Side chain orientation of residues lining the selectivity filter of epithelial Na+ channels.

Epithelial Na(+) channels (ENaCs) selectively conduct Na(+) and Li(+) but exclude K(+). A three-residue tract ((G/S)XS) present within all three subunits has been identified as a key structure forming a putative selectivity filter. We investigated the side chain orientation of residues within this tract by analyzing accessibility of the introduced sulfhydryl groups to thiophilic Cd(2+). Xenopus oocytes were used to express wild-type or mutant mouse alphabetagammaENaCs. The blocking effect of external Cd(2+) was examined by comparing amiloride-sensitive Na(+) currents measured by two-electrode voltage clamp in the absence and presence of Cd(2+) in the bath solution. The currents in mutant channels containing a single Cys substitution at the first or third position within the (G/S)XS tract (alphaG587C, alphaS589C, betaG529C, betaS531C, gammaS546C, and gammaS548C) were blocked by Cd(2+) with varying inhibitory constants (0.06-13 mm), whereas the currents in control channels were largely insensitive to Cd(2+) at concentrations up to 10 mm. The Cd(2+) blocking effects were fast, with time constants in the range of seconds, and were only partially reversible. The blocked currents were restored by 10 mm dithiothreitol. Mutant channels containing alanine or serine substitutions at these sites within the alpha subunit were only poorly and reversibly blocked by 10 mm Cd(2+). These results indicate that the introduced sulfhydryl groups face the conduction pore and suggest that serine hydroxyl groups within the selectivity filter in wild-type ENaCs face the conduction pore and may contribute to cation selectivity by participating in coordination of permeating cations.

Epithelial Na ϩ channels (ENaCs) 1 mediate Na ϩ transport across apical membranes of high resistance epithelia and participate in body fluid homeostasis, control of blood pressure, and airway fluid balance (1). Four subunits have been cloned from mammalian tissues and are termed ␣ENaC, ␤ENaC, ␥ENaC, and ␦ENaC. Na ϩ -transporting epithelial cells typically express ␣, ␤, and ␥ subunits. ENaC subunits belong to the ENaC/DEG superfamily with degenerins (DEG and MEC) from Caenorhabditis elegans mechanosensory neurons, acid-sensing ion channels, and peptide-gated Na ϩ channels (2). ENaC/DEG members have limited sequence homology and share a membrane topology similar to that of other two-transmembrane domain channels such as inward rectifier K ϩ channels and P2X receptors (3). A tetrameric structure has been reported for ␣␤␥ENaCs (two ␣ subunit, one ␤ subunit, and one ␥ subunit), ␣ subunit-only channels, and FMRF amide-gated Na ϩ channel (4 -7). However, a channel composition of more than four subunits has also been suggested (8,9).
Studies on cloned and native channels have shown that typical ENaCs are highly selective for the small alkali cations Li ϩ and Na ϩ (10). Cloned ␣␤␥ENaC has a Li ϩ /Na ϩ selectivity of 1.5-2.0 and is essentially impermeable to K ϩ , with a K ϩ /Na ϩ selectivity of Ͻ0.01 (11)(12)(13). Using site-directed mutagenesis, we and others have identified key residues that dictate ENaC cation selectivity (11)(12)(13)(14)(15). A three-residue tract ((G/S)XS) within a region (pre-M2 or pore region) preceding the second transmembrane domain (referred to as M2) of ␣, ␤, and ␥ subunits is considered to be the key element forming a putative selectivity filter based on complete or partial loss of selectivity among Na ϩ , Li ϩ , and K ϩ in mutant channels containing point mutations at the first or third position within this tract (11)(12)(13)(14)(15). The contributions of the three subunits to ion selectivity are not equal, suggesting that the selectivity filter has an asymmetric structure that is distinct from the nearly perfect symmetry of K ϩ channel selectivity filters (15). In addition to the three-residue tract, mutations of nearby residues or of selected residues in M2 or post-M2 regions also alter cation selectivity (13,(15)(16)(17). Based on results from previous studies, the general consensus is that the ENaC pore is formed by the pre-M2 and M2 domains of all three subunits; however, the orientation of these two segments (pre-M2 and M2) within the three-dimensional ENaC structure is unknown, and different models have been proposed (11-13, 15, 18).
