Computer modelling reveals new conformers of the ATP binding loop of Na⁺/K⁺-ATPase involved in the transphosphorylation process of the sodium pump

Comparative modeling of the open conformation

As the solved crystal structures of Na⁺/K⁺-ATPase are only of a non-human origin (Håkansson, 2003; Kanai et al., 2013; Laursen et al., 2015; Laursen et al., 2013; Mark Hilge et al., 2003; Morth et al., 2011; Morth et al., 2007; Nyblom et al., 2013; Ogawa et al., 2015; Ogawa et al., 2009; Shinoda et al., 2009; Yatime et al., 2011), we decided to employ the procedure of homology modeling to get its human α₂-subunit 3-D-structure. The primary amino acid sequence of the human Na⁺/K⁺-ATPase was retrieved from the ExPASy server (UniProt KB/TrEMBL; http://www.expasy.ch/). The resulting P50993 (AT1A2_HUMAN) target human sequence in the length of 1,020 amino acids for the Na⁺/K⁺-ATPase α₂-subunit precursor of sodium/potassium-transporting ATPase α₂-subunit, Homo sapiens, EC 3.6.3.9, was chosen (Shull, Pugh & Lingrel, 1989). Five amino acids at the N-terminal beginning of this sequence compared to the translated RNA sequence do not occur in the native form (Hara et al., 1987; Kawakami et al., 1985; Ovchinnikov et al., 1986; Shull, Schwartz & Lingrel, 1985). Hence, they were not included in our further numbering. For modeling, known structures of Na⁺/K⁺-ATPase deposited at the RCSB Protein Data Bank (http://www.pdb.org/) were used. In order to create the model based on the above mentioned sequence, the solved crystal structures of Na⁺/K⁺-ATPase with RCSB Protein Data Bank (http://www.pdb.org/) accession codes 3B8E (Morth et al., 2007) and 3KDP (Morth et al., 2007) were used as the templates for our modeling. The multialignment of the chosen target (P50993, AT1A2_HUMAN) sequence and the two templates (3B8E, 3KDP) for open conformation was prepared by MODELLER program (salign module) (Eswar et al., 2006; Marti-Renom et al., 2000; Šali, 1995; Šali & Blundell, 1993). The choice of templates (solved crystal structures) was based on the species proximity (pig over shark) and absence of any cardioglycosides in the solved crystal structure, in order to get as close to the native form as possible. Using this multialignment and the solved 3D crystal structures, we have generated thousand M₄M₅-loop models by the MODELLER (automodel module) (Eswar et al., 2006; Marti-Renom et al., 2000; Šali, 1995; Šali & Blundell, 1993) program. From those created models the best thirty were selected using the PROCHECK (Laskowski et al., 1993; Morris et al., 1992) and Verifi3D (Bowie, Luthy & Eisenberg, 1991; Lüthy, Bowie & Eisenberg, 1992) programs (Table 1).

Comparative modeling of the closed conformation

As in the previous comparative modeling procedure, we have used the sequence P50993 (AT1A2_HUMAN) for modeling of Na⁺/K⁺-ATPase in the closed conformation. However, the solved crystal structures of the RCSB Protein Data Bank (http://www.pdb.org/—accession codes 3WGU, 3WGV and 4HQJ (Kanai et al., 2013; Nyblom et al., 2013)) were used as the templates for our modeling. Using the above-mentioned settings for the modeling program (MODELLER see previous paragraph) we obtained ten models of Na⁺/K⁺-ATPase in the closed conformation. From these, the best model has been chosen using the above-mentioned PROCHEK and Verifi3D programs (Table 1).

Docking, using the open and closed conformations

The best thirty models corresponding to the open conformation and the best model for the closed conformation were used for docking of Mg²⁺•ATP complex, using the Vina-Autodock program (Trott & Olson, 2010). We have decided to use the whole Mg²⁺•ATP complex, which has been derived from structures containing ATP, deposited in RCSB Protein Data Bank (http://www.pdb.org/). The sequential docking of Mg²⁺, followed by ATP, has not been used, since the bond between Mg²⁺ and ATP phosphates is stronger than between Mg²⁺ and –COOH groups of amino acids (Alberty, 1969; Dudev, Cowan & Lim, 1999). In addition, the same procedure was used for the model of the closed conformation.

