Identification of an Amiloride Binding Domain within the α-Subunit of the Epithelial Na+ Channel*

Limited information is available regarding domains within the epithelial Na+ channel (ENaC) which participate in amiloride binding. We previously utilized the anti-amiloride antibody (BA7.1) as a surrogate amiloride receptor to delineate amino acid residues that contact amiloride, and identified a putative amiloride binding domain WYRFHY (residues 278–283) within the extracellular domain of αrENaC. Mutations were generated to examine the role of this sequence in amiloride binding. Functional analyses of wild type (wt) and mutant αrENaCs were performed by cRNA expression in Xenopus oocytes and by reconstitution into planar lipid bilayers. Wild type αrENaC was inhibited by amiloride with aK i of 169 nm. Deletion of the entire WYRFHY tract (αrENaC Δ278–283) resulted in a loss of sensitivity of the channel to submicromolar concentrations of amiloride (K i = 26.5 μm). Similar results were obtained when either αrENaC or αrENaC Δ278–283 were co-expressed with wt β- and γrENaC (K i values of 155 nm and 22.8 μm, respectively). Moreover, αrENaC H282D was insensitive to submicromolar concentrations of amiloride (K i = 6.52 μm), whereas αrENaC H282R was inhibited by amiloride with a K i of 29 nm. These mutations do not alter ENaC Na+:K+ selectivity nor single-channel conductance. These data suggest that residues within the tract WYRFHY participate in amiloride binding. Our results, in conjunction with recent studies demonstrating that mutations within the membrane-spanning domains of αrENaC and mutations preceding the second membrane-spanning domains of α-, β-, and γrENaC alters amiloride’s K i , suggest that selected regions of the extracellular loop of αrENaC may be in close proximity to residues within the channel pore.

Limited information is available regarding domains within the epithelial Na ؉ channel (ENaC) which participate in amiloride binding. We previously utilized the anti-amiloride antibody (BA7.1) as a surrogate amiloride receptor to delineate amino acid residues that contact amiloride, and identified a putative amiloride binding domain WYRFHY (residues 278 -283) within the extracellular domain of ␣rENaC. Mutations were generated to examine the role of this sequence in amiloride binding. Functional analyses of wild type (wt) and mutant ␣rENaCs were performed by cRNA expression in Xenopus oocytes and by reconstitution into planar lipid bilayers. Wild type ␣rENaC was inhibited by amiloride with a K i of 169 nM. Deletion of the entire WYRFHY tract (␣rENaC ⌬278 -283) resulted in a loss of sensitivity of the channel to submicromolar concentrations of amiloride (K i ‫؍‬ 26.5 M). Similar results were obtained when either ␣rENaC or ␣rENaC ⌬278 -283 were co-expressed with wt ␤and ␥rENaC (K i values of 155 nM and 22.8 M, respectively). Moreover, ␣rENaC H282D was insensitive to submicromolar concentrations of amiloride (K i ‫؍‬ 6.52 M), whereas ␣rENaC H282R was inhibited by amiloride with a K i of 29 nM. These mutations do not alter ENaC Na ؉ :K ؉ selectivity nor single-channel conductance. These data suggest that residues within the tract WYRFHY participate in amiloride binding. Our results, in conjunction with recent studies demonstrating that mutations within the membrane-spanning domains of ␣rENaC and mutations preceding the second membrane-spanning domains of ␣-, ␤-, and ␥rENaC alters amiloride's K i , suggest that selected regions of the extracellular loop of ␣rENaC may be in close proximity to residues within the channel pore.
The diuretic amiloride is a prototypic inhibitor of epithelial Na ϩ channels (ENaCs) 1 (1), although amiloride and its various derivatives inhibit many Na ϩ -selective transport proteins. Several laboratories have recently identified domains within the epithelial Na ϩ channel and the Na ϩ /H ϩ exchanger that appear to participate in amiloride binding. Residues within the second membrane-spanning domain of ␣rENaC may interact with amiloride, as mutations of a serine residue at position 589 result in a large decrease of the apparent K i for amiloride and the amiloride analog benzamil, as well as alter cation selectivity (2). Selected mutations of residues within a hydrophobic region, termed H2 (3), immediately preceding the second membrane-spanning domains of the ␣-, ␤-, and ␥-subunits of rENaC (i.e. Trp-␣582, Ser-␣583, Gly-␤525, Gly-␥537) and the ␣-subunit of bovine ENaC (Lys-504, Lys-515) affect the K i for amiloride, and several of these mutations affect single-channel conductance (4,5). Snyder and co-workers have identified splice variants of ␣rENaC in which the C-terminal 199 or 216 amino acid residues, including the second membrane-spanning domain, are truncated (6). These splice variants are not functional when expressed in Xenopus oocytes, but retain amiloride and phenamil binding activity, suggesting that at least a portion of the amiloride and phenamil binding domain is proximal to the C-terminal 216 residues of ␣rENaC.
