In situ structural insights into the excitation-contraction coupling mechanism of skeletal muscle

Abstract

Excitation-contraction coupling (ECC) is a fundamental mechanism in control of skeletal muscle contraction and occurs at triad junctions, where dihydropyridine receptors (DHPRs) on transverse tubules sense excitation signals and then cause calcium release from the sarcoplasmic reticulum via coupling to type 1 ryanodine receptors (RyR1s), inducing the subsequent contraction of muscle filaments. However, the molecular mechanism remains unclear due to the lack of structural details. Here, we explored the architecture of triad junction by cryo–electron tomography, solved the in situ structure of RyR1 in complex with FKBP12 and calmodulin with the resolution of 16.7 Angstrom, and found the intact RyR1-DHPR supercomplex. RyR1s arrange into two rows on the terminal cisternae membrane by forming right-hand corner-to-corner contacts, and tetrads of DHPRs bind to RyR1s in an alternating manner, forming another two rows on the transverse tubule membrane. This unique arrangement is important for synergistic calcium release and provides direct evidence of physical coupling in ECC.

The in situ structure of RyR1-DHPR supercomplex provides direct evidence to their mechanical coupling in skeletal muscle.

INTRODUCTION

Muscle contractions enable essential animal behaviors, such as cardiac pulsations, exercise, respiration, and maintenance of posture. Excitation-contraction coupling (ECC) is a fundamental process that links the action potential to the contraction of striated muscle fibers in vertebrates (1). Upon arrival of action potentials at the neuromuscular junction in skeletal musculature, calcium inflow into presynaptic terminals triggers exocytosis of acetylcholine-containing sacs into synaptic clefts, leading to depolarization of the plasma membrane. The depolarization propagates along transverse tubules (T-tubules), which are invaginations of the sarcolemma that penetrate deep into skeletal muscle fibers (2, 3). Triad junctions are formed in skeletal muscle by T-tubules sandwiches between the two sides of enlarged portions of the sarcoplasmic reticulum (SR), named terminal cisternae (TC), ensuring an efficient ECC process (4). Then, in triad junctions, voltage-sensitive dihydropyridine receptors (DHPRs) in the T-tubule membrane (TTM) interact with type 1 ryanodine receptors (RyR1s) in the adjacent TC membrane (TCM), leading to a rapid release of Ca²⁺ from the SR and the subsequent contraction of myofibrils (5–8).

As the high-conductance Ca²⁺ channel localized at the SR, RyR is the largest known ion channel, with ~5000 residues in each protomer and a homotetrameric organization of over 2.2 million dalton (MDa) (9, 10). It consists of a relatively small transmembrane region forming an ion-conducting pore and a large cytoplasmic region responsible for interacting with various ligands to regulate the channel (11). Mammals have three RyR isoforms, named RyR1, RyR2, and RyR3. RyR1 and RyR2 are expressed at high levels in skeletal and cardiac muscles, respectively (12–14). In triad junctions of skeletal muscle, RyR1s are anchored to TCM in close proximity, thereby constituting efficient Ca²⁺ release units (CRU) with DHPRs located on the T-tubules. Upon arrival of the action potential, DHPR triggers opening of the RyR1 channel through mechanical coupling, but the details of this process remain enigmatic.

In the 1980s, high-density “foot” structures were found in the triad junctions of frog twitch fibers (15). Further investigation demonstrated that these structures were RyR1 channels (16). In 1989, RyR1 was purified from the fast-twitch skeletal muscle in rabbit, and its distinctive mushroom-shaped structure was resolved at 3.7-nm resolution (17). With the rapid development of cryo–electron microscopy (cryo-EM) in recent years, numerous high-resolution structural studies from purified specimens have further illuminated different states of RyR1 channels (18–22). However, how RyR1s are activated and regulated in situ in skeletal muscle is unclear.

At the triad junctions of skeletal muscle, RyR1 typically appears in situ in two (sometimes three) parallel rows (23–25). This pattern is conserved across crustaceans and vertebrates (23–26). In swim bladder muscle fibers from toadfish, adjacent RyR1 channels were found to be arranged in a slightly twisted corner-to-corner manner (27). Other studies revealed that when crystallized into two-dimensional (2D) arrays, RyR1 channels form a checkerboard-like pattern with adjacent pairs bound together in a similar corner-to-corner way (28).

In addition to the arrangement of RyR1s in the triad junctions, how RyR1s interact with DHPRs at the triad junctions remains a topic of great research interest because it is related to the molecular mechanism of ECC. Previous studies using the freeze-fracture technique have revealed the presence of diamond-shaped particle clusters, referred to as “tetrads” on TTMs (29, 30). These observations suggest a direct mechanical interaction between RyR1 and DHPR. Furthermore, it has been observed that the separation between individual DHPRs within a tetrad decreases in response to pharmacological manipulation of RyR1, providing compelling evidence for the physical linkage between DHPRs and RyR1 (30). Although different possible interaction sites and modes between RyR1 and DHPR have been proposed in recent decades (31–41), the interaction between them is still not observed to confirm the existence of their physical coupling.

Recent trails of in situ structural studies using cryo–electron tomography (cryo-ET) revealed the presence of RyR1-like density on the TCM of triad junctions from cryo-lamellae of toadfish swim-bladder muscle prepared by the cryo–focused ion beam (cryo-FIB) technique (42). Besides, SR vesicles purified from rabbit skeletal muscle were used to solve the in situ structure of RyR1 at near nanometer resolution (43), and the resolution was updated to subnanometer 9.1 Å in recent years (44, 45). However, in these studies, no DHPRs were observed on the TTM due to the low resolution of the tomogram (42) or to potential specimen damage during the purification procedure (43, 44). We recently developed a workflow, VHUT-cryo-FIB, for cryo-lamellae preparation from tissue specimens and successfully observed clear densities within the triad junctions of rat skeletal muscle in the tomogram (46), providing a potential method to obtain high-resolution in situ structures of RyR1/DHPR in skeletal muscle.

In this study, we explored another specimen preparation procedure to fabricate cryo-lamellae of mouse skeletal muscle using cryo-FIB and then performed an in situ cryo-ET study of intact triad junctions. Both RyR1 and DHPR densities were observed within the triad junctions in our tomograms. Then, we resolved the in situ structures of both the RyR1 tetramer and RyR1-DHPR supercomplex using subtomogram averaging (STA) with the resolution of 16.7 and 33 Å, respectively. We further deduced the ordered spatial arrangements of RyR1s and DHPRs within the triad junctions, which were further investigated by molecular dynamics simulations, suggesting the mechanism of synergic coupling between RyR1 and DHPR and among RyR1s. Our results provide deep structural insights into the physiology of triad junctions of mammalian skeletal muscle and highlight avenues toward better understanding the ECC mechanism in skeletal muscle.

