Crystal structure of the high-affinity Na⁺,K⁺-ATPase–ouabain complex with Mg²⁺ bound in the cation binding site

Abstract

The Na⁺,K⁺-ATPase maintains electrochemical gradients for Na⁺ and K⁺ that are critical for animal cells. Cardiotonic steroids (CTSs), widely used in the clinic and recently assigned a role as endogenous regulators of intracellular processes, are highly specific inhibitors of the Na⁺,K⁺-ATPase. Here we describe a crystal structure of the phosphorylated pig kidney Na⁺,K⁺-ATPase in complex with the CTS representative ouabain, extending to 3.4 Å resolution. The structure provides key details on CTS binding, revealing an extensive hydrogen bonding network formed by the β-surface of the steroid core of ouabain and the side chains of αM1, αM2, and αM6. Furthermore, the structure reveals that cation transport site II is occupied by Mg²⁺, and crystallographic studies indicate that Rb⁺ and Mn²⁺, but not Na⁺, bind to this site. Comparison with the low-affinity [K₂]E2–MgF_x–ouabain structure [Ogawa et al. (2009) Proc Natl Acad Sci USA 106(33):13742–13747) shows that the CTS binding pocket of [Mg]E2P allows deep ouabain binding with possible long-range interactions between its polarized five-membered lactone ring and the Mg²⁺. K⁺ binding at the same site unwinds a turn of αM4, dragging residues Ile318–Val325 toward the cation site and thereby hindering deep ouabain binding. Thus, the structural data establish a basis for the interpretation of the biochemical evidence pointing at direct K⁺–Mg²⁺ competition and explain the well-known antagonistic effect of K⁺ on CTS binding.

Na⁺,K⁺-ATPase activity in the cell is defined as ouabain-sensitive ATPase, emphasizing high specificity of cardiotonic steroids (CTSs) toward the enzyme. The minimal functional unit of the Na⁺,K⁺-ATPase (1) is a heteromeric complex consisting of a large α-subunit, which is responsible for ATP hydrolysis, ion transport, and CTS binding, and a β-subunit, acting as a chaperone. A regulatory protein from the FXYD-family (denoted γ for the renal pump) often is associated with the enzyme. The ion transport of the Na⁺,K⁺-ATPase is accomplished by cycling between two major conformational states, E1 and E2 (via phosphoenzyme intermediates), which selectively bind three Na⁺ ions or two K⁺ ions, respectively (2, 3).

Structural features common for CTSs include a cis-trans-cis ring fused steroid core, which adopts a U-shaped conformation with a convex β-surface, a hydroxyl group at C14 (OH14_β), and a five- or six-membered lactone ring in a β-conformation at position C17. Some CTSs have a carbohydrate moiety of one to four residues attached to C3. Ouabain, one of the most hydrophilic CTSs, is composed of a steroid core with four hydroxyl groups at the β-surface, a hydroxyl group at the α-surface, an unsaturated five-membered lactone ring, and a rhamnose carbohydrate moiety (Fig. 1B).

Fig. 1.

