Targeted next-generation sequencing detects novel gene–phenotype associations and expands the mutational spectrum in cardiomyopathies

Cardiomyopathies are a heterogeneous group of primary diseases of the myocardium, including hypertrophic cardiomyopathy (HCM), dilated cardiomyopathy (DCM), and arrhythmogenic right ventricular cardiomyopathy (ARVC), with higher morbidity and mortality. These diseases are genetically diverse and associated with rare mutations in a large number of genes, many of which overlap among the phenotypes. To better investigate the genetic overlap between these three phenotypes and to identify new genotype–phenotype correlations, we designed a custom gene panel consisting of 115 genes known to be associated with cardiomyopathic phenotypes and channelopathies. A cohort of 38 unrelated patients, 16 affected by DCM, 14 by HCM and 8 by ARVC, was recruited for the study on the basis of more severe phenotypes and family history of cardiomyopathy and/or sudden death. We detected a total of 142 rare variants in 40 genes, and all patients were found to be carriers of at least one rare variant. Twenty-eight of the 142 rare variants were also predicted as potentially pathogenic variants and found in 26 patients. In 23 out of 38 patients, we found at least one novel potential gene–phenotype association. In particular, we detected three variants in OBSCN gene in ARVC patients, four variants in ANK2 gene and two variants in DLG1, TRPM4, and AKAP9 genes in DCM patients, two variants in PSEN2 gene and four variants in AKAP9 gene in HCM patients. Overall, our results confirmed that cardiomyopathic patients could carry multiple rare gene variants; in addition, our investigation of the genetic overlap among cardiomyopathies revealed new gene–phenotype associations. Furthermore, as our study confirms, data obtained using targeted next-generation sequencing could provide a remarkable contribution to the molecular diagnosis of cardiomyopathies, early identification of patients at risk for arrhythmia development, and better clinical management of cardiomyopathic patients.

Introduction Inherited cardiomyopathies are a group of heart muscle diseases characterized by heterogeneous phenotypes, including dilated cardiomyopathy (DCM), hypertrophic cardiomyopathy (HCM), and arrhythmogenic right ventricular cardiomyopathy (ARVC), with higher morbidity and mortality [1,2]. DCM, defined by the presence of left ventricular (LV) or biventricular dilatation and systolic dysfunction in absence of hypertension, valvular disease, or coronary artery disease [1,2], is one of the leading causes of heart failure (HF) and sudden death (SD) and represents the most common condition requiring a heart transplant [1,2]. HCM, characterized by increased LV wall thickness that is not solely explained by abnormal loading conditions [3,4], is the most common cause of SD in young, and is considered an important cause of HF and embolic stroke secondary to atrial fibrillation [5,6]. A diagnosis of ARVC is made mainly based on functional and structural abnormalities of the right ventricle (RV), fibro-fatty replacement of the myocardium, depolarization and repolarization alterations, arrhythmias with left bundle branch block (LBBB) morphology, and family history, with no single gold standard exam of diagnosis established to date [7]. ARVC represents one of the major causes high degree of genetic heterogeneity, with more than 100 cardiomyopathy-related genes identified to date, can be addressed more efficiently by high-throughput sequencing, referred to as next-generation sequencing (NGS) [36], which includes targeted resequencing, whole exome sequencing, and whole genome sequencing. Targeted enrichment appears to be the method of choice in a clinical diagnostic setting, not only because it allows focusing on the genes relevant to a particular disorder, thereby preventing unsolicited findings, but also because it provides a superior quality of exon representation and coverage.
In this study, we developed a custom "pan-cardiomyopathy panel" containing 115 genes known to be associated with DCM, HCM, and ARVC as well as to channelopathies. Based on this panel, we performed molecular screening in 38 unrelated patients, 16 affected by DCM, 14 by HCM, and 8 by ARVC, to better investigate the genetic overlap between these three phenotypes and to identify new gene-phenotype associations using the Illumina MiSeq platform. Moreover, we investigated the possible role of genes linked to channelopathies in modifying the clinical cardiomyopathic phenotype.
