Mechanical constraint imposed on plasma membrane through transverse phospholipid imbalance induces reversible actin polymerization via phosphoinositide 3-kinase activation

It is well established that the distribution of phospholipids (PL) is asymmetric between the two leaflets of the plasma membrane of mammalian cells(Zachowski, 1993) including platelets (Chap et al., 1977)and erythrocytes (Bretscher,1972; Zwaal et al.,1993). This asymmetry is maintained by the transport of aminophospholipids [phosphatidylserine (PS) and phosphatidylethanolamine (PE)]from the outer to the inner leaflet by an ATP-dependent aminophospholipid translocase, while the choline head PL remains preferentially located in the outer leaflet (Seigneuret and Devaux,1984). Loss of asymmetry is associated with several cellular processes, including cellular activation(Zwaal et al., 1989),apoptosis (Fadok et al., 1992),cellular aging (Connor et al.,1994; Gaffet et al.,1994), and sickle-cell anemia(Franck et al., 1985). During Ca²⁺-ionophore induced platelet activation, a rapid aminophospholipid exposure occurs and many filopodia appear in platelets pre-treated with calpeptin (a permeant inhibitor of calpain, which prevents proteolysis of the actin cytoskeleton)(Gaffet et al., 1995). This filopodia formation seems to be related to an excess of PL in the outer leaflet of the plasma membrane and to cytoskeleton reorganization induced by actin polymerization. In Scott syndrome, where PS exposure is significantly reduced during platelet activation, no filopodia are formed(Bettache et al., 1998). This indicates that filopodia formation tightly correlates with the specific outflux of aminophospholipids resulting in an excess of PL in the outer leaflet of plasma membrane of activated platelets. Thus, an increase in plasma membrane asymmetry could be one of the driving forces for the extension of filopodia and the change in platelet shape during activation. Likewise, the addition of aminophospholipids (PS or PE) increases the surface area of the inner leaflet and leads to an enhancement of endocytosis(Farge, 1995; Farge et al., 1999). The basic idea is that the membrane adapts to a difference in surface area between the two leaflets due to the PL excess in one of the leaflets. Addition of less than 1% of lipids to the external leaflets is sufficient to modify giant liposomes from a spherical or obloid shape into an eight-shape vesicle. Conversely, depletion of a fraction of lipids from the external leaflet (e.g. by removing lyso-PC from the outer leaflet with bovine serum albumin), induces a single invagination which is observed during endocytosis(Farge and Devaux, 1992). In the case of erythrocytes, addition of exogenous PL results in an important modification of cell shape to echynocytic shape(Fujii and Tamura, 1984). With platelets, the effect of excess of PL in the outer leaflet of plasma membrane is characterized by the formation of long filopodia(Suné and Bienvenüe,1988). However, the mechanism by which the filopodia were generated was not elucidated. In this paper, we address the question of the relationship between physical constraint acting on plasma membrane with change in cell shape and cytoskeleton reorganization. We demonstrate that a fraction of PI 3-kinase is reversibly activated by a small excess of PL in the external leaflet of plasma membrane of resting platelets. Evidence is presented that this mechanism also occurs in nucleated cells.

DNA, DNase-I and calcium ionophore A23187 were obtained from Sigma (St Louis, MO) and (³H) 5-hydroxytryptamine [(³H) 5-HT] (8.8 Ci/mmol) was from Amersham Biosciences (Saclay, France). Calpeptin and cytochalasin D were from Novabiochem (Laufelfingen, Switzerland). The spin-labeled PL were synthesized as previously described(Seigneuret and Devaux, 1984). Fluorescent PL were from Avanti Polar Lipids (Alabaster, AL) and DLPC was from Sigma. All other reagents were of the highest grade available commercially.

