The Role of Pre-H2 Domains of α- and δ-Epithelial Na+ Channels in Ion Permeation, Conductance, and Amiloride Sensitivity*

Epithelial Na+ channels (ENaC) regulate salt and water re-absorption across the apical membrane of absorptive epithelia such as the kidney, colon, and lung. Structure-function studies have suggested that the second transmembrane domain (M2) and the adjacent pre- and post-M2 regions are involved in channel pore formation, cation selectivity, and amiloride sensitivity. Because Na+ selectivity, unitary Na+ conductance (γNa), and amiloride sensitivity of δ-ENaC are strikingly different from those of α-ENaC, the hypothesis that the pre-H2 domain may contribute to these characterizations has been examined by swapping the pre-H2, H2, and both (pre-H2+H2) domains of δ- and α-ENaCs. Whole-cell and single channel results showed that the permeation ratio of Li+ and Na+ (PLi/PNa) for the swap α chimeras co-expressed with βγ-ENaC in Xenopus oocytes decreased significantly. In contrast, the ratio of PLi/PNa for the swap δ constructs was not significantly altered. Single channel studies confirmed that swapping of the H2 and the pre-H2+H2 domains increased the γNa of α-ENaC but decreased the γNa of δ-ENaC. A significant increment in the apparent inhibitory dissociation constant for amiloride (\batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(K_{i}^{\mathrm{amil}}\) \end{document}) was observed in the α chimeras by swapping the pre-H2, H2, and pre-H2+H2 domains. In contrast, a striking decline of \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(K_{i}^{\mathrm{amil}}\) \end{document} was obtained in the chimeric δ constructs with substitution of the H2 and pre-H2+H2 domains. Our results demonstrate that the pre-H2 domain, combined with the H2 domain, contributes to the PLi/PNa ratio, single channel Na+ conductance, and amiloride sensitivity of α- and δ-ENaCs.

The epithelial Na ϩ channel (ENaC) 1 was the first subgroup of the ENaC/DEG superfamily cloned from mammals. The topology of ENaC/DEG comprises two short N-and C-terminal intracellular tails, two hydrophobic membrane-spanning domains (M1 and M2), and a large, extracellular loop with two (or three) cysteine-rich domains (1,2). There is an overall ϳ37% amino acid identity between ␣-, ␦-, ␤-, and ␥-subunits. Both the ␣and ␦-subunits can form independent conducting channels with similar amiloride sensitivities and ion selectivities to those co-expressed with ␤␥-subunits. The ␤and ␥-subunits are modifying subunits that regulate the trafficking and conductance of ␣and ␦-ENaCs (1,2). The ␣␤␥-ENaC has a higher permeability to Na ϩ and Li ϩ compared with the other alkali metals. For example, ␣␤␥-ENaC has a permeability ratio of P Li /P Na up to 2, but the channel is virtually impermeant to K ϩ ions (1,2). Similar to other ENaC/DEG superfamily members, ENaC is very sensitive to amiloride, displaying an apparent inhibitory dissociation constant (K i amil ) in the nanomolar range (1,2).
Previous publications (1,2) have demonstrated that the pre-M2 region of ␣-ENaC, more precisely, the second hydrophobic domain (H2) preceding the M2 region, serves as the outer mouth of the ENaC pore and is involved in channel gating. An amiloride-binding site has also been identified functionally in the H2 domain in addition to the one more proximal to this domain (3,4). The H2 domain also forms part of the ion selectivity filter, as deduced by mutation analysis of the pre-M2 region (5)(6)(7)(8)(9)(10)(11)(12).
The 20 amino acids comprising the M2 domain mainly function as part of the conduction pathway. Studies on mice, rat, and human ␣-ENaC support this hypothesis (8,9,13,14). We have shown previously (15) that positively charged residues downstream of the M2 region of ␣-hENaC (post-M2 domain) also contributed to ion permeation and gating behavior and that the functional diameter of the channel pore, in constructs in which these post-M2 positively charged amino acids were mutated to negatively charged glutamic acids, was altered.
There is 63% diversity in the predicted amino acid sequences between ␣and ␦-ENaC subunits. ␦and ␣-ENaCs have three essential biophysical and pharmacological differences (16). First, the whole-cell amiloride-sensitive Na ϩ current of ␦␤␥-ENaC is greater than the amiloride-sensitive Li ϩ current, yielding a Li ϩ /Na ϩ permeation ratio (P Li /P Na ) of 0.6 rather than 2.0 (for ␣␤␥-ENaC). There is no difference in P Na /P K . Second, the value of K i amil for ␦␤␥-ENaC (up to 2.6 M) is 30-fold greater than that for ␣␤␥-ENaC (16). Third, the unitary Na ϩ conductance (␥ Na ) measured for ␦␤␥-ENaC in cell-attached patches is 2.4-fold greater than that of wt ␣␤␥-ENaC, but the unitary Li ϩ conductance (␥ Li ) is unchanged (16). Nothing is known about the molecular basis for the biophysical and pharmacological differences between ␣and ␦-ENaC.