RESULTS

Overall intact structure of the triad junction

To preserve the native integrity of triad junctions, we prepared cryo-lamellae from mouse skeletal muscle tissue by improving the previous protocol (47). Specifically, we separated intact skeletal muscle fibers from the extensor digitorum longus (EDL) muscle of 2- to 3-week-old mice (Fig. 1A and fig. S1A). The fibers were exposed to a 10% glycerol solution for several minutes before placing onto a cryo-EM grid, which ensured complete vitrification during subsequent plunge freezing (48–50). Then, the vitrified specimen was milled to a thickness of approximately 150 nm using cryo-FIB, and the cryo-lamellae were subjected to cryo-ET data collection by targeting the triad junctions manually (fig. S1, B to E, and table S1). The textures of muscle actin filaments were prominently visible in the raw cryo-EM micrograph, and numerous strips constituting the triad junction could be found perpendicular to the filaments and aligned parallelly and orderly (Fig. 1B). We measured the distances between two adjacent parallel triad junctions and found that they ranged from 594 nm to 1171 nm with a mean of 833 nm and an SD of 122 nm (Fig. 1C).

Fig. 1. Overall architecture of native triad junctions in mouse skeletal muscle.

(A) EDL muscle fibers were manually extracted from the mouse leg, and the black circle indicates the extraction position. (B) Low-magnification cryo-EM image of cryo-lamella of mouse skeletal muscle. Regularly arranged triad junctions are marked by black dashed line boxes. The mitochondria near triad junctions are marked by black triangles. Scale bar, 1 μm. (C) Distribution of the interval distances among the triad junctions. (Number of samples is 277, mean value is 833 nm, SD is 122 nm.) (D) A representative tomogram slice shows two perpendicular triad junctions and their surrounding environment. Scale bar, 100 nm. Horizontal and vertical triad junctions are marked by blue and red dashed line boxes, respectively. (E) Segmented tomogram shows different components around triad junctions in (D), including RyR1 (red), DHPR (pink), SR (cyan), T-tubule (blue), ribosome (purple), mitochondrion (green), thick filament (yellow), and thin filament (orange). The segmentation was performed by using the Microscopy Image Browser (MIB) tool (102). (F) Distributions of length/width and thicknesses of T-tubules in horizontal (blue) and vertical (red) triad junctions. (Length: horizontal, number of samples is 173, mean value is 498 nm, SD is 197 nm; vertical, number of samples is 25, mean value is 77 nm, SD is 16 nm. Thickness: horizontal, number of samples is 173, mean value is 53 nm, SD is 27 nm; vertical: number of samples is 25, mean value is 58 nm, SD is 27 nm.)

From a representative tomogram (Fig. 1, D and E, and movie S1), we observed useful information about the ultrastructure around the triad junctions. The T-tubule is enclosed by TCM on both sides with RyR1s clearly on the TCM and putative DHPRs on the TTM. Thick and thin filaments dominate most of the cellular space and are proximal to triad junctions, ready to perceive calcium signals from SR and then trigger dynamic sliding of myosin on actin filaments. Mitochondria were found to be strategically localized near the triad junctions, where they can offer critical energy necessary for the sliding of myosin. In addition, we found abundant ribosome particles surrounding the SR, which are important for the protein metabolism of both triad junctions and myofilaments. The presence of two perpendicular triad junctions discloses the narrow and flattened shape of T-tubules, which form a network inside the skeletal muscle and attach to the flat and elongated surfaces of TCs.

We then performed statistical analyses of the sizes of T-tubules by using a total of 198 tomograms containing the most complete triad junctions (Fig. 1F). The measured apparent lengths of T-tubules vary continuously from 49 to 996 nm because of the randomly distributed angles between the long axes of T-tubules and the direction of cryo-FIB milling. The dual-modular distribution of the apparent lengths indicates two categories of T-tubules, with one having an apparent length ranging from 160 to 996 nm, representing the tilted and horizontal T-tubules, and the other having an apparent length ranging from 49 to 118 nm, representing the close-to-vertical T-tubules. For the vertical T-tubules, there are only two RyR1 molecules located on TCMs, while for the tilt-and-horizontal T-tubules, there are more RyR1s observed.

Notably, the measured apparent lengths of vertical T-tubules represent their widths. Thus, we were able to calculate the average length of T-tubules as 498 nm with a SD of 197 nm and the average width of T-tubules as 77 nm with a SD of 16 nm (Fig. 1F). Because of the limited imaging area of each tomogram with a limited coverage of triad junctions, we believe that there would be many triad junctions much longer in length than they appear in tomograms. It is noteworthy that the thicknesses of both horizontal and vertical T-tubules are almost the same, with mean values of 53 and 58 nm, respectively (Fig. 1F), suggesting potential physiological relevance.

In situ structure of RyR1 in native triad junctions

We then manually picked out the foot-like particles that represent RyR1s located between the TTM and TCM and performed subtomogram analysis (fig. S2A), resulting in a final averaged cryo-EM map of in situ RyR1 with a resolution of 16.7 Å (Fig. 2A, fig. S2C, table S1, and movie S2). In addition to the typical square shape of RyR1 in its top view with a width of 26.7 nm, we observed both TCM and TTM in the cryo-EM map. The transmembrane region of RyR1 is embedded in the TCM, and the distance between the TCM and TTM can be measured as 16.7 nm, which leaves space for the interaction between RyR1 and DHPR.

Fig. 2. In situ structure of RyR1 in the native triad junction.

(A) Side and top views of cryo-EM map of in situ RyR1 embedded in TCM and proximal to TTM. (B) Structural model of the RyR1-FKBP12-apo-CaM complex was flexibly fitted into the cryo-EM map in (A). FKBP12s, apo-CaMs, and four RyR1 protomers are indicated accordingly. (C) Distribution and definition of mouse RyR1 domains with one protomer fitted into the cryo-EM map. Locations of FKBP12 and apo-CaM are indicated by the black dashed boxes. (D) Zoomed-in view of the FKBP12 region in (C). (E) Zoomed-in view of the apo-CaM region in (C).

The cryo-EM map of in situ RyR1 can be well fitted with the previously reported high-resolution structures of RyR1s in both open and closed states (fig. S3, A to F) (18, 22). The model-to-map Fourier shell correlation (FSC) analysis revealed that the fitting qualities of both open and closed RyR1 models into the map are remarkably similar (fig. S3G). Therefore, it is challenging to determine which specific state our in situ RyR1 map resembles, which is mainly due to the current limited resolution. In addition, we observed that the consistence resolutions between the models and map are just ~35 Å (FSC = 0.5), which are much lower than the resolution of the map (16.7 Å). This implies that there might be some conformational changes between our in situ structure of RyR1 and these two reported structures in vitro. We also observed several different pieces of densities, one piece around the clamp region and other pieces around the handle and helical domains, which would represent the binding partners of RyR1, including FK506 binding protein 12 (FKBP12) and calmodulin (CaM), respectively.

FKBP12, which is alternatively known as calstabin1 and abundantly expressed in skeletal muscle, has been reported to interact physically with RyR1 and effectively stabilize the closed state of RyR1 (51, 52). The high-resolution structure of the RyR1-FKBP12 complex [Protein Data Bank (PDB) entries 3J8H and 5TB0] (19, 21) fits appropriately into the cryo-EM map of in situ RyR1, leaving distinct pieces of densities around the handle and helical domains, which would represent the density of CaM (fig. S3, C and D).

