A novel topology model of the human Na(+)/H(+) exchanger isoform 1.

The membrane topology of the human Na(+)/H(+) exchanger isoform 1 (NHE1) was assessed by substituted cysteine accessibility analysis. Eighty-three cysteine residues were individually introduced into a functional cysteineless NHE1, and these mutants were expressed in the exchanger-deficient PS120 cells. The topological disposition of introduced cysteines was determined by labeling with a biotinylated maleimide in the presence or absence of preincubation with the membrane-impermeable sulfhydryl reagent, 2-trimethylammoniumethyl-methanethiosulfonate in streptolysin O-permeabilized or nonpermeabilized cells. We proposed a new model for the topology of NHE1 that is significantly different from the model derived from hydropathy analysis. In this model, NHE1 is composed of 12 transmembrane segments (TMs) with the N and C termini located in the cytosol. The large, last extracellular loop in the membrane domain of the original model was suggested to comprise an intracellular loop, a new transmembrane segment (TM11), and an extracellular loop in the new model. Interestingly, cysteines at 183 and 184 and at 324 and 325 mapped to intracellular loops connecting TMs 4 and 5 (IL2) and TMs 8 and 9 (IL4), respectively, were accessible to sulfhydryl reagents from the outside. Furthermore, exchange activities of two mutants, R180C and Q181C, within IL2 were markedly inhibited by external MTSET. These data suggest that part of IL2 or IL4 may be located in a pore-lining region that is accessible from either side of the membrane and involved in ion transport.

The Na ϩ /H ϩ exchanger isoform 1 (NHE1), 1 which catalyzes an electroneutral exchange of Na ϩ for H ϩ across the plasma membrane, is involved in the regulation of intracellular pH (pH i ) and cell volume (1,2). Until now, six different NHE isoforms were cloned from mammalian tissues (3)(4)(5)(6)(7)(8). Although these isoforms differ in tissue localization, sensitivity of inhib-itors, and mode of regulation, they have a similar overall structure consisting of an N-terminal membrane domain (ϳ500 amino acids) and a C-terminal cytoplasmic domain (ϳ300 amino acids). The former catalyzes an amiloride-sensitive Na ϩ /H ϩ exchange, while the latter functions as a regulatory domain (9).
Previous biochemical and molecular studies have provided important information about the structure-function of the cytoplasmic domain of NHE1 (2). We have recently found that the cytoplasmic domain consists of at least four distinct functional subdomains in terms of pH i sensitivity, some of which are involved in the regulation of the exchanger by cytosolic Ca 2ϩ , protein phosphorylation, and cell ATP (10 -12). On the other hand, little is known about the structure of the N-terminal membrane domain of NHE1, which precludes the clarification of the mechanism for ion exchange and the interaction between the cytoplasmic domain and the ion transport pathway in the membrane domain. The N-terminal membrane domain is highly conserved (70% overall identity) among NHE isoforms and is predicted to span the lipid bilayer 10 -12 times. A 12transmembrane segment (TM) model predicted by hydropathy analysis of NHE1 according to the Kyte-Doolittle algorithm (13) is shown in Fig. 1. The N-terminal region including TM1 is believed without experimental evidence to be cleaved off as a signal peptide, because its amino acid sequence is highly variable among NHE isoforms. In the human NHE1, evidence was presented showing that N-glycosylation occurs in the first putative extracellular loop (14). On the other hand, a recent chymotryptic cleavage experiment suggested that some other extracellular loops in the membrane domain of NHE1 are not fully exposed on the extracellular surface (15). Clearly, more systematic analysis of the topology of the exchanger molecule is required.
