Triple-barrel organization of ENaC, a cloned epithelial Na+ channel.

A cloned rat epithelial Na+ channel (rENaC) was studied in planar lipid bilayers. Two forms of the channel were examined: channels produced by the α subunit alone and those formed by α, β, and γ subunits. The protein was derived from two sources: either from in vitro translation reaction followed by Sephadex column purification or from heterologous expression in Xenopus oocytes and isolation of plasma membranes. We found that either α-rENaC alone or α- in combination with β- and γ-rENaC, produced highly Na+-selective (PNa/PK = 10), amiloride-sensitive (Kiamil = 170 nM), and mechanosensitive cation channels in planar bilayers. α-rENaC displayed a complicated gating mechanism: there was a nearly constitutively open 13-picosiemens (pS) state and a second 40-pS level that was achieved from the 13-pS level by a 26-pS transition. α-, β-, γ-rENaC showed primarily the 13-pS level. α-rENaC and α,β,γ-rENaC channels studied by patch clamp displayed the same gating pattern, albeit with >2-fold lowered conductance levels, i.e. 6 and 18 pS, respectively. Upon treatment of either channel with the sulfhydryl reducing agent dithiothreitol, both channels fluctuated among three independent 13-pS sublevels. Bathing each channel with a high salt solution (1.5 M NaCl) produced stochastic openings of 19 and 38 pS in magnitude between all three conductance levels. Different combinations of α-, β-, and γ-rENaC in the reconstitution mixture did not produce channels of intermediate conductance levels. These findings suggest that functional ENaC is composed of three identical conducting elements and that their gating is concerted.

It is these ␤ subunit mutations that underlie the autosomal dominant genetic hypertensive disorder, Liddle's disease (10 -14). Moreover, ␣-, ␤-, and ␥-ENaC expression has been localized to aldosterone-responsive, Na ϩ -reabsorbing epithelial tissues in the rat by in situ hybridization and immunocytochemistry using subunit-specific probes (15,16). While amiloride-sensitive Na ϩ channels have been shown also to be responsive to antidiuretic hormone in many epithelia (17,18), the rat distal colon is refractory to the influence of antidiuretic hormone (19). Thus, the ␣, ␤, and ␥ subunits may comprise only a portion of the amiloride-sensitive Na ϩ channels expressed by other cell types.
Amiloride-sensitive Na ϩ channels studied by the patch clamp technique in native epithelia have displayed a wide variety of single channel properties. For example, single channel conductances ranging from 1 to over 50 pS 1 have been reported (17). Patch clamp studies of Xenopus oocyte plasma membranes following coexpression of ␣, ␤, and ␥ cRNA reveal a channel of 5-8 pS in size (2,8), comparable with Na ϩ channels present in renal cortical collecting tubules (18). However, a recent paper by Burch et al. (16) report that the message levels of ␣, ␤, and ␥ subunits in superficial human proximal airway epithelia are not equal as they are in rat colon, kidney, and salivary gland (15) but exist in a relative ratio of ␣ Ͼ ␤ Ͼ Ͼ ␥. Interestingly, the single channel conductance of the amiloridesensitive Na ϩ channels found in this epithelium is 19 -20 pS (19). Thus, the possibility exists that different combinations of these subunits could produce channels with different unitary conductances. Another possibility is that the unitary conductances are tissue-specific, depending upon the physical state of the membrane, which in turn is dependent upon ionic conditions and membrane composition.
The purposes of the present work were 2-fold. First, we wished to develop a reconstitution system in which the single channel properties of ␣-, ␤-, and ␥-rENaC could be studied, unencumbered by problems inherent in heterologous expression systems. Second, we tested the hypothesis that variations in relative levels of ␣, ␤, and ␥ have functional consequences on single channel behavior. Our results indicate that either ␣ alone, or ␣ in combination with one or both of the other subunits of ENaC produce mechanosensitive, amiloride-inhibited, Na ϩ -selective ion channels following incorporation into planar lipid bilayers. ␣-rENaC channels showed a complex kinetic pattern: there were two main conductance transitions, one of 13 and the other of 26 pS. In contrast, ␣,␤,␥-rENaC revealed only the 13-pS state. Different combinations of ␣, ␤, or ␥ did not produce channels of intermediate conductance. Gating of ␣-rENaC was cooperative, with transitions to a 40-pS level only occurring after the 13-pS level was open. Moreover, upon treatment with the reducing agent dithiothreitol (DTT), both ␣-rENac and ␣,␤,␥-rENaC fluctuated among three independent 13-pS sublevels. We conclude that these channels are composed of three identical conduction elements and that the differences in the activity of the channels are modulated by the presence of the ␤ and ␥ subunits within the complex.