Molecular dynamics

Molecular dynamics (MD) of the M₄M₅-loop were simulated by Gromacs (Berendsen, Van der Spoel & Van Drunen, 1995; Hess et al., 2008; Lindahl, Hess & Van der Spoel, 2001; Van der Spoel et al., 2005; Van der Spoel et al., 2010), using the OPLS-AA potential (Jorgensen, Maxwell & Tirado-Rives, 1996; Jorgensen & Tirado-Rives, 1988; Kaminski et al., 2001; Meagher, Redman & Carlson, 2003; Pranata, Wierschke & Jorgensen, 1991) with combination of water model TIP3P (Jorgensen et al., 1983). Our protein model with or without 2Mg²⁺•ATP were put into a rectangular box with a 1 nm thick layer of the water molecules around and periodic boundary conditions (Berendsen, Van der Spoel & Van Drunen, 1995; Hess et al., 2008; Lindahl, Hess & Van der Spoel, 2001; Van der Spoel et al., 2005; Van der Spoel et al., 2010).

The PME method (Darden, York & Pedersen, 1993) with a length parameter of 1 nm was used to describe Coulomb type electrostatic interactions and the cut-off method with a length parameter of 1 nm for the calculation of van der Waals interactions. As the first step of the MD simulation, the system of protein and water was energetically optimized using the method of steepest descents, followed by a conjugate gradient minimization algorithm with maximum 2.5 × 10⁴ steps and maximum force smaller than 10 kj mol⁻¹ nm⁻¹ as the convergence criterion (see Supplemental Information 1). The Berendsen et al. (1984) coupling method was employed for the temperature and pressure coupling of a system to reflect the reference temperature of 310 K and the pressure of 1 bar. The leap-frog integration with 10⁴ steps was used for stabilization, with integration step of 1 fs, corresponding to 10 ps simulation time to reach the equilibrium of the rectangular box (see Supplemental Information 1). This stabilized rectangular box was used for the main thirty simulations with 5 ×10⁶ steps (2 fs single step), corresponding to 10 ns for each stabilization using the same simulation parameters as for the box stabilization. These thirty trajectories were simulated with independently generated initial conditions corresponding to a Maxwell distribution for a temperature of 310 K. The translation and rotation around the center of mass of the protein were removed, avoiding thus the simulated system distortion in the simulation box. The molecular dynamics simulation of the Na⁺/K⁺-ATPase in semi-open conformation with docked 2Mg²⁺•ATP was carried out using the same parameters as described above.

Docking to the semi-open conformation

From the charts of the time evolution of distances, the typical trajectory for the molecular dynamics of the open conformation without 2Mg²⁺•ATP has been chosen. The model showing a distance between Asp³⁶⁹ and Phe⁴⁷⁵ smaller than 2.2 nm and with the best PROCHECK and Verify3D scores were taken for docking of the 2Mg²⁺•ATP complex.

