A unique gating mechanism revealed by the cryo-EM structure of monomeric ATP9A flippase

Abstract

Among mammalian P4-ATPase flippases, only ATP9A and ATP9B do not require the auxiliary subunit CDC50 protein. Whilst its yeast homolog, Neo1, is essential for cell survival, little is known about mammalian ATP9A. We present cryo-EM structures of human monomeric ATP9A at a resolution reaching 2.2 Å, in the outward-facing E2P state. Two distinguishable conformations were obtained from a single sample, one with its outward gate open and the other in its closed form. Unlike canonical gating observed for most P-type ATPases, which is driven by the movement of transmembrane (TM) helices 1 and 2 linked to the A domain, outward gating in ATP9A is achieved by the movement of TM6-10 helices, likely initiated by the unwinding of TM6. As a result, the volume of the phospholipid binding cavity in the open state surpasses that of other flippases, which could allow binding of phospholipids with larger hydrophilic headgroups than that of phosphatidylserine. ATP9A shows an ATPase activity that is significantly increased by the addition of phospholipids that retain the overall negative charge, including phosphatidylserine, phosphatidylinositol, and its phosphorylated species, compared with other electroneutral phospholipids. The observation of spontaneous binding of phosphorylated species of phosphatidylinositol in molecular simulation reinforces this fact. Our data provide mechanistic rationales for ATP9A gating, achieved by the rearrangement of the second half of the TM helices. Since TM4-TM10 is anchored by the CDC50 protein subunit in other flippases, the here-observed outward gating mechanism is unique to P4B-type flippases, which function as a monomer.

Keywords: ATPase; cryo-EM; lipid transport; membrane protein; molecular dynamics simulations; phospholipid; transporter.

Figure 1

Cryo-EM structures of closed and open forms of human ATP9A in E2P state. Overall structures in the closed (A, gray cartoon and surface, PDB ID: 9VDL) and open (B, blue cartoon and surface, PDB ID: 9VDK) forms found in the BeF-bound E2P state viewed from the membrane plane (dotted lines) with cytoplasmic side up. Transparent surfaces represent Gaussian-filtered low-contoured maps showing molecular envelopes and micelles. A-, P-, and N-domains are indicated. BeF-bound catalytic Asp391 are shown as green spheres. Bound phospholipids (PLs) and cholesterol (CLR) in the open form are indicated with yellow sticks and surface. Inset in B represents a close-up of the phosphorylation site showing the density map (transparent surface) within 4 Å of Asp391, BeF, and the DGET loop from the A-domain. BeF, beryllium fluoride; PDB, Protein Data Bank.

Figure 2

Comparison of the molecular conformation between open and closed forms.A, superimposition of whole structures (A) of closed (gray, PDB ID: 9VDL) and open (blue, PDB ID: 9VDK) forms are shown as cartoon representations. B and C, close-up views of the TM helices from the parallel (B) or luminal side (C) of the membrane plane. Bound PS (yellow) and some of the important residues are shown as sticks. Arrows indicate the movement from closed to open forms. D and E, structure of TM6 in closed (D) and open (E) states with EM density within 4 Å from the model. EM densities for bound PS in the canonical binding site (green surface) and other lipids (yellow surface) and cholesterol (CLR) are also shown. F, a close-up of the canonical phospholipid-binding site in the open form is shown, viewed from the position where TM2 (shown as a transparent helix) is located. Red spheres represent water molecules. Dotted lines are connecting atoms within 3.5 Å distance. G, canonical phospholipid-binding site in the E2P state of ATP11C (pink, PDB ID: 7BSU) from the viewpoint similar to (F). H and I, surface representations of the TM regions in ATP9A (H) and the ATP11C–CDC50A complex (I) with their electrostatic potentials (blue: positive, red: negative). PS, phosphatidylserine; TM, transmembrane.

