Biochemical Analysis of the Membrane Topology of the Amiloride-sensitive Na' Channel*

A key protein component of the amiloride-sensitive sodium channel has been cloned from rat colon and human lung. It may represent the first member of a new family of ionic channels expressed from nematode to human. The biochemical properties of the rat protein, a 699 amino acids long polypeptide, have been analyzed. Four polyclonal antibodies raised against distinct parts of the channel immunoprecipitated a glycosylated pro- tein of 96 kDa after cRNA expression in oocytes as well as after in vitro translation. When expressed alone into oocytes, the protein was not stable; most of it remains stacked into the endoplasmic reticulum. This results in a very low yield of complete maturation of the protein at the cell surface after expression from the pure cRNA. To determine the membrane topology of the protein, in vitro translation by a rabbit reticulocyte lysate was per- formed followed by insertion into canine pancreatic microsomes and protease digestion. Analysis revealed a model with only two transmembrane a helices and a large extracellular domain of about 500 amino acids. The NH, and COOH termini are cytoplasmic. Protease digestion results suggest the possible presence of a structural element that could have a function similar to that of the Hi5 segment in K+ channels. The model indi- cates that there is no cytoplasmic site for protein kinase A phosphorylation. The well known regulation of the channel activity by hormones that

A key protein component of the amiloride-sensitive sodium channel has been cloned from rat colon and human lung. It may represent the first member of a new family of ionic channels expressed from nematode to human. The biochemical properties of the rat protein, a 699 amino acids long polypeptide, have been analyzed. Four polyclonal antibodies raised against distinct parts of the channel immunoprecipitated a glycosylated protein of 96 kDa after cRNA expression in oocytes as well as after in vitro translation. When expressed alone into oocytes, the protein was not stable; most of it remains stacked into the endoplasmic reticulum. This results in a very low yield of complete maturation of the protein at the cell surface after expression from the pure cRNA. To determine the membrane topology of the protein, in vitro translation by a rabbit reticulocyte lysate was performed followed by insertion into canine pancreatic microsomes and protease digestion. Analysis revealed a model with only two transmembrane a helices and a large extracellular domain of about 500 amino acids.
The NH, and COOH termini are cytoplasmic. Protease digestion results suggest the possible presence of a structural element that could have a function similar to that of the Hi5 segment in K+ channels. The model indicates that there is no cytoplasmic site for protein kinase A phosphorylation. The well known regulation of the channel activity by hormones that activate this kinase such as vasopressin might thus be situated on another channel component.
Amiloride-sensitive Na' channels are present in a variety of epithelia (Refs. 1-4 and reviewed in Ref. 5). The main subunit of the Na' channel protein has recently been cloned from rat colon and called RCNaCh' (or arENaCh) (6,7). The corresponding human lung sequence has also been established (8). Injection of the corresponding cRNAs into Xenopus oocytes induces an amiloride-sensitive Na' current, with the selectivity and the pharmacological properties of the native epithelial Na' channel. The primary sequence of the Na' channel protein has no homology with previously cloned channels. The purpose of this The abbreviations used are: RCNaCh, rat colon Na' channel; PAGE, polyacrylamide gel electrophoresis; pAbl, pAb2, pAb3, pAb4, polyclonal antibodies raised against RCNaCh; Endo H, endoglycosidase H. paper is to analyze the structural properties of the RCNaCh protein. I n vitro translation, protease digestions, and immunoprecipitation have provided a picture of the topology of the RCNaCh protein.

EXPERIMENTAL PROCEDURES
RNA I)-anscription-cRNA was transcribed in uitro with T7 RNA polymerase as previously described (7). The RCNaCh plasmid was linearized with AccI, MscI, and BspEI to get 3' truncated cRNAs. The COOH-terminal-deleted proteins were designated as C438, C521, and C615, respectively, the numbers 438,521, and 615 corresponding to the last encoded residues (7). The excision of anEcoRI-EcoRI fragment in 5' of the cDNA and recirculation of the plasmid have permitted a T7 RKA polymerase transcription of a 5"deleted form of RCNaCh. This last construction, called N108, was translated into protein from MeVo8 to the end.
