INTRODUCTION

Malignant hyperthermia susceptibility (MHS) is a potentially lethal inherited disorder of skeletal muscle calcium signaling, predisposing individuals to a hypermetabolic reaction triggered by exposure to inhalational anesthetics or depolarizing muscle relaxants such as succinylcholine.1,2 Inheritance of MHS is predominantly autosomal dominant, although autosomal recessive inheritance has been reported3 and non-Mendelian models proposed.4 Variants in RYR1 (MIM: 180901; MHS1, MIM: 145600) and CACNA1S (MIM: 114208; MHS5, MIM: 601887) have been associated with MH, and both genes are in the American College of Medical Genetics and Genomics (ACMG) return of secondary findings recommendations.5,6 RYR1 variants account for ~76% of MH events while ~1%7 are attributable to CACNA1S and <1% are attributable to STAC3 (MIM: 615521; Bailey–Bloch myopathy, MIM: 255995). Four additional loci have been mapped (MHS2, MIM: 154275; MHS3, MIM: 154276; MHS4, MIM: 600467; MHS6, MIM: 601888). RYR1 has a complex gene-to-phenotype relationship, being associated with several apparently distinct disorders and both autosomal dominant and autosomal recessive inheritance. Overlapping conditions include central core disease (CCD, MIM: 117000) and King–Denborough syndrome (MIM: 145600) and individuals with these disorders may be at risk for MH. Generally, these disorders result from monoallelic RYR1 variants while biallelic variants cause other myopathies; however, this correlation is evolving.8

Classification of RYR1 variants is complicated by variable expressivity, reduced penetrance, and high allelic heterogeneity. While the European Malignant Hyperthermia Group (EMHG; http://www.emhg.org/home/) has assessed 48 RYR1 variants as diagnostic of MHS, more than 165 additional variants have been reported as disease variants/pathogenic/likely pathogenic for MH in the literature and databases including HGMD9 and ClinVar.10 While the ACMG/Association for Molecular Pathology (AMP) guidelines11 provided general criteria that can be used to classify variants, many of the criteria require adaptation to be accurately applied. As part of ClinGen, we convened an RYR1-related malignant hyperthermia variant curation expert panel (https://clinicalgenome.org/affiliation/50038/) to adapt the general ACMG/AMP pathogenicity guidelines to autosomal dominantly inherited RYR1/MH, with gene-specific recommendations, to improve classification of RYR1 variant pathogenicity.

We first reviewed each ACMG/AMP criterion to determine applicability to autosomal dominantly inherited RYR1/MH and then adapted them with gene/disease-specific guidelines, if appropriate. We piloted these guidelines on 84 variants: 44 variants from the EMHG list of “diagnostic mutations” and 40 variants with MH pathogenicity classifications in ClinVar.

MATERIALS AND METHODS

ClinGen’s RYR1/MH expert panel

The RYR1/MH expert panel (EP) is composed of clinical molecular geneticists, clinical geneticists, anesthesiologists, biochemists, and physiologists to provide a balance of expertise relevant to RYR1 variant classification. The RYR1/MH EP met monthly via conference calls over a two-year period.

Evaluation and adaptation of the ACMG/AMP pathogenicity guidelines

The general ACMG/AMP pathogenicity guidelines11 were evaluated for relevance to autosomal dominantly inherited RYR1/MH and nonrelevant criteria were dropped. ClinGen-recommended amendments to the criteria were incorporated when applicable. Lastly, applicable criteria were further assessed to determine if gene-specific recommendations were warranted. Proposed changes were discussed among the full EP by emails and conference calls. Approval of revised rules required consensus of the full EP. Draft rules were piloted on a subset of RYR1 variants representing the EMHG “diagnostic mutation” list. Individual panel members scored variants using the draft guidelines and variant classifications were presented to the full panel. Areas of disagreement were used to refine the draft guidelines. Per the ClinGen FDA-approved process, rules were reviewed by the ClinGen Sequence Variant Interpretation (SVI) committee (L.G.B. recused).

