Identification of the Site of Inhibition by Omeprazole of a α-β Fusion Protein of the H,K-ATPase Using Site-directed Mutagenesis*

The α subunit of eukaryotic P-type ATPases has ten experimentally defined transmembrane or membrane inserted segments. The fifth and sixth of these are short, not predicted by hydropathy analysis, do not insert independently into microsomal membranes, and are readily removed after tryptic digestion and therefore may be membrane inserted sequences. Acid transport by the gastric H,K-ATPase is covalently inhibited by several substituted pyridyl methylsulfinyl benzimidazoles, such as omeprazole. These act as probes of accessible extracytoplasmic thiols because they are accumulated in the acid transporting gastric vesicles and then convert to thiol reactive, cationic tetracyclic sulfenamides. Inhibition is due mainly to disulfide formation with Cys813 or Cys822in M5/6 and perhaps with a contribution from Cys892 in the loop between transmembrane segment (TM) 7 and TM8. Identification of the specific cysteine responsible for inhibition should be able to define the turn between M5 and M6. The gastric H,K-ATPase α-β heterodimer was expressed as a fusion protein in HEK 293 cells. Transient transfection resulted in most of the protein being retained in the endoplasmic reticulum with only core glycosylation and minor activity of the ATPase evident. Stable transfection resulted in plasma membrane localization of the protein and complex glycosylation. The transfected but not the control cells displayed cation-stimulated, SCH 28080-inhibited ATPase activity and SCH 28080- and omeprazole-inhibited86Rb uptake. The two cysteines in M5/6 and Cys892 in the TM7/8 loop were mutated to the amino acids found in the Na,K-ATPase in order to determine which of the three cysteine residues were important for benzimidazole inhibition. Mutation of one, two, or all three cysteines did not alter enzyme activity,86Rb transport, or SCH 28080 inhibition. Only removal of Cys822 blocked omeprazole inhibition of 86Rb transport. These data suggest that Cys822 is present in a region of the enzyme most easily accessed by the cationic sulfenamide formed by omeprazole, presumably the turn between M5 and M6.

Eukaryotic small cation P-type ATPases and the Mg-ATPase of Salmonella typhimurium have 10 experimentally established transmembrane or membrane insertion sequences (1)(2)(3)(4)(5)(6). The phosphorylation and ATP-binding sites are in the large cytoplasmic loop between transmembrane segment (TM) 4 and TM5. 1 Mutational analysis has provided evidence for a role for hydrophilic and carboxylic amino acids in cation occlusion and transport particularly in the fifth and sixth membrane domain and in TM4 and TM7. From proteolysis results M5 and M6 are relatively short and do not insert independently into microsomal membranes in in vitro translation systems and in vivo expression in Xenopus oocytes (7)(8)(9). This membrane pair is relatively easily removed from the membrane after trypsinolysis (1,10). It has been suggested that this is a mobile segment in these P-type ATPases (11).
Investigation of the structure of P-type ATPases has relied largely on biochemical and molecular biological methods. Proteolysis, investigation of inhibitor-binding sites, expression of fusion proteins, and in vitro translation have been combined to provide a 10-membrane inserted segment model. (1,3,5,6). Significant progress has been made in the functional expression of different isoforms and mutants of the Na,K-ATPase (12). The expression of ouabain-insensitive forms of the enzyme in various mammalian cell lines (13)(14)(15)(16) has allowed the analysis of mutants that affect 86 Rb ϩ uptake (15,17) as well as cation affinity and ouabain binding (18,19). Extensive mutational studies to define structural elements of a P-type ATPase have also been done on the sr Ca-ATPase (20). Several hundred residues have been mutated, identifying functionally important residues for ATP-binding, phosphorylation, ion-binding, and inhibitor sites (20 -23).
The H,K-ATPase and the Na,K-ATPase consist of two tightly associated but noncovalently linked subunits, a large ␣ or catalytic subunit, which contains the phosphorylation and ATPbinding domains, and a smaller heavily glycosylated ␤ subunit. The coordinate synthesis of both subunits is obligatory for functional expression and plasma membrane targeting of the H,K-ATPase as well as the Na,K-ATPase (24 -30). Expression of the gastric H,K-ATPase in mammalian cell lines appears to require simultaneous expression of both subunits.
Inhibition of the Na,K-ATPase is used in the medical treatment of congestive heart disease, whereas inhibition of the gastric H,K-ATPase is used in the treatment of peptic ulcer disease. This latter inhibition is achieved by substituted pyridyl methylsulfinyl benzimidazoles like omeprazole, lansoprazole, pantoprazole, or rabeprazole. These are weak bases with a pK a of about 4.0 -5.0 that accumulate in the secretory canaliculus of the parietal cell, which has a pH of about 1.0. In this acidic environment the compounds convert to cationic, tetracyclic, planar sulfenamides. These are strong electrophiles that form stable disulfides with cysteine groups on proteins. The sulfenamides formed in the secretory canaliculus or on the interior face of isolated acid transporting vesicles are therefore targeted to cysteines accessible from the extracytoplasmic surface of the H,K-ATPase ␣ subunit (31). There are two cysteines (Cys 813 and Cys 822 ) in the region encompassing M5 and M6 and one cysteine (Cys 892 ) in the loop region between TM7 and TM8 that have been implicated in omeprazole inhibition of the ATPase. In the gastric H,K-ATPase of the two cysteines present in the TM5/loop/TM6 region, one is surrounded by relatively hydrophobic amino acids (Cys 813 ), and one is in the middle of a hydrophilic pocket, Glu-Leu-Cys-Thr-Asp.
