The Second Hydrophobic Domain Contributes to the Kinetic Properties of Epithelial Sodium Channels*

The epithelial sodium channel (ENaC) is the prototype of a new class of ion channels known as the ENaC/Deg family. The hallmarks of ENaC are a high selectivity for Na+, block by amiloride, small conductance, and slow kinetics that are voltage-independent. We have investigated the contribution of the second hydrophobic domain of each of the homologous subunits α, β, and γ to the kinetic properties of ENaC. Chimeric subunits were constructed between α and β subunits (α-β) and between γ and β subunits (γ-β). Chimeric and wild-type subunits were expressed in various combinations in Xenopus oocytes. Analysis of whole-cell and unitary currents made it possible to correlate functional properties with specific sequences in the subunits. Functional channels were generated without the second transmembrane domain from α subunits, indicating that it is not essential to form functional pores. The open probability and kinetics varied with the different channels and were influenced by the second hydrophobic domains. Amiloride affinity, Li+/Na+selectivity, and single channel conductance were also affected by this segment.

The epithelial sodium channel (ENaC) 1 mediates Na ϩ reabsorption in many epithelial tissues including the distal nephron, colon, lung, and secretory glands. ENaC channels are heteromultimeric proteins formed by the association of homologous subunits: ␣, ␤, and ␥. The second transmembrane domain (M2) from each subunit participates in forming the ion pathway. ␣ subunits can generate functional channels, but ␤ or ␥ cannot; nonetheless, they impart specific properties to the heterooligomeric complex. For instance, ␣ in combination with either ␤ or ␥ forms channels with levels of expression intermediate between ␣ alone and ␣␤␥ (1). The need to express ␣ subunits in order to induce amiloride-sensitive currents has suggested that they are essential to form the channel pore. The subunits of ENaC are 70 -80-kDa glycoproteins. A hydropathy profile of the amino acid sequence, gene fusion experiments, identification of glycosylation sites, and partial proteolysis have confirmed a simple structure for the subunits characterized by two transmembrane segments (M1 and M2), a large extracellular domain with multiple N-glycosylation sites, and the amino and carboxyl termini in the cytoplasmic side (2)(3)(4).
Several lines of evidence indicate that the M2 domains of the subunits determine the properties of the ion pore. This is the most conserved region among all of the channels that form the ENaC/Deg family of ion channels. Members include the mammalian acid-sensitive ion channels (5), the peptide-gated Na ϩ channel from the snail Helix aspersa (6), the degenerins from Caenorhabditis elegans (7), and two channels cloned from Drosophila: pickpocket and ripped pocket (8).
According to the predicted secondary structure, M2 includes a typical ␣-helix, long enough to traverse the plasma membrane, and a less structured segment of hydrophobic residues preceding the ␣-helix. Mutations of residues located in the initial segment of M2 alter the affinity of channels for the blocker amiloride and the ion selectivity (9 -11).
We previously showed that channels formed by ␣ with ␤ (␣␤) and ␣ with ␥ subunits (␣␥) differ in many functional properties such as in affinity for amiloride, ion selectivities, unitary conductances, and single channel kinetics (12,13). Here, we constructed chimeric subunits between ␣ and ␤ (␣-␤) and between ␤ and ␥ subunits (␥-␤) and expressed them in various combinations in Xenopus oocytes in order to examine the properties of whole-cell and unitary currents of the chimeric channels. Taking advantage of the differences in their functional properties, we have correlated specific amino acid sequences in the subunits to functional properties in order to gain insight into the structure-function of these channels.

MATERIALS AND METHODS
Construction of Chimeras and cRNA Synthesis-Two types of chimeras were made: ␥-␤ containing the amino terminus of ␥ and the carboxyl terminus of ␤, designated (␥-␤S481), (␥-␤G529), (␥-␤L535), and (␥-␤E540); and ␣-␤ containing the amino terminus of ␣ and the carboxyl terminus of ␤, designated (␣-␤S508). The letters and numbers indicate the first residue that belongs to the ␤ subunit in each chimera. Chimeras were made using polymerase chain reaction according to the protocol previously described (12) and were subcloned in pSPORT vector (Life Technologies, Inc.). Point mutations in the ␤ subunit were also generated by polymerase chain reaction: ␤G529S and ␤G530C. A stop codon was introduced in the carboxyl terminus of the ␤ subunit at position Arg 564 . This truncation has been shown before to increase the level of expression of channels at the plasma membrane without altering other properties (14). All constructs were verified by DNA sequencing. Capped cRNAs from linearized cDNAs were synthesized in vitro using T7 mMESSAGEmMACHINE (Ambion, Austin, TX) according to the provider's instructions.