Oocytes-Oocytes were prepared and injected according to Ref the RCNaCh sequence are predicted to be highly antigenic (9). They were linked to keyhole limpet hemocyanin and injected to rabbits according to standard protocols (10). An expression protein of 23 kDa corresponding to the COOH-terminal part of the RCNaCh protein (from Ileu5I0 to the COOH terminus) was obtained d e r induction of BLZl(DE3) bacteria transformed with a pET.llc vector (Novagen) containing the 1632-221-base pair fragment of RCNaCh (generated by a digestion by BcZI and BamHI). The protein was purified by electroelution, checked by NH,-terminal sequencing, and injected into rabbits. Sera were controlled monthly by enzyme-linked immunosorbent assay against pure peptides. Positive ones were those producing immunoprecipitation of the in uitro translated protein. For immunoprecipitation, the labeled material was solubilized in a buffer containing 1% Triton X-100,0.5% deoxycholate, 140 m~ NaC1,ZO m~ TridCl, pH 7.4,40 pg/ml phenylmethylsulfonyl fluoride, 10 pg/ml leupeptin, 1 1.1~ pepstatin A. After a l-h preclearing with pansorbin A (Calbiochem), the soluble extract was incubated for 1 h with antibodies diluted 100-fold and then precipitated with protein A-Sepharose CL-4B (Sigma). The beads were extensively washed with the same buffer. The immunoprecipitated material was then denatured into l% SDS, 4% 2-mercaptoethanol at 95 "C for 5 min and analyzed by SDS-PAGE. Cell-free I)-unsZation+RNAwas added to a rabbit reticulocyte lysate mixture (Promega Corp.), containing [35Slmethionine, according to the manufacturer's protocol, except that the lysate content was lowered to 40% final volume. For microsome insertion, 10 units of dog pancreatic microsomal membranes (Promega Corp.) were added at the start of the translation. For deletion mutants experiments, RNAs transcribed in uitro after linearization of the plasmid with AccI, MscI, and BspEI were used, without introduction of stop codons. This protocol was sufficient to analyze COOH-deleted mutants despite an increase in background.
Protease Digestion-Proteases were added to aliquots of the rabbit reticulocyte lysate translation mixture (final concentration, 100 pg/ml), in the presence or absence of 1% Triton X-100 and incubated for 30 min at 4 "C with trypsin or proteinase K or for 1 h at 37 "C with carboxypeptidase Y. Protease action was stopped by saturating concentrations of inhibitors (400 pg/ml phenylmethylsulfonyl fluoride and 100 pg/ml leupeptin). Some samples were treated with endoglycosidase H (Endo H) after denaturation in 1% SDS at 95 "C for 5 min (Boehringer Mannheim, 1  37 "C). Samples were then diluted into the gel loading buffer, denatured a t 95 "C, and analyzed by SDS-PAGE. After migration, gels were stained with Coomassie Blue, treated with 1 M sodium-salicylate for fluorography, dried under vacuum, and finally exposed a t -70 "C for 12 h to 7 days.

RESULTS AND DISCUSSION
We first intended to establish the topology of the Na' channel protein in the oocyte membrane, where the channel activity was initially characterized after expression of its cRNA (6,7). However, expression of RCNaCh in Xenopus oocytes results in a relatively low level of amiloride-sensitive Na' current (6,7). The neosynthesized Na' channel protein was identified after metabolic labeling with [35S]methionine of RCNaCh-injected oocytes; polyclonal antibodies raised against fragments of the protein specifically immunoprecipitated a protein of 96 kDa ( Fig. 1). After in vitro translation in the presence of rabbit reticulocyte lysate and canine pancreatic microsomes, the same antibodies precipitated two proteins of 96 and 78 kDa. The 78-kDa protein is the unglycosylated form of the protein (71, because i t was also observed when the rabbit reticulocyte lysate was incubated in the absence of canine pancreatic microsomes (see also Fig. 3A). The 96-kDa protein is sensitive to Endo H, a n enzyme that cleaves high mannose oligosaccharides from glycoproteins (11). When the immunoprecipitated material from in vitro translation or from oocytes was treated with this enzyme, the 96-kDa protein disappeared and was replaced by the 78-kDa protein (Fig. 1). The sensitivity to Endo H revealed (i) that the protein is N-glycosylated and (ii) that the 96-kDa protein is probably not completely processed, because complete glycosylation usually leads to Endo H-resistant forms of glycoproteins (12). The 96-kDa protein thus does not correspond to the mature form of the channel protein but rather to an intermediate form, most probably stacked into the endoplasmic reticulum, where it will be finally degraded. Analysis of the turnover of the 96-kDa protein supports that view. In oocytes, the half-life of the 96-kDa protein is -2 h and is not changed by treatments with known blockers of lysosomal degradation (13) such as chloroquine (800 w), NH,Cl (10 mM), brefeldin A (10 PM), or leupeptin (400 pg/ml) (not shown). The degradation of the 96-kDa protein is thus likely to take place in the endoplasmic reticulum, as previously reported for the asialoglycoprotein receptor (13) and the T cell antigen receptor (14). In these two cases of multi-subunit membrane proteins, several of the subunits have to be co-expressed to stabilize each other. When one subunit is lacking, degradation of the others takes place in the endoplasmic reticulum (13,14). In the case of the amiloridesensitive Na+ channel, it is expected that other subunits are necessary for efficient expression at the cell surface. Our current hypothesis is then that a low proportion of the RCNaCh protein reaches full maturation when RCNaCh cRNA is expressed alone, leading to a low level of functional Na+ channels (6,7).
