Cryo-EM structures of human SPCA1a reveal the mechanism of Ca²⁺/Mn²⁺ transport into the Golgi apparatus

Abstract

Secretory pathway Ca²⁺/Mn²⁺ ATPase 1 (SPCA1) actively transports cytosolic Ca²⁺ and Mn²⁺ into the Golgi lumen, playing a crucial role in cellular calcium and manganese homeostasis. Detrimental mutations of the ATP2C1 gene encoding SPCA1 cause Hailey-Hailey disease. Here, using nanobody/megabody technologies, we determined cryo–electron microscopy structures of human SPCA1a in the ATP and Ca²⁺/Mn²⁺-bound (E1-ATP) state and the metal-free phosphorylated (E2P) state at 3.1- to 3.3-Å resolutions. The structures revealed that Ca²⁺ and Mn²⁺ share the same metal ion–binding pocket with similar but notably different coordination geometries in the transmembrane domain, corresponding to the second Ca²⁺-binding site in sarco/endoplasmic reticulum Ca²⁺-ATPase (SERCA). In the E1-ATP to E2P transition, SPCA1a undergoes similar domain rearrangements to those of SERCA. Meanwhile, SPCA1a shows larger conformational and positional flexibility of the second and sixth transmembrane helices, possibly explaining its wider metal ion specificity. These structural findings illuminate the unique mechanisms of SPCA1a-mediated Ca²⁺/Mn²⁺ transport.

Cryo-EM structures of SPCA1a present the mechanisms of its wide metal ion recognition and Ca2+/Mn2+ transport.

INTRODUCTION

Newly synthesized secretory proteins are translocated and folded in the endoplasmic reticulum (ER) with the assistance of various types of ER-resident molecular chaperones and modification enzymes. Correctly folded proteins are transported to the Golgi to undergo further posttranslational modification and quality control and then sorted at the trans-Golgi network (TGN) before transportation to their destination (1). In the early secretory pathway comprising the ER and Golgi, maintaining homeostasis of luminal and cytosolic Ca²⁺ lays the foundation for proper protein synthesis (2). Sarco(endo)plasmic Ca²⁺–adenosine triphosphatase (ATPase) (SERCA) and secretory pathway Ca²⁺/Mn²⁺-ATPase (SPCA) play essential roles in sustaining higher concentrations of luminal Ca²⁺ in the ER and Golgi apparatus (2, 3). High-resolution structures of SERCA have been determined for a number of intermediate states, leading to a deep understanding of the detailed mechanisms of its Ca²⁺ transport cycle (4, 5). However, the mechanism of Ca²⁺ homeostasis in the Golgi apparatus (3), which also likely acts as an intracellular calcium storage compartment, is less well understood than that in the ER. In this context, whereas several regulins with a single-spanning transmembrane (TM) helix, such as phospholamban and sarcolipin, are known to regulate the Ca²⁺-transporting activity of SERCA (6), these regulatory TM peptides have not been identified for SPCA1a.

In contrast to SERCA localized in the ER and cis-Golgi, SPCA is present in the trans-Golgi and TGN for Ca²⁺ uptake, generating a large Ca²⁺ gradient concentration between the ER (~500 μM) and Golgi (~100 μM) (3, 7). Loss of SPCA1 is known to cause Golgi stress and apoptosis (8). In addition, loss-of-function mutations in the ATP2C1 (i.e., SPCA1) gene can cause Hailey-Hailey disease (HHD), a skin disorder characterized by persistent blisters and erosion that is inherited in an autosomal dominant manner (9). Homozygous knockout of the ATP2C1 gene in mice proved to be embryonic lethal due to failed neural tube closure (10). However, the mechanism by which decreased SPCA1 protein abundance affects keratocytes and neuroepithelial cells remains elusive.

Previous biochemical studies demonstrated that SPCA1 is unique in that it transports both Mn²⁺ and Ca²⁺, with a higher affinity for the former than the latter (11). Mn²⁺ is a crucial cofactor for many enzymes, including glutamine synthetase (12) and pyruvate carboxylase (13). O-linked glycosylation, an essential event for most secretory proteins, also requires Mn²⁺ in the Golgi (14). However, excess intracellular Mn²⁺ is toxic (15). Hitherto, SPCA1a is the most clearly identified Golgi-resident Mn²⁺ pump (11, 16–19). Previous work demonstrated that excessive Mn²⁺ uptake prevents neuronal and glial survival by inhibiting the Ca²⁺ transport activity of SPCA1 and inducing Golgi fragmentation via Golgi stress (20). Mn²⁺ accumulation to toxic levels in the mouse brain likely correlates with high expression of SPCA1 (21), suggesting that SPCA1 serves as a vital Mn²⁺ transporter in the Golgi membrane that engages in Mn²⁺ detoxification by actively pumping cytosolic Mn²⁺ into the secretory pathway (17, 22).

Although the functions of SPCA1 and the effects of its deficiency have been studied (10, 11, 23–25), the overall mechanism of operation remains elusive due to a lack of structural information. It is widely accepted that P-type ATPases such as SERCA (4, 26, 27) follow a post-Albers cycle (28, 29) during which two distinct states (E1 and E2) are generated. Numerous structural, biochemical, biophysical, and computational studies established that SERCA functions according to the scheme E1 → E1-ATP → E1P-ADP → E2P → E2-Pi → E2 (30), although the exact steps of entry/binding of adenosine triphosphate (ATP) and release of adenosine diphosphate (ADP) and Ca²⁺ remain controversial (26, 31–34). The similar functions and high amino acid sequence identity between SERCA and SPCA (fig. S1) suggest similar mechanisms of operation in metal ion transport. To gain detailed insight into the ion transport mechanism of SPCA, we here determined cryo–electron microscopy (cryo-EM) structures of human SPCA1a in two intermediate states at 3.1- to 3.3-Å resolution: the E1-ATP state using β,γ-methyleneadenosine 5′-triphosphate (AMPPCP), a nonhydrolyzable analog of ATP, in the presence of Ca²⁺ or Mn²⁺, and the E2P state using BeF₃⁻, a phosphate group mimic, in the absence of these metal ions.

As predicted, the present cryo-EM analyses revealed that the overall arrangement of the cytosolic and TM domains in SPCA1a resembles that in SERCA and that the location and amino acid residue geometry of the Ca²⁺/Mn²⁺-binding site in SPCA1a are nearly superimposable with those of the second Ca²⁺-binding site (site II) in SERCA2b. Despite these structural similarities, SPCA1a engages in fewer interactions between cytosolic domains, especially in the E1-Ca²⁺/Mn²⁺-ATP state, and undergoes similar rearrangements of both cytosolic and TM domains during the transition from E1-ATP to E2P states than occurs in SERCA. Moreover, the larger conformational and positional flexibility of TM2 and TM6 in SPCA1a compared with SERCA may help to explain the wider metal selectivity of SPCA1a. The present cryo-EM analyses provide important insights into the mechanisms of SPCA1a-mediated Ca²⁺/Mn²⁺ transport to the Golgi apparatus.

RESULTS

Cryo-EM structure analysis of human SPCA1a

Purified SPCA1a retained Ca²⁺- and Mn²⁺-dependent ATPase activity, exhibiting the release of the phosphate group as monitored by EnzCheck phosphate assay (Fig. 1A). SPCA1a showed much-enhanced ATPase activity with Co²⁺ and Ni²⁺, as well as with Ca²⁺ and Mn²⁺, compared to that without those divalent metal ions, suggesting the capacity of the Ca²⁺/Mn²⁺ pump to recognize other divalent metal ions as substrates. Although the physiological concentrations of intracellular labile Mn²⁺, Co²⁺, and Ni²⁺ remain elusive, these biochemical results suggest the wide metal specificity of SPCA1a.

Fig. 1. Cryo-EM maps and modeled structures of human SPCA1a in the E1-Ca²⁺/Mn²⁺-ATP and E2P states.

(A) ATPase activities of purified human SPCA1a in the presence of 10 μM Ca²⁺ (Ca), 10 μM Ca²⁺ + 446 nM megabody 14 (Ca + Mb14), 10 μM Mn²⁺ (Mn), 10 μM Co²⁺ (Co), or 10 μM Ni²⁺ (Ni) and in the absence of them (CTRL). All measurements were made in triplicate. (B) Domain organization of SPCA1a. SPCA1a contains three cytosolic domains, A-domain (orange), P-domain (blue), and N-domain (hot pink), and TM1–2, TM3–4, and TM5–10 helix clusters (green, violet, and wheat, respectively). (C to E) Cryo-EM density maps of SPCA1a in the E1-Ca²⁺-AMPPCP (C), E1-Mn²⁺-AMPPCP (D), and E2-BeF₃⁻ (E) states. (F to H) Ribbon diagrams of cryo-EM structures of SPCA1a in the E1-Ca²⁺-AMPPCP (F), E1-Mn²⁺-AMPPCP (G), and E2-BeF₃⁻ (H) states. The cytoplasmic domains (A-, P-, and N-domains), TM subdomains (TM1–2, TM3–4, and TM5–10), and the head domain of the megabody are colored as in (C) to (E). Ligands (AMPPCP, BeF₃⁻, Ca²⁺, and Mn²⁺) bound to SPCA1a are highlighted in circles.

