Thr-774 (Transmembrane Segment M5), Val-920 (M8), and Glu-954 (M9) Are Involved in Na+ Transport, and Gln-923 (M8) Is Essential for Na,K-ATPase Activity*

The highly conserved amino acids of rat Na,K-ATPase, Thr-774 in the transmembrane helices M5, Val-920 and Gln-923 in M8, and Glu-953 and Glu-954 in M9, the side chains of which appear to be in close proximity, were mutated, and the resulting proteins, T774A, E953A/K, and E954A/K, V920E and Q923N/E/D/L, were expressed in HeLa cells. Ouabain-resistant cell lines were obtained from T774A, V920E, E953A, and E954A, whereas Q923N/E/D/L, E953K, and E954K could only be transiently expressed as fusion proteins with an enhanced green fluorescent protein. The apparent K0.5 values for Na+, as estimated by the Na+-dependent phosphoenzyme formation (K0.5Na,EP) or Na,K-ATPase activity (K0.5Na,ATPase), were increased by around 2∼8-fold in the case of T774A, V920E, and E954A. The apparent K0.5 values for K+, as estimated by the Na,K-ATPase (K0.5K,ATPase) or p-nitrophenylphosphatase activity (K0.5K,pNPPase), were affected only slightly by the 3 mutations, except that V920E showed a 1.7-fold increase in the K0.5K,ATPase. The apparent K0.5 values for ATP (K0.5EP), as estimated by phosphorylation (a high affinity ATP effect), were increased by 1.6∼2.6-fold in the case of T774A, V920E, and E954A. Those estimated by Na,K-ATPase activity (K0.5ATPase) and ATP-induced inhibition (Ki0.5pNPPase) of K-pNPPase activity (low affinity ATP effects) were, respectively, increased by 1.8-fold and unchanged in the case of T774A but decreased by 2- and 4.8-fold in the case of V920E and were slightly changed and increased by 1.7-fold in the case of E954A. The E953A showed little significant change in the apparent affinities. These results suggest that Gln-923 in M8 is crucial for the active transport of Na+ and/or K+ across membranes and that the side chain oxygen atom of Thr-774 in M5, the methyl group(s) of Val-920 in M8, and the carboxyl oxygen(s) of Glu-954 in M9 mainly play some role in the transport of Na+ and also in the high and low affinity ATP effects rather than the transport of K+.

The similarity of M4, M5, and M6 among the three P-type ATPases is more than 50%, and the similarity between Na,K-ATPase and H,K-ATPase is in excess of 70% (Fig. 1). On the other hand, the similarity of M8 and M9 between Na,K-ATPase and Ca-ATPase is rather low, 38%, and the sequence 919 IV(V/ I)VQW 924 in M8 is highly conserved in mammalian Na,K-ATPase, indicating the possibility that these helices contribute to the third Na ϩ binding, which is only present in Na,K-ATPase. The specific requirement of a hydrophobic side chain in Val-920 of Na,K-ATPase is suggested because the corresponding residue in SERCA and H,K-ATPase is a negatively charged Glu. The Gln-923 in Na,K-ATPase is conserved in H,K-ATPase but not in SERCA (Asn-911). Q923A has been reported not to support cell growth (43). When the amino acid residue Asp-925 in M8 of the rat ␣2 isoform is mutated to Leu, a 2-fold decrease in the apparent affinity for both Na ϩ and K ϩ occurs, and for the case of mutation to Asn, a 4-fold decrease in Na ϩ affinity with a 2-fold increase in K ϩ affinity was reported (19). The presence of Na ϩ and K ϩ (14) has been reported to protect the chemical modification of Glu-953 in M9, but E953Q and E954Q showed no or only a slight effect on the apparent affinity for Na ϩ and K ϩ (17,18,27).
The purpose of this study was to obtain information concerning the third Na ϩ binding site, which may be present in M5, M8, and M9, by comparing the apparent K 0.5 values for Na ϩ with those for K ϩ and with those for ATP in mutated enzymes. The Val-920 and Gln-923 in M8 appears to face toward Glu-954 in M9 or Thr-774 in M5, respectively. Mutants carrying a single amino acid substitution in these residues and Glu-953 were constructed. The Glu-953 has been replaced with Phe in Artemia and Drosophila and has been assumed to be a Rb ϩprotectable dicylohexylcarbodiimide binding site (14). We were not able to obtain stable cell lines expressing the Gln-923 mutants. However, the amino acid substitution of Thr-774 to Ala, Val-920 to Glu (the same sequence for SERCA and H,K-ATPase at the corresponding position), Glu-953 to Ala, and Glu-954 to Ala were expressed in HeLa cells. The apparent K 0.5 for Na ϩ , K ϩ , and ATP of expressed enzymes were estimated by a Na ϩ -dependent phosphorylation reaction, the Na,K-ATPase activity, and K ϩ -dependent pNPPase activity of the enzymes. The data were subjected to a nonlinear least square regression using the Hill equation.
