Src Phosphorylates Ezrin at Tyrosine 477 and Induces a Phosphospecific Association between Ezrin and a Kelch-Repeat Protein Family Member*

Ezrin, a linker between plasma membrane and actin cytoskeleton possesses morphogenic properties and can promote dissemination of tumor cells. Ezrin is phosphorylated on tyrosine, but a detailed picture of the signaling pathways involved in this modification is lacking. The transforming tyrosine kinase Src has various cytoskeletal substrates and is involved in regulation of cellular adhesion. We studied the role of Src in tyrosine phosphorylation of ezrin in adherent cells. We show that ezrin is phosphorylated in human embryonic kidney 293 cells in a Src family-dependent way. In SYF cells lacking Src, Yes, and Fyn, ezrin was not tyrosine-phosphorylated but reintroduction of wild-type Src followed by Src activation or introduction of active Src restored phosphorylation. Mapping of the Src-catalyzed tyrosine in vitro and in vivo by site-directed mutagenesis demonstrated Tyr477 as the primary target residue. We generated a pTyr477-phosphospecific antibody, which confirmed that Tyr477 becomes phosphorylated in cells in a Src-dependent manner. Tyr477 phosphorylation did not affect ezrin head-to-tail association or phosphorylation of ezrin on threonine 566, indicating that the function of Tyr477 phosphorylation is not related to the intramolecular regulation of ezrin. A modified yeast two-hybrid screen in which ezrin bait was phosphorylated by Src identified a novel interaction with a kelch-repeat protein family member, KBTBD2 (Kelch-repeat and BTB/POZ domain containing 2). The Src dependence of the interaction was further verified by affinity precipitation assays. Identification of a functional interplay with Src opens novel avenues for further characterization of the biological activities of ezrin.

The Src family of transforming non-receptor tyrosine kinases is involved in cell adhesion, invasion, migration, cell cycle progression, and survival and is coupled to various receptor signaling pathways (reviewed in Refs. [1][2][3]. Src also has a role in control of the actin cytoskeleton. Activation of Src by a complex mechanism leads to its translocation to plasma membrane and to association with detergent-insoluble protein fractions (4). Src has various cytoskeletal substrates, and its activation induces rapid formation of lamellipodia or membrane ruffles.
One of the linker proteins between cell surface receptors, especially adhesion molecules, and the actin cytoskeleton is ezrin, a member of the ERM 1 (ezrin-radixin-moesin) protein family (reviewed in Refs. [5][6][7]. Ezrin is typically localized to specialized membrane areas such as microvilli, membrane ruffles, or the uropods of polarized lymphocytes, and it possesses morphogenic functions in a variety of cell types. In addition to membrane proteins and F-actin, ezrin interacts with cellular signaling proteins of the phosphatidylinositol 3-kinase, protein kinase A and Rho signaling pathways (8) and is thereby involved in the mechanisms of targeted signaling in cells.
Several mechanisms regulate the functional activity and molecular interactions of ERM proteins. An intramolecular association between the N terminus (N-ERMAD, N-terminal association domain) and C terminus (C-ERMAD, C-terminal association domain) keeps the proteins inactive and masks their interaction sites for other molecules (9 -11). A key regulatory switch appears to be phosphorylation of a conserved threonine in the C-terminal domain. Phosphorylation of this residue results in dissociation of the intramolecular interaction and formation of monomers with morphogenic activity (12).
The consequences of tyrosine phosphorylation of ezrin are less well known. Ezrin is a major protein phosphorylated on tyrosine after epidermal growth factor (EGF) stimulation in A431 cells (13,14), and tyrosine phosphorylation of ezrin is also induced by platelet-derived growth factor and hepatocyte growth factor (15,16). The major phosphorylated sites by the EGF and hepatocyte growth factor pathways have been mapped to Tyr 145 in the N-terminal domain and Tyr 353 in the ␣-helical region (15,17). Phosphorylation of Tyr 353 , which is not conserved in other ERM proteins, is required for the binding of the regulatory p85 subunit of phosphatidylinositol 3-kinase to ezrin and for the activation of the phosphatidylinositol 3-kinase/protein kinase Akt signaling pathway in LLC-PK1 cells (18). A defect in this signaling pathway due to a point mutation of Tyr 353 to phenylalanine sensitizes cells to apoptosis. In addition, mutagenesis of both Tyr 145 and Tyr 353 in ezrin has been shown to disturb a morphogenic response to hepatocyte growth factor (15). Early findings showed that ezrin tyrosine phosphorylation was increased in v-Src-induced cells (19). We have shown that the Src family kinase (SFK) Lck catalyzes phosphorylation of Tyr 145 in Jurkat T-cells (20). Cross-linking of intercellular adhesion molecule-1 (ICAM-1) activates Src kinases and leads to tyrosine phosphorylation of ezrin (21). Also, cross-linking of ICAM-2 induces tyrosine phosphorylation of ezrin and activation of the PKB/Akt pathway (22), which is blocked by herbimycin A, a potent inhibitor of Src and pp62 yes . Recently, it was reported that Src and ezrin act cooperatively in regulation of cell-cell contacts and scattering of mammary carcinoma cells (23). However, the relation of Src and ezrin is not known at molecular level and the Src target site in ezrin has not been characterized.
