EWI-2 and EWI-F Link the Tetraspanin Web to the Actin Cytoskeleton through Their Direct Association with Ezrin-Radixin-Moesin Proteins*

EWI-2 and EWI-F, two members of a novel subfamily of Ig proteins, are direct partners of tetraspanins CD9 (Tspan29) and CD81 (Tspan28). These EWI proteins contain a stretch of basic charged amino acids in their cytoplasmic domains that may act as binding sites for actin-linking ezrin-radixin-moesin (ERM) proteins. Confocal microscopy analysis revealed that EWI-2 and EWI-F colocalized with ERM proteins at microspikes and microvilli of adherent cells and at the cellular uropod in polarized migrating leukocytes. Immunoprecipitation studies showed the association of EWI-2 and EWI-F with ERM proteins in vivo. Moreover, pulldown experiments and protein-protein binding assays with glutathione S-transferase fusion proteins containing the cytoplasmic domains of EWI proteins corroborated the strong and direct interaction between ERMs and these proteins. The active role of ERMs was further confirmed by double transfections with the N-terminal domain of moesin, which acts as a dominant negative form of ERMs, and was able to delocalize EWIs from the uropod of polarized leukocytes. In addition, direct association of EWI partner CD81 C-terminal domain with ERMs was also demonstrated. Functionally, silencing of endogenous EWI-2 expression by short interfering RNA in lymphoid CEM cells augmented cell migration, cellular polarity, and increased phosphorylation of ERMs. Hence, EWI proteins, through their direct interaction with ERM proteins, act as linkers to connect tetraspanin-associated microdomains to actin cytoskeleton regulating cell motility and polarity.

Tetraspanins are a large family of ubiquitous membrane proteins characterized by their four transmembrane, two extracellular loops, one intracellular loop, and two short N-and C-terminal cytoplasmic domains. The second large extracellular loop contains the CCG and PXSC motifs, which are hallmarks of this family of type III proteins (9 -11). Through their large extracellular domain, tetraspanins interact with themselves and with other proteins such as growth factors, complement regulatory proteins, proteoglycans, signaling enzymes, and integrins (11)(12)(13)(14). This ability of tetraspanins allows them to organize the cellular surface into structural and functional microdomains, named tetraspanin web or tetraspanin-enriched microdomains (TEM) 2 (12)(13)(14). At this molecular network, tetraspanins are capable of modulating the functions of associated proteins. For example, tetraspanins modulate integrin-dependent migration (15) and ␣3␤1-dependent neurite outgrowth (16). Furthermore, "outside-in" signaling and adhesion strengthening through ␣6␤1 and ␣3␤1 integrins are influenced by their lateral association with CD151 (17)(18)(19). CD151 is also essential for the normal outside-in ␣IIb␤3 integrin signaling in platelets (20) and modulates the activation-dependent conformation of integrin ␣3␤1 (21). Tetraspanin microdomains also have a crucial role in the proper adhesive function of ICAM-1 and VCAM-1 during leukocyte adhesion and transendothelial migration (22).
ERM proteins belong to the 4.1 band superfamily, which is a major component of cortical cytoskeleton involved in membrane-cytoskeletal associations (24,25). They are concentrated in actin-rich surface structures such as microvilli, filopodia, and membrane ruffles (24,26). In polarized leukocytes, ERM proteins are redistributed to the cellular uropod, a membrane protrusion at the rear part of the cell body (27). Two conformational states have been described for ERMs: an inactive form, soluble in the cytoplasm, and an unfolded active state (26). Changes in conformation are regulated by the phosphorylation of a conserved threonine residue in their C-terminal domain. This phosphorylation disrupts the intramolecular association between the N-and the C-terminal domains and the protein unfolds (26). The C-terminal domain contains an actin-binding site, whereas N-terminal domain of ERMs interacts with the cytoplasmic domains of different adhesion molecules, including CD44, CD43, PSGL-1, ICAM-1-3, and VCAM-1 (28 -32). Most of these proteins contain a positively charged amino acid cluster at a juxtamembrane position that determines their association with ERMs (33).
