Clathrin-mediated Endocytosis of the Epithelial Sodium Channel

Here we present evidence that the epithelial sodium channel (ENaC), a heteromeric membrane protein whose surface expression is regulated by ubiquitination, is present in clathrin-coated vesicles in epithelial cells that natively express ENaC. The channel subunits are ubiquitinated and co-immunoprecipitate with both epsin and clathrin adaptor proteins, and epsin, as expected, co-immunoprecipitates with clathrin adaptor proteins. The functional significance of these interactions was evaluated in a Xenopus oocyte expression system where co-expression of epsin and ENaC resulted in a down-regulation of ENaC activity; conversely, co-expression of epsin sub-domains acted as dominant-negative effectors and stimulated ENaC activity. These results identify epsin as an accessory protein linking ENaC to the clathrin-based endocytic machinery thereby regulating the activity of this ion channel at the cell surface.

Here we present evidence that the epithelial sodium channel (ENaC), a heteromeric membrane protein whose surface expression is regulated by ubiquitination, is present in clathrin-coated vesicles in epithelial cells that natively express ENaC. The channel subunits are ubiquitinated and co-immunoprecipitate with both epsin and clathrin adaptor proteins, and epsin, as expected, co-immunoprecipitates with clathrin adaptor proteins. The functional significance of these interactions was evaluated in a Xenopus oocyte expression system where co-expression of epsin and ENaC resulted in a downregulation of ENaC activity; conversely, co-expression of epsin subdomains acted as dominant-negative effectors and stimulated ENaC activity. These results identify epsin as an accessory protein linking ENaC to the clathrin-based endocytic machinery thereby regulating the activity of this ion channel at the cell surface.
The epithelial Na ϩ channel (ENaC) 2 is expressed in absorptive tissues that establish low luminal Na ϩ concentrations. Abnormalities of this channel contribute to the pathophysiology of a number of diseases including Liddle's syndrome, psuedo-hypoaldosteronism and cystic fibrosis (1)(2)(3). ENaC is a heterotrimeric channel, composed of three homologous subunits, ␣, ␤, and ␥ (4). The channel complex is inefficiently processed with less than 20% of newly synthesized subunits reaching the apical membrane of endogenously expressing epithelia (3,5). The apical membrane stability of these channels is a subject of dispute but it is widely agreed that ENaC is retrieved from the apical membrane by a process involving ubiquitination (6). Early studies in oocytes demonstrated that mutation of a C-terminal PY motif, or co-expression of a dominant negative dynamin, resulted in increased ENaC activity, consistent with regulation of surface expression by clathrin-mediated endocytosis (7). Subsequently it was recognized that the ubiquitin-protein ligase Nedd4-2 binds to the consensus PPXY (PY) motifs in the cytoplasmic tails of ENaC subunits via its WW domains and negatively regulates surface expression of the channel (1)(2)(3). However, the role of clathrin-mediated endocytosis in this process is not well understood.
Clathrin-mediated endocytosis is a process by which specific membrane proteins are internalized from the plasma membrane by eukaryotic cells. Target proteins bind through classic motifs to adaptor proteins that link the cargo to clathrin lattices that deform the cell membrane to form a bud and ultimately an endocytic vesicle (8). Membrane proteins that are modified by attachment of one or more ubiquitin molecules appear to be internalized by signals unlike the classic linear internalization motifs and may require binding to accessory proteins that link them to clathrin adaptors (9). Proteins of the epsin family appear to be excellent candidates to function as adaptor coupling proteins because they contain ubiquitin interaction motifs (UIM) as well as clathrin binding motifs (10,11), and epsin has been shown to downregulate ENaC activity when co-expressed in Chinese hamster ovary cells (12). Epsin orthologs in Drosophila and Caenorhabditis elegans target monoubiquitinated ligands of the Notch pathway for endocytosis (13)(14)(15). However, direct experimental evidence for a role of epsins linking ubiquitinated cargo to clathrin-coated vesicles in mammalian cells is lacking (9). The current studies were undertaken to examine the possibility that ENaC internalization involves clathrin-mediated endocytosis and to determine whether epsin is involved in linking ubiquitinated ENaC to clathrin adaptors.

