Na+ Inhibits the Epithelial Na+ Channel by Binding to a Site in an Extracellular Acidic Cleft*

Background: External Na+ inhibits ENaC. Results: Mutations centered about a key aspartate in an acidic cleft weakened Na+ inhibition and altered inhibitor selectivity. Conclusion: The acidic cleft hosts an inhibitory Na+ binding site. Significance: The acidic cleft harbors a key Na+ binding site for ENaC and perhaps sites for ligands that regulate other members of the ENaC/degenerin family. The epithelial Na+ channel (ENaC) has a key role in the regulation of extracellular fluid volume and blood pressure. ENaC belongs to a family of ion channels that sense the external environment. These channels have large extracellular regions that are thought to interact with environmental cues, such as Na+, Cl−, protons, proteases, and shear stress, which modulate gating behavior. We sought to determine the molecular mechanism by which ENaC senses high external Na+ concentrations, resulting in an inhibition of channel activity. Both our structural model of an ENaC α subunit and the resolved structure of an acid-sensing ion channel (ASIC1) have conserved acidic pockets in the periphery of the extracellular region of the channel. We hypothesized that these acidic pockets host inhibitory allosteric Na+ binding sites. Through site-directed mutagenesis targeting the acidic pocket, we modified the inhibitory response to external Na+. Mutations at selected sites altered the cation inhibitory preference to favor Li+ or K+ rather than Na+. Channel activity was reduced in response to restraining movement within this region by cross-linking structures across the acidic pocket. Our results suggest that residues within the acidic pocket form an allosteric effector binding site for Na+. Our study supports the hypothesis that an acidic cleft is a key ligand binding locus for ENaC and perhaps other members of the ENaC/degenerin family.

The epithelial Na ϩ channel (ENaC) 2 is a highly Na ϩ -selective heterotrimeric channel whose activity is modified by a variety of extracellular factors (1,2). ENaC-mediated Na ϩ conductance through the apical membrane is rate-limiting for transepithelial Na ϩ absorption in many Na ϩ -transporting epithelia (2). In the distal nephron, this transporter has a crucial role in extracellular fluid volume regulation and is under the control of the volume-regulatory hormones aldosterone and vasopressin (2,3). The role of ENaC in transepithelial Na ϩ transport is broad because the channel is expressed in many transporting epithelia, including the distal nephron, lung airway and alveoli, distal colon, sweat glands, lingual epithelia, and inner ear (4 -6). ENaC subunits are also expressed in vascular smooth muscle and endothelial cells, where they appear to modulate vascular tone (7).
Sensitivity to extracellular factors is a characteristic of members of the ENaC/degenerin family. For ENaC, a high extracellular [Na ϩ ] inhibits the channel through at least two mechanisms. First, external Na ϩ rapidly inhibits ENaC by directly binding to extracellular allosteric effector sites within the channel. This allosteric inhibition, referred to as Na ϩ self-inhibition, reflects a low-affinity (tens of micromolar), pH-sensitive, nonvoltage-sensitive reduction in ENaC open probability (8 -13). This inhibitory effect is also cation-selective because both Na ϩ and Li ϩ inhibit the channel, whereas K ϩ has a modest inhibitory effect at best, and protons activate the channel (9,10). Second, an elevated intracellular Na ϩ concentration inhibits the channel over a slow time course (14,15).
The extracellular domains of members of the ENaC gene family likely host allosteric effector sites that sense factors in the extracellular environment. For example, acid-sensing ion channels (ASICs) are proton-activated. The resolved structure of ASIC1 features an acidic pocket in each subunit that was proposed to contain proton binding sites competent to trigger gating (16). A major challenge in elucidating ENaC gene family function has been the ability to distinguish ligand binding from downstream transduction steps. The crystal structure of ASIC1 suggests several putative proton binding sites, and mutations at select sites affect acid activation of ASIC1 (17)(18)(19). However, one cannot readily distinguish a proton binding site from sites involved in transmitting conformational changes to the channel gate. In this work, we used inhibitor specificity to tackle this challenge.
ENaCs are activated by proteases, which cleave the ␣ and ␥ subunits at specific sites, releasing imbedded inhibitory tracts (1, 20 -23). On the basis of ASIC1 homology and our work that identified the binding site of a peptide derived from the ␣ subunit inhibitory tract, we developed a structural model of the ENaC ␣ subunit (24,25). We found that the ␣ subunit inhibi-tory tract lies adjacent to an acidic region that is homologous to the ASIC1 acidic cleft. We hypothesized that specific residues within the acidic cleft of the ␣ subunit form a Na ϩ binding site required for the Na ϩ self-inhibition response.