Pouyssegur and co-workers generated a mutant NHE1 that had an apparent 30-fold decrease in its affinity for methylpropylamiloride. Sequence analysis identified a single point mutation within the putative fourth transmembrane domain, changing a leucine in position 167 to a phenylalanine. Further analysis of this region by site-directed mutagenesis identified phenylalanine residues at positions 165 and 168 that may participate in amiloride binding (7). Analysis of this leucine residue within the putative fourth transmembrane domain of NHE2 yielded similar results (8).
We have previously raised both polyclonal and monoclonal antibodies to amiloride (9 -11). Amiloride was conjugated to carrier protein with linking groups located at different positions on the amiloride molecule to allow distinct sites of the amiloride molecule to be exposed following immunization (10). One amiloride derivative was coupled to bovine serum albumin through a hydrocarbon spacer arm on a terminal nitrogen of its guanidinium moiety (9). This strategy was based on previous observations that several amiloride analogs with hydrophobic substituents at this site are potent inhibitors of epithelial Na ϩ * This work was supported by grants from the Department of Veterans Affairs, by Grants DK51391, DK09215, and DK37206 from the National Institutes of Health, and by Grant CB-11 from the American Cancer Society. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ʈ Recipient of a postdoctoral fellowship award from the Cystic Fibrosis Foundation.
¶ ¶ This work was performed during the tenure of an Established Investigatorship Award from the American Heart Association. To whom all correspondence should be addressed: Medical Research (151), VA Medical Center, University and Woodland Ave., Philadelphia, PA 19104. E-mail: kleyman@mail.med.upenn.edu. channels (1). The binding of anti-amiloride antibodies to amiloride was examined by solid phase immunoassay using amiloride-bovine serum albumin conjugates adsorbed onto a solid support. Polyclonal and several monoclonal anti-amiloride antibodies recognized both benzamil and amiloride, but did not bind ethyl isopropylamiloride (9,10), consistent with the rank order of potency of inhibition of high amiloride affinity epithelial Na ϩ channels (i.e. benzamil Ͼ amiloride Ͼ Ͼ ethyl isopropylamiloride (1)). We utilized one monoclonal anti-amiloride antibody (BA7.1) as a surrogate amiloride receptor (12) to identify the amino acid residue types that may form an amiloride binding site, as well as their topologic orientation. Analysis of structural features of this anti-amiloride antibody led to identification of a structurally related 6-residue tract within the extracellular loop of ␣rENaC (13). We now provide evidence that this 6-amino acid residue tract within the extracellular loop of ␣rENaC, identified as a putative amiloride binding site by its homology with the amiloride binding domain within the anti-amiloride antibody BA7.1 (12), is required to express an epithelial Na ϩ channel that is sensitive to nanomolar concentrations of amiloride.

EXPERIMENTAL PROCEDURES
Materials-Amiloride was a gift from Merck. Lipids were purchased from Avanti Polar Lipids (Alabaster, AL). Moloney murine leukemia virus reverse transcriptase and DNA tailing kit were obtained from Life Technologies, Inc., Taq polymerase from Perkin Elmer, dNTPs from Pharmacia Biotech Inc., Geneclean kit from Bio101, Sequenase II DNA sequencing kit from U. S. Biochemical Corp., T4 ligase from Boehringer Mannheim, Escherichia coli strains from Strategene (La Jolla, CA), restriction enzymes and cap analog from New England Biolabs (Beverly, MA), and pALTER-1 in vitro mutagenesis system, TNT™-coupled reticulocyte lysate system, Ribomax kit, and plasmid Wizard™ miniprep kits from Promega (Madison, WI). ␣rENaC, ␤rENaC, and ␥rENaC cDNAs in the vector pSPORT were a gift from Drs. B. C. Rossier and C. Canessa (University of Lousanne, Lousanne, Switzerland). Monomeric actin was purified from rabbit skeletal muscle (a kind gift from Dr. S. S. Rosenfeld, University of Alabama, Birmingham, AL) or purchased from Sigma. All other reagents were purchased from Sigma.
Preparation of ␣rENaC Mutants-A mutant of ␣rENaC in which amino acid residues 278 -283 were deleted was generated by designing PCR primers to amplify the 5Ј and the 3Ј regions of ␣rENaC flanking the region to be deleted, and introducing a unique XhoI restriction site 3Ј to the deletion to allow the two products corresponding to the 5Ј and 3Ј ends of ␣rENaC to be ligated following digestion with XhoI. The XhoI restriction site was generated by mutating nucleotide C942 to G, and C945 to G. These mutations did not alter the amino acid residues in positions 287 and 288. Primer pairs to amplify the 5Ј region of ␣rENaC were 5Ј-GTACCGGTCCGGAATTCCCGGGTCG-3Ј and 5Ј-CAGTCTC-GAGAGAATGTTGATCTCCCTCACTGCATCCACCCCAGAGGAG-3Ј. PCR was performed by denaturing the reaction mixture at 92°C for 5 min, followed by 35 cycles (0.5 min at 92°C, 1 min at 58°C, and 2 min at 72°C), and a final extension at 72°C for 7 min. A product of the predicted size of ϳ950 base pairs was isolated, purified (Geneclean), and digested with SmaI and XhoI at 37°C for 1 h. Primer pairs to amplify the 3Ј region of ␣rENaC were 5Ј-ACCGCTCGAGACTGTCGGA-CACCTCG-3Ј and 5Ј-TCTAAGGGATGCATAGACTGTGTGTTC-3Ј. PCR was performed by denaturing the reaction mixture at 92°C for 5 min, followed by 35 cycles (92°C for 1 min, 55°C for 2 min, 72°C for 3 min) and a final extension at 72°C for 7 min. A product with the predicted size of 1890 base pairs was isolated, purified (Geneclean), and digested with XhoI and NsiI for 1 h at 37°C. The 5Ј and 3Ј PCRamplified regions of ␣rENaC were ligated into pSPORT, which had been digested with SmaI and NsiI. Plasmids were amplified, purified, and subjected to DNA sequencing through the deletion site by Sanger dideoxynucleotide sequence analysis (14) to confirm that the construct was correct.
