Site-directed chemical labeling of extracellular loops in a membrane protein. The topology of the Na,K-ATPase alpha-subunit.

We have mapped the membrane topology of the renal Na,K-ATPase alpha-subunit by using a combination of introduced cysteine mutants and surface labeling with a membrane impermeable Cys-directed reagent, N-biotinylaminoethyl methanethiosulfonate. To begin our investigation, two cysteine residues (Cys(911) and Cys(964)) in the wild-type alpha-subunit were substituted to create a background mutant devoid of exposed cysteines (Lutsenko, S., Daoud, S., and Kaplan, J. H. (1997) J. Biol. Chem. 272, 5249-5255). Into this background construct were then introduced single cysteines in each of the five putative extracellular loops (P118C, T309C, L793C, L876C, and M973C) and the resulting alpha-subunit mutants were co-expressed with the beta-subunit in baculovirus-infected insect cells. All of our expressed Na,K-ATPase mutants were functionally active. Their ATPase, phosphorylation, and ouabain binding activities were measured, and the turnover of the phosphoenzyme intermediate was close to the wild-type enzyme, suggesting that they are folded properly in the infected cells. Incubation of the insect cells with the cysteine-selective reagent revealed essentially no labeling of the alpha-subunit of the background construct and labeling of all five mutants with single cysteine residues in putative extracellular loops. Two additional mutants, V969C and L976C, were created to further define the M9M10 loop. The lack of labeling for these two mutants showed that although Met(973) is apparently exposed, Val(969) and Leu(976) are not, demonstrating that this method may also be utilized to define membrane aqueous boundaries of membrane proteins. Our labeling studies are consistent with a specific 10-transmembrane segment model of the Na,K-ATPase alpha-subunit. This strategy utilized only functional Na,K-ATPase mutants to establish the membrane topology of the entire alpha-subunit, in contrast to most previously applied methods.

Na,K-ATPase (sodium pump) is an integral membrane protein that is present in most animal cells. The enzyme consists of two subunits, a large catalytic ␣-subunit (about 110 kDa) and a glycosylated ␤-subunit (ϳ55 kDa); both subunits are required for enzymatic activity (1)(2)(3). Na,K-ATPase utilizes the energy derived from ATP hydrolysis to transport Na ϩ and K ϩ ions across plasma membranes against their electrochemical gradients and is a member of the P 2 -ATPase family (4). The ion gradients generated by the sodium pump are important for regulating a variety of physiological functions such as cell excitability, contractility, and secondary active transport.
Several isoforms of the ␣and ␤-subunits have been cloned from various species, and the primary sequences have been described (5). The secondary structural information on the protein, on the other hand, remains controversial despite extensive investigation; such information is essential for understanding structure/function relationships of the Na,K-ATPase. Based on hydropathy analysis (6), protease accessibility (7,8), and immunochemical studies (9,10), the amino-terminal third of the ␣-subunit appears to contain four membrane-spanning regions. These regions are followed by a large cytoplasmic loop that has been identified as the ATP-binding domain. Chemical reagents that inactivate the Na,K-ATPase (and abolish high affinity ATP binding) in an ATP protectable manner all modify amino acid residues in this loop (reviewed in Ref. 5). This loop has recently been overexpressed in Escherichia coli and exhibits the same nucleotide-binding specificities as native Na,K-ATPase (11). Studies on the carboxyl-terminal third of the ␣-subunit have produced conflicting results (7)(8)(9)(10)(12)(13)(14)(15). Recent data using epitope accessibility claim to establish four transmembrane segments in the carboxyl-terminal region and have a total of eight transmembrane segments (12), whereas several other studies claim to establish that there are 10 transmembrane segments (13)(14)(15).
