Genomic frontiers in congenital heart disease

Abstract

The application of next-generation sequencing to study congenital heart disease (CHD) is increasingly providing new insights into the causes and mechanisms of this prevalent birth anomaly. Whole-exome sequencing analysis identifies damaging gene variants altering single or contiguous nucleotides that are assigned pathogenicity based on statistical analyses of families and cohorts with CHD, high expression in the developing heart and depletion of damaging protein-coding variants in the general population. Gene classes fulfilling these criteria are enriched in patients with CHD and extracardiac abnormalities, evidencing shared pathways in organogenesis. Developmental single-cell transcriptomic data demonstrate the expression of CHD-associated genes in particular cell lineages, and emerging insights indicate that genetic variants perturb multicellular interactions that are crucial for cardiogenesis. Whole-genome sequencing analyses extend these observations, identifying non-coding variants that influence the expression of genes associated with CHD and contribute to the estimated ~55% of unexplained cases of CHD. These approaches combined with the assessment of common and mosaic genetic variants have provided a more complete knowledge of the causes and mechanisms of CHD. Such advances provide knowledge to inform the clinical care of patients with CHD or other birth defects and deepen our understanding of the complexity of human development. In this Review, we highlight known and candidate CHD-associated human genes and discuss how the integration of advances in developmental biology research can provide new insights into the genetic contributions to CHD.

Congenital heart disease (CHD) has long been recognized as the most common and often severe anomaly at birth, with a prevalence of 6–13 in 1,000 newborn babies^1–5. CHD encompasses a broad spectrum of heart malformations that range from a solitary abnormality (atrial or ventricular septal defect or isolated dysplastic valve) to complex lesions consisting of multiple defects (tetralogy of Fallot or hypoplastic left heart syndrome) (Fig. 1). Advanced surgical and medical interventions, including palliation and definitive correction of critical (requiring early intervention for survival) malformations during infancy have significantly reduced CHD mortality⁶. Currently, >90% of patients with CHD survive to adulthood, accounting for the increased prevalence of CHD in the general population⁶. Moreover, the increased longevity of patients with CHD has led to a greater recognition of the associated comorbidities. Approximately 13% of newborn babies with CHD have extracardiac structural or functional defects and can develop neurodevelopmental delay⁷ throughout childhood. When these additional diagnoses are likely to have resulted from the same aetiology, CHD is classified as syndromic. Complex heart malformations most often occur in syndromic CHD, which can result in lifelong multisystem health issues. Understanding the aetiologies and mechanisms of CHD can therefore contribute insights into normal and aberrant developmental processes of multiple organ systems.

Fig. 1 |. Common types of congenital heart disease.

Schematic of congenital heart disease malformations grouped by anatomical subtype and showing the oxygen saturation of blood. Abnormal connections between the left and right chambers, such as atrial or ventricular septal defects or atrioventricular canal defects, are typically acyanotic owing to left-to-right shunting of oxygenated blood into the pulmonary circulation. Left ventricular outflow tract obstructions (LVOTOs) range from isolated aortic stenosis to a combination of anomalies (hypoplastic left heart syndrome with mitral and aortic stenosis) associated with cyanosis. Conotruncal malformations are frequently cyanotic, with shunting of deoxygenated blood into the systemic circulation, and include tetralogy of Fallot and transposition of the great arteries. The percentages denote the prevalence of malformation in the cohort of patients with congenital heart disease from the Paediatric Cardiac Genomics Consortium^17,18. Note that the percentages exceed 100 due to coexisting structural cardiac abnormalities in the participants. Ao, aorta; LA, left atrium; LV, left ventricle; PA, pulmonary artery; RA, right atrium; RV, right ventricle.

Genetics has a central role in CHD pathogenesis, a discovery that has benefited from considerable technical advances in examining the human genome. Early evidence came from karyotyping patients with syndromic CHD that uncovered aneuploidies, most commonly trisomy 13, 18 or 21 (Down syndrome) and monosomy X (Turner syndrome), which occur in ~12% of patients with CHD⁸. Cytogenic analyses and genome arrays extended the molecular insights into subchromosomal structural rearrangements, revealing prevalent recurrent 3-Mb deletions at 22q11.2 in DiGeorge (velocardiofacial) syndrome^9–12. Deletions at 22q11.2 are detected in approximately 2% of all patients with CHD¹³ and in 13% of patients with specific cardiac malformations, as assessed by fluorescence in situ hybridization and targeted amplification¹⁴. Karyotype-detected or microarray-detected abnormalities together account for almost 25% of cases of sporadic CHD¹⁵.

