Original Paper | Published:

Magicin, a novel cytoskeletal protein associates with the NF2 tumor suppressor merlin and Grb2


Neurofibromatosis 2 (NF2) is a dominantly inherited disorder characterized by bilateral vestibular schwannomas and meningiomas. Merlin, the neurofibromatosis 2 tumor suppressor protein, is related to the ERM (ezrin, radixin, moesin) proteins and, like its family members, is thought to play a role in plasma membrane–cytoskeletal interactions. We report a novel protein as a merlin-specific binding partner that we have named magicin (merlin and Grb2 interacting cytoskeletal protein) and show that the two proteins interact in vitro and in vivo as well as colocalize beneath the plasma membrane. Magicin is a 24 kDa protein that is expressed in many cell lines and tissues. Magicin, similar to merlin, associates with the actin cytoskeleton as determined by cofractionation, immunofluorescence and electron microscopy. Analysis of the magicin sequence reveals binding motifs for the adaptor protein Grb2. Employing affinity binding, blot overlay and co-immunoprecipitation assays, we demonstrate an interaction between Grb2 and magicin. In addition, merlin is capable of forming a ternary complex with magicin and Grb2. These results support a role for merlin in receptor-mediated signaling at the cell surface, and may have implications in the regulation of cytoskeletal reorganization.


Neurofibromatosis 2 (NF2) is an inherited autosomal dominant disorder with predisposition to vestibular schwannomas and meningiomas. Tumors are caused by inactivating mutations of the NF2 gene that result in loss of function of the encoded protein merlin, also known as schwannomin (Rouleau et al., 1993; Trofatter et al., 1993). Merlin exhibits similarity to the ERM (ezrin, radixin and moesin) proteins, which are members of the protein 4.1 superfamily. ERM proteins function as linker molecules facilitating regulated attachment of membrane proteins to the actin cytoskeleton, which is essential for several fundamental processes including cell shape, cell adhesion, motility and integration of cell surface receptors with many signaling pathways (Bretscher et al., 2002). As predicted from the structural similarity with the ERM proteins, merlin shares some functions with its family members; however, it has distinct functions as a tumor suppressor in both humans and mice (Gusella et al., 1999; Bretscher et al., 2002).

Merlin exists as two major isoforms due to alternative splicing. Merlin isoform 1 is regulated in a manner similar to the ERM proteins; the carboxy terminus of isoform 1 binds to the amino terminus of merlin and the ERMs. However, the intramolecular interaction of merlin is quite dynamic and of lower affinity than that of ezrin (Nguyen et al., 2001). In contrast, merlin isoform 2 lacks the ability to form an intramolecular head-to-tail association and exists in a constitutively open conformation (Sherman et al., 1997; Gonzalez-Agosti et al., 1999; Gronholm et al., 1999).

Similar to ERM proteins, merlin localizes to the leading or ruffling edges in many cell types, and colocalizes with F-actin in these motile regions (Gonzalez-Agosti et al., 1996; Sainio et al., 1997; Deguen et al., 1998). Merlin interacts with F-actin through actin binding sites within the FERM domain (Xu and Gutmann, 1998; Brault et al., 2001; James et al., 2001). In addition to actin, several other merlin interacting proteins have been identified, which include β-fodrin/βII-spectrin (Scoles et al., 1998; Neill and Crompton, 2001), SCHIP-1 (Goutebroze et al., 2000), NHE-RF (Murthy et al., 1998), β1-integrin (Obremski et al., 1998), CD44 (Sainio et al., 1997; Morrison et al., 2001), HRS (Scoles et al., 2000), RhoGDI (Maeda et al., 1999), syntenin (Jannatipour et al., 2001), paxillin (Fernandez-Valle et al., 2002) and RIβ subunit of the cAMP-protein kinase A (Gronholm et al., 2003). Among these proteins, NHE-RF, CD44 and RhoGDI have been independently identified as ERM binding partners (Tsukita et al., 1994; Reczek et al., 1997; Takahashi et al., 1997), while HRS and syntenin appear to be specific for merlin. It is interesting to note that many of these proteins are plasma membrane-associated proteins or adaptors that couple membrane proteins to cytoskeletal components. Expression of activated forms of Rac and Cdc42 leads to merlin phosphorylation and merlin is also capable of negatively regulating Rac (Shaw et al., 2001). Subsequent studies have shown that S518 in merlin is phosphorylated by the Rac/Cdc42 effector p21-activated kinase (Pak) (Kissil et al., 2002; Xiao et al., 2002). Merlin binds to Pak and is thought to inhibit the recruitment of Pak to focal adhesions (Kissil et al., 2003). Furthermore, merlin-deficient schwannoma cells exhibit alterations in actin cytoskeleton organization that are reversed by reintroduction of wild-type but not mutant merlin (Pelton et al., 1998; Bashour et al., 2002). Thus, studies performed on merlin collectively suggest that merlin might indeed play a role in modulating many aspects of receptor–cytoskeleton linkage as well as in signaling to the cytoskeleton that is important for cell growth and adhesion.

In our efforts to find merlin-specific interactors, we have identified magicin, a novel cytoskeletal-associated protein, which binds specifically to merlin but not to the ERM proteins. Magicin also associates with Grb2, which has been suggested to act not only as an adaptor but also as an effector of actin-dependent processes in response to cell signaling (Carlier et al., 2000). The direct interaction of magicin with both merlin and Grb2 adds further support for a role of merlin in receptor-mediated signaling and/or a role in the regulation of cytoskeletal reorganization.


Identification of a merlin binding protein by yeast two-hybrid screen

To identify merlin binding proteins, we screened a human fetal frontal cortex cDNA library with merlin baits using a LexA-based yeast two-hybrid system. A cDNA referred to as MerIntA was isolated, which specifically associates with full-length and the carboxy terminus of merlin but not with unrelated baits such as APP, Bicoid, CDC2, IQ GAP or neurofibromin (NF1) (Figure 1a and b).

