Research Article

Cryo-EM structures capture the transport cycle of the P4-ATPase flippase

See allHide authors and affiliations

Science  13 Sep 2019:
Vol. 365, Issue 6458, pp. 1149-1155
DOI: 10.1126/science.aay3353

Flipping a lipid

The membranes of eukaryotic cells have different lipid compositions in their inner and outer leaflets. Enzymes known as flippases and floppases use the energy from adenosine triphosphate (ATP) hydrolysis to translocate lipids against a concentration gradient from the outer to inner or inner to outer leaflets, respectively. Flippases are P4-type ATPases that are important in processes such as membrane trafficking, signaling, and apoptosis. Hiraizumi et al. report the cryo–electron microscopy structure of six intermediates of the human flippase ATP8A1 bound to the partner protein it requires for function, CDC50. ATP binding and autophosphorylation of ATP8A1 drive a cycle of conformations in which lipids bind differently, powering translocation.

Science, this issue p. 1149

Abstract

In eukaryotic membranes, type IV P-type adenosine triphosphatases (P4-ATPases) mediate the translocation of phospholipids from the outer to the inner leaflet and maintain lipid asymmetry, which is critical for membrane trafficking and signaling pathways. Here, we report the cryo–electron microscopy structures of six distinct intermediates of the human ATP8A1-CDC50a heterocomplex at resolutions of 2.6 to 3.3 angstroms, elucidating the lipid translocation cycle of this P4-ATPase. ATP-dependent phosphorylation induces a large rotational movement of the actuator domain around the phosphorylation site in the phosphorylation domain, accompanied by lateral shifts of the first and second transmembrane helices, thereby allowing phosphatidylserine binding. The phospholipid head group passes through the hydrophilic cleft, while the acyl chain is exposed toward the lipid environment. These findings advance our understanding of the flippase mechanism and the disease-associated mutants of P4-ATPases.

In eukaryotic cells, the phospholipid compositions differ between the outer and inner leaflets of the plasma and organellar membranes: phosphatidylcholine (PC) and sphingomyelin are enriched in the outer leaflet, whereas phosphatidylserine (PS) and phosphatidylethanolamine (PE) are confined to the inner leaflet (1). The maintenance and disruption of the asymmetric composition affect processes, such as membrane biogenesis, membrane trafficking, signaling, and apoptosis. Three types of transporters—scramblase, floppase, and flippase—have been reported to function as phospholipid translocators (24). Scramblases catalyze bidirectional phospholipid translocations that dissipate the membrane asymmetry without energy consumption. In contrast, flippase and floppase, which use the energy of adenosine triphosphate (ATP) hydrolysis to mediate specific phospholipid translocations against their concentration gradients, maintain the asymmetric phospholipid composition. ATP-binding cassette transporters function as floppases that drive the inner-to-outer translocation of lipids, whereas type IV P-type ATPases (P4-ATPase) are flippases that drive the outer-to-inner translocation of lipids.

The transport by P-type ATPases occurs essentially according to the Post-Albers mechanism (5, 6), wherein ATP hydrolysis–coupled phosphorylation and dephosphorylation within the cytoplasmic ATPase domain mediates the transition between two intermediate states, E1 and E2, which have different affinities for the substrates, enabling the substrate transport across the membrane (fig. S1A). Among the P-type ATPase family members, the P1- to P3-ATPases are ion transporters, with the most representative being the P2-ATPase family, which includes the sarcoplasmic reticulum Ca2+ pump (SERCA), Na+/K+-ATPase, and H+/K+-ATPase. The P4-ATPase is the only member that functions as a lipid transporter (7). The reaction of the P4-ATPases also follows the Post-Albers mechanism, but it simply translocates phospholipids from exoplasmic leaflet to cytoplasmic leaflet and does not require any counter-transported substrate (fig. S1B) (8). The human genome encodes 14 P4-ATPase subclasses, which differ in their lipid selectivities and tissue expression (9). For most P4-ATPases, heterodimerization with a CDC50 family protein is essential for proper expression and flippase activity (10, 11).

