Skip to main content
Advertisement
  • Loading metrics

Family Based Whole Exome Sequencing Reveals the Multifaceted Role of Notch Signaling in Congenital Heart Disease

  • Christoph Preuss ,

    Contributed equally to this work with: Christoph Preuss, Melanie Capredon, Florian Wünnemann

    Affiliation Cardiovascular Genetics, Department of Pediatrics, CHU Sainte-Justine, Université de Montréal, Montreal, Québec, Canada

  • Melanie Capredon ,

    Contributed equally to this work with: Christoph Preuss, Melanie Capredon, Florian Wünnemann

    Affiliation Cardiovascular Genetics, Department of Pediatrics, CHU Sainte-Justine, Université de Montréal, Montreal, Québec, Canada

  • Florian Wünnemann ,

    Contributed equally to this work with: Christoph Preuss, Melanie Capredon, Florian Wünnemann

    Affiliations Cardiovascular Genetics, Department of Pediatrics, CHU Sainte-Justine, Université de Montréal, Montreal, Québec, Canada, Faculty of Biology, University of Muenster, Muenster, Germany

  • Philippe Chetaille,

    Affiliation Department of Pediatrics, Centre Mère Enfants Soleil, Centre Hospitalier de l'Université (CHU) de Québec, Quebec City, Québec, Canada

  • Andrea Prince,

    Affiliation Cardiovascular Genetics, Department of Pediatrics, CHU Sainte-Justine, Université de Montréal, Montreal, Québec, Canada

  • Beatrice Godard,

    Affiliation Omics-Ethics Research Group, Research Institute of Public Health, Université de Montréal, Montréal Québec, Canada

  • Severine Leclerc,

    Affiliation Cardiovascular Genetics, Department of Pediatrics, CHU Sainte-Justine, Université de Montréal, Montreal, Québec, Canada

  • Nara Sobreira,

    Affiliation McKusick Nathans Institute of Genetic Medicine, Johns Hopkins University School of Medicine, Baltimore, Maryland, United States of America

  • Hua Ling,

    Affiliation McKusick Nathans Institute of Genetic Medicine, Johns Hopkins University School of Medicine, Baltimore, Maryland, United States of America

  • Philip Awadalla,

    Affiliation Ontario Institute for Cancer Research, Toronto, Ontario, Canada

  • Maryse Thibeault,

    Affiliation Cardiovascular Genetics, Department of Pediatrics, CHU Sainte-Justine, Université de Montréal, Montreal, Québec, Canada

  • Paul Khairy,

    Affiliation Montreal Heart Institute, Université de Montréal, Montréal, Québec, Canada

  • MIBAVA Leducq consortium,

    Affiliation Membership of the MIBAVA Leducq consortium is provided in the Acknowledgements

  • Mark E. Samuels,

    Affiliations Centre de Recherche CHU Sainte Justine, Université de Montreal, Montréal, Québec, Canada, Department of Medicine, Université de Montreal, Montréal, Québec, Canada

  • Gregor Andelfinger

    gregor.andelfinger@recherche-ste-justine.qc.ca

    Affiliation Cardiovascular Genetics, Department of Pediatrics, CHU Sainte-Justine, Université de Montréal, Montreal, Québec, Canada

Abstract

Left-ventricular outflow tract obstructions (LVOTO) encompass a wide spectrum of phenotypically heterogeneous heart malformations which frequently cluster in families. We performed family based whole-exome and targeted re-sequencing on 182 individuals from 51 families with multiple affected members. Central to our approach is the family unit which serves as a reference to identify causal genotype-phenotype correlations. Screening a multitude of 10 overlapping phenotypes revealed disease associated and co-segregating variants in 12 families. These rare or novel protein altering mutations cluster predominantly in genes (NOTCH1, ARHGAP31, MAML1, SMARCA4, JARID2, JAG1) along the Notch signaling cascade. This is in line with a significant enrichment (Wilcoxon, p< 0.05) of variants with a higher pathogenicity in the Notch signaling pathway in patients compared to controls. The significant enrichment of novel protein truncating and missense mutations in NOTCH1 highlights the allelic and phenotypic heterogeneity in our pediatric cohort. We identified novel co-segregating pathogenic mutations in NOTCH1 associated with left and right-sided cardiac malformations in three independent families with a total of 15 affected individuals. In summary, our results suggest that a small but highly pathogenic fraction of family specific mutations along the Notch cascade are a common cause of LVOTO.

Author Summary

Left-ventricular outflow tract obstructions comprise a group of developmental heart disorders that are genetically and phenotypically heterogeneous, with no single gene accounting for the majority of cases. In order to identify mutations contributing to disease, we selected patients from 51 families with a history of congenital cardiac malformations. We interrogated the entire coding sequences of 106 patients and identified a small but highly pathogenic fraction of mutations that are likely to contribute to disease in 12 families (23.5%). Furthermore, we present a strategy for identifying candidate mutations based on familial segregation in a genetically heterogeneous disorder.

Introduction

Left-ventricular outflow tract obstructions (LVOTO) comprise a group of cardiac malformations that restrict blood flow in the left portion of the heart. This heterogeneous subclass of cardiac malformations is commonly associated with aortic valve disease and early onset aortopathy, including severe stenosis and aortic dilation. These complications often result in a high disease burden later in life manifested by thoracic aortic aneurysm and valve replacement. Despite recurrent clustering of LVOTO traits among family members and a strong hereditary component [1,2], the underlying genetic cause of disease remains largely enigmatic [3]. Recent whole-exome sequencing and genotyping approaches revealed single nucleotide variants (SNVs) and copy number variations (CNVs) in genes that play an important role during the course of early outflow tract development and in chromatin modifications [4,5]. The list of known CHD disease genes is rapidly expanding and can be broadly defined by three functional groups: transcriptional regulators, cardiac structural proteins and signaling components. Disrupting mutations in critical factors regulating cardiac gene expression, such as NKX2-5, TBX5 and members of the GATA gene family (GATA4-GATA6) have been identified in family based studies for congenital heart disease (CHD) [69]. In addition, rare missense mutations in structural proteins of the cardiac muscle, including FLNA and MYH6 have been associated with single gene disorders and isolated cases of CHD [10]. However, the major fraction of patients with CHD cannot be solely explained by mutations in known disease genes [3]. Recent studies suggest that multiple genes in conserved signaling pathways mediating important cues for crucial processes during early embryonic development contribute to disease. Both familial and sporadic occurrences of LVOTO have been associated with mutations in genes of the Notch-signaling cascade [11,12]. Multifaceted Notch signaling plays a crucial role in cardiac cell fate regulation and orchestrates the morphogenesis of cardiac chambers and valves [1315]. A recent study has suggested that NOTCH1 haploinsufficiency alters specific gene networks affecting valve development and osteogenic factors which in turn result in aortic valve disease [16]. This is consistent with the identification of the NOTCH1 ligands JAG1 and DLL4 which have been shown to cause Alagille syndrome and aortic valve disease in human and animal models [1719]. However, the high functional redundancy of the Notch signaling cascade and the multitude of overlapping phenotypes between interacting genes suggest a certain level of locus heterogeneity associated with disease. Here, we report our results from a family based approach that aims to decipher the underlying genetic heterogeneity of LVOTO traits in a pediatric cohort including 182 patients from 51 families of French Canadian origin. Central to our approach is the family unit which serves as a reference to identify causal genotype-phenotype correlation.

