Of mice and men: molecular genetics of congenital heart disease

Congenital heart disease (CHD) affects nearly 1 % of the population. It is a complex disease, which may be caused by multiple genetic and environmental factors. Studies in human genetics have led to the identification of more than 50 human genes, involved in isolated CHD or genetic syndromes, where CHD is part of the phenotype. Furthermore, mapping of genomic copy number variants and exome sequencing of CHD patients have led to the identification of a large number of candidate disease genes. Experiments in animal models, particularly in mice, have been used to verify human disease genes and to gain further insight into the molecular pathology behind CHD. The picture emerging from these studies suggest that genetic lesions associated with CHD affect a broad range of cellular signaling components, from ligands and receptors, across down-stream effector molecules to transcription factors and co-factors, including chromatin modifiers.


Introduction
Congenital heart disease (CHD) is the most prevalent birth defect, with a postnatal incidence of 0.8 % [1] and an approximately tenfold higher prenatal incidence [2]. CHD is a group of structural abnormalities of the heart, which include septal defects, valve defects and lesions affecting

ACTC1
Cardiac α-actinin I [124] ELN elastin I c [184] therapy for treatment of myocardial infarction. In familial cases of CHD, identification of disease genes will benefit the genetic counseling process for CHD-afflicted families. Such knowledge is particularly important for the growing population of adults with CHD, due to the high recurrence risk of certain forms of CHD [4].
In this review, we aim to summarize current knowledge in the molecular genetics of CHD, from the perspective of, disease gene identification efforts in humans, and functional analyses of disease genes in animal models.

Part I: genes associated with congenital heart disease
Below we discuss genes associated with syndromic and isolated CHD, juxtaposing studies from multiple model systems to clarify why errors in the underlying molecular machinery manifests themselves as congenital heart defects.

Genes associated with laterality defects
The heart is the first organ to break the bilateral symmetry of the developing embryo. During early embryogenesis, left-right asymmetry of the body-axis is established via intricate cross-talk amongst signaling pathways such as Notch, Nodal, Hedgehog, FGF and BMP, ultimately restricting NODAL signaling to the left side of the embryo. The nodal cilia model is the predominant model to elucidate induction of embryonic asymmetry in the developing embryo, although other models have been proposed [5]. Briefly, nodal cilia in the node of the primitive streak produce a directional fluid flow which induces leftright asymmetry by delivering morphogens to the left side of the embryo and/or by acting on mechanosensory cilia [6][7][8][9]. Ultimately, laterality cues are relayed downstream to ensure left-sided expression of the PITX2 transcription factor in the lateral plate mesoderm [10], a critical component in determining organ laterality [11]. NODAL [12,13], LEFTY2 [14], ACVR2B [15], GDF1 [16] CFC1 [17], CITED2 [18] and ZIC3 [19] have all been localized to the laterality signaling pathway. Albeit human mutations in these genes show a wide range of heart malfunctions, many cluster around laterality defects such as heterotaxy and faulty looping of the heart. Interestingly, a recent study identified mutations in Nephronophthisis-4 (NPHP4), a cilia related gene, and linked them to a variety of cardiac laterality defects such as transposition of the great arteries (TGA), atrioventricular septal defects (AvSD), double outlet right ventricle (DORv), dextrocardia and abnormal pulmonary venous return. Laterality defects of the abdominal organs were also observed. Morpholino knock down of nphp4 in zebrafish resulted in reverse orientation or faulty looping of the heart [20]. Previous studies have also connected other NPHP family members to inborn heart defects and laterality deficiencies [21,22].
The ZIC3 gene encodes a zinc finger transcription factor known to cause cardiovascular defects when mutated The list include genes, which have been associated with CHD by identification of mutations in two or more unrelated patient and/or genes where human genetic analyses are complemented with functional analyses a TF transcription factor, HC heavy chain b I isolated CHD, S syndromic CHD, HTX heterotaxy, NS Noonan syndrome, LS LeOPARD   In humans, mutations in the transcriptional co-activator CITeD2 (Cbp/p300-interacting transactivator, with Glu/ Asp-rich carboxy-terminal domain 2) are associated with laterality defects and cardiac anomalies such as septal defects and TGA [18]. Mice deficient in Cited2 die during gestation expressing partially penetrant laterality defects and fully penetrant heart defects [31, 32]. The heart anomalies include ASD, vSD, common atrioventricular canal (CAvC), DORv and IAA type B [32, 33]. Recently, Lopes et al. [31] showed that specific deletions of Cited2 in heart progenitors do not produce heart defects and that the cardiac malfunctions seen in Cited2 −/− embryos arise during the early phases of establishing the left-right body axis in close relation to NODAL signaling.

Genes encoding components of signaling pathways
Animal models have illustrated that cardiac development involves spatial and temporal coordination of a number of signaling pathways [34,35]. The identification of disease genes in syndromic and isolated CHD has confirmed the involvement of a subgroup of these pathways in human heart development, and has further contributed new information about additional pathways.
The NOTCH signaling pathway acts locally as a cellfate regulator, and as a patterning signal effector in many developmental processes. Its activity includes left-right axis partitioning and heart morphogenesis. For an extensive review on Notch signaling in cardiac development see [36].