The ectodomains of Na/K-ATPase, gastric H/K-ATPase (18), inward-rectified and hERG K ϩ channel (19,20), and C5a receptor (21) have been shown recently to mediate ion selectivity, channel conductance, and agonist/ion affinity. The contribu-tions of the extracellular pre-H2 domains of ENaC to the ion selectivity filter and amiloride inhibition are unknown. The aim of the present study was to test the hypothesis that the pre-H2 domains of ␣and ␦-ENaC subunits may be integrally associated with amiloride affinity, ion selectivity, and conductance, therefore contributing to the different pharmacological and biophysical properties between ␣and ␦-ENaC. Our results show that swapping the pre-H2 and/or H2 regions of ␣and ␦-ENaC carried over the features associated with their parental wild type constructs partly in cation selectivity, single channel conductance, and almost entirely amiloride sensitivity to the chimeric mutations.
Dual Electrode Voltage Clamp-Whole-cell cation currents were measured in oocytes expressing ENaC using dual electrode voltage clamp 48 or 72 h post-injection (23). Oocytes were impaled with two 3 M KCl-filled electrodes, having resistances of 0.5-2 megohms. A Dagan TEV-200 voltage clamp amplifier was used to clamp oocytes with concomitant recording of currents. Two reference electrodes were connected to the bath by 3 M KCl, 3% agar bridges. The continuously perfused bathing solution was ND96 (in mM): 96 NaCl, 1 MgCl 2 , 1.8 CaCl 2 , 2 KCl, and 5 HEPES (pH 7.4). Equimolar concentrations of LiCl were used to replace NaCl to examine ion selectivity. The chamber volume was less than 100 nl, and the switching of solutions was controlled by SF-77B Perfusion Fast-Step system (Warner, Hamden, CT). Sampling protocols were given by pCLAMP 8.0 software (Axon Instruments, Union City, CA), and currents at Ϫ100 and ϩ60 mV were continuously monitored at an interval of 2 s using a strip chart recorder. Oocytes were clamped at a holding potential of 0 mV. The current-voltage (I-V) relationships were acquired by stepping the holding potential in 20-mV increments from Ϫ120 to ϩ 80 mV. I-V data were recorded after the monitoring currents were stable before and after the application of 10 M amiloride to the bath except where stated. Data were sampled at the rate of 1 kHz and filtered at 500 kHz.
Single Channel Patch Clamp-Oocytes were shrunk in a hyperosmotic medium, and the vitelline membrane was removed before patch clamping (15). The on-cell and inside-out configurations were used to record single channel currents, using an Axopatch 1B amplifier (Axon Instruments, Union City, CA) (15). The patch pipettes were pulled from fire-polished borosilicate glass (WPI Instruments) by using a multistepped micropipette puller (model M97, Flaming/Brown). The electrode tips were fire-polished. The resistance of the electrode was 5-10 megohms when filled with pipette solution (NaCl 100 mM, HEPES-Na 10 mM (pH 7.5)). Currents were collected using the software CLAMPEX 7.0 at a sampling interval of 500 s. The current traces were filtered with the 0.1 kHz built-in low pass filter of CLAMPEX 7.0 and digitized by DigiData 1200 (Axon Instrument, Union City, CA). After collecting data for the on-cell configuration, inside-out patches were excised in bath solution (100 mM LiCl, 10 mM HEPES free acid (pH 7.5), LiOH), simply by pulling the electrode back. Depolarizing potentials (0 to ϩ120 mV) were applied to the on-cell patch, and both depolarizing and hyperpolarizing potentials (Ϫ100 mV to ϩ100 mV) were applied to the inside-out patches.
Data Analysis-All macroscopic currents presented in this paper were amiloride-sensitive currents (except the traces in Fig. 6), which were derived by subtracting, by CLAMPFIT, the amiloride-resistant current measured in the presence of 10 M amiloride from the total current amplitude measured in the absence of the drug. Amiloride-sensitive currents, measured between 200 and 800 ms after application of the test potentials, were averaged.
Analysis of single channel data was performed using the FETCHAN and pSTAT programs of pCLAMP version 8.0 or Clampfit 9.0 (Axon Instruments, Union City, CA) as described previously (15). The unitary current level was computed by making all-point histograms or measuring the amplitude of the current transient. Single channel conductance was calculated by linear regression of I-V curves. Values of ␥ Na and ␥ Li under bi-ionic conditions were estimated by fitting the inward and outward current sections of I-V curves, respectively.
