scieee AI-readable full text Open interactive document viewer

Subphenotype-based characterisation of signalling pathways in congenital heart defects (CHD)

Dombrowsky, Gregor

Abstract

Congenital heart defects (CHDs) are the most common (1 - 2 %) birth defects amongst live newborns. Despite more than 200 genes having been associated with the development of CHD, the molecular basis remains partially unclear, and only a minority of patients can be assigned to an underlying genetic variant. The identification and characterisation of novel genetic variants increases diagnostic yields, improves risk stratification and, thus, the overall clinical management of CHD patients. By analysis of an exome-based case-control study comprising of 3907 CHD-cases and 5157 population matched controls, this thesis focussed on enhancing the understanding of the genetic contribution to CHD in general and on revealing sub-phenotype specific differences. Established CHD-genes were investigated with respect to deleterious variants that might account for the patient´s phenotype. Clustering based on phenotypic subtypes followed by GO-term enrichment revealed distinct genes and mechanisms underlying the different CHD subtypes. Secondly, a permutation-based analysis of signalling pathways revealed 26 of the tested combinations to be significantly enriched. Molecular dynamics was utilised to determine the molecular mechanisms of variants localised in PTPN11 and PIK3CA. The last part was focussed on NOTCH1, a gene that has been consistently determined to be one of the key contributors to CHD. Besides truncating variants, alterations in disulfide-bonds in NOTCH1 were significantly enriched. Functional analysis showed a complete impairment of the signalling capability for the tested nonsense variants and moderately reduced capabilities for missense variants. In summary, this thesis provides novel insights into the genetic background, including specific CHD subgroups, which might offer novel surveillance and treatment strategies for CHD patients in the future.

Full text

Dissertation Subphenotype-based characterisation of signalling pathways in congenital heart defects (CHD) zur Erlangung des Doktorgrades Doktor rerum naturalium der Mathematisch-Naturwissenschaftlichen Fakultät der Christian-Albrechts-Universität Kiel Author: Gregor Dombrowsky Kiel, 18th July 2024 First examiner: Prof. Dr. Hinrich Schulenburg Second examiner: Prof. Dr. Marc-Phillip Hitz Date of the disputation: 13.09.2024 “Research means that you don’t know, but are willing to find out.” Charles F. Kettering Contents Eigenständigkeitserklärung 9 Abstract 9 Introduction 13 The human circulation system ............................... 14 Development of the human heart during embryogenesis ................ 16 Main types of CHD and prevalence ............................ 18 Multifactorial causes of CHD ................................ 24 Current diagnostic approaches and yields ........................ 25 Outlook to approaches in this thesis ............................ 27 1 Application of a diagnostic framework for variant interpretation in CHD 29 1.1 Introduction ....................................... 29 1.2 Material and methods ................................. 32 1.2.1 Curation of the exome dataset ........................ 32 1.2.2 Curation of a CHD-gene set ......................... 34 1.2.3 Implementation of the ACMG-System ................... 36 1.2.4 Clustering and Enrichment Testing ..................... 41 1.3 Results .......................................... 43 1.3.1 Curation of the analysed exome cohort ................... 43 1.3.2 Curation of a CHD-gene set ......................... 47 1.3.3 Identification of causal variants in the cohort ............... 49 1.3.4 Individual gene contributions ......................... 50 1.3.5 Result stratification by phenotype ...................... 62 1.4 Discussion ........................................ 68 2 Enrichment-testing of signalling pathways in CHD sub-phenotypes 75 2.1 Introduction ....................................... 75 2.2 Material and methods ................................. 81 2.2.1 Reactome gene sets .............................. 81 2.2.2 ROC analysis .................................. 82 2.2.3 Pathway analysis ................................ 83 2.2.4 Testing at gene-level ............................. 84 2.3 Results .......................................... 85 2.3.1 Evaluation of in-silico predictors ....................... 85 5 Contents 2.3.2 Enrichment testing for signalling pathways ................. 87 2.3.3 Identification of enrichment at gene level .................. 90 2.3.4 Expression of identified genes during heart developmental stages . . . 105 2.3.5 Interaction networks ..............................106 2.4 Discussion ........................................107 3 Variant characterisation in PTPN11 and PIK3CA using molecular dynamics 117 3.1 Introduction .......................................117 3.2 Material and methods .................................121 3.3 Results ..........................................122 3.3.1 Dynamic modelling of SHP2 .........................123 3.3.2 Dynamic modelling of PI3K ..........................127 3.4 Discussion ........................................139 4 Phenotypic variability in patients with NOTCH1-variants 143 4.1 Introduction .......................................143 4.2 Material and methods .................................147 4.2.1 Material .....................................147 4.2.2 Aggregation of published NOTCH1 variants ................151 4.2.3 Enrichment testing ...............................152 4.2.4 DNA extraction from EDTA-blood samples .................152 4.2.5 Capillary sequencing .............................152 4.2.6 Site-directed mutagenesis ..........................153 4.2.7 Gateway cloning ................................154 4.2.8 Preparation of plasmid DNA .........................155 4.2.9 Cell culture assays ...............................155 4.2.10 Digenic analysis ................................158 4.2.11 Replication ...................................158 4.3 Results ..........................................158 4.3.1 Aggregation of published NOTCH1 variants ................158 4.3.2 Enrichment testing ...............................159 4.3.3 Carrier description ...............................161 4.3.4 Functional testing ...............................169 4.3.5 Non-coding variants in NOTCH1 .......................174 4.3.6 Digenic involvement ..............................175 4.3.7 Replication of findings .............................176 4.4 Discussion ........................................180 Conclusion and outlook 189 6 Contents Acknowledgements 193 Appendix 195 Vectors .............................................195 Primer .............................................197 Published variants in NOTCH1 corresponding to Section 4.2.2. ............201 Chromatograms from capillary sequencing, NOTCH1 segregation testing . . . . . . 206 Western blot replicates ....................................212 Luciferase assays raw data .................................215 Abbreviations .........................................219 Bibliography 224 List of Figures 270 List of Tables 273 7 Eigenständigkeitserklärung Hiermit bestätige ich, Gregor Dombrowsky, dass ich die vorliegende Arbeit selbständig verfasst und keine anderen als die angegebenen Hilfsmittel benutzt habe. Sie wurde zuvor nicht an einer anderen Stelle im Rahmen eines Prüfungsverfahrens vorgelegt. Die Arbeit wurde unter Einhaltung der Regeln guter wissenschaftlicher Praxis erstellt. Die Stellen der Arbeit, die dem Wortlaut oder dem Sinn nach anderen Werken entnommen sind, wurden unter Angabe der Quelle kenntlich gemacht. Ein akademischer Grad wurde nicht entzogen. Kiel, den 18.09.2024 Gregor Dombrowsky 9 Introduction protect and anchor the heart. The rigidity of the fibrous pericardium prevents overfilling of the heart. The serous membrane of the pericardium and the epicardium are separated by a layer of fluid that reduces frictions during pumping of the heart. Development of the human heart during embryogenesis As the heart occupies a central role in the human organism, it is the first organ to develop during embryogenesis. Cardiogenesis begins in the second to third week of the development and is completed around week eight. The developmental steps outlined below are described based on a review of Buijtendijk et al.. 20 Cardiogenesis is a highly complex process that is tightly coordinated by spatial and temporal regulation from different signalling events.21 The cardiac tissue is predominantly a mesodermally derived structure. One of the earliest signs of cardiac development is the migration of precardiac cells from the primitive streak toward the myocardial plate, where they form paired cardiac crescents (also termed the cardiogenic region). Through this migration, the local signalling environment changes. The crescents eventually fuse to form the primitive heart tube around day 21 of embryogenesis. This structure comprises an inner layer of endothelial cells and an outer layer of cardiomyocytes, separated by an extracellular matrix, the so called cardiac jelly. These cells in the primitive heart tube are referred to as the first heart field. Already at this stage, primitive cardiomyocytes begin contracting, and the contractures are increasingly harmonised through gap junctions. Five regions in the heart tube can be distinguished which later develop into different structures of the heart. These regions include the truncus arteriosus, which results in the ascending aorta and the pulmonary trunk, the bulbus cordis which eventually becomes the right ventricle and the primitive ventricle, which develops into the left ventricle. The primitive atrium forms the anterior part of both atria and finally the sinus venosus, which results in the posterior part of the atria, the sino-atrial node and the coronary sinus. The heart tube undergoes a right-sided looping process, leading to a loss of its symmetry. This process takes place in the fourth week of fetal development. In parallel, the size of the heart grows five-fold. Growth of the heart at this point is mainly due to newly differentiated cardiomyocytes, which are termed the secondary heart field. 22 While the eventual left ventricle is solely comprised of cells from the first heart field, the left atrium, right atrium and right ventricle are composed of cells from both the first and second heart field (Figure 2).20 16 Introduction Figure 2: Schematic representation of the human cardiogenesis. a: Cell lineages in the heart development. b: Key stages of cardiogenesis. Figure retrieved from Tan and Lewandowski 2020.23 Looping finishes around day 28 in human embryogenesis. As a next step, septation occurs to create the four chambers. This part is initiated by the process of trabeculation, which refers to the infiltration of the endocardial layer with myocardial cells. It begins with cardiomyocytes secreting extracellular matrix (ECM) components into the atrioventricular canal and outflow tracts. Four cardiac cushions form from this ECM, the posterior and anterior AV-cushions and the parietal and septal outflow cushions. The cushions consequently are invaded by endocardial cells that undergo epithelial- to-mesenchymal transition (EMT), and cells derived from the neural crest. These structures expand towards each other, and eventually fuse to separate a primitive atrium and ventricle. The atrial septum formation is initiated around day 28 of development. Through increased proliferation of atrial cells, a primary atrial septum is formed until week six. Afterwards, a secondary, muscular atrial septum is formed through folding of the muscular wall of the atrium. Until birth, an opening remains present in the atrial septum called the foramen ovale. This shunt allows the exchange of blood from the right to the left atrium. Formation of the ventricular septum occurs slightly later, and is finished around the eighth week of development. The ventricle increases and the wall thickens. The trabeculae increase in length 17 Introduction through the addition of new cardiomyocytes at their base. The left and right ventricles become separated by a two layered septum with a muscular and a membranous part. In parallel, the atria expand and develop into the later outflow tracts. Ultimately, septation of the outflow tract occurs. A controlled, unidirectional blood flow in the heart is maintained by the four valves in the developed heart. These originate from the four cardiac cushions that form during septation. The atrioventricular valves form between the fifth to eighth week, followed by the semilunar valves between week five to nine. The cushions expand, and are subsequently invaded by mesenchymal cells. These primordial structures elongate by cellular proliferation and synthesis of extracellular matrix (ECM) components, resulting in the formation of the valvular leaflets. The leaflets undergo remodelling of the ECM, forming in the multilayered structure of the mature valve.24 The valves are subjected to strong mechanical forces during systole and diastole. To fulfil their function, they require both durability but also flexibility for movement. To achieve this, the mature valves are built from a three-layered extracellular matrix interspaced by valvular interstitial cells (VICs) and surrounded by an outer layer of valvular endothelial cells. A fibrosa layer mainly containing fibrillar collagen ensures durability. It is located towards the great vessels in semilunar valves and the ventricular side for the atrioventricular valves. The inner part of the valve termed spongiosa contains mainly proteoglycans contributing to compressibility. Finally, elastin is enriched in the ventricularis layer in semilunar valves or the atrialis layer for the atrioventricular-valves. Until birth, the fetal circulation system remains connected to the mother ´ s. The child’s pulmonary circulation remains inactive as the umbilical cord supplies oxygen-rich blood. Two anatomical features are leading to a blood flow from the pulmonary circulation into the systemic one. The ductus arteriosus connects the pulmonary artery and the descending aorta and the foramen ovale, which is an intentional opening between the septae. These provide circumvention of the pulmonary circulation during the fetal development. The development of the heart is finalised with the closure of the foramen ovale and the ductus arteriosous during the first days of live. The fossa ovalis in the right atrium is the remnant of the foramen ovale. Main types of CHD and prevalence Multiple studies have reported an increase in the prevalence of CHD throughout the years. 5,6,25 This observation appears to be rooted in increased reporting of milder lesions, whereas severe lesions stagnated over the last decades. 6 Severe lesions are estimated to contribute 10 % of all CHD. 26 An explanation is most likely to be found in advances of diagnostic applications and the ability to detect milder, asymptomatic malformations. 18 Introduction Overall, CHD is a complex disorder with a broad phenotypic spectrum. Defects can range from mild and even asymptomatic forms to severe, life-threatening malformations (see Figure 3). Depending on the type of cardiac malformation, left-to-right or right-to-left shunting can occur. This leads to a mixture of oxygenated and deoxygenated blood, resulting in unphysiological pressure conditions. Cyanosis - blue colouring of tissue - is observed in forms of CHD that lead to a reduced oxygen saturation in the systemic circulation. Typical manifestations of CHD comprise cardiac insufficiency, reduced physical fitness, failure to thrive and shortness of breath, especially under physical engagement. Additional effects may include cardiomyopathy, hypertension and arrhythmic disorders.18,27 CHD can also lead to syncopes and reduced immunological resilience, resulting in higher infection rates. 14,15 Severe forms of CHD often require treatment within the first year of life. Especially for the fraction of severe malformations, the overall life expectancy is significantly reduced in comparison to the general population. 28 Overall survival rates were reported at 87 % for one-year survival and 81.4 % for ten-year survival. 29 Yet, critical heart defects, such as the hypoplastic left heart, still show a comparably poor survival rate (one-year survival 17.4%, 95% CI 0.0–54.5). 29 However, with the development and implementation of new treatment techniques, survival rates for CHD patients have increased dramatically. Consequently, an increasing share of CHD patients reach adulthood so that adults make up most CHD patients nowadays. 25 It can also be expected to increase the frequency of familially transmitted variants, shifting the spectrum away from de novo variants. The most clinically relevant forms of CHD are described below. The defined groups follow the classification scheme applied later in this thesis (see Table 1.6). The information is summarised based on "Kompendium angeborene Herzfehler bei Kindern" by Blum et al.18 19 Introduction Figure 3: Overview of important congenital heart defects. The image on top shows the normal adult human heart. Shown below are six common congenital hear defects. The labels indicate the individual lesions which result in the respective condition. Figure reproduced with permission from Morton et al.2022.30 Septal defects Septal defects represent the most common form of congenital heart defects. 3 They consist in an opening of the septal wall. Based on their anatomy, they can be divided into atrial septal defects (ASD), ventricular septal defects (VSD) and atrioventricular septal defect (AVSD). Atrial 20 Introduction septal defects are further subdivided into ASD type I (approx. 20 %), ASD type II (approx. 70 %) and Sinus-venosus defects (approx. 10 %). Ventricular septal defects lead to an interventricular communication with a left-to-right shunting due to the pressure difference between pulmonary and systemic circulation. Oxygenated blood partially returns into the pulmonary circulation. As a consequence, patients often develop pulmonary hypertension. Other symptoms include breathlessness and reduced resilience. Septal defects are a risk factor for heart insufficiency and poor growth. Small septal defects can be asymptomatic, and can even spontaneously close without treatment. Septal defects are furthermore part of more complex malformations such as Tetralogy of Fallot. 18 Atrioventricular septal defects (AVSD) result in the communication of both atria and both ventricles, leading left-to-right shunting. Depending on the size, septal defects require surgical or catheter-based intervention. Valvular phenotypes Valvular phenotypes can affect any of the four heart valves. Therefore, they can contribute to obstructions of the left and the right outflow tract. The spectrum regarding morphology and severity of malformations is broad. Depending on a stenosis or dysplasia of variable degree, the movement and ability to close can be impaired. Disability to properly close the dysplastic valve results in regurgitation (backflow) of blood, thus increasing the workload for the cardiac chamber, and posing the risk of heart failure. In its most severe form, an atresia of the valve can be present and the bloodflow may be entirely blocked. The most common valvular malformation is the bicuspid aortic valve (BAV) as compared to the normal tricuspid aortic valve. 31 BAV can often have very mild symptoms. Valvular phenotypes can also manifest in more complex conditions, such as the hypoplastic left heart, in which hypoplasia of the left ventricle is mediated through the impairment of both the aortic and the mitral valve. Valvular phenotypes can be treated by dilatation or by interventional valve replacement. Obstructions in the right ventricular outflow tract (RVOTO) Malformations altering the right ventricular outflow tract impair blood flow into the pulmonary circulation. These comprise stenosis or atresia of the pulmonary valve or artery. Stenosis occurs in 60 % of cases as a valvular form, but can also be present sub- or supravalvularly. Due to stenosis, the right ventricle cannot sufficiently transfer blood to the pulmonary circulation, leading to hypertrophy and ischaemic damage of the cardiac musculature. Treatment may include Balloon-based dilatation of the stenosis or valve replacement surgery. In the case of pulmonary atresia (PA), the valve is completely closed and bloodflow into the pulmonary vessels is blocked. Without the foetal bypassing systems, this represents a life-threatening condition. PA is often accompanied by a VSD. This combination reflects a subform of Tetralogy of Fallot. The therapeutic options include reconstruction or bypassing of the aplastic valve. Overall, there is a considerable overlap between RVOTO and conotruncal defects. 21 Introduction Impairment of the left ventricular outflow tract (LVOTO) Left ventricular outflow tract obstructions (LVOTO) restrict the bloodflow of oxygenated blood into the systemic circulation. They comprise sub-, supra- or valvular aortic stenosis, dysplasia or atresia of the aortic valve, aortic coarctation and hypoplastic left heart syndrome (HLHS) as the clinically most critical condition. LVOTOs generally result in an inadequate oxygen supply to the organism therefore causing cyanosis. Coarctation of the aorta describes a narrowing of the aorta, usually of the aorta descendens after the aortic arch. The vessels supplying the upper regions of the body originate before the coarcation and thus remain unaffected. The lower body is insufficiently supplied with oxygen, which may damage the internal organs including liver and kidneys. The narrowing leads to an extended workload for the left ventricle, increasing the risk of ventricular failure. As an extreme manifestation, the aortic arch can be fully interrupted (interrupted aortic arch; IAA). In this phenotype, the interruption must be circumvented by the presence of a persistent ductus Botalli and a VSD which results in the supply of partially oxygenated blood to the lower body. Treatment of aortic coarctation usually involves surgical anastomosis, in which the aorta is connected before and after the narrowing. One of the most severe left ventricular pathologies is the hypoplastic left heart syndrome (HLHS), where the left ventricle is severely underdeveloped due to a stenosis or atresia of both the mitral and the aortic valve. Consequently, the aorta is also poorly developed, and the body is not sufficiently supplied by oxygenated blood. To maintain at least a partial supply of oxygen, this defect requires a connection between the atria e.g. in form of a patent ductus arteriosus and an ASD to allow for the crossing of oxygenated blood into the systemic circulation. Unless treated early after birth, this heart defect shows a high lethality. As the diagnosis of HLHS is often done prenatally, intervention can and must begin early after birth. The ductal connection is kept open by application of prostaglandins. Treatment options include a series of surgical interventions to create a Fontan circulation or heart transplantation in some cases. The first is realised by three consecutive procedures. The Norwood-operation which is conducted in the first days after birth, followed by Glenn-procedure and the Fontan procedure. Ultimately, a single functional ventricle is created. Despite progress in the treatment of HLHS, it still shows a high lethality. Complex (conotruncal) heart phenotypes Conotruncal defects consist of a spectrum of morphologies that exhibit malpositioned origins of the great vessels. Tetralogy of Fallot (TOF) represents around 8 % of congenital cardiac malformations, making it one of the most common complex phenotypes. As the name suggests, TOF is characterized by the combination of four elements. The impairment of the pulmonary valve results in the obstruction of the right ventricular outflow tract. This impairment can range from stenosis up to a complete atresia of the valve. It is combined with a VSD and an overriding aorta, meaning that the origin of the aorta is located further to the right than usual. In combination, a higher amount of low-oxygenated blood enters the body ´ s circulation. Lastly, a hypertrophy of the right ventricle occurs due to the impaired pulmonary valve function. Without intervention, TOF shows a high rate of mortality. Corrective surgery 22 Introduction is performed to close the septal defect and re-establish the proper function of the pulmonary valve. Transposition of the great arteries (TGA) is a condition where the connection of the great vessels is reversed. As a consequence, two separate circulations are present. During pregnancy, this can be compensated through the foetal bypass systems. Postnatally, symptoms become apparent. The systemic circulation is supplied only by intermixed oxygen-rich and -poor blood resulting in a cyanotic situation. The reduced oxygenation in the systemic circulation furthermore results in an increased workload for the left ventricle, which results in damage of the left atrial myocardium. TGAs occur isolated in > 70 % of affected individuals, but can also be accompanied by VSDs or obstructions in the left- or right -outflow tracts. Inital treatment consists of prostaglandin administration and balloon atrial septostomy. The latter is a procedure to widen the foramen ovale in order to increase the exchange of oxygenated blood to the right heart. Subsequently, TGA can be surgically corrected by an arterial switch operation. Transposition of the great arteries can occur in a congenitally corrected form (ccTGA). In this case, the great vessels, as well as the left and right ventricles, are transposed. The morphologically left ventricle thus acts as the functional right ventricle an vice versa. Overall, cardiac circulation is function-able, however the valves and myocardial walls are not appropriately adapted to their tasks. The symptomatology is usually mild. Optionally, the correct anatomy can be restored by a double-switch operation. In a truncus arteriosus communis (abbreviated truncus), both the aorta- and the pulmonary outlet originate from a common vessel connected to the heart. Septation has not occurred correctly and the presence of a VSD connects both heart chambers. Also, both valves are present in a combined form as a truncal valve. Insufficiency of the truncal valve results in heart insufficiency. As a result of a truncus arteriosus, an increased burden is posed on the pulmonary circulation, resulting in pulmonary hypertension. If untreated, this condition shows high lethality during the first months of life, and requires surgical repair or creation of a Fontan-circulation. Double outlet right ventricle (DORV) represents a condition, in which both the aorta and the pulmonary artery originate in the right ventricle. DORV can be considered an extreme form of the overriding aorta present in TOF. Similar conditions can also occur as a double inlet left ventricle (DILV), double inlet right ventricle (DIRV) or double inlet right ventricle (DOLV). DORV is the most common of those with a fraction of 1-3 % of CHDs. 18 These conditions are usually accompanied by a VSD, which allows for the crossing of blood between the ventricles. Other conditions During the prenatal development, connections between the cardiac circulation systems are present. The ductus arteriosus provides a connection between aorta and pulmonary artery. The foramen ovale is a deliberate septal opening between the two atria. These structures bridge the inactive pulmonary circulation. When the pulmonary circulation of the child becomes 23 Introduction activated after birth, the pressure equilibrium changes altering the blood flow. As a result, these fetal connections normally close within the first days of life. However, these foetal structures can persist and result in a constant exchange of blood between the pulmonary and the body circulation. A patent ductus arteriosus (PDA) or persistent foramen ovale (PFO) can be present especially in preterm gestations. Most of these close spontaneously in the first year of life, and often only have a small functional impact. In severe forms of CHD, where the pulmonary and systemic circulation would be fully separated (e.g. Transposition of the great arteries), presence of this cross-bridging may be beneficial and the PDA is deliberately kept open on e.g. by the administration of prostaglandins. Multifactorial causes of CHD The overall aetiology of CHD is complex and remains incompletely understood. Both genetic and non-genetic factors are known to contribute to CHD. Non-genetic noxae include e.g. pre-gestational maternal diabetes, maternal phenylketonuria, as well as exposure to organic solvents, pesticides or alcohol. 