ELISA and functional assays (histogram in Fig 1⇑). The Western blotting performed on cellular lysates after TCA-precipitation revealed an enrichment of the major bands at Mr of ≈28, 35, and 66 kDa in the OGP fraction (Fig 6b⇓, EA.hy926 cells). Likewise, other molecular forms of TFPI (bands at Mr of ≈20 to 25, 46, and 75 kDa) were exclusively found in the OGP fraction.
The Triton-insolubility of TFPI is further supported by immunofluorescence. The specific staining for TFPI was present even after treatment of EA.hy926 cells with Triton X-100 (Fig 6c⇑), and a high degree of colocalization with caveolin was also observed in these conditions (not shown). Images taken at high magnification revealed a very discrete punctuate distribution of TFPI (Fig 6c⇑, inset).
Distribution of TFPI in Nycodenz Density Gradient Fractions of HUVEC and Rat Lung
Figure 7⇓ presents results of the fractionation through density gradients of Nycodenz of HUVEC and rat lung homogenates obtained in the absence of detergents. The distribution of TFPI antigen determined by ELISA (Fig 7a⇓) shows essentially similar profiles for HUVEC and rat lung, with two peaks within the range of densities of 1.05 to 1.08 and 1.105 to 1.11 g/mL. More than half of the total TFPI in the homogenates was recovered in fractions 10 to 13, while most of the caveolin in HUVEC and AP activity in rat lung were found within fractions 10 to 12, which also stained positively when the monoclonal anti-K2TFPI IgG was used in the dot-blot assay (Fig 7b⇓).
Control of the Distribution of TFPI by the Lipid Organization of the Membrane
Cholesterol Controls the Clustering of TFPI
The role of cholesterol in the organization of TFPI clusters was studied by treatment of EC with filipin or nystatin. As observed by immunofluorescence, filipin induced marked patching and fusion of TFPI clusters over the EA.hy926 cell surface into large areas of high fluorescence intensity (Fig 8a⇓, magenta-colored areas, arrows).
On nystatin-treated HUVEC, the distribution of caveolin at the level of apical plasmalemma changed from the uniform clusters observed in resting cells to a very diffuse fluorescence, while TFPI appeared as large patches (Fig 8b⇑).
Quantitative estimation of TFPI after incubation of EC with filipin showed no enhanced release of TFPI into the supernatants or modification of the anticoagulant activity on the EC surface. Nevertheless, our ELISA on EC monolayers treated with filipin showed significant redistribution of cellular TFPI, with ≈40% increase of the surface-associated antigen, while the total cellular TFPI remained unchanged.
TFPI Locates Within GSL Domains
The relationship between TFPI and glycolipid microdomains was investigated by double staining of EC with anti-TFPI IgG and CT-B conjugated with FITC. CT-B specifically binds to the monosialoganglioside GM1, which was shown to concentrate in caveolae.34 Here we illustrate serial optical sections through EA.hy926 cells, from basal to apical levels (Figs 8c⇑ through 8e) and show that clusters of TFPI colocalize with GM1 mainly in perinuclear areas presumably representing the Golgi network, in the most apical cytoplasm, and at the plasmalemma level.
Agonist-Induced Release and Redistribution of TFPI
We extended previous investigations15 by testing whether human EC in culture respond to other stimuli similarly to thrombin. Treatment of EC with A23187 for 10 minutes at 37°C induced a ≈40% increase of TFPI release into the medium and similar enhancement of anticoagulant activity associated with the cell surface. Immunostaining for TFPI on EA.hy926 cells after stimulation with calcium ionophore also revealed an enhanced intensity of the fluorescence over the cell surface, presumably achieved by fusion of the clusters (Fig 9a⇓, compare to Fig 2a⇑), together with marked intracellular redistribution of the antigen, with patching toward the cell periphery (Fig 9b⇓; compare to Fig 2b⇑). On thrombin-treated cells double stained for TFPI and GM1, the largest patches of TFPI on the cell surface were confined to GM1-rich domains (Fig 9c⇓, arrows). Double immunolabeling for TFPI and caveolin in HUVEC also showed patching of TFPI toward the cell periphery and unclustering of caveolin-positive vesicles into a diffuse fluorescence (Fig 9d⇓; compare to Fig 4d⇑ and Fig 8b⇑).