Genome-wide linkage analyses enabled the initial identification of monogenic causes of familial CHD, including variants in genes encoding transcription factors and transcriptional regulators of genes related to cardiac development¹⁶. A detailed knowledge of the human genome at base pair resolution and advances in sequencing technologies have substantially accelerated the discovery of genes associated with CHD. Contemporary whole-exome sequencing (WES) and whole-genome sequencing (WGS) platforms, bioinformatic analytical tools and large, aggregated, population-based sequencing datasets enable the identification of deleterious missense and loss-of-function (LOF) variants (together denoted as damaging variants), small insertions or deletions, copy number variants (CNVs) and structural variants. Application of these strategies has uncovered damaging variants that segregate in familial cases of CHD, reflect de novo mutations in CHD probands with unaffected parents, and occur at significantly higher rates in approximately 20% of patients with CHD in large cohort studies^15–19.

Whole-exome sequencing.

(WES). Targeted sequencing of protein-encoding regions, which comprise 1% of the genome. WES can be used to determine single-nucleotide variants, small insertions and deletions, and copy number variants, but is less sensitive than WGS for the detection of structural variants.

Whole-genome sequencing.

(WGS). Sequencing of the entire genome (protein coding and non-coding regions) that can be used to determine single-nucleotide variants, small insertions and deletions, and structural variants.

Collectively, these approaches identify damaging coding variants in definitive and candidate genes for CHD in 45% of patients with CHD^15–19. Genes with a definitive association with CHD harbour variants that significantly co-segregate in families with CHD, are significantly enriched in unrelated patients with CHD compared with controls or account for heart malformations in patients with syndromic CHD. Candidate genes for involvement in CHD have similar properties, but without sufficient statistical significance to warrant a definitive classification. Diverse types of variants (aneuploidies, CNVs, structural variants and LOF variants) alter the dosage of CHD-related genes, whereas deleterious missense variants can maintain physiological gene dosage but impair function of the encoded protein.

Genes linked to CHD provide new insights into the crucial molecules and pathways involved in cardiogenesis, complementing and extending the considerable information derived from exploring heart development in experimental models. Studies of spontaneous (de novo) gene variants in sporadic CHD can identify new genes involved in cardiogenesis, while detailed clinical evaluations allow the definition of a broader spectrum of cardiac and extracardiac phenotypes than studies in experimental models. Given that the dosage of most human variants is heterozygous, whereas studies in experimental models often explore homozygous variants, human CHD data are informative about subtler influences on heart formation. Additionally, the observation that the highest rates of de novo damaging variants in genes associated with CHD occur in patients with critical and complex malformations^17–19 provides compelling evidence for profound deleterious effects on reproductive health and the evolutionary constraint on many genes linked to CHD.

Why is CHD the most common congenital anomaly? One hypothesis is that complex developmental processes of heart formation are exquisitely sensitive to changes in gene dosage among many essential genes and pathways. When gene dosage is altered, developmental errors can occur, many of which are readily detected by contemporary screening ultrasonography that is routinely performed at 18 weeks gestational age²⁰. To highlight the exquisite molecular orchestration of heart formation, in this Review we briefly summarize cardiac development before discussing progress in the past 10 years in the genomic discoveries and strategies at the forefront of current CHD research. Several informative reviews have examined key signals that govern cardiac morphogenesis, and readers are referred to these publications for more details^21–26. Throughout the Review, we highlight in bold those genes with damaging variants identified in patients with CHD.

Cardiac development

The heart forms during the first weeks of embryogenesis through the specification of cardiac progenitor fields that coalesce into a cardiac crescent. These progenitor fields evolve into a linear and then looped heart tube that grows, undergoes regional transformation and emerges as a nearly fully formed organ by 8 weeks of gestation in humans. Throughout cardiac morphogenesis, the interplay between differentiating progenitor cell populations generates specialized cardiac cells with precise spatial organization within constrained regions.

Overview of progenitor cell populations and morphogenesis.

