Role of Cofilin in Epidermal Growth Factor–Stimulated Actin Polymerization and Lamellipod Protrusion

Cell motility plays an important role in many basic biological processes, including embryogenesis, neurite growth, wound healing, inflammation, and cancer metastasis. Motility of crawling cells is dependent on the ability to extend F-actin–rich protrusions, usually in the form of lamellipods (Abercrombie et al. 1970; Chen et al. 1994; Verschueren et al. 1994; Xie et al. 1995). Protrusion of such actin-rich lamellipods in moving cells requires cycles of actin polymerization and depolymerization (actin polymerization transients) (Lauffenburger and Horwitz 1996; Mitchison and Cramer 1996; Bailly et al. 1998a; Condeelis 1998).

Previous studies have demonstrated the requirement for free barbed ends in the control of this cycle (Handel et al. 1990; Symons and Mitchison 1991; Chan et al. 1998). We have shown previously that stimulation of metastatic MTLn3 cells with EGF causes a transient increase in actin nucleation activity resulting from the appearance of free barbed ends at the extreme leading edge of extending lamellipods (Chan et al. 1998). This increase in actin nucleation activity is associated with an increase in the number of filaments throughout the leading edge (Bailly et al. 1999; Chan et al. 1998). Coincident with the increase in barbed and pointed ends, a general decrease in filament length with a specific loss of long filaments is observed at the leading edge of EGF-stimulated MTLn3 cells (Bailly et al. 1999).

The molecular mechanism(s) for the appearance of free barbed ends at the leading edge is still unclear. The currently popular models propose that free barbed ends arise by either (A) uncapping of barbed ends, (B) severing of noncovalent bonds in F-actin to produce short filaments with free barbed ends, or (C) de novo nucleation of filaments from a nucleation template.

There is evidence both for and against model A. Studies in platelets suggest that uncapping of barbed ends by capping protein could be a major mechanism for the production of free barbed ends in response to thrombin stimulation (Hartwig et al. 1995). However, similar studies in Dictyostelium rule out uncapping by capping protein as a mechanism for the production of free barbed ends during chemotactic stimulation (Eddy et al. 1997). Furthermore, the increase in barbed end number after stimulation of tumor cells with EGF is accompanied by the shortening of actin filaments (Bailly et al. 1999), which is inconsistent with an uncapping mechanism. In support of models B and C, the decrease in filament length that accompanies the increase in filament number after EGF stimulation suggests that either a severing activity is switched on at the leading edge, resulting in the generation of short filaments with free barbed and pointed ends, or that de novo nucleation from a template like the actin-related protein (Arp) 2/3 complex has occurred, resulting in the appearance of free barbed ends. In vitro studies have shown that cofilin increases the number of free barbed ends by severing filaments, thereby increasing the rate of actin polymerization (Du and Freiden 1998; Maciver et al. 1998; Ichetovkin et al. 2000). Model C has received more attention recently with the discovery of the Arp2/3 complex and its ability to nucleate actin polymerization through regulation by members of the Wiskott-Aldrich syndrome protein (WASP) family of proteins (Mullins et al. 1998; Egile et al. 1999; Machesky et al. 1999; Rohatgi et al. 1999; Welch 1999).

The ADF/cofilin (cofilin) family of proteins is capable of binding to both G- and F-actin to increase the rate of depolymerization of actin filaments (for review see Bamburg 1999; Bamburg et al. 1999). The mechanism(s) by which depolymerization is enhanced involves both filament severing to increase the number of depolymerizing ends (Du and Freiden 1998; Maciver et al. 1998; Ichetovkin et al. 2000), and a cofilin-induced increase in the off-rate from the pointed ends (Carlier et al. 1997), potentially through a change in the twist in the actin filaments (McGough et al. 1997; Bamburg 1999; Bamburg et al. 1999). Cofilin can also increase the number of free barbed ends by severing filaments, thereby increasing the rate of actin polymerization (Du and Freiden 1998; Maciver et al. 1998; Ichetovkin et al. 2000). Hence, cofilin is capable of increasing the rate of actin polymerization, depolymerization, and the number of barbed ends in vitro. Translocation of cofilin to the plasma membrane where barbed ends appear after stimulation has been shown in studies using HL-60 cells (Suzuki et al. 1995), neutrophils (Djafarzadeh and Niggli 1997; Heyworth et al. 1997), and in flattened Dictyostelium under starvation stress (Aizawa et al. 1995), suggesting that cofilin may participate in barbed end mobilization at the membrane.

To investigate the role of cofilin in EGF-stimulated actin polymerization in MTLn3 cells, we examined in detail the temporal and spatial distribution of cofilin relative to free barbed ends, and characterized the actin dynamics by measuring changes in the number of filaments using two different techniques. We used a light microscope severing assay to determine if cofilin severing activity is turned on during EGF stimulation of MTLn3 cells. Furthermore, function-blocking antibodies that inhibit cofilin's severing activity were used to assess the role of cofilin in vivo. Our results indicate that cofilin plays a direct role in the mobilization of barbed ends in the leading edge and is necessary for the protrusion of lamellipods in response to EGF stimulation.