The absolute permeabilities for Na ϩ (P Na ) and Li ϩ (P Li ) and intraoocyte ion concentrations were retrieved from the fitting of the macroscopic current-voltage curves (I-V) with the Goldman-Hodgkin-Katz Equation 1 (15), where R, T, and F have their usual meanings; z is the valence of the cation; I amil x represents the amiloride-sensitive current carried by the cation X ϩ ; P x is the absolute permeability value for the cation X ϩ , and [X ϩ ] out and [X ϩ ] in indicate extraoocyte and intraoocyte concentrations, respectively. A cell is the calculated surface of an oocyte, assuming that the oocyte is a spheroid with a diameter of 1.0 mm.
The absolute cation permeability coefficients from single channel recordings were computed by fitting the bi-ionic I-V curves with the modified Goldman-Hodgkin-Katz current Equation 2, where i Li and i Na indicate the unitary current carried by Na ϩ in the pipette solution ([Na ϩ ] out ) and Li ϩ in the bath solution ([Li ϩ ] in ), respectively. The corresponding permeabilities were expressed as P Na and P Li , respectively.
To analyze the affinity of amiloride to the channel, perfusates containing different concentrations of amiloride (ranging from 0 to 1 mM) were switched into the chamber. Both the K i amil and Hill coefficient were retrieved by fitting the dose-response curves of amiloride with the Hill Equation 3, where K i amil is the concentration required for inhibiting half of the maximal current; I is the measured amiloride-sensitive current at the concentration of Amil; and n represents the Hill coefficient.
The voltage dependence of amiloride inhibition was also characterized by fitting the plot of K i amil as function of membrane voltage with the Woodhull Equation 4 (24), where K i amil (E test ) is the equilibrium inhibitory dissociation constant at a test potential (E test ); K i amil (0) is the K i amil at 0 mV; ␦ is the fractional distance across the electric field that can be sensed by amiloride; z is ϩ1 for amiloride; and R, T, and F have their usual meanings.
Both whole-cell and single channel current traces carried by cations moving from the extra-oocyte to the intra-oocyte side were depicted as inward (negative) currents and vice versa. Average data were reported as mean Ϯ S.E. if not stated otherwise. Statistical analysis was performed using one-way analysis of variance combined with the Bonferroni test for variance and mean of unpaired data. A p value of Ͻ0.05 between experimental groups was considered statistically significant.

RESULTS
The purpose of the study was to test the role of the pre-H2 regions of ␣and ␦-ENaC subunits in ion permeation, conductance, and amiloride inhibition. Because these properties of ␣-rENaC and ␦-hENaC are significantly different from each other at the macroscopic and microscopic current levels, these particular subunits were chosen for analysis. The chimeras were engineered by swapping the pre-H2 region as well as the H2 domain. Fig. 1 shows that there is 45.8 and 52.6% identity (37% identity for full-length␣-rENaC and ␦-hENaC) between the pre-H2 and H2 regions, respectively.

Ion Selectivity, Whole-cell and Single Channel Studies
Whole-cell Studies-To test the role of the pre-H2 region in ion selectivity, whole-cell amiloride-sensitive currents carried by extracellular Na ϩ or Li ϩ (101 mM) were recorded in oocytes with the conventional two-electrode voltage clamp technique. Wild type (wt) rat ␤␥-ENaC subunits were co-expressed with wt ␣-rENaC and its swap chimeras, whereas ␦-hENaC and the swapped constructs were expressed with wt human ␤␥-ENaC subunits. As shown in Fig. 2, the activation of the amiloridesensitive currents evoked at membrane potentials was not time-dependent between Ϫ120 mV to ϩ80 mV. Oocytes were held at 0 mV, which was close to the resting membrane potential as ENaC expression caused an overloading of the cytosolic Na ϩ and Li ϩ concentration, leading to a depolarization of the resting membrane potential. Both wt and chimeric constructs had a slightly inward rectified I-V curve with a reversal potential (E rev ) of more than ϩ10 mV (Fig. 3). The macroscopic amiloride-sensitive Na ϩ and Li ϩ currents were Ϫ1520.37 Ϯ 361 and Ϫ3524.33 Ϯ 606 nA, respectively, at a membrane potential of Ϫ120 mV for wt ␣␤␥-rENaC (n ϭ 33). Current carried by Li ϩ was markedly greater than that of Na ϩ (p Ͻ 0.001). On the other hand, the whole-cell amiloride-sensitive Na ϩ and Li ϩ currents of ␣ pH2 ␤␥-rENaC were Ϫ2970.99 Ϯ 832 (p Ͻ 0.01 compared with that of wt ␣␤␥-ENaC) and Ϫ4011.51 Ϯ 1178 nA, respectively (n ϭ 25). Conversely, the amiloridesensitive Na ϩ current in oocytes expressing ␦␤␥-hENaC was greater than the Li ϩ current (Ϫ2602.61 Ϯ 530 versus Ϫ1660.94 Ϯ 386 nA for Na ϩ and Li ϩ , respectively, n ϭ 34). The current amplitudes of both amiloride-sensitive Na ϩ and Li ϩ currents associated with ␦ pH2 ␤␥-hENaC were greater than wt construct (Ϫ3302.67 Ϯ 670 nA for Na ϩ and Ϫ2490.11 Ϯ 528 nA for Li ϩ , respectively, n ϭ 26, p Ͻ 0.05). These results for wt ␣␤␥-rENaC and ␦␤␥-hENaC were consistent with those reported previously (16,25,26).
