Epithelial sodium channel pore region. structure and role in gating.

Epithelial sodium channels (ENaC) have a crucial role in the regulation of extracellular fluid volume and blood pressure. To study the structure of the pore region of ENaC, the susceptibility of introduced cysteine residues to sulfhydryl-reactive methanethiosulfonate derivatives ((2-aminoethyl)methanethiosulfonate hydrobromide (MTSEA) and [(2-(trimethylammonium)ethyl]methanethiosulfonate bromide (MTSET)) and to Cd(2+) was determined. Selected mutants within the amino-terminal portion (alphaVal(569)-alphaTrp(582)) of the pore region responded to MTSEA, MTSET, or Cd(2+) with stimulation or inhibition of whole cell Na(+) current. The reactive residues were not contiguous but were separated by 2-3 residues where substituted cysteine residues did not respond to the reagents and line one face of an alpha-helix. The activation of alphaS580Cbetagamma mENaC by MTSET was associated with a large increase in channel open probability. Within the carboxyl-terminal portion (alphaSer(583)-alphaSer(592)) of the pore region, only one mutation (alphaS583C) conferred a rapid, nearly complete block by MTSEA, MTSET, and Cd(2+), whereas several other mutant channels were partially blocked by MTSEA or Cd(2+) but not by MTSET. Our data suggest that the outer pore of ENaC is formed by an alpha-helix, followed by an extended region that forms a selectivity filter. Furthermore, our data suggest that the pore region participates in ENaC gating.

Epithelial sodium channels (ENaCs) 1 are composed of three homologous subunits, termed ␣-, ␤-, and ␥ENaC (1,2). These subunits assemble to form a hetero-oligomeric, Na ϩ -selective ion channel with a subunit stoichiometry of 2␣:1␤:1␥ (3,4), although an alternative subunit stoichiometry has been proposed (5,6). All three Na ϩ channel subunits have cytoplasmic amino and carboxyl termini, two transmembrane domains (termed M1 and M2), and a large ectodomain (7)(8)(9). Previous studies have shown that selected point mutations within the pore region preceding M2 of each subunit altered functional properties of the channel, including cation selectivity, single channel conductance, and sensitivity to the blocker amiloride (4, 10 -14). Specific mutations of residues in a conserved threeresidue tract, (G/S)XS (where X is Ser, Gly, or Cys), within the pore region of the three ENaC subunits, rendered channels K ϩ -permeable. Snyder et al. (15) examined the accessibility of a sulfhydryl-reactive methane thiosulfonate (MTS) derivative to substituted cysteine residues within the pore region of human ␥ENaC, and they proposed a structural model of the channel pore similar to that proposed by Kellenberger et al. (11) but distinct from the resolved structure of the KcsA K ϩ channel pore (16).
We previously reported that selected cysteine substitutions within the carboxyl-terminal domain of the pore region of mouse ␣ENaC (␣Ser 580 -␣Ser 592 ) altered the cation selectivity and amiloride sensitivity of the channel and proposed that this region forms the selectivity filter of the channel (14). In the current study, we systematically examined accessibility of sulfhydryl reagents to ␣␤␥mENaCs with engineered cysteine within the 24-residue pore region of the ␣-subunit. Channels with selected cysteine mutations within the carboxyl-terminal portion of the pore region responded to the external application of MTS derivatives with an inhibition of amiloride-sensitive Na ϩ currents. In contrast, we observed a significant increase in amiloride-sensitive Na ϩ currents following the external application of MTS derivatives or Cd 2ϩ when cysteine residues were introduced at selected sites within the amino-terminal portion of the pore region of ␣mENaC. The pattern of distribution of cysteine mutations that led to MTS-induced activation of Na ϩ currents suggests that this region has an ␣-helical structure. In addition, the activation of ␣S580C␤␥ by an MTS reagent was associated with a dramatic increase in channel open probability. We propose that the ENaC pore region forms part of the outer pore vestibule with an ␣-helix followed by an extended region. ENaC may have limited structural similarities with the KcsA K ϩ channel (16,17). In addition, our results suggest that the pore region has a role in ENaC gating.

EXPERIMENTAL PROCEDURES
Reagents-All chemicals were from Sigma unless stated otherwise. Cysteine-scanning Mutagenesis-Site-directed mutagenesis was performed on mouse ␣ENaC (18) with a sequential polymerase chain reaction method using Pfu DNA polymerase (Stratagene, La Jolla, CA). Amino acids ␣Val 569 -␣Ser 592 of ␣mENaC were replaced individually with a cysteine residue, and target mutations were conformed by automated DNA sequencing, as described previously (14).
Two-electrode Voltage Clamp-Two-electrode voltage clamp was performed 20 -72 h after injection at room temperature (22-25°C) as described previously (14). Data acquisition and analyses were performed using pClamp 6.03 software (Axon Instruments) on a Pentium PC. Oocytes were maintained in a recording chamber with 1 ml of bath solution containing (in mM) 100 sodium gluconate, 2 KCl, 1.8 CaCl 2 , 5 BaCl 2 , 10 HEPES, pH 7.2, and continuously perfused at the flow rate of 4 -5 ml/min. Pipettes filled with 3 M KCl had resistances of 0.5-5 M⍀. Typically, oocytes were clamped to a series of voltage steps from Ϫ140 to ϩ40 mV in 20-mV increments for 450 ms every 2 s, and the whole cell currents were measured at 400 ms. Amiloride-sensitive Na ϩ currents were defined as the difference of Na ϩ currents in the absence and presence of 100 M amiloride in the bath solution.