The first P4-ATPase member identified, ATP8A1, was found in bovine erythrocytes and chromaffin granules as an aminophospholipid translocase (12, 13). P4-ATPases, including ATP8A1, are present in plasma and organellar membranes and sequester PS lipids from the outer to the inner leaflet in resting cells. In apoptotic cells, P4-ATPases are cleaved and inactivated by proteases such as caspases and calpains. The PS that is subsequently exposed on the cell surface acts as an “eat me” signal to induce phagocytosis (14, 15). Furthermore, the ATP8A1-catalyzed flipping of PS in the organellar membrane is necessary for the transport of recycling endosomes, membrane fission, and cell migration (16, 17). Several diseases are associated with P4-ATPases. For example, ATP8B1 mutations cause the liver diseases known as benign recurrent intrahepatic cholestasis 1 and progressive familial intrahepatic cholestasis 1, ATP10A is associated with type 2 diabetes and insulin resistance, and ATP11A is associated with cancer (18). Furthermore, ATP8A1 and ATP8A2 have been identified as causative genes for neurological disorders. ATP8A1 knockout mice show hippocampus-dependent learning deficits associated with the exposure of PS on the outer surface of the plasma membrane in hippocampal neurons (1921).

As compared with the canonical ion-transporting P-type ATPases, P4-ATPase has a large transport substrate and thus is expected to use a different mechanism for substrate recognition and translocation (2224). However, despite substantial efforts, the molecular mechanism underlying the lipid flippase activity by the P4-ATPases has remained elusive. Here, we report the cryo–electron microscopy (cryo-EM) structures of the human ATP8A1-CDC50a heterodimer complex in its six distinct intermediates: an apo state (E1), the nonhydrolyzable ATP analog β,γ-methyleneadenosine 5′-triphosphate (AMPPCP)–bound state (E1-ATP), the adenosine diphosphate–inorganic phosphate (ADP-Pi) analog AlF4-ADP–bound state (E1P-ADP), the phosphate-analog AlF4-bound transient phosphorylated state (E1P), the BeF3-bound phosphoenzyme ground state (E2P), and the AlF4-bound dephosphorylation state with the substrate phospholipid (E2Pi-PL), revealing the transport cycle along the lipid flipping reaction.

Overall structure

We performed a cryo-EM analysis of the P4-ATPase lipid translocator family to elucidate the lipid translocation mechanism (Fig. 1). We expressed full-length human ATP8A1 and human CDC50a together in mammalian human embryonic kidney–293F cells and purified the complex in glycol-diosgenin (GDN) micelles (fig. S1C). SDS–polyacrylamide gel electrophoresis analysis showed higher molecular weight bands of CDC50a, probably derived from multiple glycosylations (fig. S1D) (25, 26). The purified ATP8A1-CDC50a complex showed PS-dependent ATPase activity, with a Michaelis constant Km of 111 ± 26.4 μM and a maximum rate of reaction Vmax of 99.7 ± 9.50 nmol min−1 μg−1, as well as weak PE-dependent ATPase activity (Fig. 1C), consistent with previous reports (13, 17). The ATPase activity was inhibited by general inhibitors of P-type ATPases, such as beryllium fluoride (BeF3) and aluminum fluoride (AlF4) (fig. S1F). The purified ATP8A1-CDC50a complex was subjected to cryo-EM single-particle analyses under several different conditions; namely, without any inhibitors and in the presence of AMPPCP, ALF4-ADP, BeF3, and ALF4 (fig. S1B). The acquired movies were motion-corrected and processed in RELION 3.0 (27), which provided cryo-EM maps at overall resolutions of 2.6 to 3.3 Å, according to the gold-standard Fourier shell correlation 0.143 criterion (figs. S4 to S8). The flexible cytoplasmic ATPase domain is most stabilized in the AlF4-ADP and BeF3-bound states, allowing the de novo modeling of almost the entire ATP8A1-CDC50a complex, except for some minor disordered regions (Fig. 1, A and B, and fig. S9). The overall structure shows the typical P-type ATPase fold, composed of three large cytoplasmic domains (A, actuator; N, nucleotide binding; P, phosphorylation) and ten membrane-spanning helices (M1 to M10). CDC50a has two transmembrane helices (TM1 and TM2) at the N and C termini, an ectodomain consisting of an antiparallel β-sandwich (β1 to β8), and extensions of ~60 and 70 amino acids with less secondary structure in the β3-β4 and β5-β6 loops, respectively, which are stabilized by two intrachain disulfide bonds (fig. S3D). The three N-linked glycosylation sites of CDC50a, which are important for the proper folding as well as the membrane trafficking of the P4-ATPases, are clearly visible in the cryo-EM map (fig. S3, D and E) (11, 28).

