Ryanodine Receptors: Structure and Function*

Ryanodine receptors (RyRs) are huge ion channels that are responsible for the release of Ca2+ from the sarco/endoplasmic reticulum. RyRs form homotetramers with a mushroom-like shape, consisting of a large cytoplasmic head and transmembrane stalk. Ca2+ is a major physiological ligand that triggers opening of RyRs, but a plethora of modulatory proteins and small molecules in the cytoplasm and sarco/endoplasmic reticulum lumen have been recognized. Over 300 mutations in RyRs are associated with severe skeletal muscle disorders or triggered cardiac arrhythmias. With the advent of high-resolution structures of individual domains, many of these can be mapped onto the three-dimensional structure.

described as a mushroom, with a large cap representing ϳ80% of the volume located in the cytoplasm and the stalk crossing the membrane into the SR/ER lumen. The transmembrane region measures 120 ϫ 120 ϫ 60 Å, whereas the cytoplasmic area measures ϳ270 ϫ 270 ϫ 100 Å. These two major parts are connected via four thick columns. An interesting and important feature is that the RyR cytoplasmic head does not form a rigid block. Instead, there are many solvent-filled cavities and numerous globular masses that may correspond to individual or groups of folded domains. To aid with the structural description of RyRs, several portions have received names, including "clamps," "handles," and a "central rim" that surrounds a central cavity (Fig. 1). The globular portions have received identifying numbers and are often referred to as "subregions" (Fig. 1) (23,25). Although most cryo-EM studies have been performed on RyR1, there are reconstructions for RyR2 and RyR3 as well, albeit at much lower resolutions. These show that the overall shape is very similar for all RyR isoforms (26,27).
There has been much debate about the number of transmembrane helices, but the overall consensus now is that there are either six or eight segments per subunit (28). Five or six of these can be detected in the cryo-EM maps (23,24). The inner helices create the pore-forming region, and sequence homology suggests an arrangement similar to various tetrameric ion channel structures. A major point of interest lies with the motions the channel undergoes during opening and closing. Based on different cryo-EM reconstructions of RyR1, there is still some debate about the movements the inner helices undergo during channel opening (22)(23)(24). A systematic investigation of the open and closed states at 10.2 Å seems to indicate that the inner helices kink, thus widening the pore during channel opening (22). A 9.6 Å cryo-EM structure of RyR1, reported to be in the closed state, shows that these helices are already kinked and thus suggests an alternative mechanism for channel opening (24). It is of course possible that the extraction and purification conditions in this latter study were not favorable for a closed state, but higher resolution studies will be needed to shed light on this matter. It is also clear from these studies that the RyR is a bona fide allosteric protein, as channel opening results in substantial structural rearrangements in the cytoplasmic portion. The most prominent motions occur near the cen-tral rim and the clamp regions ( Fig. 2) (22). The movements in the transmembrane and cytoplasmic portions are most likely transmitted via the columns. Due to this allosteric coupling, binding of ligands or auxiliary proteins to any mobile portion of the cytoplasmic region can influence the ability of the pore region to open.
In a physiological context, many RyRs will not be found in isolation but rather in a cluster with neighboring channels. An interesting feature of the RyR is that it can form regular arrays at the SR-plasma membrane junctions. Using purified protein, it has been shown that RyR1 can form planar crystalline arrangements, forming checkerboard patterns in the absence of any other protein (29). Two-dimensional crystallization experiments show that subregion 6 in the clamp region is responsible for the interprotein contacts (30). As the clamp region has been shown to undergo substantial motions during opening and closing (22), motions in one RyR channel can thus be transmitted to neighboring RyRs. This may underlie the phenomenon of coupled gating, whereby opening of one channel can induce opening in neighboring channels through physical interactions (31). Although the ability to form two-dimensional crystals holds promise for improved resolutions, so far these have not yet exceeded the single-particle cryo-EM studies.

