Ezrin is a cyclic AMP‐dependent protein kinase anchoring protein

Daniel T. Dransfield, Alvin J. Bradford, Jenetta Smith, Marianne Martin, Christian Roy, Paul H. Mangeat, James R. Goldenring

Author Affiliations

  1. Daniel T. Dransfield1,
  2. Alvin J. Bradford1,
  3. Jenetta Smith1,
  4. Marianne Martin2,
  5. Christian Roy2,
  6. Paul H. Mangeat2 and
  7. James R. Goldenring1
  1. 1 Institute of Molecular Medicine and Genetics, Departments of Medicine, Surgery and Cellular Biology & Anatomy, Medical College of Georgia and the Augusta Veterans Affairs Medical Center, Augusta, GA, 30912, USA
  2. 2 CNRS UMR 5539, University of Montpellier, 34095, Montpellier, Cedex 5, France


cAMP‐dependent protein kinase (A‐kinase) anchoring proteins (AKAPs) are responsible for the subcellular sequestration of the type II A‐kinase. Previously, we identified a 78 kDa AKAP which was enriched in gastric parietal cells. We have now purified the 78 kDa AKAP to homogeneity from gastric fundic mucosal supernates using type II A‐kinase regulatory subunit (RII) affinity chromatography. The purified 78 kDa AKAP was recognized by monoclonal antibodies against ezrin, the canalicular actin‐associated protein. Recombinant ezrin produced in either Sf9 cells or bacteria also bound RII. Recombinant radixin and moesin, ezrin‐related proteins, also bound RII in blot overlay. Analysis of recombinant truncations of ezrin mapped the RII binding site to a region between amino acids 373 and 439. This region contained a 14‐amino‐acid amphipathic α‐helical putative RII binding region. A synthetic peptide containing the amphipathic helical region (ezrin409–438) blocked RII binding to ezrin, but a peptide with a leucine to proline substitution at amino acid 421 failed to inhibit RII binding. In mouse fundic mucosa, RIIimmunoreactivity redistributed from a predominantly cytosolic location in resting parietal cells, to a canalicular pattern in mucosa from animals stimulated with gastrin. These results demonstrate that ezrin is a major AKAP in gastric parietal cells and may function to tether type II A‐kinase to a region near the secretory canaliculus.


Epithelial cells, by definition, segregate the components of their apical and basolateral domains in order to express and maintain specialized functions at disparate surface environments. The proper segregation of both second messenger systems and specialized membrane–cytoskeleton complexes contributes, among other factors, to epithelial integrity and function. In the gastric parietal cell, histamine stimulates an increase in cAMP, leading to a marked morphological rearrangement, which elicits the fusion of H/K‐ATPase‐containing tubulovesicles with the intracellular canalicular target membrane (Forte et al., 1977; Soroka et al., 1993). These structural changes represent one of the largest reversible membrane–cytoskeletal rearrangements in eukaryotic cells. Several investigations have demonstrated the histamine‐induced phosphorylation of both cytosolic and particulate substrates in gastric parietal cells (Chew and Brown, 1987; Urushidani et al., 1987, 1989; Oddsdottir et al., 1988). The only well‐characterized histamine‐stimulated phosphoprotein is ezrin, which co‐localizes with both F‐actin filaments and the intracellular canalicular membranes (Hanzel et al., 1989, 1991). In this cell system, as in other epithelial cells, ezrin is thought to play a major role as both a membrane–cytoskeletal linker (Hanzel et al., 1991; Algrain et al., 1993) and a critical component in microvilli biogenesis (Takeuchi, et al., 1994; Berryman et al.,1995).

While both type I and type II cAMP‐dependent protein kinases (A‐kinases) are present in gastric parietal cells, the type I A‐kinase is predominantly cytosolic, while the type II enzyme partitions into both cytosolic and particulate fractions (Corbin et al., 1977; Chew, 1985; Goldenring et al., 1992). Chew (1985) has demonstrated that elevation of intracellular cAMP activates cytosolic type I A‐kinase. In addition, we have observed that stimulation of parietal cells with histamine also induces dephosphorylation of the particulate regulatory subunit of type II A‐kinase (RII) (Goldenring et al., 1992). The dephosphorylation of RII recently has been correlated with dissociation of the type II holoenzyme and kinase activation (Dibona et al., 1979). Together, these data suggested that subcellular activation of A‐kinase in parietal cells might be regulated through sequestration by specific A‐kinase anchoring proteins (AKAPs) (Scott and Carr, 1992). More recently, we have characterized a 78 kDa AKAP (AKAP78) which was enriched significantly in gastric parietal cells (Dransfield et al., 1995). This AKAP distributed predominantly into a Triton X‐100‐insoluble fraction (Dransfield et al., 1995), suggesting its association with the cytoskeleton. We have now sought to isolate and identify AKAP78 from parietal cells. The results indicate that AKAP78 is ezrin, a member of a family of actin‐associated proteins.