MTLn3 cells were grown in α-MEM (Gibco Laboratories) supplemented with 5% FCS, as described previously (Segall et al. 1996; Bailly et al. 1998b).

Inhibitory (Ab286) and noninhibitory (Ab287) antibodies were generated against rat cofilin peptides containing amino acids 104–117 and 57–71, respectively. Based on comparison with the atomic structure of cofilin (Fedorov et al. 1997), the peptide 104–117 (WAPESAPLKSKMIY) contains the sequence identified as both the actin-binding and phosphatidylinositol 4,5-bisphosphate (PIP2)-binding site on cofilin (Yonezawa et al. 1991). Antibodies generated against this sequence should block the interaction between cofilin and actin. On the other hand, the peptide 57–71 (VGDVGQTVDDAYTTF) is not at the actin–cofilin interface, and thus represented a candidate sequence for production of an antibody (Ab287) that would not block the interaction between actin and cofilin. The peptides were synthesized as a multiple antigen peptide system (MAPS) (Posnett et al. 1988; Posnett and Tam 1989) containing four branches, and injected into rabbits without further coupling (Covance). Immunoreactive sera were subsequently affinity purified on the same peptides immobilized on Sepharose matrices.

Specificity of cofilin antibodies was assessed by Western blot (Ab287) and immunoprecipitation analysis (Ab286, Ab287) using MTLn3 cell lysates and purified recombinant rat cofilin as a positive control. The antibody Ab287 recognizes a single polypeptide of 21 kD in Western blots of MTLn3 cell lysates and purified rat cofilin. In addition, both antibodies (Ab286, Ab287) precipitate from whole cell homogenates of MTLn3 cells a single 21-kD polypeptide that binds anticofilin (data not shown).

MTLn3 cells were plated on Mattek dishes as described previously (Chan et al. 1998), serum starved for 3 h in α-MEM supplemented with 0.35% BSA and 12 mM Hepes, pH 7.5 (starvation medium), and stimulated with a final concentration of 5 nM EGF in starvation medium, or left untreated. Cells were fixed with 3.0% formaldehyde, 0.1% (vol/vol) glutaraldehyde, and 0.02% NaN₃ in PBS, pH 7.35, (145 mM NaCl, 4 mM NaH₂PO₄, 6 mM Na₂HPO₄) for 1 h, washed four times in PBS, and permeabilized with 0.5% Triton X-100 in PBS for 15 min. They were then washed three times in PBS, once in TBS (20 mM Tris, 154 mM NaCl, pH 8), and rinsed twice with NaBH₄ in TBS for 15 min to reduce glutaraldehyde autofluorescence. The preparations were then washed four times, and blocked and stabilized by incubation for 20 min with 5 μM unlabeled phalloidin (Sigma Chemical Co.) in TBS supplemented with 1% BSA and 1% FCS. Cells were further incubated with anticofilin antibodies (Ab287, 5 μg/ml), followed by FITC-conjugated anti–rabbit antibodies (Cappel Laboratories), and mounted as described previously (Bailly et al. 1999). For F-actin visualization in the same cells, 0.5 μM rhodamine-labeled phalloidin (Molecular Probes) was used during the blocking step.

Nucleation sites (barbed ends) were localized using a previously described protocol (Chan et al. 1998; Bailly et al. 1999) with slight modifications. In brief, the cells were permeabilized for 1 min in presence of 0.45 μM of G-actin (either rhodamine-labeled or biotin-labeled) in buffer C (138 mM KCl, 10 mM Pipes, 0.1 mM ATP, 3 mM EGTA, 4 mM MgCl₂, pH 6.9) with 1% BSA and 0.025% saponin. After fixation and subsequent washes, rhodamine-actin incorporation was visualized directly on the fluorescence microscope, while biotin-actin distribution was identified using Cy3-coupled antibiotin antibodies (Sigma Chemical Co.). Previous work has shown that both techniques gave similar results in terms of localization and quantitation of free barbed ends (Bailly et al. 1999).

