Conserved cysteines in the finger domain of the epithelial Na+ channel α and γ subunits are proximal to the dynamic finger–thumb domain interface

The epithelial Na+ channel (ENaC) is a member of the ENaC/degenerin family of ion channels. In the structure of a related family member, the “thumb” domain's base interacts with the pore, and its tip interacts with the divergent “finger” domain. Between the base and tip, the thumb domain is characterized by a conserved five-rung disulfide ladder holding together two anti-parallel α helices. The ENaC α and γ subunits' finger domains harbor autoinhibitory tracts that can be proteolytically liberated to activate the channel and also host an ENaC-specific pair of cysteines. Using a crosslinking approach, we show that one of the finger domain cysteines in the α subunit (αCys-263) and both of the finger domain cysteines in the γ subunit (γCys-213 and γCys-220) lie near the dynamic finger–thumb domain interface. Our data suggest that the αCys-256/αCys-263 pair is not disulfide-bonded. In contrast, we found that the γCys-213/γCys-220 pair is disulfide-bonded. Our data also suggest that the γ subunit lacks the terminal rung in the thumb domain disulfide ladder, suggesting asymmetry between the subunits. We also observed functional asymmetry between the α and γ subunit finger–thumb domain interfaces; crosslinks bridging the α subunit finger–thumb interface only inhibited ENaC currents, whereas crosslinks bridging the γ subunit finger–thumb interface activated or inhibited currents dependent on the length of the crosslinker. Our data suggest that reactive cysteines lie at the dynamic finger–thumb interfaces of the α and γ subunits and may play a yet undefined role in channel regulation.

The epithelial Na ϩ channel (ENaC) 2 facilitates Na ϩ transport in the aldosterone-sensitive distal nephron, the respiratory tract, and the colon (1)(2)(3)(4). ENaC-mediated Na ϩ reabsorption in the distal nephron is critical to K ϩ secretion and plays a key role in extracellular fluid volume homeostasis and blood pressure maintenance (1). ENaC also plays sensory roles in the taste buds, vascular endothelia and smooth muscle, and the brain (5). At the luminal surface in its varied roles, ENaC is exposed to dynamic extracellular environments and is responsive to stimuli in those environments, including Na ϩ , Cl Ϫ , H ϩ , and shear stress. These factors directly regulate the channel by influencing the channel's open probability (6).
ENaC is a member of the ENaC/degenerin family of ion channels, which are cation-selective channels that gate in response to extracellular cues (6). There are four ENaC subunits, ␣, ␤, ␥, and ␦, that assemble to form either ␣␤␥ or ␦␤␥ heterotrimers. Channel assemblies that include ENaC subunits and acid-sensing ion channel (ASIC) subunits have also been reported (7)(8)(9). Sequence comparisons between ENaC subunits and ASIC1, whose structure has been reported, suggest that the core folds of the proteins are similar (palm and ␤-ball domains in Fig. 1A), but that the peripheral helical domains (finger, thumb, and knuckle in Fig. 1A) are divergent and may underlie functional differences between the channels (10,11). The sites of channel activating proteolytic cleavage in the ␣ and ␥ subunits map to their respective finger domains (6,12,13), underlining the notion that the divergent domains are functionally important. We previously reported a model of the ␣ subunit extracellular domains using ASIC1 homology complemented by constraints for the divergent finger domain derived from functional experiments (10).
ENaC/degenerin proteins contain 14 conserved cysteines in the extracellular domains (11,14). Based on functional experiments, Sheng et al. (14) proposed a model for the ␣ subunit where 10 of the cysteines formed a disulfide bond ladder that defined one domain, and the remaining 4 cysteines formed two disulfide bonds elsewhere in the structure. The ASIC1 structure accorded with these results and showed that the disulfide bond ladder held two helices together in an antiparallel fashion and formed the core of the thumb domain ( Fig. 1A) (11). Firsov et al. also functionally examined the extracellular domain cysteines, and proposed asymmetry between the subunits regarding the cysteine pair at the top of the thumb domain (15). The thumb domain of each subunit interacts with the finger domain at one end and the extracellular entry to the channel pore at the other. Each ENaC subunit finger domain contains two cysteines that are unique to ENaC subunits in the protein family ( Fig. 1 (A  and B); Cys-2 and Cys-3) and two additional cysteines that are unique to ENaC ␤ subunits (Fig. 1B). Although we assumed the finger domain cysteines to be disulfide-bonded in our ␣ subunit cro ARTICLE homology model, the disposition of these cysteines had not been experimentally determined.
Double cleavage of the ENaC ␣ and ␥ subunits liberates autoinhibitory tracts from each subunit (13,16). Synthetic peptides corresponding to each of the liberated tracts inhibit the channel (17,18). We have shown that the ␣ subunit-derived inhibitory peptide and the autoinhibitory tract from which it is derived probably bind the ␣ subunit at overlapping sites (10,19). One key feature of these sites is involvement of the finger-thumb domain interface. We hypothesized that ␣ subunit cleavage activates the channel by releasing an interaction between the finger and thumb domains. We proposed that this conformational change propagates through the thumb domain to the transmembrane helices that form the channel pore (see Fig.  1A). Indeed, directly crosslinking the finger-thumb interface inhibited the channel in a crosslinker length-dependent manner (20).
Here, through crosslinking experiments, we explore the finger-thumb domain interfaces of the ␣ and ␥ subunits. We show that both interfaces are dynamically linked to channel activity. We also demonstrate that the conserved cysteines in the finger domains of the ␣ and ␥ subunits (Cys-2 and Cys-3) lie close to the respective finger-thumb domain interfaces. However, our data suggest that whereas the ␥ subunit finger domain cysteines are probably disulfide-bonded, the ␣ subunit equivalents are not. Additionally, our data suggest that the ␥ subunit harbors a pair of thumb domain cysteines at the finger-thumb domain interface that are not likely to be disulfide-bonded. These data suggest that reactive cysteines lie at the dynamic finger-thumb interfaces of the ␣ and ␥ subunits.

