Moesin, like ezrin, colocalizes with actin in the cortical cytoskeleton in cultured cells, but its expression is more variable

Ezrin is a cytoskeletal protein originally identified as a com-ponent of the intestinal microvillus core and shown to be present in a wide variety of cultured cells in actin-contain-ing surface structures (Bretscher, 1983, 1989; Gould et al., 1986; Pakkanen et al., 1987, 1988; Pakkanen, 1988). It is a substrate of certain protein tyrosine kinases (Hunter and Cooper, 1981; Cooper et al., 1982; Gould et al., 1986). Phosphorylation and dephosphorylation of ezrin, at least in human carcinoma A431 cells stimulated with epidermal growth factor (EGF), parallels the appearance and disap-pearance of cell surface structures that are enriched in ezrin (Bretscher, 1989). The location of ezrin in cell surface structures suggests that it might link a membrane compo-nent with the underlying cytoskeleton. This possibility is supported by the finding that the amino-terminal approx. 260 residues of ezrin show 32% sequence identity to the glycophorin-binding domain of the membrane-cytoskeletal linking protein band 4.1 of the human erythrocyte (Ander-son and Marchesi, 1985; Leto and Marchesi, 1984; Conboy et al., 1986; Correas et al., 1986; Gould et al., 1989).

Over the past few years ezrin has been identified as a component of various structures that participate in a number of diverse processes. Ezrin becomes phosphorylated and appears to be involved in the cytoskeletal and membrane reorganization that delivers the gastric proton pump to the apical membrane of stimulated parietal cells (Hanzel et al., 1991). It is a tyrosine kinase substrate in activated T-cells (Egerton et al., 1992). Ezrin, or an immunologically related protein, is present as a component of the marginal band of chicken erythrocytes (Birgbauer and Solomon, 1989) and of the cortical cytoskeleton of neural growth cones (Goslin et al., 1989). Ezrin was independently identified using anti-bodies to a synthetic peptide based on a cloned human endogenous retroviral gag-related DNA sequence, and then isolated and called cytovillin (Suni et al., 1984; Pakkanen, 1988; Pakkanen et al., 1988; Pakkanen and Vaheri, 1989;

Turunen et al., 1989). Despite its presence in a wide vari-ety of systems, the function of ezrin and the significance of its phosphorylation remain unknown.

Proteins very closely related to ezrin have recently been described. Tsukita and colleagues (1989) isolated an F-actin barbed-end capping protein of 82 kDa from hepatic adherens junctions that they called radixin. Cloning and sequence analysis of mouse radixin cDNA revealed that it has 75% protein sequence identity to ezrin (Funayama et al., 1991). We recently described the isolation of two human placental proteins, p81 ezrin and a 77 kDa polypeptide, which cross-reacted with antibodies to chicken ezrin (Bretscher, 1989). The 77 kDa protein copurified with p81 ezrin through a number of chromatographic steps. Although the biochemical and immunological data suggest that p81 and p77 might be closely related, we were able to gener-ate antibodies specific for p81 ezrin (Bretscher, 1989). Con-currently, Lankes et al. (1988) purified a human heparin-binding protein of 77 kDa that was reported to be involved in the inhibition of smooth muscle cell proliferation. The sequence of the cDNA specifying this protein, termed moesin, revealed a protein sequence 74% identical to p81 ezrin and 81% identical to radixin. Moesin was found to be identical to the placental p77 protein that we had puri-fied (Lankes and Furthmayr, 1991).

A number of studies describing the localization of ezrin, moesin and radixin have been published. Ezrin has been reported to be enriched in surface structures that contain an actin cytoskeleton (ezrin: Bretscher, 1983, 1989; Gould et al., 1986; cytovillin: Pakkanen et al., 1987, 1988; Pakka-nen, 1988). Moesin was originally thought to be a mem-brane protein that binds to the extracellular matrix and is localized on the surface of cells (Lankes et al., 1988). How-ever, its sequence does not suggest a secreted protein (Lankes and Furthmayr, 1991) and we report here that it is an intracellular protein. Radixin has been localized to adherens junctions and to the contractile ring during cytoki-nesis (Tsukita et al., 1989; Sato et al., 1991). However, the interpretation of these data is complicated by the possibil-ity, now evident, that some antibody preparations may react with more than one of these closely related proteins. Indeed, all the antibodies used to date to localize radixin, also rec-ognize ezrin and moesin (Sato et al., 1992), casting some doubt on earlier localization studies. We have therefore attempted to generate antibodies specific for each of the two placental proteins, ezrin and moesin, that we have isolated. Here we describe the first detailed localization of moesin in cultured cells using specific antibodies, and compare its distribution with ezrin and actin. A preliminary report of this work has appeared (Franck and Bretscher, 1992).