Blood collected in 0.15 vol. of ACD (85 mM trisodium citrate, 111 mM dextrose, 71 mM citric acid), was obtained from healthy volunteers. Platelets were prepared at room temperature using the erythrocyte cushion procedure, as previously described (Gaffet et al.,1995; Valone et al.,1982). After staining with Plaxan reagent (Sobioda,Montbonnot-Saint-Martin, France), platelets were resuspended in buffer B (96 mM NaCl, 1 mM CaCl₂, 2.7 mM KCl, 2 mM MgCl₂, 5 mM dextrose, 50 mM HEPES, pH 7.4) at 2.2×10⁹ platelets/ml. Platelets were incubated for 30 minutes at 37°C (pH 7.4) before the addition of 1 mM Ca²⁺ (final concentration) and testing. Platelets were activated at 37°C by adding calcium ionophore A23187 (1 μM final concentration) in the presence of 1 mM external Ca²⁺ (pH 6.9).

Platelets at 2.2×10⁹ cells/ml previously pre-treated or not with different inhibitors (cytochalasin D, calpeptin, wortmannin or LY294002) were incubated at 37°C without shaking with a short chain spin-labeled PS or PC analogue at 1% of total PL as indicated(Suné and Bienvenüe,1988). At 2 and 30 minutes, this PL incubation was stopped for further analysis as indicated below.

Fibroblasts treated with nocodazole (2.5 μg/ml) were immediately plated onto polylysine-treated coverslips coated with human serum fibronectin for 3 hours according to Kaverina et al.(Kaverina et al., 1998). After this treatment, fibroblasts were incubated with 50 μM DLPC (corresponding to ∼2% of total phospholipids). At 2 minutes of incubation, fibroblasts were fixed with 2% paraformaldehyde for 30 minutes and observed at light microscopy.

Untreated and nocodazole-treated fiboblasts were incubated, when necessary,with 100 nM wortmannin for 15 minutes and then incubated with 50 μM DLPC for 2 minutes. After 30 minutes fixation with 2% paraformaldehyde, cells were permeabilized with 0.1% Triton X-100 and were stained with phalloidin-TRITC. The fluorescence microscopy observations were carried out with a LEICA DMRA2 microscope.

Platelets were lysed by the addition of an equal volume of buffer containing 2% Triton X-100, 10 mM EGTA, and 100 mM Tris-HCl, pH 7.4. The actin filament content was determined by Dnase I inhibition assay as described previously (Blikstad et al.,1978; Fox et al.,1981).

Platelets were pre-incubated for 30 minutes at 37°C with or without different inhibitors (5 μM cytochalasin D or 100 nM wortmannin, 100μg/ml calpeptin), and then incubated with PL analogues [(0.2)PS or (0.2)PC]or activated by Ca²⁺-ionophore A23187 (1 μM) or by PMA. For SEM,the samples were prepared as indicated(Gaffet et al., 1995). For confocal microscopy, control platelets, platelets incubated with (0.2)PC or(0.2)PS, and platelets pre-treated with calpeptin and then activated with Ca²⁺-ionophore A23187 were fixed with 2% paraformaldehyde for 30 minutes and allowed to deposit on coverslips. The deposited platelets were permeabilized with 0.025% saponin and were stained with Dnase 1 Oregon green and phalloidin-TRITC or with anti-PI-3-kinase (anti-p85, 1:30, Sigma P-8208). Anti-p85α was detected with FITC-conjugated goat anti-rabbit IgG (1 hour at room temperature). The confocal microscopy observations were carried out with a LEICA TCS 4D microscope.

Platelets were labeled with [³²P]orthophosphate (0.4 mCi/ml) for 60 minutes, washed, and resuspended at 1×10⁹ cells/ml. After incubation of platelets with excess PL or with thrombin (without shaking),reactions were stopped at the indicated time by addition of chloroform/methanol (1:1 v/v) containing 0.4 mol/l HCl, and lipids were immediately extracted and analyzed by high-performance liquid chromatography(HPLC) as described (Gratarap et al.,1998).

Platelet secretion was quantified as described previously(Suné and Bienvenüe,1988) using (³H) 5-HT as a marker for dense granules. Radioactivity was measured in the supernatant of the platelet suspension(11,000 g for 3 minutes) and expressed as a percentage of total radioactivity in the suspension. Platelet lysis, as evaluated by lactate dehydrogenase activity (Sigma kit No 500) in the supernatant of activated platelets, did not exceed 4.5%. Platelet activation was carried out either by the Ca²⁺-ionophore A23187 (1 μM for 4 minutes) in the presence of calpeptin (100 μg/ml, 30 minutes pre-incubation time) or by the phorbol 12-myristate 13-acetate (100 nM for 20 minutes). Akt phosphorylation was assessed by western blot using a specific antibody that recognizes Akt only when phosphorylated at Ser473 (pSer473Akt, 1:1000, Cell Signaling Technology,Beverly, MA).