According to the PhosphoPlus database, SPCA1a is ubiquitylated at Lys⁸ (A-domain) and Lys⁴⁹⁶ (N-domain). Consistently, we verified the marked ubiquitylation of endogenous SPCA1a by Western blotting using an anti-ubiquitin antibody (fig. S2A). By contrast, overexpressed dodecapeptide from human podoplanin (PA)-tagged SPCA1a was ubiquitylated only partially, indicating a much lower ubiquitylation efficiency compared to the endogenous protein. As expected, ubiquitylated species were completely removed after treatment with Usp2cc enzyme (35) (fig. S2, C to E). The resulting homogeneously deubiquitinated SPCA1a sample was used for cryo-EM single-particle analysis (fig. S2, F and G).

We performed cryo-EM analyses of SPCA1a in the E1-Ca²⁺/Mn²⁺-AMPPCP and E2-BeF₃⁻ states as mimics of the E1-Ca²⁺/Mn²⁺-ATP and E2P states, respectively. During data analysis, we found that whereas the A-domain of SPCA1a was visible in some two-dimensional (2D)–averaged images of the E2P state (figs. S3 and S4A), it was barely visible in any 2D-averaged images of the E1-AMPPCP state (fig. S4B). Presumably due to the lack of density of the A-domain in the 2D-averaged maps, 3D reconstruction of the E1-Ca²⁺/Mn²⁺-AMPPCP state was unsuccessful. We thus predicted that an additional protein moiety might compensate for the apparent loss of the A-domain, facilitating 3D reconstruction. We first tested human cofilin-1 (hCFL-1) as a binder protein since it is known to bind the P-domain of SPCA1a (25). However, no interaction was observed between purified hCFL-1 and SPCA1a in our in vitro assay using size exclusion chromatography (SEC; fig. S5, A and B).

We then decided to screen a nanobody that binds specifically and tightly to SPCA1a using a yeast surface-displayed nanobody library (36). The ability of nanobodies to bind SPCA1a was first evaluated by flow cytometry (fig. S6, A and B). Nanobodies with a high fluorescein isothiocyanate (FITC)–positive percentage (≥10%) were chosen for sequencing and were found to share the same sequence. Clone No.14, named Nb14, was eventually selected and used in subsequent steps. The open reading frame of Nb14 was subcloned into the pET151b vector with an N-terminal pelB signal peptide to produce the nanobody in the periplasm of Escherichia coli (fig. S7, A to C). The binding performance of the recombinant nanobody was further assessed by SEC in the presence of Ca²⁺ or Mn²⁺ (fig. S8, A and B). The complex of Nb14 and purified human SPCA1a was subjected to cryo-EM single-particle analysis.

While the E1-AMPPCP state of SPCA1a in complex with Nb14 generated a cryo-EM map at an overall resolution of 3.4 Å, particles showing clear density for the A-domain only accounted for 8.45% of total particles (fig. S9, A and D). Therefore, we performed extra rounds of classification to exclude particles without the visible A-domain and reattempted 3D reconstruction (fig. S10A). Consequently, Nb14 was found to bind the N-domain of SPCA1a (fig. S10D). We surmised that the increased A-domain visibility in the E1-AMPPCP state is likely due to the bound Nb14 with a molecular mass of ~17 kDa; hence, we predicted that increasing the nanobody size would further facilitate 2D and 3D classification, eventually generating an even higher resolution map of SPCA1a. To this end, we converted Nb14 to a megabody (37) by fusing a scaffold protein to Nb14 while retaining the binding affinity for SPCA1a. c7HopQ was used as a scaffold protein to generate megabody 14 (Mb14) with a molecular weight of ~56 kDa (fig. S7, D to F). The SEC analysis confirmed its capability to bind SPCA1a (fig. S8, C to F). The biochemical analysis demonstrated that Mb14 binding abolished the Ca²⁺-dependent ATPase activity of SPCAC1a (Fig. 1A), suggesting that bound Mb14 may have blocked the conformation and/or dynamics of the A-domain during the catalytic cycle. Consistently, the usage of Mb14 greatly increased the proportion of SPCA1a particles with a clearly visible A-domain (from 8.45 to 48.95% or 53.87%; fig. S9, B to D) and improved the map resolution (from 3.4 to 3.1 Å; figs. S11 and S12). Therefore, cryo-EM maps of the SPCA1a-Mb14 complex were used for subsequent structure modeling and refinement. By conducting further 3D variability analysis using cryoSPARC (38), we identified domain rotations for each of the Ca²⁺- and Mn²⁺-bound E1-AMPPCP states, leading to the reconstruction of at least three subclass structures (figs. S11 and S12). In this work, subclass 2 of the Ca²⁺-bound form and subclass 1 of the Mn²⁺-bound form were chosen for structure description and comparison since they showed the clearest density maps in almost the entire region among all subclasses (figs. S11D and S12D).

Overall structure of SPCA1a

The present cryo-EM analysis revealed that, like SERCA, SPCA1a shares the typical architecture of P-type ATPases, with 10 TM helices (TM1 to TM10) and three cytosolic domains, including the actuator (A), nucleotide-binding (N), and phosphorylation (P) domains (Fig. 1, B to H). Mb14 was found to bind the N-domain in the E1-Ca²⁺/Mn²⁺-AMPPCP state (Fig. 1, C and D, light green). The TM domain of SPCA1a consists of three TM helix clusters (TM1–2, TM3–4, and TM5–10), similar to SERCA (Fig. 1F and fig. S13A).

Despite the similar arrangements of the cytosolic domains and TM helices, superimposition of the cryo-EM structures of SPCA1a and SERCA2b revealed different structural features, especially in the loop regions on both the cytosolic and luminal sides, and interactions between cytosolic domains. As predicted from amino acid sequence alignment (fig. S1), the nine cytosolic loops and four luminal loops of SPCA1a are shorter than those of SERCA2b due to the deletion of SERCA2b-specific residues (Fig. 2, A and D). In this connection, whereas most of the loops are exposed to the solvent without strong interactions with other domains in SPCA1a, loop 9 in the N-domain of SERCA2b is much longer than its counterpart in SPCA1a, and Leu⁵⁷⁷ in this loop appears to make hydrophobic contacts with Leu⁴⁸⁵ within the same domain of the E1-Ca²⁺-AMPPCP state of SERCA2b (27). In addition, this longer loop forms hydrogen bonds (H-bonds) with parts of the A-domain in the Ca²⁺-unbound E2 state of SERCA (fig. S14). We also found that loop 3 in the A-domain of SERCA2b contains an insertion of Thr¹⁷¹, and this residue forms H-bonds with Glu⁴⁸⁶ of the N-domain in the E1-Ca²⁺-AMPPCP state (Fig. 2A, inset). Thus, while the longer loops in the A- and N-domains of SERCA2b appear to contribute to stabilizing interdomain interactions within the cytosolic head piece, the interdomain interactions seem much weaker in SPCA1a due to the lack of the loops.

Fig. 2. Superimposition of cryo-EM structures of human SPCA1a and SERCA2b.

(A) Cryo-EM structures of SPCA1a (blue) and SERCA2b (salmon, PDB ID: 6LLE) in the E1-Ca²⁺-AMPPCP state were aligned such that the RMSD values of their P-domains are minimized. The longer loops of SERCA2b compared with those in SPCA1a are colored fluorescent green and numbered sequentially according to amino acid sequence. The circled inset shows interactions between Glu⁴⁸⁶ (N-domain) and Thr¹⁷¹ (A-domain). (B) H-bond network formed between cytosolic domains in the E1-2Ca²⁺-AMPPCP state of SERCA2b. (C) H-bond network formed between cytosolic domains in the E1-Ca²⁺-AMPPCP state of SPCA1a. (D) Cryo-EM structures of SPCA1a and SERCA2b (PDB ID: 6LLY) in the E2P state were aligned such that the RMSD values of their P-domains are minimized. (E) H-bond network formed between cytosolic domains in the E2-BeF₃⁻ state of SERCA2b. (F) H-bond network formed between cytosolic domains in the E2-BeF₃⁻ state of SPCA1a.