Most methods used until now to establish membrane topology of the Na,K-ATPase suffer from inherent limitations. Several methods employ proteolytic digestion in sealed vesicles or with purified protein (7,15). It is not clear in such studies whether after the initial protease cleavages that the protein retains what can be considered as its native structure. In more recent cleavage methods, based on metal-ion catalyzed cleavages (15), assumptions are made that the cleavage event occurs close to the metal binding sites, and the locations of the bound metals are also unknown. Indeed with this and other cleavage approaches, interpretations based on minor fragmentation products need further confirmation by other methods. In the use of epitope accessibility, assumptions must be made about the effects of permeablilizing detergents, i.e. that such detergent treatments do not alter protein structure (9,10,16). These methods are particularly prone to provide misleading information with a protein such as the Na,K-ATPase, where abundant evidence exists demonstrating the mobility or flexibility of its carboxyl-terminal regions. In other methods originally introduced in studies of prokaryotic membrane protein topology, the amino-terminal domain truncations are attached to reporter proteins to determine the topogenic properties of the fusion region (13,14). However, because the complete structure of the protein of interest is not used and functional tests are not then available, it is assumed that the membrane orientation of the truncations faithfully reflects their orientation in the complete proteins.
In this paper, we describe a different approach to determine the correct number of transmembrane-spanning regions of the Na,K-ATPase ␣-subunit. This method uses data obtained from functional Na,K-ATPase molecules in their native states. Two cysteine residues (Cys 911 and Cys 964 ) in the ␣-subunit are changed into serine or alanine residues; the resulting ␣ "null" mutant protein no longer has available cysteine residues for extracellular chemical modification (17). Using this null protein as a starting point, our strategy is to introduce one cysteine residue into each of the putative extracellular loops so the membrane topology of the ␣-subunit can be probed by surface labeling with membrane impermeable sulfhydryl reagents. The Na,K-ATPase mutants are heterologously expressed in insect cells using a baculovirus expression system. We are able to characterize and perform labeling experiments exclusively on expressed proteins, because the insect cells (sf9 and High Five cells) contain no detectable amount of endogenous Na,K-ATPase activity (3,18). Our results are consistent with a specific 10-transmembrane segment topological model of the Na,K-ATPase ␣-subunit. A preliminary report of some of our findings has been published previously (19).

EXPERIMENTAL PROCEDURES
A full-length cDNA clone encoding the wild-type sheep Na,K-ATPase ␣1and ␤1-subunits was a gift of Dr. Elmer Price (University of Missouri, Columbia). The Bac-To-Bac TM baculovirus expression system was obtained from Life Technologies, Inc.
Plasmids and Construction of Mutant-Site-directed mutagenesis of the ␣-subunit was carried out in the pST100 vector via the polymerase chain reaction overlap extension mutagenesis method (20, 21) using primers listed in Table I. The plasmid pST100 was constructed by subcloning the wild-type sheep ␣-subunit cDNA into the multiple cloning site of pOCUS-2 vector (Novagen) as a NotI and Sse8387I fragment; two silent mutations were then introduced into the ␣-subunit cDNA to facilitate cassette mutagenesis (see Fig. 1A). Using the C911S/C964A mutant as the background construct, cysteine residues listed in Table I were introduced in the ␣-subunit individually.
Recombinant baculovirus was produced by following the Bac-To-Bac™ system protocols provided by the manufacturer. Briefly, the donor plasmid pFASTBACDUAL␣␤ (Fig. 1B) was constructed by subcloning the wild-type ␤and ␣-subunit cDNA into the multiple cloning sites I and II of the pFASTBACDUAL vector, respectively. The ␤-subunit cDNA was introduced into the vector as an EcoRI/SpeI fragment and the ␣-subunit cDNA as a SmaI/StuI fragment. To construct donor plasmids containing the cysteine mutations, the wild-type ␣-subunit cDNA in the pFASTBACDUAL␣␤ was replaced with the ␣ mutants using AflII and SacI. DH10BAC cells were transformed with pFAST-BACDUAL␣␤ vectors to obtain recombinant baculovirus shuttle vectors (bacmids), which were used to transfect insect cells to generate recombinant baculoviruses. The genomic DNA of the recombinant baculoviruses were isolated by using the Easy-DNA Kit (Invitrogen) and were sequenced to ensure the appropriate cysteine mutations in the ␣-subunits.
Cells and Viral Infections-High Five cells were maintained at 27°C in 250-ml suspension cultures and were split every 3 days with fresh Ex-Cell™405 medium (JRH Biosciences) to keep cell density between 0.5 and 4 ϫ 10 6 cells/ml. For viral amplification, log phase high viability High Five cells (Ͼ98%, as determined by trypan blue exclusion) were infected with recombinant baculovirus at a multiplicity of infection of 0.1-1. After 5 days, cells were centrifuged (500 ϫ g, 10 min), and the resulting supernatants were collected as viral stocks. For protein expression, log phase high viability High Five cells in medium containing 1% ethanol (v/v) were infected with viral stocks at a multiplicity of infection of 10 -15. High Five cells were harvested (100 ϫ g, 10 min) 4 days after infection.