We have studied the role of Src kinase in tyrosine phosphorylation of ezrin in adherent cells and characterized a novel Src-dependent phosphorylation site, Tyr 477 . We show that, on one hand, phosphorylation at Tyr 477 does not affect ezrin headto-tail association and that, on the other hand, it is independent of ezrin threonine phosphorylation, which has been linked to conformational opening of the molecule. Moreover, we show that phosphorylation of ezrin by Src induces an association with a member of the kelch-repeat superfamily protein, KBTBD2 (Kelch-repeat and BTB/POZ domain containing 2).

MATERIALS AND METHODS
Cells and Antibodies-Human embryonic kidney 293 (HEK293, ATCC, Manassas, VA) cells were maintained in RPMI 1640 medium including 10% fetal calf serum and antibiotics. A431 epidermoid carcinoma cells (ECACC, Salisbury, United Kingdom), SYF mouse embryonic fibroblasts from the triple Src knock-out mouse lacking Src, Yes, and Fyn SFKs (24) and the add-back versions of the SYF cells, which have been introduced either with the wild-type c-Src gene or the constitutively active Y527F mutant Src (25) kindly provided by Dr. J. A. Cooper (Fred Hutchinson Cancer Research Center, Seattle, WA) were maintained in Dulbecco's modified Eagle's medium including 10% fetal calf serum and antibiotics.
Phosphotyrosine was detected with the monoclonal anti-phosphotyrosine monoclonal antibody clone PT-66 (Sigma). The polyclonal rabbit antibody Ez9 against ezrin has been described previously (26). The VSVG-tag monoclonal antibody was from Roche Diagnostics (clone P5D4, Mannheim, Germany). Mouse IgG1 (Dako Cytomation, Glostrup, Denmark) or rabbit preimmune serum were used as controls. Rabbitphosphospecific antibodies against Src(pY418) and against ezrin-pThr 566 (Tyr 558 on moesin) were from BioSource International (Camarillo, CA) and Santa Cruz Biotechnology (Santa Cruz, CA), respectively. Rabbit polyclonal antibody recognizing human ezrin phosphorylated on Tyr 477 was produced by BioSource International by immunization of a 6-mer synthetic phosphopeptide PV(pY)EPV and subsequent purification of the antiserum using both negative and positive affinity purification methods.
Plasmids and Transfections-The recombinant GST-ezrin constructs encoding N-terminal (amino acids 1-309) and short C-terminal (amino acids 477-585) ezrin fusion proteins have been described previously (27). The ␣-helical (amino acids 296 -468) and C-terminal (amino acids 309 -585) sequences were amplified by PCR using human ezrin cDNA as template and cloned into pGEX4T-1 vector (Amersham Biosciences). Mutations Y145F, Y353F, Y477F, Y482F, Y498F, and Y564F were introduced by QuikChange mutagenesis kit (Stratagene, La Jolla, CA) and subcloned into pCDNA3 (Invitrogen) eukaryotic expression vector for cell transfections. The double mutant Y145F/Y477F was constructed from the two separate mutant cDNAs by conventional recombinant DNA techniques. GST fusion construct encoding full-length human KBTBD2 was cloned by PCR using the library clone IRAKp961E2391Q2 (Deutsches Ressourcenzentrum fü r Genomforschung GmbH, Berlin, Germany) as a template. All of the constructs were verified by DNA sequencing. The C-terminal GST-merlin construct (amino acids 492-595) has been described previously (26). The wild-type ezrin construct was a kind gift of Dr. M. Arpin (Institut Curie, Paris, France). The cDNAs of the constitutively activated Src containing Y529F mutation and the dominant negative kinase-dead Src K296R/Y528F in pUSEamp eukaryotic expression vector were from Upstate Biotechnology. Cells were transfected using FuGENE 6 transfection reagent (Hoffmann-La Roche Ltd., Basel, Switzerland) and used 48 h after transfection. The amino acid numbering of ezrin is according to Krieg and Hunter (17).