MATERIALS AND METHODS
Reagents-Rabbit polyclonal antibodies to ␣ and ␤ ENaC were kindly provided by Dr. Kathi Peters (University of Pittsburgh; ␣ ENaC) and Dr. Mark Knepper (Laboratory of Kidney and Electrolyte Metabolism, NHLBI, National Institutes of Health, Bethesda MD; ␤ ENaC) and their specificity in CCD cells has been described (16). Chicken polyclonal antibody to ␥ ENaC has been described (17). ␥ ENaC bands visualized by this antibody in CCD cells were competed by incubation with immunizing peptide (not shown). Mouse monoclonal antibody to ubiquitinylated proteins (clone FK2) was purchased from Biomol. Mouse monoclonal antibody to ␣ adaptin was purchased from BD Biosciences, and EEA1 rabbit polyclonal antibody was a gift from Dr. Silvia Corvera. Antibodies to heavy chain clathrin, ␤ and ␥ adaptin and epsin have been previously described (18). Rabbit polyclonal antibody to the N terminus of 2 was the generous gift of Dr. Juan Bonifacino (NIH, Bethesda, MD). ␤R564X ENaC mutant was generated as previously described (19). The full-length epsin 1 cDNA in pBluescript was a generous gift from Pietro de Camilli (Yale University School of Medicine, New Haven, CT). The epsin DPW domain (residues Arg 229 -Asp 407 with a subsequent TAA stop codon) was amplified from this by PCR and then inserted into pcDNA3.1myc using the EcoRI and XbaI sites. The ENTH domain construct (Met 1 -Ala 163 ) was prepared by introduction of an Ala 164 stop mutation in full-length epsin 1 inserted into pcDNA3.1myc as for the DPW domain. The proteins encoded by the pcDNA3.1myc vector have an N-terminal myc epitope tag. GST-epsin 1-(1-407) has been previously described (20). For construction of ENaC-ubiquitin conjugates, a plasmid containing three tandem ubiquitin (gift from G. Lukacs, Toronto) was amplified using unique primers containing a BspHI site on the 5Ј primer. The product was cloned into pTopo-Blunt II (Invitrogen). The ubiquitin containing fragment was excised using BspHI and XhoI (in linker region) and cloned in-frame to the carboxyl end of ␣-ENaC that had been cleaved using NcoI and XhoI. Cloning into the NcoI site on ␣-ENaC results in the deletion of the final two amino acids of the native protein. For both ␤ and ␥ ENaC a KpnI site was introduced before the termination codon and the ubiquitin construct was ligated into the KpnI site and a unique downstream restriction enzyme site. Lysine to arginine mutants of ␣ and ␥ ENaC (␣ K47R,K50R and ␥ K6 -13R ) (K6 -13R is representative for K6R,K8R,K10R,K12R,K13R) were generated by substituting specific lysine residues with arginines using PCRbased mutagenesis as described (6). All constructs were verified by automated dideoxynucleotide sequencing.
The mpkCCD c14 cells (kindly provided by Alain Vandewalle and Marcelle Bens) were grown in flasks, tissue culture dishes, and Transwell porous supports. Growth medium was composed of equal volumes Dulbecco's modified Eagle's medium and Ham's F-12, 60 nM sodium selenate, 5 g/ml transferrin, 50 nM dexamethasone, 1 nM triiodothyronine, 10 ng/ml epidermal growth factor, 5 g/ml insulin, 20 mM D-glucose, 2% (v/v) fetal calf serum, and 20 mM HEPES, pH 7.4, at 37°C in 5% CO 2 , 95% air atmosphere. For all experiments the CCD cells were subcultured onto permeable filter supports (0.4 M pore size, 4.5 and 44.2 cm 2 , Transwell, Corning Costar, Cambridge, MA) and were used for studies when transepithelial resistance measured with a portable EVOM (World Precision Instruments, Sarasota, Fl) had reached a stable plateau indicating complete epithelial organization.
Clathrin-coated Vesicle (CCV) Isolation-CCVs were isolated by a modification of the method of Metzler et al. (21), CCD cells cultured on 44.2-cm 2 filter supports were washed 3 times with ice-cold phosphatebuffered saline on ice. Cells were scraped and homogenized in isolation buffer containing 10 mM MES-NaOH, pH 6.5, 100 mM KCl, 1 mM EGTA, 0.5 mM MgCl 2 , 0.02% NaN 3 , with protease inhibitors, using a Dounce homogenizer, then sonicated by 3 bursts of 10 s on ice. The homogenate was centrifuged at 17,800 ϫ g for 20 min using Sorvall T865 rotor and Sorvall Discovery 90 centrifuge at 4°C and the supernatant saved. The supernatant was sedimented at 56,000 ϫ g for 1 h. The pellet was resuspended in 1.7 ml of isolation buffer and sheared through a 22-gauge needle 20 times. The redissolved pellet was added on top of an equal volume of 8% sucrose in isolation buffer made in D 2 O and centrifuged at 115,800 ϫ g for 2 h (S120AT2-0112 rotor; Sorvall RC M120EX centrifuge). The final pellet represented the CCV preparation. CCVs were resuspended in 0.2-0.5 ml of RIPA buffer, frozen in liquid nitrogen, were kept at Ϫ80°C for further study.
