Tissue factor (TF) is the most important trigger of blood coagulation in vascular pathology. Rabbit TF, with or without (ΔC) its COOH-terminal intracellular tail, has been conjugated to green fluorescent protein (GFP) to study subcellular localization and other functions of TF. TF-GFP and TFΔC-GFP are associated with Na2CO3-resistant buoyant fractions in HEK-293 cells (lipid rafts); there is no morphological difference in the surface distribution of these or other GFP-labeled membrane proteins present in or excluded from rafts (confocal microscopy, HEK-293 cells). Endogenous TF expressed by rabbit aortic smooth muscle cells (SMCs) is also raft associated. Membranes from HEK-293 cells expressing recombinant TF-GFP or wild-type TF were equipotent to clot human plasma; however, TFΔC-GFP was ∼20-fold more active (per membrane weight). Immunoblot confirmed that the deletion mutant is more abundantly expressed, and confocal microscopy showed that it has preferential membrane localization, whereas TF-GFP is mainly intracellular (nuclear lining and multiple granules). With a similar half-life (<4 h), the two constructions differ by their intracellular retention, lower for TFΔC-GFP. In serum-starved SMCs, the expression of endogenous TF was upregulated by interleukin-1β and/or FBS treatment (immunoblot, immunofluorescence, clotting assay). However, TF secretion or surface expression was not regulated by stimuli of physiological intensity (such as stimulation of the coexpressed kinin B1 receptors), although a calcium ionophore was highly active in this respect. TF is a raft-associated molecule whose surface expression (secretion) is apparently retarded or impaired by structural determinant(s) located in its COOH-terminal tail.
- smooth muscle cells
- kinin B1 receptors
- lipid rafts
tissue factor (TF, CD142) is a ∼46-kDa glycoprotein related to type II cytokine receptors that possesses one transmembrane domain and a small intracellular COOH-terminal tail. The extracellular domain of TF is the binding site for coagulation factor VII and VIIa. TF is currently considered to be the single most relevant physiological trigger of blood coagulation (46). TF is also a key regulator of embryonic angiogenesis, oncogenic neoangiogenesis, inflammatory response, and perhaps cell migration (24). TF is constitutively expressed in the brain, cardiomyocytes, organ capsules, renal glomeruli, blood vessel adventitia, etc., as a defense mechanism against hemorrhage but is synthesized de novo in response to pathological situations in vascular cells [endothelial cells, smooth muscle cells (SMCs)], monocytes, renal tubular epithelium, and malignancies (17, 45, 46). Of relevance to vascular biology, balloon injury applied to the rabbit aorta in vivo is followed by de novo TF expression in SMCs of the media and neointima during several weeks (22). Direct vascular injury in mice induces thromboses that are dependent on TF derived from the vascular wall, not from blood cells (shown notably using strains of animals expressing low levels of TF engrafted or not with bone marrow from mice with normal TF expression) (11). TF is both expressed at the cell surface and released into the supernatant of human SMCs (18).
TF regulation at the protein level is not fully understood. The molecule possesses a short COOH-terminal intracellular tail that has been the focus of much attention, with potential phosphorylation and palmytoylation sites (Fig. 1) (15, 50). Homozygous mice expressing a TF with the cytoplasmic domain deletion do not exhibit alterations in blood coagulation but rather an inhibition of immunoinflammatory responses (49) and increased tumoral angiogenesis (5). Whether TF is associated to cholesterol lipid rafts is controversial (13, 30). A fraction of the clotting activity in TF-expressing cells is latent or “encrypted.” A calcium ionophore is frequently used to increase the cell surface procoagulant activity in such systems (e.g., Refs. 9 and 29). However, this response may include the surface expression of anionic phospholipids, such as phophatidyl serine, which are obligatory cofactors of TF for the efficient binding of factor VII (13, 46). Whether the subcellular localization of TF is changed during “deencryption” is not clear (16). The characterization of subcellular coagulant microparticles containing TF is also an active research front in pathology that has often been modeled in cultured cells using calcium ionophores (3, 6, 18).
