The highly selective epithelial Na⁺ channel ENaC belongs to a new class of proteins called the ENaC/degenerin (DEG) superfamily, which includes a variety of proteins found in mammals, nematodes, flies, and snails that are involved in transepithelial Na⁺ transport, mechanotransduction, and neurotransmission (Mano and Driscoll, 1999; Kellenberger and Schild, 2002). ENaC expressed in the apical membrane of epithelial cells mediates sodium reabsorption in the distal nephron, the colon, and the lung. In the distal nephron, ENaC activity is tightly regulated by aldosterone and vasopressin, serving to fine-tune renal Na⁺ excretion (Garty and Palmer, 1997). This fine regulation of the distal sodium handling is critical for the maintenance of a balance between sodium intake and elimination. Genetic defects of ENaC activity lead either to severe hypertension in case of channel gain-of-function as found in Liddle syndrome or to salt losing nephropathy in case of loss of ENaC function in pseudohypoaldosteronism type-1 (Hansson et al., 1995; Chang et al., 1996; Lifton et al., 2001).

Sodium absorption in the distal nephron is tightly linked to potassium excretion. Pharmacological blockers of ENaC such as amiloride or triamterene increase renal Na⁺ excretion and decrease K⁺ excretion and are therefore used as K⁺-sparing diuretics. Amiloride and its derivative benzamil are high-affinity blockers of ENaC from the external side with an IC₅₀ in the submicromolar range (Kleyman and Cragoe, 1988; Garty and Palmer, 1997). Initial work on amiloride block of the epithelial Na⁺ channel showed that amiloride interacts competitively with the permeant Na⁺ or Li⁺ ions, suggesting that amiloride binds within the external vestibule of the channel pore acting as a channel pore blocker (Palmer and Andersen, 1989; Palmer, 1990).

ENaC is formed by homologous subunits that have two membrane-spanning segments (M1 and M2) linked by a large extracellular loop. ENaC is a heterotetramer made of 2α, 1β, and 1γ subunits arranged pseudosymmetrically around the channel pore in a αβαγ configuration. Structure-function experiments showed that mutations of amino acid residues in the extracellular loop close to M2 of α, β, and γ ENaC subunits dramatically decrease the channel affinity for amiloride (Schild et al., 1997). These amino acid residues, αS583, and the homologous βG525 and γG537 are accessible from the extracellular side and are located directly upstream of the selectivity filter, indicating that they represent a potential binding site for amiloride in the external channel pore.

However, direct evidence that the residues αS583, βG525, and γG537 indeed form the amiloride binding site is still lacking. It remains possible that these mutations induce steric changes that indirectly affect amiloride block by impairing the accessibility of the ligand to its binding site. In addition, it is still unknown whether amiloride analogs and triamterene share with amiloride the same receptor.

To answer these questions, we have investigated the effects of mutations of αS583, βG525, and γG537 on amiloride, benzamil, and triamterene block of ENaC current at the macroscopic and at the single-channel level. We show that these mutations equally affect ENaC block by the three agents tested, indicating that these drugs share a common blocking mechanism. We provide evidence that the mutations lower the affinity of the channel to amiloride and benzamil by increasing the dissociation rate of the blocker from its receptor. We conclude that αS583, βG525, and γG537 are part of the receptor for ENaC blockers in the external channel pore.

Materials and Methods

Site-Directed Mutagenesis and Expression in Xenopus laevis Oocytes. Site-directed mutagenesis was performed on rat ENaC cDNA as described previously (Schild et al., 1997). Complementary RNAs of each αβγ subunit were synthesized in vitro. Stage V-VI oocytes were surgically removed from the ovarian tissue of female X. laevis frogs that had been anesthetized by immersion in MS-222 (2 g/l; Sandoz, Basel, Switzerland). The oocytes were defolliculated, and healthy stage V and VI X. laevis oocytes were pressure-injected with 100 nl of a solution containing equal amounts of α, β, and γ ENaC subunits or α and β subunits at a total concentration of 100 ng/μl. To obtain a lower channel density for αβγ ENaC single-channel patches we reduced the amount of RNA injected. For simplicity, mutants are named by the mutated subunit only, although always all three subunits (α, β and γ) were coexpressed, where not noted explicitly.

