High proportion of genetic cases in patients with advanced cardiomyopathy including a novel homozygous Plakophilin 2-gene mutation

Cardiomyopathies might lead to end-stage heart disease with the requirement of drastic treatments like bridging up to transplant or heart transplantation. A not precisely known proportion of these diseases are genetically determined. We genotyped 43 index-patients (30 DCM, 10 ARVC, 3 RCM) with advanced or end stage cardiomyopathy using a gene panel which covered 46 known cardiomyopathy disease genes. Fifty-three variants with possible impact on disease in 33 patients were identified. Of these 27 (51%) were classified as likely pathogenic or pathogenic in the MYH7, MYL2, MYL3, NEXN, TNNC1, TNNI3, DES, LMNA, PKP2, PLN, RBM20, TTN, and CRYAB genes. Fifty-six percent (n = 24) of index-patients carried a likely pathogenic or pathogenic mutation. Of these 75% (n = 18) were familial and 25% (n = 6) sporadic cases. However, severe cardiomyopathy seemed to be not characterized by a specific mutation profile. Remarkably, we identified a novel homozygous PKP2-missense variant in a large consanguineous family with sudden death in early childhood and several members with heart transplantation in adolescent age.


Study population
We included patients with no-ischemic severe heart failure with need of HTx or mechanical circulatory support. We also included patients transferred to our center for the evaluation to list for HTx. A positive family history for cardiomyopathy or in case of presumed sporadic cases age at diagnosis 40 years were defined as additional study criteria. This led to the inclusion of 30 DCM (21 familial, 9 sporadic) and 3 RCM index-patients (1 familial, 2 sporadic). Additionally, for ARVC-cases an at least possible diagnosis of ARVC according the revised Task Force Criteria [13] and the modified criteria [14] in case of pediatric patients (ARVC-01 and ARVC-08) was a prerequisite. Ten ARVC index-patients (6 familial, 4 sporadic cases) met the criteria and were included in the study. Of these 4 patients had ARVC with left ventricular dilation and had been treated by HTx, and total artificial heart implantation, respectively. The remaining 6 cases were diagnosed with arrhythmias and right ventricle enlargement/failure. Four of them underwent HTx, two cases were evaluated for HTx and treated with an ICD. All patients (n = 43) were genotyped in 46 cardiomyopathy-associated genes (see below).
DCM and RCM were diagnosed according to established criteria [15,16]. Clinical characteristics of the index-patients are summarised in Table 1, S1 and S2 Tables. All patients were of European/Caucasian origin. The written informed consent of all patients was obtained. In case of minors patient written informed consent of the parents were obtained. The study was approved by the local ethics committee (Ruhr University Bochum, Bad Oeynhausen, Germany, Reg.-No. 27.1/2011). This study was complied with the Declaration of Helsinki.  [17] (for gene list see S3 Table). Enrichment, sequencing kits and sequencing system were

*
New York Heart Association functional classification of heart failure before VAD or HTx, respectively. † Classification according Task Force Criteria [13]. ‡ Familial disposition for cardiomyopathy due to pedigree analysis and anecdotal evidence. § Sporadic = isolated cases without family history. II Treated medically.

#
Progressive cardiomyopathy with familial disposition due to pedigree analysis or anecdotal evidence, respectively. **Not yet listed for HTx but in the long term surveillance program. † † Classification according to Task Force Criteria [13] and consideration of the modified pediatric criteria [14]. For pedigrees see S2 Fig [18], GATK [19], and SAMtools [20]). A minimum read depth of ! 20 reads was set as quality cut-off (quality data see S4 Table). For variant annotation the VariantStudio™ v2.1 software provided by Illumina Inc with the UCSC Genome Build hg19 as reference was used. We filtered variants located in exon-and consensus splice-sequences (± 1 and 2 position) with a global minor allele frequency (MAF) 0.0005 according to the data provided by the Exome Aggregation Consortium (Version 0.3, ExAC data and browser (beta) Version) [21]. Synonymous variants were not taken into account. Variants of interest were verified by Sanger sequencing (BigDye 1 Terminator v1.1 Cycle Sequencing Kit, ABI PRISM 1 3100 genetic analyzer, Applied Biosystems, Foster City, CA, USA).

