Matters of the heart: genetic and molecular characterisation of cardiomyopathies Posafalvi, Anna

Transcription

1 University of Groningen Matters of the heart: genetic and molecular characterisation of cardiomyopathies Posafalvi, Anna IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below. Document Version Publisher's PDF, also known as Version of record Publication date: 2015 Link to publication in University of Groningen/UMCG research database Citation for published version (APA): Posafalvi, A. (2015). Matters of the heart: genetic and molecular characterisation of cardiomyopathies [Groningen]: University of Groningen Copyright Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons). Take-down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim. Downloaded from the University of Groningen/UMCG research database (Pure): For technical reasons the number of authors shown on this cover page is limited to 10 maximum. Download date:

2 Matters of the heart: genetic and molecular characterisation of cardiomyopathies Pósafalvi Anna

3 The work described in this thesis was supported by the University Medical Center Groningen, the Jan Kornelis de Cock Foundation, the NutsOhra Foundation and the Netherlands Heart Foundation. Printing of this thesis was supported by the Graduate School of Medical Sciences and the University Library, University of Groningen, Groningen, the Netherlands. Copyright: 2015 by Anna Posafalvi All rights reserved. No parts of this book may be reproduced, stored in retrieval system, or transmitted in any form or by any means without prior written permission of the author and the publishers holding the copyrights of the published articles. Cover photograpy: Anna Posafalvi Design: dreamed of by Anna Posafalvi, dreams made come true by Joanna Smolonska Layout and printing: Lovebird Design & Printing Solutions ISBN: (printed) (electronic)

4 Matters of the heart: genetic and molecular characterisation of cardiomyopathies PhD thesis to obtain the degree of PhD at the University of Groningen on the authority of the Rector Magnificus Prof. E. Sterken and in accordance with the decision by the College of Deans. This thesis will be defended in public on Monday 20 April 2015 at 16:15 hours by Anna Posafalvi born on 17 June 1986 in Debrecen, Hungary

5 Supervisor Prof. RJ Sinke Co-supervisor Dr. JDH Jongbloed Assessment committee Prof. MH Breuning Prof. RA de Boer Prof. RMW Hofstra

6 D rá ga Na gya p á mn a k... To my dearest grandfather... Wheresoever you go, go with all your heart. (Confucius)

7 Paranymphs Ena Sokol Eva Teuling

8 TABLE OF CONTENTS Preface: about cardiomyopathies in a nutshell 9 Outline of this thesis 15 Appendix 1 List of genes 16 Frequently used abbreviations 18 Chapter 1: Introduction New clinical molecular diagnostic methods for congenital and inherited heart disease (Expert Opin Med Diagn 2011, review) Chapter 2: Candidate gene screening 2.1: Mutational characterisation of RBM20 in dilated cardiomyopathy and other cardiomyopathy subtypes 2.2: Missense variants in the rod domain of plectin increase susceptibility to arrhythmogenic right ventricular cardiomyopathy Chapter 3: Exome sequencing 3.1: Hunting for novel disease genes in autosomal dominant cardiomyopathies: elucidating a role for the sarcomeric pathway 3.2: Homozygous SOD2 mutation as a cause of lethal neonatal dilated cardiomyopathy 3.3: One family, two cardiomyopathy subtypes, three disease genes: an intriguing case Chapter 4: Targeted sequencing 4.1: Gene-panel based Next Generation Sequencing (NGS) substantially improves clinical genetic diagnostics in inherited cardiomyopathies 4.2: Titin gene mutations are common in families with both peripartum cardiomyopathy and dilated cardiomyopathy (Eur Heart J 2014) Chapter 5: Discussion and future perspectives 233 Summary 251 Appendix 2 List of authors and affiliations 268 About the author 270 Acknowledgements

9

10 about cardiomyopathies in a nutshell The disease Cardiomyopathy is an insidious disease of the heart muscle (myocardium) leading to decreased pumping capacity, and resulting in a wide range of symptoms. These range from mild (dizziness, fatigue, chest pain or oedema) to severe (heart failure, arrhythmia, embolism, or even sudden death). PREFACE Figure 1. Schematic cross-section of a healthy heart (a) and hearts with DCM (b), HCM (c) and ARVC (d) In dilated cardiomyopathy (DCM), the left ventricle becomes enlarged with a thin, weakened muscle wall, and is unable to generate enough pumping force during contractions; the myocardium is thickened in hypertrophic cardiomyopathy (HCM); in arrhythmogenic right ventricular cardiomyopathy (ARVC) fibrofatty infiltration of the myocardium leads to arrhythmia. Figure published by Wilde & Behr, Nature Reviews Cardiology, 2013; used with permission. For more information on cardiomyopathy types, see box 1. THE DISEASE 9

11 Clinical diagnosis and treatment guidelines When the symptoms of cardiomyopathy appear, the diagnosis of the disease is most frequently made by electrocardiogram (ECG), and noninvasive imaging techniques such as an X-ray of the chest, echocardiography (imaging of the heart with ultrasound), or MRI. In addition, patients receive a general medical examination combined with a simple blood test (measuring, for instance, molecular markers of heart failure or kidney function). Less regularly, cardiac catheterization or coronary angiography is used. Both these methods are minimally invasive, only a thin tube is inserted in one of the biggest veins of the body and threaded to the heart, instead of an open surgery. These help physicians acquire either a myocardial biopsy for further experimental analysis or enough information to exclude potential blockage (stenosis) of the heart and the coronary blood vessels. The therapy for cardiomyopathy largely depends on the disease type and the severity of the symptoms. Therapy aims at slowing down the progression of the disease or at disease prevention in susceptible individuals through life style changes and medical treatment using different antihypertensive, antiarrhythmic, diuretic or anticoagulant drugs (e.g., ACE inhibitors or calcium antagonists). In serious cases of arrhythmia, the implementation of an ICD (a small, implantable defibrillator) or a pacemaker may be the solution. Heart transplantation is only considered as a last resort in patients with end-stage heart failure. The genetic causes Even though there are several environmental factors that may trigger the onset of cardiomyopathy (viral infections, the use of certain drugs, alcoholism, and other cardiovascular conditions, as well as certain systemic disorders), we often see the disease running in families (30-50% of ARVC and DCM cases; see box 1 for definitions). Most of these familial cardiomyopathy cases can be explained by an autosomal dominant (AD) inheritance pattern. To date, about 76 genes are known to be involved in different types of cardiomyopathy, which often show considerable genetic overlap (figure 2), and the majority of these 76 genes show AD inheritance. Additionally, a few genes, such as DMD, EMD, GLA, LAMP2, or TAZ, are involved in the X-linked form of the disease. Exceptionally, autosomal recessive inheritance is also observed. These patients usually exhibit more severe symptoms, and the disease generally begins in infancy or early childhood (paediatric cardiomyopathies; the genes involved include ANO5, MYL2, PKP2, TNNI3). 10 CLINICAL DIAGNOSIS AND TREATMENT GUIDELINES

12 Although it is also known that abnormalities in mitochondrial DNA can contribute to the pathogenesis of different cardiomyopathies (e.g., mutations of MTTL1), this has not yet been extensively studied. The possible complex, oligogenic or multifactorial causes for cardiomyopathies have also not been investigated in detail, nor have the potential roles of risk alleles of lower effect size, copy number variations (such as those including the BAG3 or PRDM16 genes), or micrornas. To date, a significant proportion of familial cardiomyopathies (about 30-40% of HCM, 40-50% of ARVC, and around 50% of DCM cases; see box 1 for definitions) remain genetically unexplained. PREFACE Figure 2. Cardiomyopathy disease genes and the genetic overlap between subtypes of the disease (updated from Jongbloed at el, EOMD 2011) Not only is there considerable phenotypic overlap between the subtypes of cardiomyopathy, many genes are also involved in multiple forms of the disease. The official full names of the abbreviated genes, according to OMIM, are listed in appendix 1. THE GENETIC CAUSES 11

13 Types of cardiomyopathy There are various forms of cardiomyopathy, each with different underlying causes for the insufficient circulation. The cardiomyopathies investigated in this thesis include: 1. dilated cardiomyopathy (DCM): one or both of the ventricles (in most cases only the left one) become enlarged with a thin, weakened muscle wall unable to generate enough pumping force during contractions (figure 1) 2. arrhythmogenic right ventricular cardiomyopathy (ARVC): the replacement of the degenerating myocardium with scar (fibrofatty) tissue results in disturbed electrical signals and conduction in the heart (arrhythmia) 3. hypertrophic cardiomyopathy (HCM): a thickened myocardium due to abnormal growth and arrangement (hypertrophy and disarray) of muscle fibres results in smaller chamber volume and sometimes blocks the blood flow (obstruction) 4. restrictive cardiomyopathy (RCM): due to their stiffness, the ventricles do not get refilled with enough blood during relaxation, hence the heart cannot supply the organs with sufficient circulation during contraction 5. left-ventricular non-compaction cardiomyopathy (LVNC): the wall of the left ventricle is spongiform, characterized by a meshwork of muscle fibres 6. peripartum cardiomyopathy (PPCM): a special form of dilated cardiomyopathy that becomes manifest towards the end of pregnancy or within a few months following delivery 7. paediatric cardiomyopathy: this type of cardiomyopathy becomes manifest in infancy or early childhood, and is usually characterized by more severe symptoms and worse outcomes than when the disease manifests in adulthood (from a structural-functional point of view, most frequently it is DCM>HCM>RCM>ARVC) Our methods Candidate gene screening Sanger-sequencing: This method of DNA-sequencing allows us to detect single nucleotide changes and small indels of DNA fragments with an average size of base pairs. It can be used for screening candidate genes in a large cohort of patients, as well as for segregation analysis of a variant within a family, or for confirmation of DNA-variations detected by high-throughput sequencing. Disease gene mapping haplotype sharing test (HST): An ideal, SNP-genotyping-based method for small cardiomyopathy families, who are usually not suitable for classical linkage analysis. With this method, we aim to identify chromosomal regions shared among affected family members, hypothesizing that the highest chance of finding the mutation is in the largest shared region of the family. We use this method as a filtering step in exome sequencing data analysis if a variant is located in the 2 nd largest shared haplotype of 10 cm, it is more likely to be causative than a variant located in the 57 th largest shared haplotype of only 0.1 cm. High-throughput sequencing exome sequencing: Sequencing all coding parts (exons) of all genes (about 1% of the genome). Though costly and requiring intensive data analysis, this method is suitable for identifying private coding mutations of novel disease genes in families with an unknown genetic cause of cardiomyopathy. gene-panel based (targeted) sequencing: High-throughput sequencing of a DNA sample previously enriched for the small set of genes we are interested in. Since this method results in very high coverage across the regions of interest and high data quality, it has recently been implemented in routine diagnostics. 12 THE GENETIC CAUSES

14 The challenges we face Identifying a novel disease gene carrying the heterozygous causal variant (heterozygous because of the dominant inheritance) is usually more challenging than working on a recessive disease, but there are also other complications to be considered in our research. Cardiomyopathy is, in general, a late onset disease. For example, DCM usually begins between 20 and 50 years of age, while most ARVC patients are diagnosed before 40 years of age. Thus, low penetrance of the disease at young age makes it difficult to make the genetic diagnosis in a family as the disease status of young relatives is uncertain (partly due to the variety in the nature and severity of the symptoms). Furthermore, phenocopies also occur, with family members having comparable symptoms due to an independent cause (e.g. developing disease on the basis of another, often non-genetic, cardiovascular event: coronary artery disease). In consequence, the medical diagnosis of cardiomyopathy is based on exclusion criteria and performing segregation analysis for a putative pathogenic variant in families without being absolutely sure of the healthy/affected status of the screened individuals can be complicated. Since cardiomyopathy can be so difficult to diagnose, and because the chances of a successful treatment rapidly decline with time, our aims are (1) to obtain an early (molecular) diagnosis of the inherited form of the disease before severe symptoms become manifest, and (2) to enable preventive treatment (including life-style changes as well as medical treatment if necessary) of the endangered individuals, combined with regular, thorough cardiological check-ups. PREFACE THE CHALLENGES WE FACE 13

15 Recommended literature website of the National Heart, Lung, and Blood Institute, health topic on cardiomyopathies: nhlbi.nih.gov/health/health-topics/topics/cm/ website of the Children s Cardiomyopathy Foundation: main_brochure.htm Herschberger RE, Lindenfeld J, Mestroni L et al: Genetic evaluation of cardiomyopathy a Heart Failure Society of America guideline. J Cardiac Fail 2009;15:83-97 Wilde AA & Behr ER: Genetic testing for inherited cardiac disease. Nat Rev Cardiol 2013;10: Teekakirikul P, Kelly MA, Rehm HL et al: Inherited cardiomyopathies Molecular genetics and clinical genetic testing in the postgenomic era. J Mol Diagn 2013;15: Jongbloed JD, Pósafalvi A, Kerstjens-Frederikse WS et al: New clinical molecular diagnostic methods for congenital and inherited heart disease. Expert Opin Med Diagn 2011;5:9-24 Posafalvi A, Herkert JC, Sinke RJ et al: Clinical utility gene card for: dilated cardiomyopathy (CMD). Eur J Hum Genet 2013;21 doi: /ejhg Te Rijdt WP, Jongbloed JD, de Boer RA et al: Clinical utility gene card for: arrhythmogenic right ventricular cardiomyopathy (ARVC). Eur J Hum Genet 2014;22. doi: /ejhg Udeoji DU, Philip KJ, Morrissey RP et al: Left ventricular noncompaction cardiomyopathy: updated review. Ther Adv Cardiovasc Dis 2013;7: Caleshu C, Sakhuja R, Nussbaum RL et al: Furthering the link between the sarcomere and primary cardiomyopathies: restrictive cardiomyopathy associated with multiple mutations in genes previously associated with hypertrophic or dilated cardiomyopathy. Am J Med Genet A 2011;155A: Peled Y, Gramlich M, Yoskovitz G et al: Titin mutation in familial restrictive cardiomyopathy. Int J Cardiol 2014;171:24-30 Wooten EC, Hebl VB, Wolf MJ et al: Formin homology 2 domain containing 3 variants associated with hypertrophic cardiomyopathy. Circ Cardiovasc Genet 2013;6:10-8 Chang B, Nishizawa T, Furutani M et al: Identification of a novel TPM1 mutation in a family with left ventricular noncompaction and sudden death. Mol Genet Metab 2011;102:200-6 Luxán G, Casanova JC, Martínez-Poveda B et al: Mutations in the NOTCH pathway regulator MIB1 cause left ventricular noncompaction cardiomyopathy. Nat Med 2013;19: Purevjav E, Varela J, Morgado M et al: Nebulette mutations are associated with dilated cardiomyopathy and endocardial fibroelastosis. J Am Coll Cardiol 2010;56: Arndt AK, Schafer S, Drenckhahn JD et al: Fine mapping of the 1p36 deletion syndrome identifies mutation of PRDM16 as a cause of cardiomyopathy. Am J Hum Genet 2013;93:67-77 Ohno S, Omura M, Kawamura M et al: Exon 3 deletion of RYR2 encoding cardiac ryanodine receptor is associated with left ventricular non-compaction. Europace 2014;16: Pinto JR, Yang SW, Hitz MP et al: Fetal cardiac troponin isoforms rescue the increased Ca2+ sensitivity produced by a novel double deletion in cardiac troponin T linked to restrictive cardiomyopathy: a clinical, genetic, and functional approach. J Biol Chem 2011;286: van Hengel J, Calore M, Bauce B et al: Mutations in the area composita protein αt-catenin are associated with arrhythmogenic right ventricular cardiomyopathy. Eur Heart J 2013;34: Pruszczyk P, Kostera-Pruszczyk A, Shatunov A et al: Restrictive cardiomyopathy with atrioventricular conduction block resulting from a desmin mutation. Int J Cardiol 2007;117: RECOMMENDED LITERATURE

16 OUTLINE OF THIS THESIS The aims of this thesis are (1) to provide a better understanding of the genetic background and the molecular pathomechanism of familial cardiomyopathies, (2) to identify novel disease genes in unsolved families, and (3) to improve the existing methods of molecular diagnostic testing. Chapter 1 is a detailed introduction to the field of cardiogenetics. This chapter reviews congenital and late onset heart diseases (the latter referring to cardiomyopathies and arrhythmia syndromes), categorizes the genes involved in the different types of heritable heart diseases, and thoroughly describes the research methods with special attention paid to their potential future diagnostic applications in cardiovascular diseases. The subsequent chapters contain experimental data and are subdivided based on the research methods used. In chapter 2, we applied the classical candidate gene screening approach Sanger sequencing. We were interested if (and to what extent) the known DCM gene RBM20 contributes to the genetic background of the disease in Dutch patients (2.1). In addition, we hypothesized that the desmosomal PLEC gene may play a role in the development of ARVC. In an attempt to prove this, we studied the clustering of sequence variations in patients compared to that in a healthy control population (2.2). High-throughput sequencing is a recent technological development that is revolutionizing the science of genetics. We applied two different experimental designs of this method to elucidate genetic causes for cardiomyopathies. In chapter 3, we have described families where mutations in known cardiomyopathy genes had been excluded, and we successfully applied exome sequencing to identify novel disease genes in both autosomal dominant (3.1) and recessive (3.2) cardiomyopathies, while 3.3 is an interesting case report on a family suffering from both forms of the disease. In chapter 4, we applied targeted enrichment of DNA samples to a set of well-defined candidate disease genes. We address the applicability and the quantitative advantages of targeted sequencing in routine diagnostics for a cohort of 252 unselected cardiomyopathy patients in 4.1, while report our findings on targeted sequencing of PPCM/DCM families in 4.2. The work described in this thesis is then discussed in a broader context, and future perspectives for the use of high-throughput sequencing in research and diagnostic settings, as well as potential research directions in the field of cardiogenetics, are presented in chapter 5. OUTLINE OF THE THESIS ABOUT CARDIOMYOPATHIES IN A NUTSHELL 15

17 APPENDIX 1 List of cardiomyopathy genes: (official abbreviations and names of genes included in figure 2 of the preface) ABCC9 ATP-binding cassette, subfamily C (CFTR/MRP), member 9 ACTC1 actin, alpha, cardiac muscle 1 ACTN2 actinin, alpha 2 ANKRD1 ankyrin repeat domain 1 (cardiac muscle) ANO5 anoctamin 5 BAG3 BCL2-associated athanogene 3 CALR3 calreticulin 3 CAV3 caveolin 3 CRYAB crystalline, alpha B CSRP3 cysteine and glycine-rich protein 3 (cardiac LIM protein) CTNNA3 catenin (cadherin-associated protein), alpha 3 DES desmin DMD dystrophin DOLK dolichol kinase DSC2 desmocollin 2 DSG2 desmoglein 2 DSP desmoplakin DTNA dystrobrevin, alpha EMD emerin EYA4 EYA transcriptional coactivator and phosphatase 4 FHL1 four and a half LIM domains 1 FHL2 four and a half LIM domains 2 FHOD3 formin homology 2 domain containing 3 FKRP fukutin related protein FKTN fukutin FXN frataxin GATAD1 GATA zinc finger domain containing protein 1 GLA galactosidase, alpha ILK integrin-linked kinase JPH2 junctophilin 2 JUP junction plakoglobin LAMA4 laminin, alpha 4 LAMP2 lysosomal-associated membrane protein 2 LDB3 LIM domain binding 3 LMNA lamin A/C MIB1 mindbomb E3 ubiquitin protein ligase 1 MT-TL1 mitochondrially encoded trna leucine 1 (UUA/G) 16 APPENDIX 1

18 MYBPC3 myosin-binding protein C, cardiac MYH6 myosin, heavy chain 6, cardiac muscle, alpha MYH7 myosin, heavy chain 7, cardiac muscle, beta MYL2 myosin, light chain 2, regulatory, cardiac, slow MYL3 myosin, light chain 3, alkali; ventricular, skeletal, slow MYLK2 myosin light chain kinase 2 MYOZ2 myozenin 2 MYPN myopalladin NEBL nebulette NEXN nexilin (F actin binding protein) NKX2-5 NK2 homeobox 5 mtdna mitochondrial DNA PDLIM3 PDZ and LIM domain 3 PKP2 plakophilin 2 PLN phospholamban PRDM16 PR domain containing 16 PRKAG2 protein kinase, AMP-activated, gamma 2 noncatalytic subunit PSEN1 presenilin 1 PSEN2 presenilin 2 PTPN11 protein tyrosine phosphatase, non-receptor type 11 RAF1 Raf-1 proto-oncogene, serine/threonine kinase RBM20 RNA binding motif protein 20 RYR2 ryanodine receptor 2 (cardiac) SCN5A sodium channel, voltage-gated, type V, alpha subunit SDHA succinate dehydrogenase complex, subunit A, flavoprotein (Fp) SGCD sarcoglycan, delta (35kDa dystrophin-associated glycoprotein) TAZ tafazzin TBX20 T-box 20 TCAP titin-cap TGFB3 transforming growth factor, beta 3 TMEM43 transmembrane protein 43 TMPO thymopoietin TNNC1 troponin C type 1 (slow) TNNI3 troponin I type 3 (cardiac) TNNT2 troponin T type 2 (cardiac) TPM1 tropomyosin 1 (alpha) TTN titin TTR transthyretin TXNRD2 thioredoxin reductase 2 VCL vinculin APPENDIX 1 LIST OF CARDIOMYOPATHY GENES 17

19 APPENDIX 1 Frequently used abbreviations: ACE AD angiotensin convertase enzyme autosomal dominant inheritance pattern AGVGD align Grantham variation Grantham distance (pathogenicity prediction software for missense variants) AR ARVC bp CGH CHD cm CNV DCM dbsnp DMEM DNA EBS ECG ES ESP autosomal recessive arrhythmogenic right ventricular cardiomyopathy base pair comparative genomic hybridization congenital heart disorders centimorgan copy number variation dilated cardiomyopathy NCBI s SNP database Dulbecco s Modified Eagle Medium deoxyribonucleic acid epidermolysis bullosa simplex electrocardiogram exome sequencing E. coli Escherichia coli FBS GERP exome sequencing project (variant database of the NHLBI) fetal bovine serum genomic evolutionary rate profiling (a score indicating the evolutionary conservation of a nucleotide) GoNL Genome of the Netherlands (database of the genomes of 500 individuals, used as a frequency database of the Dutch wild type ) GWAS HCM HEK HF HiSeq HLA HST H 2 O H 2 O 2 ICD LDB3 LSH genome-wide association study hypertrophic cardiomyopathy human embryonic kidney 293T cells heart failure Illumina s Next Generation Sequencer system major histocompatibility complex genes haplotype-sharing test hydrogen oxide (water) hydrogen peroxide implantable cardioverter-defibrillator LIM domain binding 3 gene longest shared haplotype 18 APPENDIX 1

20 LVNC MD MiSeq MRI mrna NCBI NGS NHLBI OMIM PBS PCR PLEC PolyPhen PPCM left ventricular non-compaction cardiomyopathy muscular dystrophy Illumina s personal sequencer, the little sister of the HiSeq system in benchtop size, with faster workflow, allowing the assembly of small genomes or target regions magnetic resonance imaging messenger RNA National Center for Biotechnology Information next generation sequencing National Heart Lung and Blood Institue, a division of National Institutes of Health in the USA Online Mendelian Inheritance in Man a comprehensive database of human genes and genetic phenotypes authored and edited by the Johns Hopkins University phosphate buffered saline polymerase chain reaction plectin Polymorphism phenotype (pathogenicity prediction software for missense variants) peripartum cardiomyopathy RBM20 RNA binding motif protein 20 RCM ROS RNA RT SCD SIFT SNP restrictive cardiomyopathy reactive oxygen species ribonucleic acid reverse transcription sudden cardiac death sorting intolerant from tolerant (pathogenicity prediction software for missense variants) single nucleotide polymorphism SOD2 superoxide dismutase 2 TFC trna TTN VOUS VUS 1000G task force criteria (diagnostic criteria of ARVC) transfer RNA titin, the longest gene of the human genome variant of unknown significance variant of unknown significance 1000 Genomes catalog of human genetic variation APPENDIX 1 FREQUENTLY USED ABBREVIATIONS 19

21

22 CHAPTER 1 INTRODUCTION

23

24 Chapter 1: Introduction Novel clinical molecular diagnostic methods for congenital and inherited heart disease Jan DH Jongbloed, Anna Posafalvi, Wilhelmina S Kerstjens-Frederikse Richard J Sinke, J Peter van Tintelen Published in Expert Opinion on Medical Diagnostics, 2011

25 Importance of the field: For patients with inherited and congenital heart disorders, causative mutations are often not identified due to limitations of current screening techniques. Identifying the mutation is of major importance for genetic counseling of patients and families, facilitating the diagnosis in persons at-risk and directing clinical management. Next generation sequencing (NGS) provides unprecedented opportunities to maximize mutation yields and improve clinical management, genetic counseling and monitoring of patients. Areas covered in this review: We review recent NGS applications, focusing on methods relevant for molecular diagnostics in cardiogenetics. We discuss requirements for reliable implementation into clinical practice and challenges that clinicians, bioinfomaticians and molecular diagnosticians must deal with in analyzing resulting data. What the reader will gain: Readers will be introduced to recent developments, techniques and applications in NGS. They will learn about possibilities of using it in clinical diagnostics. They will become acquainted with difficulties and challenges in interpreting the data and considerations around communicating these issues to patients and the community. Take home message: Although several obstacles are to overcome and much still to learn, NGS will revolutionize clinical molecular diagnostics of inherited and congenital cardiac diseases, maximizing mutation yields and leading to optimized diagnostic and clinical care. Keywords: cardiogenetics, molecular clinical diagnostics, next-generation sequencing, targeted enrichment, exome sequencing, inherited and congenital heart disease Article highlights: 1. Novel clinical molecular diagnostic methods in cardiogenetic diagnostics are to be found in the field of Next generation sequencing (NGS) and novel applications that have recently become available with the launching of this technology will become part of daily diagnostic practice. 2. The main challenges of the implementation of NGS in daily diagnostic work are the assurance of good quality control and reliable data analysis and interpretation. 3. The most important consideration for clinical counseling will be the ascertainment of variants with uncertain clinical significance and the only reasonable way to deal with this problem is to pursue maximum data dissemination in the scientific community. 4. NGS provides unique solutions and will bring shorter reporting times, maximize mutation detection rates, and decrease costs if all the disease-related genes can be tested in parallel in a single experiment. 5. Despite the technological, bioinformatical and ethical problems, the use of NGS technology will lead to much improved and more effective diagnostic and preventive care for patients suffering from inherited and congenital heart disorders (CHD) and their relatives.

26 INTRODUCTION What started in the 1950s with observations of cardiac diseases segregating in families and suggesting heritable disease [1][2], has led in the last 15 years to the identification of many disease-associated genes and mutations. Advances in cardiogenetics have exceeded the level of being scientifically interesting phenomena and have major implications in genetic counseling and in directing clinical therapy [3][4][5]. Not only the expanding possibilities in DNA analyses, but also the increased awareness among cardiologists, pediatric cardiologists, and general practitioners of the potential heritability of cardiac disease has led to growing numbers of patients being referred to departments of genetics and/or cardiogenetic outpatient clinics for genetic counseling and DNA diagnostics. Diseases for which patients attend the cardiogenetics outpatient clinic are primary arrhythmia syndromes [4], cardiomyopathies [6] or familial congenital heart disorders (CHD) [7][8]. Examples of arrhythmia syndromes are the congenital long QT syndrome (LQTS), Brugada syndrome or cathecholaminergic-induced polymorphic ventricular tachycardia (CPVT), which are all associated with sudden cardiac death (SCD) at relatively young age. Most patients with a cardiomyopathy present with hypertrophic (HCM) or dilated (DCM) cardiomyopathy. Arrhythmogenic right ventricular cardiomyopathy (ARVC), restrictive (RCM) and left ventricular noncompaction cardiomyopathies (LVNC) are encountered less frequently. Cardiomyopathies frequently present with output-failure leading to fatigue, however, arrhythmias and SCD may occur. Finally, examples of CHD that may be heritable include either valvular abnormalities, such as bicuspid aortic valve/aortic valve stenosis (BAV/AVS) or pulmonary valve stenosis (PVS), septal defects (like atrial or ventricular septal defects; ASD/VSD), endocardial cushion defects (atrio-ventricular septal defect: AVSD), vascular abnormalities, such as coarctation of the aorta (CoA) or persistent ductus arteriosus (PDA), and more complex abnormalities like hypoplastic left heart syndrome (HLHS), tetralogy of Fallot (TOF), or heterotaxy-related cardiac abnormalities like transposition of the great arteries. Notably, genetics of CHD s becomes increasingly important, because due to the enormous development in surgical and cardiological care many of the 1 in 100 people born with a CHD survive to have offspring [9][10]. Interestingly, the boundaries between the different clinical entities are disappearing as overlapping clinical phenotypes are being recognized more frequently. For example, patients suffering from arrhythmia syndromes CHAPTER 1 NOVEL CLINICAL MOLECULAR DIAGNOSTIC METHODS 25

27 have been reported to also developing a cardiomyopathy [4]. In addition, patients diagnosed with inherited CHD s have been reported in which also a cardiomyopathy is identified [11][12]. As expected from this, both clinical and genetic heterogeneity is often being observed within these disorders (see also below). This review describes new developments in clinical molecular diagnostic methods for inherited and CHD s, with a focus on the novel applications that have recently become available with the launching of NGS (NGS) technologies. CURRENT CARDIOGENETIC DIAGNOSTICS In recent years, the relevance of genetic analyses in the genetic counseling and monitoring of patients and their family members having a cardiac disease with a proven, or at least suspected familial nature, has been increasingly recognized. Genetic analyses have therefore become an important part of the diagnostic activities to reach a clinical diagnosis in such patients. Since the first discovery of the MYH7 gene underlying HCM [13], a growing number of tests for heart-related disease have been introduced in DNA diagnostic laboratories worldwide, including array-cgh technology for the diagnostics of CHD s. This is exemplified by the fact that at Orphanet, the European database for rare diseases and orphan drugs [14], and or GeneTests [15] websites, the mutation analyses for the majority of known genes related to inherited cardiomyopathies, arrhythmia syndromes and congenital structural cardiac disorders are being offered in at least one of the European laboratories (see also Table 1). Important to note however is that in most inherited cardiac diseases, the genetic cause has not identified yet. For example, in at most 50% of ARVC, 25% of DCM, 60% of HCM, 25% of LVNC and ~10% of RCM patients the underlying disease gene was found. The large number of genes related to these different groups of inherited cardiac diseases underscores the fact that the genetic causes of these disorders show a high level of heterogeneity. Moreover, some of the genes have proved to be mutated in different cardiac diseases. This concept was first recognized within specific disease entities. As a result, the fact that both HCM and DCM can be caused by mutations in genes encoding components of the sarcomere, the contractile machinery of cardiomyocytes, has been known since about the year 2000 [17]. However, the boundaries between the different cardiac diseases are also fading, as there is no longer a strict separation between cardiomyopathies and channelopathies due to recent observations that mutations in ion-channel 26 INTRODUCTION

28 and related proteins can also play a role in the pathogenesis of DCM [18][19]. Moreover, a genetic overlap between cardiomyopathies/channelopathies and inherited structural cardiac disorders has also been suggested. For example, several mutations in sarcomeric proteins have been described that resulted in congenital heart malformations (Table 1) [8]. In addition, mutations in the cardiac T-box factor gene TBX20 were shown to result in cardiomyopathies in both mice and human, among other cardiovascular abnormalities [11]. The phenomenon discussed above is exemplified in Figure 1 by showing the genetic heterogeneity and overlap in genes that underlie different types of cardiomyopathies (DCM, HCM, LVNC, ARVC, and RCM), including a few genes that are also known to be involved in channelopathies (RYR2, SCN5A and PRKAG2) or CHD s (MYH7, MYBPC3). Finally, in addition to the significant heterogeneity of monogenic cardiac diseases, there is an emerging recognition that a significant proportion of patients carry two or more independent disease-causing gene mutations, which lead to more severe forms of clinical disease [20]. These might occur in the same gene (compound heterozygotes) or in different genes (bi- or multigenic). There may also be genetic modifiers present that are associated with a poorer prognosis. This concept and the fact that many genes might underlie a disease support the idea that large numbers of genes should be analyzed in parallel preferably within the same experiment in patients with inherited cardiac disorders to improve risk-assessment. Together, these observations imply that at least 110 genes are putative candidate disease genes in patients presenting at cardiogenetic outpatient clinics for a genetic diagnosis, since there are now ~60 cardiomyopathy [21], ~20 channelopathy [22], and ~30 CHD disease genes [8] known to be involved in the respective diseases (Table 1). Up to today, these genes have been analyzed at the nucleotide level on a gene-by-gene basis mainly. For this purpose, various pre-screening techniques like denaturing gradient gel electrophoresis (DGGE), denaturing high-performance liquid chromatography (dhplc), single strand conformation polymorphisms analysis (SSCP), conformation-sensitive capillary electrophoresis (CSCE), or high-resolution melting analysis (HRM) are generally being used to screen for aberrant PCR-amplified DNA sequences. The abnormal PCR fragments are then subsequently analyzed by Sanger sequencing to identify the exact nucleotide substitutions [23]. However, in a considerable number of genetic laboratories, the preferred screening approach is direct dideoxy sequencing of all exonic and adjacent intronic sequences of genes of interest without using pre-screening methods. CHAPTER 1 NOVEL CLINICAL MOLECULAR DIAGNOSTIC METHODS 27

29 Figure 1. Genetic heterogeneity and overlap in genes causing cardiomyopathies. Shown are genes underlying DCM, HCM, LVNC, ARVC, and RCM. Notably, some of these genes are also known to be involved in channelopathies and/or congenital heart malformation (based upon [16]). Genes also involved in congenital cardiac disease are indicated in bold. Genes also involved in channelopathies are underlined. The genes incorporated are: ABCC9 (ATP-sensitive potassium channel), ACTC1 (cardiac α-actin), ACTN2 (α-atinin-2), CALR3 (Calreticulin 3), CAV3 (caveolin 3), (CSRP-3 (muscle LIM protein), CRYAB (Alpha-B chrystallin) DES (desmin), DSG2 (desmoglein-2), DSC2 (desmocollin-2), DSP (desmoplakin), DTNA (dystobrevin), DMD (dystrophin), EMD (emerin), EYA4 (Eyes absent 4), GLA (α-galactosidase), ILK (Integrin-linked kinase), JPH2 (junctophilin) JUP (junctional plakoglobin), LAMA4 (laminin α4), LAMP2 (lysosome-associated membrane protein 2), LDB3 (cypher/ ZASP), LMNA (lamin A/C), mtdna (mitochondrial DNA), MYBPC3 (myosin-binding protein C), MYH6 (α-myosin heavy chain), MYH7 (β-myosin heavy chain), MYL2 (regulatory myosin light chain), MYL3 (essential myosin light chain), MYPN (myopalladin), NEXN (nexilin), PDLIM (PDZ and LIM domain protein 3), PKP2 (plakophilin-2), PLN (phospholamban), PSEN1 (Presenilin-1), PSEN2 (Presenilin-2), PRKAG2 (AMPK-γ2 subunit), RBM20 (RNA binding motif protein 20), RyR2 (ryanodine receptor 2), SCN5A (cardiac sodium channel), TAZ (Tafazzin), TCAP (titincap/telethonin), TGFb3 (transforming growth factor β3), TMPO (thymopoietin), TNNC1 (cardiac troponin C), TNNI3 (cardiac troponin I), TNNT2 (cardiac troponin T), TPM1 (α -tropomyosin), TTN (titin), VCL (metavinculin). 28 INTRODUCTION

30 If available for the respective genes, multiplex ligation-dependent probe amplification is used to screen for the deletion and/or duplication of one or more exons, as these are not identified using PCR-based techniques [24]. Also in cardiogenetics, examples have been found in arrhythmia syndromes and cardiomyopathies [25][26]. However, since using these approaches is laborious, relatively expensive and time-consuming, DNA diagnostics is often limited to a maximum of ~10 putative disease genes, as health insurance companies are not prepared to reimburse many more gene tests, if at all. It is therefore often difficult to decide which genes should be screened in a specific patient. In general, the genes being analyzed are those for which considerable mutation yields are reported in the literature. If a genotype-phenotype relationship has been identified, gene selection will of course be guided by the phenotypes identified in the respective patients and their affected family members. For example, in a patient presenting with DCM and conduction disease, the LMNA, DES and SCN5a genes are among the first genes to analyze, while patients presenting with an inherited arrhythmia syndrome should first be screened for genes encoding the respective ion channel proteins. As already mentioned above, in general, genetic testing in current cardiogenetic diagnostics is often limited to and guided by knowledge on the most common causative genes (for an overview of genes: see Table 1). The best possibilities to come to a genetic diagnosis in cardiomyopathies are in HCM, as mutations in the MYH7 and MYBPC3 genes account for ~80% of the cases in which a genetic cause is identified [21]. In HCM, genetic testing is therefore often started with these genes and, in addition, in the TNNT2 gene. When no mutation is identified in these 3 genes, the most logical option would be to analyze the other sarcomeric genes (TNNI3, TNNC1, ACTC1, TPM1, MYL2, MYL3 and TTN. The latter is very rarely screened since it is the largest human gene known). Other genetic analyses, like that of genes encoding Z-disk proteins, are often not performed because reported mutation yields are <1%, with the exception of the CSRP3 gene (1-5%) [5][21]. In DCM, up to 40 genes are known to cause disease, all with relatively low frequencies. Due to their higher frequencies, testing often starts with LMNA, MYH7 and TNNT2 [21]. For LVNC, choices in genetic testing are often comparable to those in DCM and HCM since most mutations are as yet found in sarcomeric genes. In ARVC, the group of candidate genes is relatively small, providing the opportunity to test them all. However, given the significantly higher yields reported for PKP2, mutation screening in that gene should be considered before continuing CHAPTER 1 NOVEL CLINICAL MOLECULAR DIAGNOSTIC METHODS 29

31 with other (desmosomal) genes, of which the DSC2 and DSG2 genes are the most logic next choice [27]. Whereas the success rate in genetic testing in the majority of cardiomyopathies can still be improved, that of channelopathies is high, with mutations identified in most of the patients. In particular in long QT syndrome, in over 80% of cases a genetic diagnosis is made. For cases of CHD, genetic diagnostics is currently mainly driven by phenotypic characteristics in the respective patient, because of the high diversity in non-syndromic CHD s. Moreover, these patients are often also screened by performing array-cgh analysis. Array-CGH is an assay in which DNA samples from patients and a healthy control are labeled with different fluorescent dyes and cohybridized to an array containing known DNA sequences. Differences in relative fluorescence intensities of hybridized DNA on the microarray then reflect differences in copy number between the genome of the patients and the healthy control. When, applying this method, a microdeletion or duplication is identified that is not known as a common copy number variant (CNV), this might represent the disease-causing genomic imbalance [28][29]. Mutation screening of the most promising candidate gene or genes in such a deletion or duplication in a cohort of CHD patients might result in identifying new disease genes. For example, mutation analysis in 402 patients of the topranking TAB2 candidate gene, which was one of the five genes in a critical 850 kb deleted region on 6q that was shared by 12 CHD patients resulted in finding two conserved missense mutations [30]. Finally, with respect to current testing regimes it is important to note that recently gene-chip re-sequencing technologies were implemented, providing the opportunity to analyze larger numbers of genes within one test. Although its use is limited yet since it is often only commercially available, this technology will be of importance in cardiogenetic diagnostics in the coming years (see also section 3.1). Even when using predictors like known mutation yield frequencies, phenotypes, or family history, in deciding on gene analyses, significant numbers of patients are left without a genetic diagnosis from current DNA diagnostic practice. Moreover, as also mentioned previously, a significant proportion of patients carries more than one mutation. Thus, other approaches are needed to maximize mutation yields and minimize investigation times. In the next section, highly promising possibilities that have recently become available to optimize cardiogenetic diagnostics will be presented and discussed. Notably, some of these techniques, such as re-sequencing arrays (CardioChips), have already been implemented but most are not yet being used in regular DNA diagnostics. 30 INTRODUCTION

32 Table 1. Tentative summary of genes* involved in inherited and congenital heart disease #. Genes mainly involved in cardiomyopathies cardiomyopathies arrhythmias Structural heart disease Skeletal muscle disease Remarks Gene-group Sarcomeric proteins CALR3 + DTNA + + MYBPC MYH MYH MYL2 + MYL3 + NEXN + TNNC1 ++ TNNI3 ++ TPM1 ++ Nuclear envelope EMD + ++ LMNA LAP2/TMPO ++ Cyto-architecture ACTC1 ++ ACTN2 ++ CRYAB ++ Cataract CSRP3/MLP ++ DES DMD + ++ FHL2 ++ FKRP + ++ FKTN + ++ ILK ++ LAMA4 ++ MYPN ++ PDLIM3/ALP ++ SCGD + ++ TCAP ++ + TTN VCL ++ ZASP(LDB3) ++ + Ion channels/ Calcium handling ABCC9 ++ PLN ++ SCN5A + ++ Desmosomal proteins DSC2 ++ DSG2 ++ DSP ++ syndromal JUP + Naxos disease PKP2 ++ Miscellaneous EYA4 + Hearing loss GLA + Storage disorder JPH2 + LAMP Storage disease PRKAG Storage disease mtdna PSEN1 and (possible) Alzheimer disease RBM20 + TAZ ++ Syndromal TGFB3 + TMEM43 ++ CHAPTER 1 NOVEL CLINICAL MOLECULAR DIAGNOSTIC METHODS 31

33 Genes mainly involved in arrhythmias arrhythmias Structural heart disease Skeletal muscle disease Remarks Gene-group Sodium channel related SCN1B ++ SCN4B ++ Epilepsy/seizures SCN5A + ++ SNTA1 ++ Potassium channel related AKAP9 ++ KCNA5 ++ KCNE1 ++ KCNE2 ++ KCNE3 ++ KCNH2 ++ KCNJ2 ++ syndromal KCNJ8 ++ KCNQ1 ++ Calcium metabolism CACNA1C ++ syndromal CASQ2 ++ RYR TRPM4 ++ Others CAV GJA5 ++ GPD1L ++ HCN4 ++ Genes mainly involved in structural heart disease cardiomyopathies cardiomyopathies arrhythmias Structural heart disease Skeletal muscle disease Remarks Gene-group Transcription factors ANKRD1 ++ CITED2 ++ FOXH1 ++ GATA4 ++ GATA6 ++ NKX NKX TBX1 ++ TBX TBX ZIC3 ++ Heterotaxy ZFPM2 ++ Ligands/receptors AVCR1 ++ Heterotaxy ALK2 ++ Syndromal CFC1 ++ Heterotaxy GDF1 ++ JAG1 ++ Syndromal LEFTY2 ++ Heterotaxy NODAL ++ Heterotaxy NOTCH1 ++ TDGF1 ++ Sarcomeric proteins ACTC MYH11 ++ miscellaneous BRAF + ++ syndromal 32 INTRODUCTION

34 CRELD1 ++ ELN ++ KRAS ++ syndromal MAP2K1 and 2 ++ syndromal MED13L ++ NRAS ++ syndromal PTPN11 ++ syndromal RAF syndromal SHOC2 ++ syndromal SOS1 ++ syndromal TLL1 ++ *The gene names correspond to the following proteins (in alphabetical order): ABCC9, ATP-binding cassette, subfamily C, member 9; ACTC1, alpha actin; ACTN2, actinin, alpha 2; AVCR1, activin A receptor, type I; AKAP9, A kinase (PRKA) anchor protein (yotiao) 9; ANKRD1, ankyrin repeat domain 1 (cardiac muscle); BRAF, v-raf murine sarcoma viral oncogene homolog B1; CACNA1c, calcium channel, voltage-dependent, L type, alpha 1C subunit; CALR3, calreticulin 3; CASQ2, calsequestrin 2 (cardiac); CAV3, caveolin 3; CFC1, cripto, FRL-1, cryptic family 1; CITED2, Cbp/p300-interacting transactivator, with Glu/Asp-rich carboxy-terminal domain, 2; CRELD1, cysteine-rich with EGF-like domains 1; CRYAB, chrystallin, alpha-b; CSRP3/MLP, cysteine- and glycine-rich protein 3 / cardiac LIM protein; DES, desmin; DMD, dystrophin; DSC2, desmocollin 2; DSG2, desmoglein 2; DSP, desmoplakin; DTNA, dystrobrevin, alpha; ELN, elastin; EMD, emerin; EYA4, eyes absent 4; FHL2, four-anda-half LIM domains 2; FKRP, fukutin-related protein; FKTN, fukutin; FOXH1, forkhead box H1; GATA4, GATA binding protein 4; GATA6, GATA binding protein 6; GDF1, growth differentiation factor 1; GJA5, gap junction protein, alpha 5, 40kDa; GLA, galactosidase, alpha; GPD1L, glycerol-3-phosphate dehydrogenase 1-like; HCN4, hyperpolarization activated cyclic nucleotide-gated potassium channel 4; ILK, integrin-linked kinase; JAG1, jagged 1; JPH2, junctophilin 2; JUP, junction plakoglobin; KCNA5, potassium voltage-gated channel, shakerrelated subfamily, member 5; KCNE1, potassium voltage-gated channel, Isk-related family, member 1; KCNE2, potassium voltage-gated channel, Isk-related family, member 2; KCNE3, potassium voltage-gated channel, Iskrelated family, member 3; KCNH2, potassium voltage-gated channel, subfamily H (eag-related), member 2; KCNJ2, potassium inwardly-rectifying channel, subfamily J, member 2; KCNJ8, potassium inwardly-rectifying channel, subfamily J, member 8; KCNQ1, potassium voltage-gated channel, KQT-like subfamily, member 1; KRAS, v-ki-ras2 Kirsten rat sarcoma viral oncogene homolog; LAMA4, laminin alpha-4; LAMP2, lysosomeassociated membrane protein 2; LAP2/TMPO, lamina-associated polypeptide 2 / thymopoietin; LEFTY2, leftright determination factor 2; LMNA, lamin A/C; MAP2K1 and 2, mitogen-activated protein kinase kinase 1 and 2; MED13L, mediator complex subunit 13-like; mtdna, mitochondial DNA; MYBPC3, myosin-binding protein C, cardiac; MYH11, myosin, heavy chain 11, smooth muscle; MYH6, myosin, heavy chain 6, cardiac muscle, alpha; MYH7, myosin, heavy chain 7, cardiac muscle, beta; MYL2, myosin, light chain 2, regulatory, cardiac, slow; MYL3, myosin, light chain 3, alkali, ventricular, skeletal, slow; MYPN, myopalladin; NEXN, nexilin (F actin binding protein); NKX2.5, NK2 transcription factor related, locus 5 (Drosophila); NKX2.6, NK2 transcription factor related, locus 6 (Drosophila); NODAL, nodal homolog (mouse); NOTCH1, notch 1; NRAS, neuroblastoma RAS viral (v-ras) oncogene homolog; PDLIM3/ALP, PDZ and LIM domain protein 3; PKP2, plakophilin 2; PLN, Phospholamban; PRKAG2, protein kinase, AMP-activated, gamma 2 non-catalytic subunit; PSEN 1 and 2, presenilin-1 and -2; PTPN11, protein tyrosine phosphatase, non-receptor type 11; RAF1, v-raf-1 murine leukemia viral oncogene homolog 1; RBM20, RNA binding motif protein 20; RyR2, ryanodine receptor 2 (cardiac); SCGD, delta sarcoglycan; SCN1B, sodium channel, voltage-gated, type I, beta; SCN4B, sodium channel, voltage-gated, type IV, beta; SCN5A, sodium channel, voltage-gated, type V, alpha subunit; SHOC2, soc-2 suppressor of clear homolog (C. elegans); SNTA1, syntrophin, alpha1 (dystrophin-associated protein A1, 59 kda, acidic component); SOS1, son of sevenless homolog 1 (Drosophila); TAZ, tafazzin; TBX1, T-box 1; TBX20, T-box 20; TBX5, T-box 5; TCAP, titincap (telethonin); TDGF1, teratocarcinoma-derived growth factor 1; TGFB3, transforming growth factor, beta 3; TLL1, tolloid-like 1; TMEM43, transmembrane protein 43; TNNC1, troponin C (type 1: slow); TNNI3, troponin I type 3 (cardiac); TNNT2, troponin T type 2 (cardiac); TPM1, tropomyosin 1 (alpha); TRPM4, transient receptor potential cation channel, subfamily M, member 4; TTN, titin; VCL, vinculin; ZASP(LDB3), Z-band alternatively spliced PDZ motif-containing protein; ZFPM2, zinc finger protein, multitype 2; ZIC3, Zic family member 3 (oddpaired homolog, Drosophila). # Indicated are the diseases in which a particular gene is involved: ++: generally accepted to be involved in; +: incidentally found to be involved in. Genes indicated in bold and in italics are not offered in at least one international laboratory by searching the Orphanet and GeneTests websites. CHAPTER 1 NOVEL CLINICAL MOLECULAR DIAGNOSTIC METHODS 33

35 FUTURE DIAGNOSTIC APPROACHES AND FIRST APPLICATIONS Since the early 1990s, enormous progress has been made in identifying the genetic causes of inherited or CHD s. So far, the hunt for causal genes has been performed using linkage and association techniques or array-cgh analysis and the subsequent mutational analysis of genes in the candidate region(s). These methods resulted in the identification of causal genes encoding proteins that are parts of various cellular structures or pathways. The discovery that these structures or pathways are also involved in these diseases led to candidate gene approaches to screen genes encoding other components of these structures or pathways [31]. However, to date, only a small proportion of allelic variants underlying disease have been discovered. For example, as mentioned previously, for cardiomyopathies between 40-80% of patients/families are still without a genetic diagnosis. This is because these investigations are hampered by factors such as having only a relatively small number of affected individuals within families to perform a linkagebased approach. In addition, it is not feasible to screen the genes encoding proteins that are part of the cellular structures or pathways on a gene-by-gene basis, since these contain hundreds of proteins for which the encoding genes need to be tested. Furthermore, other proteins that are not involved in these structures or pathways may also play a role in disease development. In order to maximize genetic testing for patients with inherited cardiac disorders, we therefore need approaches that enable mutational screening of cardiac disease genes in one experiment and on a large scale. The novel genomic techniques and some adaptations that will permit this are discussed below. 1. Cardiochips and arrays In recent years, several platforms were launched to facilitate parallel processing of larger numbers of genes. Low-density DNA hybridization assays were already being used in the early 2000s to identify known mutations (including small deletions/insertions) in HCM [32]. Customized resequencing assays were developed to identify mutations in cardiomyopathy genes, exploiting the Affymetrix gene-chip re-sequencing array technology. Waldmüller et al. [33] reported the use of array-based re-sequencing for testing the three most commonly affected genes in HCM (MYH7, MYBPC3 and TNNT2), while Foksteun et al. [34] demonstrated the use of a DNA re-sequencing array for detecting mutations in 16 HCM genes. In addition, Zimmerman et al. [35] 34 INTRODUCTION

