Interactions of (cid:1) and (cid:2) ENaC with Nedd4 Can Be Facilitated by an ERK-mediated Phosphorylation*

Phosphorylation of the epithelial Na (cid:3) channel (ENaC) has been suggested to play a role in its regulation. Here we demonstrate that phosphorylating the carboxyl termini of the (cid:1) and (cid:2) subunits facilitates their interactions with the ubiquitin ligase Nedd4 and inhibits channel activity. Three protein kinases, which phosphorylate the carboxyl termini of (cid:1) and (cid:2) ENaC, have been identi-fied by an in vitro assay. One of these phosphorylates (cid:1) Thr-613 and (cid:2) Thr-623, well-conserved C-tail threonines in the immediate vicinity of the PY motifs. Phosphorylation of (cid:2) Thr-623 has also been demonstrated in vivo in channels expressed in Xenopus oocytes, and mutating (cid:1) Thr-613 and (cid:2) Thr-623 into alanine increased the channel activity by 3.5-fold. Effects of the above phosphoryl-ations on interactions between ENaC and Nedd4 have been studied using surface plasmon resonance. Peptides having phospho-threonine at positions (cid:1) 613 or (cid:2) 623 bind the WW domains of Nedd4 two

Phosphorylation of the epithelial Na ؉ channel (ENaC) has been suggested to play a role in its regulation. Here we demonstrate that phosphorylating the carboxyl termini of the ␤ and ␥ subunits facilitates their interactions with the ubiquitin ligase Nedd4 and inhibits channel activity. Three protein kinases, which phosphorylate the carboxyl termini of ␤ and ␥ENaC, have been identified by an in vitro assay. One of these phosphorylates ␤Thr-613 and ␥Thr-623, well-conserved C-tail threonines in the immediate vicinity of the PY motifs. Phosphorylation of ␥Thr-623 has also been demonstrated in vivo in channels expressed in Xenopus oocytes, and mutating ␤Thr-613 and ␥Thr-623 into alanine increased the channel activity by 3.5-fold. Effects of the above phosphorylations on interactions between ENaC and Nedd4 have been studied using surface plasmon resonance. Peptides having phospho-threonine at positions ␤613 or ␥623 bind the WW domains of Nedd4 two to three times better than the non-phosphorylated analogues, due to higher association rate constants. Using a number of different approaches it was demonstrated that the protein kinase acting on ␤Thr-613 and ␥Thr-623 is the extracellular regulated kinase (ERK). It is suggested that an ERKmediated phosphorylation of ␤Thr-613 and ␥Thr-623 down-regulates the channel by facilitating its interaction with Nedd4.
Active Na ϩ reabsorption in kidney collecting duct, distal colon, lung, and exocrine glands is mediated by an apical amiloride-blockable Na ϩ channel (1)(2)(3). The channel is composed of three homologous subunits, denoted ␣, ␤, and ␥ENaC (Epithelial Na ϩ Channel). 1 Its cell surface expression is determined by interactions of the C-tails of ␤ and ␥ with the ubiquitin ligase Nedd4. The WW domains of Nedd4 bind to the proline-rich PY motifs on ␤ and ␥ENaC, leading to channel ubiquitination, internalization, and degradation (4,5). A central role of ENaC in maintaining salt and water balance has been conclusively demonstrated by identifying genetic diseases associated with mutations in channel subunits, as well as by the phenotypic analysis of ENaC knockout mice (for review see Refs. [1][2][3]. ENaC's activity is also controlled by a number of hormones such as the mineralocorticoid aldosterone, the anti-diuretic peptide vasopressin, and insulin (1,3). Previous studies have suggested the involvement of protein phosphorylation in these mechanisms. The serine/threonine kinase sgk (serum and glucocorticoid-dependent kinase) is induced by aldosterone and can activate the channel upon co-expression in Xenopus oocytes (6 -8). This response was recently found to involve phosphorylation of Nedd4-2 by sgk (9,10). The response of A6 cells to both aldosterone and insulin requires activation of phosphoinositide 3-kinase (11,12). In addition, aldosterone and insulin, as well as intracellular signaling components such as protein kinases C and A, increase the in vivo phosphorylation of the carboxyl termini of both ␤ and ␥ENaC (13).