Crystal structures of bacterial K ϩ channels have provided insight regarding mechanisms by which K ϩ channels achieve a K ϩ -selective phenotype (19,20). Four K ϩ -binding sites within the 12-Å-long selectivity filter are formed by 16 carbonyl oxygen atoms from the signature sequence TVGY and four hydroxyl oxygen atoms from the threonine residues (20). In contrast, the ion-binding sites within the selectivity filter of voltage-gated Na ϩ and Ca 2ϩ channels are thought to come from side chains of key residues (DEKA for the Na ϩ channels and EEEE for the Ca 2ϩ channels) (21). It is not clear whether backbone oxygens, side chain oxygens, or both coordinate permeating Na ϩ and Li ϩ ions in ENaCs. Kellenberger et al. (22) recently proposed that the four carbonyl oxygen atoms from four serine residues at the third position within the (G/S)XS tracts (two from ␣ENaC and one from ␤ENaC and ␥ENaC each) provide a Na ϩ -binding site and that the side chains point to the intersubunit spaces and do not interact with cations passing through the selectivity filter. The proposed side chain orientation of the Ser residues at the third position was largely based on a positive correlation between increases in K ϩ permeability and the volumes of several substituted residues at ␣Ser 589 . Our previous study suggested that the hydroxyl group of ␣Ser 589 might contribute to Na ϩ binding and cation selectivity since only a Thr substitution at ␣Ser 589 retained the ability to discriminate between Na ϩ and K ϩ (13). As the Van der Waals volume of the Thr residue (93 Å 3 ) is 27% greater than that of the Ser residue (73 Å 3 ) (23), our results did not support the model proposed by Kellenberger et al. (22) and raised the possibility that the hydroxyl side chain at position 589 of the ␣ subunit might interact with ions traversing the selectivity filter. To probe the side chain orientation of filterlining residues, we examined the accessibility of engineered sulfhydryl groups at the first and third positions in mouse ␣ENaC, ␤ENaC, and ␥ENaC to external Cd 2ϩ . Our results indicate that sulfhydryl groups from Cys residues introduced at the first and third positions of the selectivity filter tract ((G/ S)XS) interact with external Cd 2ϩ and suggest that the side chains of Ser residues within the selectivity filter, including ␣Ser 589 , may point to the pore lumen and interact with cations within the filter.

EXPERIMENTAL PROCEDURES
Site-directed Mutagenesis-Site-directed mutagenesis was performed on cDNAs for mouse ␣ENaC, ␤ENaC, and ␥ENaC cloned into pBluescript SK Ϫ (Stratagene) as reported previously (13). Target mutations were confirmed by direct sequencing. The first and third residues within the three-residue tract ((G/S)XS) of ␣ENaC, ␤ENaC, and ␥ENaC were individually substituted with Cys. To enhance functional expression in oocytes, we generated ␤ and ␥ subunits with an introduced Liddle's mutation (␤R564X or ␥R583X, denoted ␤ T or ␥ T , respectively) that truncates their intracellular C termini.
Two-electrode Voltage Clamp-All experiments were performed 20 -96 h after cRNA injections at room temperature (20 -24°C). Oocytes were placed in an oocyte recording chamber from Warner Instrument Corp. (Hamden, CT) and perfused with a bath solution containing 110 mM NaCl, 2 mM KCl, 2 mM CaCl 2 , and 10 mM HEPES, pH 7.4. Whole cell Na ϩ currents were measured by two-electrode voltage clamp as described previously (13). To obtain a current-voltage relationship, oocytes were clamped from Ϫ140 to 60 mV in 20-mV increments for 0.5 s. For time course experiments, oocytes were continuously clamped at Ϫ100 mV. Amiloride-sensitive currents are defined as the difference between total currents and the currents with 100 M amiloride in the bath solution.
Cd 2ϩ Accessibility-The effects of Cd 2ϩ on mutant ENaCs were examined by comparing currents before and after external application of CdCl 2 . Increasing concentrations of CdCl 2 (from 10 Ϫ7 to 10 Ϫ2 M) were used to determine dose-response relationships. As some mutations within the selectivity filter tract decrease amiloride sensitivity (15,24), whole cell currents (rather than amiloride-sensitive currents) were used to obtain dose-response curves. Amiloride at 100 M was typically added to the bath solution at the end of an experiment to measure the amiloride-insensitive current. Recordings from oocytes that had high amiloride-insensitive currents (Ͼ0.5 A at Ϫ100 mV) were not included in analysis to minimize current contamination from endogenous channels, with the exception of recordings from oocytes expressing ␣␤G525C␥ and ␣␤␥G542C, as these mutants show little block by 100 M amiloride due to an ϳ1000-fold increase in the amiloride K i (24). Inhibitory constants for Cd 2ϩ were estimated from the best fit of doseresponse data by nonlinear least-square curve fitting with the Hill equation: where I is the relative current obtained by normalizing the current in the presence of Cd 2ϩ to the value in the absence of Cd 2ϩ , K i is the inhibitory constant, n is the Hill coefficient, and the C refers to [Cd 2ϩ ].
Statistical Analysis-Values are expressed as means Ϯ S.E. Statistical significance was determined by Student's t test. A p value of Ͻ0.05 was considered statistically significant. Nonlinear least-square curve fittings were performed with OriginPro 7 (OriginLab Corp., Northampton, MA).

RESULTS
We previously observed low levels of functional ENaC expression in Xenopus oocytes expressing ␣␤␥mENaCs with a Cys substitution at the first or third residue within the (G/S)XS tract of the putative selectivity filter, with whole cell Na ϩ currents at Ϫ100 mV typically below 500 nA (13,15). Taking advantage of the enhanced surface expression of channels with Liddle's mutations, we investigated accessibility of external Cd 2ϩ to engineered sulfhydryl groups within the (G/S)XS tract. A sequence alignment of residues near the (G/S)XS tracts in the ENaC/DEG family is shown in Fig. 1A. Our rationale was that Cys-substituted channels would be blocked by external Cd 2ϩ if introduced sulfhydryl groups point to the filter lumen as illus-FIG. 1. Pore region sequence alignment and experimental models. A, amino acid sequences neighboring the putative selectivity filter regions of ENaC/DEG family members. Identical residues are shown in gray. The box identifies the three-residue tract that forms a putative selectivity filter. The identified amiloride-binding site is indicated by the arrow. m, mouse; r, rat; h, human; FaNaCh, FMRF amidegated Na ϩ channel. B and C, models illustrating two of the possible orientations of an engineered Cys residue within the selectivity filter. The conduction pore is shown as the open space between the two arched lines. A Cys residue is displayed in a ball-and-stick mode. The larger and smaller black spheres represent the sulfur and carbonyl oxygen atoms, respectively. Carbon and nitrogen atoms are shown as gray spheres. In B, the sulfhydryl group faces the conduction pore and binds Cd 2ϩ . The binding of Cd 2ϩ blocks Na ϩ conduction through the selectivity filter and therefore blocks Na ϩ currents. In C, the sulfhydryl group faces away from the conduction pore and would not bind Cd 2ϩ . As a result, Cd 2ϩ does not block Na ϩ currents. trated in Fig. 1B. Conversely, channels would not be blocked by Cd 2ϩ if the introduced sulfhydryl groups face away from the pore lumen as shown in Fig. 1C.