Results

Familial based whole-exome sequencing of CHD

Whole-exome sequencing was performed in 182 individuals from 51 families with a strong heritable history of congenital heart disease (CHD) on the SOLiD 5000xl and Illumina HiSeq2000 platform (Fig 1a). The average read depth for the targeted platforms was 82x with 78% of targeted regions covered at greater than 20x. The ethnicity of our cohort was evaluated using principal component analysis and revealed that all except 1 patient, who was removed from subsequent analysis, clustered closely with individuals of European descent (Fig 1b) [20]. We evaluated the overall burden of co-segregating disease mutations for protein coding genes in our families instead of performing family-based association analysis due to the modest sample size and phenotypic heterogeneity of our cohort [21]. Among all 182 samples, 72,332 potential deleterious coding SNVs were identified of which 1245 were loss-of-function SNVs, including splice site disrupting variants, stop-gain mutations, frameshift indels and 71087 non-synonymous SNVs (Fig 2). In order to identify candidate deleterious variants, we applied four prediction algorithms for variant pathogenicity (SIFT, Poly-Phen-2, Mutation Taster, Mutation Assessor) and the CADD method [22]. All non-synonymous variants predicted to be damaging by at least two algorithms were included for further downstream analysis. We then filtered for rare variants, based on minor allele frequencies from the Exome Aggregation consortium (ExAC) ExAC Total MAF < 0.1%) and for variants which co-segregated with disease within families. This family-based approach that includes disease-segregation among multiple affected as filtering criteria, significantly reduced the number of potential pathogenic variants per patient in our family cohort (S1 Fig). As a next step, we prioritized candidate genes among the remaining 321 loss of function and 13,501 non-synonymous variants under the assumption that an excess of rare deleterious variants (rdSNVs) in genes intolerant to deleterious mutations would be more likely to affect gene function and contribute to disease in our cohort.

thumbnail
Fig 1. Overview of the study cohort.

(A) Prevalence and phenotypes associated with CHD in the general population and in the presented cohort. The numbers highlighted in grey indicate the prevalence of CHD phenotypes in 1 million live births according to Fahed et al. [3]. The graph highlighted in blue represents the distribution of overlapping CHD phenotypes in our cohort, which show a similar distribution of what is expected from the general CHD population. (B) Principal component analysis of individuals from the French Canadian pediatric cohort presented in this study with 2504 individuals from the 1000 Genomes phase 3 release for the five human super populations (AFR: purple, AMR: red, EUR: blue, EAS:green, SAS:yellow). Study subjects (black dots) and controls (grey dots) predominantly cluster with individuals of European descent (blue dots).

https://doi.org/10.1371/journal.pgen.1006335.g001

thumbnail
Fig 2. Variant filtering criteria and candidate gene prioritization.

Variants were called using the GATK pipeline as outlined in the method section. Several stringent quality and allele filtering thresholds were applied prior to filtering for allele frequency and co-segregation among family members. Annotation and filtering of coding SNVs was performed using the ANNOVAR pipeline for RefSeq gene annotations [65].

https://doi.org/10.1371/journal.pgen.1006335.g002

Enrichment of deleterious variants in highly conserved genes

To reduce the noise of our enrichment analysis, we further selected genes based on the Residual Variation Intolerance Score (RVIS) statistic [23]. Fig 3a highlights the enrichment of co-segregating rdSNVs in genes towards the lower and higher end of the genome-wide genic intolerance distribution based on their RVI scores. This unbiased enrichment approach for rare variants segregating with disease revealed 14 candidate genes showing an enrichment of rdSNVs (z-scores > 2) in our cohort among the top 10% most intolerant genes genome-wide (Fig 3a, S1 Table). Among these are two promising candidate genes, NOTCH1 and KMT2D harboring rdSNVS co-segregating with disease in multiple families. NOTCH1 is associated with aortic valve disease [11,12] whereas variants in KMT2D cause the pediatric disorder Kabuki syndrome with multiple congenital malformations including aortic valve disease [2426]. Overall, we observed a significant excess (p< 0.01, Wilcoxon) of co-segregating variants among highly conserved genes (RVIS < 10th percentile) in our data set (S2 Fig). We used the Gene Set Enrichment Analysis (GSEA) to investigate the potential biological relevance of a higher genetic burden in conserved genes for 50 hallmark gene sets from the Molecular Signatures Database [2729]. Hallmark gene sets represent well-defined biological states that display coherent expression and were considered significantly enriched with a FDR multiple test correction p value of < 0.05. We found significant enrichment in the two hallmark gene sets, “myogenesis” and “epithelial-mesenchymal transition” (Fig 3b). The highly conserved genes related to “myogenesis” and “epithelial-mesenchymal transition” are likely to play a role in disease etiology due to their crucial role during cardiac development [30,31]. Contrary to the enrichment of co-segregating rdSNVs in genes intolerant to deleterious variants, an excess of rdSNVs was observed in genes with a high genome-wide tolerance of common variants (RVIS > 90th percentile) (S2 Fig). This is to be expected, since genes on this extreme end of the RVIS distribution include frequently mutated loci such as MUC6 and MUC16 (Fig 3a) that accumulated an excess of low frequency variants due to their length and exon number which makes them unlikely contributors to CHD [3234]. Finally, mutational burden testing was performed on highly conserved candidate genes (RVIS < 10th percentile) in order to determine the contribution of individual candidate genes to disease in our cohort, based on collapsing co-segregating rdSNVs into a single gene score. We compared 106 affected individuals from our family cohort with 193 non-consanguineous population controls. The control samples sequenced at the Baylor-Hopkins Center for Mendelian Genomics were selected from the Mendel cohort without a history of cardiovascular disease and cluster predominantly with our samples of French Canadian origin (Fig 2b) [35]. While we did not observe a significant enrichment of rdSNVs in a single candidate gene, testing for novel mutations revealed a significant burden in NOTCH1 (novel alleles 0/193, novel alleles in LVOTO cohort 3/106, p = 0.0462, Fisher exact test one tailed). This is in line with an earlier association of NOTCH1 with bicuspid aortic valve disease highlighting a significant excess of private mutations in the locus [36]. After gene prioritization, we retained 53 candidate genes based on the following criteria: harboring co-segregating variants observed in at least one family, pathway association with the hallmark pathways “myogenesis”, “epithelial-mesenchymal transition” and “chromatin modification” for downstream analysis (S3 Fig)

thumbnail
Fig 3. Candidate gene prioritization in family based whole exome sequencing.

(A) Cumulative sum of unique and rare (ExAC <0.1%) deleterious variants co-segregating with disease for each gene in the whole exome dataset against the genome-wide RVI score percentiles. The red line indicates a transformed z-score thresold cutoff (p-value < 0.01), highlighting genes which have acquired more than 10 rare deleterious variants in the whole-exome dataset (z-score > 2). Candidate genes NOTCH1 and KMT2D in the top 10% of intolerant genes as well as likely false positive MUC16 and MUC6 in the top 10% of tolerant genes, are highlighted in red. (B) Pathway enrichment of genes among the 10% RVI percentile which harbor two or more rdSNVs in 106 whole-exome sequenced probands. Ranked Gene Set Enrichment analysis was performed on z-score based gene ranking with the weighted parameter and 1000 permutations for 50 BROAD derived hallmark gene sets.

