Random Mutagenesis of the Sheep Na,K-ATPase (cid:97) 1 Subunit Generating the Ouabain-resistant Mutant L793P*

The polymerase chain reaction was used to randomly mutagenize a cDNA cassette encoding amino acids 691– 946 of the sheep Na,K-ATPase (cid:97) subunit. The mutagenized cassettes were used to replace the wild-type region in the full-length cDNA, and pools of mutants were transfected into HeLa cells. After the generation of resistant cells via selection in 0.5 (cid:109) M ouabain, polymerase chain reaction was used to amplify the mutagenized cassette from the genomic DNA of the stable transfec-tants. Sequence analysis of the polymerase chain reac- tion product revealed three amino acid substitutions: I729V, L793P, and K836R. Subsequent site-directed mu- tagenesis experiments showed that only L793P was important for resistance. To elucidate the role of L793 in ouabain inhibition, additional mutations at this position were prepared. L793A and L793I mutants were con- structed and expressed in HeLa cells. Only L793A sur-vived selection using ouabain, which suggested that re- sistance is not due to the specific substitution of leucine with proline. To explore the mechanism of resistance, apparent affinities of the L793P mutant for sodium and potassium were compared to the wild-type HeLa pump. Although the apparent affinities were comparable for sodium, the mutant had a 2-fold higher apparent affinity for potassium. This suggests that the mechanism of ouabain insensitivity of L793P is PCR a 1% agarose gel, excised and gel purified with the Qiaex gel extraction kit (QIAGEN). PCR (cid:59) digested Hin dIII and Bgl of wild-type digested with Hin dIII Bgl II.

Na,K-ATPase is located on the cell membrane of all eukaryotic cells and maintains the electrochemical gradient of the cell. It consists of a larger ␣ subunit and a smaller, glycosylated ␤ subunit. The former is the catalytic subunit; it pumps three Na ϩ ions out in exchange for two K ϩ ions pumped into the cell at the expense of ATP. It is also the site of interaction with cardiac glycosides, compounds that are commonly used to treat congestive heart failure. There is considerable interest in the characterization of the ouabain binding site of Na,K-ATPase to better understand the ouabain binding site with the aim of promoting rational drug design.
The site of cardiac glycoside binding is not fully understood, in part because the three-dimensional structure of the enzyme has not been elucidated. The amino-terminal half of the protein was found to be important after a chimera comprising the amino-terminal half of the ouabain-resistant rat ␣1 enzyme and the carboxyl-terminal half of the ouabain-sensitive sheep ␣1 enzyme proved to be ouabain resistant (1). Subsequently, several amino acids in the amino-terminal half of the protein were identified that affected inhibition of the enzyme by ouabain. Using the numbering of Shull et al. (2), these include the cysteine and tyrosine in the first transmembrane domain; Cys-104 (3,4) and Tyr-108 (4); Gln-111 (1, 2), Pro-118 (4), Asp-121 (5), and Asn-122 (1,6) in the first extracellular domain; and Tyr-308, located in the extracellular domain between the putative H3-H4 transmembrane domains (7).
The possibility that the carboxyl half of the protein may interact with ouabain was initially suggested by the finding that a monoclonal antibody localized to the H1-H2 extracellular domain increased ouabain binding (8). Furthermore, a chimera that was made using the amino-terminal half of the H,K-ATPase and the carboxyl-terminal half of the rat ␣1 Na,K-ATPase retained sensitivity to ouabain (9). Soon thereafter, Thr-797 (10 -12) and Arg-880 (4) were identified as being important in conferring sensitivity to ouabain.
Based on the hypothesis that the carboxyl-terminal half of this protein, like the amino-terminal half, potentially has several residues that affect the interaction with ouabain, we have reexamined a region of the sheep ␣1 subunit that includes amino acids 691-946 using PCR 1 -based random mutagenesis. This region encompasses a cytoplasmic region, the putative transmembrane domains H5 and H6, and two large predicted extracellular domains.

