Structures and characterization of digoxin- and bufalin-bound Na⁺,K⁺-ATPase compared with the ouabain-bound complex

Significance

Na⁺,K⁺-ATPase is the ion pump responsible for maintenance of the electrochemical gradients of Na⁺ and K⁺ across the membrane of animal cells. Cardiotonic steroids constitute a broad class of specific Na⁺,K⁺-ATPase inhibitors, including drugs of clinical importance with multiple physiological effects. The existence of several endogenous cardiotonic steroids suggests their involvement in health and disease mediated by various signaling pathways, but the structure–activity relationships are not yet understood. Using X-ray crystallography and analysis of binding kinetics, we characterize Na⁺,K⁺-ATPase complexes with a total of five cardiotonic steroids, showing variations in glycosylation, steroid core substituents, and structure of the lactone substituent. This insight is highly relevant for the understanding of physiological effects and future drug development based on cardiotonic steroids.

Abstract

Cardiotonic steroids (CTSs) are specific and potent inhibitors of the Na⁺,K⁺-ATPase, with highest affinity to the phosphoenzyme (E2P) forms. CTSs are comprised of a steroid core, which can be glycosylated, and a varying number of substituents, including a five- or six-membered lactone. These functionalities have specific influence on the binding properties. We report crystal structures of the Na⁺,K⁺-ATPase in the E2P form in complex with bufalin (a nonglycosylated CTS with a six-membered lactone) and digoxin (a trisaccharide-conjugated CTS with a five-membered lactone) and compare their characteristics and binding kinetics with the previously described E2P–ouabain complex to derive specific details and the general mechanism of CTS binding and inhibition. CTSs block the extracellular cation exchange pathway, and cation-binding sites I and II are differently occupied: A single Mg²⁺ is bound in site II of the digoxin and ouabain complexes, whereas both sites are occupied by K⁺ in the E2P–bufalin complex. In all complexes, αM4 adopts a wound form, characteristic for the E2P state and favorable for high-affinity CTS binding. We conclude that the occupants of the cation-binding site and the type of the lactone substituent determine the arrangement of αM4 and hypothesize that winding/unwinding of αM4 represents a trigger for high-affinity CTS binding. We find that the level of glycosylation affects the depth of CTS binding and that the steroid core substituents fine tune the configuration of transmembrane helices αM1–2.

Cardiotonic steroids (CTSs) induce diverse physiological effects on, for example, heart muscle and blood pressure regulation, but the underlying mechanisms remain unknown, despite a long history of therapeutic applications and model studies. It is widely recognized that they target Na⁺,K⁺-ATPase, and a direct consequence of their binding is an inhibition of the enzyme. Their positive inotropic effect in cardiomyocytes has been related to coupling between Na⁺,K⁺-ATPase and Na⁺/Ca²⁺-exchanger through the intracellular Na⁺ concentration, whereas numerous other outcomes observed on the cellular level have led to hypotheses of the existence of signaling cascade mechanisms with Na⁺,K⁺-ATPase acting as a receptor. The minimal functional unit of the enzyme is an αβ-complex, and because there exist four α- and three β-isoforms of the Na⁺,K⁺-ATPase, the variations in the heterodimer composition and a vast number of CTSs differing in apparent isoform specificities (1) add to the complexity and multiplicity of reported physiological responses.

The conserved structural core shared by all CTSs includes a cis-trans-cis ring-fused steroid core with two methyl substituents at steroid positions C10_β and C13_β, two hydroxyl groups (OH3_β and OH14_β), and an unsaturated lactone ring at the C17_β position, among which the lactone and OH14_β are critical for binding to Na⁺,K⁺-ATPase (2, 3). The type of lactone at the C17_β position divides natural CTSs into cardenolides (five-membered lactone rings) and bufadienolides (six-membered lactone rings). Finally, many CTSs are glycosylated by one to four carbohydrate residues at OH3_β (Fig. S1). It has been shown that glycosylation improves CTS affinity toward the Na⁺,K⁺-ATPase and contributes (at least in the case of digoxin and digitoxin) to their Na⁺,K⁺-ATPase isoform selectivity, with up to fourfold preference for α2/α3 over α1 (1).

