Extracellular intersubunit interactions modulate epithelial Na+ channel gating

Epithelial Na+ channels (ENaCs) and related channels have large extracellular domains where specific factors interact and induce conformational changes, leading to altered channel activity. However, extracellular structural transitions associated with changes in ENaC activity are not well defined. Using crosslinking and two-electrode voltage clamp in Xenopus oocytes, we identified several pairs of functional intersubunit contacts where mouse ENaC activity was modulated by inducing or breaking a disulfide bond between introduced Cys residues. Specifically, crosslinking E499C in the β-subunit palm domain and N510C in the α-subunit palm domain activated ENaC, whereas crosslinking βE499C with αQ441C in the α-subunit thumb domain inhibited ENaC. We determined that bridging βE499C to αN510C or αQ441C altered the Na+ self-inhibition response via distinct mechanisms. Similar to bridging βE499C and αQ441C, we found that crosslinking palm domain αE557C with thumb domain γQ398C strongly inhibited ENaC activity. In conclusion, we propose that certain residues at specific subunit interfaces form microswitches that convey a conformational wave during ENaC gating and its regulation.

Epithelial Na + channels (ENaCs) and related channels have large extracellular domains where specific factors interact and induce conformational changes, leading to altered channel activity. However, extracellular structural transitions associated with changes in ENaC activity are not well defined. Using crosslinking and two-electrode voltage clamp in Xenopus oocytes, we identified several pairs of functional intersubunit contacts where mouse ENaC activity was modulated by inducing or breaking a disulfide bond between introduced Cys residues. Specifically, crosslinking E499C in the β-subunit palm domain and N510C in the α-subunit palm domain activated ENaC, whereas crosslinking βE499C with αQ441C in the α-subunit thumb domain inhibited ENaC. We determined that bridging βE499C to αN510C or αQ441C altered the Na + selfinhibition response via distinct mechanisms. Similar to bridging βE499C and αQ441C, we found that crosslinking palm domain αE557C with thumb domain γQ398C strongly inhibited ENaC activity. In conclusion, we propose that certain residues at specific subunit interfaces form microswitches that convey a conformational wave during ENaC gating and its regulation.
ENaCs are members of a family of ion channels that are formed by subunits that share common structural features, including large well-organized extracellular domains. In the case of ENaC, heterotrimeric channels are formed by α (or δ), β, and γ subunits. Resolved structures of extracellular domains of ENaC and a related acid-sensing ion channel 1 (ASIC1) have been described as an outstretched hand holding a ball, with discrete regions formed by β strands or α helices, including the finger, thumb, knuckle, palm, and β ball domain (12,13). Factors in the extracellular environment, including specific ions, peptides, and proteins, interact at sites within ENaC and related family members and modulate channel gating. In the case of ENaC, channel activity is regulated by extracellular Na + , Cl − , protons, metals, proteases, and sheer stress (3,(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28). For example, Na + interacts at a site within the extracellular region of the α subunit of ENaC, resulting in conformation changes that eventually reduce channel open probability, a phenomenon referred to as Na + self-inhibition (3,15,29). However, the mechanistic details how these extracellular regulators impact ENaC gating remain elusive.
A five-residue track connecting the palm domain β11 and β12 strands, referred to as the β11-β12 linker, has been shown to have a role in ASIC1 desensitization (30)(31)(32)(33)(34)(35)(36). Specifically, structural studies suggest that a hydrophobic residue (L414) and an adjacent hydrophilic residue (N415) in this linker positions reorient toward the central vestibule in desensitized channels, compared with closed channels. This structural reorientation has been proposed to function as a molecular clutch, decoupling upper and lower parts of the extracellular domain, and facilitating a transition to the desensitized state (32). These structural transitions within the ASIC1 β11-β12 linker raised the possibility that this linker within ENaC subunits may also undergo a conformational change in association with gating transitions. Indeed, we previously suggested that L511 in human γ ENaC, homologous to L414 in ASIC1, interacts with an adjacent subunit through hydrophobic contacts. We speculated that the gain-of-function γL511Q mutation disrupted the hydrophobic interaction by introducing a hydrophilic side chain, facilitating a separation of the palm domain of γ subunit and the thumb domain of an adjacent subunit and stabilizing the channel in an open state (37). In support of this hypothesis, a recent ASIC1a study demonstrated that a hydrophobic patch stabilized L414 side chain when the channel is in a closed state and where structural transitions were deemed necessary for channel desensitization (34).
The five-residue track connecting the palm domain β11-β12 strands is conserved among ENaC subunits (Fig. 1D). A sole hydrophobic residue (Leu or Phe) is flanked by four hydrophilic residues including a negatively charged residue preceding human γL511 (mouse γL517). Mutations of this acidic residue in the α or γ subunit alter ENaC activity, with changes in either the Na + self-inhibition response or the response to extracellular Zn 2+ (38)(39)(40). We hypothesized that acidic residues in β11-β12 interact with polar residues of an adjacent subunit at the subunit interface and that these intersubunit interactions have important roles in the regulation of ENaC gating. To test this hypothesis, we engineered Cys substitutions in the β11-β12 linker and at sites in adjacent subunits that are predicted to be in close proximity. Channels were expressed in Xenopus oocytes that were treated with either oxidizing agents to induce disulfide bridges or reducing agents to break these bridges. We identified several pairs of introduced Cys residues where ENaC currents were altered in response to a reducing or oxidizing reagent.

