Experimental conditions were fine-tuned to obtain the inactivated state. The channels were prepared in 2 mM free Ca²⁺, a concentration that inactivates the channel. As RyR1 is constitutively bound to the sensitizing ATP in the muscle cell (Kushmerick et al., 1992; Meissner et al., 1986; Xu et al., 1996), we included its non-hydrolyzable form, AMP-PCP (ACP), for its structural determination. In order to visualize the direct effect of Ca²⁺ on RyR1’s conformation, FKBP12, a stabilizer of the cytoplasmic domain, was excluded. To determine the structure of RyR1 in a detergent-free, membrane-embedded natural state, RyR1 was reconstituted into membrane scaffold protein (MSP) 1E3D1 nanodiscs in the presence of phosphatidylcholine, an abundant phospholipid in membrane fraction preparations. As a control, we carried out cryo-EM and 3D reconstruction of RyR1 using identical buffer conditions and saturating ACP, and substituted Ca²⁺ by 1 mM EGTA plus 1 mM EDTA (dataset denominated RyR1-ACP/EGTA). The channels were also reconstituted into nanodiscs in the presence of phosphatidylcholine.

The Ca²⁺-induced activity profile of rabbit RyR1 in SR membranes was determined using the tritiated ryanodine binding assay, which reflects the probability of channel opening. The assay indicated RyR1 channel inactivation above 0.1 mM Ca²⁺, with an IC₅₀ of 0.6 mM (Figure 1A). The presence of 2 mM ATP increased the efficacy of Ca²⁺-induced activation by approximately threefold at peak activation with similar potency for Ca²⁺-induced inactivation (IC₅₀ of 0.7 mM), in agreement with earlier results obtained in lipid bilayer (Laver et al., 1995). We replicated the ryanodine-binding experiments over the same Ca²⁺ concentration range with 2 mM ACP. Maximal ryanodine binding was ~1.5-fold higher in the presence of ACP compared to ATP, and in this case slightly higher Ca²⁺ concentration was necessary for the same degree of RyR1 inhibition, with an IC₅₀ value for Ca²⁺ of 1.5 mM in the presence of ACP. Thus, under our experimental conditions of 2 mM free Ca²⁺, RyR1 inactivation was partial.

Cryo-EM of rabbit RyR1 embedded in nanodisc in the presence of 2 mM ACP and 2 mM free Ca²⁺ followed by single-particle analysis and 3D classification revealed two classes of particles, with their pore either fully open or fully closed. The findings, reproduced in two independent datasets (A and B), reflected the partial inhibition observed at 2 mM free Ca²⁺ (pCa 2.7) in the ryanodine-binding studies (Figure 1A). The 3D reconstructions of RyR1-ACP/Ca²⁺ open reported here are the first RyR1 open structures obtained in the presence of natural activators alone, possibly owing to the use of nanodiscs. The RyR1-ACP/Ca²⁺ reconstructions with a closed pore suggest a distinct conformation that we refer to as RyR1-ACP/Ca²⁺ inactivated henceforth, as further supported by our data.

A symmetry of C4 was imposed after ascertaining the true fourfold symmetry of the protein. The closed-pore (inactivated) channels were resolved to 3.8 Å and 4.1 Å resolution for the A and B datasets, respectively, which improved to 3.5 Å and 3.8 Å, respectively, using a phase improvement procedure (Terwilliger et al., 2020). Despite this improvement, we followed a conservative approach and used the non-density-modified maps for subsequent analysis. The open conformation represented a smaller fraction of the data (13 and 19% for the A and B datasets), which limited the resolution to 4.6 Å and 5.8 Å, respectively (Figure 1—figure supplement 1, Figure 1—figure supplement 2, Figure 1—figure supplement 3). The RyR1-ACP/Ca²⁺_A open subset improved to a resolution of 4.0 Å after a symmetry expansion step. Unless specified, the reported observations correspond to dataset A.

Classification of the control RyR1-ACP/EGTA dataset yielded a single class with a resolution of 4.3 Å (0.143 FSC gold standard), but further symmetry expansion and focused classification using a quadrant-shaped mask increased resolution to 3.9 Å (Figure 1—figure supplement 4). Despite the presence of the ACP activator, the 3D reconstruction of RyR1-ACP/EGTA yielded a closed state. Tables 1 and 2 summarize the cryo-EM data collection, single-particle image processing, model quality attributes, and database IDs for all datasets, and Table 3 compiles the main characteristics of the three conformations analyzed (closed, open, inactivated).

