Determining the Pathogenicity of Genetic Variants Associated with Cardiac Channelopathies

Advancements in genetic screening have generated massive amounts of data on genetic variation; however, a lack of clear pathogenic stratification has left most variants classified as being of unknown significance. This is a critical limitation for translating genetic data into clinical practice. Genetic screening is currently recommended in the guidelines for diagnosis and treatment of cardiac channelopathies, which are major contributors to sudden cardiac death in young people. We propose to characterize the pathogenicity of genetic variants associated with cardiac channelopathies using a stratified scoring system. The development of this system was considered by using all of the tools currently available to define pathogenicity. The use of this scoring system could help clinicians to understand the limitations of genetic associations with a disease, and help them better define the role that genetics can have in their clinical routine.

the pathogenicity of each genetic variation. However, they do not achieve an accurate numeric scale of pathogenicity, thus the pathogenic classification remains ambiguous. Another recent report highlights the current challenges in investigating causality of human sequence variants, suggesting that each variation be assessed by combining different tools to determine disease contributions 8 . In the present report we propose a scale to stratify pathogenicity for genetic variations associated with channelopathies. We integrate multiple parameters to reach a more definite conclusion for use in the clinic.
Channelopathies. Channelopathies are electrical disorders in structurally normal hearts caused by pathogenic variations in genes encoding cardiac ionic channels or regulatory proteins. These alterations modify the ionic balance of the electrical component of cardiac function and cause life-threatening arrhythmias 9 .
Long QT syndrome. LQTS is a genetic disorder characterized by a prolongation of the QT interval on the ECG (QTc . 480 ms). The clinical presentation can be variable, ranging from asymptomatic to episodes of syncope and SCD due to ventricular tachyarrhythmia (torsade de pointes) in the setting of a structurally normal heart. The syncopal episode may be induced by exertion, and emotional or auditory stimuli. SCD is the first event in 5% of asymptomatic LQTS individuals. The estimated prevalence is 1/2500, but the penetrance of the disease is not 100% 10 . In 1993, the first clinical diagnosis of LQTS was published 11 . At present, genetic analyses are included in the scoring criteria for diagnosis 5 . Thus, individuals and family members at risk with a normal ECG may be identified through genetic testing. Currently, nearly 1200 pathogenic variations have been identified in 15 genes (KCNQ1, KCNH2, SCN5A, ANK2, KCNE1, KCNE2, KCNJ2, CACNA1C, CAV3, CALM1, CALM2, SCN4B, AKAP9, SNTA1, and KCNJ5) 12 . Combined, these genes contribute to approximately 80%-85% of all LQT cases. However, 70%-75% of LQT cases are attributable to pathogenic variations only in 3 genes, KCNQ1 (LQT1), KCNH2 (LQT2), and SCN5A (LQT3), with KCNQ1 responsible for around 35% of cases. Genetic testing for LQTS has a major role in the diagnosis of index cases, risk stratification, family screening, and therapeutic strategies, but only for a small number of pathogenic variations 13 . Most identified genetic variations remain untranslated to the clinic.
Brugada syndrome. BrS is an inherited disease described 20 years ago and characterized by an ECG pattern consisting of coved-type STsegment elevation in atypical right-bundle branch block in leads V1 to V3 (type-1) 14 . BrS patients have an increased risk for SCD resulting from episodes of polymorphic ventricular tachyarrhythmia 15 . BrS accounts for 4%-12% of overall SCD and 20% of SCDs with normal hearts. The reported prevalence is 1 to 5/10000 in Europe and 12/ 10000 in Southeast Asia, with a ratio of 851 men5women affected. BrS typically affects young male adults (30-40 years old), and SCD typically occurs during sleep 15 . To date, more than 300 pathogenic variations in 16 genes (SCN5A, GPD1-L, SCN1B, SCN2B, SCN3B, KCNE3, KCNE5, KCNJ8, KCND3, CACNA1C, CACNB2b, CACNA2D1, RANGRF, HCN4, SLMAP and TRPM4) have been associated with BrS 16 . However, genetic testing only identifies the pathogenic cause in 35%-40% of clinically diagnosed cases of BrS. Approximately 25%-30% of diagnosed patients carry a pathogenic variation in the SCN5A gene 17 .
