Enhancement of branching efficiency by the actin filament-binding activity of N-WASP/WAVE2

The actin cytoskeleton is one of the primary determinants of cell shape and motility. In the process of actin cytoskeletal reorganization, the Wiskott-Aldrich syndrome protein (WASP) family plays essential roles downstream of many regulators (Takenawa and Miki, 2001). This family consists of two WASP proteins (Derry et al., 1994; Miki et al., 1996) and three verprolin-homologous proteins (WAVEs) (Miki et al., 1998; Suetsugu et al., 1999). The key role of WASP family proteins is to activate the actin-related protein (Arp) 2/3 complex in response to signals (Fukuoka et al., 2001; Miki et al., 2000; Rohatgi et al., 1999).

The architecture of WASP family proteins, WASP, neural WASP (N-WASP) and WAVEs are similar not only in their C-terminal regions but also in several other domains (Fig. 1a). All WASP family proteins have a basic region and a proline-rich region in their center as well as verprolin-homology, cofilin-homology and acidic domains (VCA domain) at the C terminus. The proline-rich region binds many signaling molecules such as Ash/Grb2 and WASP-family-interacting SH3 protein (WISH), as well as actin regulatory molecules such as profilin (Carlier et al., 2000; Fukuoka et al., 2001; Miki et al., 1996; Suetsugu et al., 1998). The VCA domain is the activating domain for Arp2/3. Basic amino acids are clustered throughout the mid-region of WASP family proteins and this basic domain was recently found to bind phosphoinositides (Prehoda et al., 2000; Rohatgi et al., 2000). However, the entire basic domain is not necessary for binding to phosphoinositides (Rohatgi et al., 2000).

Experiments performed in vitro with pure Arp2/3 complex and actin show that the C-terminal VCA domain of WASP family proteins is the activating domain of the Arp2/3 complex, which results in rapid de novo nucleation and subsequent burst of actin polymerization (Blanchoin et al., 2000; Pantaloni et al., 2000; Rohatgi et al., 1999). Therefore, the VCA peptide fragment is thought to be a constitutive activator of actin polymerization in vitro. In this process, the verprolin-homology (V) domain directly associates with actin, and the cofilin-homology acidic (CA) region interacts with the Arp2/3 complex, forming a nucleating machine for actin polymerization (Machesky et al., 1999; Miki et al., 1996; Miki et al., 1998; Rohatgi et al., 1999; Suetsugu et al., 1999).

Activated Arp2/3 not only causes actin polymerization but also induces formation of branched filaments, which are similar in morphology to actin filaments seen in the lamellipodia and actin comet of motile bacteria (Blanchoin et al., 2000; Pantaloni et al., 2000). This activity is due to an actin filament side-binding activity of Arp2/3 itself (Mullins et al., 1998; Mullins et al., 1997), because pre-existening actin filaments increased their actin polymerization even when their barbed ends were protected from elongation (Amann and Pollard, 2001), and antibody against the actin-filament-binding interface of Arp2/3 inhibited branching (Bailly et al., 2001).

In studies of actin-based motility, the formation of actin comets by intracellular pathogens such as Shigella and Listeria are well-established models for understanding the event at the leading edges of motile cells. This actin comet formation can also be reconstituted with plastic beads coated with N-WASP and five other proteins: F-actin, Arp2/3, barbed-end capping protein, profilin and actin depolymerizing factor (ADF)/cofilin (Loisel et al., 1999; Pantaloni et al., 2001; Suetsugu et al., 2001b). This observation clearly shows that N-WASP is the minimum essential molecule for actin-based movement. Previously, N-WASP itself was shown to associate with actin filaments (Egile et al., 1999). Therefore, the association of F-actin with N-WASP is likely to be important. However, the significance of this filament-binding activity is still unclear.

We previously investigated which region of N-WASP is essential for actin comet formation (Suetsugu et al., 2001b). Whereas actin comets formed on beads coated with full-length N-WASP, no comets formed on beads coated with only VCA domain, the minimum domain required for Arp2/3 activation (Suetsugu et al., 2001b). VCA-coated beads were able to induce actin clouds around the beads, but few grew to actin comets. Thus, some other region is likely to be important for actin-based movement besides the VCA domain (Suetsugu et al., 2001b). Various mutant studies indicated that the basic region is also required for actin comet formation (Suetsugu et al., 2001b).

