The NF2 Tumor Suppressor Merlin and the ERM Proteins Interact with N-WASP and Regulate Its Actin Polymerization Function*

The function of the NF2 tumor suppressor merlin has remained elusive despite increasing evidence for its role in actin cytoskeleton reorganization. The closely related ERM proteins (ezrin, radixin, and moesin) act as linkers between the cell membrane and cytoskeleton, and have also been implicated as active actin reorganizers. We report here that merlin and the ERMs can interact with and regulate N-WASP, a critical regulator of actin dynamics. Merlin and moesin were found to inhibit N-WASP-mediated actin assembly in vitro, a function that appears independent of their ability to bind actin. Furthermore, exogenous expression of a constitutively active ERM inhibits N-WASP-dependent Shigella tail formation, suggesting that the ERMs may function as inhibitors of N-WASP function in vivo. This novel function of merlin and the ERMs illustrates a mechanism by which these proteins directly exert their effects on actin reorganization and also provides new insight into N-WASP regulation.

The function of the NF2 tumor suppressor merlin has remained elusive despite increasing evidence for its role in actin cytoskeleton reorganization. The closely related ERM proteins (ezrin, radixin, and moesin) act as linkers between the cell membrane and cytoskeleton, and have also been implicated as active actin reorganizers. We report here that merlin and the ERMs can interact with and regulate N-WASP, a critical regulator of actin dynamics. Merlin and moesin were found to inhibit N-WASP-mediated actin assembly in vitro, a function that appears independent of their ability to bind actin. Furthermore, exogenous expression of a constitutively active ERM inhibits N-WASP-dependent Shigella tail formation, suggesting that the ERMs may function as inhibitors of N-WASP function in vivo. This novel function of merlin and the ERMs illustrates a mechanism by which these proteins directly exert their effects on actin reorganization and also provides new insight into N-WASP regulation.
Ezrin, radixin, and moesin, known as ERM 1 proteins, act as membrane-cytoskeletal linkers that play important roles in cell morphology, polarity, and signal transduction (1). ERMs bind membrane proteins through their N-terminal FERM (band 4.1, ERM) domain and actin filaments through their C terminus, thus linking the cell membrane to the underlying cytoskeleton. Merlin, the protein product of the Neurofibromatosis type 2 (NF2) tumor suppressor gene, is closely related to the ERMs (2,3). The FERM domain of merlin and ERMs is the region of highest homology and can interact with various membrane proteins. The C terminus of ERMs contains a high affinity actin-binding site that is not conserved in merlin. However, merlin also directly binds F-actin through its FERM domain and stabilizes actin filaments in vitro (4,5). ERMs and merlin are regulated by an intramolecular head-to-tail association between the FERM domain and the C terminus that prevents binding of membrane partners as well as F-actin (6). The autoinhibited conformation of the ERMs can be disrupted by polyphosphatidylinositol 4,5-bisphosphate (PIP 2 ) binding or phosphorylation at a conserved C-terminal threonine.
The role of ERMs and merlin in actin cytoskeleton reorganization has been well documented. In cultured cells, they localize to actin-rich regions such as microvilli, lamellipodia, filopodia, and neuronal growth cones. Antisense suppression of ERMs can disrupt microvilli and slow growth cone advancement (7,8). The absence of merlin, either in schwannoma cells from NF2 patients, or mouse embryo fibroblasts from NF2deficient mice, results in increased membrane ruffling and cell motility (9,10). Thus merlin may negatively regulate actin remodeling in vivo. These cellular phenotypes caused by merlin or ERM loss can be partially explained by their structural role as membrane-cytoskeleton linkers. In addition, merlin and ERMs are regulated by Rho family GTPases via phosphorylation and also alter GTPase function through a negative feedback mechanism, thereby regulating the actin cytoskeleton indirectly (9,11). However, it remains unknown whether these proteins can directly alter actin nucleation and assembly downstream of Rho, Rac, or Cdc42.
