Heteromeric Assembly of Acid-sensitive Ion Channel and Epithelial Sodium Channel Subunits*

Amiloride-sensitive ion channels are formed from homo- or heteromeric combinations of subunits from the epithelial Na+ channel (ENaC)/degenerin superfamily, which also includes the acid-sensitive ion channel (ASIC) family. These channel subunits share sequence homology and topology. In this study, we have demonstrated, using confocal fluorescence resonance energy transfer microscopy and co-immunoprecipitation, that ASIC and ENaC subunits are capable of forming cross-clade intermolecular interactions. We have also shown that combinations of ASIC1 with ENaC subunits exhibit novel electrophysiological characteristics compared with ASIC1 alone. The results of this study suggest that heteromeric complexes of ASIC and ENaC subunits may underlie the diversity of amiloride-sensitive cation conductances observed in a wide variety of tissues and cell types where co-expression of ASIC and ENaC subunits has been observed.

Amiloride-sensitive ion channels are formed from homo-or heteromeric combinations of subunits from the epithelial Na ؉ channel (ENaC)/degenerin superfamily, which also includes the acid-sensitive ion channel (ASIC) family. These channel subunits share sequence homology and topology. In this study, we have demonstrated, using confocal fluorescence resonance energy transfer microscopy and co-immunoprecipitation, that ASIC and ENaC subunits are capable of forming cross-clade intermolecular interactions. We have also shown that combinations of ASIC1 with ENaC subunits exhibit novel electrophysiological characteristics compared with ASIC1 alone. The results of this study suggest that heteromeric complexes of ASIC and ENaC subunits may underlie the diversity of amiloride-sensitive cation conductances observed in a wide variety of tissues and cell types where co-expression of ASIC and ENaC subunits has been observed.
The electrophysiological characteristics of cells expressing amiloride-sensitive cation conductances vary widely, and regulation of ENaCs and ASICs occurs through several mechanisms. Pharmacological inhibitors of channel function include small molecules, such as amiloride, or peptide toxins, such as psalmotoxin I (35) and APETx2 (36). Post-translational modifications, including proteolytic cleavage (37) or phosphorylation (38,39), also regulate these channels. Finally, the subunit composition of the channel has been shown to play an important role in defining the electrophysiological phenotype of the channel.
The co-expression of ASIC and ENaC subunits within various tissues, the sequence homology within this superfamily of channel subunits, and the propensity of these subunits to form heteromeric ion channels all suggest that ASIC and ENaC subunits may form cross-clade heteromeric ion channels. Previous studies in glioblastoma cells, which express an amiloride-sensitive cation current that has been linked to glioma cell motility and proliferation, have demonstrated that ASIC1 and ASIC2 can co-precipitate with ␥ENaC (29,(51)(52)(53).
In order to determine whether other combinations of ASIC and ENaC subunits can intermix to form heteromeric complexes, we have co-expressed human ASIC1b or ASIC2b with human ENaC subunits in CHO K1 cells that express neither amiloride-sensitive nor acid-activated currents. Intermolecular associations between co-expressed ion channel subunits have been demonstrated by confocal FRET microscopy and by coimmunoprecipitation. The electrophysiological properties of channels comprised of ASIC1 and ENaC subunits have been determined by two-electrode voltage clamp of cRNA-injected Xenopus laevis oocytes. We present evidence that ASIC and ENaC subunits can intermix promiscuously and that ion channel properties, including cation selectivity and pharmacological inhibition, are affected by the subunit composition of the expressed channels.
For imaging and electrophysiological recordings, cells were split into either 35-mm dishes or 6-well tissue culture plates with flame-sterilized glass coverslips and transiently transfected using 0.5 g of plasmid DNA and 1.5 l of Lipofectamine 2000 reagent (Invitrogen) for each construct, according to the manufacturer's supplied protocols. For immunoprecipitations, transfections were performed in 100-mm dishes with 3 g of DNA and 9 l of Lipofectamine 2000 for each construct.
Whole Cell Patch Clamp-Amiloride-sensitive ENaC whole cell currents were recorded in transfected and nontransfected CHO K1 cells. Cells were cultured in 35-mm dishes with flamesterilized glass coverslips and patch-clamped in serum-free RPMI medium on the coverslip. Alternatively, cells were cultured without coverslips, scraped into RPMI, and allowed to settle onto the glass-bottomed perfusion chamber before patch clamping. Micropipettes were prepared using a Narashigi PP-83 two-stage micropipette puller and were filled with an electrolyte solution (100 mM potassium gluconate, 30 mM KCl, 10 mM NaCl, 20 mM HEPES, 0.5 mM EGTA, 4 mM ATP, pH 7.2). Pipette electrical resistance was 5-10 megaohms. The whole cell patch configuration was established as described previously (52). ENaC currents were recorded by holding the cells at 0 mV and stepping from Ϫ100 to ϩ100 mV in 20-mV increments. Amiloride sensitivity was determined by superfusion with 10 M amiloride in RPMI.
ASIC currents were recorded in ASIC1-or ASIC2-transfected cells by holding the membrane voltage at Ϫ60 mV and perfusing the cells with Krebs buffer (130 mM NaCl, 2.4 mM K 2 HPO 4 , 1 mM CaSO 4 , 1 mM MgSO 4 , 10 mM D-glucose, and 10 mM HEPES, pH 7.4). Acid-induced currents were evoked by exchanging the perfusing buffer with Krebs, pH 5.0, in which HEPES was replaced with MES, and pH was adjusted with HCl.
