Mutational Analysis of Cysteine-rich Domains of the Epithelium Sodium Channel (ENaC)

One of the characteristic features of the structure of the epithelial sodium channel family (ENaC) is the presence of two highly conserved cysteine-rich domains (CRD1 and CRD2) in the large extracellular loops of the proteins. We have studied the role of CRDs in the functional expression of rat αβγ ENaC subunits by systematically mutating cysteine residues (singly or in combinations) into either serine or alanine. In the Xenopusoocyte expression system, mutations of two cysteines in CRD1 of α, β, or γ ENaC subunits led to a temperature-dependent inactivation of the channel. In CRD1, one of the cysteines of the rat αENaC subunit (Cys158) is homologous to Cys133 of the corresponding human subunit causing, when mutated to tyrosine (C133Y), pseudohypoaldosteronism type 1, a severe salt-loosing syndrome in neonates. In CRD2, mutation of two cysteines in α and β but not in the γ subunit also produced a temperature-dependent inactivation of the channel. The main features of the mutant cysteine channels are: (i) a decrease in cell surface expression of channel molecules that parallels the decrease in channel activity and (ii) a normal assembly or rate of degradation as assessed by nondenaturing co-immunoprecipitation of [35S]methionine-labeled channel protein. These data indicate that the two cysteines in CRD1 and CRD2 are not a prerequisite for subunit assembly and/or intrinsic channel activity. We propose that they play an essential role in the efficient transport of assembled channels to the plasma membrane.

fied in the nervous system of mammals; (iii) FaNaCh involved in synaptic transmission in snail; (iv) MEC-4, MEC-10, DEG-1 (degenerins), and UNC 105 of the Caenorhabditis elegans nematode expressed in sensory neurons and muscles, respectively (1). All proteins of this supergene family share a common membrane topology with two transmembrane domains, short intracellular N and C termini and a large extracellular loop (2). An heterotetrameric structure has been recently proposed for ENaC (3,4), whereas an homo-tetrameric structure has been proposed for FaNaCh (5). It is therefore likely that this gene family of cation channel is tetrameric, despite a study suggesting a nonameric architecture for ENaC (6).
Structure-function studies performed on ENaC subunits have indicated some specific functional roles of different domains of the proteins. Schild et al. (7), have demonstrated that a stretch of amino acids, preceding the second (M2) transmembrane domain (pre-M2 domain) of each of the three subunits contributes to the formation of channel pore and is a major site for amiloride binding. Schild et al. (8), Snyder et al. (9), Staub et al. (10), and Shimkets et al. (11) have identified a PY sequence in the intracellular C termini of the ␤ and ␥ subunits as possible sites responsible for recycling of the cell surface expressed ENaC, either through the binding to an ubiquitine ligase (Nedd4), and/or by clathrin-mediated endocytosis. In addition, mutations of the PY motif abolishes the feedback inhibition of ENaC by intracellular Na ϩ , an important mechanism controlling Na ϩ reabsorption (12). In intact cells (i.e. Xenopus oocytes) phosphorylation studies have also demonstrated a possible role of the C termini, as the target sequences for hormone-stimulated protein kinases A-and C-dependent phosphorylation of the ␤ and ␥ ENaC subunits (13). The intracellular N terminus of the ␣ subunit has been identified as a possible gating domain for ENaC (14). The same portion of the ␥ subunit protein has also been described as participating in ENaC subunit assembly (15). Little information is, however, available about the structural and functional role of the extracellular loop of ENaC subunits. The extracellular loop represents the largest domain (ϳ70% of amino acids) and is glycosylated for most members of the family. The most notable feature of the extracellular loop is the presence of two cysteinerich domains (CRD1 and CRD2) covering about 50% of the sequence. Among 15 ␣␤␥ ENaC subunits cloned from different species, all the extracellular cysteines (16) are conserved, suggesting that they are involved in disulfide bond formation. Most of these cysteines (14) are also conserved among the genes expressed in mammalian nervous system (MDEG 1 and 2, ASIC, and DRASIC) as well as among the degenerins and FaNaCh. The degenerins family is characterized by the presence of an additional CRD (16). Considering the number, sequence distribution, and potential disulfide bond formation, one could suggest an important role(s) of these cysteines in formation of mature ENaC/DEG channels. A particular interest into the CRDs arose from the recently identified Cys 133 to Tyr 133 mutation in the CRD1 of human ␣ENaC subunit causing the pseudohypoaldosteronism type 1 (PHA-1), an inherited disease characterized by severe neonatal salt wasting, hyperkaliaemia, metabolic acidosis, and unresponsiveness to the mineralocorticoid hormone aldosterone (17). In studies of the mechanism of this mutation we have demonstrated that ␣C133Y replacement leads to the temperature-sensitive decrease in functional activity of ENaC, usually suggesting an impaired folding and/or assembly in the endoplasmic reticulum (ER) compartment, as shown, for example, for the CFTR⌬F508 mutant (18).
