Summary from functional studies on RyR disease-associated mutations. Mutations that alter RyR functions are listed as hyperactive or hypoactive. Mutations with no discernable change compared with wild type channel are listed as “uncertain.” References to the functional studies are listed beside the corresponding mutations. (Δ) = deletion mutation. (/) = double mutation. Amino residue numbering is for human RyRs. (MH) = Malignant hyperthermia, (CCD) = central core disease, (MmD) = multi-minicore disease, (CNMDU1) = congenital neuromuscular disease with uniform type 1 fiber, (CPVT) = catecholaminergic polymorphic ventricular tachycardia, (ARVD2) = arrhythmogenic right ventricular dysplasia type 2, and (IVF) = idiopathic ventricular fibrillation.
Ca2+ ions form important intracellular messenger molecules in nearly every cell type of the human body. Under so-called “resting“ states of the cell, the concentration of Ca2+ in the cytoplasm is extremely low (~10-7M), but this can rapidly rise ~2 orders of magnitude when an appropriate signal is generated. Ca2+ ions can enter the cytoplasm either from the extracellular space or from intracellular compartments, through specialized membrane proteins. These "calcium channels" are complex proteins, often consisting of multiple subunits, and are targets for multiple regulatory events.
A major intracellular Ca2+ store is the endoplasmic reticulum (ER), or its specialized form, the sarcoplasmic reticulum (SR) in muscle tissue. The mobilization of Ca2+ from these organelles can trigger many Ca2+-dependent events, such as the contraction of muscle tissue. The coupling between electrical excitation of a muscle cell and its subsequent contraction (excitation-contraction coupling, E-C coupling) is a finely orchestrated process that requires a functional cross-talk between proteins embedded in the plasma membrane and in the SR membrane (Figure 1). An electrical signal, a depolarization of the plasma membrane, can lead to opening of L-type voltage-gated calcium channels (CaV), which results in the influx of Ca2+ into the cytoplasm. In the SR membrane, large calcium release channels called Ryanodine Receptors (RyRs) can detect this initial influx, and release more Ca2+ in a process known as Ca2+-induced Ca2+ release (CICR). The latter event provides the bulk of the Ca2+ required for contraction to occur. In this scenario, the RyRs form signal amplifiers, which both detect and augment the Ca2+ signal [1, 2]. This scenario likely predominates in cardiac tissue, but Ca2+ is not always required to open RyRs. In skeletal muscle, for example, various pieces of evidence suggest a direct interaction between the L-type calcium channel (CaV1.1) and the RyR [3-6]. In this case, the depolarization of the plasma membrane causes distinct conformational changes in CaV1.1, which are then transmitted directly to the RyR, causing them to open.
RyRs owe their name to the binding of ryanodine, a toxic compound from the South American plant Ryania speciosa . RyRs are currently the largest known ion channels, with a massive size of >2.2MDa. They form homotetrameric assemblies [8, 9], and each monomer contains ~5000 amino acid residues. In humans and all other mammalian organisms investigated so far, three different isoforms (RyR1-3) have been found to date. RyR1 is widely expressed in skeletal muscle, and was the first one to be cloned [10, 11]. RyR2 is primarily found in the heart [12, 13] and RyR3 was originally identified in the brain , although each isoform is found in many different cell types .
Since Ca2+ ions form very potent intracellular messenger molecules, it is not surprising that their entry into the cytoplasm is under intense regulation. RyRs therefore form the targets for a plethora of auxiliary proteins and small molecules that are known to regulate their ability to open or close [15-17]. The primary regulatory molecule is cytoplasmic Ca2+, which triggers the channel to open. However, higher Ca2+ levels can trigger closing, indicating that there is more than one Ca2+ binding site. Under conditions whereby the SR is overloaded with Ca2+, RyRs can also open spontaneously in a process known as store-overload-induced calcium release (SOICR) [18, 19].
An overview of several positive and negative RyR modulators is shown in Figure 2. Well-known regulators include FK506 binding proteins (FKBPs), small proteins that stabilize the closed state of the channels and can prevent the formation of subconductance states [20, 21]. Calmodulin (CaM) is a ~17kDa protein that can bind 4 Ca2+ ions in separate EF hands, well-known Ca2+-binding motifs. CaM binds RyR in a 4:1 stoichiometry and can either inhibit or stimulate RyRs depending on the isoform and Ca2+ levels [22-25]. RyRs are also the target for several kinases (PKA, PKG, CaMKII) and phosphatases (PP1, PP2A, PDE4D3), and the degree of phosphorylation seems to affect RyR activity . In this chapter, we discuss the involvement of RyRs in several genetic diseases, summarize the outcome of several years of functional studies on disease variants of the RyR, and describe the insights into disease mechanisms through low- and high-resolution structural studies.