Na ϩ self-inhibition is an allosteric response. Channel modifications that affect the apparent strength of Na ϩ self-inhibition could alter channel behavior at different steps in this regulatory process. Experimental manipulations that affect Na ϩ self-inhibition could alter the Na ϩ binding site, transduction steps subsequent to Na ϩ binding, or the stability of the open state relative to the closed state of the channel independent of Na ϩ binding. The fact that ENaC is both regulated by external Na ϩ and conducts Na ϩ complicates our ability to differentiate the Na ϩ binding step from other steps in the Na ϩ self-inhibition mechanism. To define sites within the channel periphery that participate in Na ϩ binding, we studied properties that directly reflect this event. Accordingly, we measured the effect of mutations on the effector or cation specificity of the Na ϩ self-inhibition response. Select mutations within the acidic cleft weakened Na ϩ self-inhibition and resulted in channels that were more effectively inhibited by Li ϩ or K ϩ than by Na ϩ , in contrast to the wild-type channel phenotype, where Na ϩ is the most potent inhibitor. We identified a single titratable residue whose mutation abrogated Na ϩ self-inhibition as well as channel activation by acidic pH. Through cross-linking, we demonstrate that constraining movements of adjacent structures within the acidic cleft reduced channel activity. Our data show that the acidic cleft of the ␣ subunit of ENaC hosts an allosteric effector binding site for Na ϩ and provide strong support for the concept that interactions of external factors with the extracellular region of ENaC have an important role in modulating channel activity.

EXPERIMENTAL PROCEDURES
ENaC Expression and Mutagenesis-Mouse ENaC subunits were mutated and expressed in Xenopus oocytes as described previously (26). All mutations were confirmed by direct sequencing.
Current Measurement-Electrophysiological measurements were performed using a GeneClamp 500B voltage clamp amplifier (Axon Instruments, Foster City, CA) and pClamp 10.2 software (Axon Instruments). Oocytes were placed in a 20-l recording chamber (AutoMate Scientific, Berkeley, CA) with perfusion (3-5 ml/min) controlled by an eight-channel pinch valve system (AutoMate Scientific). For experiments that did not examine the cation specificity of ENaC inhibition, we used a high Na ϩ bath solution (110 mM NaCl, 2 mM KCl, 2 mM CaCl 2 , and 10 mM HEPES (pH 7.4)). We replaced HEPES with MES for lower pH versions of the high-Na ϩ bath solution. To make a low-Na ϩ bath solution, we replaced NaCl with 109 mM NMDG and 1 mM NaCl. For experiments that examined the cation specificity of ENaC inhibition, we used buffers that included K ϩ channel blockers as described previously (9) (100 mM NMDG-Cl, 0.82 mM MgCl 2 , 0.41 mM CaCl 2 , 10 mM NMDG-HEPES, 5 mM BaCl 2 , and 10 mM TEA-Cl for the NMDG ϩ solution and with 100 mM NaCl, 100 mM LiCl, and 100 mM KCl replacing NMDG-Cl for the Na ϩ , Li ϩ , and K ϩ solutions, respectively). Amiloride was dissolved in dimethyl sulfoxide at 100 mM and diluted in bath solution to 10 M for fixed-voltage experiments or 100 M for voltage ramp experiments.
Current-Voltage Curve Analysis-The comparison of ENaC inhibition by extracellular Na ϩ , Li ϩ , and K ϩ is complicated by the fact that ENaC is permeable to these cations by a 1.6:1: Ͻ0.01 (Li ϩ :Na ϩ :K ϩ ) ratio (27). To measure ENaC inhibition by these ions, we determined the relative Na ϩ permeability (P Na ) in the presence of each ion from the amiloride-sensitive steadystate current-voltage (I-V) curves by fitting each curve to the Goldman-Hodgkin-Katz model as described previously (9) with some modifications. When Na ϩ was the only permeant ion present, we fit I-V curves to equation 1, where [Li] o is extracellular [Li ϩ ] and P Li /P Na was assumed to equal 1.6 on the basis of the ion selectivity of the pore (9,27). We assumed that the mutations examined here, far from the pore, did not affect the value of P Li /P Na . We also assumed that [Na] i remained constant throughout the experiment and that intracellular [Li ϩ ] was zero. To minimize changes in [Na] i during the experiment, we clamped oocytes at Ϫ20 mV between voltage ramps. We collected five amiloride-sensitive I-V curves for each oocyte, each using bath solutions with different cations (NMDG ϩ , Na ϩ , Li ϩ , K ϩ , and NMDG ϩ again) for 1 min, followed by the same solution supplemented with 100 M amiloride for 20 s. We fit the I-V curves from each oocyte simultaneously using a global fitting routine and Igor Pro 6 (Wavemetrics, Oswego, OR). This allowed us to use data under all conditions to determine a global value for [Na] i while independently determining local values for P Na under each condition. We assigned values for the global parameters F, R, and T and the local parameters [Na] o or [Li] o ⅐P Li /P Na . E was the independent variable. Each fit produced five P Na values, one for each condition. We normalized each P Na value to a corrected P Na determined with bath NMDG ϩ at the beginning and end of the experiment to account for channel rundown, which we assumed was linear over the course of the experiment. Statistical Analyses-Values presented are mean Ϯ S.E. We performed multiple group comparisons using analysis of variance followed by Tukey post hoc test. We adjusted the significance threshold from p ϭ 0.05 for multiple comparisons using the Bonferroni correction. We performed pairwise compari-sons using Student's t test. We considered p values of less than 0.05 significant.