Regulators
A Ca 2ϩ -selective pore could be made with just a few transmembrane ␣-helices. Why is it then that we have evolved Ͼ2-MDa giants to fulfill the same function? The answer undoubtedly lies with regulation: as Ca 2ϩ is a very potent messenger, its entry into the cytoplasm through RyRs is tightly controlled by a myriad of proteins, small molecules, and post-translational modification events that affect opening or closing of the channels. By providing a large cytoplasmic mass and many solvent-filled cavities, each RyR yields ϳ500,000 Å 2 of surface area (transmembrane area included) onto which many regulators can dock. In addition to the Ca 2ϩ and voltage-gated calcium channels mentioned above, these modulators include binding partners in both the cytoplasmic and SR/ER luminal portions that can provide either positive or negative input signals (Fig.  3A). The list is too extensive to cover every binding partner in detail, but several excellent reviews have provided comprehensive overviews (6,32). Here, I describe a number of regulators that have received a lot of attention.
Well known binding partners of RyRs are FK506-binding proteins (FKBPs). Named according to their molecular mass, both FKBP12 and FKBP12.6 can associate with all three RyR isoforms (33). The affinity seems to be quite strong, as FKBP12 co-purifies with RyR1, and FKBP12.6 co-purifies with RyR2. They stabilize the closed state of the channels and prevent the formation of subconductance states in RyR1 (34). Using cryo-EM studies, FKBPs were found to bind in a site near subdomains 3, 5, and 9 (35,36), and this was confirmed via FRET studies (37,38  Other well studied regulators are EF-hand-containing proteins that can bind Ca 2ϩ . Calmodulin (CaM) can associate with the RyR under both apo-and Ca 2ϩ -loaded conditions and finetune the effect of Ca 2ϩ . The functional effect of CaM depends on both the Ca 2ϩ concentration and the RyR isoform. At high Ca 2ϩ levels, CaM can inhibit both RyR1 and RyR2. At low Ca 2ϩ levels, however, it activates RyR1 but inhibits RyR2 (40 -43). Many studies have focused on identifying the sequences in RyRs that can bind CaM. Currently, there is one crystal structure available for Ca 2ϩ /CaM bound to a peptide from RyR1 (residues 3614 -3643). In this case, both CaM lobes wrap around an amphipathic ␣-helix in an antiparallel arrangement, with the N-terminal lobe bound to the C-terminal half and the C-terminal lobe bound to the N-terminal half of the helix (44). However, several other RyR peptides are found to bind CaM (e.g. Ref. 45), and it is likely that the Ca 2ϩ /N-terminal lobe binds a different segment in intact RyRs (44,46). Interestingly, the same RyR1 peptide can bind the EF-hand protein S100A1, which can enhance opening of both RyR1 and RyR2 (47). An NMR structure shows that its binding site overlaps with the Ca 2ϩ /CaM-binding site (48), and it is therefore likely that S100A1 acts by displacing CaM at high Ca 2ϩ levels, thus abolishing the latter's inhibitory effects. Sorcin is yet another EFhand-containing protein that can associate with RyR2 at elevated Ca 2ϩ concentrations and mediate inhibition (49).
Using cryo-EM reconstructions, the binding of both apo-CaM and Ca 2ϩ /CaM has been visualized on intact RyRs (50). These show that the location of CaM changes upon binding Ca 2ϩ , in accordance with the differential functional effects of CaM at various Ca 2ϩ levels (Fig. 3B). FRET-based measurements between CaM and FKBP suggest that the N-terminal lobe is closer to FKBP than the C-terminal lobe (37). As both CaM and FKBP are located in a region of the channel that undergoes substantial motions upon opening and closing (22), these allosteric modulators likely act by stabilizing or destabilizing individual states.
Calsequestrin is a major Ca 2ϩ -buffering protein in the SR lumen. It can form oligomers and interact with the membrane-associated proteins junctin and triadin. Together, this complex is thought to either increase or decrease the RyR activity depending on the calsequestrin isoform, although little is known about the exact molecular mechanisms (51).

