Paradoxical Stimulation of a DEG/ENaC Channel by Amiloride*

Extracellular amiloride inhibits all known DEG/ENaC ion channels, including BNC1, a proton-activated human neuronal cation channel. Earlier studies showed that protons cause a conformational change that activates BNC1 and exposes residue 430 to the extracellular solution. Here we demonstrate that, in addition to blocking BNC1, amiloride also exposes residue 430. This result suggested that, like protons, amiloride might be capable of activating the channel. To test this hypothesis, we introduced a mutation in the BNC1 pore that reduces amiloride block, and found that amiloride stimulated these channels. Amiloride inhibition was voltage-dependent, suggesting block within the pore, whereas stimulation was not, suggesting binding to an extracellular site. These data show that amiloride can have two distinct effects on BNC1, and they suggest two different interaction sites. The results suggest that extracellular amiloride binding may have a stimulatory effect similar to that of protons in BNC1 or extracellular ligands in other DEG/ENaC channels.

A characteristic shared by DEG/ENaC cation channels is that extracellular amiloride inhibits their current. The ability of amiloride to inhibit the human epithelial Na ϩ channel (ENaC) makes it clinically useful as a diuretic (1). Amiloridesensitivity is also a valuable marker of currents generated by other DEG/ENaC channels from vertebrates, insects, mollusks, and nematodes (2,3). However, the mechanism of amiloride's action is not understood. For example, some data suggest that amiloride blocks current by occluding the ion-conducting pore (4), whereas other evidence suggests that amiloride binds a predicted extracellular domain of ENaC (5,6).
The relationship between amiloride block and channel activity is interesting. For example, the Drosophila Na ϩ channel Ripped Pocket (RPK) generates a minimal Na ϩ current that is blocked by relatively high concentrations of amiloride (7). However, when RPK contains a "Deg" mutation that increases Na ϩ current 50-fold, current is blocked by 20 times lower concentrations of amiloride. In this study we further examined the relationship between amiloride and channel activity. The DEG/ ENaC channel that we studied was BNC1 (also called MDEG and BNaC1), a proton-gated channel from human neurons that may be involved in synaptic transmission and the sensation of pain and sour taste (8 -12). Under basal conditions at neutral extracellular pH, BNC1 generates only a very small whole cell Na ϩ current (8). However, lowering extracellular pH activates a large transient Na ϩ current (11).
Like RPK, BNC1 is activated by a Deg mutation, which alters a critical glycine residue at position 430 (9). Residue 430 is a site where analogous mutations in some Caenorhabditis elegans family members cause neurodegeneration (13). In BNC1, mutation of Gly 430 to valine or larger amino acids locks the channel in an activated state, even at neutral pH (9). In contrast, mutation of Gly 430 to cysteine (BNC1-G430C), generates only small constitutive currents (14), and confers a unique sensitivity to sulfhydryl-reactive methanethiosulfonate (MTS) 1 compounds (15). The ability of MTS compounds to modify Cys 430 is state-dependent. Under basal conditions at extracellular pH 7.4 with the channel closed, Cys 430 is largely protected from extracellular MTS compounds. However when the channel is activated by protons, Cys 430 becomes accessible to extracellular MTS compounds, which covalently modify Cys 430 to make its side chain larger. As a result, the channel is irreversibly activated because it is unable to return to the inactive state.
In the course of investigating the role of residue 430 in channel activation and in inhibition by amiloride, we made the surprising discovery that under certain conditions amiloride could activate current. Here we describe our investigation of the responsible mechanisms.

EXPERIMENTAL PROCEDURES
cDNA Constructs-BNC1 mutants were constructed by singlestranded mutagenesis of BNC1 (8) in pBluescript. The validity of constructs was confirmed by DNA sequencing. Constructs were cloned into pMT3 (16) for expression.