Isolation of gastric AKAP78

Since AKAP78 was present in both particulate and cytosolic fractions, we purified AKAP78 from 100 000 g gastric fundic mucosal supernates using RII affinity chromatography. Recombinant RII was bound to cAMP–agarose resin, and the resulting RII affinity resin was then used to resolve gastric fundic mucosal supernates. Figure 1 demonstrates that the majority of protein voided the RII affinity column. However, most of the AKAP78 was retained on the resin, as evidenced by a decrease in [32P]RII binding to AKAP78 in the void fraction. Examination of the eluted fraction on Coomassie blue‐stained gels revealed an essentially homogeneous fraction containing a single 78 kDa band. A minor band of ∼75 kDa, which also bound RII, was observed variably and appears to be a breakdown product of the 78 kDa band. The 250 mM NaCl eluate contained a marked enrichment in AKAP78 RII binding. Further elution of the resin with higher NaCl concentrations up to 1.0 M failed to elute any other RII binding proteins, and no significant protein elution was observed either by A280 or Coomassie blue staining. Additional elution of the column overnight with 10 mM cAMP resulted in the elution of recombinant RII (data not shown), but no further AKAPs were co‐eluted. In contrast to the results obtained with the RII affinity column, chromatography of gastric supernate over cAMP resin retained <5% of the cytosolic ezrin (data not shown). Ezrin which was retained on cAMP resin could be eluted with either 250 mM NaCl or 1 mM cAMP. These results suggested that a small population of ezrin is associated with RII in the cytosolic fraction.

Figure 1.

Purification of gastric AKAP78. Gastric fundic mucosal high speed supernates from two rabbit fundi were chromatographed over RII affinity resin. Upper panel: the majority of protein, as monitored by A280, voided the column. A sharp elution of protein was observed at 250 mM NaCl. No further elutions of protein were detectable at either higher concentrations of NaCl or with 1 mM cAMP (data not shown). Lower panels: Coomassie blue staining demonstrated, in comparison with the unresolved gastric fundic mucosal supernate (S), a decrease in the staining in the void fraction (V) for a 78 kDa protein, which was eluted as an essentially homogeneous fraction with 250 mM NaCl (E). Examination of [32P]RII binding to these three fractions also demonstrated the loss of binding to AKAP78 in the void fraction with elution of this protein in the 250 mM NaCl fraction. A minor 75 kDa breakdown product was observed variably. Examination of the three fractions for ezrin immunoreactivity showed a similar diminution of ezrin immunoreactivity in the void fraction with enrichment into the 250 mM NaCl fraction. The results are representative of three separate experiments.

Given the prominence of the eluted AKAP78 band, we investigated the possibility that AKAP78 might be the canalicular‐associated protein, ezrin (Hanzel et al., 1991). Monoclonal antibodies against ezrin documented the presence of ezrin in the gastric fundic mucosal supernates, with both a decrease in ezrin immunoreactivity in the void fraction of the RII affinity column and an enrichment of ezrin immunoreactivity in the 250 mM NaCl eluate paralleling the enrichment for RII binding (Figure 1). These results suggested that AKAP78 was indeed ezrin.

RII binding to ERM family members

Ezrin is a member of the Band 4.1 superfamily of cytoskeletal proteins which also includes talin, merlin, radixin and moesin (Gould et al., 1989; Turunen et al., 1989; Lankes and Furthmayer, 1991; Sato et al., 1992). All of the Band 4.1 proteins contain significant homologies in their amino‐terminal regions, which have been implicated in membrane association (Arpin et al., 1994). Ezrin, radixin and moesin represent a subfamily of proteins (the ERM family) possessing a higher degree of homology across their entire sequences (Sato et al., 1992). We investigated the capacity of recombinant human ezrin, radixin and moesin produced in a Sf9 baculovirus expression system to bind [32P]RII (Figure 2). All three of the recombinant ERM proteins bound RII. Previous studies have shown that synthetic peptides containing amphipathic α‐helical RII binding motifs can block [32P]RII binding to AKAPs (Carr et al., 1991, 1992a). One such peptide containing the RII binding region of the thyroid AKAP, HT‐31, binds with especially high affinity (Carr et al., 1991, 1992a) and has been used to identify authentic RII binding proteins. We have reported previously that the HT‐31 peptide inhibited RII binding to AKAP78 (Dransfield et al., 1995). A 1 μM concentration of the HT‐31 peptide inhibited binding of [32P]RII to all three ERM proteins (Figure 2).

Figure 2.

