Actin remodeling requires ERM function to facilitate AQP2 apical targeting

ERM (ezrin/radixin/moesin) proteins cross-link actin filaments with the plasma membrane. They are involved in the formation of microvilli, cell-cell adhesion, maintenance of cell shape, cell motility and membrane trafficking. Recent analyses reveal that they are not only involved in cytoskeleton organization but also in signaling pathways (Ivetic and Ridley, 2004; Speck et al., 2003).

ERM proteins share a high degree of homology among themselves. However, their tissue distribution and primary structure suggest that these are not simply redundant proteins (Tsukita and Yonemura, 1997). ERM proteins promote the organization of the cortical actin-based cytoskeleton through their C-terminal domain and of the cell membrane through the N-terminal FERM domain (Bretscher et al., 2002; Jankovics et al., 2002; Mangeat et al., 1999; Tsukita and Yonemura, 1999). The FERM domain of ERM proteins interacts directly with the intracellular domain of several membrane proteins, as well as indirectly through apical scaffolding proteins (Reczek and Bretscher, 1998). In the C-terminal region of the ERM family proteins, an actin binding site has been identified. In particular, a site interacting with F-actin and highly conserved among ERM members has been detected in the 34 C-terminal amino acids using a truncated ezrin fused with glutathione-S-transferase (Turunen et al., 1994). The association between the ERM proteins and the underlying cytoskeleton is crucial for determining cell shape, apical junction (AJ) assembly during development, motility and several plasma membrane processes, including endocytosis and exocytosis (Mangeat et al., 1999; Tsukita and Yonemura, 1999; Van Furden et al., 2004). Indeed, endocytosis and exocytosis are processes that are strictly dependent on actin dynamics, and ERM proteins participate not only in controlling actin dynamics but also in signal transduction pathways (Bretscher et al., 2002; Ivetic and Ridley, 2004). Interestingly, high level expression of the carboxyl-terminal domain (aa 319-583) of radixin disrupts normal cytoskeletal structure and function and causes a strong desegregation of ventral actin filaments in NIH-3T3 cells (Henry et al., 1995). Small GTP binding proteins of the Rho family are key players in regulating actin assembly and ERM proteins play an important role in the activation of members of the Rho family by recruiting their regulators. The FERM domain of ERM binds signaling molecules of the Rho pathway, including Rho guanine dissociation inhibitors (Rho-GDI) and the Rho GDP/GTP exchange protein Dbl (Bretscher, 1999). The observation that ERM proteins function both upstream and downstream of Rho GTPases implies that there could be a feedback loop for Rho-pathway autoregulation (Hirao et al., 1996; Mammoto et al., 2000; Takahashi et al., 1997). The ability of ERM to reversibly recruit the regulators of the Rho family is a consequence of the molecular conformation. In fact, the intramolecular interaction between the N- and C-terminal domains of ERM proteins masks several binding sites including signaling molecules of the Rho pathway, leading to a dormant protein in closed monomers and the intermolecular interaction in head-to-tail hetero- and homo-oligomers (Gary and Bretscher, 1993; Gautreau et al., 2000; Ivetic and Ridley, 2004). We have recently shown that cAMP-dependent translocation of the water channel aquaporin-2 (AQP2) into the apical membrane in renal principal cells is associated with RhoA inactivation, resulting in actin depolymerization (Klussmann et al., 2001; Tamma et al., 2001; Tamma et al., 2003). This event is a prerequisite for proper targeting of AQP2-bearing vesicles into the apical membrane of renal collecting duct cells, leading to an increase in osmotic water permeability. These observations suggest that actin could negatively regulate AQP2 targeting, forming a physical barrier that must be removed for vesicular fusion to occur. However, while actin cytoskeleton and actin-regulatory proteins have an increasingly recognized role in the assembly of vesicular transport and in their translocation, the mechanisms underlying this regulation are still largely unknown. ERM proteins proposed to link transmembrane proteins to the actin cytoskeleton in the apical domain regulate cell signalling events that affect actin organization.

This study represents the first report investigating the functional involvement of moesin, one of the ERM proteins expressed in renal principal cells in the regulation of AQP2 trafficking.

