Selective Assembly of Na,K-ATPase α2β2 Heterodimers in the Heart

The Na,K-ATPase α2 subunit plays a key role in cardiac muscle contraction by regulating intracellular Ca2+, whereas α1 has a more conventional role of maintaining ion homeostasis. The β subunit differentially regulates maturation, trafficking, and activity of α-β heterodimers. It is not known whether the distinct role of α2 in the heart is related to selective assembly with a particular one of the three β isoforms. We show here by immunofluorescence and co-immunoprecipitation that α2 is preferentially expressed with β2 in T-tubules of cardiac myocytes, forming α2β2 heterodimers. We have expressed human α1β1, α2β1, α2β2, and α2β3 in Pichia pastoris, purified the complexes, and compared their functional properties. α2β2 and α2β3 differ significantly from both α2β1 and α1β1 in having a higher K0.5K+ and lower K0.5Na+ for activating Na,K-ATPase. These features are the result of a large reduction in binding affinity for extracellular K+ and shift of the E1P-E2P conformational equilibrium toward E1P. A screen of perhydro-1,4-oxazepine derivatives of digoxin identified several derivatives (e.g. cyclobutyl) with strongly increased selectivity for inhibition of α2β2 and α2β3 over α1β1 (range 22–33-fold). Molecular modeling suggests a possible basis for isoform selectivity. The preferential assembly, specific T-tubular localization, and low K+ affinity of α2β2 could allow an acute response to raised ambient K+ concentrations in physiological conditions and explain the importance of α2β2 for cardiac muscle contractility. The high sensitivity of α2β2 to digoxin derivatives explains beneficial effects of cardiac glycosides for treatment of heart failure and potential of α2β2-selective digoxin derivatives for reducing cardiotoxicity.

The Na,K-ATPase ␣ 2 subunit plays a key role in cardiac muscle contraction by regulating intracellular Ca 2؉ , whereas ␣ 1 has a more conventional role of maintaining ion homeostasis. The ␤ subunit differentially regulates maturation, trafficking, and activity of ␣-␤ heterodimers. It is not known whether the distinct role of ␣ 2 in the heart is related to selective assembly with a particular one of the three ␤ isoforms. We show here by immunofluorescence and co-immunoprecipitation that ␣ 2 is preferentially expressed with ␤ 2 in T-tubules of cardiac myocytes, forming ␣ 2 ␤ 2 heterodimers. We have expressed human ␣ 1 ␤ 1 , ␣ 2 ␤ 1 , ␣ 2 ␤ 2 , and ␣ 2 ␤ 3 in Pichia pastoris, purified the complexes, and compared their functional properties. ␣ 2 ␤ 2 and ␣ 2 ␤ 3 differ significantly from both ␣ 2 ␤ 1 and ␣ 1 ␤ 1 in having a higher K 0.5 K ؉ and lower K 0.5 Na ؉ for activating Na,K-ATPase. These features are the result of a large reduction in binding affinity for extracellular K ؉ and shift of the E 1 P-E 2 P conformational equilibrium toward E 1 P. A screen of perhydro-1,4-oxazepine derivatives of digoxin identified several derivatives (e.g. cyclobutyl) with strongly increased selectivity for inhibition of ␣ 2 ␤ 2 and ␣ 2 ␤ 3 over ␣ 1 ␤ 1 (range 22-33-fold). Molecular modeling suggests a possible basis for isoform selectivity. The preferential assembly, specific T-tubular localization, and low K ؉ affinity of ␣ 2 ␤ 2 could allow an acute response to raised ambient K ؉ concentrations in physiological conditions and explain the importance of ␣ 2 ␤ 2 for cardiac muscle contractility. The high sensitivity of ␣ 2 ␤ 2 to digoxin derivatives explains beneficial effects of cardiac glycosides for treatment of heart failure and potential of ␣ 2 ␤ 2selective digoxin derivatives for reducing cardiotoxicity.
The Na,K-ATPase plays a key role in cardiac muscle contractility by regulating the cytosolic Ca 2ϩ concentration, via the Na ϩ /Ca 2ϩ exchanger, and hence the excitation-contraction coupling in cardiac myocytes (1). Three isoforms of the Na,K-ATPase ␣ subunit and three isoforms of the ␤ subunit are expressed in cardiac muscle, and their expression levels vary between species (2)(3)(4). In addition, the regulatory FXYD1 subunit and possibly other FXYD proteins are expressed in cardiac muscle (5,6). Although ␣ 1 is the most abundant ␣ subunit isoform in cardiomyocytes in the majority of species, the ␣ 2 isoform is functionally more important for cardiac muscle contractility (7)(8)(9)(10)(11). The mechanisms underlying distinct roles of ␣ 2 and ␣ 1 isoforms in the heart are unclear. In smooth and skeletal muscle, the ␣ 2 subunit is concentrated at the junctions between the tubular membrane and sarcoplasmic reticulum in close proximity to the Na ϩ /Ca 2ϩ exchanger and other components of the excitation-contraction complex (12), resulting in more efficient regulation of contraction by the ␣ 2 subunit in contrast to uniformly distributed ␣ 1 subunit. However, the published data on localization of the two isoforms in cardiac myocytes is inconsistent. For example, preferential T-tubular localization of ␣ 2 but uniform plasma membrane localization of the ␣ 1 have been reported by several groups (11,13), whereas quite the opposite distribution of the isoforms was reported (3), and uniform distribution of both ␣ 1 and ␣ 2 subunits has also been reported (14). Measurements of the Na,K-ATPase function of isoforms in cardiac myocytes suggest that the ␣ 2 subunit is more concentrated in T-tubular membranes than in the external sarcolemma, whereas the ␣ 1 subunit is equally distributed in the plasma membrane (11,15,16).
Assembly with the ␤ subunit is required for maturation, trafficking, membrane insertion, and transport activity of the ␣ subunit (17), and three ␤ isoforms differentially regulate these processes. Three ␤ subunit isoforms differ from each other in the number of N-linked glycans, and recombinant addition or removal of N-glycans has been shown to alter the polarized sorting of the Na,K-ATPase (18,19). The three ␤ subunit isoforms differentially modulate voltage dependence of the Na,K-ATPase activity and the apparent affinity of the enzyme for Na ϩ , K ϩ and inhibitors (20,21). ␤ 1 , but not ␤ 2 or ␤ 3 , undergoes post-translational modifications, namely glutathionylation and palmitoylation, and ␤ 1 glutathionylation induced by oxidants decreases the Na,K-ATPase activity (22,23). Thus, the specific assembly with particular ␤ isoforms could account for distinct physiological roles of ␣ 1 and ␣ 2 isoforms. However, it is not known whether the ␣ 2 subunit has a preference for a particular ␤ isoform in cardiac myocytes.
To examine possible functional differences between ␣ 2 ␤ (1-3) isoforms, we have expressed human ␣ 2 with all three human ␤ subunits in Pichia pastoris, purified the complexes, and compared their functional characteristics and inhibitor sensitivity. Previous work has demonstrated some features of ␣ 2 ␤ 2 and ␣ 2 ␤ 3 when expressed in Xenopus oocytes and SF-9 insect cells (20,24). Experimentally, the purified complexes are advantageous in that they allow characterization of the functional properties and inhibitor selectivity of each isoform separately (25)(26)(27) and also detailed mechanistic properties of ion binding and conformational changes (28 -30).