Expression and Electrophysiological Analysis in Xenopus Oocytes-cDNA constructs were expressed in defolliculated albino Xenopus laevis oocytes (Nasco, Fort Atkinson, WI) by nuclear injection of plasmid DNA. cDNA encoding BNC1-G430C was injected at 5 ng/l. cDNA encoding BNC1-G430V/G437C was injected at 40 ng/l. For coexpression of the double mutant BNC1-G430V/G437C with BNC1-G430V, we injected 10 and 2 ng/l, respectively. For coexpression of BNC1-G430V/G437C with wild-type BNC1, we injected 25 and 5 ng/l, respectively. Following injection, oocytes were incubated at 18°C in modified Barth's solution. Generally, cells were studied 12-24 h after injection; however, cells expressing BNC1-G430V/G437C (with or without wild-type BNC1) were studied 36 -48 h after injection. Whole cell currents were measured using a two-electrode voltage clamp. During recording, oocytes were bathed in frog Ringer's solution (116 mM NaCl, 0.4 mM CaCl 2 , 1 mM MgCl 2 , 5 mM HEPES, pH 7.4). MTS compounds were from Toronto Research Biochemicals (Toronto, Ontario) and included cationic reagents, 2-aminoethyl methanethiosulfonate bromide (MTSEA) and 2-(trimethylammonium)ethyl methanethiosulfonate bromide (MTSET), and an anionic reagent, sodium 2-sulfonatoethyl methanethiosulfonate (MTSES). * This work was supported by the Howard Hughes Medical Institute. 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

RESULTS
Zn 2ϩ Reversibly Activates BNC1-G430C-In earlier work (15) we showed that when Gly 430 was mutated to cysteine, covalent modification of residue 430 by MTS reagents irreversibly locked BNC1-G430C in an activated state. Therefore we hypothesized that a noncovalent interaction between Zn 2ϩ and Cys 430 might also activate the channel, but in a reversible way. Fig. 1A shows that extracellular Zn 2ϩ had no effect on current in oocytes expressing wild-type BNC1. However, at neutral extracellular pH, Zn 2ϩ reversibly activated current in oocytes expressing BNC1-G430C (Fig. 1B). The EC 50 for activation of BNC1-G430C by Zn 2ϩ was approximately 300 M (Fig. 1C).
Previous studies (15) showed that pH-dependent activation of BNC1-G430C allowed MTS compounds to modify Cys 430 . Fig. 1D shows that activation by Zn 2ϩ also allowed the MTS compound, MTSES, to modify Cys 430 and irreversibly activate the channel. How might both Zn 2ϩ and MTSES interact with Cys 430 ? Because the BNC1 channel is a homomultimer composed of multiple BNC1 subunits (17), it is likely that within one channel there are multiple cysteines with which these agents may interact. Taken together, these data indicate that Zn 2ϩ binds to Cys 430 to reversibly activate BNC1-G430C.
Amiloride Increases the Accessibility of Cys 430 to Zn 2ϩ and MTS Reagents- Fig. 2A shows that 100 M Zn 2ϩ activated BNC1-G430C and that current was blocked by 1 mM amiloride. However, as shown in Fig. 2, B and C, Zn-activated currents were not smaller but larger if a submaximal concentration of amiloride (100 M) was present. In the absence of Zn 2ϩ , amiloride either had no effect (Fig. 2B) or blocked a small amount of current (Fig. 2C), depending on the amount of basal current generated by BNC1-G430C. The combined effects of amiloride and Zn 2ϩ concentrations are shown in Fig. 2D; at low concentrations, amiloride enhanced Zn 2ϩ stimulation of current, and amiloride blocked at high concentrations. Enhanced stimulation in the presence of intermediate amiloride concentrations may also explain the transient current stimulation observed on washing out high concentrations of amiloride and Zn 2ϩ (Fig. 2A).
These data show that amiloride alone does not stimulate current, but when both amiloride (intermediate concentrations) and Zn 2ϩ are present, current is greater than with Zn 2ϩ alone. One interpretation of these results is that amiloride binds to the channel, and in so doing makes Cys 430 more accessible for interaction with Zn 2ϩ . In this respect, amiloride might be similar to protons; a decrease in extracellular pH causes a conformational change that exposes Cys 430 to the extracellular solution where it can be irreversibly modified by MTS reagents (15).
Therefore, we hypothesized that amiloride causes a conformational change that exposes Cys 430 to the extracellular solution. To test this hypothesis, we asked if amiloride would allow MTS reagents to irreversibly modify and activate the channel at pH 7.4. Fig. 3A shows that in the absence of amiloride, MTSEA had little or no effect at pH 7.4. Current remained low when amiloride and MTSEA were simultaneously present in the extracellular solution. However, once amiloride and MTSEA were removed, it was apparent that a large current had been activated; this current was blocked by subsequent treatment with amiloride. We obtained similar results with two other MTS reagents, MTSET and MTSES (Fig. 3, B and C).
These results indicate that in the presence of amiloride, MTS reagents irreversibly modified Cys 430 , locking the channel open. Because the activated channel remained sensitive to amiloride block, current was only apparent after amiloride was removed from the extracellular solution.