Protons Activate Mouse ENaC-Collier and Snyder (10) and
Collier et al. (28) reported previously that human ENaC is activated by extracellular acidification with an apparent pH 50 of activation of 6.5. They found that rat ENaC was also activated by acid but at a lower pH (28). Because acidification was associated with reduced Na ϩ self-inhibition, they suggested that extracellular acidification activated ENaC by relieving Na ϩ selfinhibition. We examined the effect of extracellular acidification on mouse ENaC expressed in Xenopus oocytes by two-electrode voltage clamp (Fig. 1A). Although modest acidification (to pH 6.5) did not alter channel activity (see below), lowering the extracellular pH from 7.4 to 4.7 stimulated currents by 44 Ϯ 3% (n ϭ 6, p Ͻ 0.05 using paired Student's t test). This is similar to the increase observed for human ENaC by lowering the pH from 7.4 to 6.5 (10). When currents are measured under voltage clamp conditions, currents decay over time. We found that the rate of decay of ENaC currents increased from 3.4 Ϯ 0.9%/min to 10.3 Ϯ 0.5%/min with acidification to pH 4.7 (p Ͻ 0.05 using paired Student's t test). Channel activation by reducing extracellular pH to the range that should titrate acidic residues suggested that residues in the acidic cleft could be involved in Na ϩ sensing.
Mutation of Sites in the Acidic Cleft Affect Na ϩ Selfinhibition-Our model of the ␣ subunit features an acidic region at the juncture of the finger, ␤-ball, and thumb domains (Fig. 1C), which has the appearance of an acidic cleft (24). A homologous region in ASIC1 has been proposed to play a role in proton sensing (16). We hypothesized that residues in the vicinity of the acidic cleft of ENaC subunits coordinate Na ϩ , the first step in the Na ϩ self-inhibition response.
To examine whether mutations of residues in the acidic cleft weakened Na ϩ self-inhibition, we performed a Trp mutagenesis screen of acidic cleft Asp and Glu residues (Fig. 1C, red spheres) and measured the effect of mutation on Na ϩ self-inhibition. Wild-type or mutant ENaCs were expressed in Xenopus laevis oocytes, and ENaC currents were measured by two-electrode voltage clamp. Because ENaC Na ϩ self-inhibition manifests slowly relative to bath solution exchange, acutely raising the bath [Na ϩ ] from a low concentration (where the channel open probability (P o ) is high) results in a peak inward current that subsequently declines (8,11), reflecting Na ϩ binding events at extracellular allosteric site(s) that are transduced to reduce channel P o (i.e. Na ϩ self-inhibition) ( Fig. 1B) (12,13,29). Of the acidic sites tested, we found that the Trp mutation at ␣Asp-176, ␣Glu-362, and ␣Asp-365 reduced Na ϩ self-inhibition (Fig. 1E, red columns). The latter two lie at the beginning of a loop connecting the sixth and seventh ␤ strands ( Fig. 1C; ␤6-␤7 loop), whereas ␣Asp-176 on the ␣1 helix interacts with the ␤6-␤7 loop ( Fig. 2A).