In line with this, while eight interdomain H-bonds are formed at the interfaces of the A-, N-, and P-domains in the E1-Ca²⁺-AMPPCP state of SERCA2b, only three are present in the corresponding state of SPCA1a (Fig. 2, B and C). The reduced interdomain interactions in SPCA1a appear consistent with the unstable nature of its A-domain in the E1-AMPPCP state. By contrast, the A-domain of SPCA1a was clearly identified in the E2-BeF₃⁻ state even without the nanobody/megabody (fig. S4A). Consistently, the number of interdomain H-bonds in SPCA1a is comparable to that in SERCA2b (Fig. 2, E and F). Thus, while the cytosolic domain arrangement is similar between SPCA1a and SERCA, the mode of interactions at the cytosolic domain interfaces is different between these two Ca²⁺ pumps, possibly explaining the unstable nature of the A-domain in the E1-AMPPCP state of SPCA1a.

Structural basis of ATP binding and phosphorylation in SPCA1a

The cryo-EM structures revealed that AMPPCP is located at the N-domain and P-domain boundary in SPCA1a (Figs. 1, F and G, and 3, A and B), like in SERCA1a (26) and SERCA2b (27). In the E1-Ca²⁺-AMPPCP state, the ATP-binding pocket consists of Thr³⁵², Glu⁴²⁷, Ser⁴⁵⁶, Lys⁴⁸⁰, Gly⁵⁷¹, Asp⁵⁷², and Arg⁶¹⁹ (Fig. 3A). An AMPPCP molecule in the Mn²⁺-bound state is located at almost the same position as in the Ca²⁺-bound state (Fig. 3B). To compare the location of AMPPCP in SPCA1a and SERCA, we focused on the geometry of residues surrounding AMPPCP bound to SPCA1a and SERCA2b (fig. S15). By measuring the distances between the center of the AMPPCP molecule and the Cα atoms of the surrounding residues (fig. S15, A and C), we found that the mode of ATP binding in SPCA1a is highly similar to that in SERCA2b (fig. S15, B and D).

Fig. 3. ATP- and metal ion–binding sites of human SPCA1a.

(A and B) ATP-binding site of SPCA1a in the E1-Ca²⁺-AMPPCP (A) and E1-Mn²⁺-AMPPCP (B) states. ATP-binding amino acid residues are shown in stick representation. Residues belonging to the A-, N-, and P-domains are colored orange, hot pink, and light blue, respectively. Electron density for the bound AMPPCP molecule is represented as mesh. (C) Phosphorylation site of SPCA1a in the E2-BeF₃⁻ state. Amino acid residues involved in interactions with the phosphoryl group are represented by sticks. Residues belonging to the A-, N-, and P-domains are colored orange, hot pink, and light blue, respectively. Cryo-EM density for AMPPCP and BeF₃⁻ is represented as mesh. (D and E) Metal ion–binding sites of SPCA1a in the E1-Ca²⁺-AMPPCP (D) and E1-Mn²⁺-AMPPCP (E) states. Amino acid residues located at the Ca²⁺- or Mn²⁺-binding pocket are shown in stick representation. (F) Comparison of the distances between the O^δ atom of Asp⁷⁴² and Ca²⁺ (or Mn²⁺) in the E1-Ca²⁺-AMPPCP (light green) and E1-Mn²⁺-AMPPCP (pink) states. Structures are aligned for residues Val³⁰³-Ile³⁰⁶. Note that the Cα and O^δ atoms of Asp⁷⁴² get closer to Mn²⁺ by 0.25 and 0.63 Å, respectively, compared to those in the Ca²⁺-bound state. (G to I) Top (i.e., cytosolic) view of the TM domain of SPCA1a in the E1-Ca²⁺-AMPPCP state (G), the E2P state (H), and SERCA2b in the E1-Ca²⁺-AMPPCP state (I). Amino acid residues involved in Ca²⁺ binding are represented by sticks. Ca²⁺ ions are shown as green spheres. Structures were aligned for the TM7–TM10 cluster in a conventional manner.

In the E2-BeF₃⁻ state (Fig. 1, E and H), no metal ion is bound to the TM domain. A BeF₃⁻ molecule is covalently bridged to Asp³⁵⁰ with the assistance of a nearby Mg²⁺ ion coordinated by Asp³⁵⁰ and Asp⁶⁴⁴ (Fig. 3C). Notably, the Thr¹⁸⁹-Gly¹⁹⁰-Glu¹⁹¹ (TGE) loop is located adjacent to Asp³⁵⁰ (Fig. 3C), as seen in other P-type ATPases in the E2P state (26, 39, 40) (also see below for a description of the movement of the TGE loop).

Structural basis of Ca²⁺/Mn²⁺ binding in SPCA1a

Well-resolved density was observed for the metal-binding pocket in the TM domain of SPCA1a in the E1-Ca²⁺/Mn²⁺-AMPPCP state, demonstrating that the Ca²⁺/Mn²⁺-binding site is primarily constituted by TM4 and TM6 (Fig. 3, F and G). In the Ca²⁺-bound state, the main-chain carbonyl groups of Val³⁰³, Ala³⁰⁴, and Ile³⁰⁶ in TM4 and the O^δ atoms of Asn⁷³⁸ and Asp⁷⁴² in TM6 coordinate to the Ca²⁺ ion with distances of 2.36 to 2.44 Å, adopting a square pyramid geometry. In addition, the O^ɛ atom of Glu³⁰⁸ in TM4 appears to weakly bind to the metal ion as a sixth ligand with a distance of 3.12 Å, forming an octahedral-like Ca²⁺ complex (Fig. 3D), as observed in one of the Ca²⁺-binding sites (site II) of SERCA1a (26) and SERCA2b (27) (Fig. 3I).

Notably, Mn²⁺ occupies the same binding site in SPCA1a as Ca²⁺ (Fig. 3, D and E), forming a similar coordination structure to the Ca²⁺-bound state. The Mn²⁺-bound state also has a square pyramid geometry, in which the main-chain carbonyl groups of Val³⁰³, Ala³⁰⁴, and Ile³⁰⁶ in TM4 and the side-chain O^δ atoms of Asn⁷³⁸ and Asp⁷⁴² coordinate to Mn²⁺ with distances of 2.15 to 2.55 Å. However, the side chain of Glu³⁰⁸ is located more apart from Mn²⁺ compared to the Ca²⁺-bound state, not allowing an octahedral-like complex. Since the ion radius of Mn²⁺ (~80 pm) is smaller than that of Ca²⁺ (~100 pm), capturing Mn²⁺ requires TM4 and TM6 to move closer to each other. In support of this, the main-chain Cα and side-chain O^δ atoms of Asp⁷⁴² in TM6 get closer to Mn²⁺ by 0.25 and 0.63 Å, respectively, than to Ca²⁺ (Fig. 3F), forming a more compact metal-binding complex. This closer geometry between Mn²⁺ and its surrounding residues likely explains the higher affinity for Mn²⁺ than for Ca²⁺, as revealed by a previous biochemical study using a liposomal SPCA1a (11). Unlike SERCA, SPCA1 lacks another metal-binding pocket (site I), likely due to amino acid substitutions in TM6 (e.g., T798M, E907D, E770A, and N767S). The lack of site I may provide TM6 with positional flexibility due to the increased vacant space and thereby allow it to more closely approach TM4 to bind Mn²⁺ at site II (Fig. 3, E to H).

In the metal unbound state (i.e. E2-BeF₃⁻ state), the metal-binding cavity gets wider as a result of the rotation and rearrangement of TM4 and TM6 (movie S1), which drives Met⁷⁴¹ and Asp⁷⁴² in TM6 away from the ion-binding site (Fig. 3, G and H, and fig. S13, B and C). This structural insight seems to nicely explain how Ca²⁺/Mn²⁺ are released from SPCA1a.

Conformational transition from E1-Ca²⁺ (Mn²⁺)-ATP to E2P states in SPCA1a

To gain further insight into the mechanisms of Ca²⁺/Mn²⁺ transport conducted by SPCA1a, we compared cryo-EM structures of its E1-AMPPCP and E2P states (Fig. 4, A to G). Given that, except for the metal-binding site, overall structures of SPCA1a in E1-Ca²⁺-AMPPCP and E1-Mn²⁺-AMPPCP states are almost identical to each other, with a root mean square deviation (RMSD) value of 0.983 Å, we used the structure of the E1-Ca²⁺-AMPPCP state exclusively for comparison with the E2P state. We found that during the transition from E1-AMPPCP to E2P states, the N-domain rotates away from the P-domain by 58° around the indicated axis (Fig. 4, B and D). In addition to N-domain rotation, the A-domain rotates by 85° around the indicated axis (Fig. 4, B and E), which drives the ¹⁸⁹TGE loop closer to Asp³⁵⁰ to prevent the rebinding of ADP (Fig. 4B, inset), as seen in other P-type ATPases. Although SPCA1a undergoes a similar cytosolic domain rearrangement to SERCA, the luminal Ca²⁺ release gate constituted by TM3–4 and TM5–6 is much narrower in the E2P state of SPCA1a than in the corresponding state of SERCA2b (Fig. 4, G and H).