The results of this study put a spotlight on novel cardiomyopathic gene-phenotype associations, extending the mutational spectrum underlying cardiomyopathies. Furthermore, our data suggest that some genes involved in channelopathies may act as genetic modifiers, modulating the clinical features and severity of cardiomyopathic phenotypes, as well as contributing to the variable cardiomyopathic phenotypic expression. Our findings support the "pan-cardiomyopathy panel" approach, since many patients carried multiple rare variants in several genes.

Ethics statement
All patients provided their written informed consent to participate in this study. We obtained written consent from parents of the minors included in the study. The project conformed to the principles of the Declaration of Helsinki (World Medical Association) and was approved by the Ethics Committee of the University Hospital Consortium, Policlinico of Bari, Italy.

Study subjects, disease criteria, and clinical evaluation
The patients who were referred to the Cardiomyopathy Unit, Cardiology Unit, Department of Emergency and Organ Transplantation, University of Bari Aldo Moro, Bari (Italy), between February 2008 and June 2014 were enrolled in this study. A total of 38 Italian unrelated patients (16 DCM, 14 HCM, and 8 ARVC) were recruited on the basis of more severe phenotypes and family history of cardiomyopathy and/or SD (S1 Table).
All patients underwent clinical work-up, including medical history, physical examination, 12-lead electrocardiogram (ECG), transthoracic echocardiography, and 24-hour ECG monitoring. Where appropriate, exercise testing and coronary angiography were performed. Subjects without contraindications, such as pacemaker, defibrillator, or severe claustrophobia, underwent cardiac magnetic resonance imaging. DCM, HCM, and ARVC as well as channelopathies, were selected for targeted gene enrichment. For genes with multiple transcripts, all exons included in transcripts expressed in cardiac muscle were considered in the gene panel design. The coordinates of genomic regions were based on NCBI build 37 (UCSC hg19). The complete list of genes and transcripts is reported in S2 Table. Targeted gene enrichment and high-throughput sequencing Total DNA was extracted from peripheral blood samples using the Wizard Genomic DNA Purification Kit (Promega, Mannheim, Germany) according to the manufacturer's instructions, quantified, and qualitatively checked using NanoDrop 2000c (Thermo Fisher Scientific, Waltham, MA, USA).
Custom targeted gene enrichment and DNA library preparation were performed using the Nextera Capture Custom Enrichment kit (Illumina) according to the manufacturer's instructions. The targeted regions were sequenced using the Illumina MiSeq platform, generating approximately two millions of 150-bp paired-end reads for each sample (Q30 !90%).
Sequencing and genotyping data were submitted to "The European Genome-phenome Archive" EGA (http://www.nature.com/ng/journal/v47/n7/full/ng.3312.html), with the accession number EGAS00001002506. S3 Table reports the accession numbers of the sequencing data for each patient.

Variant calling, filtering, and classification
Variant calling was performed using the web tool Online Deep Exome Sequencing Software Analysis (ODESSA) [46]. The pipeline, after sequencing data submission, executed the following steps: the quality checks and filter of the reads; the alignment on the reference genome hg19; the variant preprocessing; the coverage statistics and metrics; the variant calling; the variant annotation. Genetic variants predicted to alter the protein, such as non-synonymous variants, nonsense variants, canonical splicing site variants (affecting the donor or acceptor splice sites), in-frame and frameshift insertion/deletions, supported by at least 10 reads, and with a minor allele frequency (MAF) 0,4%, 0,2%, 0,05% for DCM, HCM and ARVC respectively in variant databases, chosen considering the prevalence of the phenotypes in the general population [47], were selected. To assess the potential functional impacts of variants, four bioinformatics algorithms were used: PolyPhen-2 (PP2), Sorting Tolerant From Intolerant (SIFT), Protein variation effect analyzer (Provean), and Mutation Taster. PhyloP and Phast-Cons scores were also used to analyze the degree of sequence conservation of the mutated loci, and Grantham's score was considered to categorize codon replacements. To calculate these three parameters for each rare missense variant, we employed Mutation Taster. Missense variants were considered "potentially pathogenic" if classified simultaneously as "damaging" by SIFT, "deleterious" by Provean, "possibly" or "probably damaging" by Polyphen-2, and "Disease Causing" by Mutation Taster. TTN missense variants were not considered potentially pathogenic independently from the predictors' results because, as recently reported, there is no statistical difference in frequency between cases and controls and their potentially pathogenic role can be established only by a segregation analysis in affected families [48,49]. Stopgain and stoploss variants, splicing variants, frameshift, and in-frame insertions and deletions were considered potentially pathogenic if classified as "Disease Causing" by Mutation Taster; we defined these variant types as Radical variants. TTN truncating variants were considered potentially pathogenic only if they affected TTN isoforms in the A-band region [49][50][51].