Identification of mutations in JAG1, encoding a NOTCH signaling ligand, in patients with Alagille syndrome provided the first link between NOTCH signaling and human CHD [37,38]. Alagille syndrome (AGS, OMIM #118450) is a multisystem disorder, which involves the liver, heart, eyes, face and skeleton [39]. Approximately 90 % of these patients have cardiovascular anomalies, often presenting in the form of stenosis in the pulmonary artery branch, valvular pulmonary stenosis (PS) and tetralogy of Fallot (TOF) [40]. The majority (>90 %) of AGS cases are caused by mutations in JAG1, however select (<1 %) cases are caused by mutation in NOTCH2 [41,42]. Mice homozygous for targeted deletion of Jag1 die during embryonic development possibly due to vascular defects while heterozygous Jag1 −/+ mice display ocular defects [43].
Mutations in NOTCH1 have been identified in patients with isolated CHD [45][46][47]. Patients with NOTCH1 mutations often present malfunctions of the aortic valve. NOTCH1 signaling has been linked to endothelial-to-mesenchymal transformation (eMT), a fundamental process in the early stages of cardiac valve formation, where endocardial cells detach to become a migratory mesenchyme that forms endocardial cushions, precursors of cardiac valves. Notch1 mutant mice develop hypoplastic endocardial cushions due to impaired eMT [48]. Recently, Luna-Zurita et al. and others [48][49][50] outlined that a NOTCH1, wNT4 and BMP2 signal interplay between the endocardium and myocardium underlie valve morphogenesis. Furthermore, Bosse et al. [51] showed that compound mutant Notch +/− ; Nos3 −/− mice display an accelerated bicuspid aortic valve phenotype compared to Notch +/− and Nos3 −/− alone, suggesting an interaction between nitric oxide (NO) and NOTCH signaling in the development of the aortic valve. In the same paper, these data were further supported by in vitro data, suggesting that NO regulates Notch signaling in aortic valve interstitial cells.
Signal transduction through the RAS-mitogen activated protein kinase (MAPK) pathway can stimulate cell proliferation, differentiation, survival and metabolism. Identification of disease genes in Noonan syndrome (OMIM #163950), Costello syndrome (OMIM #218040), LeOP-ARD syndrome (OMIM #151100), Cardio-facio-cutaneous (CFC) syndrome (OMIM #115150) and a few other syndromes with distinct but overlapping phenotypes, collectively known as RASopathies (see [52] for review), have firmly established a link between the RAS-MAPK signal transduction pathway and human CHD. The RASopathies are manifested by a wide range of multisystem anomalies, including CHD. In Noonan syndrome approximately 85 % of patients have a variety of cardiac defects, most commonly including pulmonary valve stenosis, ASD and hypertrophic cardiomyopathy [53,54].

Genes encoding cardiac transcription factors
Cardiac developmental signals are conveyed to transcriptional circuits that regulate gene expression during normal heart development. At the heart of these transcriptional networks lie a set of core transcription factors many of which are associated with isolated CHD (Table 1). Transcriptional focal points include NKX2-5, GATA4 and TBX5. These transcription factors interact at cardiac promoters in synergistic fashions (see below). Their function and molecular signatures have been thoroughly described and excellently reviewed elsewhere, and will only be briefly mentioned here (see [68] for a recent review).
In humans, disease causing mutations in the homeodomain protein NKX2-5 result in a plethora of CHDs including ASD, vSD, TOF and DORv [69][70][71] Septal defects and atrio-ventricular conduction defects are commonly seen in patients with a mutated NKX2-5 gene [70]. Disruption of Tinman, the homologue of NKX2-5 in Drosophila melanogaster, results in a fruit fly devoid of the dorsal vessel, a structure analogous to the human heart [72]. Similarly, Nkx2-5 functionality is crucial in mice as homozygous mutations cause embryonic lethality due to faulty cardiac looping and insufficient myocardial differentiation during chamber formation [73,74]. Mouse studies show that Nkx2-5 gene dosage is critically important for properly regulated development of the cardiac conduction system as Nkx2-5 null mice lack the primordium of the Av node and the conduction system of heterozygous mutant embryos only contain half the normal number of cells [75]. Additionally, Pashmforoush et al. [76] generated ventricular-restricted Nkx2-5 knockout mice that display progressive complete heart block and massive trabecular muscle overgrowth.
NKX2-5 ranks high in the cardiac regulatory hierarchy and is expressed in both the first-and second heart field (SHF) [77]. Its expression is closely coordinated through GATA factors, SMAD proteins and by self-autoregulation [78][79][80][81]. Proliferation of the SHF and outflow tract (OFT) morphology is regulated by Nkx2-5 feedback repression of BMP2/SMAD1 signaling [82]. It was recently shown that JARID2, which is also implicated in OFT development, is a direct target of NKX2-5 regulation [83]. Furthermore, it has been demonstrated that Nkx2-5 interacts with Gata4 within cardiac promoters, suggesting that the proteins cooperate in the transcriptional activation of cardiac target genes [84,85]. Nkx2-5 interacts with Tbx5 in vitro and the two proteins were shown to activate a cardiac-specific Nppa promoter in a synergistic fashion [86] and Nkx2-5 cooperates with Tbx5 in development of the cardiac conduction system in vivo [87].