To study the functional role(s) of CRD1 and CRD2 of ENaC, we replaced the conserved cysteine residues with either serine or alanine. Functional and cell surface expression of the mutants show that replacement of two cysteines from the CRD1 of ␣, ␤, and ␥ ENaC subunits and two cysteines from the CRD2 of the ␣ and ␤ but not the ␥ subunit prevents the normal transport of assembled channels from intracellular compartments to the plasma membrane, thereby decreasing the number of ENaC molecules at the cell surface.

EXPERIMENTAL PROCEDURES
Site-directed Mutagenesis-Cysteine to serine or alanine mutations were introduced into the rat ␣, ␤, and ␥ ENaC cDNAs by the polymerase chain reaction-based approach. The relative positions of the conserved cysteines along the extracellular loops of ␣, ␤, and ␥ subunits are shown in Fig. 1. Fig. 1 shows also an abbreviated numbering of the mutated cysteines used in this study to simplify the presentation of the data. The corresponding mutations are: has been mutated to an alanine to avoid the formation of a consensus sequence for N-linked glycosylation. The rat ␣C158 (␣C1) residue corresponds to the human ␣C133 described as causing the PHA-1 phenotype when mutated to Tyr (␣C133Y). All the ␤ and ␥ subunit mutations were introduced into the FLAG epitope-tagged ␤ and ␥ cDNAs, allowing determination of the cell-surface expressing ENaC using a binding analysis (see below). All mutations were confirmed by the dideoxynucleotide sequencing method on both strands.
Electrophysiology and Binding Analysis-Complementary RNAs of each ␣␤␥ ENaC subunits and the corresponding mutants were synthesized in vitro, and equal amounts of wild type (wt) or mutated subunit cRNA at saturating concentrations for maximal channel expression (9 ng of total cRNA) were injected into Xenopus oocytes. Macroscopic whole oocyte Na ϩ currents (I Na ) resulting from ENaC expression were measured using the two-electrode voltage-clamp technique and determined as the amiloride-sensitive inward current at Ϫ100 mV holding potential. For both macroscopic current as well as binding analysis, the oocytes were kept after injection into the low Na ϩ modified Barth saline (MBS: 10 mM NaCl, 90 mM NMDG-Cl, 5 mM KCl, 0.41 mM CaCl 2 , 0.33 mM CaNO 3 , 0.82 mM MgSO 4 ).
Temperature Dependence-Under standard conditions, the incubation temperature is kept at 19°C. However, in our recent study of human ␣ENaC C133Y mutation, this temperature has been demonstrated to be permissive for the functional expression of the mutant channel proteins that normally would not be expressed at the cell membrane when the incubation temperature was raised to 30°C (17). To test for the temperature dependence of the cysteine mutants activity into the rat sequence, we took advantage of the rapid expression of ENaC in the Xenopus oocytes system, in which significant channel activity is expressed as soon as 6 -8 h after cRNA injection. Under standard physiological temperature (19°C), the I Na of wt ␣␤␥ ENaC was typically between 10 and 15 A/oocyte. Preliminary experiments showed that the amiloride-sensitive sodium transport was temperaturedependent and increased by 50 -80% when the temperature bath was raised to 30°C, a maximum for a cell from a poikilotherm animal. Temperature sensitivity was therefore tested by comparing functional expression of wt and mutated channel protein at 19 versus 30°C.