In cell types where TF is inducible, the stimuli and mechanisms of TF expression are strikingly similar to those of the kinin B1 receptor, a G protein-coupled receptor expressed under the effect of cytokines, some MAPK pathways, and nuclear factor (NF)-κB (32, 44). The B1 receptor is also reportedly expressed in atheromas, based on immunohistochemistry (42), and thus colocalized with TF (46). TF and the B1 receptor may be essentially coexpressed in parallel in the vasculature after a sublethal injection of lipopolysaccharide to rabbits (17, 44). Because B1 receptor stimulation elicits a Ca2+ signaling response exceptionally resistant to desensitization in rabbit SMCs (33), Ca2+-mediated deencryption is at least one manner in which both molecules can interact.
We produced a green fluorescent protein (GFP) conjugate of rabbit TF (TF-GFP) along with a similar construction with the deletion of the cytoplasmic domain (TFΔC-GFP) to examine several issues: the subcellular distribution of these molecules in live cell imaging, the partition of these constructions to cholesterol-rich lipid rafts, and their possible mobilization under several types of cell stimulation and release as subcellular particles. Special attention has been given to the fate of endogenous and recombinant TF in rabbit aortic SMCs.
MATERIALS AND METHODS
Construction of TF-GFP.
The local ethics committee approved the procedures based on rabbits. RNA was extracted from a freshly removed rabbit brain using TRIzol reagent (GIBCO-BRL), and the extract was further treated with DNAse I. First-strand cDNA was synthesized using SuperScript II RNase H− reverse transcriptase (GIBCO-BRL). The entire coding region of the rabbit TF (2) (excluding the stop codon) was then amplified by PCR. The PCR sense and antisense primers used were 5′-AAATAAGCTTAATGGCGCCCCCGACCCGGCTCC-3′ and 5′-TATTGGATCCGGCGATGTTCAGGGGGGAGCTCTCC-3′, respectively. These primers contained additional HindIII and BamHI sites, respectively (boldface), for the directional cloning of the rabbit TF coding region in the eukaryotic expression vector pEGFP-N3 (Clontech Laboratories; Palo Alto, CA) encoding GFP. Both the PCR fragment and the pEGFP-N3 vector were digested with HindIII and BamHI and ligated at room temperature for 2 h. The resultant vector (pTF-GFP; Fig. 1) contained the rabbit TF coding sequence fused in frame at its carboxyl terminus with GFP, expressed under the control of the cytomegalovirus promoter. With the use of pTF-GFP as a template, PCR was performed with an alternate 38-base antisense primer containing the stop codon antisense sequence (TCA) of TF just 3′ of the BamHI site; this produced the wild-type (WT) coding sequence of rabbit TF under the control of the same promoter (Fig. 1). Finally, a fluorescent construction with the deletion of much of the intracellular COOH-terminal tail of rabbit TF (TFΔC-GFP; Fig. 1) was obtained using the same PCR sense primer described above and the antisense primer 5′-TATTGGATCCCTTTCTGCACTTGTACACGGTCACAG-3′. The sequences of the constructions were verified (RSVS Core Laboratory, Pavillon Marchand, Laval University, Quebec City, Quebec, Canada).
Cell culture and transfection.
The rabbit bradykinin (BK) B2 receptor fused to GFP (B2R-GFP) is a construction coding for an alternate membrane protein also based on pEGFP-N3; its production and properties are reported elsewhere (4, 23). The B2 receptor is reportedly enriched in lipid rafts (27). This vector and the GFP-coding vector were used as controls in some experiments. The GFP-GPI coding vector (in pCDNA3) was a gift from Dr. A. Le Bivic, Faculté des Sciences de Luminy, Marseilles, France. It possesses the GPI addressing sequence for lipid rafts (31). Fyn-GFP and GFP-KRas are lipid-modified proteins that are, respectively, enriched in and excluded from rafts (25); the corresponding expression vectors were gifts from Dr. M. Philips, New York University School of Medicine, New York, NY.
Rabbit aortic SMCs, HEK-293, and COS-1 cells were cultured as previously described (23, 43, 44). The vectors were transfected in any of the three cell types using the X-Gen 500 transfection reagent (MBI Fermentas; Flamborough, Ontario, Canada) as directed and cultured in serum-containing medium until use. TF-coding vectors were transfected as 4 μg DNA/8 cm2 cell culture surface area (or using the same proportions in other culture surface areas).
