The Carboxyl Terminus of the α-Subunit of the Amiloride-sensitive Epithelial Sodium Channel Binds to F-actin*

The activity of the amiloride-sensitive epithelial sodium channel (ENaC) is modulated by F-actin. However, it is unknown if there is a direct interaction between α-ENaC and actin. We have investigated the hypothesis that the actin cytoskeleton directly binds to the carboxyl terminus of α-ENaC using a combination of confocal microscopy, co-immunoprecipitation, and protein binding studies. Confocal microscopy of Madin-Darby canine kidney cell monolayers stably transfected with wild type, rat isoforms of α-, β-, and γ-ENaC revealed co-localization of α-ENaC with the cortical F-actin cytoskeleton both at the apical membrane and within the subapical cytoplasm. F-actin was found to co-immunoprecipitate with α-ENaC from whole cell lysates of this cell line. Gel overlay assays demonstrated that F-actin specifically binds to the carboxyl terminus of α-ENaC. A direct interaction between F-actin and the COOH terminus of α-ENaC was further corroborated by F-actin co-sedimentation studies. This is the first study to report a direct and specific biochemical interaction between F-actin and ENaC.

The amiloride-sensitive epithelial sodium channel (ENaC) 2 is a member of the degenerin/epithelial sodium channel superfamily of ion channels. ENaC is expressed at the apical surface of polarized epithelia and is in part responsible for maintaining proper salt and water homeostasis in the body. A great deal of information is known about the biophysical properties of ENaC once it is inserted into the apical surface of an epithelial cell plasma membrane. However, less is known about the proteins that interact with ENaC. Data from the literature indicate an interaction between ENaC and components of the apical membrane cytoskeleton. A partially purified ENaC complex from bovine renal epithelia copurifies with ankyrin, spectrin, and actin (1), suggesting that these cytoskeletal proteins may be associated with ENaC. In addition, ␣-rENaC has been shown to bind to ␣-spectrin, and this is mediated through direct interaction between the ␣-spectrin Src homology 3 domain and the second proline-rich region in the COOH terminus of ␣-rENaC (2). Electrophysiological data provide further support for an interaction between ENaC and the actin-based cytoskeleton. In cellattached patches of A6 renal epithelial cells treated with the actin filament disrupter cytochalasin D, an induction of ENaC activity was observed (3), thereby suggesting that changes in the actin cytoskeleton affect the activity of ENaC. ENaC activation was also observed when short F-actin filaments were added to excised patches, and this effect was increased with the addition of cytochalasin D and/or ATP. These effects were reversed by the addition of the G-actin binding protein, DNase I. In planar lipid bilayers, short F-actin filaments were demonstrated to increase the open probability of rENaC (4), whereas application of DNase I prevented the activation of rENaC. The application of gelsolin, a Ca 2ϩ -activated protein that severs actin filaments and caps the plus end of the actin filament, preventing the repolymerization of actin and keeping it in a gel-like state, was found to cause a sustained activation of rENaC. In addition, actin was required for the transient activation of rENaC by protein kinase A and ATP when ENaC was reconstituted into planar lipid bilayers. These data indicate that a direct interaction between actin and ENaC may underlie the regulation of ENaC by short actin filaments. Identification of regions involved in a direct protein interaction between actin and ␣-ENaC has only been deduced from biophysical methods. Current evidence suggests that actin interacts with the C-terminal domain of ␣-ENaC. A C-terminal truncation mutant (R613X) of ␣-rENaC was not responsive to the addition of actin. The single channel recordings of the ␣ R613X ␤␥-rENaC treated with actin were the same as wild type ␣␤␥-rENaC that was not treated with actin (5). The deletion of 14 amino acids, Glu 631 -Phe 644 , in a chimeric rat-bovine ␣-ENaC (␣-rbENaC), consisting of ␣-rENaC residues 1-615 and ␣-bENaC residues 570 -650, nullified the effect of actin normally seen with the chimeric ␣-bENaC (6). The deleted 14-amino acid sequence of ␣-bENaC has an 11-amino acid sequence identity to the same region of amino acids in the COOH terminus of ␣-rENaC. This high degree of sequence identity suggests that this amino acid sequence in ␣-rENaC may also participate in the regulation of ENaC by actin. However, to date, there is no definitive biochemical evidence for a direct interaction between F-actin and ␣-ENaC.
In order to investigate the hypothesis that the actin cytoskeleton interacts directly with the carboxyl terminus of ␣-ENaC, we have used a combination of gel overlay and F-actin co-sedimentation assays to demonstrate binding of actin to the COOH terminus of ␣-ENaC. Actin was found to co-immunoprecipitate with ␣-ENaC from MDCK cell lysates, thereby providing in vivo data supporting an association between actin and ENaC. Moreover, co-localization of actin and ␣-ENaC in the apical membrane of MDCK cells stably expressing functional ENaC was demonstrated using laser-scanning confocal microscopy. These three independent lines of evidence support an interaction between actin and ␣-ENaC, which is mediated by the direct binding of actin to the carboxyl terminus of ␣-ENaC.

MATERIALS AND METHODS
Cell Culture-Stably transfected MDCK cells expressing the rat isoforms of ␣␤␥-ENaC and MDCK parental cells were obtained as a kind gift from Drs. R. G. Morris and J. A. Schafer. The ␣-rENaC subunit was tagged with a single FLAG tag in the extracellular loop as described by Firsov et al. (7); both ␤and ␥-subunits were wild type. Cells were grown at 37°C in a 5% CO 2 humidified incubator. Cells were initially grown in T75 flasks in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum (HyClone) and antibiotics (penicillin and streptomycin (1%), G418 (800 g/ml), hygromycin (300 g/ml), and puromycin (5 g/ml)). Once the cells were at least 80% confluent, cells were seeded on poly-L-lysine-coated semipermeable supports (Transwell; 24-mm diameter, 0.4-m pore size; Costar). Cells were used ϳ5-14 days later when polarized monolayers were formed. MDCK cells used for Western blots and co-immunoprecipitations were plated in 100-mm plastic Petri dishes. In some experiments, in order to increase ENaC expression, the media were supplemented with 2 M dexamethasone (Sigma) and 2 mM sodium butyrate (Fluka) 24 h before the cells were harvested.
