Molecular Distance Measurements Reveal an (αβ)2Dimeric Structure of Na+/K+-ATPase

ATP hydrolysis by Na+/K+-ATPase proceeds via the interaction of simultaneously existing and cooperating high (E1ATP) and low (E2ATP) substrate binding sites. It is unclear whether both ATP sites reside on the same or on different catalytic α-subunits. To answer this question, we looked for a fluorescent label for the E2ATP site that would be suitable for distance measurements by Förster energy transfer after affinity labeling of the E1ATP site by fluorescein 5′-isothiocyanate (FITC). Erythrosin 5′-isothiocyanate (ErITC) inactivated, in an E1ATP site-blocked enzyme (by FITC), the residual activity of the E2ATP site, namely K+-activated p-nitrophenylphosphatase in a concentration-dependent way that was ATP-protectable. The molar ratios of FITC/α-subunit of 0.6 and of ErITC/α-subunit of 0.48 indicate 2 ATP sites per (αβ)2 diprotomer. Measurements of Förster energy transfer between the FITC-labeled E1ATP and the ErITC-labeled or Co(NH3)4ATP-inactivated E2ATP sites gave a distance of 6.45 ± 0.64 nm. This distance excludes 2 ATP sites per α-subunit since the diameter of α is 4–5 nm. Förster energy transfer between cardiac glycoside binding sites labeled with anthroylouabain and fluoresceinylethylenediamino ouabain gave a distance of 4.9 ± 0.5 nm. Hence all data are consistent with the hypothesis that Na+/K+-ATPase in cellular membranes is an (αβ)2 diprotomer and works as a functional dimer (Thoenges, D., and Schoner, W. (1997) J. Biol. Chem.272, 16315–16321).

location of the low affinity E 2 ATP binding site may be obtained by similar means, i.e. affinity labeling followed by amino acid sequencing and by energy transfer experiments after specific modification of the E 1 ATP site by a fluorophor. Affinity labeling of the low affinity E 2 ATP site needs the availability of specific ATP derivatives as probes (4). Since many derivatives of fluorescein (pyrene, eosin, or erythrosin) bind to the E 1 ATP site as well (32)(33)(34)(35), the possibility arose that fluorescent pseudo-ATP derivatives may exist that label the E 2 ATP site as well. Provided specific labeling of the E 2 ATP binding site by a fluorescent pseudo-ATP derivative can be achieved, Förster energy transfer measurements may give information on the distance of this E 2 ATP site to the E 1 ATP site, since the latter can be labeled specifically by FITC at Lys 501 (26,27,36). Hence, information may be obtained from the availability of such a system on the question as to whether both ATP binding sites are close or distant from each other. This paper shows that a pseudo-ATP binding for the E 2 ATP site exists. Erythrosin 5Јisothiocyanate (ErITC) inactivates the residual K ϩ -activated phosphatase (an activity of the E 2 ATP site) in an enzyme whose E 1 ATP had already been blocked by labeling with FITC. Energy transfer from the E 1 ATP-labeled fluorescein to the E 2 ATP-labeled erythrosin (as well to the Co 2ϩ ions sitting in the E 2 ATP site) was so low that it is unlikely that both ATP sites reside on the same catalytic ␣-subunit.

EXPERIMENTAL PROCEDURES
All chemicals were of the highest purity available and were obtained from Bio-Rad, Boehringer Mannheim, Merck, and Molecular Probes (Eugene, OR). Lab-Trol protein standard is a product of Merz & Dade (Munich, Germany). [␥-32 P]ATP was from Amersham Pharmacia Biotech. Calculation and presentation of data were performed with the program Graph-Pad-Prism 2.01 (Graph Pad Software Inc., San Diego, CA).
Preparation of the MgATP Complex Analogs-The synthesis of Cr(H 2 O) 4 ATP, Cr(H 2 O) 4 AdoPP[CH 2 ]P, and Co(NH 3 ) 4 ATP was performed by the aniline method of Cleland et al. (37) with the variations described earlier (22).
Enzyme and Assays-Na ϩ /K ϩ -ATPase from pig kidney with a specific enzymatic activity of 25-27 units/mg of protein was isolated by a modification of Jørgensen's procedure (38) and measured by coupled spectrophotometric assay (39). One enzyme unit is defined as the amount of enzyme hydrolyzing 1 mol of ATP/min at 37°C. Protein was determined by the method of Lowry et al. (40) using Lab-Trol as the protein standard. Lab-Trol is a mixture of proteins and enzymes used for the calibration of assays in clinical chemical analysis. All buffers used were made up to their respective pH value at room temperature.
K ϩ -activated p-nitrophenylphosphatase activity was measured on a multititer plate by incubating Na ϩ /K ϩ -ATPase at 37°C in a total volume of 150 l containing 61 mM Tris/HCl (pH 7.25), 6.4 mM MgCl 2 , 12 mM KCl, and 5 mM p-nitrophenylphosphate. The reaction was stopped after 15 min by the addition of 200 l of 3 N NaOH. The p-nitrophenolate formed was measured at 405 nm by an enzyme-linked immunosorbent assay reader (8).
Inactivation of Na ϩ /K ϩ -ATPase with FITC-Na ϩ /K ϩ -ATPase at a final concentration of 1 unit/ml (65 g/ml) was incubated overnight at 37°C in a solution containing 20 mM Tris/HCl (pH 7.25), 15 mM NaCl, and 10 M FITC. The inactivated enzyme was sedimented in Eppendorf tubes at 100,000 ϫ g in a 50Ti rotor using adapters of our own design. The pellet was resuspended in 1 ml of 20 mM Tris/HCl (pH 7.25), and the Na ϩ /K ϩ -ATPase activity (39) was determined. The residual activity was less than 5% initial activity.