Electrophysiological Analysis. Electrophysiological measurements were taken at 16 to 48 h after injection. Macroscopic currents for drug inhibition curves were recorded using the two-electrode voltage-clamp technique at a holding potential of −100 mV. Currents were recorded with a TEV-200 amplifier (Dagan, Minneapolis, MN) equipped with two bath electrodes. For drug inhibition curves, the solution contained 70 mM N-methyl d-glucamine, 20 mM NaCl, 20 mM KCl, 10 mM HEPES, and 0.2 mM CaCl₂, pH 7.4. An analogous solution, in which the 20 mM NaCl was replaced by 20 mM N-methyl d-glucamine, was used to determine the zero current level. Amiloride, benzamil, and triamterene were from Sigma (Buchs, Switzerland) and all other chemicals were from Sigma or Fluka (Buchs, Switzerland).

Single-channel currents and macroscopic currents for the determination of the blocker off-rate were measured in the outside-out configuration of the patch-clamp technique essentially as described previously (Kellenberger et al., 1999b). The external solution in patch-clamp experiments contained 110 mM NaCl or LiCl, 1.8 mM CaCl₂, and 10 mM HEPES-NaOH, pH 7.35. The Na⁺ concentration of this solution was different from that of solutions used for whole-cell current measurements. Because Na⁺ is known to compete with amiloride for its binding site (Garty and Palmer, 1997), we have measured amiloride block of wt ENaC at different Na⁺ concentrations with the two-electrode voltage-clamp to quantify the effect of Na⁺. The IC₅₀ of amiloride block was 0.07 ± 0.01 μM (n = 145) in Na⁺-free extracellular solution containing 20 mM Li⁺ as conducting ion, 0.11 ± 0.01 μM (n = 24) with 20 mM extracellular Na⁺, and 0.24 ± 0.01 μM (n = 17) with 110 mM extracellular Na⁺. This shows that the effect of Na⁺ concentration on amiloride block is relatively small in our experimental conditions. Changing of extracellular solutions in patch-clamp experiments was done using the Rapid Solution Changer RSC-200 (Biologic, Claix, France). The pipette solution contained 75 mM CsF, 17 mM N-methyl-d-glucamine, 10 mM EGTA, and 10 mM HEPES, pH 7.35. The inclusion of fluoride facilitated the formation of stable excised outside-out patches.

Pipettes were pulled from Borosilicate glass (WPI, Sarasota, FL). Pipettes for macropatches were used with a resistance of 1 to 3 MΩ, and pipettes for single-channel measurements had a resistance of 4 to 10 MΩ, when filled with the pipette solution. In patch-clamp experiments, currents were recorded with a List EPC-9 patch clamp amplifier using a PC-based data acquisition system (Pulse; HEKA Electronic, Lambrecht/Pfalz, Germany) and filtered at 1,000 to 2,000 Hz for single-channel analysis. Single-channel openings were detected by standard half-amplitude threshold analysis using TAC 4.09 (Bruxton Corporation, Seattle, WA). Patches for the kinetic analysis of αβ ENaC channels contained one single channel. Other patches contained one to four channels. Channel number was estimated from the maximum number of overlapping openings. Open times and closed times were analyzed from the binned data with durations >200 μs, using TACFIT (Bruxton Corporation). In channels with short closings (as the mutants αS583C and βG525C in the presence of amiloride or benzamil), open duration time constants may be overestimated because of missed short closings. This error was in all cases less than 4%, as calculated according to Colquhoun and Hawkes (1995).