Immunoprecipitation of Ezrin-The cells were lysed in cold radioimmune precipitation assay buffer (50 mM Tris-HCl, 150 mM NaCl, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS, 1 mM orthovanadate, and protease inhibitor mixture, pH 7.4) and centrifuged at 16000 ϫ g for 10 min. Endogenous ezrin was immunoprecipitated from the supernatants with Ez9 antibody, and transfected proteins were precipitated with anti-VSVG followed by protein G-Sepharose beads (Amersham Biosciences). The beads were washed four times in radioimmune precipitation assay buffer and finally suspended to Laemmli buffer. The immunoprecipitate was resolved on SDS-PAGE, blotted on nitrocellulose, and analyzed by immunoblotting with appropriate antibodies using ECL detection (Amersham Biosciences). When needed, the cells were treated before lysis with 200 M pervanadate for 15 min at 37°C or with 200 ng/ml EGF (Calbiochem, La Jolla, CA) for 5 min at 37°C. A431 cells were serum-starved for 16 h before treatments. In some experiments, the cells were preincubated with 10 M PP2 (Calbiochem) for 10 min before and during pervanadate treatment.
In Vitro Phosphorylation of Ezrin-GST fusion proteins were expressed in Escherichia coli and purified according to the manufacturer's instructions (Amersham Biosciences). The GST portion of some of the fusion proteins was cleaved off by thrombin (Sigma). 2 g of polypeptides were used for each reaction. The phosphorylation reaction was carried out in Src buffer (50 mM Tris-HCl, 10 mM MgCl 2 , 3 mM MnCl 2 , pH 7.4) including 10 Ci of [␥-33 P]ATP (PerkinElmer Life Sciences) and 1 unit of recombinant human Src (Upstate Biotechnology) for 15 min at 30°C. The reaction was stopped by boiling in reducing Laemmli buffer. The samples were resolved on SDS-PAGE, the gel was fixed and stained by Coomassie Blue, and dried and subjected to autoradiography.
Affinity Pull-down Experiments-2 g of the C-terminal GST-ezrin fusion protein (amino acids 309 -585) were phosphorylated in 20 l of Src buffer in the presence of 200 M ATP (Sigma). Two samples were incubated with 1 unit of recombinant human Src, whereas one sample was incubated with no added enzyme for 30 min at 30°C. In a scintillation analysis involving a trace amount of [␥-32 P]ATP, the phosphorylation level of Src-phosphorylated samples was ϳ12% total protein (data not shown). After reaction, 200 l of 50 mM Tris-HCl, 10 mM EDTA was added to the mixtures together with glutathione-Sepharose beads. The mixtures were incubated on a shaker for 1 h followed by extensive washes of the beads, first in 50 mM Tris-HCl, 10 mM EDTA and then three times in the calf-intestinal phosphatase (CIP) buffer (50 mM Tris-HCl, pH 7.9, 100 mM NaCl, 10 mM MgCl 2 , 1 mM dithiothreitol). In all of the washing steps during the experiments, 1 mM orthovanadate was included with the exception of the washes of one of the phosphorylated samples where orthovanadate was omitted in steps preceding the CIP treatment. The CIP reaction was preformed in 30 l of the CIP buffer with 5 units of CIP (Finnzymes Oy, Espoo, Finland) for 30 min at 37°C. Other samples were incubated in the buffer with no added enzyme. Following reaction, the beads were extensively washed, once in the CIP buffer and three times in binding buffer (50 mM Tris-HCl, 150 mM NaCl, 0.5% Nonidet P-40).
Transfected cells were lysed in binding buffer, and the lysates were added on beads and incubated on a shaker for 2 h at 25°C. Beads were washed in binding buffer four times and boiled in 30 l of Laemmli buffer. 10 l of each sample was analyzed by SDS-PAGE and immunoblotting with appropriate antibodies. GST-KBTBD2 was bound to glutathione-Sepharose beads, incubated with lysates of cells transfected with wild-type full-length or C-terminal ezrin (amino acids 306 -585) treated or untreated with pervanadate, washed, and analyzed as described above.