The subcellular fluorescence distribution, mostly in live cells and without fixation, was observed using a Bio-Rad 1024 laser beam confocal microscope (×40 or 60 objective with oil immersion; GFP: emission 488 nm, detection above 510 nm; rhodamine-labeled phalloidin: emission 568 nm, detection above 585 nm). Saponin permeabilization of SMCs was applied for the double detection of TF-GFP coupled to staining with rhodamine-labeled phalloidin (Molecular Probes).
Fixed but nonpermeabilized SMCs or COS-1 cells were used to detect surface WT or recombinant rabbit TF in immunofluorescence using the monoclonal 11F (17) (American Diagnostica; Montreal, Quebec, Canada; 20 μg/ml; general methods as in Ref. 34).
Cell fractionation and immunoblots.
Immunoblots were performed as previously reported for GFP-labeled proteins (4, 44). Briefly, transfected HEK-293 or SMCs grown in 75-cm2 flasks (∼80% confluence) were used. The material analyzed was the total cell extract, the plasma membrane fraction (third pellet, 150,000 g × 3 h, both normalized as protein content) (23), the sedimented cell culture medium (150,000 g × 3 h, normalized by cell surface area) for the evaluation of particle release, or the buoyant plasma membrane fractions containing cholesterol-rich rafts (prepared as described using Optiprep density gradients of Na2CO3-treated cells) (43). Samples (30 μg of membrane or total cell extract proteins) were analyzed after electrophoresis (9% SDS-PAGE) and protein transfer with a specific monoclonal antibody to GFP (Zymed or JL-8 clone from Clontech) or rabbit TF (clone 11F, 5 μg/ml, reducing conditions not used with this antibody). Immunoblotting for caveolin-1 was applied in some experiments as previously described (43).
Platelet-free plasma was prepared from human blood donated by healthy volunteers (Vacutainer tubes containing buffered sodium citrate). The activity of TF was quantified as the clotting time measured by observing 5-ml polypropylene tubes under constant agitation at 37°C containing freshly mixed human plasma (900 μl), 150 mM CaCl2 (50 μl), and a sample of cell plasma membrane (third pellet, 150,000 g) (23) dissolved in 50 μl tricine-sucrose buffer as a source of both TF and its adjuvant phospholipids. The membranes were prepared in an identical manner from transfected or control HEK-293 cells (recombinant forms of TF), cultured rabbit aortic SMCs, or a fresh rabbit brain (endogenously produced TF) and normalized on the basis of protein content (BCA protein assay, Pierce). Alternatively, resting or stimulated cells (75-cm2 flasks) were washed with serum-free medium and allowed to release TF-containing particles into the supernatant as a function of time; these particles were spun down (150,000 g × 3 h), and the pellet was resuspended in 50 μl tricine-sucrose buffer (23) for application to the clotting test (normalized on the basis of the cell surface). Negative controls without a source of TF or thrombin always failed to clot the plasma during the allowed observation period (10 min). Complementary experiments involved the addition of drugs or thrombin to the TF-plasma mixture before the clotting assay.
6-Amino-6-benzyl-1-ethyl-8-methoxy-5-oxo-octahydro-indolizine-3-carboxylic acid 4-carbamimidoyl-benzylamide (compound 4) from Hanessian et al. (21), a combined inhibitor of factor VIIa and thrombin, was a gift from E. Therrien and Prof. S. Hanessian (Université de Montréal). Thrombostat (bovine α-thrombin) was from Parke-Davis (Scarborough, Ontario, Canada). Active site-inhibited factor VIIa (ASIS) (41) was a gift from Dr. L. C. Petersen (Novo Nordisk A/S). The remaining drugs were purchased from Sigma (St. Louis, MO).
Functional activity of TF recombinant forms.
To determine whether TF-GFP and its variants were functional, membranes prepared from the homogenization of HEK-293 cells transiently expressing this protein, the deletion mutant TFΔC-GFP, or WT rabbit TF (WT-TF) were assayed in a coagulation test. Membranes from untransfected HEK-293 cells were used as a negative control, and those from a fresh rabbit brain were used as a positive control. When incubated with recalcified human plasma in polypropylene tubes, WT-TF and TF-GFP promoted clotting with approximately equal potencies and were nearly as potent as rabbit brain membranes when normalized for protein content (Fig. 2). Membranes from untransfected HEK-293 cells were inactive, in this respect, as those from cells expressing GFP (10 and 100 μg protein/reaction). Unexpectedly, membranes containing TFΔC-GFP were about 20-fold more active by protein content to clot plasma than those containing TF-GFP (Fig. 2).