The inhibition of the Na,K-ATPase by digitalis CGs 5 has been used for years to treat heart failure. CGs increase the force of cardiac muscle contraction by reducing the inward Na ϩ gradient that decreases Ca 2ϩ extrusion via the Na ϩ /Ca 2ϩ exchanger (NCX1), leading to increased Ca 2ϩ -induced Ca 2ϩ release from the sarcoplasmic reticulum during excitation-contraction coupling. As a toxic side effect, excessive inhibition of the Na,K-ATPase increases bulk intracellular Na ϩ concentration, excessive accumulation of Ca 2ϩ ions (i.e. "calcium overload"), and "spontaneous" Ca 2ϩ release from sarcoplasmic reticulum that can trigger cardiac arrhythmias (1,31). The preferential role of ␣ 2 in excitation-contraction coupling suggests that ␣ 2 -specific inhibitors might be able to induce an ionotropic effect without triggering Ca 2ϩ overload and arrhythmias. Some years ago, we demonstrated that some natural CGs, such as digoxin and digitoxin, exhibit a moderate intrinsic selectivity for ␣ 2 over ␣ 1 , whereas aglycones, such as digoxigenin and digitoxigenin, show no selectivity (25). Thus, the isoform selectivity was attributed to the sugar moiety, especially the third digitoxose. It was proposed that modification of the third sugar could raise selectivity for ␣ 2 . Indeed, chemical modification of the third digitoxose residue of digoxin by periodate oxidation and reductive amination by primary amines, R-NH 2 , produced perhydro-1,4-oxazepine derivatives with enhanced selectivity of inhibition for ␣ 2 ␤ 1 over ␣ 1 ␤ 1 (27). Most recently, we have described perhydro-1,4-oxazepine digoxin derivatives with various straight chain, branched, and cyclic or heterocyclic aliphatic substitutions and shown that compounds with four carbon substitutions (cyclobutyl (DcB), methyl cyclopropyl (DMcP), and isobutyl (DiB)) showed an especially high selectivity for ␣ 2 ␤ 3 /␣ 1 ␤ 1 . ␣ 2 ␤ 3 is the principal Na,K-pump isoform in non-pigmented cells of ciliary epithelium. We have shown that the digoxin derivatives with enhanced selectivity for ␣ 2 ␤ 1 and especially ␣ 2 ␤ 3 efficiently reduce intraocular pressure when applied topically to rabbit eyes (26,27).
We present evidence here that the ␣ 2 and ␤ 2 isoforms preferentially assemble with each other in the heart and reside predominantly in the T-tubules. By systematic analysis of properties of purified ␣ 2 ␤ 2 in comparison with ␣ 1 ␤ 1 , ␣ 2 ␤ 1 , and ␣ 2 ␤ 3 , we demonstrate distinctive functional properties and isoformselective inhibition of ␣ 2 ␤ 2 , which explain the important role of ␣ 2 for myocardial contractility and the pharmacological potential of ␣ 2 ␤ 2 -selective CGs.

Results
Distribution of Na,K-ATPase ␣ 1 , ␣ 2 , ␤ 1 , and ␤ 2 Subunits in Rat and Human Heart-Normal rat or human frozen heart sections were used to study the intracellular localization of the Na,K-ATPase subunit isoforms by immunofluorescence. In rat cardiomyocytes, the Na,K-ATPase subunits were differentially distributed (Fig. 1A). The ␣ 1 isoform was less abundant in the T-tubular membranes than in the external sarcolemma. Conversely, the ␣ 2 isoform was more abundant in the T-tubular membranes than in the sarcolemma, consistent with previously published results (11,13,21). Both isoforms are expressed at the intercalated discs, but ␣ 1 subunit expression is more pronounced. The immunofluorescence pattern with the Na,K-ATPase ␤ 2 antibody was similar to that of the ␣ 2 antibody with major expression observed in the T-tubules. In frozen sections of human hearts, the ␤ 1 isoform was present more abundantly in the sarcolemma than in the T-tubular membranes, whereas the ␤ 2 isoform was present exclusively in the T-tubules (Fig.  1B). These results demonstrate a differential localization of the Na,K-ATPase subunits with the ␣ 2 and ␤ 2 Na,K-ATPase subunits following the same T-tubule-specific expression pattern and the ␣ 1 and the ␤ 1 subunits ubiquitously distributed but more abundant in the sarcolemma.
Because the Na,K-ATPase is a crucial component in regulating postnatal cardiac function (32), we analyzed whether the Na,K-ATPase subunits are also selectively expressed during embryogenesis. Paraffin-embedded sections of mouse embryos (embryonic day 12.5) were analyzed by immunofluorescence (Fig. 2). The Na,K-ATPase ␣ 1 subunit was expressed ubiquitously (top and bottom left panels), whereas the ␣ 2 and ␤ 2 were mostly restricted to the heart (top and bottom right panels, respectively).
The Na,K-ATPase ␣ 2 and ␤ 2 Subunits Are Selectively Coimmunoprecipitated from Mouse Heart-To analyze the composition of the Na,K-ATPase heterodimers present in heart microsomal membranes, proteins were co-immunoprecipitated with an ␣ 2 subunit-specific antibody, and the presence of ␣ 1 , ␤ 1 , ␤ 2 , and ␤ 3 subunits was analyzed by Western blotting. 5  To prevent the overlap of the ␤ subunits bands with the band corresponding to the heavy chain of the immunoprecipitating antibody, the immunoprecipitated proteins were treated with PNGase F before SDS-PAGE. Immunoprecipitation of the ␣ 2 subunit resulted in the co-immunoprecipitation of the ␤ 2 subunit and minor amounts of ␤ 1 and ␤ 3 subunits (Fig. 3A, left). No ␣ 1 subunits were detected in the immunoprecipitation with the ␣ 2 -specific antibody. In contrast, when the ␣ isoform-nonspecific antibody was used, both ␣ 1 and ␣ 2 subunits were immunoprecipitated, and approximately equal amounts of all of the three ␤ subunits were co-immunoprecipitated (Fig. 3A, right). Conversely, immunoprecipitation with a ␤ 2 -specific antibody resulted in co-immunoprecipitation of the ␣ 2 but not the ␣ 1 subunit (Fig. 3B). Taken together, these results suggest that the ␣ 2 and ␤ 2 subunit isoforms associate predominantly with each other but not with other isoforms expressed in the heart.
Expression and Purification of Na,K-ATPase ␣ 2 ␤ 2 -␣ 2 ␤ 2 and ␣ 2 ␤ 3 were expressed in P. pastoris as described under "Experimental Procedures." Under optimal expression conditions, specific ouabain binding for the ␣ 2 ␤ 2 clone was 8 Ϯ 1 pmol/mg protein and 10 Ϯ 2 pmol/mg for the ␣ 2 ␤ 3 clone, respectively. The addition of DMSO to the culture medium, which was reported to increase expression of GPCRs in P. pastoris, did not increase expression (33).
Both isoforms were purified via the N-terminal His tag of the ␤ subunit by metal affinity chromatography on BD-Talon beads and reconstituted with purified human FXYD1 on the BD-Talon beads, as described in Experimental Procedures. 6 The purity of ␣ 2 ␤ 2 and ␣ 2 ␤ 3 was comparable with that of purified human ␣ 1 ␤ 1 and ␣ 2 ␤ 1 complexes. As depicted in Fig. 4, at least five bands of ␤ 2 were observed, in comparison with two bands for ␤ 3 and ␤ 1 . The two bands of ␤ 1 were shown previously to represent two glycosylated versions of the ␤ 1 subunit of the Man X -GlcNAc 2 type, typical for P. pastoris (34). An increase in heterogeneity of glycoforms of ␤ 2 as compared with ␤ 3 and ␤ 1 is consistent with the presence of eight N-glycosylation sites in ␤ 2 and only two and three in ␤ 3 and ␤ 1 , respectively. When deglycosylated by PNGase treatment, all three ␤ isoforms migrated at their expected molecular mass of approximately 35, 31, and 33 kDa for ␤ 1 , ␤ 3 , and ␤ 2 , respectively.