Binding of [32P]RII to ERM protein constructs. Total extracts of 10 000 Sf9 cells expressing full‐length recombinant ezrin (E), radixin (R) and moesin (M) were resolved on 8% SDS–PAGE gels and either stained with Coomassie blue, or transferred to nitrocellulose and probed with [32P]RII in the absence or presence of 1 μM HT‐31 peptide. Specific binding to all three ERM proteins that was inhibited in the presence of HT‐31 peptide was observed. The results are representative of two separate experiments.

Binding of RII to ezrin affinity resin

To confirm that the binding of RII to ezrin was not an artifact of the electroblot overlay assay, we incubated recombinant ezrin affinity resin with [32P]RII in the absence or presence of the HT‐31 inhibitory peptide. Figure 3 demonstrates that [32P]RII was retained on the ezrin affinity resin. However, the HT‐31 peptide significantly inhibited the association of RII with ezrin. These results, along with those in Figure 1, demonstrate that RII and ezrin associate under non‐denaturing conditions.

Figure 3.

Binding of [32P]RII to ezrin affinity resin. Ezrin affinity resin was incubated with [32P]RII (1) in the absence or (2) in the presence of 1 μM HT‐31 inhibitory peptide. The positions of molecular mass standards (kDa) are noted at the left. The autoradiograph of adherent proteins resolved on 10% SDS–PAGE gels demonstrates the inhibition of [32P]RII binding to ezrin by the HT‐31 peptide. The results are representative of two separate experiments.

Mapping the RII binding domain in ezrin

All three of the ERM proteins contain an amino‐terminal Band 4.1 homology domain (Sato et al., 1992; Algrain et al., 1993; Arpin et al., 1994; Martin et al.,1995). This is followed by a region of α‐helix which, in ezrin and radixin, is terminated by a polyproline domain. Following the polyproline domain is a carboxy‐terminal domain which has been implicated in actin filament association (Turunen et al., 1994). To map the region responsible for RII binding, we investigated RII binding to full‐length recombinant ezrin and to a series of ezrin truncations produced in a Sf9 baculovirus expression system. Figure 3 summarizes the results of binding to these constructs. RII binding was observed in full‐length recombinant ezrin and a carboxy‐terminal truncation (ezrin310–586), but no binding was observed in the amino‐terminal fragment (ezrin1–310). Similar results were obtained in full‐length, amino and carboxyl truncations of ezrin expressed as GST fusion proteins in a prokaryotic expression system (data not shown). Further carboxy‐terminal truncations mapped the RII binding to the region of predicted α‐helix between amino acids 373 and 478 (Figure 4). These results suggested that the RII binding region was situated within the ezrin α‐helical domain. No binding of [32P]RII was demonstrated to the recombinant subunits of the H/K‐ATPase (Figure 4), supporting the specificity of the interaction of RII with a discrete component of the parietal cell secretory machinery, ezrin.

Figure 4.

Binding of [32P]RII to recombinant ezrin constructs. (A) Total extracts of 20 000 Sf9 cells expressing full‐length recombinant ezrin (A.1) six truncated constructs (indicated as A.2–7), the α and β subunits of H/K‐ATPase (H) and wild‐type viral proteins (w) were resolved on 15% SDS–PAGE and transferred to nitrocellulose. (B) Visualization of ezrin constructs with a combination of polyclonal rabbit anti‐ezrin antibodies. The Western blot demonstrates the migration of all seven recombinant protein constructs. (C) [32P]RII overlay: the autoradiograph demonstrates RII binding to full‐length ezrin, ezrin310−586, ezrin310−478 and ezrin373−586. Minor binding to some proteolytic fragments was also noted. This pattern of binding to the truncated constructs mapped the binding site to between amino acids 373 and 478. No binding was observed either in the Sf9 extract containing the α and β subunits of H/K‐ATPase or in the extract from cells infected with wild‐type baculovirus. The results are representative of two separate experiments.

To define further the region responsible for RII binding in ezrin, a series of recombinant carboxy‐terminal truncations were constructed in Escherichia coli (Figure 5). In both purified GST fusion proteins and cleaved recombinant proteins, RII binding was observed in ezrin1−439, ezrin1−466 and ezrin1−482, but not in the shorter truncations, ezrin1−366, ezrin1−406 and ezrin1−423. No binding of [32P]RII to the GST polypeptide was observed following cleavage of the fusion proteins (Figure 5). These results, in combination with the binding to the Sf9 truncations, map the RII binding region to between amino acids 373 and 439.

Figure 5.