NHF (normal human fibroblasts) and PtK₂ (rat kangaroo epithe-lial) cells were grown in Eagle’s minimum essential medium sup-plemented with 10% fetal calf serum; NRK (normal rat kidney) cells, MDBK (Madin-Darby bovine kidney epithelial) cells and Fu5C8 (rat hepatocytes) were grown in Eagle’s minimum essen-tial medium supplemented with 10% calf serum; and A431 (human epidermoid carcinoma) cells were grown in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal calf serum. Fu5C8 rat hepatoma cells (Lawrence and Brown, 1992) were obtained from Bill Brown (Cornell University). All other cells were obtained from the American Type Culture Collection (ATCC, Rockville, MD).

Human placental ezrin and moesin were purified essentially as described (Bretscher, 1989). Partial digestion of the purified pro-teins with α-chymotrypsin was performed as follows. α-Chy-motrypsin was added to ezrin or moesin at 0.3 mg/ml in 20 mM 2-(N-morpholino)ethane sulfonic acid (MES), 300 mM KCl, 0.5 mM dithiothreitol (DTT), pH 6.7, to give a final concentration of 10 μg/ml of the protease and incubated at 18°C. At various time points, samples were withdrawn, mixed with SDS-gel elec-trophoresis sample buffer and boiled. A 3 μg or 0.3 μg sample of the digest was analyzed for each time point for the protein stain-ing and immunoblotting, respectively (shown in Fig. 2).

A 2 ml sample of 0.3 mg/ml ezrin in 20 mM MES, 300 mM KCl, 0.5 mM DTT, pH 6.7, was digested with 10 μg/ml α-chymotrypsin for 80 min at 18°C. These conditions had been chosen from analy-sis of preliminary experiments to yield optimal production of the 32 kDa fragment with no significant intermediates of higher mol-ecular mass (Fig. 2). The reaction was stopped by the addition of 5 μM phenylmethylsulfonyl fluoride (PMSF). The sample was diluted twofold with 20 mM MES, pH 6.7, to reduce the KCl con-centration and made 80 mM in phosphate from a 0.8 M K₂HPO₄/KH₂PO₄, pH 7.0, stock solution, applied to a 3 ml hydroxyapatite column (HA-Ultrogel; LKB Instruments Inc., Gaithersburg, MD) and eluted directly with a 20 ml 100 mM to 800 mM potassium phosphate gradient, pH 7.0. The 32 kDa frag-ment eluted at about 400 mM phosphate. Peak fractions were dia-lyzed against 20 mM bis-Tris-propane, 20 mM NaCl, 0.5 mM DTT, pH 6.7, and applied to a 1 ml Mono-Q column equilibrated with this buffer. The 32 kDa fragment flowed through and was applied directly to a 1 ml Mono-S column. The column was washed with 10 ml 20 mM MES, 20 mM NaCl, 1 mM DTT, pH 6.7, and eluted with a 7 ml gradient of 20 mM to 720 mM NaCl in this buffer. Pure 32 kDa fragment eluted at about 500 mM NaCl. It was then subjected to automated sequence analysis on an Applied Biosystems model 470A gas-phase peptide sequencer. The 32 kDa fragment of moesin was transferred to a PVDF mem-brane and sequenced directly.

Antigen-affinity-purified antibodies to human ezrin have been described (Bretscher, 1989). Antibodies to human placental moesin were elicited in rabbits immunized with protein that had been subjected to preparative gel electrophoresis and affinity puri-fied as described in detail (Bretscher, 1983). Antibodies that were specific for moesin were obtained after removing ezrin-reactive antibodies on ezrin immobilized on nitrocellulose. The phospho-tyrosine antibody was a mouse monoclonal IgG2b_k (cat. no. 05-321, UBI, Lake Placid, NY).