Despite their structural functions, ERMs are also implicated in cell signaling (26,34). Several studies have demonstrated that ERM activation is linked to the Rho signaling pathway (26). Furthermore, it has been reported that moesin and ezrin interact with Syk (35) and that ezrin is associated with p85, the regulatory subunit of phosphatidylinositol 3-kinase (36).
EWI-2 and EWI-F have a net positive charge in their cytoplasmic tails that makes their association with ERMs feasible. Here we have studied the possible interaction between EWI-2 and EWI-F with ERM proteins in order to gain functional insights of the EWI proteins into the tetraspanin web. Immunofluorescence and immunoprecipitation experiments demonstrated that there is indeed an association of EWI proteins with ERMs. In vitro protein-protein binding assays confirmed that this association is strong and direct, but an indirect mechanism through tetraspanin CD81 also exists. Finally, by silencing the expression of endogenous EWI-2, we show a role of EWI-2 in ERM-dependent processes, such as cell polarization and migration.
Flow Cytometry Analysis, Immunofluorescence, and Confocal Microscopy-Cells were stained as described previously (15) and analyzed in a FACScan cytofluorometer (BD Biosciences) using the CellQuest program (BD Biosciences).
Confocal microscopy analyses were performed using a Leica TCS-SP confocal laser scanning unit equipped with argon and He/Ne laser beams and attached to a Leica DMIRBE inverted epifluorescence microscope (Leica Microsystems, Heidelberg, Germany), using a ϫ63 oil immersion objective.
Cell Transfection-Transiently transfected NS1 cells were generated by electroporation. 2 ϫ 10 7 cells were electroporated with 10 g of DNA in Opti-MEM medium (Invitrogen) at 280 V and 1200 microfarads (Gene Pulser II; Bio-Rad). 24 h after transfection, cells were processed for immunofluorescence experiments as described previously (37).
Immunoprecipitation and Western Blot-For immunoprecipitation, cells were washed two times with Tris-buffered saline and lysed by incubation for 30 min in 1% Nonidet P-40, 2 mM CaCl 2 , 2 mM MgCl 2 , phosphatase inhibitors (1 mM sodium pyrophosphate, 1 mm sodium orthovanadate, 10 mM sodium fluoride), and protease inhibitors (Complete; Roche Applied Science) in Tris-buffered saline. Cell lysates were precleared 4 h with Gly-Sepharose and immunoprecipitated with mAbs coupled to protein-G-Sepharose. After washing six times with lysis buffer, proteins bound to Sepharose beads were eluted by boiling in sample buffer, subjected to 10% SDS-PAGE under reducing conditions, and transferred onto a nitrocellulose membrane (Trans-Blot transfer medium; Bio-Rad). Membranes were incubated with the polyclonal anti-ERM 90/3 antibody and a peroxidase-conjugated goat anti-rabbit IgG (Pierce). Proteins were visualized using an enhanced chemiluminescence system (Amersham Biosciences). Each peptide (30 nmol) was conjugated to 40 l of streptavidin-Sepharose (Amersham Biosciences). GST constructs were transformed into BL21 bacteria, induced using isopropyl 1-thio-␤-D-galactopyranoside (300 M) for 5 h, and purified by affinity chromatography with glutathione-Sepharose (Amersham Biosciences) (31,35). The amount of fusion protein was estimated by Coomassie Blue staining. Pulldown assays with HeLa cells extracts were carried out as described previously with either biotinylated peptides or GST constructs (35).
EWI-2 Silencing by siRNA-To selectively knockdown the expression of endogenous EWI-2 protein, two different target sequences were used as follows: oligo 1, against the target sequence GUUCUCCUAUGCUGUCUU corresponding to nucleotide sequence 253-270 of the open reading frame of EWI-2 (Eurogentec, Seraing, Belgium); and oligo 2, with the sequence GGCUUC-GAAAACGGUGAUC that pairs with nucleotide sequence 1824 -1835 of the open reading frame of EWI-2 (reference number 215098; Ambion, Austin, TX). RNA duplexes (2 M per sample) corresponding to these target sequences, as well as a negative oligonucleotide that does not pair with any human mRNA (Eurogentec), were nucleofected in CEM cells using the cell line Nucleofector kit V (Amaxa GmbH, Cologne, Germany). Cells were transfected on day 0, washed on day 1, and experiments performed on day 3.