2. Ryanodine receptors and genetic disease
2.1. Malignant hyperthermia
Malignant hyperthermia (MH) is a pharmacogenetic disorder, requiring both a genetic mutation and an external trigger to cause a disease phenotype. The condition is mostly linked to mutations in the RyR skeletal muscle isoform (RyR1), but some mutations in the skeletal muscle L-type calcium channel (CaV1.1) can also cause MH . It is typically triggered by the use of inhalational anesthetics or succinylcholine, a muscle relaxant . The condition is considered rare, with an incidence of 1:60,000 for adults or 1:15,000 for children . An MH episode is characterized by an abnormal rise in the core body temperature, skeletal muscle rigidity, acidosis, and tachycardia. Most modern surgery rooms monitor for these signs, and will make use of dantrolene, which can rapidly reverse the symptoms. Dantrolene acts by decreasing the intracellular Ca2+ concentration . Several studies suggest a direct interaction between dantrolene and RyR1, and it is thought that dantrolene directly prevents Ca2+ release through RyR1 [31, 32].
The first link between MH and RyR1 came through a related disorder in pigs, known as porcine stress syndrome (PSS). It was found that the pig RyR1 mutation R615C underlies PSS , and soon after the corresponding mutation in humans was linked to MH .
2.2. Core myopathies
Central core disease (CCD) is a congenital myopathy, characterized by progressive muscle weakness and the presence of metabolically inactive tissue (cores) in the center of muscle fibers. These cores lack mitochondria and the myofibrillar organization is disrupted. The common phenotype of CCD includes muscle atrophy, floppy infant syndrome and skeletal muscle deformities [35, 36]. It was found early on that there is a connection between CCD and MH susceptibility , and some RyR1 mutations have been found to cause both. Multi-mini core disease (MmD) is another inherited congenital myopathy. Different from CCD, whose inheritance is autosomal dominant, MmD is usually considered recessive . As its name suggests, MmD is characterized by the presence of multiple not-well-defined cores in muscle fibers. MmD can cause axial muscle weakness that can lead to severe scoliosis . Although MmD is mainly associated with mutations in RyR1 , mutations in selenoprotein N 1 , a protein required for RyR1 calcium release , and α-actin (ACTA1)  have also been found to underlie MmD.
2.3. Catecholaminergic polymorphic ventricular tachycardia
Also known as CPVT, this disorder manifests itself in young individuals, with either syncope or sudden cardiac death as the first symptom. Many affected individuals will develop cardiac arrhythmias that are triggered by exercise or emotional stress. CPVT is mostly detected through stress tests, which indicate bidirectional ventricular tachycardia. Typically no morphological abnormalities of the myocardium are detected. The disease is mostly due to mutations in the cardiac RyR2 isoform , but mutations in the associated proteins calsequestrin  and triadin  can also be the cause. As CPVT is triggered by β-adrenergic stimulation, it can be treated by β-blockers . Flecainide, a sodium channel blocker, has also been found to be beneficial for CPVT patients .
Now more than 20 years since the initial identification of the first RyR disease mutations, nearly 500 mutations have been found in RyR1 and RyR2 combined. In RyR1, most disease mutations seem to be spread across the gene, with some clustering in distinct areas, whereas in RyR2, most mutations are found in one of 3 or 4 different disease hot spots (Figure 3). The appearance of clusters may be due to bias in sequencing, as historically only the areas that had already shown to be involved in disease were being considered for sequence analysis.
3. Functional studies
Although increasing numbers of RyR mutations are identified in patients and their family members, only a handful of them have been validated as causative disease mutations . Functional studies are necessary to prove the molecular basis of the mutations as pathogenic. Without functional characterizations, the possibility that the mutations are the result of polymorphism cannot be ruled out. In this section, different methods of performing functional studies and their results are highlighted.
3.1. Model systems
There are a variety of ways to prepare model systems to study functional effects of mutations in RyRs: endogenous or recombinant expression systems and in vivo models.
In the early days, before and shortly after the discovery of RyR mutations in MH-susceptible pigs and individuals, SR vesicles were prepared from muscle biopsies [47-55]. SR vesicles from non-MH-susceptible animals or individuals were also obtained to serve as controls. These are endogenous expression systems since SR vesicles contain RyRs, and therefore, can be used as a whole or can be further purified to obtain single RyR channels. These approaches can be invasive, as they involve acquisition of native RyRs expressed in patients or knock-in mice carrying mutations. For example, skeletal myotubes have been isolated from MH-susceptible and/or CCD patients [56-59]. Both RyR1 and RyR2 can be studied by isolating skeletal myotubes and cardiomyocytes from knock-in mice, respectively [60-73]. Flexor digitorum brevis fibers, present in the feet, have also been derived from knock-in mice for the study of RyR1 mutations [67, 74-76]. In some cases, measures have been taken to alleviate some of the affliction on obtaining native RyRs from patients by using less invasive approaches. In one study of a CCD-associated RyR1 mutation, fibroblasts from a CCD patient were differentiated into muscle by myoD conversion with adenovirus . In other cases, immortalized lymphoblastoid cells were generated using B-lymphocytes isolated from blood samples [78-83].