The number of free pointed ends was determined using the DNase I inhibition assay as described previously (Podolski and Steck 1990; Eddy et al. 1997). In brief, MTLn3 cells were starved for 3 h and detached from plates (Chan et al. 1998), and concentrated to 2 × 10⁷ cells/ml in starvation buffer. At various times after stimulation, the cells were lysed by adding an equal volume of L buffer (10 mM imidazole-acetate, 3.3 mM Tris-acetate, pH 7.5, 50 mM KCl, 2.2 mM Mg-acetate, 4 mM EGTA, 0.55 mg/ml BSA, 0.5 mM ATP, 1 μg/ml each of leupeptin, pepstatin, and chymostatin, 0.25% Triton X-100, and 20 μM phalloidin). Lysates were vortexed for 2 s and immediately microfuged at 4°C for 1 min at 8,700 g. The low-speed Triton-insoluble cytoskeletal pellet was resuspended on ice in 10% lysate volume of L buffer and incubated at 37°C for 30 min to polymerize all residual G-actin into the cytoskeleton. The sample was diluted to 3.33 × 10⁶ cell equivalents/ml in L buffer and incubated with 8.8 nM DNase I for 17 min at room temperature. [³H]DNA (0.011 μg/μl) was then added to the mixture for another 17 min. The reaction was terminated and unhydrolyzed DNA was precipitated by the addition of ice-cold 10% TCA; the radioactivity in the supernatant was analyzed. The number of filament pointed ends was calculated by fitting experimental cpm values to a rabbit G-actin standard curve at a fixed DNase I concentration.

Although DNase I can be used for the staining of either G- or F-actin, the specificity of DNase I staining is determined by the buffer conditions. For staining of G-actin, the G-actin pool has to be stabilized first by 10% ice-cold glycerol and MgCl₂ before application of DNase I (Knowles and McCulloch 1992). In our study, we used DNase I to stain the pointed ends of F-actin. Thus, the cells were permeabilized in a glycerol-free saponin extraction buffer, and washed extensively in an F-actin stabilization buffer to remove all traces of G-actin before the application of DNase I. The F-actin pool was selectively stabilized by 20 μM phalloidin and 0.5 μM cytochalasin D to prevent depolymerization.

Pointed ends were visualized using immunofluorescence with antibodies against DNase I to detect incorporation of exogenous DNase I in permeabilized MTLn3 cells. In brief, MTLn3 cells grown on Mattek dishes were starved for 3 h in starvation buffer, stimulated with EGF, and permeabilized for 30 s with buffer B (5 mM KCl, 137 mM NaCl, 4 mM NaHCO₃, 0.4 mM KH₂PO₄, 1.1 mM Na₂HPO₄, 2 mM MgCl₂, 5 mM Pipes, and 5.5 mM glucose, pH 7.2) containing 0.025% saponin, 20 μM phalloidin, and 4 mM EGTA. After washes with buffer B containing 20 μM phalloidin and 4 mM EGTA to remove all G-actin, cells were incubated with 1 μM DNase I (diluted in 10 mM imidazole-acetate, pH 7.5, 3.25 mM Tris-acetate, 51 mM KCl, 2.24 mM Mg-acetate, 0.54 mM BSA, and 4.93 mM Mn-acetate) and 0.5 μM cytochalasin D for 17 min. After a brief wash with buffer B containing 20 μM phalloidin and 4 mM EGTA to remove DNase I not bound to the actin cytoskeleton, cells were fixed with 3.7% formaldehyde in buffer B for 5 min, followed by a 10-min incubation with 0.1 M glycine in buffer B. Further steps were carried out as described for cofilin immunofluorescence.

The severing of actin filaments by cell lysates prepared from MTLn3 cells was observed microscopically in a flow cell coated with 0.2% nitrocellulose in amylacetate. After a 1-h incubation with nitrocellulose, antibiotin antibodies diluted in ISAP buffer (20 mM Tris-HCl, pH 7.5, 5 mM EGTA, 2 mM MgCl₂, 50 mM KCl, 1 mM ATP, 1 mM DTT) were perfused into the chamber for 5 min. The remaining binding sites were blocked with ISAP buffer containing 0.5 mg/ml BSA. Rhodamine and biotin-labeled F-actin were prepared by incubating 0.4 μM rhodamine-labeled actin, 0.2 μM biotin-labeled actin, and 1.4 μM unlabeled actin in ISAP buffer and 0.2 μM phalloidin for no more than 1.5 h. The flow cell was filled with 0.04 μM rhodamine/biotin-labeled F-actin in antibleaching buffer (ISAP buffer containing 5 mg/ml BSA, 0.036 mg/ml catalase, 0.02 mg/ml glucose oxidase, 6 mg/ml glucose, 100 mM DTT, 1 mM ATP) for 5 min. Unbound F-actin was washed away with antibleaching buffer and severing was initiated by replacement of the flow cell solution with cell lysate prepared as described below.