Data collection methods

Population data was ascertained from gnomAD v2.1.1.12 REVEL scores (v0.19.1) were used for bioinformatic predictions for single-nucleotide variants (SNVs).13 The literature was searched for relevant data including case information and functional data. For case information, the number of unrelated probands with either a personal or family history of an MH event was recorded (see Supplemental information). Care was taken to avoid double counting cases reported multiple times. Reports were examined for instances of de novo inheritance and/or segregation.

Pathogenicity assessment

Revised ACMG/AMP criteria were used to assess 44 EMHG MH “diagnostic mutations.” Four of 48 EMHG variants were excluded because they were associated with RYR1-related myopathies and not MH. An additional 40 ClinVar RYR1 variants were also classified. Individual criteria were weighted based on available evidence and weighted criteria were combined using the Bayesian framework for variant scoring.14

RESULTS AND DISCUSSION

The ACMG/AMP guidelines11 are generic and broadly useful for all Mendelian genes and disorders. These generic rules may over- or underestimate evidence for any specific gene and must be adapted for specific implementations. As an EP, we suggest guidelines to be used/dropped, guidelines to be refined, and weight adjustments where appropriate. A summary of revised guidelines is in Table 1 and a full description is in Table S1 with gene/disease-specific adaptations highlighted below (updated versions will be maintained at clinicalgenome.org).

Table 1 Modified American College of Medical Genetics and Genomics (ACMG)/Association for Molecular Pathology (AMP) criteria suggested for autosomal dominantly inherited RYR1/MHa.
Full size table

Criteria dropped for MH/RYR1: PVS1/PM2/PM3/PM4/PP2/PP4/BS4/BP1/BP3/BP5

These criteria were dropped based on the biology of MH/RYR1. See Supplemental information for details.

Criteria used according to general guidelines: PS1/PS2/PM5/PM6/PP1/BP2/BP7

These criteria were retained in the RYR1/MH-specific guidelines including adaptations as recommended by the ClinGen SVI Committee (PS2/PM6, weighting of de novo observations, https://clinicalgenome.org/site/assets/files/3461/svi_proposal_for_de_novo_criteria_v1_0.pdf) and the Cardiomyopathy EP (PP1, weighting segregation events).15 We made further modifications to the ACMG/AMP criteria, which may not be specific to RYR1/MH. The PS1 (same amino acid change, different nucleotide change) and PM5 (different amino acid change, same codon) criteria were modified such that to use either of them, a previously classified variant must have been classified as pathogenic without the use of PS1 or PM5. Furthermore, for PM5, we added a requirement that the Grantham score difference compared with reference of the new variant must be greater than that for the previously identified pathogenic variant compared with reference. For criterion BP2 (evidence against pathogenicity based on presence of known pathogenic variant) it is suggested that only variants identified in cis with the variant under review be considered. Because the occurrence of biallelic pathogenic RYR1 variants has been described in MHS,3,16 two variants in trans is not considered evidence against pathogenicity. Finally, as RYR1/MH primarily results from missense alterations, BP7 (synonymous variant without splicing effect) is used as recommended.

