Stabilization of the α2 Isoform of Na,K-ATPase by Mutations in a Phospholipid Binding Pocket*

Background: The α2 isoform of Na,K-ATPase is unstable compared with α1 and α3. Results: Mutations in TM8–10 strongly stabilize α2. A novel phospholipid antagonist selectively inactivates α2, and mutations in TM8–10 protect against inactivation. Conclusion: A phosphatidylserine binding pocket within TM8–10 has been identified. Significance: Mechanistic insights into α2 instability and a possible physiological role have been obtained. The α2 isoform of Na,K-ATPase plays a crucial role in Ca2+ handling, muscle contraction, and inotropic effects of cardiac glycosides. Thus, structural, functional, and pharmacological comparisons of α1, α2, and α3 are of great interest. In Pichia pastoris membranes expressing human α1β1, α2β1, and α3β1 isoforms, or using the purified isoform proteins, α2 is most easily inactivated by heating and detergent (α2 ≫ α3 > α1). We have examined an hypothesis that instability of α2 is caused by weak interactions with phosphatidylserine, which stabilizes the protein. Three residues, unique to α2, in trans-membrane segments M8 (Ala-920), M9 (Leu-955), and M10 (Val-981) were replaced by equivalent residues in α1, singly or together. Judged by the sensitivity of the purified proteins to heat, detergent, “affinity” for phosphatidylserine, and stabilization by FXYD1, the triple mutant (A920V/L955F/V981P, called α2VFP) has stability properties close to α1, although single mutants have only modest or insignificant effects. Functional differences between α1 and α2 are unaffected in α2VFP. A compound, 6-pentyl-2-pyrone, isolated from the marine fungus Trichoderma gamsii is a novel probe of specific phospholipid-protein interactions. 6-Pentyl-2-pyrone inactivates the isoforms in the order α2 ≫ α3 > α1, and α2VFP and FXYD1 protect the isoforms. In native rat heart sarcolemma membranes, which contain α1, α2, and α3 isoforms, a component attributable to α2 is the least stable. The data provide clear evidence for a specific phosphatidylserine binding pocket between M8, M9, and M10 and confirm that the instability of α2 is due to suboptimal interactions with phosphatidylserine. In physiological conditions, the instability of α2 may be important for its cellular regulatory functions.

The ␣2 isoform of Na,K-ATPase plays a crucial role in Ca 2؉ handling, muscle contraction, and inotropic effects of cardiac glycosides. Thus, structural, functional, and pharmacological comparisons of ␣1, ␣2, and ␣3 are of great interest. In Pichia pastoris membranes expressing human ␣1␤1, ␣2␤1, and ␣3␤1 isoforms, or using the purified isoform proteins, ␣2 is most easily inactivated by heating and detergent (␣2 > > ␣3 > ␣1). We have examined an hypothesis that instability of ␣2 is caused by weak interactions with phosphatidylserine, which stabilizes the protein. Three residues, unique to ␣2, in trans-membrane segments M8 (Ala-920), M9 (Leu-955), and M10 (Val-981) were replaced by equivalent residues in ␣1, singly or together. Judged by the sensitivity of the purified proteins to heat, detergent, "affinity" for phosphatidylserine, and stabilization by FXYD1, the triple mutant (A920V/L955F/V981P, called ␣2VFP) has stability properties close to ␣1, although single mutants have only modest or insignificant effects. Functional differences between ␣1 and ␣2 are unaffected in ␣2VFP. A compound, 6-pentyl-2-pyrone, isolated from the marine fungus Trichoderma gamsii is a novel probe of specific phospholipid-protein interactions. 6-Pentyl-2-pyrone inactivates the isoforms in the order ␣2 > > ␣3 > ␣1, and ␣2VFP and FXYD1 protect the isoforms. In native rat heart sarcolemma membranes, which contain ␣1, ␣2, and ␣3 isoforms, a component attributable to ␣2 is the least stable. The data provide clear evidence for a specific phosphatidylserine binding pocket between M8, M9, and M10 and confirm that the instability of ␣2 is due to suboptimal interactions with phosphatidylserine. In physiological conditions, the instability of ␣2 may be important for its cellular regulatory functions.
The Na,K-ATPase or Na,K-pump is an integral membrane protein that actively transports Na ϩ and K ϩ ions across the plasma membrane of virtually all animal cells. The Na,K-pump maintains the electrochemical potential gradients of Na ϩ and K ϩ ions, required for regulation of cell volume, electrical excitability, nutrients and neurotransmitter uptake, intracellular pH, and particularly relevant to this paper, cellular Ca 2ϩ handling (1,2).
The Na,K-ATPase is a heterodimer composed of the catalytic ␣ subunit that couples ATP hydrolysis and ion transport and the ␤ subunit that stabilizes the enzyme and permits trafficking to the cytoplasmic membrane (3,4). The Na,K-ATPase usually also contains an auxiliary subunit of the FXYD protein family (5)(6)(7). Recently, the structures of pig kidney Na,K-ATPase at 3.5 Å (8) and shark rectal gland Na,K-ATPase at 2.4 Å resolution (9) have been determined. These structures resolve most of the features of ␣ and ␤ subunits and FXYD subunits. In addition, lower resolution structures of the Na,K-ATPase with bound ouabain have become available (10,11).