Endosomal Preparation-CCD cells grown on filter supports were scraped into phosphate-buffered saline, pelleted, and resuspended in 600 l of HB (250 mM sucrose, 10 mM HEPES, 0.5 mM EDTA, pH 7.4) containing protease inhibitors by passing through a 200 -1000-l pipette tip 10 times, then lysed by passing through a 22-gauge needle 20 times. Following a low speed spin, the post-nuclear supernatant was diluted 1:1 with 62% sucrose and placed at the bottom of a 4.4-ml polyclear centrifuge tube (Seton Scientific, Sunnyvale, CA). 1.5 ml of 35% sucrose was layered on top followed by 1.5 ml of 25% sucrose and 0.5 ml of HB. The gradients were centrifuged in a TST 60.1 rotor at 167,000 ϫ g for 70 min at 4°C, and the interfaces were collected. Protein-matched samples were run on SDS-PAGE gels, transferred to nitrocellulose, and probed with EEA 1 and ␣ ENaC. Quantitation of internalized horseradish peroxidase in gradient fractions was performed as described (17).
Immunoblot and Immunoprecipitation-CCD cells grown on Transwell filters were lysed in CCD lysis buffer (50 mM Tris, pH 7.4, 5 mM EDTA, 1% Nonidet P-40, 0.4% sodium deoxycholate) with protein inhibitors, incubated on ice for 10 min. For experiments involving IP or immunoblot with ubiquitin antibody, 5 mM N-ethylmaleimide and 10 M MG-132 were added to lysis buffer to inhibit deubiquitination. Without these agents in lysis buffer, we were unable to detect any interaction between ENaC and ubiquitin. After centrifugation at 4°C for 10 min at 13,000 ϫ g, the supernatant was recovered, and immunoprecipitation was performed with appropriate antibodies. The immunoprecipitated material, cell lysate, and CCV samples were analyzed by SDS-PAGE (8 -10% polyacrylamide). Separated proteins were transferred to nitrocellulose membranes and then subjected to Western blot analysis with the appropriate antibodies and visualized using chemiluminescence (PerkinElmer Life Sciences).
Protein Binding Studies-Pulldown-type assays were performed in 500 l of total volume containing 100 g of GST or the appropriate GST fusion protein first immobilized on 25 l of GSH-Sepharose by incubation at 4°C. The Sepharose beads containing the immobilized proteins were washed and resuspended in assay buffer containing 0.5% Triton X-100. [ 35 S]Methionine-labeled and in vitro translated ENaC or ENaC-Ub 3 were added to the tubes and incubated at 4°C for 2 h with continuous gentle mixing. The GSH-Sepharose beads were then recovered by centrifugation and an aliquot of each supernatant was removed. The pellets were then washed and resuspended in sample buffer.
Channel Expression in Xenopus Oocytes-Sequences for all wild type and mutant constructs were confirmed by automated DNA sequence analysis performed at the University of Pittsburgh DNA sequencing facility. Complimentary RNAs (cRNA) for wild type and mutant constructs were prepared using a cRNA synthesis kit employing T 3 RNA polymerase (mMESSAGE mMachine, Ambion Inc., Austin, TX). Xenopus oocytes (stage V-VI) were pretreated with 2 mg/ml collagenase (type IV) in calcium-free saline solution. Murine ENaC cRNAs (1-3 ng/subunit in 50 nl of H 2 O) were microinjected into all oocytes. Oocytes in the experimental group were additionally injected with 5 ng of cRNA of epsin, the ENTH domain of epsin, or the DPW domain. All oocytes were incubated at 18°C in modified Barth's saline (88 mM NaCl, 1 mM KCl, 2.4 mM NaHCO 3 , 0.3 mM Ca(NO 3 ) 2 , 0.41 mM CaCl 2 , 0.82 mM MgSO 4 , 15 mM HEPES-NaOH, pH 7.2) supplemented with 10 g/ml sodium penicillin, 10 g/ml streptomycin sulfate, and 100 g/ml gentamycin sulfate. Whole cell currents were measured 24 -46 h after cRNA injections. To determine whether the effects of epsin on ENaC were specific, we examined ROMK expressing oocytes (22) with and without epsin. Wild type ROMK cRNA (2 ng) was injected with or without epsin cRNA (5 ng).
Whole Cell Current Measurements-A two-electrode voltage clamp technique was used as previously described (22). Whole cell inward amiloride-sensitive currents were measured in control oocytes expressing ␣␤␥ ENaC alone or experimental oocytes expressing ␣␤␥ ENaC ϩ epsin, ENTH domain, or DPW domain, using a DigiData 1200 interface (Axon Instruments, Foster City, CA) and a TEV 200A Voltage Clamp amplifier (Dagan Corp., Minneapolis, MN). Data acquisition and analysis were performed using pClamp 7.0. Amiloride-sensitive currents were defined as the difference of the current in the absence and presence of 0.1 mM amiloride. Membrane potentials were clamped from Ϫ140 to ϩ60 mV in 20-mV increments with duration of 900 ms. Currents were measured at a holding potential of Ϫ100 mV, 600 ms after initiation of the clamp potential. For ROMK measurements, whole cell K currents were measured at Ϫ100 mV in the absence or presence of 5 mM BaCl 2 .