In this study, to map the transmembrane topology of NHE1, we determined the accessibility of 83 cysteine residues introduced into a cysteineless form of NHE1 (designated "Cys-less NHE1") to cysteine-directed reagents, biotin maleimide and MTSET, in normal and permeabilized cells (16,17). The former reagent is almost membrane-impermeable and covalently labels cysteine residues with a biotin group that is readily detectable by the use of streptavidin-biotin chemistry. The latter is membrane-impermeable and used to block the biotinylation. Based on the detailed analysis, we proposed a new topology model of NHE1, which consists of 12-transmembrane segments with the N and C termini located in the cytosol. We found that some introduced cysteines within the intracellular loops are also accessible from the outside, suggesting that residues at these positions may be part of the ion transport pathway.

EXPERIMENTAL PROCEDURES
Materials-Biotin maleimide was purchased from Molecular Probes, Inc. (Eugene, OR). MTSET and streptavidin-conjugated agarose were purchased from Toronto Research Chemicals Inc. and Pierce, respec-tively. Streptolysin O (SLO) was obtained from Sigma. The amiloride derivative, 5-(N-ethyl-N-isopropyl)amiloride (EIPA) was a gift from New Drug Research Laboratories of Kanebo, Ltd. (Osaka, Japan). 22 NaCl was purchased from NEN Life Science Products. A NHE1specific polyclonal antibody (RP-cd) was described previously (10). All other chemicals were of the highest purity available.
Cells and Culture Conditions and Stable Expression-The Na ϩ /H ϩ exchanger-deficient cell line (PS120) (18) kindly provided by Dr. J. Pouysségur (Nice, France) and the corresponding transfectants were maintained in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) containing 25 mM NaHCO 3 and supplemented with 7.5% (v/v) fetal calf serum, penicillin (50 units/ml), and streptomycin (50 g/ml). Cells were maintained at 37°C in the presence of 5% CO 2 . PS120 cells (5 ϫ 10 5 cells/100-mm dish) were transfected with each plasmid construct (20 g) by the calcium phosphate co-precipitation technique. Cell populations that stably express mutant NHE1 were selected by a "H ϩ killing" procedure as described (9). Among 89 substituted cysteine mutants, stable transfectants for six mutants (see Fig. 9) were not obtained.
Construction of Na ϩ /H ϩ Exchanger Mutants-The plasmid carrying cDNA coding for the Na ϩ /H ϩ exchanger (NHE1 human isoform) containing some unique restriction sites (pEAP-⌬5ЈAN) was described previously (19). In this study, we introduced several additional unique restriction sites into pEAP-⌬5ЈAN by a polymerase chain reactionbased construction strategy as described previously (11). The resulting plasmid contained multiple unique sites for XhoI, AatII, KpnI, Bsu36I, AccI, SacII, AflII, NdeI, and ApaI at the positions corresponding to amino acids 126, 159, 179, 358, 503, 599, 635, 658, and 698 of NHE1, respectively. For the construction of cDNA for Cys-less NHE1, we synthesized the oligonucleotide sense and antisense primers in which nucleotides corresponding to cysteine were changed to those for alanine. Using these mutant primers and appropriate external primers, the DNA fragments were generated by polymerase chain reaction, digested, and inserted into appropriate sites of the modified plasmid. We finally replaced nine cysteine residues with alanine. Furthermore, we introduced single cysteine residues into this Cys-less NHE1 (see Fig. 1) using a similar polymerase chain reaction-based method. DNA se-quences of polymerase chain reaction fragments were confirmed with a Perkin-Elmer ABI model 373S autosequencer.