We examined whether merlin and ERMs could directly regulate the actin polymerization pathway downstream of these GTPases. The WASP (Wiskott-Aldrich syndrome protein) family proteins, upon activation by Rac and Cdc42, are capable of activating actin assembly by the actin nucleating Arp2/3 complex (12). Cdc42 binds and activates N-WASP, a ubiquitously expressed member of the WASP family, which in turn stimulates Arp2/3 complex activity. WIP (WASP interacting protein) interacts with the WH1 domain of N-WASP and inhibits its function in vitro and in vivo (13,14). We report here that merlin and ERMs are able to directly bind N-WASP. Furthermore, merlin and ERMs can regulate N-WASP function in vitro and in vivo, demonstrating a novel, direct mechanism for their effect on actin reorganization and a new mode of N-WASP regulation.
* This work was supported by National Institutes of Health Grants NS24279 (to N. M., M. F. J., J. F. G., and V. R.) and AI052354 and HL59561 (to S. B. S.) and Department of Defense Grant DAMD17-02-0647. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

EXPERIMENTAL PROCEDURES
Materials and Antibodies-HA-radixin full-length (FL) and N-term and C-term plasmids were a gift from F. Solomon (MIT, Cambridge, MA). Mutants FL HA-radixin (T564E and T564A) were provided by S. Tsukita (Kyoto University, Kyoto, Japan). Myc-tagged N-WASP constructs were described previously (15). Mini-N-WASP was provided by W. Lim (University of California, San Francisco, CA). The anti-WAVE, anti-N-WASP, and anti-WIP antibodies have been described (13,16,17). Anti-NHE-RF IC270 and anti-merlin antibodies 1C4 and C26H were described previously (18,19). Anti-Nck and anti-Myc 9E10 antibodies were purchased from BD Transduction Laboratories and Developmental Studies Hybridoma Bank, respectively. Phospho-ERM antibody 3141 and anti-HA antibody were purchased from Cell Signaling Technologies and Covance, respectively.
Pull-down Assays-Merlin and moesin GST fusion proteins were expressed and purified as described (18). 293T lysates were incubated with 200pmol of immobilized merlin or moesin GST fusion protein.
Lysate lane shows 10% of input. For pull-down assays using in vitro transcribed and translated N-WASP truncations, the SP6 Quick Coupled kit (Promega) was used with 2 g of each N-WASP plasmid. The lysate lane shows 5% of input. Direct binding in solution was carried out with recombinant N-WASP (0.32 g) and GST-purified moesin constructs (10 g) as described (15). The input N-WASP lane contains 40% of total input.
Immunoprecipitation-IP was performed using 293T lysate and 5 g of merlin antibody 1C4 or mouse IgG followed by Western blotting with anti-N-WASP antibody. Merlin IP was detected by anti-merlin C26H antibody. The reverse IP used 5 g affinity purified N-WASP rabbit antibody or normal rabbit serum as a control and was detected by merlin antibody. For phospho-ERM interaction, HeLa cells were treated for 5 min with calyculinA (50 nM) before harvesting in lysis buffer (as above with 100 nM calyculinA). The N-WASP IP was performed with 8 g of affinity-purified N-WASP or rabbit IgG, followed by Western blotting with the phospho-ERM and N-WASP antibodies.
Actin Polymerization Assays-N-WASP and Cdc42 were purified as described (16). Actin from rabbit skeletal muscle and purified Arp2/3 complex were a generous gift of F. Nakamura and J. H. Hartwig (Brigham and Women's Hospital, Boston, MA). G-actin was freshly thawed at 4°C and centrifuged at 400,000 ϫ g for 1 h to remove F-actin. Polymerization assays were carried out as described (16). Briefly, actin was mixed with pyrene-labeled actin at a ratio of 3:2 and was used at a final concentration of 1 M in a total volume of 80 l of Xenopus buffer. Merlin and moesin proteins were preincubated with other proteins such as N-WASP, Arp2/3 complex, and Cdc42-GTP␥S before the reaction was initiated by addition of actin mixture. The increase in fluorescence was measured by a spectrofluorimeter.