The high-affinity Na⁺,K⁺-ATPase E2P–ouabain complex. The E2P–ouabain complex is depicted using the following color codes: red [nucleotide binding (N) domain], blue [phosphorylation (P) domain], light yellow [actuator (A) domain], orange (αM1–2), teal (αM3–4), purple (αM5–6), brown (αM7–10), green (β-subunit), and deep pink (γ-subunit). The CTS binding site is marked CTS. Ouabain is represented by yellow sticks, Mg²⁺ and the three water molecules by light green and red spheres, respectively. The cholesterol–phospholipid site is marked CLR, with cholesterol and phosphatidylserine depicted in orange and in black sticks, respectively. The carbohydrate moieties are depicted in light pink sticks, confirming the three glycosylation sites at Asn158, Asn193, and Asn265 at the β-subunit. (A) Cartoon representation of the high-affinity Na⁺,K⁺-ATPase E2P–ouabain complex. (B) Structural representation of ouabain, illustrating charge relocation in the lactone ring. (C) Well-defined electron density map of the occupied ouabain and cation binding sites. The initial F_o–F_c map obtained after initial rigid body refinement, contoured at 5.0 σ, is depicted in green mesh, and the 2F_o–F_c map, contoured at 1.5 σ, is depicted in gray mesh. (D) Structural alignment of the high-affinity [Mg]E2P–ouabain complex and the low-affinity [K₂]E2–MgF_x–ouabain complex (11). The [K₂]E2–MgF_x–ouabain complex in gray cartoon indicates the location of two K⁺ (dotted spheres). Residues lining ion site II are shown in sticks. (E) Surface representation (hydrophobic moieties in gray, positively charged atoms in blue, negatively charged atoms in red) of the cation and ouabain binding sites of the high-affinity ouabain complex with Mg²⁺-bound at site II (Left) and the K⁺-occluded low-affinity ouabain complex (11) (Right), visualized from αM1–2. The map has been intersected along αM4 and αM6. αM4 (depicted in cartoon) subsequently was superimposed onto the figure to demonstrate the spatial rearrangement of the CTS binding site as a consequence of the winding/unwinding of αM4. The interior of the protein is depicted in dark gray.

Measurable CTS binding to the Na⁺,K⁺-ATPase occurs in the presence of Mg²⁺, whereas high-affinity binding (K_d values in the nanomolar range) requires autophosphorylation of the Na⁺,K⁺-ATPase. In contrast, the presence of monovalent cations in the cation transport sites, conventionally denoted I–III, inhibits CTS binding to the enzyme. Ouabain–K⁺ antagonism is a well-known and well-described phenomenon in the literature.

The crystal structure of the α₁β₁γ complex of the pig kidney Na⁺,K⁺-ATPase in the P_i-induced E2P form stabilized by ouabain was described earlier (4). However, key details on high-affinity protein–ligand interactions remained concealed at 4.6 Å resolution. Furthermore, the impact of structural components of CTSs, such as glycosylation, substituents on the steroid core, and five- vs. six-membered lactone rings, on the affinity to Na⁺,K⁺-ATPase was unclear. Given the benefits and further prospects of therapeutic application of CTSs and their multiple effects observed in cell cultures (5, 6), the need for a higher-resolution structure was obvious. Detailed screens for additives and optimization of the cryoprotection process significantly improved the quality of the diffraction data and electron density maps obtained from this crystal form. The final model was refined against anisotropically truncated data extending to 3.4 Å resolution, yielding R_work and R_free values of 22.6% and 24.4%, respectively (Table S1).

Results and Discussion

The overall conformation of the protein is similar to the previously reported low-resolution structure (4). However, higher resolution has resulted in considerably improved electron density maps, revealing significant details in the transmembrane region, such as ions and side chains lining the ouabain binding cavity and membrane components (Fig. 1C). Also, side chains lining the phosphorylation site can be located, revealing the coordination of the phosphorylated catalytic Asp369 and Mg²⁺. Moreover, the electron density maps indicate a cholesterol molecule located between the transmembrane segments of the γ-subunit (γM) and αM8–10. This locus has been assigned to a lipid site and shown to influence Na⁺,K⁺-ATPase stability (7) (Fig. S1). Finally, residual electron density extending from the side chains of Asn158, Asn193, and Asn265 in the β-subunit confirms the three N-linked glycosylation sites, conserved in all mammalian β1-isoforms (Fig. 1A).

Phosphorylation Site.

The electron density maps define the locations of the side chains lining the phosphorylation site, formed by highly conserved motifs, such as the ³⁶⁹DKTGTLT³⁷⁵ motif, the ⁶⁰⁸MVTGD⁶¹² motif, Lys691, and the ⁷¹⁰DGVND⁷¹⁴ motif of the phosphorylation (P) domain together with the ²¹²TGES²¹⁵ motif of the actuator (A) domain (8, 9). Shortly, the phosphate group of the phosphorylated catalytic Asp369 is stabilized by the side chains of Thr371, Thr610, Lys691, Asp710, Asn713, the backbone amides of Thr371 and Gly611, and the stabilizing Mg²⁺. In addition, the side chains of Asp369, Thr371, Asp710, and the backbone carbonyl of Thr371 further participate in an octahedral coordination of the Mg ion (Fig. S2). Thus, the loop containing the TGES motif does not interact directly with the phosphorylation site, but rather shields the site from spontaneous hydrolysis, as seen also for the E2–BeF₃⁻ complex of the sarcoplasmic reticulum Ca²⁺-ATPase (SERCA1a) (10). This suggests that ouabain binding does not modify the phosphorylation site; rather, it stabilizes the enzyme in an E2P conformation. The ouabain-binding pocket is formed by the transmembrane segments αM1–6, covalently linked to the A- and P-domains; therefore, occupation of the site hinders further conformational rearrangements.