Migration Assays-Migration assays were performed in 5-m diameter pore Transwell cell culture chambers (Costar, Cambridge MA). CEM cells transfected with siRNA oligonucleotides were added to the upper chamber of the Transwell inserts (5 ϫ 10 5 /100 l). In the lower well, 600 l of completed RPMI medium, with or without 100 ng/ml of the chemokine SDF-1␣ (R&D, Minneapolis, MN), was added. Chambers were incubated for 2-4 h at 37°C. Migrated CEM cells were recovered from the lower chamber and estimated by flow cytometry. For transendothelial migration assays human umbilical vein endothelial cells (HUVEC) were grown to confluence on Transwell inserts precoated with 20 g/ml fibronectin and activated with tumor necrosis factor-␣ (20 ng/ml) for 20 h before adding the lymphoid cells (22).

EWI-2 and EWI-F Are Localized at ERM-enriched Subcellular Structures-
The expression pattern of EWI-2 and EWI-F was first analyzed in different cellular types by flow cytometry. We studied their expression in lymphoid, monocyte-derived dendritic cells and some adherent cell types using specific monoclonal antibodies. EWI-2 was expressed in T and B cells (Fig. 1A). Dendritic cells showed an increment of EWI-2 expression during their maturation induced with LPS (Fig. 1A, asterisk), although no increment was observed in PBLs when stimulated with the chemokine SDF-1␣ (Fig. 1A, double asterisk). In contrast, EWI-F was expressed by HeLa and HUVEC cells (Fig. 1A), with no detectable expression in the bone marrow-derived cells tested (Fig. 1A).
Next, the subcellular localization of EWI-2 and EWI-F was studied by immunofluorescence microscopy. In polarized lymphoid cell lines (CEM and HSB-2) and PBLs stimulated with SDF-1␣, EWI-2 was concentrated at the cellular uropod (Fig.   1B), in a similar pattern and frequency to that of the ERMassociated protein ICAM-3 ( Fig. 1, C and D). CD147 is included as a control protein that does not redistribute to the cellular uropod. On the other hand, EWI-F was located at the intercellular contacts of HUVEC cells, and at apical microvilli and microspikes of HeLa cells (Fig. 1B).
At a subcellular level, ERM proteins are preferentially located at the uropod of polarized leukocytes as well as cell-cell contacts and apical microvilli of adherent cells (24,26,27). Therefore, the possible colocalization of EWI proteins with ERMs was assessed by confocal double immunofluorescence microscopy. We found that in lymphoid CEM cells, EWI-2 colocalized with ERMs at the cellular uropods, whereas EWI-F showed a high degree of colocalization with ERMs at the apical surface of HeLa cells ( Fig. 2A).
NS1 cells are a constitutively polarized mouse myeloma cell line. Exogenous plasma membrane proteins that are associated with ERMs are redistributed to the cellular uropod when transfected into these cells (37). When cDNAs of human EWI-2 and EWI-F were transfected into NS1 cells, both molecules were redistributed to the cellular uropod (Fig. 2B).
EWI-2 and EWI-F Associate in Vivo and in Vitro with ERMs-To determine whether ERM proteins interact in vivo with EWI-2 and EWI-F, HeLa (for EWI-F) and HSB-2 and CEM cells (for EWI-2) were lysed in 1% Nonidet P-40 containing lysis buffer and immunoprecipitated with different mAbs. Anti- CD44 and anti-ICAM-1 were used as positive controls of interaction with ERMs (28,33). In HeLa cells, EWI-F and ERMs coprecipitated (Fig. 3A), and a similar association of EWI-2 and ERM was detected in HSB-2 and CEM cell lysates (Fig. 3A). In contrast, no coprecipitation of ERMs with CD59 (which anchors to the plasma membrane by a glycosylphosphatidylinositol linkage) or CD45 was observed. These data indicate that there is an in vivo association of ERMs with EWI-2 and EWI-F molecules.