Finger domain cysteines in the ␣ subunit are near the fingerthumb domain interface
A peptide derived from the furin-released autoinhibitory fragment of the ␣ subunit decreased channel currents by binding at the finger-thumb domain interface of the ␣ subunit (19).
Previously, we recapitulated this effect in the absence of the inhibitory peptide by constraining the distance between the finger domain (using residue 171 in helix 1; red circle in Fig. 1A) and the thumb domain (blue circle in Fig. 1A) of the ␣ subunit using short methanethiosulfonate (MTS)-based crosslinkers (20). Inhibition by MTS crosslinkers was length-dependent. To further examine the dynamics at the finger-thumb domain interface, we aimed to induce a direct disulfide bridge between cysteines introduced at positions 171 and 474 in the ␣ subunit using the oxidizing agent copper phenanthroline (Cu-Phe) (21). However, when we applied 4 M Cu-Phe for 15 s to the ␣Y474C single mutant co-expressed with wildtype ␤ and ␥ subunits in Xenopus oocytes, we observed a modest inhibition of channel currents (Fig. 2). Cu 2ϩ alone had no effect on ␣Y474C currents ( Fig. 2A, cyan traces), reducing currents by 5 Ϯ 2% (n ϭ 7, p ϭ not significant by paired t test). Inhibition by Cu-Phe reversed only upon the addition of the reducing agent DTT ( Fig. 2A), suggesting a covalent modification of the channel. This contrasted with effects on wildtype currents, which were unaffected by 4 M Cu-Phe or 4 M Cu 2ϩ addition. The absence of an effect with Cu 2ϩ alone is consistent with previous work (22). Notably, extracellular DTT treatment did not affect wildtype currents, consistent with previous studies (20,23,24). Wildtype channel insensitivity to DTT suggests either that native extracellular disulfide bonds are not susceptible to reduction by DTT or that reduction of those bonds does not affect channel function. Although Cu-Phe-induced disulfide bonds may also be DTT-insensitive, DTT reversal after oxidation in the case of ␣Y474C suggests reduction of an induced disulfide bond. We also found that Cu-Phe had no effect on neighboring ␣K476C mutant channels, showing that the Cu-Phe effect is specific to ␣Y474C (Fig. 2, A and C).
We hypothesized that Y474C formed a disulfide bond with an endogenous cysteine in the presence of Cu-Phe. ENaC subunits have cysteine pairs that correspond to all seven disulfidebridged pairs observed in the extracellular domains of the ASIC1 structure (Fig. 1). Functional experiments suggest that ENaC finger-thumb domain interface cysteines these cysteines may be similarly disulfide-bonded in ENaC subunits (14,15). ENaC subunits have an additional pair of conserved cysteines in their finger domains ( Fig. 1 (A and B); Cys-2 and Cys-3). The disulfide bonding status and position of the finger domain cysteine pair is unknown, although our ␣ subunit homology model posited Cys-256 (␣Cys-2) and Cys-263 (␣Cys-3) in a disulfide bond and placed them more than 15 Å from Tyr-474 (10). To test whether Cu-Phe induced a bond between ␣Y474C and either of the ␣ subunit finger domain cysteines, we mutated each to alanine in the background of ␣Y474C and treated oocytes expressing the double mutants with 4 M Cu-Phe. We found that Cu-Phe inhibited ␣C256A,Y474C currents in a DTT-reversible manner, similar to what we observed with the ␣Y474C single mutant and different from wildtype. In contrast, any Cu-Phe effect on ␣C263A,Y474C currents reversed upon washout of the oxidant, and DTT had no effect on measured currents. We note here that we frequently observed Cu-Phe inhibition that reversed upon washout prior to DTT addition in each of the groups tested. This suggests that an unstable conformational change may be induced by Cu-Phe, but this effect is difficult to interpret. These results show that DTTreversible Cu-Phe inhibition of ␣Y474C requires endogenous ␣Cys-263. These results suggest that ␣Cys-263 is in close proximity to the finger-thumb domain interface and is free to form a disulfide bond in the presence of ␣Y474C.

