Non-coordinate regulation of endogenous epithelial sodium channel (ENaC) subunit expression at the apical membrane of A6 cells in response to various transporting conditions.

In many epithelial tissues in the body (e.g. kidney distal nephron, colon, airways) the rate of Na(+) reabsorption is governed by the activity of the epithelial Na(+) channel (ENaC). ENaC activity in turn is regulated by a number of factors including hormones, physiological conditions, and other ion channels. To begin to understand the mechanisms by which ENaC is regulated, we have examined the trafficking and turnover of ENaC subunits in A6 cells, a polarized, hormonally responsive Xenopus kidney cell line. As previously observed by others, the half-life of newly synthesized ENaC subunits was universally short ( approximately 2 h). However, the half-lives of alpha- and gamma-ENaC subunits that reached the apical cell surface were considerably longer (t(12) > 24 h), whereas intriguingly, the half-life of cell surface beta-ENaC was only approximately 6 h. We then examined the effects of various modulators of sodium transport on cell surface levels of individual ENaC subunits. Up-regulation of ENaC-mediated sodium conductance by overnight treatment with aldosterone or by short term incubation with vasopressin dramatically increased cell surface levels of beta-ENaC without affecting alpha- or gamma-ENaC levels. Conversely, treatment with brefeldin A selectively decreased the amount of beta-ENaC at the apical membrane. Short term treatment with aldosterone or insulin had no effect on cell surface amounts of any subunits. Subcellular fractionation revealed a selective loss of beta-ENaC from early endosomal pools in response to vasopressin. Our data suggest the possibility that trafficking and turnover of individual ENaC subunits at the apical membrane of A6 cells is non-coordinately regulated. The selective trafficking of beta-ENaC may provide a mechanism for regulating sodium conductance in response to physiological stimuli.

The kidney regulates extracellular fluid volume in the body through modulation of Na ϩ reabsorption along the nephron. The ultimate regulation of Na ϩ reabsorption occurs at the apical surface of collecting ducts and is mediated by the epi-thelial sodium channel (ENaC). 1 ENaC is also located at the apical membrane of other epithelial tissues throughout the body, including colon, sweat glands, and airway (1). Abnormalities of function of this channel, linked to inherited alterations in channel structure, have been shown to be important in several human diseases, including the hypertension seen in patients with Liddle's Syndrome (2) and the salt-wasting seen with some variants of psuedohypoaldosteronism (3). In addition, apparent overactivity of this channel, seen in the presence of common mutations of the cystic fibrosis transmembrane conductance regulator, has been linked to the pathogenesis of airway disease in cystic fibrosis (4). However, little is known about how ENaC activity is regulated under different physiological states.
The primary structure of ENaC was elucidated through expression cloning (5) and revealed that the channel is formed by three homologous subunits, ␣, ␤, and ␥. When these three subunits are expressed together in Xenopus oocytes, they produce a channel with the typical biophysical and pharmacologic properties of the native channel: low conductance, high Na ϩ /K ϩ selectivity, and sensitivity to amiloride in the submicromolar range (1,5,6). However, the open probability of this channel in oocytes is much higher than in most tissues, suggesting that this expression system lacks physiologically important regulation machinery for ENaC (7). Moreover, the expression, activity, stoichiometry, and trafficking of ENaC in various tissues appears to be differentially regulated and complex. Thus, a complete understanding of ENaC trafficking and regulation requires examination of these processes in cells from tissues that express the channel endogenously.
We have investigated the trafficking and turnover of individual ENaC subunits in the hormonally responsive Xenopus kidney cell line A6. Interestingly, we find that Xenopus ENaC subunits (xENaCs) at the apical cell surface have different half-lives and that xENaC subunits traffic to the cell surface in a non-coordinate fashion in response to a subset of transportmodulating agents. In particular, our data suggest a model in which cell surface xENaC activity can be regulated by the selective insertion or removal of ␤-xENaC.