Fig. 4. Domain rearrangements of human SPCA1a during the transition from E1-ATP to E2P states.

(A and B) Relocation of cytosolic domains and TM helices during the transition from E1-Ca²⁺-AMPPCP (A) to E2P (B) states. Rotation angles and axes calculated in ChimeraX are represented as black arrows and lines, respectively. A-, N-, and P-domains and TM1–2, TM3–4, TM5–6, and TM7–10 clusters are colored orange, hot pink, light blue, green, yellow, violet, and dark yellow, respectively. (C to F) Close-up views of cytosolic domain rearrangement during the transition from E1-Ca²⁺-AMPPCP to E2P states, in which P-, N-, and A-domains and the TM7–10 cluster are highlighted in (C), (D), (E), and (F), respectively. Structures were aligned for the TM7–10 cluster in a conventional manner. Structures of SPCA1a in E1-Ca²⁺-AMPPCP and E2-BeF₃⁻ states are represented as transparent and colored models, respectively. (G and H) Cavities formed by TM1–2, TM3–4, and TM5–6 of SPCA1a (G) and SERCA2b (H) in the E2-BeF₃⁻ state. Cavities shown as blue surfaces were calculated with pyKVFinder (https://github.com/LBC-LNBio/pyKVFinder), and their sizes are represented by the number of hydrogen atoms that can be accommodated in the cavity.

To quantitatively compare the domain movements in SPCA1a and SERCA during the transition from the E1-AMPPCP to E2P states, we quantified the rotation angles of the main-chain Cα atoms in the cytosolic and TM domains and performed a statistical analysis of SERCA2b, SERCA1a, and SPCA1a (fig. S16). Since the rotational direction of each domain varies, the rotation axes were determined separately for A-, N-, and P-domains and for TM1–2, TM3–4, TM5–6, and TM7–10. Figure S17 shows a schematic representation of our rotation angle calculation. It is apparent that during the transition from the E1-AMPPCP to E2P states, the A- and N-domains undergo similar rotation to those in SERCAs, while the P-domain and TM5–6 of SPCA1a rotate by larger angles than in SERCAs. In contrast, the TM1–2 of SPCA1a rotate by a smaller angle than in SERCAs (fig. S16). These observations indicate that, overall, SPCA1a undergoes a similar domain rearrangement to SERCAs during the E1-ATP to E2P transition, while the TM1–2, TM5–6, and P-domain display different relocation behaviors between these two pumps.

Another striking difference between SPCA1a and SERCA2b was seen in TM2. TM2 of the E2P state of SPCA1a could only be modeled for residues Phe¹⁰⁰ to Glu¹²⁵, while the corresponding part of SERCA2b (Phe⁸⁸ to Lys¹²⁰) could be modeled entirely and was found to be longer than that in SPCA1a (fig. S18). In the E2P state, Tyr¹²²-Glu¹²⁵ of SPCA1a is twisted concomitant with the A-domain rotation (fig. S18A), while the corresponding helix of SERCA2b remains straight (fig. S18B). It is, however, notable that the TM2 twisting is seen in the E2·Pi state of SERCA2b, i.e., immediately after the dephosphorylation (fig. S19C) (41). Given that TM2 is directly linked to the A-domain, the observed twist around the cytosolic end of TM2 in SPCA1a suggests that TM2 is one of the key elements that modulate the metal ion transport by this Ca²⁺/Mn²⁺ ATPase (see also Discussion).

HHD mutations

Our present study also provides structural information for understanding mutations causing HHD. We mapped known missense mutations associated with HHD (23) on the cryo-EM structure of human SPCA1a in the E1-Ca²⁺-AMPPCP state (Fig. 5). Notably, the mutations are located throughout the entire SPCA1a protein chain. Most mutations occur around the metal-binding pocket (16 mutations) and in the P-domain (15 mutations), suggesting that these mutations affect the metal transport activity, efficiency of ATP hydrolysis, and/or phosphorylation of Asp³⁵⁰ of SPCA1a. In support of this, a previous study (42) reported that D742Y mutation (on TM6) at the metal-binding pocket eliminated the ability to bind both Ca²⁺ and Mn²⁺, and G309C mutation (on TM4) adjacent to the metal-binding pocket abolished Mn²⁺ binding, but the affinity for Ca²⁺ was retained, suggesting that Gly³⁰⁹ plays a role in Mn²⁺ selectivity. As an additional note, it is known that the symptoms of the HHD can develop at certain ages, mostly 19 to 65 years old (43), suggesting that patients carrying the mutations do not show obvious symptoms until the ages. In other words, the mutations are not always the sole reasons for the HHD, and other acquired environmental, accidental, or stochastic factors may be involved in this pathology. Nevertheless, the present cryo-EM structures provide molecular insight into the functional defects that can be caused by HHD mutations.

Fig. 5. HHD-associated mutations in SPCA1.

HHD-associated missense mutations (red spheres) are mapped onto the cryo-EM structure of SPCA1a in the E1-Ca²⁺-AMPPCP state.

DISCUSSION

In the present study, cryo-EM structures were determined at near-atomic resolutions for human SPCA1a, a Golgi-resident Ca²⁺/Mn²⁺ pump, in the E1-Ca²⁺/Mn²⁺-ATP and E2P states. The structures provided important insights into how Ca²⁺ and Mn²⁺ ions are specifically bound to the TM domain and transported to the Golgi lumen by SPCA1a in conjunction with marked rearrangements of the cytosolic and TM domains.

Unlike other Ca²⁺-transporting ATPases, such as SERCA and plasma membrane Ca²⁺ ATPase (PMCA), SPCA1a is ubiquitylated at Lys⁸ in the A-domain and Lys⁴⁹⁶ in the N-domain. However, the physiological significance of SPCA1a ubiquitylation is unknown. It should be noted that when SPCA1a was overexpressed, the proportion of ubiquitylated SPCA1a was much smaller than that of the nonubiquitylated form (fig. S2A). Monoubiquitylation plays a unique role in 26S proteasome degradation (44) and locking the conformation of the FANCD2-FANCI heterodimer (45). Therefore, monoubiquitylation is considered a versatile mechanism in protein quality control and conformational regulation. Together with the largely different ubiquitylation patterns observed for overexpressed and endogenous SPCA1a proteins (fig. S2A), and the mislocalization of overexpressed SPCA1a (fig. S2B), these results imply that native ubiquitylation of SPCA1a may play a role in specifying the cellular localization, binding partners, and hence physiological function of SPCA1. In this context, we failed to detect an interaction between SPCA1a and hCFL-1 in our in vitro experiments using purified proteins (fig. S5), although previous studies reported their functional interaction (25).

The Mn²⁺ transport activity of SPCA1a was first confirmed by observing the inhibitory effect of MnCl₂ on Ca²⁺ transport activity (16, 42). The present cryo-EM structure of SPCA1a in the E1-Mn²⁺-AMPPCP state revealed the mechanism of Mn²⁺ binding. Previous studies on plasma membrane ATPase-related 1 (PMR1), a yeast homolog of SPCA1a, suggested the presence of a possible selective gate controlling Mn²⁺ entry, formed by Gln⁷⁸³ and Val³³⁵ (46, 47). Although these two residues are conserved in SPCA1a (Gln⁷⁴⁷ and Val³¹⁴ in SPCA1a, respectively; fig. S1), Val³¹⁴ has been found to be located far from Gln⁷⁴⁷ (fig. S20A). Instead, Gln⁷⁴⁷ appears to form a hydrophobic gate with Val³¹³ and Val³¹⁶ in the E2P state (fig. S20B), which may serve as a Mn²⁺ selection filter. To compare TM helix interactions between SERCA1a, SERCA2b, and SPCA1a, a quantitative analysis was performed by counting the number of H-bonds formed between different TM helices (Fig. 6A and fig. S20C). Whereas TM6 forms 10 and 11 H-bonds with neighboring TMs in SERCA1a and SERCA2b, respectively, only five interdomain H-bonds were found for TM6 of SPCA1a in the E1-Mn²⁺-AMPPCP state (Fig. 6, A to D), and similar observations were made for the E2P state (Fig. 6, A and E to G). Interdomain interactions involving TM6 are depicted in Fig. 6 (B to G). It is thus obvious that smaller numbers of H-bonds are formed between TM6 and its neighboring TM helices in SPCA1a, compared to those in SERCA1a and SERCA2b. These structural features suggest that TM6 of SPCA1a may be more mobile than in SERCAs, possibly explaining the capability to bind Mn²⁺ and Ca²⁺ (Fig. 3, D to G) and show the wide metal specificity (Fig. 1A). The Q747A gain-of-function mutation enhanced the SPCA1-mediated uptake of Mn²⁺ (17), while the corresponding Q783A mutation in yeast PMR1 blocked Mn²⁺ transport (46, 47), suggesting that the Mn²⁺ transport activity of SPCA1a is highly sensitive to the residues or local structure surrounding the metal-binding site.