Figure 1

Isolation of merlin binding proteins using the yeast two-hybrid system. (a) Schematic diagram of the merlin baits used. (b) Table indicating the specificity of interactions. Controls tested under the same conditions do not show interaction with any of the candidates. (c) The protein sequence of human magicin was aligned with the predicted sequences of mouse, fish and fly magicin using multalin. The complete sequences of these proteins are available at the following GenBank/EMBL/DDBJ accession numbers: Homo sapiens (AAK32724), Mus musculus (BAB23039), Danio rerio (AI942663, AI884138), Drosophila melanogaster (AAF56470)

Sequence analysis revealed MerIntA to be a novel gene, although it did not have a candidate ATG in the reading frame, indicating that the 5′ end of this clone was missing. Therefore, rapid amplification of cDNA ends (5′ RACE) was performed from placenta and peripheral blood mononuclear cell (PBMC) cDNAs to obtain the 5′ end. The full-length product revealed a protein of 178 amino acids (aa) with a predicted molecular mass of 20 kDa that we named magicin. During the course of this study, three mRNA sequences, which match those of magicin, were deposited in the GenBank database with accession numbers AF317679, AF358829 and AF321617. These entries are identical among themselves and to magicin, differing only in the length of the untranslated regions. Information submitted to GenBank for AF317679 suggests that magicin is a tumor-related protein, while AF358829 indicates that magicin may function as a tumor angiogenesis marker (Liu et al., 2002).

Further database analyses identified anonymous, predicted orthologues in mouse (BAB23039), zebrafish (AI942663 and AI884138) and Drosophila (AAF56470), while no orthologues were immediately evident in Caenorhabditis elegans or yeast. The zebrafish and Drosophila orthologues are 76 and 43% identical to the human protein (Figure 1c). Thus, magicin is a novel highly conserved protein of unknown function. Magicin sequence was examined with Scansite (http://scansite.mit.edu) to predict protein binding sites. A binding motif for the Src homology 2 (SH2) domain of the growth factor receptor bound 2 (Grb2) protein was revealed in a high stringency scan (Y64 of magicin). The consensus sequence of this binding motif is pYV/INX (Schlessinger, 1994) and, in addition, several PxxP motifs known to bind SH3 domains were also observed in a medium stringency scan.

Magicin binds to merlin but not to moesin

To confirm and characterize the association between merlin and magicin, we tested the ability of hemagglutinin (HA)-tagged magicin (MerIntA aa 13–178) expressed in COS-7 cells to bind with different segments of merlin, expressed as bacterial glutathione S-transferase (GST) fusion proteins in affinity precipitation assays. This version of tagged magicin migrated slower than expected due to an additional 90 aa derived from the vector sequence (pJG4-5) encoded between the HA tag and magicin as a fusion protein. The bound protein, detected with an anti-HA antibody, revealed that magicin associated specifically to the carboxy-terminal merlin fusion proteins, but did not bind to either the amino terminus of merlin or GST alone (Figure 2a). The results further demonstrated that the carboxy terminus of merlin isoform 2 captured more magicin than the truncated carboxy-terminal protein common to both isoforms (aa 308–579). Moreover, less magicin was captured by the carboxy-terminal GST fusion protein of isoform 1 compared to the other two carboxy-terminal segments. Samples were assayed by Ponceau-S staining to ensure equal loading of GST fusion proteins.

Figure 2

Binding of magicin to merlin but not to moesin. (a) NP40 lysates from COS-7 cells overexpressing HA-merintA were incubated with various GST-merlin fusion proteins immobilized on glutathione Sepharose 4B beads. Bound proteins were separated on 10% SDS–PAGE and immunoblotted with an anti-HA antibody. MerintA migrates slower than expected due to additional 90 aa derived from the vector sequence (pJG4-5) encoded between the HA tag and the coding sequence. MerintA exhibits a stronger binding to the C-terminus of merlin isoform 2 (aa 340–590) when compared with either isoform 1 (aa 340–595) or a domain common to both isoforms (aa 308–579). No binding was observed with the N-terminus of merlin (aa 1–332). (b) Various GST-moesin fusion proteins were tested for binding to merintA employing the same assay as in (a). None of the moesin constructs bind to merintA. Full-length merlin isoform 2 was used as a positive control. The lysate lane represents about 1/20 of the amount used in the precipitation assays

To test whether magicin binds exclusively to merlin or also to other ERM family members, moesin was chosen as a representative, and different segments of GST-moesin fusion proteins were tested. Magicin did not bind to moesin, but bound efficiently to full-length merlin isoform 2 (Figure 2b). The affinity binding assays thus illustrate that magicin associates with the carboxy-terminal but not with the amino-terminal domain of merlin, confirming the results of the yeast two-hybrid analysis. Moreover, magicin binds specifically to merlin but not to moesin. Subsequent experiments were carried out with either endogenous magicin or HA-tagged full-length magicin (A4) without the additional 90 aa derived from the yeast two-hybrid library plasmid.

Magicin is widely expressed in cells and tissues

To investigate the expression profile of magicin, a fetal human multiple-tissue RNA blot was probed with the 32P-labeled cDNA encoding magicin. Multiple transcripts with sizes of 1, 1.3 and 3.2 kb were detected in fetal tissues (Figure 3a). The sizes of these transcripts match those deposited in GenBank (see above), indicating that they all translate into the same protein and differ only in the untranslated regions. Similar results were noted in human adult tissues examined (data not shown). Thus, magicin is widely expressed and has different transcript sizes.

Figure 3

Expression profile of magicin. (a) Analysis of magicin RNA. A multiple tissue poly(A)+ RNA blot from human fetal tissues (Clontech) was probed with 32P-labeled cDNA fragment encoding magicin as recommended by the manufacturer. (b) Characterization of anti-magicin antibodies. CAD cell lysates were subjected to immunoprecipitation with a rabbit anti-magicin polyclonal antibody (Tim3). Precipitates were immunoprobed with an anti-magicin monoclonal antibody 7E1. Control immunoprecipitations include omission of cell lysate or adding a nonspecific antibody. (c) Analysis of magicin protein. Western blot analyses of various cell lines reveal that magicin is expressed as a 24 kDa protein. In some cell lines, an additional 50 kDa protein is observed. Equal amounts of protein lysates (100 μg) were separated by 12% SDS–PAGE and immunoblotted with 7E1 antibody

At the protein level, the monoclonal antibody 7E1, raised against a GST-magicin, recognized a predominant band at 24 kDa (Figure 3b and c) in immunoblots of CAD cells, which was slightly larger than predicted (20 kDa) from the open reading frame. The 24 kDa polypeptide was not detected upon preabsorbing the antibody with purified magicin (data not shown). Tim3, a polyclonal antibody, was employed to immunoprecipitate magicin, and the monoclonal antibody 7E1 recognized the same 24 kDa protein in the Western blot analysis (Figure 3b), thus confirming specificity of these antibodies. The expression profile of endogenous magicin was examined by Western blot analysis in various cell lines. Analyses of normalized protein lysates indicated that magicin was expressed in many of the tested cell lines including CAD, Met5A, Gli238, F293, 293T, COS-7 and HeLa (Figure 3c). An additional 50 kDa band was detected in many cell lines with this antibody. As the CAD cell line revealed only the 24 kDa protein, this cell line was chosen for subsequent staining experiments.