Documented inhibitors of the proteolytic action of factor VIIa were assessed to further document the mechanism of the clotting induced by WT-TF or TF-GFP. In both cases, compound 4 (4–12 μM) (21) or nafamostat (10–30 μM) (48) considerably increased the clotting time in the reaction triggered by the TF-containing membranes and with partial selectivity because the inhibitors were less active to influence the clotting induced by α-thrombin (Table 1). ASIS (200 nM–2 μM), a form of factor VIIa inactivated by a covalently attached adduct at its active site (41), massively inhibited the coagulation induced by both recombinant TF forms but did not influence that induced by α-thrombin (Table 1). This supports that TF availability is limiting in the applied coagulation test and that the membrane preparations also supply the phospholipids necessary for this activity (46).
Confocal microscopy applied to TF-GFP.
The localization of TF in lipid rafts is a matter of current interest (see Introduction). HEK-293 cells were examined using confocal microscopy after transfection to compare the subcellular distribution of TF-GFP and its deletion mutant with other GFP-labeled proteins that are included or excluded from lipid rafts (Fig. 3). TF-GFP and TFΔC-GFP were observed as membrane-associated fluorescence with variable intracellular background (much less for the deletion mutant) in the plane that is halfway to the thickness of the cells (“equatorial”; Fig. 3). An examination of the apical plane of the cells evidenced numerous fluorescent filopodia. The formation of these apical filopodia may not be the result of TF-GFP expression or the indication of a partition to lipid rafts, because they were also present in other cells expressing any of the GFP-labeled membrane proteins whether they are known to be enriched in rafts (GPI-GFP, Fyn-GFP, B2R-GFP) or excluded from them (GFP-KRas; Fig. 3). Cells transfected for a GFP coding vector are filled with fluorescence because this protein is distributed in all the cellular water and the filopodia can be detected at the apical surface of the cells as extensions containing some cytosol (Fig. 3).
Relationship of TF with lipid rafts.
A purification of buoyant fractions of HEK-293 cells expressing either TF-GFP or GPI-GFP was applied to clarify the association of TF-GFP with lipid rafts (Fig. 4A). It was found that TF-GFP (∼73 kDa) comigrated (floatation) with the raft marker GPI-GFP (both proteins detected using anti-GFP monoclonals). The second fraction from the top (density 1.07–1.08) contained most of the immunoreactive TF-GFP and was also highly active to clot human plasma. The second fraction was slightly more active to clot than 100 μg of purified plasma membranes from HEK-293 cells expressing TF-GFP (Fig. 2A), but all fractions contained <0.4 μg protein/50 μl; thus TF-GFP is purified >250-fold in the buoyant cell fractions relative to the plasma membranes. β-Cyclodextrin (5 mM, 30 min) extracts cholesterol from cell membranes in serum-free culture media; this treatment increased the average density of rafts without disrupting them completely (Fig. 4B). This loss of uniformity is significant because fraction 3 is almost at the top of the density curve (Fig. 4A). The redistribution of TF-GFP induced by β-cyclodextrin was not parallel to a loss of the fusion protein in the plasma membrane fraction of treated HEK-293 cells (Fig. 4C).
Microparticles released by ionophore (10 min) in the serum-free supernatant of TF-GFP expressing HEK-293 (one 75-cm2 flask) were sequentially centrifuged to concentrate them (150,000 g × 3 h), resuspended in 12% Optiprep in tricine-sucrose buffer, and applied to the final Optiprep gradient centrifugation as previously described (43). Immunoblot (anti-GFP antibodies from Clontech) indicated the presence of TF-GFP only in the buoyant fractions of the supernatant of ionophore-treated cells (Fig. 4D), and the density of the released particles was similar to that of the TF-GFP-containing rafts prepared from the same type of cells (Fig. 4A).