External Cd 2ϩ Blocks ENaCs with a Single Cysteine Substitution within the Putative Selectivity Filter-Whole cell currents in oocytes expressing ␣␤␥mENaCs with a single Cys substitution and a Liddle's mutation were in the range of 1ϳ5 A at Ϫ100 mV (Table I). Amiloride at 100 M blocked the currents of the mutants containing a Cys substitution at the third Ser in the ␣, ␤, or ␥ subunit to Ͻ200 nA, a level typically observed in non-injected oocytes. However, amiloride failed to block the currents of channels containing a Cys mutation at the first position of the selectivity filter tract of the ␣ or ␤ subunit to the same level. This is consistent with previous reports that mutations at ␣Gly 587 and ␤Gly 529 decrease the amiloride sensitivity of the mutant channels (14,15). The effects of external Cd 2ϩ on ENaC currents were examined by analyzing the current changes in the presence of increasing concentrations of CdCl 2 (from 10 Ϫ7 to 10 Ϫ2 M) in the bath solution. Representative recordings are shown in Fig. 2. All six mutant channels were blocked by external Cd 2ϩ in a dose-dependent manner (Fig. 3). The estimated inhibitory constants (K i ) and Hill coefficients are listed in Table I. In contrast, control channels containing a Liddle's mutation but with no Cys substitution were not blocked by Cd 2ϩ . These results indicate that the introduced sulfhydryl groups at the selectivity filter-lining sites in all three subunits were accessible to external Cd 2ϩ . It was apparent that mutant channels with a Cys substitution at the first position within the filter tract (␣G587C, ␤G529C, and ␥S546C) showed higher sensitivity to Cd 2ϩ than channels containing a Cys mutation at the third position (␣S589C, ␤S531C, and ␥S548C) of the same subunit (p Ͻ 0.05). With Cys substitutions at the first or third position of the (G/S)XS tract, ␣ mutants were more sensitive to Cd 2ϩ than ␤ and ␥ mutants (p Ͻ 0.05), whereas ␤ and ␥ mutants showed similar Cd 2ϩ sensitivity (p Ͼ 0.05).
Cd 2ϩ Block Is Reversed by the Reducing Reagent Dithiothreitol (DTT)-As a group 12 metal, Cd 2ϩ binds free sulfhydryl groups with high affinity in a "near-covalent" manner (25). To confirm that the Cd 2ϩ -induced block of the Cys mutant ENaCs was due to Cd 2ϩ binding to the introduced sulfhydryl group, we examined the reversibility of the block. As shown in Fig. 4 (A-F), currents only partially recovered following Cd 2ϩ washout from the bath solution. We also tested the reversibility of the Cd 2ϩ block with a single voltage (Ϫ100 mV) clamp protocol (Fig. 4, G and H). Consistent with the above results, Na ϩ currents in oocytes expressing ␣G587C␤ T ␥ or ␣S589C␤␥ T were only partially restored following washout of Cd 2ϩ from the bath solution. The addition of the reducing reagent DTT (10 mM) restored currents to near control levels. Currents restored by DTT were blocked by 0.1 mM amiloride, eliminating the possibility that DTT-induced currents were mediated by endogenous channels. Moreover, application and washout of 10 mM DTT did not produce noticeable effects on currents in oocytes expressing the same mutant channels without prior application of Cd 2ϩ . These results suggest that the observed Cd 2ϩ block of the mutant ENaCs containing Cys substitutions within the putative selectivity filter was due to Cd 2ϩ interaction with thiol groups rather than with oxygens.
Effects of External Cd 2ϩ on Non-Cys Mutants-To eliminate the possibility that the Cd 2ϩ block of the Cys mutants was due to Cd 2ϩ binding to endogenous Cys in the ENaC pore, we examined the effects of Cd 2ϩ on channels containing an Ala substitution at ␣Gly 587 or ␣Ser 589 or a Ser substitution at ␣Gly 587 . External application of Cd 2ϩ at concentrations up to 1 mM did not block the currents in oocytes expressing ␣G587A, ␣G587S, or ␣S589A (Fig. 5), where either the ␤ or ␥ subunit had a Liddle's mutation. 10 mM Cd 2ϩ blocked Ͻ40% of the currents in these mutants, and the block was completely reversed by Cd 2ϩ washout from the bath solution, in contrast to the poor reversibility of the Cd 2ϩ inhibition of ␣G587C␤ T ␥ and ␣S589C␤ T ␥ (Fig. 4). The partial and low affinity block by 10 mM Cd 2ϩ likely reflects binding of divalent Cd 2ϩ to non-thiol ligands within the selectivity filter of the mutant channels and suggests that the observed Cd 2ϩ block of the Cys mutant channels at submillimolar concentrations resulted from Cd 2ϩ binding to the introduced sulfhydryl groups.