CaM is another key binding partner that can regulate the activity of RyRs, while in skeletal muscle, the effect of this regulation varies depending on the concentration of Ca²⁺ (53). At nanomolar concentrations of Ca²⁺, CaM is in its apo form (apo-CaM), which can bind to RyR1 and weakly activate the opening of RyR1, while at micromolar concentrations of Ca²⁺, CaM is in its calcium binding form (Ca²⁺-CaM), which will bind to RyR1 with a different binding site and inhibit the opening of RyR1 (52, 54, 55). In recent years, the high-resolution structures of RyR1 bound with CaM were determined with (PDB entry 7TZC) and without Ca²⁺ (PDB entry 6X33) in the buffer (56, 57). However, both structures were supposed to be closer to the apo-CaM bound state (RyR1-apo-CaM) because the binding position of CaM in the structure of Ca²⁺ loading RyR2-CaM complexes (58) shows a slight slipping down compared to these two structures. These two structures can be fitted into our cryo-EM map of in situ RyR1, including the abovementioned pieces of densities (fig. S3, E and F). In addition, we found that the density near the HD1 domain in our cryo-EM map is absent in the low-pass–filtered map of RyR2-Ca²⁺-CaM [Electron Microscopy Database (EMDB) entry EMD-9833] but exists in the low-pass–filtered map of RyR2-apo-CaM (EMDB entry EMD-9836) (fig. S3H) (58). These structural analyses suggest that the present in situ structure of RyR1 represents the apo-CaM binding state of RyR1, which is further stabilized by the binding of FKBP12. Considering that we did not trigger myofibrils with additional calcium buffer during specimen preparation, we speculate that the present in situ structure of RyR1 represents a closed steady state.

On the basis of a combination of homology modeling (fig. S4) and artificial intelligence modeling, we built a full-length model of mouse RyR1 that was fitted into the in situ cryo-EM map together with the fitting of both FKBP12 and apo-CaM. Then, the molecular dynamics flexible fitting (MDFF) approach was used to obtain an integrative in situ structural model of the RyR1-FKBP12-apo-CaM complex (Fig. 2, B and C, and movie S2). In this model, FKBP12 is encapsulated by the N-terminal domain, SPRY1 and SPRY3 domains, and handle domain (Fig. 2D), while apo-CaM is surrounded by the central, handle, and helical domains (Fig. 2E). Notably, the connection between apo-CaM and the HD2 domain at the bottom of the clamp region could only be found in the present in situ map (fig. S3I) and was missing in the previously reported cryo-EM map of the RyR1-apo-CaM complex (56, 57), indicating a stabilized state of in situ RyR1 in the present study.

In addition, we observed that the apo-CaM model appeared to be partially accommodated by the density, suggesting a potential partial occupancy of apo-CaM in the in situ condition. Therefore, we performed the C4 symmetry expansion and further classified the particles by focusing on one RyR1 protomer (fig. S5), which resulted in two classes with density differences at the region corresponding to CaM. In the first class, the local density fully accommodates the structural model of CaM, while, in the second one, the density can only partially accommodate the model of CaM. The ratio of the particle numbers in these two classes is approximately 6:4, suggesting that at least half of the in situ RyR1 particles would be in the CaM binding state.

Furthermore, we compared our in situ skeletal RyR1 map with the previously reported RyR1 maps in nanodiscs (EMDB entry EMD-2751) and SR vesicles (EMDB entry EMD-10840), respectively (fig. S6). In comparison with the density of RyR1 solved in nanodiscs, our in situ RyR1 exhibits additional densities corresponding to FKBP12 and CaM. Also, comparing with the density of RyR1 solved in the SR vesicles, we observed additional densities that would correspond to CaM. Furthermore, in comparison to our in situ RyR1 map, the map of RyR1 solved in the SR vesicles shows the difference densities around the HD2 and P2 domains, which is possibly due to the potential conformational changes between RyR1s in the SR vesicles and in the skeletal muscle in situ. The exact explanation for this difference may need a higher resolution density map in the future.

Arrangement of RyR1s in native triad junctions

To investigate the in situ organization of RyR1s, we mapped the aligned particles of RyR1 tetramers back to their original tomograms and found that RyR1 tetramers are predominantly arranged one-by-one in two parallel rows within the triad junctions (Fig. 3A, fig. S7A, and movie S3). Moreover, we found that two neighboring RyR1 tetramers interact with each other in a right-hand corner-to-corner manner (Fig. 3A and fig. S7A), which means that when viewed from the cytoplasmic side of RyR1, the interaction interfaces of neighboring RyR1 tetramers are always located at the right side of the corner of each tetramer. Such a right-hand corner-to-corner array was suggested by previous studies of 2D crystals of the RyR1 tetramer (28), a thin section study of triad junctions (59) and a low-resolution cryo-ET study based on isolated triad junctions (43). In this study, we not only resolved the right-hand corner-to-corner arrangement of RyR1 tetramers directly at high resolution but also quantified the distances between two adjacent tetramers, which ranged primarily (85%) from 29 to 33 nm with an average of 31 nm (Fig. 3B).

Fig. 3. In situ arrangement of RyR1s in the native triad junction.

(A) Two representative tomogram slices showing the embedded array of RyR1 tetramers in the triad junctions. The structures of RyR1s are plotted back into the tomograms according to their refined coordinates and Euler angles. Scale bars, 20 nm. (B) Statistical distribution of distances between two neighboring RyR1 tetramers. (C) Slice of the STA averaged map of dimeric RyR1 tetramers. Scale bar, 20 nm. (D) Diagram of the geometric arrangement of RyR1 tetramers, which are represented as square boxes. (E) STA averaged map of dimeric RyR1 tetramers in the triad junction, showing both TTM and TCM. The structural model of RyR1-FKBP12-apo-CaM is fitted into the map. The contact site between two tetramers is marked by the black dashed box. (F) The zoomed-in view of the contact site in (E). The nearest potential contacting residues Lys³¹¹⁵ and Leu³¹¹³ are labeled with their main-chain distance indicated. The position of residue Gly³¹⁹², whose mutation was reported to be relevant to premature mortality in two patients (61), is indicated in red.

In addition, we observed defects in the RyR1 lattice with occasional receptors missing. To quantify this observation, we defined a 2 × 2 arrangement of RyR1s within two rows as a 4-RyR1 unit. If one or more RyR1s are missing within one 4-RyR1 unit, we defined this unit as a defective 4-RyR1 unit. Then, we examined the raw tomograms by plotting back the aligned RyR1 models and analyzed all the 4-RyR1 units. To avoid the possibility of the cryo-FIB milling damage, we focused on analyzing RyR1 arrays that are located far away from the upper and lower surfaces of the tomogram. We managed to analyze 329 units in total and find that there are 131 units belonging to the defective ones (fig. S7B). The existence of the defective units could be due to the missing during particle picking and the particle cleaning during data processing and could also reflect the occasionally imperfect packing of RyR1 in nature.