Three major progenitor cell populations contribute to cardiac structures: the first heart field (FHF), derived from the lateral plate mesoderm, the second heart field (SHF), derived from the lateral plate splanchnic mesoderm, and migratory cardiac neural crest populating the III, IV and VI pharyngeal arches^27–30 (Fig. 2). The symmetric FHF and SHF meet anteriorly to form the cardiac crescent, which then fuses into a linear heart tube with ventricular and atrial progenitors oriented on the anterior–posterior axis. The heart tube undergoes rightward looping with input from a morphogen gradient defining the left–right axis, producing the spatial arrangement of fetal cardiac chambers. Cardiac cells derived from the FHF give rise to the atria and left ventricle, whereas cells from the SHF become the right ventricle and outflow tract and also contribute to both atria. Ballooning of the outer curvature of the heart tube via asymmetric growth increases the size of the ventricular chamber³¹. Subsequent outflow tract septation and smooth muscle differentiation occur with contribution from the cardiac neural crest.

Fig. 2 |. Human heart development.

Human heart development involves the formation of the first heart field (FHF) or cardiac crescent at approximately embryonic day 15, with chamber-specific cardiac progenitors aligned in a specified spatial distribution. Concurrently, an additional population of cardiac progenitors, the second heart field (SHF), develops posterior to the FHF. By embryonic day 21, the midline fusion of the FHF forms a linear cardiac tube and SHF-derived cells migrate (arrows) into the arterial pole (AP) and venous pole (VP) of the tube. The linear cardiac tube undergoes looping by embryonic day 28 to form early ventricular chambers. Cardiac neural crest cells migrate to populate the III, IV and VI pharyngeal arches, the aortic sac (AS) and the conotruncal (CT) ridges. By embryonic day 50, the cardiac structures largely resemble those of the mature heart, with four main chambers and two outflow tracts (OFT). A, atrium, Ao, aorta; AV, atrioventricular; DA, ductus arteriosus; LA, left atrium; LCA, left carotid artery; LSCA, left subclavian artery; LV, left ventricle; PA, pulmonary artery; RA, right atrium; RCA, right carotid artery; RSCA, right subclavian artery; RV, right ventricle; V, ventricle.

Genetic regulators of cardiac morphogenesis.

Many crucial molecules participate in the differentiation of progenitor cell populations into cardiac cell lineages and in directing cardiac morphogenesis, some of which are encoded by genes that are associated with CHD. Formation of the FHF from cardiogenic mesoderm is induced by signalling pathways including bone morphogenetic protein (BMP) and fibroblast growth factor (FGF) from the adjacent endoderm^32–35. In experimental model systems, ablation of genes encoding BMPs reduces the expression of crucial cardiac transcripts that are required for commitment to a cardiomyocyte lineage³⁶. Sarcomere formation and contraction are initiated by the expression of genes such as GATA4, GATA6, NKX2–5 and TBX5 (REFS ^21–26,37). Expression of IRX4 in both the FHF and SHF of the cardiac crescent promotes specification of ventricular cardiomyocytes³⁸. Commitment to an atrial lineage is initiated by factors including retinoic acid within mesoderm progenitor fields³⁹. Further specification results in cardiomyocytes destined to populate the left (IRX4 (REF.⁴⁰), HAND1 (REF.⁴¹) and MSX1 (REF.⁴²)) and right (IRX4 (REF.⁴⁰) and GATA6 (REF.⁴³)) ventricles and the outflow tract (GATA6 (REF.⁴³), HAND2 (REF.⁴¹), ARID3B⁴⁴ and TEAD⁴⁵). SHF contributions to the outflow tract require FGF signalling³⁵, which enables neural crest migration and is required to induce tbx1 expression in the SHF in zebrafish⁴⁶. Despite stereotyped differentiation pathways, plasticity is maintained in the progenitor fields during cardiac development, because ablation of the FHF cells in zebrafish results in SHF compensation that regenerates ventricular cardiomyocytes⁴⁷.

Signalling pathways and gene networks regulating cardiac neural crest differentiation have also been identified. Absence of retinoic acid signalling prevents normal differentiation of cardiac neural crest cells in mice⁴⁸, which can be ameliorated by maternal vitamin A supplementation⁴⁹, highlighting genotype and environmental interactions in CHD. By contrast, a GATA6 missense variant that is recurrently identified in patients with CHD and disrupts outflow track formation strikingly increases retinoic acid signalling⁴³. Cardiac neural crest migration depends on FOXC2 expression⁵⁰, which in turn mediates the activation of Sema3c expression in the SHF⁵¹, whereas Tbx1 expression results in counterbalance inhibitory signals⁵¹.

