The nucleotide switch in Cdc42 modulates coupling between the GTPase-binding and allosteric equilibria of Wiskott–Aldrich syndrome protein

Methods

Constructs and Proteins. Preparation and purification of WASP constructs and Cdc42 (residues 1–179) loaded with guanosine 5′-[β,γ-imido]triphosphate (GMPPNP) have been described elsewhere (9–11). Arp2/3 complex was isolated from bovine thymus (12). Actin was purified from rabbit skeletal muscle (13).

Chemical Denaturation. CD measurements (222 nm) were obtained on a Aviv Associates (Lakewood, NJ) 62DS spectrometer at 25°C by using 10 μM protein samples and urea or guanidine hydrochloride as denaturant. Chemical denaturation experiments were performed in 25 mM phosphate, pH 7/150 mM NaCl/2 mM MgCl₂/2 mM DTT. Experimental data were fit to a six-parameter equation (14) to obtain ΔG of folding and m. Where performed, identical ΔG values were obtained for urea and guanidine hydrochloride denaturation.

Fluorescence Binding Assays. Binding affinity of WASP constructs for Cdc42 loaded with a fluorescent nonhydrolyzable GTP analog, mantGMPPNP, was measured on a Horiba Jobin Yvon Fluorolog 3-11 fluorimeter. Binding was followed by monitoring the quenching of mant fluorescence at 432 nm (λ_ex = 350) upon addition of WASP protein. Experiments were performed in 25 mM phosphate, pH 7/150 mM NaCl/2 mM MgCl₂/2 mM DTT. Equilibrium dissociation constants (K_D) were obtained by fitting the data to a quadratic equation for a single-site binding isotherm (15).

Binding affinities of Arp2/3 for I290Q and basic region GBD (BGBD)–VCA were measured by using a rhodamine anisotropy fluorescence assay. A C-terminal cysteine residue was introduced by mutagenesis to I290Q and BGBD–VCA to facilitate labeling with tetramethylrhodamine 5′-malemide. Rhodamine fluorescence was monitored at 574 nm (λ_ex = 552 nm). Rhodamine-labeled I290Q and BGBD–VCA displayed higher florescence anisotropy upon addition of Arp2/3 complex in the presence of 4 μM actin-latrunculin. Experiments and analysis were performed in the same buffers as above.

Because mant fluorescence can interfere with pyrene fluorescence in pyrene-actin polymerization assays and because mant labeling decreases the affinity of Cdc42 for WASP (16), we also measured affinity of Cdc42–GMPPNP and Cdc42–GDP for WASP proteins by using a previously described FRET-based assay as a prelude to the actin polymerization measurements below (17). Briefly, WASP proteins were fused to cyan-fluorescent protein (CFP) and yellow fluorescent protein at their N and C termini, respectively. Binding of Cdc42 was measured through changes in the ratio of yellow fluorescent protein emission (526 nm) to CFP emission (476 nm) after CFP excitation (λ_ex = 433 nm). Experiments were performed in 20 mM Hepes, pH 7.3/150 mM KCl/2 mM MgCl₂/2 mM DTT. Data were fit to a quadratic equation for a single-site binding isotherm (17).

Pyrene Actin Polymerization Assays. Actin polymerization assays were performed as described in ref. 12. Pyrene fluorescence at 407 nm (λ_ex = 365 nm) was monitored over time. Polymerization assays were performed in 10 mM imidazole, pH 7/50 mM KCl/1 mM MgCl₂/10 mM EGTA and contained 4 μM actin (6% pyrene), 10 nM Arp2/3, and 500 nM WASP proteins. Fractional activity is determined from the concentration of filament barbed ends calculated for each WASP construct relative to the isolated VCA domain as described in ref. 12.