To confirm the exact interaction between two adjacent RyR1 tetramers, we resubtracted the subtomograms of neighboring tetramers and then performed subtomogram averaging based on the original orientation parameters of a single tetramer, resulting in a clear density at the corner-to-corner interaction site of two neighboring tetramers (Fig. 3C, table S1, and movie S3). The center-to-center separation between these two tetramers was measured as 31.1 nm (Fig. 3D), which is consistent with the above statistics. The length of the square edge of the RyR1 tetramer was also measured as 26.7 nm, and the angle between the diagonal line of the tetramer and the central line of the RyR1 array was measured as 17.9° (Fig. 3D), which is related to the exact position of the corner interaction. These precise measurements illustrate the geometric arrangement of RyR1 tetramers in the native triad junction, providing direct proof of previous suggestions from 2D crystallographic studies (28).

We then fitted the structural model of the RyR1-FKBP12-apo-CaM complex into the averaged map of dimeric RyR1 tetramers (Fig. 3E) and observed the potential contact site of RyR1 tetramers that is situated in the HD2 domain, specifically with residues ranging from 3087 to 3118 (Fig. 3F). We found that the nearest site lies between Leu³¹¹³ of one RyR1 tetramer and Lys³¹¹⁵ of another tetramer, and their main-chain distance is less than 3.5 Å, suggesting a potential direct interaction at this site (Fig. 3F). Previous studies have suggested that the contact site of neighboring RyR1 tetramers is located at residues ranging from 2540 to 3207 (28, 60). Furthermore, it has been reported that the point mutation of Gly³¹⁹¹ in RyR1, located proximal to the contact site, is causative for premature mortality in two patients (61). According to the structural model of the dimeric RyR1 tetramers here, we speculate that the alteration of Gly³¹⁹¹ would strongly and negatively affect the corner-to-corner interaction between neighboring tetramers and thus eliminate the coupling between them, which will be further studied below.

Molecular dynamics insight into the coupling of neighboring RyR1s

The specific array arrangement of RyR1 tetramers with corner-to-corner interactions suggests its essential role in the conformational change coupling of RyR1s in one triad junction, which has been hypothesized previously (28) but lacks experimental support. Here, on the basis of the in situ structural model of the RyR1 tetramer array, we performed molecular dynamics (MD) simulations (tables S2 and S3) to investigate the structural dynamics of RyR1s in the array.

First, coarse-grained (CG) MD simulations were performed to explore the assembly of four RyR1 tetramers (fig. S8A and movies S4 and S5). Starting with RyR1 tetramers more than 20 Å away from each other, the RyR1 tetramers in the open conformation (movie S4) tend to form more right-hand corner-to-corner contacts through intermolecular helical-helical domain interactions than those in the closed conformation (movie S5) during 3-μs CG simulations (Fig. 4, A and B, and fig. S8B). The assembled architecture formed by the open RyR1s is consistent with their arrangements in the triad junctions in situ (Fig. 4A and fig. S8B). This observation suggests that RyR1s in the open conformation can spontaneously self-assemble into the right-hand corner-to-corner contact pattern.

Fig. 4. Molecular dynamics simulations of RyR1-RyR1 interactions and dynamics.

(A) Representative results of CG MD simulations of four RyR1 tetramers in two states. Residues ranging from 2951 to 3240 are shown in red, and the right-hand contacts between RyR1s are marked by black circles. (B) Statistics of the counts of right-hand corner-to-corner contacts observed during CG MD simulations for different states. (Open: number of samples is 10, mean value of count number is 1.6, SD is 1.0; closed: number of samples is 10, mean value is 0.5, SD is 0.7.) (C) Representative dimeric RyR1 tetramers in the open state from AA MD simulations. (D) Magnified views of two contact RyR1 protomers, a2 in purple and c1 in yellow, in different states, respectively. (E) Statistics of the counts of all contacts, hydrogen bonds, and hydrophobic contacts during the last 110 ns of AA MD simulations. (All contacts: open, mean value of count number is 33.5, SD is 11.6; closed, mean value is 23.5, SD is 7.2; H-bond: open, mean value of count number is 18.5, SD is 5.2; closed, mean value is 8.5, SD is 3.6; hydrophobic contact: open, mean value of count number is 11.3, SD is 3.6; closed, mean value is 7.8, SD is 2.7. Number of samples is 1098.) (F) DCCMs of the interacting RyR1 protomers, a2 and c1 in (D). Correlated values above zero are represented in gradient red, and the high positive value represents a strongly correlated motion. (G) Primary motions of residues (labeled as RES) ranging from 2494 to 3611 at the contact P2 and HD2 domains of the interacting RyR1 protomers in different states, respectively. These primary motions were derived based on the first principal component from PCA investigation of AA MD simulations of dimeric RyR1 tetramers.

Then, we performed all-atom (AA) MD simulations to examine the interactions and motions of two adjacent RyR1 tetramers (fig. S8C and movies S6 and S7). Starting from the initial separation setup, these two tetramers, in both open and closed conformations, tended to quickly interact with each other through the intermolecular helical-helical domains, and the distance of their center of mass (COM) decreased from ~330 Å to 315 to 325 Å in three independent MD replicas (Fig. 4C and fig. S9A). However, we observed that the open RyR1s exhibited more intermolecular hydrogen bonds (H-bonds) and hydrophobic contacts than the closed state, resulting in tighter helical-helical domain interaction (Fig. 4, D and E, and fig. S9B). In addition, the open RyR1s showed a stronger dynamic correlation than the closed state (Fig. 4F and fig. S9C). Specifically, compared to the closed RyR1s, notable correlative motions between open RyR1s were observed, particularly at the P2 and HD2 domains, with residues ranging from 2200 to 3611. The two most distant RyR1 protomers also showed strong correlative motions in the open state compared to the closed state (fig. S9, D and E).

Furthermore, we used principal components analysis (PCA) to conduct another comprehensive investigation of the global motions of assembled RyR1s in both open and closed states (fig. S10). We found that the two open RyR1s exhibit a concerted and synergetic motion of up and down swinging at the P2 and helical domains upon contact with each other (Fig. 4G and fig. S10). In contrast, these strong correlative and symmetrical movements were not observed for the RyR1s in the closed state (Fig. 4G and fig. S10). These analyses indicate that the strong interactions between neighboring RyR1s in the open state promote the synergetic movement of their P2 and helical domains, providing the structural basis of coupling between neighboring RyR1s for coactivation and synergistic calcium release, which was previously termed the calcium spark in cardiac muscle (62).

Physical connections between RyR1 and DHPR in the native triad junction

DHPR is commonly known as the voltage-gated L-type Ca²⁺ channel, also referred to as Cav1. It comprises several subunits, including the core pore-forming and voltage-sensing α1 subunit, the transmembrane γ subunit, the extracellular α2δ subunit, and the intracellular β1a subunit (63). The α1 and β1a subunits have been identified as critical components in the ECC process (64). Unlike cardiac muscle, which relies on a “calcium-induced calcium release” mechanism (65), DHPRs appear to transmit signals to RyR1s through mechanical coupling in skeletal muscle. Despite extensive research over recent decades in search of evidence for the interaction between RyR1 and DHPR, direct evidence is still absent due to a lack of detectable DHPR densities either in isolated T-tubule vesicles or in fixed skeletal muscle tissues (42–44).

