Catecholaminergic polymorphic ventricular tachycardia patients with multiple genetic variants in the PACES CPVT Registry

Background Catecholaminergic polymorphic ventricular tachycardia (CPVT) is often a life-threatening arrhythmia disorder with variable penetrance and expressivity. Little is known about the incidence or outcomes of CPVT patients with ≥2 variants. Methods The phenotypes, genotypes and outcomes of patients in the Pediatric and Congenital Electrophysiology Society CPVT Registry with ≥2 variants in genes linked to CPVT were ascertained. The American College of Medical Genetics & Genomics (ACMG) criteria and structural mapping were used to predict the pathogenicity of variants (3D model of pig RyR2 in open-state). Results Among 237 CPVT subjects, 193 (81%) had genetic testing. Fifteen patients (8%) with a median age of 9 years (IQR 5–12) had ≥2 variants. Sudden cardiac arrest occurred in 11 children (73%), although none died during a median follow-up of 4.3 years (IQR 2.5–6.1). Thirteen patients (80%) had at least two RYR2 variants, while the remaining two patients had RYR2 variants plus variants in other CPVT-linked genes. Among all variants identified, re-classification of the commercial laboratory interpretation using ACMG criteria led to the upgrade from variant of unknown significance (VUS) to pathogenic/likely pathogenic (P/LP) for 5 variants, and downgrade from P/LP to VUS for 6 variants. For RYR2 variants, 3D mapping using the RyR2 model suggested that 2 VUS by ACMG criteria were P/LP, while 2 variants were downgraded to likely benign. Conclusions This severely affected cohort demonstrates that a minority of CPVT cases are related to ≥2 variants, which may have implications on family-based genetic counselling. While multi-variant CPVT patients were at high-risk for sudden cardiac arrest, there are insufficient data to conclude that this genetic phenomenon has prognostic implications at present. Further research is needed to determine the significance and generalizability of this observation. This study also shows that a rigorous approach to variant re-classification using the ACMG criteria and 3D mapping is important in reaching an accurate diagnosis, especially in the multi-variant population.


Introduction
Catecholaminergic polymorphic ventricular tachycardia (CPVT) is a rare inherited arrhythmia syndrome characterized by ventricular tachycardia (VT) provoked by adrenergic stress. [1] The condition is caused by excessive calcium leak from the sarcoplasmic reticulum, leading to delayed after-depolarizations and arrhythmias. [1] Most cases are attributed to mutations in RYR2-coded ryanodine receptor (RyR2) or CASQ2-coded calsequestrin-2 [1]. Although less recognized, SCN5A, TRDN, and CALM1-3 have also been implicated in catecholamine sensitive polymorphic VT. [2][3][4][5][6][7] To date, genotype-based risk predictors have not been clinically useful. In other inherited arrhythmic conditions, like long QT syndrome (LQTS) and hypertrophic and arrhythmogenic cardiomyopathies, patients with double and compound mutations fare especially poorly. [8][9][10][11][12] We used the Pediatric and Congenital Electrophysiology Society (PACES) Registry [13,14] to characterize CPVT patients with �2 variants. To systematically assess the likelihood of pathogenicity, variants were mapped on to the 3D structure of RyR2, which provides mechanistic insights into their function and enhances the analysis compared to sequence-based scoring algorithms alone.

Material and methods
This is a retrospective study derived from the PACES CPVT Registry, which is an international multicenter registry of children (�19 years) and their first-degree relatives with a diagnosis of CPVT made by consensus criteria. [1] Clinical, genotypic and outcome data were previously reported. [13,14] Participating sites received ethical approval locally and the protocol adhered to the 1975 Declaration of Helsinki. The coordinating center was responsible for data collection and analysis, and the protocol was approved by the UBC C&W Research Ethics Board. The review board did not require individual patient consent as the study was a retrospective chart review. All data were de-identified prior to entry and analysis. A stepwise bioinformatics approach was implemented to classify variant pathogenicity, including structural mapping using 3D model of pig RyR2 in open-state (Fig 1). Continuous data are presented as the median (interquartile range). Detailed methods are available in the supporting supplemental information (S1 File and S2 File).