In the present study, we investigated how full-length N-WASP/WAVE2 induces filament branching. We found that the basic domain located in the mid-region of N-WASP and WAVE2 binds with actin filaments. Increased association of Arp2/3 with filaments was achieved by the actin-filament-binding activity of N-WASP and WAVE2. The increase in association between Arp2/3 and filaments was independent of the N-WASP activators Cdc42 and PIP₂. In N-WASP, the basic region is reported to be the PIP₂-binding region. However, PIP₂ and actin filaments bind to different sites in N-WASP, and recruitment of Arp2/3 was not prevented by PIP₂. Furthermore, filament branching induced by wild-type N-WASP was more efficient than branching induced by VCA fragments, presumably owing to recruitment of Arp2/3 to filaments.

Mutant or wild-type N-WASP and WAVE2 with an N-terminal six-His tag were expressed in SF9 cells with the Bac-to-Bac baculovirus expression system (Gibco BRL). Recombinant virus-infected Sf9 cells were lysed, clarified and affinity-purified with Ni-NTA beads (Qiagen). VCA was purified as described previously (Miki et al., 1996). Recombinant Cdc42 was purified and loaded with GTPγS as described previously (Suetsugu et al., 1998). Arp2/3 was affinity purified as described previously with the CA fragment of N-WASP (Egile et al., 1999). Actin was purified from rabbit skeletal muscle and monomeric actin (G-actin) was isolated by gel filtration on Superdex 200 (Amersham Pharmacia) in G buffer (2 mM Tris-HCl, pH 8.0, 0.2 mM CaCl₂, 0.5 mM DTT, 0.2 mM ATP). Capping protein β1 (CapZ) was purified as described previously (Soeno et al., 1998). Barbed-end capping activity was determined by assaying for lowering of the end point of polymerization and pointed-end nucleating activities not inhibited by cytochalasin b.

The pyrene-actin assay was performed in XB as described previously (Rohatgi et al., 1999). Final concentrations of Arp2/3, G-actin and pyrenyl-actin were 60 nM, 0.46 μM and 0.04 μM, respectively. The wild-type or mutant N-WASP concentration was 100 nM unless otherwise indicated. PIP₂-containing vesicles were prepared as previously described (Rohatgi et al., 1999). Components without G-actin were mixed and allowed to form complexes for 5 minutes at room temperature and then G-actin was added. The generated barbed-end concentration was calculated at the point where polymerization is 80% complete, assuming that pointed elongation is minimum (Higgs et al., 1999) with the equation,

where k₊ is the rate constant for the association of G-actin with filament ends.

The G-actin solution was first clarified by ultracentrifugation at 100,000 g for 20 minutes. To make ordinary filaments, G-actin (40 μM) was polymerized for 1 hour at room temperature by addition of 5× G-to-F conversion buffer (80 mM Tris-HCl, 250 mM KCl, 1 mM ATP, 2 mM MgCl₂). For generation of ADP-F-actin, the F-actin obtained was treated with 20 units ml^–1 hexokinase and 0.2 mM glucose overnight, as described previously (Pollard, 1984). The amount of polymerized actin was confirmed by ultracentrifugation. ADP-phosphate (ADP-P_i) filaments were made by polymerizing G-actin in the presence of phalloidin (4 μM). ADP filament was mixed with phalloidin and also used in the assays. Gelsolin-capped filaments were prepared as described (Mullins et al., 1998). The molar ratio of gelsolin and actin monomer was 1:200.

The binding of F-actin to N-WASP was analyzed by co-sedimentation assay. Wild-type or mutant N-WASP and WAVE2 solutions were clarified by ultracentrifugation to remove any aggregates. These F-actin and N-WASP/WAVE2 solutions were then mixed and incubated at room temperature for 1 hour. The mixture was then centrifuged at 70,000 g for 20 minutes. The supernatants and pellets were then analyzed by SDS-PAGE.