Cell Surface Expression of ENaC-The cell surface expression of ENaC was determined by specific binding of [ 125 I]M 2 IgG 1 (M 2 Ab) to oocytes expressing FLAG-tagged ␤ and ␥ subunits as described previously (19). Anti-FLAG M 2 monoclonal antibody was iodinated using the IODO-BEADS Iodination reagent (Pierce) and carrier-free Na 125 I (Amersham), according to the Pierce protocol. Iodinated antibody had a specific activity of 5-20 ϫ 10 17 cpm/mol. Binding of the iodinated antibody to oocytes expressing the FLAG-containing wild-type or mutant ␤ and ␥ ENaC subunits was determined 16 -20 h after the cRNA injection. Twelve oocytes in each experimental group were transferred into a 2-ml Eppendorf tube containing the low sodium MBS supplemented with 10% heat-inactivated calf serum and incubated for 30 min on ice. The binding was started by the addition of 12 nM of the iodinated antibody in a final volume of 100 l. After 1 h incubation on ice, the oocytes were washed eight times with 1 ml of low sodium MBS supplemented with 5% calf serum and then transferred individually into tubes containing 250 l of low sodium MBS. The samples were counted and the same oocytes utilized for two-electrode voltage-clamp studies. Nonspecific binding was determined on oocytes injected with untagged ENaC subunits.
Immunoprecipitation-Injected with wild-type or mutant ENaC subunit oocytes were incubated in modified Bart's medium containing 0.8 mCi/ml [ 35 S]methionine (NEN Life Science Products) for the times indicated in the figure legends and then subjected to different chase periods in the presence of 10 mM cold methionine. Microsomes were prepared as described (20) and the immunoprecipitations were performed as described (21) under nondenaturing conditions, resolved by 5-8% gradient SDS-PAGE and revealed by fluorography. The polyclonal antibodies used were raised against the amino acids: 10 to 77 of the rat ␣ENaC subunit, 559 to 636 of the rat ␤ENaC subunit, and 570 to 650 of the rat ␥ENaC subunit (2). were expressed with wt FLAG-tagged ␤ and ␥ subunits in Xenopus oocytes to test for channel functional activity as well as for cell-surface expression at 19°C. As shown in Fig. 2, several cysteine mutants behave differently, both with respect to channel activity and to cell surface expression. In the CRD1, the mutations ␣C1S and ␣C6S are characterized by a low macroscopic Na ϩ current (I Na ) ( Fig. 2A), as well as a decrease in cell surface expression (Fig. 2B). For these mutants, I Na decreased significantly (35 Ϯ 5%, n ϭ 46, p Ͻ 0.001 and 25 Ϯ 4%, n ϭ 38, p Ͻ 0.001 of wt channel activity, respectively). Binding analysis showed a corresponding decrease in cell surface expression (21 Ϯ 5%, n ϭ 39, p Ͻ 0.001 and 20 Ϯ 10%, n ϭ 32, p Ͻ 0.001 of wt channel cell surface expression, respectively). Mutants of other cysteines from the CRD1 (␣C2S to ␣C5S) were indistinguishable from wild type channels (Fig. 2, A  and B).
Functional Expression of Mutant ENaC Is Temperature-dependent-Using the assay described under "Experimental Procedures," we tested the temperature sensitivity of wt and mutant channels at 30°C versus 19°C. At 30°C all mutants in CRD1 (excepting ␣C4S) showed a significant decrease in I Na (Fig. 3A). The activity (I Na ) of ␣C1S and ␣C6S mutants further decreased to near undetectable levels (5 Ϯ 1% and 2.0 Ϯ 0.5% of wt channel activity, respectively). In CRD2, no significant activity could be detected for the ␣C11S mutant (0.4 Ϯ 0.1%, n ϭ 16, p Ͻ 0.001) and ␣C12S mutant (0.2 Ϯ 0.1%, n ϭ 15, p Ͻ 0.001), respectively, of wt channel activity. For the other CRD2 mutants, a significant decrease in activity as compared with the 19°C incubation condition was observed (Fig. 3B). Only the ␣C7S mutant maintained a significant increase in channel activity (201 Ϯ 27%, n ϭ 15, p Ͻ 0.005) at 30°C compared with the wt channel.