Mouse ENaC model and its α/β subunit interface
Based on current knowledge of ENaC and ASIC structurefunction relationships, we hypothesized that specific extracellular interdomain interactions involving β11-β12 linkers and adjacent sites located at a subunit interface have important roles in modulating ENaC gating. To test this hypothesis, we introduced Cys residues at specific sites within ENaC subunits and exposed expressed channels to either oxidizing agents to induce disulfide bridges or reducing agents to break disulfide bonds.  Figure 1. Acidic residues of the β11-β12 linkers reside at subunit interfaces. A and B, locations of the acidic residues of the β11-β12 linkers in the mouse ENaC structural model. The trimeric model was previously built (41) and displayed by PyMol 2.4 (78). α, β, and γ subunits are showed in red, blue, and green, respectively. Side chains of αE557, βE499, and γD516 are presented as spheres. A, side view. B, top view. C, βE499 and its adjacent residues in the α subunit. The displayed area corresponds to the square in A. Side chains of the labeled residues are presented as sticks with carbon in cyan, oxygen in red, and nitrogen in blue. Distances between two residues were measured as the minimal distance between nonhydrogen atoms of the side chains of the two residues using PyMol. The three α subunit residues are the closest polar residues to βE499. D, sequence alignments of the α, β, and γ subunits of human (h), mouse (m), rat (r) ENaC, and chicken (c) ASIC1a. Numbers in the parenthesis represent the first residues of the amino acid sequence. Acidic residues of interest within the β11-β12 linker are framed in red rectangle. Among them, αE557, βE499, and γD516 of mouse ENaC are shown in red, blue, and green. ASIC1a, acid-sensing ion channel 1; ENaC, epithelial Na + channel. the template of trimetric human ENaC structure (41) (Fig. 1, A  and B).
At the α/β subunit interface in the mouse ENaC model, we identified three α-subunit polar residues in close proximity (less than 4.5 Å of minimal distance between side-chain oxygen or nitrogen atoms) to βE499 in the β11-β12 linker, including αN510 in β10 strand of the palm domain, αQ441 in α4 helix, and αT439 immediately preceding α4 helix of the thumb domain (Fig. 1C). These residues (βE499, αN510, αT439, and αQ441) were individually mutated to Cys, and wildtype or mutant channels bearing one or two Cys mutants were expressed in Xenopus oocytes. The functional effects of inducing or disrupting Cys crossbridges were assessed by twoelectrode voltage clamp. Oocytes were treated with an oxidizing agent (H 2 O 2 ) to induce disulfide bond formation, or with a reducing agent (DTT) to break disulfide bonds. Changes in Na + current were assessed in response to oxidizing or reducing agents, whereas oocytes were maintained at a constant potential (−100 mV). The response to amiloride (10 μM) was measured to determine the ENaC-dependent component of the whole cell Na + current. As mentioned previously, ENaCs not only conduct Na + but are inhibited by extracellular Na + , a phenomenon referred to as Na + self-inhibition (3,14). This inhibitory response to Na + is assessed by rapidly changing the solution bathing oocytes from a low (1 mM) to a high (110 mM) [Na + ] while oocytes are maintained at a constant potential (−100 mV). The change in bath [Na + ] was accompanied by a rapid increase in inward Na + current, which reaches a peak current (Ipeak) and was followed by a slow decay in current to a steady-state current (Iss), reflecting Na + self-inhibition. The percent of inhibition following Ipeak was calculated as described under the Experimental procedures section. Previous work has shown that the magnitude of the Na + self-inhibition response correlates with ENaC open probability (42). In selected experiments, the Na + self-inhibition response was examined before and after application of a reducing or oxidating reagent. In other experiments, only the later was examined.
Channels with Cys substitutions of conserved acidic residues in the β11-β12 linker respond to sulfhydryl reagents in a subunit-dependent manner To further explore the functional roles of the β11-β12 linker acidic residues, we examined the effects of sulfhydryl reagents on the activity of ENaCs with Cys substitutions at αE557, βE499, or γD516 (Fig. 1). As shown in Figure 2, sodium (2-sulfonatoethyl) methanethiosulfonate (MTSES) significantly reduced or increased the currents in oocytes expressing αE557Cβγ or αβE499Cγ, respectively, when compared with wildtype channels. It also moderately reduced αβγD516C channel activity (Fig. 2, A and B). Similarly, [2-(trimethylammonium) ethyl] methanethiosulfonate bromide (MTSET) significantly reduced αE557Cβγ activity and increased αβE499Cγ and αβγD516C activity (Fig. 2, C and D). These results suggest these β11-β12 acidic residues have roles in modulating ENaC activity, in agreement with prior studies (38)(39)(40).