The heart also contains non-chamber cardiomyocytes within the atrioventricular canal and outflow tract as well as valvular endothelial cells and smooth muscle cells. Non-chamber myocardium is specified by the repression of chamber-specific genes by regulatory proteins such as those encoded by TBX2 and TBX3 (REF.⁵²) and BMP2 and BMP4 (REF.⁵³). A subset of valvular endothelial cells undergo endothelial-to-mesenchymal transition to populate the endocardial cushions under the guidance of signalling cascades that include vascular endothelial growth factor (VEGF)⁵⁴ and transforming growth factor-β (TGFβ)⁵⁵. Studies of outflow tract development in mouse models have also demonstrated the requirement for histone deacetylase 3 (encoded by Hdac3) for smooth muscle differentiation, whereas the pulmonary arteries and veins require FGF and Sonic hedgehog signalling^56–58. Cardiomyocytes can also become non-chamber cells via transdifferentiation, and single-cell characterization of the mouse developing heart found a cardiomyocyte population that transdifferentiates into vascular smooth muscle cells of the great arteries⁵⁹.

Establishment of a left–right signalling axis in the embryo occurs within the embryonic node, where signals mediated by motile and sensory cilia lead to the asymmetric expression of genes encoding downstream regulatory signals, including the genes coding for the TGFβ family members Nodal, Lefty1, Lefty2 and Zic3 (REFS^60–62). These signals are crucial for appropriate rightward looping of the linear heart tube that properly arranges the atria, ventricles and outflow tract and establishes the sinoatrial node in the right atria. Perturbation of Sonic hedgehog or activin signalling in the chick embryo⁶³ or knockout of Nodal, Lefty1 or Lefty2 (encoding left–right axis signalling proteins and transcriptional regulators) in mice lead to defects in cardiac looping and resultant congenital heart defects^64,65.

Disruption of human cardiac morphogenesis

The anatomical study of human embryos has provided insights into human cardiogenesis and the aetiology of the cardiac malformations that result from the disruption of specific spatiotemporal processes. CHD lesions are often classified into anatomical subgroups that arise from disruption of shared embryonic processes. As such, the discovery of genes with pathogenic variants that elicit particular malformations provides clues into the downstream developmental and physiological consequences of genetic regulation in heart development.

Conotruncal defects.

Conotruncal defects, such as tetralogy of Fallot, persistent truncus arteriosus and transposition of the great arteries (Fig. 1), reflect maldevelopment of the ventricular septum in the outflow tract and/or the great arteries, or malalignment of these structures. Conotruncal defects frequently occur in autosomal dominant 22q11.2 deletion syndrome (which commonly involves deletion of the CHD-associated gene TBX1)⁶⁶ and CHARGE syndrome (which in most cases involves deletions or variants in the CHD-associated gene CHD7)⁶⁷. These syndromes are associated with abnormalities in the heart, face and head, which arise from neural crest cells, highlighting the effects of damaging gene variants on cell lineages that contribute to extracardiac malformations. Both of these CHD syndromes have variable penetrance and expressivity^68,69 (defined in Table 1), which suggests that additional factors influence the clinical phenotype. Although these additional factors are largely unknown, a genetic modifier influencing conotruncal defects in 22q11.2 deletion syndrome has been reported⁷⁰. Recurrent duplications and deletions at 1q21.1 — which spans the candidate genes for CHD GJA5 (encoding connexin 40) and CHD1L (a gene related to CHD7) — are associated with tetralogy of Fallot, often but not always with extracardiac abnormalities^71,72.

Table 1 |.

