Introduction

Sudden cardiac death (SCD) is a major cause of unexpected, non-traumatic deaths in young and apparently healthy individuals. The major aetiologies of SCD include cardiomyopathies, channelopathies and ischaemic heart diseases. In about two thirds of SCD cases, structural abnormalities are evident at autopsy [1–3], whereas no structural pathogenic variation of the heart is found in the remaining cases. These cases are called sudden unexplained deaths (SUD). Many of the SUD cases are suspected to be caused by inherited cardiac channelopathies [4] among which the most common ones are long QT syndrome (LQTS), Brugada syndrome (BS), catecholaminergic polymorphic ventricular tachycardia (CPVT) and short QT syndrome (SQTS). More than 100 genes have been shown to be associated with inherited SCD [5–8]. With modern DNA technology like next-generation sequencing (NGS), it is possible to investigate large numbers of genes and thus support the clinical investigation and increase the possibility to establish a likely cause of death. This may be of importance for legal reasons but may also be valuable information when considering screening of family members for the same disease.

In the forensic setting, SUD is a major challenge. Investigation of pathogenic variants in genes associated with cardiac diseases is possible, but the diagnostic yield is not well known. The aim of this study was to estimate the frequency of pathogenic variants in 34 of the genes most frequently associated with channelopathic diseases, comparing SUD cases and patients with clinically diagnosed inherited cardiac channelopathies.

Material and methods

Study population

A total of 15 SUD cases from the Section of Forensic Pathology, Department of Forensic Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, Denmark, were investigated. The individuals were <50 years of age and randomly selected if (i) deceased within a time period of 1 year (2009–2010), (ii) they had died suddenly, unexpectedly and with no diagnosis after thorough macro- and microscopic investigations by the forensic pathologists and (iii) the toxicology screenings were negative. The SUD cases included 11 males with a median age of 38 years (range 3–49) and four females with a median age of 19 years (range 1–42). The majority, 12 (80 %), was last seen alive and healthy less than 24 h prior to the death and none had a known family history of SCD.

A total of 29 clinically affected patients from the Department of Cardiology, National University Hospital, Rigshospitalet, Copenhagen, Denmark, were included. Twelve (41 %) patients had LQTS, nine (31 %) Brugada syndrome, one (3 %) SQTS and seven (24 %) non-diagnostic phenotypes with symptoms and findings suggestive of channelopathic diseases. The median age at time of diagnosis was 36 years (8–64) with 22 (79 %) under the age of 50 years. A total of 17 (59 %) were males. Family history of SCD was reported in 10 (34 %) patients (Table 1).

Table 1 Characteristics of cardiac patients
Full size table

Genetic investigation of single nucleotide variants and large insertion-deletions

DNA was extracted from anti-coagulated whole blood using the QIAamp DNA Mini Kit (Qiagen, Germany) and quantified with the Quantifiler Human DNA Quantification Kit (Thermo Fisher, USA) following the manufacturer’s protocol. The 15 SUD samples were investigated with the AmpFlSTR Identifiler PCR Amplification Kit (Thermo Fisher, USA) to measure the degree of DNA degradation. Full STR profiles were found, thus no SUD sample was degraded. Two patient samples were whole genome amplified due to low DNA concentrations, using the REPLI-g Midi kit (Qiagen, Germany) following the manufacturer’s protocol.

A total of 34 genes (ABCC9, AKAP9, ANK2, CACNA1C, CACNB2, CASQ2, CAV3, DPP6, GJA5, GPD1L, HCN4, KCNA5, KCNE1, KCNE1L, KCNE2,KCNE3, KCNE4, KCNH2, KCNJ2, KCNJ5, KCNJ8, KCNQ1, MYH6, MYH7, NPPA, PRKAG2, RYR2, SCN1B, SCN2B, SCN3B, SCN4B, SCN5A, SNTA1 and TRPM4) were investigated using NGS. All genes have been reported to be associated with inherited cardiac diseases often presented with arrhythmia. The genes were selected from the literature and Online Mendelian Inheritance in Man (OMIM) [9]. The gene MYH7, which is primarily associated with structural cardiac diseases, was included as a result of the similarity to MYH6 and the capture design. The total length of the target regions was 387,395 bp and involved 639 target regions. Target regions included exons, 250 bp of adjacent introns and 5′- and 3′-UTR regions. Targeted sequencing capture probes were custom designed by Roche NimbleGen (USA). Library preparation was performed by using the TruSeq LT DNA Sample Preparation Guide (Illumina, USA). This protocol involved the steps “Fragment Genomic DNA” to “Ligate Indexed Paired-End Adapters”. Following this, the preparation of the libraries were continued using the NimbleGen SeqCap EZ Library SR User’s Guide v4.0 (Roche, Germany) according to the manufacturer’s recommendations. After library build, the samples were analysed on Agilent’s 2100 Bioanalyzer and the DNA concentrations were measured using the Qubit 2.0 Fluorometer (Thermo Fisher, Germany). Libraries were sequenced on an Illumina MiSeq using the MiSeq Reagent kit v2 (500 cycles) with 250-bp paired-end sequencing.

