Integration of Multiple Signals Through Cooperative Regulation of the N-WASP-Arp2/3 Complex

See allHide authors and affiliations

Science  27 Oct 2000:
Vol. 290, Issue 5492, pp. 801-806
DOI: 10.1126/science.290.5492.801


The protein N-WASP [a homolog to the Wiskott-Aldrich syndrome protein (WASP)] regulates actin polymerization by stimulating the actin-nucleating activity of the actin-related protein 2/3 (Arp2/3) complex. N-WASP is tightly regulated by multiple signals: Only costimulation by Cdc42 and phosphatidylinositol (4,5)-bisphosphate (PIP2) yields potent polymerization. We found that regulation requires N-WASP's constitutively active output domain (VCA) and two regulatory domains: a Cdc42-binding domain and a previously undescribed PIP2-binding domain. In the absence of stimuli, the regulatory modules together hold the VCA-Arp2/3 complex in an inactive “closed” conformation. In this state, both the Cdc42- and PIP2-binding sites are masked. Binding of either input destabilizes the closed state and enhances binding of the other input. This cooperative activation mechanism shows how combinations of simple binding domains can be used to integrate and amplify coincident signals.

Many cellular processes are controlled by networks of interacting signaling pathways (1, 2). For example, during directed cell motility, multiple pathways converge to precisely target actin polymerization to the cell's leading edge. Little is known, however, about the molecular mechanisms by which the relevant signaling proteins integrate these multiple inputs to yield a coordinated response.

WASP and its homolog N-WASP link multiple signaling pathways to actin assembly (3–5). N-WASP interacts with the Arp2/3 complex and activates its ability to nucleate actin filaments (5,6). Activation only occurs, however, when N-WASP is stimulated by the proper set of upstream signals. The two best characterized inputs are the rho family guanosine triphosphatase (GTPase) Cdc42 and phosphatidylinositol 4,5-bisphosphate (PIP2), both of which are regulated by upstream pathways critical for motility (1, 7, 8). Individually, Cdc42 and PIP2 are weak activators of N-WASP. Together, however, the inputs act synergistically: costimulation with low concentrations of both yields potent activation (5). Thus, N-WASP acts as a signal integration device that can precisely target actin polymerization to sites on the membrane at which both PIP2 and activated Cdc42 are present.

Several domains within N-WASP have been implicated in this signal processing behavior (Fig. 1A). Output is controlled by a COOH-terminal domain that directly interacts with and activates the Arp2/3 complex (5). This domain is referred to as the VCA domain (also called WWA) because it has a verprolin homology motif (V), a cofilin homology motif (C), and an acidic motif (A). The acidic motif binds Arp2/3, whereas the verprolin motif binds actin monomers, probably delivering them to Arp2/3 (5,6). The isolated VCA domain is constitutively active; however, this activity is suppressed in full-length N-WASP, indicating that NH2-terminal regions play a regulatory role (5). One key regulatory domain is the GTPase-binding domain (GBD), which forms an intramolecular interaction with the cofilin homology motif (Fig. 1A) (9, 10). Because activated Cdc42 can disrupt the intramolecular interaction by binding the GBD, a simple model for N-WASP regulation has been proposed: the intramolecular interaction between the GBD and the cofilin motif blocks binding of Arp2/3 to the VCA domain, and Cdc42 relieves this autoinhibitory interaction (3, 5, 9). There are several problems with this model. First, the GBD has never experimentally been shown to block Arp2/3 binding or to suffice as a functional repressor of the VCA domain. Second, this simple autoinhibition model fails to explain how PIP2 is detected as an input, and how its effects are synergistically integrated with those of Cdc42.

Figure 1

Both the basic motif and GBD are required for repression and regulation of N-WASP. (A) N-WASP contains the following domains: EVH1 domain, basic motif (B), GBD, proline-rich motif (Pro), and verprolin/cofilin/acidic domain (VCA). Curved arrow indicates intramolecular interaction between GBD and the cofilin motif (within the VCA domain) (9, 10). (B) Repression of VCA-Arp2/3–mediated actin polymerization by control region fragments (25), tested using in vitro pyrene actin polymerization assay (26). Assays contain 50 nM VCA, 50 nM Arp2/3, and 2.5 μM actin (2% pyrene actin), along with the indicated concentration of variable fragment. Binding of the VCA domain to these control region fragments fused to GST (27) is shown below. (C) Mini–N-WASP (178–274-GSGSGSGSG-392–501) mimics the regulatory behavior of N-WASP (28). Pyrene actin polymerization assays are shown for mini–N-WASP alone and mini–N-WASP plus Cdc42·GTPγS (0.15 μM), PIP2 (∼0.4 μM in PS:PC:PIP2 vesicles at 48:48:2), or both (same concentrations as above). Activity of VCA domain alone (dotted line) is shown for comparison. Assays contain 50 nM mini–N-WASP, 50 nM Arp2/3, and 2.5 μM actin. Bar graph shows maximal polymerization rate for each assay.