We hypothesized that ␤6-␤7 loop residues participate in Na ϩ binding. We performed Trp scanning mutagenesis of the residues lining the ␤6-␤7 loop and determined the effects of individual mutations on Na ϩ self-inhibition (Fig. 1E). In our model, the ␤6-␤7 loop extends from the ␤-ball toward helix ␣1 of the finger domain, with ␣Leu-369 at the apex of the loop interacting with ␣Asp-176 from helix ␣1 of the finger domain (Figs. 1C and 2A). We found that a Trp mutation at multiple residues flanking ␣Leu-369 reduced Na ϩ self-inhibition, whereas ␣L369W enhanced Na ϩ self-inhibition. Mutations of acidic cleft residues affect Na ؉ self-inhibition. A, acidification activates mouse ENaC. Recordings were normalized to the average current prior to acidification. The average of six normalized current tracings is shown, with periodic error bars indicating the mean Ϯ S.E. The average starting current was 1.8 Ϯ 0.4 A. B, representative recording of the Na ϩ self-inhibition response of wild-type mouse ENaC (Ϫ100 mV holding potential). Bathing solutions and the addition of amiloride (10 M) are noted. I Peak and I SS currents following the transition to a high [Na ϩ ] solution are also noted. C, model of the acidic cleft. Acidic residues (red spheres) and the ␤6-␤7 loop (blue) are highlighted. An imbedded inhibitory tract from which an inhibitory peptide can be derived (␣-8, pink) is located near the acidic cleft. D, sequence alignment of the ␤6-␤7 loop region. ␣Asp-365 and ASIC1 residues implicated in the response to H ϩ are highlighted (red boxes). E, Na ϩ selfinhibition of the Trp mutants is reported as 100 ϫ [1 Ϫ (I SS Ϫ I amiloride )/ (I Peak Ϫ I amiloride )]. Trp mutation at sites flanking ␣Leu-369 strongly reduced Na ϩ self-inhibition, including at acidic residue ␣Asp-365. *, p Ͻ 0.002 versus the wild type (n ϭ 6 -15 for mutants and 115 for the wild type). Model positions of mutations are noted.
If Na ϩ binding to residues in the ␤6-␤7 loop results in a decrease in channel activity, then the conformation of the ␤6-␤7 loop must be linked to the conformation of the channel gate. It follows that constraining movement of the loop should alter channel activity. We used a cross-linking approach to test this hypothesis. Our model suggested that ␣Leu-369 at the apex of the ␤6-␤7 loop interacts with ␣Asp-176 from helix ␣1 of the finger domain ( Fig. 2A). We introduced Cys at ␣Leu-369 in the ␤6-␤7 loop and, simultaneously, to ␣Asp-176 in helix ␣1 of the finger domain. Because ␣L369C and ␣D176C are predicted to be in close proximity, we examined whether these introduced Cys residues formed a natural disulfide bond, linking the ␤6-␤7 loop with the ␣1 helix. DTT addition activated ENaCbearing Cys residues at both sites (Fig. 2, B and C). This activation required the presence of both introduced Cys residues because DTT did not affect the wild type or channels with a single Cys substitution. The DTT-treated double mutant had modestly reduced Na ϩ self-inhibition compared with DTTtreated wild-type channels (22.8 Ϯ 0.5% versus 38 Ϯ 2%, n ϭ 4, p Ͻ 0.05), suggesting that the introduced cysteines residues, per se, modestly altered channel function. These data suggest that conformational changes that involve the ␤6-␤7 loop affect channel activity.
Residues in the Acidic Cleft Determine the Cation Specificity of Inhibition-To identify residues involved in the Na ϩ binding step, we measured the effect of specific mutations on the ligand specificity of Na ϩ self-inhibition. As noted above, mutations that alter the Na ϩ self-inhibition response could reflect changes in the Na ϩ binding step, transduction steps subsequent to Na ϩ binding, or the relative stability of the open  A, ENaC expressing oocytes were perfused for 1 min with each solution, after which currents were determined at varying holding potentials (Ϫ140 to 60 mV in 20-mV steps, 0.5 s/step). Amiloride (amil)-supplemented solutions were then perfused for 20 s, followed by measurement of currents at the different holding potentials. B, resultant I-V curves were fit with equation 1 or 2 to derive P Na values in each solution (red, initial NMDG ϩ ; green, Na ϩ ; blue, Li ϩ ; black, K ϩ ; orange, final NMDG ϩ ). Measurements in NMDG ϩ were taken at the beginning and end of each experiment to estimate channel rundown (ϳ2-7%/min), which was used to adjust P Na values for channel rundown. C, correlation between Na ϩ self-inhibition measured under a constant voltage clamp (Fig. 1B) and the ratio of Na ϩ permeabilities in Na ϩ and NMDG ϩ . D, ratio of Na ϩ permeabilities measured in different bath solutions: Li ϩ versus Na ϩ (top panel) and K ϩ versus Na ϩ (bottom panel). *, p Ͻ 0.003 versus wild type (n ϭ 40 for the wild type and 5-12 for mutants). and/or closed states. We reasoned that the ability to discriminate Na ϩ and Li ϩ from other cations reflects ligand binding.