We have recently demonstrated phosphorylation of the carboxyl termini of ENaC subunits, expressed as glutathione Stransferase (GST) fusion proteins by crude cytosolic fractions (14). The current study characterizes conserved residues phosphorylated in the carboxyl termini of ENaC subunits, explores their physiological role, and identifies the kinase involved. The data indicate that an extracellular regulated kinase (ERK)-dependent phosphorylation of ␤Thr-613 and ␥Thr-623 may be important in controlling interactions between the channel and Nedd4.

EXPERIMENTAL PROCEDURES
Recombinant DNA and Proteins-The carboxyl termini of the rat ␤ and ␥ENaC (␤ 557-638 and ␥ 564 -650) were subcloned downstream GST in the bacterial expression vector pGEX3X as described in a previous study (14). cDNA constructs expressing fusion proteins between GST and the three WW domains of rat Nedd4 were kindly provided by D. Rotin (Hospital for Sick Children, Toronto, Canada) and are described in a previous study (5). GST fusion proteins were expressed in a protease-deficient Escherichia coli strain and purified on glutathione beads as detailed previously (14). Functional expression in Xenopus oocytes was done using ENaC clones in pSPORT-1, obtained from B. C. Rossier (Institute of Pharmacology, University of Lausanne). Immunoprecipitation of phosphorylated ENaC was performed using a hemagglutinin A (HA)-tagged ␤ and ␥ construct kindly provided by B. Schwappach (Zentrum fü r Molekulare Biologie, Universitä t Heidelberg). The HA epitope was introduced in the ecto domains of these subunits at a position shown before not to affect channel activity (15). Point mutations in the various constructs were introduced using a QuikChange site-directed mutagenesis kit and verified by sequencing.
Extraction, Fractionation, and Assay of Cytosolic Protein Kinases-Cytosol was extracted from rat distal colon and fractionated by ion exchange chromatography using the following protocol. Rats (Wistar, 9 -11 weeks old) were sacrificed by cervical dislocation. The distal colon was excised, cut open, and rinsed first in phosphate-buffered saline and then in buffer A composed of: 50 mM ␤-glycerophosphate, pH 7.3, 1.5 mM * This work was supported by research grants from the Israel Science Foundation and the United States-Israel Binational Science Foundation (to H. G. and E. R.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ Both authors contributed equally to this work. EGTA, 1.0 mM EDTA, 1.0 mM dithiothreitol, and 0.1 mM de-aerated sodium orthovanadate (16). The epithelial cells were scraped off the connective tissue using a glass slide and suspended in buffer A ϩ a mixture of protease inhibitors (1 mM phenylmethylsulfonyl fluoride, 1.0 mM benzamidine, 10 g/ml aprotinin, 10 g/ml leupeptin, and 2.0 g/ml pepstatin-A). Cells were washed by centrifugation and then disrupted on ice by sonication (3 ϫ 5 s). Cell homogenates were centrifuged for 30 min at 30,000 ϫ g, and the supernatants were collected and further fractionated by ion exchange chromatography. 10 mg of protein in 50 ml of buffer A was loaded onto a MonoQ column (Amersham Biosciences, Inc.) at a rate of 1 ml/min. Bound proteins were eluted at the same rate by a linear NaCl gradient (0 -30%) in buffer A. Ninety 1-ml fractions were collected, stored at 4°C, and assayed for kinase activity within 24 h.
Chinese hamster ovary (CHO) cells stably transfected with the human insulin receptor were kindly provided by Dr. Yehiel Zick (Department of Molecular Cell Biology, The Weizmann Institute of Science, Rehovot, Israel). Cells were cultivated in a 1:1 mixture of Dulbecco's modified Eagle's medium and F12 supplemented with 10% fetal calf serum and 1% glutamine. After achieving 80% confluency, the medium was replaced by serum-free Dulbecco's modified Eagle's medium for a period of 24 -36 h. Serum-depleted cells were incubated with 100 nM insulin for 5-120 min in the presence and absence of various inhibitors. They were washed in ice-cold phosphate-buffered saline, harvested, and suspended in buffer A ϩ protease inhibitors. Cells were disrupted by 7-s sonication on ice. The homogenates were centrifuged for 60 min at 15,000 ϫ g, and the supernatants were collected.