Pore analysis using the program HOLE (Figure 1B and C) of RyR1-ACP/EGTA indicated a closed pore with a radius constricting to 1 Å at the known hydrophobic gate Ile4937. In both RyR1-ACP/Ca²⁺ open A and B datasets, the hydrophobic gate as appraised by the pore profile (Figure 1B, Figure 1—figure supplement 5) and as measured at the S6 helix backbone C_α atoms (for more accurate measurement according to the lower resolution of the two datasets) (Figure 1C, Figure 1—figure supplement 5) had diameters of 15.8 Å and 16.0 Å, which indicates an open, Ca²⁺-permeable pore in both cases. Pore dimensions are comparable to the pore diameter of RyR1 open structures obtained in the presence of activators such as Ca²⁺/PCB95 or Ca²⁺/caffeine/ATP (16.7 Å; PDB ID: 5TAL; des Georges et al., 2016). The slightly narrower pore in our case could be attributed to the millimolar instead of submicromolar Ca²⁺ in our case and/or extra activators in addition to endogenous physiological activators in earlier open structures.

The reconstructions corresponding to RyR1-ACP/Ca²⁺ inactivated of the A and B datasets displayed a pore diameter at the Ile4937 gate of ~2 Å when considering the side chains, and 10.7 Å and 11.3 Å diameter, respectively, when measured at the backbone C_α atoms, making it impermeable to Ca²⁺ ions (Figure 1B and C, Figure 1—figure supplement 5). The position of the S6 backbone at Ile4937 is similar to our closed structure (r = 10.4 Å) and to RyR1-FKBP12/EGTA (r = 10.3 Å; PDB ID: 5TB0; des Georges et al., 2016). Interestingly, the channel pore was narrower than 1 Å at Gln4933 in RyR1-ACP/Ca²⁺_A inactivated (Figure 1B and C).

We previously established that the large square-shaped cytoplasmic shell of the RyR undergoes a conformational change upon opening. The periphery of its four quadrants tilts downward (toward the membrane), while its inner corner tilts away from it, rotating around a pivot point (Samsó et al., 2009). Such tilt can be quantified with the flexion angle, whereby negative angle (downward) correlates with opening, positive angle with closing, and absence of FKBP lowers the flexion angle of closed states (Steele and Samsó, 2019). In general, the approximate ranges of flexion angles are +1° to +2° for RyR1-FKBP12/EGTA closed, –1° to –3° for RyR1-EGTA closed, 0° to –3° for RyR1-FKBP12 ‘primed’ (with activating ligands and closed pore), and –1.5° to –5° for RyR1-Ca²⁺ open with or without FKBP12 (Iyer et al., 2020; Steele and Samsó, 2019). Here, RyR1-ACP/Ca²⁺ open channels have flexion angles of –5.1° and –4.8° for the A and B datasets, within the expected range. The consensus 3D reconstruction of RyR1-ACP/EGTA had flexion angle of –2.2°, and its two classes had flexion angles of - 1.8° (24%) and –2.9° (76%) (Figure 1—figure supplement 4). But unexpectedly for closed-pore channels, RyR1-ACP/Ca²⁺ inactivated datasets have a negative flexion angle (–4.6° and –4.2° for A and B datasets, respectively) (Figure 1—figure supplements 1 and 3). As there was an indication of variability in the cytoplasmic shell of RyR1-ACP/Ca²⁺_A-inactivated dataset, we carried out fourfold symmetry expansion and focused classification for the monomer and obtained three subclasses, all showing downward motion: class 1 (22% of particles, 3.7 Å resolution, –3.6° flexion angle), class 2 (44% of particles, 3.3 Å resolution, –4.4° flexion angle), and class 3 (20% of particles, 3.8 Å resolution, –5.6° flexion angle) (Figure 1—figure supplement 1). Such pronounced downward flexion angles in a closed channel cannot be explained just by the lack of FKBP12 in our preparations. The distinct conformation of RyR1-ACP/Ca²⁺ inactivated, consisting of a closed pore and extreme-downward cytoplasmic assembly, prompted further analysis of the central region that joins the cytoplasmic and transmembrane domains.

In RyR1-ACP/Ca²⁺ open, the high-affinity Ca²⁺-binding site was formed by Glu3967, Glu3893 (from the CD; residues 3668–4070), and Thr5001 (from the CTD (C-terminal domain); residues 4957–5037) as reported earlier (des Georges et al., 2016), and additionally Gln3970 in our case, which was visible up to 20σ. The Ca²⁺-induced reorientation of CTD with respect to CD and subsequent separation of S6 was obvious when compared to the RyR1-ACP/EGTA structure (Figure 2A, Figure 2—figure supplement 1), consistent with previous reports of Ca²⁺-induced activation (Bai et al., 2016; des Georges et al., 2016).