https://doi.org/10.1371/journal.pgen.1006335.g003

Detailed analysis of families with NOTCH1 mutations

In the next step of our family-based approach, we performed a detailed analysis of affected family members in three families sharing novel and deleterious mutations in NOTCH1 that are likely to contribute to disease (Fig 4a, S2 Table). Phenotypic analysis of affected family members revealed the absence of syndromic disease and indicated a high rate of cardiac valve anomalies and vascular obstructions (Table 1). Two novel NOTCH1 nonsense mutations in two independent families were identified, both of which are located in the extracellular domain of the protein that has been associated with aortic valve disease [11] (Fig 4b). Both stop-gain mutations (Family 1: c.C3765A:p.C1255*; Family 2: c.C2439G:p.Y813*) are located in the N-terminal EGF domain repeats 14–36 of NOTCH1, potentially leaving the ligand binding site of the protein intact if escaping nonsense-mediated decay. The affected mother in family 1 (Family1, II-5) (Fig 4) had a bicuspid aortic valve and severe aortic stenosis. She is a mutation (c.C3765A:p.C1255*) carrier and had been remarried after conceiving two siblings who died of Tetralogy of Fallot (TOF) shortly after birth. Unfortunately, DNA was not available for genetic characterization of the variant in these patients. The affected half-brother had valve dysfunction and a bicuspid aortic valve, which indicates a strong penetrance of the Mendelian variant within this family. The inheritance patterns of valve disease in family 1 highlight the importance of different genetic backgrounds for the expressivity of the associated phenotype. While members of family 1 and 2 had primarily left-sided cardiac malformations, affected members in family 3 suffered primarily from right-sided cardiac defects. We identified two highly conserved NOTCH1 missense mutations in cis in family 3, of which one is novel (c.G578A:p.G193D) and one is rare (7/73804 ExAC alleles, c.G3860A:p.R1287H) in the general population. Both mutations segregate with disease on the same haplotype with TOF or ventricular septal defects in five affected family members. The novel mutation c.G578A:p.G193D is predicted to be highly deleterious by all major algorithms (SIFT score = 0, Polyphen-2 score = 1, CADD score = 32) [22,37,38]. Both mutations resided in the EGF repeat domains 4 and 33 and are located in the extracellular NOTCH1 domain which mediates ligand binding to JAG1 and DLL4 [39]. In order to gain additional evidence of disease association in family 3, we performed linkage analysis on the extended family (9 individuals). The analysis supported the whole-exome based association with NOTCH1 with a genome-wide linkage signal of LOD > 1.5 at four genomic loci (chr1, chr9, chr10 and chr15) (S4a Fig). When only considering rare co-segregating variants under the linkage peaks, solely NOTCH1 remains as a potential candidate gene for this family (S4b Fig). While protein disrupting mutations in NOTCH1 are likely to contribute to disease, interpreting the role of additional deleterious mutations in genes of uncertain function that may act epistatically remains challenging in a family based context (S2 Table). Apart from the highly pathogenic variants in NOTCH1, we identified a family trio with a private co-segregating nonsense mutation in ARHGAP31 (Family IV: c.A4222T:p.K1408*) in our cohort (S5 Fig, S2 Table). ARHGAP31, a Rho GTPase and specific regulator of the RAC1/CDC42 axis has been associated with valve disease and is strongly and specifically expressed in the course of early heart development in mouse [40] and in chicken (S6 Fig). The father, carrying the loss of function variant in ARHGAP31 in family 4 (Family 4, I-1) (S5 Fig) was diagnosed with aortic valve disease, manifested as aortic regurgitation, bicuspid aortic valve and a slight dilation of the ascending aorta. His son had been diagnosed with aortic stenosis early in life, associated with a bicuspid aortic valve, and developed a dilation of the ascending aorta later in life. Both patients showed no history of syndromic disease. Notably, NOTCH1 and ARHGAP31 mutation carriers overlap in their phenotypic spectrum, showing a clustering of valvular phenotypes and aortopathies. Furthermore, ARHGAP31 and NOTCH1 have both been associated with Adams-Oliver syndrome (OMIM#100300, OMIM#616028), which is associated with severe cardiac malformations such as valvular defects and TOF [40,41]. These findings suggest a certain degree of locus heterogeneity in our cohort.

thumbnail
Fig 4. NOTCH1 mutations in three independent families.

(A) Pedigrees of families harboring high impact NOTCH1 mutations. Colors represent the different phenotype associations. The (+/-) symbols indicate mutation carrier status. (B) Schematic representation of the NOTCH1 locus. Clustering of protein truncating mutations is observed in the extracellular part of the protein in the EGF domains 1–36.

https://doi.org/10.1371/journal.pgen.1006335.g004

thumbnail
Table 1. Clinical features of three families carrying pathogenic NOTCH1 mutations.

https://doi.org/10.1371/journal.pgen.1006335.t001

Targeted re-sequencing of candidate genes associated with CHD

To test the hypothesis that multiple candidate loci contribute to disease and eliminate gene coverage biases in our 53 previously identified candidate genes, we performed Nimblegen targeted re-sequencing for 153 patients from our initial whole-exome sequencing cohort in a trio design which included 82 affected individuals and 71 unaffected family members (Fig 1). We achieved an average coverage of 71.4x and 99% of the target exons were covered at 20x. Principal component analysis of our candidate genes against the ExAC dataset revealed that several genes including MYH6, KMT2D, EP300 and EP400 show an enrichment of rare private mutations despite their high intolerance to common mutations (RVIS: MYH6 = -2.78, KMT2D = -5.29) in the ExAC cohort. This is also reflected by an enrichment of rare benign mutations in these intrinsically variable genes (S7 Fig) based on ClinVar annotation. Despite the significant reduction of benign mutations by applying stringent allele filtering thresholds (p = 0.021; Fisher exact test), it remains a challenging task to discriminate between causative and non-causative missense mutations in common and often asymptomatic diseases of the aortic valve, such as BAV. For this reason, we used the recently developed Combined Annotation Dependent Depletion (CADD) framework to characterize and rank co-segregating mutations in our candidate genes [22]. This framework provides a genome-wide estimate of protein altering mutations and can be used as a meta-annotation tool to rank variants according to their pathogenicity scores. Based on the unbiased CADD score statistics, we observed a significant (p< 0.05, Wilcoxon) enrichment of deleterious mutations in patients compared to unaffected family members and 193 population controls from the Mendel dataset (Fig 5a). This enrichment in variant pathogenicity in our family cohort is due to a small fraction of variants exclusively observed in patients exceeding CADD scores >20 and ranking among the top 1% most pathogenic mutations genome-wide (p < 0.05, chi-square, Fig 5b). In contrast to the enrichment of pathogenic variants in patients, we did not observe an excess of rare or novel variants in cases when compared to controls (p > 0.05, chi-square, Fig 5b). These results are consistent with the notion that natural selection removes rare protein altering mutations from the general population and implies that only a small but highly pathogenic fraction of familial transmitted mutations contributes to disease in our cohort.

thumbnail
Fig 5. Association between variant pathogenicity and allele frequencies.

(A) Variant pathogenicity based on scaled CADD scores is significantly higher in familial cases compared to unaffected family members (p < 0.05, Wilcoxon-rank sum test) and controls from the Mendel dataset (p < 0.01, Wilcoxon-rank sum test). (B) This enrichment of variant pathogenicity is due to a small fraction of variants exclusively observed in cases exceeding CADD scores >20 (p < 0.05, chi-square test). In contrast, there was no significant enrichment observed for rare (ExAC MAF < 0.1%) or novel deleterious variants in cases compared to unaffected family members or population controls.

