Biochemical characterization of ezrin-actin interaction.

The highly related actin isoforms are thought to have different functions. We recently demonstrated a polarized distribution of actin isoforms in gastric parietal cells and association of gastric ezrin with the cytoplasmic beta-actin isoform (Yao, X., Chaponnier, C., Gabbiani, G., and Forte, J. G. (1995) Mol. Biol. Cell. 6, 541-557). Here we used ultrastructural immunocytochemistry to verify that beta-actin is located within canalicular microvilli and the apical cortex of parietal cells, similar to the localization reported for ezrin. Furthermore, we tested whether ezrin binds preferentially to cytoplasmic beta-actin compared with the skeletal muscle alpha-actin isoform. Purified cytoplasmic beta-actin (from erythrocytes) and skeletal alpha-actin were assembled with gastric ezrin. Co-sedimentation experiments showed that gastric ezrin selectively co-pelleted with the beta-actin isoform and only very poorly with alpha-actin. Binding of erythrocytic beta-actin to ezrin is saturable with a molar ratio of approximately 1:10 (ezrin:actin) and a dissociation constant approximately 4.6 x 10(-8) M. In addition, ezrin promoted pyrene-labeled actin assembly, with predominant effects on filament elongation and a distinct preference for beta-actin compared with alpha-actin. Given these isoform-selective associations, we speculate that actin isoforms might segregate into different functional domains and exert specificity by interacting with isoform-orientated binding proteins.

Although actin is a highly conserved protein, several distinct tissue-specific isoforms exist. The actin isoforms are encoded by separate genes and differ by less than 10% in amino acid sequence (1) and are generally believed to exert different functions. For example, profilin has different affinities for each of the cytoplasmic actin isoforms, ␤-actin and ␥-actin, and for sarcomeric actin (2,3). We showed recently that there is a polarized distribution of cytoplasmic ␤and ␥-actin isoforms in gastric parietal cells (4), consistent with some difference in functional activity and/or preferential interaction with localized actin-binding proteins. The ␤-actin isoform is concentrated near the apical membrane of all gastric epithelial cells, including the apical secretory canalicular membrane of parietal cells, while the ␥-actin isoform is primarily distributed toward the basolateral surface, with minor deposition in the region of the secretory canaliculi of parietal cells.
Ezrin is an actin-binding protein of the ezrin/radixin/moesin (ERM) family of cytoskeleton-membrane linker proteins (5). Within the gastric epithelium, ezrin has been localized exclusively to the apical canalicular membrane of parietal cells (6,7). Because of its cytolocalization and stimulation-dependent phosphorylation, an implied role for ezrin has been suggested in the apical surface membrane remodeling associated with parietal cell activation via the protein kinase A pathway (6,8).
Phosphorylation of ezrin has also been associated with surface membrane remodeling of A431 cells stimulated by epidermal growth factor (9), although activation in this case was via protein tyrosine kinase. Our previous studies showed that the F-actin which co-localized with ezrin in parietal cells was primarily comprised of the ␤-actin isoform, and that the ␤-actin isoform was preferentially co-immunoprecipitated with ezrin from extracts of gastric membranes (4). Recently, Shuster and Herman (10) reported that ezrin, contained within lysates of retinal pericytes, bound to immobilized matrices of cytoplasmic ␤-actin, but not to the skeletal ␣-actin isoform. These authors further demonstrated a co-localization of antibodies against the ␤-actin isoform and ezrin in leading lamellae of motile cells. Thus a body of evidence suggests interaction of ezrin with the actin cytoskeleton may be specific for the ␤-actin isoform, but much controversy remains regarding the nature of the interaction and whether it is direct or via intermediary proteins.
The purpose of the present experiments was to test if gastric ezrin exerts isoform-specific association with cytoplasmic ␤-actin in relatively simple reconstituted systems. Accordingly, we separately assembled skeletal ␣-actin and cytoplasmic ␤-actin in the presence of gastric ezrin and subjected them to centrifugation. Gastric ezrin selectively co-sedimented with cytoplasmic ␤-actin compared with skeletal ␣-actin, which is consistent with an earlier observation that showed a weak interaction between skeletal ␣-actin and intestinal ezrin (11), and supports the notion of isoform specificity. Further characterization of ezrin-actin interaction by using erythrocytic ␤-actin revealed that gastric ezrin binds to actin in a saturable manner with a molar ratio of about 1:10 (ezrin:actin) and a dissociation constant (K d ) ϳ4.6 ϫ 10 Ϫ8 M for the ezrin-actin binding relationship.