In Vitro Translation and Reconstitution of ENaC Subunits-The
ENaC plasmids (pSport) were linearized overnight with NotI. The linearized DNA was purified using GeneClean kit (Bio101, La Jolla, CA), followed by in vitro transcription using T7 RNA polymerase according to the manufacturer's instructions (Ribomax kit, Promega Corp., Madison, WI). A 2:1 molar ratio of cap analog m 7 G(5Ј)ppp(5Ј)G (NEB, Beverly, MA) to GTP was used in the translation reaction as described previously (20).
RNA was in vitro translated using a rabbit nuclease-treated cell-free lysate system (Promega) according to the manufacturer's instructions and as described previously (20). 1.5 units of canine microsomal membranes (Promega) were added to the translation reaction. This resulted in the core glycosylation of the de novo synthesized protein (7). In vitro translated proteins were purified on a G-75 Sephadex (Pharmacia Biotech Inc.) column as described previously (20). Fractions enriched in the appropriate ENaC subunits were reconstituted into phospholipid liposomes as described earlier (20). For liposomes containing ␣,␤,␥-rENaC, subunits were in vitro translated separately, and each was purified over a gel filtration column. Identical elution fractions were collected and assayed for total 35 S incorporation. The relative ratios of ␣, ␤, and ␥ subunits reconstituted into vesicles were determined by these 35 S measurements, assuming similar methionine compositions for each subunit. Control liposomes were also prepared from a mock in vitro translation/ Sephadex column purification run, following an identical protocol. In this case, ENaC RNA was simply omitted from the reaction mixture. These liposomes were used as control material for the bilayer experiments.
Channel Expression in Xenopus Oocytes and Oocyte Membrane Vesicle Preparation-Membrane vesicles from ␣-rENaCand ␣,␤,␥-rENaC mRNA-injected and water-injected oocytes were made essentially as described by Perez et al. (21). Thirty oocytes in each group were rinsed and homogenized in high [K ϩ ]/sucrose medium containing multiple protease inhibitors. Membranes were isolated by discontinuous sucrose gradient centrifugation and resuspended in 300 mM sucrose, 100 mM/ KCl, and 5 mM MOPS at pH 6.8. This material was aliquoted into 50-l fractions and stored at Ϫ80°C until use.
Bilayer Experiments and Data Analysis-Lipid bilayers were cast from a phospholipid solution in n-octane containing a 2:1:2 mixture of diphytanoyl-phosphatidylethanolamine/diphytanoyl-phophatidylserine/ oxidized cholesterol (25 mg/ml). Lipids were purchased from Avanti Polar Lipids (Alabaster, AL). Bilayer capacitances ranged from 300 to 400 picofarads. The solution bathing the bilayers consisted of 100 mM NaCl and 10 mM MOPS-Tris buffer (pH 7.4) unless otherwise noted. All solutions were made with Milli-Q water and were filter-sterilized through 0.22-m Sterivex-GS filters (Millipore, Bedford, MA). The reconstituted vesicles or oocyte membranes were applied with a glass rod to one side (trans) of the preformed bilayer with the membrane voltage held at Ϫ40 mV. Under these conditions, channels oriented (Ͼ90% of the time, n ϭ 250) asymmetrically, with the amiloride-sensitive (extracellular) side facing the trans solution. Voltage was applied to the cis chamber, and the trans chamber was virtual ground.
Single channel recordings were acquired and analyzed using pCLAMP software as described previously (22,23). The recordings were acquired and stored unfiltered. For analysis, they were filtered at 300 Hz with an 8-pole Bessel filter and acquired at 1 ms/point. The 50% threshold-crossing technique was employed to produce events lists. Open and closed dwell time histograms were logarithmically binned and fitted by a sum of exponential functions using maximum likelihood. All data analysis was performed in bilayers containing a single active channel. Application of a hydrostatic pressure difference across a bilayer containing reconstituted ENaC increased open probability to near 1 (20). This effect was independent of the direction of the hydrostatic pressure gradient, i.e. it was equally effective when the bathing solution level was lowered or raised on either side of the channel-containing bilayer. In multichannel membranes, hydrostatic pressure could thus reveal the total number of channels present in the membrane. Therefore, at the beginning of every experiment, 1 ml of the trans bathing solution was removed to determine the total number of Na ϩ channels present. If more than one channel was detected, the membrane was broken and the incorporation procedure repeated.