EXPERIMENTAL PROCEDURES
Materials-Molecular biology reagents were from Amersham Corp., Boehringer Mannheim, Pharmacia Biotech Inc., Promega, QIAGEN Inc., Stratagene, and U. S. Biochemical Corp. Reagents used in enzyme assays were obtained from Aldrich, and National Diagnostics (Atlanta, GA). Tris-ATP was from Sigma, ␥-[ 32 P]ATP from DuPont NEN, and ouabain from Boehringer Mannheim. Bio-Rad was the source for electroporation equipment. Tissue culture supplies were obtained from Life Technologies, Inc., and the Lineberger Cancer Center Tissue Culture Facility at the University of North Carolina. All other reagents were of the highest quality available. The eukaryotic expression vector constructs were described previously (1). The vector, pKC4, has the complete cDNA inserts of either sheep ␣1 (S␣1-pKC4) or rat ␣1 (rat ␣1-pKC4) Na,K-ATPase subunits. The sheep ␣1 cDNA is divided by endonuclease restriction sites into discrete cassettes that are compatible with the M13 multiple cloning site.
Random Mutagenesis-PCR-based random mutagenesis, based on a protocol by Leung et al. (13), is outlined in Fig. 1. The most salient features are that the reactions were conducting using a 1.0 mM concentration of each of the four deoxynucleotide triphosphates except for dATP, which was at a concentration of 0.2 mM. Furthermore, 0.5 mM MnCl 2 was added so that, with MgCl 2 , the total ion concentration was 6.0 mM. Under these conditions, the fidelity of the Taq DNA polymerase is compromised. It was estimated that the error rate increased from 0.4% under normal conditions to approximately 2% in the altered con-* This work was supported by National Institutes of Health Grant R01HL46469 (to E. M. P.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ditions. The reaction template was the eukaryotic expression vector pKC4 containing the sheep ␣1 cDNA. The vector was linearized using a unique Sca I site, which cuts outside of the sheep coding sequence. Two oligonucleotide primers were designed to amplify a region that includes the restriction enzyme sites HindIII and BglII, which define a region of the cDNA corresponding to amino acids 691-946 in the sheep ␣1 subunit. The forward primer was 5Ј-AAGTACCACACGGAGATTGT-GTTC-3Ј and the reverse primer was 5Ј-GAAAGCAGCAAGGGCT-GTCTCTTC-3Ј. The negative control was deionized distilled water. Both primers were used at a concentration of 20 pmol per 100 l of reaction volume. Amplification was performed on an Eri-Comp Twin-Bloc thermal cycler and consisted of an initial denaturation (94°C for 6 min) after which Taq polymerase was added, followed by 30 cycles of annealing (50°C for 1 min), elongation (70°C for 4 min), and denaturation (94°C for 1 min), followed by 2 min at 55°C, and finally, holding at 72°C for 10 min. The PCR product was size-fractionated by electrophoresis on a 1% agarose gel, then excised and gel purified with the Qiaex gel extraction kit (QIAGEN). The purified PCR fragment (ϳ780 base pairs) was digested with HindIII and BglII, and the pool of mutant cassettes was used to replace the wild-type cassette in the expression vector S␣1-pKC4 that had likewise been digested with HindIII and BglII.
Site-directed Mutagenesis-Mutagenesis of the HindIII/BglII cassette was performed as described previously (1) and according to the method of Kunkel (14). Several oligonucleotides were designed to reproduce the changes found in the original resistant mutant when annealed to single-stranded M13 containing the cassette. The base at 2464, according to the numbering of Shull et al. (2), was changed from A to G; likewise, base 2657 was changed from T to C and base 2786 from A to G. Several plaques from each mutagenesis experiment were picked and sequenced. Those cassettes with the mutation were sequenced in their entirety. Double-stranded (replicative form) form of the mutant cassette in M13 was double digested with HindIII and BglII and ligated back into the S␣1-pKC4 that had also been double digested. DH5␣FЈ cells were transformed and several colonies were screened using restriction endonuclease digests. Positive colonies had the 763-bp cassette sequenced in its entirety before the construct was purified on a cesium chloride gradient in preparation for electroporation into wild-type HeLa cells.
