Bayesian modeling to predict malignant hyperthermia susceptibility and pathogenicity of RYR1, CACNA1S and STAC3 variants

Malignant hyperthermia (MH) is a potentially lethal pharmacogenetic sensitivity to volatile anesthetics and succinylcholine. Most patients have more than one anesthetic exposure before an MH event [1]; therefore, a negative personal and/or family history does not exclude MH susceptibility (MHS). Additionally, the diagnosis of MH may be challenging because the signs and symptoms are not specific. The Clinical Grading Scale (CGS) [2] defines the clinical features of ‘very likely’ and ‘almost certain’ MH cases, but the CGS score may also be high during anesthetics complicated by sepsis, thyroid storm, respiratory failure, adverse drug reactions unrelated to MHS, neurologic disease, or other conditions producing hyperthermia, muscle rigidity and/or hypercarbia [3]. Muscle bioassay [4–6] has been used for identification of MHS, using strips of excised muscle for an in vitro bioassay – known as the caffeine-halothane contracture test (CHCT) in North America and the in vitro contracture test in Europe.

More than 25 years ago, researchers identified RYR1 as the gene most frequently associated with MHS [7,8]. Early genetic studies were conducted in nonmyopathic subjects with family history of clinical MH, including many individuals who had undergone muscle contracture testing. Depending on study selection criteria and testing methods, RYR1 variants occur in as few as 23% or as many as 86% of MHS North American or European individuals [9–12]. Variants are much less common in the CACNA1S gene [13–15] that encodes the α1 subunit of the skeletal muscle voltage-gated calcium channel (interacting with RYR1 in the process of excitation–contraction coupling). Variants in the STAC3 gene (encoding Stac3) were recently found in patients with congenital myopathy associated with MHS [16,17]. The Stac3 protein has an essential role in skeletal muscle excitation–contraction coupling [18].

Many patients with suspected MH events never undergo muscle contracture testing since the assay is costly, invasive and often inconvenient (requiring travel to a testing facility). Clinical genetic testing for MHS has been available since 2005, and the European Malignant Hyperthermia Group (EMHG) provides an updated list of diagnostic MH mutations (currently 48 RYR1 and 2 CACNA1S variants) [19]. An individual carrying one of these variants should be considered MHS (at increased risk of developing MH during anesthesia), and family members may be screened by genetic testing. An individual who does not carry a familial variant cannot be regarded as MH negative (MHN) since other pathogenic variants could be involved, and the genetics of MH are not fully known. Numerous other variants in RYR1 have been reported but are not functionally characterized and thus are not considered to be diagnostic for MHS. Muscle contracture testing, supplemented by recombinant in vitro expression in HEK 293 cells or myotubes of RYR1 knockout mice demonstrating functional change, remains a critical part of the classification of the variants based on EMHG guidelines [20].

Next-generation sequencing (NGS) for MH was described in 2012 [21], and whole exome sequencing has been performed on the known MH-associated genes, as well as genes involved in excitation–contraction coupling and skeletal muscle calcium homeostasis [22–24]. This approach is less costly and time consuming than Sanger sequencing, and it allows for identification of novel genes and/or variants [24]. The population frequency of an allele can be estimated by searching public databases containing sequencing data from >67,000 human exomes. Since most RYR1 variants are rare, having frequencies <0.001, a low frequency cannot be used alone as a predictor of pathogenicity [25].

In 2015, the American College of Medical Genetics and Genomics (ACMG) and the Association for Molecular Pathology (AMP) published a joint consensus recommendation for interpreting sequence variants observed in patients with suspected inherited disorders [26]. This included weighted criteria for classifying variants as pathogenic or benign, and proposed rules for combining those criteria into five categories (pathogenic, likely pathogenic, variant of uncertain significance [VUS], likely benign and benign). Tavtigian et al. described Bayesian modeling using the ACMGG/AMP criteria to produce a quantitative probability of pathogenicity [27]. Bayesian reasoning uses a prior probability and modifies that based on conditional factors, resulting in a posterior probability. This methodology allows for integration of data from multiple sources to make predictions about the significance of rare genetic variants.

