The M4M5 Cytoplasmic Loop of the Na,K-ATPase, Overexpressed in Escherichia coli, Binds Nucleoside Triphosphates with the Same Selectivity as the Intact Native Protein*

Escherichia coli was used to overexpress the large cytoplasmic loop of the rat Na,K-ATPase. A 1260-base DNA segment encoding Lys354–Lys774of the rat α1-subunit was constructed via polymerase chain reaction. The polymerase chain reaction product was successfully subcloned into the expression vector pET-28 (Novagen), which produces an N-terminal 6-histidine-tagged fusion protein. The pET-28 vector containing rat α-loop, i.e. pAN, was used to transform calcium-competentE. coli BL21(DE3) cells, and positive clones were selected by kanamycin resistance. Bacterial cultures were grown, and protein synthesis was induced with isopropyl β-d-thiogalactoside. Cells were harvested and lysed, revealing production of the His-tagged fusion protein (∼46 kDa). The fusion protein was affinity-purified from other soluble cellular proteins via a Ni-NTA column, which routinely yielded ∼20 mg of soluble His6-α-loop/L cell culture. The His6-α-loop retained significant native structure, as evidenced by the ability of ATP and ADP (but not AMP, CTP, GTP, or UTP) to protect against chemical modification by either fluorescein isothiocyanate or maleimidylanilinonapthalene sulfonic acid. More specifically, circular dichroism spectroscopy was used to estimate the secondary structure of the His6 loop, revealing an ordered folding composed of 23% α-helix, 23% antiparallel β-sheet, 4% parallel β-sheet, 19% β-turn, and 32% random coil. The 6-histidine loop bound the fluorescent ATP analog trinitrophenyl-ATP with high affinity, as determined by measuring the fluorescence changes associated with binding. Affinities for ATP (∼350 μm) and ADP (∼550 μm) were determined by their ability to compete with and displace 2′,3′-O-[2,4,6,-trinitrophenyl]-ATP. These nucleotide affinities are similar to those observed for the E2conformation of the intact Na,K-ATPase.

The Na,K-ATPase (3.6.1.37) is an integral membrane protein that is responsible for maintaining ionic homeostasis in animal cells by mediating the active translocation of sodium and potassium ions against their electrochemical gradients across the plasma membrane. The ion gradients produced by the sodium pump are important for cell excitability and contractility, as well as for regulation of other intracellular ion and solute concentrations. The current model for the sodium pump cycle is conserved for other ion pumps, e.g. the plasma membrane and sarcoplasmic reticulum calcium pumps, the gastric hydrogen/ potassium pump, and the Neurospora proton pump. These ion pumps make up the enzyme class known as the P-type ATPases. Although much information has been gathered about the kinetic mechanism of the sodium pump, there is still rather little detail known about the structure of the protein.
The sodium pump exists as a functional heterodimeric protein consisting of a catalytic ␣-subunit (ϳ110 kDa) and a smaller, glycosylated ␤-subunit (ϳ55 kDa). Both subunits have been cloned, and the primary structure has been determined in several isoforms from a variety of species (1). However, even after extensive study for several decades, information about the specific amino acids involved in formation of the binding sites for physiological ligands is rudimentary. Although the detailed topology of the ␣-subunit of the sodium pump is still the subject of investigation, an overall consensus on structure is emerging; it shows 10 transmembrane segments (3). All the residues implicated with ATP binding thus far have been localized to the major cytoplasmic loop, which is composed of about 430 amino acid residues between transmembrane segments M4 and M5 (2). Moreover, this loop contains four of the most highly conserved P-type ATPase sequences (3). Much of the information suggesting the involvement of various amino acid residues with pump function comes from chemical modification experiments. Such studies have identified a number of residues in the M4M5 loop thought to be involved in ATP binding. For example, Lys 480 , Lys 501 , Gly 502 , Asp 710 , Asp 714 and Lys 719 are all modified by a variety of chemical agents in the absence of ATP but not in its presence, suggesting a role in ATP binding (4). In subsequent studies, some of these residues were changed via mutagenesis with little or no effect on enzyme activity (1). However, measurements of overall enzyme activity only provide information on whether the mutated residue is absolutely essential for function, but they do not necessarily address whether the residue is involved directly in ATP binding. For example, mutation of Asp 369 (i.e. the site of phosphorylation), not surprisingly, abolishes enzyme activity (5), but interestingly, mutation of this residue increases the ATP binding affinity (6). Consequently, to determine whether a residue is important for ATP binding one must directly measure ATP binding and characterize its affinity. The only approach that will identify ATP contact residues is to crystallize the protein in the presence of substrate. Unfortunately, crystallographic analysis of integral membrane proteins still remains difficult because generally applicable strategies for obtaining x-ray quality crystals of these molecules do not yet exist.