In our tomograms featuring native triad junctions, we observed protrusion densities on the extracellular side of TTMs (Fig. 5A). These densities exhibited a symmetrical arrangement and were accompanied by RyR1 density on the corresponding TCM, some of which possessed an intracellular portion linked to the RyR1 corner (Fig. 5B). Considering previous results (29), it is highly likely that these densities anchored to the TTM belong to DHPRs. Intriguingly, at least 50% of all RyR1s lack this accompanying density, suggesting that the copy number of tetrads formed by DHPRs is lower than that of RyR1s in triad junctions (Fig. 5C).

Fig. 5. The in situ structure of the RyR1-DHPR supercomplex in the native triad junction.

(A) Representative tomogram slice shows that both RyR1 and DHPR densities existed in the triad junction. The pink box marks a RyR1-DHPR supercomplex, and the red box marks RyR1 without DHPR bound. Scale bar, 100 nm. (B) Magnified view of the RyR1-DHPR supercomplex in (A). (C) Magnified view of RyR1 without DHPR bound in (A). (D) Final averaged map of the RyR1-DHPR supercomplex with TTM and TCM. The distance between TTM and TCM is indicated. (E) Structural model of RyR1-FKBP12-apo-CaM and the previously reported structural model of DHPR (PDB entry 5GJW) (63) are fitted into the RyR1-DHPR supercomplex map. (F) Top view of the RyR1-DHPR supercomplex, showing the twist angle between the “tetrad” formed by four DHPRs and the RyR1 tetramer. (G) Two connection densities between RyR1 and DHPR are shown in green and yellow. The RyR1 domains and DHPR components are colored differently. (H) Magnified view of the density map around connection 1 in (G). (I) Magnified view of the density map around connection 2 in (G).

We manually picked RyR1 particles accompanied by DHPR-like densities and then performed STA with C4 symmetry (fig. S2B), resulting in a final cryo-EM map of the RyR1-DHPR supercomplex with a resolution of 33 Å (fig. S2D), which features RyR1 on the TCM and DHPRs on the opposing TTM (Fig. 5D and movie S8). The alignment procedure using C1 symmetry yielded a similar map with four extruding densities on TTM, showing that the DHPRs were arranged in a nearly C4 symmetrical pattern corresponding to the RyR1 tetramer (fig. S11A). The extracellular domains of DHPRs within the T-tubular lumen resemble the tetrad that was observed on TTMs in previous freeze-fracture studies (29) and can be fitted well with the model of the α2δ subunit of reported DHPR structures (PDB entry 5GJW) (63) in terms of both shape and size (fig. S11B).

To further investigate the interactions between RyR1 and DHPR, we fitted the above full-length mouse RyR1 model and the previously reported DHPR structure (PDB entry 5GJW) into the map and derived the structural model of the RyR1-DHPR supercomplex (Fig. 5E and movie S8). In this model, the main body of each DHPR molecule, consisting of α1, α2δ, and γ subunits, lies on top of the handle region of RyR1 instead of the RyR1 tetramer corner (Fig. 5F). Notably, in both raw tomograms (Fig. 5B) and the averaged map (Fig. 5G), there are two connection densities observed for each RyR1 and DHPR, denoted as connection 1 near the P1 subdomain of RyR1 and connection 2 near the P2 subdomain of RyR1 (Fig. 5, H and I), suggesting the physical interactions between RyR1 and DHPR directly or indirectly via other protein partners, such as STAC3 (SH3 and cysteine rich domain 3) (66, 67) and JP2 (junctophilin-2) (68, 69).

Arrangement of the RyR1-DHPR supercomplex in the native triad junction

By plotting the aligned RyR1-DHPR map back to the corresponding tomograms, we found that tetrads do not appear opposite every RyR1. Instead, they tend to skip one of two adjacent RyR1s in the same row and bind to the other RyR1 (Fig. 6A and movie S8). This result is consistent with our statistics of the distances between two neighboring RyR1-DHPR supercomplexes (Fig. 6B). Two peaks at ~44 and ~62 nm were found in the distribution histogram. The 44-nm peak coincides with the distance between two adjacent RyR1-DHPR complexes from different rows, and the peak of 62 nm matches the distance between two adjacent complexes in the same row (Fig. 6C). This unique RyR1-DHPR arrangement provides a good explanation for the previous observation in freeze-fracture experiments where the dispositions of tetrads in two adjacent rows were found in an alternating way (29).

Fig. 6. The arrangement of the RyR1-DHPR supercomplex in the native triad junction.

(A) Representative tomogram slice of the adjacent RyR1-DHPR supercomplex in one triad junction. The STA averaged maps of RyR1-DHPR supercomplexes are plotted back into the tomograms and shown as the side and top views, respectively. Scale bars, 20 nm. (B) Distribution histogram of distances between adjacent RyR1-DHPR supercomplexes in the tomograms. (C) Arrangement of RyR1-DHPR supercomplexes in the triad junction with their nearest distances measured. The locations of RyR1 tetramers without DHPR are indicated as red squares. (D) Architecture of the native triad junction is shown as a diagram with TCM and TTM colored light cyan and light blue, respectively. RyR1s without DHPR bound are shown in pink, and the RyR1-DHPR supercomplex is shown in pale magenta. The coordinates of the T-tubule are indicated. The top view of TCM from the cytoplasmic side and the top view of TTM from the lumen side of the T-tubule are shown below with the TTM set to partial transparency to show the arrangement of RyR1s.

To further explore the specific organization of RyR1-DHPR supercomplexes, we re-extracted the adjacent RyR1-only particles based on the coordinates of RyR1-DHPR supercomplexes and then performed subtomogram analysis. The resulting STA map showed the absence of noticeable DHPR protrusion densities on the TTM (fig. S11C), confirming the 1:2 stoichiometry between DHPR and RyR1 on the skeleton muscle triad junction. Besides, by comparing the density maps of RyR1s between RyR1-DHPR supercomplex and RyR1-only complex, we observed their potential conformational differences near the P2 and HD2 domain (movie S9), suggesting a presumable different functional state, which needs further investigations with future higher resolution maps.

By combining the above observations of arrangements of RyR1s and RyR1-DHPR supercomplexes as well as the statistical measurement of triad junction morphology, we were able to build an architectural model of the skeleton muscle triad junction (Fig. 6D). The entire triad junction is characterized by a long and narrow T-tubule sandwiched by two TCMs. The average length of the triad junction can be more than 500 nm with a width of ~80 nm and a thickness of ~55 nm. The width of the triad junction can accommodate only two rows of the RyR1 array embedded in the TCM. The distance between TCM and TTM is similar to the interstitial space of the triad junction, which allows the interaction between the cytoplasmic region of RyR1 and that of DHPR. At the TCM, RyR1s interact with each other in a right-hand corner-to-corner way to form a two-row array. At the TTM, every four DHPRs form a tetrad that interacts with the handle regions of the RyR1 tetramer at the interstitial space, forming the RyR1-DHPR supercomplex. The tetrads of DHPR also form an ordered arrangement at the TTM; however, unlike the array of RyR1s, these tetrads are not proximal to each other because one tetrad appears every two RyR1 tetramers. Therefore, the RyR1-DHPR supercomplexes and RyR1s without bound DHPR are arranged in an alternating checkboard-like pattern within the triad junction.