After 15 minutes of actin polymerization, 1 μl of the reaction mixture was diluted in 100 μl of staining solution composed of 0.002% BSA, 0.5% methylcellulose, 3 mg ml^–1 glucose, and 4 units ml^–1 Alexa 488 phalloidin (Molecular Probes) in XB. Then, 2 μl of solution containing stained actin filaments was placed between a polylysine-coated glass slide and an 11×22 mm cover glass. Samples were observed on a Zeiss Axiovert S100 with an intensified charge-coupled device camera (Photometrics). F-actin for pre-incubation was generated by co-polymerization of G-actin and rhodamine-labeled G-actin (Theriot and Fung, 1998) at a ratio of 7:3. This labeled F-actin was added to the reaction mixture to a final concentration of 300 nM. Capping protein was added at a final concentration of 100 nM during pre-incubation.

The basic clusters are often the sites for actin filament binding. Therefore, to determine the role of the basic region in detail, we tested the binding of wild-type or Δbasic N-WASP, in which the basic region is deleted, and WAVE2 to actin filament (F-actin). We performed co-sedimentation assays of wild-type or Δbasic N-WASP and F-actin. Whereas wild-type N-WASP co-sedimented with F-actin, Δbasic mutant N-WASP did not. These results indicates that the basic region of N-WASP binds with F-actin (Fig. 1b). The Δbasic WAVE2 also did not associate with F-actin through the basic region (Fig. 1c).

To determine how N-WASP or WAVE2 associates with actin filament through the basic region, we assayed for association of actin filaments with isolated basic region by co-sedimentation (Fig. 2). We first tested actin filament binding of N-WASP fragment spanning amino acids 125-273, which contains the entire basic region. As shown in Fig. 2a, this fragment associated with actin filaments. Basic clusters at amino acids 186-195 were previously shown to be the PIP₂ binding region (Rohatgi et al., 2000). Therefore, we next searched for a region in the basic clusters that contributes to actin filament association. N-WASP amino acids 125-165 (first half of the basic region) strongly associated with filaments. By contrast, N-WASP amino acids 166-275 (latter half of the basic region) had a low affinity for filaments (Fig. 2a).

We next examined whether these basic fragments can compete out with full-length N-WASP for binding with actin filaments. In good agreement with the above experiment, both fragments from the basic region of N-WASP, the two halves of the basic region, inhibited the association of N-WASP with filaments. Also, inhibiting activity was stronger with the first half of the basic region than the second half (Fig. 3). Therefore, association of N-WASP with actin filaments involves the entire basic region. Because the second half of the basic region is involved in PIP₂ binding, the modes of association of N-WASP to filaments and to PIP₂ are different. Indeed, neither PIP₂ nor Cdc42 prevented the association of N-WASP to filaments (not shown).

The isolated basic region of WAVE2 also associated with actin filaments, demonstrating that the basic region is the actin-filament-binding region (Fig. 2a). For both N-WASP and WAVE2, association of the isolated basic region with filaments was saturated as filament concentration increased. From a plot of the association of the basic regions and filaments, the K_d for actin filament association was estimated to be ∼1 μM, which agrees with that of a previous study (Fig. 2b) (Egile et al., 1999).

The actin filaments used in the above experiments were made by polymerization of G-actin (ordinary filaments). When actin polymerizes, the state of the actin filaments changes with time. First, ATP-G-actin polymerizes to form ATP-F-actin. However, the ATP on the filaments is soon hydrolyzed to ADP, but the phosphate generated by ATP hydrolysis is not released from the filaments immediately, resulting in transient ADP-P_i filaments. ADP-P_i filaments, which have a half-life of ∼2 minutes, release P_i, becoming the more-unstable ADP filaments. Therefore, most of the filaments used in the above study (ordinary filaments) were ADP filaments with very short ATP and ADP-P_i actin caps at the ends (Carlier et al., 1984). In the place of active actin polymerization, where N-WASP or WAVEs might function, it is thought that significant amounts of ADP-P_i filaments exist, whereas in the resting state, most filaments are thought to be ADP filaments.