␣rENaC site-directed mutants were generated using the Altered Sites™ in vitro mutagenesis system according to the manufacturer's instructions. Wild type ␣rENaC was excised from pSPORT-1 by digesting with SalI and SphI, and then ligated into pALTER-1 via EcoRI and SphI sites with an EcoRI-NotI-SalI adaptor. Single-stranded pALTER-1-␣rENaC template was prepared, and appropriate mutagenic oligonucleotide primers, including an ampicillin repair primer, were annealed to the template and second strand synthesis was performed with T4 polymerase and T4 DNA ligase. Following transformation of E. coli ES1301 mutS, cells containing mutated pALTER-1-␣rENaC were selected on the basis of ampicillin resistance. Plasmid was then purified from individual colonies and subjected to DNA sequencing to confirm the mutation of ␣rENaC. The primer used to generate H282D was 5Ј-ACCGCTTCGATTACATCAAC-3Ј; the primer used to generate H282R was 5Ј-TACCGCTTCCGCTACATCAAC-3Ј; the primer used to generate R280G was 5Ј-TACGGCTTCCATTACATCAAC-3Ј. Plasmids were linearized by overnight digestion with SphI, and cRNA was synthesized using T7 RNA polymerase in the presence of the cap analog m 7 G(5Ј)ppp(5Ј)G using Promega's Ribomax kit according to the manufacturer's instructions.
Channel Expression in Xenopus Oocytes-Oocytes were injected with a total of 25 ng of either ␣,␤,␥rENaC cRNA or ␣⌬278 -283,␤,␥rENaC cRNA, and current recordings were performed 2 days later, as described previously (15). Briefly, recordings were carried out at room temperature (20°C). Oocytes were impaled with recording current and voltage electrodes filled with 3 M KCl (resistances ranging from 0.4 to 3 megohms). Two external electrodes (1 voltage and 1 current) made from Ag-AgCl and were connected to the chamber via 3% agar bridges filled with 3 M KCl. The recording and references electrodes were connected to a four-electrode voltage clamp (TEV-200, Dagan, Minneapolis, MN). Oocytes were voltage-clamped to 0 mV, and their membrane voltage was stepped for 450 ms from Ϫ100 to ϩ100 mV in 20-mV increments to measure whole-cell currents. The potential was returned to 0 mV for 50 ms between each voltage step. Current-voltage (I/V) curves were constructed as described previously (16). Increasing concentrations of amiloride were sequentially added to the bath solution, and current was allowed to stabilize for 5 min between bath additions. Currents measured in the presence of varying concentrations of amiloride were normalized to the currents obtained in the absence of amiloride.
Planar Lipid Bilayer Experiments and Channel Expression-Planar lipid bilayers were made as described previously (17) using a phospholipid solution containing a 2:1:2 mixture of diphytanoyl-phosphatidylethanolamine/diphytanoyl-phosphatidylserine/oxidized cholesterol in n-octane (final lipid concentration ϭ 25 mg/ml). Bilayers were bathed either with symmetric 100 mM NaCl containing 10 mM MOPS-Tris buffer (pH 7.4) or with this solution in the trans compartment and a buffer containing 100 mM KCl, 10 mM MOPS-Tris (pH 7.4) in the cis compartment. In selected experiments, actin was added to the cis compartment. Actin was diluted prior to use to a final concentration of 4 -10 g/ml with a buffer containing 2 mM Tris, 0.2 mM CaCl 2 , 0.2 mM MgATP, 0.2 mM ␤-mercaptoethanol (pH 8.0). The mixture of actin with recombinant plasma gelsolin (dissolved at a concentration of 10 g/ml in 100 mM KCl, 10 mM HEPES buffer (pH 7.4), and dialyzed against 0.2 mM EGTA buffer) at 1:1 ratio was added to both sides of the bilayer as described previously (18). All solutions were made with Milli-Q water and were filter-sterilized by passing the solution through 0.22-m filters (Sterivax-GS filters, Millipore Corporation, Bedford, MA). Current measurements were performed using a high gain amplifier circuit, as described previously (17). Here and elsewhere throughout this report, applied voltage is referred to the trans chamber, which was connected to the current-to-voltage converter and therefore was held at virtual ground. Under these experimental conditions, the channels were oriented with their amiloride-sensitive (extracellular) surface exposed to the trans compartment in over 90% of the incorporations. The probability of successful incorporation of single channels versus multi-channel incorporations depends upon the density of channels in oocyte membranes. In our experience, these probabilities were highly variable, and we have developed a procedure of empirical dilution of oocyte vesicles yielding predominately single-channel incorporations (17). The experiments when no channels were evident in the membrane were considered unsuccessful incorporations as described previously. The rate of successful incorporations from oocyte membrane was 1 in 50 -200 attempts. Amiloride was added to the trans chamber at concentrations indicated in the figures.
