Comparative study of ezrin phosphorylation among different tissues: more is good; too much is bad

Abstract

In a comparison of three different tissues, the membrane cytoskeleton linker protein ezrin was found to assume high levels of phosphorylation on threonine-567 (T567) in the brush border membranes of renal proximal tubule cells and small intestine enterocytes, in contrast to the apical canalicular membrane of gastric parietal cells. Together with an earlier observation that increased T567 phosphorylation is associated with more elaborate microvilli in parietal cells, this comparative study suggested a higher phosphorylation level requirement for the denser and more uniform distribution of microvilli at brush border surfaces. Using a kinase inhibitor, staurosporin, and metabolic inhibitor, sodium azide, relatively high turnover of ezrin T567 phosphorylation was observed in all three epithelia. Aiming to understand the role of phosphorylation turnover in these tissues, detergent extraction analysis of gastric glands and proximal tubules revealed that an increased phosphorylation on ezrin T567 greatly enhanced its association with F-actin, while ezrin-membrane interaction persisted regardless of the changes of phosphorylation level on ezrin T567. Finally, expression of Thr567Asp mutant ezrin, which mimics the phospho-ezrin state but does not allow turnover, caused aberrant growth of membrane projections in cultured proximal tubule cells, consistent with what had previously been observed in several cell lines and gastric parietal cells. These results fit into a model of surface plasticity, which posits that the turnover of phosphorylation on T567 empowers ezrin to relax and reposition membrane to the underlying cytoskeleton under varying conditions of filament growth or rapid membrane expansion (or depletion).

the membrane-cytoskeleton linker protein ezrin is a major component of membrane projections in many cell types (6). As a linker protein, ezrin is invested with two binding domains: the COOH-terminal F-actin binding domain (C-term), and the NH₂-terminal membrane binding domain (N-term), which has also been called the FERM domain because of a conserved membrane binding domain found in all members of the FERM (4.1, ezrin/radixin/moesin) protein family (1, 10, 16, 17). It seems that the N-term can be further divided into two subdomains because 1) phosphorylation-dependent and -independent membrane binding was observed for ezrin (49); and 2) ezrin can bind two different types of membrane components: phospholipid phosphatidylinositol 4,5-bisphosphate (PIP₂) (32) and membrane proteins. The binding of ezrin to membrane proteins was reported to be either direct or through adaptor proteins, such as sodium/hydrogen exchanger (NHE) regulatory factors 1 and 2 (5).

The intramolecular interaction of the N-term with C-term provides a means for regulating ezrin activity. In gastric parietal cells, where ezrin was first reported as a stimulation-dependent 80 K phosphoprotein (39, 40), N-C binding of ezrin was visualized by in vitro fluorescence resonance energy transfer (FRET) analysis in situ (48), complementing evidence from blot overlay experiments (14). Ezrin in the N-C conformation did not show significant binding to other proteins, because coimmunoprecipitation with anti-ezrin did not pull down any binding proteins from gastric gland lysates (48). Phosphorylation of ezrin on T567 was found to break the N-C binding and turn ezrin into the active conformation, which allows F-actin binding and stronger membrane binding (49). Similar observations were also observed with other members of the ERM protein family, moesin (20, 30, 34) and radixin (21, 25).

The development of a specific antibody against threonine-567-phosphorylated ezrin isoform (T567 in ezrin, T558 in moesin, and T564 in radixin) (25), together with the introduction of the T567D mutant (15), greatly facilitated the study of the physiological significance of this phosphorylation event. A common observation is that the phosphorylation on T567 is often accompanied by enhanced cellular activity, for instance, the translocation and activation of NHE3 in an intestinal cell line (36), and the formation of microvilli, lamellipodia, and membrane ruffles in several other systems (11, 15, 33). Naturally, phosphorylation on T567 is regarded as an activation mechanism for ezrin (12), and the phosphorylated ezrin is considered as an active form of ezrin (45).

The expression of ezrin T567D mutant also caused elongated projections on the plasma membrane of cultured gastric parietal cells (47). However, the elongated projections and the associated T567D mutant were not located on the apical membrane where the majority of native ezrin is usually located. Instead, T567D mutant expression caused abnormal growth of membrane projections on the basolateral membrane and thus changed the polarity of gastric parietal cells (47). We attributed this phenomenon to the blockage of dephosphorylation after our more recent discovery that there is a high turnover of phosphorylation on ezrin T567 (49). Fluorescence recovery after photobleaching (FRAP) analysis revealed that the ezrin T567D mutant was tightly associated with its basolateral membrane/cytoskeleton locus. Thus dephosphorylation is as important as the phosphorylation event for ezrin function in gastric parietal cells.

Is this phosphorylation turnover mechanism for the regulation of ezrin activity also present in other cell systems? We describe here surprising results, comparing the state of ezrin phosphorylation in gastric parietal cells and other freshly isolated epithelial cells, mainly renal proximal tubule cells and intestinal enterocytes, where ezrin is enriched in the brush border microvillar membranes. We found that the steady-state phosphorylation level of ezrin T567 in the latter brush border systems is much higher than that in gastric parietal cells; nevertheless, ezrin T567 phosphorylation is regulated by a turnover mechanism in all these epithelia. In addition, we provide evidence to support the notion that ezrin can bring membrane to filamentous actin binding sites, thus high turnover of ezrin phosphorylation in the actin binding domain empowers ezrin to reposition the membrane along the filamentous actin length and provide dynamic surface plasticity.