This study analyzes the mechanism(s) by which TFPI in resting human EC in culture is exposed/anchored in the plasma membrane. Since removal of heparan-sulfate from the cell surface did not affect the distribution of cellular TFPI and it is known that TFPI in the fluid phase inhibits TF•FVIIa relatively slowly when used at concentrations equivalent to its plasma level,35 we reasoned that TFPI should be more specifically anchored in the plasma membrane to meet the requirements for the effective action of the inhibitor as endothelial cell-bound anticoagulant.
The granular distribution that we first described for TFPI15 highly resembles the clustered appearance of GPI-anchored proteins, such as alkaline phosphatase,36 folate receptor,22 or uPAR.26 The lipid content of the membrane is responsible for the aggregated state of certain GPI-anchored proteins within microdomains,36 as well as for their insolubility in Triton X-100.25
On the basis of biochemical and morphological evidence, we propose that cellular TFPI locates within similar microdomains in EC, probably via a GPI-linkage. We found that more than half of the cellular TFPI partitioned into Triton-insoluble extracts, which is in accordance with previously reported data.12 We reinforce this observation by immunofluorescence showing that TFPI stays with the EC monolayers even after Triton-extraction, which suggests that TFPI is intrinsically bound to the plasma membrane. TFPI detected by Western blotting was heterogeneous, but this is not surprising, since multiple forms for TFPI in plasma,37 endothelial cells,12 or conditioned media from hepatoma cell lines38 have been described and explained by posttranslational modifications and/or different degrees of proteolytic carboxy-terminal truncation.38,39 Even so, we observed a significant enrichment of TFPI, together with the presence of additional specific bands, in the OGP lysates. Using radiolabeling and fluorography, we found that [1-3H]-ethanolamine, a specific precursor of the glycolipid anchor, incorporated into TFPI, especially into the bands at molecular weights of ≈45 and 66 to 70 kDa, which are also enriched in the OGP fraction, and mannosamine inhibited this incorporation. PI-PLC cleaved approximately half of the TFPI from native EC surface, and by detergent extraction of PI-PLC-treated cells, an equivalent decrease of TFPI was found in the Triton-insoluble material. One could expect a higher proportion of TFPI to be cleaved out by PI-PLC, but this is not imperative, since a great variability in the sensitivity of GPI-anchored proteins toward the enzyme’s action has been described, depending to a large extent on the cell type and the protein studied.20 This also can explain the slightly different decrease of the intensity of staining observed for TFPI and uPAR after treatment of EC with PI-PLC. The enzymatically cleaved TFPI was found after both Western blotting and fluorography at Mr of approximately 46 kDa and 70 kDa, which coincides with some of the molecular species that are exclusively present in the OGP fractions (compare Fig 3a⇑, lane B, and Fig 3b⇑, lane D, with Fig 6b⇑, lane OGP). Finally, as shown by immunofluorescence, mannosamine affected the normal exposure of TFPI on the cell surface, inhibited the incorporation of [1-3H]ethanolamine, and induced the accumulation of TFPI into the secretory pathway with consecutive enhanced release into the medium. All of these modifications are in accordance with the effects that mannosamine has been proved to exert on GPI-anchored proteins.21 Although the cDNA-deduced sequence of the COOH terminus of TFPI is unusually basic for processing of the protein to a GPI-form,30 we consider that, taken together, our data show that TFPI bears the hallmarks of a GPI-anchored protein.
