Main

Cardiac muscle contraction is triggered by Ca2+ flux into the cytosol, initially from the extracellular environment, mediated by Cav1.2, and subsequently from the sarcoplasmic reticulum Ca2+ store, mediated by RyR21,2,3. Ryanodine receptors are the largest known ion channels and consist of a homotetramer with a molecular mass of more than 2 megadaltons. More than 80% of the protein folds into a multi-domain cytoplasmic assembly that senses interactions with a variety of modulators, which range from ions to proteins4,5,6. The precise regulation of RyR2 activity is critical for each heartbeat. Aberrant activity of RyR2 is associated with life-threatening cardiac arrhythmias7,8,9,10.

The 17-kDa protein CaM is an essential calcium sensor that has a central role in most calcium signalling events11. CaM consists of roughly symmetrical N- and C-terminal lobes (N- and C-lobes hereafter), joined by a flexible hinge12,13. Each lobe can cooperatively bind to two Ca2+ ions, with a micromolar-range binding affinity, via two EF-hand (helices E and F-hand) motifs. Upon Ca2+ binding, the exposure of several hydrophobic residues in both lobes facilitates CaM binding to the target sequence. CaM interacts directly with ryanodine receptors with a 1:1 stoichiometry of the CaM–RyR protomers14,15 and binding affinity at nanomolar range14.

Regulation of ryanodine receptors by CaM, however, is isoform-specific. CaM shows biphasic regulation of RyR1, acting as a weak activator at nanomolar levels of Ca2+ (apo-CaM) and an inhibitor at micromolar levels of Ca2+ (Ca2+-CaM)14,16. By contrast, apo-CaM has no effect17 or an inhibitory effect on RyR214, whereas Ca2+-CaM inhibits RyR214. CaM has also been shown to facilitate the termination of store-overload-induced Ca2 + release (SOICR)18. Aberrant interactions between CaM and RyR2 are associated with heart failure19,20,21,22, and correction of impaired CaM–RyR2 interactions may serve as a therapy for lethal arrhythmia in pressure-overload-induced heart failure23.

Structural characterization of RyR–CaM complexes has been limited to low-resolution electron microscopy maps that suggest two overlapping, but distinct, binding sites in RyR1 for apo- and Ca2+-CaM24,25,26. A peptide that corresponds to residues 3614–3643 of RyR1 (residues 3581–3612 in the central domain of RyR2) binds to both apo- and Ca2+-CaM15,27. The crystal structure of Ca2+-CaM bound to the peptide revealed hydrophobic anchors in the N and C termini of the peptide that accommodate the C- and N-lobes of Ca2+-CaM, respectively28.

To elucidate the modulation of RyR2 by CaM, we report eight cryo-electron microscopy (cryo-EM) structures of RyR2 that collectively reveal molecular recognition characteristics for different forms of CaM and provide insights into the regulation of RyR2 channel gating by CaM.

Structures of RyR2 under eight conditions

To achieve a better understanding of RyR2 modulation by CaM (Extended Data Fig. 1a, b), we determined the cryo-EM structures of the porcine RyR2 (Extended Data Fig. 1c) under the following eight conditions.

Condition (1) consisted of RyR2 bound to FKBP12.6 and apo-CaM (hereafter FKBP12.6/apo-CaM) and was used to assess the apo-CaM binding site. Condition (2) consisted of RyR2 bound to FKBP12.6 and a Ca2+-binding-deficient CaM mutant that mimics apo-CaM24,29 (CaM-M) in the presence of ATP, caffeine and low [Ca2+] (hereafter FKBP12.6/ATP/caffeine/low-[Ca2+]/CaM-M), this structure was used to investigate the mechanism for the binding-location switch of CaM. Condition (3) consisted of RyR2 bound to FKBP12.6 in the presence of ATP, caffeine and low [Ca2+] (FKBP12.6/ATP/caffeine/low-[Ca2+]), the presence of which maximizes the open state. Condition (4) consisted of RyR2 bound to FKBP12.6 and Ca2+-CaM in the presence of ATP, caffeine and low [Ca2+] (hereafter FKBP12.6/ATP/caffeine/low-[Ca2+]/Ca2+-CaM); this condition was used to examine the effect of Ca2+-CaM on the open RyR2 channel in the presence of FKBP12.6, ATP, caffeine and low [Ca2+]. Conditions (5) and (6) corresponded to conditions (3) and (4), respectively, but were treated with CHAPS and DOPC instead of digitonin. Condition (7) consisted of RyR2 in high [Ca2+] in the presence of FKBP12.6, ATP, caffeine and Ca2+-CaM (hereafter FKBP12.6/ATP/caffeine/high-[Ca2+]/Ca2+-CaM); this environment was used to achieve a better resolution for Ca2+-CaM. Condition (8) consisted of RyR2 bound to Ca2+-CaM in the presence of 2,2′,3,5′,6-pentachlorobiphenyl (PCB95) and low [Ca2+] (hereafter PCB95/low-[Ca2+]/Ca2+-CaM); this condition was used to investigate the effect of Ca2+-CaM on PCB95 and the Ca2+-activated RyR2 channel30. The eight conditions and corresponding structures are summarized in Supplementary Table 1 and Extended Data Table 1, respectively.