MATERIALS AND METHODS

Reagents.

Phosphatase inhibitor calyculin A (CLA; Biomol International, Plymouth Meeting, PA) and kinase inhibitor staurosporin (Alexis, Lausen, Switzerland) was used at 1 μM final concentration. Stock solutions of these drugs were made in DMSO. When CLA and staurosporin were used, control samples were mock treated with DMSO.

The monoclonal anti-ezrin (4A5) antibody used in this study was purchased from Covance (Berkeley, CA). Horseradish peroxidase (HRP)-conjugated goat anti-mouse and HRP-conjugated goat anti-rabbit were purchased from Jackson ImmunoResearch (West Grove, PA). FITC-labeled phalloidin is from Sigma. Alexa 555-conjugated goat anti-mouse was a product of Invitrogen (Eugene, OR). Phospho-ezrin (Thr567)/radixin (Thr564)/moesin (Thr558) rabbit antibody (anti-T567P) was purchased from Cell Signaling Technology (Danvers, MA). Mouse monoclonal anti-moesin antibody is from Santa Cruz Biotechnology (Santa Cruz, CA). Monoclonal anti-actin (clone C4) was purchased from MP Biomedicals (Solon, OH).

Recombinant adenoviruses rAD/EzWT-CFP, rAD/EzT567A-CFP, and rAD/EzT567D-CFP were produced and amplified in our lab as described previously (47). These viruses express wild-type ezrin, its T567A mutant, or its T567D mutant, with a cyan fluorescence protein (CFP) tag inserted at the carboxy terminus.

Isolation of gastric glands, renal proximal tubules, renal proximal tubule cells, and small intestine organoids.

All procedures and treatments for handling animals were reviewed and approved by the Berkeley Animal Care and Use Committee. Gastric glands were isolated from New Zealand White rabbits (Oryctolagus cuniculus) as previously described (47). About 60% of the mass of isolated gastric glands consists of parietal cells, and these contain virtually all of the glandular ezrin.

Isolation of the renal proximal tubules followed the procedure of Mandel's laboratory with modifications (9). Briefly, a kidney was taken immediately after the stomach was removed from the rabbit. After removal of the renal capsule, the cortex was trimmed from the excised kidney, minced, and digested for 30 min at 37°C in Eagle's minimum essential medium (MEM) supplemented with 20 mM HEPES (pH 7.3), 1 mg/ml collagenase, and 0.8 mg/ml bovine serum albumin (BSA). All solutions used for mincing, collagenase digestion, and subsequent centrifugations were all gassed with 100% O₂ to minimize hypoxia during tissue preparation. The digested mixture was filtered through cheesecloth to remove large connective tissue and undigested materials and sedimented by centrifugation at 50 g for 2 min at room temperature. The brief centrifugation efficiently removed single cells, small fragments of tubules, and other small structures. The pellet was washed twice with oxygenated MEM (repeat resuspension and centrifugation) at room temperature. The pellet was then resuspended in oxygenated MEM supplemented with 20 mM HEPES and 4 mM glycine until ready for use. Microscopic examination of the preparation indicated that ∼80% of the preparation is derived from proximal tubule. The remaining 20% consisted of other parts of the nephron and connective tissue, which have much lower levels of ezrin expression (3).

To isolate single proximal tubule cells, the isolated proximal tubules were first subjected to the same collagenase-enriched MEM described above. The digested materials were then filtered through 40 μM nylon cell strainers (Becton Dickinson, Franklin Lakes, NJ) to remove large tubules. The cells in the filtrate were pelleted by centrifugation at 200 g for 5 min at room temperature. Cells were washed twice with MEM before plating onto Matrigel (Collaborative Biomedical, Stony Brook, NY)-coated coverslips or dishes and incubated at 37°C in chemically defined culture medium A, which consists of DMEM/F12 (GIBCO BRL), 20 mM HEPES, 0.2% BSA, 10 mM glucose, 8 nM epidermal growth factor, 1× SITE medium (containing selenite, insulin, transferrin, and ethanolamine; S4920, Sigma), 1 mM glutamine, 100 U/ml penicillin/streptomycin, and 400 g/ml gentamycin sulfate, pH 7.4. After 5 h of incubation, the cells became attached to the vessel surfaces and adenoviral infection was started.

Approximately 30 cm of small intestine was removed from the rabbit, cut open, washed with saline, and the mucosa scraped. Fragments of mucosal scrapings were subjected to the same collagenase digestion procedure described for isolation of renal proximal tubules. The final pellet was resuspended in oxygenated MEM supplemented with 20 mM HEPES and consisted of small groups of villi that we call intestinal organoids.

Immunoblot analysis.

Protein samples were separated by SDS-PAGE before transferring onto nitrocellulose membranes. Membranes were blocked with 2% BSA in Tris-buffered saline (10 mM Tris, pH 7.0, 150 mM NaCl) containing 0.05% Tween 20. The membranes were then probed with primary and secondary (HRP-conjugated) antibodies. Results were then recorded by X-ray films with the Western Lightning Chemiluminescence substrate (PerkinElmer Life Sciences, Boston, MA). When reprobing was needed, the blot was stripped (in 2% SDS, 1% 2-mercaptoethanol, and 62.5 mM Tris·HCl, pH 6.8) at 50°C for 30 min. The blot was then blocked and reprobed with another primary antibody. The signals from anti-T567P and anti-ezrin probing were found not to interfere with each other or subsequent Western blot results.

