Electrophysiological and Biochemical Evidence That DEG/ENaC Cation Channels Are Composed of Nine Subunits*

Members of the DEG/ENaC protein family form ion channels with diverse functions. DEG/ENaC subunits associate as hetero- and homomultimers to generate channels; however the stoichiometry of these complexes is unknown. To determine the subunit stoichiometry of the human epithelial Na+ channel (hENaC), we expressed the three wild-type hENaC subunits (α, β, and γ) with subunits containing mutations that alter channel inhibition by methanethiosulfonates. The data indicate that hENaC contains three α, three β, and three γ subunits. Sucrose gradient sedimentation of αhENaC translated in vitro, as well as α-, β-, and γhENaC coexpressed in cells, was consistent with complexes containing nine subunits. FaNaCh and BNC1, two related DEG/ENaC channels, produced complexes of similar mass. Our results suggest a novel nine-subunit stoichiometry for the DEG/ENaC family of ion channels.

expression of all three subunits is required to generate maximal Na ϩ current, although expression of ␣ENaC alone can produce small currents. In addition, biochemical data show that the three human ENaC (hENaC) subunits associate (19). Genetic evidence suggests that MEC-4, MEC-10, and DEG-1 also function as heteromultimers (7,20). However, the subunit stoichiometry is not known for any DEG/ENaC channel. EXPERIMENTAL PROCEDURES cDNAs and mutations were generated as described previously (9,17,18). FaNaCh was amplified by polymerase chain reaction following reverse transcription of RNA from H. aspersa. We tagged the C terminus of ␣hENaC with the sequence DYKDDDDK (␣ Flag ) for immunoprecipitation by anti-Flag M2 monoclonal antibody. This did not alter the function of the ␣ subunit in Xenopus oocytes or epithelia or its ability to associate with other subunits (19).
Wild-type or mutant ␣-, ␤-, and ␥hENaC (0.2 ng each) were expressed in Xenopus oocytes by nuclear injection of cDNA (18). When a mixture of wild-type and mutant cDNAs for a subunit was coinjected, the total amount of cDNA for the subunit remained constant. 16 -24 h after injection, whole-cell Na ϩ current was measured by two-electrode voltage clamp at Ϫ60 mV (bathing solution, 116 mM NaCl, 2 mM KCl, 0.4 mM CaCl 2 , 1 mM MgCl 2 , 5 mM HEPES, pH 7.4). To increase the affinity for amiloride of channels containing ␥ G536C , we decreased the Na ϩ concentration for experiments in Fig. 1 (25 mM NaCl, 93 mM KCl). Under these conditions, 3 mM amiloride completely inhibits hENaC current. The fraction of Na ϩ current inhibited by MTS 1 reagents, Inh, was determined by measuring current blocked by a maximal concentration of amiloride before and after addition of MTSET (Toronto Research Chemicals) or MTSEA to the bathing solution for 100 s. Following coexpression of wild-type and mutant hENaC subunits, we determined the number of ␣, ␤, or ␥ subunits in the channel complex, n, as described previously by MacKinnon (21) using the equation Inh ϭ f wt n Inh wt ϩ ͑1 Ϫ f wt n ͒Inh mut where Inh wt is the fraction of current inhibited in channels containing only wild-type subunits, Inh mut is the fraction of current inhibited in channels containing n mutant subunits, and f is the fraction of subunits expressed that are wild-type or mutant, as indicated.
cDNAs were transcribed and translated in vitro in the presence of canine pancreatic microsomal membranes, as described previously (22). COS-7 cells were electroporated as described previously (19), pulselabeled with 100 Ci/ml [ 35 S]methionine (NEN Life Science Products) at 37°C for 30 min, and then chased for 0 -7 h at 15°C. We have found that hENaC subunits assemble in the ER (19), then become insoluble to detergents prior to transport to the golgi. 2 To minimize insolubility, we incubated cells at 15°C to inhibit transport out of the ER. Proteins were solubilized in cold Tris-buffered saline (50 mM Tris, 150 mM NaCl) containing 1% digitonin, 0.4 mM phenylmethylsulfonyl fluoride, 20 g/ml aprotonin, 20 g/ml leupeptin, and 10 g/ml pepstatin A and sedimented on sucrose gradients as described in figure legends.