Data Analysis-Acquisition and analysis of single-channel recordings were performed using pCLAMP software and hardware (Axon Instruments) as described previously (17). Data were stored digitally, and were filtered at 300 Hz with a 8-pole Bessel filter prior to acquisition at 1 ms/point. All the analyses were performed for membranes containing only single Na ϩ channels. The single-channel open probability was calculated for at least 3 min of continuous recording using Equation 1.
N is total number of channels (always equal to 1 in these experiments, as determined by activating all of the channels present in the bilayer by imposing a hydrostatic pressure gradient (see Ref. 17 for details)), I is the mean current over the period of observation, and i is the main (highest observed) state unitary current determined from all points current amplitude histograms produced by pCLAMP. The mean current (I) over the period of observation was calculated using the events list generated by pCLAMP software and Equation 2.
i m is an event current (all levels, including the zero current level); t m is an event dwell time, and M is the total number of events. Data are expressed as mean value Ϯ 1 standard deviation for n experiments.

RESULTS
Limited information is available regarding ENaC domains that participate in amiloride binding. We previously utilized the anti-amiloride antibody BA7.1 as a surrogate amiloride receptor (12) to delineate amino acid residues that contact amiloride. We observed that the sequence YYGHY contained in the CDR3 domain of the heavy chain of mAb BA7.1 aligned with the sequence tract WYRFHY, corresponding to residues 278 -283 within ␣rENaC (13,19). It is likely that the ␣-subunit of the epithelial Na ϩ channel possesses an amiloride binding site, as expression of the ␣-subunit alone is sufficient to induce expression of a Na ϩ current in Xenopus oocytes, which is inhibited by amiloride with apparent inhibitory constants (K i ) nearly identical to that observed with expression of ␣,␤,␥rENaC heterotrimers (K i values of 100 and 104 nM, respectively). Similar results have been reported with expression of ␣rENaC alone or expression of ␣,␤,␥rENaC heterotrimers in planar lipid bilayers (K i values of 170 nM for both channels) (3,15,17,19). The putative amiloride binding tract WYRFHY is within the extracellular domain of ␣rENaC (20 -22). To examine the role of this sequence in amiloride binding, we generated several mutants of ␣rENaC. One mutant, ␣rENaC ⌬278 -283, has a deletion of residues 278 -283 (i.e. WYRFHY). Our analysis of the binding of amiloride to BA7.1 suggested that the histidine within the CDR3 region of the heavy chain stabilized amiloride binding via electrostatic interactions with the halide (Cl) on the pyrazine ring of amiloride (12). Therefore, two site-directed mutants were generated: ␣rENaC H282D, in which the histidine at position 282 was mutated to aspartic acid, and ␣rENaC H282R, in which the histidine at position 282 was mutated to arginine. An additional site-directed mutant, ␣rENaC R280G, was also generated.
Functional Properties of Mutations within the Sequence Tract WYRFHY-A major technical difficulty in examining the functional properties of ␣rENaC ⌬278 -283 in the Xenopus oocyte expression system is the low level of Na ϩ current observed when the ␣-subunit is expressed alone (i.e. without co-expression of ␤and ␥rENaC) (3). Therefore, experiments using this system were performed with the heterotrimeric channel. Fig. 1 shows the results of measurements of macroscopic currents in Xenopus oocytes expressing wt ␣,␤,␥rENaC or ␣⌬278 -283,␤,␥rENaC. Interestingly, oocytes expressing ␣⌬278 -283,␤,␥rENaC displayed a current that was ϳ40% smaller than that in oocytes expressing wt ␣,␤,␥rENaC. The deletion of residues 278 -283 within ␣rENaC may affect the association of subunits forming the channel and alter subsequent transit of the channel to the cell surface, or alter singlechannel properties. The amiloride-sensitive portion of the cur-  8). These data suggest that ␣⌬278 -283,␤,␥rENaC expressed in Xenopus oocytes is much less sensitive to amiloride than the wt counterpart. rent (10 M amiloride added to the bath) was much larger in the oocytes expressing wt ␣,␤,␥rENaC (63 Ϯ 8%) than in oocytes expressing ␣⌬278 -283,␤,␥rENaC (19 Ϯ 3%). These observations are consistent with those reported by Busch et al. (23). A portion of the base-line current in water-injected oocytes was also found to be amiloride-sensitive (17 Ϯ 4%). Inhibition of wt ␣,␤,␥rENaC by amiloride can be described in terms of Michaelis-Menten kinetics (Fig. 2) with a K i of 231 Ϯ 46 nM (n ϭ 4), in reasonable agreement with previous observations (3). The low current induced by ␣⌬278 -283,␤,␥rENaC and the limited inhibition by 10 M amiloride precluded the detailed analysis of amiloride-induced inhibition of ␣⌬278 -283,␤,␥rENaC expressed in oocytes. A rough estimation of Michaelis-Menten kinetics, based on analyses of the normalized current and assuming that 60% of the total current is mediated by the Na ϩ channel, gave a K i of 35 Ϯ 10 M (n ϭ 9) for the mutant channel. In view of the low levels of macroscopic current observed in Xenopus oocytes expressing ␣rENaC alone, further studies examining the properties of ␣rENaC ␣⌬278 -283rENaC, and of ␣rENaC with selected mutations within the WYRFHY tract (residues 278 -283) were performed in planar lipid bilayers.