MATERIALS AND METHODS
Cell Lines-A6 cells were maintained in amphibian medium containing 10% fetal bovine serum (BioWhittaker, Walkersville, MD) at 28°C in 5% CO 2 . For experiments, cells were seeded at high density on permeable supports (0.4-m pore, Costar, Cambridge, MA or Millipore, Bedford, MA) and used at least 8 days post-plating. HeLa cells were maintained in Dulbecco's modified Eagle's medium supplemented with * This work was supported by National Institutes of Health Grants R01DK47874 (to J. P. J.) and R01DK54407 (to O. A. W.) and by a grant from the Cystic Fibrosis Foundation (to O. A. W.). The Laboratory of Epithelial Cell Biology is supported in part by Dialysis Clinic Inc. 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  10% fetal bovine serum (Atlanta Biologicals, Norcross, GA), streptomycin (100 g/ml), and penicillin (100 units/ml).
Antibodies and Reagents-Generation and affinity purification of anti ␤and ␥-xENaC antibodies is described in Rokaw et al. (8). Affinitypurified anti ␣-xENaC antibody, generated against a peptide encoding residues 107-125 of ␣-xENaC, was a generous gift of Dr. Thomas Kleyman (University of Pennsylvania and Veterans Affairs Medical Center, Philadelphia, PA) and is described and characterized in Zuckerman et al. (9). An anti ␣-xENaC antibody, generated against this region of the protein by our laboratory, that recognizes proteins of the same molecular weights was also used in some experiments. Insulin (humulin R) was purchased from Eli Lilly (Indianapolis, IN), and aldosterone, vasopressin (ADH), brefeldin A (BFA), and amiloride were from Sigma.
Expression of xENaC Subunits in HeLa Cells-Expression of xENaC subunits in HeLa cells was performed essentially as described in Weisz et al. (10). HeLa cells were plated on 35-mm dishes at approximately 40% confluence. The following day, cells were infected at 37°C with recombinant vaccinia virus encoding T7 RNA polymerase (vTF7.3) at a multiplicity of infection of 10 -20. After adsorption for 30 min, the inoculum was replaced with 0.75 ml of serum-free Dulbecco's modified Eagle's medium containing 3 g of total DNA and 5 l of Lipo-fectAMINE (Life Technologies, Inc.). At 3 h post-infection, cells were starved for 15 min in cysteine-free, methionine-free Dulbecco's modified Eagle's medium, then labeled for 2 h with 100 Ci/ml [ 35 S]methionine (in vitro labeling mix; PerkinElmer Life Sciences). After rinsing with phosphate-buffered saline, cells were lysed in detergent solution (50 mM Tris, pH 8.0, 1% Nonidet P-40, 0.4% deoxycholate, 62.5 mM EDTA, and 1 g/ml aprotinin). After brief centrifugation (1 min at maximal speed in a microcentrifuge) to remove nuclei, SDS was added to a final concentration of 0.2%. Lysates were immunoprecipitated using the appropriate anti-xENaC antibodies, eluted, divided into two aliquots, and mock-treated or treated with N-glycanase (New England BioLabs, Beverly, MA). Samples were immunoprecipitated using the appropriate anti-xENaC antibodies and electrophoresed on 10% SDS-PAGE gels. Radiolabeled bands were visualized using phosphorimaging with Quantity One software (Personal Molecular Imager FX; Bio-Rad).
Modulation of Amiloride-sensitive Short Circuit Current-A6 cells were mock-treated or treated with aldosterone (1 M for 18 h or 3 h), ADH (100 milliunits/ml for 30 min), insulin (100 milliunits/ml for 30 min), BFA (5 g/ml for 3 h) or amiloride (100 M for 18 h). Transepithelial potential difference and short circuit current (I sc ) were measured using a sterile in-hood short-circuiting device as described previously (11).
Half-life of ENaC Subunits-A6 cells were starved for 30 min, then radiolabeled for 1 h and chased for 0 -4 h. Individual filters were solubilized in detergent solution, and ENaC subunits were immunoprecipitated as described above and analyzed on SDS-PAGE gels. To determine the half-life of ENaC subunits that had reached the apical membrane, A6 cells grown on 6-well filter inserts were chilled and biotinylated on ice with sulfo-NHS-SS-biotin (Pierce). Cells were subsequently returned to 28°C for various chase periods. At each time point, cells were solubilized, biotinylated subunits were recovered by binding to streptavidin-agarose, and samples were washed and electrophoresed on SDS-PAGE gels. Two filter inserts were combined for each time point. After transfer to nitrocellulose, gels were probed with anti-xENaC antibodies, and the half-life of each subunit tagged at the cell surface was quantitated by measuring the decrease in biotinylated subunit recovered with time. For ␣-xENaC quantitation, the intensities of the 150-and 180-kDa bands were summed; for ␤-xENaC quantitation, both the 97-kDa band and the minor higher M r species (when observed) were summed; and for ␥-xENaC quantitation, the intensity of the 95-kDa band was measured (see Table I).