Ezrin extraction analyses with Triton X-100 and digitonin.

Triton X-100 (TX-100) extraction analyses were performed to evaluate the association of ezrin with F-actin (9, 12). Freshly isolated gastric glands or renal proximal tubules were treated with 1 μM CLA for 5 min to block protein dephosphorylation and thus maximize phosphorylation levels, or treated with 10 mM sodium azide for 30 min to block metabolism and reduce ATP (and phosphorylation) levels. The tissues were then extracted with 1% TX-100 for 5 min at room temperature. After centrifugation at 100,000 g for 15 min at 4°C, samples of supernatant extracts and cytoskeleton pellet were collected and examined by Western blots successively probed to determine the relative phosphorylation level on ezrin T567 (T567-P), total ezrin, and actin.

Digitonin was used to perforate cells by way of cholesterol extraction for the study of cytosolic proteins with minimum interference to membrane proteins (24, 35). Freshly isolated gastric glands were compared with an experimental set treated with 1 μM CLA; control proximal tubules were compared with an experimental set treated with 10 mM NaN₃. The respective glands and tubules were permeabilized by treatment with 40 μg/ml digitonin and extracted for 5 min, either in low-salt buffer (250 mM sucrose, 1 mM EDTA, and 10 mM HEPES, pH 7.3) or normal phosphate-buffered saline with 1 mM EDTA. Samples were centrifuged at 10,000 g for 5 min at 4°C. Blots of supernatant extracts and residual pellet fractions were probed for ezrin T567-P and ezrin. Additional gels were run in parallel for Coomassie blue staining.

Time course of the phosphorylation turnover on ezrin T567.

Gastric glands aliquoted into individual tubes were treated with 1 μM CLA for various time periods as indicated. Aliquoted renal proximal tubules were treated with a membrane-permeable broad-spectrum kinase inhibitor, staurosporin, at 1 μM for various time periods as indicated. Reactions were stopped by immediate boiling in SDS-loading buffer. Samples were analyzed by Western blots probed for ezrin T567-P and ezrin.

Immunofluorescence microscopy.

Proximal tubules and small intestine organoids treated with sodium azide and control samples were attached to the poly-l-lysine (P1399, Sigma)-coated coverslips, fixed by 3.7% formaldehyde, and permeabilized with 0.5% TX-100. Samples were then probed with anti-ezrin antibody. Afterward, the cells were incubated with Alexa 555-conjugated anti-mouse antibody together with FITC-phalloidin. Images of Alexa 555 (excitation with 543-nm laser, emission from 590–655 nm) and FITC (excitation with 488-nm laser, emission from 505–580 nm) were collected at 1 Airy unit pinhole with Plan-Neofluar ×40/1.3 oil differential interference contrast objective on a Zeiss LSM 510 meta confocal microscope. Cells grown on Matrigel-coated coverslips were fixed by 3.7% formaldehyde and permeabilized with 0.1% TX-100, and the same procedure was then followed as described above.

Live cell imaging.

Proximal tubule cells were grown on Matrigel-coated coverslips, infected with recombinant adenovirus expressing CFP-tagged wild-type ezrin, T567A mutant, or T567D mutant for 2 days. CFP images (excitation with 458-nm laser, emission from 473–515 nm) were then collected with a water immersion objective Achroplan ×40/0.8 W at 1 Airy unit pinhole on a Zeiss LSM 510 meta confocal microscope.

RESULTS

Ezrin T567 phosphorylation in renal proximal tubule and small intestine enterocytes is higher than that in gastric parietal cells.

Two types of cells well known for the rich brush border microvilli on the apical plasma membrane, renal proximal tubule cells and small intestinal enterocytes, are a rich source of ezrin. Our first objective was to determine the relative phosphorylation level on ezrin T567 (T567-P) in these cells and compare that with parietal cells. For this purpose the relative levels of total ezrin and of T567-P ezrin were analyzed in isolated gastric glands, renal proximal tubules, and small intestine organoids by Western blot. To approach maximal and minimal levels of T567 phosphorylation, the various tissues were respectively treated with the protein phosphatase inhibitor CLA or the metabolic inhibitor sodium azide. The expectation was that inhibition of protein phosphatases by CLA would increase T567-P level if there were sustained kinase activity. On the other hand, azide shuts off ATP production by the oxidative phosphorylation, thus depriving the protein kinases of their substrate and causing decreased phosphorylation level. NaN₃ has long been used to achieve chemical anoxia in physiological studies (22, 41).

Western blots of treated and control tissues were probed with anti-T567-P followed by stripping and reprobing with anti-ezrin (Fig. 1). Two proteins were revealed by T567-P antibody. The band with higher apparent molecular mass (M_r) is T567-P ezrin because it was recognized by a highly specific ezrin antibody. The other band is likely moesin judging from its slightly lower M_r and subsequent tests. Interestingly, the results for gastric glands differed markedly from those obtained for proximal tubules and small intestine. For gastric glands, the ezrin T567-P level increased greatly when incubated with CLA, indicating that T567 phosphorylation level is low in the native condition, whereas azide did not produce any detectable effect on ezrin T567-P level, again reflecting the fact that gastric ezrin carries a low level of T567-P in the steady state. A contrasting result was observed with renal proximal tubules. CLA had virtually no effect on ezrin T567-P level, whereas azide greatly depressed T567-P. The results for small intestinal enterocytes were similar to those for proximal tubules, indicating that both of these brush border-rich tissues have a higher steady-state level of ezrin T567-P than that of gastric glands.