All cryo-EM datasets were processed following the same procedure (Extended Data Fig. 2). The FKBP12.6/apo-CaM RyR2 structure was determined at an overall resolution of 3.6 Å, the highest among all of the available RyR2 structures (Fig. 1a and Extended Data Figs. 1i, 3a, 4a–i). The secondary structural elements of apo-CaM were clearly resolved (Fig. 1a and Extended Data Fig. 3b). The FKBP12.6/ATP/caffeine/low-[Ca2+]/CaM-M RyR2 structure was determined at an overall resolution of 4.2 Å, in which the well-resolved CaM-M is positioned similarly to apo-CaM on RyR2 (Fig. 1b and Extended Data Figs. 1i, 3c, d). The densities for both lobes of Ca2+-CaM are visible in the FKBP12.6/ATP/caffeine/high-[Ca2+]/Ca2+-CaM RyR2 structure (3.9 Å resolution) and PCB95/low-[Ca2+]/Ca2+-CaM RyR2 structure (4.4 Å resolution) in which the N-lobe is better-resolved than the C-lobe. By contrast, only one lobe of Ca2+-CaM—the N lobe as judged from the comparison with the structure of FKBP12.6/ATP/caffeine/high-[Ca2+]/Ca2+-CaM at 3.9 Å—is discernible in the FKBP12.6/ATP/caffeine/low-[Ca2+]/Ca2+-CaM structure with a resolution of 4.2 Å (Fig. 1c, d and Extended Data Figs. 1, 3).

Fig. 1: Cryo-EM structures of the RyR2–CaM complexes.
figure 1

a–d, Overall electron microscopy maps for four indicated complexes. a, RyR2 in the presence of FKBP12.6/apo-CaM (3.6 Å). b, RyR2 in the presence of FKBP12.6/ATP/caffeine/low [Ca2+]/CaM-M (4.2 Å). c, RyR2 in the presence of FKBP12.6/ATP/caffeine/high [Ca2+]/Ca2+-CaM (3.9 Å). d, RyR2 in the presence of PCB95/low-[Ca2+]/Ca2+-CaM (4.4 Å). Insets, cytoplasmic views of the channel gates. The dashed circles indicate the distances between the Cα atoms of the gating residues in the diagonal protomers. See Extended Data Table 1 for details of the structures. SR, sarcoplasmic reticulum. e, The pore of RyR2 remains closed in the FKBP12.6/apo-CaM structure. The ion permeation path, calculated by HOLE41, is illustrated as yellow dots. SF, selectivity filter. f, g, Open pores of RyR2 in FKBP12.6/ATP/caffeine/low-[Ca2+]/CaM-M (f) and FKBP12.6/ATP/caffeine/high-[Ca2+]/Ca2+-CaM (g). h, The corresponding pore radii of RyR2 for the three structures shown in e–g. Electron microscopy maps were generated in Chimera and contoured at levels of 0.027, 0.022, 0.02 and 0.015 for a–d, respectively. All structures were prepared using PyMOL (http://www.pymol.org).

Full size image

With regard to the gating state of the eight structures reported here, the clearly resolved Ile4868 residues on the S6 helical bundle of FKBP12.6/apo-CaM constitute the constriction site with a radius of approximately 1 Å, which is identical to the previously reported apo-RyR2 in the closed state30 (Fig. 1e, h). The constriction site appears to shift to Gln4864 with an expanded radius of around 3 Å in conditions (2), (3) and (5)–(7) (Fig. 1f–h and Extended Data Fig. 1d, h), similar to that in the open PCB95/low-[Ca2+] structure30. However, the density for the side chain of Gln4864 in FKBP12.6/ATP/caffeine/low-[Ca2+]/Ca2+-CaM is not well-resolved. We, therefore, compared the distances of Cα atoms of the gating residues in the diagonal protomers, which are approximately 16 Å for FKBP12.6/ATP/caffeine/low-[Ca2+]/Ca2+-CaM and 11 Å for FKBP12.6/apo-CaM (Fig. 1a and Extended Data Fig. 1e). The constriction site in PCB95/low-[Ca2+]/Ca2+-CaM is constituted by Ile4868, for which the diagonal distance of the Cα atoms is approximately 11 Å—similar to that in FKBP12.6/apo-CaM (Fig. 1a, d).

As seen in RyR1, the Ca2+-, ATP- and caffeine-binding sites are located at the interfaces between the central and channel domains of RyR231 (Extended Data Fig. 4j–m).

Location of apo-CaM in RyR2

Consistent with the low-resolution structure of RyR1, apo-CaM is located in an elongated cleft formed by the handle, helical and central domains of RyR2 in FKBP12.6/apo-CaM24,25 (Fig. 2a). The N-lobe is stuck in the upper half of the cleft formed by helical domain 1 (HD1), whereas the C-lobe is located at the bottom edge of the cleft surrounded by the handle and central domains of RyR2 (Fig. 2a).

Fig. 2: Interfaces between CaM and RyR2.
figure 2

a, Apo-CaM is located in a cleft formed by the handle, helical and central domains of RyR2. One RyR2 protomer is shown in domain-coloured surface view. b, c, Multiple interfaces between RyR2 and apo-CaM. The red, dashed boxes indicate the interfaces. d, A previously unresolved helix α−1 on RyR2 serves as the primary docking site for both lobes of Ca2+-CaM. e, Functional validation of the observed interfaces between mouse RyR2 and human CaM. Open probabilities of single RyR2 channels before (control) and after addition of CaM(WT) (1 μM). Data are mean ± s.e.m. from RyR2(WT) (n = 9), RyR2(Y2156A) (n = 8), RyR2(V3599A) (n = 7), RyR2(W3587A) (n = 8) and RyR2(L3590A) (n = 9 single channels) and analysed by paired, two-sided Student’s t-test (versus its own control) with P values shown in blue and by one-way analysis of variance (ANOVA) with a Dunnett’s post hoc test (versus RyR2(WT) control and RyR2(WT) with CaM(WT), respectively) with adjusted P values shown in black. f, The termination threshold of Ca2+ release in Ryr2WT- and Ryr2-mutant-expressing HEK293 cells transfected with no CaM (control), CaM(WT) or CaM-M. Data are mean ± s.e.m. with the number of independent experiments for each condition shown and analysed by one-way ANOVA with a Dunnett’s post hoc test with adjusted P values shown in blue (versus its own control) and in black (versus RyR2(WT) control).