Shigella Infections and Immunofluorescence-SV40 transformed mouse embryo fibroblast-like cells from N-WASP-deficient animals (28) were transduced with pLNCX-GFP-N-WASP retrovirus and sorted to enrich rescued cells. 18 h after radixin transfection, cells were infected with Shigella flexneri strain 2457T as described by Shibata et al. (33). Immunofluorescence analysis was performed after lysis with 0.1% Nonidet P-40, using anti-HA antibody to detect transfected radixin, followed by incubation with Cy3-conjugated secondary antibody (Jackson Laboratories) and Alexa-fluor 488-conjugated phalloidin (Molecular Probes) to detect actin. Coverslips were mounted in ProFade Gold medium containing DAPI (Molecular Probes) to visualize bacteria. Infected cells (with at least five bacteria) were examined and scored for transfection and at least one actin comet tail. The percentage of transfected cells with actin tails was normalized to untransfected cells with tails on the same coverslip. Values were means from three independent experiments in which a total of Ͼ150 transfected cells were examined for each construct (except for the constitutively active radixin T564E where 346 transfected cells were counted). The samples were analyzed by Olympus Provis AX70 microscope, 100ϫ oil immersion lens, Magna-Fire S99806 camera, using MagnaFIRE acquisition software and pseudo-colored using Adobe Photoshop.

RESULTS AND DISCUSSION
To investigate the potential interaction between merlin/ ERMs and proteins involved in Rac/Cdc42 signaling to actin, we performed pull-down assays using GST fusions encoding various merlin and moesin domains. Both merlin and moesin precipitated endogenous N-WASP from fibroblast and 293T cell lysates ( Fig. 1A and supplemental Fig. S1). WAVE, another WASP family member, and the Arp2/3 complex were not de-tected in these precipitates (not depicted). The full-length (FL) and N-terminal FERM domains (N-term) of both merlin and moesin precipitated N-WASP, whereas the C-terminal domains (C-term) and GST alone did not. The reduced binding of FL moesin is presumably a result of its closed conformation. The intramolecular association of merlin, unlike moesin, is more dynamic, which allows greater accessibility (1). We compared binding to NHE-RF, an established FERM interactor, in the same assay. As expected, both merlin and moesin FERM domains precipitated NHE-RF, whereas C-term and GST alone did not (Fig. 1A). Interestingly, NHE-RF and N-WASP bind the FERM domain in distinct ways. NHE-RF displayed higher affinity for moesin FERM, as has been reported previously (18). On the contrary, N-WASP bound both FERM domains similarly, suggesting that it probably binds different residues than NHE-RF, which are highly conserved between the two FERM domains. We also tested whether other proteins that associate with N-WASP precipitated in the same assay. Nck, an adaptor protein that regulates N-WASP function through its prolinerich domain (20,21), did not precipitate along with N-WASP (Fig. 1A, lower panel). WIP exists in a complex with N-WASP but is expressed mostly in hematopoietic cells, and we could not detect a significant amount in the cells used in this assay (not depicted).
N-WASP integrates signals from various sources by interacting with distinct partners through each of its domains. To investigate which of these domains mediates the FERM domain interaction, we used in vitro transcribed and translated (IVTT) N-WASP peptides in a GST pull-down assay. Three Myc-tagged N-WASP IVTT products were tested for binding to the merlin FERM domain GST fusion protein (Fig. 1B). WH1 and WGP (WH1, G protein-binding region, proline-rich domain) polypeptides were capable of binding, whereas the VCA (verprolin homology, cofilin homology, acidic) was not, indicating that the WH1 domain is sufficient. This is a region conserved between N-WASP and WASP, but not WAVE/Scar proteins, which supports our finding that WAVE does not interact with merlin or moesin.
To determine whether the FERM-N-WASP interaction is direct, we performed a solution binding assay using purified GST-moesin and purified, untagged N-WASP, followed by precipitation of GST fusion proteins. Both moesin FERM domain and FL moesin showed direct binding as compared with C-term alone (Fig. 1C) and FERM moesin bound FL N-WASP in a dose-dependent manner (Fig. 1D).