GATA binding protein 4 (GATA4) plays a pivotal role near the top of the transcriptional cascades that control heart development (see [88] for a recent review). In humans, the cardiac defects found in patients with interstitial deletions in 8p23.1 are attributed to haploinsufficiency of GATA4 (see below). Intragenic GATA4 mutations can also cause isolated CHDs, primarily cardiac septal defects, but PS, TOF and other defects have been reported [89-91]. In a recent paper human missense mutations in GATA4 were shown to disrupt GATA4-SMAD4 interactions in the BMP/TGF-β signaling pathway, likely causing AvSD and valve abnormalities in the affected patients [92]. embryonic development in Gata4 deficient mice is arrested at e10.5 with incorrect ventral folding, endodermal malfunctions and an inability to establish a primitive heart tube [93,94]. Correct Gata4 dosage is critically important for normal heart development, as mice homozygous for a hypomorphic allele develop CAvC, DORv and a hypoplastic ventricular myocardium [95]. Furthermore, it has also been shown that mice heterozygous for GATA4 mutations develop septation and endocardial cushion defects [90]. Heterozygous knock-in mice harbouring a Gata4 G296S mutation, previously identified in patients with septum defects and pulmonary valve stenosis, display ASD and semilunar valve stenosis [96].
GATA4 interacts with numerous transcription factors that promote cardiogenesis [88]. Direct downstream targets of GATA4 include HAND2 and MEF2C required for SHF development [97,98]. GATA4 and TBX5 proteins physically interact and this interaction is disrupted by mutations in GATA4 [89]. Furthermore, Gata4 and Tbx5 double heterozygous mice develop cardiovascular defects, which point towards a genetic interaction between the two [99].
The T-box transcription factors are important cardiac transcription factors. They are involved in fundamental cardiac developmental processes, including development of the chamber myocardium, outflow tract and the conduction system [100]. TBX1 regulates proliferation of cardiac progenitors in the SHF and haploinsufficiency of TBX1 is considered the primary cause of CHD in patients with DiGeorge syndrome (see below). TBX5 participates in regulation of gene expression in the developing chamber myocardium and conduction system [101]. Mutations in TBX5 cause Holt-Oram syndrome (OMIM #142900), a syndrome distinguished by upper limb defects and heart defects-primarily septal and conduction defects [102,103]. Tbx5 null mice possess a deformed linear heart tube and underdeveloped atria while heterozygous Tbx5 mice model heart and limb abnormalities observed in Holt-Oram syndrome, potentially explaining cardiac conduction system defects seen in these patients [101].
Human mutations in TBX20 cause aberrant valvulogenesis, septal defects, TOF and cardiomyopathy [104]. Deletion of Tbx20 in mice generates a linear heart tube which fails to loop properly and exhibits insufficient chamber formation [105]. Heterozygous Tbx20 mice show onset of dilated cardiomyopathy recapitulating some of the human defects [106]. Recent papers by Cai and co-workers [107,108] place TXB20 in the formation of the cardiac atrioventricular canal in a complex signaling network involving TBX20, TBX2 and BMP2.
In vitro, transcription factor AP-2gamma (TFAP2C) has been shown to bind the TBX20 promoter site and repress TBX20 expression [109]. Interestingly, mutations in TFAP2B causes Char syndrome (OMIM #169100) characterized by facial dysmorphism, anomalies of the fifth finger and patent ductus arteriosus (PDA) [110]. Isolated PDA has also been linked to mutations in TFAP2B [111,112] and a recent Tfap2B knock out study in mice reported phenotypes resembling the characteristics of Char syndrome [113]. TFAP2 isoforms form a complex with CITeD2, CReBBP and eP300 [114,115]. Mutations in the transcriptional coactivators CReBBP and eP300 are associated with Rubinstein-Taybi syndrome (OMIM #180849) displaying mental retardation, broad thumbs and toes, facial abnormalities, and in some cases, CHD [116,117].
Genes encoding components of the cardiac sarcomere Mutations in genes encoding cardiac structural proteins have also been connected to CHD. Several studies link mutations in the cardiac sarcomeric protein MYH6 (myosin heavy chain 6) to ASD [118][119][120]. Morpholino knock down of myh6 in the developing chicken heart implies that its functionality is required in the formation of the atrial septum [120]. Molecular regulation of MYH6 expression involves transcription factors such as GATA4 [121], TBX5 and MeF2C [122].
Other members of the contractile units in cardiovascular muscle include MYH7 and ACTC1. A mutation in MYH7 encoding myosin heavy chain 7 was shown to cause CHDs such as ebstein's anomaly and septal defects [123]. Mutations in ACTC1 encoding the human α-cardiac muscle actin can cause ASD and morpholino knock down of Actc1 causes looping and atrial septal anomalies in chicken embryos [124].
Mutations in MYH11, encoding the major contractile protein of smooth muscle cells can cause thoracic aortic aneurysm and/or aortic dissection and PDA [125]. Mice homozygous for deletion of Myh11 show a delayed closure of the ductus arteriosus [126], which is connected to the shunting functions of smooth muscle cells upon birth [127,128].