Source Data

Full size image

The FKBP12.6/apo-CaM structure reveals five surface patches on RyR2 that interact with apo-CaM. The N-lobe interacts with RyR2 through three interfaces that are mainly located in HD1 (Fig. 2b and Extended Data Fig. 5a). The most prominent interface is formed between the N terminus of helix 4 (N4) in the N-lobe and the C termini of helices 2b and α1 in HD1, and is mainly mediated by extensive hydrophobic residues. Phe66, Pro67 and Leu70 on N4 probably interact with Tyr2203 in helix 2b and Tyr2157 (human Tyr2156) in helix α1. Ile10 in N1 may also interact with Tyr2157 (Extended Data Fig. 5a, g). The human Y2156C variant is linked to catecholaminergic polymorphic ventricular tachycardia8. The second interface is mediated by charged residues between the N terminus of N1 in the N-lobe and helix α1 in HD1 and the N terminus of helix α0 in the central domain. The third interface is formed between a region rich in acidic residues in N3 of the N-lobe and Lys2558 in helix 8b of HD1 (Fig. 2b and Extended Data Fig. 5a).

The C-lobe interacts with RyR2 through two interfaces (Fig. 2c). One is consistent with previous reports15,27. Residues 3593–3607 in the central domain folds into a newly resolved helix α ‘minus 1’ (helix α−1) that is enclosed by the hydrophobic cavity of the C-lobe, representing the primary interface. Phe3604 serves as a hydrophobic anchor for the hydrophobic cavity of the C-lobe. A minor interface is formed between helix 12 and the C terminus of helix 11 in the handle domain with the C terminus of C1 and the loop between C2 and C3 in the C-lobe, also through hydrophobic interactions (Fig. 2c and Extended Data Fig. 5b, c, h).

Shift of CaM-binding site in RyR2 after Ca2+ loading

CaM markedly slips down along the cleft after Ca2+ loading, making extensive interactions with the central domain (Fig. 2d). The N-lobe is anchored by the central domain and the C-lobe drops beyond the cleft, coordinated only by helix α−1 (Fig. 2d). The limited contact may explain the structural flexibility of the C-lobe.

The binding of Ca2+-CaM with intact RyR2 is similar to that of Ca2+-CaM with the RyR1 peptide28. The N- and C-lobes of CaM interact with the C- and N termini of helix α−1, respectively (Fig. 2d). Phe3604 and Trp3588 anchor the hydrophobic cavities of the N- and C-lobes, respectively (Extended Data Fig. 5d–f, i). An additional interface is formed between the N terminus of N3 and the C terminus of helix α9 in the central domain to further stabilize the binding of N-lobe. Asp51 in N3 probably interacts with Arg2209 in helix 2b of HD1 (Fig. 2d and Extended Data Fig. 5e).

Functional validation of RyR2–CaM interfaces

We assessed the effect of mutations in RyR2 and CaM that are located in the structurally revealed interfaces on CaM regulation of RyR2. Wild-type CaM (CaM(WT)) strongly reduced the open probability of single wild-type mouse RyR2 (RyR2(WT)) channels. The following mutations markedly reduced CaM inhibition of single RyR2 channels: RyR2(Y2156A) (porcine Y2157A) near an interface between the apo-CaM N-lobe and RyR2, RyR2(V3599A) near both the apo-CaM C-lobe–RyR2 and Ca2+-CaM N-lobe–RyR2 interfaces, and RyR2(W3587A) and RyR2(L3590A) near the Ca2+-CaM C-lobe–RyR2 interface (Fig. 2e and Extended Data Fig. 6).

We next examined the effect of these RyR2 mutations on the termination of RyR2-mediated SOICR. As previously shown18, CaM(WT) increased Ca2+ release termination in RyR2(WT)-expressing HEK293 cells, whereas CaM-M reduced Ca2+ release termination (that is, a longer calcium release) (Fig. 2f and Extended Data Fig. 7a–c). Consistent with their effect on single RyR2 channels, all four mutations in RyR2 significantly reduced Ca2+ release termination in HEK293 cells, probably by impairing the effect of endogenous CaM on RyR2 inhibition (Fig. 2f). These RyR2 mutations also reduced or abolished the effect of exogenously expressed CaM(WT) and CaM-M on Ca2+ release termination (Fig. 2f), but had little or no effect on SOICR activation or store capacity (Extended Data Fig. 7d, e).

We also assessed the functional importance of CaM residues near the RyR2–CaM interfaces. Mutations in CaM near the apo-CaM N-lobe–RyR2(K2153/Y2156) interface (CaM(E15A), CaM(F66A) and CaM(L70A)), near the apo-CaM C-lobe–RyR2(V3599) interface (CaM(M110A) and CaM(F142A)), near the Ca2+-CaM N-lobe–RyR2(V3599) interface (CaM(F20A) and CaM(F69A)) and near the Ca2+-CaM C-lobe–RyR2(W3587/L3590) interface (CaM(F93A), CaM(L106A) and CaM(M146A))—as with CaM-M—significantly reduced the effect of CaM on Ca2+ release termination compared to CaM(WT) (Extended Data Fig. 7f). All CaM mutations except for CaM(F20A) and CaM(F142A) had little or no effect on SOICR activation or store capacity (Extended Data Fig. 7g, h). Note that some mutations in CaM may induce conformational changes, thus affecting CaM–RyR2 interactions allosterically. Collectively, these functional studies support the importance of the newly identified RyR2–CaM interfaces in CaM regulation of RyR2.