36 demonstrated the efficient analysis of 19 genes implicated in DCM using a CardioChip. Nowadays, these cardiochips are being used in a significant number of molecular diagnostic laboratories and until the applicability of NGS technology for molecular diagnostics has convincingly been proven, resequencing array methods provide the best approach to analyze multiple genes within one test. For example, Partners Healthcare, in cooperation with Harvard Medical School, offers genetic testing applying their DCM CardioChip TM TEST, the design of which is based on the CardioChip described by Zimmerman et al. [35]. However, this type of technology is still not very widespread in daily diagnostic practice and NGS applications that are currently being developed will, most likely, replace such array techniques in the near future. CHAPTER 1 2. Next generation sequencing NGS techniques have recently become available that provide the opportunity to identify every unique variant in an individual genome via whole-genome re-sequencing [36][37]. The molecular basis of each type of technology is a DNA library preparation (including shearing the DNA, adapter ligation, and gel purification of DNA fragments of the desired size), the amplification of the resulting single strands and performing sequencing reactions on the amplified strands. Using reaction chambers that contain huge amounts of such oligonucleotides, a large number of these arbitrary nucleotide strands can be analyzed in parallel in a single run. As a result, the nucleotide sequences (the so called reads ) of millions of different DNA fragments can be determined within a relatively short time. Depending on the question to be answered, subsequent bioinformatic analyses can translate these nucleotide sequences into useful information, e.g. reports on variants/ mutations found in genes of interest, de novo assembly of genomic regions (up to a full genome), or copy number variations in parts of the genome or in the full genome. There are various companies offering machines and solutions that use this highly promising technique (for extensive overviews see: [38][39]). Although they are still being too expensive to be introduced at a diagnostic level, a personalized genome can now be produced. This was recently shown by Ashley and co-workers [40], who reported on the full genome of a patient with a family history of vascular disease and early SCD. However, in most cases targeted approaches will need to be applied to identify disease-causing mutations. NOVEL CLINICAL MOLECULAR DIAGNOSTIC METHODS 35

37 3. Disease-specific targeted enrichment and re-sequencing The human genome contains about putative genes [41]. Logically, not all of these are associated with a certain disease and to apply NGS without having to sequence a full genome, we need methods for targeted enrichment of DNA fragments encoding the known or suspected disease genes. As also mentioned above, at least ~110 genes have been implicated in monogenic cardiac disorders and mutation analysis on a gene-by-gene basis is not feasible. Several methods that may help to enrich these genes have become available in the last years, basically making use of either hybridization or PCR-based capturing. Hybridization-based enrichment generally utilize probes complimentary to the sequences of interest that are either presented in a solid phase, such as oligonucleotide microarrays, or in a solution phase, applying molecular inversion probe (MIP)-based or biotynylated RNA-based approaches [42][43]. To enrich for sequences of interest, the total DNA is applied to the probes and the desired fragments hybridize. The non-targeted fragments are subsequently washed away, and the enriched DNA eluted for resequencing. A recent proof-of-principle study convincingly demonstrated the applicability of solid-phase enrichment and subsequent NGS in a diagnostic context, using autosomal recessive ataxia as a prototypical heterogeneous monogenic disorder [44]. In this study, the complete genomic sequence (coding and noncoding regions) of seven genes known to cause autosomal recessive ataxia were presented on a NimbleGen sequence capture array. By hybridizing diagnostic samples onto this array, these were enriched for DNA fragments encoding parts of the seven genes. Subsequent re-sequencing using Roche 454 Titanium shotgun sequencing was used to determine the sensitivity and specificity of NGS of enriched samples for the identification of pathogenic mutations. The enrichment showed high specificity: 80% of the sequences obtained were on target, which means that these could be mapped back to the targeted gene regions. In addition, also high sensitivity was demonstrated: pathogenic mutations for 6/7 studied mutant alleles and more than 99% of known SNP variants were identified. Mutation and SNP detection accuracy was shown to be limited by sequence coverage and misalignment rather than sequencing errors [44]. Methods have also been developed that enable the specific PCR-driven amplification of DNA fragments of interest. Of these, the microdroplet-based PCR enrichment technique of RainDance technologies has been shown to be effective in the simultaneous amplification of almost 4,000 products [45]. In addition to this commercially available 36 INTRODUCTION

38 technology, several laboratories have developed their own approaches. For example, in a case study long range PCR enrichment was used to amplify 16 HCM genes for subsequent NGS [46]. For this purpose, primers and reactions were used that PCR-amplified DNA fragments of ~5100 nucleotides with overlaps averaging 550 nucleotides and together encompass the genes in full in 14/16 genes. Resulting PCR fragments were gel purified, an equimolar pool of fragments generated, and Roche 454 and Illumina DNA libraries were prepared. Subsequent sequencing on the respective machines showed that 95% and 90%, respectively, of the sequencing reads were on target, but with a pattern of variable coverage. The latter emphasizes the need to have sufficient sequencing depth (see also sections 4 and 6). Variants identified could be confirmed by Sanger sequencing [46]. When using selected enrichment it is important to realize that for the efficient use of NGS machine capacity, the parallel sequencing of multiple patient samples in a single run is preferred. In such cases, barcoding the patient-specific samples prior to sequencing will aid in distinguishing the different patient sample data after their joint sequencing run. Barcoding is the simple technique of adding a unique nucleotide sequence to the adapter sequences that are ligated to DNA fragments during the patients library preparation [47][48]. CHAPTER 1 4. Exome sequencing Although selectively enriching a panel of genes will lead to the identification of the disease-causing gene or genes in significantly more cases, the fact that this panel is still a selection of genes encoded from the genome implies that the causal gene may still not be identified in each individual patient using this approach. A more comprehensive alternative would therefore be to enrich an individual s DNA for all the protein-encoding regions ( the exome ) of the genome - the exome encompasses ~1% of the whole genome - and then perform NGS (exome sequencing; [42]). The method is the same as for targeted re-sequencing (see section 3.3), however instead of using probes complimentary to the coding sequences of a subset of genes, all known coding DNA fragments are presented as probes. Moreover, in contrast to sequencing a full genome, this approach is currently feasible for patients seeking a genetic diagnosis at cardiogenetic outpatient clinics. Exome sequencing was recently shown to be a powerful tool for identifying candidate genes in a proof-of-concept experiment by Ng et al. that used four unrelated, affected individuals with the rare, autosomally dominant Freeman- NOVEL CLINICAL MOLECULAR DIAGNOSTIC METHODS 37

39 Sheldon syndrome, which is known to be caused by mutations in the MYH3 gene [49]. Exome sequencing of these four patients and subsequent data analyses indeed led to the identification of causative mutations in MYH3. To evaluate the effectiveness of the exome sequencing method, Ng et al. then applied the same approach to find the gene responsible for Miller syndrome, a rare disorder characterized by facial dysmorphia and abnormalities of the extremities. As a result, a single candidate gene was identified. The subsequent screening of this gene by conventional Sanger sequencing in other, unrelated kindreds led to the identification of additional disease-causing mutations [50]. Other recent reports have demonstrated the successful application of exome sequencing in making a genetic diagnosis for various disorders [51] [52][53]. Together, these studies show that exome sequencing is a powerful tool for identifying the causative genes in monogenic disorders. Thus, exome sequencing, rather than custom-designed enrichment techniques, might soon be the method of choice for DNA diagnostic purposes. 5. Combined approaches The above examples demonstrate the potential power of exome sequencing. However, these cases describe the hunt for the causal disease gene in patients and families with very rare syndromes for which a thorough phenotypic classification was possible and had been performed. More importantly, finding the gene was simplified by the fact that the disease showed recessive inheritance was caused by mutations in the same gene or by de novo mutations (although not certain for every disease at the time the analysis was started). Hunting disease genes in disorders that are known to be genetically very heterogeneous, like the cardiac disorders for which the development of dedicated diagnostics is the subject of this review, will probably be more challenging and more sophisticated bioinformatic analysis techniques might be needed (see the section on Challenges of future diagnostics ). Therefore, methods to narrow down the genomic regions of interests in specific patient cohorts or families might support the identification of causal genes in these disorders. For example, in families showing an X-linked inheritance pattern, their mutation might be identified by applying X-exome capture and sequencing, as demonstrated for terminal osseous dysplasia by Sun et al. [54]. When array-cgh analysis in a patient with a certain type of CHD resulted in the identification of a small genomic deletion or duplication, targeted re-sequencing of the, in general, limited number of genes in this region in 38 INTRODUCTION

40 a cohort of patients having the same disease, could be performed. Finding mutations in one of these genes would then confirm the role of this candidate gene in CHD s. The availability of linkage data or association peaks will provide the possibility of focusing on that specific part of the exome that originates from those regions, instead of analyzing the complete exome. As an example, exome sequencing in conjunction with homozygosity mapping led to the rapid identification of the causative allele for non-syndromic hearing loss in a consanguineous Palestinian family [55]. No such example has been reported for a monogenic cardiac disorder yet. Interestingly, however, using a haplotype sharing test, we were recently able to identify the causal MYH7 and PKP2 mutations in the shared regions of a single DCM and multiple ARVC families, respectively [56]. Combining this haplotype approach with exome sequencing and data analysis of genes in regions identified in such families that do not encode already known cardiac disease genes will most likely lead to the identification of the causal allele. As exemplified in the Insulin Resistance Atherosclerosis Family Study (IRASFS), exome sequencing has also been useful in finding rare variants that may be a common explanation for linkage peaks observed in complex trait genetics [57]. It is important to note, however, that this could only be achieved because only a few families in the sample contributed significantly to a linkage signal and these families all carried the same rare variant. Thus, exploiting such a combined approach will often be limited to families with sufficient affected individuals to enable haplotype sharing analyses or the application of other linkage techniques. Nevertheless, exome sequencing of larger groups of likely unrelated patients and subsequent data analysis and comparison will undoubtedly result in the identification of causative genes that are shared by two or more of these patients. Success will either be based on the presence of founder mutations in specific populations and, as a result, the presence of two or more patients within a cohort carrying the same gene mutation, or on the increasing chance of encountering two or more patients who carry different mutations, but lying in the same gene, when more exomes of patients with the same disease are sequenced. CHAPTER 1 6. Other applications In addition to DNA re-sequencing, other NGS applications will become available for clinical diagnostic purposes. Four of these are described below. (1) Coverage information (the number of reads covering a specific DNA sequence) of DNA re-sequencing runs can be used to identify copy number NOVEL CLINICAL MOLECULAR DIAGNOSTIC METHODS 39

41 variations. Higher or lower coverage numbers of successive DNA sequence reads indicate duplications or deletions, respectively, of chromosomal regions, isolated genes or smaller parts thereof. This information is not often used at the moment, although it is automatically incorporated in the results of ordinary DNA re-sequencing [58][59]. (2) The NGS technique can support studies on biological interactions between DNA and proteins, like transcription factors, chromatin, or other DNA-binding proteins. By using chromatin immunoprecipitation (ChIP) and subsequent NGS of DNA fragments ChIP derived sequences can be determined (ChIP-seq). ChIP-seq entails a series of steps: i, chemical cross-linking of DNA and associated proteins; ii, isolation and lysis of nuclei and subsequent DNA fragmentation; iii, the use of an antibody against the DNA-binding protein of interest, to specifically immunoprecipitate the associated protein:dna complex; iv, reverse the chemical crosslink and isolate the DNA; v, sequence the resulting DNA fragments applying NGS.. Although a direct role of this technique in genetic diagnostics might as yet not be envisaged, this method will, for example, be important in identifying new candidate disease genes or in finding genes that are regulated by known disease genes and that may form targets for disease treatment. For example, this approach was used with the enhancer-associated protein p300 from mouse heart tissue (embryonic day 11.5) to identify over 3,000 candidate heart enhancers genome-wide [60]. (3) Instead of sequencing parts of the genome, the direct sequencing of mrna molecules on a large scale can be performed using NGS platforms (RNA-seq; [61]). Since these molecules represent the nucleotide sequences that are transcribed into proteins, the probability that mutations identified at this level (the transcriptome) are truly expressed is higher than those identified at the DNA level. Moreover, using this approach, sequencing of tissue-specific RNA molecules can be performed, for example, enabling the identification of mutations specifically expressed in the heart or even in pre-determined cardiac cell types. Likewise, the nucleotide sequences of non-coding RNAs and/or micrornas can be determined. (4) Comparable to copy number determination using coverage statistics of DNA re-sequencing results, figures on RNA expression levels can be discovered by using RNA sequencing data [62][63]. However, to identify mutations as well as to determine RNA expression levels in patients suffering from cardiac disease, myocardial sampling would be required. As this would require the use of invasive interventions, it is unlikely to become a regular application in standardized diagnostic work. 40 INTRODUCTION

42 CHALLENGES OF FUTURE DIAGNOSTICS Several NGS applications are now available to broaden the DNA diagnostic possibilities in cardiogenetics. These will certainly lead to maximized identification of the disease-causing mutations. However, the challenge in applying NGS is not so much producing the data, but its subsequent quality control, analysis and interpretation. In NGS experiments large amounts of data (up to gigabases of nucleotide sequences in a single run) are being produced. Therefore, where data quality is concerned, it is of the utmost importance for diagnostic purposes to have absolute confidence that every exon of interest, together with the flanking intronic sequence containing consensus splice site sequences, is being analyzed, and that fully reliable data is being produced. This probably implies that more stringent quality control criteria will be needed to fulfill clinical diagnostic requirements, than those needed for purely research projects. Logically, although of utmost importance, this does not apply to the proper distinction between true- and false-positives, since these are inherent in both research and diagnostic applications. However, particularly when hybridization-based capturing approaches are being applied, the minimum exon coverage needed to obtain nearly complete certainty that a heterozygous mutation will be detected has to be carefully established. This is in particular challenging when GC-rich regions are concerned, as was demonstrated when the performance of NimbleGen 385K custom arrays for the re-sequencing of 22 genes most of which associated with hereditary colorectal cancer was evaluated [64]. Since this might be of importance in certain disease entities, the minimum coverage has to be even more carefully determined if a mosaic situation is suspected and mutations or variants might be present at percentages <50%, as is expected at heterozygosity. In addition to taking care to obtain enough coverage of every DNA sequence of interest, there has to be absolute certainty that deletion/insertion mutations will be identified, as the difficulty in tracing these mutations is intrinsic to their characteristics. When custom-designed enrichment techniques are applied, the pooling of patient samples is needed to ensure the efficient use of sequencing flow cells. This implies the use of patient-specific barcoding and thus comprehensive monitoring of the patients material trace. In addition to this, efficient processing of larger numbers of patient material, as is daily practice in diagnostics laboratories, will benefit from automated library preparation (as an example see: [65]). Moving these laboratory procedures to a robotic workstation will be an important next step CHAPTER 1 NOVEL CLINICAL MOLECULAR DIAGNOSTIC METHODS 41

43 in getting NGS into clinical molecular diagnostics. Finally, since the careful archiving of patient-related experimental results over longer periods of time are a prerequisite for good quality diagnostic care, methods to handle the storage of the huge datasets produced by NGS, as well as the minimal requirements for their storage must be discussed. Undoubtedly, these quality control issues will be solved satisfactorily and reliable results from NGS analyses can then be communicated to the respective patients. More challenging, however, will be the interpretation of these results. Even when the number of analyzed genes is limited because of the use of targeted capturing or amplification, large numbers of variants will be identified for which putative pathogenicity has to be determined. When these variants concern nonsense or frameshift mutations, the origin of the affected sequences will need to be verified. Do these originate from pseudo-genes or pseudo-exons and can the respective variants therefore be discarded, or are these true truncating mutations? If the latter is the case, the disease-causing mutation might have been found. However, the largest number of variants identified will be missense mutations; the substitution of only one amino acid residue in the protein sequence concerned. The first step in analyzing the list of variants will be to use the correct methods and tools to reliably separate the possibly disease-related from benign variants. The view now emerging in the field is that the most important reference database needed to perform such analysis is the one compiling all the results of NGS and/or Sanger sequencing experiments already performed. This might be an in-house collection (probably the preferred starting point), but ideally will be a databases with sequencing data from several laboratories performing this type of work. In addition, NCBI s SNP database might be used as a reference database, although this also incorporates variants for which a pathogenic nature is not excluded. Moreover, NCBI s SNP database might contain variants that are harmless for carriers in the heterozygous state, but disease-causing in a homozygous or compound heterozygous carrier. Taken together, the big challenge in identifying disease causing mutations from rare, but benign variant is to get to know the frequencies in which these rare variants are present in patient cohorts compared to their presence in the general population. Re-sequencing studies of disease genes have typically not subjected control populations to the same level as patient cohorts and therefore these rare variants go undetected. Therefore, to value variants identified in re-sequencing studies it is of major importance to know how many new variants can be expected when a new set 42 INTRODUCTION

44 of individuals of a given size is being sequenced. Interesting in this respect is that a recent study calculated that 350 individuals have to be sequenced to find all common variants (frequency at least 1%), whereas >3,000 individuals have to be analyzed to identify all variants with a frequency of at least 0.1% [66]. This underscores the importance of compiling open source databases containing large datasets of variants identified in re-sequencing experiments. After omitting all the variants identified more often in other sequenced individuals, the putative pathogenicity of the remaining variants needs to be determined. Several software tools can be applied, of which those that calculate the level of conservation of the affected nucleotide (i.e. GERP; [67]) and/or amino acid residue (i.e. phylop; [68]), so far seem to provide the most important discriminating factor [53][69]. In addition, aspects like differences in the physico-chemical properties of the amino acids involved, the presence of the affected amino acid in a known functional domain, the known involvement of the affected gene in a comparable disease or other diseases in general, and/ or the expression of the respective gene in the tissue or tissues of interest. Prediction programs like SIFT, Polyphen or MutPred combine knowledge on aspects useful in predicting the putative pathogenicity of variants [70][71]. Such programs might be incorporated into the pipelines that are being complied for analyzing NGS data. Several data analysis pipelines have recently been published (i.e. [72][73][74]). Together, this information will result in a ranking of variants, in which the highest ranked variants will represent the most likely disease-causing ones. Next, additional experiments will have to be performed to verify the pathogenicity of the variants. First, carriership in the patient has to be confirmed by Sanger sequencing. Second, if needed, the absence of the variant(s) in a large number of healthy controls has to be verified. Third, when possible and appropriate, co-segregation analysis within the relevant family has to be performed. And finally, certainly where exome sequencing approaches are concerned, the presence of the variants in other patients suffering from the same disease should be determined. Ideally, these combined approaches should facilitate the identification of either the main suspect or of a few suspects. However, in a lot of cases this will most likely not be the situation and further analyses, e.g. at the functional level, will be needed to identify the causal gene conclusively. It is also possible that several of the ranked variants may contribute to disease development, since this concept has now been well-established by studies in a considerable proportion of patients suffering from inherited cardiac diseases [18]. CHAPTER 1 NOVEL CLINICAL MOLECULAR DIAGNOSTIC METHODS 43

45 CONCLUSION Recent developments in genome-wide screening techniques have created exciting possibilities for taking genetic diagnostics and research to a higher level. The availability SNP arrays is enabling not only the hunt for associations in larger groups of patients with multifactorial or polygenic diseases, but also the identification of disease-causing genomic regions in affected families. More importantly, using NGS techniques, the content of an individual s complete genome, or larger parts of the genome, can now be determined at the single nucleotide level, certainly with capturing techniques that allow one to zoom in on all protein-encoding DNA fragments (the exome) or specific subsets of the exome. Applying NGS technology will greatly enhance the possibilities to identify new disease genes and should provide a unique way to reduce screening times and maximize mutation detection rates in clinical molecular diagnostics. But perhaps even more importantly, it may help decrease the costs of genetic testing per individual if all the relevant disease genes can be tested in parallel. Since a large number of putative disease genes may underlie disease in the field of cardiogenetics and multiple genes might contribute to disease development, the exploitation of NGS techniques will provide the field with the optimal diagnostic genetic toolbox now available. As discussed here and in the next section, it is important to realize that although NGS is a highly promising techniques, the results must be treated with great care and there is still much to learn. EXPERT OPINION Current clinical care and molecular diagnostics of inherited and CHD s suffers from laborious, time-consuming, costly procedures and the limited possibilities to screen all the known genes involved in the respective disease. This results in incomplete, expensive diagnostic work and long reporting times. NGS technology provides unique solutions and will bring shorter reporting times, maximize mutation detection rates, and decrease costs if all the disease-related genes can be tested in parallel. Although the re-sequencing of a whole genome could technically already be applied, this is not yet financially feasible. However, we expect is the genetic diagnostics of cardiac diseases to gradually grow to a point at which a personalized cardio-related genome is produced for every patient visiting a cardiogenetics outpatient clinic. Laboratories involved in the diagnostics 44 INTRODUCTION

46 of cardiogenetics will probably first focus on developing procedures for targeted enrichment of cardiac disease genes. Depending on the design, these procedures will facilitate the parallel analysis of a minimum of ten (in ARVD/C ~10 causative genes are now known) and up to several hundred genes. Notably, cost analysis of re-sequencing many of the common genes has shown that this is much more cost effective than other current methods [27][28]. However, since the mutation detection rate will still be limited when targeted capturing is applied, we expect the genetic diagnostics of cardiac disease to quickly move towards the exome sequencing of patient samples. As soon as exome sequencing and its data analysis and interpretation becomes daily practice and the $1000 genome comes within reach [75][76], the re-sequencing of a complete genome will enter the field of clinical diagnostics. However, there are ethical and practical considerations around re-sequencing an entire exome or genome that should not be ignored (see also: [77]) and these are elaborated below. Potentially, although such investigations will be implemented to test patients referred for a particular disease (in our case, a cardiac disorder), exome or whole genome sequencing will also reveal mutations related to completely different diseases, or variants of unknown significance that might cause unnecessary anxiety. Thus, methods to either mask results or filter only those results relevant to the diagnostic request should be considered and the analysis of data outside the known disease genes should only be performed after informed consent given by the patient. Of course, we could choose to simply discard the irrelevant data, but the danger is that variants that appear unimportant today may be shown to be disease-related in the future. Together, these new technical advances demand that patients be comprehensively counseled. On the practical side, the most important consideration concerns the ascertainment of variants with uncertain clinical significance [78]. Undoubtedly, the only reasonable way to deal with this problem is to pursue maximum data dissemination in the scientific community. This will require the construction of databases for all or much of the data from exome and whole genome sequencing projects, like the recent 1000 genomes project [77][78]. This would serve as a reference database for benign variants, although the difficulty remains of how to decide whether a variant is benign or diseasecausing. This means we also need databases to compile mutations and variants that are identified in patients and can be related to disease. These should preferably be disease-specific databases, rather than locus-specific databases, as the information should not be limited to information on the specific CHAPTER 1 NOVEL CLINICAL MOLECULAR DIAGNOSTIC METHODS 45

47 variant. In addition, information about the context in which the variant was identified should be documented, e.g. the phenotype/characteristics of the patient carrying the variant, the co-existence of variants in the same or other disease genes, the co-segregation of the variant with disease in the patient s family, phenotypic details on other affected family members, the carriership frequency of the variant in a larger patient cohort, etc. Early initiatives in this direction have been reported in the last decade in the field of cardiogenetics, e.g. the database on genetic mutations in inherited arrhythmias [79]; the human intermediate filament database [80][81]; and the ARVD/C database [82][83]. In addition, data analysis pipelines should be developed and/or improved to combine all the available data on a specific variant, including the results of in silico prediction programs, to support the deciphering of a variant s clinical significance. In order to accomplish this, clinical and molecular geneticists and bioinformaticians should collaborate closely to reach this major goal. In conclusion, although there are still several hurdles to be taken before NGS can be implemented in the clinical molecular diagnostics of genetic cardiac diseases, this new technology will soon be making an impact on molecular cardiogenetics. It is therefore necessary that molecular diagnosticians, clinical geneticists and genetic counselors, cardiologists and pediatric cardiologists and other physicians involved in the diagnosis and treatment of inherited cardiac diseases should start learning about NGS and get comfortable with this new technique. Finally, despite the technological, bioinformatical and ethical problems discussed here, the use of NGS technology will certainly lead to much improved and more effective diagnostic and preventive care for patients suffering from inherited and CHD and their relatives. ACKNOWLEDGEMENTS We thank Jackie Senior for editorial assistance to the authors during the preparation of this manuscript. REFERENCES 1. Gaunt RT, Lecutier MA. Familial cardiomegaly. Br Heart J 1956;18: Paley DH, Familial cardiac arrhythmia. Trans Am Coll Cardiol 1952;2: Schwartz PJ. The congenital long QT syndromes from genotype to phenotype: clinical implications. J Intern Med 2006;259: Lehnart SE, Ackerman MJ, Benson DW Jr et al. Inherited arrhythmias: a National Heart, Lung, and Blood Institute and Offi ce of Rare Diseases workshop consensus report about the diagnosis, phenotyping, molecular mechanisms, and therapeutic approaches for primary cardiomyopathies of gene mutations affecting ion channel function. Circulation 2007;116: INTRODUCTION

48 5. Bos JM, Towbin JA, Ackerman MJ. Diagnostic, prognostic, and therapeutic implications of genetic testing for hypertrophic cardiomyopathy. J Am Coll Cardiol 2009;54: Ashrafian H, Watkins H. Reviews of translational medicine and genomics in cardiovascular disease: new disease taxonomy and therapeutic implications cardiomyopathies: therapeutics based on molecular phenotype. J Am Coll Cardiol 2007;49: Pierpont ME, Basson CT, Benson DW Jr et al. Genetic basis for congenital heart defects: current knowledge: a scientific statement from the American Heart Association Congenital Cardiac Defects Committee, Council on Cardiovascular Disease in the Young: endorsed by the American Academy of Pediatrics. Circulation 2007;115: Wessels M, Willems P. Genetic factors in non-syndromic congenital heart malformations. Clin Genet 2010;78: Drenthen W, Boersma E, Balci A et al. Predictors of pregnancy complications in women with congenital heart disease. Eur Heart J 2010;31: Botto LD, Lin AE, Riehle-Colarusso T et al. Birth Defects Seeking causes: Classifying and evaluating congenital heart defects in etiologic studies. Res A Clin Mol Teratol 2007;79: Kirk EP, Sunde M, Costa MW et al. Mutations in cardiac T-box factor gene TBX20 are associated with diverse cardiac pathologies, including defects of septation and valvulogenesis and cardiomyopathy. Am J Hum Genet 2007;81: Monserrat L, Hermida-Prieto M, Fernandez X et al. Mutation in the alpha-cardiac actin gene associated with apical hypertrophic cardiomyopathy, left ventricular non-compaction, and septal defects. Eur Heart J 2007;28: Geisterfer-Lowrance AA, Kass S, Tanigawa G et al. A molecular basis for familial hypertrophic cardiomyopathy: a beta cardiac myosin heavy chain gene missense mutation. Cell 1990;62: van Spaendonck-Zwarts KY, van den Berg MP, van Tintelen JP. DNA analysis in inherited cardiomyopathies: current status and clinical relevance. Pacing Clin Electrophysiol 2008;31 Suppl 1:S Kamisago M, Sharma SD, DePalma SR et al. Mutations in sarcomere protein genes as a cause of dilated cardiomyopathy. N Engl J Med 2000;343: Bienengraeber M, Olson TM, Selivanov VA et al. ABCC9 mutations identified in human dilated cardiomyopathy disrupt catalytic KATP channel gating. Nat Genet 2004;36: McNair WP, Ku L, Taylor MR et al. SCN5A mutation associated with dilated cardiomyopathy, conduction disorder, and arrhythmia. Circulation 2004;110: Kelly M, Semsarian C. Multiple mutations in genetic cardiovascular disease: a marker of disease severity? Circ Cardiovasc Genet 2009;2: Hershberger RE, Cowan J, Morales A et al. Progress with genetic cardiomyopathies: screening, counseling, and testing in dilated, hypertrophic, and arrhythmogenic right ventricular dysplasia/cardiomyopathy. Circ Heart Fail 2009;2: Campuzano O, Beltrán-Alvarez P, Iglesias A et al. Genetics and cardiac channelopathies. Genet Med 2010;12: Hutchison CA 3rd. DNA sequencing: bench to bedside and beyond. Nucleic Acids Res 2007;35: Schouten JP, McElgunn CJ, Waaijer R et al. Relative quantification of 40 nucleic acid sequences by multiplex ligation-dependent probe amplification. Nucleic Acids Res 2002;30:e Koopmann TT, Alders M, Jongbloed RJ et al. Long QT syndrome caused by a large duplication in the KCNH2 (HERG) gene undetectable by current polymerase chain reaction-based exon-scanning methodologies. Heart Rhythm 2006;3: Bhuiyan ZA, van den Berg MP, van Tintelen JP et al. Expanding spectrum of human RYR2-related disease: new electrocardiographic, structural, and genetic features. Circulation 2007;116: Sen-Chowdhry S, Morgan RD, Chambers JC et al. Arrhythmogenic cardiomyopathy: etiology, diagnosis, and treatment. Annu Rev Med 2010;61: Thienpont B, Mertens L, de Ravel T et al Submicroscopic chromosomal imbalances detected by array-cgh are a frequent cause of congenital heart defects in selected patients. Eur Heart J 2007;28: Li F, Lisi EC, Wohler ES et al. 3q29 interstitial microdeletion syndrome: an inherited case associated with cardiac defect and normal cognition. Eur J Med Genet 2009;52: CHAPTER 1 NOVEL CLINICAL MOLECULAR DIAGNOSTIC METHODS 47

49 30. Thienpont B, Zhang L, Postma AV et al. Haploinsufficiency of TAB2 causes congenital heart defects in humans. 31. Ahmad F, Seidman JG, Seidman CE. The genetic basis for cardiac remodeling. Annu Rev Genomics Hum Genet 2005;6: Waldmüller S, Freund P, Mauch S et al. Low-density DNA microarrays are versatile tools to screen for known mutations in hypertrophic cardiomyopathy. Hum Mutat : Waldmüller S, Müller M, Rackebrandt K et al. Array-based resequencing assay for mutations causing hypertrophic cardiomyopathy. Clin Chem 2008;54: Fokstuen S, Lyle R, Munoz A et al. A DNA resequencing array for pathogenic mutation detection in hypertrophic cardiomyopathy. Hum Mutat 2008;29: Zimmerman RS, Cox S, Lakdawala NK et al. A novel custom resequencing array for dilated cardiomyopathy. Genet Med 2010;12: Wheeler DA, Srinivasan M, Egholm M et al. The complete genome of an individual by massively parallel DNA sequencing. Nature 2008;452: Shendure J, Ji H. Next-generation DNA sequencing. Nat Biotechnol 2008;26: Mardis ER. Next generation DNA sequencing methods. Annu Rev Genomics Hum Genet 2008;9: Metzker ML. Sequencing technologies - the next generation. Nat Rev Genet 2010;11: Ashley EA, Butte AJ, Wheeler MT et al. Clinical assessment incorporating a personal genome. Lancet. 2010;375: Venter JC, Adams MD, Myers EW et al. The sequence of the human genome. Science 2001;291: Hodges E, Xuan Z, Balija V et al. Genome-wide in situ exon capture for selective resequencing. Nat Genet 2007;39: Gnirke A, Melnikov A, Maguire J et al. Solution hybrid selection with ultra-long oligonucleotides for massively parallel targeted sequencing. Nat Biotechnol 2009;27: Hoischen A, Gilissen C, Arts P et al. Massively parallel sequencing of ataxia genes after array-based enrichment. Hum Mutat 2010;31: Tewhey R, Warner JB, Nakano M et al. Microdroplet-based PCR enrichment for largescale targeted sequencing. Nat Biotechnol 2009;27: Voelkerding KV, Dames S, Durtschi JD. Next generation sequencing for clinical diagnostics-principles and application to targeted resequencing for hypertrophic cardiomyopathy: a paper from the 2009 William Beaumont Hospital Symposium on Molecular Pathology. J Mol Diagn 2010;12: Meyer M, Stenzel U, Myles S et al. Targeted high-throughput sequencing of tagged nucleic acid samples. Nucleic Acids Res 2007;35:e Parameswaran P, Jalili R, Tao L et al. A pyrosequencing-tailored nucleotide barcode design unveils opportunities for large-scale sample multiplexing. Nucleic Acids Res 2007;35:e Ng SB, Turner EH, Robertson PD et al. Targeted capture and massively parallel sequencing of 12 human exomes. Nature 2009;461: Ng SB, Buckingham KJ, Lee C et al. Exome sequencing identifies the cause of a mendelian disorder. Nat Genet 2010;42: Choi M, Scholl UI, Ji W et al. Genetic diagnosis by whole exome capture and massively parallel DNA sequencing. Proc Natl Acad Sci USA 2009;106: Lalonde E, Albrecht S, Ha KC et al. Unexpected allelic heterogeneity and spectrum of mutations in Fowler syndrome revealed by next-generation exome sequencing. Hum Mutat 2010;31: Hoischen A, van Bon BW, Gilissen C et al. De novo mutations of SETBP1 cause Schinzel-Giedion syndrome. Nat Genet 2010;42: Sun Y, Almomani R, Aten E et al. Terminal osseous dysplasia is caused by a single recurrent mutation in the FLNA gene. Am J Hum Genet 2010;87: Walsh T, Shahin H, Elkan-Miller T et al. Whole exome sequencing and homozygosity mapping identify mutation in the cell polarity protein GPSM2 as the cause of nonsyndromic hearing loss DFNB82. Am J Hum Genet 2010;87: van der Zwaag PA, van Tintelen JP, Gerbens F et al. Haplotype sharing test maps genes for familial cardiomyopathies. Clin Genet 2010, in press 57. Bowden DW, An SS, Palmer ND et al. Molecular basis of a linkage peak: exome sequencing and family-based analysis identify a rare genetic variant in the ADIPOQ gene in the IRAS Family Study. Hum Mol Genet. 2010, in press 58. Yoon S, Xuan Z, Makarov V et al. Sensitive and accurate detection of copy number variants using read depth of coverage. Genome Res 2009;19: INTRODUCTION

50 59. Alkan C, Kidd JM, Marques-Bonet T et al. Personalized copy number and segmental duplication maps using next-generation sequencing. Nat Genet 2009;41: Blow MJ, McCulley DJ, Li Z et al. ChIP-Seq identification of weakly conserved heart enhancers. Nat Genet 2010;42: Cirulli ET, Singh A, Shianna KV et al. Screening the human exome: a comparison of whole genome and whole transcriptome sequencing. Genome Biol 2010;11:R Marioni JC, Mason CE, Mane SM et al. RNAseq: an assessment of technical reproducibility and comparison with gene expression arrays. Genome Res 2008;18: Torres TT, Metta M, Ottenwälder B et al. Gene expression profiling by massively parallel sequencing. Genome Res 2008;18: Hoppman-Chaney N, Peterson LM, Klee EW et al. Evaluation of oligonucleotide sequence capture arrays and comparison of next-generation sequencing platforms for use in molecular diagnostics. Clin Chem 2010;56: Farias-Hesson E, Erikson J, Atkins A et al. Semi-automated library preparation for high-throughput DNA sequencing platforms. J Biomed Biotechnol 2010;2010: Ionita-Laza I, Lange C, M Laird N. Estimating the number of unseen variants in the human genome. Proc Natl Acad Sci U S A 2009;106: Cooper GM, Goode DL, Ng SB et al. Single-nucleotide evolutionary constraint scores highlight disease-causing mutations. Nat Methods 2010;7: Pollard KS, Hubisz MJ, Rosenbloom KR et al. Detection of nonneutral substitution rates on mammalian phylogenies. Genome Res 2010;20: Ng SB, Bigham AW, Buckingham KJ et al. Exome sequencing identifies MLL2 mutations as a cause of Kabuki syndrome. Nat Genet 2010;42: Xi T, Jones IM, Mohrenweiser HW. Many amino acid substitution variants identified in DNA repair genes during human population screenings are predicted to impact protein function. Genomics 2004;83: Li B, Krishnan VG, Mort ME et al. Automated inference of molecular mechanisms of disease from amino acid substitutions. Bioinformatics 2009;25: De Schrijver JM, De Leeneer K, Lefever S et al. Analysing 454 amplicon resequencing experiments using the modular and database oriented Variant Identification Pipeline. BMC Bioinformatics 2010;11: Blankenberg D, Gordon A, Von Kuster G et al. Manipulation of FASTQ data with Galaxy. Bioinformatics 2010;26: Wang K, Li M, Hakonarson H. ANNOVAR: functional annotation of genetic variants from high-throughput sequencing data. Nucleic Acids Res 2010;38:e Qiu J, Hayden EC. Genomics sizes up. Nature 2008;451: Hayden EC. International genome project launched. Nature 2008;451: ten Bosch JR, Grody WW. Keeping up with the next generation: massively parallel sequencing in clinical diagnostics. J Mol Diagn 2008;10: Ho CY, MacRae CA. Defining the pathogenicity of DNA sequence variation. Circ Cardiovasc Genet 2009;2: Szeverenyi I, Cassidy AJ, Chung CW et al. The Human Intermediate Filament Database: comprehensive information on a gene family involved in many human diseases. Hum Mutat 2008;29: van der Zwaag PA, Jongbloed JD, van den Berg MP et al. A genetic variants database for arrhythmogenic right ventricular dysplasia/cardiomyopathy. Hum Mutat 2009;30: CHAPTER 1 NOVEL CLINICAL MOLECULAR DIAGNOSTIC METHODS 49

51

52 CHAPTER 2 CANDIDATE GENE SCREENING

53

54 Chapter 2.1 Mutational characterisation of RBM20 in dilated cardiomyopathy and other cardiomyopathy subtypes Anna Posafalvi, Ludolf G Boven, Cindy Weidijk, Paul A van der Zwaag, Jan G Post, Karin Y van Spaendonck-Zwarts, Imke Christiaans, Maarten P van den Berg, Robert MW Hofstra, Gerard J te Meerman, Richard J Sinke, J Peter van Tintelen, Jan DH Jongbloed Manuscript in preparation

55 ABSTRACT Introduction: Dilated cardiomyopathy (DCM) is an insidious disease of the myocardium, leading to impaired heart function. RBM20, a recently discovered gene associated with DCM encodes the RNA-binding motif protein 20. It is involved in the tissue-specific splicing of titin and several other proteins with an essential function in the heart, and is known to have a five amino acid mutation hotspot in exon 9. Objectives: Our aim was to perform mutation screening of RBM20 in a Dutch cardiomyopathy patient cohort, and to design a spicing assay for evaluation of the pathogenicity of the identified variants. Methods and results: We performed mutation screening of RBM20 by Sanger (DCM patients only; n=436) and gene-panel based (targeted) next generation sequencing (all cardiomyopathy subtypes included; n=1311 patients). This resulted in the identification of 18 novel and 5 known, rare missense variants in 35 probands. In total 10 likely pathogenic or pathogenic missense mutations were identified: 7 missense variants in the RS-rich domain, 1 novel missense variant of the RNA-recognition motif and 2 outside of these important domains. Peripartum cardiomyopathy (PPCM) was observed in two DCM families carrying RBM20 mutations (p.r636h and p.s637n). A differential splicing assay of the known RBM20-target LDB3 was developed to evaluate the predicted pathogenicity of 11 missense variants, but the results were inconclusive. Conclusion: Our study identified a significant number of novel and potentially pathogenic RBM20 mutations, particularly in the RS domain. In addition, known hotspot mutations were identified. Analysis of all cardiomyopathy subtypes by next generation sequencing revealed that likely pathogenic or pathogenic mutations were found almost exclusively in DCM patients. The RBM20 mutations that we found in patients with PPCM were not unexpected as previous studies had shown TTN mutations are a frequent cause of PPCM. They lead to its shifted isoform composition, and TTN is the best characterized splicing target of RBM20. Keywords: dilated cardiomyopathy, RBM20, genetic screening, differential splicing assay

56 INTRODUCTION The RNA-binding motif protein 20 gene, RBM20, is a known disease gene in dilated cardiomyopathy (DCM) (Brauch et al, Li et al 2010, Millat et al, Refaat et al, Guo et al 2012, Wells et al). It was initially reported as missense mutations clustering in a hot-spot between amino acids , encoding an RS-rich domain of unknown function. It has also been observed that RBM20 carriers sometimes exhibit an unusually severe phenotype, or early onset of the disease (Brauch et al). Recent studies showed that RBM20 is involved in the heart-specific splicing of the mrna molecules of several target genes. Some of these targets, such as the sarcomeric giant titin gene (TTN), as well as the cytoskeletal LIM domain binding 3 gene (LDB3), the calcium/calmodulin-dependent protein kinase II delta gene (CAMK2D), and the gene encoding one of the voltagedependent calcium channel proteins involved in membrane polarisation (CACNA1C) (Guo et al 2012) have been linked to cardiomyopathy but also to other cardiac phenotypes. Li et al (2013) demonstrated complex exon skipping and shuffling patterns of titin in Rbm20 -/- rats were demonstrated. These animals were shown to suffer from ultrastructural changes in the heart, including Z line streaming, abnormal myofibril width and orientation, and aggregation of mitochondria (Guo et al 2013). A more recent study crosslinking RBM20 molecules to their RNA targets, followed by immunoprecipitation and RNA sequencing (CLIP-Seq) suggests a role for RBM20 in differential splicing of PDLIM3, LMO7, RTN4 and RYR2 as well, while confirming its previously observed role in splicing of CAMK2D, LDB3 and TTN (Maatz et al). Finally, the phenotypic influence of RBM20 knock down was studied in a mouse stem cell model, which led to the identification of intracellular enrichment of actin stress fibres and thin, elongated sarcomeric ultrastructures during cardiac differentiation, an observation that agrees with the abnormal sarcomere geometry in the thin, weakened myocardium of DCM patients (Beraldi et al). This study also showed disturbed Ca 2+ handling in shrbm20 cardiomyocytes and abnormal splicing of TTN and CAMK2D. To date, the extent to which RBM20 contributes to DCM in the Netherlands has not been studied, nor has the putative role of the gene in other cardiomyopathy subtypes. The aim of this study was to determine if RBM20 mutations are responsible for DCM and other types of cardiomyopathy in the Dutch population and to gain functional evidence for the pathogenicity of some of the identified variants. CHAPTER 2.1 RBM20 IN DILATED CARDIOMYOPATHY 55

57 METHODS Mutational screening Patients: Written informed consent was obtained from index patients and their relatives involved in the first phase of the study with approval of the medical ethics committees of the participating hospitals. Patients included in this phase fulfilled the formal diagnostic criteria for DCM (Mestroni et al). Cardiomyopathy patients in the second phase of this study were evaluated in a standard diagnostic setting. The clinical diagnosis of these patients was inferred from the referral diagnosis to our laboratory, and included a variety of cardiomyopathy subtypes. DNA sequencing: Peripheral blood was collected from cardiomyopathy patients and their relatives when applicable. DNA was isolated according to the standard protocols. In the first phase of this study, Sanger sequencing of whole or part of the RBM20 gene was performed in 436 DCM patients. For this purpose, 19 amplicons covering all exons and the flanking intronic regions of the RBM20 gene were amplified by AmpliTaq Gold (Applied Biosystems, Thermo Fisher Scientific, Waltham, MA, USA) using a standardized PCR protocol (for primer sequences see also table S1). After purification, the PCR products were sequenced on an ABI3730xI sequencer (Applied Biosystems, Thermo Fisher Scientific, Waltham, MA, USA). Resulting sequences were visualized for analysis by Mutation Surveyor software (SoftGenetics LLC, State College, PA, USA). In the second phase of this study, patients DNA (n=1311) was analyzed using our gene-panel-based targeted next generation sequencing (NGS) method as described previously (Sikkema-Raddatz et al, chapters 4.1 and 4.2), and including RBM20. Variant classification: The identified genetic variants were classified as benign (B), likely benign (LB), variant of unknown significance (VOUS), likely pathogenic (LP), and pathogenic (P) as described in chapter 4.1. In short, our classification was based on the changes in the physicochemical nature of the affected amino acid residues, the evolutionary conservation of the respective residue and the region harbouring the variant, the predicted pathogenicity using various software (such as AGVGD, PolyPhen, SIFT, and MutationTaster), and data from literature and online databases when available. Variant frequency information from additional databases (dbsnp, ExAC (including 1000 Genomes and ESP6500), and GoNL) was taken into consideration. Marker analysis: Several genetic markers (and their respective primers) were selected in the region both 5cM up- and down-stream of the R634W 56 SANGER SEQUENCING

58 mutation from the decode high-resolution genetic map, as shown in table S2a. Lengths of PCR products were determined on the ABI 3730xl DNA Analyzer (Applied Biosystems, Thermo Fisher Scientific, Waltham, MA, USA) and the results analyzed using the GeneMapper v 4.1 software (Applied Biosystems, Thermo Fisher Scientific, Waltham, MA, USA). Functional analysis Plasmids: Flag-tagged, kanamycine-resistant pcmv6-entry vectors containing the human RBM20 cdna sequences (wild type, and two mutants: R634P and Y681C) were obtained from OriGene Technologies Inc (Rockville, MD, USA). Site-directed mutagenesis was applied to introduce additional variants as follows: (1) primer pairs were designed that could anneal back to back to the plasmid and for this purpose were phosphorylated at the 5 end, then the desired mutation was introduced in the middle of the forward or reverse primer, with about perfectly matched nucleotides upstream and downstream of the mutated nucleotide; (2) the mutation was introduced with a PCR reaction, using 2.5ng wild-type vector, the mutation specific primers and the Phusion Site-Directed Mutagenesis Kit (Thermo Fisher Scientific, Waltham, MA, USA); (3) PCR products were visualized on 1% agarose gel and, when produced in sufficient amounts, circularized by Quick T4 DNA ligase (New England Biolabs, Ipswich, MA, USA); (4) plasmids were transformed into competent DH5α E. coli cells, colonies selected for kanamycine resistance were mini-cultured and isolated using the GeneJET Plasmid Miniprep Kit (Thermo Fisher Scientific, Waltham, MA, USA) kit; (5) plasmids were Sanger-sequenced and the presence of the introduced mutations confirmed. Tables S3 and S4 show the primers used for site-directed mutagenesis and prior and subsequent plasmid sequencing. Transfection, cell lines: HEK293 cells were cultured in a 6-well culture plate in DMEM containing 10% foetal bovine serum and 1% penicillinstreptomycin. Transfection was performed using 150 µl serum-free DMEM including 1 µg plasmid DNA and 5µl polyethilenimine per sample. Cells were cultured at 37 o C. RNA was isolated with the standard Trizol method 48 hours after transfection, and subsequently reverse transcribed into cdna by using oligo(dt) 18 primers and the RevertAid H Minus First Strand cdna Synthesis Kit (Fermentas, Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer s instruction. Differential splicing assay: First, PCR and gel electrophoresis of cdna samples was performed to enable several quality checks, including CHAPTER 2.1 RBM20 IN DILATED CARDIOMYOPATHY 57

59 analysis of the expression of house-keeping genes and RBM20 mrna, and sequencing RBM20 to confirm the presence of introduced mutations (PCR and sequencing reactions were performed as described above; the primers used for the analysis of RBM20 expression are shown in table S5). As several putative target mrna molecules were available for analyses of potential effects of RBM20 mutations on differential splicing, results of these molecules in non-transfected (endogenous), wild type and hotspot mutation (R634P) transfected HEK293 cells were evaluated (primers designed for amplification of target mrna molecules are shown in table S6). Based on this evaluation, we selected LDB3 for further follow up. As a start, we extracted separated PCR products of various lengths from agarose gel and sequenced them. Next, PCR conditions for amplification of LDB3 were optimised to enable termination of amplification before reaching saturating conditions (36 cycles; 62 o C annealing), while allowing parallel quantification of several PCR products of different fragment sizes. Then, 16 cdna samples originating from transfected HEK293 cells (2 of non-transfected HEK293 cells, 2 of cells transfected with plasmid containing wild type RBM20, 11 of cells transfected with plasmids carrying a single nucleotide variant each, and 1 non-template control) were placed on a 96 well PCR plate in triplo (PCR experimental replicates). To avoid possible plate-position-related effects, samples were distributed according to a computer generated random position. All LDB3 mrna products were PCRamplified. Resulting PCR products were then purified using the High Pure PCR Product Purification Kit (Roche Applied Science, Penzberg, Germany). Subsequently, aliquots of these samples were transferred to a 384-well plate and loaded on the Caliper GX capillary electrophoresis system (Perkin Elmer, Waltham, MA, USA), which separates PCR products by size and quantifies and visualizes the separated products. Each sample was quantified three times (capillary electrophoresis experimental replicates). Data analysis: We first performed peak detection of the three dominant product sizes that correspond to the previously sequenced fragments (the peaks were detected at an approximate fragment length corresponding to the 245, 472 and 613 bp long PCR products). The position of the peak was determined and an area under the curve (AUC) was determined by adding the intensity values +/- 6 bases from the peak position. Thus 9 data points were available for each peak of each wild type/mutant/control transfected sample (3x PCR and 3x capillary electrophoresis experimental replicates). Normalization was performed by setting the value of the AUC to 100 and assigning the proportional value to the 58 SANGER SEQUENCING

60 other peaks. We used the group of certainly benign (wild-type) and the group of certainly pathogenic (R634P, R634W, R636H) transfected cells as controls during the data analysis, and compared the splicing pattern of cells transfected with the remaining variants of uncertain significance to these two control groups. Analysis of variance was used to evaluate differences between cells. RESULTS Mutational screening To study the contribution of RBM20 mutations to the development of cardiomyopathy, specifically DCM, in the Dutch population, we analyzed this gene by Sanger sequencing and targeted NGS (as part of a panel of 55 cardiomyopathy genes). Initially, we studied RBM20 in 62 DCM patients by Sanger-sequencing all exons and the flanking intronic sequence. In an additional cohort of 374 patients, we performed sequencing of amplicons in which potentially interesting genetic variants were previously identified (in the literature or in our 62 patients). Exon 9 was chosen since it encodes the RS-rich (Arginine and Serine-rich) domain and carries the known mutation hot-spot of five neighbouring amino acids between positions (Brauch et al). Exons 6 and 7 were chosen because they encode the RNA-recognition motif (RRM) domain and a putative nuclear localization signal between c (Filippello et al). The RRM domain, in particular, is highly homologous to RRMs of other spliceosomal and SR-proteins, is expected to play an essential role in the protein function and reported by Li et al (2010) to carry one missense variant. Finally, exon 11 was chosen because it harbours the missense variant D888N reported by Refaat et al as causative mutation. However, when our targeted NGS method was implemented into our routine diagnostics (Sikkema-Raddatz; chapter 4.1), we also evaluated the contribution of RBM20 mutations to both DCM and other cardiomyopathy subtypes for 1311 patients. In this study, we focus only on those patients who had interesting RBM20 variants. In 35 patients, we identified 23 missense variants (18 novel) that were either absent or present at very low frequency (MAF<0.0005) in the ExAC ( and/or the GoNL ( databases. Table 1 contains clinical information is given on RBM20 patients and their family members who were identified within the initial phase (Sanger CHAPTER 2.1 RBM20 IN DILATED CARDIOMYOPATHY 59