Rat colon cytosol MonoQ fractions, whole CHO cytosol, and purified activated ERK2 were used in an in vitro phosphorylation assay of GST-ENaC fusion proteins. The fusion proteins were immobilized on glutathione-agarose beads and suspended in buffer A. Aliquots of 47 l were mixed with 50-l volumes of cytosolic fractions, whole cytosol (ϳ1 mg/ml protein), or purified ERK2 (1.5 ng/ml, specific activity of 0.45 mmol/min/mg), all in buffer A. Phosphorylation was initiated by the addition of 3 l of ATP mixture composed of: 1.8 l of 1 M MgCl 2 , 0.9 l of 0.2 mM ATP, and 0.3 l of [ 32 P]ATP (10 mCi/ml, 3000 Ci/mmol). Suspensions were shaken for 30 min at 30°C, pelleted, and washed several times in HB1B buffer (20 mM HEPES, pH 7.7, 50 mM NaCl, 0.1 mM EDTA, 25 mM MgCl 2 , and 0.05% Triton X-100). Pellets were suspended in Laemmli sample buffer, boiled for 5 min, resolved on 12% SDS-PAGE, and exposed to a PhosphorImager plate and/or x-ray film.
Specific binding of cytosolic kinases to ENaC C-tail fusion proteins was determined by co-precipitation, as described previously (17). In brief, fusion proteins immobilized on glutathione beads were incubated with either cytosolic fractions or purified kinases for 2 h. at 4°C. Incubation was done in a binding buffer composed of 150 mM NaCl, 22 mM HEPES, pH 7.7, 2 mM MgCl 2 , 0.075% Triton X-100, 20 mM ␤-glycerophosphate, 0.1 mM EDTA, 0.1 mM sodium orthovanadate, and protease inhibitors. The beads were sedimented and washed twice in HB1B buffer, a third wash in buffer A, and a final wash in a kinase assay buffer (20 mM HEPES, pH 7.7, 20 mM MgCl 2 , 20 mM ␤-glycerophosphate, 2 mM dithiothreitol, and 0.1 mM sodium orthovanadate). Beads were suspended in 30-l volumes of the above kinase assay buffer, and phosphorylation was initiated by the addition of 2 M ATP plus 2 Ci of [␥-32 P]ATP.
BIAcore Experiments-Binding of the WW domains of Nedd4 to ␤ and ␥ PY peptides was monitored by surface plasmon resonance in a BIAcore 2000 sensor, as described previously (18). In brief, peptides serving as substrates were immobilized on the sensor chip (SA, BIAcore, Uppsala Sweden) through amino termini biotin residues. In each chip, the peptides to be compared were immobilized on two of the channels. Free biotin and ␥ peptide carrying the mutationY628A were immobilized on the other two channels and served as negative controls. The three WW domains of rat Nedd4 were expressed as GST fusion proteins and used as analytes. They were suspended in HBS buffer (10 mM HEPES, pH 7.4, 140 mM NaCl, 3.4 mM EDTA, and 0.005% P20) and injected at a flow rate of 20 l/min. Binding was monitored simultaneously in all four channels for at least 4 min and terminated by the application of HBS buffer without analyte (dissociation phase). The chip was regenerated by a subsequent injection of 10 l of HBS ϩ 0.05% SDS and extensively washed in HBS, and then the next analyte was applied. Each observation was confirmed by at least three measurements in two different chips.
Functional Expression and Phosphorylation in Xenopus Oocytes-The ENaC-mediated Na ϩ current was determined in the oocyte expression system as described before (8,14). In brief, stage V-VI oocytes were injected with cRNA mixtures containing 2.5 ng of each of the ENaC subunits. Oocytes were incubated at 17°C in a medium that contained 96 mM NaCl and 10 M amiloride. Electrophysiological measurements were performed 48 -72 h later by means of a two-electrode voltage clamp technique. Channel activity was determined as the amiloridesensitive current amplitudes monitored at Ϫ100 mV. Phosphorylation was examined in oocytes injected with cRNA mixtures in which either ␤ or ␥ were HA-tagged. Groups of ϳ40 ENaC-expressing oocytes were incubated for 4 h with 3.3 mCi of orthophosphate ( 32 P i ). A mixture of orthovanadate (0.5 mM) and H 2 O 2 (1 mM) was added in the last 60 min of the incubation to stimulate endogenous kinases. Another aliquot of ϳ40 injected oocytes was metabolically labeled by a 2-day incubation with 100 Ci of [ 35 S]methionine. Oocytes labeled with 32 P or 35 S were washed and homogenized in buffer A ϩ protease inhibitors, and membranes were isolated by centrifugation through a sucrose cushion. Membranes were solubilized in 1% Triton X-100 in buffer A and centrifuged for 5 min at 11,000 ϫ g to remove insoluble material. Aliquots of ϳ200 l of detergent-soluble membrane protein extracts were incubated for 12-16 h at 4°C with 2 g of a mouse monoclonal anti-HA antibody (clone 12CA5, Roche Molecular Biochemicals) and then for another 2 h with protein A-Sepharose beads. The beads were sedimented, washed twice in buffer A ϩ 0.1% Triton X-100, and a third time in buffer A ϩ 0.5 M LiCl. Immunopellets were suspended in SDS sample buffer, resolved on 8% SDS-PAGE, and assayed for radioactivity.