Cell Culture and Transfection-HeLa cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% bovine calf serum and antibiotics. Cells were prepared for transfection as described previously by Burns and Price (10). Each electroporation combined 20 g of linearized plasmid with 1 ϫ 10 6 HeLa cells. Electroporation of the randomly mutagenized plasmid used a total of 1 mg of pooled mutant plasmid that was transfected into a total of 5 ϫ 10 7 cells in 50 separate electroporations. Each electroporation was plated to a 100-mm tissue culture plate. Cells that were resistant to ouabain were isolated using a cloning cylinder. These cell lines were expanded in medium containing 0.5 M ouabain.
RNA Analysis-Analysis of all ouabain-resistant cell lines used the protocol described by Price and Lingrel (1).
Recovery of Mutagenized Cassettes from Stably Transfected Ouabainresistant Cell Lines and Sequencing-Genomic DNA was prepared for amplification as described previously (10) and was essentially as described by Higuchi (15). The two oligonucleotide primers were the same as those used to amplify the cassette for random mutagenesis. The PCR reactions were essentially the same as above, except that the dNTPs were used in equimolar concentrations and no MnCl 2 was added to the reaction. PCR products were gel purified with the Qiaex gel extraction kit (QIAGEN) and sequenced directly by adapting the methods of Kretz et al. (16) and Kusukawa et al. (17) to more readily accommodate the annealing reaction as outlined in the Sequenase 2.0 kit from U. S. Biochemical Corp. The primer:template ratio was set at approximately 100:1. After the addition of annealing buffer and deionized distilled water, the total volume was 10 l. This was incubated at 95°C for 2 min, then put on ice immediately. Sequencing followed, according to the instructions in the Sequenase 2.0 kit, which is based on the method of Sanger et al. (18). To control for the possibility of errors introduced by Taq DNA polymerase, PCR products from eight separate reactions based on the original resistant cell line were pooled. Sequence of sitedirected mutants was based on a pool of five separate PCR reactions.
Preparation of Crude Plasma Membranes from HeLa Cells-Membranes were prepared from wild-type HeLa cells and cells transfected with Lys-mutants or rat ␣1. This was done essentially as described by Price and Lingrel (1), except that cells from five confluent T150 flasks were harvested and cells were homogenized using 30 -40 strokes of a Dounce-type homogenizer, and the final preparation was suspended in  (19) using bovine serum albumin as the standard. Microsome preparations were snap frozen in liquid nitrogen, stored at Ϫ70°C, and used within 3 months.
Na,K-ATPase Activity Assay and Data Analysis-Enzyme activity was determined by the release of ␥-32 P from ␥-[ 32 P]ATP in an assay that was essentially as described by Jewell and Lingrel (20) in a method modified from Brown (21). In brief, assays were prepared in 1.7-ml tubes, split into triplicates, and incubated at 37°C for 1 h. Tubes were placed on ice, and 250 l of acid molybdate (5% ammonium molybdate in 2 N H 2 SO 4 containing 9.17% trichloroacetic acid and 2 mM phosphate that was diluted 2:3 with H 2 O) and 800 l of a 1:1 mix of isobutanol: ethyl acetate were added. Tubes were shaken for 15 s three times at intervals of 5 min and centrifuged briefly; then 500 l of organic phase was transferred to 5 ml of scintillation fluid. Counts were taken to give an error of Ϯ1%. Less than 20% of the ATP was hydrolyzed over the course of 1 h in any tube. Release of phosphate over the hour was linear in all assays.