The amount of mature form was thus too low to permit the analysis of its topology with protocols applicable in situ. Be- Kyte and Doolittle (16) hydropathy profiles of rat colon and human lung Na' channels. The window for hydrophobicity analysis is n = 19 residues (16). B , description of the different constructs used in this study. Boxes indicate the hydrophobic domains, M, the putative first methionine, and V, the potential N-glycosylation sites. "Potential" glycosylation sites at AsnW and Asngl are not shown, because they are located before the first hydrophobic domain (intracellular).
cause the protein insertion in the membrane is expected to remain unchanged during processing of RCNaCh, we analyzed the topology of the RCNaCh form detected during in vitro translation experiments. This was done in the presence of rabbit reticulocyte lysate and canine pancreatic microsomes (15). The RCNaCh sequence contains two hydrophobic domains of more than 20 residues (located between residues 110 and 139 and between residues 588 and 613) characterized by a Kyte and Doolittle (16) index of more than 1.6 ( Fig. 2A). One simple hypothesis is that they are the only transmembrane domains in the RCNaCh structure. To test that model, different RCNaCh constructs, corresponding to different truncated forms of the protein, were obtained by removal of either one NH,-terminal segment, leading to the construct called N108, or three COOHterminal segments leading to constructs called C438, C521, and C615 (numbers are those of the extreme residue (7), MetIos for N108, Lys438 for C438, T r p S z 1 for C521, for C615). Fig. 2B presents the different constructs used in this work. Translation of the truncated form N108 begins just before the first putative transmembrane domain. Deletions in C438 and C521 remove the second putative transmembrane domain, whereas deletion in C615 occurs immediately after it. The five constructs were translated in vitro in the presence of canine pancreatic microsomes. Treatment by proteinase K revealed fragments that were protected from degradation because of their localization within the lumen of microsomes. Endo H treatment was used to evaluate the extent of the N-glycosylation for each fragment.
Four distinct polyclonal antibodies were raised against four distinct parts of the RCNaCh protein. pAbl was raised against a peptide located between residue 44 and residue 58 (i.e. before the first putative transmembrane segment). pAb2 was raised against a peptide located between residue 167 and residue 187 (i.e. between the two putative transmembrane segments). pAb3 was raised against a peptide located between residue 625 and residue 636 (located after the second putative transmembrane segment). PAM was raised against a protein produced in Escherichia coli and corresponding to the COOH-terminal part of the protein (beginning before the second putative transmembrane segment, at the presence of canine pancreatic microsomes, two major products were observed at 96 and 78 kDa (Fig. 3 A , Total ). These two bands were immunoprecipitated by the four antibodies (Fig.   3A ). When the total translation product was treated with proteinase K, a fragment of 70 kDa was found to be protected from protease degradation (proteinase K, Total, Fig. 3A). pAbl, raised against a peptide located before the first putative transmembrane segment, and pAb3, raised against a peptide locatedafter the second putative transmembrane segment, failed to recognize the 70-kDa peptide. The same observation was made with pAb4, an antibody raised against the COOH-terminal segment of RCNaCh. pAb2 was the only serum that could immunoprecipitate this 70-kDa peptide. The segment between the two transmembrane domains is thus likely to be located into the microsomes, whereas the NH, and COOH termini of the protein probably lie outside the microsomal lumen (Fig. 3B).
