Inhibition of the epithelial sodium channel (ENaC) by connexin 30 involves stimulation of clathrin-mediated endocytosis

Mice lacking connexin 30 (Cx30) display increased epithelial sodium channel (ENaC) activity in the distal nephron and develop salt-sensitive hypertension. This indicates a functional link between Cx30 and ENaC, which remains incompletely understood. Here, we explore the effect of Cx30 on ENaC function using the Xenopus laevis oocyte expression system. Coexpression of human Cx30 with human αβγENaC significantly reduced ENaC-mediated whole-cell currents. The size of the inhibitory effect on ENaC depended on the expression level of Cx30 and required Cx30 ion channel activity. ENaC inhibition by Cx30 was mainly due to reduced cell surface ENaC expression resulting from enhanced ENaC retrieval without discernible effects on proteolytic channel activation and single-channel properties. ENaC retrieval from the cell surface involves the interaction of the ubiquitin ligase Nedd4-2 with PPPxY-motifs in the C-termini of ENaC. Truncating the C- termini of β- or γENaC significantly reduced the inhibitory effect of Cx30 on ENaC. In contrast, mutating the prolines belonging to the PPPxY-motif in γENaC or coexpressing a dominant-negative Xenopus Nedd4 (xNedd4-CS) did not significantly alter ENaC inhibition by Cx30. Importantly, the inhibitory effect of Cx30 on ENaC was significantly reduced by Pitstop-2, an inhibitor of clathrin-mediated endocytosis, or by mutating putative clathrin adaptor protein 2 (AP-2) recognition motifs (YxxФ) in the C termini of β- or γ-ENaC. In conclusion, our findings suggest that Cx30 inhibits ENaC by promoting channel retrieval from the plasma membrane via clathrin-dependent endocytosis. Lack of this inhibition may contribute to increased ENaC activity and salt-sensitive hypertension in mice with Cx30 deficiency.


Edited by Mike Shipston
Mice lacking connexin 30 (Cx30) display increased epithelial sodium channel (ENaC) activity in the distal nephron and develop salt-sensitive hypertension. This indicates a functional link between Cx30 and ENaC, which remains incompletely understood. Here, we explore the effect of Cx30 on ENaC function using the Xenopus laevis oocyte expression system. Coexpression of human Cx30 with human αβγENaC significantly reduced ENaC-mediated whole-cell currents. The size of the inhibitory effect on ENaC depended on the expression level of Cx30 and required Cx30 ion channel activity. ENaC inhibition by Cx30 was mainly due to reduced cell surface ENaC expression resulting from enhanced ENaC retrieval without discernible effects on proteolytic channel activation and singlechannel properties. ENaC retrieval from the cell surface involves the interaction of the ubiquitin ligase Nedd4-2 with PPPxY-motifs in the C-termini of ENaC. Truncating the Ctermini of βor γENaC significantly reduced the inhibitory effect of Cx30 on ENaC. In contrast, mutating the prolines belonging to the PPPxY-motif in γENaC or coexpressing a dominant-negative Xenopus Nedd4 (xNedd4-CS) did not significantly alter ENaC inhibition by Cx30. Importantly, the inhibitory effect of Cx30 on ENaC was significantly reduced by Pitstop-2, an inhibitor of clathrin-mediated endocytosis, or by mutating putative clathrin adaptor protein 2 (AP-2) recognition motifs (YxxФ) in the C termini of βor γ-ENaC. In conclusion, our findings suggest that Cx30 inhibits ENaC by promoting channel retrieval from the plasma membrane via clathrin-dependent endocytosis. Lack of this inhibition may contribute to increased ENaC activity and salt-sensitive hypertension in mice with Cx30 deficiency.
The epithelial sodium channel (ENaC) belongs to the ENaC/ degenerin family of ion channels. ENaC is a heterotrimer consisting of three homologous subunits (αβγ) (1). Each subunit consists of two transmembrane domains connected by a large extracellular loop and has short intracellular N-and Ctermini. In several sodium transporting epithelia, ENaC provides the rate-limiting step for apical sodium entry (2). In particular, ENaC is expressed in the apical membrane of principal cells in the aldosterone-sensitive distal nephron (ASDN). This nephron segment consists of the late distal convoluted tubule (DCT2), the connecting tubule (CNT), and the collecting duct (CD). The regulation of ENaC activity in the ASDN plays a key role in fine-tuning renal sodium excretion and hence in long-term control of arterial blood pressure (3,4). ENaC activity in the ASDN critically depends on channel abundance in the apical membrane, which is determined by the balance between channel insertion into the plasma membrane and endocytic channel retrieval (4)(5)(6)(7). An essential hormonal regulator of ENaC activity in the ASDN is the mineralocorticoid aldosterone, but aldosterone-dependent and -independent ENaC regulation has been described in mouse distal nephron (8)(9)(10). The mechanisms involved in mediating the stimulatory effect of aldosterone on ENaC are highly complex and probably include increased forward trafficking of ENaC (11,12) and reduced channel retrieval resulting in increased channel expression at the cell surface. There is good evidence that phosphorylation of the ubiquitinprotein ligase Nedd4-2 by serum and glucocorticoid-induced kinase 1 (Sgk1) mediates at least in part the stimulatory effect of aldosterone (13,14). The C-termini of all three ENaC subunits contain proline-rich PPPxY (PY) motifs. Nedd4-2 interacts with PY-motifs via its WW-domains, promoting the ubiquitination, endocytosis, and proteasomal degradation of the channel (13,(15)(16)(17)(18)(19)(20)(21). Sgk1 is stimulated by aldosterone, and phosphorylation of Nedd4-2 by Sgk1 reduces the affinity of the PPPxY-motif to Nedd4-2, which attenuates channel retrieval from the cell surface, thereby increasing channel surface expression (5,7,20). The mechanisms by which Nedd4-2-dependent ubiquitination promotes ENaC internalization are not yet fully understood, but probably include clathrin-mediated endocytosis (22)(23)(24). It has been suggested that epsin, which has both ubiquitin-interacting and clathrinbinding motifs, may serve as a link between ubiquitinated ENaC subunits and the clathrin-dependent endocytic machinery (23). Additionally, clathrin-interacting proteins may directly bind ENaC and induce ENaC internalization. In particular, it has been shown that ENaC interacts with the μ-subunit of the clathrin adaptor protein 2 (AP-2) (23,24). AP2 is a multisubunit complex formed of α, β, μ, and σ-subunits, which functions as a major hub in clathrin-mediated endocytosis binding clathrin itself and a plethora of accessory and cargo proteins (25)(26)(27). It has been shown that μAP2 initiates clathrin-mediated endocytosis by binding to a YxxΦmotif present in numerous transmembrane proteins (26,28,29). Interestingly, C-termini of all three ENaC subunits contain YxxΦ-motifs, which partially overlap with the PPPxY motifs. It has been proposed that ENaC is regulated by two independent retrieval pathways, one involving Nedd4-2 and the other depending on μAP2 as key component (30). However, the functional role of the YxxΦ internalization motif in ENaC endocytosis remains to be elucidated. In addition to aldosterone and several other hormones, numerous extracellular and intracellular factors including regulatory proteins can modulate ENaC activity in the ASDN. They include proteases, which activate ENaC by cleaving the channel at specific sites in the extracellular domains of the αand γsubunit. This releases inhibitory domains and probably causes a conformational change leading to channel activation (31,32).
Importantly, Cx30 is expressed in the apical membrane of distal tubular epithelial cells in rodents suggesting that it may function there as a hemichannel (33). Apical Cx30 expression in the distal tubule was found to be upregulated by high-salt diet suggesting that Cx30 may be involved in the regulation of salt absorption in the distal nephron (33). Further studies supported this hypothesis. Cx30 knockout mice demonstrated impaired pressure natriuresis and salt-sensitive hypertension, which can be prevented by the ENaC inhibitor benzamil. This supports the hypothesis that Cx30 and ENaC may functionally interact in the distal nephron (34). Indeed, it has been shown that the open probability of ENaC and the number of active channels were increased in CCD of Cx30 −/− mice compared with wild-type mice (36). This may be due to a tonic inhibitory effect of Cx30 on ENaC possibly mediated by ATP release and purinergic signaling (47). Absence of this tonic inhibition in Cx30 −/− mice is thought to be responsible for increased ENaC activity. We hypothesized that additional molecular mechanisms may be involved in mediating the inhibitory effect of Cx30 on ENaC.
The aim of this study was to investigate this further and to elucidate the underlying mechanisms of ENaC inhibition by Cx30 using the Xenopus laevis oocyte expression system.