Electrophysiology and Data Evaluation-Channel activity was recorded using either two-microelectrode voltage clamp or patch clamp techniques. For two-microelectrode recordings, current and voltage electrodes were pulled from borosilicate glass, had resistances Ͻ1 megaohm, and were filled with 3 M KCl. ENaC activity was calculated from the difference in whole-cell current before and after the addition of 50 M amiloride to the bathing solution. Currents were recorded with an OC-725B oocyte clamp (Warner Instrument Corp., Hamden, CT), digitized at 0.1 kHz (ITC-16; HEKA, Lambrecht, Germany), and stored on a hard disc. Membrane potential was held at Ϫ60 mV. Currentvoltage relations were generated by changing the membrane potential from Ϫ180 to 80 mV in 20-mV incremental steps of 200-ms duration. I/V curves were fitted to the constant field equation. The standard bath solution was composed as follows: 100 mM sodium gluconate, 4 mM KCl, 2 mM CaCl 2 , and 10 mM HEPES, pH adjusted to 7.4. In some experiments, 100 mM Na ϩ was replaced by 100 mM Li ϩ or K ϩ . The K i values of amiloride were calculated by measuring the fractional inhibition of whole-cell currents produced by increasing concentrations of amiloride in the bath solution. The data were fitted to the equation, An Axopatch 200A amplifier and Digidata 1200A (Axon Instruments, Foster City, CA) interfaced to a PC were used to acquire data at 5 kHz. The data were filtered at 200 Hz during acquisition using an eight-pole Bessel filter (Frequency Devices, Inc., Haverford, MA) and stored directly on the hard drive of a PC. Axon's pClamp 6 software was used for data analysis. Data were filtered digitally at 50 Hz for analysis and display. I/V relations were constructed by measuring current passing through channels between 0 and Ϫ80 mV, and the single channel conductance was subsequently estimated by linear regression between Ϫ20 and Ϫ80 mV.
Fetchan was used to generate events list files (EVL files) from the data files. A 50% threshold was set to determine transitions between the open and closed levels from at least 10 min of data recorded at Ϫ40 mV. Channel open probability (P o ) was calculated from the EVL files generated by Fetchan using a specialized NP o program written by Jinliang Sui (Mount Sinai School of Medicine, New York, NY) (15) and available on the World Wide Web. The NP o program reads Fetchan EVL files of any size and expresses these files as the product of N (number of channels in the membrane patch; maximum N ϭ 5) and P o versus time. The resolution of changes in NP o is determined by a user-selectable bin width ranging from 0.1 to 30 s/bin. To express the NP o of each patch, we used the mean of successive 30-s measurements made over the entire EVL file. Patches containing single and multiple channels were used to calculate P o .
Open and closed times were calculated from histograms generated using Fetchan EVL files and the Pstat program within pClamp in logarithmic binning mode (i.e. a plot of number of observations versus log dwell time in ms). The fitting method was simplex-least squares. Results are expressed as mean Ϯ S.E. Differences between groups were assessed using Student's t test. p Ͻ 0.05 was considered to be statistically significant.

Expression of Chimeras in Xenopus
Oocytes-In order to define the domains responsible for conferring the functional differences between ␣␤ and ␣␥, we generated several chimeric subunits that contained the amino terminus of ␥ and the second hydrophobic domain (M2) of the ␤ subunit. We made four ␥-␤ chimeras, (␥-␤S481), (␥-␤G529), (␥-␤L535), and (␥-␤E540), each of which contained progressively fewer sequences from ␤. The letter and number indicate the first residue from the ␤ subunit at the junction of the chimera. An ␣-␤ chimera was also made; it contained most of the ␣ sequences and the M2 from ␤, (␣-␤S508). A schematic representation of the chimeras is shown in Fig. 1.
None of the ␥-␤ chimeras injected alone or coinjected with ␤ or ␥ subunits induced a significant increase in oocyte whole-cell conductance. However, coinjection with wild-type ␣ subunits induced several A of amiloride-sensitive current that were in the range of 2-6 A/oocyte.
Injection of ␣-␤ chimera with wild-type ␥ induced currents in the range of 3-8 A (4.8 Ϯ 1.2 A/oocyte), but no currents were observed when the ␣-␤ chimera was injected alone or when coinjected with wild-type ␤ or ␥ subunits.