Subcellular Fractionation of Endocytic Compartments in A6 Cells-A6 cells grown on filter inserts were treated with hormone or diluent for the specified period of time. The cells were washed twice in Dulbecco's modified Eagle's medium, scraped into phosphate-buffered saline, and pelleted at 15,000 ϫ g in a microcentrifuge. Cells were 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. The nuclei were pelleted at 600 ϫ g, and 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 anti-xENaC antibodies. Quantitation of internalized horseradish peroxidase in gradient fractions was performed using the assay described in Steinman et al. (12).

RESULTS
Characterization of anti-xENaC Antibodies-We previously generated antibodies against ␤and ␥-xENaC and confirmed that these antibodies recognize proteins of the appropriate molecular mass in A6 cells by immunoprecipitation and Western blotting (8). In addition, we recently acquired an antibody against ␣-xENaC from Dr. Thomas Kleyman. By Western blotting, this antibody recognizes bands of 70, 150, and 180 kDa in A6 cells (9). The 180-kDa band likely represents a homodimer of mature ␣-xENaC and is preferentially recovered from apical membrane preparations of A6 cells (9). The identity of the 150-kDa band is unknown but appears to represent a form of ␣-xENaC that is primarily intracellular, as it is frequently in lower abundance when we blot ENaC biotinylated at the apical membrane of A6 cells. All three bands are competed away by preincubating the antibody with immunizing peptide (9). Another group has also observed migration of ␣-ENaC at 150 and 180 kDa (13), and an antibody generated in this laboratory against the carboxyl-terminal sequence of ␣-xENaC recognizes the 180-kDa protein (8,9). To verify that our antibodies react specifically with xENaC subunits, we expressed ␣-, ␤-, and ␥-xENaC from cDNA in HeLa cells using recombinant vaccinia virus and immunoprecipitated radiolabeled cells using anti ␣-, ␤-, or ␥-xENaC antibodies (Fig. 1A). None of the antibodies recognized proteins from mock-infected HeLa cells. Anti-␣-xENaC immunoprecipitated a protein of approximately 65 and 75 kDa from HeLa cells transfected with ␣-xENaC (Table 1). Anti ␤-xENaC recognized two bands of molecular masses ϳ70 and 97 kDa from cells transfected with ␤-xENaC, and anti ␥-xENaC recognized 2 bands of molecular masses 75 and 95 kDa from cells transfected with ␥-xENaC. In the case of ␣and ␥-xENaC, treatment with N-glycanase resulted in collapse of the more slowly migrating band into the band with faster mobility. Treatment of ␤-xENaC immunoprecipitates with N- glycanase caused a shift in the migration of the higher molecular mass band; however, this band did not migrate with the lower molecular mass band, suggesting a further post-translational modification of this subunit. The nonglycosylated subunits that we observed in this overexpression system have previously been observed in other heterologous expression systems, and are not membrane associated nor do they represent degradation intermediates (14). It is likely that these forms are an aberration induced by overexpression. To confirm that endogenous ENaC expressed in A6 cells was glycosylated, we biotinylated the apical membrane of polarized cells, recovered surface proteins using streptavidin-agarose, and treated the samples with N-glycanase. ENaC subunits were then detected by Western blotting (Fig. 1b). The mobilities of both ␤and ␥-xENaC shifted to 70 and 75 kDa as expected. However, neither N-glycanase nor endoglycosidase H treatment affected the mobility of the ␣-xENaC bands. We suspect that these high molecular mass forms of ␣-xENaC may be aggregated or modified in such a way that they are resistant to deglycosylation. However, another group has reported that ENaC subunits at the cell surface of CHO cells appear to be deglycosylated (75).