Genes encoding chromatin modifiers
Analysis of model organisms has shown that dynamic modification of chromatin structure serves as an important regulator of gene expression during heart development (reviewed in [129]). Genes that encode proteins which modify or bind to histones have been implicated as disease genes in syndromes causing heart defects. This evidence supports a functional link between chromatin modification and human heart development and defects.
Kabuki syndrome (OMIM #147920) is characterized by intellectual disability, craniofacial anomalies, skeletal and hand malformations. Abnormal organ development is also recurrent and includes CHD in approximately 50 % of the cases [130]. Heart defects usually present in the form of septal defects and CoA [131]. Recently, Ng et al. [132] used exome sequencing to identify mutation of KMT2D (MLL2) as a major cause of Kabuki syndrome. KMT2D encodes a histone methyltransferase involved in di-and tri-methylation of the Lys-4 position of histone H3, which marks actively transcribed genes [133].
CHARGe syndrome (OMIM #214800) is characterized by growth retardation and malformation of eyes, ears, genitals, choanae and heart defects-often in the form of outflow tract malformations [134][135][136]. Approximately twothirds of the cases are caused by mutation of CHD7, which encodes a member of the chromodomain helicase DNA binding (CHD) family [134,135,137]. In vitro studies have shown that CHD7 binds DNA regions which correlate closely to regions of H3K4 methylation and regions with characteristics of enhancer elements. This hints that the protein is involved in transcriptional activation [138]. Recently, it was shown that CHD7 controls core components of the transcriptional circuit of neural crest cells and that CHD7 is essential for neural crest cell migration [139]. This function may explain the high frequency of outflow tract defects in CHARGe syndrome, as neural crest cells are known to play a crucial role in septation of the cardiac outflow tract [140].
In a recent study, Zaidi et al. [141] conducted a comprehensive screening of all protein coding genes in hundreds of patients with severe forms of CHD. In this study, whole-exomes of 362 children with CHD and their healthy parents were screened for de novo nucleotide variants by next-generation sequencing (NGS). De novo variants from these parent-offspring trios were compared to de novo variants identified in 264 healthy parent-offspring trios. The authors performed transcriptome profiling experiments to identify genes with a high expression in mouse embryonic hearts (HHe genes). In the trio datasets, they compared the number of de novo variants in genes, homologous to HHe genes. This comparison yielded a significant higher rate of de novo mutations in CHD patients compared to controls. when they compared the frequencies of damaging mutations (i.e. splice-site mutations, nonsense mutations and mutations introducing frameshift) between the two groups, the differences were even more pronounced, with an odds ratio of 7.5 (p = 0.001). Interestingly, GeneOntology analysis of 249 de novo mutations identified in CHD patients revealed significant enrichment for mutations in genes involved in H3K4 methylation. Moreover, in CHD patients 27 % of the damaging mutations within HHe genes were affecting proteins involved in H2K4 or H3K27 histone modification. These data suggest that genes involved in histone-modification are significant in the pathogenesis of isolated CHD.
Human genome analysis in combination with functional analysis of candidate genes in animal models has been instrumental in identifying the genes responsible for heart defects in several microdeletion syndromes (see below). Interestingly, several of these genes also encode chromatin modifying proteins, which support the potentially significant role of epigenetic mechanisms in both isolated and syndromic CHD.

Part II: chromosomal aberrations in congenital heart disease
Microscopically visible chromosomal aberrations are present in 8-18 % of CHD patients [142][143][144]. Furthermore, CHD is a characteristic part of the clinical spectrum in a significant number of syndromes caused by a chromosome abnormality. The most common chromosome syndrome associated with CHD is Down syndrome [145]. Congenital heart defects are seen in 45 % of individuals with Down syndrome, with the majority of cardiac defects being AvSD, ASD and vSD [146]. Cardiac defects are also found at a high frequency in other aneuploidy syndromes, including Turner syndrome (monosomy X), edward syndrome (trisomy 18) and Patau syndrome (trisomy 13) [147][148][149][150].
CHD is a component of the clinical spectrum in a number of syndromes caused by submicroscopic chromosomal deletions or duplications (listed in Table 2). Some of these syndromes are well-studied microdeletion syndromes, for which the molecular defect has been known for many years. In addition, several novel microdeletion and microduplication syndromes associated with CHD have recently been discovered due to the widespread use of molecular cytogenetic methods.
CHD candidate genes identified from microdeletion and microduplication syndromes Genotype-phenotype comparisons in patients with microdeletion and microduplication syndromes have identified candidate CHD disease genes. Subsequent mutation screening of candidate genes in patients and studies of the genes and their product in animal models have substantially added to the understanding of CHD and cardiac developmental biology.
The majority of DiGeorge syndrome (DGS, OMIM #188400, also known as 22q11.2 deletion syndrome and velocardiofacial syndrome) are caused by a 3 Mb deletion in 22q11.2 [151]. 22q11.2 duplication syndrome (OMIM #608363) is caused by duplication of genomic material in 22q11.2. Most of the 22q11.2 duplications that have been reported are reciprocal to the common 3 Mb deletion involved in DGS [152].