Fig. 6. H-bond analysis based on cryo-EM structures of SPCA1a.

(A) Number of H-bonds formed between each TM helix and its neighboring regions in the E1-Ca²⁺/Mn²⁺-AMPPCP (blue) and E2-BeF₃⁻ (salmon) states of SERCA1a (s1a), SERCA2b (s2b), and SPCA1a (spca). (B to G) Close-up views of interdomain H-bonds involving TM6 in the E1-Mn²⁺-AMPPCP state of SPCA1a (B), the E1-2Ca²⁺-AMPPCP state of SERCA1a (C; PDB ID: 3AR2), the E1-2Ca²⁺-AMPPCP state of SERCA2 (D; PDB ID: 6LLE), the E2P state of SPCA1a (E), the E2P state of SERCA1a (F; PDB ID: 3B9B), and the E2P state of SERCA2b (G; PDB ID: 6LLY). Ca²⁺ and Mn²⁺ ions are represented as green and purple spheres, respectively. Residues of TM5, TM6, and TM9 are colored yellow, green, and magenta, respectively. TM helix numbers are indicated in parentheses: 5–6 in Asn⁷³⁰ (5–6) in (B) refers to the loop between TM5 and TM6. 9−10 in Phe⁹⁵⁷ (9−10) in (C) and Phe⁹⁵⁶ (9−10) in (D) refer to the loop between TM9 and TM10.

We compared the subclass structures of the E1-Ca²⁺-AMPPCP state determined by 3D variability analysis using cryoSPARC (fig. S21). Of note, three cytosolic domains move jointly in the same directions in the E1-Ca²⁺-AMPPCP state (fig. S21, A to D), mimicking the onset of the transition from E1-ATP to E2P. Similar behavior was also observed for SERCA2b T1032stop (27), a SERCA2b mutant lacking the luminal C-terminal tail (C-tail) and displaying the higher ATPase activity than the wild type (48). Thus, the removal of the C-tail resulted in heterogeneous conformational subclasses in both E1-2Ca²⁺-ATP and E2P states of SERCA2b. SPCA1a with 10 TM helices and no subsequent extension segments is unlikely to undergo conformational regulation observed with SERCA2b. In this context, the density of the A-domain was less clear in the Mn²⁺-bound state than in the Ca²⁺-bound one (figs. S11 and S12), suggesting the greater dynamics of the A-domain in the former state. Given that the location and position of the cytosolic domains are closely coupled to those of the TM helices in SPCA1a, as seen in other ATPases (4), the different dynamic properties of the A-domain in the Ca²⁺- and Mn²⁺-bound states may explain the different turnover rates in the Ca²⁺- and Mn²⁺-dependent ATPase cycles of SPCA1a (Fig. 1A).

In the transition from E1-ATP to E2P states, we found that TM2 of SPCA1a is partially twisted on the cytosolic side, while TM2 of SERCA2b remains straight (fig. S18). Detailed comparison of TM helix movements induced by the E1-ATP to E2P transition revealed that whereas TM3 of SERCA2b rotates with its cytosolic end anchored (fig. S22, A to C), the whole part of TM3 rotates away from the original position in SPCA1a (fig. S22, D to F). In addition, a loop of the P-domain approaches the twisted site of TM2 in the E2P state of SPCA1a, forming a close contact between Lys³³³ (P-domain) and Ser¹²⁴ (twisted site of TM2; fig. S22E, inset). Although SERCA2b also contains a similar contact between Val³³³ (P-domain) and Gln¹⁰⁸ (TM2; fig. S22B, inset), its TM2 remains straight in the E2P state.

This variable behavior of TM2 may suggest its potential regulatory role in the Ca²⁺/Mn²⁺ transport of SPCA1. TM2 twisting also appears to happen in SERCA1 in E2-derivative states (E2, E2·Pi, E2-ATP, E2ADP, etc.; fig. S22, G to P). A salt bridge is formed between Glu¹¹³ and Arg³³⁴ (fig. S22, G to P, inset), possibly compensating for the structural stability impaired by TM2 twisting in the E2-derivative states. According to the structure-based sequence alignment, Glu¹¹³ and Arg³³⁴ in SERCA1 correspond to Glu¹²⁵ and Arg³³³ of SPCA1a, respectively, indicating that these two charged residues are conserved in SPCA1a (fig. S1).

We note that TM2 twisting happens in Na⁺/K⁺ ATPase and H⁺/K⁺ ATPase as in SPCA1a and SERCA1 (fig. S23, A and B), and that, in almost all cases, the twist is seen in the E2-derivative states. By contrast, TM2 twisting was not observed in ATPases with other functions than metal ion transport, such as ATP13A with polyamine transport activity and P4-ATPase with lipid translocation activity (fig. S23, C and D) (39, 49). Thus, the TM2 twisting seems likely to be an essential event of metal transporting P-type ATPases.

Previous molecular dynamics simulation works demonstrated that in SERCA, the kinked part of TM1 and the middle region of TM4 constitute a primary Ca²⁺ access point, where Asp⁵⁹ (TM1) and Glu³⁰⁹ (TM4) likely serve as critical residues for Ca²⁺ attraction (50–52). The superimpositions of TM1, TM2, and TM4 between SERCA1a, SERCA2b, and SPCA1a in the E1-Ca²⁺-AMPPCP state demonstrate that the relative positions of TM1, TM2, and TM4 are almost identical among these three pumps and that Asp⁵⁹ and Glu³⁰⁹ in SERCA correspond to Asn⁸¹ and Glu³⁰⁸ in SPCA1a (fig. S19A). In the E2P state, however, TM1 moves closer to TM4 in SPCA1a than in SERCA1a and SERCA2b (fig. S19B). In the meantime, the cytosolic side of TM2 gets twisted in SPCA1a, whereas the corresponding part of SERCA1a and SERCA2b remain intact (fig. S19B). Notably, the TM domain arrangement of SPCA1a in the E2P state is highly similar to that in the E2·Pi state of SERCA. Thus, the TM2 twisting seems to be closely coupled to the rearrangement of its neighboring TM1 and TM4, and thereby regulates the entry and release of Ca²⁺ in both SPCA1a and SERCA. In this context, the closure of the luminal Ca²⁺ release gate is seen in the E2P state of SPCA1a (Fig. 4G), whereas SERCA2b undergoes this gate closure in the subsequent E2·Pi state (53). These structural findings may suggest that SPCA1a releases the metal ions at earlier steps than SERCA. Future structural studies of SPCA1a in other intermediate states, in combination with computational and mutational analyses based on structural and phenotypic information, will thoroughly reveal the mechanisms of metal ion recognition and release by SPCA1a.

MATERIALS AND METHODS

Expression of SPCA1a

A PiggyBac Cumate Switch Inducible Vector harboring the coding region of human SPCA1 isoform 1a, with a PA-tag (GVAMPGAEDDVV) and a tobacco etch virus (TEV) cleavage site at the N terminus, was introduced into human embryonic kidney (HEK) 293T cells along with the Super PiggyBac Transposase Expression Vector (System Bioscience, LLC, CA, USA). The resultant cell line inducibly expressing SPCA1a was cultured in Dulbecco’s modified Eagle’s medium supplemented with 4% inactivated fetal calf serum and incubated in a humidified incubator with 5% CO₂ at 37°C. After 2 days of incubation, expression of SPCA1a was induced with cumate (150 μg/ml) and 50 nM phorbol 12-myristate 13-acetate (PMA). Cells were incubated at 37°C for another 30 to 50 hours, harvested by centrifugation at 1000 to 2000g for 15 min, frozen in liquid N₂, and stored at −80°C before purification.

Immunofluorescence

To analyze the localization of overexpressed SPCA1a in cells, the inducible HEK293T cells were plated onto poly-l-lysine–coated coverslips. Expression of SPCA1a was induced with cumate (150 μg/ml) and 50 nM PMA. Cells were fixed with phosphate-buffered saline (PBS) containing 4% paraformaldehyde, permeabilized with PBS containing 0.1% Triton X-100, and blocked with PBS containing 2% fetal calf serum in PBS for 45 min at room temperature. Cells were then incubated with antibodies against PA-tag [Fujifilm-Wako, Sendai, Miyagi, Japan; 1:5000 for immunofluorescence (IF)] and GM130 (PM061; MBL, Tokyo, Japan; 1:2000 for IF) overnight at 4°C. CF488-conjugated anti-rabbit immunoglobulin G (IgG) and CF568-conjugated anti-rat IgG antibodies (Biotium, San Francisco Bay Area, CA, USA; 1:2000 for IF) were used as secondary antibodies. Nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI) (Nacalai, Kyoto, Japan; 1:10,000 for IF). Fluorescent images were obtained using an FV1000 confocal microscope (Olympus, Tokyo, Japan) equipped with a UPLSAPO 60× silicon oil-immersion objective lens (numerical aperture, 1.30).