Catecholaminergic Polymorphic Ventricular Tachycardia. CPVT is a familial disease characterized by severe arrhythmias under adrenergic stimulation, such as exercise or emotional stress. The resting ECG is usually normal, but exercise testing induces ventricular arrhythmia in 75%-100% of patients. In many cases, the first manifestation of CPVT is the death of the patient. CPVT is associated with high mortality (around 30% by the age of 30 years), and the estimated prevalence is 1/10000 18 . To date, more than 100 pathogenic variations have been identified in five genes, contributing to nearly 60% of all clinically diagnosed cases (RYR2, CASQ2, KCNJ2, TRDN, CALM1, and CALM2). The RYR2 gene, encoding the ryanodine receptor (autosomal dominant), is responsible for nearly 50% of all cases 12 .
Short QT syndrome. First reported in 2000 19 , SQTS is considered the most lethal channelopathy; it has a high familial incidence of palpitations or syncope, and SCD, typically during childhood, is often the only phenotypic manifestation. It is characterized by a short QT interval on the ECG (QTc , 325 ms) with a high sharp T wave 20 . The incidence and prevalence are difficult to determine due to limited data. So far, pathogenic variations have been reported in 4 genes (KCNQ1, KCNH2, KCNJ2, and CACNA2D1), and these account for nearly 50% of clinically diagnosed SQT cases 12 .

Methods
In 2008, the American College of Medical Genetics and Genomics (ACMG) published guidelines for genetic variant interpretation 21 . However, most of the genetic tools outlined have been improved since 2008, and additional ones have been developed. Most of the current genetic data generated by high-throughput technologies has identified exonic common variants previously identified in the general population (minor allele frequency, MAF . 1%). Much of the remainder is classified as being of unknown significance, with only a small number of variants classified as pathogenic. To supplement the ACMG guidelines, we propose to incorporate new information to stratify the level of pathogenicity into 5 groups: pathogenic, probably pathogenic, unknown significance, probably non-pathogenic, and non-causal/benign.
Our scale of pathogenicity ( Figure 1) is based on the analysis of several items: -Clinical ascertainment (disease-associated gene or candidate gene) and determination of the genetic isoform in which the variation was identified; -Whether the variation has been previously identified in international databases; -Family segregation (identification of de novo variation may occur); and -In vitro, in vivo, and in silico evaluation.
We have categorized the scoring using a similar approach to that in the LQTS Diagnostic Criteria 11 . Our scores range from 0 to 15 points ( Figure 1). Each item receives a score between 0 and 3 points, and the total of the assessed items determines whether the exonic genetic variation could be considered pathogenic ($ 12 points), probably pathogenic (from 9 to 11 points), unknown significance (from 5 to 8 points), probably non-pathogenic (# 4 points), and benign. This last group comprises genetic variations considered as common in the global population (MAF . 1%). This fact implies that our score is useful only for non-silent variants identified in dominant disorders, in concordance to the diseases assessed in the present report. Finally, before final genetic interpretation of a variant, it is crucial to know the genetic isoform and tissue expression in order to perform an appropriate translation into clinical practice.

Results
Clinical ascertainment. The use of genetics as a clinical diagnostic tool must always consider the context of the clinical case. Therefore, genetics should complement, and never supplant, the clinical investigation. The advent of NGS, with its constant discovery of new associations of genes with diseases, prompts the need for the confirmation of pathogenicity in more patients and families. This is especially true in the study of SCD, in which segregation analyses are limited by the sizes of the families due to the sudden death of some affected individuals at a young age 22 . Without a clear clinical ascertainment, and a clear correlation between the disease and the genetic variant, causality cannot be conclusive.
Some proportion of the SCD cases arising from unknown genetic causes could be explained by pathogenic genetic variations in genes not yet associated with the disease. We hypothesize that the genes most likely to be associated with such cases are other genes encoding cardiac ion channels and/or regulatory proteins associated with them, independently of the cardiac channelopathy diagnosed. Recently, through the development and improvement of NGS technologies, several genes previously associated with only one channelopathy have been linked to other channelopathies; for example, the SCN5A gene contributes to both BrS and LQTS 23 .