Stoichiometry of ␥hENaC
Using the G536C Mutation-To investigate the stoichiometry of hENaC, we first asked how many ␥ subunits contribute to the channel complex. Our strategy, similar to that described by MacKinnon (21), was to coexpress mixtures of wild-type and mutant ␥ subunits in Xenopus oocytes and then determine the sensitivity to an inhibitor. We used MTSET, 3 an agent that covalently modifies cysteines. Wild-type hENaC is relatively insensitive to MTSET (Fig. 1A).
Previous studies indicate that Gly 536 in the ␥ subunit lines the * 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  channel pore (23). 3 When we replaced Gly 536 with cysteine (␥ G536C ), MTSET irreversibly decreased current 87% (Fig. 1B) by covalently modifying the introduced cysteine. Wild-type ␥ (␥ wt ) and ␥ G536C produced equal Na ϩ currents (Fig. 1E). Interestingly, when we coexpressed a 0.5:0.5 mixture of ␥ wt and ␥ G536C (with wild-type ␣ and ␤hENaC), most of the Na ϩ current (77%) was inhibited by MTSET (Fig. 1C). When we increased the contribution of ␥ wt (0.8:0.2 ␥ wt :␥ G536C ), MTSET decreased Na ϩ current 48% (Fig. 1D). The finding that MTSET decreased Na ϩ current out of proportion to the fraction of mutant subunits suggested that hENaC contains more than one ␥ subunit and the ␥ G536C phenotype is dominant.
To determine the number of ␥ subunits in hENaC, we make two assumptions. First, we assume that ␥ wt and ␥ G536C express equally and associate randomly in the channel complex. This seems reasonable since expression of ␥ wt or ␥ G536C with ␣and ␤hENaC produced equal amounts of Na ϩ current. If this is correct, then channel composition will be determined by a binomial distribution. Second, we assume that a single mutant ␥ subunit is sufficient to make a channel sensitive to MTSET. The dominant effect of ␥ G536C suggests this is correct.
Consider the examples illustrated in Fig. 1G (expression of 0.5:0.5 ␥ wt :␥ G536C ). If hENaC has one ␥ subunit, there will be two channel populations; one will contain a wild-type ␥ subunit and the other will contain ␥ G536C . In contrast, if hENaC has three ␥ subunits, then one-eighth of the channels will have three wild-type ␥ subunits and one-eighth will have three mutant subunits. The remainder (six-eighths) will contain both wild-type and mutant ␥ subunits. If a single mutant ␥ makes a channel sensitive to MTSET, then the fraction of Na ϩ current sensitive to MTSET will be determined by the fraction of channels that have at least one mutant ␥ subunit. Thus, we can calculate the number of ␥ subunits in the hENaC complex (n) by measuring the fraction of Na ϩ current inhibited by MTSET (Inh). Fig. 1F shows a plot of the values measured for Inh when we expressed ␥ wt (f wt ϭ 1.0), ␥ G536C (f wt ϭ 0), or mixtures of both. Curves for expected values of Inh if n is 1-4 are superimposed. The data suggest hENaC has three ␥ subunits.
Stoichiometry of ␣-, ␤-, and ␥hENaC Using the "Deg" Mutation-In contrast to MTSET, MTSEA inhibited current in cells expressing wild-type subunits by 54% ( Fig. 2A). Our unpublished data suggest that MTSEA inhibits by covalently modifying one or more pore-lining cysteines in the ␥ subunit. As a second independent test of the number of ␥ subunits and to determine the number of ␣ and ␤ subunits, we took advantage of a mutation in the Deg residue that prevents inhibition. In MEC-4, mutation of Ala 442 to bulky amino acids causes neurodegeneration (20), and mutation of the equivalent residue in  Fig. 1. E, amiloride-sensitive Na ϩ current (relative to wild-type) in oocytes expressing two wild-type subunits with the indicated mutant subunit (mean Ϯ S.E., n ϭ 7-8).