The recently published crystal structure of the Na⁺,K⁺-ATPase phosphoenzyme (E2P) in complex with the widely studied CTS ouabain (4, 5) showed that the high-affinity CTS-binding site is constituted by the transmembrane helices αM1–6 of the catalytic α-subunit, forming a pocket exposed to the extracellular side and overlapping with the extracellular ion exchange pathway (6). The E2P–ouabain structure also revealed details on protein–ligand interactions facilitating high-affinity CTS binding compared with a low-affinity ouabain complex (7). Among the important features brought to view by the high-affinity complex structure were (i) a Mg²⁺ ion occupying cation-binding site II, (ii) the rearrangement of αM4, forming the structural basis for the well-known antagonistic effect of K⁺ on ouabain binding, and (iii) an E2P-specific configuration of αM1–2 on the cytoplasmic side, whereas the extracellular end of this helix pair closes in on the CTS-binding site (4). Biochemical experiments showing competitive interactions between K⁺ and Mg²⁺ suggested that the nature of the cation in site II is a determinant for ouabain affinity. In addition, long-range interactions between the unsaturated, polarized five-membered lactone ring of ouabain and the Mg²⁺ ion were suggested as a factor for CTS recognition and differentiation. Despite previous reports showing that glycosylated CTSs have higher Na⁺,K⁺-ATPase affinity than their aglycones, no specific interactions were observed between the sugar moiety of ouabain and the protein to explain that effect.

To gain a better understanding of the structure–activity relationship of the CTSs, we have crystallized the E2P form of the pig kidney Na⁺,K⁺-ATPase (α₁β₁γ) in complex with two CTSs: bufalin (a nonglycosylated bufadienolide) and digoxin (a trisaccharide-conjugated cardenolide) (Fig. 1A), which also are pharmacological agents. We further performed experiments on CTS binding to Na⁺,K⁺-ATPase, including the aglycones digitoxigenin and ouabagenin (Fig. S1). The data revealed notable qualitative differences in kinetics of the enzyme interactions with the glycosylated vs. nonglycosylated CTSs as well as a remarkable insensitivity of bufalin binding to K⁺. The time course of Na⁺,K⁺-ATPase inhibition under steady-state conditions, mimicking the interactions with CTSs in vivo, revealed that binding occurs in two steps. The impact of separate structural components, such as sugar and lactone moieties, on the individual steps of CTS binding is discussed on the basis of our structural and biochemical data.

Fig. 1.

Structural comparison of the crystal structures of the high-affinity Na⁺,K⁺-ATPase α₁β₁γ E2P–CTS complexes. The phosphoenzyme stabilized by bufalin, digoxin, and ouabain (5) is depicted in blue, green, and gray cartoons, respectively, and the bufalin, digoxin, and ouabain molecules are represented by magenta, orange, and dark gray sticks, respectively. The K⁺ and Mg²⁺ ions are represented by purple and yellow spheres, respectively. (A) Structural representation of the CTSs digoxin, bufalin, and ouabain. (B and C) The final 2F_o-F_c electron density maps of the E2P–bufalin and E2P–digoxin, respectively, complexes (contoured at 1.0σ level). The maps are represented by gray mesh. (D) Structural alignment of the E2P–bufalin, E2P–digoxin, and E2P–ouabain complexes performed on the segments αM7–10 showing a high degree of overall structural similarity. (E) The CTS-binding site visualized from the extracellular site based on the same alignment as above. The alignment reveals similar hydrophobic interactions between the α-surface of the CTS core and αM4–6. In contrast, different interactions are formed between the substituents at the β-surface of the CTS core and αM1–2, leading to minor CTS-induced rearrangements. (F) The CTS-binding site visualized from αM1–2. αM4 overlays well for Mg²⁺-bound complexes of E2P–digoxin and E2P–ouabain as well as the E2P–bufalin complex, despite potassium bound in the cation-binding sites.