Crosslinking across the ␥ subunit finger-thumb domain interface potentiates ENaC currents
Sequence homology suggests significant structural and functional symmetry between ENaC subunits (25), yet functional differences are clearly evident. For example, the ␣ and ␥ subunits are subject to activating proteolysis, whereas the ␤ and ␦ subunits are not (12,26,27). We hypothesized that the conformations of the channel pore and the ␥ subunit's finger-thumb domain interface were coupled, analogous to the ␣ subunit (20). We introduced a cysteine at position 115 in the ␥ subunit (␥D115C), which is equivalent to site 171 in helix 1 of the ␣ subunit at the finger-thumb domain interface (see Fig. 1A). We previously reported evidence of crosslinking between this site and Y474C at the top of the thumb domain in the ␣ subunit (20). When we perfused bifunctional MTS-2-MTS for 1 min, we observed an increase in channel currents that remained elevated after washout of the crosslinker (Fig. 3A). DTT appeared to slowly reverse the activation, although its slow kinetics made it difficult to distinguish it from channel run-down. Furthermore, the effect was dependent on the length of the crosslinker. Neither MTS-1-MTS nor MTS-3-MTS had an effect on ␥D115C currents (Fig. 3B). In contrast, 1-min treatment of each of these MTS crosslinkers irreversibly inhibited wildtype ENaC currents by an average of 24% and in each case was different from the mutant tested with the same crosslinker. We and others have previously reported this effect on wildtype ENaC and that it is attenuated by amiloride (20,23,28). Our data showing MTS-2-MTS activation of ␥D115C in contrast to inhibition of wildtype channels suggest covalent modification of ␥D115C. We hypothesized that the length-dependent activation and DTT reversibility observed with ␥D115C resulted from crosslinking between ␥D115C on helix 1 in the finger domain and an endogenous cysteine.

ENaC finger-thumb domain interface cysteines
not test ␥D115C,C413A due to poor functional expression. Our results show that MTS-2-MTS activation of ␥D115C requires endogenous ␥Cys-415. The results of these experiments suggest that in the presence of ␥D115C, ␥Cys-415 is free and in close proximity to ␥D115C.

Endogenous ␥ subunit finger domain cysteines are near the finger-thumb domain interface
In the course of performing the experiments shown in Fig. 4, we observed that both the ␥C213A (␥Cys-2) and ␥C220A (␥Cys-3) single mutants responded to bifunctional MTS crosslinkers differently than wildtype (Fig. 5). Whereas wildtype currents were consistently inhibited by each of the MTS crosslinkers tested and were not DTT-sensitive, currents from ␥C213A and ␥C220A channels both responded in a crosslinker length-dependent manner ( Fig. 5 and Table 1). MTS-1-MTS had little effect on ␥C213A and ␥C220A channels. MTS-2-MTS stimulated currents from both mutant channels, which reversed upon DTT addition. This potentiation was in stark contrast to MTS-2-MTS inhibition of wildtype channels. MTS-4-MTS had no effect on either mutant channel; this non-effect was also different from MTS-4-MTS inhibition of wildtype channels. In contrast, MTS-8-MTS inhibited both ␥C213A and ␥C220A in a DTT-insensitive manner, reminiscent of the effect of MTS-8-MTS on wildtype channels. Clearly, the effects of the shorter MTS crosslinkers were length-dependent for ␥C213A and ␥C220A (see Table 1) and slowly reversed with DTT. These data suggest that the shorter MTS compounds bridged the remaining ␥ subunit finger domain cysteine and a nearby endogenous cysteine, whereas MTS-8-MTS may have inhibited mutant channel currents analogously to wildtype channels. Our data suggest that ␥Cys-213 and ␥Cys-220 are ordinarily disulfide-bonded but that when either is mutated to alanine, the other becomes capable of crosslinking to an endogenous cysteine. In support of this notion, when we mutated both ␥Cys-213 and ␥Cys-220 to alanine, the current response to both MTS-2-MTS and MTS-8-MTS reverted to wildtype (Fig. 5).
We hypothesized that like the ␣ subunit finger domain cysteines, the ␥ subunit finger domain cysteines lie near the finger-thumb domain interface and that MTS-2-MTS activation of both ␥C213A and ␥C220A mutant channels may require residue ␥Cys-415 or possibly ␥Cys-429. In the background of each single mutant, we mutated either ␥Cys-415 or ␥Cys-429 to alanine. We then tested the four double mutants with each of the bifunctional MTS crosslinkers (  Table 1). Further, ␥C213A,C415A and ␥C220A,C415A double mutants had attenuated responses to MTS-8-MTS compared with the corresponding single mutant. These data suggest that although MTS-8-MTS inhibition of ␥C213A and ␥C220A was not DTT-reversible, the effect requires endogenous ␥Cys-415. On its own, mutation of ␥Cys-415 to alanine did not change the effects of MTS-2-MTS or MTS-8-MTS on the channel (data not shown). Altogether, our results show that the crosslinker effects on the endogenous ␥ subunit finger domain cysteine mutants (␥Cys-2/Cys-3) depend on the presence of endogenous ␥Cys-415 (␥Cys-11) at the top of the thumb domain.