Two different chimeras injected together, ␣-␤ with ␥-␤, induced functional channels with cells expressing 2-4 A of whole-cell current. The magnitude of the amiloride-sensitive currents expressed by these channels appears in the right column of Fig. 1.
These experiments showed that sodium channels, sensitive to amiloride block, can be formed by combinations of chimeras and subunits other than ␣, ␤, and ␥. Most significantly, ␥ with ␣-␤ chimera and two chimeras ␣-␤ with ␥-␤ induced sodium currents although these channels do not have the M2 from ␣ subunits.
Open Probability and Kinetics of Single Channels-The level of expression of all of the functional constructs was consistently above 1 A/oocyte, making possible the characterization of the properties of unitary currents of these channels.
We examined the kinetics and P o of all of the functional channels generated by injection of various combinations of subunits and chimeras. For kinetic studies, only patches containing single channels were included in the analysis, while for calculation of P o , patches with single and multiple channels were considered. Fig. 2 shows representative examples of patches containing single channels for the four different ␣(␥-␤) channels. ␣(␥-␤S481) channels exhibited very high P o (0.94 Ϯ 0.03) that was indistinguishable from our previous report of ␣␤ channels (P o ϳ 1). The next three ␥-␤ chimeras, that contain progressively less sequence from the ␤ subunit, had significantly lower P o values that were 0.62 Ϯ 0.04 for ␣(␥-␤G529), 0.21 Ϯ 0.02 for ␣(␥-␤L537), and 0.12 Ϯ 0.04 for ␣(␥-␤E540) ( Table I).
The four types of channels containing ␥-␤ chimeras had kinetics described by only one open and one closed state. Fig. 3 shows histograms of the distribution of dwell times of open and closed events. The histograms were constructed with data from several patches containing single channels in order to accumulate a large enough number of events. In some cases, such as for ␣(␥-␤E540) channels, the long duration of the closed state prevented the accumulation of a large number of events. All histograms were well fitted with a single exponential probability density function. The time constants for the open ( o ) and closed ( c ) states are also shown in Fig. 3. The values of the  durations of open and closed events were also calculated from individual patches, and the mean Ϯ S.E. of all of these values appear in Table I. The progressive decrease in lower P o exhibited by the ␥-␤ chimeric channels was mainly due to longer closed times, from 48 Ϯ 15 ms for ␣(␥-␤S481) channels to 14.  Table I and Fig. 6 result from taking the mean Ϯ S.E. from each individual patch (Table I) or from pooling the data from all patches to construct the dwell time histograms (Fig. 6). Amiloride Affinity of Chimeric Channels-In addition to the differences in P o and kinetics, other properties were noticed to be different among the chimeric channels. The affinity for amiloride was examined by measuring the fractional block of whole-cell currents by increasing concentrations of amiloride in the perfusate. Measurements were done with the two-electrode voltage clamp in the presence of 100 mM Na ϩ in the bath and at a membrane potential of Ϫ60 mV. The K i values for ␣(␥-␤S481) and ␣(␥-␤G529) were 1.75 Ϯ 0.24 and 1.2 Ϯ 0.11 M, respectively (Table II). For ␣(␥-␤L535) and ␣(␥-␤E540), the K i values were 0.25 Ϯ 0.05 and 0.2 Ϯ 0.02, respectively. For the first two chimeric channels, the K i values were similar to that of ␣␤, whereas the K i values of the last two channels were equal to that of ␣␥ channels. These results are in agreement with the notion that residues located in the segment preceding M2 determine amiloride affinity. Since the first two channels have the preceding M2 segment from ␤ subunits, they had the K i of ␣␤ channels (1 M); in contrast, the last two chimeras have the preceding M2 segment from ␥, and therefore they exhibit a K i similar to ␣␥ channels (0.1 M).
Ion Selectivity and Single Channel Conductance-Ion selectivity and single channel conductance were also affected by the M2 subunit composition. The conductance and selectivities for Na ϩ and Li ϩ were determined by the ratio of unitary currents in the presence of a 150 mM concentration of each one of the cations in the pipette solution (I Li ϩ /Na ϩ). When the magnitude of the unitary currents was too small to be measured accurately, the I Li ϩ /Na ϩ ratio was calculated from whole-cell currents in the presence of 150 mM Na ϩ or Li ϩ in the bath solution. All channels were also examined for permeability to K ϩ using bath solutions containing 150 mM K ϩ .
