Investigation of the Filamin A–Dependent Mechanisms of Tissue Factor Incorporation into Microvesicles

Abstract We have previously shown that phosphorylation of tissue factor (TF) at Ser253 increases the incorporation of TF into microvesicles (MVs) following protease-activated receptor 2 (PAR2) activation through a process involving filamin A, whereas phosphorylation of TF at Ser258 suppresses this process. Here, we examined the contribution of the individual phosphorylation of these serine residues to the interaction between filamin A and TF, and further examined how filamin A regulates the incorporation of TF into MVs. In vitro binding assays using recombinant filamin A C-terminal repeats 22–24 with biotinylated phospho-TF cytoplasmic domain peptides as bait showed that filamin A had the highest binding affinities for phospho-Ser253 and double-phosphorylated TF peptides, while the phospho-Ser258 TF peptide had the lowest affinity. Analysis of MDA-MB-231 cells using an in situ proximity ligation assay revealed increased proximity between the C-terminus of filamin A and TF following PAR2 activation, which was concurrent with Ser253 phosphorylation and TF-positive MV release from these cells. Knock-down of filamin A expression suppressed PAR2-mediated increases in cell surface TF procoagulant activity without reducing cell surface TF antigen expression. Disrupting lipid rafts by pre-incubation with methyl-β-cyclodextrin prior to PAR2 activation reduced TF-positive MV release and cell surface TF procoagulant activity to the same extent as filamin A knock-down. In conclusion, this study shows that the interaction between TF and filamin A is dependent on the differential phosphorylation of Ser253 and Ser258. Furthermore, the interaction of TF with filamin A may translocate cell surface TF to cholesterol-rich lipid rafts, increasing cell surface TF activity as well as TF incorporation and release into MVs.


Introduction
Tissue factor (TF) is a transmembrane protein that is expressed on the cell surface and released from cells in microvesicles (MVs). TF has a small cytoplasmic domain, which has no intrinsic kinase activity, and is not required for the procoagulant activity of TF. 1 However, Ser253 and Ser258 within the cytoplasmic domain of TF are phosphorylated following the activation of protein kinase C-α (PKCα) 2 and p38, 3 respectively. The phosphorylation of Ser253 occurs first, which subsequently allows Ser258 to be phosphorylated. 2 TF phosphorylation mediates several cellular processes including cell migration, 4 angiogenesis, 5 vascular endothelial growth factor production 6 and metastasis. 7 We have previously shown that the incorporation of TF into MVs following protease-activated receptor 2 (PAR2) activation is differentially regulated by the phosphorylation of Ser253 and Ser258, whereby phosphorylation of Ser253 promotes the incorporation of TF into MVs and phosphorylation of Ser258 suppresses this process. 8 We have also demonstrated that the presence of filamin A is crucial for TF to be incorporated into MVs in response to PAR2 activation. 9 Filamin A is a 280-kDa protein composed of an N-terminal actin-binding domain and 24 immunoglobulin (Ig)-like repeats. 10 Repeat 24 allows filamin A molecules to dimerize to form a 'leaf spring' structure, 10 while the actin-binding domain cross-links actin filaments. Filamin A also binds to over 90 transmembrane and cytosolic proteins, 11 most of which interact with Ig repeats 16-24 of filamin A, which regulates the function and cellular localization of these proteins. 11 TF also binds to this region of filamin A through a direct interaction between the cytoplasmic domain of TF and the C-terminus repeats 22-24 of filamin A. 12 Furthermore, phosphorylation of the cytoplasmic domain of TF enhances the binding to filamin A compared with nonphosphorylated TF. 12 However, the contribution of the individual phosphorylation at Ser253 and Ser258 to this interaction has not been explored. The aim of this study was to determine whether the individual phosphorylation of TF at Ser253 and Ser258 differentially modulates the binding affinity of TF for filamin A, in an attempt to further clarify the mechanism by which the sequential phosphorylation of Ser253 and Ser258 may regulate the incorporation of TF into MVs in a filamin A-dependent mechanism.