Phosphorylation
RyRs are the target for several kinases (PKA, PKG, and Ca 2ϩ / CaM-dependent protein kinase II (CaMKII)) and phosphatases (PP1, PP2A, and PDE4D3). Some of these enzymes are anchored to RyRs via scaffolding proteins, allowing for specific and compartmentalized regulation (32). Phosphorylation by PKA has received a lot of attention, as it forms the link between physiological stress and RyRs via activation of ␤-adrenergic receptors. PKA targets several cytoplasmic proteins, and at least two RyR residues (Ser-2843 in human RyR1 and Ser-2030 and Ser-2808 in RyR2) can be phosphorylated by PKA. However, a great deal of controversy exists about which phosphorylation event predominates and the role of this in heart failure. On the one hand, Ser-2808 was found to be hyperphosphorylated by PKA in heart failure, leading to the dissociation of FKBP12.6. As the latter stabilizes the closed state, the dissociation would then lead to enhanced activity of the RyR (52). However, several groups have failed to detect PKA hyperphosphorylation in heart failure and did not observe a dissociation of FKBPs by RyR phosphorylation (e.g. Refs. 53 and 54). For example, one report found that Ser-2808 is already highly phosphorylated in the basal state and that instead Ser-2030 is the major PKA target residue in RyR2 (55). Most likely, much of the controversy has to do with details in the preparations. For example, RyRs are sensitive to redox conditions, and a different amount of oxidation in the various studies could therefore influence the affinity for FKBPs (56).
CaMKII is regulated by intracellular Ca 2ϩ concentrations through CaM. Like PKA, it can phosphorylate Ser-2843 in RyR1 and Ser-2808 in RyR2 but also seems to have a unique phosphorylation site in RyR2 (Ser-2814). CaMKII was found to increase the open probability and Ca 2ϩ sensitivity of the channel (57) and has also been shown to contribute to cardiac arrhythmia and contractile dysfunction (58).

Disease Mutations
Although the roles and mechanisms of RyRs in heart failure are still under scrutiny, it has become clear that mishandling of Ca 2ϩ in the cytoplasm due to mutations in the ryr genes can give rise to severe conditions. No disease phenotype has been associated with RyR3 mutations, but both RyR1 and RyR2 have been linked to a number of genetic diseases that are due mostly to their prominent role in muscle contraction (59 -61). Malignant hyperthermia (MH) is a pharmacogenetic disorder, characterized by muscle rigidity and fatal rises in body temperature (62,63). It is typically triggered by the combination of a RyR1 mutation and an external compound such as a volatile anesthetic or succinylcholine, a muscle relaxant. In some cases, stress may serve as an alternative external trigger (64). In pigs, the RyR1 mutation R615C was found to cause the related porcine stress syndrome (65,66), and the corresponding mutation in humans was soon found to underlie MH (67). In a MH event, an excessive leak of Ca 2ϩ from the SR results in a hypermetabolic state, depleting the ATP pool and leading to acidosis. Dantrolene is a clinically approved drug to treat MH and acts by decreasing the intracellular Ca 2ϩ concentration (68). Several studies suggest a direct interaction between dantrolene and RyR1 (69,70). Dantrolene inhibits RyR1 even when expressed in a heterologous system such as HEK293 cells and appears to inhibit SOICR (71). However, the single-channel behavior of RyR1 incorporated in planar lipid bilayers appears to be unaffected, and the precise mechanism of action is therefore still unknown.
In addition to MH, RyR1 mutations can cause congenital myopathies such as central core disease (CCD) (72,73). The latter is characterized by cores of metabolically inactive tissue in the center of muscle fibers, which can lead to muscle weakness. In contrast to MH, CCD does not require an external trigger. Although RyR1 is abundant in skeletal muscle, it is also widely expressed in the brain, and a CCD mutation was recently shown to have a neuronal phenotype (74). Other defects associated with RyR1 mutations include heat/exercise-induced exertional rhabdomyolysis (75), multiminicore disease (76), and atypical periodic paralysis (77).
Because RyR2 plays a major role in cardiac excitation-contraction coupling, mutations in this isoform can give rise to cardiac arrhythmias. Catecholaminergic polymorphic ventricular tachycardia (CPVT) is a condition often triggered by emotional or physical stress, leading to bidirectional ventricular tachycardia that may result in sudden cardiac death (78). A popular model assumes that the leakage of Ca 2ϩ through RyR2 increases the activity of the Na ϩ /Ca 2ϩ exchanger, which allows for the electrogenic influx of three Na ϩ ions in exchange for efflux of one Ca 2ϩ ion, resulting in delayed afterdepolarizations (59). In addition to CPVT, RyR2 mutations associate with arrhythmogenic right ventricular dysplasia, in which the right ventricular muscle is gradually replaced by fibrofatty deposits (79). As RyR2 is also widely expressed in the brain, there may be neuronal effects of the mutations as well, and a mouse model has shown that RyR2 mutations can also give rise to seizures (80).