FRET Microscopy-Transfected cells grown on glass coverslips were fixed in 2% paraformaldehyde (Tousimis, Rockville, MD) for 10 min and mounted on glass microscope slides using 0.1% p-phenylenediamine (Sigma) in 9:1 glycerol/phosphatebuffered saline. Cells were imaged on a Leica confocal SP2 microscope using an HCX 100ϫ plan APO 1.40 numerical aperture oil immersion objective. An argon laser was used for excitation of ECFP (458 nm) and EYFP (514 nm). A double dichroic mirror (458 nm/514 nm) was used to exclude both excitation beams from the imaging pathway. ECFP emission was 465-505 nm. EYFP emission was 525-600 nm. Images were obtained for both ECFP and EYFP emission before and after regional photobleaching of EYFP with high intensity 514-nm light. EYFP photobleaching was continued until the measured EYFP emission was less than 30% of the original intensity.
The observed FRET intensity was computed as FRET% ϭ (I post Ϫ I pre )/(I post ) ϫ 100, where I is the measured ECFP fluorescence intensity before or after regional photobleaching. For each cell, observed FRET was measured both inside and outside the photobleached region. Cells occasionally demonstrated an increase in ECFP intensity in the unbleached regions that appeared as an anomalous FRET signal in the nonbleached regions. This was often due to motion of the cell or changes in the fluorophore distribution during the photobleaching procedure. Cells that exhibited greater than 3% anomalous FRET in the nonbleached regions were excluded from the data set. Immunoprecipitation-CHO K1 cells were plated in 100-mm tissue culture dishes and transfected as described previously. Twenty-four hours post-transfection, the cells were lysed for 1 h in 1% Triton X-100 (Pierce) in phosphate-buffered saline with Complete protease inhibitor mixture (Roche Applied Science) at 4°C. Nonsoluble material was removed by centrifugation, and the supernatants were assayed for protein concentration by BCA assay (Pierce). Supernatants containing 250 g of soluble protein were incubated overnight at 4°C with 5 g of primary antibody (rabbit anti-ASIC1 or rabbit anti-ASIC2; Alomone, Jerusalem, Israel) and 25 l of protein A-agarose beads (Pierce). After incubation, immunopurified protein complexes were separated by centrifugation and were washed extensively with 0.1% Triton X-100 in phosphate-buffered saline.
In order to observe co-precipitation of fluorescent tagged ASIC and ENaC subunits with nontagged ASIC1 or ASIC2, protein complexes were imunoprecipitated with anti-ASIC1 or anti-ASIC2 on protein A-agarose beads as described above. Washed beads were resuspended in 50 l of p-phenylenediamine mounting medium and mounted on a glass microscope slide with a coverslip. Protein A beads were imaged with differential interference contrast optics. EYFP fluorescence was imaged on the TCS SP inverted confocal microscope with a ϫ60 plan APO 1.40 numerical aperture oil immersion objective. EYFP fluorescence was imaged with excitation at 488 nm and emission from 520 to 585 nm.
Two-electrode Voltage Clamp-Whole cell conductances of X. laevis oocytes were recorded by two-electrode voltage clamp. Oocytes were harvested from mature frogs and prepared as described previously (45). Oocytes were microinjected with a total of 0.25 ng of cRNA encoding ASIC subunits, ENaC subunits, or combinations of both. Oocytes were impaled with borosilicate glass electrodes (53432-921; VWR) pulled to a resistance of 0.5-2.0 megaohms on a Kopf 700D (Tujunga, CA) micropipette puller. Electrodes were back-filled with 3 M KCl. Whole cell currents were recorded with a holding potential of Ϫ60 mV using a Geneclamp 500 amplifier (Axon Instruments).
Oocytes were perfused with ND96 buffer (96 mM NaCl, 2 mM KCl, 1.8 mM CaCl 2 , 1 mM MgCl 2 , 5 mM HEPES, pH 7.4). Typical acid activated currents were induced by rapid switching, using a Warner-SF77B Perfusion Fast-step, to ND96, pH 5.0 (HEPES replaced with MES and pH adjusted with HCl). pH activation curves were recorded by sequentially switching from pH 7.4 to pH 7.0, 6.5, 6.0, 5.5, 5.0, and 4.0 ND96 buffers. Acid-induced currents were normalized to pH 4.0 peak currents for analysis. Na ϩ /K ϩ permeability ratios were measured as the ratio of pH 5.0 peak currents measured in ND96 and KD98 buffers (96 mM NaCl replaced with 96 mM KCl). Amiloride inhibition dose response of acid induced current was recorded by superfusing individual oocytes with increasing concentrations of amiloride (0, 5, 10, 50, 100, and 500 M) present in pH 5.0 ND96 buffer. P. cambridgei venom (PC venom; Spider Pharm, Yarnell, AZ) inhibition of acid-induced currents was recorded similarly, using increasing concentrations of venom (0, 0.005, 0.01, 0.02, 0.5, and 0.1 l venom/ml) diluted into ND96, pH 7.4. For both inhibition experiments, peak currents were normalized to the maximal current amplitude recorded prior to application of amiloride or PC venom. The results of multiple titration experiments were pooled, and inhibition curves were fitted to the Hill equation in Sigmaplot 9.0 (SyStat). IC 50 values of inhibition curves were reported with S.E. values derived from the curve fits.