To compare the inhibition of ENaC by Na ϩ , Li ϩ , and K ϩ , we examined steady-state Na ϩ permeability in the presence of different extracellular cations using a method adapted from Bize and Horisberger (9). Amiloride sensitive I-V curves were obtained at steady state in oocytes sequentially perfused with buffers containing NMDG ϩ , Na ϩ , Li ϩ , K ϩ , and then NMDG ϩ again as the predominant cation (Figs. 3, A and B). We then fit the I-V curves to equations derived from the Goldman-Hodgkin-Katz model of ion flux to determine the relative permeability of ENaC to Na ϩ (P Na ) under each condition (Fig. 3B and equations 1 and 2, see "Experimental Procedures"). Reductions in P Na in Na ϩ , Li ϩ , or K ϩ relative to P Na in NMDG ϩ reflect ENaC inhibition by these effectors. We assumed that the Li ϩ / Na ϩ permeability ratio (P Li /P Na , i.e. the pore selectivity) is 1.6 on the basis of published reports (27,30) and is not affected by these mutations far from the pore. We assumed that the K ϩ permeability is 0. We also assumed that intracellular [Na ϩ ] ([Na] i ) remains steady through the course of the experiment. With these assumptions, fitting each curve to the Goldman-Hodgkin-Katz model depends on two variables, P Na and [Na] i , which have little interdependency for curve fitting purposes (Fig. 3B). We corrected for channel rundown using P Na measurements in NMDG ϩ at the beginning and end of the experiment and assuming a linear rundown.
The extent of Na ϩ inhibition measured by this method correlated with Na ϩ self-inhibition measured by rapidly switching solutions from NMDG ϩ to Na ϩ (Fig. 1B). Na ϩ self-inhibition is given by the ratio of P Na with extracellular Na ϩ to P Na with extracellular NMDG ϩ (P Na (Na)/P Na (NMDG)). We plotted the results for the wild type, ␣D176W, and the ␤6-␤7 loop Trp mutants (Fig. 3C). The fitted line gave a Pearson's correlation coefficient of Ϫ0.75 (p Ͻ 0.005). Notably, the fitted line is shifted from its theoretical position (the diagonal from (0, 100) to (1, 0)) toward the origin. This may reflect, in part, an underestimation of the peak current (e.g. Fig. 1B), which depends upon the efficiency of solution exchange and the kinetics of channel inhibition.
Wild-type mouse ENaC showed a slightly different effector specificity for channel inhibition (Na ϩ Ͼ Li ϩ Ͼ Ͼ K ϩ , Fig. 3D) than what has been reported for human ENaC (Na ϩ Ϸ Li ϩ Ͼ K ϩ ) (9). Trp substitutions at sites toward the beginning or end of the ␤6-␤7 loop changed this order. For ␣Q363W, ␣D365W, and ␣V373W, Li ϩ was a more potent inhibitor than Na ϩ , whereas K ϩ and Na ϩ had similar inhibitory effects. The mutants ␣I367W, ␣S371W, and ␣T374W exhibited weakened Na ϩ inhibition relative to K ϩ , whereas ␣L370W exhibited weakened Li ϩ inhibition relative to Na ϩ . Trp mutation at other sites neighboring ␣Gln-363 and ␣Val-373 showed trends toward changing inhibitor specificity but were not significantly different from the wild type.
To corroborate these results, we measured effector specificity using a second method that adapted a fixed-voltage protocol for measuring Na ϩ self-inhibition (Fig. 1B). First we examined Na ϩ inhibition by rapidly switching from an NMDG ϩ solution to a Na ϩ solution while voltage-clamping the oocyte at Ϫ100 mV (Fig. 4). As described above, this leads to a rapid inward peak current (I Peak ) that declines to a steady-state level (I SS ). The fall from I Peak to I SS reflects Na ϩ self-inhibition. Because ENaC also conducts Li ϩ , we measured Li ϩ inhibition in a similar manner by rapidly switching from an NMDG ϩ solution to a Li ϩ solution. Because ENaC does not appreciably conduct K ϩ , we measured K ϩ inhibition indirectly by examining the effect of K ϩ pretreatment on Na ϩ inhibition. Instead of initially bathing the oocyte in an NMDG ϩ solution, where ENaC P o is high (12), we used a K ϩ solution. If channel inhibition by K ϩ is much weaker than by Na ϩ , switching from NMDG ϩ or K ϩ to Na ϩ should produce similar Na ϩ self-inhibition responses. This is what we observed for wild-type channels (Fig. 4). If channel inhibition by K ϩ and Na ϩ are similar, then pretreatment with K ϩ should attenuate the inhibitory response to Na ϩ , as we observed with the ␣V373W and ␣D365W mutants (Fig. 4). If K ϩ inhibits an ENaC mutant more effectively than Na ϩ , then switching from a K ϩ to an Na ϩ bath should induce a rise in current that represents Na ϩ relief of K ϩ inhibition. This was seen with the ␣Q363W mutant (Fig. 4). Using this fixed-voltage protocol, we also found that Na ϩ was a more effective inhibitor than Li ϩ of wild-type mouse ENaC. ␣N364W exhibited a pattern similar to the