MTLn3 cell lysates were prepared by lysing one part of cell suspension (8 × 10⁶ cells/ml) in 12 parts of lysis buffer (20 mM Tris-HCl, pH 7.5, 5 mM EGTA, 0.5 mM MgCl₂, 0.5% Triton X-100, 0.5 mM ATP, 5 mg/ml BSA, 0.036 mg/ml catalase, 0.02 mg/ml glucose oxidase, 6 mg/ml glucose, 100 mM DTT, and protease inhibitors) at the times indicated after stimulation of cells with EGF. For antibody experiments, cell lysates were incubated with a molar excess of anticofilin or nonimmune IgG for 1 h on ice. The relative number of filaments was calculated as the ratio of the number of filaments at 20 min divided by the number of filaments 30 s after perfusion in the same field. The first measurement was made after perfusion so that filament breakage caused by perfusion would not be scored. Filament breakage was never observed due to perfusion of lysis buffer alone.

Cells were grown on Mattek dishes as described above and placed in L15 medium (Life Technologies) with 0.35% BSA for 1–2 h before injection. Microinjection was conducted using an Eppendorf semiautomated microinjection system using needles pulled on a Sutter p87 micropipette puller. Antibodies (nonimmune rabbit IgG or Ab286 anticofilin at 6 mg/ml in PBS) were mixed with FITC-labeled dextran (0.5–1 mg/ml final concentration; Molecular Probes). Cells were allowed to recover for 1–2 h before further manipulations. For time lapse video microscopy, injected cells were identified directly using the FITC fluorescence of the injected dextran. For F-actin quantitation and barbed end localization, the injected cells were further localized after fixation using FITC-conjugated anti–rabbit antibodies (Cappel Laboratories). From the settings used for microinjection, we estimated a volume of injection ranging between 1 and 10% of the total cell volume, based on calibration of this microinjection system as reported elsewhere (Minaschek et al. 1989). This would deliver between 0.8 and 8 μM of anticofilin Fab binding sites per cell. Since the concentration of cofilin in MTLn3 cells is 6.0 ± 0.5 μM (data not shown), we calculated that under these conditions, the microinjection delivered between 0.13 and 1.33 anticofilin Fab binding sites per molecule of cofilin.

Protrusive activity after EGF stimulation was recorded using time lapse video microscopy and quantified as an increase in total cell area as described previously (Segall et al. 1996; Bailly et al. 1998b).

Images were taken using constant settings on an Olympus I ×70 microscope with 60× NA 1.4 infinity-corrected optics coupled to a computer-driven cooled CCD camera using IPLab Spectrum software (Vaytek). The digitized images were converted linearly in NIH Image and analyzed using a program described previously (Bailly et al. 1999). Fluorescence in the cell cortex due to cofilin, rhodamine-actin, or DNase I staining was expressed as the mean pixel intensity within 1.1 μm of the cell membrane and covering the whole cell perimeter. The intensity of DNase I staining in stress fibers was quantitated as relative pixel intensity in random spots along stress fibers. The percentage of cofilin at the leading edge was calculated as the intensity of cofilin at the leading edge divided by the total cofilin intensity of the whole cell.

For F-actin quantification, the relative amounts of F-actin present in the cells were evaluated using rhodamine-phalloidin (Molecular Probes) staining as described previously (Bailly et al. 1999). Quantitation of fluorescence intensity was done as described above, using a derivative of the same program to obtain the cell area and the mean intensity of the whole cell.

To study the relationship between cofilin and the appearance of free barbed ends during stimulated cell motility, the distribution of cofilin in MTLn3 cells was examined at various times after EGF stimulation. To localize free barbed ends, cells were permeabilized in detergent containing rhodamine-actin at concentrations sufficient to support only barbed end polymerization (Chan et al. 1998). However, under these conditions, we found that permeabilization before fixation extracted cofilin from the leading edge. The extent of this extraction is shown in Fig. 1e and Fig. f (see also Fig. 3). To preserve the cofilin distribution in MTLn3 cells, it was necessary to fix cells with a mixture of formaldehyde and glutaraldehyde for 1 h before detergent extraction (Bamburg and Bray 1987). Although this prevented the localization of cofilin and barbed ends in the same cell, it permitted the localization of cofilin in cells at various times after stimulation without the artefact of cofilin extraction. In resting MTLn3 cells, cofilin is distributed diffusely in the cytoplasm (Fig. 1 b). Upon EGF stimulation, cofilin is recruited to the leading edge (Fig. 1 d), colocalizing with F-actin (Fig. 2).

To compare the distribution of cofilin relative to free barbed ends, parallel cultures were prepared to detect free barbed ends using rhodamine-actin incorporation (Chan et al. 1998), and cofilin using anticofilin antibody (Ab287). Quantitative analysis of the distribution of cofilin and free barbed ends in lamellipods shows overlapping localization, i.e., both localize very close to the plasma membrane (Fig. 3). The percentage of total cellular cofilin present in the leading edge increases from 4% in resting cells to 8% after EGF stimulation. In kinetic studies (Fig. 4), localization of cofilin to the leading edge peaks at 50 s after EGF stimulation, corresponding with the peak of barbed end appearance in the leading edge after stimulation. Concomitantly, accumulation of F-actin in the leading edge reaches a maximum by 60 s after stimulation (Chan et al. 1998).