It is known that ENaC channels can be found with very different open probabilities (P_o). Experiments were performed on αβγ ENaC channels with P_o ≥ 0.2, which constituted the majority of the ENaC in our patches. In addition, changes in P_o during the course of the experiment may affect the result. In our conditions of excised outside-out patch-clamp recording, ENaC activity in some experiments slightly decreased during the first 2 to 3 min after formation of the excised patch. Then, it generally stabilized at a new level. To reduce errors caused by a possible slow rundown of channel activity, we alternated the different experimental conditions (i.e., absence of drug, presence of drug at different concentrations) several times during the lifetime of the patch. The following analysis of P_o in experiments with αβ ENaC illustrates the stability of channel activity over time. The P_o of unblocked channels was 0.93 ± 0.06 at the beginning and 0.95 ± 0.02 at the end of the experiments. Averaged data are presented as mean ± S.E.M.

Results

A Common Binding Site for Amiloride and Its Analogs. Individual substitutions by cysteine of the residues αS583, βG525, and γG537, located in the external pore entry of ENaC, shift the concentration for half-maximal inhibition (IC₅₀) to a ∼20-fold higher concentration for the mutation in the α subunit, and to a ∼1,000-fold higher concentration for the β and γ mutations (Schild et al., 1997) (see alignment Fig. 1A). To determine whether these mutations have the same effect on other blockers of ENaC, we have tested the current inhibition by the amiloride-derivative benzamil and the analog triamterene (Fig. 1B). In ENaC wt, the IC₅₀ for benzamil block is 11 nM and the IC₅₀ for triamterene block is 4.5 μM (Fig. 1C). Thus, compared with amiloride, which has an IC₅₀ of ∼0.1 to 0.2 μM (Garty and Palmer, 1997; Schild et al., 1997), ∼10-fold lower concentrations of benzamil, and ∼50-fold higher concentrations of triamterene are needed to block ENaC. The 10-fold higher affinity for benzamil relative to amiloride is probably caused by an additional hydrophobic interaction as a result of the added benzyl group. It should also be noted that only the protonated form of amiloride blocks ENaC (Garty and Palmer, 1997). The pK_a of triamterene is 6.2, compared with 8.7 for amiloride. Thus, at pH 7.4, 95% of amiloride, but only 6% of triamteren is protonated, explaining in part the different concentration dependence of ENaC block. The mutations in the α, β, and γ subunit (αS583, βG525, and γG537) shifted the IC₅₀ values for benzamil to 5-, 1180-, and 320-fold higher concentrations (Fig. 1C). The triamterene IC₅₀ was shifted by a factor of 5 by the αS583C mutation. Triamterene at a concentration of 1 mM inhibited only ∼20% of currents carried by ENaC containing the mutations βG525C and γG537C. Because of the limited solubility of triamterene, it was not possible to test higher concentrations. Thus, it was not possible to determine exactly the IC₅₀ of triamterene block of ENaC containing the β or γ mutation; however, the data suggest that the mutations shift the inhibition curve of triamterene by a similar extent to higher concentrations as observed with amiloride and benzamil inhibition (Fig. 1C). These data show that αS583, βG525, and γG537 mutations affect similarly the block of ENaC by amiloride, benzamil, and triamterene, suggesting overlapping binding sites for these ligands.

We have verified the importance of the αS583, βG525, and γG537 residues on ENaC block by amiloride, by generating more conservative amino acid substitutions than cysteine substitutions, i.e., mutations of Ser (αENaC) to Ala and Gly, and of Gly (β, γ) to Ser and/or Ala. As shown in Fig. 1D, the mutations αS583G and αS583A shifted the IC₅₀ of amiloride block by a factor of 6 and 19, respectively; the mutation βG525A by a factor of 1,220; and the mutation γG537S by a factor of 136. This indicates that mutations to Ala cause the same shift of the inhibition curves as mutations to Cys, whereas mutations to Ser or Gly have a smaller, but significant effect on channel block by amiloride.