Therefore, we examined association of these isolated basic regions with ADP-P_i filaments and ADP filaments (Fig. 2c). Because phalloidin inhibits phosphate release from filaments, actin filaments made by polymerization reaction including phalloidin were used as ADP-P_i filaments. ADP filaments were made by treating ordinary filaments with hexokinase (Pollard, 1984). These ADP filaments were used with and without phalloidin for comparison of ADP-P_i filaments. The affinity against both N-WASP and WAVE2 seems to be stronger for ADP filaments than for ADP-P_i filaments (Fig. 2c). Association of phalloidin with filaments did not significantly affect association with basic regions (Fig. 2c).

To examine the meaning of the filament-binding activity of N-WASP/WAVE2, we assessed the influence of N-WASP/WAVE2 on the association of Arp2/3 with filaments. Arp2/3 alone was reported to have actin filament side-binding activity (Mullins et al., 1998; Mullins et al., 1997). Thus, we examined the effect of N-WASP/WAVE2 on the association of actin filaments with Arp2/3 by the co-sedimentation assay. Addition of VCA or Δbasic N-WASP increased the association of Arp2/3 with filaments, as reported previously (Amann and Pollard, 2001), suggesting that activation of Arp2/3 by VCA increases the affinity of Arp2/3 to filaments. In parallel with stronger association with ADP filaments, the amount of Arp2/3 in the actin filament pellet was increased about threefold by addition of wild-type N-WASP/WAVE2 (Fig. 4). This increase was not observed upon addition of the Δbasic mutant or VCA fragment, suggesting that actin filament binding of N-WASP/WAVE2 is essential for recruitment of Arp2/3 to filaments. Association of N-WASP-Arp2/3 with ADP-P_i filaments stabilized with phalloidin is less than that with ordinary filaments stabilized with phalloidin, and ordinary filaments without phalloidin (Fig. 4). Therefore, the ability of N-WASP to recruit Arp2/3 to the F-actin pellet was greater with ADP-F-actin than with ADP-P_i-F-actin. In agreement with a different mode of association for PIP₂ and filaments, addition of PIP₂ at a concentration sufficient to activate N-WASP did not prevent recruitment of Arp2/3 to actin filaments. Therefore, PIP₂ is not likely to compete with actin filaments for the same binding sites on N-WASP. Rather, this result indicates that PIP₂ and actin filaments bind to different sites within the basic region of N-WASP (Fig. 4).

Recent studies showed that VCA in the presence of Arp2/3 simultaneously stimulates nucleation and Y-branching of actin filaments in vitro (Blanchoin et al., 2000; Pantaloni et al., 2000). To assess the effect of binding of actin filaments with N-WASP or WAVE2, we visualized actin filaments formed in the presence of preformed actin filaments. All actin filaments were visualized by staining with green phalloidin and actin filaments used for pre-incubation (preformed filaments), which were made with rhodamine-labeled G-actin, were detected by rhodamine (red) fluorescence (Fig. 5a).

As reported previously (Blanchoin et al., 2000), activation of Arp2/3 by either VCA, Δbasic N-WASP and WAVE2, or wild-type N-WASP and WAVE2 induced branches much more efficiently than did Arp2/3 alone (Fig. 5). However, the branch rate was affected by actin filament binding activity (Fig. 5). We first examined branch formation by wild-type N-WASP/WAVE2 on pre-existing filaments formed from rhodamine-labeled, red G-actin. On these pre-existing filaments, induction of branched filament formation was two times greater with wild-type N-WASP or WAVE2 (0.09-0.10 branches per μm filament) than with Δbasic mutant N-WASP/WAVE2 or VCA fragment (0.03-0.04 branches per μm filament) (Fig. 5d). Decrease in activation of Arp2/3 in the Δbasic mutant is unlikely because Δbasic mutant N-WASP/WAVE2 is a potent Arp2/3 activator (Fig. 6c). Wild-type N-WASP or WAVE2 induced much more branching than did the VCA fragment alone. The similar branch-inducing activity of VCA and the Δbasic mutant N-WASP/WAVE2, which is about half as active as that of wild-type proteins, indicates that the association of N-WASP or WAVE2 with filaments is important for efficient branch formation.