Fig. 1 Biochemical and cryo-EM studies of the ATP8A1-CDC50a complex.

(A) Topology diagram of ATP8A1-CDC50a. Conserved domains and TM helices are schematically illustrated. In the cytoplasmic regions, the A, N, and P domains and the C-terminal regulatory domain are colored yellow, red, blue, and green, respectively. M1-M2 and M3 to M10 of ATP8A1 are purple and orange, respectively, and CDC50a is pink. The N-glycosylation sites are shown as sticks. cyto, cytoplasmic side; exo, exoplasmic side. (B) Overall structure of ATP8A1-CDC50a complex. Cryo-EM maps (top) and ribbon models (bottom). The same color scheme is used throughout the manuscript. (C) Phospholipid-dependent ATPase activity of ATP8A1. Data points represent the mean ± SEM of three to six measurements at 37°C. By nonlinear regression of the Michaelis-Menten equation, ATP8A1-CDC50a in GDN micelles has a Km of 111.0 ± 26.4 μM for 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-l-serine (POPS) and a maximal ATPase activity of 99.7 ± 9.5 nmol min−1 μg−1. POPE, 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphatidylethanolamine.

Interaction between ATP8A1 and CDC50a

CDC50a is an essential component for P4-ATPases and is required for the proper expression and folding of ATP8A1 (fig. S1E) (10, 11). CDC50a and ATP8A1 interact extensively through the extracellular TM and intracellular regions (fig. S3). In the extracellular region, the CDC50a ectodomain covers all of the extracellular loops of ATP8A1, except for the M1-M2 loop, interacting in an complementary electrostatic manner: the extracellular loops of ATP8A1 bear negative charges, whereas CDC50a bears positive charges (fig. S3, A and B). In particular, Asp961 and Glu1026 of ATP8A1 form a salt bridge with Arg262 of CDC50a. In addition, the M3-M4 loop of ATP8A1 extends toward CDC50a, and the two bulky residues at the tip of the loop, Trp328 and Tyr329, form hydrophobic interactions involving Phe127, Tyr299, Pro300, Val301, and the N-glycan attached to Asn180 of CDC50a (fig. S3E). In the TM region, several bulky residues, such as Trp942, Ala947 (M9), Met1038, Phe1042, and Leu1049 (M10) of ATP8A1, and Phe54, Ile57, Phe61 (TM1), Phe324, Leu325, Ala328, Tyr329, and Val332 (TM2) of CDC50, are engaged in the complex interaction (fig. S3G). Furthermore, we observed a strong planar density at the interface between M7 and M10 of ATP8A1 and TM2 of CDC50a (fig. S3C), which could be assigned to the cholesteryl hemisuccinate added during solubilization. Therefore, cholesterol may bind to the same site and facilitate the heterodimeric interaction of ATP8A1 and CDC50a. In the cytoplasmic region, the N-terminal tail of CDC50a adopts an unstructured loop conformation that extends parallel to the plasma membrane and interacts with the M6-M7 and M8-M9 loops and the short segment connecting M4 and the P domain (fig. S3F). Overall, CDC50a envelops the bulk of the TM segments and forms extensive interactions with ATP8A1, which explains the chaperone activity of CDC50a for the P4-type ATPases.

Entire transport cycle of P4-ATPase

The cryo-EM structures revealed the clear densities of the inhibitors in their respective maps, bound at the catalytic site of ATP8A1 and stabilizing the ATPase domain in different conformations (Fig. 2), whereas CDC50a adopted almost the same conformation in all of these states [root mean square deviation (RMSD) (Å) = 0.24 to 0.66]. Most notably, the TM region of ATP8A1 remains structurally rigid throughout the transport cycle, probably because of the tight association with CDC50a, in contrast to other P-type ATPases, such as SERCA (figs. S10 and S11 and movie S1).