Half-lives and Steady State Distribution of xENaC Subunits in A6 Cells-We used these antibodies to measure the half-life of newly synthesized xENaC subunits as well as xENaC subunits that had reached the apical plasma membrane of polarized A6 cells. Consistent with a previously published study (15) we found that the half-life of newly synthesized ENaC subunits is relatively short in A6 cells, with a half-life of approximately 2 h for each subunit (Fig. 2). In addition, we measured the relative amounts of individual ENaC subunits at the plasma membrane at steady state. Biotinylated proteins were precipitated from 200 g of cell lysate using streptavidin-agarose, electrophoresed, and Western-blotted using antibodies against the individual xENaC subunits (Fig. 3, right lane). The signal was compared with the amount of each ENaC subunit in 40 g of cell lysate (Fig. 3, left lane). Based on this and other similar experiments, we estimate that that roughly 20% of each subunit is at the cell surface under steady state conditions. Importantly, this steady state distribution of ENaC subunits is considerably higher than that reported in oocytes (14), where less than 1% of the total subunits were detected at the plasma membrane. This observation suggests that there could be significant differences in the regulation of ENaC traffic between this overexpression system and cells that endogenously express the channel.  A6 cells were starved, radiolabeled for 1 h, then chased for the indicated times. Cells were solubilized, and individual ENaC subunits were immunoprecipitated as described under "Materials and Methods." Halflives were quantitated from SDS-PAGE gels using phosphorimaging. Quantitation of the gels shown in the inset is shown. Similar results were obtained in at least three experiments for each subunit.

FIG. 3. Steady state distribution of xENaC subunits in A6 cells.
Polarized A6 cells were biotinylated using a membrane-impermeant reagent. Cells were solubilized, and 40 g of the lysate was loaded in the left-hand lanes. 200 g of protein was incubated with streptavidinagarose to recover cell surface xENaC subunits (right lanes). After electrophoresis and transfer to nitrocellulose, the samples were blotted to detect ␣-,␤-, or ␥-xENaC. Approximately 20% of the total xENaC in the cells could be biotinylated.
Because a large fraction of newly synthesized ENaC is rapidly degraded and does not reach the cell surface, the short half-life we measured for newly synthesized subunits does not accurately reflect the half-life of ENaC that has reached the cell surface. Therefore, we biotinylated the apical membrane of polarized A6 cells and measured the rate of degradation of individual ENaC subunits over time (Fig. 4). Interestingly, all of the ENaC subunits at the plasma membrane had considerably longer half-lives compared with newly synthesized subunits. Surprisingly, however, we consistently observed that the rate of degradation of cell surface ␤-xENaC was considerably faster (ϳ6-h half-life) than that of ␣or ␥-xENaC, which remained relatively stable over the 24-h time course. The long half-life of ␣-xENaC that has reached the cell surface has also been independently verified in Dr. Thomas Kleyman's laboratory. 2 The surprising observation that ENaC subunits have different rates of degradation upon reaching the cell surface suggests the possibility that individual subunits might traffic to or from the plasma membrane independently (non-coordinately).
Effect of Altered Transporting Conditions on ENaC Cell Surface Subunit Levels and Membrane Trafficking-ENaC activity has been reported to be up-regulated by various hormone treatments including aldosterone, ADH, and insulin. Stimulation of ENaC channel activity by aldosterone occurs in two phases: an initial, acute phase (within 1-3 h of stimulation), thought to involve activation of silent or inactive channels resident in the apical membrane, and a late phase, (upon overnight stimulation) whose mechanism is unknown. The mechanism of action of these agents in kidney remains unclear but does not appear to involve synthesis of new channel subunits. By contrast, hyperpolarization with the sodium channel inhibitor amiloride 2 T. Kleyman, personal communication.