Na,K-ATPase has four isoforms of the ␣ subunit (␣1-4) and three isoforms of the ␤ subunit (␤1-3), expressed and regulated in a tissue-and development-specific fashion, together with one of seven different mammalian FXYD regulatory proteins (1)(2)(3)(4)(5)(6)(7)12). ␣1 is almost ubiquitously distributed, whereas ␣2 is expressed strongly in muscle (skeletal, smooth, and cardiac) and astrocytes, ␣3 primarily in nerve cells, and ␣4 only in spermatozoa. Human heart expresses ␣1, ␣2, and ␣3 isoforms and ␤1 (13). In cardiac myocytes, ␣2 is concentrated in the T-tubules adjacent to the SR, 3 whereas ␣1 is more evenly distributed in T-tubules and SR (14). In smooth muscle and brain astrocytes, ␣1 is uniformly distributed but ␣2 and ␣3 show a punctated distribution, overlying the SR (15). ␣2 carries a large fraction of the Na,K-pump current in T-tubules of rat and mouse cardiac myocytes (16) and preferentially modulates Ca 2ϩ tran-sients and SR Ca 2ϩ release in mouse cardiac myocytes compared with ␣1 (17). Also, ␣2 is the major regulator of cardiac contractility in rat cardiomyocytes (18). In particular, ␣2 is functionally coupled to the Na/Ca exchanger and regulates Ca 2ϩ handling without changing the [Na] i (19). Compared with ␣1, ␣2 shows a higher K 0.5 K o (20 -22) and steeper voltage dependence (18,20), properties that allow a rapid change in rate with change in K o and membrane potential. This fits well with a specific regulatory role of ␣2 in muscle Ca 2ϩ handling. Similarly, there is good evidence that ␣2 plays an important role in contractility of vascular smooth muscle and control of blood pressure and regulation of Ca 2ϩ handling (23,24). Experiments with mice engineered to have ouabain-sensitive ␣1 and -insensitive ␣2 isoforms (␣1 S/S ␣2 R/R ) or ouabain-insensitive ␣1 and ␣2 (␣1 R/R ␣2 R/R ), instead of the wild-type with ouabain-insensitive ␣1 and ouabain-sensitive ␣2 (␣1 R/R ␣2 S/S ), have shown that ␣2 plays a predominant role in cardiac glycoside-induced positive inotropy (17,25) and also provide strong evidence for endogenous mammalian cardiac glycosides (25,26). ␣2 also plays the major role in cardiac glycoside-induced positive inotropy in rat cardiomyocytes (18). Recent observations that human ␣2 shows a moderate selectivity over ␣1 for some digitalis glycosides, such as digoxin, support the notion that cardiotonic effects of digoxin in humans are exerted via ␣2 (27).
Familial hemiplegic migraine is a rare but severe autosomal dominant subtype of migraine with aura (28). About 25% of familial hemiplegic migraine cases (type 2) are attributed to mutations to ␣2 (29,30). ␣2 is expressed in brain astrocytes, and most FHM2 ␣2 mutations cause functional defects in active Na ϩ and K ϩ transport, and compromised clearing of extracellular K ϩ or glutamate, with a tendency to cortical spreading depression (31).
In view of the above and much more information available in the literature, there is obviously a great interest in comparing the structural, functional, and pharmacological differences of the ␣1, ␣2, and ␣3 isoforms. We have described extensively the Pichia pastoris expression system for Na,K-ATPase isoforms and purification of ␣1␤1, ␣2␤1, and ␣3␤1 complexes (21,27,(32)(33)(34), as well as reconstitution, in vitro, of ␣␤FXYD complexes using purified FXYD proteins (21,35). A fundamental condition for maintenance of Na,K-ATPase activity of all the purified detergent-soluble ␣␤ or ␣␤FXYD complexes is the presence of an acidic phospholipid, especially phosphatidylserine (33,34). A requirement of detergent-soluble native Na,K-ATPase for an acidic phospholipid has been known for many years (36). Recent work shows that phosphatidylserine undergoes a specific interaction with the protein and protects against thermally mediated or detergent-mediated inactivation (33)(34)(35). Together with the phosphatidylserine, cholesterol also provides strong additional stabilization (34). One indication of a specific interaction of the phosphatidylserine with the protein is the structural selectivity of both headgroups and fatty acyl chains. Of the many phospholipids tested, 1-stearoyl-2-oleoylsn-glycero-3-phospho-L-serine (SOPS) is optimal (34). More recently, the FXYD proteins (FXYD1, FXYD2, and FXYD4) have been shown to strongly stabilize the ␣␤ complexes (21) by a mechanism involving amplification of the SOPS-protein interaction. This confirmed the specific nature of the SOPS-protein interaction and suggested two possible locations for an SOPS molecule, close to the FXYD protein, but no definitive answer on the location of an SOPS binding pocket was obtained (35).
In experiments to express and purify the ␣2 isoform of Na,K-ATPase, we came across a serious problem of instability (21). This problem has been noted previously in attempts to express ␣2 in Saccharomyces cerevisiae and is the cause of a very low expression level (37). This low expression is likely to compromise experiments to analyze cardiac glycoside binding properties of ␣2 in comparison with ␣1 and ␣3 (38). In the P. pastoris system, expression of ␣2 was significantly raised by reducing the growth temperature of the yeast to 20°C and selecting for high copy number clones. Sufficiently stable purified ␣2␤1 protein for biochemical work was obtained by washing and eluting complexes in the presence of SOPS/cholesterol, and the ␣2␤1 complex could be further stabilized by FXYD1 (21). Nevertheless, the ␣2␤1 or ␣2␤1FXYD1 complexes are still much more sensitive to thermally mediated and detergent-mediated inactivation than the ␣1␤1 or ␣1␤1FXYD1 complexes. Importantly, it was observed that the ␣2␤1 complex binds SOPS less well than ␣1␤1, and as discussed (21), this feature might be the key to explaining the increased thermally mediated and detergentmediated inactivation of ␣2.
This study addresses two questions, namely the origin of instability of ␣2 and the location of a specific SOPS binding pocket. We have exchanged residues in trans-membrane segments of ␣2, which could be responsible for the weaker interaction with SOPS, with the equivalent residues in ␣1. The result is that a stabilized ␣2␤1 complex and a clear-cut answer on the location of an SOPS binding pocket have been obtained.
Expression in P. pastoris and Purification of Na,K-ATPase Isoforms and Mutants-pHIL-D2 vector, containing Na,K-ATPase genes, was linearized by NotI digestion and used to transform spheroplasts of P. pastoris SMD1165, and His ϩ Mut s transformants were selected as described previously (32). Membrane preparations were made, and clones were screened for optimal protein expression by Western blotting using the anti-KETYY antibody. Large scale growth of isoforms was done in Bellco Spinner Flasks TM in 10-liter volumes of growth medium, as described previously (21,27). Expression of the Na,K-ATPase was induced by adding 0.5% methanol daily for 4 days or 42 hours (␣ 2 mutants) at 20°C. Cells were collected, washed, and broken with glass beads, and membranes were prepared as described previously (21,34). Procedures for solubilization of membranes in DDM followed by purification of the Na,K-ATPase enzymes have been described in detail (21,27,33,34). All isoform complexes were eluted from BD Talon beads in a solution containing 250 mM imidazole, 100 mM NaCl, 20 mM Tricine-HCl, pH 7.4, 0.1 mg/ml C 12 E 8 , 0.08 mg/ml SOPS, 0.01 mg/ml cholesterol, and 25% glycerol. The proteins were stored in aliquots in Ϫ80°C. In the experiment of Fig. 4, ␣1␤1, ␣2␤1, and ␣2VFP complexes were purified and eluted with different concentrations of SOPS (from 0 to 0.1 mg/ml). Protein complexes were analyzed by SDS-PAGE, and immunoblots were probed with anti-KETYY (anti-all ␣ isoforms).