To examine the interaction between merlin and N-WASP in vivo, we performed co-immunoprecipitation using 293T cell lysates. Anti-merlin antibody specifically precipitated endogenous N-WASP along with endogenous merlin and conversely, the N-WASP antibody precipitated both merlin and N-WASP (Fig. 1E). Although moesin and merlin were both capable of interacting with N-WASP in pull-down assays, we were unable to detect any ERMs in N-WASP immunoprecipitates, suggesting that their tightly closed conformation in vivo renders them inaccessible. Phosphorylation of a conserved C-terminal threonine has been shown to stabilize the open conformation of ERMs (22,23). To enrich phosphorylated ERMs, we treated HeLa cells with a Ser/Thr phosphatase inhibitor (calyculinA) and used an ERM antibody that recognizes the conserved C-terminal phospho-Thr. Using the N-WASP antibody for immunoprecipitation, we detected a significant association between the phosphorylated, active ERMs and N-WASP in vivo (Fig. 1F). We also performed co-immunoprecipitation experiments using FLAG-merlin FERM domain and Myc-N-WASP WH1 domain, which confirmed that these regions were sufficient to mediate the in-teraction in vivo. 2 The ERMs are more similar to each other within the FERM domain (87% identity) than to merlin (63%). Given this high degree of similarity, and the fact that two ERMs and merlin can co-immunoprecipitate with N-WASP, it is likely that all three ERMs interact similarly with N-WASP. Therefore, we assume that any functional consequence of FERM domain-N-WASP binding will apply to ezrin, radixin, and moesin.
To assess the relevance of the interaction between merlin/ ERMs and N-WASP, we examined the effect of these proteins on N-WASP induced Arp2/3 complex-mediated actin polymerization in vitro. The kinetics of actin polymerization can be monitored using pyrene-labeled actin, which displays an increase in fluorescence intensity when incorporated into fila-ments. Full-length N-WASP, being auto-inhibited in vitro, requires binding of activators such as Cdc42 or PIP 2 (16). We examined the effect of merlin/ERMs on full-length N-WASP activated by Cdc42 and found that merlin and moesin proteins that contained the FERM domain inhibited the rate of actin assembly. The merlin FERM domain caused a dose-dependent decrease in the rate of actin polymerization when added to actin, Arp2/3 complex, N-WASP, and Cdc42 ( Fig.  2A). FERM domain and FL moesin also functioned similarly in this assay (Fig. 2B) as did FL merlin (not depicted). Despite FERM domain binding N-WASP more robustly than FL proteins in the pull-down assays, their effect appears similar in kinetics, probably due to the higher sensitivity of the actin polymerization assay. Thus, all proteins that demonstrated binding to N-WASP in the pull-down assay were able to inhibit N-WASP-mediated actin polymerization. The C ter-2 N. Manchanda and V. Ramesh, unpublished data.

FIG. 1. Interaction of merlin and ERMs with N-WASP.
A, Western blot detects N-WASP in 293T lysate and in the material precipitated by various GST fusion proteins. GST-N-merlin and N-moesin (1-332 aa) bound N-WASP, as did FL GST merlin (1-595 aa) and moesin (1-577 aa), whereas GST alone, C-merlin (340 -595 aa), and C-moesin (305-577 aa) did not. The middle panel shows the same blot re-probed with anti-NHE-RF antibody IC270, which detected differentially phosphorylated NHE-RF. The lower panel displays the same blot probed for Nck. Supplemental Fig. S1 shows loading of GST fusion proteins. B, three Myctagged IVTT N-WASP peptides were used as input in pull-down assays with GSTmerlin FERM domain (1-332 aa). WH1 and WGP domains both precipitated, whereas VCA did not, as detected by anti-Myc (9E10) antibody. WH1 protein showed a 3.2-fold increase in precipitation with FERM domain as compared with GST alone, and WGP showed a 3.4fold increase. These values are the mean of three independent experiments, which were quantified by densitometry. C, solution binding of recombinant N-WASP with purified GST-moesin was detected by pull-down. The top panel shows Western blot that detects N-WASP binding to moesin N-term and FL proteins but not C-term. Typically 10% of total amount was bound to moesin N-term. The lower panel shows Ponceau-S staining, which detects equal loading of purified moesin N and C proteins (FL is slightly higher). D, dose-dependent binding of N-WASP to moesin N-term. A Western blot representative of three independent experiments which were performed as in C above. E, co-immunoprecipitation of endogenous merlin and N-WASP. F, Western blot with anti-phospho-ERM antibody detected endogenous, phosphorylated ERMs in N-WASP immunoprecipitates. mini of moesin (Fig. 2C) and merlin (not depicted), which cannot bind N-WASP, had no effect, confirming the specificity of this result. The C terminus of the ERMs contains the F-actin-binding site. The fact that we observed no effect of this region in our assay indicates that F-actin binding does not contribute to the inhibition we observed, and indeed it is the FERM domain binding to N-WASP that mediates this function.