Functional expression of human Cx30 in X. laevis oocytes
In a first set of experiments, we established the functional expression of human Cx30 in X. laevis oocytes. It has been reported that connexin hemichannels are inhibited by divalent cations (Ca 2+ , Mg 2+ ) from the extracellular side (40,48,49). Therefore, we hypothesized that currents mediated by heterologously expressed Cx30 hemichannels may be revealed by removing divalent cations from the bath solution to disinhibit the channels. To test this, it was necessary to silence interfering endogenous connexin 38 (Cx38) hemichannels known to be expressed at the cell surface of X. laevis oocytes and to be activated by removal of extracellular Ca 2+ and Mg 2+ (50)(51)(52). As expected, concomitant removal of Ca 2+ and Mg 2+ from the bath solution reversibly stimulated inward whole-cell currents (ΔI Ca 2+ Mg 2+ -Removal ) in noninjected oocytes confirming endogenous expression of functional Cx38 hemichannels (Fig. 1A, upper panels). It has been reported that endogenous expression of Cx38 in oocytes can be efficiently suppressed by antisense oligoDNA (AS Cx38) (50, 51) (see Experimental procedures). Using this established approach, we confirmed that in oocytes injected with AS Cx38, no inward currents were detectable upon Ca 2+ and Mg 2+ removal (Fig. 1A, middle panels). In contrast, in oocytes coinjected with AS Cx38 and cRNA for human Cx30, removal of Ca 2+ and Mg 2+ from the bath solution activated large inward currents consistent with the functional expression of Cx30 hemichannels at the cell surface (Fig. 1A, lower panels). Readdition of Ca 2+ and Mg 2+ rapidly returned the stimulated currents back to baseline. This inhibitory effect of Ca 2+ and Mg 2+ was usually preceded by a brief and variable inward current peak (Fig. 1A, left lower panel; Fig. 3A) probably due to a transient stimulation of Ca 2+ -activated chloride channels known to be endogenously expressed in oocytes (53). On average, inward currents elicited by Ca 2+ and Mg 2+ removal were not detectable in AS Cx38 injected control oocytes and were much higher in oocytes coinjected with AS Cx38 and cRNA for human Cx30 than those in noninjected oocytes expressing endogenous Cx38 (Fig. 1B). Thus, these initial experiments confirmed that endogenous Cx38 can be suppressed efficiently with AS Cx38. Moreover, they demonstrate that the oocyte system is suitable for heterologously expressing functional Cx30 hemichannels, which are present at the cell surface and can be activated by removing divalent cations from the bath solution.
Interestingly, in the presence of divalent cations, baseline inward currents were significantly higher in oocytes coinjected with Cx30 cRNA and AS Cx38 than those in noninjected or AS Cx38-injected oocytes (Fig. 1C). This finding suggests that Cx30 hemichannels are not fully blocked in the presence of physiological concentrations of Ca 2+ (1.8 mM) and Mg 2+ (1 mM) in the extracellular solution but exhibit a small residual channel activity under these conditions. To investigate whether these baseline inward currents in Cx30 expressing oocytes are carried by Na + , we substituted Na + in the bath solution with the large organic cation NMDG + . In Cx30 expressing oocytes (coinjected with AS Cx38), the baseline inward currents at a holding potential of −60 mV were significantly reduced by replacing Na + with NMDG + (Fig. 2A). This indicates that the current is mainly carried by Na + as expected for a Cx30-mediated current. This is further supported by the leftward shift of the reversal potential from −15.3 ± 0.7 to −35.1 ± 0.8 mV (p < 0.001) observed upon switching from Na + to NMDG + in the bath solution (Fig. 2C). In contrast, in control oocytes without Cx30 (but injected with AS Cx38), replacing Na + by NMDG + had almost no effect on the magnitude of baseline inward currents (Fig. 2, B and D). Analysis of the reversal potential shifts upon switching from Na + to NMDG + -containing solution in control oocytes revealed a marginal leftward shift from −30.4 ± 0.6 mV to −35.2 ± 0.7 mV (p < 0.001; Fig. 2D). The magnitude of this shift was significantly smaller than that observed in oocytes expressing Cx30 (−4.8 ± 0.6 in control oocytes versus −19.8 ± 0.9 mV in Cx30-expressing oocytes; p < 0.001). This indicates that the Na + conductance of control oocytes is negligible. Thus, expression of Cx30 leads to a significant increase of a baseline cation conductance in oocytes. To estimate the cation permeability ratio (P Na :P NMDG ) for Cx30, we had to correct for the contribution of endogenous oocyte channels to the observed currents. Therefore, we subtracted the average whole-cell current values measured in control oocytes at each holding potentials from the corresponding individual wholecell current values measured in oocytes from the same batch expressing Cx30. From the resulting I/V curves, we determined the reversal potential shifts due to the replacement of Na + in the bath solution by NMDG + (Fig. 2E). From the average shift (−33.1 ± 2.2 mV), we estimated a permeability ratio P Na :P NMDG of about 1.0 : 0.27. To the best of our knowledge, there are no data reported in the literature regarding the cation selectivity of Cx30 hemichannels. However, our estimate is in good agreement with previously published data on the ion selectivity of other connexin hemichannels, e.g., Cx50 or Cx38 (40,54,55).
Next, we tested whether the Na + conductance in Cx30-expressing oocytes can be blocked by the nonspecific connexin inhibitor carbenoxolone (56). Importantly, carbenoxolonesensitive currents (ΔI carb ) were detected in oocytes expressing Cx30 (Fig. 2, F and H) but not in control oocytes injected only with AS Cx38 (Fig. 2, G and H). This supports the Inhibition of the epithelial sodium channel by connexin 30 conclusion that the increased baseline conductance in oocytes injected with Cx30 cRNA is due to Cx30 hemichannels expressed at the cell surface, which have a residual activity even in the presence of divalent cations in the bath solution.
Taken together, these experiments demonstrate that oocytes with silenced endogenous Cx38 are a suitable system for heterologously expressing functional human Cx30 hemichannels. These channels have a small baseline activity, mediate Na + inward currents, and can be disinhibited by extracellular Ca 2+ and Mg 2+ removal. This latter maneuver can be used to assess functional Cx30 expression at the cell surface (see below). In all subsequent experiments, oocytes injected with cRNAs were routinely coinjected with AS Cx38 to suppress endogenous Cx38 expression.