The susceptibility of mutant channel with engineered sulfhydryl groups to sulfhydryl reagents was examined with the sulfhydryl reagents (2-aminoethyl) methanethiosulfonate hydrobromide (MTSEA), [(2-(trimethylammonium) ethyl] methanethiosulfonate bromide (MT-SET), sodium (2-sulfonatoethyl) methanethiosulfonate (MTSES), and Cd 2ϩ . MTSEA (2.5 mM), MTSET (1 mM), and MTSES (5 mM) were prepared in bath solution immediately prior to use. Base-line Na ϩ currents and Na ϩ currents at 1-3 min following perfusion of the oocytes with a reagent were measured. Oocytes were then washed with bath solution for 3 min, and the currents were monitored for observing the reversibility of the response to the reagent. Bath solution containing 100 M amiloride was then applied to the oocyte to determine the amiloride-insensitive current. Responses are expressed as ratios of amiloride-sensitive Na ϩ currents after and before addition of the reagent (I/I 0 ).
Single Channel Recordings-Patch clamp was performed in cellattached mode as described previously (14). Bath and pipette solutions were identical and contained (in mM) 110 LiCl or NaCl, 2 CaCl 2 , 10 HEPES, pH 7.4, adjusted with LiOH or NaOH, respectively. Na ϩ or Li ϩ currents were recorded at Ϫ100 mV (membrane potential). To test the effects of MTS on single channel properties of ␣S580C␤␥ mENaC, 10 mM MTSET prepared in Na ϩ bath solution was added into the bath solution giving a 1 mM final concentration of MTSET, and currents were recorded within 1 h. Event files were generated from long current traces (Ͼ5 min) using Fetchan 6.05 (Axon Instruments), and open probability was estimated with Histogen 6.05 (Axon Instruments).
Statistical Analyses-Values are expressed as mean Ϯ S.E. Student's t test was used for significance analysis between wild type and mutant channel using MS Excel 97 (Microsoft, Inc.).

Responses of Mutant mENaCs to External MTS Deriva-
tives-ENaC pore regions are highly conserved among members of the ENaC/degenerin family (Fig. 1A). Secondary structure predictions of the pore region suggest that the aminoterminal portion may exist as either ␣-helix or ␤-sheet, whereas the carboxyl-terminal portion appears to be more irregular in structure. The center portion is predicted to be a turn region (Fig. 1B). To probe the pore region structure, all residues within the ␣mENaC pore region (␣Val 569 -Ser 592 ) were systematically mutated to cysteine and coexpressed with WT ␤and ␥mENaC subunits in Xenopus oocytes. We previously observed that all mutants with cysteine substitutions within the pore region of ␣mENaC retained channel activity, although low levels of expressed currents (Ͻ200 nA) were observed with two mutants (␣G587C␤␥ and ␣S589C␤␥) (14).
Wild type ␣␤␥mENaC responded to 2.5 mM MTSEA with a partial inhibition of whole cell Na ϩ currents (I/I 0 ϭ 0.75 Ϯ 0.05, n ϭ 17; Fig. 2A). This reduction in current is similar to that reported by other investigators (5,15). The partial inhibition of wild type ENaC by MTSEA is likely due to covalent modification of Cys 547 in ␥mENaC that aligns to Ser 588 in ␣mENaC, as proposed by Snyder et al. (15). Mutation of ␥Cys 547 to serine largely eliminated the MTSEA-induced partial inhibition of Na ϩ currents, whereas mutation of an adjacent cysteine (␥C551S) had no effect on the partial inhibition of Na ϩ currents by MTSEA (data not show). Amiloride-sensitive whole cell Na ϩ currents in oocytes expressing wild type ␣␤␥mENaC were not significantly altered following addition of 1 mM MTSET or 5 mM MTSES to the bath solution ( Figs. 2A and 4A).
We observed distinct effects of MTS reagents on ENaCs with cysteine substitutions within the amino-terminal (␣Val 569 -␣Trp 582 ) and carboxyl-terminal (␣Ser 583 -␣Ser 592 ) domains of the pore region of the mouse ␣-subunit ( Fig. 2A). Within the carboxyl-terminal portion of the pore region of ␣mENaC, only 1 of 8 mutants examined (␣S583C␤␥) responded to both MTSEA (I/I 0 ϭ 0.04 Ϯ 0.01, n ϭ 4) and MTSET (I/I 0 ϭ 0.31 Ϯ 0.09, n ϭ 4) with a large inhibition of the amiloride-sensitive inward Na ϩ current ( Fig. 2A and Fig. 3, B and C). Several mutants (␣S588C, ␣V590C, and ␣L591C) responded to MTSEA with an inhibition of the amiloride-sensitive inward Na ϩ current that was significantly greater than WT. These residues are located in close proximity to and on either side of a three-residue tract (␣Gly 587 -␣Ser 589 ) that has a critical role in restricting K ϩ permeation through the channel (11,12,14,15). One mutant (␣S588C) that responded to MTSEA with a partial inhibition of whole cell Na ϩ current is located within this three-residue tract. The MTSEA-and MTSET-induced inhibition of whole cell Na ϩ currents remained after these reagents were removed from the bath solution (Fig. 3C).
