Structural determination of hNKA in three conformations
To solve the human NKA (hNKA) structure and investigate the cytoplasmic gating mechanism, we overexpressed and purified hNKA that contained α1, β1 and γ2 subunits (Supplementary Fig. 2a). To trap hNKA in the E1 state, only Na+ and Mg2+ were added in buffer when purification16. In the same way, existence of only K+ in solution can stabilize the conformation in E2 state16. We determined the cryo-EM structures of hNKA in three conditions: 1) 150 mM NaCl, 3mM Mg2+, 1 mM ATP analogue (adenosine 5’-O-(3-thio) triphosphate, ATPγS), 2) 150 mM NaCl, 3mM Mg2+, and 3) 100 mM KCl, 3mM Mg2+ (Supplementary Fig. 2-4 and Supplementary Table 1). The cryo-EM structures solved in these conditions represent E1·3Na·ATP, E1·3Na, and E2·[2K] states with resolution of 3.1, 3.4, and 3.1 Å respectively, good enough for model building and subsequent structural analysis.
Overall structure of the E1 state preceding ATP hydrolysis
We successfully solved the cryo-EM structures of hNKA in the E1·3Na and E1·3Na·ATP state at 3.4 Å and 3.1 Å resolution in the absence or presence of a slowly hydrolyzed ATP analogue, respectively, representing two E1 structures of hNKA preceding ATP hydrolysis previously unknown according to our knowledges. Intriguingly, we observed an opening K+/Na+-exchanging path in our E1 state structures between transmembrane Na+-binding site and cytoplasm (Fig. 1, Supplementary Movie 1).
The structures of the E1·3Na·ATP and E1·3Na states are virtually identical with a root mean square deviation (RMSD) of 0.348 Å, except a slightly difference in ATP binding pocket in the cytoplasmic headpiece and the Na+-binding sites in transmembrane region of α1 subunit (Fig. 2 and Supplementary Fig. 5). We found that the configuration of cytosolic headpiece of NKA changes in E1 state. Both the E1·3Na·ATP and E1·3Na structures are in open headpiece conformation, in which the A domain is separated from the N domain and stabilized only by the eighth helix at the end of the P domain (Pα8), (Supplementary Fig. 5a, asterisks location). In the E1~P·[3Na]·ADP state (PDB code: 3WGU), in which ATP is hydrolyzed to ADP but not released, the A domain interacts with the N domain with two additional sites (Supplementary Fig. 5a, plus symbols), stabilizing the closed conformation of the headpiece14. The movement and rotation of the A domain pull up the M1 helix after phosphorylation, to close the cytoplasmic gate (Fig. 3a).
In E1·3Na·ATP and E1·3Na states, NKA have three Na+-binding sites I, II and III, which are located in the middle of the membrane (Fig. 1 and 4a-c). As compared with the E1~P·[3Na]·ADP state of pig NKA, the positions of three Na+ are different (Fig. 4d-f), of which site Ⅰ and Ⅱ locations are changed largely and site Ⅲ seems very restricted throughout the E1 basic state (Fig. 4e,f; Supplementary Fig. 5b, c). During the E1·3Na/E1·3Na·ATP to E1~P·[3Na]·ADP transition, Na+-binding sites move toward the depth of cation binding cavity (Fig. 4e, f).
The cytoplasmic gate closure coupled with ATP hydrolysis
In the physiological situation, the binding of ATP accelerates the transition from E2·[2K] to E1 state4,5. During this process, the cytoplasmic gate of NKA must open for releasing K+ to the cytoplasm and then fill transmembrane Na+-binding sites sequential with three intracellular Na+ 14,17. Then the cytoplasmic gate close and three Na+ are occluded within the NKA. Our samples reveal two similar E1 states (E1·3Na and E1·3Na·ATP) structures preceding ATP hydrolysis with opening cytoplasmic pathway (Fig. 1, 2, Supplementary Fig. 5b). Compared to the E1·3Na·ATP state, large conformational changes are observed in the E1~P·[3Na]·ADP state (Fig. 3a). The closing of the cytoplasmic headpiece induces the closing of the M1 sliding door and occlusion of the bound Na+. ADP has a smaller pocket by the tilt with N and P domain aligning with PN (Fig. 3b, c). These changes bring the A domain further toward the N domain and close the headpiece (Fig. 3a, Supplementary Fig. 5a). As a result, gate closing signal would be transmitted to the A domain sit on Pα8 helix and make it rotates about 26° to pull up the M1 sliding door (Fig. 3a red arrow). Notably, M1e is moved towards the cytoplasmic side for about 5 Å, and close NKA cytosolic door directly (Supplementary Fig. 5c).