61 sequencing of DCM patients), while table 2 summarizes information on all the potentially interesting variants identified in this study, including their classification. As expected from their low population frequency, all variants were classified, at minimum, as VOUS (table 2). Together, 13 variants were classified as VOUS, 7 as LP and 3 as P (hot-spot mutations). Six variants were identified in two or more patients, including the pathogenic R634W mutation identified in 3 different patients and the likely pathogenic mutations Y681C and Y1193C identified in 5 and 3 patients, respectively. The variants found in the known mutation hot-spot (R634P, R634W, and R636H) of the RS-rich domain (residues R632-S654) were classified as pathogenic not only as a consequence of their high evolutionary conservation, complete absence in control populations, and pathogenic predictions by multiple in silico programmes, but also due to several previous reports on hotspot mutations leading to an unusually severe phenotype and cosegregation with disease in several families, including the R634P mutation in our studies (table 1). In addition to the mutations mentioned above, we identified four other novel, likely pathogenic mutations in residues of the RS domain: R623Q, R632K, P633L, and S637N. We also identified likely pathogenic mutations outside the RS domain: V535L, Y681C, and Y1193C. The novel missense variant V535L resides in the RRM, a region in which only one genetic variant affecting the same amino acid position (V535I) was previously reported. We considered this variant as likely pathogenic, because it is located in an area of very high evolutionary conservation, is not found in any population database, and is predicted to be pathogenic by three of the four prediction programmes we used (PolyPhen, SIFT, MutationTaster). The Y681C mutation is located just outside the RSrich domain, affects a very conserved residue residing in a significantly conserved region and is predominantly predicted to be a harmful variant. It has been identified in one individual in the GoNL population and twice (MAF= ) in the ExAC population. Finally, Y1193C resides at the end of a Zinc-finger(like) domain (residues ), affects a conserved residue and is surrounded by conserved residues, is predicted to be pathogenic by three out of four prediction programs we used, and has never been identified in the control populations of GoNL and ExAC. In addition to the likely pathogenic and pathogenic mutations described above, 13 missense variants were detected outside the RBM20 domains or regions of known importance and were classified as VOUS (table 2). It is 60 SANGER SEQUENCING

62 Table 1. Clinical information of RBM20 families identified during phase I; Sanger sequencing. The mutations as well as the result of segregation analysis, when applicable, are indicated. Family Year of birth gender RBM20 mutation Age at diagnosis clinical status family M unknown 66 affected DCM 1961 M p.l100f 40 affected DCM clinical details family F p.r634p 33 affected LV-dysfunction; LVEF 11% (63yr); NSVT; ICD (64yr) 1946 M no sample 58 affected DCM; died at age 62yr 1938 M no sample na affected? SCD 41yr 1967 M p.r634p 39 affected DCM; NSVT; PVCs; ICD (39yr) 1969 M p.r634p 31 affected? NSVT; LVEF 53%; PVCs 1910 F no sample na affected? SCD 46yr family M no sample 50 affected DCM; HTX (53yr); hypertrophic cardiomyocytes on histology 1929 M none 72 unclear severe atherosclerosis; LVEF 35% (72yr); VT, MI (72yr), ICD (72yr); AF (79yr) 1959 F p.s637n 35 affected PPCM (35yr); VT; HTX (37yr) 1992 F p.s637n 16 affected DCM (16yr); NSVT; PVCs; ICD and MI (17yr); HTX (19yr) family M p.y681c 56 affected DCM; NSVT; PVCs; AF (59yr) 1974 M non-carrier 21 affected DCM; AF (20); ICD (30) 1942 M no sample 44 affected DCM; AF; died at age 45yr; interstitial fibrosis on histology 1907 M no sample healthy Died at age 59yr family p.v535l 57 affected LV dilatation; AVB; sporadic case family p.r634w 55 affected LV dysfunction; LVEF 35% (67yr); died at age 70yr no sample affected LV dilatation; died at age 67yr no sample affected LV dilatation; died at age 40yr unknown affected DCM family p.r634w and p.g672s 20 affected DCM family F p.r636h 47 affected DCM leading to HTX 1968 F p.r636h 23 affected Died at age 23yr (PPCM); cardiomegaly (post mortem) Abbreviations: AF atrial fibrillation, AVB atrioventricular block, DCM - dilated cardiomyopathy, HTX heart transplantation, ICD implantable cardioverter-defibrillator, LV left ventricle, LVEF left ventricular ejection fraction, MI - myocardial infarction, NSVT non-sustained ventricular tachycardia, PPCM peripartum cardiomyopathy, PVC - premature ventricular contraction, SCD sudden cardiac death, VT ventricular tachycardia CHAPTER 2.1 important to note that a considerable subset of these 13 variants were identified in non-dcm patients, mainly those with hypertrophic cardiomyopathy (HCM). This is in contrast to our likely pathogenic and pathogenic mutations, which were almost exclusively identified in DCM patients, with the only exception being the Y681C and Y1193C mutations identified in both DCM and HCM patients. In addition, two mutations, the likely pathogenic S637N and the pathogenic R636H, were identified in DCM families in which peripartum RBM20 IN DILATED CARDIOMYOPATHY 61

63 Table 2. RBM20 variants identified by Sanger sequencing (S) and targeted sequencing (T) The table includes information on each mutation, the method by which the mutation was identified, the number of patients carrying the mutation and their particular cardiomyopathy subtype, putative carriership of other likely pathogenic or pathogenic mutations, the frequency in the ExAC population and the respective classification of the mutations. Evolutionary conservation, predicted pathogenicity (by AGVGD, SIFT, PolyPhen2 and MutationTaster softwares) and allele frequencies (in dbsnp, ExAC, and GoNL) were taken into consideration for classification. Protein change # of patients (CM cdna coord Genomic coord (37) method subtype) other mutations MAF ExAC dbase classification HGMD L100F c.298c>t 10: S/T 2 (DCM) LP SCN5A & MYH6 (in 1 pat) 0 VOUS G284R c.850g>a 10: T 2 (DCM; HCM) VOUS G309E c.926g>a 10: T 1 (HCM) 0 VOUS V487M c.1459g>a 10: T 1 (HCM) VOUS V535L c.1603g>c 10: S 1 (DCM) 0 LP V545I c.1633g>a 10: T 1 (DCM) VOUS R623Q c.1868g>a 10: T 1 (unsp) LP MYH7 0 LP R632K c.1895g>a 10: T 1 (unsp) 0 LP P633L c.1898c>t 10: T 1 (unsp) LP R634P c1901g>c 10: S/T 1 (DCM) 0 P yes; other aa change R634W c1900c>t 10: S/T 3 (DCM) 0 P yes R636H c1907g>a 10: S 1 (PPCM/DCM) 0 P yes S637N c1910g>a 10: T 1 (PPCM/DCM) 0 LP yes; other aa change G672S c2014g>a 10: S/T 3 (2 DCM; 1 HCM) 0 VOUS R673Q c.2018g>a 10: T 1 (DCM) VOUS yes; other aa change S675T c.2023t>a 10: T 1 (DCM) 0 VOUS Y681C c2042a>g 10: S/T 5 (4 DCM; 1 HCM) P PLN (in 1 pat) LP G758S c.2272g>a 10: T 1 (HCM) LP MYH7 0 VOUS D786G c.2357a>g 10: T 1 (DCM) VOUS I921F c.2761a>t 10: T 2 (1 DCM; 1 HCM) LP NEXN & TNNT2 (in 1 pat) 0 VOUS V1102A c.3305t>c 10: T 1 (HCM) 0 VOUS Y1193C c.3578a>g 10: T 2 (1 DCM; 1 HCM) 0 LP A1208V c.3623c>t 10: T 1 (DCM) 0 VOUS Abbreviations: CM cardiomyopathy, DCM dilated cardiomyopathy, HCM hypertrophic cardiomyopathy, HGMD Human Gene Mutation Database, LP: likely pathogenic, MAF minor allele frequency, P pathogenic, PPCM peripartum cardiomyopathy, VOUS variant of unknown significance. Gene symbols: MYH6 Myosin, heavy chain 6, cardiac muscle, alpha, MYH7 Myosin, heavy chain 7, cardiac muscle, beta, NEXN Nexilin (F actin binding protein), PLN Phospholamban, SCN5A sodium channel, voltage gated, type V alpha subunit and TNNT2 Troponin T type 2 (cardiac). 62 SANGER SEQUENCING

64 cardiomyopathy (PPCM) was also diagnosed. We also identified the missense variant D888N, which was formerly considered to be pathogenic (Refaat et al). However, we anticipate that D888N is a rare polymorphism, as it was only identified in 6 of our 374 Sanger-sequenced DCM patients (0.16%), in 13 of our 1200 NGS-analyzed patients (0.11%), and in comparably low frequencies in control populations: 0.18% in Dutch controls (GoNL; 9 in 996 alleles, which means 9/498 healthy individuals) and 0.28% in the ExAC database, D888N is therefore not included in summary table 2. Cosegregation and haplotype analysis In cases where DNA of affected family members was available, we studied the carriership status of those individuals (for details, see table 1). We were able to show co-segregation of mutations R634P and R636H with the disease phenotype in the respective families. We did not find co-segregation of Y681C in the one small family that was available for testing. We also identified unrelated index patients carrying the same variants: three patients carrying the pathogenic mutation R634W and five patients having the likely pathogenic mutation Y681C. Although no family relation between these patients was known, we hypothesized that these mutations were inherited from common ancestors (founders). Therefore, we performed haplotype analysis using markers within the approximately ±5cM region surrounding the respective mutations. These studies revealed that the patients carrying the R634W mutation share a relatively large haplotype (see table S2b; results shown for two of the three patients), suggesting that the mutation originated from a common founder. In contrast, no shared haplotype could be identified for the Y681C mutation (data not shown). Together with lack of cosegregation of this variant in the small family studied, the fact that it was also found in 1/996 alleles in the GoNL database, and that one of the Y681C patients is also carrying a certainly pathogenic PLN deletion, these results suggest that Y681C is less likely to be pathogenic. However, more data is needed to verify this conclusion. CHAPTER 2.1 Functional evaluation In order to evaluate our classification of variants L100F, V535L, R634P, R634W, R636H, G672S and Y681C using a functional approach, we designed a differential splicing assay. In addition, we included the likely benign variants W768S, W768L and the D888N variant, which was formerly reported RBM20 IN DILATED CARDIOMYOPATHY 63

65 as pathogenic but that we now designate as benign. For this purpose, we transfected HEK293 cells with wild type and mutated human RBM20 cdna expressing vectors. The RNA isolated from these cells was reverse transcribed, and we then studied potential differences in the composition of isoforms of putative splicing targets of RBM20 (Guo et al 2012, Maatz et al) or other spliceosomal proteins. For this purpose, primers were designed for the CAMK2D, CAMK2G, LDB3, SH3KBP1, SORBS1(1), SORBS1(2), TNNT2, TPM1 and TRDN targets, and differences in splicing patterns in presence of wild type, R634P, or endogenous RBM20 production were analyzed. Out of the nine potential targets, only LDB3 showed clear effects in this assay. We therefore decided to continue evaluating differential splicing of the known cardiomyopathy gene (Vatta et al) and RBM20 target (Guo et al 2012, Maatz et al) LDB3. We first sequenced the three different length PCR products of LDB3 detected on agarose gel, using the primers corresponding to 3 sequences of exon 3 and 5 sequences of exon 7, respectively. The two longer products we identified were shown to correspond to transcripts NM_ and NM_ (product of 472 nucleotides, including exon 3, 5, 6 and 7), and transcripts NM_ and NM_ (product of 614 nucleotides, including exon 3, 4 and 7). However, the shorter 245 nucleotide product did not correspond to the remaining isoforms NM_ and NM_ and only contained sequences of exon 3 and 7 (see also figure 1). Next, we analyzed the resulting LDB3 products in cdna derived from HEK293 cells expressing the different RBM20 mutations/variants. A preliminary screening (a PCR under saturating conditions followed by gel electrophoresis) did not show an obvious presence/absence of certain LDB3 products when comparing wild type cells with cells expressing one of the hotspot mutations. However, as we observed subtle differences in LDB3 product intensities between the different samples (data not shown), subsequent analyses were aimed at quantifying the respective products under non-saturating conditions. Unfortunately, this did not lead to the identification of significant differences in product intensities between cells carrying the vector expressing wild type or certainly pathogenic mutations of RBM20 (figure 2). As shown in figure 2, we did observe some differences in the product intensity patterns between non-transfected, wild type transfected and mutation (R634P(1), R634P(2), R634W and R636H) transfected cells, although this was less apparent for the R636H transfected cells. The most prominent effect we saw was relatively high amounts of the 613bp product in non- 64 SANGER SEQUENCING

66 CHAPTER 2.1 Figure 1. LDB3 splicing products identified in transfected HEK293 cells. PCR amplified and sequenced LDB3 transcripts expressed by the HEK293 cells are shown. Two fragments correspond to known LDB3 transcripts/isoforms (472bp and 613bp), while the exon composition of the shortest one (245bp) does not correspond to known transcripts. The known transcripts of LDB3 have the following NCBI IDs: transcript/isoform 1 NM_ , 2 NM_ , 3 NM_ , 4 NM_ transfected cells, while we observed high proportions (90-100%) of the 245 bp product in wild type transfected cells and slightly smaller proportions (<80%) of this short product in mutation transfected cells (with the exception of R636H) (figure 2). In most of the other cases (G672S, W768S, W768L and D888N transfected cells) the product intensity patterns resembled that of wild type transfected cells. The only exception to this was V535L transfected cells, which looked more like those of the certainly pathogenic mutation expressing cells, suggesting that this variant may be pathogenic as well. During our analysis of variants we grouped together the certainly pathogenic, wild type, non-transfected and unknown variants. A significant effect of the base position was present, but there was no significant effect of the baseposition x mutation type interaction. As an illustration, the results for the interaction are shown in figure 3. In summary, we were not able to show the presence of different splicing products or a shift in isoform composition among the samples using our in vitro splicing assay. RBM20 IN DILATED CARDIOMYOPATHY 65

67 Figure 2. Individual LDB3 splicing plots of transfected and non-transfected HEK293 cells. On each plot the X axis shows the basepositions (corresponding to the 245, 473 and 613 bp long fragments), while the associated area under the curve (AUC) values are indicated on the Y axis. Samples marked with * are biological replicates. DISCUSSION In this study, we aimed to identify and functionally evaluate mutations in the RBM20 gene in DCM patients, as well as in patients suffering from other subtypes of cardiomyopathy. This led to the identification of 23 different rare variants that might be involved in disease development. Importantly, the 10 mutations in 17 patients that were classified as likely pathogenic or pathogenic were almost exclusively found in DCM patients, underscoring the idea that RBM20 mutations are associated with the development of this cardiomyopathy 66 SANGER SEQUENCING

68 Figure 3. Grouped LDB3 splicing plots of transfected HEK293 cells. CHAPTER 2.1 subtype. Notably, a few non-hotspot variants of unknown significance or likely pathogenicity were found in HCM, or in both HCM and DCM patients. Moreover, the identification of two likely pathogenic or pathogenic mutations in PPCM cases suggests that RBM20 mutations could be specifically involved in the development of this particular manifestation of familial DCM as previously reported for the TTN gene (van Spaendonck-Zwarts et al). Of the 23 variants we identified, 3 were classified as pathogenic mutations, 7 as likely pathogenic mutations and 13 as variants with unknown clinical significance. Based on its frequency in controls, we classified variant D888N, which was formerly reported to be pathogenic, as a rare polymorphism. To gain more insight into the putative pathogenicity of the promising variants we found, we performed additional tests such as segregation and haplotype analyses. When DNA sample from family members was available, we performed segregation analysis for the mutations found in the index patients. We found the mutations R634P and R636H co-segregated with the disease phenotype, while the Y681C variant did not co-segregate in the one family we could further investigate. One of the most interesting findings was the identification of the S637N mutation in a family with PPCM and DCM. However, the segregation analysis in this family could not prove that this novel RBM20 hot spot mutation (found in the PPCM patient and her DCM-affected daughter, but not in the grandfather who had possible DCM; Spaendonck-Zwarts et al) played and exclusive disease-causing role, leading to our classification of it as a VOUS. It is possible, that the RBM20 mutation together with another not- RBM20 IN DILATED CARDIOMYOPATHY 67

69 yet-identified genetic mutation could explain the disease in this family, with the RBM20 mutation contributing to the PPCM phenotype. In a few other cases, the RBM20 variant was shown to be inherited in combination with other likely pathogenic mutations, e.g. in MYH7 or SCN5A (table 2). Such digenic and oligogenic inheritance is increasingly observed in various types of cardiomyopathy (Bauce et al, Xu et al 2010, Nakajima et al, Roncarati et al, Bao et al, Rigato et al, Pugh et al, Haas et al, chapter 2.2 and chapter 4.1). Likewise, the RBM20 mutation carriers published by Li et al (2010) carry other variants in LDB3 and LMNA as well. For unrelated patients who carried the same mutation, we performed marker analysis in order to check if the mutation is part of the same relatively large haplotype. This resulted in linking three index patients carrying the same R634W mutation, which suggests that they have inherited it from a common ancestor. Finally, in order to be able to further evaluate the pathogenicity level of the mutations found in our patients, we performed functional experiments. Several recent papers indicated that RBM20 has an important role in splicing dynamics of RNA molecules encoding cardiomyopathy-related proteins, mainly by studying these processes under in vivo conditions both in the presence of wild type or mutated RBM20 (Guo et al 2012, Guo et al 2013, Maatz et al). However, there were no myocardial biopsies available from the patients with RBM20 mutations in our study that could be used to evaluate the effect of these mutations on differential splicing in vivo. Therefore, it was our aim to design an in vitro assay in which this could be easily tested but not requiring an invasive procedure. Unfortunately, although our first experiments on transcripts of the validated RBM20-target LDB3 in HEK293 cells indicated minor differences in isoform intensities between samples, no significant differences were apparent upon quantification of the data and the results were not conclusive. One explanation for this could be that the cell line used in this assay has an embryonic kidney origin rather than being derived from embryonic or adult cardiomyocytes. If this were the case, splicing products resulting from compensating mechanisms and/or splicing procedures diverging from regular cardiac splicing may have concealed the true RBM20-related splicing effects. For example, one of the LDB3 transcripts we identified both in non-transfected and in wild type and mutated RBM20- transfected HEK293 cells does not correspond to the known isoforms, and it lacks the alternative Zasp-like motifs (exons 4, 5 or 6) which play a role in 68 SANGER SEQUENCING

70 actin-binding (Huang et al, Klaavuniemi et al). Hence, it might not be the ideal system for modelling cardiomyopathy. On the other hand, it is likely that RBM20 overexpression conditions in our assay influence the outcome more than the effects of the RBM20 mutations, as Maatz et al have recently shown dramatic differences in the splicing of TTN, RYR2, CAMK2D and LDB3 between heart failure patients with relatively high and low expression levels of endogenous RBM20. Moreover, the same study also showed that patients exhibiting high expression of wild type RBM20 had similar LDB3 splicing patterns to a DCM patient who carried the RBM20 mutation S635A. The fact that differences could be observed between LDB3 splicing patterns of the non-transfected (endogenous RBM20 expressing) and the transfected (wild type or mutant RBM20 overexpressing) HEK293 cells during our functional in vitro assay suggests that RBM20 overexpression, regardless of its wild type or mutant sequence, affects differential LDB3 splicing to an extent that conceals mutation effects, and thus hampers the evaluation of pathogenicity levels of the different RBM20 variants. An intriguing finding is that in two RBM20-mutation-carrying families in our study, those carrying the S637N and the R636H mutations, family members were diagnosed with PPCM. Previously, no association had been reported between PPCM and RBM20 mutations. However, we recently showed that PPCM/DCM can be frequently caused by truncating TTN mutations (van Spaendonck-Zwarts et al). In fact, our functional studies on explanted tissue of a TTN truncating mutation carrier demonstrated that the passive force generation of the single cardiomyocytes is drastically decreased, but also showed that the TTN isoform composition (which provides the molecular basis for this passive tension) is shifted towards more long, compliant N2BA isoform production in place of short N2B isoform production (van Spaendonck-Zwarts et al). The splicing of TTN is known to be regulated by RBM20-mediated exon skipping, a process which is disrupted by mutations of RBM20 (Guo et al 2012). Moreover, Rbm20 -/- rats were found to exclusively express the N2BA isoform at all ages, while this isoform was shown to only account for about 15% of expressed titin 20 days after birth in healthy animals (Guo et al 2013). Together, these findings suggest that titin isoform shift may be a common molecular pathway in PPCM/DCM patients with either RBM20 or TTN mutations. RBM20 mutation carriers have been reported to have unusually severe symptoms with very early onset (Brauch et al, Li et al 2010, Wells et al). We have only observed this in two cases within the two families with PPCM/DCM: CHAPTER 2.1 RBM20 IN DILATED CARDIOMYOPATHY 69

71 a patient carrying R636H passed away at the age of 23 while another with the S637N mutation had to go through heart transplantation at the age of 19. It is likely that the more severe nature of the disease in the earliest reported RBM20 families was the result of patient selection bias as more recent papers reported contradictory results, although including patients carrying RBM20 mutations of which the pathogenic nature is questionable (like the S455L and D888N variants) may have affected the outcomes of these studies (Haas et al; Refaat et al). Aberrant splicing, which is the key pathomechanism leading to cardiomyopathy and DCM in RBM20-mutation-carrying patients, has been associated with various diseases including, increasingly, cardiac diseases. For example, it was recently shown that heart failure (due to aortic stenosis, ischemic and DCM) is associated with altered splicing of sarcomeric genes, such as TNNT2, TNNI3, MYH7 or FLNC (Kong et al). Indeed, it is not only RBM20 that is known to contribute to the splicing machinery and splicing events in the physiological processes of the heart, several other splicing factors have also been connected to DCM in various animal models; SC35, ASF/SF2, and FXR1 are involved in the processing of calcium homeostasis or desmosomal proteins (Ding et al, Xu et al 2005, Whitman et al). There are also structural homologues of RBM20 that are known to play a role in the heart. One of these homologues is MATR3, which is involved in adult onset myopathy (Senderek et al), and which has been shown to interact with wild type RBM20, but to lose this ability in the presence of the S635A mutant (Maatz et al). RBM4 and PTB are antagonists of each other competing for the recognition element of TPM1 mrna and regulating its exon selection (Lin & Tarn), while RBM24 is involved in cardiac differentiation and its targets are important for sarcomere assembly (Poon et al). According to the co-translational assembly model, the muscular RNA-binding proteins play an extremely important role in filament assembly, because they bind and stabilize their targets in the nucleus, and mediate the transports of those to close proximity to the sarcomere for immediate use (Zarnescu & Gregorio). It would therefore be worthwhile to study whether some of these spliceosomal proteins/genes could also contribute to human cardiac diseases, e.g. cardiomyopathy or heart failure in general. Remarkable in this respect is that, apart from RBM20, no clearly Mendelian inherited mutations in other spliceosomal proteins causing DCM have yet been reported. It may be that these mutations are rare and private familial mutations that still have to be 70 SANGER SEQUENCING

72 discovered. On the other hand, as the splicing machinery is generally quite complex, involving a number of analogous proteins, it may be able to accept mutations in individual components of the machine without very drastic effect on its function. If this is the case, it will be even more intriguing to find out why RBM20 mutations specifically result in the development of DCM. Since the expression of RBM20 is not strictly limited to the heart (Filippello et al), and the RBM20 RNA-recognition element containing an UCUU core sequence (Maatz et al) is quite short and not very specific, neither of these can sufficiently explain the purely cardiac phenotype of the patients. Further research on how the interconnections of RBM20 with other spliceosomal proteins ultimately mediate the splicing process in the heart in a tissue- and life-stage specific manner is needed. ACKNOWLEDGEMENTS CHAPTER 2.1 The authors would like to thank Jackie Senior and Kate Mc Intyre for editing the manuscript. Anna Pósafalvi was supported by grants from the Jan Kornelis de Cock Foundation. REFERENCES Bao JR, Wang JZ, Yao Y et al. Screening of pathogenic genes in Chinese patients with arrhythmogenic right ventricular cardiomyopathy. Chin Med J (Engl) 2013;126: Bauce B, Nava A, Beffagna G et al. Multiple mutations in desmosomal proteins encoding genes in arrhythmogenic right ventricular cardiomyopathy/dysplasia. Heart Rhythm 2010;7:22-9 Beraldi R, Li X, Martinez Fernandez A et al. Rbm20-deficient cardiogenesis reveals early disruption of RNA processing and sarcomere remodeling establishing a developmental etiology for dilated cardiomyopathy. Hum Mol Genet 2014;23: Brauch KM, Karst ML, Herron KJ et al. Mutations in ribonucleic acid binding protein gene cause familial dilated cardiomyopathy. J Am Coll Cardiol 2009;54: Ding JH, Xu X, Yang D et al. Dilated cardiomyopathy caused by tissue-specific ablation of SC35 in the heart. EMBO J 2004;23: Filippello A, Lorenzi P, Bergamo E et al. Identification of nuclear retention domains in the RBM20 protein. FEBS Lett 2013;587: Guo W, Pleitner JM, Saupe KW et al. Pathophysiological defects and transcriptional profiling in the RBM20-/- rat model. PLoS One 2013;8:e84281 Guo W, Schafer S, Greaser ML et al. RBM20, a gene for hereditary cardiomyopathy, regulates titin splicing. Nat Med 2012;18: Haas J, Frese KS, Peil B et al. Atlas of the clinical genetics of human dilated cardiomyopathy. Eur Heart J 2014; pii: ehu301 [epub ahead of print] Huang C, Zhou Q, Liang P et al. Characterization and in vivo functional analysis of splice variants of cypher. J Biol Chem 2003;278: Klaavuniemi T, Kelloniemi A, Ylannet J: The ZASPlike motif in actinin-associated LIM protein is required for interaction with the α-actinin rod and for targeting to the muscle Z-line. J Biol Chem 2004;279: Kong SW, Hu YW, Ho JW et al. Heart failure-associated changes in RNA splicing of sarcomere genes. Circ Cardiovasc Genet 2010;3: Li D, Morales A, Gonzalez-Quintina J et al. Identification of novel mutations in RBM20 in patients with dilated cardiomyopathy. Clin Trans Sci 2010;3:90-97 RBM20 IN DILATED CARDIOMYOPATHY 71

73 Li S, Guo W, Dewey CN et al. RBM20 regulates titin alternative splicing as a splicing repressor. Nucleic Acids Res 2013;41: Lin JC & Tarn WY: Exon selection in alpha-tropomyosin mrna is regulated by the antagonistic action of RBM4 and PTB. Mol Cell Biol 2005;25: Maatz H, Jens M, Liss M et al. RNA-binding protein RBM20 represses splicing to orchestrate cardiac pre-mrna processing. J Clin Invest 2014 Jun 24. pii: Mestroni L, Rocco C, Gregori D et al. Familial dilated cardiomyopathy: evidence for genetic and phenotypic heterogeneity. Heart Muscle Disease Study Group. J Am Coll Cardiol 1999;34: Millat G, Bouvagnet P, Chevalier P et al. Clinical and mutational spectrum in a cohort of 105 unrelated patients with dilated cardiomyopathy. Eur J Med Genet 2011;54:e570-5 Nakajima T, Kaneko Y, Irie T et al. Compound and digenic heterozygosity in desmosome genes as a cause of arrhythmogenic right ventricular cardiomyopathy in Japanese patients. Circ J 2012;76: Poon KL, TanKT, Wei YY et al. RNA-binding protein RBM24 is required for sarcomere assembly and heart contractility. Cardiovasc Res 2012;94: Pugh TJ, Kelly MA, Gowrisankar S et al. The landscape of genetic variation in dilated cardiomyopathy as surveyed by clinical DNA sequencing. Genet Med 2014;16(8):601-8 Refaat MM, Lubitz SA, Makino S et al. Genetic variation in the alternative splicing regulator RBM20 is associated with dilated cardiomyopathy. Heart Rhythm 2012;9:390-6 Rigato I, Bauce B, Rampazzo A et al. Compound and digenic heterozygosity predicts life-time arrhythmic outcome and sudden cardiac death in desmosomal gene-related arrhythmogenic right ventricular cardiomyopathy. Circ Cardiovasc Genet 2013;6: Roncarati P, Viviani Anselmi C, Krawitz P et al. Doubly heterozygous LMNA and TTN mutations revealed by exome sequencing in a severe form of dilated cardiomyopathy. Eur J Hum Genet 2013;21(10): Senderek J, Garvey SM, Krieger M et al. Autosomal-dominant distal myopathy associated with a recurrent missense mutation in the gene encoding the nuclear matrix protein, matrin 3. Am J Hum Genet 2009;84:511-8 van Spaendonck-Zwarts KY, Posafalvi A, van den Berg MP et al. Titin gene mutations are common in families with both peripartum cardiomyopathy and dilated cardiomyopathy. Eur Heart J 2014;35: Vatta M, Mohapatra B, Jimenez S et al. Mutations in Cypher/ZASP in patients with dilated cardiomyopathy and left ventricular non-compaction. J Am Coll Cardiol 2003;42: Wells QS, Becker JR, Su YR et al. Whole exome sequencing identifies a causal RBM20 mutation in a large pedigree with familial dilated cardiomyopathy. Circ Cardiovasc Genetics 2013;6: Whitman SA, Cover C, Yu L et al. Desmoplakin and Talin2 are novel mrna targets of Fragile X-related protein-1 in cardiac muscle. Circ Res 2011;109: Xu X, Yang D, Ding JH et al. ASF/SF2-regulated CaMKIIdelta alternative splicing temporally reprograms excitation-contraction coupling in cardiac muscle. Cell 2005;120:59-72 Xu T, Yang Z, Vatta M et al. Compound and digenic heterozygosity contributes to arrhythmogenic right ventricular cardiomyopathy. J Am Coll Cardiol 2010;55: Zarnescu DC & Gregorio CC: Fragile hearts: New insights into translational control in cardiac muscle. Trends Cardiovasc Med 2013;23: SANGER SEQUENCING

74 Supplementary table 1. RBM20 sequencing primers. PT-tails were attached to all primers that facilitated sequencing using PT-primers and resulted in better quality sequencing data. amplicon direction primer sequence (5'-> 3') exon 1 F GCCACCGGGAAGGACAAGGG R TCAGCAGGGACGGGAAATGAAGC exon 2/1 F GACCAGTGTGGGAAGGTCTTG R CATTAGAGGGAAACCGGGTACTG exon 2/2 F AATCAACTGAGGCATCCGTCTG R AAGGCAGCTTGACCATCCTG exon 2/3 F TATGGCCCTGAAACAGATGG R GATAGCAACGCCTGGTCCTC exon 2/4 F CCCGAGGAACCAACCTCAGAC R CTTGCCCATTGGCAGCGTGAG exon 3 F CAGCCTCTGGGCGCTCTGTG R ACTTTGCCTGTTCTGGTCTCTG exon 4 F GGGCTGCTAGGAAGGTTTGG R TGCTTTCTACATCCGTGAGAAGG exon 5 F CAGCATGTCCAGAGGTACAATC R GCATTCCAGCCTGGCTGTCTC exon 6 F CCTGGGTGATGGGAGTGAGAC R GTGGTCTGTGGCATATACACTG exon 7 F GACGTGGAATCATGCCTTGTG R GCAATGGTTTGCCTCGAGATCC exon 8 F TGGTGGACCAGGCAATGAATG R GAACAGGGCACAGCATGACTC exon 9/1 F TGCACAGTATATCTAAGACAGAGAC R AGACCCAGATCTCGGGTACTTC exon 9/2 F CCGGCAACTGGACAAGGCTG R CTCCTTCAGGGCCTGCCTCG exon 10 F GCTGGGACCTGCATTCAATATC R GGGTCTCAGCCATATTCCATCC exon 11/1 F ATGGCCAAGTCTTGTGCCTTCC R CCGCTCAGCATCCAGATTTAGG exon 11/2 F AGCCTCAAGTCACCCAGAGAAC R GGTGAGCAGGAGTCCAATCAAC exon 12 F GCTCCTAATGACAGTGCTTTGG R CAAGCTCTTGAGGTTGCTATGG exon 13 F TGGAGCTCGTGGCTCCCATTTC R AAACAGCCTGGTGTGCTTGG exon 14 F GCACAGATGCCAGGAGAGGGATG R TGGGTGACTTGCTCCTGGAGAC CHAPTER 2.1 RBM20 IN DILATED CARDIOMYOPATHY 73

75 Supplementary table 2a. Marker analysis primers. The markers used to screen for a potentially shared haplotype inherited from a common ancestor were selected from the decode high resolution genetic map. The names and primers can be found in the first two columns. The RBM20 gene can be found at M (NC_ ) and markers in a region of 5 cm at each side were selected (see Location and Distance coloumns). Marker Primer sequences 5'- 3' Location (NC_ ) D10S530 TCTAGCAGTAAGAGTTGTGTCTCC 107.5M / cm 4.4 TTGACAAGGCCATCAAAAC D10S1778 CTTGGTTATGATCTCACATGGTCT 108.1M / cm 3.7 CTGCTCTGGATTGAATGTTT D10S521 CTCCAGAGAAAACAGACCAA 109.3M / cm 2.1 CCTACCATCAATCAACTGAG D10S597 GAATGAAGACATCCAGAGG 111.2M / cm 1.3 GCAAGTATCAGAAACCCAA D10S543 AAAGATGTTCAGGTAGATAACACAC 111.8M / cm 0.6 ATCCCTCAGCCCCACT MUTATION (R634W) D10S1760 GCGAGACTCCATCTCCATAG 113.7M / cm 0.6 CCATATAGTGGGTGGCTTAAA D10S1429 GCTCGTAATAGCTTTGTCCA 114.1M / cm 1.4 ATGAAACCATATATGTGACTTTTTG D10S168 CATGGCACTAATAGAGTTAAC 114.7M / cm 2 TTCACTTGGGATGGAGGCA D10S554 GGAGGACTCATGTCAGACTT 116.0M / cm 4.4 CCTACCTTTAATTCAGCCCT D10S468 CAGGCATGTCCATGTAGGTA 117.2M / cm 6.7 TCTGTAAATAACTCATTTGTCCG Distance to RBM20 gene (cm) Supplementary table 2b. Results of marker analysis for the R634W mutation carriers Markers Expected product range (length in bp) Patient 1 Patient 2 Control 1 Control 2 allele 1 allele 2 allele 1 allele 2 allele 1 allele 2 allele 1 allele 2 D10S D10S D10S D10S D10S MUTATION (R634W) D10S D10S D10S D10S SANGER SEQUENCING

76 Supplementary table 3. RBM20 vector sequencing primers amplicon VP1.5_F 02A_R_PT2 02B_F_PT1 02C_R_PT2 02D_F_PT1 exon5_r exon3_f 09A_R_PT2 09B_F_PT1 11A_R_PT2 11B_F_PT1 XL39_R primer sequence (5'-> 3') GGACTTTCCAAAATGTCG CATTAGAGGGAAACCGGGTACTG AATCAACTGAGGCATCCGTCTG AGATAGCAACGCCTGGTCCTC CCCGAGGAACCAACCTCAGAC CCACAGAAGCCAAAGGAAATG CTGGGAGCTGCATGTGAAAG AGACCCAGATCTCGGGTACTTC CCGGCAACTGGACAAGGCTG CCGCTCAGCATCCAGATTTAGG AGCCTCAAGTCACCCAGAGAAC ATTAGGACAAGGCTGGTGGG Supplementary table 4. Site-directed mutagenesis primers. The 5 primer ends were phosphorylated. CHAPTER 2.1 Variant Primer sequences 5'- 3' TM Calculator Annealing L100F_F CAGCTCACCTTCCACCGGC 70,84 71 L100F_R GGCCTGCAGTTGAGCCAGC 71,09 71 V535L_F GGCTGCCCTTTGGAAAGGT 67,73 68 V535L_R CCAGGTTAATGAGGTCATTCTCAGTG 67,73 68 R634W_F AGAAAGGCCGTGGTCTCGTAG 66,85 67 R634W_R GGGCCATATCTGTCTGCTTCC 67,13 67 R636H_F CGCGGTCTCATAGTCCGGT 67,79 68 R636H_R GCCTTTCTGGGCCATATCTGTC 67,99 68 G672S_F TCCTGGGAGCACTCTCCCTATGC 71,74 71,5 G672S_R GTCCCGGCTATTGCCCCAGT 71,45 71,5 W768S_F CAAAGCCAAGTCGGACAAGTATCTG 68,72 69 W768S_R GGCTCTTTCCGGTAGTAGCCGT 68,54 69 W768L_F CAAAGCCAAGTTGGACAAGTATCTGA 68,14 68 W768L_R GGCTCTTTCCGGTAGTAGCCGT 68,54 68 S637N_F CGGTCTCGTAATCCGGTGAGC 70,16 70 S637N_R CGGCCTTTCTGGGCCATATC 69,82 70 D888N_F AAAGTGAGGCAGAGGGGGAG 67,08 67 D888N_R CACTCTCCCAATTTTGTTCCTTCTT 66,44 67 RBM20 IN DILATED CARDIOMYOPATHY 75

77 Supplementary table 5. RBM20 cdna primers amplicon direction primer sequence (5'-> 3') exon 3 F CTGGGAGCTGCATGTGAAAG exon 5 R CCCACAGAAGCCAAAGGAAATG exon 9 F GGAGAAGTACCCGAGATCTG exon 9 R TGGCCTCGTCTTTCCTCCTG exon 10 F AACAGGAGGGCATGGAAGAAAG exon 11 R CATTCCCTACGGCCTTGACTC Supplementary table 6. Target cdna primers. The following primer pairs were used for differential splicing analysis of the target molecules of RBM20. Abbreviations: CAMK2D calcium/calmodulin-dependent protein kinase II delta; CAMK2G - calcium/calmodulindependent protein kinase II gamma; LDB3 LIM domain binding 3; SH3KBP1 SH3- domain kinase binding protein 1; SORBS1 sorbin and SH3 domain containing 1; TNNT2 troponin T type 2 (cardiac); TPM1 tropomyosin 1 (alpha); TRDN triadin. target gene direction primer sequence (5'-> 3') CAMK2D F AAGGGTGCCATCTTGACAAC R TCAAAGTCCCCATTGTTGAT CAMK2G F CAACGATCCACGGTGGCATCC R GTGTAGGCCTCAAAGTCCCCA LDB3 F TCAAAGCGTCCCATTCCCATC R TGAATTCTGTCCCCGTCATCTG SH3KBP1 F GTAGAGGAAGGATGGTGGGAA R CTACTTCAATTGACCTTGGTC SORBS1 /1 F AGAGCACTCAGGACTTAAGC R AGATGCAGGAAACTGGTAGG SORBS1 /2 F CGCTCTTCCTCACTGAAGTC R GAGTAGGGCTGATGGCTGAG TNNT2 F CATAGAAGAGGTGGTGGAAGAG R TCCTTCTCCCGCTCATTCC TPM1 F TGGACGCCATCAAGAAGAAG R TCATATTTGCGGTCGGCATC TRDN F CAAAGACACTGGCGAAAG R GCTTGTTCTGTCGGTAAGG 76 SANGER SEQUENCING

78

79

80 Chapter 2.2 Missense variants in the rod domain of plectin increase susceptibility to arrhythmogenic right ventricular cardiomyopathy Anna Posafalvi, Petros Syrris, Vincent Plagnol, Ludolf G Boven, Marieke C Bolling, Judith A Groeneweg, MP van den Berg, Marcel F Jonkman, Arthur AM Wilde, Richard NW Hauer, Richard J Sinke, William McKenna, J Peter van Tintelen, Jan DH Jongbloed Manuscript in preparation

81 ABSTRACT Aims: Although mutations causing arrhythmogenic right ventricular cardiomyopathy (ARVC) have been identified in several genes, they do not explain the disease in all patients. Since the majority of these genes encode desmosomal components, we hypothesized that proteins known to be physically linked to the desmosome might contain ARVC-causing genetic variations. Thus, we screened patients for mutations in the PLEC gene encoding the cytolinker protein plectin that anchors intermediate filaments to the desmosomes. Methods & Results: We sequenced 107 ARVC patients from the Netherlands and 358 patients from the UK with either Sanger or high-throughput sequencing for the coding regions of PLEC, and identified 96 novel or low frequency (<2%) variants scattered across the gene. A comparison with the genetic variation of PLEC seen in the GoNL control population revealed an area of the rod domain harbouring multiple missense variants that are only present in ARVC patients. Careful classification of the variants based on their conservation, frequency, and predicted pathogenicity level, combined with the lack of genetic variability of this region in healthy controls, suggests that these variants, which are located in the domain responsible for homodimerization, may affect normal plectin function. Conclusions: Although PLEC has been hypothesised as a promising candidate gene for ARVC, our current studies do not support mutations in this gene as the primary cause of ARVC. Our data do, however, suggest that PLEC missense mutations, in particular those in regions of the protein lacking variation in healthy controls, could play a risk factor role in an oligogenic inheritance model of ARVC. Keywords: ARVC, oligogenic inheritance, plectin, rod domain

82 INTRODUCTION Arrhythmogenic right ventricular cardiomyopathy (ARVC) is a progressive heart disease characterized histologically by fibrofatty infiltration of the ventricular myocardium. Clinically, ARVC patients may suffer from ventricular arrhythmias, syncope, and sudden cardiac death as early as young adulthood (Sen-Chowdhry et al), with the majority of cases diagnosed before the age of 40 (Teekakirikul et al). Familial ARVC is a genetically heterogeneous disease and is most commonly transmitted as an autosomal dominant trait with variable expression and incomplete penetrance (te Rijdt et al). However, oligogenic inheritance is increasingly observed (Bauce et al, Xu et al, Nakajima et al, Bao et al), and carriership of multiple rare variants was recently suggested to be related to a more severe disease outcome (Marcus et al 2013, Rigato et al). Currently, ARVC is considered a disease of the desmosome, an important cell-cell adhesion structure. Genes encoding desmosomal proteins such as plakophillin-2 (PKP2), desmoglein-2 (DSG2), desmocollin-2 (DSC2), junctional plakoglobin (JUP), and desmoplakin (DSP) have been reported to carry ARVC-related mutations (te Rijdt et al). Most mutations causing familial ARVC have been found in the PKP2 gene (found in up to 70% of the patients) (van Tintelen et al). Mutations have also been identified in the DSC2 and DSG2 genes although in lower proportions (in up to 15% of patients). Mutations in the JUP and DSP genes are less frequent in ARVC patients, and are more often associated with the cardiocutaneous syndromes Naxos disease (McKoy et al) and Carvajal syndrome (Norgett et al), respectively. Finally, desmin (DES), the intermediate filament protein that builds up the cytoskeleton of cardiomyocytes and is anchored to the membrane via desmosomes has been implicated in ARVC as well (te Rijdt et al). Recently, more ARVC-associated genes have been discovered: the cellcell adhesion molecule α-3 catenin (CTNNA3); the nuclear envelope protein lamin A/C (LMNA); titin (TTN), which is involved in sarcomere assembly and passive elasticity; regulators of intracellular Ca-levels, namely ryanodine receptor 2 (RYR2) and phospholamban (PLN); transforming growth-factor β-3 (TGFB3) and transmembrane protein 43 (TMEM43), of which the function is unknown (te Rijdt et al). These new genes still only contribute to a small proportion of this predominantly desmosomal disease, and roughly half of ARVC cases remain unexplained (Jacob et al, Marcus et al 2013). It was therefore the purpose of this study to explore whether the putative candidate CHAPTER 2 PLEC IN ARRHYTHMOGENIC CARDIOMYOPATHY 81

83 gene PLEC, which encodes a protein (plectin) physically connecting the cardiac desmosome, is also involved in ARVC. Plectin is a large cytolinker protein and belongs to the plakin family of proteins. Among other functions, plectin is believed to connect the desmosomes to the cytoskeleton by binding the DSP protein of the desmosomes and the desmin/keratin filaments in the myocardium/skin epithelium, respectively. In cardiac tissue, plectin is mainly localized at the intercalated disk and the sarcomeric Z-line, whereas in skin it is located in desmosomes and hemi-desmosomes. This means that plectin potentially has a general and fundamental function in junctional complexes (Wiche et al, Zernig et al). When fully knocked out in mice, the lack of plectin causes severe skin blistering, a phenotype similar to the symptoms of epidermolysis bullosa patients. Although this blistering condition in mice seemed lethal in itself a few days after birth, the animals also exhibited more phenotypic changes, such as generalized myopathies of the skeletal muscle, while ultrastructural differences (e.g. aberrant myofibril bundles and focal loss of Z-lines) were observed in the heart (Andrä et al). Over the past two decades, PLEC has been shown to harbour mutations in patients suffering from various forms of inherited epidermolysis bullosa. Clinical phenotypes associated with mostly homozygous nonsense/frameshift mutations in the gene encompass autosomal recessive epidermolysis bullosa simplex (EBS) with muscular dystrophy (MD) (MIM#226670), EBS-MD with myasthenic symptoms (Winter & Wiche; not in OMIM), EBS-Ogna (autosomal dominant, heterozygous; MIM#131950), EBS with pyloric atresia (MIM#612138), and limb-girdle muscular dystrophy (MIM#613723) (reviewed by Winter & Wiche, Sonnenberg & Liem). Heterozygous carriership of missense mutations was recently reported in mild cases of EBS without cardio-muscular manifestation (Bolling et al 2013). Moving beyond these skin and muscle abnormalities, and based on the symptoms of the plectin KO mouse, PLEC was also suggested to be involved in further disease processes such as neurodegeneration or cardiomyopathy due to increased fragility of the desmosomes (Andrä et al). Nonetheless, only two incidental cases of cardiac involvement have been reported: a plectin mutation-carrying EBS-MD patient with ventricular hypertrophy (Schröder et al) and an EBS-MD patient who, by the age of 30, was discovered to have asymptomatic dilated cardiomyopathy (DCM) which later progressed to right ventricle involvement including (septal) fibrosis as well as features associated with arrhythmogenic cardiomyopathy 82 SANGER SEQUENCING

84 (Bolling et al 2010). Very recently, left ventricular non-compaction cardiomyopathy developed in an EBS-MD patient with a homozygous plectin truncation (Villa et al). An intriguing related phenomenon was also observed in a striated-muscle-specific conditional knock-out mouse model of plectin: the mice showed progressively declining endurance performance and, by the age of 16 months, a remarkable increase in connective tissue formation in the heart, indicating cardiomyocyte degeneration (Konieczny et al). In fact, the fibrofatty replacement of cardiomyocytes is the key underlying factor for cardiac conductivity problems in arrhythmogenic cardiomyopathy, and is one of the major task force criteria for the diagnosis of the disease (Marcus et al 2010, Elliott et al). Unfortunately, no electrophysiological studies addressing potential arrhythmias had been carried out in these knock-out mice. Based on these observations, we hypothesized that plectin might play a role in cardiac pathophysiology, and analysed a large cohort of ARVC patients for carriership of PLEC variants. In many of these patients, novel or low-frequency variants with a potential pathogenic nature were identified. Comparing the genetic variation of PLEC in the Genome of the Netherlands (GoNL) database with that in our patients suggests that novel variants in a region of the rod domain may underlie, or at least increase the susceptibility to, ARVC. Although our data does not prove a major causal role for PLEC in ARVC development, it does suggest that variants in the PLEC gene may contribute to the development of an ARVC phenotype in a disease model in which a combination of several factors, both genetic and non-genetic, are needed to reach the threshold levels above which disease development is initiated. CHAPTER 2 MATERIALS AND METHODS Patients Patients were clinically evaluated and the diagnosis of ARVC was based on the criteria of the consensus-based international Task Force Criteria (TFC) (Marcus et al 2010). Our Dutch Sanger-sequenced cohort included 88 patients who fulfilled these clinical criteria for ARVC (TFC+) and 19 patients who did not completely fulfil them. The British gene-panel sequenced cohort included 123 TFC+ and 235 TFC- index patients. All patients were evaluated for genetic variations in PLEC. Many of the patients had also been screened for mutations in other ARVC-related genes. PLEC IN ARRHYTHMOGENIC CARDIOMYOPATHY 83

85 Genetic analyses Genomic DNA was isolated from blood samples using standardized procedures. Written informed consent was obtained from the index patients and their relatives according to the participating hospitals medical ethics committee guidelines. For the Dutch cohort, primers for PCR amplification of the coding regions of the PLEC gene (about a 14 kb long sequence) were designed to encompass the coding exons as well as adjacent intronic sequences as described previously (Bolling et al 2013), using the sequence obtained from the GenBank database. Variant annotation is according to NM_ unless otherwise indicated. Amplifications were conducted following a standard PCR protocol and PCR products were analysed by direct Sanger sequencing. For the UK cohort, a next generation sequencing (NGS) protocol was designed to screen 2.1 Mbp of genomic DNA sequence per patient that covered the coding regions of genes known to be associated with inherited cardiomyopathies (including ARVC) and arrhythmia syndromes. The gene panel also included PLEC as a candidate gene for ARVC. The sequencing methodology has been reported in detail (Lopes et al). Data analysis: variant classification In this study, the pathogenic nature of the identified missense mutations were judged based on: (1) the differences in the physico-chemical properties of the affected amino acids, (2) the evolutionary conservation of the affected amino acids across orthologues, (3) the frequency of the variant in control populations and databases (such as dbsnp ( SNP/), 1000Genomes ( and ESP ( gs.washington.edu/evs/)), and (4) the predicted pathogenic or benign nature of the variant (identified using the Alamut software; Interactive Biosoftware, Rouen). Every variant was then classified as either benign, likely benign, variant of unknown significance, likely pathogenic, or pathogenic (see chapter 4.1 for a detailed description). Data analysis: frequency-based variant clustering Chromosomal positions of single nucleotide variants of the PLEC gene identified in our Sanger- and gene-panel-sequenced patient cohorts were annotated using information from the 1000 Genomes (1000G), the Exome Sequencing Project (ESP) and GoNL ( control 84 SANGER SEQUENCING

86 cohort databases using an in-house developed script (to be published elsewhere), and frequency information on these variants was collected. PLEC variants with an allele frequency <2% in 1000G, GoNL, and ESP (considering the European ancestry population only) were uploaded into SeattleSeq and only non-synonymous coding variants were analysed further (variants with an allele frequency 2% were considered as definite polymorphisms). The resulting list of low frequency and novel ARVC variants was compared with a list of low frequency and novel coding variations extracted from the GoNL database, which was filtered for allele frequencies as described above. Regions consisting of consecutive novel or rare (<0.5%) genetic variants of the PLEC gene identified exclusively in our ARVC TFC+ or TFC- patients, but lacking novel and/or rare variants in the GoNL control population, were subsequently checked for the putative presence or absence of any variation with allele frequency of >2% in GoNL. CHAPTER 2 RESULTS Genetic screening of plectin As a result of the genetic analysis of 465 patients (211 TFC+ and 254 TFC-) for the coding sequences of PLEC, we identified 96 variants that were either novel (not known from control populations) or low frequency (<2% in control populations), the majority of which were missense mutations (tables 1 and 2). Next, all variants were classified, assessing their potential pathogenic nature with the help of in silico prediction tools, as described elsewhere (for the classification criteria: see chapter 4.1; for the classification outcome: see supplementary table 1). For some of the PLEC variants we studied their possible cosegregation with the disease phenotype in the families, but this yielded no conclusive results (data not shown). In order to further assess the potential involvement of the variants in the disease, we compared the localization of these variants with the localization of genetic variants of the PLEC gene in a healthy population by extracting data from the GoNL whole genome sequencing database of 500 individuals (1000 alleles/genomes). This led to the identification of clustering of consecutive novel and low frequency variants in probable and known ARVC patients and the lack of these variants in healthy controls. PLEC IN ARRHYTHMOGENIC CARDIOMYOPATHY 85