Variants not classified as potentially pathogenic by predictors, but for which a pathogenic role was supported by published data [52] and/or by evidences in ClinVar and HGMD databases were considered as potentially pathogenic. Variants classified as Variants of Uncertain Significance ("VUS"), "Likely Benign" or "Benign" in ClinVar database were not considered as potentially pathogenic variants (S1 Fig).
It should be considered that in our study the designation of potentially pathogenic variants derived from a genetic test aimed to assess new gene-phenotype associations. The clinical relevance of these potentially pathogenic variants should instead be assessed on the basis of the current guidelines [53], but it was not the aim of the study.

Clinical features
The clinical characteristics of all evaluated patients are shown in S1 Table. A total of 16 DCM patients (14 men, 2 women; mean age 38 ± 15 years; range: 15 to 67 years) were included in this study, 11 of which had a family history of DCM and/or SD.
Fourteen patients (7 men and 7 women; mean age 35 ± 14 years; range: 16 to 58 years) presented with either obstructive (n = 7) or non-obstructive (n = 7) form of HCM, 10 of which had a family history of the condition and/or SD.
Eight patients (5 men and 3 women; mean age 40 ± 21 years; range: 12 to 72 years) presented with ARVC, 7 of which had a family history of arrhythmogenic cardiomyopathy and/or SD.
Overall, 31 patients out of 38 received an implantable cardioverter defibrillator (ICD), 24 of which in primary prevention and 7 in secondary prevention; one DCM patient aged 45 years underwent a heart transplant; and one DCM patient aged 71 years carrying an ICD died due to refractory HF.

Pan-cardiomyopathy gene panel sequencing and variant calling
We designed a custom pan-cardiomyopathy panel containing 115 genes known to be associated with DCM, HCM, and ARVC, as well as to channelopathies (S2 Table). The custom gene panel, encompassing all exons of each gene and splicing sites, was based on the Illumina strategy of targeted enrichment of these regions. Using this custom panel, we performed genetic screening of the 38 unrelated patients described above, obtaining an average of two millions of 150-bp paired-end reads per sample.
A total of 3286 missense variants, 18 nonsense variants, 23 splicing site variants, 29 inframe and 6 frameshift insertion/deletions were identified using the ODESSA pipeline, providing an average of 85 variants per DCM patient, 89 per HCM patient, and 92 per ARCV patient (S4 Table). Overall, the mean read depth of coverage for these selected variants was 222X (range, 12-565X). Variants that fulfilled the inclusion criteria described in the Materials and Methods were annotated as rare variants and classified using standard nomenclature (S1 Fig).

Variant assessment
In total, we detected 142 rare variants in 40 genes, of which 137 variants had a heterozygous status and 5 variants on the X chromosome had a hemizygous status. Thirty-three (23,2%) of these 142 rare variants were absent from both dbSNP and gnomAD (S5 Table). One hundred and twenty-four rare variants were missense variants (87%), while 5 were frameshift insertion/ deletion variants, 3 in-frame insertion/deletion variants, 8 nonsense variants, and 2 involved a splicing site (S6 Table). The TTN gene had the largest number of rare variants (42 variants), followed by OBSCN (16 variants) and RYR2 (6 variants).
Considering rare variants for each cardiomyopathy type, in ARVC patients, the highest number of rare variants was found in TTN and RYR2 genes (4 variants) followed by OBSCN and PKP2 (3 variants) (S7 Table). For DCM patients, the highest number of rare variants was found in TTN gene (26 variants), followed by OBSCN (8 variants) and ANK2 (4 variants) genes (S8 Table). For HCM patients, the highest number of rare variants was found in TTN gene (12 variants), followed by OBSCN (5 variants) and MYBPC3 (4 variants) genes (S9 Table).
All patients carried at least one rare variant, while most of them (94.7%) carried more than one rare variant; on average, 3.7 rare variants per patient (range, 1-9) were detected, with DCM patients carrying the highest average number of rare variants (4.5 per patient), followed by HCM (3.6 per patient), and ARVC (2.4 per patient) (S10 Table).