Fig. 2 Entire transport cycle of ATP8A1-CDC50a.

(A) The six different intermediates of ATP8A1-CDC50a during the phospholipid translocation cycle are shown, arranged clockwise as in the Post-Albers reaction cycle: E1, E1-ATP, E1P-ADP, E1P, E2P, and E2Pi-PL. The bound inhibitors are shown in space-filling model representations. (B) Comparison of the phosphorylation sites in each intermediate. AMPPCP and ADP are shown as sticks, and AlF4 and BeF3 are shown as spheres. Densities are shown as green mesh, contoured at 3.5σ. Single-letter abbreviations for the amino acid residues are as follows: A, Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met; N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr.

The structures obtained under three conditions, namely without inhibitors, with AMPPCP, and with AlF4-ADP, describe the conformational changes upon ATP binding and autophosphorylation, which correspond to the E1, E1-ATP, and E1P-ADP conformations, respectively, in the Post-Albers scheme (Fig. 2A). The densities of the N and A domains are only weakly visible in the E1 state, indicating the highly flexible motion of these domains without any ligand (E1 in Fig. 2 and fig. S4). The particles were then classified according to the densities of the N and P domains, and these domains were modeled into the class with the strongest densities, which probably represents the most likely arrangement in the E1 state (class 2 in fig. S4). The particles of the AMPPCP-bound state can be classified into three similar conformations, wherein the N and P domains adopt slightly different orientations (fig. S5). The comparison of these classes indicated that ATP binding at the N domain induces the mutual approach of the N and P domains. The density for the AMPPCP is most clearly visible within the class where these domains are proximal and bridged by AMPPCP (E1-ATP in Fig. 2B and fig. S5D): The adenine ring interacts with Phe534 of the N domain, whereas the phosphate group interacts with Asp409 and Thr411 (the DKTG motif), Asn789, and Asp790 at the phosphorylation site of the P domain, in cooperation with a Mg2+ ion. The AlF4-ADP–bound state is similar to the E1-AMPPCP conformation, but the N and P domains are more tightly bridged by ADP and AlF4 (E1P-ADP in Fig. 2 and fig. S6), captured in the phosphoryl transfer intermediate (E1P-ADP).

Overall, ATP binding and the subsequent phosphoryl transfer reaction induce the proximal arrangement of the N and P domains, which is accompanied by a slight outward shift of the A domain by ~6.5 Å (E1, E1-ATP, and E1P-ADP in Fig. 2A, fig. S10, and movie S1). The phosphorylation reaction is mediated by the motions of the ATPase domain and does not require any changes in the TM region. The TM segments of ATP8A1 adopt almost the same conformation throughout the transition (fig. S11), which is consistent with the substrate-independent autophosphorylation of P4-type ATPases (8).

The two phosphate analogs, BeF3 and AlF4, occupy the phosphorylation site in a similar manner, but their coordination geometries are slightly different (E1P, E2P, and E2Pi-PL in Fig. 2). BeF3 is covalently attached to the carboxylate side chain of Asp409, in coordination with a Mg2+ ion, and captures the phosphoenzyme ground state (E2P in Fig. 2B). The A domain is tightly fixed to the phosphorylation site (E2P in Fig. 2A and fig. S7) through the backbone carbonyls of Asp189 and Gly190 in the conserved DGET motif (residues 189 to 192) (E2P in Fig. 2B). The N domain is pushed apart from the P domain and no longer has access to the phosphorylation site, thus representing the ADP-insensitive E2P state (8). The particles of the AlF4-bound state could be separated into two different classes (fig. S8), and both showed clear AlF4 density at the phosphorylation site. In the first class, the bound AlF4 does not mediate any interdomain interactions, and the catalytic domains adopt a conformation similar to the AlF4-ADP bound state (E1P-ADP and E1P in Fig. 2B), likely representing the transient phosphorylated state (E1P) immediately after the ADP release. In the second class, AlF4 mediates the interaction between the N and A domains through the DGET motif in a similar manner to the BeF3-bound state, but the A domain is rotated by ~22° around the phosphorylation site, as compared with the BeF3-bound state (Fig. 3A). This allows the repositioning of the carboxyl side chain of Glu191 in the DGET motif to provide a catalytic base for the dephosphorylation reaction (E2P and E2Pi-PL in Fig. 2B) (8), thereby mimicking the dephosphorylation transition–like intermediate (29). The rearrangement of the A domain accompanies the swing-out motion of the TM1-TM2 segment, which is directly connected to the A domain, consequently creating a large cleft between the M1-M2 and M4-M5 segments, in which the clear density of a glycerophospholipid is observed (Fig. 3, B to D). Therefore, this second AlF4 class structure represents the substrate-bound E2Pi state (E2Pi-PL). The rearrangement of the A domain is likely to be coupled to the binding of the substrate lipid, as it occupies the cleft and pushes out the M1-M2 segment (figs. S10 and S11), which explains the substrate-dependent dephosphorylation of P4-ATPases (Fig. 1C) (8).