-Genetic variation identified in a disease-related gene 3 points -Genetic variation identified in a novel gene 2 points (Encoding cardiac ion channels and/or associated regulatory proteins) -Genetic variation identified in a novel gene 1 point (Not encoding cardiac ion channels and/or associated regulatory proteins) Genetic databases. Several genetic databases are available, such as PubMed (www.ncbi.nlm.nih.gov/pubmed), the Exome Variant Server (http://evs.gs.washington.edu/EVS/), the 1000 Genomes Project (www.1000genomes.org), and the ClinSeq Project (www. genome.gov/clinseq). These databases catalog all the genetic data generated by published large-scale population studies, including pathogenic, potentially pathogenic, genetic variant of unknown significance (GVUS), and common genetic variations in the population. Other databases focus only on potential pathogenic variations [such as ClinVar (www.clinvar.com) and HGMD (www. hgmd.cf.ac.uk)]; however, caution should be exercised when regarding these classifications as not all characterizations are exhaustive and some of these classified variants could be benign. We believe that, for each reported pathogenic variation, the published source should be carefully analysed and interpreted. Importantly, only if a genetic variation has been reported as pathogenic in more than one study could it be suggested as damaging.
In our proposed stratification, this item determines if the variation is novel (never identified so far) or already reported. If reported, the variation could be considered pathogenic, of unknown significance, or neutral/non-causal. In addition, the genetic variation could be classified as a common (MAF . 1%) or rare (MAF , 1%) variation in the global population. (MAF , 1% does not imply certain pathogenicity).  www.nature.com/scientificreports Family segregation. Family history of syncope/SCD is a significant risk factor for other syncope/SCD in the family 5 . For this reason, familial genetic screening is recommended. However, the lack of information about pathogenicity of most known genetic variants implies that genetic screening does not always help to clarify the risk of syncope/SCD in a family.
In our proposed stratification, familial testing enables discernment of whether a variation is de novo or inherited. If de novo, the score should be high (2 points) because de novo variations are strongly associated with pathogenicity 27 . If a variant is suspected to be inherited, as many relatives as possible should be tested to clarify the role of variant. We believe that family segregation is the most important parameter that helps to clarify a variant's contribution, but should not be the only one investigated. For cases in which a clinically affected family member does not carry the potentially pathogenic variation identified in the index case (negative segregation), the genetic variation could be discarded, or at least ruled out as the main contributor to the disease, since it could be a modifier of the phenotype. If an asymptomatic family member carries the potentially pathogenic variation, the variation should not be discarded because of the possibility for incomplete penetrance. Therefore, in our stratification to determine pathogenicity, the presence of a minimum of 5 clinically affected carriers in at least 3 generations 28 denotes high genotype-phenotype correlation (3 points). A lower number of affected carriers is given a score of 2 points; when the index case alone carries the variant, 1 point is assigned. In vitro studies. In vitro studies are one of the main experimental confirmations of pathogenicity in cardiac channelopathies. However, it is not feasible to perform in vitro studies to characterize each identified genetic variation. Even when available, some in vitro functional data not accurately reflect in vivo physiology 29 . This phenomenon occurs because in vitro studies are performed in heterologous systems that do not include all the necessary biological partners that modify the final phenotype. Therefore, in our proposed stratification, if a cellular model is available for a variant and it mimics the biological mechanism(s) observed in vivo in humans (positive result), 2 points are tallied. In particular, ''radical'' genetic variations (nonsense and indels) have been assumed as pathogenic/deleterious because the protein will be altered; in consequence, 2 points are given in all these cases. As with the animal models, a score of 3 points is excluded from this category because in vitro studies cannot faithfully reflect the in vivo mechanistic pathway(s) in humans. If a cellular model is available but does not recapitulate the biological mechanism(s) observed in vivo in humans (negative result) or no model is available, 0 points are given. In our proposed scale, 0 points are assigned if bioinformatic analyses are not available for a variant in any of the computational databases, or, if they are available but all in silico analyses showed a neutral/benign prediction. To include the largest quantity of bioinformatic data possible, 2 points are assigned if more than 4 different databases agree on the predicted pathogenicity (deleterious, probably, and/or possibly pathogenic) and 1 point if less than 4 databases agree on the predicted pathogenicity of the variant. This last point includes, for example, 3 pathogenic and 1 benign, 2 pathogenic and 2 benign, or 4 benign predictions. This threshold is based on the fact that computational systems do not include all biological elements existing in humans. Regarding ''radical'' genetic variations, including nonsense genetic variations and frameshift/in-frame genetic variations, no database analyses them because it is assumed that the protein product will be altered; 2 points are given in all these cases. We do not assign a score 3 points because an in silico prediction cannot faithfully reflect a human mechanistic pathway, and the use of bioinformatic tools can produce erroneous conclusions regarding pathogenicity, as we reported recently 30 . A positive in silico result means that pathogenicity is predicted; in contrast, a negative result means that a neutral/benign variation is predicted.