Yeast Two-hybrid Screening and ␤-Galactosidase Filter Assay-The C-terminal ezrin cDNA encoding for amino acids 309 -585 was introduced by PCR and conventional cloning to a modified yeast two-hybrid bait plasmid pBTM116Src, which also encodes for active Src from a separate transcript (28). The bait was used in screening of a human fetal brain library in pACT2 (Clontech). To verify the Src dependence of the interaction, each prey clone was tested in a mating assay (29) against C-terminal ezrin bait in pBTM116Src vector or, as a control, in BTM116 without Src. The mated yeast cells expressing the bait and the prey plasmids were transferred to nitrocellulose filters and quick-frozen in liquid nitrogen, and the filter was incubated for 30 min at 37°C on a wet Whatman filter moistened with Z-buffer (60 mM Na 2 HPO 4 , 40 mM NaH 2 PO 4 , 10 mM KCl, 1 mM Mg 2 SO 4 ) including 0.7 mg/ml X-gal (5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside).

Src Catalyzes Tyrosine Phosphorylation of Ezrin in Vivo-
We previously demonstrated that ezrin is dynamically phosphorylated on tyrosine in T-lymphocytes by the Src family kinase Lck (20). To find out whether ezrin is a substrate of SFK also in adherent cells, we investigated the phosphorylation status of ezrin in HEK293 cells in which SFK activity was stimulated by pervanadate. Following stimulation, endogenous ezrin was immunoprecipitated and blotted with a phosphotyrosine antibody. To specify the role of SFK, we treated cells with the Src family-selective inhibitor PP2 (30) before and during pervanadate stimulation. Immunoblotting of the whole cell lysates with an antibody recognizing active Src phosphorylated on Tyr 418 confirmed that pervanadate treatment activated Src. As shown in Fig. 1, ezrin was tyrosine phosphorylated following pervanadate treatment and tyrosine phosphorylation was inhibited if cells were preincubated with PP2.
To provide further evidence for the involvement of Src kinase in ezrin phosphorylation, we used SYF embryonic fibroblast cells from the triple Src knock-out mouse lacking Src, Yes, and Fyn SFKs (24) and the add-back versions of SYF cells reconstituted either with the wild-type c-Src gene or the constitutively activated Y527F mutant Src (25). When endogenous ezrin was immunoprecipitated from SYF or SYF/SrcWT cells, no phosphorylation was detected with the anti-phosphotyrosine antibody ( Fig. 2A). On the contrary, in the SYF cells, which express the constitutively active mutant Src (SYF/ Y527FSrc), ezrin was phosphorylated on tyrosine ( Fig. 2A). Immunoblotting of the whole cell lysates with the antibody recognizing active (pY418)Src confirmed that only SYF/ Y527FSrc cells contained active Src ( Fig. 2A, right panel), whereas SYF/SrcWT expressed low levels of wild-type Src, which is not active in basal cell culture conditions (25). When SYF/SrcWT cells were treated with pervanadate, ezrin became tyrosine-phosphorylated, whereas pervanadate had no effect on SYF cells (Fig. 2B). These results confirm that Src kinase activity is explicitly involved in phosphorylation of ezrin.
Mapping of the Phosphotyrosine Residue in Vitro-To map the tyrosine residue(s) phosphorylated by Src, we generated recombinant ezrin fragments fused with GST, produced them in E. coli, purified them, and subjected to in vitro phosphorylation reaction with recombinant Src kinase and [␥-33 P]ATP. Previously identified ezrin phosphorylation sites, Tyr 145 and Tyr 353 (17), were mutated to phenylalanine in some of the fusion peptides to directly analyze whether they are the target sites for Src. Autoradiographs of the reaction mixtures separated in SDS-PAGE showed that the C-terminal half of ezrin (amino acids 309 -585) and a short C-terminal construct (amino acids 477-585) had incorporated 33 P (Fig. 3A). The fusion pro-FIG. 1. Tyrosine phosphorylation of ezrin in 293 cells. Cells were treated (ϩ) or untreated (Ϫ) with pervanadate and PP2. Ezrin was immunoprecipitated, and its phosphorylation status was analyzed by immunoblotting with an anti-phosphotyrosine antibody (left panel). Subsequently, the same blot was stripped and reprobed with the anti-ezrin antibody (middle panel). Src is activated during pervanadate treatment but not after preincubation with PP2 as shown by immunoblotting of whole cell lysates with an anti-Src-pY418 antibody (right panel). Ezrin is phosphorylated on tyrosine after pervanadate treatment, and phosphorylation is inhibited by preincubation with PP2. WCL, immunoblot of whole cell lysate; WB, Western Blot; pre-imm, immunoprecipitation (IP) with pre-immune serum. teins in which the previously identified phosphotyrosines were mutated did not differ from the corresponding wild-type fusion proteins. Interestingly, a C-terminal construct of merlin (amino acids 492-595), homologous to the short ezrin C-terminal construct, was not phosphorylated in a parallel experiment.
