ERM (ezrin-radixin-moesin) proteins are components of thecortical cytoskeleton and play a role in linking the actin-containing cytoskeleton to cell membrane molecules. TheERM family consists of three closely related proteins; ezrin(Gould et al., 1989; Turunen et al., 1989), radixin (Funayamaet al., 1991) and moesin (Lankes and Furthmayr, 1991). Theyshare 75-80% sequence homology and consist of threedomains: a globular amino-terminal domain, an extended α-helical domain and a charged carboxy-terminal domain (Vaheriet al., 1997). The amino-terminal domain has considerablehomology with members of the band 4.1 family of membrane-organizing proteins.
The amino-terminal domain of ERM proteins interacts withplasma membrane molecules such as CD43, CD44,intercellular adhesion molecule-1 (ICAM-1) and ICAM-2(Helander et al., 1996; Hirao et al., 1996; Heiska et al., 1998;Yonemura et al., 1998), whereas binding sites for actin arecontained both in the amino- and carboxy-terminal domain(Turunen et al., 1994; Pestonjamasp et al., 1995; Roy et al.,1997). ERM proteins regulate the cell surface distribution of
adhesion molecules, organization of cell membrane structuresand maintenance of cell shape (Lamb et al., 1997; Vaheri et al.,1997). These functions are partially redundant within the ERMfamily members. When the expression of all three ERMproteins is blocked by antisense oligonucleotides microvillidisappear and cellular adhesion is disrupted (Takeuchi et al.,1994).
Homotypic and heterotypic association of ERM proteinshas been described (Gary and Bretscher, 1993, 1995;Andreoli et al., 1994; Magendantz et al., 1995; Pestonjamaspet al., 1995). The N-ERMAD (ezrin-radixin-moesinassociation domain) in ezrin has been mapped to amino acids1-296 and the C-ERMAD to amino acids 479-585 (Gary andBretscher, 1995). The carboxy-terminal functional domain,including the F-actin binding site, is masked in the nativemonomer but can be exposed by sodium dodecyl sulphate orif the domain is expressed as a truncated protein (Gary andBretscher, 1995). It has been suggested that intramolecular orintermolecular association of ERM proteins, in a head totail manner, would be a way to regulate their functionalactivity (Gary and Bretscher, 1995; Henry et al., 1995;Martin et al., 1995). According to this model, the folded state
Ezrin, radixin and moesin (ERM) are homologous proteins,which are linkers between plasma membrane componentsand the actin-containing cytoskeleton. The ERM proteinfamily members associate with each other in a homotypicand heterotypic manner. The neurofibromatosis 2 (NF2)tumor suppressor protein merlin (schwannomin) isstructurally related to ERM members. Merlin is involvedin tumorigenesis of NF2-associated and sporadicschwannomas and meningiomas, but the tumor suppressormechanism is poorly understood. We have studied theability of merlin to self-associate and bind ezrin. Ezrin wascoimmunoprecipitated with merlin from lysates of humanU251 glioma cells and from COS-1 cells transfected withcDNA encoding for merlin isoform I. The interaction wasfurther studied and the association domains were mappedwith the yeast two-hybrid system and with blot overlay andaffinity precipitation experiments. The heterotypic bindingof merlin and ezrin and the homotypic association of merlin
involves interaction between the amino- and carboxy-termini. The amino-terminal association domain of merlininvolves residues 1-339 and has similar features with theamino-terminal association domain of ezrin. The carboxy-terminal association domain cannot be mapped as preciselyas in ezrin, but it requires residues 585-595 and a moreamino-terminal segment. Unlike ezrin, merlin does notrequire activation for self-association but native merlinmolecules can interact with each other. Heterodimerizationbetween merlin and ezrin, however, occurs only followingconformational alterations in both proteins. These resultsbiochemically connect merlin to the cortical cytoskeletonand indicate differential regulation of merlin from ERMproteins.
Homotypic and heterotypic interaction of the neurofibromatosis 2 tumor
suppressor protein merlin and the ERM protein ezrin
Mikaela Grönholm 1,*, Markku Sainio 1, Fang Zhao 1, Leena Heiska 1, Antti Vaheri 2 and Olli Carpén 1
Departments of 1Pathology and 2Virology, University of Helsinki, Haartman Institute, PO Box 21 (Haartmaninkatu 3), FIN-00014Helsinki*Author for correspondence (e-mail: [email protected])
Accepted 23 December 1998; published on WWW 25 February 1999
896
of the monomers and the homotypic/heterotypic oligomers inthe cytoplasm represent the inactive form of the protein.Signals disrupting this intramolecular or intermolecularinteraction may expose masked functional sites, such asmembrane-binding and actin-binding domains, allowingERM proteins to bind to other partners and function as cross-linkers.
The neurofibromatosis 2 (NF2) tumor suppressor proteinmerlin (schwannomin) is structurally related to ERM proteins(Rouleau et al., 1993; Trofatter et al., 1993). Inactivation of theNF2gene, which encodes for merlin, leads to the developmentof schwannomas and meningiomas in the dominantly inheritedNF2 disease. NF2 gene mutations are also found in sporadicschwannomas and a proportion of meningiomas (Louis et al.,1995). In addition to the overall ERM-like domain structure,merlin possesses some functional properties of ERM familymembers. The amino acid identity between merlin and ezrin is61% in the amino-terminal domain, but the carboxy-terminaldomain shares only 22% identity with ezrin (Turunen et al.,1998).
Two major alternatively spliced NF2 variants are expressedin vivo. The isoform I, lacking exon 16, encodes for a 595amino acid protein with a predicted molecular mass of 66 kDa(Rouleau et al., 1993; Trofatter et al., 1993). Isoform IIcontains exon 16 which inserts 11 unique carboxy-terminalamino acids followed by a termination codon that preventstranslation of exon 17 (Bianchi et al., 1994). In cultured cells,merlin is localized underneath the plasma membrane in apattern typical of ERM proteins (Gonzalez-Agosti et al.,1996; Sainio et al., 1997). Overexpression of full-lengthisoform I induces morphogenic changes, such as cell surfaceprotrusions and elongation of cell body (Sainio et al., 1997).Transfected and endogenous merlin colocalizes with ezrin,although in cells with a poorly developed actin cytoskeletonmerlin replaces ezrin in filopodia and ruffling edges (Sainioet al., 1997). Based on these findings, merlin, similar to ERMproteins, is a cytoskeleton-associated membrane organizingprotein and thus a unique type of tumor suppressor. Thetumor suppressor mechanism of merlin is, however, poorlyunderstood. Overexpression of isoform I in rat schwannomacells and NIH 3T3 cells inhibits cell proliferation, whiletruncated constructs or isoform II fails to influenceschwannoma growth (Lutchman and Rouleau, 1995; Shermanet al., 1997).