As an approach to achieving this aim, we describe the bac-terial production and purification of a soluble polypeptide corresponding to the large cytoplasmic loop of the Na,K-ATPase. This soluble protein is able to bind ATP, as evidenced by the ability of ATP to prevent modification by both fluorescein 5Јisothiocyanate (FITC) 1 and 2-[4Јmaleimidylanilino]napthalene-6-sulfonic acid (MIANS), two fluorescent probes previously demonstrated to label the M4M5 loop of the full-length Na,K-ATPase in an ATP-protectable fashion. In addition, we used 2Ј,3Ј-O-[2,4,6,-trinitrophenyl]adenosine 5Ј-triphosphate (TNP-ATP) fluorescence to estimate, by competition with ATP, the ATP binding affinity to this segment of the sodium pump. A preliminary report of some of this work was presented at the Eighth International Meeting on the sodium pump (7).

EXPERIMENTAL PROCEDURES
Reagents, Media, and Bacterial Strains-NaCl, ATP, ADP, AMP, CTP, GTP, UTP, ethanol, ethidium bromide, bovine serum albumin, ultrapure urea, 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside, and Tris base were purchased from Sigma. ␤-mercaptoethanol, SDS, ammonium persulfate, Coomassie Brilliant Blue R-250, DNA miniprep kit, and low molecular weight standards were from Bio-Rad. 4-(2-Aminoethyl)-benzenesulfonyl fluoride hydrochloride was from ICN. Acrylamide and bisacrylamide were from Boehringer Mannheim. Rainbow gel electrophoresis standards were from Amersham Pharmacia Biotech. FITC, MIANS, and TNP-ATP were from Molecular Probes. Polyvinylidene difluoride electroblotting membrane was from Millipore. Tryptone, granulated agar, and yeast extract were from Difco. Agarose and restriction endonucleases were from Life Technologies, Inc. The pCR-Script cloning kit and Pfu DNA polymerase were from Stratagene. The pET-28 expression vector and Escherichia coli BL21(DE3) cells were from Novagen. Ampicillin and kanamycin were obtained from the University Hospital Pharmacy (Oregon Health Sciences University). E. coli DH5␣ cells were a generous gift from Dr. Linda Kenney (Molecular Microbiology and Immunology, Oregon Health Sciences University). DNA miniprep kits, DNA gel extraction kits, IPTG, and His 6 antibody were from Qiagen. DNA sequencing was perfomed at the core facility at Oregon Health Sciences University.
Construction of the Protein Expression Vector pAN-The portion of the rat ␣1-subunit encoding the M4M5 cytoplasmic loop (Lys 354 -Lys 774 ) was amplified via polymerase chain reaction (PCR) in the presence of 10 M oligonucleotide primers (shown below), 1.2 mM of the four deoxynucleoside triphosphates, and 5 units of Pfu DNA polymerase in 50 l of the manufacturer's buffer. The template was pGEM-rat␣1, a generous gift from Dr. Robert Mercer (Washington University, St. Louis, MO). Twenty-five PCR cycles (30 s at 94°C, 1 min at 53°C, and 2 min at 72°C) were performed in a PTC-100 thermal cycler (MJ Research Inc.). Agarose gel electrophoresis of the PCR revealed a single DNA fragment of 1.2 kilobases.
The forward primer was as follows.

5Ј-GC ATG CAT ATG AAG AAC TGC CTG G-3Ј
NdeI rat ␣1 loop The reverse primer was as follows. TAA GAA TTC TAT TTC TTC AAG TTA TC-

5Ј-GG GTG
After gel purification of the PCR product, a blunt end ligation was performed with the Srf I-digested pCR-Script plasmid according to the manufacturer's protocol (Stratagene) (Fig. 1). (pCR-Script confers ampicillin resistance and contains a portion of the lacZ gene encoding the ␤-galactosidase gene product providing blue-white color selection of recombinant plasmids.) Calcium competent (8) E. coli DH5␣ cells were transformed with the ligation mixture, and positive colonies were selected on LB amp agar plates containing 5-bromo-4-chloro-3-indolyl ␤-D-galactopyranoside and IPTG. White colonies were selected and grown overnight in LB amp , and the DNA was isolated via minipreparation techniques (Bio-Rad and Qiagen DNA miniprep kits). Restriction endonuclease mapping revealed positive clones, which were subsequently sequenced to verify that no random mutations occurred during the PCR. The cloned M4M5 loop was released from pCR-Script by digesting with NdeI and EcoRI; these unique restriction sites were engineered into the oligonucleotide primers (see above). The 6-histidine fusion protein vector, pET-28 (Novagen), was also digested with EcoRI and NdeI. A ligation reaction was performed after gel purifying both DNA fragments at a ratio of 10:1 (insert to vector) using T4 DNA ligase (1 unit/10 ng of DNA) at 25°C for 1 h. The cloned product, called pAN, was subsequently transformed into competent DH5␣ cells and selected by the kanamycin resistance conferred by pET-28. DH5␣ cells have a higher efficiency for transformation with ligation reactions than do BL21 cells. Therefore, we used DH5␣ cells for initial transformations and long term storage of vectors as glycerol stocks, whereas we preferred BL21(DE3) cells for protein expression.