Recent studies have demonstrated that ␣ 2 ␤ 1 is less stable to thermal and detergent-mediated inactivation than ␣ 1 ␤ 1 , due to suboptimal interaction with phosphatidylserine (35). We investigated the relative effect of the ␤ subunit on Na,K-ATPase isoform stability by thermal inactivation of [ 3 H]ouabain binding to the membranes and Na,K-ATPase activity of the purified proteins, as described in previous publications (see for example Refs. 29 and 65). By both criteria, the ␣ 2 complexes were significantly less thermally stable than ␣ 1 , and ␣ 2 ␤ 2 was somewhat less stable than ␣ 2 ␤ 1 and ␣ 2 ␤ 3 . Thus, the relative instability of the ␣ 2 ␤ (1-3) complexes is a feature attributable primarily to the ␣ 2 subunit. The differences between ␤ isoforms, ␤ 2 Ͻ ␤ 3 Ϸ ␤ 1 , are rather small.
Functional Properties of Purified ␣ 1 ␤ 1 , ␣ 2 ␤ 1 , ␣ 2 ␤ 2 , and ␣ 2 ␤ 3 Complexes-The specific Na,K-ATPase activity of the purified isoforms was highest for ␣ 1 ␤ 1 (16.4 Ϯ 0.7 mol/mg/min), followed by similar values for ␣ 2 ␤ 1 (10.9 Ϯ 0.6), ␣ 2 ␤ 3 (10.7 Ϯ 1.9), and ␣ 2 ␤ 2 (8.4 Ϯ 1.4; Table 1, column 2). Note that FXYD1 itself inhibits the Na,K-ATPase activity of the purified human ␣ 1 ␤ 1 by about 25% and ␣ 2 ␤ 1 by about 15% compared with the ␣␤ complexes alone (30). One kinetic property showing large differences between the isoform complexes was the K 0.5 K ϩ for activation of Na,K-ATPase activity, with values of 1.5 Ϯ 0.1 mM for ␣ 1 ␤ 1 and 2.7 Ϯ 0.1 mM for ␣ 2 ␤ 1 , whereas the K 0.5 K ϩ for both ␣ 2 ␤ 2 and ␣ 2 ␤ 3 was much higher, with apparent K 0.5 values of 7.4 Ϯ 0.2 and 6.4 Ϯ 0.5 mM, respectively (Table 1, column 4). Sodium titrations for activation of Na,K-ATPase activity revealed that K 0.5 Na ϩ was not different between ␣ 1 ␤ 1 and ␣ 2 ␤ 1 , whereas K 0.5 Na ϩ for ␣ 2 ␤ 2 and ␣ 2 ␤ 3 was significantly lower ( Table 1, column 3). We have also looked at the affinity for FIGURE 1. Na,K-ATPase ␤ 2 and ␣ 2 subunits are localized almost exclusively in T-tubules in cardiomyocytes, whereas the ␣ 1 and ␤ 1 subunits are localized in both sarcolemma and T-tubules. A, frozen sections of rat heart were double-stained by using mouse antibodies against ␣ 1 subunit (green) and rabbit antibodies against ␣ 2 subunit (red) (top panels) or by using mouse antibodies against ␣ 1 subunit (green) and rabbit antibodies against ␤ 2 subunit (red) (bottom panels). Anti-mouse Alexa Fluor 488-conjugated secondary antibodies were used to detect anti-␣ 1 primary antibodies, and antirabbit Alexa Fluor 633-conjugated secondary antibodies were used to detect anti-␤ 2 and anti-␣ 2 primary antibodies. The arrows show localization of the ␣ 1 subunits, but not of ␣ 2 and ␤ 2 subunits, in the sarcolemma. The arrowheads show co-localization of ␣ 1 and ␤ 2 subunits or ␣ 1 and ␣ 2 subunits in T-tubules. B, frozen sections of human heart were double-stained by using mouse antibodies against ␤ 1 subunit (green) and rabbit antibodies against ␤ 2 subunit (red). Anti-mouse Alexa Fluor 488-conjugated secondary antibodies were used to detect anti-␤ 1 primary antibodies, and anti-rabbit Alexa Fluor 633conjugated secondary antibodies were used to detect anti-␤ 2 primary antibodies. The arrows show localization of the ␤ 1 subunits, but not of ␤ 2 subunits, in the sarcolemma. The arrowheads show co-localization of ␤ 1 and ␤ 2 subunits in T-tubules. The stealth-like arrowheads show the T-tubules in which the ␤ 2 subunits, but not the ␤ 1 subunits, are present. inhibition of Na,K-ATPase activity by vanadate, which is a phosphate analogue that binds to the E 2 (2K) conformation, mimicking the transition state E 2 P2K during dephosphorylation (36). All three ␣ 2 isoforms have a higher K i vanadate compared with ␣ 1 ␤ 1 (K i ϭ 0.48 M), and the effects are greatest with ␤ 2 and ␤ 3 in the order Table 1,

column 5).
A simple explanation of the raised K 0.5 K ϩ in Na,K-ATPase activity assays could be that the ␤ 2 and ␤ 3 subunit reduce the binding affinity of 2K ions for their extracellular sites. K ϩ and Na ϩ binding was determined by using the electrochromic shift dye RH421 in a medium of fixed ionic strength (containing also 5 mM magnesium ions) ( Fig. 5 and Table 2) (29,30,37). The inset of Fig. 5A shows the typical RH421 responses upon the addition of Na ϩ ions (E 1 -E 1 (3Na)) and then ATP (to E 2 P) and K ϩ ions (E 2 (2K)) for the ␣ 2 ␤ 2 complex, as explained in recent papers (29). By varying Na ϩ or K ϩ concentrations, equilibrium titrations of either 3Na ϩ binding to cytoplasmic sites or 2K ϩ FIGURE 2. The Na,K-ATPase ␣ 2 subunit and ␤ 2 subunit are preferentially expressed in the mouse embryonic heart in contrast to the ubiquitously expressed ␣ 1 subunit. A, paraffin-embedded sections of mouse embryos (embryonic day 12.5) were double-stained by using mouse antibodies against ␣ 1 subunit (left panels) and rabbit antibodies against ␣ 2 subunit or ␤ 2 subunit (right panels). Anti-mouse Alexa Fluor 488-conjugated secondary antibodies were used to detect anti-␣ 1 primary antibodies, and anti-rabbit Alexa Fluor 633-conjugated secondary antibodies were used to detect anti-␣ 2 or anti-␤ 2 primary antibodies. The rectangles show the embryonic hearts. B, a scheme demonstrating the assembly of the images shown in A. Each rectangle outlined by a dotted line represents an individual confocal microscopy image taken at ϫ10 magnification. FIGURE 3. The Na,K-ATPase ␣ 2 subunit and ␤ 2 subunit preferentially interact with each other in mouse heart. A, Western blotting analysis of proteins immunoprecipitated and co-immunoprecipitated from the detergent extracts of mouse heart microsome membranes by using either the ␣ 2 -specific antibodies (left panels) or the ␣-nonspecific antibodies (right panels) shows preferential co-immunoprecipitation of the ␤ 2 subunit with the ␣ 2 subunit. B, Western blotting analysis of the immunoprecipitated ␤ 2 subunit and co-immunoprecipitated ␣ 2 subunit isoforms shows that the ␣ 2 subunit is preferentially co-precipitated with the ␤ 2 subunit. Input lanes contain 10% of the extract used for immunoprecipitation. To prevent an overlap of the ␤ subunit bands with the heavy chain band of the antibodies used for immunoprecipitation, the immunoprecipitated proteins were treated with PNGase F before SDS-PAGE. IP, immunoprecipitation; DG, deglycosylated.
binding to extracellular sites are readily obtained, leading to curves such as those in Fig. 5, A and B. The binding parameters derived from best fits of the curves to the Hill equation are collected in Table 2, where K 0.5 Na ϩ and K 0.5 K ϩ represent the intrinsic binding affinity for 3Na ϩ and 2K ϩ ions, respectively. Evidently, the intrinsic binding affinity for 3Na ϩ ions is the same for all of the isoform complexes. By contrast, in these conditions, the binding affinity of 2K ϩ ions to ␣ 2 ␤ 1 is significantly lower than to ␣ 1 ␤ 1 , and the binding affinity for both ␣ 2 ␤ 2 and ␣ 2 ␤ 3 is further strongly reduced compared with ␣ 2 ␤ 1 . Note that, compared with K 0.5 K ϩ for ␣ 1 ␤ 1 , the K 0.5 K ϩ for ␣ 2 ␤ 2 or ␣ 2 ␤ 3 is almost an order of magnitude higher. Because vanadate binds primarily to the E 2 (2K) conformation, the differences in K i for vanadate inhibition of Na,K-ATPase activity could be secondary to the differences in K 0.5 K ϩ values, with the same order for the K i as for Table 1).