Cd 2ϩ Block and the Endogenous Cys Residue-The ␥ subunit has a natural Cys within the (G/S)XS tract (Cys 547 ). This residue is not conserved among ENaC/DEG family members ( Fig.  1) and is not required for cation selectivity, as point mutations of this Cys do not alter Li ϩ /Na ϩ or K ϩ /Na ϩ selectivity (12). To investigate whether ␥Cys 547 was involved in the Cd 2ϩ block of the Cys-substituted mutant channels, we expressed channels with Cys introduced into the (G/S)XS tract in the setting of ␥C547S. Unfortunately, the additional mutation (i.e. ␥C547S) in the presence of a Liddle's mutation decreased functional channel expression to levels that were too low to allow us to perform a typical Cd 2ϩ dose-response experiment. As an alternative, we examined the blocking effects of 0.1 and 1 mM Cd 2ϩ on ␣G587C␤ T ␥C547S and ␣S589C␤ T ␥C547S. We observed that 0.1 mM Cd 2ϩ blocked about half of the current in oocytes expressing ␣G587C␤ T ␥C547S and that 1 mM Cd 2ϩ blocked approximately half of the current in oocytes expressing ␣S589C␤ T ␥C547S. These results indicate that the ␥C547S mutation did not affect the Cd 2ϩ block of the Cys mutants, sug- gesting that the observed Cd 2ϩ block was due to its interaction with the engineered sulfhydryl groups.
Rates of Cd 2ϩ Block-Rates of Cd 2ϩ block are often used to compare the relative Cd 2ϩ accessibility to engineered Cys residues. We measured the rate of 1 or 10 mM Cd 2ϩ block of Na ϩ currents in oocytes expressing channels with a single Cys mu-tant in the (G/S)XS tract using a voltage clamp protocol as described under "Experimental Procedures." Fig. 6 shows representative recordings from these experiments. Time constants () for Cd 2ϩ -dependent inhibition of mutant channels and the calculated rates of Cd 2ϩ block are listed in Table II. In agreement with the relative affinities of the Cd 2ϩ block, Cd 2ϩ inhib- itory rates for mutant channels with a Cys substitution at the first position of the (G/S)XS tract were greater than those observed for channels with a Cys substitution at the third position (␣G587C versus ␣S589C, p Ͻ 0.001; ␤G529C versus ␤S531C, p Ͻ 0.01; and ␥S546C versus ␥S548C, p Ͻ 0.001). Interestingly, the rate of Cd 2ϩ block of the ␣G587C mutant was similar to that of ␤G529C channels (p Ͼ 0.05) and greater than that of the ␥S546C mutant (p Ͻ 0.001), despite the order of  Table I. FIG. 4. Reversibility of the Cd 2؉ block following Cd 2؉ washout and following DTT treatment. A-F, reversibility of the Cd 2ϩ block of Cys mutants following Cd 2ϩ washout. Oocytes were clamped following the current-voltage protocol described under "Experimental Procedures," and whole cell Na ϩ currents at Ϫ100 mV were used to obtain relative currents. The relative currents in the bath solution prior to Cd 2ϩ addition (Control), in the bath solution containing 10 mM CdCl 2 , following Cd 2ϩ washout from the bath solution for 2 min, and in the bath solution perfused with 0.1 mM amiloride (Amil.) are shown as means Ϯ S.E., with the number of observations in parentheses. *, p Ͻ 0.05 compared with control values from paired Student's t test. G and H, reversibility of the Cd 2ϩ block of ␣G587C␤ T ␥ and ␣S589C␤ T ␥ with DTT. Oocytes were continuously clamped at Ϫ100 mV. 1 mM CdCl 2 , 10 mM DTT, and 0.1 mM amiloride were sequentially applied by switching to the appropriate bath solution at the time periods indicated by the bars. Each panel is representative of six independent observations. Cd 2ϩ affinity (␣G587C Ͼ ␤G529C Ϸ ␥S546C). For the mutants with Cys introduced at the third position of the (G/S)XS tract, the rate of Cd 2ϩ block of ␣S589C was greater than those of ␤S531C and ␥S548C (p Ͻ 0.001) and is in agreement with the order of Cd 2ϩ affinity (␣S589C Ͼ ␤S531C Ϸ ␥S548C). The Cd 2ϩ block rate of ␥S548C was moderately higher than that of ␤S531C (p Ͻ 0.01).
External Cd 2ϩ Block of Channels Containing Cys Mutations at the Putative Amiloride-binding Sites-␣Ser 583 , ␤Gly 525 , and ␥Gly 542 have been proposed to form an amiloride-binding site (24). These residues likely line the ENaC conduction pore, as mutations at these sites significantly alter both amiloride sensitivity and single channel conductance, and Cys introductions render the channel sensitive to sulfhydryl-reactive Zn 2ϩ and methanethiosulfonate reagents (5,12,18,24). We examined the Cd 2ϩ -dependent inhibition of ␣S583C␤␥, ␣␤G525C␥, and ␣␤␥G542C. The three mutant channels were blocked by external Cd 2ϩ in a dose-dependent manner ( Fig. 7 and Table I).