To investigate whether EWI-2 and EWI-F directly bind to ERMs, GST fusion proteins containing the cytoplasmic domains of EWI-2 or EWI-F were generated and incubated with HeLa cell lysates. These pulldown biochemical analyses showed that ERMs associate with the cytoplasmic domain of both EWI proteins (Fig. 3B), whereas no significant binding was observed with the negative control GST.
To determine whether EWI-ERM interaction was direct, N-ezrin and N-moesin domains, obtained by in vitro translation, were incubated with EWI-Cyt-GST fusion proteins. A direct interaction was found between the cytoplasmic domain of EWI-2, EWI-F, and the N-terminal domain of ezrin and moesin, whereas a negligible signal was observed with GST ( Fig.   3, C and D, and quantified in E). Wild type PSGL-1 and a mutant of PSGL-1 (R345Stop or 1-11), which does not bind to ERMs (31,35), were also included as additional positive and negative controls (Fig. 3D). Therefore, EWI-2 and EWI-F bind directly to the N-terminal domain of ezrin and moesin.
ERMs Are Necessary for EWI-2 Localization at the Cellular Uropod-To further corroborate the role of ERM proteins in EWI subcellular localization, double transfections of EWI cDNAs and constructs of moesin coupled to GFP protein (fulllength moesin-GFP and N-terminal moesin-GFP) were performed. N-moesin domain, which lacks the actin-binding motif of the molecule, is able to interact with transmembrane proteins but not with the cortical cytoskeleton, so that remains evenly distributed throughout the cell periphery, acting as a dominant negative mutant of ERMs (37,41). EWI proteins and full-length moesin-GFP accumulated at the cellular uropod. However, when EWI proteins were cotransfected with N-moesin-GFP, they become uniformly distributed throughout the plasma membrane, confirming the active role of ERMs in their subcellular localization (Fig. 4A).
Association of ERMs with EWI Protein Direct Partner CD81-To determine whether the localization of EWI-2 relies only on its association with ERMs through its cytoplasmic domain, NS1 cells were transfected with a cytoplasmic truncated version of EWI-2, tagged with FLAG. The uropod localization of truncated EWI-2 was still observed, suggesting that other ERM-binding proteins, laterally associated with EWI-2, could also be involved in its subcellular localization (Fig. 4B).
Tetraspanins, direct partners of EWI proteins, appear to be good candidates for this putative indirect mechanism. Experiments of pulldown of HeLa lysates were performed, using as baits the biotinylated peptides comprising the C-terminal cytoplasmic domains of CD9, CD81, and CD151. The cytoplasmic C-terminal tail of CD81 associated with ERMs (Fig. 5A), whereas neither CD9 nor CD151 cytoplasmic domains were able to pull down ERMs. A biotinylated peptide corresponding to the sequence of the EWI-2 cytoplasmic domain was included as positive control (Fig. 5A). Likewise, CD81-ERM association was detected when pulldown experiments were carried out using a GST fusion protein containing the C-terminal cytoplasmic region of this tetraspanin (Fig. 5B). Protein-protein binding assays confirmed that there is an in vitro interaction between the C-terminal domain of CD81 and the N-terminal domains of ezrin and moesin (Fig. 5C). In vivo association between CD81 and ERM proteins was also detected in coimmunoprecipitation experiments (Fig. 5D). Interestingly, anti-CD9 but not anti-CD151 mAbs were also able to coimmunoprecipitate ERMs. Finally, CD81 and CD9 colocalized with ERM proteins at the cellular uropods of CEM cells and PBLs (Fig. 5E and data not shown). When transfected into NS1 cells, these tetraspanins were redistributed to the cellular uropod (Fig. 5F ). (7). However, the role of endogenously expressed EWI-2 in ERMdependent cell functions has not been addressed. Interference by two different siRNA oligonucleotides directed against EWI-2 in CEM cells that express high levels of this protein (see Fig. 1A) decreased its expression down to 20 (oligo 1) or 40% (oligo 2), respectively (Fig. 6A). Expression levels of direct and indirect partners (CD81, CD151, and ␤1 integrins), other ERMbinding proteins (ICAM-3, CD43, and CD44), and unrelated proteins such as ␤2 integrins were not significantly altered ( Fig.  6B and data not shown).