Kinetics of αβENaC Block. The ENaC block by amiloride or analogs corresponds to a reversible binding process according to a simple model as follows: where α and β are the opening and closing rate constants of the channel, respectively, and k_on and k_off are the association and dissociation rate constants, respectively, for the ligand L (e.g., amiloride). This model implies that binding of the ligand to a single site of the open channel, O, results in a blocked state B of the channel complexed with the ligand L. We assume that the blocker binds to the open channel, although no experimental evidence excludes the possibility that amiloride can also bind to the closed channel C. Simulation with this and a more complex model involving a blocked state of the closed channel showed that both models gave essentially the same result (data not shown). The scheme predicts that the dwell times of the open state, τ_o, and the dwell times of the blocked state of the channel, τ_B, are inversely proportional to k_on and k_off according to the following relationships: and These relationships indicate that the time constant for the blocked state of the channel τ_B is independent of the amiloride concentration. It should be pointed out that according to eq. 2, the time constant for the open state depends not only on the k_on for amiloride but also on the rate constant of the channel closing, β.

To determine the association rate of the blocker, we analyzed single-channel recordings of ENaC activity obtained from excised outside-out patches. We first used ENaC formed by α and β subunits only, which are almost equally sensitive to inhibition by amiloride or by benzamil compared with αβγ ENaC (Fig. 2A). The advantage of using αβ ENaC is that this channel is almost constantly open; therefore, blocking events can be readily distinguished from the fast channel closings (McNicholas and Canessa, 1997; Fyfe and Canessa, 1998). Figure 2B shows single-channel traces of αβ EnaC, illustrating that channel closings are rare and extremely short (uppermost trace in Fig. 2B). Addition of benzamil to the extracellular solution induced transitions to the zero current level of longer duration than the short closures, indicating that they represent benzamil-induced blocking events. The dwell time of the channel open state shortens with increasing concentration of benzamil, as predicted by eq. 2, and depends at high drug concentrations essentially on the drug association rate constant k_on. This is shown qualitatively by the representative current traces in Fig. 2B and quantitatively in Fig. 2C, where the reciprocal of the open state lifetime is plotted against the benzamil concentration. The k_on for benzamil can be calculated according to eq. 2 and corresponds to the slope of 1/τ_open in Fig. 2C as 57 ± 3 μM⁻¹s⁻¹. The corresponding value for the amiloride k_on was 68 ± 15 μM⁻¹s⁻¹ (Table 1), indicating that the association rate of both ligands to the receptor on αβ ENaC is basically the same.

As shown in eq. 3, the dissociation rate constant for amiloride/benzamil corresponds to the reciprocal of the channel blocked-state lifetime. According to eq. 1, the nonconducting states of the channel represent either closed channels or channels blocked by the drug.

The closed time histogram in the absence of benzamil shows a main component with a time constant of 4 ms (Fig. 3A). In the presence of 10 nM benzamil, longer nonconducting states occur (Fig. 3B) that are clearly different from the short closed states of the channel and therefore represent blocking events. The closed time histogram shows the presence of this new component of closed times that represents the blocking events with a time constant τ_B of 244 ± 82 ms. As mentioned previously, the time constant τ_B corresponds to the average residency time of benzamil on its binding site in the channel pore. The reciprocal of τ_B corresponds to the benzamil k_off (eq. 3) and is independent of the benzamil concentration as shown in Fig. 3C. The equilibrium dissociation rate constant can be calculated from the microscopic k_off and k_on values: The value obtained matches very well the inhibitory constant IC₅₀ determined from current inhibition curves in Fig. 2A (both 0.07 μM). To further confirm that ENaC block is the same in whole oocytes and in excised patches, we have determined benzamil block in single-channel patches by plotting the channel open probability P_o versus the benzamil concentration (Fig. 2A, right). Both approaches yielded the same concentration dependence for benzamil block.

Block of wt and Mutant αβγ ENaC. The gating of αβγ ENaC typically consists of spontaneous transitions between long open and closed states as illustrated by the traces in Fig. 4A. Typical open state and closed state dwell times are of the order of hundreds of milliseconds to seconds (Garty and Palmer, 1997; Kellenberger et al., 2002). This slow gating mode complicates the kinetic analysis of the ENaC block by amiloride and benzamil because blocking events by amiloride are of about the same dwell time duration as channel closed states. In Fig. 4A, a single-channel trace of αβγ ENaC in the presence of 1 μM amiloride shows short channel openings separated by transitions to nonconducting states, which could be either closings or blocking events by amiloride. Thus, according to eq. 1 the rate constants for channel opening (β) and for amiloride dissociation (k_off) cannot be distinguished and therefore not precisely determined.