A more commonly used model is a recombinant expression system. By using cells that lack the native expression of RyRs, all results will arise only from the mutant RyR. One of the most commonly used systems are dyspedic myotubes, which are derived from mice that lack RyR1, called "RyR1-null mice" [80, 84-90]. Also commonly used are human embryonic kidney (HEK)-293 cells. HEK-293 cells not only lack native RyR expression but also lack other complexes involved in E-C coupling and yet have been demonstrated to express functional RyR when its cDNA is transfected [19, 79, 80, 89-115]. Other cell lines used in functional studies include COS-7 , myoblastic C2C12 , HL-1 cardiomyocytes [108, 110, 115, 118], and CHO cells .
Finally, knock-in mice have served as valuable models for the study of mutations in RyR1 and RyR2 [60-64, 66-76, 120-128]. The comparison between mice that are homozygous or heterozygous for the mutation in question has provided clues about the gene dosage effects. Studying the effect of mutations on the body as a whole provides insights that are more relevant to clinical phenotypes.
Multiple techniques have been implemented to study mutant RyRs from different expression systems. Five major methods are introduced here.
One way to measure Ca2+ release through RyR channels is through [45Ca2+]-uptake and -release assays [50, 52, 98]. Microsomal vesicles expressing recombinant RyR and SERCA1a are prepared. The vesicles are then incubated with 45Ca2+, during which SERCA1a pumps 45Ca2+ into the vesicles. By comparing the amount of 45Ca2+ accumulation in the presence or absence of the potent RyR inhibitor Ruthenium Red, the amount of Ca2+ release through RyR can be determined.
A more widely used technique is a Ca2+ fluorescence assay involving whole cell samples (Figure 4A,B) [19, 57-59, 70, 71, 73, 77, 79, 89, 91-93, 95, 102, 104-106, 108, 109, 112, 113, 115-118]. Cells with either endogenous or recombinant expression of RyR are loaded with a fluorescent Ca2+ indicator, such as fura-2, which diffuses throughout the cytoplasm. By measuring fluorescence with confocal microscopy, resting [Ca2+] levels in the cytoplasm, as well as frequency, duration, and extent of Ca2+ release from RyR can be calculated.
Commonly used is a [3H]ryanodine binding assay (Figure 4C) [19, 51, 54, 60, 67, 77, 79, 86, 90, 92, 93, 95, 96, 103, 105, 108, 113, 129]. SR vesicles containing endogenous or recombinant RyR are incubated with radioactive [3H]ryanodine. Since ryanodine binds to the channels in the open state, binding of [3H]ryanodine is indicative of the measure of open channels. The extent of RyR activation is then measured by radioactive counting.
As opposed to whole-cell measurements, activities of a single channel can be recorded using planar lipid bilayer electrophysiology (Figure 4D) [19, 53, 54, 67, 69, 77, 79, 99-101, 105, 107, 108, 111, 112, 114]. Recombinant or endogenous RyR from crude SR vesicle preparations or from purified material are fused into an artificial lipid bilayer formed across two chambers. The current generated by ions passing through the channel is recorded. This method allows various properties of the single channel to be determined, including the open probability, duration of the open and closed time, and conductance. The two chambers can resemble cytoplasmic and luminal sides of the channel, allowing the studies of Ca2+-induced Ca2+ release (CICR) from the cytoplasmic side or store-overload-induced Ca2+ release (SOICR) from the luminal side. The planar lipid bilayer technique also allows for additional control because purified auxiliary proteins or pharmacological modulators can be added to either side.