Ca2+-dependent shift in CaM-binding sites on RyR2

The location and conformation of CaM-M and apo-CaM are identical in the structures (Fig. 3a), which suggests that the positional switch for apo-CaM and Ca2+-CaM results from the distinct conformations of CaM after Ca2+ loading instead of a direct effect of Ca2+ on RyR2. The CaM lobes have previously been reported in three conformations—open, semi-open and closed—when interacting with other proteins32,33,34. Comparative docking analysis of these conformations into the electron microscopy reconstruction for FKBP12.6/ATP/caffeine/high-[Ca2+]/Ca2+-CaM suggests an open conformation for the N-lobe (Extended Data Fig. 8a, b). An important distinction between the open and semi-open or closed C-lobes is the slightly larger helical angle between C1 and C4, both of which were resolved in the map and conformed to the open state (Extended Data Fig. 8c, d). These analyses are consistent with the observations in the crystal structure of RyR1 peptide-bound Ca2+-CaM28. Upon Ca2+ loading, the compact structure of apo-CaM is relaxed to Ca2+-CaM35, which slips down towards the N-terminal part of helix α−1, consistent with previous studies36.

Fig. 3: Molecular basis for the shift of binding site for CaM after Ca2+ loading.
figure 3

a, Ca2+ binding to RyR2 is not responsible for the positional shift in apo-CaM and Ca2+-CaM. In the presence of 20 μM Ca2+, the binding site for CaM-M remains the same as for apo-CaM. The two structures are superimposed relative to CaM. b, Upon Ca2+ loading, the expansion of CaM structure may lead to steric hindrance between the N-lobe and HD1. The N-lobes are superimposed. Red arrows indicate directions of conformational changes from apo-CaM to Ca2+-CaM. c, The shift in the CaM-binding site is accompanied by marked conformational changes in helices α−1 and 12 in RyR2. Red arrows indicate the conformational changes in RyR2 from FKBP12.6/apo-CaM to FKBP12.6/ATP/caffeine/high-[Ca2+]/Ca2+-CaM. d, In FKBP12.6/apo-CaM, helix α−1 contacts helix α9. e, In FKBP12.6/ATP/caffeine/high-[Ca2+]/Ca2+-CaM, helix α−1 of RyR2 is positioned away from helix α9. Helix N3 in the N-lobe contacts helix α9 instead. f, Helix α−1 of RyR2 serves as an essential anchor for CaM. Red arrow indicates the direction of conformational change of helix α−1 from FKBP12.6/apo-CaM to FKBP12.6/ATP/caffeine/high-[Ca2+]/Ca2+-CaM.

Full size image

Helix α−1 and helix 12 are two helices that were only resolved in the structures with CaM, probably owing to stabilization of these segments through CaM binding. Pronounced shifts in these two helices are observed between the structures bound to apo-CaM and Ca2+-CaM. In the presence of Ca2+-CaM, helix α−1 swings by around 90° and the N-terminal half of helix 12 also bends by nearly 90° (Fig. 3c). Accordingly, helix α−1, which contacts helix α9 of the central domain in FKBP12.6/apo-CaM, is positioned away from helix α9 in FKBP12.6/ATP/caffeine/high-[Ca2+]/Ca2+-CaM. Now helix α9 is in contact with helix 3 in the N-lobe of Ca2+-CaM (Fig. 3d–f). The CaM-bound RyR2 structures shown here reveal that helix α−1 serves as an essential anchor for CaM (Fig. 3f).

Inhibitory modulation of RyR2 by Ca2+-CaM

Both caffeine and ATP are located at the interfaces between the U-motif and O-ring, locking them into a stable unit that stabilizes the open state (Extended Data Fig. 9a). Caffeine and ATP counteract inhibitory effect of Ca2+-CaM on RyR2, manifested by the lack of intradomain change of individual central domains. The pore remains open (Extended Data Figs. 1e, 9b). Nevertheless, Ca2+-CaM induces an anticlockwise rotation of the central domains in the cytoplasmic view in the same direction as that from the open to the closed state30,37 (Extended Data Fig. 9c and Supplementary Video 1). The central domains undergo similar shifts in the presence of high concentrations of Ca2+ and CaM, but not in the CHAPS plus DOPC condition (Extended Data Fig. 10a, b).

By contrast, RyR2 activated by PCB95 and Ca2+ is closed after addition of Ca2+-CaM (Fig. 1d). Detailed structural examination shows an anticlockwise rotation of the central domains and outward motions of the auxiliary motifs of the individual central domain, including helices α0, α1, α4 and the U-motif, with respect to the centre of the concave surface (Extended Data Figs. 9d, 10c). The motion of the U-motif appears to release the pulling force for the dilation of the S6 helix, resulting in closure of the pore (Extended Data Fig. 9e). Taken together, these results indicate that the inhibitory force of Ca2+-CaM is sufficient to overcome the synergistic activation of RyR2 by PCB95 and Ca2+, but not by the collective effect of ATP, caffeine and Ca2+ (Fig. 4).