␣ENaC Cysteine Mutants and the Intrinsic Channel Properties-In Fig. 2, C and F, the amiloride-sensitive sodium current (I Na )/cell surface expression (fmol) ratios are shown. If a mutation decreases I Na without changing the current/binding ratio, one can deduce that channel delivery to the cell surface has been impeded with no change in its intrinsic activity. Such was the case for ␣C1S and ␣C6S in CRD1 (Fig. 2C) and ␣C11S and ␣C12S in CRD2 (Fig. 2F). If a mutation increases the current/ binding ratio, one can deduce that the intrinsic channel activity has increased. An example of this situation is the ␣C7S mutant (98.0 Ϯ 14.3 A/fmol, n ϭ 27, p Ͻ 0.001) and, to a lesser extent, ␣C8S (61.9 Ϯ 8.5 A/fmol, n ϭ 29, p Ͻ 0.001), ␣C13S (44.2 Ϯ 6.1 A/fmol, n ϭ 27, p Ͻ 0.001), ␣C15S (50.5 Ϯ 9.5 A/fmol, n ϭ 29, p Ͻ 0.005), and ␣C16S (69.9 Ϯ 5.3 A/fmol, n ϭ 29, p Ͻ 0.001), compared with wt channel current/binding ratio (22.5 Ϯ 2.0 A/fmol, n ϭ 30) (Fig. 2F). Changes in several biophysical properties of the channel (ion conductance, gating, ion selectivity) could account for the differences in intrinsic channel activity (see discussion in Ref. 22). In our experimental conditions, it could be either the channel unitary conductance (gNa) or the mean open probability (P o ). In addition, alteration of affinity to amiloride could influence the correct determination of I Na in loss of function mutations. To distinguish between these possibilities, we measured the apparent inhibitory constant for amiloride (K i ) and the unitary conductance for the ␣C1S, ␣C6S, ␣C11S, and ␣C12S as well as for the ␣C7S mutant. These experiments showed that neither K i to amiloride nor unitary conductance were affected by these mutations (data not shown). In addition, measurements of ion selectivity did not reveal any differences between wt and mutated channels. The order of ion permeability (Li ϩ Ͼ Na ϩ Ͼ ϾϾ Ͼ K ϩ ), characteristic of ENaC, was respected (data not shown).
Interestingly differential sensitivity to these agents of wt and mutant channels. The affinity to amiloride is not changed and the I Na have an inhibition rate (ϳ20% with MTSEA and ϳ60% with pyrene-7-MTS) similar for the wild type and all the 16 ␣ENaC mutants (data not shown). These experiments suggest that in the threedimensional structure of ENaC, positions of the cysteines are distant enough from the channel pore and amino acids determining the binding of amiloride.
Finally, the studies of activity decreasing mutants indicate that the delivery of the ␣C1S, ␣C6S, ␣C11S, and ␣C12S containing channels to the plasma membrane is severely impeded at 19°C and even more at 30°C. In contrast, incubation at 30°C abolishes (␣C8S, ␣C10S, ␣C13S, ␣C15S, and ␣C16S) or decreases (␣C7S) the increased activity of these mutants. We have therefore focused our studies on the mutants where the functional effects of the mutations were more pronounced with increased temperature.
Functional Properties of Double Cysteine Mutants-Various double mutants (␣C1/C6S, ␣C6/C11S, ␣C11/C12S, ␣C1/C11S, ␣C1/C12S, ␣C6/C12S) were constructed in order to verify whether these mutations affect the same functional domain of the channel (CRD1 or CRD2) or not. Additivity of the functional effect should indicate that the two cysteines are in two distinct functional domains. Fig. 4 shows that this was not the case for the ␣C1/C6S and ␣C11/C12S double mutants. Within CRD1 the double mutant ␣C1S/C6S is not functionally different from the single ␣C1S and ␣C6S mutants. Within CRD2, the double mutant ␣C11S/C12S was not additive but rather more active than the ␣C11S and ␣C12S mutants alone (Fig. 4). In contrast, the double mutations across CRD1 and CRD2 (␣C1S/C11S, ␣C1S/C12S, ␣C6S/C11S, and ␣C6S/C12S) show strong additivity and are no longer functional (Fig. 4). These data indicate that the Cys 1 and Cys 6 in CRD1 on the one hand and the Cys 11 and Cys 12 in CRD2 on the other hand participate in the formation of two different functional domains of the ␣ENaC subunit. Inactivation of both leads to complete loss of channel activity.