Kinetic Analysis of the Interaction of Na ϩ /K ϩ -ATPase with ErITC from a Study of the Overall Na ϩ /K ϩ -ATPase Activity-Na ϩ /K ϩ -ATPase was inactivated with varying concentrations of ErITC as follows. 1 unit of Na ϩ /K ϩ -ATPase was incubated in a total volume of 1 ml with ErITC (0 -1 mM) in 20 mM Tris/HCl (pH 7.25) and 15 mM NaCl at 37°C. The residual Na ϩ /K ϩ -ATPase activity was assayed by transferring aliquots of 50 l to the optical assay (39). To detect whether ErITC binds to the ATP site, the enzyme was inactivated with 200 nM ErITC in the presence of 0, 2.5, or 25 mM Na 2 ATP (adjusted to pH 7.25 with Tris base). The inactivation constant K i was calculated according to Piszkiewics and Smith (41).
Kinetic Analysis of the Inactivation of K ϩ -activated p-Nitrophenylphosphatase by ErITC in a FITC-inactivated Na ϩ /K ϩ -ATPase-FITC is a covalent label for the high affinity E 1 ATP binding site of Na ϩ /K ϩ -ATPase that leaves the activity of the E 2 ATP site almost unaffected (8,11,25,36). Therefore, the effects of ErITC on K ϩ -activated p-nitrophenylphosphatase, an activity of the low affinity E 2 ATP site, were tested in the following way.
Na ϩ /K ϩ -ATPase at a final concentration of 1 unit/ml (65 g/ml) was incubated overnight at 37°C in a solution containing 20 mM Tris/HCl (pH 7.25) and 10 M FITC. A control enzyme was treated in the same way but without FITC. This control was set at 100%. The inactivated enzyme (residual Na ϩ /K ϩ -ATPase activity of 1%) was spun down in Eppendorf tubes at 100,000 ϫ g. The pellet was washed in 20 mM Tris/HCl (pH 7.25) and resuspended in a solution of 20 mM Tris/HCl (pH 7.25), 15 mM NaCl, and various concentrations (0 -10 M) of ErITC. After incubation for 15 min at 37°C, the K ϩ -activated p-nitrophenylphosphatase activity was estimated.
Analysis of the Effect of Co(NH 3 ) 4 ATP on the Incorporation of ErITC in FITC-treated Na ϩ /K ϩ -ATPase-Co(NH 3 ) 4 ATP is a specific inhibitor of the E 2 ATP site of Na ϩ /K ϩ -ATPase (9,22). It was used, therefore, to detect interferences of ErITC with this site.
Na ϩ /K ϩ -ATPase at a final concentration of 2.5 units/ml (163 g/ml) was incubated overnight at 37°C in a solution containing 20 mM Tris/ HCl (pH 7.25), 15 mM NaCl, and 10 M FITC. The enzyme was spun down at 100,000 ϫ g, and washed with 20 mM Tris/HCl (pH 7.25), and resuspended in 20 mM Tris/HCl (pH 7.25), 15 mM NaCl, and 1 mM Co(NH 3 ) 4 ATP (the control contained no Co(NH 3 ) 4 ATP). After a 1-h incubation at 37°C, the enzyme was spun down and washed again. After incubation at 37°C for 1 additional h in a solution containing 20 mM Tris/HCl, 15 mM NaCl, and different concentrations of ErITC (0 -1 M), the enzyme was spun down at 100,000 ϫ g and washed three times with 20 mM Tris/HCl (pH 7.25). The fluorescence of ErITC was detected in a Hitachi F-3000 spectrofluorometer at ex 530 nm and em 555 nm. A probe labeled with FITC served as control.
Effect of ErITC on Na ϩ -dependent Phosphorylation from [␥-32 P]ATP into a Co(NH 3 )4PO4-pretreated Enzyme-Co(NH 3 ) 4 PO 4 is an inhibitor of the E 2 ATP site (11,12), and Na ϩ -dependent phosphorylation (frontdoor phosphorylation) detects an activity of the E 1 ATP site (1,4,21). Interference by ErITC of the E 1 ATP site can be detected from the change of the Na ϩ -dependent phosphorylation of Na ϩ /K ϩ -ATPase by the chosen isothiocyanate in an enzyme whose E 2 ATP site had been blocked with Co(NH 3 ) 4 PO 4 .
Na ϩ /K ϩ -ATPase at a final concentration of 1 units/ml (65 g/ml) was incubated overnight at 37°C in a solution containing 20 mM Tris/HCl (pH 7.25) and 50 mM Co(NH 3 ) 4 PO 4 (a control without Co(NH 3 ) 4 PO 4 was run in parallel). The inactivated enzyme was sedimented at 100,000 ϫ g, washed in 20 mM Tris/HCl (pH 7.25), and incubated in 20 mM Tris/HCl, 15 mM NaCl, and 2 M ErITC. After different incubation times at 37°C, the enzyme was spun down and washed twice with 20 mM Tris/HCl (pH 7.25), and front-door phosphorylation was measured as follows. The enzyme in 100 mM NaCl, 1 mM MgCl 2 , and 10 mM imidazole/HCl (pH 7.25) was placed on ice. The reaction was started by the addition of 100 l of 1 mM [␥-32 P]ATP (200 cpm/pmol) so that the final concentration was 0.1 mM. The reaction was stopped after 1 min by the addition of 250 l of 25% trichloroacetic acid, 10 mM Na 2 HPO 4 , and 1 mM unlabeled ATP. The mixture was centrifuged at 100,000 ϫ g for 30 min. The pellet was washed three times with 500 l of 5% trichloroacetic acid containing 2 mM Na 2 HPO 4 and 0.2 mM unlabeled ATP. Background labeling of 1 unit of Na ϩ /K ϩ -ATPase that was quenched first with 250 l of 25% trichloroacetic acid prior to the addition of the phosphorylation mixture was subtracted from all samples.