3.3. Highlight of the results
Functional studies of mutant RyR channels have generated multitudes of insights into the molecular basis of the RyR channelopathies. Accordingly, complex sets of results have been generated that require careful interpretation and have resulted in a lot of debate. For example, many mutations seem to make the RyR channels more active, but the exact nature of this change is still under scrutiny. In some cases, mutations seem to be more sensitive to cytosolic activators, while other evidence suggests they are more sensitive to luminal Ca2+ levels. Others suggest that altered phosphorylation states of the channel or affinity to auxiliary proteins is the causative mechanism. For this section, results of the functional studies are highlighted as mutations that cause RyRs to be either hypersensitive (more prone to open upon stimulation) or hyposensitive (requiring more stimulus to open). RyR mutations can be pathogenic either through hyper- or hypo-activity since alteration in Ca2+ homeostasis in either direction can cause aberrant muscle functions. Yet other mutations have exhibited no apparent alterations in the functional studies performed so far. The results are summarized in Table 1.
|R163C||MH/CCD/MmD||[61, 65-67, 85, 86, 91, 94, 103]||E2371G||MH|||
|G248R||MH||[91, 94, 103]||G2434R||MH||[1,2,36]|
|R328W||MH||||R2435H||MH/CCD||[85, 91, 94]|
|G341R||MH||[86, 91, 94]||R2435L||MH/CCD||[77, 88]|
|Y522S||MH/CCD||[60, 63, 64, 85, 87, 89, 91, 94, 120]||R2458C||MH||[91, 94]|
|R2458H||MH||[86, 91, 94]|
|R614C||MH||[47-52, 54, 86, 88, 89, 91,|
94, 101, 116, 117, 129]
|R614L||MH||[91, 94]||Y4796C||CCD||[87, 88]|
|E1058K||MH||||T4826I||MH||[76, 86, 103, 122]|
|R2163C||MH||[86, 88, 91, 94]||ΔF4863_D4869||CCD|||
|R2163H||MH/CCD||[85, 88, 91, 94]||R4893W||CCD|||
|V2168M||MH||||I4898T||CCD||[78, 89, 93]|
|RyR2||ΔExon 3||CPVT||||N2386I||ARVD2||[106, 109]|
|R176Q/T2504M||ARVD2||[108, 109, 112, 115]||N4104K||CPVT||[19, 118, 119]|
|E189D||CPVT||||Q4201R||CPVT||[107, 108, 112]|
|G230C||CPVT||||R4497C||CPVT||[19, 68, 71, 72, 105, 118, 119, 123, 124]|
|L433P||ARVD2||[108, 109, 115]||S4565R||CPVT|||
|P2328S||CPVT||[52, 54, 55]||N4895D||CPVT|||
|R109W/M485V||CCD/MmD||[99, 100]||G4899R||CCD||[87, 100, 103]|
|Uncertain effects||RyR1||I403M A1577T/G2060C||MH/CCD|
The majority of the mutations in both RyR1 and RyR2 have been shown to make the channels hyperactive. This includes the first disease-associated mutation found in porcine stress syndrome, RyR1 R615C. Functional studies have revealed that RyR1 channels with this mutation have significantly lowered threshold for activation by [Ca2+], caffeine, and halothane compared to wild-type channels, while having markedly increased threshold for inactivation by [Ca2+], Ruthenium Red, or Mg2+ ions [47-52, 101, 116, 117, 129]. These results are characteristic of mutations that alter RyRs into hypersensitive channels [19, 55-59, 61, 63, 65, 67-69, 71, 73, 76-83, 85-89, 91, 93, 94, 96, 97, 102-109, 111-115, 118, 119]. RyRs that are hyperactive are often described as “leaky,” and on top of their increased sensitivity towards activators or reduced sensitivity towards inhibitors, they often exhibit increased resting [Ca2+] in the cytoplasm, reduced SR Ca2+ store content, or altered maximal Ca2+ release.
Although less common, some mutations in RyR have been demonstrated to reduce channel activity. [62, 74, 75, 80, 82, 84, 87, 89, 98, 100, 103, 104, 106, 110, 112, 115]. For example, the I4898T mutation in RyR1, associated with an unusually severe and highly penetrant form of CCD, has been shown to cause complete uncoupling of sarcolemmal excitation from SR Ca2+ release, in which activating signals from L-type CaV channels to RyR1 are uncoupled . It has also been shown that the I4898T mutation leads to a complete loss of Ca2+ release induced by caffeine stimulation . Many other CCD-associated mutations that are located at the predicted pore region of RyR1 (G4890R, R4892W, A4894P, I4897T, G4898R, G4898E, A4905V, and R4913G) have also been shown to abolish E-C coupling and/or exhibit reduced sensitivity to activation by [Ca2+], caffeine, or 4-CmC [62, 87, 98, 100, 103, 104]. Mutations that disrupt signaling between the L-type CaV and RyR are known as “E-C uncoupling” mutations. In RyR2, fewer mutations have thus far been associated with a reduction in activity, suggesting that a loss of function in RyR2 may be less tolerated than in RyR1. The A4860G mutation in RyR2 is associated with idiopathic ventricular fibrillation (IVF) and is located at the predicted inner pore helix. It has been shown to diminish response or increase the threshold for activation by SOICR [110, 112]. Furthermore, the mouse RyR2 A1107M mutation (corresponding to the T1107M mutation that is associated with hypertrophic cardiomyopathy in humans) has been shown to increase the threshold for Ca2+ release termination .