Merlin binds to magicin in vivo

To confirm the potential association between merlin and magicin in vivo, we performed co-immunoprecipitation experiments in T-47D cells. Immunoprecipitating merlin with a rabbit polyclonal antibody followed by immunoblot analysis showed that endogenous magicin co-immunoprecipitated with merlin (Figure 4a, left). Reciprocal immunoprecipitations performed with an anti-magicin antibody followed by immunoblotting with anti-merlin antibody further confirmed the interaction of these two proteins (Figure 4a, right). Neither merlin nor magicin was detected in the negative control, which included either normal rabbit serum or preimmune serum. Thus, magicin and merlin interact both in vitro and in vivo. A recent study showed that merlin mutant S518D, which mimics constitutive phosphorylation, abolished the ability of merlin to suppress cell growth in cultured cells, and S518A, a mutant that lacks phosphorylation, can effectively suppress growth (Surace et al., 2004). We therefore tested the ability of magicin to bind the S518A and S518D mutants of merlin and observed an enhanced binding of magicin with the S518A mutant of both merlin isoforms compared with S518D (Figure 4b).

Figure 4

Magicin binds to merlin in vivo. (a) Endogenous merlin was immunoprecipitated from confluent T-47D cell lysates using rabbit polyclonal antibody (C26), magicin antibody Tim3, or normal rabbit serum (NRS) and the preimmune serum as controls. The immunoprecipitates were separated on 10% SDS–PAGE and immunoblotted for merlin and magicin employing a chicken polyclonal antibody (C26) and monoclonal antibody 7E1, respectively. The lysate lane represents about 1/20 of the amount used in the immunoprecipitation assays. (b) Magicin exhibits enhanced binding with S518A than S518D for both merlin isoforms 1 and 2. 293T cells were cotransfected with HA-tagged magicin (Mag.) and Flag-tagged S518A or S518D mutant of merlin isoform 1 or 2 (Iso1 or Iso2). The cell lysates were subjected to immunoprecipitation with the Flag-tagged M2 antibody or normal mouse IgG as a control. Immunoprecipitates were fractionated by 10.5–14% SDS–PAGE and immunoblotted for Flag-merlin and HA-magicin. (c) Magicin associates with the insoluble/cytoskeletal cell fractions. CAD cells were subjected to cytoskeletal fractionation using Triton X-100 as the detergent and total amounts of soluble (S) and insoluble (I) protein were separated on 12% SDS–PAGE and immunoblotted for magicin and merlin. Endogenous merlin and magicin are predominantly present in the insoluble fraction

Magicin associates with the actin cytoskeleton

Merlin can bind actin either directly through low-affinity binding sites reported in the FERM domain (residues 178–367) (Xu and Gutmann, 1998; Brault et al., 2001; James et al., 2001), or indirectly via the actin binding protein β-fodrin/βII-spectrin (Scoles et al., 1998; Neill and Crompton, 2001). Merlin is detergent (Triton X-100)-insoluble, which is indicative of its capability to interact with the cytoskeleton (Deguen et al., 1998; Stokowski and Cox, 2000). We therefore, tested magicin for its ability to associate with the cytoskeleton using similar fractionation experiments in CAD cells. Endogenous magicin, similar to merlin, was primarily associated with the detergent-insoluble cytoskeletal fraction (Figure 4c). This observation is consistent with enhanced binding of magicin with S518A merlin since S518D merlin has been reported to be less tightly associated with the cytoskeleton (Shaw et al., 2001).

Magicin colocalizes with merlin and actin

Merlin localizes to the submembranous region of cells, particularly to the motile regions, such as leading or ruffling edges where it colocalizes with actin (Gonzalez-Agosti et al., 1996). We transiently transfected CAD cells with HA-tagged magicin (A4 aa 1–178) and green fluorescent protein (GFP)-merlin isoform 1 or 2 in order to examine the localization of these two proteins. In these cells, magicin specifically localized to submembranous regions and neurites where it colocalized precisely with merlin isoform 2 (Figure 5a–c). Similar results were obtained in cells examined for colocalization of magicin and GFP-merlin isoform 1 (data not shown). This localization pattern was confirmed in CAD cells stained with 7E1 to visualize endogenous magicin. The endogenous protein was also observed to localize to submembranous regions and neurites (Figure 5d and e), whereas no specific staining was obtained when the primary antibody was omitted (Figure 5f). Double labeling of CAD cells for magicin and actin showed strong colocalization in neurites and beneath the plasma membrane (Figure 5g–i). Thus magicin colocalizes with merlin and actin, consistent with its function as a merlin binding protein in vivo.

Figure 5

Colocalization of magicin with merlin and actin. (a, b) GFP-merlin isoform 2 was directly visualized in CAD cells that were also stained for overexpressed magicin with an HA antibody (12CA5). (c) Merged images. Merlin colocalizes with magicin in neurites and beneath the plasma membrane. (d, e) CAD cells were stained with 7E1 for magicin or (f) without primary antibody. Endogenous magicin localizes in neurites and beneath the plasma membrane. Bar, 10 μm. (g) CAD cells were stained with 7E1 for magicin and (h) with Alexa Fluor® 488 phalloidin for actin. (i) Merged images. Magicin colocalizes with actin in neurites and underneath the plasma membrane. Bar, 20 μm

The association of magicin with merlin was further investigated by immunoelectron microscopy. Immunogold staining of CAD cells with 7E1 and a polyclonal antibody to merlin (A19) showed that endogenous magicin resides on cortical actin filaments and colocalizes with merlin (5 nm gold) in many cases (areas of colocalization are circled). Gold (10 nm)-identifying magicin is evenly distributed along filaments and does not localize preferentially to substructures such as branching points or filament ends. Extensive staining of magicin can also be seen along the actin cytoskeleton within neurites (Figure 6). When the primary antibody was omitted as a negative control, no labeling with gold particles was found (data not shown).