Analysis of the amino acid sequence of the short C-terminal construct revealed altogether four tyrosines, each of which was mutated to phenylalanine one by one and introduced into the GST fusion construct containing the C-terminal half of ezrin (amino acids 309 -585). In vitro phosphorylation reaction of the mutated fusion proteins showed that mutation of Tyr 477 but not the three other tyrosines abolished the phosphorylation induced by Src (Fig. 3B). The phosphorylation achieved a stoichiometry of 0.87 mol of phosphate/mol wild-type fusion protein (data not shown). These results indicate that Tyr 477 is the major site of phosphorylation in vitro.
Tyr 477 Is the Major Src-catalyzed Phosphotyrosine in Vivo-To see whether Tyr 477 is phosphorylated by Src also in vivo, ezrin expression plasmids carrying the Y477F mutation were constructed and introduced to HEK293 cells by transfection. VSVG-tagged wild-type and mutant ezrin from control or pervanadate-treated cells was immunoprecipitated with an anti-VSVG-tag antibody and phosphorylation was probed with anti-pTyr antibody. The expressed wild-type protein was readily phosphorylated after pervanadate treatment (Fig. 4). Interestingly, also the Y477F ezrin showed a low level of pTyr reactivity (Ͻ10% wild-type ezrin) after pervanadate treatment. To further characterize this remaining pTyr reactivity, we introduced a second mutation to the same expression construct, namely Y145F, because tyrosine 145 has been shown to be a phosphorylation target in ezrin (15,17,20). The double mutant Y145F/Y477F showed similar pTyr reactivity as ezrin-Y477F when the cells were stimulated with pervanadate (Fig. 4). The result indicated that Tyr 477 is the major site of phosphorylation in ezrin during pervanadate treatment.
To better study the phosphorylation of Tyr 477 in vivo, we generated a phosphospecific ezrin-Tyr 477 antibody. Its specificity was characterized with several methods. The antibody recognized recombinant C-terminal ezrin only when the peptide was phosphorylated in vitro by recombinant Src. The reactivity was lost if the ezrin preparation was subsequently dephosphorylated by CIP (Fig. 5A). The anti-ezrin-pTyr 477 antibody recognized immunoprecipitated wild-type ezrin after pervanadate treatment, whereas ezrin mutants Y477F or Y145F/Y477F were not recognized under similar conditions (Fig. 5B). Immunoblotting of 293 cell lysates containing endogenous or trans- the antigenic phosphopeptide (Fig. 5C). Furthermore, pervanadate-treated lysates containing transfected Y145F or Y353F ezrin mutants reacted with the anti-ezrin-pTyr 477 antibody, whereas the lysates expressing Y477F or Y145F/Y477F did not react (Fig. 5D). In conclusion, these results indicate that the antibody is indeed specific for ezrin-Tyr 477 when phosphorylated by Src.
Tyr 477 Is phosphorylated in SYF/SrcY527F Cells-To see whether ezrin can be constitutively phosphorylated on Tyr 477 in cells with up-regulated Src activity, we analyzed SYF cells and SYF/SrcWT and SYF/SrcY527F variants with anti-ezrin-pTyr 477 antibody. In these experiments, the antibody reacted with ezrin in unstimulated SYF/SrcY527F cells (Fig. 5E). The result further verified that Tyr 477 is a target for Src in these cells.
Phosphorylation of Tyr 477 and Thr 566 Are Independently Regulated-Ezrin undergoes conformational regulation, which is dependent on the phosphorylation of Thr 566 . Previous studies have shown that T566D substitution renders ezrin in a constitutively active form (12,31). We wanted to analyze whether Tyr 477 phosphorylation affects the phosphorylation level of Thr 566 . HEK293 cells were transfected with VSVG-tagged wild-type and Y477F mutant ezrin, and lysates of transfected or untransfected cells treated with pervanadate were probed with a phosphospecific antibody against phosphorylated Thr 566 ezrin. Both forms of ezrin reacted with anti-pThr 566 antibody at a similar level (Fig. 6A). The result suggests that Tyr 477 phosphorylation does not promote Thr 566 phosphorylation.