Immunoprecipitation

HEK293T cells from a 6-cm dish were harvested, washed twice with PBS, and resuspended in 1 ml of lysis buffer comprising 50 mM Hepes-NaOH (pH 7.0), 100 mM NaCl, 20% glycerol, 1 mM CaCl₂, 1 mM MgCl₂, 1% n-dodecyl-β-d-maltoside (DDM), and 1/100 Protease Inhibitor Cocktail (Nacalai). The solution was gently mixed by rotation at 4°C for 1 hour and then centrifuged at 17,000g for 10 min. Supernatant (800 μl) was collected and gently mixed with 1 μg of anti-ATP2C1 antibody (WH0027032M1; Sigma-Aldrich, Burlington, MA, USA) by rotation at 4°C for 1 hour. Forty-microliter slurry of protein G Sepharose, prewashed with Wash buffer comprising 50 mM Hepes-NaOH (pH 7.0), 100 mM NaCl, 20% glycerol, 1 mM CaCl₂, 1 mM MgCl₂, 0.02% glyco-diosgenin (GDN), and 1/100 Protease Inhibitor Cocktail (Nacalai), was added into the supernatant and gently mixed by rotation at 4°C for 1 hour. The protein G Sepharose was then washed five times with 500 μl of Wash buffer and terminated by adding 40 μl of SDS buffer for the subsequent SDS–polyacrylamide gel electrophoresis.

Purification of Usp2cc

The Usp2cc deubiquitination enzyme was purified according to a previous report (35). In brief, referring to the amino acid sequence of GenBank accession AAV27298.1, the DNA sequence of Usp2cc was optimized for expression in E. coli, ordered from Eurofins Genomics (Tokyo, Japan), and inserted into the pET15b vector by Gibson Assembly. After purification using an Ni-NTA affinity column (35), SEC was carried out on a Superdex 75 10/300 GL column (GE Healthcare, Chicago, IL, USA) at 4°C with buffer containing 50 mM MES (pH 7.4), 300 mM NaCl, 20% glycerol, and 1 mM dithiothreitol (DTT). Peak fractions were pooled, concentrated, and stored at −80°C before use.

Purification of SPCA1a

All purification steps were conducted at 4°C. To prepare Ca²⁺-bound SPCA1a, harvested cells were resuspended in Buffer A comprising 50 mM Hepes-NaOH (pH 7.0), 100 mM NaCl, 20% glycerol, 1 mM CaCl₂, 1 mM MgCl₂, 1 mM DTT, 1 mM phenylmethylsulfonyl fluoride (PMSF), and 1/100 Protease Inhibitor Cocktail (Nacalai) and were homogenized using a Dounce homogenizer. After homogenization, 1% (w/v) DDM was added to solubilize membranes. The solution was gently mixed by rotation at 4°C for 1.5 hours and then centrifuged at 22,300g for 1 hour to remove insoluble material. The supernatant was centrifuged again for another 15 min in a new tube to remove remaining impurities, and the supernatant was incubated for 3 hours with 5 ml of anti–PA-tag Antibody Beads (Fujifilm-Wako), preequilibrated with Buffer B [50 mM Hepes-NaOH (pH 7.0), 100 mM NaCl, 20% glycerol, 1 mM CaCl₂, 1 mM MgCl₂, and 1 mM DTT]. Beads were washed with 50 ml of Buffer B + 0.02% GDN (Anatrace, Maumee, OH, USA), followed by elution with 10 ml of Buffer B + 0.02% GDN and PA peptide (0.2 mg/ml), followed by 40 ml of Buffer B + 0.02% GDN. The elution fraction was concentrated using an Amicon Ultracel -50K centrifugal filter (Merck, Darmstadt, Germany). The concentrated SPCA1a was treated with 200 to 300 pmol of Usp2cc at 4°C for 1 hour to cleave off ubiquitin, and the ubiquitin-free SPCA1a was purified by SEC on a Superose 6 Increase 10/300 GL column (GE Healthcare) at 4°C preequilibrated with buffer containing 50 mM MES (pH 6.0), 100 mM NaCl, 1 mM CaCl₂, 1 mM MgCl₂, 1 mM DTT, and 0.02% GDN. The nanobody (or megabody) complex with SPCA1a was prepared by mixing SPCA1a with nanobody (or megabody) at a quantity ratio of 1:1.5 to 1:2 (a 1:1 molar ratio should also work), along with 200 to 300 pmol of Usp2cc at 4°C for 1 hour, and this was subjected to SEC purification. Peak fractions were pooled, concentrated to 2 to 6 mg/ml using an Amicon Ultracel -100K centrifugal filter (Merck), and kept on ice before use.

To prepare Mn²⁺-bound SPCA1a for cryo-EM analysis, the purification procedures were slightly modified as follows. Buffer A was replaced by Buffer A_Mn [50 mM Hepes-NaOH (pH 7.0), 100 mM NaCl, 20% glycerol, 2 mM EGTA, 1 mM MgCl₂, 1 mM DTT, 1 mM PMSF, and 1/100 Protease Inhibitor Cocktail]. After solubilization with Buffer A_Mn + 1% DDM, 1 mM MnCl₂ was added to the sample before centrifugation. During the purification with anti–PA-tag Antibody Beads (Fujifilm-Wako), Buffer B was replaced by Buffer B_Mn [Hepes-NaOH (pH 7.0), 100 mM NaCl, 20% glycerol, 1 mM MnCl₂, 1 mM MgCl₂, and 1 mM DTT]. At the last purification step by SEC, the running SEC buffer was replaced by the SEC_Mn buffer [50 mM MES (pH 6.0), 100 mM NaCl, 1 mM MnCl₂, 1 mM MgCl₂, 1 mM DTT, and 0.02% GDN].

Nanobody selection from the library

Nanobody selection was performed following the recommended protocol (36) with some modifications. In brief, 5 × 10⁹, 3 × 10⁷, 2 × 10⁷, and 2 × 10⁷ induced yeast cells were used for the first, second, third, and fourth rounds, respectively. The working volume was 5 ml for the first round and 1 ml for subsequent rounds. Before each round of magnetic-activated cell sorting (MACS) selection, a preclearing step was performed, in which induced yeast cells were washed and resuspended in buffer [50 mM MES (pH 6.0), 100 mM NaCl, 1 mM CaCl₂, 1 mM MgCl₂, 1 mM DTT, 0.02% GDN, and 20 μM AMPPCP], incubated with 200 nM anti–PA-FITC [Fujifilm-Wako, in-house FITC-labeled with the Pierce FITC Antibody Labeling Kit (Thermo Fisher Scientific, Waltham, MA, USA)] antibody and anti-FITC microbeads (Miltenyi, Bergisch Gladbach, North Rhine-Westphalia, Germany) at 4°C for 40 min, and then passed through an LD column to remove clones nonspecifically bound to beads and anti–PA-FITC antibody. The precleared yeast cells were then incubated with a preincubated complex of PA-SPCA1a and anti–PA-FITC at 4°C for 1 hour and passed through an LS column (Miltenyi) to collect binders. Successively lower concentrations of PA-SPCA1a from the second round (500, 500, 100, and 20 nM) were used for selection to enrich nanobodies with higher binding affinity. The collected yeast cells were then recovered and induced for subsequent selections. After four rounds of MACS selection, yeast cells were stained with anti–PA-FITC and sorted into 96-well plates by fluorescence-activated cell sorting (FACS) using a BD FACSMelody cell sorter (BD Biosciences, Franklin Lakes, NJ, USA). Clones were stained with anti–PA-FITC and analyzed by flow cytometry with an Accuri C6 flow cytometer (BD Biosciences) to screen for binders.

Expression and purification of nanobodies

Nanobody genes were isolated from yeast cells using the GenTLE (from Yeast) High Recovery Kit (Takara, Kyoto, Japan), sequenced, and cloned into the expression vector pET151 with a pelB signal peptide inserted at the N terminus, resulting in the pET151_pelB_Nb14 construct. The construct was transformed into E. coli BL21 (DE3), and cells were grown in Terrific Broth and induced with 1 mM isopropyl β-d-thiogalactopyranoside (IPTG) at 25°C for 15 hours when the absorbance at 660 nm (OD₆₆₀) reached 0.6 to 0.8. After culturing, cells were washed with 1× PBS, pelleted, and resuspended in 100 ml of sucrose buffer [500 mM sucrose, 200 mM tris (pH 8), and 0.5 mM EDTA]. After stirring on ice for 15 min, 200 ml of chilled deionized water was added to release periplasmic nanobody by stirring for a further 45 min. The lysate was then mixed with 150 mM NaCl, 2 mM MgCl₂, and 20 mM imidazole and centrifuged at 22,300g at 4°C for 20 min. The supernatant was mixed with 5 ml of Ni-NTA resin (Qiagen, Hilden, Düsseldorf, Germany), preequilibrated with normal buffer (150 mM NaCl and 2 mM MgCl₂), and stirred on ice for 1 hour. The mixture was then applied to a column and washed twice with a high-salt buffer [20 mM Hepes (pH 7.5), 500 mM NaCl, and 20 mM imidazole] and then twice with NH7.5 buffer [20 mM Hepes (pH 7.5) and 100 mM NaCl] + 20 mM imidazole. Nanobodies were eluted with NH7.5 buffer + 400 mM imidazole and purified by SEC on a Superdex200 10/300 GL increase column (GE Healthcare).