wild type. For ␣V373W and ␣D365W, this pattern was reversed. For ␣Q363W, Li ϩ and Na ϩ were similarly effective inhibitors. For the wild type and ␣N364W, K ϩ pretreatment did not alter Na ϩ self-inhibition when compared with NMDG ϩ pretreatment. Sites Adjacent to the ␤6-␤7 Loop Affect Inhibitor Specificity-Mutations in the ␤6-␤7 loop affect the cation selectivity of channel inhibition. Our model of a hypothetical ENaC ␣ trimer suggested that adjacent sites could coordinate a bound ion in concert with sites in the ␤6-␤7 loop (Fig. 5A). ␣Asn-161 is at the end of a highly conserved sequence in the ␤2-␣1 loop connecting the ␤ ball and the finger domain in the ␣ subunit and lies in proximity to the ␤6-␤7 loop (Fig. 5B). The first Ser within a WPSXXS motif in the knuckle domain of a neighboring subunit lies across the plane of the ␣ subunit ␤6-␤7 loop (Fig. 5C). According to recent reports, the neighboring knuckle domain is part of the ␤ subunit (31,32). Notably, a polymorphism of the Trp residue in the WPSXXS motif of the human ␣ subunit (␣W493R) was associated with a loss of Na ϩ self-inhibition and an increased channel P o (33). We mutated ␣Asn-161 to Ala and ␤Ser-462 to Cys and assessed the resultant inhibitor specificity. We compared P Na values derived from steady state I-V curves using different perfusion buffers as described above. Both ␣N161A and ␤S462C weakened Na ϩ inhibition but did not affect Li ϩ or K ϩ inhibition (Fig. 5D). We corroborated these results for ␣N161A using the fixed-voltage protocol for measuring self-inhibition induced by Na ϩ or Li ϩ (Fig. 5E), which showed a pattern of inhibition similar to ␣D365W and ␣V373W and different from wild-type ENaC (Fig. 4).
Protonation of ␣Asp-365 Activates ENaC-Acidification of the extracellular solution increased mouse ENaC currents (Fig.  1A). We hypothesized that protonation of a Na ϩ -coordinating residue weakens Na ϩ binding, diminishing Na ϩ self-inhibition and activating the channel. To identify titratable residues responsible for this activation, we performed titrations of wild-type channels and three mutants: ␣D176W, ␣E362W, and ␣D365W (Fig. 6B). We performed each experiment as described for Fig. 1A and measured the acute amiloride-sensitive change in current upon acidification from pH 7.4. Although each mutant blunts Na ϩ self-inhibition (Fig. 1), we observed that only ␣D365W reversed the effect of acidification on ENaC currents, causing currents to modestly fall instead of rise (Figs. 1A and 6, A and B). Channels bearing ␣D176W or ␣E362W responded to acidification similarly to wild-type. For all groups tested, effects at pH 5.5 were intermediate to effects at pH 4.7. In contrast to human ENaC (10), effects at pH 6.5 were minor for all groups tested.
We also examined the effect of acidification on the Na ϩ selfinhibition response. At pH 4.7, the wild type and the three Trp mutants exhibited low Na ϩ self-inhibition responses (Fig. 6, C  and D). These data, taken together, suggest that reducing the pH to 4.7 or altering the Asp group at ␣Asp-365 removes an important coordinating group from one of the putative Na ϩ FIGURE 5. Specific conserved sites on structures near ␣D365 weaken Na ؉ inhibition. A, structures near the ␤6-␤7 loop in an ENaC ␣ trimer model, including the ␤2-␣1 loop of the ␣ subunit and the beginning of helix ␣6 of the neighboring ␤ subunit. B and C, sequence alignments of the ␤2-␣1 loop (B) and the ␤10-␣6 loop (C), with sites of interest indicated by arrows. D, P Na in various bath solutions (X) was determined as described for Fig. 3. P Na values where Na ϩ , Li ϩ , or K ϩ were the predominant cation were normalized to P Na determined in NMDG ϩ . *, p Ͻ 0.003 versus wild (n ϭ 40 for the wild type and 8 for mutants). Values for the wild type are indicated by dashed lines. E, representative experiment examining the effector specificity of inhibition of ␣N161A under a constant voltage clamp. Quantification of eight experiments gave values of 10.9 Ϯ 0.9% for Na ϩ inhibition following NMDG ϩ pretreatment, 25 Ϯ 1% for Li ϩ inhibition, and 7 Ϯ 1% for Na ϩ inhibition following K ϩ pretreatment. Comparable wild type values were 36 Ϯ 3%, 19 Ϯ 2%, and 34 Ϯ 3%, respectively (n ϭ 6). Na ϩ inhibition measurements following NMDG ϩ or K ϩ perfusion was weaker in both cases for ␣N161A compared with the wild type (p Ͻ 0.05, Student's t test). K ϩ pretreatment did not affect Na ϩ inhibition of either the wild type or ␣N161A compared with NMDG ϩ pretreatment (p ϭ not significant, paired Student's t test). In contrast to the wild type, Li ϩ inhibition was greater than Na ϩ inhibition for ␣N161A channels (p Ͻ 0.05, paired Student's t test).