Assembly of the static 3D computational model of Na⁺/K⁺-ATPase

The main goal of this work was to describe the molecular mechanism of γ-phosphate transfer from the ATP in the binding site to the phosphorylation site (Asp³⁶⁹) of Na⁺/K⁺-ATPase. To achieve this, first, a static three-dimensional model of Na⁺/K⁺-ATPase was developed based on the latest data and information. The P50993 (AT1A2_HUMAN) human target sequence of the Na⁺/K⁺-ATPase α₂-isoform (sodium/potassium-transporting ATPase α₂-subunit, Homo sapiens) was used for modeling and 3D model assembling. Two templates of Na⁺/K⁺-ATPase structures of accession codes 3B8E (Morth et al., 2007) and 3KDP (Morth et al., 2007) were retrieved from the Protein Data Bank which were proposed by UniProt server as sequence P05024 (sodium/potassium-transporting ATPase α₁-subunit, Sus scrofa) with sequences identity 86.5% of the target sequence. The alignments for open conformation of the M₄M₅-loop were prepared with identity of the 85.6% between the corresponding sequences of the human α₂-isoform and pig α₁-isoform for the M4M5-loop. The alignment and template structures were used for comparative modeling using the MODELLER program. The obtained 3D models were verified, applying the PROCHECK and Verifi3D programs (Table 1). Our modeling procedure resulted in static structures of the M₄M₅-loop of human α₂ isoform of Na⁺/K⁺-ATPase between Thr³³⁸ and Ile⁷⁶⁰ (see Fig. 1, with docked 2Mg²⁺•ATP). These models show distances around 3.26 nm between Phe⁴⁷⁵ as part of the ATP-binding site and the α-carbon of Asp³⁶⁹, the acceptor site for the phosphointermediate in ATP hydrolysis.

Two ATP binding sites exist in the open conformation of Na⁺/K⁺-ATPase

The obtained models of the open conformation of Na⁺/K⁺-ATPase were tested by an in silico ATP-docking experiment for its ability to bind Mg²⁺•ATP. Surprisingly, we identified two possible docking sites (Fig. 1): the first one is in closest vicinity to Phe⁴⁷⁵ (“the Phe⁴⁷⁵ location”) and the second one is close to Asp³⁶⁹, (“the Asp³⁶⁹ location”). Both binding sites showed only slightly different docking energies. While the docking energy at the Phe⁴⁷⁵ location was E_b =  − 7.6 kcal/mol, the docking energy at the Asp³⁶⁹ location was E_b =  − 8.6 kcal/mol. A closer insight into our model clearly indicated interactions among π-electrons between Phe⁴⁷⁵ and the ATP adenine ring at the Phe⁴⁷⁵ location, but the interaction between ATP’s phosphates with bound magnesium and the negatively charged aspartate residue was responsible for the 2Mg²⁺•ATP binding at the Asp³⁶⁹ location.

The amino acids found in the neighborhood of docked Mg ²⁺•ATP in both binding sites were identified and found to be in agreement with already published data (Jacobsen, Pedersen & Jorgensen, 2002; Jorgensen, Hakansson & Karlish, 2003; Jorgensen, Jorgensen & Pedersen, 2001; Jorgensen & Pedersen, 2001; Kubala et al., 2003; Pedersen, Jorgensen & Jorgensen, 2000).

High ATP concentration hinders the enzyme cycle and keeps the Na⁺/K⁺-ATPase at the open conformation

The structure with best docking energy for both 2Mg²⁺•ATP docking sites was the starting point for a molecular dynamics simulation, which revealed another surprising result. The molecular dynamics simulation in the presence of 2Mg²⁺•ATP (we ran two simulations series, one with 2Mg²⁺•ATP docked in the ATP binding site and the other in the phosphorylation site) did not result in a stable close conformation needed to phorphorylate Asp³⁶⁹ during Na⁺/K⁺-ATP hydrolysis. Interestingly, the enzyme preferentially remained in the open conformation in both simulations as is evident from the resulting distance distribution between α-carbons of Asp³⁶⁹ and Phe⁴⁷⁵ (Figs. 2A, 2B, 3A and 3B, the distance varied from 2.5 nm to 3.4 nm, with maxima 2.9 nm and 3.1 nm respectively).

However, molecular dynamics experiments in the absence of 2Mg²⁺•ATP (Fig. 3C) in nanosecond timescale exhibited a different pattern. This conformation was characterized by shortening of the distance between α-carbons of Asp³⁶⁹ and Phe⁴⁷⁵ to about d ∼ 2.00 nm (Fig. 3C).