Of the 142 rare variants, 28 were classified as potentially pathogenic, 9 of which already present in ClinVar and/or HGMD databases and/or supported by published data (Table 1 and S11 Table). Fifteen potentially pathogenic variants were missense mutations, 4 were frameshift insertion/deletion mutations, 2 were in-frame insertion/deletion mutations, 5 were nonsense mutations, and 2 affecting a splicing site (Table 1). We excluded from the list of the potentially pathogenic variants, all the TTN missense variants indicated as potentially pathogenic by the analysis with the bioinformatics tools because, as recently reported, there is no statistical difference in frequency between cases and controls and their potential pathogenic role can be established only by a segregation analysis in affected families [48,49]. Regarding the TTN truncating variants, we followed the recent evidences that suggest to consider as potentially pathogenic only those variants affecting all TTN isoforms in the A-band region [50,51]. We identified three TTN truncating variants in DCM patients. We considered as potentially pathogenic only the TTN truncating variant V16477fs, referred to the N2B isoform (NM_ 003319) as it maps in the TTN A-band region and was also not present in gnomAD database. The R2490fs variant, referred to the Novex-3 transcript (NM_1333379) and the L24944X variant, referred to the N2B transcript (NM_003319) were not considered potentially pathogenic as they mapped in the I and M-band regions of the TTN isoforms, respectively.

New gene-phenotype associations
Most of the rare variants identified in our study showed a well-established association with a specific cardiomyopathic phenotype. In addition, in 23 of 38 patients, we found at least one novel potential gene-phenotype association (Table 2 and S10 Table). As shown in Fig 1, considering the most represented genes, we found six new associations for AKAP9 gene, four for ANK2 gene, three for DLG1, OBSCN and TRPM4 genes and two for PSEN2 and SYNE1 genes. In particular, we focused on potentially pathogenic rare variants to examine the novel genephenotype associations (Table 2 and S11 Table). For the ARVC phenotype, we found two Expanded mutational spectrum in cardiomyopathies through targeted next-generation sequencing potentially pathogenic variants involving DMD and OBSCN genes, not previously associated with this phenotype. Indeed, we found one novel potentially pathogenic missense variant in DMD gene in patient identified as 1751ARVC (DMD_R1320H) and one potentially pathogenic missense variant in OBSCN gene.
Regarding the HCM phenotype, we identified two potentially pathogenic variants involving DLG1 and AKAP9 genes, not previously associated with this phenotype. In DLG1 gene, a potentially pathogenic missense variant was found in patient identified as 1776HCM (DLG1_ G639R); in AKAP9 gene, a novel potentially pathogenic nonsense variant was found in patient identified as 1833HCM (AKAP9_Y3870X).
With regard to DCM, the association between potentially disease-causing variants in OBSCN gene and this phenotype was recently described [54]. Our findings were in agreement with this new association, as we found two potentially pathogenic variants in OBSCN gene, one in patient identified as 682DCM (OBSCN_R6669H) and the other in patient identified as 1718DCM (OBSCN_A5660V). Moreover, for DCM phenotype, we identified one potentially pathogenic variant involving the AKAP9 gene, not previously associated with DCM, in patient identified as 76DCM (AKAP9_R3435X).

