GRB2 Links Signaling to Actin Assembly by Enhancing Interaction of Neural Wiskott-Aldrich Syndrome Protein (N-WASp) with Actin-related Protein (ARP2/3) Complex*

Proteins of the Wiskott-Aldrich Syndrome protein (WASp) family connect signaling pathways to the actin polymerization-driven cell motility. The ubiquitous homolog of WASp, N-WASp, is a multidomain protein that interacts with the Arp2/3 complex and G-actin via its C-terminal WA domain to stimulate actin polymerization. The activity of N-WASp is enhanced by the binding of effectors like Cdc42-guanosine 5′-3-O-(thio)triphosphate, phosphatidylinositol bisphosphate, or the Shigella IcsA protein. Here we show that the SH3-SH2-SH3 adaptor Grb2 is another activator of N-WASp that stimulates actin polymerization by increasing the amount of N-WASp·Arp2/3 complex. The concentration dependence of N-WASp activity, sedimentation velocity and cross-linking experiments together suggest that N-WASp is subject to self-association, and Grb2 enhances N-WASp activity by binding preferentially to its active monomeric form. Use of peptide inhibitors, mutated Grb2, and isolated SH3 domains demonstrate that the effect of Grb2 is mediated by the interaction of its C-terminal SH3 domain with the proline-rich region of N-WASp. Cdc42 and Grb2 bind simultaneously to N-WASp and enhance actin polymerization synergistically. Grb2 shortens the delay preceding the onset of Escherichia coli (IcsA) actin-based reconstituted movement. These results suggest that Grb2 may activate Arp2/3 complex-mediated actin polymerization downstream from the receptor tyrosine kinase signaling pathway.

N-WASp activates Arp2/3 in an enhanced fashion (6), leading to a dramatic enhancement of actin polymerization in vitro. This result suggested that the WA domain may be buried in the N-WASp protein due to some internal interaction, and N-WASp would switch to an active state, in which the WA domain is exposed, upon binding Cdc42 or PIP 2 . Very recent NMR studies (31) demonstrate that the so-called "cofilin-homology region" in the WA domain of WASp interacts with the CRIB domain, in an autoinhibited complex. Activation of N-WASp/Arp2/3 by the Shigella flexneri bacterial pathogen has also been described (9). S. flexneri undergoes actin-based propulsion in host cells and is one of the best tools for a molecular approach of actin-based motility. Initiation of actin assembly at the surface of Shigella is mediated by a single protein, IcsA (32), that harnesses N-WASp (33). IcsA mimics Cdc42 in inducing a structural change in N-WASp, exposing WA and stimulating Arp2/3-induced actin assembly and bacterial motility (9). The structural basis for activation of N-WASp by IcsA is not yet known.
The above results suggest that the function of WASp proteins in a physiological context is to link Cdc42 activation to de novo actin assembly. However, the activation of N-WASp may be driven by other ligands binding to the proline-rich region of N-WASp via SH3 domains, such as the adaptors Grb2 and Nck (22,23,34,35). The SH2/SH3 adaptor Grb2 (36,37) is conventionally considered to link the tyrosine-phosphorylated EGF receptor (via its SH2 domain) and the Ras guanine nucleotide exchange factor Sos (via its SH3 domains), without affecting the nucleotide exchange function (38 -42). On the other hand, several studies suggest that Grb2 binds to a large number of other targets (see Ref. 43 for a recent review), including proteins involved in the organization of the actin cytoskeleton. Therefore, Grb2 could act not only as an adaptor but also as an effector of actin-dependent processes in response to signaling. Microinjection of anti-Grb2 antibodies abolishes actin-based membrane ruffling in response to EGF (44), binding of Grb2 to dynamin (45) activates its GTPase activity (46), and inhibitors of the SH2 domain of Grb2 block cell motility in hepatocyte growth factor-stimulated Madin-Darby canine kidney cells (47). The role of these interactions in the regulation of cytoskeletal reorganization has not been clearly established; however, addition of Grb2 to cell lysates increases the amount of N-WASp that coimmunoprecipitates with the activated EGF receptor (23). Here we use in vitro actin polymerization assays with pure Arp2/3 and N-WASp and motility assays of Shigella in a medium reconstituted from pure proteins (3) to show that Grb2 directly activates N-WASp and mediates Arp2/3 complexdependent actin nucleation.