Criteria specified for RYR1/MH: BA1/BS1/PS4/BS2/PS3/BS3/PM1/PP3/BP4

Allele frequency specificiations: BA1/BS1/PS4

BA1 and BS1 use minor allele frequencies (MAF) in population data sets to support benign classification for common variants. The BA1 criterion is considered standalone and was originally set to 0.05 (5%) MAF.11 It has been suggested that BA1 can be defined as the combined MAF for all pathogenic variants in the population for the gene/disease dyad with the understanding that any one variant should have a lower MAF than the combined total. To determine a gene/disease-specific cutoff for BA1, disease prevalence, penetrance, and gene contribution need to be considered. This can be estimated by the formula: \((\frac{{\left[ {{\mathrm{disease}}\;{\mathrm{prevalence}}} \right]{\mathrm{x}}\left[ {{\mathrm{\% }}\;{\mathrm{gene}}\;{\mathrm{contribution}}} \right]}}{{\left[ {{\mathrm{penetrance}}} \right]}})\).15 The prevalence of MH (defining the disorder as MH, not MHS) in the population can be estimated using the frequency of MH events in individuals exposed to triggering agents. The frequency of events is as high as 1/10,000 pediatric anesthesias.2 The rate of adult MH events seems lower than that of children17 but the underlying genetic risk is assumed to be the same. The gene contribution of RYR1 to MH is ~76% depending on ethnicity.7 Calculating thresholds for BA1 relies on an accurate estimate of penetrance, which is difficult to determine for MHS.18 In lieu of using an estimate for MHS penetrance, we instead substituted a value of 1%, as it is a reasonable boundary between the penetrance of a Mendelian disorder variant and that of a risk allele. This value is nearly certain to be lower than the actual penetrance of MHS, but underestimating this value is conservative with respect to the outcome in that it will numerically raise BA1, which would lead to fewer variants being classified as benign based on this single criterion. Using 0.01 to adjust our calculated BA1 allows for a BA1 MAF of 0.0038 (0.38%).

In addition to a standalone MAF (BA1), BS1 defines the MAF at which a variant is considered to have strong evidence against pathogenicity. The field has been moving to define BS1 based on the contribution of the most common pathogenic allele for a disorder. For RYR1/MH, we calculated BS1 considering the frequency of MH reactions in children (1/10,000) a value of 0.01 substituted for penetrance (as explained above), and a maximum individual allele contribution of 16% (variant c.7300G>A was identified in 118/722 MH families, 16.3%).7 Correcting for alleles/person gives a BS1 value of 0.0008 (0.08%).

While a high MAF of a variant in controls can be used to refute pathogenicity, criterion PM2 gives weight for absence or very low frequency in control populations. Based on observations that the majority of possible RYR1 missense variants (~30,000 variants) are not represented in gnomAD v2.1.1 (2,800 RYR1 missense variants) and many known pathogenic variants (classified without the use of PM2) are present in gnomAD, it is unlikely that the absence of a variant in gnomAD is support for pathogenicity. While the absence or low frequency of a variant in gnomAD has little value alone, it is important in weighting PS4. PS4 takes into consideration the prevalence of the variant in affected individuals compared with controls. For RYR1/MH, we modified the PS4 criterion using a point system, awarding 0.5 case points for each unrelated proband reported to have undergone an MH event and awarding an additional 0.5 case points for a positive in vitro contracture test (IVCT) or caffeine–halothane contracture test (CHCT) in either the proband or a variant-positive family member. The strength level of PS4 is based on odds ratios comparing total case points, an approximation of the total number of cases of MH investigated in the literature (3,000) and the number of alleles in the gnomAD continental population with the highest MAF (popmax). When the popmax MAF is ≤0.00006 (~7/113,000 alleles), strength levels are awarded according to the following system: PS4 for ≥7 MH case points, PS4_Mod for 2–6 MH case points, and PS4_Sup for one MH case point. When gnomAD popmax MAF is >0.00006, case points can be counted and compared with alleles in the gnomAD population with the highest MAF by calculating an odds ratio (OR, MedCalcs online calculator (https://www.medcalc.org/calc/odds_ratio.php), awarding PS4 for an OR ≥ 18.7; PS4_Mod for an OR ≥ 4.33; and PS4_Sup for an OR ≥ 2.08. Every effort needs to be made to avoid double counting of cases reported in multiple studies. The Bayesian framework for the classification of variants using the ACMG/AMP criteria was used to set the OR value for each strength level.14

Disease-specific phenotype: BS2

The IVCT/CHCT diagnostic tests have low false negative rates19,20 and can be used to determine MHS status in individuals who carry RYR1 variants. A negative IVCT or CHCT result supports benign status. Two or more unrelated individuals with a negative result allow BS2 to be applied. One individual with a negative result allows BS2_Mod.