MATERIALS AND METHODS
Purification of Actin Isoforms-An acetone powder was prepared from rabbit muscle using the method of Pardee and Spudich (12). Typically, we obtained 20 mg of pure skeletal actin from 1 g of acetone powder. Cytoplasmic ␤-actin was purified from bovine red blood cells as described by Puszkin et al. (13). Briefly, bovine red blood cells (RBC) 1 were osmotically hemolyzed, and an acetone powder was made from the harvested RBC ghosts. The resulting acetone powder was then extracted twice using 15 ml buffer A (5 mM Tris, pH 8.0, 0.2 mM CaCl 2 , 0.2 mM ATP, 0.2 mM DTT, 0.02% sodium azide)/mg of acetone powder for 3 h at 4°C. After clarification at 27,000 ϫ g for 1 h at 4°C, the extract was fractionated over a DE52 anion exchange column pre-equilibrated with buffer A, and developed with a linear gradient of 50 mM to 1 M NaCl. ␤-Actin eluted from the column between 0.15 and 0.25 M NaCl, identified by dot-blot. These fractions were pooled, concentrated to 4 ml, and applied to a 2 ϫ 40-cm Sephadex G-150 column equilibrated with buffer A. The peak actin fractions were then pooled in buffer A containing 2 * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  mM MgCl 2 and 100 mM KCl at room temperature in order to polymerize native actin. After pelleting filamentous actin at 100,000 ϫ g, F-actin was depolymerized in buffer A and rechromatographed on G-150 Sephadex prior to storage at Ϫ80°C. The yield of ␤-actin was relatively poor, thus we switched from human to bovine RBC for more voluminous starting material; typically, we obtained about 4 mg of pure ␤-actin from 4 liters of whole bovine blood. SDS-gel electrophoresis demonstrated that each actin isoform was greater than 98% pure, both at the level of actin isoforms (two-dimensional electrophoresis) and with respect to other contaminating proteins.
Purification of Ezrin-Gastric ezrin was purified from rabbit gastric mucosal homogenates using a procedure described by Bretscher (14) with modification. Gastric ezrin was extracted with buffer containing 0.5% Triton X-100, 300 mM NaCl, 1 mM EGTA, 0.25 mM phenylmethylsulfonyl fluoride, 10 M E64, 1 mM benzamide, 20 mM Tris-Cl, pH 7.4, with stirring on ice for 60 min to maximize the extraction of ezrin, and centrifuged at 20,000 ϫ g for 20 min. The supernatant was brought to 40% (NH 4 ) 2 SO 4 and the precipitate removed by centrifugation at 12,000 ϫ g for 20 min. The resulting supernatant was precipitated with 75% (NH 4 ) 2 SO 4 . The precipitates were solubilized in 10 mM imidazole, pH 6.7, 1 mM DTT and dialyzed against the same buffer overnight. After clarification at 30,000 ϫ g for 20 min, the supernatant was applied to a hydroxyapatite column pre-equilibrated with 100 mM potassium phosphate, pH 7.0, and developed with a 40-ml linear gradient of 100 -800 mM potassium phosphate. Fractions containing ezrin were identified by dot-immunoblot and pooled for dialysis for 3 h against 10 mM bis-Trispropane (BTP buffer), pH 6.7, 20 mM NaCl, 1 mM DTT. The material was then centrifuged and applied onto a Q-Sepharose column and developed with 40-ml linear gradient from 20 mM to 1 M NaCl in BTP buffer. Fractions rich in ezrin were pooled, dialyzed against 20 mM MES, pH 6.7 (MES buffer), 20 mM NaCl, 1 mM DTT, and applied to an S-Sepharose column and again developed with a 40-ml linear 20 mM to 1 M NaCl gradient in MES buffer. Homogeneous ezrin eluted at ϳ650 mM NaCl. The ezrin-enriched fractions were desalted in microconcentration tubes (Amicon, Beverly, MA) and stored in MES buffer at 0 -4°C until use. Densitometric analysis of the sample run on 9% SDS-acrylamide gel demonstrated that ezrin was greater than 95% pure with respect to other contaminating proteins and degradation products of ezrin.