Ouabain binding to yeast membranes using [ 3 H]ouabain was done as described in Ref. 40. For thermal inactivation experiments, yeast membranes were incubated at 45°C for indicated times, and aliquots of 200 g were moved to ice-cold tubes, before measurement of ouabain binding.
Na,K-ATPase Activity of Purified Isoform Complexes and Mutants, Thermal Detergent-mediated and 6-Pentyl-2-pyronemediated Inactivation-Purified enzymes were diluted with elution buffer to 0.05-0.15 g/l and then preincubated 10 min at 30°C. The standard reaction medium consisted of 400 l of 130 mM NaCl, 20 mM KCl, 3 mM MgCl 2 , 25 mM histidine, pH 7.4, 1 mM EGTA, 1 mM ATP with added lipids 0.005 mg/ml C 12 E 8 , 0.01 mg/ml SOPS, 0.001 mg/ml cholesterol. The reaction was started by addition of enzymes (0.1-0.2 g). Phosphate release was measured over 2, 4, and 6 min at 37°C, using a malachite green dye to detect the phosphate (P i Color Lock, Innova Biosciences). Na,K-ATPase activity was calculated from the slope of the time course by linear regression analysis (r Ͼ0.99). For measurements of K 0.5 Na ϩ and K 0.5 K ϩ , the Na ϩ concentration was varied with 150 mM (Na ϩ ϩ choline) and constant 20 mM KCl, or K ϩ was varied at constant 130 mM Na ϩ , respectively (21).
For thermal inactivation experiments, the different isoform complexes were diluted with the elution buffer to equal concentrations, 0.1-0.15 mg/ml, and then incubated at 37°C for the indicated times. Alternatively, C 12 E 8 or 6-pentyl-2-pyrone was added to the enzymes at the indicated concentrations and incubated for 5 min at 37°C. Aliquots of 0.02-0.1 g were removed at the indicated times to ice-cold tubes containing the standard reaction medium (ϪATP). The percent of control was calculated from specific activities obtained before and after heating or in the absence or presence of detergent.
Expression of FXYD1 in E. coli, Purification of FXYD1, and Reconstitution of ␣␤FXYD1 Complexes-Human FXYD1 was cloned into the pET28-TevH E. coli expression vector as described previously (35). Protein expression of FXYD1 was performed in CD41 E. coli cells. Cells were grown in LB medium containing kanamycin (50 g/ml) overnight at 37°C and then diluted 1:1000 in to 200 ml of medium and grown overnight. The next day, the culture was diluted into a 10-liter fermentor to an A 600 of 0.05 and then grown again at 37°C to an A 600 0.8. Induction was performed overnight with 0.5 mM IPTG, with gradual reduction of temperature to 20°C, and the cell density reached an A 600 of 5.6. The cells were harvested by centrifugation and washed twice with lysis buffer containing 50 mM Tricine/Tris, pH 7.4, 250 mM NaCl, 5 mM MgCl 2 , 1 mM EDTA, 10 g/ml DNase, and 0.1 mM of PMSF. The cells were disrupted by three passages through a high pressure emulsifier (Emulsiflex C5, AVESTIN). The lysate was harvested at 200,000 ϫ g for 90 min at 4°C. The pellet was resuspended in solubilization buffer containing 50 mM Tricine/Tris, pH 7.4, 250 mM NaCl, 10% glycerol, 4 mg/ml of n-dodecyl ␤-maltoside (final DDM/protein ratio was 2:1) and 8 M urea, followed by homogenization (glass-Teflon) at 25°C. The solution was diluted 2-fold with a buffer containing 50 mM Tricine/Tris, pH 7.4, 250 mM NaCl, 10% glycerol, and 20 mM imidazole and centrifuged at 200,000 ϫ g for 90 min at 4°C. The supernatant was incubated for 1 h with BD Talon beads (Co 2ϩ -chelate) at a cells/bead ratio of 1:4 (w/w). The beads were washed three times with a buffer containing 50 mM Tricine/Tris, pH 7.4, 250 mM NaCl, 10% glycerol, 2 mg/ml DDM, 4 M urea, and 40 mM imidazole. The FXYD1 was eluted with a 1-bead volume of buffer containing 50 mM Tricine/Tris, pH 7.4, 500 mM NaCl, 10% glycerol, 250 mM imidazole, 4 M urea, and 0.1 mg/ml C 12 E 8 . Urea was removed from purified FXYD1 by dialysis in gradual steps, using a buffer containing 50 mM Tricine/Tris, pH 7.4, 500 mM NaCl, 10% glycerol, and 3, 2, and 1 M urea, and finally a buffer without urea. After dialysis, the protein concentration was 0.15 mg/ml, and 0.1 mg/ml C 12 E 8 , 0.08 mg/ml SOPS, 0.01 mg/ml cholesterol, and glycerol to 25% were then added. The protein was separated into aliquots and stored at Ϫ80°C. For reconstitution of ␣␤FXYD complexes, pure FXYD1 was added to the ␣␤ complex at a molar ratio of 2.5:1-30:1 (FXYD/ Na,K-ATPase) and incubated 4 h on ice (35).