Coexpression of human Cx30 with human ENaC reduces amiloride-sensitive currents
To investigate whether Cx30 can modify ENaC function, we expressed human αβγENaC in oocytes with or without coexpression of human Cx30. Amiloride-sensitive inward currents representative whole-cell current traces recorded in a human Cx30 expressing oocyte (A) or in a control oocyte (B). Oocytes were superfused with standard bath solution (NaCl, open bars) or solution in which Na + was replaced by NMDG + (NMDG-Cl, grey bar). Dashed lines indicate zero current level. Continuous holding potential was −60 mV. Before, during, and after exposure to NMDG + bath solution, voltage step protocols were performed with nine consecutive 2 s voltage steps in 20 mV increments starting with a hyperpolarizing pulse to −120 mV. Insets show overlays of resulting whole-cell current traces obtained at different holding potentials. C and D, Left panels, current data from the final 100 ms portion of each pulse were taken from similar experiments as shown in A and B, respectively, to construct corresponding average I/V curves. Mean and SEM (n = 34; N = 4) are shown. Right panels, reversal potentials obtained from the same experiments summarized in left panels. Individual data points and mean ± SEM are shown. Lines connect data points from individual oocytes. ***p < 0.001; paired Student's t-test. E, Left panel, average I/V plots corrected for endogenous oocyte currents using the data from the experiments summarized in C and D. The average whole-cell current values measured in control oocytes were subtracted from the corresponding individual whole-cell current values measured in oocytes from the same batch expressing Cx30. Right panel, reversal potentials obtained in the same experiments as summarized in the left panel. Individual data points and mean ± SEM are shown. ***p < 0.001; paired Student's t-test. F and G, representative overlays of whole-cell current traces resulting from similar voltage step protocols as described in A and B are shown from an oocyte expressing Cx30 (F) and from a control oocyte (G). For each oocyte current overlays are shown before (−carb) and 1 min after switching to a bath solution containing 1 mM carbenoxolone (+carb). H, current data from the final 100 ms portion of the pulses were taken from similar experiments as shown in F and G. Average I/V plots were constructed using carbenoxolone-sensitive current values (ΔI carb ), which were calculated by subtracting the current values recorded in the presence of carbenoxolone (+carb) from the corresponding current values recorded in its absence (−carb). Mean and SEM (Cx30: n = 21; N = 2; control: n = 27; N = 2) are shown.
Inhibition of the epithelial sodium channel by connexin 30 (ΔI Ami ) and inward currents elicited by extracellular Ca 2+ and Mg 2+ removal (ΔI Ca 2+ Mg 2+ -Removal ) were determined to assess ENaC and Cx30 function, respectively. As shown in Figure 3A, continuous whole-cell current recordings were performed at a holding potential of −60 mV and were started in the presence of amiloride (2 μM) to inhibit ENaC. Washout of amiloride revealed an ENaC-mediated sodium inward current (ΔI Ami ). Reapplication of amiloride returned the current to its initial level. In control recordings from oocytes expressing ENaC without Cx30 (but coinjected with AS Cx38), subsequent removal of Ca 2+ and Mg 2+ had no detectable effect on the current level (Fig. 3A, control). In contrast, in oocytes coexpressing ENaC and Cx30, removal of Ca 2+ and Mg 2+ from the extracellular bath solution in the presence of amiloride increased the inward current by disinhibiting Cx30 hemichannels as described above. As expected, ΔI Ca  where ΔI Ami is the amiloride sensitive current of an individual oocyte coexpressing Cx30 and ENaC, whereas ΔI Ami ðmean; controlÞ is the mean ΔI Ami measured in control oocytes from the same batch but expressing ENaC alone (control). Original data are the same as in C. Mean ± SEM and data points for individual oocytes are shown; **p < 0.01; ***p < 0.001; n.s., not significant; compared with control (markers above the columns) or to another comparison group as indicated; Kruskall-Wallis with Dunn's post hoc test.
Inhibition of the epithelial sodium channel by connexin 30 increased when the amount of injected cRNA for Cx30 was increased from 1 ng to 2 ng or 4 ng (Fig. 3, A and B). This is not surprising, because it can be expected that increasing the amount of injected cRNA will increase channel expression. Importantly, increased expression of Cx30 resulted in progressive reduction of ΔI Ami (Fig. 3, A and C). We calculated the relative inhibitory effect of Cx30 on ENaC (in %) for different amounts of injected Cx30 cRNA. The results are summarized in Figure 3D and demonstrate that Cx30 significantly downregulates ENaC in a dose-dependent manner. A significant inhibitory effect of Cx30 on ENaC was consistently observed in 33 batches of oocytes coinjected with 2 ng of Cx30 cRNA per oocyte and 0.1 ng of cRNA for each ENaC subunit. In these experiments, the relative inhibitory effect of Cx30 on ENaC averaged 54 ± 3% (N = 33, n = 496).
In conclusion, these data demonstrate that coexpression of Cx30 has a robust inhibitory effect on ENaC currents. The inhibitory effect of Cx30 increases with increasing expression of Cx30.
Inhibitory effect of Cx30 on ENaC requires ion channel activity of Cx30 As described above, Cx30 expressing oocytes exhibit a carbenoxolone-sensitive conductance in the presence of physiological concentrations of Ca 2+ and Mg 2+ in the bath solution (Fig. 2). To test whether this channel activity is relevant for the inhibitory effect of Cx30 on ENaC, we incubated oocytes for 48 h after cRNA injection with the Cx30 blocker carbenoxolone (100 μM). In oocytes expressing ENaC and Cx30, preincubation with carbenoxolone prevented the stimulation of Cx30 mediated inward currents upon removal of Ca 2+ and Mg 2+ from the extracellular solution (Fig. 4A, right trace). Interestingly, this inhibitory effect of carbenoxolone preincubation was observed in the absence of carbenoxolone during the current measurements. This may be due to its intracellular accumulation and/or incomplete washout. In contrast, a stimulatory response to Ca 2+ and Mg 2+ removal was readily observed in matched ENaC and Cx30 coexpressing oocytes not treated with carbenoxolone ( Fig. 4B, right trace). Importantly, average ENaC currents were more than twofold larger in carbenoxolone treated oocytes coexpressing ENaC and Cx30 than those in matched oocytes coexpressing ENaC and Cx30 but without carbenoxolone treatment (Fig. 4C). Without coexpression of Cx30, carbenoxolone had no significant effect on ENaC-mediated currents in ENaC expressing oocytes from the same batch (Fig. 4, A and B, left traces, and Fig. 4C). Thus, inhibition of Cx30 by incubation of oocytes in the presence of carbenoxolone significantly reduced the inhibitory effect of Cx30 on ENaC (Fig. 4D). In an additional set of experiments, we demonstrated that in oocytes coexpressing ENaC and Cx30, acute application of carbenoxolone (1 mM) for 1 min had no apparent effect on ΔI Ami . Compared with ΔI Ami in oocytes expressing ENaC alone, ΔI Ami was reduced in oocytes coexpressing ENaC and Cx30 by 70 ± 4% before and by 69 ± 5% after application of carbenoxolone (n.s.; N = 1, n = 10). Thus, acute application of carbenoxolone had no detectable effect on ENaC inhibition by Cx30 coexpression.
These findings indicate that a prolonged ion channel function of Cx30 is needed to cause the observed inhibitory effect of Cx30 coexpression on ENaC and that short-term inhibition of Cx30 by carbenoxolone cannot reverse this effect.
Carbenoxolone is a rather nonspecific drug and may have additional side effects besides blocking Cx30. Therefore, we established an alternative approach to silence the ion channel activity of Cx30. It has been shown for Cx26, a close homologue of Cx30, that introducing a point mutation into the pore-forming transmembrane domain 1 (M34T or M34A) renders Cx26 inactive by constricting the channel pore (57,58). Interestingly, the M34T mutation of Cx26 was found to be associated with hereditary deafness in humans (59,60). We hypothesized that due to the high degree of sequence similarity between Cx26 and Cx30 (approx. 77%), a homologous M34A substitution in Cx30 would also result in nonfunctional channels. Indeed, as shown in Figure 5, A and B, divalent Inhibition of the epithelial sodium channel by connexin 30 cation removal from the bath solution did not elicit an inward current response in oocytes expressing ENaC together with the mutant Cx30 (ENaC+Cx30 M34A ). This lack of response to divalent cation removal is similar to that in control oocytes expressing ENaC alone and confirms the loss-of-function effect of the M34A mutation on Cx30 ion channel function. In contrast, in matched oocytes coexpressing ENaC with wildtype Cx30 (ENaC+Cx30), the usual inward current response to divalent cation removal was detected. Importantly, in contrast to Cx30, Cx30 M34A did not significantly reduce ΔI Ami (Fig. 5, C and D). Western blot analysis revealed that intracellular and cell surface expression of Cx30 M34A was similar to that of Cx30 (Fig. 5, E-H). Thus, the M34A mutation effectively suppresses ion channel function of Cx30 without compromising its level of protein expression or its localization at the cell surface. Taken together these findings support our conclusion that the ion channel function of Cx30 is essential for its inhibitory effect on ENaC. They also argue against the possibility that Cx30 inhibits ENaC simply by competing for the transcription or expression machinery of the oocyte.
It has been reported that decreasing or increasing the extracellular Ca 2+ concentration enhances or reduces ion permeation through Cx30 hemichannels, respectively (61). Therefore, we hypothesized that the inhibitory effect of Cx30 on ENaC may be modified by varying the extracellular Ca 2+ concentration. To test this, we incubated oocytes expressing -Removal values were obtained from similar experiments as shown in A and calculated as described in Figure 1. Mean ± SEM and individual data points for each experiment are shown; ***p < 0.001; n.s., not significant; one-way ANOVA with Bonferroni post hoc test (n = 36, N = 3). C, ΔI Ami were calculated from similar experiments as shown in A as described in Figure 3. Mean ± SEM and data points for individual oocytes are shown; ***p < 0.001; n.s., not significant; Kruskall-Wallis with Dunn's post hoc test. D, relative inhibitory effect of Cx30 on ENaC calculated from the data shown in C essentially as described in Figure 3. Mean ± SEM and data points for individual oocytes are shown; ***p < 0.001; n.s., not significant; Kruskall-Wallis with Dunn's post hoc test. E and F, representative western blots showing cell surface (E) or intracellular (F) expression of C-terminally V5-tagged wild-type Cx30 and mutant Cx30 M34A in oocytes from the same batch as used in A. No specific signal was detected with the anti-V5 antibody in oocytes expressing ENaC alone. To validate separation of cell surface proteins from intracellular proteins, blots were stripped and reprobed using an antibody against β-actin. G and H, densitometric evaluation of Cx30 and Cx30 M34A expression from three similar blots as shown in E and F. In each blot the density value of the Cx30 M34A band was normalized to that of the Cx30 band. Lines connect data points obtained in the same experiment, and mean ± SEM are shown; n.s. not significant; one sample Wilcoxon signed-rank test (N = 3).
ENaC with or without Cx30 for 48 h after cRNA injection in ND9 with a standard (1.8 mM), decreased (0.9 mM), or increased (3.6 mM) Ca 2+ concentration. As shown in Figure 6, A and C, varying the Ca 2+ concentration in the incubation medium had no effect on ΔI Ami in oocytes expressing ENaC without Cx30. In contrast, the inhibitory effect of Cx30 coexpression on ΔI Ami was slightly more pronounced with 0.9 mM and substantially attenuated with 3.6 mM extracellular Ca 2+ compared with its effect under control conditions with 1.8 mM extracellular Ca 2+ (Fig. 6, B and C). Thus, there appears to be an inverse relationship between the Ca 2+ concentration of the incubation solution and the relative inhibitory effect of Cx30 on ENaC (Fig. 6D). As stated above, Cx30 hemichannels are inhibited by extracellular Ca 2+ . Therefore, these finding suggests that the degree of ENaC inhibition by Cx30 correlates with the ion channel activity of Cx30, which is modulated by the extracellular Ca 2+ concentration. However, the inhibitory effect of extracellular Ca 2+ on Cx30 does not rule out the possibility that Cx30 hemichannels mediate Ca 2+ influx and modulate intracellular Ca 2+ signaling. To investigate whether the inhibitory effect of coexpressed Cx30 on ENaC depends on the presence of extracellular Ca 2+ , we substituted Ca 2+ in the incubation solution by Ba 2+ . In oocytes expressing ENaC without Cx30, replacement of Ca 2+ by Ba 2+ had no significant effect on ΔI Ami . Importantly, in oocytes incubated in the presence of Ba 2+ instead of Ca 2+ , the inhibitory effect of Cx30 coexpression on ΔI Ami was not only preserved but even enhanced (Fig. 7, A-D). Thus, extracellular Ca 2+ is not necessary for mediating the inhibitory effect of Cx30 on ENaC.