Glossary of human genetics terms

Domain	Term	Definition
Mode of inheritance	Autosomal dominant	A single copy of a gene variant on a non-sex chromosome is sufficient to cause disease
	Autosomal recessive	A variant on each gene copy is required to cause disease
	X-linked	Gene variants encoded on the X chromosome can exhibit dominant or recessive transmission
	Hemizygous	Presence of a single gene copy in diploid cells, such as the X chromosome in males
	Mosaicism	The presence of at least two cell populations with differing genotypes, resulting in a gene variant being present in only a subset of cells
	De novo variant	A gene variant in an individual that is absent from both parents, and arises due to new mutation of parental gametes or after fertilization
	Monogenic inheritance	A single gene is causative for a trait
	Polygenic inheritance	Many genes contribute to a trait
	Oligogenic inheritance	A few genes influence the trait
Variant types	Indel	An insertion or deletion of a sequence of 1–49 base pairs
	Aneuploidy	An abnormal number of chromosomes
	Structural variant	A broad class of variants encompassing ≥50 base pairs of sequence that includes deletions, duplications, inversions, insertions or transversions
	Copy number variant	An unbalanced structural variant (deletion, duplication or insertion) that alters the copy number of a gene or genes
Molecular consequences of genetic variants	Damaging	A general descriptive term indicating that a variant alters the expression level or function of a gene
	Loss-of-function	A variant that extinguishes the expression of one gene copy or abrogates its function
	Gain-of-function	A variant that enhances gene function, produces a new function or inappropriately increases gene expression
	Frameshift	Insertion or deletion of nucleotides that shifts the triplet amino acid code used for translation of mRNA to protein
	Missense	A variant that substitutes one amino acid for another in the protein sequence
	Synonymous	A variant that alters a nucleotide but does not alter the amino acid sequence of the encoded protein
	Splicing	A gene variant that alters canonical nucleotide signals to excise introns or exclude exons from mature mRNAs; splicing variants can abrogate existing splice signals or create new splice signals
Genotype–phenotype relationship	Variable expressivity	Variation in phenotype among individuals with the causal genetic variant
	Penetrance	The proportion of individuals who manifest the phenotype associated with a genetic variant

Non-syndromic conotruncal malformations are also mainly associated with monogenic LOF variants in genes related to two important signalling pathways orchestrating developmental programmes in the SHF and cardiac neural crest cells: the Notch pathway (NOTCH1) and the VEGF pathway (KDR, FLT4, VEGFA, FGD5, BCAR1, IQGAP1, FOXO1 and PRDM1)^73,74. Why these progenitor fields and developing tissues have a unique susceptibility to alterations in the VEGF pathway remains an open question. Autosomal dominant damaging variants in ZFPM2 (encoding the transcription factor ZFPM2) and in the genes encoding its binding partners GATA4 and GATA6 also frequently result in conotruncal defects^75–77. Both GATA genes encode transcription regulators that act as pioneer factors, binding and influencing closed chromatin regions to enable subsequent transcription factor binding, highlighting the importance of this transcriptional regulatory network in conotruncal development.

Atrial and ventricular septal defects.

Abnormal development of the endocardial cushions and/or the SHF-derived dorsal mesenchymal protrusion^78,79 results in atrial and ventricular septation defects, atrioventricular valve abnormalities and atrioventricular canal defects that include septation and valve defects. Complete atrioventricular canal defects are found in only ~5% of individuals with CHD⁸⁰, but notably occur in almost half of the ~40% of individuals with trisomy 21 who have CHD^81,82. This remarkable association indicates that genes in chromosome 21, particularly those localized within a 0.96-Mb region⁸³, are crucially involved in endocardial cushion development, but the responsible gene(s) remain unknown. As in other syndromes, CHD penetrance is incomplete in individuals with trisomy 21, and 60% of individuals have a structurally normal heart^81,82.

Isolated atrial or ventricular septal defects are the most common congenital heart lesions; however, among patients with sporadic occurrence of these malformations in the absence of a genetic syndrome, the causal aetiology is rarely defined. Damaging variants in NKX2–5 and GATA4 were initially identified in patients with rare autosomal dominant familial septal defects, whereas TBX5 variants were found to cause Holt–Oram syndrome in families with septal defects and associated limb defects¹⁶. In model systems, the proteins coded by these genes have been shown to interact physically to promote proper atrial and ventricular septation^84–86.

Left-sided obstructive lesions.

This group of CHDs includes the spectrum of stenotic and atretic anomalies of the left-side chambers and valves that occur in isolation (mitral stenosis, bicommissural aortic valve, aortic stenosis and coarctation of the aorta) or in combination, such as Shone complex (annulo-leaflet mitral ring, parachute mitral valve, subaortic stenosis and coarctation of the aorta) and hypoplastic left heart syndrome (mitral and aortic stenosis or atresia, and hypoplastic left ventricle). Left-sided obstructive lesions have higher rates of heritability than other CHDs, and a range of malformations is observed within families, indicating that the development of the involved structures shares genetic pathways⁸⁷.