Crosslinking αN510C and βE499C at the α/β subunit interface activates ENaC To explore the functional interactions between αN510C in the α-subunit palm domain β10 strand and βE499C in the β-subunit β11-β12 linker (Fig. 3C), channels with 0, 1 or 2 Cys-substituted subunits were coexpressed with complementary wildtype subunits in Xenopus oocytes. Whole-cell currents in oocytes expressing wildtype or mutant ENaCs were measured at −100 mV (membrane potential). After assessing the Na + self-inhibition response, 10 mM DTT was added to the bath solution. DTT did not significantly alter currents in oocytes expressing wildtype ENaC (Fig. 3A), consistent with previous observations (43). The slow decrease in current over time likely reflects the well-known rundown in ENaC activity under a continuous clamping at hyperpolarizing potential (44). Interestingly, in contrast to the continuous rundown in WT and single Cys mutants, DTT treatment caused a modest but rapid inhibition of the double Cys mutant (αN510CβE499Cγ, Fig. 3A). The current in the presence of DTT relative to basal current (I DTT /I) was significantly lower than that observed with wildtype ENaC or channels with a single Cys substitution (Fig. 3B). These results suggest that DTT inhibited αN510CβE499Cγ channels by breaking a disulfide bond spontaneously formed between the two Cys residues.
The aforementioned results suggested that a disulfide bond between αN510C and βE499C activates ENaC. Nevertheless, the DTT effect on αN510CβE499Cγ channel was modest, raising the possibility that only some channels formed disulfide bridges between αN510C and βE499C. We suspected that an oxidizing reagent would activate αN510CβE499Cγ channels by promoting the disulfide bridging and would enhance the subsequent response to DTT. To enhance disulfide bond crosslinking of αN510C and βE499C, we exposed ENaCexpressing oocytes to hydrogen peroxide (H 2 O 2 ; 0.045%) prior to DTT (10 mM) treatment. Modest decreases in wholecell currents were observed in oocytes expressing either wildtype, αN510Cβγ, or αβE499Cγ channels in response to 0.045% H 2 O 2 (Fig. 3, D and E). In contrast, oocytes expressing αN510CβE499Cγ channels responded to H 2 O 2 with a large and significant increase in current (I H2O2 /I = 1.78 ± 0.16, n = 11, p < 0.0001 versus 0.70 ± 0.04, n = 11 for WT; p < 0.0001 versus 0.69 ± 0.04, n = 12 for αN510Cβγ; p < 0.0001 versus 0.91 ± 0.06, n = 12 for αβE499Cγ), which was sustained following H 2 O 2 washout (Fig. 3, D and E). Subsequent treatment with 10 mM DTT led to a large fall in current only in oocytes expressing αN510CβE499Cγ channels (Fig. 3, D and F). Moreover, the effect of DTT on αN510CβE499Cγ channels was greater than what would be expected from a simple reversal of the H 2 O 2 -induced current increase. This likely reflected an additive effect of DTT on both populations of channels with pre-existing and H 2 O 2 -induced disulfide bonds. These results suggest that crosslinking these two residues activated the mutant channels, whereas breaking the bond inhibited the channels.
We predicted that the activated channels would exhibit a reduced Na + self-inhibition response. This is what we observed. The Na + self-inhibition response of αN510CβE499Cγ channels was significantly blunted after treatment with H 2 O 2 . Following DTT treatment, the Na + self-inhibition response was more robust than the response under baseline conditions (Fig. 3, G and H). Again, the later likely reflected the effect of DTT on H 2 O 2 -oxidized channels as well as naturally bridged channels. Basal channel currents prior to either DTT or H 2 O 2 did not significantly differ in oocytes expressing wildtype or mutant channels (p > 0.05, Fig. 3I).
Crosslinking αQ441C and βE499C at the α/β subunit interface favors low channel activity We examined whether other residues with Cys substitutions in the vicinity of βE499C formed functional crosslinks. αQ441 is in the thumb domain α4 helix and in proximity to βE499 in the β11-β12 linker (Fig. 4J). Oocytes expressing αQ441C-βE499Cγ channels, responded to the reducing agent DTT with a significant increase in current, which was not seen in oocytes expressing wildtype or channels with a single Cys substitution (I DTT /I = 1.76 ± 0.23, n = 12, p < 0. 0001 versus 0.91 ± 0.03, n = 12 for WT; p < 0.0001 versus I DTT /I = 0.68 ± 0.08, n = 12 for αQ441Cβγ; p < 0.0001 versus I DTT /I = 0.86 ± 0.04, n = 12 for αβE499Cγ; Fig. 4, A and B). A small inhibition of current in response to DTT was observed in αQ441Cβγ channels (I DTT / I = 0.68 ± 0.08, n = 12, p = 0.0002 for αQ441Cβγ versus I DTT / I = 0.91 ± 0.03, n = 12 for WT, Fig. 4, A and B) for an unknown reason. The large effect of DTT specifically in the double Cys mutant channels was consistent with a disulfide bridge that occurred naturally in the absence of an oxidizing reagent.