BNC1 activates the channel (10). In hENaC subunits, the Deg residues are serines. We found that mutation of a Deg serine to cysteine in any of the subunits (␣ S549C , ␤ S520C , and ␥ S529C ) abolished inhibition by MTSEA. Instead, MTSEA and MTSET ( Fig. 2A) stimulated a small increase in Na ϩ current. In this regard, modification of this residue in hENaC may be similar to the activation that occurs with a bulky residue at the Deg position in MEC-4 and BNC1. Perhaps after the cysteine is modified, it provides a steric barrier that prevents MTSEA from entering the pore where it could alter the ␥-cysteine(s) and inhibit current.
We expressed mixtures of these mutant and wild-type subunits and measured current inhibition by MTSEA. Because MTSEA stimulates channels containing the Deg Ser-to-Cys mutation, it complicates the quantitative assessment of MTSEA's inhibition of wild-type channels. To circumvent this problem, we took advantage of the fact that MTSET also stimulates these mutant channels ( Fig. 2A), but it does not inhibit wild-type channels (Figs. 1A and 2A). We performed the experiments in two steps as shown in Fig. 2A. First, we added MTSET. It stimulated channels containing ␣ S549C , but more importantly, it prevented stimulation on subsequent addition of MTSEA. Second, after washing out MTSET, we applied MTSEA and measured the inhibitory effect on amiloride-sensitive Na ϩ current ( Fig. 2A). In contrast to wild-type hENaC, when we expressed ␣ S549C , MTSEA produced a minimal decrease in current (6.9%). Studies with mixtures suggest that ␣ S549C prevents channel inhibition by MTSEA in a dominant manner, since Na ϩ current was inhibited less than the proportion of wild-type ␣ subunits. Similar results were obtained with ␤ S520C and ␥ S529C , although MTSET stimulated ␥ S529C less than the other mutants. Each mutant produced Na ϩ currents equal to wild-type (Fig. 2E), supporting the assumption that wild-type and mutant subunits express equally. Fig. 2, B-D, shows the fraction of Na ϩ current inhibited by MTSEA versus the fraction of subunits that were wild-type. The results suggest that hENaC contains three ␣, three ␤, and three ␥ subunits.
Sucrose Gradient Analysis of Channel Mass-To further test stoichiometry, we determined the molecular mass of hENaC by sucrose gradient sedimentation. Because ␣hENaC can form a homomeric channel, we first determined the mass of channels containing only the ␣ subunit. We translated ␣hENaC in vitro with microsomal membranes. This assay allows subunit multimerization in the ER, while minimizing the possibility that other cellular proteins will associate with the hENaC channel complex and alter its molecular mass. Fig. 3A shows that unglycosylated ␣ subunits sedimented mainly in fractions 4 and 5, similar to a standard of 240 kDa. This migration suggested that most of the unglycosylated subunits were in the form of dimers or trimers, consistent with our previous finding that subunit interactions begin prior to glycosylation (19). In contrast, glycosylated subunits sedimented in the same fractions (8 -10) as a 950-kDa standard (Fig. 3A), consistent with a channel complex containing at least nine subunits. Because most of the ␣ subunits in fractions 8 -10 were glycosylated, the data suggest that subunits oligomerize during processing in the ER. Further, multimerization is probably very efficient, since little glycosylated ␣ was found in fractions 4 and 5. Because most multimeric proteins multimerize in the ER (24), this assay probably provides an accurate representation of the size of the complex at the plasma membrane.
We asked whether other members of the DEG/ENaC family form a complex of similar size. The glycosylated form of FaNaCh sedimented in the same fractions as ␣hENaC (Fig.  3B). BNC1 sedimented in lighter fractions (peak in fraction 8) (Fig. 3B), consistent with the lower molecular mass of glycosylated BNC1 monomer (70 kDa) compared with ␣hENaC (87 kDa) and FaNaCh (90 kDa). These results are consistent with FaNaCh and BNC1 complexes that contain at least nine subunits.