These results and others suggest that Gln-923 in M8 is crucial for the active transport of Na ϩ and/or K ϩ across membranes and that Thr-774 in M5, Val-920 in M8, and Glu-954 in M9 are related mainly to the active transport of Na ϩ , namely, the recognition and/or the occlusion-deocclusion of Na ϩ , and also to the high and low affinity ATP effects rather than to the transport process of K ϩ .

MATERIALS AND METHODS
Site-directed Mutagenesis-Plasmids pGEM-NaK and pCDL-NaK containing the entire coding region of rat Na,K-ATPase ␣1 cDNA have been described previously (50). Site-directed mutagenesis was performed using the megaprimer PCR method (51) using synthetic oligonucleotides. The amplified cDNA fragment, including the region encoding amino acid residues 674 -936 (between KpnI and EcoRI sites) for the Thr 774 , Val 920 , and Gln 923 mutants or 935-1016 (between EcoRI and SacI (vector) sites) for the Glu 953 and Glu 954 was subcloned back into the wild type rat Na,K-ATPase ␣1 subunit cDNA in the pGEM vector. The mutations were verified by nucleotide sequencing (52). The entire coding region of the mutated rat ␣1 cDNA was transferred to the expression vector pCDL as described previously (50) or the fusion protein expression vector pEGFP-C3 (Clontech. BD Biosciences).
Transfection of HeLa Cells Expressing the Rat ␤ Subunit with Rat Na,K-ATPase ␣1 Subunit-HeLa cells expressing the rat ␤ subunit (53) were cultured in a 35-mm dish in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. Cells were transfected with 2 g of plasmid DNA containing a wild type or mutant Na,K-ATPase ␣1 subunit cDNA with the GenePORTER transfection reagent (Gene Therapy Systems) and subjected to selection in 10 M ouabain. Ouabain-resistant cells were then expanded into stable cell lines.
Localization of EGFP Fusion Protein in HeLa Cells-The fusion protein expression vector pEGFP carrying the entire coding region of the wild type and mutated rat ␣1 cDNA was transfected to HeLa cells as described above. Within 24 -36 h after transfection, cells were washed twice with phosphate-buffered saline. Laser-scanning confocal microscopy was performed using a confocal microscope LSM510 (Zeiss). The EGFP fluorescent signals were visualized and recorded through an 40ϫ water immersion objective.
Preparation of SDS-purified Membrane Vesicles-Crude plasma membranes were prepared from HeLa cells (53). Crude plasma membranes at a concentration of 1 mg/ml were incubated with 0.15 mg/ml SDS in a buffer containing 0.25 M sucrose, 10 mM dithiothreitol, and 10 mM Tris-HCl (pH 7.4) for 5 min at room temperature. The sample was loaded onto stepwise sucrose density gradients consisting of 10 and 40% sucrose layers and centrifuged at 350,000 ϫ g in a Beckman TLA-100.3 rotor for 10 min at 4°C. The fractions at the 10%:40% interface were collected and pooled. The pooled sample was diluted with 3 volumes of the sucrose buffer and centrifuged at 350,000 ϫ g in a Beckman TLA-100.3 rotor for 10 min. The pellet was suspended in the sucrose buffer containing 10 mM dithiothreitol, and the protein concentration was measured using bovine serum albumin as a standard as described in (54).
Sodium-dependent Phosphoenzyme (EP) Formation from ATP-SDStreated membrane vesicles (4 g) were incubated at 0°C for 10 s in 40 l of 0-16 mM NaCl, 0.43 mM MgCl 2 , 25 mM imidazole/HCl (pH 7.2), and 2 M [␥-32 P]ATP. The reaction was terminated by adding 500 l of an ice-cold solution of 10% trichloroacetic acid solution containing 10 mM inorganic phosphate. The samples were centrifuged at 15,000 ϫ g for 10 min at 4°C. The precipitates were washed with an ice-cold 10% trichloroacetic acid solution by centrifugation. The resulting precipitates were washed with ice-cold water by centrifugation. The precipitates were dissolved in sample buffer containing a trace amount of a bromphenol blue, 10% glycerol, 1% SDS, 5% ␤-mercaptoethanol, and 10 mM sodium phosphate buffer (pH 6.0) and subjected to SDS-PAGE at pH 6.0 (55). After drying the gels, the amount of 32 P incorporated into the Na,K-ATPase ␣ subunit was quantitated by means of a BAS 2000 FIG. 1. Sequence homology among the P-type ATPase family in the transmembrane helices M4-M9. A, the sequences of the transmembrane helices M4, M5, M6, M8, and M9 in rabbit SERCA (47), rat Na,K-ATPase (48), and pig gastric H,K-ATPase (49) are shown. B, the sequence alignment of the M8 helices used in the Ogawa and Toyoshima homology modeling (46). The identical and similar amino acid residues between two types of ATPase are shown in white letters against a black background and in black letters against a gray background, respectively. The amino acid residues, in which a side chain or main chain oxygen atoms contribute to Ca 2ϩ binding in SERCA (9) are shown in black or white arrowheads, respectively. The sequence similarity is shown in parentheses. The position of the amino acid residues, Thr-774, Val-920, Gln-923, Glu-953, and Glu-954, are shown by arrows. system (Fuji), and the amount present relative to the wild type was determined.