It is well established that a variety of cell types contain clustered arrays of GPI-anchored proteins and that a subset of these is in caveolae.40 Caveolae, or plasmalemmal vesicles, are small invaginations of the plasma membrane with a flask shape and the marker protein caveolin as structural component.40,41 The core of caveolae is composed of sphingolipids (such as GM1)34 and cholesterol,42 and their normal morphology is maintained by rings of sterols surrounding the openings of caveolae.43
We present here the first ultrastructural evidence for the constitutive localization of TFPI in caveolae in resting endothelium. We examined a broad range of specimens (HUVEC, EA.hy926 cells, and mouse lung endothelium) and frequently found clusters of immunogold-labeled TFPI located within or at the neck of membrane invaginations/vesicles that proved to be caveolae by double staining for caveolin. Since there is still controversy concerning the native location of some GPI-anchored proteins in caveolae,23,44 our postembedding immunogold staining approach, using fixed cells embedded at low temperature in Lowicryl, entirely rules out any possible antibody-induced redistribution and nonspecific clustering of TFPI in caveolae. No labeling for TFPI was observed in clathrin-coated pits, even when live cells were immunostained for TFPI at 4°C and then warmed at 37°C (results not shown). Therefore, like Sevinsky et al,12 we exclude a potential binding of TFPI in EC to α2-macroglobulin receptor/low-density lipoprotein receptor–related protein, which was shown to mediate the cellular degradation of TFPI in hepatoma cell lines.45 Since caveolae constitute plasmalemma microdomains that are devoid of anionic binding sites,46 particularly heparan sulfate and/or heparin,47 our results also indicate that TFPI does not remain merely anchored to sulfated glycosaminoglycans after secretion, as was inferred until now.
The study of TFPI distribution was further extended by immunofluorescence and confocal microscopy. TFPI and caveolin exhibited a high degree of colocalization in clusters on fixed HUVEC, mainly at the plasmalemma level and the trans-Golgi network. On live cells immunostained for TFPI at 6°C to 8°C, its distribution was firmly punctuated and superimposed over caveolin. Although after warming of the cells at 37°C to allow for endocytosis, TFPI appeared in larger patches, probably owing to antibody-induced redistribution, colocalization with caveolin was still apparent for every cluster of TFPI. Accordingly, the internalization of the anti-TFPI IgG bound to TFPI partially occurred via caveolae, a process previously described both for another GPI-anchored protein, alkaline phosphatase,24 and for a transmembrane protein, TF, detected in caveolae in SMC.48
In the attempt to further substantiate the location of TFPI in caveolae, we analyzed the distribution of TFPI, caveolin, and AP in the Nycodenz gradient fractions obtained by centrifugation of HUVEC and rat lung detergent-free homogenates. To discuss only two of the contradictory reports regarding the clustering of GPI-anchored proteins in caveolae, Schnitzer et al stated that normally, and even more after cross-linking with antibodies, GPI-linked proteins partition into diffusion-restrictive domains, some of which form annular regions at the opening of caveolae and coisolate with caveolae during detergent solubilization.49 On the other hand, Smart et al50 found GPI-linked proteins associated with caveolae even when the latter were isolated in the absence of detergent. The same authors showed that caveolae, because of their specific lipidic content, have a unique buoyant density that causes them to separate from other membrane domains during detergent-free fractionation and thus proved that the specific partitioning of GPI-linked proteins is not simply an artifact of detergent solubilization.50 Also working with detergent-free fractions, we found most of TFPI, caveolin, and AP at the density of 1.06 to 1.08 g/mL, with a smaller peak at 1.105 to 1.11 g/mL. The equilibration of all three proteins at low densities suggests that they are confined to the same type of vesicle that behaves this way on the gradient because of a high lipid-to-protein ratio (caveolae?). The higher density peak may have resulted from interactions of caveolae with endosomal or Golgi domains24 or represents immature vesicles within the biosynthetic pathway, before aquiring the final specific proteolipidic composition on exit from the trans-Golgi network.51
The alteration of the membrane lipid organization by cholesterol-complexant drugs (filipin, nystatin) induced unclustering of TFPI and formation of large patches on the cell surface. The steady-state enrichment of caveolin in caveolae is sensitive to cholesterol oxidation,52 and filipin treatment or metabolic depletion of cholesterol disrupts the structure of caveolae, probably by affecting the membrane fluidity or reducing the ability of caveolae to seal themselves off from the extracellular space.22,53 Since this was shown to produce unclustering of GPI-anchored proteins with formation of patches and enhanced exposure on the cell surface,53 we believe that filipin/nystatin brought about patching of TFPI by inducing flattening of caveolae with consequent enhanced exposure of the inhibitor on the cell surface.