Results and Discussion

Crystals of the E2P–bufalin and the E2P–digoxin complexes were obtained in two distinct crystal forms exhibiting space group symmetries P2₁ and P2₁2₁2₁, respectively, but each had two αβγ-complexes per asymmetric unit (Table S1). The E2P–digoxin complex was obtained in the same K⁺-depleted crystal form as previously reported for the E2P–ouabain complex (5), whereas crystals of the E2P–bufalin complex were only obtained in the presence of K⁺ and displayed a different crystal packing not seen earlier (Fig. S2). Because of a severe X-ray diffraction anisotropy, the structure of the E2P–bufalin complex was refined against anisotropically truncated data extending to 3.4-Å resolution, yielding R_work and R_free values of 24.5% and 28.8%, respectively. The refinement of the E2P–digoxin complex was also carried out against anisotropically truncated data extending to 3.9-Å resolution, resulting in final R_work and R_free values of 22.1% and 25.3%, respectively.

Despite moderate resolution, the structures of the two E2P–CTS complexes are, overall, well-resolved in the electron density maps (Fig. 1 B and C). The two complexes are structurally similar to the previously reported E2P–ouabain complex (5) (Fig. 1D). An exception is the nucleotide-binding domain of E2P–bufalin, which is loosely associated with the rest of the enzyme in E2P states, shifted here by ∼10° toward the actuator domain, most likely because of a different crystal packing and with no consequence to the overall conformation otherwise. The fact that a similar organization of E2P–CTS complexes is observed, as obtained from two different crystal forms and with three different compounds, validates the structural characteristics of high-affinity E2P–CTS complexes of the Na⁺,K⁺-ATPase described here (4, 5).

Based on initial F_o-F_c difference maps, bufalin and digoxin were located in the cavity of the extracellular ion pathway formed by the transmembrane segments αM1–6. Ions occupying the cation-binding sites were also identified (Fig. 2 and Fig. S3). Thus, for the E2P–bufalin complex, a model of two potassium ions occupying cation-binding sites I and II superimposed well with residual electron density and an anomalous difference Fourier map obtained from an Rb⁺-substituted crystal. For the E2P–digoxin complex, residual electron density matched a single Mg²⁺ bound in cation-binding site II, as observed for the ouabain complex (5), while also providing a clear trace of the trisaccharide moiety named tridigitoxose.

Fig. 2.

The CTS- and cation-binding sites. The Na⁺,K⁺-ATPase α- and β-subunits are depicted in gray and black cartoons, respectively. The F_o-F_c map (contoured at 3.5σ level) obtained after initial rigid body refinement is depicted by beige mesh. (A) The digoxin-binding site. Digoxin and the side chains lining the binding cavity are depicted by orange and green sticks, respectively. The Mg²⁺ ion occupying cation site II is represented by a yellow sphere. (B) The bufalin-binding site. Bufalin and the side chains interacting with bufalin or the two bound potassium ions (purple spheres) are depicted by magenta and blue sticks, respectively. The anomalous difference Fourier map (contoured at 3.5σ level) obtained by replacing 100 mM KCl with 100 mM RbCl is depicted by orange mesh.

From the structural alignment depicted in Fig. 1 E and F, the overall binding mode of bufalin and digoxin appears similar to that of ouabain (5): (i) the concave apolar α-surface of the steroid core interacts with bulky hydrophobic side chains of αM4–6, supporting a previous note that these residues constitute a general, conserved docking platform for all CTSs (5, 7); (ii) the β-surface of the steroid is facing the polar side chains of αM1–2; (iii) the lactone is located in a hydrophobic funnel formed by αM4–6, leading to the cation-binding sites; and (iv) similar to the rhamnose moiety of ouabain, the tridigitoxose moiety of digoxin is found in a wide hydrophilic cavity, with the two distal digitoxose units extending into the extracellular matrix.

For ouabain, four hydroxyl substituents together with the glycosylated OH3_β at the β-surface of the steroid core participate in an extensive hydrogen-bonding network with polar side chains of αM1–2 and αM6 (5), imposing species-specific sensitivity (8, 9). Bufalin and digoxin, however, are much more hydrophobic and contain only the two highly conserved hydroxyl groups OH14_β and OH3_β (OH3_β is glycosylated in digoxin and other glycosylated CTSs). Digoxin, in addition, also has OH12_β. Similar to the E2P–ouabain complex, OH14_β of bufalin and digoxin are likely within hydrogen-bond distance to Thr797 (αM6) and further stabilized by Asp121 (αM2). OH3_β of bufalin and the corresponding glycosidic bond of digoxin, located near the solvent-exposed glycosyl-binding area of the CTS-binding site, do not appear to form any protein–ligand interactions. OH12_β of digoxin may form hydrogen bonds with Asn122 (αM2) (Figs. 1E and 2). Thus, unlike ouabain, bufalin and digoxin do not form hydrogen bonds with either Gln111 or Glu117 (loop αM1–2). Comparison of the binding modes of the three CTSs shows similar interactions between the conserved OH14_β and Thr797 (αM6), whereas variation in the substituents at the steroid core results in different interaction networks with αM1–2.