Cloning and Expression of Repeats 22-24 of Filamin A
A synthetic gene encoding repeats 22-24 of filamin A (amino acids 2329-2647; GeneArt, Thermo Fisher Scientific, Paisley, United Kingdom) was sub-cloned into a bacterial expression vector in tandem with an N-terminal His (Â6) tag by the Protein Expression Laboratory (University of Leicester, United Kingdom). The insert sequence was confirmed and the plasmid was transformed into Escherichia coli BL21(DE3). Recombinant protein was isolated and purified on a nickel column using an ÄKTA purifier (GE Healthcare Life Sciences, Bucks, United Kingdom). The recombinant filamin A was dialysed overnight in phosphate buffer saline (PBS) and then examined by SDS-PAGE (sodium dodecyl sulphate polyacrylamide gel electrophoresis) and western blot analysis to ensure it had the correct molecular weight (36 kDa) and was recognized by a filamin A antibody (EP2405Y; GeneTex Inc., California, United States) directed against the 24th C-terminal repeat of filamin A (►Supplementary Fig. 1).

TF Cytoplasmic Domain/Recombinant Filamin A Binding Assays
High-binding 96-well plates (Greiner Bio-One Ltd., Gloucester, United Kingdom) were coated with 1 μM of recombinant filamin A repeats 22-24 or 1 μM of bovine serum albumin (BSA) in 50-mM carbonate buffer, pH 9.6 overnight at 4°C. The plates were then washed three times with wash buffer (PBS, pH 7.4, 1-mM CaCl 2 , 1-mM MgCl 2 , 0.05% [v/v] Tween 20) and blocked with 5% (w/v) BSA in PBS for 1.5 hours. The plates were washed three times and incubated for 2 hours with 0 to 10 μM of biotinylated TF peptides corresponding to the final 19 amino acids of the cytoplasmic domain of TF (biotin-CRKAGVGQSW-KENSPLNVS) with phosphorylation at Ser253 and/or Ser258 (shown as underlined), or non-phosphorylated (95% purity; Biomatik, Ontario, Canada). The plates were washed three times and incubated for 1 hour with streptavidin-HRP (horseradish peroxidase) diluted 1:2,000 (GE Healthcare Life Sciences). The plates were washed again and TMB (3,3′,5,5′tetramethylbenzidine) One Solution (Promega, Southampton, United Kingdom; 100 μL) added and developed for 10 minutes. Reactions were stopped with 2-M sulphuric acid (50 μL) and the absorbance was recorded at 450 nm. Data were analysed using GraphPad Prism 6 using non-linear regression and onesite binding, and K d values were calculated. The specific binding of TF peptides to recombinant filamin A was calculated as the absorption values for the binding of the TF peptides to filamin A following subtraction of the values for the binding of the TF peptides to BSA at each respective concentration of TF peptide.

Cell Culture and Transfection
The breast cancer cell line MDA-MB-231 was cultured in DMEM (Dulbecco's Modified Eagle's Medium) containing 10% (v/v) heat-inactivated foetal calf serum (FCS) and maintained at 37°C under 5% (v/v) CO 2 . Cells were seeded out into appropriate plates 48 hours prior to experiments and adapted to serum-free medium (SFM) by 24-hour incubation in 5% (v/v) FCS medium, followed by two washes with PBS and 1-hour incubation in SFM. Cells were then treated with PAR2 agonist peptide (PAR2-AP; SLIGKV-NH 2 ; 20 μM; Tocris Bioscience, Bristol, United Kingdom) or human factor VIIa (FVIIa; 5 nM; Enzyme Research Laboratories Ltd., Swansea, United Kingdom). To knock down filamin A expression, cells were reverse transfected with Silencer Select filamin A siRNA (2 nM) or Silencer Select negative control siRNA1 (2 nM) using Lipofectamine RNAiMAX (all from Thermo Fisher Scientific) as previously described. 10