After addition to platelets, any short-chain analogue of PL (spin labeled,fluorescently labeled or di lauroyl) was very rapidly incorporated (<1 minute) into the outer leaflet of the plasma membrane of resting platelets(Suné et al., 1987) and generated an imbalance in the amount of PL between the two leaflets. Albeit the imbalance was small (less than 1% of total PL), it resulted in an increase of the area of the outer leaflet compared to the inner leaflet, generating a physical constraint and the extension of several long filopodia (1-10 μm; Fig. 1). However, platelets were not activated by such addition, since neither increase in the intracellular concentration of calcium nor release of serotonin were detected(data not shown). Furthermore, the filopodia formation induced by PL excess is also observed by scanning electron microscopy in the presence of creatine kinase and phosphocreatine (ADP scavenging system). This observation excludes the possibility of an effect of PL excess on platelet activation induced by ADP release. Moreover, 30 minutes of incubation with excess PL, platelets responded fully to thrombin or Ca²⁺-ionophore activation. Since the addition of excess PL did not induce any `classical' platelet activation, we can assume that the effect is primarily physical and exclude receptor activation such as the Edg (endothelial differentiation genes) receptors,which are activated by sphingosine-1-phosphate and lysophosphatidic acid(Siess, 2002) and which induced irreversible platelet activation. Indeed, short chain PLs interact preferentially with the lipid bilayer of the membrane and not with membrane proteins (Hauser, 2000). In the case of a paramagnetic PS analogue [(0.2)PS], the platelets recovered their initial shape within 20 minutes (Fig. 1c). Conversely, no shape relaxation occurred with the PC analogue[(0.2)PC] (Fig. 1d-e). As expected (Bettache et al.,1998), Ca²⁺-ionophore activation of platelets pre-treated with calpeptin induced changes in shape with filopodia extension(Fig. 1f). In all cases,extension of filopodia induced by the addition of PL analogues was inhibited by cytochalasin D (an inhibitor of actin polymerization)(Fig. 1g-h) but not by nocodazole (a microtubule disrupting agent that did not induce serotonin secretion) (Fig. 1i). Taken together, these data indicated that the actin cytoskeleton rather than the microtubule cytoskeleton is implicated in the formation of filopodia induced by physical constraint.

Observations by confocal microscopy of platelets stained with fluorescently labeled phalloidin and DNase I (labeling F- and G-actin, respectively) showed that G- and F-actin strongly redistributed after PL addition: long filopodia contained mainly F-actin while G-actin was concentrated in the cell body. This demonstrated that actin filaments sustained the filopodia induced by external addition of PL in a manner similar to that described for platelet activation by a Ca²⁺ ionophore (Fig. 2j-l). Finally, similar results were obtained with different short chain PL (spin labeled-, dilauroyl-, or NBD-PC or -PS) demonstrating that the fatty acid composition was not responsible for filopodia formation (not shown).

In a previous paper (Suné and Bienvenüe, 1988), where the shape change was first described,no data were provided to demonstrate the implication of actin cytoskeleton in response to PL excess. Two hypotheses could explain this actin remodeling event occurring in response to excess PL: (1) platelet recruitment of pre-existing actin filaments at some points of the plasma membrane; or (2) de novo actin polymerization from the monomeric actin pool. To distinguish between these hypothesis, we measured the G- and F-actin content of platelets under various conditions. Actually, PL addition clearly triggered actin polymerization, inhibited by cytochalasin D(Fig. 3).