Recent work has shown that one of the two cysteines in M5/6 is probably the cysteine essential for inhibition. Labeling of the gastric H,K-ATPase with radioactive omeprazole under acid transport conditions, digestion with trypsin, and SDS solubilization with labeling of unreacted SH groups with fluorescein 5-maleimide was followed by Tricine gradient SDS-polyacrylamide gel electrophoresis. Then isolation of the fragment containing M5/6, thermolysin digestion, and SDS Tricine gradient gel electrophoresis was used in order to define which cysteine of the two present in M5/6 was labeled. These data were interpreted as showing that Cys 813 was the stable covalent binding site for omeprazole (31). Similar data were obtained using reaction with N-ethylmaleimide after omeprazole binding, reduction, and FMI labeling. 2 Because protonated omeprazole and its reactive sulfenamide are relatively bulky cations, it is likely that the labeled cysteine is exposed to the luminal surface of the ATPase. Mutational analysis of hydrophilic amino acids in the Na,K-and Ca-AT-Pases has provided the basis for a topographic model of the M5/6 region of these enzymes. Alignment of the H,K-ATPase with these models of the Na,K-and Ca-ATPases would make Cys 813 the more exposed of the two in the loop connecting M5 and M6, consistent with the labeling data.
In the absence of N-ethylmaleimide added during digestion and solubilization, the incorporated label is relatively unstable. It is possible that either cysteine could react and that the disulfide bond on Cys 813 survives more effectively than the disulfide bond on Cys 822 during the isolation procedure. Disulfide exchange during the different steps in the isolation procedure or unrecognized variability in cleavage of the region between the two cysteines by thermolysin are other possibilities raising questions as to the role or location of Cys 813 . In the case of pantoprazole, it appears that both cysteines are able to react (32). It is also possible that both cysteines are accessible and that a transmembrane hairpin model for M5 and M6 is an oversimplification.
An alternative strategy for identification of the cysteine important for omeprazole inhibition is to remove the cysteines in extracytoplasmic region of the C-terminal domain by site-directed mutagenesis and determine which mutation prevents omeprazole inhibition of 86 Rb transport in cells transfected with the H,K-ATPase while retaining cation stimulated ATPase activity and inhibition of ATPase and cation transport by the K competitive reagent, SCH 28080.
The construction of a gastric H,K-ATPase composed of a fusion protein of the ␣ and ␤ subunit, done successfully for the Na,K-ATPase (33), was accomplished in order to bypass the complexities of co-translational assembly and post-translational processing of a two-subunit heterodimer. These ␣-␤ fusion vectors were made with different mutations in Cys 813 , Cys 822 , and Cys 892 . These fusion vectors were stably expressed in HEK 293 cells. NH 4 ϩ -stimulated, SCH 28080-inhibited ATPase activity and Western blot analysis as well as confocal microscopy were used to confirm enzyme activity and plasma membrane expression. Inhibition of 86 Rb uptake by acid-activated omeprazole as compared with inhibition of uptake by SCH 28080 was used to determine which cysteine was essential for inhibition by the compounds after acidification of the drugs. None of the cysteines were essential for enzyme activity or cation transport, but the cysteine at position 822 must be present for omeprazole inhibition of transport by the gastric H,K-ATPase. If this cysteine is the more exposed of the two and is in the loop connecting M5 and M6, an unconventional model for this region of the enzyme ensues.

EXPERIMENTAL PROCEDURES
Construction of the Vector for the H,K-ATPase ␣-␤ Fusion Protein-cDNAs coding for the rabbit H,K-ATPase ␣ and ␤ subunit (34, 35) (GenBank accession numbers X64694 and M35544) were inserted into the multiple cloning site of the mammalian expression vector pcDNA3.1 (Invitrogen). A modified DNA coding for the full-length ␤ subunit with Glu instead of Met at amino acid position 1 was generated by using polymerase chain reaction and a mutagenic sense primer. This resulted in the introduction of XhoI restriction site at the beginning of the ␤ sequence. Using the XhoI/XbaI restriction sites, the polymerase chain reaction product was then ligated into the pcDNA3.1 vector containing the cDNA for the ␣ subunit. A synthetic oligonucleotide coding for the last 17 amino acids of the ␣ subunit was extended by DNA encoding an additional 18 amino acids as had been done for the Na,K-ATPase (33). Ligation into the HindIII/XhoI restriction site resulted in a cDNA coding for a polypeptide with ␣ and ␤ subunits linked by a hydrophilic sequence of 18 amino acids (Fig. 1).
Site-directed Mutagenesis of Cys 813 , Cys 822 , and Cys 892 -The coding sequence of the H,K ␣-␤ fusion protein was cloned into a CLONTECH pALTER-EX1 vector. Amino acid substitutions for these cysteines were selected based on the amino acids in the Na,K-ATPase. Mutagenesis was performed according to the manufacturer (CLONTECH, pALTER Mutagenesis System). The mutagenic primer used for Cys 813 3 Thr was GGGCACCATCACCATCCTCTTCAT, that for Cys 822 3 Gly was TCATAGAACTCGGCACCGACATT, and that for Cys 892 3 Leu was CTGGTTCCCGCTGCTGTTAGTGG. Mutants were prepared containing single mutations in each of the three cysteines. Four additional mutants were prepared containing the mutations Cys 813 3 Thr ϩ Cys 822 3 Gly, Cys 813 3 Thr ϩ Cys 892 3 Leu, Cys 822 3 Gly ϩ Cys 892 3 Leu, and Cys 813 3 Thr ϩ Cys 822 3 Gly ϩ Cys 892 3 Leu. The mutant cDNAs were cloned back into pcDNA3.1. Each final construct was sequenced before use for transfection using dideoxy sequencing.
