Cryo-EM structure of KdpFABC in nanodiscs.

(A) Overview of the density map with subunits rendered in conventional colors corresponding to those in panel B. The Post-Albers reaction scheme is shown in the inset. (B) Diagram illustrating the transmembrane topology of KdpFABC. (C) Rendering of the unsharpened map at a lower density threshold illustrates the presence of the membrane scaffolding protein surrounding the nanodisc. (D) Detail of the newly resolved N-terminus of KdpB revealing a potential interaction between Glu161, Arg3, Asp552 and Lys557. (E) Detail of the nucleotide binding site of KdpB showing ADP associated with two Mg2+ ions and the phosphorylated catalytic residue: Asp307. Panels A, C and D show the unsharpened map at thresholds of 7, 3 and 4 σ; panel E shows the sharpened map at 3.9 σ.

Biochemical preparation of KdpFABC.

(A) SDS gel of nanodisc sample used for cryo-EM showing the presence of KdpA, KdpB, KdpC and the membrane scaffolding protein, spNW25. KdpF is not visible due to its small size (3 kD); typically, it stains poorly and runs very near the die front of the gel. (B) Elution profile from FPLC size-exclusion column of the sample used for making cryo-EM grids; the main peak is composed of nanodiscs containing KdpFABC and the lower peak contains empty nanodiscs. Slight asymmetry of the main peak is consistent with the presence of single- and double-occupancy nanodiscs seem during image processing. (C-D) MALS analysis during HPLC elution of nanodiscs reconstituted with POPC/DOPA lipids and E. coli polar lipids, respectively. This analysis indicates DOPC/DOPA lipids produce smaller nanodiscs containing ∼1 KdpFABC molecule (panel c), whereas E. coli polar lipids produce larger, more heterogeneous nanodiscs (panel d). (E) ATPase activity is preserved after nanodisc reconstitution, though at a reduced rate attributable to the lipid environment. (F) Mg2+ dependence of ATPase activity indicates a Hill coefficient (n) of ∼1.8, which is consistent with the two Mg2+ ions associated with ADP in the structure. (G) Example of transport current measured by solid supported membrane electrophysiology on WT KdpFABC. (H) Inductively coupled plasma mass spectrometry analysis of stock solutions used for ATPase assay indicate substantial K+ contamination from ATP and MgCl2 solutions. These data represent the mean and std. dev. of triplicate measurements.

Cryo-EM processing pipeline.

Four grids from a single sample were imaged to produce a total of ∼50,000 micrographs. These were divided into ten groups of 2000-6000 micrographs for initial processing. After motion correction and determination of defocus parameters, particles were picked using TOPAZ and extracted with a box size of 335Å and 3-fold binning. Multiple rounds of 2D classification (a few examples of 2D classes are shown) and ab initio reconstruction were used to remove false positive picks and empty nanodiscs. A subset of these particles were further classified to generate four main classes representing single occupancy nanodiscs with E1 and E2 conformations, double occupancy nanodiscs which were mainly E1 conformation, and junk. These low-resolution structures were used as references for hetero-refinement of particles partitioned in seven groups. Particles conforming to each conformation were combined and used for masked 3D classification, where the mask encompassed the cytoplasmic domains of KdpB. Particles representing the E1∼P·ADP conformation, which represented the largest class, were combined for a final non-uniform refinement job.

Cryo-EM densities.

Correlation between map densities and model for individual transmembrane helices from KdpA as well as transmembrane helices and cytoplasmic domains of KdpB. The sharpened map was used for helices and the P-domain with a threshold of 5.7 σ, whereas the unsharpened map was used for the more flexible A- and N-domains at 2.5 σ.

Structure determination of KdpFABC.

Selectivity filter of KdpA.