87 Table 1. PLEC variants identified in ARVC patients from the Netherlands. Overview of all novel and low frequency (<2%) coding variants identified in the PLEC gene in Dutch ARVC patients. Frequencies in the healthy control population of the GoNL database are indicated. The variants have been classified on the basis of differences in physicochemical nature, conservation, frequency, and predicted pathogenic effect. Frequencies of variants found in patients are indicated in red, while of variants found in controls in green. Frequency-based clustering of variants in the rod domain is shown in light yellow. variant position Sanger NGS data classification genomic cdna protein ARVC GoNL freq (%) G>A R141C VOUS C>T A7V 1x - LB G>C D42H 1x - VOUS C>T Q10* 1x - VOUS _ c69_70instac D23_N24insY 1x - VOUS C>T R316Q VOUS C>T V402M VOUS G>A R422Q 1x - LB C>T R433Q 3x LB C>T V512M VOUS G>C S523R 1x VOUS G>A A582V LP C>G Y612* 1x - LP A>G Y625C 1x - LP C>G Q639H LB G>A R1044C VOUS G>A R1365W 2x LB T>C K1427R 1x - LB G>C T1467R VOUS C>T R1480H LP G>A R1496C LP C>T V1593M LB G>A R1646H 1x - VOUS C>T R1648Q VOUS C>T R1670Q 1x LB G>A A1685V LP C>T T1859M 1x - VOUS C>T R1875W 1x - LP G>A R1934H 1x - LP C>G R1954G 2x - LP G>A R1963W 1x - LP G>A R1963Q 3x - VOUS C>G L1972V 1x - LP G>A E1974K 1x - LP G>A A1978T 1x - LP A>G K2037E 1x - LP G>A R2110Q 1x - LP G>A A2158V 1x LP G>A A2241V VOUS A>C V2248G VOUS G>A R2276H 1x - LP G>A A2311T 1x - VOUS G>A R2419Q 1x - LP C>T A2560T 2x LB G>A R2694W LB C>G A2744G 2x - LB 86 SANGER SEQUENCING

88 C>T R2808Q LB C>T A2820T LB C>T R2821Q LB G>A A2871V 1x VOUS C>T V2961M VOUS T>C Y2967C LB C>T E2974K 3x VOUS C>T E2975K 1x - VOUS C>T R3076Q VOUS G>C D3077E B C>T D3130N LB C>T R3155Q LB C>T D3320N VOUS G>A R3446C 1x VOUS C>T G3458R LB G>C G3490A LB G>A R3637C VOUS C>A A3686S LB C>T E3761K LP G>A A3775V LB C>T E3914K LB C>T G4008S VOUS G>A T4044M LB C>T R4146H LP C>T V4148I LP C>T V4200M LP C>T E4201K LP G>A V4399I 1x - LB CHAPTER 2 Abbreviations B: benign; LB: likely benign; LP: likely pathogenic; P: pathogenic; VOUS: variant of unknown significance Clustering of novel, likely pathogenic variants in the rod domain We have identified one large, potentially disease-associated cluster of novel, missense genetic variants in the rod domain of the PLEC gene in the Dutch patient cohort (variants T1859M-R2110Q, see table 1). Interestingly, another, significantly overlapping ARVC-associated region was found in the same domain in the UK cohort (variants R1688C-E2157A, see table 2). This cluster of variants was not only promising due to its presence in ARVC patients and absence in control populations, but also because all variants within the cluster were classified as likely pathogenic or variant of unknown significance (VOUS) (for details of the variant classification, see supplementary table 1). Therefore, on the basis of the clustering of variants in patients and their predicted pathogenicity, we considered this region as potentially pathogenic. PLEC IN ARRHYTHMOGENIC CARDIOMYOPATHY 87

89 Table 2. PLEC variants identified in ARVC patients from the United Kingdom. Overview of all novel and low frequency (<2%) coding variants identified in the PLEC gene in British ARVC patients. Frequencies in the healthy control population of the GoNL database are indicated. The variants have been classified on the basis of differences in physicochemical nature, conservation, frequency, and predicted pathogenic effect. Frequencies of variants found in patients are indicated in red, while of variants found in controls in green. Frequency-based clustering of variants in the rod domain is shown in light yellow. variant position NGS data classification genomic cdna protein TFC+ TFC- GoNL freq (%) G>C D10E 1x - LB G>A P89L 1x - VOUS C>T R102H 1x - VOUS G>A R141C VOUS G>A P168L 1x - LB C>T R248Q 1x - VOUS C>T R316Q VOUS C>T V402M VOUS C>T R433Q 5x 8x LB C>T V512M VOUS G>C S523R 1x VOUS C>T C545Y 1x - LP G>A A582V LP C>G Q639H LB G>T P825Q 1x - LP C>G C850S 1x - LP C>T A987T 1x 1x - B C>T V988M 1x - VOUS G>A R1044C VOUS G>A R1118C 1x - LB C>T R1291Q 1x - LB G>A R1365W LB G>C R1397G 1x - LB T>C K1427R 1x - LB G>C T1467R VOUS C>T R1480H LP G>A R1496C LP A>C V1546G 1x - LP C>T V1593M 1x LB C>T R1648Q VOUS C>T S1656L - LB C>T R1670Q 4x LB G>A A1685V LP G>A R1688C 2x - VOUS G>A A1747V 1x - LP C>T E1762K 2x - LP G>A T1884M 1x - VOUS G>A A1909V 1x - VOUS G>A A1961V 1x - LP G>A R1963W 2x - LP G>A R1963Q - VOUS G>A A2144V 1x - LP T>G E2157A 1x - VOUS G>A A2158V LP C>A Q2203H 1x - LB G>A A2241V VOUS G>A R2246W 1x - VOUS A>C V2248G VOUS 88 SANGER SEQUENCING

90 C>T R2286Q 1x - LP C>T R2316Q 1x - VOUS C>T A2560T 1x 2x LB G>A R2694W 1x LB C>T R2808Q 2x LB C>T E2818K 1x - LP C>T A2820T LB C>T R2821Q 1x LB T>C I2865V 1x - LB G>A A2871V 2x VOUS T>C K2915R 1x - VOUS C>T V2961M VOUS T>C Y2967C 2x 13x LB C>T D2973N 2x - VOUS C>T E2974K 1x 2x VOUS C>T E2975K 1x - VOUS C>T A2981T 1x - LB C>T R3076Q VOUS G>C D3077E 2x 3x B C>T D3130N LB C>T E3149K 1x - VOUS T>C T3152A 1x - LB C>T R3155Q LB C>T D3320N VOUS C>T R3335Q 1x - LB T>C K3352E 1x - LB G>A R3368C 1x - VOUS G>A R3409C 1x - VOUS G>A R3446C 1x 3x VOUS C>T G3458R 1x 6x LB C>T R3485Q 1x - LB G>C G3490A - LB C>T R3514Q 1x - LB A>G S3583P 1x - LP G>A R3637C 1x VOUS C>A A3686S LB A>G S3720P 1x - LB C>T E3761K LP G>A A3775V LB G>A A3813V 1x - LB G>A A3816V 1x - LB C>T E3914K LB T>A Q3921L 1x - LP C>G D4004H 2x 1x - B C>T G4008S VOUS G>A T4044M 2x 5x LB C>T R4146H LP C>T V4148I LP C>T V4200M LP C>T E4201K 1x 1x LP G>A R4206C 1x - LP C>T D4219N 1x - LP C>T A4377T 1x - LB G>A P4410S 1x - LP C>T G4629S 1x - LP G>A R4667C 1x - LP CHAPTER 2 Abbreviations B: benign; LB: likely benign; LP: likely pathogenic; P: pathogenic; VOUS: variant of unknown significance PLEC IN ARRHYTHMOGENIC CARDIOMYOPATHY 89

91 Absence of frequent non-synonymous variations in the rod domain region in controls Next, we investigated whether frequent (MAF>2%) SNPs reside in the potentially pathogenic region in the rod domain in the GoNL control population. For this purpose, all variants of PLEC between chromosomal positions were extracted from the GoNL database then uploaded to SeattleSeq for annotation with protein coding features (table 3). Only a few synonymous variants were identified, with the exception of one additional missense variant (K2047E) reported with the frequency of 8%. This variant, however, had a very low quality score, which indicated that this was a sequencing artefact rather than a true variant. Hence, it seems that missense mutations in this region are not tolerated without phenotypic consequences. Further interesting regions of plectin Based on the absence of low frequency variants in patients versus controls, we identified additional interesting, potentially ARVC-associated regions of PLEC. Notably, these were only found in the UK cohort (probably due to the larger number of patients involved in the study) and spanned much shorter regions of the gene than the one in the rod domain. One of these ARVCassociated regions was the spectrin repeat region (P825Q-V988M): though this region only contained two likely pathogenic variants in our patients, it did not contain any non-synonymous variant in the GoNL control population (data not shown). The other two variants, despite being classified as VOUS, might also be more damaging, since this segment of the encoded plectin protein is known to be responsible for interactions with, for example, actin, nesprin, and costameric proteins. Likewise, the region of variants R3335Q- R3409C, partially residing in the intermediate filament-binding plectinrepeats of the protein, may also contribute to the development of ARVC. Though this latter region was also free of coding non-synonymous genetic variants in Dutch controls (GoNL), two of the four variants found in patients were classified as likely benign (primarily because they were predicted to be harmless) (see classification in supplementary table 1). Additionally, the C-terminus of PLEC (R4206C-R4667C) could be potentially interesting, but one relatively frequent missense variant (T4539M, 2.381% in GoNL), which was also found in 2/107 Dutch and 18/358 British patients, resides in this otherwise likely ARVC-related plectin-repeat region. 90 SANGER SEQUENCING

92 The potential role of other desmosomal mutations and/or external factors: is ARVC a multifactorial disease? Of the 30 patients who were carriers of PLEC missense variants in the potentially disease-associated region of the rod domain, the vast majority had a TFC+ cardiac phenotype (true ARVC). Of these, 14 patients were found to carry other ARVC-related potentially pathogenic mutations or VOUS in addition to their PLEC cluster variants (table 4, only likely pathogenic and pathogenic variants included), mostly in the PKP2 gene but also in DSC2, DSP, JUP, SCN5A or TMEM43 for some cases. Moreover, five patients had multiple low frequency or novel genetic variants in PLEC (four of which were additional variants in the same rod cluster). While exercise (Perrin et al, Saberniak et al) and certain viral infections (Grumbach et al) remain important contributors to the onset of an ARVC phenotype, our study indicates that a potential oligogenic inheritance might complicate the seemingly multifactorial disease background and cause variable penetrance of ARVC. CHAPTER 2 Table 3. High frequency (>2%) PLEC control variants localized in the ARVCassociated potentially pathogenic region of the rod domain. Variants of PLEC localized in the potentially pathogenic cluster of missense variants associated with ARVC. This region was found to be enriched for missense variants in both the Dutch and UK ARVC patient cohorts and lacking low frequency (<2%) variation in the healthy population represented by GoNL. No frequent coding non-synonymous PLEC variants, except for the c.6319t>c; p.k2047e variant (highlighted in black), which most likely is an artefact, were identified in the GoNL control population. Synonymous variants are indicated in gray. variant position dbsnp GoNL data remark genomic cdna protein rs number quality frequency (%) C>T A1697A rs ,24 47, C>T A1880A - 222,26 8, C>T A1998A - 43,31 4,1096 most likely artefact T>C K2047E - 41,57 8,0645 most likely artefact A>G A2106A rs ,12 31, C>T A2113A rs ,93 35,9177 PLEC IN ARRHYTHMOGENIC CARDIOMYOPATHY 91

93 Table 4. Carriership of additional potentially pathogenic variants in patients with PLEC variants in the ARVC-associated cluster of the rod domain. Of the patients carrying plectin variants in the putatively ARVC-associated region identified in the rod domain, 17 of the 43 were found to be carriers of additional genetic mutations in desmosomal genes or other known ARVC genes. Only genetic variants classified as pathogenic or likely pathogenic are included. patient ID TFC status PLEC cluster variants other PLEC variant other ARVC-related mutation 1 other ARVC-related mutation 2 1 TFC- R1688C (VOUS) DSP c.7994c>t; p.t2665m (VOUS) - 2 TFC- R1688C (VOUS) TFC- R1688C (VOUS) JUP c.902a>g; p.e301g (VOUS) - 4 TFC- A1747V (LP) SCN5A c.665g>a; p.r222q (P) - 5 TFC- E1762K (LP) TFC- E1762K (LP) TFC+ T1859M (LB), K2037E (LP) R422Q (LB), A2744G (LB) PKP2 c.2062t>c; S688P (LP) - 8 TFC+ R1875W (LP) - PKP2 c.1848c>a; Y616X (P) - 9 TFC+ T1884M (VOUS) TFC- A1909V (VOUS) PKP2 c.1132c>t; p.q378x (P) - 11 TFC+ R1934H (LP) - PKP2 del exon 1-4 (P) - 12 TFC+ R1954G (LP) - PKP2 c.397c>t; p.q133x (P) TMEM c.934c>t; p.r312w (VOUS) 13 TFC- A1961V (LP) TFC- R1963W (LP) TFC- R1963W (LP) TFC- R1963W (LP) TFC+ R1963W (LP) - DSP c.4775a>g; p.k1992r (VOUS) - 18 TFC- R1963Q (VOUS) TFC- R1963Q (VOUS) TFC- R1963Q (VOUS) TFC- R1963Q (VOUS) SCN5A c.3206c>t; p.t1069m (LP) - 22 TFC- R1963Q (VOUS) TFC+ R1963Q (VOUS) Q10* (VOUS) PKP2 c.2386t>c; C796R (P) - 24 TFC+ R1963Q (VOUS), K2099R (LP) D42H (VOUS) TFC- L1972V (LP) R433Q (LB) TFC+ E1974K (LP) - PKP2 c.1211dup; L404fs (P) DSP c.269g>a; p.q90r (VOUS) 27 TFC+ A1978T (LP) TFC+ R2110Q (LP) A2242V (B) PKP2 c.235c>t; R79X (P) - 29 TFC- A2144V (LP) PKP2 c.1597_1600delatcc; p.p533fsx561 (P) - 30 TFC- E2157A (VOUS) - - Abbreviations B: benign; LB: likely benign; LP: likely pathogenic; P: pathogenic; TFC task force criteria (diagnostic criteria of ARVC); VOUS: variant of unknown significance 92 SANGER SEQUENCING

94 DISCUSSION We hypothesized that the cytolinker protein plectin, which supports the binding of intermediate filaments to the desmosomes, might carry genetic variants that contribute to the development of, or at least the susceptibility for ARVC, which is known as a disease of the cardiac desmosome. Our reasons were that (1) plectin is highly expressed in the myocardium and is physically connected to the cell junctions which are known to be involved in the pathomechanism of ARVC, (2) its knock down in various mouse models leads to cardiac pathology, and (3) late-onset cardiac symptoms have recently been reported for a couple of mutated plectin-carrying EBS patients. Our Sanger sequencing and NGS-based analysis of the PLEC gene resulted in the identification of 96 novel or low frequency (<2%), mostly heterozygous variants in the PLEC gene in patients with ARVC. Previously, multiple homozygous or compound heterozygous truncating nonsense and frameshift variants of PLEC had been shown to lead to different manifestations of epidermolysis bullosa simplex; only a couple of missense variants were identified in unusually mild cases of EBS (Bolling et al 2013). The majority of the variants now identified are missense by nature, except for a small deletion and two heterozygous truncating variants. The Q10* variant is located in one of the multiple 5 exon 1 sequences of the gene, and upon RNA splicing, it ends up only in one isoform (isoform 1a, transcript NM_ ) that was previously not found to be of importance in the heart (Fuchs et al). The same exon was shown to have a small, in-frame deletion in another patient. The possible pathogenic nature of these two variants is unclear, yet not likely. Moreover, in a third patient, we identified another truncating variant, p.y612* in exon 14, which is part of all currently known plectin isoforms. The respective patient did not show any skin phenotype, i.e. blistering. This is consistent with heterozygous carriers of truncating PLEC mutations generally reported as being healthy, while a homozygous or compound heterozygous form causes EBS. Therefore, it seems that heterozygous truncating PLEC mutations are not disease-related, although we cannot exclude that mild phenotypes might have been overlooked. All remaining variants (n=93) detected in our patient cohorts are missense variants of which a substantial number were classified as VOUS or likely pathogenic based on conservation, predicted effects on protein function and differences in physico-chemical properties of the respective amino acid residues. When analyzing the distribution of these missense variations, we CHAPTER 2 PLEC IN ARRHYTHMOGENIC CARDIOMYOPATHY 93

95 noted that the variants are scattered around the entire gene and do not show an obvious, exclusive clustering any smaller area or domain. Moreover, when studying the presence of novel or rare variants in the control population of the Genome of the Netherlands, various missense mutations were also identified (n=48), and of these a subset were also classified as VOUS or likely pathogenic. This raised the question whether such missense mutations are all harmless variants or whether a subset might have disease-associated effects, as was previously shown for a couple of heterozygous missense mutations, such as the p.r2110w mutation, that led to mild forms of EBS (Bolling et al 2013). For this reason we compared the distribution of PLEC variations in affected and healthy individuals, searching for regions rich in genetic variation in patients, but that show variant deserts in the control group. We subsequently identified one such region in the coiled-coil rod domain (between amino acids ). This region exhibits almost exclusively synonymous (low AND high frequency) genetic variation in the GoNL control population, but contains many missense variants in our ARVC patients. According to their conservation, frequency in SNP databases and predicted pathogenicity, the majority of these variants were classified as likely pathogenic, or, in a few cases, as VOUS. Notably, due to coverage and/or mapping difficulties for some parts of this region, variant calling was hampered and in a few cases based on less than 1000 alleles. We identified an additional short region of the plectin-repeat domains which had ARVC-associated variants clustering in the UK patient cohort but did not see these variations in the GoNL cohort. In addition, several likely pathogenic variants found at positions outside the clustering in the rod domain were identified. Unfortunately, it is not possible to interpret the functional consequences of these missense variants and make conclusions about their potential causative nature without performing further followup experiments. However, a modifier role in ARVC can still be anticipated. What our findings, however, suggest is that the homo-dimerizing rod domain, which seems to contain a well-defined region of disease-associated missense variants in both the British and Dutch patient cohorts (see also figure 1), may be of structural importance for plectin molecules. Indeed, it has been observed that the dimerized rod domains are able to form remarkably stable polymers via further lateral connections with each other (Walko et al). Moreover, in EBS-Ogna mouse and patient keratinocytes, the missense variants of the rod domain were found to increase the proteolysis of plectin 94 SANGER SEQUENCING

96 in the hemidesmosomes, a mechanism which was rescued by treatment with calpain inhibitors (Walko et al). Interestingly, there are a few exceptional yet mild EBS cases recently reported to be due to heterozygous missense variants of PLEC (Bolling et al 2013). In the skin, PLEC homozygous truncations (and as a consequence the lack of plectin protein expression) seem to predominantly cause a dysfunction of the hemidesmosomes, leading to insufficient attachment of keratinocytes to the dermis at the basement membrane zone and manifesting as the basal blistering of the skin (McLean et al, Smith et al). The missense heterozygous variants of the rod domain also cause basal blistering by increasing the vulnerability of plectin to calpain-mediated proteolysis (Walko et al), yet their effect is not as drastic and the phenotype is limited to the hands and feet, areas subject to more mechanical stress. CHAPTER 2 Figure 1. Schematic illustration of regions of interest in PLEC. The marked areas were found to have a number of missense variants found in ARVC patients, yet showing variant deserts in the GoNL control population. Orange-yellow colour indicates the region found in the Dutch patient cohort, while regions coloured in red were found in the British patient cohort. Exon 31 encodes the homodimerizing rod domain of plectin. The exact mechanism by which heterozygous missense variants of PLEC could lead to cardiomyopathy are unknown. However, we know that plectin contains a number of plakin repeat domains typical of desmosomal proteins and responsible for binding intermediate filaments, as well as a highly variable N-terminal actin-binding domain. Thus we anticipate that by binding both intermediate filaments and myofilaments, PLEC may play a role in attracting the structurally robust desmosomes around the more fragile adherens junctions at the cardiac intercalated disks. Adherens junctions anchor the more dynamic and fragile thin (actin) filaments of the neighbouring cardiomyocytes together PLEC IN ARRHYTHMOGENIC CARDIOMYOPATHY 95

97 and provide the continuity between myofibrils of these neighbouring cells. We hypothesize that the presence of missense variants leads to increased proteolysis of plectin in the heart (similar to what occurs in the skin), which could in turn decrease its interjunctional linking capacity, make the adherens junctions less supported by desmosomes and thus more sensitive to stress and damage. This could explain why mutations in PLEC do not seem to be involved in a monogenic, classical Mendelian form of ARVC, since the intact, strong desmosomes could still keep the cardiomyocytes aligned together. Thus, an additional desmosomal mutation or external stressor may be necessary to pass the thresholds for disease development. This would also fit the concept of an oligogenic/multifactorial disease mechanism of ARVC that has been suggested by Marcus et al (2013) and Perrin et al, and would better explain the low penetrance observed in ARVC. The pathogenicity of previously identified genetic variants is increasingly being questioned because they are found in healthy individuals in much larger percentage than would be expected based on the estimated 1:2000 incidence and late-onset nature of the disease. Even radical mutations can be found in 0.5% of ostensibly healthy controls, while missense mutations were identified in nearly 16% of controls that were screened for just the five prominent ARVC genes (Kapplinger et al). These observations would also fit in an oligogenic disease model. In this study, we have shown that about half of the patients carrying a PLEC missense variant in the pathogenic region of the rod domain are also carriers of other definitive mutations in known ARVC genes. Certain novel missense variants identified in our ARVC patient cohort have been identified in EBS patients as well, but without any obvious overlap in clinical phenotypes. For instance, variants R1963W and p.v4399i were identified in both EBS and ARVC patients, but absent from GoNL (EBS patients; unpublished results). Moreover, the mutation p.r2110w (in dermatological context reported as p.r2000w) that was detected in seven EBS index patients (Bolling et al 2013, Kiritsi et al) affects the same amino acid as the likely pathogenic p.r2110q variant in one of our ARVC patients. Surprisingly, however, our ARVC patients exhibited no striking skin blistering disorder or muscular dystrophy, while the respective EBS patients did not show ARVC features. It is currently unclear what could cause some patients with missense variants to develop either mild EBS or ARVC, although one might expect that the stochastic distribution of wt/wt, wt/vous and VOUS/ VOUS plectin dimers, as well as the tissue-specific factors influencing the 96 SANGER SEQUENCING

98 calpain-mediated proteolysis, could to some extent explain this phenotypic variability. It is also possible that some EBS patients are suffering from an undiscovered cardiomyopathy. Another explanation is that genetic modifiers or a multifactorial background might also influence the phenotype of EBS (Padalon-Brauch et al), in which case PLEC missense mutations are not the sole cause of the disease, although they still contribute substantially to the phenotype. Importantly, EBS has much younger age of onset (and is more easily diagnosed) than ARVC, and EBS patients might still have the chance of developing cardiomyopathy later in life. For this reason, EBS patients (at least those with an identified genetic variant of PLEC) might benefit from cardiological evaluation and/or regular follow-up (Bolling et al 2010). The referral of ARVC patients to dermatologists, on the contrary, does not seem to be essential based on our current knowledge. The exact mechanical role of plectin in different tissues, as well as the mechanism via which truncating and missense mutations cause different disease phenotypes, however, requires further in-depth research in the future. There are a few examples of patients and families suffering from a combination of skin blistering and cardiac phenotypes due to PLEC mutations (Schröder et al, Bolling et al 2010). One patient had a homozygous frameshift mutation in PLEC and exhibited EBS-MD with cardiac hypertrophy, while another was compound heterozygous for a truncating mutation (p.e1724x) and a missense variant (p.r433q). Though this missense variant was detected in 3 Dutch and 13 UK patients in this study, it was also present in relatively high frequency in GoNL controls (1.36%) and thus classified as likely benign. However, in the patient carrying both PLEC variants, the p.r433q variant is most likely the predominant form produced due to diminished protein expression from the allele carrying the nonsense mutation, and therefore it might still contribute to the cardiac pathology in this patient. In addition to these two previous case reports, we recently also observed cardiac involvement in two brothers affected by severe EBS-MD carrying double homozygous PLEC variants (p.e1914x and p.y2967c) (Koss-Harnes et al). Since then, both patients died due to hypertrophic dilated cardiomyopathy. Additionally, one of their non-ebs brothers suffered from sudden cardiac death, a well-known feature of ARVC, and his son was affected by DCM without any signs of skin blistering or muscular dystrophy. Unfortunately, the PLEC carriership status of these family members is unknown. Villa et al have also just reported a case of EBS-MD developing left ventricular non-compaction cardiomyopathy. CHAPTER 2 PLEC IN ARRHYTHMOGENIC CARDIOMYOPATHY 97

99 Taken together, these findings suggest that it might be advisable to perform genetic screening of PLEC in other types of cardiomyopathy as well. The exact mechanical role of plectin in different tissues, as well as the mechanism via which truncating and missense mutations cause different disease phenotypes requires in-depth research in the future. CONCLUSIONS We identified 96 novel or low frequency (<2%), mostly heterozygous, missense variants of PLEC in ARVC patients. The facts that these variants were identified in addition to pathogenic desmosomal mutations in a subset of these patients, that co-segregation with disease could not be established for those variants that were analysed for this study, and that novel or rare variants with predicted pathogenic nature were also identified in the healthy GoNL cohort led us to conclude that PLEC cannot be considered the primary genetic cause of inherited ARVC. By comparing the natural genetic variation of the gene, by collecting all variants with an allele frequency <2% identified in the GoNL population with that found in patients, we were able to identify a region of probably ARVC-associated missense variants in the rod domain, which is thought to mediate homodimerization/dimerization of the protein and might become more vulnerable to proteolysis. We hypothesize that genetic variations in this domain of PLEC may make the cellular junctions more fragile, thus increasing the susceptibility to ARVC. This result underscores the previously suggested multifactorial nature of ARVC and suggests that PLEC variations, at least when present in specific regions of the protein, contribute to the number of risk factors to reach the threshold levels needed to initiate the development of this disease. ACKNOWLEDGEMENTS The authors would like to acknowledge Jackie Senior and Kate Mc Intyre for editing this manuscript, as well as Rudi Alberts for designing the script for the GoNL, 1000 Genomes and ESP allele frequency annotation of the genetic variants. Part of this work was undertaken at University College London Hospital and University College London (UCLH/UCL), which received some funding from the UK Department of Health s NIHR Biomedical Research Centres funding scheme. This study made use of data generated by the Genome of the Netherlands Project. 98 SANGER SEQUENCING

100 REFERENCES Andrä K, Lassmann H, Bittner R et al. Targeted inactivation of plectin reveals essential function in maintaining the integrity of skin, muscle, and heart cytoarchitecture. Genes Dev 1997;11: Bao JR, Wang JZ, Yao Y et al. Screening of pathogenic genes in Chinese patients with arrhythmogenic right ventricular cardiomyopathy. Chin Med J (Engl) 2013;126: Bauce B, Nava A, Beffagna G et al. Multiple mutations in desmosomal proteins encoding genes in arrhythmogenic right ventricular cardiomyopathy/dysplasia. Heart Rhythm 2010;7:22-9 Bolling MC, Jongbloed JDH, Boven LG et al. Plectin mutations underlie epidermolysis bullosa simplex in 8% of patients. J Invest Dermatol 2013; doi: /jid Bolling MC, Pas HH, de Visser M et al. PLEC1 mutations underlie adult-onset dilated cardiomyopathy in epidermolysis bullosa simplex with muscular dystrophy. J Invest Dermatol 2010;130: Elliott P, Andersson B, Arbustini E et al. Classification of the cardiomyopathies: a position statement from the European Cociety of Cardiology Working Group on Myocardial and Pericardial Diseases. Eur Heart J 2008;29:270-6 Fuchs P, Zörer M, Rezniczek GA et al. Unusual 5 transcript complexity of plectin isoforms: novel tissue-specific exons modulate actin binding activity. Hum Mol Genet 1999;8: Grumbach IM, Heim A, Vonhof S et al. Coxsackievirus genome in myocardium of patients with arrhythmogenic right ventricular dysplasia/ cardiomyopathy. Cardiology 1998;89:241-5 Jacob KA, Noorman M, Cox MGPJ et al. Geographical distribution of plakophilin-2 mutation prevalence in patients with arrhythmogenic cardiomyopathy. Neth Heart J 2012;20:234-9 Kapplinger JD, Landstrom AP, Salisbury BA et al. Distinguishing arrhythmogenic right ventricular cardiomyopathy/dysplasia-associated mutations from background genetic noise. J Am Coll Cardiol 2011;57: Kiritsi D, Pigors M, Tantcheva-Poor I et al. Epidermolysis bullosa simplex ogna revisited. J Invest Dermatol 2013;133:270-3 Konieczny P, Fuchs P, Reipert S et al. Myofiber integrity depends on desmin network targeting to Z-disks and costameres via distinct plectin isoforms. J Cell Biol 2008;181: Koss-Harnes D, Hoyheim B, Jonkman MF et al. Lifelong course and molecular characterization of the original Dutch family with epidermolysis bullosa simplex with muscular dystrophy due to a homozygous novel plectin point mutation. Acta Derm Venereol 2004;84: Lopes LR, Zekavati A, Syrris P et al. Genetic complexity in hypertrophic cardiomyopathy revealed by high-throughput sequencing. J Med Genet 2013;50(4): Marcus FI, McKenna WJ, Sherrill D et al. Diagnosis of arrhythmogenic right ventricular cardiomyopathy/dysplasia: proposed modification of the task force criteria. Eur Heart J 2010;31: Marcus FI, Edson S, Towbin JA: Genetics of arrhythmogenic right ventricular cardiomyopathy: a practical guide for physicians. J Am Coll Cardiol 2013;61: McKoy G, Protonotarios N, Crosby A et al. Identification of a deletion in plakoglobin in arrhythmogenic right ventricular cardiomyopathy with palmoplantar keratoderma and woolly hair (Naxos disease). Lancet 2000;355: McLean WH, Pulkkinen L, Smith FJ et al. Loss of plectin causes epidermolysis bullosa with muscular dystrophy: cdna cloning and genomic organization. Genes Dev 1996;10: Nakajima T, Kaneko Y, Irie T et al. Compound and digenic heterozygosity in desmosome genes as a cause of arrhythmogenic right ventricular cardiomyopathy in Japanese patients. Circ J 2012;76: Norgett EE, Hatsell SJ, Carvajal-Huerta L et al. Recessive mutation in desmoplakin disrupts desmoplakin intermediate filament interactions and causes dilated cardiomyopathy, woolly hair and keratoderma. Hum Mol Genet 2000;9(18): Padalon-Brauch G, Ben Amiati D, Vodo D et al. Digenic inheritance in epidermolysis bullosa simplex. J Invest Dermatol 2012;132(12): Perrin MJ, Angaran P, Laksman Z et al. Exercise testing in asymptomatic gene carriers exposes a latent electrical substrate of arrhythmogenic right ventricular cardiomyopathy. J Am Coll Cardiol 2013;62: Rigato I, Bauce B, Rampazzo A et al. Compound and digenic heterozygosity predicts life-time arrhythmic outcome and sudden cardiac death in desmosomal gene-related arrhythmogenic right ventricular cardiomyopathy. Circ Cardiovasc Genet 2013;6: Saberniak J, Hasselberg NE, Borqquist R et al. Vigorous physical activity impairs myocardial function in patients with arrhythmogenic CHAPTER 2 PLEC IN ARRHYTHMOGENIC CARDIOMYOPATHY 99

101 right ventricular cardiomyopathy and in mutation positive family members. Eur J Heart Fail 2014;16: Schröder R, Kunz WS, Rouan F et al. Disorganization of the desmin cytoskeleton and mitochondrial dysfunction in plectin-related epidermolysis bullosa simplex with muscular dystropyhy. J Neuropathol Exp Neurol 2002;61(6): Sen-Chowdhry S, Morgan RD, Chambers JC et al: Arrhythmogenic cardiomyopathy: etiology, diagnosis, and treatment. Annu Rev Med 2010;61: Smith FJ, Eady RA, Leigh IM et al. Plectin deficiency results in muscular dystrophy with epidermolysis bullosa. Nat Genet 1996;13:450-7 Sonnenberg A & Liem RKH: Plakins in development and disease. Exp Cell Res 2007;313: Teekakirikul P, Kelly MA, Rehm HL et al. Inherited cardiomyopathies: molecular genetics and clinical genetic testing in the postgenomic era. J Mol Diagn 2013;15: te Rijdt WP, Jongbloed JDH, de Boer RA et al. Clinical utility gene card for: arrhythmogenic right ventricular cardiomyopathy (ARVC) Eur J Hum Genet 2013; doi: /ejhg Villa CR, Ryan TD, Collins JJ et al. Left ventricular non-compaction cardiomyopathy associated with epidermolysis bullosa simplex with muscular dystrophy and PLEC1 mutation. Neuromuscular Disord 2015;25:165-8 Walko G, Vukasinovic N, Gross K et al. Targeted proteolysis of plectin isoform 1a accounts for hemidesmosome dysfunction in mice mimicking the dominant skin blistering disease EBS- Ogna. PLoS Genet 2011;7(12):e Wiche G, Krepler R, Artlieb U et al. Occurrence and immunolocalization of plectin in tissues. J Cell Biol 1983;97: Winter L & Wiche G: The many faces of plectin and plectinopathies; pathology and mechanisms. Acta Neuropathol 2013;125(1):77-93 Zernig G & Wiche G: Morphological integrity of single adult cardiac myocytes isolated by collagenase treatment: immunolocalization of tubulin, microtubule-associated proteins 1 and 2, plectin, vimentin, and vinculin. Eur J Cell Biol 1985;38(1): van Tintelen JP, Entius MM, Bhuiyan ZA et al. Plakophilin-2 mutations are the major determinant of familial arrhythmogenic right ventricular dysplasia/cardiomyopathy. Circulation 2006;113: Xu T, Yang Z, Vatta M et al. Compound and digenic heterozygosity contributes to arrhythmogenic right ventricular cardiomyopathy. J Am Coll Cardiol 2010;55: SANGER SEQUENCING

102 Supplementary table 1: Classification of PLEC variants Abbreviations ESP (EA; AA): allele frequency in European Ancestry and African Ancestry individuals in the Exome Sequencing Project database; GoNL: allele counts in the GoNL database; n.a.: not applicable/available VARIANT PREDICTED EFFECT CONTROL DATABASE FREQUENCIES CLASSIFICATION coordinates Grantham distance conservation (up to...) PhyloP [-14.1;6.4] AGVGD PolyPhen SIFT MutationTaster splicing 1000 Genomes c.30g>c, D10E 45 weak (cow) 0.53 C0 benign deleterious polymorphism - 0/2184 0/~ % LIKELY BENIGN deleterious disease causing - 0/2184 0/~ % VOUS c.266g>a, P89L 98 weak (cow) 5.21 C0 probably damaging c.305c>t, R102H 29 weak (cow) 2.38 C0 probably deleterious disease causing - 0/2184 4/8296; 1/4116 0,00% VOUS damaging c.421g>a, R141C 180 weak (cow) 4.24 C0 probably deleterious disease causing - 0/2184 0/~ ,10% VOUS damaging c.503g>a, P168L 98 weak (horse) 0.29 C0 benign tolerated disease causing - 0/2184 0; 6/3732 0,00% LIKELY BENIGN c.20c>t, A7V 64 weak (region) 0.37 C0 benign tolerated polymorphism - 0/2184 0/~ % LIKELY BENIGN (NM_ ) c.124g>c, D42H 81 high (zebrafish) 1.17 C0 possibly deleterious disease causing strong effect 0/2184 0/~ % VOUS (NM_ ) damaging (but large exon) c.28c>t, Q10* n.a. weak (region) n.a. n.a. n.a. n.a. n.a. - 0/2184 0/~ % VOUS (NM_ ) c.69_70instac, n.a. weak n.a. n.a. n.a. n.a. n.a. - 0/2184 0/~ % VOUS D23_N24insY (NM_ ) c.743c>t, R248Q 43 weak (cow) 1,82 C0 benign deleterious disease causing minor effect 0/2184 0/~ ,00% VOUS c.947c>t, R316Q 43 high (zebrafish) 1,58 C0 benign deleterious disease causing - 0/2184 0/~ ,10% VOUS c.1204c>t, V402M 21 high (zebrafish) 5,37 C0 possibly deleterious disease causing - 0/2184 0; 1/4212 0,20% VOUS damaging c.1265g>a, R422Q 43 weak (cow) 1.42 C0 benign tolerated polymorphism - 7/2184 1/8410; 57/ % LIKELY BENIGN c.1298c>t, R433Q 43 high (zebrafish) 5.37 C0 probably deleterious disease causing - 13/ /8466; 14/ % LIKELY BENIGN damaging c.1534c>t, V512M 21 high (zebrafish) 4,08 C0 possibly deleterious disease causing - 0/2184 2/8416; 0 0,20% VOUS damaging c.1569g>c, S523R 110 high (zebrafish) 0.37 C65 benign deleterious disease causing - 0/2184 9/8342; 0 0.2% VOUS c.1634c>t, C545Y 194 high (zebrafish) 5,61 C65 probably deleterious disease causing - 0/2184 0/~ ,00% LIKELY PATHOGENIC damaging c.1745g>a, A582V 64 high (zebrafish) 2,87 C65 benign deleterious disease causing minor effect 0/2184 0/~ ,10% LIKELY PATHOGENIC c.1836c>g, Y612* n.a. high (region) n.a. n.a. n.a. n.a. n.a. - 0/2184 0/~ % LIKELY PATHOGENIC c.1874a>g, Y625C 194 high (Xenopus) 2.06 C55 probably damaging ESP (EA; AA) GoNL deleterious disease causing - 0/2184 0/~ % LIKELY PATHOGENIC c.1917c>g, Q639H 24 weak 1,42 C0 benign tolerated polymorphism minor effect 0/2184 0/~ ,10% LIKELY BENIGN CHAPTER 2 PLEC IN ARRHYTHMOGENIC CARDIOMYOPATHY 101

103 VARIANT PREDICTED EFFECT CONTROL DATABASE FREQUENCIES CLASSIFICATION coordinates Grantham distance conservation (up to...) PhyloP [-14.1;6.4] c.2474g>t, P825Q 76 high (zebrafish) 5.13 C65 probably damaging c.2549c>g, C850S 112 high (zebrafish) 5.13 C65 probably damaging AGVGD PolyPhen SIFT MutationTaster splicing 1000 Genomes ESP (EA; AA) GoNL deleterious disease causing - 0/2184 0/~ % LIKELY PATHOGENIC deleterious disease causing - 0/2184 0/~ % LIKELY PATHOGENIC c.2959c>t, A987T 58 weak (cow) 0,53 C0 benign tolerated polymorphism - 77/2184 3/8370; 442/4402 0,00% BENIGN c.2962c>t, V988M 21 high (zebrafish) 2.14 C15 benign deleterious disease causing - 0/2184 0; 13/ % VOUS c.3130g>a, R1044C 180 weak (cow) 2,22 C25 possibly damaging deleterious disease causing - 0/2184 3/8480; 0 0,20% VOUS c.3352g>a, R1118C 180 weak (cow) 0.21 C25 benign tolerated polymorphism - 0/2184 1/8494; 0 0.0% LIKELY BENIGN c.3872c>t, R1291Q 43 weak (cow) 2.63 C0 benign tolerated polymorphism - 0/2184 0/~ % LIKELY BENIGN c.4093g>a, R1365W 101 weak 0.93 C15 benign deleterious polymorphism minor effect 9/ /8098; 2/ % LIKELY BENIGN c.4189g>c, R1397G 125 weak (cow) 1,01 C45 benign deleterious disease causing - 9/2184 0; 23/4302 0,00% LIKELY BENIGN c.4280t>c, K1427R 26 high (zebrafish) 2.22 C25 benign deleterious disease causing - 1/ /8436; 44/ % LIKELY BENIGN c.4400g>c, T1467R 71 weak (cow) 5,61 C15 possibly damaging c.4439c>t, R1480H 29 high (zebrafish) 5,61 C25 possibly damaging c.4486g>a, R1496C 180 high (zebrafish) 0,69 C65 possibly damaging deleterious disease causing - 0/2184 0/~ ,20% VOUS deleterious disease causing - 0/2184 1/8548; 0 0,10% LIKELY PATHOGENIC deleterious disease causing - 0/2184 0/~ ,40% LIKELY PATHOGENIC c.4637t>g, V1546G 109 high (zebrafish) 2,06 C65 benign deleterious disease causing - 0/2184 0/~ ,00% LIKELY PATHOGENIC c.4777c>t, V1593M 21 weak (cow) 0.85 C0 benign deleterious polymorphism minor effect 6/ /7890; 4/ % LIKELY BENIGN c.4937g>a, R1646H 29 moderately (only R or K) 2.30 C0 benign deleterious disease causing minor effect 0/ % VOUS c.4943c>t, R1648Q 43 high (zebrafish) 0,77 C35 benign deleterious disease causing - 0/2184 0/~ ,20% VOUS c.4967c>t, S1656L 145 weak 1.74 C0 benign tolerated polymorphism - 6/ % LIKELY BENIGN c.5009c>t, R1670Q 43 moderate 2.06 C0 benign deleterious polymorphism - 0/ /6166; 2/ % LIKELY BENIGN c.5054g>a, A1685V 64 high (zebrafish) 5,45 C65 probably deleterious disease causing - 0/2184 0/~ ,20% LIKELY PATHOGENIC damaging c.5062g>a, R1688C 180 weak (cow) 1,74 C45 possibly deleterious disease causing - 10/2184 0/~ ,00% VOUS damaging c.5240g>a, A1747V 64 high (zebrafish) 5,53 C65 possibly deleterious disease causing - 0/2184 0/~ ,00% LIKELY PATHOGENIC damaging c.5284c>t, E1762K 56 high (zebrafish) 5,53 C65 benign deleterious disease causing - 1/2184 0/~ ,00% LIKELY PATHOGENIC c.5576c>t, T1859M 81 weak (cow) 3.19 C0 possibly tolerated disease causing - 0/2184 1/6312; 22/ % VOUS damaging c.5623c>t, R1875W 101 high (zebrafish) 2.63 C65 possibly deleterious disease causing - 0/2184 0/~ % LIKELY PATHOGENIC damaging 102 SANGER SEQUENCING

104 c.5651g>a, T1884M 81 weak (cow) 0,93 C0 probably damaging deleterious polymorphism - 0/2184 0/~ ,00% VOUS c.5726g>a, A1909V 64 weak (cow) 1,82 C0 possibly deleterious polymorphism minor effect 0/2184 0/~ ,00% VOUS damaging c.5801g>a, R1934H 29 high (zebrafish) 1.82 C25 probably deleterious disease causing - 0/2184 0/~ % LIKELY PATHOGENIC damaging c.5860c>g, R1954G 125 high (zebrafish) 0.85 C65 probably deleterious disease causing - 0/2184 0/~ % LIKELY PATHOGENIC damaging c.5882g>a, A1961V 64 high (zebrafish) 2,71 C65 possibly deleterious disease causing - 0/2184 0/~ ,00% LIKELY PATHOGENIC damaging c.5887g>a, R1963W 101 high (zebrafish) 1.90 C65 probably deleterious disease causing - 0/2184 0/~ % LIKELY PATHOGENIC damaging c.5888g>a, R1963Q 43 high (zebrafish) 2.06 C35 possibly deleterious disease causing - 19/218 35/4580; 4/ % VOUS damaging c.5914c>g, L1972V 32 high (zebrafish) 5.37 C25 probably deleterious disease causing - 0/ % LIKELY PATHOGENIC damaging c.5920g>a, E1974K 56 high (zebrafish) 5.37 C55 probably deleterious disease causing - 0/2184 0/~ % LIKELY PATHOGENIC damaging c.5932g>a, A1978T 58 high (zebrafish) 2.71 C55 possibly deleterious disease causing - 1/2184 1/6216; 6/ % LIKELY PATHOGENIC damaging c.6109a>g, K2037E 56 high (zebrafish) 1.82 C55 probably deleterious disease causing - 0/ % LIKELY PATHOGENIC damaging c.6329g>a, R2110Q 43 high (zebrafish) 1.90 C35 benign deleterious disease causing - 0/ % LIKELY PATHOGENIC c.6431g>a, A2144V 64 high (zebrafish) 5,21 C0 probably deleterious disease causing minor effect 1/2184 0/~ ,00% LIKELY PATHOGENIC damaging c.6470t>g, E2157A 107 high (zebrafish) 4,32 C0 benign tolerated disease causing - 0/2184 1/7912; 1/3826 0,00% VOUS c.6473g>a, A2158V 64 high (zebrafish) 5.21 C0 possibly damaging deleterious disease causing strong effect (but large exon) 0/2184 1/7880; 0 0.1% LIKELY PATHOGENIC c.6609c>a, Q2203H 24 weak (cow) 0,45 C0 benign tolerated polymorphism - 10/2184 0/7114; 13/3336 0,00% LIKELY BENIGN c.6722g>a, A2241V 64 high (zebrafish) 3,03 C0 benign deleterious disease causing - 5/2184 2/7676; 3/3563 0,60% VOUS c.6736g>a, R2246W 101 weak (dog) 1,82 C25 benign deleterious polymorphism minor effect 0/2184 0/~ ,00% VOUS c.6743a>c, V2248G 109 weak (cow) 2,3 C0 benign deleterious disease causing - 0/2184 0/~ ,70% VOUS c.6827g>a, R2276H 29 high (zebrafish) 4.00 C25 probably deleterious disease causing - 0/2184 1/8102; 0 0.0% LIKELY PATHOGENIC damaging c.6857c>t, R2286Q 43 high (zebrafish) 5,77 C35 probably deleterious disease causing - 0/2184 0/~ ,00% LIKELY PATHOGENIC damaging c.6931g>a, A2311T 58 high (zebrafish) 2.14 C0 benign deleterious disease causing - 0/ % VOUS CHAPTER 2 PLEC IN ARRHYTHMOGENIC CARDIOMYOPATHY 103

105 VARIANT PREDICTED EFFECT CONTROL DATABASE FREQUENCIES CLASSIFICATION coordinates Grantham distance conservation (up to...) PhyloP [-14.1;6.4] AGVGD PolyPhen SIFT MutationTaster splicing 1000 Genomes c.6947c>t, R2316Q 43 high (frog) 3,43 C0 benign deleterious disease causing - 7/2184 6/8470; 1/4256 0,00% VOUS ESP (EA; AA) GoNL c.7256g>a, R2419Q 43 high (zebrafish) 4.08 C35 possibly damaging deleterious disease causing - 0/ % LIKELY PATHOGENIC c.7678c>t, A2560T 58 weak (cow) 0.53 C0 benign tolerated polymorphism - 5/ /8568; 9/ % LIKELY BENIGN c.8080g>a, R2694W 101 weak (mouse) 2,06 C0 benign deleterious polymorphism minor effect 0/2184 4/8198; 1/4296 0,40% LIKELY BENIGN c.8231c>g, A2744G 60 weak (cow) 2.30 C0 benign tolerated polymorphism - 0/ % LIKELY BENIGN c.8423c>t, R2808Q 43 weak (cow) 1,5 C0 benign deleterious disease causing - 23/ /8494; 131/4296 0,90% LIKELY BENIGN c.8452c>t, E2818K 56 high (zebrafish) 3.68 C55 benign deleterious disease causing - 0/2184 3/8524; 0 0.0% LIKELY PATHOGENIC c.8458c>t, A2820T 58 weak (lemur) C0 benign tolerated polymorphism - 0/2184 0/~ ,10% LIKELY BENIGN c.8462c>t, R2821Q 43 weak (lemur) 0,12 C0 benign tolerated polymorphism - 2/ /8518; 4/4312 0,30% LIKELY BENIGN c.8539t>c, I2865V 29 weak (cow) 0,12 C0 benign tolerated polymorphism - 0/2184 6/8380; 1/4152 0,00% LIKELY BENIGN c.8612g>a, A2871V 64 high (zebrafish) 5.45 C65 probably damaging deleterious disease causing - 1/ /8348; 3/ % VOUS c.8744t>c, K2915R 26 high (zebrafish) C25 benign deleterious polymorphism minor effect 0/2184 0/~ ,00% VOUS c.8881c>t, V2961M 21 high (zebrafish) C0 benign deleterious polymorphism - 0/2184 0/~ ,20% VOUS c.8900t>c, Y2967C 194 weak (cow) 1.58 C0 benign tolerated disease causing - 9/ /8272; 10/ % LIKELY BENIGN c.8917c>t, D2973N 23 high (zebrafish) 3.11 C15 benign deleterious disease causing - 0/ /8328; 3/ % VOUS c.8920c>t, E2974K 56 moderate 1.66 C55 benign deleterious polymorphism - 0/ /8334; 0 0.5% VOUS (zebrafish) c.8923c>t, E2975K 56 moderate 4.16 C15 benign deleterious disease causing - 0/2184 4/8332; 0 0.0% VOUS c.8941c>t, A2981T 58 weak C0 benign tolerated polymorphism - 0/ % LIKELY BENIGN c.9227c>t, R3076Q 43 weak (cow) 1,98 C0 benign deleterious disease causing - 0/2184 1/8538; 1/4330 0,10% VOUS c.9231g>c, D3077E 45 high (zebrafish) 0,29 C35 benign deleterious polymorphism - 117/ /8532; 927/4330 0,90% BENIGN c.9388c>t, D3130N 23 weak (lemur) 0,04 C0 benign tolerated polymorphism - 0/2184 1/8206; 0 0,10% LIKELY BENIGN c.9445c>t, E3149K 56 high (zebrafish) 2,47 C55 benign deleterious disease causing - 4/ /8308; 2/4118 0,00% VOUS c.9454t>c, T3152A 58 weak C0 benign tolerated polymorphism - 0/ % LIKELY BENIGN c.9464c>t, R3155Q 43 weak C0 benign tolerated polymorphism minor effect 1/2184 0/~ ,10% LIKELY BENIGN c.9958c>t, D3320N 23 high (zebrafish) 2,55 C0 benign tolerated disease causing - 0/2184 0/~ ,20% VOUS c.10004c>t, R3335Q 43 weak (mouse) C0 benign tolerated polymorphism - 0/2184 0/~ ,00% LIKELY BENIGN c.10054t>c, K3352E 56 weak (cow) C0 benign tolerated polymorphism minor effect 0/2184 0/~ ,00% LIKELY BENIGN c.10102g>a, R3368C 180 weak (cow) 1,25 C55 benign deleterious polymorphism - 0/2184 0/~ ,00% VOUS c.10225g>a, R3409C 180 high (zebrafish) 1,17 C0 possibly damaging c.10336g>a, R3446C 180 high (zebrafish) 1.09 C65 probably damaging c.10372c>t, G3458R 125 weak C0 benign tolerated polymorphism strong effect (but large exon) tolerated disease causing - 0/2184 2/8540; 0 0,00% VOUS deleterious disease causing - 7/ /8412; 12/ % VOUS 11/ /8338; 7/ % LIKELY BENIGN 104 SANGER SEQUENCING

106 c.10454c>t, R3485Q 43 weak (cow) 0,53 C0 benign deleterious polymorphism - 0/2184 0/~ ,00% LIKELY BENIGN c.10469g>c, G3490A 60 high (zebrafish) 5.86 C55 possibly deleterious disease causing - 21/ /8390; 8/ % LIKELY BENIGN damaging c.10541c>t, R3514Q 43 weak (cow) 2,55 C0 benign tolerated disease causing - 0/2184 1/8291; 0 0,00% LIKELY BENIGN c.10747a>g, S3583P 74 high (zebrafish) 4,81 C65 probably deleterious disease causing - 0/2184 0/~ ,00% LIKELY PATHOGENIC damaging c.10909g>a, R3637C 180 weak (cow) 0.93 C35 possibly damaging deleterious disease causing - 1/ /8372; 0 0.1% VOUS c.11056c>a, A3686S 99 weak (mouse) C0 benign tolerated polymorphism - 0/2184 0/~ ,10% LIKELY BENIGN c.11158a>g, S3720P 74 weak (cow) 0,12 C0 benign deleterious polymorphism - 0/2184 0/~ ,00% LIKELY BENIGN c.11281c>t, E3761K 56 high (zebrafish) 5,29 C55 probably deleterious disease causing - 0/2184 0/~ ,10% LIKELY PATHOGENIC damaging c.11324g>a, A3775V 64 weak (cow) 0,77 C0 benign tolerated polymorphism - 0/2184 0; 1/4360 0,10% LIKELY BENIGN c.11438g>a, A3813V 64 weak 0,53 C0 benign tolerated polymorphism - 0/2184 2/8272; 0/3880 0,00% LIKELY BENIGN c.11447g>a, A3816V 64 weak (lemur) 0,61 C0 benign tolerated polymorphism minor effect 0/2184 4/8290; 0 0,00% LIKELY BENIGN c.11740c>t, E3914K 56 weak (cow) 3,51 C0 benign tolerated disease causing - 0/2184 0/~ ,30% LIKELY BENIGN c.11762t>a, Q3921L 113 high (zebrafish) 4,24 C65 probably deleterious disease causing - 0/2184 2/8376; 0 0,00% LIKELY PATHOGENIC damaging c.12010c>g, D4004H 81 high (zebrafish) 1,74 C0 benign tolerated polymorphism - 82/2184 2/8026; 408/3760 0,00% BENIGN c.12022c>t, G4008S 56 high (zebrafish) 2,47 C55 benign deleterious disease causing - 0/2184 7/7656; 3/3544 0,30% VOUS c.12131g>a, T4044M 81 high (zebrafish) 2.79 C65 possibly deleterious disease causing - 18/ /8382; 3/ % LIKELY BENIGN damaging c.12437c>t, R4146H 29 high (zebrafish) 5,86 C25 probably damaging c.12442c>t, V4148I 29 high (zebrafish) 5,86 C25 probably damaging c.12598c>t, V4200M 21 high (zebrafish) 5,86 C15 probably damaging c.12601c>t, E4201K 56 high (zebrafish) 5,86 C15 possibly damaging c.12616g>a, R4206C 180 high (zebrafish) 1,98 C65 probably damaging c.12655c>t, D4219N 23 high (zebrafish) 5,86 C15 probably damaging deleterious disease causing - 0/2184 0/~ ,10% LIKELY PATHOGENIC deleterious disease causing - 0/2184 0/~ ,10% LIKELY PATHOGENIC deleterious disease causing - 0/2184 0/8362; 1/4056 0,20% LIKELY PATHOGENIC deleterious disease causing - 0/2184 4/8368; 1/4073 0,20% LIKELY PATHOGENIC deleterious disease causing minor effect 0/2184 0/~ ,00% LIKELY PATHOGENIC deleterious disease causing - 0/2184 0/~ ,00% LIKELY PATHOGENIC c.13129c>t, A4377T 58 weak (cow) 0,93 C0 benign tolerated polymorphism - 0/2184 6/8446; 0 0,00% LIKELY BENIGN c.13195g>a, V4399I 29 weak 0.69 CC0 benign tolerated polymorphism - 12/2184 0/8330; 2/ % LIKELY BENIGN c.13228g>a, P4410S 74 high (zebrafish) 5.86 C65 probably deleterious disease causing - 1/2184 0/8304; 13/ % LIKELY PATHOGENIC damaging c.13885c>t, G4629S 56 high (zebrafish) 4.00 C55 benign deleterious disease causing minor effect 0/2184 0/~ ,00% LIKELY PATHOGENIC deleterious disease causing - 0/2184 0/~ ,00% LIKELY PATHOGENIC c.13999g>a, R4667C 180 high (zebrafish) 2,79 C65 probably damaging CHAPTER 2 PLEC IN ARRHYTHMOGENIC CARDIOMYOPATHY 105

107

108 CHAPTER 3 EXOME SEQUENCING

109

110 Chapter 3.1 Hunting for novel disease genes in autosomal dominant cardiomyopathies: elucidating a role for the sarcomeric pathway Rowida Almomani*, Anna Posafalvi*, Paul A van der Zwaag, Carlo L Marcelis, Bert Baars, Johanna C Herkert, Rudolf A de Boer, Karin Y van Spaendonck-Zwarts, Maarten P van den Berg, Richard J Sinke, J Peter van Tintelen, Jan DH Jongbloed * The first two authors contributed equally The last two authors contributed equally

111 ABSTRACT We performed exome sequencing and a haplotype sharing test on a group of twelve families with autosomal dominant cardiomyopathy and no previous genetic diagnosis in order to identify potentially novel disease genes. Our approach resulted in the identification of the genetic cause of disease in 6/12 families. We found truncating variants in TTN in two dilated cardiomyopathy families, a frame-shift mutation in FLNC and a double missense mutation in FHL2 in two arrhythmogenic cardiomyopathy families, and missense variants in the COBL and STARD13 genes in two dilated cardiomyopathy families, respectively, both of which are genes that have not been related to cardiac pathology before. Thorough data-mining suggests a possible role for all of these genes in the disease mechanism of late onset cardiomyopathies. By creating a co-expression network of the five genes using an expression-arraybased bioinformatics database and software created in the department, we show that 100 of the 166 proteins included in our network have been annotated with a potential function in cardiac development and physiology. Of these 100 proteins, 28 are known as disease genes in various types of cardiomyopathy, and a role in sarcomere assembly seems to be the common molecular pathway connecting a large proportion of these genes.