Cx30 reduces surface expression of ENaC
The inhibitory effect of Cx30 on ENaC may be due to reduced channel expression at the cell surface, reduced singlechannel current amplitude, and/or reduced channel open probability. To investigate whether Cx30 changes ENaC surface expression, we used a chemiluminescence assay based on the detection of a FLAG tag inserted into the extracellular domain of the β-subunit of ENaC (62)(63)(64). As shown in Figure 8A, coexpression of Cx30 reduced ΔI Ami by about 50% indicating that the inhibitory effect of Cx30 on ENaC including a β-subunit with a FLAG tag (ENaC-βFLAG) was similar to that observed for wild-type ENaC (Fig. 3). Importantly, Cx30 caused a substantial decrease of ENaC-βFLAG surface expression (Fig. 8B), which can fully explain the observed 50% reduction of ΔI Ami . To confirm that the chemiluminescence assay can reliably detect a decrease in channel surface expression under our experimental conditions, we performed control experiments in which the amount of cRNA injected per oocyte was reduced by a factor of 3 (1/3 ENaC). As expected, this significantly reduced both ΔI Ami and the chemiluminescence signal, because fewer channels are expressed and reach the cell surface (Fig. 8, A and B). The inhibitory effect of reducing the amount of injected cRNA on ΔI Ami and ENaC surface expression was similar to that observed with Cx30 coexpression. In control oocytes, the luminescent signal was very small, which confirmed that the chemiluminescence assay specifically detected ENaC at the cell surface. Additionally, using western blot analysis of whole-cell lysates, we tested whether Cx30 altered total protein expression of α-ENaC (Fig. 8, C and F), β-ENaC (Fig. 8, D and G) and γ-ENaC subunits (Fig. 8, E and H). A single band corresponding to the expected size of full-length α-, β-, and γ-ENaC was detected at about 95 kDa in ENaC-expressing but not in control oocytes. Additionally, a second band corresponding to the partially cleaved γ-ENaC was observed at about 74 kDa. Importantly, densitometric evaluation of ENaC expression demonstrated that the overall protein expression levels of the three ENaC subunits were not significantly altered by Cx30 coexpression (Fig. 8, F-H). In oocytes from the same batches, the significant inhibitory effect of Cx30 coexpression on ΔI Ami was confirmed (Fig. 8I). Thus, Cx30 significantly reduces ENaC expression at the cell surface without detectable effect on overall ENaC protein expression.

Cx30 stimulates internalization of ENaC
Cx30 may reduce ENaC expression at the cell surface by stimulating ENaC retrieval from the plasma membrane and/or Inhibition of the epithelial sodium channel by connexin 30 by hampering the insertion of newly synthesized or recycled channels into the plasma membrane. It is well established that mechanisms altering the internalization rate of ENaC have a strong impact on the abundance of ENaC at the cell surface (13,(15)(16)(17)(18)(19)(20)(21). Therefore, we hypothesized that Cx30 decreases ENaC cell surface expression by stimulating ENaC retrieval.
To estimate the rate of ENaC retrieval, we used a strategy similar to that described in previous studies (65,66). This strategy is based on converting acutely and irreversibly the channels present at the cell surface into fully active channels and on measuring the subsequent current decline as surrogate parameter for the rate of channel retrieval. For this purpose, we used a mutant β-ENaC subunit carrying a cysteinesubstitution at the so-called degenerin position (β S520C ) and coexpressed it with wild-type αand γ-ENaC with or without additional coexpression of Cx30. This cysteine residue can be covalently modified by application of the sulfhydryl reagent (2-(trimethylammonium)ethyl) methanethiosulfonate bromide (MTSET). This modification by MTSET increases the P o of ENaC at the cell surface close to 1 (67). After MTSET treatment, fully active ENaC channels are continuously retrieved from the cell surface and are gradually replaced by newly inserted unmodified channels with lower P o . This results in a decay of ΔI Ami over time toward the initial current level before MTSET application. Thus, measuring the rate of ΔI Ami decline after MTSET treatment allows to estimate the apparent rate of ENaC retrieval from the plasma membrane.
As illustrated in Figure 9, A and B and summarized in Figure 9C, application of MTSET for 5 min strongly stimulated ΔI Ami in oocytes expressing αβ S520C γENaC with or without Cx30 coexpression. In preliminary experiments, we confirmed that application of MTSET for 5 min was sufficient for ΔI Ami to reach a plateau value (data not shown). Assuming that MTSET increased ENaC P o close to 1, we estimated a baseline ENaC P o averaging 0.2 and 0.3 in oocytes with or without coexpression of Cx30, respectively. These values are in good agreement with previously published baseline P o values for human ENaC expressed in oocytes (68)(69)(70) and with the P o values determined in our single channel recordings and estimated from our experiments with chymotrypsin (see below). The slightly lower P o value for ENaC in oocytes coexpressing Cx30 may indicate a small inhibitory effect of Cx30 on P o . Importantly, the MTSET experiments confirm our conclusion that the large inhibitory effect of Cx30 coexpression on ΔI Ami cannot be attributed to a reduction in P o but must be caused mainly by reduced channel expression at the cell surface. After reaching a maximal value due to exposure to MTSET, ΔI Ami gradually decreased over time in both groups of oocytes. ΔI Ami 70 min after MTSET treatment (ΔI Ami (70 min) ) in oocytes expressing ENaC alone was reduced to 62 ± 5% of its value measured directly after MTSET treatment (ΔI Ami (0 min) ). In contrast, in oocytes coexpressing ENaC and Cx30, ΔI Ami decreased within 70 min to 39 ± 4% of its value after MTSET stimulation. Thus, the relative rate of ΔI Ami decline was significantly higher in oocytes coexpressing ENaC and Cx30 compared with that in oocytes expressing ENaC alone (Fig. 9D). This finding suggests that Cx30 coexpression stimulates ENaC retrieval from the cell surface.
However, to draw this conclusion, it has to be considered that insertion of newly synthesized and recycled ENaC contributes to ΔI Ami and will partially compensate for its decline due to channel retrieval. Thus, the observed decline of ΔI Ami is likely to underestimate ENaC retrieval rate. Moreover, differences in ENaC insertion rates and P o may at least in part account for the different relative rates of ΔI Ami decline in oocytes with and without Cx30 coexpression. Therefore, to estimate more accurately the contribution of ENaC retrieval to the decline of ΔI Ami within 70 min (ΔI Retrieval ), it was necessary to correct for the concomitant increase of ΔI Ami resulting from channel insertion (ΔI Insertion ) during this time period. To make this correction, MTSET was applied again after 70 min assuming that only newly inserted channels can be stimulated by this second MTSET application. As shown in Figure 9, A-D, in both groups of oocytes, reapplication of MTSET increased ENaC currents almost back to the level reached after the first application of MTSET. This indicates that during the This correction revealed that without concomitant channel insertion, ΔI Ami would have been reduced by ENaC retrieval within 70 min to 46 ± 6% (n = 11; N = 3) of ΔI Ami (0 min) in oocytes expressing ENaC alone and to 26 ± 4% (n = 12; N = 3; p < 0.01) in oocytes coexpressing ENaC and Cx30. Using these corrected values, we performed a regression analysis and demonstrated that ΔI Retreival correlated linearly with ΔI Ami (0 min) in both groups of oocytes (Fig. 9, E and F), which indicates that their endocytic machinery was not saturated. Importantly, in oocytes coexpressing ENaC and Cx30, the linear regression coefficient was higher (k = 0.68 ± 0.02; Fig. 9F) than that in oocytes expressing ENaC alone (k = 0.42 ± 0.04; Fig. 9E).
In summary, this further supports the conclusion that Cx30 coexpression significantly stimulates the rate of ENaC retrieval from the plasma membrane, thereby decreasing channel expression at the cell surface.
Cx30 has no effect on proteolytic activation of ENaC To investigate a possible effect of Cx30 on proteolytic ENaC activation, we used chymotrypsin (2 μg/ml) to activate ENaC in oocytes expressing ENaC with or without Cx30. We have previously shown that chymotrypsin in this  β S520C +Cx30, B). To determine ΔI Ami at different time points, oocytes were repeatedly clamped for a short period of time at a holding potential of −60 mV and current traces were recorded as illustrated. Oocytes were unclamped during the remaining time of the recording to minimize sodium loading of the oocytes. Impaling microelectrodes were not removed from the oocyte until the end of the experiment. Application of amiloride (Ami, 2 μM) during each current measurement is indicated by filled bars. Dashed lines indicate zero current level. After determining initial ΔI Ami , oocytes were exposed to MTSET for 5 min by exchanging the bath solution from ND96 with 2 μM amiloride to a solution containing in addition 1 mM MTSET (application of MTSET is indicated by arrows). Before the second current measurement (time 0 min) MTSET was washed out with ND96 containing 2 μM amiloride. ΔI Ami was determined every 10 min as indicated (min: 10, 20, …, 70). Between the measurements, oocytes were maintained in ND96 with 2 μM amiloride. At the end of the recordings, oocytes were exposed to MTSET for a second time. C, summary of ΔI Ami values from similar experiments as shown in the representative traces A and B. Mean ± SEM current values for ENaC-β S520C oocytes (black line and symbols; n = 11; N = 3) or ENaC-β S520C +Cx30 oocytes (red line and symbols; n = 12; N = 3). D, data shown in C were normalized as ratio of ΔI Ami measured at different time points to ΔI Ami measured at time 0 min after incubation with MTSET (ΔI Ami /ΔI Ami (0 min) ). **p < 0.01; *p < 0.05; n.s., not significant; ENaC-β S520C versus ENaC-β S520C +Cx30; two-way ANOVA with Bonferroni post hoc test. E and F, linear regression analysis of ΔI Retrieval versus ΔI Ami (0 min) in oocytes expressing ENaC alone (E) or coexpressing ENaC together with Cx30 (F). Each point represents an individual measurement from the same experiments summarized in C and D. Calculated linear regressions with proportionality coefficients (k) are depicted by dashed lines. For better comparison, the diagonal dashed lines depict the hypothetical linear regression with k = 1. G, normalized ΔI Retrieval (ΔI Retrieval /ΔI Ami (0 min) ) calculated from the same experiments shown in E and F. **p < 0.01; Student's t-test.
Inhibition of the epithelial sodium channel by connexin 30 concentration maximally activates ENaC by proteolytic cleavage, increasing average channel open probability (P o ) from about 0.3 to almost 1 consistent with a threefold increase of ENaC-mediated whole-cell currents (64,68,71). As expected, washout of amiloride revealed larger ENaCmediated baseline currents in oocytes expressing ENaC alone (Fig. 10A) than in oocytes coexpressing Cx30 and ENaC (Fig. 10B) confirming the inhibitory effect of Cx30 on ENaC. Importantly, chymotrypsin stimulated baseline ENaC currents in both groups of oocytes by approximately, threefold (Fig. 10C). This indicates that Cx30 does not interfere with proteolytic ENaC activation and probably has no major effect on channel P o , which was about 0.3 in both groups.