Electrophysiology-Whole cell clamp bath solution was RPMI culture medium. The pipette solution was 100 mM potassium gluconate, 30 mM KCl, 10 mM NaCl, 20 mM HEPES, 0.5 mM EGTA, free Ca 2ϩ less than 10 nM, 4 mM ATP, at a pH of 7.2. The bath contained serum-free RPMI 1640 cell culture medium (133 mM Na ϩ , 5.3 mM K ϩ , 108.3 mM Cl Ϫ ). These solutions approximate normal ionic gradients in situ. After formation of a gigaohm seal, the membrane within the seal was ruptured by an additional suction pulse. The whole-cell configuration was confirmed by an increase in the capacitance with no change in resistance. After the additional capacitance was balanced, the cells were held at Ϫ60 mV and clamped to membrane potentials ranging from Ϫ160 to ϩ40 mV sequentially for 1 s, returning to the holding potential (Ϫ60 mV) for 1 s between each test voltage. Currents were recorded digitally using pClamp hardware and software (Axon Instruments, Sunnyvale, CA). During bath perfusion to change to amiloride-supplemented medium, the cells were held at Ϫ60 mV and pulsed sequentially to Ϫ120 and ϩ60 mV for 0.5 s, returning to the holding potential between each test pulse (see Fig. 1C). This provided a continuous record and shows in real time the inhibition of inward current by amiloride. Single channel currents were recorded using the patch clamp technique. For cell-attached patches the pipette solution was RPMI culture medium, and for outside-out patches the pipette solution was 150 mM KCl. In all cases, the bath solution was RPMI culture medium. Data were recorded and analyzed using fetchex, fetchan, and pstat software (Axon Instruments, Sunnyvale, CA).
Immunocytochemistry-MDCK cells were grown on poly-L-lysinecoated semipermeable supports as described above. Culture medium was aspirated from the monolayer, cells were rinsed twice with 1ϫ phosphate-buffered saline (PBS) (137 mM NaCl, 27.7 mM KCl, 1.5 mM KH 2 PO 4 , 8 mM Na 2 HPO 4 ), and monolayers were then fixed with 3% paraformaldehyde (prepared from 20% EM Grade solution; Electron Microscopy Services) for 15 min at 37°C. Cells were rinsed three times with 1ϫ PBS and then permeabilized using 1ϫ PBS with 0.1% Triton X-100 (PBST) for 5 min at room temperature. The blocking step was done with 1ϫ PBS plus 10% normal serum (Jackson ImmunoResearch Laboratories) for 1 h at room temperature. The wheat germ agglutinin/ Alexa594 conjugate (WGA594) (Molecular Probes, Inc., Eugene, OR) was used at a concentration of 5 g/ml diluted in 1ϫ PBS and incubated with the samples for 10 min at room temperature. The anti-FLAG M2 antibody (Stratagene) was diluted 1:100 with 1ϫ PBST from a stock concentration of 2 mg/ml with 1ϫ PBST and incubated with the samples for 1 h at room temperature. Specificity of the anti-FLAG M2 antibody was demonstrated using the FLAG peptide (Sigma). The antibody was preabsorbed with the FLAG peptide, 10 g of peptide, 1 g of anti-FLAG M2 antibody for a minimum of 15 h prior to use. All phalloidin conjugates (Molecular Probes) were diluted 1:100 with 1ϫ PBST from a stock concentration of ϳ6.6 M and incubated with the samples for 1 h at room temperature. Monolayers were then washed with 1ϫ PBS, three times for 10 min at room temperature. Secondary antibody conjugated to Alexa488 or Alexa594 (Molecular Probes) was diluted 1:100 with 1ϫ PBST from a stock of 2 mg/ml and incubated with sample for 1 h at room temperature. After application of primary/probe and secondary antibody, samples were washed in 1ϫ PBS three times for 10 min each at room temperature. Counterstaining was performed using Hoechst 33258 (20 g/ml in 1ϫ PBS) for 3-5 min at room temperature. Cells were rinsed once with 1ϫ PBS. Monolayers were mounted with 1% para-phenylenediamine in 1:9 (v/v) 1ϫ PBS/glycerol and coverslipped. Images were viewed using laser-scanning confocal microscopy (Leica DM IRBE microscope, mounted with a Leica TCS SP scanhead). The Leica DM IRBE microscope was equipped with oil, PlanApochromat 40ϫ, 63ϫ, and 100ϫ objectives. The 100ϫ objective with a numerical aperture of 1.4 was used to acquire the final images. Visualization of blue fluorophores (Hoechst 33258) was achieved by using a dedicated UV laser (Coherent) for excitation at 350 nm. Green fluorophore (Alexa488) excitation at 488 nm was achieved by using an argon laser (Leica). Red fluorophore (Alexa 594) excitation at 568 nm was achieved by using a krypton laser (Leica). Energy emission in the form of light by blue fluorophores (380 -494 nm), green fluorophores (500 -575 nm), and red fluorophores (596 -722 nm) was detected using three independent photomultiplier tubes. Color channels for the final double label images were captured sequentially and then merged using the Leica TCS NT software. Adobe Photoshop version 7.0 was used for image processing. All confocal microscopy was done at the University of Alabama High Resolution Imaging Facility and the Veterans Affairs Hospital (Birmingham, AL).