Preparation of Rat Heart Sarcolemma Microsomes-Sarcolemma microsomes were prepared as described previously for preparation of kidney plasma membranes (43). Briefly, left ventricular rat cardiac tissue was homogenized in ISE buffer (25 mM imidazole, pH 7.3, 1 mM EDTA, and 250 mM sucrose). The homogenate was centrifuged at 3500 ϫ g for 20 min. The super-natant from this centrifugation was centrifuged again at 6,000 ϫ g, and the resulting supernatant was centrifuged at 48,000 ϫ g for 1 h. The resulting pellets were homogenized in ISE buffer and represent the microsomal fraction. Microsomes were further purified by layering on a discontinuous sucrose density gradient (from 8.5 to 60%) and centrifuged for 3 h at 100,000 ϫ g. Bands of interest were aspirated, washed, suspended in ISE buffer, and kept at Ϫ80°C. The Na,K-ATPaserich fraction was further purified by incubation with SDS after titration to estimate the optimum SDS concentration producing maximum stimulation of ouabain-inhibitable ATP hydrolysis (44). Following several washings, the membranes were suspended in ISE buffer and kept at Ϫ80°C. The ouabain-inhibited Na,K-ATPase activity ranged between 50 and 100 mol⅐h Ϫ1 ⅐mg Ϫ1 at 37°C. Full details will be published elsewhere. 4 Immunoblots of the sarcolemma membranes were probed with the following isoform-specific antibodies: ␣1 monoclonal 464.6 (6H) (GeneTex); ␣2 polyclonal antibody raised synthetic peptide against 432-445 amino acids, (Upstate Cell Signaling Solutions 07-674), and ␣3 monoclonal raised against cardiac microsomes (Santa Cruz Biotechnology, sc Ϫ58631). The blots were calibrated using purified human ␣1␤1, ␣2␤1, and ␣3␤1 0.1-2 g per lane.
Thermally Mediated and Detergent-mediated Inactivation of Na,K-ATPase Activity of Rat Heart Sarcolemma-Rat sarcolemma membranes were diluted with buffer containing 10 mM Mops/Tris, pH 7.2, and 25% glycerol to 0.34 mg/ml. Aliquot (ϳ0.25 g) was diluted into 500 l of the standard reaction medium with varying concentrations of ouabain (from 0 to 5 mM) and incubated for 30 min at 37°C. For thermally mediated or detergent-mediated inactivation experiments, the sarcolemma membranes were diluted to 0.34 mg/ml and either incubated at 45°C for the indicated times or 200 M C 12 E 8 was added, and the suspension was incubated at 37°C for the indicated times. Aliquots of ϳ0.25 g were then added to the standard reaction medium containing 1 mM ATP and 0, 30 M, or 5 mM of ouabain and incubated 30 min at 37°C. Na,K-ATPase activity for ␣2 ϩ ␣3 was calculated as (activity control Ϫ activity 30 M ouabain), and activity of ␣1 was calculated as (activity 30 M ouabain Ϫ activity 5 mM ouabain).

RESULTS
Instability of ␣2-As described in the Introduction, when human ␣1 and ␣2 isoforms were expressed in P. pastoris and purified, it was found that ␣2 is much less stable than ␣1, an effect attributed to suboptimal interaction with SOPS (21). As a first step to explore this finding, we have compared the thermal inactivation of three human isoform complexes, ␣1␤1, ␣2␤1, and ␣3␤1. This was done in measurements of both ouabain binding to the intact yeast membranes (Fig. 1A) and Na,K-ATPase activity of the purified isoform proteins (Fig. 1B). In both types of assay, the pattern is similar, namely the ␣2 isoform is thermally inactivated much more quickly compared with ␣3 and ␣1. It is also obvious that the detergent-soluble purified isoforms are all inactivated more quickly than the membranebound proteins. Nevertheless, the rank order of thermal inactivation, ␣2 Ͼ Ͼ ␣3 Ͼ ␣1, is the same for both membrane-bound and detergent-soluble proteins. Thus, either ouabain binding on membranes or Na,K-ATPase activity of the purified proteins can be used to assess loss of active pump units upon heating, i.e. thermal stability.
It is worth pointing out that the yeasts expressing ␣1, ␣2, and ␣3, used in Fig. 1, were all grown at 20°C, which is optimal for expression of ␣2, whereas ␣1 is expressed optimally at 25°C (21). It turns out that the growth temperature affects both expression levels and stability of ␣2 (see supplemental Fig. 1). Thus, comparisons of stability of ouabain binding to isoforms, such as in Fig. 1, must be done with membranes prepared from cells grown at the same temperature.
Design and Expression of Stabilizing Mutants of ␣2-Knowing that (a) the order of thermal stability is ␣1 Ͼ ␣3 Ͼ Ͼ ␣2, (b) that ␣2 binds SOPS less well than ␣1 (21), and (c) that the protein is destabilized by a displacement of bound phosphatidylserine by the detergent (34,35), one could propose that instability of ␣2 is associated with suboptimal binding of SOPS to specific trans-membrane residues, leading to easy displacement of the SOPS by the detergent. The test of this hypothesis is to identify residues in trans-membrane segments that are unique to ␣2, mutate them to the equivalent residues in ␣1, 4 Y. Mahmmoud, manuscript in preparation. FIGURE 1. Thermal inactivation of ␣1␤1, ␣2␤1, and ␣3␤1 in P. pastoris membranes and purified isoform complexes. A, membranes expressing ␣1␤1, ␣2␤1, and ␣3␤1 were heated at 45°C for the indicated times and then ouabain binding was measured. B, purified complexes ␣1␤1, ␣2␤1, and ␣3␤1 were diluted to the same concentration and heated at 37°C for the indicated times, prior to measurement of Na,K-ATPase activity. These data represent the average of two experiments. The ouabain binding capacities in A were ␣1␤1 25-30 pmol/mg protein, ␣2␤1 20 -25 pmol/mg protein, and ␣3␤1 20 pmol/mg protein, and Na,K-ATPase activities in B were in the range ␣1␤1 12-15 mol of P i /mg/min, ␣2␤1 6 -8 mol of P i /mg/min, and ␣3␤1 Ϸ10 mol of P i /mg/min, respectively. DECEMBER  singly or together, with the prediction that the changes should stabilize the altered ␣2 isoform. Inspection of human ␣1, ␣2, and ␣3 sequences reveals that there are only a few candidate residues unique to ␣2, including four in M8, M9, and M10, one in L9 -10, and one each in M2 and M4. These are collected in Table 1, together with the mutants prepared and the names used for this paper. Apart from these residues, the trans-membrane segments of ␣1, ␣2, and ␣3 are fully conserved, except for a few differences that are not unique to ␣2 and are therefore considered irrelevant to the differential instability of ␣2. The mutations studied here have focused on the residues located in TM8, TM9, and TM10, as detailed in Table 1, because they are all located in a pocket, which could bind a phospholipid. 