Variant classification
This study was primarily designed to identify dominant single gene mutations with clear clinical impact on ARVC or other cardiomyopathy subtypes. Therefore the guidelines [12] of the American College of Medical Genetics and Genomics (ACMG) were considered for variant classification. The ACMG defined five variant classes: class 5, pathogenic; class 4, likely pathogenic; class 3, variant of uncertain significance (VUS); class 2, likely benign; class 1, benign. For use on mainly private variants the criteria and rules for classification have been adjusted (for details see S1 Methods). The cut-off for the minor allele frequency (MAF) of variants of interest was set to a value 0.0005 and thus within the range of the respective disease prevalence (http://www. orpha.net) [22]. The ExAC browser [21] was used as reference for allele frequencies. Only frequencies of the non-Finnish European population were taken into account. Original reports listed in the Human Mutation Database Professional 2015.1 (HGMD 1 ) [8] and the ARVC Genetic Variants Database [23,24] were reviewed concerning variant classification. Deleterious effect of a variant was considered when !3 prediction tools forecasted the variant to be deleterious (MutPred [25], SNP&GO [26,27], Fathmm [28], and PolyPhen-2 (Polymorphism Phenotyping v2, humVar) [29]). In case of the TTN-gene only variants affecting the cardiac titin isoforms N2B and/or N2BA were further investigated. In this manuscript only (likely) pathogenic gene variations are referred to as 'mutation' while (likely) benign variants, variants of uncertain significance and gene variations in a general context are called 'variant'.

Gene and variant nomenclature
The nomenclature of sequence variants based on the rules of the Human Genome Variation Society (HGVS, www.hgvs.org). Nucleotide and amino acid numbering refers to the reference sequences given by the National Center of Biotechnology Information (NCBI, http://www. ncbi.nlm.nih.gov) (S3 Table). In case of TTN the nomenclature refers to the longest predicted isoform (NM_001267550.1, NP_001254479.1). All variations listed in this manuscript were submitted to ClinVar (http://www.ncbi.nlm.nih.gov/clinvar) using the submission name HDZNRW. An already existing listing of the study variants in the ClinVar database was checked. Some of them have been reported by others, a few with conflicting interpretation of pathogenicity. The listing of TTN variants was only partially verified because of inconsistent nomenclature based on different isoforms.

Patient cohort
The study cohort comprised 43 unrelated index-patients (29 male, 14 female) with cardiomyopathy of different types (Fig 1). Of these, 33 patients presented with end-stage disease and already underwent implantation of VAD and/or HTx. Ten patients were evaluated for HTx due to clinical signs of advanced cardiomyopathy. Seventy-nine percent of DCM-and RCMpatients were in New York Heart Association (NYHA) classification III or IV (Table 1). Six ARVC-patients were classified with definite diagnosis (DD), 2 patients with possible (PD), and 2 patients with borderline diagnosis (BL) ( Table 1). The mean age at diagnosis was not significantly different in DCM-(0031±17 years) and ARVC-cases (35±17 years) (S1 Fig). Due to the small number statistical evaluations for RCM-cases were not possible in this study (Table 1). Familial disposition for the cardiomyopathy due to pedigree analysis and anecdotal evidence was suspected in 21of 30 DCM, 6 of 10 ARVC and once in RCM (n = 3) cases (Table 1). Sporadic cardiomyopathy was suspected in 15 patients. Where available the pedigrees illustrating heart disease within the families and variant co-segregation/penetrance are shown in S2 Fig.

Genotyping results
In the entire study population we identified (likely) pathogenic mutations in 56% of indexpatients (18 familial, 6 sporadic cases). Nine patients were carriers of a pathogenic and 15 carried a likely pathogenic mutation. VUS were found in 21% of cases (4 familial, 5 sporadic cases). An only likely benign variant was identified in one patient and variants with an allele frequency exceeding the disease prevalence and therefore not further investigated were identified in 9 patients. This means that in total in 23% (n = 10, 6 familial, 4 sporadic cases) of patients no disease associated variants were identified (Fig 2).