Na,K-ATPase and Potassium-dependent pNPPase Activity-The ATPase activity of SDS-treated membrane vesicles from HeLa cells was measured in duplicate at 37°C for 30 min in a reaction medium containing 0.5ϳ2 g of enzyme protein, 0 -160 mM NaCl, 4 mM ATP, 16 mM KCl, 5 mM MgCl 2 , 125 mM sucrose, 0.5 mM EDTA, and 20 mM Tris-HCl (pH 7.4) in the presence of 5 M or 5 mM ouabain. The reactions were terminated by adding an equal volume of 12% SDS. Inorganic phosphate was determined colorimetrically by measurement of inorganic phosphate as ammonium molybdate complexes (56). Rat Na,K-ATPase activity was determined as the difference between the activities in the presence of 5 M and 5 mM ouabain. The relative turnover number was obtained by dividing rat Na,K-ATPase activity by the relative amount of EP max , estimated as described above. The value of the wild type was taken as 240 s Ϫ1 , which was estimated in a previous paper (57).
To investigate the effect of ATP and K ϩ on Na,K-ATPase activity, 4 mM ATP and 16 mM K ϩ were replaced with 0 -4 mM ATP and 0 -16 mM K ϩ , respectively. The effect of vanadate was measured after the addition of 0 -25 M sodium vanadate.
The pNPPase activity of the SDS-treated membrane vesicles was measured in the presence of 0 -16 mM KCl at 37°C for 30 min in a solution containing 1ϳ2 g of enzyme protein, 0.5-10 mM pNPP, 6 mM MgCl 2 , 125 mM sucrose, 0.5 mM EDTA, and 20 mM imidazole/HCl (pH 7.2). The K-pNPPase activity was calculated as the difference between the activities with and without KCl. The K-pNPPase activity was also measured in the presence of 5 mM ouabain.
Curve Fitting of the Data-The data were subjected to curve fitting to estimate each apparent K 0.5 and n for a ligand (L) by a nonlinear least square regression using the Hill equation (GraphPad Prism; GraphPad Ln L ϩ [L] n L ), in which V max , K 0.5 L , and n L are the maximum amounts of EP formed or the maximum Na/K-or K-pNPPase activity, the concentration of ligand required for the halfmaximum formation of EP or the half-maximum velocity, and the Hill coefficient, respectively. The value shown in the figures represent each K 0.5 and n estimated from three independent duplicated experiments (the means Ϯ S.D., n ϭ 6) with the correlation coefficient, r.

Construction of the Mutants of Na,K-ATPase Carrying a Single Amino Acid Substitution Near the Cation Binding Site-
Homology modeling of Na,K-ATPase based on the three-dimensional structure of Ca-ATPase (9) was carried out with the SWISS-MODEL software program (www.expasy.ch/spabv) (Fig. 2, A and B) using the alignment shown in Fig. 1A. The amino acid residues Thr-774 (M5), Val-920 and Gln-923 (M8), and Glu-953 and Glu-954 (M9) appear to be close together and also in close proximity to the two Na ϩ binding sites estimated from homology modeling. Na,K-ATPase transports three Na ϩ ions from the cytosol to the extracellular space, whereas the Ca-ATPase and H,K-ATPase transports two Ca 2ϩ ions and H ϩ ions, respectively. Thus, some amino acid residues that are not conserved between Na,K-ATPase and Ca-ATPase or H,K-ATPase might be related to the third Na ϩ ion binding conformation. Val-920 is conserved in all Na,K-ATPase enzymes, except for the fungal Blastocladiella emersonii, but the corresponding amino acid in Ca-ATPase and H,K-ATPase is Glu. . The figures were prepared using RasMol. The homologues amino acid residues, of which the side-chain oxygen atom contributes to Ca 2ϩ binding in SERCA, and several amino acid residues, which had been reported to contribute to binding the cation in Na,K-ATPase, are shown in a stick-and-ball model. The side chains of the amino acid residues mutated in this paper are also shown in a stick-and-ball model in yellow. The oxygen and nitrogen atoms are colored red and blue, respectively. Spheres (cyan) represent the position of two Ca 2ϩ (Na ϩ ) ions.
The Gln-923 is highly conserved in Na,K-ATPase and H,K-ATPase, so this amino acid may be related in some manner to the transport of monovalent cations. To investigate this point, mutant proteins carrying a single amino acid substitution of these residues (T774A, V920E, Q923N/D/E/L, E953A/K, and E954A/K) were constructed.
Each of the mutant Na,K-ATPase molecules was expressed in HeLa cells expressing a rat Na,K-ATPase ␤ subunit as described in a previous paper (53). In the case of Q923N/D/E/L, E953K, and E954K, stable cell lines surviving in the presence of 10 M ouabain could not be obtained, but some cells transfected with E953K survived for 24 -36 h after transfection in the presence of 10 M ouabain.