We then measured the lengths of the original filaments (mother filaments) from branch points, which grew in the original direction, and the length of the branched filaments (daughter filaments), which grew at about 70° angles to the direction of mother filament growth (Fig. 5c). Branching on pre-existing filaments occurred randomly with no preference of branch formation near the end of pre-formed filaments (Fig. 6c). This result suggests that actin filament binding activity of N-WASP/WAVE2, which presumably links Arp2/3 and the actin filament, did not create the preference of branch position on pre-existing filaments near either the barbed end or the pointed end.

Filaments polymerized only by G-actin have short ATP- and ADP-P_i-actin caps, which usually consist of several actin monomers (<0.6 μm, or <100 actin monomers) (Carlier et al., 1984). In our preparation, the average length of pre-existing filaments was ∼5 μm, so most filaments were thought to be in the ADP form. We confirmed that branching occurs on ADP filaments, which are produced by treatment of ordinary filaments with hexokinase (Pollard, 1984). The extent of branching on ADP filaments made by hexokinase treatment was similar to that on ordinary filaments, confirming the effectiveness of branching on ADP filaments by wild-type N-WASP/WAVE2 (Fig. 5a,d).

We next examined branching on de novo filaments. Actin polymerization ceases within 2 minutes of the burst of polymerization under this condition (2 μM G-actin) (data not shown). Therefore, branches formed on de novo, green filaments are thought to represent branches on ADP-P_i or ATP filaments, because the half-life of phosphate release is ∼2 minutes, which is sufficient to cease actin polymerization. The branches on de novo green filaments were observed in all cases at a rate of 0.06 branches per μm filament (Fig. 5a,d). On de novo filaments, which are presumably ADP-P_i filaments, branching also occurred randomly (Fig. 5b). Because wild-type N-WASP and mutant N-WASP lacking the basic domain induced branching equally on ADP-P_i filaments, branching on de novo ADP-P_i filaments does not appear to be influenced by the actin filament binding activity of WASP family proteins, in agreement with the binding assay (Fig. 2).

Therefore, when Arp2/3 is activated by wild-type N-WASP/WAVE2, branching is generated more efficiently on pre-existing filaments than on de novo filaments. This phenomenon is not observed without filament binding activity or with VCA only, suggesting that the filament binding activity of N-WASP/WAVE2 is essential for this preferential branching on ADP filaments.

In all experiments described above, preformed red filaments were pre-incubated with Arp2/3 and N-WASP before addition of G-actin to allow complex formation. After the pre-incubation process, G-actin was added and then polymerization proceeded. Pre-incubation of N-WASP/WAVE2 and filaments might stabilize the filaments, which would increase the rate of branching, because many actin-filament-binding proteins stabilize filaments. However, full-length N-WASP did not stabilize the filaments at all as measured by depolymerization of gelsolin-capped filaments (Fig. 6i). Neither activation of N-WASP with Cdc42 and PIP₂ nor incorporation of Arp2/3 stabilized actin filaments. Therefore, N-WASP and Arp2/3 did not seem to stabilize the pre-existing filaments (Fig. 6i).

To elucidate whether the increase of branching by intact N-WASP and WAVE2 accompany the increase of the number of barbed ends, we analyzed the kinetics of actin polymerization in the presence of pre-existing filaments. To minimize the background polymerization at pointed ends, the G-actin concentration was set at 0.5 μM, which is below the pointed end critical concentration. To eliminate polymerization from the free barbed ends of pre-existing filaments, gelsolin-capped filaments were used. Gelsolin binds barbed ends with high affinity against barbed ends (K_d < 1 nM) and blocks barbed end elongation of pre-existing filaments (see control curves in Fig. 6a-c). We measured actin polymerization induced by intact N-WASP (with actin filament binding activity), VCA domain alone or Δbasic mutant (without actin filament binding activity) in the presence and absence of pre-existing filaments. The polymerization curves obtained were differentiated with time (dF/dt) and the kinetics of the increase of actin polymerization were compared (Fig. 6a-g). The differentiated polymerization curve (dF/dt) indicates the rate of increase of actin filaments.