The G437C Mutation Reduces Sensitivity to Amiloride Block-Extracellular amiloride and protons shared the ability to expose Cys 430 to MTS compounds. However, an important difference between the two interventions was that protons activated current (15), whereas amiloride alone did not. We hypothesized that amiloride might have two effects; like protons, amiloride might bind to one site to stimulate current, but in addition, amiloride might bind to a second site to block the channel. Thus, any stimulatory effect would be obscured by the We previously showed that MTSES alone has no effect on BNC1-G430C when extracellular pH is 7.4 (see Ref. 15 and Fig. 3C). In all figures, the 0 indicates zero current and membrane voltage was Ϫ60 mV. ability of amiloride to simultaneously block the channel. To test this hypothesis, we asked whether the inhibitory effect of amiloride could be selectively abolished by mutation. We tested a mutation equivalent to a mutation that reduces amiloride block of ENaC: substitution of cysteine for Gly 437 (Fig. 4A). Residue 437 lies within the predicted pore-forming second membrane-spanning domain and is highly conserved among DEG/ENaC channels (9). In ENaC subunits, mutations equivalent to G437C had several effects; if present in the ␤ or ␥ subunits of the channel, the mutations reduced amiloride block and Na ϩ conductance. Also, the introduced cysteine could be modified by MTSET to inhibit current (17,18). However, if present in all ENaC subunits, the mutations abolished channel activity (18).
To determine whether the G437C mutation reduced amiloride inhibition, we studied it in the context of BNC1-G430V, which generates large constitutive currents (9, 15). BNC1-G430V generated large currents (904 Ϯ 84 nA; n ϭ 9) that were blocked by amiloride (Fig. 4B). Expression of the double mutant BNC1-G430V/G437C failed to generate current; this was consistent with the observation that mutations abolish ENaC function if present in all subunits of that channel (18). Therefore, we coexpressed BNC1-G430V/G437C with a lesser amount of BNC1-G430V (5:1 ratio). This combination generated large constitutive currents (773 Ϯ 63 nA; n ϭ 10; Fig. 4C) like BNC1-G430V. However in contrast to BNC1-G430V, 3 mM amiloride blocked only 38 Ϯ 4% of the current (n ϭ 6). MTSEA blocked a larger portion of the current (69 Ϯ 3%; n ϭ 6, Fig. 4C). MTSEA block was likely due to covalent modification of Cys 437 , because block was irreversible, and MTSEA did not block channels that lacked the G437C mutation (wild-type BNC1 or BNC1-G430V; not shown). These data indicate that when subunits containing the G437C mutation are included in a BNC1 channel complex, the sensitivity to block by amiloride is reduced.
A hint of amiloride stimulation came from examining the response to increasing amiloride concentrations (Fig. 4C, left part of trace). Amiloride produced a biphasic response, with inhibition at 100 M, an increase in current at 1 mM, followed by inhibition at 3 mM. However, because the G430V Deg mutation was present in every subunit, a stimulatory effect of amiloride would be minimized. Amiloride Stimulates Current in Cells Expressing BNC1-G430V/G437C with Wild-type BNC1-To test whether amiloride might stimulate BNC1-G430V/G437C, we coexpressed it with a small amount of wild-type BNC1 (5:1 ratio). This combination generated no current on its own; basal current (133 Ϯ 13 nA; n ϭ 10) was not different from that in uninjected cells (not shown). However, addition of amiloride stimulated current (Fig. 4D). In contrast, amiloride did not stimulate wild-type BNC1 (Fig. 4E). Simulation was reversible and evident only at concentrations greater than 100 M. 3 mM amiloride increased current by 279 Ϯ 40 nA (n ϭ 10). This result, coupled with the ability of amiloride to increase access of residue 430, indicated that amiloride, like protons, could stimulate BNC1.
Amiloride May Interact with Two Different Sites in BNC1-Because amiloride had two effects (stimulation and inhibition), we hypothesized it might interact with two separate sites in BNC1. To test this idea, we determined the effect of voltage on inhibition and stimulation. Because amiloride is positively charged, effects should be voltage-dependent if the binding site lies within the membrane electrical field. To test inhibition, we studied BNC1-G430V. Block was steeply voltage-dependent, suggesting that amiloride blocked within the membrane electrical field (Fig. 5, A-D). Fig. 5, B and C, shows that block depended on the reversal potential of BNC1 current, suggesting that outward flow of cations prevented amiloride from reaching a blocking site. These results suggest that when amiloride blocks, it binds within the channel pore.
To test stimulation, we studied oocytes expressing BNC1-G430V/G437C and wild-type BNC1 (5:1 ratio). Fig. 5, E and F, shows that stimulation was not appreciably voltage-dependent. This result suggested that amiloride stimulated the channel by binding a site outside the membrane electrical field, different from the site of amiloride block. DISCUSSION Our data show that amiloride can have two distinct effects on the BNC1 channel and suggest that there may be two different sites of interaction. One site appears to lie within the channel pore; when amiloride interacts with this site it blocks the channel. Three pieces of evidence suggest that amiloride blocks within the channel pore: the voltage dependence suggested block within the membrane electrical field; alteration of the reversal potential and the outward flow of cations reduced block by extracellular amiloride; and block was attenuated by mutation of residue 437, which other data suggest lies within the second membrane-spanning helix (19,20).