Discussion
To validate our scale, we analysed several genetic variants localised in diverse genes associated and non-associated with different channelopathies (Table 1). For LQTS, we analysed genetic variations localized in two different genes, KCNQ1 and KCNH2. The first variation analysed was KCNQ1_p.R452W (c.1354C . T). This genetic variation has been previously associated with the disease 31 -CM055335-(3 points), reported as pathogenic but only detected in the index case (2 point). No in vivo or in vitro studies were performed or identified. In silico analysis revealed four databases with damaging prediction (2 points). The total score of 8 points in this case indicates that the variant remains classified as a GVUS; this is in accordance with a recent report classifying KCNQ1_p.R452W as a variant of unknown signifi-www.nature.com/scientificreports SCIENTIFIC REPORTS | 5 : 7953 | DOI: 10.1038/srep07953 cance 32 . Thus, this GVUS should be considered carefully in clinical practice. We believe that GVUS should be further analysed in relatives, since family segregation and genotype-phenotype correlation are required to clarify the pathogenic role of a variant.
The second variation was KCNQ1_p.Y315S (c.944A . C), previously associated with the disease 33 -CM970823-(3 points) and reported as pathogenic in different studies (3 points). Familial segregation was detected in less than 4 relatives (2 points), and expression of the variant in rabbit heart produced transgenic rabbits with an LQT phenotype 34 (2 points). In vitro analysis revealed a pathogenic role (2 points), and in silico analysis revealed 4 databases with damaging prediction (2 points). The total score of 14 points supports its classification as pathogenic in LQTS. In cases in which a pathogenic variation has been identified, current clinical/genetic guidelines should be followed both in the index case and in family members.
In KCNH2, also associated with LQTS, we analysed 2 genetic variations. The first one was KCNH2_p.E58A (c.173A . C), which was previously associated with the disease 35 -CM057129-(3 points), reported as pathogenic only in one study (2 points), and was detected only in the index case (1 point). No in vivo or in vitro studies were identified, and in silico analysis revealed 3 databases with pathogenic prediction (1 point). The total score of 7 points indicates that the variant is likely a GVUS. Hence, further studies such as family segregation and in vivo and/or in vitro analyses should be performed to clarify its role in LQTS. The last variation associated with LQTS that we analysed was KCNH2_p.G628S (c.1882G . A), which was reported to be associated with the disease 36 -CM950710-(3 points) and as pathogenic in more than one study (2 points). Familial segregation was identified in , 4 relatives (2 points), and expression in rabbit heart produced transgenic rabbits with an LQT phenotype 34 (2 points). In vitro analysis revealed a pathogenic role (2 points), and in silico analysis revealed 4 databases with damaging prediction (2 points). A total score of 13 points supports the classification of this variant as pathogenic in LQTS.