Discussion
The technology of NGS is now emerging as a powerful approach to comprehensively explore genetic mutations in a wide range of human pathologies [36,55]. In this study, we used targeted resequencing to identify novel genetic variations associated with three inherited cardiomyopathies, namely HCM, DCM, and ARVC, which are common causes of mortality before the fifth decade of life. These inherited cardiomyopathies are genetically heterogeneous and are associated with rare mutations in a large number of genes, many of which overlap among the phenotypes. Moreover, considerable phenotypic and genetic overlaps between cardiomyopathies and arrhythmic syndromes have been documented [22][23][24][25][26][27][28][29]. Considering these overlaps, our study aimed to look for new potential gene-phenotype associations with a view to expand the mutational spectrum underlying cardiomyopathies, with possible implications for better diagnosis and clinical management. We designed a custom gene panel including 115 genes, known to be associated with the three cardiomyopathic phenotypes and/or arrhythmic syndromes. Using this panel, we performed molecular screening of a cohort of 38 unrelated patients affected by cardiomyopathies (16 DCM, 14 HCM, and 8 ARVC). Overall, our analysis identified a total of 142 rare variants in 40 genes and confirmed 11 rare variants (ACTC1_S236F, LAMP2_V310I, LMNA_Leu140-Ala146dup, LMNA_Q517X, MYBPC3_A364T, MYBPC3_c.821 +1G>A, MYBPC3_E728X, MYH7_A1763T, MYH7_G701D, MYL2_E134A, and TNNT2_ R102Q) which we previously identified in some of our patients by Sanger sequencing [52,56], thereby demonstrating that our approach, that combined NGS and different in silico prediction Expanded mutational spectrum in cardiomyopathies through targeted next-generation sequencing Our results confirm previous data indicating that cardiomyopathic patients could carry multiple specific genetic variants [33,34,57]. In fact, we found on average more than three rare variants per proband, with DCM patients carrying the highest average number of rare variants, followed by HCM and ARVC patients. Furthermore, since our subjects were selected on the basis of phenotype severity and family history of cardiomyopathy and/or SD, our results support the evidence that multiple mutations are likely to elicit a more severe phenotype. Expanded mutational spectrum in cardiomyopathies through targeted next-generation sequencing Indeed, previous studies suggested that patients carrying multiple mutations are more likely to develop the disease at an early age and with more severe clinical features, as poor prognoses are associated with compound or double heterozygotes [33,34]. Titin is the largest described human protein, constituting one of the most abundant types of filaments in both cardiac and skeletal muscle and plays a central role in sarcomere organization. TTN has been recognized as the main disease-causing gene in DCM and is also involved in the pathogenesis of HCM and ARVC [17,49,58]. Our study confirmed the high prevalence of TTN rare variants in cardiomyopathies, most of which are missense. TTN missense variants are very common and often frequently benign and it was reported that TTN missense variants bioinformatically classified as "severe" did not associate with differences in clinical phenotypes in DCM patients, so arguing that many TTN missense variants are not disease-causing variants [48]. For these reasons, many TTN missense variants identified in our study were not classified as potentially pathogenic, despite the results of predictors' analysis. Regarding TTN truncating variants, recent studies supported that not all these variants are equals in terms of clinical relevance. In particular, the TTN truncating variants with the highest probability of pathogenicity are those affecting TTN isoforms in the A-band region [50,51]. In our study, we identified three TTN truncating variants in DCM patients: V16477fs (referred to NM_003319 transcript), L24944X (referred to NM_003319 transcript) and R2490fs (referred to NM_133379 transcript). According to recent evidences, we considered as potentially pathogenic, with a possible clinical relevance, only the V16477fs variant, as it maps in the A-band region and affects all TTN transcripts. However, it cannot be ruled out that multiple variants (missense and truncating) on the TTN protein may have a cumulative effect on a cardiomyopathic phenotype.
In this study, we extended the number of overlapping genes among cardiomyopathies and also between cardiomyopathies and arrhythmic syndromes, identifying novel potential genephenotype associations (Fig 1 and S2 Table). As reported in Fig 2, which shows the distribution of specific and overlapping genes carrying rare variants identified in this study, among the three cardiomyopathic phenotypes, we detected five genes (DMD, OBSCN, SYNE1, TNNC1, VCL) carrying rare variants in the three phenotypes, six genes (AKAP9, CACNA1C, CHRM2, DLG1, MYO6, PSEN2) carrying rare variants in HCM and DCM, and one gene (TRPM4) carrying rare variants in ARVC and DCM. As several patients carried variants in genes not previously related to their phenotype, the use of our "pan-cardiomyopathy panel" may prove beneficial to identify the novel potential gene-phenotype associations. In fact, a phenotypespecific targeted gene testing [30] might miss to identify several variants in cardiomyopathic patients. Conversely, a comprehensive gene testing could enhance the diagnostic and prognostic implications of genetic screening by not only allowing an early diagnosis of inherited cardiomyopathic forms but also by contributing to a more accurate prognostic stratification and patient management [31].