CD8 cells were established by stably transfecting the RC.SV3 rabbit cortical collecting duct cells with cDNA encoding rat AQP2. CD8 cells were grown at 37°C as described in a hormonally defined medium (Valenti et al., 1996). CD8 cells respond to cAMP-elevating compounds like vasopressin or forskolin with an increase in the osmotic water permeability coefficient (Pf) and the subsequent AQP2 redistribution from an intracellular pool to the apical surface (Tamma et al., 2001; Valenti et al., 1996; Valenti et al., 2000; Valenti et al., 1998). Confluent monolayers were used at days 3-5 after plating.

A peptide (ERM peptide) corresponding to a short sequence in the carboxyl terminal region of the ERM family (RDKYKTLRQIRQGN), including the threonine 558 residue, was synthesized (Fig. 2). This peptide is connected by a disulphuric bond to the cysteine of a chain comprising 20 hydrophobic amino acids which may constitute a permeable helix through the plasma membrane. A dansyl residue was added to the hydrophobic chain to visualize the presence of the peptide in the cells. In addition, a control peptide with a reversed sequence with respect to the peptide described above was synthesized (NGQRIQRLTKYKDR). To get the peptide into the cells, it was employed at a final concentration of 2 μM in the cell culture medium and incubated for 3 hours. Fluorescence analysis then confirmed that it was internalized into the cells, probably by endocytosis.

CD8 cells were grown on glass coverslips and fixed with 4% paraformaldehyde in phosphate buffer saline (PBS) for 20 minutes. Cells were washed 3 times for 5 minutes in PBS and permeabilized with 0.1% Triton X-100 in PBS for 10 minutes. After blocking in 0.1% gelatin in PBS for 20 minutes, cells were incubated at room temperature for 1 hour, either with affinity-purified anti-AQP2 antibodies or with monoclonal anti-moesin or with anti-phosphorylated moesin (Thr 558) antibodies. Cells were washed three times for 5 minutes with 0.1% gelatin in PBS and incubated for 60 minutes with fluorescein-conjugated goat anti-rabbit/mouse IgG (10 μg/ml in PBS), followed by washing twice for 1 minute in PBS containing 2.7% NaCl (PBS high salt) and twice in regular PBS. Alternatively, after blocking, the actin cytoskeleton was visualized by incubation with phalloidin-TRITC (100 μg/ml, 45 minutes). The coverslips were mounted in 50% glycerol in 0.2 M Tris-HCl, pH 8.0, containing 2.5% n-propyl gallate to retard quenching of the fluorescence. Moesin or the actin cytoskeleton were detected by epifluorescence microscopy (Leica DMRXA microscope). Alternatively, AQP2 and moesin were detected with an epifluorescence microscope (TE 2000S, Nikon Instruments, Florence, Italy) equipped with a CCD camera (Princeton Instruments MicroMax 512BFT, NJ) using a Delta RAM Highspeed Multiwavelength Illuminator for excitation (Photo Technology International PTI, NJ). The xz planes were obtained by deconvolution using Autodeblur software (Universal Imaging Corporation, PA). The fluorescence intensity was analyzed by using Metamorph software and the statistical analysis was performed by using a one-way Anova and Tukey's multiple comparison test.

Actin polymerization was analyzed as previously described (Hall et al., 1988; Knetsch et al., 2001; Peracino et al., 1998). Briefly, CD8 cells were left untreated or stimulated with forskolin (10^–4 M). Alternatively, cells were either preincubated with the ERM peptide (2 μM) for 3 hours or pretreated with the control peptide. The treatments were stopped by adding 450 μl of 3.7% paraformaldehyde, 0.1% Triton X-100, 0.25 μM TRITC-phalloidin in 20 mM potassium phosphate, 10 mM PIPES, 5 mM EGTA and 2 mM MgCl₂, pH 6.8. After staining for 1 hour the cells were washed three times with PBS and 800 μl of methanol were added overnight. The fluorescence (540/565 nm) was read in a RF-5301PC fluorimeter. The values obtained were analyzed by Student's t-test.