In simple epithelia, the apical membrane polarizes high concentrations of glycolipids and GPI-anchored proteins54 as a combined result of sorting in the Golgi apparatus and vectorial apical delivery. We observed that TFPI associates in clusters with GM1, from the Golgi level throughout the apical cytoplasm, and on the plasmalemma surface. Since a key event in the apical sorting is the formation of GSL/GPI-anchored protein microdomains in the trans-Golgi network,51 we suggest that clustering plays a significant role in the proper targeting of TFPI to the apical cell surface and its exposure in specific microdomains, partially located in caveolae.
Caveolae mediate many different functions: endocytosis and transcytosis,55,56 potocytosis,57 regulation of cell surface-associated proteolysis,12,26 and calcium regulation and signaling.58 Caveolae are enriched in signaling molecules, thus being able to compartmentalize the signal transduction machinery and provide spatial and temporal information as cells respond to stimuli.57 On thrombin-stimulated EC, the fusion of individual clusters of TFPI is mediated through patching within GM1-rich microdomains. In view of the recently reported detection of the functional thrombin receptor in caveolae,59 we suggest that the signal(s) generated by thrombin’s binding to its receptor on EC is transduced in caveolae, leading to unclustering and patching of TFPI. Whether this process is mediated by thrombin-induced biphasic [Ca2+]i transients mediated by IP3 receptor(s) and involving actin cytoskeleton reorganization58 or through tyrosine kinases60 or protein kinase C,61 is an issue under active investigation in our group. A preliminary indication that [Ca2+]i transients in caveolae might be involved arises from the described effects of calcium ionophore stimulation on TFPI redistribution in EC. Whatever the mechanism(s), results are consistent with the localization of TFPI in caveolae/glycolipid domains, which will supply the inhibitor with the mobility required for lateral migration in the membrane and with the most proper environment for exerting its inhibitory activity toward TF•FVIIa.12 It remains to be answered whether the modifications observed for TFPI after stimulation of EC represent a process similar to the reported enlargement of caveolae bearing TF in SMC after detachment of the cells,48 which was suggested to make available a latent pool of procoagulant activity at sites of vascular injury.48 Conversely, in EC in which TF is inducible only by stimulation,4–6 the enhanced exposure of TFPI on the cell surface might be meant to counterbalance the newly expressed TF, a process that would further extend the role already assigned to caveolae in the maintenance and upregulation of the anticoagulant properties of endothelium in vivo.
Selected Abbreviations and Acronyms
|BSA||=||bovine serum albumin|
|CT-B||=||cholera toxin–binding subunit|
|HUVEC||=||human umbilical vein endothelial cells|
|K2TFPI||=||Kunitz2 domain of TFPI|
|PI-PLC||=||phosphatidylinositol phospholipase C|
|SMC||=||smooth muscle cells|
|TFPI||=||tissue factor pathway inhibitor|
|tPA||=||tissue-type plasminogen activator|
|TEA buffer||=||triethanolamine buffer|
|uPAR||=||urokinase-type plasminogen activator receptor|
This work was supported by stipends from the British Heart Foundation (grant PG/95027), Stanley Thomas Johnson Foundation, and the Thrombosis Research Trust. We highly acknowledge the skillful help of Ulla Dennehy and Sally Mill (cell culture), and we thank Dr A. Esmail in our institute for helping us with the animal work. We also thank Professor Fedor Bachmann for the critical reading of the manuscript.
Stern D, Handley D, Nawroth PP. Endothelium and the regulation of coagulation. In Simionescu N, Simionescu M, eds. Endothelial Cell Biology in Health and Disease. NY: Plenum Press; 1988:275–306.
Bevilacqua MP, Pober JS, Majeau GR, Cotran RS, Gimbrone MA Jr. Interleukin 1(IL-1) induces biosynthesis and cell surface expression of procoagulant activity in human vascular endothelial cells. J Exp Med. 1984;160:618–623.