Functional Effects of the C1S, C6S, C11S, and C12S Mutations in the ␤ and ␥ ENaC Subunits-The conservation of the extracellular cysteines among all the members of ENaC/DEG family suggests the potential importance of the corresponding cysteine residues in the function of ␤ and ␥ ENaC subunits. To verify this hypothesis, we mutated the cysteines in the ␤ and ␥ ENaC subunits corresponding to the ␣ENaC Cys 1 , Cys 6 , Cys 11 , and Cys 12 to either serine or alanine and then expressed them in Xenopus oocytes. The nonpermissive incubation temperature of 30°C was chosen in order to maximize the functional effects. Fig. 5A demonstrates that the ␤C1S and ␤C6S mutations cause similar effects compared with the ␣C1S and ␣C6S mutants. The ␥C1S and ␥C6S mutations provoked an even more dramatic decrease in channel activity with almost no detectable I Na . The oocytes injected with triple ␣C1S/␤C1S/␥C1S and ␣C6S/␤C6S/␥C6S mutant subunits have no detectable I Na .
These data indicate a similar functional role for Cys 1  in each of the three ENaC subunits in CRD1. In CRD2, mutation of the Cys 11 and Cys 12 in the ␤ and ␥ subunits revealed a different pattern of expression. As shown in Fig. 5B, the effect of the double ␤C11S/C12S mutant is similar to that of the ␣C11S/C12S mutant. In contrast, the double ␥C11S/C12A mutant did not significantly alter the functional activity. Under nonpermissive condition (30°C, 8 h of incubation) the activity of ␣␤, ␣␤C11S/C12S, ␣␥, ␣␥C11S/C12S channels was undetectable (Fig. 5B). Mutation of the Cys 11 and Cys 12 in all three ␣␤␥ ENaC subunits led to unfunctional channels. These data suggest an asymmetric role of cysteines 11 and 12 in the ␣, ␤, and ␥ subunits. The absence of any detectable I Na even after 76 h of expression of the triple ␣C1S-C6S/␤C1S-C6S/␥C1S-C6S and ␣C11S-C12S/␤C11S-C12S/␥C11S-C12A mutants (Fig. 5C) also shows that the cause of the impaired cell surface expression is not due to a slow kinetic of intracellular transport but rather to a complete blockade of routing to the cell surface.
Protein Assembly and Stability of Mutant ␣, ␤, and ␥ ENaC Subunits-The possible mechanisms which could explain the decreased cell surface expression of the cysteine mutants are the following: (i) misfolding during translation, (ii) lack of proper oligomerization due to abnormal conformations of assembled complexes, (iii) disruption of the normal transport of the mutant proteins from ER to Golgi and/or from Golgi to the cell surface, or (iv) increased rate of degradation of the mutant proteins leading to a decrease of the total channel protein pool.

DISCUSSION
One of the distinct features of the ENaC/DEG gene family is its membrane topology which predicts that more than 70% of the protein mass is exposed to the extracellular space which is facing the external compartment of the organism (urine in the distal nephron and feces in the colon or airspace in the lung). This suggests that the extracellular loop has a possible role in specific protein-protein interaction (i.e with extracellular matrix proteins in the case of degenerins (23)) and/or ligandreceptor interaction.
The extracellular loop of ENaC/DEG channel family proteins is characterized by the presence of two or three cysteine-rich domains (CRDs). It has been demonstrated that during protein folding and maturation, extracellular cysteine residues are rapidly oxidized, and that enzyme-catalyzed disulfide exchange continues until the most thermodynamically stable conformation is reached. Among the transport proteins, the formation of disulfide bonds in the extracellular loop or N and C terminus of the protein is usually critical for proper co-translational folding of the protein and subsequent assembly and oligomerization in the ER compartment. For instance, mutation of conserved cysteines forming disulfide bonds in the ectodomain of the ␤ subunit of Na,K-ATPase prevents assembly with the ␣ subunit without changing the half-life of the ␤ protein (24). Likewise, cysteine bonds in the ectodomain of the low density lipoprotein receptor allow the folding of this protein domain (repeated many times) into an octahedral cage, binding of calcium molecules, required for apoE and apoB binding and receptor recycling (25). Mutations of one or two cysteines forming a disulfide bond in the N termini of either the ␣ or ␤ subunit of the acetylcholin receptor abolish channel subunit assembly and ligand binding (26).