The interaction of moesin with actin cytoskeleton was measured by its solubility in Triton X-100. Briefly, CD8 cells were seeded 2 days before the experiments and grown to confluence. The Triton soluble fraction was extracted by a 1 minute incubation with 500 μl of a Triton X-100 buffer (80 mM PIPES, 5 mM EGTA, 1 mM MgCl₂, 0.5% Triton X-100, 50 mM NaF, pH 6.4, and calyculin 1 μM), which preserves the cytoskeleton association of proteins. Proteins were resolved in a 13% polyacrylamide gel and then transferred onto immobilon-P (Millipore) by standard procedures.

CD8 cells were homogenized with a glass/Teflon homogenizer in ice-cold buffer containing 130 mM KCl, 20 mM NaCl, 1 mM MgCl₂, 10 mM Hepes, pH 7.5, and 1 mM PMSF, 2 mg/ml leupeptin and 2 mg/ml pepstatin A. Nuclei were removed by centrifugation at 800 g for 10 minutes. Membrane and cytosol fractions were obtained by centrifugation for 1 hour at 4°C at 150,000 g in Beckman Instruments ultracentrifuge. The pellets (particulate fractions) were resuspended in homogenization buffer and the supernatant (soluble fractions) was concentrated using centricon tubes (10,000 Da cut-off).

Because ERM protein can reversibly interact with actin, and Rho-dependent actin rearrangements occur during forskolin stimulation in renal CD8 cells, the localization of the ERM protein moesin was analyzed in CD8 cells at rest and in forskolin-stimulated cells.

Immunofluorescence localization of moesin in CD8 cells revealed a predominant expression at cell adhesion sites (Fig. 1A, CTR). These results are in agreement with previous observations showing that ERM proteins are localized at specific regions where actin filaments interact with plasma membranes, organizing actin enriched structures like microvilli, cleavage furrows and ruffling membranes. In forskolin-stimulated cells, the antibody stained mainly cell-surface structures that may represent apical microvilli, in agreement with similar localization in other cell lines (Hayashi et al., 1999; Henry et al., 1995; Yonemura et al., 1998) (Fig. 1A, FK). Western blotting analysis was performed following cell fractionation. In control cells, the majority of immunodetectable moesin was associated with the soluble fraction (S), compared with the particulate fraction (P). Forskolin stimulation resulted in a significant increase in moesin in the particulate fraction and was associated with a parallel decrease in the soluble fraction (Fig. 1B and densitometric analysis shown on the right). These findings suggest a relocalization of moesin from the soluble to the particulate fraction upon forskolin stimulation.

ERM proteins consist of three domains: a globular NH₂-terminal membrane-binding domain followed by an extended α-helical domain and a positively charged COOH-terminal domain. The C-terminal region of ERM proteins has a very high affinity for F-actin through its major actin-binding sites, which consist of 34 amino-acids and which are highly conserved among these proteins. Nevertheless, the lack of six C-terminal amino acids completely prevents any actin binding (Turunen et al., 1994). To investigate the role of moesin in regulating the actin cytoskeleton in CD8 cells, a peptide corresponding to a highly conserved sequence in the ERM family at the C-terminal domain (ERM peptide) was synthesized and introduced into CD8 cells as described by Oehlke (Oehlke et al., 1998; Scheller et al., 1999) (Fig. 2A). A peptide having a reversed sequence with respect to the ERM peptide was also synthesized, and was used as a control (Fig. 2A, control peptide).