Overexpression of a 6-Histidine-tagged M4M5 Loop-The constructed pAN expression vector was used to transform calcium competent BL21 (DE3) cells. The E. coli transformants were selected on LB kan (30 g/ml) agar plates. A single colony was picked to grow overnight in 5 ml of LB kan , and this culture was subsequently used to inoculate 1 liter of LB kan containing a final concentration of 2% ethanol. After the culture grew to an A 600 of 0.8 -1.0, 1 mM IPTG was added to induce the synthesis of protein from the lac promoter and grown further at 25°C to an A of ϳ2.0.
Cells were then collected and suspended in 30 ml of a lysis buffer containing 50 mM Tris, 100 mM NaCl, pH 8.0, with a hand-held homogenizer. A 10-mg quantity of lysozyme was added in the presence of 150 M 4-(2-aminoethyl)-benzenesulfonyl fluoride hydrochloride (serine protease inhibitor), and the mixture was incubated on ice for 30 min, with occasional plunging of the homogenizer. After addition of 40 mg of deoxycholic acid, the suspension was heated to 37°C with constant stirring. Once the suspension became viscous and difficult to stir, 200 l of deoxyribonuclease (1 mg/ml) was added, and the mixture was incubated at 25°C until the suspension was no longer viscous (ϳ30 min). A final concentration of 1% Triton X-100 was added, and the cell lysate was incubated for an additional 30 min at 25°C. The soluble fraction was separated from the membranous fractions by centrifugation at 12,000 ϫ g for 45 min at 4°C. Expression of the His 6 loop was verified by running a sample of the soluble cell lysate on a 10% Laemmli gel and electroblotting onto polyvinylidene difluoride membrane (10 mM CAPS, pH 11.0, 45 min). The polyvinylidene difluoride membrane was screened with a 1:1000 dilution of a 6-histidine antibody (Qiagen) revealing a protein (ϳ46 kDa) from the lysate of IPTG-induced cells only.
Purification of His 6 Loop via Ni 2ϩ -Column Chromatography-A 4-ml slurry of His-Bind resin (nitriloacetic acid-agarose, Novagen) was placed into a 5-ml column; after the resin settled, a bed volume of 2 ml remained. The column was washed with 10 bed volumes of sterile deionized H 2 O and then charged with nickel by adding 5 bed volumes of 50 mM NiSO 4 . Unbound Ni 2ϩ was washed away with 5 bed volumes of binding buffer (5 mM imidazole, 500 mM NaCl, 20 mM Tris-HCl, pH 7.9). One-third of the soluble cell lysate (ϳ10 ml) was applied to the Ni 2ϩcolumn. The column was subsequently washed with 5 bed volumes of binding buffer to remove nonspecifically bound E. coli proteins. The desired His 6 loop protein was eluted from the Ni 2ϩ -column with a linear imidazole gradient from 50 to 400 mM in a buffer containing 500 mM NaCl and 20 mM Tris (pH 7.9). Twenty-five 1.5-ml fractions were collected (0.5-0.75 ml/min flow rate, gravity flow) over 45-60 min and analyzed for protein content by the method of Bradford (9). Proteincontaining fractions were further analyzed via SDS-polyacrylamide gel electrophoresis (i.e. 10% Laemmli gel). The desired product was routinely recovered over 5-6 fractions ranging from ϳ125 to 250 mM imidazole. The purity varied slightly between preparations; only preparations that were Ն95% pure (based upon densitometry of the Coomassie-stained peptide bands) were used for experimental measurements.