(A) The map reveals densities at all sites (S1-S4) of the selectivity filter. Because the density at the S3 site is the highest in the entire map (Fig. 2 - figure suppl. 1a), it has been assigned as a K+ ion (purple sphere). Grey mesh corresponds to the sharpened map at 6.5 σ. (B-C) ATPase and transport assays (respectively) illustrate that Q116R and G232D mutations lower the apparent affinity of the pump (Km values in mM shown in the legend). Background activity is consistently observed for the WT, which is likely due to K+ contamination in stock solutions (Fig. 1 – figure suppl. 1h). The location of Gln116 is shown in panel a, G232 is identified in Fig. 2 – figure suppl. 1b. (D) Electrostatic charge at the entrance to the selectivity filter of the WT pump shows strong negative charge. A K+ ion is visible in the selectivity filter as a purple sphere. (E-F). Substitution of Glu116 has differing effects that reflect changes in net charge at the entrance to the selectivity filter. (G-I) Transport assays with various monovalent cations show that the strong selectivity of the WT pump is compromised by the G232D mutation but not by the Q116R mutation. Table 2 shows that measurements from transport assays were broadly consistent with these ATPase data.

Selectivity filter of KdpA.

(A) Distribution of raw densities for the entire map. The arrow indicates the threshold used for rendering the map in Fig. 2 (0.5 which corresponds to 6.5 σ). The inset shows the extreme tail in the density distribution; the highest four densities (circled) correspond to the K+ ion modeled at the S3 site. (B) Coordination of densities in the selectivity filter in the refined model. The regular network of bonds at the S3 site suggest that this is the high affinity binding site for K+, whereas less regular coordination at other sites is consistent with water.

Summary of apparent affinities of the selectivity filter mutations measured by ATPase and transport assays.

Vestibule of KdpA.

(A) The map reveals numerous spherical densities immediately below the selectivity filter and leading into the tunnel that connects KdpA with KdpB. Glu370 and Arg493 are the only charged residues in the transmembrane domain of KdpA and reside on either side of this vestibule. Grey mesh corresponds to the sharpened map at 6.5 σ. (B) ATPase assays of E370Q and E370H show that neutralization of charge on Glu370 reduces the apparent affinity without affecting Vmax, but that reversing the charge with E370K almost completely abolishes activity. The double mutant E370K/R493E partially restores activity, but with much lower apparent affinity and Vmax. (C) ATPase assays show that mutation of Arg493 has relatively modest effects on apparent affinity without affecting Vmax. Note that the axis with Km is logarithmic. Table 3 shows that measurements from transport assays were broadly consistent with these ATPase data.

Summary of apparent affinities of the vestibule mutations measured by ATPase transport assaus.

Tunnel connecting KdpA and KdpB.

(A) The intersubunit tunnel is shown as a transparent blue surface running from the selectivity filter on the right (S3) to the canonical binding site on the left (near N624). Non-protein densities within the tunnel have been modeled as water molecules (red spheres). Asp370, Arg493 in the vestibule and Phe232 at the subunit interface are shown as space-filling models and K+ as purple spheres. As in other figures, KdpA helices are green and KdpB helices are brown; KdpF is cyan. (B) The distribution of water molecules within the tunnel shows that the vestibule is well populated whereas the subunit interface is quite hydrophobic. Values along the x-axis correspond to the contour length along the tunnel. (C) Tunnel profiles for the experimentally determined model (WT) and predicted models for the Val496 mutants. The horizontal dotted line at a radius of 1.4 Å represents the conventional limit for passage of water or K+ ions. Note that both V496R and V496W constrict the tunnel. SF and CBS indicate locations of the selectivity filter and canonical binding site, respectively. (D) ATPase and transport assays of the Val496 mutants in 150 mM K+ showing that introduction of positive charge (V496R) abolishes activity.

Canonical binding site in KdpB.