Cells were grown on 10 mm coverslips, washed in PBS and fixed in freshly prepared 4% paraformaldehyde in PBS for 15 min at room temperature. They were permeabilized with 0.2% Triton X-100 in PBS for 5 min, and then incubated with 3 μg/ml of ezrin-or moesin-specific antibody for 60 min. After washing in PBS, the cells were incubated in 3.3 μg/ml fluorescein-labeled goat anti-rabbit IgG (ICN Biochemicals, Costa Mesa, CA) and 33 ng/ml rhodamine-phalloidin (Molecular Probes, Eugene, OR). After washing, the coverslips were mounted in glycerol containing 1 mg/ml phenylenediamine. Cells were observed on a Zeiss Uni-versal microscope and images photographed on Kodak T-max 3200 film (Eastman Kodak Co., Rochester, NY). Alternatively, cells were viewed in a Zeiss Axiovert 10 microscope and confo-cal images collected in a BioRad MRC600 confocal imaging system.

For immunoblotting of total cell extracts, cell lysates were pre-pared as described (Dinsmore and Solomon, 1991). Alternative methods employing extraction in RIPA-VF buffer (0.1% SDS, 1% Triton X-100, 1% deoxycholate, 0.15 M NaCl, 1 mM EDTA, 25 mM Tris-HCl, pH 7.4, 1 μg/ml leupeptin, 1 μg/ml pepstatin, 1 μg/ml aprotinin, 10 μg/ml soybean trypsin inhibitor, 0.25 mM PMSF and 100 μM Nα₃VO₄), or when cells were scraped directly into SDS-gel sample buffer (Laemmli, 1970), boiled for 2 min and clarified by centrifugation, gave identical results. Protein sam-ples were analyzed by 10% or 12% SDS-PAGE (Laemmli, 1970) and immunoblotted essentially as described by Towbin et al. (1979) except that transfer was performed using a semi-dry blot-ting apparatus according to the manufacturer’s instructions (Inte-grated Separation Systems, Hyde Park, MA). After transfer, blots were blocked with 3% bovine serum albumin and incubated with 30 or 50 ng/ml specific antibody. After washing and incubation with 1.9 μg/ml horseradish peroxidase-labeled goat anti-rabbit IgG (Cappel), bound IgG was visualized using an enhanced chemilu-minescence detection system (Amersham).

Immunoprecipitations on A431 cells were performed as follows. Subconfluent cells two days after passage were pretreated with either 10 μM phenylarsine oxide (PAO; Aldrich, Milwaukee, WI) in 0.01% DMSO, or 0.01% DMSO alone, for 20 min at 37°C; 20 nM EGF was then added to all dishes except those at the zero time points, and incubation at 37°C was continued until lysis at various times as indicated. Cells were lysed with 0.5% SDS in TBS containing 0.25 mM PMSF and 0.5 mM benzamidine, then diluted with SDS-deficient RIPA-VF (Gould et al., 1986) to give 0.1% SDS final concentration, and clarified by centrifugation at 100,000 g for 20 min; 300 μl samples of the supernatant were mixed with 200 μl of 25% (v/v) suspension of Protein A-Sepharose beads (Sigma, St. Louis, MO) plus either 1.1 μg of affinity-purified ezrin antibody or 15 μl of anti-moesin antiserum, and incubated with gentle shaking for 2 h at 4°C. Immunopre-cipitates were washed with RIPA-VF and then eluted from the beads with SDS-gel sample buffer. Proteins were separated by 8% SDS-PAGE and transferred to a PVDF membrane (Immobilon-P; Millipore) that was blocked with 10% non-fat dry milk. Blots were incubated with 1 μg/ml phosphotyrosine antibody, washed, then incubated with 0.5 μg/ml horseradish peroxidase-conjugated goat anti-mouse IgG + IgA + IgM (Gibco BRL, Grand Island, NY) and visualized by the enhanced chemiluminescence detection system (Amersham).