The whole-cell Na ϩ and Li ϩ current levels associated with the pre-H2 swapped chimeras (Figs. 2 and 3) were not identical to those of wt constructs, indicating that their perme-abilities to Na ϩ (P Na ) and Li ϩ (P Li ) have been altered. To address whether the pre-H2 domain regulates Na ϩ or Li ϩ permeation, we calculated the absolute permeation constant.
The retrieved values were listed in Table I. Only ␣ pH2 and ␣ pH2H2 chimeras shifted the E rev significantly to the hyperpolarizing direction. The value of P Na for wt rat ENaC was about half of P Li , yielding a P Li /P Na ratio of 1.8. However, the P Na of wt human ENaC was greater and the P Li /P Na ratio was less compared with its rENaC counterpart. In contrast, the P Li for wt ␦-hENaC was smaller (approximately up to 50% that of rENaC), whereas P Na was greater than these of wt rENaC, yielding a P Li /P Na ratio of 0.62. The selectivity ratios of wt rat and human ENaC constructs calculated from the permeation constant were completely consistent with previous studies (16,25).
An essential G(S)XS tract has been identified in the H2 region of ENaC, which controls Na ϩ permeability (10,12). As shown in Fig. 1, there is a serine-rich domain (564 -568, SESPS) within the pre-H2 region of ␣-ENaC, which is not present in the equivalent stretch of ␦-hENaC subunit. To determine the role of this SXS tract in ion selectivity, Na ϩ and Li ϩ permeabilities of triple mutants ␣564E,S566A,S568V␤␥-rENaC (␣ trip ) and ␦E514S,A516S,V518S␤␥-hENaC (␦ trip ) were examined. For comparison, chimeras swapping the H2 and pre-H2ϩH2 domains were constructed. A significant increment in P Na was observed for the ␣ trip , ␣ pH2 , and ␣ pH2H2 chimeras, but there was a significant decrease in P Li for the ␣ H2 chimera ( Table I). The P Li /P Na ratio reduced markedly for all four chimeric ␣-rENaC constructs. In contrast, only the ␦ trip mutation exhibited an increment in P Li , whereas the P Li /P Na ratio was not statistically different from that of wt ␦-ENaC.
Single Channel Studies-In order to re-confirm the macroscopic results under strictly controlled ionic conditions, we performed single channel studies under bi-ionic conditions. The on-cell and inside-out patches were obtained from injected oocytes, and single channel traces were digitized (Fig. 4). The average unitary currents were plotted as I-V curves shown in Figs. 5 and 6. Wild type ␣␤␥-ENaC and ␣ pH2 ␤␥ chimera function as slight outward rectifiers compared with the macroscopic I-V relationships (Figs. 3 and 5A). Replacement of the pre-H2 region with the corresponding regions of ␦-ENaC shifted the E rev to Ϫ5 from Ϫ15 mV. By comparison, the I-V relationship of ␦␤␥-hENaC revealed an inwardly rectified curve with an E rev of 15 mV (Fig. 5B).
We also applied the Goldman-Hodgkin-Katz equation to the I-V relationships generated under bi-ionic conditions to determine the permeability constant and ratio. The permeability ratio was also recalculated using the E rev (Table II). Consistent with the whole-cell observations, the P Li was greater than P Na for wt ␣␤␥-rENaC, yielding a P Li /P Na ratio of 1.82 (Table II). Substitution of the pre-H2 and pre-H2ϩH2 domains of ␣-ENaC with the corresponding regions of ␦-ENaC resulted in an increment in P Na , whereas swapping of the H2 domain with that of ␦-ENaC led to a decrease in P Li . Both the pre-H2 and/or H2 domain substitution of ␣-ENaC resulted in reduced discrimination between Na ϩ and Li ϩ . With respect to wt ␦-hENaC, the P Li was less than its P Na with a calculated P Li /P Na ratio of 0.55. The chimeric ␦-ENaC constructs with replacement of the pre-H2 and/or H2 domains of ␣-ENaC showed no marked change in Na ϩ and Li ϩ permeabilities.