Equation 2 predicts that in the presence of amiloride, the reciprocal of the open-state lifetime should be linearly related to the concentration of the blocker as already shown for benzamil block of αβ ENaC (Fig. 2B). This is confirmed for amiloride block of αβγ ENaC in Fig. 4B, which plots the relationship between the reciprocal of the open-state lifetime and the amiloride concentration. The k_on values for amiloride and benzamil determined from such experiments are 80 ± 8 and 68 ± 8 μM⁻¹s⁻¹ (Table 1) and are not significantly different from the values obtained in the αβ ENaC channel. These experiments confirm that the higher affinity of benzamil relative to amiloride for ENaC is not caused by differences in their respective k_on. The measured association rate for amiloride to αβγ ENaC is similar to that estimated from noise analysis of cloned αβγ ENaC (80 μM⁻¹s⁻¹; Segal et al., 2002) and higher than the association rate estimated by the same technique for block of Na⁺ current in frog skin (13 μM⁻¹s⁻¹; Li et al., 1985).

We then asked whether the lower affinity of the αS583C and βG525C ENaC mutants to amiloride and benzamil (Fig. 1) could have been caused by a lower association rate constant of the blockers to their receptor. The αS583Cβγ ENaC in the presence of 1 μM amiloride shows bursts of short openings and closings separated by longer closed times (Fig. 4A). For comparison, the wt αβγ ENaC shows similar short openings in the presence of 1 μM amiloride (compare Fig. 4A and Table 1); however, the closed/blocked dwell times are shorter in the αS583C mutant. The αβG525Cγ mutant in the presence of 1 μM benzamil displays open times that are substantially longer than those of ENaC wt with 1 μM benzamil, indicating that this mutation slows the blocker on-rate to some extent. With 10 μM benzamil, however, the open times are further reduced (Fig. 4A). The quantitative analysis of channel open times (Table 1) shows that the drug on-rate is slightly increased in ENaC containing the mutation αS583C with regard to ENaC wt, whereas it is ∼4 times decreased in βG525C mutant channels. The determination of the amiloride and/or benzamil association rates with low-affinity ENaC mutants indicates that the small changes in the blocker on-rate that we observe cannot account for the 30- to 1,000-fold decrease in the mutant channel affinity for the blocker.

As mentioned previously, the determination of the apparent dissociation rate constants of amiloride and benzamil from single channel analysis of ENaC wt are problematic because channel blocking events are undistinguishable from channel closures. However, for the ENaC mutants the duration of the nonconducting states induced by amiloride or benzamil; i.e., the nonconductive states within bursts (compare traces obtained in the presence/absence of the blocker) are much shorter than in the wt ENaC (Fig. 4A). This suggests that the mutations αS583C and βG525C may increase the dissociation of the blocker from its receptor site, probably by destabilizing the binding interaction between the ligand and the receptor. We measured the duration of the nonconductive states within bursts of single channels in the αS583Cβγ and αβG525Cγ mutants. The analysis of the dwell time distribution of these nonconductive states yielded time constants of 26 ms for amiloride block of αS583Cβγ and 27 ± 2 ms for benzamil block of αβG525Cγ ENaC, corresponding to k_off values of 38 and 37 s⁻¹, respectively. Thus, these values are clearly increased over the k_off of benzamil from αβ ENaC (4 ± 1.4 s⁻¹; Fig. 3B).