Figure 3

Role of Arg849 in TM5 upon gating.A–D, close-up of phospholipid-binding site in ATP9A E2P closed form (A, PDB ID: 9VDL), its open form (B, PDB ID: 9VDK), ATP11C PS-bound E2P state (C, PDB ID: 7BSU), ATP8B1 PI-bound E2-Pi state (D, PDB ID: 8OXC) and Neo1 E2P state (D, PDB ID: 7RD6), viewed from the luminal/extracellular side. Some of the important residues are shown as sticks and indicated in the figures. In C–E, their corresponding amino acids in ATP9A are indicated in parentheses. Dotted lines are connecting atoms within 3.5 Å distance. E, the ATP9A (blue ribbons) open form is superimposed on the complex of ATP11C (pink) and CDC50A (gray). Exoplasmic loops of ATP11C, which form intimate interactions with CDC50A, are shown as orange ribbons. PDB, Protein Data Bank; PS, phosphatidylserine; TM, transmembrane.

Figure 4

Dimensions of the phospholipid-binding site in different flippases. Amino acid models of indicated flippases are shown as surface representations. Their whole structures viewed from the membrane plane are shown in the upper panel. Membrane slices, viewed from the luminal/extracellular side of the membrane, at the position indicated with an orange dotted line, are shown in the lower panel. Bound phospholipids are shown as spheres. Black arrowheads indicate the phospholipid-binding site. PDB codes for each structure are indicated in the figure. PDB, Protein Data Bank.

Figure 5

Specific ATPase activities of purified ATP9A in the presence of phospholipids. ATP9A is purified in the absence of phosphate analog and measured its ATPase activity in the absence (W/O) or presence of indicated phospholipids added (final concentrations of 20 μM). Background ATPase activity is subtracted by setting the BeF-inhibited sample as a blank. Data plotted are mean ± SD from three independent experiments (∗p < ∗0.01 and ∗∗p < 0.0001, one-way ANOVA, compared with ATPase activity without phospholipids, W/O). BeF, beryllium fluoride; W/O, without.

Figure 6

Molecular simulations.A, simulation snapshot from the coarse-grained simulations showing a PI(3,5)P₂ lipid bound to the lipid-binding site. B, radial distribution functions (RDFs, normalized distance histograms) between the indicated lipid headgroup and the center of mass (COM) of the binding site residues (Gln92, Leu99, Thr104, Tyr105, Ser340, Asn341, Ile345, Ser346, Arg849, Ser850, Gln857, Tyr872, and Thr882) of the protein. The curves correspond to simulations in the presence of PI(3,4,5)P₃ (black), PI(3,5)P₂ (red), and PI(4,5)P₂ (green), respectively. The strongest interactions are observed for PIPs, which is the only lipid showing significant peaks at a distance of less than 1 nm, indicating strong interactions of the headgroup with the binding site. PI(3,4,5)P₃ (black curve) has the strongest interactions with the binding site. Single representative simulation replicas were used for these plots. The data show that PIPs are most likely to bind to the lipid-binding site. C, after finding that PIPs are most likely to bind the protein, we placed both PI(3,5)P₂ and PI(3,4,5)P₂ in the membrane in the same simulation setup and simulated 10 replicas of this system. Each color in the panels in plot C represents the RDF for a single replica. The data show that the lipid-binding site has no specific preference for PI(3,4)P₂ or PI(3,4,5)P₃. See Table S2 for the lipid composition of the membranes. D, representative snapshots from all-atom simulation for PI(3,4,5)P₃ (left) and DOPS (right) bound to the phospholipid-binding site, viewed from the luminal side. Potential hydrogen bonds are shown by connecting atoms within 3.5 Å distance. Amino acids within 4 Å distance from either of the phospholipids are shown as sticks as candidates for van der Waals interaction. E, RDFs between the side chains of binding site residues, which interact most with PI(3,4,5)P₃ and DOPS in the all-atom simulations. The RDFs between indicated residues and the closest phosphate atom of the PI(3,4,5)P₃ headgroup (left) or the carboxylate carbon atom of the PS headgroup (right) were plotted, respectively. The different colors correspond to interactions of different amino acids. F, the interaction between Arg849 and Thr885 quantified using RDFs. The RDF is constructed between the hydroxyl oxygen of the Thr885 side chain and the terminal carbon atom of the Arg849 side chain. All RDFs for the all-atom simulations are collected from the last 200 ns of three simulation replicas each for DOPS and PI(3,4,5)P₃. DOPS, 1,2-dioleoyl-sn-glycero-3-phospho-l-serine; PI(3,5)P₂, phosphatidylinositol 3,5-bisphosphate; PI(4,5)P₂, phosphatidylinositol 4,5-bisphosphate.