Isolation and Analysis of MVs from Cell Culture Medium
MDA-MB-231 cells were cultured in T75 flasks until 80% confluent, adapted to SFM as descried previously and activated with PAR2-AP (20 μM) or FVIIa (5 nM) for 0 to 60 minutes. The culture medium was centrifuged at 1,500 g for 15 minutes, followed by centrifugation of the resultant supernatant at 40,000 g for 30 minutes at 20°C. The pelleted MVs were washed with HEPES (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid) buffered saline (HBS), pH 7.4 (10 mL) and centrifuged again. MVs were then re-suspended in HBS (100 μL) and stored at -80°C.
The concentrations of MVs and the TF activity of the MVs were determined using the calibrated automated thrombogram (CAT) assay. Platelet-poor pooled human plasma was prepared from blood from 15 healthy donors (ethical approval obtained from the University of Leicester) taken into 3.2% (w/v) tri-sodium citrate, supplemented with corn trypsin inhibitor (CTI; Cambridge Biosciences; 18.3 μg/mL), and centrifuged at 1,800 g for 30 minutes at room temperature. The plasma was aspirated and equal volumes from each donor pooled and then passed through 0.22-μm filters to remove endogenous MVs. The cell-derived MV samples were mixed with filtered pooled plasma at a ratio of 1:4. The MP-Reagent (4-µM phospholipids, no TF; Diagnostica Stago Ltd., Ontario, Canada) was used to measure the TF activity of the MVs and the PRP-Reagent (no phospholipids, 1-pM TF) was used to measure the phospholipid-mediated procoagulant activity of the MV samples. MP-or PRP-Reagent (20 μL) was added to each well in 96-well plates followed by the addition of the MV/human plasma mix (80 μL). Thrombin generation was then measured using the CAT assay on a Fluoroskan Ascent Fluorometer. Lag times and endogenous thrombin potential (ETP) were calculated using the Thrombinoscope software. MVs isolated from cells treated with FVIIa were measured using the Zymuphen MP-Activity assay (Hyphen-Bio Med, Neuville sur Oise, France) according to the manufacturers' instructions, due to interference from the added FVIIa in the CAT assay.
To measure the TF antigen content of the MVs by enzymelinked immunosorbent assay (ELISA), MV samples were first supplemented with 1% (v/v) Triton X-100 (to extract TF out of MVs) and then diluted 1:1 in diluent buffer and quantified using the Tissue Factor Quantikine ELISA kit (R&D Systems, Abingdon, United Kingdom) according to the manufacturer's instructions.

Immunoprecipitation of Tissue Factor
MDA-MB-231 cells were cultured to 80% confluence in T25 flasks, adapted to SFM as described previously and activated with PAR2-AP (20 µM) or FVIIa (5 nM). Cells were washed with PBS and lysed in PhosphoSafe lysis buffer (Merck-Millipore, Nottingham, United Kingdom; 500 µL) containing protease inhibitor cocktail (1% v/v; P8340; Sigma Chemical Company Ltd., Poole, United Kingdom). The lysates were centrifuged at 10,000 g on a microcentrifuge for 10 minutes and the supernatant incubated with the mouse anti-TF antibody HTF1 (10 µg; Affymetrix eBioscience, Altrincham, United Kingdom) or mouse control IgG (Cell Signaling Technology, Leiden, The Netherlands) at 4°C overnight. PureProteome protein A magnetic beads (15 µL; Merck-Millipore) were added to the samples and incubated at 4°C for 90 minutes. The beads were then washed five times with PBS, pH 7.2 (1 mL) containing 0.05% (v/v) Tween-20 and denatured by boiling in Laemmli's buffer (70 µL; Sigma Chemical Company Ltd.) for 10 minutes. The supernatant was removed from the beads and proteins were separated by 12% (w/v) SDS-PAGE. Membranes were probed with a rabbit anti-TF antibody (FL-295; Santa Cruz Biotechnology Inc., Heidelberg, Germany), a rabbit anti-phospho-serine PKC substrate antibody (New England Biolabs, Hitchin, UK) or a rabbit anti-phospho-Ser258 TF antibody (Abcam, Cambridge, United Kingdom) and detected using an anti-rabbit alkaline phosphatase-conjugated secondary antibody (Santa Cruz Biotechnology Inc.). The membranes were developed with Western Blue substrate for alkaline phosphatase (Promega). Images were acquired on an Ima-geQuant LAS-4000 imager (GE Healthcare Life Sciences) and densitometry was performed using the ImageQuant software.