In contrast, channels with cysteine substitutions at multiple sites within the amino-terminal portion of the pore region of ␣mENaC responded to MTSEA (i.e. ␣V572C, ␣S576C, ␣S580C, ␣Q581C, and ␣W582C) or MTSET (i.e. ␣V572C, ␣S576C, ␣N577C, and ␣S580C) with a significant increase in amiloridesensitive inward Na ϩ currents ( Fig. 2A). Only one mutant (␣S573C␤␥) responded to MTSET with an inhibition of whole cell Na ϩ current (I/I 0 ϭ 0.64 Ϯ 0.03, n ϭ 5). The residues within the amino-terminal portion of the pore region where cysteine substitutions responded to MTS reagents with a large change in whole cell Na ϩ current line one face of an ␣-helix, with the exception of ␣W582C (Fig. 2B), suggesting that this region is ␣-helical in structure.
These increases in whole cell currents occurred rapidly (within a minute) after application of MTS reagents and were not reversible after MTSEA or MTSET were removed from the FIG. 1. Sequence alignments and secondary structure prediction of ␣ENaC pore region. A, pore region amino acid sequence alignments of ␣, ␤, and ␥mENaC, deg-1 (degenerin from C. elegans), mec-4 (mechanosensitive protein from C. elegans), hBNC1 (human brain sodium channel), and ASIC1 (acid-sensing ion channel 1). Identical amino acids are shaded, and similar residues are boxed. *, analogous site within the C. elegans proteins deg-1 and mec-4 where mutations result in neurodegeneration; ࡗ, proposed amiloride-binding site in ␣-, ␤-, and ␥ENaC; q, key residues that limit K ϩ permeation. B, secondary structure prediction of the ␣mENaC pore region was performed with DNASis 2.6 for Windows 95 (Hitachi Software Engineering Co., Ltd., South San Francisco, CA) using Chou-Fasman algorithm. Uppercase indicates probability, and lowercase indicates a possibility that the residue occurs in the indicated conformation.
bath solution (Fig. 3C), indicating that the introduced cysteine residues were irreversibly modified. The inward Na ϩ currents following stimulation or inhibition by MTSEA or MTSET were completely inhibited by 100 M amiloride (Fig. 3C). These data suggest that MTS modification of the ␣mENaC pore residues did not interfere with the blocking effect of 100 M amiloride, although it has been proposed that amiloride blocked ENaC by binding to a site formed by ␣Ser 583 , ␤Gly 525 , and ␥Gly 542 (10). Fig. 3A shows representative recordings of the response of ␣S580C␤␥ to 1 mM MTSET.
MTSEA and MTSET introduce positively charged groups to targeted cysteine residues. We tested whether the charges car-ried by MTS reagents were related to the stimulatory or inhibitory effects on mutant channels by examining the effects of a negatively charged reagent MTSES (5 mM) on ␣S576C␤␥, ␣S580C␤␥, and ␣S583C␤␥. ␣S576C␤␥ responded to MTSES with a modest increase in whole cell Na ϩ current (I/I 0 ϭ 1.21 Ϯ 0.04, n ϭ 4), and ␣S580C␤␥ responded to MTSES with a large increase in whole cell Na ϩ current (I/I 0 ϭ 3.3 Ϯ 0.1, n ϭ 6, Fig.  4A). In contrast, whole cell currents in oocytes expressing ␣S583C␤␥ were unchanged in response to MTSES. When oocytes expressing ␣S583C␤␥ were pretreated with MTSES, subsequent treatment with MTSEA still greatly inhibited whole cell Na ϩ currents (I/I 0 ϭ 0.25 Ϯ 0.04, n ϭ 4), suggesting that ␣S583C was still accessible to MTSEA. When either ␣S576C␤␥ or ␣S580C␤␥ was pretreated with MTSES, the subsequent increase in whole cell Na ϩ current in response to MTSEA was significantly less than that observed without MTSES pretreatment (Fig. 4B). These data suggest that MTSES reacted efficiently with both ␣S576C␤␥ and ␣S580C␤␥ but that ␣S583C was largely unmodified by MTSES. Snyder et al. (15) also observed that human ENaC with a cysteine substitution at the site analogous to ␣Ser 583 (i.e. ␥G536C) was modified by MTSET but not by MTSES.