The results of immunofluorescence and kinetic studies shown here, combined with previous reports showing that cofilin can increase the number of free barbed ends in vitro (Du and Freiden 1998; Maciver et al. 1998; Ichetovkin et al., 1999), suggest that cofilin could be associated with the production of free barbed ends as they appear after EGF stimulation. If cofilin-mediated severing of preexisting actin filaments occurs transiently and locally at the leading edge after EGF stimulation, free barbed and pointed ends would be exposed and the number of actin filaments should increase in the leading edge. We investigated this possibility by quantitating the number of pointed ends of actin filaments in low-speed Triton-insoluble cytoskeletons using a DNase I binding assay. DNase I binds the pointed ends of actin filaments with high affinity, and can thus be used to quantify directly the number of pointed ends in cells (Podolski and Steck 1990; Eddy et al. 1997). As shown in Fig. 5, the number of filaments quantitated in low-speed Triton-insoluble cytoskeletons increases by 1.5-fold, which corresponds to a net increase of 6 × 10⁴ pointed ends per cell within 50 s after EGF stimulation.

To identify where these new pointed ends are localized during stimulation, we bound DNase I to saponin-permeabilized cells, then immunostained for incorporation of exogenous DNase I with anti–DNase I IgG. As shown in Fig. 6 A, in resting cells DNase I localizes to DNA in the nucleus and to actin filaments in stress fibers (Fig. 6 A, arrow) as well as in the cell cortex. After EGF stimulation, the DNase I staining pattern is identical to that in the resting cells, except that the filaments in the cell cortex (Fig. 6 A, arrowheads) are stained at much higher intensity, indicating an increase in the number of pointed ends near the plasma membrane. Fig. 6 B shows the leading edge of cells like those in Fig. 6 A at higher magnification, clearly demonstrating the increase of pointed end staining at the leading edge (Fig. 6 B, arrowhead) of EGF-stimulated cells (EGF 50 sec) compared with that of resting cells (EGF 0).

Quantitation of the DNase I binding by measuring pixel intensity (Fig. 7) demonstrates a 1.5-fold increase in the number of pointed ends at the leading edge compartment, but not in stress fibers, after EGF stimulation. In control experiments, where DNase I binding to the pointed end of F-actin was blocked by pretreatment with equal molar amounts of G-actin, cytoplasmic binding sites for DNase I are blocked, indicating that DNase I is binding to the pointed ends of actin filaments in the cytoplasm (Fig. 7, inset). These results are consistent with our previously reported twofold increase after EGF stimulation in pointed end incorporation of 2 μM rhodamine-actin at the leading edge in gelsolin-capped cytoskeletons (Chan et al. 1998). They are also consistent with the 1.5-fold increase in filament density measured after stimulation at the leading edge using EM (Bailly et al. 1999). The fact that a 1.5–2-fold increase in pointed ends at the leading edge of stimulated cells is measured using three different methods (Chan et al. 1998; Bailly et al. 1999) (Fig. 5 and Fig. 7) indicates that the number of DNase I molecules bound is a direct measurement of the number of filaments present in the cytoplasm. These results also indicate that the number of filaments is increasing at the leading edge after stimulation.

To determine if severing could contribute to the increase in filament number, a light microscope severing assay was used to determine if a cofilin-mediated severing activity is turned on during EGF stimulation of MTLn3 cells. This assay allows the experimenter to visualize the cutting of filaments directly in the light microscope using fluorescent F-actin. As shown in Fig. 8 A, only a small amount of filament severing is observed when filaments are perfused with cell lysates prepared from resting cells. However, when filaments are perfused with cell lysate from EGF-stimulated cells, severing activity is increased dramatically.

To identify the type of severing protein involved in this stimulated severing activity, we took advantage of a difference between the severing activities of the gelsolin and cofilin families of severing proteins. Phalloidin inhibits severing of cofilin-like proteins by competing with cofilin for binding to actin filaments, but does not block the severing activity of gelsolin-like proteins. We confirmed this by measuring severing of phalloidin-saturated filaments with either gelsolin or cofilin. Only gelsolin severed filaments under these conditions (data not shown). Next, we measured the ability of the severing activity in cell extracts to sever phalloidin-saturated F-actin. As shown in Fig. 8 A, saturation of actin filaments with phalloidin inhibits the severing activity present in extracts from stimulated cells, indicating that a cofilin-mediated severing reaction operates in the EGF-stimulated cells. The severing activity present in MTLn3 cell lysates was 1.3-fold in resting lysates compared with buffer, and increased to 1.9-fold in lysates from EGF-stimulated cells (Fig. 8 B). The EGF-stimulated severing activity decreases to background levels when the actin filaments are saturated with phalloidin.