Cell Culture, Transfection, and Selection of Stable Cell Lines-HEK 293 (ATCC CRL 1573) cells were maintained in a humidified incubator at 37°C in 95% O 2 /5% CO 2 in Dulbecco's essential medium supplemented with 10% fetal calf serum, 50 units/ml penicillin, 50 g/ml streptomycin, and 2 mM L-glutamine. Single cells were plated on 100-mm diameter tissue culture dishes. Cells were grown until the monolayer covered ϳ70% of the surface of the dish. Transfection was done by the calcium phosphate method (36) using 10 g of plasmid for each culture dish. For transient transfections, maximal expression of recombinant protein was achieved after 48 -72 h.
Stable cell lines were selected by addition of 0.4 mg/ml Zeocin (Invitrogen) 60 h after transfection. This concentration was maintained until single colonies appeared. Colonies were isolated and grown in 6-well Falcon plates containing culture medium supplemented with 0.1 mg/ml Zeocin. Each clone was screened by Western blot analysis as described below.
Membrane Isolation, SDS-Polyacrylamide Gel Electrophoresis, and Western Blot Analysis-Cells from transient transfections or stable cell lines, grown to confluence, were resuspended in buffer A (10 mM PIPES/2 mM EGTA, pH 7.4). Homogenization was performed using a N 2 cavitation bomb. The cell homogenate was centrifuged at 900 ϫ g for 10 min. The supernatant was layered onto 40% (w/v) sucrose and spun in a Beckman SW41 swinging bucket rotor at 25,000 rpm for 1.5 h at 4°C. The fraction at the interface was collected and resuspended in 15 ml of the buffer A. Membranes were collected by centrifugation in a Beckman 75Ti rotor (35, Gel sample buffer (4% SDS, 0.05% bromphenol blue (w/v), 20% glycerol, 1% ␤-mercaptoethanol (v/v) in 0.1 M Tris buffer, pH 6.8) was mixed with 20 g of membranes and incubated for 5 min at room temperature. The sample was loaded on a 5% SDS-polyacrylamide gel. As a standard for native ␣ and ␤ subunits, 200 ng of purified gastric vesicles (G 1 fraction; Ref. 40) were used. High molecular mass standards (Bio-Rad) were loaded on each gel.
Quantitative Western Blot Analysis-An estimate of the specific activity of the expressed fusion protein can be achieved by optical quantitation of Western blots of expressed protein as compared with a hog gastric H,K-ATPase standard. Various amounts (100 ng to 1 g) of purified hog gastric H,K-ATPase (40) were loaded side by side with 20 g of membranes containing the expressed ␣-␤ fusion protein onto a 5% SDS-polyacrylamide gel for quantitative Western blot analysis. Protein transfer on to nitrocellulose membranes and Western analysis was carried out as described above. All blots were scanned using the AMBIS optical imaging system (AMBIS Inc., San Diego, CA), and the optical density of the bands was quantified. The amount of the ␣-␤ fusion protein was calculated by comparison with the densities of the H,K-ATPase standards on the same blot. NH 4 ϩ -stimulated ATPase Activity-ATPase assays were performed by the method of Yoda and Hokin (41). Membranes (20 g of protein) were added to a reaction mixture containing (final concentrations) 40 mM Tris-HCl, pH 7.4, and 2 mM MgCl 2 with or without 60 mM NH 4 Cl. NH 4 ϩ is an effective surrogate for K ϩ in the gastric H,K-ATPase and does not require an ionophore for access to the extracytoplasmic face of vesicular enzyme (6). Inhibitors including 1 mM EGTA (Ca-ATPases), 1 mM ouabain (Na,K-ATPase), 1 M oligomycin (mitochondrial ATPase), 10 nM bafilomycin (V-type ATPases), and 100 nM thapsigargin (endoplasmic reticulum Ca-ATPase) were added to suppress the activities of possible contaminating ATPases. Specific H,K-ATPase activity was measured for basal enzyme and the NH 4 ϩ -stimulated enzyme by addition of 10 M SCH 28080, a selective competitive inhibitor of the gastric H,K-ATPase (42). All reactions were started by adding 2 mM ATP. The reactions were terminated after 10 min at 37°C by adding an equal volume of stop solution containing 4.5% ammonium molybdate (w/v) and 14% perchloric acid. Reaction mixtures were extracted with butyl acetate, and P i released was quantified by A 320 nm . The specific activity was expressed in mol P i /mg/h. The results of quantitative Western blot analysis were used to calculate the approximate specific activity of the expressed ␣-␤ fusion protein.