Phosphorylation of Tyr 477 Does Not Affect Ezrin Head-to-Tail
Interaction-To find out the effect of ezrin Tyr 477 phosphorylation on the intramolecular N-ERMAD to C-ERMAD association, we tested how tyrosine-phosphorylated ezrin self-associates in an affinity precipitation assay. The C-terminal ezrin GST fusion protein (amino acids 309 -595) was either nonphosphorylated or phosphorylated in vitro with recombinant Src and bound to glutathione-Sepharose. One phosphorylated sample was further treated with CIP to verify the effect of phosphorylation on binding. Lysates of transfected cells expressing N-terminal ezrin-(1-309) protein were incubated with the treated C-ezrin-GST beads. N-ezrin bound to beads was detected by immunoblotting of the washed and eluted samples separated on SDS-PAGE. The results show that non-phosphorylated, phosphorylated, or phosphorylated and CIP-treated C-terminal ezrin-GST bound N-ezrin in a similar manner (Fig.  6B). This indicates that the phosphorylation of Tyr 477 does not affect the ability of ezrin to form an intramolecular or intermolecular association via N-and C-ERMADs.
Tyr 477 Is Not Phosphorylated after EGF Stimulation in A431 Cells-Ezrin is a major cellular protein phosphorylated on ty-rosine after EGF stimulation (14). To see whether EGF stimulates Tyr 477 phosphorylation on ezrin, we stimulated A431 epidermoid carcinoma cells with EGF. After immunoprecipitation of endogenous ezrin, we probed the samples either with a general anti-pY antibody or with the anti-ezrin-pTyr 477 antibody. The results shown in Fig. 7A indicate that, although EGF stimulates ezrin tyrosine phosphorylation, it does not activate Src and does not lead either directly or indirectly to the phosphorylation of Tyr 477 . In parallel experiments, pervanadate treatment of the cells resulted in Src activation and phosphorylation of ezrin at Tyr 477 .
Tyr 477 Is Not Phosphorylated in Jurkat T-cells-We have shown earlier that the Src family kinase Lck phosphorylates ezrin at Tyr 145 in Jurkat T-cells (20) and wanted to test whether Tyr 477 was phosphorylated also in these non-adherent cells during pervanadate treatment. Although immunoblotting of the whole cell lysates showed that pervanadate induced a prominent phosphorylation on tyrosine as detected with anti-pY antibody, there was no signal when the lysates were probed with anti-ezrin-pTyr 477 (Fig. 7B).
Src-dependent Phosphorylation of Ezrin Induces Association with KBTBD2-To identify Src phosphorylation-dependent binding partners for ezrin, we performed a yeast two-hybrid screen of a fetal brain library with bait encoding C-terminal ezrin together with active Src. The positive clones of the screen were further analyzed to verify the dependence of the interaction of Src activity by a filter ␤-galactosidase assay in which the interaction was detected by the blue color of the yeast colony. In this assay, the identified clones were tested against C-terminal ezrin in a vector encoding for Src or a similar vector without Src. This assay distinguished between interactions that occurred independent of Src (e.g. clone A) and others that required the presence of Src (Fig. 8A). Sequencing of the clones, whose interaction was Src-dependent followed by data base searches, identified two independent clones encoding for KBTBD2. Both clones contained the sequence of the 3Ј end of KBTBD2 cDNA and were missing ϳ560 nucleotides from the 5Ј end of the open reading frame. To further verify the interaction and demonstrate its dependence on Src activity, full-length KBTBD2 cDNA was obtained and fused to GST and the fusion protein was used in an affinity precipitation assay. Lysates containing VSVG-tagged C-terminal ezrin or fulllength ezrin from control cells or cells treated with pervanadate were allowed to bind to GST or GST-KBTBD2. The results showed significant binding between ezrin and KBTBD2 only under conditions in which ezrin was phosphorylated by Src (Fig. 8B).
FIG. 6. Phosphorylation of Tyr 477 does not affect Thr 566 phosphorylation or N-ERMAD-C-ERMAD interaction. A, 293 cells expressing endogenous, VSVG-tagged wild-type, or VSVG-tagged Y477F ezrin were treated or untreated with pervanadate. Whole cell lysates were probed with the anti-pThr 566 antibody. The wild-type and Y477F ezrin show a similar level of reactivity independently of pervanadate treatment. B, C-terminal GST-ezrin-(309 -585) was untreated, phosphorylated with Src, or phosphorylated, and phosphatase-treated (CIP) as in Fig. 5A. Glutathione-Sepharose bead-bound GST peptides were used to precipitate N-terminal VSVG-ezrin-(1-309) from cellular lysates. One half of the sample was analyzed in Western blots with anti-VSVG antibody (left panel), and the other half was analyzed with anti-pY antibody (right panel). Phosphorylation by Src does not affect ezrin N-ERMAD-C-ERMAD interaction. untransf., untransfected; WB, Western blot.