FIG. 5. Altered transport conditions modulate the surface density of ␤but not ␣or ␥-xENaC. A6 cells were mock-treated or treated with aldosterone (Aldo, 18 h or 3 h), ADH, insulin, or BFA as described under "Materials and Methods." Cells were then rapidly chilled to 4°C, the apical membranes were biotinylated, and the cells were solubilized. Biotinylated proteins were collected using streptavidin-agarose, samples were electrophoresed and transferred to nitrocellulose, and blots were sequentially probed using anti-xENaC subunit antibodies; representative blots for each condition are shown. The blots on the left were sequentially probed with anti-␤and ␥-xENaC; a blot from a separate experiment was probed with anti-␣-xENaC. The blot on the right was probed sequentially with all three anti-xENaC antibodies. Quantitation of six to nine paired experiments (mock-treated versus experimental) for each condition (mean Ϯ S.E.) is shown in the lower panel. Statistical significance was calculated from the raw data by paired t test.  4. Half-lives of cell surface ␣-, ␤-, and ␥-xENaC in A6. A6 cells were rapidly chilled to 4°C, and the apical surfaces were biotinylated as described in under "Materials and Methods." The cells were then warmed to 28°C for various periods and solubilized, and biotinylated proteins were collected using streptavidin-agarose. Samples were electrophoresed, transferred to nitrocellulose, and probed using anti-xENaC antibodies. A representative blot for each subunit is shown. Quantitation (mean Ϯ S.E.) of four experiments for ␤and ␥-xENaC and three experiments for ␣-xENaC is plotted. Statistical significance was calculated by paired t test.
appears to decrease ENaC activity as an adaptive response (16). We therefore tested the effects of these treatments on ENaC activity in A6 cells. As expected, treatment with aldosterone (both short (3 h) and long (18 h) term), ADH (30 min), and insulin (30 min) increased short circuit current in A6 cells (Table II). By contrast, treatment for 3 h with the fungal metabolite BFA, which inhibits delivery of newly synthesized proteins to the plasma membrane, decreased amiloride-sensitive current (Table II). Next, we tested whether the same treatments affected the surface density of xENaC subunits in the apical membrane. To do this, we incubated A6 cells with aldosterone (3 or 18 h), ADH (30 min), insulin (30 min), or BFA (3 h), then biotinylated the apical domains of the cells. The cells were then solubilized, and biotinylated proteins were collected with streptavidin-agarose, electrophoresed, transferred to nitrocellulose, and blotted with anti-xENaC antibodies (Fig. 5). The level of ␣-xENaC recovered after these treatments was generally unaffected, although a slight (ϳ10%) but statistically significant decrease in apical ␣-xENaC was observed upon treatment with ADH. By contrast, the amount of ␤-xENaC at the apical surface was dramatically elevated in cells treated for 18 h with aldosterone (ϳ60% increase; p Ͻ 0.01) or for 30 min with ADH (ϳ20% increase; p Ͻ 0.01). Treatment with insulin or short term incubation with aldosterone had no effect on the density of ␤-xENaC at the apical membrane, whereas incubation with BFA reduced apical ␤-xENaC levels. These striking results were reproducibly observed in multiple experiments, including several experiments in which the same blots were stripped and sequentially probed with individual anti-xENaC antibodies in random order.
The altered cell surface distribution of ␤-xENaC in response to short term ADH might be due to increased delivery of newly synthesized or recycling proteins, decreased endocytosis or degradation, or to a combination of these effects. Internalization of ENaC expressed in Xenopus oocytes (measured as a decrease in current density) has been shown to occur via a dynamin-dependent mechanism, suggesting that ENaC may traffic to endosomes (17). Changes in ENaC internalization and recycling would be expected to result in reduced levels of ␤-xENaC in endosomal compartments. To examine this possibility, we isolated endosomal populations from non-stimulated or ADHstimulated A6 cells and measured the levels of xENaC subunits recovered in these fractions. For this purpose, we used a flotation gradient developed to isolate endosomal fractions from other polarized epithelial cell lines. To test this method, we incubated A6 cells with the fluid phase marker horseradish peroxidase for 10 min, then homogenized the cells and isolated endosomal fractions