112 INTRODUCTION Dilated cardiomyopathy (DCM) is a progressive heart disease mainly characterised by left ventricular dilatation and impaired cardiac contraction, while arrhythmogenic right ventricular cardiomyopathy (ARVC) is a common cause of sudden cardiac death because of its association with ventricular arrhythmias (Hershberger et al, Basso et al). Currently, there are more than 50 genes linked to the pathogenesis of familial DCM. In the pre-ngs era these genes could only explain up to 20% of Dutch DCM cases (25% in familial cases and 8% in sporadic cases) (van Spaendonck et al, 2013), while including screening of the titin (TTN) gene improved this to 45-50% (Wilde & Behr, Posafalvi et al, van Spaendonck et al 2014). Our gene-panel-based Next Generation Sequencing (NGS) method, which was recently implemented into routine DNA diagnostics, resulted in the identification of mutations and likely pathogenic variants in up to 55% of DCM index patients (see chapter 4.1). On the other hand, to date there are still only 13 ARVC genes known (te Rijdt et al). One of these is the desmosomal plakophilin 2 gene (PKP2), and mutations in this gene are the most frequent cause of familial ARVC, occurring in up to 70% of the patients (van Tintelen et al). Until recently, the yield of attempts to identify genetic mutations in ARVC patient cohorts via traditional sequencing was only ~50% (Cox et al, Quarta et al). In this study our aim was to identify the disease gene in families currently considered unsolved (without a known genetic factor potentially explaining the phenotype). For this purpose, we used exome sequencing (ES), i.e. sequencing of all protein-coding regions of the genome, to identify (potentially novel) disease genes in inherited cardiomyopathy patients/families. Since the inheritance pattern in the families studied was most likely to be autosomal dominant, and ES is well known to result in a huge number of heterozygous variants (potential mutations as well as benign variants), the data analysis was much more challenging than identifying the cause of the disease in a recessive form of the disease (such as in the rare cases of consanguinity). Hence, it was of special importance to narrow down the search for causal variants into chromosomal regions of particular interest. For this purpose, we combined ES with a haplotype sharing test (HST). HST has previously been shown to be a crucial step for successfully identifying regions carrying causative genes in cardiomyopathy families that are too small for classical linkage analysis (van der Zwaag et al). We applied HST as a filtering method during data analysis, and this helped us to prioritize the long list of genes containing heterozygous variants. CHAPTER 3.1 AUTOSOMAL DOMINANT CARDIOMYOPATHIES 111

113 Using this combined approach of ES and HST, we succeeded in identifying the disease gene or putative disease gene in six out of our twelve families with autosomal dominant cardiomyopathies. We identified five potential disease genes, of which three were novel, one had occasionally been associated with cardiomyopathies, and one was the known cardiomyopathy gene TTN, which was not routinely screened for at that time due to its enormous size. MATERIALS AND METHODS Patients Families were selected because multiple affected members were available for HST analysis and because, in all cases, previous Sanger sequencing approaches and, in most cases, gene-panel-based NGS had not resulted in the identification of a pathogenic mutation or likely pathogenic variant. Eleven families were recruited from the cardiomyopathy cohort of the University Medical Center Groningen, the Netherlands, and one family was recruited from the University Medical Center Nijmegen, the Netherlands. The DCM patients were diagnosed according to established clinical criteria (Mestroni et al). One family had ARVC fulfilling the task force criteria (TFC) (Marcus et al), and one family had five family members with suspected ARVC, but not yet fulfilling all of those criteria. Our approach included (1) for most families, pre-screening of patients using gene-panel-based NGS targeting 55 known cardiomyopathy genes, and subsequent selection of candidate patients/families (some families were analysed using our gene-panel-based approach during the course of this study); (2) HST of all available affected family members and subsequent data analysis; (3) ES of at least two family members who are as distantly related as possible; (4) identification of probable disease regions and genes; (5) confirmation and co-segregation analysis; (6) mutational screening of probable disease genes in additional patients; and (7) co-expression network analysis to obtain supportive evidence of pathogenicity. Targeted sequencing DNA samples isolated from peripheral blood of patients were sequenced for 55 known cardiomyopathy disease genes as formerly described by Sikkema- Raddatz et al and Posafalvi et al (manuscript in preparation, see also chapter 4.1). Data analysis was performed using the MiSeq reporter program (Illumina, San 112 EXOME SEQUENCING

114 Diego, CA, USA), Next Gene software (v2.2.1, Softgenetics, State College, PA, USA) and Cartagenia software (Cartagenia, Leuven, Belgium), as described (Sikkema-Raddatz et al, chapter 4.1). Haplotype sharing test To establish haplotypes and to identify possible shared haplotypes, single nucleotide polymorphism (SNP) genotyping of the DNA samples was performed using the Human 610-quad beadchip 610K SNP array (Illumina) according to the manufacturer s protocols. The data was analysed using Microsoft Office Excel 2007 (Microsoft, Redmond, WA, USA) software as previously described by van der Zwaag et al. The longest shared haplotypes (LSH) identified were used for ranking candidate variants in the last step of the exome sequencing data analysis. In this step we assume that the longer a shared region is between affected family members, the higher the chance that it contains the mutual causative mutation. In the cases in which the mutation identified was not localised in the 1 st LSH, we checked if those chromosomal regions which ranked better than the one carrying the mutation contained any cardiomyopathy candidate genes, and mutations in those genes were excluded. Additionally, the array data was also used for quality control purposes: we performed a concordance check between the genotyping and exome sequencing datasets to exclude potential sample-swaps during the experimental procedures. CHAPTER 3.1 Exome sequencing Exome sequencing was performed on Illumina HiSeq2000 sequencers in paired end mode and 100bp read lengths following sample preparations using SureSelect exome capture kit All Exon V4 or V5 (Agilent Technologies, Inc., Santa Clara, CA, USA) enrichment according to the manufacturer s protocol. The raw Fastq files were aligned by using bwa to the human reference genome (hg 19, NCBI build 37) (Li et al, 2009a), SAM/BAM files were manipulated by Samtools and Picard-1.57 (Li et al, 2009b). Then the Genome Analysis Toolkit was used to perform base quality score recalibration, duplicate removal and INDEL realignment (McKenna et al). The output vcf files were annotated by our in-house bioinformatics pipeline and SeattleSeq ( AUTOSOMAL DOMINANT CARDIOMYOPATHIES 113

115 Data analysis, filtering and prioritization After quality filtering of the data and checking the concordance of SNP calls from the genotyping and sequencing platforms, we used various, generally accepted, data filters in our analysis. These included filtering data for a minimal read depth, checking the allele balance (and only keeping heterozygous variant calls), and using a population frequency filter. From the remaining list of variants, we focused on those novel or rare coding variants that were shared among affected family members. At this step, we implemented both negative and positive filters. For instance, it is well known that olfactory receptor genes exhibit unusually high genetic variability between individuals (Waszak et al), hence those variants do not seem relevant in DCM (negative filter). On the other hand, we looked carefully at variants in genes which had been previously associated with cardiomyopathy or heart-specific phenotypes. For this purpose, we not only focussed on known cardiomyopathy genes, but also included genes known to show cardiac expression or found to be important for abnormal cardiomyocyte proliferation, or associated with a thin myocardial wall or other cardiac phenotypes in a heart-specific protein network built purely upon functional data (such as mouse models, yeast-twohybrid screening or other sources of experimental proteomics data, Lage et al, 2010) (positive filter). In addition, we performed thorough data-mining taking into consideration everything known about those genes that remained at the end of the analysis: their known function, their potential cardiac expression, and the existence of any pseudogenes. In parallel with this last step, the remaining variants were ranked according to their localization into one of the shared haplotypes of considerable size within the family and their putative pathogenicity (for details on variant classification, see also chapter 4.1; for a decision tree during our exome sequencing data analysis, see figure 1). Mutation screening Sanger sequencing was performed for validation of the ES results in the DNA of the index patients, for segregation analysis of the identified genetic variants (potential mutations) within families, as well as for screening in larger patient cohorts (where appropriate). Primer sequences are available upon request. In order to screen for additional mutation carriers of the COBL mutation c.998g>a; p.(arg333gln) identified in family 5, restriction digestion analysis 114 EXOME SEQUENCING

116 SEQUENCING DATA read depth (exclude nucleotides of coverage <20x) allele balance (homozygous or heterozygous) filtering FREQUENCY FAMILY in-house sample pool filter (exclude polymorphisms & artefacts) population frequency filter (dbsnp, 1000 Genomes) create list of variants shared between affected family members exclude variants from non-affected relatives FUNCTIONALITY keep only coding variants (frameshift, nonsense, missense) splice site variants CHAPTER 3.1 prioritization GENE SELECTION Does the gene fit? blacklisted genes to be neglected special attention given to known disease genes literature to determine: expression function pseudogenes pathways CLASSIFICATION Does the variant fit? physicochemical changes conservation affected domain frequency predicted pathogenicity GENETIC SUPPORT Does the family background fit? re-check the mode of inheritance combine with HST or homozygosity mapping results check co-segregation within the family Figure 1. Decision tree of exome sequencing data analysis Abbreviation: HST haplotype sharing test. AUTOSOMAL DOMINANT CARDIOMYOPATHIES 115

117 was performed on COBL PCR products using BseXI according to the protocols provided by the manufacturer (Thermo Scientific, Waltham, MA, USA) and results visualized by 1.5% agarose gel electrophoresis. Co-expression network An online software using publicly available data from expression array depositories has been created (see Results section). Approximately 80,000 human, mouse and rat microarrays archived in the Gene Expression Omnibus were subjected to principal component analysis. The resulting components hypothetically reflect on transcriptomes, which are often well conserved across species and are enriched for known biological phenomena. The software uses statistical methods on the combined, multi-species gene network built upon these components in order to predict biological functions of candidate genes. Furthermore, it is able to visualise the co-expression network of a set of genes of interest in Cytoscape. A detailed description of the method can be found elsewhere (Fehrmann et al, in press, Nat Genet). RESULTS & DISCUSSION Our approach of applying ES combined with HST as a filtering step to facilitate more focused variant prioritization during data analysis resulted in the identification of the potentially causative mutation in six out of twelve unsolved cardiomyopathy families, affecting five different genes in total. We describe our findings for these six families in detail below, with the pedigrees and haplotype sharing results of the respective families shown in figure 2. Six families in which the cause of the disease was identified Family 1, DCM: TTN c.82117c>t; p.(arg27373*) and Family 2, DCM: c.75607dela; p.(ser25203valfs*29), (NM_ ) We have identified a disease-causing variant, c.82117c>t; p.(arg27373*), in the TTN in family 1. The variant was a novel nonsense mutation encoded in the 3 rd longest shared haplotype (2q14.1q31.1), and was shown to co-segregate with the disease phenotype in all six affected family members for whom a DNA sample was available (this family is also described in chapter 4.2). Mutations of TPM1 (candidate gene localised in the 2 nd LSH) were excluded by Sanger sequencing. Likewise, in family 2 another novel TTN mutation, c.75607dela; p.(ser25203valfs*29), was identified in the two family members who were analysed by ES and in this case the 116 EXOME SEQUENCING

118 TTN gene was located in the second largest haplotype (2p11.2q33.1). No additional affected family members were available for this analysis. TTN is known to play a role in sarcomere assembly and stabilization (Granzier & Wang) and has been associated with heart failure (Hein et al) and cardiomyopathy (Gerull et al) for decades, but has not been extensively sequenced in patients due to its huge genomic size (TTN is the largest gene of the human genome with a length of ~0.3 Mbp). Currently, TTN is suggested to be involved in dilated, restrictive, hypertrophic, and arrhythmogenic right ventricular cardiomyopathies (Gerull et al, Peled et al, Satoh et al, Taylor et al) and was recently reported to carry truncating mutations in up to 25% of familial DCM cases (Herman et al). Furthermore, our group has performed functional analysis of the TTN isoform composition combined with single cardiomyocyte passive force measurements on another truncating TTN variant (p.lys15664valfs*13) recently identified in a peripartum cardiomyopathy patient (van Spaendonck-Zwarts et al, 2014). What we showed based on these analyses is that the physiological function of the sarcomeres was affected by the presence of the TTN variant. Due to the technical advances made in parallel with our initial exome sequencing studies, we have implemented a gene-panel-based NGS approach in order to analyse 55 known cardiomyopathy genes in the genome diagnostics laboratory of our department during the course of this project. This method is currently used as a routine screening step before applying exome sequencing on gene-panel negative patients only (see also Sikkema-Raddatz et al, chapter 4.1 of this thesis, and figure 1 in chapter 5). A remarkable advantage of this technical improvement is that all 363 exons of the TTN gene have now also been included in our targeted diagnostic approach. This resulted in the identification of TTN truncating mutations in up to 15% of criteria-positive DCM cases (see also chapter 4.1). CHAPTER 3.1 Family 3, ARVC: FHL2 c.698_699delinsaa; p.(gly233glu) (NM_ ) In this family, we identified a putative mutation in the four and a half LIM domains 2 gene (FHL2), which is known to be much more prominently expressed in the heart than in other organs (Chan et al). Even though FHL2 seems not to be required for the embryonic development of the heart and its full knock out in mice does not cause any cardiac phenotype up to 15 months of age (Chu et al), the stress of sustained β-adrenergic stimulation by soproterenol treatment lead to cardiac hypertrophy in these animals (Kong et al, 2001). AUTOSOMAL DOMINANT CARDIOMYOPATHIES 117

119 Family 1 A) B) 118 EXOME SEQUENCING

120 Family 2 A) B) CHAPTER 3.1 AUTOSOMAL DOMINANT CARDIOMYOPATHIES 119

121 Family 3 Family 4 A) B) 120 EXOME SEQUENCING

122 Family 5 A) B) CHAPTER 3.1 AUTOSOMAL DOMINANT CARDIOMYOPATHIES 121

123 Family 6 A) B) Figure 2. Pedigrees including results of co-segregation analyses (A) and HST (haplotype sharing test) results (B) of the six solved families 122 EXOME SEQUENCING

124 Recent studies have also shown that FHL2 is able to prevent pathological growth of the heart via the suppression of calcineurin activation that is induced by stress (Hojayev et al). Also, the overexpression of FHL2 might be the reason why ROCK2 conditional knock out mice were rescued from cardiac hypertrophy (Okamoto et al). Most importantly, a missense variant of FHL2 (p.gly48ser) found in a DCM patient has been reported to affect the binding of titin to the encoded protein (Arimura et al). Our putative FHL2 mutation was found in all three affected (and exome sequenced) siblings in this family. Unfortunately, the unaffected parents were not available for carriership analysis, nor were further affected family members available for co-segregation analysis, and HST was not performed in this family. Nonetheless, we classified this mutation as likely pathogenic because it is novel (i.e. not present in any control populations), the affected residue is localised in an evolutionarily highly conserved region of the 4 th LIM zinc-binding domain, and the mutation is suggested to be deleterious by most protein effect prediction programs. Family 4, ARVC-like: FLNC c.6864_6867dup; p.(val2290argfs*23) (NM_ ) We identified a potentially causative mutation in a gene-panel negative family with several family members suspected of ARVC, yet none fulfilling TFC in the filamin C gene (FLNC), which encodes an actin-crosslinking phosphoprotein (van der Flier & Sonnenberg). FLNC is highly expressed in murine cardiac and skeletal muscle during embryonic development and regeneration (Goetsch et al) and localizes to the Z-disk of striated muscle and to the intercalated disks in the heart (van der Ven et al, 2000). It is expected to have an essential role in the maintenance of the structural integrity of the cell and to protect it against mechanical stress as was observed in mutant zebrafish that suffered from enlarged hearts (Fujita et al). Moreover, FLNC was also shown to have interactions with delta- and gamma-sarcoglycan, in particular in the muscles (Thompson et al). FLNC mutations are known to cause distal and myofibrillar myopathy and might also affect the heart (reviewed by Selcen & Carpén), but had not thus far been associated with cardiomyopathy (or ARVC in particular), although a FLNC mutation associated with arrhythmia and late onset myofibrillar myopathy has been reported (Avila-Smirnow et al). Moreover, it is known that FLNC, along with other sarcomere genes (MYH7, TNNI3, TNNT2), shows differential splicing in failing heart, DCM and aortic stenosis (Kong et al, 2010). The insertion identified in this family causes a frameshift in exon 41 (which encodes filamin repeat 20 and mediates interaction with XIRP1 according to CHAPTER 3.1 AUTOSOMAL DOMINANT CARDIOMYOPATHIES 123

125 the Uniprot database, leading to a premature stop 23 codons after the affected codon, and, as a consequence, the loss of filamin repeats (unless the full sequence is subject to nonsense-mediated decay). The mutation was absent from control populations. Moreover, in the 6500 exomes of the ESP project, no truncating mutations were identified except for two truncations in the last but one exon of the gene, which probably do not have a large effect on the protein level and might escape nonsense-mediated decay. All affected family members for whom material was available were shown to carry the mutation. The mutation is located in the 26 th longest shared haplotype (7q31.32q35), which is still a shared haplotype of considerable size (29.73cM), although in this particular case HST was not used for variant prioritization. The shared mutation was identified after exome sequencing and data analysis of four affected siblings (II:2, II:5, II:6, II:8) and filtering using exome data of one unaffected sibling (II:3). Family 5, DCM: COBL c.998g>a; p.(arg333gln) (NM_ ) In this family suffering from an unusually mild and low penetrant form of DCM, we identified a putative missense mutation in the cordon bleu WH2 repeat protein gene (COBL) localized in the second longest shared haplotype (7p14.1q11.22). At the same time, no mutations were found in the FKTN gene located in the 1 st LSH, nor in the cardiomyopathy gene-panel. Even though this mutation affects a highly conserved region of the protein, we classified it as a variant of unknown significance (VOUS) due to the contradictory pathogenicity predictions and the fact that the variant was found with an allele frequency of 0.12% in the ESP database and present in only one individual within the genome of the Netherlands project. This VOUS co-segregated with the mild and low penetrance, late-onset DCM phenotype in the family. The paediatric patient (V:1 in the pedigree, see figure 2) was not carrying the same VOUS, but her severe symptoms and early onset of disease might indicate an independent cause of disease, perhaps according to a recessive inheritance pattern. The COBL protein is known to be of key importance in cytoskeletal dynamics as a very potent actin nucleator promoting the construction of long, unbranched filaments by elongation at the barbed ends (Ahuja et al). The knock out of the COBL homologue in zebrafish was previously found to cause developmental problems of the nervous system due to the inhibition of motile cilia causing insufficient determination of the left-right asymmetry axis. Interestingly, zebrafish also exhibited problems in the embryonic 124 EXOME SEQUENCING

126 development of the heart (the direction of heart looping was disturbed), which is not unexpected given that the heart, just like the nervous system, develops from a ciliated cell layer called Kupffer s vesicle. Unfortunately, there is no information available about whether there were any microscopic changes in the ultrastructure of muscle filaments in the hearts of these knock out animals (Ravanelli & Klingensmith). Curiously, another actin nucleator that, similar to COBL, promotes the growth of actin filaments at the barbed ends (though with somewhat weaker activity; Ahuja et al) was shown to play a role in sarcomere assembly in cardiomyocytes (Taniguchi et al). Also, a recent study showed significant association between hypertrophic cardiomyopathy and a missense variant of this gene, FHOD3 (formin homology 2 domain containing 3), and demonstrated the importance of its Drosophila homologue in normal systolic contractions of the adult heart in a knock down model (Wooten et al). To date, COBL has not been connected to heart diseases, yet it is known to be highly expressed in the heart according to the GeneCards database (www. genecards.org), and its interaction with actin filaments makes it an interesting candidate disease gene for DCM. Though a recent study investigated the functional consequences of mutating certain amino acids of the first two actin monomer binding WH2 domains of COBL by electron microscopy (Jiao et al), the potential role of the evolutionarily highly conserved KRAP motifs of the protein have yet to be discovered; one such motif is affected by the missense variant identified in our patient. In addition to the identification of this missense VOUS in affected members of family 4, we have screened a further 183 DCM index patients for carriership of this variant, and have identified one more, unrelated, paediatric patient carrying the same putative mutation. Due to the severity of the symptoms in this child, and the very early onset of the disease (at age 1 year), we anticipated that compound heterozygosity could explain her phenotype, yet no additional coding COBL variant was identified by Sanger sequencing. However, gene-panel-based NGS for 55 cardiomyopathy-related genes and the subsequent stringent variant classification in this patient resulted in the identification of a likely pathogenic missense variant c.263a>c p.(glu88ala) of the myosin light chain 2 gene (MYL2, NM_ ). The patient was confirmed to carry both mutations and we expected to identify their paternal and maternal origin, respectively. However, co-segregation analysis proved the maternal origin of both COBL and MYL2 mutations, raising the question if CHAPTER 3.1 AUTOSOMAL DOMINANT CARDIOMYOPATHIES 125

127 further genetic or other external factors may be behind the early manifestation of symptoms in the child. Family 6, DCM: STARD13 c.3017c>t; p.(pro1006leu) (NM_ ) A genetic variant in the START domain containing 13 (STARD13) gene was found in this family (START is the abbreviation of StAR-related lipid transfer; StAR stands for steroidogenic acute regulatory protein). The encoded protein is expected to be responsible for the binding of negatively charged small lipids such as phosphatidylcholine and fatty acids (Thorsell et al). STARD13 has been previously linked to several phenotypes including intracranial aneurysm (Yasuno et al) and insulin resistance related to metabolic syndrome (Nock et al). Combined with the facts that (1) irregular myocardial lipid turnover is a known phenomenon in dilated cardiomyopathy (Feinendegen et al) and (2) perturbed lipid metabolism, myocardial lipid accumulation, and a shift to the use of fatty acids instead of glucose as the predominant source of energy is observed in (and prior to the onset of) cardiomyopathy in diabetic patients and model animals (reviewed by Bayeva et al), these associations suggest that the genetic variant in STARD13 reported in this study could be related to disease in this family. The variant was identified in the 2 nd longest shared haplotype of the family (13p13q13.3) and was classified as likely pathogenic due to its novelty, the high evolutionary conservation of the affected amino acid and the respective lipid-binding START domain, and predicted pathogenicity according to all available software. Mutations of known candidate genes were excluded by gene-panel-based sequencing, and of those affected family members tested all were found to be carriers of the STARD13 mutation. Upon the identification of this novel candidate gene, the medical records of the family were re-checked for possible signs of the diabetes mellitus or metabolic syndrome potentially associated with this mutation, but no such symptoms have been observed thus far in the patients (aged 80, 74, 67 and 53 years). Network of the identified genes In order to gain insight into possible cardiac functions of the five genes identified in our exome sequenced families (COBL, FHL2, FLNC, STARD13, TTN), their HGNC approved gene symbols were uploaded to the Gene Network website ( which predicted that all of them except STARD13 might potentially be involved in different cardiomyopathies using data from the Kyoto Encyclopedia of Genes 126 EXOME SEQUENCING

128 and Genomes pathways (data not shown). Subsequently, the co-expressional network of these five genes was visualized in Cytoscape ( nl:8080/genenetwork cytoscape.html) (figure 3). The resulting network consists of 166 genes, of which 27 were already well known to be involved in the pathogenesis of various types of cardiomyopathy: ACTC1, ACTN2, ANKRD1, CAV3, CRYAB, CSRP3, FHL1, FHL2, LDB3, MYBPC3, MYH6, MYH7, MYL2, MYL3, MYOZ2, MYPN, NEXN, PDLIM3, PLN, SCN5A, TCAP, TNNC1, TNNI3, TNNT2, TPM1, TTN, and VCL. In addition to these well characterised disease genes, the novel neonatal DCM-associated ALPK3 (alpha-kinase 3) was also present in the network. The knock out model for the mouse homologue of this gene encoding a nuclear protein kinase is known to suffer from cardiomyopathy (van Sligtenhorst et al), and we have recently discovered a homozygous mutation of the gene in the DCM-affected child of a consanguineous family (manuscript submitted). More importantly, we applied an unprecedented approach to putting the genes in a functionally meaningful perspective. While searching for this list of co-expressed genes in the database of the Cardiovascular Gene Ontology Annotation Initiative ( UCL), we discovered that about 60% of the genes (100/166) have previously been manually annotated with a potential role in the physiological and/or pathological mechanisms of the cardiovascular system (table 1) based on the literature, and this is underscored by previous functional studies, as will be discussed below. For instance, triadin (TRDN) and xin actin-binding repeat containing 1 (XIRP1) are both known to be subject to tissue-specific splicing in the heart via RNA-binding motif protein 20 (RBM20), a known dilated cardiomyopathy protein that is part of the spliceosomal complex in the heart (Guo et al). TRDN is known to stimulate the ryanodine receptor-2 (RYR2) that functions as a sarcoplasmic Ca 2+ release channel with the help of calsequestrin (CASQ2, also featured in the co-expression network), and in this way play a role in excitation-contraction coupling in the heart (Morad et al, Terentyev et al, 2005; Terentyev et al, 2007). Mutations of TRDN have been identified in patients with catecholaminergic polymorphic ventricular tachycardia (Roux-Buisson et al). Furthermore, the XIRP1 gene, which was formerly known as cardiomyopathy associated 1 (CMYA1), is connected in the expression network to FLNC, and the respective protein was also shown to bind with the FLNC protein and participate in the process of sarcomere assembly and actin dynamics in cardiomyocytes (van der Ven et al, 2006). CHAPTER 3.1 AUTOSOMAL DOMINANT CARDIOMYOPATHIES 127

129 Figure 3. Co-expression network built upon the five genes (TTN, FHL2, FLNC, COBL, and STARD13) identified in the exome sequenced cardimyopathy families Red circles indicate the five genes identified by exome sequencing. Green circles indicate the genes co-expressed with those five identified genes. Only those genes having multiple connections within the network are indicated: the lighter green genes are co-expressed with two of the five candidates, while the darker green ones are expressed with three of them. 128 EXOME SEQUENCING

130 Curiously, the FLNC frameshift mutation identified in family 4 affects filamin repeat 20, which is known to mediate the binding of XIRP1. Moreover, proline-rich regions of XIRP1 were recently discovered to bind the SH3 domains of nebulin (NEB) and nebulette (NEBL), the myofibrillar proteins involved in the pathomechanism of nemaline myopathy and cardiomyopathy, respectively (Eulitz et al, Lehtokari et al, Purevjav et al). It is rather remarkable that several genes in the network are shown to be important in sarcomere assembly, and this also applies to the proteins encoded by three of the five genes we identified: COBL also functions as an actin nucleator (Ahuja et al), TTN is a known structural component of the sarcomere (Horowits et al), and FLNC is also expected to play a role in the assembly process (van der Ven et al, 2000; Bönnemann et al, Fujita et al). Comparably, tropomodulin-1 (TMOD1) and leiomodin (LMOD) were shown to be involved in sarcomere assembly, as they have a role in fine-tuning the length of thin filaments in cardiomyocytes. TMOD1 caps the pointed end of actin filaments in the M-line of sarcomeres, while the competing LMOD2 is an actin nucleation factor that promotes sarcomere assembly in a tropomyosin-dependent way (Chereau et al, Skwarek-Maruszewska et al; Tsukada et al). In line with this, the gene encoding the cardiomyopathy-related tropomyosin (TPM1) also appears in the co-expression network of the five genes. This suggests that these genes are part of a putative common molecular pathway. The fact that the disease genes we identified by exome sequencing are connected within such a functionally meaningful co-expression network, and that it is enriched for known cardiomyopathy genes as well as genes expected to play an essential role in the heart, underscores the usefulness of such databases in interpreting highthroughput genetic findings. Furthermore, our finding that this network is enriched for the sarcomeric components is in line with the recent observation in a cohort of 639 DCM patients that 14% of the known pathogenic mutations were related to the sarcomeric structure, making this the most frequently mutated cellular compartment in the disease (Haas et al). One of the limitations of our study is that we have not yet found additional patients with the same mutations, or with other relevant genetic variants in some of the candidate genes. Although our approach of combining HST and ES did help deal with the relatively small size of families, the limited number of affected individuals available in this study might have influenced our findings. Segregation analysis supported putative pathogenicity of the identified variants in most families, yet it is always challenging to perform CHAPTER 3.1 AUTOSOMAL DOMINANT CARDIOMYOPATHIES 129

131 Table 1. List of the five genes identified in the exome sequenced cardiomyopathy families and the genes of the additional 161 co-expressed proteins Genes in black have been previously studied in the context of cardiomyopathies and connected to the disease, or have been included in the Cardiovascular Gene Annotation Ontology Initiative database on the basis of data-mining and functional studies suggesting a putative role in cardiovascular physiology or pathophysiology. Genes in grey have not been previously connected to cardiomyopathies or annotated with a putative role in cardiovascular physiology or pathophysiology. gene name of gene ABRA ACTA1 Actin, alpha skeletal muscle ACTC1 Actin, alpha cardiac muscle 1 ACTN2 Alpha-actinin-2 ADPRHL1 ADSSL1 ALPK3 AMOTL2 AMPD1 AMP deaminase 1 ANKRD1 ANKRD2 Ankyrin repeat domain-containing protein 2 ANXA3 APOBEC2 ASB2 ASB5 ATP1A2 Sodium/potassium-transporting ATPase subunit alpha-2 ATP2A1 Sarcoplasmic/endoplasmic reticulum calcium ATPase 1 ATP2A2 Sarcoplasmic/endoplasmic reticulum calcium ATPase 2 AXL BAG2 BDNF Brain-derived neurotrophic factor C10orf7 CA3 CACNA1S Voltage-dependent L-type calcium channel subunit alpha-1s CACNB1 Voltage-dependent L-type calcium channel subunit beta-1 CACNG1 Voltage-dependent calcium channel gamma-1 subunit CALD1 Caldesmon CAND2 CAP2 CASQ1 Calsequestrin-1 CASQ2 Calsequestrin-2 CAV3 Caveolin-3 CFL2 CHRNA1 Acetylcholine receptor subunit alpha CHRNB1 Acetylcholine receptor subunit beta CHRND Acetylcholine receptor subunit delta CKB Creatine kinase B-type CKM Creatine kinase M-type CMYA5 CNN1 Calponin-1 COBL Protein cordon-bleu CORO6 COX6A2 Cytochrome c oxidase subunit 6A2, mitochondrial CRYAB Alpha-crystallin B chain CSRP3 Cysteine and glycine-rich protein 3 DMPK Myotonin-protein kinase DUSP13 many other DUSPs DUSP27 many other DUSPs EEF1A2 ENO3 Beta-enolase FABP3 Fatty acid-binding protein, heart FHL1 Four and a half LIM domains protein 1 FHL2 Four and a half LIM domains protein 2 FLNC Filamin-C HFE2 HRC Sarcoplasmic reticulum histidine-rich calcium-binding protein HSPB3 HSPB6 HSPB7 Heat shock protein beta-7 HSPB8 IP6K3 Inositol hexakisphosphate kinase 3 ITGB1BP2 Integrin beta-1-binding protein 2 ITGB1BP3 KBTBD5 KERA Keratocan LDB3 LMOD1 Leiomodin-1 LMOD3 LRRC2 MB Myoglobin MLIP MURC MUSK Muscle, skeletal receptor tyrosine-protein kinase found by exome sequencing known in cardiomyopathy listed in Cardiovascular Gene Ontology Annotation Initiative 130 EXOME SEQUENCING

132 MYBPC1 Myosin-binding protein C, slow-type MYBPC2 Myosin-binding protein C, fast-type MYBPC3 Myosin-binding protein C, cardiac-type MYBPH Myosin-binding protein H MYF6 Myogenic factor 6 MYH1 Myosin-1 MYH11 Myosin-11 MYH2 Myosin-2 MYH3 Myosin-3 MYH6 Myosin-6 MYH7 Myosin-7 MYH8 Myosin-8 MYL2 Myosin regulatory light chain 2, ventricular/cardiac muscle isoform MYL3 Myosin light chain 3 MYL4 Myosin light chain 4 MYL7 Myosin regulatory light chain 2, atrial isoform MYLK3 Myosin light chain kinase 3 MYLPF Myosin regulatory light chain 2, skeletal muscle isoform MYO18B MYOD1 Myoblast determination protein 1 MYOF Myoferlin MYOG Myogenin MYOM1 Myomesin-1 MYOM2 Myomesin-2 MYOT Myotilin MYOZ1 Myozenin-1 MYOZ2 Myozenin-2 MYPN NEXN Nexilin NPHS2 Podocin NPPA Natriuretic peptides A NPPB Natriuretic peptides B NRAP OBSCN PACSIN3 PDLIM3 PDLIM5 PFKM PKIA PLN Cardiac phospholamban POPDC2 PPP1R27 many other PPP1Rs PPP2R3A many other PPP2Rs PRKAA2 5 -AMP-activated protein kinase catalytic subunit alpha-2 PYGM RAPSN RBFOX RBM24 RP11-59J5.1 RP11-766F14.2 RRAD RTN2 RYR1 Ryanodine receptor 1 SCN4A Sodium channel protein type 4 subunit alpha SCN5A Sodium channel protein type 5 subunit alpha SGCA Alpha-sarcoglycan SGCG Gamma-sarcoglycan SH3BGR SLN Sarcolipin SMPX Small muscular protein SMTNL1 SMTNL2 SOX10 SRL SRPK3 SRSF protein kinase 3 STAC3 STARD13 StAR-related lipid transfer protein 13 SYNPO2 SYNPO2L TAGLN Transgelin TCAP Telethonin TECRL TGFB1I1 TMOD1 Tropomodulin-1 TNFRSF12A TNNC1 Troponin C, slow skeletal and cardiac muscles TNNC2 Troponin C, skeletal muscle TNNI1 Troponin I, slow skeletal muscle TNNI2 Troponin I, fast skeletal muscle TNNI3 Troponin I, cardiac muscle TNNT1 Troponin T, slow skeletal muscle TNNT2 Troponin T, cardiac muscle TNNT3 Troponin T, fast skeletal muscle TPM1 Tropomyosin alpha-1 chain TPM2 Tropomyosin beta chain TRDN Triadin TRIM63 TTN Titin UNC45B Protein unc-45 homolog B VCL Vinculin VGLL2 Transcription cofactor vestigial-like protein 2 XIRP1 Xin actin-binding repeat-containing protein 1 ZFP106 CHAPTER 3.1 IN TOTAL: 166 genes KNOWN IN CARDIOMYOPATHY: 28 genes (17.47%) ANNOTATED WITH PUTATIVE CARDIOVASCULAR ROLE: 100 genes (60.24%) AUTOSOMAL DOMINANT CARDIOMYOPATHIES 131

133 this for a late-onset disease such as cardiomyopathy because the healthy or affected status of family members is sometimes questionable. This fact, combined with the occasional presence of phenocopies, makes the accurate phenotyping of relatives sometimes difficult and might hamper the accurate analysis of ES and HST data. We cannot fully exclude the possibility that these two issues might have affected the outcome (especially the lack of any mutation being identified) in some of our families. On the other hand, having no genetic cause of disease identified in half of our families might also be due to other, technical problems, such as a lack of sufficient coverage of the respective mutation/disease gene in the available ES data. It has been anticipated that the revolutionary development of new genetics methods necessitate the application of appropriate bioinformatic tools and functional follow up to better interpret the respective results (Singleton, 2014). A very appealing, recent, example of combining exome sequencing with the creation of networks in neurodegeneration was published by Novarino et al. This group identified mutations of novel candidate disease genes in consanguineous families of hereditary spastic paraplegias (HSP), and validated their findings by discovering additional novel genes (and their mutations) selected from the protein interaction network of these novel candidate genes and already known HSP disease genes. Although there have been protein-protein interaction networks created for certain cardiac phenotypic traits (but not for cardiomyopathies) (Lage et al, 2010; Lage et al, 2012), this is the first example of combining exome sequencing and the use of a co-expression based network for the interpretation of the role of potential cardiovascular disease genes and pathways in inherited cardiomyopathy. Admittedly, the network of genes created in this study is based on shared mrna expression patterns instead of interactions at the protein level. However, in comparison with protein interaction networks, it has the advantage of not creating a bias through exclusion of those genes from the network analysis that have not yet been functionally studied or otherwise shown to interact with heart-specific proteins. CONCLUSIONS We have performed haplotype sharing tests and exome sequencing in twelve families suffering from DCM or ARVC with no identified genetic cause of the disorder. This resulted in the identification of potentially causative, heterozygous variants in six of the twelve families sequenced. 132 EXOME SEQUENCING

134 Their involvement in disease was supported by the fact that the mutations identified co-segregated with the disease; most genes were located in one of the longest shared haplotypes and were absent or present at very low frequency in control populations. Moreover, the fact that 2/3 of the genes co-expressed with these five genes (TTN, FHL2, FLNC, COBL and STARD13) are annotated with a potential function in the heart, and many are related to the process of sarcomere assembly and reorganization of the cytoskeleton, suggests that a potential common molecular pathway may connect them in cardiomyopathy. Since one of the genes discovered in two families is the well-known DCM gene TTN, it has become part of the routine in our department to first perform targeted sequencing for 55 cardiomyopathy genes (including TTN; chapter 4.1) and then to only perform exome sequencing after excluding mutations in all these known disease genes. In the future, it will be of great importance to investigate the cellular function of the COBL and STARD13 genes, as well as the molecular pathways they play a role in, and the potential involvement of the six identified mutations in the pathomechanism of DCM and ARVC. Moreover, we will try to identify underlying disease genes in the other six families by (1) reanalysing the data, (2) incorporating exome sequence data of additional affected and unaffected family members, (3) analysing the data for putative large deletions/duplications, and/or (4) applying other genomic techniques, such as RNA sequencing or whole genome sequencing. CHAPTER 3.1 ACKNOWLEDGEMENTS The authors would like to thank the families for participating in this study; Ludolf Boven and Elisabetta Lazzarini for technical assistance; members of the Genomics Coordination Centre, UMCG, for assistance in data analysis; Ellen Otten, Gerdien Bosman, Sandra Hermers, Rina Keupink, Jolien Klein-Wassink-Ruiter, Karin Nieuwhof, Wilma van der Roest and Marijke Wasielewski for counselling of families; and Jackie Senior and Kate Mc Intyre for editing this manuscript. Rowida Almomani was supported by the Netherlands Heart Foundation (grant 2010B164) and Anna Pósafalvi was supported by grants from the Jan Kornelis de Cock Foundation. AUTOSOMAL DOMINANT CARDIOMYOPATHIES 133

135 REFERENCES Ahuja R, Pinyol R, Reichenbach N et al. Cordon-bleu is an actin nucleation factor and controls neuronal morphology. Cell 2007;131(2): Arimura T, Hayashi T, Matsumoto Y et al. Structural analysis of four and half LIM protein-2 in dilated cardiomyopathy. Biochem Biophys Res Commun 2007;357(1):162-7 Avila-Smirnow D, Béhin A, Gueneau L et al. A novel missense FLNC mutation causes arrhythmia and late onset myofibrillar myopathy with particular histopathology features. Abstract for poster P2.18 presented at the 15 th International Congress of The World Muscle Society, 2010 ( Basso C, Corrado D, Marcus FI et al. Arrhythmogenic right ventricular cardiomyopathy. Lancet 2009;373: Bayeva M, Sawicki KT and Ardehali H. Taking diabetes to heart deregulation of myocardial lipid metabolism in diabetic cardiomyopathy. J Am Heart Assoc 2013;2:e Bönnemann CG, Thompson TG, van der Ven PF et al. Filamin C accumulation is a strong but nonspecific immunohistochemical marker of core formation in muscle. J Neurol Sci 2003;206:71-8 Chan KK, Tsui SK, Lee SM et al. Molecular cloning and characterization of FHL2, a novel LIM domain protein preferentially expressed in human heart. Gene 1998;210(2): Chereau D, Boczkowska M, Skwarek-Maruszewska A et al. Leiomodin is an actin filament nucleator in muscle cells. Science 2008;320(5873): Chu PH, Bardwell WM, Gu Y et al. FHL2 (SLIM3) is not essential for cardiac development and function. Mol Cell Biol 2000;20(20): Cox MG, van der Zwaag PA, van der Werf C et al. Arrhythmogenic right ventricular dysplasia/ cardiomyopathy: pathogenic desmosome mutations in index-patients predict outcome of family screening: Dutch arrhythmogenic right ventricular dysplasia/cardiomyopathy genotype-phenotype follow-up study. Circulation 2011;123(23): Eulitz S, Sauer F, Pelissier MC et al. Identification of Xin-repeat proteins as novel ligands of the SH3 domains of nebulin and nebulette and analysis of their interaction during myofibril formation and remodeling. Mol Biol Cell 2013;24(20): Feinendegen LE, Henrich MM, Kuikka JT et al. Myocardial lipid turnover in dilated cardiomyopathy: a dual in vivo tracer approach. J Nucl Cardiol 1995;2:42-52 Fujita M, Mitsuhashi H, Isogai S et al. Filamin C plays an essential role in the maintenance of the structural integrity of cardiac and skeletal muscles, revealed by the medaka mutant zacro. Dev Biol 2012;361:79-89 Gerull B, Gramlich M, Atherton J et al. Mutations of TTN, encoding the giant muscle filament titin, cause familial dilated cardiomyopathy. Nat Genet 2002;30(2):201-4 Goetsch SC, Martin CM, Embree LJ et al. Myogenic progenitor cells express filamin C in developing and regenerating skeletal muscle. Stem Cell Dev 2005;14:181-7 Granzier HL and Wang K. Gel electrophoresis of giant proteins: solubilization and silver-staining of titin and nebulin from single muscle fiber segments. Electrophoresis 1993;14:56-64 Guo W, Schafer S, Greaser ML et al. RBM20, a gene for hereditary cardiomyopathy, regulates titin splicing. Nat Med 2012;18(5): Haas J, Frese KS, Peil B et al. Atlas of the clinical genetics of human dilated cardiomyopathy. Eur Heart J 2014; pii: ehu301 Hein S, Scholz D, Fujitani N et al. Altered expression of titin and contractile proteins in failing human myocardium. J Mol Cell Cardiol 1994;26(10): Herman DS, Lam L, Taylor MR et al. Truncations of titin causing dilated cardiomyopathy. N Engl J Med 2012;366(7): Hershberger RE, Hedges DJ and Morales A: Dilated cardiomyopathy: the complexity of a diverse genetic architecture. Nat Rev Cardiol 2013;10: Hojayev B, Rothermel BA, Gillette TG et al. FHL2 binds calcineurin and represses pathological cardiac growth. Mol Cell Biol 2012;32(19): Horowits R, Kempner ES, Bisher ME et al. A physiological role for titin and nebulin in skeletal muscle. Nature 1986;323(6084):160-4 Jiao Y, Walker M, Trinick J et al. Mutagenetic and electron microscopy analysis of actin filament severing by Cordon-Bleu, a WH2 domain protein. Cytoskeleton 2014;71(3): Kong SW, Hu YW, Ho JW et al. Heart failure-associated changes in RNA splicing of sarcomere genes. Circ Cardiovasc Genet 2010;3: Kong Y, Shelton JM, Rothermel B et al. Cardiac-specific LIM protein FHL2 modifies the hypertrophic response to beta-adrenergic stimulation. Circulation 2001;103(22): EXOME SEQUENCING