Cell Surface ENaC Labeling-Surface expression of ENaC was measured using a luminometric technique as described (22). Control oocytes expressing mouse ␣, ␤-FLAG, and ␥ ENaC subunits and experimental oocytes co-injected with ␣␤ flag ␥ ENaC and epsin were blocked with MBS supplemented with 1 mg/ml of bovine serum albumin (MBS-BSA) after 2 days of incubation. Oocytes were then exposed to MBS-BSA with 1 mg/ml of mouse monoclonal anti-FLAG antibody (M2, Sigma) at 4°C for 1 h. Of note, ␤ ENaC containing the FLAG epitope (DYDKKKD) at the extracellular loop does not alter I Na relative to wild type ENaC expression. After the oocytes were labeled with primary antibody, they were washed six times in MBS-BSA at 4°C and incubated in MBS-BSA supplemented with 1 mg/ml horseradish peroxidase-conjugated secondary antibody (peroxidase-conjugated AffiniPure F(abЈ)2 fragment goat anti-mouse IgG, Jackson ImmunoResearch Laboratories (West Grove, PA)) for 1 h at 4°C. After 12 additional washes, individual oocytes were placed in 100 ml of SuperSignal ELISA Femto solution (Pierce) and incubated at room temperature for 1 min. Chemiluminescence was quantified in arbitrary light units using a TD-20/20 luminometer with signal integration over a 60-s interval.

RESULTS
To determine whether ENaC subunits were present in CCV, we employed a standard method (21) for isolating these vesicles from a continuous line of mouse cortical collecting duct cells that express the channel endogenously (23,24). CCVs prepared from mpkCCD c14 (CCD cells) were first characterized in terms of enrichment of coat proteins and adaptors. A typical example of such a preparation is shown in Fig.  1A. Compared with the post-nuclear supernatant and crude microsomes, CCV were enriched in clathrin heavy chain, and adaptors found at Golgi (␥ adaptin) and the plasma membrane (2). Typically, epsin was present but not markedly enriched in CCVs. Fig. 1B demonstrates that this CCV preparation was highly enriched in all three ENaC subunits. The pattern of subunit enrichment is interesting and deserves comment. We have recently observed that ENaC subunits expressed alone produce single bands on SDS gels by immunoblot: ␣ at 95 kDa, ␤ at 96 kDa, and ␥ at 93 kDa. However, when expressed together, a second band appears for each subunit (65 kDa for ␣, 110 kDa for ␤, and 75 kDa for ␥). These appear to represent the processed forms of the subunits as they exhibit complex N-glycans and the ␣ and ␥ subunits appear to undergo proteolytic processing (25). Both mature and immature forms of ENaC appear at the apical membrane (26). The experiments describing ENaC subunit processing were carried out in Madin-Darby canine kidney cells overexpressing double epitope-tagged subunits (25). In endogenously expressing epithelia, shorter forms of ␥ but not ␣ ENaC have been reported following stimulation of ENaC activity by aldosterone (27). Complex glycosylation of ENaC has been reported in A6 cells, although with full-length ␥ ENaC (28). In CCD cells, CCVs appear to be enriched in the mature forms of ␤ and ␥ ENaC, but only full-length ␣ is demonstrated.
To demonstrate endocytosis of the channel, we have taken advantage of the relatively rapid insertion and retrieval of ENaC in response to elevated cAMP. In CCD epithelia, forskolin stimulates rapid and parallel increases in ENaC currents and membrane capacitance (t1 ⁄ 2 ϳ 4.5 min). Its withdrawal results in equally rapid reversal of these capacitance and current responses, which are blocked by chloroquine, consistent with endocytic retrieval of ENaC (16). We have examined the appearance of ␣ ENaC in early endosomes following forskolin withdrawal. Early endosomes were identified by their internalization of horseradish peroxidase ( Fig. 2A) and they were isolated on a discontinuous sucrose gradient as previously described (17). Fractions enriched in EEA1, a marker for early endosomes, were enriched in the three ENaC subunits as well (Fig.  2B). Following forskolin withdrawal, there is a rapid increase in the ␣ subunit of ENaC in early endosomes over a 10-min time course, which is typical of ENaC current decay kinetics (16) (Fig. 2C). These data indicate that ENaC traffics through an early endosomal compartment following its endocytosis from the apical membrane within CCVs.