Labeling of Na ϩ /K ϩ -ATPase for Lifetime Measurements of Fluorescence-Na ϩ /K ϩ -ATPase (6 units, 300 g) was incubated in a total volume of 1 ml overnight at 37°C in a solution containing 20 mM Tris/HCl (pH 7.25), 15 mM NaCl, and 10 M FITC (a control contained no FITC). This inactivated enzyme was centrifuged in Eppendorf tubes at 100,000 ϫ g. The pellet was resuspended in 1 ml of a solution containing 20 mM Tris/HCl (pH 7.25), 15 mM KCl, and 2 M ErITC or 1 mM Co(NH 3 ) 4 ATP. This preparation was incubated additionally for 3 h at 37°C, centrifuged, and washed twice in 20 mM Tris/HCl (pH 7.25). After the last centrifugation step, the protein was resuspended in 0.3 ml of 20 mM Tris/HCl buffer (pH 7.25) (final concentration 1.8 mg/ml). All lifetime measurements of FITC-labeled Na ϩ /K ϩ -ATPase were performed in the presence of 5 g/ml antifluorescein antibodies to correct for contributions by free and nonspecifically attached FITC molecules (34).
To study the effects on the individual ␣-subunit, the labeled enzyme was solubilized by incubation with 5% SDS at 37°C for 15 min to break any protein-protein interactions of the ␣-subunits. The energy transfer was measured in the presence of SDS in the same way.
Specific binding of the ouabain derivatives anthroylouabain (AO) and FEDO to Na ϩ /K ϩ -ATPase proceeded under conditions of backdoor phosphorylation (8,42). Na ϩ /K ϩ -ATPase (6 units) was incubated in 200 mM imidazole/HCl buffer (pH 7.25), 3 mM MgCl 2 , 10 mM imidazole/phosphate (pH 7.25), and 150 nM AO for 1 h at 37°C. To prepare the sample labeled with donor (AO) and acceptor (FEDO), the enzyme was incubated under the same conditions but in the presence of 150 nM AO and 300 nM FEDO. This concentration was used because the affinity of Na ϩ /K ϩ -ATPase for AO was twice as high as that for FEDO (42).
Determination of the Amount of FITC, ErITC, AO, FEDO, and Co(NH 3 ) 4 ATP Bound to Na ϩ /K ϩ -ATPase-A Specol 211 spectrophotometer was used for absorbance measurements. Steady-state fluorescence data were collected in quartz cuvettes on a Perkin-Elmer LS-5 fluorometer equipped with monochromators (34). Excitation and emission wavelengths were 500 and 520 nm, respectively, for FITC and FEDO, 362 and 471 nm for AO, and 530 and 555 nm for ErITC, respectively. Two Glan-Thompson polarizers were used for determination of the steady-state anisotropy values. All measurements were performed at 25°C. The molar ratio of bound fluorophores per ␣-subunit was determined from steady-state fluorescence measurements based on the known quantum yield of standards of known concentrations. Quantum yield of bound FITC was compared with free FITC in ethanol, and bound ErITC was compared with free ErITC in water.
Although the former increases about 10%, the quantum yield of the latter increases four times (34). For the determination of the molar FITC/ErITC ratio per mol of ␣-subunit, we labeled Na ϩ /K ϩ -ATPase at a specific activity of 25-27 units/mg of protein first with 10 M FITC at pH 7.25 and subsequently with 2 M ErITC (as described above). Steady-state fluorescence of the protein-bound fluorescein was corrected for unspecific fluorescence (outside the E 1 ATP binding site) by the addition of 5 g/ml antifluorescein antibodies (34). The molar concentration of Na ϩ /K ϩ -ATPase and its ␣-subunit was calculated from the protein concentration using the molecular weight of 113 kDa for the ␣and 55 kDa for the ␤-subunit (3).
Fluorescence Lifetime Measurements-The samples were measured in a total volume of 600 l at a final protein concentration of 1.8 mg/ml. The apparatus for lifetime measurements was based on a laser excitation source and on time-correlated single photon counting as the detection system. The excitation source consisted of a cavity-dumped dye laser (model 375, Spectra Physics, Mountain View, CA) synchronously pumped by the argon ion laser (model 171, Spectra Physics) and a frequency doubler. The excitation pulses (full width at half maximum about 10 ps) were generated at 356 nm with pyridine 1 as the laser dye. The required emission wavelength was selected by a monochromator with a proper cut-off filter in front of the input slit. The fluorescence decays were measured with the emission polarizer oriented at the "magic angle" of 54.7 degrees to the direction of the excitation polarization vector. The response function of the apparatus was determined by the REF procedure of Vecer (43) with FITC and potassium iodidequenched FITC as reference compounds. In the case of AO as fluorescence donor, 1,4-bis[2-(5-phenyloxazolyl)]benzene was used as the reference compound. All experiments were repeated with an unlabeled sample to correct for the fluorescence background and light scattering. The decay data were analyzed by a nonlinear least squares deconvolution procedure. The fit quality was evaluated from the randomness of the residual plot and the autocorrelation function together with the R 2 value.
Calculation of the Förster Resonance Energy Transfer and of Molecular Distances-Distances (R) between donor and acceptor pairs were derived from the apparent efficiencies of Förster energy transfer (E), which was calculated from the quenching of the steady-state fluorescence intensity of the donor or from the decrease of donor lifetime ( D , presence of donor only; DA , presence of donor and acceptor). The apparent efficiency of energy transfer is related to the absolute rate of energy transfer (k T ) as shown below.