Steady-state ³⁵S labeling of A431 cells was achieved by growing the cells in methionine-free MEM (ICN, Costa Mesa, CA) supplemented with 10% fetal bovine serum, 10 μM L-methionine and 80 μCi/ml [³⁵S]methionine (Tran³⁵S-Label; ICN) for 10 h prior to lysis. Immunoprecipitated ³⁵S-labeled proteins from untreated cells were visualized after electrophoresis and transfer to the blot by autoradiography for 6 days at −70°C using Kodak X-OMAT film.

Affinity-purified antibodies to ezrin that do not recognize moesin (Fig. 1) have been described (Bretscher, 1989). To obtain antibodies specific for moesin, rabbits were immunized with purified human moesin that had been subjected to preparative gel electrophoresis. The antiserum to moesin reacted well with moesin, but also to some extent with ezrin (Fig. 1). Ezrin and moesin degradation products are also recognized. To obtain antibodies specific for moesin, the antibodies that eluted from immobilized moesin were absorbed on ezrin. The resulting antibody preparation rec-ognized moesin but not ezrin (Fig. 1).

Limited proteolytic digestion of ezrin and moesin with a variety of proteases indicated that both proteins contain a relatively resistant 32 kDa domain, whereas the rest of the fragments generated during proteolysis were remarkably different. An example of a digestion series with α-chy-motrypsin is shown in Fig. 2. Longer digestion of both pro-teins resulted in the accumulation of the 32 kDa domains as the largest protease resistant fragment (Fig. 2, lanes 8 and 9). The 32 kDa domains from ezrin and moesin were subjected to amino-terminal sequencing. The resulting sequences (PKPINVRVTTMDA for ezrin and PKTISVRVTT for moesin) correspond to the first amino acid residues of the intact proteins (Gould et al., 1989; Lankes and Furthmayr, 1991), indicating that the 32 kDa fragments are derived from their amino-terminal domains. Immunoblotting of the ezrin and moesin digests with anti-bodies specific to each of the two proteins revealed that nei-ther antibody recognized the highly related 32 kDa domains (Fig. 2).

In earlier studies (Bretscher, 1983, 1989; Gould et al., 1986; Pakkanen et al., 1987, 1988; Pakkanen, 1988) ezrin was found to be present in a wide variety of cell lines. With the availability of antibodies that specifically recognized ezrin and moesin, we wished to determine whether these proteins were found in the same or different cells, and to examine their relative abundance. The concentration of antibodies used for immunoblotting was adjusted so that they gave an equivalent signal for equal amounts of the purified human proteins. In this way, a direct comparison could be made for relative abundance of the two proteins within a cell line. Extracts of cells of fibroblastic or epithelial origin from various species were subjected to SDS-PAGE followed by immunoblotting with the ezrin- and moesin-specific anti-bodies (Fig. 3). All cells were found to contain ezrin, but their content of moesin varied considerably. Human fibrob-lasts and NRK cells contained about equal amounts of ezrin and moesin. In contrast, human A431 epidermoid carci-noma cells were rich in ezrin and contained little moesin, and moesin was not detectable in rat Fu5C8 hepatocytes or rat kangaroo PtK₂ cells. The variability in moesin content was not due to species specificity of the antibody, as large differences in moesin expression were found within one species (compare rat NRK cells and rat hepatocytes, or human fibroblasts and human A431 cells).

We used our specific antibodies in immunofluorescence microscopy to localize ezrin and moesin in several cultured cell lines. In all cases, the distribution was compared with that of F-actin, as revealed by staining with rhodamine-phalloidin, to determine whether ezrin or moesin colocalized with a specific subset of actin-containing structures. Here we document the distributions found in rat NRK cells and normal human fibroblasts. These cell lines contain ezrin and moesin in comparable amounts, and have either abun-dant or few surface structures, respectively. No staining of cells was seen unless they were permeabilized, indicating that both ezrin and moesin are intracellular proteins.

The localizations of actin and ezrin (Fig. 4A-D) and actin and moesin (Fig. 4E-H) by conventional fluorescence microscopy in NRK cells are shown. Micrographs taken at two different focal planes document that both ezrin and moesin are enriched and colocalize with actin in the abun-dant surface microvilli present on these cells. Neither pro-tein associates with the actin-containing stress fibers. Moesin and ezrin were highly enriched in other types of surface structures such as blebs, retraction fibers and some ruffles (not shown). The ezrin antibody also gave a general diffuse staining, even at low antibody concentrations (Fig. 4B,D).