␣S583C channels were 7-and 6-fold more sensitive to Cd 2ϩ than the ␤G525C and ␥G542C mutants, respectively (p Ͻ 0.01). The rate of Cd 2ϩ block of ␣S583C channels was higher than that of ␥G542C mutants (p Ͻ 0.01), but similar to that of ␤G525C (p Ͼ 0.05) (Table II).

Effects of Mutations within the Selectivity Filter on the Amiloride Block of mENaC Currents-Amiloride blocks ENaCs
with an inhibitory constant of ϳ100 nM. ␣Ser 583 , ␤Gly 525 , and ␥Gly 542 have been identified as key residues within a putative amiloride-binding site based on changes in the amiloride K i that were observed with point mutations (24) and are four residues away from the (G/S)XS tract. ϳ1000-fold increases in the amiloride K i have been reported with a Cys substitution at either ␤Gly 525 or ␥Gly 542 , although Cys substitutions at selected sites within the pore region of ␣mENaC, including ␣Ser 583 , result in only small increases (Ͻ3.5-fold) in the amiloride K i (13). Mutations of the first residue within the (G/S)XS tract also decrease amiloride sensitivity (14,15). The ␣G587S mutant increases the amiloride K i by 6.8-fold, although ␣G587A slightly decreases the amiloride K i (14). In oocytes expressing ␣G587C␤ T ␥, we observed that the reduction in whole cell current with 1 mM Cd 2ϩ was greater than that observed with 100 M amiloride. These data suggested that the block of ␣G587C␤ T ␥ by 100 M amiloride was incomplete. We therefore examined the response of ␣G587C␤ T ␥ T and ␣S589C␤ T ␥ T to increasing concentrations of amiloride (Fig. 8). The estimated K i for ␣G587C␤ T ␥ T was 3.16 Ϯ 0.24 M (n ϭ 8), and the Hill coefficient was 0.76 Ϯ 0.03 (n ϭ 8). This decrease in amiloride sensitivity was much greater than has been observed with Cys substitutions at other sites within the ␣ pore region. In contrast, the K i and Hill coefficient for ␣S589C␤ T ␥ T were 41.8 Ϯ 6.5 nM (n ϭ 4) and 0.85 Ϯ 0.01 (n ϭ 4), respectively.

DISCUSSION
Our study addresses a key question regarding the structure of the ENaC selectivity filter and the mechanisms of cation selectivity. Do hydroxyl oxygen atoms of the Ser residues at the first and third positions within the (G/S)XS tract point to the pore to coordinate permeant cations (Li ϩ or Na ϩ )? Our results indicate that the engineered sulfhydryl groups within the putative selectivity filter are accessible to externally applied Cd 2ϩ and suggest that hydroxyl groups from wild-type Ser residues within the putative selectivity filter are also accessible to cations traversing the pore. These results are consistent with the

FIG. 5. Effects of external Cd 2؉ on non-Cys mutant ENaCs.
Oocytes were injected with cRNAs for ␣G587A␤ T ␥ (or ␣G587A␤␥ T ), ␣G587S␤ T ␥, or ␣S589A␤ T ␥ (or ␣S589A␤␥ T ) and clamped following the current-voltage protocol described under "Experimental Procedures." A, Cd 2ϩ dose-response curves for ␣G587A, ␣G587S, and ␣S589A. Relative currents are shown. The number of observations was 8, 5, and 7 for ␣G587A, ␣G587S, and ␣S589A, respectively. The results obtained with mutant channels containing either the ␤R564X or ␥R583X Liddle's mutation were identical and were pooled. B-D, reversibility of Cd 2ϩ effects. The relative currents prior to Cd 2ϩ addition (Control), following perfusion with 10 mM CdCl 2 , following Cd 2ϩ washout for 2 min, and following perfusion with 0.1 mM amiloride are shown as means Ϯ S.E., with the numbers of observations in parentheses. The relative currents after Cd 2ϩ washout in B-D were not significantly different from control values (p Ͼ 0.05, paired Student's t test). notion that hydroxyl groups from the Ser residues within the putative selectivity filter contribute to cation binding and therefore to selectivity.