Effects of some other disease-associated mutations are less apparent. For instance, single channel measurements of RyR1 A1577T and G2060C mutant channel produced no discernible change in the activity of the channel compared to wild type . Similarly, [3H] ryanodine binding assays of RyR1 R2939K expressed in HEK-293 cells failed to show alteration in Ca2+ dependence and caffeine activation compared to wild-type channel . In addition, RyR1 I404M expressed in dyspedic mice myotubes exhibited resting [Ca2+] levels and SR Ca2+ content comparable to that of the wild type . Furthermore, expressions of RyR1 D3987A and G3939A rabbit cDNA in HEK-293 cells, which correspond to the positions of MH-associated mutations D3986E and G3938D in humans, respectively, showed responses to activation by [Ca2+], caffeine, and 4-CmC comparable to that of wild type [92, 95]. For these mutations where no clear functional effect has been observed, it is of course possible that they are not causative of the disease but, instead, simply represent polymorphisms. Alternatively, their effect may only become apparent in the native context, where particular regulatory mechanisms exist that are not captured by the model systems.
Overall, there is an apparent theme between RyR activity and disease phenotype. MH and CPVT are associated with RyRs that have an overall gain-of-function phenotype. On the other hand, CCD can be due to either a gain or a loss of RyR1 activity. Loss of function results in impaired Ca2+ release, and hence decreased contractility observed in CCD. The gain of RyR1 activity can lead to a general “leak“ of Ca2+, resulting in an overall lowered concentration of Ca2+ in the SR. The result is then an insufficient amount of Ca2+ being available for an E-C coupling event, again resulting in decreased contractility.
For those mutations with less clear effects, as well as those which have been shown to be causative for aberrant Ca2+ homeostasis, combining functional studies with structural investigations would provide further useful insight into the molecular basis of disease-associated mutations in RyRs.
4. Structural studies of RyRs
Structural biology has thrived in recent decades and provides researchers with powerful tools for exploring all aspects of life. Different structural techniques, covering the resolution range from 50 Å down to 1 Å, can reveal the structural details of various bio-molecules, from large protein/DNA/RNA complex to small peptides, which help to improve our understanding of the physiology and pathology of many biological systems.
However, the large size of the RyR and its membrane protein nature make it a very challenging target for structural studies. Most structural techniques require protein samples with high yield and purity, which is normally achieved by overexpression in recombinant systems combined with multi-step chromatography purification processes. The large size of the RyR makes the recombinant expression difficult because the folding of this giant requires the assistance of many chaperones which are often absent in non-native expression systems. The differences in the protein trafficking and post-translational modification systems may also affect the maturation of the recombinant RyR. The alternative way is to purify RyRs from native tissues, but the purification is not easy without artificial affinity tags. To further complicate the issue, the presence of a large exposed hydrophobic surface in the transmembrane region requires the addition of detergents to make the protein soluble during the purification. The denaturing property of the detergents usually makes the sample less homogenous and their presence also affects the crystal packing, both of which are undesirable for x-ray crystallography. The intrinsic dynamic feature of this multi-domain channel further increases heterogeneity of the sample and makes it harder to study.
Despite all the technical difficulties, during the last two decades, several structures of RyRs have been obtained using cryo-electron microscopy (cryo-EM) [130-141], NMR  and X-ray crystallography [143-148]. Some of these structures [142, 143] describe how regulatory proteins interact with small fragments (peptides) from RyRs and are not relevant to the theme of this book. Here we will focus on the structures of the full-length channel or large domains where disease mutations are localized, and show how the structural studies can reveal the molecular mechanism of a genetic disease. More details on the structural insights into RyR function have been described in a recent review paper .
4.1. Cryo-EM studies of RyRs
Electron microscopy has been able to obtain structural information on RyRs at near sub-nanometer resolution. The early studies using thin-section or negative stain electron microscopy discovered some "feet" structures (square-shaped densities) in the junction region between the SR and T-tubules in muscle tissues [150-157]. The identities of these "feet" were confirmed later when the purified RyRs were imaged by EM, showing similar densities . The negative stain structures with improved resolution revealed that RyRs have a four-fold symmetry and can interact with four neighbouring molecules to form two-dimensional checkerboard-like lattices both in native tissue and in vitro [158-162], suggesting that RyRs may interact cooperatively during gating. The resolution was further improved by the cryo-EM technique [130-141], in which the protein samples are studied under cryogenic temperatures. So far the highest resolution reaches 9.6 Å for RyR1 in a reported closed state.