32–35 In addition, exposure to viral infections during early pregnancy has been shown to contribute to CHD. 36 It has furthermore been shown that genetic and environmental modifiers do not affect the risk for CHD independently. Instead, there are additive effects and extensive crosstalk.37,38 A genetic contribution to CHD can be deducted from an increased recurrence risk in families with previously affected individuals. 7,39 Monozygotic twins show a higher concordance for the presence of CHD than dizygotic twins. 39 Yet, discordant twins in CHD are reported, therefore suggesting an additional impact of environmental and epigenetic contribution. 40 Furthermore, due to the increased likelihood of homozygous carriership for recessive variants, offsprings from consanguineous couples are at an increased risk to be affected by CHD. 41 The risk of a child to be affected by CHD is 4.1 % when a sibling is affected by CHD compared to 1.1 % in unaffected families. 42 . Recurrent familial CHD of similar phenotypes was found in 50% of these sibling pairs. Still, the phenotype differs even in families with a clear segregation of a single variant with CHD. This variable phenotypic expressivity is a characteristic of CHD that is yet to be fully understood. The complexity of CHD is illustrated by the reduced penetrance and variable expressivity. 43 Although variants in causative genes are strongly associated with CHD and show an increased burden, unaffected carriers are also observed. 39 Furthermore, individuals carrying identical or highly similar variants often show different morphological phenotypes. The mechanisms underlying these observations remain partially unclear. The variable expressivity, taken together with the reduced penetrance, contradicts a simplified model of a monogenetic disorder. The current view suggests a multifactorial aetiology with both environmental factors and a spectrum of genetic variation.7,44 24 Introduction Variations with a large effect size and strong penetrance are at one end of this genetic spectrum. These variants tend to follow a classical Mendelian inheritance pattern. Due to a high selective pressure, they are characterised by a low population allele frequency and often arise de novo. On the other end of the spectrum, additive effects of common variants have been identified through genome wide association studies (GWAS). Such variations individually have a small effect size, but contribute a relevant share to the overall disease risk when cumulating their combination. Risk estimation based on these common SNPs can be done through the development of polygenic risk scores. An intermediate group of variations results in a moderate effect size that requires modulation through the presence of additional variants. These variants are increasingly integrated into concepts of digenic and oligogenic inheritance patterns. All types and sizes of genetic alterations have been associated to the development of congenital heart defects. 7 The spectrum ranges from aneuploidies affecting whole chromosomes, structural variations of different sizes, to single nucleotide variants affecting genes involved in heart development. Regarding aneuploidies, Trisomy 21 is the most common finding in context with CHD. Approximately 40-50 % of affected individuals show CHD with enrichment of septal defects. 45 Additionally, individuals with Turner-syndrome (X0-karyotype) show cardiovascular abnormalities in 50 % of cases.46 Large structural variants (SVs) are also often detected in patients with CHD. 12 Given that these alterations usually affect multiple genetic loci, of which only a subset is involved in cardiogenesis, SV-linked CHD mainly occurs in a syndromic fashion. 22q11.2 deletion syndrome accounts for up to 2 % of CHD with an excess of conotruncal defects. 47 TBX1 coding for a transcription factor required for the accurate heart development is located in the deleted region. Williams-Beuren Syndrome (deletion of 7q11.23 including the ELN-locus) presents often with arterial abnormalities such as supravalvular aortic stenosis. 7 Other examples include the deletion or duplication of 1q21.1 affecting the GJA5 genelocus as well as deletion of 8p23.1 (GATA4).7 Finally, single-nucleotide variants (SNVs) and small indels as well as copy number variations (CNVs) in dosage sensitive CHD genes may contribute to the development of the disease. Numerous genes have been associated with syndromic and non-syndromic forms of CHD. 7,48 Examples include CHD7 (CHARGE-syndrome) 49 ,NKX2.5 50 , or genes of the RASopathy-spectrum 51 . In general, causal genes often code for transcription factors, proteins in signalling pathways or structural components of the developing heart.7 Current diagnostic approaches and yields Currently, a genetic or environmental cause for CHD can be identified in 15 - 35 % of cases. 52–54 Congenital heart defects have been associated with a wide range of genetic malformations, ranging from aneuploidies and large chromosomal rearrangements to single nucleotide exchanges. Since not 25 1 Application of a diagnostic framework for variant interpretation in CHD (warm VUS) or benignancy (cold VUS).73 Aim The goal of this Chapter was to establish a subset of the analysed cohort, in which CHD could be attributed to a rare variant identified by ES. To that end, a curation of associated genes followed by the application of ACMG classification to the variant data was performed. The observations of established CHD-genes were to be quantified and a stratified analysis based on sub-phenotypes, analysis of GO-term enrichment and STRING-interactions was employed to provide insights on specific genotype-phenotype correlations. 1.2 Material and methods 1.2.1 Curation of the exome dataset Background The work presented herein is primarily based on a cohort of 3907 ES-patients with CHD and 5157 population-matched controls. This cohort is part of previous publications. 79,80 The inclusion criterion for patients was the presence of a CHD that required intervention within the first year of life. Samples with known structural variations, such as 22q11 syndrome, or chromosomal aneuploidies such as trisomy 21 were excluded if such a diagnosis was reported. The cohort consisted of samples collected or available through the German Center for Cardiovascular Research (DZHK), the German Competence Network for Congenital Heart Defects, the National Register for Congenital Heart Defects and the University of Nottingham (UK). Additionally, samples previously published 79 have been included. This comprised a subset of samples with CHD from the Deciphering Developmental Disorders-study (DDD) was included. The DDD project, carried out by the Wellcome Sanger Institute in the UK, performed ES on a cohort of children with undiagnosed developmental disorders.81 A subset of these individuals also displayed congenital heart defects. These samples had been subjected to ES on DNA from peripheral blood using different versions of the SureSelect Exome chips (Agilent). Enriched libraries were subjected to 75-base paired-end sequencing (Illumina HiSeq) following the manufacturer ´ s instructions.79 Data curation and quality filtering of the sequencing data were performed in accordance with previous work of our group. 79,82 In short, alignment was initially done using the Burrows-Wheeler Aligner (BWA) based on the Ensembl reference build of the human genome GRCh37. 83 Variant calling was performed using the Genome Analysis Toolkit (GATK V4.1) following the GATK best-practice guidelines. 84 Quality control based on genotype quality, coverage, allelic balance and Variant Quality Score Recalibration (VQSR) were performed using Hail 0.2.7. Samples were restricted to European ancestry (probabilityEUR > 0.8). Provided metadata at sample-level In addition to variant information, available metadata at sample level was supplied, including gender, relatedness to other samples in the cohort, information regarding age at sample-taking and, if applicable, age at death. Phenotypic information on cardiac 32 1.2 Material and methods and extracardiac malformations was provided in the human phenotype ontology format (HPO format). The HPO nomenclature system is based on a hierarchical classification of terms in which more general terms are iteratively split into more precise terms. 85 Each HPO entry is uniquely identifiable with a name and a corresponding number. Processing as part of this thesis Following the quality control steps outlined above, variants were functionally annotated using the Variant Effect Predictor tool (VEP v.104). 86 The annotation of positions and functional predictions were based on the canonical transcript as defined by Ensembl (https://www.ensembl.org). Using the Ensembl Assembly converter, variant positions were lifted from the initial GRCh37 build to GRCh38. 87 Gene names were used in accordance to the HUGO Gene Nomenclature Committee.88 The VEP tool functionalities were extended using the plug-ins CADD (version 1.6) and dbNSFP (version 4.1a) 89 to annotate different in-silico prediction algorithms. In detail, CADD 90 , MPC 91 , REVEL 92 and MVP 93 were subsequently used to assess deleteriousness of missense variants. For the classification of splice site variants, SpliceAI was used. 94 As suggested, a cut-off of ≥ 0.5 was applied to identify variants with an altering effect on splicing patterns.94 In addition to information annotated by the VEP-algorithm, the amino acid length of the protein corresponding to the transcript used (i.e. the canonical transcript) was retrieved from UniProt as of April 2020. 95 Furthermore, functional domains and sites of post-translational modification were retrieved from UniProt. Based on the protein-level consequences annotated by VEP, variants were collapsed into groups of similar mechanisms (Table 1.2). In order of decreasing impact, those were defined as proteintruncating variants (PTV), variants affecting splice sites (SPLICE), protein-altering variants (PAV) and synonymous variants (SYN). Variants that would fulfil multiple groups were assigned to the most severe consequence. The o/e score derived from the gnomAD-project allows an estimation of gene ´ s tolerance towards different types of variation, namely PTVs, PAVs and SYN. 96 Scores together with the 95 % confidence intervals were retrieved and annotated as of November 2019. Table 1.2: Grouping of consequences at protein level. Variants were grouped based on the highest applicable consequence listed in the table. Included consequencesVariant group Frameshift variant, Start lost, Stop gained, Stop lostProtein truncating variant (PTV) Splice region variant, Splice acceptor variant, Splice donor variant Splice variant (SPLICE) Frameshift variant, Start lost, Stop gained, Stop lostProtein truncating variant (PTV) Stop retained variant, Synonymous variantSynonymous variant (SYN) 33 1 Application of a diagnostic framework for variant interpretation in CHD Frequency filters Minor allele frequencies (MAF) and minor allele counts (MAC) were assessed based on gnomAD v2.1. Compared to the analysed cohort, samples of German origin are underrepresented in gnomAD, resulting in slightly biased approximations of allele frequencies. To account for this, allele frequencies from a set of internal unpublished control samples of German origin were included. As a further extension to the original sample set, allel frequencies from the UK-BioBank (UKBB) was included in the analysis. 97 Due to the differences in the applied enrichment kits, namely the Agilent SureSelect for the initial cohort and the IDT xGEN for UKBB, discrepancies in the covered variants were observed. Throughout the presented data, carrier information from the UKBB was only considered when both kits covered the affected position. Count information within the UKBB samples and coverage information was added as additional information per variant. Internal allele counts (AC) were generated based on the full cohort as shown in Equation 1.1. Internal allele frequencies were retrieved by dividing the AC by the number of samples (n= 9064). Internal allele frequencies and allele counts were restricted to European ancestry samples. ACint =nhet cases +nhet controls + 2·nhom cases + 2·nhom controls (1.1) Variants were considered rare if the MAF in gnomAD v2.1, the internal cohort, the added German control set and UKBB, if covered were below 0.1 %. A variant was considered ultrarare if the allele count was additionally ≤ 2 in all above-mentioned cohorts. The variants used were generally restricted to rare variants with an exonic consequence on the canonical transcript. Phenotype grouping The submitting istitution provided phenotype grouping for initial phenotype information. Annotation was done separately for cardiac and non-cardiac phenotypes defined based on the HPO-nomenclature. 85 All supplied cardiac phenotypes from the cohort were reviewed and equivalent terms were unified. At the beginning of the data curation, samples with unclear or conflicting phenotype information were removed (n= 19). Of note, only the term for the primary CHD was used. This means that complex malformations such as Tetralogy of Fallot (TOF) or the hypoplastic left heart syndrome (HLHS) were not taken into account for their individual parts. A total of 222 distinct cardiac phenotypes remained after this curation step. These phenotypes were sorted into 8 main-groups and 25 subgroups which further refine those main groups. Terms were allowed to be part of multiple categories, e.g. aortic valve stenosis would be considered as both a valvular and a left-ventricular outflow tract obstruction. Each case was assigned to all appropriate phenotypic main- and subgroups based on the reported phenotypes. 1.2.2 Curation of a CHD-gene set A list of disease-associated genes was compiled and further manually curated to reflect a set of high confidence CHD-related genes. Genes were retrieved from three sources and overlapped. These include Genepanelapp (https://panelapp.genomicsengland.co.uk/), hosted by the Genomics England 34 1.2 Material and methods project 98 , Gene2Phenotype (G2P) (https://www.ebi.ac.uk/ gene2phenotype/g2p_vep_plugin), jointly managed by the Wellcome Sanger Institute and the European Bioinformatics Institute (EBI) 99 and ClinGen (https://clinicalgenome.org/), an NIH-funded database 100 . All data sources were used as of May 2022. Genepanelapp is hosted by the Genomics England project and provides genepanels for numerous diseases with an expert-panel revised confidence rating for each gene-disease association. Genes with a “Green” status were considered high confidence. The panel from Genepanelapp included in the process are describe in Table 1.3 Table 1.3: Panels from UK Panelapp included for curation of the CHD-gene set. Panel included • Arrhythmogenic cardiomyopathy • Brugada syndrome • Cardiac arrhythmias • Cardiac arrhythmias - additional genes • Cardiomyopathies - including childhood onset • Catecholaminergic polymorphic Ventricular Tachycardia • CHARGE syndrome • Dilated cardiomyopathy - adult and teen • Dilated Cardiomyopathy and conduction defects • Familial non syndromic congenital heart disease • Hypertrophic cardiomyopathy - teen and adult • Idiopathic ventricular fibrillation • Kabuki syndrome • Left Ventricular Noncompaction Cardiomyopathy • Long QT syndrome • Progressive cardiac conduction disease • Pulmonary arterial hypertension • RASopathies • Sudden cardiac death • Sudden death in young people • Thoracic aortic aneurysm or dissection G2P is a freely available online resource providing information for gene-disease associations, including inheritance patterns and disease mechanisms. The confidence of the gene-disease association is assessed through expert panels and classified as "definitive", "strong", "moderate" or "limited". Genes with a "definitive" or "strong" status were considered high confidence. ClinGen is a database that integrates gene-disease associations from different sources and provides an assessment of confidence for the association. All gene-disease associations were downloaded and overlapped with the gene’s presence in one or both of Genepanelapp and G2P. A "definitive" or "strong" classification status was considered high confidence. Based on the classified strength of the association in the sources, genes were split into high- 35 1 Application of a diagnostic framework for variant interpretation in CHD confidence genes (Tier 1) and lower-confidence genes (Tier 2). Genes were considered for Tier 1 if they were concordantly reported with strong evidence in all of the used resources or were published in more than three independent references linking it to the disease. Tier 2 represents associations reported in at least one of the three sets with strong or moderate evidence but not in all three databases. In cases where the assessment of the different sources varied, further manual curation was carried out based on information collected from OMIM 101 , Clinvar 102 , and gene-specific publications retrieved via NCBI-PubMed, as well as the curated list provided by Kahlert et al. 2017.103 Based on the mentioned sources, genes were also reviewed for the mode of inheritance, disease mechanism and the (main) phenotypic association if this information was missing from the initially assembled dataset. Genes were accordingly reclassified if deemed appropriate after manual review of available literature. In total, 241 genes were assembled. These were divided into subgroups based on their most common disease association. Besides genes mainly associated with CHD, genes that are associated with a syndromic disorder that are frequently accompanied by CHD (herein termed "Optional CHD genes") were collected. Aside from CHD, genes reported with congenital cardiomyopathies and arrhythmias were also included. Genes were subdivided as follows: 1. genes confidently associated with CHD 2. genes associated with a syndrome where CHD is a known aspect of the phenotypic spectrum ("Optional CHD") 3. genes that are mainly reported in the context of cardiomyopathies 4. genes that are mainly reported in the context of arrhythmias 1.2.3 Implementation of the ACMG-System Assessment of variants identified in the cohort was performed in accordance with the ACMG guidelines and recent adaptations and appending guidelines. The selection of applied criteria was primarily based on the initial ACMG guideline by Richards et al. 2015 72 and the best practice guidelines for variant classification in rare diseases by Ellard et al. 2020. 73 The latter proposed further subclassification of variants of uncertain significance into six groups. Additional amendments were also included.74–78 Given the incompleteness of the available data for the ES-cases, mainly regarding haplotype information and unavailable segregation data within the family, not all criteria were applicable to the cohort. Details are depicted in the following sections. Criteria based on population data A notable indication of the pathogenicity of a variant is the frequency in which it is observed in the general population. Population frequencies were herein assessed by integrating data from gnomAD- 36 1.2 Material and methods exomes (V2.1), gnomAD-genomes (V3.1), the allele frequency within the tested cohort itself and in additional control cohorts, including the UKBB and an internal cohort of northern-european samples. PS4, PM2 Evidence of pathogenicity can be considered based on the increased prevalence of a variant in affected individuals compared to controls (PS4) and low allele frequencies or absence of carriers in population databases (PM2). Taking into account the reduced penetrance observed in CHD, PS4 was applied if a variant was present less or equal to two times in both gnomAD and the UKBB. PM2 was applied for variants with a minor allele frequency of ≤ 0.1 % for dominant and ≤ 1 % for recessive inheritance. Following the amendment issued by the ClinGen working group, the strength of PM2 was downgraded to supporting.104 BA1, BS1, BS2 Conversely, high carrier frequencies in controls indicate the absence of a functional impact of a variant. The BA1 was applied above an MAF of 5 % in any of the included datasets. BS1 was applied at > 0.1 % for dominant and > 1 % for recessive disorders. Variants where the BA1 criterium is not applicable were considered according to the published list of exceptions. 74 BS2 normally applies if a healthy carrier is observed for a gene with complete penetrance. To account for incomplete penetrance, the threshold was increased to ≥ 2 control carriers with the correct zygosity. BS2 was applied with a reduced strength (moderate) if the variant was observed in one control carrier regardless of zygosity for dominant inheritance. The variant had to be observed in one homozygous or hemizygous control carrier for recessive disorders. BA1, BS1, PS4 and PM2 were applied mutually exclusive. Criteria based on computational and predictive data PVS1 PVS1 represents the strongest criterium for pathogenicity in the ACMG guideline. It is applicable for null variants in genes where loss of function is a known disease mechanism. PVS1 was applied in line with the updated guideline. 75 Based on this publication, a decision tree was implemented to apply PVS1 at modified strengths. To estimate the intolerance of a gene towards loss of function, the observed/expected score from gnomAD was utilized as a reference. 96 It compares the observed frequency of variants against the number of observations theoretically expected if the variant had no selectional effect. The score is generated for loss-of-function, missense and synonymous variants and ranges from 0 to 1. Lower values indicate a lower tolerance against a variant type since the number of observations is lower than expected, suggesting a selection influence. As suggested in the reference publication, a cut-off of ≤ 0.35 was used to define constraint genes. 96 Using Decipher (https://www.deciphergenomics.org/) 105 , the location of the observed null-variant was reviewed for escaping nonsense-mediated decay (NMD). The effect of variants in splice sites was pre-screened for variants with a high effect prediction (SpliceAI ≥0.5) . These splice variants were evaluated as PVS1_very_strong unless they were 37 1 Application of a diagnostic framework for variant interpretation in CHD present in the last 10 % of the protein, which was classified as PVS1_moderate. PS1, PM5 PS1 is applicable when the same variant at the amino acid level has previously been observed and reported as (likely) pathogenic. This was implemented using Clinvar as a reference database. For applying PS1, the report had to be classified without conflicts by multiple submitters, be reviewed by an expert panel, or be part of a guideline. If phenotypic details were incomplete for entries or only a single submission was present, PS1 was downgraded to supporting. If the same position but with a different amino acid substitution was previously reported as (likely) pathogenic, PM5 was considered applicable. Similar to PS1, the entry was required to contain multiple concordant submissions, be reviewed by an expert panel or be part of a guideline. PM4 PM4 applies for inframe deletions or insertions not located in a repetitive region. PM4 was applied as moderate if in silico prediction reported a deleterious effect (PP3). Otherwise it was downgraded to supporting evidence. PM4 was also applied as moderate for stop-loss variants. PP3, BP4 The criteria PP3 and BP4 integrate the variant assessment from in-silico prediction tools estimating the deleteriousness of missense variants. Initially, the strength of those criteria was only considered to be supportive. A recent publication proposed a refinement to the PP3 and BP4 criteria. 76 CADD, MPC and REVEL scores were applied to evaluate variants according to Table 1.4. As a starting point, each score was separately evaluated per variant. In cases where conflicting results were observed, neither PP3 nor BP4 were applied. If the strength of the prediction differed, the median evaluation (rounded up) was used (e.g. strong + supportive = moderate; strong + moderate = strong). Table 1.4: In silico predictors used for determining PP3 and BP4. Shown are cut-off values applied for CADD, MPC and REVEL score. For pathogenic criteria, the upper boarder is included in the range for the applied strength, for benign criteria, the lower barrier is included. The table was generated based on Pejaver et al. 2022.76 BP4 (benign)PP3 (pathogenic) Method Very StrongStrongModerateSupportingSupportingModerateStrong -≤ 0.150.15 - 17.317.3 - 22.725.3 - 28.1≥ 28.1-CADD ----1.360 - 1.828≥ 1.828-MPC ≤ 0.0030.003 - 0.0160.016 - 0.1830.183 – 0.2900.644 - 0.7730.773 - 0.932≥ 0.932REVEL BP3 Inframe insertions and deletions in repetitive regions without a known function were not assessed. Thus, the corresponding criterium (BP3) was not applied. 38 1.2 Material and methods BP7 Synonymous variants were considered for BP7, unless they were classified as (likely) pathogenic in Clinvar or predicted to alter splicing based on SpliceAI. Criteria based on functional data PS3, BS3 PS3 and BS3 reflect evidence generated by functional data on the variant. Functional data generated from in vivo or in vitro experiments was not feasible for all variants in an automated and comparable fashion. The corresponding criteria (PS3, BS3) were not applied. PP2, BP1 To evaluate the gene constraint against missense and truncating variants with respect to a potential mechanism of disease, the observed expected score (o/e) was retrieved from gnomAD. 96 As suggested in the reference publication, a cut-off of ≤ 0.35 was used to define constraint. 96 In addition, the local, region-specific constraint was checked using Decipher. Conversely, altering variants observed in a gene, which does not show constraint against such variants, can be classified benign using BP1. PM1 PM1 applies to variants located in a mutational hot spot or critical and well-established functional domain without benign variation. Hotspot regions of high missense constraint were identified and retrieved using the Decipher protein browser. Functional domains were annotated using the InterPro database as annotated in the variant effect predictor output. 106 Annotations containing compositional bias were not considered as functionally relevant domains. The presence of previously reported variants in those domains, especially the absence of variants consistently reported as benign, was verified using Decipher. Overlapping altering variants were classified as PM1_moderate in the absence of benign variants. If in-silico predictors suggested a functional impact (PP3), this was upgraded to PM1_strong. If isolated benign variants had been reported in the domain in question but did not overlap with the variant, PM1_supporting was applied if PP3 was also applicable. Criteria based on segregation and inheritance data PS2, PM6, BS4 Segregation information is generally a strong indication of whether a variant might be relevant to the phenotype. A large fraction of the cohort represents single-index patients. Phenotypic and especially genotypic information on the parents and other family members were therefore unavailable for the majority of samples. Consequently, criteria concerning the inheritance or variant segregation in the family were not broadly applied to the whole set (PS2, PM6, BS4). In cases where parental genotype information was available either through trio-ES-data or Sanger-sequencing based segregation analysis, those criteria could be evaluated for specific variants. Since Sangersequencing based segregation of the specific variant is insufficient to confirm maternity or paternity, PM6 was applied instead of PS2. 39 1 Application of a diagnostic framework for variant interpretation in CHD PP1 Co-segregation of the variant with the disease in multiple affected family members was not applicable since the available data was generally not covering sufficient affected members per family. Criteria based on haplotype data Due to the technical limitations of short-read ES, information on the allelic haplotypes was generally unavailable. Therefore, no information on whether two variants are present in cis or trans are retrieved. If indeed two distinct variants in the same gene were identified, those were considered as possibly compound heterozygous. The ACMG criteria covering haplotype information (PM3 and BP2) were not applied to the dataset. Criteria based on other criteria PP4 As outlined before, congenital heart defects are a highly complex group of malformations and follow a multifactorial aetiology. PP4, which is applicable when the observed phenotype is highly specific for a disease with a single gene aetiology, was thus not applied. PP5/BP6 PP5 and BS6 have been discontinued in the current guidelines and were not applied. 