Fig. 4: Schematic of RyR2 modulation by CaM.
figure 4

The two RyR2 structures on the left were obtained in the presence of CHAPS and DOPC instead of digitonin, which was used for all other structural determinations. Despite a rotation (indicated by red arrows) of the central domains, the pore of ATP, caffeine and Ca2+-activated RyR2 channel remains open in the presence of Ca2+-CaM under these two conditions (left four). By contrast, Ca2+-CaM leads to closure of PCB95 and Ca2+-activated RyR2 channel (right two). Our study demonstrates that the inhibitory force of Ca2+-CaM can overcome the synergistic activation by PCB95 and Ca2+ but not by ATP, caffeine and Ca2+ (bottom right). The top right structure PCB95/low-Ca2+ was obtained from a previous study (RCSB Protein Data Bank (PDB) code 5GOA).

Full size image

Discussion

Among the reported RyR2 sequences for CaM-binding38, residues 4247–4277 are invisible in our structure. The segment (residues 2023–2039) that contains the protein kinase A phosphorylation site Ser2032 is resolved in our structures, but shows no interaction with CaM even at high concentrations39. The other two sequences (1942–1966 and 3582–3608) bind to CaM in our structures (Extended Data Fig. 10d). Despite the distinct effects of apo-CaM on RyR1 and RyR214, the primary apo-CaM-binding sequences are invariant in these two isoforms (Extended Data Fig. 10e). The molecular determinants for the functional difference are yet to be revealed.

The inhibitory mechanisms by which CaM regulates RyR2 must be investigated on an already opened channel by structural biology. As the presence of micromolar-range Ca2+ is required for opening ryanodine receptors by cryo-EM studies30,31,32,37, it is impractical to obtain an open structure in the absence of micromolar concentrations of Ca2+ (apo-CaM form). Although the location and conformation of CaM-M appear identical to those of apo-CaM, it has previously been reported that modulation of RyR2 by CaM-M is distinct from that by apo-CaM40—although the mechanism needs to be investigated further.

Owing to extensive interactions between the U-motif and O-ring, the two undergo coupled motions during channel gating30,37. The presence of caffeine and ATP locks them into a more rigid structure, probably increasing the energy barrier for inhibiting RyR2 by Ca2+-CaM. By contrast, the PCB95- and Ca2+-activated RyR2 channel can be effectively closed by Ca2+-CaM. Therefore, the gating state of RyR2 is defined by the combined effect of competing stimulatory and inhibitory regulators (Fig. 4). It remains to be investigated whether the conclusions presented here can be recapitulated for other ryanodine receptor isoforms or in lipid bilayers and the relevance to disease-related mutations (Extended Data Fig. 10f, g).

Methods

Expression and purification of GST–FKBP12.6

Because the sequence of porcine FKBP12.6 is not available in the public domain, human FKBP12.6 was applied to pull-down porcine RyR2 (pRyR2)30. The complementary DNA of full-length human FKBP12.6 (also known as FKBP1B) was cloned into the pGEX-4T-2 vector with a C-terminal 6×His tag and an N-terminal glutathione S-transferase (GST) tag. Protein was overexpressed in the Escherichia coli BL21 (DE3) strain at 18 °C for 12–15 h after the addition of 0.2 mM isopropyl-β-d-thiogalactoside (IPTG) to cells with an optical density at 600 nm (OD600) of 1.0. Cells collected by centrifugation were resuspended in lysis buffer (25 mM Tris, pH 8.0, 150 mM NaCl). Cell debris was removed by centrifugation at 22,000g for 1 h, and the supernatant was applied to Ni2+-NTA resin (Qiagen). The resin was washed with both W1 buffer (25 mM Tris, pH 8.0, 500 mM NaCl) and W2 buffer (25 mM Tris, pH 8.0, 20 mM imidazole) and eluted with 25 mM Tris, pH 8.0 and 300 mM imidazole. The elution was further purified by anion-exchange chromatography (SOURCE 15Q, GE Healthcare).

Expression and purification of the wild-type CaM and CaM mutant

In mammals, three independent genes (CALM1–CALM3) with approximately 80% identity42 are transcribed into at least eight mRNAs that encode identical CaM proteins43. It has previously been reported that the first methionine residue of CaM was removed under physiological conditions44. The complementary DNA of human CALM3 without the initial Met was cloned into the pET21 vector with an N-terminal 6×His tag followed by an N-terminal SUMO tag and a stop codon in the C terminus, preventing the translation of a C-terminal 6×His tag in the original pET21 vector (Extended Data Fig. 1a). The expression and purification protocol was similar to that of GST–FKBP12.6 mentioned above. Specifically, the N-terminal 6×His tag and SUMO tag were removed together by the SUMO protease UlP1p45 during purification. The CaM protein was further purified by anion-exchange chromatography (SOURCE 15Q, GE Healthcare) using buffer 1 (25 mM Tris, pH 8.0) and buffer 2 (1 M NaCl, 25 mM Tris, pH 8.0). Finally, the protein was applied to size-exclusion chromatography (SEC; Superdex-200, GE Healthcare) in buffer F (20 mM HEPES, pH 7.4, 200 mM NaCl, 0.1% digitonin, 1.3 μg ml−1 aprotinin, 1 μg ml−1 pepstatin, 5 μg ml−1 leupeptin, 0.2 mM PMSF and 2 mM DTT), which is the same as that used for the last-step purification of RyR2. The N-terminal boundary of wild-type CaM was confirmed by N-terminal sequencing (Extended Data Fig. 1b). The expression and purification of the CaM mutant that is deficient in Ca2+ binding at all four EF-hand Ca2+-binding sites (E32A, E68A, E105A and E141A) (denoted as CaM-M) were the same as for the wild-type CaM.