Additionally, we performed 30 simulations in the absence of 2Mg²⁺•ATP. Yet, there was no stable result: sometimes, the molecular dynamic simulation led to the new conformation (we will call this conformation “semi-open” conformation), but sometimes the enzyme remained in the open conformation, with the ratio open/semi-open conformation being approximately 1:1.4 (Figs. 2C and 3C). Clearly, conformational transitions between the “open” and “semi-open conformations” seem to be rather a stochastic process (Figs. 2C and 3C). Consequently, we decided to call this newly identified conformation representing a distance of 2.3 nm between Phe⁴⁷⁵ and Asp³⁶⁹ the “semi-open” conformation of the catalytic site of Na⁺/K⁺-ATPase. The semi-open conformation is characterized by the formation of a hydrogen bond between Arg⁵⁴³ and Asp⁶¹¹ (see Fig. 4).

Hinge movement, ATP binding and enzyme phosphorylation

Release and re-binding of 2Mg²⁺•ATP complexes at two different binding sites in the open conformation may have huge consequences for the molecular mechanism of the transphosphorylation process to Asp³⁶⁹ as part of the ATP hydrolysis of Na⁺/K⁺-ATPase. Most importantly is the fact that the α-subunit of Na⁺/K-ATPase can wobble between the “open” and “semi-open conformations” in the absence of 2Mg²⁺•ATP.

Naturally, the obvious question arises, whether and how the 2Mg²⁺•ATP complex interacts with the “semi-open conformation.” Therefore, 2Mg²⁺•ATP molecule has been docked into the “semi-open conformation” (Fig. 5), revealing only a single 2Mg²⁺•ATP binding site exists. This 2Mg²⁺•ATP binding site in the “semi-open conformation” was formed as a sandwich structure from both, “the Phe⁴⁷⁵ location” and “the Asp³⁶⁹ location” as they were revealed and identified at the open conformation. Both sites have approached each other due the stochastic process in the absence of 2Mg²⁺•ATP, probably due to the preceding hinge movement in the absence of 2Mg²⁺•ATP. This binding pocket for a single 2Mg²⁺•ATP is characterized by the most favorable and highest docking energy of E_b =  − 8.8 kcal/mol.

Furthermore, docking of 2Mg²⁺•ATP to the “semi-open conformation” results in a further substantial shortening of the distance between the Asp³⁶⁹ and Phe⁴⁷⁵. Consequently, the γ-phosphate of 2Mg²⁺•ATP was attracted to Asp³⁶⁹ and the mutual distance between the α-carbons of Asp³⁶⁹ and Phe⁴⁷⁵ decreased to approximately d = 1.8 nm (Figs. 2D and 3D, majority falling in the interval 1.5–2.0 nm). This shortening can be explained as a consequence of the second phase of the hinge movement: bending of the N-domain toward the P-domain, which completes “the hinge mechanism” (Fig. 6).

In order to verify our conclusions, the closed conformation of Na⁺/K⁺-ATPase structure was prepared by homology modeling as well using crystallography templates (Fig. 7). The docking experiment of 2Mg²⁺•ATP to the “closed” conformation, revealed the existence of a single ATP binding site as well (Fig. 7). Moreover, the molecular dynamic experiment with the “semi-open” sub-conformation shows that it’s conformation differs from that one of the “closed” conformation with overall RMSD < 0.3 nm, which is within the experimental error of crystallographic data.

Multiple ATP binding sites are found in the open and semi-open conformations of the cytoplasmic M₄–M₅-loop of Na⁺/K⁺-ATPase

The intention of this work was to learn by inspection of a large number of related and crystallized P-type ATPases and analogous computer modeling of the cytoplasmic M₄–M₅-loop of the human α₂ isoform of Na⁺/K⁺-ATPase, how on a molecular level the distance is shortened between the nucleotide binding site (the N-domain) and the phosphorylation site Asp³⁶⁹ at the P-domain. The distance of 3.26 nm between both sites (in the “open” state, Fig. 6) is too high to support either the Na⁺ + Mg²⁺ or the Mg²⁺-dependent transphosphorylation process or the ATP–ADP exchange reaction (Fahn, Koval & Albers, 1966), both the partial reactions of Na⁺/K⁺-ATPase. Evidently, any changes by binding of Na⁺ or K⁺ to their respective membrane sites must be excluded, since our analysis was restricted exclusively to the molecular events at the large cytoplasmic M₄–M₅-loop: We intended to understand the bending mechanism of the N-domain towards the P-domain.