77 BP5 BP5 applies for variants in a case where an alternative molecular basis can be identified. Since in the setting of a research-based ES, the suspected cause is not known, a prioritisation of one variant over the other is not feasible. BP5 was not applied. Evaluation of variants in Titin TTN codes for the Titin protein, a component of the sarcomeric muscle structure. Heterozygous variants in TTN can be causative for different types of cardiomyopathy. Coding variants are widespread as this is one of the largest proteins produced in the human organism. Variants in TTN were only considered if they were previously reported as pathogenic or were located within the A-Band, which is considered the hotspot location for functionally relevant variants.78 Summary Variants were classified based on the applicable criteria (see Table 1.1). Table 1.5 provides a summary of the workflow used for the CHD-cohort. For all (likely) pathogenic variants or VUS, the plausibility of a link to the reported phenotype was manually checked by evaluating the following aspects: • The zygosity fits to the described inheritance pattern for the gene. Heterozygous variants in a recessive gene were noted as carriership. • Trio information was included wherever available to determine de novo status of the variant. Variants in dominant genes that were inherited from a healthy parent were usually not considered except when there was clear evidence for a pathogenicity of the variant and incomplete penetrance was reported in the gene. 40 1.2 Material and methods • The overlap between the phenotype of the patient and the reported phenotypes associated with the disease was also considered. Genes exclusively reported in the context of cardiomyopathies or arrhythmias were only considered for patients who displayed such features. 1.2.4 Clustering and Enrichment Testing Based on the subset of pathogenic, likely pathogenic and tepid, warm or hot VUS ( ≥ 3.3) variants identified in the ES cohort were subjected to a hierarchical clustering analysis to examine whether certain genes associate more closely with specific CHD subtypes. Only genes with at least five independent carriers were considered, reflecting the median number of individuals across the evaluated genes. Fifty-four genes were thus included. The variant frequency per phenotypic subgroup was corrected for the gene size to reflect the number of carriers per 1000 amino acids. The data was scaled and centered using the scale()-function in R. Clustering was applied using the pheatmap()-function based on Euclidean distance and the Wardclustering algorithm. Phenotype-specific clusters were manually defined. The web tool version of Enrichr (https://maayanlab.cloud/Enrichr/) was used to retrieve enriched terms per phenotype cluster. 107,108 Enrichments regarding the "GO_biological process" and "Reactome pathways" were evaluated. The top ten associations per phenotype were identified based on the "combined score", which represents the product of the p-value derived from the Fisher exact test and the z-score of the deviation from the expected rank. 41 1 Application of a diagnostic framework for variant interpretation in CHD Table 1.8: Genes associated with diseases present in the analyesd patient cohort. Tier1: Genes with high evidence, Tier2: Genes with moderate evidence. XLD: X-linked dominant, XLR: X-linked recessive. Genes Mode of inheritanceTier ABCC9, ABL1, ACTC1, ARHGAP31, ARID1A, ASXL1, BRAF, CBL, CDK13, CFC1, CHD4, CHD7, EHMT1, FBN1, FBN2, FGFR2, FLT4, FOXC1, GATA4, GATA6, GDF1, GJA1, HRAS, JAG1, KANSL1, KDR, KMT2D, KRAS, LZTR1, MAP2K1, MAP2K2, MED13L, MRAS, NF1, NKX2-5, NODAL, NOTCH1, NOTCH2, NR2F2, NRAS, PPP1CB, PRKD1, PTPN11, RAF1, RIT1, RRAS2, SETBP1, SHOC2, SKI, SMAD3, SMAD4, SOS1, SOS2, TAB2, TBX1, TBX20, TBX5, TGFBR1, TGFBR2, TRAF7, ZEB2 XLD: BCOR, FLNA Dominant CHD Tier1 ADAMTS19, B3GALT6, CFAP53, DDX11, ESCO2, EVC, EVC2, HYLS1, MMP21, PIGL, SH3PXD2B, SLC2A10, STRA6 XLR: MED12, MID1, NONO Recessive ACVR2B, COL1A2, CRELD1, GATA5, RBPJ, ROBO4, RRAS, TFAP2B, TLL1, CITED2, FOXH1, HAND2, IRX4, SMAD6 XLD: NAA10 Dominant CHD Tier2 ALDH1A2, HYAL2, IGFBP7, NKX2-6, PLD1, RBM8A, SPRED2 Recessive MYOM2, CRKL, SEMA3D Unclear ADNP, COL1A1, CREBBP, ELN, EP300, NSD1, RAI1, SALL1 XLD: KDM6A Dominant Optional CHD Tier1 GUSB, IDUA XLR: ZIC3 Recessive CSNK2A1, KAT6B, KMT2A, NAA15, NIPBL, SALL4, SF3B4, SMC3 XLD: SMC1A Dominant Optional CHD Tier2 ALX3, B3GAT3, NEK1, CHST3, EOGT, FKBP14, IPO8, LTBP2, MEGF8, MKKS, RAB23 XLR: ATRX, GLA, GPC3, RBM10 Recessive ACTA1, ACTN2, ACVRL1, BAG3, CPT2, CSRP3, DES, DSC2, DSG2, DSP, EYA4, FHOD3, FLNC, IDH2, JPH2, JUP, LDB3, LMNA, MYBPC3, MYH7, MYL2, MYL3, MYPN, NEXN, PKP2, PLN, RBM20, RYR2, SGCD, SLC40A1, TCAP, TMEM43, TNNC1, TNNI3, TNNT2, TPM1, TTN, TTR, VCL Dominant Cardiomyopathy Tier1 ACAD9, ACADVL, ALPK3, COX15, DNAJC19, DOLK, GAA, HAMP, HFE, HFE2, MLYCD, MUT, PCCA, PCCB, PPCS, SLC22A5, SPEG XLR: DMD, L AMP2, TAZ Recessive PRDM16, CDH2, MIB1 Dominant Cardiomyopathy Tier2 MRPL44, C1QBP, ELAC2, GTPBP3, AARS2, AGK, AGL, ALMS1, EPG5, HADHA, MTO1, PPA2 XLR: IDS Recessive ANK2, CACNA1C, CALM1, CALM2, CALM3, HCN4, KCNE2, KCNH2, KCNJ2, KCNQ1, MYH6, PRKAG2, SCN5A, TNNI3K Dominant Arrhythmia Tier1 CASQ2, KCNE1, TECRL, TRDN XLR: EMD, FHL1 Recessive CACNB2, CAV3, TGFB3, TRPM4, SCN1B, SLC4A3 Dominant Arrhythmia Tier 2 48 1.3 Results 1.3.3 Identification of causal variants in the cohort Overall, 49637 variant-carrier pairs (comprising 26089 distinct variants) in genes associated with congenital cardiac malformations, arrhythmias or cardiomyopathies were initially identified in the analysed cohort. Of these, 17725 variants were only present in a single individual, while the rest was harboured by multiple samples. This set of variants was evaluated as per the classification workflow described in the methods (Section 1.2.3). The results can be found in Table 1.9. Most of variants (86 %) were classified as benign or likely benign, while 2 % were evaluated as (likely) pathogenic. Table 1.9: Classified were 49637 variants identified in the cohort affecting genes related to congenital cardiac phenotypes (see Table 1.8.) Shown in the lower part is the subset of CHD and optional CHD-genes. VUS = variant of uncertain significance, AD = autosomal dominant, AR = autosomal recessive, n.a. = unclear inheritance pattern, XL = X-linked. Pathogenic (0.4%) Likely pathogenic (1.6%) VUS (3.5) (0.4%) VUS (3.4) (0.7%) VUS (3.3) (1.3%) VUS (3.2) (3.0%) VUS (3.1) (2.5%) VUS (3.0) (4.1%) Likely benign (65.6%) Benign (20.4%) ControlCaseControlCaseControlCaseControlCaseControlCaseControlCaseControlCaseControlCaseControlCaseControlCase 415510031117721001076911912940917042526856913021956449464126AD All genes 9151961604444715725316133923731425057840036433052277261AR 234221977936922339226126101n.a. 1732306220142110123326301147816641037187123XL 14177301497611261751863373105768865197419691069186671387955364611Sum 2118392238515575446983217952341402596433466522231888AD CHD and optional CHD genes only 4979771620352588611158712710127818517071369148135AR 234221977936922339226126101n.a 162200501681039135131959501054648165114XL 71331223232480921181411492444462443904865169533690826622238Sum When focussing on variants linked only to syndromic or non-syndromic structural congenital heart defects (CHD Tier1 + 2, optional CHD Tier1 + 2), 24545 variant-carrier pairs (comprising 13024 distinct variants) were found. Within this subset, 405 variants in 386 individuals (346 CHD-cases, 40 controls) were classified as pathogenic or likely pathogenic. Furthermore, (likely) pathogenic variants were identified in genes associated to cardiomyopathy or arrhythmia in eight patients that match the corresponding phenotype. In total, this amounts to 8.9 % of cases where a possibly causal variant could be identified. Including hot, warm and tepid VUS variants extends this fraction to 14.2 % based on a total of 731 variants in 689 samples (544 CHD cases and 145 controls). Monoallelic carriership of a variant (class 3.3 or higher) in a recessive gene was identified for 431 samples, of which 204 were cases, and 227 were controls. At the time of the initial annotation, 248 of 732 (33.8 %) of the identified variants (VUS > class 3.2 and pathogenic in CHD genes) had not been reported previously in publications or databases (Clinvar, dbSNP). For (likely) pathogenic variants, unreported variants accounted for 146 of 405 findings (36.0 %). 49 1 Application of a diagnostic framework for variant interpretation in CHD 1.3.4 Individual gene contributions Next, the identified variants were stratified based on the affected genes. Figure 1.2 shows the contribution per individual genes. When considering only (likely) pathogenic variants, PTPN11 and NOTCH1 were the most commonly identified genes with 28 and 20 cases, respectively. For both genes, no control carriers were identified. NOTCH1 was most frequently observed in non-syndromic CHD cases. Other recurrent genes such as SMAD4 and KMT2A were found mostly in syndromic cases. As expected, including variants of uncertain significance (VUS class ≥ 3.3) increases the identification of variants in cases, but also in controls. Again, PTPN11 and NOTCH1 remained the most common genes (see Figure 1.2 B). 50 Figure 1.2: Potentially causal variants identified in known CHD genes. A: Pathogenic and likely pathogenic variants identified. B: Pathogenic, likely pathogenic variants as well as VUS (class ≥ 3.3) identified. 1 Application of a diagnostic framework for variant interpretation in CHD In Figures 1.3 to 1.8, details on six genes are depicted to illustrate the findings of this analysis. All of these genes are linked to conditions following an autosomal dominant inheritance pattern. The examples were selected to present genes that were consistent with the previously published phenotype spectrum (NKX2-5,TAB2), which had a high frequency of (likely) pathogenic variants (PTPN11,KMT2D) and those in which the observed phenotype was broader than thus far published (KMT2A,SMAD6). NKX2-5 codes for the transcription factor NK2 Homeobox 5. It is one of the first genes in which variants were linked to non-syndromic CHD. 68 All of the five identified carriers showed isolated septal defects, in one sample further in co-occurence with coarctation of the aorta (see Figure 1.3 and Table 1.10). Variants clustered in the Homeodomain and specifically at the protein position 178. The p.(Thr178Met) variant as well as the frameshift variant p.(Leu202fs) have previously been reported as pathogenic.68,112 In addition, a novel variant, p.(Thr178Met) in the Homeodomain was identified. Figure 1.3: (Likely) pathogenic variants identified in NKX2-5, colour-coded by the phenotypic maingroup. Table 1.10: Summary table of variant and carrier information for (likely) pathogenic variants identified in NKX2-5. Colours indicate the phenotypic maingroup. PhenotypeIdentifier Protein consequenceVariant (hg38)Sample Atrial septal defect|Aortic coarctation rs104893900; Clinvar:9004 p.(Thr178Met)5:173233011:G:AC02126 Secundum atrial septal defectnovelp.(Lys140Met)5:173233125:T:AC08116 Atrial septal defect rs104893900; Clinvar:9004 p.(Thr178Met)5:173233011:G:AC02127 Atrial septal defect rs104893900; Clinvar:9004 p.(Thr178Met)5:173233011:G:AC02128 Secundum atrial septal defect rs1554093461; 536136 p.(Leu202fs*49)5:173232937:CCA:CC08670 TAB2 codes for the TGF-Beta Activated Kinase 1 (MAP3K7) Binding Protein 2. The protein plays a role in mediating signalling events in multiple pathways. It forms a complex with TRAF6, 52 1.3 Results TAB1 and TAK1 in response to Il-1 and TNF α -signalling 113 This leads to activation of JNK and NF κ B 114 . Signalling involving TAB2 has also been implicated in ciliogenesis and cilia maintenance. 115 Therefore, TAB2 plays a role in regulation of cell survival or cell death in response to inflammatory signals. TAB2 is also implicated in astrogenesis during brain development through the formation of a complex with NCOR and the cleaved HER4-receptor. 116 Monoallelic loss of function variants in TAB2 have been associated with isolated CHD and also with cardiomyopathies, skeletal and connective tissue defects. 117,118 TAB2 also causes polyvalvular heart disease, characterized by cardiac (valvular) anomalies as well as short stature, facial dysmorphism and mild developmental delay. 117,118 In this cohort, TAB2 showed six carriers with truncating variants (see Figure 1.4 and Table 1.11). Four of them displayed facial dysmorphism, which is in line with the published phenotype spectrum (OMIM 614980). 101 Five of the six identified patients showed valvular phenotypes, affecting the aortic, tricuspid or mitral valve. Three of the identified variants (p.(Val514fs), p.(Gln238Ter) and p.(Gln325Ter)) have not been described before. Figure 1.4: (Likely) pathogenic variants identified in TAB2, colour-coded by the phenotypic maingroup. 53 1 Application of a diagnostic framework for variant interpretation in CHD Table 1.11: Summary table of variant and carrier information for (likely) pathogenic variants identified in TAB2. Colours indicate the phenotypic maingroup. PhenotypeIdentifier Protein consequence Variant (hg38)Sample Ventricular septal defect|Aortic valve stenosisnovelp.(Ser293Ter)6:149378793:C:GC00053 Aortic coarctation|Bicuspid aortic valvenovelp.(Val514fs*4)6:149379454:TG:TC08428 Hypertelorism|Ventricular septal defect|Frontal bossing rs1554263321; Clinvar:488615 p.(Arg441Ter)6:149379236:C:TC07103 Long philtrum|Anteverted nares|Rod-cone dystrophy|Muscular hypotonia|Joint hypermobility|Short stature|Tricuspid regurgitation|Disproportionate shortlimb short stature|Mild global developmental delay|Feeding difficulties novelp.(Gln238Ter)6:149378627:C:TC01801 Facial asymmetry|Strabismus|Joint hypermobility|Mitral valve prolapse|Constipation|Craniofacial asymmetry|Cognitive impairment novelp.(Gln325Ter)6:149378888:C:TC00210 Frontal bossing|Short stature|Tricuspid regurgitation|Depressed nasal bridge|Abnormality of digit Clinvar:1199210p.(Arg546Ter)6:149397636:C:TC00815 (Likely) pathogenic variants in PTPN11 (shown in Figure 1.5 and Table 1.12) were the most common finding in this cohort. This gene codes for the Tyrosine-protein phosphatase non-receptor type 11 (SHP2), an enzyme implicated in the intracellular mediation of different signalling pathways. 119–123 Its implication in CHD is well established and various pathogenic variants have already been reported. These reported variants generally display a gain-of-function mechanism resulting in increased phosphatase activity. SHP2 has an activating role in RAS/MAPK-signalling. 121 Other modulated signalling pathways include signalling by FGFR 124 , EGFR 125 or PDGFR 126 . Variants in PTPN11 have been implicated in multiple diseases including Leopard-Syndrome (OMIM 151100), Noonan-syndrome (OMIM 163950), juvenile myelomonocytic leukemia (OMIM 607785) and metachondromatosis (OMIM 156250). 122,127 Leopard-syndrome and Noonan-syndrome are subforms of the so called RASopathies. Cardiac defects, especially pulmonary stenosis are a common feature of these diseases, present in 50-80 % of affected. 128,129 In contrast to most of the other described genes, PTPN11 shows a constraint towards missense variants (o/e 0.52) and typically pathogenic variants have a missense consequence leading to a gain of function effect.127 In this cohort, 28 cases with (likely) pathogenic variants in PTPN11 were identified. Patients were found with both syndromic (n= 18; 64 %) as well as non-syndromic phenotypes (n= 10; 36 %). A high specificity for RVOTO and Septal defects was observed across all patients. The majority of the identified variants were previously reported as pathogenic. In Chapter Two of this work, dynamic molecular modelling was applied to the previously published p.(V518F) and the uncharacterised variant p.(M504V). 54 1.3 Results Figure 1.5: (Likely) pathogenic variants identified in PTPN11, colour-coded by the phenotypic maingroup. 55 1 Application of a diagnostic framework for variant interpretation in CHD Table 1.12: Summary table of variant and carrier information for (likely) pathogenic variants identified in PTPN11. Colours indicate the phenotypic maingroup. PhenotypeIdentifier Protein consequence Variant (hg38)Sample Cryptorchidism|High forehead|Low-set ears|Abnormality of the hip joint|Laryngomalacia|Abnormal mitral valve morphology|Pulmonic stenosis|Sparse scalp hair|Flat capital femoral epiphysis|Short stature|Lactose intolerance|Increased nuchal translucency|Perimembranous ventricular septal defect|Right aortic arch rs121918454; Clinvar:13325 p.(Ala72Gly)12:112450395:C:GC00321 Ventricular septal defect|Tricuspid regurgitation|Pulmonic stenosis|Short stature|Hydrocephalus rs121918455; Clinvar:13327 p.(Asn308Ser)12:112477720:A:GC02135 Pulmonary arterial abnormality|Calvarial skull defect rs397507501; Clinvar:40482 p.(Thr42Ala)12:112446385:A:GC08419 Pulmonic stenosis|Short stature|Dyslexia rs727503381; Clinvar:164998 p.(Asn200Tyr)12:112454636:A:TC00611 Pulmonic stenosis rs397507529; Clinvar:40525 p.(Ile282Val)12:112473031:A:GC07992 Epicanthus|Hypertelorism|Low-set ears|Anteverted nares|Brachydactyly|Pulmonic stenosis|Wide intermamillary distance rs397507529; Clinvar:40525 p.(Ile282Val)12:112473031:A:GC00345 Lop ear|Conductive hearing impairment|Delayed speech and language development|Pulmonic stenosis|Proportionate short stature|Microtia|Moderate global developmental delay rs397507529; Clinvar:40525 p.(Ile282Val)12:112473031:A:GC00790 Cryptorchidism|Epicanthus|Pulmonic stenosis|Short stature rs397507529; Clinvar:40525 p.(Ile282Val)12:112473031:A:GC06957 Strabismus|Seizures|Pulmonic stenosis rs397507529; Clinvar:40525 p.(Ile282Val)12:112473031:A:GC00805 Pulmonary valvar abnormality rs28933386; Clinvar:13326 p.(Asn308Asp)12:112477719:A:GC01805 Pulmonic stenosis rs28933386; Clinvar:13326 p.(Asn308Asp)12:112477719:A:GC08246 Thick lower lip vermilion|Retrognathia|Epicanthus|Lowset ears|Downslanted palpebral fissures|Delayed speech and language development|Pulmonic stenosis|Coarse hair|Short nose|Proportionate short stature rs28933386; Clinvar:13326 p.(Asn308Asp)12:112477719:A:GC00288 Hypertelorism|Low-set ears|Intellectual disability, mild|Plagiocephaly|Umbilical hernia|Pulmonic stenosis|Right ventricular hypertrophy|Frontal bossing|Widely patent fontanelles and sutures|Rhizomelic arm shortening|Unilateral ptosis|Progressive ptosis|Mild global developmental delay|Rhizomelic leg shortening rs28933386; Clinvar:13326 p.(Asn308Asp)12:112477719:A:GC00913 Pulmonic stenosis rs121918457; Clinvar:13331 p.(Thr468Met)12:112488466:C:TC08062 Pulmonic stenosis rs121918457; Clinvar:13331 p.(Thr468Met)12:112488466:C:TC08063 Pulmonic stenosis rs121918457; Clinvar:13331 p.(Thr468Met)12:112488466:C:TC08064 Cryptorchidism|Renal duplication|Microcephaly|Downslanted palpebral fissures|Keratoconus|Pectus carinatum|Pulmonic stenosis|Short stature rs397507539; Clinvar:40550 p.(Pro491Ser)12:112489047:C:TC07077 Pulmonic stenosis|Abnormal facial shape|Coronal craniosynostosis|Mild global developmental delay rs397507546; Clinvar:40561 p.(Gly503Glu)12:112489084:G:AC07070 Cryptorchidism|Epicanthus|Hypertelorism|Low-set, posteriorly rotated ears|Abnormality of the sternum|Pulmonic stenosis|Woolly hair|Clinodactyly of the 5th finger|Cognitive impairment rs397507547; Clinvar:40562 p.(Met504Val)12:112489086:A:GC07095 Right ventricular outlet obstruction rs1488010874; Clinvar:1775089 p.(Val518Phe)12:112489128:G:TC08284 56 1.3 Results Table 1.12 continued. Summary table of variant and carrier information for (likely) pathogenic variants identified in PTPN11. Colours indicate the phenotypic maingroup. PhenotypeIdentifier Protein consequence Variant (hg38)Sample Swiss cheese atrial septal defect|Hydrocephalus|Inguinal hernia rs121918466; Clinvar:13340 p.(Gln79Arg)12:112450416:A:GC02519 Secundum atrial septal defect|Talipes equinovarus rs397507518; Clinvar:40507 p.(Glu110Lys)12:112450508:G:AC08037 Atrioventricular septal defect rs397507526; Clinvar:40521 p.(Leu262Arg)12:112472972:T:GC08741 Thin upper lip vermilion|Smooth philtrum|Behavioral abnormality|Delayed speech and language development|Prominent fingertip pads|Intellectual disability, mild|Plagiocephaly|Ventricular septal defect|Atrial septal defect|Generalized seizures|Sleep disturbance|Abnormality of the cerebral cortex|Arnold- Chiari type I malformation|Prominent forehead rs376607329; Clinvar:40522 p.(Arg265Gln)12:112472981:G:AC00970 Joint hypermobility|Ventricular septal defect rs121918455; Clinvar:13327 p.(Asn308Ser)12:112477720:A:GC06927 Atrioventricular septal defect rs28933386; Clinvar:13326 p.(Asn308Asp)12:112477719:A:GC08907 Ventricular septal defect rs397507540; Clinvar:40552 p.(Pro491Leu)12:112489048:C:TC08035 Cryptorchidism|Progressive sensorineural hearing impairment|Hepatosplenomegaly|Ventricular septal defect|Prominent forehead rs397507547; Clinvar:40562 p.(Met504Val)12:112489086:A:GC06938 KMT2D is encoding for the histone methyltransferase MLL2. It is involved in chromatin regulation by methylating the K4 of histone H3. Amongst other functions, it mediates the regulation of target genes i.a. c-Myc in response to canonical WNT-signalling. 130 Additionally, it is a co-activator to the oestrogen receptor.131 The underlying mechanism is haploinsufficiency of KMT2D, both truncating as well as missense variants throughout the whole transcript have been reported. 132 Variants in KMT2D are associated to the Kabuki syndrome (OMIM 602113). This syndrome is a multi-organ disorder characterised by short stature, auditory and ophthalmologic symptoms, craniofacial dysmorphia and developmental delay. 133 Congenital heart defects are common amongst patients with Kabuki-Syndrome, with a prevalence of approx. 70 %.134 Eleven patients with (likely) pathogenic variants in KMT2D were identified (see Figure 1.6 and Table 1.13). The majority of findings represent truncating variants or splice variants. Furthermore, a recurrent missense variant in the C-terminal SET-domain (p.(H5459R)) was observed. This region shows a very high missense constraint suggesting that this variant might also confer a loss-of-function. The observed phenotypic spectrum was fairly heterogeneous, containing both syndromic and isolated cases as well as various forms of CHD. 57 1 Application of a diagnostic framework for variant interpretation in CHD garding patients with ductal defects, heterotaxia or unspecified forms of CHD ("other"), no meaningful clusters could be identified. A strong cluster consisting of 20 genes was observed for septal defects. As can be expected, the hierarchical clustering revealed high overall similarities between RVOTO and conotruncal defects. Still, the main clusters of associated genes differed substantially. Only KDR and FLT4 showed a strong relative contribution to the conotruncal-related and RVOTO-cluster. Since the phenotypic spectrum of both genes have been implicated in conotruncal phenotypes such as TOF 140,141 , these genes were also included in further analyses for the RVOTO-set. LVOTO interestingly showed a divergent pattern compared to RVOTO. Lastly, valvular phenotypes were observed in two distinct clusters. The first of those clusters largely overlapped with the genes mainly found in RVOTO, whereas the second partially overlapped with LVOTO-enriched genes, suggestive that there might be distinct clusters for the different heart valves. The yellow highlights in Figure 1.10 indicate which genes were considered as major constituents of the respective clusters. To provide an intrinsic validation of the observed results, the distribution of benign and likely benign variants in the same set of genes was analysed. As a validation step, all variants predicted to be benign or likely-benign in the same genes (n= 4749) were plotted in the order of the clusters determined trough the clustering of pathogenic variants (Figure 1.10 B). This negative control showed complete disparity of the observed clustering. 64 1.3 Results Relative frequency of variants per 1000 AA Relative frequency of variants per 1000 AA Figure 1.10: Heatmap clustering of (likely) pathogenic variants per phenotype maingroup. The upper plot shows the clustering of (likely) pathogenic variants and VUS (class ≥ 3.3) in the cohort. Yellow frames indicate clusters considered for the subsequent GO-enrichment analysis. The lower plot shows distribution of (likely) benign variants in the same genes. GO-enrichment Enrichment of GO-terms was evaluated per phenotype based on the gene clusters identified in the previous step. Enrichment was evaluated with respect to the biological function as well as contribution to Reactome pathways. 107,108 In Figures 1.11 and 1.12, the ten most enriched terms per phenotype group are displayed. Enrichment in biological processes as shown in Figure 1.11 suggested different mechanisms predominantly involved in the phenotypic groups. Left-ventricular phenotypes scored highest for processes involved in the regulation of histone-methylation. In contrast, right-ventricular phenotypes were mainly connected to VEGFR- and EGFR-signalling. Enrichment was only observed at a moderate magnitude. The conotruncal cluster was found to be enriched for multiple terms connected to cell differentiation and proliferation processes, especially endothelial cell differentiation. Valvular phenotypes did not display strong specificity. Septal defects showed most enrichment for the regulation of cardioblast differentiation, suggesting that genes involved in the aetiology of septal defects act early in the cardiac development. 65 1 Application of a diagnostic framework for variant interpretation in CHD Figure 1.11: Ten most enriched terms per main phenotype for biological processes. 66 1.3 Results Figure 1.12: Ten most enriched terms per main phenotype for Reactome pathways. As for enrichment regarding Reactome pathway terms, LVOTO and septal defects were not found to be strongly enriched for specific pathways (see Figure 1.12). Consistent with the biological processes, RVOTO again showed enrichment for VEGF-signalling as well as PTPN11-mediated signalling. Interestingly, RVOTO showed overlapping terms with valvular phenotypes. Both groups share terms involving members of the neurotrophic tyrosine receptor kinase (NTRK) family. Activation of these receptors result in activation of the MAPK-pathway and are involved in cellular differentiation processes. Taken together, these results show distinction of participating genes and pathways and underlying mechanisms that contribute to the development of defined CHD subpopulations. 67 1 Application of a diagnostic framework for variant interpretation in CHD 1.4 Discussion The knowledge of the molecular mechanisms involved in congenital heart defects is constantly expanding and major advances have been made in understanding the genetic basis of CHD. Still, novel gene-associations are likely yet to be discovered and even if a gene has been established as a cause of CHD, the functional and clinical implications of novel variants found in such a gene are often unclear. Accurate classification of likely pathogenic variants in known CHD genes allows to filter out cases whose phenotype is likely to be explained. This helps to prioritise further efforts on the analysis of unsolved cases and serves as a first preparative step to discover novel associations. Here, an ES cohort of CHD cases was analysed. Although this sample size is a good basis for detecting and interpreting variants when integrating information from population databases, the statistical power remains a limitation. In the first publication of this cohort, a 20-fold larger cohort size was estimated to be needed to discover the majority of dominant S-CHD-associated genes. 79 Of note, this estimation was calculated based on trio sequencing data. As trio sequencing is more informative compared to singletons, this number can be expected to be higher, when parental information is unavailable. Concerning extracardiac malformations, roughly two-thirds of the patients displayed isolated heart defects. In this group, the diagnostic yield is typically lower than for syndromic cases. 