Actin polymerization initiated by Arp2/3 and mutant N-WASP or wild-type N-WASP proceeds in an autocatalytic manner as polymerization proceeds (Pantaloni et al., 2000). By contrast, the polymerization nucleated by actin filament seeds without gelsolin did not display any auto-catalytic polymerization (not shown). In polymerization mediated by VCA, Δbasic N-WASP or wild-type N-WASP, addition of filaments shifted the peak of polymerization (dF/dt) earlier, indicating that the lag time of polymerization decreased. Therefore, addition of actin filaments clearly activated Arp2/3-mediated actin polymerization (Fig. 6). Because the barbed ends of these actin filaments are protected by gelsolin, the sides of filaments are important for enhancing actin polymerization, as previously reported (Amann and Pollard, 2001). Addition of pre-existing actin filaments makes the maximum rate of increase of filaments (the height of dF/dt) greater for wild-type N-WASP. By contrast, the maximum rate of increase of filaments with VCA or Δbasic was not increased (Fig. 6e-g). Thus, pre-existing filaments confer more co-operativity on polymerization induced by wild-type N-WASP, presumably through actin filament binding activity and recruitment of Arp2/3 to filaments. These results indicate that pre-existing filaments upregulate actin polymerization mediated by wild-type N-WASP by (1) decreasing the lag time (depends on Arp2/3 because the filament-side binding activity of N-WASP is not involved) and (2) increasing the rate of polymerization (depends on actin-filament-side binding activity of N-WASP).

We next examined the dF/dt when 80% of polymerization was completed. At this point, the number of barbed ends is proportional to the rate of increase of actin filaments because pointed end elongation of filaments is assumed to be minimal (Higgs et al., 1999). Pre-existence of actin filaments increased the number of barbed ends by ∼30% only with wild type N-WASP, suggesting that actin filament binding activity contributes to the increase in the number of barbed ends (Fig. 6h). The rate increase might be due to a decrease in the rate of depolymerization because many actin-filament-binding proteins stabilize actin filaments. However, N-WASP did not inhibit pointed-end depolymerization (Fig. 6i).

In agreement with preferential binding to ADP filaments, the increase in the number of barbed ends by wild-type N-WASP was specific for ADP-F-actin and was not seen with ADP-P_i-F-actin (actin polymerized in the presence of phalloidin). Association of phalloidin with ADP-F-actin did not affect the rate of polymerization (data not shown). However it is possible that the rapid formation of ADP-P_i filaments by N-WASP and Arp2/3 might mask the effect of pre-existing ADP-P_i filaments. Although most of the actin filaments used in the above experiment made by polymerization of G-actin are thought to be in the ADP form, enhancement of polymerization rates by ADP-F-actin was confirmed with ADP-F-actin generated by hexokinase treatment (data not shown).

Actin polymerization after the polymerization lag was faster with the Δbasic mutant of N-WASP than with wild-type N-WASP (Fig. 5c,g). This phenomenon is explained in two ways. First, the release of auto-inhibition of N-WASP by the deletion of the basic region contributed to the unmasking of the catalytic VCA domain of N-WASP, because basic-acidic interaction is important for autoinhibition of N-WASP (Suetsugu et al., 2001b). Second, the presence of the WH1 domain further activates Arp2/3 (Suetsugu et al., 2001a). Although the Δbasic mutant of N-WASP is able to activate Arp2/3 more efficiently, neither the efficiency of branching nor the number of barbed ends was increased by the pre-existence of actin filaments, presumably owing to the deficiency in actin filament binding activity (Fig. 5; Fig. 6).

Wild-type WAVE2 but not the Δbasic mutant WAVE2 also increased the number of the barbed ends ∼30% in the presence of pre-existing actin filaments, indicating that association of actin filaments plays a similar role among WASP family proteins (data not shown). Therefore, addition of actin filaments in the ADP form enhanced actin nucleation and pre-existing filaments provide the scaffold for induction of rapid actin polymerization.