Immunoelectron Microscopy-Gastric glands were isolated from rabbit stomach as described previously (15). The isolated glands were either maintained in a resting (nonsecreting) state by treatment with 10 Ϫ4 M cimetidine, or they were maximally stimulated to secrete acid by treatment with 10 Ϫ4 M histamine plus 5 ϫ 10 Ϫ5 M isobutylmethylxanthine. For fixation, the glands were harvested by a brief centrifugation, fixed in 4% paraformaldehyde plus 0.05% glutaraldehyde, and infiltrated according to a standard protocol described by Berryman et al. (16) using LR-gold (Ted Pella, Inc., Redding, CA). Thin sections of the LR-gold-embedded glands were picked up on Formvar-coated nickel grids. Grids were blocked in phosphate-buffered saline ϩ 0.05% Tween 20 ϩ 1% bovine serum albumin ϩ 0.1% cold water fish gelatin ϩ 0.02% NaN 3 for 10 min and then incubated with affinity-purified anti-␤-actin antibody for 1 h and washed with four successive washes (phosphatebuffered saline ϩ 0.05% Tween 20) of 5 min each. Antibody against the cytoplasmic ␤-actin isoform was a gift from Dr. Christine Chaponnier, with characteristics as described by Yao et al. (4). Grids were then incubated for 45 min with goat anti-rabbit IgG conjugated with 5-nm colloidal gold. Grids were rinsed in wash buffer (3 ϫ 5 min), phosphatebuffered saline (2 ϫ 5 min), H 2 O (2 ϫ 5 min) and then stained with 2% aqueous uranyl acetate for 90 s. In the case of low power images, immunostained grids were incubated with a silver enhancement kit for approximately 4 min (Ted Pella, Inc.) followed by the uranyl acetate staining. A number of controls were routinely performed for immunolabeling: omission of the primary antibody and incubation with preimmune serum.
Co-sedimentation of Ezrin with Actin Filaments-Stored actin samples were thawed and centrifuged at 312,000 ϫ g (90,000 rpm, TLA 100 rotor; Beckman Instruments) for 40 min to remove any polymerized or aggregated actin. Aliquots of monomeric actin isoforms were incubated for 2 h at 23, with or without ezrin or myosin S1 fragment, in polymerizing buffer (5 mM Tris, pH 7.5, 0.5 mM ATP, 2 mM MgCl 2 , 100 mM KCl, and 0.2 mM DTT) to promote polymerization. The filaments were then sedimented by centrifugation at 312,000 ϫ g for 40 min. Myosin S1 subfragment, a generous gift from Roger Cooke (University of California, San Francisco), was used as a positive control in co-sedimentation experiments. For visualization and quantitation of actin and its binding proteins, pellets and supernatants were solubilized in SDS gel sample buffer and subjected to electrophoresis. Some gels were stained with Coomassie Blue and dried between sheets of cellulose for visualization and quantitative scanning. Other were transblotted to nitrocellulose and probed with ezrin antibodies.
Pyrene-labeled Actin Assembly-Pyrene-labeled actin was prepared by modifying the method of Kouyama and Mihashi (17). Briefly, 100 mg of polymerized actin was pelleted and homogenized in buffer P (10 mM Hepes, pH 7.4, 2 mM MgCl 2 , 100 mM KCl, 0.5 mM ATP), 5.6 mg of N-(1-pyrenyl)iodoacetamide (Molecular Probes, Inc., Eugene, OR) was added with agitation, and the solution was incubated at 4°C in the dark overnight. The labeled actin was dialyzed against three changes of buffer A for 24 h in order to depolymerize the F-actin. Any resulting F-actin and/or precipitated N-(1-pyrenyl)iodoacetamide were removed by centrifugation at 100,000 ϫ g for 2 h. Aliquots of pyrene-labeled G-actin were quickly frozen in liquid nitrogen and stored at Ϫ80°C until use.