An outstanding question is how higher Ca²⁺ concentration may result in inactivation. Potential mechanisms are an additional allosteric change that dampens the affinity of the Ca²⁺ activation site, or that the high-affinity Ca²⁺ site remains occupied while an additional conformational change overcomes activation. Consistent with the second scenario, the high-affinity Ca²⁺-binding site in RyR1-ACP/Ca²⁺ inactivated remained fully occupied by Ca²⁺ up to 20σ, although with additional out-of-plane tilting of the CD and CTD around the Ca²⁺-binding site. The Ca²⁺ ion was coordinated by Glu3893, Glu3967 (CD), and Thr5001 (CTD) within contact distances of 2.8 Å, consistent with the open structures. However, on the CD-C' side, Gln3970 lost coordination to Ca²⁺, interacting with Ser4029 instead (Figure 2A, Figure 2—figure supplement 1). Mutations Q3970E/K in RyR1 are implicated in central core disease and equivalent RyR2 mutations in cardiac arrhythmia (Chirasani et al., 2019), which highlights the important Ca²⁺ sensing role for this residue.

To understand how the fully occupied high-affinity Ca²⁺-binding site (CD/CTD interface) led to a closed pore in RyR1-ACP/Ca²⁺ inactivated, we looked for differences between RyR1-ACP/Ca²⁺ open and RyR1-ACP/Ca²⁺ inactivated in the domains spanning from the CD/CTD interface to the pore. Specifically, we focused on domains around the high-affinity Ca²⁺-binding site (CD, EF hand, U-motif, CTD, and the C′ section of S6 or S6C′; domain boundaries indicated in Figure 2), which we term ‘central block.’ A comparison was done for dataset A after confirming similarity for the central block of 0.95 Å Cα root mean square deviation (RMSD) for the encompassed 672 resolved residues, between inactivated A and B datasets (Figure 2—figure supplement 2A). The RMSD (Cα backbone) between individual domains from RyR1-ACP/Ca²⁺ inactivated versus open was below 1.5 Å for all comparisons when the domains were pre-aligned pairwise (Figure 2B, Figure 2—figure supplement 2B and C), indicating little conformational change. The N′ section of S6 was also included in the comparisons. Without pre-alignment, RMSD was higher (Figure 2B), indicating domain repositioning, the degree of which was estimated by subtracting the pre-aligned from the unaligned RMSDs. This yielded an average shift of ~2 Å per domain in going from RyR1-ACP/Ca²⁺ open to RyR1-ACP/Ca²⁺ inactivated, except for S6N’ where the shift was ~0.7 Å (Figure 2B). Comparing the same domains between RyR1-ACP/Ca²⁺ inactivated and RyR1-ACP/EGTA (closed) yielded smaller repositioning with average shifts of ~1 Å, reflecting similarity in the domain’s locations, with the exception of the EF hand domain that undergoes an ~5 Å repositioning (Figure 2B, Figure 2—figure supplement 2C).

When the central blocks were pre-aligned as a unit using their respective CD domains, and the individual domains were compared pairwise (CD-aligned; Figure 2—figure supplement 2A), there was a good overlap between RyR1-ACP/Ca²⁺ open and RyR1-ACP/Ca²⁺ inactivated (RMSD below 1.5 Å) except for S6N′ (3.8 Å), revealing the bend in the middle of S6 upon opening. In contrast, the CD-aligned RMSD of RyR1-ACP/Ca²⁺ inactivated vs. RyR1-ACP/EGTA closed was higher, between 3 and 4 Å for the EF hands, U-motif, and CTD, and increases toward the C-terminus, 5 Å for S6C′ and almost 7 Å for S6N′ (Figure 2B, ‘CD-aligned’). This indicated that domains CD, EF hands, U-motif, S6C′, and CTD relocated together, and that the central block became more compact in the presence of Ca²⁺ acting at the CD/CTD interface. Taken together, these results imply that conformational changes from CD are transmitted to S6 only in the presence of Ca²⁺ (Figure 2C and D).