5 This concept was proposed previously (21). Fig. 2 presents a Western blot and ouabain binding capacities of the yeast membranes expressing the wild-type proteins ␣1␤1, ␣2␤1, and all the mutants ␣2F␤1, ␣2V␤1,␣2P␤1, ␣2VF␤1, and ␣2VFP␤1. The triple mutant ␣2PVR␤1 was expressed only at very low levels and was not studied further. The wild-type ␣1␤1, ␣2␤1, and all the mutants were purified as described under "Experimental Procedures" and our recent papers (21,27). A representative purification of ␣1␤1, ␣2␤1, ␣2F␤1, and ␣2VFP␤1 is shown in supplemental Fig. 2, and purification of the other mutants ␣2V␤1, ␣2P␤1, and ␣2VF␤1 was essentially identical. The specific Na,K-ATPase activities of the different purified proteins were as follows: ␣1, 15-25; ␣2, 10 -15; ␣2V, 10 -15; ␣2F, 12-16; ␣2P, 8 -9; ␣2VF, 16; ␣2VFP, 10 -12 mol/min/mg. Stability, SOPS "Affinity," and FXYD1 Interactions of ␣2 Mutants-Thermal stability of the purified ␣1, ␣2 and ␣2F, ␣2V,␣2P, ␣2VF, and ␣2VFP mutants was compared at 37°C (Fig. 3A). Values of half-times calculated from fits to exponential decay curves are given in the legend. It is striking that ␣2 is much less stable than ␣1, whereas the triple mutant ␣2VFP was much more stable than ␣2 and only slightly less stable than ␣1.

Stabilization of the ␣2 Isoform of Na/K-ATPase
The stability of single mutant ␣2V and double mutant ␣2VF was also greater than that of ␣2 but lower than the ␣2VFP mutant, whereas the stability of the single mutants ␣2F and ␣2P

Stabilization of the ␣2 Isoform of Na/K-ATPase
was similar to that of ␣2. Inactivation by excess detergent, C 12 E 8 , showed essentially the same features (Fig. 3B) except that, in this case, the single mutants ␣2F and ␣2P were also a little more resistant to C 12 E 8 than ␣2 itself. The most striking feature in both types of experiments is that there is a cooperative effect of the three mutations in stabilizing the ␣2, whereas single mutations have smaller or only minor effects (see "Discussion" and Fig. 11 for a possible interpretation). Instability of ␣2 compared with ␣1 is associated with a lower "apparent affinity" for the SOPS (ϳ4-fold) (21). Accordingly, the stabilizing effect of the ␣2VFP mutant could be predicted to increase the apparent affinity for SOPS compared with ␣2. This feature was shown clearly in a series of experiments that examined the specific Na,K-ATPase activities of several preparations of ␣1, ␣2, and ␣2VFP at increasing concentrations of SOPS (Fig.  4). Because the difference between ␣2 and ␣1 is itself relatively small, it was necessary to measure the effect of increasing SOPS concentrations in at least three separate experiments, with each point representing the average value of (v/v) 0.1 SOPS Ϯ S.E. The important point is that the curve for ␣2VFP clearly lies above that for ␣2 and below that for ␣1. The other mutants were not examined in this type of experiment, because it would be very difficult to detect the partial effects of ␣2V and ␣2VF.
Another aspect of stabilization concerns the effects of FXYD1 on ␣1, ␣2, and the mutants (Fig. 5 and Table 2). As described recently, purified FXYD1 and other FXYD proteins, expressed in E. coli, stabilize the ␣1␤1 complexes by amplifying the protein-SOPS interaction (35). Both ␣1␤1 and ␣2␤1 were also found previously to be stabilized by unpurified FXYD1 expressed in P. pastoris (21). Thus, mutants of ␣2 that are predicted to affect protein-SOPS interaction could also affect the protein-FXYD1 interaction. The stabilizing effect of FXYD1 was assessed by reconstituting the purified ␣1␤1, ␣2␤1, and ␣2VFP␤1 and ␣2F␤1 complexes with purified FXYD1 in solution, to yield the ␣␤FXYD1 complexes, see "Experimental Pro-cedures." The purified FXYD1 was added to the purified ␣␤ complexes at increasing molar ratios, and after 4 h on ice, the complexes were incubated for 5 min at 37°C in the presence of 500 M C 12 E 8 , and Na,K-ATPase activity was then measured. FXYD1 protected against the C 12 E 8 , and from the hyperbolic curves in Fig. 5a the maximal degree of protection (v max / v(ϪFXYD1)), and the K 0.5 for FXYD1 were obtained by fitting the data as described in the legend to Table 2, and are collected together in Table 2. The normalized curves in Fig. 5b allow easy visualization of the different apparent affinities of the FXYD1. In these conditions there are two significant effects that distinguish between ␣1, ␣2, and ␣2VFP. First, the apparent affinity of ␣1 for FXYD1 is higher than for ␣2 and that for ␣2VFP is similar to ␣1. Second, the degree of stabilization of ␣2 by FXYD1 is significantly higher than for ␣1 and again ␣2VFP is closer to ␣1. The latter feature may arise because the initial degree of inactivation of ␣2 is greater than for either ␣1 or ␣2VFP, but in any case, the experiments show that the interaction of FXYD1 with ␣2VFP is more similar to that of ␣1 than ␣2. We have also looked at the effects of FXYD1 on the single mutant ␣2F  . Protection of ␣1␤1, ␣2␤1, ␣2VFP␤1, and ␣2F␤1 by FXYD1 against detergent-mediated inactivation. ␣1␤1, ␣2␤1, ␣2VFP␤1, and ␣2F␤1 isoform complexes were diluted to 0.11 g/l and incubated on ice for 4 h with FXYD1 at 0 -30:1 molar ratio (FXYD1/Na,K-ATPase). C 12 E 8 0.5 mM was added, and after incubation for 5 min at 37°C, Na,K-ATPase was measured. Rates of ATP hydrolysis at different (FXYD1/Na,K-ATPase) ratios were fitted to the function v ϭ ((V max ⅐X)/(X ϩ K 0.5 )) ϩ V(ϪFXYD1). X-(FXYD1/Na,K-ATPase) and V(ϪFXYD1) is the rate in the absence of FXYD1. The upper plot represents v/v(ϪFXYD1) versus (FXYD1/Na,K-ATPase), and the lower plot normalized function represents v/V max versus (FXYD1/Na,K-ATPase).