142,143 The analysed cohort was biased towards isolated CHDs, as some of the contributing clinical centres preferentially included non-syndromic cases. The assignment of non-syndromic versus syndromic cases was based on the phenotypes reported by the local submitter. It has to be highlighted, that subtle syndromic features might have remained undetected. Furthermore, a considerable fraction of the patients were recruited and phenotyped at a very young age. Syndromic features might develop later on and were not present at the time of recruitment. The examined cohort includes index-parent trios as well as a few families with multiple affected members. Typically related samples are removed from case-control analyses to avoid over-representation of familially transmitted variants. In contrast, this analysis was not restricted to unrelated samples. Integration of sequencing data from unaffected family members can give valuable insights regarding a variant ´ s potential relevancy e.g. by determining the inheritance from a healthy parent. Removal of related parents would lead to a loss of this information and might result in false-positive enrichment signals. The restriction to ultra rare variants ensures that one family can at maximum contribute twice to the enrichment. This minimizes a possible distortion of the signal through over-representation of a familial variants. Variants carried by multiple individuals were typically found in both cases and controls. Multiple ACMG-criteria such as BS2 or BS4 reflect the presence of the variant in multiple controls. In the subset of variant classes 3.3 or higher in genes dominantly linked to CHD, 45 variants were carried by more than one sample. 68 1.4 Discussion One aspect explored in the analysis was how the phenotypic outcome regarding cardiac subphenotypes can be linked to the underlying genetic aspects. In total, eight phenotypic main groups were defined to consolidate the broad variety of structural cardiac malformations. For conotruncal, valvular, LVOTO and RVOTO phenotypes, approximately similar sized groups were present. Compared to epidemiological data derived from previous studies, the analysed cohort reflects the general prevalence of sub-phenotypes.109,110 Given that the PAN study, which was used to estimate the prevalence of CHD subtypes was recruited through similar centres to the presented study, potential overlaps regarding samples were assessed. 109 Only 55 samples of the analysed cohort were also included in the PAN study, representing 1.4 % of the cases in this cohort. The phenotypic distribution of those 55 samples aligns with the overall distribution within the whole cohort. An over-representation of (non-syndromic) conotruncal malformations and LVOTO, especially HLHS and to a lesser extend valvular and RVOTO was observed. As these are conditions with more severe clinical implications and these patients are genetically often underdiagnosed compared to syndromic CHD, these groups were chosen as a focus of the analysis. As mentioned above, the recruitment of participants was slightly biased in comparison with the general population prevalence towards NS-CHD. Management of patients with HLHS is a focus at the University Hospital Schleswig-Holstein (UKSH) in Kiel, providing an explanation for the over-representation of this phenotype. Further aspects that influenced the composition of the cohort and therefore the outcome of the analysis consist of the inclusion criterium that patients were required to have had interventional measures during the first year of life to be eligible for the analysed cohort as well as exclusion of samples with a known structural variant. Mild septal defects, which account for a major part of CHDs do not necessarily require early interventional measures. These aspects have to be considered when interpreting the results. ES typically yields multiple ten-thousand of variants per sample, resulting in the need for variant prioritisation. 65 Variants that might be responsible for the observed phenotype in the patient were explored using the variant interpretation framework by the American College of Medical Genetics.72 As this workflow is primarily intended for variant detection in individual patients, adaptations had to be made to apply the workflow to a larger cohort with limited phenotypic and familial history information. One of the more considerable alterations applied to the ACMG criteria was the reduced weight attributed to the absence of a variant from control samples (criteria PS4, PM2). Regularly applying these criteria would require enrichment in case-control studies or a complete absence of the variant in population databases. Due to the incomplete penetrance observed in CHD, this was deemed inappropriately restrictive. Single healthy carriers can be found even for well-established pathogenic variants in CHD and were thus not considered as concluding evidence to exclude pathogenicity.43 As in-silico predictors to assess the severity of missense variants CADD, MPC and REVEL were applied. They were selected from the wide variety of available predictors as they were found to show the 69 1 Application of a diagnostic framework for variant interpretation in CHD best results based on a comparison using Clinvar variants (see also Chapter 2, Figure 2.5). Furthermore, they were integrated into the grading scheme proposed as part of the ACMG-adaptations. 76 The CADD score showed the most exhaustive annotation of variants in the dataset. MPC was additionally applied as it focusses on a different aspect, namely regional depletion of variants. A combination of these predictors can be expected to result in a good computational approximation of a variant ´ s effect. Further evidence for the variant assessment can be provided by segregation information in relatives of the index patient. A relation between de novo occurrence and enrichment has been established for truncating variants, especially for but not limited to syndromic CHD. 48,79,143 Lack of knowledge on the parental genotypes in the majority of cases is a limitation of the current dataset. Individual follow up may still allow for reclassification of VUSes. In addition to the segregation testing in the family, this might reveal additional phenotypes that were not present at the time, allowing for a better genotype-phenotype correlation and estimation of possible underlying syndromic conditions. While this was not performed as part of this analysis, the presented classification would suggest a set of variants and cases in which detailed follow-up could significantly help further elucidate the variant and upgrade its classification. The focus of the analysis was based on detecting variants that were found in well-established genes related to CHD. Unsurprisingly, a majority of 86 % amongst the identified variants were evaluated to be benign or likely benign, compared to only 2 % of (likely) pathogenic variants. In 8.9 % of the CHD samples, a potentially causative variant could be assigned. This figure is in line with previous reports of diagnostic rates of 5-10 % based on ES data. 7,11,39 These reports primarily emphasize severe, damaging variants that tend to occur de novo. Genes involved in chromatin modification represent are frequently found to be affected in CHD. 11,39 The genes that explain the largest fraction of cases in the presented cohort, such as PTPN11, NOTCH1, FLT4 or SMAD6 also displayed high enrichment in case-control studies or were prominently discussed in the cited publications. In summary, both the fraction of ES-positive cases and the distribution of the identified genes appear consistent with the overall published picture. By applying the sub-classification scheme introduced by Ellard et al. 2020, VUSes with a higher probability of pathogenicity could be isolated. 73 Although lacking evidence for a classification as likely pathogenic, these variants have at least some evidence in favour of a deleterious effect, e.g. based on low population frequencies or high in-silico prediction values. When including these variants of uncertain significance, the fraction of samples with a candidate variant could be increased to 14.5 %. These variants can be considered as a selection of variants which would be favourable for further follow-up. However, the inclusion of VUSes is also likely to increase the number of false-positive associations. The increase in the number of healthy carriers observed is a likely indication of this uncertainty. Still, these VUSes were found almost two-fold more often within cases versus controls. In addition, a considerable number of these variants in controls are found in genes that follow a recessive inheritance. These variants would thus represent carrierships at best, which is consistent 70 1.4 Discussion with an absence of cardiac symptoms. As the likelihood is much lower to detect compound-heterozygous or homozygous variants, most variants deemed causal were found for dominant genes. Still, (likely) pathogenic variants in genes following a recessive inheritance pattern were also detected in around 5 % of samples. A heterozygous carriership in a recessive gene alone is insufficient to cause a phenotype. However, it might suggest the presence of a second, undetected variant. ES has limitations regarding the detection of copy number variants, non-coding variants such as those in promoter or enhancer elements as well as variants that generate a cryptic splice site in non-coding regions. Such variants are therefore likely not captured in the current dataset. Samples in which a heterozygous deleterious variant was identified might be promising candidates for further follow-up. Only in two samples a likely compound-heterozygosity was identified based on the ES-data. The curation of the variant-set was focussed towards identification of monoallelic associations, especially regarding prefiltering for a population allele frequency of less than 0.1 %. For recessive inheritance patterns, a higher frequency could be allowed, typically by filtering for allele frequencies below 1 %. Relaxing this parameter might add additional candidates for homozygous and compound-heterozgous associations. When identifying multiple variants in the same gene, resolution of the haplotype in the affected sample would be required to determine, whether the variants are located in cis or trans. The phenotypic spectrum of genes associated to CHD is considerably heterogeneous. Nonsyndromic forms make up roughly 70 % of CHD cases, but CHD is also a common phenotype in a variety of syndromic conditions. 7 Similarly, associated genes are detected in isolated CHD but also when CHD is present as an optional feature of various syndromic conditions. As illustrated by the genes outlined herein, most genes display a broader spectrum of associations as established thus far. The data underlines that a strict distinction between non-syndromic and syndromic genes is usually not possible, as most variants are found in a variable phenotypic spectrum. The data especially in this cohort is primarily focused on the heart and thus potentially neglects a share of associated patients without CHD. In general, most of the analysed genes were found in both syndromic and non-syndromic carriers. Overall, a phenotypic overlap of reported phenotypes and the published spectrum was found as exemplified by NKX2-5 or PTPN11. Consistent with previous publications, the diagnostic yield for syndromic patients was higher than that of non-syndromic patients. 142,143 This was generally consistent across all analysed main- and subgroups. When stratifying the results based on phenotypic groups, differences were found. Generally, a low diagnostic yield for isolated cardiac heterotaxia was observed. This can partially be attributed to the low number of samples included displaying this phenotype. Furthermore, heterotaxia typically presents as a systemic phenotype in which the cardiac aspect is only one manifestation. The full genetic background of this type of disorder is not fully captured when focussing on CHD-related genes. 71 1 Application of a diagnostic framework for variant interpretation in CHD Overall, roughly one third of the variants herein linked to CHD were novel, providing over hundred pathogenic or likely pathogenic variants that extend the picture of genetic contribution to CHD. Six genes were discussed in further detail to illustrate consistencies and differences in comparison to the existing literature. These were selected to illustrate the concurrence with previously published phenotype spectra, but also show the novel findings generated in this analysis. NOTCH1 and PTPN11 were confirmed as key players representing the most common findings amongst the tested samples. Detailed workup of NOTCH1 is further presented of Chapter Four. PTPN11 was the most common finding in the cohort with a phenotypic focus for septal defects and RVOTO, but with both syndromic and non-syndromic carriers. The majority of variants has previously been reported as pathogenic based on independent cases. In contrast to other described genes, PTPN11 shows a constraint towards missense variants (o/e = 0.52) and missense variants are wellcharacterised as pathogenic. 127,129 Consistent with gain of function as the established mechanism of disease, only missense variants were observed in this gene. NKX2-5 was reviewed as a gene with an defined spectrum of isolated septal defects. This is in agreement with the established disease associations.68,112 TAB2 was shown to be affected only by stop-gain variants but not altering variants. It showed a interestingly high specificity towards syndromic valvular phenotypes with two thirds of samples displaying extracardiac features. A recent publication has highlighted the contribution of TAB2 variants to extracardiac connective tissue disorders. 144 TAB2 is furthermore relevant to the primary cilia during cardiogenesis, where it regulates the Transforming Growth Factor-Beta-Activated Kinase 1 (TAK1). 145 Defective primary cilia have been linked to mitral valve prolapse, bicuspid aortic valve and aortic stenosis. 146,147 A mitral valve prolapse was observed in one case, as well as two cases with aortic valve defects (see Table 1.11). This illustrates a likely mechanism underlying the identified variants, and might also extend to defects of the tricuspid valve, as two cases with this defect were also identified to carry variants in TAB2. KMT2D and KMT2A are genes involved in histone methylation processes. Both are normally associated solely to syndromic conditions involving CHD. 133,136 While this expectation was largely confirmed for variants in KMT2A, a considerable fraction of patients in KMT2D showed isolated CHD. This might imply that variation in KMT2D might extend to milder phenotypes with presence of CHD without pronounced extracardiac malformations. Alternatively, certain phenotypes might have been missed due to inaccurate phenotyping or manifestation at a later stage in life. Histone modification and chromatin remodelling are processes known to strongly contribute to CHD.39 SMAD6 previously associated to aortic valve disease, showed a highly heterogeneous spectrum that differed strongly from initial publications. Enrichment for LoF-variants in SMAD6 was previously published in a broader phenotypic context comprising TOF, HLHS, TGA and coarctation. 11 In that study, eight out of nine variants were reported as inherited. Given its involvement in BMP-signalling, which is a key signalling pathway during cardiogenesis, SMAD6 could here be highlighted as an important genetic contributor to CHD. 139 In summary, the analysis has confirmed previously made genotype-phenotype correlations for CHD-genes but also suggests divergent or extended phenotypic 72 1.4 Discussion associations in a subset of the analysed genes. Based on the distribution of variants identified, a clustering approach was applied to further correlate individual genes to phenotypic sub-spectrum of CHD. To account for a potential bias due to differences in the size of the gene product, the carrier number was normalised to the frequency per 1000 amino acids. Clusters were observed for five of the tested sub phenotypes. Interestingly, the involved genes did not extensively overlap between different subtypes. Especially LVOTO and RVOTO showed distinct enrichment. Variants that were classified as (likely) benign were used as a comparison. The absence of the clustering pattern build based on the pathogenic variants can serve as a control for the validity of the observations. The subsequent GO-enrichment for biological processes revealed a pronounced focus of LVOTO in histone modification. The key function of histone modifiers for cardiogenesis has been well studied and is reinforced by the high number of variants found in KMT2D and KMT2A. 48,148 For RVOTO, a moderate signal for VEGF-signalling was observed. VEGF-signalling has been implicated in epithelial-to-mesenchymal-transition at the outflowtract as part of the valve formation. 149 Conotruncal phenotypes mainly correlated to terms of cell growth and differentiation. While these are important processes throughout the whole cardiogenesis, correct formation of the outflow tract is heavily dependant on cell proliferation considering the need for septation, OFT-elongation and valve formation. The high ranking of multiple terms involving differentiation of endothelial and endocardial cells as well as the term "positive regulation of cardiac epithelial to mesenchymal transition" indicates a specific importance of EMT in the development of conotruncal defects. For septal defects, regulatory processes such as cardioblast differentiation and regulation of cardiomyocyte apoptosis were the strongest associations. The latter is in line with the intentional apoptotic processes involved in septation. Valve defects displayed two distinct clusters, partially overlapping with LVOTO and RVOTO respectively. No specific enrichment was evident for this group regarding GO-terms. Taken together with the division observed in the clustering heatmap, this would suggest that different mechanisms contribute to cardiac valve pathogenesis. Further refinement of these mechanisms could possibly be achieved with an increased sample size. In addition to biological processes, contributions of individual Reactome pathways were evaluated. The aspect of cell-cell signalling, which is essential in heart development and considerably contributes to CHD is also further expanded in Chapter Two. LVOTO and septal defects did not show particularly strong specificity, indicating that their aetiology might be rooted in a more complex crosstalk of multiple processes. For the other conditions, enrichment in specific and distinct processes was observed. Again, VEGF-signalling was found enriched for RVOTO. The highest score was observed for "PTPN11 activation by MET-signalling". MET-signalling is part of the broad groups of signalling pathways that involves membrane bound receptor tyrosine kinases. PTPN11 is one of multiple 73 2 Enrichment-testing of signalling pathways in CHD sub-phenotypes trabeculation, and thus is important for developing the ventricular chamber. 197 Endocardial activation of signalling by NOTCH and NRG1 also regulates the process of trabeculation. The endocardium expresses NRG1 as a member of the EGF family and interacts with HER2/HER4 receptors at the myocardium contact. 197 Implications of NOTCH-signalling and particularly NOTCH1 in CHD will be discussed further in the last Chapter of the Thesis. Further signalling mechanisms involved in cardiogenesis Cardiogenesis initially begins with the movement of mesodermal cells towards the cardiac crescent. As these cells migrate, the surrounding signalling environment is redefined. The effect of WNT-signalling, which previously blocked entering of the cardiac cell lineage, is succeeded by the activation of BMP signalling. 20 Migration is furthermore regulated by fibroblast growth factor (FGF)-signalling. 159 Containment of the cardiogenic region is mediated by BMP inhibition from neural tube cells and FGF from the endodermal cells. 159 The fibroblast growth factor receptors (FGFRs) are receptor tyrosine kinases (RTKs). RTKs constitute a broad group of single-pass membrane proteins which also comprise the epidermal growth factor receptors (EGFRs and HERs), the insulin and the insulin-like growth factor receptors (IR and IGFR), the platelet-derived growth factor receptors (PDGFRs) and the vascular endothelial growth factor receptors (VEGFRs). VEGFR-signalling is important for angiogenesis, but is also implicated in EMT at the outflow tract as part of the valve formation. 149,198 Numerous well-established genes involved in the development of CHD are part of or regulated by receptor tyrosine kinases. Generally, RTKs have a ligand binding domain on the extracellular side of the protein and a kinase domain on the intracellular side. Upon ligand binding, RTKs form dimers that phosphorylate each other, allowing for subsequent phosphorylation of other substrates that mediate the signal. A well-studied early marker for the cardiac progenitor cells is the expression of bHLH-type transcription factors MESP1 and MESP2. 199 MESP1 defines the specification of cardiovascular progenitor cells by altering the transcriptional profile, i.a. repressing NODAL expression and upregulating cardiac transcription genes NKX2-5 and GATA4. 200 Mediated through the WNT-antagonist DKK1, MESP1 furthermore downregulates canonical WNT-signalling thus promoting differentiation of early cardiomyocytes.201 Nuclear receptors are involved in sensing of steroid-derived hormones such as estrogens and retinoic acids, vitamins or fatty acids. 202 Since those ligands are lipid-based and can traverse the membrane, the receptors are nuclearly localised or can translocate to the nucleus upon ligand binding. Nuclear receptors share common functional domains and can bind to regulatory regions on the DNA, either in a monomeric or an oligomeric form. This mediates the signalling information through modification of target gene transcription.202 An additional crucial element in the development of the heart is the breakage of symmetry. Leftright patterning is mainly coordinated by the TGF β family members NODAL and LEFTY1/2 and the transcription factor PITX2 which acts downstream of Nodal-signalling.203,204 80 2.2 Material and methods Taken together, the different steps of heart formation are orchestrated by a complex interplay of different signals expressed in a cell-type and timepoint-dependent manner. Aim This chapter aims to investigate the individual implications of signalling cascades in congenital heart defects. The first aim was to define the involvement of particular signalling pathways to CHD by conducting a permutation-based enrichment testing approach. Within enriched pathways, individual contributors at gene-level were to be investigated. Finally, specific associations between CHD subphenotypes and distinct signalling pathways were to be addressed by a stratified analysis of the respective subgroups. 2.2 Material and methods The analyses in this chapter were performed on the dataset from the exome-sequenced (ES) cohort as outlined in Chapter One (see Section 1.2.1). 2.2.1 Reactome gene sets The Reactome database (https://reactome.org/) is a collection of open-access gene sets for metabolic and signalling components. 205 It covers signalling in humans and in several model organisms. Reactome also includes interactions involving non-protein molecules, such as nucleotides or small molecules. Entries are manually curated and reviewed by an expert panel. The data within Reactome is hierarchically structured, ultimately reaching down to binary reactions between single interaction partners. Top-level signalling pathways were retrieved as of May 2021. A total of 17 pathways were used, comprising 2541 distinct proteins. Non-protein interaction partners were discarded. A notable subset of those proteins was found to be involved in multiple signalling pathways (see Figure 2.4). 81 2 Enrichment-testing of signalling pathways in CHD sub-phenotypes Death Receptor Signalling Integrin Signalling Intracellular signalling by second messengers MAPK family signalling cascades MTOR signalling Signalling by Erythropoietin Signalling by GPCR Signalling by Hedgehog Signalling by Hippo Signalling by Leptin Signalling by Non-Receptor Tyrosine Kinases Signalling by NOTCH Signalling by Nuclear Receptors Signalling by Receptor Tyrosine Kinases Signalling by Rho GTPases, Miro GTPases and RHOBTB3 Signalling by TGFB family members Signalling by WNT Signalling by WNT 9 1 87 76 2 1 68 60 1 0 8 97 67 31 60 12 301 Signalling by TGFB family members 7 0 6 8 2 0 3 6 1 0 6 15 9 11 9 102 Signalling by Rho GTPases, Miro GTPases and RHOBTB3 52 6 32 43 2 6 75 2 4 0 8 33 48 84 668 Signalling by Receptor Tyrosine Kinases 21 14 109 110 5 19 56 9 2 7 33 22 60 493 Signalling by Nuclear Receptors 2 5 43 36 2 6 43 0 0 0 12 55 272 Signalling by NOTCH 15 1 67 63 1 0 6 55 1 0 9 217 Signalling by Non-Receptor Tyrosine Kinases 6 4 19 21 1 3 7 4 0 2 55 Signalling by Leptin 0 0 3 4 0 4 0 0 0 11 Signalling by Hippo 2 0 0 1 1 0 0 0 20 Signalling by Hedgehog 1 0 67 56 0 0 26 137 Signalling by GPCR 52 7 60 34 2 10 701 Signalling by Erythropoietin 2 3 9 11 0 25 MTOR signalling 0 1 18 3 41 MAPK family signalling cascades 8 18 141 326 Intracellular signalling by second messengers 15 4 326 Integrin Signalling 127 Death Receptor Signalling 144 Figure 2.4: Overlap between the tested pathways from Reactome database. Numbers indicate number of overlapping proteins in the pathways. Mean = 227 ± 215 genes. Darker colours indicate larger pathways. 2.2.2 ROC analysis The accuracy of the following eight in-silico algorithms was tested with a focus on variants in CHD genes. CADD The CADD-score (Combined Annotation Dependent Depletion) represents a commonly used metascore that integrates a total of 63 features, such as previous prediction scores (SpliceAI; MMSplice; GRANTHAM, SIFT, PolyPhen, GERP, phastCons, phyloP), DNase hypersensitivity sites, transcription factor binding sites, distance to exon-intron-boundaries and expression levels in studied cell lines. 90 It is commonly applied as a phred-score representing 10 x -log of the ranked score. CADD can be calculated for an extensive fraction of variants, whereas other scores were explicitly designed for specific subtypes, thus lacking annotation for some variants. REVEL The Rare Exome Variant Ensemble Learner (REVEL) score is a metascore, integrating multiple previous algorithms including MutPred, FATHMM, VEST, PolyPhen, SIFT, PROVEAN, MutationAssessor, MutationTaster, LRT, GERP, SiPhy, phyloP, and phastCons. 92 It was built using a random forest model based on neutral and disease-related missense variants retrieved from the Human Gene Mutation Database (HGMD). 82 2.2 Material and methods MPC The MPC score (Missense badness, PolyPhen-2, and Constraint) aims to identify local sub-genic regions depleted of missense variants indicating a potential functional relevance. The score was developed by comparing the observed frequency of rare (MAF < 0.1 %) missense variants within the ExAc dataset against the frequency that would be expected under the assumed absence of selective pressure.91 MVP The MVP (Missense Variant Pathogenicity prediction) score is a metascore that specifically discerns between high- and lowconstraint genes measured by the pLI value. 93 It integrates multiple information such as conservation, variant and protein structure elements as well as previous prediction scores such as Eigen, VEST, MutationTaster, PolyPhen2, SIFT, PROVEAN, fathmm-MKL, fathmm, MutationAssessor, and LRT. Other in-silico predictors DANN, MutationTaster, GERP++ and FATHMM were also tested but not further applied due to low performance in the ROC analysis.206–209 Apart from the CADD score, values were used as a rankscore, which reflects a normalisation to a 0 to 1 scale, with 1 indicating the most deleterious effect. The selection of suitable prediction algorithms and cut-offs was determined using ROC (receiver operating characteristic) analysis. The performance was assessed using the Roc()-function from the pRoc-package in R. A set of true positive variants with a pathogenic or likely pathogenic assessment (n= 20144) and a true negative variant set with likely benign or benign assessment (n= 31469) was defined. Variants in genes with a known association with CHD were retrieved from the Clinvar database as of July 2022 (see CHD Tier 1 and 2, Optional CHD Tier 1 and 2 in Table 1.8). Variants were only considered when they were concordantly assessed by multiple submitters without conflicts, reviewed by expert panels or were noted in guideline publications. Benign variants were set as control and pathogenic variants as case. Missing values were removed. For the accurate interpretation of variant severity, high specificity was deemed preferable to high sensitivity. A cut-off at a specificity of 0.90 was used to classify pathogenic variants. In addition, a cut-off at a specificity of 0.50 was also included, reflecting the point at which the fraction of benign variants surpasses those from the pathogenic set. This value was defined as an upper border for variants of (likely) benign consequences for further testing. 