Data Analysis and Statistics-Unless stated otherwise, all data are presented as mean Ϯ S.D. Statistical significance of FRET microscopy data was determined using a one-tailed Student's t test comparing each condition against the ECFP-, EYFP-cotransfected negative control. For electrophysiological experiments, statistical significance was computed using a twotailed Student's t test comparing each condition against the ASIC1 controls. For all experiments, significance was accepted at the p Ͻ 0.05 level.

RESULTS
In order to determine heteromultimeric associations between ASIC and ENaC subunits, combinations of fluorescent tagged subunits were co-expressed in CHO K1 cells. CHO K1 cells were selected for expression studies, because they exhibit neither amiloride-sensitive nor acid-inducible currents (supplemental Fig. S1A). CHO K1 cells have been used extensively for the study of ion channel function in general and ENaC/degenerin channels specifically (56,57).

Expression of Fluorescent Tagged ASIC and ENaC Subunits in CHO K1
Cells-The addition of C-or N-terminal fusion proteins may potentially disrupt the folding or function of expressed chimeric proteins. Before using the ASIC and ENaC fusion proteins for determining heteromeric complex formation, the function of transfected ASIC and ENaC ion channels was assessed using whole cell patch clamp (see Fig. S1).
CHO K1 cells were transfected with combinations of ASIC and ENaC subunits such that traditional ASIC homomultimers or heteromeric ENaCs would be expressed. Amiloride-sensitive currents in CHO K1 cells triply transfected with ␣␤␥ or ␦␤␥ subunits were recorded using a voltage step protocol. The addition of C-terminal fluorescent proteins to all of the expressed subunits simultaneously did not appear to inhibit functional channel formation (Fig. S1B). However, when the fluorescent protein was attached to the N terminus, the ␣␤␥ combination formed a functional amiloride-sensitive channel, but the ␦␤␥ combination did not, thus indicating that the EYFP-␦ENaC construct was not functional (Fig. S1C).
Acid-activated currents were recorded in CHO K1 cells transfected either with ASIC1 or with ASIC2. The addition of N-or C-terminal fluorescent proteins to ASIC subunits did not disrupt formation of functional acid-activated channels, although the kinetics of channel activation and inactivation appeared to differ from those of the nontagged channels (Fig.  S1, D and E). Surprisingly, cells transfected with ASIC1 with both N-terminal ECFP and C-terminal EYFP also formed an acid-activated ion channel. The acid-activated current in these cells, however, did not spontaneously inactivate in the presence of low pH (Fig. S1D).
The results of the whole cell patch clamp experiments indicated that ASIC and ENaC subunits tolerate the addition of Nor C-terminal fluorescent proteins well. In almost all cases, functional ion channels were formed, despite the addition of four or more fluorescent proteins to the intracellular face of the channel. The expression of functional ion channels also indicated that the addition of fluorescent proteins did not dramatically affect the plasma membrane incorporation of the expressed channels; some population of functional, properly assembled ion channels must be present at the plasma membrane to observe whole cell amiloride-sensitive or acid-induced currents.
FRET Microscopy Demonstrates That ASIC and ENaC Subunits Co-aggregate-Fluorescence resonance energy transfer microscopy was used to determine intermolecular interactions between various combinations of ASIC and ENaC subunits. FRET was measured as an increase in ECFP fluorescence intensity upon selective photobleaching of EYFP using high intensity illumination at 514 nm. Cells were sequentially photobleached until EYFP fluorescence intensity was reduced to less than 30% of the initial fluorescence, typically within 2 min. Exposing cells transfected with ECFP alone to 514-nm illumination for the same time had no effect on ECFP fluorescence (data not shown).
For positive FRET controls, cells transfected with either the ECFP-EYFP fusion protein or with the ECFP-ASIC1-EYFP construct were imaged (Fig. 1, A and B). With a seven-amino acid linker between the donor and acceptor fluorophores, the ECFP-EYFP construct would be expected to yield the highest possible FRET, excluding the effects of fluorophore dipole orientation and close-packing interactions. This construct demonstrated 27.4 Ϯ 3.8% FRET (Fig. 1). The ECFP-ASIC1-EYFP fusion served as a positive FRET control for membranebound proteins and had 20.1 Ϯ 3.8% FRET. For negative FRET controls, cells were cotransfected with either ECFP and EYFP or ECFP-CLC1 and EYFP-ASIC1. The ECFP-, EYFP-transfected cells displayed cytoplasmic and nuclear distribution of fluorophores, as seen with the ECFP-EYFP fusion proteins, but did not exhibit significant FRET (6.6 Ϯ 3.5%).
A concern when measuring FRET between membranebound fluorescent tagged proteins is that diffusion of the pro- teins is limited, and therefore nonassociating proteins may display nonspecific FRET due to localized close packing (58,59). To address this issue, a membrane-bound negative control was used. Cells were transfected with fluorescent tagged ASIC1 and CLC-1, a chloride channel not expected to interact with ASIC or ENaC subunits. These cells demonstrated co-localization of the ECFP-and EYFP-tagged proteins, particularly in the endoplasmic reticulum of the cell, but had FRET values not different from the ECFP-, EYFP-co-transfected cells (6.1 Ϯ 1.7%).