Coexpression of Cx30 does not affect the single-channel current amplitude of ENaC and has no apparent effect on single-channel P o
To test whether Cx30 expression alters the singlechannel current amplitude of ENaC, we performed singlechannel recordings in outside-out patches of oocytes expressing ENaC alone or ENaC together with Cx30. Figure 11, A and B show representative recordings from an outside-out patch excised from an oocyte expressing only ENaC or ENaC with Cx30, respectively. The initial washout of amiloride revealed single-channel activity with up to two apparent channel levels. The single-channel current amplitude was about 0.38 pA, which is typical for ENaC at this holding potential (and corresponds to a single-channel conductance of about 5 pS). Importantly, the average single-channel current amplitude was not significantly different in ENaC versus ENaC+Cx30 expressing oocytes (Fig. 11C). The estimated apparent single-channel P o in these recordings was rather variable, which is a typical finding in single-channel recordings of ENaC. On average, ENaC P o was similar in both groups (Fig. 11D). However, due to the variability of single-channel P o , we cannot exclude the possibility that Cx30 has a minor inhibitory effect on ENaC P o as suggested by our experiments with MTSET (see above; Fig. 9). Collectively, our data indicate that the inhibitory effect of Cx30 is mainly mediated by a reduction of the number of ENaC channels in the plasma membrane and not by an effect on channel P o or singlechannel conductance.

C-termini of βand γ-ENaC are critically involved in Cx30mediated downregulation of ENaC
It is well documented that C-termini of ENaC subunits (βand γENaC in particular) are critically involved in the regulation of channel retrieval from the plasma membrane (4). Therefore, we tested whether the C-termini of α-, β-, or γ-ENaC are involved in the Cx30-mediated downregulation of ENaC. For this purpose, channels with one C-terminally truncated ENaC subunit (α T : P619X, β T : R566X, or γ T : K576X) and two wild-type subunits (α T βγ, αβ T γ, or αβγ T ) were expressed alone or coexpressed with Cx30. The truncations eliminated highly conserved C-terminal sequences (72). As shown in Figure 12A, C-terminal truncations of βor γENaC significantly increased ENaC currents compared with the wild-type channel. Truncation of all three C-termini had a similar stimulatory effect as that of truncating the C-terminus of the γ-subunit alone. These findings are consistent with the well-established concept that channels with C-terminal truncations are internalized and degraded more slowly than wild-type ENaC. Interestingly, ΔI Ami was not increased in oocytes expressing α T ENaC, which is in agreement with previously reported findings that C-terminal truncations of βor γENaC but not of αENaC substantially increased ENaC-mediated currents (15,72). The relative inhibitory effect of Cx30 on ENaC-mediated currents was preserved in oocytes expressing α T βγENaC (42 ± 3%) and was similar to that observed in matched oocytes expressing wild-type ENaC (52 ± 2%; Fig. 12B). In contrast, the relative inhibitory effect of Cx30 on ΔI Ami was significantly reduced in oocytes expressing αβ T γENaC (17 ± 5%) and essentially abolished in oocytes expressing αβγ T ENaC (0 ± 5%) or α T β T γ T ENaC (4 ± 4%; Fig. 12B). In additional control experiments, it was confirmed that in oocytes coexpressing Cx30 and αβγ T ENaC, the Cx30-mediated inward currents elicited by Ca 2+ and Mg 2+ removal were similar to those observed in matched oocytes coexpressing Cx30 and wild-type ENaC (data not shown). Thus, the absence of the inhibitory effect of Cx30 on αβγ T ENaC was not due to impaired expression of Cx30 hemichannels at the cell surface. In conclusion, these data indicate that Cx30-mediated downregulation of ENaC critically depends on the presence of intact C-termini of βand γ-ENaC.

Nedd4-2 is not involved in Cx30-mediated ENaC downregulation
Sequence alignment of ENaC C-termini (Fig. 13) reveals the presence of highly conserved PPPxY-motifs in all three ENaC subunits. The role of these PPPxY-motifs in the regulation of ENaC cell surface expression has been characterized in detail (13,15,16,(19)(20)(21). The PPPxY-motifs are recognized by WW domains of the ubiquitin ligase Nedd4-2, which leads to Nedd4-2 binding to ENaC and subsequent channel ubiquitination and internalization. Importantly, mutating the  , βENaC (R566X), or γENaC (K576X) with corresponding wild-type ENaC subunits (α T βγ, αβ T γ or αβγ T ) or triple-truncated ENaC (α T β T γ T ) in each case with or without Cx30. Individual ΔI Ami values were normalized to the mean ΔI Ami obtained in matched oocytes from the same batch expressing wildtype ENaC without Cx30. Mean ± SEM and data points for individual oocytes are shown; ***p < 0.001; **p < 0.01; n.s. not significant; Kruskall-Wallis with Dunn's post hoc test (67 ≤ n ≤ 153, 5 ≤ N ≤ 7). B, relative inhibitory effect of Cx30 on wild-type (αβγ) or truncated (α T βγ, αβ T γ, αβγ T or α T β T γ T ) ENaC calculated as described in Figure 3 using original data shown in A. ΔI Ami obtained in oocytes expressing wild-type (αβγ) or truncated (α T βγ, αβ T γ, αβγ T , or α T β T γ T ) ENaC together with Cx30 was normalized to the corresponding mean ΔI Ami recorded in matched control oocytes from the same batch expressing wild-type (αβγ) or truncated (α T βγ, αβ T γ, αβγ T or α T β T γ T ) ENaC alone. Mean ± SEM and data points for individual oocytes are shown; ***p < 0.001; n.s., not significant; Kruskall-Wallis with Dunn's post hoc test.