Duolink Proximity Ligation Assays
MDA-MB-231 cells (10 4 ) were seeded out into 96-well plates (Greiner Bio-One Ltd.), adapted to SFM as described previously and incubated with PAR2-AP (20 µM) or FVIIa (5 nM). The cells were fixed for 15 minutes with 4% (v/v) paraformaldehyde, washed three times with PBS, permeabilized with saponin (0.2% [v/v]) for 5 minutes and blocked with Duolink blocking buffer (Sigma Chemical Company Ltd.) for 1 hour. Samples were incubated overnight at 4°C with the mouse anti-TF antibody HTF1 (5 µg/mL) together with the rabbit anti-filamin A antibody EP2405Y (raised against the C-terminus of filamin A; 2.6 µg/mL) in Trisbuffered saline (TBS; 20-mM Tris-HCl, pH7.4, 150-mM NaCl) supplemented with 1% (v/v) blocking buffer. The anti-filamin A antibody was replaced with rabbit IgG (New England Biolabs Inc.; 2.6 µg/mL), or the anti-TF antibody was replaced with normal mouse IgG (5 µg/mL) in negative control samples. The cells were washed three times with TBS containing 0.05% (v/v) Tween-20 (TBST) and the proximity ligation assay (PLA) procedure performed according to the manufacturer's instructions. The actin cytoskeleton was labelled by incubation with Acti-stain 488 (100 nM; Cytoskeleton Inc., Denver, CO, United States) and nuclei were labelled with DAPI (4′,6-diamidino-2-phenylindole; 2 µg/mL; Sigma Chemical Company Ltd.). Images were acquired using an EVOS fluorescent microscope (Thermo Fisher Scientific). The number of red fluorescent events and nuclei was determined using ImageJ. Flow Cytometry for Cell Surface TF MDA-MB-231 cells (2 Â 10 5 ) were seeded out into 6-well plates, adapted to SFM as described previously and incubated with PAR2-AP (20 µM) or FVIIa (5 nM) for 30 minutes. Cells were then detached using TrypLE Select (Thermo Fisher Scientific), pelleted by centrifugation at 260 g for 5 minutes on a microcentrifuge and washed with HBS containing 0.1% (w/v) BSA and 0.1% (v/v) sodium azide. Cells were incubated for 60 minutes on ice in HBS/BSA/sodium azide containing either the mouse anti-TF Alexa Fluor-488 antibody (R&D Systems) or the mouse IgG Alexa Fluor-488 isotype control antibody (R&D Systems) at a final concentration of 1 μg/mL. Cells were analysed using a CyAn flow cytometer (Beckman Coulter, High Wycombe, UK) with a gate set at 2% of the control IgG. Data were analysed using the Summit software (Beckman Coulter Inc.).

Cell Surface TF Activity Assay
MDA-MB-231 cells (5 Â 10 3 ) were seeded out into 96-well plates, adapted to SFM as described previously and incubated with PAR2-AP (20 µM) for 30 minutes. Cells were then washed twice with HBS containing 1 mg/mL BSA and 0.1% (v/v) sodium azide (200 μL), and incubated either with the mouse anti-TF inhibitory antibody HTF1 (40 μg/mL) to block TF procoagulant activity or with the mouse control IgG antibody (40 μg/mL) at room temperature for 15 minutes. Filtered platelet-poor, pooled human plasma containing CTI (18.3 µg/mL; 60 μL) was then added to each well and thrombin generation was measured using the CAT assay in the absence of any exogenous reagent. TF-specific thrombin generation was calculated as the ETP of samples incubated with control IgG minus the ETP of samples incubated with the inhibitory TF antibody HTF1.