ATP binding and hydrolysis adjust the sodium ion binding sites
E1·3Na and E1·3Na·ATP are largely similar. However, it seems evident that the reaction with ATP causes different extents of rearrangements of the three Na+. Na+ in site Ⅰ and Ⅱ of E1·3Na and E1·3Na·ATP states are juxtaposed at the interhelix space between M4-M6, and closer to the cytoplasmic surface (Fig. 4a-c). We observe a gate residue Glu334 on M4e at the door, whose orientation of side chain tends to open and close accompanied by M1 sliding door conformational changes (Fig. 4d-f). Glu334 (Glu327 in pig) only coordinate site Ⅱ Na+ to help close the cytosolic door when ATP hydrolysis to ADP (Fig. 4d). Glu334 is highly conserved among P2-ATPases (Supplementary Fig. 10). The side chains of Asp811 and Asp815 (Asp804 and Asp808 in pig) in M6 are play an important role in site Ⅰ and Ⅱ formation. Asp811 side chain coordinates site Ⅰ Na+ at E1·3Na state but stabilizes site Ⅱ Na+ when it moves deeper in E1·3Na·ATP state (Fig. 4b,c). Obviously, the Na+ of site Ⅲ is the initial binding at the deepest position of funnel shaped Na+-binding cavity, and site Ⅱ Na+ is the last one binding near the gate. This Na+ ions binding sequential scenario is in good agreement with the hypothesis in pig NKA E1~P·[3Na]·ADP state structure studies14.
All Na+-binding sites are slightly offset to the M5 and M8 side during E1·3Na/E1·3Na·ATP to E1~P·[3Na]·ADP transition, raising the question how the ATP binding and hydrolysis affect the Na+-binding site. Indeed, ATP and Na+ bind to spatially distant locations from cytosolic headpiece to transmembrane domain. The binding of ATP induces rotation of the N domain for about 7° (Fig. 2a) with Arg692 on the short loop connecting the third strand (Pβ3) and the sixth helix (Pα6) functioning as a pivot (Supplementary Fig. 6a). The adenine ring of the bound nucleotide is stacked with Phe482, and interacts with Glu453, Asp450 and Arg551 (Fig. 2d). The γ-phosphate of ATPγS inserts into the P domain, stabilized by a cofactor Mg2+, Thr378 and Gly618 (Fig. 2b, d), decreasing distance between the N domain and the P domain. As a result, ATP analogue is delivered to the phosphorylation site with a proper orientation that facilitates phosphoryl transfer. In E1·3Na, Mg2+ stabilizes the important 376DKTGTLT motif and the 715TGDGVND motif (Fig. 2c). The ATP binding changes the location of cofactor Mg2+ (Fig. 2e), disrupting the hydrogen bond network formed with these two motifs. These changes make 376DKTGTLT motif loosely and move subtly (Fig. 2e arrow; Supplementary Fig. 6b), which is connected to M4 through Pβ1, Pα1, Pβ0 and interacts with Pβ6-Pα8, Pβ7-M5 loops near M5 (Supplementary Fig. 6b, c). Therefore, ATP binding induce conformational changes in 376DKTGTLT motif and affects the binding of cofactor Mg2+ and might trigger the orientation of M4 and M5 and adjust the position of the Na+-binding sites. The binding of ATP coupled with the movement of M4 and M5 might result in the displacement of Na+-binding site (Supplementary Fig. 6c). The rearrangement three Na+-binding site to move toward the depth of cation binding cavity is to prepare for the next step to close the inner cell door (Fig. 4). After binding of three Na+ and ATP molecule, M1 sliding door rearranges to a position that blocks the cytoplasmic entrance pathway.
K+-occluded E2 basic state
In addition to the E1·3Na·ATP and E1·3Na structures described above, we also obtained a cryo-EM structure of hNKA in the E2·[2K] state at 3.1 Å resolution (Supplementary Figs. 1), which is a basic state following dephosphorylation with two K+ ions in occluded conformation (Fig. 5a, b). A cytoplasmic K+ site (Site C) is the third K+-binding site (Fig. 5a, red circle), which is implicated in activation of dephosphorylation18. As compared to the E2·[2K]·Pi state (a preceding state in Post-Albers cycle, Supplementary Fig. 1; PDB code: 3KDP)8, the two structures are very similar with an RMSD of 1.087 Å ( Supplementary Fig. 7a). The K+-binding sites Ⅰ and Ⅱ between the M4, M5 and M6 are almost identical (Supplementary Fig. 7b). K+-coordinating residues at site I (Asp815, Asp811, Ser782, Glu786 and Asn783) and site Ⅱ (Asp811, Glu334, Val332, Val329, Ala330 and Asn783) are similar to that of Na+-binding site I and Ⅱ in corresponding E1·3Na·ATP state (Fig. 2b, Supplementary Fig. 7b, 8b). The 219TGES motif of A domain, which plays important role in dephosphorylation of E2P to E2 transition19-21, is further stabilized by Mg2+ and 376DKTGTLT motif in P domain (Fig. 5c, d, Supplementary Fig. 7c, d). Compared with E1·3Na·ATP state, the α1 subunit undergoes a large conformational change (Supplementary Fig. 8). In TM region, M1 to M6 rotate toward the opposite side for about 20°, whereas M7 to M10 are relatively rigid. When the PN domain is aligned, the N domain tilts for 99°, and the A domain rotates for about 71° relative to E1·3Na·ATP state (Supplementary Fig. 8a). From the E2·[2K] to E1·3Na·ATP state, the 219TGES motif moves for 11.7 Å to expose the conserved Asp376 phosphorylation site (Supplementary Fig. 8c).