SDS-PAGE, Immunoprecipitations/Co-immunoprecipitations, and Immunoblotting-Whole cell lysates prepared from MDCK cells stably expressing ENaC and MDCK parental cells were used. Cells were grown in three 100-mm plastic Petri dishes until confluent. Cells were washed twice with 2 ml of cold 1ϫ PBS. Petri dishes were placed on ice for 10 min with 1 ml each of 1ϫ lysis buffer (50 mM Tris, pH 7.4, 150 mM NaCl, and 1% Triton X-100) supplemented with Complete protease inhibitor mixture (Roche Molecular Biochemicals). Cells were scraped from the Petri dishes and placed into microcentrifuge tubes on ice. Lysates were sheared a minimum of three times with a 22-gauge needle, and incubated on ice for 1 h. Sheared lysates were then spun at 15,800 ϫ g at 4°C for 5 min. The supernatant was removed, and a BCA protein assay (Pierce) was performed to quantify the amount of total protein in the samples. A maximum of 200 g of whole cell lysate diluted in 1ϫ PBS was used per immunoprecipitation reaction and incubated overnight at 4°C on a rotator with 3 g of anti-FLAG mAb (Stratagene), 40 l of a 50% slurry of protein G-agarose beads (Roche Applied Science) and Complete protease inhibitor mixture (final volume of 500 l). After overnight incubation, beads were centrifuged for 2 min at 15,800 ϫ g. The supernatant was aspirated, and beads were washed and pelleted three times (2 min at 15,800 ϫ g) in 1ϫ lysis buffer supplemented with Complete protease inhibitor mixture. Samples were diluted 1:1 (v/v) with Laemmli sample buffer (62.5 mM Tris-HCl, pH 6.8, 2% SDS, 25% glycerol, 0.01% bromphenol blue, 5% ␤-mercaptoethanol), heated at 95°C for 5 min, and separated by SDS-PAGE with constant voltage at room temperature. Proteins were transferred to polyvinylidene difluoride (PVDF) membranes (Bio-Rad). The Western transfer was done in running buffer (25 mM Tris base, pH 8.0, 192 mM glycine, 1% (w/v) SDS, 20% (v/v) methanol) at 4°C for 1 h at constant voltage. Blots were either blocked in 1ϫ TBST (10 mM Tris base, pH 8.0, 150 mM NaCl, 0.1% Tween 20) with 5% nonfat dry milk overnight at 4°C or for 1 h at room temperature. Following immunoprecipitation with the mAb anti-FLAG antibody, Western blot detection of ␣-rENaC was performed using anti-FLAG mAb (1:500 dilution) or a rabbit polyclonal anti-␣-ENaC antibody (2 g/ml final concentration) (Affinity BioReagents). Specificity of the anti-␣-ENaC antibody was demonstrated using its immunizing peptide (Affinity BioReagents). The antibody was preabsorbed with its immunizing peptide 2 g peptide/1 g of ␣-ENaC antibody for a minimum of 15 h prior to use. Primary antibodies were diluted with 1ϫ TBST, 1% nonfat dry milk and incubated with blots for 1 h at room temperature. Blots were washed a minimum of 3 times with 1ϫ TBST, 10 min each at room temperature. Secondary antibodies conjugated to horseradish peroxidase (Jackson ImmunoResearch Laboratories) were used at dilutions of 1:10,000 or 1:20,000 in 1ϫ TBST and incubated for 1 h at room temperature. Blots were washed a minimum of three times with 1ϫ TBST, 10 min each at room temperature. Immunoreactivity was visualized using enhanced chemiluminescence (Super Signal West Pico; Pierce) and imaged onto Eastman Kodak Co. X-OMAT AR film (Fisher). Controls for immunoprecipitations consisted of ChromPure normal IgG serum (Jackson ImmunoResearch Laboratories) added in a concentration equivalent to the primary antibody. Co-immunoprecipitation of ␣-rENaC with actin was performed as above with the following exceptions: 1) the lysis buffer used contained 50 mM Tris base, 15 mM Na 2 HPO 4 , 150 mM NaCl, 0.5% Triton X-100, 0.5% deoxycholate, pH 7.41; 2) following the BCA assay, the 200 g of supernatant was precleared with 50 l of protein A-agarose beads and 50 l of protein G-agarose beads for 2 h at 4°C; 3) after preclearing, the samples were spun, and the anti-FLAG antibody alone was added for overnight incubation at 4°C; 4) the ␣-rENaC/FLAG complex was captured by the addition of 40 l of protein G-agarose and incubated for 2 h at 4°C; 5) the beads were washed three times with the lysis buffer plus 500 mM KCl, pH 9.1; 6) the co-immunoprecipitation control consisted of an anti-His antibody (Santa Cruz Biotechnology, Inc., Santa Cruz, CA) added in a concentration equivalent to that of the primary anti-FLAG antibody. The monoclonal anti-actin antibody (Chemicon) was used at a 1:1,500 dilution for 1 h at room temperature. Immunoblotting and primary antibody detection for the co-immunoprecipitations was done exactly as described above.
Generation and Purification of GST Fusion Proteins-GST-␣-hENaC C-terminal fusion protein (a generous gift of Dr. F. J. McDonald, Dunedin, New Zealand), GST-␥-rENaC C-terminal, GST-␣-rENaC NH 2 -terminal, and GST were produced in E. coli BL21-CodonPlus-RP (Stratagene). Cells were induced with 0.1 mM isopropyl-␤-D-thiogalactopyranoside for 3 h at 37°C. Bacterial cultures were spun at 10,000 ϫ g for 10 min at 4°C. Bacterial pellets were snap frozen with liquid N 2 and stored overnight at Ϫ80°C. Pellets were lysed with BugBuster (Novagen) in the presence of benzonase (Novagen) and Complete protease inhibitor mixture (Roche Molecular Biochemicals) for 20 min at room temperature or at 4°C for 45 min and then spun at 16,000 ϫ g for 20 min at 4°C. Supernatant was poured over a glutathione-Sepharose 4B (AP Biotech) gravity flow column five times. Bound GST fusion protein or GST was eluted with 10 mM reduced glutathione (Sigma) in three 2-ml fractions. The purified fractions were assayed by SDS-PAGE, and the gel was stained with Gel Code (Pierce). The fractions were then pooled, concentrated, and dialyzed overnight at 4°C in 1ϫ PBS with three buffer changes.
Labeling and Polymerization of Nonmuscle Actin-Nonmuscle actin (Cytoskeleton) was labeled with Alexa Fluor 488 carboxylic acid succinimidyl ester (Molecular Probes) according to the manufacturer. The polymerization of the nonmuscle actin was done as previously described by Hitt et al. (8).