https://doi.org/10.1371/journal.pgen.1006335.g005

Excess of inherited missense variants along the NOTCH signaling axis

To assess whether highly ranked pathogenic mutations are enriched in disease relevant gene sets, we compared their pathogenic burden for 50 hallmark gene sets in patients and controls using the Wilcoxon rank sum statistic. We found a significant enrichment of CADD scores in members of the Notch signaling and the previously highlighted myogenesis gene sets (p < 0.05, FDR adjusted Wilcoxon) (S3 Table). This enrichment was due to mutations residing in genes related to both pathways, including NOTCH1, MAML1, JAG1, MYH6, ARHGAP31 and EPHB4 with CADD scores > 20. Finally, by ranking mutations based on their predicted pathogenicity we were able to exclude mutations in intrinsically variable genes and identified the small but highly pathogenic fraction of transmitted missense mutations that are likely to contribute to CHD in 12 families of our 51 families (S3 Table). Notably, our highest ranked disease-segregating mutations clustered predominantly in genes along the Notch signaling axis (Table 2). We identified two families with mutations in MAML1, a transcriptional co-activator of NOTCH1 involved in outflow tract defects [42,43]. Both rare mutations with CADD scores > 20 (c.G2129A:p.R710Q; c.G913A:p.A305T) segregated with aortic valve disease in five affected patients in both families, including a distantly related cousin. In addition, a family trio with a private mutation (c.G1540A:p.G959A) in JARID2 was identified which segregated with bicuspid aortic valve in the index case and his father. JARID2 has been reported to cause outflow tract malformations in a mouse model and regulates NOTCH1 expression via histone modifications in the developing heart [44]. This is in line with the identification of a patient with an overlapping phenotype and a compound heterozygous mutation (maternal: c.C3830T:p.P1277L/paternal: T4211G:p.V1404G) in the Notch signaling repressor SMARCA4 (BRG1). This analysis is however insufficient to conclusively assess novel gene associations with CHD based on family association alone, especially in the light of emerging evidence supporting a role of deleterious mutations inherited from unaffected family members [45].

thumbnail
Table 2. Candidate genes in familial LVOTO cluster predominantly along the Notch signaling axis.

https://doi.org/10.1371/journal.pgen.1006335.t002

Sensitized genetic background in a family with pleiotropic CHD

The CADD based pathogenic ranking revealed a compound heterozygous mutation in MYH6 in a family with pleiotropic congenital heart disease (S8 Fig). The index case in Family 5 (Family 5, III-3) recessive for rare maternal and paternal MYH6 mutations (maternal:c.G5653A:p.E1885K/paternal: c.G1482A:p.M494I) died shortly after birth of hypoplastic left heart syndrome. This is consistent with a previously described disease association of recessive MYH6 mutations in two patients with HLHs [46]. Notably, four affected family members in Family 5 are carriers of the paternal mutation that co-segregated in incomplete penetrance with aortopathy, septal defects and bicuspid aortic valve. This finding overlaps with a study, reporting incomplete penetrant autosomal dominant MYH6 mutations in a family with an identical clustering of LVOTO subtypes [47]. This instance highlights the complexity of assessing causal genotype-phenotype correlations in families with CHD where the deleterious missense mutation with a higher pathogenicity in the index case (maternal CADD = 34/paternal CADD = 17.7) is inherited from the maternal branch of the family without a prevalent history of disease (S8 Fig).

Discussion

In this study, we performed whole-exome and targeted candidate gene sequencing for a cohort of families with LVOTO in order to dissect the underlying genetic architecture of CHD. We identified a clustering of rare co-segregating mutations in highly conserved genes along the Notch signaling axis.

NOTCH1 mutations in LVOTO

While both gain and loss of function mutations have been reported for NOTCH1 [11,12,48], identification of families with multiple affected members carrying novel nonsense mutations remains rare in the context of congenital heart disease. In our cohort, we observed a mutational clustering in the extracellular NOTCH1 terminus in three families with 15 affected mutation carriers. This supports the notion that disruption of the cytosolic domain, but not the ligand binding part of the protein contributes to valve disease. Our findings are consistent with recent studies highlighting that the disruption of NOTCH1 downstream signaling can affect epithelial-mesenchymal transition in the course of outflow tract development [30,49]. During this developmental process, that is enriched for rare deleterious mutations in our LVOTO cohort, endocardial cells detach to become a migratory mesenchyme which in turn is crucial for the proper formation of cardiac valves [30]. Notably, we did not observe a stringent genotype-phenotype correlation with NOTCH1 deficiency and LVOTO subtypes across families in our cohort. The presence of right-sided outflow tract obstructions in families 1 and 3 indicates a degree of phenotypic heterogeneity for this locus. This is consistent with multiple disease associations of NOTCH1 with right-sided lesions reporting that up to 18% of NOTCH1 mutation carriers show right sided cardiac malformations [11,50]. The phenotypic expansion in our family cohort is intriguing due to the previous association of NOTCH1 and ARHGAP31 with Adams-Oliver syndrome (AOS) [40,41]. When analyzing the molecular pathways disrupted in AOS and LVOTO, it seems likely that the specific CDC42/RAC1 regulator ARHGAP31, which is highly expressed in the developing mouse heart [40], and the Notch pathway converge in a common signaling route that is critical for outflow tract development. Notably we did not observe characteristics of Adams-Oliver syndrome in patients with NOTCH1 nonsense mutations but one incidence of aortic valve disease.

Enrichment of deleterious mutations along the NOTCH signaling cascade

Furthermore, chromatin remodeling is an important mechanism in neural crest cells in the process of valve cushion formation and required for outflow tract formation. This process is mediated by several of the candidate genes along the Notch signaling axis identified in this study including SMARCA4 JARID2 and MAML1. We speculate that the disruption of chromatin modifiers, such as MAML1, which functions as a transcriptional co-factor for NOTCH1, might predispose patients for the observed phenotype of outflow tract defects in our cohort. Targeted disruption of Notch downstream signaling in neural crest cells by dominant negative MAML1 (DNMAML) results in similar outflow tract malformations in an animal model of disease [43]. This is in line with the tissue specific role of transcriptional Notch repressors JARID2 and SMARCA4 (BRG1). Depletion of Brg1 in endothelial cells frequently results in bicuspid aortic valve in mice and endothelial JARID2 is required for normal cardiac development [44,51,52]. These findings support the role of locus heterogeneity in aortic valve disease where disruption of functionally related members of the same pathway can result in similar phenotypes in mice and men. This is emphasized by the identification of a highly pathogenic and rare disease segregating variant in EPHB4. In a mouse model, Ephb4 loss of function phenocopies the effect of Notch1 mutations resulting in similar aortic malformations [53].

Sensitized genetic background in CHD

Our study highlights that familial segregation and weighing a variant in favor of pathogenicity can serve as an important resource for discriminating potential disease causing variants in CHD in the absence of large sample sizes. Stringent allele filtering thresholds (ExAC < 0.1% MAF) and CADD score based pathogenicity ranking (scaled CADD scores > 15) are particularly useful in analyzing intrinsically variable genes such as KMT2D and MYH6. Such filtering allowed us to identify two recessive MYH6 mutations in a family with pleiotropic cardiac malformations including HLHs, septal defects and aortopathy. Dominant MYH6 mutations are known to have a variable penetrance and despite the absence of MYH6 expression in the outflow tract, impaired vascular function and decreased blood flow have been reported in an animal model [54]. This highlights how autosomal dominant transmitted missense mutations may create a sensitized genetic background which predisposes a family member to CHD despite being completely penetrant. The presence of a second mutation in the same locus or environmental perturbations would then be required to reach a disease state. This is likely the situation in the index patient with HLHs in family 5 (Family 5, III-3) (S5a Fig), where the maternal MYH6 mutation with a higher pathogenicity is inherited from the branch of the family without a prevalent history of CHD.

Limitations and conclusions

While we observed an enrichment of deleterious mutations in highly conserved genes along the multifaceted NOTCH signaling cascade, our study has limitations. First, none of our candidate loci showed a genome wide significant association with disease. This might be due to the sample size of our cohort which provides limited statistical power, the differences in sequencing enrichment kits and sequencing platforms used or more importantly the underlying complex genetic architecture of CHD.