Patch Clamp Experiments-Cell-attached patch clamp recordings were obtained from defolliculated oocytes previously injected with ␣-rENaC and ␣,␤,␥-rENaC cRNA (12.5 ng for oocytes expressing ␣-rENaC alone, and 1 ng of each subunit for oocytes expressing all three rENaC subunits) or 50 nl of nuclease-free water at room temperature. The composition of the pipette solution (ND-96) was (in mM): 96 NaCl, 2.4 KCl, 1.8 CaCl 2 , 1 MgCl 2 , and 5.0 HEPES (pH 7.4). The bathing solution had K ϩ substituted for Na ϩ to eliminate the resting membrane potential. Blunt-tipped patch micropipettes with a tip resistance of approximately 1-2 M⍀ were fabricated using a Narashigi pp73 twostage micropipette puller. Under these conditions, inward single channel currents (downward current openings) in the cell-attached patches were carried by Na ϩ when the pipette was positive relative to the bath. Single channel currents were examined using an EPC-7 patch clamp amplifier (List Electronics, Darmstadt, Germany). After sealing the pipettes to the cells (seal resistance Ͼ10 gigaohms), the cell-attached patches were clamped to a variety of voltages using an S-95 trilevel stimulator (Medical Systems; Greenville, NY). The amplifier gain was 100 mV/pA. This degree of amplification was sufficient to resolve clearly single channel openings Ͼ0.3 pA. The resultant ionic currents at each membrane potential were filtered at 300 Hz, digitally recorded at 1 kHz, and filed for analysis by pCLAMP software (Axon Instruments, Sunnyvale, CA). All patch clamp experiments were performed at 21 Ϯ 2°C.

RESULTS
In Vitro Translation of rENaC-We have previously demonstrated that in vitro translated ␣-bENaC, the bovine ENaC homolog, migrates as a dimer on SDS-polyacrylamide gel electrophoresis under nonreducing conditions (7). As seen from the autoradiograph shown in Fig. 1 (lane 1), a similar gel pattern was observed for ␣-rENaC in vitro translated in the presence of microsomes and run under denatured but nonreduced conditions. In seven experiments of this type, the presence of a dimer at 172 Ϯ 11 kDa was always observed. However, the relative distribution of ␣-rENaC in the monomeric versus dimeric form differed from one batch of translation reaction to the other. The highest yield of a dimer to a monomer form is shown in the example in Fig. 1. Despite this variability there were never any observed differences in the biophysical properties of the in vitro translated channel incorporated into planar lipid bilayers (see below). Moreover, there was no evidence for the presence of a trimer form of ␣-rENaC (i.e. a product that migrated at ϳ270 kDa was not observed; see lane 1). These observations suggest that the ␣ subunit of ENaC may occur in its native state as a covalently linked dimer. As seen from lane 2, this pattern of interaction was not modified by the addition of the ␤ and ␥ subunits. Lane 3 shows that the 180-kDa band is absent following reduction with 20 mM DTT and that the highest molecular mass band observed was one at 92 kDa. This size protein was expected for monomeric and core-glycosylated ␣-rENaC and was similar to previous observations on ␣-bENaC (7). Co-translation of all three subunits simultaneously did not reveal any forms between 200 and 300 kDa. Lane 4 shows a control, run in the absence of any ENaC message. Thus, the polypeptides seen in lanes 1-3 at molecular masses lower than 92 kDa probably represent partial translation products. The absence of a protein that migrated at a relative molecular mass in the range of the cumulative masses of one ␣, one ␤, and one ␥ subunit suggests that ␣, ␤, and ␥ do not covalently interact in a simple 1:1:1 ratio and that electrostatic interactions may be important in stabilizing the native ENaC complex.
Conductance Properties of rENaC-The column-purified in vitro translated polypeptides of ␣-rENaC, or ␣,␤,␥-rENaC were reconstituted into liposomes and then incorporated into planar lipid bilayer membranes. As shown in Fig. 2, these in vitro translated proteins form functional ion channels (n ϭ 7 for each). This figure depicts channel activity at different applied potentials. There was no effect of voltage on P o of either channel. Fig. 2B shows the associated main state single channel current versus voltage curves for each channel. Both ␣-rENaC and ␣,␤,␥-rENaC channels were linear. In over 1,000 attempted incorporations using control liposomes (see "Materials and Methods"), no channels of any kind were detected.