Fig. 1.

Differential phosphorylation level (P) on Thr567 of ezrin in small intestine epithelial cells (SI), kidney proximal tubule cells (K), and gastric glands (GG). Small intestine organoids, kidney proximal tubules, and gastric glands were isolated from rabbit by the collagenase digestion methods described in materials and methods. These tissues were mock treated with vehicle (ctrl), treated with sodium azide (N₃) or calyculin A (CLA). Western blot analyses of these samples were done with anti-T567-P, and the samples were then stripped and reprobed with anti-ezrin. The arrowhead indicates the position of ezrin T567-P. The band at the slightly lower M_r is moesin T558-P.

Relatively fast turnover of ezrin T567 phosphorylation.

To determine the relative turnover of ezrin T567-P, time-course experiments were performed to study the changes in ezrin T567-P after the addition of agents that block phosphorylation or dephosphorylation. For gastric glands, CLA was used because the phosphorylation level is low in the native state. Changes in the relative level of ezrin T567-P after addition of CLA are shown in Fig. 2A, where it appears that the time for 50% increase of ezrin T567-P was less than a minute for gastric glands. Since ezrin in proximal tubules is normally highly phosphorylated, we used agents to block phosphorylation (either azide or the membrane permeable kinase inhibitor staurosporin) and subsequently observed the time course of T567-P depletion from ezrin. In Fig. 2B, staurosporin was used to decrease the phosphorylation level. Staurosporin effectively reduced the phosphorylation of ezrin, with the time required for 50% decrease in T567-P being about 2–3 min. When doing the similar experiment with azide treatment, the half time for depletion was found to be somewhat slower (∼4 min; data not shown). Because of inherent problems with respect to permeation and multiplicity of action for the various inhibitors, these data cannot be taken as definitive values for reaction rates, but they do indicate that the T567-P turnover is relatively fast in these tissues.

Fig. 2.

Turnover of ezrin T567 phosphorylation in gastric glands and renal proximal tubules. Gastric glands were treated with CLA (A) and proximal tubules were treated with staurosporin (B). At various time points, as indicated on top of the lanes, samples were taken and boiled in SDS-PAGE loading buffer. Samples were then analyzed by Western blot with anti-T567-P, stripped, and reprobed with anti-ezrin. For proximal tubule samples, the blot was further probed with anti-moesin antibody without stripping.

Since moesin existed in a similar amount as ezrin in the brush border of proximal tubules, it was also of interest to examine the phosphorylation turnover of moesin in these samples. Thus the blots with proximal tubules were also probed with anti-moesin antibody. The result shown in Fig. 2B indicated that moesin, like ezrin, assumed a highly phosphorylated steady state in proximal tubule, and the phosphorylation level on T558 dropped rapidly upon inhibition of kinases with staurosporin. The time required for half decrease in T558-P was about 2–3 min, similar to that of ezrin T567-P.

Ezrin-cytoskeleton binding is enhanced by phosphorylation.

To evaluate T567-P turnover as a regulatory mechanism for ezrin activity in tissues, two extraction methods were used to study the interactions of ezrin with F-actin and plasma membrane in renal proximal tubules and gastric glands. Extraction with TX-100 removes soluble and membrane bound proteins, leaving behind the F-actin cytoskeleton pellet and many actin bound proteins (15). Isolated gastric glands and proximal tubules were taken fresh as control tissue, or treated with CLA to block protein dephosphorylation and thus maximize phosphorylation levels, or treated with sodium azide to reduce ATP phosphorylation levels. The tissues were then extracted with 1% TX-100 for 5 min at room temperature. Samples of supernatant extracts and cytoskeleton pellet were examined by Western blots probed for T567-P, total ezrin, and actin.

For gastric glands, total actin was predominantly distributed to the cytoskeleton pellet over the supernatant (∼4:1) and was not significantly different between the various experimental treatments (Fig. 3C). This is consistent with previous findings that actin in gastric glands is predominantly in the filamentous form (2). As expected, treatment with CLA resulted in a large increase (∼4–5 fold) in T567-P level compared with control glands (Fig. 3B). Treatment with N₃ produced no significant change in T567-P compared with fresh glands (Fig. 3B). For all treatments, ezrin T567-P was always predominantly distributed toward the pellet, which was always about four times greater than that extracted into the TX-100 supernatant (Fig. 3B). In control glands, total ezrin was distributed between supernatant and pellet in approximately 60:40 ratio (Fig. 3A). Treatment with CLA greatly altered the TX-100 extraction so that the majority of ezrin (>80%) remained with the pellet (Fig. 3A). For glands treated with N₃, there was very little difference from control, either with respect to the relative amount of T567-P or in the supernatant/pellet distribution ratio of total ezrin (Fig. 3A).

Fig. 3.