This rate has been defined by Förster (44) as where ⌽ D is the quantum yield of the donor in the absence of acceptor, n is the refractive index of the solution, R is the distance between donor and acceptor (in cm), J is the spectral overlap integral (cm 3 /M) defined as where ⑀ A () is the extinction coefficient of the acceptor, and F D ( ex ,) is the fluorescence of the donor (excited at ex ) that is emitted at wavelength . Finally, 2 is the orientation factor of the donor, which is defined as and ␣ and ␤ are, respectively, the angles that the emission dipole of the donor and the absorption dipole acceptor's form with vector R. Vector R connects these two dipoles. P is then the angle between the planes that contain ␣ and ␤. The resulting Förster critical distance (R 0 ), which is the distance in the case of E ϭ 0.5, is Labeling of the enzyme used in this work (pH at 7.25) resulted in the donor and acceptor anisotropy values low enough to take the orientational factor 2 for the mutual distance calculations as 2/3, assuming random mutual orientation of fluorophores (45). For determination of the overlap integral (J), the values of ⑀ and ⌽ D were taken from Amler et al. (34). An excess of donor concentration over acceptor concentration was corrected for using rates of energy transfer (34). Briefly, the corrected rate of energy transfer k C was calculated as where k is the observed rate of energy transfer, and n D and n A are the molar ratios of bound donor and acceptor, respectively.

Characterization of Erythrosin
Isothiocyanate as a Label of ATP Sites-ErITC has been formerly used as a label for the high affinity E 1 ATP site (34). To certify that ErITC is a useful label for both ATP sites of Na ϩ /K ϩ -ATPase, the action of ErITC on the overall Na ϩ /K ϩ -activated ATP hydrolysis as well as on partial activities of the enzyme was studied.
When the enzyme was incubated with increasing micromolar concentrations of ErITC at 37°C, Na ϩ /K ϩ -ATPase activity was lost as a time-and concentration-dependent process (Fig. 1A). The data were fitted with a two-site model (4), which revealed K i values of 0.66 and 0.78 M (Fig. 1B). Millimolar concentrations of ATP (0 -25 mM) protected Na ϩ /K ϩ -ATPase against the inactivation by 0.2 M ErITC (Fig. 2). Na ϩ -dependent phosphorylation of the catalytic ␣-subunit from the E 1 ATP site in Na ϩ /K ϩ -ATPase is a specific property of this site. To verify that ErITC interacts in fact with the E 1 ATP site, we blocked the E 2 ATP site with Co(NH 3 ) 4 PO 4 (4,12) and studied the effect of 2 M ErITC on the Na ϩ -dependent formation of a phospho intermediate. Table I shows that the velocity of inactivation of the Na ϩ -dependent autophosphorylation of a Co(NH 3 ) 4 PO 4treated and and -untreated control enzyme is the same. Hence, one of the two sites interacting with ErITC (Fig. 1B) is the E 1 ATP site.
Furthermore, to learn whether ErITC may also affect the partial activity of the E 2 ATP site, we blocked the activity of the E 1 ATP site by incubation with FITC first (8,36) and then studied the action of ErITC on the remaining activity of the K ϩ -activated p-nitrophenylphosphatase (Figs. 3 and 4), which is an enzymatic property of the E 2 conformation and the E 2 ATP site (4,8). It was inactivated by ErITC as well. The kinetics of the inactivation of K ϩ -activated p-nitrophenylphosphatase by ErITC gave a straight line in a reciprocal plot of inactivation velocity constant versus the ErITC concentration (41) (Fig. 3), indicating thereby the interaction of ErITC with a single site Fig. 3, inset). The modification of this site was prevented by 10 mM ATP (Fig. 4). Hence, ErITC labeled under these specific conditions a site with low affinity for ATP.
It has been shown formerly that the MgATP complex analog Co(NH 3 ) 4 ATP is a specific label of the E 2 ATP binding site (4,9). To certify that ErITC labels in fact the E 2 ATP site in a FITCpretreated Na ϩ /K ϩ -ATPase, we additionally blocked the free E 2 ATP site in a FITC-pretreated Na ϩ /K ϩ -ATPase (whose E 1 ATP site was blocked by FITC) by treatment with 1 mM Co(NH 3 ) 4 ATP (Fig. 5). No incorporation of erythrosin from ErITC into such a double-modified enzyme (whose E 1 ATP and E 2 ATP sites had been blocked) was seen. Only a control enzyme, whose E 2 ATP site was accessible (no pretreatment with Co(NH 3 ) 4 ATP), was modified by ErITC (Fig. 5). In conclusion, ErITC can modify both ATP sites (Fig. 1B), but after specific blockade of the E 1 ATP site, it is a modifier of the E 2 ATP site (Figs. [3][4][5]. Steady-state Fluorescence of FITC-and ErITC-labeled Na ϩ / K ϩ -ATPase-Unlike previous reports (25,34,36), enzyme labeling with isothiocyanates was performed for a longer time period and at pH 7.25 (see "Experimental Procedures"). Similar to the FITC labeling conditions at pH 9.0 (25,34,36), inactivation of Na ϩ /K ϩ -ATPase by FITC at pH 7.25 was protectable by ATP, and the resultant protein was sensitive to Na ϩ and K ϩ (18) (but another residue may be modified within the E 2 ATP site at pH 7.25 besides Lys 501 at pH 9 (26,27)). When the enzyme was investigated for any labeling by FITC outside the E 1 ATP site by the use of specific antifluorescein antibodies (46), the quenching of fluorescence intensity was found to be about 25%, indicating a relatively high specificity of labeling (data not shown). Hence, all labeling with ErITC and FITC was performed at pH 7.25. As expected, fluorescence emission spec-FIG. 1. Inactivation of the Na ؉ /K ؉ -ATPase activity by ErITC. A, Na ϩ /K ϩ -ATPase was inactivated with ErITC as follows. 