The common 3 Mb deletion affects more than 50 genes, including the gene encoding the T-box transcription factor TBX1. It is generally accepted that haploinsufficiency of TBX1 significantly contributes to the CHD phenotype in DGS patients. Tbx1 −/− mice display similar cardiac phenotypes to individuals with 22q11.2 DS [153,154]. Conditional knock-out experiments in mice have shown that Tbx1 is required for proliferation of cardiac progenitors in the SHF-a cell population which contributes to the development of the cardiac outflow tract [155]. Additionally, point mutations in TBX1 have been reported in patients without the 22q11.2 deletion, but exhibit a clinical presentation similar to DGS [156,157].
Transgenic mice overexpressing Tbx1 display phenotypic similarities consistent with 22q11 duplication patients, including cardiac outflow tract defects. This suggests that correct gene-dosage of TBX1 is important for normal cardiac development [152,158,159].
Conversely, cases with cardiac defects carrying smaller (1.5 Mb) deletions within the common 3 Mb region, distal to TBX1 have also been reported [160][161][162], which implies that other genes in the 3 Mb region may contribute to the cardiac phenotype of DGS patients. An interesting candidate gene within this region is CRKL, encoding a protein kinase. Mice with targeted deletion of Crkl exhibit defective OFT development and vSDs [160,163]. Furthermore, experiments with compound heterozygous Tbx1 +/− , Crkl +/− mice indicate a possible genetic interaction between the two genes, leading to the increased severity of the cardiac phenotypes in the double mutants [164]. A genetic interaction between Crkl and Fgf8 has also been shown [165], supporting a link among FGF8, TBX1 and CRKL in the pathogenesis of DGS (see below).
A high degree of phenotypic variability is a characteristic feature of DGS and 22q11.2 duplication syndrome. Parts of this variation may perhaps be explained by variations in genes located within 22q11.2, with TBX1 as the most likely candidate. An alternative explanation involves epistasis, as TBX1 has been shown to regulate or interact with several proteins and signaling networks. Gene Table 2 Microdeletion and microduplication syndromes with genomic copy number variation and CHD a Genes causing heart defects when deleted in mice (http://www.informatics.jax.org) (bold) and/or by point mutations in additional patients with CHD as part of the clinical spectrum (underlined) b Point mutations in RAI1 causes Smith-Magenis syndrome, but CHD have not been reported in patients with point mutations in RAI1 and cardiac defects are not observed in the Rai1 mouse model [286] c One out of four KDvS patients with point mutations in KANSL1 had CHD d The lissencephaly phenotype of MDLS is caused by haploinsufficiency of the PAFAH1B1 gene (also known as LIS1) [287,288] expression profiling of tissues in the pharyngeal region from mouse models with targeted deletion of Tbx1, have identified several Tbx1 target genes, which include genes involved in homeostasis of retinoic acid (RA) [166][167][168]. Interestingly, RA regulates Tbx1 expression [169], thus there seems to be a dual relationship between TBX1 expression and RA signaling. During development of the pharyngeal arches TBX1 expression in the pharyngeal endoderm is regulated by the Hedgehog signaling pathway though the action of forkhead transcription factors [170,171]. TBX1 itself regulates the expression of FGF8 in the SHF and in the pharyngeal endoderm [172,173]. A genetic interaction among Tbx1, Six1/Eya1 and Fgf8 was recently demonstrated in mouse models [174]. Further, TBX1 can act as a negative modulator of BMP signaling by binding SMAD1 and hereby interfere with the SMAD1/SMAD4 interaction [175]. williams-Beuren syndrome (wBS, OMIM #194050) is caused by deletion of genomic material in 7q11.23. Most patients with wBS are heterozygous for a 1.5-1.8 Mb deletion encompassing ~28 genes [176][177][178][179]. Cardiovascular abnormalities are present in 75 % of individuals with wBS, predominantly in the form of supravalvular aortic stenosis (SvAS) and pulmonary arterial stenosis [180]. In 6-10 % of cases aortic or mitral valve defects are also seen, and other so-called "atypical" cardiac defects in the form of ASD, vSD and TOF are observed in a significant fraction of the patients [180,181].
The ELN gene, encoding elastin, is believed to be the gene responsible for SvAS in wBS. Patients with atypical deletions including only ELN and LIMK1 genes and SvAS have been reported [182,183]. In addition, point mutations in ELN are associated with familial and sporadic SvAS [184,185]. Targeted deletion of the Eln gene in mice results in reduced aortic lumen diameter due to subendothelial accumulation of smooth muscle cells [186].