Plasma Membrane Isolation-Harvested High Five cells (ϳ4 ϫ 10 8 cells) were frozen at Ϫ20°C for an hour and then were resuspended in 6 ml of ice-cold homogenizing buffer (250 mM sucrose, 10 mM Tris-HCl, pH 7.4) containing a protease inhibitor mixture (1 g/ml antipain, 1 g/ml pepstatin, 1 g/ml leupeptin, 20 g/ml phenylmethanesulfonyl fluoride). The cells were disrupted via Dounce homogenization, and the cell mixture was centrifuged for 15 min at 500 ϫ g to remove intact cells and debris. Into the supernatant containing cell membranes was added an equal volume of 2.55 M sucrose (made in 2 mM EDTA and 10 mM Tris-HCl, pH 7.4), and the membrane mixture was fractionated on a five-step sucrose gradient as described (22). Briefly, the sucrose gradients (3 ml of 2 M sucrose, 6 ml of 1.6 M sucrose, 12 ml of cell membrane mixture, 12 ml of 1.2 M sucrose, and 6 ml of 0.8 M sucrose) were centrifuged in a Beckman SW 28 rotor at 25,000 rpm for 2.5 h, and the endoplasmic reticulum, Golgi apparatus, and plasma membrane fractions were discernible at the 1.5 M, 1.3 M, and 1 M density regions, respectively. The collected plasma membrane fraction was diluted with buffer (25 mM imidazole, 1 mM EDTA, pH 7.4), concentrated by centrifugation (Beckman 60Ti rotor, 40,000 rpm, 30 min) and resuspended in a small volume of the homogenizing buffer. The plasma membrane proteins were kept at 4°C for short term storage and at Ϫ20°C for long term storage (ϳ4 weeks).
ATPase Assay-A typical Na,K-ATPase assay contained 500 l of assay medium (0.6 mM EGTA, 156 mM NaCl, 24 mM KCl, 3.6 mM MgCl 2 , 3.6 mM ATP, 60 mM imidazole, 10 mM sodium azide, pH 7.2) and 100 l of cell membrane preparations containing 8 g of protein. The protein concentration was measured by the method of Lowry et al. (23) using bovine serum albumin as a standard. The assay mixture was incubated at 37°C for 30 min, and the phosphate release was determined as reported by Brotherus et al. (24). The Na,K-ATPase activity was the difference between the ATP hydrolysis measured in the presence and absence of 83 M ouabain.
Phosphorylation with [ 32 P]ATP-This procedure was carried out essentially as described (25)  Topology of the Na,K-ATPase ␣-Subunit calculated from the difference between [ 32 P]phosphate incorporation in the medium above and that measured in a medium containing 1 M KCl instead of NaCl.

Equilibrium [ 3 H]Ouabain
Binding-Ouabain binding of the expressed proteins was measured as described (26)  At the end of the incubation period, the cell mixture was briefly treated with 2-mercaptoethanol (14 mM) to remove excess reagent and then was washed twice with 50 ml of the incubation buffer. When the 2-mercaptoethanol quench was omitted, the outcome of the labeling reactions was unaltered (data not shown). Cells were harvested, and the cell membranes were fractionated on the five-step sucrose gradient as described above. For immunoprecipitations, 100 -300 g of plasma membrane proteins were solubilized in 500 l of 2% CHAPS in the incubation buffer for 15 min at room temperature, and the insoluble material was pelleted (600 ϫ g, 6 min). The CHAPS mixture was diluted to 1% and precleared overnight with rabbit pre-immune serum and protein G-Sepharose beads (Amersham Pharmacia Biotech). The Sepharose beads were pelleted (600 ϫ g, 3 min), and the resulting supernatant was incubated with antibody raised against the ATP-binding domain of the ␣-subunit (11) and fresh protein G-Sepharose beads overnight at 4°C. The Sepharose beads were washed twice with 1% CHAPS and eluted with a 30-l sample buffer (equal volumes of 8 M urea, 10% SDS, and 125 mM Tris buffer, pH 6.8). The protein samples were separated in 7.5% acrylamide gel (27), transferred onto nitrocellulose membrane (MSI Micron Separations, Inc.) by electroblotting in 10 mM CAPS (pH 11) containing 10% methanol, and immunostained with peroxidaselinked streptavidin (Amersham Pharmacia Biotech). An identical nitrocellulose blot was stained with mouse monoclonal anti-␣1 antibody (ABR Bioaffinity Reagent, Inc.) and peroxidase-linked sheep antimouse secondary antibody (Amersham Pharmacia Biotech). The signal intensity levels on the blots were analyzed by the computer software NIH Image.