on a discontinuous sucrose gradient as described under "Materials and Methods" (Fig. 6A). Late endosomes are concentrated at the 8.5%, 25% sucrose interface, whereas early endosomes are found at the 25%, 35% sucrose interface, very similar to the pattern observed using mammalian cells (18). Upon fractionation of untreated A6 cells, we found ENaC subunits in early endosomes but not at the interface containing late endosomes (not shown), consistent with the presence of a recycling population of channel subunits. In response to ADH, where apical membrane ␤-xENaC increased, endosomal levels of ␤-xENaC decreased, suggesting either an increase in exocytosis or a decrease in recycling in response to ADH (Fig. 6, B and C). By contrast, the levels of ␣and ␥-xE-NaC typically increased in these endosomal fractions, although the changes were not statistically significant compared with mock-treated cells (Fig. 6, B and C). This difference in the response of ␤-xENaC to ADH was detected in multiple experiments and was observed when the same blot was sequentially probed with all three anti-xENaC antibodies (Fig. 6B). Thus, our data suggest that acute stimulation with ADH results in a dramatic redistribution of ␤-xENaC from an endosomal pool to the apical surface of A6 cells. Altogether, our data suggest that selective delivery and/or retrieval of individual subunits can be regulated in a non-coordinate manner and that a selective increase in the level of cell surface ␤-xENaC compared with the other ENaC subunits could account for the increased sodium conductance observed in hormonally stimulated cells. DISCUSSION We have investigated the trafficking and turnover of endogenous ENaC in the hormonally responsive, well characterized kidney cell line A6. As previously observed in these and other cells (14,15,19), we found that the majority of newly synthesized ENaC is rapidly degraded shortly after synthesis. However, by contrast to results reported in Xenopus oocytes, in which Ͻ1% of the total ENaC was present at the plasma membrane at steady state (14), a significant amount of total cellular ENaC subunits are present at the plasma membrane of A6 cells at steady state. Moreover, we found that the half-life of ENaC that reaches the plasma membrane is considerably longer than that of newly synthesized subunits and, interestingly, differs among subunits, with ␤-xENaC having the shortest half-life. Up-regulation of ENaC activity by long term treatment with aldosterone or acute treatment with ADH, both predicted by electrophysiological studies to increase cell surface channel number (20 -23), caused a dramatic and selective elevation in the level of ␤-xENaC recovered at the apical plasma membrane. Brief treatment with aldosterone or insu-FIG. 6. Redistribution of ␤-xENaC from endosomes in response to ADH. Panel A, Isolation of early and late endosomes from A6 cells by subcellular fractionation. A6 cells were allowed to internalize horseradish peroxidase (HRP) for 10 min at 28°C, then homogenized and fractionated as described under "Materials and Methods." Individual fractions were collected and assayed for horseradish peroxidase activity. The migration of early and late endosomes in these gradients is noted. Panels B and C, A6 cells were mock-treated or treated with ADH for 30 min, then fractionated as above. Early endosomal fractions were blotted with anti xENaC antibodies. A single blot that was probed sequentially with anti-xENaC antibodies is shown in panel B, and quantitation of three independent experiments (mean Ϯ S.E.) is shown in panel C. Statistical significance was calculated by paired analysis of the raw data; * p Յ 0.05. lin, where evidence for an increase in surface channel number is more conjectural (22,24,25), had no effect on the surface levels of any ENaC subunits. Incubation for 3 h with brefeldin A, which prevents delivery of newly synthesized proteins to the plasma membrane, resulted in the selective reduction of ␤-xENaC at the cell surface as well as a decrease in I sc . Our data are consistent with a model in which changes in ENaC activity in response to stimulation by some hormones or by alterations in sodium influx in A6 cells is caused by selective ␤-xENaC insertion into or retrieval from the apical membrane. We suggest the intriguing possibility that assembly and disassembly of ENaC channels can occur at multiple intracellular locations and that delivery and/or retrieval of individual subunits can be regulated in a non-coordinate fashion.