ATP binding into the semi-open conformation leads to the hinge movement and triggers enzyme phosphorylation

We identified by molecular modeling of the cytoplasmic loop structure the existence of 3 different conformational states with the ability to bind ATP (Fig. 6). In the absence of ATP and 2Mg²⁺•ATP the “open” and “semi-open” conformational states are freely interconverted. The open state binds ATP in the absence of Mg²⁺ to the N-domain as previously shown (References). It may bind, however, also 2Mg2⁺•ATP at 2 sites, the “the Phe⁴⁷⁵ location” and the “the Asp³⁶⁹ location” (Fig. 1). Yet, in this open conformation, no transphosphorylation of the gamma phosphate group of ATP to Asp³⁶⁹ residing on the P-domain is possible: the terminal phosphate of ATP is much too remote from the carboxyl group of Asp³⁶⁹. Molecular modeling clearly showed that it is rather the newly identified semi-open conformation which binds 2Mg²⁺•ATP to a single site in such a way that the terminal phosphate approaches the phosphate acceptor site Asp³⁶⁹ on the P-domain leads and via a further shift to the “occluded” state may achieve its phosphorylation.

On a molecular level our model describes and is in agreement with the vast majority of published structures for the ATP binding domains of P-type ATPases (Bublitz et al., 2013; Bueno-Orovio et al., 2014; Castillo et al., 2011; Castillo et al., 2015; Fuller et al., 2013; Howland, 1991; Jacobsen, Pedersen & Jorgensen, 2002; Jensen et al., 2006; Jorgensen, Hakansson & Karlish, 2003; Jorgensen, Jorgensen & Pedersen, 2001; Jorgensen & Pedersen, 2001; Kanai et al., 2013; Kaplan, 2002; Laursen et al., 2015; MacLennan & Green, 2000; Morth et al., 2007; Nyblom et al., 2013; Obara et al., 2005; Ogawa et al., 2009; Olesen et al., 2007; Pedersen, Jorgensen & Jorgensen, 2000; Sacchetto et al., 2012; Shinoda et al., 2009; Toyoshima, 2008; Toyoshima et al., 2013; Toyoshima & Mizutani, 2004; Toyoshima et al., 2000). Importantly, Arg⁵⁴³ is located in the N domain near the interface to the P domain (Fig. 1). This residue has been shown to be essential for nucleotide binding; its substitution by Gln abolishes high-affinity binding of ATP (in the absence of Mg²⁺) and also Na⁺/K⁺-ATPase activity (Pedersen, Jorgensen & Jorgensen, 2000). The free energy required overcoming the electrostatic interactions between the γ-phosphate of 2Mg²⁺•ATP and the carboxylate groups amounts to 7.9 kcal/mol for Asp³⁶⁹. This value supports our model exactly. In addition, the increased binding energy of 2Mg²⁺•ATP is connected with a conformational transition constituting the driving force for transport of K⁺ across the membrane (Howland, 1991). Additionally, our molecular modeling experiments showed in docking experiments a very favorable binding energy of 2Mg²⁺ATP at the semi-open conformation. The strong electrostatic interaction with the negative charges of Asp³⁶⁹, Asp⁷⁰⁹ and Asp⁷¹³ with 2Mg²⁺•ATP shows that the γ-phosphate of the tightly bound ATP are important to approach the surface of the P domain in Na⁺/K⁺-ATPase (Jorgensen, Hakansson & Karlish, 2003; Jorgensen, Jorgensen & Pedersen, 2001; Jorgensen & Pedersen, 2001). This certainly leads to further bridging the gap between the N- and P-domains and the formation of a “closed conformation” (Fig. 6C) resulting in a type of “occluded 2Mg²⁺•ATP” preceding the formation of a phosphointermediate in the ATP-E₁ form of the α subunit of Na⁺/K⁺-ATPase.