Examination of the single channel records over several min-utes for both ␣-rENaC and ␣,␤,␥-rENaC indicate continuous activity with no run-down (n ϭ 16; Fig. 3). These channels were kinetically different from ␣-bENaC in bilayers, in that ␣-bENaC channel activity was punctuated by long closed periods lasting up to several minutes (20). Long (Ͼ1-s) closures were never observed for the rENaC channels in over 2 h of recording. ␣-rENaC channels displayed a specific gating pattern; with symmetrical 100 mM NaCl, the largest conductance level was 40 pS, but the channel appeared to fluctuate among 0-, 13-, 26-, and 40-pS levels (see associated all points amplitude histograms, Fig. 3B). These histograms did not fit the binomial algorithm for independent gating, indicating that these conductance levels cannot be produced by three independent channels. Neither the 13-nor 26-pS current transitions were ever observed independently of each other in over 300 separate experiments. In the small number of cases (19/ 330) in which multiple channels were incorporated into the bilayer, the pattern shown at the top of Fig. 3 was simply repeated, i.e. for two channels in the bilayer, 4, 5, and 6 additional states were seen. Moreover, the single channel records indicated that the gating properties of the ␣-rENaC channel were not independent. Openings of the 26-pS level were only observed after the 13-pS state was open, never before. However, albeit infrequently, the 13-pS state would close prior to closure of the 26-pS transition (see Fig. 2A, top). When ␣,␤,␥-rENaC, in a 1:1:1 (w/w/w) combination, was incorporated into bilayers, only a 13-pS conductance level was observed (Figs. 2 and 3, bottom). However, brief openings of Ͻ250 ms in a duration to 40 pS were occasionally seen. Again, these openings occurred on the top of 13-pS conductance level that was, in essence, constitutively open. Event dwell time histograms were constructed for all the conductance levels of ␣-rENaC and ␣,␤,␥-rENaC channels by setting a threshold at 50% of the open level of each substate (Fig. 3C). The dwell histograms in each sublevel for each channel were all fitted by a single exponential function. The closed state time constants were 70 Ϯ 9 and 40 Ϯ 5 ms for ␣-rENaC and ␣,␤,␥-rENaC, exit from the closed state is nearly twice as long for ␣-rENaC than for ␣,␤,␥-rENaC.
Because ␣,␤,␥-rENaC expressed in Xenopus oocytes displays a Na ϩ -selective, 5-pS single channel conductance with relatively long lived open and closed conductance states as measured by patch clamp (2, 8), we wanted to assess directly whether bilayer reconstitution protocol utilizing in vitro translated polypeptides could affect conductance and/or open and closed times. Therefore, we compared rENaC single channel properties determined from patch clamp measurements of rENaC-expressing oocytes with those made in bilayers using in vitro translated proteins or subsequent to fusion of rENaCexpressing oocyte plasma membranes. Oocytes were injected either with ␣-rENaC or ␣,␤,␥-rENaC cRNA and then either they were patch-clamped or their plasma membranes were used for fusion to planar bilayers. Fig. 4 shows the results of A, records are shown for ϩ100 mV holding potential and are representative of nine separate experiments. Record was filtered at 300 Hz using an 8-pole Bessel filter prior to the acquisition and were sampled at 1000 Hz using a Digidata 1200 interface. Bathing solutions contained symmetrical 100 mM NaCl plus 10 mM MOPS-Tris (pH 7.4). Dotted lines indicate zero current. B, all point amplitude histograms. Histograms were generated by pCLAMP software from a record of 5 min in length. C, single channel events dwell time histograms. Time constants were calculated from single exponential fits for each state. This experiment is representative of nine and seven separate trials for ␣-rENaC and ␣,␤,␥-rENaC, respectively. Bin width was 1 ms. these maneuvers. Patch clamp recordings of oocytes expressing ␣-rENaC revealed channels with a large conductance of approximately 18 pS. Interposed among the large transitions were two additional conductance levels of 6 and 12 pS each. In recordings made from three separate oocytes, these ␣-rENaC channels behaved in this manner. Except for the absolute values of the conductance states, this kinetic behavior was very similar to that observed for ␣-rENaC in the bilayer (Fig. 2). However, in contrast to the bilayer, channel activity occurred in bursts rather than continuously. Patch clamp recordings made from oocytes expressing ␣,␤,␥-rENaC typically showed long lived 6-pS channels, similar to what was previously reported (2). Native oocyte membranes from Xenopus oocyte expressing either ␣-rENaC or ␣,␤,␥-rENaC fused to bilayer membranes revealed channels with similar kinetic behavior to those measured by patch clamp, but only with larger conductance states. In all other respects, however, these channels behaved identically to those observed for in vitro translated protein incorporated into bilayers. We conclude, therefore, that the bilayer is an appropriate system in which to study rENaC.
Effects of Hydrostatic Pressure on rENaC Channel Activity-Based upon our experience with ␣-bENaC (20), we next tested the hypothesis that both ␣-rENaC and ␣,␤,␥-rENaC could be activated by an imposition of a hydrostatic pressure gradient across the bilayer membrane. As shown in Fig. 5, channel activities of both ␣-rENaC and ␣,␤,␥-rENaC were significantly enhanced by membrane stretch. A hydrostatic pressure gradient (⌬P) was established by removal of 1 ml of bathing solution from one compartment of the bilayer chamber, which was equivalent to 0.26 mm Hg. For ␣,␤,␥-rENaC, single channel open probability (P O ) increased from 0.15 Ϯ 0.04 to 0.65 Ϯ 0.08 (n ϭ 7), and for ␣-rENaC a ⌬P of 0.26 mm Hg increased P O from 0.60 Ϯ 0.08 to 0.91 Ϯ 0.08 (n ϭ 8). There was no apparent change in single channel conductances following stretch, but the relative frequencies of channel conductance levels were altered. In the case of ␣-rENaC, the 40-pS level predominated after stretch, while for ␣,␤,␥-rENaC, it appeared that three 13-pS conductance levels were present. All point amplitude histogram analysis of these records after stretch revealed that these three conductances could be fit by a binominal distribution, indicating that stretch induced independent gating of the component sublevels.