The aims of this study were to: identify novel variants (not previously associated with MH) by sequencing of RYR1, CACNA1S and STAC3; and to assess pathogenicity of identified variants (both novel and previously described in association with MH) using application of Bayesian statistics [27] to integrate data from the current and prior studies.

Materials & methods

Subjects

The study design was a case–control genetic risk assessment study. Study subjects living in the continental USA were recruited from the North American Malignant Hyperthermia Registry (NAMHR; Supplementary Material). Registrants are primarily of European descent, with <10% reporting mixed race with Native American, Asian or African ancestry. Once identified from the registry, study candidates were contacted by mail as many as three-times to request their participation. The subject population included three groups: MHS individuals with abnormal CHCT results in the NAMHR; subjects with suspected MH episodes, abnormalities on clinical genetic testing of RYR1 or rhabdomyolysis, and their first-degree relatives with similar history who had joined the NAMHR without CHCT results; and MHN subjects with normal CHCT results in the NAMHR. Initially, only subjects meeting criteria for Group 1 were contacted to participate in the study; however, the response rate was low. Group 2 was included to expand the study and evaluate individuals with a clinical suspicion of MH or those who might have care altered based on personal/family history. Subjects were included from families with a previously reported MH variant. Blood samples for DNA analysis were collected, stored and processed according to protocols approved by the University of Pittsburgh Medical Center and Cincinnati Children's Hospital Medical Center Institutional Review Boards. All subjects gave written informed consent.

Gene sequencing

Whole exome sequencing by NGS was performed by Prevention Genetics, WI, USA. Results are reported here only for RYR1, CACNA1S and STAC3 genes. For this NGS panel, the full coding regions plus approximately ten base pairs of noncoding DNA flanking each exon were sequenced for each gene. Sequencing was accomplished by capturing specific regions with an optimized solution-based hybridization method, followed by massively parallel sequencing of the captured DNA fragments. Additional Sanger sequencing was performed for any regions not captured or those having an insufficient number of sequence reads. All pathogenic, undocumented and questionable variant results were confirmed by Sanger sequencing. All sequencing data are available on request.

Initial pathogenicity classification of previously identified variants

Identified gene variants were referenced against the EMHG registry of diagnostic MH variants. The ‘previously reported pathogenicity’ classification of non-EMHG variants was based on data in the National Center for Biotechnology Information (NCBI) ClinVar website (www.ncbi.nlm.nih.gov/clinvar/docs/linking) [28] and reported by Prevention Genetics. Variants were classified as pathogenic, likely pathogenic, VUS, likely benign, or benign as presented in the consensus statement of the ACMG/AMP [26].

In silico analysis of nucleotide/amino acid substitutions

PolyPhen-2 [29] (individual and batch modes; http://genetics.bwh.harvard.edu/pph2) was used to predict the possible impact of amino acid substitutions on protein structure and function. PolyPhen-2 produces a qualitative prediction of probably damaging, possibly damaging, or benign, along with a numerical score ranging from 0.0 (benign) to 1.0 (damaging). The HumDiv and HumVar scores and the sensitivity and specificity estimate for the scores were recorded. The Genomic Evolutionary Rate Profiling (GERP++) score, the Combined Annotation-Dependent Depletion (CADD) [30] score, the Grantham distance and the Genome Aggregation Database (gnomAD; gnomAD.broadinstitute.org) [31] data were obtained from SeattleSeq Annotation 150 (http://snp.gs.washington.edu/SeattleSeqAnnotation150/) using the exact location of the nucleotide substitution in the individual chromosomes of interest (1, 12 and 19). These positions were obtained from the NCBI ClinVar [28] website (www.ncbi.nlm.nih.gov/clinvar/docs/linking), if the variant had been previously reported and included in ClinVar. In other cases, the location was determined using nearby previously reported variants and verifying the location on the chromosomal map for NCBI Assembly Genome Reference Consortium human build patch 38 (GRCh 38).