Enhanced ezrin binding to F-actin upon phosphorylation on ezrin T567 in both gastric glands (A–C) and renal proximal tubules (D–F). Control samples and samples treated with CLA or sodium azide (N₃) were extracted with Triton X-100 buffer for 5 min at room temperature. The extractions were separated from the pellets by centrifugation. Supernatant (S or Sup) and pellet (P or Pell) fractions were subjected to Western blot analyses with anti-ezrin (A and D), anti-T567-P (B and E), and anti-actin (C and F). Ezrin data are presented as the percentage of total ezrin distributed between the S and P. The T567-P data are presented as the percentage of T567-P compared with the S plus P in CLA-treated sample for both gastric glands and proximal tubules. The actin data are shown as the relative distribution of actin between S and P. Density data are plotted as means ± SE; for gastric glands, n = 4; for proximal tubules, n = 5.

As in gastric glands, proximal tubular actin was predominantly distributed to the TX-100 pellet for all treatments (Fig. 3F). Treatment of proximal tubules with CLA produced no significant change in the level of total T567-P ezrin (P > 0.05), although there was a tendency for a shift in T567-P distribution from pellet to supernatant (Fig. 3E). After N₃ treatment, the levels of T567-P ezrin were significantly (P < 0.01) decreased in the supernatant and pellet fractions (Fig. 3E), and total ezrin was redistributed toward the supernatant compared with control (Fig. 3D). The distribution of T567-P ezrin was always predominant in the pellet under all conditions for both proximal tubules and gastric glands (cf. Fig. 3, B and E). The most obvious difference between proximal tubules and gastric glands occurred in the respective control preparations where the proportion of total ezrin was predominantly associated with the cytoskeleton for proximal tubules and reversed for gastric glands (cf. Fig. 3, A and D).

For both gastric glands and renal proximal tubules, T567-P ezrin was always more distributed toward the cytoskeletal pellet than the TX-100 supernatant, consistent with studies on various cell lines (15, 25) and supporting the notion that T567 phosphorylation exposes the actin binding site on ezrin. The distribution data also predictably show that when T567 phosphorylation is high (e.g., CLA treatment in gastric glands and steady-state control for proximal tubules), total ezrin is also associated with the cytoskeletal pellet, and these trends are reversed on lowering phosphorylation with N₃.

Ezrin remains bound to membrane in the nonphosphorylated form: biochemical evidence.

In another extraction method, we used digitonin to form pores in the plasma membrane, allowing soluble cytoplasmic proteins to exit while retaining cytoskeleton, membrane proteins, and large protein complexes associated with membranes or cytoskeletal structures (24). Freshly isolated gastric glands were compared with an experimental set treated with CLA; control proximal tubules were compared with an experimental set treated with NaN₃. The respective glands and tubules were permeabilized by treatment with digitonin and extracted for 5 min, either in low-salt buffer or normal phosphate-buffered saline. Blots of supernatant extracts and residual pellet fractions were probed for ezrin T567-P, total ezrin, and actin. Additional gels were run in parallel for Coomassie blue staining to ascertain that cytosolic proteins did leak out after digitonin permeabilization (data not shown).

Similar to earlier experiments, control levels of ezrin T567-P were relatively low in gastric glands and relatively high in proximal tubules; treatment of glands with CLA caused a large increase (∼4-fold) in T567-P (Fig. 4B), and treatment of tubules with N₃ caused a large decrease in T567-P (Fig. 4D). When control preparations of either gastric glands or proximal tubules were permeabilized in low-salt buffer, relatively little of the total ezrin was released to the supernatant, e.g., <20% for glands and <5% for tubules (Fig. 4, A and C). There was little change in the pattern of ezrin released from permeabilized gastric glands after CLA treatment or the pattern of ezrin released by tubules after treatment with N₃ (Fig. 4, A and C). These data contrasted sharply with those for TX-100 cytoskeletal extraction, where low activity of T567 phosphorylation was correlated with a higher distribution of total ezrin to the supernatant. Since control glands or N₃-treated tubules demonstrate relatively little ezrin binding to actin cytoskeleton when the bulk of ezrin is in the dephosphorylated state, the general retention of ezrin in the digitonin-permeabilized preparations appears to be due to another binding site, likely the NH₂-terminal membrane binding site.

Fig. 4.

Obstinate binding of ezrin to membrane in gastric glands (A and B) and renal proximal tubules (C and D). Tissues were treated with either CLA (gastric glands) or with NaN₃ (renal proximal tubules) to obtain phosphorylation levels different from the native conditions (ctrl). Samples were then treated with digitonin at room temperature for 5 min to permeabilize the cells. Permeabilizations were performed both in low-salt buffer (sucrose) or normal phosphate-buffered saline (NaCl). Samples were separated into supernatant and pellet fractions for Western blot analyses. Blots were probed with anti-ezrin (A and C) and anti-T567P (B and D). Ezrin data are presented as the percentage of total ezrin distributed between the S and P. The T567-P data are presented as the percentage of T567-P compared with the S plus P in CLA-treated sample for gastric glands and are compared with the S plus P of the control sample for proximal tubules. Density data are plotted as means ± SE; for gastric glands, n = 3; for proximal tubules, n = 5.

The pattern of ezrin release by digitonin-permeabilized glands was considerably altered when the extraction medium included relatively high ionic strength (PBS). In PBS the majority of total ezrin was released by the control glands, and the distribution ratio was reversed by the increased T567-P associated with CLA treatment (Fig. 4A). On the other hand, for control proximal tubules with their high steady-state level of ezrin T567-P, the elevated ionic strength medium accounted for only a slight loss of total ezrin to the supernatant, and this was reversed when N₃ was used to lower T567-P (Fig. 4C).