In addition, environmental factors, such as aortic flow, or differences in the genetic background of family members were not considered in our study design. However, these factors are likely to contribute to the phenotypic heterogeneity and incomplete penetrance observed in families with CHD. Further functional assessment of the identified mutations in multiple loci, that may act epistatically, will be needed to unravel their role during the course of disease. Taken together, our study provides a broader insight in the genetic and phenotypic heterogenic landscape of CHD. We identified co-segregating mutations that are likely to contribute to disease in 12 of our 51 families (23.5%) in multiple loci, which is in line with several recent studies highlighting oligogenic inheritance patterns of non-syndromic CHD in families [45,5557]. While the number of CHD associated genes increases with each exome study, there is also a growing body of evidence that these genes converge in a complex, yet discrete network driving outflow tract and vascular development [16,58,59].

Material and Methods

Patients

Families were recruited at CHU-Sainte Justine hospital within the French Canadian population. Participants were evaluated by clinical examination, standard 12 lead electrocardiography measurements as well as two-dimensional echocardiography. Written and informed consent was obtained from all patients or from parents in the case of pediatric patients. The study was approved by the CHU-Sainte Justine ethical board commission. The 193 unrelated control samples used for the burden test were selected among 865 samples from the Baylor-Hopkins Center for Mendelian Genomics (BHCMG) that were not from consanguineous family, of North American descent, had no known cardiovascular phenotypes and are unrelated [35].

Whole-exome and targeted sequencing

DNA was extracted from peripheral blood using the Qiagen Gentra Puregene blood kit (QIAGEN, Toronto, Canada). 96 samples (55 affected individuals) were sequenced using standard library preparation protocols with the Agilent SureSelect 50Mb exome enrichment kit and subsequently subjected to 150 base pair, paired-end sequencing on the Illumina HiSeq2000 platform at the McGill Genome Center. Post-quality control reads were aligned to the reference human genome version 19 (hg19) using bwa [60]. A total of 86 samples (51 affected individuals) were sequenced using the TargetSeq enrichment kit developed for the SOLiD5500xl sequencing platform at CHU-Sainte Justine hospital. Sequencing was performed using bar coded multiplex runs of twelve samples on eight flowchips. Mapping was performed using LifeScopeTM 2.0 Genomic Analysis Software, on the hg19 genome builds. Nimblegene target re-sequencing was performed using a custom designed SeqCap panel for candidate genes and sequenced on the Illumina HiSeq2000 platform. Aligned reads were subsequently processed using Picard (http://picard.sourceforge.net) duplicate removal and Genome Analysis Toolkit (GATK v3.2), INDEL realignment, base quality score recalibration and SNP and INDEL discovery using Haplotypecaller according to GATK best practices [61]. The 193 controls from the Mendel Genomics cohort were sequenced on the Illumina HiSeq2000 platform and aligned using the GATK pipeline as described elsewhere [62].

Variant prioritization and quality control

Variants were filtered based on Exome Aggregation Consortium (ExAC) allele filter thresholds for the general population [63]. Variants were prioritized on co-segregation patterns in pedigrees and absence of variants in control individuals within the dataset for which echocardiograms and ECGs were available. Cumulative sums of co-segregating rdSNVs were computed for all protein coding genes based on RefSeq gene annotation and transformed to z-scores. These counts were summed for 10 equal sized successive bins for the Residual Variation Intolerance Scores based on the ExAC population dataset. Significance was assessed under the null hypothesis that there should be no significant relationship between gene-based rdSNVs count and RVI scores (RVIS). In the presence of co-segregating rare mutations in genes that are intolerant or tolerant to mutations, larger counts in the low or high RVI score bins towards the left and right side of the distribution were tested for enrichment using the Wilcoxon test statistic. We further assessed the departure from the null hypothesis by evaluating the significance of the cumulative distribution by transforming z-scores to p-values. Adapted gene set enrichment analysis was performed on z-score based gene ranking for co-segregating rdSNVs with the weighted parameter and 1000 permutations against 50 hallmark gene sets from the Molecular Signatures Database [28,29]. Population stratification and IBD analysis was performed for whole-exome and target enrichment variant calls in order to exclude sample duplications and mismatches within family pedigrees. Principal components analysis on joined variant calls from 1000 genomes, study subjects and control samples from the Mendel cohort was performed using the SNPRelate package (version 1.0.1). Common variants (MAF > 5%). for 2504 individuals from the 1000 Genomes dataset phase 3 release and study subjects were plotted in the R software environment. All variants were LD pruned in SNPrelate with an LD threshold of 0.2 to produce a set of SNPs in approximate linkage equilibrium. PCA plots using the first two principal components were created using the ggplot2 package in R [64]. Prediction on variant deleteriousness for SIFT, Poly-Phen-2, Mutation Taster, Mutation Assessor and CADD scores were obtained through the ANNOVAR [65] annotation toolset for RefSeq gene annotation models.

Variant validation

All prioritized variants for the presented families were validated using PCR-based bidirectional Sanger sequencing using BigDye 3.1 chemistry on an ABI 3130xl genetic analyzer (Applied Biosystems) to confirm mutations detected by whole exome sequencing. PCRs for family members were performed using primer pairs in S4 Table. Electropherograms were aligned to the NCBI build 37 reference sequence of the respective candidate locus and analysed using SeqMan Pro, which is part of the Lasergene suite (DNASTAR).

Linkage analysis

DNA from whole blood from drawn from family 3 was used for genotyping on the Infinium OmniExpressExome-8 Bead Chip in order to validate the whole-exome sequencing association with NOTCH1. Genotypes were called with 99.8% accuracy. Raw genotypes were filtered using plink [66]. LD based pruning was applied using a window size of 50 SNPs and a total of 24551 markers were used for linkage analysis using Superlink Online [67]. Multipoint analysis was performed in Superlink using an autosomal-dominant mode of inheritance with a Disease Frequency of 0.01 and a MOI of 0, 0.99, 0.99 (S2 Fig). Results from linkage analysis were extracted from Superlink web-based output and subsequently visualized in Rstudio using the ggplot2 package [64].

Web resources

The URLs for data presented herein are as follows:

1000Genomes, www.browser.1000genomes.org

RVIS, www.genic-intolerance.org/download.jsp

ANNOVAR, www.annovar.openbioinformatics.org

dbSNP, www.cbi.nlm.nih.gov./projects/SNP

Exome Aggregation Consortium (ExAC), www.exac.broadinstitute.org

Gene Set Enrichment analysis (GSEA) tool, www.broadinstitute.org/gsea

Broad derived hallmark gene sets, www.broadinstitute.org/gsea/msigdb/collection

Combined Annotation Dependent Depletion scores, www.cadd.gs.washington.edu

Supporting Information

S1 Fig. Candidate gene filtering in family based whole exome sequencing.

Boxplots comparing distribution of remaining deleterious (missense and nonsense) variants in whole-exomes of patients for different ExAC frequency filtering thresholds. Co-segregation in multiple affected family members significantly reduced (p < 0.01, Wilcoxon) potential candidate variants independent of allele filtering thresholds

https://doi.org/10.1371/journal.pgen.1006335.s001

(TIF)

S2 Fig. Excess of deleterious variants in highly conserved genes.

Excess of rdSNVs along the upper and lower 10th percentile of the genome-wide RVI score scale (p < 0.01, Wilcoxon) for rare variants (ExAC MF < 0.1%).

https://doi.org/10.1371/journal.pgen.1006335.s002

(TIF)

S3 Fig. Candidate genes retained from whole-exome sequencing.

Candidate genes retained after whole-exome variant filtering and gene prioritization based on co-segregation and pathway analysis. Candidate genes are grouped in three broad functional categories based on pathway and literature association.

https://doi.org/10.1371/journal.pgen.1006335.s003

(TIF)

S4 Fig. Genome wide linkage analysis results for Family 3.