136 Lage K, Møllgård K, Greenway S et al. Dissecting spatio-temporal protein networks driving human heart development and related disorders. Mol Syst Biol 2010;6:381 Lage K, Greenway SC, Rosenfeld JA et al. Genetic and environmental risk factors in congenital heart disease functionally converge in protein networks driving heart development. Proc Natl Acad Sci USA 2012;109(35): Lehtokari VL, Kiiski K, Sandaradura SA et al. Mutation update: The spectra of nebulin variants and associated myopathies. Hum Mutat 2014; doi: /humu Li H & Durbin R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 2009;25: Li H, Handsaker B, Wysoker A et al. The Sequence Alignment/Map format and SAMtools. Bioinformatics 2009;25: Marcus FI, McKenna WJ, Sherrill D et al. Diagnosis of arrhythmogenic right ventricular cardiomyopathy/dysplasia: proposed modification of the task force criteria. Circulation 2010;121: McKenna A, Hanna M, Banks E et al. The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res 2010;20: Mestroni L, Maisch B, McKenna WJ et al. Guidelines for the study of familial dilated cardiomyopathy. Eur Heart J 1999;20: Morad M, Cleemann L, Knollmann BC. Triadin: the new player on excitation-contraction coupling block. Circ Res 2005;96(6):607-9 Nock NL, Wang X, Thompson CL et al. Defining genetic determinants of the Metabolic Syndrome in the Framingham Heart Study using association and structural equation modeling methods. BMC Proc 2009;3(Suppl 7): S50 Novarino G, Fenstermaker AG, Zaki MS et al. Exome sequencing links corticospinal motor neuron disease to common neurodegenerative disorders. Science 2014;343(6170): Okamoto R, Li Y, Noma K et al. FHL2 prevents cardiac hypertrophy in mice with cardiac-specific deletion of ROCK2. FASEB J 2013;27: Peled Y, Gramlich M, Yoskovitz G et al. Titin mutation in familial restrictive cardiomyopathy. Int J Cardiol 2014;171(1):24-30 Posafalvi A, Herkert JC, Sinke RJ et al. Clinical utility gene card for: dilated cardiomyopathy (CMD). Eur J Hum Genet 2013;21. doi: /ejhg Purevjav E, Varela J, Morgado M et al. Nebulette mutations are associated with dilated cardiomyopathy and endocardial fibroelastosis. J Am Coll Cardiol 2010;56(18): Quarta G, Muir A, Pantazis A et al. Familial evaluation in arrhythmogenic right ventricular cardiomyopathy: impact of genetics and revised task force criteria. Circulation 2011;123: Ravanelli AM & Klingensmith J. The actin nucleator Cordon-bleu is required for development of motile cilia in zebrafish. Dev Biol 2011;350(1): Roux-Buisson N, Cacheux M, Fourest-Lieuvin A et al. Absence of triadin, a protein of the calcium release complex, is responsible for cardiac arrhythmia with sudden death in human. Hum Mol Genet 2012;21(12): Satoh M, Takahashi M, Sakamoto T et al. Structural analysis of the titin gene in hypertrophic cardiomyopathy: identification of a novel disease gene. Biochem Biophys Res Commun 1999;262(2):411-7 Selcen D and Carpén O. The Z-disk diseases. Adv Exp Med Biol 2008;642: Sikkema-Raddatz B, Johansson LF, de Boer EN et al. Targeted next-generation sequencing can replace Sanger sequencing in clinical diagnostics. Hum Mutat 2013;34: Singleton AB. Genetics. A unified process for neurological disease. Science 2014;343(6170):497-8 Skwarek-Maruszewska A, Boczkowska M, Zajac AL et al. Different localizations and cellular behaviors of leiomodin and tropomodulin in mature cardiomyocyte sarcomeres. Mol Biol Cell 2010;21(19): Taniguchi K, Takeya R, Suetsugu S et al. Mammalian formin fhod3 regulates actin assembly and sarcomere organization in striated muscles. J Biol Chem 2009;284: Taylor M, Graw S, Sinagra G et al. Genetic variation in titin in arrhythmogenic right ventricular cardiomyopathy-overlap syndromes. Circulation 2011;124(8): te Rijdt WP, Jongbloed JD, de Boer RA et al. Clinical utility gene card for: arrhythmogenic right ventricular cardiomyopathy (ARVC). Eur J Hum Genet 2014;22(2). doi: /ejhg Terentyev D, Cala SE, Houle TD et al. Triadin overexpression stimulates excitation-contraction coupling and increases predisposition to cellular arrhythmia in cardiac myocytes. Circ Res 2005;96(6):651-8 CHAPTER 3.1 AUTOSOMAL DOMINANT CARDIOMYOPATHIES 135

137 Terentyev D, Viatchenko-Karpinski S, Vedamoorthyrao S et al. Protein protein interactions between triadin and calsequestrin are involved in modulation of sarcoplasmic reticulum calcium release in cardiac myocytes. J Physiol 2007;583(Pt 1):71-80 Thompson TG, Chan YM, Hack AA et al. Filamin 2 (FLN2): A muscle-specific sarcoglycan interacting protein. J Cell Biol 2000;148: Thorsell AG, Lee WH, Persson C et al. Comparative structural analysis of lipid binding START domains. PLoS One 2011;6(6):e19521 Tsukada T, Pappas CT, Moroz N et al. Leiomodin-2 is an antagonist of tropomodulin-1 at the pointed end of the thin filaments in cardiac muscle. J Cell Sci 2010;123(Pt 18): van der Flier A & Sonnenberg A: Structural and functional aspects of filamins. Biochim Biophys Acta 2001;1538: van der Ven PF, Obermann WM, Lemke B et al. Characterization of muscle filamin isoforms suggests a possible role of gamma-filamin/ ABP-L in sarcomeric Z-disc formation. Cell Motil Cytoskeleton 2000;45: van der Ven PF, Ehler E, Vakeel P et al. Unusual splicing events result in distinct Xin isoforms that associate differentially with filamin c and Mena/VASP. Exp Cell Res 2006;312(11): van der Zwaag PA, van Tintelen JP, Gerbens F et al. Haplotype sharing test maps genes for familial cardiomyopathies. Clin Genet 2011;79: van Sligtenhorst I, Ding ZM, Shi ZZ et al. Cardiomyopathy in α-kinase 3 (ALPK3)-deficient mice. Vet Pathol 2012;49: van Spaendonck-Zwarts KY, van Rijsingen IA, van den Berg MP et al. Genetic analysis in 418 index patients with idiopathic dilated cardiomyopathy: overview of 10 years experience. Eur J Heart Fail 2013;15: van Spaendonck-Zwarts KY, Posafalvi A, van den Berg MP et al. Titin gene mutations are common in families with both peripartum cardiomyopathy and dilated cardiomyopathy. Eur Heart J 2014; doi: /eurheartj/ehu050 van Tintelen JP, Entius MM, Bhuiyan ZA et al. Plakophilin-2 mutations are the major determinant of familial arrhythmogenic right ventricular dysplasia/cardiomyopathy. Circulation 2006;113: Waszak SM, Hasin Y, Zichner T et al. Systematic inference of copy-number genotypes from personal genome sequencing data reveals extensive olfactory receptor gene content diversity. PLoS Comput Biol 2010;6(11):e Wilde AA & Behr ER: Genetic testing for inherited cardiac disease. Nat Rev Cardiol 2013;10: Wooten EC, Hebl VB, Wolf MJ et al. Formin homology 2 domain containing 3 variants associated with hypertrophic cardiomyopathy. Circ Cardiovasc Genet 2013;6:10-18 Yasuno K, Bilguvar K, Bijlenga P et al. Genome-wide association study of intracranial aneurysm identifies three new risk loci. Nat Genet 2010;42: EXOME SEQUENCING

138

139

140 Chapter 3.2 Homozygous SOD2 mutation as a cause of lethal neonatal dilated cardiomyopathy Rowida Almomani 1, *, Anna Posafalvi 1, *, Johanna C Herkert 1, Jan G Post 2, Paul A van der Zwaag 1, Peter Willems 3, Cindy Weidijk 1, Peter GJ Nikkels 4, Richard J Rodenburg 5, Richard J Sinke 1, J Peter van Tintelen 1, Jan DH Jongbloed 1 *These authors contributed equally to these studies. Manuscript in preparation

141 ABSTRACT Although cases are rare, neonatal and paediatric dilated cardiomyopathy (DCM) is a severe and often lethal disease, in which a genetic factor plays an important role in disease development. Identifying this genetic component is of major importance for parents as it enables prenatal diagnostics to be performed in their future pregnancies. Here, we report the results of homozygosity mapping followed by exome sequencing in a DCM-affected neonate in whom autosomal recessive inheritance was anticipated. This approach revealed a potentially pathogenic, homozygous missense mutation, c.542g>t, p.(gly181val), in the gene encoding Superoxide dismutase 2 (SOD2). SOD2 is a mitochondrial matrix protein that converts the reactive oxygen species (ROS) superoxide anion (O 2 ) into H 2 O 2, and is therefore important for preventing cellular damage due to oxidative stress. We measured the oxidation of hydroethidine and detected a significant increase in O 2 levels in the fibroblasts of the patient compared with controls. This indicates that the mutation affects the catalytic activity of SOD2, which could lead to a drastic increase in damaging oxygen radical levels in the neonatal heart and result in rapidly developing heart failure and death. In conclusion, we have identified a novel mitochondrial gene involved in severe neonatal cardiomyopathy, thus expanding the wide range of genetic factors involved in paediatric cardiomyopathies.

142 INTRODUCTION Dilated cardiomyopathy (DCM) is characterized by left ventricular enlargement and systolic dysfunction, which can lead to heart failure and sudden cardiac death (Fatkin et al). It is the most common type of cardiomyopathy and the major reason for heart transplantations in children. The incidence of DCM in children is estimated to be 0.57/100,000 per year, and is even higher in children below the age of one year (8.34/100,000) (Towbin et al). Approximately 25-50% of DCM cases are familial, and mutations in more than 50 genes have been reported to be associated with adult-onset familial DCM, some of which are observed in paediatric DCM as well (Somsen et al, Dellefave & McNally, Posafalvi et al). DCM-associated genes encode diverse groups of proteins including cytoskeletal, sarcomeric, ion transport, nuclear membrane and mitochondrial proteins (Somsen et al, Dellefave & McNally, Posafalvi et al). In contrast to adult DCM, knowledge about the underlying genetic causes of paediatric cases is still limited. In familial cases, mutations are regularly found in the known DCM genes (Rampersaud et al). However, these neither explain the majority of pediatric cases in which rare mutations in autosomal recessive inherited genes underlie disease, nor the cases of children whose DCM is part of a syndromic or metabolic disease (Kindel et al). Therefore, Burns et al recently concluded that approaches using gene-panel based applications targeting adult DCM disease genes are less appropriate for the severe infantile forms of the disease, and they suggested that gene discovery is likely to proceed more rapidly when exome sequencing (ES) or genome sequencing are applied. Successful application of ES to identify the causal mutations in paediatric DCM has been recently demonstrated (Theis et al 2011, 2014; Louw et al). Here we have used homozygosity mapping followed by ES to identify the genetic cause of lethal DCM in a three-day-old Dutch girl. The homozygous mutation, c.542g>t, p.(gly181val), we found in the SOD2 gene (NM_ ) most likely affects the catalytic activity of the protein, leading to excess oxygen radical levels with strongly damaging effects in the neonatal heart. CHAPTER 3.2 METHODS Case report The female patient was born at weeks gestation after a caesarean delivery due to breech presentation and meconium staining of the amniotic fluid. The pregnancy was complicated by maternal nephrotic syndrome SOD2 IN AUTOSOMAL RECESSIVE CARDIOMYOPATHY 141

143 at 19 weeks gestation and treated with prednisone. Her Apgar scores were 2-3 and 9, her birth weight was 2240 g (<p2.3), length at birth was 49 cm (p25) and head circumference was 33.0 cm (p5). Umbilical artery ph was 7.14 with a base excess of -4 mmol/l. The day after birth she presented with apnoeas, poor circulation and mild tachycardia. A chest X-ray was normal. Echocardiography showed a structurally normal heart, but left ventricular function seemed poor. Cardiac troponin and BNP were elevated, 0.28 µg/l (n < 0.16 µg/l) and 2819 pmol/l, respectively. On the third day after birth she developed cardiogenic shock with frequent ventricular extrasystoles and tachycardia. Both ventricles showed dilatation and she died three days postpartum. Biochemical studies showed a high level of lactate in the blood, possibly due to poor circulation; increased amino acids, including proline and alanine; and increased organic acids, including 3-methylglutaconic acid. Pompe disease was excluded by normal plasma alpha-glycosidase levels. Disorders of N-glycosylation and peroxisomal metabolism were also excluded. Viral serology showed no abnormality. At autopsy, macroscopic examination of the heart revealed severe dilatation of both ventricles, without any histological abnormalities. The cardiac weight was consistent with 41 weeks gestation, while the weights of other organs were consistent with weeks gestation. Skeletal muscle showed no abnormalities and there were no indications of disorders of fatty acid oxidation or of mitochondrial disease, although these could not be fully excluded by histological examination. Actin, dystrophin, sarcoglycan, dystroglycan, dysferlin, caveolin-3, merosin, myosin and spectrin-1 staining were normal, and respiratory chain complexes were measured and also normal. Intracranial examination showed small cerebral subependymal cysts. Genomic DNA of the child and her parents was extracted from peripheral blood using standard protocols. The parents provided informed consent for DNA studies, and for diagnostic procedures. The UMCG ethical committee approved this study. Homozygosity mapping Genome-wide genotyping with the HumanCytoSNP-12 BeadChip 300K SNP array (Illumina, San Diego, CA, USA) was performed according to the manufacturer s protocols. Data from the arrays were converted to genotypes using the GenomeStudio data analysis software (Illumina). The genotype data was subject to homozygosity mapping using Microsoft Office Excel 142 EXOME SEQUENCING

144 2010 (Version 14.0; Microsoft, Redmond, WA, USA) software by searching for homozygous regions in the patient s DNA, allowing for a 1% genotyping error margin. The size of the homozygous regions was calculated in megabases (Mb) and in centimorgans (cm), based on the decode genetic map (Kong et al). Exome sequencing ES on the patient s DNA was performed using the SureSelect 50Mb exome capture kit (Agilent, Santa Clara, CA, USA) following the manufacturer s protocol. The enriched fragments captured were sequenced using the Illumina HiSeq platform in paired-end mode, with a read length of 100 bp following the manufacturer s protocol. The raw Fastq files were aligned by using bwa to the human reference genome (hg 19, NCBI build 37) (Li et al, 2009a), SAM/BAM files were manipulated by Samtools , and Picard-1.57 (Li et al, 2009b). Then the Genome Analysis Toolkit (GATK) was used to perform base quality score recalibration, duplicate removal and INDEL realignment (McKenna et al). The output vcf files were annotated by our in-house Bioinformatics pipeline and Seattleseq ( Subsequent mutation analysis Sanger sequencing was used to confirm the presence/absence of the SOD2 mutation in the patient and her family members. In addition, screening of all exons and exon/intron junctions of the SOD2 gene was performed in other patients. PCR was performed by using AmpliTaq Gold PCR Master Mix (Invitrogen Life Science Technologies, Carlsbad, CA, USA) following the official protocol and resulting fragments were sequenced by Applied Biosystems 96-capillary 3730XL system (Carlsbad, CA, USA). CHAPTER 3.2 RNA extraction and Reverse Transcriptase-PCR (RT-PCR) product analysis RNA was isolated from cultured fibroblasts from the patient. Cells were cultured in standard medium for human fibroblasts (Dulbecco s modified Eagle s medium with 10% FBS, 1% penicillin/streptomycin, 1% glucose, 1% glutamax) with 5% CO 2 at 37 C. RNA was extracted with the RNeasy Mini Kit (QIAGEN, Venlo, the Netherlands) following the manufacturer s protocol. cdna was synthesized from 500 ng of total RNA by RevertAid RNaseH- M-MuLV reverse transcriptase in a total volume of 20 μl according to the SOD2 IN AUTOSOMAL RECESSIVE CARDIOMYOPATHY 143

145 protocol provided by the supplier (MBI-Fermentas, St Leon-Rot, Germany). To investigate whether the c.542g>t mutation could have an effect on mrna splicing, we performed RT-PCR with primers specific for SOD2 and designed to amplify the exon that was expected to be affected by the mutation and flanking sequences (primers are available upon request). Target regions were amplified by PCR and the PCR products were examined by 2% agarose gel and analysed by Sanger sequencing. To test for effects of nonsense-mediated decay, fibroblasts were treated with cycloheximide for 4.5 hr, followed by RNA analysis using the same procedures as those for RNA from untreated cells. Measurement of superoxide substrate levels Fibroblasts, cultured to 70% confluence, were incubated in HEPES-Tris medium containing 10 μm hydroethidine (HEt) for 10 min at 37 C. Within the cell, HEt reacts with O2 to form the fluorescent and positively charged product ethidium (Et) or oxyethidium. The reaction was stopped by thoroughly washing the cells with PBS to remove excess HEt. For quantitative analysis of Et emission signals, coverslips were mounted in an incubation chamber placed on the stage of an inverted microscope (Axiovert 200 M; Carl Zeiss, Jena, Germany) equipped with a Zeiss 40/1.3 NA fluor lens objective. Et was excited at 490 nm using a monochromator (Polychrome IV; TILL Photonics, Gräfelfing, Germany). Fluorescence emission was directed using a 525DRLP dichroic mirror (Omega Optical, Brattleboro, VT) through a 565ALP emission filter (Omega Optical) onto a CoolSNAP HQ monochrome charge-coupled device camera (Roper Scientific, Vianen, the Netherlands). The imagecapturing time was 100 ms. Routinely, 10 fields of view per coverslip were analysed. SOD2 protein s 3D structure As the 3D-structure of the SOD2 protein is known, HOPE software was applied to predict the potential effect of the p.(gly181val) missense mutation on the 3D structure of the protein (Venselaar et al). Additionally, the Uniprot protein database ( was used to search for known functional features within the mitochondrial Superoxide dismutase [Mn] protein (accession number: P04179) in the region affected by the genetic variation. 144 EXOME SEQUENCING

146 RESULTS Case report Genealogical analysis found a distant relationship between the parents 6 to 8 generations previously, suggesting an autosomal recessive inheritance (figure 1). Array-CGH showed no pathogenic copy number variations. Diagnostic Sanger sequencing results of mitochondrial DNA, isolated from fibroblasts, and of the POLG, MYL2, MYH7, LMNA, DES, SUCLA2 and RYR2 genes were normal. Respiratory chain complexes were found to function normally. Echocardiography revealed no abnormalities in the mother or father (aged 27 and 29, respectively) or in the patient s younger brother (cardiologically evaluated aged 1 week). CHAPTER 3.2 Figure 1. Pedigree of a Dutch family with a child with severe, lethal DCM, in whom autosomal recessive inheritance was expected due to the pedigree composition. The patient is marked with a black symbol. SOD2 IN AUTOSOMAL RECESSIVE CARDIOMYOPATHY 145

147 Figure 2. Homozygosity mapping results show the second longest homozygous region (the longest autosomal homozygous region) on chromosome 6, where the SOD2 gene is located. Homozygosity mapping Homozygosity mapping in the patient (figure 1; X:1) revealed the longest homozygous region was on the X chromosome (figure 2). The longest autosomal region of homozygosity was located on chromosome 6, between rs and rs (159,949, ,713,427 bp; UCSC Genome Browser, build hg19), spanning 268 SNPs and 4.26 cm. This homozygous region contains 26 genes, including the SOD2 gene. Exome sequencing ES was performed to target all exons and exon/intron junction sequences of the known genes in the human genome to identify potentially pathogenic, disease-causing mutations. Using the sequence analysis pipeline from GATK, we identified 41,621 different variants in the patient s exome data. Data filtering was performed to exclude all known variants with a high frequency (> 1%) in the dbsnp129, the 1000 Genomes Project, GoNL, ESP6500 databases and in our in-house database. We then selected for coding variants in the remaining 325 variants and subsequently for nonsense, missense, splice site, and frame shift variants in concordance with autosomal recessive inheritance 146 EXOME SEQUENCING

148 (i.e. homozygous or compound heterozygous variants in one gene). This resulted in the identification of a homozygous mutation, c.542g>t; p.(gly181val) (NM_ ), in the SOD2 gene located in the second longest homozygous region on chromosome 6 (figure 2). This mutation was absent from known control populations (ESP6500, GoNL, and 1000 Genomes). Our ES data was also analysed for potential causal mutations in known cardiomyopathy genes, relevant metabolic and syndromic genes, and nuclear encoded mitochondrial genes, but no putative pathogenic mutations were identified. Sanger Sequencing, gene-panel-based resequencing and RT-PCR product analysis Using Sanger sequencing, the homozygous mutation c.542g>t; p.(gly181val) was confirmed in the affected child (figure 3) and in heterozygous form in her parents, but it was absent in her brother (data not shown). Furthermore, Sanger sequencing of the SOD2 gene in an additional DCM cohort of 27 different paediatric patients and 161 adult patients, and genepanel-based resequencing of the gene in more than 1,000 adult cardiomyopathy patients revealed no pathogenic SOD2 mutations. RT-PCR product analysis of RNA isolated from patient fibroblasts, and cultured both with and without cycloheximide, showed only a transcript of wild type size, indicating that this mutation has no effect on splicing. CHAPTER 3.2 Superoxide (O 2 ) substrate levels For superoxide substrate level measurements, hydroethidine was used as an intracellular probe to measure the levels of superoxide (O 2 ) in the patient fibroblasts. Notably, hydroethidine is not sensitive to H 2 O 2. Hydroethidine is a cell-permeable compound that interacts with O 2 to form ethidium or oxyethidium. The oxidation levels of hydroethidine measured in our in vitro assay indicated a significant increase of superoxide (O 2 ) levels in the fibroblasts of the patient comparable to the order of magnitude seen in complex I deficient fibroblasts (figure 4). What we could not directly determine from this data was whether the significant increase of O 2 levels resulted from a complex I deficiency or from abnormal SOD2 enzyme activity. However, mitochondrial respiratory chain enzyme activities (complexes I, II, III, IV, and V) were also measured and revealed no differences in the activity, suggesting SOD2 activity as the likely mechanism. SOD2 IN AUTOSOMAL RECESSIVE CARDIOMYOPATHY 147

149 control patient Figure 3. Sanger sequencing confirmed the presence of the homozygous SOD2 variant c.542g>t, p.(gly181val) in the affected patient (bottom) compared to control (top) and in heterozygous form in her parents (not shown). 148 EXOME SEQUENCING

150 Figure 4. The oxidation of hydroethidine analysis shows a significant increase of ROS (O 2 ) level as measured in both the nuclear and mitochondrial fractions in the fibroblasts of the patient compared to control fibroblasts. SOD2 3D structure: predicting the effect of the p.(gly181val) mutation Using the HOPE software we retrieved the 3D structure information of the SOD2 protein through the WHAT IF Web services, the Uniprot database and a series of DAS-servers, in order to predict the effect of the p.(gly181val) mutation on the protein structure. The Gly181 residue is part of a manganese/ iron superoxide dismutase domain, which is important for the main activity of the protein. The domain has a function in superoxide dismutase activity (oxidoreductase activity) and metal ion binding. According to the Uniprot database, four important amino acid residues are involved in the formation of the Mn-binding pocket that binds the manganese co-factor of the enzyme (accession number: P04179). These residues are His50, His98, Asp183 and His187. Interestingly, the aspartic acid residue of key importance (Asp183) is only two amino acids away from the Gly181 residue that was mutated in our patient. The increased size of the mutant residue is predicted to disturb the core structure of the manganese/iron superoxide dismutase domain and, as a consequence, the catalytic activity of the enzyme (figure 5). CHAPTER 3.2 DISCUSSION Using a combination of homozygosity mapping and ES in the patient, we detected a novel homozygous missense mutation, c.542g>t; p.(gly181val), in an evolutionarily highly conserved domain of the SOD2 gene located in the SOD2 IN AUTOSOMAL RECESSIVE CARDIOMYOPATHY 149

151 A B Figure 5. 3D structure of the SOD2 protein: (A) Overview of the SOD2 protein in ribbonpresentation. (B) Magnification of the part of the manganese/iron superoxide dismutase domain where the mutated residue is located. The protein backbone (grey) and the side chains of both the wild-type (green) and the mutant residue (red) are shown. The mutant residue is bigger than the wild-type residue, which may disturb the core structure of this domain and affect the catalytic activity of the enzyme. second longest homozygous region on chromosome 6. To our knowledge, this is the first report of a major role for mutated SOD2 in human disease. Two facts support the potential pathogenicity of this mutation. The first is that the mutation is located in the functionally important C-terminal manganese/ iron superoxide dismutase region of the respective protein. The second is that drastic differences between the size and the physico chemical characteristics of the wild-type glycine (which is the smallest of all residues and its presence is known to often provide flexibility to protein structures) and the mutant valine residues are predicted to disturb the core structure in this crucially important domain. Furthermore, according to the Uniprot database, the mutation is localized only two amino acids away from one of the four Histidine/Aspartic acid residues that are involved in the binding of the manganese co-factor. The role of the mutation Hydroethidine oxidation measurements indicated a significant increase in the levels of O 2 (one of the major ROS which are the physiological substrate of the SOD2 enzyme) in the fibroblasts of the patient; this substrate level was comparable in order of magnitude to the levels seen in complex-i-deficient fibroblasts. Since no deficiency in any of the mitochondrial respiratory chain 150 EXOME SEQUENCING

152 complexes I-V was seen, this significant increase in O 2 could probably be explained by the pathogenic effect of the c.542g>t; p.(gly181val) SOD2 mutation on the function of the encoded enzyme, leading to malfunctioning and accumulation of damaging oxygen radicals in the cells and increased oxidative stress. The role of superoxide dismutase in disease SOD2 belongs to the manganese/iron superoxide dismutase family which is one of the primary families of antioxidant enzymes in mammalian cells. These antioxidant enzymes protect cells from the damage caused by ROS. In eukaryotic cells, there are three SOD homologs: Cu/ZnSOD (SOD1), Mn/ FeSOD (manganese superoxide dismutase 2; SOD2) and extracellular SOD3. SOD2 is a mitochondrial matrix protein which converts superoxide anion (O 2 ) to H 2 O 2 which is then metabolized by glutathione peroxidase into H 2 O (Alscher et al). Oxidative stress is a deleterious process mediated by ROS, and it can lead to severe damage of cellular structures and their building blocks, including proteins, DNA and lipids (Valko et al). ROS are naturally formed during mitochondrial metabolism, and cells self-regulate their ROS levels by producing antioxidant enzymes (Starkov, 2008). Deficiency of one the antioxidant enzymes, such as SOD2, may affect any organ at any age, but most often affect organs with a high energy demand, such as the heart and brain, as is commonly observed in mitochondrial disorders (Meyers et al). Furthermore, it has been reported that oxidative stress and mutations in the SOD2 gene are involved in the pathogenesis of several diseases such as mitochondrial dysfunction, cancer, neurological disorders, diabetes, and many cardiovascular diseases including hypertension, atherosclerosis, and restenosis (Hedskog et al, Jenner, 2003, Louzao & van Hest, Cai & Harrison, Griendling & FitzGerald). There have also been reports of the involvement of other nuclear genes, such as TAZ (D Adamo et al), TXNRD2 (Conrad et al, Sibbing et al), DNAJC19 (Davey et al, Ojala et al), and SDHA (Levitas et al), in mitochondrial cardiomyopathy, and this also seems applicable to the current case. CHAPTER 3.2 Superoxide dismutase in cardiomyopathy Oxidative stress and disturbed mitochondrial respiratory function are known to play a substantial role in the development of heart failure (Huss & Kelly) and the role of the SOD2 protein in cardiomyopathy has previously been demonstrated in mice. Homozygous Sod2 knockout mice showed SOD2 IN AUTOSOMAL RECESSIVE CARDIOMYOPATHY 151

153 neonatal lethality due to neurodegeneration and cardiomyopathy (Li et al 1995). In addition, the intake of antioxidants improved their phenotypes of dilated cardiomyopathy and muscle fatigue and had beneficial effects on electrophysiological disturbances in heart and muscle (Koyama et al, Sunagawa et al). Interestingly, heterozygous SOD2+/- mice showed reduced SOD2 enzyme activity, yet did not exhibit any disease phenotype at 9 months of age (Li et al 1995). Likewise, the parents of the severely affected child described here, who are heterozygous carriers of the SOD2 mutation, did not show any cardiac abnormalities. Finally, chemotherapeutic (anthracyclin-induced) cardiomyopathy and heart failure is believed to be a side effect of superoxide radical accumulation leading to the induction of mitochondrial dysfunction in the heart (Thayer, 1988). In fact, this phenotype was successfully rescued in transgenic mice by the overexpression of SOD2 (Yen et al), underscoring the cardioprotective role of this enzyme in healthy individuals. CONCLUSIONS Here we have reported the successful use of a combined approach using homozygosity mapping and exome sequencing to identify the causal mutation in the mitochondrial protein, SOD2, in a child with severe neonatal cardiomyopathy. Protein conformation predictions and functional evaluation support the role of SOD2 deficiency in the abnormally elevated levels of oxidative stress found in our patient. Oxidative stress itself is known to be involved in the development of various diseases, including cardiomyopathies. The result from our patient adds a novel, nuclear-encoded disease gene to the list of genes involved in severe mitochondrial cardiomyopathies. ACKNOWLEDGEMENTS We thank the parents and sibling of the patient for participating in this study; Ludolf Boven and Sander Grefte for technical assistance; members of the Genomics Coordination Centre, UMCG, for assistance in data analysis; and Jackie Senior and Kate Mc Intyre for editing this manuscript. Rowida Almomani was supported by the Netherlands Heart Foundation (grant 2010B164). 152 EXOME SEQUENCING

154 REFERENCES Alscher RG, Erturk N, Heath LS: Role of superoxide dismutases (SODs) in controlling oxidative stress in plants. J Exp Bot. 2002;53: Burns KM, Byrne BJ, Gelb BD et al. New mechanistic and therapeutic targets for pediatric heart failure: report from a National Heart, Lung, and Blood Institute working group. Circulation. 2014;130:79-86 Cai H & Harrison DG: Endothelial dysfunction in cardiovascular diseases: the role of oxidant stress, Circ. Res. 2000;87:840 4 Conrad M, Jakupoglu C, Moreno SG et al. Essential role for mitochondrial thioredoxin reductase in hematopoiesis, heart development, and heart function. Mol Cell Biol. 2004;24: D Adamo P, Fassone L, Gedeon A et al. The X-linked gene G4.5 is responsible for different infantile dilated cardiomyopathies. Am J Hum Genet. 1997;61:862-7 Davey KM, Parboosingh JS, McLeod DR et al. Mutation of DNAJC19, a human homologue of yeast inner mitochondrial membrane co-chaperones, causes DCMA syndrome, a novel autosomal recessive Barth syndrome-like condition. J Med Genet. 2006;43: Dellefave L & McNally EM: The genetics of dilated cardiomyopathy. Curr Opin Cardiol. 2010;25: Fatkin D, Otway R, Richmond Z: Genetics of dilated cardiomyopathy. Heart Fail Clin. 2010;6: Griendling KK & FitzGerald GA: Oxidative stress and cardiovascular injury: part I: basic mechanisms and in vivo monitoring of ROS, Circulation 2003;108: Hedskog L, Zhang S, Ankarcrona M: Strategic role for mitochondria in Alzheimer s disease and cancer. Antioxid Redox Signal. 2012;16: Huss JM & Kelly DP: Mitochondrial energy metabolism in heart failure: a question of balance. J Clin Invest. 2005;115: Jenner P: Oxidative stress in Parkinson s disease. Ann. Neurol. 2003;53:S26 S38 Kindel SJ, Miller EM, Gupta R et al. Pediatric cardiomyopathy: importance of genetic and metabolic evaluation. J Card Fail (5): Kong A, Gudbjartsson DF, Sainz J et al. A high-resolution recombination map of the human genome. Nat Genet. 2002;31:241-7 Koyama H, Nojiri H, Kawakami S et al. Antioxidants improve the phenotypes of dilated cardiomyopathy and muscle fatigue in mitochondrial superoxide dismutase-deficient mice. Molecules : Levitas A, Muhammad E, Harel G et al. Familial neonatal isolated cardiomyopathy caused by a mutation in the flavoprotein subunit of succinate dehydrogenase. Eur J Hum Genet. 2010;18: Li H & Durbin R: Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics, 2009;25: Li H, Handsaker B, Wysoker A et al. The Sequence Alignment/Map format and SAMtools. Bioinformatics. 2009;25: Li Y, Huang TT, Carlson EJ et al. Dilated cardiomyopathy and neonatal lethality in mutant mice lacking manganese superoxide dismutase. Nat Genet. 1995;11: Louw JJ, Corveleyn A, Jia Y et al. Homozygous lossof-function mutation in ALMS1 causes the lethal disorder mitogenic cardiomyopathy in two siblings. Eur J Med Genet 2014; pii: S (14) doi: /j.ejmg Louzao I & van Hest JC: Permeability effects on the efficiency of antioxidant nanoreactors. Biomacromolecules. 2013;14: McKenna A, Hanna M, Banks E et al. The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res. 2010;20: Meyers DE, Basha HI, Koenig MK: Mitochondrial cardiomyopathy: pathophysiology, diagnosis, and management. Tex Heart Inst J. 2013;40: Ojala T, Polinati P, Manninen T et al. New mutation of mitochondrial DNAJC19 causing dilated and noncompaction cardiomyopathy, anemia, ataxia, and male genital anomalies. Pediatr Res 2012;72:432-7 Posafalvi A, Herkert JC, Sinke RJ et al. Clinical utility gene card for: dilated cardiomyopathy (CMD). Eur J Hum Genet. 2013;21. doi: / ejhg Rampersaud E, Siegfried JD, Norton N et al. Rare variant mutations identified in pediatric patients with dilated cardiomyopathy. Prog Pediatr Cardiol. 2011;31(1):39-47 Sibbing D, Pfeufer A, Perisic T et al. Mutations in the mitochondrial thioredoxin reductase gene TXNRD2 cause dilated cardiomyopathy. Eur Heart J 2011;32: Somsen G, Hovingh G, Tulevski I et al. Familial dilated cardiomyopathy. In: Cinical Cardioge- CHAPTER 3.2 SOD2 IN AUTOSOMAL RECESSIVE CARDIOMYOPATHY 153

155 netics. Baars H, Doevendans P, Smagt J, eds. Springer 2011, Starkov AA: The role of mitochondria in reactive oxygen species metabolism and signaling. Ann N Y Acad Sci 2008;1147:37-52 Sunagawa T, Shimizu T, Matsumoto A et al. Cardiac electrophysiological alterations in heart/ muscle-specific manganese-superoxide dismutase-deficient mice: prevention by a dietary antioxidant polyphenol. Biomed Res Int 2014: doi: /2014/ Thayer WS: Evaluation of tissue indicators of oxidative stress in rats treated chronically with adriamycin. Biochem Pharmacol 1988;37: Theis JL, Sharpe KM, Matsumoto ME et al. Homozygosity mapping and exome sequencing reveal GATAD1 mutation in autosomal recessive dilated cardiomyopathy. Circ Cardiovasc Genet. 2011;4(6): Theis JL, Zimmermann MT, Larsen BT et al. TN- NI3K mutation in familial syndrome of conduction system disease, atrial tachyarrhythmia and dilated cardiomyopathy. Hum Mol Genet 2014; pii:ddu297 Towbin JA, Lowe AM, Colan SD et al. Incidence, causes, and outcomes of dilated cardiomyopathy in children. JAMA 2006;296: Valko M, Leibfritz D, Moncol J et al. Free radicals and antioxidants in normal physiological functions and human disease. Int J Biochem Cell Biol 2007;39:44-84 Venselaar H, Te Beek TA, Kuipers RK et al. Protein structure analysis of mutations causing inheritable diseases. An e-science approach with life scientist friendly interfaces. BMC Bioinformatics 2010;11:548 Yen HC, Oberley TD, Gairola CG et al. Manganese superoxide dismutase protects mitochondrial complex I against adriamycin-induced cardiomyopathy in transgenic mice. Arch Biochem Biophys 1999;362: EXOME SEQUENCING

156

157

158 Chapter 3.3 One family, two cardiomyopathy subtypes, three disease genes: an intriguing case Anna Posafalvi, Nicole Corsten-Janssen, Paul A van der Zwaag, Jan G Post, Richard J Sinke, J Peter van Tintelen, Jan DH Jongbloed

159 ABSTRACT Pedigree information is often crucial in making decisions in clinical genetic counselling and diagnostics. Here we report on how pedigree information guided genetic analysis in a large, complex family with three affected individuals suffering from neonatal or late-onset dilated cardiomyopathy. Exome sequencing in combination with haplotype sharing tests led to causal mutations in the MYL2 (myosin light chain 2) and SOD2 (superoxide dismutase 2) genes in two deceased babies, respectively. Now targeted next generation sequencing based on a cardiomyopathy gene panel has revealed the possible role of another gene, JUP (junction plakoglobin), in one of the grandmothers affected with adult DCM. We present the 10-generation family pedigree that was constructed during the course of continuing genetic analyses and discuss aspects that directed diagnostic routing. We show the benefit of using pedigree data for the clinical genetic work on an intriguing familial cardiomyopathy case.

160 INTRODUCTION Idiopathic dilated cardiomyopathy (DCM) is a rare, progressive disease of the myocardium, usually exhibiting an autosomal dominant inheritance pattern and late onset of symptoms of heart failure (such as dyspnoea, syncope, and oedema), arrhythmias and thromboembolism. In some cases, however, cardiomyopathy may start at a very young age or just after birth, when it often proves to be lethal. This form of the disease (called neonatal or paediatric cardiomyopathy) is believed to be caused by autosomal recessive mutations. There are more than 50 genes known to be involved in cardiomyopathies, but since they only explain the disease in a relatively small proportion of patients, there must be novel genes to be discovered in many unsolved families (Teekakirikul et al, Almomani et al, see also chapter 3.2; Posafalvi et al). Pedigree information is often very important for genetic screening decisions in cardiomyopathy families. Here we report on how the family s pedigree guided our genetic analyses in an unusual case of an extended consanguineous family that is affected by two types of dilated cardiomyopathy. The family shows a regular adult-onset disease (putatively autosomal dominant (AD)) and a severe neonatal form (putatively autosomal recessive (AR)), with three possible disease-causing genes underlying the condition. CHAPTER 3.3 MATERIALS AND METHODS Patients The family pedigree is shown in figure 1. Patient X:1 died at the age of 6 months from a severe neonatal form of dilated cardiomyopathy, and was later found to carry a homozygous mutation (c.403-1g>c) in an acceptor splice site of intron 6 of the known DCM gene, myosin regulatory light chain 2 (MYL2). This mutation leads to the activation of a cryptic splice site, causing a frameshift in the C-terminal EF-hand motif of the encoded protein. Functional followup experiments showed that the calcium-binding properties of the mutant molecule were perturbed (Weterman et al). Parents IX:1 and IX:2 have since had a second affected baby who died from the same disease at age 4 months. This child was also homozygous for the MYL2 mutation, which caused huge emotional distress to the family. There is one healthy older sibling (not shown in pedigree) and the mother (IX:2) had three miscarriages before patient X:1 was born. The grandmother of patient X:1, VIII:2, was diagnosed with heart failure due to dilated cardiomyopathy at the age of 54 years. Patient X:2 also 1 FAMILY, 2 TYPES OF CARDIOMYOPATHY, 3 DISEASE GENES 159

161 Figure 1. The 10-generation pedigree of a family with both neonatal and lateonset cardiomyopathy. Square symbols (men), circles (women); black symbol (child affected by neonatal dilated cardiomyopathy), grey symbol (person affected by adultonset dilated cardiomyopathy), diagonal line through symbol (deceased). The pedigree is incomplete; it only indicates the degree of relationship between patients VIII:2, X:1 and X:2. The genealogical cross-links within the family were discovered by Eric Hennekam. suffered from a lethal, neonatal form of DCM and died three days after birth (see also chapter 3.1). Her parents, IX:3 and IX:4, and her brother (not shown in pedigree) were found to be unaffected. Homozygosity mapping SNP genotyping on a HumanCytoSNP-12 BeadChip 300K SNP array (Illumina, San Diego, CA, USA) and data analysis by genomestudio (Illumina) 160 EXOME SEQUENCING

162 and Microsoft Office Excel 2010 (Version 14.0; Microsoft, Redmond, WA, USA) software was performed as described earlier by van der Zwaag et al. We aimed to identify chromosomal regions which are homozygous in the patients or shared by the patients. Targeted NGS Sample preparation and targeted enrichment of a panel of 55 cardiomyopathy-related genes were performed according to the manufacturer s instructions (SureSelect XT Custom library, SureSelect Library prep kit, Agilent Technologies, Inc., Santa Clara, CA, USA), and as recently described in more detail by Sikkema-Raddatz et al. Sequencing was performed on a MiSeq sequencer (Illumina, San Diego, CA, USA) using 151 bp paired-end sequencing. Subsequent data analysis and variant filtering were performed with Next Gene (v2.2.1, Softgenetics, State College, PA) and Cartagenia (Cartagenia, Leuven, Belgium) software, as described in chapter 4.1. Variant classification To classify the variants identified, we performed a comprehensive analysis using information on the type of variation, the evolutionary conservation of the affected residue and the residing protein-region, the frequency of the variant in numerous control populations (such as 1000 Genomes, GoNL, and ESP6500), and the pathogenicity predicted by Alamut software (version 2.3.6), PolyPhen2, AGVGD, SIFT and MutationTaster. In addition, literature and database searches for further information were implemented. Finally, we uploaded the list of variants to the Combined Annotation Dependent Depletion online variant prioritizer tool (CADD, edu/info) to obtain a list of top candidate variants that could be considered to be likely pathogenic. CHAPTER 3.3 RESULTS The role of the pedigree in making genetic analysis decisions The initial three-generation pedigree of this family (not shown) indicated several possible modes of inheritance for the disease, including autosomal recessive or di-/oligogenic inheritance in particular. Initially, we considered the involvement of the same mutation in homozygous form in the neonatal cases and in heterozygous form in the late-onset DCM that affected the 1 FAMILY, 2 TYPES OF CARDIOMYOPATHY, 3 DISEASE GENES 161

163 grandmother. Since there were two cousins affected by the same, very rare, lethal (supposedly recessive) disease in the same family, we were curious if there was a more complex relation between the family members. After successfully extending the pedigree to 10 generations (figure 1), and discovering the multiple genealogical cross-links and distant consanguinity between the four parents of the affected babies, the inheritance model shifted towards a combination of the two forms mentioned above, i.e. the same gene, carrying an autosomal recessive inherited mutation, was anticipated to underlie disease in the neonatal cases X:1 and X:2, while the genetic cause of the disease would be independent and autosomal dominant in the grandmother (VIII:2). To find shared homozygous regions, homozygosity mapping was performed on DNA samples of the two neonatal patients. However, there was also still a possibility of finding homozygosity in X:2 and compound heterozygosity in X:1, carrying the same mutation as X:2 but in heterozygous form and combined with another, independent mutation, inherited from the grandmother (VIII:2) and causing the late-onset of her phenotype. Surprisingly, this mapping approach did not result in the identification of particularly large, shared homozygous regions (the longest such region was only 3.59 cm). Moreover, our analyses with the goal of identifying a homozygous region in X:2 that was heterozygously present in X:1 did not reveal any putative candidate gene regions either. These negative results were later supported by the identification of the homozygous MYL2 mutation in X:1 and the exclusion of this mutation in patient X:2. Subsequently, homozygosity mapping on the individual samples was performed to identify independent homozygous regions in the genome of X:2 that were not shared with X:1. When conducting this analysis for X:1, the MYL2 splice site mutation was found to be located in the 3 rd largest homozygous chromosomal region of 5.89 cm (12q24.11-q24.13), which happened to be the second longest autosomal homozygous fragment in the patient. The search for independent homozygous regions in the genome of X:2 that were not shared with X:1 was combined with exome sequencing. This resulted in the discovery of a causative, recessively inherited variant in the nuclear encoded mitochondrial enzyme superoxide dismutase (SOD2). This is located in the longest autosomal homozygous region (6q25.3-q26; the 2 nd longest such chromosomal region), spanning 4.26cM. Functional studies performed on the fibroblast samples of patient X:2 confirmed elevated levels of the substrate (oxygen free radicals), while the possible defect of any of the mitochondrial 162 EXOME SEQUENCING

164 complexes was excluded. Together, these findings strongly support the expected pathogenic role of the recessive mutation, c.542g>t; p.(gly181val), in this novel DCM gene (Almomani et al, see also chapter 3.2). Finally, as the 10-generation pedigree indicated that the DCM in the grandmother (VIII:2) could not be genetically related to the disease in X:1 or X:2, targeted NGS was performed on her DNA. Nevertheless, we excluded her as a carrier of either the MYL2 or SOD2 variants. Targeted sequencing identifies the third gene in VIII:2 The grandmother of the MYL2 patient, VIII:2, was shown to not carry the MYL2 mutation, neither could she potentially carry the SOD2 mutation. Targeted sequencing using our cardiomyopathy gene-panel revealed 152 variants in total in the 55 genes covered by the panel, of which four variants remained after filtering the data with the standard parameters of our analysis pipeline. Our routine classification method pointed to RYR2 (ryanodine receptor 2), c.3517a>g, p.(met1173val) and/or JUP (junction plakoglobin), c.746c>t, p.(thr249met) as likely pathogenic variants (see details in table 1). In both cases, the respective amino acid residues are highly conserved (the Met1173 of RYR2 at least up to chicken, and the Thr249 residue of JUP up to Drosophila), and all the protein effect prediction programs we used supported the likely pathogenicity. The RYR2 variant was absent from well-known population databases, while the JUP variant was reported only once in 8,600 European American alleles in the ESP database. Subsequent application of the CADD tool strongly suggested a primary pathogenic role, as the scaled C-score for the JUP variant is an extremely high Since CADD uses a logarithmic scale, this means a much higher predicted deleteriousness than that of the RYR2 missense variant, which scored only Additionally, a score of ~30 indicates that the variant belongs to the top 0.1% most deleterious of all substitutions that might theoretically occur in the human genome (CADD website and Kircher et al). Hence, without any further functional confirmation of the effects of the four variants, it seems most probable, from the currently available data, that JUP is the cause of the DCM in patient VIII:2 (although a digenic background for the disease development and causative roles for both variants cannot be excluded). CHAPTER FAMILY, 2 TYPES OF CARDIOMYOPATHY, 3 DISEASE GENES 163

165 Table 1 (A-D). Interpretation of putative pathogenicity of variants found in the adult DCM patient VIII:2 by gene-panel-based NGS. Only four variants remained after filtering for the region of interest, read depth, known polymorphisms and artefacts, and variant frequency in various databases. According to our standardized classification system based on evolutionary conservation (table A), prediction software (table B) and allele frequencies (table C), the RYR2/JUP missense variants were classified as the most likely to be causative. According to the CADD tool, JUP p.(thr249met) has the highest pathogenicity rank. Table D shows our final classification of the four variants. TABLE A variant conservation gene genomic, cdna and protein coordinate transcript ID nucleotide / amino acid PhyloP [-14.1;6.4] GERP [-12.36;6.18] conserved domain RYR2 SCN5A PKP2 JUP chr1: a>g c.3517a>g p.(met1173val) chr3: c>t c.3304g>a p.(ala1102thr) chr12: g>a c.419c>t p.(ser140phe) chr17: g>a c.746c>t p.(thr249met) NM_ weak /high SPRY (SPIa/Ryanodine receptor) NM_ not conserved / weak sodium transport-associated NM_ weak /moderate NM_ high / high armadillo TABLE B variant effect gene Grantham distance AGVGD PolyPhen2 SIFT Mutation Taster splicing RYR2 21 C15 possibly damaging (0.920) deleterious SCN5A 58 C0 benign (0.000) tolerated PKP2 155 C0 benign (0.007) tolerated JUP 81 C0 probably damaging (1.000) deleterious disease causing (p-value: 0.998) polymorphism (p-value: 0.996) polymorphism (p-value: 0.93) disease causing (p-value: 0.999) no effect no effect no effect no effect 164 EXOME SEQUENCING

166 TABLE C mutation database control database frequencies gene HGMD dbsnp rs number ESP GoNL RYR2 not present not present - not present (in about 13000) not present SCN5A not present - - PKP2 present (as disease causing) JUP not present db134, validated, clinical significance: probable non-pathogenic allele; MAF: 3/2184=0.001 db138, no validation, no frequency data (0/2184 in 1000genomes) rs rs not present (in about 13000) EA: A=26/G=8575 (ALL: 26/6503=0.004) EA: A=1/G=8599 (ALL: 1/6503=0) not present 6/996=0.006 not present TABLE D our classification CADD results final conclusion gene RawScore PHRED RYR2 LIKELY PATHOGENIC 2,746, SCN5A VOUS ,385 PKP2 Likely Benign 1,673, JUP LIKELY PATHOGENIC 5,033, CAUSATIVE VARIANT CHAPTER FAMILY, 2 TYPES OF CARDIOMYOPATHY, 3 DISEASE GENES 165

167 DISCUSSION Here we report on the unusual and rare example of a multigeneration, triple-consanguineous family that is affected by two distinct types of cardiomyopathy (namely, an adult onset form and a lethal neonatal form). Three different genes have been associated with the respective phenotypes in the three patients. The family s large, ten-generation pedigree played an important role in guiding the genetic analyses to uncover the mutations causing the cardiomyopathy disease types observed in the family. Although it was at first unexpected that the two affected babies would have different genetic causes of their disease, the genealogical reconstruction of the pedigree clearly indicated that they could have independent causes. This observation was further substantiated by the homozygosity mapping on the neonates DNA samples, which did not result in the identification of any obvious candidate regions. Based on the pedigree composition, one can quickly appreciate that the founder of the SOD2 mutation and the founder of the MYL2 mutation are likely to be ancestors from different branches of the family. The MYL2 mutation was most likely inherited from I:2, while potentially I:2, I:3, III:5 or III:6 could have been the individuals with the initial SOD2 mutation. Although the cardiac symptoms of both recessive patients seemed to be comparable at first glance, a systematic evaluation revealed clear differences in their phenotypes. Both were suffering from lethal neonatal cardiomyopathy, but the baby with the homozygous MYL2 mutation had myopathy with fibre type disproportion type 1, while the baby with the homozygous SOD2 mutation suffered from subependymal cysts (which is a possible manifestation of mitochondrial disease affecting the central nervous system), and she did not have typical skeletal muscle problems. For both genes, homozygosity mapping on the individual DNA samples (as shown in figure 2) supported their involvement in disease: MYL2 and SOD2 are localized in one of the longest autosomal homozygous regions of the patients (SOD2 in the 1 st, MYL2 in the 3 rd longest region). In fact, this was a major determinant in the completion of the exome sequencing data analysis for X:2. The fact that both patients carried relatively small homozygous regions (including those relatively large ones harbouring the causal mutations) supports the idea that both mutations are quite old and have been inherited from a founder many generations ago. We have identified the variant p.(thr249met) of JUP as the most likely cause of disease in VIII:2. Upon stringent filtering of the respective gene- 166 EXOME SEQUENCING