The absence of reactivity of pAb4 against the 70-kDa peptide was puzzling at first, because this peptide contains a fragment that is located between the two potential transmembrane domains ( Fig. 2A). The most probable interpretation is that PAM is directed specifically against the COOH-terminal region of the immunogen. Other bands were observed in the total translation product. Because several of them were recognized by pAb2, pAb3, and pAb4 but not by pAbl, whose epitope is located very near the NH, terminus, they probably correspond to translation products beginning at some internal methionine (false starts). For convenience, the following part of the paper deals with the total translation product. Only the highest molecular weight corresponding to the protein translated from the first methionine was considered for structural analyses. The apparent M, then shifted to 50,000, whereas the calculated molecular mass of the peptide located between residue 110 and residue 613 is 58 kDa (Fig. 4B). The difference between observed and calculated M, values (8 kDa) could be due to abnormal mobility of the polypeptide in SDS-PAGE. Another possibility is that the peptide obtained after proteolysis is shorter than the 110-613 fragment.
The protection from proteolysis of the 50-and 70-kDa fragments is probably due to microsomal incorporation, because it was not observed when proteinase K treatment was carried out on microsomes solubilized with Triton X-100 or when translation was performed in the absence of microsomes (Fig. 4A).
Carboxypeptidase Y is an exopeptidase starting its catalytic action at the COOH terminus. Its action is known to be stopped at a glycine residue. For RCNaCh, the digestion by carboxypeptidase Y is expected to be stopped by G~Y~'~, resulting in an -2-kDa amputation of the 96-kDa protein. Fig. 4A shows that the mobility of the RCNaCh protein is indeed slightly increased after treatment by carboxypeptidase Y. Because microsomes did not protect the protein against carboxypeptidase Y action, the COOH-terminal part of RCNaCh has to be located outside the microsomal structure, i e . pointing toward the cell cytoplasm. Similar treatments were camed out with the different truncated forms of RCNaCh. Fig. 5A shows the results obtained with N108. N-glycosylated N108 has a molecular weight of 84,000 (65,000 a h r Endo H deglycosylation). N108 shares most of the topological properties of RCNaCh. For instance, (i) its COOH-terminal segment is cytoplasmic, as demonstrated by a shift in its mobility after carboxypeptidase Y treatment; (ii) proteinase K digestion generates a 70-kDa peptide (50 kDa after deglycosylation), as previously observed with the fulllength RCNaCh protein. Thus the removal of the -100 first NH,-terminal residues in NlO8 does not alter the M, of the final fragment protected in microsomes from proteinase K digestion.
The difference between the M, values of N108 before and after treatment with proteinase K represents the M, of the segment(.$ exposed to the protease, i.e. 14,500. Because Met''' is located at the very beginning of the first putative transmembrane segment, one should expect that only the 8.9-kDa fragment going from A r g 6 I 4 to the COOH-terminal end would be digested by proteinase K in N108. A 5.6-kDa difference is observed between this calculated mass and the experimental value. On the other hand, the difference between M, values of RCNaCh and N108 provides the M, of the deleted NH, segment from Met' to Arg107, i.e. 12,500. In that case, there is a very good agreement with the calculated M, of 12.4 kDa. The possibility of an abnormal mobility of these proteins in SDS-PAGE is then excluded. The shifts between theoretical masses of the proteinase K peptide and of the COOH-terminal cytoplasmic peptide would then indicate a transmembrane topology different from the one depicted in Fig. 3B. Fig. 5B summarizes the different results obtained with N108. Three different COOH-terminal-deleted constructs were analyzed and presented in Fig. 6. C615 includes the two hydrophobic segments (Fig. 2) but lacks most of the COOH-terminal part. It has a M, of 87,000 (67,000 after deglycosylation) (Fig. 6). Proteinase K treatment generates a 69-kDa peptide, which moves at 50 kDa after deglycosylation, and which, therefore, has the same characteristics as the peptide generated by proteinase K from RCNaCh and N108. Thus the removal of the 84 last COOH-terminal residues does not alter the size of the fragment protected in microsomes from proteinase K digestion. Moreover, because C615 is sensitive to carboxypeptidase Y, has to be accessible to the exopeptidase and is then cytoplasmic. The difference between the M, values of C615 before and after treatment with proteinase K is 18,000. Because A r g 6 1 5 is located immediately after the second putative transmembrane domain (see the model presented in Fig. 3B), this value is expected to correspond to the size of the NH,terminal segment (from NH, terminus to LyslOg, M, = 12,500).

Topology of a New Q p e of Ionic Channel
In that case, as with RCNaCh and N108, there is a difference of 5.5 kDa between the calculated mass of the exposed segment (12.5 kDa) and the experimental M, (18,000).