So far, only few binding partners for merlin have beencharacterized. Merlin has been shown to associate with CD44but it is not known whether the interaction is direct or indirectthrough other molecules (Sainio et al., 1997). In addition,merlin, as well as ezrin, binds to the regulatory cofactor forNa+-H+ exchange, hNHE-RF or EBP50 (Reczek et al., 1997;Murthy et al., 1998) and merlin has also been shown to interactwith β-spectrin (Scoles et al., 1998) and associate with actinand microtubules (Xu and Gutmann, 1998). Whether theseinteractions are relevant to merlin’s antiproliferative effect is,however, unresolved. Since the carboxy terminus of merlin isnot as conserved as in other ERM proteins, it is also unclear ifthe activity of merlin is regulated by head-to-tail association asis the case with other ERM family members. In this paper wehave tested whether merlin is able to form homotypicinteractions and to interact with the prototypic ERM familymember, ezrin.
MATERIALS AND METHODS
AntibodiesAll antibodies have been characterized. Ezrin was detected with 3C12mAb (Böhling et al., 1996). Two merlin-specific rabbit antisera raisedagainst synthetic peptides were used. Anti-schwannomin detectsisoform I (Lutchman and Rouleau, 1995), whereas 1398NF2 detectsboth isoform I and II (den Bakker et al., 1995). Polyclonal LexAantibody was kindly provided by Dr E. Golemis, Fox Chase CancerCenter, Philadelphia, PA (Samson et al., 1989). 12CA5 mAb, whichreacts with the influenza virus hemagglutinin epitope (HA), was fromBoehringer-Mannheim, GmbH, Mannheim, Germany. As controls,preimmune serum of 1398NF2 rabbit (pre) and X63 mAb (ATCC,Maryland, USA) were used.
Confocal microscopyHuman U251 glioma cells (Westermark, 1973), expressingendogenous merlin and ezrin, were grown on glass coverslips inMEM/10% fetal bovine serum and serum starved for 20 hours beforefixation in −20°C methanol. For double staining of merlin and ezrin,fixed cells were incubated simultaneously with 1398NF2 antiserum(1:100 dilution) and 3C12 mAb (1:200 dilution), followed by TRITC-conjugated goat anti-rabbit IgG (Dako A/S, Copenhagen, Denmark)and FITC-conjugated goat anti-mouse IgG (Dako). Specimens wereviewed with a confocal 410 Invert Laser Scan Microscope (Carl Zeiss,Oberkochen, Germany).
Coimmunoprecipitation of ezrin with merlin in U251glioma cells and transfected COS-1 cellsU251 glioma cells from a 10 cm plate were lysed in 500 µl ELB-buffer (50 mM Hepes, pH 7.4, 150 mM NaCl, 5 mM EDTA), 1% NP-40 and protease inhibitors, and centrifuged at 15,000 g for 1 hour at4°C. The supernatant was incubated with 3C12 mAb, anti-schwannomin antiserum or pre-immune rabbit serum and Protein G-Sepharose beads (Pharmacia, Uppsala, Sweden) for 4 hours at 4°C.Immunoprecipitates were washed with ELB-0.1% NP-40 and boundproteins were eluted from the beads by boiling in non-reducingLaemmli sample buffer. Samples were separated on SDS-PAGE,transferred to nitrocellulose filters (Schleicher & Schuell, Dassel,Germany) and immunoblotted with 3C12 mAb (1:3000 dilution) and1398NF2 rabbit antiserum (1:3000 dilution) for 2 hours andperoxidase conjugated sheep anti-mouse antibody (Ig-POD) or sheepanti-rabbit antibody (IgG-POD) (Boehringer-Mannheim GmbH)(1:1500 dilution) for 30 minutes. The bound antibodies were detectedby enhanced chemiluminescence (Boehringer-Mannheim GmbH).
COS-1 cells were transfected with a cDNA encoding for full lengthmerlin isoform I kindly provided by Dr J. Gusella (Trofatter et al.,1993) subcloned in the pcDNA3 expression vector (Invitrogen, SanDiego, CA) (Sainio et al., 1997) or a control SVβ-Galactosidaseexpression vector (Clontech Laboratories, Inc., Palo Alto, CA) usingSuperfect (Qiagen, GmbH, Hilden, Germany). After 70 hours, cellswere lysed. Coimmunoprecipitation of ezrin with merlin was done asdescribed for U251 glioma cells but merlin was immunoprecipitatedwith 1398NF2 rabbit antiserum.
Chemical cross-linkingU251 glioma cells from a 10 cm plate were rinsed in PBS andincubated for 30 minutes at room temperature in PBS containing athiol-cleavable cross-linker dithiobis-succinimidyl-propionate (DSP,0.2 mM) (Pierce Chemical Co., Rockford, IL), protease inhibitors andphosphatase inhibitors (200 µM Na3VO4, 50 mM NaF and 50 mM β-glycerophosphate). The reaction was quenched for 15 minutes byaddition of 1 M Tris-Cl, pH 7.5, to obtain a final concentration of 50mM. The cells were lysed in non-reducing Laemmli sample buffer.Samples were separated on SDS-PAGE under non-reducing andreducing conditions and transferred to nitrocellulose filters. The filterswere subsequently immunoblotted for merlin with anti-schwannomin
M. Grönholm and others
897Homo- and heterotypic interaction of merlin and ezrin
antiserum (1:2000 dilution), and after stripping according tomanufacturer’s instructions, for ezrin with 3C12 mAb (1:3000dilution).
Recombinant DNA constructsThe merlin fusion protein constructs were generated from the cDNAencoding for merlin isoform I. For the yeast two-hybrid method, insertswere subcloned into the bait vector, EG202, which contains a LexA-DNA binding domain (Gyuris et al., 1993) or the prey vector JG 4-5,which contains the HA epitope tag (Gyuris et al., 1993). Merlin 1-100,1-167, 1-546, 1-585, 1-595, 339-585, 252-595 and 339-595 werecreated by digestion with restriction endonucleases and ligation into thebait and prey vector directly or via subcloning into pGEM expressionvector (Promega, Madison, WI). Merlin 1-339 was amplified by thePCR method and subcloned into the two-hybrid vectors. Ezrin 1-309and 1-585 were amplified by PCR, using the pCV6 clone (Turunen etal., 1989) as a template, and subcloned into the two-hybrid vectors.Ezrin 1-170 and 278-585 were generated by digestion with restrictionendonucleases and ligation into the two-hybrid vectors. The authentityof all constructs was verified by sequencing.
Yeast two-hybrid systemThe genotype of the Saccharomyces cerevisiaestrain BOY1, kindlyprovided by P. Ljungdahl, Ludwig Institute for Cancer Research,Stockholm, Sweden, is MATα his3 trp1 leu2::6LexAop-LEU2URA3::8LexAop-Gal1-LacZ. BOY1 mating type a was made usingthe YCpHO CUT4 plasmid (Raghuraman et al., 1994). Yeast strainswere grown at 30°C in rich medium or in synthetic minimal mediumwith appropriate amino acid supplements. Bait and prey constructswere transformed into BOY1-yeast of both a and α mating type usingthe TRAFO protocol (www.manitoba.ca/faculties/medicine/human-genetics/gietz/trafo.hmtl) and plated on selection plates. Clones weregrown to late logarithmic phase in selective medium. For analysis offusion protein expression, yeast cells from 1 ml of overnight culturewere lysed in reducing Laemmli sample buffer, the samples wereboiled and analyzed by SDS-PAGE and immunoblotting. Baits andpreys were grown on selection plates, replica plated together on richmedia plates for mating overnight and replica plated on double(tryptophane and histidine) selection with 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-gal) (Boehringer-MannheimGmbH) for detection of interactions.