The other site appears to be different from the blocking site; when amiloride interacts there it can cause a conformational change that activates the channel and exposes residue 430 to extracellular MTS reagents. Our data make no predictions about the location of the stimulatory site, except that it is not likely deep within the pore because the transmembrane electrical field has little effect on amiloride-dependent stimulation.
We considered the possibility that amiloride-dependent stimulation might represent the interaction of amiloride at a single site producing open channel block. However, this mechanism does not explain stimulation because net current increased in the presence of amiloride, and amiloride stimulated current in channels that were largely insensitive to amiloride block. Moreover, membrane voltage had an effect on block, but not on stimulation. We also considered the possibility that amiloride might interact with a single binding site that might move during channel gating to give both voltage-dependent and -independent effects. This also seems unlikely, because the G437C mutation selectively attenuated block. Nevertheless, we cannot exclude an interaction at a single site that somehow both blocks and activates the channel; the multimeric nature of BNC1 might allow for such an effect.
Our data suggesting that amiloride may interact with two sites may help explain earlier studies of other DEG/ENaC channels. Studies of ENaC suggest that amiloride might block the pore. First, amiloride block is voltage-dependent (4). Second, amiloride is a more potent blocker in reduced extracellular Na ϩ , suggesting that amiloride and Na ϩ may compete for a binding site in the channel pore (4). Third, mutations that alter ion conduction or selectivity also alter amiloride block (4,18,20,21). On the other hand, studies of ENaC identified residues in the predicted extracellular domain that bind an amiloride anti-idiotypic antibody; mutation of those sites altered amiloride sensitivity (5). In addition, nonfunctional ENaC splice variants that lack the pore-forming second transmembrane domain but retain much of the extracellular domain retain binding to phenamil (an amiloride derivative) with high specificity (6). These findings are consistent with an extracellular amiloride-binding site on ENaC. The suggestion of two interaction sites in BNC1 raises the question of whether amiloride will have two effects on all DEG/ENaC family members. Alternatively, some family members may lack one or both sites and may not be inhibited by amiloride.
The large extracellular domain may regulate the function of DEG/ENaC channels. In the FMRFamide-activated Na ϩ channel (FaNaCh), this region may form a ligand-binding site for channel activation (22). Consistent with this possibility, mutations in the predicted extracellular domain reduce sensitivity to FMRFamide. 2 In BNC1, ASIC, and DRASIC, this region may be responsible for proton activation (2). In C. elegans DEG/ ENaC proteins, genetic studies suggest that this region may interact with extracellular collagen to affect channel gating (23,24). Our data suggest that there is also an amiloridebinding site in BNC1, perhaps located in the extracellular domain, that can modulate channel activity. Future studies that identify this site may help our understanding of the function of the extracellular domain and could possibly identify agents that activate or inhibit the channels for therapeutic purposes.
FIG. 5. Voltage dependence of amiloride block and amiloride stimulation. Panels A-D, amiloride block of BNC1-G430V. Panel A, families of BNC1-G430V currents at voltages ranging from Ϫ120 to ϩ60 mV. Holding voltage (Vh) was Ϫ20 mV. Amiloride concentration is indicated. Panel B, current-voltage relationship from an oocyte expressing large BNC1-G430V currents. For several minutes before current measurements were made, holding voltage was clamped to 0, Ϫ40, or Ϫ80 mV, as indicated. When holding voltage was Ϫ40 or Ϫ80 mV, intracellular Na ϩ accumulated and the reversal potential became more negative (27). In oocytes expressing ENaC, a similar increase in intracellular [Na ϩ ] occurs (26). Once the reversal potential no longer drifted, current was measured during brief 1 s voltage steps in the absence (open symbols) or presence (closed symbols) of 30 M amiloride. Panel C, current values from panel B normalized to the reversal potential in each condition. Panel D, current-voltage relationship of amiloride-inhibited current. Amiloride concentration was 1 M (squares), 10 M (inverted triangles), 100 M (circles), or 1 mM (triangles). Data are average Ϯ S.E. from six oocytes where Vh was Ϫ20 mV. Panels E and F, amiloride stimulation of current in oocytes expressing BNC1-G430V/G437C with wild-type BNC1. Panel E, current-voltage relationship of amiloridestimulated current. Amiloride concentration was 1 mM (f) or 3 mM (E). Data are average Ϯ S.E. from three oocytes. Panel F, families of currents at voltages ranging from Ϫ120 mV to ϩ40 mV. Amiloride concentration is indicated.
preciate the discussions and help of Drs. Margaret Price, Joseph Cotten, and John Rogers. We thank the University of Iowa DNA Core Facility for assistance with sequencing and oligonucleotide synthesis.