Fig. 3 Phospholipid recognition.

(A) Structural comparison of the E2P and E2Pi-PL states, showing the large rearrangement of M1-M2 and the N and A domains upon phospholipid binding. (B and C) Phospholipid binding site, viewed (B) parallel to the membrane plane and (C) as a close-up of the head group. Residues within 4 Å of the bound phospholipid are shown as sticks. Hydrogen bond interactions are shown as black dashed lines. (D) Cryo-EM density showing the bound endogenous phospholipid (green mesh, 2.5σ). (E) Residues constituting the hydrophobic gate are shown. The putative translocation pathway is indicated by an orange arrow.

Phospholipid recognition

Given that the substrate lipids (such as PS or PE) were not added during purification, it is likely that endogenous phospholipid contained in the GDN micelles is specifically bound to ATP8A1 in the AlF4-bound state (Fig. 3, A to D, and fig. S12A). ATP8A1 shows PG-, PE-, and PS-dependent ATPase activity, with the highest preference for PS (13, 17), and the size and shape of the head group density are in good agreement with those of the serine moiety (fig. S1H). Therefore, we modeled PS into the density. PS is recognized within the open cleft, in which the phosphate group is coordinated by the backbone amide groups of Ile357 and Ser358 in the conserved PISL motif at the unwound kink of M4 and further stabilized by the Gln88, Asn353, and Asn882 side chains (Fig. 3, C and E), whereas the attached acyl chains are exposed to the bulk lipid environment and partly accommodated in the hydrophobic pocket formed by the conserved residues in TM2 and TM4, such as Val103, Pro104, Phe107 (M2), Val361, Val365 (M4), Val883, and Leu891 (M6) (Fig. 3C). In the current cryo-EM map, the acyl chains are most visible near the attached glycerol moiety, and PS molecules with shorter acyl chains showed weaker ATPase activity (fig. S1G), indicating that the acyl chains, as well as the hydrophilic head group, are specifically recognized in the substrate binding pocket of ATP8A1.

The head group of PS is situated within a small cavity on the extracellular half of the cleft and is surrounded by hydrophilic residues, such as Gln88, Asn352, Asn353, and Asn882 (Fig. 3C), with which the serine moiety forms hydrogen bonding interactions. Consistently, mutational studies have shown the importance of the uncharged polar residues Gln88, Gln89, Asn352, and Asn353 for PS selectivity (2224). Such interactions explain the head group preferences of ATP8A1, which has weak selectivities for PE and PG, with head groups that can form similar hydrogen bonding interactions, and no selectivity for PC, with head methyl groups that cannot form such hydrogen bonding interactions. The PC selective P4-ATPases have nonpolar residues, such as Ala and Gly, at the corresponding positions (fig. S2), also supporting the proposal that the residues constituting this exoplasmic cavity primarily define the head group selectivity.