Immunoblot analysisProtein samples were separated in 10% SDS-PAGE, transferred ontonitrocellulose sheets and blocked overnight using 5% nonfat milkpowder in PBS-0.1% Tween-20. Primary antibodies, anti-LexAantiserum (1:5000 dilution) and 12CA5 mAb (1:1500 dilution) inPBS-0.1% Tween-20 were incubated for 1 hour. Secondary antibodiesand detection were as described in the coimmunoprecipitationexperiment.
Determination of β-galactosidase activityFor quantitative studies, yeast strains were grown to an approximateOD600 of 1.0 in selective medium lacking histidine and tryptophan. β-Galactosidase activity was determined as described (Harshman et al.,1988). β-Galactosidase units were calculated using the formula 1000× OD420/cell volume (ml) × time of reaction (minutes) × OD600. Thefinal values were the result of four independent determinations.Variation was less than 15%.
Coimmunoprecipitation of two-hybrid constructsBOY1 yeast cotransformed with bait and prey constructs were grownovernight at 30°C in selective medium to an OD600 of 0.8-1.0. Cellswere washed once with PBS and lysed with a mini beadbeater(BioSpec Products, Inc., Bartlesville, OK) in the presence of 1 mlacid-washed glass beads (Sigma Chemical Co., St Louis, MO) in 200µl of ELB-buffer, 1% NP-40 and protease inhibitors. The debris was
removed by centrifugation, and the supernatant diluted to an NP-40concentration of 0.5%. Protein concentration was measured at A280nm. 250 µg of total protein lysate was incubated for 30 minutes onice with 12CA5 mAb (1:1000 dilution) followed by incubation withProtein G-Sepharose beads for 2 hours. The beads were washed inELB-0.1% NP-40 and bound proteins were eluted in reducingLaemmli sample buffer, boiled and analyzed by 10% SDS-PAGE andimmunoblotting.
Production of recombinant proteinsThe baculovirally expressed GST-merlin fusion protein, GST-m 1-595was produced as described (Sainio et al., 1997). Additional carboxy-terminal deletion constructs, GST-m 1-339 and GST-m 1-546, werecreated by digestion with restriction endonucleases and ligation intopAcG2T vector (Pharmingen, San Diego, CA). The proteins wereexpressed using the BaculoGold (Pharmingen) baculovirus expressionsystem and purified by glutathione agarose beads (Pharmingen). Theezrin baculovirus expression construct, a kind gift of P. Mangeat(Université Montpellier II, Montpellier, France) (Andreoli et al.,1994), was purified as described (Hirao et al., 1996). GST-ez 279-531and GST-ez 477-585 in pGEX vectors (Pharmacia) were expressed inEscherichia coliDH5α cells and purified by glutathione-Sepharosebeads as described (Turunen et al., 1994).
Blot overlayRecombinant wild-type ezrin (ez 1-585) and GST-ez 477-585 werebiotinylated as described (Gary and Bretscher, 1993). Ezrin, GST-merlin and GST-ezrin constructs, GST and BSA (1-2 µg/lane) wererun in 10% SDS-PAGE, blotted onto nitrocellulose filters and blockedovernight using 3% BSA in TBS-0.1% Tween-20. The blots wereincubated with biotin-labeled ezrin (0.1 µg/ml) in 1% BSA in TBS-0.1% Tween-20 for 4 hours. Peroxidase-conjugated avidin (1:10000dilution) (Extravidin, Sigma) was detected using enhancedchemiluminescence.
Affinity precipitationYeast lysates were prepared as above. 200 µg of total protein fromlysates was incubated with purified GST-fusion protein bound toglutathione-Sepharose beads (0.5-1 µg) for 45 minutes. The beadswere pelleted, the supernatant removed and the beads washed in ELB-0.1% NP-40-buffer. Bound proteins were eluted from the beads byboiling in non-reducing Laemmli sample buffer separated in 10%SDS-PAGE gels and analyzed by immunoblotting.
RESULTS
Distribution of both merlin and ezrin in human U251glioma cells depends on confluencyWe compared the distribution of merlin and ezrin in U251glioma cells, which express these proteins endogenously, insparse and confluent growth conditions. Previously, an increasein cell density has been reported to upregulate merlinexpression (Shaw et al., 1998). The subcellular distribution ofboth merlin and ezrin was affected by an increase inconfluency. In subconfluent U251 glioma cells, double stainingof merlin and ezrin revealed a highly overlapping subcellulardistribution at cell surface projections resembling rufflingedges (Fig. 1A-C). In confluent cell cultures, only few cellswith accumulation of merlin and ezrin at the cell peripherywere present. Instead, most of the cells had lost thesubmembranous staining pattern of both proteins, and stainingshowed a diffuse or punctate cytoplasmic pattern (Fig. 1D-F).The results demonstrate a concominant regulation for merlinand ezrin distribution under different growth conditions.
898
Association of merlin and ezrin in vivoThe colocalization of merlin and ezrin in U251 glioma cellsand in COS-1 cells (Sainio et al., 1997) and the previouslyreported heterodimerization between ERM proteins suggestedthat an interaction could occur between merlin and ezrin. Thepossible interaction was tested by coimmunoprecipitationexperiments. The cDNA encoding for merlin isoform I or β-galactosidase was transfected into COS-1 cells that do notexpress detectable amounts of endogenous merlin. After 70hours, merlin was detected by western blotting from lysatesof cells transfected with specific but not the control DNA (Fig.2A,B). Cell lysates were immunoprecipitated with merlin,ezrin (as a positive control) or an irrelevant control antiserumand coprecipitating ezrin was detected by western blotting.Fig. 2A shows that ezrin was present in the precipitateobtained by ezrin or merlin antibodies but not with controlantiserum. The specificity was confirmed by lack of ezrinreactivity in precipitates obtained from β-galactosidasetransfectants with merlin antiserum (Fig. 2B). In additionalcoimmunoprecipitation experiments from U251 glioma celllysates ezrin could be detected in precipitates by ezrin and
merlin antiserum but not by control antiserum (Fig. 2C). Theseresults indicate that merlin and ezrin form a complex in vivo.This is not, however, a proof for a direct interaction betweenthe proteins.
In further experiments, we tested whether merlin and ezrinexist as dimers in U251 cells. After treatment with a chemicalcross-linker, novel bands, at sizes between 140-170 kDa,appeared in merlin and ezrin immunoblots of cell lysates (Fig.2D). The bands were not present in lysates from cells that werenot cross-linked or in lysates separated under reducingconditions. One of the separated bands was identically blottedwith both merlin and ezrin antibodies, raising the possibilitythat it represents the merlin-ezrin heterodimer. The size of thelower band in the merlin immunoblot is consistent with amerlin homodimer. However, the possibility that merlin andezrin would form a complex with other partners cannot be ruledout.