Gel Overlays-Samples were heated at 95°C for 5 min and separated using 12% Tris/glycine gels with constant voltage at room temperature. Proteins were transferred onto PVDF membrane in running buffer (25 mM Tris base, pH 8.0, 192 mM glycine, 20% (v/v) methanol) at 4°C for 1 h at constant voltage. Gel overlays were blocked in 1ϫ TBST, 5% nonfat dry milk overnight at 4°C. The blots were probed with 25 g/ml F-actin/Alexa488 in 1ϫ TBST, 5% nonfat dry milk for 2 h at room temperature, washed briefly in 1ϫ TBST, allowed to air-dry, and scanned using a FujiFilm FLA-5100 Scanner (FujiFilm). Images were processed using Adobe Photoshop 7.0.
Nonmuscle F-Actin Co-sedimentation-Nonmuscle F-actin co-sedimentation experiments were done according to the manufacturer's instructions (Cytoskeleton). All samples were incubated at room temperature for 30 min. Briefly, 40 l of a 23 M nonmuscle F-actin stock was incubated with 10 l of GST-␣-hENaC C-terminal fusion protein (0.4 g of full-length fusion protein). Negative controls consisted of 40 l of F-actin buffer incubated with 10 l of GST-␣-hENaC C-terminal fusion protein (0.4 g of full-length fusion protein) and an equal amount of GST protein that was used for the GST-␣-hENaC C-terminal fusion protein sample. Positive controls consisted of 1 l of ␣-actinin (2.5 g), 40 l of nonmuscle F-actin, 8 l of F-actin buffer, and 1 l of Tris-HCl. Following the 30-min incubation, samples were subjected to ultracentrifugation for 1.5 h, 150,000 ϫ g at 24°C. Supernatant was removed, and pellets were resuspended in 50 l of Laemmli sample buffer and heated at 95°C for 5 min. Equal volumes of supernatant and pellet were loaded into the wells, separated using 12% Tris/glycine gels with constant voltage at room temperature. Gels were stained using GelCode (Pierce).

RESULTS
Determination of Functional ENaC-Wild type MDCK cells do not endogenously express amiloride-sensitive epithelial sodium channel (9,10). MDCK cells stably transfected with rat ␣␤␥-ENaC were examined by whole-cell patch clamp. Three stably transfected MDCK cells (induced to increase expression of ENaC) were whole-cell-clamped. Each cell had inward sodium current that was inhibited by 2 M amiloride, thereby confirming the presence of ENaC in the plasma mem-␣-ENaC Binds to F-actin brane. Fig. 1 (A and B) shows a representative example of whole-cell currents before and after treatment with amiloride. Fig. 1C shows the amiloride block in a real time pulse record. Fig. 1D is the summary of the amiloride-blockable current-voltage data. The amiloride-sensitive current reversed at a positive membrane potential, indicating sodium selectivity. Fig. 1B shows the slight reduction in outward current after amiloride treatment. This may be due to inhibition of outward potassium flux through ENaC. The result is a shift to the left in the I-V relation.
In the cell-attached configuration, small single channel currents with relatively long open times were observed (Fig. 1E). Moreover, cAMP characteristically increased channel activity (NP o 0.5 versus 1.5). The single channel conductance was 8 picosiemens, which is typical of human ENaC channels (11). To confirm independently the identity of the single channels, outside-out patches were formed, and amiloride (100 nM) was added to the bath solution (Fig. 1F). 100 nM is slightly above the IC 50 (75 nM) for inhibition of ENaC. This concentration was chosen, because it would produce a partial inhibition of the channels, changing the long open times into rapidly transitioning channels due to the binding and unbinding of the inhibitor. Based on these observations (conductance and appropriate inhibition by amiloride), we conclude that the stably transfected MDCK cells expressed functional ENaC in their apical plasma membrane.
Co-localization of ␣-ENaC with the Cortical F-actin Cytoskeleton-Initially, stably transfected MDCK cells expressing the ␣-rENaC subunit with a FLAG epitope tag on the extracellular loop and wild type ␤and ␥-rENaC subunits, were labeled with a monoclonal anti-FLAG epitope antibody to determine the membrane localization of ␣-ENaC. Fig. 2, A-E, show representative images of the pattern of expression for ␣-rENaC in this cell line. As these images illustrate, the level of ␣-rENaC expression varied between cells. Fig. 2E is a XZ reconstruction, illustrating that ENaC immunoreactivity was primarily situated in the apical domain of this cell line. In contrast to the anti-FLAG antibody, the  MARCH 10, 2006 • VOLUME 281 • NUMBER 10

␣-ENaC Binds to F-actin
peptide competition controls (Fig. 2, F and G) exhibited little or no detectable immunoreactivity at the surface of the monolayer (Fig. 2F) or within the cytoplasm (Fig. 2G). In order to determine where the apical anti-FLAG immunoreactivity was located relative to the apical plasma membrane, double labeling experiments were performed. Fig. 2H is an XZ reconstruction of a monolayer labeled with anti-FLAG antibody to detect ␣-rENaC. The monolayer was labeled with wheat germ agglutinin (WGA) conjugated to Alexa594 (WGA594) prior to fixation to delineate the apical plasma membrane (Fig. 2I). Fig. 2J is an overlay of H and I, which demonstrates co-localization of ␣-rENaC with WGA within the apical plasma membrane.
To localize the cortical F-actin cytoskeleton with respect to the apical membrane in MDCK cells stably expressing ENaC, monolayers were labeled with WGA594 and the specific F-actin probe, phalloidin, conjugated to Alexa488. Fig. 3, A-D, are optical sections from a Z-stack of MDCK ENaC cells. As the optical sections were taken deeper into the the nuclei (D). E, an XZ reconstruction taken from the region demarcated by the white line in A. Immunoreactivity for ␣-rENaC is localized primarily in the apical domain. F and G, FLAG peptide competition control images. F, the signal due to the FLAG antibody was competed away when the FLAG antibody was preincubated with the FLAG peptide, demonstrating the specificity of this antibody. The XZ reconstruction in G has no detectable immunoreactivity in the apical domain when compared with E. H-J, confocal micrographs of an XZ reconstruction of a monolayer dual labeled for FLAG epitope-tagged ␣-rENaC (anti-FLAG antibody) (H) and apical plasma membrane (WGA/Alexa594) (I), merged in J. J shows co-localization of ␣-rENaC with the apical plasma membrane, revealing that ␣-rENaC is expressed in the apical plasma membrane of the MDCK ENaC cells. Nuclei were stained with Hoechst 33258. All settings for the lasers, Leica TCS SP scanhead, and Leica TCS NT software were the same when acquiring the images for the control and experimental samples. Scale bar, 10 m. The single label experiment is representative of work repeated five times, and the double label experiment is representative of work repeated three times. cells, the fluorescence from the WGA594 largely disappeared, leaving only the basolateral labeling of the phalloidin/Alexa488 in the majority of the field of view. An XZ reconstruction (Fig. 3, E-G) revealed colocalization of actin and WGA in some of the cells, demonstrating that a population of F-actin is associated with the apical membrane.