ASIC1 and ASIC2 form homomultimeric ion channels, and whole cell patch-clamp experiments demonstrated that the fluorescent tagged ASIC subunits appeared functional when transfected in CHO K1 cells. Cells co-transfected with ECFPand EYFP-tagged ASIC subunits were therefore expected to exhibit significant FRET, since ion channels would assemble with a mixture of ECFP-and EYFP-tagged subunits. In fact, the FRET efficiencies observed in cells co-transfected with ASIC1-ECFP and ASIC1-EYFP were as high as those observed for the positive FRET control (N-terminal tag, 32.9 Ϯ 8.4%; C-terminal tag, 28.6 Ϯ 8.1%). FRET efficiencies for the ASIC2-cotransfected cells were somewhat lower (N-terminal tag, 19.6 Ϯ 3.4%; C-terminal tag, 23.2 Ϯ 4.3%). The FRET efficiencies measured for the N-terminal tagged subunits were not statistically different from those of the C-terminal tagged samples.
ASIC1 has also been shown to form heteromeric channels with ASIC2, and CHO K1 cells transfected with combinations of fluorescent tagged ASIC1 and ASIC2 subunits demonstrated a range of FRET efficiencies from 20.2 Ϯ 8.4% to 25.6 Ϯ 7.6%. Again, the FRET efficiencies measured for the N-terminal tagged subunits were not statistically different from those of the C-terminal tagged samples.
Combinations of ASIC and ENaC subunits exhibited a range of FRET efficiencies from 8.1 Ϯ 5.3% (ASIC1, ␦ENaC with N-terminal tags) to 24.0 Ϯ 5.7% (ASIC1, ␣ENaC with N-terminal tags). Of the ENaC subunits co-expressed with ASIC subunits, ␣and ␥ENaC had higher FRET efficiencies with ASIC1 and ASIC2 than did ␤ and ␦ subunits. Even the combination of ASIC1 with ␦ENaC with gave FRET efficiencies statistically higher than those observed for the negative FRET control (p ϭ 0.046, one-tailed t test). It was notable that this was the same ␦ENaC subunit that did not form amiloride-sensitive ion channels as observed by whole cell patch clamp.
Under typical cell culture conditions, the majority of the expressed fluorescence was found in the endoplasmic reticulum of the transfected cells. In order to address concerns that the observed FRET was due to improperly assembled ASIC and ENaC subunits trapped in the ER of the cell, transfected cells were cultured in the presence of 5 mM Na ϩ phenylbutyrate to enhance translocation of the expressed protein to the plasma membrane (Fig. 2). For all combinations of C-terminally tagged ASIC and ENaC subunits, with the exception of ASIC2 combined with ␣ENaC, statistically significant FRET was observed in regions of interest specifically localized at the plasma membrane of the cell. Overall FRET efficiencies appeared lower than those observed in the ERlocalized samples, suggesting that close packing interactions enhance the observed FRET in ER-localized ASIC and ENaC subunits. Alternatively, the presence of homomeric ASIC channels in the plasma membrane could also reduce the observed FRET efficiency.
Immunoprecipitation-As a complementary method to FRET microscopy, co-immunoprecipitation was used to determine whether heteromeric assembly occurs between ASIC and ENaC subunits in transfected CHO K1 cells. We first demonstrated that ASIC1 and ASIC2 could be co-precipitated from transfected CHO K1 cells using traditional immunoprecipitation and Western blot methods (Fig. 3A). Cells were transfected with nontagged ASIC1 and ASIC2-YFP. The cells were lysed and immunoprecipitated with protein A beads and anti-ASIC1 antibody. Precipitated proteins were then transferred to a polyvinylidene difluoride membrane, and Western blots were resolved first with anti-ASIC1 (Fig. 3A, lane 2). An ϳ70 kDa band was present, consistent with the molecular mass of the untagged ASIC1 subunit. The blot was then stripped and reprobed using anti-ASIC2, exposing an ϳ90 kDa band, consistent with the molecular mass of ASIC2 with a 27-kDa EYFP fusion. The 50 kDa band for IgG was also present. This initial blot demonstrated that EYFP-tagged subunits could be immunoprecipitated with nontagged subunits.
We then took advantage of the attached fluorescent protein to assess whether each of the fluorescent tagged ASIC and ENaC subunits could associate with nontagged ASIC1 or ASIC2. Rather than resolving the precipitated proteins on SDS-PAGE and Western blotting, the protein A-conjugated agarose beads with the captured immunocomplexes were imaged directly on a confocal microscope, using differential interference contrast optics to image beads and EYFP fluorescence to determine the presence of fluorescent proteintagged subunits.
This method was first tested under control conditions. Cells expressing either ASIC1-EYFP or ASIC2-EYFP were solubilized and immunoprecipitated with the appropriate specific antibody. The imaged beads demonstrated bright yellow fluorescence at the perimeter of the beads (Fig. 3B). When the specific antibody was excluded from the immunoprecipitation, no yellow fluorescence was observed, indicating that the EYFP-tagged subunits did not interact directly with the protein A beads. Similarly, when preimmune rabbit IgG was used instead of the specific antibody, no fluorescence was observed, indicating the specificity of the antibodies used. No yellow fluorescence was seen when the cells were transfected with nontagged ASIC subunits, demonstrating the lack of intrinsic autofluorescence of the transfected CHO K1 lysates. Last, when cells were cotransfected with nontagged ASIC subunits and with soluble EYFP,  3). B, control immunoprecipitations resolved by differential interference contrast imaging of agarose beads (gray scale images) and yellow fluorescence as described under "Experimental Procedures." C, experimental co-immunoprecipitations demonstrate intermolecular interactions between ASIC and ENaC subunits using differential interference contrast and yellow fluorescence imaging of immunocomplexed agarose beads. Data are representative of three replicate experiments.
no fluorescence was observed when precipitations were performed using specific antibodies. These controls indicated that the EYFP protein itself did not associate with the protein A beads, the antibody, or the expressed ASIC subunits themselves.