Expression and Enzymatic Characterization of Cysteine
Mutants-Using the C911S/C964A construct as the null back-ground, a panel of Na,K-ATPase ␣ mutants was generated. Each mutant contained one cysteine residue in each of the putative extracellular loops (Table I) and was co-expressed with the wild-type sheep ␤-subunit using infection with baculovirus. The expressed plasma membrane proteins were purified by a five-step sucrose gradient centrifugation and characterized to confirm functional integrity. As shown in Table II, the mutants displayed specific activities in the range of 0.1-0.6 mol P i ⅐ mg Ϫ1 protein ⅐ min Ϫ1 , had equal ouabain-binding and phosphorylation levels, which ranged from 10 -100 pmol ⅐ mg Ϫ1 protein, and the turnover number, based on these specific activities and the ligand-binding levels, were all between 7000 and 10,000 min Ϫ1 . In other words, all the expressed mutants showed normal functional activity.
Labeling of Cysteine Mutants-High Five cells expressing the Na,K-ATPase cysteine mutants were incubated with MT-SEA-biotin so that a biotin group would be introduced into the ␣-subunits that contain extracellularly exposed cysteines. After the MTSEA-biotin treatment, the mutant proteins were immunoprecipitated and resolved by SDS-polyacrylamide gel electrophoresis, and the protein biotinylation levels were determined on a blot using peroxidase-linked streptavidin and chemiluminescence. An identical blot was stained with anti-␣1 antibody to determine the amount of ␣-subunit immunoprecipitated. Fig. 2A shows the typical ␣-subunit biotinylation patterns for the expressed proteins, and Fig. 2B shows the amount of ␣-subunit on the blot. A plot of the specific activity of the mutants against their respective ␣-subunit intensities on the blot is shown in Fig. 3A. A strong correlation between the two variables is found in this plot, demonstrating that the difference in the specific activity is because of the difference in the expression level. The very low level (if any) of labeling of the null mutant can be readily seen by visual examination of the relative intensity of staining by the streptavidin compared with the ␣-subunit activity (Fig. 2, A and B, respectively). To allow for the different expression levels of the mutants, the ratio of the ␣-subunit biotinylation level to the ␣-subunit intensity level on the blot was calculated for each mutant. Fig. 3B shows such ratios normalized against that of the null construct. The relative biotinylation signals for the P118C, T309C, L793C, L876C, and M973C mutants are 4 -12 times stronger than that of the null mutant. The ratios demonstrate the significantly greater access of the cysteine residues in these constructs than in the null mutant. To further define the M9M10 extracellular loop, V969C and L976C mutants were constructed. The ␣-subunit biotinylation and intensity levels for these and the M973C mutant are shown in Fig. 4, A and B, respectively. Good ex- Topology of the Na,K-ATPase ␣-Subunit pression levels for the three mutants in the M9M10 regions are observed (Fig. 4B); however, the labeling data of the V969C and L976C mutants showed that these two residues are not accessible from the extracellular medium (Fig. 4A). DISCUSSION We have established the membrane topology of the Na,K-ATPase ␣-subunit by testing the accessibility of cysteine residues introduced in the predicted extracellular loops. Our method rejects all models consisting of less than 10 transmembrane segments and confirms a specific 10 transmembrane segment model (Fig. 5). Our strategy is based on the earlier observations that Cys 911 and Cys 964 are the only two cysteine residues exposed to nonpenetrating cysteine-specific reagents in the extracellular medium (17). In the current study, the C911S and C964A substitutions were made to construct an ␣-subunit background and the lack of labeling of this mutant confirmed the removal of all exposed sulfhydryl groups (Fig. 2), which supports the conclusion of our earlier labeling studies (17). Cysteine was then introduced at residues 118, 309, 793, 876, 969, 973, or 976 of the background construct in each of seven mutants and probed with a membrane impermeable cysteine-directed reagent.