Polymerization Assays-Polymerization of actin was monitored by the increase in fluorescence of pyrenyl-labeled actin. The assays were carried out at 20°C in a Spex fluorolog2 or a Safas flx spectrofluorim-eter. Samples (80 l) contained typically 2.5 M MgATP-G-actin (10% pyrenyl-labeled), 10 -20 nM Arp2/3 complex, and N-WASp, Grb2, and other reagents as indicated in F buffer (G buffer supplemented with 0.1 M KCl and 1 mM MgCl 2 ). The maximum slope of the polymerization curves was taken as a measure of the activation of Arp2/3 complex.
Modeling of the Kinetic Data in Terms of N-WASp Monomer-Dimer Equilibrium-N-WASp (W) was assumed to bind Arp2/3 (Y) and Grb2 (X) and to associate into a dimer W 2 with an equilibrium dissociation constant K D . The different equilibria to be considered (Fig. 6) were described by the following Equations 1-6.
The modeling of data shown in Fig. 2a (54), was tested by examining the motile behavior of bacteria in the motility medium reconstituted from pure proteins (3). Bacteria (6ϫ10 9 bacteria/ ml) were precoated with N-WASp by preincubation in X buffer (20 mM HEPES, pH 7.5, 0.1 M KCl, 1 mM MgCl 2 , 0.2 mM CaCl 2 , 50 mM sucrose) containing 0.2 M N-WASp and Grb2 as indicated, at 20°C for 10 min. Bacteria were then sedimented at 13,000 ϫ g for 4 min, resuspended in the original volume of X buffer, and diluted 30-fold in the motility medium. The motility medium consisted of 8 M F-actin, 5 M actin depolymerizing factor, 50 nM Capping protein, 2.5 M profilin, and 0.1 M Arp2/3 complex and was supplemented or not with Grb2 at the same concentration as in the preincubation step. Motility was observed in phase contrast optical microscopy as described (3). The time course of generation of actin tails and establishment of the steady rate of propulsion was monitored for each sample, and the average rates were measured over at least 10 bacteria.
EDC/NHS Cross-linking of N-WASp-N-WASp was equilibrated in 20 mM HEPES, 60 mM KCl buffer, pH 7.5, and incubated at 1 M with 5 mM EDC and 5 mM NHS for 45 min at 0°C. Samples were immediately denatured and submitted to SDS-PAGE (9% acrylamide), electrotransferred onto nitrocellulose, and Western blotted using a polyclonal anti-N-WASp at a 1:500 dilution (9) as primary antibody and peroxidase-coupled anti-rabbit IgG as secondary antibody, followed by ECL chemiluminescent detection (Amersham Pharmacia Biotech).

Grb2 Enhances the Interaction of N-WASp with Arp2/3, Leading to the Stimulation of Actin Polymerization-
The binding of Cdc42-GTP␥S or of the Shigella protein IcsA to N-WASp leads to the activation of actin polymerization by Arp2/3. The molecular mechanism by which activated Arp2/3 complex induces the seemingly autocatalytic polymerization of actin is not fully understood. The current interpretation of the activation of Arp2/3 by N-WASp is that when N-WASp binds effectors like Cdc42 or IcsA, the C-terminal WA domain of N-WASp is exposed and is able to interact with Arp2/3 and G-actin to stim-ulate actin filament nucleation and barbed end growth (6,9). The pyrene fluorescence assay for actin polymerization provides a rapid, accurate, and quantitative test of the function of putative effectors leading to the structural change and activation of N-WASp (or of antagonists of this reaction).