Preparation of sarcoplasmic reticulum membranes from porcine heart

The procedures for preparing the membranes of the sarcoplasmic reticulum from porcine hearts were similar to previously described procedures30. A single porcine heart was cut into small pieces and then resuspended in five volumes of homogenization buffer A (20 mM HEPES, pH 7.4, 150 mM NaCl, 5 mM EDTA, 1.3 μg ml−1 aprotinin, 1 μg ml−1 pepstatin, 5 μg ml−1 leupeptin and 0.2 mM PMSF). Homogenization was performed in a blender (JYL-C010, Joyoung) for fifteen cycles. The debris was removed by low-speed centrifugation (6,000g) for 10 min. The supernatant was further centrifuged at high speed (20,000g) for 1 h. The pellet was then resuspended in two volumes of homogenization buffer B (20 mM HEPES, pH 7.4, 1 M NaCl, 1.3 μg ml−1 aprotinin, 1 μg ml−1 pepstatin, 5 μg ml−1 leupeptin, 0.2 mM PMSF and 2 mM DTT) and flash-frozen in liquid nitrogen.

Purification of pRyR2 by GST–FKBP12.6

The pRyR2–FKBP12.6 complex was purified based on previously described procedures30 with slight modifications. The membrane of the sarcoplasmic reticulum from a single porcine heart was solubilized at 4 °C for 2 h in homogenization buffer B supplemented with 5% CHAPS and 1.25% soy bean lecithin. After solubilization, the final concentration of NaCl in the system was diluted to 200 mM by homogenization buffer B without NaCl. Approximately 5–6 mg of purified GST–FKBP12.6 was then added to the system and further incubated for 1 h at 4 °C. After ultrahigh-speed centrifugation (200,000g), the supernatant was loaded onto a GS4B column (GE Healthcare). The resin was washed with buffer similar to the homogenization buffer B, except that the NaCl concentration was 200 mM and 0.1% digitonin was added. The complex was eluted by a solution containing 80 mM Tris, pH 8.0, 200 mM NaCl, 10 mM GSH, 0.1% digitonin, 1.3 μg ml−1 aprotinin, 1 μg ml−1 pepstatin, 5 μg ml−1 leupeptin, 0.2 mM PMSF and 2 mM DTT. The eluted protein was further purified through SEC (Superose 6, 10/300 GL, GE Healthcare) in buffer F. The pRyR2–FKBP12.6 complex fractions were concentrated to approximately 0.1 mg ml−1 for electron microscopy sample preparation. Specifically, for the FKBP12.6/apo-CaM sample, 5 mM EDTA, which has no effect on the zinc finger structure of RyR230, was included throughout purification of RyR2. For the CHAPS- and DOPC-treated FKBP12.6/ATP/caffeine/low-[Ca2+] and CHAPS- and DOPC-treated FKBP12.6/ATP/caffeine/low-[Ca2+]/Ca2+-CaM samples, the proteins were extracted only by CHAPS and washed and eluted by a buffer containing 0.5% CHAPS plus 0.002% DOPC. The eluted proteins were further purified through SEC in buffer F except 0.1% digitonin was replaced by 0.25% CHAPS plus 0.001% DOPC. For the PCB95/low-[Ca2+]/Ca2+-CaM sample, the proteins were purified using GST–FKBP12 as a bait and the RyR2–FKBP12 (containing GST–FKBP12) complex fell apart during SEC purification30.

Cryo-EM sample preparation

The cryo-EM samples of RyR2–CaM complexes were prepared as follows. FKBP12.6/apo-CaM: 5 mM EDTA was added to CaM (in buffer F) before sample preparation, and CaM with a final concentration of 250 μM was added to RyR2 (in buffer F plus 5 mM EDTA). FKBP12.6/ATP/caffeine/low-[Ca2+]/CaM-M: CaM-M (in buffer F) and RyR2 (in buffer F) were separately added with 20 μM Ca2+ (low-Ca2+ concentration), 5 mM ATP and 5 mM caffeine and CaM-M with a final concentration of 250 μM was added to RyR2. FKBP12.6/ATP/caffeine/low-[Ca2+]/Ca2+-CaM: CaM (in buffer F) and RyR2 (in buffer F) were separately added to 20 μM Ca2+, 5 mM ATP and 5 mM caffeine and CaM with a final concentration of 2.5 μM was added to RyR2. The other samples were prepared by the same procedure. Note that 5 mM Ca2+ represents high-Ca2+ concentration. Vitrobot Mark IV (FEI) was used for the preparation of cryo-EM grids. The procedures for preparing the eight samples were the same. Aliquots (3 μl each) of pRyR2 samples were placed on glow-discharged lacey carbon grids (Ted Pella). Grids were blotted for 2 s and flash-frozen in liquid ethane. Owing to the presence of high concentrations of Ca2+ in the filter paper used for blotting, it has previously been reported that the final concentration of free Ca2+ may be much higher than those used during sample preparation46. The low- and high-Ca2+ concentrations presented here only indicate those used during sample preparation and may be lower than the true Ca2+ concentrations.