Ouabain Binding Site.

Ouabain binding to an extracellular site of Na⁺,K⁺-ATPase is associated with a downward movement (toward the cytoplasmic side) and lateral shift of αM1–2 toward αM3–10, as well as an ∼15° lateral tilt of the extracellular segment of αM4 away from αM6 (4, 11). The present structure reveals that this rearrangement is supported by an extension of the extracellular part of αM4 by one helix turn (at residues Ile318–Val325), causing an ∼3-Å displacement of this segment toward αM3. The windup of αM4 causes an extension of the ouabain-binding cavity toward cation site II, allowing ouabain to move ∼2 Å deeper into the cavity (Fig. 1D). The electron density maps unambiguously define ouabain located between the transmembrane segments αM1–6, with its lactone facing down (Fig. 1C). Interactions between the concave apolar α-surface of the ouabain steroid core and the complementary scaffold, formed by bulky hydrophobic side chains of Ile315, Phe316, Gly319 (αM4), Phe783, Phe786 (αM5), and Leu793 (loop αM5–6), are similar to those in the low-affinity [K₂]E2–MgF_x–ouabain complex (11) and constitute a generally conserved docking platform for all CTSs. In contrast, the β-surface of the steroid core participates in an extensive network of interactions with polar side chains of αM1, αM2, and αM6 (Fig. 1C). Thus, the hydroxyl group OH19_β of ouabain forms a bifurcated hydrogen bond with Gln111 (αM1) and Asn122 (αM2). Gln111 further interacts with OH1_β and forms stabilizing hydrogen bonds to Asn122. Together, these form the structural basis for the crucial role of Gln111 and Asn122 in species-specific ouabain sensitivity, in which substitution of these residues, e.g., naturally occurring in rodent α1 isoforms, leads to a significant decrease in ouabain affinity (12). Furthermore, hydrogen bonds are formed between OH5_β and Glu117 (αM2) and between the highly conserved OH14_β and Thr797 (αM6), a critical residue for ouabain binding (13, 14), which is stabilized further by Asp121 (αM2). Thus, all hydroxyl groups located on the β-surface participate in a hydrogen-bonding network, in contrast to OH11_α on the α-surface (Fig. S3). However, it seems that inversion of OH11_α into an 11-epi-ouabain form would enable hydrogen bonding with Asn122, explaining its higher affinity (15). The above residues, with the exception of Glu117, have been identified by mutagenesis as crucial for ouabain binding (16, 17). Our structure further confirms that OH12_β of digoxin may decrease the affinity relative to that of digitoxin because of a steric interference with the backbone of αM4 (11, 18).

The sugar moiety of ouabain is located in a wide cavity exposed to the extracellular environment and lined with polar residues, such as Glu116 (loop αM1–2), Glu312 (αM4), Arg880, and Asp884 (loop αM7–8) (Fig. S3). The electron density maps do not support any intimate interaction with the rhamnose moiety exposed to the aqueous environment. This correlates well with the fact that glycosylation of CTSs is not essential for Na⁺,K⁺-ATPase inhibition but rather affects CTS affinity and specificity toward enzyme isoforms (19, 20). The reason for the higher affinities of the glycosylated CTSs compared with their aglycones may be in the steric and hydrophilic nature of the sugar moiety, which facilitates proper orientation of the molecule in the preliminary complex, increasing the probability of transition to the final, high-affinity complex.