NMR Acquisition. NMR data were collected on 500- or 600-MHz Varian Innova spectrometers at 25°C. To probe the structural details of WASP in complexes with Cdc42, ¹H/¹⁵N transverse relaxation-optimized spectroscopy (TROSY) heteronuclear sequential quantum correlation (HSQC) spectra were acquired on 100 μM ²H/¹⁵N GBD–VCA protein in 10 mM imidazole, pH 7/50 mM KCl/1 mM MgCl₂/10 mM EGTA in the presence of either 500 μM Cdc42–GMPPNP or 750 μM Cdc42–GDP. To determine the binding interface between Cdc42–GDP and GBD–VCA, ¹H/¹⁵N TROSY HSQC spectra were acquired on a 100 μM sample of ²H/¹⁵N Cdc42–GDP in 20 mM Tris, pH 7/100 mM NaCl/2 mM MgCl₂/5 mM DTT in the presence of increasing concentrations of GBD–VCA (up to 300 μM). Backbone ¹HN, ¹⁵N, ¹³Cα, ¹³Cβ, and ¹³CO assignments were obtained for the Cdc42–GDP domain through the following ²H-decoupled, triple resonance spectra appliedtoa ²H/¹⁵N/¹³C-labeled sample: CT-HNCa, CT-HN(CO)Ca, HN(Ca)Cb, and HN(COCa)Cb (18).

All data sets were processed identically with nmrpipe (19) and analyzed with nmrview (20).

Formulation of the Two-State Allosteric Model. The development of a two-state allosteric model for WASP has been described in detail in ref. 10. Briefly, the fraction of WASP proteins in the active state is described by

[1]

where

[2]

[A] is the concentration of activator (Cdc42); [WASP_i,0] and [WASP_a,0] are the concentrations of inactive and active WASP in the absence of activator; K_D,a and K_D,i are the binding affinities of WASP_a and WASP_i for activator, respectively; ΔG_eq is the free energy difference between the inactive and active states of WASP, which we approximate by the folding stability of the autoinhibited domain; and R and T are the ideal gas constant and temperature. In the absence of any activator, Eq. 1 simplifies to

[3]

In the presence of infinite activator, Eq. 1 can be described by

[4]

The measured binding affinities reflect an equilibrium mixture of inactive and active WASP states and can be described by the following equation (10):

[5]

Results

Development of a Two-State Model for Allostery in WASP. To understand the activation of WASP by Cdc42, we have developed a simple two-state model of WASP based on classical descriptions of allostery (10, 21, 22) (Fig. 2). In such models, the state of the system can be quantitatively described by only three parameters: K_eq, the equilibrium constant between the inactive and active states of the free protein (WASP); C, the ratio of affinities of activator (Cdc42) for the active and inactive states of WASP (K_D,a/K_D,i); and A, the concentration of activator. Binding is coupled to activation through the product K_eq*C, which is the equilibrium constant between the inactive and active states of the saturated protein–activator complex. Thus, activator binding shifts the equilibrium by the amount C. Briefly, our model proposes, based on a large body of biophysical data, that the allosteric equilibrium in WASP can be quantitatively represented by the folding equilibrium of the regulatory domain formed by intramolecular interactions between the GBD and the activity-bearing (toward the Arp2/3 complex) VCA elements of the protein. Additionally, the parameter C can be determined from the affinities of Cdc42 for engineered WASP proteins that are significantly stabilized and significantly destabilized relative to wild type. Previous work showed that the model quantitatively predicts the hydrogen exchange behavior and Cdc42–GTP binding affinity of WASP proteins over a range of folding stabilities (K_eq values) (10). To extend this work across a wider range of K_eq values and, more importantly, to examine the activity of different WASP proteins toward Arp2/3 complex, we generated a series of WASP constructs by introducing mutations in the GBD outside of the CRIB motif (Figs. 1 and 3). Proteins in this series have folding stabilities ranging from –5.8 to more than –1.5 kcal mol^–1 (1 cal = 4.18 J) (K_eq = 17,000 to <12), including M307A, which has a folding stability comparable to the protein most representative of full-length WASP (GBDPVCA) (Table 1) and two mutants, I294A and I290Q, whose stabilities against chemical denaturation were too low to measure accurately.