Responses of Mutant mENaCs to External Cd 2ϩ -Group IIB divalent cations such as Cd 2ϩ and Zn 2ϩ are able to bind free sulfhydryls with high affinity and therefore have been used as biophysical probes to study the pore structure of ion channels (19 -22). Cd 2ϩ was used in this study as its crystal radius (0.92 Å) (21) is nearly same as that of Na ϩ (0.95 Å) (23). We examined whether extracellular Cd 2ϩ (5 mM) altered whole cell Na ϩ currents in oocytes expressing either WT or mutant mENaCs. A modest increase in amiloride-sensitive whole cell Na ϩ current was observed in response to Cd 2ϩ in oocytes expressing WT ␣␤␥mENaC. A similar response to extracellular Cd 2ϩ was observed with 16 of the 22 mENaC mutants examined. Several mutations within the ␣mENaC pore region responded to Cd 2ϩ with a significant increase (␣N577C, ␣S580C, and ␣W582C) or a modest decrease (␣G579C and ␣L584C) in whole cell Na ϩ current. Similar to MTSEA, Cd 2ϩ abolished whole cell amiloride-sensitive Na ϩ currents in oocytes expressing ␣S583C␤␥ (I/I 0 ϭ 0.03 Ϯ 0.01, n ϭ 4, Fig. 2A). The blocking effect of Cd 2ϩ on ␣S583C␤␥ was both fast and voltage-dependent, as evidenced by the minimal block of outward currents, compared with the large inhibition of inward currents (Fig. 3B, right panel). This is consistent with the observation of voltage-dependent block of rat ␣S583C␤␥ ENaC by external Zn 2ϩ (10).
Role of Pore Region in ENaC Gating-The introduction of cysteine residues at ␣Val 572 , ␣Ser 576 , ␣Asn 577 , ␣Ser 580 , ␣Gln 581 , or ␣Trp 582 led to channels that responded to MTSEA, MTSET, MTSES, or Cd 2ϩ with an increase in whole cell Na ϩ current. These increases in whole cell currents occurred rapidly (within a minute) after external application of the MTS reagent or Cd 2ϩ (Fig. 3C), suggesting that changes in single channel Na ϩ conductance or open probability occurred. Previous studies demonstrated that the introduction of residues with large side chains at the site analogous to ␣Ser 576 of mENaC led to a significant increase of currents in oocytes expressing ASIC2 (or BNC1), an ENaC-related H ϩ -gated ion channel (24). Similar mutations of deg-1 or mec-4 (mechanosensitive proteins in Caenorhabditis elegans) led to neuronal degeneration (25,26). We examined whether the introduction of a residue with a large side chain at or in proximity to ␣Ser 576 increased whole cell Na ϩ currents. Whole cell Na ϩ currents in oocytes expressing either ␣V572F␤␥ or ␣S576C␤␥ were significantly greater than that observed in oocytes expressing WT ENaC, although Na ϩ currents measured in oocytes expressing ␣V572C␤␥ or ␣S576F␤␥ were similar in magnitude to WT. In contrast, whole FIG. 2. A, accessibility to external 2.5 mM MTSEA, 1 mM MTSET, and 5 mM Cd 2ϩ of wild type and cysteine-substituted mENaCs. Bars (I/I 0 ) indicate ratios of the remaining amiloride-sensitive Na ϩ current at 2 min after external application of the reagent to the amiloride-sensitive Na ϩ current before delivery of the reagent into the bath. Data are presented as mean Ϯ S.E., from 4 to 8 oocytes except for wild type response to MTSEA (17 oocytes). Filled bars indicate statistical significance (p Ͻ 0.05, mutant versus WT). ND indicates not determined due to low level of expressed currents with ␣G587C␤␥ and ␣S589C␤␥. B, helical wheel analysis of ␣mENaC pore residues Thr 570 -Ser 583 . Residues where substitution with cysteine resulted in significant changes in amiloride-sensitive Na ϩ currents following external application of the sulfhydryl reagents are marked with *. Residue Gly 579 is not marked based on the small inhibition by Cd 2ϩ . cell currents measured in oocytes expressing ␣S580F␤␥ or ␣S580C␤␥ were significantly less than that observed in oocytes expressing WT ENaC (Fig. 5).
We performed single channel analyses of ␣S580C␤␥ before and after treatment with MTSET, to test whether the modification of the introduced cysteine residue altered channel gating (Fig. 6). The single channel slope conductance for Na ϩ of ␣S580C␤␥ was 3.7 Ϯ 0.3 pS (n ϭ 11), slightly less than the conductance of wild type mENaC (4.3 pS) (14). This is consistent with our previous observation that the single channel slope conductance for Li ϩ of ␣S580C␤␥ was nearly identical to WT ENaC and that the Li ϩ /Na ϩ current ratio for ␣S580C␤␥ was 1.36-fold greater than that of WT (14). The open probability of ␣S580C␤␥ was 0.07 Ϯ 0.02 (n ϭ 5), determined at potentials between Ϫ100 and Ϫ60 mV. The reduced single channel Na ϩ conductance and open probability were consistent with the reduced whole cell Na ϩ currents observed in oocytes expressing ␣S580C␤␥, when compared with oocytes expressing WT ENaC (Fig. 5). Following treatment with MTSET, ␣S580C␤␥ exhibited a dramatic change in gating characteristics. When patches were made shortly (within minutes) following MTSET treatment, channels were primarily open but exhibited frequent transitions to the closed state (Fig. 6B). However, when patches were made minutes later following MTSET treatment, channels remained open, and very few brief closures were observed (Fig. 6, C and D). The open probability was not determined due to too few transitions between open and closed states despite long (Ͼ10 min) recordings; however, open probability was clearly Ͼ0.9 (Fig. 6, C and D). The single channel conductance of MTSET-modified ␣S580C␤␥ was 2.3 Ϯ 0.1 pS (n ϭ 4), lower than that of unmodified channels (equaling a 38% decrease). In many recordings we only observed noise, comparable to open channel noise, with no clear transitions suggesting that the channel remained open over a recording period of 5-10 min (data not shown). These data indicate that MTSET converted ␣S580C␤␥ to a lower conductance channel, but one that was nearly continuously open. These MTSET-induced changes in conductance and open probability are the likely mechanisms of the MTSET-induced increase in whole cell Na ϩ current observed with this mutant channel ( Figs. 2A and 3A).