Many structural details can be appreciated from the cryo-EM structures. The overall shape of the RyR is like a mushroom with a large cap in the cytoplasmic domain (27 × 27 × 12 nm) and a small stem (12 × 12 × 6 nm) crossing the ER/SR membrane. Instead of a single rigid block, the large "cap" consists of fifteen individual globular regions per subunit with many solvent channels in between [163, 164] (Figure 5A,B). This arrangement maximizes the total exposed surface area, and RyRs thus form perfect scaffolds for many small molecules and protein regulators to dock. The number of transmembrane (TM) helices in the "stem" is still controversial at the current resolution. Five or six TM helices can be detected in the cryoEM maps [134, 141], but it has been suggested that the total number should be either 6 or 8 . The last two TM helices create the pore-forming domain, which is homologous to the ones from tetrameric potassium channels and bacterial sodium channels .
Comparison of the RyR structures in the closed and open states reveals movement in both the "cap" region and "stem" region upon channel gating. An iris-like movement around the pore switches the channel between dilated and constricted conformations  (Figure 5C,D). Samso et al. systematically produced RyR structures in the closed and open states and found the inner helices in the TM region to kink in the open state, increasing the diameter of the pore by ~4 Å . However, another group reported that the inner helics are already kinked in the closed state . The discrepancy can be attributed to differences in sample preparation. Although the conditions may favor a closed or open channel when the RyR is in a membrane, this may not be the case after extraction with detergents. In addition, the amount of oxidation in the sample can also affect the open probability of the channel. It is therefore possible that the 9.6Å map represents an open, rather than a closed channel, and this would reconcile the discrepancies between both studies. So far, most cryo-EM studies are done on RyR1 due to its relatively high abundance. RyR2 and RyR3 have also been studied via cryo-EM [137, 138], albeit at lower resolution. The overall folding is similar for all three isoforms. The major difference between RyR1 and RyR2 is the clamp region on the corner of the "cap" , while a central domain (possibly corresponding to the residues 1303-1406 in RyR1), which is missing in RyR3, is the most divergent part between RyR1 and RyR3 . These structural elements very likely contribute to the different properties of the three subtypes.
The cryo-EM studies of RyRs thus far do not provide the resolution to locate individual residues as X-ray crystallography does, but it can give clues about the approximate locations of individual domains through analysis of the difference densities from insertion of fusion proteins (such as GFP) or binding of antibodies for the particular domain. The locations of the three disease hot spots have drawn particular interest since these parts are certainly important for channel function, as the clustering of disease-associated mutations suggests. From the primary sequence analysis, it is clear that the third hot spot is in the transmembrane region ("stem") of the channel. Several insertion studies localize the N-terminal disease hot spot at the clamp region between the subdomain 5 and 9 [31, 167-169], and the central hot spot between the subdomain 5 and 6 , suggesting that these two hot spots interact with each other  (Figure 5A,B). Similar methods were also used to localize the divergent regions [172-176] and the phosphorylation sites [177, 178]. The accuracy of these studies varies depending on the resolution, linker length, insertion site selection and the size of the fusion protein or antibody. Difference densities analysis was also used to identify the binding sites of protein regulators, such as FKBP12/12.6 [133, 179-181], calmodulin [180, 182, 183], chloride intracellular channel 2 (CLIC-2)  and various toxins [185, 186].
4.2. X-ray crystallographic studies of RyRs
The crystallization of full-length RyRs is a very challenging task due to its large size, membrane protein nature and intrinsic flexibility. Improvement in techniques will likely reveal a high-resolution structure of intact RyRs in the future. The cryo-EM structures have revealed that the RyR is a large modular protein that consists of many small domains. Many efforts have therefore been made to crystallize individual domains or domain clusters that carry important functions.
The first few crystal structures of RyRs that have been solved are of the N-terminal domains (domain A) in RyR1 (residues 1-205) and RyR2 (residues 1-217) [144, 145]. Both display a β-trefoil folding motif (three groups of β-strands that form a triangle structure) that consists of twelve β strands and contains an extra α helix (Figure 6A,B). Domain A spans about one third of the N-terminal disease hot spot. A total of 23 and 9 mutations have been found so far in domain A of RyR1 and RyR2, respectively, many of which cluster in a loop between β strands 8 and 9. Several disease mutants of this domain have been analyzed either by NMR or x-ray crystallography, but none of them showed major structural changes [144, 145] (Figure 7A,B). The observation that most mutations are located on the surface of domain A indicates that they likely affect the interaction of domain A with other domains in RyR or auxiliary proteins.