Cdc42-GTP Affinity for WASP Constructs. The binding affinity of each WASP construct for active Cdc42 was determined by monitoring the fluorescence quenching of mantGMPPNP-loaded Cdc42 in a titration (Fig. 4). The measured binding affinities displayed an inverse correlation with protein folding stability. For example, the most stable BGBD-C protein, which approximates the inactive state of WASP, has ≈500-fold lower affinity for Cdc42 than the least stable protein I290Q, which emulates the fully active state of WASP. Using the measured values of K_D,a (I290Q) and K_D,i (BGBD-C) to estimate C_Cdc42,GTP yields: C_Cdc42,GTP = K_D,a/K_D,i = 1.73 × 10^–3. The greatest changes in K_D occur in the transition region between stabilities of –4 and –1 kcal mol^–1. Outside of this region, changes in K_D are very small. The data can be fit well to the equation relating K_D to ΔG (r² = 0.97), yielding a C_Cdc42,GTP value of 2.35 × 10^–3, similar to that determined from the endpoints alone, 1.73 × 10^–3, and to the value of 3.4 × 10^–3 previously measured under slightly different conditions (10).

Activation of Arp2/3 Complex by WASP Constructs. We next examined WASP activity toward Arp2/3 complex as a function of K_eq by using a pyrene-actin assembly assay. As in Cdc42 binding, the WASP mutants have a range of Arp2/3-stimulatory activities (Fig. 5 A and C), which correlate inversely with folding stability. Highly destabilized proteins, such as I290A, display higher levels of activity (>0.5 relative to VCA), whereas more stable proteins, such as M307A, have significantly lower activity (<0.1). In addition, a second series of WASP constructs containing the basic sequence of amino acids [basic region (BR)] immediately N-terminal to the GBD (Fig. 1) was generated because this element has been shown to contribute to autoinhibition in the WASP homolog, N-WASP (23, 24). For all mutants, we find that addition of the BR increases the folding stability of WASP by ≈0.5 kcal·mol^–1 when compared with non-BR WASP constructs (Table 1). This increase in stability causes a concomitant decrease in activity toward the Arp2/3 complex, such that BR and non-BR proteins fall on essentially the same fitted curve (see below) (Figs. 5 B and C). This result suggests that the BR decreases WASP activity through modulating the interaction between GBD and VCA rather than through an independent mechanism involving direct interaction with the Arp2/3 complex (23). Thus, WASP activity toward the Arp2/3 complex is limited by the folding stability of the autoinhibited domain and reflects the equilibrium between inactive and active states.

To evaluate these data in terms of our two-state model, we first tested whether the Arp2/3 complex itself could bias the WASP allosteric equilibrium (i.e., whether C_Arp2/3 ≠ 1) (Supporting Methods, which is published as supporting information on the PNAS web site). By using a fluorescence binding assay, we measured the Arp2/3 complex affinity of rhodamine-labeled I290Q and BGBD–VCA, the active WASP proteins with the lowest and highest folding stabilities. Affinity was measured in the presence of 4 μM actin-latrunculin to approximate the conditions of the assembly assay. These measurements determined that C_Arp2/3 = 0.053. By using this value, the data in the assembly assay fit well (r² = 0.97) (Fig. 5C) to the equation describing the fraction of the activated Arp2/3 complex and actin bound to active WASP, yielding a binding affinity of active WASP for Arp2/3 and actin (K_D(Arp2/3),a = 21 nM) similar to that measured in a titration of VCA activity toward the Arp2/3 complex (data not shown) (K_D(Arp2/3),a = 85 nM). Thus, the two-state model also reasonably describes WASP activity toward the Arp2/3 complex as a function of the allosteric equilibrium.

The Cdc42 Nucleotide Switch Modulates Coupling Between the GTPase-Binding and Allosteric Equilibria in WASP. Our model predicts that activators will shift the WASP allosteric equilibrium toward the activated state according to their C parameters (Fig. 5C). The magnitude of C dictates the maximum increase in activity that can be achieved by a single agonist. Fig. 6A demonstrates that for several mutants, activity toward the Arp2/3 complex is a linear function of the percent saturation of WASP by Cdc42, as expected. Highly destabilized WASP proteins, such as I294A, are predominantly unfolded so that the addition of Cdc42–GMPPNP has little or no further effect on the equilibrium or protein activity (Fig. 6A). However, addition of increasing amounts of Cdc42–GMPPNP can achieve full activation in even the most stabilized WASP proteins, such as GBD–VCA and M307A. Thus, C_42,GTP, when coupled with C_Arp2/3, provides sufficient driving force to fully overcome WASP autoinhibition for K_eq ≤ 1,200.