General comments of the in vitro interaction studiesTo further study the interaction between merlin and ezrin andto map the interacting domains, several different approaches
M. Grönholm and others
Fig. 1. Distribution of ezrin and merlin in U251 glioma cells depends on confluency. U251 glioma cells of subconfluent (A-C) or confluent (D-E) density were fixed and double stained with 3C12 mAb for ezrin (ez) (A,D) and a 1398NF2 rabbit antiserum for merlin (m) (B,E). A controlstaining was done with X63 mAb (G) and preimmune 1398NF2 serum (pre) (F). The cells were viewed by confocal microscopy. Horizontalsections at the plane 1 µm above the growth substratum are shown. In C and F, the composite images of sections A and B, and D and E,respectively, show ezrin in green, merlin in red and areas where the proteins codistribute in yellow. Bars, 25 µm.
899Homo- and heterotypic interaction of merlin and ezrin
were used. A set of amino- and carboxy-terminal deletionconstructs of merlin and ezrin was generated and expressedusing the yeast two-hybrid system (Fig. 3). By the yeast matingsystem, combinations between all bait and prey constructswere analyzed, expressing the baits in the yeast mating type aand the preys in α mating type or vice versa. Interactions weredetected by the survival of yeast in leucine-deficient medium(not shown) and by induction of β-galactosidase production.The yeast coexpressing a bait (LexA-fusion) and a prey (HA-fusion protein) were also used for coimmunoprecipitation.Yeast lysates were coimmunoprecipitated using an anti-HAmAb and coprecipitating bait constructs were immunoblottedwith the anti-LexA antibody. The yeast two-hybrid systemshowed slight variation in the expression levels of the proteins,when expressed in bait or prey vectors (Fig. 4). Regardless ofwhether the constructs were expressed as baits or preys similarbinding results were obtained. Merlin constructs 1-585 and339-595 were transactivating as baits in the two-hybrid systemand therefore not suitable for scoring of β-galactosidasevalues. They could, however, be used in thecoimmunoprecipitation experiment. Shorter carboxy-terminalconstructs of merlin than m 339-595 could not be introducedinto yeast cells since the cells failed to grow aftertransformation. The constructs may be toxic to the yeast orthey may affect the cell proliferation.
Two additional independent methods were used to confirmthe yeast two-hybrid results. We used the blot overlayexperiments with purified recombinant protein constructs, as ithas been successful in mapping the association domains ofezrin (Gary and Bretscher, 1995). In these experimentsimmobilized full-length ezrin is denatured and therefore the C-ERMAD is active (Gary and Bretscher, 1995). We also tested,using the affinity precipitation method, whether carboxy-terminal domains of merlin and ezrin from total cell lysatesbind to GST-merlin.
Fig. 2. Coimmunoprecipitation of ezrin with merlin in transfectedCOS-1 cells and U251 glioma cells, and the effect of chemical cross-linking. (A-B) COS-1 cells were transfected with cDNA encoding formerlin isoform I (A) or, as a control, with with β-galactosidasecDNA(B). Merlin was precipitated from cell lysates with 1398NF2 rabbitantiserum bound to Protein G-Sepharose beads. Rabbit preimmuneserum was used as a negative control. Bound material was separatedby SDS-PAGE and transferred to nitrocellulose filters. Coprecipitatingezrin was detected with 3C12 mAb and enhancedchemiluminescence. IP indicates the antibody used forimmunoprecipitation and BLOT the antibody used for westernblotting. WB indicates immunoblotting of ezrin with 3C12 mAb andmerlin with 1398NF2 rabbit antiserum. Note the lack of endogenousmerlin in cells transfected with β-galactosidasecDNA (B) and thepresence of transfected protein in NF2cDNA transfectants (A).(C) Total lysates of U251 glioma cells were immunoprecipitated with3C12 mAb for ezrin, anti-schwannomin antiserum for merlin or withpreimmune serum. Coprecipitating ezrin was detected byimmunoblotting. WB shows immunoblot detection of endogenousezrin and merlin in U251 lysate. (D) U251 cells were exposed to thecross-linker DSP and subsequently lysed in non-reducing Laemmlisample buffer. Identical samples with DSP (+) or without DSP (−)cross-linking were separated on SDS-PAGE under non-reducing (non-red) and reducing (red) conditions. The filters were immunoblottedfor merlin with anti-schwannomin antiserum, and after stripping, forezrin with 3C12 mAb. In the absence of cross-linking and underreducing conditions, where the cross-linker is cleaved, only the bandsrepresenting the monomeric forms of merlin and ezrin are present. Incross-linked U251 samples additional bands between 140-170 kDaconsistent with merlin homodimers and heterodimers are detected.
Fig. 3. Domain structure of merlin and ezrin, and merlin and ezrinconstructs used in two-hybrid experiments. Merlin and ezrin consistof a globular amino terminus , an α-helical region and a
Acarboxy-terminal domain . A charged actin-binding regionis present in the carboxy terminus of ezrin. The fusion proteinconstructs contain a LexA-DNA binding domain in baits and an HA-epitope tag in preys.
900
Homotypic binding of merlin in vitroAn interaction between merlin molecules (m 1-595) wasdetected in the two-hybrid system by the activation of the β-galactosidasereceptor gene (Fig. 5). This self-associationcould be further confirmed by coimmunoprecipitation of theexpressed proteins from the yeast lysates (Fig. 6). If the last 10or 49 carboxy-terminal amino acids were removed from oneof the binding partners, the interaction was retained (Fig. 5).However, after removal of the residues from both proteins nobinding was seen. This is very evident in Fig. 6, in which m1-595 coimmunoprecipitates m 1-595 and m 1-585 whereas nointeraction is seen between m 1-585 and m 1-585.
Mapping of the interaction domains indicated that m 1-595interacts with the amino-terminal part of merlin. The first 339amino acids, which contain the amino-terminal globulardomain, are sufficient for binding, while amino acids 1-167 didnot mediate the interaction. The carboxy-terminal domain (m252-595), which contains the last 343 amino acids, bound toall four constructs (1-339, 1-546, 1-585, 1-595) that contain theintact amino-terminal domain (Figs 5, 6). A shorter carboxy-terminal construct, m 339-595, however, did not bind m 1-595(Fig. 6) or m 1-339 (not shown). The merlin carboxy-terminalconstruct m 339-585, with the last 10 amino acids deleted,bound the short merlin amino-terminal construct m 1-167, butnot constructs with an intact amino-terminal domain (Fig. 5).M 339-585 also bound merlin carboxy-terminal construct m252-595, whereas carboxy-terminal constructs with an intactcarboxy terminus, m 252-595 (Fig. 5) and m 339-595 (notshown) did not bind m 252-595.
Based on the β-galactosidase activity, the strongest bindingwas detected between two full-length proteins (Fig. 5B). Highβ-galactosidase values were also seen in interactions betweenm 1-595 and m 1-339, between m 1-595 and m 1-546, andbetween m 1-167 and m 339-585.
The homotypic amino-terminal to carboxy-terminalassociation of merlin was also demonstrated by an independentassay, in which affinity precipitation was performed withrecombinant proteins (Fig. 7). In these experiments, totallysates of yeast cells expressing merlin HA-fusion proteins,were allowed to bind beads containing GST-merlin constructs.GST-fusion proteins expressing merlin amino acids 1-595, 1-546 and 1-339, but not GST alone, bound to merlin 252-595present in a cell lysate. These results further support an amino-to carboxy-terminal binding of merlin.