To determine whether the mass of an hENaC complex containing ␣, ␤, and ␥ was the same as a homomeric complex, we coexpressed the three subunits in COS-7 cells. Sedimentation was determined at times from 0 to 7 h after pulse labeling. Immediately after the 30-min pulse, immunoprecipitated ␣ subunits peaked in fraction 4, and very little ␣ was present in fractions 8 and 9 (Fig. 4). This suggests a complex containing two-three subunits. With time, the fraction of ␣ in denser fractions increased, and at 7 h, a significant amount of ␣ subunits were in a complex that sedimented in fractions 8 and 9. We also determined the sedimentation of ␤ subunits that coimmunoprecipitated with ␣hENaC. At 7 h, a large fraction of ␤ subunits were also in a complex sedimenting in fractions 8 and 9. Because this assay only detected ␤ subunits that were associated with ␣, the results suggest that fractions 8 and 9 contain heteromeric complexes and likely represent the oligo- Following sedimentation on 10 -45% gradient, ␣hENaC was immunoprecipitated and separated by SDS-PAGE, and the glycosylated form in each fraction was quantitated by phosphorimaging. ␤hENaC that coprecipitated with ␣hENaC was also quantitated (at 0 h, insufficient ␤ was present for quantitation). meric state of the functional channel complex. These data suggest that the channel complex contains at least nine subunits. We cannot exclude the possibility that other cellular proteins tightly associate with hENaC and influence the density. However, this seems unlikely because sedimentation of the channel complex in cells and in vitro was similar. DISCUSSION Our functional and our biochemical data support a model of hENaC containing nine subunits; three ␣, three ␤, and three ␥. These results suggest a novel stoichiometry for an ion channel and contrast with the stoichiometry of voltage-gated K ϩ channels (four subunits) (21) and Na ϩ /Ca 2ϩ channels (four repeats), nicotinic acetylcholine receptors (five subunits), and gap junction hemichannels (six subunits) (25).
The finding that BNC1 and FaNaCh migrated on sucrose gradients similar to hENaC suggests that those channels may also be constructed from nine subunits. Thus, a multimeric complex composed of nine subunits may be a conserved feature of the DEG/ENaC family. In addition, genetic evidence suggests that some C. elegans family members contain more than one copy of each subunit. For example, MEC-4, MEC-10, and probably one unidentified subunit may form a functional complex. A loss-of-function mutation on one mec-4 allele suppressed a dominant neurodegeneration-associated mutation on the other mec-4 allele (26). Similar interactions appear to occur with MEC-10 subunits (27). These results suggest that the functional complex contains two or more MEC-4 and two or more MEC-10 subunits (7).
The approach we used has several advantages. First, the electrophysiological assays allowed us to selectively determine the stoichiometry of functional channels. If different combinations of subunits produced nonfunctional channels or channels not delivered to the cell surface, they would not be detected. Second, our approach allowed us to determine the absolute number of ␣, ␤, and ␥ subunits in the channel complex, rather than a ratio of one subunit relative to another. Third, use of two independent electrophysiological assays for the ␥ subunit strengthen our conclusions. Fourth, we were able to avoid the use of concatameric constructs, which could disrupt the normal association of subunits, as previously reported for Shaker K ϩ channels (28). Finally, we used both functional and biochemical approaches. Both supported the same conclusion.
Our approach also has limitations. First, our electrophysiological assays would not detect inactive channels. Second, our sucrose gradient assay detected channels in the ER, rather than channels at the plasma membrane. However, earlier work suggested that the hENaC subunits assemble in the ER (19). Finally, our electrophysiological approach relies on two assumptions; that wild-type and mutant subunits express equally and associate randomly and that a single mutant subunit has a dominant effect on MTS sensitivity. As discussed above, these assumptions are likely valid. However, if the assumption that a single mutant subunit is dominant is incorrect, our calculations will underestimate the number of subunits. In the case of all three limitations, the concordance between biochemical and functional data support the validity of our approach.
It is interesting to speculate how nine subunits might assemble to form a highly selective channel. Based on theoretical considerations, Guy and Durell (29) independently proposed a model for ENaC that contains three ␣, three ␤, and three ␥ subunits. In their model, the pore is formed by ␣␤ barrels derived from the residues immediately preceding M1 and M2, and they predicted that all nine subunits contribute to the channel pore. Our experimental data are consistent with this model.