(A) Several non-protein densities are visible at the canonical binding site, the most prominent of which occupies the conserved Na2/Ca2 site from Na,K-ATPase and SERCA (figure suppl. 1). This site has the highest density in KdpB (figure suppl. 2b) and thus has been modeled as K+ (purple sphere), whereas the others are modeled as water (red spheres). Grey mesh corresponds to the sharpened map at 5.8 σ. (B) Asp583 and Lys586 are the only charged residues in the transmembrane domain of KdpB. ATPase and transport assays in 150 mM K+ show that mutation of these residues is generally not tolerated, except for the D583A mutation which generates an uncoupled phenotype (ATPase activity without transport).

Canonical binding sites of P-type ATPases.

The primary K+ binding site in KdpB is structurally conserved with ion binding sites for Na+ and Ca2+ in Na,K-ATPase and SERCA. Whereas multiple ions are transported in each cycle by these related P-type ATPases, only a single ion appears to bind to KdpB at a position equivalent to the Na2 and Ca2 sites. Like those sites, the K+ in KdpB is coordinated by carbonyl oxygens exposed by the unwinding of the M4 helix near a conserved proline (figure suppl. 2). Additional ions bind in a pocket between M4, M5 and M6 in Na,K-ATPase and SERCA, which is occupied by weaker densities in KdpB that are modeled as water.

Properties of the canonical binding site in KdpB.

(A) Coordination network involving the K+ and water molecules in the canonical binding site. (B) Distribution of raw densities for KdpB portion of the map, with the inset highlighting the highest two densities, both of which correspond to the K+ ion modeled at the canonical binding site (CBS). The arrow indicates the density threshold used for rendering the map in Fig. 5 (0.45 which corresponds to 5.8 σ). (C) ATPase activity of the KdpB-D538A mutant showing uncoupled activity at low K+ that is inhibited at higher K+ concentrations.

Exit site from KdpB.

(A) The map shows a water filled cavity next to Thr75 that leads to the cytoplasm (sharpened map at 6.5 σ). (B-C) pH dependence of the T75D mutant is shifted relative to WT, consistent with the idea that high pH generates negative charge in this cavity and inhibits K+ release. (D) The T75K substitution not only interferes with transport, but also produces an uncoupled phenotype in which ATPase activity is not dependent on K+. At high pH there is a small recovery of transport activity as well as K+-dependence of ATPase activity, suggesting that these effects result from the positive charge from the lysine substitution.

Exit site for K+ from KdpB.

(A) Comparison of the canonical binding site from E1∼P (this study) and E2·Pi determined previously by Sweet et al. (Sweet et al., 2021) (PDB code 7BGY). In particular, movements of M5 cause Lys586 to swing into the canonical binding site in E2·Pi, thus displacing the K+. (B) Comparison of the exit site from wild-type E1∼P model with a predicted model for the T75K mutant. In this predicted structure, the lysine side chain swings into the canonical binding site, thus potentially acting as a built in ion to stimulate ATPase activity in the absence of K+. (C) pH dependence of both ATPase activity and transport from the KdpB-L72D mutant. Like the T75D mutant, the pH dependence of activity is shifted towards low pH. Although the rates are lower than WT and T75D, the K+ dependence of ATPase activity indicates that energy coupling is intact.

Conduction pathway of KdpFABC.

Diagram illustrating features of the conduction pathway seen in the E1∼P·ADP map together with residues that have been mutated in this study. Ions (purple spheres) are attracted from the periplasm to the mouth of the selectivity filter of KdpA by negative surface potential. Selectivity and affinity of the pump is governed by tight coordination of the ion at the S3 site. Ions become rehydrated upon release into a vestibule filled with water (red spheres) next to charged residues Glu370 and Arg493. Val496 resides in the widest part of the tunnel where introduction of positive charge abolishes transport. Phe232 resides at the subunit interface where there is a hydrophobic barrier that is essential to ion occlusion. K+ binding to the canonical binding site next to Pro264 on M4 in KdpB triggers phosphorylation of Asp307 followed by allosteric movements of M5 that drive Lys586 to displace the ion into the exit site next to Thr75. A general lack of ligands in this exit site is consistent with low affinity, thus promoting K+ release to the cytoplasm against the electrochemical gradient.