Classification of gene variants by Bayesian posterior probability method

Variants were initially classified based on their posterior probabilities from the Bayesian calculator described by Tavtigian et al. [27] (available in the Supplementary Material of that article). Assignment of evidence strength was done using the ACMG/AMP criteria presented by Richards et al. [26]. Briefly, pathogenic criteria were weighted as strong (PS1-4), moderate (PM1-6) or supporting (PP1-5), and benign criteria were weighted as standalone (BA1), strong (BS1-4) or supporting (BP1-7). Critical changes were made to the Tavtigian et al. calculator in the Bayesian prior probability and the exponential progression coefficient. The Bayesian prior probability was set to 0.02, instead of 0.10, based on the number of nonintronic variants in RYR1 that are pathogenic and the total number of nonintronic variants present in a European-based sample (gnomAD database). The exponential progression coefficient was set to 1.9 instead of 2.0. This yields a posterior probability of 0.996 for RYR1 p.Arg614Cys (a known EMHG pathogenic variant) given two strong (PS3, PS4), one medium (PM2) and two supporting (PP1, PP3) criteria with a Bayesian prior probability of 0.02. All EMHG pathogenic variants met criteria to achieve a posterior probability of 0.991 or greater.

A threshold for the maximum frequency of the variant allele was set at 2.0 × 10^-4 (the frequency of the RYR1 p.Arg614Cys minor allele frequency [MAF] in Europeans from the gnomAD database). Two variant frequencies were determined, including: the MAF for the European non-Finnish (EnF) cohort reported in version 2.1.1 of gnomAD; and the highest frequency observed in any subpopulation in gnomAD 2.1.1. If the variant was reported in a subpopulation but not in the EnF population, the frequency was calculated as less than one/number of EnF exomes in the dataset. If the variant was not listed in gnomAD, the value from the National Heart, Lung, and Blood Institute Exome Sequencing Project, as reported by Seattle Seq Annotation 150, was used as the denominator. If there was a numeric value for the number of times the major allele was observed in the sample but there was no value for the minor allele, the MAF was reported as one/number of observations of the major allele. For example, the RYR1 p.Gln4837Glu major allele (C) was reported 13,006-times and there was no value for the minor allele (G); therefore, the MAF was reported as 7.69 × 10^-5. If the variant frequency in the EnF subpopulation was above threshold or the frequency in any subpopulation listed in gnomAD was more than double the threshold (true for RYR1 p.Gly893Ser and RYR1 p.Val4842Met), a BS1 was assigned, otherwise, a PM2 was assigned. If the allele frequency was reported as ‘unknown’, the entry in Supplementary Table 4 was unknown (e.g., RYR1 p.Ala4295Val).

An abnormal CHCT was not used as the sole evidence for pathogenicity of a specific variant, as multiple variants may be present in individuals with abnormal CHCT. Assessing the impact of an abnormal CHCT on the probability that a given variant in RYR1, CACNA1S or STAC3 is pathogenic is not straightforward. Taken alone, an abnormal CHCT has almost no impact on the conditional probability that a given variant is pathogenic. Adding sequencing information can significantly alter the conditional probability, provided that certain conditions are satisfied, including:

• A strongly positive CHCT (caffeine contracture >0.6 g; halothane contracture >1.1 g).

• Identification of all variants in RYR1, CACNA1S and STAC3 by sequencing.

• No ‘known’ pathogenic variants are present in any of the three genes above.

• All but one of the variants in those three genes are known to be benign or meet strong criteria, such as allele frequency, for a benign variant.

The mean probability that the remaining VUS in RYR1, CACNA1S or STAC3 is pathogenic is 0.755 (95% CI = 0.688–0.823; Supplementary Material). This is a Bayesian posterior, as it is a conditional probability. A variant can be classified as pathogenic without specific functional testing (PS3), based on evidence gathered from multiple families and full gene sequencing of RYR1, CACNA1S and STAC3 as shown in Supplementary Material.