When reconstituted in planar lipid bilayers, ␣rENaC ⌬278 -283, ␣rENaC R280G, ␣rENaC H282D, and ␣rENaC H282R formed Na ϩ channels essentially indistinguishable by conductance or gating from those produced by wt ␣rENaC (Fig. 3). All mutated ␣rENaC channels display a concerted type gating between 13 and 39 pS states consistent with what was reported previously (17). We have previously observed this gating pattern for ␣rENaC alone, and for ␣,␤rENaC or ␣,␥rENaC heterodimers, which also display a concerted type gating between the 13 pS and 39 pS states and which is quite distinct from the gating pattern observed with heterotrimeric channels (17). However, similar to the experiments with heterotrimeric channels containing ␣rENaC ⌬278 -283, the mutant channels exhibited altered sensitivities to amiloride (Figs. 3 and 4). Wild type ␣rENaC was inhibited by amiloride with a K i of 169 Ϯ 15  Table I). Hill coefficients (n) were obtained using a best fit approach and are indicated for each plot.  Table I). However, a conservative mutation of His-282 (␣rENaC H282R) led to a decrease in the apparent K i for amiloride by 5.8-fold. The ␣rENaC mutant R280G had an apparent K i for amiloride of 830 Ϯ 70 nM (n ϭ 6), a 4.9-fold increase when compared with wt ␣rENaC. Similar apparent K i values for amiloride were obtained by fitting the amiloride dose-response data to either the first order Michaelis-Menten equation or the Michaelis-Menten equation with the Hill coefficient (Table I). These data support the hypothesis that residues within the tract WYRFHY are required for expression of epithelial Na ϩ channels that are sensitive to nanomolar concentrations of amiloride, and suggest that amiloride binds to, or interacts with, residues within this tract.
Single-channel properties of ␣rENaC channels studied in planar lipid bilayers under these conditions differ from heterotrimeric rENaC channels expressed in Xenopus oocytes and studied by the patch-clamp technique (3,17). Therefore, we examined the properties of wt ␣,␤,␥rENaC and of ␣⌬278 -283,␤,␥rENaC in planar lipid bilayers. Incorporation of mem-brane vesicles obtained from oocytes co-expressing wt ␣,␤,␥rENaC or ␣⌬278 -283,␤,␥rENaC in planar lipid bilayers generated channels which displayed an essentially identical gating pattern with a predominant residence in a 13-pS state and occasional openings to 39 pS (Fig. 5, top traces) and consistent with our previous findings of the heterotrimeric ENaC currents observed in bilayers (17). However, the sensitivity of these channels to inhibition by amiloride was significantly different. The heterotrimeric channel containing ␣rENaC ⌬278 -283 was inhibited by amiloride at concentrations more than 2 orders of magnitude higher than wt ␣,␤,␥rENaC (Table  I). Amiloride dose-response curves are illustrated in Fig. 6. The amiloride inhibitory constants (K i values) for the deletion mutant (i.e. ␣⌬278 -283,␤,␥rENaC) and wt ␣,␤,␥rENaC  Table I). These results are similar to the apparent K i for amiloride of wild type ␣,␤,␥rENaC and the estimated apparent K i for amiloride of ␣⌬278 -283,␤,␥rENaC expressed in oocytes.
We have previously demonstrated that the addition of actin  to the cis compartment alters properties of wt ␣,␤,␥rENaC reconstituted into planar lipid bilayers, by reducing singlechannel conductance to 6 pS and by increasing mean open and closed times, characteristics similar to those observed for ENaCs expressed in native tissues and analyzed by patchclamp (18,24). The sensitivities to amiloride of wt ␣,␤,␥rENaC and of ␣⌬278 -283,␤,␥rENaC were not altered when actin was added to the cis compartment (Figs. 6 and 7), although singlechannel conductance was reduced to 6 pS.