Labeling with Biotin Maleimide-Biotin maleimide labeling of the NHE1 mutant molecules was carried out as described previously (16,17) with modification. Confluent cells cultured on 60-mm dishes were washed twice with 5 ml of PBSCM (phosphate-buffered saline containing 0.1 mM CaCl 2 and 1 mM MgCl 2 at pH 7.2) and incubated in 1 ml of PBSCM containing 0 or 5 mM MTSET for 30 min at room temperature. Cells were washed twice with 5 ml of PBSCM and then incubated in 1 ml of PBSCM containing 0.05 or 0.5 mM biotin maleimide (100 mM stock in Me 2 SO) for 30 min at room temperature. Cells were washed once with PBSCM containing 1% 2-mercaptoethanol and once with PBSCM and collected in the tube by centrifugation. Cells were solubilized with 1 ml of the lysis buffer containing 1% Triton X-100, 150 mM NaCl, 20 mM Hepes/Tris (pH 7.4), 1 mM phenylmethylsulfonyl fluoride, and 1 mM benzamidine. After centrifugation for 10 min at 15,000 rpm, the supernatant was mixed with streptavidin-agarose beads (30 l of resin) and incubated for 1 h at 4°C with mild agitation. Agarose beads were washed more than five times with 1 ml of lysis buffer, mixed with 50 l of SDS-PAGE sample buffer containing 3% SDS, and boiled for 10 min at 100°C. The proteins were separated on an 8.5% gel by SDS-PAGE and analyzed by the immunoblot with the NHE1 antibody as described previously (10). The blots were developed using the ECL detection system (Amersham Pharmacia Biotech).
When indicated, cells were permeabilized with SLO (Sigma) immediately before the biotinylation, as described previously (17). Before the experiment, an aliquot of the SLO stock solution (25,000 units/ml of water) was diluted 70-fold into SLO buffer containing 2.5 mM MgCl 2 , 1 mM dithiothreitol, 115 mM potassium acetate, and 25 mM Hepes/KOH (pH 7.4), and the diluted solution was kept on ice for at least 10 min. Cells were washed twice with SLO buffer and then incubated for 15 min on ice in 1 ml of the above SLO-containing solution. Cells were washed twice with 5 ml of ice-cold SLO buffer and then incubated with 5 ml of prewarmed SLO buffer for 30 min at 37°C. Cells were washed with SLO buffer and then with PBSCM, which was followed by the biotinylation step.
Measurement of 22 Na ϩ Uptake- 22 Na ϩ uptake activity and its pH i  (14). Nine endogenous cysteines mutated to alanines and 89 cysteines newly introduced into Cys-less NHE1 are shown by filled squares and filled circles, respectively. EL1-EL6, putative extracellular loops connecting TMs 1 and 2, 3 and 4, 5 and 6, 7 and 8, 9 and 10, and 11 and 12, respectively; IL1-IL5, putative intracellular loops connecting TMs 2 and 3, 4 and 5, 6 and 7, 8 and 9, and 10 and 11, respectively. dependence were measured by K ϩ /nigericin pH i clamp method (12). Serum-depleted cells in 24-well dishes were preincubated for 30 min at 37°C in Na ϩ -free choline chloride/KCl medium containing 20 mM Hepes/Tris (pH 7.4), 1.2-140 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 5 mM glucose, and 5 M nigericin. Choline chloride was added to medium to maintain the total concentration of KCl plus choline chloride at 140 mM. 22 Na ϩ uptake was started by adding the same choline chloride/KCl solution containing 22 NaCl (37 kBq/ml) (final concentration, 1 mM), 1 mM ouabain, and 100 M bumetanide. In some wells, the uptake solution contained 0.1 mM EIPA. After 1 min, cells were rapidly washed four times with ice-cold phosphate-buffered saline to terminate 22 Na ϩ uptake. pH i was calculated from the imposed [K ϩ ] gradient by assuming the equilibrium and an intracellular [K ϩ ] of 120 mM. As indicated, cells were washed three times with PBSCM and incubated with 10 mM MTSET for 30 min at room temperature before the 22 Na ϩ uptake measurement.
Protein Determination-Protein concentration was measured by the bicinchoninic assay system (Pierce Chemical Co.) using bovine serum albumin as a standard.

