Binding of 2 (cid:42) (3 (cid:42) )- O -(2,4,6-Trinitrophenyl)ADP to Soluble (cid:97)(cid:98) Protomers of Na,K-ATPase Modified with Fluorescein Isothiocyanate EVIDENCE FOR TWO DISTINCT NUCLEOTIDE SITES*

The overall reaction of well-defined solubilized pro- tomers of Na,K-ATPase (one (cid:97) plus one (cid:98) subunit) re-tains the dual ATP dependence observed with the mem- brane-bound enzyme, with distinctive ATP effects in the submicromolar and submillimolar ranges (Ward, D. G., and Cavieres, J. D. (1993) Proc. Natl. Acad. Sci. U. S. A . 90, 5332–5336). We have now found that the K (cid:49) -phospha-tase activity of the (cid:97)(cid:98) protomers is still inhibited by 2 (cid:42) (3 (cid:42) )- O -(2,4,6-trinitrophenyl)adenosine 5 (cid:42) -diphosphate (TNP-ADP). What is most significant is that the TNP- ADP effect can be observed clearly with protomeric enzyme whose high affinity ATP site has been blocked covalently with fluorescein isothiocyanate. We conclude that nucleotides can bind at two discrete sites in each protomeric unit of Na,K-ATPase. The sodium pump or Na,K-ATPase 1 uses the energy of ATP hydrolysis to power the active transport of Na (cid:49) and K (cid:49) ions across the plasma membrane (1). This integral membrane enzyme consists of (cid:97) and (cid:98) subunits. The (cid:97) chain

The sodium pump or Na,K-ATPase 1 uses the energy of ATP hydrolysis to power the active transport of Na ϩ and K ϩ ions across the plasma membrane (1). This integral membrane enzyme consists of ␣ and ␤ subunits. The ␣ chain presents an ATP-binding site, a phosphorylation site, and a ouabain-binding site. The ␤ subunit is a glycoprotein of unclear function, found in equimolar proportions with the ␣ chain. We recently reported (2) that solubilized ␣␤ protomers of highly purified sodium pump responded to ATP with high and low affinity effects, and also that the complex behavior did not result from protomer association to form soluble (␣␤) 2 dimers. Those experiments demonstrated that a complex dependence on nucleotides was intrinsic to the ␣␤ protomer. They could not decide, however, whether this arose from the presence of more than one ATP site per ␣␤ protomer or from a single site whose function and affinity changed round the catalytic cycle. That is the question addressed in this paper.
It has long been recognized that ATP has two types of activatory effect during the sodium pump cycle (3). Na,K-ATPase becomes phosphorylated by ATP during the course of the reac-tion, and the small high affinity activation (K 0.5(high) Ͻ 1 M) correlates well with the ATP requirement for phosphorylation of the E 1 form of the enzyme (4). The considerable low affinity stimulation of the overall cycle (K 0.5(low) approximately 150 -300 M) seems to result from accelerating a rate-limiting step in the E 2 form of the dephosphoenzyme (4,5). ADP (6), nonphosphorylating ATP analogues (7), and acyl coenzymes A (8) can replace ATP in this low affinity effect. For these reasons, this is regarded as a regulatory nucleotide effect, in contrast with the "catalytic" (high affinity) ATP action that results in enzyme phosphorylation in the presence of Na ϩ ions. Affinity probes have so far returned what seems to amount to a single, high affinity ATP site on the ␣ chain. The purine ring subsite has been mapped with 2-azido-ATP and 8-azido-ATP (9, 10), which label peptides identifying Gly 502 and Lys 480 , respectively, as the anchoring points. This is a region which had earlier been found to be labeled by FITC, after binding covalently to Lys 501 (11,12). Probes for the 5Ј-triphosphate moiety of ATP, on the other hand, have identified a sequence between Asp 710 and Lys 719 on the ␣ chain (13,14).
The structural findings have polarized the rationalization of the two ATP actions toward two basic models: (i) a unique ATPbinding site whose affinity and function change round the reaction cycle and (ii) a membrane (␣␤) 2 dimer, with one ATP site per ␣ subunit, where the interaction of ␣␤ units leads to a degree of half-of-the-sites reactivity (15) during turnover. Evidence for the first arises from studies with the vanadate-inhibited enzyme (16), from experiments with TNP-ATP, a high-affinity ATP analogue (17), and, at least superficially, from the enzymatic competence of the ␣␤ protomer in solution (2). There are a number of observations, however, that cannot be explained by a single site (18), and it now seems increasingly likely that TNP-ADP can access more than one class of binding sites (19).