Similar single channel analyses were performed with Li ϩ as the conducting ion in the pipette, as well as in the bath solution. We observed that MTSET treatment of oocytes expressing ␣S580C␤␥ reduced the unitary Li ϩ current from 0.7 to 0.3 pA at Ϫ80 mV and locked the channel in an open state as only brief closures were observed (Fig. 6F). The Li ϩ current reduction in response to MTSET was larger than Na ϩ current reduction following MTSET treatment. These data obtained from single channel analyses with Li ϩ as the conducting ion were consistent with the effect of MTSET on amiloride-sensitive whole cell Li ϩ currents measured in oocytes expressing ␣S580C␤␥. Unlike the effects of MTSET on whole cell ␣S580C␤␥ Na ϩ currents, Li ϩ currents were not significantly increased by MTSET (I/I 0 ϭ 0.93 Ϯ 0.06, n ϭ 4). It is likely that the MTSET-induced reduction in Li ϩ unitary current balances the MTSET-induced increase in ␣S580C␤␥ open probability. DISCUSSION The substituted cysteine accessibility method has been used to probe pore structure of various ion channels (27). In this study, we used this approach to examine the pore structure of ␣ENaC. Several distinct effects of MTS reagents and Cd 2ϩ on ␣␤␥mENaCs were observed with cysteine substitutions within the pore region of the ␣-subunit. MTSEA, MTSET, and Cd 2ϩ inhibited whole cell Na ϩ currents in oocytes expressing ␣S583C␤␥, as previously reported (4,10). Surprisingly, this is the only mutant that displayed significant block by these three reagents. The lack of an inhibitory effect of MTS reagents on engineered cysteines near ␣Ser 583 indicates that this residue is located within a restricted site. Interestingly, another MTS reagent (MTSES) with a negative charge did not inhibit ␣S583C␤␥. Furthermore, MTSES pretreatment of oocytes expressing this mutant channel failed to prevent the subsequent inhibition by MTSEA (Fig. 4B), indicating that MTSES did not efficiently modify ␣S583C␤␥. Schild et al. (10) proposed that ␣Ser 583 is located in the electrical field of the ENaC pore, as Zn 2ϩ -induced inhibition of ␣S583C␤␥ was voltage-dependent. We also observed that Cd 2ϩ -induced block of ␣S583C␤␥ mENaC was voltage-dependent (Fig. 3B). These results suggest that a negative potential within the vicinity of ␣Ser 583 hinders the access of MTSES. This notion is reminiscent of the proposal that the pore helix of KcsA K ϩ channels generates a negative potential at its carboxyl terminus that contributes to the stabilization of cations in the pore cavity (28).
At sites amino-terminal to ␣Ser 583 , only ␣S573C␤␥ and ␣G579C␤␥ responded to MTSET or Cd 2ϩ with a partial inhibition of the whole cell Na ϩ current. Several mutant channels with cysteine mutations within the amino-terminal portion of the pore region (i.e. amino-terminal to ␣Ser 583 ) responded to sulfhydryl reagents with significant increases in whole cell current. These residues (␣Val 572 , ␣Ser 576 , ␣Asn 577 , ␣Ser 580 , ␣Gln 581 , and ␣Trp 582 ) line one face of an ␣-helix, with the exception of ␣W582C (Fig. 2B), consistent with an ␣-helical structure as suggested by secondary structure predictions (Fig.  1B). These stimulatory effects do not appear to rely on the positive charges carried by these reagents, as the negatively charged MTSES also stimulated ␣S576C␤␥ and ␣S580C␤␥ (Fig. 4A). Snyder et al. (15) also observed MTSET-induced activation of Na ϩ currents with cysteine mutations in this region of human ␥ENaC, although only three mutants responded with a gain of function in response to MTSET, and the location of these residues was not suggestive of an ␣-helical structure. These differences in responses of ␣and ␥-subunit mutants to MTS reagents may reflect, in part, the presence of two ␣-subunits and only one ␥-subunit within each channel protein. Another possibility is that the three ENaC subunits may not be arranged symmetrically along pore axis. This latter notion is supported by the observations that mutations at homologous sites in different subunits led to different changes in channel selectivity, amiloride sensitivity, and divalent cation sensitivity (10,12,15). Although KcsA core pore is formed by four identical subunits in a symmetrical manner (16), studies on voltage-gated Na ϩ channels suggested that the four-pore segments from Domains I-IV are arranged asymmetrically (19,20).
ENaCs with cysteine substitutions at sites carboxyl-terminal to ␣Ser 583 either did not responded to MTS reagents and Cd 2ϩ or, alternatively, were partially inhibited (i.e. ␣L584C␤␥, ␣S588C␤␥, ␣V590C␤␥, and ␣L591C␤␥). MTSET did not inhibit channels with introduced cysteines carboxyl-terminal to ␣Ser 583 , a region encompassing the proposed selectivity filter. Given several reports indicating ␣Gly 587 and ␣Ser 589 have an important role in conferring cation selectivity and restricting K ϩ permeation (11,12,14), residues ␣Ser 587 -␣Ser 589 and adjacent residues are likely facing the conducting pore. A simple explanation for a lack of an inhibitory effect of MTSET on channels with cysteine substitutions in this region is that these sites are not accessible to the reagent.