RESULTS
Characterization of Cys-less NHE1-In this study, we first characterized Cys-less NHE1 expressed in the exchanger-deficient PS120 cells. Fig. 2A shows pH i dependence of 22 Na ϩ uptake in cells expressing the wild-type and Cys-less mutant exchangers. Cys-less NHE1 exhibited a high 22 Na ϩ uptake activity (Ͼ40 nmol/mg/min) similar to the wild-type, and the pK values for both exchangers were found to be similar (ϳ6.7). Hill coefficients were also similar (1.22 or 1.25 for the wild-type or Cys-less NHE1). Fig. 2B shows the concentration dependence of EIPA on 22 Na ϩ uptake by both exchangers, which gave similar IC 50 values (60 -80 nM). The amounts of the exchanger proteins expressed were also similar ( Fig. 2C). Thus, Cys-less NHE1 is fully functional and has almost the same basic properties as the wild-type NHE1.
We compared the biotinylation efficiency among the wildtype, Cys-less, and other mutant exchangers that were well expressed in PS120 cells (Fig. 2C). Cells were incubated with 0.05 or 0.5 mM biotin maleimide, and biotinylated proteins were subsequently recovered with streptavidin-agarose. After SDS-PAGE, the biotinylated exchangers were visualized by immunoblot analysis with an anti-NHE1 antibody (Fig. 2D). As expected, recovery of the Cys-less NHE1 protein was minimal, indicating that the mutant was only slightly labeled with externally applied biotin maleimide. In contrast, heavy antibody staining was observed for the cysteine-substituted mutant E151C, indicating its strong biotinylation (Fig. 2D). We observed very weak labeling of the wild-type and M-Cys-less NHE1. In the latter mutant, we replaced 6 endogenous cysteine residues in the N-terminal membrane domain with alanine while retaining 3 endogenous cysteines in the C-terminal cytoplasmic domain. Thus, the endogenous cysteine residues in the N-terminal membrane domain are not accessible from the external surface, and a relatively low concentration (0.5 mM) of biotin maleimide is virtually membrane-impermeable under the conditions used.
Biotinylation of Substituted Cysteines in Putative Extracellular Loops-We introduced 15 cysteine residues individually into the putative first extracellular loop. All of the mutants were well expressed in PS120 cells (Fig. 3, A and C). We observed two forms of the exchanger with a high (ϳ110 kDa) or low (ϳ90 kDa) apparent molecular mass that are thought to be the mature proteins with N-and O-linked glycosylation or immature proteins containing only N-linked high mannose oligosaccharide, respectively (14). Mutant exchangers with a high molecular weight were labeled with externally applied biotin maleimide (Fig. 3, B and D). In addition, an aggregated form of the exchanger was also biotinylated. We found that 22 Na ϩ uptake into cells expressing the wild-type or Cys-less NHE1 was measured using the K ϩ /nigericin method. B, pH i of cells expressing the wild-type or Cys-less NHE1 was clamped at 5.6 by the use of K ϩ /nigericin, and 22 Na ϩ uptake into these cells was measured as a function of EIPA concentration. C and D, cells expressing the wild-type or NHE1 mutants were incubated with 0.05 or 0.5 mM biotin maleimide, solubilized with the lysis buffer, and then treated with streptavidin-agarose as described under "Experimental Procedures." Total cell lysate (40 g) (C) or the proteins recovered with streptavidinagarose (D) were analyzed by SDS-PAGE. Wild-type or mutant NHE1 proteins were visualized by immunoblot analysis with the NHE1 antibody. these mutant exchangers were biotinylated with variable efficiency. For example, H35C, S56C, T79C, and H81C were strongly biotinylated, while S40C, T68C, R72C, and H76C were only weakly labeled. Preincubation of cells with the membraneimpermeable SH modifier MTSET significantly reduced the biotinylation of all these mutants, indicating that substituted cysteines are accessible to external MTSET. However, the variable efficiency of biotinylation suggests that the first extracellular loop may form a structure in which some amino acid residues are not readily accessible to biotin maleimide.