We analyzed how the central block may evolve from RyR1-ACP/EGTA (closed), to RyR1-ACP/Ca²⁺ open, to RyR1-ACP/Ca²⁺ inactivated. Upon Ca²⁺-induced opening, ‘engagement’ of CD and CTD by Ca²⁺ tilts the CD out of plane (such that helix h tilts 3° inwards), while the CTD tilts up and outwards. This pulls the S6 helices (which are directly connected to the CTD) away from the pore axis, opening the ion pathway (Figure 2C and D). To reach the inactivated state, the central block tilts further (see rotation axis in Figure 2—figure supplement 2B) in a movement similar to pushing down the levers of a winged corkscrew, which pushes each S6 helix 2.5 Å toward the pore axis, closing the channel. Alpha helices within the CTD act as a lever coupling this out-of-plane rotation to S6C′ and as described further below, ATP reinforces the connection between CTD-S6C′ in the presence of Ca²⁺. These conformational changes involved in activation, inactivation, and transition from the inactivated to the closed state are shown in the morph among conformations in Figure 2—video 1 and illustrated in the schematics in Figure 2D.

The EF hand domain of RyR1, a candidate Ca²⁺ regulation site (Du et al., 1998; Gomez et al., 2016; Gomez and Yamaguchi, 2014; Xiong et al., 1998), did not show significant changes, with maximum Cα pre-aligned RMSD of 1.5 Å among the different states analyzed (closed, open, inactivated). The EF hand loops (residues 4081–4090, 4116–4123) were empty in our high Ca²⁺ conditions, which is consistent with the low affinity of Ca²⁺ to this site (K_d 3.7 mM, Xiong et al., 1998). Nonetheless, the entire EF hand domain, which protrudes from the CD-C′, repositions noticeably during activation, with a 3.4° counterclockwise in-plane rotation (as seen from the cytoplasmic side), and 6.8° upward out-of-plane rotation. This movement brings the EF hand domain in closer proximity to the S2–S3 loop (cytoplasmic loop between S2 and S3 TMD helices; residues 4664–4786) (Figure 3). With inactivation, an additional 1° counterclockwise rotation and 2.7° upward out-of-plane rotations define a physical limit to the counterclockwise motion, forming two salt bridges at this inter-subunit interface: Glu4075-Arg4736 and Lys4101-Asp4730, with their side chains within 3.5 Å (Figure 3A). This interaction is critical to support inactivation, as demonstrated by the fact that MH/CCD mutations facing this interface F4732D, G4733E, and R4736W/Q, with the latter including the Arg directly involved in the salt bridge (Figure 3B), greatly reduced channel inactivation (Gomez et al., 2016). On the other hand, MH/CCD mutations T4082M, S4113L, and N4120Y, in regions of the EF hand domain away from the interface (Figure 3B), did not affect RyR1 inactivation (Gomez et al., 2016), serving as a negative control for this hypothesis.

The ATP-binding site showed full occupancy in the three conformations – open, closed, and inactivated. Under closed-state conditions, ACP bound to the pocket formed by the U-motif, S6C′, and CTD (Figure 4A, ‘closed’), in the same position as ATP (des Georges et al., 2016), with a high map significance (7σ). An additional elongated non-protein density was associated to ACP in the RyR1-ACP/Ca²⁺ open and inactivated maps. This distinct density has high map significance (15σ) (Figure 4A, Figure 4—figure supplement 1), suggestive of a putative Ca²⁺ ion. Well-resolved local density in RyR1-ACP/Ca²⁺_A-inactivated class 1 map allowed tentative modeling of two waters on either side of the putative Ca²⁺ that connect on either side to ACP’s γ-phosphate and Thr4979 (CTD) (Figure 4A, ‘inactivated’), suggesting potential coordination through Ca²⁺’s first layer of hydration. Analysis of protein-ligand interactions based on the atomic coordinates (Laskowski and Swindells, 2011) showed an increased network of electrostatic and hydrophobic interactions, and three predicted hydrogen bonds, that the inactivated conformation gained with respect to closed conformation (Figure 4A). Thus, the nucleotide nestled deeper into the cavity in going from closed to open, and then to inactivated, increasing connectivity between S6C′ and CTD and reducing connectivity to the U-motif (Figure 4B). Interestingly, ACP in RyR1-ACP/Ca²⁺_A inactivated acquired an interaction with the backbone carbonyl of His4983, a residue that participates in the C2H2 zinc motif that is central to the CTD (Figure 4). Based on the higher map significance of the putative Ca²⁺ ion in our cryo-EM maps obtained at higher Ca²⁺ concentration (Figure 4—figure supplement 1) and the enhanced inactivation when ATP is present (Sitsapesan and Williams, 2000), the interface between the nucleotide, the CTD, and S6 could be thought of as an atypical low-affinity Ca²⁺ inactivation site. The picture is bound to be more complex as presumably Mg²⁺ present in the cytoplasm would also bind to ATP. Thus, under physiological conditions of high local Ca²⁺, competition between the two divalent cations for ATP may take place.