Treatment of wildtype or mutant channels with the oxidizing agent H 2 O 2 was associated with minimal changes in currents (Fig. 4, D and E), whereas subsequent treatment with DTT led to a selective increase in current of αQ441CβE499Cγ channels (Fig. 4, D and F). When cells expressing αQ441C-βE499Cγ channels were pretreated with DTT, subsequent treatment with H 2 O 2 led to a large decrease in current (Fig. 4, G and H). These results suggest that a disulfide bond was spontaneously formed between αQ441C and βE499C in the absence of an added oxidant, and the bridging rendered the mutant channels in a low-activity state. To demonstrate that   (blue), αβE499Cγ (orange), and αN510CβE499Cγ (red) ENaCs showing the effects of DTT and Na + self-inhibition responses before and after DTT. Na + selfinhibition was examined by switching Na + bath solution from 1 mM (white bar) to 110 mM (black bar). Ipeak and Iss represent the peak current after switching from low to high [Na + ], and the steady-state current measured 40 s after Ipeak, respectively. 10 mM DTT (dark gray bar) was applied for 5 min. Amiloride (10 μM) was added to the bath as indicated by an arrow. Traces were superimposed by aligning the currents prior to DTT application. B, scatter plot of I DTT /I. I and I DTT were amiloride-sensitive current before and after DTT. Dot plots of WT, αN510Cβγ, αβE499Cγ, and αN510CβE499Cγ are presented in colors matching their traces. Bars are mean ± SD. Numbers in the parentheses are numbers of oocytes used in the experiment. C, a mouse ENaC model showing the relative position of βE499C and αN510C. Side chains of βE499C and αN510C were modeled by PyMol using the default rotamers. The distance was measured from sulfur to sulfur on the two Cys residues. Carbon, oxygen, nitrogen, and sulfur are shown in cyan, red, blue, and yellow, respectively. D, current traces from oocytes expressing WT, αN510Cβγ, αβE499Cγ, and αN510CβE499Cγ ENaCs are shown in colors consistent with A. H 2 O 2 (0.045%, light gray bar) was applied in 110 mM Na + bath solution for 3 min and washed out for 1 min before adding 10 mM DTT (dark gray bar) for 2 min and amiloride (10 μM, arrow). E, dot plots of I H2O2 /I. I and I H2O2 were currents measured before and after 0.045% H 2 O 2 , respectively. F, dot plots of I DTT /Iwash. Iwash and I DTT were measured immediately prior to and at the end of DTT application. Bars are mean ± SD. G, representative trace presenting the change of Na + selfinhibition responses before and after 0.045% H 2 O 2 (3 min) and after 10 mM DTT (2 min) in an oocyte expressing αN510CβE499Cγ. H, dot plots of Na + self-inhibition (%) measured prior to H 2 O 2 (self-inhibition (SI)-1), after H 2 O 2 (SI-2) and after DTT (SI-3) in G. Values were calculated using the formula: 100 × (Ipeak-Iss)/Ipeak. The p values were obtained using repeated-measures ANOVA and Tukey's post hoc test. I, normalized amiloride-sensitive currents represent basal channel activity prior to either DTT or H 2 O 2 treatment. Data were from experiments as shown in A and D. Amiloride-sensitive currents from all oocytes of the same batch were normalized to the mean of the currents in oocytes expressing wildtype ENaCs. The amiloride-sensitive currents from the three batches of oocytes expressing wildtype channels were 1.9 ± 1.5 μA (n = 5), 5.6 ± 1.5 μA (n = 6), and 6.1 ± 1.8 μA (n = 5). Since data from two groups did not pass normality test, Kruskal-Wallis nonparametric test was used for statistical analysis. There was no significant difference among the four groups (p > 0.05). The p values in B, E, and F were from one-way ANOVA and Tukey's post hoc tests. ENaC, epithelial Na + channel.  Current traces of WT (purple), αQ441Cβγ (blue), αβE499Cγ (orange), and αQ441CβE499Cγ (red) are superimposed by aligning the currents prior to DTT application. About 10 mM DTT was applied for 5 min. Ipeak and Iss represent the peak current after switching from low to high [Na + ] and the steady-state αQ441CβE499Cγ channels indeed have low ENaC activity, we expressed wildtype, αQ441Cβγ, αβE499Cγ, and αQ441C-βE499Cγ channels in oocytes and measured amiloridesensitive currents. As shown in Figure 4I, oocytes expressing αQ441CβE499Cγ channels had significantly lower currents than oocytes expressing wildtype and single mutant channels. The Na + self-inhibition response of αQ441CβE499Cγ channels (21 ± 11%, n = 12) was significantly less than that of wildtype (32 ± 9%, n = 12), αQ441Cβγ (32 ± 11%, n = 12), and αβE499Cγ (34 ± 9%, n = 12, Fig. 4, A and C). For αQ441C-βE499Cγ channels, DTT not only activated the channel but also significantly increased the Na + self-inhibition magnitude of αQ441CβE499Cγ (36 ± 12%, n = 12, p < 0.0001 versus the first response prior to DTT (see aforementioned) from paired Student's t test. These results are in contrast to the inverse relationship between the magnitude of Na + self-inhibition and ENaC activity that has been frequently observed in mutagenesis studies (see Discussion section) (41,45,46).