The effect of Grb2, an SH3-SH2-SH3 multidomain adaptor, on the function of N-WASp was addressed. Grb2 stimulated actin polymerization in the presence of Arp2/3 complex and N-WASp (Fig. 1a). The effect of Grb2 was maximum at a concentration of 1 M, and half-effect was observed at 0.1-0.2 M (data not shown). Under conditions of maximum activation, in the presence of 1 M Grb2, the N-WASp concentration dependence of the rate of actin polymerization was consistent with the binding of Grb2⅐N-WASp complex to Arp2/3 with an equilibrium dissociation constant of 5 nM (Fig. 1b). In the absence of Grb2, however, the activity measurements did not reflect normal saturation of Arp2/3 by N-WASp. First, the half-maximum was obtained at a concentration of N-WASp representing 25% of the Arp2/3 concentration. Second, activation of polymerization first increased with N-WASp and then decreased at high concentration (Fig. 2a). Third, the percent activation provided by Grb2 increased with the concentration of N-WASp (Fig. 2a, inset). The decrease in activation of polymerization upon increasing N-WASp was observed in a range of N-WASp concentration (0.1-0.8 M) in which no appreciable sequestration of G-actin by WA can occur. The two opposite effects of N-WASp therefore suggest that N-WASp exists in two states that can interact with Arp2/3. At low concentration (10 Ϫ9 -10 Ϫ8 M), N-WASp binds Arp2/3 in a complex that stimulates polymerization, whereas at high concentration (10 Ϫ7 -10 Ϫ6 M) it also binds Arp2/3, but in a state that does not stimulate polymerization. The stronger stimulating effect of Grb2 at high N-WASp concentration suggests that Grb2 shifts N-WASp to its active state. To get more insight into the possible association state of N-WASp and its control by Grb2, sedimentation velocity and cross-linking experiments were carried out. At 4 M N-WASp had a sedimentation coefficient s 20,w of 5.45 S. This value is much higher than the one expected (3.98 S) for a monomeric 56-kDa globular protein and is suggestive of a dimeric structure of N-WASp. Cross-linking of N-WASp by EDC/NHS showed evidence for a triplet of covalently crosslinked polypeptides migrating with apparent molecular masses of 150 -190 kDa in SDS-PAGE (Fig. 2b), supporting the view that N-WASp is in a self-association equilibrium. The different cross-linked species may correspond to different locations of the cross-links in the amino acid sequence. The amount of crosslinked polypeptides decreased in the presence of Grb2, suggesting that Grb2 stabilizes the monomeric form of N-WASp. No covalently cross-linked Grb2-N-WASp polypeptide was de- tected. As will be discussed under "Discussion," the polymerization data shown in Fig. 2a are satisfactorily modeled (continuous line through the data points) assuming that N-WASp undergoes self-association upon increasing concentration and that the monomeric form activates actin polymerization and is favored by Grb2 binding.
Interaction of the C-terminal SH3 Domain of Grb2 with N-WASp Is Responsible for the Stimulation of Actin Assembly by Arp2/3 Complex-The above results suggest that one or both of the two SH3 domains of Grb2 activate N-WASp by binding to the proline-rich region of the protein. This possibility was tested using mutated forms of Grb2 and specific targets of the SH3 domains as follows (Fig. 3).
First, the activating effect of Grb2 wild type (curve f in Fig.  3A) was totally inhibited by 2.5 M of the Sos-derived peptide dimer VPPPVPPRRR-Aha-K-Aha-RRRPPVPPPV (curve h in Fig. 3A), which binds both SH3 domains (Fig. 3B) and blocks the Ras signaling pathway (52). The pYpYN phosphorylated tripeptide, which mimics the targets of the SH2 domain of Grb2, caused a weak inhibition in the absence as well as in the presence of Grb2. However, activation of N-WASp by Grb2 in itself was not affected by pYpYN tripeptide (compare curves j and k). In conclusion, the SH2 domain of Grb2 is not responsible for the activation of N-WASp, and binding of the SH2 domain of Grb2 to a phosphotyrosine peptide does not affect the affinity of its SH3 domains for proline-rich targets, in agreement with Cussac et al. (55).
The G203R point mutation in the C-terminal SH3 domain of Grb2 is located at the interface of the two SH3 domains (Fig.  3B). It has been shown to impair sex myoblast migration (56). The G203R Grb2 mutant was totally unable to activate N-WASp (curve l in Fig. 3A). This effect may be due to the arginine bulging off the surface of the SH3 domain or to a local structural perturbation.