binding sites of ENaC. To further address this hypothesis, we examined the response of the channel to acidification after introducing conservative Glu or Asn substitutions at ␣Asp-365 or after reversing the charge by Lys substitution. All three ␣Asp-365 mutants (Glu, Asn, and Lys) prevented channel activation by acidification, echoing our result with ␣D365W (Fig.  6B). We also examined the effect of these mutations on Na ϩ , Li ϩ , or K ϩ inhibition at steady state using the voltage ramp method (Fig. 3). We found that each mutant specifically weakened Na ϩ inhibition (Fig. 6E). Our data suggest that the specific moiety at position ␣365, rather than charge per se, is needed to confer "full" sensitivity to external Na ϩ .

DISCUSSION
ENaCs are members of a family of ion channels that respond to factors in the extracellular environment. One of the external factors that regulate ENaC is Na ϩ . There are multiple lines of evidence supporting the notion that Na ϩ binding to one or more effector sites within the extracellular domains results in the inhibition of channel activity in an allosteric manner (1,12,13,29,34). Given that allosteric transitions are required for this process, it is not surprising that many substitutions throughout the extracellular regions of ENaC subunits altered the Na ϩ selfinhibition response. For example, we and others have reported mutations that weaken Na ϩ self-inhibition in the palm domain, thumb domain, ␤ ball-thumb domain interface, knuckle domain, finger domain, and subunit interfaces and among the Cys involved in disulfide bridges (13,25,28,31,33,(35)(36)(37)(38)(39). Furthermore, it is not surprising that mutations in transmembrane helical residues that interfere with pore closure and increase channel P o (e.g. ␤S518K) significantly weaken Na ϩ self-inhibition (38,40). Likewise, other maneuvers that increase channel P o , e.g. Zn 2ϩ addition (41) or Cl Ϫ removal (42), weaken apparent Na ϩ self-inhibition. In summary, on the basis of previous studies, it has been difficult to discern whether specific mutations that change Na ϩ self-inhibition do so by altering Na ϩ binding, downstream transduction steps, or the baseline stability of the open or closed states of the channel.
We postulated that Na ϩ binds to a site defined, in part, by acidic residues, consistent with Na ϩ -bound protein structures (43,44) and the predominance of oxygen atoms coordinating Na ϩ in model small molecules (45). We also reasoned that the preference of the channel for Na ϩ over other cations as an inhibitory effector originates from the binding site(s) rather than from subsequent transduction steps. On these premises, we tested the ability of ␣ subunit acidic cleft mutants to both affect Na ϩ self-inhibition and alter the relative inhibitory efficacy of Na ϩ , Li ϩ , and K ϩ . Mutation of ␣Asp-365 or neighboring residues in our ␣ subunit model weakened inhibition by external Na ϩ and altered the cation selectivity of channel inhibition. Even conservative ␣Asp-365 mutations (Glu or Asn) had this effect. ␣Asp-365 substitutions also prevented channel activation by acidification regardless of whether a conservative or non-conservative substitution was used. These results support the notion that ␣Asp-365 has a role in Na ϩ binding and the subsequent reduction in channel P o , and that acidification activates ENaC by protonating ␣Asp-365 and disrupting Na ϩ binding at an effector site that includes ␣Asp-365.