Statistical Analysis
Data represent the calculated mean values from the number of experiments stated in the figure legends AE the calculated standard error of the mean. The data were analysed using the statistical package for the social sciences (SPSS). Significance was determined using one-way ANOVA (analysis of variance) and Tukey's honest significance test, and values of p < 0.05 were considered significant. To confirm these data, non-biotinylated phospho-Ser253 and non-biotinylated double phosphorylated TF peptides were used to compete out binding of biotinylated TF peptides in different phosphorylation states. In the presence of the highest concentration of the non-biotinylated phospho-Ser253 peptide, the binding of the phospho-Ser258 peptide was reduced by up to 68.5 AE 12.7%, and the non-phosphorylated TF peptide was competed out by 19 AE 11.4%; however, there was no significant influence on the binding of the double phosphorylated TF peptide (►Fig. 1B). Similar results were observed with increasing concentrations of the nonbiotinylated phospho-Ser253/258 peptide to compete out binding of the biotinylated phospho-Ser258 peptide to recombinant filamin A (►Fig. 1C). A polyclonal anti-filamin A antibody also competed out the binding of the phospho-Ser253 TF peptide, indicating that the binding was mediated through repeats 22-24 of filamin A (►Fig. 1D).  Fig. 2A), a trend was observed for cells to release increased levels of phosphatidylserine-positive MVs over 60 minutes, which was accompanied by higher MV-associated TF antigen levels (►Fig. 2B). MVs released in response to PAR2-AP also had increased TF activity as shown by a decrease in lag time in the CAT assay (n ¼ 3; ►Supplementary Fig. 2B). The procoagulant activity associated with the MVs was TF dependent because pre-incubation of the samples with the inhibitory TF antibody HTF1 reduced the ETP and extended the lag time to more than 25 minutes (►Supplementary Fig. 3). Incubation of MDA-MB-231 cells with PAR2-AP also resulted in a significant increase in the phosphorylation of TF at Ser253, with maximal levels observed at 30-minute activation (►Fig. 2C, E). Ser253 phosphorylation was followed by phosphorylation of Ser258, which peaked at 50 minutes post-activation with PAR2-AP (►Fig. 2D, E). Data from two experiments showed a similar pattern of TF phosphorylation in cells activated with FVIIa (►Supplementary Fig. 4). The release of TF-positive MVs and the phosphorylation of TF at Ser253 and Ser258 in endothelial cells following PAR2 activation has been shown previously. 8

Activation of PAR2 Results in Increased Proximity of TF and Filamin A in MDA-MB-231 Cells and Primary Endothelial Cells
The proximity of TF and filamin A was examined by an in situ procedure employing the Duolink PLA, which permits the detection of two molecules within close ( 40 nm) proximity of each other. 13 Close proximity between TF and the C-terminus of filamin A was detectable in non-activated MDA-MB-231 cells, and a trend was observed for the number of interactions per cell to increase following PAR2 activation with the agonist peptide, reaching the maximum at 40 minutes (►Fig. between TF and filamin A (n ¼ 3, p ¼ 0.027 vs. untreated cells; ►Supplementary Fig. 5B). In contrast, few TF-filamin-A interactions were detected in non-activated primary endothelial cells transfected to express low levels of TF (►Supplementary Fig. 5C). However, as with MDA-MB-231 cells, there was a trend for the number of interactions to increase in both types of endothelial cells following PAR2 activation (►Supplementary Fig. 5C). Fewer TF-filamin-A interactions were detected in endothelial cells compared with MDA-MB-231 cells, with only a few events observed per cell, probably as a result of low levels of TF expression. Co-immunoprecipitation of TF and filamin A using either anti-TF or anti-filamin A antibodies was not possible, and pull-down assays using the TF cytoplasmic domain peptides conjugated to magnetic beads did not pull down filamin A from cell lysates.

Phosphorylation of Ser253 within the TF Cytoplasmic Domain Enhances the Proximity of TF and Filamin A in Endothelial Cells
To examine whether serine 253 phosphorylation enhances the interaction between TF and filamin A in cells, primary endothelial cells (HDBEC) were transfected to express mutant forms of TF with substitution of serine 253 with an alanine residue (to prevent phosphorylation) or an aspartate residue (to mimic phospho-serine) and analysed by PLA. Transfected endothelial cells expressed similar levels of wildtype and mutant forms of TF (►Supplementary Fig. 6). Alanine substitution of residue 253 resulted in lower proximity between TF and filamin A in PAR2-activated cells compared with activated cells expressing wild-type TF (►Fig. 3B). In contrast, increased proximity between TF and filamin A was observed in cells expressing the aspartate 253-substituted TF above levels observed in cells expressing wild-type TF, in both non-activated and PAR2-activated cells (►Fig. 3B).