The actual peptide internalization into the cells was confirmed by its fluorescence localization (Fig. 2B). Preincubation of the cells with the ERM peptide was associated with a strong, and time-dependent, decrease in moesin content in the soluble fraction. By contrast, preincubation with the control peptide for 3 hours had no effect on moesin abundance in the soluble fraction. In addition, forskolin treatment resulted in a decrease in moesin abundance in the soluble fraction (Fig. 2C). As we have previously shown that forskolin stimulation results in a partial depolymerization of actin cytoskeleton (Klussmann et al., 2001; Tamma et al., 2001; Tamma et al., 2003), we next tested whether the effect of the ERM peptide is associated with actin remodeling. Incubation of CD8 cells with the ERM peptide for 3 hours resulted in a dramatic reduction in stress fibers in the absence of forskolin stimulation (Fig. 3A, +ERM peptide), an effect similar to that obtained in forskolin-stimulated cells (Fig. 3A, FK, –peptide). Pretreatment with the control peptide had no effect on actin organization (Fig. 3A, +control peptide). Semi-quantitative analysis of the amount of F-actin evaluated with the actin polymerization assay confirmed that F-actin content significantly decreased on preincubation with the ERM peptide (58.04±8.49 peptide; 85.13±2.33 control, P<0.01), an effect similar to that obtained with forskolin stimulation (forskolin 62.58±2.90, P<0.001) (Fig. 3B). Introduction of a control peptide with a reversed sequence had no effect on F-actin content (Fig. 3B, control peptide).

Actin depolymerization is an important prerequisite for cAMP-dependent translocation of the water channel protein AQP2 to the plasma membranes in renal principal cells (Klussmann et al., 2001; Tamma et al., 2001; Valenti et al., 2000). Therefore, we next tested whether a plasma membrane relocalization of the water channel AQP2 occurs upon peptide preincubation. Immunofluorescence experiments were performed in the presence of the ERM and the control peptides. Preincubation with the ERM peptide induced a partial relocalization of AQP2 to the plasma membrane, which was completed upon forskolin treatment (Fig. 4, ERM peptide and xz reconstruction in the inset). As expected, forskolin stimulation caused AQP2 translocation from an intracellular pool to the apical membrane coincident with a decrease in intracellular staining due to the relocalization of AQP2 to the apical plasma membrane (Fig. 4, FK and xz reconstruction in the inset). Pretreatment with the control peptide did not affect the cellular localization of AQP2 (Fig. 4, control peptide and xz reconstruction in the inset).

AQP2 translocation, in the presence of the ERM peptide, might be due to the uncoupling effect of endogenous moesin from its effector/s, leading to actin depolymerization. Because phosphorylation of conserved threonine in ERM proteins influences the differential capacity to associate with actin cytoskeleton, we next evaluated the association dynamics between actin filaments and moesin in response to the introduction of the ERM peptide and during forskolin stimulation. To this end, we used an extraction procedure with Triton X-100 buffer, which preserves the cytoskeleton and the cytoskeleton-associated proteins. When Triton X-100 buffer-extracted material was compared by immunoblotting, we found that, relative to the control, the amount of actin-dissociated moesin was significantly increased both upon forskolin stimulation (densitometry of the signals: 2.667±0.48 fold, n=9, P<0.05 compared with control) and also after the ERM peptide treatment (3.84±0.64 fold, n=4, P<0.001). Interestingly, a similar effect was observed in response to the selective inhibitor of Rho kinase, Y27632 (3.083±0.31 fold, n=3, P<0.05 compared with control) (Fig. 5A). Compared with the control, preincubation with the control peptide had no effect on the amount of actin dissociated moesin (1.11±0.11 fold, n=3, P>0.05). The distribution of endogenous P-Thr-moesin was then examined in the Triton-soluble fractions. Compared with the control condition no significative difference in the band intensity was observed when the cells were either stimulated with forskolin or preincubated with the ERM peptide, whereas the phosphorylation level decreased significantly on Y27632 treatment (Fig. 5B). When the amount of P-Thr-moesin was normalized to the total amount of moesin of each Triton-soluble fraction, it seemed clear that overall, dissociation of moesin from actin cytoskeleton was associated with a consistent reduction in P-Thr-moesin abundance (Fig. 5C). In addition, these data confirm that actin depolymerization is paralleled with a dissociation of moesin from the actin cytoskeleton. Immunofluorescence localization of P-Thr-moesin revealed that at rest, it was localized not only at the cell-cell adhesion sites but also along intracellular filaments depicting actin stress fibers (Fig. 5D, CTR, arrowheads). This particular staining, as well as that at the cell-cell adhesion sites, disappeared after forskolin stimulation (Fig. 5D, FK).