Indeed, structural alignment based on the transmembrane segments αM7–10 of the three E2P–CTS complexes shows that the transmembrane helices forming the CTS-binding cavity are superimposable, except for αM1–2 (Fig. 1 E and F). In the E2P–bufalin and E2P–digoxin complexes, the extracellular part of αM1–2 seems to move slightly further into the CTS-binding site compared with the E2P–ouabain complex, suggesting that the position of αM1–2 is affected by the number, size, and properties of the substituents at the CTS core. Thus, in addition to a prominent movement toward the cytoplasmic side and lateral shift of αM1–2 promoted by autophosphorylation (4), a fine-tuning of αM1–2 seems to be induced by the bound CTS, leading to a unique E2P subconformation, which may also explain a distinct proteolytic pattern (10). Previous findings, showing that isoform selectivity of various CTSs is related to the residues in the loop of αM1–2 (e.g., Thr114 and Gln119) (Fig. 1E), although not directly involved in CTS binding, may follow a similar rationale (11–13).

As depicted in Figs. 1E and 2A, the first proximal digitoxose unit in the digoxin complex is located in a wide cavity exposed to the extracellular environment lined with polar residues, such as Gln111, Thr114, Glu116, Gln117 (loop αM1–2), Glu312 (αM4), Arg880 (loop αM7–8), and Gln84 (β-ectodomain). Of these residues, Glu312 may interact directly with the glycosidic bond between the first and second digitoxose units, and Arg880 can form hydrogen bonds with OH3 of the second digitoxose unit. Other residues of both α- and β-subunits may also be in a position to form temporary water-mediated interactions with the tridigitoxose moiety exposed to the solvent (12). The impact of the aqueous environment on the protein coordination of the CTS sugar moiety has been discussed for the high-affinity E2P–ouabain complex, for which the electron density maps did not support any strong coordination of the rhamnose moiety (5).

The structures of the E2P–CTS complexes reveal differences in the cation-binding sites. Residual electron density in binding site II of the E2P–digoxin complex (Fig. 2A) suggests the presence of one Mg²⁺, similar to the E2P–ouabain complex. As depicted in Fig. 1F, the Mg²⁺ ions located in cation-binding site II of the cardenolide complexes are superimposable with similar coordination (5).

The bufalin complex was crystallized in the presence of ∼150 mM K⁺, and residual electron density covered both cation-binding sites I and II (Fig. 2B). For the E2P–ouabain complex, K⁺ binding to sites I and II unwinds and drags αM4 toward the cation site, hindering tight ouabain binding (5). This rearrangement does not occur in the bufalin-bound complex: structural alignment of the cation-binding sites with the K⁺-occluded [K₂]E2–MgF_x form (14, 15) reveals that αM4 remains wound up as seen in the high-affinity ouabain complex, even in presence of K⁺. As a consequence, K⁺ bound to site II is shifted by about 1.5 Å toward αM4, coordinated by the six-membered lactone carbonyl group of bufalin (see below) along with the backbone carbonyls of Ala323 and Val325 (αM4), the carboxyl groups of Glu327 (αM4), and possibly, also Asp804 (αM6) (Fig. S4). The coordination of the K⁺ ion in site I seems to involve the same residues as for the occluded [K₂]E2–MgF_x complex, but we cannot rule out that the ion here also may shift slightly in position.

Similar to ouabain, the lactone moieties of both bufalin and digoxin are located in a hydrophobic funnel formed by Leu125 (αM2), Ala323 (αM4), and Ile800 (αM6), leading to the cation-binding sites I and II. No polar interactions are found between the lactone groups and surrounding residues. However, comparison of the three E2P–CTS complexes indicates that the six-membered lactone of bufalin reaches ∼1.5 Å deeper into the site than the five-membered lactone of the two cardenolides (Fig. 1F), suggesting that K⁺ may drag bufalin toward the cation-binding site. The lack of hydroxyl substituents and glycosylation at the steroid core may facilitate bufalin sliding deeper into the site as the lactone carbonyl in paraposition coordinates directly to K⁺ at site II. The bulky, six-membered lactone group hinders αM4 unwinding.