Rubidium Transport Assay-Ion transport in stably transfected HEK 293 cells (or untransfected cells) was measured by uptake of 86 Rb ϩ as a surrogate for K ϩ . A similar method has been used by Blostein et al. (43). Cells were grown until confluence in 6-well Falcon plates, the medium was removed, and the wells were washed once with 4 ml of ice-cold wash solution containing 144 mM NaCl and 5 mM Hepes, pH 7.4. Each well was incubated in the presence of 1 mM ouabain (to inhibit Na,K-ATPase) and 10 M furosemide (to inhibit NaKCl 2 co-transport) for 10 min at 37°C with 1 ml of a solution containing 144 mM NaCl, 5 mM Hepes, 0.5 mM MgCl 2 , 0.5 mM CaCl 2 , 50 M phenolphthalein, 1 mM Rb ϩ , and 3 ϫ 10 6 cpm 86 Rb ϩ . After 10 min the supernatant was removed, and each well was washed three times with 4 ml of ice-cold wash solution. Rubidium transport was carried out in the absence and in the presence of 100 M SCH 28080. Omeprazole is 5-methoxy-2-[(4methoxy-3,5-dimethyl-2-pyridyl)methylsulfinyl]-1 H-benzimidazole and requires an acidic pH for rapid conversion to the active sulfenamide. Acid activation was achieved by dissolving 1 mM omeprazole in 1 N HCl. The sulfenamide is relatively stable at pH 1.0. 50 l was added to the appropriate wells to give a final concentration of 10 M and after 5 s a pH of 7.4 was restored by addition of 50 l of 1 M Tris. Rb ϩ uptake experiments were then carried out as described above.
At the end of the experiments, cells were dissolved by incubation for 1 h with 2 ml of 0.5 N NaOH. Solutions (1 ml) were counted using Cerenkov radiation. Total protein/well was measured using the Bradford assay as described above. Rb ϩ uptake by the cells was expressed as nmol 86 Rb ϩ /min/mg total protein. Inhibitor-sensitive flux was calculated by subtracting Rb ϩ uptake with inhibitor for the H,K-ATPase from Rb ϩ uptake without inhibitor.
Immunofluorescence and Confocal Laser Scanning Microscopy-Cells showing expression of the ␣-␤ fusion protein as shown by Western blot analysis were chosen for immunofluorescence studies. Cells were transferred to chamber slides (LabTek, Nunc Inc., Naperville, IL) and grown to confluence in culture medium containing 0.1 mg/ml Zeocin. Cells were fixed for 10 min in methanol at Ϫ20°C. Slides were washed three times in phosphate buffer (0.1 M, pH 7.4) and incubated for 2 h at room temperature in blocking solution containing 5% bovine serum albumin (w/v), 0.5% Triton X-100, and 10% normal goat serum (Life Technologies, Inc.) (v/v) dissolved in 0.1 M phosphate buffer. Slides were then incubated overnight at 4°C in the primary antibody solution, which was either monoclonal antibody 12.18 or monoclonal antibody 2B6, diluted 1:250 in phosphate buffer. The solution was removed, and the slides were washed for 20 min in phosphate buffer. The secondary antibody solution (anti-mouse IgG, raised in goat against heavy and light chains, conjugated to fluorescein isothiocyanate Calbiochem, San Diego, CA, 1:50 diluted in phosphate buffer) was added, and the slides were incubated at room temperature for 2 h. After a final wash with phosphate buffer for 20 min, coverslips were mounted using 90% glycerol (v/v) in phosphate buffer as mounting medium. Immunofluorescent images were analyzed using a Leitz/Leica Fluovert FU inverted microscope fitted with the Leica CLSM and a confocal fluorescence scanning adapter. The objective used was a PL Fluotar 100x/1.32. Photomultiplier amplification was kept constant in each experiment. Images were normalized over 16 scans, and data collection was operated by a Motorola MVME147 (MC68030) computer system running an OS-9 operating system (V2.4). Normalized images were processed and printed out using Adobe Photoshop 2.01 (Adobe Systems, Inc.) imaging software.
Materials-All chemicals were analytical grade. The Taq polymerase (AmpliTaq) was purchased from Perkin-Elmer. The pcDNA3.1 vector and Zeocin were obtained from Invitrogen. The pALTER-EX1 mutagenesis system was purchased from CLONTECH.

RESULTS
In Vitro Translation of the ␣-␤ Fusion Vector-The ␣-␤ fusion protein should be core glycosylated in the endoplasmic reticulum and processed in the Golgi during its passage to the plasma membrane. Core glycosylation should be observed during in vitro translation in the presence of microsomes due to transfer of carbohydrate from the high mannose dolichol precursor to the glycosylation consensus sequences on the ␤ subunit that terminates the ␣-␤ fusion protein. This provided a convenient means of testing the competence of the vector prior to transfection.
The left-hand panel of Fig. 2 shows an autoradiogram from an in vitro translation of the vector containing the cDNA for the fusion protein in the presence of [ 35 S]methionine as described elsewhere (7). When the technique is carried out without microsomes, a single band appears with a molecular mass of 135 kDa (Fig. 2, lane 1). With the addition of microsomes a second band with a higher molecular mass at 152 kDa is seen (Fig. 2,  lane 2). An increase in mass of 15 or 19 kDa would be expected for glucose-trimmed or untrimmed core glycosylation of seven N-linked glycosylation sites, respectively. The observed increase in mass therefore shows the presence of core glycosylation indicating a competent vector.
Transfection of HEK 293 Cells-Western blot analysis of membranes of transiently transfected HEK 293 cells with the H,K-ATPase ␣-␤ fusion protein cDNA in pcDNA3.1 using either mAb 12.18 against the ␣ subunit or mAb 2B6 against the ␤ subunit (Fig. 2, lane 3) shows essentially the same pattern as in in vitro translation experiments with only core glycosylation. The majority of the protein following transient transfection is therefore probably retained in the endoplasmic reticulum. The results with the two antibodies were identical; hence results with mAb 2B6 are shown.