because, in the molecular structures, the phenylalanine residue makes close contact with the FXYD protein, and it might be thought that it would affect the FXYD1 interaction on its own. In fact, the data of Fig. 5 show that the affinity of the ␣2F mutant for FXYD1 is similar or lower than that of ␣2, and the degree of protection by FXYD1 against the detergent is significantly lower than that of ␣1, ␣2, or the triple mutant ␣2VFP. In other words, the difference between ␣2VFP and ␣2 results from a collaborative effect of the three mutations, as found also in the experiments without FXYD1. The absolute thermal stability of the purified ␣␤FXYD1 complexes, measured at 37 and 45°C, also shows clearly that the order is ␣1␤1FXYD1 Ͼ ␣2VFP␤1FXYD1 Ͼ ␣2␤1FXYD1 (data not shown).
A Phospholipid Antagonist-A novel way to study proteinlipid interactions and the mechanistic origin of instability of ␣2 is described in Figs. 6 and 7. In a search for natural compounds with selectivity for ␣2 compared with ␣1 and ␣3 isoforms, we have screened organic solvent extracts from marine organisms (see "Experimental Procedures"). One extract (from a fungus, Trichoderma gamsii strain NF-26) was found to selectively inactivate ouabain binding to P. pastoris membranes expressing ␣2 by comparison with membranes expressing ␣1 and ␣3 (Fig.   6). This extract was subjected to a series of purification steps, and eventually, the active principle was isolated and found to be a hydrophobic aromatic ␦-lactone with the structure shown in Fig. 6. The compound, 6-pentyl-2-pyrone (also known as 6-amyl-␣-pyrone), is well known and has been used as an artificial flavor in the food industry. The selectivity sequence for inactivating the Na,K-pump isoforms seen in Fig. 6 ␣2 Ͼ ␣3 Ͼ ␣1 suggested that it might act as an antagonist of the protein-SOPS interaction. To test this hypothesis, we have looked at inactivation of the purified ␣1␤1, ␣2␤1, and ␣2VFP␤1 complexes, by commercially available pure 6-pentyl-2-pyrone. The data in Fig. 7A show clearly that 6-pentyl-2-pyrone, at concentrations in the millimolar range, strongly inactivates Na,K-ATPase activity of ␣2␤1, and ␣1␤1 is insensitive while the triple mutant ␣2VFP is partially inactivated. The protection of ␣2VFP against this compound is similar to the effects observed for this mutant in thermally mediated and detergent-mediated inactivation experiments and dependence of activity on SOPS concentra- FIGURE 7. Inactivation of purified ␣1␤1, ␣2␤1 ␣3␤1, and ␣2VFP␤1 complexes by 6-pentyl-2-pyrone. FXYD1 protects ␣2 against 6-pentyl-2-pyrone. The enzymes were incubated with 6-pentyl-2-pyrone as indicated under "Experimental Procedures," and the Na,K-ATPase activity was then measured.

TABLE 2
Stabilizing effects of FXYD1 on ␣1␤1, ␣2␤1, ␣2VFP␤1, and ␣2F␤1 K 0.5 FXYD1 was obtained by fitting rates of ATP hydrolysis at different (FXYD1/Na,K-ATPase) ratios to the function v ϭ ((V max ⅐X)/(X ϩ K 0.5 )) ϩ V(ϪFXYD1). X is (FXYD1/Na,K-ATPase) and V(ϪFXYD1) is the rate in the absence of FXYD1. The degree of protection is calculated as the ratio V max /V(ϪFXYD1) ϮS.E. S.E. was calculated as tions (seen in Figs. 3 and 4). Therefore, the conclusion is that the compound interferes with the specific SOPS-protein interaction, within the M8, M9, M10 pocket, and preferentially inac-tivates ␣2 because the SOPS-␣2 interaction is weaker than the SOPS-␣3 and SOPS-␣1 interaction. In other experiments, it was also found that FXYD1 significantly protects against the 6-pentyl-2-pyrone (Fig. 7B). Because the FXYD1 stabilizes the SOPS in its binding pocket (35), these data also indicate that 6-pentyl-2-pyrone acts as an antagonist of the SOPS in the binding pocket. Fig. 8 and Table 3 together collect the functional properties of the ␣2VFP mutant in comparison with ␣1 and ␣2 and show that the triple mutant behaves almost exactly like ␣2. This is particularly striking in the case of inhibition of Na,K-ATPase activity by digoxin, which displays a 3-4-fold preference for ␣2/␣1, as reported recently (27), and an almost identical preference for ␣2VFP/␣1. Similarly, small but significant differences in K 0.5 K ϩ and K 0.5 Na ϩ between ␣2 and ␣1, which have also been described previously (20 -22), are preserved between ␣2VFP and ␣1. In short, the large differences in stability of ␣2 and ␣2VFP, described above, are not paralleled by any detectable differences in functional properties. This is consistent with localized and highly specific effects of the mutations on phospholipid-protein interactions predicted above.