The P49L point mutation in the N-terminal SH3 domain of Grb2 is known to greatly weaken the interaction between Grb2 and Sos (42). This mutation is lethal in Caenorhabditis elegans (56). However, the P49L Grb2 mutant activated N-WASp (curve g) with an efficiency that varied, among different preparations, between 50 and 100% of the one observed with wild type Grb2. The effect was abolished by the peptide dimer inhibitor. This result suggests that the C-terminal domain of Grb2 is mainly responsible for its interaction with and activation of N-WASp. In agreement with this conclusion, the P206L point mutation in the C-terminal SH3 domain of Grb2 yielded a protein that failed to activate N-WASp (curve m).
Control assays showed that Grb2 does not affect the polymerization of actin alone and that in the absence of N-WASp, Grb2 and P49L Grb2 activate Arp2/3 to a small extent (curves c and d), indicating that in the overall complex comprising N-WASp, Grb2, and Arp2/3, contacts exist between Arp2/3 and N-WASp, Grb2 and N-WASp, and Grb2 and Arp2/3. A similar conclusion was reached concerning the activation of N-WASp by the Shigella protein IcsA (9).
The activation of N-WASp by the isolated N-terminal and C-terminal SH3 domains of Grb2 was tested next. Fig. 4 shows that in the presence of 40 nM N-WASp, half-maximal activation was observed with 0.25 and 2.5 M C-terminal and N-terminal SH3 domains, respectively. In conclusion, the C-terminal domain of Grb2 activated N-WASp with an apparent affinity comparable to that of Grb2 and about 10-fold higher than the N-terminal domain. This result corroborates the conclusions derived from the studies with the P49L and P206L Grb2 mutants. In contrast, in the association of Grb2 with Sos and with dynamin, the N-terminal SH3 domain of Grb2 was found to be the most efficient (57).
Finally, the isolated SH2 domain of Grb2 did not activate actin polymerization in the presence of Arp2/3 and N-WASp. In contrast, a concentration-dependent decrease in the rate of polymerization was observed (data not shown). The inhibitory effect of the SH2 domain was not abolished by the pYpYN peptide, which rules out the possibility that the inhibition be mediated by binding of SH2 to a phospho-tyrosine motif in N-WASp.
Cdc42 and Grb2 Bind to N-WASp Simultaneously, Causing Superactivation of Actin Polymerization-Cdc42 binds to the CRIB domain of N-WASp, and Grb2 binds the proline-rich region. Hence it is theoretically possible that these two effec- tors of N-WASp, which both independently activate the protein, bind simultaneously to the same N-WASp molecule. This possibility was tested in the experiment shown in Fig. 5. Actin was polymerized in the presence of 20 nM Arp2/3, 30 nM N-WASp, and saturating amounts (1 M) of either Cdc42-GTP␥S or Grb2 then in the presence of both effectors together. The same level of Arp2/3 activation was reached with either Grb2 or Cdc42-GTP␥S. A higher activation was obtained when both effectors were added together, indicating that Grb2 and Cdc42 can be bound simultaneously to N-WASp and that the structure/activity of the N-WASp molecule is different when the two effectors are bound, allowing synergistic behavior.
Grb2 Promotes the Onset of Actin-based Motility of E. coli (IcsA)-The actin-based propulsion of E. coli (IcsA) is acknowledged as a model for lamellipodium extension, which is suitable for biochemical analysis, and has recently been reconstituted from a set of pure proteins (3). The effect of Grb2 on actin-based motility of E. coli (IcsA) was tested as described under "Materials and Methods." Briefly, bacteria were first preincubated in the presence of N-WASp to allow tight association of N-WASp with IcsA. N-WASp-coated bacteria were then placed in the motility medium. The binding of N-WASp to IcsA at the surface of E. coli was not inhibited by the presence of Grb2 in the preincubation step. Both the rate of propulsion and actin tail length were unchanged upon addition of Grb2 to the motility medium. On the other hand, the time required for bacteria to start moving, i.e. the pre-steady-state step, was shortened from 30 to 5 min when Grb2 was present at concentrations 1-4 M in the N-WASp-precoating step. These observations, summarized in Table I, indicate first that Grb2 and IcsA do not compete for binding to the same site on N-WASp. Second, the effect of Grb2 on the delay preceding the onset of movement appears linked to the enhanced interaction between N-WASp and IcsA during the precoating of E. coli (IcsA) by N-WASp, suggesting that, in stabilizing the active monomeric form of N-WASp, Grb2 facilitates the interaction of N-WASp with IcsA at the bacterium surface.