To elucidate the mechanism of N-WASP regulation and signal integration, we identified the minimal domains required to repress the VCA domain, mapped their interactions, and determined how they communicate. We found that two adjacent regulatory domains, the GBD and a novel PIP2-binding motif, are necessary and sufficient for proper repression and regulation of N-WASP. In the absence of stimuli, the two regulatory motifs together lock the VCA-Arp2/3 complex in an inactive “closed” conformation. The mechanism of repression allows for highly cooperative activation: Cdc42 and PIP2disrupt the closed state in a thermodynamically coupled fashion, providing the basis for potent signal integration by N-WASP.

VCA domain activity was potently repressed (inhibition constantK i = ∼1 μM) by a minimal fragment containing both the GBD and an adjacent, highly basic motif in an in vitro actin polymerization assay (Fig. 1B). We refer to this composite domain as the “control region.” The basic motif is only ∼20 residues in length and contains nine lysine residues (11). In contrast, a second fragment consisting solely of the GBD, although able to interact strongly with the VCA domain, had virtually no inhibitory effect (Fig. 1B), even at concentrations (100 μM) well above saturation (12). The GBD failed to inhibit even when covalently tethered (cis) to the VCA domain (13). A third fragment that contains the basic motif and only half of the GBD was also unable to repress (Fig. 1B). These data show that current models for autoinhibition are incorrect: although required, the GBD alone is insufficient to repress VCA activation of the Arp2/3 complex; rather, the composite control region is the minimal repressive element.

Remarkably a “mini–N-WASP” that contains only the control region and the VCA domain attached by a nine-residue linker is also sufficient to recapitulate the hallmark regulatory behavior of N-WASP. Its Arp2/3 stimulatory activity is highly repressed but can be synergistically activated by costimulation with PIP2 and Cdc42 complexed with guanosine 5′-O-(3′-thiotriphosphate) (Cdc42·GTPγS) (Fig. 1C) (14).

Contrary to previous models, we found that neither the control region nor GBD blocks binding of Arp2/3 to the VCA domain (Fig. 2A) (15). Instead, the basic motif participates in two previously uncharacterized interactions critical for regulation (Fig. 2, B through D). First, the basic motif is an essential part of a novel Arp2/3 binding site (Fig. 2B). Although this Arp2/3 interaction region appears to be crucial for repression, it is not sufficient, because a fragment that binds Arp2/3 but lacks the intact GBD (residues 178 through 244) also fails to inhibit VCA activity (Fig. 1B). Arp2/3 therefore interacts with two sites within N-WASP: the acidic motif within the VCA domain and the basic motif within the control region. Second, we found that the basic region is also a novel phospholipid-binding module that specifically recognizes PIP2 (Fig. 2C). This 20-residue motif is distinct from larger canonical phospholipid-binding domains, such as pleckstrin homology (16) or FYVE finger domains (17).

Figure 2

Repression of the N-WASP–Arp2/3 complex requires a network of concerted interactions. (A) Arp2/3 binding to the VCA domain is not blocked by the GBD (residues 196–274) or control region (178–274). Arp2/3 (1 μm) can bind to VCA (10 μm) precomplexed with a GST-GBD fusion (27). Arp2/3 can also bind GST–mini-N-WASP, which contains the full control region and is repressed. (B) Direct binding of Arp2/3 to GST fusions of control region fragments (27). (C) PIP and PIP2 vesicle spin-down binding assays (29). S indicates supernatant (unbound) and P indicates pellet (bound). Control region (residues 178–274) can selectively bind PIP2whereas the GBD alone (residues 196–274) cannot. (D) Summary of control region and output region interactions mapped in this study [complete supplementary data is given in Web fig. 1 (30)]. Solid lines are essential regions; dotted lines are important but nonessential regions. Also shown is a cartoon of the repressed, “closed” state of the N-WASP–Arp2/3 complex (31). The interactions in the closed state could repress by inducing a conformational change in Arp2/3 and/or the VCA domain, or a change in the arrangement of the two components. Disruption of these interactions allows conversion to the active, “open” state. (E) Artificial disruption of the repressive network leads to activation. Control region subfragments consisting of either the GBD alone, or the basic motif with a truncated GBD (+B-G), activate mini–N-WASP. Actin polymerization assays were performed with 80 nM mini–N-WASP ± 10 μM of either subfragment. The postulated mechanism of artificial activation is shown.