Homotypic interactions of ezrin in vitroThe yeast two-hybrid experiments with ezrin constructsprovided results that were in accordance with previous studies(Gary and Bretscher, 1995). Unlike the results with merlin, twofull-length ezrin molecules did not interact, if the proteins werein a native form (Figs 5, 6). In the blot overlay experiments(Fig. 8), where the immobilized protein had been exposed to
M. Grönholm and others
Fig. 4.Characterization of bait and prey fusion proteins. Merlin (m)and ezrin (ez) constructs in the two-hybrid vectors were expressed inyeast cells, separated by SDS-PAGE, transferred to nitrocellulosefilters and immunoblotted. (A) Bait constructs were detected withanti-LexA antibody. (B) Prey constructs were detected with anti-HAmAb. All constructs migrate at expected sizes.
Fig. 5.Two-hybrid matingexperiment, color selection and β-galactosidase values. (A) Yeast cellsexpressing the indicated merlin (m)or ezrin (ez) constructs (left = baits,top = preys) were grown on plateslacking histidine and tryptophane,but containing leucine and X-Gal.Blue color is a quantitativemeasurement of an interaction.(B) β-Galactosidase values for two-hybrid interactions were analyzed asdescribed in Materials and Methods.The values were categorized asfollows: – = <20; + = 21-150; ++ =151-300; +++ = 301-450; ++++ =451-600. EG202 and JG4-5 areempty bait and prey vectors,respectively.
901Homo- and heterotypic interaction of merlin and ezrin
SDS, an interaction was detected as described earlier (Gary andBretscher, 1995). Ez 1-585 and a carboxy-terminal constructez 278-585 bound strongly in the two-hybrid experiment (Fig.5) and coimmunoprecipitation (Fig. 6), and ez 1-585 bound toGST-ez 477-585 in the blot overlay (Fig. 8). The amino-terminal residues ez 1-309 did not interact with ez 1-585.Apparently, the C-ERMAD of the ez 1-585 protein is maskedand cannot interact with ez 1-309 that contains the N-ERMAD.The amino-terminal residues 1-309 but not 1-170 bound thecarboxy-terminal residues 278-585 in which the C-ERMAD isexposed (Figs 5, 6). In the blot overlay experiment the carboxy-terminal biotin-labeled probe (GST-ez 477-585) also bound theα-helical domain of ezrin (GST-ez 279-531), which could bea result of the two α-helices binding to each other. It also showsa weak band with GST-ez 477-585, which does not fit themodel of head-to-tail binding. However, this band is clearly notas strong as the other ezrin interactions, and could represent abackground signal (Fig. 8).
Heterotypic interactions of merlin and ezrin in vitroThe full-length ezrin and merlin did not bind to each other inthe yeast two-hybrid system (Figs 5, 6). The result is thusanalogous with findings for two ezrin monomers. However,constructs of merlin or ezrin with amino-terminal deletions (m252-595, ez 278-585) heterodimerized with amino-terminaldomains of reciprocal proteins. Very strong interaction wasseen between ez 1-309 and m 252-595 (Figs 5, 6). M 1-339
and m 1-585 showed weak binding while m 1-546 boundstronger to ez 278-585 (Fig. 5). These results indicate thatassociation sites in ezrin and merlin must be unmasked forheterodimerization. The regulation and/or binding affinitymight slightly differ between ezrin and merlin, since full-length merlin did not bind ez 278-585 whereas full-length ezrinweakly bound m 252-595. However, full length ezrin (ez 1-585) did not bind a shorter carboxy-terminal merlin construct,m 339-585 which lacked the 10 most amino-terminal residues(Fig. 5).
The affinity precipitation results show that the GST-fusionm 1-546 and 1-339 but not m 1-595 binds ez 278-585 in a yeast
Fig. 6. Coimmunoprecipitation of baitconstructs with prey constructs. Yeastcells cotransformed with the two-hybrid bait and prey vectors containingthe indicated merlin (m) and ezrin (ez)residues were grown in selectivemedium and lysed. The prey fusionproteins were immunoprecipitatedwith the anti-HA mAb 12CA5. Boundmaterial was separated by SDS-PAGEand transferred to nitrocellulose filters.Coprecipitating bait fusion proteinswere immunodetected with anti-LexAantibody.
Fig. 7. Affinity precipitation of carboxy-terminal constructs of merlinand ezrin with GST-merlin fusion proteins. Lysates from yeastexpressing carboxy-terminal regions of merlin (m) or ezrin (ez) fusedwith an HA-tag were incubated with glutathione agarose-coupledGST-merlin 1-595, 1-546, 1-339 and GST. The agarose-boundmaterial was eluted by boiling in Laemmli sample buffer, separatedby SDS-PAGE and transferred to nitrocellulose filters. Proteins weredetected by anti-HA immunoblotting using 12CA5 mAb.
Fig. 8. Binding of biotin-labeled ezrin probes to recombinant merlinand ezrin proteins. Recombinant ezrin and GST-merlin and -ezrinfusion proteins were separated by SDS-PAGE and transferred tonitrocellulose filters. The filters were incubated with biotin-ezrinprobes ez 1-585 and GST-ez-477-585. Bound protein was detectedwith peroxidase-conjugated avidin and enhancedchemiluminescence. (A) Coomassie Blue staining of SDS-PAGEindicating the location of immobilized proteins. (B) Blot overlayprobed with biotinylated ez 1-585. (C) Blot overlay probed withbiotinylated GST-ez 477-585.
902
total lysate (Fig. 7). The results are in line with the two-hybridexperiments. In the blot overlay experiment, GST-m 1-339 aswell as GST-m 1-546 bound to the biotin-labelled GST-ez 477-585 probe (Fig. 8). The result extends the two-hybridexperiments by demonstrating that the last 109 carboxy-terminal residues of ezrin are sufficient for heterodimerizationwith merlin.
DISCUSSION
The understanding of the biological functions and tumorsuppressor mechanism of merlin has been hampered byinsufficient knowledge of its molecular interactions. So far,only a few direct interaction partners to merlin have beenreported. We now show that merlin interacts with an ERMfamily member, ezrin, and that merlin shows homotypicbinding. The evidence for an interaction between merlin andezrin in vivo includes subcellular colocalization, concomitantredistribution, and most importantly, coimmunoprecipitation ofendogenous or transfected merlin and endogenous ezrin fromcell lysates and the presence of a band consistent with a merlin-ezrin heterodimer after chemical cross-linking. Aheterodimerization between the two proteins could, in addition,be detected by several in vitro techniques. While merlinappears to be involved in regulation of ERM-dependent events,it may also have separate functions from ERM proteins anddifferent means for functional regulation. This is suggested byour evidence that merlin, in contrast to ezrin, does not requireexposure of the association domains for homotypic binding.