Addition of PS or PC analogues to resting platelets resulted in a rapid and marked increase in F-actin concentrations from ∼50% of the total amount of cell actin in untreated platelets to ∼64% in treated platelets(Fig. 3). When a PS analogue was added to resting platelets, actin polymerization was reversible: the F-actin content raised 2 minutes after PS addition (∼64%) and progressively decreased to the control value (∼51%) 30 minutes later. This reversibility correlated with the translocation of aminophospholipids to the inner leaflet by aminophospholipid translocase(Suné et al., 1987). By contrast, after PC addition, F-actin content remained high (∼63%) during the entire period of the incubation, correlating with the stable insertion of PC in the outer leaflet. The transient 13% increase in actin polymerization induced by excess PL in the outer leaflet was highly significant, and corresponded to about one half of what was observed (21%) in the case of Ca²⁺-ionophore activation of platelets. In all cases, addition of 5μM cytochalasin D, 15 minutes before platelet treatments almost completely abolished actin polymerization response(Fig. 3).

If actin polymerization is essential for membrane extension, the original cause in this study is obviously physical, being strictly dependent on transient PL excess on the external leaflet. However, PL insertion and translocation per se can hardly explain the overall increase in actin polymerization. Therefore, we next aimed to characterize how a physical constraint and/or curvature could activate some signals (such as protein or PL kinase activation) that lead to actin polymerization. A comparable example of physical force is the fluid shear stress, which plays determinant roles in maintaining cardiovascular homeostasis. During this mechanical stimulation,focal adhesion kinase (FAK) (Li et al.,1997) and the mitogen-activated protein kinases, such as extracellular signal-regulated kinase (ERK) and c-Jun N-terminal kinase (JNK)are activated (Jo et al.,1997; Li et al.,1996; Tseng et al.,1995). In addition, it has been recently reported that the phosphoinositide 3-kinase (PI 3-kinase) mediates shear stress-dependent activation of JNK in endothelial cells (Go et al., 1998). Furthermore, PI 3-kinase activity is strongly dependent on membrane curvature(Hübner et al., 1998). Interestingly, it was demonstrated that the enzymatic products of PI 3-kinase mediate platelet actin assembly and filopodial extension after PMA stimulation(Hartwig et al., 1996a). During phagocytosis, PI 3-kinase activity is also reported to be involved in pseudopodial extension (Cox et al.,1999). Since we observed changes in platelet shape with a strong membrane curvature in filopodia emerging in response to excess PL, we tested the possible involvement of PI 3-kinase in this mechanism by using two unrelated and potent PI 3-kinase inhibitors wortmannin and LY294002. Pre-treatment of platelets with wortmannin significantly reduced the number of filopodia induced by excess PL (Fig. 4c). We also tested the effect of wortmannin and LY294002 on actin polymerization. Fig. 4d shows that the increase in F-actin content induced by PL addition was significantly reduced by these two inhibitors. Wortmannin also inhibited actin polymerization of platelets activated by PMA(Fig. 4d) in agreement with results obtained by Hartwig et al.(Hartwig et al., 1996b). Among the various PI 3-kinase products,[³²P]PtdIns(3,4)P₂ was the only one whose content was found to be significantly increased when excess PL was added to the outer leaflet (Fig. 4e). In the same conditions, [³²P]PtdIns(4)P and[³²P]PtdIns(4,5)P₂ did not change significantly(not shown). Pre-incubation of platelets with wortmannin or LY294002, prior to adding PL, completely inhibited the increase of PtdIns(3,4)P₂.

PI 3-kinase involvement in actin polymerization and filopodia formation was further assessed by (1) the membrane recruitment of PI 3-kinase in response to the PL excess; and (2) the phosphorylation of Akt, a downstream protein kinase target of activated PI 3-kinase (Franke et al., 1997). During the physical constraint applied by PL excess,PI 3-kinase was translocated from punctuate sites in the cytoplasm to the plasma membrane (Fig. 5A). As far as the constraint persisted, PI 3-kinase remained localized to the plasma membrane (Fig. 5A, PC 1min, PC 30min, PS 1min). PI 3-kinase translocation from the cytoplasm to the plasma membrane was already also observed in various cells in response to growth factor receptor signaling (Gillham et al.,1999). However, when the PL excess disappeared, the distribution of PI 3-kinase reversed to a cytoplasmic localization indistinguishable to that observed in control platelets (Fig. 5A, PS 30min, Control). Concomitantly, the PL excess induced a net increase in Akt phosphorylation that was abolished by the presence of PI 3-kinase inhibitors and reversed when PL excess was nullified(Fig. 5B,C). Akt was previously demonstrated to be a required intermediate between PI 3-kinase activation and filopodia extension in Dictyostelium discoideum(Meili et al., 1999) and in the formation of stress fibers in endothelial cells(Morales-Ruiz et al.,2000).