DISCUSSION

In the present study, to further investigate the molecular mechanism of ECC of skeletal muscle, we resolved the native architecture of the mouse skeletal muscle triad junction at molecular resolution by using in situ cryo-ET techniques. With subtomogram analysis, the in situ structure of RyR1 in complex with FKBP12 and apo-CaM was solved to 16.7 Å and found to form a two-row array with a right-hand corner-to-corner interaction pattern on the TCM. Furthermore, DHPRs were found on the TTM, and every four DHPRs formed a tetrad that interacted with the RyR1 tetramer in an alternating way. From the in situ structure of the RyR1-DHPR supercomplex, two connection densities between RyR1 and DHPR were observed, with one located at the P1 domain of RyR1 and another located at the P2 domain. Overall, the RyR1-DHPR supercomplexes and RyR1s without DHPR bound are arranged in an alternating checkboard-like pattern within the triad junction. To understand the biological importance of the right-hand corner-to-corner arrangement of RyR1s, we performed molecular dynamics simulations and found that this unique arrangement favors strong interactions among RyR1s in the open state, thus enabling synergic calcium release within the array of RyR1s.

ECC of muscle has been proposed for many years, and a vast amount of research has been performed in decades to explore its molecular mechanism. At the triad junction, the coupling procedure includes two steps: The first step is the coupling between the activation of DHPR and that of RyR, and the second step is the coupling of the synergic calcium release of neighboring RyRs.

For the first step of coupling where DHPR is involved, two different mechanisms were proposed. One proposed mechanism was chemical coupling between DHPR and RyR2 in cardiac muscle, in which the calcium released by the voltage-gated DHPR triggers the subsequent activation of RyR2 (70). The second proposed mechanism was physical coupling between DHPR and RyR1 in skeletal muscle (29), but this suggestion lacked direct evidence. Our results included direct observation of the physical interaction between DHPR and RyR1 with two connection sites, connections 1 and 2.

The α1 subunit of DHPR plays a crucial role in the ECC process. It not only functions as a voltage sensor but also plays a potential role in delivering signals to RyR1. We predicted the complete structural models of the DHPR α1 and β1a subunits together using AlphaFold2 (71) combined with trRosetta server (72), followed with relaxation in AA simulations. It was found that α1 subunit could potentially occupy the density of connection 2 by its II-III loop between the II and III repeats and its C-terminal domain (CTD) and then make contact with RyR1 (fig. S11D). This is consistent with previous studies showing that the II-III loop might interact directly with RyR1 (38, 73) and is critical for ECC (31). It has been reported that the majority of DHPR α1 in skeletal muscle is truncated at residue 1664 of the C terminus (74), so we also predicted structure model of DHPR with a C-terminal truncated α1 subunit (fig. S11E). It showed that the truncated α1 subunit could also occupy the density of connection 2. In addition, fluorescence resonance energy transfer experiments showed that the association of the DHPR with RyR1 in resting muscle cells causes the rearrangement of the α1 subunit cytoplasmic domain including the conformational changes of II-III loop, revealing its potency for interacting with RyR1 (40).

STAC3 was predicted to be a crucial component for the interaction between RyR1 and DHPR (66, 67). Its PKC (protein kinase C–like) domain has been shown to interact with the CTD and II-III loop of the α1 subunit (75, 76). In addition, JP2 was reported as a minimum requirement for the ECC process, and its N terminus was found to bind to the last 12 residues of the α1 subunit (68, 69). Therefore, it is highly likely that STAC3 and JP2 are also included in the formation of the RyR1-DHPR supercomplex and located at the sites of connection 2.

Previous studies tried to identify the important region of RyR1 for the ECC process by constructing the RyR1-RyR2 or RyR1-RyR3 chimeric proteins, and one of them found that residues 1272 to 1455 of RyR1, which belong to the SPRY3 domain, are crucial for the formation of the DHPR tetrad (36), suggesting that its role lies in interacting with DHPR. Our study found that such a region is located near the density of connection 1, which is consistent with the previous study. In addition, another study determined that residues 1635 to 3720 of RyR1 form the region required for ECC restoration (34), which is situated near the density of connection 2 in our study. Therefore, both previous functional studies and our structural study suggest that RyR1 and DHPR engage in rich interaction modes and have multiple binding sites to ensure efficient direct signal transduction during the ECC process, which will be further validated by a higher-resolution structural study of the RyR1-DHPR complex.

The second step of coupling was proposed previously (77, 78) based on the observation of an array localization pattern of RyR1/2 channels in cardiac cells (25) and muscle tissues (59) and the observation of a spontaneous local increase in the concentration of intracellular calcium, commonly known as “Ca²⁺ sparks” (62). This coupling is important to facilitate the propagation of the gating signal of RyR1/2s and thereby trigger synchronized muscle contraction. In cardiac cells, the gating kinetics of RyR2 are believed to be related to its array-based interaction, as RyR2s tend to form large clusters in cardiac cells (65). This interaction within RyR2 arrays not only changes the gating of RyR2 from a thermodynamically reversible mode to an irreversible mode but also decreases the opening time of RyR2 channels (65). In the triad junction of skeletal muscle, RyR1 tetramers are also well organized into two parallel rows with a right-hand corner-to-corner interaction mode, suggesting a potential array-based coupling mechanism consistent with RyR2s in cardiac cells. Our study solved these array-based interactions, which enabled us to build a reliable structural model of the RyR1 array and their interaction interfaces, providing a molecular explanation of previous disease-related mutations (61), which could potentially abolish the coupled gating among RyR1s, thereby attenuating the effect of “Ca²⁺ sparks”. In addition, our MD simulations allowed us to observe the synergistic conformational changes of neighboring RyR1s in the open state, providing an in-depth understanding of the mechanism of array-based coupling by RyR1s. However, the limitation of our current MD simulations is the computational expense and challenge of investigating these large complex biological systems on a large time and space scale. Future work on the interactions and correlative motions coupling between combinations of mixed closed and open RyR1s as well as combinations of different states of RyR1 and DHPR (or with STAC3) are expected to advance the understanding on the molecular level regulation of the ECC process.

In previous studies, the in situ structure of RyR1 was resolved at near nanometer resolution using SR vesicles purified from rabbit skeletal muscle (44). While no DHPRs were observed on the TTM in their data, the contact sites between two adjacent RyR1s were reported, covering residues 2950 to 2976, 3134 to 3163, 3199 to 3212, and 3241 to 3254 in rabbits (44). To assess the consistency of these receptor-to-receptor contact sites between RyR1s under different sample preparation methods, we aligned their reported sites with our structural model through sequence alignment (fig. S12). It revealed that the contact sites they identified closely correspond to those in our model. For instance, the residues 3070 to 3113 and 2949 to 2987 of mouse RyR1 align well with our observed nearest potential contact site. This suggests that the structural model we built is relatively accurate, and the RyR1 interaction pattern is conserved across different types of samples. In the future, the higher-resolution analysis of RyR1’s in situ structure will provide more detailed information about receptor-to-receptor contact sites, leading to a deeper understanding of the coupling mechanisms of neighboring RyR1s.