Patients with monosomy X or Turner syndrome (45,XO karyotype) often have bicommissural aortic valve (15–30%), aortic coarctation (7–20%) and, less commonly, hypoplastic left heart syndrome^88,89. The increased risk of left-sided obstruction lesions in patients with Turner syndrome relates to altered dosages of genes on chromosome X and copy number polymorphisms at the 12q13.31 locus⁹⁰. However, the mechanisms by which these genotypes influence the development of left-sided structures is unknown. The autosomal dominant Jacobsen syndrome (11q23 deletion), which causes diverse and severe left-sided obstructive lesions in 33% of the patients⁹¹, is mediated in part by a requirement for protein C-ets-1 (encoded by ETS1) function in the developing endocardium to support cardiomyocyte proliferation and extracellular matrix organization⁹². Monogenic damaging variants in GATA5 cause bicuspid aortic valve^93,94 and in RBFOX2 cause hypoplastic left heart syndrome⁹⁵. Rare recessive missense variants in MYH6 (encoding the cardiac α-myosin heavy chain) cause Shone complex¹⁷ and hypoplastic left heart syndrome⁹⁶. Given that cardiac α-myosin heavy chain is a contractile protein in the sarcomere that is highly expressed during cardiogenesis, MYH6 is one of a very few CHD-associated genes that is not related to transcription or translation.

Right-sided obstructive lesions.

Obstruction to the flow of systemic venous blood as it travels from the right side of the heart to the pulmonary capillary beds can occur in tricuspid atresia, pulmonary atresia with intact ventricular septum, isolated pulmonary valve stenosis and supravalvular and branch pulmonary artery stenosis, and reflects a wide range of embryological abnormalities. Tricuspid atresia results from agenesis of the right-sided atrioventricular valve and is associated with chromosomal trisomy in humans⁹⁷ and Zfpm2 LOF in mice⁹⁸. Pulmonary valve stenosis is present in >50% of children with Noonan syndrome and is more frequently associated with variants in the PTPN11 gene⁹⁹. The pulmonary valve stenosis associated with PTPN11 variants results from increased activation of mitogen-activated protein kinases (MAPKs) in endocardial cells that leads to excess endothelial-to-mesenchymal transition in the developing pulmonary valve¹⁰⁰. Pulmonary valve stenosis is a common feature of other syndromes (such as Costello and cardiofaciocutaneous syndromes) that result from excessive activation of the RAS–MAPK signalling pathway¹⁰¹. Autosomal dominant and recessive forms of Robinow syndrome are associated with isolated pulmonary valvular stenosis and conotruncal lesions, such as tetralogy of Fallot^102,103, and arise from pathogenic variants in genes encoding receptors and ligands controlling planar cell polarity (WNT5A, ROR2, DVL1 and DVL3)^{102,104–106}, demonstrating the importance of the planar cell polarity pathway in the development of the SHF-derived right ventricular infundibulum and the outflow tract cushions. Individuals with autosomal dominant Alagille syndrome (JAG1 and NOTCH2) can have tetralogy of Fallot, pulmonary valve stenosis and/or stenosis of the peripheral pulmonary arterial branches, demonstrating the dose sensitivity of Notch signalling in right ventricular outflow tract development, pulmonary valve formation and vasculogenesis that contributes to the pulmonary arterial tree^107,108.

Left–right patterning defects.

Defects in left–right axis patterning cause heterotaxy syndrome, resulting in abnormal positioning of the viscera as well as CHD that includes structural malformations and aberrant arterial and venous connections. Normal cilia function is crucial for left–right axis patterning, and a subset of patients with primary ciliary dyskinesia have heterotaxy syndrome with CHD. Patients with non-syndromic CHD and damaging variants in cilia-related genes generally have transposition of the great arteries¹⁰⁹. Damaging variants in Nodal and TGFβ signalling pathway genes that participate in laterality (NODAL, FOXH1, ZIC3, ACVR2B and SMAD2) are also frequent among patients with cardiac left–right patterning defects^109,110. Variants in cilia-related genes that are responsible for laterality defects cause CHD by autosomal dominant or autosomal recessive mechanisms. CHD-associated variants in genes encoding Nodal and TGFβ signalling proteins are dominant, whereas ZIC3 variants cause X-linked heterotaxy syndrome in hemizygous males^111,112.