RESULTS
In Vitro and in Vivo Phosphorylation of ENaC-Cytosol extracted from rat distal colon has been used to identify protein kinases capable of phosphorylating ENaC subunits. The colonic tissue was selected, because it is relatively homogenous and has high Na ϩ channel abundance. Proteins were fractionated on a MonoQ column, and fractions were tested for their ability to phosphorylate fusion proteins having the carboxyl termini of ␥ ENaC. Fig. 1 depicts a typical assay demonstrating the existence of at least three protein kinase-enriched fractions phosphorylating GST-␥. The first peak of protein kinase activity was eluted around fractions 38 -41 (ϳ0.07-0.09 M NaCl) and is characterized in this study. A second peak, seen around fraction 59 (ϳ0.19 M NaCl), was found to phosphorylate residue ␥Thr-630 (data not shown). Hence, it is likely to be the same kinase detected in a crude DE-52 fraction reported before (14). A third, major peak was seen around fractions 74 -77 (ϳ0.25-0.27 M NaCl) and will be described elsewhere.
Initial experiments have used mutagenesis to identify ENaC residues phosphorylated by the kinase eluted around fraction 40. Phosphorylation of GST-␥ appeared to be on Thr-623, and mutating this site to Ala residue completely inhibited incorpo-FIG. 1. Fractionation of protein kinases that phosphorylate the carboxyl tail of ␥ENaC. Rat colon cytosol was fractionated on a MonoQ column, and different fractions were tested for their ability to incorporate 32 P into GST-␥ C-tail, as described under "Experimental Procedures." Three major peaks of kinase activity are marked by horizontal bars. ration of 32 P into this fusion protein ( Fig. 2A). Mutating ␥Thr-630, another residue found to be phosphorylated by crude cytosol, had no effect. Cytosolic fractions 38 -41 phosphorylated GST-␤ as well. In this case, however, mutating the ␤ residue analogous to ␥Thr-623 (␤Thr-613) into an Ala residue had only a minor effect on the incorporation of 32 P into this fusion protein (not shown). This may be due to the phosphorylation of additional residues on the carboxyl tail of ␤. Phosphorylation of ␤Thr-613 could be convincingly shown in subsequent experiments using purified kinase and cytosol from CHO cells.
It was further demonstrated that the protein kinase eluted in fraction 40 tightly binds to the ␥ C-tail and can be coprecipitated with it. In this assay, GST-␥ immobilized on glutathione beads was incubated with the above cytosolic fraction in the absence of ATP. The beads were washed several times and then incubated with [␥-32 P]ATP with no added protein kinases. Substantial phosphorylation of the fusion protein can take place under these conditions only if some of the kinase is precipitated with its substrate. As shown in Fig. 2B, GST-␥ could be effectively phosphorylated following such co-precipitation, and the phosphorylation occurred on Thr-623.