Genetic investigations of large insertion-deletions (indels) in KCNQ1, KCNH2, KCNE1, KCNE2 and SCN5A were performed with multiplex ligation-dependent probe amplification (MLPA) [10] using pre-designed kits (SALSA MLPA P114 Long-QT probe mix, SALSA MLPA P108 SCN5A probe mix (MRC-Holland, Holland). The analyses were performed according to the protocol provided by the supplier. In brief, the procedure involved denaturation of the DNA, hybridisation, ligase reaction, PCR and fragment separation by means of electrophoresis. Each sample was investigated in duplicate. If a rearrangement was observed, a third investigation was performed. Only concordant results were reported.

Data analysis and annotation

MiSeq Reporter (Illumina, USA) was used for the analysis of fastq-files generated by the MiSeq with default settings for PCR amplicon analysis. The human genome sequence UCSC/hg19 [11] was used as reference sequence. The following steps were performed using the MiSeq Reporter: Reads were aligned to the reference genome using BWA [12]. Variant calling was carried out using The Genome Analysis Toolkit (GATK) [13]. Aligned reads were provided in BAM-files (Binary compressed SAM format) and lists of variants in variant caller format (VCF). A total of 98.4 % of the target regions were covered more than 10 times. The median coverage for all regions and samples was 1216 (range 619–3191). In total, all exons in the investigated 34 genes were sequenced except for exon 1 and the 5′-UTR region in SCN1B. Moreover, the first exons and 5′-UTR regions of GPD1L, KCNH2, SCN5A and SNTA1 were not sufficiently covered (maximum coverage <50). Variants were included according to the following criteria: (1) passing the MiSeq Reporter quality filter on default settings based on coverage depth, variant score, mapping quality and quality score and (2) the coverage was above 50.

Alamut-HT v. 1.1.8 (Interactive Biosoftware, France) and Alamut v.2.2 (Interactive Biosoftware, France) was used for the annotation and evaluation of variants. In brief (Fig. 1), variants were selected by (1) frequency in reference populations [14, 15] and associations to cardiac diseases in HGMD [16], (2) in silico analysis of the likelihood of pathogenicity (Tables 2 and 3) and (3) evaluation and classification of each variant by two medical doctors. Variants were classified as likely pathogenic or of unknown significance based on the allele frequency of the variant, segregation evidence, in vitro functional studies, number of affected individuals found with the variant, computational prediction of the functional effect of amino acid substitutions and status as a de novo mutation. The algorithm used for in silico analysis was inspired by van Spendonck-Zwarts et al. [17].

Fig. 1
figure 1

Selection and evaluation of variants. After annotation (step 1), variants were selected manually (step 2) or through the in silico analysis (step 3). Using Alamut (Version 2.2, Interactive Biosoftware, LLC, France), variants were evaluated in terms of clinical relevance (step 4) and categorised as “likely pathogenic”, “unknown pathogenicity” or “unlikely pathogenic”. ESP (EA) minor-allele frequency in European American population in NHLBI GO Exome Sequencing Project (ESP) [14]; dbSNP variation global minor allele in the single nucleotide polymorphism database [15], HGMD Human Gene Mutation Database [16], VUS variant of unknown significance

Full size image
Table 2 In silico analysis scoring scheme for missense and nonsense variants
Full size table
Table 3 Sub-classifications of variants based on in silico analysis
Full size table

The fragment lengths of the MLPA investigations were analysed with GeneMapper v. 4.0 (Thermo Fisher, USA) and visual inspection of the electropherograms. Probe ratios below 0.7 or above 1.3 were regarded as heterozygous deletions or duplications, respectively.

The software, R v. 2.11.0 [18] was used to perform the statistical analysis of medians, means and standard deviations. The genetic results of the two investigated groups were compared with Fisher’s exact test.

Ethical standards

The study was approved by the Committees on Health Research Ethics in the Capital Region of Denmark (H-2-2012-017), local ethics committee and the Danish Data Protection Agency (2011-54-1262).

Results

Possibly pathogenic variants in the SUD victims and cardiac patients

Among the 44 investigated samples, the mean number of variants per sample that passed the MiSeq Reporter quality filter was 528 (s.d. 124). After selection of candidate variants by the process described in Fig. 1 and Tables 2 and 3, a mean of 40 (s.d. 8) possibly pathogenic variants per sample were further evaluated and classified.

Among the 15 SUD victims, three likely pathogenic variants including one novel, were identified in three (20 %) cases (Table 4). No larger rearrangement (indels) was found. In 12 (41 %) of the 29 patients, 18 likely pathogenic variants including three novel, were identified. Large deletions were found in a Brugada patient (SCN5A deletion in exon 23) and a LQTS patient (KCNQ1 deletion in exon 7, 8 and 9). The coverage in the deleted areas was approximately 50 % of the average coverage. Both deletions were verified with MLPA. No other insertion or deletion of large gene segments was observed. Details of all possibly pathogenic variants are described in Table 4. The difference in the number of observed likely pathogenic variants in the SUD cases and the patients was not statistically significant (p = 0.1).