(A) Each panel represents an autosomal chromosome. LOD scores above 1.5 are colored red. LOD = log-of-odds, cM = centimorgan. (B) List of all rare (ExAC MAF < 0.1%), deleterious variants co-segregating with TOF from 3 whole-exome sequenced probands (Family 3: III-4, IV-2, V-1) across three generations. The linkage peak on chromosome 9 (chr9:137352393–140964937) overlaps with co-segregating pathogenic mutations in NOTCH1.

https://doi.org/10.1371/journal.pgen.1006335.s004

(TIF)

S5 Fig. Pedigree of Family 4 with co-segregating ARHGAP31 mutation.

(A) Pedigree of family 4 harboring a protein truncating mutation in ARHGAP31. Colors represent the different phenotype associations. The (+/-) symbols indicate mutation carrier status. (B) Clinical features of patients carrying the stop-gain mutation (c.A4222T:p.K1408*) in ARHGAP31.

https://doi.org/10.1371/journal.pgen.1006335.s005

(TIF)

S6 Fig. ARHGAP31 expression pattern during embryo development in Gallus gallus.

Each panel represents in situ hybridization pictures for ARHGAP31 for different stages (9, 15, 19) in the course of chicken development obtained from the online repository GEISHA (www.geisha.arizona.edu) The expression pattern observed in chicken corresponds to a strong and specific expression of ARHGAP31 in the developing heart and vasculature. This strong expression pattern overlaps with ARHGAP31 expression in the developing mouse heart as reported by Southgate et al [40].

https://doi.org/10.1371/journal.pgen.1006335.s006

(TIF)

S7 Fig. Genetic Heterogeneity in candidate genes.

(A) Principal component analysis of candidate genes against the ExAC dataset for rare stop-gain (PC1) and missense variants (PC2). Genes which have acquired a higher frequency of rare variants such as KMT2D (MLL2), EP400 and MYH6 cluster outside the majority of genes. (B) The ClinVar annotation [68] of candidate loci reflects the distribution of rare deleterious and benign mutations in genes which have acquired an excess of rare variants.

https://doi.org/10.1371/journal.pgen.1006335.s007

(TIF)

S8 Fig. Pedigree of Family 5 with compound heterozygous MYH6 mutations.

(A) Pedigree of family 5 harboring a compound heterozygous mutation in MYH6. Notably, the paternal variant p.M494I shows incomplete penetrance at II-2 and leads in combination with the maternal variant p.E1885K to a more severe phenotype (Hypoplastic left heart). Colors represent the different phenotype associations. The (+/-) symbols indicate mutation carrier status. (B) Clinical features of MYH6 mutation carriers.

https://doi.org/10.1371/journal.pgen.1006335.s008

(TIF)

S1 Table. Excess of rare deleterious variants in highly conserved genes.

Genes among the top 10% RVIs genome-wide scale are highlighted harboring an excess of rare (ExAC < 0.1%) deleterious variants (>10 rdSNVs). Z-scores are calculated based on variant distribution for 106 affected individuals and transformed in p-values.

https://doi.org/10.1371/journal.pgen.1006335.s009

(XLS)

S2 Table. Whole-exome filtering strategies for the identification of pathogenic variants in Families 1–4.

The tables depict the individual filtering strategies for Families 1–3 with novel pathogenic protein disrupting mutations in NOTCH1 and ARHGAP31 (Family 4). Filtering was performed based on allele frequency cutoffs (ExAC MF < 0.1%) and segregation with disease among multiple affected family members.

https://doi.org/10.1371/journal.pgen.1006335.s010

(XLS)

S3 Table. CADD ranked variant pathogenicity in candidate genes.

Co-segregating, private variants sorted based on scaled CADD score pathogenicity in targeted re-sequenced candidate genes.

https://doi.org/10.1371/journal.pgen.1006335.s011

(XLS)

S4 Table. Variant pathogenicity in Broad derived hallmark gene sets.

The table lists the excess of pathogenicity based on scaled CADD scores in affected individuals for candidate loci. Significance was determined by comparing variants in affected individuals and unrelated and unaffected parents using Wilcoxon rank-sum tests with subsequent FDR p-value correction for multiple testing.

https://doi.org/10.1371/journal.pgen.1006335.s012

(XLS)

S5 Table. Primer pairs used for Sanger Sequencing Validation.

Primers used for Sanger sequencing validation of rare deleterious and co-segregating SNVs in the study.

https://doi.org/10.1371/journal.pgen.1006335.s013

(XLS)

Acknowledgments

We are deeply indebted to all participants and their family members for participation in this study. We would like to thank the Exome Aggregation Consortium and the groups that provided exome variant data for comparison as well as the McGill Genome center for providing the sequencing data. We thank all referring physicians and support staff for their assistance as well as Lut Van Laer for critically reading the manuscript.

Consortia

List of MIBAVA Leducq consortium members:

Bart Loeys M.D., Harry Dietz M.D., Anders Franco-Cereceda M.D., Per Eriksson Ph.D., Salah A. Mohamed Ph.D., Andrew S. McCallion Ph.D., Luc Mertens Ph.D., Lut Van Laer Ph.D., Seema Mital M.D.

URL: www.fondationleducq.org/network/mibava-mechanistic-interrogation-of-bicuspid-aortic-valve-associated-aortopathy

Author Contributions

  1. Conceived and designed the experiments: GA CP MC FW.
  2. Performed the experiments: CP MC FW.
  3. Analyzed the data: CP MC FW AP HL GA.
  4. Contributed reagents/materials/analysis tools: PK SL MT PC NS GA.
  5. Wrote the paper: CP FW GA.
  6. Provided expert advice and leadership for the research activity: BG PA MES NS.