Expression and purification of megabodies

The scaffold protein c7HopQ was used to generate a megabody, as previously reported (37). Briefly, the DNA sequence of c7HopQ was inserted into the pET151_pelB_Nb14 vector as described in the Supplementary Materials. After transforming the construct into E. coli BL21 (DE3), cells were grown in Terrific Broth and induced with 1 mM IPTG at 25°C for 20 hours after reaching OD₆₆₀ ~ 1. Induced cells were washed with 1× PBS, pelleted, and resuspended in 100 ml of sucrose buffer [500 mM sucrose, 200 mM tris (pH 8), and 0.5 mM EDTA]. After stirring on ice for 20 min, 200 ml of chilled deionized water was added to release periplasmic megabody by stirring for a further 45 min. The lysate was mixed with 150 mM NaCl, 2 mM MgCl₂, and 20 mM imidazole and centrifuged at 22,300g at 4°C for 20 min. The supernatant was mixed with 5 ml of Ni-NTA resin (Qiagen) preequilibrated with normal buffer (150 mM NaCl and 2 mM MgCl₂) and stirred on ice for 1 hour. The mixture was applied to an open column and washed twice with a high-salt buffer [20 mM Hepes (pH 7.5), 500 mM NaCl, and 20 mM imidazole] and then twice with NH7.5 buffer [20 mM Hepes (pH 7.5) and 100 mM NaCl] + 20 mM imidazole. Megabodies were eluted with NH7.5 buffer + 400 mM imidazole and purified by SEC on a Superdex200 10/300 GL increase column (GE Healthcare).

ATPase activity assay

The SPCA1a protein used for the ATPase activity assay was purified essentially as described above. However, we used the buffer deprived of CaCl₂ throughout the purification processes. The reaction mixture composed of 100 mM Hepes (pH 7.0), 100 mM NaCl, 0.02% GDN, 1 mM EGTA, 5 mM ATP, and appropriate combinations of MgCl₂ and CaCl₂ (0.000944 M CaCl₂ and 0.011916 M MgCl₂), MnCl₂ (0.001037 M MnCl₂ and 0.011907 M MgCl₂), CoCl₂ (0.001029 M CoCl₂ and 0.0119148 M MgCl₂), or NiSO₄ (0.0010375 M NiSO₄ and 0.0119073 M MgCl₂) to fix the final Ca²⁺, Mn²⁺, Co²⁺, or Ni²⁺ concentration to 10 μM. After preincubation at 37°C for 10 min, the ATP hydrolysis reaction was initiated by adding 200 nM SPCA1a proteins. To generate free metal ions at 10 μM, a publicly available program WEBMAX EXTENDED (https://somapp.ucdmc.ucdavis.edu/pharmacology/bers/maxchelator/webmaxc/webmaxcE.htm) was used for calculation. The amount of inorganic phosphate released from the reactions was quantified using the EnzCheck Phosphate Assay Kit (Thermo Fisher Scientific). The absorbance at 360 nm (A360) was measured with a U-3900 spectrophotometer (Hitachi, Chiyoda, Tokyo, Japan). For the ATPase activity with 10 μM MnCl₂, CoCl₂, and NiCl₂, 100 nM SPCA1a protein was used due to the highest limit of A360 of the spectrophotometer used.

Grid preparation for cryo-EM

Purified SPCA1a proteins were mixed with inhibitor solutions at final concentrations of 1 mM AMPPCP for the E1-Ca²⁺/Mn²⁺-AMPPCP form and 1 mM BeSO₄, 5 mM NaF, and 5 mM EGTA for the E2-BeF₃⁻ form. After incubation for 1 hour on ice, protein solutions (3 μl) were applied to a freshly glow-discharged Quantifoil honey carbon grid (R 1.2/1.3, Cu, 300 mesh) using a Vitrobot Mark IV instrument (Thermo Fisher Scientific) at 4°C with a blotting time of 4 s under 100% humidity conditions. Grids were plunge-frozen in liquid ethane.

Cryo-EM data collection

For the E1-Ca²⁺-AMPPCP state of SPCA1a complexed with Nb14, data were collected using a Krios G4 microscope (Thermo Fisher Scientific) operated at 300 kV; 3789 movies were acquired using a Gatan K3 Summit direct electron detector with a Gatan Quantum-LS Energy Filter (Gatan, Pleasanton, California, USA) in electron counting mode. Movies were collected at a magnification of ×105,000 with a calibrated pixel size of 0.83 Å per pixel (University of Tokyo, Japan). The electron flux was set to an accumulated exposure of 49 e⁻/Ų at the specimen. Data were automatically acquired by the image shift method using SerialEM software (54) with a defocus range from −1.8 to −0.84 μm.

For the E1-Ca²⁺-AMPPCP state of SPCA1a complexed with Mb14, data were collected using a CRYOARM 300 II microscope (JEOL, Akishima, Tokyo, Japan) operated at 300 kV; 5600 movies were acquired using a Gatan K3 Summit direct electron detector with an in-column Omega energy filter in electron counting mode. Movies were collected at a magnification of ×60,000 with a calibrated pixel size of 0.788 Å per pixel (Tohoku Medical Megabank Organization, Tohoku University, Japan). The electron flux was set to an accumulated exposure of 50 e⁻/Ų at the specimen. Data were automatically acquired by the image shift method using SerialEM software, with a defocus range from −1.6 to −0.8 μm.

For the E1-Mn²⁺-AMPPCP state of SPCA1a complexed with Mb14, data were collected using a CRYOARM 300 II microscope (JEOL) operated at 300 kV; 6075 movies were acquired using a Gatan K3 Summit direct electron detector with an in-column Omega energy filter in electron counting mode. Movies were collected at a magnification of ×60,000 with a calibrated pixel size of 0.788 Å per pixel (Mega bank). The electron flux was set to an accumulated exposure of 50 e⁻/Ų at the specimen. Data were automatically acquired by the image shift method using SerialEM software, with a defocus range from −1.6 to −0.8 μm.

For the E2-BeF₃⁻ form of SPCA1a, data were collected using a Titan Krios G3i microscope (Thermo Fisher Scientific) operated at 300 kV; 3501 movies were acquired using a Gatan K3 Summit direct electron detector with a Gatan Quantum-LS Energy Filter (GIF) in electron counting mode. Movies were collected at a magnification of ×105,000 with a calibrated pixel size of 0.83 Å per pixel (University of Tokyo, Japan). The electron flux was set to an accumulated exposure of 49 e⁻/Ų at the specimen. Data were automatically acquired by the image shift method using SerialEM software, with a defocus range from −1.8 to −0.8 μm.

Cryo-EM data processing

Motion correction and Bayesian polishing were performed in RELION-3 (55), and other steps were performed in cryoSPARC (56). Dose-fractionated movies were subjected to beam-induced motion correction (implemented in RELION-3) and then imported into cryoSPARC.

For the E1-Ca²⁺-AMPPCP state of SPCA1a complexed with Nb14, contrast transfer function (CTF) estimation was performed through patch CTF estimation implemented in cryoSPARC. Particles were picked using a cryoSPARC Blob picker and extracted with downsampling to a pixel size of 3.32 Å per pixel. These particles were subjected to several rounds of 2D classification. The best particles were selected and subjected to cryoSPARC template picker, and again extracted for several rounds of 2D classification. Best particles were then subjected to ab initio reconstruction and heterogeneous refinement. Particles belonging to the best class were reextracted with a pixel size of 1.0375 Å per pixel and subjected to another three rounds of ab initio reconstruction and heterogeneous refinement, resulting in 142,884 particles and a map at a global resolution of 3.4 Å. These particles were exported to RELION-3 for Bayesian polishing and imported back into cryoSPARC for nonuniform refinement along with local defocus refinement and global CTF refinement, resulting in an improved resolution of 3.3 Å. The improved map was then subjected to local resolution estimation and filtered using a local filtering function.