We used the sequencing data available from this study and existing class assignments of known variants to identify variants that met the four conditions above. The method described by Tavtigian et al. [27] for testing the internal validity of the ACMG-AMP combining criteria did not apply, without modification, to our data and to the study of MH in general. The variants found in the present study were scored using the Bayesian approach with the following modification. Novel variants meeting all of the four conditions above were assigned PS3 with a probability of 0.755. The posterior probability for a variant meeting the four criteria above was computed by adding (1 - 0.755) multiplied by the posterior without PS3 and 0.755 multiplied by the posterior including PS3.

where α = 0.755 and the * indicates all of the evidence of strong (other than PS3), moderate or supporting that applies to the variant.

Results

Genotyping results

Complete sequencing of RYR1, CACNA1S and STAC3 by NGS was performed for 64 subjects (57 proven or suspected MHS subjects or first-degree relatives and 7 MHN control subjects; Figure 1). Many previously reported variants (Table 1) were identified, including 8 EMHG pathogenic variants in RYR1 [19] found in 15 subjects and 7 variants in RYR1 or CACNA1S [9,11,12,14,22,24,32–43] classified as likely pathogenic. One of these likely pathogenic variants for MH is pathogenic for Central Core Disease [36,38–41]. Many other previously reported variants [6,9,11,32,34,38,43–48] were identified, and had been classified as VUS or benign/likely benign (Table 1). Additional benign genetic variants (not listed) were found in RYR1 and CACNA1S from every subject and in STAC3 from 25 subjects. The CACNA1S variant p.Pro1839Ser [43] was initially reported as likely pathogenic by Prevention Genetics; however, we changed the initial classification to VUS based on the Bayesian posterior probability from previously reported data. Fourteen novel variants were found in 12 subjects, including 12 RYR1 variants (1 duplication), a CACNA1S variant and a STAC3 variant (Table 2).

All genetic variants were missense mutations, except for the single duplication. An intronic variant in RYR1 was found in one subject (Supplementary Table 1); this person also had a recognized pathogenic variant and a VUS in RYR1.

Clinical details of subjects with abnormal CHCT

A total of 38 subjects had abnormal CHCT results (performed between 1970 and 2015; Supplementary Table 1). Four of these CHCT results were not quantitatively comparable to the others, as they involved nonstandard methods [5]. Genotyping revealed pathogenic RYR1 variants in 12 of the 38 subjects (32%), and likely pathogenic variants in RYR1 or CACNA1S in 7 additional subjects; therefore, 50% of subjects with CHCT results diagnostic of MHS had at least likely pathogenic previously described variants.

After application of Bayesian methods, 11 of these 38 subjects had only benign/likely benign variants in RYR1, CACNA1S and/or STAC3, or no potentially pathogenic variants identified. Muscle contracture testing (between 1984 and 2015) was performed according to the standard method for 8 of these 11 subjects and produced maximal contractures in the presence of 3% halothane of 0, 1.05, 1.1, 1.2, 2.2, 2.58, 3 and 3.94 gm. The diagnostic cutoff for the assay is 0.7 gm.

Twenty six of these 38 subjects reported a personal experience of MH during general anesthesia. The CGS scores of these MH events ranged from 10 to 70 for subjects with pathogenic or likely pathogenic variants in RYR1 or CACNA1S. In the six subjects with only VUSs based on previous reports, the highest CGS score was 53. Of the ten subjects with only benign or likely benign genetic variants by previous reports, six individuals reported suspected MH events with CGS scores between 0 and 15.

Eleven subjects underwent muscle contracture testing only because of family history of MH. Pathogenic or likely pathogenic RYR1 variants were found in seven of these subjects and two VUSs in RYR1 were found in one other subject.

Seven subjects reported a fatal MH event in a first- or second-degree relative, and five had either a pathogenic or likely pathogenic RYR1 variant (p.Gly341Arg, p.Thr2206Met, p.Arg2435His or p.Leu4824Pro) or the VUS p.Gln4837Glu. The remaining two subjects had only benign variants in RYR1, CACNA1S and STAC3.