We introduced 3, 5, 8, and 12 cysteine residues into the putative second, third, fourth, and fifth extracellular loops, respectively (see Fig. 1). These mutants were well expressed in PS120 cells (data not shown). The mutants for the putative second, third, and fourth extracellular loops were all strongly labeled with biotin maleimide, which was significantly reduced by preincubation with MTSET (Fig. 4, A-C). In the putative fifth loop, all of the mutants were strongly biotinylated in a MTSET-sensitive manner (Fig. 4D). However, some mutants (E368C, S375C, H376C, and T377C) were less sensitive to MTSET (Fig. 4D), suggesting that the fifth extracellular loop may form a structure in which some amino acid residues are not readily accessible to MTSET. K384C and S388C were not labeled with biotin maleimide from either side of the membrane (data not shown). Thus substituted cysteines except for those at positions 384 and 388 in these putative loops are exposed to extracellular medium.
In order to test if substituted cysteines not readily accessible from the outside in Fig. 5 are accessible from the cytosol, we permeabilized cells with streptolysin O. As shown in Fig. 6A, cell permeabilization extensively increased the biotinylation of M-Cys-less NHE1 that contains three endogenous cysteines in the cytoplasmic domain. However, the same treatment did not enhance the biotinylation of Cys-less NHE1 and E151C (data not shown). The streptolysin O treatment also increased biotinylation of R180C, Q181C, F182C, E248C, I249C, H250C, T251C, R321C, or R323C, each of which was inhibited by preincubation with MTSET (Fig. 6, B-D; data not shown for some mutants). The data suggest that these substituted cysteines are accessible from the cytosol. In contrast, two mutants, S126C (data not shown) and S127C (Fig. 6A), were not labeled even in permeabilized cells, suggesting that these residues may be embedded in the lipid bilayer and thus inaccessible from either side of the membrane.
Intriguingly, we found that T183C was strongly biotinylated in a MTSET-sensitive manner in nonpermeabilized cells (Fig.  5B), although biotinylation of R180C and Q181C was strongly enhanced by the streptolysin O treatment as stated above (Fig.  6B). Similar aberrant biotinylation occurred in nonpermeabilized cells expressing T323C, S324C, or H325C (Fig. 5D), although the labeling is not so strong. The data suggest that these cysteines are accessible from outside, despite their suggested intracellular localization.
We next determined the localization of the N terminus by utilizing a cysteine mutant designated C8*, in which 8 endog- enous cysteines except for the most N-terminal one (see Fig. 1) were replaced by alanine. C8* was strongly biotinylated in a MTSET-sensitive manner only after cells were permeabilized (Fig. 6E), suggesting that the N terminus is localized in the cytosol.
Biotinylation of Cysteine Residues in the Last Putative Extracellular Loop-We had a particular interest in the last putative extracellular loop, not only because a previous report (15) suggested that this region may not be fully exposed to the outside but also because our preliminary study revealed that mutation of some residues within this region was able to alter the function of NHE1 (e.g. see K472C in Fig. 8). We introduced 26 cysteines individually into this region. All of the single cysteine mutants except for K438C, R440C, I451C, G456C, and R458C were well expressed in PS120 cells (Fig. 7A, but data not shown for some mutants) and fully functional (see Fig. 8). The mutants with extremely reduced exchange activity did not produce stable transfectants, suggesting that the native residues at these positions might be important for the ion transport or expression of the exchanger in the plasma membrane.
We found that in nonpermeabilized cells, 14 single cysteine mutants (from K447C to S464C) were not efficiently labeled with biotin maleimide in the absence or presence of external MTSET (Fig. 7, B and C). However, four mutants, K471C, K472C, H473C, and F474C, in the C-terminal portion of the last putative extracellular loop (see Fig. 1) were biotinylated in nonpermeabilized cells in a MTSET-sensitive manner (Fig.  7C), indicating that these cysteines are accessible from the outside. In contrast, the labeling of K443C, K447C, and D448C in the N-terminal portion of the last putative extracellular loop (see Fig. 1) was strongly enhanced by treatment with streptolysin O (Fig. 7D), indicating the accessibility of these cysteines from the inside. The streptolysin O treatment did not increase the biotinylation of Q449C (Fig. 7D) and other mutants, F450C to S464C (data not shown). These results suggest that the region from Gln 449 to Ser 464 are not accessible from both extracellular and intracellular biotin maleimide. Thus, the last putative extracellular loop appears to consist of an intracellu-lar loop, a transmembrane segment, and an extracellular loop.