Single Channel Conductance
The results of on-cell single channel experiments further verified the whole-cell observations (Fig. 4A). The unitary current level recorded at Ϫ100 mV in on-cell patches displayed differences when the pipette was filled with Na ϩ or Li ϩ (100 mM). For the wt ␣␤␥-rENaC and swap chimeric constructs, the Na ϩ current was much smaller than Li ϩ current (Fig. 4A, top panel and Table III). In contrast, the single channel Na ϩ current of wt ␦␤␥-hENaC was much greater (12 versus 4 pS for wt ␣␤␥-ENaCs), whereas the Li ϩ current associated with ␦␤␥-hENaC was almost identical to that of ␣␤␥-ENaCs (7 pS). Swapping the H2 and pre-H2ϩH2 domains increased ␥ Na of ␣ chimeras, but the ␥ Na was significantly reduced for the swap ␦-ENaC constructs in on-cell patches.
For wt ␣␤␥-rENaC, there was no significant difference in ␥ Na and ␥ Li between the on-cell and inside-out configurations. In contrast, both the ␥ Na and ␥ Li of ␦-ENaC in the inside-out patches were strikingly smaller than those from the cell-attached patches (57 and 66% of the on-cell conductance, respectively, see Table III). With regard to wt ␣␤␥-hENaC, the ␥ Li was reduced by 44%. Consistent with the results in the on-cell patches, an increment for the ␣ H2 and a decrease for the ␦ H2 and ␦ pH2H2 constructs in ␥ Na were observed in the inside-out patches (Table III). Inconsistent with data obtained from the on-cell configuration, both the ␥ Na and ␥ Li for the other constructs declined significantly in the inside-out patches (Table III).

Amiloride Sensitivity
To test the hypothesis that the pre-H2 region preceding the amiloride-binding site located in the H2 domain modified the kinetics of amiloride inhibition of the ENaC channel, we used the whole-cell mode to record amiloride-sensitive Na ϩ currents in the presence of various amiloride concentrations (range, 0 -1 mM). Fig. 7A shows representative traces of the time course and concentration dependence of the amiloride inhibitory effect for wt and pre-H2 swap constructs. Generally speaking, stable currents could be achieved within a few seconds following the application of amiloride. Fig. 7, B and C shows the dose-response curves at Ϫ100 mV fitted with the Hill equation. The value of K i amil for wt ␣␤␥ rENaC was 54.7 Ϯ 19 nM (n ϭ 7). For the ␣ pH2 ␤␥ chimeric mutation, the K i amil increased approximately by 3-fold to 150.0 Ϯ 16.5 nM (n ϭ 5, p Ͻ 0.01). Consistent with a previous report (16), the K i amil for ␦␤␥-hENaC was 52-fold greater than that of ␣␤␥-rENaC (2860 Ϯ 501 nM, n ϭ 12). Exchanging the pre-H2 region of ␦-hENaC with that of ␣-rENaC significantly decreased the K i amil to 745 Ϯ 247 nM for the ␦ pH2 ␤␥-hENaC (n ϭ 6).
We constructed swap chimeras by switching the H2 domain and the pre-H2ϩH2 domains, in which an amiloride-binding site is located as positive controls (1, 2). Fig. 8 presents the computed K i amil and Hill coefficient as a function of membrane potential. The voltage dependence of K i amil was fitted with the Woodhull equation (24). The retrieved parameters including the fractional distance across the electrical field sensed by amiloride (␦), the experimental K i amil at 0 mV, and the esti-   Table IV.
The experimental K i amil at 0 mV for the ␦ pH2 construct decreased significantly as compared with that for wt ␦-ENaC, whereas an increment was observed for the ␣ pH2 chimera (Table IV). Swapping of the H2 domains between ␣and ␦-ENaCs almost completely exchanged the values of K i amil for each other (Fig. 8, A and B). However, the pre-H2ϩH2 domain swap channels did not show an additional shift in amiloride affinity to the others, instead the change in K i amil was smaller compared with those of the H2 swap chimeras. These results suggest that there are intramolecular domain-domain interactions between the pre-H2 and H2 domains and that their interactions with amiloride are not synergistic. In comparison to hyperpolarizing membrane potentials, the K i amil was more voltage-dependent at depolarizing membrane potentials from Ϫ20 to ϩ80 mV (Fig. 8, A and B). All estimated values of the K i amil at 0 mV (K i amil (0)) were pretty close to those computed from the experimental data. For example, the calculated K i amil (0) for wt ␣␤␥-rENaC was 354 and 330 nM for the experimental data. The value of ␦ was 0.35 for wt rENaC, close to those yielded by noise analysis of the macroscopic current  (28,33). In contrast, the value of ␦ for ␦␤␥-hENaC was greater than that for ␣␤␥-rENaC (Table IV).