Lipid translocation pathway of ATP8A1

In the P2-ATPases, conformational changes in the cytoplasmic ATPase domains are coupled to rearrangement of the core TM helices that constitute the cation binding sites (Fig. 4A) (30, 31). Especially in SERCA, Glu309 of the conserved PEGL motif, located in the unwound M4 kink, constitutes part of the ion binding sites, enabling coupling between ion binding and release and rearrangement of the ATPase domain. In ATP8A1, the M4 segment is similarly kinked at the PISL motif, but the ion binding sites are lost by the substitution with hydrophobic residues Ile357, Leu854, and Val977 (Fig. 4B). Although the PS binding site partially overlaps the Ca2+ binding site in SERCA (site II), the arrangement of the surrounding residues remains almost unchanged throughout the transport cycle. It has been suggested that the P4-ATPases use a different translocation pathway for a large lipid substrate (9), and according to previous mutation studies on the yeast and bovine P4-ATPases, the residues associated with the head group selectivity are mapped along the hydrophilic cleft between the M1-M2 and M3-M4 segments (fig. S12B). In the PS-bound structure, the head group enters from the exoplasmic leaflet and stays occluded in the middle of the pathway by the side chain of Ile357 in the PISL motif. The mutation of Ile357 to bulky residues, such as Met and Phe, drastically reduced the lipid transport activity, whereas the mutations to smaller residues only moderately affected the transport activity (22), suggesting that Ile357 constitutes a central hydrophobic gate for the lipid translocation, together with other residues on the M1-M2 segment, such as Phe81 and Ile108 (Fig. 3E). Because the exoplasmic end is also closed by the M1-M2 loop (fig. S12B), the current structure probably represents a partially occluded state. Although PS binding induces a slight reorientation of the Ile357 side chain toward the M1-M2 segments (Fig. 4B), the translocation of the hydrophilic head group requires further rearrangement of the central gate residues, which is probably coupled with the phosphate release from the P domain. By analogy to the E2P-to-E2 transition in SERCA, the phosphate release “unlocks” the A domain and allows a further outward shift of the M1-M2 segment, thus inducing the opening of the central hydrophobic gate (32). Previous structures of P2-ATPases revealed several lipid binding sites; for example, the E2 structure of SERCA stabilized by thapsigargin and an inhibitor, 2,5-di-tert-butyl-1,4-dihydroxybenzene (BHQ) (33), showed PE binding between the M2 and M4 segments at the intracellular leaflet, corresponding to the putative exit of the lipid translocation pathway in ATP8A1 (fig. S13A). Furthermore, phospholipids are anchored to the positively charged residues at the protein-lipid interface and interplay with the protein conformational changes during the transport cycle in SERCA (34). The ATP8A1-CDC50A complex has clusters of positively charged residues at both the entrance and exit of the translocation pathway, which may play important roles in lipid translocation (fig. S12C).

Fig. 4 Comparison of the phospholipid binding sites.

(A) Ca2+ binding site of SERCA in the Ca2+ binding state (right: PDB ID 1T5S) and the H+ binding state (left: PDB ID 3B9R), viewed from the cytoplasmic side. Residues involved in Ca2+ and H+ transport are shown as ball-stick representations. Hydrogen bonds are shown as black dashed lines and the bound Ca2+ are pale blue spheres. (B) Phospholipid binding site of ATP8A1 in the unbound state (right: E1-ATP) and the phospholipid-bound state (left: E2Pi-PL) from the same viewpoint as in (A). Residues involved in phospholipid translocation and other residues corresponding to those coordinating H+ and Ca2+ in SERCA are shown as ball-and-stick representations. Hydrogen bonds are shown as black dashed lines. Cryo-EM density showing the side chain of Ile357 in unwound M4 kink (green mesh, 3.0σ).

C-terminal autoregulatory domain

In the BeF3-stabilized E2P state, we observed an extra density extending through the cytoplasmic catalytic domains (Fig. 5, A and B), which we assigned as the C-terminal autoregulatory domain (residues 1117 to 1140) (35, 36), consisting of the conserved GYAFS motif (residues 1119 to 1123) and a short helical domain (residues 1131 to 1137), although the ~50–amino acid linker connected to the M10 helix was disordered. The regulatory domain interacts with the N domain, and the GYAFS motif is specifically recognized by a short loop region of the N domain (residues 533 to 539) (Fig. 5B). Notably, Phe1122 occupies the ATP binding site and stacks with Phe534. The densities of the C-terminal residues are only visible in the BeF3-stabilized E2P conformation and are completely disordered in the other conformations, including the ligand-unbound E1 state. This suggests that the regulatory domain specifically stabilizes ATP8A1 in the E2P conformation, in which the N domain is somewhat farther apart (Fig. 5C and fig. S10).