Introduced Cys at other sites adjacent to αQ441C, αN510C, and βE499C at the α/β subunit interface do not respond to reducing or oxidizing reagent We examined the effects of reducing and oxidizing reagents on channels with introduced Cys residues at other adjacent sites at the α/β subunit interface, including αN510CβF500Cγ, αN510CβM85Cγ, and αT439CβE499Cγ (Figs. 5, A and B and  6A). Significant changes of current were not observed following treatments with either H 2 O 2 or DTT (Figs. 5 and 6), highlighting the specificity of the functional crosslinking we observed between αN510, αQ441, and βE499. We also assessed if a pair of Cys residues within β subunit (βE499C and βG86C) could form spontaneous or induced disulfide bond (Fig. 6B). A lack of effect of either DTT or H 2 O 2 (Fig. 6, G-J) on αβG86C-E499Cγ channels suggested this was not the case. A similar pair of Cys residues within β11-β12 and β1-β2 linkers in ASIC1a were reported to form a disulfide bond, leading to dramatically slowed desensitization (32).
To validate an induced disulfide bond between αE557C and γQ398C and further explore interactions between the two residues, we took advantage of the fact that two introduced Cys residues in close proximity can coordinate a metal ion, forming a metal-mediated crosslink (47). We examined whether Cu 2+ could bridge αE557C and γQ398C. Treatment of oocytes expressing αE557CβγQ398C channels with 1 μM Cu 2+ led to a rapid and large reduction in current that persisted after removing Cu 2+ from the bath solution. In contrast, Cu 2+ did not reduce currents in oocytes expressing WT, αE557Cβγ, or αβγQ398C channels (αE557CβγQ398C I Cu2+ /I = 0.46 ± 0.15, n = 10, p = 0.0005 versus 1.07 ± 0.08, n = 9 for WT; p = 0.026 versus 1.03 ± 0.08, n = 8 for αE557Cβγ; p = 0. 0003 versus 1.07 ± 0.04, n = 8 for αβγQ398C, Fig. 7, I and J). Subsequent treatment with 10 mM DTT reversed the Cu 2+ -induced inhibition of αE557CβγQ398C channels, whereas little or no change in current was observed in oocytes expressing WT, αE557Cβγ, or αβγQ398C channels (Fig. 7, I and K). These results suggest that crosslinking αE557C and γQ398C at the α/γ interface maintains the channel in a low-activity state. Interestingly, current measured 40 s after Ipeak, respectively. B, dot plots of the I DTT /I for WT and mutant channels. The I DTT and I were the currents measured 5 min after and immediately before DTT treatment, respectively. Data were from the experiments shown in A. WT, αQ441Cβγ, αβE499Cγ, and αQ441CβE499Cγ are presented in color consistent with their traces. C, dot plots of Na + self-inhibition (%) of WT and mutant channels shown in colors matching traces in A. D, representative current recordings showing the effect of 0.045% H 2 O 2 and 10 mM DTT on WT, αQ441Cβγ, αβE499Cγ, and αQ441CβE499Cγ channels. H 2 O 2 was applied for 3 min, followed by 0.5 min wash out and then 5 min DTT treatment. E, dot plots of I H2O2 /I. The I H2O2 and I were the currents measured after and immediately before H 2 O 2 treatment, respectively. F, dot plots of I DTT /Iwash. The I DTT and Iwash were measured after and immediately before DTT application (i.e., after washout of H 2 O 2 ), respectively. G, current trace showing the current changes responding to DTT (4 min), DTT washout (0.5 min), and H 2 O 2 (3 min) in the same oocyte expressing αQ441CβE499Cγ. H, normalized amiloride-sensitive currents (n = 8) were the amiloride-sensitive currents measured immediately before DTT (I), after DTT (I DTT ), after DTT washout (Iwash), and after H 2 O 2 (I H2O2 ) that were divided by the mean of the current immediately before DTT. p Values were calculated by repeated-measures ANOVA and Tukey's post hoc test. In B, C, E, and F, bars are mean ± SD, with numbers of oocytes shown in the parentheses, and one-way ANOVA and Tukey's post hoc test were used to obtain p values. I, normalized amiloridesensitive currents were amiloride-sensitive currents divided by mean of the amiloride-sensitive current in the same batch of oocytes expressing wildtype ENaC. Data were from two batches of oocytes in which the amiloride-sensitive currents in wildtype-expressing oocytes were 5.9 ± 0.6 μA (n = 27) and 4.7 ± 0.9 μA (n = 30). The p values were from Kruskal-Wallis nonparametric test with Dunn's multiple comparisons test as data from three groups did not pass a normality test. The experiment was performed to compare expressed currents of the wildtype and mutant channels, and the data were not from experiments shown in A and D. J, mouse ENaC model showing the relative locations of βE499C and its adjacent αENaC residues. ENaC, epithelial Na + channel.
αE557C and γQ398C are homologous to the βE499C and αQ441C within the β/α interface, where crosslinking also kept the channel in a low-activity state (Fig. 4).