Na,K-ATPase activity was determined against increasing concentrations of ouabain, NaCl, and KCl. In ouabain sensitivity experiments, specific enzyme activity was taken as the difference between the activity at 10 Ϫ2 M ouabain and the test concentrations of ouabain. Activity in ouabain-free medium was taken as 100% activity. The rate of phosphate release was linear at each concentration of ouabain. The specific conditions for the ouabain experiments were as detailed by Jewell and Lingrel (20). All experimental data were analyzed for goodness of fit to either a one-site or two-site Michaelis-Menten equation for microsome preparations expressing mutant enzyme, as well as those from the host cell line (22). Cell lines expressing mutants fit best to the two-site model, so the lower affinity site was taken as the IC 50 for the mutant enzyme. The higher affinity site was within the standard error for the IC 50 of wild-type HeLa cells using the one-site model.
In the ion dependence experiments, specific Na,K-ATPase activity was taken as the difference between activity in ion-free medium and the test concentrations of ion. ATPases that were non-Na,K-ATPases were not dependent on ion concentration. When preparations containing mutant enzyme were assayed, 1 M ouabain was in the assay medium to inhibit the native ouabain-sensitive pump of HeLa cells. The specific conditions for the ion dependence experiments were as detailed by Jewell and Lingrel (20). In these experiments, 100% activity was activity at V max according to the equation, where n ϭ 3 for Na ϩ and n ϭ 2 for K ϩ ; this equation assumes that each ion binding site is independent. Data were also fit to the equation, v ϭ V max ͓ion͔ n /͑EC 50 (ion) ϩ ͓ion͔ n ͒ with EC 50 ϭ EC 50 (ion) 1/n (Eq. 2) where n ϭ 3 for Na ϩ and n ϭ 2 for K ϩ ; this equation does not assume independence between ion binding sites. Both equations are from Garay and Garrahan (23). These models were fit using nonlinear regression with SigmaPlot (Jandel Scientific), which fits by minimizing the sum of squares of differences between data and model.

RESULTS
The pool of mutated cDNAs resulted in one resistant clone from a total of 50 tissue culture dishes when transfected cells were selected in 0.5 M ouabain. This clone, designated RM-PCR1 was expanded into a stable cell line and maintained in 0.5 M ouabain. Genomic DNA was isolated from these cells and PCR was used to amplify the integrated, randomly mutated cassette. The amplicon was of the expected size based on the cDNA (ϳ780 base pairs; data not shown). Product from endogenous Na,K-ATPase gene of the HeLa genome would be expected to be significantly larger because introns would be included in the PCR product. Northern analysis using a probe targeted to a region coding for the small t intron of the SV40 vector that is inserted between the cDNA and the polyadenylation sequence indicated that the original resistant clone expressed this vector-derived RNA (data not shown). The PCR product recovered from RMPCR1 genomic DNA is therefore probably the result of stable integration of the randomly mutagenized vector into the HeLa genome. The resistant phenotype is likely due to the presence and expression of the mutant sheep ␣1 cDNA.
Sequencing revealed that RMPCR1 contains a sheep-specific sequence with four base pair changes that code for three amino acid changes. Using the numbering of Shull et al. (2), the base at 2464 was changed from A to G, producing the amino acid change I729V; base 2541 was changed from A to G but did not change the code for glycine; base 2657 was changed from T to C, producing L793P; and base 2786 was changed from A to G, which produced K836R. According to a predicted two-dimensional structure based, again, on Shull et al. (2), I729V is predicted to be intracellular, L793P lies in the putative fifth transmembrane domain, and K836R is predicted to be extracellular.