Ion Selectivity of rENaC in Planar Bilayers-When ␣-rENaC or ␣,␤,␥-rENaC channels were bathed with asymmetric solutions of NaCl (a 10-fold gradient) under nonstretched conditions, a reversal potential of 57 Ϯ 3 mV was measured, and a permeability ratio for Na ϩ versus Cl Ϫ of 10:1 for each channel type was calculated. These values did not change upon stretch (n ϭ 9 for each channel type). These results are similar to those previously reported for ␣-bENaC (20). Likewise, P Na ϩ/P K ϩ for each channel was determined from reversal potential measurements made under biionic conditions in the absence of and in the presence of a 0.26 mm Hg hydrostatic pressure gradient across the membrane (Fig. 6). Under nonstretched conditions, both channels were 10-fold more permeable to Na ϩ than to K ϩ . However, upon stretch, each channel lost some of its ability to discriminate between Na ϩ and K ϩ , decreasing to 3:1 and 4:1 for ␣-rENaC and ␣,␤,␥-rENaC, respectively.
Gating Properties of rENaC-The records depicted in Figs. 2 and 3 indicate that the apparent single channel kinetic behavior of ␣,␤,␥-rENaC was different from that of ␣-rENaC. ␣-rENaC has one small 13-pS conductance level that is almost continuously open and a larger 40-pS level, while ␣,␤,␥-rENaC primarily displays the 13 pS level with only brief openings to the 40-pS level. Based on the observations that both ␣-rENaC and ␣,␤,␥-rENaC display kinetic properties indicative of subconductive type behavior and that ␤ and ␥ themselves do not form channels, it is likely that the conduction pathway(s) of rENaC is formed by the ␣ subunit(s). We next examined whether these channels were comprised of more than one conduction element. The results presented in Fig. 1 indicate that disulfide bond formation occurs between two of the subunits. Thus, we postulated that disrupting disulfide bonds would reveal more fundamental kinetic behavior. Therefore, we tested the effects of 25 M of the reducing agent DTT on ␣-rENaC and ␣,␤,␥-rENaC in bilayers. Fig. 7 summarizes the results of a typical experiment. First, either ENaC was sensitive to DTT only from the trans (i.e. outside) bathing solution. Second, concentrations of DTT below 25 M had no effect on single channel properties, and concentrations larger than 50 M irreversibly damaged coordinated channel activity (data not shown). As can be seen in the figure, treatment of either ␣-rENaC or ␣,␤,␥-rENaC with DTT resulted in the appearance of three indistinguishable 13-pS conductance states. As ␤ and ␥, independently or together, do not induce channel activity, at least as expressed in oocytes (2,7), it is plausible that both ␣-rENaC and ␣,␤,␥-rENaC consist of a minimum of three protochannels formed by ␣ subunits that gate in a concerted fashion. In the presence of DTT, this synchronous gating may be disrupted, thus permitting independent operation of these three conductive elements.
To further test the hypothesis that ENaC consists of three individual protochannels formed by ␣ subunits, we cross-linked all of the subunits present in the functional complex together. The sulfhydryl reactive reagent 5,5-dithiobis(2-nitrobenzoate) (DTNB) was used as the cross-linking reagent. Again, DTNB was only effective from the trans side of the bilayer. Fig. 5 shows that DTNB treatment of either ␣-rENaC or ␣,␤,␥-rENaC produced channels that fluctuated between a 0-and 40-pS level. Thus, the kinetic behavior of the channels indicated that the three putative individual ␣ subunit proto-channels may operate in concert. However, the complete opening of this channel complex when in its native form occurred in two steps, the second twice the size of the first (Fig. 3A). Thus, we hypothesized that one of the ␣ subunit protochannels was anchored to the complex by a noncovalent interaction. To test this idea, we exposed rENaC to elevated salt concentrations in the hope of minimizing electrostatic interactions between subunits (the bulk of the amino acids comprising each ENaC subunit lies in a large extracellular loop (24 -26)). Thus, the prediction was that both ␣-rENaC and ␣,␤,␥-rENaC should gate in a very similar manner, with one of the protochannels behaving as an independent lower conductance channel and the two disulfide-linked protochannels operating in effect as a single higher conductance unit. This experiment has been performed a total of six times each for ␣-rENaC and ␣,␤,␥-rENaC, and the results are summarized in Fig. 8. Raising the NaCl concentration of the bilayer bathing solution from 0.1 to 1.5 M resulted in the appearance of three conductance levels for both ␣-rENaC and ␣,␤,␥-rENaC. The conductance levels of rENaC were increased to 19, 38, and 57 pS by the elevated [Na]. Moreover, the current transitions appeared to fluctuate randomly from the zero conductance level to each of the three higher levels, unlike the kinetic behavior of the channels at 0.1 M salt. In addition, both ␣-rENaC and ␣,␤,␥-rENaC responded identically to elevated salt and to DTNB treatment, suggesting that the core conduction elements of both channels are identical.