OBSCN variants were firstly described in HCM patients [59] and, more recently, OBSCN variants associated with DCM and Left Ventricular Noncompaction patients were reported [54,60]. Our results are in agreement with the association between OBSCN gene variants and DCM phenotype, since we found eight different rare variants in 8 DCM patients. When our analysis was restricted to potentially pathogenic variants, the association between OBSCN variants and DCM remained, because two potentially pathogenic variants, OBSCN_R6669H and OBSCN_A5660V, were detected in patients identified as 682DCM and 1718DCM, respectively (Table 2). Furthermore, for the first time to our knowledge a possible association between OBSCN gene and the ARVC phenotype is evidenced. In fact, we found three different OBSCN rare variants in two out of eight ARVC probands. One of these variants, OBSCN_A5791P was classified as potentially pathogenic (Table 1 and S11 Table) and was detected in patient identified as 1662ARVC (Table 2). A possible pathogenic role for OBSCN variants detected in some of our DCM and ARVC patients is supported by the fact that these variants involve OBSCN domains having functional and/or structural importance and affect highly conserved amino acid regions across species. Obscurin is a member of a family of giant proteins that is expressed in striated muscle and interacts with titin, myomesin and small ankyrin-1 and has been proposed to be a structural protein linking the M-line of the sarcomere to the sarcoplasmatic reticulum [61,62]. Different obscurin isoforms are known: obscurin A (~720 kDa); obscurin B (~870 kDa), which is similar to obscurin A but lacking the nonmodular COOH-terminal region and including two Ser/Thr kinase domains, called SK2 and SK1; and two smaller obscurin isoforms, one containing a full-length SK1 and a partial SK2 and the other containing only the SK1 domain. The expression of these obscurin isoforms differs; while giant obscurins A and B are expressed in higher amounts in skeletal muscle, the tandem and single kinase isoforms are largely present in cardiac muscle [62]. Variant OBSCN_R6669H involves the SK1 domain, and variant OBSCN_A5791P falls in a functionally important domain of obscurin, the tandem Rho-guanidine nucleotide exchange factor (RhoGEF) present in Rho/Rac/Cdc42-like GTPases, and thus these variants may affect obscurin cell signaling pathways. Variant OBSCN_A5660V is present in the src homology-3 (SH3)-conserved domain, a protein interaction domain. Therefore, OBSCN appears to be a good candidate gene for both DCM and ARVC phenotypes. Recent data supported OBSCN mutations as a significant causal factor for DCM pathogenesis, reporting that disease-related OBSCN mutations cause haploinsufficiency that accounts for the development of the DCM phenotype [54]. Moreover, it was suggested that OBSCN mutations could act alone or in concert with other mutations, contributing to cause DCM [54,63]. Our data are in agreement with this last hypothesis, since patients carrying a potentially pathogenic Expanded mutational spectrum in cardiomyopathies through targeted next-generation sequencing variant in the OBSCN gene also carried other rare variants in different genes, such as PKP2, or TTN, suggesting a possible cumulative effect on cardiomyopathic phenotype.
We identified the DMD_R1320H variant in the patient identified as 1751ARVC, who was diagnosed with ARVC and associated LV systolic dysfunction at the age of 42 years. The novel variant is localized in the spectrin domain number 9 of dystrophin and is predicted to be pathogenic in silico because it is believed to affect two DMD transcripts. Mutations in dystrophin affecting the plasma membrane and causing myofiber loss could lead to the pathologic substrate favoring myocardial atrophy and fibrofatty replacement, which are the main pathogenetic mechanisms described in ARVC [78]. However, more evidences are necessary to consider DMD as a new gene associated with the ARVC phenotype.