The interactions of the control region (Fig. 2D) and the requirements for repression (Fig. 1B) support a concerted repression model. Neither the GBD nor the basic motif is intrinsically repressive; rather, each alone is a neutral anchor point that, only when topologically linked to and acting in concert with the other, locks the N-WASP–Arp2/3 complex in an inactive “closed” state (Fig. 2D). These interactions may alter the conformation of Arp2/3 and/or its relationship with the VCA domain, rendering it inactive. This model is supported by the finding that when the GBD and basic modules are added together, but as covalently distinct elements, they fail to repress VCA domain activity (13). Moreover, these control region subfragments are found to activate mini–N-WASP (Fig. 2E), behavior incompatible with subfragments that are intrinsically repressive (these would repress or be neutral). The behavior is best explained as uncoupling of concerted interactions required for repression (Fig. 2E).

The novel mechanism of N-WASP repression suggests a reciprocal mechanism for activation by both Cdc42 and PIP2. These inputs bind the GBD and basic motifs, respectively, and thus either could disrupt the “closed” state and release the active VCA-Arp2/3 complex. Activated Cdc42 is known to disrupt the intramolecular GBD-VCA interaction (9, 10). We find that Cdc42 also disrupts the control region-Arp2/3 interaction (Fig. 3A). However, this alone cannot explain synergistic activation by PIP2 and Cdc42.

Figure 3

Cdc42 and PIP2 bind cooperatively to N-WASP when competing against repressive interactions. (A) Cdc42·GTPγS inhibits Arp2/3 binding to a GST–control region fragment (27). (B) PIP2 vesicle binding assays (29) show that Cdc42·GTPγS and PIP2 can simultaneously bind the control region (B-GBD). S indicates supernatant (unbound) and P indicates pellet (bound). Cdc42·GTPγS does not bind to PIP2 vesicles in the absence of the control region (not shown). (C) Cdc42·GTPγS enhances mini–N-WASP binding to PIP2. In the absence of Cdc42, mini–N-WASP binds poorly to PIP2, most likely because the GBD-VCA interaction sterically masks the PIP2-binding site. High-affinity PIP2 binding is observed if Cdc42 is present, or if the VCA domain is removed. The cooperative effects between Cdc42 and PIP2 are not observed for binding to the control region alone (B-GBD). (D) PIP2 enhances apparent affinity of Cdc42 for the control region–Arp2/3 complex. Dissociation constants for the control region interaction with fluorescently tagged Cdc42 (32), either alone (K d) or with additional factors (K d app), were measured as described in Web fig. 2 (30). We added 2.5 μM of Arp2/3 or 10 μM of PIP2 (PC:PS:PIP2 at ratios of 48:48:2), or both. The presence of Arp2/3 significantly decreases the apparent affinity, as expected, because Arp2/3 competes against Cdc42 for binding to the control region (A). However, addition of PIP2 with Arp2/3 restores the higher apparent affinity, indicating that PIP2 and Cdc42 cooperate to compete against Arp2/3 binding. Cooperation is not observed in the absence of Arp2/3.

Synergistic activation could be explained, however, if PIP2and Cdc42 act cooperatively to disrupt the closed state. Therefore, we tested if binding of the two inputs is thermodynamically coupled. Vesicle-binding studies show that PIP2 and Cdc42 can bind the control region simultaneously (Fig. 3B), and that in the context of mini–N-WASP, Cdc42 can dramatically enhance PIP2 binding (Fig. 3C). Moreover, parallel fluorescence binding studies show that PIP2 can enhance Cdc42 binding to the control region, in the presence of Arp2/3 (Fig. 3D). Thus, cooperative binding of PIP2 and Cdc42 to N-WASP has been observed.