Figure 6

Colocalization of merlin and magicin in the cortex of a CAD cell determined by electron microscopy. Merlin and magicin colocalized in the Triton X-100-insoluble cytoskeleton from CAD cells using a rabbit polyclonal anti-merlin antibody A19 (Santa Cruz Biotechnology) and anti- magicin mouse monoclonal antibody 7E1, followed by 5 nm anti-rabbit immunogold and 10 nm anti-mouse immunogold particles. Magicin and merlin colocalize in many areas along the cytoskeleton (circles). No gold labeling was found when primary antibodies were omitted or when preimmune IgG was used as a control. The inset (lower right) shows a lower magnification view of a portion of a CAD cell. The bracketed area indicates the region shown at high magnification. Bar, 100 nm

Magicin binds directly to Grb2

Based on the Scansite prediction that magicin has a consensus sequence (pYVNG) for binding to the SH2 domain of the adaptor protein Grb2 (Schlessinger, 1994), we tested the interaction of magicin with endogenous Grb2 by affinity binding assays. The results of three independent experiments demonstrate that magicin specifically captured Grb2, and did not bind to the GST control protein (Figure 7a). The in vivo association between magicin and Grb2 was also confirmed by co-immunoprecipitation experiments. Heavy membrane/cytosolic fractions prepared from T-47D cells, which show an enrichment of these proteins, were immunoprecipitated with anti-magicin antibody (Tim3) or preimmune serum and subjected to immunoblot analysis. The results indicate that Grb2 can co-immunoprecipitate with magicin (Figure 7b). Moreover, we could also detect Grb2 along with magicin when anti-merlin antibodies were employed in co-immunoprecipitation experiments. Neither protein was detected when an anti-ezrin antibody was employed as a negative control (Figure 7c), confirming the specific interaction between magicin and merlin. Thus, magicin, merlin and Grb2 interact at the endogenous level in a physiologically relevant manner and may exist together in a complex.

Figure 7

Magicin binds Grb2 in vitro and in vivo. (a) NP-40 lysates from CAD cells were incubated with GST or GST-magicin fusion proteins and bound proteins were detected with an anti-Grb2 polyclonal antibody. (b) Magicin and Grb2 co-immunoprecipitate. Cytosolic/heavy membrane fractions (Cyt/HMF) of T-47D cell lysates were subjected to immunoprecipitation with the anti-magicin polyclonal antibody Tim3 or the preimmune serum as a control. Immunoprecipitates were detected for magicin and Grb2 employing respective monoclonal antibodies together. The lysate lane represents about 1/20 of the amount used in the immunoprecipitation assays. (c) Merlin, magicin and Grb2 exist as a complex. Cytosolic/heavy membrane fractions of T-47D cell lysates were subjected to immunoprecipitation with anti-merlin rabbit polyclonal antibody (C26) or an anti-ezrin rabbit polyclonal antibody as a control. Immunoprecipitates were fractionated by 8–12% gradient SDS–PAGE and immunoblotted for merlin, ezrin, magicin and Grb2. The membrane fraction lane represents about 1/20 of the amount used in the immunoprecipitation assays. (d) His-magicin and BSA were subjected to 12% SDS–PAGE. Ponceau-S staining shows equal amounts of His-tagged magicin and BSA (left panel). GST-Grb2 fusion protein or GST alone was incubated with the membrane and then immunoblotted with an anti-GST antibody. Full-length GST-Grb2 and the SH3 domains of GST-Grb2 bind specifically and directly to magicin (right panel)

The interaction of magicin and Grb2 was further characterized to define the binding domain in Grb2. Upon induction with growth factors such as α-FGF, β-FGF, EGF, NGF, PDGF or insulin, magicin did not reveal tyrosine phosphorylation (data not shown), suggesting that binding may not occur through the SH2 domain of Grb2. Blot overlay assays were then employed to test a direct interaction between magicin and Grb2 and to examine if this association could be mediated through the SH3 domains of Grb2. Full-length Grb2, as well as the first and second SH3 domains of Grb2, bound specifically to magicin but not to bovine serum albumin (BSA) used as a control protein (Figure 7d). These experiments demonstrate that magicin can bind directly to Grb2 through the SH3 domains.

Under medium stringency, Scansite predicted binding of the polyproline region of magicin with other SH3 domain-containing proteins; therefore, the interaction of magicin with c-Abl, Nck, cortactin, Crk, RasGAP and PLC-γ was also examined by affinity binding assays. Although the SH3 domains of Nck and Abl were predicted to bind better with magicin than Grb2, we did not observe binding of these proteins by affinity binding assays, and therefore did not pursue further. The SH3 domain of cortactin showed a relatively weak interaction, while others showed no evidence of an interaction (data not shown).


Merlin, like the other ERM family members, provides a link between plasma membrane proteins and the cortical actin cytoskeleton. This study extends the role of merlin via its interaction with magicin to signal transduction pathways mediated through the adaptor protein Grb2. Magicin, named for merlin and Grb2 interacting cytoskeletal protein, is a novel protein that binds merlin but not the other ERM proteins. The colocalization of magicin with merlin and the observed binding between these two proteins indicate that magicin is a biologically relevant partner of merlin. These results, along with the initial finding that merlin and magicin interact in a yeast two-hybrid assay, are consistent with a direct interaction between the two proteins. We showed earlier that compared to merlin isoform 1, isoform 2 bound to a greater extent NHE-RF as well as F-actin, possibly due to masking of the amino-terminal binding sites in isoform 1 resulting from inter- and intramolecular associations (Gonzalez-Agosti et al., 1999; James et al., 2001). In this study, when free carboxy-terminal domains of merlin isoform 1, isoform 2, or the common region to both isoforms, were employed, an enhanced binding to magicin was still observed with isoform 2 (Figure 2), indicating that differential binding may occur as a result of specific sequence differences between isoforms 1 and 2 and not simply conformational changes.