The lipid raft distribution of TF-GFP was compared with that of other proteins that are mostly GFP labeled (Fig. 5). The four left-most lanes were the lightest fractions from the second Optiprep density centrifugation (fractions designated with exponent 2), and the following tracks were loaded from the six bottom fractions of the first Optiprep centrifugation (exponent 1) that are not carried over in the second centrifugation and contain either cytosolic or solubilized nonraft membrane proteins. TF-GFP and TFΔC-GFP (∼81 kDa) behave precisely as the known raft-partitioning proteins GPI-GFP, Fyn-GFP, B2R-GFP, and caveolin-1 in this analytic system, with all proteins being enriched in the fractions with a density close to 1.08. The cytosolic protein GFP and the membrane protein GFP-KRas are excluded from these buoyant fractions (Fig. 5).
Comparison of TF-GFP and TFΔC-GFP.
The experiments shown in Fig. 6 aim to explain the diversity of TF-GFP found in fractions of HEK-293 cells and obtain hints about their metabolism. Anti-GFP monoclonals (Clontech) were not reactive with the total cell extract of untransfected HEK-293 cells but revealed GFP (27 kDa) and TF-GFP (73 kDa) in cells transfected with the corresponding vectors (Fig. 6A). TF-GFP was reactive with anti-rabbit TF monoclonals as well (data not shown). In cells expressing TF-GFP, the presence of an immunoreactive band with a molecular weight close to that of GFP is an indication of a limited half-life for the conjugate, because GFP is relatively stable in mammalian cells (19). TFΔC-GFP was consistently more abundant than TF-GFP in cells transfected with the same amount of DNA (Fig. 6, B–D). The secretory pathway was interrupted by the Golgi-disrupting drug brefeldin A (14); this agent mainly increases the molecular weight dispersion of both TF-GFP (average molecular weight of 73 kDa) and of its deletion mutant TFΔC-GFP (paradoxically heavier, at 81 kDa; Fig. 6B). In the presence of the glycosylation inhibitor tunicamycin, discrete lower molecular bands of 57 and 54 kDa, respectively, appearred as the lower limit of the distribution for these two proteins (arrow in Fig. 6B); these molecular masses are precisely those predicted for the mature but unglycosylated constructions. Thus, on the average, molecules of TFΔC-GFP may be more heavily glycosylated than those of TF-GFP. In cells treated with the protein synthesis inhibitor anisomycin (0–4 h), there was a time-dependent decrease of content for either GFP conjugate, suggesting a relatively brief half-life similar for both constructions (Fig. 6C). Accordingly, the lysosomal acidification inhibitor bafilomycin A1 (4 h) similarly increased the total cell contents of the two GFP conjugates (Fig. 6D).
Confocal microscopy revealed that the subcellular localization of TF-GFP and TFΔC-GFP differs (Fig. 7A). Congruent with its more effective clotting efficacy, the latter form has preferential membrane localization in both HEK-293 cells and SMCs relative to TF-GFP. The latter fusion protein comparatively exhibits strong intracellular labeling, with a consistent positive perinuclear lining and heterogeneous granular material present in the cytosol. The frequency of cells of either type presenting a heavier intracellular fluorescence than their membrane labeling was significantly higher for TF-GFP compared with its deletion mutant in both cell types (Fig. 7B). In cells expressing TF-GFP and treated with brefeldin A, large fluorescent conglomerates were observed and often contiguous with a prominently labeled nuclear lining (Fig. 7A).
Nature of the procoagulant activity from rabbit aortic SMC membranes.
A membrane preparation from SMCs maintained in culture in the presence of FBS (10%) was highly active to clot human citrated and recalcified plasma (Table 1; 30 μg/reaction). Documented inhibitors of the proteolytic action of factor VIIa, compound 4, nafamostat, and ASIS, were highly active to increase the clotting time, supporting that the presence of endogenous TF is limiting in the clotting assay applied to SMC membranes.
Regulated expression of endogenous TF in rabbit aortic SMCs.