Clinical details of subjects with no CHCT results

No muscle contracture testing had been performed on 19 subjects (Supplementary Table 2), although the majority (15/19) of these individuals had reported MH events (CGS scores of 3–73) between 1986 and 2017. Pathogenic variants recognized by the EMHG were found in three of these subjects (CGS scores 73, 33 and 15). Four subjects had not experienced MH events, but they joined the NAMHR due to family history. Two of these subjects had a first-degree relative with a history of repeated episodes of rhabdomyolysis without exposure to anesthetics, and they had RYR1 variants identified by clinical genetic testing (Supplementary Material). The other two subjects had strong family histories of MH, and only benign variants were found in these individuals.

Clinical details of MHN subjects

Seven subjects, serving as MHN controls, had a record of previous normal CHCT between 1990 and 2016 (Supplementary Table 3). Six of these individuals had experienced suspected MH events (1979–2014, CGS scores 10–58), and the remaining subject had an MHS sibling. In the subject with CGS score of 10, the RYR1 variant p.Glu1424Lys (predicted to be likely benign by Bayesian methodology) was identified. All other variants were benign or likely benign variants in RYR1.

Bayesian statistics

Predictions for the seven novel variants (not previously associated with MH) from subjects with abnormal CHCT included four VUSs and three benign or likely benign (Table 2; Supplementary Table 4). All VUSs (RYR1 p.His460Arg, p.Arg3629Trp, and p.Glu3853Lys, and STAC3 p.Met187Thr) had posterior probabilities of 0.649. This is the maximum Bayesian posterior probability for a novel variant, achieved with in silico evidence of a damaging mutation, MAF below threshold and PS3 per Equation 1. The PS3 criterion was assigned for these variants given the abnormal CHCT and the lack of other variants on full gene sequencing of the three genes in that subject. Novel variants from patients with no CHCT results were VUSs or likely benign variants with posterior probabilities from 0.094 to 0.344.

Using Bayesian statistics to combine the case details from this report with previous reports, we were able to predict pathogenicity for the previously identified variants (Tables 1 & 3; Supplementary Table 4). Modeling results predicted likely pathogenicity for the CACNA1S variant p.Thr1009Lys and the RYR1 variants p.Ser1728Phe and p.Leu4824Pro with posterior probabilities of 0.978, 0.979 and 0.926, respectively. The CACNA1S variant p.Pro1839Ser had been reported as VUS [43], but application of Bayesian methods produced posterior probability of pathogenicity of <0.001 (consistent with benign status).

Case details for subjects with STAC3 & CACNA1S variants

The variable penetrance of MH and wide range of symptoms and signs in MHS patients with RYR1 variants have been recognized [38,49–51]. Less information is available from MHS patients with other gene variants; therefore, we include the case details of these subjects with STAC3 variant p.Met187Thr and CACNA1S variants p.Pro1839Ser and p.Thr1009Lys (with RYR1 variant p.Arg274Cys in the same subject; Supplementary Material).

Discussion

In this study, we examined three genes, RYR1, CACNA1S and STAC3, believed to harbor 75% or more of MH-causative variants [14]. Since pathogenic variants in STAC3 were found relatively recently, few published studies have described complete sequencing of these three genes in the same population sample. Miller et al. recently published a study from the UK in which these three genes were sequenced from 722 MH families [14]. Our study evaluated subjects living in the USA, identified additional novel variants and case details for select variants, and provided Bayesian statistics to predict pathogenicity. Our results did confirm previous reports, revealing many recognized pathogenic variants and many VUSs in RYR1 [10,11,48,52,53].

Bayesian modeling of the ACMG/AMP variant classification guidelines was proposed by Tavtigian et al. in 2018, and we applied that framework to variants associated with MH. The original article presented a probability threshold of 99% for pathogenicity, indicating >99% certainty that a variant is disease causing. The posterior probability cutoff for pathogenic was thus 0.99. The ACMG/AMP guidelines are less clear, as they defined ‘likely pathogenic’ to mean >90% certainty that a variant causes disease, and the threshold for ‘pathogenic’ is higher (but not clearly defined). We maintained the threshold for pathogenic at 0.99, but an argument could be made for a lower threshold. The threshold should likely be modified based on the effective balance of risk and benefit to patients for the disease of interest. The 0.99 threshold may be appropriate for diseases such as cancer when prophylactic surgery may be chosen in the presence of certain pathogenic variants. However, a lower threshold may be more appropriate for MH due to the risks associated with underclassification of a variant as pathogenic. Another recent study described a quantitative approach to variant classification in hypertrophic cardiomyopathy, and the authors chose an even lower pathogenic threshold of 0.95 [54].