To investigate if the severing activity in cell lysates is specifically due to cofilin, we characterized a monospecific polyclonal antibody (Ab286) that was designed to recognize the actin binding site on cofilin (amino acids 104–117) and block cofilin's severing activity. Affinity purified Ab286 was added to purified rat recombinant cofilin and cofilin's severing activity in the light microscope assay was compared with that in the presence of nonimmune IgG. As shown in Fig. 9 A, addition of a molar excess (30:1) of affinity purified Ab286, but not nonimmune IgG, to purified recombinant rat cofilin inhibits its severing activity.

Ab286 was then used as a function-blocking antibody to inhibit cofilin's severing activity in cell extracts. As shown in Fig. 9 B, addition of Ab286 (but not nonimmune IgG) to cell extracts completely inhibited severing activity in extracts from both resting and stimulated cells. Therefore, cofilin is responsible for the severing activity detected in resting and stimulated cell lysates.

To investigate the function of cofilin in vivo, antibody Ab286, which inhibits the severing activity of pure recombinant cofilin as well as cofilin in cell extracts (see above), was microinjected into MTLn3 cells. Microinjection of these function-blocking antibodies against cofilin did not alter cell morphology for at least up to 2 h after injection (in particular, cell spreading as evaluated by area measurements was unaffected [Table]). However, the ability of cells to extend lamellipods after EGF stimulation was strongly inhibited (Fig. 10). On the other hand, control cells injected with comparable amounts of nonimmune rabbit IgG extended lamellipods after stimulation with EGF with the expected kinetics and amplitude (Segall et al. 1996; Bailly et al. 1998b).

The ability of function-blocking antibodies against cofilin to inhibit lamellipod extension suggests that cofilin activity is required for actin polymerization at the leading edge after EGF stimulation, since actin polymerization is known to be required for lamellipod extension. However, these results did not distinguish how cofilin is functioning to affect actin polymerization at the leading edge. Cofilin could affect polymerization either: (a) indirectly, by increasing the rate of filament turnover to supply monomers for polymerization (Carlier et al. 1997; Ressad et al. 1999; Svitkina and Borisy 1999); or (b) directly, by severing filaments to generate new barbed ends used to nucleate actin polymerization (Du and Freiden 1998; Maciver et al. 1998; Ichetovkin et al., 1999). To distinguish these possibilities, we measured the ability of cells microinjected with function-blocking antibodies against cofilin to mobilize barbed ends after EGF stimulation.

Free barbed ends can be localized and quantified directly in cells by light permeabilization in the presence of concentrations of labeled G-actin sufficient to polymerize onto barbed but not pointed ends (Symons and Mitchison 1991; Chan et al. 1998; Bailly et al. 1999). We have shown previously that the kinetics of barbed end appearance after stimulation are tightly controlled, with a maximum at 50–60 s after stimulation with EGF (Fig. 4; Bailly et al. 1999). We thus analyzed barbed end distribution in cells microinjected with function-blocking Ab286 or control nonimmune IgG at peak stimulation (stimulation with EGF for 1 min). As shown in Fig. 11 B, cells microinjected with control antibodies did not display any alteration in the distribution of barbed ends after stimulation, with typical incorporation of labeled actin at the leading edge of the cells (Chan et al. 1998; Bailly et al. 1999). However, a large proportion of the cells injected with function-blocking antibodies against cofilin totally failed to generate barbed ends in the leading edge after EGF stimulation (Fig. 11A and A′). Indeed, the proportion of cells that did not show any labeled actin incorporation at the leading edge after stimulation more than doubled in cells microinjected with the anticofilin compared with control cells injected with nonimmune IgG (Fig. 11 C). Moreover, this number is likely to be an underestimate of the inhibition, since only cells with complete inhibition of barbed ends at the leading edge (typical examples shown in Fig. 11A and A′) were scored in this assay, whereas cells with partial decreases in barbed end intensity were not included.

Finally, to determine if the failure of cells microinjected with Ab286 to generate barbed ends was due to the inhibition of cofilin's severing activity, and not simply the consequence of a shortage of actin monomers due to the blocking of actin turnover by cofilin, we evaluated the amounts of F-actin present in the microinjected cells at the time barbed ends were measured (1 min after EGF stimulation). The amount of F-actin in cells microinjected with anticofilin was quantified by pixel intensity after staining with rhodamine-labeled phalloidin, and compared with control cells either not injected or injected with control IgG. We found that 1–2 h after microinjection, there is no significant increase in the amount of F-actin in cells injected with cofilin antibodies compared with control cells (Table). This indicates that the concentration of actin monomers in vivo was unlikely to be significantly affected by microinjection of blocking antibodies against cofilin for the brief period used in these experiments.