Knock-Down of Filamin A Expression Prevents PAR2-Mediated Increases in Cell Surface TF Activity on MDA-MB-231 Cells
We have previously shown that knock-down of filamin A expression suppresses the incorporation of TF into MVs in response to PAR2 activation, 9 although the underlying mechanism is unclear. Therefore, here we examined whether filamin A regulates the translocation of TF to the cell surface, or alternatively if it controls the cell surface localization of TF, facilitating the incorporation of TF into MVs. As shown previously, 9 reverse transfection of MDA-MB-231 cells with 2 nM of filamin A Silencer Select siRNA resulted in the partial knockdown of filamin A expression, but did not alter TF expression (►Supplementary Fig. 7). Suppression of filamin A expression using specific siRNA abolished increases in TF-specific thrombin generation following PAR2 activation in transfected MDA-MB-231 cells compared with that observed on the surface of MDA-MB-231 cells transfected with control siRNA (►Fig. 4A). The increase in thrombin generation was shown to be predominantly TF specific and was suppressed by incubation of the cells with the inhibitory TF antibody HTF1 (40 μg/mL; ►Supplementary  compared with non-transfected cells, indicating that transfection with siRNA itself had no effect on overall cellular behaviour (►Supplementary Fig. 9).

Disruption of Lipid Rafts Reduces TF-MV Release and Suppresses PAR2-mediated Increases in Cell Surface TF Activity
A recent study has shown that macrophage-derived TF-positive MVs originate from lipid rafts. 14 Furthermore, TF associated with cholesterol-rich microdomains has been shown to possess increased procoagulant activity. 15 Therefore, since filamin A knock-down suppressed PAR2-mediated increases in cell surface TF activity (►Fig. 4A), we tested whether filamin A is required for the translocation of TF into cholesterol-rich microdomains and subsequent incorporation into MVs. Depletion of membrane cholesterol from MDA-MB-231 cells with methyl-β-cyclodextrin (MβCD) significantly reduced the release of MVs from MDA-MB-231 cells (►Fig. 5A). Similarly, treatment of cells with MβCD also reduced the associated TF activity of these MVs (►Fig. 5B). Consequently, the outcome of concurrent filamin A knock-down and lipid raft disruption on TF activity on the surface of cells was examined. Cell surface TF activity was reduced in the presence of MβCD alone (►Fig. 5C), in agreement with previous reports that lipid rafts support TF procoagulant activity. 15,16 Both pre-incubation of the cells with MβCD and the suppression of filamin A expression reduced cell surface TF activity to similar levels following PAR2 activation. In addition, no additional or cooperative reduction in TF activity was observed on pre-incubation of filamin A knock-down cells with MβCD (►Fig. 5C). To examine whether TF phosphorylation alone is sufficient for TF to be translocated to lipid rafts and therefore increase TF procoagulant activity, MDA-MB-231 cells were incubated with phorbol myristate acetate (PMA; 100 nM) for 30 minutes to directly activate PKCα and induce TF phosphorylation. 17 Analysis of PMA-treated and -untreated cells showed no differences in levels of cell surface TF activity, which was also unaltered by suppression of filamin A expression (►Supplementary Fig. 10). Increased levels of phosphatidylserine (PS) on the outer leaflet of the plasma membrane can also augment TF procoagulant activity. 18 Therefore, the effect of the suppression of filamin A expression on levels of cell surface PS was examined by flow cytometry for annexin V binding, and showed that cells with filamin A knock-down had marginally higher levels of cell surface PS compared with cells transfected with control siRNA (►Fig. 6).