We then performed double labeling experiments to determine whether ␣-rENaC and F-actin are co-localized in this model system. The representative XZ reconstructions in Fig. 4, C and F, shows the overlay of the FLAG (Fig. 4, A and D) and phalloidin (Fig. 4, B and E) signals. When the two signals are merged, partial co-localization can be observed, indicating that both F-actin and ␣-rENaC segregate to the same region of the apical domain.
Western Blotting of ␣-ENaC-In order to examine whether ␣-rENaC and actin are found in the same protein complex, we attempted to co-immunoprecipitate FLAG-tagged ␣-ENaC and actin from the MDCK cell line. To first demonstrate that we were able to immunoprecipitate specifically FLAG-tagged ␣-ENaC, we used the monoclonal anti-FLAG antibody; the ␣-subunit was immunoprecipitated using the anti-FLAG antibody, and the resulting immunoblot was then probed with the same primary antibody (Fig. 5A). As shown in lane 3, FLAG tagged ␣-ENaC was only detected by the anti-FLAG antibody in lysates immunoprecipitated with the anti-FLAG antibody. It was not detected in the beads only control (lane 1) or the nonimmune mouse IgG control (lane 2) or when blots were probed with the secondary antibody alone (lanes 4 -6). To further demonstrate that the 100-kDa polypeptide immunoprecipitated by the anti-FLAG antibody was indeed ␣-rENaC, samples were immunoprecipitated with the anti-FLAG antibody and the blot was probed using a commercially available anti-␣-rENaC antibody (Fig. 5B). In addition to the stably transfected MDCK ENaC cell lysate (designated E), lysates prepared from the original MDCK parental strain (designated P) were subjected to immunoprecipitation with the anti-FLAG antibody. The parental strain has previously been reported not to express ENaC as assayed by reverse transcription-PCR (12). As shown in lane 1 of Fig. 5B, a band at ϳ100 kDa was observed, whereas no signal was detected in lane 2 from the parental cell line. In order to demonstrate the specificity of this antibody, a peptide competition experiment was performed using replicates in lanes 3 and 4. The samples were immunoprecipitated with the anti-FLAG antibody, and the blot was probed with the primary antibody, which had been preabsorbed with its immunizing peptide. Upon development of the blot, the ϳ100-kDa band previously observed in lane 1 from the ENaC-expressing MDCK cell line was not detected.
Association of ␣-ENaC with F-actin in Vivo-In order to determine if F-actin is found in the same protein complex as the ␣-subunit within the MDCK cell line stably expressing FLAG-tagged ENaC, we attempted to co-immunoprecipitate actin with FLAG-tagged ␣-ENaC. Cell lysates

␣-ENaC Binds to F-actin
prepared from the MDCK cell line stably expressing FLAG-tagged ␣-ENaC were immunoprecipitated with anti-FLAG antibody, an irrelevant mouse monoclonal antibody, or beads alone. Immunoprecipitates were separated by SDS-PAGE and transferred to PVDF membrane, and blots were probed with a monoclonal anti-actin antibody. As can be seen in Fig. 6, no actin signal was detected in the beads alone control (lane 1). The anti-actin antibody detected a strong signal at about 45 kDa (lane 3), corresponding to the molecular mass of actin, in the cell lysate immunoprecipitated with the anti-FLAG antibody. A faint signal was observed when an irrelevant mouse monoclonal antibody (mAb anti-His) was substituted for the anti-FLAG antibody in the immunoprecipitation. We anticipated that we would observe a small amount of "background" in lane 2 in the anti-His control lane (lane 2), since actin interacts with IgG (12). This report demonstrated that the so-called "nonspecific background" often seen in this type of experiment is actually an interaction between actin and IgG. Furthermore, the authors of the study showed that the amount of IgG that would coprecipitate with actin would decrease as the pH of the buffer was increased (12). There-fore, in order to minimize this interaction in our experiments, the beads were washed two times with a high salt buffer with a pH of 9.1. An additional control experiment done exactly as the co-immunoprecipitation experiments described above, but using MDCK parental cell lysates, revealed no detectable actin following treatment with anti-FLAG antibody. 3 This observation suggests the actin signal seen in the anti-FLAG lane is due to the co-immunoprecipitation of actin with ␣-rENaC rather than an interaction between actin and IgG. These data demonstrate that F-actin and ␣-rENaC are found within the same protein complex.
The Carboxyl Terminus of ␣-hENaC Binds Directly to Nonmuscle F-actin in Gel Overlays-Electrophysiological studies have suggested that F-actin interacts with the COOH terminus of the ␣-subunit of ENaC (5,6). However, biochemical data demonstrating a direct interaction between ENaC and actin are lacking. To determine if F-actin binds to the COOH terminus of ␣-ENaC, we performed gel overlay assays, a widely used biochemical technique to examine whether actin binds directly to another protein.