After identification of the mutated base, site-directed mutagenesis was used to reproduce pairs of mutations in the wild-type sheep ␣1 cDNA (Table I). Because the three-dimensional structure of this enzyme is not known, it was possible that resistance to ouabain was due to any individual change or to the combined conformational effects of any two or all three amino acids. To discriminate the contribution of each mutation or combination to resistance, pairs of mutations were made and expressed in HeLa cells. Depending on what pair(s) survived, the individual or combination of important amino acid(s) could be deduced. The L793P mutant was constructed individually because it was the least conservative change of the three and therefore was thought the most likely candidate for causing resistance. Results from transfections (Table I) show that only when L793P was in the construct (i.e. either the double mutants I729V/L793P and L793P/K836R or the single mutant L793P) were the cells resistant to ouabain. These data indicate that the single mutant L793P was solely responsible for the ouabain-resistant character of the original resistant HeLa cells. Two clones from the transfection experiment with the individual mutant L793P were isolated and expanded into stable cell lines designated L793P1 and L793P2. Northern analysis on these cell lines was positive for the vector-derived se- None 12  14  16  ---L793P  31  16  27  ---L793I - a Construction of sheep ␣1 NaK-ATPase mutants is described under "Experimental Procedures." Each transfection used 20 g of linearized plasmid and 1 ϫ 10 6 HeLa cells. b -, not performed.
quence described above (data not shown). Furthermore, sequencing of PCR product derived from genomic DNA verified that the cDNA that was integrated into the genomic DNA of the host cell contained the site-directed L793P change.
Crude membranes from the two L793P cell lines, as well as from wild-type HeLa cells, and a stable cell line expressing rat ␣1 were prepared and assayed for the ability of ouabain to inhibit Na,K-ATPase activity (Fig. 2). The L793P1 and L793P2 cell lines are comparable in their resistance to ouabain, which approaches that of the cell line expressing the rat ␣1 subunit (Fig. 2, inset).
These mutant Na,K-ATPases were also characterized for their dependence on Na ϩ and K ϩ ions for activity (Table II) and were compared to wild-type HeLa Na,K-ATPase. These data show that little difference in the dependence of Na ϩ exists between the L793P mutants and the wild-type enzyme, whereas significant differences are seen in the dependence on K ϩ on activity. The measured dependence on K ϩ suggests that the mutant ouabain-resistant enzymes have a higher affinity for K ϩ than does the wild-type enzyme. Data were fit to both independent and cooperative models, but there was no difference in conclusions derived from these models.
To test the hypothesis that the mechanism of resistance introduced by this substitution is due to a localized change in conformation that may be specific to substitution with proline, two other amino acids with nonpolar side chains were substituted: isoleucine to maintain the size and polarity of leucine, and alanine, to vary the size. As with all site-directed mutants, the mutagenized cassette was sequenced in its entirety before being transfected into the host cells. Table I indicates that resistance of the Leu-793 substitution is not specific to proline substitution, because L793A generated ouabain-resistant colonies. The identities of the L793A clones were verified, as de-scribed previously, by collecting genomic DNA from three clonal cell lines, using PCR to amplify the pertinent cassette, and sequencing the PCR product. All three cell lines produced the correct size of amplicon, and sequence analysis revealed a sheep-specific sequence that contained the engineered base changes.

DISCUSSION
There were three amino acid changes produced using this PCR-based random mutagenesis method. Interestingly, each amino acid is conserved in all Na,K-ATPases reported to date (24). However, L793P accounted for most if not all of the resistance of the original triple mutant. This was supported by the fact that the IC 50 values for the L793P stable cell lines were comparable to the IC 50 for rat ␣1 construct, a sodium pump that is highly resistant to ouabain (20) (Fig. 2). Although neither the I729V nor the K836R mutations could confer resistance to ouabain, this does not exclude these residues from contributing to ouabain sensitivity, as might have been seen if less conservative substitutions had occurred at these sites.