Effect of External MTSET on Na ϩ /H ϩ Exchange Activity of Substituted Cysteine Mutants-We investigated the effect of externally applied MTSET (10 mM) on 22 Na ϩ uptake by almost all functional cysteine mutants (Fig. 8). We found that uptake activities of many mutants were not affected by MTSET. However, a slight but significant MTSET-dependent inhibition of activity was observed in P153C, A281C, N282C, H373C, and K471C (activity decreased to 82 Ϯ 2, 91 Ϯ 3, 87 Ϯ 3, 77 Ϯ 3, and 89 Ϯ 3% of control, respectively, after MTSET treatment). These cysteine mutants were labeled in nonpermeabilized cells with biotin maleimide in a MTSET-sensitive manner (see above). Intriguingly, MTSET markedly inhibited uptake activities of R180C and Q181C (residual activity of 32 Ϯ 4 and 39 Ϯ 10%, respectively, after MTSET treatment), indicating that these two cysteines are accessible to externally added MTSET. Of note, these substituted cysteines were not efficiently labeled in nonpermeabilized cells with biotin maleimide in an MTSETsensitive manner (Fig. 5B). Thus, the loop between TM4 and TM5 could be involved in the ion transport. DISCUSSION In this study, we have utilized the reactivity of introduced cysteines toward biotinylating reagent to determine the membrane topology of NHE1. This approach has been used to determine the topology of human P-glycoprotein (16) and several other transporters (17,20,21). We found that externally applied biotin maleimide did not virtually label the wild-type, Cys-less, and M-Cys-less NHE1 (Figs. 2, C and D) but strongly labeled M-Cys-less NHE1, in which 3 native cysteines were retained in the C-terminal cytoplasmic domain, when cells were permeabilized (Fig. 6A). These results confirmed the pre- vious observations (16) that this reagent is relatively membrane-impermeable and highly cysteine-directed if it is used in low concentrations (0.5 mM). Therefore, labeling with biotin maleimide in a manner sensitive to the membrane-impermeable MTSET primarily defines the introduced cysteine as being exposed on the cell surface, while similar labeling with biotin maleimide only in permeabilized cells defines the cysteines as being localized intracellularly.
Using Cys-less NHE1, we introduced single cysteine residues into the regions of the exchanger modeled to be localized extracellularly or intracellularly from hydropathy analysis. Cysless NHE1 has basic properties similar to the wild-type NHE1 in that it was highly expressed in the transfectants and exhibited Na ϩ /H ϩ exchange activity comparable with that of the wild-type exchanger ( Figs. 2A and 8). In addition, both Cys-less and wild-type exchangers exhibited essentially the same pH i and EIPA concentration dependences of exchange activity (Fig.  2, A and B). We found that most of single cysteine mutants exhibited relatively high exchange activity (Fig. 8), although stable transfectants could not be obtained with some mutants. We observed two bands reactive with the NHE1-antibody in the transfected cells (Figs. 3A and 7A), i.e. high and low molecular weight forms of the exchanger that are thought to be the mature proteins with N-and O-linked glycosylation or the immature proteins only containing high mannose oligosaccharide, respectively (14). As expected, in most of the cases, the cysteine residues introduced into the putative extracellular loops were strongly labeled with biotin maleimide only in the high molecular weight form of the exchanger. However, the low molecular weight form of NHE1 was also slightly biotinylated in some cases (e.g. see Fig. 4D). Although we do not know the precise reason for that, it is possible that part of the lower form is expressed in the plasma membrane under some conditions. Fig. 9 shows a new topological model of NHE1 predicted on the basis of the present results. The model consists of 12 TMs and intracellular and extracellular loops connecting TMs with both the N and C termini located in the cytoplasm. Interestingly, a 12-TM model with the intracellular N and C termini has been proposed for a Na ϩ /H ϩ antiporter of Escherichia coli (22). The assignment of TM1 to TM9 of NHE1 in the new model is not different from that in the original hydropathy model (Fig.  1). We could not obtain evidence for the existence of the intracellular loop connecting TM2 and TM3 (IL1), because the introduced cysteines at positions 126 (data not shown) and 127 (Fig. 6A) were not labeled in permeabilized as well as in nonpermeabilized cells. However, since the hydrophobic region of amino acids 105-150 is long enough to span the lipid bilayer twice, we consider this hydrophobic region to form TM2 and TM3.