References

  1. 1. Cripe L, Andelfinger G, Martin LJ, Shooner K, Benson DW. Bicuspid aortic valve is heritable. J Am Coll Cardiol. 2004;44: 138–43. pmid:15234422
  2. 2. McBride KL, Pignatelli R, Lewin M, Ho T, Fernbach S, Menesses A, et al. Inheritance analysis of congenital left ventricular outflow tract obstruction malformations: Segregation, multiplex relative risk, and heritability. Am J Med Genet A. 2005;134A: 180–6. pmid:15690347
  3. 3. Fahed AC, Gelb BD, Seidman JG, Seidman CE. Genetics of congenital heart disease: the glass half empty. Circ Res. 2013;112: 707–20. pmid:23410880
  4. 4. Zaidi S, Choi M, Wakimoto H, Ma L, Jiang J, Overton JD, et al. De novo mutations in histone-modifying genes in congenital heart disease. Nature. Nature Publishing Group, a division of Macmillan Publishers Limited. All Rights Reserved.; 2013;498: 220–3. pmid:23665959
  5. 5. Hitz M-P, Lemieux-Perreault L-P, Marshall C, Feroz-Zada Y, Davies R, Yang SW, et al. Rare Copy Number Variants Contribute to Congenital Left-Sided Heart Disease. Spinner NB, editor. PLoS Genet. 2012;8: e1002903. https://doi.org/10.1371/journal.pgen.1002903 pmid:22969434
  6. 6. Laforest B, Andelfinger G, Nemer M. Loss of Gata5 in mice leads to bicuspid aortic valve. J Clin Invest. American Society for Clinical Investigation; 2011;121: 2876–87. pmid:21633169
  7. 7. Schott JJ, Benson DW, Basson CT, Pease W, Silberbach GM, Moak JP, et al. Congenital heart disease caused by mutations in the transcription factor NKX2-5. Science. 1998;281: 108–11. Available: http://www.ncbi.nlm.nih.gov/pubmed/9651244 pmid:9651244
  8. 8. McCulley DJ, Black BL. Transcription factor pathways and congenital heart disease. Curr Top Dev Biol. 2012;100: 253–77. pmid:22449847
  9. 9. Basson CT, Huang T, Lin RC, Bachinsky DR, Weremowicz S, Vaglio A, et al. Different TBX5 interactions in heart and limb defined by Holt-Oram syndrome mutations. Proc Natl Acad Sci U S A. 1999;96: 2919–24. Available: http://www.ncbi.nlm.nih.gov/pubmed/10077612 pmid:10077612
  10. 10. Robertson SP, Twigg SRF, Sutherland-Smith AJ, Biancalana V, Gorlin RJ, Horn D, et al. Localized mutations in the gene encoding the cytoskeletal protein filamin A cause diverse malformations in humans. Nat Genet. 2003;33: 487–91. pmid:12612583
  11. 11. Garg V, Muth AN, Ransom JF, Schluterman MK, Barnes R, King IN, et al. Mutations in NOTCH1 cause aortic valve disease. Nature. 2005;437: 270–4. pmid:16025100
  12. 12. McBride KL, Riley MF, Zender GA, Fitzgerald-Butt SM, Towbin JA, Belmont JW, et al. NOTCH1 mutations in individuals with left ventricular outflow tract malformations reduce ligand-induced signaling. Hum Mol Genet. 2008;17: 2886–93. pmid:18593716
  13. 13. Niessen K, Karsan A. Notch signaling in cardiac development. Circ Res. 2008;102: 1169–81. pmid:18497317
  14. 14. High FA, Epstein JA. The multifaceted role of Notch in cardiac development and disease. Nat Rev Genet. Nature Publishing Group; 2008;9: 49–61. pmid:18071321
  15. 15. Penton AL, Leonard LD, Spinner NB. Notch signaling in human development and disease. Semin Cell Dev Biol. 2012;23: 450–7. pmid:22306179
  16. 16. Theodoris C V, Li M, White MP, Liu L, He D, Pollard KS, et al. Human disease modeling reveals integrated transcriptional and epigenetic mechanisms of NOTCH1 haploinsufficiency. Cell. 2015;160: 1072–86. pmid:25768904
  17. 17. Eldadah ZA, Hamosh A, Biery NJ, Montgomery RA, Duke M, Elkins R, et al. Familial Tetralogy of Fallot caused by mutation in the jagged1 gene. Hum Mol Genet. 2001;10: 163–9. Available: http://www.ncbi.nlm.nih.gov/pubmed/11152664 pmid:11152664
  18. 18. Hofmann JJ, Briot A, Enciso J, Zovein AC, Ren S, Zhang ZW, et al. Endothelial deletion of murine Jag1 leads to valve calcification and congenital heart defects associated with Alagille syndrome. Development. 2012;139: 4449–60. pmid:23095891
  19. 19. D’Amato G, Luxán G, Del Monte-Nieto G, Martínez-Poveda B, Torroja C, Walter W, et al. Sequential Notch activation regulates ventricular chamber development. Nat Cell Biol. Nature Publishing Group; 2015;18: 7–20. pmid:26641715
  20. 20. The 1000 Genomes Project Consortium. A map of human genome variation from population-scale sequencing. Nature. 2010;467: 1061–73. pmid:20981092
  21. 21. Risch N. Linkage strategies for genetically complex traits. I. Multilocus models. Am J Hum Genet. 1990;46: 222–228. http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=2912646&tool=pmcentrez&rendertype=abstract pmid:2301392
  22. 22. Kircher M, Witten DM, Jain P, O’Roak BJ, Cooper GM, Shendure J. A general framework for estimating the relative pathogenicity of human genetic variants. Nat Genet. 2014;46: 310–5. pmid:24487276
  23. 23. Petrovski S, Wang Q, Heinzen EL, Allen AS, Goldstein DB. Genic intolerance to functional variation and the interpretation of personal genomes. PLoS Genet. Public Library of Science; 2013;9: e1003709. pmid:23990802
  24. 24. Ng SB, Bigham AW, Buckingham KJ, Hannibal MC, McMillin MJ, Gildersleeve HI, et al. Exome sequencing identifies MLL2 mutations as a cause of Kabuki syndrome. Nat Genet. 2010;42: 790–3. pmid:20711175
  25. 25. Makrythanasis P, van Bon BW, Steehouwer M, Rodríguez-Santiago B, Simpson M, Dias P, et al. MLL2 mutation detection in 86 patients with Kabuki syndrome: a genotype-phenotype study. Clin Genet. 2013; pmid:23320472
  26. 26. Prakash SK, Bossé Y, Muehlschlegel JD, Michelena HI, Limongelli G, Della Corte A, et al. A roadmap to investigate the genetic basis of bicuspid aortic valve and its complications: insights from the International BAVCon (Bicuspid Aortic Valve Consortium). J Am Coll Cardiol. 2014;64: 832–9. pmid:25145529
  27. 27. Liberzon A, Subramanian A, Pinchback R, Thorvaldsdóttir H, Tamayo P, Mesirov JP. Molecular signatures database (MSigDB) 3.0. Bioinformatics. 2011;27: 1739–40. pmid:21546393
  28. 28. Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci U S A. 2005;102: 15545–50. pmid:16199517
  29. 29. Liberzon A, Birger C, Thorvaldsdóttir H, Ghandi M, Mesirov JP, Tamayo P. The Molecular Signatures Database Hallmark Gene Set Collection. Cell Syst. 2015;1: 417–425. pmid:26771021
  30. 30. Timmerman LA, Grego-Bessa J, Raya A, Bertrán E, Pérez-Pomares JM, Díez J, et al. Notch promotes epithelial-mesenchymal transition during cardiac development and oncogenic transformation. Genes Dev. 2004;18: 99–115. pmid:14701881
  31. 31. Kovacic JC, Mercader N, Torres M, Boehm M, Fuster V. Epithelial-to-mesenchymal and endothelial-to-mesenchymal transition: from cardiovascular development to disease. Circulation. 2012;125: 1795–808. pmid:22492947
  32. 32. Fuentes Fajardo K V, Adams D, Mason CE, Sincan M, Tifft C, Toro C, et al. Detecting false-positive signals in exome sequencing. Hum Mutat. 2012;33: 609–13. pmid:22294350
  33. 33. Shyr C, Tarailo-Graovac M, Gottlieb M, Lee JJY, van Karnebeek C, Wasserman WW. FLAGS, frequently mutated genes in public exomes. BMC Med Genomics. BioMed Central; 2014;7: 64. pmid:25466818
  34. 34. MacArthur DG, Manolio TA, Dimmock DP, Rehm HL, Shendure J, Abecasis GR, et al. Guidelines for investigating causality of sequence variants in human disease. Nature. Nature Publishing Group, a division of Macmillan Publishers Limited. All Rights Reserved.; 2014;508: 469–76. pmid:24759409