Stability of ␣2 in Rat Heart Membranes-The instability of ␣2 expressed in the yeast membrane and in the purified ␣2␤1 complexes raises a question whether ␣2 expressed in a native membrane shows the same stability characteristics. To address this question, we have utilized a partially purified preparation of Na,K-ATPase prepared from rat heart sarcolemma. The specific Na,K-ATPase activity was ϳ50 mol/mg/h. Fig. 9a assessed the isoform content of these membranes using ␣1-, ␣2-, and ␣3-specific antibodies in immunoblots, calibrated with known amounts of the purified human ␣1␤1, ␣2␤1, and ␣3␤1 complexes. This showed that the heart enzyme contains all three isoforms in the following fractions: ␣1, 0.77; ␣2, 0.14; and ␣3, 0.089. In Na,K-ATPase assays, the activity of the rodent ␣1, with very low affinity for ouabain, can be easily distinguished from the activity of ␣2 and ␣3, which have high affinity for ouabain. Fig. 9b shows that about 30% of the ATPase activity was inhibited at ouabain concentrations, 30 -50 M, representing the activity of ␣2 plus ␣3, whereas a further 50% was inhibited at 4 -7.5 mM ouabain, representing the activity of ␣1. Over- FIGURE 8. Inhibition of Na,K-ATPase activity ␣1␤1, ␣2␤1, and ␣2VFP␤1 by digoxin. ␣1␤1, ␣2␤1, and ␣2 VFP ␤1 were incubated at 25°C for 30 min for activation. The enzyme (Ϸ0.08 -0.2 g of protein) was added to 400 l of the reaction medium containing the following: 1 mM ATP, NaCl; 130 mM, KCl; 5 mM, MgCl 2 ; 3 mM, histidine; 25 mM, pH 7.4, EGTA; 1 mM, SOPS 0.01 mg/ml; cholesterol 0.001 mg/ml; C 12 E 8 0.005 mg/ml; the indicated concentrations of digoxin, and were incubated for 1 h at 37°C. Na,K-ATPase activity was measured as described under "Experimental Procedures." FIGURE 9. Isoform content and Na,K-ATPase activity of ␣2/␣3 versus ␣1 isoforms in rat heart sarcolemma membranes. a, immunoblots using isoformspecific antibodies of ␣1, ␣2, and ␣3 to determine isoform content of rat heart sarcolemma membranes. Calibration of the signals used was 0.05-2 g of purified human ␣1␤1, ␣2␤1, and ␣3␤1 complexes. The calculated amounts of the isoforms are given in micrograms. b, ATPase activity of rat heart sarcolemma membranes measured in the standard reaction medium containing 0, 30 M, or 5 mM ouabain. Na,K-ATPase attributable to ␣2 ϩ ␣3 and ␣1 were calculated from control activity Ϫ activity at 30 M and activity at 30 M Ϫ activity at 5 mM ouabain, respectively. The data represent one of several similar experiments.

TABLE 3
Functional properties of ␣1␤1, ␣2␤1, and ␣2VFP␤1 K 0.5 K ϩ and K 0.5 Na ϩ values were obtained by measuring Na,K-ATPase activity at varying [K ϩ ] or [Na ϩ ], as described under "Experimental Procedures," and fitting the data to the function v ϭ V 0 ϩ V max ⅐X n /(K n ϩ X n ), where K n ϭ K 0.5 K ϩ or K 0.5 Na ϩ . For different digoxin concentrations in Fig. 8, the percent inhibition v CG /v 0 was calculated, and K i digoxin was obtained by fitting the data to the function v CG /v 0 ϭ K i /([CG] ϩ K i ) ϩ c, where CG indicates cardiac glycoside. The constant c represents a small fraction of Na,K-ATPase activity that is not inhibited (Ͻ5%). all about 80% of the activity was inhibited at 5 mM ouabain. In the experiments of Fig. 10, the membranes were either heated at 45°C for several hours or incubated over 90 min with 200 M C 12 E 8, as described under "Experimental Procedures," and ATP hydrolysis was then measured without ouabain, with 30 M ouabain, or with 5 mM ouabain. The activity of ␣2 ϩ ␣3 was measured as the control activity Ϫ activity at 30 M ouabain and that of ␣1 by activity at 30 M activity Ϫ activity at 5 mM ouabain. Evidently, a significant proportion of the ␣2 plus ␣3 was inactivated more rapidly compared with ␣1. The differences are significant and quite reproducible, but are smaller than observed between ␣1 and ␣2 in the yeast membranes. This is not surprising because the proportion of ␣2/␣3 is about 60:40 (Fig. 9a), and a significant fraction of the activity inhibited at 30 M ouabain is contributed by ␣3, which is closer in stability to ␣1 than to ␣2 (see Fig. 1). Another possible factor is that a small percentage of ␣1 may also be inhibited by 30 M ouabain, because the reported K i value for inhibition of the all ␣1 rat kidney Na,K-ATPase is reported to be 100 -150 M (45). Nevertheless, the finding that a significant fraction of activity, attributed to ␣2 ϩ ␣3, is more sensitive to thermal inactivation and detergent than activity ascribed solely to ␣1 provides a strong indication that the ␣2 in the rat heart membranes is a less stable protein than ␣1.

DISCUSSION
Binding Pocket for Phosphatidylserine, the Origin of Instability of ␣2- Fig. 11 highlights trans-membrane segments M8, M9, and M10 of the ␣ subunit and the FXYD protein in the structure of Na,K-ATPase, with the detail depicting the residues in ␣2 (Fig. 11B), which have been mutated to the equivalent residues of ␣1 and other residues of interest (Fig. 11C). The mutations, ␣2A920V in M8, L955F in M9, and V981P at the end of M10, were predicted, singly or together, to strengthen the interaction of ␣2 with SOPS and to stabilize the protein. Our findings indeed provide a clear indication that 1) the SOPS binds specifically within a pocket created by M8, M9, and M10, in proximity to the FXYD protein; 2) a weak interaction of SOPS in this pocket is the source of the instability of ␣2, compared with ␣1 and ␣3; and 3) the triple mutation ␣2VFP strongly stabilizes the protein. The detailed evidence is presented below. FIGURE 10. Thermally mediated and detergent-mediated inactivation of Na,K-ATPase isoforms in rat heart sarcolemma membranes. Rat heart sarcolemma membranes were heated at 45°C (a) or incubated at 37°C with 200 M C 12 E 8 (b) for the indicated times, and Na,K-ATPase activities of ␣2 ϩ ␣3 or ␣1 isoforms were then measured as described in Fig. 9. The data represent the average Ϯ S.E. of three separate experiments. Evidence for the SOPS Binding Pocket-The triple mutant ␣2VFP is strongly protected against thermally mediated and detergent-mediated inactivation, with stability closer to ␣1 than to ␣2 (Figs. 3 and 4). Similarly, the "affinity " of ␣2VFP for SOPS lies between that of ␣1 and ␣2. These features are consistent with the prediction that the residues (Ala-920, Leu-955, and Val-981) facing the lipid bilayer in M8, M9, and M10 are responsible for the weak interaction of ␣2 with SOPS, and instability of ␣2 compared with ␣1 (or ␣3). In contrast to the triple mutant, effects of the single mutants, where tested on stability, were less pronounced or even insignificant. The implication is that there are cooperative effects of the triple mutant in binding SOPS, as could be imagined for binding of the head group and two fatty acyl chains. Note in Fig. 11 that there are also two lysine residues, Lys-944 and Lys-946, at the entrance to M9 and membrane-water interface that are well placed to interact with the negatively charged phosphate and serine carboxylate of the headgroup. These conclusions do not exclude the possibility that there are additional specific phospholipid-binding sites. However, they do suggest that SOPS binding in the pocket between M8, M9, and M10 is of major importance for stabilizing the protein. Fig. 11 also depicts a second proline residue, Pro-978, in the loop between M9 and M10, adjacent to Pro-981. An ␣2 P978L mutation has been reported in FHM2. The interest of this mutant is that the protein is expressed in mammalian cells at 28°C but not at 37°C, indicating that it is unstable and is degraded at 37°C (47). By contrast, the many other ␣2 FHM2 mutants are expressed at 37°C but inactivate pump function in various ways (30,48,49). The location of the P978L, and its property of destabilizing the protein, fits very well with the other evidence presented here for the specific SOPS-FXYD-␣ stabilizing interactions.