For CPVT, 2 variations were analysed. The first one was RyR2_p.R1013Q (c.3038G . A), previously associated with the disease 37 -CM097930-(3 points) reported as pathogenic but only detected in the index case (2 point). Familial segregation was identified only in the index case (1 point) and no in vivo or in vitro studies were performed. In silico analysis revealed two databases with damaging prediction (1 point). A score of 7 points suggests classification as a GVUS, in accordance with a recent report 32 . The second was RyR2_p.V2475F (c.7423G . T), previously reported associated with the disease 37 -CM097960-(3 points), and with a deleterious effect in more than one study (3 points). Familial segregation was identified only in the index case (1 point), and in vivo studies yielded a positive result (2 points) 38 . In vitro studies also showed a positive effect (2 points), and in silico analysis revealed 4 databases with damaging prediction (2 points). The total score of 13 points supports classification of this variant as pathogenic in CPVT.
Finally, three variations were analysed for BrS. The first one was SCN5A_p.V1340I (c.4018G . A), previously associated with the disease 17 -CM100703-(3 points), and with a deleterious effect reported in only one publication (2 points). It was identified only in the index case (2 points), and no in vivo studies were performed. In vitro studies showed a functional effect (2 points), and in silico analysis revealed three databases with damaging prediction (1 point). A score of 9 points indicates classification of this variant as probably pathogenic. Therefore, further studies should be performed to clarify its role in BrS.
The second variation analysed in BrS was SCN5A_p.R104Q (c.311G . A), previously reported associated with the disease 39 -CM014904-(3 points), and with a deleterious effect in some studies (3 points). Positive familial segregation was identified in less than 4 relatives (2 points) but no in vivo studies were performed. In vitro studies showed a functional effect (2 points), and in silico analysis revealed four databases with damaging prediction (2 points). The total score of 12 points indicates pathogenicity.
The last variation showing a disease-association gene was SCN5A_p.N70K (c.210T . G), previously reported to be associated with the disease 17 -CM100623-(3 points), and with a deleterious effect but only in one report (2 points). It was identified only in the index case (1 point), and no in vivo or in vitro studies were performed. In silico analysis revealed two databases with damaging prediction (1 point). The total score of 7 points supports classification of this variant as a GVUS.
To further assess our scale of pathogenicity, we include some variation in genes showing a reported association with the disease. Hence, we analyze PKP2_p.Q62K in a case of suspicious BrS. The genetic variation is located in a novel gene which encodes a desmosomal protein (1 point). The variation is already reported only in one study as pathogenic (associated with arrhythmogenic cardiomyopathy) (2 points), but only identified in the index case (1 point). Neither in vivo nor in vitro studies were performed (0 points respectively), and in silico analysis revealed two databases with damaging prediction (1 point). The total score of 5 points indicates GVUS, in concordance to published studies 40 .
We also analyzed SCN1Bb_p.P213T in a case of suspicious LQT. The genetic variation is located in a novel gene which encodes an ion channel regulatory protein (2 points). The variation is novel but the gene is linked to functional pathways of cardiac electrical disorders (2 points). No segregation was reported (1 point), and no in vivo studies are available (0 points). Both in vitro and in silico studies showed positive results (2 points). Therefore, total score of 9 points indicates probably pathogenic, as suggested in a recent publication 41 .

Conclusions
Determining the pathogenic role of genetic variants in channelopathies associated with SCD is necessary for improved clinical diagnosis and therapy of inherited arrhythmogenic diseases. To date, most identified genetic variants are considered rare variants of ambiguous clinical significance.
In the present report we propose a first approach in stratification of pathogenicity for genetic variations associated with channelopathies. Using several genetic tools available in the biomedical field, we propose a stratified scale of pathogenicity for variants identified in genes associated with SCD. We bring together multiple well accepted parameters that should be used together to reach a more definite clinical conclusion. We believe that this stratification may help determine the pathogenicity of genetic variants and will help clinicians understand the limitations of genetics and to better decide on therapeutic measures to prevent syncopal episodes and SCD.

Limitations
However, some limitations have to be acknowledged in the use of this approach. First, we propose the scale using several tools, but not all tools are available for each variant. It is not prudent to classify a variant using only one of these tools. Second, as genetic information is progressively becoming available, it is possible that new genetic/ clinical data can change the pathogenicity level in a near future. We recommend reassessing the results on a regular basis. Finally, expertise is crucial in the handling of the databases and interpretation of the information. Thus, translation into clinical practice should be performed by consensus of a group of experts in different clinical and basic disciplines.