The homotypic binding of merlin and heterotypic bindingbetween merlin and ezrin, in analogy with ERM proteins,occured via amino-terminal interaction with the carboxyterminus. This is indicated by the fact that the carboxy-terminaldeletions m 1-339 and m 1-546 did not show binding unlessthe partner contained an intact carboxy terminus. In ezrin, N-ERMAD contains residues 1-296 and further deletion of a fewcarboxy-terminal amino acids results in loss of activity (Garyand Bretscher, 1995) (Fig. 9). Homotypic binding of merlinoccured via residues 1-339, while 1-167 was inactive. Thus,the amino-terminal association domain of merlin seems to bevery similar to ezrin.
Based on the fact that the residues of the carboxy-terminalassociation domain of ezrin are poorly conserved in merlin, ithas been suggested that the carboxy terminus of merlin wouldnot allow self-association (Gary and Bretscher, 1995). Ourresults indicate that this is not the case. In fact, m 252-595 notonly bound to the amino terminus of merlin but also to ezrin.The carboxy-terminal binding region of merlin could not bemapped as precisely as the C-ERMAD of ezrin which iscontained in residues 479-585. The C-ERMAD of ezrinmediates not only intramolecular binding to N-ERMAD ofezrin, but also to moesin (Gary and Bretscher, 1995), and asshown here, binding to the amino-terminal association domainof merlin. This result may reflect the conservation of theamino-terminal domain between ERM proteins and merlin,which apparently includes the residues critical for C-ERMADbinding. While the construct m 252-595 binds to the amino-terminal domain of merlin, m 339-595 did not mediate theinteraction. This construct, m 339-595, disrupts the α-helicaldomain which could abolish a conformation needed for the
interaction. However, overlapping carboxy-terminal constructsof ezrin, GST-ezrin fusion proteins 325-585 and 368-585,interact with the full length ezrin molecule (Gary andBretscher, 1995). Alternatively, the result indicates that merlinneeds a longer carboxy-terminal domain for binding than ezrin.A similar feature for the binding domain of merlin and ezrinis the absolute requirement of the most carboxy-terminalresidues. In ezrin, deletion of the last two amino acids resultedin loss of activity (Gary and Bretscher, 1995), in merlin the last10 residues were needed for interaction with the amino-terminal association domain. However, these ten residues arenot sufficient for the interaction, which apparently involves aninterplay between two or more different regions. If the last 10amino acids in the merlin caboxy-terminal construct wereremoved the binding characteristics became strikinglydifferent. This construct, m 339-585, did not bind the amino-terminal association domain in merlin, m 1-339, but did binda shorter amino-terminal construct, m 1-167, and surprisinglythe intact carboxy-terminal construct, m 252-595. This couldindicate that the properties of the interactions are morecomplex than portrayed here, or that the constructs, m 339-585and m 1-167, that disrupt the α-helix and the amino-terminalglobular domain, respectively, do not present the foldingpatterns and charge of the intact association domains orintact molecule. This could allow the deviant binding to takeplace.
Our study indicates that the regulation of homotypic head totail association of merlin is different from ezrin. Thisdifference may have functional consequences. Based on thecurrent model, an intramolecular association between N-ERMAD and C-ERMAD retains ERM proteins in a dormantstate, in which binding sites for cell membrane componentsand actin are masked (Gary and Bretscher, 1995; Martin et al.,1997; Tsukita et al., 1997). In line with this idea, full lengthezrin, when overexpressed in SF9 cells, does not inducemorphogenic cell surface alterations (Martin et al., 1995,1997). Unmasking of these binding sites by conformationalactivation (Bretscher, 1989) or by amino- or carboxy-terminaldeletions (Martin et al., 1995, 1997), results in prominent cellsurface alterations. In contrast to ezrin, expression of merlin inmammalian cells (Sainio et al., 1997) or SF9 cells (F. Zhao,
M. Grönholm and others
167 339
286 2961
1EZRIN
MERLIN
AMINO-TERMINAL ASSOCIATION DOMAIN
EZRIN
MERLIN339
479 530 583 585
252 585 595
CARBOXY-TERMINAL ASSOCIATION DOMAIN
Fig. 9. A schematic diagram of amino- and carboxy-terminalassociation domains in ezrin and merlin. On the top are shown thestructural domain of ezrin and merlin described in Fig. 3. Boxesdepict the association domains. represent regions that are not sufficient for an interaction. Deletion of carboxy-terminal aminoacids represented by abolish the interaction. Data on the association domains of ezrin is taken from Gary and Bretscher(1995).
903Homo- and heterotypic interaction of merlin and ezrin
M. Sainio and O. Carpén, unpublished) results in cell surfacealterations reminiscent of those caused by ‘active’ ezrinconstructs. It is possible that this difference reflects thedifferential regulation of homotypic association betweenmerlin and ezrin. In the yeast two-hybrid experiments, in whichthe proteins should retain a native conformation, full-lengthmerlin (1-595) can bind another full-length molecule, whilefull-length ezrin (1-585) cannot bind another native ezrin ormerlin molecule. A possible explanation for this is that theintramolecular association in ezrin is so stable thatdimerization between expressed proteins does not occur. Inmerlin, the intramolecular association is not of high affinityand dimerization can occur. Dimerization or oligomerizationhas previously been shown to correlate with functional activityof ERM proteins (Berryman et al., 1995).
Full-length merlin binds amino- and carboxy-terminalconstructs of merlin, but not amino- or carboxy-terminalconstructs of ezrin. For the heterotypic interaction betweenmerlin and ezrin, merlin has to be expressed as a truncatedprotein. Our interpretation of the results is the following. In thenative merlin, the intramolecular binding sites, which allowhomodimerization are exposed, whereas the binding sites forezrin are masked. Heterodimerization between merlin andezrin occurs only following conformational alternations in bothproteins. Thus, also merlin apparently undergoesconformational activation, which is a prerequisite forheterodimerization but not for homodimerization. Thedifferential regulation of homodimerization versusheterodimerization of merlin suggests that the rank of order forbinding partners depends on cellular signals that affectactivation of merlin and ezrin. The in vivocoimmunoprecipitation experiments strongly suggest that incells at least a fraction of merlin and ezrin are in an activeconformation and can associate heterotypically.
Merlin’s tumor suppressor mechanism involves an ability tosuppress cell proliferation with an unknown mechanism.Studies of the expression of merlin and ERM proteins inschwannomas, the tumors associated with NF2 gene defects,showed loss of merlin in all tumors, whereas the expression ofezrin, radixin and moesin is retained (Stemmer-Rachamimov etal., 1997). The ERM proteins are associated with cell growth,as they are downstream targets of the Rho family of small Gproteins, that are involved in regulation of cell growth (Mackayet al., 1997). Moreover, abnormal regulation or high expressionof ezrin has been associated with cell transformation andincreased proliferation (Jooss and Muller, 1995; Kaul et al.,1996; Lamb et al., 1997). In this regard, the possible oppositeeffects of merlin and ezrin on cell proliferation may be linkedto differential regulation of their association ability. Sherman etal. (1997) suggested that tumor growth inhibition by merlindepends on an interdomain association that occurs either in cisor in trans between the amino- and the carboxy-terminaldomains. An alternative explanation is that the intact amino-and carboxy-terminal domains of merlin are required forheterotypic association with ezrin and other ERM proteins.Analysis of functional consequences of heterotypic bindingbetween merlin and ezrin could provide novel information ofmerlin’s tumor suppressor function.