Next, it was determined that the above phosphorylation takes place also in Xenopus oocytes expressing the three ENaC subunits. Accordingly, HA-tagged wild type and mutated subunits were translated in the oocyte system and metabolically labeled with either [ 35 S]methionine or 32 P i . Orthovanadate and H 2 O 2 were added during the incubation with 32 P to achieve activation of endogenous protein kinases. Immunoprecipitation of HA-tagged ␤ and ␥ demonstrated phosphorylation of both subunits (Fig. 3A). Mutating ␥Thr-623 into an Ala residue significantly decreased incorporation of 32 P into this subunit without affecting labeling by [ 35 S]methionine. In three different experiments using ϳ40 oocytes each, the mutation of Thr-623 lowered the 32 P/ 35 S ratio of the immunoprecipitated subunit by 47 Ϯ 15%. Thus, Thr-623 is one of the ␥ residues phosphorylated in the oocyte system. Phosphorylation of the ␤ subunit on the other hand, was not significantly affected by mutating Thr-613 into Ala. Thus, this residue is either not phosphorylated in oocytes or its phosphorylation is masked by 32 P incorporation into other residues. This may be similar to the phenomenon described above where mutating ␤Thr-613 into Ala had a minor effect on the overall ␤ phosphorylation by fractionated cytosol, even though it is clearly one of the phosphorylated residues (see subsequent experiments in CHO-T cells). Expressing the double mutant ␤T613A/␥T623A in oocytes resulted in macroscopic Na ϩ currents that were 3.5-fold higher than that evoked by the wild type channel (Fig. 3B). Mutating the same residues into glutamic acid had a smaller but nevertheless significant stimulatory effect. Thus, ␥Thr-623 and/or ␤Thr-613 have a functional role and their mutation largely activates the channel. However, because mutation to a neutral or negatively charged amino acid evoked the same response, it is not certain that this activation is due to the inability to phosphorylate threonine 613/623.
Effects of Phosphorylation on the Channel/Nedd4 Interactions-␥Thr-623 and ␤Thr-613 are well-conserved residues located immediately before the PY motifs (Fig. 2C). Their phosphorylation could affect the channel by altering its interaction with Nedd4. Assessment of such a mechanism has been done by determining association of the WW regions of Nedd4 with phosphorylated and non-phosphorylated PY peptides using BIAcore (18). Accordingly, various ␤ and ␥ peptides, listed in Table I,  were synthesized and attached to streptavidin-coated sensor chips via an amino-terminal biotin moiety. The three WW domains of rat Nedd4, expressed as GST fusion proteins, were passed over the chip and binding was recorded as a change in refractive index. A typical sensogram-monitoring interactions of WW2 to ␥PY peptides is illustrated in Fig. 4A. Binding of the recombinant protein to the phosphorylated peptide (␥ pThr-623) was 2-to 3-fold higher than that recorded for the nonphosphorylated analogue (␥). No significant binding was observed with a peptide carrying a point mutation that impairs Nedd4-ENaC interactions (␥ Y628A), or to free biotin. The small rapid signal seen in these channels is probably due to bulk or nonspecific effects. Therefore, in all subsequent experiments the signal obtained in the biotin channel was subtracted from readings in other channels to obtain ENaC-specific signals. We have also tested a ␥ peptide with phospho-threonine at position 630, shown to be phosphorylated by another cytosolic fraction (14). This peptide (␥ pThr-630) appeared to bind WW2 at a rate that was close to that of the wild type peptide (Fig. 4B).
The above protocol was repeated for different concentrations of all three WW domains using both ␤ and ␥ peptides (Fig. 5). All three WW domains were found to bind better to PY peptides that have phospho-threonine at position ␤613 or ␥623. The experiment of Fig. 5 also confirms a previous observation that WW2 and WW3 bind PY sequences much better than WW1 (19,20). Additional competition experiments summarized in Fig. 6, further establishes the above findings. In this experiment the WW3 fusion protein was preincubated in solution with different non-biotinylated peptides, and then applied to the sensor chip. The competing soluble peptide largely inhibited association of WW3 with the immobilized peptide, and phospho-peptides were more effective competitors then the non-phosphorylated analogues. This result was observed irrespective of the immobilized substrate (␤ or ␥) or analyte (WW2 or WW3, data not shown). It demonstrates that the phosphorylated peptides better associate with the WW fusion proteins also in solution; i.e. it rules out the possibility that their larger binding reflects differences in peptide immobilization or packing on the chip.
Although the above data clearly show effects of phosphorylation on the interactions between WW proteins and PY peptides, kinetic analysis of these effects was not trivial. As reported previously, the association could not be well fitted to simple 1:1 Langmuir binding and a considerable fraction of the response was attributed to a very slow and non-saturable component (18). This phase was preceded by a faster saturable component that was particularly noticeable at high concentrations of the analytes. This phenomenon was independent of the flow rate (20 -75 l/min) and hence not likely to reflect mass transfer effects. Data could in principle be fitted to a sum of two exponents, but the fit was satisfactory only for the interaction of ␤ and ␤ pThr-613 with WW3. In this case, the faster event had the characteristics of a simple 1:1 binding and its apparent association rate constants (k s ) showed the expected linear dependence on the analyte concentration (Fig. 7). The dissociation rate constants (k d ) were concentration independent. The kinetics parameters extracted from this analysis, and the equi-librium constants calculated as ratios of the association and dissociation rate constants, are summarized in Table II. The phospho-␤ peptide was found to have ϳ4-fold higher affinity toward WW3, due to an increased association rate constant. The second, slower phase could not be analyzed kinetically, but this component too was markedly affected by phosphorylation.