To explore the relationship of ENaC subunits to adaptor proteins, co-immunoprecipitation studies were performed. CCD cell lysates were FIGURE 1. ENaC is present in clathrin-coated vesicles. A, typical example of a CCV preparation probed for heavy chain clathrin, ␤ and ␥ adaptin, 2, and epsin. All lanes were loaded with equal amounts of protein: lane 1 represents the post-nuclear supernatant, lane 2 the resuspended pellet from the 56,000 ϫ g spin that is loaded on the D 2 O gradient, and lane 3 the final CCV preparation. The proteins were visualized at the following molecular mass by markers: heavy chain clathrin, ϳ200 kDa; ␤ adaptin and ␥ adaptin, ϳ100 kDa; 2, ϳ40 kDa; epsin, ϳ170 kDa (n ϭ 3). B, the same three CCD preparations (lanes 1-3 from A) probed for the three ENaC subunits showing enrichment of apparent mature forms of ␤ and ␥ ENaC as described in the text (n ϭ 3). FIGURE 2. ENaC is present in early endosomes. A, endosomal preparation isolated on discontinuous sucrose gradients as described under "Materials and Methods." CCD cells were preincubated with horseradish peroxidase (HRP) on the apical membrane and allowed to take up the enzyme at 37°C for 10 min. Cells were then gently lysed as described and the post-nuclear supernatant centrifuged on the sucrose gradient. Sequential fractions were taken from the top of the gradient and assayed for horseradish peroxidase. A sharp increase in horseradish peroxidase was noted in fraction 4, which corresponded to the interface between the 25 and 35% sucrose (n ϭ 2). B, equivalent amounts of protein from each fraction were loaded on SDS-PAGE and probed for EEA1 and ENaC subunits (n ϭ 3). C, appearance of ␣ ENaC in EEA1 endosomes following forskolin withdrawal. CCD cells were incubated with 10 M forskolin, which induced a rapid increase in I sc and apical membrane capacitance consistent with exocytic insertion of ENaC (16). Washout of forskolin results in a rapid and parallel decline in I sc and capacitance consistent with retrieval of ENaC via endocytosis with a t1 ⁄2 of 4.5 min. EEA1 fractions from discontinuous gradients were isolated from CCD cells at the time of forskolin removal (0) and at 2, 5, and 10 min following removal. Equivalent amounts of protein were loaded from each time point, separated by SDS-PAGE, and probed for EEA1 and ␣ ENaC. As shown in the upper panel, amounts of EEA1, a control for loading, stayed relatively constant or decreased slightly. Amounts of ␣ ENaC detected in early endosomes progressively increased over the time course of current decline following forskolin withdrawal (n ϭ 3).
subjected to immunoprecipitation with antibodies specific for ␣, ␤, and ␥ ENaC and analyzed for the presence of epsin by immunoblot. Fig. 3A shows that epsin was efficiently co-immunoprecipitated by all three ENaC subunits. Furthermore, the subunits of ENaC were found to be ubiquitinated under these assay conditions. Fig. 3B demonstrates that both ␣ and ␤ ENaC when immunoprecipitated from CCD lysates can be detected with antibodies specific for ubiquitin. In addition, immunoprecipitations performed with anti-ubiquitin are positive for both ␣ and ␤ ENaC when probed with ENaC specific antibodies. IP with ␥ ENaC was also positive for ubiquitin, but we could not demonstrate that IP with ubiquitin could show the presence of ␥ ENaC (not shown). As a control, we demonstrated that the ubiquitin antibody could also co-IP epsin, which is known to be ubiquitinated (29). With regard to ENaC, the results are somewhat surprising because previous studies in overexpression systems have shown that ␣ and ␥ ENaC, but not ␤ ENaC are polyubiquitinated (1). In A6 cells that endogenously express the channel, ubiquitination of all three subunits has recently been shown, including 1-3 ubiquitins on ␣ ENaC and 2-3 ubiquitins on ␤ ENaC. Higher levels of ubiquitination were seen following incubation of cells with MG-132 (30). Our results demonstrate that at least two of the subunits of this heteromeric channel are ubiquitinated indicating that ENaC is a multiubiquitinated hetero-oligomer. Fig. 3C demonstrates that epsin antibody also co-immunoprecipitated ␣ adaptin. An antibody to the N terminus of the 2 subunit of adaptin immunoprecipitated ␣ and ␤ ENaC (Fig. 3D). Taken together, these studies show direct interactions between ENaC and ubiquitin, ENaC and epsin, and both ENaC and epsin with adaptors. This multisubunit complex is in full accord with the known binding determinants of these proteins.