168 A) B) CHAPTER 3.3 Figure 2. SNP genotyping results for patients X:1 and X:2. Homozygous regions identified in patient X:1 (A) and patient X:2 (B) are shown. The genes identified as causative are marked in the 2 nd and 3 rd longest homozygous regions, respectively. 1 FAMILY, 2 TYPES OF CARDIOMYOPATHY, 3 DISEASE GENES 167

169 panel-based targeted NGS data, two of the remaining four variants (the above-mentioned JUP and a missense variant of RYR2) were predicted and classified as likely pathogenic. The same JUP variant was previously reported as an incidental finding in 1/1236 alleles of exome-sequenced individuals (non-selected for cardiomyopathy, arrhythmia, or sudden death, though some of them having an increased risk or a history of coronary artery disease), and was also classified as a variant of unknown clinical significance (Ng et al). This finding neither supports, nor excludes the putative role of the variant in cardiomyopathy development. However, the use of the new CADD online variant prioritizer tool pinpointed this JUP p.(thr249met) variant as an order of magnitude more likely to be disease-causing than the second best candidate, an RYR2 missense variant. According to the Uniprot database (accession number: P14923), the JUP variant is located in the third ARM repeat of the encoded junction plakoglobin protein, which spans amino acids It is involved in the interaction with desmocollin and desmoglein (Witcher et al), the cadherins known to play a role in cell adhesion and desmosome formation (Garrod et al). It is important to note that heterozygous missense mutations of neither JUP nor RYR2 have been associated with DCM so far, although both have been associated with arrhythmogenic right ventricular cardiomyopathy (ARVC). At this stage, we cannot exclude the possibility that the combination of both variants were the trigger to the development of DCM, or that an unidentified gene was also involved in the disease. However, according to our current data, the role of the JUP variant seems the most probable, and this could be easily followed up by functional experiments investigating the potentially impaired binding of desmocollin and desmoglein in the presence of the variant. Additionally, a recent study on 639 DCM patients suggested that the genetic overlap between various types of cardiomyopathy is much more extensive than previously estimated; it reported that 31% of the truly pathogenic mutations of DCM patients are mutations of typical ARVC-related genes and have been previously associated with ARVC (Haas et al). This intriguing family nicely exemplifies the importance of extensive analysis of the family history by pedigree reconstruction in genetic counselling. These genealogical studies led to an easier interpretation of why the MYL2 mutation was not found in patients VIII:2 and X:2, as well as of the rare recessive phenotypes caused by the two different genes. The common ancestor carrying the founder MYL2 mutation can be identified nine generations ago, and the one carrying the SOD2 mutation either nine or seven generations ago. 168 EXOME SEQUENCING

170 Our homozygosity mapping data supports the idea that both mutations are old and were transmitted through multiple generations. The improved understanding of the genetic background of the family has essential practical implications too. The parents of both X:1 and X:2 have been counselled that they have a 25% recurrence risk due to their carriership of an autosomal recessive disease. With the identification of the causative gene, they can now consider the reproductive options available (e.g. prenatal screening). This is of outmost importance - especially given that IX:1 and IX:2 have, in the meantime, had a second baby who was also affected by the same lethal disorder. Fortunately, prenatal diagnosis in a very recent pregnancy of IX:4 indicated that the foetus was not a homozygous carrier of the SOD2 mutation. Thus, with good genetic counselling and prenatal screening, this family should be able to avoid having any more seriously affected children. ACKNOWLEDGEMENTS We would like to acknowledge all those involved in counselling the distinct branches of this family. We thank Eric Hennekam, UMCU, for the pedigree construction; Jos Dijkhuis and his team in the Genome Diagnostics laboratory, Department of Genetics, UMCG, for technical support and performing the molecular genetic tests; and Jackie Senior and Kate Mc Intyre for editing this manuscript. CHAPTER 3.3 REFERENCES Almomani R, Posafalvi A, Herkert JC et al. Homozygous SOD2 mutation as a cause of severe neonatal dilated cardiomyopathy (manuscript in preparation, see also chapter 3.2) Garrod DR, Merritt AJ, Nie Z. Desmosomal cadherins. Curr Opin Cell Biol 2002;14: Haas J, Frese KS, Peil B et al. Atlas of the clinical genetics of human dilated cardiomyopathy. Eur Heart J 2014; pii: ehu301 Kircher M, Witten DM, Jain P et al. A general framework for estimating the relative pathogenicity of human genetic variants. Nat Genet 2014;46:310-5 Ng D, Johnston JJ, Teer JK et al. Interpreting secondary cardiac disease variants in an exome cohort. Circ Cardiovasc Genet 2013;6: Posafalvi A, Herkert JC, Sinke RJ et al. Clinical utility gene card for: dilated cardiomyopathy (CMD). Eur J Hum Genet 2013;21(10). doi: /ejhg Sikkema-Raddatz B, Johansson LF, de Boer EN et al. Targeted next-generation sequencing can replace Sanger sequencing in clinical diagnostics. Hum Mutat 2013;34: Teekakirikul P, Kelly MA, Rehm HL et al. Inherited cardiomyopathies: molecular genetics and clinical genetic testing in the postgenomic era. J Mol Diagn 2013;15: van der Zwaag PA, van Tintelen JP, Gerbens F et al. Haplotype sharing test maps genes for familial cardiomyopathies. Clin Genet 2011;79: Weterman MA, Barth PG, van Spaendonck-Zwarts KY et al. Recessive MYL2 mutations cause infantile type I muscle fibre disease and cardiomyopathy. Brain 2013:136; Witcher LL, Collins R, Puttagunta S et al. Desmosomal cadherin binding domains of plakoglobin. J Biol Chem 1996;271: FAMILY, 2 TYPES OF CARDIOMYOPATHY, 3 DISEASE GENES 169

171

172 CHAPTER 4 TARGETED SEQUENCING

173

174 Chapter 4.1 Gene-panel-based Next Generation Sequencing (NGS) substantially improves clinical genetic diagnostics in inherited cardiomyopathies Anna Posafalvi *, Jan DH Jongbloed *, Renee C Niessen, Paul A van der Zwaag, Yvonne Hoedemaekers, Birgit Sikkema-Raddatz, Jos Dijkhuis, Sebastiaan RD Piers, Katja Zeppenfeld, Rudolf A de Boer, Paul L van Haelst, Daniela QCM Barge-Schaapveld, Folkert W Asselbergs, Jasper J van der Smagt, Maarten P van den Berg, J Peter van Tintelen, Richard J Sinke *The first two authors contributed equally The last two authors contributed equally Manuscript submitted

175 ABSTRACT Background: Targeted next generation sequencing (NGS) is an attractive approach for the screening of multiple genes underlying genetic heterogeneous diseases, such as cardiomyopathies. We implemented an enrichment kit targeting 55 cardiomyopathy-related genes in our routine diagnostics work. The aim of this study was to determine the diagnostic yield, to evaluate the contribution of mutations in genes that were previously only infrequently or never screened for, and to obtain more insight into the suggested bigenic or multigenic inheritance patterns in a subset of patients. Methods and Results: DNA samples of 252 cardiomyopathy patients were analysed and their clinical characteristics collected. Patients with one or more variants labelled as likely pathogenic or pathogenic were considered to be resolved. Retrospective phenotype evaluation showed that of these 252 patients, 125 fulfilled the formal clinical criteria for a cardiomyopathy disease, 44 were suspected of having cardiomyopathy, and 37 had an unconfirmed diagnosis. We excluded 46 from further analysis. We identified pathogenic or likely pathogenic mutations in 107/206 (52%) patients: in 56% (40/72) of dilated cardiomyopathy (DCM) patients fulfilling the clinical criteria, and in 52% (12/23) of DCM-like patients. Truncating mutations in TTN were found in 14% of DCM patients. The yield in hypertrophic cardiomyopathy (HCM) and HCM-like patients was 46% (21/46) and 36% (4/11), respectively. In >50% of all our cardiomyopathy cases, we identified mutations in genes that were previously rarely analysed, and in 15% of cases, we found two or more pathogenic or likely pathogenic mutations. Conclusions: Targeted sequencing of cardiomyopathy genes results in a diagnostic yield of over 50%. In particular, our yield for genetic testing of DCM patients was substantially increased (approx. 55% vs % earlier). As this NGS method enables a large set of genes to be screened, including some infrequently studied genes, it opens up new avenues for exploring the role of rare genes and/or multiple mutations underlying inherited cardiomyopathies. Key Words: Next Generation Sequencing, targeted enrichment, clinical diagnostics, diagnostic yield, cardiomyopathy, genetics

176 INTRODUCTION Next Generation Sequencing (NGS) is one of the most promising developments in clinical genetics, including cardiogenetics, of the past few years (Jongbloed et al). This technique enables clinicians to make a genetic diagnosis within a short time frame for diseases which potentially have multiple genes underlying the phenotype. To apply NGS in a clinical diagnostics setting, the currently preferred method appears to be dedicated and reliable targeted enrichment, which provides sufficient specificity and sensitivity to replace the gold standard of Sanger sequencing (Sikkema-Raddatz et al). The use of several targeted enrichment methods (putatively) applicable for clinical diagnostics have been reported recently, with most of them using array-based enrichment and targeting a relatively small subset of genes (Harakalova et al, Almomani et al, Mook et al). However, approaches applying in-solution enrichment methods are also becoming increasingly popular (Sikkema-Raddatz et al, Lopes et al): these require smaller amounts of input DNA, while providing higher efficiency and better reproducibility, and being easier to handle (Querfurth et al, Shearer et al). Cardiomyopathies are a group of genetically and sometimes phenotypically overlapping heterogeneous disorders. The major subforms, in which over 50 disease genes have been identified, include arrhythmogenic right ventricular (ARVC), dilated (DCM), hypertrophic (HCM), left-ventricular non-compaction (LVNC), and restrictive (RCM) cardiomyopathies. Many of these genes are involved in different types of the disease (Teekakirikul et al, van Tintelen et al). In the pre-ngs era, the yield of diagnostic screening in well-defined patient cohorts varied widely: 35-70% in HCM (Christiaans et al, Pinto et al, Wilde & Behr), 20-25% in DCM (Wilde & Behr, Posafalvi et al, van Spaendonck-Zwarts et al, 2013), approximately 50% in ARVC (Cox et al, Quarta et al, te Rijdt et al), 25-40% in LVNC (Teekakirikul et al, Hoedemaekers et al), and approximately 35% in RCM (Teekakirikul et al). Since the number of genes associated with cardiomyopathies is large and still growing, this disease is an ideal candidate for the implementation of the rapidly developing NGS-based diagnostic tools. Several studies have already reported the screening of multiple cardiomyopathy genes (range 5 to 84 genes) within one experiment using NGS (Voelkerding et al, Gowrisankar et al, Zimmerman et al, Meder et al, Mook et al, Lopes et al, Pugh et al, Haas et al). Some of these studies only focused on cohorts of one type of cardiomyopathy. CHAPTER 4.1 DIAGNOSTIC YIELD OF CARDIOMYOPATHIES 175

177 We have recently demonstrated that the sensitivity, specificity and robustness of targeted NGS for cardiomyopathies is equal to those of Sanger sequencing (SS) (Sikkema-Raddatz et al). Subsequently, we constructed an improved enrichment kit targeting 55 cardiomyopathy genes and implemented this into our routine clinical diagnostic work. Here we report on the outcome and yield when we used this gene panel in a large cohort of cardiomyopathy patients. Patients were diagnosed with various types of cardiomyopathies including DCM, ARVC, HCM, LVNC, RCM, or with phenotypic characteristics related to cardiomyopathy, but not yet classified as a specific subtype. Their DNA was screened for variants using our 55 gene-panel-based method and, after data analysis, variant filtering and prioritization, we classified the variants found with the help of a strategy developed in-house. Our hypothesis was that implementing this test into routine diagnostics would lead to: (1) higher diagnostic yield, (2) identification of mutations in genes that were previously infrequently or never screened, and (3) provide more insight into the suggested bigenic or multigenic inheritance in a subset of cardiomyopathy patients. METHODS Patient material DNA was isolated according to standard operating procedures from peripheral blood samples obtained from 252 cardiomyopathy patients, who were referred to our laboratory for gene-panel-based genetic analysis. Informed consent to perform the diagnostic screening was obtained from all patients. They were referred to our department by four Dutch clinical genetics centres: Groningen, Leiden, Nijmegen and Utrecht. Targeted sequencing DNA fragment libraries were prepared according to the manufacturer s instructions (SureSelect XT Custom library, SureSelect Library prep kit, Agilent Technologies Inc., Santa Clara, CA, USA). The following experimental steps were performed: fragmentation of genomic DNA samples, end-repair, adapter ligation, size selection, and amplification of the purified product. Targeted enrichment was performed according to the manufacturer s instructions (Sureselect XT Custom library, Agilent Target Enrichment kit & Agilent SureSelect MP Capture Library kit, Agilent Technologies Inc.). 176 TARGETED SEQUENCING

178 Hybridization of the DNA fragment libraries with the capture probes for 55 selected genes was performed, followed by purification and barcoding of the captured fragments. Finally, equimolar pools of 12 samples were prepared. Sequencing was performed on a MiSeq sequencer (Illumina, San Diego, CA, USA) using 151 bp paired-end reads according to the manufacturer s instructions. The sample preparation, targeted enrichment and sequencing method has been described in detail by Sikkema-Raddatz et al. Capture probes of the following 55 cardiomyopathy-related genes were included in the custom designed, targeted enrichment kit. Their respective OMIM IDs are given in brackets, and genes marked by # were recently added to the improved version of the 48-gene enrichment kit described by Sikkema- Raddatz et al: ABCC9 (*601439), ACTC1 (*102540), ACTN2 (*102573), ANKRD1 (*609599), BAG3 (*603883), CALR3 (*611414), CAV3 # (*601253), CRYAB (*123590), CSRP3 (*600824), DES (*125660), DMD (*300377), DSC2 (*125645), DSG2 (*125671), DSP (*125647), DTNA # (*601239), EMD (*300384), EYA4 # (*603550), GATAD1 # (*614518), GLA (*300644), JPH2 (*605267), JUP (*173325), LAMA4 (*600133), LAMP2 (*309060), LDB3 (*605906), LMNA (*150330), MYBPC3 (*600958), MYH6 (*160710), MYH7 (*160760), MYL2 (*160781), MYL3 (*160790), MYPN (*608517), MYOZ1 (*605603), MYOZ2 (*605602), NEXN # (*613121), PKP2 (*602861), PLN (*172405), PRKAG2 (*602743), PSEN1 (*104311), PSEN2 (*600759), RBM20 (*613171), RYR2 (*180902), SCN5A (*600163), SGCD (*601411), SOD2 # (*147460), TAZ (*300394), TBX20 (*606061), TCAP (*604488), TMEM43 (*612048), TNNC1 (*191040), TNNI3 (*191044), TNNT2 (*191045), TPM1 (*191010), TTN (*188840), TXNRD2 # (*606448), VCL (*193065). CHAPTER 4.1 Sequence annotation and variant calling Data analysis was performed using the MiSeq reporter program (Illumina, San Diego, CA, USA) to generate fastq.gz output files. These were uploaded to the NextGene software (v2.2.1, Softgenetics, State College, PA, USA) and upon quality filtering, aligned to the reference genome (Human_v37.2). SNPs and indels were called, and the respective variant list was converted into the *.vcf file format for further analysis. Variant filtering, interpretation and prioritization The *.vcf files obtained from NextGene were uploaded into the Cartagenia software (Cartagenia, Leuven, Belgium), with which variant filtering and DIAGNOSTIC YIELD OF CARDIOMYOPATHIES 177

179 classification was performed (as summarized in figure 1 and described in the supplementary methods). Remaining variants were evaluated for their potential pathogenicity using in silico prediction tools and data, available via the Cartagenia and Alamut programs (versions and 2.3.6, respectively; Interactive Biosoftware, Rouen, France) and/or other resources (see table 1). We took various factors into account, such as the nature and location of the variants, the conservation of this area, the frequency of the variant in the general population (when it was available in any of the healthy or patient population databases), and the predicted pathogenicity of the variant according to multiple prediction programmes. Moreover, data on variants available from the scientific literature and from disease and variant databases (such as the Leiden Open Variant Databases (LOVDs) and the ARVD/C genetic variants Figure 1: Flowchart of the Cartagenia filtering tree used to determine our final variant list for analysis. Variant filtering strategy as used by the Cartagenia software. The input variant list contained on average 168 (± 24) variants per patient. Details of the filtering steps and strategy are described in the Supplementary methods. After performing the filtering steps, an average of 8 (± 6) variants per patient remained on the final variant list. These were classified after data-mining using Cartagenia and Alamut. Variants which were classified as benign or likely benign are regularly being added to our in-house database of managed variants (grey feed-back loop). The respective population control cohorts were: GoNL Genome of the Netherlands; 1000G 1000 Genomes project; ESP exomes from the NHLBI Exome Sequencing Project (ESP); and dbsnp the dbsnp database of NCBI. 178 TARGETED SEQUENCING

180 Table 1. Criteria for variant classification Classification Mutation type Criteria Benign (B) any MAF* >0.02 Likely benign (LB) intronic or synonymous No predicted # significant changes in RNA splicing Variant of uncertain significance (VOUS) Likely pathogenic (LP) Pathogenic (P) missense any intronic missense or synonymous missense nonsense or frame-shift intronic missense nonsense or frame-shift No predicted # significant changes in RNA splicing AND No, or only ¼ of prediction programs^ used suggest pathogenicity AND residue and surrounding residues not evolutionary conserved Variants which do not fit into any of the other categories, or for which the available information is contradictory Large effect on recognition of consensus splice site (±1 and 2) predicted # in gene for which association of such mutation with phenotype has not yet been established AND MAF* <0.001 or novel Prediction program # suggests that mutation creates cryptic splice site with large effect AND MAF* <0.001 or novel 3/4 or all 4 prediction programs^ used suggest pathogenicity AND Residue and surrounding residues are evolutionary conserved (at least up to chicken) AND MAF* <0.001 or novel OR variant does not completely fulfil the above criteria but there is other evidence available, such as functional proof or co-segregation data Truncating mutation in gene for which association of such mutation with phenotype has not yet been established AND MAF* <0.001 or novel Large effect recognition of consensus splice site (±1 and 2) predicted # in gene for which association of such mutation with phenotype has been established AND MAF* <0.001 or novel 3/4 or all 4 prediction programs^ used suggest pathogenicity AND Residue and surrounding residues are evolutionary conserved (at least up to chicken) AND MAF* <0.001 or novel AND Additional evidence like functional proof or co-segregation data Truncating mutation in gene for which association of such mutation with phenotype has been established AND MAF* <0.001 or novel Abbreviations: MAF: Minor Allele Frequency.*In population control cohorts (with at least 200 allele counts): 1000 Genomes, GoNL (Genome of the Netherlands), NHLBI Exome Sequencing Project (ESP); # RNA splicing prediction programs as provided by the Alamut software; ^Protein effect prediction programs as provided by the Cartagenia and Alamut software: SIFT, Polyphen, AGVGD, Mutation Taster CHAPTER 4.1 DIAGNOSTIC YIELD OF CARDIOMYOPATHIES 179

181 database ( was also taken into consideration for the classification. Based on all the available data, a final classification was performed. Our classification criteria are summarized in table 1: variants were classified as benign (B), likely benign (LB), variant of uncertain significance (VOUS), likely pathogenic (LP), or pathogenic (P). Finally, for the purpose of this study, we considered patients who were shown to carry one or more likely pathogenic and/or pathogenic variants as resolved cases. Patient inclusion We collected clinical data on 252 patients to retrospectively evaluate whether they fulfilled the formal diagnostic criteria for the respective subtypes of cardiomyopathy, as published for ARVC (Marcus et al, DCM (Mestroni et al), HCM (Gersh et al), and LVNC (Jenni et al). Patients were categorized as fulfilling criteria, suspected (not fulfilling criteria, but showing features of the respective subtype), or unconfirmed diagnosis (retrospective phenotype confirmation was not possible because no clinical data were available; patient categorized on the basis of the referral diagnosis to our laboratory), or excluded from the analysis if the clinical data could not confirm the suspected diagnosis of cardiomyopathy. Statistics Statistical calculations were performed using the Statistical Package for Social Sciences software, version 22.0 (IBM SPSS Statistics, Inc., Chicago, Illinois, USA). Descriptive statistics are reported as mean ± SD or number (percentage). Continuous variables were compared using the unpaired Student s t test or One-way ANOVA. Discrete variables were compared using Fisher s Exact test. Values of P<0.05 were considered statistically significant. RESULTS Summary of sequencing data We conducted targeted resequencing of 55 cardiomyopathy-related genes and the subsequent data analysis, variant interpretation and prioritization (see Methods and supplementary methods) in 252 patients. We were able to analyse, on average, 99.2 % of all targeted nucleotides with a coverage of at least 20x. The average coverage per nucleotide was 180 TARGETED SEQUENCING

182 433 (± 161) and varied between 134 (± 52) and 1126 (± 451). After MiSeq reporter quality filtering, vcf files were uploaded into the Cartagenia software and the regions of interest were selected, after which, on average, 168 (± 24) variants per patient remained. Upon using our filtering strategy, on average, 8 (± 6) variants per patient remained (range 1-49 variants), which were subjected to further interpretation and prioritization. Phenotype evaluation, patient inclusion and categorization The clinical data of the 252 patients were retrospectively evaluated (see figure 2 for inclusion/exclusion criteria, phenotype evaluation, cardiomyopathy subtype categorization, and genetic diagnostic outcome). We excluded 46 patients because they had: (1) a primary arrhythmia or conduction disorder evaluated with the purpose of identifying a potentially related, but late developing, cardiomyopathy (n=19), (2) a family history of cardiomyopathy, but they did not fulfil the criteria or were not suspected of having familial cardiomyopathy (n=12), (3) vascular disease (n=4), (4) syndromal cardiomyopathy (n=2), (5) congenital heart disease (n=4), or metabolic cardiomyopathy (n=2), or because these patients had been published elsewhere (n=3). Of the remaining 206 patients, 125 fulfilled the clinical criteria of the respective cardiomyopathy subtype ( fulfilling criteria ), 44 did not fulfil the criteria, but showed features of the disease and were thus suspected, and 37 patients had no detailed clinical records available, hence we analysed these 37 on the basis of the referral diagnosis to our laboratory ( unconfirmed diagnosis ) (figure 2). The group of 125 patients who fulfilled the criteria consisted of 71 DCM, 47 HCM, 3 ARVC, 3 LVNC, and 1 RCM cases (figure 2). The group of 44 suspected patients consisted of 23 DCM, 11 HCM, 6 ARVC, 1 LVNC, and 3 unspecified CM cases (figure 2). And the group of 37 patients with an unconfirmed diagnosis consisted of 16 DCM, 17 HCM, 1 ARVC, 1 LVNC, and 2 unspecified CM cases (figure 2). For an overview of all patients, including the mutations they carry, see supplementary table 2. CHAPTER 4.1 Mutation spectrum We identified 142 pathogenic or likely pathogenic mutations (113 different mutations in total) in 34 genes. Of the 113 different mutations, 13 were classified as pathogenic, while the remaining 100 were classified as likely pathogenic (supplementary table 2). DIAGNOSTIC YIELD OF CARDIOMYOPATHIES 181

183 Figure 2: Flow chart of patient inclusion, phenotype evaluation and categorization, and genetic diagnosis. After referral to our laboratory, patients were retrospectively phenotyped and categorized into cardiomyopathy subtypes (fulfilling criteria, suspected disease, or unconfirmed) and were either genetically resolved or remained without a genetic diagnosis. The number of patients in the different groups/categories are indicated. Among those mutations identified in multiple patients, we saw several well-known, Dutch founder mutations: c.2373dupg (n=5), c.2864_2865delct (n=2) and c.2827c>t; p.r943* (n=1) in MYBPC3 (Christiaans et al) and c.40_42delaga (n=4) in PLN (van der Zwaag et al, 2012). We identified putatively truncating TTN mutations (i.e. mutations leading to a premature stop codon or nonsense, frame shift, or consensus splice site mutations) in 21/206 (10%) patients. Of these, 11/125 (9%) were identified in patients who fulfilled our criteria, with 10/71 (14%) in DCM patients and 1/47 (2%) HCM patients. In addition, 4/44 (9%) were identified in suspected patients, with 3/23 (13%) in DCM-like patients and 1/23 (9%) in HCM-like patients. 182 TARGETED SEQUENCING

184 Table 2: Diagnostic yield in patients who fulfilled the clinical criteria for a cardiomyopathy subtype CM* subtype negative P LP positive (P + LP) total ARVC 2 (67%) 1 (33%) 0 1 (33%) 3 DCM 32 (45%) 4 (6%) 36 (50%) 40 (56%) 72 HCM 25 (53%) 6 (13%) 15 (33%) 21 (46%) 46 LVNC 2 (67%) 1 (33%) 0 1 (33%) 3 RCM 0 1 (100%) 0 1 (100%) 1 CM total 61 (49%) 13 (10%) 51 (41%) 64 (51%) 125 Table 3: Diagnostic yield in patients with a suspected cardiomyopathy subtype CM* subtype negative P LP positive (P + LP) total ARVC 3 (50%) 2 (33 %) 1 (17%) 3 (50%) 6 DCM 11 (48%) 2 (9%) 10 (43%) 12 (52%) 23 HCM 7 (64%) 0 4 (36%) 4 (36%) 11 LVNC (100%) 1 (100%) 1 RCM CM 1 (33%) 0 2 (67%) 2 (67%) 3 total 22 (50%) 4 (9%) 18 (41%) 22 (50%) 44 Table 4: Diagnostic yield in patients with an unconfirmed diagnosis CM* subtype negative P LP positive (P + LP) total ARVC 1 (100%) DCM 7 (41%) 3 (18%) 7 (41%) 10 (59%) 17 HCM 7 (41%) 2 (12%) 8 (47%) 10 (59%) 17 LVNC 1 (100%) RCM CM (100%) 1 (100%) 1 total 16 (43%) 5 (14%) 16(43%) 21 (57%) 37 *Abbreviations: ARVC: arrhythmogenic right ventricular cardiomyopathy, CM: cardiomyopathy unspecified, DCM: dilated cardiomyopathy, HCM: hypertrophic cardiomyopathy, LVNC: left ventricular non-compaction, LP: likely pathogenic (sometimes together with one or more LPs), P: pathogenic (sometimes together with one or more LPs), RCM: restrictive cardiomyopathy. CHAPTER 4.1 Finally, we found TTN mutations in 6/37 (16%) patients with an unconfirmed diagnosis, with 5/16 (31%) in DCM patients and 1/17 (6%) HCM patients (see supplementary table 2). When categorized according to mutation type, we had 69 different missense mutations, 27 different truncating mutations (frame shift and nonsense mutations) and 17 different splice site mutations. In the 206 patients described here, we did not identify pathogenic or likely pathogenic mutations DIAGNOSTIC YIELD OF CARDIOMYOPATHIES 183

185 in the ACTC1, BAG3, CAV3, CRYAB, DSG2, EYA4, GATAD1, LAMP2, MYL2, MYOZ1, MYOZ2, PRKAG2, PSEN1, PSEN2, SGCD, SOD2, TAZ, TBX20, TCAP, TMEM43, and TXNRD2 genes. Out of the 113 different disease-associated mutations that we have identified, only 32 (28%) were found in the Human Gene Mutation Database (HGMD; see supplementary table 2); in addition, six of the mutated nucleotides were known in HGMD, but reported as mutated into another nucleotide than the one seen in our patients. Diagnostic yield The overall diagnostic yield for our patient cohort was 52% (107/206) (supplementary table 3). Of these, 77/206 (37%) patients carried one pathogenic or likely pathogenic variant, and 30/206 (15%) patients carried more than one mutation. At least one pathogenic mutation was identified in 22/206 (11%) patients. In the group of patients fulfilling the clinical criteria for cardiomyopathy, the diagnostic yield was 64/125 (51%) (table 2; figure 2). In patients suspected of having cardiomyopathy, this was 22/44 (50%) (table 3; figure 2). Interestingly, the diagnostic yield was the highest (although this was not significant) in our 37 patients with an unconfirmed diagnosis: 21/37 (57%) (table 4; figure 2). This could be attributed mainly to the higher yield for HCM patients (10/17; 59%) compared to the yield for patients who fulfilled the clinical criteria for HCM and the yield for suspected HCM cases. Pathogenic mutations were identified in 13/125 (10%) patients who fulfilled the criteria, in 4/44 (9%) suspected cases, and in 5/37 (14%) patients with an unconfirmed diagnosis. When cardiomyopathy subtypes were taken into consideration, the differences between the fulfilling criteria and suspected subcategories were observed for both DCM and DCM-like, and HCM and HCM-like patients: for both subtypes, we observed a higher, although insignificant, diagnostic yield in the group of patients fulfilling the clinical criteria than in those who were only suspected of having the subtype. In 40/72 (56%) DCM patients who fulfilled the criteria and in 12/23 (52%) suspected DCM patients, we identified pathogenic or likely pathogenic mutations (figure 2; table 2). Likewise, the diagnostic yield was 21/46 (46%) for HCM patients who fulfilled the criteria and 4/11 (36%) for HCM-like patients (figure 2; table 2). Unfortunately, we could not make similar comparisons for the other cardiomyopathy subtypes because the numbers of patients were too low. It is important to note, however, that we found an underlying genetic cause of disease in 3/4 (75%) of the 184 TARGETED SEQUENCING

186 patients with an unspecified cardiomyopathy (3 from the suspected group and 1 from the unconfirmed diagnosis group). Effect of including genes that were previously infrequently or never screened for To gain insight into the effect of including less prevalent genes in the panel on our diagnostic yield, we investigated how many patients were found to carry mutations in genes that were routinely screened in the pre-ngs era (the pre- NGS genes ), and compared this with the number of patients with mutations in the less prevalent genes (i.e. those genes previously not included in routine diagnostic screening). The following genes were considered as the pre-ngs genes in the DCM, HCM and ARVC subtypes: LMNA, MYBPC3, MYH6, MYH7, PLN, SCN5A and TNNT2 in DCM (Wilde & Behr, Posafalvi et al, Hershberger & Siegfried); ACTC1, MYBPC3, MYH7, MYL2, MYL3, TNNI3, TNNT2 and TPM1 in HCM (Wilde & Behr, and Pinto et al); and DSC2, DSG2, DSP, JUP, PKP2 and PLN in ARVC (Quarta et al, Wilde & Behr, te Rijdt et al). This study revealed that mutations were identified in the pre-ngs genes in 34% (21/62) of DCM and DCM-like cases (38% in criteria-positive cases), 40% (14/35) of HCM and HCMlike cases (48% in criteria-positive cases), and 75% (3/4) of ARVC and ARVC-like cases, which means that 25-66% of cases are explained by mutations in genes previously rarely investigated in the particular type of cardiomyopathy. Notably, when multiple mutations were identified in a patient and these included a pre- NGS gene, he or she was considered to be resolved. We also evaluated how many patients were now found to carry mutations (i.e. at least one if multiple mutations were found) in genes that were previously not reported as being involved in that specific subtype (based upon van Tintelen et al). As expected, for the large number of genes reported to be involved in DCM, only two patients (2/62; 3%) were found to carry a mutation in a non-dcm gene (CALR3 and MYL3, both in criteria-positive cases). In HCM and HCM-like cases, 6/35 (17%) patients carried mutations in genes not previously reported in HCM (in 3 criteria-positive cases: LAMA4, ABCC9, and DSP; one suspected case: DES & DTNA (both in one patient); and 2 unconfirmed cases: DSP & RYR2 (both in one patient) and RYR2). In the other subtypes, this was 1/4 (25%) in ARVC and ARVC-like patients (DMD in the ARVC-like patient), and 1/2 (50%) in LVNC and LVNC-like cases (JUP in the LVNC-like case). CHAPTER 4.1 DIAGNOSTIC YIELD OF CARDIOMYOPATHIES 185

187 Bigenic or multigenic inheritance Finally, a group of 30/206 patients (15%) carried multiple pathogenic or likely pathogenic variants: 15/125 (12%) of patients who fulfilled the criteria, 7/44 (16%) of the suspected patients, and 8/37 (22%) of patients with an unconfirmed diagnosis. Of these, 12 patients carried two pathogenic/likely pathogenic mutations (2 patients had one pathogenic and one likely pathogenic mutation) and 3 patients carried three pathogenic/likely pathogenic mutations (1 with one pathogenic and two likely pathogenic mutations; two with three likely pathogenic mutations) in the category of criteria-positive cases; seven patients in the suspected category all carried two likely pathogenic mutations; and in the unconfirmed category, six patients carried two pathogenic/ likely pathogenic mutations (3 patients with one pathogenic and one likely pathogenic mutation, and 3 with two likely pathogenic mutations) and two patients carried three likely pathogenic mutations. Patient characteristics Of the 206 patients we included, 126 (61%) were male and 80 (39%) female. There were no significant differences in sex distribution between the three patient categories (see supplementary table 4). The mean age at diagnosis of patients without a mutation compared to those carrying one mutation, and those with multiple mutations, did not differ significantly: 52 (± 13) years, 48 (± 18) years and 48 (± 19) years, respectively (for age of diagnosis see supplementary table 2). In addition, no significant differences regarding the age of diagnosis were noted among patients who fulfilled the clinical criteria when comparing the subgroups without a mutation, carrying a single mutation, or carrying multiple mutations. This was also true for the suspected and unconfirmed patient groups. In the subcategory of patients who fulfilled the clinical criteria, we evaluated the Left Ventricular Ejection Fraction (LVEF) in DCM patients, and the InterVentricular Septum thickness (IVS) in HCM patients, and investigated potential differences between patients carrying no, one or multiple mutations (see supplementary tables 5 and 6). Notably, comparable analyses within the other subcategories was not relevant as patient numbers were too low. In the DCM subcategory there was no significant difference between the mean LVEF for patients carrying no, one or multiple mutations (supplementary table 5). We also observed no differences with respect to the mean IVS in HCM patients 186 TARGETED SEQUENCING

188 who carried either no or one mutation: 19.5 (± 4.4) mm versus 18.2 (± 2.9) mm (p=0.29) (supplementary table 6). The IVS was available for only 2/4 HCM cases with multiple mutations, so that a comparison was not appropriate. DISCUSSION The diagnostic yield using our gene-panel-based NGS method is more than half (52%) for all the cases of hereditary cardiomyopathy: 51% in patients who fulfilled the clinical criteria, 50% in patients suspected of having a cardiomyopathy, and 57% in patients with an unconfirmed diagnosis. A substantial increase in diagnostic yield was observed particularly for DCM and DCM-like patients: 56% for criteria fulfilling, 52% for suspected and 56% for unconfirmed cases, in comparison to the yield of 20% in a cohort of 418 Sanger-sequenced patients (van Spaendonck et al, 2013). Our yield is comparable to that reported by two recent studies which used NGS in large DCM cohorts: 37% (121 patients; Pugh et al) and 73% (639 patients; Haas et al). The difference in these two studies may, in part, be explained by the number of genes that they screened for: 46 and 84 genes, respectively. In addition, the differences might also reflect differences in the pathogenicity classification criteria used. Pugh et al included variants in an additional class, the VOUS-favour pathogenic class, which we would mostly have included in our likely pathogenic class. Moreover, Haas et al included disease causing mutations according to the HGMD database, but this means they may also have included variants that were later proven not to be causal, like the c.419c>t, p.ser140phe variant in PKP2, which was recently shown not to co-segregate with disease in a Dutch ARVC family (Groeneweg et al). Compared to previous results obtained by Sanger sequencing of the most prevalent genes, our approach leads to a significantly higher diagnostic yield, which can be largely attributed to truncating mutations identified in the TTN gene. In cases which fulfilled the criteria, the percentage of TTN mutations was 14%, comparable to the cohorts studied by Pugh et al and Haas et al, but slightly lower than previously reported (18% in sporadic cases and 25% in familial DCM cases; Herman et al). The diagnostic yield for HCM and HCM-like cases was 46% for those that fulfilled the clinical criteria, 36% for suspected cases, and 59% for unconfirmed cases. The yield from criteria-positive cases is comparable to that seen in Dutch HCM patients in the pre-ngs era (approx. 50%; Christiaans et al). Notably, Lopes et al has also reported a slightly higher diagnostic yield of 57% after high-throughput sequencing of 41 genes in 223 HCM patients. The fact CHAPTER 4.1 DIAGNOSTIC YIELD OF CARDIOMYOPATHIES 187

189 that the diagnostic yield in criteria-positive cases is not significantly higher after the gene-panel-based sequencing in our study is rather surprising, certainly given the fact that mutations were identified in the most prevalent HCM genes in only 48% of these patients (i.e. the genes which were also screened in the pre-ngs era). This lack of increase in yield can probably be explained by the fact that HCM screening has been available in the Netherlands since 1996, and many of the more severe HCM cases have already undergone genetic diagnostic screening. In line with this observation, Hofman et al reported that the yield of DNA testing in arrhythmia syndromes has dropped over a 15-year period. They thought this was due, over time, to more patients with unclear diagnoses being referred for DNA testing, often coming from relatively small families. Slightly higher diagnostic yields are being observed in patients who fulfil the clinical criteria for DCM and HCM (see tables 2 and 3), compared to the yield in the suspected patients: 56% versus 52% for DCM and DCMlike cases, and 46% versus 36% for HCM and HCM-like cases. However, as these differences are not statistically significant, larger cohorts need to be studied to reveal whether diagnostic yields might be higher in patients who fulfil the clinical criteria. Importantly, a pathogenic or likely pathogenic mutation was also identified in a relatively large number of patients suspected of having cardiomyopathy. In a 2010 position paper, the European Society of Cardiology working group on Myocardial and Pericardial Diseases stated that genetic testing is not indicated for the diagnosis of a borderline or doubtful cardiomyopathy except for selected cases in the setting of expert teams (Charron et al). However, the identification of a mutation in criteria-negative cases may be helpful in directing further diagnostic work and family-screening. In ARVC, for example, the identification of a pathogenic mutation is one of the task force s major criteria (Marcus et al). One remarkable observation made in this study is that a high diagnostic yield was found within the subcategory of unconfirmed diagnoses, mainly in HCM and DCM patients. This might be explained by the fact that these patients were referred by other Dutch clinical genetics centres or by our own cardiogenetics team, who offer consultations to patients in other (regional) hospitals who have been selected by the local cardiologists. It is possible that the local physicians used higher phenotypic thresholds before referring patients for genetic testing. A substantial number of pathogenic or likely pathogenic mutations were identified in genes that would rarely or never have been analysed in the pre- 188 TARGETED SEQUENCING

190 NGS era: only 40-50% of our mutations were identified in pre-ngs genes. A comparable finding was reported by Haas et al as they frequently found mutations in desmosomal, channelopathy and HCM genes in their DCM cohort; this is a subset of genes that would not have been screened for in this patient group, or only irregularly, in the pre-ngs era. A considerable part of these mutations were identified in the titin (TTN) gene, i.e. in 10% in our total cohort and in up to 15% in DCM patients who fulfilled the clinical criteria. Because of its large size, the TTN gene was hardly ever examined in detail in the pre-ngs era, but new technologies now enable its routine screening. As reported before by Herman et al, we also identified truncating TTN variants in HCM and HCM-like cases, but this was in a percentage that did not differ from healthy controls. This underscores the importance of cosegregation analyses in larger patient cohorts to find further support for the pathogenicity of these mutations, as recently published for some families with truncating TTN mutations (van Spaendonck et al, 2014). Similarly, the role of likely pathogenic mutations in other rarely studied genes should be further investigated, including through co-segregation analyses. As expected, we also identified known Dutch founder mutations, for instance, the c.40_42delaga; p.(arg14del) mutation in the PLN gene (van der Zwaag et al, 2012; 2013) was identified in four cases and the c.2373dupg mutation in MYBPC3 (Alders et al, Michels et al, Christiaans et al) was identified in five cases. In contrast, only 30% of the mutations we identified were already known from the HGMD database, which underscores the importance of thorough data-mining and interpretation, and of the sharing of data for the careful classification of variants. Another interesting observation was that 30/206 index patients (15%) were found to carry two or more pathogenic or likely pathogenic variants. Since earlier DNA diagnostic work was stopped once a pathogenic or likely pathogenic variant was found in one of the candidate genes, this is a highly interesting finding, and it provides support for the multigenic background of cardiomyopathies, which has recently been addressed by Roncarati et al, Bauce et al, Xu et al, Bao et al, and Rigato et al. The phenotype in multiple mutation carriers is generally believed to be more severe and/or to manifest at a younger age. However, we did not observe either of these. On the one hand, this might be because our group of patients was too small, or because several severe cases in which we only identified one mutation using our approach were actually bi- or multigenic, but they might have had other mutations in CHAPTER 4.1 DIAGNOSTIC YIELD OF CARDIOMYOPATHIES 189

191 genes not included in our gene panel. We cannot therefore exclude that some of these variants were wrongly classified and they are not truly pathogenic. Future co-segregation and functional analysis should offer more insight into this. On the other hand, we cannot rule out that in some patients with multiple likely pathogenic mutations, their cardiomyopathy only developed because of the presence of more than one mutation, and that these mutations individually would not have resulted in disease development, or only at very old age (bi-/oligogenic inheritance). Finally, a genetic diagnosis is still lacking for approximately 45% of our patients. There are several possible reasons for this: (1) the major cause of disease in some patients is not genetic; (2) mutations lie in regulatory regions of currently screened genes or in regulatory RNAs; (3) patients might carry deletions/duplications of one or more exons and these cannot be detected with the current approach; (4) patients may have causal mutations in other unknown or rare cardiomyopathy genes; and (5) the underlying cause is truly oligogenic and therefore difficult or impossible to deduce from gene-panel-based NGS analyses. Other strategies are needed to identify the missing inheritance in those patients, including exome or genome sequencing strategies, and/or RNA sequencing, and where applicable these strategies could be combined with linkage or linkage-like methods (van der Zwaag et al, 2011). Together, our gene-panel-based approach allows a more complete identification of disease-causing mutations in cardiomyopathy patients. We show that this is a valuable tool for routine diagnostics, and that it will facilitate more accurate and/or personalized counselling of patients and their families. Our approach results in a substantial increase in the diagnostic yield for DCM patients compared to the results from Sanger sequencing of the most prevalent genes (>50% vs %). In addition, slightly higher diagnostic yields were achieved for patients fulfilling the DCM and HCM clinical criteria compared to patients with suspected disease; however, this must be confirmed in larger cohorts. Finally, our gene-panel-based approach enables the large-scale exploration of rare genes and multiple mutations underlying the inherited cardiomyopathies, for which the clinical relevance now has to be validated. ACKNOWLEDGEMENTS We would like to thank the clinical geneticists, genetic counsellors and cardiologists for counselling and referring their patients to the Department of Genetics, UMCG, for 190 TARGETED SEQUENCING

192 routine diagnostic screening; the molecular genetics team of the Genome Diagnostics section for technical assistance; staff members of the Genome Diagnostics section for help in variant interpretation and classification; and Jackie Senior for editing this manuscript. SOURCES OF FUNDING This study was supported by a grant from the Doelmatigheidsfonds of the University Medical Center Groningen (to JDH Jongbloed, JP van Tintelen and RJ Sinke); a grant from the NutsOhra foundation (project to JDH Jongbloed, MP van den Berg, JP van Tintelen and RJ Sinke), and grants 2007B132 and 2010B164 from the Netherlands Heart Foundation (to JDH Jongbloed, PA van der Zwaag and JP van Tintelen). Disclosures: The authors declare no conflicts of interest. REFERENCES Alders M, Jongbloed R, Deelen Wet al. The 2373insG mutation in the MYBPC3 gene is a founder mutation, which accounts for nearly one-fourth of the HCM cases in the Netherlands. Eur Heart J 2003;24: Almomani R, can der Heijden J, Ariyurek Y et al. Experiences with array-based sequence capture; toward clinical applications. Eur J Hum Genet 2011;19:50-55 Bao JR, Wang JZ, Yao Y et al. Screening of pathogenic genes in Chinese patients with arrhythmogenic right ventricular cardiomyopathy. Chin Med J (Engl) 2013;126: Bauce B, Nava A, Beffagna G et al. Multiple mutations in desmosomal proteins encoding genes in arrhythmogenic right ventricular cardiomyopathy/dysplasia. Heart Rhythm 2010;7:22-29 Charron P, Arad M, Arbustini E et al. European Society of Cardiology Working Group on Myocardial and Pericardial Diseases Genetic counselling and testing in cardiomyopathies: a position statement of the European Society of Cardiology Working Group on Myocardial and Pericardial Diseases. Eur Heart J 2010;31(22): Christiaans I, Nannenber EA, Dooijes D et al. Founder mutations in hypertrophic cardiomyopathy patients in the Netherlands. Neth Heart J 2010;18: Cox MG, van der Zwaag PA, van der Werf C et al. Arrhythmogenic right ventricular dysplasia/ cardiomyopathy: pathogenic desmosome mutations in index-patients predict outcome of family screening: Dutch arrhythmogenic right ventricular dysplasia/cardiomyopathy genotype-phenotype follow-up study. Circulation 2011;123(23): Gersh BJ, Maron BJ, Bonow RO et al. ACCF/AHA guideline for the diagnosis and treatment of hypertrophic cardiomyopathy: a report of the American College of Cardiology Foundation/ American Heart Association Task Force on Practice Guidelines. Circulation 2011;124:e783 e831 Gowrisankar S, Lerner-Ellis JP, Cox S et al. Evaluation of second-generation sequencing of 19 dilated cardiomyopathy genes for clinical applications. J Mol Diagn 2010;12: Groeneweg JA, van der Zwaag PA, Jongbloed JD et al. Left-dominant arrhythmogenic cardiomyopathy in a large family: associated desmosomal or nondesmosomal genotype? Heart Rhythm 2013;10: Haas J, Frese KS, Peil B et al. Atlas of the clinical genetics of human dilated cardiomyopathy. Eur Heart J 2014; pii: ehu301. [Epub ahead of print] Harakalova M, Mokry M, Hrdlickova B et al. Multiplexed array-based and in-solution genomic enrichment for flexible and cost-effective targeted next-generation sequencing. Nat Protoc 2011;6(12): Herman DS, Lam L, Taylor MRG et al. Truncations of titin causing dilated cardiomyopathy. N Engl J Med 2012;366: Hershberger RE & Siegfried DE. Update 2011: Clinical and genetic issues in familial cardiomyopathy. J Am Coll Cardiol 2011;57: CHAPTER 4.1 DIAGNOSTIC YIELD OF CARDIOMYOPATHIES 191

193 Hoedemaekers YM, Caliskan K, Michels M et al. The importance of genetic counseling, DNA diagnostics, and cardiologic family screening in left ventricular noncompaction cardiomyopathy. Circ Cardiovasc Genet 2010;3(3):232-9 Hofman N, Tan HL, Alders M et al. Yield of molecular and clinical testing for arrhythmia syndromes: report of 15 years experience. Circulation 2013;128: Jenni R, Oechslin E, Schneider J et al. Echocardiographic and pathoanatomical characteristics of isolated left ventricular non-compaction: a step towards classification as a distinct cardiomyopathy. Heart 2001;86(6): Jongbloed JDH, Pósafalvi A, Kerstjens-Frederikse WS et al. New clinical molecular diagnostic methods for congenital and inherited heart disease. Expert Opin Med Diagn 2011;5(1):9-24 Lopes LR, Zekavati A, Syrris P et al. Genetic complexity in hypertrophic cardiomyopathy revealed by high-throughput sequencing. J Med Genet 2013;50(4): Marcus FI, McKenna WJ, Sherrill D et al. Diagnosis of arrhythmogenic right ventricular cardiomyopathy/dysplasia: proposed modification of the task force criteria. Circulation 2010; 121: Meder B, Haas J, Keller A et al. Targeted next-generation sequencing for the molecular genetic diagnostics of cardiomyopathies. Circ Cardiovasc Genet 2011;4: Mestroni L, Maisch B, McKenna WJ et al. Guidelines for the study of familial dilated cardiomyopathy. Eur Heart J 1999;20: Michels M, Soliman OII, Kofflard MJ et al. Diastolic abnormalities as the first feature of hypertrophic cardiomyopathy in Dutch myosin-binding protein C founder mutations. JACC Cardiovasc Imaging 2009;2:58-64 Mook ORF, Haagmans MA, Soucy JF et al. Targeted sequence capture and GS-FLX Titanium sequencing of 23 hypertrophic and dilated cardiomyopathy genes: implementation into diagnostics. J Med Genet 2013; doi: /jmedgenet Pinto YM, Wilde AA, van Rijsingen IA et al. Clinical utility gene card for: hypertrophic cardiomyopathy (type 1-14). Eur J Hum Genet 2011;19(8). doi: /ejhg Posafalvi A, Herkert JC, Sinke RJ et al. Clinical utility gene card for: dilated cardiomyopathy (CMD). Eur J Hum Genet 2013;21(10), doi: /ejhg Pugh TJ, Kelly MA, Gowrisankar S et al. The landscape of genetic variation in dilated cardiomyopathy as surveyed by clinical DNA sequencing. Genet Med 2014;16(8):601-8 Quarta G, Muir A, Pantazis A et al. Familial evaluation in arrhythmogenic right ventricular cardiomyopathy: impact of genetics and revised task force criteria. Circulation 2011;123: Querfurth R, Fischer A, Schweiger MR et al. Creation and application of immortalized bait libraries for targeted enrichment and next-generation sequencing. Biotechniques 2012;52(6): Rigato I, Bauce B, Rampazzo A et al. Compound and digenic heterozygosity predicts life-time arrhythmic outcome and sudden cardiac death in desmosomal gene-related arrhythmogenic right ventricular cardiomyopathy. Circ Cardiovasc Genet 2013;6: Roncarati P, Viviani Anselmi C, Krawitz P et al. Doubly heterozygous LMNA and TTN mutations revealed by exome sequencing in a severe form of dilated cardiomyopathy. Eur J Hum Genet 2013;21(10): Shearer AE, Hildebrand MS, Smith RJ. Solutionbased targeted genomic enrichment for precious DNA samples. BMC Biotechnol 2012;4;12:20 Sikkema-Raddatz B, Johansson LF, de Boer EN et al. Targeted next generation sequencing can replace Sanger sequencing in clinical diagnostics. Hum Mut 2013;34(7): Teekakirikul P, Kelly MA, Rehm HL et al. Inherited cardiomyopathies Molecular genetics and clinical genetic testing in the postgenomic era. J Mol Diagn 2013;15: te Rijdt WP, Jongbloed JD, de Boer RA et al. Clinical utility gene card for: arrhythmogenic right ventricular cardiomyopathy (ARVC). Eur J Hum Genet 2014;22(2). doi: / ejhg van Tintelen JP, Pieper PG, van Spaendonck-Zwarts KY. Pregnancy, cardiomyopathies, and genetics. Cardiovasc Res 2014;101(4): doi: /cvr/cvu014 van der Zwaag PA, van Tintelen JP, Gerbens F et al. Haplotype sharing test maps genes for familial cardiomyopathies. Clin Genet 2011;79: van der Zwaag PA, van Rijsningen IAW, Asimaki A et al. Phospholamban R14del mutation in patients diagnosed with dilated cardiomyopathy or arrhthmogenic right ventricular cardiomyopathy: evidence supporting the concept of arrhythmogenic cardiomyopathy. Eur J Heart Fail 2012;14: van der Zwaag PA, van Rijsingen IAW, de Ruiter R et al. Recurrent and founder mutations in the Netherlands Phospolamban p.arg14del mutation causes arrhythmogenic cardiomyopathy. Neth Heart J 2013;21: TARGETED SEQUENCING