Emerging CHD genes from cohort studies

Genes linked to CHD are increasingly being identified from analyses of CHD cohorts, recruited from single centres or by national and international collaborations^17,19,113. Most studies enrol patients with severe or critical CHD, under-represent patients with prevalent simple CHD lesions and exclude patients with clinically established genetic syndromes¹¹⁴. Variants are typically discovered by WES and analyses of case–control comparisons. These studies show that patients with CHD have the same rates of total de novo variants but significantly more rare (allele frequency ≤ 1 × 10⁻⁵ in the general population) damaging variants compared with the general population, including 9% excess of de novo damaging variants, 7% excess of dominant inherited variants and 1% recessive inherited variants^17–19. Among patients with unexplained sporadic CHD, damaging variants are identified in approximately 8–10% of patients^17,19. The rate of de novo damaging variants is highest among patients with syndromic CHD, whereas the rate of inherited LOF variants is highest among patients with isolated CHD¹⁹.

Cohort studies have confirmed the importance of previously identified biological pathways in CHD pathogenesis, including genes related to the Notch¹¹⁵ and VEGF¹¹⁵ signalling cascades. However, the most prominent functional class of newly identified CHD-associated genes is chromatin remodelling^18,116. Damaging variants in these genes encoding chromatin remodelling proteins occurred most often in patients with CHD who had extracardiac anomalies and neurodevelopmental delay. The broad expression pattern of these genes indicates that shared epigenetic and transcriptional regulatory pathways are employed across organ development.

WES data from patients with CHD supports previous evidence indicating the importance of ciliary function in cardiogenesis. As has been identified in other recessive disorders, damaging variants in cilia-related genes are often inherited from an unaffected parent¹¹⁷. By contrast, damaging variants in genes related to chromatin modification are more often de novo variants¹¹⁷, indicating the substantial deleteriousness of even one mutant allele in this gene class.

Definitive genes for CHD.

Analyses of WES data from two cohorts of patients with CHD, totalling 2,056 patients with isolated CHD, 2,994 with syndromic CHD and 808 with CHD and unknown extracardiac status, were of sufficient statistical power to detect a significant excess of damaging variants in patients with CHD compared with matched controls, with correction for the analyses of all protein-coding genes (P ≤ 2 × 10⁻⁶)^17,19,118. The identified genes from these cohorts have a definitive causal association with CHD (FiGS 3,4, genes shown in white font), as are the genes associated with rare familial CHD, but the genes discovered in large study cohorts have greater contribution to the overall population prevalence of CHD. A total of 18 genes were significantly enriched in these two WES cohorts: ADNP, ANKRD11, CDK13, CHD4, CHD7, DDX3X, DYRK1A, FLT4, KMT2A, KMT2D, NOTCH1, NSD1, PACS1, PRKD1, PTPN11, RBFOX2, SMAD6 and TAB2 (REFS^{17, 19, 73}). Notably, despite exclusion of patients with syndromic CHD at enrolment, the analysis identified variants in CHD7 (CHARGE syndrome), KMT2D (Kabuki syndrome), NSD1 (Sotos syndrome) and PTPN11 (Noonan syndrome), highlighting the possibility that some variants associated with syndromic CHD result in formes frustes of syndromic phenotypes that escape clinical recognition. Variants in CDK13, CHD4 and PRKD1 are enriched in syndromic CHD¹⁹, whereas SMAD6 variants are more prevalent in isolated CHD¹¹⁹.

Fig. 3 |. Definitive and candidate genes for congenital heart disease: genes with LOF variants.

Genes with a definitive association with congenital heart disease (CHD) identified in whole-exome sequencing analyses of patients with CHD^17,19 are significantly enriched for damaging variants in patients with CHD compared with controls (binomial P ≤ 2 × 10⁻⁶) after Bonferroni correction (for 20,440 human genes). Candidate genes for CHD (P < 0.05) were identified by binomial comparison of per-gene de novo loss-of-function (LOF) variants identified in combined data from 4,891 CHD trios^17,19,118 (patients and unaffected parents) and 1,789 unaffected trios (from the Simons Simplex Collection cohort¹²⁸). Enrichment for rare, transmitted LOF variants was determined by comparing 4,443 patients with CHD with 141,456 individuals without CHD included in the Genome Aggregation Database cohort¹²⁹. Candidate genes were restricted by genetic constraint (probability of LOF intolerance (pLI) >0.75), high expression (>50th percentile) in the developing heart^18,130, and identification of three or more LOF variants in CHD cohorts^17,19,118 (accounting for differences in variant frequencies). Definitive (white font) and candidate (black font) genes with a nominal increase in LOF variants (scale 0–15 variants per gene; listed in Supplementary Table 1) among patients with CHD compared with participants without CHD are shown, ordered by percentile rank for expression in the mouse developing heart and by genetic constraint (pLI). Genes without expression data are not shown. See the Online Mendelian Inheritance in Man database for the clinical phenotypes associated with the genetic variants.