We first examined the role of actin filament binding activity of N-WASP/WAVE2 in formation of actin branching and polymerization. The actin filament binding activity of N-WASP/WAVE2 recruits Arp2/3 more effectively to ADP filaments than to ADP-P_i filaments. Therefore, the pre-existence of ADP filaments increased both the rate of branch formation and the rate of actin polymerization when activated by wild-type N-WASP/WAVE2 but not by VCA or Δbasic. Therefore, the VCA domain, which is the smallest peptide fragment capable of activating Arp2/3-mediated actin polymerization in vitro, does not represent all of the activities of N-WASP/WAVE2 for regulating actin filaments through Arp2/3. ADP filaments are required for optimum activation of Arp2/3 by N-WASP/WAVE2.

When actin polymerizes, ATP-G-actin first forms ATP filaments. ATP on actin filaments is rapidly hydrolyzed to ADP (half-life is ∼10 seconds). The phosphate from ATP is not released from actin filaments immediately after hydrolysis of ATP (dissociation half-life is about 2 minutes). In cells and in extracts, ADF/cofilin enhances the release of phosphate more than 10 times (Blanchoin and Pollard, 1999). Indeed, the most dynamic actin filaments turn over with a half-life of 23 seconds (Theriot and Mitchison, 1991). Therefore, in cells, the filaments formed de novo are likely rapidly to become ADP filaments. Thus, the pre-existing filaments in cells are likely to be in the ADP form. Therefore, branched filament formation at the side of pre-existing ADP filaments is probably important for actin cytoskeletal re-organizations.

There is evidence that supports the importance of efficient side-branching from pre-existing filaments. Actin comet formation in vitro using plastic microspheres coated with N-WASP or its mutant is a good system for studying the mechanism of actin cytoskeletal formation. We previously showed that VCA alone is not sufficient for actin comet formation but VCA-coated beads were able to induce actin filament formation around the beads. The basic region, which we found to be the actin filament binding region, was required to form actin comets (Suetsugu et al., 2001b).

Because N-WASP-coated bead movement was reconstituted by only five proteins [profilin, ADP/cofilin, capping protein, Arp2/3 and actin (Pantaloni et al., 2001)], the lower branching efficiency of VCA or Δbasic mutant of N-WASP is the likely cause for the inability of VCA- or Δbasic-mutant-coated beads in actin comet formation. Before initiation of actin comet formation and directional movement, actin polymerizes randomly around the beads to form a dense actin cloud on which a moving force for beads should be generated (Cameron et al., 1999). After forming an actin cloud, the beads suddenly form actin comets and begin to move. In this period of actin cloud formation, there should be large amounts of ADP-actin filaments around beads. Therefore, higher branching rate on pre-existing ADP filament is thought to be important for formation of dense actin cloud. Because generation of force for movement requires a scaffold, the formation of dense actin clouds by efficient branching, on which many actin filaments can elongate, is likely to be important for initiation of directional movement of actin comets.

Here, we show that WASP family proteins have two mechanisms to activate Arp2/3-mediated side branching. Activation of Arp2/3 by VCA is the first identified mechanism for actin filament branching (Blanchoin et al., 2000; Pantaloni et al., 2000). Both rate of branching and association of Arp2/3 to filaments were increased by the association of Arp2/3 with VCA (Fig. 3) (Marchand et al., 2001). The enhancement of filament association is presumably due to VCA-induced changes inside Arp2/3, because VCA itself does not associate with actin filaments (Machesky and Insall, 1998; Miki and Takenawa, 1998). Branching induced by VCA-activated Arp2/3 occurs preferentially on newly formed filaments, which are composed mainly of ADP-P_i filaments, because polymerization ceases within the half-life of P_i release (Fig. 6) (Blanchoin et al., 2000). This observation is consistent with the hypothesis that autocatalytic Arp2/3 activation is mediated by ADP-P_i filaments, in which filaments formed de novo by one Arp2/3 complex in turn activate other Arp2/3 complexes (Amann and Pollard, 2001; Pantaloni et al., 2000).