1 unit of Na ϩ / K ϩ -ATPase was incubated in a total volume of 1 ml with 0 nM (f), 50 nM (Ⅺ), 100 nM (ࡗ), 250 nM (छ), 500 nM (q), and 1 mM (E) ErITC in 20 mM Tris, HCl (pH 7.25) and 15 mM NaCl at 37°C. The residual Na ϩ /K ϩ -ATPase activity was assayed by transferring aliquots of 50 l to the optical assay (39). B, the apparent velocity constants of inactivation were calculated as monoexponential decay and plotted against ErITC concentration. The K i values were calculated by a fit of a two-site binding hyperbola Y ϭ ((v max1 ϫ X)/(K i1 ϩ X) ϩ (v max2 ϫ X)/(K i2 ϩ X)).   tra of the FITC-and ErITC-labeled Na ϩ /K ϩ -ATPase were both independent of the pH during the labeling procedure. Covalent modification of Na ϩ /K ϩ -ATPase by ErITC led to a substantial increase in the quantum yield and lifetime of the excited state of erythrosin as compared with ErITC (without any shift of the emission spectrum) as a free label in water. The sites labeled by FITC and ErITC differed in their accessibility to iodide ions. Iodide quenched the fluorescence of ErITC-labeled Na ϩ /K ϩ -ATPase with a quenching constant of K q ϭ 0.4 M Ϫ1 , whereas the quenching constant of the FITC-labeled enzyme was K q ϭ 3.5 M Ϫ1 . These results are quite similar to those under the conditions of labeling at pH 9 (a 30-min incubation of 1 mg of protein/ml with 10 mM isothiocyanate in the dark and at room temperature in 50 mM Tris/HCl, 2 mM MgCl 2 ) (34).
FITC and ErITC have a similar structure and the same reactive chemical group. Hence, one may assume that Na ϩ /K ϩ -ATPase labeled by the two isothiocyanates may record K ϩ -and Na ϩ -induced conformational changes in the same way. However, this is not the case. Although FITC-labeled Na ϩ /K ϩ -ATPase showed (without using antifluorescein antibodies) a 12% decrease in its fluorescence intensity upon E 1 Ϫ E 2 transition, very little if any change occurred for the ErITC-labeled enzyme. Fluorescence response of FITC-treated enzyme to Na ϩ and K ϩ ions, respectively, was quite similar and insensitive to pH of labeling. Labeling of Na ϩ /K ϩ -ATPase by ErITC at pH 9 did not show specific inhibition of the K ϩ -activated phosphatase in FITC-pretreated enzyme. An E 1 Ϫ E 2 transition was seen when the effects of Na ϩ and K ϩ ions were studied in such an enzyme preparation, observing the FITC fluorescence (with or without ErITC bound to the E 2 ATP binding site (data not shown)).
Since the nature of the fluorophor is important for the ATP site specificity, we also tried to learn whether the change of the reactive group may affect the interaction with the E 2 ATP site. When cysteine-reactive erythrosin 5Ј-iodacetamide was used, a modification of the E 2 ATP site was also seen, as checked by the loss of the activity of K ϩ -activated p-nitrophenylphosphatase in a FITC-inactivated Na ϩ /K ϩ -ATPase. The apparent affinity of the drug, however, was 10 times lower, and about 20 mol of SH groups/mole of Na ϩ /K ϩ -ATPase were labeled. Hence, this substance is not useful for Förster energy transfer measurements. Interestingly, previous studies with 5-iodacetamidofluorescein revealed that it is bound outside the ATP binding site (47,48) at Cys 457 (49).
Distance between the High Affinity and Low Affinity ATP Binding Sites-As is evident from the experiment shown in Fig. 4, labeling of the high affinity E 1 ATP site by FITC and of the low affinity E 2 ATP site by ErITC should allow determination of the distance between both ATP binding sites. The donoracceptor pair FITC/ErITC is well known for its high overlap integral, its high quantum yield of FITC, and the high extinction coefficient of ErITC. Thus, this donor-acceptor pair is a suitable tool to study long distance interactions of enzyme subunit as well as short distances. Short distances would be indicated by a high Förster energy transfer, whereas low energy transfer would indicate long distances between the fluorophores. Therefore, Na ϩ /K ϩ -ATPase was labeled at a FITC concentration of 30 nmol/mg of protein. This procedure achieved a total inhibition of Na ϩ /K ϩ -ATPase activity (Fig. 4). Half of the FITC-labeled enzyme preparation was subsequently labeled with ErITC (6 nmol/mg of protein). The K ϩ -activated p-nitrophenylphosphatase was inhibited, thereby, by more than 99%. After removal of free fluorophore by 3-fold centrifugation after each labeling step, the steady-state anisotropy values were measured (r ϭ 0.27 for FITC, and r ϭ 0.23 for ErITC, respectively, at the doubly labeled sample), and the fluorescence lifetime of the covalently attached FITC was determined in a FITC/ErITC-doubly labeled and a control FITC-labeled enzyme. As is evident from Fig. 6, energy transfer from the fluorescein residue to the erythrosin residue attached to Na ϩ / K ϩ -ATPase was not very pronounced. The average lifetime of the FITC-excited state was ϭ 3.21 ns (Table II). A twocomponent lifetime decay was found to fit best to the experimental data obtained in the absence of ErITC as the energy acceptor. However, the fluorescence intensity decay became more heterogeneous in the presence of ErITC. In this case, only a three-component fit was adequate. The experimental data did not support more complex theoretical models of the fluorescence decay. Under these conditions, the average lifetime of the attached FITC decreased very slightly to only ϭ 2.88 ns (Table II), due to Förster resonance energy transfer. The tiny decrease by about 10.5% of the average lifetime of the excited state of fluorescein in the double-labeled FITC/ErITC enzyme (as compared with a control enzyme) indicates a very long distance between the FITC label attached to the E 1 ATP binding site and the ErITC attached to the E 2 ATP site. Some slightly elevated R 2 values in Table II could result as a consequence of differences in background fluorescence between the untreated sample, which was taken as a background correction, and the treated enzyme (50). Such minor differences can be hardly controlled and avoided.