Based on three E2P–CTS crystal structures and detailed kinetic analysis of CTS binding, we propose a sequence of events in binding and discuss specific roles of the individual structural components of CTSs in this process. The general structural similarity of the three CTS–enzyme complexes is confirmed by CTS-binding experiments under steady-state conditions of ATP hydrolysis (Fig. 3A, Inset). The inhibition of the ATPase activity by varying ouabain or bufalin concentration was followed in time and analyzed as described in Materials and Methods. A hyperbolic dependence of the observed rate constant on the CTS concentration (Fig. 3A) is indicative of a classical two-step ligand-binding scheme:

$$E\mathrm{+}L\stackrel{K_{\text{i}}}{\iff }E\mathrm{\cdot }L\underset{k_{\mathrm{-}1}}{\overset{k_{\mathrm{+}1}}{\rightleftarrows }}E\mathrm{-}L,$$

where E represents the phosphoenzyme and L is a CTS ligand. Briefly, the fast equilibrium initial binding of the ligand is followed by a conformational rearrangement (induced fit) of the complex, as previously hypothesized on the basis of the crystal structure of the high-affinity E2P–ouabain complex (4, 5). The binding process is fully described by the following scheme, including (i) the equilibrium characterized by dissociation constant K_i for the preliminary complex formation and (ii) conformational transition with the rate constants k₊₁ and k₋₁ (Table S2). The intercept of the hyperbolic curve with the y axis at CST→0 in Fig. 3A reflects k₋₁ and the properties of the tight E–L complex. These values are very similar for bufalin and ouabain, consistent with the crystal structures showing overall identical complexes. The forward rate constant k₊₁ reflects how fast the transition to E–L occurs. The backward rate constant k₋₁ describes the strength of finally established protein–ligand contacts.

Fig. 3.

Interactions of the Na⁺,K⁺-ATPase with CTSs. (A) Binding of CTSs under steady-state conditions of the Na⁺,K⁺-ATPase reaction followed by a coupled enzymes method. Inset shows changes in the degree of ATPase inhibition over time after ouabain addition (second arrow). The rate constants k_obs were calculated as described in Materials and Methods and presented for ouabain (●) and bufalin (■) as functions of CTS concentrations. The data were fitted to a hyperbolic function, and parameters of the best approximation are presented in Table S2. (B) Effect of glycosylation on CTS binding to the Na⁺,K⁺-ATPase. The decrease of Na⁺,K⁺-ATPase activity is shown as a function of digoxin (△), ouabain (○), or ouabagenin (■) concentrations in the preincubation medium containing 3 mM P_i and 3 mM MgCl₂. The data points for digoxin and ouabain are almost overlapping. The data were fitted to a square root equation as described in ref. 4. The calculated K_d values are 2.8 ± 2 nM (digoxin), 1.1 ± 1 nM (ouabain), and 844 ± 100 nM (ouabagenin). (C) The effect of the size of lactone on binding of aglycones to the Na⁺,K⁺-ATPase. The decrease of Na⁺,K⁺-ATPase activity is shown as function of digitoxigenin (●) or bufalin (■) concentrations. Presence of 200 mM K⁺ affects the inhibitory properties of the aglycones (○ and □, respectively). The calculated K_d values are 14 ± 5 nM (bufalin), 9 ± 10 nM (bufalin; 200 mM K⁺), 26 ± 15 nM (digitoxigenin), and 650 ± 400 nM (digitoxigenin; 200 mM K⁺).

The above scheme also offers an explanation for the only partial inhibition observed for the aglycone ouabagenin in the experiments on equilibrium binding of CTSs to the E2P conformation in the absence of monovalent cations (Fig. 3B) (conditions similar to crystallization experiments). The amount of the fast dissociating complex E∙L at saturating ligand concentrations is determined by the ratio k₋₁/k₊₁. On dilution with the reaction media for activity measurements, E∙L will promptly dissociate and maintain a pool of fully active protein E if the ratio k₋₁/k₊₁ is high. Glycosylated CTSs (Fig. 3B), however, show 100% inhibition in the same experimental setup, suggesting that the presence of a sugar moiety shifts the equilibrium toward the tight E–L complex and is advantageous for the induction of the conformational transition.