DISCUSSION
The present results show that in vitro the SH2/SH3 adaptor Grb2 acts as an effector of N-WASp, leading to the stimulation of actin assembly via N-WASp⅐Arp2/3 complex. The present data, involving sedimentation velocity, EDC cross-linking, together with the concentration dependence of N-WASp activity in actin polymerization and the effect of Grb2 on N-WASp⅐Arp2/3-mediated actin polymerization and actinbased motility of E. coli (IcsA) suggest that the structural basis for N-WASp activity involves at least a monomer-dimer equilibrium, shifted toward the active monomer form by Grb2, according to the minimum scheme displayed in Fig. 6. The data shown in Fig. 2a could be satisfactorily accounted for by this scheme (see "Materials and Methods" for detailed equations), assuming that 1) N-WASp dimerizes with an equilibrium dissociation constant K D of 90 nM; 2) Grb2 binds preferentially monomeric N-WASp; 3) both monomeric and dimeric forms of N-WASp bind Arp2/3 with the same affinity (K Y ϭ 5 nM), independently of Grb2 binding (Fig. 1b, inset), but only the monomeric forms (N-WASp-Arp2/3 and Grb2-N-WASp-Arp2/3) are active in filament branching. This model explains why in the absence of Grb2 full activation of Arp2/3 is not reached upon increasing N-WASp concentration; the formation of the inactive Arp2/3-N-WASp dimer (W 2 Y 2 ) is thought to overlap the saturation of Arp2/3 by monomeric N-WASp. Consistently, Grb2 is more effective at higher N-WASp concentration (Fig.  2a, inset) because it is thought to shift the inactive dimer toward the active monomer form. In the reports published so far, it has been puzzling to observe quantitatively different effects of IcsA and Cdc42 (9) on N-WASp-mediated activation of Arp2/3, as well as additive effects of different ligands (Cdc42 and PIP 2 (6); Cdc42 and Grb2 in the present work). Such data are well accommodated by a monomer-dimer equilibrium of N-WASp that could be shifted to different extents to the active monomer form by ligand binding. The proposal of regulation of N-WASp by self-association is not incompatible with the evidence for an interaction between peptides of the C-terminal domain of N-WASp and peptides of the CRIB domain (26,31). Such an interaction may take place via internal folding of a monomer as well as via monomer-dimer interaction (or a higher degree of self-association), as observed in Ezrin, another linker protein of the cytoskeleton (58). However, only the selfassociation hypothesis can account for the lower activity of N-WASp at high concentration reported here (Fig. 2a). The present functional study is understood as preliminary and calls for more extensive structural studies to understand fully the regulation of N-WASp.
Overall, our results give full biochemical/molecular support to the view (44,23,47) that Grb2 may play a role in actin-based motility, and they suggest that it could act also in vivo by modulating the interaction between N-WASp and Arp2/3 complex. As indicated by She et al. (23), Grb2 could mediate the motile response of cells to EGF by directly linking the tyrosinephosphorylated EGF receptor to the N-WASp⅐Arp2/3 complex. In addition to the PH and the CRIB domains, the proline-rich region of N-WASp, which contains five consensus PPPPPXR motifs, now appears to be another signal-responsive region of this fascinating connector in actin-based motility. It is possible that the other SH2/SH3 domain adaptor Nck, which has recently been demonstrated to be involved in the recruitment of In the absence of N-WASp, Grb2 elicited a weak activation of Arp2/3. It is possible that the ridge of negatively charged amino acids created by the relative positions of the two SH3 domains at the surface of Grb2 participates in electrostatic interactions with Arp2/3 complex, like the acidic C terminus of N-WASp, resulting in stimulation of the nucleating activity. A similar observation has been made with IcsA (9).