We constructed the expression vector encoding the EGFP fusion protein with the wild type enzyme and transfected it in HeLa cells. The resulting EGFP wild type fusion Na,K-ATPase had an ATPase activity similar to the wild type enzyme, suggesting that the fusion with EGFP had a negligible effect on ATPase activity (data not shown). The EGFP fusion protein with the Q923N/D/E/L, E953K, and E954K mutants were transiently expressed in HeLa cells. Green fluorescent signals were observed on the plasma membrane in HeLa cells expressing the wild type and Q923N/D/E/L, E953K, and E954K mutants. When the HeLa cells were transfected with the expression vector without Na,K-ATPase cDNA, the green fluorescence was uniformly observed in the cytosol. The fusion protein expression vector contains the neomycin-resistant gene, so HeLa cells transfected with the EGFP mutant fusion constructs were selected in the presence of G418. We obtained G418-resistant stable cell lines, but these cells did not express the green fluorescent fusion proteins. These results suggest that, although these mutant enzymes were normally expressed on the plasma membrane, the Na ϩ /K ϩ -ATPase activity and/or the transport activity of Na ϩ /K ϩ across the membranes was too low to permit stable cell lines to survive in the presence of 10 M ouabain.
Effect of Each Mutation on the Apparent Affinity for Na ϩ Registered by ATP-dependent EP Formation and by Na,K-ATPase Activity-To estimate the apparent K 0.5 for Na ϩ in the absence of K ϩ , the Na ϩ -dependent EP (E2P) formation of a SDS-treated enzyme preparation under steady state conditions was measured in the presence of 0 -16 mM NaCl, 0.43 mM Mg 2ϩ , and 2 M [␥ 32 -P]ATP (Fig. 3). The data were fitted to the Hill equation with the correlation coefficient 0.956 -0.986, which gave values for the maximum amount of phosphoenzyme (EP max ), the Hill coefficient n Na,EP , the half-maximum concentration of Na ϩ (K 0.5 Na,EP ), and the correlation coefficient (r). The K 0. 5 Na,EP of the wild type was 0.59 Ϯ 0.03 mM, with an n value of 1.8 Ϯ 0.2. Mutation increased the values of K 0.5 Na,EP in T774A, V920E, and E954A to 2.4-, 3.4-, and 1.8-fold, respectively (Table I, a) without significant change in the K 0.5 Na,EP in E953A. These mutant enzymes showed about a 20ϳ30% reduction in n value, a slight reduction of positive cooperativity for Na ϩ . These data indicate that the T774A, V920E, and E954A mainly increased the apparent K 0.5 for Na ϩ that accompanies the phosphorylation reaction (Table I, a).
To estimate the apparent K 0.5 and Hill coefficient (n) for Na ϩ in the presence of K ϩ , the Na,K-ATPase activities of the rat wild type and mutants were measured in the presence of 0 -160 mM NaCl, 4 mM ATP, 16 mM KCl, and 5 mM MgCl 2 and either 5 M or 5 mM ouabain. The difference between these activities was assumed to be the Na,K-ATPase activity of the expressed enzyme. The values for the maximum velocity (V max ), Hill coefficient (n Na,ATPase ), and the half-maximum concentration of Na ϩ (K 0.5 Na,ATPase ) were estimated by fitting the data to the Hill equation with a correlation coefficient 0.992-0.994 (Fig. 4). The K 0.5 Na,ATPase for the wild type was 5.7 Ϯ 0.2 mM with an n value of 2.0 Ϯ 0.1. The value of K 0.5 Na, ATPase for T774A, V920E, E953A, and E954A increased to 7.5-, 1.8-, 1.2-, and 3.1-fold, respectively, without a significant change in positive cooperativity for Na ϩ . The turnover number, as estimated from V max and EP max , for all mutants was rather close to those of the wild type (Table I, h). These data show that the main effect of each of the three mutations (T774A, V920E, and E954A) was a clear increase in the apparent K 0.5 values for Na ϩ , as estimated by both phosphorylation and Na,K-ATPase activity (Table I,

a and b).
Effect of Substitution on the Apparent Affinity for K ϩ , as Evidenced by Na,K-ATPase Activity and K-pNPPase Activity-To estimate the apparent K 0.5 for K ϩ , Na,K-ATPase activities of the wild type and mutants were measured with increasing KCl concentrations up to 16 mM KCl in the presence of 4 mM ATP, 100 mM NaCl, and 5 mM MgCl 2 and either 5 M or 5 mM ouabain (Fig.  5). The data were fitted as described above, and the values for the Hill coefficient (n K,ATPase ) and the half-maximum concentration of K ϩ (K 0.5 K,ATPase ) were estimated with the correlation coefficient 0.990 -0.994 (Fig. 5). The wild type enzyme showed a value of 1.24 Ϯ 0.05 mM for K 0.5 K,ATPase and 1.5 Ϯ 0.1 for n K,ATPase . The T774A and E954A showed values similar to those of the wild type enzyme, which were slightly lower than that of E953A (Table I, c). However the K 0.5 K,ATPase of V920E showed a 1.7-fold increase in K 0.5 K,ATPase without significant change in the n K,ATPase . The K ϩ -dependent hydrolysis of pNPP (K-pNPPase) by Na,K-ATPase is generally thought to be catalyzed by the K ϩoccluded or -bound enzyme form (58). To estimate the apparent K 0.5 for K ϩ in the absence of Na ϩ , K ϩ -dependent pNPPase activity was measured in the presence of 0 -16 mM KCl, 5 mM pNPP, 6 mM MgCl 2 , 125 mM sucrose, 0.5 mM EDTA, and 20 mM imidazole/HCl (pH 7.2) (Fig. 6). The data obtained from the difference in values in the absence and presence of K ϩ were fitted as described above with the correlation coefficient 0.982-0.990 except in the case of V920E, in which the value was 0.881. The half-maximum concentration of K ϩ (K 0.5 K,pNPPase ) and Hill coefficient (n K,pNPPase ) were estimated to be 1.24 Ϯ 0.05 mM and 2.2 Ϯ 0.2, respectively, for the wild type. Each mutation appeared to induce little significant effect on the apparent K 0.5 K,pNPPase (Table I,  The incorporated 32 P in the Na,K-ATPase ␣ subunit was detected and quantitated, and the data were subjected to curve-fitting to estimate K 0.5 Na,EP (mM) and n Na,EP using the Hill equation (dotted line, wild; broken line, V920E; solid lines, T774A, E963A, and E954A) as described above, where 100% of EP was the value of EP max was estimated from the fitting. Each value of K 0.5 and n shown in this paper represent the means Ϯ S.D. obtained from three independent duplicated experiments, and r is the correlation coefficient.