Next, the binding properties of the aglycones bufalin (bufadienolide) and digitoxigenin (cardenolide) were compared (Fig. 3C). These two CTSs are identical, except for the size of their lactone ring (Fig. S1). As discussed above, both aglycones exhibit incomplete inactivation, although to different levels. The affinities for both CTSs were high, with apparent equilibrium dissociation constant K_d of 14 nM for bufalin and 26 nM for digitoxigenin. As shown earlier for other cardenolides (e.g., ouabain) (5), K⁺ in the preincubation media shifts the digitoxigenin-binding curve to higher ligand concentrations (i.e., decreases apparent affinity; K_d as high as 650 nM).

The K⁺ effect on bufalin binding is, however, peculiar, because K⁺ rather improves the affinity (K_d ∼ 9 nM) and makes the inhibition complete. This effect is not caused by ionic strength favoring hydrophobic interactions, because 200 mM N-methyl-d-glucamine (NMDG⁺) did not induce such changes (Fig. S5). As described above, the lactone of bufalin seems to interact directly with K⁺ in binding site II, an interaction that is prohibited with a hydrated but deeper-bound Mg²⁺ ion involved in digoxin and ouabain binding. A direct coordination to K⁺ at site II enables bufalin to retain the K⁺-bound phosphoenzyme in an E2P-like form. However, the bulky, six-membered lactone, at the same time, prevents protein rearrangements associated with K⁺ occlusion.

The effect of K⁺ on the equilibrium between bufalin-bound E2P complexes (E∙L and E–L) revealed in the kinetic experiments is also consistent with our observation that crystallization was successful only in the presence of K⁺. At the same time, we observe that Mg²⁺, obligatory for measurable binding of cardenolides, is not required for bufalin binding. The experiments leading to the above conclusion were based on interactions of the fluorescent probe eosin with Na⁺,K⁺-ATPase and followed well-established facts: (i) eosin binds with high affinity to the E1 conformation with an increase in fluorescence; (ii) binding of CTS produces the E2 form; and (iii) the E2→E1 transition is a rate-limiting step. Fig. S6 depicts changes of eosin fluorescence after its addition to the Na⁺,K⁺-ATPase preincubated with either 1 mM ouabain or 0.5 mM bufalin in the presence of 20 mM 1,2-diaminocyclohexanetetraacetic acid (a Mg²⁺ chelator) overnight. The fact that the rate of eosin binding to enzyme preincubated with ouabain was the same as in the control experiment excludes the formation of an ouabain–enzyme complex. In contrast, a significant decrease in the rate of fluorescence change in a similar experiment with the bufalin-exposed enzyme revealed the presence of E2 conformation (i.e., on preincubation, bufalin did form a complex with the enzyme, even in the absence of Mg²⁺).

Compilation of the structural data and kinetics outlines major principles for CTS binding. (i) An initial low-affinity complex forms by binding of the hydrophobic α-surface of the steroid core to the scaffold formed by αM4–5 and the loop αM5–6; these contacts are similar in the complexes based on both E2P and [K₂]E2–MgF_x forms. (ii) In the autophosphorylated state of the enzyme the closure of the site around the CTS is enabled, with hydrogen bonds forming between the β-surface of the CTS core and polar side chains of αM1, αM2, and αM6. The type and number of these interactions are important for K_d values and fine-tuning of the αM1–2 configuration. (iii) The origin of the cation in the cation-binding site and the type of CTS lactone determine the arrangement of residues Ile318–Pro326 (αM4) at the interface of the CTS-binding site and cation site II. We hypothesize that winding/unwinding of this stretch of αM4 has critical influence on the spatial organization of the CTS-binding site and serves as a triggering mechanism for a CTS-induced fit. The type and number of substituents on the steroid core, including the degree of glycosylation, affect the depth of CTS binding within the site and together with the size of lactone ring, determine the extent of K⁺ antagonism. Note that digitoxigenin is less sensitive to K⁺ than, for example, ouabain. In terms of kinetics, these factors have an impact on the rates of the conformational transitions k₋₁ and k₊₁. These principles explain the binding modes of the CTSs tested and can be used to rationalize the search for isoform-specific CTSs.