Magicin is a 178 aa evolutionarily conserved protein with no predicted transmembrane domains or signal sequences. The coiled-coil domain spanning aa 109–145 may play a role in mediating protein–protein interactions. Magicin migrates at 24 kDa, which is slightly larger than the predicted size of 20 kDa. In some cell types, an additional 50 kDa protein is detected with one of our antibodies and this species needs further characterization. Our results also document that magicin binds directly to Grb2 in vitro as well as in vivo. The Grb2 adaptor consists of a single SH2 domain flanked by two SH3 domains. Recruitment of Grb2 to activated and autophosphorylated receptors is mediated by the SH2 domain and is known to occur either directly or via linker proteins such as SHP-2 and Shc (Buday, 1999). Each SH3 domain preferentially binds proteins containing proline-rich motifs that adopt a left-handed polyproline type II helix with the minimal consensus sequence PxxP (Mayer, 2001).

Although magicin has a consensus sequence YVNG (aa 64–67) for binding the SH2 domain of Grb2, we have not observed Y64 to be phosphorylated after various mitogenic stimuli. Furthermore, we have shown that amino-terminal SH3 domain of Grb2 binds stronger than the carboxy-terminal SH3 domain to magicin; however, full-length Grb2 displays the strongest binding (Figure 7d). Thus, interaction of magicin with Grb2 might occur exclusively via the SH3 domains. However, it cannot be ruled out that the binding of magicin to Grb2 is mediated through both the SH2 and SH3 domains of Grb2 and further studies are necessary to clarify this. One of the well-characterized partners of the Grb2 SH3 domain is the nucleotide exchange factor Sos, and the mechanism of Ras activation through the formation of a ternary complex EGFR/Sos/Grb2 is well established (Buday, 1999). Nonetheless, many studies have clearly demonstrated that in addition to Sos, other effector molecules can bind to the SH3 domains of Grb2, which include adaptor proteins, phosphotyrosine phosphatases, serine/threonine kinases, and more importantly proteins implicated in cytoskeletal reorganization such as dynamin, dystroglycan, Wiskott–Aldrich syndrome proteins WASP and N-WASP (Gout et al., 1993; Yang et al., 1995; Miki et al., 1996; She et al., 1997). Therefore, magicin can be added to this growing list of cytoskeletal proteins that bind to the SH3 domains of Grb2.

During completion of this work, a report appeared showing the identification of a novel endothelial-derived gene (EG-1), isolated by suppression subtractive hybridization performed on control endothelial cells versus endothelial cells exposed to tumor conditioned media. The sequence of the EG-1 gene is identical to magicin (Liu et al., 2002). In that study, the authors did not characterize the protein encoded by the EG-1 sequence, but showed an upregulation of transcripts in endothelial cells when exposed to tumor conditioned media, suggesting that EG-1 may play a role in tumor angiogenesis.

Merlin shares many properties in common with the ERM proteins, yet has a unique and important tumor suppressor function. The identification of proteins that bind differentially to merlin and ERM proteins is one way to tease apart the distinct roles they play. Magicin, a merlin-specific interactor, through its association with Grb2 could mediate merlin's role as a growth regulator. The direct binding of Grb2 to activated and phosphorylated receptors, particularly EGFR, and its indirect association with PDGFR through linker protein SHP-2, potentially places the merlin–magicin complex in these signaling pathways. In this context, it is interesting to note that a study in Drosophila identified negative regulators of both the epidermal growth factor and transforming growth factor β pathways as genetic modifiers of a dominant-negative merlin phenotype, indicating that merlin might negatively regulate these pathways (LaJeunesse et al., 1998). Studies examining the interaction between Drosophila homologues of merlin and magicin will further clarify their role in the EGFR pathway. Merlin's ability to suppress growth has been suggested to act through its interaction with CD44 (Morrison et al., 2001) and HRS (Scoles et al., 2000) as well as its function to regulate Rac negatively (Shaw et al., 2001). Thus, merlin might function at different levels as growth regulator through its participation in a variety of signaling pathways.

We and others have shown that merlin can bind to F-actin through actin binding sites within the FERM domain (Xu and Gutmann, 1998; Brault et al., 2001; James et al., 2001). NF2-deficient schwannoma cells exhibit cytoskeletal abnormalities manifested by membrane ruffling and large surface areas (Pelton et al., 1998), and reintroduction of merlin reverses these abnormalities (Bashour et al., 2002) supporting a role for merlin in reorganization of the actin cytoskeleton. Both merlin and magicin are present in the detergent-insoluble cytoskeletal fraction. The interaction of magicin with merlin and Grb2 might enable its association with actin. In this context, it is highly relevant to note that Grb2 through its SH3-mediated interaction with N-WASP is suggested to activate Arp2/3 complex-mediated actin polymerization downstream from the receptor tyrosine kinase signaling pathway (Carlier et al., 2000). Taken together, these results suggest that merlin, along with magicin, could be part of a multiprotein complex that is essential for cellular signaling to the actin cytoskeleton downstream of receptor tyrosine kinase signaling pathways. Determining the functional implications of merlin–magicin and magicin–Grb2 interactions in the context of observed cytoskeletal abnormalities in merlin-deficient cells, as well as merlin's tumor suppressor function in NF2-related benign tumors and non-NF2-related malignant tumors will be a challenge for future studies.

Materials and methods

Molecular cloning of magicin

Merlin constructs representing full-length (aa 1–595), amino terminus (aa 1–340) and carboxy terminus (aa 341–595) of isoform 1 were used as baits to screen a human fetal frontal cortex interaction library as previously described (Murthy et al., 1998). The cDNA clone obtained (aa 13–178) from this screen was referred to as MerIntA. The Marathon™ cDNA Amplification Kit (Clontech) was utilized according to the manufacturer's recommendations to obtain a full-length clone. The control bait plasmids included bicoid, CDC2, APP, IQGAP and neurofibromin.

Northern blot analysis

Human multiple tissue blots (Clontech) were hybridized according to the manufacturer's recommendations using 32P-labeled and ultrapurified cDNA encoding magicin as a probe. Blots were hybridized at 42°C for 16 h, washed and exposed to X-Omat AR film (Kodak Eastman Co.).

DNA constructs and expression in Escherichia coli

Full-length merlin isoform 1 (aa 1–595) and isoform 2 (aa 1–590) as well as various merlin truncations (aa 1–332 and 308–579, common to both isoforms; 340–595 of isoform 1; 340–590 of isoform 2) were expressed as GST fusion proteins using the vector pGEX2T (Amersham Biosciences). Full-length (aa 1–577), amino terminus (aa 1–332) and carboxy terminus (aa 307–577) of human moesin were expressed as GST fusion proteins using the vector pGEX4T1 (Amersham Biosciences). Expression and purification of all the GST fusion proteins were performed as described previously (Murthy et al., 1998). To generate GST-magicin fusion protein, magicin (aa 13–178) was cloned into pGEX-4T1, expressed and purified as for merlin. To generate His-tagged magicin, aa 13–178 was cloned into pQE30 (Qiagen), expressed and purified using the denaturing conditions as recommended by the manufacturer.