For the ENaC/DEG channel family, any cysteine of a given CRD could make disulfide bonds with a cysteine located in the same CRD, with a CRD on the same subunit (intrasubunit bonding), or with a cysteine of CRD located on another subunit (intersubunit bonding). However, the intersubunit bonding for these proteins is unlikely for the following reasons: (i) immunoprecipitation of ENaC subunits under denaturing and nonreducing conditions did not reveal intersubunit association in the A6 cells (27), Madin-Darby canine kidney cells (13), Chinese hamster ovary cells, and Xenopus oocytes 2 ; (ii) FaNaC, which is a homotetramer for which no evidence for intersubunit disulfide bonding could be obtained (5). This indicates that the most probable structural feature of the conserved cysteines is the formation of the intrasubunit disulfide bonds. However, if all of the 16 cysteines are involved in disulfide bonds within the ENaC subunits, an enormous number of permutations of disulfides is possible.
Several approaches have been used in order to determine the exact pattern of disulfide bond formation for membrane proteins. A possible biochemical approach uses the differential electrophoretic mobility by SDS-PAGE of the wild-type and the mutant proteins in their nonreduced forms. However, for ENaC, the shift in the electrophoretic mobility between the wild-type channel in its reduced and nonreduced state was not high enough in our experiments to hope to identify as much as eight possible disulfide bonds (data not shown). The similar small shift between the reduced and nonreduced forms was also observed for in vitro translated ␣ENaC (28). Another broadly used approach involves the modification of free cysteines liberated after the mutation of their disulfide bond counterpart using the sulfhydryl reagents. Analysis of differential sensitivity of wt and mutants to these agents helped to identify the cysteine bond patterns in several studies (29). However, our experiments with two MTS reagents have demonstrated identical sensitivity of wt and mutant channels to these reagents. Another tested approach is based on an assumption that the effect of Cys 3 Ser or Cys 3 Ala mutations is limited to the elimination of disulfide bonds which involved that Cys residue. Therefore, mutation of either or both Cys involved in a given disulfide bond results in channel mutants with the same functional properties. For this, double cysteine mutants with mutated Cys 1 , Cys 6 , Cys 11 , and Cys 12 in ␣ENaC were constructed. The similarity in decrease in I Na , sensitivity to temperature, and absence of additivity in the loss-of-function, suggested that the Cys 1 and Cys 6 , as well as Cys 11 and Cys 12 are pairs forming two disulfide bonds in each ENaC subunit.
Among 16 mutated cysteines in the ␣ENaC subunit, four (Cys 1 , Cys 6 from CRD1, and Cys 11 and Cys 12 from CRD2) have demonstrated the critical importance in functional expression of ENaC. The parallel decrease in cell surface expression (Fig.  2, B and E) and the lack of effect of the mutations on intrinsic channel properties indicate that the primary cause of the impaired channel activity for the mutants is the routing of the channels to the cell surface. The temperature sensitivity of the mutants provides an additional indirect evidence that the native protein structures formed with participation of the cysteines are necessary for the proper accomplishment of one or several steps in ENaC processing. In this study, we have considered two possibilities to explain the decreased number of ENaCs containing the mutant subunits at the cell surface: (i) affected assembly of the channel subunits and (ii) their increased rate of degradation. However, metabolic labeling studies have demonstrated that neither assembly nor degradation rate were affected in the mutant subunits (Figs. 6 and 7).