However, this observed cooperativity discussed above cannot result from direct interaction between Cdc42 and PIP2, because the two inputs do not bind cooperatively to the isolated control region (Fig. 3, C and D). Instead, cooperativity must result from coordinated competition with other control region (B-GBD) ligands. For example, interaction of the control region with the VCA domain, present in mini–N-WASP, appears to mask both the PIP2- and Cdc42-binding sites (Fig. 3C). Thus, Cdc42 strongly enhances PIP2 binding in this context. Arp2/3 binding to the control region also appears to mask both the Cdc42- and PIP2-binding sites, explaining the cooperative effects of the two inputs in this context (Fig. 3D). In summary, it is the structure and interactions within the closed state that makes binding of the two inputs highly cooperative.

The activation behavior of N-WASP can thus be modeled, using the thermodynamic cycle shown in Fig. 4A. In the repressed state, a network of interactions holds the N-WASP–Arp2/3 complex in a closed state. Cdc42 and PIP2 can individually disrupt this network and activate the complex. However, because their binding sites are masked by repressive interactions, binding of either input alone to the closed state is relatively weak. In contrast, if one input molecule is prebound, the closed state is destabilized, and binding of the second molecule is considerably enhanced. The degree to which binding of one ligand enhances binding of the other is determined by the cooperativity factor, c. This behavior is analogous to that of any cooperative binding protein, such as hemoglobin (18, 19), although in this case, cooperativity is observed between heterotropic rather than homotropic ligands. Nonetheless, both mechanisms involve increasing stabilization of an open or relaxed state by binding of successive ligands.

Figure 4

Highly cooperative activation mechanism of N-WASP allows for potent signal integration. (A) Thermodynamic cycle modeling states of Cdc42 (green) and PIP2 (purple) binding to mini–N-WASP. Binding of a second input molecule is more favorable, because the energetic cost of disrupting the closed state is paid by binding of the first input molecule. The degree of enhancement in binding is given by the cooperativity factor, c. (B) Theoretical curves indicate the concentration of two input molecules required to achieve 50% activation of a protein switch such as that modeled in (A) (33). Intercepts give the dissociation constants for binding of each input molecule alone. Dependence of the curves on the cooperativity constant, c, is also shown, as are experimentally determined concentrations of input molecules required to achieve 50% activation of mini–N-WASP (open circles). Each point represents a series of five or six actin polymerization assays in which the concentration of one activator is fixed and the concentration of the second activator is varied [Web fig. 3 (30)]. The 50% activation points (K act) are determined based on maximal rates of actin polymerization. Data fit the model with c = ∼350. (C) Cooperativity provides a mechanism for signal integration. Top graph shows hypothetical, spatially overlapping concentration gradients of Cdc42·GTP (green) and PIP2(purple). The bottom graph shows the calculated response of N-WASP (fraction maximal activity) assuming the model in (A) and a cooperativity of either 1 or 350. With high cooperativity, coincident signals are integrated and amplified.

This framework reveals that a highly cooperative activation mechanism allows for potent signal integration. A series of simulated contours indicating the concentrations of two activating inputs required to achieve 50% activation is shown in Fig. 4B. If two input ligands are completely independent (c = 1), their combined effects are nearly additive and the 50% activation contour will be close to linear. In contrast, if c is high, the two ligands will act synergistically. The 50% activation contour will be concave, because when added together, significantly lower concentrations of each ligand are required for activation.

We experimentally measured activation of mini–N-WASP as a function of both Cdc42 and PIP2 concentration (Fig. 4B). High concentrations of either input alone (∼3 μM Cdc42 or ∼8 μM PIP2) are required to yield 50% activation. However, costimulation with 10-fold less of each input yields the same degree of activation. The observed behavior closely fits that predicted by our model with a cooperativity factor of >100 linking Cdc42 and PIP2 activation (20).

These results indicate that N-WASP can exist in a primed state that, although repressed, is preloaded with Arp2/3 and ready for immediate actin filament assembly upon activation (21). The high cooperativity by which this state of N-WASP is activated allows it to detect and amplify weak but coincident signals of both Cdc42 and PIP2 (Fig. 4C) (22). Thus, N-WASP approximates a coincidence detector or a logical “AND” gate, devices whose output is dependent on stimulation by the proper combination of inputs (23). The mechanism of N-WASP regulation reveals general principles by which simple protein interaction modules, if combined in the proper cooperative fashion, can yield a sophisticated signal-integrating machine.

Note added in proof: Rohatgi et al. (34) have recently also identified the basic motif of N-WASP as the PIP2 responsive element.

  • * To whom correspondence should be addressed. E-mail: wlim{at}


View Abstract

Navigate This Article