Fig. 5 ATP8A1 autoregulation by the C-terminal domain.

(A) In the BeF3-stabilized E2P state, an extra density is observed around the cytoplasmic catalytic domains, corresponding to the C-terminal autoregulatory domain. The density is shown as a green mesh, contoured at 3σ. (B) Close-up view of the interaction between a short loop region (residues 533 to 539) in the N domain and the regulatory domain. An atomic model of the GYAFS motif and a short helical region in the regulatory domain are modeled into the density. (C and D) Arrangements of the N and A domains and the regulatory domain, shown for the E2P (C) and E1-ATP (D) states, viewed from the cytoplasmic side. The N and A domains of the E2Pi-PL state are superimposed in a transparency representation. The regulatory domain keeps the N domain apart from the A domain and thus facilitates the rotational movement of the A domain around the phosphorylation site in the E2P state (C), whereas the similar rearrangement is hindered by the N domain in the E1-ATP state (D).

The C-terminal regulatory domain has different effects between the yeast and mammalian P4-ATPases. In the yeast Drs2p flippase, it exerts an autoinhibitory effect on ATPase activity (37). However, in the mammalian ATP8A2 flippase, it mediates a rather complicated regulation mode. The partial truncation of the GYAFS motif and the short helical domain results in decreased ATPase activity, whereas the complete loss of the C-terminal residues, including the disordered loop region, restores the ATPase activity to the same level as the wild-type enzyme (38), indicating that the GYAFS motif and the short helical domain observed in the current cryo-EM map positively modulate the enzymatic reaction. We hypothesize that the regulatory domain keeps the N domain apart from the A domain in the E2P state and thus facilitates the rotational rearrangement of the A domain that is required for PS binding. The conformation of the N domain in the E1 and E1-ATP states would sterically prevent this rotational motion (Fig. 5D).

Mechanism of the P4-type ATPase

The current cryo-EM structures revealed six different intermediates of ATP8A1, namely, E1, E1-ATP, E1P-ADP, E1P, E2P, and E2Pi-PL, demonstrating the transport cycle of the lipid flippase reaction (Fig. 6). Although ATP8A1 shares a similar ATP hydrolysis–dependent catalytic reaction with the ion-transporting P2-type ATPases, such as SERCA (30, 31), Na+/K+-ATPase (39), and H+/K+-ATPase (40), there are notable differences in their transport mechanisms and substrate transporting pathways (fig. S14). These canonical ion-transporting P-type ATPases undergo extensive rearrangements of the TM region, especially in the M1 to M6 segments that constitute the ion translocating pathway (fig. S14A). In contrast, ATP8A1 maintains the structural rigidity of the TM region throughout the transport cycle, as the core TM segments (M3 to M10) could be superimposed well in all of the intermediates [RMSD (Å) = 0.30 to 0.69]. Consequently, ATP8A1 has a distinct pathway for the lipid head group, between the M1-M2 and M3-M4 segments, and the lipid translocation is essentially accomplished by the mobile segments of M1-M2 (fig. S14B).

Fig. 6 Proposed mechanism of phospholipid translocation.

Schematic model of the phospholipid translocation cycle by ATP8A1-CDC50a, according to the Post-Albers mechanism. The model is depicted with the same colors as in Fig. 1A. ATP binding induces the proximal arrangement of the N and P domains, by bridging these domains and slightly forcing out the A domain. After the phosphoryl transfer reaction, ADP is released from the N domain, and the A domain approaches the N domain and interacts with it, through the DGET motif, to form the E2P state. The C-terminal regulatory domain penetrates between the P and N domains and stabilizes the E2P state. The rearrangement of the A domain induces flexibility in the M1-M2 segments, thus allowing phospholipid binding at the interface between the M1-M2 and bulk TM segments. Phospholipid binding induces further rearrangement of the A domain, thereby facilitating the dephosphorylation reaction (E2Pi-PL). Ile357 constitutes a hydrophobic gate that occludes the middle of the translocation pathway. Phospholipid translocation to the cytoplasmic leaflet is probably coupled to the phosphate release at the P domain, allowing the further outward shift of the M1-M2 segment (E2-PL). The translocated phospholipid laterally diffuses to the cytoplasmic leaflet, and the enzyme adopts the E1 conformation, ready to initiate another reaction cycle.­­