n value in the V920E mutant, 3.0 Ϯ 0.7, could be due to the lower r value, 0.881 (Fig. 6).
Effect of the Substitution on the Apparent High Affinity ATP Effects, as Evidenced by Na ϩ -dependent EP Formation-To estimate the apparent K 0.5 for the high affinity ATP effect, Na ϩ -dependent EP formation was measured in the presence of 0.001-10 M [␥ 32 -P]ATP, 0.43 mM Mg 2ϩ , and 16 mM NaCl (Fig.   7). The data were fitted to the Hill equation with a correlation coefficient 0.984 -0.991. The wild type enzyme showed 21 Ϯ 1 nM for the K 0.5 EP with little cooperativity, n ϭ 1.1 Ϯ 0.1. Each mutation induced a 1.3-2.6-fold increase in the K 0.5 EP without any significant change in the Hill coefficient (n) except that the V920E showed a negative cooperativity for ATP, 0.6.
Effect of the Substitution on the Apparent Low Affinity ATP Effects, as Evidenced by Na,K-ATPase Activity and by ATPinduced Inhibition of K-pNPPase Activity-To estimate the apparent K 0.5 for the low affinity ATP effect, Na,K-ATPase activities were measured in the presence of increasing ATP concentrations up to 5 mM, 100 mM NaCl, 16 mM KCl, and 5 mM MgCl 2 and either 5 M or 5 mM ouabain. The data were fitted to the Hill equation with a correlation coefficient 0.970 -0.997. The wild type enzyme showed 0.50 Ϯ 0.03 mM for K 0.5 ATPase and slight positive cooperativity for ATP (n ϭ 1.3 Ϯ 0.1), which may be related with a high and a low affinity ATP effects as widely accepted. The T774A and V920E, respectively, showed a 1.8-fold increase and 2-fold decrease in the apparent K 0.5 ATPase with a similar n value, 0.9 ( Fig. 8A and Table I, m and n). The E953A and E954A showed little change in the apparent K 0.5 ATPase , with an n value of around 1. Submillimolar ATP is known to inhibit K-pNPPase activity by shifting the equilibrium from KE2 to E1ATP. Fig. 8B shows the ATP-dependent inhibition of K-pNPPase, which was measured in the presence of 0 -4 mM ATP, 16 mM KCl, 5 mM pNPP, 6 mM MgCl 2 , 125 mM sucrose, 0.5 mM EDTA, 20 mM imidazole/ HCl (pH 7.2). The data were fitted as describe above with the correlation coefficient 0.957-0.982. The significant changes ob-TABLE I Comparison of the relative change in the apparent K 0. 5 The relative values for the change induced by each mutation were calculated where each relative value was the ratio (the apparent K 0.5 of the mutant)/(that of the wild type) taken from Figs. 3-8. The relative mean value of each mutant is shown, and the minimum ((mean Ϫ S.D.) mutant / (mean ϩ S.D.) wild ) and maximum ((mean ϩ S.D.) mutant /(mean Ϫ S.D.) wild ) values are calculated and shown in parentheses; thus, the relative value of the wild type is 1 Ϯ S.D. The relative values of V max /EP max are also shown, where the value of the wild type enzyme was 240 Ϯ 14 s Ϫ1 . Bold type indicates parameters significantly affected by the mutation.  was measured, and the data were subjected to curve fitting to estimate K 0.5 Na,ATPase (mM) and n Na,ATPase using the Hill equation as described under "Materials and Methods," where 100% of the Na,K-ATPase activity was the value of V max estimated from the fitting. served were that the V920E showed a 4.8-fold increase in the apparent affinity for ATP (K i,0.5 pNPPase ), and the E954A showed a 1.7-fold decrease. DISCUSSION Present data were fitted to the Hill equation, and the correlation coefficients (r) in each fitting were 0.95-0.99, except for one case, 0.88 of the K 0.5 K,pNPPase for V920E, showing that the data fitted reasonably well. Each Hill equation gave a better correlation coefficient than the coefficient from the Michaelis-Menten equation for estimating the high and low affinity ATP effects (not shown), which would be reasonable, because of the oligomeric properties of Na,K-ATPase (59).