Inhibition of the epithelial sodium channel by connexin 30
PPPxY-motifs in βor γ-ENaC increased ENaC-mediated currents to a similar degree as truncating the corresponding C-termini (15,16). Therefore, we investigated whether mutating the PPPxY-motif in γENaC also abolished the inhibitory effect of Cx30 on ENaC. In oocytes expressing a mutant channel with the three critical proline residues ( 623 PPP 625 ) in the PPPxY-motif of γENaC mutated to alanine (αβγ P623A-P625A ), average ΔI Ami was almost doubled compared with ΔI Ami in matched oocytes expressing wild-type ENaC (Fig. 14A). This confirmed previously reported data and is consistent with the concept that mutating the PPPxY-motif of γENaC impedes the interaction of endogenous Nedd4 with ENaC. However, the relative inhibitory effect of Cx30 on ΔI Ami was not significantly different in oocytes expressing γ P623A-P625A mutant ENaC (49 ± 3%) compared with that observed in oocytes expressing wild-type ENaC (57 ± 2%) (Fig. 14B). This finding suggests that Nedd4-2 is not involved in Cx30-mediated ENaC downregulation. In further experiments, we mutated the tyrosine residue (Y627) in the PPPxYmotif of γENaC to alanine, which is also thought to be critical for Nedd4-2 binding to ENaC (13,(19)(20)(21). In good agreement with this concept, we found that the γ Y627A mutation increased ΔI Ami to a similar extent as the γ P623A-P625A mutation (Fig. 14A). Interestingly, the relative inhibitory effect of Cx30 on γ Y627A mutant ENaC was slightly reduced (41 ± 5%) compared with that on wild-type ENaC (Fig. 14B). Mutating the homologous tyrosine residues in all three ENaC subunits (α Y644A β Y620A γ Y627A ENaC) resulted in even more pronounced reduction of Cx30-mediated effect on ENaC (36 ± 5%). Thus, disruption of the PPPxY-motif by alanine substitutions of the proline residues in γENaC did not significantly alter the inhibitory effect of Cx30 on ENaC. In contrast, substitution of the tyrosine residue by an alanine had a small but significant effect on Cx30-mediated ENaC inhibition. We also investigated the effect of coexpressing catalytically inactive Xenopus xNedd4-CS (bearing a C938S mutation in the Hect domain) on Cx30-mediated ENaC downregulation. As expected from previously reported findings (73), coexpression of xNedd4-CS significantly increased ENaC-mediated currents by about approximately twofold probably due to an inhibitory effect on endogenous Nedd4 (Fig. 14A). However, the relative inhibitory effect of Cx30 on ΔI Ami was not significantly affected by coexpressing xNedd4-CS (52 ± 4%; Fig. 14B). Taken together, these data support the conclusion that Nedd4-2 is not involved in mediating the inhibitory effect of Cx30 on ENaC.
Cx30-mediated ENaC downregulation is significantly reduced by inhibiting clathrin-mediated endocytosis or mutating the YxxΦ-motif in the C-termini of βor γ-ENaC Primary sequence alignment of ENaC C-termini (Fig. 13) reveals that all three ENaC subunits have highly conserved YxxΦ-motifs (where Φ indicates an amino acid with a bulky hydrophobic side chain like leucine), which partially overlap with their PPPxY-motifs. YxxΦ-motifs are known to be involved in binding the μ-subunit of the clathrin adaptor protein 2 (AP-2), which is responsible for clathrin-mediated endocytosis of many cell surface proteins (26,28,29). Thus, the YxxΦ-motifs are likely to be important for clathrinmediated ENaC internalization and may be involved in mediating the inhibitory effect of Cx30 on ENaC.
To explore a possible role of clathrin-mediated endocytosis in Cx30-mediated inhibition of ENaC, we used a pharmacological tool, Pitstop-2, which specifically binds to clathrin, thereby inhibiting clathrin-mediated endocytosis (74). Oocytes expressing ENaC with or without Cx30 were preincubated with Pitstop-2 (30 μM) or with 0.02% dimethyl sulfoxide (mock incubation) for 1 h before determining ΔI Ami . In ENaC expressing oocytes without coexpression of Cx30, Pitstop-2 treatment significantly increased ΔI Ami by about twofold compared with ΔI Ami measured in oocytes after mock incubation (Fig. 15, A and C). This increase of ΔI Ami may be explained by an effective inhibition of clathrin-mediated ENaC retrieval by Pitstop-2 while channel insertion into the plasma membrane continues to occur resulting in a net increase of channel expression at the cell surface. Importantly, in oocytes coexpressing ENaC with Cx30, the relative stimulatory effect of Pitstop-2 on ΔI Ami was much more pronounced with an average increase of about fivefold (Fig. 15, B and C). Moreover, ΔI Ami values in Pitstop-2-treated oocytes coexpressing ENaC and Cx30 reached a level similar to that observed in mocktreated oocytes expressing ENaC alone. Thus, inhibition of clathrin-mediated endocytosis by Pitstop-2 significantly reduced the inhibitory effect of Cx30 on ENaC (Fig. 15D). Taken together these findings indicate that clathrin-mediated endocytosis is critically involved in the downregulation of ENaC by Cx30. Next, we introduced mutations in γENaC resulting in truncations before (γ K626X ) or after (γ A635X ) the YxxΦ-motif. Importantly, deleting the YxxΦ-motif by the γ K626X mutation increased ΔI Ami and significantly reduced the relative inhibitory effect of Cx30 on ENaC (Fig. 16). In contrast, truncating the C terminus after the YxxΦ-motif (γ A635X ) affected neither the basal ΔI Ami nor the relative inhibitory effect of Cx30 on ENaC (Fig. 16). These data indicate that the intact YxxΦ-motif in γENaC is critically important for the Cx30-mediated ENaC inhibition. To investigate this further, we substituted the critical leucine (γ L630 ) to a negatively charged aspartate (γ L630D ). This increased basal ΔI Ami (by about 44%, Fig. 16A) and reduced the degree of Cx30-mediated ENaC inhibition (Fig. 16B). Substitution of the homologous leucine in the YxxΦ-motif of β-ENaC to aspartate (β L623D ) or the simultaneous substitution of both leucine residues in βand γ-ENaC (β L623D γ L630D ) also increased basal ΔI Ami and reduced the inhibitory effect of Cx30 on ENaC (Fig. 16). Thus, our data highlight the crucial role of the C-terminal YxxΦ-motifs of βand γENaC in Cx30-mediated ENaC inhibition. Collectively, our findings demonstrate that enhanced clathrin-mediated channel retrieval is mainly responsible for the inhibitory effect of Cx30 on ENaC.