An alternative or additional explanation to that just given is that raised K 0.5 K ϩ of the ␣ 2 ␤ 2 and ␣ 2 ␤ 3 complexes is caused by an ␣ 2 ␤ 2 -or ␣ 2 ␤ 3 -induced shift in poise of the E 1 -E 2 conformational equilibrium toward E 1 or E 1 -P. This explanation would also be consistent with the parallel reduction in apparent K 0.5 Na ϩ and increase in K i for vanadate (Table 1, column 3). A direct test of the relative effects of ␤ 1 , ␤ 2 , and ␤ 3 on the conformational changes was made by measuring the rates of the E 1 P(3Na) 3 E 2 P and E 2 (2Rb)ATP 3 E 1 3Na⅐ATP transitions using RH421 in stopped-flow experiments, as described in our recent publications (29,30). Traces for the ␣ 2 ␤ 2 complex are shown in Fig. 5, C and D, and the rate constants for all of the isoform complexes are collected in Table 3. These data show directly that for ␣ 2 ␤ 1 , the rates of E 1 P(3Na) 3 E 2 P are indeed slower than for ␣ 1 ␤ 1 , and for both ␣ 2 ␤ 2 and ␣ 2 ␤ 3 , the rate is still slower than for ␣ 2 ␤ 1 .
The rates of E 2 (2Rb)ATP 3 E 1 3Na⅐ATP for ␣ 2 ␤ 1 , ␣ 2 ␤ 2 , and ␣ 2 ␤ 3 are all significantly slower than for ␣ 1 ␤ 1 , but they are indistinguishable from each other. The turnover rate in s Ϫ1 for each isoform was calculated from the expression (a ϫ b)/(a ϩ b), where a is the rate of E 1 P(3Na) 3 E 2 P and b is the rate of E 2 (2Rb)ATP 3 E 1 3Na⅐ATP (Table 3), assuming that the rates of phosphorylation E 1 Na ϩ ATP 3 E 1 P(3Na) and dephosphorylation E 2 P2Rb 3 E 2 (2Rb) are fast and do not significantly limit the turnover rate. The calculated value for ␣ 1 ␤ 1 (14 s Ϫ1 ) is greater than for ␣ 2 ␤ 1 , ␣ 2 ␤ 2 , and ␣ 2 ␤ 3 , which are not significantly different from each other (9.17, 8.76, and 8.75 s Ϫ1 , respectively; average of 8.89). The ratios of the calculated turnover rates for ␣ 1 ␤ 1 /␣ 2 ␤ 1 , ␣ 1 ␤ 1 /␣ 2 ␤ 2 , and ␣ 1 ␤ 1 /␣ 2 ␤ 3 are essentially the same, with an average of 1.6. These ratios are close to those of Na,K-ATPase activities in Table 1, validating the assumptions that underlie the calculation of the turnover rates.

TABLE 1 Functional properties of purified Na,K-ATPase complexes
Specific activity, K 0.5 Na ϩ , K 0.5 K ϩ , and K i vanadate of purified Na,K-ATPase complexes are shown. Values represent averages of at least three different experiments Ϯ S.E. The reaction medium contained sodium plus potassium as indicated, 1 mM ATP, 3 mM MgCl 2 , 25 mM histidine, pH 7.4, 1 mM EGTA, 0.01 mg/ml SOPS, 0.001 mg/ml cholesterol, and 0.005 mg/ml C 12 E 8 . Maximal Na,K-ATPase activity was measured in the presence of 120 mM NaCl, 20 mM KCl. For ion titrations, ATPase activity was measured in medium containing 80 mM KCl and 0 -120 mM NaCl or 120 mM NaCl and 0 -40 mM KCl. Ionic strength was maintained constant with choline chloride. K 0.5 values were obtained from least square fits of the data points to the Hill equation. K i vanadate was determined in medium containing 120 mM NaCl, 20 mM KCl, and 1 mM ATP, and data points were fitted to a one-side inhibition model.

Isoform complexes
Specific Na,K-ATPase activity values and selectivity ratios for all of these compounds are summarized in Table 4. Compounds that have the highest selectivity, compared with digoxin itself, are indicated by double asterisks, and compounds with significant but lower selectivity are shown with a single asterisk. There is a clear peak for the cyclobutyl derivative with four carbon atoms, DcB, with 16.9-, 22.2-, and 33.6-fold selectivity for ␣ 2 ␤ 1 , ␣ 2 ␤ 2 , and ␣ 2 ␤ 3 over ␣ 1 ␤ 1 , respectively. Overall, Table 4 shows that aliphatic substituents with four carbon are most selective (DcB Ͼ DMcP Ͼ to the E 1 conformation. The inset shows a standard experiment and fluorescence changes associated with ion binding and release to ␣ 2 ␤ 2 (sodium binding to E 1 , sodium release and conformational transition to E 2 P, and potassium binding). B, equilibrium titration of potassium binding to E 2 P. C, stopped-flow trace of the E 1 Na3 3 E 2 P transition of ␣ 2 ␤ 2 fitted to a double exponential function. The average of 15 traces is shown. D, stopped-flow trace of the ␣ 2 ␤ 2 E 2 (Rb2)ATP 3 E 1 Na3ATP transition fitted to a single exponential function.