Ezrin remains bound to membrane in the nonphosphorylated form: imaging evidence.

To study the localization of nonphosphorylated ezrin in an in situ setting, renal proximal tubules were incubated either with nitrogen for hypoxia or with sodium azide for chemical anoxia before immunofluorescence staining with ezrin antibody. Phosphorylation on ezrin T567 was found greatly decreased by hypoxia treatment and further decreased by anoxia (NaN₃) treatment (Fig. 5A). Nevertheless, proximal tubules in each of the three different treatments were stained similarly (Fig. 5, B–D). There was always a high degree of colocalization of ezrin with F-actin. Most of the staining was with the apical brush border membrane facing the lumen of the tubule; a weak but clear staining of the basal membrane was also detected; cytoplasm and lateral membrane were not stained. These results indicated that ezrin is localized to a membrane surface regardless of the phosphorylation level on T567.

Fig. 5.

The pattern of ezrin immunostaining is unchanged in kidney proximal tubule cells upon hypoxia or anoxia treatment. Proximal tubules from rabbit kidney were incubated in nitrogen-saturated media (hypoxia treatment, H) for 60 min or incubated with sodium azide (chemical anoxia, N₃) for 30 min. Control samples were incubated in oxygen-saturated media for 60 min. A fraction of each sample was analyzed by Western blot with anti-T567-P, stripped, and reprobed with anti-ezrin (A). The rest of the samples [control (B), hypoxia (C), and anoxia (D)] were stained with FITC-phalloidin (left) and mouse-anti-ezrin (center); the differential interference contrast is at right. For each treatment, higher magnifications are shown for the areas in the respective blue boxes. BL, basolateral membrane; Cyt, cytosol; BB, brush border. Bar equals 20 μm.

Similar experiments were performed with small intestine organoids. Samples of normoxia and anoxia conditions (treated with azide) were compared after double staining for ezrin and F-actin. Both stains were largely colocalized with the majority of the staining at the brush border membrane (Fig. 6, A and B). While faint staining of F-actin is visible on lateral cell membrane, ezrin seems exclusively localized on the microvilli at the brush border. No staining was detected on basal membrane for both molecules. Again, no change of localization occurred when ezrin was dephosphorylated by the azide treatment.

Fig. 6.

The pattern of ezrin immunostaining is unchanged in small intestine enterocytes after anoxia treatment. Small intestine organoids, both control (A) and sodium azide-treated samples (B), were stained with FITC-phalloidin and mouse anti-ezrin. The dephosphorylation of ezrin T567 was shown previously in Fig. 1. CB, cell body. Bar equals 20 μm.

Unregulated membrane projections with the expression of T567D mutant ezrin.

Although ezrin in renal proximal tubule is in a highly phosphorylated form, fast and high turnover of the T567-P still exists. Thus, it was of interest to determine whether this phosphorylation turnover is required for its normal function. For this purpose, isolated renal tubule cells were cultured on Matrigel-coated coverslips and infected with recombinant adenovirus expressing CFP-tagged constructs of ezrin, including wild-type, the T567A mutant, or the T567D mutant. After 48 h of infection, expression of the three CFP-tagged ezrin constructs was directly examined by fluorescence microscopy. In addition, uninfected control cells were stained for both ezrin and F-actin. As seen in Fig. 7A, these control cells maintain cell surface microvilli of about 3-μm length after 2 days in culture. Just like the freshly prepared proximal tubules, the majority of ezrin and F-actin was found to be on the microvillar membranes. The ezrin T567-P level in these cultured tubule cells was also examined. As shown in Fig. 7B, with similar amount of ezrin, the T567-P level in the control sample is similar to that in the CLA-treated sample, whereas the azide-treated sample showed a diminished level of ezrin T567-P. These results indicate that the ezrin activity is kept similar in these cultured cells compared with the freshly isolated tubules.

Fig. 7.

Characterization of cultured proximal tubule cells. A: for immunofluorescence staining of ezrin, cells grown on Matrigel-coated coverslips for 2 days were fixed and stained with FITC-phalloidin (F-actin) and mouse-anti-ezrin (native ezrin). Bar equals 5 μm. B: to examine relative ezrin T567-P level, cells were treated with CLA or azide before Western blot analysis with anti-T567P and anti-ezrin antibodies.

Live cell imaging of CFP fluorescence (Fig. 8A) showed that wild-type ezrin had a similar staining pattern to that of native ezrin. A similar staining pattern was also observed for cells expressing the ezrin T567A mutant (Fig. 8B). However, the distribution and appearance of the T567D mutant differed from wild-type or T567A mutant, although it was still localized on cell surface membrane structures (Fig. 8C). T567D mutant ezrin usually tended to accumulate at one end of a cell, compared with wild-type or the T567A mutant, which had a more even distribution at the cell surface. An even more marked difference was that cells expressing T567D mutant often carried cell surface projections longer than 5 μm (many of them longer than 10 μm), whereas cells expressing wild-type or T567A mutant carried microvilli of the more normal 2- to 3-μm length.

Fig. 8.