Structure of the ENaC Pore Region-The effects of MTS reagents and Cd 2ϩ were most prominent when cysteine residues were placed at sites amino-terminal to ␣Ser 583 . Only modest changes in whole cell currents were observed in response to MTSEA or Cd 2ϩ when cysteine residues were placed carboxyl-terminal to ␣Ser 583 . As these accessible residues (defined by a large response to MTS reagents or Cd 2ϩ ) preceded ␣Leu 584 , our results support a model of the pore region of the channel that has been proposed by both Kellenberger et al. (11) and Snyder et al. (15). These groups suggested that pore is formed by residues that enter the membrane spanning region from an extracellular site and transitions to an ␣-helical second membrane spanning domain. As the pore enters the membrane, it gradually narrows to form a site that restricts K ϩ , analogous to a funnel that narrows to its spout. This model is based, in part, on the observation that the substitution of cysteine residues in the ␤or ␥-subunits at a position analogous to ␣Ser 583 (␤G525C and ␥G542C) resulted in a large increase in the K i values of amiloride. Schild and co-workers (10, 11) proposed that amiloride interacts directly with these residues and that ␣Ser 583 must be external to ␣Ser 589 . A model of the outer vestibule of the pore, incorporating an ␣-helical structure that transitions to a narrow selectivity filter, is illustrated in Fig.  7B. This model is consistent with that proposed by Palmer (29) 10 years ago. Schild et al. (10) previously demonstrated that ENaCs with acidic residues at position ␣Ser 580 (or at the analogous positions ␤Gly 522 and ␥Gly 534 ) were inhibited by extracellular Ca 2ϩ in a voltage-dependent manner, suggesting that these residues (i.e. amino-terminal to ␣Ser 583 , ␤Gly 525 , and ␥Gly 537 , respectively) were accessible to extracellular Ca 2ϩ . These data support the proposed pore structure of Kellenberger et al. (11) and Snyder et al. (15), whereby the interaction of Ca 2ϩ with ␣S580D would block the pore of the channel (see Fig. 7B). However, we observed that oocytes expressing ␣S580C␤␥ mENaC responded to MTSEA, MTSET, and Cd 2ϩ with a large increase in amiloride-sensitive currents. If the ␣S580C side chain extends into the pore lumen, its modification by MTS reagents or Cd 2ϩ would be expected to inhibit the channel and not activate it. Furthermore, channels with substituted cysteine residues at selected sites near ␣Ser 580 (i.e. ␣Val 572 , ␣Ser 576 , ␣Asn 577 , ␣Gln 581 , or ␣Trp 582 ) also responded to MT-SEA, MTSET, or Cd 2ϩ with an activation of amiloride-sensitive currents. If this proposed pore structure is correct, it is reasonable to predict that channels with cysteine residues substituted at multiple sites amino-terminal to the selectivity filter would respond to MTSEA, MTSET, or Cd 2ϩ with a large inhibition of whole cell currents. However, of the mutants we have examined, only ␣S583C␤␥ responded to these reagents with a large inhibition of Na ϩ current. Our results suggest that periodic residues within the amino-terminal portion of the ␣mENaC pore region are accessible to sulfhydryl reagents externally applied but do not directly face the conducting pore.
The outer pore of the KcsA K ϩ channel is formed by an ␣-helix that enters the membrane followed by an extended region directed toward the extracellular space (16) (Fig. 7A). Is it necessary to propose a pore structure for ENaC that basically differs from other highly selective cation channels? Previous studies have suggested that the pore regions of the voltagegated Na ϩ and Ca 2ϩ channels are similar in structure to voltage-gated K ϩ channels, although they clearly differ in mechanisms for achieving cation selectivity (30). Our data suggest that pore region residues amino-terminal to ␣Gln 581 form an ␣-helix, and our previous results (14) suggested that residues extending from ␣Ser 580 -␣Ser 589 form an extended selectivity filter. This secondary structure is similar to that of the KcsA K ϩ channel. Furthermore, within the pore regions of K ϩ channels (Shaker and inward rectifier Kir2.1) and ␣mENaC, the pattern of accessibility to substituted cysteine residues to cysteine-reactive reagents is strikingly similar (Fig. 7C) (31)(32)(33)(34). Our previous observation that mutation of the GSS tract within the selectivity filter of the ␣-subunit of mENaC (␣Gly 587 -␣Ser 589 ) to the K ϩ channel selectivity filter signature sequence GYG rendered the mutant channel K ϩ -selective (14) is also consistent with the notion that K ϩ channels and ENaC may have similar pore structures. Fig. 7A illustrates a model of the ENaC ␣-subunit pore region using the structure of the KscA K ϩ channel, aligning the GYG tract within KscA with the GSS tract within ␣ENaC. Residues ␣Val 572 -␣Ser 580 form an ␣-helix; ␣Leu 584 -␣Ser 589 form an extended selectivity filter; and ␣Gln 581 -␣Ser 583 are located at the turn region where the ␣-helix transitions into the selectivity filter. Residues where substituted cysteines responded to MTS reagents or Cd 2ϩ line one face of the helix, with the exception of ␣Trp 582 (Fig. 2B). Our model places ␣Ser 592 at a location external to the GSS track, consistent with the previous observation that the mutant ␣S592I␤␥ rat ENaC was blocked by external Ca 2ϩ in a voltage-dependent manner (13). This model (Fig. 7A) is not consistent with amiloride interacting directly with ␤G525C or ␥G542C, residues proposed to interact directly with amiloride (10). These residues would be within the internal portion of the selectivity filter. However, these mutations (i.e. ␤G525C or ␥G542C) might change the structure of the pore and indirectly alter the K i values of amiloride, as suggested by Schild et al. (10).