MTS compound reduction of wildtype ENaC currents and effects on finger-thumb interface mutants are separate phenomena
If inhibition of wildtype ENaC by MTS crosslinkers was due to modification of a pore cysteine that blocks the pore, as we and others have proposed (20,23,28), we would not expect that mutants far from the pore could negate that inhibition. Introduction of a free cysteine at the ␥ subunit finger-thumb domain interface does exactly that, either through introduction of a novel cysteine (␥D115C) or through mutation of one-half of a putative disulfide pair (␥C213A or ␥C220A) (Figs. 3 and 5). This suggests that an alternate mechanism underlies bifunctional MTS compound inhibition of wildtype ENaC. To test Wildtype or D115C ␥ subunits were co-expressed with wildtype ␣ and ␤ ENaC subunits in Xenopus oocytes, and currents were measured by two-electrode voltage clamp at Ϫ100 mV. Oocytes were treated with 10 M MTS-n-MTS for 60 s, followed by 110-Na buffer for 60 s and 10 mM DTT for 90 s. A, representative recordings are shown. MTS-2-MTS inhibited currents of wildtype (p Ͻ 0.05 by paired Student's t test), and subsequent DTT addition had no effect. In contrast, MTS-2-MTS addition increased currents of ␥D115C (p Ͻ 0.0001 by paired Student's t test), and subsequent DTT addition reversed the activation (p Ͻ 0.001 by paired Student's t test). B, the change in current due to MTS-n-MTS was determined by comparing the amiloride-sensitive current at the end of MTS-n-MTS washout with the amiloride-sensitive current before MTS-n-MTS treatment. Bars, mean of each group. **, p Ͻ 0.01; ****, p Ͻ 0.0001 by one-way ANOVA followed by Tukey post hoc analysis.

ENaC finger-thumb domain interface cysteines
whether residues that are involved in MTS compound-mediated inhibition of wildtype ENaC overlap with residues involved in activation in the cases of ␥C213A and ␥C220A, we performed a protection experiment (Fig. 6). Before a 1-min treatment with MTS-2-MTS, we treated wildtype and mutant channels with monofunctional MTS-2 for 1 min or alternatively with vehicle. Whereas mock-pretreated mutants were activated by MTS-2-MTS as seen before in Fig. 5, MTS-2 pretreatment attenuated activation by MTS-2-MTS for both ␥C213A and ␥C220A by 61 and 81%, respectively. In contrast, MTS-2 pretreatment of wildtype channels had no effect on subsequent inhibition by MTS-2-MTS (Fig. 6). These data show that MTS-2 protects the ␥C213A and ␥C220A mutants from reaction with MTS-2-MTS but that MTS-2 cannot protect wildtype ENaC from irreversible inhibition by MTS-2-MTS. These data suggest that the mechanisms underlying wildtype channel inhibition and ␥C213A or ␥C220A activation by MTS compounds are distinct.

Mutation of finger-thumb domain interface cysteines reduces channel expression at the cell surface
We observed reduced currents for channels with mutated finger-thumb domain interface cysteines. Changes in wholecell currents may reflect changes in the number of channels at the cell surface and/or the biophysical characteristics of those channels. To determine whether these mutations affected the number of channels at the cell surface, we incorporated a ␤-subunit with an extracellular FLAG tag to perform a chemiluminescence-based assay. One day after cRNA injection of oocytes, we measured amiloride-sensitive currents by two-electrode voltage clamp at Ϫ100 mV (Fig. 7A). Currents were greatly reduced for ␣C256A,Y474C and ␣C263A,Y474C, moderately reduced for ␥C415A and ␥C429A, and modestly reduced for ␥C220A. Two days after cRNA injection, we used an anti-FLAG antibody to quantify channels on the cell surface (Fig. 7B). Surface expression measurements largely paralleled current measurements. Similar to the effect on currents, ␣C256A,Y474C and ␣C263A,Y474C exhibited the greatest reductions in surface expression. Surface expression of ␥C213A was modestly reduced where currents were unchanged, and surface expression of ␥C429A was unchanged where currents were moderately reduced. These data suggest that changes in cell-surface expression largely account for the differences in currents that we observed but that there are probably differences at the single-channel level as well.