Variant classification
In 19 of 46 genes, a total of 57 variants were found (Fig 1). We identified 41 missense variants in addition to 3 nonsense, 7 frameshift, 3 in frame insertion/deletion, and 3 canonical splicesite variants. Out of these, 15 had entries in the mutation-databases [8,24], whereas 42 variants were previously not described in context with a disease. We evaluated the pathogenic evidence of all variants (S5 and S6 Tables) and classified four variants likely benign, whereas fifty-three variants were classified (likely) pathogenic or VUS, respectively. The potentially disease related variants affected 18 genes. In total, we identified 9 pathogenic mutations in 9 index-patients (8 familial cases, 1 sporadic case) in 6 different genes (LMNA, DES, RBM20, CRYAB, PKP2, PLN). We identified two or more variants in 14 DCM-and 3 ARVC-cases.
In DCM-patients TTN was the most frequently affected gene (51% of variants; Fig 1). In the DCM-cohort, pathogenic mutations were identified in DES, LMNA, and RBM20 (Table 1,  S5 Table), whereas likely pathogenic mutations were found in MYL2, MYH7, TTN, TNNC1, DES, LMNA, RBM20, NEXN, and PKP2. Remarkably, a homozygous PKP2-gene mutation (c.2035C>T, p.His679Tyr) was identified in a large Turkish family (DCM-23). As expected, PKP2 was the most frequently affected gene in ARVC-patients. Pathogenic mutations in the ARVC-cohort were found in PKP2 and PLN, a likely pathogenic mutation was found in LMNA (Table 1, S5 Table). Surprisingly, no further desmosomal genes were affected in ARVC-cases. In the RCM-cases two patients with pathogenic or likely pathogenic mutations in TNNI3 and CRYAB, respectively, were identified (Table 1, S5 Table). The median age at diagnosis in DCM-and ARVC-cases (33 and 36 years, respectively) was not significantly different in dependence of variant classification (S3 Fig). Twenty-one TTN variants (7 TTN-tv (premature stop codon with/without preceded frameshift, and canonical ± 1 or 2 splice site variants), 13 missense, 1 in-frame insertion) were identified in 16 (37%) of 43 index-patients (Table 1, S5 Table). In DCM-patients TTN-variants were more frequent (p<0.05; S4 Fig), whereas TTN-tv were increased by trend (p = 0.05, data not shown) when compared to the other cardiomyopathies. Most of the TTN-variants (5 TTNtv, 7 missense) affected the titin domains spanning the A-band of the sarcomere. But also variants in the I-band (3x missense, 2x TTN-tv), M-line (2x missense, 1x in frame insertion), and Z-band (1x missense) region of the protein were identified (S7 Table). All but one TTN-variant were located within exons that are estimated to be incorporated in 100% of the cardiac TTNtranscripts (PSI 100%, [6], http://cardiodb.org/titin) and might be expressed to a notable amount. The TTN-variant found in patient DCM-2 is located within the I-band and is estimated to be expressed only in 8% of the transcripts and is therefore believed to be of low disease relevance [6]. Seven (21%) of the 34 variant carriers of the present study had only TTNvariant(s). Of these four patients had a single TTN missense variant affecting the I-, A, and Zband of titin, respectively. One patient was carrier of a single TTN-tv in the A-band of the protein. DCM-16 had a splice and a missense TTN-variant in the A-and I-band, respectively. DCM-24 is carrier of one in-frame duplication and two missense variants affecting the M-line and the A-band, respectively (S5 and S7 Tables).

Family with a novel homozygous PKP2-gene mutation
We identified a novel homozygous PKP2-mutation c.2035C>T, p.His679Tyr in index-patient DCM-23 (III/4, Fig 4A) who is a member of a large non-consanguineous Yezidi family. The affected family members did not fulfill the ARVC Task Force criteria although PKP2 is generally associated with ARVC. Three genetically untested brothers died suddenly in childhood (III/2, III/3, III/13; 11 years, 2 years, 7 days, respectively). She and two siblings (male (III/1), female (III/10)) who are also homozygous for the PKP2-mutation underwent HTx in adolescent age (14, 13 and 17 years, respectively) because of DCM (ECGs shown in S5 Fig.). The 57 years old mother (II/8), the 66 years old father (II/9) and two other siblings (III/8, III/12) who are heterozygous mutation carriers have no signs of cardiac disease at all, suggesting a variant co-segregation in an autosomal recessive mode. Surprisingly, the youngest sister (III/11, 18 years old) who is also a homozygous mutation carrier has currently no detectable signs of cardiomyopathy as examined by ECG, TTE and MRI (data not shown). The suspected pathogenic impact of PKP2 c.2035C>T was verified by analyzing whole exome data of the parents, an unaffected sister (III/8), and the index-patient (III/4) (S1 Methods, S6 Fig). No other co-segregating variant with an allele frequency 0.0005 was found in cardiomyopathy genes listed in HGMD (n = 156, S8 Table). A further indication of pathogenicity is the absence of PKP2 c.2035C>T in the ExAc browser with sufficient coverage of the gene region (read depth of 50 in 80% of the alleles). Analysis of the PKP2 locus (Fig 4B, S1 Methods) of the index-patient (III/4) and her transplanted brother (III/1) revealed that they are homozygous for a haplotype surrounding the PKP2-gene which largely overlaps between father (II/9) and mother (II/8), suggesting that they are probably (distantly) related and the mutation might be inherited from a common ancestor. In order to study the pathogenic mechanism of the mutation we investigated plakophilin-2 mRNA and protein levels in explanted myocardium, but no differences compared to controls were detected (S1 Methods, S7 and S8 Figs). Immunofluorescence staining of myocardial sections of the explanted heart of individual III/10 showed normal localization of PKP2 at the intercalated disc (S1 Methods, S9 Fig). Thus, the mutation appears to have no influence on the stability or subcellular localization of PKP2 within the myocardium. No segregation with disease was observed for the nonsense variant in LAMA4 (S5 Table).