Role of EWI-2 in Cell Migration and Polarization-Exogenous expression of EWI-2 in Du145 metastatic prostate cells inhibits cell migration on both fibronectin and laminin
Cell migration assays with CEM cells transfected with both siRNA-EWI-2 oligonucleotides showed an increased cell migration compared with CEM cells transfected with a negative control siRNA oligonucleotide (Fig. 7, A and B). Similar results FIGURE 4. ERMs are necessary for EWI-2 localization at the cellular uropod. A, N-moesin delocalizes EWI proteins from the cellular uropod. NS1 cells were double transfected with cDNAs coding for human EWI proteins and two GFP constructs of moesin (full-length moesin and N-moesin-GFP), seeded onto fibronectin and stained for EWI proteins. Merged images of the GFP signal and the EWI staining are shown, together with the corresponding DIC images. Bars ϭ 10 m. B, the cytoplasmic domain of EWI-2 is not strictly required for its localization at the cellular uropod. NS1 cells were transfected with FLAG-tagged constructs of EWI-2 wild type (WT) and a mutant of EWI-2 lacking the cytoplasmic domain (⌬cyt), seeded onto fibronectin, and stained with a fluorescein isothiocyanate-conjugated anti-FLAG mAb. The corresponding DIC image is also shown. Arrowheads point to uropods of polarized NS1 cells. Bars ϭ 10 m.
were observed when chemotaxis toward the chemokine SDF-1␣ was assessed (Fig. 7, A and B). Moreover, transendothelial migration was also augmented in EWI-2-interfered CEM cells (data not shown). This functional effect in lymphocyte motility pointed us to analyze the morphology of EWI-2-interfered CEM cells showing that the frequency of polarized cells bearing uropods was also significantly increased (Fig. 7C).
It has been described that ERM phosphorylation has a potential function in the induction of the uropod in T-lymphocytes (42). We therefore analyzed whether interference of EWI-2 expression affects ERM phosphorylation. Lysates of CEM cells interfered with the negative oligo siRNA, and each one of the two EWI-2 siRNA oligonucleotides were blotted with an anti-phospho-ERM antibody, thus showing an increased phosphorylation of ERMs in EWI-2-interfered cells (Fig. 7D). These results point to an EWI-ERM interaction as a key regulator in lymphocyte polarity and motility.

DISCUSSION
We demonstrate herein the direct association of EWI-2 and EWI-F and one of their direct partners, tetraspanin CD81, with ERM actin-connecting molecules and their role in migration and cell polarization. These data point to a novel function for EWIs in providing a link to the tetraspanin web with the actin cytoskeleton and certain signal transduction pathways.
Membrane interactions with the cytoskeleton appear to be a key phenomenon in the regulation of many cellular functions. Because very few integral membrane proteins have been found to directly interact with actin, it is feasible that different accessory proteins would connect the actin cytoskeleton with the plasma membrane (24,43). Some of these adaptor proteins are ERMs (24,25) that bind to the actin cytoskeleton through their C-terminal domain and to some plasma membrane proteins through their N-terminal domain. It has been reported that CD43, CD44, ICAM-1 and -2, and VCAM-1, which bind to ERM proteins, bear positively charged amino acid clusters in their cytoplasmic domains, near the juxta-membrane domain (28,33). CD44 and ICAM-1 additionally share a basic isoelectric point based on the balance of positive and negatively charged amino acid residues in the whole cytoplasmic domain  (Table 1). However, not only net charge but three-dimensional structure is important for ERM binding (33). Thus, the cytoplasmic domain of ICAM-3, which also binds to ERM proteins, does not have a basic amino acid cluster and has an acid isoelectric point. Instead, ICAM-3 contains a serine-rich motif (Ser 487 and Ser 489 ) that is critical for ERM binding (37). Phosphorylation of these residues regulates ERM-directed polarization of ICAM-3 (37).