The introduction of cysteine residues carboxyl-terminal to the GYG tract of K ϩ channels conferred sensitivity to MTS reagents (Fig. 7, A and C). If ENaC and K ϩ channels share a similar pore structure, the introduction of cysteine residues carboxyl-terminal to ␣Ser 589 would be expected to confer sensitivity to MTS reagents. However, we only observed a modest block of Na ϩ currents in response to MTSEA when cysteine residues were present at ␣Val 590 or ␣Leu 591 , and no changes in whole cell Na ϩ currents were observed in response to MTSET or Cd 2ϩ . These results suggest that K ϩ channels and ENaC differ in their pore structure, although the nature of these differences has yet to be defined. Based on our data and previous studies, we propose that the amino-terminal portion of the ␣ENaC pore region forms an ␣-helix, and the carboxyl-terminal FIG. 7. Structural models of the ␣ENaC pore region. A, structure of KcsA pore region (right) and a model for ␣mENaC pore region (left) are presented in stick model with ribbon rendering (9 shin green lines) using the modeling software HyperChem 5.1 (Hypercube Inc., Gainesville, FL). For KcsA K ϩ channel, structure of residues P41-Y60 was generated from coordinates obtained from the Protein Data Bank (code: 1BL8) and the numbering of the residues is shown in C of this figure. The structural model for ENaC pore region was generated by mutating KcsA pore region residues to Val 572 -Ser 592 of ␣mENaC according to the alignments shown in C. Residue ␣Phe 586 was omitted and energy minimization was not performed. Element colors are as followings: cyan for carbon, blue for nitrogen, and red for oxygen. Residues corresponding to Val 572 , Ser 573 , Ser 576 , Asp 577 , Ser 580 , Gln 581 , Trp 582 , and Ser 583 of ␣mENaC are displayed in violet color. Substituted cysteines at these sites were accessible to sulfhydryl reagents from extracellular side judged by significant changes in amiloride-sensitive Na ϩ currents in response to these reagents. KcsA residues homologous to the residues in voltage-gated and inward rectifier K ϩ channels that are accessible to external sulfhydryl reagents are also shown in violet color. B, an alternative model for the ␣mENaC pore region was generated by rotating residues Ser 12 -Ser 20 (corresponding to Ser 583 -Ser 592 of ␣mENaC) from the model A (left panel) by 180°along the X axis. Externally accessible residues in ␣mENaC are also highlighted in violet color as in A. This model includes all 24 residues (Val 569 -Ser 592 ) of ␣mENaC pore region. The key residues retaining K ϩ /Na ϩ selectivity (Gly 19 and Ser 21 , corresponding to Gly 587 and Ser 589 of ␣mENaC) are located in the narrowest region of the pore. C, sequence alignments and comparison of accessibility to external sulfhydryl reagents between K ϩ channels and ENaC. The alignments of the pore region residues of KcsA, Shaker K ϩ channel, inward rectifier K ϩ channel Kir 2.1 and ␣mENaC were performed by aligning the Gly 587 -Ser 588 -Ser 589 of ␣mENaC to the K ϩ channel GYG track. A gap was introduced in K ϩ channels to enhance overall alignments. The plus sign (ϩ) indicates sites where cysteine substitutions within the pore region result in channels that were inhibited (or activated) by MTS reagents, Ag ϩ , or Cd 2ϩ and the minus sign (Ϫ) shows residues not accessible to the reagents. The plus/minus sign (Ϯ) indicates sites where cysteine substitutions result in channels that were modestly inhibited by MTSEA or Cd 2ϩ . The X indicates the sites not tested for sulfhydryl reagent accessibility. portion forms a selectivity filter, like that of K ϩ channels. Furthermore, we propose that the transition from selectivity filter to ␣-helical second membrane spanning domain within ENaC has a structure that differs from K ϩ channels.