Serum-starved cells expressed less clotting activity (Fig. 8A) and immunoreactive rabbit TF (immunoblot, ∼46-kDa band; Fig. 8B) relative to cells that were starved for 24 h and further stimulated with either 10% FBS or IL-1β for 4 h. These two types of stimuli exerted no additive effects (Fig. 8). TF was detected in an immunofluorescence assay based on anti-rabbit monoclonal antibody (Fig. 8C). The antibody appeared specific as nonpermeabilized COS-1 expressing recombinant rabbit WT-TF (transient transfection) exhibited a membrane expression of the protein (data not shown). Rabbit cultured aortic SMCs expressed more or less membrane TF as a function of tissue culture conditions, the surface protein being induced by 4-h IL-1β or FBS treatments (Fig. 8C, effects not additive, as in Fig. 8A). Endogenous TF was associated with low-density lipid rafts and comigrated with caveolin-1 in Na2CO3 extracts of SMCs (Fig. 5, bottom).
The expression of another inflammatory gene product has been studied under the same set of culture conditions. The kinin B1 receptor (binding assay, saturating radioligand concentration, Fig. 8D) behaved similarly as TF, with both FBS and IL-1β being stimulatory (the latter less than the former).
Effect of ionophore on the functional TF expression and distribution in rabbit cultured SMCs.
A first series of experiments was related to the concept of acute “deencryption” of TF activity, here applied to the SMC model that expresses endogenous TF (cells maintained in 10% FBS). The calcium ionophore A23187 (5 μM) is a rapid and drastic stimulant of TF clotting activity in the sedimentable particles from washed SMCs treated with the ionophore for 10 min (Fig. 9A). The membrane clotting activity was also increased by this treatment; the ionophore was active in this respect in other cell types (29, 46). The effects of the agent were abated by concurrent EGTA treatment, supporting the effect of extracellular calcium entry in the effect of A23187 (Fig. 9A). Rabbit aortic SMCs transiently expressing TF-GFP illustrated filopodium labeling (Fig. 9, C–E) as the HEK-293 cells (Fig. 3). SMCs exhibited a strong, finely granular cellular labeling in the plane near the surface that contains the filopodia (Fig. 9, C–E). The TF-GFP-positive filopodia are elaborate and branching; they are fragmented after ionophore treatment (Fig. 9C, top right), consistent with the finding of sedimentable particles with clotting activity in the supernatant of treated cells (Fig. 9A). A further effect of ionophore treatment was to increase the number of protrusions at the cell surface (Fig. 9C, bottom right); some contraction of the cell was also observed, but not the translocation of the finely granular intracellular fluorescence to the cell surface. TF-GFP-positive filopodia coexpressed polymerized actin, at least at the base of the filopodia, as shown in dual labeling experiments that exploit red-labeled phalloidin (Fig. 9D, matched frames from the same cells, but only some cells in the field expressed TF-GFP). Several other types of treatment failed to translocate the intracellular granular fluorescence associated with TF-GFP to the cell surface in 15–60 min: the B1 receptor agonist Lys-des-Arg9-BK (100 nM; Fig. 9E), forskolin (50 nM), iloprost (50 nM), and phorbol myristate acetate (1 μM) (data not shown).
The possible expression of enzymes whose synthesis is induced by inflammation, like inducible nitric oxide (NO) synthase (NOS) and cyclooxygenase (COX)-2, may be parallel to that of TF and of the B1 receptor in SMCs. Because endogenous prostanoids or NO may interfere with the effect of kinins on TF procoagulant activity (as previously observed for the B1 receptor-mediated stimulation of DNA synthesis masked by the endogenous COX products in rabbit aortic SMCs; Ref. 28), we combined the kinin agonist stimulation with NOS or COX inhibition. Exogenous IL-1β (5 ng/ml) increased the clotting activity of the membranes in serum-starved SMCs and adding the nonselective NOS or COX inhibitors NG-nitro-l-arginine (10 μM) or diclofenac (500 nM), respectively, to the cytokine treatment (4 h) did not conclusively further increased the activity; the clotting activity in the cell supernatants were not measurable in this series of experiments (data not shown). FBS restoration (10%, 4 h) is a stimulus for both kinin B1 receptor and TF expression (Fig. 8); in FBS-treated cells, diclofenac only slightly increased the membrane clotting activity in the membranes (Fig. 9B; not in the supernatant, data not shown). Adding the peptidase-resistant B1 receptor agonist Sar-[d-Phe8]des-Arg9-BK at a concentration known to maximally stimulate the receptors (28) had no consistent effect on FBS-stimulated SMCs, whether treated or not with diclofenac (Fig. 9B shows the activity in membranes; data not shown for supernatant activities). Concurrent treatment with NG-nitro-l-arginine did not modify the clotting effect of FBS and also failed to unmask an effect of the B1 receptor agonist (Fig. 9B).