Discussion
We introduced Cys residues at specific sites at ENaC subunit interfaces to explore the structural transitions during ENaC gating. Using the oxidizing agent H 2 O 2 or metal ion Cu 2+ to facilitate Cys crosslinks and the reducing agent DTT to release disulfide bonds, we examined the functional effects of intersubunit interactions at selected sites. Our results indicate that crosslinking αN510C and βE449C activated ENaC. In contrast, crosslinking αQ441C and βE499C, αE557C and γQ398C inhibited ENaC. We propose that residues, at or in the vicinity of these sites, facilitate interactions across subunit interfaces that help stabilize the channel in an open state or closed state. In essence, these residues constitute microswitches that can be triggered to transmit conformational changes that occur during ENaC gating transitions (Fig. 8). While we have not directly assessed channel gating, the changes in current that we are examining, in all likelihood, reflect changes in channel gating. It is difficult to explain the rapid and often reversible changes we are seeing on changes in ENaC trafficking and number of channels at the cell surface, and we are not perturbing the transmembrane regions where we might observe changes in single channel conductance.
Inducing an α/β-subunit crosslink between αN510C in palm domain β10 strand and βE499C in the palm domain β11-β12 linker promoted channel activation, whereas crosslinking αQ441C in the α4 helix of thumb domain and βE499C led to channel inhibition. This shift in bridging partners of βE499C (with αN510C or αQ441C) had a dramatic effect on the conformational states of ENaC, suggesting that intersubunit interactions involving βE499 and neighboring residues modulate channel gating. We propose that βE499, αN510, and αQ441 form a microswitch at α/β subunit interface that help stabilize ENaC in an open state or closed state, respectively. This switch can be triggered during intrinsic gating and in response to extracellular Na + and other factors that impact ENaC gating behavior. A function role of the β11-β12 linker βE499 in regulating ENaC gating is also supported by the observed activation of αβE499Cγ channel by MTS reagents (Fig. 2).
The exact structural changes that occur in response to a Cys crosslink (or loss of a crosslink) that maintain channels in a specific (high or low) activity state are not known and will require high-resolution structures of ENaC in closed and open states to elucidate structural transitions associated with changes in ENaC gating that can then be confirmed by additional functional studies. As outlined in Figure 8, we hypothesize that crosslinking specific sites in the α-subunit palm domain β10 strand and the β-subunit palm domain β11-β12 linker (αN510C and βE499C) constrain these subunits at the mid palm domain level, altering interactions between α-subunit thumb domain and β-subunit palm domain and facilitating a transition in the transmembrane helices to an open conformation state. In contrast, crosslinking the α-subunit thumb domain α4 helix and β-subunit palm domain β11-β12 linker (αQ441C-βE499C) strengthen the interaction between these two domains, prompting a transmembrane helical conformation that stabilizes the channel in a closed state. We found that the specific functional interactions at the α/β subunit interface were conserved at the α/γ subunit interface. Similar to an αQ441C and βE499C bridge, crosslinking αE557C in the palm domain β11-β12 linker and γQ398C in α4 helix of thumb domain trapped channels at a low-activity state. We speculate that β11-β12 linker αE557 and thumb domain γQ398 may form another microswitch at the α/γ subunit interface in ENaC gating. The functional changes in both pairs at α/β and α/γ interfaces are consistent with the notion that constraining thumb domain and palm domain at subunit interfaces favors a closed state (48)(49)(50). In ASIC1a, introduction of a disulfide bridge between thumb domain N357C of one subunit and palm domain β1-β2 linker T84C of another subunit blocked channel activation by protons, apparently favoring a closed state (32). These findings suggest that thumb and palm domain interactions at a subunit interface are a common feature in the regulated gating of ENaCs and ASICs.
In the assessment of crosslinking αE557C and γQ398C, H 2 O 2 induced an unexpected inhibition on αE557C channels (Fig. 7, F and G). Given its insensitivity to DTT, the H 2 O 2induced inhibition might reflect oxidation of the thiol group (-SH) of αE557C to sulfinic acid (-S=O(OH)) or sulfonic acid (-S(=O) 2 (OH)). These oxidation reactions are not reversed by reducing agents (51). While it is not clear why αE557C is particularly sensitive to oxidation, it is known that susceptibility to oxidation of thiols is usually correlated to its acidity, which is influenced by the presence of polar residues or hydrogen bonds (52). As shown in Figure 7A, αE557 is in proximity to several polar residues including γS396, γQ398, and γE467. Interestingly, we observed a similar H 2 O 2 -induced inhibition in γE467C channels (L. Z. et al., unpublished observation). The effects of H 2 O 2 on both αE557C and γE467C preclude a clear assessment of a potential bridge between these Cys residues. Nevertheless, the surprising effect highlights the significant role of αE557 in regulating ENaC activity. The same may be true for γE467, as suggested in a recent report (41).
Members of the ENaC/degenerin family have evolved as ion channels with large extracellular domains with complex folds, where channel activity is highly regulated by extracellular factors. For ENaC, these factors include Na + and other metals, H + , Cl − , proteases, and shear stress (3,14). ASICs are regulated by extracellular H + and specific small proteins that bind to and regulate channel activity (53). Degenerin channels in Caenorhabditis elegans are mechanoactivated channels (54)(55)(56). It is important to understand the extracellular structural transitions that occur in response to these channel activators or inhibitors and lead to transitions in the transmembrane helices that stabilize channels in a high-or low-activity state. Our results provide insights regarding possible structural transitions within the extracellular ENaC subunits that affect channel gating in response to extracellular factors. Furthermore, our observations in this study may advance an understanding of the allosteric regulations of ENaC-related channels.