The lactone is located in a hydrophobic funnel lined by Leu125 (αM2), Ala323 (αM4), and Ile800 (αM6), 5 Å away from cation site II. No polar interactions are found between the lactone ring and the surrounding residues. However, charge distribution, promoted by relaxation of the highly strained five-membered unsaturated lactone ring, favors an increased partial negative charge on the carbonyl group, unlike six-membered unsaturated lactones of bufadienolides. Facilitated by the hydrophobic environment, the carbonyl group therefore may be engaged in long-range electrostatic interactions with Mg²⁺ bound in site II (see below), improving affinity to cardenolides, and add to CTS differentiation.

The orientation of the ouabain molecule within the site has been a subject for discussion. The crystal structure (with the lactone buried in the transmembrane domain) opposes interpretations of lanthanide-based resonance energy transfer (LRET) experiments using fluorescently labeled ouabain (BODIPY-ouabain) that reach a model with the lactone pointing toward the extracellular side (4, 21). In our opinion, the contradiction is apparent, because the terms “high-” and “low-affinity binding” essentially have different meanings between studies. K_d for the interactions between BODIPY-ouabain and Na⁺,K⁺-ATPase constructs is in the micromolar range (∼10 µM), as revealed by inhibition of the ATPase activity (figure S3 in ref. 21). Obviously, the interactions studied by LRET have very different characteristics compared with the complex used for crystallization with K_d around 3 nM (ref. 4 and present report). The reason for the dramatic change in binding properties is unclear, but a 1,000-fold difference in affinity excludes a direct comparison with the current study. It also is possible that different orientations of ouabain are allowed at initial stages, because binding succeeds in several steps. The snapshots obtained by LRET therefore may represent an intermediate complex, whereas the crystallographic data refer to the final, high-affinity Na⁺,K⁺-ATPase–ouabain complex.

Cation Binding Site.

Protein sample and crystallization buffers were Na⁺- and K⁺-free, yet a strong peak (7.3 σ) was observed in the F_o–F_c difference Fourier map, representing a bound ion at cation site II (22, 23). Anomalous difference Fourier maps, based on several datasets collected at different wavelengths (λ = 1.8–1.2 Å), showed no peak for this site, discouraging models of, e.g., K⁺, Zn²⁺ or Ca²⁺ as bound ions under native crystallization conditions. Also, no site II replacement was found in Ca²⁺- or Zn²⁺-soaked crystals (although Zn²⁺-specific sites were found at the membrane interface; Fig. S4). In contrast, an anomalous difference peak (5.3 σ) at site II appeared upon cocrystallization with 20 mM Mn²⁺ (Fig. 2A), suggesting that the peak observed in the native dataset originates from Mg²⁺ present in the crystallization buffer at concentrations saturating site occupancy (Materials and Methods). Indeed, structural refinement with Mg²⁺ at site II and three water molecules, completing an octahedral coordination, complemented the residual density (Fig. S5). Anomalous difference Fourier maps obtained from cocrystallization with Rb⁺ further demonstrates how Rb⁺ outcompetes Mg²⁺ and occupies both cation sites I and II (Fig. 2 B and C). Na⁺, on the other hand, shows no potent binding and does not suppress the signal of Rb⁺ (Fig. 2D). Residues lining the Mg²⁺ site are identical to those of the K⁺ occluded site II in the [K₂]E2–MgF_x complex (22, 23) with the exception of Val322, which has moved to the ouabain site (Fig. 1D). However, the ionic radius of Mg²⁺ is significantly smaller than that of K⁺; therefore, only Asn776, Glu779 (αM5), and Asp804 (αM6) participate directly in an octahedral coordination of Mg²⁺. Structural alignment of the high-affinity [Mg]E2P–ouabain complex presented here and the low-affinity [K₂]E2–MgF_x–ouabain complex (11) reveals that this tight association of αM4 to Mg²⁺ in the extracellular ion pathway grants space for ouabain to bind at a deeper site with long-range electrostatic interactions, as described above (Fig. 1D). In contrast, K⁺ occlusion winds up and drags Ile318–Val325 toward site II, closing the cation binding site from the extracellular side and hindering tight ouabain binding (Fig. 1E).