To provide a more physiological environment to the TMD and avoid the presence of detergents, we embedded the protein in scaffold protein MSP1E3D1 that assembled into nanodiscs. The reconstituting lipid was phosphatidylcholine. Refinement focused on the CD-TMD region of RyR1-ACP/Ca²⁺_A inactivated and open, resulting in reconstructions with 3.5 Å and 4.4 Å resolution, respectively, yielded a visible electronfor the nanodisc. Two molecules of the MSP1D1E3 scaffold protein wrapped closely around the TMD in a double-belt arrangement (Figure 5A). The larger top belt adopts a quatrefoil shape, with one voltage sensor (S1–S4) in each lobe, while the lower belt is rounder and smaller, following the tapered TMD. The larger footprint of the top half of RyR1’s TMD, reflected in the surrounding nanodisc, appears to correlate with the curvature of the membrane around RyR1 observed by electron tomography in its native membrane (Chen and Kudryashev, 2020; Renken et al., 2009). Even considering the tight fit between the nanodisc and RyR1’s TMD, conformational changes were unhindered and the channel opened within the nanodisc environment. The slight expansion of the top belt of the nanodisc at the level of the ion gate upon opening (Figure 5—video 1) reveals a certain degree of plasticity of the scaffold protein.

We resolved two lipid densities at the four inter-subunit interfaces in the RyR1-ACP/Ca²⁺_A-inactivated CD-TMD-focused map (Figure 5B). The two lipid densities, observed up to 8σ map significance, spanned ~20 Å across the inner leaflet of the SR membrane, at a hydrophobic pocket in the domain-swapped inter-subunit space formed by the S1–S4 bundle and core TMD helices (S5 and S6) (Figure 5B and C). One density (lipid 1), encompassing two fatty acyl tail moieties of 16 and 11 carbons with a polar head compatible with a phosphatidylcholine molecule, was resolved near S5/S6 (Figure 5C). The lipid likely originates from a longer unsaturated PC (16:0–18:1) used in 50-fold molar excess while embedding RyR1 into nanodiscs. Lipid 2, with a resolved 16-carbon fatty acid tail, was sandwiched between lipid 1 and S3/S4. The lipid’s fatty-acyl tails interact extensively with 21 amino acids in the lipophilic pocket formed by S3/S4 and S5/S6 of neighboring subunits, while the density corresponding to the lipid head moiety was positioned within 5 Å from Tyr4629, Asn4857, and Tyr4909 (Figure 5C). Together, the lipids covered ~622 Ų of the 1990 Ų surface area of the S3/S4-S5/S6 subunit interface forming a hydrophobic crevice (Figure 5D). No lipid density was resolved in the RyR1-ACP/Ca²⁺_A open map: besides the lower resolution of this map, its crevice is narrower, as S3 and S4 remodeled the lipid-binding pocket by tilting ~4.4° and ~3.9°, respectively. Furthermore, reorientation in Phe4808 (S4) and Tyr4912 (S6) as modeled for the open state would result in steric clash with lipid 2 and lipid 1, respectively (Figure 5—figure supplement 1A and B; lipids superimposed from the RyR1-ACP/Ca²⁺-inactivated reconstruction – shown in Figure 5—figure supplement 1C in the same orientation), rearranging or even excluding the lipids in the open channel. A similar observation was reported for the TRPV3 channel, whereby lipids get squeezed out in going from the closed to the open state (Singh et al., 2018). We did not resolve lipids within the crevice of RyR1-ACP/EGTA, probably owing to the lower resolution of this reconstruction (4.30 Å), but examination of the crevice indicates no expected hindrance to lipid entry offered by residues Phe4808 and Tyr4912 in the closed state (Figure 5—figure supplement 1D). Moreover, we resolved lipids in a higher-resolution (3.27 Å) 3D reconstruction of closed RyR reconstituted in lipids, in this case for the RyR2 isoform (Figure 5—figure supplement 1E; PDB ID: 6WOU; Iyer et al., 2020). The analogous appearance of the lipid densities in the two isoforms is remarkable (three tubular densities with similar lengths and orientations; compare Figure 5—figure supplement 1C and E), suggesting that the interaction of lipids with the RyR TMD crevice is conserved among isoforms, and characteristic of closed and inactivated states.