Isolation of membranes of Zeocin-resistant cell clones with stable transfection showed positive Western blot staining for the H,K-ATPase with mAb 2B6 as well as mAB 12.18. In contrast to the results with transient expression and in vitro translation, the gel now shows an additional band with a molecular mass centered at 163 kDa, and the band corresponding to nonglycosylated protein disappears (Fig. 2, lane 4). The most likely interpretation is that the 163-kDa band represents processed glycosylated ␣-␤ fusion protein. Treatment of these membranes with glycosidases showed a endoglycosidase H-insensitive but endoglycosidase F-sensitive deglycosylation of the expressed fusion protein, which confirms complex glycosylation (Fig. 3).
Localization of the ATPase-To determine the localization of the expressed ATPase, we used confocal laser scanning microscopy following immunostaining with either mAb 12.18 recognizing an epitope on the cytoplasmic surface of the ␣ subunit or mAb 2B6, which recognizes an extracytoplasmic epitope in the ␤ subunit (amino acids 236 -281). Fig. 4 shows four sections of a typical HEK 293 cell that has been co-transfected with two pcDNA3.1 vectors containing the rabbit gastric H,K-ATPase ␣ and ␤ subunit. These cells were stained with mAb 12.18. This image shows that the majority of the ␣ subunits is retained in the endoplasmic reticulum. This image is similar to that found previously using co-transfection of the two subunits in LLC-PK cells (28).   5 shows four sections through a cluster of HEK 293 cells stably transfected with the ␣-␤ fusion protein vector and then stained with mAb 2B6. It can be seen that a major fraction of the immunofluorescence is present at the periphery of the cells presumably at or close to the plasma membrane in contrast to the results shown above with co-transfection. Some fluorescence is also visible intracellularly. Fig. 6 shows four sections through a single HEK 293 cell, again showing significant peripheral localization of the ␣-␤ fusion protein. No immunostaining was observed with untransfected HEK 293 cells or when the primary antibody was absent (data not shown).

Content of Expressed H,K-ATPase in the Membrane
Fraction-Western blots using mAb 2B6 against the ␤ subunit were quantified in order to estimate the amount of labeled H,K-ATPase ␣-␤ fusion protein. This enabled comparison of the specific activity of the ␣-␤ fusion protein as compared with the ␣ ␤ heterodimer. The hog gastric H,K-ATPase fraction (G 1 ) was used at several concentrations (200, 400, 600, 800, and 1000 ng) as standard, assuming that this fraction is about 85% pure ATPase protein (40). The calibration curve was linear (data not shown). The amount of ␣-␤ fusion protein in the membrane fraction was equivalent to 11.6 ng of ␣-␤ fusion protein/g membrane protein). Similar amounts were found for each of the expressed mutants (Fig. 7). NH 4 ϩ -stimulated ATPase Activity- Fig. 8 shows the amount of NH 4 ϩ -stimulated, SCH 28080-sensitive ATPase activity of total membrane protein of cells, transfected with the wild type ␣-␤ fusion protein or each of the mutants. No inhibition by SCH 28080 was found in untransfected cells. The NH 4 ϩ -stimulated, SCH 28080-inhibited ATPase activity of the wild type or mutated fusion proteins were between 1.0 and 1.2 mol/mg protein/h. Given the amount of H,K-ATPase protein in the membrane fraction, the specific activity of the heterologous enzyme is calculated to be 86 mol/mg/h for the NH 4 ϩ stimulation, about one-half to two-thirds of the activity measured for the hog H,K-ATPase preparation in the same experiment. Each of the mutants expressed an activity comparable with that of the native fusion protein; hence ATPase activity was not affected by introduction of the mutated amino acids.
Rubidium Uptake- Fig. 9 details the results of 86 Rb ϩ uptake experiments. 100 M SCH 28080 or preincubation for 5 s with 10 M acid-activated omeprazole had no significant effect on rubidium uptake in untransfected cells. The addition of 100 M SCH 28080 inhibited the rubidium uptake of all cells transfected with either the wild type fusion protein or the mutants. Omeprazole is a prodrug and after acid activation forms a cationic sulfenamide that is a sided covalent thiol reagent (31). Addition of acid-activated omeprazole to these intact cells allows selective reaction with the outside face of the enzyme, given the relative membrane impermeability of the cationic sulfenamide and protective intracellular SH moieties such as those in glutathione. Preincubation for 5 s with 10 M acidactivated omeprazole inhibited the 86 Rb uptake of the cells transfected with the wild type fusion protein or the mutant Cys 813 3 Thr as well as the mutant Cys 892 3 Leu. In contrast, preincubation with 10 M acid-activated omeprazole had no significant effect on the 86 Rb uptake of cells transfected with the mutants Cys 822 3 Gly, Cys 822 3 Gly ϩ Cys 892 3 Leu, or Cys 813 3 Thr ϩ Cys 822 3 Gly ϩ Cys 892 Leu. DISCUSSION Whereas much has been learned about the secondary structure of the gastric H,K-ATPase by biochemical means such as proteolysis and site-specific labeling as well as in vitro translation (6,7,31), an effective model for effects of mutagenesis on transport has not been available, although changes in enzyme activity have been described as a result of site-specific mutations (44,45) In our laboratory, the co-transfection of ␣ and ␤ subunits of the gastric H,K-ATPase was not successful in producing a plasma membrane heterodimer in quantities sufficient for transport assay. It had been possible to express functional Na,K-ATPase in the plasma membrane by generating a fusion protein between its ␣ and ␤ subunits (33). Based on this observation, it seemed that functional expression might be achieved by constructing a similar fusion protein using the two subunits of the gastric H,K-ATPase. This would avoid the need for transfection with separate vectors and measurement of the levels of each subunit in each experiment and was expected to produce a more uniformly transfected cell population. There are as yet no known methods for selecting cells expressing the H,K-ATPase, in contrast to the ouabain inhibition methods used for the Na,K-ATPase (13).