194 van Spaendonck-Zwarts KY, Posafalvi A, van den Berg MP et al. Titin gene mutations are common in families with both peripartum cardiomyopathy and dilated cardiomyopathy. European Heart Journal 2014;35(32): van Spaendonck-Zwarts KY, van Tintelen JP, van Veldhuisen DJ et al. Peripartum cardiomyopathy as a part of familial dilated cardiomyopathy. Circulation 2010;121(20): van Spaendonck-Zwarts KY, van Rijsingen IA, van den Berg MP et al. Genetic analysis in 418 index patients with idiopathic dilated cardiomyopathy: overview of 10 years experience. Eur J Heart Fail 2013;15: Voelkerding KV, Dames S, Durtschi JD. Next generation sequencing for clinical diagnostics-principles and application to targeted resequencing for hypertrophic cardiomyopathy: a paper from the 2009 William Beaumont Hospital Symposium on Molecular Pathology. J Mol Diagn 2010;12: Wilde AA & Behr ER. Genetic testing for inherited cardiac disease. Nat Rev Cardiol 2013;10: Xu T, Yang Z, Vatta M et al. Compound and digenic heterozygosity contributes to arrhythmogenic right ventricular cardiomyopathy. J Am Coll Cardiol 2010;55: Zimmerman RS, Cox S, Lakdawala NK et al. A novel custom resequencing array for dilated cardiomyopathy. Genet Med 2010;12: CHAPTER 4.1 DIAGNOSTIC YIELD OF CARDIOMYOPATHIES 193

195 SUPPLEMENTARY MATERIAL Supplementary methods: Variant filtering A classification tree was developed using software from Cartagenia (Leuven, Belgium) in which subsequent filtering steps were used as described below. First, variants were filtered for the regions of interest, which included all exons of the targeted 55 cardiomyopathy genes with their respective +/- 20 bp flanking intronic sequences. Next, quality filtering of the called variants was performed, excluding all those which were identified with a read depth <20x. In the next step, variants were filtered against our in-house list of managed variants, which is regularly updated and contains previously identified and validated variants, including polymorphisms ( benign variants) and sequencing artefacts. This was followed by excluding any variants that were present with an allele frequency 2%, with a minimum of 200 alleles screened in cohorts of ostensibly healthy controls: (1) the Genome of the Netherlands: the 1000 genome database of healthy Dutch individuals ( (2) the 1000 Genomes project ( variants identified in the 2184 genomes from the 1000 Genomes project, and (3) the dbsnp database ( status: validated). In addition, those variants present with an allele frequency 5% (again with a minimum of 200 alleles screened) in the ESP6500 database (NHLBI Exome Sequencing Project (ESP); (variants identified during exome sequencing of 6500 individuals) were excluded. In the latter case, a higher frequency cut-off was selected, as this database contains the exomes of patients with cardiovascular diseases. As the final filter, we used an additional managed variant list containing likely benign variants, which had been previously identified as such in our validation series of targeted enrichment sequencing (Sikkema-Raddatz et al). Or they were frequently seen in our patient samples, but not in more than 20% of those samples, and predicted in silico to be likely benign (variants identified in 20% of patient samples were incorporated into our in-house database of polymorphisms, benign variants, or artefacts, depending on the nature of the variant). 194 TARGETED SEQUENCING

196 Supplementary Table 1: Targeted cardiomyopathy genes Gene Chromosome Basepair position* (start - end) NEXN** LMNA TNNT PSEN ACTN RYR TTN DES CAV TMEM SCN5A MYL TNNC MYOZ SGCD DSP LAMA PLN EYA SOD TBX GATAD PRKAG MYPN MYOZ VCL LDB ANKRD RBM BAG CSRP MYBPC CRYAB ABCC PKP MYL MYH MYH PSEN ACTC TPM CHAPTER 4.1 DIAGNOSTIC YIELD OF CARDIOMYOPATHIES 195

197 TCAP JUP DSC DSG DTNA CALR TNNI JPH TXNRD DMD X GLA X LAMP2 X EMD X TAZ X List of genes included in the targeted Sure Select enrichment kit. *base pair position according to NCBI build 37 (UCSC hg19); **NEXN Nexilin, rat, homolog of, LMNA Lamin A/C, TNNT2 Troponin T2, cardiac, PSEN2 Presenilin 2, ACTN2 Actinin, alpha-2, RYR2 Ryanodine receptor 2, TTN Titin, DES Desmin, CAV3 Caveolin 3, TMEM43 Transmembrane protein 43, SCN5A Sodium channel, voltage gated, type V, alpha subunit, MYL3 Myosin, light chain 3, alkali, ventricular, skeletal, slow, TNNC1 Troponin C, slow, MYOZ2 Myozenin 2, SGCD Sarcoglycan, delta, DSP Desmoplakin, LAMA4 Laminin alpha-4, PLN Phospholamban, EYA4 eye absent 4, SOD2 superoxide dismutase 2, mitochondrial, TBX20 T-box 20, GATAD1 GATA zinc finger domain-containing protein 1, PRKAG2 Protein kinase, AMP-activated, noncatalytic, MYPN gamma-2, Myopalladin, MYOZ1 Myozenin 1, VCL Vinculin, LDB3 LIM domain-binding 3, ANKRD1 Ankyrin repeat domain-containing protein 1, RBM20 RNA-binding protein 20, BAG3 BCL2-associated athanogene 3, CSRP3 Cysteine- and glycine-rich protein 3, MYBPC3 Myosinbinding protein C, cardiac, CRYAB Crystalline, alpha-b, ABCC9 ATP-binding cassette, subfamily C, member 9, PKP2 Plakophilin 2, MYL2 Myosin, light chain 2, regulatory, cardiac, slow, MYH6 Myosin, heavy chain 6, cardiac muscle, alpha, MYH7 Myosin, heavy chain 7, cardiac muscle, beta, PSEN1 Presenilin 1, ACTC1 Actin, alpha, cardiac muscle, TPM1 Tropomyosin 1, TCAP Titin-cap, JUP Junction plakoglobin, DSC2 Desmocollin 2, DSG2 Desmoglein 2, DTNA Dystrobrevin, alpha, CALR3 Calreticulin 3, TNNI3 Troponin I, cardiac, JPH2 Junctophilin 2, TXNRD2 Thioredoxin reductase 2, DMD Dystrophin, GLA Galactosidase, alpha, LAMP2 Lysosome-associated membrane protein 2, EMD Emerin, TAZ Tafazzin 196 TARGETED SEQUENCING

198 Supplementary Table 2: Patients and mutations Summary of clinical data and the mutations identified in 206 patients, including their diagnosis (Dx) at referral, gender, age of onset, family history, diagnosis (Dx) after phenotype evaluation, pathogenic and or likely pathogenic mutations identified and the number of mutations per patient. Abbreviations: ARVC arrhythmogenic right ventricular cardiomyopathy, CM (unspecified) cardiomyopathy, DCM dilated cardiomyopathy, Dx diagnosis, F female, HCM hypertrophic cardiomyopathy, LVNC left ventricular non-compaction, M male, n.a. not available, neg negative, pos positive, RCM restrictive cardiomyopathy, WPW Wolff-Parkinson-White syndrome. Symbols used * stop codon, ^ mutation known in HGMD, mutated nucleotide (also) known in HGMD, but substituted for another nucleotide. The following transcripts were used for the nomenclature of the mutations ABCC9 (NM_ ), ACTN2 (NM_ ), ANKRD1 (NM_ ), CALR3 (NM_ ), CAV3 (NM_ ), CSRP3 (NM_ ), DES (NM_ ), DMD (NM_ ), DSC2 (NM_ ), DSP (NM_ ), DTNA (NM_ ), EMD (NM_ ), GLA (NM_ ), JPH2 (NM_ ), JUP (NM_ ), LAMA4 (NM_ ), LDB3 (NM_ ), LMNA (NM_ ), MYBPC3 (NM_ ), MYH6 (NM_ ), MYH7 (NM_ ), MYL3 (NM_ ), MYPN (NM_ ), NEXN (NM_ ), PKP2 (NM_ ), PLN (NM_ ), RBM20 (NM_ ), RYR2 (NM_ ), SCN5A (NM_ ), TNNC1 (NM_ ), TNNI3 (NM_ ), TNNT2 (NM_ ), TPM1 (NM_ ), TTN (NM_ ) and VCL (NM_ ). Patient Dx at referal Gender Age diagnosis Family History Dx after phenotype evaluation patient categorisation pathogenic mutation(s): gene, cdna; protein likely pathogenic mutation(s): gene, cdna; protein number of mutations 1 DCM M 47 neg DCM fulfilling criteria no no 0 2 DCM M 50 neg DCM fulfilling criteria no MYPN, c.211_213delgaa; p.e71del 1 3 DCM M 48 neg DCM fulfilling criteria no TTN, c.75391delg; p.v25131lfs* DCM F n.a. pos DCM fulfilling criteria no no 0 5 DCM M 58 neg DCMlike suspected no ANKRD1, c.222dupa; p.l75tfs*8 1 6 DCM F 64 neg? DCMlike suspected no no 0 7 DCM F n.a. neg DCMlike suspected no no 0 8 HCM M 61 neg HCM fulfilling criteria no LAMA4, c.4624a>t; p.n1542y 1 9 DCM M 65 neg ARVC fulfilling criteria PKP2, c g>c^ no 1 10 DCM F 49 neg DCM fulfilling criteria no no 0 11 DCM F 45 pos DCM fulfilling criteria no no 0 12 HCM M 65 pos HCM fulfilling criteria no no 0 13 HCM F 56 pos? HCM fulfilling criteria MYBPC3, c.2864_2865delct; p.p955rfs*95^ ACTN2, c.2386c>t; p.r796c & LDB3, c.608c>t; p.s203l 3 CHAPTER 4.1 DIAGNOSTIC YIELD OF CARDIOMYOPATHIES 197

199 Patient Dx at referal Gender Age diagnosis Family History Dx after phenotype evaluation patient categorisation pathogenic mutation(s): gene, cdna; protein likely pathogenic mutation(s): gene, cdna; protein number of mutations 14 HCM M 60 neg? HCM fulfilling criteria no no 0 15 DCM M 44 neg DCM fulfilling criteria no no 0 16 DCM F 42 pos DCM fulfilling criteria no no 0 17 DCM F 47 neg? DCM unconfirmed no no 0 18 DCM M 49 pos DCM fulfilling criteria no no 0 19 HCM M 61 pos HCMlike suspected no no 0 20 DCM F 54 pos DCMlike suspected no no 0 21 CM F 59 n.a. CM suspected no RYR2, c.9454c>t; p.r3152c & LAMA4, c.3335c>a; p.p1112h 22 DCM M 39 pos DCMlike suspected no LDB3, c.608c>t; p.s203l 1 23 HCM F 58 pos? HCMlike suspected no no 0 24 HCM M 23 n.a. HCMlike suspected no TTN, c.89426g>a; p.r29809q (splice) 1 25 HCM F 53 n.a. HCM unconfirmed no ACTN2, c.2035a>c; p.k679q 1 26 HCM F 57 pos HCM fulfilling criteria no no 0 27 DCM F 55 neg DCM fulfilling criteria no NEXN, c.1174c>t; p.r392* & MYH6, c.961g>c; p.v321l 28 RCM M 0 neg RCM fulfilling criteria TNNI3, c.527g>a; p.w191* NEXN, c.1453g>a; p.e485k 2 29 HCM M 13 pos? HCM fulfilling criteria MYBPC3, c.2373dupg; p.w792vfs*41^ no 1 30 DCM F 61 neg DCM fulfilling criteria no MYH6, c.3010g>t; p.a1004s^ 1 31 DCM F 58 neg DCM fulfilling criteria no no 0 32 DCM F 42 pos DCM fulfilling criteria no DSP, c.4679a>g; p.q1560r & MYH7, c.4084t>c; p.s1362p 33 HCM M 73 n.a. HCM fulfilling criteria no MYH6, c.4328c>a; p.a1443d^ 1 34 HCM M 27 neg HCMlike suspected no no 0 35 DCM F 35 n.a. DCMlike suspected no TTN, c.91152t>a; p.y30384* 1 36 HCM M 45 neg HCM unconfirmed no TTN, c t>c 1 37 ARVC M n.a. n.a. ARVC unconfirmed no no 0 38 DCM F 55 pos DCM fulfilling criteria no no TARGETED SEQUENCING

200 39 HCM M 62 neg HCM fulfilling criteria no no 0 40 HCM F 46 pos HCM fulfilling criteria no MYBPC3, c.3065g>c; p.r1022p^ & EMD, c.149c>a; p.p50h 41 HCM M n.a. n.a. HCM fulfilling criteria no no 0 42 DCM M 60 neg DCM fulfilling criteria no LMNA, c.992g>a; p.r331q^ 1 43 HCM M 44 pos HCM fulfilling criteria no no 0 44 HCM M 65 neg HCM fulfilling criteria MYBPC3, c.2373dupg; p.w792vfs*41^ no 1 45 HCM F 64 pos HCM fulfilling criteria no no 0 46 HCM M 65 neg DCM fulfilling criteria MYBPC3, c.3776dela; p.q1259rfs*72^ no DCM M 59 neg DCMlike suspected MYBPC3, c.3776dela; p.q1259rfs*72^ no 1 48 HCM M 49 neg HCM fulfilling criteria no ABCC9, c.4516c>t; p.r1506c 1 49 DCM M 50 pos DCM fulfilling criteria PLN, c.40_42delaga; p.r14del^ DSP, c.1778a>g; p.n593s 2 50 LVNC M n.a. neg LVNClike suspected no JUP, c.849g>t; p.k283n 1 51 HCM M 50 neg HCMlike suspected no no 0 52 DCM F 38 Pos DCM unconfirmed no MYBPC3, c.841c>a; p.= (splice) 1 53 HCM F n.a. n.a. HCM unconfirmed no TNNI3 c.626a>c; p.glu209ala^ 1 54 CM F n.a. n.a. CM unconfirmed no MYPN, c.2242c>t; p.r748c & DMD, c.2827c>t; p.r943c 55 DCM F 73 pos? DCM fulfilling criteria no SCN5A, c.659c>t; p.t220i^ 1 56 DCM F 67 pos DCM fulfilling criteria no no 0 57 HCM M 51 pos? HCM fulfilling criteria no no 0 58 HCM M 27 pos HCM fulfilling criteria MYBPC3, c.2373dupg; p.w792vfs*41^ no 1 59 HCM F n.a. n.a. HCM fulfilling criteria no GLA, c.1153a>g; p.t385a^ 1 60 DCM M 54 n.a. DCM fulfilling criteria no no 0 61 DCM/ARVC M 47 pos DCMlike suspected no NEXN, c.995a>c; p.e332a & DES, c.1193t>c; p.l398p 62 DCM/ARVC F 55 pos ARVClike suspected DSP, c t>a no CHAPTER 4.1 DIAGNOSTIC YIELD OF CARDIOMYOPATHIES 199

201 Patient Dx at referal Gender Age diagnosis Family History Dx after phenotype evaluation patient categorisation pathogenic mutation(s): gene, cdna; protein likely pathogenic mutation(s): gene, cdna; protein number of mutations 63 HCM F 45 neg HCMlike suspected no no 0 64 DCM M n.a. n.a. DCM unconfirmed no no 0 65 DCM M 56 pos DCM fulfilling criteria no MYL3, c.517a>g; p.m173v^ 1 66 DCM M 41 neg DCM fulfilling criteria no LAMA4, c.4624a>t; p.n1542y 1 67 DCM F 66 neg DCM fulfilling criteria no DSP, c.273+5g>a^ 1 68 HCM F 65 n.a. HCM fulfilling criteria MYBPC3, c.2864_2865delct; p.p955rfs*95^ 69 CM F 34 neg DCMlike suspected no no 0 70 DCM M 50 neg DCMlike suspected no DSP, c.1778a>g; p.n593s 1 71 DCM M 55 n.a. DCMlike suspected no TTN, c.75332_75335duptaag; p.m25113kfs*16 72 HCM F 70 n.a. HCMlike suspected no CSRP3, c.131t>c; p.l44p^ 1 73 DCM M 42 pos DCM unconfirmed no no 0 74 DCM M 43 pos DCM fulfilling criteria no no 0 75 HCM F 4 pos HCM fulfilling criteria no MYH7, c.2890g>c; p.v964l^ 1 76 DCM M 44 pos DCM fulfilling criteria no VCL, c.2467c>t; p.r823w 1 77 HCM F 54 n.a. HCM fulfilling criteria no no 0 78 DCM M n.a. n.a. DCM fulfilling criteria no no 0 79 DCM F 50 n.a. DCM fulfilling criteria no no 0 80 DCM/HCM M n.a. n.a. DCMlike suspected no no 0 81 HCM F 59 Neg HCM unconfirmed no no 0 82 DCM M n.a. n.a. DCM unconfirmed TNNI3 c.292c>t; p.r98*^ no 1 83 DCM M n.a. n.a. DCM unconfirmed no no 0 84 HCM M n.a. n.a. HCM unconfirmed no no 0 85 DCM M 66 neg DCM fulfilling criteria no no 0 86 HCM M 56 pos HCM fulfilling criteria no no 0 87 DCM M 41 pos? DCM fulfilling criteria no no 0 88 DCM M 54 neg DCM fulfilling criteria no no 0 89 HCM F 69 neg HCM fulfilling criteria no no TARGETED SEQUENCING

202 90 HCM M n.a. neg HCM fulfilling criteria no ANKRD1, c.368c>t; p.t123m^ & TTN, c t>g 91 LVNC M 61 pos LVNC fulfilling criteria no no 0 92 ARVC F n.a. n.a. ARVClike suspected no no 0 93 DCM F 38 pos DCMlike suspected no PKP2, c.1288a>g; p.k430e & DSP, c.939+1g>a^ 94 HCM M n.a. n.a. HCM unconfirmed no no 0 95 (D)CM F 28 pos DCM fulfilling criteria no TTN, c.86872dupa; p.s28958kfs* DCM F 36 pos DCM fulfilling criteria no TTN, c.45616g>t; p.e15206* 1 97 HCM F 57 neg? HCM fulfilling criteria no no 0 98 DCM M 42 n.a. DCM fulfilling criteria no TTN, c g>c 1 99 ARVC/CM M n.a. pos ARVClike suspected PLN, c.40_42delaga; p.r14del^ no ARVC F 63 pos ARVClike suspected no no HCM M n.a. n.a. HCM unconfirmed MYBPC3, c.654+1g>a SCN5A, c.2423g>a; p.r808h DCM F 68 neg DCM fulfilling criteria SCN5A, c.2582_2583deltt; p.f861wfs*90^ no HCM M 55 pos HCM fulfilling criteria no ACTN2, c.690t>a; p.d230e DCM M 40 pos DCM fulfilling criteria no no DCM F 46 n.a. DCM fulfilling criteria no LMNA, c.437c>a; p.a146d & DMD, c.2827c>t; p.r943c & MYPN, c.59a>g; p.y20c^ 106 DCM F 35 pos DCM fulfilling criteria no no HCM M 59 pos HCM fulfilling criteria no no DCM M 35 neg DCM fulfilling criteria no MYH7, c.4377g>t; p.k1459n^ DCM M 23 neg DCMlike suspected no no CM F 60 pos DCMlike suspected no ABCC9, c.2324c>a; p.p775h ARVC/DCM M 22 pos CM suspected no DSP, c.4915g>a; p.v1639m & ANKRD1, c.599_600delat; p.d200gfs*8 112 DCM M 72 pos DCM fulfilling criteria no TTN, c.12897dupa p.g4300rfs* HCM M 8 pos HCM fulfilling criteria no no DCM F 49 pos? DCM fulfilling criteria no no CHAPTER 4.1 DIAGNOSTIC YIELD OF CARDIOMYOPATHIES 201

203 Patient Dx at referal Gender Age diagnosis Family History Dx after phenotype evaluation patient categorisation pathogenic mutation(s): gene, cdna; protein likely pathogenic mutation(s): gene, cdna; protein number of mutations 115 HCM F n.a. n.a. HCM fulfilling criteria TNNT2, c.814c>t; p.q272* no DCM M n.a. n.a. DCM unconfirmed no no DCM F n.a. n.a. DCM unconfirmed PLN, c.40_42delaga; p.r14del^ no HCM M n.a. n.a. HCM unconfirmed no no HCM F 60 pos HCM fulfilling criteria no no DCM M 50 n.a. DCM fulfilling criteria no no DCM M 17 neg DCM fulfilling criteria no RYR2, c.3152g>a; p.r1051h HCM M 56 n.a. HCM fulfilling criteria no no HCM M n.a. n.a. HCM unconfirmed no no DCM M n.a. n.a. DCM unconfirmed no TTN, c g>t DCM M n.a. n.a. DCM fulfilling criteria no MYH7, c.5773c>g: p.r1925g^ & MYBPC3, c.3392t>c; p.i1131t^ 126 HCM F n.a. n.a. HCM fulfilling criteria no JPH2, c.723c>g; p.s241r HCM M n.a. n.a. HCM unconfirmed no MYH6, c.2354g>a; p.r785h DCM F n.a. n.a. DCM unconfirmed no TTN, c g>t HCM M n.a. n.a. HCM unconfirmed no RYR2, c.9418t>g; p.l3140v HCM F n.a. n.a. HCM unconfirmed no TNNT2, c.821+2dupt & GLA, c.427g>a; p.a143t^ & RYR2, c.8162t>c; p.i2721t 131 HCM M n.a. n.a. HCM unconfirmed no ANKRD1, c.222dupa; p.l75tfs* DCM M n.a. n.a. DCM unconfirmed no no ARVC F 57 n.a. ARVC fulfilling criteria no no DCM M 42 pos DCM fulfilling criteria no TTN, c.54339dela; p.e18113dfs* HCM M 56 neg HCM fulfilling criteria no no HCM M n.a. n.a. HCM fulfilling criteria no DSP, c.944g>a; p.r315h HCM F 62 pos HCMlike suspected no no HCM M n.a. n.a. HCM unconfirmed no no HCM F n.a. n.a. HCM unconfirmed no no DCM M n.a. n.a. DCM unconfirmed no no TARGETED SEQUENCING

204 141 DCM M n.a. n.a. DCM unconfirmed MYBPC3, c.2827c>t; p.r943*^ankrd1, c.222dupa; p.l75tfs* DCM M 57 neg DCM fulfilling criteria no DSP, c.1778a>g p.n593s DCM M 56 pos DCM fulfilling criteria no TTN, c g>a HCM M 7 neg HCM fulfilling criteria no MYH7, c.2156g>a; p.r719q^ ARVC F 38 n.a. ARVC fulfilling criteria no no HCM M 1 pos HCM fulfilling criteria no MYH7, c.1816g>a; p.v606m^ DCM F 61 pos DCM fulfilling criteria no TTN, c.80314_80315del; p.v26772* DCM F 50 pos DCM fulfilling criteria no no DCM F 49 pos DCM fulfilling criteria no no HCM M 63 neg HCM fulfilling criteria no no DCM M 46 pos DCM unconfirmed no TNNC1, c.439c>t; p.r147c^ DCM F 37 n.a. DCM fulfilling criteria no DSP, c.8500c>t; p.r2834c DCM F n.a. pos DCM fulfilling criteria no ACTN2, c.1426g>t; p.a476s & DMD, c.343a>g; p.i115v 154 HCM M 56 pos HCM fulfilling criteria no no DCM M 20 neg DCM fulfilling criteria no no DCM F 43 neg DCM fulfilling criteria no no HCM M 59 pos HCM fulfilling criteria no no HCM F 66 n.a. HCM fulfilling criteria no no DCM M 57 pos DCMlike suspected no TTN, c a>g & DSC2, c.943-1g>a^ 160 ARVC M n.a. n.a. ARVClike suspected no DMD, c.7988c>g; p.t2663r DCM M n.a. n.a. DCM unconfirmed no TTN, c.85809dela; p.k28603nfs* HCM F n.a. n.a. HCM unconfirmed no DSP, c.1714c>t; p.r572w & RYR2, c.8162t>c; p.i2721t 163 DCM F 34 pos DCM fulfilling criteria no RBM20, c.1910g>a; p.s637n LVNC M n.a. n.a. LVNC fulfilling criteria no no DCM F 25 pos DCM fulfilling criteria no DMD, c.8255a>g; p.y2752c LVNC F 76 pos LVNC fulfilling criteria MYBPC3, c.2373dupg; p.w792vfs*41^ no HCM M 48 neg HCM fulfilling criteria no no CHAPTER 4.1 DIAGNOSTIC YIELD OF CARDIOMYOPATHIES 203

205 Patient Dx at referal Gender Age diagnosis Family History Dx after phenotype evaluation patient categorisation pathogenic mutation(s): gene, cdna; protein likely pathogenic mutation(s): gene, cdna; protein number of mutations 168 HCM F 50 n.a. HCM fulfilling criteria no CSRP3, c.208g>t; p.g70w HCM M 61 pos HCM fulfilling criteria no no DCM F 51 pos DCM fulfilling criteria no DMD, c.2827c>t; p.r943c DCM M 14 pos DCM fulfilling criteria no no DCM F 70 neg DCMlike suspected no no CM F 49 pos? DCMlike suspected no no LVNC M n.a. neg LVNC unconfirmed no no DCM F 49 pos DCM fulfilling criteria no no DCM M 68 pos DCM fulfilling criteria no no DCM M 58 pos DCM fulfilling criteria no NEXN c.1453g>a; p.e485k & MYBPC3 c.649a>g; p.s217g^ 178 HCM F 64 n.a. HCM fulfilling criteria no no DCM M 36 pos DCM fulfilling criteria no DSP, c.4608_4612delacgcc; p.r1537efs*5 & MYPN, c.59a>g; p.y20c^ & MYH7, c.5773c>g; p.r1925g^ 180 HCM M 77 pos? HCM fulfilling criteria no MYH6, c.4264c>t; p.r1422w DCM M 69 pos? DCMlike suspected no no ARVC M n.a. n.a. ARVClike suspected no no DCM F 31 pos DCMlike suspected no RYR2, c.1939c>t; p.r647c DCM F 48 pos DCMlike suspected no no HCM M 68 pos HCMlike suspected no 186 DCM M 53 n.a. DCM unconfirmed no 187 HCM F 70 pos? HCM fulfilling criteria no DES, c.170c>t; p.s57l & DTNA, c.1778c>a; p.s593y ACTN2, c.690t>a; p.d230e & TTN, c g>a & MYH6, c.3010g>t; p.a1004s^ LAMA4, c.3286c>t; p.r1096c & MYPN, c.2951g>a; p.r984q DCM M 54 neg? DCM fulfilling criteria no LMNA, c.992g>a; p.r331q^ & MYH6, c.3809g>a; p.r1270h TARGETED SEQUENCING

206 189 HCM + WPW M 69 neg? HCM fulfilling criteria no no DCM M 22 n.a. DCM fulfilling criteria no TPM1, c.853t>c; p.*285glnext*20 & JPH2, c.8g>a; p.g3e 191 HCM M 49 neg HCMlike suspected no no HCM M 49 neg? HCMlike suspected no ANKRD1, c.222dupa; p.l75tfs*8 & LAMA4, c.3335c>a; p.p1112h DCM F n.a. n.a. DCM unconfirmed no 194 HCM M n.a. n.a. HCM unconfirmed MYBPC3, c.2373dupg; p.w792vfs*41^ TTN, c.17823dela; p.i5941mfs*8 & DSP, c.3294c>g p.d1098e RYR2 c.8162t>c; p.i2721t DCM M 51 n.a. DCM fulfilling criteria no TTN, c.3100g>a; p.v1034m (splice)^ DCM M 23 n.a. DCM fulfilling criteria no DES, c.1193t>c; p.l398p DCM M 45 n.a. DCM fulfilling criteria no no DCM M 61 n.a. DCM fulfilling criteria no no DCM M 42 n.a. DCM fulfilling criteria no no DCM M 59 n.a. DCM fulfilling criteria no TTN, c.54406_54409delcagt; p.q18136mfs*8 201 DCM M 55 n.a. DCM fulfilling criteria no CALR3, c.147dupt; p.r50* DCM M 57 n.a. DCM fulfilling criteria MYH6, c.3607dupg; p.a1203gfs*30 no DCM M 53 n.a. DCM fulfilling criteria no no HCM/DCM M 40 n.a. CM suspected no no DCM M 56 n.a. DCM like suspected PLN, c.40_42delaga; p.r14del^ no DCM M 77 n.a. DCM like suspected no no CHAPTER 4.1 DIAGNOSTIC YIELD OF CARDIOMYOPATHIES 205

207 Supplementary Table 3: Total diagnostic yield CM subtype neg P LP pos (P + LP) total ARVC 6 (60%) 3 (30%) 1 (10%) 4 (40%) 10 DCM 50 (45%) 9 (8%) 53 (47%) 62 (55%) 112 HCM 39 (52%) 8 (11%) 27 (36%) 35 (47%) 74 LVNC 3 (60%) 1 (20%) 1 (20%) 2 (40%) 5 RCM 0 1 (100%) 0 1 (100%) 1 CM 1 (25%) 0 3 (75%) 3 (75%) 4 total 99 (48%) 22 (11%) 85 (41%) 107 (52%) 206 *Abbreviations: ARVC: arrhythmogenic right ventricular cardiomyopathy, CM: cardiomyopathy, DCM: dilated cardiomyopathy, HCM: hypertrophic cardiomyopathy, LVNC: left ventricular non-compaction, LP: likely pathogenic (sometimes together with one or more LP s), neg: negative, P: pathogenic (sometimes together with one or more LPs), pos: positive, RCM: restrictive cardiomyopathy. Supplementary Table 4: Gender and genetic diagnosis. Fulfilling Suspected Unconfirmed Total M* F p-value M F p-value M F p-value M F p-value Total mut 39 (51%) 25 (52%) 1 15 (60%) 8 (41%) (48%) 9 (69%) (52%) no mut (53%) Comparison of the sex distribution in the three patient categories related to their mutation carrier status (no mut = no mutation identified; 1 mut = one or multiple mutations identified. *M = male; F = female). Supplementary Table 5: LVEF in DCM patients fulfilling criteria related to presence or absence of single and/or multiple mutations. Mutations P-value 0 1 >1 0vs 1 0vs1 0vs>1 1vs>1 Data available 32/32 27/30 9/10 LVEF[%]±SD 29.3± ± ± * Groups were compared using one-way ANOVA Supplementary Table 6: IVS in HCM patients fulfilling criteria related to presence or absence of single and/or multiple mutations. Mutations P-value 0 1 >1 0vs 1 0vs1 0vs>1 1vs>1 Data available 23/25 16/17 2/4 IVS[mm]±SD 19.5± ± ± * Groups were compared using one-way ANOVA TARGETED SEQUENCING

208

209

210 Chapter 4.2 Titin gene mutations are common in families with both peripartum cardiomyopathy and dilated cardiomyopathy Karin Y van Spaendonck-Zwarts, Anna Posafalvi, Maarten P van den Berg, Denise Hilfiker-Kleiner, Ilse AE Bollen, Karen Sliwa, Mariëlle Alders, Rowida Almomani, Irene M van Langen, Peter van der Meer, Richard J Sinke, Jolanda van der Velden, Dirk J Van Veldhuisen, J Peter van Tintelen *, Jan DH Jongbloed * * The last two authors contributed equally Published in European Heart Journal, 2014

211 ABSTRACT Aims: Peripartum cardiomyopathy (PPCM) can be an initial manifestation of familial dilated cardiomyopathy (DCM). We aimed to identify mutations in families that could underlie their PPCM and DCM. Methods and Results: We collected 18 families with PPCM and DCM cases from various countries. We studied the clinical characteristics of the PPCM patients and affected relatives, and applied a targeted next-generation sequencing (NGS) approach to detect mutations in 48 genes known to be involved in inherited cardiomyopathies. We identified 4 pathogenic mutations in 4/18 families (22%): 3 in TTN and 1 in BAG3. In addition, we identified 6 variants of unknown clinical significance that are likely to be pathogenic in 6 other families (33%): 4 in TTN, 1 in TNNC1, and 1 in MYH7. Measurements of passive force in single cardiomyocytes and titin isoform composition potentially support an upgrade of one of the variants of unknown clinical significance in TTN to a pathogenic mutation. Only 2/20 PPCM cases in these families showed recovery of left ventricular function. Conclusion: Targeted NGS shows that potentially causal mutations in cardiomyopathy-related genes are common in families with both PPCM and DCM. This supports the earlier finding that PPCM can be part of familial DCM. Our cohort is particularly characterised by a high proportion of TTN mutations and a low recovery rate in PPCM cases. Keywords: cardiomyopathy, peripartum cardiomyopathy, genetics, pregnancy, titin

212 INTRODUCTION Peripartum cardiomyopathy (PPCM) is an idiopathic cardiomyopathy presenting with heart failure secondary to left ventricular systolic dysfunction towards the end of pregnancy or in the first months following delivery, where no other cause of heart failure is found. The left ventricle may not be dilated but the ejection fraction is nearly always reduced below 45%. 1 According to this recent definition, the time frame is not strictly defined, in contrast to previous definitions. 2-4 The severity of PPCM is highly variable, ranging from complete recovery to rapid progression to end-stage heart failure. PPCM affects 1:300 to approximately 1:3000 pregnancies, with geographic hot spots of high incidence such as in Haiti and Nigeria. 4,5 The precise mechanisms that lead to PPCM are not fully known. Several risk factors and possible underlying pathological processes have received attention, such as abnormal autoimmune responses, apoptosis, and impaired cardiovascular microvasculature. 5,6 Recent work into the pathogenesis of PPCM has shown involvement of a cascade with oxidative stress, the prolactin-cleaving protease cathepsin D, and the nursing hormone prolactin, which may lead to a target for a disease-specific therapy, namely pharmacological blockade of prolactin by bromocriptine. 7-9 In addition, involvement of cardiac angiogenic imbalance may explain why PPCM is a disease seen in late pregnancy and why pre-eclampsia and multiple gestation are important risk factors. 10 PPCM is probably caused by a complex interaction of more than one pathogenic mechanism. The large variation in incidence and clinical characteristics may reflect the involvement of specific mechanisms, or combinations thereof, in certain subgroups of PPCM. We and others recently reported that PPCM can be an initial manifestation of familial dilated cardiomyopathy (DCM), 11,12 indicating that, at least in a subset of cases, genetic predisposition plays a role in the pathophysiology of pregnancy-associated heart failure. Accordingly, Haghikia et al. reported a positive family history for cardiomyopathy in 16.5% (19/115) of PPCM cases from a German PPCM cohort. 13 So far, eight cases with underlying mutations in DCM-related genes have been published 11,12,14,15 and several other cases with familial occurrences of PPCM and DCM, as well as familial clustering of PPCM, have been reported Here, we describe our extensive genetic analysis using next-generation sequencing (NGS) technology to identify potentially causal mutations in families with both PPCM and DCM from various parts of the world. CHAPTER 4.2 TTN IN PERIPARTUM CARDIOMYOPATHY 211

213 METHODS Subjects and Clinical Evaluation We collected a cohort of families with cases of both PPCM and DCM from various parts of the world (the Netherlands, Germany, and South Africa) and studied their clinical characteristics by reviewing medical reports. The local institutional review committees approved the study, and all participants gave their informed consent. PPCM was diagnosed when a patient had an idiopathic cardiomyopathy presenting with heart failure secondary to left ventricular systolic dysfunction towards the end of pregnancy or in the first months following delivery, where no other cause of heart failure was found. 1 DCM was diagnosed when a patient had both a reduced systolic function of the left ventricle (left ventricular systolic ejection fraction <0.45) and dilation of the left ventricle (left ventricular enddiastolic dimension >117% of the predicted value corrected for body surface area and age) and only after other identifiable causes like severe hypertension, coronary artery disease, and systemic disease had been excluded. 25 If only one of the two criteria was fulfilled, the patient was labeled with mild DCM. If the family history suggested DCM in a relative but there were no medical reports to confirm this, the relative was labeled as having possible DCM. Familial PPCM/ DCM was diagnosed when there were 2 affected family members, at least one with PPCM and one with DCM or sudden cardiac death (SCD) 35 years. Targeted Next-Generation Sequencing of 48 Cardiomyopathy- Related Genes Genomic deoxyribonucleic acid (DNA) was extracted from blood samples obtained from all the available PPCM patients and their affected relatives. Targeted NGS was performed in one or two affected relatives in the selected families (these individuals are marked with an arrow in Figures 1 and 2). We developed a kit based on Agilent Sure Select Target Enrichment for mutation detection in 48 genes (all exonic and ± 20 bp of exon-flanking intronic sequences) known to be involved in inherited cardiomyopathies (ABCC9, ACTC1, ACTN2, ANKRD1, BAG3, CALR3, CRYAB, CSRP3/MLP, DES, DMD, DSC2, DSG2, DSP, EMD, GLA, JPH2, JUP, LAMA4, LAMP2, LMNA, MYBPC3, MYH6, MYH7, MYL2, MYL3, MYPN, MYOZ1, MYOZ2, PKP2, PLN, PRKAG2, PSEN1, PSEN2, RBM20, RYR2, SCN5A, SGCD, TAZ, TBX20, TCAP, TMEM43, TNNC1, TNNI3, TNNT2, TPM1, TTN, VCL, ZASP/LDB3). 26 Samples 212 TARGETED SEQUENCING

214 were prepared according to the manufacturer s protocols and multiplexed to an amount still permitting a theoretical coverage of 100 reads per targeted sequence/per patient. All samples were sequenced using 151 bp paired-end reads on an Illumina MiSeq sequencer and analyzed using the MiSeq Reporter pipeline and Nextgene software. 27 Eleven amplicons with low coverage were also analyzed by Sanger sequencing. Identified mutations were confirmed by Sanger sequencing. To study co-segregation, affected relatives were screened for carriership of the identified mutations by Sanger sequencing. Sanger Sequencing STAT3 Gene The STAT3 gene (all coding exons and flanking intronic sequences) was analysed by Sanger sequencing in PPCM patients of the collected families. Classification of Identified Mutations The criteria used to classify mutations were published recently. 28 Briefly, we used a list of mutation-specific features based on in silico analysis using the mutation interpretation software Alamut (version 2.2.1). A score was given depending on the outcome of a prediction test for each feature (i.e. the PolyPhen-2 prediction tool). Then, depending on the total score and the presence/absence of the mutation in at least 300 ethnically matched control alleles (data obtained from the literature and/or available databases, e.g. and or from our own control alleles), we classified mutations as: pathogenic, not pathogenic, or as a variant of unknown clinical significance (VUS; VUS1, unlikely to be pathogenic; VUS2, uncertain; VUS3, likely to be pathogenic). Cosegregation data and/or functional analysis were needed to classify a mutation as pathogenic. CHAPTER 4.2 Functional Analysis of TTN mutation Passive force was measured in single membrane-permeabilized cardiomyocytes mechanically isolated from the heart tissue. 29,30 Titin isoform composition was analysed as described previously. 30 TTN IN PERIPARTUM CARDIOMYOPATHY 213

215 Figure 1. Pedigrees of the Dutch families (NL1-11). Square symbols indicate men; circles, women; diamonds, unknown sex; and triangles, miscarriage. Blue symbols indicate a clinical diagnosis of PPCM; black symbols, (mild) DCM; grey symbols, possible DCM; orange symbols, sudden cardiac death (SCD). Diagonal lines through symbols indicate deceased; arrows indicate patients selected for targeted next-generation sequencing; and the number in a symbol indicates the number of individuals with this symbol (question mark if unknown). 214 TARGETED SEQUENCING

216 Figure 2. Pedigrees of the South African (SA1) and German families (GER1-6). Square symbols indicate men; circles, women; diamonds, unknown sex. Blue symbols indicate clinical diagnosis of PPCM; black symbols, (mild) DCM; orange symbol, sudden cardiac death (SCD). Diagonal lines through symbols indicate deceased; arrows indicate patients selected for targeted next-generation sequencing; the number in a symbol indicates the number of individuals with this symbol (question mark if unknown); and SB indicates still birth. RESULTS Clinical Characteristics: Low Rate of Full Recovery in PPCM Cases of Familial PPCM/DCM We collected 18 families with familial PPCM/DCM. These families originated from the Netherlands (n=11), Germany (n=6), and South Africa (n=1; black). Clinical data of the PPCM cases in these families are summarised in Table 1 and of all (likely) affected relatives in Supplemental Table S1. The pedigrees of all the families are shown in Figures 1 (NL1-11) and 2 (SA1 and GER1-6). In two families there were two cases of PPCM (NL1 and SA1). Eight families (NL1-7 and SA1) have been described previously. 11,31 The median age at diagnosis in PPCM patients was 29 years (n=15; range years), with mean parity 2 (n=13; range 1-4). PPCM diagnosis was postpartum in 12/14 patients. Only 2/20 PPCM patients showed a full recovery of left ventricular function, one of them even had an uneventful next CHAPTER 4.2 TTN IN PERIPARTUM CARDIOMYOPATHY 215

217 Table 1. Clinical characteristics of confirmed PPCM cases Family Patient Referred for Diagnosis (age in yrs) Timing at Diagnosis Pregnancy LVEF at Diagnosis at LVEF Follow-Up NL1 II:6 HF PPCM (29) Just after delivery P4 D (31) NL1 III:4 Cardiogenic shock Cardiological remarks and outcome (age in yrs) Pathology and other remarks PPCM (27) 3 days after delivery P1 20% D MOF (27) Myocyte hypertrophy NL2 III:3 HF PPCM (26) Few days after delivery P4 NL3 III:1 HF PPCM (33) 37th week of pregnancy P2 CS 25% 3 months 33% NL4 III:2 HF PPCM (30) 3 months after delivery 21% NL5 III:1 HF PPCM (33) 35th week of pregnancy P1 AI CS 23% 9 months no recovery 6 months 44%, 7 years 42% NL6 III:2 HF PPCM (29) 2 months after delivery P3 23% 6 months 23% NL7 III:1 HF PPCM (23) Just after delivery P1 CS 29th week, eclampsia NL8 III:3 HF PPCM (35) 2 weeks after delivery P2 18% NL9 III:1 HF, respiratory insufficiency PPCM (30) Just after delivery P1 CS, twin pregnancy 25% 8 years 10% 30% 6 months 55%, 3 years normal LBBB, D asthma cardiale (26) Thrombus LV apex, tachycardia AF (30), PVCs, VTs (46), D HF (51) ICD/CRT (31), LVAD, VF, D cardiogenic shock (34) Thrombus LV apex, TIA, VT (35), PM, HTX (37), normal LVEF (51) Dilated heart, myocyte hypertrophy, fibrosis Signs of acute myocarditis (EMB), suspicion of vasculitis New pregnancy, terminated (35) NL10 III:6 NL11 III:1 Chest pain, coughing Dyspnea, tachycardia PPCM (36) 3 weeks after delivery P2 CS 29th week, HELLP PPCM (20) Just after delivery Poor 6 months 55%, 20-30% 2 years 45%, 3 years 50-55% 4 months 30-35% Tachycardia SA1 II:5 PPCM (23) 1 month after delivery P2 22% No recovery SA1 II:6 Screening, asymptomatic PPCM (22) P1 43% 24% No recovery GER1 II:1 PPCM SB 27 weeks 20% 6 months 37%, 2 years normal Full recovery with uneventful 2nd pregnancy 2 years later GER2 II:1 PPCM 25% ICD, HTX Suspicion of neurodermitis 216 TARGETED SEQUENCING

218 GER3 II:1 PPCM P1 25% 6 months 30% Subsequent pregnancy entered with 30% LVEF, VAD after 2nd pregnancy, no recovery GER4 II:1 PPCM 25% 6 months 25% GER5 II:1 PPCM 25% GER6 II:1 PPCM (33) 3 months after delivery <30% 6 months 36%, >1 year 47% 6 months no recovery BiVAD, no recovery, D after 2 years Graves disease, nicotin and drug abuse AF indicates atrial fibrillation; AI, artificial insemination; AT, atrial tachycardia; (Bi)(L)VAD, (bi)(left) ventricular assist device; CRT, cardiac resynchronization therapy; CS, caesarean section; D, death; EMB, endomyocardial biopsy; HELPP, hemolysis, elevated liver enzymes, low platelet count; HF, heart failure; HTX, heart transplantation; ICD, implantable cardiac defibrillator; LBBB, left bundle branch block; LV, left ventricle; LVEF, left ventricular ejection fraction; MOF, multiple organ failure; P, pregnancy; PM, pacemaker; PPCM, peripartum cardiomyopathy; PVC, premature ventricular contraction; RV, right ventricle; SB, still birth; TIA, transient ischemic attack; VF, ventricular fibrillation; VT, ventricular tachycardia. Table 2. Potentially causal mutations identified in 10/18 families Family Tested patient Gene Amino acid change Nucleotide change Classification Co-segregation Yes/Unknown Affected relatives carrier NL1 II:3 TTN p.arg27373* c.82117c>t Pathogenic Yes II:1, II:3, II:4, III:4, III:5, III:6 NL3 II:2 BAG3 p.gln340* c.1018c>t Pathogenic Yes II:2, III:1 NL4 III:2 TNNC1 p.gln50arg c.149t>c VUS3 Yes II:5, III:2, III:5, IV:1 NL6 II:3 TTN p.asn28726lysfs*3 c.86171_86174dupaaag VUS3 Yes II:1, II:3 NL9 III:1 TTN p.arg17599* c.52795c>t Pathogenic Yes III:1, III:5 NL10 II:6 TTN p.arg23956thrfs*9 c.71867_71876delgagttctgga Pathogenic Yes II:1, II:2, II:6, III:2, III:5, III:6 NL11 III:1 TTN p.ser27317lysfs*10 c.81949dupa VUS3 Yes II:1, III:1 GER1 II:1 TTN p.trp18357* c.55070g>a VUS3 Unknown No samples available GER4 II:1 TTN p.lys15664valfs*13 c.46990_46993delaagg VUS3 Unknown No samples available GER5 II:1 MYH7 p.arg1303gly c.3907c>g VUS3 Yes I:1, II:1 Nomenclature according to HGVS (Human Genome Variation Society) using the reference sequences: TTN (NM_ ; Q8WZ42-1), BAG3 (NM_ ), TNNC1 (NM_ ), MYH7 (NM_ ). VUS indicates variant of unknown clinical significance (VUS3, likely to be pathogenic, VUS2, uncertain). VUS2 p.arg279trp (c.835c>t) on same allele CHAPTER 4.2 TTN IN PERIPARTUM CARDIOMYOPATHY 217

219 pregnancy (NL9 III:1, LVEF still normal 3 years after diagnosis; and GER1 II:1, full recovery with uneventful second pregnancy two years later). Another PPCM patient showed recovery of left ventricular function, but only under treatment with a beta-blocker and ACE inhibitor (NL10 III:6). In addition to 20 confirmed PPCM patients in these families, five relatives show clinical characteristics suggestive for PPCM (NL4 II:2, GER1 I:1, GER3 I:1, GER4 I:1, GER5 I:1; Table S1). PPCM could not be confirmed because clinical data of these relatives was lacking. In addition, two relatives with DCM showed a decline of left ventricular function after delivery (NL2 IV:8 and SA1 II:3; Table S1). Targeted Next-Generation Sequencing: Potential Causal Mutations in Cardiomyopathy-Related Genes, in particular TTN, are Common in Familial PPCM/DCM Using our validated NGS approach, 27 a mean coverage of 220x per individual patient was reached and, on average, 98.5% of all targeted nucleotides were covered at least 20x. In 4/18 families (22%) pathogenic mutations in cardiomyopathy-related genes were identified (3 in TTN and 1 in BAG3). In addition, in 6 other families (33%) VUS3s were identified (4 in TTN, 1 in TNNC1, and 1 in MYH7). An overview of these mutations and VUS3s and the respective co-segregation analyses are shown in Table 2. All 7 TTN mutations/vus3s were located in the titin A-band, for which over-representation of mutations in DCM patients was reported previously. 32 No potential mutations were identified in 8 families (NL2, NL5, NL7, NL8, SA1, GER2, GER3, and GER6). An overview of the 26 mutations that were not classified as potentially disease-causing (VUS1s and VUS2s) identified in the 18 families is shown in Supplemental Table S2. No STAT3 Mutations in PPCM Cases No STAT3 mutations were identified in 15 PPCM cases (DNA was available from 15/20 cases). Functional and Protein Analyses Support the Pathogenicity of a Likely Pathogenic TTN Mutation Heart tissue from PPCM patient GER4 II:1 with a VUS3 in TTN was available for functional and protein analyses. Passive force was measured in 218 TARGETED SEQUENCING

220 single cardiomyocytes (n=4) at sarcomere lengths of 1.8 to 2.2 μm (see Figure 3). Our functional measurements of passive stiffness, which is largely based on titin composition in the heart, revealed a very low passive force development (1.0±0.3 kn/m 2 ) at a sarcomere length of 2.2 μm in the PPCM sample compared to previously reported values in control hearts (~2.5 kn/m 2 ). 29,30 Analysis of titin isoform composition showed a shift towards the more compliant N2BA isoform evident from a higher N2BA/N2B ratio (0.72±0.02; mean of triplo) in the PPCM heart compared to the previously reported ratio (0.39±0.05) in control hearts. 30 DISCUSSION This is the first report of a comprehensive genetic analysis in a large series of cases with familial occurrences of PPCM and DCM. We identified pathogenic mutations in cardiomyopathy-related genes in 4/18 families (22%) and VUSs that are likely to be pathogenic in 6 other families (33%). These data support the earlier finding that PPCM can be part of familial DCM. 11,12 Cascade genetic screening can identify relatives at risk in those families in which an underlying mutation has been identified. Our data also specifically show a low recovery rate in our cohort (only 10%) compared to reports in other groups not selected for familial cases (recovery rates of around 25 to 50%), indicating that the presence of an underlying mutation or positive family history for cardiomyopathy in a patient with PPCM may be a prognostic factor for a low recovery rate. The targeted NGS approach that we have developed provides highthroughput, rapid and affordable molecular analysis for cardiomyopathies. 27 As accurate annotation of mutations in cardiomyopathies is of the utmost importance, 37 we were extremely careful in classifying these. 28 Our study has CHAPTER 4.2 Figure 3. Force measurements in heart tissue of GER4 II:1. Single cardiomyocyte of the PPCM heart sample (A). Passive force development was measured at sarcomere lengths of 1.8, 2.0 and 2.2 μm. (B) TTN IN PERIPARTUM CARDIOMYOPATHY 219