The analysis of the subcellular distribution of ezrin and moesin in NRK cells was complicated by the thickness of these cells, so further examination was done using confo-cal fluorescence microscopy. Fig. 5 shows three optical sec-tions through the same NRK cell stained for actin and ezrin. Ezrin is clearly enriched in actin-containing surface struc-tures, such as small ruffles and the surface microvilli. The diffuse staining seen by conventional microscopy seems to originate from cytoplasmic staining, which is excluded from the nucleus. We also detected staining on or near the ven-tral surface of the cell. Cells stained for actin and moesin, and examined in various optical sections (Fig. 6), revealed colocalization of moesin with actin in both membrane ruf-fles and surface microvilli.

A rather different distribution of ezrin and moesin was found in interphase normal human fibroblasts, which do not have abundant surface structures (Fig. 7). Here, a patchy staining was seen for both ezrin and moesin. We often observed a diffuse F-actin staining with few or no stress fibers in those areas of cells where ezrin and moesin were concentrated. In early stages of mitosis these cells develop abundant microvilli that are enriched in ezrin, moesin and actin. This results in very bright staining of mitotic cells (not shown). During furrowing and cytokine-sis, ezrin and moesin-containing microvilli tend to con-centrate near the cleavage furrow to give an intense signal in this region of the cell as well as staining the ruffling membranes and microspikes in the polar regions of the spreading daughter cells (data not shown). Sato et al. (1991) reported that radixin is enriched in the cleavage furrow and polar regions of dividing cells. However, since it is now clear that the particular antibody used in this study crossreacts with moesin (Sato et al., 1992), the suggestion that radixin is present in these structures needs to be re-evaluated.

In earlier studies Tsukita et al. (1989) and Sato et al. (1991, 1992) reported that antibodies to radixin, subse-quently found to cross-react with ezrin and moesin (Sato et al., 1992), stained cell-to-cell adherens junctions and cell-to-substratum focal contacts. Using both conventional and their procedures to unmask antigenic sites in the same cell lines (Volberg et al., 1986; Avnur and Geiger, 1981), we did not detect ezrin or moesin in adherens junctions or in focal contacts.

Treatment of A431 cells with EGF leads to a progression of morphological changes, including the rapid and transient formation of microvilli and membrane ruffles. Ezrin is recruited into the induced surface structures and is tran-siently phosphorylated on tyrosine and serine in a time course that corresponds to their formation and disappear-ance (Bretscher, 1989). A431 cells are rich in ezrin and also contain a small but detectable amount of moesin (Fig. 3). Therefore, we compared the effect of EGF treatment on the subcellular distribution of the two proteins and examined whether moesin also becomes phosphorylated on tyrosine in response to EGF.

First, we compared the distributions of actin and ezrin (Fig. 8) and actin and moesin (Fig. 9) by confocal microscopy in A431 cells after EGF treatment. Examina-tion of three different optical sections through these cells revealed that, as expected, ezrin is highly enriched and colo-calizes with actin in the structures on the upper surface of the cells. Moesin is also highly enriched in all the actin-containing surface structures on the dorsal surface of the cell. No cytoplasmic staining was seen in A431 cells with the moesin antibody, whereas a general diffuse ezrin stain-ing was consistently seen.