FITC and MIANS Modification Experiments-An aliquot (10 -20 g) of the purified dog kidney Na,K-ATPase or the bacterially produced His 6 loop was incubated with either 5 M FITC or 50 M MIANS for 10 min at 25°C in 20 l of a buffer containing 50 mM Tris (pH 7.4), 2 mM EDTA, and 5 mM of either ATP, ADP, AMP, CTP, GTP, or UTP as indicated in the corresponding figure (Fig. 3, A and B). For experiments testing whether FITC or MIANS could label denatured protein, the protein was incubated in the presence of 1% SDS prior to the chemical treatment. The chemical modification reactions were stopped by the TNP-ATP Binding to Isolated His 6 Loop and Competition by Adenine Nucleotides-TNP-ATP binding to the His 6 loop was performed essentially as described by Moczydlowski and Fortes (18) for TNP-ATP binding to Na,K-ATPase with minor changes. Briefly, fluorescence changes were measured in quartz cuvettes on a model PTI-QM1 (Photon Technology International, Monmouth Junction, NJ) steady state fluorometer. The excitation wavelength was 410 nm (5 nm width), and the emission wavelength was 545 nm (2 nm width). Aliquots of a 1 mM TNP-ATP stock solution were titrated into a 1-ml solution of 50 mM MOPS (pH 7.5) containing 100 g of His 6 loop maintained at 25°C. The same TNP-ATP additions were made into buffer only. The difference between the protein-containing and protein-free solution revealed the fluorescence increases due to interactions between TNP-ATP and the His 6 loop.
Competition with nucleotides was measured by titrating aliquots from a 25 mM stock of ATP, ADP, or AMP into a 1-ml solution of 50 mM MOPS (pH 7.5) containing 1 M TNP-ATP and 10 M His 6 loop. The high protein to TNP-ATP ratio was to ensure that essentially all of the TNP-ATP was bound prior to the nucleotide additions. If there is a significant portion of TNP-ATP in solution, it is difficult to detect the TNP-ATP that is displaced from the His 6 loop by the competing nucleotides. Titrations of the TNP-ATP/His 6 loop solution with buffer were made to determine the fluorescence changes caused by dilution. Subtraction of the dilution effect from the nucleotide titration experiments revealed the fluorescence changes associated with the nucleotide-induced TNP-ATP displacement from the His 6 loop.
Circular Dichroism Measurements-CD spectra were taken on a JASCO J-500 A spectrophotometer. Measurements were made using a 0.1-mm path length cell (Helma) at a constant temperature of 20°C. Data were collected on an IBM/PC-XT using the IF-2 interface; software was provided by Jasco. Spectra and buffer baselines were the average of 10 scans, each recorded at 0.1-nm intervals, using a scanning rate of 5 nm/min and a 2-s time constant. The protein concentration was determined by amino acid analysis and was approximately 2.0 mg/ml. Before spectral deconvolution for secondary structure analysis, the buffer baseline was subtracted and the resulting spectrum was smoothed using the program provided by Jasco. The secondary structure of the His 6 loop was computed using the singular-value and variable-selection methods of Compton et al. (11).

RESULTS
Our goal was to characterize the isolated ATP binding domain of the Na,K-ATPase. As a means to this end, we devel-oped a method to produce large quantities of a soluble peptide corresponding to the cytoplasmic loop between transmembrane segments M4 and M5 (i.e. from Lys 354 to Lys 774 , rat ␣1 subunit).
Overexpression and Purification of the His 6 ␣-Loop Fusion Protein-The plasmid pAN was constructed by inserting the cDNA encoding the rat ␣1 M4M5 cytoplasmic loop into pET-28 multiple cloning site at the ndeI (5Ј) and the EcoRI (3Ј) restriction site locations downstream from the histidine coding sequence (Fig. 1). pAN was used to transform a BL21(DE3) E. coli strain. For induction of gene expression, cells were grown in 1 liter of LB kan medium containing a final ethanol concentration of 2%, and fusion protein production was induced with IPTG.
(The presence of ethanol significantly increased the amount of fusion protein in the soluble fraction.) This process routinely yielded significant production of the His 6 loop fusion protein (molecular mass, ϳ46 kDa) with approximately 40% associated with the soluble fraction (Fig. 2). The His 6 loop was purified via a Ni-NTA affinity column (ϳ20 -25 mg/liter; Fig. 2).
FITC Labeling of the His 6 Loop-FITC is a fluorescent amine-reactive molecule that labels Lys 501 in the purified Na,K-ATPase; this reaction is prevented by the simultaneous presence of ATP (12,13). In this study, we used ATP protection against FITC labeling as a tool to demonstrate that ATP binds to His 6 loop (Fig. 3A). It is clear that after incubation with 5 M FITC, both the purified Na,K-ATPase and the His 6 loop were labeled by FITC. Moreover, when the FITC incubation was performed in the presence of either 5 mM ATP or 5 mM AMP, only ATP (not AMP) prevented FITC modification of both the intact purified sodium pump and the His 6 loop (Fig. 3A).