Functional assay specifications: PS3/BS3

Functional characterization is considered a crucial determinant of the pathogenicity of RYR1 variants.21 Within the ACMG/AMP guidelines, functional assay results are used for PS3 (well-established in vitro or in vivo functional studies supportive of a damaging effect) and BS3 (well-established in vitro or in vivo functional studies show no damaging effect on protein function or splicing). RYR1 is a homotetrameric calcium channel in the sarcoplasmic reticulum (SR) of skeletal muscle important in excitation–contraction coupling. Volatile anesthetics and depolarizing muscle relaxants can cause increased release of SR calcium in a dysfunctional RYR1 channel resulting in MH. When considering functional assays for variant assessment it is desirable to identify assays that are closely related to the physiologic defect causative of disease. For RYR1/MH, assays that measure release of calcium in response to pharmacologic agents are considered good representations of the disease mechanism. Well-recognized assays include transfection of RYR1 complementary DNA (cDNA) into HEK293 cells, CHO cells, or RYR1 knockout myotubes followed by SR calcium release measurements in response to caffeine, halothane, voltage/potassium, or 4-chloro-m-cresol. A significant decrease in the EC50 for the sensitivity of calcium release compared with wild-type (WT) RYR1, is considered evidence for pathogenicity. Multiple replicates for each variant within a single instance of the assay are necessary to determine significance of these values. Positive (pathogenic) and negative (benign) controls support that the assay categorizes the variants accurately. For the purpose of assessing RYR1 transfection studies to weight PS3, results were dichotomized into pathogenic EC50 values that are significantly decreased as compared to WT versus benign EC50 values that are not significantly decreased. For RYR1 pathogenicity assessment, the whole of prior published work (Fig. 1, Table S2)22 allows us to consider transfection assays in HEK293 cells using photometry/imaging to measure calcium release a well-defined functional test. However, recommendations for increased stringency in analyses of functional data have recently been suggested.23 To determine the appropriate PS3 weight based on HEK293 transfection assays we have considered published results including results for 35 variants assessed to be likely pathogenic or pathogenic (LP/P) without the use of functional data, and ten control variants including eight variants associated with CCD and two common variants. Of the 35 LP/P variants, 29 have been shown to reduce the calcium release EC50 in response to RYR1 agonists. Five variants have shown discordant results across assays, and one variant has shown an EC50 increase. Of the ten control variants, one variant has shown an EC50 reduction in response to agonist and nine variants have either shown no response to agonist (6) or a response similar to WT (3). This set of variants suggests a likelihood ratio for an EC50 reduction of 9.11:1 with a 95% confidence interval of 1.4:1 to 59:1. This level of support is above the threshold for moderate evidence (4.33:1 odds). We suggest that functional evidence supporting pathogenicity from HEK293 cells be used at the level of moderate. When the field generates additional data for control variants the weighting of PS3 for this assay should be reconsidered.

Fig. 1: Cumulative HEK293 transfection assay data for RYR1 variants from the literature.
figure 1

Variants are grouped according to pathogenicity assessment without consideration of PS3/BS3 (functional data). CCD central core disease, MHS malignant hyperthermia susceptibility, P/LP pathogenic/likely pathogenic, SNP single-nucleotide polymorphism, VUS variant of uncertain significance.

Full size image

While positive evidence (reduced EC50) is considered moderate support for pathogenicity, reduced penetrance and the limitations of expression systems24 suggest a nonsignificant change in EC50 values may not support benign status at a moderate level (Fig. 2). It was decided that lack of response to agonists be weighted as supporting evidence (BS3_Sup). Regarding other in vitro assays that test calcium release in response to agonists, where historical data were limited, we suggest that multiple controls be run in parallel and statistical analyses be used to determine the level of strength for PS3 according to the Bayesian framework.

Fig. 2: Decision tree for weighting functional evidence PS3/BS3.
figure 2

cDNA complementary DNA, MH malignant hyperthermia, SR sarcoplasmic reticulum.