Discussion
This study focused on the genotyping of patients with advanced cardiomyopathy. Our inclusion criteria lead to a study cohort mainly composed of patients who have been bridged for heart transplantation and those already transplanted. The study design was set for the identification of highly penetrant monogenic causes of cardiomyopathies. In total 57 variants were identified of which a high proportion were classified (likely) pathogenic (51%). Surprisingly, (likely) pathogenic mutations were more frequently present in DCM-than in ARVC-cases (60% versus 40%) although the majority of currently known ARVC-genes with the exception of the minor disease genes TGFB3, CTNNA3 (see www. arvcdatabase.info), [4], FLNC [30], and SCN5A [31] were analyzed. No variants were identified even in one definite and one familial ARVC-case. This strongly suggests that other currently unknown genes and/or genotypes are involved in the development of ARVC. Especially future screening of the FLNC-gene could increase the variant detection because a prevalence of 1% of truncating FLNC-mutations was reported in dilated and arrhythmic cardiomyopathies [30].
The mutation rate and profile in ARVC patients is in discrepancy to what have been expected and published before by others. For example Groeneweg et al. [32] found a mutation rate of 63% in a large ARVC-cohort, and in desmosomal genes a mutation rate of 40% was reported [33]. The high mutation rate in our DCM-cohort is in the upper range of values reported so far [2,4,5], in case of familial DCM it is in agreement with the findings of Cuenca et al. [7]. Apparently, the mutation profiles differ notably between the different studies. As expected, screening for gene variants is more successful in familial cases, but also in presumably sporadic cases (likely) pathogenic mutations were identified in a noteworthy high number (25%). In previous studies the prevalence of (potentially) pathogenic mutations in sporadic DCM-cases was 8% and 68%, respectively [4,5]. Our study cohort includes 12 patients with disease onset at 20 years of age assuming a genetic cause of disease although familial cardiomyopathy was reported only for four cases. In eight patients an at least likely pathogenic gene mutation was identified. These findings underscore the relevance to perform genetic testing not only in suspected familial or sporadic cases with clinical clues suggestive of a particular/ rare genetic disease as recommended by the European Society of Cardiology [11] but in all index-patients with an unknown etiology of the disease. Nevertheless, a characteristic mutation profile for severe cardiomyopathy could not be defined.
Twenty five percent of our DCM cohort carried a TTN-tv mutation as reported for endstage DCM patients by Roberts et al. [6]. All TTN-tv mutations were classified as likely pathogenic considering a high prevalence of TTN-tv in DCM-cohorts. TTN alternative splicing, position-independent nonsense-mediated mRNA decay, an internal promoter, allele-specific expression, structural and functional effects on sarcomere formation and force generation as well as negative effects by incorporation of truncated titin into the sarcomere were discussed as pathomechanisms for truncated titin [34][35][36][37][38][39]. Nevertheless, many additional data are required for a better understanding of the mechanisms. Because of the background noise in the general population TTN missense variants were classified as VUS although recent largescale sequencing and functional analysis indicated strong impact on cardiomyopathies [40][41][42]. Protein unfolding and domain destabilization was shown and/or predicted for TTN missense mutations located in the Z-disk [42], A/I junction region [43], and in the I-band [41] of the protein, respectively. Decreased contractile force generation of a missense mutation located in the Z/I-band junction was reported by Hinson et al. [35]. We identified 7 indexpatients with only TTN variants of different types affecting different protein regions. But although functional data are available a final classification of TTN variants is not possible. Classification still based mostly on populations genetic data and one should keep in mind that new functional data could lead to variant re-classification. Two patients carried multiple TTN variants. We did not investigate whether the gene was affected in a compound heterozygous or homozygous way. Regardless, no data are available about patients with several TTN variants and about a possibly increasing impact on disease. In agreement with recent data [6,40] most TTN-variants affected the inextensible A-band of titin and are incorporated in 100% of the cardiac TTN-transcripts. [44,45]. Patient DCM-16 is carrier of a TTN-tv and missense variant, had disease onset at the age of 20 years and VAD implantation when she was 32 years old. Therefore, we suggest that TTN associated cardiomyopathy is present also in young adults even though mostly it has its onset in the mid-late life period [40].