To determine whether the interaction between ENaC subunits and epsin required ENaC ubiquitination, we performed pulldown experiments examining the interaction between ENaC and the UIM motif of epsin. We in vitro translated ␣, ␤, and ␥ ENaC, or each ENaC subunit fused to three tandem ubiquitins in the presence of [ 35 S]methionine and incubated with GST alone, GST-UIM, or GST-epsin, then pulled down the GST constructs with glutathione-Sepharose beads. Interactions between translated subunits and fusion proteins will shift the translated proteins from supernatant to pellet. As shown in Fig. 4, each ENaC subunit is found in pellet when conjugated to ubiquitin. There is a small amount of ENaC alone that is bound to GST-epsin, but the major shift of subunits from supernatant to pellet is seen with ubiquitinated forms. These experiments demonstrate that ENaC interacts with epsin strongly when ubiquitinated and that the UIM domain of epsin is sufficient for this interaction.
To examine the significance of the interaction between ENaC and epsin, the Xenopus oocyte expression system was employed. Although previous studies have demonstrated that ENaC activity in oocytes is negatively regulated by Nedd4-2, we examined this first as a baseline for further experiments. All three ENaC subunits were expressed, with or without Nedd4-2, and ENaC activity was measured as amiloride-sensitive current using the two-electrode voltage clamp as described (22). Fig. 5A demonstrates that co-expression of Nedd4-2 markedly down-regulated ENaC activity in oocytes as previously shown (1,6). We next examined the effect of epsin on ENaC activity in oocytes by expressing ␣␤␥ ENaC with or without fulllength epsin. As shown in Fig. 5B, epsin down-regulated ENaC activity almost to the extent seen with expression of Nedd4-2, consistent with a role for epsin in promoting endocytosis of the channel. The effect of epsin on ENaC surface expression was measured using a FLAG-tagged ␤ ENaC coexpressed with wild-type ␣␥ ENaC. We have previously shown that this construct behaves similarly to wild type ␣␤␥ ENaC in oocytes (22). As shown in Fig. 5C, epsin co-expression decreased ENaC surface expression to a similar degree as it down reduced the amiloride-sensitive current (Fig.  5B). Thus, the effect of epsin on ENaC activity is totally explained by decreased ENaC expression at the plasma membrane. To determine whether ENaC ubiquitination was required for the action of epsin on ENaC activity in oocytes, we coexpressed epsin with a truncation mutant of ␤ ENaC along with wild type ␣ and ␥ ENaC. Such mutations of C-terminal ␤ Non-immune serum did not immunoprecipitate epsin. B, CCD cell lysate was immunoprecipitated with antibodies to ␣ ENaC, ␤ ENaC, or ubiquitin (Ub). Both ␣ and ␤ ENaC co-IP with ubiquitin, and ubiquitin antibody co-IP with both ENaC subunits. IP with ubiquitin also immunoblots with epsin antibody. C, CCD cell lysate was immunoprecipitated with epsin antibody and immunoblotted with antibody to ␣ adaptin. D, CCV preparation was solubilized and IP with 2 antibody and probed for ␣ and ␤ ENaC. E, negative controls. We incubated beads alone, pre-immune serum (PIS), or ␣ ENaC antibody with cell lysate then blotted for ␣ ENaC. No bands are apparent with beads alone or nonspecific IgG (n ϭ 3). WB, Western blot. . ENaC interaction with epsin UIM domains is enhanced by their conjugation to ubiquitin. Each ENaC subunit alone, or conjugated to three tandem ubiquitins was in vitro translated in the presence of [ 35 S]methionine, and equal amounts of isotope were then incubated with GST, GST-UIM, or GST-epsin constructs as described under "Materials and Methods." Samples were then incubated with glutathione-Sepharose beads and beads were spun down. Samples from both pellet (P) (bound to beads) and supernatant (S) were then separated by SDS-PAGE and analyzed on a phosphorimager. Ubiquitin-conjugated ␤-ENaC conjugate aggregated on entry to the 10% SDS gels and did not fully migrate, therefore, ␤ ENaC samples were separated on 8% SDS-PAGE with 8 M urea. In vitro translated ␣, ␤, and ␥ ENaC were visualized at 79, 72, and 74 kDa, respectively, and were largely recovered in supernatant. There is a slight, background interaction between unmodified ENaC and both GST and GST-UIM, and all three subunits have a weak interaction with GST-epsin, but the majority of the in vitro translate is in the supernatant. ENaC fused to ubiquitin runs typically higher in the gels as seen with ␣, ␤, and ␥ subunits visualized at 107, 100, and 102 kDa, respectively. All three ENaC subunits fused to 3 tandem ubiquitins show significant accumulation in the pellet of GST-UIM and GST epsin, indicating an interaction (n ϭ 3).