Nawroth PP, Stern DM. Modulation of endothelial cell hemostatic properties by tumor necrosis factor. J Exp Med. 1986;163:740–745.
Broze GJ Jr., Warren LA, Novotny WF, Higuchi DA, Girard JJ, Miletich JP. The lipoprotein-associated coagulation inhibitor that inhibits the factor VII-tissue factor complex also inhibits factor Xa: insight into its possible mechanism of action. Blood. 1988;71:335–343.
Huang ZF, Wun TC, Broze GJ, Jr. Kinetics of factor Xa inhibition by tissue factor pathway inhibitor. J Biol Chem. 1993;268:26950–26955.
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.
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.
Bajaj MS, Kuppuswamy MN, Saito H, Spitzer SG, Bajaj SP. Cultured normal human hepatocytes do not synthesize lipoprotein-associated coagulaton inhibitor: evidence that endothelium is the principal site of its synthesis. Proc Natl Acad Sci U. S. A. 1990;87:8869–8873.
Abildgaard U. Heparin/low molecular weight heparin and tissue factor pathway inhibitor. Haemostasis. 1993;23(suppl. 1):103–106.
Lupu C, Lupu F, Dennehy U, Kakkar VV, Scully MF. Thrombin induces the redistribution and acute release of tissue factor pathway inhibitor from specific granules within human endothelial cells in culture. Arterioscler Thromb Vasc Biol. 1995;15:2055–2062.
Edgwell CJS, McDonald CG, Graham JB. Permanent cell line expressing human factor VIII-related antigen established by hybridization. Proc Natl Acad Sci U. S. A.. 1983;80:3734–3737.
Jaffe EA, Nachman RL, Becker CG, Minick CR. Culture of human endothelial cells derived from human umbilical veins: identification by morphologic and immunologic criteria. J Clin Invest. 1973;52:2745–2756.
Brunner G, Gabrilove J, Rifkin DB, Wilson EL. Phospholipase C release of basic fibroblast growth factor from human bone marrow cultures as a biologically active complex with a phosphatidylinositol-anchored heparan sulfate proteoglycan. J Cell Biol. 1991;114:1275–1283.
Rothberg KG, Ying Y-S, Kamen BA, Anderson RGW. Cholesterol controls the clustering of glycophospholipid-anchored membrane receptor for 5-methyltetrahydrofolate. J Cell Biol. 1990;111:2931–2938.
Mayor S, Maxfield FR. Insolubility and redistribution of GPI-anchored proteins at the cell surface after detergent treatment. Mol Biol Cell. 1995;6:929–944.
Parton RG, Joggerst B, Simons K. Regulated internalization of caveolae. J Cell Biol. 1994;127:1199–1215.
Stahl A, Mueller BM. The urokinase-type plasminogen activator receptor, a GPI-linked protein, is localized in caveolae. J Cell Biol. 1995;129:335–344.
Liu P, Anderson RGW. Compartmentalized production of ceramide at the cell surface. J Biol Chem.. 1995;270:27179–27185.
Lupu F, Bergonzelli GE, Heim DA, Cousin E, Genton CY, Bachmann F, Kruithof EKO. Localization and production of plasminogen activator inhibitor-1 in human healthy and atherosclerotic arteries. Arterioscler Thromb. 1993;13:1090–1100.
Emeis JJ, van den Eijnden-Schrauwen Y, van den Hoogen CM, de Priester W, Westmuckett AD, Lupu F. An endothelial storage granule for tissue-type plasminogen activator. J Cell Biol.. 1997;139:245–256.
Parton RG. Ultrastructural localization of ganglioside:; GM1 is concentrated in caveolae. J Histochem Cytochem. 1994;42:155–166.
Rao LV, Nordfang O, Hoang AD, Pendurthi UR. 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.
Cerneus DP, Ueffing E, Posthuma G, Strous GJ, van der Ende A. Detergent insolubility of alkaline phosphatase during biosynthetic transport and endocytosis. Role of cholesterol. J Biol Chem. 1993;268:3150–3155.
van’t Veer C, Hackeng TM, Delahaye C, Sixma JJ, Bouma BN. Activated factor X and thrombin formation triggered by tissue factor on endothelial cell matrix in a flow model: effect of tissue factor pathway inhibitor. Blood. 1994;84:1132–1142.