Because EWI-2 and EWI-F possess positively charged amino acids in their cytoplasmic domains (Table  1), we hypothesized a possible interaction with ERM proteins. In fact, the cytoplasmic domain of EWI-2 shows five basic residues that are near the juxta-membrane domain and confer with the cytoplasmic domain a highly positive net charge (isoelectric point ϭ 10.93). EWI-F cytoplasmic sequence also has some basic residues, although they are not so close to the juxta-membrane region and some acid residues give the cytoplasmic domain a clearly lower isoelectric point (8.68). Immunoprecipitation and proteinprotein binding assays revealed that there is indeed an in vivo association with ERMs through the cytoplasmic domain of these proteins. The lower isoelectric point of EWI-F could account for the different binding capacities between EWI proteins observed in direct binding assays.
Double transfection experiments revealed that EWI proteins were delocalized from the cellular uropod when cotransfected in NS1 cells with a dominant negative form of moesin that cannot bind to F-actin, confirming the crucial role of ERMs in EWI subcellular localization. However, when a cytoplasmic truncated version of EWI-2 was transfected, it was still found at the cellular uropod. Interestingly enough, another ERM-binding protein that is also included in the tetraspanin microdomains, such as CD44, shows the same behavior (44). All these data suggested that a correct ERM association is necessary for EWI localization, but in the absence of the primary association site (the cytoplasmic domain) other ERM-binding proteins, probably included into the tetraspanin web, could also relocalize EWI proteins to ERM-enriched sites. In this regard, we have confirmed that CD81 binds directly to ERM proteins. In the case of CD9, anti-CD9 mAbs were able to coimmunoprecipitate ERMs from cell lysates, but experiments of pulldown with a biotinylated peptide containing the C-terminal cytoplasmic domain of CD9 did not show a direct interaction. This association might again occur via an indirect mechanism involving EWI proteins or CD81 in the context of the tetraspanin web. In fact, heterodimerization of tetraspanins through a conserved hydro-phobic interface has been proposed (45). Alternatively, tetraspanin intracellular loop or the N-terminal domain could be responsible for CD9/ERM association ( Table 1).
The association of EWI proteins and CD81 with ERMs inside the TEM has potential relevant consequences in several cellular processes. It is well known that tetraspanins are involved in intercellular adhesion and fusion events (14,46). An essential role for CD9 and CD81 in mouse sperm-egg fusion has been reported (47)(48)(49)(50). Oocytes fuse to sperm predominantly at their microvilli-rich region, where CD9 is located (47)(48)(49), and the phosphorylation/dephosphorylation balance of ERM proteins contributes to the formation of microvilli (24). Under such conditions, EWI and CD81 interactions with ERMs could regulate the formation of microvilli, and the absence of tetraspanins CD9 or CD81 might inhibit fusion just by disruption of the microvillar structure. At this localization, a cis interaction of CD9 and CD81 with other proteins on the oocyte plasma membrane might be also important for fusion, a possibility that is consistent with the putative role of these tetraspanins as organizers of the tetraspanin web on these cells. The recent demonstration that none of the integrins known to be present on the mouse egg are essential for sperm-egg binding and fusion suggests that the actual sperm receptor in mice is not an integrin (51), so an alternative tetraspanin partner might be the key, for example EWI-F, which is also expressed in oocytes (50). On the other hand, the fusion of human gametes is more dependent on ␣6␤1 integrin than mouse gamete fusion (52). This integrin role is modulated by tetraspanin CD9, which controls its necessary redistribution into clusters (52). Another interesting possibility is that tetraspanin complexes may work in trans. In this regard, a mouse sperm fusion-related molecule called Izumo has been recently identified (53), so a tetraspanin partner could be a possible ligand for Izumo.
The expression of CD81 in immune cells is involved in the pathogenesis of hepatitis C virus (HCV) infection (54). This tetraspanin has been identified as a cell-surface receptor for HCV in a screening using the soluble form of the E2 envelope protein of this virus (55). Although CD81 is required, it is not sufficient for HCV infection, and additional cell-surface coreceptors are involved in viral internalization (13). On the other hand, it has been described that HCV replication also requires microtubule and actin polymerization (56) that could be triggered by CD81 cross-linking. In this context, CD81/EWI-2 complexes and their association with ERM proteins could have an important role during both viral infection and replication. In fact, both EWI-2 and EWI-F as well as CD81 are detected in HCV target cells, including human hepatocytes and lymphoid B cells (1).