For assembly experiments, actin was used at a concentration of 5 M. Because some polymerization of actin occurs during the freezing of samples, thawed actin was spun at 312,000 ϫ g for 40 min before initiating assembly. Polymerization was initiated at the following final conditions: 5 mM Tris, pH 7.5, 0.5 mM ATP, 2 mM MgCl 2 , 10 mM KCl, and 0.2 mM DTT. Fluorescence was monitored continuously: excitation wavelength ϭ 355 nm; emission wavelength ϭ 407 nm. Purified yeast cofilin was used as a control according to Moon et al. (18).

Immunogold Labeling of ␤-Actin
Isoform-Previous studies revealed a polarized distribution of ezrin in the apical and canalicular plasma membranes of gastric parietal cells (6) and that ezrin was spatially co-distributed with F-actin filaments, primarily composed of the ␤-actin isoform (4). In the present experiments we sought to more closely define the location of the ␤-actin isoform within gastric glandular cells using immunogold labeling at the level of the electron microscope. A micrograph from resting, or nonsecreting, gastric glands is shown in Thin sections were stained with affinity-purified anti-␤-actin antibody followed by colloidal gold-conjugated (5 nm) goat anti-rabbit IgG. Immunostained grids were incubated with silver enhancer solution followed by post-staining with uranyl acetate. The micrograph shows a cross-section through a gastric gland including several parietal cells (PC) and mucous neck cells (MNC) surrounding the gland lumen. There is a dense distribution of enhanced gold particles along the apical surfaces (Ap) of all epithelial cells, especially within parietal cells along the intracellular canaliculi (IC), which are invaginations of the apical surface membrane coursing throughout the parietal cell. Gold particles are also seen along the basolateral region (Bl) of parietal cells, with a somewhat lower density than along canaliculi. Large mitochondria throughout the cytoplasm are characteristic of parietal cells; numerous cytoplasmic tubulovesicles and canalicular microvilli, also characteristic of parietal cells, are not easily visualized at this magnification. Bar marker is 2 m. cells, mucous neck cells, and chief cells) and along the canalicular surfaces within parietal cells. The secretory canaliculi are invaginations of the apical surface membrane coursing throughout the parietal cell; at the level of magnification shown in Fig. 1, the collapsed canaliculi are easily visualized by the dense trail of enhanced gold particles. Gold particles are also clearly seen along the basolateral region of parietal cells, with a somewhat lower density than along the canaliculi. With the exception of discrete staining along their apical borders, there was virtually no gold staining within chief cells or at their basolateral surfaces. For stimulated gastric glands (Fig. 2), enhanced immunogold staining of ␤-actin can readily be seen along the dilated canaliculi of parietal cells and the apical surfaces of neighboring epithelial cells. Mitochondria within the stimulated parietal cells appeared to be highly concentrated in the cytoplasm, due to the fusion of tubulovesicles with the apical canalicular membrane. At the higher resolution shown in Fig. 3, specific gold particle labeling of ␤-actin can clearly be localized to apical microvilli and the region of the terminal web just beneath the apical canalicular surface of parietal cells. Very few gold particles extend deep into parietal cell cytoplasm, although lateral folds and basal membranes show distinct gold labeling.
Gastric Ezrin Is Preferentially Co-sedimented with Cytoplasmic ␤-Actin-The terminal step in our procedure for purifying ezrin from gastric homogenates is represented in Fig. 4. Two major peaks of ezrin were eluted from the S-Sepharose, one in the range of 450 -550 mM NaCl with several ezrin breakdown products in the 40 -55 kDa range (identified by blot) and a second peak eluting at 625-675 mM NaCl containing a single 80-kDa ezrin peak and some low molecular weight peptides migrating near the dye front. This second highly purified ezrin preparation was used in our tests of ezrin-actin interaction.