We next sought to determine how much of the membrane-cytoskeleton reactivity that observed in platelets following a mechanical constraint was generalized to other cellular systems, notably nucleated cells. Addition of DLPC to untreated L929 fibroblasts induced no gross morphological change (not shown). However, when L929 cells were pre-treated with nocodazole in order to disrupt the microtubule cytoskeleton (Fig. 6b), subsequent addition of 50 μM DLPC resulted in drastic membrane extensions (lamellipodia and short filopodia)(Fig. 6c). Therefore, under these specific conditions, as observed with platelets, a similar membrane reorganization was induced by phospholipid imbalance in nucleated cells.

Microscopic observations of L929 fibroblasts stained with fluorescently labeled phalloidin showed that F-actin was enriched in membrane extensions when nocodazole-treated cells were incubated with a phospholipid excess(Fig. 7c). This drastic enrichment in actin was not observed in untreated cells(Fig. 7a) or in cells uniquely treated with nocodazole (Fig. 7b). As in platelets, pre-incubation of nocodazole-treated cells by wortmannin abrogated the effect of phospholipid addition on F-actin recruitment (Fig. 7d).

It is now well established that an excess of PL in the outer leaflet of erythrocytes causes the formation of spicules(Fujii and Tamura, 1984). The mechanism of this change in shape may be explained by the `bilayer couple'hypothesis (Sheetz and Singer,1974). According to this model, a surface expansion of one of the two membrane leaflets due to PL incorporation induced a physical constraint as a result of differential tension, which can be accommodated by a membrane curvature with its convexity turned to the side containing the PL excess. Differential tension and membrane curvature are indissociable processes. Membrane tension was known to be related to mechanosensitive receptor activation, leading to weakly specific ion channel opening(Hamill and Martinac, 2001)and local ionic composition changes. It cannot be excluded that this change could induce actin polymerization, but it seems that filopodia growth was never associated with membrane tension induced by shear or mechanical stresses. Thus, we infer that curvature was probably the main cause of the effects described in this paper. In erythrocytes(Franck et al., 1985), the length of spicule extensions seems to be limited by the spectrin network underlying the cytosolic face of the membrane. Here, a PL imbalance between the two leaflets of platelet plasma membrane generated highly curved structures stabilized as very long (1-10 μm) and very thin (∼200 nm diameter) filopodia by de novo actin polymerization.

The emergence of filopodia extension and actin polymerization may be rationalized by a generalization of the stochastic theory(van Oudenaarden and Theriot,1999). According to this theory, owing to the actin network dynamics, a bead symmetrically surrounded by actin can move in one direction when the symmetry of the system is broken by some random process, leading to a polarization of the actin mantle and formation of an actin tail, a biological driving force capable of pushing the bead ahead. In this context, the shape of a resting state platelet would be the result of the dynamic equilibrium between actin polymerization (controlled by actin and accessory protein concentrations) and the applied physical constraint. When the transverse phospholipid content is unbalanced by external addition of a minimal amount of PL, some small exvaginations occurred that could become nucleation points for fast actin polymerization by elongation of initially small filaments at their barbed end, known to be in proximity of the membrane surface(Wilkins and Lin, 1981). At these points, the equilibrium between the two forces is displaced, leading to elongation until the external excess in PL is accommodated by these highly curved structures.