To study the significance of the apparent change in the K 0.5 value, the relative values for the change induced by each mutation were calculated (Table I). These values permitted us to assume that up to a 1.6-fold change (bold character) in the apparent K 0.5 would be significantly different from that of the wild type.
The data suggest that T774A, V920E, and E954A showed a significant increase in K 0. 5 Na,EP and the K 0.5 Na,ATPase values for Na ϩ and the K 0.5 EP for the high affinity ATP effect, compared with the wild type enzyme. The T774A and V920E showed, respectively, an increase and a decrease in the K 0.5 ATPase for the low affinity ATP effect. The V920E and E954A showed, respectively, a decrease and an increase in K i,0.5 pNPPase for another low affinity ATP effect. No significant change was detected in either K 0.5 K,ATPase and K 0.5 K,pNPPase for the apparent affinity for K ϩ , except for the K 0.5 K,ATPase in V920E. The E953A showed little significant change in all K 0.5 values examined. The turnover number of these mutants, V max /EP max , was nearly the same as the wild type enzyme (Table I, h), which indicates the present mutations mainly affected the apparent K 0.5 values.
To compare the change in apparent K 0.5 for Na ϩ with that for K ϩ and that for ATP, each ratio was calculated (Table II). The calculated data containing the minimum and the maximum values suggest that the T774A, V920E, and E954A showed a significant (bold character) decrease in the apparent affinities for Na ϩ relative to that for K ϩ (Table II, a/c, a/d, b/c, and b/d), except that the value K 0.5 Na,ATPase /K 0.5 K,ATPase of V920E was 1, within experimental error. These three mutations also showed a significant decrease in the apparent affinities for Na ϩ relative to those for the high and low affinity ATP effects (Table II, b/e, a/f, b/f, a/g, and b/g), except that the value of K 0.5 Na,EP / K 0.5 ATPase of the T774A and the K 0.5 Na,EP /K i,0.5 pNPPase of E954A were rather close to 1. The E953A showed little change in the relative ratios except a 1.6-fold increase in the value of K 0.5 Na,ATPase /K i,0.5 pNPPase . These results indicate that three residues, Thr-774, Val-920, and Glu-954, are related to the apparent affinities for mainly Na ϩ and possibly for ATP rather than those for K ϩ . Some decreases in the n values for the high and low affinity ATP effects (Figs. 7 and 8, A and B), such as the case of V920E, may indicate that the mutation induced the negative cooperativity for ATP, but more precise experiments will be required to clarify this point.
It has been reported that the Q923A-mutated enzyme could not support the cell growth possibly because of an impairment in the maturation process or targeting to the plasma membrane (35). The present study showed that Q923N/D/E/L could be transiently expressed as a fusion protein with EGFP on the plasma membrane as seen in the case of the EGFP wild type fusion protein at 24 -36 h after transfection, although they did not support sustained cell growth, as has been reported previously (35). The selection in the presence of G418 for 2-3 weeks led to a negligible green fluorescent signal in the G418-resistant HeLa cells. These data indicate that the EGFP mutant fusion protein was able to normally form an ␣␤ complex and depress endogenous Na,K-ATPase activity to deplete the ␤ subunit. Therefore, the data suggest that the Gln-923 mutation had little effect on either the maturation process or the targeting of the enzyme to the plasma membrane. These considerations support a scenario in which Gln-923 is prerequisite for 2), and 16 mM of NaCl. After SDS-PAGE, the incorporated 32 P in the Na,K-ATPase ␣ subunit was detected and quantitated. The data were subjected to curve fitting to estimate K 0.5 EP (nM) and n EP using the Hill equation (dotted line, wild; broken line, V920E; solid lines, T774A, E953A, and E954A) as described above, where 100% of EP was the value of EP max estimated from the fitting.
FIG. 8. ATP concentration dependence of Na,K-ATPase activity. A, the ATPase activity of SDS-treated membrane vesicles of the rat wild type (E), T774A (F), V920E (OE), E953A (■), and E954A (‚) was measured at 37°C for 30 min in a reaction medium containing 0.5ϳ2 g of enzyme protein, 0 -5 mM ATP, 100 mM NaCl, 16 mM KCl, 5 mM MgCl 2 , 125 mM sucrose, 0.5 mM EDTA and 20 mM Tris-HCl (pH 7.4) in the presence of 5 M or 5 mM ouabain. The Na,K-ATPase activity was measured, and the data were subjected to curve fitting to estimate K 0.5 ATPase (mM) and n ATPase using the Hill equation (dotted line, wild; broken line, V920E; solid lines, T774A, E953A, and E954A), where 100% of Na,K-ATPase activity was the value of V max estimated from the fitting. B, the potassium-dependent pNPPase activity was measured as described in Fig. 6 except that 16 mM KCl was present with 0 -4 mM ATP. The data were subjected to curve fitting to estimate K i,0.5 pNPPase (mM) and n pNPPase using the Hill equation as described above, where 100% of the activity represents the value without ATP. some step required for Na ϩ /K ϩ -dependent ATP hydrolysis and/or Na ϩ /K ϩ transport across the membranes. Thus, the transport activity of Na ϩ and/or K ϩ in Q923N/D/E/L might be too low to permit sustained cell growth.