Aside from transporting Na + , extracellular Na + is one of the key regulators of ENaC through binding to an extracellular site in the α subunit based on functional as well as structural studies (15,29). The inhibitory response of extracellular Na + , referred to as Na + self-inhibition, transitions channels to a lowactivity state. Open probability of ENaC is, in general, C-F, current traces and summary data to probe if βE499C and αT439C could be crosslinked. G-J, current traces and summary data to probe if βE499C and βM85C could be crosslinked. All C-J panels are presented as for Figures 3 and 4. In C and G, DTT was applied for 2 min. In E and I, H 2 O 2 was applied for 3 min, after wash out for 1 min, DTT was applied for 2 min. The p values were from one-way ANOVA and Tukey's post hoc analysis. ENaC, epithelial Na + channel; H 2 O 2 , hydrogen peroxide. proportional to the magnitude of Na + self-inhibition (3,14,42). Other extracellular regulators, including proteases, modulate ENaC activity in part by altering Na + self-inhibition (3, 14-16, 26, 29, 37, 46, 49, 57-72). We examined the Na + self-inhibition response of the double Cys mutant channels before and after inducing or breaking a disulfide bond to determine if it was altered. Interestingly, we found that these functional crosslinks differentially affected the Na + selfinhibition response. For αN510CβE499Cγ channels, crosslinking increased channel current with a blunted Na + selfinhibition response, whereas releasing the disulfide bond reduced channel current and enhanced Na + self-inhibition (Fig. 3, G and H). This relationship between channel activity and the magnitude of the Na + self-inhibition response is strikingly similar to what has been observed in gain-offunction mutations and specific ENaC activators that affect channel gating (20, 21, 26, 37-40, 42, 45, 46, 49, 62-65, 68, 70, 73-77). In contrast to αN510CβE499Cγ channels, crosslinked αQ441C-βE499C channels exhibited both reduced channel currents and a blunted Na + self-inhibition (Fig. 4C). These channels were activated by DTT, along with a restored Na + self-inhibition response. Our results suggest that αQ441C-βE499Cγ channels are crosslinked under baseline conditions, and that this crosslink largely prevents conformational changes induced by extracellular Na + . This is reminiscent of the previously described αK477C-βV85C crosslinked channels (50). The Na + self-inhibition response can be readily observed by monitoring current changes in response to a rapid change in the bath [Na + ], reflecting a conformational change that impacts the open probability of ENaC. Our results suggest that crosslinked αQ441CβE499Cγ channels are trapped in a low-activity (i.e., low open probability) state, even in the presence of a low (1 mM) bath [Na + ]. Consistent with this notion, Collier et al. (50) reported a low open probability of crosslinked αK477CβV85C (also within the α/β subunit interface) in the presence of 1 mM extracellular Li + , along with a reduced Na + self-inhibition response. DTT relieved this locked low-activity state, permitting conformational changes in response to bath [Na + ] changes, similar to the response of wildtype ENaC.
In summary, we identified multiple pairs of extracellular domain residues where introduced Cys side chains across Purple arrows indicate potential movement of transmembrane (TM) domains and the lower thumb in response to βE499C and αN510C crosslinking. We propose that crosslinking βE499C and αN510C strengths interaction between the two palm domain residues and weakens the interaction between βE499 and thumb domain αQ441. As a result, the α-subunit lower thumb domain swings away from the α/β interface, favoring an open state TM conformation. B, a model showing a possible mechanism behind ENaC inhibition by crosslinking βE499C and αQ441C. We propose that crosslinking palm domain βE499C and thumb domain αQ441C strengths the interaction between the two residues and enhances contacts between the α thumb domain and β palm domain. This crosslink swings the α subunit lower thumb toward the α/β interface, favoring a closed TM conformation. ENaC, epithelial Na + channel.
DTT or H 2 O 2 treatment as shown in C and F. Amiloride-sensitive currents from all oocytes of the same batch were normalized to the mean current in oocytes of the same batch expressing wildtype ENaC. The amiloride-sensitive currents from the four batches of oocytes expressing wildtype channels were 4.1 ± 0.9 μA (n = 3), 1.2 ± 0.2 μA (n = 4), 2.6 ± 0.6 μA (n = 3), and 8.2 ± 1.7 μA (n = 5). Since data from one group did not pass normality test, Kruskal-Wallis nonparametric test was used for statistical analysis. There was no significant difference among the four groups (p > 0.05 ENaC gating subunit interfaces formed either spontaneous or oxidantinduced disulfide bridges. These crosslinks resulted in significant changes in ENaC activity that were attributable to altered channel gating. Interestingly, certain crosslinking events disrupted Na + self-inhibition response, whereas disulfide bridge formations at other sites retained the inhibitory response. We propose certain endogenous residues at or near these crosslinking sites interact across these interfaces forming key microswitches to facilitate structural changes during channel transitions between closed and open states. There are likely additional sites present in the extracellular domains that mediate domain-domain interactions. Identification and characterization of these interacting sites will greatly enhance our understanding of the mechanisms of ENaC gating and its regulation and should facilitate the discovery of specific therapeutic agents selectively targeting ENaC gating.