Calculation of the distance between the E 1 ATP and E 2 ATP sites needs, besides knowledge of the efficiency of the Förster energy transfer, the molar ratio of the attached fluorophores per ␣-subunit. This ratio was determined fluorescently based on the increases in quantum yield of FITC and ErITC, respectively (see "Experimental Procedures"). Known concentrations of FITC in ethanol, ErITC in water, and of the enzyme allowed us to calculate the molar ratio from steady-state fluorescence data. The calculated molar ratio of total FITC/␣-subunit was 0.8. There was, however, a 25% unspecific labeling by FITC outside the E 1 ATP binding site in the sample, as was evident from the fluorescence quench by the addition of antifluorescein antibodies (34). By correction for this unspecifically attached FITC, a molar binding ratio of 0.6 mol of specifically bound FITC (within the ATP site)/mol of ␣-subunit was obtained. The molar ratio of ErITC/␣-subunit in an FITC-prelabeled enzyme was 0.48. Hence, 1 mol of ErITC and FITC were bound/2 mol of ␣-subunits under the above conditions. However, when ErITC labeling was performed in an enzyme preparation that had not been previously preincubated with FITC, a molar ratio of 1.3 mol ErITC/mol of ␣-subunit was obtained. In lifetime measurements, similar corrections of the molar donor to acceptor ratio using antifluorescein antibodies were performed (34,46). The corrected rate of energy transfer was thus, 0.108 ns Ϫ1 , and the calculated distance between the donor and the acceptor was 6.9 Ϯ 0.9 nm. These results were calculated from the energy transfer rate according to Amler et al. (34,51). Inclusion of 5% SDS into the assay system decreased the energy transfer be-tween FITC and ErITC to 3% of the original. Thus, subunit separation resulted in the disappearance of Förster energy transfer.
An alternative approach to determine the distance between the low and high affinity ATP binding sites on Na ϩ /K ϩ -ATPase was to measure the distance between FITC attached to the high affinity E 1 ATP binding site and Co(NH 3 ) 4 ATP bound to the E 2 ATP site (8,9). Since the absorption spectrum of Co(NH 3 ) 4 ATP overlaps well with the fluorescence emission spectrum of FITC, this was possible. We calculated a critical distance for this donor-acceptor pair as R 0 ϭ 4.0 nm. We have to take into account that Co(NH 3 ) 4 ATP does not bind covalently. Hence, the presence of some free Co(NH 3 ) 4 ATP or Co 2ϩ ions has to be considered. Since Co 2ϩ ions are known to be potent collisional fluorescence quenchers, the lifetime of the excited state of FITC has always been measured in the presence of 1 mM Co 2ϩ ions. The lifetime of FITC-labeled Na ϩ /K ϩ -ATPase in the presence of 1 mM CoCl 2 was found to be 2.3 ns, whereas after inactivation with 1 mM Co(NH 3 ) 4 ATP, it decreased to 2.0 ns (Table II). The data in both cases were collected in the presence of antifluorescein antibodies to avoid contribution of nonspecifically attached fluorophores (34,46). This decrease was due to Förster resonance energy transfer between FITC and Co(NH 3 ) 4 ATP bound to the E 2 ATP site. Clearly, the 12.2% resonance energy transfer efficiency obtained points to a remote molecular distance between the protein-bound FITC and Co(NH 3 ) 4 ATP. Taking into account the critical distance of the energy donor-acceptor pair, we calculated a distance between E 1 ATP and E 2 ATP sites of 6.0 Ϯ 0.9 nm from this experiment (Table II).
Distance Measurements between Ouabain Binding Sites-To verify that the calculation of the distance between the E 1 ATP and E 2 ATP binding sites represents the actual distance between two ␣-subunits, we tried to measure the distance between ouabain binding sites as well. It is generally accepted that each ␣-subunit has a molar ratio of ATP binding sites to phosphorylation sites to cardiac glycoside binding sites of 1:1:1 (1-3, 52). Therefore, the cardiac glycoside binding site of Na ϩ / K ϩ -ATPase was labeled first to 50% with AO under the conditions of backdoor phosphorylation and also additionally with the fluoresceinylated ouabain derivative, FEDO (8,42). Steady-state fluorescence measurements of Na ϩ /K ϩ -ATPase labeled with both ouabain derivatives revealed a binding ratio of AO to FEDO of 0.53 AO:0.55 FEDO per ␣-subunit, which are values within the theoretical error of 10% per label and result in a ratio of 1:1:2. This ratio was sufficient for distance measurements between two similar ouabain binding sites. On average, only half of the enzyme molecules had one labeled ␣-subunit with AO and the other one with FEDO. The rest of the functional units of the enzyme had both ␣-subunit labeled by either by FEDO or AO only.