Discussion
We previously reported that the incorporation of TF into MVs in response to PAR2 activation is promoted by the phosphorylation of TF at Ser253, and subsequently suppressed by phosphorylation of Ser258. 8 Furthermore, we have shown that filamin A is essential for the increased incorporation of TF into MVs following PAR2 activation. 9 However, the underlying mechanisms of filamin A-dependent incorporation of TF into MVs are not known. Since filamin A binds to the cytoplasmic domain of TF in a phosphorylation-dependent manner, 12 the main aim of this study was to examine the individual role of Ser253 and Ser258 in the interaction between filamin A and TF. In vitro binding assays revealed that both the phospho-Ser253 TF peptide and dual phosphorylated TF peptide had the highest binding affinities for the C-terminus repeats 22-24 of filamin A (►Fig. 1A). In contrast, the phospho-Ser258 TF peptide showed a low binding affinity for repeats 22-24 of filamin A, even below that of the non-phosphorylated TF peptide. This indicates that the phosphorylation of Ser253 is essential for the interaction with repeats 22-24 of filamin A and that additional phosphorylation of Ser258 does not interfere with this process per se. However, phosphorylation of Ser258 and the subsequent de-phosphorylation of Ser253 acts as a agonist peptide (PAR2-AP; SLIGKV, 20 μM) for 0 to 60 minutes, fixed and then incubated with anti-TF antibody (HTF1) together with the anti-filamin A antibody EP2405Y. In control reactions, HTF1 and EP2405Y were replaced with mouse and rabbit IgG, respectively. The proximity ligation assays were performed using the DuoLink assay. Nuclei were stained with DAPI (4′,6-diamidino-2-phenylindole) and the actin cytoskeleton was stained using Acti-stain phalloidin-488 (n ¼ 4). (B) HDBEC (human dermal blood endothelial cells) were transfected to express wild-type TF (WT), or TF with alanine (Ala253) or aspartate (Asp253) substitutions of serine 253. Cells were activated with PAR2-AP (20 µM) for 45 minutes and the proximity of TF and filamin A was examined using the PLA (proximity ligation assay) procedure. IgG ¼ TF antibody replaced with mouse control IgG (n ¼ 4, Ã p < 0.05 vs. WT without PAR2 activation. ÃÃ p < 0.01 vs. WT with PAR2-AP, # p < 0.0001 vs. Asp253 with PAR2 activation).
Thrombosis and Haemostasis Vol. 117 No. 11/2017 molecular switch, and therefore the two phosphorylation sites have opposing roles in regulating the binding of TF to the C-terminus of filamin A. Since filamin A is essential for the incorporation of TF into MVs, this may also explain the contrasting effects of Ser253 and Ser258 phosphorylation on TF incorporation into MVs. 8 Because the values for the binding affinities of the TF cytoplasmic domain peptides to filamin A were in the μM range, this low binding affinity and transient nature of the interaction may explain why co-immunoprecipitation of TF and filamin A was not possible. Therefore, PLA were employed to detect TF: filamin A interactions in situ. We used MDA-MB-231 cells throughout the study as they release TF in a filamin A-dependent manner in response to PAR2 activation (►Fig. 2A, B). 9 Increased proximity between TF and filamin A following PAR2 activation was shown to be transient in both MDA-MB-231 cells (►Fig. 3) and endothelial cells (►Supplementary Fig. 5C). The direct interaction of the C-terminus of filamin A with the cytoplasmic domain of TF phosphorylated at both Ser253 and Ser258 has been demonstrated previously. 12 Also, TF has been shown to co-localize with filamin A at the leading edge of spreading epithelial cells 19 and in migrating smooth muscle cells. 20 However, this is the first study to demonstrate the opposing roles of the individual phosphorylation of TF at Ser253 and Ser258 on its interaction with filamin A. The dissimilar binding affinities between filamin A and TF in different phosphorylation states may arise due to structural changes within the cytoplasmic domain of TF which occur following phosphorylation of Ser253 and Ser258, 21 and/or changes in the structure of filamin A following cellular activa-tion. 22,23 Further investigation of the structure of filamin A in complex with the phospho-TF peptides would be required to determine the conformation of filamin A with TF in different phosphorylation states.