Our conclusions appear to conflict with the pore model proposed by Kellenberger et al. (22), in which Na ϩ within the selectivity filter is coordinated exclusively by four carbonyl oxygen atoms contributed from four Ser residues within the selectivity filter tract, whereas the Ser side chains point to intersubunit interfaces. This hypothesis was primarily based on the positive correlation between the relative changes in cation selectivity and the volumes of the residues introduced at ␣Ser 589 . However, this positive correlation was observed for a

Rates of Cd 2ϩ block
Oocytes were continuously clamped at Ϫ100 mV, and the currents were recorded before (20 s) and after (20 s) application of 1 or 10 mM CdCl 2 . The decline of the inward current from the basal current was fit with a single exponential equation to obtain a time constant ( ). Rates of inhibition were calculated as follows: rate ϭ 1/(⅐C), where C is the concentration of Cd 2ϩ . . Specifically, substitution of ␣Ser 589 with either Gly or Ala reduces the residual volume, but increases the K ϩ /Na ϩ current ratio. Furthermore, channels with a Thr substitution at ␣Ser 589 retain the ability to discriminate Na ϩ and K ϩ (13), although the volume of Thr is ϳ27% greater than that of Ser (23). Nevertheless, molecular sieving is clearly one of the determining factors governing ENaC selectivity (3, 11, 22, 26 -29). However, it is still unclear whether size exclusion is sufficient for ENaCs to achieve ion selectivity. Other factors (such as a high field strength site within the channel pore that might preferentially bind smaller cations such as Li ϩ and Na ϩ ) might influence cation selectivity (26 -28). We speculate that ENaC cation selectivity is achieved through multiple mechanisms, including size exclusion, electrostatic interactions between ions and amino acid residues lining the pore, preferential cation coordination number and geometry, and pore flexibility. We propose that a mixture of carbonyl and hydroxyl oxygen atoms within the ENaC narrow selectivity filter coordinates perme-ating Na ϩ or Li ϩ ions and that the selectivity filter has an asymmetric design.
The third serine residue within the selectivity filter tract is conserved among all ENaC/DEG family members, consistent with a crucial role within the filter. Indeed, ␣Ser 589 is highly sensitive to point mutations. Substitution of ␣Ser 589 with Gly, Ala, Cys, Asp, Asn, Gln, or His results in K ϩ -permeable channels with a reduced discrimination between Na ϩ and K ϩ (11,13,22). Nonfunctional channels were observed when ␣Ser 589 was mutated to Val, Leu, Met, Glu, Lys, Arg, Phe, or Trp (22). The only substitution at ␣Ser 589 that retains K ϩ exclusion is Thr, which also bears a hydroxyl group (13). Although mutations of this Ser residue in ␥ENaC have also been reported to result in K ϩ -permeable channels (12), the Ser at the third position in ␤ENaC does not appear to have a key role in conserving ENaC cation selectivity, as ␤S531A and ␤S531C are not K ϩ -permeable (11,15).
In contrast, the first residue within the three-residue tract of ␤ENaC (Gly) contributes to K ϩ /Na ϩ selectivity, as channels with mutation of ␥Gly 529 to Ser, Cys, or Asp are K ϩ -permeable (14,15). Although we reported that channels with mutations of the first residue within the (G/S)XS tract in ␣ENaC (␣G587C and ␣G587A) are K ϩ -permeable (13) , Kellenberger et al. (14) reported no change in the K ϩ /Na ϩ current ratio in the ␣G587A mutant. In addition, mutations of the first residue in ␥ENaC (Ser 546 in mENaC) are not K ϩ -permeable (12,14). In summary, mutagenesis data suggest that the third Ser in both ␣ENaC and ␥ENaC and the first Gly in ␤ENaC and perhaps also in ␣ENaC are important in restricting K ϩ permeation.
The apparent requirement of a Ser residue at the third position in the selectivity filter of ␣ENaC and ␥ENaC is interesting. Ser is not present in the selectivity signature sequence (DEKA) of voltage-gated Na ϩ channels. However, Ser residues in specific Na ϩ /substrate symporters have been found to have important roles in determining Na ϩ binding affinity (30). Wheat HKT1 is a Na ϩ -coupled K ϩ transporter that is related to fungal Trk and bacterial TrkH K ϩ transporters. These HKT1related transporters are predicted to have four membrane domain-pore region-membrane domain repeats, and all four pore loop-like regions contain a highly conserved Gly that is considered crucial for K ϩ selectivity, analogous to the first Gly in the K ϩ channel selectivity filter sequence GYG (31). Interestingly, Arabidopsis thaliana HKT1 functions as a selective Na ϩ transporter (32). It was recently demonstrated that the Na ϩ -selective phenotype is due to a Ser residue that replaces a Gly residue within the first membrane domain-pore region-membrane domain (33). This is reminiscent of the weaver mouse mutation, where a Gly-to-Ser mutation (G156S) within the selectivity filter of the G protein-gated inward rectifier K ϩ channel GIRK2 results in an increase in GIRK2 Na ϩ permeability (34). These observations suggest that the presence of Ser residues within selectivity filters or Na ϩ -binding sites favors interactions with Na ϩ rather than K ϩ . The retained K ϩ / Na ϩ selectivity of ␣S589T␤␥ suggests that Na ϩ selectivity may be dependent on the presence of hydroxyl groups. In light of our current results regarding the external Cd 2ϩ block of ␣S589C␤ T ␥ and ␣␤ T ␥S546C, we propose that hydroxyl oxygens from selected Ser residues within the (G/S)XS tract participate in coordinating Na ϩ or Li ϩ during ion permeation.