Identifying the Protein Kinase Phosphorylating ␤Thr-613 and ␥Thr-623-Next, we studied the identity of the cytosolic kinase eluted in fraction 40, which phosphorylates ␥Thr-623. The presence of multiple proline residues near the phosphorylation site suggested involvement of a proline-directed kinase, e.g. a member of the MAPK family such as ERK, p38, or c-Jun amino-terminal kinase (JNK). In particular, involvement of ERK seemed likely, because the sequence PXTP is a typical ERK phosphorylation motif (21). This possibility was assessed in the experiments depicted in Figs. 8 and 9. First, the active rat colon fractions were tested for the presence of this kinase using antibodies against total and active (phosphorylated) ERK. Indeed, both ERK1 and -2 were present in the same MonoQ fractions that incorporate 32 P into ␥Thr-623, and the abundance of active ERK in the various fractions correlated with their phosphorylation activity (Fig. 8, top and middle  panels). Other experiments tested phosphorylation of the ␤ and ␥ fusion proteins by purified activated ERK2. This protein kinase effectively incorporated 32 P from ATP into both GST-␤  and -␥, and phosphorylation took place on residues Thr-613 and Thr-623, respectively (Fig. 9A). ERK2 was also tightly bound to the two fusion proteins and could be co-precipitated by each of them (Fig. 9B).
Subsequent experiments have used CHO cells overexpressing the insulin receptor (CHO-T) as a means to further establish the involvement of ERK in the above phosphorylation. Stimulating these cells by insulin evokes a transient activation of ERK as well as other intracellular signaling kinases (22).
Hence, they provide a convenient tool for examining the role of these kinases in the phosphorylation of GST-␥. We found that the C-tail fusion proteins of ENaC were phosphorylated mainly by two protein kinases present in CHO-T; ERK and casein kinase 2 (CK2). 2 Therefore, it was possible to study phosphorylation of ␤613 and ␥623 using whole, non-fractionated, cytosol providing that CK2 was completely inhibited by heparin or that its major phosphorylation sites (␥Thr-599, ␤Ser-631) were mutated. Under these conditions phosphorylation of GST-␤ and -␥ by whole CHO-T cytosol took place only on residues ␤Thr-613 and ␥Thr-623, respectively (Fig. 10A) 3 It should be stressed that, in the experiments of Fig. 10 and 11, insulin was used as a means to achieve regulated activation of MAPK FIG. 6. Competition of phosphorylated and non-phosphorylated peptides. Aliquots of WW3 (0.6 M) were preincubated with either diluent or different peptides (250 g/ml). The mixtures were applied onto the sensor chip, and binding to phosphorylated ␤ peptide was monitored. A, sequential recordings following the injections of: WW3, WW3 ϩ ␤, WW3 ϩ ␤p, WW3 ϩ ␥, WW3 ϩ ␥p, and WW3. Injection of the analyte (first arrow) was followed by injection of HBS (second arrow). The chip was then regenerated with 10 l of HSB ϩ 0.05% SDS (third arrow) and finally equilibrated with HBS (fourth arrow). B, means Ϯ S.E. of the maximal bindings recorded in three independent experiments. Data are presented as fraction of WW3 binding in the absence of competing peptide.

TABLE II
Kinetic parameters of the interaction between WW3 and ␤ peptides The association rate constants (k a ) were calculated from slopes of the best-fitted lines from plots of ks versus the analyte concentration (Fig.  7, rϽ0.99). The dissociation rate constants (k d ) are means Ϯ S.D. of values obtained for different concentrations of the analyte. The equilibrium constants (K D ) are k a /k d . FIG. 9. Phosphorylation of GST-␤ and ␥ by ERK2. A, wild type and mutated GST-␤ and ␥ were phosphorylated by purified activated ERK2 (final concentration, 0.7 ng/ml; specific activity, 0.45 mmol of phosphate/min/mg of protein). B, phosphorylation was done following co-precipitation of the fusion proteins and ERK2 as described under "Experimental Procedures." Phosphorylation of both subunits was strongly stimulated by insulin (Fig. 10B). This response was blocked by PD98059 but slightly activated by SB203580 (Fig. 10, B and C). PD98059 is a specific inhibitor of ERK activation, interacting with the upstream kinase (MAPK kinase) (23). SB203580, on the other hand, specifically inhibits p38 MAPK. The slight increase in phosphorylation evoked by this compound probably reflects the activation of Raf-1 and a subsequent stimulation of ERK (24). The above interpretation was further confirmed by immunoblots of CHO-T cytosol with antibodies against phosphorylated and total ERK (Fig. 10D). Insulin, PD98059, and SB203580 produced the expected changes in the abundance of phosphorylated ERK without affecting the total abundance of this protein kinase.