To determine whether gastric ezrin stably binds to actin filaments in an isoform-specific manner, we assayed the ability of ezrin to co-sediment with skeletal muscle ␣-actin and RBC cytoplasmic ␤-actin. Gastric ezrin was incubated with the respective actin isoforms under polymerizing conditions for 2 h, and the filaments were pelleted by centrifugation at 312,000 ϫ g for 40 min. Actin filaments longer than 10 subunits in length will sediment under these conditions (19). The Coomassie Blue-FIG. 2. ␤-Actin is primarily localized to apical canalicular surfaces of secreting parietal cells. Gastric glands stimulated with histamine plus isobutylmethylxanthine were prepared and immunostained as described in Fig. 1. Enhanced gold particles are clearly seen along the dilated intracellular canaliculi (IC) of parietal cells and near the apical surfaces (Ap) of neighboring chief cells (CC). There is somewhat less deposition of gold particles at basolateral surfaces (Bl) of parietal cells and virtually no basolateral staining of chief cells. In these stimulated parietal cells, the mitochondria are highly concentrated in the cytoplasm due to the fusion of tubulovesicles to the canalicular membrane, and the canalicular spaces are extended and readily seen. The bar marker is 2 m.

FIG. 3. Magnified view shows that ␤-actin is heavily localized to apical and canalicular microvilli of parietal cells.
Secreting gastric glands were processed as described in Fig. 1. Thin sections were stained with anti-␤-actin antibody followed by colloidal gold-conjugated goat anti-rabbit IgG (5 nm); no silver enhancement was applied. The gold particles are primarily distributed to the apical microvilli (mv) and to the cortical region (cor) subadjacent to the apical membrane. Some gold particles are also seen near lateral membrane folds (lat). Bar marker is 0.5 m.
FIG. 4. Chromatographic purification of gastric ezrin on S-Sepharose. Ezrin was extracted from gastric mucosal homogenates, precipitated by 75% (NH 4 ) 2 SO 4 , and partially purified by successive chromatography on hydroxyapatite and Q-Sepharose, as described under "Materials and Methods." The ezrin peak (identified by dot blot) eluting from Q-Sepharose was applied as starting material (SM) to a column of S-Sepharose and eluted with a 20 mM to 1 M NaCl gradient (tube numbers indicated above). Eluting fractions were subjected to SDS-polyacrylamide gel electrophoresis and stained by Coomassie Blue. One peak, eluting in the range of 450 -550 mM NaCl (tubes 18 -22), was rich in ezrin but also had several additional peptides, including some ezrin hydrolytic products (identified by separate Western blot). The second ezrin peak eluting in the range of 625-675 mM NaCl (tubes 26 -28) was virtually free of contaminating peptides except for some small molecular weight peptides migrating near the dye front. This second highly purified peak of ezrin was used for subsequent binding studies. Molecular weight standards (mw) are shown to the left. stained gel in Fig. 5A shows that gastric ezrin did not appear in the pellet when incubated without actin and pelleted very poorly when incubated with ␣-actin. However, ezrin obviously appeared in the pellet when ␤-actin was polymerized (Fig. 5B). The S1 tryptic fragment of myosin II, a known actin-binding protein that decorates actin filaments side-wise, was used as a control. The myosin subfragment sedimented with either the skeletal ␣-actin or the RBC ␤-actin without isoform specificity and with an apparent molar ratio of 1:1 (myosin/actin). It is also apparent that neither gastric ezrin nor the myosin subfragment affected the amount of either ␣or ␤-actin that sedimented.