We identified a non-synonymous PKP2-mutation that in homozygous carriers led to manifestation of end-stage DCM at young age. This mutation might be characterized by a reduced penetrance [46][47][48] since at present a homozygous 18 years old mutation carrier has no signs of cardiac disease. We cannot rule out additional genetic factor(s) underlying the disease of the family, although screening of all known cardiomyopathy associated genes gave no indication for this. For PKP2 truncating mutations and interestingly, also for missense mutations haploinsufficiency was indicated as a possible pathomechanism. Functional data to support this assumption are limited. Kirchner et al. [49] reported decreased in vivo and in vitro protein expression, impairment of protein structure and stability, and mislocation of plakophilin 2 in the cytoplasm in case of PKP2 missense mutations. In addition for other PKP2 missense mutations impairment of in vitro de novo desmosome assembly was described [50] and influence on the expression of heat shock protein 90 was observed in case of two PKP2 truncating mutations [51]. In our study the investigation of heart tissue of a homozygous mutation carrier revealed no impact of PKP2 p.His679Tyr on plakophilin expression, stability and cell localization. This might indicate a different pathomechanism as described above, although it must clearly be pointed out, that in contrast to our in vivo data the characteristics of the mutations described by Kirchner et al. based mainly on cell culture experiments. To our knowledge, this is the first report of a PKP2-missense mutation causing recessively inherited DCM. No signs of hair or skin abnormalities that were reported with other recessive desmosomal gene mutations [52,53] were observed.
We detected a missense mutation in the Z-disk protein NEXN that was already described to have a disruptive effect on the cardiac sarcomere in a DCM patient and a dominant negative effect on Z-disk stability in zebrafish [54]. In accordance, we found this mutation in a male DCM patient (DCM-19) who underwent VAD implantation at the age of 22 years. Based on the functional studies described by Hassel et al. [54], an allele frequency clearly below the prevalence of DCM, and multiple computational evidence for a deleterious effect on the protein this NEXN mutation must be classified as likely pathogenic according to ACMG guidelines. Our study is the second report that links the NEXN gene to DCM and thereby strengthens the importance of NEXN as a (minor) disease gene.
In our study, the usage of a sequencing panel with a comparatively small number of myopathy associated genes for genotyping patients has historical reasons. But despite the limited gene number it is important to point out that the majority of variants identified here were novel variants in known major disease genes which are listed in the Mendelian Inheritance in Man (MIM) compendium and which are covered by the 46-gene panel used here. As expected besides these also variants in minor disease genes were identified.
Extensive variation screening in other (minor) disease genes in the whole exome and genome, respectively, might provide further insights. However, variant classification with respect to the impact on disease development and clinical course still remains the real challenge [12]. Due to limited availability of case reports, small families with limited co-segregation data, and rarely available functional data, some ACMG [12] criteria were not applicable and others had to be adjusted. But apart from this limitation, the ACMG guidelines were a suitable tool for a uniform and differentiated variant classification. At present there are no guidelines available how to treat non-symptomatic mutation carriers, due to the heterogeneous genetic background of cardiomyopathies and the lack of pharmacological evidence for preventive treatment. In order to overcome this limitation in the future more studies like the recent trials that investigate the effect of drug treatment of pre-symptomatic PLN and LMNA mutation carriers are of impact (Clinical trials identifier: NCT01857856 and European Society of Cardiology Congress 2016, Rome, Poster P4981, respectively).

Limitation
We excluded variants with an allele frequency above the disease prevalence. Therefore, this study is not suited for the identification of a complex genotype or modifiers as well as variants with reduced penetrance. Copy number variants and large DNA-rearrangements have not been investigated. Only allele frequencies of the ethnically narrow non-Finnish European population provided by the Exome Aggregation Consortium were taken into account to define variants of further interest. Variant screening was limited to 46 myopathy associated genes covered by the TruSight™ Cardiomyopathy gene panel of Illumina. The recently described ARVC and DCM disease gene FLNC and the ARVC disease gene SCN5A were not screened for variants.