DiB), whereas the substituents with three carbons (DiP Ϸ DcP Ͼ DP) are also quite selective, and the selectivity ratios decline for substituents with 5, 6, or 7 carbon atoms. The sulfonyl derivatives DMSM, DEMS, and DESA also showed strongly increased selectivity over Potassium-Cardiac Glycoside Antagonism-CGs bind with high affinity to E 2 P, and potassium binding causes rapid dephosphorylation to E 2 (2K), which has a much lower affinity for CGs than E 2 P (38). Because K 0.5 K ϩ of ␣ 2 ␤ 2 , ␣ 2 ␤ 3 and also ␣ 2 ␤ 1 for activating Na,K-ATPase are higher than for ␣ 1 ␤ 1 in the order (Table 1), a higher selectivity for ␣ 2 ␤ 2 and ␣ 2 ␤ 3 and also ␣ 2 ␤ 1 might be due, at least partially, to a weaker K ϩ -CG antagonism. To assess the effect of K ϩ -CG antagonism more systematically, we have measured K i values for digoxin, the digoxin derivatives and ouabain at increasing K ϩ concentrations from 2.5, 5, 10, and 20 mM K ϩ . Fig. 7 presents the data for digoxin, but very similar data were obtained for the isobutyl derivative of digoxin (DiB) and also ouabain (not shown). For ␣ 1 ␤ 1 , the K i values increased by 3-4-fold in the range of 2.5-20 mM K ϩ , respectively. By contrast, the K i values for ␣ 2 ␤ 1 were much less affected by the increased potassium concentration, and inhibition of ␣ 2 ␤ 2 and ␣ 2 ␤ 3 was essentially unaffected by potassium ions in the range 2.5-20 mM. Consequently, the selectivity for inhibition of ␣ 2 ␤ 2 over ␣ 1 ␤ 1 by digoxin (or other CGs) significantly increased over the range 2.5, 5, 10, and 20 mM potassium from 3.67, 4.77, 11.6, to 15.0, respectively. The differences in K ϩ -digoxin antagonism were also investigated more directly by K ϩ -[ 3 H]digoxin displacement assays (Fig. 8). Yeast membranes harboring either ␣ 1 ␤ 1 , ␣ 2 ␤ 1 , ␣ 2 ␤ 2 , or ␣ 2 ␤ 3 were equilibrated with [ 3 H]digoxin and subsequently incubated with increasing amounts of KCl (up to 150 mM). In the conditions of digoxin binding (with vanadate/magnesium but sodium-free), the apparent K 0.5 K ϩ for digoxin displacement was 0.62 mM for ␣ 1 ␤ 1 (see also Ref. 25), 1.8 Ϯ 0.1 mM for ␣ 2 ␤ 1 , 2.8 Ϯ 0.1 mM for ␣ 2 ␤ 2 , and 3.2 Ϯ 0.5 mM for ␣ 2 ␤ 3 . Furthermore, at high K ϩ concentrations, displacement of digoxin binding was incomplete for ␣ 2 ␤ 1 , ␣ 2 ␤ 2 , and ␣ 2 ␤ 3 , the remaining fraction at saturating potassium being 20.3 Ϯ 1.5, 26 Ϯ 1.8, and 30.3 Ϯ 1.5% of control, respectively. Thus, K ϩ -CG antagonism reflects both the affinity for potassium ions and the maximal degree of displacement by potassium ions.

Rates of conformational changes for the different isoform complexes determined by stopped-flow measurements
All rates were measured at 23°C as described under "Experimental Procedures." The turnover rate was calculated from the function (a ϫ b)/(a ϩ b).   Overall, in conditions similar to the extracellular physiological medium (140 mM sodium, 5 mM potassium or with potassium elevated to 20 mM), K-CG antagonism is prominent for ␣ 1 ␤ 1 , weak for ␣ 2 ␤ 1 and negligible for ␣ 2 ␤ 2 and ␣ 2 ␤ 3 . Note, however, that, in similar conditions, the selectivity ratios for all six derivatives DiB, DMcP, DcB, DMSM, DEMS, and DESA for ␣ 2 ␤ 3 /␣ 1 ␤ 1 Ͼ ␣ 2 ␤ 2 /␣ 1 ␤ 1 Ͼ ␣ 2 ␤ 1 /␣ 1 ␤ 1 are significantly greater than for digoxin itself (Table 4). This shows that reduced K-CG antagonism can indeed account only partially for the increased selectivity for ␣ 2 ␤ 3 and ␣ 2 ␤ 2 , whereas the structure of the derivatives is a crucial element determining the isoform selectivity.

Discussion
We discuss here the data showing selective assembly of ␣ 2 ␤ 2 in cardiac myocytes and distinct functional properties and iso-form-selective inhibition of human ␣ 2 ␤ 2 and ␣ 2 ␤ 3 , together with possible molecular explanations and physiological or pharmacological implications. Comparisons of functional properties and inhibition of ␣ 2 ␤ 1 with ␣ 1 ␤ 1 reveal the differences between ␣ 2 and ␣ 1 , whereas comparisons of ␣ 2 ␤ 2 or ␣ 2 ␤ 3 with ␣ 2 ␤ 1 reveal the influence of ␤ 2 or ␤ 3 compared with ␤ 1 . As one general conclusion, it is the combined effects of ␣ 2 and ␤ 2 (or ␤ 3 ) that give rise to the distinctive functional properties and isoform-selective inhibition of ␣ 2 ␤ 2 (or ␣ 2 ␤ 3 ).
Specific Assembly of ␣ 2 ␤ 2 in Heart-Three of the four Na,K-ATPase ␣ subunit isoforms and all three ␤ subunit isoforms are expressed in the heart. Although ␣ 1 is the most abundant ␣ subunit isoform, the ␣ 2 isoform rather than the ␣ 1 isoform plays a key role in cardiac muscle contractility by regulating the cytosolic Ca 2ϩ concentration in cardiac myocytes. The transient

TABLE 4 K i values and selectivity ratios for the inhibition of purified Na,K-ATPase isoforms by perhydro-1,4-oxazepine derivatives of digoxin
The reaction medium contained 130 mM NaCl, 5 mM KCl, 3 mM MgCl 2 , 25 mM histidine, pH 7.4, 1 mM EGTA, 0.01 mg/ml SOPS, 0.001 mg/ml cholesterol, and 0.005 mg/ml C 12 E 8 . Each experiment was carried out at least three times. The calculated K i values represent averages Ϯ S.E. *, compounds with significantly higher selectivity (K i ␣ 1 ␤ 1 /␣ 2 ␤ 1-3 ) than digoxin (Ͼ4-and Ͻ10-fold). **, compounds with the highest selectivity (K i ␣ 1 ␤ 1 /␣ 2 ␤ 1-3 ) compared with digoxin (Ͼ12-fold). . Potassium-digoxin antagonism depicted for ␣ 1 ␤ 1 , ␣ 2 ␤ 1 , ␣ 2 ␤ 2 , and ␣ 2 ␤ 3 . Inhibition of purified Na,K-ATPase isoforms by digoxin was measured, and K i values were plotted against the potassium concentration in the assay medium. Error bars, S.E. rise in Ca 2ϩ concentration associated with electrical excitability then triggers cardiac muscle contraction (1). Despite extensive studies of isoform-specific properties of the Na,K-ATPase, the reasons for differential roles of ␣ 2 and ␣ 1 isoforms in cardiac muscle contractility are not fully understood. The data presented here indicate that the ␣ 2 and ␤ 2 isoforms preferentially assemble with each other in the heart as detected by co-immunoprecipitation (Fig. 3). These data are consistent with previous reports on selective assembly (17) and co-purification of ␣ 2 and ␤ 2 isoforms in the brain (39). Immunofluorescence data in rat and human heart sections indicate that both ␣ 2 and ␤ 2 isoforms specifically localize to the T-tubular membranes, whereas ␣ 1 and ␤ 1 isoforms are distributed in both T-tubular and external sarcolemma membranes (Fig. 1). The data on preferential T-tubular localization of the ␣ 2 and ubiquitous distribution of the ␣ 1 are in agreement with previous reports on measurements of the isoform-specific Na,K-ATPase activity in cardiac myocytes (11,15,16,40) and several immunofluorescence reports (11,13). On the other hand, our data contrast with several other reports on immunodetection of the Na,K-ATPase ␣ isoforms in the heart (3,14). The uniform distribution of the ␤ 1 isoform has been reported previously (3), whereas this is the first report describing the specific localization of the ␤ 2 isoform in the heart.

CG K i ؎ S.E. Selectivity
It seems paradoxical that there are two isoform complexes, ␣ 2 ␤ 2 and ␣ 2 ␤ 3 , with similar functional properties, as shown in this paper, but in the heart, ␣ 2 assembles with ␤ 2 rather than with ␤ 3 . Assembly with a particular ␤ isoform is known to affect trafficking and polarized sorting of the ␣ subunit (19,41,42), suggesting that association with the ␤ 2 can be important for the specific location of the ␣ 2 isoform in the T-tubules. Particularly, recombinant addition of N-glycosylation sites to the ␤ 1 subunit has been shown to alter localization of the Na,K-ATPase from the basolateral to the apical domain of the plasma membrane in gastric epithelial cells (19). It is possible that eight N-glycans versus two or three in the ␤ 3 or ␤ 1 isoform play a role in this specific targeting of the ␣ 2 to the microdomains of T-tubular membranes that contain Na ϩ /Ca 2ϩ exchanger and are proximal to other components of Ca 2ϩ -regulating complex, which can explain the importance of the ␣ 2 isoform in cardiac muscle contraction.