For mammalian expression, several epitope-tagged constructs were used. Flag-tagged full-length merlin isoforms 1 and 2 were described previously (Gonzalez-Agosti et al., 1999). S518A and S518D mutants were generated from Flag-tagged merlin isoforms 1 and 2 using the QuickChange site-directed mutagenesis kit (Stratagene) and were sequenced to verify the changes. GFP-tagged merlin isoforms 1 and 2 were made in pEGFP-N1 (Clontech). MerIntA, a hemagglutinin-epitope (HA)-tagged construct of magicin (aa 13–178) containing 90 additional amino acids of the pJG4-5 vector, was derived from the cDNA library of the yeast two-hybrid system. Full-length HA-tagged magicin (A4) (aa 1–178) was engineered to have a 5′ HA tag (YPYDVPDYA) and was cloned into pcDNA3 (Invitrogen).

Cell lines and transfections

CAD cells (mouse neuronal cells) were cultured in Dulbecco's modified Eagle's medium: Nutrient Mix F-12, D-MEM/F-12 (GibcoBRL) containing 10% fetal bovine serum (FBS; Hyclone) and 1% penicillin–streptomycin (GibcoBRL). Other cell lines, including COS-7 (monkey kidney fibroblast), F293 (fibroblast), Gli238 (glioma), HeLa (human adenocarcinoma epithelial), Met5A (mesothelioma) and 293T (human embryonic kidney), T-47D (human breast carcinoma epithelial), were maintained in MEM Eagle (BioWhittaker) or D-MEM (GibcoBRL) supplemented with 10% FBS (Sigma) and 1% penicillin–streptomycin (GibcoBRL).

CAD, COS-7 and 293T cells were transfected using LipofectAMINE 2000 (Invitrogen) according to the manufacturer's recommendations.

Generation of magicin antibodies

The rabbit polyclonal antibody (Tim3) against aa 165–178 of human magicin was generated by Research Genetics Inc. For a monoclonal antibody, mice were injected with approximately 50 μg of GST-magicin (aa 13–178). After a final boost, spelenocytes were harvested and fused to SP2 myeloma cells. Fusions and selection were carried out as described before (Harlow and Lane, 1988). The resulting monoclonal antibody (7E1) was determined as IgG2b.

Antibodies and reagents

The in-house antibodies used in this study are as follows: anti-merlin, C26 (aa 308–579 of isoform 1) rabbit and chicken polyclonal antibodies and N21 (aa 1–332 of merlin) rabbit polyclonal, which recognizes ezrin, moesin and merlin. Commercial antibodies used in this study include anti-GST, sc-138 (Santa Cruz Biotechnology); anti-Grb2, rabbit polyclonal antibody sc-255 (Santa Cruz Biotechnology) and mouse monoclonal antibody #610111 (BD Biosciences); anti-Flag M2 monoclonal antibody F 3165 (Sigma); anti-merlin, A19 (sc-331, Santa Cruz Biotechnology); anti-HA, 12CA5 (Roche). The anti-ezrin rabbit polyclonal antibody was kindly provided by Dr Anthony Bretscher (Cornell University, NY).

Horseradish peroxidase (HRP)-conjugated anti-mouse or anti-rabbit antibodies (Amersham Biosciences) were used at 1 : 10 000. HRP-conjugated anti-chicken antibodies (Sigma) were used at 1 : 15 000. Western blot analyses were visualized using the ECL system (Amersham Biosciences). Full-length, SH2 domain and two SH3 domains of GST-Grb2 fusion proteins were purchased from Santa Cruz Biotechnology.


Total cell lysates were extracted by ice-cold Nonidet P-40 Tris-based (NP-40/Tris) lysis buffer containing 0.5 or 1% NP-40, 150 mM NaCl, 50 mM Tris (pH 8.0) and 1 × Complete™ protease inhibitor cocktail (Roche) or 1% NP-40/HEPES lysis buffer containing 50 mM HEPES (pH 7.4), 150 mM NaCl, 1 mM Na3VO4, 2 mM EDTA and 1 × Complete™. The lysates were subjected to immunoprecipitation or affinity binding assays. Western analyses were performed as described (Gonzalez-Agosti et al., 1996).

Affinity binding assays

COS-7 cells transfected with 10 μg of HA-tagged magicin (aa 13–178) were lysed in 0.5% NP-40/Tris lysis buffer. Lysates were incubated with 600 pmol of various GST-merlin or GST-moesin fusion proteins or GST control and immobilized on glutathione Sepharose 4B beads (Amersham Biosciences) overnight at 4°C. Beads were washed, and the bound proteins were resolved by 10% SDS–PAGE. Magicin was detected with an anti-HA antibody.

CAD cells were lysed as described above except that lysates were incubated with GST-magicin (aa 13–178) or GST alone, and the blot was probed with anti-Grb2 polyclonal (1 : 1000; Santa Cruz Biotechnology).

Blot overlay assay

Purified His-magicin (3 μg) or BSA (3 μg) was subjected to 10% SDS–PAGE. After blocking in 10% milk, the membranes were incubated with respective GST-Grb2 proteins (3 μg/ml; Santa Cruz) for 16 h. Membranes were incubated with the anti-GST antibody (1 : 1000; Santa Cruz) for 1 h, and incubated with HRP-conjugated anti-mouse antibodies (1 : 10 000) for 1 h. Proteins were visualized using ECL system (Amersham Biosciences).

Co-immunoprecipitation assays

Confluent T-47D cells were lysed in ice-cold 1% NP-40/HEPES lysis buffer. Merlin was immunoprecipitated using C26 rabbit serum and protein A–agarose beads (Roche). Merlin and magicin were detected with C26 chicken antibody (1 : 200) and 7E1 (1 : 500), respectively. Reciprocal immunoprecipitations were performed using anti-magicin antibody Tim3 and immunodetection was carried out with C26 and 7E1. 293T cells were cotransfected with 20 μg of HA-tagged magicin A4 (aa 1–178) and 20 μg of Flag-tagged S518A or S518D mutants of merlin isoform 1 or 2. The cells were harvested and lysed as described above, and immunoprecipitated with anti-Flag antibody (Sigma) or normal mouse IgG (Santa Cruz Biotechnology) and protein G–agarose beads, and detected with anti-HA antibody (1 : 1000; Covance) and anti-Flag antibody (1 : 2000; Sigma).