Since the initial links between RyR mutations and disease were established, Ͼ300 disease mutations have been identified. In both RyR1 and RyR2, most of these cluster in three or four different "hot spots," located in the N-terminal region (first ϳ600 amino acids), a central region (amino acids ϳ2100 -2500), and the C-terminal area (amino acid ϳ3900 -end) (Fig.  4A) (60,81). However, the appearance of clusters may be due, in part, to sequencing bias, and mutations are increasingly found outside of these three segments. The C-terminal portion encodes the transmembrane area, and a subset of mutations could thus interfere directly with the passage of Ca 2ϩ . However, this third hot spot also encodes a substantial amount of cytoplasmic components. Interestingly, many CCD mutations are found in the C-terminal region.
A general observation is that most disease mutations cause a gain of function, although there are exceptions (82). These mutations then lead to premature or prolonged release of Ca 2ϩ in the cytoplasm. It has been shown that CPVT mutations in RyR2 lower the threshold for activation by luminal Ca 2ϩ , thus affecting SOICR (16,83). In addition, many studies have shown that the mutations increase the sensitivity of the channels to activating agents (e.g. Refs. 84 and 85). One hypothesis suggests that there is a direct interaction between the N-terminal and central hot spots and that this interaction is "unzipped" during channel opening through allosteric coupling (86,87). Disease mutations clustered at this interface would thus weaken the interaction and facilitate channel opening. This theory is based on spectroscopic measurements and the ability of peptides to interfere with channel function, but a direct interaction between individually folded domains of the N-terminal and central hot spots has not yet been shown. Another theory, similar to the effect of PKA phosphorylation, proposes that the disease mutations weaken the interaction with FKBPs, and the resulting dissociation of FKBP then leads to increased singlechannel activity under conditions that simulate stress (88). Despite the differences, the combined data converge on a gain of function for most disease mutations, with a concomitant increased sensitivity toward cytoplasmic or luminal activation mechanisms and decreased binding of modulators that preferentially bind to the closed state.

Crystal Structures and Domain Architecture
Despite the availability of high-quality cryo-EM maps, the current resolution is too small to pinpoint the location of individual amino acids. In trying to match primary with tertiary structure, many studies have benefited from raising antibodies against particular stretches of the RyR sequence or from inserting GFP at various locations in the sequence. Cryo-EM reconstructions were then made for the RyR-antibody complexes and fusion proteins, which created restraints on the position of the sequence in the three-dimensional structure (for a review, see Ref. 21). Using this approach, it was shown, for example, that insertions in the N-terminal and central disease hot spots both yield difference density in the clamp region, suggesting that these segments may be in close proximity (89), in agreement with the zipper hypothesis.
Although no well diffracting crystals have been reported for the intact RyR, four reports have provided high-resolution information on individually folded RyR domains, corresponding to the N-terminal domains of RyR1 (residues 1-205) (90,91) and RyR2 (91,92) and a larger N-terminal region of RyR1 (residues 1-559) (93). The latter study represents the bulk of the N-terminal disease hot spot and shows that it is built up by three individually folded domains (A-C) that form a compact arrangement (Fig. 4B). Domains A and B form ␤-trefoils, containing 12 ␤-strands each, whereas domain C consists of a fivehelix bundle. Docking of this hot spot in several RyR1 cryo-EM maps shows that it is located in the cytoplasmic portion, forming a vestibule around the 4-fold symmetry axis (Fig. 4, C and D). Although this position may seem at odds with a GFP insertion study (89), the data can be reconciled when the length of the linkers for the fusion protein is taken into consideration (93).