Additional studies were performed to examine whether deletion of the WYRFHY tract (residues 278 -283) or mutations within this tract affect selectivity properties of the channel. The Na ϩ :K ϩ selectivity of ␣rENaC ⌬278 -283, ␣rENaC R280G, ␣rENaC H282D, and ␣rENaC H282R did not differ from wt ␣rENaC (Fig. 8), as determined in symmetrical NaCl and biionic (NaCl trans:KCl cis) conditions in planar lipid bilayers. In addition, Na ϩ :K ϩ selectivity ratio of the heterotrimeric Na ϩ channel was not altered by deletion of residues 278 -283, as determined by the incorporation of channels in planar lipid bilayers (Fig. 9), or by expression of channels in Xenopus oocytes using the two electrode voltage clamp technique (Fig. 10). The plots in these graphs represent fit of the data obtained in Na ϩ to K ϩ substitution experiments using the Goldman-Hodgkin-Katz (GHK) equation, where I Na ϭ P Na e EF 2 /RT ͓͑Na ϩ ͔ cis Ϫ ͓Na ϩ ͔ trans e ϪEF/RT ͒/͑1 Ϫ e ϪEF/RT ͒ (Eq. 4) and I X ϭ P X e EF 2 /RT ͓͑X ϩ ͔ cis Ϫ ͓X ϩ ͔ trans e ϪEF/RT ͒/͑1 Ϫ e ϪEF/RT ͒ (Eq. 5) As the intracellular Na ϩ and K ϩ concentrations in Xenopus oocytes were not directly measured, these concentrations were set as adjustable parameters when determining Na ϩ :K ϩ selectivity of rENaC expressed in oocytes by the GHK equation (equations [3][4][5]. The control I/V curves for wt ␣,␤,␥rENaC and mutant ␣⌬278 -283,␤,␥rENaC rENaC were obtained in a bath solution containing 96 mM NaCl and 2 mM KCl. Under these conditions the best curve fit was achieved with the ratios of P Na :P K of 4:1 and 8:1 for wt ␣,␤,␥rENaC and for the WYRFHY tract deletion mutant, respectively (Fig. 10). The intracellular Na ϩ and K ϩ concentrations of 14 mM and 148 mM for wt ␣,␤,␥rENaC, respectively, and 31 mM and 158 mM for ␣⌬278 -283,␤,␥rENaC, respectively, were computed as the parameters for best fit to the GHK equation (equations [3][4][5]. Substitution of 96 NaCl in the bath with 15 mM Na ϩ /83 mM K ϩ shifted the I/V curve for both wt and the WYRFHY tract deletion mutant (Fig. 10). The best fit to GHK equation under these conditions was obtained with P Na :P K ratios of 31:1 and 20:1 for wt ␣,␤,␥rENaC and ␣⌬278 -283,␤,␥rENaC, respectively. Intracellular Na ϩ and K ϩ concentrations of 11 mM and 186 mM for wt ␣,␤,␥rENaC, respectively, and of 12 mM and 223 for ␣⌬278 -283,␤,␥rENaC, respectively, were computed as the parameters for best fit to the GHK equation. These results are consistent with the findings in planar lipid bilayers, suggesting that the mutations we have generated within ␣ENaC do not change its Na ϩ to K ϩ permeability ratio. Substituting all the Na ϩ in the bath for K ϩ (100 mM KCl buffer) resulted in a dramatic shift of the I/V curve for both wt ␣,␤,␥rENaC and for the WYRFHY deletion mutant (Fig. 10). Fitting these curves to GHK results in estimation of the Na ϩ to K ϩ permeability ratios of 1.6 ϫ 10 10 for wt ␣,␤,␥rENaC, and 3.1 ϫ 10 10 for ␣⌬278 -283,␤,␥rENaC, values that are close to infinity. Intracellular Na ϩ and K ϩ concentrations computed as parameters for these fits were found to be 1.2 mM and 198 mM for wt ␣,␤,␥rENaC, respectively, and 0.9 mM and 167 mM for ␣⌬278 -283,␤,␥rENaC, respectively. Ion concentration dependence of the selectivity properties has been previously shown for channels that can accommodate multiple ions at the time (25)(26)(27), and ␣,␤,␥rENaC indeed is a multi-ion channel (28). In this case a close proximity in the of ion concentration dependence of the selectivity properties of the wt and mutant channel may suggest that deletion of the WYRFHY tract does not affect the number of ions that the channel can accommodate at the same time. On the other hand these computer-generated estimates are rough, especially for the complete Na ϩ to K ϩ substitution experiments. Nonetheless, the relative changes in cation selectivity associated with changes of the ionic composition are similar among wt and mutant channels, consistent with previous measurements made with the use of bilayer system. DISCUSSION Amiloride is the prototypic inhibitor of epithelial Na ϩ channels. Amiloride analogs have been used by a number of investigators as tools to isolate and characterize amiloride-sensitive Na ϩ channels, and may have an important role as Na ϩ channel inhibitors in the treatment of selected forms of hypertension (29,30). Previous studies demonstrated that it is the charged, protonated species of amiloride that inhibits Na ϩ channels (31)(32)(33)(34). This channel block is dependent, in part, on the apical membrane potential. Analysis of the kinetics of amiloride binding to the Na ϩ channel in the presence of a varying apical plasma membrane potential suggests that amiloride senses between 10% and 45% of the membrane electric field (31,34). This observation, when taken together with studies utilizing voltage-clamped cells to demonstrate that amiloride binding kinetics are altered by Na ϩ or Li ϩ loading cells to generate an outward current through the channel (35,36), supports the idea that amiloride interacts within the channel pore. This hypothesis is further supported by recent studies of Waldmann et al. demonstrating that substitutions of the first or second putative transmembrane region of ␣rENaC with the corresponding domains within Mec-4 led to a decrease amiloride sensitivity by 3-and 14-fold, respectively (2). Mec-4 has significant sequence similarity with ENaCs and is a member of the Caenorhabditis elegans degenerin family that is associated ). Hill coefficients (n) were obtained using a best fit approach and are indicated for each plot. The addition of actin to the cis compartment did not alter the K i for amiloride.