The transmembrane disposition of TM10 to TM12 of the original model is extensively modified in the new model. We found that introduced cysteines at positions 373-377, 381, and 407-409 were accessible from the outside. The intervening hydrophobic stretch of amino acids 385-406, which mostly corresponds to TM10 of the original model, is not long enough to span the lipid bilayer twice if it forms an ␣-helix. Because two mutants, K384C and S388C, were not labeled with biotin maleimide from either side of the membrane (see "Results"), we tentatively placed most of this hydrophobic segment within the lipid bilayer. Cysteines at positions 471-474 in the last extracellular loop of the original model were labeled with biotin maleimide from the outside in a MTSET-sensitive manner, while cysteines at positions 443, 447, and 448 were labeled from the inside. Thus, the hydrophobic stretch of amino acids 449 -470 appears to cross the lipid bilayer with an inside to outside orientation. Such an interpretation is consistent with our finding that introduced cysteines in this region were not readily biotinylated from either side of the membrane (Fig. 7, B and C). Since cysteine at position 448, but not that at position 449, was biotinylated from the inside, the new TM11 could start from Gln 449 (Fig. 7D).
The N terminus has been predicted to be cleaved off as a signal peptide at a site somewhere before the glycosylation sites in the first extracellular loop. A typical structure of the signal peptide has three distinct domains, i.e. an N-terminal positively charged region (1-5 residues long); a central, hydro- phobic region (7-15 residues); and a more polar C-terminal region (3-7 residues long) (23). Although NHE1 was indeed predicted to possess a signal peptide, 2 cleavage of the N terminus of NHE1 has not yet been experimentally proven. We observed that all of the introduced cysteines in the first extracellular loop were biotinylated in a MTSET-sensitive manner (Fig. 3, B and D), suggesting that these residues including cysteine at position 35 near the extracellular interface of TM1 were retained in the mature NHE1 protein. Furthermore, mutant C8* was labeled only after cells were permeabilized (Fig.  6E), suggesting that Cys 8 is also retained in the mature protein and localized in the cytosol. It is therefore most likely that NHE1 is not cleaved during protein processing, unless introduced individual cysteines somehow prevent the cleavage of protein by signal peptidases. Thus, the central hydrophobic stretch in the N-terminal signal peptide-like structure is likely to form the first transmembrane ␣-helix.
As described above, we observed that M-Cys-less NHE1 containing three endogenous cysteines (amino acids 538, 561, and 794) were labeled with biotin maleimide only after cells were permeabilized (Fig. 6A). This result is consistent with much previous biochemical data showing that most of the C-terminal domain of NHE1 (ϳ300 amino acids) is located intracellularly (10, 24 -26). Recently, Shrode et al. (15) have shown that hemagglutinin epitope tagged at the C-terminal end of NHE1 can be recognized by the hemagglutinin-specific antibody only after cells expressing NHE1-hemagglutinin are permeabilized. All these results strongly suggest that the C-terminal tail (amino acids 500 -815) of NHE1 is entirely located intracellularly, which is in contrast to the recent report that part of the C terminus of NHE3 is exposed on the extracellular side (27).