As shown in the right panel in Fig. 8A, all swap chimeric ␣-ENaC constructs had an increased Hill coefficient. On the other hand, the Hill coefficient reduced markedly for all chimeric ␦ constructs at depolarizing membrane potentials (Fig.  8B, right panel). DISCUSSION The goal of the present study was to examine whether the pre-H2 domain regulates ion selectivity, unitary conductance, and amiloride sensitivity of ENaC. Our experimental strategy was to swap 24 amino acid residues preceding the H2 regions of ␣and ␦-ENaCs. The main findings of our studies are as follows. 1) The pre-H2 and H2 domains and an SXSXS motif of ␣and ␦-ENaCs are involved in Na ϩ and/or Li ϩ permeation. 2) Single channel studies confirmed that the ␥ Na of ␦␤␥-hENaC increased for the ␣ H2 and ␣ pH2H2 chimeras, supportive of the whole-cell results; the ␥ Na of chimeric ␦-ENaC constructs was reduced in on-cell patches.
3) The ␥ Li of all chimeric ␣ and ␦ constructs decreased in the inside-out mode with the exception of ␥ Na for the ␣ H2 and ␦ pH2 chimeras. 4) The pre-H2 clusters combined with the H2 domains of ␣and ␦-subunits contribute to amiloride inhibition by modifying the K i amil , the Hill coefficient, and voltage dependence.
Ion Permeation-Our results suggest that the swap chimeric ␣ pH2 -, ␣ H2 -, and ␣ pH2H2 -ENaC channels discriminate less between Na ϩ and Li ϩ due to the elevated Na ϩ permeation, which was inherited from the wild type ␦␤␥-hENaC (Tables I and II). These results for wild type ␦␤␥-hENaC are consistent with the observations Waldmann et al. (17). Replacement of hydrophobic region 2 (H2) of ␣-hENaC with the corresponding sequence of MEC-4 significantly increased ␥ Na to 14 pS and the ␥ Na /␥ Li ratio increased to 1.6 (17). Serines 589 and 593 located in the pore region of ␣-ENaC were further identified as modifying ␥ Na to the same extent (17). Additionally, mutations of the pre-M2 region revealed that the H2 domain regulated ion permeation; in particular, ␥ Na was altered (4,8,9,11,14). However, these two key residues are highly conserved in both ␣and ␦-ENaC isoforms. The present studies with an ␣-triple mutant suggested that an SXSXS motif partially contributes to the increased Na ϩ permeability of the ␣ pH2 -ENaC chimera. In comparison to the ␣ pH2 -chimera, the swap ␦ constructs did not significantly affect Na ϩ permeability.
Our results suggest that multiple domains integrally control Li ϩ and Na ϩ permeation. They support the idea that the selectivity filter consists of a chain of amino acid residues (i.e.

FIG. 4. Single channel current traces recorded from on-cell and inside-out patches.
A, the abscissa is 2 s and the ordinate is 1 pA. These traces were digitized at membrane potential of Ϫ100 mV. The upper trace was recorded with 100 mM Na ϩ , and the bottom trace was made with 100 mM Li ϩ . The icon in the middle represents the patch clamping configuration: the on-cell mode. B, single channel traces obtained from the inside-out patches under bi-ionic conditions. On the right side are the patch potentials (V m ). Left panel is from the wt constructs, and the right panel is from the swap chimeras. Insideout patches were clamped at command potentials (V p ) from Ϫ100 mV to ϩ 100 mV. The scale bar for the x axis is 10 s and for the y axis is 1 pA.

TABLE I Ion permeation ratio estimated by fitted the macroscopic currents with the Goldman-Hodgkin-Katz Equation 1
The pre-H2 (␣ pH2 and ␦ pH2 ), H2 (␣ H2 and ␦ H2 ), and pre-H2H2 (␣ pH2H2 and ␦ pH2H2 ) swap chimeras as well as triple mutants (␣ trip and ␦ trip ) are compared with wild type constructs. E rev , reversal potential of Na ϩ current; P Na and P Li , absolute permeation constants for Na ϩ and Li ϩ ; and P Li /P Na , permeation ratio of Na ϩ and Li ϩ . Numbers in parentheses in the 1st column are oocytes tested.

FIG. 5. Unitary current-voltage (I-V) relationships for wild type and swap ENaC constructs from on-cell and inside-out patches.
A, I-V curves of wt and the swap rENaC from the on-cell patches at pipette voltage (V p ) from Ϫ120 to Ϫ20 mV. The dotted line was drawn by fitting the data with a linear fitter to calculate single channel conductance of Na ϩ (square) and Li ϩ (circle). Bottom panel shows I-V curves obtained from the inside-out patches for wt (n ϭ 11) and the swap rENaC (n ϭ 10). Reversal potential is labeled by arrow and number. The dotted lines were created by fitting the current data with the Goldman-Hodgkin-Katz equation strictly for bi-ionic conditions (Equation 2). The unitary Na ϩ (␥ Na ) and Li ϩ (␥ Li ) conductances under bi-ionic conditions were computed by linear regression of inward and outward currents, respectively. B, I-V curves for wt and the swap ␦-hENaC constructs from the on-cell and inside-out patches.
the pre-H2, H2, M2, and post-M2 domains) and that the selectivity to Li ϩ , Na ϩ , and K ϩ is determined by different residues. Moreover, more than one residue can influence the permeation for each individual alkali cation species (8 -10, 12). This idea explains why the Na ϩ permeation of the swap chimeric ␦ pH2H2 -ENaC channels was not changed and why the P Na of the ␣ pH2H2 is not identical to that of wt ␦␤␥-hENaC. Our observa-tions suggested that the pre-H2 and H2 domains of ␦-ENaC are not the only key regions for its greater Na ϩ permeability over ␣-ENaC.