Another difference between ␤ 3 and ␤ 2 is a role of ␤ 2 in cell adhesion (39,43). Whether this adhesive role is required for specific location of ␣ 2 ␤ 2 in T-tubules is not clear, but ␣ 2 ␤ 2 differs from ␣ 2 ␤ 3 in this crucial aspect.
Functional Properties of the ␣ 2 ␤ 1 , ␣ 2 ␤ 2 , and ␣ 2 ␤ 3 Isoform Complexes-The major functional differences between ␣ 2 ␤ 2 and ␣ 2 ␤ 3 when compared with ␣ 2 ␤ 1 , and especially ␣ 1 ␤ 1 , is the raised K 0.5 K ϩ for activating Na,K-ATPase, reduced K 0.5 Na ϩ , and turnover rate ( Table 1). These results confirm previous evidence that ␣ 2 raises K 0.5 K ϩ compared with ␣ 1 when it is complexed with ␤ 1 (20,24,44) and now show that both ␤ 2 and ␤ 3 lower K 0.5 Na ϩ as well as raising K 0.5 K ϩ when complexed with ␣ 2 , in comparison with complexes with ␤ 1 . The RH421 experiments show that the principal mechanism of the effects of ␣ 2 versus ␣ 1 and ␤ 2 and ␤ 3 versus ␤ 1 to raise K 0.5 K ϩ is reduction of the intrinsic binding affinity for potassium ions at the extracellular surface, with K 0.5 K ϩ values in the order Fig. 5 and Table 2). In RH421 experiments, the potassium affinity is determined in the presence of 50 mM sodium, and, in principle, the reduced affinity for K exc ions could reflect an increased affinity for Na exc and competition with K exc . Nevertheless, displacement of digoxin by potassium ions in the absence of sodium ions (K 0.5 K ϩ 0.6 Ϯ 0.07 mM for ␣ 1 ␤ 1 , 1.8 Ϯ 0.1 mM for ␣ 2 ␤ 1 , 2.8 Ϯ 0.1 mM for ␣ 2 ␤ 2 , and 3.2 Ϯ 0.5 mM for ␣ 2 ␤ 3 ) shows that there is a true and large reduction in intrinsic K exc affinity (Fig. 8). The vanadate titrations with K i values in the order Table 1) are consistent with the order of decreasing affinities for potassium ions.
Compared with ␣ 1 ␤ 1 , ␣ 2 ␤ 1 , ␣ 2 ␤ 2 , and ␣ 2 ␤ 3 significantly reduced the rate of the conformational transition E 1 P(3Na) 3 E 2 P, with the order Fig. 5 and Table 3). In steady-state Na,K-ATPase conditions, a shift of the E 1 P(3Na) 3 E 2 P conformational equilibrium toward E 1 P would contribute to the higher K 0.5 K ϩ as well as lower K 0.5 Na ϩ (and higher K i vanadate) ( Table 1). Note that the reduced K 0.5 Na ϩ of ␣ 2 ␤ 2 and ␣ 2 ␤ 3 compared with ␣ 1 ␤ 1 and ␣ 2 ␤ 1 is not explained by a change in intrinsic binding affinity for sodium ions (see Fig. 5A).
␣ 2 ␤ 1 also displayed a reduced rate of E 2 (2Rb)ATP3 E 1 (3Na)ATP when compared with ␣ 1 ␤ 1 , but there was no further decrease in ␣ 2 ␤ 2 and ␣ 2 ␤ 3 . This finding explains the reduced turnover rate for all ␣ 2 complexes compared with ␣ 1 . The different ␤ isoforms appear to have no isoform-selective influence on the turnover rate, which is governed by the slow rate-determining step E 2 (2Rb)ATP 3 E 1 (3Na)ATP, reduced by ␣ 2 , and not by the faster E 1 P(3Na) 3 E 2 P transition.
In the absence of high resolution structures of the ␣ 2 complexes, one can only speculate on possible explanations of the kinetic effects of ␣ 2 and ␤ 2 or ␤ 3 , such as the reduced K exc binding affinity. Using chimeras of Na,K-ATPase and H,K-ATPase, the extracellular domain was identified as the main modifier of the apparent potassium affinity (for displacing bound ouabain) (45), and the main interaction site was shown to be an SYGQ motif in the M7-8 loop of ␣ (46). In an older study (47), the extracellular domain of ␤ 1 was suggested to cover the extracellular domain of the ␣ subunit and control access to the Rb(K) binding sites. In general, this concept appears compatible with the molecular structures of Na,K-AT-Pase (␣ 1 ␤ 1 ) (48 -50). Thus, compared with ␣ 1 ␤ 1 , the K exc entry and exit pathway may be more accessible in ␣ 2 ␤ 1 itself and even more so in ␣ 2 ␤ 2 and ␣ 2 ␤ 3 , leading to the large decrease in K exc binding affinity.
Digoxin Derivatives with Strong Selectivity for ␣ 2 ␤ 2 and ␣ 2 ␤ 3 Complexes-The conclusion from all of the experiments in Table 4 is that digoxin perhydro-1,4-oxazepine derivatives with four carbon substitutions (cyclobutyl (DcB), methyl cyclopropyl (DMcP), and isobutyl (DiB)) have the highest selectivity for ␣ 2 ␤ 3 /␣ 1 ␤ 1 and ␣ 2 ␤ 2 /␣ 1 ␤ 1 compared with compounds with 1-3 or 5 or a greater number of carbon atom substitutions. This confirms the data in Ref. 26 for ␣ 2 ␤ 3 /␣ 1 ␤ 1 and extends them to ␣ 2 ␤ 2 /␣ 1 ␤. The optimal size of the aliphatic substituents (four carbon atoms), cyclic and non-cyclic, may indicate a size restriction of the space between the ␣ and ␤ subunits. Replacement of one methylene group in the cyclobutyl DcB with a single NH (DAz) or sulfur (DTh) atom strongly reduces selectivity, whereas the three new sulfonyl derivatives, DMSM, DEMS, and DESA, showed enhanced selectivity for ␣ 2 ␤ 2 and ␣ 2 ␤ 3 and to some extent also ␣ 2 ␤ 1 , relative to digoxin itself (Table 4). Thus, the highest selectivity for the ␣ 2 ␤ 2 and ␣ 2 ␤ 3 isoform complexes depends crucially on the structure of the derivatives, providing the strongest evidence for selective interactions of the substituent groups with ␤ 2 and ␤ 3 , respectively.
We have attempted to explain the selectivity of DcB for ␣ 2 ␤ 3 and ␣ 2 ␤ 2 with molecular docking models, starting with a structure of renal Na,K-ATPase (␣ 1 ␤ 1 ) with bound digoxin (51) (Protein Data Bank code 4RET) (Fig. 9). The models display the optimal positions found for DcB relative to digoxin and emphasize the major attractive interactions, particularly with the ␤ subunit. The modeling supports the experimental result whereby DcB binds to ␣ 2 ␤ 3 with the highest potency and selectivity, showing a hydrophobic interaction between the DcBcyclobutyl moiety and ␤ 3 Val-88. There is also an electrostatic interaction of the protonated nitrogen of perhydro-1,4-oxazepine ring with Asp-889 of ␣ 2 . In the case of ␣ 2 ␤ 2 , to which DcB also binds with good potency and selectivity, the modeling shows a hydrogen bond interaction between the protonated perhydro-1,4-oxazepine ring nitrogen and ␤ 2 Glu-89 and also an additional hydrogen bond between the lactone ring and the structural water molecule located at the bottom of the binding site. It is noticeable that the steroid and lactone moieties are somewhat rotated compared with the digoxin itself. By contrast with the ␣ 2 ␤ 3 and ␣ 2 ␤ 2 models, in the ␣ 1 ␤ 1 model, the steroidlactone moiety of DcB almost exactly overlaps that of digoxin and displays only a weak interaction with ␤ 1 Gln-84. This result also appears to be consistent with the experimental result for ␣ 1 ␤ 1 which displays a much lower potency for DcB and only a small difference from digoxin itself.