Expression of ezrin T567D mutant produces irregular membrane projections on the surface of renal proximal tubule cells in culture. Cells infected with recombinant adenovirus expressing wild-type (WT) ezrin-cyan fluorescence protein (CFP) (A), the T567A mutant of ezrin-CFP (B), or the T567D mutant of ezrin-CFP (C) were directly imaged with CFP excitation. CFP images were collected together with phase images. Bar equals 20 μm.

DISCUSSION

The steady state of ezrin phosphorylation in renal proximal tubules and small intestinal enterocytes is higher than that in gastric parietal cells.

Ezrin was identified as a major component on the apical canalicular membrane in gastric parietal cells stimulated to secrete acid (18, 39). Studies with the ezrin gene knockdown in mice demonstrated that ezrin is required for the stimulation-associated elaboration of the microvilli on the apical canalicular membrane of parietal cells (38). On the basis of the relatively high turnover of phosphorylation on ezrin T567 and the drastic change in its binding affinity to filamentous actin associated with that phosphorylation, a rolling motor model was hypothesized to describe how ezrin repositions membrane along the filamentous actin length, thus maintaining appropriate tensile force between membrane and cytoskeleton during dynamic membrane turnover associated with parietal cell secretion (49). At present, no other theory explains how ezrin works in other cells. Thus we tested the rolling motor model, focusing on other microvilli-rich cells: renal proximal tubule cells and enterocytes of small intestine.

Ezrin from both renal proximal tubule cells and intestinal enterocytes were found to be highly phosphorylated, compared with the low steady-state level of phosphorylation in the nonsecreting gastric parietal cells (Fig. 1). Both the renal proximal tubule cells and small intestinal enterocytes are known for the presence of their characteristic apical brush border membranes, which are densely packed with uniformly sized microvilli. In contrast, the microvilli found in nonsecreting gastric parietal cells normally exist in various lengths and sparse distribution (13). Upon stimulation, there is a massive membrane recruitment from intracellular tubulovesicles onto the apical canalicular membrane, forming a much denser distribution of elongated apical microvilli (13). Previously, we reported a significant increase of ezrin phosphorylation (40) and, specifically, phosphorylation of ezrin on T567 (49), in gastric parietal cells when these cells were physiologically stimulated. Thus, a higher steady-state level of ezrin T567-P seems to be correlated with a denser distribution of microvilli, whether from the same cell (e.g., the different physiological conditions of parietal cell) or from different types of tissue.

T567 phosphorylation turnover is required for ezrin function in renal proximal tubule cells.

The prompt decrease in ezrin T567-P with azide treatment in renal proximal tubule cells and small intestine enterocytes indicates that continuous kinase activity is needed to maintain the high level of phosphorylation on ezrin T567. The importance of T567-P turnover for the regulation of ezrin function in renal proximal tubule cells is demonstrated with the expression of the T567D mutant, which mimics permanent phosphorylation of T567, not allowing the turnover of ezrin between phosphorylated and dephosphorylated forms. When T567D mutant ezrin was overexpressed in primary kidney cell cultures, surface membrane structures were abnormally long and tended to accumulate at a small area on the plasma membrane, where the T567D mutant was localized. This morphological change obviously affected the fine structure of microvilli and may even disrupt the polarized distribution of the membrane proteins. Since the precise localization of channels and pumps on the plasma membrane(s) is essential for the functional transport of nutrients from the glomerular filtrate back into circulation, it is conceivable that the abnormal morphology of the renal proximal tubule cells would result in a decreased efficiency of function. Unfortunately, this experiment cannot be done with the primary cultures of proximal tubule cells expressing T567D mutant since the cells do not maintain their polarity in culture. However, experiments done with a polarized kidney epithelial cell line NRK-52E demonstrated an abnormal relocation of Na-K-ATPase onto the apical membrane upon treatment with RhoA (23), which is known to induce phosphorylation of ERM proteins on the conserved T567/T564/T558 site (25, 33).

The T567-P high turnover mechanism for regulation of ezrin activity is thus not limited to gastric parietal cells but also applies to ezrin in renal proximal tubule cells and small intestine enterocytes. In addition, moesin, which is expressed at relatively high levels in renal proximal tubules, was also subjected to the high turnover regulation on the phosphorylation of the conserved T558 (Fig. 2B). Thus it may be that this high turnover mechanism is a universal one for the regulation of all ERM protein activity.

Ezrin brings membrane to cytoskeleton.

Studies with transformed cell lines often described ezrin as a diffusely localized, or cytosolic, protein, in its inactive form (4, 5, 42). Since ezrin has increased membrane and cytoskeleton binding affinity upon activation (by phosphorylation), the major function of ezrin was believed to be “linking F-actin to membrane” (6). However, our examination of the state and localization of ezrin in three freshly isolated tissues suggests a modification of this concept. Phosphorylation on T567 was observed to induce tighter membrane binding of ezrin in gastric parietal cells (49), as expected from many other studies (12, 19, 45), since the membrane-binding NH₂ terminus is partially masked by the COOH terminus in the nonphosphorylated N-C binding conformation (14, 48). However, ezrin in its nonphosphorylated form still localizes to the microvilli-rich apical membrane of parietal cells as shown previously (49) and in Fig. 4A in the present study. Similarly, in renal proximal tubule cells and small intestine enterocytes, the dephosphorylated form of ezrin remains localized on the brush border microvilli-rich membrane (Figs. 5 and 6). Binding analyses showed that dephosphorylation by anoxia treatment would only slightly shift the distribution of ezrin toward the cytosolic pool, which is very small (Fig. 4C). Severe alterations in the cytoskeleton and the repolarization of many membrane proteins occur rapidly after ischemia and reperfusion in the kidney (26–29). Although the mechanism of these transitions is not completely understood, it is known that ischemia alone does not cause the rearrangement of F-actin cytoskeleton. While reperfusion after 40 min of ischemic treatment caused significant decrease in F-actin, the remaining F-actin signal still largely remained on the brush border membrane (7). Our procedures to induce anoxia or hypoxia, and the observed results, were obviously similar to the renal ischemia treatments without reperfusion.