2.2.3 Pathway analysis The 17 signalling pathways, as defined based on Reactome, were tested for enrichment in the ES cohort. Statistical testing was performed using a one-sided Fisher ´ s exact test comparing the ratio of carriers vs. non-carriers in CHD patients or respective sub-groups against the control set. Carriership was defined as a sample having at least one variant in any component of the pathway fulfilling the applied filtering scenario. Carriership of multiple variants in the same pathway was therefore not considered multiple times. Only ultrarare variants (MAC ≤ 2) with an exonic consequence on the 83 2 Enrichment-testing of signalling pathways in CHD sub-phenotypes canonical transcript (based on Ensembl) were included in the analysis. The following six filtering scenarios were tested: 1. Ultrarare synonymous variants. Synonymous variants served as a negative scenario, in which no enrichment was to be expected. Synonymous variants with a predicted effect on splicing were excluded from this scenario. 2. Ultrarare PTVs. All truncating variants and variants with a predicted effect on splicing based on SpliceAI were included. 3. Ultrarare Tier2 PAVs.Variants were restricted to protein-altering variants and required to be predicted as deleterious by at least one of the four prediction algorithms (CADD, REVEL, MPC, MVP) based on the 0.90 specificity cut-off defined. 4. Ultrarare Tier1 PAVs. At least two in-silico algorithms had to predict the variant as pathogenic based on the 0.90 specificity cut-off defined. Variants with one or more benign predictions were removed. 5. Ultrarare PTVs + Tier2 PAVs. This scenario represents the combination of variants in the truncating as well as in the Tier2 PAV variant set. 6. Ultrarare PTVs + Tier1 PAVs. This scenario represented the union between truncating and high confidence PAVs. Permutations for each of the tested pathways were tested to reduce false positive signals in the study. To that end, 20000 random permutations were created that reflected the same composition as the original pathway. In detail, each permutation had the same number of genes as the original pathway, and the size of the protein coded by the permuted gene did not differ by more than ± 5 % to the original. Within one permutation, no duplicate genes or genes contained in the original pathway set were allowed. For the comparison of the permutations, the odds ratio (OR) of each permutation was compared to that of the original pathway. A significance threshold of 0.01 was applied, so that a scenario was considered significant if no more than 20000 x 0.01 (= 200) permutations showed a higher odds ratio than the original pathway. The p-value was adjusted using the false discovery rate correction (FDR) based on 918 tests, reflecting 17 tested pathways multiplied by 9 different phenotype groups and by 6 different filtering scenarios. A significance threshold of padj ≤0.01 was used. 2.2.4 Testing at gene-level An additional one-sided Fisher ´ s exact test was performed at the gene-level within pathways deemed significantly enriched. FDR correction was performed based on the number of genes within the tested pathway. Significance levels were defined as p adj ≤ 0.01, p adj ≤ 0.05 and p ≤ 0.01. All tests were performed using R. 84 2.3 Results The filter with the most significant p-value is used to visualise those genes identified in multiple filtering scenarios. Multiple resources have been included to provide further evidence and assess the relevancy of the findings. Conservation over the whole protein and for the specific amino acid positions that were found to be affected in the samples, inter-species conservation was retrieved using Aminode.210 Evaluation of expression during heart developmental stages Cardiac expression of the identified genes throughout and after embryogenesis was estimated using a published dataset by Cardoso-Moreira et al. 2019. 211 Reads per kilobase of transcript per Million mapped reads (RPKM) values in the cardiac tissue were used to quantify the gene expression. The developmental stage was defined between the fourth week post conception (wpc) until the eighth, the maturation phase between weeks nine to 38 and postnatal as all stages after birth. STRING-interaction To reveal potential interaction networks, a network was constructed using STRING. 212 As input, the genes identified as enriched in the pathway analysis were combined with known cardiac genes in the same pathways. Included interaction indicators were restricted to co-expression, experimental data and co-mentioning in abstracts. A high-confidence cut-off of ≥0.7 was applied to restrict the results to trustworthy interactions. The visualisation of the results was generated using Cytoscape.213 2.3 Results 2.3.1 Evaluation of in-silico predictors The performance of eight in-silico predictors and their suitable cut-offs were determined to be applied specifically to CHD-related variants. Eight commonly used algorithms were tested for their ability to discriminate between (likely) pathogenic and (likely) benign variants using a receiver operating characteristic (ROC) curve. A set of benign and pathogenic variants in CHD-related genes retrieved from Clinvar were used as a training dataset. To filter only for variants with high confidence for deleteriousness, a stringent cut-off at a specificity of ≥ 0.9 was selected. A score at a specificity of ≤ 0.5 was set as a threshold for benign variants. ROC analysis revealed the best discriminative abilities for REVEL_rs, MVP_rs, MPC_rs and CADD (see Figure 2.5). 85 2 Enrichment-testing of signalling pathways in CHD sub-phenotypes Figure 2.5: ROC-analysis performed over common in-silico prediction scores. Marked points refer to the used thresholds for pathogenic and benign variants. The numbers in the brackets correspond to the specificity and sensitivity at the respective position. 86 2.3 Results Table 2.1: Summary of ROC results for the four in-silico predictors used. ES = exome sequenced cohort Fraction of variants annotated in ES [%] Benign cut-off (≤ 0.5 specificity) Pathogenic cut-off (≥ 0.9 specificity)Score 59,6 ≤ 0.566≥ 0.851REVEL_rankscore 98,6 ≤ 22.05≥ 26.55CADD_phred 59,5 ≤ 0.502≥ 0.852MPC_rankscore 54,5 ≤ 0.610≥ 0.911MVP_rankscore Regarding the coverage of variants in the ES data, CADD_annotation was available for the majority of the variants. All other scores showed a similar range of approximately 55-60% of the variants, as these primarily are annotated for missense-variants only. Inframe insertions and deletions are generally annotated by CADD but lack anntoations from the other algorithms. Taken together, the combination of CADD, REVEL_rs, MPC_rs and MVP_rs with the cut-offs specified in Table 2.1 were applied in further analyses. 2.3.2 Enrichment testing for signalling pathways Seventeen signalling pathways were defined based on the Reactome database and tested for enrichment in the ES cohort. Testing was performed over all CHD cases along with individually for each of the eight defined phenotypic main groups. Furthermore, variants were filtered based on six different scenarios, resulting in 918 combinations. A correlation between the p-value and size of the tested pathway was observed in initial runs (see Figure 2.6) due to a high variability of signalling pathway size as shown in Figure 2.4. 87 2 Enrichment-testing of signalling pathways in CHD sub-phenotypes Figure 2.6: Size correlation of tested pathways. Each dot represents a tested pathway-phenotype combination. A permutation-based approach was conducted to account for this correlation of enrichment with pathway size. Per pathway, 20000 random permutations were created and used for correction. A total of 26 different combinations showed significant enrichment based on the defined criteria (see Figure 2.7). Synonymous variants, which served as a negative control did not reach significance in any combination. Heterotaxia, ductal phenotypes and “other” CHDs did not yield significant results in any scenario, likely due to these groups ´ relatively small sample size. Of note, different patterns were observed of the LVOTO and valvular cases on the one hand and RVOTO, septal and conotruncal phenotypes on the other hand, suggesting differences in their aetiologies. Generally, most significant enrichments were obtained for signalling by receptor tyrosine kinases, signalling by NOTCH and signalling by WNT. 88 2.3 Results Figure 2.7: Results of the permutation-based testing of signalling pathways. Each of the 17 signalling pathways was tested for enrichment in all CHD cases or individual sub-phenotypes. A scenario was considered significant, if the odds ratio of the actual pathway was higher than at least 200 of the permutations and the p-value was below 0.01 after correcting for the number of performed tests (n= 918). Significant scenarios are marked in blue and the number indicates the -log (10) of the adjusted p-value. 89 2 Enrichment-testing of signalling pathways in CHD sub-phenotypes Table 2.5: Carrier information for identified variants in JAK2. The applied testing scenario included PTVs and PAVs of Tier2. n.a.: not available JAK2 Alternative genetic finding Inheritance PhenotypeExisting variation Protein consequence VariantSample LZTR1: p.829S/X (VUS) n.a.Double outlet right ventriclers1299892808, COSV104701179 , COSV67600242 146 D/N9:5044488:G:AC07826 NAA10: Splicevariant (VUS) n.a.Secundum atrial septal defect-206 Y/*9:5054566:C:GC08193 - n.a.Control sampleCOSV67679851Splice_variant9:5070053:G:CC01496 - n.a.Tetralogy of FallotCOSV104701467773 P/R9:5080567:C:GC08487 - Inherited from mother Tetralogy of Fallotrs1337247004775 P/A9:5080572:C:GC01059 - n.a.Control samplers1337247004775 P/A9:5080572:C:GC01060 PTPN11: p.308N/D (P) n.a.Atrioventricular septal defect-799 R/G9:5080644:A:GC08907 - n.a.Bicuspid aortic valve-925 L/V9:5090457:C:GC01377 - n.a.Control sampleCM141662, COSV67602992 938 R/P9:5090497:G:CC05679 - n.a.Truncus arteriosusrs767683708958 S/C9:5090557:C:GC08413 - n.a.Transposition of great arteries & intact ventricular septum COSV675789921012 E/K9:5090886:G:AC09056 - n.a.Double outlet right ventricle-1092 D/Y9:5126429:G:TC08697 - n.a.Progressive microcephaly, Epicanthus, Hypertelorism, Abnormality of skin pigmentation, Ventricular septal defect, Recurrent respiratory infections, Sleep disturbance, Clinodactyly of the 2nd toe rs13678510651124 D/A9:5126763:A:CC00327 GATA4: p.280T/M (LP) n.a.Ventricular septal defect, Atrial septal defect, Subvalvular aortic stenosis, Abnormal atrioventricular conduction COSV676374751133 */W9:5126791:A:GC00762 96 2.3 Results Figure 2.10: Variants identified in JAK2. The top frame shows the distribution of variants found in the ES cohort at protein level. Functional domains shown are annotated based on UniProt. The middle frame indicates the frequency of missense variants at a given amino acid position based on gnomAD V3.1. The bottom frame represents the substitution score as generated with Aminode. High values represent lower conservation. Missing regions lack sufficient data on homologous proteins. Red lines in the middle and bottom frame indicate the position of the variants found in the ES cohort. PDGFRB PDGFRB codes for the Platelet-derived growth factor receptor beta, a tyrosine kinase receptor. A total of 16 CHD cases and three controls were identified in PDGFRB (see Table 2.6). Most patients displayed septal or conotruncal defects and did not show extracardiac phenotypes. Mainly, two clusters were observed, located at the first ATP binding site of the Protein kinase as well as at the end of the protein kinase (Figure 2.11). 97 2 Enrichment-testing of signalling pathways in CHD sub-phenotypes Table 2.6: Carrier information for identified variants in PDGFRB. The applied testing scenario included PTVs and PAVs of Tier2. n.a.: not available PDGFRB Alternative genetic finding Inheritance PhenotypeExisting variation Protein consequence VariantSample FBN2:p.539C/R (VUS) n.a.Double inlet left ventricleCOSV5580835477 Q/*5:150135690:G:AC07985 - n.a.Microcephaly, Camptodactyly of 2nd-5th fingers, Ventricular septal defect, Coarctation of aorta, Thrombocytopenia, Neutropenia, IgA deficiency, Dysplastic aortic valve, Fluctuating splenomegaly, Dysplastic corpus callosum, Progressive pes cavus, Bilateral sensorineural hearing impairment, Cognitive impairment rs374239531, COSV55809278 228 G/D5:150133957:C:TC00800 - n.a.Hypoplastic left heart syndrome -298 N/S5:150133627:T:CC08176 - n.a.Control samplers756150324507 R/C5:150129817:G:AC05283 - n.a.Control samplers752228799604 R/H5:150124828:C:TC05091 - n.a.Transposition of great arteries & intact ventricular septum -606 L/I5:150124823:G:TC08369 - n.a.Secundum atrial septal defectCOSV55811777607 G/S5:150124820:C:TC08283 - n.a.Hypoplastic left heart syndrome rs139554380, COSV55807667 618 T/M5:150124786:G:AC08268 - n.a.Transposition of great arteries & intact ventricular septum rs758280032, COSV55805130 650 S/L5:150124324:G:AC08405 - n.a.Secundum atrial septal defect-824 H/P5:150121003:T:GC08217 - n.a.Control samplers921756253, COSV55802364, COSV55802430 932 E/K5:150119471:C:TC04787 - n.a.Vesicoureteral reflux, Large for gestational age, Ventricular septal defect, Abnormality of esophagus morphology -936 I/T5:150118844:A:GC00490 - n.a.Double outlet right ventricle, Heterotaxy COSV55809727948 R/W5:150118809:G:AC08826 - n.a.Congenitally corrected transposition of great arteries rs1163477683974 D/V5:150117834:T:AC08900 n.a.Tricuspid atresiars1561984983, COSV53844961 980 S/N5:150117816:C:TC07581 TLL1:p.627G/S (VUS) n.a.Ventricular septal defectrs1383355454, COSV53837276 1030 P/R5:150117666:G:CC07981 - n.a.Discordant ventriculo-arterial connections rs1383355454, COSV53837276 1030 P/R5:150117666:G:CC08015 n.a.Double inlet left ventricleCOSV538453501031 E/K5:150117664:C:TC08079 n.a.Right ventricular outlet obstruction COSV538453501031 E/K5:150117664:C:TC08080 98 2.3 Results Figure 2.11: Variants identified in PDGFRB. The top frame shows the distribution of variants found in the ES cohort at protein level. Functional domains shown are annotated based on UniProt. The middle frame indicates the frequency of missense variants at a given amino acid position based on gnomAD V3.1. The bottom frame represents the substitution score as generated with Aminode. High values represent lower conservation. Missing regions lack sufficient data on homologous proteins. Red lines in the middle and bottom frame indicate the position of the variants found in the ES cohort. KAT2B KAT2B coding for the Lysine Acetyltransferase 2B represents another member of the HAT-enzyme family. Seven CHD cases and two control carriers were identified (see Table 2.7). Most variants identified are located at the N-terminus of the protein, where the N-acetyltransferase domain and the BROMO-domain are located (Figure 2.12). The BROMO-domain is necessary to recognise acetylated lysins. For all samples with variants in the N-acetyltransferase- or the BROMO-domain, no other pathogenic variants in established CHD genes were identified that might alternatively explain the phenotype. A case carrying the p.(A745T) variant is affected by an interrupted aortic arch and multiple ventricular septal defects in a syndromic context. All other cases did not show extracardiac malformations based on the available data. Two of the cases with variants in the BROMO-domain showed a ccTGA phenotype, one a DORV. The samples in the N-acetyltransferase-domain showed an ASD type II and an aortic coarctation with BAV, respectively. 99 2 Enrichment-testing of signalling pathways in CHD sub-phenotypes Table 2.7: Carrier information for identified variants in KAT2B. The applied testing scenario included PTVs and PAVs of Tier2. n.a.: not available KAT2B Alternative genetic finding Inheritance PhenotypeExisting variation Protein consequence VariantSample KMT2A: p.1292R/X (LP) n.a.Thin vermilion border, Hypertelorism, Long philtrum, Bulbous nose, Downslanted palpebral fissures, Long eyelashes, Sacral dimple, Brachydactyly, Global developmental delay, Tetralogy of Fallot, Scoliosis, Congenital nystagmus, Dilatation of the renal pelvis COSV55430474137 R/Q3:20072439:G:AC00371 - n.a.Control samplers1205273470, COSV55434569 528 R/Q3:20126074:G:AC05710 - n.a.Secundum atrial septal defect-619 K/R3:20137048:A:GC08608 - n.a.Aortic coarctation, Bicuspid aortic valve -638 K/R3:20140273:A:GC07801 - n.a.Male pseudohermaphroditism, Duplicated collecting system, Brachycephaly, Microcephaly, Pointed chin, Cupped ear, Anteverted nares, 11 pairs of ribs, Postaxial hand polydactyly, Growth delay, Umbilical hernia, Ventricular septal defect, Sparse scalp hair, Curly hair, Intestinal malrotation, Interrupted aortic arch -745 A/T3:20148415:G:AC00241 - n.a.Control sample-757 A/T3:20148451:G:AC05428 n.a.Congenitally corrected transposition of great arteries rs769126094776 R/C3:20152352:C:TC08143 - n.a.Double outlet right ventricle, Hypoplastic aortic arch -798 N/S3:20152419:A:GC07786 - n.a.Congenitally corrected transposition of great arteries -805 P/X3:20152439:CCT:CC08235 100 2.3 Results Figure 2.12: Variants identified in KATB2. The top frame shows the distribution of variants found in the ES cohort at protein level. Functional domains shown are annotated based on UniProt. The middle frame indicates the frequency of missense variants at a given amino acid position based on gnomAD V3.1. The bottom frame represents the substitution score as generated with Aminode. High values represent lower conservation. Missing regions lack sufficient data on homologous proteins. Red lines in the middle and bottom frame indicate the position of the variants found in the ES cohort. RBFOX2 RBFOX2 codes for the RNA Binding Fox-1 Homologue 2, a protein involved in RNA binding in the context of cell type specific alternative splicing. In this analysis, a total of eight patients were identified to harbour variants in RBFOX2 (see. Table 2.8). A stop gain variant p.(R447*) at the very end of the protein was recurrently identified in two patients, one with an isolated ASD type II, one with a syndromic VSD, microcephaly, abnormal facial shape and learning disability. Given the very late position in the protein, a functional consequence of this variant remains unclear. Further five missense variants were identified in RBFOX2. Of note, and contrary to the previous data available, these missense variants account for the majority of identified deleterious variants. Three of these cluster in the RNA-recognition motif domain, a region which is the interaction site, at which the protein binds to the RNA (Figure 2.13). 216 Most interestingly, two variants (p.(L181P) and p.(S187F)) fall very close to a RNA-Interaction site at position 188. Three other missense variants were identified further c-terminally and are not within currently known functional domains. The carriers of those variants are affected by AVSD, hypoplastic left heart syndrome (HLHS) and an unicuspid aortic valve respectively, which fits the tendency towards mostly left-sided phenotypes. The patient carrying the p.(V279I) variant also harbours a rare missense variant in LZTR1. For all other samples, no genetic alterations were identified in known CHD-related genes. 101 2 Enrichment-testing of signalling pathways in CHD sub-phenotypes Table 2.8: Carrier information for identified variants in RBFOX2. The applied testing scenario included PTVs and PAVs of Tier2. n.a.: not available RBFOX2 Alternative genetic finding Inheritance PhenotypeExisting variation Protein consequence VariantSample - n.a.Unicuspid aortic valvers756914441180 T/A22:35781671:T:CC08727 - n.a.Atrioventricular septal defectCOSV53265502, COSV53267222 181 P/L22:35781667:G:AC01316 - n.a.Hypoplastic left heart syndrome -187 S/F22:35781649:G:AC07782 LZTR1:p.692F/V (VUS) n.a.Trigonocephaly, Behavioral abnormality, Autistic behavior, Seizures, Global developmental delay, Specific learning disability, Bicuspid aortic valve, Short nose, Clinodactyly of the 5th finger -279 V/I22:35761451:C:TC00211 - n.a.Hypoplastic left heart syndrome -368 Y/C22:35759894:T:CC08268 - n.a.Hypogonadotrophic hypogonadism, Micropenis, Thick lower lip vermilion, Thick upper lip vermilion, Short philtrum, Anosmia, Abnormality of saccadic eye movements, Blepharophimosis, Oculomotor apraxia, Autistic behavior, Intellectual disability, mild, Agenesis of corpus callosum, Cerebellar vermis hypoplasia, Pes planus, Constipation, Abnormality of the cerebral cortex, Highly arched eyebrow, Mild short stature, Clinodactyly of the 5th finger, Bilateral cryptorchidism, Feeding difficulties in infancy, Enuresis nocturna, Incisor macrodontia, Moderate global developmental delay, Perimembranous ventricular septal defect, Right bundle branch block, Shortening of all phalanges of fingers -392 A/V22:35746536:G:AC00742 - n.a.Microcephaly, Specific learning disability, Ventricular septal defect, Abnormal facial shape -447 R/*22:35744222:G:AC00065 - n.a.Secundum atrial septal defect-447 R/*22:35744222:G:AC08493 102 2.3 Results Figure 2.13: Variants identified in RBFOX2. The top frame shows the distribution of variants found in the ES cohort at protein level. Functional domains shown are annotated based on UniProt. The middle frame indicates the frequency of missense variants at a given amino acid position based on gnomAD V3.1. The bottom frame represents the substitution score as generated with Aminode. High values represent lower conservation. Missing regions lack sufficient data on homologous proteins. Red lines in the middle and bottom frame indicate the position of the variants found in the ES cohort. PIK3CA PIK3CA codes for the catalytic subunit of the phosphoinositid-3-kinase heterodimer. This phosphorylates phosphatidylinositol (4,5)-bisphosphate (PIP2) to phosphatidylinositol (3,4,5)- trisphosphate (PIP3) on the membrane, to which downstream kinases, including Akt, bind for signal transduction.217 As shown in Table 2.9, multiple missense variants in PIK3CA were identified. Most variants are located within the N-terminal adaptor-binding domain (ABD) and the catalytic domain (Figure 2.14). While the first is important for the interaction with the regulatory subunit, the latter region is essential to the enzymatic capabilities. Interestingly, recurrent missense variant were found at two positions. The p.(H59P) in the adaptor binding domain is carried by two individuals. One case is affected by an isolated truncus arteriosus, the other case by an AVSD. P.(H994R) is also shared by two affected individuals from the same family. Both show a coarctation of the aorta. Most of the displayed variants fall in regions of high evolutionary conservation and low carrier frequencies in the main population. 103 2 Enrichment-testing of signalling pathways in CHD sub-phenotypes Table 2.9: Carrier information for identified variants in PIK3CA. The applied testing scenario included PAVs of Tier 2, n.a.: not available. PIK3CA Alternative genetic finding Inheritance PhenotypeExisting variation Protein consequence VariantSample - n.a.Truncus arteriosusrs148062665459 H/P3:179199001:A:CC08824 - n.a.Atrioventricular septal defectrs148062665459 H/P3:179199001:A:CC09021 - n.a.Control sample-78 E/G3:179199058:A:GC01917 - n.a.Patent ductus arteriosusrs886042002, CM162636, COSV55873442, COSV55926654, COSV55949642 108 R/H3:179199148:G:AC08123 - n.a.Control sample-133 D/Y3:179199734:G:TC05143 - n.a.Hypoplastic right heart-431 D/G3:179210226:A:GC01246 - n.a.Double inlet left ventricle-474 E/K3:179210446:G:AC09034 - n.a.Right ventricular outlet obstruction -557 Y/H3:179219200:T:CC08112 - n.a.Hypospadias, Micrognathia, Macrotia, Ventricular septal defect, Secundum atrial septal defect, Double outlet right ventricle, Hypoplasia of right ventricle, Broad thumb, Congenital microcephaly, Interrupted aortic arch type B -913 I/T3:179230075:T:CC00543 - n.a.Inguinal hernia, Strabismus, Autism, Laryngomalacia, Ventricular septal defect, Atrial septal defect, Gastroesophageal reflux, Laryngeal cleft, Echogenic fetal bowel, Mild global developmental delay, Patent ductus arteriosus after birth at term rs773300933, COSV55886677 926 D/N3:179230113:G:AC06870 - n.a.Aortic coarctationrs373295359994 H/R3:179234138:A:GC08004 RAF1:p.483K/T (LP) n.a.Aortic coarctation, Bicuspid aortic valve rs373295359994 H/R3:179234138:A:GC08005 104 2.3 Results Figure 2.14: Variants identified in PIK3CA. The top frame shows the distribution of variants found in the ES cohort at the protein level. Functional domains shown are annotated based on UniProt. The middle frame indicates the frequency of missense variants at a given amino acid position based on gnomAD V3.1. The bottom frame represents the substitution score as generated with Aminode. High values represent lower conservation. Missing regions lack sufficient data on homologous proteins. Red lines in the middle and bottom frame indicate the position of the variants found in the ES cohort. 2.3.4 Expression of identified genes during heart developmental stages Expression of the identified genes during developmental, maturation and postnatal stages of cardiogenesis was evaluated using a dataset published by Cardoso-Moreira et al. 2019. 211 Results are depicted in Figure 2.15. LAMC3,NRG2,FLT4 and ATP6V1B1 showed overall low cardiac expression. A large fraction of genes showed their main expression during early timepoints of the developmental stage and decreased over time. Other genes such as JAK2,TAB2,KDR and PIK3CA peaked during maturation. 105 2 Enrichment-testing of signalling pathways in CHD sub-phenotypes development. Kat2b-knockdown experiments in zebrafish showed perturbation of these developmental processes. 246 Indeed, two of the KAT2B-variant carriers in this cohort displayed a syndromic phenotype encompassing skeletal symptoms, a brachy- or polydactyly (see Table 2.7; cases C00371 and C00241). Furthermore, an interaction of the N-terminal part of PCAF and the intracellular NOTCH1-fragment has been shown, resulting in enhancement of target gene transcription.247,248 These interactions provide indications of how KAT2B might contribute to CHD. The suspected association is further supported by the fact that PCAF also functions in histone modification and chromatin remodelling, which are particularly relevant for CHD. Aside from an isolated study, which reported enrichment of two SNPs within KAT2B in a small Chinese cohort 249 , an association to CHD at the human level has thus far not been made. KAT2B thus represents an attractive novel candidate identified in this analysis. RBFOX2 has previously been found enriched in patients with left-sided congenital heart defects, especially for HLHS, mainly attributed to truncating variants. 143 Only one missense variant, p.(K220R) was reported, presenting with a complex cardiac malformation including a hypoplastic left ventricle and aortic arch, a DORV, an AVSD and unspecified extracardiac malfomations. 11 In contrast to these publications, severe missense variants were found to be the primary variant type affecting RBFOX2, indicating that the underlying mechanism of disease might also extend to missense variants. In particular, multiple variants were found within the RNA-binding domain of RBFOX2, suggesting this site as a potential hotspot. RBFOX2-dependant splicing patterns have been shown in multiple experiments, but also especially during the epithelial to mesenchymal transition (EMT). 250,251 Since this represents an important step of cardiogenesis, alterations in RBFOX2 activity are likely to result in CHD. Knock-out of rbfox2 in mice has been shown to recapitulate the HLHS phenotype and revealed disorganised interactions between cells and the extracellular matrix.252 The phenotypic spectrum found the identified cases of this cohort is mainly composed of septal defects and left-sided phenotypes, which is consistent with published observations for the truncating variants in patients and the available animal models. Multiple cases showed extracardiac symptoms, including skeletal defects and neurological abnormalities. Neurodevelopmental delay was reported for some of the published cases, in general, extracardiac aspects were not specifically addressed in the study.143 KDR coding for the Vascular Endothelial Growth Factor Receptor 2 (VEGFR2), has been identified with a high burden of loss of function variants in patients with conotruncal phenotypes. In fact, all but one of the identified patients showed a TOF phenotype, which represents a uncommonly specific phenotypic pattern. The findings of KDR variants in this cohort have recently been published in a joint publication. 141 KDR is involved in signalling by the vascular endothelial growth factor (VEGF). Ligand binding to the VEGF receptor is intracellularly mediated through several cascades including RAS/MAPK and PI3K/AKT. This signalling is essential for the regulation of endothelial growth during 112 2.4 Discussion angiogenesis and vasculogenesis as well as for formation and maturation of the valves in the outflow tract. 