A third possibility, i.e. the idea of two distinct nucleotide sites per ␣␤ unit, has not had much currency because of the lack of supporting structural data, the controversy above, and a certain sense of economy. This was, however, a viable alternative hypothesis in the case of the soluble ␣␤ protomer (2). We have approached the question by examining the behavior of soluble protomers whose high affinity ATP site has been irreversibly blocked by FITC. The covalently bound fluorescein suppresses phosphorylation by ATP and obliterates the overall pump cycle, but does not affect backwards phosphorylation by inorganic phosphate, or the K ϩ -phosphatase activity of the enzyme (19 -21). A realistic setting was that the K ϩ -phosphatase substrate binds at a place other than the FITC-blocked site.
The K ϩ -phosphatase activity is the ability of the sodium pump to hydrolyze substrates like pNPP, carbamyl phosphate, and 3-OMFP, hydrolyses which do not support cation fluxes. It is an E 2 -type of activity on account of not only its K ϩ require-ment but the effect of nucleotides on the reaction also. In fact, in the absence of Na ϩ ions, the K ϩ -phosphatase activity is inhibited, with low affinity, by ATP and other nucleotides (22,23), including analogues like TNP-ADP (19). We have used this effect to probe for a low affinity nucleotide site in FITC-modified, soluble ␣␤ protomers and their parallel FITC-free controls. We found that TNP-ADP had a clear inhibitory effect, as would only be expected if one high affinity and one low affinity nucleotide site co-existed on the same protomeric unit. A preliminary report has been published (24).

EXPERIMENTAL PROCEDURES
Enzyme Purification-Sodium pump was purified from pig kidney outer medulla with the zonal-rotor procedure of Jørgensen (25). The specific Na,K-ATPase activity ranged from 20 to 35 mol⅐min Ϫ1 ⅐mg Ϫ1 at 37°C.
FITC Treatment-A sample of purified membrane-bound sodium pump was incubated at 20°C, in the dark, either (i) at 2.4 mg of protein/ml and with 20 M FITC for periods up to 3 h, or (ii) at 1 mg of protein/ml and with 50 M FITC for 30 min. This was done in a medium containing 100 mM NaCl, 50 mM Tris/Cl Ϫ (pH 9.2), 5 mM EDTA and 0.5% dimethyl sulfoxide. A parallel control sample was handled similarly, omitting the FITC. The membranes were chilled and spun out at 356,000 ϫ g for 15 min and at 4°C in a Beckman TL-100 ultracentrifuge. They were washed, and resuspended, with a solution ("2 ϫ S") containing 300 mM KCl, 20 mM TES/K ϩ (pH 7.5), 2 mM EDTA, and 2 mM dithiothreitol. Ultracentrifugation and wash were omitted in the experiments of Fig. 1 (see below, under Enzyme Activities). Solubilization to ␣␤ Protomers-This was carried out essentially as described (2,26), by mixing equal volumes of FITC-treated (or control) membranes in 2 ϫ S medium and aqueous C 12 E 8 , at a detergent:protein mass ratio of 2.9:1. Insoluble material was removed by ultracentrifugation for 15 min at 356,000 ϫ g. When assaying the enzymatic activities of soluble protomers, C 12 E 8 was added to all media (2) at a concentration of 50 M (critical micelle concentration).