Wun T-C, Huang MD, Kretzmer KK, Palmier MO, Day KC, Bulock JW, Fok KF, Broze GJ Jr. Immunoaffinity purification and characterization of lipoprotein-associated coagulation inhibitors from Hep G2 hepatoma, Chang Liver, and SK hepatoma cells. J Biol Chem. 1990;265:16096–16101.
Chang WJ, Ying YS, Rothberg KG, Hooper NM, Turner AJ, Gambliel HA, De Gunzburg J, Mumby SM, Gilman AG, Anderson RG. Purification and characterization of smooth muscle cell caveolae. J Cell Biol. 1994;126:127–138.
Parton RG, Simons K. Digging into caveolae [comment]. Science. 1995;269:1398–1399.
Simionescu N, Lupu F, Simionescu M. Rings of membrane sterols surround the openings of vesicles and fenestrae, in capillary endothelium. J Cell Biol. 1983;97:1592–1600.
Mayor S, Rothberg KG, Maxfield FR. Sequestration of GPI-anchored proteins in caveolae triggered by cross-linking. Science. 1994;264:1948–1951.
Warshawsky I, Broze GJ Jr., 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.
Simionescu N, Simionescu M, Palade GE. Differentiated microdomains on the luminal surface of the capillary endothelium: I. Preferential distribution of anionic sites. J Cell Biol. 1981;90:605–613.
Simionescu M, Simionescu N, Silbert JE, Palade GE. Differentiated microdomains on the luminal surface of the capillary endothelium: II. Partial characterization of their anionic sites. J Cell Biol. 1981;90:614–621.
Mulder AB, Smit JW, Bom VJJ, Blom NR, Ruiters MHJ, Ruud Halie M, van der Meer J. Association of smooth muscle cell tissue factor with caveolae. Blood. 1996;88:1306–1313.
Schnitzer JE, McIntosh DP, Dvorak AM, Liu J, Oh P. Separation of caveolae from associated microdomains of GPI-anchored proteins [see comments]. Science. 1995;269:1435–1439.
Smart EJ, Ying YS, Mineo C, Anderson RG. A detergent-free method for purifying caveolae membrane from tissue culture cells. Proc Natl Acad Sci U. S. A. 1995;92:10104–10108.
Conrad PA, Smart EJ, Ying YS, Anderson RG, Bloom GS. Caveolin cycles between plasma membrane caveolae and the Golgi complex by microtubule-dependent and microtubule-independent steps. J Cell Biol. 1995;131:1421–1433.
Chang WJ, Rothberg KG, Kamen BA, Anderson RGW. Lowering, the cholesterol content of MA104 cells inhibits receptor-mediated transport of folate. J Cell Biol. 1992;118:63–69.
Ghitescu L, Fixmamn A, Simionescu M, Simionescu N. Specific binding sites for albumin restricted to plasmalemmal vesicles of continuous capillary endothelium: receptor-mediated transcytosis. J Cell Biol. 1986;102:1304–1311.
Anderson RG. Caveolae: where incoming and outgoing messengers meet. [Review]. Proc Natl Acad Sci U. S. A. 1993;90:10909–10913.
Fujimoto T, Miyawaki A, Mikoshiba K. Inositol 1,4,5-trisphosphate receptor-like protein in plasmalemmal caveolae is linked to actin filaments. J Cell Sci. 1995;108:7–15.
Stan R-V, Roberts WG, Predescu D, Ihida K, Saucan L, Ghitescu L, Palade GE. Immunoisolation and partial characterization of endothelial plasmalemmal vesicles (caveolae). Mol Biol Cell. 1997;8:595–605.
Smart EJ, Foster DC, Ying Y-S, Kamen BA, Anderson RGW. Protein kinase C activators inhibit receptor-mediated potocytosis by preventing internalization of caveolae. J Cell Biol. 1994;124:307–313.