A representative sample of the eluted GST-␣-hENaC C-terminal fusion protein (1.0 g of full-length fusion) stained with Gel Code is shown in Fig. 7A. In Fig. 7A, the arrow points to the full-length fusion protein, and the asterisk marks degradation products from the fusion protein. As shown in Fig. 7B, full-length GST-␣-hENaC C-terminal (lane 3) (denoted by the arrow just below the 37-kDa molecular mass marker) but not GST alone (lane 2) bound to nonmuscle F-actin/Al-exa488 in vitro. Nonmuscle F-actin/Alexa488 did not specifically bind to either GST-␣-rENaC NH 2 -terminal (lane 1) or GST-␥-rENaC C-terminal (lane 5) fusion proteins, which were used as additional negative controls. Whole cell MDCK ENaC lysates, which contain numerous actin-binding proteins, were used as positive controls for actin binding (lane 4). In Fig. 7, B (lanes 1, 2, 3, and 5) and C, binding of nonmuscle actin/Alexa488 to a protein below the 75-kDa molecular mass marker was observed. The identity of this protein was not determined, but it is believed to be a bacterial protein that is commonly co-purified with the 3 C. Mazzochi, J. K. Bubien, P. R. Smith, and D. J. Benos, unpublished data. . This lane has a band at ϳ100 kDa, which corresponds to ␣-rENaC. The Western blot (IB) with lanes 1-3 was probed with mouse mAb anti-FLAG (1:500). Lanes 4, 5, and 6 were immunoprecipitated as described for lanes 1, 2, and 3, respectively; however, the blot was probed only with the polyclonal goat anti-mouse horseradish peroxidase (GAM-HRP) conjugate. Notice in lane 6, the 100-kDa signal is not detected when the blot is probed with the secondary antibody alone. Samples were immunoprecipitated overnight at 4°C on a rotator with Protein G-agarose beads. Samples were separated on an 8% SDS-polyacrylamide gel, transferred to PVDF, and blocked with 1ϫ TBST plus 5% nonfat dry milk. Immunoblots were probed with mouse mAb anti-FLAG primary antibody, or the primary antibody was omitted. The secondary antibody used was a polyclonal goat anti-mouse horseradish peroxidase conjugate. The signal was visualized using enhanced chemiluminescence. B, detection of ␣-rENaC using an alternative antibody. Lane 1, ␣-rENaC was immunoprecipitated from MDCK ENaC lysates (E) exactly as described in A. Lane 2, MDCK parental lysates (P) were used as a control for antibody specificity. Lanes 1 and 2 were probed with a rabbit polyclonal ␣-ENaC (2 g/ml). In lane 1, a band at ϳ100 kDa, corresponding to ␣-ENaC, was detected in the MDCK ENaC lysate (E) but not in the MDCK parental lane (P). Lanes 3 and 4, replicates of lanes 1 and 2, respectively. The blot was probed with anti-␣-rENaC antibody preincubated with the immunizing peptide. The signal of the 100-kDa band detected in lane 1 was competed by the immunizing peptide (lane 3). In both blots, primary antibody was detected using a polyclonal goat anti-rabbit horseradish peroxidase conjugate. The signal was visualized using enhanced chemiluminescence. These experiments are representative of work repeated three times. MsIgG, mouse immunoglobulin. FIGURE 6. Co-immunoprecipitation of ␣-rENaC with actin in vivo. ␣-rENaC was immunoprecipitated as described in the legend to Fig. 2A. Lane 1, the protein G-agarose beads control; lane 2, the mAb anti-His antibody control; lane 3, immunoprecipitated with mAb anti-FLAG antibody for the ␣-ENaC subunit. The arrow denotes the detection of actin at ϳ45 kDa. All samples were washed with a pH of 9.1. The blot was probed with mAb anti-actin (1:1,500), and the primary antibody was detected using a polyclonal goat antimouse horseradish peroxidase conjugate. The signal was visualized using enhanced chemiluminescence. This experiment is representative of work repeated three times.
GST fusion proteins (13). Fig. 7C shows a representative microtubule and nonmuscle actin/Alexa488 competition assay, which was used to determine if the binding of actin is specific or nonspecific for the COOH terminus. Actin, like microtubules, is a negatively charged biopolymer. If the binding is due to nonspecific interactions, and then the microtubules will bind to the fusion protein and no signal would be observed, whereas if the binding interaction is specific, then a signal from the nonmuscle actin/Alexa488 would still be observed. As shown in Fig. 7C, nonmuscle actin/Alexa488 did bind to the GST-␣-hENaC C-terminal fusion protein in the presence of microtubules. A second competition control (Fig. 7D) was conducted to establish that binding was not a result of the Alexa488 and its linker group. Gel overlay membrane strips were incubated with nonmuscle F-actin/Alexa488 (25 g/ml) in the presence of two different amounts of unlabeled nonmuscle F-actin (lane 1, 100 g/ml; lane 2, 1 mg/ml) The binding of the F-actin labeled with Alexa488 to the GST-␣-hENaC C-terminal fusion protein was observed to decrease with an increased amount of unlabeled actin, as shown in Fig. 7. As shown in Fig. 7E, full-length GST-␣-rENaC C-terminal (1.0 g of full-length fusion denoted by the arrow; an asterisk denotes a degradation product) was observed to bind nonmuscle F-actin in gel overlays as well.