The other mechanism found here is that branching occurs preferentially on ADP filaments when Arp2/3 is activated with wild-type N-WASP or WAVE2. This preference of branching on ADP filaments is due to the recruitment of activated Arp2/3 to filaments or activation of Arp2/3 on filaments, which then increases the rate of branching up to twofold (Fig. 4; Fig. 5). Thus, recruitment of Arp2/3 to filaments results in greater branching on ADP filaments than on new filaments. The increase in branching by N-WASP/WAVE2 presumably increases the number of elongating filaments near the region of N-WASP/WAVE2 activation and helps to generate the force for movement.

Recently, cortactin has been shown to tighten branches formed by Arp2/3, presumably through filament binding as well as Arp2/3-associating activity of cortactin (Uruno et al., 2001; Weaver et al., 2001; Weed et al., 2000). However, the distributions in actin comets of N-WASP and cortactin are very different. N-WASP is located exclusively at the surface of bacteria, where actin polymerization occurs. In contrast to N-WASP, cortactin is localized throughout the actin comet (Zettl and Way, 2001). So, what is the importance of higher branching efficiency on pre-existing filaments? One possibility is that intact N-WASP or WAVE2 makes more branches on pre-existing filaments, which are presumably tightened by cortactin to make the actin cloud tighter before the formation of actin comets, and thereby promotes the initiation of actin-comet-based movements.

In this study, we identified multiple roles for the basic region of N-WASP. This region was shown to bind PIP₂ and actin filaments (Rohatgi et al., 2000). These two activities are separate in both binding site and activity. The presence of a sufficient amount of PIP₂ for N-WASP activation did not decrease enhancement of Arp2/3 association with filaments (Fig. 4). The PIP₂-binding region is restricted to only 20 amino acids of the latter half of the basic region. However, most of the basic region contributes to association with actin filaments (Fig. 2) (Rohatgi et al., 2000). This region is also proposed to be involved in association of N-WASP with Arp2/3 (Prehoda et al., 2000). However, the association of VCA with Arp2/3 is far greater than that of the basic region (Suetsugu et al., 2001a). Thus, the basic region itself is not likely to be involved in the recruitment of Arp2/3 to filaments. Although the basic region-Arp2/3 complex association will be important for inhibition of Arp2/3 (Prehoda et al., 2000).

Deletion of the basic region of N-WASP leads to constitutively active N-WASP in Arp2/3-mediated actin polymerization (Fig. 4c) (Suetsugu et al., 2001b). This result suggests that the basic region is also involved in the formation of an autoinhibited structure in addition to the region around the CRIB motif (Kim et al., 2000). Positive charges in the basic region and negative charges in the C-terminal acidic region are thought to interact by electrostatic force. In support of this hypothesis, partial deletion of acidic residues led to the activation of N-WASP, indicating that the basic-acidic interaction plays an indispensable role in the formation of the autoinhibited structure (Suetsugu et al., 2001a).

Therefore, the basic region of N-WASP appears to have multiple roles including PIP₂ binding, actin filament binding, Arp2/3 association and autoinhibition. The three-dimensional structure of N-WASP will be required to verify these functions.

Prior to this study, the role of the basic region in WAVE was not known. Like the basic region of N-WASP, the basic region of WAVE binds actin filaments (Fig. 1). Therefore, the binding with actin filaments at the basic region is common to WASP family proteins including N-WASP and WAVE2. The basic region of WAVE, like that of N-WASP, may also be involved in phosphoinositide binding. Further study, however, is needed to determine the detailed role of the basic region of WASP family proteins in regulation of actin filament formation.

We thank S. H. Zigmond (University of Pennsylvania) and M.-F. Carlier (Centre National de la Recherche Scientifique) for critical reading and valuable discussion of our manuscript. We also thank T. Azuma (Keio University) for providing us with gelsolin. We thank T. Suzuki and C. Sasakawa (Institute of Medical Science, University of Tokyo) for providing antibody against Arp3. This work was supported by grant-in-aids from the Ministry of Education, Culture, Sports, and Technology of Japan and from JST.