Since ezrin and moesin are both recruited into EGF-induced cell surface structures, we determined whether moesin was also phosphorylated on tyrosine under these conditions. Replicate samples of A431 cells before and at various times after EGF stimulation were harvested and subjected to immunoprecipitation. The moesin-specific anti-bodies worked poorly for immunoprecipitations, so an anti-serum that immunoprecipitates both moesin and some ezrin was used, as well as antibodies specific for ezrin. A por-tion of each immunoprecipitate was run on SDS-PAGE and silver stained in order to verify that comparable amounts of protein were immunoprecipitated at each time point for each antibody. A second portion of each sample was run on SDS-PAGE, transferred to membranes and probed with antibodies to phosphotyrosine. The experiment was run in duplicate, on cells not treated, or pretreated, with the tyro-sine-phosphatase inhibitor phenylarsine oxide (PAO; Garcia-Morales et al., 1990), and the results are presented in Fig. 10. Moesin and/or ezrin were identified by co-elec-trophoresis of immunoprecipitates of A431 cells labeled to steady state with [³⁵S]methionine. Ezrin was rapidly and transiently phosphorylated on tyrosine, as reported previ-ously (Bretscher, 1989). Maximal tyrosine phosphorylation of ezrin occurred within 1 min after EGF addition, or at about 3 min in cells pretreated with PAO. PAO significantly increased the tyrosine phosphorylation stoichiometry of ezrin at 1 to 20 min. The moesin antiserum immunopre-cipitates about equal amounts of moesin and ezrin from A431 cells, as seen by silver staining and autoradiography of ³⁵S-labeled proteins. Thus the tyrosine phosphorylation of ezrin seen in the anti-moesin immunoprecipitates pro-vides a convenient internal standard for comparing the rel-ative efficiency of phosphorylation of the two proteins. In the anti-moesin immunoprecipitates, the time course of ezrin phosphorylation exactly parallels that seen in the anti-ezrin immunoprecipitates, while tyrosine phosphorylation of a comparable amount of moesin is undetectable even in the presence of PAO.

In an earlier report, we described the purification of two proteins, designated p81 ezrin and p77, from human pla-centa that were immunologically related to chicken brush border ezrin (Bretscher, 1983, 1989). Comparison of the partial protein sequence of chicken ezrin with protein and cDNA-predicted sequences from human p81 ezrin revealed that they were homologous and very closely related pro-teins (Gould et al., 1989). The placental p77 protein had very similar biochemical properties to p81 ezrin (Bretscher, 1989, 1991). Lankes et al. (1988) and Lankes and Furth-mayr (1991) purified a heparin-binding protein from bovine uterus and human HL60 cells, and isolated and sequenced the human cDNA encoding this protein, which is called moesin. Moesin is identical to our placental p77 protein (Lankes and Furthmayr, 1991). Ezrin and moesin show 74% sequence identity (Fig. 11). A third closely related protein, radixin, has also been identified. It shows 75% sequence identity to ezrin and 81% identity to moesin (Funayama et al., 1991). The amino-terminal regions of all three proteins show about 32% sequence identify with the membrane binding domain of human erythrocyte band 4.1. More dis-tant members of the protein family that share homology with the amino-terminal domain of band 4.1 have also been described recently (Fig. 11). These include the membrane-cytoskeletal linking protein talin (Rees et al., 1989) and two putative protein phosphotyrosine phosphatases (Yang and Tonks, 1991; Gu et al., 1991). Band 4.1 is a well charac-terized membrane-cytoskeletal linking protein that binds to glycophorin through its amino-terminal 30 kDa domain (Leto and Marchesi, 1984; Marchesi, 1985) and to the actin-spectrin complex through an 8 kDa domain located in the carboxyl-terminal half of the molecule (Correas et al., 1986), as shown schematically in Fig. 11. The finding that ezrin has a domain homologous to the membrane-binding domain of band 4.1 and is enriched in cell surface struc-tures has led to the hypothesis that it too might be a mem-brane-cytoskeletal linking protein (Gould et al., 1989). On the basis of sequence homology with ezrin, a similar func-tion was suggested for moesin (Lankes and Furthmayr, 1991) and radixin (Funayama et al., 1991).

The membrane-binding 30 kDa domain of band 4.1 is relatively resistant to digestion by α-chymotrypsin and can be recovered as a functional domain (Leto and Marchesi, 1984). Here we show that ezrin and moesin both have a 32 kDa α-chymotrypsin-resistant domain that makes up the amino-terminal regions of the proteins. This further strengthens the concept that these proteins have a mem-brane binding domain analogous to the corresponding domain of band 4.1. Importantly, the carboxyl-terminal halves, which have been suggested as binding to cytoskele-tal components, give rise to different proteolytic fragments, indicating that these parts of the proteins have different con-formations.