The Carboxyl Terminus of ␣-hENaC Co-sediments with Nonmuscle F-actin in Actin Co-sedimentation Assays-Actin co-sedimentation assays were used to corroborate the results of the actin overlay assays. In order to investigate whether the COOH terminus of ␣-ENaC directly binds to F-actin, we used a human ␣-ENaC fusion protein. The F-actin binding experiments were carried out with the human ␣-ENaC fusion protein, because the full-length rat ␣-ENaC fusion protein undergoes a severe amount of degradation. This results in a very small yield of fulllength GST-␣-rENaC C-terminal fusion protein. The 14-amino acid putative actin binding domain has a 78% sequence identity between the rat and human isoforms of ␣-rENaC. Only three amino acids in the putative actin binding domain are different between the human and rat ENaC isoforms. All three amino acids are conserved substitutions; the conserved substitutions do not change the charge in this region. Actin co-sedimentation assays were performed to demonstrate that the binding between the GST-␣-hENaC C-terminal fusion protein and nonmuscle F-actin was not due to the linear conformation that the fusion protein adopted when run through the SDS-polyacrylamide gels used in the gel overlay experiments. The experiment illustrated in Fig. 8A is a control sedimentation of the GST-␣-hENaC C-terminal fusion protein in the absence of actin to determine in which fraction the fusion protein comes down following sedimentation. The majority of the fusion protein was detected in the soluble fraction (S), and to a lesser extent some fusion protein was detected in the pellet (P). The arrow denotes the full-length GST-␣-hENaC C-terminal fusion with a mass of ϳ35 kDa. An asterisk denotes degradation products of the full-length length GST-␣-hENaC C-terminal fusion protein. The protein detected shortly below the 75-kDa molecular mass marker, which is detected in all of the soluble fractions, was unidentified. However, when a GST fusion protein is eluted from a column during purification, often an assortment of bacterial proteins can co-purify with the GST fusion protein (13). DnaK is a bacterial protein with a mass of ϳ70 kDa and is often observed when Binding of nonmuscle F-actin/Alexa488 to GST-␣-hENaC C-terminal fusion protein was observed. Samples were separated on a 12% SDS-PAGE gel, transferred, and blocked overnight in 1ϫ TBST, 5% nonfat dry milk. B was probed with nonmuscle F-actin/Al-exa488 (25 g/ml) in 1ϫ TBST, 5% nonfat dry milk containing 5 M phalloidin for 2 h in the dark. C was probed with an equal molar concentration of microtubules and nonmuscle F-actin/Alexa488 to determine if the binding of F-actin to the GST-␣-hENaC C-terminal fusion proteins was specific. The arrow denotes F-actin/Alexa488 conjugate binding to the full-length GST-␣-hENaC C-terminal fusion protein in both B and C. D, labeled and unlabeled F-actin control. An increased amount of unlabeled nonmuscle F-actin decreased binding of labeled nonmuscle F-actin/Alexa488. GST-␣-hENaC C-terminal fusion protein was separated, transferred, and blocked as above in B and C. Lane 1 was probed simultaneously with nonmuscle F-actin/Alexa488 (25 g/ml) and unlabeled nonmuscle F-actin (100 g/ml). Lane 2 was probed with an increased concentration of unlabeled nonmuscle F-actin (1 mg/ml), whereas the concentration of labeled nonmuscle F-actin/Alexa488 was the same as in lane 1 at 25 g/ml. The arrow points to full-length GST-␣-hENaC C-terminal fusion protein. The overlay buffer used in D was the same as used in B and C. E, nonmuscle F-actin binds to 1.0 g of GST-␣-rENaC C-terminal fusion protein. The arrow points to full-length GST-␣-rENaC C-terminal fusion protein, and the asterisk indicates a degradation product of this fusion protein. This overlay was done in the same manner as in B. A is representative of work repeated four times, and B is representative of work repeated five times. C is representative of work repeated three times, D is representative of work repeated two times, and E is representative of work repeated three times. GST fusion proteins are eluted from columns (13). As shown in Fig. 8B, the GST-␣-hENaC C-terminal fusion protein but not GST alone was found to co-sediment with nonmuscle F-actin. A small amount of GST-␣-hENaC C-terminal fusion protein was observed in the P lane of the GST sample. This was a result of a very small amount of spillover from sample loading of an adjacent S lane of GST-␣-hENaC C-terminal fusion protein. F-actin was detected as the large bands observed in the P lanes with a mass of ϳ45 kDa. The positive control, ␣-actinin, a known actin-binding protein with a mass of ϳ100 kDa, was observed to cosediment with nonmuscle F-actin. These data demonstrate that F-actin binds to the GST-␣-hENaC C-terminal fusion protein in vitro.

DISCUSSION
The mammalian cytoskeleton is composed of three major protein families: microfilaments, microtubules, and intermediate filaments, as well as numerous cytoskeleton-associated proteins (14). Together they contribute toward cell shape, motility, signal transduction, vesicular trafficking, and ion channel/transporter/receptor function. The actin cytoskeleton has been shown to interact both directly and indirectly with ion channels and membrane transport proteins. In addition to functioning as an anchor for ion channels/transporters within the plasma membrane (15), the actin cytoskeleton participates in the vesicular trafficking of ion channels and transporter proteins to the membrane (16,17). To date, the majority of the literature has described indirect protein interactions between the actin cytoskeleton and ion channels and transporters. These protein interactions are often mediated through a protein linking actin to the integral membrane protein.
For example, cystic fibrosis transmembrane conductance regulator is linked to the cortical actin cytoskeleton through a protein scaffolding complex composed of EBP-50 (ezrin-radixin-moesin binding phosphoprotein 50) and ezrin (18). However, there is increasing evidence that the actin-based cytoskeleton also directly interacts with a number of ion channels, transporters, and receptors (19 -22).
Ion channels and transporters perform a variety of functions, from action potential propagation in the nervous system to allowing the passage of ions across a cell's membranes to maintain extracellular fluid homeostasis. However, whether and how the actin cytoskeleton influences the function of an ion channel or transporter by either indirect or direct binding to the integral membrane protein is poorly understood.
Although there are several reports regarding the biophysical affects of actin on ENaC (3)(4)(5)(6)(23)(24)(25), there are no reports in the literature that have attempted to elucidate biochemically the nature of the ENaC/actin relationship. Previous biophysical studies offer insight into the nature of the ENaC/actin interaction. The COOH terminus of ␣-ENaC has been shown to contribute to the modulation of the channel by actin (5). Through the use of chimeric constructs composed of ␣-rENaC and ␣-bENaC along with site-directed mutagenesis, a conserved 14-amino acid region in the COOH terminus of ␣-bENaC has been shown to contribute specifically to actin's modulation of ENaC (6). In the present study, we have investigated the question of whether or not actin binds to the COOH terminus of ␣-ENaC.
ENaC is expressed in the apical domain of polarized epithelia. However, because ENaC is found in such small amounts in native epithelia (26), we used a MDCK cell line stably transfected with wild type, rat isoforms of ␣-ENaC with a FLAG tag on its extracellular domain and ␤and ␥-ENaC for our studies. Fig. 1 demonstrates the ENaC expressed in this MDCK cell line forms functional amiloride-sensitive sodium channels at the apical surface of the plasma membrane domain and that they resembled native ␣␤␥-ENaC channels in their biophysical properties. The conductance was 8 picosiemens at a transmembrane potential of 60 mV, NP o was increased by cAMP, and the channels exhibited the appropriate inhibition at the appropriate concentration by amiloride.