Finger-thumb domain interface cysteines do not mediate inhibition by SIN-1, a peroxynitrite generator
Our data suggest that neither ␣Cys-256/␣Cys-263 (␣Cys-2/ Cys-3) in the finger domain nor ␥Cys-415/␥Cys-429 (␥Cys-11/   in the thumb domain are disulfide-bonded. That their modification can modulate channel activity raises the possibility that oxidants or reactive oxygen species can directly regulate ENaC through these sites. Reactive oxygen species have been shown to regulate ENaC activity (29 -31). DuVall et al. reported that peroxynitrite produced from SIN-1 inhibited currents from ENaC expressed in oocytes (32). SIN-1 reacts at physiological pH to produce nitric oxide and superoxide, which subsequently react to form peroxynitrite. SIN-1 treatment of oocytes expressing ENaC rapidly inhibited amiloride-sensitive currents (76 Ϯ 16%) (Fig. 8, A and B). Subsequent removal of SIN-1 from the perfusate rapidly restored currents to 77 Ϯ 12%  (open circles). B, the effect of MTS-n-MTS was determined by comparing the amiloride-sensitive current at the end of MTS-n-MTS washout to the amiloride-sensitive current before MTS-n-MTS treatment. Bars, mean of each group. Data were analyzed by two-way ANOVA, which indicated differences between channels and between crosslinkers (p Ͻ 0.0001). We compared individual pairs using Tukey's multiple comparison test. Indicated in B are comparisons between ␥C213A or ␥C220A and wildtype channels with a given MTS compound (left column) and between double mutants bearing ␥C415A or ␥C429A and the corresponding single mutants, ␥C213A or ␥C220A, with a given MTS compound (across the top or center rows, respectively). *, p Ͻ 0.05; ****, p Ͻ 0.0001. Comparisons of the effects of different crosslinkers on a given channel are shown in Table 1.

ENaC finger-thumb domain interface cysteines
of their starting values (Fig. 8, A and C). We then tested alanine mutants of ␣Cys-256, ␣Cys-263, ␥Cys-415, and ␥Cys-429 to determine whether SIN-1-dependent inhibition required cysteine residues at any of these sites. For each of the mutants tested, the reduction of current in the presence of SIN-1 was similar to wildtype. Nor did any of the mutants tested reduce the sustained inhibition observed after SIN-1 removal. On the contrary, ␣C256A increased inhibition observed after SIN-1 removal, compared with wildtype. Our data suggest that SIN-1-dependent ENaC inhibition does not require any of the cysteines tested here.

Discussion
Two groups previously identified several disulfide pairs in the extracellular domains of the ␣, ␤, and ␥ subunits using a double-mutant cycle approach (14,15). In accordance with these results, the structure of homologous ASIC1 showed that the 14 extracellular cysteines that are conserved between ENaC and ASIC subunits formed seven disulfide bonds (11). ENaC subunits have a unique pair of cysteines in the divergent finger domain. Experiments by Sheng et al. (14) to determine whether the conserved finger domain cysteines are disulfide-bonded to each other were either inconclusive or unsupportive of the presence of such a bond. Regarding the ␥Cys-415/␥Cys-429 (␥Cys-11/Cys-12) pair, previous experiments suggested a disulfide bond for the equivalent pairs in the ␣ (14, 15) and ␤ subunits (15). However, neither study concluded that ␥Cys-415 and ␥Cys-429 are disulfide-bonded. Indeed, Firsov et al. (15) noted functional asymmetry regarding this cysteine pair in the ␥ subunit.
For the ␣ subunit, our results suggest that ␣Cys-263 (␣Cys-3) is near the finger-thumb domain interface and is free to form a disulfide bond with an introduced cysteine. We cannot rule out the possibility that ␣Cys-256 (␣Cys-2) and ␣Cys-263 (␣Cys-3) are a disulfide pair in the absence of an introduced cysteine in the thumb domain (e.g. ␣Y474C). In contrast to the endogenous cysteines in the ␣ subunit finger domain, our results suggest that ␥Cys-213 (␥Cys-2) and ␥Cys-220 (␥Cys-3) are disulfidebonded. The removal of one by mutagenesis renders channel currents susceptible to cysteine-reactive MTS reagents. Furthermore, our data suggest that ␥Cys-415 (␥Cys-11) and ␥Cys-429 (␥Cys-12) are not disulfide-bonded, at least in the context of the ␥D115C, ␥C213A, or ␥C220A mutants, in accord with earlier studies (14,15). This was somewhat surprising because earlier studies were ambiguous on the status of the ␥Cys-415/ ␥Cys-429 pair, these residues are conserved across the protein family, and the equivalent sites in ASIC1 are disulfide-bonded (11). This finding suggests that thiol-reactive molecules could directly regulate the channel via an available cysteine near the finger-thumb interface of either the ␣ or ␥ subunits. We tested Table 1 Crosslinker