Several lines of evidence suggest that apical membrane surface expression and stoichiometry of ENaC channels in polarized tissues and cells is complex. First, various stoichiometries of the subunits have been described in terms of the ultimate channel expressed in membranes. Two groups have determined, using several approaches, that the stoichiometry is 2 ␣, 1 ␤, and 1 ␥ subunit (27,28). By contrast, Welsh and co-workers (29 -31) describe a channel made up of 9 subunits, trimers of the three individual subunits. Interestingly, this group has also reported that individual subunits of human ENaC, when expressed independently, can oligomerize into homomultimers that efficiently traffic to the cell surface (32,33). It is known that the ␣ subunit alone is capable of forming a channel (34) and that dimer channels of either ␣␤ or ␣␥ can be expressed that demonstrate modest differences in ion selectivity, amiloride sensitivity, and open probability in relation to trimer channels (35). However, these heterodimeric channels generate only a fraction (10 -15%) of the current seen when all three subunits are expressed (35); this suggests that remodeling of surface channel subunit composition could serve as a rapid and efficient mechanism to dramatically alter Na ϩ absorption at the apical membrane of cells. Generally such channels have not been observed in nature, possibly because of their low level of activity. Nonetheless, the recent demonstration that, unlike ␣-ENaC-deficient mice, ␤and ␥-ENaC deficient mice do not die due to failure to clear lung liquid at birth suggests that ␣␤ and ␣␥ dimers may have sufficient channel activity for pulmonary clearance (36 -40). These observations suggest that there may be differing physiologic responses to varying expression of individual subunits of ENaC. Moreover, an additional subunit (␦-ENaC) has been identified and may contribute to the diversity of channel stoichiometries in vivo (41).
Second, the considerable variability in expression and regulation of ENaC subunit mRNA and protein across epithelial tissues that express the channel may also be taken to reflect a plasticity of channel composition in the apical membrane. In rat kidney, message for all three subunits is expressed constitutively and changes little with salt restriction (42). Other investigators described modest changes primarily in ␣ subunit mRNA in rat kidney in response to steroid stimulation or adrenalectomy (42,43). A recent study demonstrated significant increases in total ␣-rENaC protein in response to elevated aldosterone levels in rat kidney (44). However, other studies in the same tissue found no significant effects of steroid treatment on protein expression (42). By contrast, colon seems to express mRNA for ␣-xENaC constitutively, whereas ␤and ␥-xENaC message are selectively induced by mineralocorticoids (42,43,45). In lung tissue, mRNA for the three subunits is primarily regulated by glucocorticoids (42), but as in colon, non-coordinate expression of subunits has been described (5,46,47). These observations suggest that new heterotrimeric channels could be created in response to alteration in message levels of only one or two of the subunits in vivo.
A third line of evidence suggesting channel diversity is the variability of single channel properties of ENaC expressed in cells and tissues (reviewed in Ref. 1). Single channel open probability of ENaC in A6 cells varied widely (16) (48), and within patches of channels in cortical collecting tubules, the open probabilities resolved into two distinct populations of channels with either high or low values (49,50), both different from the rather high open probability seen with trimer channels expressed in oocytes (5). The data on variability of message expression and microscopic properties of ENaC across tissues has led Garty and Palmer (1) and McNicholas and Canessa (35) to speculate that there may be varying modular arrangements of channels expressed in these membranes.
Finally, several other instances have been described in which individual subunit combinations can associate to generate distinct heteromultimeric channels with unique functional properties. Excellent examples of this are the P2X receptor subunits, which have similar topology to ENaC subunits and that coassemble to form channels with widely distinct properties (51)(52)(53)(54). Other heteromultimeric channels, such as ligandgated receptors, cAMP-gated channels, ␥-aminobutyric acid receptors, and some K ϩ channels, can also combine to generate different channels with unique functional properties (55)(56)(57)(58)(59)(60)(61). In addition, in the case of gap junction assembly, it is clear that heteromeric channels composed of different connexin subunits can form and that at least some connexin subunits can assemble into ion channels after exiting the endoplasmic reticulum (62,63). Thus, although our suggestion that ENaC subunits can recombine into multiple heteromeric channels possibly at or near the apical membrane runs contrary to the generally accepted notion that all heteromeric channels combine with fixed stiochiometry in the endophasmic reticulum, some evidence for this paradigm already exists.