Enzyme Activities-These were measured in washed membranebound enzyme, or in soluble protomers, as linear 6-point time courses at 20°C. Na,K-ATPase activity was determined (2) over a 2.5-min period, in 50 l and at 1 mM [␥-32 P]ATP, 2 mM MgCl 2 , 1 mM EDTA, 118 mM NaCl, 32 mM KCl, 10 mM TES/Na ϩ (pH 7.5), and 1 mM dithiothreitol. The reactions were stopped by freezing on a dry ice/ethanol bath. K ϩ -phosphatase activity was routinely measured from six time points over 2.5 min in 50 l of 1 ϫ S medium (pH 7.5) supplemented with 6 mM MgCl 2 plus pNPP (Tris salt). The release of p-nitrophenol was measured spectrophotometrically at 410 nm, after stopping the reactions with 0.3 ml of 1 M NaOH. In experiments in the presence of TNP-ADP or TNP-ATP, p-nitrophenol calibration curves were constructed for each of the TNP-nucleotide concentrations. All enzyme activities and errors were obtained from least squares linear fitting of the time courses. Most K ϩ -phosphatase assays were done at several pNPP concentrations, ranging from 1.5 to 20 mM, in order to estimate V max by nonlinear regression (Sigmaplot package, Corte Madera, CA). In the experiments in Fig. 1, the incubation with (or without) FITC was arrested at the times indicated by diluting (i) a 10-l sample into 990 l of chilled medium containing (mM) NaCl 130, KCl 20, histidine (pH 7.2) 25, [␥-32 P]ATP 1, and MgCl 2 1 (Fig. 1A) or (ii) 10 l each into 700 l of chilled assay media suitable for K ϩ -phosphatase or Na,K-ATPase activity determinations. Similar reaction media were used when 3-OMFP replaced pNPP as K ϩ -phosphatase substrate; product release was then measured against 3-O-methylfluorescein calibration curves, from the Solid symbols, an experiment, under similar inactivation conditions, to measure the K ϩ -phosphatase activity (solid circles) as well as ATP hydrolysis (solid squares). The substrate concentrations were 1 mM [␥-32 P]ATP for Na,K-ATPase assays and 40 mM pNPP (Tris) for the K ϩ -phosphatase assays.

FIG. 2. Block of individual sodium pump protomers by FITC.
K ϩ -activated phosphatase activity (V max ), Na,K-ATPase activity, and protein concentration were measured in FITC-treated and control sodium pump, in their original membrane-bound state (MB, stippled bars) and in the ␣␤ protomers arising after solubilizing with C 12 E 8 (␣␤, solid bars). The specific activities of the FITC-treated enzyme are presented as percent of the specific activities of the respective control enzyme. Open bar, rescue of Na,K-ATPase specific activity to be expected after solubilization, had FITC inactivated only one-half of a functional (␣␤) 2 membrane dimer. At 20°C, the control K ϩ -phosphatase specific activities (mol⅐min Ϫ1 ⅐mg Ϫ1 ) of the membrane-bound and solubilized enzymes were 3.41 Ϯ 0.08 and 1.28 Ϯ 0.02, and the control Na,K-ATPase specific activities, 2.68 Ϯ 0.11 and 1.54 Ϯ 0.02, respectively. The initial treatment was done for 3 h, at 2.4 mg of protein/ml, and with or without 30 M FITC, as detailed under "Experimental Procedures." absorbance increase at the 475 nm peak of the product.
Analytical Ultracentrifugation-Sedimentation velocity analyses of the solubilized Na,K-ATPase were done at 40,000 rpm and 20°C on a Beckman Optima XL-A analytical ultracentrifuge, using double-sectored centerpieces. Sample and reference cells were radially scanned at 280 nm every 20 min. Sedimentation coefficients were calculated from linear plots of the logarithm of the radial position of the inflexion points versus time and corrected (28) to s 20,w .
Materials-TNP-ADP and TNP-ATP (disodium salts) were purchased from Molecular Probes; ATP (disodium salt) was from Boehringer; ATP (Tris salt), pNPP (Tris salt), 3-OMFP, and FITC (obtained as isomer I or as the mixed isomers) were from Sigma. C 12 E 8 was from Nikko Chemicals, dimethyl sulfoxide from Aldrich, and [␥-32 P]ATP from DuPont NEN. All other reagents were of the highest purity available.

RESULTS
Our results for the inactivation of the Na,K-ATPase activity of the sodium pump by FITC, the protective effect of ATP (20,29), and the survival of the K ϩ -phosphatase activity (20,21) are shown in Fig. 1, for illustration purposes. We could also confirm that the K 0.5 for pNPP increases (approximately 4-fold in our hands) following irreversible FITC binding. The maximal K ϩ -phosphatase rate remained high, at about 80% of the control activity (see Table II and legend to Fig. 4). In the experiments of Figs. 2 and 3, therefore, the K ϩ -phosphatase activities used in the calculations were the fitted V max values (estimated from determinations at 6 or more pNPP concentrations).