The checkboard-like arrangement of RyR1-DHPR supercomplexes and RyR1s without DHPR bound is a unique architectural feature of the triad junction of skeleton muscle, and this result provides structural basis for the unique ECC process in skeletal muscle. We speculate that there might be some structural hindrances from adjacent DHPR tetrads or allosteric unfavorable conformational changes of neighboring RyR1 tetramers to prevent the formation of the RyR1-DHPR supercomplex for every RyR1 tetramer. Moreover, it is not necessary to form the RyR1-DHPR supercomplex for every RyR1 tetramer because the corner-to-corner interaction between adjacent RyR1 tetramers ensures synergistic gating once one neighboring RyR1 tetramer is activated. In addition, the coupling efficiency by the physical corner-to-corner interaction would decrease along the propagation distance because the gating signal could not be amplified by the physical coupling in terms of the energy view. Therefore, the arrangement of DHPR tetrads for every two RyR1 tetramers is important to facilitate a fast and efficient ECC process and achieve spontaneous activation of RyR1s in the triad junction to ensure synergistic contraction of the entire fibril.

In conclusion, our study elucidated the inherent architecture of triad junctions in mammalian skeletal muscle and included the direct observation of the arrangement of RyR1 and the RyR1-DHPR supercomplex within the triad junction. This architecture is suspected to be related to the fast and efficient ECC process in skeletal muscle: A long and narrow triad junction network spreading over the long skeletal muscle fibers guarantees rapid signal transmission. Once a DHPR tetrad in the T-tubule is activated by voltage potential, it activates the RyR1 underneath instantly through physical coupling; the activated RyR1 then activates the neighboring RyR1s instantly, also through physical coupling. In this way, a vast amount of calcium is released from the SR, and muscle contraction is induced. This intricate and ingenious construction of triad junctions ensures rapid signal transmission during the ECC process, thereby preserving the normal physiological functions of skeletal muscle.

MATERIALS AND METHODS

Extraction and vitrification of mouse skeletal muscle fibers

The EDL muscles were separated from mice (Charles River, KM mouse, male, 2- to 3-week-old) and incubated in Dulbecco’s modified Eagle’s medium (DMEM) [high glucose, l-glutamine with sodium pyruvate (110 mg/liter)] containing 0.2% collagenase at 37°C for digestion. During this time, the condition of muscles should be regularly checked to avoid overdigestion. After ~30 min, when the muscles began to loosen, they were transferred to a dish containing DMEM without collagenase to stop the digestion. Then, the pipette was used to gently flush the muscles with medium until the muscle was naturally released, and the live muscle fibers were spread in the medium. During this process, the flushing time when the sample was exposed to room temperature was less than 10 min. Then, the dish was transferred into an incubator for incubation at 37°C and 5% CO₂ for at least 5 min to keep the muscle fibers alive before another round of flushing.

Before plunge freezing, the isolated single muscle fiber was incubated in medium containing 10% glycerol for 3 to 5 min to avoid incomplete vitrification. Then, 3 μl of medium containing one isolated muscle fiber was placed on the glow-discharged grid (Quantifoil R2/1 Au 200 mesh). After incubation for 60 s at 37°C, the grid was blotted from the opposite side for 10 s using EMGP2 (Leica) and then plunge-frozen into liquid ethane.

The animal experiments were performed in the Laboratory of Animal Center of Institute of Biophysics, Chinese Academy of Sciences, in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and according to guidelines approved by the Institutional Animal Care and Use Committee at Institute of Biophysics. The reference number for approval is SYXK2022139.

Cryo-lamella preparation using cryo-FIB

The vitrified grid containing the isolated muscle fiber was clipped into a special AutoGrid (Thermo Fisher Scientific) designed for cryo-FIB milling. After being loaded into the transfer shuttle (79), the grid was transferred into a cryo-FIB/scanning electron microscopy (SEM) dual-beam microscope Aquilos2 (Thermo Fisher Scientific). Before milling, the grid was sputter-coated with platinum for 10 s to reduce the charging effects of the sample. Then, it was injected with organometallic platinum for 12 s to prevent damage by gallium ion beam milling. The grid was pretilted 8° to expose the muscle fibers on the grid. The sample was milled to ~300 nm following the regular steps. Then, the cryo-lamella was polished further using a 10-pA current and examined using cryo-SEM images to estimate the thickness by charging propensity. The process was continued until the thickness reached ~120 nm.

Tilt series acquisition

The grids were inserted into the autogrid cassette at an angle that was 90° rotated relative to the angle when inserted into the cryo-FIB/SEM microscope to ensure that the lamellae milling direction was perpendicular to the tilt axis of the electron microscope stage so that the nearby sample that was not milled would not block the lamellae during rotation. Afterward, the cassette was loaded into a transmission electron microscope Titan Krios G3 (Thermo Fisher Scientific) equipped with a zero-loss energy filter and a K2 Summit direct electron detector (Gatan). Serial EM software (80) was used to acquire images. Target lamellae containing areas of interest were mapped at a low magnification of ×6300 to check and select the collection sites. Tilt series were acquired at ×64,000 magnification (pixel size of 2.22 Å) with a dose-symmetric tilt scheme using the beam-shift method (81). The total dose was set to 140 to 160 e⁻/Ų, and the tilt range was approximately −40° to +40° relative to the lamella pretilt angle (8°) with an increment of 2°. The actual range may be adjusted according to the condition of each lamella. The defocus was set to range between 4 and 5 μm.

Image processing

Raw images of the tilt series were sent to Warp (82) for motion correction and CTF estimation. After manually filtering out the bad tilt images, the preprocessed tilt series were aligned using the patch-tracking method included in the IMOD software package (83). Then, the alignment parameters were sent back to Warp, and the tomograms were reconstructed. To visualize the densities of interest more clearly, we performed nonlinear anisotropic diffusion (NAD) filtering on the tomograms (84). Each RyR1 particle center was manually picked in Dynamo (85), and its initial orientation was determined based on the vector between the manually selected RyR1 transmembrane region and cytoplasmic region.

A total of 4379 particles were picked from 264 tomograms. Then, relion3.1 (86) was used for the STA assay. First, the subtomogram particles in a box size of 48³ voxels were extracted from 8× binned tomograms in Warp. The direct reconstruction map using predetermined orientations was used as the initial reference for alignment. C4 symmetry was always applied, and a mask surrounding the cytoplasmic region of RyR1 was used to help align. After global refinement, a 3D classification was applied, and a better class containing 51% of the particles (2241) was selected for subsequent refinement. After re-extracting from 2× binned tomograms with a box size of 128³ voxels, a local refinement was performed, and a 3D classification followed. Then, 61% of particles (1365) in a better class were used for final local refinement, and the resolution was estimated 16.7 Å based on the gold-standard FSC with a 0.143 criterion after postprocessing.

For the RyR1-DHPR supercomplex, 1421 particles were manually picked from 220 tomograms. First, a box size of 48³ voxels was used to extract the particles from 8× binned tomograms in Warp. The low-pass–filtered RyR1 density map was used as the initial reference for alignment. Both C4 and C1 symmetry were applied during the process, confirming the four extruding densities for DHPR. A 3D classification without a mask was performed after global refinement, and a class containing 18% particles (251) was selected for local refinement. This density map contained four extruding densities assumed to be DHPRs. Then, all 1421 particles were extracted using a box size of 96³ voxels from 4× binned tomograms. After local refinement, a mask surrounding the RyR1 and DHPR α2δ subunit densities was used in the 3D classification procedure, and a class containing 11% particles (154) was selected from eight classes. The density map containing one RyR1 and four DHPRs was lastly obtained after local refinement and postprocessing.