In order to determine whether ASIC and ENaC subunits heteromultimerize, CHO K1 cells were transfected with either nontagged ASIC1 or ASIC2 and with one of the following EYFPtagged subunits: ASIC1, ASIC2, ␣ENaC, ␤ENaC, ␥ENaC, or ␦ENaC. Complexes of ASIC and ENaC subunits were then precipitated with protein A beads and specific antibodies against either ASIC1 or ASIC2 (Fig. 3C). In every combination of subunits tested, yellow fluorescence was observed at the perimeter of the agarose beads. To confirm that the observed fluorescence was due to the fluorescent tagged subunits interacting with the nontagged subunits, each of these precipitations was repeated using cells transfected with only the fluorescent tagged subunits, not the nontagged ASIC1 or ASIC2. For each of these controls, no fluorescence was observed, indicating that the anti-ASIC1 and ASIC2 antibodies do not cross-react with the other ASIC and ENaC subunits.
Combinations of ASIC and ENaC Subunits Form Ion Channels with Novel Electrophysiological Properties-In order to measure electrophysiological properties of defined combinations of ASIC and ENaC subunits, cRNA encoding the ASIC and ENaC subunits was injected into X. laevis oocytes. Several electrophysiological properties were measured, including channel activation and inactivation kinetics, channel activation in response to extracellular acidification, relative permeability of Na ϩ and K ϩ , and pharmacological inhibition of current using amiloride or the ASIC1 selective venom of P. cambridgei.
To limit the number of combinations of ASIC and ENaC subunits to be examined, only ASIC1 was expressed in combination with ENaC subunits, and these combinations were compared with homomeric ASIC1. ASIC2 currents were also recorded for comparison.
Oocytes injected with combinations of ASIC1 and ENaC subunits exhibited acid-activated currents similar to those observed in oocytes injected with ASIC1 alone (Fig. 4). Oocytes injected with individual ␣-, ␤-, or ␥ENaC subunits did not exhibit any acid-activated currents. Previous studies have demonstrated that ␦ENaC can form an acid-activated ion channel, particularly when co-expressed with ␥ and ␦ subunits (60). Our results were consistent with this observation, but the recorded current was miniscule compared with that observed for ASIC1 expressed alone or in combination with ␦ENaC.
Channel-gating kinetics were determined from the acid-induced currents. ASIC1 was typified by rapid activation (116 Ϯ 40 ms) and inactivation (523 Ϯ 174 ms), with a narrow peak half-width (743 Ϯ 253 ms). In contrast, ASIC2 activation was much slower (3501 Ϯ 2060 ms), and the inactivation kinetics and peak halfwidth were primarily dictated by the time of acid application throughout the course of the experiment. Combinations of ␣-, ␤-, or ␥ENaC subunits with ASIC1 resulted in kinetics similar to those observed for ASIC1 alone. ASIC1 expressed in combination with ␦ENaC, however, resulted in acid-induced currents with significantly longer half-width (1299 Ϯ 456 ms) and inactivation kinetics (932 Ϯ 443 ms) than observed in ASIC1 alone.
The sensitivity of expressed ASIC and ASIC⅐ENaC combinations to activation by extracellular protons was determined by recording current amplitudes in response to activation with sequentially lower pH buffers (Fig. S2). In these experiments, ASIC1 displayed a typical pH 50 of 6.0 Ϯ 0.2. ASIC2 was much less sensitive to acid activation (pH 50 4.4 Ϯ 1.5), although this value was not well defined because the peak current did not saturate over the range of pH values tested. Combinations of ASIC1 and ENaC subunits again behaved similarly to ASIC1 alone, with no significant difference seen in pH 50 .
The relative permeability of Na ϩ and K ϩ were next determined for oocytes expressing ASIC1, ASIC2, or ASIC1 in combination with ENaC subunits. When the extracellular Na ϩ was replaced with K ϩ in the low and high pH perfusion buffers, the oocytes exhibited smaller peak current amplitudes when channel opening was induced by a shift from pH 7.4 to pH 5.0 (Fig.  5). This effect was completely reversible upon returning to Na ϩ -containing buffers. The relative permeability of K ϩ to Na ϩ was taken as the ratio of the K ϩ peak current to the Na ϩ FIGURE 4. Gating kinetics of heteromeric ASIC⅐ENaC ion channels. Whole cell acid-induced currents were recorded from X. laevis oocytes injected with 25 ng of cRNA for ASIC1, ASIC2, or combinations of ASIC1 with ␣, ␤, ␥, or ␦ ENaC. Acid-activated currents were induced by rapid switching of perfusion buffer from pH 7.4 to pH 5.0 (A). Gating kinetic parameters including peak width (half-width, half-maximal activation) (B), activation (C), and inactivation (D) were determined from n oocytes. Error is S.D. *, significant difference from value observed from ASIC1 alone (p Ͻ 0.013, two-tailed t test). peak current. The K ϩ /Na ϩ permeability of ASIC1 was 0.41 Ϯ 0.09. ASIC2 was significantly less selective for Na ϩ when activated at pH 5.0 (I K /I Na ϭ 0.57 Ϯ 0.07). When co-expressed with ␣-, ␤-, or ␦ENaC subunits, the permeability of ASIC1 became significantly more selective for Na ϩ compared with K ϩ (I K /I Na ϭ 0.21 Ϯ 0.06 to 0.22 Ϯ 0.07). These results indicate that subunit composition influences the ion conductive pathway of the heteromeric ASIC⅐ENaC complex.