To directly compare the mutants with the wt ENaC in αβγ channels, we used in addition an alternative approach to determine the dissociation rate of amiloride and benzamil from ENaC, which consisted in measuring the time course of Li⁺ current recovery after rapid removal of the ENaC blocker. Excised outside-out macropatches were obtained from oocytes expressing ENaC wt or mutants, and placed in the solution stream of a rapid solution exchanger device. This setting allowed a rapid and complete washout of benzamil or amiloride in the external solution. The speed of the perfusion change was monitored for each experiment by measuring the appearance of an inward current after the rapid substitution of the external nonconducting K⁺ ion by conducting Li⁺ ions. On average, this yielded a time constant of 17 ± 4 ms for the exchange of the external solution. Current traces in macropatches of αβγ ENaC wt and mutant, generated by the rapid washout of amiloride or benzamil from the bath are shown in Fig. 5. With wt ENaC macropatches, the appearance of the inward Li⁺ current after washout of amiloride follows a single exponential, and the rate of appearance determined by the best fit of the tracing gave a value of 2.1 ± 0.4 s⁻¹ (τ ≅ 0.5 s; Table 2). The rate of current appearance corresponds to the dissociation of bound amiloride, and this dissociation rate is close to the dissociation rates estimated by noise analysis from cloned ENaC and from Na⁺ current in frog skin, which were 9 and 4 s⁻¹ (Li et al., 1985; Segal et al., 2002). The rate of appearance of the Li⁺ current after rapid removal of external benzamil was slower with a rate constant of 0.11 ± 0.03 s^−1. This is consistent with the lower dissociation rate constant for benzamil compared with amiloride that we have measured from single channel activity of αβ ENaC. This lower dissociation rate can account for the higher affinity for ENaC of benzamil relative to amiloride. The recovery of the inward Li⁺ current after amiloride washout was extremely fast in αS583Cβγ and αβG525Cγ ENaC mutants compared with the wt. Under these conditions, the speed of the fluid exchange becomes limiting for the determination of the rate constant of current recovery. Therefore, we have analyzed in the ENaC mutants the time course of Li⁺ current recovery from benzamil inhibition because of its slower time course compared with amiloride. Figure 5 clearly shows that in the αS583Cβγ and αβG525Cγ mutants the current recovery from benzamil inhibition was much faster than for the wild type. In Fig. 5, the insets displaying the recording in a 10-fold expanded time scale show that current recovery is faster for αβG525Cγ than for αS583Cβγ. The dissociation rate constants k_off obtained from this analysis are comparable with those determined from single-channel kinetics. They are 20- and 200-fold increased in the mutants αS583Cβγ and αβG525Cγ relative to ENaC wt (Table 2).

We can conclude from these experiments that the increase in the dissociation rate can account for most of the differences in benzamil affinity between the ENaC wt and the ENaC αS583Cβγ and αβG525C mutants. The k_off of αS583Cβγ ENaC is increased by a factor of 10 and 24 for amiloride and benzamil, respectively, explaining the 10- to 20-fold shift in macroscopic inhibition curves. The βG525C mutation induces a 187-fold increase in k_off and a 4-fold decrease in the k_on for benzamil. The resulting calculated 750-fold shift in IC₅₀ is close to the ∼1,000-fold shift observed.

Discussion

We have investigated the effects of ENaC αS583, βG525, and γG537 mutations on the kinetics of channel block by amiloride and benzamil. These two high-affinity blockers differ in their affinity to ENaC. We show here that this is because of their different dissociation rates from the receptor in the ENaC pore. Mutations of αS583, βG525, and γG537 decrease the affinity of ENaC for amiloride and benzamil essentially by increasing the dissociation rate of the ligands, with little changes in their association rate to the receptor. In other words, these mutations destabilize the binding interaction between the blockers and the receptor on the channel, favoring the unbinding of the ligand. These effects on the ligand-receptor interactions indicate that residues αS583, βG525, and γG537 are part of the receptor for amiloride and benzamil located in the external entry of the channel pore. Because mutations of αS583, βG525, and γG537 have similar effects on amiloride, benzamil, and triamterene block, we conclude that these three ENaC blockers share a common receptor.

The Amiloride Binding Site. Early experiments on frog skin showed that amiloride block is voltage-dependent, suggesting that the binding site on ENaC is within the transmembrane electrical field. These experiments also provided evidence for an interaction between extracellular Na⁺ and amiloride block (Palmer and Andersen, 1989; Garty and Palmer, 1997). Based on these observations, it was suggested that amiloride may block the channel by a simple physical occlusion of the pore.