We thank G. Rouleau for anti-schwannomin and E. Zwarthoff for1398NF2 rabbit antiserum, E. Golemis for anti-LexA antibody, J.
Gusella for merlin cDNA, R. Brent for plasmids, P. Mangeat for theezrin expression construct, P. Ljungdahl for BOY1 yeast and H.Pihlaja and T. Halmesvaara for skillful technical assistance. This workwas supported by the Academy of Finland, the Finnish CancerFoundation and the Sigrid Jusélius Foundation, Helsinki.
REFERENCES
Andreoli, C., Martin, M., Le Borgne, R., Reggio, H. and Mangeat, P.(1994). Ezrin has properties to self-associate at the plasma membrane. J.Cell Sci. 107, 2509-2521.
Berryman, M., Gary, R. and Bretscher, A. (1995). Ezrin oligomers are majorcytoskeletal components of placental microvilli: a proposal for theirinvolvement in cortical morphogenesis. J. Cell Biol. 131, 1231-1242.
Bianchi, A. B., Hara, T., Ramesh, V., Gao, J., Klein-Szanto, A. J., Morin,F., Menon, A. G., Trofatter, J. A., Gusella, J. F., Seizinger, B. R., et al.(1994). Mutations in transcript isoforms of the neurofibromatosis 2 gene inmultiple human tumour types. Nat. Genet. 6, 185-192.
Böhling, T., Turunen, O., Jääskeläinen, J., Carpén, O., Sainio, M.,Wahlström, T., Vaheri, A. and Haltia, M. (1996). Ezrin expression instromal cells of capillary hemangioblastoma. An immunohistochemicalsurvey of brain tumors. Am. J. Pathol. 148, 367-373.
Bretscher, A. (1989). Rapid phosphorylation and reorganization of ezrin andspectrin accompany morphological changes induced in A-431 cells byepidermal growth factor. J. Cell Biol. 108, 921-930.
den Bakker, M. A., Riegman, P. H., Hekman, R. A., Boersma, W., Janssen,P. J., van der Kwast, T. H. and Zwarthoff, E. C. (1995). The product ofthe NF2 tumour suppressor gene localizes near the plasma membrane andis highly expressed in muscle cells. Oncogene 10, 757-763.
Funayama, N., Nagafuchi, A., Sato, N., Tsukita, Sa. and Tsukita, Sh.(1991). Radixin is a novel member of the band 4. 1 family. J. Cell Biol. 115,1039-1048.
Gary, R. and Bretscher, A. (1993). Heterotypic and homotypic associationsbetween ezrin and moesin, two putative membrane-cytoskeletal linkingproteins. Proc. Nat. Acad. Sci. USA 90, 10846-10850.
Gary, R. and Bretscher, A. (1995). Ezrin self-association involves binding ofan N-terminal domain to a normally masked C-terminal domain thatincludes the F-actin binding site. Mol. Biol. Cell 6, 1061-1075.
Gonzalez-Agosti, C., Xu, L., Pinney, D., Beauchamp, R., Hobbs, W.,Gusella, J. and Ramesh, V. (1996). The merlin tumor suppressor localizespreferentially in membrane ruffles. Oncogene 13, 1239-1247.
Gould, K. L., Bretscher, A., Esch, F. S. and Hunter, T. (1989). cDNAcloning and sequencing of the protein-tyrosine kinase substrate, ezrin,reveals homology to band 4. 1. EMBO J. 8, 4133-4142.
Gyuris, J., Golemis, E., Chertkov, H. and Brent, R. (1993). Cdi1, a humanG1 and S phase protein phosphatase that associates with Cdk2. Cell 75, 791-803.
Harshman, K. D., Scott Moye-Rowley, W. and Parker, C. S. (1988).Transcriptional activation by the SV 40 AP-1 recognition element in yeastis mediated by a factor similar to AP-1 that is distinct from GCN4. Cell 53,321-330.
Heiska, L., Alfthan, K., Grönholm, M., Vilja, P., Vaheri, A. and Carpén,O. (1998). Association of Ezrin with Intercellular Adhesion Molecule-1 and-2 (ICAM-1 and ICAM-2). Regulation by phosphatidylinositol 4,5-bisphosphate. J. Biol. Chem. 273, 21893-21900.
Helander, T. S., Carpén, O., Turunen, O., Kovanen, P. E., Vaheri, A. andTimonen, T. (1996). ICAM-2 redistributed by ezrin as a target for killercells. Nature 382, 265-268.
Henry, M. D., Gonzalez-Agosti, C. and Solomon, F. (1995). Moleculardissection of radixin: distinct and interdependent functions of the amino-and carboxy-terminal domains. J. Cell Biol. 129, 1007-1022.
Hirao, M., Sato, N., Kondo, T., Yonemura, S., Monden, M., Sasaki, T.,Takai, Y., Tsukita, Sh. and Tsukita, Sa. (1996). Regulation mechanism ofERM (ezrin/radixin/moesin) protein/plasma membrane association: possibleinvolvement of phosphatidylinositol turnover and Rho-dependent signalingpathway. J. Cell Biol. 135, 37-51.
Huang, L. Q., Ichimaru, E., Pestonjamasp, K., Cui, X. M., Nakamura,H., Lo, G. Y. H., Lin, F. I. K., Luna, E. J. and Furthmayr, H. (1998).Merlin differs from moesin in bindint to F-actin and in its intra- andintermolecular interactions. Biochem. Biophys. Res. Commun. 248, 548-553.
Jooss, K. U. and Muller, R. (1995). Deregulation of genes encoding
904
microfilament-associated proteins during Fos-induced morphologicaltransformation. Oncogene 10, 603-608.
Kaul, S. C., Mitsui, Y., Komatsu, Y., Reddel, R. R. and Wadhwa, R. (1996).A highly expressed 81 kDa protein in immortalized mouse fibroblast: itsproliferative function and identity with ezrin. Oncogene 13, 1231-1237.
Lamb, R. F., Ozanne, B. W., Roy, C., McGarry, L., Stipp, C., Mangeat, P.and Jay, D. G. (1997). Essential functions of ezrin in maintenance of cellshape and lamellipodial extension in normal and transformed fibroblasts.Curr. Biol. 7, 682-688.
Lankes, W. T. and Furthmayr, H. (1991). Moesin: a member of the protein4. 1-talin-ezrin family of proteins. Proc. Nat. Acad. Sci. USA 88, 8297-8301.
Louis, D. N., Ramesh, V. and Gusella, J. F. (1995). Neuropathology andmolecular genetics of neurofibromatosis 2 and related tumors. Brain Pathol.5, 163-172.
Lutchman, M. and Rouleau, G. (1995). The neurofibromatosis type 2 geneproduct, schwannomin, suppresses growth of NIH 3T3 cells. Cancer Res.55, 2270-2274.
Mackay, D. J. G., Esch, F., Furthmayr, H. and Hall, A. (1997). Rho- andRac-dependent assembly of focal adhesion complexes and actin filamentsin permeabilized fibroblasts: an essential role for Ezrin/Radixin/Moesinproteins. J. Cell Biol. 138, 927-938.