149,198 This might explain the observed link to TOF, whose leading manifestation is the pulmonary valve dysplasia. The ATP6V1B1-gene codes for a subunit of an intracellular (vesicular) ATPase, which transport protons against a gradient therefore regulating (sub-)cellular acidification. 253 In particular, this gene is predominantly expressed in the kidney. Recessive variants in this gene are associated with distal renal tubular acidosis and sensorineural deafness. 254 Cardiac phenotypes have thus far not been described. Given the low cardiac expression during embryogenesis and the absence of previous reports in the context of congenital (cardiac) malformations, it remains to be explored, if and how variants in this gene might be linked to CHD. However, a recent publication has reported twins with a heterozygous, de novo inframe deletion in ATP6AP1, another gene coding for a accessory subunit of the same V- APTase complex. 255 Both individuals showed a congenital atrial septal defect, and literature research from the authors revealed additional cases also presenting with cardiac phenotypes. LAMB1 and LAMC3 are two members of the laminin protein family. Laminins are glycoproteins present in the extracellular matrix. A number of laminins are encoded in the human genome which are combined in different heterotrimers. 256 Laminins play a role in cell adhesion and migration processes. Both LAMB1 and LAMC3 are linked to MET-signalling. MET is a receptor tyrosine kinase activated by binding to its ligand, the Hepatocyte growth factor (HGF). Signaling by MET contributes to cell growth, cell survival and motility during embryonic development and to tissue homeostasis. 257 Recessive variants in these genes are associated with lisencephaly (OMIM:615191). Cardiac malformations have thus far not been reported. Interestingly, LAMB1 translation was shown to be under regulation of TGF β -signalling mediated by the PDGF-receptor. 258 Signalling via a PDGF-receptor (beta) is also a mechanism suggested by the results of this analysis. However, neither of these genes have thus far been linked to CHD and evidence remains inconclusive. Neuregulin 2 encoded by NRG2 is one of four neuregulins, a family of growth factors implicated in the development of the nervous system and the heart. 259 Like its paralog NRG1,NRG2 has been implicated in glioma, which is one of the most common tumour types of the nervous system, where it contributes to increased migration of the tumour cells. 260 Neuregulins signal via HER tyrosine kinase receptors and regulate cell growth and differentiation. NRG2 binds explicitly to HER3 and HER4 and signalling is intracellularly transduced via PI3K and MAPK cascades.259 So far, variants in NRG2 have not been reported in the context of CHD. However, mouse models targeting its receptors have shown a decrease in mesenchymal cell numbers and density in AV cushions linked to a deficiency of ErbB3 and ErbB4, which encode HER3 and HER4, as well as the absence of ventricular trabeculation due to deficiency of ErbB4. 261,262 Cardiac expression of NRG2 was shown to be focussed in the atrium and in lower levels in the ventricle and outflow 113 2 Enrichment-testing of signalling pathways in CHD sub-phenotypes tract. 259 Homozygous deficiency of ErbB3 leads to intrauterine lethality and the fetus displays a small right ventricle. 262 In this analysis, NRG2 was found enriched in RVOTO patients. It could thus be hypothesised that impaired NRG2-signalling might affect the formation of the right ventricle and outflowtract in a HER-dependent manner. In the majority of genes identified, the combination of ultrarare truncating and severe altering variants yielded the strongest enrichment. Contrary to that observation, PTPN11,NRG2,PIK3CA and SMAD6 showed enrichment exclusively for altering variants. As truncating variants tend to result in a loss-of-function mechanism, the absence of PTVs points to a gain-of-function mechanism resulting from the missense variants in these four genes. Indeed, this concurs with previous publications as exemplified for PTPN11 and somatic variants found in PIK3CA.127,263 Interestingly, no additional pathogenic variant in a known CHD gene was observed in most of the identified cases. While this is undoubtedly not sufficient to rule out other undetected contributors, it agrees with a suspected role of the signalling pathway components. Replication of these results in a larger cohort would strengthen the observations and potentially allow to further sub-define the relevant processes within the identified pathways. The cardiac-specific expression pattern, retrieved from an external dataset, was included to interpret differences throughout the timeline of heart formation. While the majority of candidate genes were found to be well expressed in the heart, LAMC3,NRG2 and ATP6V1B1 showed overall little expression throughout all time points. Of note, low expression levels do not necessarily relate to absent relevancy for cardiogenesis, as exemplified by the low expression apparent for FLT4 despite its role as a well characterised CHD gene. The expression data were furthermore used to assess, at which time-points of the heart development the genes were mainly involved. Many genes, such as VANGL2,RBFOX2 or SMARCA4 were found to be expressed most strongly in early time-points and decrease throughout time, reflecting their importance in the early processes of cardiogenesis. Other genes, such as TAB2,JAK2 or KDR peaked during the maturation phase, suggesting a stronger involvement in later occurring processes such as valve and outflow tract formation. However, a clear correlation with the mainly observed phenotype-group was not apparent for those genes. In a complementary analysis, interactions of the identified candidates with established genes involved in CHD were explored. Overall, two distinct, larger subnetworks could be identified, one containing genes of the RAS/MAPK-pathway and a second one that contains NOTCH- and BMP- signalling as well as histone-modifier genes. Of note, as the Reactome database represents a default input for STRING interactomes, the parameters were restricted to experimental and co-expression data. Higher confidence thresholds were applied to restrict to the most credible interactions. This, however, explains the lack of interactions observed even for well-known CHD-genes. 114 2.4 Discussion Despite the used stringency, the analysis was able to uncover multiple interactions between the candidates and other established CHD genes. PIK3CA has been shown to interact with members of the RAS-family, including KRAS and HRAS. 264 Association between JAK2 and SHP2, encoded by PTPN11, was reported as critical for the functional capacity of JAK2 during mediation of angiotensin II signalling. 265,266 Activation of SHP2 by the activated receptor PDGFRB occurs in the context of receptor tyrosine kinase signalling. 126,267 Interaction between the bromodomain region of SMARCA4 and the histone deacetylase CHD4 has been shown to suppress trans-activation capabilities of the N-terminal region of CHD4 strongly. 268 KAT2B mediates actetylation of SMAD3 thus interacting with TGF β -signalling. 269 High expression of KAT2B correlated with strong activity of TGF β -signalling. It is furthermore linked to the histone deacetylases EP300 and CREBBP, which both showed enrichment in the current analysis but also have been extensively described before.270,271 These interactions further reinforce the importance of the presented genes. The analysis presents suggestions, of how these novel genes are embedded in the existing cardiac regulatory networks and through which interactions their effect towards CHD might occur. In summary, this analysis illustrates the landscape of variants signalling pathways in CHD. Subphenotype-specific enrichment in signalling was revealed both at a pathway and gene-specific level. While some genes were only found in the collective of all CHD samples, several were found to contribute to specific subtypes. The latter were most frequently attributed to right-sided and conotruncal malformations as compared to left-sided lesions, suggesting a more defined set of contributors. Multiple novel or incompletely established candidate genes could be linked to the occurrence of heart defects and previously published associations could be reaffirmed. These results help to facilitate the diagnostic of genetic contributions in CHD and might eventually contribute to the expansion of therapeutic options. 115 3 Variant characterisation in PTPN11 and PIK3CA using molecular dynamics 3.1 Introduction The analysis of a genetic variant is commonly based on the level of DNA-, mRNA-sequences or primary structure protein data. The location of the sequence change within the three dimensional structure in which the protein occurs in its native environment however is normally not considered. This poses the risk of neglecting important information that can contribute to a more accurate variant classification and a better understanding of the functional effects. As an example, potential proximity to other variants, which are located at remote positions in the primary amino acid sequence, but cluster together as consequence of the folding might be overlooked. Testing the functional effect of individual variants is often costly, potentially demanding a considerable effort regarding resources and time. The ability to estimate potential functional consequences based on structural data is therefore of high value. In this chapter, the method of molecular dynamics was employed to characterise variants suspected to contribute to CHD. In vivo, a protein is surrounded by water, ions and various small molecules. All proteins thus constantly interact with other molecules, and some are furthermore constitutive part of an oligomeric protein complex. At a temperature around 37 °C, there are considerable atomic movements, which results in a dynamic structural flexibility in living cells. Structural data is mainly generated by experimental techniques such as cryo electron microscopy, X-ray crystallography, NMR or mass spectrometry. 272,273 These techniques provide essential information on the structure and function on a molecular level, which can be further complemented with information on the dynamic behaviour of the protein in its native surrounding generated by molecular dynamic simulations. The conditions under which the measurements are carried out are to a certain extent artificial. Depending on the technique used, they require large amounts of isolated, purified protein, unphysiological concentrations of ions and potentially ultra low temperatures. Molecular dynamics (MD) is a technique to simulate the dynamic behaviour of a system - in this case a protein of interest including its surrounding solvation - by stepwise iteration of atomic movements over time. MD is a highly resource intensive process where the computation time is dependant on both the size of the system to be modelled and the duration of the simulation. Thus, MD simulations are usually performed at a timescale of nanoseconds, but can also extend to longer time-frames. As some protein movements such as enzymatic catalysis, ligand binding, or allosteric regulations occur at larger timescales of micro- to milliseconds, not all biologically relevant processes can be fully 117 3 Variant characterisation in PTPN11 and PIK3CA using molecular dynamics captured with typical MD simulations. 274 The limited timescale currently poses an important limitation to the technique. The underlying basis of MD is rooted in classical Newtonian mechanics. 275 A position in threedimensional space, a direction and a velocity defines each atom. Interatomic, covalent interactions are approximated by harmonic potentials. The summarisation of interactions between the individual atoms of a model is expressed by a forcefield, that accounts for bonds, angles, torsions and non-covalent interactions such as electrostatic forces and van-der-Waals interactions. Different force field models such as AMBER, CHARMM, and GROMOS are in use.276–278 MD can be performed under defined thermodynamic environments. 279 These differ by the fixation of specific characteristics, such as the volume of the simulation box, the total energy, the temperature or the pressure within the system. The microcanonical ensemble (NVE) describes a fully isolated system which is not exchanging volume, energy or atoms with its environment. The canonical ensemble (NVT) is characterised by constant volume, number of atoms and temperature of the system. This is realised using an external thermostat such as a Langevin thermostat. In an isothermal-isobaric ensemble (NPT), both temperature and pressure are kept constant through a thermostat and barostat, so that the system’s volume becomes variable. The starting point of molecular dynamics is a protein structure usually generated by structural biology. Solvation is initially carried out by randomly placing of solvent (e.g. water molecules) in a surrounding box as well as ions to counter potential charges of the protein of interest. The temperature of the system is realised by random assignment of velocities to the atoms following the Maxwell-Bolzmann-distribution for the desired temperature.279 Next, the energy and acting forces in the system are calculated following the force field model. MD is commonly carried out in incremental timesteps of 1 to 2 fs. For each timestep, the positions are updated accordingly and the calculation of the force field is adapted to the new setting of the system. After the solvation and ionisation of a system, an initial equilibration phase is usually conducted to allow the system to adapt to the changes introduced by the addition of solvent and the application of a temperature. Once a system is equilibrated, a production simulation is carried out to assess the behaviour of the simulated structure, typically for a period of ≥ 100 ns. The atomic movement within the system is documented in frequent intervals, which can be used to evaluate the (differential) behaviour of the structures. The root-mean-square deviation (RMSD) and root mean square fluctuation (RMSF) values are meaningful characteristics to quantify the behaviour of a model. 280,281 The RMSD represents the relative position of the atoms in the structure compared to a reference, e.g. the initial protein conformation. RMSD allows tracking of the relaxation of the protein structure during the equilibration step. While an increasing RMSD reflects a movement away from the initial state, the flattening of the RMSD indicates that the structure has reached an equilibrated conformation. RMSF is a measure of how strong a subset of the structure, e.g. an individual side-chain, moves over time. It corresponds to the time average of the RMSD. As demonstrated herein, the RMSF can 118 3.1 Introduction be applied to determine more flexible regions within the structure compared to the overall movement as calculated by the RMSD.281 PTPN11 PTPN11 codes for a non-receptor tyrosine phosphatase (SHP2) involved in the intracellular mediation of various signalling pathways. This gene was already addressed in Chapter One (see Section 1.3.4). In its inactive configuration, the protein ´ s catalytic function is auto-inhibited through a blocking interaction between the N-SH2 and PTP domains. Once the N-SH2 domain binds to a phosphotyrosine peptide, conformational changes result in the active state. Most pathogenic variants reported are located within or around the interacting surfaces of the N-SH2 and PTP domains. 219 In confirmation of the findings in Chapter One, PTPN11 and interacting genes are suggested to be one of the core clusters in CHD. PIK3CA Frequently, signals generated by receptor tyrosine kinases are intracellularly mediated by the Phosphoinositide 3-kinase (PI3K). PIK3CA codes for the catalytic subunit of the Phosphoinositide 3-kinase (p110α). Together with a regulatory subunit, p110αforms a heterodimer. p110 α contains five functionally different domains. These comprise the N-terminal adaptor-binding domain (ABD) that binds to p85, the Ras-binding domain (RBD), the putative membrane-binding domain (C2), the helical domain of unknown function, and the kinase catalytic domain. AKT/PI3K-signalling is involved in processes such as proliferation, angiogenesis, survival and metabolism. 282 PIK3CA is needed for insulin-mediated blood glucose regulation as well as signal transduction of insulin-like growth factor 1. PI3K-signalling contributes to the regulation of cardiomyocyte size, cell survival and angiogenesis and the pathway has been linked to cardiac hypertrophy. 283 Transgenic mice with a cardiac-specific overexpression or knock out of PI3K revealed a correlation to the heart size. 284 While over-activation of PI3K-signalling resulted in enlarged heart, loss of function had the opposite effect. Involvement in structural congenital heart defects is so far not commonly reported for PIK3CA- related conditions. However, in a study on PIK3CA-related overgrowth disorders by Mirzaa et al., cardiac phenotypes were reported in 19 individuals. 285 The mainly observed phenotypes were atrial and ventricular septal defects, but also included arrhythmias, PDA, PFO, cardiomyopathies and Tetralogy of Fallot. Interestingly, the authors confirmed ten of the identified 60 variants to be present in a constitutive fashion.285 Seven of these ten overlap with cases that exhibit cardiac symptoms. For the patients of the cohort analysed in this thesis, it cannot be definitively confirmed, if the variants are indeed present in a constitutive status. The sequencing data was generated from peripheral blood samples. Given accessibility, validating the variant in a different tissue such as saliva could be used to resolve this uncertainty. The age of identified patients is not known for all carriers of PIK3CA. However, at least four carriers were recruited as newborns and an additional three samples during early childhood. As the rate of acquired somatic variants in blood increases with age, this 119 3 Variant characterisation in PTPN11 and PIK3CA using molecular dynamics raises the likelihood of a constitutional variant state in these cases. AKT/PI3K-signalling is a common mechanism regulating the cell cycle and cellular proliferation. While PI3K, a heterodimer formed including PIK3CA, is at the beginning of this intracellular cascade, MTOR, which was also found as part of the pathway analysis, is implicated in this signalling. Unfortunately, the majority of samples that harboured PIK3CA-variants in this cohort were isolated single index patients. Subsequently, no segregation data could be retrieved. PIK3CA represents a well-known oncogene. Somatic variation in PIK3CA is reported in various types of cancer and results in a gain of function. Two hotspots, p.(E545K) (rs104886003) and the closely co-located p.(E542K) (rs121913273) as well as p.(H1047R) (rs121913279) are most frequently observed in PIKCA-related disorders and account for approximately 80 % of PIK3CA- related cancers. 286 P.(E545K) and p.(E542K) are located in the helical domain of p110 α and lead to a reduction of the inhibitory capacity of the regulatory subunit. 287,288 p.(H1047R) is located at the C-terminus of the protein in the kinase domain. This variant leads to a constitutive activation of PI3K activity. 288 Their positions identified in this analysis do not overlap with the main mutational hotspots outlined above. Apart from its involvement in cancer, variants in PIK3CA lead to a spectrum of disorders including PROS (phosphoinositide-3 kinase-related overgroth spectrum), PIK3CA-related vascular malformations and PIK3CA-related nonvascular lesions. 282 In these conditions, variants in PIK3CA are primarily found in postzygotic mosaicisms. Dependant on which tissue is carrying the variant, different subforms of PROS can be present which result in asymmetric tissue overgrowth. Important subforms of this disease spectrum are the MCAP (megalencephaly-capillary malformation) syndrome and CLOVES (congenital lipomatous asymmetric overgrowth of the trunk, lymphatic, capillary, venous, and combined-type vascular malformations, epidermal nevi, skeletal and spinal anomalies) syndrome. Lipomatous overgrowth, intellectual disability as well as hypoglycemia, hypothyroidism or growth hormone deficiency may be present in affected patients. Again, the hotspot variants, p.(E545K) (rs104886003) and the closely co-localised p.(E542K) (rs121913273) as well as p.(H1047R) (rs121913279) are frequently observed. p.(E545K) and p.(E542K) are located in the helical domain of p110 α and lead to a reduction of the inhibitory capacity of the regulatory subunit. 287,288 p.(H1047R) is located at the C-terminus of the protein in the kinase domain. This variant leads a constitutive activation of p110αkinase activity.288 Aim In this chapter, molecular dynamics simulations were applied to characterise variants in PTPN11 and PIK3CA, which were the identified through the analysis in the previous chapter. Using this approach, structural effects should be elucidated to provide a better classification of their pathogenicity. A second aim was to provide insights into potential functional mechanisms underlying these variants, especially in comparison to previously known variants. 120 3.2 Material and methods 3.2 Material and methods SHP2, encoded by PTPN11 and PI3K, whose catalytic subunit is encoded by PIK3CA were subjected to a molecular dynamic simulation approach. Out of the genes identified in the pathway analysis in Chapter 2, these genes were selected primarily based on availability of cystallographic data, which could be used as an initial template. MD simulations were carried out using VMD Version 1.9.3 and NAMD Version 2.13.280,289 All simulations were performed at 310 K using a CHARMM-based force field model. 290,291 Periodic boundary conditions were adopted in all MD simulations. The particle-mesh-Ewald summation method was used to determine the long-range electrostatic forces. The van der Waals forces were calculated using a smooth cut-off of 12 Å with a switching distance of ten Å. For temperature control, a thermostat using the Nosé-Hoover algorithm was applied.292,293 Model preparation Initial PDB structures were generated using Alphafold V2. 294 For the two structures to be analysed, the amino acid sequences were retrieved from Uniprot (SHP2: Q06124, p110 α : P42336). Experimental data for both proteins were retrieved from the PDB database. 295 For PTPN11, 2SHP served as a reference structure (PDB-ID: 2SHP). 296 For PIK3CA, a crystal structure containing the full protein of p110 α in complex with a partial structure of p85 α is published (PDB-ID: 7MYN) and was used as a template. 297 After alignment of the Alphafold-generated structure of p110 α to the corresponding structure in 7MYN, p85 α was added to the PDB file of p110 α to obtain the heterodimeric structure. To verify the accuracy of the PDB files generated by Alphafold, alignment to the reference crystal structure was done and visually inspected. Due to low confidence in the prediction, the N-terminal fraction of SHP2 was removed after amino acid 541. Protonation states for the side chains of Histidine, Aspartic acid and Glutamic acid residues were initially predicted by Alphafold and were manually verified. A disulfide bond between Cys 333 and Cys 367 of SHP2 was patched using psfgen. Histidines 53,169 and 419 were ε -protonated. For p110 α , disulfide bonds between Cys 24 and Cys 36 , Cys 695 and Cys 692 as well as between Cys 905 and Cys 984 were introduced using psfgen. p85 α was protonated at position Asp421, Glu507 and Glu515. His450 was ε-protonated. Generation of an initial structure file, solvation and ionisation were carried out using the Tk-console plugin of VMD. The corresponding scripts were provided by the Quantum Biology and Computational Physics Group and adapted as required. The total charge of the system was verified to be balanced by summarising the charges of all atoms. The size of the simulation box was calculated based on the dimensions of the protein using an excess of at least 6 Å. The structures were put in a water box with a NaCl concentration of 0.15 M resulting in a total of 156403 atoms for the SHP2-model and 287532 atoms for the p110α-p85αmodel. Equilibration Equilibration of the system was carried out under constant temperature (310 K) and pressure (one bar) using a Langevin thermostat and Langevin-piston barostat respectively. 298,299 121 3 Variant characterisation in PTPN11 and PIK3CA using molecular dynamics 180° H59P R108H E78G D133Y D431G E474K H1047R – COSMIC-hotspot E545K – COSMIC-hotspot I913T D926N Y557H H994R Domains: █ ABD █RBD █C2 PI3K type █PIK helical █G loop/Activation loop/Catalytic loop █ Catalytic site Figure 3.7: Overview of the modelled variants in PIK3CA. All variants identified in the pathway analysis were modelled into the PIK3CA structure as well as the two established hotspot variants p.(H1047R) and p.(E545K). 128 3.3 Results A published crystal structure of the p110 α and p85 α heterodimer was used as a template to generate a full protein model of p110 α using Alphafold to provide predictions of the incompletely crystallised loop regions (PDB-ID: 7MYN; see Figure 3.8). The p85 α subunit was retained from the 7MYN file. Figure 3.8: Overlay of Alphafold generated structure and initial crystal sturcture 7MYN. Green: p110αgener- ated by Alphafold; Darkblue: p110αbased on 7MYN; Cyan: p85αbased on 7MYN Initial solvatation, ionisation and equilibration of the structure were performed. After equilibration of the structure was achieved as indicated by a stable RMSD, variants were introduced. Each production simulation was performed for 100 ns and repeated three times. The RMSD of all structures suggested an overall stability of the systems (see Figure 3.9). 129 3 Variant characterisation in PTPN11 and PIK3CA using molecular dynamics Figure 3.9: RMSD of the modelling simulations for p110 α . Each graph shows the RMSD for an individual replicate for the wild-type or mutant protein. Darkblue, green and cyan display the individual replicates. 130 3.3 Results Figure 3.9 continuation of the RMSD of the individual modelling simulations for p110α. Interpretation of the dynamic behaviour of the models was performed based on the RMSF value. The raw RMSF value for the wild-type showed high flexibility for regions between the functional domains as well as the C-terminal end of p110 α and the N-terminal start of p85 α (see Figure 3.10). While the catalytic site overall showed small RMSF-values, two peaks at around residue 880 and 970 were observed. Those residues are located at the outer regions of the catalytic site. 131 3 Variant characterisation in PTPN11 and PIK3CA using molecular dynamics Figure 3.10: RMSF of the wild-type p110 α -protein. Shown is a representative RMSF distribution for the widltype protein. Functional domains are shown by the coloured boxes. Black: averages RMSF over three replicates, gray: standard deviation of RMSF. 132 Figure 3.11: Relative RMSF values for the tested variants in p110 α . The line indicates the average of three replicas. Shaded area indicates the average of three replicas ± 1 standard deviation. The red dot shows the position of the modelled variants. Blue lines indicate variant identified in CHD cases in this study, green lines indicate variants identified in controls. Red indicates somatic hotspot variants. 3 Variant characterisation in PTPN11 and PIK3CA using molecular dynamics Figure 3.11 continuation. Relative RMSF values for the tested variants in p110α. 134 3.3 Results In order to compare the RMSF of the variants to that of the wild-type, a normalisation of the value to that of the wild-type was performed (see Figure 3.11). When comparing those normalised values, differences were specifically observed in the catalytic site of p110 α . While p.(E545K) showed an increased RMSF briefly after the G-loop at around residue 800, p.(H1047R) showed an increase of RMSF in the activation loop. The tested novel variants substantially recapitulated those patterns. Particularly strong effects were observed for p.(H59P), p.(D431G) and p.(I913T). The two missense variants p.(E78G) and p.(D133Y) identified in healthy carriers showed similar patterns to the wild-type. The two COSMIC-variants were used to identify regions which deviated most from the wild-type with respect to the RMSF value. Indeed, multiple regions of interest were identified (see Figures 3.13, 3.12). Interestingly, p.(E545) and p.(H1047R) showed distinct patterns for the majority of residues. p.(H1047R) was mostly characterised by an increased RMSF in residues 941 to 948 (Figure 3.14). p.