To examine the biophysical basis of the Cdc42 nucleotide switch, we applied the allosteric model to the GDP state of the GTPase. We measured the affinity of Cdc42–GDP for the most stable (50 μM) and least stable (2.7 μM) mutants, by using a FRET-based fluorescence assay (17), yielding a value of C_42,GDP = 5.4 × 10^–2. This is 5-fold less than the value measured for Cdc42–GMPPNP using the same assay (K_D,i = 2.6 μM; K_D,a = 29 nM; C_42,GTP = 1.1 × 10^–2). This result suggested that the GDP form of Cdc42 should be a poorer activator of WASP than the GTP form of Cdc42 at comparable levels of WASP saturation. Consistent with this prediction, addition of Cdc42–GDP to polymerization assays did increase the rates of actin polymerization achieved by each WASP mutant but to lower levels than achieved by Cdc42–GMPPNP (data not shown). Parallel titrations of Cdc42–GMPPNP or Cdc42–GDP to M307A show that Cdc42–GMPPNP can stimulate maximum WASP activity at saturation whereas Cdc42–GDP can only partially activate WASP even at saturation (Fig. 6B and Fig. 9, which is published as supporting information on the PNAS web site). Thus, the coupling of the GTPase binding equilibrium to the allosteric equilibrium in WASP is less efficient for Cdc42–GDP than for Cdc42–GMPPNP.

To better understand the structural details of WASP interaction with Cdc42, we acquired ¹H/¹⁵N TROSY HSQC spectra of ²H/¹⁵N GBD–VCA free and in the presence of Cdc42–GMPPNP and Cdc42–GDP (Fig. 7). In free GBD–VCA, peaks corresponding to the folded autoinhibited core are well dispersed in the 2D plane (Fig. 7A). Peaks representing unstructured elements, including the CRIB motif and the V and A regions of the VCA domain, are clustered in the center of the spectrum with high intensity. Addition of Cdc42–GMPPNP to 99% saturation (Fig. 7B) causes substantial broadening of the resonances representing the folded core, including the upfield-shifted indole ring HN resonance of Trp-252, which is a good indicator of the structured autoinhibited domain (9, 25). Severe global broadening of the WASP resonances is consistent with a conformational equilibrium between folded and unfolded states on the surface of the GTPase. We also observe the appearance of downfield-shifted peaks characteristic of the bound CRIB motif (10, 11), suggesting that this element anchors the protein to the GTPase. In contrast, in the presence of 90% saturating Cdc42–GDP (limited by Cdc42 solubility) (Fig. 7C), peaks representing the autoinhibited core are virtually unaffected, indicating that the folded structure is heavily populated on the surface of the GTPase. Resonances from the CRIB motif disappear from the center of the spectrum, and some new downfield shifted peaks are now observed. In the reverse titration of unlabeled GBD–VCA into ²H/¹⁵N Cdc42–GDP, significant line-broadening at substoichiometric ligand is observed primarily for peaks corresponding to the Cdc42 α1-helix, β2/β3-hairpin, and α5-helix (Fig. 8B). These elements are all close to the bound CRIB motif in the Cdc42–GTP–GBD complex structure (Fig. 8A), suggesting that this element is similarly bound in the Cdc42–GDP–GBD–VCA complex. Most resonances corresponding to the switch 1 and switch 2 regions of Cdc42–GDP are not observed for either the free protein or the GBD–VCA complex because of exchange broadening. Thus, it is unclear how extensively the bound GBD–VCA contacts these elements. These results suggest that the CRIB motif interacts with the GTPase in a similar manner in the Cdc42–GTP and Cdc42–GDP complexes. In both cases, the autoinhibited domain is likely in equilibrium on the Cdc42 surface between the folded and unfolded states, with the latter probably bound as observed in the Cdc42–GTP–GBD structure (Fig. 8A). In both complexes, the folding equilibrium of WASP is shifted toward the unfolded state relative to the free protein. But because of their different C parameters, the two nucleotides lead to different proportions of the folded and unfolded states, with GDP favoring the folded state and GTP favoring the unfolded state. Thus, Cdc42–GTP is a full agonist, whereas Cdc42–GDP is only a partial agonist.