Inspection of the linker sequence used in the ␣-␤ fusion protein of the Na,K-ATPase (33) showed both hydrophilicity and cytoplasmic anchoring properties due to the three arginines present. The length of 18 amino acids also allowed both subunits to associate competently. A previous construct we had made of the gastric H,K-ATPase containing a linker of only two amino acids was not expressed in the plasma membrane. Apparently the 39 amino acids predicted to precede the transmembrane segment of the ␤ subunit require more than 2 amino acids in the linker region to allow functional association between the two subunits. Therefore we used the identical sequence to construct the cDNA encoding an ␣-␤ fusion protein of the H,K-ATPase as had been used for the Na,K-ATPase. This fusion protein is expressed in HEK 293 cells in the plasma membrane and is functional in terms of ATPase activity and ion transport. The latter property allowed investigation of the cysteines important for omeprazole inhibition.
Western blots of membranes derived from transiently transfected cells with the ␣-␤ fusion protein (Fig. 2, lane 3) and radiographs from in vitro translation products of the ␣-␤ fusion protein vector had two bands of immunoreactivity in transiently transfected cells. One band is unglycosylated with a relative molecular mass of 135 kDa, and the other core glycosylated with a molecular mass of 155 kDa. Most of the expressed fusion protein is still located in the endoplasmic reticulum in these transiently transfected cells.
In the stable cell lines, the pattern of the antibody-labeled bands in the Western blot analysis changed. No band was found with the molecular mass of 135 kDa, but an additional band was found with a molecular mass of 163 kDa. Given the anomalous molecular mass of the ␣ subunit, we would expect the glycosylated ␣-␤ fusion protein to display a relative molecular mass of about 155-180 kDa, because the relative molecular mass for the ␣ subunit is 95 kDa (46), and that for the mature ␤ subunit is 60 -85 kDa (47,48). The relative molecular mass found for chemically cross-linked ␣-␤ heterodimers of the native hog gastric H,K-ATPase appeared to be 170 kDa (49), close to the 163-kDa band found in the stable cell lines. En-doglycosidase H-insensitive but endoglycosidase F-sensitive deglycosylation of the expressed fusion protein confirms the presence of complex glycosylation. Hence the ␣-␤ fusion product in stably transfected cells is within the range expected for the fully glycosylated protein.
We concluded from these findings that the ␣-␤ fusion protein is able to leave the endoplasmic reticulum and be transported to distal Golgi compartments and/or the plasma membrane in the stable cell lines. It also seems that a two-step maturation of the protein occurs in these transfects. The first step is a core glycosylation, which appears throughout the time of the transient transfection (60 -72 h), the second step is the transport of the protein to distal Golgi compartments and the plasma membrane with full glycosylation. Longer culture times appear to be necessary to express adequate levels of the mature protein in the plasma membrane.
In order to confirm that the ␣-␤ fusion protein was located in the plasma membrane, confocal laser scanning microscopy was applied to antibody-labeled cells stably transfected with ␣-␤ fusion protein. Images obtained from this study showed a localization of the protein that was consistent with its presence on or close to the plasma membrane.
The membrane fraction prepared from stably transfected

cells expressed NH 4
ϩ -stimulated enzyme activity. Addition of 10 M SCH 28080 inhibited the ammonium-stimulated enzyme activity by 95% in the presence of a large number of inhibitors of other ATPases. This concentration of SCH 28080 is selective for the gastric H,K-ATPase (42), and no inhibition of ATPase activity was observed in membranes obtained from untransfected cells. Given the fact that the membrane fraction with the expressed heterologous protein contains approximately 1/100th of the amount of the wild type ␣-␤ fusion protein as shown by quantitative Western blot analysis, the specific activity of expressed rabbit H,K-ATPase ␣-␤ fusion protein is calculated to be 86 mol/mg/h. This is about one-half to two-thirds of that generally found for hog H,K-ATPase but is similar to that seen for the rabbit preparation, which is, however, less pure than that from the hog (50). Hence the fusion of ␣ and ␤ polypeptides (and the presence of the 18 residues) not only enables expression but produces a protein with a specific activity similar to that of the ␣ and ␤ heterodimer.
Rubidium flux assays in the transfected HEK 293 cells showed that the enzyme was functionally expressed in the plasma membrane and catalyzed ion transport. HEK 293 cells stably transfected with the ␣-␤ fusion protein showed both a SCH 28080-and omeprazole-inhibitable Rb ϩ transport into the cells in the presence of 1 mM ouabain and 10 M furosemide. This was not observed with untransfected HEK 293 cells. Rabeprazole, a newer inhibitor of acid secretion, also inhibited Rb transport (data not shown). Hence the ATPase expressed in these cells appears to be functional in terms of NH 4 ϩ -stimulated ATPase activity and ion transport.