The TMD helices were examined in the membrane-like environment provided by the nanodisc. Significant perturbations in the α-helical configuration were observed. S4 displays near-3₁₀ helix in residues 4807–4813 and (only for the open structure) near-π-helix in residues 4814–4819. S5 displays π-helix in residues 4856–4859. S6 displays π-helix in residues 4921–4928 of S6 (Figure 6A and B). In S4, comparison of the three conformations suggests that the degree of over-coiling of the N′ half of S4 is balanced by the uncoiling of its C′ half to near-π-helix in the open structure (Figure 6A), which causes the reorientation of Phe4808 into the hydrophobic crevice. In S6, the π-helix region introduces non-canonical i ← i + 5 hydrogen bonding that is energetically less stable than a regular α-helix (Fodje and Al-Karadaghi, 2002; Kumar and Bansal, 2015; Figure 6B), and may facilitate the flexibility of S6 required for its structural transitions.

The three segments departing from α-helical configuration contain Phe clusters (FFF 4807–4809, FF 4858–4859, FFFF 4920–4923) conserved in all RyR isoforms and IP₃R intracellular Ca²⁺ release channels; with some of them lining the hydrophobic crevice and forming contacts with the lipids (Figure 5C). Deletion of FF 4923–4924 (human sequence; equivalent to rabbit FF 4922–4923) in S6 of RyR1, which is associated to fetal akinesia deformation syndrome or FADS, resulted in loss of Ca²⁺ conductance (Xu et al., 2020), which further supports a role for the Phe clusters in the stability of the pore.

The SR luminal loop (4860–4878) between S5 and the P-loop helix (also termed pore helix; residues 4879–4893), proposed to act as a luminal Ca²⁺ sensor (Sitsapesan and Williams, 1995) and as an anchorage point for proteins within the SR lumen (Beard et al., 2009), contains six acidic residues (EDEDEPD 4867–4873). This luminal loop is separated from the rest of the TMD, and its acidic residues point to the lumen interior. As our experimental high Ca²⁺ concentrations correspond to these in the SR lumen, this conformation of the luminal loop (Figure 6C) is probably a close representation of the native state. The higher-resolution RyR1-ACP/Ca²⁺-inactivated map revealed a disulfide bond (Cys4876-Cys4882) between S5 and the P-loop (Figure 6C), and two inter-subunit hydrogen bonds (Tyr4888-Arg4913 and Tyr4888-Asp4917) between the P-loop and S6. These interactions may stabilize the luminal mouth of the channel, keeping the proper arrangement of the four SR luminal loops, as well as holding S6 through its conformational transitions.

All chemicals were purchased from Thermo Fisher or Sigma-Aldrich except where indicated.

RyR1 activity was estimated by measuring the extent of bound ³[H]ryanodine in microsomes isolated from rabbit skeletal muscle when incubated with free Ca²⁺ alone (10 μM to 2 mM range), or in the presence of 2 mM ATP or 2 mM AMP-PCP (ACP) sodium salts. Concentrations of total Ca²⁺ added to the reaction mixture were estimated in Maxchelator (https://somapp.ucdmc.ucdavis.edu/pharmacology/bers/maxchelator). Preincubated membrane vesicles (~40 μg) were allowed to bind 5 nM [³H]ryanodine (PerkinElmer) in a buffer containing 50 mM MOPS (pH 7.4), 0.15 M KCl, 0.3 mM EGTA, protease inhibitors, and 2 mM DTT for 3 hr at 37°C. Sample aliquots were diluted sevenfold with an ice-cold wash buffer (0.1 M KCl) before placing onto Whatman GF/B filter papers in a vacuum-operated filtration apparatus. The remaining radioactivity in the filter papers after washing three times with the wash buffer was measured by liquid scintillation counting. Nonspecific ryanodine binding was estimated in the presence of 250 μM unlabeled ryanodine (Calbiochem) and subtracted from the total binding. Data represent the mean specific [³H]ryanodine binding from four independent experiments.