It was further shown that the capacity of CHO-T cytosol to phosphorylate GST-␥ correlates in time with the activation of ERK (Fig. 11, A and B). In this experiment, cytosol was extracted from CHO-T cells that were incubated for different periods of time with insulin. It was then assayed for the activation of ERK (using the anti-phospho-ERK antibody) and the phosphorylation of ␥Thr-623. Both events showed the same time dependence indicating involvement of the same kinase. Finally, it was demonstrated that GST-␥ precipitates ERK from insulin-stimulated CHO-T cells. The co-precipitation protocol was carried out using GST and GST-␥, and the precipitated proteins were analyzed by immunoblotting with the antiphospho ERK antibody. The antibody did recognize a band of ϳ42 kDa the abundance of which was affected by insulin treatment (Fig. 11C). This protein was precipitated by GST-␥ but not by GST. DISCUSSION The current study describes phosphorylation of a conserved threonine in the carboxyl termini of ␤ and ␥ENaC, identifies the protein kinase involved, and demonstrates a potential physiological role for such phosphorylation. An in vitro phosphorylation assay using GST fusion proteins has demonstrated incorporation of 32 P into ␤Thr-613 and ␥Thr-623. These residues and the prolines around them are well conserved through evolution. In fact the position ␤Thr-613/␥The-623 is one of only three cytoplasmic serine/threonines fully conserved in all ␤ and ␥ENaC subunits cloned so far (the other two are ␤Ser-620/ ␥Thr-630 and ␤Ser-631/␥Thr-644). The following evidence suggests that the protein kinase phosphorylating ␤Thr-613 and ␥Thr-623 is ERK. (i) ERK is co-precipitated from whole cytosol by GST-␥ and -␤ and phosphorylates them at ␥Thr-623/␤Thr-613. (ii) This protein kinase is eluted from the MonoQ column in the same fractions found to incorporate 32 P into ␥Thr-623. (iii) Purified activated ERK2 can bind the carboxyl termini of ␤ and ␥ and phosphorylate ␤Thr-613 and ␥Thr-623. (iv) The insulin-induced phosphorylation of GST-␥ by CHO-T cytosol has a kinetic and pharmacological profile that fits well activation of ERK1/2. Because identification of ERK as the kinase phosphorylating ␤Thr-613 and ␥Thr-623 is based on in vitro measurements only, it is in principle possible that other kinases too may phosphorylate these residues in vivo.
It was also demonstrated that at least ␥Thr-623 is phosphorylated in Xenopus oocytes. In addition, injecting oocytes with doubly mutated channel T613A/T623A elevated the amiloride-cascades. Its effects on the phosphorylation of ENaC are not necessarily related to its ability to activate this channel in kidney derived cells (23). blockable Na ϩ current by 3.4 Ϯ 0.8-fold. This is consistent with previous data showing a ϳ2-fold increase of Na ϩ current in oocytes expressing the singly mutated channel ␤T613A (25). Mutating these residues into glutamic acid (which may mimic phospho-threonine) failed to produce the opposite effect (Fig.  3). These data may indicate that 1) glutamic acid behaves differently than phospho-threonine in this particular case and 2) the activation observed upon mutating ␤Thr-613 and ␥Thr-623 is independent of their phosphorylation.
Because the phosphorylated threonine flanks the PY motifs of ␤ and ␥, we have tested the possibility that it influences interaction with Nedd4. Binding was assessed by surface plasmon resonance using phosphorylated and non-phosphorylated synthetic peptides corresponding to the PY domains. The use of synthetic peptides was required, because the GST fusion proteins were only ϳ10% phosphorylated. The following three criteria have established that the observed associations of GST-WW with the immobilized PY peptides are specific and resemble the physiological interaction between ENaC and Nedd4. (i) Binding was largely inhibited by mutating the essential tyrosine ␥Y628 into alanine. (ii) Preincubating WW2 and 3 with ␤ or ␥ peptides blocked their association to the immobilized substrate. (iii) In agreement with previous studies using other methods (19,20), the PY peptides could bind WW2 and 3 much better than WW1.