Experimental procedures
Chemicals MTSES and MTSET were from Toronto Research Chemicals. All other chemicals were from Sigma-Aldrich unless otherwise stated.

Site-directed mutagenesis in ENaC
Point mutations were introduced into mouse α, β, and γ ENaC plasmid DNAs using QuickChange II XL site-directed mutagenesis kit (Agilent). The DNA sequences, including those of wildtype and mutant ENaC, were confirmed by direct sequencing at the Genomics Research Core of University of Pittsburgh. Wildtype and mutant RNAs were made from the linearized DNA using T3 RNA in vitro transcription kit and purified with RNA Cleanup Kit (Thermo Fisher Scientific Inc). RNA quality was verified by denaturing RNA gel analysis, and RNA concentrations were estimated by a spectrophotometer. One mutation (αN510C) was made previously (41).

Xenopus oocyte isolation
Oocytes harvested from female Xenopus laevis were separated into small pieces and treated with 2 mg/ml type II collagenase plus 2 mg/ml trypsin inhibitor for 55 to 60 min. Dispersed oocytes were immersed in solution containing 100 mM K 2 HPO 4 solution plus 1 mg/ml bovine serum albumin for 15 min before wash and incubation with modified Barth's solution (88 mM NaCl, 1 mM KCl, 2.4 mM NaHCO 3 , 15 mM Hepes, 0.3 mM Ca(NO 3 ) 2 , 0.41 mM CaCl 2 , 0.82 mM MgSO 4 , 10 μg/ml streptomycin sulfate, 100 μg/ml gentamycin sulfate, and 10 μg/ml sodium penicillin, pH 7.4). The frog surgery protocol was approved by the University of Pittsburgh's Institutional Animal Care and Use Committee.

Two-electrode voltage clamp
Whole cell Na + currents were measured using a twoelectrode voltage clamp at room temperature 1 or 2 days after injection. Oocytes were placed in a chamber and perfused with constant flow (NaCl-110: 110 mM NaCl, 2 mM KCl, 2 mM CaCl 2 , and 10 mM Hepes, pH 7.4) and clamped at −100 mV (membrane potential) using glass pipettes filled with 3 M KCl solution. Recorded signals were amplified and digitized using TEV200A Voltage Clamp Amplifier (Dagan Corporation), DigiData 1440A, and Clampex 10.4 software (Molecular Devices).

Effects of MTSET and MTSES
MTSES and MTSET were freshly dissolved in regular bath solution (NaCl-110). MTSES (2 mM) solution was used for experiments within 2 h. MTSET (1 mM) solution was prepared immediately prior to use. MTSES and MTSET solutions were perfused for 2 min. Amiloride-sensitive Na + currents prior to and after MTSES or MTSET application were measured as I and I MTSES or I MTSET , respectively. Na + -self inhibition response Na + -self inhibition was examined by perfusing oocytes with NaCl-1 bath solution (1 mM NaCl, 109 mM N-methyl-Dglucamine, 2 mM KCl, 2 mM CaCl 2 , and 10 mM Hepes, pH 7.4) for 60 s and rapidly switching to NaCl-110 bath solution described previously and perfusing for 60 s. During this process, current recorded increased rapidly to reach a peak value (Ipeak) and then gradually decreased to a steady-state level (Iss, measured at 40 s after Ipeak). Na + self-inhibition magnitude was calculated as percentage of (Ipeak-Iss)/Ipeak.

Crosslinking introduced Cys residues and reducing disulfide bonds
DTT (200 mM) stock solutions were prepared on the day of experiment in NaCl-110. DTT stock solutions were diluted with NaCl-110 to 10 mM preceding recording. H 2 O 2 (3%) solutions were stored at room temperature and protected from light. H 2 O 2 (0.045%) solutions were prepared by diluting 3% H 2 O 2 with NaCl-110 before recording. A 40 mM CuSO 4 stock solution was prepared in ethanol and diluted to 1 μM with NaCl-110 before use. A 10 mM DTT, 0.045% H 2 O 2 , or 1 μM Cu 2+ solution was applied extracellularly after the current became stable. At the end of each recording, 10 μM amiloride was applied to determine the amiloride-sensitive component of the whole-cell current. Amiloride-sensitive current before and after the treatment was measured using Clampfit 10.4 (Molecular Devices) and presented as I and I DTT , I H2O2 , and I Cu . The ratios I DTT /I, I H2O2 /I, and I Cu /I were calculated to show the effects of these reagents.

Statistical analyses
Data are presented as mean ± SD. Shapiro-Wilk test was performed for normality determination. Significance was examined by Student's t test with Welch's correction for comparing two groups, one-way ANOVA followed by Tukey's post hoc test for comparing multiple groups. p < 0.05 was regarded statistically significant. p Values were shown unless less than 0.0001. All analyses were performed using Prism 7 or 9 (GraphPad Software, Inc).

Data availability
All data generated or analyzed during this study are included in the article and supporting information.
Funding and additional information-This work was supported by the National Institutes of Health grants R01 HL147818 and P30 DK079307 and the Xiangya Scholar Fund (to L. Z. and X. W.) from The Third Xiangya Hospital, Central South University, China. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.