Population
Of 237 patients entered in the PACES CPVT Registry, 193 (81%) underwent genetic testing. There were 15 patients (8%) from 12 families with �2 variants. Table 1 summarizes the genotypes and phenotypes of these multi-variant carriers. The median age at presentation was 9 years (IQR 5-12) and 9 (60%) were female. Thirteen children (87%) had �2 RYR2 variants, one had CASQ2 and RYR2 variants, and one had RYR2 and SCN5A variants ( Table 1, Table 2 and S2 File). There were 12 children (80%) who were probands. Inheritance could not be established in 6 children (40%) owing to a lack of parental genetic data. A family history of  [15] The ribbon diagram shows a RyR2 dimer from the side. Red circles represent large alpha solenoid regions which are unstructured in the cryoelectron micrograph (CryoEM) structure. All variants discussed in the manuscript are shown in sphere representations. Red squares highlight location of the most likely damaging variants (expanded in Fig 2). Domains are named according to nomenclature used by des Georges, et al., 2016. A (dark orange), B (green), and C (light blue) domains: form part of the N-terminal domain (NTD); NSol (dark blue): alpha-solenoid region near the NTD; SPRY1 (light orange), SPRY2 (dark green), and SPRY3 (dark grey): three domains named after splA kinase and RyRs where they were first identified; Sol2 (light grey): second alpha-solenoid region centrally located on RyR2; Rep12 (yellow) and Rep23 (brown): four repeats (~100 aa each) in two tandem arrangements, Repeats 1 and 2 located between SPRY 1 and SPRY2, and Repeats 3 and 4 located within Sol2; CSol3 (magenta): third alpha-solenoid region located near the C-terminal; EFs (red): pair of EF hand-like motifs located within CSol3 region; TaF

Therapies & outcomes
Anti-arrhythmic therapy was instituted in 13 of 15 patients (87%). One patient received no treatment (subject #2), and in one patient, only an implantable cardioverter-defibrillator (ICD) was used (subject #6). A beta-blocker was prescribed in all treated patients. Therapeutic escalation with flecainide and left cardiac sympathectomy was necessary in subject #8 due to arrhythmias on beta-blockers. He continued to have events despite these ancillary treatments. Subject #12 also received flecainide for refractory arrhythmias and valproate for seizures, while subject #15 eventually had a left cardiac sympathectomy. An ICD was implanted in 10 of 15 patients (67%), 9 of which were for secondary prevention after SCA. Treatment failure occurred in 4 of 13 subjects (31%) on medication. No patients died during a median follow-up of 4.3 years (IQR 2.5-6.1).

Genetic analysis
Based on American College of Medical Genetics and Genomics (ACMG) criteria, 13 of 29 variants (45%) were defined as pathogenic/likely pathogenic (P/LP) after a review of the literature and population allele frequencies in the Exome Aggregation Consortium (ExAC) browser [16]. A remaining 16 (65%) were variants of unknown significance (VUS) by ACMG criteria. Table 2 and the supplemental (S2 File) summarize the variants in this population and the clinical and molecular data to support variant pathogenicity using our stepwise approach to classification. We then undertook a detailed analysis of RYR2 variants using the 3D structure of pig and mouse RyR2 [15,17] (Figs 1 & 2) based on several rationales. Firstly, predicting the pathogenicity of variants on sequence alone does not consider the chemical environment of the affected residues. Substitutions of amino acid residues involved in protein folding, domaindomain interactions, and interactions with auxiliary ligands are much more likely to affect function than residues simply pointing to solvent. This type of information is not available from sequence-based algorithms like the Polyphen score. Secondly, knowledge of the 3D environment can give possible clues on the disease mechanism. [