This system provided a tool to identify the functional site of inhibition by omeprazole of the H,K-ATPase. From earlier studies, binding to cysteines in M5/6 appeared to correlate with inhibition of the enzyme. Treatment with thiol reducing agents reversed the inhibition, showing that the covalent bond was a disulfide (6). Omeprazole was also able to react with Cys 892 in the loop between TM7 and TM8 (31). We therefore replaced the two cysteine residues shown to bind omeprazole in the M5/6 domain (Cys 813 and Cys 822 ) and in the extracytosolic loop between TM7 and TM8 (Cys 892 ) with the amino acids found in these positions in the Na,K-ATPase.
All mutant ␣-␤ fusion proteins expressed NH 4 ϩ -stimulated and SCH 28080-inhibited ATPase activity at a level similar to that of unmutated protein. In addition, the rubidium uptake in cells transfected with the wild type or mutant ␣-␤ fusion protein was inhibited by SCH 28080 and was quantitatively similar.
The mutants containing the exchange of Cys 813 3 Thr or Cys 892 3 Leu showed inhibition by omeprazole of the rubidium uptake equal to that of the unaltered fusion protein. In contrast, rubidium uptake by monolayers of cells transfected with the mutants containing the replacement Cys 822 with Gly alone or in combination with exchanges of Cys 813 or Cys 892 could not be inhibited by omeprazole under our assay conditions.
If the residue Cys 822 is the important residue for inhibition of the H,K-ATPase with omeprazole or rabeprazole, the residue has to be accessible from the luminal side of the pump. The best indication for the cytoplasmic boundaries of the membrane pair 5 and 6 is given by protease digestion sites of the H,K-ATPase and the endoplasmic reticulum Ca-ATPase, which are readily prepared as cytoplasmic side out vesicles. After digestion of the H,K-ATPase with trypsin, two peptides of 5 and 5.9 kDa are found. The 5.9-kDa band is found predominantly after a short digestion of 20 min. The N-terminal sequence showed that there was a digestion site after Lys 783 ((K/S)IAY). The 5.9-kDa fragment then should extend to Lys 835 (LAYEK) based on molecular mass. After longer digestion, after the addition of dithiothreitol, or after digestion following omeprazole inhibition, a second shorter peptide is found. The N-terminal sequence of this 5.0-kDa peptide showed that the digestion site was after Lys 791 of the H,K-ATPase ((K/N)IPE) probably extending again to Lys 835 (LAYEK) (31). Placement of Cys 822 in the hairpin loop between transmembrane sequences representing M5 and M6 would place tryptic cleavage at Lys 835 within the transmembrane domain.
These data are similar to those found for the sr Ca-ATPase (51). Digestion of the sr Ca-ATPase with proteinase K resulted in digestion after asp 818 . The latter experiment showed also a small percentage of cleavage after Gly 808 . Trypsin digestion of the same ATPase (5) also showed a digestion site after Arg 762 ((R/Y)LISS).
If the membrane pair 5 and 6 consists of the amino acid sequence 763-808 for the sr Ca-ATPase and 791-835 for the H,K-ATPase, the usual helix-loop-helix structure is relatively short with only a small joining loop. In vitro translation experiments have failed to show either signal anchor or stop transfer properties for M5 or M6 for the H,K-ATPase (7) and weak stop transfer properties for the TM6 region of the endoplasmic reticulum Ca-ATPase (8). Similarly, in vivo translation of constructs for these regions of either the Na,K or H,K-ATPases in Xenopus oocytes did not show any membrane inserting properties (52). 3 Furthermore, the variability of the trypsin digestion sites before M5 in the case of the H,K-ATPase and of the proteinase K digestion sites after M6 in the case of the sr Ca-ATPase suggests varying accessibility of the proteases to these locations. Perhaps these are due to conformational changes happening after the initial scission of this region of the protein, and the actual membrane regions are present between amino acids 763 and 817 for the sr Ca-ATPase and between amino acids 784 and 833 for the H,K-ATPase.
An importance of the carboxylic acid residues after the putative M6 has been reported for the H,K-ATPase (44). Mutation of the three negatively charged residues in this part of the sequence (LAYEKAESDIM) abolished the SCH 28080-sensitive phosphorylation of the H,K-ATPase. One study reported loss of activity of the sr Ca-ATPase with concomitant replacement of these conserved carboxylic acids as well (9). However, most of the reported mutational studies for the sr Ca-ATPase and the Na,K-ATPase investigated residues thought to be within the putative transmembrane segment 6 but not after (16,(21)(22)(23)53).
The negatively charged aspartate at position 824 after Cys 822 (ELCTD) is conserved throughout the eukaryotic P-type ATPase family. The carboxylic acid preceding leucine at position 820 is an aspartate in the Na,K-ATPases and is a hydrophilic asparagine in the sr Ca-ATPase. This is therefore a retained hydrophilic sector in this region of these P-type AT-Pases. Mutation of these residues abolishes ion binding and activity of the Na,K-ATPase as well as the sr Ca-ATPase (22,23,53). Even exchange of Asp 3 Glu abolished calcium occlusion and function of the endoplasmic reticulum Ca-ATPase (23). It seems that binding of a bulky charged inhibitor to Cys 822 in the middle of this hydrophilic pocket would have an inhibitory effect by blocking the entry or binding of the transported ion or a conformation change in the M5/6 region. Binding to Cys 813 would also be expected to inhibit enzyme function.