Binding of [32P]RII to carboxy‐terminal truncations of ezrin. Purified recombinant truncations as GST fusion proteins (GST–ezrin) or after cleavage of the GST extension (ezrin) were separated on 8% SDS–PAGE gels (1, ezrin1−366; 2, ezrin1−406; 3, ezrin1−423; 4, ezrin1−439; 5, ezrin1−466; 6, ezrin1−482). (A) Coomassie blue‐stained gel of proteins. The position of the GST cleavage product is noted at the right. Ezrin1−439 repeatedly demonstrated a slightly higher mobility than predicted. (B) Ezrin immunoblot. Proteins were transferred to nitrocellulose and probed with a combination of anti‐ezrin polyclonal antibodies. Ezrin immunoreactivity coincided with the Coomassie blue staining bands observed in (A). (C) Binding of [32P]RII. Proteins were transferred to nitrocellulose and probed with [32P]RII. Only the three largest ezrin constructs bound RII. The results are representative of two separate experiments.

Since previous investigations have suggested that RII association with AKAPs is mediated through an amphipathic α‐helical domain (Carr et al., 1991), we examined the region between amino acids 373 and 439 for a putative RII binding site. Figure 6A depicts the helical wheel analysis of a putative amphipathic 14‐amino‐acid sequence between amino acids 417 and 430. While AKAPs in general have little sequence homology across their entire sequences, significant homologies have been noted in their RII binding regions (Scott and McCartney,1994). A comparison of the region spanning the RII binding motif of ezrin with analogous regions from MAP2, HT‐31 and AKAP79 (Figure 6B) shows significant homologies throughout the sequence.

Figure 6.

Analysis of the putative RII binding, amphipathic α‐helical region of ezrin. (A) Helical wheel analysis of the ezrin amino acid sequence between 373 and 478 demonstrated a single 14‐amino‐acid region which displayed a putative amphipathic α‐helical structure. *L indicates the position of the leucine to proline substitution which disrupts the amphipathic helix. (B) The putative RII binding region of ezrin was compared with analogous regions from MAP‐2, HT‐31 and AKAP79. Positions of significant ezrin homology with other AKAP sequences are designated with asterisks (*). Positions of identities with the ezrin sequence are noted in bold. (C) To analyze the amphipathic α‐helical structure, purified gastric AKAP78/ezrin was resolved on 8% SDS–PAGE gels and transferred to nitrocellulose. Strips of electroblot (∼0.5 μg of AKAP78/ezrin protein) were incubated with [32P]RII in the absence of blocking peptide (0) or in the presence of 1 μM HT‐31 peptide or ezrin409−438 peptide (EZpep) in concentrations of 20, 30 and 60 μM. The HT‐31 peptide abolished binding to AKAP78. Ezrin409−438 peptide blocked binding to ezrin with an approximate IC50 of 30 μM. The results are representative of four separate experiments. (D) Disruption of the α‐helical domain with a leucine to proline substitution. A synthetic peptide with a leucine to proline substitution at amino acid 421 in ezrin409−438 (ezrin409−438L≥P) was constructed. Strips of electroblot (∼0.5 μg of AKAP78/ezrin protein) were incubated with [32P]RII in the absence of blocking peptide (1), in the presence of 60 μM ezrin409−438 (2) or in the presence of 60 μM ezrin409−438L≥P (3). Ezrin409−438L≥P failed to inhibit binding to AKAP78/ezrin. The results are representative of three separate experiments.

As noted above, peptides constructed against the amphipathic helical regions of AKAPs can inhibit RII binding to native AKAPs. The HT‐31 amphipathic helix peptide inhibited RII binding to purified gastric AKAP78/ezrin (Figure 6C). We synthesized a 30‐amino‐acid polypeptide encompassing the predicted ezrin amphipathic α‐helix (ezrin409−438). Figure 5C demonstrates that ezrin409−438 could block all RII binding to purified gastric ezrin (IC50= 30 μM). While the IC50 for ezrin409−438 was higher than that for the HT‐31 peptide, this figure is similar to inhibitory concentrations for other AKAP RII binding domain peptides (Carr et al., 1992b). In addition, previous studies have also demonstrated that disruption of the amphipathic helical region with leucine to proline substitutions abolishes RII binding. We, therefore, tested the ability of a synthetic ezrin409−438 peptide with a leucine to proline substitution at amino acid 421 (ezrin409−438L≥P) to inhibit RII binding to purified gastric AKAP78/ezrin (Figure 6). While 60 μM ezrin409−438 completely inhibited RII binding, ezrin409−438L≥P failed to inhibit at concentrations of up to 100 μM. These data support the presence of an amphipathic α‐helical RII binding site in a specific region of the ezrin α‐helical domain.

Localization of RII in parietal cells

To investigate the localization of RII in parietal cells, we developed a polyclonal rabbit antiserum against recombinant murine RII. In Western blots, the polyclonal antibodies recognized a single 55 kDa species in murine fundic mucosa which co‐migrated with recombinant murine RII (Figure 7).

Figure 7.