Table 4 Specifications of likely pathogenic and novel variants identified in the study
Full size table

Variants of unknown significance

In all of the SUD victims and in 26 (90 %) of the patients, variants previously associated with cardiac diseases but with an unknown significance, were identified (Supplementary material online, Table S1). Two previously identified variants (K1459N and A1777T in MYH7) in patients diagnosed with hypertrophic cardiomyopathy (HCM) were found in two of our BS patients. Both variants were categorised as being of unknown significance because of the phenotypic variation between the patients.

Discussion

With NGS, we identified likely pathogenic variants in three of the 15 SUD victims (20 %). In the patients, the corresponding frequency was 41 %. In addition, two larger deletions of whole exons were identified in the patient cohort, increasing the diagnostic rate of the genetic investigation to 48 %. The frequency of identified likely pathogenic variants in the clinically diagnosed patients was >2-fold compared with that of the SUD victims, although not statistically significant (p = 0.1). The difference is most likely due to the reduced possibility of phenotypic characterisation of the forensic victims [19].

Previous genetic studies of SUD cases showed pathogenic variants in 11–17 % of the deceased [20–22] when investigating up to five genes associated with channelopathies (KCNQ1, KCNH2, SCN5A, KCNE1, KCNE2). However, Tester et al. [4] found putative pathogenic variants in 26 % of the individuals in the same five genes when investigating a large cohort comprising 173 SUD victims. In the 26 % with a pathogenic variant, the median age was 16 years. The corresponding number in our study was 37 years. Moreover, Tester et al. [4] found a higher frequency of pathogenic variants in the female SUD cases compared to that of the male SUDs (39 and 18 %, respectively). In our study, the males constituted 11 of the 15 SUD cases. Disease-causing variants in patients diagnosed with LQTS, BS and CPVT were previously identified in 19–50 % of the investigated cases [23–27]. None of the studies were performed with NGS and the number of investigated genes was smaller.

Compared to previous studies with similar cohorts, we found higher frequencies of likely pathogenic variants in both deceased and living patients when investigating 34 genes associated with channelopathies. This suggests that the gene panel used in this study is relevant in both clinical and forensic investigations. Due to the wide screening, we found that five of the cardiac patients carried several likely pathogenic variants in two or more genes. This may result in a compound effect, worsening the phenotypic trait of the disease. We also observed an overlap in the established phenotypes and the identified genotypes, specifically two BS diagnosed patients were identified with two variants previously found in patients diagnosed with hypertrophic cardiomyopathy [28, 29]. This could suggest that the diseases are expressed differently among individuals or not related to the described phenotypes and low frequent variants. In both the clinical and forensic settings, many cases have unspecific symptoms or presentation that makes it impossible to establish a clear phenotypic diagnosis. Hence, a wider genetic screening would be preferable in these cases and could contribute to the clinical investigation. Papadakis et al. [30] evaluated family members of victims of either sudden arrhythmic death syndrome (SADS) (n = 163) or with unspecific abnormalities of the heart (n = 41). They found a similar proportion (47 and 51 %, respectively) of channelopathic diseases in the relatives of the two groups. The predominant diagnoses were LQTS and BS in both cohorts. This demonstrates that there is diversity in both phenotypic and genotypic expression and that we should also include genes associated with structural cardiac diseases when investigating arrhythmogenic diseases.

We found a high frequency of variants associated with a cardiac disease with unknown significance. The variants were present in ESP and dbSNP in higher frequencies than the expected phenotype prevalence or present in several individuals in our study cohort. Similarly, Andreasen et al. [31] recently investigated the frequency of previously cardiomyopathy-associated variants in population-based exome data. They found a significant overrepresentation of variants previously thought to be pathogenic in the control population which indicated that the genetic variants may not be disease-causing but may have a modifying effect or may even be benign.

To establish the real pathogenicity of the described variants in this study, family co-segregation studies and functional testing should be performed. Non-coding and regulatory regions were not analysed unless they had previously been found to be associated with a cardiac disease or symptom. A large amount of genetic information that may be of importance has not yet been published. For future work, focus ought to be extended to the intronic and UTR regions in order to fully understand the genetics behind the diseases causing SCD. Also, epigenetic factors may have a possible influence on the pathogenesis and the medical relevance of epigenetic regulation remains largely unexplored in cardiac pathology [32].

This study shows that NGS is a valuable tool in a postmortem forensic setting with enhanced results to previous first-generation sequencing studies. Additionally, it is possible to use NGS to detect large deletions of entire exons. As described above, we found inconsistencies between phenotypes and genotypes, suggesting that there is an overlap among the channelopathic and cardiomyopathic diseases. For future studies, wide genetic screening of both channelopathic and cardiomyopatic associated genes should be performed in suspected SCD victims to fully understand the phenotype-genotype correlation.