Gastric Ezrin Binds Stably to Cytoplasmic ␤-Actin Filaments in a Saturable Manner-To probe the stoichiometry of ezrinactin association, we carried out separate sets of co-sedimentation experiments: one case in which the ezrin concentration was held constant while varying the ␤-actin concentration; in the other case the ␤-actin concentration was fixed while ezrin concentration was varied. Coomassie Blue-stained gels representing these respective experiments are shown in Fig. 6, A and B. As actin was increased more ezrin was removed from the supernatant; when the molar ratio of actin:ezrin was 20:1, or greater, virtually no ezrin remained in the supernatant (Fig.  6A). When ␤-actin was held constant at 5 M, the proportion of ezrin that sedimented with the ␤-actin pellet deceased as ezrin concentration was increased, consistent with a saturation isotherm for ezrin binding (Figs. 6B). At low concentration virtually all ezrin is removed to the ␤-actin pellet; as the ratio of actin:ezrin falls below 10:1, ezrin content in the supernatant increases. These data also support the idea that ezrin binds to actin polymers. Densitometric data for bound and total ezrin are shown in Fig. 7 for four separate co-sedimentation experiments where ezrin was varied between 0.05 and 2.0 M and F-actin was polymerized at a constant 5 M ␤-actin concentration. Linear transformation of the data, plotting ezrin bound/ ezrin free against ezrin bound (Scatchard plot), provided a K d for ezrin binding to ␤-actin of ϳ4.6 ϫ 10 Ϫ8 M. The calculated stoichiometric molar ratio of 0.084 (ezrin/actin) suggested that about 1 ezrin was bound per 10 actin molecules.
Gastric Ezrin Affects Pyrene-labeled Actin Polymerization in Vitro-To evaluate the quality of our purified actin isoforms from bovine RBCs and rabbit skeletal muscle, the time course of polymerization for pyrene-labeled actin was monitored, as described by Cooper et al. (20). Pyrene fluorescence increases when a pyrene-labeled actin subunit is incorporated into a polymerizing filament. Both skeletal ␣-actin and cytoplasmic ␤-actin isoforms readily polymerized when KCl and MgCl 2 were included as demonstrated by the respective standard curves in Fig. 8, A and B. To validate the properties of our pyrene-labeled actin, we tested the influence of yeast cofilin on pyrene-labeled actin (18). Yeast cofilin increased the rate of assembly of both skeletal ␣-actin and cytoplasmic ␤-actin isoforms. The effects of cofilin were most prominent in the late stages of polymerization, after the lag phase (nucleation phase), consistent with the report of Moon et al. (18).
When gastric ezrin was included with the pyrene-labeled skeletal ␣-actin, the curve shifted slightly to the left in the late FIG. 5. Gastric ezrin is preferentially co-pelleted with the ␤-actin isoform, but not the skeletal ␣-actin isoform. Samples of ezrin alone, actin alone, actin plus ezrin, and actin plus myosin S1 subfragment were incubated for 2 h in polymerization buffer and centrifuged as described under "Materials and Methods." Equal volumes of supernatant (S) and pellet (P) fractions were resolved by electrophoresis and proteins visualized by Coomassie blue staining. A, skeletal ␣-actin (5 M) was incubated with 1 M ezrin or 5 M myosin subfragment, as indicated. B, RBC ␤-actin (5 M) was incubated with 1 M ezrin or 5 M myosin subfragment, as indicated. In the myosin-containing lanes, the 95-kDa band is the S1 fragment, while the lower molecular mass band (ϳ70 kDa) is probably derived from the further tryptic degradation of the S1 fragment. Ezrin band densities were measured in the supernatants and pellet, and the ezrin bound was calculated as the fractional amount of ezrin in the pellet. Data are the mean Ϯ S.E. from four separate experiments, two derived from Coomassie Blue-stained protein and two from Western blots probed for ezrin. stage of polymerization, but not in the initial nucleation stage, suggesting that ezrin might bind to the actin filaments or possibly promote the assembly (Fig. 8A). An increased molar ratio of ezrin/actin further shifted the curve to the left.
The addition of ezrin also promoted the assembly of pyrenelabeled ␤-actin, with effects that were more pronounced than for skeletal ␣-actin (Fig. 8B). Ezrin reduced the time for assembly to achieve a steady state in a concentration-dependent manner, e.g. at a molar ratio of 1:10 (ezrin:actin) the time to achieve steady state decreased from ϳ60 to ϳ30 min. The profile of ezrin on RBC ␤-actin assembly is somewhat like that of cofilin, that is, ezrin does not seem to modulate the lag associated with the nucleation phase while it promotes the filament elongation phase as evidenced by increased rate of assembly after the initial lag phase. Thus the pyrene-labeled actin assembly assay indicated that ezrin might directly interact with actin isoforms, with a distinct preference for RBC ␤-actin compared with skeletal ␣-actin.