(E545K) was found to have an increased RMSF at residues 784 to 797 and a decreased RMSF at residues 507 to 514 (Figures 3.15, 3.16)). While the latter is not part of a defined functional domain, it showed close proximity to the regulatory subunit. Residues 941 - 948 Residues 784 - 797 Residues 507 - 514 Figure 3.12: Regions with strong deviations of RMSF compared to the p110 α wild-type. Displayed is the wild-type PI3K model. Transparent blue: p85 regulatory subunit. Ribbon structure: p110 α subunit. Coloured regions were defined based on differences found in Figure 3.13. Displayed is a smoothened trajectory over the course of the modelling (blue = beginning of simulation, red = end of simulation). 135 3 Variant characterisation in PTPN11 and PIK3CA using molecular dynamics Figure 3.13: Heatmap of relative RMSF values for the tested molecules, averaged over three replicates each. Values are normalised compared to the average of the wild-type (RMSF = 1). Red arrows indicate regions, in which the external hotspot variants (darkblue on the x-axis) deviated most from the wild-type. Green variants on the x-axis are variants found in controls, light blue those found in CHD-cases. The colouring on the y-axis indicates the location of functional domains. 136 3.3 Results wildtype p.(H1047R) p.(H59P) p.(E78G) p.(R108H) p.(D133Y) p.(D431G) p.(E474K) p.(Y557H) p.(I913T) p.(D926N) p.(H994R) p.(E545K) Figure 3.14: Comparative trajectories of residues 941 to 948 of p110 α in the wild-type and mutated structures. 137 4 Phenotypic variability in patients with NOTCH1-variants has furthermore been implicated in the regulation of the epidermal 329 , alveolar 330 , immunologic 331 and pancreatic 332 development as well as for osteoblastic differentiation 333 . Notch-signalling is a central regulator during somitogenesis and coordinates the expression of Nodal, thus governing the establishment of the left-right axis formation.334,335 Under hypoxic conditions, HIF1A can bind to ICN, delay protein degradation and thus prolong signalling. 336 NOTCH-mediated hypoxia responses are an important mechanism for developmental processes such as angiogenesis and EMT, but also for the growth of malignant tumours.336–338 Clinical aspects and associations to NOTCH1-signalling Due to its essential role in different processes during embryogenesis, alterations in NOTCH1 are linked to multiple clinical conditions. NOTCH1-associated conditions generally include a cardiac involvement, although it is also implicated in non-cardiac phenotypes. Based on OMIM, NOTCH1 is autosomal dominantly associated with the aortic valve disease type 1 (OMIM 109730) as well as the Adams-Oliver syndrome type 5 (AOS; OMIM 616028). 226,339 Aortic valve disease type 1 manifests by the presence of a bicuspid aortic valve, calcification or stenosis of the aortic valve. Rarely, it can also be accompanied by other cardiac phenotypes such as a VSD or a Mitral valve stenosis. Furthermore, patients with this condition may develop aneurysms of the ascending aorta.226,340 Adams-Oliver syndrome is mainly characterised by terminal transverse limb defects, aplasia cutis congenita, and various forms of CHD. 339 This syndrome shows an incomplete penetrance. The frequency of CHD amongst AOS is estimated at 23 %. 339 Furthermore, cutis marmorata telangiectatica congenita and neurologic findings can be present. Besides NOTCH1, the Adams-Oliver syndrome can occur due to defects in ARHGAP31,DLL4 and RBPJ in an autosomal dominant manner and DOCK6 and EOGT in an autosomal recessive manner. The risk for cardiac involvement is increased for NOTCH1,DOCK6,DLL4 and EOGT-associated AOS compared to ARHGAP31 and RBPJ.339 While the cardiac spectrum of NOTCH1 variants is considered to be rooted in loss of function variants, direct or indirect over-activation of NOTCH1 occurs in the tumour-related context. Gain-of function variants in NOTCH1 are frequently found in T-cell Acute Lymphoblastic Leukemia (T-ALL). 341 Other cancers which have been related to altered NOTCH1 activity include head and neck squamous cell carcinomas342, breast cancer343, lung cancer344 or pancreatic cancer.345 Especially the cardiac spectrum of NOTCH1 is evidently broader than the two associations mentioned above. As illustrated by an evaluation of previous publications concerning NOTCH1 at the beginning of the result section in this chapter, alterations have been identified in cases with a broad spectrum of various CHD phenotypes (see Section 4.2.2). It is still unresolved, which factors contribute to this variety of outcomes. In this Chapter, the genotypic and phenotypic spectrum of NOTCH1 was analysed to better understand, which types of variants contribute to NOTCH1-related CHD and what distinctions can be made regarding the different patients ´ groups. However, coding variants alone appear insufficient to fully explain the observed spectrum alone, suggesting additional modifiers 144 4.1 Introduction within or outside of NOTCH1. Protein structure of NOTCH1 NOTCH1 is a large protein of 272 kDA comprised of 2555 amino acids. It is a single-pass membrane receptor, thus consisting of an extracellular, transmembrane and an intracellular region containing different functional domains (see Figure 4.1). The extracellular region of the protein primarily consists of 36 EGF-like domains. These domains have a high sequence similarity with the epidermal growth factor (EGF) and their three-dimensional structure is organised by disulfide-bridges. The ligand-binding motif, which is essential for the recognition of the ligands is located within the 11th and 12th EGF-like repeat.346 The remainder of the extracellular region contains the three Lin12/Notch repeats (LNR) and the heterodimerisation domain (HD). These regions of NOTCH1 are most commonly found to be altered in T-cell acute lymphoblastic leukaemia lymphoma (T-ALL). 347 Both the LNR and the HD are considered to negatively regulate the NOTCH1 activity by maintaining the receptor in its resting conformation prior to ligand binding. 347 Alterations in the HD domain can lead to ligand-independent release of the ICN fragment.348 The PSEN1-interaction site is located directly before the transmembrane region and is required for processing by the γ -secretase complex, an oligomeric complex capable of intramembrane cleavage.187 This step releases the intracellular NOTCH1-fragment (ICN). Figure 4.1: Structural domains currently described within the NOTCH1 protein. EGF = Epidermal growth factor, LNR = Lin12/Notch repeats, HD = heterodimerisation domain, RAM = RBP-J κ -associated module, ANK = Ankyrin domain, TAD = transcriptional activation domain, PEST = PEST domain.The black bar in the middle indicates the location of the cell membrane. At the intracellular site, NOTCH1 contains the RBP-J κ -associated module (RAM). This domain mediates the interaction between the ICN and the CSL (CBF1, Suppressor of Hairless, Lag-1) family of transcription factors. 349 The assembly of the transcriptional complex is furthermore coordinated by the TAD domain (transcriptional activation domain) located further towards the C-terminus. 350 Additionally, the six Ankyrin domains (ANK) are required for NOTCH-mediated signal transduction. 351 Two locations within the ANK domains interact with HIF1AN (Hypoxia-inducible factor 1-alpha inhibitor), allowing cross talk between cellular responses to hypoxia and NOTCH-signalling.336 The C-terminal PEST domain, named after the sequence motif (proline (P), glutamic acid (E), serine (S) and threonine (T)) is a reported hotspot region for activating variants in aggressive phenotypes in various human cancers, including triple-negative breast cancer.352 145 4 Phenotypic variability in patients with NOTCH1-variants Post-translational modifications furthermore modulate the functionality of the protein. Especially residues within the extracellular EGF-like repeat regions undergo O-glycosylation via POFUT1 or EOGT, followed by further glycosylations by glycosyl-transferases such as B4GALT1 or members of the Fringe family. 353 The glycosylation pattern of NOTCH1 shows a modulating effect on its ligand binding.353–355 CDK8-mediated phosphorylation of ICN in the PEST domain leads to ubiquitin E3 ligase-mediated protein degradation and therefore inactivation of the signalling. 356 This process can be further regulated by the kinase DYRK2.357 Ligand specificity in NOTCH1-Signalling NOTCH1 like all members of the NOTCH family receives signals in a juxtacrine fashion from adjacent cells. The receptor is then proteolytically processed to transduce the signal into the regulation of target genes. NOTCH-signalling may be activated by the binding of five potential ligands, DLL1, DLL3, DLL4, JAG1 and JAG2. It has been demonstrated that distinct binding preferences exist for the different NOTCH receptors and that distinct activation patterns arise from binding to these ligands.355,358,359 Regulation of NOTCH-signalling by cell-specific DLL1 expression is important for boundary formation during somite segmentation. 360 NOTCH effects on angiogenesis are modulated by a competition between the pro-angiogenic JAG1 and the repressing DLL4. 358 This is further refined through an extensive crosstalk between NOTCH- and VEGF-signalling. Vegf induces the Dll4 expression, leading to Dll4-mediated Notch-signalling at the early stage. 361 Dll1-mediated Notch-signalling in turn up-regulates the Vegfr2, suggesting that the Dll1-Notch axis contributes to the responsiveness of arterial cells to Vegf.359,361 In-vivo experiments demonstrate that Dll4-Notch1 interaction occurs in a constant fashion and triggers the expression of HEY1 and HEYL, while Dll1-Notch1 interactions in contrast occur in time-specific pulses and induce specifically HES1. 362 These processes were suggested for distinct outcomes in modulating cell fate during myogenesis.362 During cardiac valve formation, Dll4-Notch1-signalling induces epithelial-mesenchymal transition, leading to cushion formation. This is followed by Jag1-Notch1-signalling, which restrains Bmpmediated valve mesenchyme proliferation by sustaining Hbegf-EGF receptor signalling.363 Aim As NOTCH1 was found to be the most frequently affected gene in CHD, this Chapter is intended to provide a deeper analysis of the genetic and phenotypic landscape of variants found in NOTCH1-related cases. The approaches used included statistical enrichment testing, phenotypic follow up and segregation testing in affected families as well as a functional evaluation of identified variants. Further aspects that might account for the inherently heterogeneous spectrum are to be addressed by the investigation of non-coding variants and digenic interactions. 146 4.2 Material and methods 4.2 Material and methods 4.2.1 Material Table 4.1: Antibiotics used. Final concentration usedReferenceSupplierName 100 µg/mlA9393-5gSigma-AldrichAmpicillin 50 µg/mlK1377-5gSigma-AldrichKanamycin sulfate 75 µg/mlS9007Sigma-AldrichSpectinomycin dihydrochloride pentahydrate Table 4.2: Antibodies used. Used DilutionHost speciesReferenceSupplierName 1:2500MouseAB18184Abcamα-His6x 1:10000MouseMA5-15739ThermoFisher Scientific β-Actin loading control monoclonal antibody 1:15000Mouse26183ThermoFisher Scientific α-HA Tag Monoclonal Antibody (2-2.2.14) 1:20000MouseMA1-91878InvitrogenDYKDDDDK Tag Monoclonal Antibody (FG4R) (α-FLAG) 1:10000Goat115-035-003DianovaPeroxidase-conjugated AffiniPure Goat Anti-MouseIgG (H+L) Table 4.3: Cells used. ReferenceSupplierName ATCC CRL-3216™Kindly provided by AG Schade; Pharmazeutisches Institut, CAU Kiel HEK293T C404003Invitrogen™One Shot™ TOP10 chemically competent E. coli C862003Invitrogen™One Shot™ Mach1™ T1-phage- resistant, chemically competent E. coli 147 4 Phenotypic variability in patients with NOTCH1-variants Table 4.4: Buffers used. RecepieUseBuffer 16 g ad 500 ml ddH2O, autoclaveBacteria cultivationLB Agar 10 g ad 500 ml ddH2O, autoclaveBacteria cultivationLB liquid medium 450 ml DMEM, high glucose, GlutaMAX™ Supplement 50 ml FCS, heat inactivated Cell cultivationDMEM +/- 500 ml 10x TAE Ad 5 l ddH2O Running buffer agarose-gels TAE 50 mM TRIS base 150 mM NaCl 2 mM EDTA 1 % IGEPAL 0.1 % SDS 0.5 % Dodecylsulfate, Na-salt freshly supplied with cOmplete™ protease inhibitor cocktail Cell lysisRIPA-buffer 1 tablet PBS ad 500 ml ddH2O 500 µl Tween20 Mebrane washingPBST 5 g milk-powder ad 100ml PBST Membrane blocking; Antibody dilution 5% MMP 50 ml 20X Bolt™ MOPS SDS Running Buffer (LifeTech) ad 1 l H2O Running buffer SDS- gels MOPS running buffer 48 mM Tris Base 39 mM Glycine 10 % (v/v) EtOH Proteintransfer with Semidry system Bjerrum Schafer- Nielsen Buffer 25 mM Tris Base 192 mM Glycine 10 % (v/v) methanol 0.03 % SDS Proteintransfer with Tankblot system Towbin Buffer 7.5 g Glycin 0.5 g SDS 5 ml TWEEN20 10 % (v/v) EtOH ad 500 ml H2O adjust to pH 2.2 with HCl Membrane strippingStripping buffer 148 4.2 Material and methods Table 4.5: Material used. ReferenceSupplierName R1131thermo scientific3M Sodium Acetate solution R-03021-D50advansta AdvanStain Ponceau Protein blot Staining Solution AM9937 Thermo Fisher Scientific (invitrogen) AmbionTM Nuclease-free water 106000002GE HealthcareAmershamTM ProtronTM 0.45µm NC 4390939 applied biosystems (via. Thermo fisher) AmpliTaq Gold Fast PCR Master Mix X89900.2BodeBacillol AF 840004BiozymBiozym LE Agarose NW04125BOXinvitrogenBolt®4-12% BIS-TRIS Plus B0007Novex by life technologiesBoltTM LDS Sample Buffer 3030-153WhatmanChromatography paper 11836170001Sigma-Aldrich cOmpleteTM Mini, EDTA-free Protease Inhibitor R0141thermo scientificdATP A3672,0100PanReac AppliChemDimethyl Sulfoxid for cell culture 646563-10X.5MLSigma-AldrichDL-Dithiothreitol-solution 61965-026Gibco (via Life Technologies) DMEM, high glucose, GlutaMAX Supplement 20765ServaDodecylsulfate, Na-salt 14190-094Gibco (via Life Technologies) DPBS 1x (Dulbecco´s Phosphate buffered saline) 03690-100MLSigma-Aldrich Ethylendiaminetetraacetic acid solution, pH8.0, 0.5M in H2O EF0654thermo scientific FastAP Thermosensitive Alkaline Phosphatase F9665Sigma-AldrichFetal Calf Serum (FCS) heat inactivated G5516-100MLSigma-AldrichGlycerol A1067,0500PanReac AppliChemGlycine for molecular biology I8896-50mlSigma-AldrichIGEPAL®CA-630 IEVH00005MerckImmobilon-E Transfer membrane 22700025invitrogenLB Agar 12780052invitrogenLB Broth Base #BCBS1154VSigma-AldrichNonidet P 40 Substitute 31985-062Gibco (via Life Technologies)OPTI-MEM®I #26616thermo scientificPageRuler Prestained Protein Ladder 18912-014gibco PBS tablets 18912Gibco (via Life Technologies)PBS w/o Tween Tablets 149 4 Phenotypic variability in patients with NOTCH1-variants Table 4.5 continuation. Material used. ReferenceSupplierName 12368010invitrogenPlatinum SuperFi II PCR MasterMix N0468SNew England BiolabsQuickLoad 1kB DNA Ladder 15544034InvitrogenS.O.C. Medium 6331.1RothHCl 25 % S5150-1lSigma-AldrichSodium chloride solution, 5M 30435-1kgSigma-AldrichSodium phosphate dibasic dihydrate 26625thermo scientific Spectra Multicolor High Range Protein Ladder 10481220001Roche Diagnostics GmbHT4 DNA Ligase N3271SNew England BiolabsTriDye 100bp DNA-Ladder 648311-1kgMerckTris Base, ULTROL®Grade T8787-50MLSigma-AldrichTritonTM X-100 #R0531thermo scientificTurbofect Transfection reagent P1379-250mlSigma-AldrichTWEEN 20 T143RothWater, nuclease free, DEPC-treated Equipment: 58816-121VWRAnalog Vortex mixer 1851196 BIORADC100 Touch Thermal cycler 5427000015EppendorfCentriuge 5430 5100-0001NALGENECyro 1 C Freezing container #12009077BioradGelDoc Go Imaging system 980123001HLC DitabisHeating block 10027-242VWRIncubating orbital shaker 3500I 521-1647VWRMicro Star 17 R centrifuge A25977InvitrogenMinicell running chamber 521-2161VWRMiniStar table centrifuge ND-LIE-PRThermo Fisher ScientificNanoDrop™ lite Spectrometer 612-4552VWRPipetting aid Powerpette Pro 9907200HirschmannPipetus® 1645052BioradPower Pack™ HC Power Supply 1861096BioradT100 Thermal cycler 1703940BioradTrans-Blot SD 849-30001-3analytik JenaUVP Hybridizer Oven 824827VWRVWR Ergonomic high perfomance multichannel pipet EI0001InvitrogenXcell II Blot module EI9051InvitrogenXcell Sure Lock (TM) 150 4.2 Material and methods Table 4.5 continuation. Material used. ReferenceSupplierName Kits: 1705062BioRadClarity Max™ Western ECL Substrate E1910PromegaDual Luciferase Reporter Assay System 12663QiagenHiSpeed® Plasmid Maxi Kit (25) L3000-001InvitrogenLipofectamine(R) 3000 11791020InvitrogenGateway™ LR Clonase™ II enzyme-mix 28004QiagenMinElute® PCR Purification Kit (50) T1020New England BiolabsMonarch DNA Gel Extraction Kit 27106QiagenQIAprep® Spin Minipred Kit (250) Consumables: 65.989.002Sarstedt8er lidchain, flat 300912BDBD 10ml syringe LuerLok™ tip #3693CorningCorning Plate 72.379SarstedtCryopure tube 1.8ml white 357551FalconSerological pipet SLMP025SSMerckMillex-MP 25mm sterile filter unit 72.985.002SarstedtMultiply® µ-strip 0.2ml chain 861.569.005SarstedtInoculation spreader 72.695.700SarstedtReaction tube 2ml, LowBind 70.760.502SarstedtStacked pipette tips 02.010.00.015pfm medicalSurgical blade No.15 pfm 833.922SarstedtTC-Platte 24-well standard F 833.920SarstedtTC-Platte 6-well standard F 83.392.4005SarstedtTC-Platte 96-well standard F 83.3902SarstedtTC-Scale 100, standard Further information on the used primers and vectors can be found in the appendix (see Section A.1). 4.2.2 Aggregation of published NOTCH1 variants Entries concerning NOTCH1 from the Cosmic database were retrieved as of June 2020 (cancer.sanger.ac.uk). 364 Variants in Clinvar were retrieved as of April 2021 using the SimpleClinvar Online tool. 365 Variants in gnomAD (V3.1) were used based on the freeze of April 2021. Publications reporting NOTCH1-related CHD patients were collected from PubMed as of June 2024. Search parameters were "NOTCH1 and (CHD or congenital heart defects)". Publications were manually 151 4 Phenotypic variability in patients with NOTCH1-variants revised. Variants that were explicitly cited from previous publications or that were missing information regarding position or patients phenotype were excluded from the analysis. Furthermore, synonymous variants and variants that were considered benign by the authors were excluded from the collection. Two-hundred-forty-two variants from 49 publications were eventually considered.226–229,366–410 4.2.3 Enrichment testing Enrichment testing for NOTCH1-variants was performed on the cohort as outlined in Chapter 1 (Section 1.2.1). Samples were restricted to European ancestry, and related samples were kept in the analysis. Testing was generally based on ultrarare variants (MAC ≤ 2). Domains and modification sites in NOTCH1 were defined based on Uniprot (Acc.no. P46531), Phosphosite (www.phosphosite.org/proteinAction.action?id=8548) and Interpro (https://www.ebi.ac.uk/interpro/) (see Appendix, Table A.5). Enrichment testing was performed in R studio using a one-sided Fisher ´ s exact test. P-values were corrected for multiple testing using false-discovery-rate (FDR). 4.2.4 DNA extraction from EDTA-blood samples The Gentra Puregene Blood Kit (Qiagen Ref: 158445) was used for DNA extraction from fresh EDTA- blood-samples. 2 ml blood was used as input, and the extraction was performed according to the manufacturer ´ s protocol. Resuspension of the DNA was performed in 50 µ l H 2 O. The concentration and purity of the extracted DNA was measured using a NanoDrop™ Lite. 4.2.5 Capillary sequencing Variants identified in the NGS-based data were confirmed using capillary sequencing. In addition, segregation of the variant in family members was performed when such samples were available. Reference sequences of the consensus coding sequence (CCDS) were retrieved from Ensembl (www.ensembl.org) and the genomic sequence was based on RefSeq (https://www.ncbi.nlm.nih.gov/nuccore). Primer pairs were designed to include the variant of interest in an amplicon of approx. 250-1000 bp. Primer candidates were checked for potential unspecific binding sites in the human genome and formation of hairpin structures using the freely available online tools Primer3 (https://primer3.ut.ee/) and OligoCalc (http://biotools.nubic.northwestern.edu/OligoCalc.html). PCR Amplification was done using Applied Biosystems™ AmpliTaq Gold™ Fast PCR Master Mix (ThermoFisher) in a 50 µ l scale. The manufacturer ´ s protocol was followed using 75 ng genomic DNA and 1.5 µ l of a 10 µ M-dilution per primer. Annealing temperature and elongation time were calculated based on the primer properties and amplicon size. The PCR was performed for 40 cycles. To verify amplification, 5 µ l PCR product was separated using a 1 % agarose gel. Upon successful amplification, purification of the remaining volume was performed using the QIAquick PCR Purification 152 4.2 Material and methods Kit. In the presence of additional products, the entire volume was separated on a 0.5 % agarose gel. The correct band was cut and the desired fragment was extracted using the Monarch® DNA Gel Extraction Kit. The quantity and purity of the PCR product were evaluated using a NanoDrop™ Lite. 10 to 20 ng of purified PCR product and 5 µ l of the 10 µ M primer dilution were prepared and adjusted to a total volume of 10 µ l. Capillary sequencing was carried out through Eurofins/GATC (https://eurofinsgenomics.eu/). The sequencing results were evaluated using the SnapGene software (www.snapgene.com). 4.2.6 Site-directed mutagenesis The introduction of the variants into the cDNA-plasmids was carried out by PCR-based site-directed mutagenesis (SDM). Using primers carrying the desired alteration, a linear amplicon of the full vector was generated. For mutagenesis of the ICN-cDNA, Platinum™ SuperFi™ DNA Polymerase was used. Forward and reverse primers were designed to overlap for > 15 bases and incorporate the variant mismatch in the middle of the sequence. Standard design parameters such as high GC content at the 3 ´ -position of the primer as well as comparable melting temperatures and overall GC-contents were considered, undesired binding outside the target region was excluded using NCBI-blast. Successful amplification was tested on a 1 % agarose gel using 5 µ l product. In order to remove the input DNA, DpnI digestion was carried out using 1 µ l per reaction and incubation at 37 °C for at least one hour. Subsequently, the digested PCR product was separated on a 0.5 % agarose gel and the appropriate band was cut. The extraction was carried out using the Monarch Gel Extraction kit with an elution volume of 8-10 µ l H 2 O. Transformation of the purified PCR product was done in TOP10-E.coli as described below. Plasmid DNA isolated from grown colonies was checked for correct size using a 1 % agarose gel and verification of the sequence surrounding the targeted site was carried out by capillary sequencing through Eurofins/GATC (https://eurofinsgenomics.eu/). Due to the size of the plasmid (13065 bp), this approach required adaptation for the mutagenesis of the full-length NOTCH1 construct. Amplification with the regular protocol yielded only very low amplifications and no successful bacterial colonies. SDM on the pcDNA-NOTCH1-FLAG vector (13kb) was carried out using the SuperFi II-MasterMix based on the manufacturer ´ s protocol. Mutagenesis primers were designed to be directly adjacent to each other on the 5 ´ end but not overlapping. One primer (either the forward or the reverse) carried the desired variant as a mismatch in the middle of its sequence. PCR was then performed to generate a linearised product of the complete vector with the variant included. The PCR protocol was adapted from the manufacturers instructions, the elongation time was calculated with 1min per 1kB (see Table 4.6). 153 Bibliography [302] F. Schuhmann, D. R. Kattnig, and I. A. Solov’yov, “Exploring Post-activation Conformational Changes in Pigeon Cryptochrome 4,” The Journal of Physical Chemistry. B, vol. 125, no. 34, pp. 9652–9659, Sep. 2, 2021, issn: 1520-5207. doi: 10.1021/acs.jpcb.1c02795. pmid: 34327996. [303] A. Kohlmeyer, TopoTools: Release 1.6 with CHARMM export in topogromacs, version v1.6, Zenodo, Apr. 22, 2016. doi: 10.5281/ZENODO.50249. [Online]. Available: https://zenodo.org/ record/50249 (visited on 07/21/2023). [304] A. Bakan, L. M. Meireles, and I. Bahar, “ProDy : Protein Dynamics Inferred from Theory and Experiments,” Bioinformatics, vol. 27, no. 11, pp. 1575–1577, Jun. 1, 2011, issn: 1367-4811, 1367-4803. doi: 10.1093/bioinformatics/btr168. [Online]. Available: https://academic.oup.com/ bioinformatics/article/27/11/1575/217006 (visited on 07/21/2023). [305] A. W. Poole and M. L. Jones, “A SHPing tale: Perspectives on the regulation of SHP-1 and SHP-2 tyrosine phosphatases by the C-terminal tail,” Cellular Signalling, vol. 17, no. 11, pp. 1323–1332, Nov. 2005, issn: 08986568. doi: 10.1016/j.cellsig.2005.05.016. [Online]. Available: https://linkinghub.elsevier.com/retrieve/pii/S0898656805001233 (visited on 09/29/2023). [306] M. Anselmi and J. S. Hub, “Revealing allostery in PTPN11 SH2 domains from MD simulations,” Biophysics, preprint, Jun. 30, 2022. doi: 10.1101/2022.06.29.498118. [Online]. Available: http://biorxiv.org/lookup/doi/10.1101/2022.06.29.498118 (visited on 09/29/2023). [307] T. Niihori et al., “Functional analysis of PTPN11/SHP-2 mutants identified in Noonan syndrome and childhood leukemia,” Journal of Human Genetics, vol. 50, no. 4, pp. 192–202, Apr. 2005, issn: 1434-5161, 1435-232X. doi: 10.1007/s10038-005-0239-7. [Online]. Available: https: //www.nature.com/articles/jhg200530 (visited on 05/28/2023). [308] P. Gkeka et al., “Investigating the Structure and Dynamics of the PIK3CA Wild-Type and H1047R Oncogenic Mutant,” PLoS Computational Biology, vol. 10, no. 10, B. L. De Groot, Ed., e1003895, Oct. 23, 2014, issn: 1553-7358. doi: 10.1371/journal.pcbi.1003895. [Online]. Available: https://dx.plos.org/10.1371/journal.pcbi.1003895 (visited on 10/03/2023). [309] H. Leontiadou, I. Galdadas, C. Athanasiou, and Z. Cournia, “Insights into the mechanism of the PIK3CA E545K activating mutation using MD simulations,” Scientific Reports, vol. 8, no. 1, p. 15544, Oct. 19, 2018, issn: 2045-2322. doi: 10.1038/s41598-018-27044-6. [Online]. Available: https://www.nature.com/articles/s41598-018-27044-6 (visited on 10/03/2023). [310] M. Mente¸s, B. B. Karakuzulu, G. B. Uçar, and C. Yandım, “Comparative molecular dynamics analyses on PIK3CA hotspot mutations with PI3K α specific inhibitors and ATP,” Computational Biology and Chemistry, vol. 99, p. 107726, Aug. 2022, issn: 14769271. doi: 10.1016/j. compbiolchem.2022.107726. [Online]. Available: https://linkinghub.elsevier.com/retrieve/pii/ S1476927122001062 (visited on 10/03/2023). 256 Bibliography [311] M. Zhang, H. Jang, and R. Nussinov, “The mechanism of PI3K α activation at the atomic level,” Chemical Science, vol. 10, no. 12, pp. 3671–3680, Mar. 28, 2019, issn: 2041-6520. doi: 10.1039/c8sc04498h. pmid: 30996962. [312] N. Jin et al., “Therapeutic implications of activating noncanonical PIK3CA mutations in head and neck squamous cell carcinoma,” Journal of Clinical Investigation, vol. 131, no. 22, e150335, Nov. 15, 2021, issn: 1558-8238. doi: 10.1172/JCI150335. [Online]. Available: https://www.jci. org/articles/view/150335 (visited on 08/08/2023). [313] C. Fritsch et al., “Characterization of the Novel and Specific PI3K α Inhibitor NVP-BYL719 and Development of the Patient Stratification Strategy for Clinical Trials,” Molecular Cancer Therapeutics, vol. 13, no. 5, pp. 1117–1129, May 1, 2014, issn: 1535-7163, 1538-8514. doi: 10.1158/1535-7163.MCT-13-0865. [Online]. Available: https://aacrjournals.org/mct/article/13/ 5/1117/91923/Characterization-of-the-Novel-and-Specific-PI3K (visited on 06/29/2024). [314] E. Gazave et al., “Origin and evolution of the Notch signalling pathway: An overview from eukaryotic genomes,” BMC evolutionary biology, vol. 9, p. 249, Oct. 13, 2009, issn: 1471-2148. doi: 10.1186/1471-2148-9-249. pmid: 19825158. [315] Morgan, Thomas Hunt, “Sex-linked inheritance in Drosophila.,” Carnegie institution of Washington No. 237., 1916. [316] J. Chen and J. D. Griffin, “Using a zebrafish Danio rerio model system to study NOTCH1 -induced T-cell leukaemia,” Journal of Fish Biology, vol. 73, no. 5, pp. 1121–1138, Oct. 2008, issn: 00221112, 10958649. doi: 10.1111/j.1095-8649.2008.01979.x. [Online]. Available: https://onlinelibrary.wiley.com/doi/10.1111/j.1095-8649.2008.01979.x (visited on 10/01/2023). [317] L. Zhao et al., “Notch signaling regulates cardiomyocyte proliferation during zebrafish heart regeneration,” Proceedings of the National Academy of Sciences, vol. 111, no. 4, pp. 1403– 1408, Jan. 28, 2014, issn: 0027-8424, 1091-6490. doi: 10.1073/pnas.1311705111. [Online]. Available: https://pnas.org/doi/full/10.1073/pnas.1311705111 (visited on 10/01/2023). [318] R. A. Conlon, A. G. Reaume, and J. Rossant, “Notch1 is required for the coordinate segmentation of somites,” Development (Cambridge, England), vol. 121, no. 5, pp. 1533–1545, May 1995, issn: 0950-1991. doi: 10.1242/dev.121.5.1533. pmid: 7789282. [319] S. S. Huppert et al., “Embryonic lethality in mice homozygous for a processing-deficient allele of Notch1,” Nature, vol. 405, no. 6789, pp. 966–970, Jun. 22, 2000, issn: 0028-0836. doi: 10.1038/35016111. pmid: 10879540. [320] S. N. Koenig, K. Bosse, U. Majumdar, E. M. Bonachea, F. Radtke, and V. Garg, “Endothelial Notch1 Is Required for Proper Development of the Semilunar Valves and Cardiac Outflow Tract,” Journal of the American Heart Association, vol. 5, no. 4, e003075, Apr. 3, 2016, issn: 2047-9980. doi: 10.1161/JAHA.115.003075. [Online]. Available: https://www.ahajournals.org/ doi/10.1161/JAHA.115.003075 (visited on 10/01/2023). 