To further investigate the nucleotide specificity of the His 6 loop, we tested the ability of several nucleotides to protect against FITC modification (Fig. 3B). Purified His 6 loop was incubated with 5 M FITC in the presence of 5 mM ATP, ADP, AMP, GTP, and UTP. It is clear that both ATP and ADP protected against FITC, whereas AMP, CTP, GTP, and UTP were unable to prevent FITC modification. These data are in agreement with the nucleotide specificity demonstrated for the purified Na,K-ATPase, which has been shown to bind ATP and ADP with high affinity, (14). Therefore, it appears that the His 6 loop retains the structural parameters that confer nucleotide specificity in the intact enzyme.
Structural Analysis of the His 6 Loop-The structural integ- rity of the His 6 loop was directly reflected by the structural requirement for FITC modification, i.e. after we denatured the His 6 loop with 1% SDS, incubation with FITC failed to modify the peptide (Fig. 4). Similar results were observed when we denatured purified dog Na,K-ATPase with 1% SDS (Fig. 4). These results are not due to an anomalous effect of SDS on the FITC reaction; FITC was also unable to modify the His 6 loop after denaturation with 6 M guanidine-HCl or by repeated freezing and thawing (data not shown).
More directly, the overall secondary structure of the His 6 loop fusion protein was estimated by CD spectroscopy. As shown in Fig. 5, the resultant spectrum, which summarizes the mean residue molar ellipticity as a function of wavelength, exhibits a maximum at 188 nm and two minima, one at 206 nm and the other at 221 nm. The spectrum is the average of 10 scans, the general shape of which did not change. Analysis of the CD spectrum of the His 6 loop, calculated using the singular-value decomposition method (11), predicts secondary structural elements distributed approximately as 23% ␣-helix, 23% antiparallel ␤-sheet, 4% parallel ␤-sheet, 19% ␤-turn, and 32% random coil. The total of the fractions in this method is not constrained to be 100%, but rather should lie between 95 and 105%. The observation that the components sum to close to 100% indicates that the fusion protein is highly structured as assessed by this technique.
MIANS Labeling of the His 6 Loop-Recently, the fluorescent sulfhydryl reagent MIANS was shown to inactivate the dog kidney Na,K-ATPase by specifically labeling a cysteine residue in the M4M5 loop of the Na,K-ATPase (15). Both MIANS labeling and enzyme inactivation were prevented by the simultaneous presence of ATP. 2 In this study, we tested whether ATP was able to protect against MIANS modification of the His 6 loop. Indeed, ATP, but not AMP prevented MIANS labeling of the bacterially produced His 6 loop (Fig. 6). However, unlike FITC, MIANS does label the denatured peptide, but ATP no longer protects against its modification (Fig. 6). The simplest explanation for a lack of ATP protection is that there is no longer significant structure to bind ATP.
TNP-ATP Binding to the His 6 Loop and Competition with Nucleotides-The interactions of the His 6 loop and TNP-ATP were studied. This fluorescent ATP analog has been shown to bind with high affinity to the native Na,K-ATPase, as well as to other members of the P-type ATPase family (17)(18)(19). TNP-ATP is a useful fluorescent probe for studying the nucleotide binding  site because its fluorescence changes significantly on binding. When an aliquot of His 6 loop was added to a solution containing 5 M TNP-ATP, it resulted in an enhancement of the TNP-ATP fluorescence consistent with TNP-ATP binding to the engineered protein (Fig. 7). The addition of ATP to a solution containing His 6 loop and TNP-ATP resulted in a decrease in the fluorescent signal, as would be expected if ATP displaced TNP-ATP from the His 6 loop (Fig. 7).
The apparent binding affinity of the His 6 loop for TNP-ATP was determined by titrating small aliquots of TNP-ATP into a solution containing His 6 loop. Plotting the fluorescence change against [TNP-ATP] showed that the binding of TNP-ATP to the His-tagged construct was composed of two parts: a saturable component and a nonsaturable component (Fig. 8A). Subtraction of the nonsaturable component from all the data revealed the fluorescence changes due to specific TNP-ATP binding (Fig.  8B). The apparent affinity of the His 6 loop for TNP-ATP was ϳ3 M (Table I).