Our study provides a detailed analysis in vivo and in vitro of the role of cofilin in EGF-stimulated lamellipod extension and actin polymerization. We have demonstrated that: (a) cofilin is recruited to the leading edge at the same time and in the same location as free barbed ends in response to stimulation with EGF; (b) stimulation with EGF results in an increase in the number of filaments (Chan et al. 1998; Bailly et al. 1999; this study), as well as a decrease in filament length at the leading edge (Bailly et al. 1999); (c) EGF stimulates cofilin's severing activity in cell lysates; and (d) function-blocking antibodies against cofilin suppress the appearance of free barbed ends and lamellipod protrusion after stimulation with EGF.

Altogether, these results directly implicate cofilin in the dynamics of actin polymerization at the leading edge of cells after stimulation with EGF. They also indicate that cofilin's severing activity is likely to play a direct role in the mobilization of free barbed ends at the leading edge after stimulation.

The results shown here demonstrate that cofilin is recruited to the leading edge of the lamellipod, where barbed ends appear after EGF stimulation. However, we have also found, as previously reported for other types of cells (Bamburg and Bray 1987), that the visualization of cofilin in the leading edge is highly sensitive to detergent extraction. The ease with which cofilin is extracted from cells even during light permeabilization before fixation suggests that cofilin might be only transiently associated with the actin cytoskeleton in situ. This is consistent with the biochemical properties of cofilin in vitro, where purified cofilin bound to F-actin is released upon stimulation of cofilin's severing and depolymerizing activities (Hayden et al. 1993). Our results support this conclusion, since the EGF-stimulated recruitment of cofilin that we observed at the leading edge is only transient and cotemporal with barbed end appearance. They also suggest that transient activation of cofilin plays a role in the mobilization of barbed ends and their polymerization at the leading edge.

The localization of cofilin at the extreme leading edge of extending lamellipods shown here is inconsistent with previous elegant studies in Xenopus keratocytes, where cofilin was localized starting ∼1 μm back from the edge (Svitkina and Borisy 1999). However, as described before (Bailly et al. 1999), the ultrastructure of the cytoskeleton at the leading edge in MTLn3 cells is quite distinct from that of keratocytes, and more comparable to the general organization of the cytoskeleton reported for mammalian fibroblasts and chemotactic ameboid cells such as macrophages and leukocytes (Ryder et al. 1984; Hartwig and Shevlin 1986). Furthermore, when cofilin was localized in the lamellipods of fibroblasts in parallel with keratocytes (Svitkina and Borisy 1999), it was found to be distributed at the extreme leading edge of the lamellipod of fibroblasts, a pattern comparable to that described here in MTLn3 cells. Finally, the precise location of free barbed ends has not been determined in keratocytes as done for MTLn3 cells (Bailly et al. 1999) and fibroblasts (Symons and Mitchison 1991), making the relative locations of cofilin and free barbed ends uncertain in keratocytes.

Altogether, these results suggest that the precise localization of cofilin at the leading edge may vary depending on cell type and the type of motility being exhibited. For example, keratocytes exhibit the continuous extension of the leading edge without cycles of extension and retraction as seen in fibroblasts and MTLn3 cells. In addition, F-actin in keratocytes does not exhibit the characteristic centripetal flow relative to the substratum at the leading edge seen in fibroblasts (Theriot and Mitchison 1992), suggesting differences between fibroblasts and keratocytes in the dynamics and anchoring of the F-actin network at the leading edge (Smilenov et al. 1999). Since the relative retention of cofilin in extracted cells may depend on the relative rate of turnover of the filaments at the leading edge, it is possible that variations in the localization of cofilin in the leading edge occur among the different cell types due to differences in actin dynamics.

To identify the mechanisms responsible for the generation of free barbed ends at the leading edge, we determined the number of filaments in the leading edge relative to the whole cell. We found that the number of filaments increases by 1.5-fold, an increase of 60,000 filaments per cell, within 50 s after EGF stimulation. In situ localization of pointed ends using DNase I staining demonstrates that the increase in the number of filaments is confined to the leading edge and does not occur within stress fibers or in the general cytoplasm. These results are consistent with previous indirect measurements of actin filament number in cells after EGF stimulation, using the incorporation of 2 μM rhodamine-actin to detect the number of pointed ends in cytoskeletons capped with gelsolin–actin complex (Chan et al. 1998), as well as the direct measurement of filament density at the leading edge using quantitative EM (Bailly et al. 1999). Thus, these three different assays for measuring actin filament numbers agree that the number of filaments increases significantly after EGF stimulation, and that the increase in filament number occurs specifically in the leading edge.

The increase in filament number at the same time and place as the increase in free barbed ends after stimulation is inconsistent with a pure uncapping mechanism. In a pure uncapping mechanism, filament length would increase as a result of polymerization from free barbed ends, but the number of filaments before and after EGF stimulation would remain constant. However, an increase in filament number alone does not distinguish between severing and de novo nucleation as the dominant mechanism responsible for the increase in free barbed ends at the leading edge in MTLn3 cells.