Cytosol/heavy membrane fractions from confluent T-47D cells were obtained as described (Yashiro-Ohtani et al., 2000). Briefly, T-47D cells were lysed with 1% Triton/MBS buffer (1% Triton X-100, 25 mM MES, 150 mM NaCl, pH 6.5, 50 mM NaF, 1 mM Na2VO3, protease inhibitor cocktail) on ice for 30 min, followed by 20 strokes of loose-fitting homogenization. The lysates were subjected to sucrose gradient centrifugation and the pool of cytosol/heavy membrane fraction was lysed with 1% N-octyl-β-D-glucoside (Roche). These samples were incubated with C26 (anti-merlin), anti-ezrin or normal rabbit serum and protein A–agarose beads. The immunoprecipitates were detected with N21 (1 : 200), which recognizes both ezrin and merlin, anti-magicin 7E1 (1 : 500) and anti-Grb2 monoclonal (1 : 500).

Cell fractionation assay

Cell fractionations were performed as previously described (Stokowski and Cox, 2000). The detergent-soluble material was precipitated with 85% acetone at −20°C for 16 h and recovered by centrifugation at 14 000 g for 15 min. Detergent-insoluble material was scraped in PBS containing protease inhibitors and centrifuged at 14 000 g for 15 min. All samples were resuspended in the same volume of sample buffer, and examined by Western blot analysis to detect merlin and magicin expression.


CAD cells grown on glass coverslips were cotransfected (16 h) with full-length HA-magicin (A4) and GFP-merlin isoform 2 and stained as described earlier (Gonzalez-Agosti et al., 1996). The primary anti-HA antibody 12CA5 (1 : 200; Roche) was incubated at 37°C for 1 h and the secondary antibody, rhodamine-conjugated anti-mouse antibody (BioSource International), was incubated at 37°C for 30 min. For endogenous localization of magicin, CAD cells were treated with 0.1% SDS for 1 min to augment antigenicity and stained with 7E1 (1 : 10). Magicin was then visualized using a Tetramethyl-Rhodamine TSA Fluorescence System (NEN Life Science Products) according to the manufacturer's recommendations. To visualize F-actin, Alexa Fluor® 488 phalloidin (1 : 1000; Molecular Probes) was incubated at 37°C for 30 min after the incubation of primary antibodies. Coverslips were washed extensively between antibody incubations and mounted using the ProLong® Antifade Kit (Molecular Probes). Confocal images were obtained with a Nikon TE 300 microscope and the BioRad MRC 100 laser imaging system.

Electron microscopy

Immunogold labeling of merlin and magicin was conducted as previously described (James et al., 2001). Glass-adherent cytoskeletons of CAD cells were treated with 0.1% SDS for 1 min to augment antigenicity after fixation with 1% glutaraldehyde for 10 min and incubated with rabbit anti-merlin antibody A19 (1 : 10) and with mouse anti-magicin antibody 7E1 (1 : 10) at 25°C for 1 h. After washing, coverslips were incubated with goat anti-rabbit IgG-coated 5 nm colloidal gold particles for merlin staining and with goat anti-mouse IgG-coated 10 nm colloidal gold particles (1 : 20) for magicin staining. Coverslips were washed, fixed, washed in distilled water, rapidly frozen, freeze dried and metal coated. Replicas were analysed and photographed in a JEOL 1200-EX electron microscope at 100 kV accelerating voltage.


  1. Bashour AM, Meng JJ, Ip W, MacCollin M and Ratner N . (2002). Mol. Cell. Biol., 22, 1150–1157.

  2. Brault E, Gautreau A, Lamarine M, Callebaut I, Thomas G and Goutebroze L . (2001). J. Cell Sci., 114, 1901–1912.

  3. Bretscher A, Edwards K and Fehon RG . (2002). Nat. Rev. Mol. Cell Biol., 3, 586–599.

  4. Buday L . (1999). Biochim. Biophys. Acta, 1422, 187–204.

  5. Carlier MF, Nioche P, Broutin-L’Hermite I, Boujemaa R, Le Clainche C, Egile C, Garbay C, Ducruix A, Sansonetti P and Pantaloni D . (2000). J. Biol. Chem., 275, 21946–21952.

  6. Deguen B, Merel P, Goutebroze L, Giovannini M, Reggio H, Arpin M and Thomas G . (1998). Hum. Mol. Genet., 7, 217–226.

  7. Fernandez-Valle C, Tang Y, Ricard J, Rodenas-Ruano A, Taylor A, Hackler E, Biggerstaff J and Iacovelli J . (2002). Nat. Genet., 31, 354–362.

  8. Gonzalez-Agosti C, Wiederhold T, Herndon ME, Gusella J and Ramesh V . (1999). J. Biol. Chem., 274, 34438–34442.

  9. Gonzalez-Agosti C, Xu L, Pinney D, Beauchamp R, Hobbs W, Gusella J and Ramesh V . (1996). Oncogene, 13, 1239–1247.

  10. Gout I, Dhand R, Hiles I, Fry M, Panayotou G, Das P, Truong O, Totty N, Hsuan J and Booker G . (1993). Cell, 75, 25–36.

  11. Goutebroze L, Brault E, Muchardt C, Camonis J and Thomas G . (2000). Mol. Cell. Biol., 20, 1699–1712.

  12. Gronholm M, Sainio M, Zhao F, Heiska L, Vaheri A and Carpen O . (1999). J. Cell Sci., 112, 895–904.

  13. Gronholm M, Vossebein L, Carlson CR, Kuja-Panula J, Teesalu T, Alfthan K, Vaheri A, Rauvala H, Herberg FW, Tasken K and Carpen O . (2003). J. Biol. Chem., 278, 41167–41172.