One possible mechanism of resistance could have been an increase in affinity for both ions, resulting in a more efficient enzyme. Because enzyme activities in response to increasing concentrations of sodium were not significantly different between the mutant and the host cell, any increase in the efficiency of this enzyme with regard to Na ϩ is probably negligible. Alternatively, the conformational change that accompanies the catalytic cycle may occur more readily, even in the presence of ouabain. Activity in response to an increasing concentration of potassium showed that the mutant pump could achieve comparable rates of activity at half the concentration of potassium when compared to the host cell response, suggesting that the affinity for potassium is increased in the L793P mutant. Johnson et al. (25) have shown that an increase in K ϩ concentrations correlates with decreased ouabain binding in the sheep Na,K-ATPase. It is possible that increased affinity for K ϩ causes increased competition with ouabain at the ouabain binding site, and this may be the mechanism of resistance in the L793P mutant. Although K ϩ and ouabain binding sites are not thought to be identical (25), leucine 793 may be a residue that these two sites share.
The L793P inserts a proline between two other prolines in a region that is predicted to be a transmembrane domain (2). Because a polyproline sequence forms a left-handed helix with three prolines per turn, it is expected that the local conforma- Membranes were prepared from wild-type HeLa cells and clones were transfected with site-directed mutants, as described under "Experimental Procedures." All values were corrected for background activity measured in 10 Ϫ2 M ouabain. Percent activity was determined by setting the activity in the ouabain-free assay to 100%. G, wild type (wt) HeLa; f, rat ␣1; å, L793P-1; ç, L793P-2. The specific Na,K-ATPase activity (mol of ATP hydrolyzed per mg of protein/h) of each microsome preparation without ouabain was: wild type HeLa ϭ 33.9 ϩ 0.54; rat ␣1 ϭ 29.35 ϩ 0.73; L793P1 ϭ 11.63 ϩ 0.13; and L793P2 ϭ 27.55 ϩ 0.60. Plots for mutants are representative of three assays from at least two separate microsome preparations. Solid lines indicate data fit to the one-site Michaelis-Menten equation; dashed lines indicate data fit to the twosite equation. a Data fit to the noncooperative model: v ϭ V max /(1 ϩ EC 50 (ion)/[ion] n . n ϭ 3 for Na ϩ ; n ϭ 2 for K ϩ .
b Numbers in brackets represent the mean of the standard deviation corresponding to the fit of the data for each particular model. Generated using SigmaPlot by s ϭ (CV%/100)(K 0.5 ).
c Data fit to the cooperative model: v ϭ V max [ion] n /EC 50 (ion) ϩ [ion] n ) with EC 50 ϭ EC 50 (ion) 1/n , n ϭ 3 for Na ϩ ; nb ϭ 2 for K ϩ . tion of the protein will be disrupted regardless of the original conformation. However, because the host HeLa cells cannot survive in 0.5 M ouabain, the L793P mutant must be a functioning protein. To determine whether resistance was specific to the L793P substitution, two conservative substitutions, L793I and L793A, were constructed, expressed in HeLa cells, and selected for their ability to resist inhibition by ouabain. L793I is the most conservative change, so it was not expected that this construct would survive. Because L793A did survive (Table I), resistance is not isolated to any conformational changes inherent to substitution with proline alone. It is difficult to predict what the effects of the alanine substitution may have on this area. It is possible that resistance is due to conformational shifts in the protein that disrupt the relative position of at least part of the fifth transmembrane domain so that interactions between residues of juxtaposed transmembrane domains may redefine either or both of the ouabain and K ϩ binding site.
In summary, leucine 793 has been identified as an amino acid in the carboxyl-terminal half of the protein, which plays a significant role in the interaction between ouabain and the sodium pump. This residue resides in the putative fifth transmembrane domain and lies very close to threonine 797, an additional residue recently shown to be involved in ouabain inhibition (10 -12). Therefore, this region is clearly important for the interaction between ouabain and the Na,K-ATPase. These findings also support the utility of PCR-based random mutagenesis in the study of proteins, the tertiary structure of which is not available.