To determine if cofilin's severing activity is turned on by EGF stimulation, we used a light microscope severing assay to directly visualize filament severing activity in lysates from resting and stimulated MTLn3 cells. Results from the light microscope severing assay show that cofilin's severing activity is turned on by the stimulation of cells with EGF. These results in cell lysates are consistent with the EGF-induced decrease in filament length, with the specific loss of long filaments, observed in the electron microscope in cells at the leading edge as reported previously (Bailly et al. 1999). Together, these results predict that cofilin-mediated severing activity is present at the leading edge after stimulation with EGF.

The light microscope assay used to investigate the severing activity of cofilin in cell lysates was also useful in demonstrating the function-blocking activity of Ab286 toward cofilin's severing activity. Microinjection of these antibodies into cells inhibited the EGF-stimulated extension of lamellipods, supporting a role for cofilin in the turnover of actin filaments at the leading edge to supply a monomer for new assembly as proposed elsewhere (Carlier et al. 1997; Loisel et al. 1999; Ressad et al. 1999; Svitkina and Borisy 1999).

However, the direct involvement of cofilin in generating barbed ends after stimulation is supported by the finding that cells microinjected with anticofilin were inhibited in their ability to generate new barbed ends at the leading edge after stimulation with EGF. The amounts of G- (77 μM) and F-actin (76 μM) in MTLn3 cells are known (Edmonds et al. 1996). Since the amount of F-actin in cells was not changed within 2 h of microinjection of anticofilin, the failure of cells to generate free barbed ends was unlikely to be due to depletion of G-actin in vivo after the inhibition of cofilin. Furthermore, since the G-actin used to visualize free barbed ends in the assay described here was supplied exogenously, any activated nucleation template that was made before permeabilization would incorporate the exogenous G-actin and be detected regardless of whether there was enough endogenous G-actin in the cells to support polymerization. Therefore, the inhibition of lamellipod extension by function-blocking antibodies against cofilin is consistent with a more direct involvement of cofilin in barbed end mobilization, since the polymerization of free barbed ends is believed to supply the force for protrusion of the leading edge (Condeelis 1993).

The results presented in this study demonstrate for the first time that cofilin's severing activity is turned on during stimulated lamellipod extension and that cofilin activity is required for the appearance of barbed ends. Cofilin could contribute to barbed end formation by two different mechanisms. The first mechanism, which has received a lot of attention recently in the literature, proposes that cofilin enhances indirectly the nucleation activity of the Arp2/3 complex by supplying actin monomers through its depolymerizing activity (Egile et al. 1999; Loisel et al. 1999; Ressad et al. 1999; Svitkina and Borisy 1999). The second mechanism, based on biochemical studies (Bamburg 1999; Du and Freiden 1998; Maciver et al. 1998; Ichetovkin et al. 2000), proposes that cofilin can directly contribute barbed ends by severing actin filaments (Bailly et al. 1999).

The first mechanism is consistent with the known biochemical activities of Arp2/3 (Mullins et al. 1998; Machesky et al. 1999; Welch 1999) and its location at the leading edge (Bailly et al. 1999; Svitkina and Borisy 1999). However, it is inconsistent with the inhibition of barbed ends by anticofilin in the presence of either endogenous levels of G-actin or exogenous polymerization-competent G-actin sufficient for polymerization as discussed above. In addition, it is difficult to reconcile with the fact that polymerization-competent G-actin is present in vivo at concentrations exceeding 75 μM in MTLn3 cells before the stimulation of protrusion and motility. The second mechanism is consistent with increases in cofilin's severing activity and transient recruitment to the leading edge after stimulation. However, it is inconsistent with the rapid depolymerization of filaments in the presence of cofilin (Carlier et al. 1997). The synergistic interaction between cofilin and Arp2/3, where Arp2/3 caps and stabilizes short filaments produced by cofilin severing activity, may resolve the inconsistencies inherent in the above two mechanisms (Bailly et al. 1999). Further work will be required to measure the relative contributions of cofilin and Arp2/3 to the bursts of actin nucleation observed at the leading edge in response to chemotactic stimulation.

The authors acknowledge Jeff Wyckoff and the Albert Einstein College of Medicine Analytic Imaging Facility for help in image analysis, and Susan Welti for help in antibody production.

This work was supported by United States Army AMRDC 2466 (J.E. Segall), GM38511 (J.S. Condeelis), the Harrison Trust administered by the Einstein Comprehensive Cancer Center (N. Zebda), and National Institutes of Health training grants 5T32 HL07675-07 (A.Y. Chan) and 2-T32-CA09475 (M. Bailly). J.E. Segall is supported by an Established Scientist Award from the New York City Affiliate of the American Heart Association.