DISCUSSION
Two different actin isoforms have been identified within parietal cells, cytoplasmic ␤-actin and cytoplasmic ␥-actin, which are polarized to the apical and basolateral membranes, respectively (4). In addition these studies demonstrated a preferential interaction between ezrin and the ␤-actin isoform extracted from native parietal cells. In the present studies, we extended our earlier finding by using immunoelectron micros-copy to localize ␤-actin to the apical microvilli and in vitro reconstitution of ezrin-␤-actin interaction using purified gastric ezrin and erythrocytic ␤-actin.
Using immunoelectron microscopy, we localized ␤-actin to the canalicular surface and apical microvilli where gastric ezrin is enriched (6). Despite the dramatic elongation of apical microvilli and dilation of the canalicular lumina of parietal cells during acid secretion, the immunogold labeling of ␤-actin did not reveal obvious stimulation-mediated redistribution of ␤-actin. In fact, a recent report on the state of actin in resting and stimulated gastric glands using the DNase I assay did not reveal any significant change in either filamentous or monomeric actin pool (21), suggesting that stimulation-mediated elongation of microvilli could be due mainly to an invagination of plasma membrane resulting from the fusion of tubulovesicular membrane. It is possible that the stability and integrity of microfilaments are required for providing a structural support for the dynamic extension of apical plasma membrane and growth of microvilli.
The three-dimensional structure of cytoplasmic ␤-actin was recently solved in complex with profilin (22). Although its primary sequence is generally similar to skeletal ␣-actin, cytoplasmic ␤-actin displays several structural differences. These include: the N-terminal conformation of ␤-actin bearing a turn rather than the helical structure in skeletal ␣-actin, distinct rotational differences within subdomains of the isoforms, and differences in side chain orientation at residues 38 -52. Physiological interpretation of these structural differences is still under debate (23), but functional distinctions for the interaction of actin isoforms with actin binding proteins have been reported. Larsson and Lindberg (2) showed that cytoplasmic ␤-actin and ␥-actin have higher affinity to bind profilin (K d ϳ10 Ϫ8 M) than that of sarcomeric actin (K d ϳ4 ϫ 10 Ϫ6 M) and that profilin interaction with the non-muscle isoforms was regulated by Mg 2ϩ . Rozycki et al. (3) further demonstrated that profilin preferentially binds to cytoplasmic ␤-actin compared with ␥-actin. The increased ratio of ␤to ␥-actin isoforms during the course of co-purifying ezrin and actin by immunoprecipitation hinted that ␤-actin filaments might preferentially bind to ezrin (4). Bretscher observed (11) that ezrin co-sedimented with filamentous skeletal muscle ␣-actin in vitro only at low ionic strength. Our studies with the comparison of two different actin isoforms clearly show that gastric ezrin preferentially binds to cytoplasmic ␤-actin while ezrin binds poorly to the skeletal ␣-actin.