The important role of Cys 1 in ENaC function is supported by the recent report of a novel mutation causing PHA-1, a severe renal salt loosing syndrome caused by loss-of-function mutations of ENaC subunits. The recently reported novel mutation is a cysteine from the CRD1 in human ␣ENaC subunit (␣C133Y) corresponding to Cys 1 (Cys 158 ) in the rat sequence. The temperature sensitivity of ␣C1S mutation correlates well with the temperature sensitivity of the human counterpart (␣C158Y), described by Gruender et al. (17) and with the severity of the clinical phenotype. 3 Similarly, the ␣, ␤, and ␥ C6S 2 D. Firsov, unpublished data. 3 R. Lifton, personal communication. and C11S, and ␣ and ␤ C12S mutants, when incubated at 19°C, expressed the significantly reduced I Na as compared with the wt channel (Fig. 2D), and demonstrated almost no activity when incubated at 30°C (Fig. 3B). These results are reminiscent of a loss of function mutation observed most frequently in cystic fibrosis (CFTR⌬F508). This mutant is temperature-sensitive in vitro and, decreasing the incubation temperature to 20°C, rescues channel activity to a significant level (18). The mechanism described for the CFTR⌬F508 is, however, apparently distinct from that reported here. For the CFTR⌬F508 mutant, the corresponding protein failed to traffic to the plasma membrane, because it was stuck in the pre-Golgi compartment and degraded in a ER/proteasome compartment. Loss of function of the water channel aquaporin 2 causes nephrogenic insipidus diabetes. Studies of the molecular cause of aquaporin 2 dysfunction in the aquaporin 2 E258K mutant revealed that an assembled tetrameric channel containing the mutant subunits is retained in the Golgi complex, probably due to the introduction by this mutation of a Golgi retention signal (30). We may have a similar situation for the C1S mutant but, in the present study, we did not look for the precise intracellular localization of the mutant protein. However, the correct assembly and the normal glycosylation patterns of the mutated subunits suggest that a similar mechanism is likely.
Interestingly, despite of the absolute conservation of the Cys 11 and Cys 12 among ␣, ␤, and ␥ ENaC subunits, the ␥C11S-C12A mutant did not decrease channel function at 19°C nor at 30°C, while the corresponding pairs of the ␤ and ␣ subunits drastically decreased channel function in a temperature-sensitive manner. This functional difference could be a consequence of either the preferential assembly order of ENaC subunits where assembly of the ␥ subunit occurs in a latter step which is not critical to complete the assembly or, of the different roles of ENaC subunits in trafficking of assembled channels to the cell surface or, of the differential conformational changes caused by these mutations and allowing for the channels containing ␥ but not ␣ and ␤ mutated subunits to reach the cell surface. For ENaC, the functional differences between subunits have already been observed for (i) a differential cell surface expression of the different subunit combinations: ␣␤␥ Ͼ ␣␤ ϭ ␣␥ Ͼ Ͼ ␣ with no expression for ␤␥, ␤, and ␥ injected oocytes (19) or, (ii) the differential contribution of each subunit in functional domains such as the amiloride binding and the sensitivity to divalent cations (7). In other systems, such a difference has been described, for example, for the AChR subunits, where assembly of the channel is blocked at different steps depending on which conserved subunit the cysteine is eliminated (26). The different behavior of the ␥ subunit mutants gives us an additional example of the asymmetric role of each ENaC subunits in the function of the heterotetrameric protein.
The absence of functional effects of other cysteine mutations (12) could signify that they do not participate in the channel cellular trafficking and/or in formation of protein structures necessary for intrinsic channel activity. Therefore, a possible function of the cysteines in formation of a ligand-binding site(s) can be considered. In the absence of identified extracellular ligands for ENaC, the proteins interacting with ENaC from the extracellular side can be either the proteins of the extracellular matrix or channel modulating proteins. The only known extracellular modulators of ENaC activity are the serine proteases CAP1 (31) and trypsin (32). Both enzymes act through the extracellular side and increase ENaC activity as high as 2 to 10 times without changing the number of channels expressed at the cell surface. We tested the effect of trypsin on all the 16 ␣ENaC mutants and did not find a difference from the wildtype channel in the level of stimulation of channel activity (data not shown). However, elimination of more than one cysteine might be necessary to abolish the possible protein-protein interaction. Finally, we cannot exclude the presence of an as yet unknown extracellular ENaC ligand function which cannot be identified with the Xenopus oocyte expression system.
To conclude, this study identifies a pair of cysteines in the CRDI of ␣, ␤, and ␥ ENaC subunits and a pair of cysteines in the CRDII of ␣ and ␤ subunits, essential for the channel routing to the plasma membrane. It also establishes the molecular cause of human ␣C133S mutation causing the pseudohypoaldosteronism type I pathophysiology, which results in a reduced number of ENaCs at the cell surface.