The Cd 2ϩ block of Cys mutant channels was fast, with time constants of Ͻ3.2 s (Table II), which is consistent with the direct interaction of Cd 2ϩ with the ENaC pore. Both carbonyl oxygen and side chain sulfur atoms may provide Cd 2ϩ -binding sites and allow for the Cd 2ϩ block of the channel currents. We believe that the Cd 2ϩ block primarily results from Cd 2ϩ interaction with sulfhydryl groups rather than carbonyl oxygen  Table I. FIG. 8. Mutation ␣G587C reduces amiloride affinity. The amiloride dose-response curves for ␣G587C␤ T ␥ T and ␣S589C␤ T ␥ T are shown. Increasing concentrations of amiloride were sequentially added to the bath solution. Currents measured at Ϫ100 mV were normalized to the basal currents prior to amiloride application to obtain relative currents. The lines are from the best fit of the data with the Hill equation. atoms based on the following observations. First, minimal Cd 2ϩ sensitivity was observed with Ala substitutions of ␣Gly 587 and ␣Ser 589 . In these cases, currents were restored when Cd 2ϩ was simply removed from the bath solution, consistent with weak and reversible binding to hard donors such as oxygens. Second, the Cd 2ϩ block of channels with Cys mutants in the (G/S)XS tract was only partially reversed following washout of Cd 2ϩ from the bath solution, but was nearly completely reversed upon the addition of 10 mM DTT. This is consistent with the predominantly covalent bond characteristics of Cd 2ϩ interaction with free sulfhydryl groups (25).
Channels with a Cys substitution at either the first or third position of the (G/S)XS tract in ␣ENaC showed significantly higher Cd 2ϩ sensitivity than channels with the corresponding Cys substitution in ␤ENaC or ␥ENaC. The Cd 2ϩ sensitivity of channels with a Cys substitution in either ␤ENaC or ␥ENaC was similar. These differences may reflect the presence of more than one sulfhydryl group that could participate in Cd 2ϩ binding when Cys was introduced into ␣ENaC, whereas only one sulfhydryl group would participate in Cd 2ϩ binding when Cys was introduced into either the ␤ or ␥ subunit, assuming a tetrameric ␣ 2 ␤ 1 ␥ 1 architecture (4,5). Alternatively, sulfhydryl groups introduced at homologous sites in different subunits might be located at different depths within the conduction pore, as we previously suggested (15).
For all three subunits, the Cd 2ϩ sensitivities of ENaCs with a Cys substitution at the first position of the (G/S)XS tract were at least 10-fold greater than those observed for channels with a Cys substitution at the third position. Furthermore, the rates of Cd 2ϩ block of channels with a Cys substitution at the first position were also more rapid than those observed with a Cys substitution at the third position. These differences likely reflect the relative depth of the introduced sulfhydryl groups within the channel pore and suggest that the first residue in the (G/S)XS tract is external to the third residue. The putative amiloride-binding site has been proposed to be external to the selectivity filter (24). Channels containing a Cys substitution at the amiloride-binding site (␣S583C␤␥, ␣␤G525C␥, and ␣␤␥G542C) were blocked by external Cd 2ϩ , indicating that the introduced sulfhydryl groups were accessible to Cd 2ϩ . The result is consistent with previous observations that Zn 2ϩ and methanethiosulfonate reagents block these mutant channels (5,12,15,18,24). The greater Cd 2ϩ sensitivities and block rates of the amiloride-binding site mutants compared with those of the selectivity filter mutants are also consistent with the notion that the amiloride-binding site in the ENaC pore is external to the selectivity filter and with a gradually narrowing pore structure as proposed by Snyder et al. (12), Kellenberger et al. (14), and Palmer (29).
Based on our current observations and previous reports (11)(12)(13)(14)(15), we propose that the ENaC selectivity filter has an asymmetric structure that differs from the symmetric selectivity filter present in K ϩ channels. Our working model for an Na ϩ coordination shell within the selectivity filter is shown in Fig.  9B. In contrast to the model proposed by Kellenberger et al. (22), a Na ϩ -binding site at ␣Ser 589 -␤Ser 531 -␥Ser 548 involves four carbonyl oxygens from the four Ser residues (assuming an ␣ 2 ␤ 1 ␥ 1 subunit stoichiometry) plus two hydroxyl oxygens from two ␣Ser 589 residues. This model agrees with the following observations regarding ENaCs and Na ϩ coordination geometry. (i) A sulfhydryl group introduced at ␣Ser 589 is accessible to external Cd 2ϩ , suggesting that the hydroxyl group of ␣Ser 589 is accessible to permeating cations. (ii) K ϩ /Na ϩ discrimination is retained only when ␣Ser 589 is replaced with Thr (13), but not with other amino acids (Gly, Ala, Cys, Asp, Asn, Gln, and His) (11,13,22). These data suggest that the hydroxyl group of ␣Ser 589 has a key role in conferring K ϩ /Na ϩ selectivity. (iii) The side chain of ␤Ser 531 is dispensable with regard to K ϩ /Na ϩ selectivity (11,15). Although channels with an Ala substitution FIG. 9. Working models of the ENaC selectivity filter. A, K ϩ coordination shell in the KcsA K ϩ channel. The K ϩ (red sphere) at the K4 site within the KcsA selectivity filter is coordinated by four carbonyl and four hydroxyl oxygen atoms contributed by a Thr residue (T53) present within each of the four subunits. The coordinates are from the crystal structure of KcsA in 150 mM K ϩ (Protein Data Bank code 1K4C) (20). The coordinating Thr residues are shown in ball-and-stick mode. B, a working model for Na ϩ coordination within the ENaC selectivity filter. The model was generated by mutating the Thr residues in A to Ser