Our experimental strategy involves the functional characterization of mutants to ensure that the topology of active molecules is determined and uses the same strategy for the topology TABLE II Characteristics of Na,K-ATPases isolated from dog kidney inner medulla and from membrane preparations of baculovirus-infected High Five cells Dog kidney enzyme was obtained as described (39,40). Crude membrane preparations from baculovirus-infected High Five cells were fractionated on a five-step sucrose gradient (22), and the resulting plasma membrane fractions were collected for activity assays. The specific activity (ouabain-sensitive ATPase activity) and ligand binding properties of the mutants were determined as described under "Experimental Procedures." The turnover number, which is the turnover rate of phosphoenzyme/minute, was calculated by taking the ratio of specific activity over phosphorylation number. Each value below represents the mean of triplicate determinations of specific activity and ligand binding from at least three different membrane preparations and has a standard error of less than 10%. Specific 2. Labeling of introduced cysteine residues in the putative extracellular loops of the ␣-subunit. High Five cells expressing the Na,K-ATPase cysteine mutants were incubated with 200 M MT-SEA-biotin. After sucrose gradient fractionation, Na,K-ATPase in the plasma membrane fraction was immunoprecipitated, resolved by SDSpolyacrylamide gel electrophoresis, and transferred to nitrocellulose membrane. A, biotin incorporation into the ␣-subunit was detected with peroxidase-linked streptavidin. B, an identical blot was probed with a monoclonal anti-␣1 antibody. Data shown here represent results of three independent labeling studies. WT, wild type. FIG. 3. A, the ␣-subunit intensity levels in Fig. 2B were quantified with the computer software NIH Image and were plotted against the Na,K-ATPase activities of the respective mutants. A linear fit of the two sets of variables yielded an R-square value of 0.92. B, the ratio of the biotin incorporation level over the ␣-subunit expression level was calculated for each mutant and was normalized against such a ratio of the null construct. WT, wild type. of the entire ␣-subunit. It should also be pointed out that this strategy contains both positive and negative controls. Our previous work established the accessibility of Cys 911 (in M8) and Cys 964 (in M9) from the extracellular compartment in studies of native canine renal Na,K-ATPase (17). This provides a positive control for our heterologously expressed wild-type enzyme (Fig.  2, lane 1). The removal of these two cysteine residues provides a negative control where little or no labeling from the extracellular medium is observed (Fig. 2, lane 2). All of our mutants are functionally active. The turnover number for the phosphoenzyme intermediates of the mutants is close to the value for the wild-type phosphoenzyme heterologously expressed in High Five cell membranes (Table II). This suggests that their processing and folding is not affected by the substitutions. In contrast, previous topology studies employed ␣-subunit mutants that were less well characterized (due to endogenous Na,K-ATPase activity) (9, 10) or nonfunctional (12,13), when assumptions about the "normal" orientations of the expressed proteins were made. This is particularly relevant to approaches that use carboxyl-terminal truncations and reporter groups (such as glycosylation status) to establish topology. If trunca-tions are used, usually functional activity is lost so that no reliable test can be made to demonstrate that the truncated polypeptide is processed or folded in the same way as the intact native protein.
The amino-terminal third of the Na,K-ATPase ␣-subunit has been proposed to contain four membrane-spanning regions (7)(8)(9)(10), and the accessibility profiles of the P118C and T309C mutants confirmed the locations of Pro 118 and Thr 309 in the respective M1M2 and M3M4 extracellular loops (Figs. 2 and 3). The region around the M1M2 extracellular loop has been identified as a primary determinant of ouabain-binding affinity in several studies (28 -31). We found that the P118C substitution did not greatly alter the ouabain sensitivity of the mutant (Table II), suggesting that the amino acid residue at this position does not make direct contact with the cardiac glycoside. Furthermore, because the substitution replaces a proline residue, it is unlikely that the precise structure of the M1M2 loop in the region is critical for the ouabain binding interaction.