AO in the ouabain binding site revealed two different lifetimes, which were an intrinsic property of the label. A longlived component of 10.7 ns and a short-lived component of 3.4 ns. The average lifetime was calculated to be 7.7 ns (Table II). The average lifetime of the excited state of AO decreased in the presence of FEDO to the value of 6.05 ns (Table II). In addition, the decay became more complicated, and a three component decay had to be used to fit the measured data. A calculation of the average distance between attached AO and FEDO from the average lifetime resulted in a molecular distance of 4.9 Ϯ 0.5 nm.

DISCUSSION
This study shows, consistent with the idea that the interaction of the E 1 ATP site is not specific for a distinct fluorophore of the eosin type (53), that in addition to eosin (32, 33) and FIG. 6. Energy transfer between FITC and ErITC on Na ؉ /K ؉ -ATPase. Trace 1, fluorescence intensity decay of fluorescein of FITClabeled Na ϩ /K ϩ -ATPase. Trace 2, fluorescence intensity decay of fluorescein of doubly FITC/ErITC-labeled Na ϩ /K ϩ -ATPase. The excitation was set at 363 nm, and the fluorescence emission was taken at 520 nm. Under the figure are the residual plots of the upper trace (FITC-labeled Na ϩ /K ϩ -ATPase) and the lower trace (FITC/ErITC-labeled Na ϩ /K ϩ -ATPase) for the decay analysis presented in Table II. fluorescein (36), erythrosin isothiocyanate also binds to the E 1 ATP site ( Table I). The modification of this site is clearly evident after the blockade of the E 2 ATP site by Co(NH 3 ) 4 PO 4 (Table I). ErITC inactivated the Na ϩ -dependent phosphorylation of the ␣-subunit from the E 1 ATP site with the same rate constant as the overall Na ϩ /K ϩ -ATPase activity ( Table I). The inactivation is prohibited by an excess of ATP (Fig. 3). This overall activity is, however, affected by ErITC with two different inactivation constants of 0.66 M and 0.78 M (Fig. 1), indicating a possible interaction with a second ATP site. In fact, after specific blockade of the E 1 ATP site by FITC, the proteinreactive ErITC inactivated the K ϩ -activated p-nitrophenylphosphatase, a partial activity of the E 2 ATP site, with a K i of 0.74 M (Fig. 3). Also, labeling of this site is prevented by an additional blockade of the E 2 ATP site by Co(NH 3 ) 4 ATP (Fig. 5). Hence, after specific protection of the E 1 ATP site by FITC (8), ErITC covalently labels the E 2 ATP site.
FITC and ErITC are very similar in their chemical structures, except for the four bulky iodides in the erythrosin molecule. It is surprising, therefore, that FITC is a label of the E 1 ATP site and not of the E 2 ATP site, whereas ErITC binds to and modifies both ATP sites. The idea has been put forward more recently that the E 2 ATP site has a broader binding pocket than the E 1 ATP site, since after modification of the E 2 ATP site with 7-chloro-4-nitrobenzo-2-oxa-1,3-diazole chloride, the latter site was easily accessible to a number of fluorescence quenchers (54). This explanation, however, cannot be applied to the differential labeling of ATP sites by ErITC and FITC, since the more bulky ErITC should not bind to the E 1 ATP site, which is what actually happens (Table I). Due to the close similarity of both fluorophores, the possibility exists that the electron system of the aromatic ring system contributes to the interaction as well. It is certainly less displaced in the case of FITC as compared with ErITC.
It has been known for a long while that labeling of the E 1 ATP site at Lys 501 by FITC records the interaction of Na ϩ /K ϩ -ATPase with Na ϩ and K ϩ ions (36). Since ErITC is labeling the E 1 ATP site as well (Table I), one might assume that addition of K ϩ ions to a ErITC-labeled Na ϩ /K ϩ -ATPase should lead to a decrease of fluorescence as well as it does with a FITC-labeled enzyme. This does not occur, however, as very careful studies with such an enzyme have revealed (34). It has been shown recently that labeling of the E 2 ATP site by fluorescent ethylanthroyl-Co(NH 3 ) 4 ATP freezes the conformational flexibility of the E 1 ATP site of Na ϩ /K ϩ -ATPase, as studied by the fluorescence change of the FITC-labeled enzyme upon the addition of K ϩ ions (18). Therefore, to exclude the possibility that labeling of the E 1 ATP and E 2 ATP sites by ErITC may lead to a loss of the conformational flexibility in the presence of the transport substrates K ϩ and Na ϩ as well, additional studies on an enzyme were carried out whose E 1 ATP site was labeled by FITC followed subsequently by the covalent modification of the E 2 ATP site with ErITC. When the effect of K ϩ ions was studied in such an enzyme preparation, an E 1 -E 2 transition was seen when observing the FITC fluorescence. Hence, the conformational flexibility of the E 1 ATP site is lost only when ATP and not a pseudo-ATP analog is bound to the E 2 ATP site. These studies, moreover, seem to indicate that FITC and ErITC may not modify exactly the same site, even though they are modifying the E 1 ATP site. Consistent with the idea of a different microenvironment of the ATP sites labeled by FITC or ErITC is the observation that the quenching constants of iodide differ for FITC-(K q ϭ 0.4 M Ϫ1 ) and ErITC-modified (K q ϭ 3.5 M Ϫ1 ) enzymes. A definitive answer to this question can only be given by the demonstration that ErITC labels Lys 501 or another amino acid. Experiments to answer this question have been started.