DISCUSSION
The oncogenic kinase Src is involved in multiple signal transduction cascades. Many of these pathways are coupled to cytoskeletal alterations, and various Src substrate proteins are indeed functionally linked to cytoskeleton. Cells transformed by Src display reduced actin bundling and fewer cellular contacts. On one hand, Src is activated by extracellular signals via cell surface receptors such as growth factor receptors or integrins. On the other hand, the activation of Src affects cell adhesion to extracellular matrix and other cells via focal adhesions and adherens junctions. In this study, we examined the connection between Src and the cytoskeleton-cell membrane linker ezrin and showed that Src phosphorylates ezrin at Tyr 477 . This is a previously unknown form of ezrin regulation and might have important functional consequences, because it connects ezrin to an oncogenic signaling pathway. Further-more, we identified a novel molecular interaction, which is dependent on the Src-driven ezrin regulation.
Regulation of ezrin by Src is unique among ERM proteins, because Tyr 477 is not conserved in other ERM members or merlin. Tyr 477 resides in close proximity to a seven-proline stretch, which is also present in radixin sequence. This region is partly included in the crystal structure of the complexed FERM and tail domain of human moesin (32). Tyrosine at position 477 is located between the stretch of prolines and the C-ERMAD, which includes an actin binding site. An isolated radixin ␣-helical domain can form a monomeric and stable helical rod structure, which is stabilized by extensive salt bridge interactions (33). Based on structural studies on other ERM proteins, it is possible that Tyr 477 , located in the vicinity of the end of the ␣-domain rod and the stretch of prolines, could be available for a kinase even when ezrin is in closed confor- mation via the association of N-to C-ERMAD. Furthermore, the finding that pervanadate treatment of cells expressing wild-type or Y477F mutant ezrin does not affect phosphorylation of the conserved threonine 566 suggests that conformational opening of the molecule and phosphorylation of Tyr 477 are independently regulated.
In addition to Src, stimulation by EGF, platelet-derived growth factor, or hepatocyte growth factor results in tyrosine phosphorylation of ezrin (14 -16). However, the outcome in regard to target residues is dependent on the type of stimulation and, despite close sequence similarity, various ERM proteins display tight specificity for tyrosine phosphorylation. Both ezrin and radixin, but not moesin, are phosphorylated on tyrosine following platelet-derived growth factor receptor activation (16). After EGF stimulation, ezrin is phosphorylated at Tyr 145 and Tyr 353 , whereas radixin or moesin is not (16,17), although Tyr 145 is conserved in all family members. In agreement with previous studies, we detected tyrosine phosphorylation of ezrin after 5 min of EGF stimulation in A431 cells but we found no evidence of Src induction or phosphorylation of ezrin at Tyr 477 . Interestingly, also Krieg and Hunter (17) reported that the tryptic phosphopeptide map of an ezrin construct phosphorylated in vitro by pp60 c-Src did not show similarity with the map derived from EGF-stimulated A431 cells (17). We conclude that EGF and Src stimulation differentially regulate ezrin.
In our study, we did not detect any change in Src-induced phosphorylation of ezrin in vivo when Tyr 145 or Tyr 353 was mutated to a phenylalanine. Both in vitro and in vivo experiments demonstrated that Tyr 477 is the major target residue for Src; however, a low level of tyrosine phosphorylation was present in lysates from cells expressing tagged Y477F mutant ezrin. This weak phosphorylation remained also in cells expressing the Y145F/Y477F double mutant. A plausible explanation for the result is that immunoprecipitation of transfected ezrin mutant coprecipitates phosphorylated endogenous ezrin because of the formation of homodimers or oligomers with transfected protein (12,34). However, we cannot rule out the possibility of a second target site(s) for Src in vivo, although it seems unlikely that this site would be Tyr 145 or Tyr 353 . Previously, we found that, in T-lymphocytes, the Src family kinase Lck catalyzes the phosphorylation of Tyr 145 (20). In this study, we did not see any phosphorylation on Tyr 477 in Jurkat cells after pervanadate treatment. The differences between adherent cells and cells growing in suspension may account for the distinct phosphorylation. The signaling pathways involving Src family kinases regulate not only adhesion but also various other cellular responses involving cytoskeletal changes, which are differentially regulated in circulating lymphatic cells and in adherent cells. The results in A431 and Jurkat cells suggest a specific role for ezrin-Tyr 477 phosphorylation in cell signaling, because they point out to a cell type-and stimulation-dependent regulation.