The binding of T567 unphosphorylated ezrin to the membrane seems to be mediated by PIP₂. Tsukita and colleagues (44) found that treatment of two cell lines A431 and MDCK II with staurosporin could decrease the T567-P without affecting the membrane localization of ezrin. However, when the PIP₂ levels in A431 cells were decreased by microinjection of C3 transferase (which blocks the activation of Rho and PI4,5-kinase), ezrin was found dephosphorylated on T567 and diffused away from microvillus membranes. The PIP₂ level in MDCK II did not respond to C3 transferase but could be blocked with neomycin, which also caused a decreased T567-P level and diffusive staining of ezrin. The dependence of ezrin function on PIP₂ binding was also reported by Arpin's group (12). Their results with the LLC-PK1 cell line indicated that ezrin binding to PIP₂, through its NH₂-terminal domain, is required for membrane localization and T567 phosphorylation. It may well be that the membrane binding site on ezrin has some dependence on ionic bonding because the depletion of ezrin in permeabilized glands and tubules was increased by high salt (Fig. 3), consistent with previous observations in gastric glands (18), and as might be inferred by cytosolic divalent ion competition for ezrin release in glands (43) and in LLC-PK1 cells (12); however, in the latter case the authors have suggested that increased cytosolic Ca²⁺ might catalyze PIP₂ hydrolysis.

In contrast to the continuous binding to the membrane, the binding of ezrin to F-actin was greatly enhanced when the COOH-terminal actin-binding site was unmasked by phosphorylation at T567. Mandel et al. (9) demonstrated that dissociation of ezrin from actin cytoskeleton occurred in renal proximal tubules upon anoxia treatment. They later showed evidence that anoxia treatment induced ezrin dephosphorylation by two-dimensional electrophoresis (8). Very likely this dephosphorylation occurs on T567, although we cannot exclude the possibility of other Ser/Thr phosphorylation site(s). By F-actin cosedimentation assays (49) and Triton extraction analyses (Fig. 3), dissociation of ezrin from F-actin was found to be induced by dephosphorylation on T567 in gastric parietal cells. Similar data were obtained from many other cell models (12, 20, 30, 31, 37, 45, 49).

A relatively constant membrane binding and the ever changing F-actin binding gives the ezrin molecule the power to associate and reassociate membrane and cytoskeleton, thus enabling a dynamic reposition of membrane along the filamentous actin length, in a fashion similar to a rolling molecular motor attached to cell membrane. A cartoon depicting the relaxation-reattachment hypothesis to promote dynamic rearrangement within defined surface morphology is shown in Fig. 9. Such a model is logical and essential for surface plasticity in view of the growth and repositioning required for cytoskeleton expansion, as well as the reestablishment of membrane surface forces as processes of recruitment and endocytosis occur. This model also explains why blocking moesin dephosphorylation with CLA caused the impairment of neutrophil motility (46), which is crucial to effective host defenses against microorganisms.

Fig. 9.

Cartoon of relaxation-reattachment hypothesis for the function of the ezrin/radixin/moesin protein family. Plasma membranes (thick lines) are shown with microvillar projections and underlying actin microfilaments (thin lines) and ezrin molecules (red dots) in the membrane-filament attachment mode. At any given site or time, regulatory kinase and/or phosphatase activity would relax membrane filament attachments. A: filament elongation and growth of actin cytoskeleton such as occur with microvilli, filopodia, of lamellipodial extension. The growth of actin filaments without insertion of new membrane imposes surface tension on the plasma membrane. Dynamic ezrin-cytoskeleton dissociation and reassociation would occur via turnover of T567-P, enabling relaxation of tensile forces and reattachment to mold the new surface. B: membrane recruitment provides increased membrane surface area that is remolded by the volatile turnover of ezrin T567-P, promoting a dynamic reposition of membrane along the filamentous actin substructure.

In summary, higher steady-state phosphorylation levels on ezrin T567 were detected with freshly isolated renal proximal tubule cells and small intestine enterocytes, as compared with gastric parietal cells. However, high levels of phosphorylation did not exempt ezrin in these tissues from the high turnover regulation on T567-P, which assures a relaxation change in the association of ezrin between the F-actin-bound and -unbound state, without necessarily affecting the binding of ezrin to membrane. Interruption of the regulation of high turnover of T567-P, such as introduction of the ezrin T567D phosphorylation mimic, has been observed to cause abnormal growth and rearrangement of membrane projections (15, 33), sometimes causing local changes in polarity (23, 47), and likely to interfere with the normal membrane transport processes of the affected cells.

GRANTS

This work was supported by a grant from the National Institutes of Health, 5RO1DK10141-42.