Our results also suggest that there are multiple effector sites for Na ϩ self-inhibition. All mutants tested at or near ␣Asp-365 retained at least a small Na ϩ self-inhibition response. Acidifi-FIGURE 6. ␣Asp-365 mutants lack acid-dependent channel activation. A, acidification inhibits ␣D365W currents. B, effects of pH titration on the wild type and select mutants of acidic residues in the acidic cleft. Each experiment was performed as in Figs. 1A and 6A while varying the final pH. The acute amiloridesensitive change in current upon acidification to pH 6.5, 5.5, or 4.7 was quantified. For experiments in which currents increased as a result of acidification, we used the maximal increase in current, which occurred at 10 Ϯ 1 s after initiating solution exchange (Fig. 1A). For experiments in which the current decreased as a result of acidification, there was no transient nadir because both the fast and slow phases of the change affected the current in the same direction (A). For quantification of the acute effect of changing pH on current, we used the current value at 10 s after initiating solution exchange. *, p Ͻ 0.01 versus the wild type at pH 5.5; #, p Ͻ 0.005 versus the wild type at pH 4.7. For all groups, n ϭ 5-8. C, acidification weakens Na ϩ self-inhibition. Shown are representative current recordings of oocytes expressing the wild type or ␣ subunit mutants and voltage-clamped at Ϫ100 mV. Solutions were exchanged rapidly as indicated. D, Na ϩ self-inhibition was quantified at pH 4.7 as for Fig. 1E. No significant differences were detected (n ϭ 5-8). E, experiments were performed as described for Fig.  3. P Na values where Na ϩ , Li ϩ , or K ϩ were the principal cation were normalized to P Na determined in NMDG ϩ . *, p Ͻ 0.003 versus the wild type (n ϭ 40 for the wild type and 5-8 for mutants).
cation had no effect on the Na ϩ self-inhibition response of ␣D365W, but reduced the response of the wild type and other mutants to similarly low but measurable values (compare Figs. 1E and 6D). We propose that there are additional Na ϩ effector sites. Because ␣D365W has a small but measurable Na ϩ selfinhibition response, at least one of the additional effector sites does not depend on residues titratable in the pH 7.4 -4.7 range. Accordingly, conservative mutations of the ␤ and ␥ sites that correspond to ␣Asp-365 (␤D302N and ␥E322Q) exhibited Na ϩ self-inhibition similar to the wild type (␤D302N, 33 Ϯ 4%; ␥E322Q, 27 Ϯ 3%; wild type, 32 Ϯ 2%; n ϭ 5-6; p ϭ not significant).
Other members of the ENaC/Deg family have an Asp or Glu residue at the position equivalent to mouse ␣Asp-365 (e.g. Glu-229 in chicken ASIC1, Fig. 1D). This residue lies at the beginning of the ␤6-␤7 loop, which connects the ␤-ball and palm domains (1,16). Electrostatic calculations for the ASIC1 Glu (equivalent to ␣Asp-365) predicted a pK a of 7.4, a prerequisite for proton sensing in the physiological range (17). However, mutating this residue in ASIC1 to Gln had little effect on the pH sensitivity of channel activation (17,18). The ␦ subunit of ENaC replaces the ␣ subunit in the channel complex expressed in some tissues of several species, including humans (but not rodents). Interestingly, human ␦␤␥ channels are poorly inhibited by external Na ϩ (46). The ␦ subunit has a Pro residue at the equivalent site to ␣Asp-365. However, mutating ␦Pro-314 to Asp did not sensitize human ␦␤␥ ENaC to external Na ϩ (data not shown). This observation suggests that there are residues in addition to ␣Asp-365 that are needed to form a Na ϩ binding site in the ␣ subunit.
The ␤6-␤7 loop of ASIC1 interacts with the finger and thumb domains and is six residues longer than its ENaC subunit counterparts (Fig. 1D). In our ␣ subunit model, the shorter ␤6-␤7 loop approaches the ␣1 helix in the finger domain but does not abut thumb domain residues (Fig. 1C) (24). Our results demonstrating that ␣D176C in the ␣1 helix and ␣L369C in the ␤6-␤7 loop form a disulfide bond suggest that these structures are indeed in close proximity and that stabilizing interactions between the ␣1 helix and the ␤6-␤7 loop is inhibitory. The ASIC1 ␤6-␤7 loop includes Asp-238 and Glu-239 (cASIC1 numbering, Fig. 1D) that pair with thumb domain residues Asp-350 and Asp-346, respectively. These thumb domain residues have been proposed to sense protons (16). Krauson et al. (19) showed that mutating Asp-346 in mouse ASIC1a results in a biphasic pH activation curve, suggesting that this mutation specifically weakened of one of multiple proton sensing sites. ASIC1 residues Asp-238 and Glu-239 and ENaC residues ␣371-373 are located in a similar region of the ␤6-␤7 loop (Fig.  1D), where ␣ subunit Trp substitutions weaken Na ϩ self-inhibition and alter the cation selectivity of this process (Figs. 1, 3,  and 4). Baconguis and Gouaux (47) provided additional evidence for the role of the ␤6-␤7 loop in modulating channel gating in response to external factors. They found that psalmotoxin, an ASIC1 inhibitor, extends an Arg-rich hairpin into the acidic pocket and forms hydrogen bonds with the ␤6-␤7 loop backbone (47). These previous observations and our current findings, taken together, suggest that the ␤6-␤7 loop has evolved within members of the ENaC/degenerin family members to facilitate the sensing of specific factors.