We found that cysteines at positions 180 -182 were readily biotinylated from the inside but not from the outside (Figs. 5B and 6B). This suggests that these positions are accessible from the inside. Intriguingly, however, 22 Na ϩ uptake by two mutants R180C and Q181C was strongly inhibited by externally applied MTSET (Fig. 8), suggesting that these residues are accessible from the outside. We also found that cysteine at position 183 was strongly biotinylated from the outside in a MTSET-sensitive manner (Fig. 5B). These data suggest that at least a portion of the loop connecting TM4 and TM5 (IL2) is localized within the membrane, raising an intriguing possibility that it could form a structure lining the aqueous pore that is accessible from either side of the membrane and involved in the ion transport. As described above, cysteines at positions 180 and 181 are accessible to MTSET from the outside as evidenced by its inhibition of 22 Na ϩ uptake but were not readily accessible to biotin maleimide from the same side. Such a difference in the accessibility may simply reflect the difference in the reagent size, because biotin maleimide is much bulkier than MTSET.
It is important to note that TM4 has been suggested to contain part of the amiloride-binding site (28). Single amino acid substitutions at Phe 161 and Leu 163 within the 160 VF-FLFLLPPI 169 segment of TM4 were initially found to cause reduction of the amiloride sensitivity (29), while more recently the G174S mutation in the adjacent region in TM4 was also found to cause reduction in the amiloride affinity (29). Double mutant L163F/G174S exhibited markedly reduced affinity not only for amiloride (or its derivatives) but also for Na ϩ (29). These previous results fit nicely with the present results and together support the view that TM4 and the adjacent loop (IL2) may form part of an ion transport pathway in the exchanger. As pointed out previously (29), TM4 is a proline-rich segment 2 Prediction of the signal peptide was made on the World Wide Web. FIG. 9. A new topology model of NHE1. The model is based on the data obtained in this study. Red circles, accessible to external SH-directed reagents; blue circles, accessible to cytosolic SH-directed reagents; violet circles, not readily accessible to biotin maleimide from either side of the membrane; black circles, not analyzed because of low level expression of the functional exchanger; rectangles, positions for native cysteines replaced with alanine; blue rectangle, a native cysteine (Cys 8 ) accessible to cytosolic SH-directed reagents.
containing three highly conservative proline residues (Pro 167 , Pro 168 , and Pro 178 ). It is likely that these proline residues produce a kink in ␣-helix, which could create a space to accommodate part of IL2 in the membrane.
Cysteines introduced into the putative intracellular loop connecting TM8 and TM9 (IL4) exhibited topological dispositions similar to those in the loop connecting TM4 and TM5 (IL2) in that positions 321 and 323 were accessible from the inside, while positions 324 and 325 were accessible from the outside (Figs. 5D and 6D). However, 22 Na ϩ uptake by six cysteine mutants of IL4 was not significantly influenced by external MTSET (Fig. 8). On the other hand, recent molecular studies including one with chimeric exchangers between NHE1 and NHE3 revealed the functional importance of a 66-amino acid segment encompassing IL4, TM9, and EL5 for the interaction of the exchanger with inhibitors such as amiloride derivatives and HOE694 (30) as well as of His 349 in TM9 for the amiloride sensitivity (31). These previous results, together with our data showing that at least a portion of IL4 is localized within the membrane, raise a possibility that this loop may also be involved in ion transport. Clearly, more detailed analysis is required to clarify the functional importance of the region encompassing IL4 and TM9.
In summary, based on the results from cysteine accessibility analysis, we proposed a novel 12-TM model with the N and C termini located in the cytoplasm. This new topology model would provide a basis for the further study of the structurefunction relationship of NHE1.