ENaC eliminate the binding motif for the third WW domain of Nedd4-2 to ␤ ENaC and block ubiquitin-dependent down-regulation of the channel (6,31). As shown in Fig. 5D, epsin had no effect on the activity of ␣␤ R564X ␥ ENaC constructs when co-expressed in Xenopus oocytes. This result is consistent with the notion that epsin-mediated down-regulation of ENaC activity is dependent on interaction of the channel with Nedd4-2. To further demonstrate the requirement for ubiquitination in the ENaC-epsin interaction, we coexpressed wild type ENaC with a mutant epsin in which the UIM motif had been deleted. As shown in Fig. 5E, deletion of the epsin UIM motif abolished the epsin-mediated reduction of ENaC activity. We also examined the lysine to arginine mutations in N-terminal ␣ and ␥ ENaC that Staub and colleagues (6) described as inhibiting ubiquitination of the channel and increasing surface expression and activity of ENaC in oocytes. As shown in Fig. 6A, lysine to arginine mutations in the N termini of either ␣ or ␥ ENaC lead to modest increases in ENaC activity in oocytes compared with wild type channels, whereas expression of both mutant subunits along with wild type ␤ ENaC results in more substantial enhancement of ENaC activity as described by Staub et al. (6) and pointing to the possible significance of multiubiquitination of the channel (32,33). If the epsin-ENaC interaction is dependent on subunit ubiquitination, then we would expect that epsin would have less effect on channels expressing lysine to arginine mutations that inhibit ubiquitination. This result is seen in Fig. 6B. As shown previously, epsin decreased the activity of wild type ␣␤␥ ENaC but had no effect on the activity of ␣ K47R,K50F ␤␥ K6 -13R channels. As a further control, we examined the effects of epsin on activity of a distinct ion channel, ROMK. ROMK activity in oocytes was unaffected by co expression of epsin (Fig. 7A).
Finally, we examined several of the domains of epsin to determine whether they would affect ENaC activity in oocytes. The central clathrin-binding domain of epsin is characterized by a clathrin box motif and multiple Asp-Pro-Trp (DPW) repeats (10) (Fig. 7B). If epsin were involved in endocytosis of the channel, this domain should act as a dominant negative and stimulate ENaC activity. The ENTH domain of epsin is a phosphoinositide binding domain thought to be important for either membrane localization of epsin, and/or to facilitate membrane curvature by its interaction with phosphoinositides (34,35). Both domains were co-expressed with ENaC subunits and activity was compared with ENaC alone (Fig. 7C). As expected, the DPW domain acted as a dominant negative and stimulated ENaC activity, and a similar result was obtained with co-expression of the ENTH domain.

DISCUSSION
The apical membrane residence of ENaC has been shown to be primarily regulated by the ubiquitin ligase Nedd4-2 (3). This process can be Whole cell amiloride-sensitive currents were measured 24 h following injection. Results are displayed as normalized ␣␤␥ alone mean currents and show that Nedd4-2 significantly inhibits ENaC activity (p Ͻ 0.001 n ϭ 10). B, oocytes were injected with RNA for ␣␤␥ ENaC alone or in combination with RNA for epsin. Epsin significantly inhibits ENaC activity ( p Ͻ 0.001, n ϭ 14). C, ENaC composed of wild type ␣ and ␥ co-expressed with ␤-FLAG (␤ f ) were expressed with and without epsin with wild type ␣␤␥ ENaC as a control. After 24 h incubation, surface expression was measured using anti-FLAG antibody as described. No luminescence was seen with non-FLAG expressing ENaC. Epsin significantly decreased surface expression ( p Ͻ 0.001, n ϭ 25). D, effect of epsin on C-terminal truncation ␤ ENaC (␤R564X ) coexpressed with wild type ␣␥ ENaC. Epsin had no effect on ENaC containing C-terminal truncation of ␤ ENaC (p ϭ NS, n ϭ 9 -12). E, effect of epsin with UIM deletion on ENaC activity. Epsin lacking its UIM had no effect on the amiloride-sensitive current ( p ϭ NS, n ϭ 19).

FIGURE 6. Effect of N-terminal lysine mutation of ␣ and ␥ ENaC on functional expression and regulation by epsin.
A, lysine sites of ubiquitination on N-terminal ␣ and ␥ ENaC were mutated to arginines as described under "Materials and Methods" and co-expressed with wild type non-mutated subunits (␣ K47R,K50R with wild type ␤␥, ␥ K6 -13R ␣␤) or both mutated subunits expressed together with wild type ␤ ENaC and their activity in oocytes compared with that of the native channel subunits. Each mutation resulted in a significant enhancement of current, but enhancement was significantly greater when both mutations were co-expressed (p Ͻ 0.001, n ϭ 20 -22). B, N-terminal lysine mutations in ␣ and ␥ ENaC were coexpressed with wild type ␤ ENaC and the activity of the channels and their response to co-injected epsin were compared. Epsin significantly down-regulated native channels (p Ͻ 0.001, n ϭ 20) and mutant ␣␥ subunits significantly enhance the amiloride-sensitive current (p Ͻ 0.001, n ϭ 20). Epsin has no significant effect on ENaC containing mutated subunits (#, p ϭ NS, n ϭ 22). regulated by kinases that alter the interaction between Nedd4-2 and ENaC (2,3) or by mutations of the ␤ and ␥ subunits of the channel that block ubiquitination and lead to enhanced surface expression, and, in humans, hypertension (2,31). The mechanism(s) by which ubiquitinated ENaC is internalized from the apical membrane are, however, not entirely known. Evidence from oocyte studies demonstrate that the process is dynamin-dependent, consistent with a role for clathrin-mediated endocytosis (7). The partners that link ubiquitinated cargo, such as ENaC, to clathrin-coated vesicles are a subject of considerable study and interest (8,9). Multimodular proteins such as epsin, which have domains that interact with ubiquitin and domains that interact with clathrin adaptors, have been proposed to play such a role (10,11), and epsin does down-regulate ENaC activity when co-expressed with ENaC in Chinese hamster ovary cells (12). We sought to determine whether ENaC surface expression might be regulated by clathrin-mediated endocytosis, and whether epsin is involved in this process.