To our knowledge, this study is the first to take advantage of a GFP conjugate of TF. In reference to the function of TF as the main initiator of the blood coagulation cascade (46), TF-GFP functionally compares well with WT-TF (Fig. 2 and Table 1). TF-GFP is approximately equipotent with WT-TF to clot human plasma but may be somewhat more sensitive than the natural molecule to direct (ASIS) or indirect inhibitory drugs (the factor VIIa inhibitors compound 4 and nafamostat; Table 1). A minor conformational change induced by the fusion with GFP may determine these subtle changes. However, HEK-293 cell membranes containing TFΔC-GFP were more potent than those containing TF-GFP to clot plasma (Fig. 2), a fact that we interpret as the result of preferential membrane expression, as shown in confocal microscopy in two cell types (Fig. 7). TFΔC-GFP also is globally more abundant in HEK-293 cells transfected with equal amount of plasmids for the two GFP conjugates (immunoblots, Fig. 6). Paborsky et al. (39) previously observed that a similar truncation in human TF increased the clotting activity in the total cell extract relative to WT-TF expressed in HEK-293 cells without affecting the specific clotting activity of the two TF forms; they provided no explanation for the expression difference. A divergence of posttranscriptional processing between TF-GFP and TFΔC-GFP is suggested by the apparent hyperglycosylation of the deletion mutant (Fig. 6) and may also involve a lower proportion of rejected proteins during their maturation owing to molecular properties that remain to be determined. Another laboratory (8) showed that the membrane clotting activity is higher in MDCK cells expressing truncated TF than that determined by WT-TF; our data based on fluorescence imaging support that the preferential membrane localization is a property that may be distinct from the absolute abundance in the total cell extract. Because the two fluorescent forms apparently exhibit a similar half life (<4 h) in cells in which the protein synthesis has been blocked (Fig. 6C), our results suggest that TF-GFP is retained in the secretory pathway relative to its deletion mutant. Previous ultrastructural studies have evidenced a strong labeling of the rough endoplasmic reticulum in various cell types expressing TF (37, 47), including rabbit aortic SMCs injured in vivo (22). Thus the granular microscopic appearance of TF-GFP in the cytosol and also in the nuclear lining may reflect the slow or ineffective maturation process of TF in the early secretory pathway (as the nuclear envelope communicates with the endoplasmic reticulum). Disruption of the Golgi network with brefeldin A in the two tested cell types shows further increase of the nuclear lining labeling with TF-GFP, with the formation of large aggregates that are at least in part contiguous to the nuclei (Fig. 7). Brefeldin A is known to redistribute Golgi proteins to the endoplasmic reticulum (14). TFΔC-GFP, apparently retained to a lesser extent in the secretory pathway, is also more heavily glycosylated based on apparent molecular weight and the effect of tunicamycin (Fig. 6); this construction does not lack the potential palmitoylation site of TF, but two serine residues proposed to be phosphorylated were deleted (Fig. 1). The precise mechanism of intracellular retention of TF remains to be determined, but the phosphorylated tail of TF has been proposed to bind to the cell actin skeleton via a filamin adaptor (38). Although no known spontaneous mutation reproduces the COOH-terminal tail truncation of TF, considerable experimental efforts have been devoted to elucidate the role of this potentially regulatory domain of TF in embryonic development and pathophysiological animal models (1, 5, 7, 40, 49). For instance, “knockin” mice expressing TFΔC in a regulated manner exhibit increased angiogenesis in response to tumor graft (5), a fact that may be related to more effective cell surface expression.