Prior to this study, ezrin was reported to be present in cell surface structures and radixin in adherens junctions, cell substratum focal contacts and the contractile ring (see studies cited above). The detailed localization of moesin in cultured cells has not been described. It is now apparent that antibodies to one protein can cross-react with another member of this protein family. For example, our original antibody to chicken ezrin reacts equally well with human ezrin and moesin (Bretscher, 1989), and all antibodies raised against radixin and used in earlier localization studies recognize all three members (Sato et al., 1992). Despite the similarities between these proteins we have been able to generate polyclonal antibodies that specifically recognize and distinguish between human p81 ezrin and human p77 moesin. Immunoblots with these specific antibodies did not detect any immuno-reactive material in liver (Berryman et al., 1993), the tissue from which radixin was originally puri-fied (Tsukita et al., 1989). Immunolocalization data on frozen sections of intestinal tissue with the ezrin and moesin antibodies reveals no colocalization with adherens junctions (Berryman et al., 1993), which are enriched in radixin (Tsukita et al., 1989). Furthermore, we did not detect stain-ing of focal contacts or adherens junctions with either anti-body. For these reasons it appears that our ezrin and moesin antibodies do not cross-react significantly with radixin.

Having specific antibodies against ezrin and moesin we were able to compare their relative abundance and subcel-lular localization. We found that the expression of moesin is highly variable, whereas ezrin is present in all cell lines investigated. Both proteins are enriched and colocalize with actin in cell surface structures, such as lamellipodia, microspikes, microvilli, retraction fibers and membrane ruf-fles. No association is seen with large actin cables, such as stress fibers. In cells with few surface structures, both pro-teins show patchy staining that often colocalizes with a dif-fuse actin staining in areas of the cell containing few stress fibers. This probably represents a fine meshwork of actin filaments, perhaps similar to that found in ruffling mem-branes. In addition, the ezrin antibody shows general cyto-plasmic staining. We have seen this with different cell lines and with different preparation methods and believe that it is specific staining. The similar distribution of ezrin and moesin in cell surface structures is consistent with our find-ing that ezrin binds moesin in vitro and in vivo (Gary et al., 1992).

Interest in ezrin was originally generated by the finding that it becomes phosphorylated on tyrosine in A431 cells treated with epidermal growth factor (Cooper et al., 1982). Subsequently, a close temporal correlation was found between the EGF-induced appearance of cell surface struc-tures enriched in ezrin and the phosphorylation of the pro-tein (Bretscher, 1989). Since A431 cells also contain moesin, it was of interest to follow moesin’s localization and phosphorylation following EGF treatment. Like ezrin, moesin is also present in the induced actin-containing sur-face structures, which are therefore enriched in actin, ezrin, moesin and the activated EGF receptor (Carpentier et al., 1987). Despite their similar localization and high sequence identity, we found that ezrin becomes phosphorylated on tyrosine, as determined using phosphotyrosine antibodies, yet moesin does not. This is surprising, as one of the two major sites of ezrin tyrosine phosphorylation in this system is tyrosine 145 in the 32 kDa domain (Krieg and Hunter, 1992) and this tyrosine residue is conserved in moesin. We do not yet know the basis of this discrimination; it could be that tyrosine 145 in moesin is not accessible to the kinase that phosphorylates ezrin, even though ezrin and moesin are present in the same structures in this system. This result suggests that if tyrosine phosphorylation of either protein in some way contributes to the assembly or disassembly of EGF-induced surface structures, it is the modification of ezrin but not of moesin that may be important.

In summary, we have shown that ezrin and moesin are present in a number of cell lines in actin-containing surface structures. Combining our data with those of others (Tsukita et al., 1989; Sato et al., 1992) leads to the conclusion that radixin is an authentic component of adherens junctions and focal contacts; it is not yet clear whether it might also be present in actin-containing surface structures. The reported localization of radixin in the cleavage furrow (Sato et al., 1991) also needs to be re-evaluated in light of the results described here. The similar localizations of ezrin and moesin, yet their different proteolytic digestion patterns, kinase substrate specificity and tissue distributions (Berry-man et al., 1993), imply that these two related proteins have distinct but related functions. Similarly, the distinct local-izations reported and inferred for radixin, point to a dis-crete function for this protein. It seems that ezrin, moesin and radixin are not simply redundant homologs, but pro-teins with specific, but related, functions. The challenge now is to elucidate these functions.

We are very grateful to Janet Krizek for help in preparing the antibodies and to the NIH for support (GM36652).