A severe form of CPVT is caused by the deletion of the entire exon 3 (residues 57-91) from the ryr2 gene [187, 188]. The structure of wild type domain A shows that exon 3 contributes to a loop region, the 310 helix, β4 strand, and the α helix. One would expect that the complete absence of these structural elements will cause the misfolding of the domain and a loss-of-function phenotype. However, surprisingly, Δexon 3 is not strictly lethal. CPVT symptoms and functional studies suggest a gain-of-function phenotype instead . Even more surprising, the mutant has a higher thermal stability compared to the wild type domain . A crystal structure of RyR2 domain A with Δexon 3 revealed an unusual "structural rescue" scheme: a flexible loop following the exon 3 region replaces the deleted β4 strand and results in a similar overall fold of the domain  (Figure 7C). The α helix, however, is lost. The ability to "rescue" a β strand with a random loop segment is highly unusual and suggests that there has been evolutionary pressure to maintain the exact sequence of the flexible loop. Most likely, there is the possibility of alternative splicing, whereby exon 3 is naturally spliced in some cell types. This would represent a mechanism whereby the RyR activity can be fine-tuned in a tissue-specific manner . So far, such a splice variant has not yet been reported.
Up to now, the largest reported crystal structure for all RyRs is for the N-terminal three domains (domain A, B and C, residues 1-559) of RyR1, which covers the majority of the N-terminal disease hot spot . It shows a triangular arrangement of the three domains that interact with each other, mainly through hydrophilic interfaces. Domains A and B have similar β-trefoil folding motifs, while domain C has an armadillo-repeat structure consisting of a bundle of five α-helices (Figure 6C). A total of 76 mutations (RyR1+RyR2) have been identified in these domains. 17 of these are localized at the interfaces between domains A, B and C (within one subunit). 9 mutations are buried inside the individual domains, and the remaining are on the surface of the ABC structure.
Although the EM structure itself usually does not provide the resolution to resolve the atomic positions, the combination of crystal structures and EM structures allows pseudo-atomic models to be built. Individual amino acid residues can thus be located within the overall 3D structure. The high-resolution crystal structures can be docked into the EM maps by matching their electron densities using complex 6 dimensional search algorithms [189-191]. The confidence of the solutions depends on the quality of the EM map, the size of the docked fragments and the agreement of conformations from the two structures. Sometimes a Laplacian filter, which uses an edge-detection algorithm to increase the sensitivity can increase the success rate when a small fragment is used as a search model . Docking of the RyR1 ABC crystal structure into three different cryo-EM maps places it at the same location, the central rim region, forming a cytoplasmic vestibule around the four-fold symmetry axis (Figure 10) . This solution contradicts previous results from insertion experiments which localize the N-terminal hot spot at the clamp region near the corner [167, 168]. The authors suggested that in the GFP-insertion study  the long linker and the native flexible loop make it possible for the inserted GFP to be present in a remote location . In addition, the difference density observed in a GST-insertion study splits into two regions, one of which is indeed right beside the central rim region . Interestingly, the location and orientation of domains ABC from the IP3 receptor, a closely related Ca2+-release channel, are remarkably similar to the docking solution of RyR1 ABC, which further validates the location of the N-terminal region .
Locating the N-terminal region within intact RyRs allows several new interfaces to be identified. As the ABC domains dock near the 4-fold symmetry axis, there is a direct interaction between domains A and B from neighboring subunits (Figure 10). In addition, there are five other interfaces between the ABC domains and other regions of RyR1. All the peripheric disease mutations identified in domains ABC are located in one of these six interfaces and none of them are exposed to the surface of the full-length channel, suggesting that they mainly affect the inter-domain interactions rather than interactions between the RyR and other proteins.
The activity of RyRs can be regulated by different types of kinases and phosphotases (Figure 1 and 2). It is believed that the chronic hyperphosphorylation of RyR2 by PKA can dissociate the binding of FKBP12.6 and cause heart failure [193-195], although this result has not been confirmed by other laboratories [118, 196, 197]. Two recent papers reported the structures of a domain containing the phosphorylation sites from all three RyR isoforms , and from RyR1  (Figure 8A,B,C). The domain consists of two symmetric halves. Each half contains two α helices, one or more short 310 helices, and a C-terminal β strand. Multiple phosphorylation sites were identified in this domain of RyR2  and most cluster within the same loop. The RyR2 domain doesn’t contain any reported disease mutations, but up to 11 disease mutations are found in the RyR1 domain. Interestingly, several disease mutations from this domain are also located close to the phosphorylation sites, suggesting that phosphorylation and disease mutations at this site may have the same effect (Figure 8A). One mutation, L2867G, reduces the melting temperature of the individual domain by 13°C and causes the aggregation of the domain at room temperature (Figure 7D). Several crystal structures for the other disease mutants were also reported in this study, showing that they affect surface properties and intradomain salt bridges (Figure 7E,F,G) . Docking into the EM map places this domain in the clamp region, but none of the neighbouring domains have been identified thus far (Figure 10). Solving structures of other domains may place more pieces into this big jigsaw and increase our understanding of the isolated domains.