In this study we propose that moesin, a member of the ERM family known to cross-link actin filaments with plasma membranes, and to be implicated in signalling pathways, is a new player regulating cAMP-induced AQP2 trafficking in renal principal cells. We report that forskolin-induced elevation of intracellular cAMP is associated with a redistribution of the endogenous ERM protein moesin, from the soluble (inactive) to the membrane (active) compartment. The functional role of moesin in cAMP-induced AQP2 translocation was analyzed by uncoupling moesin from its putative downstream effector. To this end, a peptide of 14 amino acids within the binding site for F-actin, reproducing the C-terminal domain of moesin, was synthesized and introduced into the cells. The designed peptide contains threonine 558, a residue that is thought to stabilize moesin in an active (open) state when phosphorylated.

Peptide introduction in CD8 cells induced a decrease in F-actin content, an effect similar to that obtained with forskolin stimulation. Notably, this caused AQP2 relocation to the plasma membrane which was associated with a time-dependent decrease of moesin abundance in the soluble fraction.

Therefore, incubation with the ERM peptide mimicked some of the key effects associated with hormonal or forskolin stimulation, including a decrease in F-actin content, moesin translocation from a soluble compartment to a particulate fraction, and AQP2 translocation to the cell surface. This suggests that moesin itself is involved in the signal transduction cascade leading to AQP2 targeting by having a possible role in the organization of cortical actin cytoskeleton.

Actin remodeling plays a crucial role in vesicular AQP2 trafficking. Indeed, we have previously shown that RhoA inhibition with specific toxins also causes depolymerization of the actin network, resulting in AQP2 translocation in the absence of hormonal stimulation (Klussmann et al., 2001; Tamma et al., 2001). ERM proteins are activated by signals mediated through Rho (Matsui et al., 1998; Matsui et al., 1999; Ramesh, 2004; Shaw et al., 1998). ERM proteins positively regulate Rho activity by binding to RhoGDI and releasing the inactive Rho from GDI, thereby allowing activation of Rho (Ramesh, 2004). The interaction between the regulatory factor RhoGDI occurs at the N-terminal domain of ERM proteins only when they are active (Hirao et al., 1996; Takahashi et al., 1997). As mentioned previously, phosphorylation of ERM stabilizes the proteins in an open (active) conformation which does not affect the ability to interact with F-actin but rather prevents intramolecular interaction with the FERM domain, by stabilizing the activated ERM proteins (Matsui et al., 1998; Yonemura et al., 2002). We might expect that the introduction of the exogenous peptide, which contains threonine 558, phosphorylated by either Rho kinase or by protein kinase C, prevents the phosphorylation of the endogenous and functional proteins. The impairment of endogenous phosphorylation might influence the balance between the active and inactive forms of the ERM family proteins with the consequent release of regulatory factors such as RhoGDI. The RhoGDI released could directly interact with and inactivate Rho proteins leading to partial depolymerization of the actin cytoskeleton which facilitates AQP2 insertion (Klussmann et al., 2001; Tamma et al., 2001).

To conclude, we provide here the first evidence for a putative functional involvement of the ERM family protein moesin in AQP2 trafficking. Our data point to a dual role of moesin in this process: it might modulate actin depolymerization through Rho signalling due to its ability to reversibly bind RhoGDI and it participates in the reorganization of F-actin-containing cytoskeletal structures close to the AQP2-bearing vesicles fusion sites. Both processes seem to be reversibly modulated by moesin phosphorylation.

This work was supported by a grant from the EU-TMR network (proposal n° ERB 4061 PL 97-0406), from the Italian `Ministero della Ricerca Scientifica e Tecnologica' and from the CEGBA (Centro di Eccellenza di Genomica in campo Biomedico ed Agrario), from the Vigoni program (1999-2001). W.R. and E.K. were supported by the Deutsche Forschungsgemeinschaft (Ro 597/6, Ro597/9-1 and Kl1415/1), the European Union (QLK3-CT-2002-02149 and QLRT-2000-00987), and the Fonds der Chemischen Industrie. We thank our colleague Anthony Green for proofreading and providing linguistic advice.