Discussion
In the current paper, we investigated the functional interaction of human αβγENaC with human Cx30 heterologously expressed in X. laevis oocytes. We made the following observations: 1) coexpression of Cx30 significantly reduced ENaCmediated currents by decreasing channel surface expression due to enhanced channel retrieval; 2) Cx30 had no noticeable effect on proteolytic activation and single-channel properties of ENaC; 3) ENaC inhibition by Cx30 was prevented by inhibiting Cx30 channel function with carbenoxolone or a point mutation; 4) the inhibitory effect of Cx30 depended on the C-termini of βand γENaC without involving a Nedd4-2dependent mechanism; 5) inhibition of clathrin-mediated endocytosis with Pitstop-2 or by mutating putative AP-2 recognition motifs (YxxФ) in the C-termini of βand/or γENaC significantly reduced the inhibitory effect of Cx30 on ENaC. These findings indicate that enhanced clathrinmediated channel retrieval from the cell surface is the major mechanism of ENaC inhibition by Cx30.
Inhibition of the epithelial sodium channel by connexin 30 manner via P2Y 2 -dependent purinergic signaling causing a rise in intracellular Ca 2+ . However, endogenous expression of purinergic receptors, and hence the responsiveness to extracellular ATP, is negligible in X. laevis oocytes (76,77). Moreover, it has been reported that application of ATP does not inhibit ENaC-mediated currents in the oocyte expression system (78). This argues against a role of Cx30-mediated ATP release in mediating the inhibitory effect of Cx30 on ENaC observed in the present study, but does not rule out the possibility that purinergic signaling contributes to the inhibitory effect of Cx30 on ENaC in native renal tubules. Interestingly, Cx30 hemichannels mediate not only the efflux of ATP but also of a variety of other signaling molecules including NAD + , glutathione, glutamate, PGE2, or polyamines (43)(44)(45). Thus, it is conceivable that in addition to ATP, other signaling molecules released by Cx30 may affect ENaC activity. Future studies are needed to explore the potential relevance of such factors.
Expression of Cx30 hemichannels in oocytes resulted in the appearance of a small carbenoxolone-sensitive cation conductance in the presence of divalent cations. This indicates that Cx30 has residual hemichannel activity in the plasma membrane even in the presence of physiological concentrations of Ca 2+ and Mg 2+ in the extracellular solution. Inhibiting Cx30 channel activity by carbenoxolone largely prevented the inhibitory effect of Cx30 on ENaC. Moreover, a point mutation (M34A) that abolishes Cx30 ion channel function, without affecting its intracellular and cell surface expression, prevented its inhibitory effect on ENaC. These findings indicate that channel activity of Cx30 is required for its inhibitory effect on ENaC. Moreover, increasing the Ca 2+ concentration from 1.8 mM to 3.6 mM in the incubation medium significantly reduced the inhibitory effect of Cx30 coexpression on ENaC, whereas decreasing the Ca 2+ concentration from 1.8 mM to 0.9 mM had an opposite effect. This further supports the conclusion that the inhibitory effect of Cx30 depends on its ion channel function, because it is known to be inhibited by extracellular Ca 2+ (61). The reduction of Cx30 hemichannel activity may prevent the passage of regulatory factors (e.g., ions and/or larger signaling molecules) through Cx30, thereby attenuating its inhibitory effect on ENaC. It is well established that ENaC can be inhibited by intracellular Na + due to so-called Na + feedback inhibition (79)(80)(81)(82)(83). In the oocyte system, this mechanism is activated by intracellular Na + concentrations above 30 to 40 mM and probably involves the PPPxY-motifs and Nedd4-2-dependent mechanisms (80,81,83,84). In the present study, oocytes were incubated after cRNA injection in a solution with a relatively low Na + concentration (approximately 14 mM), which we routinely use to minimize ENaC-mediated Na + influx and subsequent Na + feedback inhibition (63,82,85). Moreover, we demonstrated that the inhibitory effect of Cx30 was not affected by coexpressing catalytically inactive Xenopus xNedd4-CS. Thus, under our experimental conditions, it is unlikely that Cx30-mediated Na + influx and Nedd4-2dependent Na + feedback inhibition contribute to the inhibitory effect of Cx30 on ENaC. However, a minor contribution cannot be ruled out and may play a role under conditions with enhanced Cx30-mediated Na + influx.
It has been shown that Cx26, a close homologue of Cx30, is permeable for Ca 2+ (86). Interestingly, in Cx30 expressing oocytes, reapplication of divalent cations after divalent cation removal regularly elicited a rapid and transient inward current peak (Figs. 1A and 3A), which preceded the inhibitory effect of divalent cations on Cx30-mediated inward currents. Such a peak response is a well-known phenomenon in oocytes and most likely caused by a transient stimulation of Ca 2+ -activated chloride channels (53,64,87). This suggests that Cx30 may not be fully blocked in the presence of physiological concentrations of divalent cations on the extracellular side, but may be slightly permeable for Ca 2+ and therefore stimulate Ca 2+ signaling pathways. Previous studies provide evidence that an increased intracellular Ca 2+ concentration downregulates ENaC by complex mechanisms (79,(88)(89)(90)(91)(92)(93). However, as mentioned above, there was an inverse relationship between the extracellular Ca 2+ concentration and the inhibitory effect of Cx30 on ENaC. Moreover, replacing extracellular Ca 2+ with equimolar Ba 2+ did not prevent the inhibitory effect of Cx30 on ENaC, but rather enhanced it. Thus, the inhibitory effect of Cx30 on ENaC does not require the presence of Ca 2+ in the extracellular solution. Therefore, it is unlikely that Cx30mediated Ca 2+ influx is essential for causing ENaC inhibition. Nevertheless, we cannot exclude the possibility that Ba 2+ may pass through Cx30 and stimulate intracellular signaling pathways, which are normally activated by Ca 2+ . Examples can be found in the literature that Ba 2+ can substitute for Ca 2+ in mediating biological responses (94)(95)(96).
Inhibition of the epithelial sodium channel by connexin 30 dependent mechanisms are not critically involved in mediating ENaC inhibition by Cx30. In contrast, we demonstrate that pharmacological inhibition of clathrin-mediated endocytosis by Pitstop-2 largely reduced ENaC inhibition by Cx30. Importantly, truncating the YxxΦ-motif in γ-ENaC or replacing leucine residues by negatively charged residues in the YxxΦ-motifs of γ-ENaC and/or β-ENaC not only increased baseline ENaC currents but also significantly reduced the inhibitory effect of Cx30 on ENaC. Furthermore, Cx30mediated ENaC inhibition was slightly but significantly reduced by mutating the critical tyrosine residues, which are part of the PPPxY-motifs but also of the YxxΦ-motifs. Taken together, these findings suggest that coexpression of Cx30 stimulates YxxΦ-motif-dependent clathrin-mediated retrieval of ENaC from the cell surface. YxxΦ-motifs are well-known binding sites for μAP2 and are critically involved in clathrinmediated endocytosis of various cell surface proteins including ionotropic GABA A and P2X4 receptors (26,28,29). Our finding that the inhibitory effect of Cx30 was not affected by catalytically inactive Xenopus xNedd4-CS indicates that the stimulatory effect of Cx30 on clathrin-mediated endocytosis of ENaC is independent of Nedd4-2. This supports the previously proposed concept of two independent pathways of ENaC retrieval, i.e., Nedd4-2-mediated and AP2-mediated (30). Interestingly, in a previous study it has been demonstrated that the disruption of the Nedd4-2-binding motifs dramatically reduced ENaC endocytosis, whereas mutating the YxxΦ-motif had no effect on the rate of ENaC retrieval (65). However, in the latter study, a threonine residue within the YxxΦ-motif of γENaC was mutated to alanine (γT629A). This mutation has been reported to disrupt ENaC precipitation with AP2 and to prevent ENaC stimulation by coexpression of dominant negative dynamin, which blocks endocytosis via clathrincoated pits (30). Interestingly, this threonine residue precedes the critical leucine residue mutated in the current study, and its substitution by alanine may not fully abolish the function of the YxxΦ-motif. Moreover, different expression systems may account for some of the discrepant findings. For example, Nedd4-2 overexpression increased ENaC endocytosis in human embryonic kidney 293 cells but not in Fischer rat thyroid epithelial cells (65). Thus, it is tempting to speculate that an alternative usage of Nedd4-2-mediated or AP2mediated ENaC retrieval may be physiologically important to regulate ENaC expression at the cell surface in a tissue-specific manner. In Liddle patients, C-terminal truncations of βand γENaC not only disrupt the PPPxY-motif of the affected subunit but can also affect the YxxΦ-motifs. Therefore, both pathways involved in ENaC retrieval may contribute to the pathophysiology of Liddle syndrome.
The direct mechanistic link of how Cx30 ion channel function promotes clathrin-mediated ENaC endocytosis remains to be elucidated. In this context, additional studies are needed to identify the relevant ions and/or signaling molecules, which enter the cell and/or are released from the cell through Cx30 hemichannels and are responsible for stimulating the endocytic machinery. It is also still unclear whether Cx30 mediates its inhibitory effect on ENaC mainly in a paracrine/autocrine manner or whether Cx30 has to be coexpressed in the same cell to stimulate clathrin-mediated ENaC retrieval. Moreover, it will be important to determine the specificity of the effect of Cx30 on ENaC, because a stimulation of clathrin-mediated endocytosis by Cx30 may affect the retrieval of a wide range of membrane proteins. For example, we observed in pilot experiments that Cx30 coexpression significantly reduced ROMK-mediated currents in the oocyte expression system (data not shown). ROMK is a secretory K + channel colocalized with ENaC in the apical membrane of CD principal cells and known to be internalized in a clathrin-dependent manner (103,104). However, there is no indication for increased ROMK function in Cx30 −/− mice due to reduced clathrin-mediated ROMK retrieval. In particular, Cx30 −/− mice are not hypokalemic (34), which would be an expected consequence of an increased ROMK activity in the context of an increased ENaC activity enhancing the electrochemical driving force for renal potassium secretion through ROMK. Moreover, pressure-induced natriuresis was significantly blunted in Cx30 −/− mice, but there was no difference in pressure-induced kaliuresis in wild-type and Cx30 −/− mice (34). These findings suggest that Cx30 deficiency enhances ENaC activity without affecting ROMK activity, which indicates some degree of specificity of the functional interaction between Cx30 and ENaC in vivo. The precise localization of Cx30 in the luminal membrane of distal tubular epithelium varies in different species (33). Interestingly, in mice, Cx30 expression was found to be particularly strong in the apical membrane of intercalated cells of CDs (33). Therefore, Cx30 may also regulate ion channels expressed in intercalated cells, which remains to be investigated.
It will also be of interest to explore whether other connexins may affect clathrin-mediated endocytosis. Indeed, in preliminary experiments, we observed that coexpression of Cx37 also significantly reduced ENaC-mediated currents in the oocyte expression system (data not shown). However, only basolateral localization of Cx37 has been described in various nephron segments (105) and Cx37 −/− mice lack an obvious renal phenotype (35,106). Thus, Cx30 is a more likely candidate than Cx37 to modulate ENaC function in vivo. Nevertheless, modulation of ion channel function by connexin hemichannels via stimulation of clathrin-mediated endocytosis may not be limited to ENaC and Cx30 but may represent a more general regulatory mechanism.
In conclusion, the in vitro findings of the present study reveal a novel mechanism of ENaC inhibition by Cx30. Absence of this inhibitory effect may contribute to increased ENaC activity and salt-sensitive hypertension in mice with Cx30 deficiency (33)(34)(35)(36)(37). Moreover, this study highlights the regulatory potential of modifying the rate of clathrin-mediated channel retrieval to adjust ENaC expression at the cell surface.

Isolation of oocytes and two-electrode voltage clamp experiments
Isolation of oocytes and two-electrode voltage clamp experiments were performed essentially as described previously (64,68,69,71,109). Defolliculated stage V to VI oocytes were obtained from ovarian lobes of adult female X. laevis in accordance with the principles of German legislation, with approval by the animal welfare officer for the University of Erlangen-Nürnberg and under the governance of the state veterinary health inspectorate (approval number Az. 55.2-2532-2-527). Animals were anesthetized in 0.2% MS222 (Sigma), and ovarian lobes were obtained by a small abdominal incision. Oocytes were injected with the same amount (0.1 ng unless stated otherwise) of cRNA per ENaC subunit (α, β and γ) per oocyte. Unless stated otherwise, 2 ng of human Cx30 cRNA was injected. The amount of X. laevis xNedd4-CS cRNA was 3 ng per oocyte. To suppress the expression and possible interference of endogenous connexin 38 (Cx38) hemichannels, an established approach (50, 51) was used, i.e., 3 ng of the antisense phosphothioate-oligoDNA against Cx38 (AS Cx38) corresponding to nucleotides −5 to +25 of the coding region of Cx38 (5 0 -GCT TTA GTA ATT CCC ATC CTG CCA TGT TTC-3 0 ) was routinely coinjected into the oocytes together with ENaC and/or Cx30 cRNA. To prevent Na + overloading, which would reduce ENaC cell surface expression, injected oocytes were maintained in a low-sodium ND9 solution (in mM: 9 NaCl, 2 KCl, 87 N-methyl-D-glutamine-Cl, 1.8 CaCl 2 , 1 MgCl 2 , 5 HEPES, pH 7.4 adjusted with Tris) supplemented with 100 units/ml sodium penicillin and 100 μg/ml streptomycin sulphate. Unless stated otherwise, concentration of CaCl 2 in ND9 incubation solution was 1.8 mM. Penicillin and streptomycin were omitted from ND9 when 1.8 mM CaCl 2 was replaced by 1.8 mM BaCl 2 to avoid precipitation of insoluble salts. Oocytes were studied 48 h after cRNA and AS Cx38 injection. Bath solution exchanges with a gravity-fed system were controlled by a magnetic valve system (ALA BPS-8) in combination with a TIB14 interface (HEKA). An individual oocyte was placed in an experimental chamber with a narrow flow channel (length: 45 mm; height: 3 mm; width: 3 mm) with a U-shaped cross section of 8 mm 2 . The oocyte was positioned in the experimental chamber close to the site of solution inflow and was held in place by the impaling microelectrodes. To achieve rapid and reproducible solution exchanges at the oocyte, the perfusion rate was carefully adjusted for each experimental solution to 10 ml/min, resulting in a flow velocity of 20 mm/s. The flow channel drained into a reservoir (2 cm × 1 cm) from which the solution was continuously removed via a suction tube. The suction tube was adjusted to maintain the fluid level in the flow channel at 2 mm. Oocytes were clamped at a holding potential of −60 mV using an OC-725C amplifier (Warner Instruments) connected by a LIH-1600 (HEKA) to a personal computer. Pulse 8.78 software (HEKA) was used for data acquisition. ENaC-mediated whole-cell currents (ΔI Ami ) were determined by washing out amiloride (2 μM) with amiloride-free bath solution and subtracting the whole-cell currents measured in the presence of amiloride from the corresponding whole-cell currents recorded in the absence of amiloride. ND96 solution was used as standard bath solution (in mM: 96  The cation permeability ratio, P NMDG /P Na , was calculated using the following equation (110): where ΔE rev is the shift of the reversal potential ðE rev Þ caused by replacing Na + in the bath solution with NMDG + , F is Faraday's constant, R is the universal gas constant, and T is the absolute temperature. ΔI Retrieval (see Fig. 9) was calculated according to the following equation: ΔI Retrieval ¼ ΔI Amið0 minÞ −ðΔI Amið70 minÞ − ΔI Insertion Þ ΔI Insertion , i.e., the component of ΔI Ami (70 min) mediated by MTSET responsive channels newly inserted into the plasma membrane within 70 min, was calculated from the ΔI Ami increase due to the second application of MTSET (ΔI 2nd MTSET ) Inhibition of the epithelial sodium channel by connexin 30 according to the following equation, which takes into consideration the increase in P o due to MTSET: where P o (initial) was calculated as follows: This calculation is based on the assumption that average P o of ENaC before the first exposure to MTSET is similar to that of newly inserted channels before the second MTSET application. Moreover, it was assumed that when MTSET was applied for a second time, it only increased P o of newly inserted channels to 1 and had no effect on channels covalently modified by the first MTSET application.