To better understand the biophysical consequences of DTT and high salt treatment of rENaC, amplitude histogram analyses of current records made under these experimental conditions were performed. All points and events amplitude histograms are presented in Fig. 9. The overall distribution of the all points amplitude histograms was binomial, thus suggesting equal probability of the channel residing in any conductance level (cf., Fig. 3B). The events amplitude histogram revealed a disruption of concerted gating following DTT or high salt treatment in that transitions to all conductance levels occurred independently. Such an outcome of events amplitude histogram may be due to counting channel openings "from" a fixed zero current level "to" a conductance sublevel. In order to overcome this limitation we have constructed amplitude histograms with a gradually sliding zero level, the level that channel resides in becomes an "apparent zero-current level" for the next transition. This maneuver permits the construction of a histogram of absolute values of amplitudes of transitions. The resulting histograms (Fig. 10) show that the predominant transition in the case of DTT-treated rENaC was 13 pS, while high salt treatment produced, in equal probability, transitions of 19 and 38 pS. Taking into account that the increased conductance of the channels in this latter case was due to elevated [Na ϩ ], these results support the hypothesis that rENaC consists of a minimum of three conductive elements, two of which may be linked by disulfide bonds and the third noncovalently anchored to the covalently linked complex. Histograms for DTT and high salt-treated ␣-rENaC and ␣,␤,␥-rENaC were almost identical (Figs. 9 and 10), indicating that the conductive pore of these channels was formed by ␣-rENaC.
Effect of Amiloride on rENaC-The effect of amiloride on ␣-rENaC and ␣,␤,␥-rENaC is summarized in the dose-response curves presented in Fig. 11. Amiloride inhibited both of these channels with very similar efficacy. The apparent amilorideinhibitory constant (K i amil ) was 170 Ϯ 25 nM (n ϭ 12) for both channels. Treatment of ␣-rENaC or ␣,␤,␥-rENaC with either DTT or DTNB had no significant effect on K i amil . Likewise, bathing either ␣-rENaC or ␣,␤,␥-rENaC with symmetrical 1.5 M NaCl solutions had only a minor effect on K i amil , shifting it to 250 Ϯ 30 nM (n ϭ 3) and 230 Ϯ 20 nM (n ϭ 3) for each channel, respectively, consistent with amiloride acting as a competitive inhibitor of these channels (18,27). The best fit of these amiloride dose-response curves was achieved with a Michaelis-Men- Experimental conditions were the same as described in the legends to Figs. 4 and 5, respectively. Events lists were produced by pCLAMP software using 50% amplitude threshold technique with a minimum event duration. All points amplitude histograms are shown in gray, while the events amplitude histograms are shown in black.
ten formalism using a Hill coefficient of 3. All other channel properties, i.e. ion selectivity, amiloride sensitivity, and substate behavior were preserved in the bilayer.