Characterization of polyclonal anti‐murine RII. The rabbit polyclonal antiserum raised against recombinant murine RII (1:6000) was used to probe Western blots of (1) recombinant murine RII (50 ng) and (2) post‐nuclear supernate from homogenates of murine fundic mucosa (100 μg). Specific labeling was detected with horseradish peroxidase‐coupled secondary antibody and enhanced chemiluminescence. The positions of molecular mass standards (kDa) are noted at the left. A single 55 kDa immunoreactive band was observed in mouse fundic mucosa which co‐migrated with recombinant murine RII. The results are representative of two separate experiments.

The polyclonal antibodies were used to study the localization of RII in parietal cells from vehicle‐treated and gastrin‐stimulated mice. Strong staining for RII was observed in mouse parietal cells in a punctate pattern throughout the cell (Figure 8a). No staining of parietal cells was observed with pre‐immune serum (data not shown). Double labeling of F‐actin with bodipy‐phallacidin demonstrated scattered areas of overlap. Only occasional parietal cells in vehicle animals also showed co‐localization of RII, and these were found predominantly near the tops of glands (data not shown). In gastrin‐stimulated animals, many parietal cells exhibited co‐localization of RII and bodipy‐phallacidin staining (Figure 8d–i). No labeling was observed in the nuclei of parietal cells from either vehicle or gastrin‐treated animals. These data suggested an association of RII with the secretory canaliculus, at least in stimulated parietal cells.

Figure 8.

Localization of RII in murine gastric parietal cells. Polyclonal antibodies against RII (1:6000) were used to stain sections of murine fundic mucosa from vehicle‐treated mice (ac) and mice stimulated with gastrin (di). RII staining was detected with Cy5‐rabbit IgG (a, d and g). All sections were double labeled with bodipy‐phallacidin (b, e and h) to detect F‐actin which underlies the secretory canaliculus. Anaglyph superimpositions of the double‐staining patterns are shown in (c), (f) and (i). RII co‐localization with F‐actin lining the secretory canaliculus is observed prominently in parietal cells from animals stimulated with gastrin (f and i). The results are representative of four separate experiments. Bar = 20 μm.


In many cellular systems, type II A‐kinase is sequestered in specific regions of the cell. This compartmentalization may regulate the spatial and temporal access of the kinase to specific substrates as well as to its activator, cAMP. Investigations over the past several years have pointed to the presence of a diverse group of AKAPs which are responsible for immobilizing type II A‐kinase in discrete particulate fractions in association with either membrane organelles or the cytoskeleton (Scott et al., 1992). In gastric parietal cells, histamine, through the elevation of intracellular cAMP, stimulates the fusion of intracellular tubulovesicles containing H/K‐ATPase with an apically oriented intracellular canalicular target membrane system (Forte et al., 1981). In parietal cells, we previously have provided evidence that elevations in cAMP lead to dephosphorylation of RII in particulate fractions (Goldenring et al., 1992). These studies had suggested an association of type II A‐kinase with the intracellular canalicular membranes. In addition, our preliminary investigations demonstrated an association of the major parietal cell AKAP, AKAP78, with the Triton X‐100 cytoskeleton (Dransfield et al., 1995). The present results indicate that RII relocates to the secretory canaliculus during stimulation of parietal cells. These data provided further support for the association of the type II A‐kinase with the canalicular membrane, which is intimately associated with a dense F‐actin cytoskeleton (Hanzel et al., 1989; Soroka et al., 1993). The present studies demonstrate that AKAP78 is ezrin, the actin‐associated protein which is highly enriched in the F‐actin cytoskeleton underlying the parietal cell canaliculus. Thus, ezrin may anchor the type II A‐kinase into the critical zone for regulated fusion of intracellular vesicles with the canalicular target membranes.

A great deal of study has focused on the functions of the amino and carboxyl domains of ezrin. The Band 4.1 homology region has been implicated in the association of ezrin with membrane surfaces. In particular, recent investigations have suggested that, in some systems, the ezrin interacts with the ubiquitous cell surface glycoprotein, CD44 (Tsukita et al., 1994). The carboxy‐terminal region of ezrin has been implicated in the association of ezrin with F‐actin filaments. Ezrin may interact preferentially with F‐actin filaments containing b‐actin (Yao et al., 1995). While co‐localization of ezrin with actin filaments has been noted in a number of systems (Hanzel et al., 1991; Algrain et al., 1993; Arpin et al., 1994; Martin et al., 1995), the exact nature of the ezrin–F‐actin interaction is controversial. Turunen et al. (1994) have supported a direct association of ezrin with F‐actin through its terminal 35 amino acids, while other studies (Shuster and Herman, 1995) suggest an indirect association with F‐actin. Several investigations also suggest that the amino‐ and carboxy‐terminal regions interact to regulate their respective activities. Martin et al. (1995) have demonstrated that the ezrin amino‐terminal domain inhibits the capacity of the carboxy‐terminus to induce cellular extensions. This inhibition may, in part, stem from a capacity of the protein to form head‐to‐tail homodimers (Gary and Bretscher, 1993, 1995; Andreoli et al., 1994).