  35. 35. Bamshad MJ, Shendure JA, Valle D, Hamosh A, Lupski JR, Gibbs RA, et al. The Centers for Mendelian Genomics: a new large-scale initiative to identify the genes underlying rare Mendelian conditions. Am J Med Genet A. 2012;158A: 1523–5. pmid:22628075
  36. 36. McKellar SH, Tester DJ, Yagubyan M, Majumdar R, Ackerman MJ, Sundt TM. Novel NOTCH1 mutations in patients with bicuspid aortic valve disease and thoracic aortic aneurysms. J Thorac Cardiovasc Surg. 2007;134: 290–6. pmid:17662764
  37. 37. Ng PC, Henikoff S. SIFT: Predicting amino acid changes that affect protein function. Nucleic Acids Res. 2003;31: 3812–4. Available: http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=168916&tool=pmcentrez&rendertype=abstract pmid:12824425
  38. 38. Adzhubei I, Jordan DM, Sunyaev SR. Predicting functional effect of human missense mutations using PolyPhen-2. Curr Protoc Hum Genet. 2013;Chapter 7: Unit7.20. pmid:23315928
  39. 39. Luca VC, Jude KM, Pierce NW, Nachury M V, Fischer S, Garcia KC. Structural biology. Structural basis for Notch1 engagement of Delta-like 4. Science. 2015;347: 847–53. pmid:25700513
  40. 40. Southgate L, Machado RD, Snape KM, Primeau M, Dafou D, Ruddy DM, et al. Gain-of-function mutations of ARHGAP31, a Cdc42/Rac1 GTPase regulator, cause syndromic cutis aplasia and limb anomalies. Am J Hum Genet. 2011;88: 574–85. pmid:21565291
  41. 41. Stittrich A-B, Lehman A, Bodian DL, Ashworth J, Zong Z, Li H, et al. Mutations in NOTCH1 cause Adams-Oliver syndrome. Am J Hum Genet. 2014;95: 275–84. pmid:25132448
  42. 42. Nam Y, Sliz P, Song L, Aster JC, Blacklow SC. Structural Basis for Cooperativity in Recruitment of MAML Coactivators to Notch Transcription Complexes. Cell. 2006;124: 973–983. pmid:16530044
  43. 43. High FA, Zhang M, Proweller A, Tu L, Parmacek MS, Pear WS, et al. An essential role for Notch in neural crest during cardiovascular development and smooth muscle differentiation. J Clin Invest. American Society for Clinical Investigation; 2007;117: 353–63. pmid:17273555
  44. 44. Mysliwiec MR, Carlson CD, Tietjen J, Hung H, Ansari AZ, Lee Y. Jarid2 (Jumonji, AT rich interactive domain 2) regulates NOTCH1 expression via histone modification in the developing heart. J Biol Chem. 2012;287: 1235–41. pmid:22110129
  45. 45. Sifrim A, Hitz M-P, Wilsdon A, Breckpot J, Turki SH, Thienpont B, et al. Distinct genetic architectures for syndromic and nonsyndromic congenital heart defects identified by exome sequencing. Nat Genet. Nature Research; 2016; pmid:27479907
  46. 46. Theis JL, Zimmermann MT, Evans JM, Eckloff BW, Wieben ED, Qureshi MY, et al. Recessive MYH6 Mutations in Hypoplastic Left Heart With Reduced Ejection Fraction. Circ Cardiovasc Genet. 2015;8: 564–71. pmid:26085007
  47. 47. Arrington CB, Bleyl SB, Matsunami N, Bonnell GD, Otterud BEM, Nielsen DC, et al. Exome analysis of a family with pleiotropic congenital heart disease. Circ Cardiovasc Genet. 2012;5: 175–82. pmid:22337856
  48. 48. Stittrich A-B, Lehman A, Bodian DL, Ashworth J, Zong Z, Li H, et al. Mutations in NOTCH1 cause Adams-Oliver syndrome. Am J Hum Genet. 2014;95: 275–84. pmid:25132448
  49. 49. Armstrong EJ, Bischoff J. Heart valve development: endothelial cell signaling and differentiation. Circ Res. 2004;95: 459–70. pmid:15345668
  50. 50. Kerstjens-Frederikse WS, van de Laar IMBH, Vos YJ, Verhagen JMA, Berger RMF, Lichtenbelt KD, et al. Cardiovascular malformations caused by NOTCH1 mutations do not keep left: data on 428 probands with left-sided CHD and their families. Genet Med. 2016; pmid:26820064
  51. 51. Akerberg BN, Sarangam ML, Stankunas K. Endocardial Brg1 disruption illustrates the developmental origins of semilunar valve disease. Dev Biol. 2015;407: 158–72. pmid:26100917
  52. 52. Mysliwiec MR, Bresnick EH, Lee Y. Endothelial Jarid2/Jumonji is required for normal cardiac development and proper Notch1 expression. J Biol Chem. 2011;286: 17193–204. pmid:21402699
  53. 53. Krebs LT, Starling C, Chervonsky A V, Gridley T. Notch1 activation in mice causes arteriovenous malformations phenocopied by ephrinB2 and EphB4 mutants. Genesis. 2010;48: 146–50. pmid:20101599
  54. 54. Rutland C, Warner L, Thorpe A, Alibhai A, Robinson T, Shaw B, et al. Knockdown of alpha myosin heavy chain disrupts the cytoskeleton and leads to multiple defects during chick cardiogenesis. J Anat. 2009;214: 905–15. pmid:19538634
  55. 55. Priest JR, Osoegawa K, Mohammed N, Nanda V, Kundu R, Schultz K, et al. De Novo and Rare Variants at Multiple Loci Support the Oligogenic Origins of Atrioventricular Septal Heart Defects. PLoS Genet. Public Library of Science; 2016;12: e1005963. pmid:27058611
  56. 56. Granados-Riveron JT, Pope M, Bu’lock FA, Thornborough C, Eason J, Setchfield K, et al. Combined mutation screening of NKX2-5, GATA4, and TBX5 in congenital heart disease: multiple heterozygosity and novel mutations. Congenit Heart Dis. Wiley-Blackwell; 2012;7: 151–9. pmid:22011241
  57. 57. D’Alessandro LCA, Al Turki S, Manickaraj AK, Manase D, Mulder BJM, Bergin L, et al. Exome sequencing identifies rare variants in multiple genes in atrioventricular septal defect. Genet Med. Springer Nature; 2016;18: 189–198. pmid:25996639
  58. 58. Lage K, Møllgård K, Greenway S, Wakimoto H, Gorham JM, Workman CT, et al. Dissecting spatio-temporal protein networks driving human heart development and related disorders. Mol Syst Biol. Nature Publishing Group; 2010;6: 381. pmid:20571530
  59. 59. Andersen TA, Troelsen K de LL, Larsen LA. Of mice and men: molecular genetics of congenital heart disease. Cell Mol Life Sci. 2014;71: 1327–52. pmid:23934094
  60. 60. Li H, Durbin R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics. 2009;25: 1754–60. pmid:19451168
  61. 61. McKenna A, Hanna M, Banks E, Sivachenko A, Cibulskis K, Kernytsky A, et al. The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res. 2010;20: 1297–303. pmid:20644199
  62. 62. Jurgens J, Ling H, Hetrick K, Pugh E, Schiettecatte F, Doheny K, et al. Assessment of incidental findings in 232 whole-exome sequences from the Baylor—Hopkins Center for Mendelian Genomics. Genet Med. Nature Publishing Group; 2015;17: 782–788. pmid:25569433
  63. 63. Consortium EA, Lek M, Karczewski K, Minikel E, Samocha K, Banks E, et al. Analysis of protein-coding genetic variation in 60,706 humans [Internet]. bioRxiv. Cold Spring Harbor Labs Journals; 2015 Oct. pmid:27535533
  64. 64. Wickham H. ggplot2. Wiley Interdiscip Rev Comput Stat. 2011;3: 180–185.
  65. 65. Wang K, Li M, Hakonarson H. ANNOVAR: functional annotation of genetic variants from high-throughput sequencing data. Nucleic Acids Res. 2010;38: e164. pmid:20601685
  66. 66. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MAR, Bender D, et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet. 2007;81: 559–75. pmid:17701901
  67. 67. Silberstein M, Weissbrod O, Otten L, Tzemach A, Anisenia A, Shtark O, et al. A system for exact and approximate genetic linkage analysis of SNP data in large pedigrees. Bioinformatics. 2013;29: 197–205. pmid:23162081
  68. 68. Landrum MJ, Lee JM, Riley GR, Jang W, Rubinstein WS, Church DM, et al. ClinVar: public archive of relationships among sequence variation and human phenotype. Nucleic Acids Res. 2014;42: D980–5. pmid:24234437