We have introduced new considerations for assessing the pathogenicity of genetic variants in MH beyond the standard application of the ACMG/AMP criteria. Data from sequencing and the ex vivo CHCT were applied to the computation of probability that a variant is pathogenic. While not elevating any previously identified variants to pathogenic, this modification did increase the posterior probability from the range of 0.15–0.20 to 0.6 or greater. The higher probability of pathogenicity suggests that a variant merits more attention.

Importantly, we were able to combine new and previously published case data to predict that three variants, CACNA1S p.Thr1009Lys and RYR1 p.Ser1728Phe and p.Leu4824Pro, are likely pathogenic for MH. The posterior probability for two of these variants exceeded 0.975 after applying data from other studies. Miller et al. [14] provided evidence for assignment of a PS4 criteria for the two RYR1 variants showing that the frequency is significantly greater in the MHS population than the general population (p < 1 × 10^-7).

The RYR1 variant, p.Thr3711Arg, was found in a subject with a known EMHG pathogenic variant (p.Arg614Cys) and no CHCT, but a very high CGS score of 73. The same variant was found by Brandom et al. [38] as a single variant with a strongly positive CHCT in a second individual. These two subjects with the p.Thr3711Arg variant were siblings (as the current study and the 2013 Brandom et al. study shared a few subjects); therefore, it is possible that some other shared genetic element contributed to their MH risk. The PS3 criterion was assigned to this variant with an α = 0.677 because there was sequence data for RYR1 only from the sibling with the positive CHCT.

The CACNA1S variant Thr1009Lys was described in a family by Kim et al. [24], and the pedigree analysis showed MHS transitions to MH equivocal and even MHN over three to four generations (suggesting the variant is subject to epigenetic or other modification in successive generations). Additional data will be required to determine if this is a consistent pattern for this variant or restricted to the reported family.

About 10% of the subjects in this report had multiple variants that might be responsible for the MH phenotype. Similarly, Miller et al. [14] reported that 9.2% (27/293) families, in which full 3-gene sequencing was done, carried two or more potentially pathogenic RYR1 variants. These findings demonstrate the need for complete sequencing of RYR1, and preferably all three of these genes, in an MH proband, as opposed to targeted sequencing restricted to specific loci. Relatives who could be identified as MHS by genetic testing might conclude that they are not at increased risk of MH if only one exon of RYR1 was to be examined for a familial mutation [38,55]. If full sequencing of the genes known to be associated with MH is performed for the first degree relatives of probands, then those with one of the two variants found in the proband will be identified. This reinforces the importance of collecting broad genetic data from families. Unfortunately when relatives have healthcare provided by different systems, personal data cannot be shared without the initiative of the proband. All MH probands and their relatives should be encouraged to send their medical records with their genetic reports to a research registry such as the North American Malignant Hyperthermia Registry.

Only benign or likely benign, or no potentially pathogenic variants were found in RYR1, CACNA1S or STAC3 in about a third (11/38) of the MHS subjects in this study, as well as 10/19 subjects with MH history/MH relative but lacking CHCT results. Two possibilities may explain these findings – either the subjects were not truly MHS or other, yet unidentified, genetic loci are involved in MH. Misdiagnosis of MHS is possible for some subjects since specificity estimates for CHCT are as low as 78% [56], and specificity decreases as the force of the contractures approaches the lower limits of a positive test. In two previous studies in different geographic regions, MHS subjects with no RYR1 variants had lower muscle contracture results than did those with RYR1 pathogenic variants [38,53]. Seven of the 11 subjects with only benign variants in our study had strong evidence by CHCT for MHS with halothane contracture results >1 g; however, caffeine contracture results were weaker for some subjects. CHCT specificity might best be determined by the weaker of the halothane or caffeine contractures [56]. The binomial probability of >4 false positives among 38 tests, with a specificity of 80%, is 0.96; thus, half of these cases may be false positives. The CHCT results from the remaining four subjects with benign variants were not definitive, as three had nonstandard test protocols or qualitative results, and one subject was positive only on the caffeine contracture test. One individual, with both halothane and caffeine contractures greater than the median in this small group, had a sibling with a recognized MH causative mutation. Perhaps a second locus of MHS exists in this family [57].