Taken together, it is evident that both the ␣ and the ␤ subunits determine the selectivity of the digoxin derivatives, and the ␤ subunit, in particular, has favorable interactions with the substituted third sugar residue.
Physiological Role of ␣ 2 ␤ 2 -In relation to the physiological function, major differences of ␣ 2 ␤ 2 compared with ␣ 1 ␤ 1 include the high K 0.5 K ϩ values (low affinity) for extracellular potassium ions, a somewhat lower turnover rate, and a significantly lower K 0.5 Na ϩ . Although the current experiments have been done with detergent-soluble purified proteins, a very similar effect of ␣ 2 ␤ 2 to strongly raise K 0.5 K ϩ values for extracellular potassium ions compared with ␣ 1 ␤ 1 and also ␣ 2 ␤ 1 has been described with the proteins expressed in Xenopus oocytes (20,21,52). A property that we cannot assess in experiments with detergent-soluble proteins is the dependence of activity on membrane potential. As described previously (20), ␣ 2 ␤ 1 shows a steeper dependence on voltage than ␣ 1 ␤ 1 or ␣ 3 ␤ 1 , and recent work shows that ␣ 2 ␤ 2 shows a particularly steep dependence (21). The voltage dependence of the pump current in physiological conditions is mainly a reflection of Na ϩ exc -mediated FIGURE 9. Models for docking of DcB to ␣ 1 ␤ 1 , ␣ 2 ␤ 2 , and ␣ 2 ␤ 3 . DcB was docked into homology models of human Na,K-ATPase isoforms derived from the porcine E2P⅐Mg⅐digoxin structure (Protein Data Bank code 4RET) as described under "Experimental Procedures." ␣ and ␤ subunits are shown in green and blue ribbon representations. Transmembrane helices 3 and 5 of the ␣-subunit are removed for clarity. Digoxin (yellow) and DcB (purple) are shown in a stick representation, and water and a magnesium ion are shown as spheres. The residues of the ␤ subunit closest to the cyclobutyl moiety, ␤ 1 Gln-84, ␤ 2 Glu-89, and ␤ 3 Val-88, are shown as sticks, and distances to DcB are indicated.
competition with K ϩ exc leading to inhibition at negative potentials. The conditions of our experiments are equivalent to those in cells at 0 mV membrane potential, and compared with physiological conditions, the raised K 0.5 K ϩ of ␣ 2 ␤ 2 represents, if anything, an underestimated value compared with ␣ 1 ␤ 1 . At physiological K ϩ exc of 4.5 mM and resting membrane potentials in the heart or skeletal muscle or brain glial cells (Ϫ70 to 90 mV), ␣ 2 ␤ 2 is largely inactive. Thus, ␣ 2 ␤ 2 acts as a "reserve pump," which responds to acutely increased K ϩ exc or positive membrane potentials by increasing its rate and moderating the changes in K ϩ exc and then restoring the ion gradients after the change (see Refs. 52 and 53). For example, increased activity of skeletal muscles leads to loss of potassium ions and a large increase in K ϩ exc , which can reach as high as 8.3 mM in serum, 10 -12 mM in muscle interstitial fluid, and locally as high as 25 mM in T-tubules (54). In heart muscle, the ␣ 2 ␤ 2 is also almost inactive at resting potentials but becomes acutely activated during raised cardiac activity due to the fact that the membrane potential is positive for a significant fraction of the time during the extended action potentials (21). As in skeletal muscle, the potassium concentration in T-tubules can be presumed to rise significantly. The lower K 0.5 Na ϩ of ␣ 2 ␤ 2 should also contribute significantly to rapid restoration of the sodium gradient associated with increased activity and raised cytoplasmic sodium. Due to coupling of the sodium fluxes mediated by ␣ 2 ␤ 2 with sodium and calcium fluxes mediated by NCX1, the kinetic features of ␣ 2 ␤ 2 described are expected to play an important role in regulation of the calcium dynamics of active versus resting cardiac muscle. Indeed, an important role of ␣ 2 overexpression in calcium dynamics in myocytes, associated with a decreased K 0.5 Na ϩ , has been proposed recently (55), although it was not known which ␤ isoform is coupled with ␣ 2 . The ␣ 1 ␤ 1 complex maintains ion gradients in resting conditions but is not suited for the regulatory role of ␣ 2 ␤ 2 just discussed because the potassium sites (K 0.5 K ϩ ϭ 1.5 Ϯ 0.1 mM) are almost saturated even in resting conditions, and the voltage dependence of ␣ 1 ␤ 1 is shallow.
Pharmacological Implications-As discussed in the Introduction, an ␣ 2 -selective CG could be an efficient inotropic agent. The present findings have the additional interesting implication that an ␣ 2 ␤ 2 -selective CG, such as the digoxin derivatives described here, could also have reduced cardiotoxicity compared with digoxin.
For many years, digoxin was used routinely to treat heart failure, due to its inotropic and chronotropic effects, but it is now used much less on account of the narrow therapeutic range and cardiotoxicity resulting from the well known phenomena of calcium overload and cardiac arrhythmias (31). The incidence of digitalis toxicity has decreased in parallel with its decreased use (56). The main cause of digitalis toxicity is accumulation of digoxin, secondary to decreased renal function, and a major exacerbating factor is hypokalemia, which is prevalent in subjects treated with diuretics in addition to digitalis (56) (for a recent study, see Ref. 57). Indeed, due to the potentiation of digitalis toxicity by hypokalemia, it was even recommended, in the past, to treat severe cases by raising serum potassium (58).
The salient present finding is that inhibition of ␣ 1 ␤ 1 by digoxin and all other CGs is strongly antagonized by raising potassium in the range of 2.5-20 mM, whereas ␣ 2 ␤ 2 is unaffected (Fig. 7). Assuming that digoxin toxicity in vivo, exacerbated by hypokalemia, is associated with excessive Na,K-pump inhibition, this implies that toxicity is indeed mediated by ␣ 1 ␤ 1 . Conversely, the insensitivity of inhibition of ␣ 2 ␤ 2 to potassium (2.5-20 mM) would imply that ␣ 2 ␤ 2 does not play a major role in digitalis toxicity. This conclusion fits well with the rationale that ␣ 2 ␤ 2 -selective derivatives could be effective positive inotropic agents and also have reduced toxic effects.
Considerable evidence exists for the presence of endogenous CG-like compounds in mammalian tissues, such as ouabain or marinobufagenin, that may serve to regulate Na,K-ATPase activity (59 -61). Our unpublished experiments show a higher sensitivity of ␣ 2 ␤ 2 over ␣ 1 ␤ 1 for ouabain (K i ϭ 63.7 Ϯ 9.8 versus 153.0 Ϯ 11.6 nM, respectively), 7 suggesting that endogenous ouabain-like compounds may bind more efficiently to ␣ 2 ␤ 2 isoform and thus specifically regulate the ␣ 2 ␤ 2 ion pumping activity or ␣ 2 -dependent signaling pathways. In particular, ␣ 2 -mediated signaling may explain a specific role of the ␣ 2 subunit in the modulation of blood pressure under stress conditions (60 -63).