4.3. Homology models for RyR domains
So far all the available crystal structures cover ~15% of the full-length RyR in total. There is still a big portion missing. One alternative way to obtain structural informaiton is to create homology models based on known structures using bioinformatics methods. Generally the reliability of a homology model depends on the sequence identities between two proteins, but protein 3D structures tend to be more conserved than their primary sequences during evolution . As the largest ion channel, the evolution of modular RyRs involves the incorporation, loss and duplication of different domains. Protein sequence analysis identifies several domains that are present multiple times in RyR genes, including SPRY domains and tandem repeat domains.
The spore lysis A and RyR (SPRY) domains have been found in many different proteins from bacteria, archaea, viruses and eukaryotes. They are generally known as protein interaction domains . There are three SPRY domains identified in metazoan RyR genes (SPRY1,2,3) (Figure 3). In skeletal muscle, RyR1 (located in the SR membrane) and CaV1.1 (located in the plasma membrane) are thought to interact directly. Some biochemical evidence suggest that the RyR1 SPRY2 domain forms an interaction with one of the intracellular loops of CaV1.1, the so-called II-III loop . Using the phyre2 engine , we built homology models for all three SPRY domains of RyR1, based on several known SPRY protein structures (Figure 9A,B,C). All these models show a signature β-sandwich (two stacked β-sheets) folding motif. Interestingly, all ten disease mutations identified in these three SPRYs are located peripherally when mapped into the models (Figure 9A,B,C). As before, it seems that mutations cluster at the surfaces of individual domains, likely affecting interactions with neighboring domains or RyR binding partners. This observation agrees with the known protein interacting properties of the SPRY domain. The mutations in these domains may weaken the interaction with other domains or proteins and change the channel activities.
Another domain fold present in RyRs is the so-called "RyR domain". The latter name is unfortunate, as it obviously is not the only domain present in RyRs. These domains are arranged in tandem repeats, and we will hereafter refer to them as "tandem repeat domains". Each repeat unit is about 100 residues in length, and there are two tandem repeats within each human ryr gene: one between SPRY1 and SPRY2, and the other at the central region (Figure 3). These repeat units are also present in bacteria, archaea and viruses, either in single form or in pair with itself or other proteins . The phosphorylation domain mentioned above makes up the second tandem repeat . Interestingly, in a crystal structure of a bacterial repeat domain (pdb code: 3NRT), a single domain dimerizes and presents a similar fold as the paired repeat domain shown in the phosphorylation domain structure. We created a homology model for the first tandem repeats from RyR1 based on these two protein structures . (Figure 9D). Seven mutations are found on this tandem repeat. Five of these are located on the surface, likely affecting interactions with other domains or RyR binding partners. The remaining two are buried within the domain, and may thus affect the overall folding stability.
Although the total number of the TM helices in RyRs is not clear, from the secondary structure prediction, their last two TM helices clearly share homology with the pore-forming regions of potassium channels and bacterial voltage-gated sodium channels (NaV). Because the pore-forming region is directly responsible for passage of Ca2+ ions through the SR membrane, any minor structural alteration in this region may directly interfere with the ion conduction. Indeed, up to 50 mutations (RyR1+RyR2) have been identified in this region, which consists of only 125 residues. The occurrence of one mutation per 2.5 residues makes the pore-forming region into a "super-hot" disease hot spot. We created a model for the RyR1 pore-forming domain based on several bacterial NaV structures (Figure 9E). Most mutations cluster in the top half (ER luminal side) of the domain. Interestingly, 31 out of 41 RyR1 mutations in this domain are associated with CCD, which can be caused by the loss-of-function of RyRs. Therefore, it seems that many mutations in the TM domain can directly block the ion conduction pathway and reduce channel activity, while mutations in the cytoplasmic domains tend to destabilize the closed state of the channel and make it "leaky".
Like the previous studies with crystal structures, knowing the exact locations of these domains in the full length channel will deepen our understanding of the channel in the normal and disease states. But technically it is dangerous to dock homology models into the EM maps due to their limited accuracy. However, these models can still provide a good template to guide the design of the insertion sites for the EM studies, which will help to reveal the location of these domains and the corresponding disease mutations.
The available crystal structures and homology models represent ~32% of the full length RyRs. In the years to come, further crystallographic studies on predicted domains, and their localization into full-length RyRs will likely shed more light. Although obtaining well-diffracting crystals of intact RyRs may be an arduous task, this will ultimately be necessary to fully understand the function of RyRs in health and disease.