Although hereditary cardiomyopathies and arrhythmic syndromes are considered distinct groups of genetic disorders, a considerable phenotypic overlap between these two apparently distinct cardiac disorders has been documented. In fact, the clinical cardiomyopathic picture often includes a wide range of features related to arrhythmias and conduction abnormalities; conversely, ion channel defects linked to arrhythmic syndromes may be associated with morphological changes and structural abnormalities that characterize some types of cardiomyopathy [5,[24][25][26][27][28][29]79]. In a recent study that screened HCM probands for both sarcomere and non-sarcomere protein gene mutations, ANK2, PLN, and SCN5A gene variants were suggested to have modifying effects on HCM phenotypic expression [26]. Moreover, novel associations between three genes (KCNQ1, KCNH2, and KCNE2) associated with Long QT syndrome and HCM phenotype were proposed [26]. As the main goal of our study was to look for new genephenotype associations, we included in our custom panel some genes associated with inherited arrhythmic syndromes in order to support the evidence that genes implicated in channelopathies could modify the clinical characteristics and severity of cardiomyopathic phenotypes. We found novel potential gene-phenotype associations between genes typically related to arrhythmic syndromes (AKAP9, ANK2, CACNA1C, DLG1, HCN4, KCNJ8, NUP155 and TRPM4 genes) and the cardiomyopathic phenotypes discussed in our study (S2, S9 and S11 Tables). In particular, focusing on the potentially pathogenic variants, we detected new associations between AKAP9 variants and both DCM and HCM and between DLG1 variant and HCM (Table 2 and S11 Table).
A novel association between AKAP9 gene and DCM was observed in patient identified as 76DCM. This patient demonstrated an early-onset (15 years) DCM phenotype and presented with a history of palpitations, syncope on effort, and evidence of frequent multifocal premature ventricular complexes, non-sustained VT episodes, and atrioventricular conduction abnormalities recorded on Holter monitoring. He received an ICD. During follow-up, he suffered from recurrent episodes of sustained VT and VF discontinued by ICD shocks. This patient carried two potentially pathogenic variants: DSP_V508fs variants and AKAP9_R3435X (Table 2 and  S11 Table). DSP_V508fs is a novel variant, predicted to produce a truncated protein and to reduce DSP expression. AKAP9_R3435X is predicted to produce a truncated protein from three different AKAP9 transcripts, thus suggesting either a dominant negative mechanism or a nonsense-mediated mRNA decay mechanism. We suppose that the DSP_V508fs variant is the primary disease-causing mutant since the DSP gene has been demonstrated to be associated with DCM (OMIM: 615821) and biventricular arrhythmogenic cardiomyopathy (OMIM: 607450) and that AKAP9 may be a modifying gene causing severe ventricular tachyarrhythmias, such as those occurring in the patient identified as 76DCM. This hypothesis is supported by the evidence that AKAP9 has been considered as a genetic modifier of congenital LQTS, increasing cardiac risk and disease severity in AKAP9 mutation carriers [80].
Another potentially pathogenic variant (AKAP9_Y3870X) is present in the patient identified as 1833HCM, who also carries the MYL2_E134A variant (S10 Table). The MYL2 variant may have a pathogenic role in 1833HCM patient, as previously described for the same phenotype [81], although it is classified as VUS in ClinVar database. The AKAP9_Y3870X variant may act as a disease severity modifier, even though the stop codon introduced by this mutation is located in the C-terminal part of the protein; so, its effect should be analyzed in functional studies and/or segregation in families.
DLG1 encodes the synapse-associated protein 97 protein (SAP97), a protein belonging to the membrane-associated guanylate kinase (MAGUK) family of proteins that are important for localization and organization of ion channels. SAP97 interacts with and modulates cardiac Kv4.3 channels that account for a large part of the transient outward potassium current (Ito) [82], but also binds and regulates the expression of NaV1.5, the main cardiac inward sodium channel [83]. Both channels are implicated in Brugada syndrome (BRGDA1, OMIM: 601144; BRGDA9, OMIM: 616399), and consequently a DLG1 variant may confer susceptibility to arrhythmias as demonstrated by a putative DLG1 pathogenic mutation in a patient affected by this syndrome [84]. We found the potentially pathogenic variant DLG1_G639R in patient identified as 1776 HCM, also carrying the potentially pathogenic variant MYBPC3_A364T (Table 2, S10 and S11 Tables). The MYBPC3_A364T variant may be considered pathogenic for this patient, as MYBPC3 is one of the main genes associated with HCM. The DLG1_G639R variant involves the DLG1 SH3 domain, which mediates the assembly of specific protein complexes via binding to proline-rich peptides. It is conceivable that this mutation could affect DLG1-Kv4.3 or DLG1-NaV1.5 interactions, modulating their functions and favoring susceptibility for arrhythmias.