Fig. 2.

Binding sites for Mn²⁺ and Rb⁺ as revealed by anomalous difference Fourier maps. The transmembrane segments of αM1–6 are depicted in gray, and the ouabain molecule is represented by dark gray sticks. Anomalous difference Fourier maps obtained from cocrystallization with either Mn²⁺ or Rb⁺ are depicted in pink and yellow-green mesh, respectively. (A) Thus, cocrystallization with 20 mM Mn²⁺ gave rise to a 5.3-σ peak in cation site II, showing that Mn²⁺ enters cation site II. (B) Cocrystallization with 50 mM Rb⁺ gave rise to a strong 7.1-σ peak, emerging from both cation sites I and II. (C) Cocrystallization with 10 mM Rb⁺ gave rise to a 4.9-σ peak, significantly lower than that obtained from cocrystallization with 50 mM Rb⁺, as a result of competition with 200 mM Mg²⁺ (also present in the crystallization buffer). (D) Ten millimolar Rb⁺ in the presence of 50 mM Na⁺ shows that Na⁺ does not suppress the Rb⁺ signal.

Biochemical Evidence Supporting K⁺/Mg²⁺ Competition for Site II.

The identification of an Mg²⁺ at site II implies competition with K⁺. Another assumption is that both ions enter the site from the extracellular side. Several previously published observations support the possibility and sidedness of Mg²⁺ binding within the transmembrane domain of the Na⁺,K⁺-ATPase: (i) Schwappach et al. (24) observed that Mg²⁺ is required for ouabain binding to tryptic fragments of Na⁺,K⁺-ATPase lacking the extramembranous domains. (ii) Mg²⁺ bound within the extracellular release pathway also is observed in the structure of the E2–BeF₃⁻ complex of SERCA1a (10). This Mg²⁺ site overlaps the site observed here for the Na⁺,K⁺-ATPase. (iii) It is known that high concentrations of free Mg²⁺ convert K⁺-sensitive phosphoenzyme into insensitive phosphoenzyme (25), and according to the Post–Albers scheme, the dephosphorylation is stimulated by extracellular K⁺ (2). (iv) Finally, it was shown that an Asp804Glu substitution disrupts ouabain binding (14), emphasizing the importance of a conserved cation site II (unlike cation site I). Thus, the above facts support a competition between the extracellular K⁺ and Mg²⁺, as suggested by the structural data.

Fig. 3A illustrates the competition between Mg²⁺ and K⁺ for the P_i-phosphorylated form. The ions exert opposite effects on ouabain binding to this enzyme form under equilibrium conditions. High ouabain affinity toward E2P in the presence of 3 mM Mg²⁺ alone was decreased significantly by 10 mM K⁺. An increase in Mg²⁺ concentration to 12 mM, however, reversed the inhibiting effect of K⁺ almost completely. The experiment was performed on an “open-membrane” preparation; therefore, the sidedness of the ion effects cannot be inferred from these data. The arguments, however, are found in the literature. Forbush (26) suggested that the cation binding sites in the P_i-phosphorylated form are facing outward. It also is well known that in cell cultures, a decrease in ouabain affinity is induced by extracellular K⁺ (27). Therefore, it is plausible that the well-known effects of both K⁺ and Mg²⁺ on ouabain affinity involve their binding to the outward-facing site II.

Fig. 3.

The effect of Mg²⁺ on ouabain binding and dephosphorylation. The data points represent the average of at least three experiments ± SEM. (A) Competition between Mg²⁺ and K⁺. Inhibition of the ouabain binding by K⁺ is reversed by increasing Mg²⁺ concentration. The calculated K_d values for ouabain are 3 nM (3 mM MgCl₂), 18 nM (12 mM MgCl₂ and 10 mM KCl), and 171 nM (3 mM MgCl₂ and 10 mM KCl). (B) Spontaneous dephosphorylation of the E2P phosphoenzyme (squares) or its ouabain-stabilized complex (circles). Dephosphorylation was induced by addition of either 50 mM nonradioactive P_i (filled symbols) or a mixture of 50 mM nonradioactive Pi and 10 mM EDTA (open symbols).