Cryo-EM image acquisition

With regard to the FKBP12.6/ATP/caffeine/low-[Ca2+], FKBP12.6/ATP/caffeine/low-[Ca2+]/CaM-M, CHAPS- and DOPC-treated FKBP12.6/ATP/caffeine/low-[Ca2+], CHAPS- and DOPC-treated FKBP12.6/ATP/caffeine/low-[Ca2+]/Ca2+-CaM, FKBP12.6/ATP/caffeine/high-[Ca2+]/Ca2+-CaM and PCB95/low-[Ca2+]/Ca2+-CaM datasets, grids were transferred to a Titan Krios (Thermo Fisher Scientific) electron microscope operating at 300 kV equipped with a Cs-corrector (Thermo Fisher Scientific), Gatan K2 Summit detector and GIF Quantum energy filter. Zero-loss movie stacks were automatically collected using AutoEMationII47,48 with a slit width of 20 eV on the energy filter and a defocus range from −1.3 μm to −1.7 μm in super-resolution mode at a nominal magnification of 105,000×. Each stack was exposed for 5.6 s with an exposure time of 0.175 s per frame, resulting in 32 frames per stack. The total dose was approximately 50 e− Å−2 for each stack. The stacks were motion-corrected with MotionCor249 and binned twofold, resulting in a pixel size of 1.091 Å per pixel. With regard to the FKBP12.6/apo-CaM and FKBP12.6/ATP/caffeine/low-[Ca2+]/Ca2+-CaM datasets, micrographs were collected using a Gatan K2 Summit detector mounted on a Titan Krios electron microscope (FEI Company) operating at 300 kV and equipped with a GIF Quantum energy filter (slit width 20 eV). Micrographs were recorded in the super-resolution mode with a normal magnification of 105,000 ×, resulting in a calibrated pixel size of 0.669 Å. Each stack of 32 frames was exposed for 8 s, with an exposing time of 0.25 s per frame. The total dose rate was about 45.6 e− Å−2 for each stack. All 32 frames in each stack were motion-corrected with MotionCor2 and binned to a pixel size of 1.338 Å. The defocus value of each image was set from −0.8 μm to −1.8 μm. In addition, dose weighting was performed50. The defocus values were estimated with Gctf51.

Image processing

Image-processing procedures were similar to those previously reported30. Diagrams of the procedures used in data processing are presented in Extended Data Fig. 2. For the FKBP12.6/apo-CaM dataset, 1,180,104 particles were picked from 7,800 micrographs by RELION 2.052 using templates low-pass-filtered to 20 Å to limit reference bias. After two rounds of two-dimensional classification, 832,833 particles were selected and subjected to global angular search three-dimensional classification using RELION 2.0 with one class and a step size of 7.5°. The electron microscopy map of the previously published open structure of RyR230, which was low-pass-filtered to 60 Å, was used as the initial model. After global angular search three-dimensional classification, the particles were further subjected to three-dimensional classification with 10 classes and a local angular search step of 3.75°. The local angular search three-dimensional classification was performed several times with the output from different iterations of the global angular search three-dimensional classification as input. After the merging of all good classes and removal of the duplicated particles, the particles were subjected to three-dimensional autorefinement using THUNDER software53. The final particle number for the three-dimensional autorefinement was 208,715, resulting in a 3.6 Å resolution map after post-processing. The same procedures were performed for the other datasets. The resolution was estimated with the gold-standard Fourier shell correlation 0.143 criterion54 with the high-resolution noise-substitution method55.

Model building and structure refinement

The model of the RyR2 open structure (PDB code 5GOA)30 was fitted into the maps of the eight conditions by Chimera56 and manually adjusted in COOT57. FKBP12 from the rabbit RyR1/FKBP12 complex structure (PDB code 3J8H)58 was used for homologous model building of FKBP12.6. The apo-CaM from the crystal structure 3WFN was fitted into the maps obtained in the presence of CaM-M or apo-CaM and manually adjusted in COOT. Similarly, the crystal structure of Ca2+-CaM in complex with the RyR1 peptide (PDB code 2BCX) was fitted into the maps obtained in the presence of Ca2+-CaM and manually adjusted in COOT. Structure refinement was performed using PHENIX59 in real space with restrained secondary structure and geometry. The statistics of the three-dimensional reconstruction and model refinement are summarized in Extended Data Table 1.

Evaluation of the conformations of N- and C-lobes of CaM

Different conformations of N- and C-lobes were docked into the electron microscopy reconstruction for FKBP12.6/ATP/caffeine/high-[Ca2+]/Ca2+-CaM using the ‘Fit in Map’ tool of Chimera, selecting the options that include ‘Real-time correlation’, ‘7-Å resolution of Use map simulated from atoms’, ‘Use only data above contour level from first map’, ‘Optimize correlation’, ‘Correlation calculated about mean data value’, ‘Allow rotation and shift’ and ‘Move whole molecules’.

Site-directed mutagenesis

Point mutations in mouse Ryr2 and in human CALM1 were generated with the overlap extension method using PCR. In brief, a EcoRV/HpaI DNA fragment containing the RyR2(Y2156A) mutation and an AgeI/SalI fragment containing the RyR2(V3599A), RyR2(W3587A) or RyR2(L3590A) mutations were obtained by overlapping PCR and used to replace the corresponding wild-type fragment in the NheI/BsiWI fragment of Ryr2, The mutated NheI/BsiWI fragment was then used to replace the corresponding wild-type fragment in the full-length Ryr2 cDNA in pcDNA5. A HindIII/XhoI full-length CALM1 DNA fragment containing various point mutations was generated by overlapping PCR, which was then subcloned into pcDNA3. All point mutations in Ryr2 and CALM1 were confirmed by DNA sequencing.