Conclusions-In summary, our data demonstrate the specific association of the Na,K-ATPase ␣ 2 isoform with the ␤ 2 and the specific intracellular location of the ␣ 2 ␤ 2 heterodimer. The distinct functional properties of human ␣ 2 ␤ 2 are consistent with an important regulatory role in cardiac muscle contraction. Furthermore, isoform-selective inhibition by digoxin derivatives, such as DcB, suggests that they could be safer cardiac inotropic agents compared with digoxin itself.
Immunoprecipitation-Protein extracts from mouse heart membrane fractions (100 -300 g of protein) were incubated with 30 l of the protein A-agarose suspension (Roche Diagnostics) in a total volume 1 ml of the extraction buffer at 4°C with continuous rotation for at least 3 h (or overnight) to remove the components that non-specifically bind to protein A. The precleared cell extract was mixed with 10 l of polyclonal antibodies against the Na,K-ATPase ␣ 2 subunit (Millipore) or 10 l of polyclonal antibodies against the Na,K-ATPase ␣ subunit (64) or 10 l of polyclonal antibodies against the Na,K-ATPase ␤ 2 subunit (Millipore) and incubated with continuous rotation at 4°C for 60 min. After the addition of 30 l of the protein A-agarose suspension, the mixture was incubated at 4°C with continuous rotation overnight. The bead-adherent complexes were washed three times on the beads and then eluted as described previously (65).
Where indicated, the bead-adherent proteins were treated with PNGase F. Deglycosylation by PNGase F was performed by incubation of the bead-adherent proteins with 1 l of PNGase F in 30 l of 50 mM sodium phosphate, pH 7.5, containing 1% Nonidet P-40 at 37°C for 1 h. After incubation with glycosidases, the reaction mixture was separated from the beads. The adherent proteins were eluted from the beads by incubation in 30 l of 2ϫ SDS-PAGE sample buffer for 5 min at 80°C. To account for possible dissociation of immunoprecipitated proteins from the beads during deglycosylation, the eluted proteins were combined with the reaction mixture. After separation by SDS-PAGE, the immunoprecipitated and co-immunoprecipitated proteins were analyzed by Western blotting by using appropriate antibodies.
Western Blotting Analysis-1-10 g of proteins extracted from mouse heart membranes or 5-20 l of proteins eluted from the protein A-conjugated agarose beads were loaded onto cells grown in baffled spinner flasks showed Ͼ30% higher specific ouabain binding when compared with membranes from cells that did not undergo fed batch cultivation. The same increase was observed in Western blots of these membranes, indicating that most of the expressed protein was properly folded and functional.
Biochemical Assays-ATPase activity assays as well as titrations with NaCl, KCl, and vanadate were performed as described previously (29,35) using PiColorLock TM malachite green assay (Innova Bioscience). For ion titrations, activity was measured at varying concentrations of sodium plus choline chloride and KCl (80 mM), with constant total ionic strength (170 mM total). K 0.5 Na ϩ and K 0.5 K ϩ values were obtained by fitting the data to the Hill equation using KaleidaGraph (Synergy Software) (30).
Inhibition of Na,K-ATPase activity and [ 3 H]ouabain binding and K-[ 3 H]digoxin displacement assays were performed as reported (25). For derivation of K i values, percentage inhibition of Na,K-ATPase activity, VCG/V0, was calculated, and K i values were obtained by fitting the data to the function, VCG/ V0 ϭ K i /([CG] ϩ K i ) ϩ c. Inhibition was estimated in 3-7 separate experiments, and average K i values Ϯ S.E. were calculated (27).
Equilibrium and Stopped-flow Fluorescence Measurements Using RH421-Equilibrium fluorescence experiments were carried out in a Varian fluorimeter at room temperature. 10 g of purified Na,K-ATPase was added to 1 ml of 20 mM MOPS/ Tris, pH 7.2, 5 mM MgCl 2 , and 200 nM RH421. NaCl, ATP, and KCl where then added successively. NaCl and KCl titrations were performed as described previously (37) with emission and excitation wavelength set to 580 and 680 nm with a 5-nm slit width.
Stopped-flow measurements were made using an Applied Photophysics SX20 system (see Ref. 29). Using a combined xenon/mercury lamp, excitation wavelength was 577 nm, and fluorescence was measured at Ն665 nm using cut-off filters. Solutions were mixed 1:1 using ϳ120 l/syringe, and the temperature of the measurements was set to 23°C. All solutions were buffered to pH 7.2 using MOPS/Tris, and ionic strength was kept constant at 120 mM for all measurements using choline chloride.
For measurement of E2(2Rb)ATP 3 E1⅐3NaATP, syringe 1 contained 20 g/ml enzyme non-covalently labeled with 200 nM RH421 in 20 mM RbCl, 1 mM EDTA and was mixed with 80 mM NaCl, 2 mM ATP, 1 mM EDTA in syringe 2. For measurement of E1⅐3Na 3 E2P, 10 g/ml Na,K-ATPase non-covalently labeled with 200 nM RH421 in 100 mM NaCl and 4 mM MgCl 2 in syringe 1 was mixed with 1 mM ATP in the same solution in syringe 2.
Data were fitted using KaleidaGraph (Synergy Software). Ion titrations were fitted using the Hill equation, where n is the Hill coefficient, [ion] is the free concentration of the respective ion, and K1 ⁄ 2 is the concentration required to obtain the half-maximal fluorescence signal. Stopped-flow traces were fitted to a monoexponential function, F ϭ A ⅐ e Ϫk ⅐ t ϩ c (Eq. 2) or double exponential function, where A is the amplitude of the fluorescence signal, k is the rate of the reaction, and c is the equilibrium fluorescence level after the reaction is complete. All values are expressed as averages of 2-4 experiments.
Molecular Modeling-Homology modeling of human Na,K-ATPase ␣ 2 ␤ 1 , ␣ 2 ␤ 2 , and ␣ 2 ␤ 3 isoform complexes was carried out using the template ␣ 1 ␤ 1 with bound digoxin (Protein Data Bank code 4RET) as described previously (26). The final model exhibits the highest Profiles 3D score (72) and the lowest number of Ramachandran violations (73). Rather consistent profiles were observed for each model (␣ 2 ␤ 1 , ␣ 2 ␤ 2 , and ␣ 2 ␤ 3 ), demonstrating that the human Na,K-ATPase models were reasonable and could be employed for the further docking study. The magnesium ion and three structural waters were positioned in each final model. Models were eventually refined by energy minimization using the CHARMm force field (74).
Molecular Docking-Molecular docking of DcB to the different human Na,K-ATPase isoform models was carried out in Discovery Studio version 4.0 (Biovia, Dassault Systemes, San Diego, CA) with CDOCKER, which is an implementation of a CHARMm-based docking tool using a rigid receptor (75). The DcB ligand was prepared before docking using the Prepare Ligand module to evaluate ionization states for a given pH, isomers, and tautomers, correct bad valences, and generate three-dimensional conformations. The digoxin bioactive binding conformation was copied from the 4RET crystal structure and positioned into each model. The model binding site was defined as a sphere with radius that stays within 15 Å from the geometric centroid of the digoxin ligand using the Define and Edit Binding Site tool.
DcB was docked into the active site of each Na,K-ATPase human model. Different ligand orientations were generated, and for each orientation, the CHARMm energy (interaction energy plus ligand strain) and the interaction energy alone were calculated. The ligand orientations were sorted by CHARMM energy, and the top scoring (most negative, thus favorable to binding) orientations were retained. The final orientations selected were chosen based on their docking scores, which favor interactions with amino acids from the ␤ subunit.