This model system allowed us to examine if functional ENaC at, or near, the apical domain of the MDCK monolayers co-segregated with the cortical F-actin cytoskeleton. Initial single label experiments were performed to examine the normal pattern of expression and localization of the FLAG-tagged ␣-ENaC subunit. The pattern of ␣-ENaC expression within this cell line is heterogeneous, with some cells having higher expression levels than others. However, the localization of ␣-ENaC remained consistent through all samples, as seen in Fig. 2E, where the immunofluorescence is apical and subapical. We also performed double label experiments to determine if the apparent apical labeling of ␣-ENaC was in fact in the apical plasma membrane, or if the labeling was completely below the cell surface. We observed ␣-ENaC in the apical membrane as well as below the apical domain, as shown in Fig. 2J. The localization of the ␣-subunit, found both in the apical plasma membrane and subapically, is consistent with previous studies utilizing other monolayer-forming epithelia stably transfected with ENaC (10,27). In addition, the ␣-ENaC localization we observed is also consistent with observations reported in native epithelia from the cortical collecting duct of the kidney (28 -31). Immunoprecipitations for ␣-ENaC (Fig. 5,  A and B) were carried out to confirm the single label experiments. We did a second series of double label experiments (Fig. 3, A-G) to examine the position of the F-actin immunofluorescence signal relative the apical plasma membrane itself. As expected, the cortical F-actin cytoskeleton was observed just beneath the apical plasma membrane in our model system. Fig. 4, A-F, shows two different double label experiments that were performed to see if ␣-ENaC co-localized with the cortical F-actin cytoskeleton. In Fig. 4, C and D both demonstrate that partial co-localization can be observed, indicating that both F-actin and ␣-ENaC segregate to the same region of the apical domain. This is a significant observation, because this places these two proteins in the same place, suggesting the possibility of an interaction between ␣-ENaC and the cortical actin cytoskeleton. Whereas the immunocytochemistry was encouraging, we needed to see if ␣-ENaC and F-actin exist as a complex, and in fact they do. We used whole cell lysates in these experiments for immunoprecipitating ␣-ENaC. Fig. 6 shows an actin-like band at ϳ45 kDa. The ␣-ENaC that was immunoprecipitated was in all likelihood a mixture of both ␣-ENaC inserted into the apical membrane and ␣-ENaC residing in intracellular pools; this datum supports the subapical co-localization of F-actin and ␣-ENaC observed in Fig. 4F. These data demonstrate that ␣-ENaC is located in the apical domain with the cortical actin cytoskeleton, and the co-immunoprecipitation of ␣-ENaC and actin from MDCK cell lysates corroborates an in vivo interaction between these two proteins.
The exact functional role(s) of interaction between the cortical actin cytoskeleton and ENaC is unresolved. However, there are some possibilities from other studies that may apply here. Actin has been shown to function structurally as an anchor for ion channels/transporters (15,32) and to participate in the vesicular trafficking of ion channels/transporters (16,17) in response to hormonal stimulation. This may be a possibility for ENaC as well, since studies have shown ENaC trafficking in response to hormonal stimulation (31,(33)(34)(35). Hormone-mediated ENaC trafficking may involve the actin cytoskeleton. Another line of thought is that the cortical actin cytoskeleton functions as a physical barrier to vesicular exocytosis (36 -39). In doing so, it remains in a polymerized state and prevents vesicles containing ion channels/transporters from coming in contact with the apical membrane and thus preventing vesicular fusion and channel insertion into the surface of a cell. Alternatively, vesicles containing ENaC could be trafficked or held in place by the cortical actin cytoskeleton. In the toad urinary bladder, the cortical actin cytoskeleton forms a meshwork of short filaments just under the apical surface when observed with an electron microscope (40); this is probably true for cortical collecting duct cells in the mammalian kidney. This would in turn lend support to previously reported observations that demonstrated that short actin filaments had an effect on ENaC by increasing the channel's open probability and decreasing its conductance (4,6). It may be that short actin filaments located subapically directly bind to ENaC in response to hormonal influences. In the toad urinary bladder, antidiuretic hormone causes an increase in the water permeability of epithelia in both mammals and anurans (41). The antidiuretic hormone also causes a reorganization of F-actin, which is believed to allow vesicular fusion to occur. A similar mechanism may occur in ENaC-containing epithelia. In addition, GLUT4 (glucose transporter-4) translocation has been reported to occur via an insulin-stimulated remodeling of the cortical F-actin cytoskeleton (17), and insulinstimulated GLUT4 translocation requires phosphatidylinositol 3-kinase function as well (17). Blazer-Yost et al. (33) have observed in renal cells that insulin-stimulated trafficking of ENaC required phosphatidylinositol 3-kinase. The authors of that study did not report on any aspects of the cytoskeleton; a possible mechanism of ENaC translocation may in part involve the remodeling of the cortical F-actin cytoskeleton in response to hormonal stimuli.
In order to answer the question of whether the COOH terminus of ␣-ENaC binds directly to F-actin, we carried out gel overlays and F-actin co-sedimentation assays using a GST-␣-hENaC C-terminal fusion. The 14-amino acid putative actin binding domain contained within the COOH terminus of ␣ has a 78% sequence identity between the rat and human isoforms, with only three conserved amino acid substitutions that are different in this region. Fig. 7B, lane 3, shows a strong signal at ϳ35 kDa, where F-actin/Alexa488 is bound to the full-length GST-␣-hENaC C-terminal fusion protein. A weaker signal was observed below the full-length fusion protein, where the F-actin/Alexa488 was bound to degradation products of the fusion protein. In Fig. 7E, we also observed binding of F-actin to a full-length GST-␣-rENaC C-terminal fusion protein. The full-length fusion proteins are denoted by the arrowheads. In Fig. 8, in the negative control (no actin) the GST-␣-hENaC C-terminal fusion is almost entirely in the soluble fraction, with a small amount in the pellet. When this fusion is incubated and spun down with F-actin, approximately half of the fusion appears in the pellet with F-actin, and the negative GST control is found almost entirely in the soluble fraction. One explanation for this observation could be that not all of the GST-␣-hENaC fusion protein made is in the proper conformation, thus preventing it from binding to the F-actin. These binding studies demonstrate that F-actin binds directly and specifically to the COOH terminus of ␣-ENaC.
Our results support the biophysical data that suggest a functional interaction between the ␣-ENaC COOH terminus and F-actin. We demonstrate for the first time a direct and specific interaction between F-actin and ENaC.