Discussion

Our data here and reported previously (10) indicate that the activity and Cdc42-regulation of WASP can be explained accurately and quantitatively by a two-state model. Predictions of the model that WASP's affinity for Cdc42, activity toward Arp2/3 complex, and activation by Cdc42 are functions of the allosteric equilibrium in WASP are borne out experimentally. The nucleotide switch in Cdc42 modulates not only the absolute affinity of the GTPase for WASP, but also the ability of the GTPase to distinguish between the two states of WASP. That is, the C parameter of the allosteric model is nucleotide-dependent and smaller for Cdc42–GTP than Cdc42–GDP. This property leads to the unexpected result that Cdc42–GDP can only partially activate WASP, even at saturation, whereas Cdc42–GTP can activate fully. With the truncated WASP protein used in these studies, we observe a 2-fold difference in activity between the two nucleotide states of Cdc42, as predicted by our model (Fig. 5C). The model also predicts that this difference will increase for more stable WASP proteins or when WASP is stabilized through interactions involving the BR or perhaps other binding partners, such as WIP (26, 27), with a maximum differential of C_42,GDP/C_42,GTP = 5-fold (Eq. 12 in Supporting Methods).

The biophysical basis of the nucleotide switch is generally thought to be simply a change in GTPase affinity for effectors. Our data here point toward a more complex mechanism involving modulation of the coupling (or linkage) between two equilibria, the Cdc42–WASP binding equilibrium, and a conformational equilibrium in WASP. What is the physical basis for the stronger coupling of Cdc42–GTP relative to Cdc42–GDP? Hatley et al. (28) recently attempted to identify the critical residue interactions that mediate communication between the nucleotide and effector binding sites in the G protein family by using an approach termed Statistical Coupling Analysis. This method identifies coevolving sequence positions in large alignments of conserved domains. The authors argue that such coevolution is driven by energetic coupling of residues in 3D structures of the domain and, thus, that clusters of mutually coevolving sites represent energetically cooperative units. In the case of G proteins, they find a large cluster of coevolving residues that surround the nucleotide-binding pocket and effector-binding elements. In the active GTP-bound structures of several GTPases, including Ras, these residues form a continuously connected core, whereas, in the GDP-bound structures, the connectivity of the network is broken and disorganized. Hatley et al. (28) argue that an essential component of the switch many thus be formation of this larger, energetically cooperative unit. Assuming similar behavior in Cdc42, it is possible that only active WASP, which interacts strongly with switch 1 and switch 2 (11, 16), can make contacts that propagate energetically through this large cooperative network, affording the GTP state significant differential binding to the two conformations of WASP. Because the network is predicted to be disorganized in the GDP state, such larger cooperativity may not be accessed, and this state may discriminate more poorly between the two conformations of WASP. Certainly further physical studies are needed to explore this issue.

Activation of WASP by Cdc42 plays an important role in activation of T cells during engagement by antigen-presenting cells (29, 30). During this process, WASP and Cdc42–GTP are independently concentrated at the T cell–antigen presenting cell interface (31). Locally high concentrations of the two molecules result in spatially restricted WASP activation. We have postulated that the independence of WASP and Cdc42–GTP recruitment plays an important role in mediating the choice of this particular effector (among many known effectors) by the GTPase during T cell activation (31). Thus, Cdc42–GTP selects the effectors that are locally concentrated at the site of its own activation by guanine nucleotide exchange factor, resulting in a specific signaling event. A nucleotide switch based solely on a change in affinity could be overridden in such a scenario. That is, a high local concentration of WASP could overcome a weak GTPase affinity, enabling its inappropriate activation by Cdc42–GDP, degrading the fidelity of Cdc42 signaling. The properties of WASP and Cdc42 that we have discovered here would mitigate this effect by ensuring that the Cdc42–GDP complex with WASP has a weak activity toward the Arp2/3 complex. Thus, a nucleotide switch based on both a change in affinity and a change in the efficiency of coupling between the binding and allosteric equilibria of effectors ensures both high specificity and high fidelity in GTPase (including Cdc42) signaling pathways. Given the preponderance of autoinhibited GTPase effectors, and regulatory balances between partial and full agonists throughout biology, we predict that such behaviors will be widely observed in GTPase signaling pathways.

Supplementary Material