Although the precise mechanism governing the changes in cell shape and actin polymerization that are induced by membrane curvature is not known, our results indicate that biochemical signaling pathways are triggered by this initial physical event. Through direct binding to specific protein domains the products of PI 3-kinase may recruit and modulate the activity of several proteins involved in the control of cell shape and actin cytoskeleton reorganization. Many PI 3-kinase products as well as PtdIns(4,5)P₂ act as regulators of actin polymerization(1) by uncapping the barbed end of microfilaments from gelsolin(Janmey and Stossel, 1987),(2) by dissociating the profilactin complex providing a fresh pool of monomeric actin (Lassing and Lindberg,1985), (3) by associating with α-actinin(Fukami et al., 1992) and (4)by strengthening the cytoskeleton-membrane interactions(Raucher et al., 2000).

In the case of platelets, a striking observation is that this process is fully reversible, since filopodia formation, increased actin polymerization,translocation of PI 3-kinase to the plasma membrane, and Akt phosphorylation disappeared when the imbalance was neutralized by specific aminophospholipid transport. PL imbalance creates a curvature in the plasma membrane and induces actin polymerization. Actin polymerization pushes forwards to form long filopodia until a new equilibrium occurs between F-actin pressure and membrane tension induced by transverse phospholipid imbalance. The change in F-/G-actin ratio appears to be modulated by PI 3-kinase activation, known to be related to membrane curvature (Hübner et al.,1998). Subsequently, the aminophospholipid translocase in platelets can equilibrate the two membrane leaflets, allowing filopodia to retract and actin to depolymerize. Similar dynamic changes in shape are known to be induced by upstream signals involving rac and cdc42 activation(Nobes and Hall, 1995).

The physical and mechanical constraint exerted on membrane through experimental phospholipid excess is not limited to the sole platelet model,where it is sometimes difficult to assess the actual state of activation of the cellular model. Here, it is important to point out that the PL excess provided by aminophospholipid addition induced a reversible phenotypic change in shape that cannot be explained by standard platelet activation. It was,however, very important to reproduce these phenotypic changes in a different cell model, notably within nucleated cells. In order to do so, we have had to simplify the nucleated cell model by slowing down the intense membrane dynamics that are normally recorded at the surface of dynamic fibroblasts. This was performed by disrupting the microtubule network in L929 cells with nocodazole, leaving the actin cytoskeleton functional. Under those experimental conditions, we were able to trigger major membrane extensions(lamellipodia together with short filopodia) induced by PL imbalance in L929 fibroblasts. As previously observed in platelets, the membrane extensions in fibroblasts were also actin- and PI 3-kinase-dependent, supporting a generalized concept linking physical membrane constraint with a signaling pathway leading to actin polymerization via PI 3-kinase activation. How the physical signal (i.e. the introduction of a few phospholipids to the outer phospholipid leaflet of the membrane) triggers the signaling cascade remains unknown. When growth factors activate their receptors, phosphorylated tyrosine residues on the latter are sites of PI 3-kinase recruitment(Bjorge et al., 1990). What kind of receptors are activated when excess phospholipids are inserted in the outer leaflet of the membrane remains to be determined.

Recent work (Funamoto et al.,2002; Iijima and Devreotes,2002) has demonstrated that PI 3-kinase is recruited and activated in the leading edge of cells exposed to chemoattractants gradients. The localization of PI 3-kinase could explain the generation of 3-phosphoinositides in the leading edge and the reversibility of these events may be achieved by PTEN (3-phosphoinositide phosphatase), the antagonist of PI 3-kinase.

The effect of lipid addition was also observed in NIH 3T3 fibroblasts. In these cells, the lipid excess modulates cell spreading and lamellipodial extension (Raucher and Sheetz,2000). In that study (Raucher and Sheetz, 2000), the concentration of the lipid added (2 mM) was much higher than the one used in the experiments reported here (8 μM for platelets and 50 μM for fibroblasts). However, both studies clearly show that physical constraint and/or membrane tension modulates cellular dynamics. Nevertheless, one should keep in mind that the mechanism induced by lipid addition may be subtly different depending on the cell type characteristics.

In conclusion, although the effect of tension could not be completely excluded, the data presented here indicate that cells may respond to physical forces exerted by membrane curvature, a form of mechanical stress, by activation of the PI 3-kinase/Akt signaling pathway, followed by actin polymerization and change in cell shape.

We are grateful to Patrick Carroll for carefully reading this manuscript and Brigitte Nguyen-Dao for technical assistance.