However, deletion of ELN does not explain the occurrence of the atypical heart defects in a proportion of wBS patients. Results gained from a recently reported mouse model with targeted deletion of Baz1b, indicate that deletion of this gene may account for these defects. BAZ1B is located within the wBS common deleted region, and homozygous Baz1b −/− mice exhibit a range of cardiovascular defects, which include ASD, vSD, trabeculation defects, coarctation of the aorta (COA), hypoplastic pharyngeal arch artery and a low frequency of DORv [187]. BAZ1B (also known as WSTF) acts as a subunit in three ATP-dependent chromatin remodeling complexes; the wSTF including nucleosome assembly complex (wINAC) [188], the wICH complex (wSTF-ISwI chromatin remodeling complex) [189] and the B-wICH complex [190]. These complexes are important for gene regulation, DNA replication and DNA repair [189,191]. Thus, the cardiac phenotypes of Baz1b knockout mice and the chromatin remodeling function of BAZ1B suggests that some of the phenotypes involved in wBS, including "atypical" heart defects, may be caused by epigenetic effects.
wolf-Hirschhorn syndrome (wHS, OMIM #194190) is caused by microdeletions in 4p16.3. Genotype-phenotype comparisons in patients with submicroscopic deletions suggest that haploinsufficiency of the gene encoding the histone lysine methyl transferase wHSC1 (also known as NSD2) contributes significantly to the wHS phenotype [152,192]. A recently published investigation of mice with targeted deletion of the H3K36me3specific histone methyltransferase gene Whsc1 puts forth Whsc1as another component in heart development [193]. The Whsc1 −/− mutant mice displayed ASD and vSD, and co-immunoprecipitation experiments with nuclear extracts prepared from embryonic hearts showed that whsc1 interacts with the cardiac transcription factor Nkx2-5. Furthermore, ChIP assays demonstrated that whsc1 cooperates with Nkx2-5 in the transcriptional regulation of target genes. Cross-breeding experiments with Whsc1 −/+ and Nkx2-5 −/+ mice suggested a genetic interaction between the two genes during cardiac septal formation. Another candidate gene for heart defects in wHS is the FGFRL1 gene, which encode a member of the fibroblast growth factor receptor family. During mouse development Fgfrl1 is expressed in the brain, cranial placodes, pharyngeal arches, somites and heart [194]. Targeted deletion of Fgfrl1 in mice can result in a range of developmental defects, including heart defects in the form of vSD, and both semilunar and atrioventricular valve deformation [194].
Cardiac defects are observed in 94 % of cases with interstitial deletions in 8p23.1 [195]. The defects range from isolated septal defects to complex heart defects like TOF and hypoplastic left heart syndrome (HLHS). A proportion of the patients carry a ~3.7 Mb recurrent deletion flanked by low copy repeats, although some patients have larger deletions that may extend to the 8p telomere. The gene encoding the cardiac zinc finger transcription factor GATA4 is located within the recurrent deletion, and it is well documented that this gene is associated with congenital heart defects. Mutations in GATA4 cause human CHD, often in the form of septal defects, but other defects have been reported [89,91,196]  . Therefore it is possible that the complex cardiac phenotypes observed in a subset of 8p23 deletion patients are evoked by epistatic effects from genes encoding GATA4 binding partners.
Other microdeletion and microduplication syndromes which comprise CHD The widespread use of molecular cytogenetic methods like fluorescent in situ hybridization (FISH) and especially array comparative genome hybridization (array CGH) in clinical genetics laboratories has led to the recent delineation of a number of microdeletion and microduplication syndromes, which incorporate CHD as a component of their clinical spectrum ( Table 2).
The minimal deleted region in 1q21.1 deletion syndrome contains the GJA5 gene [197]. A recent screen of 807 TOF cases revealed significant enrichment of small duplications encompassing GJA5, thus providing convincing evidence for a link between GJA5 and CHD [152,192,198]. Cardiac defects have been reported in a proportion of mice with targeted deletion of Gja5 [199,200], suggesting that haploinsufficiency of GJA5 may be responsible for cardiac defects in some individuals with 1q21.1 deletions. GJA5 encodes the cardiac gap junction subunit Connexin 40, which is expressed in the atrial myocardium and the atrioventricular conduction system [201,202]. Gap junctions are cell membrane channels that interconnect the cytoplasm of neighboring cells. In the heart, these channels contribute to the atrioventricular conduction [203,204], but at present there is no proposed mechanism describing how GJA5 haploinsufficiency results in structural heart defects.
Molecular delineation of 9q34 microdeletions and mapping of the chromosomal breakpoints in a patient with a t(X;9) translocation suggested that the EHMT1 gene is responsible for Kleefstra syndrome (KS, OMIM #610253) [205][206][207]. Mutation screening in patients without deletions in 9q34 subsequently confirmed that haploinsufficiency of EHMT1 causes KS [205,208]. Approximately 40 % of patients with KS and deletion of 9q34 have CHD, and the presence of CHD in five out of eleven KS patients with point mutations in EHMT1 confirm that this gene is responsible for CHD in KS. EHMT1 encodes euchromatic histone-lysine N-methyltransferase 1, which regulates transcription by methylation of histone H3 lysine 9 (H3K9Me2) in euchromatic DNA [209].
Koolen-De vries syndrome (KDvS, OMIM #610443) is caused by recurrent deletions in 17q21.31. Between 27 and 36 % of KDvS patients have CHD [210,211]. Recent delineation of the critical region of 17q21.31 and mutation screening of KDvS patients without deletion of 17q21 revealed that KDvS is caused by haploinsufficiency of KANSL1 [212,213]. One out of four patients with point mutations in KDvS has CHD, hinting that KANSL1 is a CHD disease gene, although further patient data is needed to confirm this link. KANSL1 encodes a member of the male specific lethal (MSL) complex initially described in Drosophila (reviewed in [214]). within the MSL complex KANSL1 interacts with KAT8, a histone acetyltransferase which regulates gene expression through acetylation of H4 lysine 16 (H4K16) [215].