Properties of (2438 and C521 are analyzed in Fig. 6. Proteinase K treatment generated a 53-kDa peptide (39 kDa after deglycosylation) from C438 (Fig. 6) and a 58-kDa peptide (46 kDa after deglycosylation) from C521 (Fig. 6). The calculated molecular mass of the peptide (located between residue 110 and residue 438) generated from C438 after proteinase K digestion is 38 kDa, very close to the observed M, of 39,000 after deglycosylation. Similarly, the calculated molecular mass for the peptide generated from C521, located between residue 110 and residue 521, is 47 kDa, very close to the observed M, value of 46,000. C438 and C521 are not sensitive to carboxypeptidase Y, in good agreement with the model shown in Fig. 3B, which puts their COOH-terminal end within the microsomal lumen. Dif-ferences between M, values before and after treatment with proteinase K are 15,000 for C438 and 13,000 for C521. These values are very close to the calculated mass of 12.5 kDa for the fragment between Met' and Lyslog, the exposed NH,-terminal segment of RCNaCh.
The extracellular domain of RCNaCh (Fig. 2)  nately not possible from the available data to provide a n exact value. The number of N-glycosylated sites in RCNaCh could vary between four and six. The secondary structure of the two hydrophobic segments analyzed according to Ref. 17 suggests that they most likely correspond to a helices. Accumulations of positive charges were observed immediately before the first transmembrane segment (_K,,,HN~MK,,,) as well as immediately after the second transmembrane segment (&,SFRS&,,), as previously reported for many other membrane proteins (18). It is generally believed that segments with an excess of positive charges do not tend to be integrated within the membrane bilayer and that they stabilize adjacent transmembrane segments.
The model, which is presented in Fig. 3B, has cytoplasmic NH,-and COOH-terminal segments (both composed of about 100 residues) and a single large extracellular domain (with at least four N-glycosylation sites).
Inward rectifier K+ channels are other examples of ionic channels that have only two transmembrane a-helical segments (19,20). An additional H5 segment, also called a P segment, is found in all known K+ channels (21, 22). This region constitutes a crucial part of the ionic pore. It was therefore of interest to analyze our own data in light of the K+ channel model. Because all known H5 structures are known to be located in the vicinity of classical transmembrane segments (23), it was logical to think that if such structures exist for the Na' channel they would be located either immediately after the first transmembrane domain or immediately before the second transmembrane domain.
The analysis of our M , data has revealed a number of differences between the calculated values for digestion products and the experimental ones that seem to provide indications that a H5-like structure might also exist in the Na' channel protein. Vasopressin is a well known activator of the epithelial Na' channel. It acts via phosphorylation by a CAMP-dependent protein kinase. Interestingly, Se2.2°4 and Se?05 from the rat colon sequence are consensus sites for phosphorylation by CAMPdependent protein kinase. However the channel topology presented in Figs. 4B and 5B indicates that these two residues have an extracellular localization. Therefore intracellular CAMP-dependent protein kinase probably cannot directly phosphorylate the RCNaCh protein, and phosphorylation induced by vasopressin would have to take place on another protein component, which would either be another subunit of the amiloride-sensitive Na+ channel or perhaps be a distinct cellular component that would indirectly regulate channel activity.
Protein kinase C can act as a negative regulator of Na+ channel activity in A6 cells (24,251. The model presented in Figs. 4B and 5B indicates that Sera' is the only site for protein kinase C that is conserved between rat and human (8). Its localization, near the cytoplasmic side of the first transmembrane a-helix, suggests a possible involvement in the negative regulation of Na+ channel activity by protein kinase C.
The primary structure of RCNaCh as well as its membrane topology clearly define a new ionic channel family. Interestingly, RCNaCh is related to the nematode proteins Mec,, Mec,,, and Deg, (26, 27). These three proteins called degenerins (26) are present in Caenorhabditis elegans mechanosensory neurons. Point mutations in their sequences induce mechanoinsensitivity and degenerescence. Their overall homology with RCNaCh is only of -E%, but the two hydrophobic segments depicted in Fig. 2 are conserved (26). The highest homology between RCNaCh and degenerins has been found around a cysteine-rich domain located extracellularly between Cys4,, and Cys507. Because of their homology with RCNaCh one proposed hypothesis (7,27) for the physiological role of C. elegans degenerins in touch neurons is that they may correspond to mechanosensitive cationic channels. The different parent degenerins (Mec,, Deg,, and Mec,,), all analogous to RCNaCh and with a high homology between them, may be the different subunits forming one single type of ionic channel.