Effect of Different Combinations of ␣,␤,␥-rENaC Subunits on Channel Activity in Bilayers-Our observations, made both in patch clamp studies of rENaC-expressing oocytes and in planar bilayers, indicate that ␣-rENaC and ␣,␤,␥-rENaC display characteristic single channel properties; namely, ␣-rENaC consisted of a small (13 pS in bilayers, 6 pS in patch) and a large (2 times the conductance of the small) conductance state, while ␣,␤,␥-rENaC primarily exhibited only the small conductance level. Because amiloride-sensitive Na ϩ channels recorded in both native epithelial and cultured cells exhibit a wide range of single channel conductances (17,18), and because different relative levels of ␣, ␤, ␥-rENaC mRNA have been detected within a given tissue (15, 16, 28), we utilized the in vitro translation-bilayer reconstitution system for rENaC subunits to test the hypothesis that varying ratios of rENaC subunit components may modify the resultant single channel kinetic properties and/or conductance. Table I presents the results of these experiments. First, ␤ or ␥ alone or in combination, do not produce channels of any kind, confirming the observations initially made in oocytes (2, 6, 7). Likewise, ␣␤ or ␣␥ yielded channels identical to ␣-rENaC described above. An excess of ␤ and ␥ relative to ␣ in the reconstitution mixture produced channels indistinguishable from those of ␣,␤,␥ in a 1:1:1 ratio. If ␣ exceeded ␤ and ␥ by more than 7-fold, only ␣-type channels were seen. Although these experiments do not address the actual stoichiometry of the subunit composition of functional rENaC, they do, nonetheless, suggest that the variable conductances observed in different cells and tissues cannot be attributed solely to different ratios of ENaC subunits comprising the functional channel complex. DISCUSSION In this work, we report the successful incorporation of ␣-rENaC and ␣,␤,␥-rENaC into planar lipid bilayers. ENaC protein was obtained either from a rabbit reticulocyte lysate in vitro translation system or following expression in Xenopus oocytes and isolation of oocyte plasma membranes. The results obtained using either of these preparations were identical. Our experiments also indicate that the ␣-rENaC subunit alone or in combination with other ␣ subunits acts as the conductive element of the channel complex. However, a high degree of concerted gating occurs between these putative conduction elements and those covalently linked by disulfide bonds. The kinetic behavior of ENaC suggests that a functional channel unit is comprised of a minimum of three conductive elements formed by ␣ subunits. ENaCs are highly Na ϩ -selective, are inhibited with high affinity by the diuretic amiloride, and are mechanosensitive.  level differed (13 pS in the bilayer and 6 pS in the patch). For ␣-rENaC and ␣,␤,␥-rENaC, a 6-pS (patch) or 13-pS (bilayers) level was routinely measured. That these amiloride-sensitive channels expressed in oocytes are ENaCs is supported by the fact that they were never observed in water-injected oocytes. We conclude, therefore, that the microenvironment in which ENaC resides determines in large measure its conductance and mean open and closed times (31)(32)(33). Aside from these changes, the channels displayed comparable amiloride sensitivities, ion selectivities, and gating patterns. The existence of subconductive levels within a single ion channel has been reported for many ion channels including the acetylcholine receptor (34), the glycine, GABA, and glutamate receptors (35)(36)(37), the dihydropyridine-sensitive Ca 2ϩ channel (38), inwardly rectifying K ϩ channels (39,40), the ryanodine receptor cation channel (41), and gramicidin (33). It is not clear why subconductive behavior has not been observed in patch records of ␣,␤,␥-rENaC channels. One possible explanation is that these channels have not yet been analyzed at high time resolution. Another reason may be that upon drawing the oocyte membrane into the tip of a patch electrode, sufficient tension may have already been applied to produce what appear to be three independent, small conductance channels (cf., Figs. 3 and 4; Refs. 2 and 8). The fine details of channel conductances appear to be influenced by the methods of observation.

Comparison of rENaC in Bilayers and by Patch Clamp-
Kinetic Behavior of ENaC in Bilayers-Visual inspection of ␣-rENaC transitions in bilayers reveals that the channel fluctuates either between 0 and 13 pS or between 13 and 40 pS but never (in at least 2 h of recording) between 0 and 26 pS. Residence of the channel in its 26-pS level was rare and short lived and occurred only when the channel transited from its 40-pS level (see current trace and associated amplitude and dwell time histograms, Fig. 2, A, B, and C). Moreover, transition to the highest conductance level (40 pS) occurred in a 26-pS step and only when the channel occupied its 13-pS level. This pattern of channel behavior was different when the bilayer was bathed with high salt-containing solutions (1.5 M, Fig. 6). Under these conditions, the channel flickered between 0 and 13 pS, and 0 and 26 pS stochastically and episodically reached its 40-pS level. Likewise, when ␣-rENaC was treated with DTT, transitions occurred independently between all three equally spaced conductance levels. High salt or DTT treatment of ␣,␤,␥-rENaC produced an identical pattern of channel activity as for ␣-rENaC (Figs. 5 and 6).
Both high salt and DTT disrupt protein-protein interactions. Thus, the change in biophysical properties associated with these treatments implies that a multimeric form of ␣-rENaC underlies channel behavior. Because these three levels represent subconductance states of a single channel entity (23), this kinetic behavior strongly suggests that ENaC is composed of a minimum of three conductive elements and that a pore is formed within each one of these elements. The observations that the same kind of channel activity following DTT treatment is seen for ENaC composed of only ␣ or of ␣,␤,␥ and that ␤ and ␥ cannot form ion channels by themselves (Table I) suggest that the conduction element is the ␣ subunit of ENaC. Whether a monomer or dimer (or higher form) of ␣-ENaC acts as the unit conduction element cannot be deduced from these experiments.