Another interesting CHD candidate gene is TBX2, which is located within the deleted region in 17q23 deletion syndrome [216]. TBX2 is expressed in non-chamber myocardium of the developing heart, and mice with targeted mutation in Tbx2 have defects in the development of the atrioventricular canal (AvC) and the OFT [217]. It has been hypothesized that TBX2 is involved in cardiac chamber development and functions as a local repressor of the chamber-specific gene program in non-chamber regions like the AvC and OFT [218,219].

Pathogenic copy number variants identified in cohorts of CHD patients
Array CGH and similar methods have been used to screen cohorts of CHD patients for pathogenic CNvs in the form of duplications and deletions. Since 2007, 14 wholegenome CNv screening studies have been reported, comprising more than 5,000 patients (Table 3, [220][221][222][223][224][225][226][227][228][229][230][231][232][233][234]). The reported studies show large differences, which include size and phenotypic composition of patient cohorts and the experimental and analytical setup, thus it is somewhat difficult to compare the results. Nevertheless, we find it safe to conclude that pathogenic CNvs are found among a significant portion of CHD patients.
The highest frequency of pathogenic CNvs is found among patients with CHD and extra-cardiac anomalies. Based on the current reports [220,224,228,230,233,234] we estimate that pathogenic CNvs are present in 15-20 % of patients with CHD and extra-cardiac anomalies.
Among patients with isolated CHD, the frequency of pathogenic CNvs is significantly lower. Here, we estimate the frequency to be between 4 and 14 % [221,222,225,231,232]. However, this estimate should be treated with caution, due to the aforementioned large differences in study design.

Identification of disease genes and pathways from CNvs detected in cohorts of CHD patients
In principle, CNvs identified as pathogenic in CHD patients should contain one or more dosage sensitive cardiac developmental genes. Thus, each pathogenic CNv, or at least overlapping CNvs should define a disease locus for  Genes known to cause CHD in humans are underlined, genes with reported cardiovascular system involvement (e.g., from targeted deletion in mice) are bold a Includes 60 patients from Thienpont et al. , and Triple X syndrome (n = 1) g Candidate genes suggested by the authors CHD and it therefore should be possible to use CNvs to detect CHD disease genes. Several groups have reported identification of CNvs spanning genes, which span genes previously recognized to cause CHD in animal models (Table 3), thereby providing a plausible link between these genes and CHD in humans. However, many of the identified CNvs do not contain a well-established cardiac developmental gene. These CNvs often contain several genes, although only one is likely to be the gene responsible for CHD. Three approaches have been utilized to identify the causal genes in such cases: (1) narrowing of the locus by comparison of multiple samples with overlapping CNvs, (2) in silico gene prioritization and (3) functional investigations of candidate genes within the CNvs.
Hitz et al. [226] used endeavour [235] to test for enrichment of angiogenesis-associated genes within 73 CNvs identified in patients with left-sided CHD. They also searched for genes with expression in the developing heart in serial analysis of gene expression (SAGe) and public databases. By combining these prioritization methods, they identified 25 CHD candidate genes (Table 3). Soemedi et al. [232] performed genomic region annotation enrichment analysis on rare deletions and duplications identified in 2,256 CHD cases. They found enrichment of 13 genes encoding proteins involved in the wNT signaling pathway (Table 3). Silversides et al. [231] performed a systematic review of genes within rare CNvs identified among 433 cases with TOF and identified 62 CHD candidate genes (Table 3). They also assessed whether genes, in predefined gene-sets derived from GeneOntology (GO) annotations and pathway and protein domain databases, were significantly overrepresented in CNvs detected in TOF cases compared to controls. They found enrichment of gene-sets belonging to five functional clusters: vasculature development, chromosome organization, cell motility, chemotaxis and neuron projection and development. Lalani et al. [228] identified eight candidate genes in CNvs identified in patients with CHD and extracardiac anomalies (Table 3). They grouped genes within enriched CNvs based on GO categories and found enrichment for genes encoding proteins involved in G-protein coupled receptor internalization, hemopoiesis and cytoskeleton organization. Furthermore they analyzed protein-protein interactions between proteins encoded by candidate genes in CNvs identified in patients and a set of 276 proteins from GO cardiac development categories. They identified 11 proteins with at least one connection with a human cardiac-specific protein (significant at p = 0.03).
Thienpont et al. [236] identified TAB2 as a dosage sensitive CHD disease gene by comparing overlapping CNvs within 6q25. The overlapping region of seven CNvs identified in CHD patients revealed a CHD locus containing 11 candidate genes, including TAB2. For prioritization of the candidate genes in the locus and surrounding genomic region, the authors performed in silico analyses of 105 genes in 6q24-25, using an adapted version of endeavour [235]. This analysis predicted TAB2 as the highest-ranking candidate gene in 6q24-25. Further functional analyses of TAB2 in human embryonic heart tissues and zebrafish suggested that TAB2 is a cardiac developmental gene. Point mutations localized within TAB2 in two unrelated CHD patients and mapping of a translocation breakpoints within TAB2 in a CHD family segregating a t(2;6) translocation, further verified TAB2 as a CHD disease gene.