with mechanotransduction (19). Mutation of a serine to phenylalanine in position 589 of the second membrane-spanning domain of ␣rENaC increases the K i for amiloride and alters cation selectivity and single-channel conductance, suggesting that serine 589 participates in amiloride binding and resides within the channel pore (2). In addition, residues preceding the second membrane-spanning domains of ␣-, ␤-, and ␥rENaC and of bovine ␣ENaC may also form part of the channel pore, and selected mutations within these regions dramatically affect amiloride sensitivity (4,5).
Organic cations other than amiloride and related analogs, such as 2,4,6-triaminopyrimidine, also function as epithelial Na ϩ channel inhibitors, although with K i values much greater than that of amiloride (37), suggesting that it is the guanidine moiety of amiloride that is interacting with the channel pore. However, the substituted pyrazine ring of amiloride is required for the drug to inhibit ENaCs with a submicromolar K i , and may have a critical role in stabilizing amiloride bound to the channel (1,38). The putative amiloride binding site on the anti-amiloride antibody BA7.1 that we previously characterized primarily interacts with the substituted pyrazine ring moiety of amiloride (12). By analogy, the 6-amino acid track within the extracellular loop of ␣ENaC we identified based on its homology with the amiloride binding site on BA7.1 likely binds amiloride via interactions with the substituted pyrazine moiety of amiloride, and is not necessarily associated with the pore region of the channel. Our results demonstrating that the single-channel conductances and Na ϩ :K ϩ selectivity ratios of ␣rENaC ⌬278 -283, ␣rENaC R280G, ␣rENaC H282D, and ␣rENaC H282D are indistinguishable from wt ␣rENaC suggest that residues 278 -283 do not form part of the Na ϩ channel pore. Li and co-workers have identified splice variants of ␣rENaC, in which the C-terminal 199 or 216 amino acid residues are truncated, including the second membrane-spanning domain (6). These splice variants are not functional when expressed in Xenopus oocytes, but retain amiloride and phenamil binding activity, suggesting that part of the amiloride and phenamil binding site is proximal to the C-terminal 216 residues of ␣rENaC, again consistent with our findings. The previously published results suggesting that residues within the second membrane-spanning domain of ␣rENaC and within a hydrophobic (putative pore) region preceding the second membrane-spanning domains of ␣-, ␤-, and ␥rENaC bind amiloride, and our results, indicating that residues within the extracellular loop of ␣rENaC (i.e. residues 278 -283) bind amiloride, suggest that amiloride contact residues are derived from different regions of rENaC and that selected regions of the extracellular loop of ␣rENaC may be in close proximity to residues within the Na ϩ channel pore.
Our previous analysis of the amiloride binding site on the anti-amiloride antibody BA7.1 suggested that a histidine residue within the CDR3 region of the heavy chain primarily interacts with the Cl atom on the pyrazine ring moiety of amiloride through an electrostatic interaction (12). Therefore, we examined the effects of mutations of the histidine residue (His-282) present within the 6 amino acid putative amiloride binding domain on the extracellular loop of ␣ENaC. Mutagenesis of this histidine to aspartic acid (H282D) resulted in a change in charge of this residue from cationic to anionic and was associated with a 39-fold increase in the apparent K i for amiloride. Alternatively, mutagenesis of this histidine to arginine (H282R) with conservation of the cationic charge was associated with a 6-fold decrease in the apparent K i for amiloride. H282R is the first mutation of ␣ENaC described that is associated with a large decrease in the amiloride K i . These data suggest that His-282 may have an important role in stabilizing the binding of amiloride to the Na ϩ channel. Although it is possible that His-282 primarily interacts with the Cl atom on the pyrazine ring moiety of amiloride, there is no direct evidence to support this hypothesis. H282R and H282D have a difference in their apparent K i values for amiloride of 225-fold. The ␣rENaC mutant R280G was inhibited by amiloride with a K i of 830 nM, a 4.9-fold increase in the apparent K i for amiloride when compared with wt ␣rENaC suggesting that residues within the tract 278 -283, other than His-282, may bind amiloride or affect the topology of this site.