The Gln-923 in M8 appeared to face toward the Thr-774 in M5 in our model (Fig. 2, A and B). The distance between the side-chain oxygen atom of Thr-774, Ser-775, and Gln-923 and the carbonyl oxygen atom of Tyr-771 is 3.3-4.1 Å, so Na ϩ could fit into this space (shown as a black circle in Fig. 2, A and B).
The data shown in Table I (compare both a and b with c and d of T774A) are consistent with the report that T774A reduced the apparent affinity for Na ϩ , as estimated by Na ϩ -dependent phosphorylation without affecting the affinity of Tl ϩ (K ϩ ) binding in a yeast expression system (37). The hydroxyl group of Tyr-771 has also been reported to be important for binding Na ϩ without participating in K ϩ -enzyme interaction (42). The direct coordination of Gln-923 to the first Na ϩ and the first K ϩ without coordination from Thr-774 to any Na ϩ and K ϩ has been proposed based on homology modeling (46). Present data are consistent with a scenario in which the oxygen atom in the Thr-774 side chain (CH 3 CH(OH)) directly coordinates to the third Na ϩ (Fig. 2, A and B).
The importance of the hydroxyl group of Ser-775 in the binding of both cations has also been reported (37). Homology modeling (46) suggests that Tyr-771 and Ser-775 coordinates, respectively, to the third Na ϩ and the first K ϩ . The Ser-775 might be important for the coordination to the first and third Na ϩ as shown in our model (shown as a cyan sphere and a black circle in Fig. 2, A and B). Tyr-771, Thr-774, and Gln-923 are present in the corresponding sequence in H,K-ATPase. However the Ser-775 is replaced by Lys-792 in H,K-ATPase, which may detract from the above possibility, because gastric H,K-ATPase is believed to bind two H ϩ ions and two K ϩ ions.
Val-920 is a highly conserved amino acid residue in Na,K-ATPase, and the corresponding amino acid residue in SERCA and H,K-ATPase is Glu, indicating the possibility of some unique role for Val-920 in Na,K-ATPase, which appears not to have been studied. The direct coordination of the Val side chain to Na ϩ is unlikely. The introduction of a negative charge at the Val-920, which is situated near Glu-954 (Fig. 2, A and B), may lead to repulsion between these residues and bent the M8 and M9 helices. This distortion in the V920E in the transmembrane segment might be related to the decrease in the affinity for Na ϩ and influence the conformational state of the N-domain such as to decrease the affinity for high affinity ATP effect and increase the affinity for low affinity ATP effects differently (Table I, e, f, and g). If these two apparent low affinity ATP effects structurally reflect the same single low affinity ATP binding conformation, the ratio K i,0.5 pNPPase /K 0.5 ATPase would be the same as the ratios in the data obtained from the mutagenesis of the ATP binding pocket as has already been reported (53,57). The ratio for the wild type and V920E was, respectively, 0.86 (0.43/0.50) and 0.42 (0.21/0.5). The relative ratio, taken as 1 in the case of the wild type enzyme, increased to 2 and 4, respectively, in R544A and K501E, whereas it was 0.7ϳ1.4 in K480A, K501A, G442A, G442P, D443A, and S445A. These data suggest the presence of two different low affinity ATP binding conformations, distinguishable based on affinities for the ATP-dependent activation of Na,K-ATPase activity and the ATP-dependent inhibition of K-pNPPase activity.
A complete understanding of the role of Val-920 will require additional experiments, but these results suggest that the role of the Val-920 ((CH 3 ) 2 CH) side chain in M8 may be not only to stabilize the third Na ϩ binding indirectly but also to influence the conformational state of the ATP binding states (59) in the N-domain (9), separated by ϳ50 Å. The high resolution NMR structure of the 213-residue nucleotide binding domain of rat ␣1 Na,K-ATPase showed only a very low ATP binding affinity, K d ϭ 5 mM (60), as in the case of the low affinity TNP-ATP binding to the expressed M4-M5 loop of Na,K-ATPase (61).
The question arises as to whether a change in the E1-E2 equilibrium state (62,63) exists for the V920E mutant. Vanadate is an analogue of P i that binds to the E2 form of Na,K-ATPase, and the change in the sensitivity of the Na,K-ATPase and K-pNPPase activity to vanadate is used to a type of indicator of the equilibrium state (62,63). However, the sensitivity of the Na,K-ATPase and pNPPase activity to vanadate did not change in the V920E mutant (data not shown).