Full size image

In addition to in vitro assays, the RYR1/MH field has established ex vivo assays measuring calcium release in patient cells. These assays do not isolate the RYR1 variant from other potential variants (in RYR1, CACNA1S, or other MHS-associated genes), which may affect calcium release. Rather, these assays are a measure of the cellular phenotype in the patient. Although we recognize this limitation of ex vivo studies, we also recognize that they have utility. As the main concern for such assays is the potential presence of other variants, this concern is mitigated if multiple unrelated individuals with the same primary variant are shown to exhibit enhanced ex vivo sensitivity to agonist. Two unrelated individuals with ex vivo tests showing increased sensitivity of calcium release in response to agonist allow PS3_Sup. For variants where ≥3 unrelated individuals had ex vivo tests showing increased sensitivity of calcium release, PS3_Mod can be applied. Ex vivo tests that do not show increased sensitivity of calcium release in response to agonist (negative result) support a benign classification of the variant. BS3_Sup can be applied if one or two unrelated individuals are tested with negative results, when ≥3 unrelated individuals are tested and all results are negative BS3_Mod can be applied.

Knock-in mouse models created to date to test RYR1 variants have shown MH reactions in response to volatile anesthetic and ex vivo studies of muscle samples from these mice show increased ligand sensitivity of calcium release as compared with WT.25,26,27,28 When knock-in mice have an MH reaction in response to agonist, and where ex vivo studies show increased calcium release compared with WT in response to agonist, PS3 can be awarded. For mouse models where either an MH crisis can be triggered by agonist or ex vivo assays show increased calcium release, but both conditions are not met, PS3_Mod can be awarded. For mouse models that do not exhibit an MH reaction when exposed to agonist and ex vivo studies do not show increased release of calcium, BS3_Sup can be awarded.

Hotspot specifications: PM1

The ACMG/AMP criteria includes moderate weight for variation in critical protein domains or mutational hotspots (PM1). While critical domains may be well-defined for a protein, the concept of mutational hotspots is less clearly defined. A general rule for consideration of a mutational hotspot would be an excess of pathogenic variation as compared with benign variation. In MH, variants have been noted to cluster in three regions of RYR1 identified as hotspots historically: the N-terminal region (residues 1–552), the central region (residues 2,101–2,458) and the C-terminal region (4,631–4,991).29 Rather than defining clear functional domains, these regions are defined by an increase in variation identified in individuals with MH. We assessed this criterion using a test set of 19 variants (Table S3) assessed to be pathogenic for MH without the use of PM1 and 27 benign variants (Table S4) that met criterion BA1. This set of variants suggests a likelihood ratio for hotspots of 24:1 with a 95% confidence interval of 3.4:1 to 163:1 (Table 2). This level of support is above the threshold for strong evidence (18.7:1 odds) and the lower bound of that confidence interval is above supporting (2.1:1). This would suggest that PM1 could be modified to PM1_strong. However, because there is a significant bias in the literature toward identifying pathogenic variants in the hotspots, and to avoid the possibility of overestimating pathogenicity, we suggest using PM1 at its default level of moderate for variants in the N-terminal and central regions. As variants in the C-terminal region may be associated with CCD and not cause MH, we suggest using PM1_supporting for variants in this region. Future studies that interrogate the gene without these biases should provide additional data on the positional skewing of pathogenic variants, which could allow upgrading PM1 to strong in the future.

Table 2 Distribution of 19 pathogenic and 27 benign variants in relation to position of defined RYR1/MH hotspots.
Full size table

Computational evidence: PP3/BP4

The PP3 and BP4 criteria consider computational evidence estimating the impact of a variant on protein function. REVEL is an ensemble method based on a number of individual tools and precomputed scores are available for all missense variants (https://omictools.com/revel-tool).13 Importantly, REVEL does not consider population frequency, which reduces double counting of evidence. Using a set of 20 pathogenic variants determined to be pathogenic without the use of PP3 and 27 benign variants described above, we tested the likelihood ratios of the predictive power of REVEL in several iterations. We settled on a trichotomization of scores with PP3, (computational evidence supporting pathogenicity), requiring a REVEL score of ≥0.85 and BP4, (computational evidence against pathogenicity), requiring a REVEL score of ≤0.5 (Table 3). Based on the Bayesian model for weighting criteria, these results suggest that PP3 and BP4 could be employed at the strong level. However, based on wide confidence intervals of the likelihood ratios for this conditional probability, we chose to weight PP3 as moderate and BP4 as supporting.14 Based on piloting these criteria it was determined that BP4 should only be implemented with other criteria. Using the Bayesian framework, BP4 in isolation results in an assessment of likely benign (LB) and it was determined that additional evidence should be available for a LB classification. For a fuller explanation of deriving such likelihood ratios, see Supplemental information.