The present study reveals that TF-GFP is highly enriched in lipid rafts isolated from HEK-293 cells; this also applied to endogenous TF in SMCs (Figs. 4 and 5). The controversy about TF partitioning into rafts may be related to methodology because detergent-based isolation is likely to reveal more stringent association than the sodium carbonate (detergent free) method used in the present study. It has been recently found that TF presence in cholesterol-rich rafts increases its affinity for factor VIIa in fibloblasts but that cholesterol depletion of cells does not per se inhibit TF membrane expression (30). We show that β-cyclodextrin treatment applied to HEK-293 cells expressing TF-GFP determines a loss of raft density homogeneity (Fig. 4B) with no loss of whole membrane abundance (Fig. 4C), consistent with these findings. A comparison of GFP-labeled membrane proteins included in rafts (TF-GFP, TFΔC-GFP, GPI-GFP, Fyn-GFP, B2R-GFP) or excluded from them (GFP-KRas; Fig. 5) does not reliably predict preferential labeling of HEK-293 cell regions, such as the apical filopods (Fig. 3), using photonic microscopy, despite the suggestion that such cell protrusions are enriched in both TF (10, 36) and lipid rafts (20).
The release of TF-rich procoagulant particles is believed to be a medically important phenomenon (see Introduction). The shedding of such particles from cultured cells has been modelled in a number of studies by ionophore treatment (29, 46). Human SMCs slowly (≥6 h) release TF-containing particles into their supernatant (≤200 nm, density of 1.10) (45). In the present study, the low-density value of at least a fraction of the TF-GFP-containing particles released by ionophore from HEK-293 cells (Fig. 4D) is practically undistinguishable from that of purified lipid rafts (Figs. 4 and 5). TF may possess a natural affinity for cholesterol-rich rafts and may be sorted in exosomes, as other lipid raft-associated proteins (12). The exosomes, the internal vesicles of multivesicular bodies, have been reported in some systems to contain TF (16). Exosomes are released in the extracellular medium upon fusion of these bodies with the plasma membrane. The lipid raft domains of the exosomes confer to them a low density comparable with that observed for the TF-GFP-containing particles released from HEK-293 cells. Further ultrastructural work will be needed to confirm the identity of intracellular TF-containing particles as exosomes. It is also likely that larger segmented membrane protrusions (Fig. 9C) that contain actin can also be released after ionophore treatment and contribute to the clotting activity released in cell supernatants (Fig. 9A). The ionophore failed to clear the considerable intracellular content of TF-GFP in transfected SMCs (Fig. 9), suggesting that this pool is not rapidly secretable. As discussed elsewhere, a sizeable proportion of the deencryption caused by the ionophore may derive from the activation of a phospholipid “scramblase” independently from TF translocation (13).
We have found that the expression of endogenous TF in rabbit aortic SMCs is a regulated process largely parallel to the expression of an inducible G protein-coupled receptor, the kinin B1 receptor (Fig. 8). TF gene expression is known to be controlled by both activator protein (AP)-1 and NF-κB in murine fibroblasts (18). Thus IL-1β and FBS, documented activators of NF-κB in rabbit aortic SMCs (44), are nonadditive stimulant of the expression of TF (immunoblot, immunofluorescence, clotting activity in membranes; Fig. 8). The greater effect of FBS relative to that of IL-1 may be related to the additional recruitment of AP-1 by serum (18). B1 receptor stimulation was hypothesized to interact with TF surface expression in rabbit vascular SMCs because it is linked to phospholipase C and calcium signaling in these cells (27, 28). An agonist of this receptor did not reproduce the effect of ionophore on SMCs, because no consistent clotting activity was deencrypted in up to 4 h (Fig. 9B) and the filopodia were not segmented (Fig. 9E). Various times of exposure failed to reveal a consistent effect for B1 receptor agonists on these end points (data not shown). This may only reflect the lack of a regulation of the secretion process in this cell type by stimuli of physiological intensity. In vascular endothelial cells, TF is also localized in intracellular secretory organelles (Weibel-Palade bodies) and cell surface caveolae-related rafts (35). The secretion of Weibel-Palade bodies after the stimulation of G protein-coupled receptors (protease-activated receptors) (26) in endothelial cells suggests that the B1 receptor-TF functional interaction may exist this type of vascular cell, which can express this type of kinin receptors (32).
TF is a raft-associated molecule whose surface expression (secretion) is apparently retarded or impaired by structural determinant(s) located in its intracellular COOH-terminal tail.
This work was supported by Canadian Institutes of Health Research Grant MOP-14077 (to F. Marceau and A. Adam) and by a Fonds de la Recherche en Santé du Québec Studentship Award (to J.-P. Fortin).
We thank Johanne Bouthillier for technical help.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2005 by the American Physiological Society