There has not been a clear consensus in the numbers of membrane-spanning domains in the carboxyl-terminal third of the Na,K-ATPase ␣-subunit. Indeed the carboxyl-terminal topology of the ␣-subunit has been the source of most debate. This is undoubtedly because of the extra mobility and flexibility of the region compared with the amino-terminal region, so that methods that involve disruptive procedures (such as proteolysis and peptide bond cleavage) are prone to generate artifactual observations. We began our investigation in this region by introducing a cysteine at residue 793, which has been predicted to be a part of the M5M6 external loop (7,14,17). Previous studies have shown that the M5M6 hairpin, following proteolysis and removal of K ϩ ions (at 37°C), is released from the membrane (32) to the extracellular space (33). Such release was not observed in the presence of K ϩ ions, suggesting that the M5M6 hairpin in the intact protein is dynamic (possibility moving in a way which is perpendicular to the plane of the membrane) during the conformational shifts of the Na,K-ATPase catalytic cycle. The present study has identified Leu 793 as one of the residues that are exposed to the extracellular phase in the native protein (Figs. 2 and 3). This provides the first evidence for the extracellular location between the loops M5 and M6 in the native Na,K-ATPase. Studies using a cysteine-specific inhibitor, omeprazole, have previously provided evidence for such a configuration in the gastric H,K-ATPase (34).
Our demonstration of the exposure of Leu 793 at the outside surface enabled us to distinguish between two 10 transmembrane models recently proposed for the ␣-subunit. In one study using carboxyl-terminal truncations of the ␣-subunit expressed in yeast (14), the authors place Ala 789 close to the cytoplasmic side and Met 809 in the extracellular loop between M5 and M6. Their results are compared with another 10 membrane-spanning model proposed earlier by Karlish et al. (7), who placed Ala 789 close to the extracellular part of M5. Our results support the latter model, because we find Leu 793 exposed in the extracellular loop, which would place Ala 789 close to the extracellular phase in M5.
Lemas et al. (35) have shown that a 26-sequence peptide (Asn 889 -Ala 914 ) in the predicted M7M8 extracellular loop interacts directly with an external fragment of the ␤-subunit. It should be pointed out that their study utilized a co-immunoprecipitation protocol, and the functional importance of the association was not determined. Our labeling result with the L876C mutant provides evidence for the extracellular location of this loop in the intact protein (Figs. 2 and 3). It is interesting to note that previous immunochemical labeling studies using an antibody raised against the Trp 887 -Arg 904 sequence have produced conflicting results. In one study, the antibody detected Na,K-ATPase from the extracellular domain after prolonged incubation (36). In that study three new models were suggested for the transmembrane organization of the Na,K-ATPase ␣-subunit, consisting of either 8 or 10 transmembrane segments. All three of these models (models A, B, and C in Ref. 36) are rejected by our observations of the extracellular localization of Leu 793 and Leu 876 . In another study, the anti-Trp 887 -Arg 904 antibody failed to label Na,K-ATPase without cell permeablization by detergent (37). An explanation that might account for the discrepancy would be that this region of the M7M8 loop is not freely accessible but rather associates with the ␤-subunit in the extracellular space. Prolonged incubation with the antibody or treatment with detergent disrupted the ␣/␤ interaction, thus enabled the binding of the antibody.
In a recent article, Møller et al. (16) carried out a very thorough study of the membrane topology of the sarcoplasmic reticulum Ca-ATPase. Their work employed a combination of sequence-specific antibodies and proteolysis methods, and they emphasized the extreme caution needed when using detergents or similar treatments for probing intravesicular locations with the antibodies. The most striking conclusion of their studies was that the extracellular loop between M7 and M8 in the previously accepted model was proposed to be cytosolic. Two different models were then suggested that might account for the cytosolic location of this loop: (i) M7 does not completely protrude across the membrane but folds backs to the cytosol from within the membrane or (ii) the existence of an additional membrane return after M7 but before the usual M8 segment. Our data provide clear evidence for the extracellular location of a residue (Leu 876 ) immediately after M7 and for the extracellular accessibility of Cys 911 close to the beginning of M8. This establishes that the region Lys 876 -Cys 911 may be accessed from the extracellular space. These data contradict the conclusion of Møller et al. (16) or else there is a basic difference between the membrane topology of the Na,K-ATPase ␣-subunit and the Ca-ATPase. Furthermore, Møller et al. (16) raised the possibility that there may be a "plasticity" in the membrane location of the M8 segment in the Ca-ATPase. They discuss earlier work on the Na,K-ATPase (8,17,32) to suggest the importance of carboxyl-terminal plasticity in ion pumping by P-type ATPases. Although our current data do not address this specific issue, modifications of our approach may in the future be employed to investigate such plasticity.