Ion channels composed of members of the ENaC/degenerin superfamily are typified by inhibition with the diuretic amiloride. To determine the relative inhibition of ASIC and ASIC⅐ENaC combinations with amiloride, acid-activated currents were recorded in the absence or presence of increasing concentrations of amiloride (Fig. S3). For oocytes injected with ASIC1 alone, the observed IC 50 for amiloride was 61.2 Ϯ 7.9 M. For ASIC2, the observed IC 50 was 18.6 Ϯ 2.2 M. Amiloride IC 50 values for combinations of ASIC1 with ENaC subunits varied from 34.8 Ϯ 6.6 (ASIC1 ␥ENaC) to 67.58 Ϯ 11.0 (ASIC1 ␤ENaC), not significantly different from ASIC1 expressed alone. These results indicate that the observed inhibition of acid-activated currents is not affected by channel subunit composition.
Unique among the ion channels of the ENaC/degenerin family, ASIC1 can also be inhibited by psalmotoxin 1, a 40-amino acid inhibitor cysteine knot toxin isolated for the venom of P. cambridgei, a tarantula found in the West Indies (35). Previous reports indicate that heteromeric complexes of ASIC subunits or ENaCs are not inhibited by psalmotoxin, but amiloride-sensitive ion channels found in glioma cells are blocked with subnanomolar concentrations of PcTX1 (51). Lacking purified, properly folded PcTX1, the raw venom of P. cambridgei was used here as a specific inhibitor of ASIC1.
When expressed in Xenopus oocytes, ASIC1 displays a modest peak current run down observed with repeated activation at pH 4.0 (Fig. 6A). Application of a high concentration of PC venom (0.2 l of venom/ml) caused complete inhibition of ASIC1 currents over the course of about 2 min. The acid-activated currents of oocytes expressing ASIC2 were not affected with this concentration of PC venom. ASIC1 currents were also not affected by application of high concentrations of the venom of another tarantula, Grammostola rosea (GS toxin; Spider Pharm, Yarnell, AZ).
Inhibition of ASIC1 currents using PC toxin was dose-dependent; serial application of increasing concentrations of PC venom resulted in decreasing current amplitudes in oocytes injected with ASIC1 or combinations of ASIC1 and ENaC subunits (Fig. 6B). Because the concentration of PcTX1 was not known in the raw venom, relative IC 50 values were reported as the dilution of venom in l/ml of perfusion buffer. The relative IC 50 of PC venom for ASIC1 was 36.2 ϫ 10 Ϫ3 Ϯ 5.7 ϫ 10 Ϫ3 l venom/ml). IC 50 for ASIC2 could not be determined, because there was no observed inhibition of ASIC2 currents over the range of concentrations tested. For every tested combination of ASIC1 with ENaC subunits, the observed IC 50 for PC venom was significantly reduced from that of ASIC1 alone, ranging from 8.1 ϫ 10 Ϫ3 Ϯ 0.6 ϫ 10 Ϫ3 (ASIC1 ␣ENaC) to 11.3 ϫ 10 Ϫ3 Ϯ 1.2 ϫ 10 Ϫ3 (ASIC1 ␦ENaC). The Hill coefficient of PC venom inhibition of ASIC1 expressed alone was 1.16 Ϯ 0.17, indicating noncooperative binding. When ASIC1 was expressed in combination with ENaC subunits, the Hill coefficients rose, ranging from 1.57 Ϯ 0.35 (ASIC1 ␦ENaC) to 2.40 Ϯ 0.36 (ASIC1 ␣ENaC). These results indicate that the addition of ENaC subunits to ASIC subunits results in channels with cooperativity in PC venom binding sites.
In summary, two-electrode voltage clamp recordings in X. laevis oocytes expressing combinations of ASIC1 and ENaC subunits provide evidence that ASIC1 and ENaC subunits can functionally interact. Specific combinations of subunits resulted in ion channels with significant differences in channel gating kinetics, ion permeability, and pharmacological inhibition using P. cambridgei venom.

DISCUSSION
The experiments presented here address the question of whether ASIC and ENaC subunits are capable of forming heteromeric complexes and whether those complexes can form FIGURE 5. Relative permeability of Na ؉ and K ؉ in heteromeric combinations of ASIC and ENaC subunits. Acid-induced currents (pH 5.0) were recorded in oocytes injected with cRNA for ASIC1, ASIC2, and ASIC 1 co-injected with ENaC subunits. Na ϩ currents were recorded in standard ND96 buffer, and K ϩ currents were recorded in KD98 buffer (see "Experimental Procedures" for buffer composition). Both Na ϩ and K ϩ currents were recorded from the same oocyte, and peak current amplitudes were normalized to maximal activation in ND96. Representative traces are shown for ASIC1 expressed alone (A) and for ASIC1 co-expressed with ␣ENaC (B). Black lines, ND96 current traces; gray lines, KD98 traces. The ratio of K ϩ current compared with Na ϩ current was averaged from n oocytes (C). Error is S.D. *, significant difference from ASIC1 (p Յ 0.019, two-tailed t test).