residues. The two ␣Ser 589 residues were modeled as D-isomers, whereas ␤Ser 531 and ␥Ser 548 were modeled as L-isomers. A Na ϩ (magenta sphere) is coordinated by four carbonyl oxygen atoms from the four Ser residues and two hydroxyl oxygen atoms from two ␣Ser 589 residues. C and D, models for ␣S589T and ␣S589C, respectively. These models were similarly generated by mutating ␣Ser 589 to the appropriate residue without altering ␤Ser 531 and ␥Ser 548 . Cd 2ϩ is shown in D as an orange sphere. E, a structural model for the ENaC outer pore. The N-terminal portion of the ␣ pore region (␣Val 579 -␣Trp 582 ) was modeled as an ␣ helix that was equivalent to the pore helix in the KcsA structure (19). The other portion (␣Ser 583 -␣Ser 589 ) was modeled as a turn or random coil structure. The secondary structures are shown as green lines, and the selected residues are displayed in stick mode. The doubleheaded arrows indicate hypothesized widths of the pore at the levels of ␣Ser 583 and ␣Ser 589 . F, a model for amiloride binding within the ENaC pore. An amiloride molecule displayed in ball-and-stick mode was placed within the ENaC pore model. We hypothesize that amiloride interacts with ␣Ser 583 and ␣Gly 587 , but not with ␣Ser 589 . All models in this figure were generated with HyperChem 7.1 (Hypercube, Inc., Gainesville, FL). The color codes are red for oxygen, blue for nitrogen, cyan for carbon, white or gray for hydrogen, yellow for sulfur, and magenta for chloride (in the amiloride structure). of ␥Ser 548 retain K ϩ /Na ϩ selectivity, ␣␤␥S548C is K ϩ -permeable (11,12). In addition, the low affinities and rates of Cd 2ϩ block of the ␤S531C and ␥S548C mutants suggest a limited accessibility of the introduced sulfhydryl groups to Cd 2ϩ . (iv) The most common coordination number for Na ϩ in proteins and small molecules is six (35). As the roles of ␤Ser 531 and ␥Ser 548 in conferring ENaC cation selectivity have been addressed by analyzing a limited number of mutants, the possibility that their hydroxyl groups face the filter lumen and are involved in Na ϩ binding cannot be excluded. The major difference between our model and that of Kellenberger et al. (22) is the orientation of the hydroxyl groups of ␣Ser 589 . In our model, the ␣Ser 589 hydroxyl groups are positioned toward the conduction pore and participate in Na ϩ binding. According to the model of Kellenberger et al., these side chains point away from the pore lumen.
Mutagenesis studies suggest that there are both structural similarities and differences between ENaC and K ϩ channel pore regions. Two models of ENaC pore regions have been proposed based on mutation analyses. One model is based on the fold of the KcsA K ϩ channels (13). However, several groups have suggested that this model is unlikely, as it places the putative amiloride-binding site below the selectivity filter. Furthermore, substituted Cys residues at sites immediately following the (G/S)XS tracts are not modified by sulfhydryl-reactive reagents (12,14,15,18). An alternative pore model with a gradually narrowing outer pore extending from the amiloridebinding site (consisting of ␣Ser 583 , ␤Gly 525 , and ␥Gly 542 ) to the selectivity filter is consistent with most mutagenesis studies (3). Our results in this study largely agree with this second model. We propose an updated model for the pore region structure (see Ref. 18 and Fig. 9E). The pore region has two distinct structures that form the outer pore. The N-terminal portions are ␣ helices that are tilted at an angle with respect to the membrane normal, and the C-terminal portions are non-helical. We previously proposed that ␣Ser 580 -␣Ser 583 reside at the transition point between the helical and non-helical structures (18). Our model in Fig. 9E instead illustrates a transition at ␣Ser 583 . The diameter of the pore at ␣Ser 583 is proposed to be ϳ6 Å based on observations on the modifications of ␣S583C␤␥, ␣␤G525C␥, and ␣␤␥G542C by sulfhydryl reagents or thiophilic Zn 2ϩ and Cd 2ϩ (5,12,15,18,24). The diameter of the selectivity filter space at the level of ␣Ser 589 is proposed to have a dimension of ϳ2 Å, estimated from the averaged Na ϩ -O distance of 2.42 Å in protein structures (35).
We also observed that ␣G587C significantly reduced the amiloride block. This change in amiloride affinity was greater than that observed for any other Cys substitution within the pore region of ␣ENaC, including ␣S583C. As some mutations at the first position in ␣ and ␤ subunits reduce the amiloride block (14,15), it is possible that part of the amiloride molecule may interact directly with residues in the selectivity filter. We propose that the positively charged guanidine moiety interacts with the first residue within the (G/S)XS tract of the selectivity filter and that the pyrazine ring interacts with a more external stretch of residues, including ␣Ser 583 -␤Gly 525 -␥Gly 542 (Fig.  9F). The first residue within the (G/S)XS tract may also provide a binding site for Na ϩ , consistent with a previous report suggesting that Na ϩ and amiloride interact with ENaCs at a common site (36). Our model is also consistent with our previ-ous observation that Cys substitutions at multiple sites neighboring ␣Ser 583 or ␤Gly 525 moderately reduce the amiloride sensitivity of mutant ENaCs (13,15).
In summary, we observed that sulfhydryl groups introduced into the putative selectivity filter of ENaCs are exposed to the conduction pore. Our results are consistent with a pore structure in which some of the hydroxyl groups of filter-lining Ser residues (especially ␣Ser 589 ) point to the filter lumen and participate in coordinating Na ϩ or Li ϩ and in determining cation selectivity.