To determine whether interaction with the WH1 domain mediated this inhibition, we utilized a mini-N-WASP protein. This protein lacks the WH1 and proline-rich domains but is still capable of auto-inhibitory binding of the GBD to the VCA, which is relieved by Cdc42, enabling Arp2/3 complex activation (24). Unlike full-length N-WASP which was inhibited strongly (Fig. 2, A and B), mini-N-WASP was unaffected by the same concentration of FERM merlin and FL moesin (Fig. 2D). We also tested the ability of merlin and moesin to inhibit VCA domain function and found none (not depicted), confirming that N-WASP polypeptides that are able to activate Arp2/3 complex but lack the WH1 domain are not inhibited. This also demonstrates that merlin and moesin have no effect on other components including Cdc42, the Arp2/3 complex, and actin itself in this assay. Interestingly, WIP and its related proteins (CR16 and WICH) also bind the WH1 domain and WIP inhibits N-WASP mediated actin polymerization in a similar assay (13,25,26). In summary, merlin and moesin inhibit N-WASP function in actin assembly in vitro through binding of their FERM domains to the WH1 domain of N-WASP. These results provide the first example of proteins unrelated to WIP binding the WH1 domain and suppressing N-WASP activity.
The enteroinvasive pathogen S. flexneri uses host cell actin polymerization machinery to propel itself within and between cells. The Shigella surface protein IcsA recruits host cell N-WASP to activate the Arp2/3 complex and form an actin comet tail (27). This process is completely dependent on the presence of N-WASP as N-WASP-deficient cells permit Shigella entry but not subsequent tail formation and motility (28,29) providing a reliable assay to study the effect of ERMs on N-WASP function in vivo. Ezrin, a member of the ERM family, is required for Shigella entry into HeLa cells. Overexpressing its FERM domain, which exerts a dominant negative effect on ezrin function, causes a decrease in Shigella entry (30).
We evaluated the in vivo relevance of the ERM-N-WASP interaction by examining the effect of exogenous ERMs on actin tail formation. ERMs were chosen because merlin and moesin FERM domains behaved similarly in binding and inhibiting N-WASP in vitro. Moreover, the physical separation of N-WASP and actin-binding domains in ERMs, unlike merlin, ensured that any effect of direct actin binding on tail formation would be evident. Radixin was chosen as we had access to several expression constructs and because we observed no difference between the various FERM domains with respect to N-WASP binding and regulation (Figs. 1 and 2). We used HA-tagged FL, N-term, C-term, and active and inactive mutants of radixin. Constitutively active radixin T564E is a phosphomimetic mutant that renders the protein open, whereas the inactive T564A mutant cannot be phosphorylated and is closed. After transfection of radixin constructs into fibroblasts and infection with Shigella, cells were fixed and processed for immunofluorescence. Due to the marked inhibition of Shigella entry caused by FERM domain overexpression (30), which we also confirmed in our experiments, we were unable to evaluate whether this construct inhibited tail formation. All other radixin constructs did not significantly alter Shigella entry and were evaluated for their effect on tail formation. We found that constitutively active radixin T564E inhibited actin tail formation by 60% (Fig. 3A). Constitutively inactive T564A had no such inhibitory effect (Fig. 3B). FL, wild-type radixin, which is probably in the closed conformation in vivo, did not significantly inhibit tail formation (Fig. 3C). Our coimmunoprecipitation results indicated that only phospho-ERMs bind N-WASP (Fig. 1F), suggesting that the phosphomimetic ERM is the only transfected construct able to bind N-WASP and thereby inhibit tail formation. The free radixin C-term, which binds actin had no effect despite co-localizing with F-actin in the tails (supplemental Fig. S2), supporting our conclusion that the inhibition is probably mediated through N-WASP binding.