257 Bibliography [321] S. N. Koenig et al., “Notch1 haploinsufficiency causes ascending aortic aneurysms in mice,” JCI insight, vol. 2, no. 21, e91353, 91353, Nov. 2, 2017, issn: 2379-3708. doi: 10.1172/jci. insight.91353. pmid: 29093270. [322] C. V. Theodoris et al., “Long telomeres protect against age-dependent cardiac disease caused by NOTCH1 haploinsufficiency,” Journal of Clinical Investigation, vol. 127, no. 5, pp. 1683– 1688, Mar. 27, 2017, issn: 0021-9738, 1558-8238. doi: 10.1172/JCI90338. [Online]. Available: https://www.jci.org/articles/view/90338 (visited on 10/01/2023). [323] F. P. Limbourg, K. Takeshita, F. Radtke, R. T. Bronson, M. T. Chin, and J. K. Liao, “Essential role of endothelial Notch1 in angiogenesis,” Circulation, vol. 111, no. 14, pp. 1826–1832, Apr. 12, 2005, issn: 1524-4539. doi: 10.1161/01.CIR.0000160870.93058.DD. pmid: 15809373. [324] T. Gridley, “Notch Signaling in the Vasculature,” in Current Topics in Developmental Biology, vol. 92, Elsevier, 2010, pp. 277–309, isbn: 978-0-12-380914-8. doi: 10.1016/S0070-2153(10) 92009-7. [Online]. Available: https://linkinghub.elsevier.com/retrieve/pii/S0070215310920097 (visited on 10/03/2023). [325] C. Siebel and U. Lendahl, “Notch Signaling in Development, Tissue Homeostasis, and Disease,” Physiological Reviews, vol. 97, no. 4, pp. 1235–1294, Oct. 1, 2017, issn: 0031-9333, 1522-1210. doi: 10.1152/physrev.00005.2017. [Online]. Available: https://www.physiology.org/ doi/10.1152/physrev.00005.2017 (visited on 10/03/2023). [326] N. Gaiano and G. Fishell, “The Role of Notch in Promoting Glial and Neural Stem Cell Fates,” Annual Review of Neuroscience, vol. 25, no. 1, pp. 471–490, Mar. 2002, issn: 0147- 006X, 1545-4126. doi: 10.1146/annurev.neuro.25.030702.130823. [Online]. Available: https://www.annualreviews.org/doi/10.1146/annurev.neuro.25.030702.130823 (visited on 10/01/2023). [327] S. Lütolf, F. Radtke, M. Aguet, U. Suter, and V. Taylor, “Notch1 is required for neuronal and glial differentiation in the cerebellum,” Development, vol. 129, no. 2, pp. 373–385, Jan. 15, 2002, issn: 1477-9129, 0950-1991. doi: 10.1242/dev.129.2.373. [Online]. Available: https: //journals.biologists.com/dev/article/129/2/373/17987/Notch1-is-required-for-neuronal-and- glial (visited on 10/02/2023). [328] J. Hatakeyama and R. Kageyama, “Notch1 Expression Is Spatiotemporally Correlated with Neurogenesis and Negatively Regulated by Notch1-Independent Hes Genes in the Developing Nervous System,” Cerebral Cortex, vol. 16, pp. i132–i137, suppl_1 Jul. 1, 2006, issn: 1460- 2199, 1047-3211. doi: 10.1093/cercor/bhj166. [Online]. Available: http://academic.oup.com/ cercor/article/16/suppl_1/i132/270125/Notch1-Expression-Is-Spatiotemporally-Correlated (visited on 10/02/2023). [329] S. Lowell, P. Jones, I. Le Roux, J. Dunne, and F. M. Watt, “Stimulation of human epidermal differentiation by Delta–Notch signalling at the boundaries of stem-cell clusters,” Current Biology, vol. 10, no. 9, pp. 491–500, May 2000, issn: 09609822. doi: 10.1016/S0960-9822(00) 258 Bibliography 00451-6. [Online]. Available: https://linkinghub.elsevier.com/retrieve/pii/S0960982200004516 (visited on 10/01/2023). [330] J. S. Guseh et al., “Notch signaling promotes airway mucous metaplasia and inhibits alveolar development,” Development, vol. 136, no. 10, pp. 1751–1759, May 15, 2009, issn: 1477-9129, 0950-1991. doi: 10.1242/dev.029249. [Online]. Available: https://journals.biologists.com/dev/ article/136/10/1751/43709/Notch-signaling-promotes-airway-mucous-metaplasia (visited on 10/01/2023). [331] K. Laky and B. Fowlkes, “Notch signaling in CD4 and CD8 T cell development,” Current Opinion in Immunology, vol. 20, no. 2, pp. 197–202, Apr. 2008, issn: 09527915. doi: 10.1016/j.coi.2008. 03.004. [Online]. Available: https://linkinghub.elsevier.com/retrieve/pii/S0952791508000290 (visited on 10/01/2023). [332] L. C. Murtaugh, B. Z. Stanger, K. M. Kwan, and D. A. Melton, “Notch signaling controls multiple steps of pancreatic differentiation,” Proceedings of the National Academy of Sciences, vol. 100, no. 25, pp. 14 920–14 925, Dec. 9, 2003, issn: 0027-8424, 1091-6490. doi: 10.1073/ pnas.2436557100. [Online]. Available: https://pnas.org/doi/full/10.1073/pnas.2436557100 (visited on 10/01/2023). [333] M. Nobta et al., “Critical regulation of bone morphogenetic protein-induced osteoblastic differentiation by Delta1/Jagged1-activated Notch1 signaling,” The Journal of Biological Chemistry, vol. 280, no. 16, pp. 15842–15848, Apr. 22, 2005, issn: 0021-9258. doi: 10.1074/jbc. M412891200. pmid: 15695512. [334] Y. J. Jiang, B. L. Aerne, L. Smithers, C. Haddon, D. Ish-Horowicz, and J. Lewis, “Notch signalling and the synchronization of the somite segmentation clock,” Nature, vol. 408, no. 6811, pp. 475–479, Nov. 23, 2000, issn: 0028-0836. doi: 10.1038/35044091. pmid: 11100729. [335] Á. Raya et al., “Notch activity induces Nodal expression and mediates the establishment of left–right asymmetry in vertebrate embryos,” Genes & Development, vol. 17, no. 10, pp. 1213– 1218, May 15, 2003, issn: 0890-9369, 1549-5477. doi: 10.1101/gad.1084403. [Online]. Available: http://genesdev.cshlp.org/lookup/doi/10.1101/gad.1084403 (visited on 10/02/2023). [336] M. V. Gustafsson et al., “Hypoxia Requires Notch Signaling to Maintain the Undifferentiated Cell State,” Developmental Cell, vol. 9, no. 5, pp. 617–628, Nov. 2005. doi: 10.1016/j.devcel. 2005.09.010. [337] C. Sahlgren, M. V. Gustafsson, S. Jin, L. Poellinger, and U. Lendahl, “Notch signaling mediates hypoxia-induced tumor cell migration and invasion,” Proceedings of the National Academy of Sciences, vol. 105, no. 17, pp. 6392–6397, Apr. 29, 2008, issn: 0027-8424, 1091-6490. doi: 10.1073/pnas.0802047105. [Online]. Available: https://pnas.org/doi/full/10.1073/pnas. 0802047105 (visited on 11/08/2023). 259 Bibliography [338] M. Guo, Y. Niu, M. Xie, X. Liu, and X. Li, “Notch signaling, hypoxia, and cancer,” Frontiers in Oncology, vol. 13, p. 1078768, Jan. 31, 2023, issn: 2234-943X. doi: 10.3389/fonc.2023. 1078768. [Online]. Available: https://www.frontiersin.org/articles/10.3389/fonc.2023.1078768/ full (visited on 07/02/2023). [339] M. P. Adam et al., “Adams-Oliver Syndrome,” GeneReviews, 2016. pmid: 27077170. [340] M. L. Loscalzo, D. L. M. Goh, B. Loeys, K. C. Kent, P. J. Spevak, and H. C. Dietz, “Familial thoracic aortic dilation and bicommissural aortic valve: A prospective analysis of natural history and inheritance,” American Journal of Medical Genetics. Part A, vol. 143A, no. 17, pp. 1960–1967, Sep. 1, 2007, issn: 1552-4825. doi: 10.1002/ajmg.a.31872. pmid: 17676603. [341] A. P. Weng et al., “Activating Mutations of NOTCH1 in Human T Cell Acute Lymphoblastic Leukemia,” Science, vol. 306, no. 5694, pp. 269–271, Oct. 8, 2004, issn: 0036-8075, 1095- 9203. doi: 10.1126/science.1102160. [Online]. Available: https://www.science.org/doi/10. 1126/science.1102160 (visited on 09/20/2023). [342] T. Fukusumi and J. Califano, “The NOTCH Pathway in Head and Neck Squamous Cell Carcinoma,” Journal of Dental Research, vol. 97, no. 6, pp. 645–653, Jun. 2018, issn: 0022- 0345, 1544-0591. doi: 10.1177/0022034518760297. [Online]. Available: http://journals. sagepub.com/doi/10.1177/0022034518760297 (visited on 10/01/2023). [343] M. Reedijk et al., “High-level coexpression of JAG1 and NOTCH1 is observed in human breast cancer and is associated with poor overall survival,” Cancer Research, vol. 65, no. 18, pp. 8530–8537, Sep. 15, 2005, issn: 0008-5472. doi: 10.1158/0008-5472.CAN-05-1069. pmid: 16166334. [344] B. Westhoff et al., “Alterations of the Notch pathway in lung cancer,” Proceedings of the National Academy of Sciences of the United States of America, vol. 106, no. 52, pp. 22293– 22298, Dec. 29, 2009, issn: 1091-6490. doi: 10.1073/pnas.0907781106. pmid: 20007775. [345] Y. Miyamoto et al., “Notch mediates TGF alpha-induced changes in epithelial differentiation during pancreatic tumorigenesis,” Cancer Cell, vol. 3, no. 6, pp. 565–576, Jun. 2003, issn: 1535-6108. doi: 10.1016/s1535-6108(03)00140-5. pmid: 12842085. [346] Y. Tashima and T. Okajima. “Congenital diseases caused by defective O-glycosylation of Notch receptors.” Nagoya University Graduate School of Medicine, School of Medicine: 3. (Aug. 2018), [Online]. Available: https://doi.org/10.18999/nagjms.80.3.299 (visited on 06/30/2023), pre-published. [347] W. R. Gordon et al., “Structure of the Notch1-negative regulatory region: Implications for normal activation and pathogenic signaling in T-ALL,” Blood, vol. 113, no. 18, pp. 4381–4390, Apr. 30, 2009, issn: 0006-4971, 1528-0020. doi: 10.1182/blood-2008-08-174748. [Online]. Available: https://ashpublications.org/blood/article/113/18/4381/25743/Structure-of-the- Notch1negative-regulatory-region (visited on 06/30/2023). 260 Bibliography [348] M. J. Malecki et al., “Leukemia-associated mutations within the NOTCH1 heterodimerization domain fall into at least two distinct mechanistic classes,” Molecular and Cellular Biology, vol. 26, no. 12, pp. 4642–4651, Jun. 2006, issn: 0270-7306. doi: 10.1128/MCB.01655-05. pmid: 16738328. [349] V. Deregowski, E. Gazzerro, L. Priest, S. Rydziel, and E. Canalis, “Role of the Ram Domain and Ankyrin Repeats on Notch Signaling and Activity in Cells of Osteoblastic Lineage,” Journal of Bone and Mineral Research, vol. 21, no. 8, pp. 1317–1326, May 22, 2006, issn: 08840431. doi: 10.1359/jbmr.060505. [Online]. Available: http://doi.wiley.com/10.1359/jbmr.060505 (visited on 06/30/2023). [350] J. C. Aster, L. Xu, F. G. Karnell, V. Patriub, J. C. Pui, and W. S. Pear, “Essential roles for ankyrin repeat and transactivation domains in induction of T-cell leukemia by notch1,” Molecular and Cellular Biology, vol. 20, no. 20, pp. 7505–7515, Oct. 2000, issn: 0270-7306. doi: 10.1128/MCB.20.20.7505-7515.2000. pmid: 11003647. [351] M. T. Ehebauer, D. Y. Chirgadze, P. Hayward, A. Martinez Arias, and T. L. Blundell, “Highresolution crystal structure of the human Notch 1 ankyrin domain,” Biochemical Journal, vol. 392, no. 1, pp. 13–20, Nov. 15, 2005, issn: 0264-6021, 1470-8728. doi: 10 .1042/ BJ20050515. [Online]. Available: https://portlandpress.com/biochemj/article/392/1/13/44076/ High-resolution-crystal-structure-of-the-human (visited on 06/30/2023). [352] K. Cravero et al., “NOTCH1 PEST domain variants are responsive to standard of care treatments despite distinct transformative properties in a breast cancer model,” Oncotarget, vol. 13, no. 1, pp. 373–386, Feb. 16, 2022, issn: 1949-2553. doi: 10.18632/oncotarget.28200. [Online]. Available: https://www.oncotarget.com/lookup/doi/10.18632/oncotarget.28200 (visited on 06/30/2023). [353] W. Saiki, C. Ma, T. Okajima, and H. Takeuchi, “Current Views on the Roles of O-Glycosylation in Controlling Notch-Ligand Interactions,” Biomolecules, vol. 11, no. 2, p. 309, Feb. 18, 2021, issn: 2218-273X. doi: 10.3390/biom11020309. pmid: 33670724. [354] V. M. Panin, V. Papayannopoulos, R. Wilson, and K. D. Irvine, “Fringe modulates Notch–ligand interactions,” Nature, vol. 387, no. 6636, pp. 908–912, Jun. 1997, issn: 0028-0836, 1476-4687. doi: 10.1038/43191. [Online]. Available: https://www.nature.com/articles/43191 (visited on 09/05/2023). [355] M. B. Andrawes et al., “Intrinsic Selectivity of Notch 1 for Delta-like 4 Over Delta-like 1,” Journal of Biological Chemistry, vol. 288, no. 35, pp. 25477–25489, Aug. 2013, issn: 00219258. doi: 10.1074/jbc.M113.454850. [Online]. Available: https://linkinghub.elsevier.com/retrieve/pii/ S0021925820491513 (visited on 09/05/2023). [356] C. J. Fryer, J. B. White, and K. A. Jones, “Mastermind Recruits CycC-CDK8 to Phosphorylate the Notch ICD and Coordinate Activation with Turnover,” Molecular Cell, vol. 16, no. 4, pp. 509– 520, Nov. 2004. doi: 10.1016/j.molcel.2004.10.014. 261 Bibliography [357] R. Morrugares et al., “Phosphorylation-dependent regulation of the NOTCH1 intracellular domain by dual-specificity tyrosine-regulated kinase 2,” Cellular and Molecular Life Sciences, vol. 77, no. 13, pp. 2621–2639, Jul. 2020, issn: 1420-682X, 1420-9071. doi: 10.1007/s00018- 019-03309-9. [Online]. Available: http://link.springer.com/10.1007/s00018-019-03309-9 (visited on 09/05/2023). [358] R. Benedito et al., “The Notch Ligands Dll4 and Jagged1 Have Opposing Effects on Angiogenesis,” Cell, vol. 137, no. 6, pp. 1124–1135, Jun. 2009, issn: 00928674. doi: 10.1016/j.cell.2009. 03.025. [Online]. Available: https://linkinghub.elsevier.com/retrieve/pii/S0092867409003249 (visited on 09/05/2023). [359] K. Hozumi, “Distinctive properties of the interactions between Notch and Notch ligands,” Development, Growth & Differentiation, vol. 62, no. 1, pp. 49–58, Jan. 2020, issn: 0012-1592, 1440-169X. doi: 10.1111/dgd.12641. [Online]. Available: https://onlinelibrary.wiley.com/doi/10. 1111/dgd.12641 (visited on 09/05/2023). [360] B. - K. Liao and A. C. Oates, “Delta-Notch signalling in segmentation,” Arthropod Structure & Development, vol. 46, no. 3, pp. 429–447, May 2017, issn: 14678039. doi: 10.1016/j.asd.2016. 11.007. [Online]. Available: https://linkinghub.elsevier.com/retrieve/pii/S1467803916301918 (visited on 10/01/2023). [361] T. Kume, “Novel insights into the differential functions of Notch ligands in vascular formation,” Journal of Angiogenesis Research, vol. 1, no. 1, p. 8, 2009, issn: 2040-2384. doi: 10.1186/ 2040-2384-1-8. [Online]. Available: http://vascularcell.com/index.php/vc/article/view/10.1186- 2040-2384-1-8 (visited on 10/01/2023). [362] N. Nandagopal, L. A. Santat, L. LeBon, D. Sprinzak, M. E. Bronner, and M. B. Elowitz, “Dynamic Ligand Discrimination in the Notch Signaling Pathway,” Cell, vol. 172, no. 4, 869– 880.e19, Feb. 2018, issn: 00928674. doi: 10.1016/j.cell.2018.01.002. [Online]. Available: https://linkinghub.elsevier.com/retrieve/pii/S0092867418300023 (visited on 09/05/2023). [363] D. MacGrogan et al., “Sequential Ligand-Dependent Notch Signaling Activation Regulates Valve Primordium Formation and Morphogenesis,” Circulation Research, vol. 118, no. 10, pp. 1480–1497, May 13, 2016, issn: 0009-7330, 1524-4571. doi: 10.1161/CIRCRESAHA.115. 308077. [Online]. Available: https://www.ahajournals.org/doi/10.1161/CIRCRESAHA.115. 308077 (visited on 09/05/2023). [364] J. G. Tate et al., “COSMIC: The Catalogue Of Somatic Mutations In Cancer,” Nucleic Acids Research, vol. 47, no. D1, pp. D941–D947, Jan. 8, 2019, issn: 0305-1048, 1362-4962. doi: 10.1093/nar/gky1015. [Online]. Available: https://academic.oup.com/nar/article/47/D1/D941/ 5146192 (visited on 07/02/2023). [365] E. Pérez-Palma, M. Gramm, P. Nürnberg, P. May, and D. Lal, “Simple ClinVar: An interactive web server to explore and retrieve gene and disease variants aggregated in ClinVar database,” Nucleic Acids Research, vol. 47, no. W1, W99–W105, Jul. 2, 2019, issn: 0305-1048, 1362- 262 Bibliography 4962. doi: 10.1093/nar/gkz411. [Online]. Available: https://academic.oup.com/nar/article/47/ W1/W99/5494761 (visited on 05/26/2023). [366] D. Alankarage et al., “Identification of clinically actionable variants from genome sequencing of families with congenital heart disease,” Genetics in Medicine, vol. 21, no. 5, pp. 1111– 1120, May 2019, issn: 10983600. doi: 10.1038/s41436-018-0296-x. [Online]. Available: https://linkinghub.elsevier.com/retrieve/pii/S1098360021014726 (visited on 06/08/2024). [367] V. Bansal et al., “Outlier-Based Identification of Copy Number Variations Using Targeted Resequencing in a Small Cohort of Patients with Tetralogy of Fallot,” PLoS ONE, vol. 9, no. 1, C. Liu, Ed., e85375, Jan. 6, 2014, issn: 1932-6203. doi: 10.1371/journal.pone.0085375. [Online]. Available: https://dx.plos.org/10.1371/journal.pone.0085375 (visited on 06/08/2024). [368] B. Bai et al., “The pregnancy-associated spontaneous coronary artery dissection in a young woman with a novel missense mutation in NOTCH1: A case report.,” BMC medical genetics, vol. 21, no. 1, p. 119, Jun. 2020, issn: 1471-2350. doi: 10.1186/s12881-020-01058-2. pmid: 32487031. [369] S. Bea et al., “Landscape of somatic mutations and clonal evolution in mantle cell lymphoma.,” Proceedings of the National Academy of Sciences of the United States of America, vol. 110, no. 45, pp. 18250–18 255, Nov. 2013, issn: 1091-6490. doi: 10.1073/pnas.1314608110. pmid: 24145436. [370] G. M. Blue et al., “The promises and challenges of exome sequencing in familial, nonsyndromic congenital heart disease.,” International journal of cardiology, vol. 230, pp. 155– 163, Mar. 2017, issn: 1874-1754. doi: 10.1016/j.ijcard.2016.12.024. pmid: 27989580. [371] P. Boerkoel et al., “ NOTCH1 loss of the TAD and PEST domain: An antimorph?” American Journal of Medical Genetics Part A, vol. 191, no. 6, pp. 1593–1598, Jun. 2023, issn: 1552- 4825, 1552-4833. doi: 10.1002/ajmg.a.63167. [Online]. Available: https://onlinelibrary.wiley. com/doi/10.1002/ajmg.a.63167 (visited on 06/08/2024). [372] E. M. Bonachea et al., “Use of a targeted, combinatorial next-generation sequencing approach for the study of bicuspid aortic valve.,” BMC medical genomics, vol. 7, p. 56, Sep. 2014, issn: 1755-8794. doi: 10.1186/1755-8794-7-56. pmid: 25260786. [373] G. Chapman et al., “Functional genomics and gene-environment interaction highlight the complexity of congenital heart disease caused by Notch pathway variants.,” Human molecular genetics, vol. 29, no. 4, pp. 566–579, Mar. 2020, issn: 1460-2083. doi: 10.1093/hmg/ddz270. pmid: 31813956. [374] N. Dargis et al., “Identification of Gender-Specific Genetic Variants in Patients With Bicuspid Aortic Valve.,” The American journal of cardiology, vol. 117, no. 3, pp. 420–426, Feb. 2016, issn: 1879-1913. doi: 10.1016/j.amjcard.2015.10.058. pmid: 26708639. 263 Bibliography [375] H. Daum et al., “Grandparental genotyping enhances exome variant interpretation.,” American journal of medical genetics. Part A, vol. 182, no. 4, pp. 689–696, Apr. 2020, issn: 1552-4833. doi: 10.1002/ajmg.a.61511. pmid: 32027463. [376] R. Debiec et al., “Novel loss of function mutation in NOTCH1 in a family with bicuspid aortic valve, ventricular septal defect, thoracic aortic aneurysm, and aortic valve stenosis.,” Molecular genetics and genomic medicine, vol. 8, no. 10, e1437, Oct. 2020, issn: 2324-9269. doi: 10.1002/mgg3.1437. pmid: 32720365. [377] M. C. Digilio et al., “Familial aggregation of "apple peel" intestinal atresia and cardiac leftsided obstructive lesions: A possible causal relationship with NOTCH1 gene mutations.,” American journal of medical genetics. Part A, vol. 179, no. 8, pp. 1570–1574, Aug. 2019, issn: 1552-4833. doi: 10.1002/ajmg.a.61195. pmid: 31111652. [378] M. Di Ianni et al., “A new genetic lesion in B-CLL: A NOTCH1 PEST domain mutation.,” British journal of haematology, vol. 146, no. 6, pp. 689–691, Sep. 2009, issn: 1365-2141. doi: 10.1111/j.1365-2141.2009.07816.x. pmid: 19604236. [379] B. Dudoignon et al., “Expanding the phenotype in Adams-Oliver syndrome correlating with the genotype.,” American journal of medical genetics. Part A, vol. 182, no. 1, pp. 29–37, Jan. 2020, issn: 1552-4833. doi: 10.1002/ajmg.a.61364. pmid: 31654484. [380] I. Foffa et al., “Sequencing of NOTCH1, GATA5, TGFBR1 and TGFBR2genes in familial cases of bicuspid aortic valve,” BMC Medical Genetics, vol. 14, no. 1, p. 44, Dec. 2013, issn: 1471-2350. doi: 10.1186/1471-2350-14-44. [Online]. Available: https://bmcmedgenet. biomedcentral.com/articles/10.1186/1471-2350-14-44 (visited on 06/08/2024). [381] E. Gillis et al., “Candidate Gene Resequencing in a Large Bicuspid Aortic Valve-Associated Thoracic Aortic Aneurysm Cohort: SMAD6 as an Important Contributor,” Frontiers in physiology, vol. 8, p. 400, 2017, issn: 1664-042X. doi: 10.3389/fphys.2017.00400. pmid: 28659821. [382] E. Helle et al., “Loss of function, missense, and intronic variants in NOTCH1 confer different risks for left ventricular outflow tract obstructive heart defects in two European cohorts.,” Genetic epidemiology, vol. 43, no. 2, pp. 215–226, Mar. 2019, issn: 1098-2272. doi: 10.1002/ gepi.22176. pmid: 30511478. [383] S. Huang, L. Yang, L. Zhao, R. Xu, and Y. Wu, “Novel In-Frame Deletion Mutation in NOTCH1 in a Chinese Sporadic Case of Adams–Oliver Syndrome,” DNA and Cell Biology, vol. 39, no. 5, pp. 783–789, May 1, 2020, issn: 1044-5498, 1557-7430. doi: 10.1089/dna.2019.5200. [Online]. Available: https://www.liebertpub.com/doi/10.1089/dna.2019.5200 (visited on 06/08/2024). [384] M. Iascone et al., “Identification of de novo mutations and rare variants in hypoplastic left heart syndrome.,” Clinical genetics, vol. 81, no. 6, pp. 542–554, Jun. 2012, issn: 1399-0004. doi: 10.1111/j.1399-0004.2011.01674.x. pmid: 21457232. 264 Bibliography [385] Y. Jia et al., “The diagnostic value of next generation sequencing in familial nonsyndromic congenital heart defects.,” American journal of medical genetics. Part A, vol. 167A, no. 8, pp. 1822–1829, Aug. 2015, issn: 1552-4833. doi: 10.1002/ajmg.a.37108. pmid: 25931334. [386] S. Kalayinia, M. Maleki, M. Mahdavi, and N. Mahdieh, “A novel de novo dominant mutation of NOTCH1 gene in an Iranian family with non-syndromic congenital heart disease.,” Journal of clinical laboratory analysis, vol. 34, no. 4, e23147, Apr. 2020, issn: 1098-2825. doi: 10.1002/ jcla.23147. pmid: 31867804. [387] R. Kridel et al., “Whole transcriptome sequencing reveals recurrent NOTCH1 mutations in mantle cell lymphoma.,” Blood, vol. 119, no. 9, pp. 1963–1971, Mar. 2012, issn: 1528-0020. doi: 10.1182/blood-2011-11-391474. pmid: 22210878. [388] S. Luo et al., “Prenatal Genetic Diagnosis in Three Fetuses With Left Heart Hypoplasia (LHH) From Three Unrelated Families,” Frontiers in Cardiovascular Medicine, vol. 8, p. 631374, Apr. 9, 2021, issn: 2297-055X. doi: 10.3389/fcvm.2021.631374. [Online]. Available: https: //www.frontiersin.org/articles/10.3389/fcvm.2021.631374/full (visited on 06/08/2024). [389] S. H. McKellar, D. J. Tester, M. Yagubyan, R. Majumdar, M. J. Ackerman, and T. M. Sundt, “Novel NOTCH1 mutations in patients with bicuspid aortic valve disease and thoracic aortic aneurysms.,” The Journal of thoracic and cardiovascular surgery, vol. 134, no. 2, pp. 290–296, Aug. 2007, issn: 1097-685X. doi: 10.1016/j.jtcvs.2007.02.041. pmid: 17662764. [390] J. A. N. Meester et al., “Elucidating the genetic architecture of Adams-Oliver syndrome in a large European cohort.,” Human mutation, vol. 39, no. 9, pp. 1246–1261, Sep. 2018, issn: 1098-1004. doi: 10.1002/humu.23567. pmid: 29924900. [391] S. A. Mohamed et al., “Novel missense mutations (p.T596M and p.P1797H) in NOTCH1 in patients with bicuspid aortic valve.,” Biochemical and biophysical research communications, vol. 345, no. 4, pp. 1460–1465, Jul. 2006, issn: 0006-291X. doi: 10.1016/j.bbrc.2006.05.046. pmid: 16729972. [392] F. I. Musfee et al., “Rare deleterious variants of NOTCH1 ,GATA4 ,SMAD6 , and ROBO4 are enriched in BAV with early onset complications but not in BAV with heritable thoracic aortic disease,” Molecular Genetics & Genomic Medicine, vol. 8, no. 10, e1406, Oct. 2020, issn: 2324-9269, 2324-9269. doi: 10.1002/mgg3.1406. [Online]. Available: https://onlinelibrary. wiley.com/doi/10.1002/mgg3.1406 (visited on 06/08/2024). [393] E. Overwater et al., “Results of next-generation sequencing gene panel diagnostics including copy-number variation analysis in 810 patients suspected of heritable thoracic aortic disorders.,” Human mutation, vol. 39, no. 9, pp. 1173–1192, Sep. 2018, issn: 1098-1004. doi: 10.1002/humu.23565. pmid: 29907982. 265 List of Figures A.6 Vectormap of pDONR223-ICN. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197 A.7 Vectormap of pcDNA3.1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197 A.8 Vectormap of pcDNA-NOTCH1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197 A.9 Vectormap of pDEST-MHHA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197 A.10 Capillary sequencing of multinucleotide variants found in patients C07635 and C07558. 210 A.11 Replicate of the western blot against the NOTCH1-ICN. . . . . . . . . . . . . . . . . . 212 A.12 Replicate of the western blot against the NOTCH1-ICN. . . . . . . . . . . . . . . . . . 212 A.13 Replicate of the western blot against the NOTCH1-full length. . . . . . . . . . . . . . 213 A.14 Replicate of the western blot against the NOTCH1-full length. . . . . . . . . . . . . . 213 272 List of Tables 1.1 Variant classes according to the ACMG scheme. . . . . . . . . . . . . . . . . . . . . 31 1.2 Grouping of consequences at protein level. . . . . . . . . . . . . . . . . . . . . . . . 33 1.3 Panels from UK Panelapp included for curation of the CHD-gene set. . . . . . . . . . 35 1.4 In silico predictors used for determining PP3 and BP4. . . . . . . . . . . . . . . . . . 38 1.5 Overview of applied ACMG criteria. . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 1.6 Grouping scheme of cardiac HPO terms. . . . . . . . . . . . . . . . . . . . . . . . . 44 1.7 Comparison of CHD subtype prevalences. . . . . . . . . . . . . . . . . . . . . . . . . 47 1.8 Genes associated with diseases present in the analysed patient cohort. . . . . . . . . 48 1.9 Variants per ACMG group. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 1.10 Summary table for (likely) pathogenic variants identified in NKX2-5. . . . . . . . . . . 52 1.11 Summary table for (likely) pathogenic variants identified in TAB2. . . . . . . . . . . . 54 1.12 Summary table for (likely) pathogenic variants identified in PTPN11. . . . . . . . . . . 56 1.13 Summary table for (likely) pathogenic variants identified in KMT2D. . . . . . . . . . . 58 1.14 Summary table for (likely) pathogenic variants identified in KMT2A. . . . . . . . . . . 60 1.15 Summary table for (likely) pathogenic variants identified in SMAD6. . . . . . . . . . . 62 2.1 Summary of ROC results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 2.2 Summary table for pathway enrichment at gene level. . . . . . . . . . . . . . . . . . 91 2.3 Published phenotypes of identified genes. . . . . . . . . . . . . . . . . . . . . . . . . 92 2.4 Variant carriers for VANGL2 ............................... 94 2.5 Variant carriers for JAK2 ................................. 96 2.6 Variant carriers for PDGFRB ............................... 98 2.7 Variant carriers for KAT2B ................................100 2.8 Variant carriers for RBFOX2 ...............................102 2.9 Carrier information for identified variants in PIK3CA. ..................104 4.1 Antibiotics used. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 4.2 Antibodies used. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 4.3 Cells used. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 4.4 Buffers used. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148 4.5 Material used. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 4.6 PCR program used for site directed mutagenesis. . . . . . . . . . . . . . . . . . . . . 154 4.7 Treatment after site directed mutagenesis of pcDNA-NOTCH1-FLAG. ............154 4.8 Enrichment testing in NOTCH1...............................160 4.9 Enrichment testing for domains and modification sites in NOTCH1. . . . . . . . . . . 160 4.10 Summary of samples with severe NOTCH1 variants. . . . . . . . . . . . . . . . . . . 164 273 List of Tables 4.11 Non-coding variants in a putative enhancer element in intron 27 of NOTCH1. . . . . . 175 4.12 Additional cases with NOTCH1 variants aggregated from collaboration partners. . . . 179 A.1 Vectors used. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195 A.2 Primers used . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198 A.3 Published variants in NOTCH1 used in Section 4.2.2. . . . . . . . . . . . . . . . . . . 201 A.4 Chromatograms of the variant validation and segregation testing by capillary sequencing. 206 A.5 Residues in NOTCH1 overlapping with domains and modification sites. . . . . . . . . 211 A.6 Raw data for NOTCH1-ICN luciferase assays. . . . . . . . . . . . . . . . . . . . . . . 215 A.7 Raw data for NOTCH1-FLAG luciferase assay without JAG1 co-transfection. . . . . . 216 A.8 Raw data for luciferase assays of NOTCH1 full length missense-variants. . . . . . . . 216 A.9 Raw data for luciferase assays of NOTCH1 full length stopgain-variants. . . . . . . . 217 A.10 Abbreviations used . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219 274