In regard to functional insights of EWI proteins into the tetraspanin web, we have demonstrated that EWI-2 is involved in lymphocyte migration. Previous studies demonstrated that exogenous expression of EWI-2 inhibited cell motility in Du145 metastatic cells (7). However, the role of endogenous EWI-2 had not been explored before. Our data show that silencing of EWI-2 expres-sion clearly augmented cell motility and chemotaxis. Moreover, it appears to be a relationship between the higher motility and an augmented cell polarization. In fact, EWI-2 silenced CEM cells presented higher frequency of cells bearing cellular uropods and increased ERM phosphorylation. These data support the notion that phosphorylation of ERM proteins is crucial in both morphological polarization and migration processes (42) and point to EWI-2 as a negative regulatory molecule.  's t test). B, cell migration and chemotaxis toward SDF-1␣ (100 ng/ml) assays were performed on Transwell chambers with CEM cells transfected with the negative control siRNA oligo and EWI-2 siRNA oligo 2. Data represent the mean Ϯ S.D. of migrated cells in a representative experiment. C, CEM cells interfered with negative siRNA oligo, and EWI-2 siRNA oligo 1 was seeded onto fibronectin (20 g/ml), fixed with 4% formaldehyde, and cellular morphology assessed by microscopy. Data represent the percentage of cellular uropods in 10 63ϫ fields of three independent experiments Ϯ S.D. (*, p Ͻ 0.002 in a Student's t test). D. CEM cells interfered with negative siRNA oligo, and EWI-2 siRNA oligos 1 and 2 were lysed in 1% Nonidet P-40 lysis buffer, resolved in a 10% SDS-PAGE, and blotted sequentially with anti-phospho-ERM pAb, anti-total-ERM pAb (90/3), and anti-␣-tubulin mAb.

TABLE 1 Comparison of the sequences and isoelectric points of the cytoplasmic domains of EWI proteins, tetraspanins, and some ERM-associated proteins
Isoelectric points of cytoplasmic domains were calculated by using the EXPASY software package Comput pI/Mw tool. The sequence data were obtained from EMBL/ GenBank/DDBJ under accession numbers AAL01052 (human EWI-2), NP_065173 (human EWI-F), AAQ14902 (human ICAM-1), AAQ14918 (human ICAM-3), and P16070 (human precursor CD44). For tetraspanins, sequences of their cytoplasmic C-and N-terminal domains and intracellular loops were obtained from Hemler et al. (57). Boldface letters depict basic residues.

Protein
Cytoplasmic

Association of EWIs with ERM Proteins
EWI-2-ERM association seems to be constitutive. However, it is feasible that this association is modulating ERM activation. In this sense, it is remarkable that increased ERM phosphorylation because of silencing of EWI-2 is not accompanied by an alteration of the expression of other ERM-binding proteins such as CD43, CD44 or ICAM-3. Nevertheless, it cannot be ruled out that the association of these proteins to ERMs could be affected. Thus far, there are no functional reports on silencing studies of any other ERM-binding protein, which would help to understand how the balance of ERM association with different adhesion receptors is regulated. On the other hand, association with EWI-2 could favor or inhibit the recruitment of different signaling proteins downstream of the ERMs. For example, PSGL-1 associates with the tyrosine kinase Syk, through its interaction with moesin (35). Moreover, there is a functional link between the p-ERM-mediated structural framework of the plasma membrane and Rho-ROCK signaling (26,34). ERMs also associate with phosphatidylinositol 4,5bisphosphate and play an important role in the activation of Rho GTPases by recruiting their positive and negative regulators (26,34).
In conclusion, we have demonstrated here that there is a direct association of EWI-2 and EWI-F and one of their direct partners, tetraspanin CD81, with ERM adaptor molecules. Moreover, an important role of EWI-2 in migration and cell polarization has been found. Therefore, EWIs could provide a link to the tetraspanin web with the actin cytoskeleton, acting as regulatory proteins and facilitating signal transduction downstream from these TEM.