Materials and Methods

Enzyme Preparation and Biochemical Experiments.

Purified pig kidney Na⁺,K⁺-ATPase was prepared as previously described (16). The specific activity was about 1,800 µmol/mg protein per hour at 37 °C. Dissociation constants for the CTSs and effects of K⁺ and NMDG⁺ on their affinity toward the E2P conformational state were estimated from residual Na⁺,K⁺-ATPase activity after preincubating each of these inhibitors with the enzyme as described by Yatime et al. (4). The approach assumes that (i) CTS binding leads to a complete inhibition of the enzymatic activity, (ii) binding reaction is effectively stopped by a 40-fold dilution of the CTS with the reaction media, and (iii) no dissociation of the inhibitor occurs in the timescale of activity measurements (2 min). Binding of CTSs to the Na⁺,K⁺-ATPase under steady-state conditions was followed in time for varying concentrations of CTSs, and the k_obs values for the inhibitor binding were extracted from exponential fits of the first derivative of the curves describing the product formation in the coupled enzyme essay system (17).

Eosin fluorescence was assessed by a stopped-flow technique essentially as described by Fedosova et al. (18). Briefly, the enzyme was preincubated in 20 mM histidine (pH 7.0), 20 mM 1,2-diaminocyclohexanetetraacetic acid, and CTS (no CTS, 1 mM ouabain, or 0.5 mM bufalin) at 0 °C overnight. Eosin binding was followed immediately after mixing of the preincubated enzyme with an equal volume of 0.4 µM eosin, 100 mM NaCl, and all of the components of the enzyme sample, excluding Na⁺,K⁺-ATPase.

The commercial programs KyPlot 5 (Kyence Inc.) and Origin 8.5 (OriginLab Corp.) were used for data analysis.

Crystallization and Data Collection.

For crystallization of the E2P–bufalin complex, Na⁺,K⁺-ATPase membrane fraction was preincubated with 20 mM histidine/0.9 mM EDTA (pH 7.0), 4 mM MgCl₂, and 4 mM H₃PO₄ (titrated with KOH, pH 7.0) and solubilized with a solution of the nonionic detergent octaethyleneglycol mono-n-dodecylether (C₁₂E₈; 180 mg/mL) at a ratio of 0.9 mg C₁₂E₈ per 1 mg protein containing bufalin (final bufalin concentration after solubilization was ∼0.5 mM). Solubilization was performed in the presence of a secondary detergent 4-cyclohexyl-1-butyl-β-D-maltoside at a detergent–protein ratio of 0.13. For crystallization of the E2P–digoxin complex, Na⁺,K⁺-ATPase was preincubated with 20 mM histidine/0.9 mM EDTA (pH 7.0), 4 mM MgCl₂, 4 mM H₃PO₄ (titrated with NMDG, pH 7.0), and 0.5 mM digoxin. The digoxin-stabilized E2P complex was subsequently solubilized with C₁₂E₈ at a ratio of 0.9 mg C₁₂E₈ per 1 mg protein. After solubilization of both E2P–CTS complexes, insoluble material was removed by ultracentrifugation. The final protein concentration was 8–9 mg/mL.

Crystals were grown by vapor diffusion from hanging drops at 19 °C. For the E2P–bufalin complex, crystals were obtained by mixing protein and reservoir solution in a ratio of 1:1 together with 0.02% (wt/vol) n-dodecyl-β-D-maltoside and 2 mM DTT. The reservoir solution (500 µL) contained 17% (wt/vol) PEG 2000 monomethyl ether, 10% (vol/vol) glycerol, 220 mM MgCl₂, 100 mM KCl, 100 mM MES (titrated with KOH, pH 6.2), 5% (vol/vol) 2-methyl-2,4-pentanediol, and 6% (vol/vol) Jeffamine M-600 (titrated with KOH, pH 6.2). For cocrystallization with Rb⁺, 100 mM KCl of the reservoir solution was replaced by 100 mM RbCl. Before flash-cooling, the crystals were dehydrated overnight against a reservoir solution containing 5% (vol/vol) ethylene glycol. Crystals of the E2P–digoxin complex were obtained by adding 25 mM DTT to the protein supernatant and subsequently, mixing 1.5 µL protein solution with 1.5 µL reservoir solution (400 µL) containing 16% (wt/vol) PEG 2000 monomethyl ether, 10% (vol/vol) glycerol, 200 mM MgCl₂, and 100 mM MES (titrated with NMDG, pH 6.2). The crystals were cryoprotected by soaking in additional 0.3 µL 80% (vol/vol) glycerol for 20 min. Crystals of both Na⁺,K⁺-ATPase–CTS complexes were mounted in litho loops (Molecular Dimensions Ltd.) and flash-cooled in liquid nitrogen. The final datasets were collected at 100 K on the Beamline X06SA at Swiss Light Source in Villigen, Switzerland.