length-dependent (MTS-n-MTS) effects on ␥ subunit cysteine mutants near the finger-thumb domain
For each channel in Fig. 5, we compared the effect of using a crosslinker of a given length with all other crosslinkers tested. The number of carbon atoms between MTS functional groups for the tested pair is indicated. Pairwise comparisons were performed by two-way ANOVA followed by Tukey's post hoc test. *, p Ͻ 0.05, **, p Ͻ 0.01, ***, p Ͻ 0.001, ****, p Ͻ 0.0001. Figure 6. Monofunctional MTS-2 protects mutant channels but not wildtype channels from bifunctional MTS-2-MTS. Mutant or wildtype ␥ subunits were co-expressed with wildtype ␣ and ␤ ENaC subunits in Xenopus oocytes. A, ENaC currents were measured using two-electrode voltage clamp at Ϫ100 mV. Oocytes were mock-treated or treated with 10 M MTS-2 for 60 s followed by washout with 110-Na buffer for 60 s. Oocytes were then treated with 10 M MTS-2-MTS for 60 s followed by washout with 110-Na buffer for 60 s. B, the effect of MTS-2-MTS treatment after mock or MTS-2 treatment was determined by comparing current after washout of MTS-2-MTS with current just before MTS-2-MTS addition. Bars, mean. **, p Ͻ 0.01; ****, p Ͻ 0.0001, by Student's t test.

ENaC finger-thumb domain interface cysteines
SIN-1, a nitric oxide and peroxynitrite generator that has been reported to inhibit ENaC currents in oocytes (32,33). We found that inhibition rapidly reversed when we removed SIN-1, suggesting either a non-covalent interaction or an unstable modification of the channel by SIN-1 or its decomposition products. We also found that none of the cysteines at the finger-thumb domain interface that our data suggest are free were required for SIN-1-dependent inhibition.
We also observed functional asymmetry between the ␣ and ␥ subunits. We previously reported that constraining the fingerthumb interface of the ␣ subunit using short bifunctional MTS crosslinkers inhibited ENaC (20). In contrast, here we show that constraining the finger-thumb interface through a number of different sites on the ␥ subunit finger domain using a short bifunctional MTS crosslinker activated ENaC. The disparity may reflect important differences in the experiment (e.g. involvement of endogenous cysteines in the ␥ thumb domain that are close but do not align with the thumb domain site we used in the ␣ subunit). Alternatively, equivalent movements at the finger-thumb interface in the ␣ and ␥ subunits may not have equivalent effects on the conformation of the channel pore. For example, the ␥ subunit's TM1-␤1 linker, which may transduce conformational changes from the thumb domain to the pore (see Fig. 1A), is 2 residues shorter than those of either the ␣ or ␤ subunits. We previously used normal mode analysis to examine motions in the ectodomains that resonated with gating-like motions in the pore (10). Despite using a symmetric ␣ 3 ENaC homology model with or without the autoinhibitory tract in all subunits, we observed asymmetry in the finger and thumb domains. Plainly, a channel composed of distinct but related subunits could give rise to functional asymmetry. Other ENaC functional properties demonstrate asymmetry as well, including proteolytic activation (27), Cl Ϫ inhibition (34), Na ϩ self-inhibition (35,36), and those relating to the pore (37,38).
We have revised our model of the ENaC ␣ subunit extracellular domains to account for our data. Our results suggest that ␣Cys-263 and ␣Cys-256 are not disulfide-bonded. Our data also suggest that ␣Cys-263 (␣Cys-2) lies close to ␣Tyr-474 in the finger-thumb domain interface, in contrast to our previous Extracellular FLAG-epitope tagged ␤ subunits (␤ F ) were co-expressed with wildtype and mutant ENaC subunits, as indicated. A, 1 day after cRNA injection of oocytes, amiloride-sensitive (10 M) currents were measured using a two-electrode voltage clamp at Ϫ100 mV. B, 2 days after injection, the remaining oocytes were placed on ice, and anti-FLAG antibodies were used to detect channels at the cell surface by chemiluminescence. Bars, mean for each group. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; ****, p Ͻ 0.0001 versus ␣␤ F ␥ by one-way ANOVA followed by Tukey post hoc analysis.