The apparent affinities for ATP, ADP, and AMP were determined from their ability to displace bound TNP-ATP from the His 6 loop. For these experiments, a 10-fold molar excess of the His 6 loop protein compared with TNP-ATP was used to ensure that all (i.e. Ͼ99%) of the TNP-ATP was bound and that the contribution from the nonsaturable TNP-ATP binding component was zero (20). The competition of TNP-ATP binding by nucleotides is evidenced by a decrease in fluorescence intensity upon the release of TNP-ATP from the His 6 loop. A plot of [nucleotide] versus percentage of change in fluorescence from a typical experiment is shown in Fig. 9. The apparent K d values for ATP and ADP were ϳ350 M and ϳ550 M, respectively (Table I). AMP showed no saturable binding in the concentration range examined. For each nucleotide, the binding experiments were performed on three separate preparations of His 6 loop with comparable results. DISCUSSION We have produced a soluble polypeptide by bacterial overexpression that is identical in sequence with the central loop of the rat ␣1-subunit. We have shown that the peptide binds ATP and ADP, but not other nucleoside triphosphates, with the same specificity as native Na,K-ATPase. This highly ordered peptide also shows labeling and protection reactions characteristic of the intact Na,K-ATPase.
Chemical modification experiments (4, 2) of the Na,K-ATPase, as well as site-directed mutagenesis studies (1), suggest that several amino acids located between the fourth and fifth putative transmembrane segments participate in the coordination of ATP. However, in some instances, the data generated from these two methods appear to be contradictory. For example, labeling Lys 501 with a number of reagents (e.g. FITC, SITS, and N-(2-nitro-4-isothiocyanophenyl)-imidazole) completely inactivates the enzyme; the prior binding of ATP prevents both modification and inactivation, consistent with Lys 501 playing a role in ATP binding. However, when Lys 501 was changed to methionine via site-directed mutagenesis (21), the enzyme retained activity, demonstrating that Lys 501 is not essential for enzyme activity. How could Lys 501 participate in ATP binding and yet not be required for enzyme activity? It is possible that Lys 501 is one of several residues that form the . There was a significant increase in the fluorescence intensity of TNP-ATP after the addition of protein, consistent with TNP-ATP binding to the His 6 loop. Also, subsequent additions of ATP compete with bound TNP-ATP, causing a decrease in the fluorescence signal. These data are from a single experiment in which ATP decreased the TNP-ATP fluorescence by ϳ80%. Over the course of all our experiments, the maximal TNP-ATP fluorescence change ranged from 30 to 80% and was typically between 40 and 50%. ATP binding pocket. Loss of a single contact residue might result in a lower ATP affinity, but not in a loss of ATP binding. Therefore, in the presence of saturating ATP, Na,K-ATPase activity might remain normal. However, attaching a chemical reagent to a contact residue or a nearby residue not only removes that residue from coordination but also occupies space in the vicinity of that residue; thus, modification of Lys 501 may prevent ATP binding via steric factors. Consequently, it is becoming readily apparent that a detailed three-dimensional structure is needed to conclusively identify the contact sites for ATP and to adequately describe its binding pocket.
To approach such experiments, the major cytoplasmic do-main between M4 and M5 of the Na,K-ATPase was overexpressed in E. coli. Similar approaches have been employed by other laboratories to isolate the ATP binding domains of the yeast proton pump (22), the sarcoplasmic reticulum calcium pump (23), the sodium pump (24), and the cystic fibrosis transmembrane conductance regulator (25). The cDNA from the rat ␣1 subunit, encoding 420 amino acids from Lys 354 -Lys 774 , was cloned into a fusion protein vector (His 6 tag, pET-28) and transformed into E. coli. This method routinely yields approximately 20 mg of pure soluble peptide per liter of cell culture. We demonstrated that this expressed and purified peptide 1) retains an ordered structure, 2) can be labeled by both FITC and MIANS, 3) binds both ATP and ADP, and 4) binds the fluorescent ATP analog, TNP-ATP. Experiments are currently under way to crystalize the His 6 loop. Structural Analysis of the His 6 Loop-Detailed structural topology of the sodium pump awaits three-dimensional x-ray crystallographic analysis. However, some assessments have been made that suggest an ␣-subunit structure with 10 transmembrane segments. Also, all the residues associated with ATP binding have been localized to the major cytoplasmic loop between transmembrane segments M4 and M5 (2). Furthermore, the residues thus far implicated in nucleotide binding among all the members of the P-type ATPase family have been assigned to this cytosolic domain. Structural models of the nucleotide binding pocket of P-type ATPases have been proposed based upon sequence homology with different kinases (e.g. adenylate kinase and phosphoglycerate kinase; see Ref. 26). According to these models, the large cytoplasmic loop is divided into three domains: a phosphorylation domain, a nucleotide binding domain, and a central domain. We calculated the ␣-helix and ␤-sheet content of the nucleotide binding domain portion (Arg 524 -Ile 654 ) of the sheep Na,K-ATPase cytoplasmic loop from the data previously reported (26). According to their predictions using the method of Chou and Fasman (27), we obtained a composition of about 35% ␣-helix and 22.5% ␤-sheet. These values agree reasonably well with our current findings of 23% ␣-helix and 27% ␤-sheet for the secondary structure of the bacterially expressed M4M5 loop. The difference in ␣-helical content may suggest that the phosphorylation domain and the central domain (included in our analysis and not in that of Taylor and Green (26)) have significantly less helical structure than does the nucleotide binding domain. Alternatively, it may simply reflect the difference between CD analysis (this study) and primary structure comparison (26) or a less than completely retained structure of our His 6 loop compared with native intact Na,K-ATPase.