Microsomes were purified from rabbit back and hind leg muscles through differential centrifugation as previously described (Hu et al., 2021; Samsó et al., 2009). 100 mg frozen membranes were thawed and solubilized in buffer A containing 20 mM MOPS pH 7.4, 1 M NaCl, 9.2% (w/v) CHAPS, 2.3% (w/v) phosphatdylcholine (PC; Sigma), 2 mM DTT, and protease inhibitor cocktail for 15 min at 4°C. The solubilized membranes were centrifuged at 100,000 × g for 60 min and the pellet was discarded. Supernatant was layered onto 10–20% (w/v) discontinuous sucrose gradients, prepared in buffer B (buffer A plus 0.5% CHAPS and 0.125% PC). The layered sucrose gradient tubes were ultracentrifuged at 120,000 × g for ~20 hr at 4°C to allow RyR1 separation. Fractions containing >95% pure RyR1 were pooled and further purified with a HiTrap Heparin HP Agarose column (GE Healthcare) after a fivefold dilution in salt-free buffer and filtration steps. RyR1 was eluted with buffer B containing 0.9 M NaCl, after washing with 20 column volumes of buffer B with 200 mM NaCl. Peak fractions were flash-frozen and stored at –80°C until reconstitution into nanodiscs and cryo-EM. 1.5–2 mg of RyR1 was purified from 100 mg of SR membrane vesicles. RyR1 purity was estimated with 12.5% SDS-PAGE and negative staining with 0.75% uranyl formate. Protein concentration in purified microsomes and RyR1 fractions was measured with Quick Start Bradford Protein Assay (Bio-Rad). The plasmid encoding for MSP1E3D1, pMSP1E3D1, was purchased from Addgene, and recombinant MSP1E3D1 was purified in Escherichia coli using the manufacturer’s instructions. RyR1-nanodiscs were obtained by mixing purified RyR1, MSP1E3D1, and POPC (Avanti polar lipids) at a 1:2:50 molar ratio. The mixture was incubated for 1 hr 30 min at 4°C before an overnight dialysis in a CHAPS-free buffer (20 mM MOPS pH 7.4, 635 mM KCl, 2 mM DTT), which contained either 1 mM EGTA +1 mM EDTA for the control ‘RyR1-ACP/EGTA’ dataset or 3.7 mM CaCl₂ for the RyR1-ACP/Ca²⁺ dataset. The dialyzed RyR1-nanodisc preparations were incubated with ACP (sodium salt) for 30 min prior to plunge freezing, at concentrations of 5 mM ACP (control RyR1-ACP/EGTA dataset) or 2 mM ACP (RyR1-ACP/Ca²⁺ datasets). Free Ca²⁺ was estimated with Maxchelator. Integrity of the nanodisc-embedded channels was examined by negative staining.

Cryo-EM grids were cleaned with a customized protocol (Passmore and Russo, 2016) prior to glow discharge. Aliquots of 1.25–1.5 µl RyR1-nanodisc were applied onto each side of glow-discharged 300 mesh UltraAufoil –1.2/1.3 Holey-Gold (Quantifoil, Germany). The grids were blotted for 1–1.5 s with an ashless Whatman Grade 540 filter paper in a Vitrobot Mark IV (Thermo Fisher Scientific) and rapidly plunged into liquid ethane. Grid quality and RyR1 sample distribution were assessed on a Tecnai F20 (Thermo Fisher Scientific) electron microscope. Data acquisition was carried out in a Titan Krios transmission electron microscope (Thermo Fisher Scientific) operated at 300 kV and counting mode, with a K3 or K2 detector (Gatan) for the ACP/Ca²⁺_A and ACP/Ca²⁺_B datasets, respectively. A Gatan Quantum Energy Filter (GIF) with a slit width of 20 eV was employed. The ACP/EGTA dataset was collected on a K2 detector and a 20 eV GIF. Datasets were collected in automated mode with the program Latitude (Gatan) with a cumulative electron dose of 70 e^-/Ų applied over 50–60 frames. Image acquisition parameters for RyR1-ACP/EGTA, RyR1-ACP/Ca²⁺_A, and RyR2-ACP/Ca²⁺_B datasets are summarized in Table 1.