Since MVs are derived from the cell surface and incorporate cell surface antigens, the role of filamin A in localizing TF to the cell surface or to specific membrane microdomains for incorporation into MVs was examined. Here we report for the first time that filamin A is required for the increased cell surface TF activity in response to PAR2 activation in these cells, since knock-down of filamin A suppressed this process (►Fig. 4A). In contrast, cell surface TF antigen increased in cells with suppression of filamin A expression, possibly because TF could not be incorporated into MVs and therefore accumulated on the cell surface. Filamin A links receptors to the actin cytoskeleton to stabilize receptors at the cell surface 24,25 and regulate receptor internalization. 26,27 It is possible that filamin A links TF to the actin cytoskeleton to translocate TF to sites of MV release. Furthermore, the interaction between filamin A and TF is unlikely to have a role in the transport of TF from intracellular stores to the cell surface, 28 since suppression of filamin A expression resulted in increased levels of cell surface TF antigen. This is in agreement with an observation by Rothmeier and colleagues that cell surface TF, rather than TF transported from the Golgi, is incorporated into macrophage-derived MVs. 14 It has been reported that TF-positive MVs are derived from lipid rafts. 14,[29][30][31] In agreement with this, we found that cholesterol depletion reduced TF-positive MV release (►Fig. 5A, B). Furthermore, disruption of lipid rafts using MβCD reduced cell surface TF-specific procoagulant activity (►Fig. 5C). This is in agreement with previous studies that have shown that cholesterol depletion using MβCD reduces TF procoagulant activity by lowering the binding affinity of cell surface TF for FVIIa. 15,16 Importantly, cell surface PS levels were not reduced by knock-down of filamin A expression (►Fig. 6), discounting the reduced exposure of PS on the cell surface as the reason for the observed reduction in cell surface TF procoagulant activity on cells with suppression of filamin A expression. Furthermore, MβCD did not alter the effect of filamin A knock-down on TF cell surface activity following PAR2 activation. In fact, MβCD had a similar effect as filamin A knock-down, but there was not any additive or cooperative reduction in TF activity (►Fig. 5C). This indicates that both filamin A knock-down and treatment with MβCD work through a similar mechanism, interfering with the association of TF with cholesterol-rich microdomains necessary for optimal TF activity. Therefore, the interaction with filamin A is likely to be required for the translocation of cell surface TF following PAR2 activation to cholesterol-rich microdomains, which subsequently allows incorporation of active TF into MVs. This is compatible with the concept that transmembrane proteins can translocate to lipid rafts following cellular activation, 32,33 and filamin A has been shown to cluster and retain receptors at the cell surface. 24,25,34 TF is known to be associated within distinct fractions following membrane fractionation by sucrose density gradient ultracentrifugation. 16 Therefore, in future experiments we aim to examine the lipid raft distribution of cell surface TF following suppression of filamin A expression.
In conclusion, this study has, for the first time, demonstrated the opposing roles of Ser253 and Ser258 phosphorylation in the interaction between TF and filamin A. Furthermore, the data suggest the filamin A-dependent translocation of TF into lipid-rich microdomains to be the mechanism by which filamin A mediates the incorporation and release of TF into MVs following PAR2 activation. Disclosures None. Fig. 6 Filamin A knock-down does not reduce cell surface phosphatidylserine. MDA-MB-231 cells were reverse transfected with filamin A or control siRNA (2 nM). Following 48-hour incubation, the cells were adapted to serum-free medium and activated with protease-activated receptor 2 agonist peptide (PAR2-AP; SLIGKV, 20 µM) for 30 minutes. Cells were detached using TrypLE Select, washed and then re-suspended in HEPES (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid) buffered saline (HBS), pH 7.4 containing calcium chloride (2 mM). Negative controls were incubated with HBS lacking calcium chloride. Cells were incubated with 5 µL of annexin V-FITC (BD Biosciences) for 20 minutes on ice. The cells were then examined by flow cytometry using a Gallios flow cytometer. A gate was set at 2% of cells incubated with annexin V-FITC in the absence of calcium chloride (n ¼ 4).

What is known on this topic
• Tissue factor (TF) is released from the cell surface in microvesicles (MVs) following the activation of protease-activated receptor 2 (PAR2). • Phosphorylation of the cytoplasmic domain of TF at Ser253 promotes the incorporation of TF into MVs in response to PAR2 activation, whereas Ser258 phosphorylation suppresses this process. • The cytoskeletal protein filamin A is also required for the incorporation of TF into MVs. Phosphorylation of both Ser253 and Ser258 within the cytoplasmic domain of TF is known to enhance the interaction between TF and filamin A.

What this paper adds
• Phosphorylation of TF at Ser253 alone enhanced TF binding to filamin A, whereas phosphorylation of Ser258 alone reduced binding to filamin A below that of non-phosphorylated TF. • The interaction between TF and filamin A increased following PAR2 activation and occurred at the same time as TF phosphorylation at Ser253 and TF-positive MV release. • Filamin A does not regulate the transport of TF to the cell surface, but it may translocate TF to cholesterol-rich microdomains where TF procoagulant activity is enhanced and where TF is incorporated into and released in MVs.