As a first approximation, a simple kinetic model of ENaC can be described as follows: where C represents the closed state and O 1 , O 2 , and O 3 the 13-, 26-, and 40-pS open states, respectively. As indicated above, there were only a few transitions to 26 pS that were observed, and these only occurred from the 40-pS conductance level and had a time constant of 35 Ϯ 8 ms (Fig. 2, B and C). There is certain probability P that the channel will reside in any of the given states, and because at any given time the channel must be in one of them, the sum of these probabilities must equal one.
Also, for a system in equilibrium the percentage of channels in any given state must remain constant. Therefore, the rate of transition out of one state must equal the rate of transition into it. The constants k, m, k 1 , and m 1 are measures of transition rates between these states. Therefore, the net transition rate out of a state is the product of the rate constant and the probability of the channel being in that state.
If the values for each of the rate constants are determined, it will be possible to calculate the values for the probabilities P C , P O 1 , and P O 2 using Equations 3, 4, and 5. The probability of the channel residing in a given state for time t or less is as follows, where T equals one divided by the sum of all of the rate constants leading away from the state (42). Thus, Substituting with the experimental data (Fig. 2C), we can calculate the following: k ϭ 14.5 s Ϫ1 ; k 1 ϩ m ϭ 13.5 s Ϫ1 ; and m 1 ϭ 11.5 s Ϫ1 . This gives a unique solution for the rate constants k and m 1 . However, to complete the model, we need values for k 1 and m. The appropriate equation can be obtained by calculating the probability of the channel proceeding to the 40-pS state from the 13-pS state (i.e. P O 1 3 O 3 ). This has been calculated from 10 min of single channel recording by counting the number of times that the channel switched from the 13-to the 40-pS state and dividing by the total number of times the channel switched out of the 13-pS state. This probability was found to be 0.66 Ϯ 0.05. We know that this probability must equal the rate of transition into the O 3 state divided by the total rate of transition out of the O 1 state. P O13O3 ϭ m/͑k 1 ϩ m͒ (Eq. 11) Thus, we calculate the values of the rate constants k 1 ϭ 4.95 s Ϫ1 and m ϭ 9.0 s Ϫ1 . Using these calculated values, we can now solve equations 3, 4, and 6 to calculate the probability of the channel being in any of the possible three states: P C ϭ 0.14; P O 1 ϭ 0.40; and P O 3 ϭ 0.46. From the all points amplitude histograms (Fig. 3B), we can compute the probability of finding a channel in any given state by calculating the area under each individual curve and dividing by the total area under the histogram. This analysis yields the following results: P C ϭ 0.14 Ϯ 0.05; P O 1 ϭ 0.35 Ϯ 0.05; and P O 3 ϭ 0.51 Ϯ 0.05.
A comparison of the values for the probabilities calculated from the histogram analysis and those derived from the kinetic simulation are in good agreement. This simulation thus formalizes the kinetic behavior of a triple-barrel model for ENaC. This triple-barrel model of ENaC is similar to that proposed for inwardly rectifying K ϩ channels (39,40). Interestingly, while there is little homology between ENaC and IRK1 or ROMK1 at the nucleotide and amino acid levels (43,44), the membrane topology of both classes of ion channel are similar in that they each have only two putative membrane-spanning domains (24 -26).
Are Biophysically Distinct Amiloride-sensitive Na ϩ channels Referable to Different Combinations of ␣, ␤, and ␥ Subunits?-Steroid hormones increase transepithelial Na ϩ transport in target epithelia such as colon, lung, and renal cortical collecting tubule (45). These tissues contain message for all three subunits of ENaC (2,5,6,15,16,46). Steroid hormones have also been found to increase the abundance of ␣-rENaC mRNA relative to ␤ and ␥ in lung (28,30). In rats fed a low Na ϩ diet to elevate circulating aldosterone levels, ␥-rENaC mRNA (in contrast to ␣ and ␤) was found to be raised specifically (29).
Although expression of ␣-rENaC in oocytes produced an amiloride-sensitive current (1), the absolute magnitude of this current was greatly augmented by co-expression with ␤and ␥-rENaC (2). The roles that each subunit plays in channel formation are unknown as is the stoichiometry of ␣,␤,␥ comprising the functional rENaC. Nonetheless, our results indicate that co-reconstituting different relative quantities of ␣,␤,␥-rENaC does not produce amiloride-sensitive Na ϩ channels with altered kinetic signatures. The large diversity in kinetic and conductance properties of these native channels in cells as measured via patch clamp (17,18) is thus likely to result, at least in part, from tissue-specific factors such as auxillary or regulatory proteins. In fact, biochemical purification studies of amiloride-sensitive Na ϩ channels have revealed a different pattern of polypeptide composition of the channel complex, depending upon the source material (47)(48)(49). However, a thorough analysis of variations in ␣, ␤, and ␥ subunit ratios on single channel properties will only be achieved once the functional significance of any channel-associated proteins are elucidated.