In order to test for independent high affinity and low affinity ATP sites, we wished to find out whether the K ϩ -phosphatase activity of well-defined ␣␤ protomers was inhibited by nucleotides when the high-affinity ATP site had been blocked by the modified fluorescein. The limited thermal stability of the solubilized protomers, even at 20°C (30), thwarted attempts at treating the enzyme with FITC at alkaline pH after detergent solubilization; lower pH values would only prolong the incubation period. The alternative was to solubilize the FITC-treated membrane enzyme. It has been reported that C 12 E 8 solubilization of the FITC-modified sodium pump releases particles (30) whose sedimentation coefficient (s 20,w ) is 6.9 S, i.e. as expected for the ␣␤ protomer (2,26). The results presented in Table I average to an identical s 20,w value and confirm that, in our hands also, solubilization of the FITC-treated enzyme leads to ␣␤ protomeric particles.
Before opting for the approach of solubilizing after FITC treatment, however, certainty was needed that the high affinity ATP site of every ␣␤ protomeric unit was occupied by FITC when complete inactivation of the overall sodium pump reaction had been achieved. The degree of confidence was not as high as desirable for the present purpose (see "Discussion"), and the endurance of the K ϩ -phosphatase activity in the FITCinactivated membrane enzyme seemed open, therefore, to two alternative interpretations. One possible explanation was that, in each ␣␤ protomeric unit, there was a site that bound the phosphatase substrate that was different from the high affinity ATP site (which was blocked by FITC). In this case, solubilization of the membrane-bound, FITC-modified sodium pump with C 12 E 8 would release ␣␤ protomers whose Na,K-ATPase specific activity (relative to the FITC-free control enzyme) was no more, and no less, than the (relative) Na,K-ATPase specific activity of the parent membrane-bound enzyme. The alternative possibility was that the high affinity ATP site and the pNPP-binding site were topographically one and the same, the different catalytic properties arising from different enzyme conformations in the presence of Na ϩ or K ϩ ions (31). If the membrane-bound sodium pump behaved as a dimer of interacting ␣␤ protomers in these conditions, one could hypothesize that only one of the protomers in the pair bound FITC covalently at its single site. Although the FITC-blocked protomer should still be able to adopt E 1 and E 2 conformations reversibly (20), the unliganded protomer in the dimer might become locked in an E 2 conformation, because of subunit interactions. In that case, its K ϩ -phosphatase activity would be spared but ATP hydrolysis, which requires cyclical E 1 -E 2 transitions, would be prevented just as in the FITC-blocked protomer. As solubilization with C 12 E 8 , in our conditions, dissociates any (␣␤) 2 dimers (2, 26), the good protomer should now be unencumbered by its FITC-blocked neighbor, and its Na,K-ATPase activity would be restored. Accordingly, the specific Na,K-ATPase activity of the soluble protomers arising from the FITC-modified enzyme should be higher than that of the parent membrane-bound enzyme, by a magnitude equal to half the loss observed in the latter. Fig. 2 shows the result of the experiment to decide between the alternative hypotheses. This was conducted with side-byside FITC-free control enzyme samples. Na,K-ATPase and K ϩphosphatase activities, as well as protein concentration, were determined in both inactivated and control samples, before and In agreement with previous observations, and in spite of the very low level of Na,K-ATPase activity left over, the K ϩphosphatase activity remains at a high 80% in both the membrane-bound and solubilized enzymes. The crucial feature, however, is that the percent Na,K-ATPase activity left in the soluble protomers does not differ from the percent Na,K-ATPase activity left in the parent membrane-bound enzyme. In the case of the "one-site-plus-dimer" hypothesis, one should have expected that the Na,K-ATPase increase to 51.6 Ϯ 6.5% of the controls, after solubilization (clear plus filled bars on right). The experiment was repeated at several levels of inactivation of the Na,K-ATPase activity, and the result is shown in Fig. 3. It is apparent that, despite some dispersion in the data, the results are firmly anchored on the 1:1 correlation line expected for the "two-sites-per-protomer" hypothesis. Table II shows that the protomeric FITC-enzyme can efficiently utilize the bulky 3-OMFP as a K ϩ -phosphatase substrate; this makes it quite unlikely that the phosphatase-binding site be but a small region of a single ATP-binding site. In order to test the possibility that high ATP concentrations could overcome the FITC block of the overall reaction, the Na,K-ATPase activities of FITC-modified and control enzymes were measured in one experiment at 2 and 20 mM [␥-32 P]ATP. The specific activities of the FITC-modified enzyme (relative to the control enzyme) were 0.9 Ϯ 0.1% at 2 mM ATP and 1.1 Ϯ 0.1% at 20 mM ATP.