The data presented in Fig. 3B confirm that the Mg ion, stimulating ouabain binding, is different from the one bound at the phosphorylation site. The time courses of spontaneous dephosphorylation of the E2P and E2P–ouabain forms were followed in experiments with dilution of radioactive phosphate in the presence and absence of EDTA. For the E2P complex, the rate of dephosphorylation was the same in both cases, implying that P_i (and not Mg²⁺) was the first ligand to dissociate. Ouabain binding had a stabilizing effect on the phosphoenzyme; in addition, its presence made the rate of dephosphorylation dependent on the composition of the chase solution. The rate of dephosphorylation of the E2P–ouabain complex in the presence of 10 mM EDTA was faster than that induced by simple dilution of radioactive phosphate, but slower than the spontaneous dephosphorylation of E2P without ouabain. This observation indicates that another Mg²⁺ (other than the Mg²⁺ bound in the phosphorylation site and dissociating after P_i) is involved in the formation of the E2P–ouabain complex. This Mg²⁺ is the first to leave, and its dissociation diminishes the stability of the entire complex. Mg²⁺ bound at site II explains these data, because the presence of ouabain does not block the access channel completely but allows ions to pass, although at reduced rates (28).

Taken together, the above observations make a strong argument for the functional significance of the Mg²⁺ bound at site II, as revealed by the crystal structure of the E2P–ouabain complex.

Conclusion.

The crystallographic and biochemical data obtained in vitro demonstrate that the nature of the cation bound in site II affects the functional properties of the phosphoenzyme, e.g., the affinity toward the CTSs. Na⁺, Ca²⁺, or Zn²⁺ does not seem to interact significantly with this site. The physiological relevance of the competition between extracellular Mg²⁺ (or Mn²⁺, Fig. S6) and K⁺ as yet another factor affecting the Na⁺,K⁺-ATPase activity awaits its confirmation from cell culture experiments. This may be particularly important for interpretations of pathophysiological mechanisms linked to mutations of Na⁺,K⁺-ATPase.

Materials and Methods

Enzyme Preparation and Biochemical Experiments.

Pig kidney Na⁺,K⁺-ATPase was purified as previously described (29). The specific ATPase activity of the Na⁺,K⁺-ATPase purified membrane preparations was about 1,800 µmol P_i per h per mg of membrane protein at 37 °C. Effects of Mg²⁺ and K⁺ on equilibrium ouabain binding to the E2P state of the Na⁺,K⁺-ATPase were measured as described by Yatime et al. (4). Dephosphorylation experiments were performed essentially as described by Fedosova et al. (30).

Crystallization and Data Collection.

For crystallization, the Na⁺,K⁺-ATPase purified membranes were preincubated with 20 mM histidine/0.9 mM EDTA (pH 7.0), 4 mM H₃PO₄ (titrated with N-methyl-d-glutamine, pH 7.0), 4 mM MgCl₂, and 0.5 mM ouabain. The ouabain-stabilized E2P complex subsequently was solubilized with the nonionic detergent octaethyleneglycol mono-n-dodecylether [C₁₂E₈] at a ratio of 0.9 mg C₁₂E₈ per mg of protein, and insoluble material was removed by ultracentrifugation. The final concentration of solubilized protein was 9–10 mg/mL.