Table 3 REVEL score distribution for 20 pathogenic and 27 benign variants for RYR1/MH.
Full size table

Piloting RYR1/MH classification criteria

We applied these modified criteria to 44 variants EMHG determined to be “diagnostic mutations” and 40 RYR1 variants with pathogenicity classifications for MH in ClinVar. The classification of each of the variants is shown in Table S3 and Table S5. Of the 44 EMHG variants, we classified 29 as P, 13 as LP, and 2 as variants of uncertain significance (VUS). Variant c.1589G>A p.(Arg530His) was classified as VUS and had limited functional data including a single ex vivo sample,30 which did not meet PS3_Sup based on the requirement for a minimum of two unrelated individuals. Variant c.1598G>A p.(Arg533His) was classified as VUS based on functional data (PS3_Mod) and presence in a hotspot (PM1). PS4 was not met by this variant based on a high allele count (32 alleles) in gnomAD.

The revised criteria were applied to 40 additional variants with pathogenicity classifications for MH in ClinVar. Ten variants had conflicting pathogenicity classifications for MH (pathogenicity classifications for disorders other than MH were not considered), nine B/LB/VUS, and one P/LP/VUS. Five variants with B/LB/VUS classifications in ClinVar were determined to be B/LB based on BA1/BS1. The remaining five discordant variants were classified as VUS. Of the remaining 30 variants, 14 were classified as P/LP, 11 as B/LB, and 5 as VUS. Applying the revised ACMG/AMP criteria 12/14 variants with a classification of P/LP in ClinVar and 3/11 variants with an classification of B/LB in ClinVar were classified as VUS. All variants classified as B/LB (13) using our criteria had ether BA1 or BS1 applied. The 19/24 variants classified as VUS had limited data; only 5 VUS variants had data that refuted pathogenicity (5/24, 21%).

Conclusions

As a ClinGen expert panel, we set out to adapt the ACMG/AMP pathogenicity criteria for classification of RYR1 variants as related to autosomal dominantly inherited MH. Combining expertise of anesthesiologists, physiologists, biochemists, and geneticists allowed for a thorough evaluation of factors that should be considered. It is also important to recognize that we successfully unified the efforts of the American-based ACMG/AMP criteria with the extensive expertise and experience of the EMHG, benefiting from both. In revising these guidelines, we have considered the statistical evidence weight as it relates to the Bayesian adaptation of the ACMG/AMP scoring system. Weighting of evidence using statistical measures should allow for a more robust and consistent pathogenicity classification framework and is broadly applicable to other disease/gene systems. The revised RYR1/MHS specific criteria should allow clinical laboratories to more consistently classify these variants based on expert guidelines and should increase the consistency of classifications, as has been demonstrated for the generic ACMG/AMP pathogenicity recommendations.31 These recommendations should be especially useful to laboratories that classify RYR1 variants as secondary findings. That MH is a pharmacogenetic trait with relatively low penetrance makes it especially challenging to classify for laboratories that do not perform a high volume of diagnostic RYR1 testing. The availability of three star ClinGen classifications in ClinVar should significantly reduce the amount of time that secondary findings evaluations consume. As well, the RYR1/MH expert panel will continue to curate variants and deposit classifications into ClinVar. Moving forward, the field should strive to increase relevant data through functional studies and shared case documentation allowing variants to move from a classification of VUS to either LB/B or LP/P. Beyond secondary findings, ClinGen classifications of RYR1 variant pathogenicity will allow the field to consider presurgical screening of patients toward elimination of MH morbidity and mortality.32