Gain reference normalization, movie frame alignment, dose weighting, and motion correction of the collected movie stacks were carried out with Motioncor2 (Zheng et al., 2017). Contrast transfer function parameters were estimated from non-dose-weighted motion-corrected images using Gctf (Zhang, 2016). All subsequent image processing operations were carried out using dose-weighted, motion-corrected micrographs in RELION 3.0 (Scheres, 2012). The micrographs were low-pass filtered to 20 Å before automated particle picking. 2D class average templates for autopicking were generated by reference-free 2D classification of 1000 manually picked particles. Autopicked particles with ethane and hexagonal ice-contaminated areas were removed by visual inspection. Particle image sub-stacks required for the focused reconstructions residues 3668–5037, encompassing the CD, U-motif, TMD, and CTDs, were generated using a signal subtraction procedure employed in relion_project module (Bai et al., 2015). Particle image stacks of quarter sub-volumes of RyR1 corresponding to a single subunit were generated by particle subtraction following a symmetry expansion step with relion_particle_symmetry_expand tool (Bai et al., 2015; Scheres, 2016). Composite tetrameric maps were generated from the symmetry-expanded monomeric maps with Chimera (Pettersen et al., 2004) vop maximum tool. B-factor applied to the reconstructed maps was estimated with relion_postprocess. Analysis of the TMD was carried out on a focused map of the CD-TMD region. Unfiltered half maps obtained from the final 3D-refinement step in RELION 3.0 were further density modified with PHENIX.Resolve (Terwilliger et al., 2020). The reported resolutions of the cryo-EM maps are based on FSC 0.143 criterion (Scheres and Chen, 2012). Local resolution was estimated with ResMap (Kucukelbir et al., 2014). Pixel size calibration of postprocessed maps was carried out using real space correlation metric of UCSF Chimera based on a published RyR1 cryo-EM map (des Georges et al., 2016). Pixel size maxima of 1.07, 1.105, and 1.07 Å were obtained in RyR1-ACP/EGTA, RyR1-ACP/Ca²⁺_A, and RyR1-ACP/Ca²⁺_B respectively. Image processing schemes of the RyR1-ACP/Ca²⁺_A, RyR1-ACP/Ca²⁺_B, and RyR1-ACP/EGTA datasets are summarized in Figure 1—figure supplement 1, Figure 1—figure supplement 3, and Figure 1—figure supplement 4, respectively.

The cryo-EM-based atomic models of RyR1 (PDB ID: 5tb3 for RyR1-ACP/EGTA, RyR1-ACP/Ca²⁺ inactivated and 5ta3 for RyR1-ACP/Ca²⁺ open) were used as the initial models for model building. The best resolved symmetry-expanded cryo-EM map for a single subunit in inactivated or open conformation was docked within a RyR1 monomer model with Chimera Fit in map tool. Local density fit of the RyR1 sequence was improved over an iterative process of amino acid fitting in Coot (Emsley et al., 2010) alternated with real space refinement in PHENIX (Afonine et al., 2018). Four copies of the monomers were docked to the whole RyR1 reconstructions. Real space refinement of the tetrameric models was carried out with secondary structure and Ramachandran restraints. Further manual fitting of the CD, TMD, and CTD (3668–5037) of RyR1 was carried out in Coot. Comprehensive model validation was carried out with PHENIX and PDB validation server at https://validate-rcsb-2.wwpdb.org/ and is summarized in Table 2. Molecular lipophilicity potential surfaces were drawn in ChimeraX (Ghose et al., 1998; Laguerre et al., 1997; Pettersen et al., 2021). Figures were generated with PyMOL (Schrodinger, 2015) and Chimera programs (Pettersen et al., 2004; Pettersen et al., 2021).

Pore radii were measured for the refined atomic model coordinates of the RyR1 pore region (residues 4821–5037) with the HOLE program (Smart et al., 1993). Dot surfaces representing the channel ion permeation pathway were generated with HOLE implemented in Coot, which were reformatted to enable visualization in UCSF Chimera. Residues per turn of S4 and S6 transmembrane helices of RyR1 were calculated with HELANAL (Bansal et al., 2000).

The cytoplasmic shell flexion angles of RyR1 in different conformations were estimated using the procedure indicated in Steele and Samsó, 2019 with minor changes. The angle was calculated between a diagonal running from the N-terminal domain (residue 348) to the P1 domain (residue 984) from the same subunit, and the horizontal plane.

Tetrameric and focused cryo-EM maps of RyR1-ACP/EGTA, RyR1-ACP/Ca²⁺ inactivated, and RyR1-ACP/Ca²⁺ open have been deposited in the Electron Microscopy Databank (EMDB) with the following accession codes: 22616, 22597 (RyR1-ACP/EGTA tetrameric and focused maps), 25828 (RyR1-ACP/Ca²⁺_A inactivated; focused map as an additional map), 25830, 25831, 25832 (three subclasses of RyR1-ACP/Ca²⁺_A inactivated), 25829 (RyR1-ACP/Ca²⁺_A open), and 25833 (RyR1-ACP/Ca²⁺_B inactivated). Atomic models generated from the cryo-EM maps have been deposited in the RCSB PDB database with the following accession codes: 7K0T (RyR1-ACP/EGTA), 7TDG (RyR1-ACP/Ca²⁺_A inactivated), 7TDJ, 7TDI, 7TDK (three subclasses of RyR1-ACP/Ca²⁺_A inactivated), and 7TDH (RyR1-ACP/Ca²⁺_A open).