The finding that ␣2VFP shows stability properties closer to ␣1 but functional and inhibition properties identical to ␣2 ( Fig.  8 and Table 3) indicates that the mutations have a strong local effect but no general effects on the protein folding. This is entirely compatible with the notion of a selective SOPS binding pocket. These features also suggest that the stabilized ␣2VFP mutant could become an important experimental tool, e.g. for structural work or development of ␣2-selective inhibitors (27).
It is a significant observation that the order of stability of the isoforms in the intact yeast membrane is the same as for the purified detergent-soluble proteins, i.e. ␣1 Ͼ ␣3 Ͼ ␣2 (Fig. 1). By contrast, the thermal stability of all three isoforms in the intact membrane is much higher than that of the purified proteins, due presumably to protection by the more ordered phospholipid bilayer, compared with the mobile mixed detergentprotein-lipid micelle. Thus, the relative stabilities of the isoforms, ␣1 Ͼ ␣3 Ͼ ␣2, are not explained by the different environments of the bilayer versus micelles. With the understanding obtained from the purified proteins, it is natural to propose that the instability of ␣2 in the intact membranes arises from suboptimal specific interactions with phosphatidylserine, which are easily disrupted by heat or detergents, leading to irreversible inactivation of the protein. Similarly, the lower stability of ␣2 compared with the ␣1 in the rat heart membrane (Fig. 10) might also reflect different specific phosphatidylserine-protein interactions.
Effects of FXYD1-The effects of FXYD1 seen in Fig. 5 and Table 2 fit in well with the other recently described stabilizing effects of FXYD proteins. As mentioned, FXYD1 expressed in P. pastoris was found to stabilize both ␣1␤1 and ␣2␤1 (21), and FXYD1, FXYD2 and FXYD4, expressed in E. coli and then purified, all stabilized ␣1␤1 with the order FXYD1 Ͼ FXYD2 Ͼ FXYD4 (35). The mechanism of stabilization appears similar for all the FXYD proteins and was worked out in greatest detail for FXYD1. It involves amplification of the specific SOPS-protein interaction as detected by a higher affinity for SOPS, protection against thermally mediated and detergent-mediated inactivation and detergent-SOPS competition, and also protection against inactivating effects of a specific phosphatidylserine decarboxylase. The FXYD proteins were suggested to interact with both the SOPS and the ␣ subunit to stabilize the proteins. Two possible locations of the SOPS were proposed, either between M8, M9, and M10 or between M2, M6, and M9, in proximity to the trans-membrane segment of the FXYD protein (35). In the present series of experiments, the stabilizing effects of FXYD1 were characterized by the following: (a) the lower apparent affinity for FXYD1 and higher maximal degree of stabilization for ␣2 compared with ␣1; (b) behavior of ␣2VFP almost identical to ␣1, and (c) affinity for FXYD1 of ␣2F similar to ␣2 and a lower maximal degree of stabilization. It is remarkable that the ␣2VFP triple mutant largely mimics ␣1 in the binding and stabilization properties of FXYD1, whereas the single mutant ␣2F binds FXYD1 like ␣2 and is even less stabilized. This behavior is similar to the stabilizing effects of triple versus single mutants, without the FXYD1, and strongly supports the concept that the FXYD1 interacts with the SOPS in the pocket between M8, M9, and M10, stabilizing the SOPS-protein interaction. The affinity of FXYD1 for ␣2 may be lower than for ␣1 because ␣2 binds SOPS less well. Previously, co-immunoprecipitation experiments suggested that FXYD1 associates less well with ␣2 than with ␣1 (21,46). Thus, the present reconstitution experiments confirm that conclusion directly and show that the mechanism involves the phosphatidylserine-protein interaction.
Phospholipid Antagonist-The inference that the 6-pentyl-2pyrone behaves as a "phospholipid antagonist" is based on the following observations: (a) this compound inactivates the isoforms in both the intact membrane and purified proteins with the order ␣2 Ͼ ␣3 Ͼ ␣1, similar to the effect of heating or excess detergent; (b) both the triple mutant ␣2VFP and reconstitution of the ␣␤FXYD1 complex protect the protein against the 6-pentyl-2-pyrone, similar, again, to their effects against heating and excess detergent (Figs. 6 and 7). Because both the ␣2VFP triple mutant and FXYD1 stabilize the specific SOPSprotein interaction, protection against inactivation by 6-pentyl-2-pyrone shows that the 6-pentyl-2-pyrone antagonizes the SOPS binding and provides independent confirmation for the specific SOPS binding pocket. Obviously, phospholipid antagonists such as 6-pentyl-2-pyrone, or similar compounds yet to be discovered, may become useful tools to analyze specific phospholipid-protein interactions in Na,K-ATPase or other membrane proteins.