A recent series of studies has provided further information concerning the structural and functional relevance of closely related ERM protein family members. Martin et al. (24) showed that the N-terminal domain of ezrin inhibits the functional activity of cell surface protrusion exerted by the C-terminal domain. These authors further mapped the inhibitory domain to the first 115 N-terminal residues. A similar interaction between the N-and C-terminal domains was also observed by Henry et al. (25) in their study of the functional relevance of radixin. Gary and Bretscher (26,27) carried out a meticulous characterization of homotypic and heterotypic interaction among the ERM protein family members, specifically ezrin and moesin. They established conditions under which there is heterotypic interaction between the N-terminal domain (amino acids 1-296) and C-terminal domain (amino acids 479 -585) to form dimers and possibly oligomers. These terminal "interactive domains" were clearly demonstrated in an extensive set of test probings using fusion protein constructs of the respective N-and C-terminal interactive domains. However, for the fulllength fusion protein, or for isolated native ezrin, they interpreted their results to suggest that the monomer exists in a form in which the C-terminal interactive domain is masked by a folding that also partially obscures the N-terminal interactive domain. When full-length ezrin was denatured the C-terminal interactive domain became accessible, but the N-terminal domain was inactivated by denaturation. Because the actin binding domain of ezrin is near the C terminus (28), and the apparent masking of the C-terminal interactive domain in the full-length monomer, Gary and Bretscher offered these data as a basis to explain the lack of ezrin association with ␣-actin filaments in vitro (11). It is possible that modification (e.g. phosphorylation) and accessory proteins might perturb this masking process and expose the F-actin binding site in the C-terminal domain. In fact, purified ezrin from gastric mucosa contains multiple spots as resolved by two-dimensional electrophoresis, which suggests that native gastric ezrin might contain a pool of phosphoezrin and/or multiple isoforms. Since there are three consensus phosphorylation sites for protein kinase A in ezrin (Ser 65 , Thr 213 , and Thr 332 ), and protein kinase A-mediated phosphorylation has been implicated in hormone-stimulated acid secretion (8), it is conceivable that the phosphorylation might alter the intramolecular masking effect and expose the C-terminal domain for actin binding. In fact, phosphorylation of ezrin by epidermal growth factor receptor tyrosine kinase triggered the dimerization, although the functional activity of this dimer is unknown at present (27). Alternatively, ezrin isoforms might form functional oligomers that might exert F-actin-binding action.
Shuster and Herman (10) recently reported that ezrin preferentially binds to an affinity column made of filamentous erythrocytic ␤-actin, but not to a skeletal ␣-actin. Because these authors were unable to reconstitute erythrocytic ␤-actinezrin interaction in vitro, they concluded that their observed interaction was indirect, and suggested that a 73-kDa polypeptide served to link the interaction between ezrin and filaments of ␤-actin isoform, although other candidates were also possible. Data presented here would argue against a requirement for the 73-kDa component, since there was no such peptide present in our reaction mixtures. On the other hand, it is not possible to rule out the participation of a low molecular weight component in the ezrin-␤-actin filament interaction.
There are some apparent contradictions concerning the nature of ezrin-actin interaction. Based on the displacement of ezrin from actin filaments by cytochalasin D, Shuster and Herman (10) proposed that ezrin binds to the barbed ends of actin filaments. Using a blot overlay assay, Pestonjamasp et al. (29) found that immobilized ezrin bound to actin filaments and the binding was minimized by myosin S1 subfragment but not by gelsolin or capping protein, suggesting that binding occurs at the filament sides and not at barbed ends. Moreover, the parallel localizations of ezrin and F-actin at the light and EM levels (6,7,11) are consistent with side binding. Our studies of ezrin on pyrene-labeled actin assembly in vitro also favors the idea that ezrin interacts with actin filaments along the side since ezrin dose-dependently promotes actin assembly, which is typically seen for a side-binding polypeptide, myosin S1 fragment (30,17). Recently, Turunen et al. (28) revealed that the region 558 -578 in ezrin, moesin, and radixin shows high sequence homology with the actin binding site of CapZ ␤-subunit (31). Despite the fact that both radixin and CapZ ␤-subunit bind to the barbed ends of actin filaments (31,32), it is not clear whether the homologous region which exerts actin binding for ezrin is also responsible for the end-binding in the case of CapZ and radixin. Recent studies showed that EF1␣, a transcription factor, has typical barbed end capping activity in addition to the known actin filament side-binding property (33). Thus, we cannot rule out the possibility that ezrin might interact with the ends of actin filaments at present.
In summary, our immunoelectron microscopic data show cytoplasmic ␤-actin is indeed located to the apical microvilli and terminal web of parietal cells, in a pattern identical to what has been reported for ezrin (4,6,7). We have verified that gastric ezrin is preferentially associated with cytoplasmic ␤-actin filaments in vitro, compared with skeletal ␣-actin, and that gastric ezrin binds to ␤-actin filaments in a saturable manner. Finally, based on the pyrene-labeled actin assembly data, we speculate that ezrin might modulate the elongation phase of the actin assembly. Because of the isopreferential associations demonstrated in this study and morphological separations seen in several systems (34 -36), we suggest that actin isoforms might segregate into different functional domains and exert their specificities by interacting with isoform-orientated actinbinding proteins.