Fakhro et al. [223] performed whole-genome CNv screening of 262 patients with Heterotaxy and isolated TGA. They identified 45 unrecorded gene-containing CNvs, including two different CNvs affecting TGFBR2. evaluation of candidate genes using in situ hybridization and Morpholino-based gene knock-down in X. tropicalis showed that the genes tgfbr2, rock2, galnt11, nek2 and nup188 are involved in left-right patterning of the heart (Table 3). ROCK2 and NeK2 are ciliary proteins, thus this study confirmed the importance of cilia and TGF-β signaling in LR patterning [237]. In addition, this study identified two novel genes (GALNT11 and NUP188) with unknown functions in LR development.
The molecular pathology of congenital heart disease Cardiac development is controlled by a large number of signaling pathways, which are tightly regulated in time and space, and interact in complex developmental networks [34]. The CHD disease genes, which have been identified to date, suggest that all aspects of developmental signaling pathways may be involved in human CHD: from ligands (e.g. JAG1) and receptors (e.g. NOTCH, PDGFRA), across down-stream signaling effectors (e.g. PTPN11, SMAD6), to transcription factors (e.g. GATA4, NKX2-5) and targets (e.g. ACTC1, MYH6) (Fig. 1). Moreover, discoveries of disease genes encoding histone-modifying proteins (e.g. CHD7, KMT2D), suggest that epigenetic regulation of an unknown number of target genes, may add an additional layer of regulation on consensus cardiac developmental networks.
Lage et al. [238] have recently shown that a wide range of CHD risk factors, functionally converge in complex, yet discrete, protein networks driving heart development. These findings, combined with the potentially hundreds of CHD disease genes [141], suggest that CHD may be caused by a very large number of combinations of mutations and environmental risk factors.
Reduced penetrance of CHD are often observed in human pedigrees (e.g. [239], unpublished observations in Danish pedigrees) and in carriers of CNvs known to cause CHD (see Table 2 and discussions in text). Some of the reduced penetrance may simply be due to unidentified asymptomatic heart defects in some carriers, but may also be caused by epistasis. Several examples of epistasis in mouse models are mentioned in the text above. winston et al. [240] performed a systematic study of the influence of genetic background on the expression of heart defects in Nkx2-5 +/− heterozygous mice. The authors compared Nkx2-5 +/− heterozygous C57Bl/6 mice with Nkx2-5 +/− heterozygous F1 progeny from crosses with two other mouse strains. The data showed that the F1 hybrid mice presented with a significantly lower incidence of septal defects compared to mice with the original C57Bl/6 background.The authors suggest that modifying alleles can either direct the manifestation of a cardiac developmental defect or buffer the effect from perturbations. The latter situation, which was the case in the study, may ensure robustness of normal heart development.
How the large heterogeneity in CHD and potentially large epistatic effects translates into lesions in the personal genome of the individual patient remains to be investigated. One possible scenario could be that individual combinations of several risk alleles may be the cause of CHD in part of the patients.

Future perspectives
It was recently demonstrated that exome-sequencing is a powerful tool for identification of de novo mutations in CHD [141]. It is very likely that more studies based on exome-sequencing will reveal new CHD disease genes in the near future. The high number of variants identified in exome-sequencing experiments is a big challenge in very heterogeneous disorders like CHD. Thus, it is likely that such studies will be performed on large numbers of parentoffspring trios or on families with dominant or recessive inherited CHD.
The past 5 years have shown that analysis of genome rearrangements in the form of CNvs, translocations or inversions can lead to detection of new CHD disease genes or loci. The technology for mapping such rearrangements is continuously improving, and breakpoints in balanced translocations and inversions can now be mapped within days using NGS [241].
Untreated, CHD is a disorder with a high mortality rate, therefore a large part of the disease causing mutations are likely rare in populations due to negative selection. However, it is also possible that some variants associated with CHD may escape negative selection. Such variants may be discovered through genome wide association studies, as has recently been demonstrated [242][243][244].
Interesting therapeutic opportunities could arise from the current knowledge of the molecular pathogenesis of CHD. A significant part of CHD seems to be caused by mutations which perturb complex developmental networks. These networks are characterized by extensive communication within and between specific signaling pathways, and with the environment. Thus, given the apparent epistatic effects observed in patients and animal models, it should be possible to manipulate the signaling pathways in the developmental networks with synthetic agonists or antagonists, and thereby alleviate effects from mutations or redirect signaling events towards normal heart development. A recent study suggests that this could be possible someday. Tian et al. [245] showed that defects in the cardiac inflow tract and Av canal (resembling complete CAvC in humans) in Wnt2 −/− mice, could be rescued by transient pharmacological activation of wnt signaling with LiCl.
Naturally, such therapeutic opportunities are presently very hypothetical, and to become reality, much more knowledge about the molecular genetics and the molecular pathology of CHD are needed. Combining human genetics/ genomics with functional studies in cell models or animal models like zebrafish, Xenopus frogs, chicken or mice are likely to have the greatest impact on our understanding of the molecular pathology in human CHD.  (1)