Surface expression assay and western blot analysis
Surface ENaC expression was determined using a chemiluminescence assay as described previously (62)(63)(64). For these experiments, cRNAs (0.6 ng per oocyte per subunit) encoding wild-type αand γ-ENaC were coinjected with cRNA encoding β-ENaC with a FLAG tag inserted into its extracellular loop. Mouse monoclonal anti-FLAG M2 antibody (Sigma-Aldrich) as primary antibody and peroxidase-conjugated sheep antimouse IgG (Chemicon) as secondary antibody were used. Individual oocytes were placed in a white U-bottom 96-well plate, and 50 μl of SuperSignal ELISA femto maximum sensitivity substrate (Thermo Scientific) was added to each oocyte. Chemiluminescence was quantified with a Tecan GENios microplate reader (TECAN). Results are given in relative light units (RLU).
To separate cell surface expressed Cx30 from intracellular Cx30, cell surface proteins were labeled with biotin essentially as described previously (64,102). Oocytes (30/group) were incubated for 15 min in a biotinylation buffer (2.5 ml; pH 9.5) containing 10 mM triethanolamine, 150 mM NaCl, 2 mM CaCl 2 , and 1 mg ml −1 EZ-linked sulfo-NHS-SS-Biotin (Pierce). After biotinylation was completed, oocytes were lysed in homogenization buffer (1 ml; pH 7.9) containing 83 mM NaCl, 10 mM HEPES, 1 mM MgCl 2 , supplemented with 0.5% Triton X-100, 0.5% Igepal CA-630 (Sigma-Aldrich) and a protease inhibitor cocktail (complete EDTA-free, Roche Diagnostics). The lysate was centrifuged for 10 min at 1000g to remove cell debris and yolk proteins. The protein concentration of the resulting cell lysate was determined with a BCA-assay (Pierce). By adding 100 μl of immunopure immobilized NeutrAvidin Agarose beads (Pierce) to 850 μl of this lysate, biotinylated proteins were precipitated on a rotating wheel at 4 C overnight. After subsequent centrifugation at 1000g for 3 min, beads (with bound cell surface proteins) and supernatant (intracellular proteins) were boiled for 5 min at 95 C with sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer (Rotiload 1, Roth). Cell surface proteins were separated from beads by centrifugation at 16,000g for 3 min. The resulting preparation of cell surface proteins (30 μl of the total volume of 90 μl) and intracellular proteins (30 μg of total protein) was loaded on the 10% SDS-PAGE and transferred to polyvinylidene fluoride (PVDF) membranes by semidry electroblotting. V5-tagged Cx30 was detected using mouse monoclonal anti-V5 antibody (Invitrogen) at a dilution of 1:1000 and a secondary horseradishperoxidase-labeled goat anti-mouse antibody (Abcam) at a dilution of 1:50,000. To validate separation of cell surface proteins from intracellular proteins by biotinylation, blots were stripped and reprobed using a polyclonal rabbit anti-β-actin antiserum (Sigma-Aldrich) at a dilution of 1:5000.
To estimate total ENaC expression in whole-cell lysates, 25 oocytes per group were homogenized with a 27-gauge needle in 1 ml of homogenization buffer (in mM: 10 HEPES, 83 NaCl, 1 MgCl 2 , pH 7.9) supplemented with protease inhibitor mixture ("Complete EDTA-free" protease inhibitor mixture tablets, Roche Diagnostics). The lysates were centrifuged for 10 min at 1000g and supernatants were incubated with 1% Triton X-100 on ice for 40 min. All samples were boiled for 5 min at 95 C and subjected to 10% SDS-PAGE using 40 μg of total protein per lane. After separation, proteins were transferred to PVDF membranes by semidry electroblotting and probed with a subunit-specific antibodies against human α-(1:5000), β-(1:10,000), or γ-ENaC (1:5000) obtained from Pineda Antibody Service as described previously (64,70,71). Horseradish-peroxidase-labeled secondary goat anti-rabbit antibodies were purchased from Santa Cruz Biotechnology and used at a dilution of 1:50,000. Chemiluminescence signals were detected using Super Signal West Femto (Thermo Scientific). Densitometry was performed using ImageJ (National Institutes of Health, Bethesda, MD). ATX Ponceau S (Fluka) membrane staining was used to control protein loading.

Single-channel recordings in outside-out patches
Single-channel recordings in outside-out membrane patches of αβγENaC expressing oocytes were performed 48 h after cRNA injection essentially as described previously (64,68,109) using conventional patch clamp technique. Patch pipettes were pulled from borosilicate glass capillaries and had a tip diameter of about 1 to 1.5 μm after fire polishing. Pipettes were filled with K-gluconate pipette solution (in mM: 90 K-gluconate, 5 NaCl, 2 Mg-ATP, 2 EGTA, and 10 HEPES, pH 7.2 with Tris). Modified ND96 solution was used as bath solution (in mM: 96 NaCl, 4 KCl, 1 CaCl 2 , 1 MgCl 2 , and 10 HEPES, pH 7.4 adjusted with Tris). Seals were routinely formed in low-sodium NMDG-Cl bath solution in which NaCl (95 mM) was replaced by NMDG-Cl (95 mM). In this bath solution, the pipette resistance averaged about 7 MΩ. After seal formation, the NMDG-Cl solution was switched to a standard NaCl bath solution. For continuous current recordings, the holding potential was set to −70 mV using an EPC9 amplifier (HEKA). Using a 3 M KCl flowing boundary electrode, the liquid junction (LJ) potential occurring at the pipette/NaCl bath junction was measured to be 12 mV (bath positive) (111). Thus, at a holding potential of −70 mV, the effective transpatch potential was −82 mV. This value is close to the calculated equilibrium potential of Cl − (E Cl − =−77.4 mV) and K + (E K + =−79.4 mV) under our experimental conditions. Experiments were performed at room temperature. To change from one bath solution to another, a conventional gravity-fed system controlled by a magnetic valve system (ALA BPS-8) was used in combination with a TIB14 interface (HEKA). Pulse 8.78 software (HEKA) was used for data acquisition. Single-channel current data were initially filtered at 1.25 kHz and sampled at 5 kHz. The current traces were digitally refiltered at 250 Hz to resolve the single-channel current amplitude (i) and channel activity. The latter was derived from binned amplitude histograms as the product NP o (64,109,111,112). The current level at which all channels are closed was determined in the presence of 2 μM amiloride. The apparent number of active channels (apparent N) in a patch was determined by visual inspection of the current traces. Apparent channel open probability P o(app) was calculated by dividing NP o by apparent N for recordings with no more than three active channels in the patch (1 ≤ N ≤ 3). Single-channel data were analyzed using the program Nest-o-Patch written by Dr V. Nesterov (Institut für Zelluläre und Molekulare Physiologie, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany).

Statistical analysis
Data are presented as mean ± SEM. Normal distribution of data was assessed using D'Agostino-Pearson omnibus test. Statistical significance was assessed by an appropriate parametric test: ANOVA (with Bonferroni post hoc test) or Student's t-test or by a nonparametric test: Kruskal-Wallis (with Dunn's post hoc test), Mann-Whitney, or Wilcoxon signed-rank test as indicated. N indicates the number of different batches of oocytes, and n indicates the number of individual oocytes studied per experimental group. Statistical analysis was performed using Graph Pad Prism 5.04.

Data availability
All data are contained within the article.
Acknowledgments-The expert technical assistance of Ralf Rinke is gratefully acknowledged.
Author contributions-A. V. I., A. D., and C. K. designed the study and interpreted the data. A. V. I. and A. D. performed the experiments, analyzed the raw data, prepared the figures, and drafted the article. C. K. coordinated the study and finalized the article. All the authors approved the final version of the article. Conflict of interest-The authors declare that they have no conflicts of interest with the contents of this article.