Over 55 disease mutations (RyR1 and RyR2 combined) can be located in the crystal structure and pseudo-atomic model. Most of these are found at domain-domain boundaries, either in between the three domains or at interfaces with neighboring RyR domains. This suggests that some domain interactions may be disrupted during channel opening and that mutations at such functional interfaces facilitate the process by weakening the contacts. Six mutations were found to be buried within individual domains, and these could alter the local folding of the domains, thereby having a less specific effect. Most of the mutations are at interfaces with other N-terminal hot spot domains, either within or across subunits. The zipper hypothesis, involving interactions with the central hot spot region, can therefore apply to only less than one-third of the N-terminal mutations.
Most disease mutations consist of point substitutions, but a very severe form of CPVT is caused by deletion of the entire third exon of RyR2, consisting of 35 amino acids (94). This exon encodes a ␤-strand and ␣-helix in RyR2 domain A, but rather than causing misfolding, the deletion increases the thermal stability of the domain (91,92). A crystal structure of the mutant domain shows that this is accomplished by insertion of a flexible loop, encoded by the fourth exon, into the ␤-trefoil core (92). In non-diseased individuals, this may represent a case of alternative splicing, whereby two short stretches "compete" for a ␤-strand position. This allows for the formation of an alternative N-terminal domain, which further fine-tunes the RyR activity.
RyRs also encode tandem repeats that are present at least two times in the sequence (Fig. 4A). Crystal structures of this domain in the central region of RyR1, RyR2, and RyR3 show a mostly ␣ helical structure with 2-fold symmetry (Fig. 4E) (95). The repeats are separated by a flexible loop that contains up to 7 different phosphorylation target sites, including the wellstudied Ser-2808 and Ser-2814 sites in RyR2. In addition to the crystallized portions, the RyR genes encode a number of pre-dicted cytoplasmic domains (Fig. 4A). This includes three SPRY domains encoded in the N-terminal one-third of the sequence. SPRY domains were first detected in the SplA kinase and in RyRs. The core of the SPRY structure consists of a sandwich formed by two four-stranded antiparallel ␤-sheets. They generally form protein-protein interaction domains and thus may serve as docking sites for auxiliary proteins in RyRs. The second SPRY domain in RyR1 has been suggested to form a docking site for the Ca V 1.1 II-III loop (96). Tandem EF-hands, forming likely Ca 2ϩ -binding sites, are encoded closer to the C terminus (97).

RyR and the Inositol 1,4,5-Trisphosphate Receptor
The ER membrane is home to another Ca 2ϩ release channel, modulated by inositol 1,4,5-trisphosphate (IP 3 ). Although much smaller (ϳ1 MDa), this IP 3 receptor (IP 3 R) also forms a tetrameric assembly and shares many structural features with RyRs, predominantly in the transmembrane and N-terminal regions. A recent cryo-EM structure at ϳ9.5 Å of IP 3 R isoform 1 displays an overall mushroom shape, although with a smaller cap than for the RyR (98). Several crystallographic studies on the N-terminal region of IP 3 Rs show a striking similarity to domains A-C in RyRs (99,100). IP 3 R domain A (also known as the "suppressor domain") and domains B and C (together known as the IP 3 -binding core) display a similar overall arrangement and domain interactions as in the RyR1 ABC structure. Docking in the 9.5 Å IP 3 R cryo-EM map yields a similar position, whereby four N-terminal ABC domains form a continuous ring around the 4-fold symmetry axis (99). Binding of IP 3 between domains B and C causes a rearrangement of the three N-terminal domains, with domain A moving ϳ3-4 Å toward domains B and C (99,100). An overall theme thus seems to emerge in both Ca 2ϩ release channels. Affecting the interfaces between the ABC domains, either by ligand binding in IP 3 Rs or by disease mutations in RyRs, seems to facilitate channel opening in both receptors. The overall similarity in structure and mechanisms between both channels is further corroborated by the fact that both domain A and the transmembrane area can be functionally swapped (99).

Conclusion
Much remains unknown about the molecular mechanisms that underlie opening and closing of the RyR. In the absence of full-length crystal structures, the construction of pseudoatomic models holds much promise. Smaller domains may prove more cumbersome to dock reliably, but together with restraints provided by difference cryo-EM and FRET measurements, there is great promise to solve this three-dimensional puzzle in the years to come.