Figure 8. SIN-1-mediated inhibition of ENaC does not require free cysteines at finger-thumb interface of the ␣ or ␥ subunits. Mutant
ENaC subunits were co-expressed with wildtype cognate subunits in Xenopus oocytes, and currents were measured by two-electrode voltage clamp at Ϫ100 mV. Solutions of 1 mM SIN-1 in 110-Na buffer were perfused for 10 min, preceded and followed by brief pulses with 10 M amiloride. A, representative recordings are shown. B, rapid inhibition by SIN-1 was calculated by comparing the steady-state current in the presence of SIN-1 with the amiloride-sensitive current before adding SIN-1. C, sustained inhibition was calculated by comparing the steady-state current after washing out SIN-1 with the amiloride-sensitive current before adding SIN-1. *, p Ͻ 0.05 by one-way ANOVA followed by Dunnett's post hoc test.

ENaC finger-thumb domain interface cysteines
model. Our earlier model was based on homology to ASIC1 with the exception of the finger domain (10). We constrained the finger domain using predicted secondary structure and constraints derived from functional experiments examining inhibition by an ␣ subunit-derived inhibitory peptide. These included scanning mutagenesis and double-mutant cycle experiments. We have subsequently described crosslinking experiments here (Fig. 2) and in previous reports (10,20,24). These allow for a straightforward derivation of new structural constraints. We have included these ( Table 2) and generated a new model of the ENaC ␣ subunit extracellular domains using Modeler (39). Several of the new constraints provide more specificity than those based on scanning mutagenesis. As a result, we removed previously derived constraints that included residues Glu-174, Ser-269, and Ser-270, leaving a total of 49 experimentally derived constraints. We generated 20 models (Fig. 9), and scored each by summing the violation distances of experimentally derived constraints (Fig. 9B). The best scoring models satisfy nearly all of the constraints that we imposed, with nine models exhibiting Ͻ1 Å of cumulative violations. Superimposing the models using backbone C ␣ atoms outside of the finger domain illustrates that the structure of the finger domain is better constrained near the finger-thumb interface that defines the N-terminal end of the inhibitory peptide binding pocket as compared with the C-terminal end of the peptide-binding pocket (Fig. 9A) or residues near either of the furin cleavage sites. Our best scoring model places ␣Cys-263 close to ␣Tyr-474 (Fig. 9, C and D), consistent with our data (see Fig. 2).
In summary, our data suggest that the finger domain cysteines in both the ␣ and ␥ subunits lie close to the finger-thumb domain interface. In the ␣ subunit, they are probably free. In the ␥ subunit, they are probably disulfide-bonded. In the ␥ subunit, the terminal pair in the thumb domain cysteine ladder is probably not disulfide-bonded. These free cysteines lie at dynamic domain interfaces and may play a yet undefined role in channel regulation. In contrast to constraining the ␣ subunit fingerthumb domain interface, constraining the ␥ subunit finger-thumb interface activated the channel, suggesting important dynamic differences between the subunits.

Experimental procedures
Plasmids and site-directed mutagenesis pBluescriptSK (Ϫ) (Stratagene, San Diego, CA) vectors containing ␣, ␤, or ␥ mouse ENaC subunits were previously described (40). Site-directed mutagenesis was performed with the QuikChange II XL site-directed mutagenesis kit (Agilent, Santa Clara, CA). Primers were designed using QuikChange Primer Design (Agilent) and were obtained from Integrated DNA Technologies (Coralville, IA) to modify mouse ENaC subunits. Direct sequencing was used to confirm constructs.