To confirm that the active mutant inhibited tail formation by inhibiting N-WASP, we co-transfected it with N-WASP and observed a large increase in tail formation compared with mutant T564E alone (Fig. 3C). Exogenous N-WASP alone has been shown to have no significant effect on Shigella tail formation (31). Thus the inhibition we observe is likely a consequence of ERM inhibition of N-WASP and not an indirect effect of overexpressing active radixin. Rho, Rac, and Cdc42 activity is dispensable for tail formation (32,33), supporting the conclusion that active ERM is not functioning by altering the activity of these GTPases. Together, these results suggest that constitutively active radixin inhibits N-WASP dependent Shigella tail formation through their interaction in vivo.
These results confirm the observations made using the actin assembly assays. The in vitro assays facilitated our ability to assess free FERM domain function because the dominant negative effect seen in vivo was eliminated. Interestingly, FL ERM was able to inhibit N-WASP in vitro (Fig. 2B), whereas FL wild-type ERM did not have a significant effect in vivo. This apparent contradiction can be explained by the fact that GST fusion ERMs can be less tightly closed than ERMs in vivo. Therefore, we observe GST-FL ERM binding and inhibiting N-WASP in vitro, whereas in vivo, there is no detectable effect on tail formation due to the inability of N-WASP to bind FL ERMs unless they are phosphorylated. We were able to address this problem by using the phosphomimetic, active FL ERM, which was found to inhibit tail formation. Use of this T564E mutant also enabled us to assess the effect of the FERM domain on tail formation because, unlike the free FERM domain, this construct does not inhibit bacterial entry.
N-WASP recruitment to the surface of Shigella is mediated through the interaction of its WH1 and GBD domains with IcsA (29). Since the FERM domains of merlin/ERMs bind the WH1 domain of N-WASP, we examined whether active radixin disrupted the recruitment of N-WASP, thereby inhibiting tail formation. However, N-WASP was still able to localize correctly to the pole of bacteria in cells transfected with active mutant. 2 This does not, therefore, appear to be the mechanism of ERM inhibition. It is possible that FERM domain binding causes a conformational change within N-WASP and/or interferes with binding of specific activators.
N-WASP activity was previously thought to be regulated by the auto-inhibitory interactions between the GBD and VCA domains. It has recently emerged, however, that WIP-WH1 interaction negatively regulates N-WASP function in vivo, indicating that N-WASP is subject to additional levels of regulation through the WH1 domain (13,14). To our knowledge, this report represents the first example of a protein unrelated to WIP functioning as an N-WASP inhibitor. Our finding that merlin/ERM inhibition of N-WASP function may also be mediated through this domain raises the possibility that the WH1 domain may serve as a general negative regulatory module for N-WASP. Most mutations in Wiskott-Aldrich syndrome patients map to the WH1 domain of WASP, suggesting that loss of the inhibition imposed by this domain may be the mechanism of disease initiation.
Previously, the effect of merlin and ERMs on the actin cytoskeleton was attributed to either binding to actin or to an indirect effect on Rho GTPase signaling. Here, we report that direct binding between the FERM domain of merlin/ERMs and N-WASP can regulate actin assembly independent of their actin binding and GTPase regulation functions. Thus our finding that merlin and ERM proteins can directly regulate N-WASP provides a novel mechanism by which these proteins effect actin reorganization and also represents an alternate means of N-WASP regulation.