Src phosphorylates a variety of cytoskeletal proteins, but the consequences of phosphorylation are variable. The epithelial cell actin-binding protein villin is phosphorylated on several tyrosines by Src, and phosphorylation of villin plays an essential role in cell migration and actin organization (35). Phosphorylation decreases the binding affinity of villin for F-actin (36). It is also reported that tyrosine phosphorylation of villin releases an autoinhibited conformation (37). Another substrate for Src is vinculin (38), an actin-binding protein regulated by phosphatidylinositol 4,5-bisphosphate binding. Tyrosine phosphorylation of vinculin does not affect its actin binding but decreases the intramolecular binding of the vinculin head domain to the vinculin tail (39). In the case of ezrin, it was recently shown that phosphatidylinositol 4,5-bisphosphate binding precedes threonine phosphorylation and is needed for the targeting of ezrin to cell membrane (31). The sequence of events concerning tyrosine phosphorylation, phosphatidylinositol 4,5-bisphosphate binding, and threonine phosphorylation of ezrin and the functions of the different heteromeric and homomeric molecular forms of the ERM proteins have yet to be clarified. However, our results suggest that Tyr 477 phosphorylation and opening of the molecule are separate events.
Another interesting cytoskeletal substrate for Src is cortactin (reviewed in Ref. 41). Although tyrosine phosphorylation does not affect F-actin binding of cortactin (42), it promotes cell migration (43). Interestingly, phosphorylation of platelet cortactin by Src induces proteolysis of cortactin by calpain (44). Ezrin is known as a calpain substrate (45,46), and we have observed an increase in proteolysis of tyrosine-phosphorylated ezrin by calpain in vitro (data not shown). Regulation of calpain-mediated proteolysis of ezrin could provide a mechanism to dynamically modulate cytoskeleton-membrane linkage during cell migration and invasion.
We identified a novel phosphospecific interaction with KBTBD2 protein as a functional consequence of Src-induced phosphorylation of ezrin. We used a modified yeast two-hybrid bait system containing active Src to find specific binding partners for tyrosine-phosphorylated ezrin. A similar approach has previously been successfully used to find interactions that depend on the phosphorylation of the bait protein by Src (28,47). KBTB2 has been discovered in screens of novel human cDNAs (48 -50), and in databases, EST sequences of KBTB2 are found ubiquitously in different tissues (UniGene, NCBI). However, its functions or molecular interactions have not been reported. KBTB2 contains an N-terminal BTB/POZ domain and five kelch motifs. BTB/POZ domains (reviewed in Refs. 51 and 52) are present in several zinc-finger transcription factors and kelch-repeat proteins and are conserved in evolution. The BTB/ POZ domains mediate homomeric or heteromeric dimerization. Kelch family proteins typically contain multiple (four to seven) Kelch motifs that form a kelch-repeat domain, which folds as a conserved tertiary structure, a ␤-propeller (reviewed in Refs. 53 and 54). Kelch-repeat superfamily members are widespread in evolution from viruses to mammals (54) and have versatile functions and cellular distribution. Interestingly, many actinbinding and actin-bundling proteins are found in this superfamily (55)(56)(57)(58)(59)(60)(61)(62)(63). Moreover, several kelch-repeat proteins are involved in the regulation of cell adhesion or cell morphology (64 -66).
Our results open novel avenues for analysis of the biological functions of ezrin in normal and abnormal cells. Several recent studies have demonstrated that ezrin is up-regulated in malignant tumors as is Src (reviewed in Refs. [67][68][69] and that this up-regulation may play a role in tumor dissemination and metastasis (70 -77). Recent generation of an ezrin knock-out mouse model (40) should provide tools to further study the role of Tyr 477 in these processes and dissect the signaling pathways in which ezrin and Src cooperate. In addition, a detailed characterization of KBTBD2 and the cellular complexes in which it is found is needed to understand how ezrin, Src, and KBTBD2 function in regulation of actin cytoskeleton, adhesion, or cellular signaling.