Nucleotide Protection against Chemical Modification of the His 6 Loop-For almost two decades, it has been known that FITC irreversibly inhibits the Na,K-ATPase in an ATP-protect-  Table I. able manner (28). The site of FITC modification was identified as Lys 501 (12,13). Considerable evidence suggests that Lys 501 resides in the ATP binding site of the sodium pump; this evidence includes the following: 1) ATP prevents modification of Lys 501 (28), 2) the reactivity of Lys 501 is sensitive to cation binding in a way similar to the effect that cation occupancy has on direct ATP binding (29 -31; 15), and 3) an equivalent lysine residue exists in a conserved sequence among most members of the P-type II ATPase family (3). In a fashion similar to the intact Na,K-ATPase, ATP and ADP (but not AMP) protect against FITC modification of our purified His 6 loop (Fig. 3). Moreover, nucleotides that do not bind with high affinity to the native enzyme also do not protect the His 6 loop from FITC modification (Fig. 3). These findings demonstrate that the His 6 loop has folded sufficiently to allow formation of a nucleotide binding pocket that shows selectivity. Indeed, denaturing the His 6 loop prior to FITC treatment results in no FITC labeling; the same finding was observed with the native Na,K-ATPase (Fig. 4). Thus, it appears that the selective reactivity of Lys 501 toward FITC is a product of the special environment in the folded central loop. The appropriate folding of this ATP binding loop generates a highly reactive lysine at position 501. This is in contrast with the behavior of MIANS and Cys 577 (see below).
We have observed that ATP, but not AMP, can also protect the loop against modification by the sulfhydryl reagent MIANS (Fig. 6). Recently, MIANS has been shown to modify a specific cysteine residue in the large cytoplasmic loop of purified Na,K-ATPase; ATP protects against MIANS modification. Proteolytic digestion and N-terminal amino acid sequencing has identified the MIANS-modified residue as either Cys 549 or Cys 577 . 2 ATP protection against modification at Lys 501 and Cys 577 (or Cys 549 ) suggests that a compact folding of the loop may bring the distant segments of the peptide together. It is interesting, however, that after denaturation of this loop, modification still occurs with MIANS, but such modification is no longer affected by the simultaneous presence of ATP.
Nucleotide Binding Affinity-The sodium pump and most other P-type ATPases have a complex ATP dependence; a high affinity ATP effect and a low affinity ATP effect. In the sodium pump, phosphorylation of E 1 to E 1 ϳP requires Na, Mg and ATP. The K m value for Na-ATPase (K m for ATP Յ1 M) agrees well with the measured ATP binding affinity (32,33). At low ATP concentrations, the rate of hydrolysis is slow and limited by the release of potassium (34). However, higher ATP concentrations (K m(ATP) Ն 100 M) facilitate the deocclusion of potassium. This low affinity effect is seen as the K m for ATP under (Na ϩ K)-ATPase conditions. ATP activation of the sodium pump, with both high and low apparent affinities, has been interpreted as being due to either two distinct ATP sites or to a single ATP binding region that alters its affinity in different pump conformations.
Using the fluorescent ATP analog, TNP-ATP, we were able to estimate the ATP affinity of the His 6 loop. In experiments designed to determine the His 6 loop affinity for TNP-ATP, we discovered that there were two components to the binding of this ATP analog, a saturable and nonsaturable component. Hellen and Pratap (20) reported a similar two-component binding curve for TNP-ATP binding to the native Na,K-ATPase. Subtracting the nonspecific binding from the data, revealed a specific saturable component with an apparent K d of ϳ3 M for TNP-ATP ( Fig. 8B; Table I). This value is about an order of magnitude higher than the Na,K-ATPase binding affinity for TNP-ATP (ϳ0.5 M at 25°C; Ref. 17), i.e. it appears to represent a low affinity site.