Our study demonstrates that a targeted resequencing approach can provide a relevant support for a genetic diagnosis in cardiomyopathic patients. However, in many patients, in particular, those affected by DCM, it was not possible to obtain a clear genetic diagnosis on the basis of our gene panel, thus suggesting a greater genetic heterogeneity in this cardiomyopathy respect to the others and/or a considerable role of modifier genes in determining this phenotype. Gene discovery by whole exome or genome sequencing would be required for those probands that are still without a molecular diagnosis. However, the application of this approach would make the interpretation of genetic screening results difficult for clinical purposes due to the large amount of sequencing data that need to be analysed.
We are aware that our study would need to be supported by evidences in favor of a pathogenic role of the variants reported here (e.g., lack of segregation analysis due to sporadic cardiomyopathy, and/or unavailability and/or small number of relatives available for genetic investigations and lack of functional studies) [85], not allowing us to assess the clinical relevance of the identified variants. Segregation analysis together with functional studies may concur to establish a more definite diagnosis, allowing the findings to be interpreted in the context of a clinical setting. However, families with multiple affected individuals are a rare occurrence and often a novel variant is present in a single individual with only few or no other clinically affected relatives. Consequently, many novel NGS variants are classified as VUS and cannot represent a conclusive test result at present, as recently highlighted [85].
On the other hand, we believe that the rigorous approaches used in this study to identify the cardiomyopathies associated variants (e.g. recruitment of patients with more severe phenotypes and family history of cardiomyopathy and/or sudden death, selection of variants using a MAF lower than the frequency of each cardiomyopathy, application of different in silico prediction tools to select the potentially pathogenic variants) reliably support our hypothesis that many of them may contribute to cardiomyopathy development and/or its progression. In particular, in agreement with results obtained by Marston et al. [54], mutations in the OBSCN gene could be considered a significant causal factor, alone or in concert with other mutations, for DCM and ARVC pathogenesis. Furthermore, variants in AKAP9 and DLG1 genes could act as genetic modifiers of arrhythmic risk and phenotype severity in cardiomyopathies. In addition, variants of unknown pathogenicity, sometimes complicating the clinical picture, may contribute to the disease development and/or may have modifier effects. It is possible to hypothesize that the genetic component underlying cardiomyopathic phenotypes and/ or modifying disease phenotype severity could be explained by the cumulative functional effects of coding and non-coding DNA variants.
To conclude, this work highlights the challenges faced in the diagnosis of cardiomyopathies. NGS is a highly accurate and reproducible approach for routine molecular screening of patients with cardiomyopathies, but the identification of genetic defects in the proband is not sufficient in isolation for clinical decision-making or genetic counselling.
Finally, our data support the remarkable contribution that a targeted NGS approach could offer to genetic overlap knowledge between different cardiomyopathies, expanding previous results and ameliorating molecular diagnosis.
In the brackets, we report the number of the new potential gene-phenotype associations (number on the left) and the number of the overlapping genes (number on the right) detected in this study (see also underlined associations in S2 Table). Missense variants were analyzed using Mutation taster, SIFT, Provean and Polyphen-2 and retained if predicted to be "disease causing" by Mutation Taster, "damaging" by SIFT, "deleterious" by Provean and "probably" or "possibly damaging" by Polyphen-2 and considered Potentially Pathogenic missense variants if not classified as VUS, Likely Benign or Benign in ClinVar database. Stop-gain, stop-loss and splicing site variants were analyzed using Mutation Taster and considered Potentially Pathogenic radical variants if predicted to be "disease causing" and if not classified as VUS, Likely Benign or Benign in ClinVar database. B. Insertion/deletion variant analysis. Frameshift and in-frame insertion/deletion variants supported by at least 10 reads with a minor allele frequency 0.004, 0.002 and 0.0005 for DCM, HCM and ARVC patients respectively in variant databases, were considered and classified as rare radical variants. Radical variants were analyzed using Mutation Taster and considered Potentially Pathogenic if predicted to be "disease causing" and if not classified as VUS, Likely Benign or Benign in ClinVar database. Missense and radical variants were also considered as Potentially Pathogenic variants, independently from the results obtained with the above predictors, if classified as "Pathogenic" or "Likely Pathogenic" in published data and/or in ClinVar or HGMD database, in association with cardiomyopathies. TTN missense variants were not considered potentially pathogenic independently from predictors' responses. TTN truncating variants were considered potentially pathogenic if they affected all TTN isoforms in the A-band region. (TIF) S1