functional ion channels with electrophysiological properties distinct from the individual ASIC and ENaC channels. To this end, FRET microscopy and co-immunoprecipitation experiments were used to assess intermolecular interactions between ASIC and ENaC subunits, and two-electrode voltage clamp of cRNA-injected X. laevis oocytes was used to determine the electrophysiological properties of expressed channels. The results of these experiments suggest that ASIC and ENaC subunits are capable of promiscuous intermixing; any tested combination of ASIC with ENaC subunits resulted in a heteromeric complex assembly detected with both FRET and co-immunoprecipitation. Heteromeric assembly of ASIC1 and ENaC subunits also led to novel channel formation with unique gating, ion conductance, and pharmacological sensitivities observed in heteromeric complexes of ASIC and ENaC subunits.
In these experiments, ASIC and ENaC subunits were expressed in either CHO K1 cells or Xenopus oocytes rather than in cells that typically express these ion channels. This was done specifically so that endogenous channel subunits would not compete with transiently transfected subunits for proteinprotein interactions. Additionally, the presence of endogenous ASICs or ENaCs would have interfered with characterization of the fluorescent tagged channels by whole cell patch clamp.
The addition of the 27-kDa fluorescent protein to either the N or C terminus of the ASIC and ENaC subunits did not appear to inhibit ion channel function in homomeric ASICs or hetero-meric ENaCs. In the transfected cells, fluorescent tags were expected to be present on every component subunit in the expressed ion channels. Although differences were observed in the electrophysiological characteristics of cells transfected with tagged subunits, functional ion channels were still expressed. The one exception to this was using the ␦ENaC subunit with N-terminal tagged EYFP.
ASIC and ENaC subunits expressed in the CHO K1 cells localized primarily to the endoplasmic reticulum of the cell. Despite the preponderance of channel subunits trapped in intracellular structures, some population of assembled subunits must be present at the surface of the cells, because ASIC and ENaC currents can be recorded from the transfected cells by whole cell patch clamp. ASIC and ENaC subunits could also be forced to the plasma membrane of the cell using Na ϩ phenylbutyrate in the cell culture medium. This allowed for FRET microscopy experiments to measure intramolecular associations both within the ER and at the plasma membrane of the cell. Fluorescence resonance energy transfer occurs when two fluorophores, one donor and one acceptor, have spectral overlap from the emission of the donor to the absorbance of the acceptor. When the two fluorophores are in close physical proximity, nonradiative energy transfer occurs from the excited fluorescence donor to the fluorescence acceptor according to the Förster equation, R ϭ R 0 (1/E Ϫ 1) 1 ⁄ 6 where the distance between a fluorescence donor and acceptor R is related to the characteristic distance for resonant energy transfer R 0 and the measured efficiency of energy transfer E. Although the spatial resolution of traditional fluorescence colocalization microscopy is limited by diffraction to ϳ2000 -3000 Å, FRET is highly sensitive to the distance between fluorophores on a scale of 10 -100 Å, an ideal range for the study of macromolecular complexes (61).
Several complicating factors can make FRET results difficult to interpret as a true quantitative measurement of distance (61).
In these experiments, all samples were bleached to within 30% of the original yellow fluorescent protein fluorescence intensity, but the extent of bleaching varied from ϳ20 to 30%. This made extrapolation of the true FRET efficiency, at which point the acceptor fluorophore was completely photobleached, difficult. There also was no control of relative concentrations of ECFP to EYFP present in the cells. In all experiments, cells were transfected with equal concentrations of vectors incorporating ECFP-and EYFP-tagged subunits, but the total fluorescence expression and relative intensities of the two fluorophores varied widely from cell to cell. In the case of excess donor fluorophore, the FRET signal will be diminished when donor fluorophores are not located near acceptor fluorophores.
The use of green fluorescent protein-derived fluorescent proteins has vastly enhanced the use of FRET imaging in transfected cells (61). Using ECFP and EYFP as the fluorescence donor and acceptor fluorophores, however, also presents some difficulties. Green fluorescent protein-derived fluorescent proteins have a potential for self-aggregation when present in high concentrations (62). This becomes particularly problematic when using ECFP as the donor fluorophore. The relatively lower fluorescence intensity of this fluorophore requires high expression levels for FRET imaging. These factors may explain the FRET observed in the ECFP-, EYFP-cotransfected cells used as the negative FRET control (63,64).
The issue of close packing interactions yielding high background FRET intensities is further complicated in membrane-bound systems, in which diffusion of the fluorophores is limited to two dimensions as opposed to three. To build confidence that the soluble ECFP and EYFP controls were valid for membrane-bound systems, additional, membranebound positive and negative controls were used (ECFP-ASIC1-EYFP and ECFP-CLC1, EYFP-ASIC1 cotransfections, respectively). These controls yielded similar FRET values as the soluble FRET controls.
In our FRET experiments, we observed a wide range of FRET efficiencies. Several combinations of ASIC and ENaC subunits (ASIC1 or ASIC2 homomultimers, ASIC1/ASIC heteromultimers, or combinations of ASIC with ␣or ␥ENaC) demonstrated FRET efficiencies similar to those observed with the ECFP-EYFP positive control. The remaining combinations of ASIC and ␤and ␦ENaC subunits demonstrated intermediate FRET efficiencies of ϳ8 -18% that could indicate a true heteromeric interaction but could also potentially be explained by some of the previously mentioned complicating factors in FRET microscopy. When expressed ASIC and ENaC subunits were forced to the plasma membrane, again a wide range of intermediate FRET values were observed, with only a single combination of subunits not statistically higher than the observed negative control value (ECFP-ASIC2, EYFP-␣ENaC). These results suggested that ASIC and ENaC subunits could intermix promiscuously, both within the endoplasmic reticulum and at the plasma membrane of transfected CHO K1 cells.