When the enzyme under acid transporting conditions is inhibited by radioactive omeprazole, it is still possible to label the M5/6 region after trypsin digestion with the fluorescent SH reagent, FMI. This indicates that only one of the two cysteines is labeled with omeprazole. In contrast another inhibitory compound, pantoprazole, is able to react with both cysteines (32).
Subsequent digestion of the fragment with thermolysin had indicated that Cys 813 is the primary site for omeprazole binding (31). This conclusion was based on the finding that ther-molysin digestion of the omeprazole-and FMI-labeled M5/6 fragment produced a smaller fragment labeled with omeprazole. This fragment had lost the FMI labeling and had NIPE as its N-terminal sequence. From the molecular mass and the absence of FMI labeling we had concluded that Cys 813 was the major site of reaction with omeprazole.
The present data show that Cys 822 is the functionally important residue for inhibition of the H,K-ATPase fusion protein by omeprazole. There are several possible explanations for the discrepancy of the labeling and mutation data. One rejects labeling on Cys 813 as being correct. Another suggests that binding in the region of Cys 822 is prevented by the Cys 822 3 Gly mutation and that noncovalent binding is a prerequisite for covalent inhibition of the enzyme by omeprazole at Cys 813 . The Cys 3 Gly mutation could distort this region, preventing reaction with the activated omeprazole. However, Cys 813 appears not to be essential for omeprazole inhibition in the fusion protein.
The unspecific protease thermolysin (semi-preferential for hydrophobic amino acids like leucine, isoleucine, and phenylalanine) may cut M5/6 into two equal sized fragments of about 20 amino acids, removing the sequence GCI or GCITI or GCITIL between these fragments. It may not cut the N-terminal region of the M5/6 peptide, leaving the NIPE sequence. This could lead to a loss of the fluorescent label on Cys 813 , although N-terminal sequencing of the entire band of the SDS gel could still give the sequence NIPE as the predominant peptide in the mixture. The second fragment would then contain the C-terminal part of the sequence but with variable N-terminal amino acids, which would not appear as unique sequences given the low yields of sequence. Variable N-terminal cutting sites at the beginning of the C-terminal fragment would obscure this second sequence. Another possibility is migration of omeprazole between Cys 822 and Cys 813 by disulfide exchange occurring after tryptic digestion of the enzyme with retention of the label at Cys 813 and loss of labeling at Cys 822 . Subsequent labeling of the fragment with FMI would then label Cys 822 , and radioactivity would be retained on Cys 813 with NIPE as the N-terminal sequence of the labeled thermolysin product.
If labeling is indeed on Cys 813 , then the mutational results show that the Cys 822 3 Gly mutation prevents effective interaction of the sulfenamide derivative of omeprazole or of protonated omeprazole with the enzyme and that this interaction is a necessary step preceding covalent inhibition of the enzyme by derivatization of Cys 813 . 3 K. Geering, personal communication. The present data suggest that Cys 822 is readily accessible to a bulky cation from the luminal side of the membrane because mutation of this residue abolishes inhibition of the enzyme by substituted pyridyl methylsulfinyl benzimidazoles. All of these bind to this region of the enzyme during inhibition (31). However, there is also some evidence that inhibition of the enzyme depends on conversion of the bound protonated form to the sulfenamide in situ rather than depending on reaction with bulk solution sulfenamide (31,54). For example, Cys 892 is in the large extracytoplasmic loop between TM7 and TM8 and is not labeled at all by pantoprazole (32), an unexpected finding if the free sulfenamide were the reactant. The sulfide of omeprazole is unreactive but still blocks omeprazole inhibition, suggesting that there is a reversible binding step before covalent inhibition (31). Addition of omeprazole to the H,K-ATPase fused to black lipid bilayers inhibits the proton current found in the absence of K ϩ (54), again as if inhibition occurs in situ on the enzyme surface.
The mutational data suggest that Cys 822 is accessible on the extracytoplasmic surface between M5 and M6 rather than within M6, and the labeling data suggest that Cys 813 is also accessible. Current models placing the carboxylic acids in M6 as being within the intramembranal region may have to be revised. These models are based on mutational results showing loss of ion binding. Translation (7), expression (52), and stability data (55) suggest that these fifth and sixth transmembrane regions interact more with the other transmembrane segments than with the hydrophobic lipid core of the membrane. The hydrophilic residues in this loop could still form ion-binding sites in association with other amino acids in other membrane segments. With this interpretation, prior mutational results giving, for example, loss of ion occlusion do not have to be interpreted as showing that these residues must reside within the hydrophobic core of the transmembrane domain of these proteins.
The data may be interpreted as showing that M5 and M6 do not form a transmembrane hairpin as illustrated in most models of these P-type ATPases. They are clearly membrane inserted, however. The model illustrated in Fig. 10 suggests that both Cys 813 and Cys 822 are accessible from the extracytoplasmic surface. In this model, the PLPL sequence found in both the Na,K and H,K-ATPases forms the beginning of the turn between M5 and M6. Mutation of these prolines in the Na,K and H,K-ATPases results in transmembrane insertion of M5 (56). 3 The two carboxylic acids at positions 820 and 824 are on the surface of M6, and Cys 813 is within a hydrophobic loop between M5 and M6.