Structure Determination and Analysis.

The X-ray diffraction data were processed and scaled with XDS (19). The E2P–bufalin and E2P–digoxin complexes displayed two different crystal forms. Crystals of the E2P–bufalin complex showed a P2₁ space group symmetry with unit cell dimensions a = 65.9 Å, b = 240.3 Å, and c = 152.7 Å, angles α = γ = 90.0° and β = 102.3°, two αβγ-heterotrimers in the asymmetric unit, and a solvent content of 68% (Matthews’ V_m coefficient = 3.81 ųDa⁻¹). The E2P–digoxin complex was crystallized in space group P2₁2₁2₁ with unit cell dimensions a = 118.2 Å, b = 118.3 Å, and c = 494.1 Å and two αβγ-heterotrimers per asymmetric unit, similar to the previously reported Na⁺,K⁺-ATPase E2P–ouabain complex (5). Both crystal forms showed severe diffraction anisotropy. Thus, to take advantage of the structural information arising from reflections at higher resolution along the well-diffracting directions, the datasets were submitted to the Diffraction Anisotropy Server (www.doe-mbi.ucla.edu/∼sawaya/anisoscale) (20) to perform an ellipsoidal truncation and anisotropic scaling of the data. Thus, data of the E2P–bufalin complex were truncated with upper resolution limit cuts at 5.6, 4.1, and 3.4 Å (along a*, b*, and c*, respectively), and data of the E2P–digoxin complex were truncated at 5.2-, 4.3-, and 3.9-Å resolution, respectively. Initial phases were obtained by molecular replacement using PHASER (21), and the crystal structure of the Na⁺,K⁺-ATPase E2P–ouabain complex (Protein Data Bank ID code 4HYT) (5) was used as a search model. Subsequently, atomic displacement parameter factors were reset to the estimated Wilson B value, and rigid body and grouped atomic displacement parameter refinement was performed in PHENIX (22). Model building was carried out manually in COOT (23), and additional model refinement was performed in PHENIX using noncrystallographic symmetry, translation–libration–screw parameterization, and grouped atomic displacement parameter refinement. Because of moderate resolution of the data, tight geometry restraints were imposed on the model to stabilize the refinement. One atomic displacement parameter group per residue was used for the refinement of the E2P–bufalin complex, and two atomic displacement parameter groups (residue main chain and side chains separately) per residue were used for the E2P–digoxin complex. Rigid body groups were defined by the A, N, and P domains along with the αM1–2, αM3–4, αM5–10/βM/γM, and β-ectodomain. Noncrystallographic symmetry and translation–libration–screw parameterization groups were defined by the A, N, and P domains, the transmembrane domain αM1–10/βM/γM, and the β-ectodomain. Because the E2P–digoxin complex displayed the same crystal form as the Na⁺,K⁺-ATPase E2P–ouabain complex (5), the same set of R_free flags was used for refinement. The geometry of the structures was evaluated using the MOLPROBITY server (24). The programs SIGMA-A and FFT from the CCP4 program suite (25) were used for generating anomalous difference Fourier maps. All structural representations in this paper were prepared with PyMOL (The PyMOL Molecular Graphics System; Schrödinger, LLC).

The final model of the E2P–bufalin complex consists of α-subunit residues 21–1,016 (complete C terminus), β-subunit residues 16–303 (complete C terminus), and γ-subunit residues 17–48 (only the transmembrane segment). For the E2P–digoxin complex, the final model consists of α-subunit residues 21–1,016, β-subunit residues 14–303, and γ-subunit residues 17–48.