An affinity of 300 -400 M was found for ATP by competition FIG. 9. Competition of TNP-ATP binding by adenine nucleotides. The fluorescence of His 6 loop bound TNP-ATP was observed at 545 nm after excitation at 410 nm. Aliquots from concentrated solutions of either ATP (A), ADP (B), and AMP (C) were then added to mixture, and the fluorescence changes were monitored. The decrease in fluorescence due to dilution were subtracted from the data. The change in fluorescence due to nucleotide binding was plotted against the concentration of the corresponding nucleotide. Data were fit to the same equation as in Fig. 8B. The mean Ϯ S.E. for the K d from three separate experiments are shown in Table I. Measurements were also made in the absence of protein, and no effects of the nucleotides were observed on the free TNP-ATP fluorescence. and displacement of TNP-ATP (Table I). This value is in reasonable agreement with the low ATP affinity for the E 2 (K 2 ) state of the native Na,K-ATPase associated with potassium deocclusion. In addition, the expressed domain has a similar affinity (K d 500 -600 M; Table I) for ADP. This is consistent with the observation that in the low affinity E 2 -state of the sodium pump, ADP has been shown to facilitate potassium deocclusion and transport (16, [35][36]. AMP concentrations up to 3 mM were not sufficient to displace TNP-ATP (Fig. 9C), and concentrations of up to 5 mM were unable to protect against FITC modification (Fig. 3).
Recently, the nucleotide binding domains of the sarcoplasmic reticulum calcium pump (23), the yeast proton pump (22), and the sodium pump (24) have been expressed in E. coli. In the sodium pump-expressed domain, ATP binding was demonstrated by measuring MgATP protection against photolabeling with 2-N 3 -ATP 32 ; however, Mg alone appeared to protect as well as MgATP (24). All our experiments were performed in the absence of magnesium; thus, the effects are solely due to the nucleotides themselves. ATP binding to expressed loops from the proton pump and sarcoplasmic reticulum calcium pump was estimated by ATP competition of TNP-ATP, as in this study. The affinities for TNP-ATP binding to the calcium pump and proton pump domains were 2 and 6 M, respectively, similar to our value of ϳ3 M for the sodium pump domain. In contrast, the ATP affinity reported for the proton pump domain was ϳ3 mM (22), significantly different from that reported for the calcium pump domain (ϳ200 M; Ref. 23) and here for the sodium pump domain (ϳ300 M; Table I). The considerably lower affinity for the proton pump ATP binding domain may be due to an inherent property of the domain or possibly because the proton pump construct was a glutathione S-transferase fusion protein, whereas the calcium pump and sodium pump constructs were both His-tagged proteins. Indeed, we were unable to successfully measure the ATP affinity for the glutathione S-transferase version of our construct even though we were able to demonstrate that ATP protected against FITC modification. It turned out that ATP (5 mM) also protected against FITC labeling of glutathione S-transferase alone. 3 It appears, then, that expression of the isolated central loops of these P-type ATPases produces a protein that is able to selectively bind ATP (or ADP) but with an affinity close to that seen in E 2 forms. Because the isolated loop has considerable secondary structure, it seems reasonable to suppose that the high affinity binding form is generated by interactions with other parts of the ATPase (probably the cation binding domains). One way of modeling these changes in the ATP binding site would be to suppose that there are two states or forms, an R (relaxed) form and a T (tense) form, the R form having lower affinity for ATP and the T form, higher affinity (Ͻ1 M). The changes in structure of the intact protein that we identify as E 1 (high sodium affinity and high ATP affinity) and E 2 (high potassium affinity and low ATP affinity) are mirrored by changes in T and R forms, respectively, in the ATP binding loop. It is interactions between the ATP binding domain in the loop and other segments of the protein that hold the loop in the tense form, which has a high affinity for ATP. In isolation, the loop is not constrained by these interactions, and a relaxed (R) form exists with low substrate affinity.
The present study of ATP binding to the purified M4M5 loop of the Na,K-ATPase provides a basis for future mutagenesis studies of the residues thought to play a role in ATP binding. Furthermore, the ability to obtain large quantities of pure soluble protein makes this method valuable for detailed structural analyses of the wild-type and mutant ATP binding domains of P-type ATPases.