All the functional chimeric channels conducted in the presence of Na ϩ or Li ϩ but none in the presence of K ϩ . ␣(␥-␤S481) and ␣(␥-␤G529) channels were equally permeable to both cations (I Li ϩ /Na ϩ of 1:1). ␣(␥-␤L535) and ␣(␥-␤E540) channels were more permeable to Li ϩ with I Li ϩ /Na ϩ of 1.2:1 and 1.4:1, respectively ( Fig. 7 and Table II). Differences in the I Li ϩ /Na ϩ were due mainly to a decreases in the Na ϩ conductance from 4.7 to 3.7 picosiemens with little change in the Li ϩ conductance, which remained close to 5 picosiemens for all ␣(␥-␤) channels.

DISCUSSION
The Role of ␣ Subunit M2 in Channel Function-Until now, the presence of the ␣ subunit M2 was thought to be critical for expression of functional amiloride-sensitive Na ϩ channels. However, we have shown that the chimeric channels, composed of ␥(␣-␤S508) and (␣-␤S508)(␥-␤S481), both of which lacked the M2 from the ␣ subunit, gave rise to functional channels. When compared with wild-type channels, the properties of ␥(␣-␤S508) and ␥(␣-␤S508)(␥-␤S481) were different. Nonetheless, these channels still possessed the general basic properties of ENaC in that they were selective for Na ϩ and not for K ϩ (but I Na ϩ Ͼ I Li ϩ) and in that channel activity could still be inhibited by amiloride, although at slightly higher doses.
Our observations indicate that the pore forming region of the ␣ subunit is not a prerequisite for the expression of functional

TABLE II
Summary of single channel conductance, Li ϩ /Na ϩ selectivity, and amiloride K i from chimeric channels Single channel conductances were estimated by linear regression between Ϫ20 and Ϫ80 mV. The Li ϩ /Na ϩ selectivity was determined as the ratio of unitary or whole-cell currents in the presence of 150 mM Li ϩ or Na ϩ . The amiloride K i was calculated from the fractional inhibition of whole-cell currents produced by increasing concentrations of amiloride in the perfusate. Data were fitted to Equation 1. All values are mean Ϯ S.E. The number of observations for the different experimental conditions is given in parentheses.

Conductance
Li ϩ /Na ϩ selectivity channels, since we have pores formed by the M2 of ␥ and ␤ (␥(␣-␤S508)) and of ␤ alone ((␣-␤S508)(␥-␤S481)). These observations taken with those of the ␥-␤ chimeras expressed with ␣ subunits demonstrate that all three M2 regions can form functional pores, the properties of which depend on the subunit type(s) present. We also tested chimeras and combinations of subunits in which there were no sequences from the ␣ subunit; none of them increased the Na ϩ permeability when injected in oocytes. Since the chimeras ␥(␣-␤S508) and (␣-␤S508)(␥-␤S481) contain the first 2 ⁄3 of ␣, this suggests that domains within this region of the ␣ subunit are necessary for channel function. Most likely, they contain elements necessary for expression of channels either by promoting assembly of the subunits and/or of channels to the plasma membrane.
There are other lines of evidence that support a potential role for the N-terminal region of the subunits participating in channel function. Residues preceding the first hydrophobic domain of acid-sensitive ion channel 2, a member of the ENaC/degenerin family of ion channels, have been shown to have a role in ion permeation, since mutations within this region dramatically increased the K ϩ permeability of these channels (16). We were unable to address the contribution of the N-terminal region to channel function, because reverse chimeras comprising the N-terminal portion of ␤ and the C-terminal of ␥ were all found to be nonfunctional.
Kinetics of Single Channels and P o -There have been numerous studies to demonstrate the putative roles of various residues in ion permeation and amiloride block (10,11). However, none of these studies attempted to characterize if the various mutations altered channel kinetics or P o . One study performed on bovine ␣ ENaC in lipid bilayers demonstrated that point mutations, in addition to altering amiloride sensitivity and ionic selectivity, increased channel P o (17).
Little is known about the structural domains that form or control the gating mechanism in these channels. ENaC exhibits characteristic slow and voltage independent kinetics with long openings and closures that can last several seconds. Previously, we had reported that channels formed by ␣␤ or ␣␥ displayed different kinetics (13). ␣␤ channels had long open times with brief closures and a P o close to 1, whereas ␣␥ channels had shorter openings with a mean P o of approximately 0.5. These results suggested that domains within the ␥ subunit were responsible for the long closures. To address this question, we analyzed the kinetics of several ␥-␤ chimeras that contained only part of the M2 from ␥ and also channels that did not have any sequences from ␥ M2 (␣-␤S508)(␥-␤S481). The results showed that the M2 region of the subunits markedly influenced the P o and kinetics of channels. However, regardless of the subunit composition, all channels exhibited closures (although the c was of different duration for each type of channel), indicating that the gating mechanism is not exclusively located in any one subunit but that most likely all three subunits contribute to it.