Whereas a specific interaction between PY peptides and WW fusion proteins were readily demonstrated, their kinetic analysis turned out to be complicated. Both associations and dissociations were biphasic, and the slow phase did not saturate at the times and concentrations used. The nature of this slow phase is not clear. It may represent a slow aggregation of the fusion protein on the chip, following its initial binding to the peptide or an induced-fit phenomenon. Nevertheless, because the slow phase is also abolished by preincubation with peptides, blocked by the mutationY628A, and not seen for WW1, it is probably physiologically relevant.
A major finding was the fact that the phosphorylated peptides bind the WW fusion proteins more than three times more than do their non-phosphorylated analogues. This result was obtained for both ␤ and ␥ and was independent of the WW domain used. Kinetic analysis of the early phase demonstrated that this elevation reflects an increase in the association rate constant with no significant change in the dissociation rate constant. Thus, the above data are consistent with the model that phosphorylation of ␤Thr-613 and ␥Thr-623 by ERK and maybe other kinases facilitates the ENaC⅐Nedd4 interaction and down-regulates the channel. Mutating these residues inhibits this interaction and increases channel activity in Xenopus oocytes. It should, however, be mentioned that recent studies revealed functional differences between Nedd4-1 used in this study and Nedd4-2, which is phosphorylated by sgk (9,10). Thus, the situation may be a more complex one.
The above mechanism is also in agreement with the suggestion that WW domains act as binding modules for phosphoserines and threonines to help determine specificity in signaling cascades (26). The structure of a complex between WW3 and a ␤ PY peptide in solution has been recently solved by NMR (27). According to this structure, ␤Thr-613 is not directly involved in WW binding and is part of a region that is disordered in solution. The difference between the two sets of data may stem from the different peptide structures used in both cases.
Yet unknown is the upstream signaling pathway that mediates its effect on ENaC through the phosphorylation of ␤613/ ␥623. At least three extra-or intracellular signals known to regulate ENaC may in principle activate ERK. A first possibil-ity is that the mechanism described in this study plays a role in a stress response and in particular in the hypotonic-induced inhibition of channels. Hypotonic shock is a well-established activator of ERK, which is of particular importance in epithelial cells that normally experience extreme osmolarities (28). A hypotonic-induced inhibition of ENaC has been demonstrated (29), and it may be mediated by phosphorylating ␤Thr-613 and/or ␥Thr-623. The second involves transforming growth factor ␤, which was shown to inhibit the aldosterone-induced Na ϩ conductance at a stage distal to channel transcription (30). Because transforming growth factors ␤1 and 2 are well-established activators of ERK (31), desensitization of the response to aldosterone may be mediated by channel phosphorylation. A third option is participation of ERK in the Ca 2ϩ -induced inhibition of Na ϩ transport. It is well-established that epithelial cells down-regulate their apical Na ϩ permeability in response to increased cell Na ϩ and that this "feedback" inhibition is mediated by a rise in cell Ca 2ϩ (1). An increase in cell Ca 2ϩ translocates Nedd4 to the apical membrane and decreases cell surface expression of ENaC (32,33). Ca 2ϩ can activate ERK through either Rap or Ras (34) and, hence, mediate its inhibitory action also by a channel phosphorylation.
Other studies have related components of the Ras pathway to the regulation of ENaC by adrenal corticosteroids (35)(36)(37). These findings, however, predict effects that are opposite to those reported in the current study; i.e. an increase in Na ϩ conductance as a result of ERK activation. It was also reported that ERK antagonizes induction of ENaC by glucocorticoids (38,39). In this case, regulation involves channel transcriptional and not its post-translational modification. Finally, it is possible that enhanced phosphorylation of ENaC is triggered by the inhibition of a specific phosphatase rather than the activation of a kinase. This was found to be the case in the phosphorylation-dependent ubiquitination of the cAMP response element in hypoxia (40).
In conclusion, the current study identifies a mechanism by which ERK down-regulates ENaC by enhancing channel association with Nedd4. The exact role of this process in the physiological regulation of Na ϩ transport awaits further studies.