Why are the ␥ Na and ␥ Li of wt and chimeric constructs in the inside-out patches lower than these of the on-cell patches (Table III)? Our previous studies have demonstrated that cytoskeletal elements physically interact with and functionally regu- late ␣␤␥-ENaC activity (34,35). In the excised patches, the precise architectural arrangement of the cytoskeleton is undoubtedly disrupted, and the channel conformation and/or interactions with these cytoskeletal proteins may be changed. Also, tension on ␦-ENaC during the formation of the inside-out patches may contribute to an alteration in cation conductance. We cannot rule out the involvement of soluble intracellular elements and salt components. In the on-cell patches, swapping the pre-H2 domain with the corresponding sequence of ␣-ENaC resulted in a decrease in ␥ Na of ␦-ENaC, suggesting that the pre-H2 region of ␦-ENaC may be an interactive motif, which directly cross-talks with the extracellular loop of other membrane proteins or serves as a regulator of other protein-protein interactive motifs. Finally, the cytosolic Li ϩ concentration (100 mM) used for the inside-out configuration is much greater than the concentrations for the cell-attached mode. As predicted by the reversal potential of the macroscopic I-V curves, intraoocyte Na ϩ content is approximately 60 mM for ␣-ENaC and less for ␦-ENaC. The decreased ␥ Na and ␥ Li in the inside-out patches excised from ␦-ENaC-expressing cells may thus be due to the feedback regulation and self-regulation, respectively (2). If so, wt ␦␤␥-hENaC is more sensitive to the inhibitory regulation than wt ␣␤␥-rENaC.
As indicated by the Goldman-Hodgkin-Katz current equation (Equation 1), the permeability to cations positively correlates to whole-cell current. Whole-cell current is determined by the number of electrically detected active channels (N), channel open probability (P o ), as well as single channel conductance. Our results, namely that the P Na of the ␣ pH2 ␤␥-rENaC is doubled whereas the single channel conductance is not changed, imply that the ␣ pH2 ␤␥-rENaC may increase either or both active channel number or channel open probability.
As shown in Figs. 1F and 6E, as well as Table III, wt ␣␤␥-hENaC differs from the rat counterpart in their biophysical features including the P Na /P Li ratio, E rev of the macroscopic Na ϩ current, and ␥ Li in the inside-out patches. If these differences are mainly determined by wt ␤and ␥-hENaC subunits, then the properties of the swap ␦-ENaC constructs may be affected by ␤␥-hENaC subunits.
Amiloride Inhibition-Multiple amiloride-binding sites at the extracellular loops of ␣␤␥ subunits have been biochemically and functionally identified (3,4). However, no amiloride binding or regulatory sites have yet been localized to the pre-H2 region. Our observations on amiloride inhibition of the wt and swap constructs suggested that the pre-H2 domain is also important for amiloride interaction with the channel, particularly for ␦-ENaC. The values of K i amil for the chimeras that swap the H2 and pre-H2ϩH2 regions are not identical to that of the other parental constructs, suggesting that domains beside these two regions influence amiloride affinity. Because ␤and ␥-subunits also contribute to amiloride affinity, co-expression of wt ␤␥-ENaC subunits may compensate for the effects of swapping pre-H2 and H2 regions on amiloride inhibition.
Our observations also confirm previous studies (16) that the K i amil of ␦␤␥-ENaC is in the micromolar rather than the nanomolar range. Amiloride inhibition of both ␣and ␦-ENaCs is voltage-dependent, especially at depolarizing membrane potentials above 0 mV (Fig. 7). Amiloride is a positively charged molecule when protonated and is thus sensitive to membrane potential (36). Cations compete with amiloride for at least one of these two binding sites. Although amiloride affinity is regulated by extracellular cation concentration, external protons, membrane potential, and intracel-

TABLE III Single channel conductance of the wt and swap ENaC constructs, on-cell and inside-out patches
The unitary Na ϩ (␥ Na ) and Li ϩ (␥ Li ) conductance was estimated by linear fitting of unitary I-V curves. Numbers in parentheses indicate the number of patches measured. lular amiloride accumulation, the fact that the identical experimental conditions were used regardless of the ENaC constructs under investigation excluded contributions from these other factors (36).