For the E1-Ca²⁺-AMPPCP state of SPCA1a complexed with Mb14, CTF estimation was performed through patch CTF estimation implemented in cryoSPARC. Particles were picked using a cryoSPARC Blob picker and extracted with downsampling to a pixel size of 3.32 Å per pixel. These particles were subjected to several rounds of 2D classification. The best particles were selected, subjected to cryoSPARC template picker, and again extracted for several rounds of 2D classification. The best particles were then subjected to ab initio reconstruction and heterogeneous refinement. Particles belonging to the best class were reextracted with a pixel size of 0.985 Å per pixel and subjected to nonuniform refinement, resulting in 385,430 particles and a map at a global resolution of 3.1 Å. These particles were exported to RELION-3 for Bayesian polishing and imported back into cryoSPARC for nonuniform refinement and 3D variability analysis, divided into seven frames using the intermediate mode in 3D variability analysis. The first, fourth, and seventh frames resulting from 3D variability analysis were chosen as subclass 1, subclass 2, and subclass 3, respectively. The chosen particles were then subjected to nonuniform refinement along with local defocus refinement and global CTF refinement, resulting in resolutions of 3.1, 3.12, and 3.14, respectively. The improved maps were subjected to local resolution estimation and filtered using a local filtering function.

For the E1-Mn²⁺-AMPPCP state of SPCA1a complexed with Mb14, CTF estimation was performed through patch CTF estimation implemented in cryoSPARC. Particles were picked using a cryoSPARC Blob picker and extracted with downsampling to a pixel size of 3.32 Å per pixel. These particles were subjected to several rounds of 2D classification. The best particles were selected and subjected to cryoSPARC template picker and again extracted for several rounds of 2D classification. The best particles were then subjected to ab initio reconstruction and heterogeneous refinement. Particles belonging to the best class were reextracted with a pixel size of 0.985 Å per pixel and subjected to nonuniform refinement, resulting in 318,234 particles and a map at a global resolution of 3.20 Å. These particles were exported to RELION-3 for Bayesian polishing and imported back into cryoSPARC for nonuniform refinement and 3D variability analysis, divided into five frames by the intermediate mode in 3D variability analysis. The first, third, and fifth frames resulting from 3D variability analysis were chosen as subclass 1, subclass 2, and subclass 3, respectively. The chosen particles were then subjected to nonuniform refinement along with global CTF refinement, resulting in resolutions of 3.19, 3.11, and 3.25 Å, respectively. The improved maps were subjected to local resolution estimation and filtered using a local filtering function.

For the E2-BeF₃⁻ form, CTF estimation was performed with CTFFIND4 (57). Particles were picked using the Blob picker in cryoSPARC and extracted with downsampling to a pixel size of 4.4 Å per pixel. These particles were subjected to five rounds of 2D classification and one round of ab initio reconstruction. The best particles were selected, subjected to the template picker in cryoSPARC, and again extracted for five rounds of 2D classification and one round of ab initio reconstruction. The best particles were reextracted with a pixel size of 1.17 Å per pixel and subjected to local CTF refinement and nonuniform refinement (58), resulting in a map at a global resolution of 3.4 Å. These particles were exported to RELION-3 for Bayesian polishing using pyem scripts (59) and imported back into cryoSPARC for nonuniform refinement, resulting in an improved resolution of 3.3 Å.

Model building and validation

The SPCA1a structure predicted by AlphaFold (60) (UniProt: P98194) for the E2-BeF₃⁻ state, and a Swiss homology model for the E1-Ca²⁺/Mn²⁺-AMPPCP state were used as the initial atomic models. The models were fitted into the refined density maps using UCSF Chimera (61) and built using Coot (62). Because of the poor density, the following sequence regions were not modeled: N–19, 67–69, 205–210, and 905–919 of SPCA1a and 135–139 of the nanobody head piece in the E1-Ca²⁺-AMPPCP state subclass 2; N–22, 67–68, 205–210, and 905–919 of SPCA1a and 135–139 of the nanobody head piece in the E1-Mn²⁺-AMPPCP state subclass 1; N–16, 65–70, 128–134, 205–208, and 904–919 in the E2-BeF₃⁻ state. Model refinement was performed using phenix.real_space_refine (63) and validated with phenix.molprobity (64) in Phenix, as summarized in Table 1. Structure figures were prepared using UCSF ChimeraX (65).

Table 1. Cryo-EM data statistics.

	Nano-Ca2+-AMPPCP	Mega-Ca2+-AMPPCP			Mega-Mn2+-AMPPCP			BeF3−
		Subclass 1	Subclass 2	Subclass 3	Subclass 1	Subclass 2	Subclass 3
Data collection and processing
Magnification	105,000	60,000						105,000
Voltage (kV)		300
Microscope	Titan Krios G4	CRYOARM 300 II						Titan Krios G3i
Detector		K3 summit
Electron exposure (e−/Å)	49	50						49
Defocus range (μm)	−1.8 to −0.84	−1.6 to −0.8						−1.8 to −0.8
Pixel size (Å)	0.83	0.788						0.83
Symmetry imposed		C1
Number of movies	3789	5600			6075			3501
Initial particle images (no.)	2,892,219	4,104,763			5,030,147			1,945,070
Final particle images (no.)	142,884	91,849	305,692	102,741	151,371	298,318	161,692	260,172
Map resolution (Å)	3.36	3.1	3.12	3.14	3.19	3.11	3.25	3.3
FSC threshold		0.143
Map resolution range (Å)	2.84–30	2.71–30	2.13–30	2.71–30	2.72–30	2.66–30	2.7–30	2.80–30
Map sharpening B factor (Å2)	−104.88	−81.86	−111.48	−85.61	−100	−163.59	−132.53	−115.5
Model building and refinement
Model resolution (Å)	BLANK	3.3	3.3	3.3	3.5	BLANK		3.3
FSC threshold		0.5
Model composition
Atoms	BLANK	15,470 (H: 7804)	7666	7678	7649	BLANK		6672
Amino acid residues		999	999	1001	997			871
Ligands		CA:1 ACP:1			MN:1 ACP:1			MG:1 BEF:1
Root mean square deviations
Bond lengths (Å)	BLANK	0.005	0.006	0.005	0.005	BLANK		0.003
Bond angles (°)		1.115	1.075	1.029	0.912			0.529
Validation
Molprobity score	BLANK	1.28	1.6	1.42	1.77	BLANK		1.33
Clashscore		4.52	8.86	7.62	9.01			6.05
Rotamer outliers (%)		0.12	0.35	0.35	1.66			0.13
EMRinger		1.88	1.63	1.64	0.39			2.11
Ramachandran plot
Favored (%)	BLANK	97.78	97.37	98.18	97.37	BLANK		98.38
Allowed (%)		2.22	2.63	1.82	2.63			1.62
Outliers (%)		0	0	0	0			0
CaBLAM outliers (%)		1.23	1.33	1.43	1.02			1.29

Main-chain rotation analysis

SPCA1a (Mb14, Ca²⁺, subclass 2), SERCA1a [Protein Data Bank (PDB) ID: 3AR2], and SERCA2b (PDB ID: 6LLE) were used as structures in the E1-Ca²⁺-AMPPCP state. SPCA1a (E2-BeF₃⁻), SERCA1a (PDB ID: 3B9B), and SERCA2b (PDB ID: 6LLY) were used as structures in the E2-BeF₃⁻ state. Domains were defined as A-, N-, and P-domains and TM1–2, TM3–4, TM5–6, and TM7–10. Structures were aligned over TM7–10 for calculating the main-chain rotation. The rotation axes and the center points of the axes were calculated in ChimeraX. 3D coordinates of the E1-Ca²⁺-AMPPCP models and E2P state models, and those of the center point of rotation axes, were merged and cleaned using in-house scripts. Angles were calculated between two vectors from the center point of the rotation axes to Cα atoms in E1-Ca²⁺-AMPPCP and E2-BeF₃⁻ states.

H-bond analysis

H-bonds were detected in UCSF ChimeraX (version 1.4). H-bonds formed between side-chain atoms and other atoms in different domains were counted, while H-bonds formed between side-chain (or main-chain) atoms and other atoms in the same domains were excluded using in-house scripts.

Other software and packages used in this study

Statistical analysis and figure preparation were performed in RStudio version 2021.9.0.351 (RStudio Inc.) with R version 4.1.0 installed. Various R packages were used, including tidyverse (1.3.1), ggplot2 (3.3.5), ggpubr (0.4.0), rstatix (0.7.0), and CytoExploreR (1.1.0). pyKVFinder (66) was used for cavity calculation. center_of_geometry() in the MDAnalysis package (www.mdanalysis.org/) was used for determining the center of AMPPCP (67, 68).

Supplementary Materials

This PDF file includes:

Supplementary Text

Figs. S1 to S23

Other Supplementary Material for this : manuscript includes the following:

Movie S1

View/request a protocol for this paper from Bio-protocol.