The post-tryptic 19-kDa membrane fragment of the ␣-subunit, consisting of residues Asn 831 -Tyr 1016 , was first described by Karlish et al. (38). The amino and carboxyl termini of this protein fragment have been assigned cytoplasmic locations (7), suggesting that it should contain an even number of transmembrane segments (at least two). As mentioned above, the detection of the L876C mutant at the extracellular domain provided evidence for the M7M8 transmembrane segments in the 19-kDa membrane fragment. The labeling data of the M973C mutant revealed an additional extracellular site after M8 (Figs.  2 and 3), validating the presence of the M9 and M10 transmembrane domains.
In the present study and in contrast to earlier studies (9, 10, 12), we have obtained unequivocal evidence for exposure of the M9M10 loop. Presumably the short length of this loop (Figs. 2 and 3) precluded this finding in earlier studies. In a study by Yoon and Guidotti (9) using epitope insertion and expression in COS-7 cells, it was concluded that the carboxyl-terminal region contained four transmembrane segments. Evidence was lacking for the final two segments after Val 939 ; thus it was concluded that the ␣-subunit had eight transmembrane segments. This conclusion was repeated in a more recent work from Lee and Guidotti (12) who inserted a tag between residues 973 and 974 and determined, using immunofluorescence, that residue 973 was internal. This recent finding is contradicted by our current observation that Met 973 is exposed at the extracellular surface, as well as by our earlier suggestion about the membrane location of a segment beginning at Met 973 based on tryptic digestion studies of membrane-associated peptides (8). Similar conclusions, that the ␣-subunit contained eight transmembrane segments, were also reached by Canfield et al. (10) based on their epitope addition studies showing that residues beyond 978 were all intracellular, whereas residue 953 was extracellular. Their study also found that there were no transmembrane segments between 832 (cytoplasmic) and 895 (cytoplasmic), which is at odds with our finding of the extracellular locations of residues Leu 876 and Met 973 . It seems likely that all eight transmembrane segment models can now be disregarded.
The V969C and L976C mutants were generated to further define the M9M10 loop, and our data showed that these two residues are not freely accessible from the extracellular medium (Fig. 4). These observations establish the very small size of this extracellular loop (it probably only consists of residues Ala 970 -Pro 975 ) and show that the M9 and M10 helices must be closely apposed. The relative lack of accessibility of V969C and L976C compared with M973C demonstrates that our approach is capable of distinguishing accessibility of residues separated by only three or four residues and may be useful in establishing the locations of the membrane interface in the protein structure. The location of these important membrane-interface regions is difficult to predict from limited structural knowledge and has proven to be difficult to define experimentally using other methods.
Our previous study on the labeling of purified renal Na,K-ATPase with 4-acetamino-4Ј-maleimidylstilbene-2,2Ј-disulfonic acid (a charged and nonpenetrating maleimide) and 7-diethylamino-3-(4Ј-maleimidyl)-4-methylcoumarin (a penetrating maleimide) showed that Cys 964 is accessible from the extracellular space when the enzyme is in a phosphorylated form (17). Upon dephosphorylation and K ϩ occlusion, however, Cys 964 is no longer exposed to the aqueous phase, suggesting the dynamic nature of the M9M10 loop region. It is quite possible that the positions of Val 969 and Leu 976 are also sensitive to the protein conformation and that their accessibility alters as the different ligands bind and dissociate. The present studies utilized labeling in intact cells where the protein is turning over and control of protein conformation is limited. Nevertheless, the M973C mutant is consistently labeled, and in future studies where labeling will be performed in isolated membranes, it will be possible to test the issues of conformation-dependent mobility of specific residues.
In summary, we have utilized a Cys substitution strategy to establish the membrane topology of the ␣-subunit of the Na,K-ATPase. The ␣-subunit has 10 transmembrane segments. Our approaches can be utilized to test whether or not the accessibility of a particular residue close to the membrane aqueous interface alters during the reactive cycle.