Our data also suggest the possibility that ENaC channels of differing stoichiometry may normally exist at the cell surface. This hypothesis in no way alters the likelihood that a channel containing all three subunits is the entity mainly responsible for Na ϩ reabsorption. Most of these partial channels would be physiologically unimportant in most circumstances due to their low level of activity. However, if channels can assemble as some site beyond the endophasmic reticulum, then insertion into or retrieval from the apical membrane of individual ENaC subunits could contribute to the rapid modulation of surface ENaC activity in response to extracellular Na ϩ changes or hormonal stimuli and could represent one mechanism of activation of "electrically silent" channels resident in the apical membrane. The strongest evidence against this hypothesis comes from the studies of Firsov et al. (64), which demonstrate that variability in expression of one subunit in oocytes does not change the apparent heterotrimeric structure at the oocyte membrane. However, ENaC assembly may be differentially regulated in this overexpression system compared with its assembly in polarized epithelial tissues that endogenously express the channel. Even in endogenously expressing epithelia, normal trafficking of ENaC depends on establishment of the complete epithelial phenotype. For example, there is a marked difference between ENaC activity in A6 cells grown on glass as opposed to permeable supports that allow exposure to media on both sides of the cell (65).
It is interesting to compare our direct measures of apical membrane ENaC subunits with the predictions made by other techniques concerning channel density under varying conditions of transport. We found that both ADH and long term aldosterone increased apical membrane ␤-xENaC. In both of these conditions, estimates of channel activity by electrophysi-ologic techniques or biochemical methods suggest an increase in the number of active apical membrane holochannels (20,21,67,68,73). This suggests that ␤-xENaC delivered from intracellular compartments may assemble with ␣␥ heterodimers already present at the cell surface. The observation that mice deficient in ␤or ␥-ENaC can clear lung liquid at birth suggests that ␣␤ and ␣␥ dimers may form in vivo as well (36, 38 -40). In contrast to the ADH and long term aldosterone effect, both short term aldosterone and insulin resulted in no detectable change in apical membrane xENaC subunits. This is consistent with patch-clamp estimates of a primary effect on channel kinetics rather than number under these conditions (1,24,72) as well as with biochemical estimates of channel number with short term aldosterone. On the other hand, studies using noise analysis suggest that both insulin and short term aldosterone markedly increase channel density (22,23). It is possible that this represents a qualitative rather than a quantitative difference, since channels with very low open probability that increased following hormonal stimulation could appear as new channels using this approach (49,50). With regard to the BFA experiments, our finding of a selective decrease in apical membrane ␤-ENaC subunit levels is consistent with the observation that this subunit has the shortest half-life in untreated cells; this would be expected to appear as a decrease in apical channel density as measured by noise analysis (74).
An alternative explanation for our data is that the increase in cell surface ␤-xENaC does not directly contribute to aldosterone-or ADH-stimulated current. It is possible that the biotinylation efficiency of ENaC subunits in fully active versus partially assembled channels could be dramatically different, and this could lead to incorrect estimates of turnover rates for each subunit. We feel that this is unlikely because we can efficiently biotinylate a large fraction of each ENaC subunit in A6 cells (approximately 20% of the total at steady state, Fig. 2) and because there are numerous lysine residues on the extracellular domain of all three subunits (at least 12 per subunit). Moreover, the changes we measured in cell surface ␤-xENaC upon physiologic manipulation mirrored the changes in I sc in all cases. In addition, the selective decrease in cell surface ␤-xENaC in BFA-treated cells is consistent with the comparatively short cell surface half-life that we measured for this subunit in untreated cells. Finally, although noise analysis predicts a 2-3-fold increase in channel number in response to ADH and long term aldosterone treatment, it is unlikely that we would have missed a change of this magnitude in all subunits. Importantly, insertion of even a few ␤-ENaC subunits into the membrane could cause a significantly greater proportional increase in channel activity because the activity of ␣␥ heterodimeric channels is only 15-20% that of the holochannel (35). Thus we feel that the most parsimonious explanation for our results is that ENaC channel stoichiometry is dynamic.
In summary, our data suggest that in contrast to newly synthesized subunits, ENaC subunits reaching the plasma membrane of A6 cells are normally long-lived. Furthermore, ENaC subunits appear to be differentially inserted into or retrieved from the cell surface both under steady state conditions and in response to physiologic modulation. Our data are consistent with a model in which changes in ENaC activity in response to stimulation by some hormones or by alterations in sodium influx is caused by selective ␤-xENaC insertion into or retrieval from the apical membrane. However, more studies are clearly necessary to understand the mechanism behind this selective trafficking, as well as its contribution to regulation of ENaC in different species and tissues.