  14. Gusella JF, Ramesh V, MacCollin M and Jacoby LB . (1999). Biochim. Biophys. Acta, 1423, M29–M36.

  15. Harlow E and Lane D . (1988). Cold Spring Harbor Press: Cold Spring Harbor, NY.

  16. James MF, Manchanda N, Gonzalez-Agosti C, Hartwig JH and Ramesh V . (2001). Biochem. J., 356, 377–386.

  17. Jannatipour M, Dion P, Khan S, Jindal H, Fan X, Laganiere J, Chishti AH and Rouleau GA . (2001). J. Biol. Chem., 29, 29.

  18. Kissil JL, Johnson KC, Eckman MS and Jacks T . (2002). J. Biol. Chem., 277, 10394–10399.

  19. Kissil JL, Wilker EW, Johnson KC, Eckman MS, Yaffe MB and Jacks T . (2003). Mol. Cell, 12, 841–849.

  20. LaJeunesse DR, McCartney BM and Fehon RG . (1998). J. Cell Biol., 141, 1589–1599.

  21. Liu C, Zhang L, Shao ZM, Beatty P, Sartippour M, Lane TF, Barsky SH, Livingston E and Nguyen M . (2002). Biochem. Biophys. Res. Commun., 290, 602–612.

  22. Maeda M, Matsui T, Imamura M and Tsukita S . (1999). Oncogene, 18, 4788–4797.

  23. Mayer BJ . (2001). J. Cell Sci., 114, 1253–1263.

  24. Miki H, Miura K and Takenawa T . (1996). EMBO J., 15, 5326–5335.

  25. Morrison H, Sherman LS, Legg J, Banine F, Isacke C, Haipek CA, Gutmann DH, Ponta H and Herrlich P . (2001). Genes Dev., 15, 968–980.

  26. Murthy A, Gonzalez-Agosti C, Cordero E, Pinney D, Candia C, Solomon F, Gusella J and Ramesh V . (1998). J. Biol. Chem., 273, 1273–1276.

  27. Neill GW and Crompton MR . (2001). Biochem. J., 358, 727–735.

  28. Nguyen R, Reczek D and Bretscher A . (2001). J. Biol. Chem., 276, 7621–7629.

  29. Obremski VJ, Hall AM and Fernandez-Valle C . (1998). J. Neurobiol., 37, 487–501.

  30. Pelton PD, Sherman LS, Rizvi TA, Marchionni MA, Wood P, Friedman RA and Ratner N . (1998). Oncogene, 17, 2195–2209.

  31. Reczek D, Berryman M and Bretscher A . (1997). J. Cell Biol., 139, 169–179.

  32. Rouleau GA, Merel P, Lutchman M, Sanson M, Zucman J, Marineau C, Hoang-Xuan K, Demczuk S, Desmaze C, Plougastel B, Pulst SM, Lenoir G, Bijlsma E, Fashold R, Dumanski J, de Jong P, Parry D, Eldridge R, Aurias A, Delattre O and Thomas G . (1993). Nature, 363, 515–521.

  33. Sainio M, Zhao F, Heiska L, Turunen O, den Bakker M, Zwarthoff E, Lutchman M, Rouleau GA, Jaaskelainen J, Vaheri A and Carpen O . (1997). J. Cell Sci., 110, 2249–2260.

  34. Schlessinger J . (1994). Curr. Opin. Genet. Dev., 4, 25–30.

  35. Scoles DR, Huynh DP, Chen MS, Burke SP, Gutmann DH and Pulst SM . (2000). Hum. Mol. Genet., 9, 1567–1574.

  36. Scoles DR, Huynh DP, Morcos PA, Coulsell ER, Robinson NG, Tamanoi F and Pulst SM . (1998). Nat. Genet., 18, 354–359.

  37. Shaw RJ, Paez JG, Curto M, Yaktine A, Pruitt WM, Saotome I, O’Bryan JP, Gupta V, Ratner N, Der CJ, Jacks T and McClatchey AI . (2001). Dev. Cell, 1, 63–72.

  38. She H, Rockow S, Tang J, Nishimura R, Skolnik E, Chen M, Margolis B and Li W . (1997). Mol. Biol. Cell, 8, 1709–1721.

  39. Sherman L, Xu HM, Geist RT, Saporito-Irwin S, Howells N, Ponta H, Herrlich P and Gutmann DH . (1997). Oncogene, 15, 2505–2509.

  40. Stokowski RP and Cox DR . (2000). Am. J. Hum. Genet., 66, 873–891.

  41. Surace EI, Haipek CA and Gutmann DH . (2004). Oncogene, 23, 580–587.

  42. Takahashi K, Sasaki T, Mammoto A, Takaishi K, Kameyama T, Tsukita S and Takai Y . (1997). J. Biol. Chem., 272, 23371–23375.

  43. Trofatter JA, MacCollin MM, Rutter JL, Murrell JR, Duyao MP, Parry DM, Eldridge R, Kley N, Menon AG, Pulaski K, Haase VH, Ambrose CM, Munroe D, Bove C, Haines JL, Martuza RL, MacDonald ME, Seizinger BR, Short MP, Buckler A and Gusella JF . (1993). Cell, 72, 791–800.

  44. Tsukita S, Oishi K, Sato N, Sagara J and Kawai A . (1994). J. Cell Biol., 126, 391–401.

  45. Xiao GH, Beeser A, Chernoff J and Testa JR . (2002). J. Biol. Chem., 277, 883–886.

  46. Xu HM and Gutmann DH . (1998). J. Neurosci. Res., 51, 403–415.

  47. Yang B, Jung D, Motto D, Meyer J, Koretzky G and Campbell K . (1995). J. Biol. Chem., 270, 11711–11714.

  48. Yashiro-Ohtani Y, Zhou XY, Toyo-Oka K, Tai XG, Park CS, Hamaoka T, Abe R, Miyake K and Fujiwara H . (2000). J. Immunol., 164, 1251–1259.

Download references


We thank Drs J Settleman, N Ramesh and members of our laboratory for helpful comments on the manuscript. Our sincere thanks to Roberta Beauchamp for her help with the cytoskeletal preparations as well as helping with the figures. We also thank Johanna Lelke for her help with the mutant merlin constructs. This work was supported in part by National Institutes of Health Grant NS24279, Department of Defense Grant DAMD17-02-0647 and by funds from Neurofibromatosis Inc. (Massachusetts Chapter) and the Texas NF Foundation.

Author information

Correspondence to Vijaya Ramesh.

Rights and permissions

Reprints and Permissions

About this article


  • NF2
  • merlin
  • tumor suppressor
  • Grb2
  • cytoskeleton

Further reading