Model building

The preliminary model of mouse RyR1 was forecasted using the SWISS-MODEL homology modeling method (87). Alphafold2 was used to predict the partial structure of RyR1 because the sequence exceeded the maximum allowed by homology modeling. Subsequently, the complete structure of RyR1 was finalized using Coot software (88). The models of RyR1, FKBP12, and CaM are fitted in the reconstruction map using chimera (89).

Then, we performed stepwise MDFF simulations to refine the complex model according to the density map. A time step of 1 fs was used throughout the simulation. Langevin dynamics were adopted at a temperature of 310 K. The equilibration step for energy minimization was performed on the initial model for 1000 steps before the refinement run. The refinement runs were performed for 3000 ps, corresponding to 3,000,000 simulation steps, and the gridForceScale values were gradually increased from 0.3 to 0.7 during the refinement. All simulations were performed using CHARMM36m force fields (90). Electrostatic calculations were treated with particle mesh Ewald (PME). A cutoff of 12 Å was chosen for short-range van der Waals interactions. NAMD (91) was used as the MD engine throughout all simulations.

Molecular dynamics simulation

The open and closed RyR1 structures were constructed using homology modeling in Maestro (Schrödinger, LLC) with the sequence information and major templates displayed in table S2; missing segments were built using the trRosetta server (72). The membrane builder in CHARMM-GUI (92) was used to build the protein-membrane complex system. To mimic the endoplasmic reticulum membrane, the membrane for RyR1 consisted of 54% phosphocholine, 18% phosphatidylethanolamine, 7% cholesterol, 5% phosphatidylserine, 8% phosphatidylinositol, and 5% N-palmitoylsphingomyelin in the molar ratio (93). The protein-membrane complex was solvated in an aqueous phase containing TIP3P water molecules, K⁺ or Cl⁻ as counter ions, and 0.05 M CaCl₂ in a periodic box. A system with one RyR1 was first constructed and equilibrated and then further built into a two-RyR1 system using Visual Molecular Dynamics (VMD) program (94). AA RyR1 systems contain up to 6 million atoms, which are computationally expensive to simulate on a long time scale. AA simulations were applied to explore the interactions and motions of two-RyR1 assembled by intermolecular helical-helical domain interactions, while CG simulations were used to explore the assembly of the four-RyR1 system.

For CG simulations, we used the Martini Bilayer Maker in CHARMM-GUI (92) to construct the protein-membrane complex systems, each of which contains CG models of ion channels, a CG bilayer with similar lipid components as in the AA systems, the nonpolarizable water models, Na⁺ or Cl⁻ as counter ions, and 0.05 M NaCl. A system with one RyR1 was first constructed and equilibrated, which was further built into a four-RyR1 system using VMD (94). A summary of the simulation systems is provided in table S3.

AA simulations were performed with the CHARMM36m-cmap force field (95). Each system was subjected to energy minimization, 2-ns multistep equilibration under isothermal-isovolumetric (NVT) and isothermal-isobaric ensemble with semi-isotropic pressure coupling (NPγT) with the force constant of dihedral restraint on proteins and lipids gradually reduced from 250 to 0 kcal/mol in six steps. Then, the system continued to relax in MD simulations under the NPγT ensemble (310 K, 1 bar, Langevin dynamics thermostat and semi-isotropic Monte Carlo barostat). A time step of 2 fs was used for the initial 7-ns relaxation and then increased to 3 fs under hydrogen mass repartitioning. All lengths of bonds to hydrogen atoms in protein or lipid molecules were constrained with SHAKE algorithm. The PME technique (96) was used to calculate long-range electrostatic interactions. The van der Waals and short-range electrostatics were cut off at 12.0 Å with a switch at 10.0 Å. The CG simulations were carried out with Martini 2.2 with an elastic network (elnedyn22) force field using the Gromacs 2019 program (97) with a time step of 20 fs under the NPγT ensemble (310 K, 1 bar, v-rescale thermostat and Parrinello-Rahman barostat).

Trajectory analyses such as distances, short-distance contacts, hydrogen bonds, PCA, and dynamic cross-correlation map (DCCM) were performed with “cpptraj” (98) and then further processed and plotted using matplotlib (99). Structures were shown by PyMOL (Schrödinger, LLC), and movies were generated in VMD (94). Solvent-accessible pore radii along a channel were calculated by HOLE (100) and visualized and rendered in VMD. The intermolecular helical-helical domain with a minimum distance less than 8 Å at the end of 3-μs CG simulations is considered to form a right-hand contact. We counted the contact numbers from 10 copies. The numbers of all contacts (<6 Å), H-bond pairs, and hydrophobic contacts (<6 Å) were counted from the last 110 ns of all three AA simulations for the open and closed RyR1-RyR1 systems. For both the PCA and DCCM analyses, we used the trajectories from the last 110 ns of all three copies. We first aligned the Cα of the interacting RyR1-RyR1 structure on the first frame by the sequences that are largely nonloop (i.e., 12 to 1297, 1432 to 1871, 1924 to 2042, 2092 to 2826, 2852 to 4255, 4541 to 4576, and 4639 to 5035 of each RyR1 monomer) to generate an averaged coordinate, on which we aligned again and then used it for the PCA and DCCM analyses. The cross-correlation matrix elements, C_ij, are defined as (101)

$$C_{\mathit{ij}}=\frac{C_{\mathit{ij}}}{C_{\mathit{ii}}^{1/2}{C}_{\mathit{jj}}^{1/2}}=\frac{\langle r_i{r}_j\rangle -\langle r_i\rangle \langle r_j\rangle }{{[(\langle r_i^2\rangle -{\langle r_i\rangle }^2)(\langle r_j^2\rangle -{\langle r_j\rangle }^2)]}^{1/2}}$$

where i and j represent two atoms with the position vectors in the structure at time t as r_i(t) and r_j(t), respectively; angle brackets denote time averages; and c_ij represents the corresponding covariance matrix element.

$$C_{\mathit{ij}}=\langle (r_i-\langle r_i\rangle )(r_j-\langle r_j\rangle )\rangle =\langle r_j{r}_j\rangle -\langle r_i\rangle \langle r_j\rangle$$

The values of the cross-correlation coefficients range from −1.0 (completely anticorrelated motions) to 0.0 (no correlated motions) to 1.0 (completely correlated motions). The DCCM of the two interacting RyR1s at the contact interface or most distant RyR1s was studied. We calculated the first principal component of residues 2494 to 3611 (which contains the P2 domain and HD2 domain) from dimerized RyR1 in both open and closed states.

Supplementary Materials

This PDF file includes:

Figs. S1 to S12

Tables S1 to S3

Legends for movies S1 to S9

References

Other Supplementary Material for this manuscript includes the following:

Movies S1 to S9