The cytoplasmic region preceding M1 can also alter the kinetics of ENaC channels. Mutations of residues ␤G37S and ␣G95S in this region decrease the P o (18). However, it is unlikely that the N-terminal domain constitutes the gate. The amino acid sequence of the N-terminal domain is one of the least conserved among all members of the ENaC/degenerin family. In addition, a large segment from the N termini of ␣, ␤, and ␥ can be deleted without affecting activity. 2 Most likely, ␤-Gly 37 and ␣-Gly 95 are not located in the gate, but their mutations produce an allosteric effect that alters the P o . On the other hand, we can rule out the gate being located at the level of the C terminus, because deletions of the C terminus of ␤ and ␥ do not alter channel gating (13).
Determinants of Amiloride Affinity and Selectivity-Amiloride blocks all of the channels belonging to the ENaC/degenerin family, although the affinities for amiloride vary widely among each of the individual members. Amiloride blocks the open channel by occluding the pore in a voltage-dependent manner; it senses approximately 15% of the membrane electrical field in wild-type ENaC. Therefore, the site of action of amiloride is most likely at the outer part of the pore, which is formed by the initial segment of M2 (10). The residues ␣-Ser 583 , ␤-Gly 525 , and ␥-Gly 537 all contribute to form a high affinity site (K i ϭ 0.1 M), maybe by forming a ring to accommodate the amiloride molecule. In agreement with this notion, our analysis of the amiloride affinity of ␣(␥-␤) channels showed that the K i was influenced by the first segment of M2. However, the results also showed that, in addition, other factors contribute to define the amiloride affinity. ␣␤, ␣␥, and all four chimeric ␣(␥-␤) channels have rings formed by identical residues, yet their amiloride affinities are different. Among the ␣(␥-␤) channels, we found that the transition from low to high amiloride K i occurred 2 G. K. Fyfe, P. Zhang, and C. M. Canessa, unpublished observations. between chimera (␥-␤G529) (K i ϭ 1.2 Ϯ 0.04) and chimera (␥-␤L537) (K i ϭ 0.25 Ϯ 0.01). The main difference between these chimeras is the presence of residues Gly 529 and Gly 530 in ␤ and of Ser 541 and Cys 542 in ␥. To examine whether these two residues were responsible for conferring the differences in K i , we expressed the mutant channels ␣␤G529S and ␣␤G530C, expecting to obtain channels with low K i . Unfortunately, these mutants were not functional when expressed with ␣ subunits alone but induced large currents when coexpressed with wildtype ␣ and ␥. The amiloride K i values of ␣␤G529S␥ and ␣␤G530C␥ were ϳ0.1 M, indicating that these residues, although essential for ion permeation, do not change the amiloride affinity. The observations suggest that, in addition to the composition of the side chains forming the putative binding site for amiloride, other surrounding amino acids influence the structure of the site and thus alter amiloride affinity.
As expected from the small differences between ␣␤ and ␣␥ channels, all chimeric channels were permeable to Na ϩ and Li ϩ but not to K ϩ . Experiments with ␥-␤ chimeras showed that M2 changes the I Li ϩ /Na ϩ ratio from 1:1 (␣␤ channels) to 1.5:1 (␣␥ channels). The change was due to a progressive decrease in Na ϩ currents without a significant change in the magnitude of Li ϩ currents as indicated in Table II. These results show that the critical residues that determine selectivity and conductance are located in the initial segment of M2 as has recently been shown by Kellenberger et al. (11,19). The corresponding residues in the three subunits (␣-Gly 587 , ␤-Gly 529 , and ␥-Ser 541 ) and a ring of conserved serines (␣-Ser 589 , ␤-Ser 531 , and ␥-Ser 543 ) form the putative selectivity filter. However, differences in I Li ϩ /Na ϩ ratio and in single channel conductance were observed among channels that had identical residues in the selectivity filter. For instance, pores formed only by ␣ subunits (homomeric ␣ channels) have I Li ϩ /Na ϩ of Ͼ1, whereas pores formed only by ␤ subunits ((␣-␤S508)(␥-␤S481)) have I Li ϩ /Na ϩ of Ͻ1. The same arguments presented for the amiloride affinity seem to apply for the determinants of selectivity in that other elements besides these residues affect the structure of the selectivity filter.