Role of Pore Region in ENaC Gating-Our observation that mutant mENaCs with cysteine substitutions within the aminoterminal portion of the pore region (preceding ␣Ser 583 ) responded to MTSEA, MTSET, or Cd 2ϩ with an increase in whole cell Na ϩ current led us to examine whether this region has a role in ENaC gating. In agreement with our observation that MTS reagents activated whole cell Na ϩ currents in oocytes expressing ␣S580C␤␥, MTSET induced a large increase in open probability of ␣S580C␤␥, indicating that this residue is within a domain that controls ENaC gating. Our data suggest ␣Ser 576 and ␣Ser 580 are two crucial residues in this gating domain, as all three MTS reagents dramatically stimulated whole cell Na ϩ currents in oocytes expressing these mutant channels. This proposed role of ␣Ser 576 in channel gating is in agreement with previous observations suggesting a similar role of the residue at the analogous position in degenerins and in H ϩ -gated Na ϩ channels. The introduction of bulky residues at a site analogous to ␣S576C within the C. elegans proteins deg-1 and mec-4 lead to neurodegeneration that is proposed to occur as a result of an unregulated activation of a putative mechanosensitive cation channel (35,36). Similarly, selected mutations at an analogous site within this pore region of related H ϩ -gated Na ϩ channels, termed BNC (or ASIC2), resulted in sustained channel activation that was independent of acidification (24). Based on these observations, we propose that the ENaC pore region participates in channel gating.
Aside from increasing open probability, external application of MTSET reduced both Na ϩ and Li ϩ unitary currents of ␣S580C␤␥. The observed changes in whole cell amiloride-sensitive Na ϩ currents of ␣S580C␤␥ in response to MTSET reflects opposing effects on open probability (increased) and unitary current (decreased). The mechanism of MTSET-induced reduction in unitary current is unknown. This may reflect changes in the conformation of the open channel due to covalent binding of MTSET to ␣S580C; alternatively, the positively charged MT-SET may partially block the pore.
Waldmann et al. (13) reported that ␣S589I and ␣S589F (analogous to ␣Ser 588 of mENaC) reduced rat ENaC open probability from 0.89 to 0.09 and 0.04, respectively, and increased Na ϩ unitary conductance from 4.6 to 10 pS without changing Li ϩ conductance. This study provided evidence that ␣Ser 588 , located within the selectivity filter of ENaC, has a role in ENaC gating. We have also observed that ␣S588C␤␥ mENaC has very low open probability resulting from short open and long close times (data not show). Fyfe et al. (37) studied the functional properties of ENaCs with chimeric ␥-␤ subunits. Their results also suggested that this region regulates ENaC gating. These observations that mutations within both amino-terminal and carboxyl-terminal portions of the pore region resulted in changes in channel gating suggest a collaborative role of these two regions in ENaC gating through intradomain interactions. We propose a working model for ENaC gating. In this model, the pore helix (amino-terminal portion of pore region) undergoes rotational movement. This movement could be in response to conformational changes at other sites within ENaC, such as the ectodomain, the M2 domains, or cytoplasmic domains (i.e. the amino terminus (38)). The rotation of the pore helix leads to changes in the diameter of the selectivity filter, which in turn allows ion translocation through the pore. A recent study of KscA gating observed movement of reporter cysteines introduced at the carboxyl-terminal end of the pore helix in association with the channel gating (45). If the ENaC pore structure shares the fundamental design of the KcsA K ϩ channel pore, ␣Ser 576 and ␣Ser 580 are expected to be close to the carboxylterminal end of the pore helix of ENaC. Substitution of large side chains at these two sites might hinder rotational movement of the pore helix, essentially locking the channel in an open state. Our model suggests that there are extensive interactions between the amino-terminal and carboxyl-terminal portions of the pore region, and we propose that this occurs, in part, through hydrogen bonding. ENaC pore regions have a large number of serine residues (7 in ␣mENaC, 2 in ␤mENaC, and 4 in ␥mENaC). Given the proposed subunit stoichiometry of 2␣, 1␤, and 1␥, a total of 20 serine residues are present within the pore regions of the channel. These serine residues, as well as other polar residues and backbone carbonyl oxygens, are capable of forming a network of hydrogen bonds. A sliding model of gating is also plausible. A relative sliding movement between ␣Ser 576 -Ser 580 and ␣Ser 588 -Ser 592 could lead to channel transitions between open and closed states.
Our model predicts a connection between ENaC gating and permeation and favors a dynamic selectivity filter. Gating and permeation have been generally considered as independent processes of ionic channels since 1952 (39). However, recent studies have challenged this concept (40,41). The notion that pore regions function as ion channel gates has been proposed for cyclic nucleotide-gated channel (42). Recent studies support a dynamic selectivity filter challenging another long term notion of a rigid selectivity filter (43).
Alternatively, if the ENaC pore is similar to that proposed by Kellenberger et al. (11) and Snyder et al. (15) (Fig. 7B), gating might also involve rotational movement of the ␣-helical region preceding the selectivity filter. This gating mechanism was proposed by Adams et al. (24) in their studies of activation of the H ϩ -gated Na ϩ channel BNC (or ASIC2).
We observed that several ␣ENaC mutants in the pore helix, including ␣V572F␤␥ and ␣S576C␤␥, expressed whole cell currents in oocytes that were significantly greater than that observed with WT ENaC, consistent with the notion that mutations within this region affect channel gating. We anticipate that selected mutations within this domain of human ENaC are likely to be found in the clinical setting of salt-sensitive hypertension due to enhanced ENaC-mediated Na ϩ transport in the distal nephron. In this context, Melander et al. (44) reported a mutation within human ENaC (␥N530K), a position analogous to mouse ␣Asn 577 , in a patient with hypertension and diabetes, although a causal relationship was not established. In summary, our data suggest that the amino-terminal pore region of ␣ENaC is ␣-helical in structure and that this region is involved in channel gating.