The present studies suggest that the α‐helical region of ezrin contains a domain for association with type II A‐kinase. Both recombinant radixin and moesin also bound RII. Since the RII binding to all three ERM proteins was inhibited by the HT‐31 amphipathic helix peptide, anchoring of A‐kinase may be a general property of the ERM proteins. Our results have delineated a specific putative amphipathic α‐helical region in ezrin between amino acids 417 and 430. The requirement for the α‐helix in binding is confirmed by the inability of the proline‐substituted ezrin peptide (ezrin409−438L≥P) to block RII binding to ezrin. Since the amphipathic α‐helix peptide from HT‐31 associates with the amino‐terminus of RII (Carr et al., 1991; Carr and Scott, 1992), its ability to block RII binding to ezrin indicates that ezrin also binds to the same AKAP association domain. The specificity and structural basis for the interaction of ezrin with RII was established by the analysis of recombinant truncations and binding inhibition with synthetic peptides spanning the putative amphipathic α‐helix. Although these studies indicate that ezrin is the major AKAP in parietal cells, the elution of cytosolic ezrin from RII affinity resin with 250 mM NaCl suggests that it has a lower affinity than many previously described AKAPs (Scott and McCartney, 1994). While these studies focused on soluble ezrin isolated following homogenization, in situ most of the ezrin in parietal cells is associated with the F‐actin cytoskeleton (Soroka et al., 1993). Mandel and colleagues (Chen et al., 1994) have observed that dephosphorylation of ezrin during renal hypoxia correlates with dissociation of ezrin from cortical actin filaments. Post‐translational modifications in cytoskeletally associated ezrin might alter affinities for RII. Since terminal truncations of ezrin can bind RII, dimerization of ezrin is not required. Nevertheless, it is not clear whether dimerization could inhibit binding or whether binding of RII could inhibit dimerization. Association with non‐dimerized or phosphorylated ezrin might result in altered affinities for the interaction of RII with ezrin.

In addition to its established role as a membrane–actin bridging protein, several studies have suggested important interactions between ezrin and second messenger systems. Bretscher (1989) demonstrated that epidermal growth factor (EGF) stimulated the phosphorylation of both tyrosine and serine residues on ezrin, in concert with its stimulated recruitment into zones of membrane ruffling in A‐431 cells. Histamine elicits the phosphorylation of ezrin in situ in parietal cells on serine residues (Urushidani et al., 1989). In addition, EGF stimulates ezrin phosphorylation in parietal cells on a serine residue distinct from that phosphorylated in response to histamine (Whitney et al., 1990). While tyrosine phosphorylation of ezrin in A‐431 cells appears to be due to phosphorylation by the EGF receptor tyrosine kinase (Bretscher, 1989), the kinases responsible for serine phosphorylation are as yet unclear. In addition to its status as a substrate for protein kinases, recent investigations have demonstrated that a PIP2 binding site is contained within the first 309 amino acids of ezrin (Niggli et al., 1995). While the status of ezrin as a target for protein phosphorylation is well established, the present data represent the first indication that ezrin, as well as other ERM proteins, directly associates with a specific protein kinase. Thus, ezrin may serve to sequester type II cAMP‐dependent protein kinase within critical sub‐apical membrane domains providing for specific localization of A‐kinase action within epithelial cells. Such regionalization in parietal cells to the intracellular canaliculus would position A‐kinase in the zone where membrane fusion and recycling are thought to be regulated by elevations of intracellular cAMP. Although widely accepted, the membrane fusion and recycling hypothesis (Forte et al., 1981) is still in dispute and has been challenged recently by an alternative mechanism involving an expansion of the secretory membrane (Pettitt et al., 1995). Nevertheless, in both mechanisms proposed, ezrin phosphorylation is considered a key factor for membrane reorganization during parietal cell stimulation. In other epithelial cells, sequestration of A‐kinase by ezrin might provide for specific regulation of plasma membrane pumps and channels located in the proximity of sub‐plasmalemmal F‐actin. Thus, ezrin may not only link plasma membrane with the cytoskeleton, but also act as a nidus for the localization of critical regulatory enzyme systems.

Materials and methods


[γ‐32P]ATP was purchased from New England Nuclear (Boston, MA). New Zealand White rabbits (2–3 kg) were purchased from Shelton's Bunny Barn (Waverly Hall, GA). A‐kinase catalytic subunit was purchased from Promega (Madison, WI). All other reagents were from standard suppliers and were of the highest purity available.