Although, a definitive answer as to which amino acid might be modified by ErITC is lacking, there is no doubt as to the conclusion that ErITC labels ATP sites (Figs. 3 and 4) and that, in an FITC-treated Na ϩ /K ϩ -ATPase, ErITC interacts with the E 2 ATP site (Figs. 4 and 5). Labeling of Na ϩ /K ϩ -ATPase by ErITC is suppressed by ATP (Figs. 3-5) (34) as is the labeling of the enzyme by FITC (36). There were 0.48 mol of ErITC/␣subunit incorporated into an enzyme whose E 1 ATP site had been modified already by FITC. Moreover, the molar ratio of specifically incorporated FITC/␣-subunit was 0.6. Hence, 1 mol of ErITC and FITC were bound/2 mol of ␣-subunits when both ATP sites were modified by different fluorescent isothiocyanates. This stoichiometry is consistent with the previous assumption that the two interacting ATP sites (4, 5) reside on different ␣-subunits (22). Consistent with this assumption are also the data obtained by Förster energy transfer measurements (Fig. 6, Table II). Fluorescein and erythrosin show a high overlap integral. Hence a close location of the two interacting ATP sites on the same ␣-subunit, as postulated by Ward and Cavieres (13), should lead to a high fluorescence quench. This is apparently not the case (Fig. 6, Table II). The data obtained  4 ATP as acceptor and AO as donor with FEDO as acceptor Na ϩ /K ϩ -ATPase (6 units, 300 g) was inactivated overnight at 37°C with 10 M FITC. This modified enzyme was additionally exposed to 2 M ErITC or 1 mM Co(NH 3 ) 4 ATP. The different washed and modified protein fractions (final concentration 1.8 mg/ml) were resuspended in 0.3 ml of 20 mM Tris/HCl buffer (pH 7.25), and fluorescence lifetimes were measured. In the case of energy transfer measurements between FITC and Co(NH 3 ) 4 ATP, the donor fluorescence lifetime was measured in the presence of 1 mM CoCl 2 . To measure the distances between ouabain binding sites, 6 units of Na ϩ /K ϩ -ATPase (300 g) were incubated in a total volume of 1 ml under backdoor conditions [8,42] with 150 nM AO and, in the case of present acceptor, with 150 nM AO and 300 nM FEDO (control background, no AO or FEDO). The washed enzyme preparations were resuspended in 0.3 ml of 20 mM Tris/HCl buffer (pH 7.25) (final concentration 1.8 mg/ml), and fluorescence lifetimes were measured. For details see "Experimental Procedures." i values (1-3) are the calculated lifetime components; f i (1-3) is the fractional intensity; E is the energy transfer efficiency (%); R 2 is the goodness of fit [63]. The lifetime was calculated by total ϭ ͐f i 2 i d, as defined by Amler et al. [34].  (Table II) show that the distance between FITC-and ErITClabeled ATP sites is 6.9 Ϯ 0.9 nm. Because of different conditions of labeling in previous experiments (pH 9 versus pH 7.25, this study), the distance between FITC and ErITC (Table II) was somewhat higher than formerly published (r ϭ 5.6 nm (34)). But as stressed previously, such long distances can only be explained by assuming a (␣␤)2 diprotomeric structure, since the diameter of the ␣-subunit is 4.5 nm (34,55). Independent measurements on the distance between the E 1 ATP site labeled by FITC and the E 2 ATP site labeled by Co(NH 3 ) 4 ATP gave a value of 6.0 Ϯ 0.9 nm (Table II). Both distances agree favorably well within the S.E. and reveal a total distance of 6.45 Ϯ 0.64 nm (according to the law of propagation of errors) between both ATP-sites. In support of the conclusion that the E 1 ATP and E 2 ATP binding sites reside on different ␣-subunits is also the observation that solubilization of an FITC/ErITC doubly labeled enzyme showed no energy transfer any more. Hence, all data do not support the postulate that the two interacting ATP sites reside on the same ␣-subunit (13). Consistent with an (␣␤) 2 diprotomeric structure is also the finding that the distance between the cardiac glycoside receptor sites labeled with either AO or FEDO is 4.9 Ϯ 0.5 nm (Table II). An (␣␤) 2 diprotomeric structure is also consistent with the longer distance (3.2 nm) between ErITC and 5-iodacetamide than between TNP-ATP and 5-iodacetamide (2.4 nm (34) or 2.9 Ϯ 0.6 nm (56)) ( Fig. 7). It had been found previously by energy transfer measurements that ATP and cardiac glycoside binding sites are 7.2 nm apart (57) and that the cytosolic part carrying the ATP site sits 3 nm above the inner side of the plasma membrane (57,58). Additionally, energy transfer measurements between the AOcontaining ouabain binding site and the lucifer yellow-modified ␤-subunit (51,59) give an overall arrangement of distances (Fig. 7). The large difference between AO and lucifer yellow attached to ␤-subunit was found in the presence and the absence, respectively, of Mg 2ϩ ions, which could reflect the interaction of both subunits in the presence of Mg 2ϩ . To complete this survey, the distance between N-(p-(2-benzimidazol)phenyl)maleimide bound to Cys 964 and FITC bound to Lys 501 was observed to be 3.6 nm (60). The distance of N-(p-(2-benzimidazol)phenyl)maleimide to AO in the ouabain binding pocket is about 3.9 nm (60). In summary, the data from this study and previous publications presented in the overview (Fig. 7) support in the context of kinetic data on ATP hydrolysis (4, 5, 7, 8, 11, 21, 22, 61, 62) the concept that E 1 ATP and E 2 ATP sites reside on adjacent and interacting catalytic ␣-subunits and an (␣␤)2 diprotomeric model of Na ϩ /K ϩ -ATPase. They are inconsistent with a model containing two separate and interacting ATP sites/␣-subunit.