It being clear that FITC can block an ATP site in every protomeric unit of the membrane-bound enzyme, we tested the effect of nucleotides on the K ϩ -phosphatase activity of the soluble protomers. Experiments using ATP (Tris salt) to 5 mM (not shown) produced clear effects with the control membranebound and solubilized enzymes, but its effect on the FITCmodified enzyme seemed to be obscured by a much reduced affinity (cf. Refs. 19 and 21). Lower pNPP or higher ATP concentrations imposed unacceptably high spectrophotometric errors or ionic strength compensations, respectively, and we resorted to using TNP-ADP, which has proved useful in experiments with the membrane-bound FITC-modified enzyme (19). The result of one of three similar experiments is presented in Fig. 4, as Dixon plots, for the soluble protomers. Here, again, the affinity is very much reduced in the FITC-treated protomers. K i works out as 65 Ϯ 8 M, compared with the native protomers at 0.084 Ϯ 0.003 M, but it is evident that TNP-ADP inhibits the K ϩ -phosphatase activity competitively well beyond what could be expected if it merely suppressed the 10% of the sodium pump that escaped FITC modification (arrows). Dixon plots of the K ϩ -phosphatase activity of the parent membranebound FITC-modified enzyme (not shown) also reflected the decreased TNP-ADP affinity, as has been observed earlier (19), with K i values of 42 Ϯ 9 M and 0.52 Ϯ 0.05 M for the FITC-treated and control enzymes, respectively. TNP-ATP was also effective with the FITC-modified membrane enzyme, but the estimated inhibition constant was 4-fold that for TNP-ADP, and it was thereby not used with the soluble protomers. Picrate (2,4,6-trinitrophenolate) at 500 M inhibited the K ϩ -phosphatase activity of the FITC-modified enzyme by 13% in conditions that 250 M TNP-ADP caused a decrease of over 70%. It seems safe to conclude, therefore, that the bulk of the TNP-ADP inhibition does not represent nonspecific effects of the trinitrophenyl group. DISCUSSION Our test for a discrete low affinity nucleotide site in Na,K-ATPase is based on searching for a known E 2 effect of nucleotides on ␣␤ protomers with a blocked high affinity ATP site. We chose FITC as the blocking agent and the K ϩ -phosphatase activity as the E 2 function. The covalent FITC binding to Lys 501 suppresses phosphorylation by ATP and the overall pump reaction, but does not prevent backwards phosphorylation by inorganic phosphate or abolish the K ϩ -phosphatase activity of the enzyme. Obviously, the possibility had to be excluded that any observable low affinity nucleotide effects could arise from binding at a unique nucleotide site that was left vacant in a neighboring and interacting subunit.
It has been reported that, at 100% inactivation of the Na,K-ATPase activity, the protein mass that can incorporate 1 mol of FITC can otherwise bind 1 mol of ouabain (32) or form 1 mol of acid-stable phosphoenzyme (33). Our difficulty with this approach was that, with the best preparations of native enzyme, it had not been possible to achieve acid-stable phosphorylation or ouabain binding levels of 1 mol/mol protomeric unit (6.9 nmol/mg of protein) using the same standard phosphorylation or ouabain binding assays, and that this did not seem just a matter of protein assay method (34). For instance, the standard ouabain/phosphorylation ratio is subject to some uncertainty (35,36), and it is now apparent that, depending on the experimental conditions, measurements of acid-stable phosphorylation can grossly underestimate the number of active sodium pump units (37,38). In addition, it has been reported that FITC can also bind to Lys 766 on the ␣ chain, which is not necessarily within the high affinity ATP-binding pocket (39). Any one, or a combination, of these complications could cause an overestimate of the number of functional high affinity ATP sites that are occupied by FITC. This possibility seemed the more plausible when considering the evidence that the membrane-bound Na,K-ATPase behaved, at least in some conditions, as a functional dimer of ␣␤ protomers (7, 18, 40 -43).