The critical rearrangement in SERCA occurs during the E1P-to-E2P transition, in which the A-domain rearrangement toward the phosphorylation site induces the opening of the “luminal gate” composed of the M1 to M4 segments and alters the affinity for the substrate ions (fig. S15A) (31). Although the A domain of ATP8A1 undergoes a similar rearrangement during this transition, the conformational change is limited to the M1 and M2 segments in the region proximal to the A domain, and the luminal side remains unchanged (fig. S15B). This rigidity is probably achieved by the tight association with CDC50a, which holds the M3 to M10 segments of ATP8A1 on both the luminal and cytoplasmic sides. Most notably, the loop connecting M3-M4 and the cytoplasmic end of M4 is constrained by the interaction with CDC50a, which probably hinders the M3 and M4 rearrangement (figs. S3, E and F, and S15B). The deletion of the CDC50a N-terminal tail, which interacts with the cytoplasmic end of the M4 segment, decreased the flippase activities of P4-ATPases (25). Therefore, the rigidity of the TM segments is important for the transport activity of P4-ATPases. Although the phosphorylation-induced A-domain rearrangement in ATP8A1 causes only minor changes on the luminal side, the density of the M1-M2 segment near the A domain is more disordered in the E2P conformation (fig. S16), suggesting higher flexibility in the linker region. This flexibility may facilitate the subsequent binding of the phospholipid between the M1-M2 and M3-M4 segments by allowing the swing-out motion of the M1-M2 segment, as observed in the AlF4-stabilized dephosphorylation transition-like state. Overall, the P4-ATPases have evolved a distinct mechanism for the lipid translocation, while sharing the similar rearrangement of the cytoplasmic domains with the canonical ion-transporting P-type ATPases.

Supplementary Materials

science.sciencemag.org/content/365/6458/1149/suppl/DC1

Materials and Methods

Figs. S1 to S16

Table S1

References (4158)

Movie S1

References and Notes

Acknowledgments: We thank H. Hirano for assistance in generating the movie, H. Nishimasu for fruitful discussions, T. Nakane for assistance with the single-particle analysis, and the structural biophysics team at Mitsubishi Tanabe Pharma Corporation, especially H. Kishida, for technical advice about model building. We also thank the staff scientists at the University of Tokyo’s cryo-EM facility, especially K. Kobayashi, T. Kusakizako, H. Yanagisawa, A. Tsutsumi, M. Kikkawa, and R. Danev. Funding: This work was supported by a MEXT Grant-in-Aid for Specially Promoted Research (grant 16H06294) to O.N. Author contributions: M.H. prepared the cryo-EM samples and performed the functional analyses. M.H. and T.N. collected and processed the cryo-EM data and built the structures. K.Y. assisted data processing and structure refinement. M.H., T.N., and O.N. wrote the manuscript. T.N. and O.N. supervised the research. Competing interests: M.H. is a graduate student at Mitsubishi Tanabe Pharma Corporation and is supported by the company with nonresearch funds. The company has no financial or other interest in this research. Data and materials availability: Cryo-EM density maps have been deposited in the Electron Microscopy Data Bank under the accession codes EMD-9931 (E1 class1), EMD-9932 (E1 class2), EMD-9933 (E1 class3), EMD-9935 (E1-ATP class1), EMD-9934 (E1-ATP class2), EMD-9936 (E1-ATP class3), EMD-9937 (E1P-ADP), EMD-9938 (E2P class1), EMD-9939 (E2P class2), EMD-9940 (E2P class3), EMD-9941 (E2Pi-PL), and EMD-9942 (E1P). Atomic coordinates have been deposited in the Protein Data Bank under IDs 6K7G (E1 class1), 6K7H (E1 class2), 6K7J (E1-ATP class1), 6K7I (E1-ATP class2), 6K7K (E1P-ADP), 6K7L (E2P-class2), 6K7M (E2Pi-PL), and 6K7N (E1P). The raw images have been deposited in the Electron Microscopy Public Image Archive, under accession code EMPIAR-10303.
View Abstract

Stay Connected to Science

Navigate This Article