ENaC functional expression in Xenopus oocytes
Plasmids were linearized with restriction endonucleases (New England Biolabs, Ipswich, MA) at 37°C overnight and  Figure 9. Model of the ␣ subunit extracellular domains. Modeler was used to create a homology model of the ENaC ␣ subunit extracellular domains, using cASIC1 (PDB code 4NYK) as a template. Distance constraints derived from functional data and predicted secondary structure were used to constrain the finger domain (green), which shares 8% sequence identity with cASIC1. A, overlay of nine models with Ͻ1 Å of cumulative distance violations of constraints derived from functional data, shown as a schematic. Models were superposed using non-finger domain residues: palm (gray), ␤-ball (pink), thumb (yellow), and knuckle (tan). The overlay shows that the highest-scoring models were better constrained at the finger-thumb domain interface near the N terminus of the bound inhibitory peptide (fuchsia, see center panel) and less well constrained near the C terminus of the bound inhibitory peptide and near the furin cleavage sites. B, plot of models sorted by cumulative violation of distance constraints derived from functional data. 9 of 20 models constructed showed less than 1 Å of cumulative violations. C, best scoring model is shown as a schematic, bound peptide is shown as sticks, and disulfide bonds are shown as orange sticks. D, close-up of best scoring model finger-thumb interface shown as a schematic with positions of key residues highlighted and shown as spheres.
ENaC finger-thumb domain interface cysteines purified using the QIAquick PCR purification kit (Qiagen, Hilden, Germany). Linearized plasmids were transcribed into RNA using the mMessage mMachine T3 transcription kit (Invitrogen) and purified using the RNeasy MiniElute cleanup kit (Qiagen). Oocytes from Xenopus laevis were harvested and defolliculated using type II collagenase (Sigma-Aldrich). 1-4 ng of RNA per ENaC subunit was injected into stage V or VI Xenopus oocytes using a Nanoject II (Drummond, Broomall, PA). Xenopus oocytes were stored in modified Barth's saline (MBS) at 18°C. The protocol for harvesting oocytes from X. laevis was approved by the University of Pittsburgh's institutional animal care and use committee.

Measurement of ENaC currents
ENaC current was measured 20 -30 h postinjection using the two-electrode voltage clamp technique using a GeneClamp 500B voltage clamp amplifier (Molecular Devices, Sunnyvale, CA) and pClamp version 10.2 software (Molecular Devices). A 20-l recording chamber (AutoMate Scientific, Berkeley, CA) was used with perfusion (3-5 ml/min) controlled by an eightchannel pinch valve system (AutoMate Scientific). Measurements of ENaC currents were made in Na-110 buffer (110 mM NaCl, 2 mM KCl, 2 mM CaCl 2 , and 10 mM HEPES (pH 7.4)). At the end of each experiment, currents were measured in 10 M amiloride (stock solution of 100 mM amiloride in DMSO stored at room temperature diluted into Na-110 buffer) to determine the amiloride-sensitive component of the current. This was defined as the ENaC-mediated current.

ENaC cell-surface expression
ENaC cell-surface expression in oocytes was measured as described previously (41)(42)(43). Mutant ␣ or ␥ subunits were coexpressed with a ␤ subunit with an extracellular FLAG tag. One day following cRNA injection of oocytes, amiloride-sensitive currents were measured. Two days following cRNA injection of oocytes, oocytes were placed on ice and incubated for 30 min in antibiotic-free MBS and 1% bovine serum albumin (MBS/BSA) and then for 1 h with MBS/BSA with 1 g/ml mouse anti-FLAG antibody (M2; Sigma). After wash with MBS/BSA, cells were incubated for 1 h with 1 g/ml horseradish peroxidaseconjugated goat anti-mouse IgG (Jackson ImmunoResearch, West Grove, PA). After wash, oocytes were transferred to a 96-well plate. SuperSignal ELISA Femto maximum sensitivity substrate (100 l; Thermo Scientific, Rockford, IL) was added to each well. Chemiluminescence was quantified using a GloMax-Multiϩ detection system (Promega, Madison, WI). The mean value measured from four wells with no cells was subtracted from the measurements of cells with ENaC-injected oocytes.

Statistical analyses
Comparisons between groups were performed by ANOVA followed by a Tukey multiple-comparison test, by Student's t test, or by a Kruskal-Wallis rank sum test followed by the Dunn test for multiple comparisons with p values adjusted with the Benjamini-Hochberg method, as indicated. We used R (R: A Language and Environment for Statistical Computing, R Foundation for Statistical Computing, Vienna, Austria) or Prism version 7 (GraphPad, Inc., La Jolla, CA) to perform statistical calculations. Adjusted values of p Ͻ 0.05 were considered significant.

Homology modeling
Modeling was performed using MODELLER version 9.13 (39). We used the ENaC ␣ subunit-cASIC1 alignment described previously (10) and the cASIC1 structural model (PDB code 4NYK). In addition to the constraints we previously used, we added new constraints derived from crosslinking experiments reported here and in previous work (see Table 2) (10,20,24). We constrained distances of C ␣-C␣ pairs using an upper bound constraint dependent on the crosslinking agent used: Cu-Phe, 7.  19.61 Å, calculated using the Molecule Calculator (44). Using our previous model (10) as an initial input, we generated 20 models, which were scored by summing deviations greater than the set distance of the upper bound constraint. We visually inspected the highest scoring models for structural violations (e.g. knots) and selected a high-scoring model with minimal violations. Superpositioning models and root mean square deviation calculations were performed using Swiss pdb viewer version 4.1.0_OSX (45). Molecular figures were generated using PyMOL (Schroedinger, LLC, New York, NY).