Crystals were grown by vapor diffusion at 19 °C in 3-µL hanging drops. Next, 1.6 mM sucrose monodecanoate and 5 mM DTT were added to the solubilized protein before it was mixed at a 1:1 ratio with the reservoir solution containing 16–17% (wt/vol) polyethylene glycol 2,000 monomethyl ether (PEG 2000 MME), 10% (vol/vol) glycerol, 200 mM MgCl₂, 100 mM MES (titrated with N-methyl-d-glutamine, pH 6.2), 100 mM urea, and 5% (vol/vol) tert-butanol. The crystals typically appeared after a few days and grew to a maximum size of ∼150 × 300 × 100 µm³ in 2–3 wk. The crystals were cryoprotected by soaking in additional 5% (vol/vol) glycerol for 1 h. Before flash cooling in liquid nitrogen, excess mother liquor was removed by gently touching the cover side surface with the edge of the loop. Because of a unit cell c-axis of nearly 500 Å, the crystals were mounted in bent LithoLoops (Molecular Dimensions) for an optimal projection of the c*-axis on the detector during data collection. All datasets were collected at 100 K on the beamlines X06SA and X06DA at the Swiss Light Source in Villigen, Switzerland. For cocrystallization of the Rb⁺, Mn²⁺, and Na⁺/Rb⁺ derivatives, 10 mM and 50 mM RbCl, 20 mM MnCl₂, and 50 mM NaCl/10 mM Rb⁺, respectively, were used and added to the crystallization buffer. The MgCl₂ concentration was adjusted to a final divalent cation concentration of 200 mM. For the Ca²⁺ and Zn²⁺ soaks, 100 mM CaCl₂ and 10 mM ZnCl₂, respectively, were added to the crystallization drop and equilibrated for 48 h.

Structure Determination and Analysis.

The diffraction data were processed and scaled with XDS software (31). The crystals displayed a P2₁2₁2₁ space-group symmetry with the unit cell dimensions a = 117.2 Å, b = 118.1 Å, and c = 495.3 Å, containing two αβγ heterotrimers per asymmetric unit and a solvent content of 77% (Matthews V_m coefficient, 5.36 ųDa⁻¹). The crystals showed low mosaicity, and diffraction up to 3.3 Å resolution along the c*-axis. However, because of severe anisotropy, the data could be processed to only 3.9 Å resolution when spherical truncation of the data was performed. To take advantage of the structural information arising from the reflections at higher resolution, the dataset was submitted to the Diffraction Anisotropy Server (http://services.mbi.ucla.edu/anisoscale) (32) to perform an ellipsoidal truncation and anisotropic scaling of the data. The ellipsoidal truncated dataset, with the upper resolution limit cut at 3.9 Å, 3.9 Å, and 3.4 Å along a*, b*, and c*, respectively, was used in later stages of the model refinement. The initial phases were obtained by molecular replacement using PHASER (33) and the previously determined structure of the high-affinity Na⁺,K⁺-ATPase E2P–ouabain complex (Protein Data Bank ID code 3N23) (4) as a search model. Rigid body refinement followed by a simulated annealing refinement protocol was performed in PHENIX (34). Manual model building was carried out in Coot (35), and the model was refined further in PHENIX using noncrystallographic symmetry (NCS), translation–libration–screw parameterization (TLS), and grouped atomic displacement parameter (ADP) refinement (residue main chain and side chains, separately). Because of the rather low resolution of the data, tight geometry restraints were imposed on the model to stabilize the refinement. Rigid body, NCS, and TLS groups were defined by the A-, N-, and P-domains along with the αM1–2, αM3–4, αM5–10/βM/γM, and β-ectodomain, respectively. The quality of the structure was assessed using the MolProbity server (36). Programs from the CCP4 program suite (37) were used to generate the anomalous difference Fourier maps (Collaborative Computational Project, Number 4, 1994).

The crystal lattice has a type 1 crystal packing with layers of membranes stacked on top of each other (38). The asymmetric unit of the unit cell consists of two αβγ-units oppositely oriented relative to each other, which interact via crystal contacts between the loops αM8–9 and αM9–10 in the transmembrane regions, further mediated by the head group of the detergent sucrose monodecanoate. Other crystal contacts between neighboring molecules occur through interactions between the cytoplasmic actuator (A) and phosphorylation (P) domains and the β-ectodomains. The overall structure was well defined by the electron density compared with the modest resolution of the diffraction data. The N-terminal tail of the α-subunit and residues 548–581 of the nucleotide binding (N) domain show a very high degree of disorder, and the N-terminal end of the α-subunit was not visible in the electron density maps. The final model consists of α-subunit residues 21–1,016 (complete C terminus), β-subunit residues 15–303 (complete C terminus), and γ-subunit residues 17–48 (only the transmembrane segment). All structural representations in this paper were prepared with PyMOL (www.pymol.org).