Our results demonstrate that ENaC is present in clathrin-coated vesicles in CCD cells, and is efficiently endocytosed, consistent with previous observations that channel activity is regulated in a dynamin-dependent manner (7). ENaC is ubiquitinated, as would be expected for a protein whose surface expression is negatively regulated by ubiquitin ligase (1). There are some differences in the degree of ubiquitination of ENaC shown here for the CCD cells and that demonstrated by others in either human embryonic kidney cells overexpressing ENaC (6) or endogenously expressing A6 cells (36). These differences may be due to the different cell systems or antibodies employed or to rates of deubiquitination. Nevertheless, all studies have shown ubiquitination of multiple subunits of this channel. . Effect of epsin on ROMK activity and effect of epsin domains on ENaC activity. A, RNA for ROMK was injected into oocytes with and without full-length epsin and activity was measured at 24 h. Co-expression of epsin had no effect on ROMK activity. B, schematic of full-length epsin 1 and the ENTH and DPW domains used for oocyte injections. C, amiloride-sensitive currents from ␣␤␥ ENaC alone were compared with currents in the presence of either DPW or ENTH domains of epsin. Co-expression of each domain with ENaC resulted in a significant enhancement of ENaC activity (p Ͻ 0.005, n ϭ 32). Statistical comparisons were made by unpaired t test, or analysis of variance followed by t test for multiple comparisons.
ENaC is a tetrameric channel composed of two ␣, one ␤, and one ␥ subunit. Studies by Staub and colleagues (6) demonstrated ubiquitination of ␣ and ␥ ENaC expressed in human embryonic kidney cells and inhibition of ubiquitination of only ␥ ENaC by mutations of lysine residues in the N terminus resulted in increased surface expression due to decreased retrieval of the channel. Additional mutations to the ␣ subunit augmented this effect, suggesting that multiubiquitination of the subunits was important to ubiquitin-dependent retrieval (see Fig. 6A).
These observations leave open the question of how ubiquitination of multiple ENaC subunits leads to endocytosis that would be dynamindependent. We show that the channel subunits interact with epsin in CCD cells, and ENaC and epsin interact with clathrin adaptors in these cells. Epsin negatively regulates channel activity in oocytes, as it does in Chinese hamster ovary cells (12) and this effect is dependent on the UIM motif of epsin. Moreover, the effect of epsin on ENaC activity can be abolished by mutations in the channel that block its ability to interact with the ubiquitin ligase Nedd4-2 or that remove the sites for its ubiquitination. Because the effect of epsin on ENaC activity was dependent both on ubiquitination of ENaC and the presence of the UIM domain of epsin, we examined the interaction in vitro between ENaC and epsin. Only ubiquitinated ENaC is seen to interact significantly with epsin, and the same result is seen with expression of the UIM domain of epsin alone. Our results are therefore consistent with previous observations that surface expression of ENaC is regulated in both a dynamin and ubiquitin-dependent manner (6, 7), but they implicate epsin as the link between ubiquitinated ENaC on the cell surface and clathrin adaptors. Our results do not rule out the possibility that some portion of ENaC internalization may also be mediated by caveolae as has been described for another ubiquitinated membrane protein, the epidermal growth factor receptor (37).
Epsin appears to have no effect of a distinct ion channel, ROMK, when co-expressed in oocytes, suggesting that the effect of epsin is not to globally stimulate endocytosis. The effects of the non-UIM domains of epsin on ENaC are also of interest. The dominant negative effect of the DPW domain of epsin is consistent with AP-2 dependence of this process, simply titrating adaptor proteins; and would be expected for any protein whose surface expression is regulated by clathrin-mediated endocytosis. The similar effect of the ENTH domain suggests that phosphoinositide interactions are important in regulating ENaC internalization. Taken together, the results are consistent with a model where epsin links multiubiquitinated cargo to adaptor complexes and regulates endocytosis.