Other studies have also identified subjects diagnosed MHS by contracture testing that had no variants in RYR1 [11,38]. Miller et al. found no pathogenic variants in RYR1, CACNA1S, or STAC3 in 103/722 MH families (14%) and estimated non-RYR1/CACNA1S/STAC3 MHS between 14 and 23% [14]. Whole exome and Sanger sequencing targeted to genes involved in calcium movement in skeletal muscle have been used to search for other MH genes [23,58,59]. Variants of low frequency were found in the CACNB1, CASQ1, CASQ2, SERCA1, KCNA1 and other genes; however, a causative role in MH has not been proven yet. Kraeva et al. reported that variants in the CASQ1 gene are unlikely to be a cause of MH in the North American population, despite evidence that CASQ1 deletion is a candidate for MHS in mice [60]. Additional studies are needed to search for other genes involved in MH.

For two of the subjects with RYR1 VUSs (a previously identified variant, p.Val2280Ile and a novel variant, p.Arg4321Leu4323dup), we reported exertional myopathy (Supplementary Material). Neither family had a history of suspected MH events or muscle contracture testing, so MHS could not be defined in these individuals. Additionally, a novel STAC3 VUS was found in one study subject. This individual had an anesthetic-induced episode that was consistent with either MH or hyperkalemic cardiac arrest due to an occult myopathy. Because these VUSs were identified in single subjects, and relatives were not available to study, it is impossible to assess segregation of these genetic variants with MH. Nevertheless, documentation of these variants with clinical details may prove useful for future reference as new methods become available. Additional clinical reports and functional testing might lead to reclassification of these variants as pathogenic for MH. For example, RYR1 mutations at amino acid 2508 were added to the EMHG list of MH causative variants after functional testing [61] and multiple reports of MH events in patients with the variant.

Applying the ACMG/AMP criteria to estimate the pathogenicity of genetic variants in MH is controversial. At the time of writing of this manuscript, other researchers are actively working in this area. The concepts presented here are independent of any process or product of international debates on estimating pathogenicity, except in so far as we have responded to reviewers’ comments. Some reviewers may be part of these groups working on developing guidelines for interpretation of pathogenicity. As standards are introduced, some of the decisions made in our data analysis may become obsolete. We aimed to present enough details that the reader could re-analyze using different judgments regarding the contribution of elements of the criteria first published by Richards et al. [26].

Conclusion

We applied in silico analysis and Bayesian statistics to genetic variants from an MHS population in the USA, thus aggregating new and older data to assign a probability of pathogenicity to variants in RYR1, CACNA1S and STAC3. Three previously described variants, CACNA1s p.Thr1009Lys and RYR1 p.Ser1728Phe and p.Leu4824Pro, were predicted to be likely pathogenic for MHS with very high probability. This method could help to accelerate expansion of the library of MH pathogenic variants, and it leverages the data obtained from whole gene sequencing. As nearly a third of MHS subjects had only benign variants in RYR1, CACNA1S and STAC3, other genes are likely involved in MH.

Future perspective

Additional studies in MHS populations will likely identify more genetic variants in RYR1, CACNA1s and STAC3, as well as other genes (e.g., those involved in excitation–contraction coupling and skeletal muscle calcium homeostasis). Bayesian modeling will be used to integrate data and predict pathogenicity for certain variants, therefore supporting/expediting the use of functional studies. Complete or nearly complete characterization of the library of pathogenic MH variants would allow for clinical genetic testing to replace muscle bioassays as the most widely applied diagnostic tool for identification of MHS patients. Genetic testing performed after a suspected MH event will provide information to facilitate the future care of the proband and family members.