The replacement of Glu-954 or Glu-953 with Lys had little influence on the localization of the protein in the cell, when the mutant proteins were expressed as the EGFP fusion proteins. However, stable cell lines expressing these mutant proteins with and without EGFP could not be obtained, although cells expressing E953K survived in the presence of 10 M ouabain for several days after transfection, but this was not the case for E954K. In our homology modeling the Glu-954 faces M6, and Glu-953 may face another transmembrane segment of an adjacent protomer or a lipid phase. The chemical modification of Glu-953 by carbodiimides has been reported to cause inactivation of Na,K-ATPase (14), although Glu-953 is less conserved in Na,K-ATPase, and the E953A showed little significant effect on the kinetic parameters examined (Ref. 17 and Table I). The ). This table shows the relative ratios ((the ratio from the mean value of the mutant enzymes)/(the ratio from the mean value of the control enzyme)) with each minimum ((mean Ϫ S.D.)/(mean ϩ S.D.)) and maximum ((mean ϩ S.D.)/(mean Ϫ S.D.)) relative ratios in parentheses; thus, each ratio of the mean value of the wild type enzyme is 1. Bold type indicates parameters significantly affected by the mutation. homology modeling (46) suggests that the Glu-954 coordinates to the third Na ϩ . These considerations suggest that the Glu-954 contributes to the formation of the Na ϩ binding conformation but appears not to be essential amino acid even though it is highly conserved. The third Na ϩ binding site has been reported to be surrounded by M5, M6, and M9 from the valence map calculated from the constructed model (46), and the proposed third Na ϩ coordinates to the side chain of Tyr-771 (M5), the carbonyls of Gly-806 and Thr-807 (M6), and the carboxyl of Glu-954 (M9) (shown as a circle with a black broken line in Fig. 2, C and D).
To determine the sodium binding sites, Ogawa and Toyoshima (46) aligned the Glu-908 and Asn-911 in SERCA with Gln-923 and Asp-926 in Na,K-ATPase or Gln-943 and Asp-946 in H,K-ATPase, respectively, which resulted in a 29% (the 7 amino acids are similar to each other in the 24 amino acids) or 25% (6 in 24) similarity in M8 between SERCA and Na,K-ATPase or H,K-ATPase, respectively. When we aligned the Glu-908 and Asn-911 in SERCA with Val-920 and Gln-923 in Na,K-ATPase (Fig. 1B), the similarity increased to 38% (9 in 24). In our alignment (Fig. 1A) the stretch was composed of seven amino acids, VTIEMCN, in SERCA was aligned with the sequence ISIEMCQ in H,K-ATPase, in which the four amino acids in the central region (underlined letters) are identical and the three amino acids at both ends are conserved. In their alignment (Fig. 1B), however, three amino acids are shifted to the left in H,K-ATPase and Na,K-ATPase by about one turn of the ␣ helix to the extracellular side, and the position of Gln-923 in their model corresponds to the position of Val-920 in our model. We carried out the same calculation as described in Ogawa and Toyoshima (46) using the structures based on both our and their alignments, and the first and second Na ϩ binding sites could be identified, but the third could not in either of the structures. Based on their model (46), the side-chain oxygen atom of Gln-923 (M8) coordinates to both the first Na ϩ ion and the first K ϩ ion. Active Na,K-ATPase appeared not to be expressed in the Gln-923/Asn/Asp/Glu/Leu mutants as described above, as in the case of substitution by Ala (35). In SERCA the amino acid substitution Glu-908 to Ala, Met, and Asp abolished Ca 2ϩ transport in the sarcoplasmic reticulum, but substitution to Gln exhibited a slight increase in the affinity for Ca 2ϩ without any effect on the rate of Ca 2ϩ transport (64,65). The Thr-774 doesn't coordinate to any Na ϩ ions in their model (46), but T774A showed significant reduction in the apparent affinity for Na ϩ (Table I) (37). The Val-920 is situated to the extracellular side of the Glu-954 in their model (Fig. 2, C and D), but V920E reduced the apparent affinity for Na ϩ , K ϩ , and high affinity ATP effect and increased the apparent affinity for low affinity ATP effects (Table I). These considerations may support the model based on our alignment.
The first and second Na ϩ binding sites are situated in about one-third of the membrane bilayer from cytoplasmic face (9,13). Domaszewicz and Apell (31) reported that a dielectric coefficient for the electrogenic binding of the third Na ϩ is about 0.25 and that the dielectric depth of the access channel is about 25% within the membrane/protein dielectric. Although this finding does not necessarily mean that it is spatially 25% inside the bilayer from the cytosol, the third Na ϩ binding site might be located in the more cytosolic side in the membrane compared with the other Na ϩ binding sites. In our homology modeling (Fig. 2, A and B), the carbonyl of Thr-771 and the side chain oxygen atoms of Thr-774 and Gln-923 are located about a quarter of the membrane bilayer from the cytosolic side.
A number of mutational and biochemical studies have been carried out in attempts to elucidate the location of the cation binding sites in Na,K-ATPase (10 -45). The third Na ϩ binding site, which is only present in Na,K-ATPase, has not been clearly defined. The results presented in this study indicate that the third Na ϩ binding site may be located between M5 and M8 and in close proximity to Thr-774 and Gln-923, although further experiments will be required to clarify the amino acid residues that contribute to the third Na ϩ binding site.