Discussion
In this study, multi-variant CPVT occurred in 8% of the PACES CPVT Registry. Nearly threequarters of these children were SCA survivors and many had variants that were likely pathogenic. While this cohort of multi-variant CPVT patients were severely affected, there are insufficient data to conclude whether this genetic phenomenon has prognostic implications. Even in the absence of prognostic utility, multi-variant CPVT is relevant to decision-making around cascade family screening. In total, there were 3 possible situations observed: (1) double variants in cis, (2) compound heterozygous variants in trans, and (3) digenic heterozygous variants (different genes). As will be discussed, each scenario creates a unique set of diagnostics, cascade screening and prognostic implications. The cumulative gene dosage phenomenon is similarly rare for hypertrophic cardiomyopathy, [9,10] arrhythmogenic right ventricular cardiomyopathy, [11] and LQTS [8,12], compared to CPVT in the PACES Registry. Multi-variant CPVT children were usually severely affected (73% with SCA). In comparison to the rest of the Registry population (38% with SCA), the severity of multi-variant CPVT appeared greater, but these data are not sufficient to imply that disease severity is directly influenced by the number of mutations. Statistical comparisons between single vs. multiple variant phenotypes were not undertaken due to the relatively small population described here, and the historical uncertainty around the genetic testing protocols/techniques which pre-dated enrollment in the registry. For example, early on, commercial genetic testing companies were only testing select RYR2 exons and older reports in the Registry were sometimes incomplete. Additionally, the second variant was sometimes a VUS, and/or potentially benign based on our mapping and/or found in cis phase. In such settings, CPVT may be driven by a single variant or the second variant could be a risk modifier. An example is RYR2-p.G1886S, which occurred in 2 of our patients, and was also seen in the general population. [16] A recent study has shown that this variant is a significant risk factor for ventricular arrhythmias in heart failure patients [21], suggesting that it could be a candidate risk modifier in CPVT. A parallel effect exists in the LQTS genes where common variants underlie a susceptibility to drug-induced QT prolongation [22,23] but do not cause overt LQTS. For example, KCNE1-p.D85N is too common in the population to independently cause LQTS, but significantly increases risk if a second LQTS mutation occurs. [24,25] In our cohort, subject #14 carried a CACNA1C VUS and a pathogenic RYR2 mutation (plus several other RYR2 VUS), and had a classic CPVT phenotype (catecholamine triggered bidirectional VT) with some QT prolongation. We hypothesize that a possible LQTS variant, plus multiple variants linked to CPVT, may have collectively contributed to his severe overlap phenotype (ie. digenic heterozygosity). Quite remarkably, another boy had two forms of CPVT (type-1 due to RYR2-p.T1107M and type-2 due to CASQ2-c.IVS5+1G>C). While CPVT type-2 alone can be especially dangerous, the RYR2-p.T1107M variant also likely has a damaging role based on the present data, and previous studies showing both a clinical and in vitro phenotype [19,26]. The growing international CPVT registry and prospective data are needed to clarify the risk and incidence of multi-variant CPVT.
The presence �2 variants creates other logistical problems. We could not differentiate between cis and trans variants in some cases owing to the inconsistent screening of the parents. Incomplete parental screening may be due to the clinicians' uncertainty regarding the diseasecausing variant in the family, thus demonstrating another challenge in this circumstance. Variants in cis phase similarly confounds screening in hypertrophic cardiomyopathy. [27] In cis phase, the CPVT phenotype would not necessarily be worse than any given single variant. However, the presence of �2 cis variants is relevant, as it demonstrates the complexities around family screening in the setting of CPVT. RYR2 variants in trans phase have a theoretical mechanism for increased severity. RyR2 is a large, homotetrameric protein made up of 4 subunits. Two variants in trans phase would mean that all 4 subunits making up the channel would be mutated. In contrast, in the typical case of autosomal dominant CPVT, half of the four subunits would be wildtype. Theoretically, this could account for a more severe phenotype in compound multi-variant CPVT. Based on the present study, we propose that targeted sequencing for both variants be performed in the clinical setting, and that relatives, especially parents, need to be evaluated by an expert to clarify the role of each variant.
The unclear pathogenicity of RYR2 variants is a growing concern in CPVT. [28] We used a standardized bioinformatics approach to variant interpretation to avoid overcalling pathogenicity. After applying the ACMG criteria, we mapped variants on the open-state structure of RyR2 to see if any other insights could be obtained. This technique relies on the 3D structure derived from high resolution studies of the ryanodine receptor. A good example of this is the N-terminal region of RyR2, which consists of three domains: domain A (residues 1-217) and domain B (residues 218-409), and domain C (residues 410-543) (Fig 1). [29][30][31] A chloride ion is coordinated by residues of all three domains and disruption of this binding site via disease-causing variants results in domain reorientations. [17,29,32] These observations suggest that CPVT variants may destabilize domain interfaces or disrupt the folding of individual domains, which would impair domain-domain interactions and cause adverse effects on channel gating. [17,20,29,[31][32][33][34] Structural analyses supported downgrading p.A1136V and p.R2028H to likely benign. RYR2-p.A1136 is located in a relatively non-conserved region within the SPRY2 domain, where its equivalent residue in both RyR1 and RyR3 is a valine, thus substitution to valine is unlikely to alter the function of RyR2 significantly. Further, structural mapping showed that valine substitution can be easily accommodated without the formation of steric clashes. The p.R2028H mutation is located in a flexible region of RyR2, where the side chain is pointing towards the solvent, and thus the mutation is unlikely to have a major impact on channel function. These are not functional assays, so the conclusions are predictive in nature.
This study is limited by its retrospective design. Genetic testing spanned nearly 15 years, and not all commercial testing companies provided technical details as would be required in the contemporary era. Results for family members were sometimes not available (often if followed by a non-participating center). Early commercial sequencing methods could not differentiate between two variants in the cis vs. trans position of RYR2 (unpublished communication). Limitations also exist in the structural analysis of variants, whereby some portions of RyR2 structure are poorly defined in the CryoEM structure. The best-defined regions are domains whose structures have been determined via X-ray crystallography (N-terminal domains, SPRY1/2, Rep12, and Rep34 domains), followed by the C-terminal and transmembrane regions, for which the resolution of CryoEM studies is the highest [18]. As such, direct analysis of variants in the N-terminal and C-terminal hotspots is the most reliable. For most other sections, direct analysis of hydrogen bonds and ionic interactions of the variants is not yet possible, however their general location in the 3D structure can be determined at the current resolution for RyR2. A detailed supplemental disclosing all the supporting data is provided to facilitate re-classification by future researchers as the field advances.

Conclusions
More than one variant may underlie a minority of CPVT cases. This poses challenges with respect to diagnosis and family counselling. While multi-variant CPVT patients were usually severely affected, further research is needed to determine the significance and generalizability of this observation. We demonstrate that a rigorous approach to variant re-classification using the ACMG criteria and 3D mapping is important in reaching an accurate diagnosis, especially in the multiple variant population.