Comparing data from ␣␤␥-rENaC and ␣␤-rENaC revealed that ␣␤-ENaC channel was more permeable to Na ϩ than Li ϩ (P Na Ͼ P Li ), had an identical fractional distance across the electrical field sensed by amiloride, and an increased K i amil (4 M, Refs. 6 and 28). These results raised the question that the chimeric ␣ pH2 -ENaC may assemble a channel with only a ␤-subunit (and not a ␥-subunit). However, the ␣ pH2 ␤␥-ENaC has a greater permeability to Li (␥ Na 4.2 versus ␥ Li 6.9 pS) and a smaller K i amil (0.13 M), which were similar to those of wt ␣␤␥-ENaC. Our studies also showed that the whole-cell Na ϩ current for the ␣ pH2 ␤-ENaC at a holding potential of Ϫ120 mV was less than 200 nA (not shown), ϳ10% of the amplitude of ␣ pH2 ␤␥-ENaC (Fig. 2). Thus, the possibility that chimeric ␣-construct may assemble with only a ␤-subunit is unlikely.
The voltage dependence of amiloride inhibition in toad urinary bladder (31, 32) and cloned ENaC (27, 33) has been confirmed. Furthermore, depolarization led to an increment in amiloride k off but a decrease in k on for the native epithelial Na ϩ conductance (31,32), whereas only a weak voltage dependence of k on and/or k off was observed for ␣␤␥and ␣␤-rENaC (28,33). A simple plug-type blocking model was proposed whereby the positively charged protonated amiloride is attracted by negatively charged amino acids located in the vestibule of the channel (31). Because the negative charges in the M2 domain did not contribute to amiloride affinity as evidenced by pore-region truncation mutants (37), and the two amiloride-binding sites identified by Ismailov et al. (3) and Snyder et al. (12) are not negatively charged, it was speculated that the negative charges of the pre-H2 region preceding the H2 cluster contributed to amiloride affinity. However, because the pre-H2 domain of ␦-hENaC contains more negative charges than the corresponding region within ␣-rENaC and yet has a lower K i amil , it is unlikely that these negative charges are important in amiloride-ENaC interactions.
We also found that the fractional distance across the electrical field sensed by amiloride for ␦-ENaC is greater (0.48) than that of ␣-ENaC (0.35). As summarized in Table IV

TABLE IV
The parameters retrieved by fitting of the voltage-dependent plots of K i amil with the Woodhull equation ␦ is the fractional distance across the electrical field sensed by amiloride. K i unchanged electrical distance (Table IV), do not support either of these two interpretations. Swapping of the H2 domains with or without combination of the pre-H2 segment suggested that some key residues within the H2 domain contribute to 90% of the difference in amiloride affinity, because the identified amiloride-binding site is identical between ␣and ␦-ENaCs. Changes in the fractional distance sensed by amiloride of the H2 swap chimeras also suggested that the H2 domain determines the diversity in amiloride sensitivity between ␣and ␦-ENaCs. The pre-H2 domain secondarily mediates amiloride inhibition.
It is well known that ENaC has a different closing conformation from DEG members (1,2). Swapping of the M2 domain of ␣-ENaC with that of MEC-4 decreased amiloride affinity accompanied by an alteration in closing conformation (17). Regulation of channel conformation and ligand binding has been found in other channels as well (18 -21). We recently found that the conformational change of ␦-ENaC induced by H ϩ , Zn 2ϩ , and other external cations was not akin to ␣-ENaC. 2 Other indirect effects subsequent to swapping of the pre-H2 and H2 regions, i.e. interactions with associated proteins, channel structure, and external cation affinity, cannot be excluded by our results. Therefore, whether the changes in ion selectivity, single channel conductance, and amiloride affinity associated with chimeric ␣and ␦-ENaC constructs are due to direct effects or conformation alterations is not clear.
Physiological Relevance-The possible physiological relevance for the pre-H2 region involvement of amiloride sensitivity, unitary conductance, and ion selectivity is not clear. The extraordinarily large M1-M2 connecting loop, comprising over 70% of the mass of the entire channel sequence, implies that the extracellular loop may function as a receptor or sensor of different stimuli. The ENaC/DEG channels are regulated by many external ligands, for example, channel-activating protease 1, protons, amiloride, bivalent cations, and mechanical stretch (10,22,36,38). The regulatory sites for most of these external inputs are likely located in the extracellular loop. The ectodomains of ENaC may also play a role in the interactions with other cytoplasmic membrane proteins (i.e. cystic fibrosis transmembrane regulator, the cytoskeleton) (35,39). Our results indicate that the pre-H2 domain of ENaC may mediate the regulation of chemical and physical stimuli on amiloride affinity and cation selectivity.
In conclusion, our studies demonstrate that the pre-H2 regions of both ␣-ENaC and ␦-ENaCs contribute to ion selectivity, single channel conductance, and amiloride inhibition. The pre-H2 domain functionally interacts with the H2 domain as evidenced by our pre-H2ϩH2 domain swap studies.