Gastric mucosal fractionation

Scraped fundic gastric mucosa from the rabbit was minced and diluted 10‐fold (v/v) with homogenization buffer (buffer A) in (mM): mannitol (113), sucrose (37), EDTA (0.4), MES (5), pH 6.7, containing benzamidine (5), AEBSF (0.1), 0.5 mg/ml leupeptin, 0.2 mg/ml chymostatin, 0.2 mg/ml pepstatin, 0.5 mg/ml soybean trypsin inhibitor and 1.4 TIU aprotinin. The tissue samples were homogenized in a Potter Teflon‐on‐glass homogenizer by five up and down strokes at 300 r.p.m. with a Cole Parmer Master Servodyne homogenizer and sequentially centrifuged at 50, 1000, 10 000 and 100 000 g to prepare four particulate pellets and a final high‐speed supernate fraction.

Purification of gastric AKAP78

Recombinant RII produced in BL21(DE3)pLysS bacteria possessing an RII‐pET11d expression plasmid (a gift of Dr John Scott, Vollum Institute, Portland, OR) was purified by cAMP affinity chromatography. For RII affinity chromatography, RII was not eluted from 0.5 ml of cAMP resin and, after extensive washing in buffer containing 1 M NaCl, the column was equilibrated in 50 mM HEPES, pH 7.4 with the previously noted protease inhibitors (Equilibration buffer). Gastric supernates from two rabbit stomachs were passed over the column and, following extensive washing with Equilibration buffer with monitoring of A280, the column was washed sequentially with Equilibration buffer containing 250 mM, 500 mM and 1 M NaCl, followed by an overnight incubation in the presence of 10 mM cAMP. Fractions of 1 ml were collected.

RII overlay

32P‐Labeled RII was used for the overlay procedure. Recombinant RII was phosphorylated by incubation with purified bovine A‐kinase catalytic subunit and [γ‐32P]ATP, as described previously (Dransfield et al., 1995). Gastric proteins were resolved on 8% SDS–PAGE gels and transferred to nitrocellulose (0.2 μm) overnight at 100 mA. Protein electroblots were blocked with 5% non‐fat dry milk, 0.1% bovine serum albumin (BSA) in 25 mM Tris–HCl, pH 7.5, 150 mM NaCl, 0.05% NaN3 (Blotto) for 16–24 h at 4°C prior to the addition of [32P]RII (100 000 c.p.m./ml Blotto, supplemented with 5 mM benzamidine) for 4 h at room temperature. Following five 15 min washes in Blotto, binding was visualized by exposure of blots to a Phosphorimaging screen (Molecular Dynamics, Sunnyvale, CA).

Recombinant protein constructs

Recombinant baculoviruses coding for human ezrin, ezrin truncations and rat H/K‐ATPase were utilized as previously described (Martin and Mangeat, 1994; Martin et al., 1995). New ezrin constructs (ezrin310−478, ezrin1−310 and ezrin1−373) and viruses coding for human radixin (carboxy‐tagged with 11 amino acids of the vesicular stomatitis virus glycoprotein G) and moesin (with amino acids 1–11 deleted) were obtained using similar protocols (Martin and Mangeat, 1994; Martin et al., 1995). Total extracts from Sf9 cells expressing the various recombinant proteins were resolved on SDS–PAGE and transferred to nitrocellulose for 100 min at 0.8 mA/cm2 in a semi‐dry apparatus.

Carboxy‐terminal truncations of ezrin were prepared as prokaryotic recombinant proteins. The subcloning of full‐length human ezrin cDNA into pGEX‐2T vector was described previously (Andreoli et al., 1994). Carboxy‐terminal deletions were obtained using the double‐stranded nested deletion kit from Pharmacia. The ezrin‐containing pGEX‐2T vector was cleaved with EcoRI, blunted with thio‐dNTPs and cleaved with EspI. Exonuclease III was then added and the reaction was stopped every 30 s for 10 min. After nuclease S1 action, filling‐in with the Klenow DNA polymerase fragment and overnight ligation, DNA was transformed into TG1 bacteria. Randomly picked clones were grown overnight in LB medium, induced for 1 h with 0.5 mM IPTG and analyzed by Western blotting for protein expression after adsorption of the bacterial lysate to glutathione–agarose. Antibodies against the ezrin moiety or the GST part of the fusion protein were used to select clones according to their relative mobility in SDS–PAGE. Selected clones were sequenced using the T7 Sequenase version 2.0 (Amersham). The full‐length ezrin used its own stop codon. The pGEX‐2T vector which contains stop codons in all three reading frames provided the stop codon for the truncated proteins. As a consequence, 2–7 additional amino acids encoded by the vector were added after the carboxy‐terminal deletion site.