Magendantz, M., Henry, M. D., Lander, A. and Solomon, F. (1995).Interdomain interactions of radixin in vitro. J. Biol. Chem. 270, 25324-25327.
Martin, M., Andreoli, C., Sahuquet, A., Montcourrier, P., Algrain, M. andMangeat, P. (1995). Ezrin NH2-terminal domain inhibits the cell extensionactivity of the COOH-terminal domain. J. Cell Biol. 128, 1081-1093.
Martin, M., Roy, C., Montcourrier, P., Sahuquet, A. and Mangeat, P.(1997). Three determinants in ezrin are responsible for cell extensionactivity. Mol. Biol. Cell 8, 1543-1557.
Murthy, A., Gonzalez-Agosti, C., Cordero, E., Pinney, D., Candia, C.,Solomon, F., Gusella, J. and Ramesh, V. (1998). NHE-RF, a regulatorycofactor for Na+-H+ exchange, is a common interactor for merlin and ERM(MERM) proteins. J. Biol. Chem. 273, 1273-1276.
Pestonjamasp, K., Amieva, M. R., Strassel, C. P., Nauseef, W. M.,Furthmayr, H. and Luna, E. (1995). Moesin, ezrin, and p205 are actin-binding proteins associated with neutrophil plasma membranes. Mol. Biol.Cell 6, 247-259.
Raghuraman, M., Brewer, B. and Fangman, W. (1994). Activation of a yeastreplication origin near a double-stranded DNA break. Genes Dev. 8, 554-562.
Reczek, D., Berryman, M. and Bretscher, A. (1997). Identification ofEBP50: A PDZ-containing phosphoprotein that associates with members ofthe ezrin-radixin-moesin family. J. Cell Biol. 139, 169-179.
Rouleau, G. A., Merel, P., Lutchman, M., Sanson, M., Zucman, J.,Marineau, C., Hoang-Xuan, K., Demczuk, S., Desmaze, C., Plougastel,B., et al. (1993). Alteration in a new gene encoding a putative membrane-organizing protein causes neuro-fibromatosis type 2. Nature 363, 515-521.
Roy, C., Martin, M. and Mangeat, P. (1997). A dual involvement of theamino-terminal domain of ezrin in F-and G-actin binding. J. Biol. Chem.272, 20088-20095.
Sainio, M., Zhao, F., Heiska, L., Turunen, O., den Bakker, M., Zwarthoff,E., Lutchman, M., Rouleau, G. A., Jääskeläinen, J., Vaheri, A., et al.(1997). Neurofibromatosis 2 tumor suppressor protein colocalizes with ezrinand CD44 and associates with actin-containing cytoskeleton. J. Cell Sci.110, 2249-2260.
Samson, M.-L., Jackson-Grusby, L. and Brent, R. (1989). Gene activation
and DNA binding by Drosophila Ubx and abd-A proteins. Cell 57, 1045-1052.
Scoles, D. R., Huynh, D. P., Morcos, P. A., Coulsell, E. R., Robinson, N. G.G., Tamanoi, F. and Pulst, S. M. (1998). Neurofibromatosis 2 tumoursuppressor schwannomin interacts with beta-II-spectrin. Nat. Genet. 18,354-359.
Shaw, R. J., Mcclatchey, A. I. and Jacks, T. (1998). Regulation of theneurofibromatosis type 2 tumor suppressor protein, merlin, by adhesion andgrowth arrest stimuli. J. Biol. Chem. 273, 7757-7764.
Sherman, L., Xu, H. M., Geist, R. T., Saporito-Irwin, S., Howells, N.,Ponta, H., Herrlich, P. and Gutmann, D. H. (1997). Interdomain bindingmediates tumor growth suppression by the NF2 gene product. Oncogene 15,2505-2509.
Stemmer-Rachamimov, A. O., Xu, L., Gonzalez-Agosti, C., Burwick, J. A.,Pinney, D., Beauchamp, R., Jacoby, L. B., Gusella, J. F., Ramesh, V. andLouis, D. N. (1997). Universal absence of merlin, but not other ERM familymembers, in schwannomas. Am. J. Pathol. 151, 1649-1654.
Takeuchi, K., Sato, N., Kasahara, H., Funayama, N., Nagafuchi, A.,Yonemura, S., Tsukita, Sa. and Tsukita, Sh. (1994). Perturbation of celladhesion and microvilli formation by antisense oligonucleotides to ERMfamily members. J. Cell Biol. 125, 1371-1384.
Trofatter, J. A., MacCollin, M. M., Rutter, J. L., Murrell, J. R., Duyao, M.P., Parry, D. M., Eldridge, R., Kley, N., Menon, A. G., Pulaski, K., et al.(1993). A novel moesin-, ezrin-, radixin-like gene is a candidate for theneurofibromatosis 2 tumor suppressor. Cell 72, 791-800.
Tsukita, Sa., Yonemura, S. and Tsukita, Sh. (1997). ERM proteins: head-to-tail regulation of actin-plasma membrane interaction. Trends Biochem.Sci. 22, 53-58.
Turunen, O., Winqvist, R., Pakkanen, R., Grzeschik, K. H., Wahlström,T. and Vaheri, A. (1989). Cytovillin, a microvillar Mr 75, 000 protein.cDNA sequence, prokaryotic expression and chromosomal localization. J.Biol. Chem. 264, 16727-16732.
Turunen, O., Wahlström, T. and Vaheri, A. (1994). Ezrin has a COOH-terminal actin-binding site that is conserved in the ezrin protein family. J.Cell Biol. 126, 1445-1453.
Turunen, O., Sainio, M., Jääskeläinen, J., Carpén, O. and Vaheri, A.(1998). Structure-function relationships in the ezrin family and the effect oftumor-associated point mutations in neurofibromatosis 2 protein. Biochim.Biophys. Acta 1387, 1-16.
Vaheri, A., Carpén, O., Heiska, L., Helander, T. S., Jääskeläinen, J.,Majander-Nordenswan, P., Sainio, M., Timonen, T. and Turunen, O.(1997). The ezrin protein family: membrane-cytoskeleton interactions anddisease associations. [review] [69 refs]. Curr. Opin. Cell Biol. 9, 659-666.
Westermark, B. (1973). Determinants for the establishment of permanenttissue culture lines from human gliomas. Int. J. Cancer 12, 438-451.
Xu, H. M. and Gutmann, D. H. (1998). Merlin differentially associates withthe microtubule and actin cytoskeleton. J. Neurosci. Res. 51, 403-415.
Yonemura, S., Hirao, M., Doi, Y., Takahashi, N., Kondo, T., Tsukita, Sa.and Tsukita, Sh. (1998). Ezrin/Radixin/Moesin (ERM) proteins bind to apositively charged amino acid cluster in the juxta-membrane cytoplasmicdomain of CD44, CD43, and ICAM-2. J. Cell Biol. 140, 885-895.
Note added in proofWhile this manuscript was under review, another publicationdemonstrating a homodimeric association of merlin appeared(Huang et al., 1998).