This study together with other recent molecular studies places the amiloride binding site in the extracellular pore region. First, residues αS583, βG525, and γG537 are located only four amino acid residues away, in the N-terminal direction, from the selectivity filter sequence G/SxS in the ion permeation pathway (Kellenberger et al., 1999a,b). Second, evidence that residues αS583, βG525, and γG537 are extracellular comes from the observation that substitution by cysteine and subsequent modification by hydrophilic sulfhydryl reagents results in channel block (Schild et al., 1997; Kosari et al., 1998; Snyder et al., 1999). The αS583C mutant is blocked by Zn²⁺ with an affinity in the micromolar range (Firsov et al., 1998). This high-affinity block requires the αS583C mutations on the two αENaC subunits of the heteromultimeric channel. If the αENaC subunits face each other across the channel pore as suggested by a study using concatameric constructs (Firsov et al., 1998), this would imply that the optimal distance for coordinated ligation of Zn²⁺ of ∼5 Å (Krovetz et al., 1997) would correspond to the distance between the two αS583 residues across the pore.

The observation that substituting cysteine residues at the amiloride binding site can form a high-affinity Zn²⁺ binding site and that they can be modified by sulfhydryl reagents indicates that their side chains are oriented toward the lumen of the pore entry. Our finding that very conservative mutations of these residues affect amiloride block further supports this conclusion.

After the identification of the critical role of the homologous αS583, βG525, and γG537 in amiloride binding, additional mutations were made in this region of the protein in the search of other residues that contribute to the amiloride binding site. In the α subunit, the mutation αW582L increased the IC₅₀ by a factor of ∼30 (Schild et al., 1997). Several mutations in the pore region were shown to increase the IC₅₀ by 2- to 4-fold (Sheng et al., 2000). Other mutations may affect indirectly channel block by amiloride. Mutation of αS588 to Iso but not Ala shifted the IC₅₀ to ∼20-fold higher concentrations, and some mutations of βG529 also reduced the sensitivity to amiloride block (Waldmann et al., 1995b; Kellenberger et al., 1999b). Thus, residues near αS583 and βG525 probably participate directly or indirectly in amiloride block but to a much lesser extent than αS583, βG525, and γG537.

Amiloride Block in Other ENaC/DEG Family Members. The amino acid sequence constituting the amiloride binding site is highly conserved among the members of the ENaC/degenerin channel family. However, ENaC/degenerin channels differ in their respective sensitivity to block by amiloride, with ENaC having the highest affinity. Interestingly, ENaC differs from the other channels of this family by a Ser residue at position αS583, whereas other ENaC/degenerin channels such as acid-sensing ion channels and degenerins carry a Gly residue. This Ser-Gly-Ser-Gly arrangement is unique for ENaC and might in part account for its high affinity for amiloride because the αS583G mutation decreases its affinity (Fig. 1D). However, the heteromeric and functional δβγ ENaC that also carries a Ser residue at the position homologous to α583 in the δ subunits, has a lower affinity for amiloride than αβγ ENaC (Waldmann et al., 1995a). Inversely, degenerin channels formed by MEC-4 and MEC-10, which both have a glycine at the homologous position, display an IC₅₀ for amiloride of 0.4 μM, thus very close to that of ENaC (Goodman et al., 2002). Differences between ENaC/DEG channels in the sequence near the amiloride binding site may account for their different pharmacological profile. For example, X. laevis ∈βγ ENaC shows a 10-fold higher IC₅₀ than X. laevis αβγ ENaC because of a Leu-to-Trp substitution at the position corresponding to residue 584 in rat αENaC (Babini et al., 2003). Interaction between amiloride and ENaC for high-affinity block is complex and may involve other, still unidentified residues in the extracellular loop of ENaC/DEG channels.

Structure-activity studies of different amiloride analogs on wild-type and mutant ENaC will certainly refine our understanding of the binding interactions between blockers and the amiloride receptor. This will provide the basis for the future understanding of the structure of the amiloride binding site and the external pore of ENaC. We hope that such studies will help the design and the identification of molecules able to block other members of the ENaC/DEG channel family such as acid-sensing ion channels for which high-affinity blockers are still needed.