The Hill coefficient of 2.4 that we observed in our studies of amiloride inhibition of wt ␣,␤,␥rENaC and wt ␣rENaC reconstituted into planar lipid bilayers is in reasonable agreement with previous observations (17). ENaC stoichiometry has not been determined. If amiloride interacts primarily with the ␣-subunit, a Hill coefficient of 2.4 indicating cooperative binding suggests that ENaCs have more than one ␣-subunit or, alternatively, that there are multiple sites, or domains, within the ␣-subunit which participate in amiloride binding. Interestingly, the Hill coefficient decreased with ␣rENaC mutations that increased amiloride's K i , suggesting that the amiloride binding domain we have identified may affect subunit-subunit interactions, may affect intramolecular interactions within the ␣-subunit, or alternatively may alter ␣-subunit stoichiometry.
The amiloride binding domain WYRFHY is conserved within ␣ENaC in all species that have been cloned and sequenced to date, including rat, human, bovine, Xenopus, and mouse (15, 19, 39 -41). A nearly identical tract, WYHFHY, is present in the recently cloned ␦ subunit of a Na ϩ channel that appears to be expressed in both epithelial and nonepithelial tissues (42). Both ␣ENaC and ␦ENaC are sufficient, by themselves, to induce the expression of Na ϩ -selective, amiloride-sensitive channels in Xenopus oocytes. The current levels observed with expression of ␣ENaC or ␦ENaC increase by approximately 100fold when co-expressed with ␤and ␥ENaC. An amiloride-and benzamil-sensitive FMRFamide peptide-gated Na ϩ channel (FaNaCh) was recently cloned from marine snail neurons (43). Interestingly, FaNaCh does not have a WYRFHY tract, although a related tract WLRFIQKF is present in the putative extracellular domain of FaNaCh that shares some features . Currents were measured in each oocyte with the ND96, 15 mM Na ϩ , and 100 mM KCl buffers. Oocytes were sequentially bathed in the following buffers: ND96, 15 mM Na ϩ , or 100 mM KCl. The chamber was washed for 10 min with the buffer and a voltage clamp protocol performed. Ten M amiloride was added, and after 4 min, the voltage clamp protocol was repeated. Data points represent amiloride-sensitive currents. The measurements were performed in Xenopus oocytes expressing wt ␣,␤,␥rENaC (n ϭ 3) and in Xenopus oocytes expressing ␣⌬278 -283,␤,␥rENaC (n ϭ 4). Lines through data points represent best fits of the Goldman-Hodgkin-Katz equation (Equations 3-5) with intracellular ion concentrations and Na ϩ :K ϩ permeability ratios set adjustable.
with the amiloride binding domain we have identified within ␣ENaC, including the presence of planar and cationic amino acid residues. Further studies are required to examine whether residues within the tracts WYHFHY in ␦ENaC and WLR-FIQKF in FaNaCh participate in amiloride binding.
The sensitivity of ␣rENaC to amiloride does not appear to be dependent upon co-expression with ␤and ␥rENaC, as wt ␣rENaC and wt ␣,␤,␥rENaC reconstituted into planar lipid bilayers have similar K i values for amiloride (155 nM and 169 nM, respectively (Table I)). This was also observed with ␣rENaC ⌬278 -283, as ␣rENaC ⌬278 -283 and ␣⌬278 -283,␤,␥rENaC have nearly identical sensitivities to amiloride (K i values of 22.8 M and 26.5 M, respectively; Table I). These data support the hypothesis that residues required to form high affinity amiloride binding domains reside within the ␣-subunit. However, ␤and ␥ENaC also participate in amiloride binding (4). Interestingly, a tract WYKLHY (residues 230 -235) within the extracellular loop of ␥rENaC bears striking similarity to the amiloride binding domain we identified within the extracellular domain of ␣ENaC (3). Additional studies are required to determine whether this region within ␥rENaC (i.e. residues 230 -235) participates in amiloride binding. It is conceivable that mutations we have generated alter the stoichiometry of subunit association, which might affect amiloride binding. Previous studies from our laboratory suggest that ␣rENaC, reconstituted alone or as ␣/␤ or ␣/␥ heterodimers, primarily exhibits 13-pS and 39-pS conductance states, whereas the ␣/␤/␥ heterotrimer primarily exhibits a 13-pS conductance state (17). The conductance states observed with both wt ␣,␤,␥rENaC and with ␣⌬278 -283,␤,␥rENaC (Fig. 3A) indicate that ␣⌬278 -283,␤,␥rENaC is reconstituted into the lipid bilayer as a heterotrimer.
In summary, the analysis of an anti-amiloride antibody resulted in the identification of an amiloride binding domain on ␣ENaC. Although there are other sites within ␣ENaC, as well as within other ENaC subunits that participate in amiloride binding, our data clearly suggest that residues 278 -283 within ␣rENaC, particularly His-282, are part of an amiloride binding site. In addition, these studies support the hypothesis that selected anti-ligand antibodies may serve as surrogate ligand receptors, and that in selected systems these antibodies may provide useful tools to develop models of tertiary structural features of naturally occurring ligand receptors.