The concerns above prompted the experiments shown in Figs. 2 and 3, which exclude the possibility that only one-half of a putative membrane dimer could bind FITC. If all active ␣␤ protomers in the membrane-bound enzyme are potential targets for the binding, the corollary would be that the K ϩ -phosphatase substrate site should be different from the blocked high affinity ATP site. It would follow also that the competitive TNP-ADP inhibition of the K ϩ -phosphatase activity of FITCmodified soluble protomers (Fig. 4) should result from TNP-ADP binding at a location different from the blocked site, probably at the pNPP-binding site. We have confirmed that, in the membrane-bound enzyme, FITC treatment decreases the affinities of pNPP, ATP, and TNP-ADP for their E 2 effects (19,21) and find that this effect persists after solubilization to ␣␤ protomers. In the control enzyme, this decreases K i for TNP-ADP from 0.52 M (cf. Ref. 17) to 0.084 M, which agrees with the 8-fold decrease of K 0.5 observed for the low affinity ATP activation of the Na,K-ATPase activity upon solubilization (2). This downward shift did not occur with the FITC-modified enzyme, as K i changed from 42 M in the membrane-bound enzyme (compare with 35 M in Ref. 19) to 65 M after solubilization. The possibility should be considered, therefore, that all that FITC does is greatly lowering the affinity of two states of a single ATP site. If that were so, the FITC block of the Na,K-ATPase activity should be relieved by increasing the ATP concentration. This did not occur at 8 mM ATP when using the coupled enzyme spectrophotometric assay (19), and we now find that it does not happen at 20 mM [␥-32 P]ATP either. Had FITC lowered both high and low affinities by, say, a thousandfold (so that K 0.5(low) Ϸ 200 mM), we would have easily noticed an increase from 1% (at 2 mM) to 9% (at 20 mM) of the control Na,K-ATPase activity. By far, the most likely explanation is that ATP is utterly prevented from accessing the catalytic pocket by the bound fluorescein which also induces a distortion in the low-affinity nucleotide/K ϩ -phosphatase site. The block of the catalytic ATP site probably results from the restricted rotational motion of the bound fluorescein, as observed even through E 1 -E 2 conformational changes (44).
It is conceivable that the catalytic ATP site and the phosphatase site could share binding groups, perhaps their phosphate subsites. At any rate, such overlap must be such that access be still allowed to the fluorescein moiety of 3-OMFP when another fluorescein blocks the catalytic ATP site of the FITC-enzyme (Table II). On the other hand, it seems probable that the second nucleotide site be the same as the phosphatase site, a question that can be explored in more elaborate experiments. Fully competitive inhibition (same site) of the K ϩphosphatase by nucleotides should lead to linear Dixon plots whereas partially competitive (allosteric) inhibition should show as hyperbolae (45). The plots in Fig. 4 look convincingly linear, but higher inhibitor and lower substrate concentrations might lead to a curvature (46). Downward-bent Dixon plots do appear when using ATP as the inhibitor with the membranebound sodium pump (21), and we have confirmed this, also with the ␣␤ protomers. Caution is needed in their interpretation, though, as in a partially modified enzyme the difference in kinetic parameters between FITC-inactivated and outlasting enzyme populations could lead to an artifactual downward curvature. For instance, had the FITC-modified enzyme been impervious to TNP-ADP inhibition, the plots in Fig. 4 would have bent down and become parallel to the abscissa, at maximal ordinate levels given by the arrows in panel B (and at much lower values of the abscissa, compare with panel A).
The test of Figs. 2 and 3 does not contradict the possibility that the membrane-bound sodium pump behave, at least in some conditions, as a dimer of ␣␤ and protomers. What it proves is the absence of half-of-the-sites reactivity toward FITC. However, some of the evidence for a dimeric Na,K-ATPase may need reinterpreting in consideration of a second nucleotide site per protomeric unit, and the added element of complexity should be taken into account. The functional relationships between the two nucleotide sites, and their behavior during the reaction cycle, remain to be explored.