Generation of stable, inducible cell lines expressing RyR2(WT) and mutants

Stable, inducible HEK293 cell lines expressing RyR2(WT), RyR2(Y2156A), RyR2(V3599A), RyR2(W3587A) and RyR2(L3590A) were generated using the Flp-In T-REx Core Kit from Invitrogen. These cell lines were not authenticated. These cells tested negative for mycoplasma contamination. In brief, Flp-In T-REx HEK293 cells were co-transfected with the inducible expression vector pcDNA5/FRT/TO containing the Ryr2WT or Ryr2-mutant cDNA and the pOG44 vector encoding the Flp recombinase in 1:5 ratios using the calcium phosphate precipitation method. The transfected cells were washed with phosphate buffered saline (PBS; 137 mM NaCl, 8 mM Na2HPO4, 1.5 mM KH2PO4 and 2.7 mM KCl, pH 7.4) 24 h after transfection followed by a change into fresh medium for 24 h. The cells were then washed again with PBS, collected and plated onto new dishes. After the cells had attached (around 4 h), the growth medium was replaced with a selection medium containing 200 μg ml−1 hygromycin (Invitrogen). The selection medium was changed every 3–4 days until the desired number of cells was grown. The hygromycin-resistant cells were pooled, aliquoted (1 ml) and stored at −80 °C. These positive cells are believed to be isogenic, because the integration of Ryr2 cDNA is mediated by the Flp recombinase at a single FRT site.

Single-cell luminal Ca2+ imaging

Luminal Ca2+ levels in RyR2WT- or Ryr2-mutant-expressing HEK293 cells transfected with or without CaM(WT) or CaM mutants were measured using single-cell Ca2+ imaging and the fluorescence resonance energy transfer (FRET)-based endoplasmic-reticulum luminal Ca2+-sensitive chameleon protein D1ER as previously described60,61. The cells were grown to 95% confluence in a 75-cm2 flask, dissociated with PBS (137 mM NaCl, 8 mM Na2HPO4, 1.5 mM KH2PO4 and 2.7 mM KCl, pH 7.4) and plated on glass coverslips placed on tissue culture dishes at approximately 10% confluence 18–20 h before transfection with cDNA for D1ER and cDNAs for CaM(WT) or CaM mutants using the calcium phosphate precipitation method. After transfection for 24 h, the growth medium was then changed to an induction medium containing 1 μg ml−1 tetracycline. After induction for around 22 h, the coverslip was mounted onto an inverted microscope (Nikon TE2000-S) and the cells on the coverslip were perfused continuously with Krebs–Ringer–HEPES buffer (125 mM NaCl, 5 mM KCl, 1.2 mM KH2PO4, 6 mM glucose, 1.2 mM MgCl2 and 25 mM HEPES, pH 7.4) containing various concentrations of CaCl2 (0, 1 and 2 mM) to induce SOICR, followed by the addition of 1.0 mM tetracaine, which was used to estimate the store capacity, and caffeine (20 mM), which was used to estimate the minimum store level by depleting the endoplasmic-reticulum Ca2+ stores at room temperature (23 °C). Images were captured with Compix Simple PCI 6 software every 2 s using the Nikon TE2000-S inverted microscope equipped with an S-Fluor 20×/0.75 NA objective. The filters used for D1ER imaging were λex = 436 ± 20 nm for CFP and λex = 500 ± 20 nm for YFP, and λem = 465 ± 30 nm for CFP and λem = 535 ± 30 nm for YFP with a dichroic mirror (500 nm). The amount of FRET in individual cells was determined from the ratio of the light emission at 535 and 465 nm. FSOICR is defined as the FRET level at which SOICR occurs, and Ftermi is defined as the FRET level at which SOICR terminates. The maximum FRET signal Fmax is defined as the FRET level after tetracaine treatment. The minimum FRET signal Fmin is defined as the FRET level after caffeine treatment. The termination and activation thresholds of SOICR in individual cells were determined using the equations shown in Extended Data Fig. 7a. The store capacity is calculated by subtracting Fmin from Fmax. Individual data points represent the average measurements of around 10–30 cells from one coverslip in one set of experiment. The number of experiments and coverslips for each condition is used as the sample size for data analyses.

Single-channel recordings in planar lipid bilayers

Recombinant RyR2(WT) and mutant channels were purified from cell lysates prepared from HEK293 cells transfected with the Ryr2WT or Ryr2-mutant (Y2156A, V3599A, W3587A or L3590A) cDNA by sucrose density gradient centrifugation as previously described62. Heart phosphatidylethanolamine (50%) and brain phosphatidylserine (50%) (Avanti Polar Lipids), dissolved in chloroform, were combined and dried under nitrogen gas and resuspended in 30 μl of n-decane at a concentration of 12 mg lipid per ml. Bilayers were formed across a 250-μm hole in a Delrin partition separating two chambers. The trans chamber (800 μl) was connected to the head stage input of an Axopatch 200A amplifier (Axon Instruments). The cis chamber (1.2 ml) was held at virtual ground. A symmetrical solution containing 250 mM KCl and 25 mM HEPES, pH 7.4 was used for all recordings, unless indicated otherwise. A 4-μl aliquot (around 1 μg protein) of the sucrose density gradient-purified recombinant RyR2(WT) or mutant channels was added to the cis chamber. Spontaneous channel activity was always tested for sensitivity to EGTA and Ca2+. The chamber to which the addition of EGTA inhibited the activity of the incorporated channel presumably corresponds to the cytosolic side of the Ca2+ release channel. The direction of single channel currents was always measured from the luminal to the cytosolic side of the channel, unless mentioned otherwise. Recordings were filtered at 2,500 Hz. Data analyses were carried out using the pCLAMP 8.1 software package (Axon Instruments). Free Ca2+ concentrations were calculated using a computer program that has previously been described63.

Statistical analysis

Data are mean ± s.e.m., derived from independent samples or independent experiments. All experiments were performed with at least five biological replicates. The GraphPad Prism 8.1 software was used to test for differences between groups. We used Student’s t-test (paired, two-tailed) or one-way ANOVA with a Dunnett’s post hoc test. P < 0.05 was considered to be statistically significant.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.