Association of RII with ezrin under non‐denaturing conditions

GST–ezrin was purified from E.coli as described by Andreoli et al. (1994). After extensive dialysis against 100 mM NaHCO3, 15 mg of fusion protein was coupled overnight at 4°C to 6 ml of CNBr‐activated Sepharose 4B (Pharmacia) according to the manufacturer's recommendations. After blocking of free sites and washing, the resin was stored in phosphate‐buffered saline (PBS) with 0.05% sodium azide. Ezrin affinity beads were blocked with 1 mg/ml BSA in PBS (BSA–PBS) for 1 h at room temperature. Thirty μl of affinity resin was incubated with 250 000 c.p.m. of [32P]RIIin 1 ml of BSA–PBS for 1 h at room temperature with constant gentle agitation. Incubations were performed in the presence or absence of 1 μM HT‐31 peptide. Following the incubation, the beads were washed four times in 0.5 ml of PBS and adherent proteins were removed with boiling 1.5% SDS stop solution. Proteins were resolved on 10% SDS–PAGE gels, and dried gels were exposed to Phosphorimaging screens (Molecular Dynamics, Sunnyvale, CA) for 1 h.


For the detection of ezrin, proteins were resolved on 8% SDS–PAGE gels and transferred to Immobilon P (Millipore, Bedford, MA). The blots were blocked for 1 h at room temperature with 5% non‐fat dry milk in 25 mM Tris–HCl, pH 7.5, 150 mM NaCl, 0.05% NaN3 (5% Blotto). The blots were then probed in 3% non‐fat dry milk in 25 mM Tris–HCl, pH 7.5, 150 mM NaCl, 0.05% NaN3, 0.05% Tween‐20 (3% Blotto) for 2 h at room temperature with a monoclonal antibody against ezrin (4C4; 1:2000; a gift from Dr John Forte, University of California, Berkeley). The antibody is specific for ezrin, and does not detect either radixin or moesin. For detection of recombinant ezrin protein constructs, blots were probed with a mixture of polyclonal anti‐ezrin antibodies (predominantly anti‐carboxy‐terminal domain) and anti‐ezrin1−310, as previously described (Andreoli et al., 1994). After the primary incubation, the blots were washed three times for 15 min each with 3% Blotto and then incubated with horseradish peroxidase‐conjugated goat anti‐mouse IgG (1:5000) for 1 h at room temperature. The blots were finally washed three times for 15 min each with 3% Blotto and immunoreactivity was detected with chemiluminescence (Renaissance, NEN) and autoradiography.

AKAP peptide competition

HT‐31 peptide, ezrin409−438 peptide (DQIKSQEQLAAELAEYTAKIALLEEARRRK) and ezrin409−438L≥P (DQIKSQEQLAAE P AEYTAKIALLEEARRRK) were synthesized as free acids by the University of Georgia, Athens Peptide Synthesis Core Laboratory. Peptides were dissolved (1 mM) in distilled water. For peptide competition experiments, [32P]RII was pre‐incubated for 30 min either in the absence or presence of different concentrations of competitive peptide, prior to its addition to strips of electroblots of purified gastric ezrin.

RII antibodies

A polyclonal antiserum against recombinant murine RII was developed in New Zealand White rabbits. To verify the specificity of the antiserum, Western blots of recombinant RII and homogenates of mouse gastric mucosa were probed at a 1:6000 dilution. Specific binding was detected with anti‐rabbit horseradish peroxidase secondary antibodies and enhanced chemiluminescent detection.

RII immunocytochemistry

C57/Bl6 mice (Jackson Laboratories) were injected intraperitoneally with either saline vehicle or gastrin (50 μg/kg). After 30 min, animals were sacrificed by carbon dioxide inhalation, and the stomach was removed rapidly, incised along the greater curvature and fixed in 4% paraformaldehyde. Frozen sections of stomachs were cut at 5 μm thickness and stained with either pre‐immune serum or the anti‐RII antiserum (1:6000) for 2 h at room temperature. Specific labeling was visualized with Cy5‐conjugated donkey anti‐rabbit IgG (Jackson Laboratories). Sections were also labeled with bodipy‐phallacidin (Molecular Probes) to visualize F‐actin underlying the secretory canaliculus. Sections were examined on a Molecular Dynamics confocal fluorescence microscope with dual imaging using Imagespace software.


We thank Dr John Scott for the gift of recombinant murine RII, Dr John Forte for the gift of a monoclonal antibody against ezrin, Dr Monique Arpin for the human radixin cDNA, and Dr Gilles Thomas for the human moesin cDNA. The authors thank Dr Richard Cameron for assistance with the production of polyclonal antibodies against RII. This work was supported by National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) FIRST Award Grant DK43405, and a Veterans Affairs Merit Award (to J.R.G.). D.T.D. was the recipient of a National Research Service Award DK‐09026. P.H.M. is supported by a grant from l'Association pour la Recherche sur le Cancer (Contract 6844) and grant BMH4 CT950090 from the European Commission.