Use of mutated forms of Grb2, of specific inhibitors of SH2 and SH3 domains, and of the isolated N-terminal and C-terminal SH3 domains of Grb2 demonstrates that the C-terminal SH3 domain of Grb2 is the main active target of the prolinerich region on N-WASp. This result is surprising when compared with binding studies of Grb2 to the homologous WASp (23), which indicated that the N-terminal SH3 domain of Grb2 bound slightly more strongly than the C-terminal SH3 domain. This discrepancy may be due either to differences in the proline-rich regions of WASp and N-WASp or to the difference in the methods used. Binding studies relied on pull down assays of WASp from whole cell lysates of hemagglutinin-WASp-transfected kidney cells using glutathione S-transferase fusion proteins coupled to glutathione-agarose beads, and here the biological activity of pure proteins or isolated domains was measured.
The physiological significance of multiple SH3 domains in adaptors like Grb2 (two SH3 domains) or Nck (three SH3 domains) has been discussed (43,57). It is generally thought that the presence of several SH3 domains strengthens the binding of the whole protein, even though individual SH3 domains may display differential affinity. The N-terminal SH3 domain of Grb2 is the only one that is essential in mediating high affinity binding to dynamin (54), and it displays a predominant contribution in binding to Sos (37,39), whereas here it is the C-terminal SH3 domain of Grb2 that plays a predominant role in activation of N-WASp. The two SH3 domains of Grb2 may interact differentially in different signaling pathways, making Grb2 a highly versatile adaptor. Our results suggest that transfection of cells with the P49L Grb2 mutant, which blocks the Ras signaling pathway, should not inhibit the function of Grb2 in actin-based motility.
The N-WASp molecule can be activated by Cdc42 and Grb2 in a synergistic fashion. Similarly, PIP 2 and Cdc42 have been shown to cooperate in activating N-WASp (6). These results support the view that the multidomain organization of N-WASp has functional significance. PIP 2 and Grb2 also cooperate in activation of dynamin GTPase (43).
The possible regulation of the activity of N-WASp in vivo by  (IcsA) in the reconstituted motility assay In the pre-coating step, bacteria were incubated for 15 min at room temperature in the presence of 0.1 M N-WASp and in the absence or presence of Grb2 at the indicated concentration. Bacteria were then centrifuged and resuspended in buffer before being placed in the motility medium that was supplemented or not with Grb2 at the indicated concentration. Samples were prepared for phase contrast optical microscopy observation. The lag time before the onset of movement was recorded. Rate measurements were performed as described (3) (13) FIG. 6. Thermodynamic scheme describing the structural basis for N-WASp-Arp2/3 activation and its regulation by Grb2. The model proposes that N-WASp (W) undergoes monomer-dimer equilibrium 2W^232 W 2 , with K D ϭ 90 nM. Both monomer and dimer forms bind Arp2/3 (Y) with the same affinity K Y ϭ 3 nM, and K YY ϭ (K Y ) 2 . Only the complex of monomeric N-WASp with Arp2/3 (WY) is active to stimulate actin polymerization by filament branching. Grb2 (X) binds preferentially the monomeric form of N-WASp with K X ϭ 5 nM, and the dimeric form of N-WASp with a 32-fold lower affinity (K XX ϭ 5000 (nM) 2 ). The WX complex associates with Arp2/3 to form a WXY active complex. The net effect of Grb2 on the stimulation of actin polymerization is to increase the proportion of N-WASp in the active monomer state (WXY). The active complexes of monomeric N-WASp and Arp2/3 (WY and WXY) are boxed.
Grb2 is an open issue. Since both Grb2 and N-WASp are soluble proteins, the simple in vitro approach suggests that Grb2 could activate N-WASp and stimulate actin polymerization in the cytosol, whereas in principle in the cell N-WASp is activated at the plasma membrane by signaling. It is possible that in the in vivo context, the cytosolic N-WASp is maintained inactive, perhaps in a self-associated form, and is unable to bind Grb2, and it binds Grb2 when it is recruited to the membrane by other effectors. The possible regulation of N-WASp by phosphorylation of a tyrosine in the N-terminal domain has not been considered yet. Further mutagenesis and biochemical and structural studies of N-WASp are required to understand fully the complex behavior of this important connector between signaling molecules and actin.