Analogously, in the crystal structure of Ca²⁺-ATPase in the E₁[2Ca²⁺-] form (Clausen et al., 2013; Toyoshima et al., 2013; Winther et al., 2013), the N domain is separated from the P domain by a distance of 2.0–2.5 nm. Such a distance is also seen in our model in the “open” conformational state: the P domain of the human α₂ isoform of Na⁺/K⁺-ATPase is separated from the N domain by a distance of less than 2 nm. Additionally, Lys⁶⁹⁰ appears to create a salt linkage with the phosphate group as has been found in previous experiments. Mg²⁺ is essential for all phosphoryl transfer reactions. The experience from Mg²⁺ binding studies is that the binding affinity and the coordination pattern depend strongly on the conformational state (Pedersen, Jorgensen & Jorgensen, 2000) . Our model shows this as well. Importantly, fluctuation in between the “open” and “semi-open conformations” is connected with binding/unbinding of 2Mg²⁺•ATP to the three above mentioned negatively charged residues of Asp⁷⁰⁹, Asp⁷¹³ and Asp³⁶⁹ (Figs. 1B and 4). However, binding of 2Mg²⁺•ATP to any site of the “open conformation” cannot lead to phosphorylation. The phosphorylation process can be triggered only when the “semi-open conformation” in the absence of ATP is formed. Once the semi-open conformation has been created, by forming a hydrogen bond between Arg⁵⁴³ and Asp⁶¹¹ (Fig. 4), the affinity for ATP peaks, facilitating thus ATP binding. The distance between Phe⁴⁷⁵ and Asp³⁶⁹ decreased to about 1.8 nm (corresponding to 1.83 in our “closed conformation” model). This average value of the experimentally reported distances for the “closed conformation”, enables the phosphorylation of Asp³⁶⁹, and is in accordance with the measured data (Jacobsen, Pedersen & Jorgensen, 2002; Jorgensen, Jorgensen & Pedersen, 2001).

To the best of our knowledge, this is the first report on the existence of three conformers of the big cytoplasmic loop binding ATP. Our finding may have important consequences for understanding the molecular mechanism of the Na⁺/K⁺-ATPase function. Na⁺ ions have been reported to increase the activity of transphosphorylation process (Kaplan, 2002). It is unclear at present, where Na⁺ binds to its transport site in the transmembranal part of the enzyme and how this may affect the conformational transitions of the hinge region in the closing process approaching N- and P-domains such a way that the phosporylation of Asp³⁶⁹ as an intermediate may happen. It needs to be investigated in further studies, how, at a molecular level, high ATP concentrations lead to the release of E₂-occluded K⁺. Micromolar ATP concentrations are sufficient for ATP binding (in the absence of Mg²⁺) to the N-domain in the open state (Kubala et al., 2003; Schoner, Beusch & Kramer, 1968; Tran & Farley, 1999). The effect of Mg²⁺ on the binding of ATP to the isolated N-domain has never been studied. Micromolar ATP concentrations are sufficient for the Na⁺ + Mg²⁺-dependent formation of the Asp³⁶⁹-phosphointermediate (Hegyvary & Post, 1971; Moczydlowski & Fortes, 1981). Millimolar ATP concentrations are necessary for the overall Na⁺/K⁺-activated ATP hydrolysis necessary for Na⁺/K⁺-transport. High (millimolar) ATP concentrations are necessary to result in the de-occlusion of K⁺ from its transmembrane site. Might it be that the existence of 2 ATP sites in the “open conformation” of the cytoplasmic loop (Fig. 1) represents a situation of opening of the closed catalytic site for MgATP at high concentrations of the energy substrate. It is well known that K⁺ ions are on its way from the outside to the inside of the cell included into the transmembrane part of Na⁺/K⁺-ATPase. High concentrations of MgATP are necessary to release occludes K⁺ from the sodium pump into the cytoplasm. One may speculate that binding of MgATP at millimolar concentrations may lead to a shift of the “closed conformation” to the “open conformation,” i.e., the displacement of the N-domain via the hinge mechanism from the P-domain due to binding of millimolar MgATP to the N-domain (Fig. 8).