Because of the number of complicating factors in FRET analysis, it was important to demonstrate the heteromeric assembly of subunits using an alternative method. Co-immunoprecipitation is a traditional technique for examining interactions between physically associating proteins. We used antibodies directed against nonfluorescent tagged ASIC subunits to capture fluorescent tagged interacting subunits. Typically, the interacting proteins are resolved by Western blot, using either a specific antibody for the interacting protein or against a convenient peptide tag introduced as a fusion (i.e. hemagglutinin, His 6 , and c-Myc). In our experiments, EYFP served as an excellent fusion tag with the added benefit of being a fluorescent indicator. This intrinsic fluorescence was used to eliminate the need for Western blotting by simply imaging immunoprecipi-tated EYFP-tagged subunits, which were captured to protein A beads with antibodies directed against the co-expressed, nontagged ASIC subunits. The immunoprecipitation experiments recapitulated the FRET observations in that ASIC1 or ASIC2 were capable of binding to one another as well as binding any of the ENaC subunits.
With evidence strongly suggesting physical interactions between co-expressed ASIC and ENaC subunits, the functional interaction between subunits was next addressed by characterizing the electrophysiological footprint of co-expressed ASIC⅐ENaC combinations compared with homomultimeric ASIC channels. An advantage of the oocyte expression system used here, compared with transient transfection of CHO K1 cells, was that combinations of cRNAs could be co-injected into each oocyte with precise control of composition and concentration. By contrast, transient transfection in CHO K1 cells had relatively low efficiency, and the expression levels of ECFP and EYFP tagged subunits varied from cell to cell, as observed by confocal microscopy. When expressed individually, ENaC subunits displayed minimal, if any, acid-activated channel activity. When co-expressed with ASIC1, however, acid-activated currents were apparent that had similar activation and inactivation kinetics as well as pH sensitivities compared with homomeric ASIC1. Only the combination of ASIC1 with ␦ENaC exhibited a current that was significantly slower to desensitize than ASIC1 alone. These results did not strongly suggest a subunit composition-dependent effect on ion channel function. However, the relative permeability of K ϩ to Na ϩ was strongly affected by channel composition, with ␣, ␤, and ␦ subunits causing significant decreases in the I K /I Na permeability ratios. Similarly, the pharmacological inhibition of heteromeric complexes of ASIC1 and ENaC subunits was strongly influenced by subunit composition; ASIC1 in combination with any of the four tested ENaC subunits caused increased sensitivity to inhibition using venom from the tarantula P. cambridgei.
FRET microscopy and immunoprecipitation experiments demonstrated that ASIC and ENaC subunits can intermix. The electrophysiological experiments demonstrated that these interactions result in functional ion channels with distinct electrophysiological properties. Acid-activated channel gating in multimeric complexes appeared to be dominated by the presence of ASIC1. The conductance through the ion-conductive pore of the channel, however, was influenced by channel composition, suggesting that ENaC subunits contribute to the pore of the ion-conductive channel. Amiloride inhibition was not affected by channel composition, suggesting an interaction with a binding site primarily composed of only a single subunit. PC venom inhibition and cooperativity, however, was enhanced in all combinations of ASIC and ENaC subunits. Salinas et al. (65) have shown that the binding site of PcTX1 has contributions from two cysteine-rich domains in the extracellular loop of rat ASIC1a. We propose that these regions are contributed from neighboring subunits rather than from the same subunit. The presence of ASIC⅐ENaC interfaces could therefore explain the observed change in inhibitory concentration and Hill coefficient. An alternative explanation for enhanced inhibition of acid-induced currents in ASIC⅐ENaC heteromers could be that ENaC subunits may provide sensitiv-ity of the heteromeric complex to additional components of the raw venom, potentiating ASIC inhibition by PcTX1. Regardless of the mechanism of increased venom sensitivity, the PC venom data strongly suggest that heteromeric ion channels are formed from combinations of ASIC and ENaC subunits.
Emerging evidence suggests that ASIC and ENaC subunits may co-assemble into heteromeric complexes in a wide variety of tissues with diverse electrophysiological characteristics. Small diameter dorsal root ganglion cells (associated with nociception) harvested from adult male rats express as many as eight distinct acid-activated ASIC-mediated currents (66). These currents differ in such electrophysiological properties as gating kinetics, pH dependence of activation and inactivation, and sensitivity to pharmacological inhibition using PC venom or Zn 2ϩ . The authors of this study suggest that the diversity of acid-activated currents observed are probably due to heteromeric combinations of ASIC subunits, but the previously reported co-expression of ENaC subunits in these cells suggests that heteromeric ASIC⅐ENaC complexes may also be present (32)(33)(34). The acid-activated currents observed in oocytes expressing ASIC1-ENaC combinations are particularly reminiscent of the currents described in these DRG cells. Further roles for heteromeric channels composed of ASIC and ENaC subunits may become apparent with additional study of the electrophysiological characteristics of heteromeric combinations of ASIC2, -3, or -4 with various ENaC subunits.