Fig. 4 |. Definitive and candidate genes for congenital heart disease: genes with damaging missense variants.

Genes with a definitive association with congenital heart disease (CHD) identified in whole-exome sequencing analyses of patients with CHD^17,19 are significantly enriched for damaging variants in patients with CHD compared with controls (binomial P ≤ 2 × 10⁻⁶) after Bonferroni correction (for 20,440 human genes). Candidate genes for CHD (P < 0.05) were identified by binomial comparison of per-gene de novo deleterious missense variants identified in combined data from 4,891 CHD trios^17,19,118 (patients and unaffected parents) and 1,789 unaffected trios (from the Simons Simplex Collection cohort¹²⁸). Enrichment for rare, transmitted damaging missense variants was determined by comparing 4,443 patients with CHD with 141,456 individuals without CHD included in the Genome Aggregation Database cohort¹²⁹. Candidate genes were restricted by genetic constraint (probability of loss-of-function intolerance (pLI) >0.75), high expression (>50th percentile) in the developing heart^18,130, and identification of eight or more damaging missense variants in patients with CHD^17,19,118 (accounting for differences in variant frequencies). Definitive (white font) and candidate (black font) genes with a nominal increase in damaging missense variants (scale 0–50 variants per gene, except RYR3, with 51 variants identified; all genes listed in Supplementary Table 1) among patients with CHD compared with participants without CHD are shown, ordered by percentile rank for expression in the mouse developing heart and by genetic constraint (pLI). Genes without expression data are not shown. See the Online Mendelian Inheritance in Man database for the clinical phenotypes associated with the genetic variants.

Previous studies highlight the biological relevance of some definitive CHD-associated genes. Null mutations of the murine homologues of SMAD6 and NOTCH1 cause heart malformations in mice^120,121. The levels of RBFOX2 transcripts are reduced in some tissues from patients with CHD compared with control individuals¹²². Finally, the protein encoded by DYRK1A is a therapeutic target for neurodevelopmental phenotypes in individuals with trisomy 21 (REF¹²³). Recurrent variants in other genes have yielded new phenotype correlations. NOTCH1 and FLT4 variants cause tetralogy of Fallot, accounting for 7% of patients in one study⁷³. Genes previously associated with intellectual disability, including ADNP, ANKRD11 (KBG syndrome), CDK13 and DDX3X (REFS^124–127), are now recognized also to be associated with CHD^124–127. However, the roles of many definitive CHD-associated genes in cardiac development remain unknown.

Candidate CHD-associated genes with damaging variants.

WES data predict that approximately 400 genes contribute to CHD¹¹⁶, far more than are known today. Some of these genes probably harbour damaging variants in patients with CHD, but without statistical significance for classification as a definitive CHD-associated gene. To address the limitations in statistical power due to small sample sizes, we performed meta-analyses of the WES data from the two large cohorts of patients with CHD described in the previous section^17,19,73,118. Patients with aneuploidies or 22q11 deletions were excluded from the analysis. WES data were jointly analysed and compared with WES data from a cohort of individuals without CHD¹²⁸ and the Genome Aggregation Database cohort¹²⁹. We identified 132 candidate genes for CHD (FiGS 3,4), including 66 enriched for LOF variants and 78 enriched for deleterious missense variants.

The findings of this meta-analysis provide further support for the roles of CHD-associated genes in transcriptional regulation (TABLE 2; Supplementary Tables 1, 2). Many of the candidate genes for CHD with LOF variants have functional annotations as chromatin modification, transcription factors or RNA processing, and some have been previously implicated in CHD or in organogenesis, accounting for associated extracardiac phenotypes. In the group of definitive and candidate genes for CHD related to the regulation of gene expression, the presence of extracardiac phenotypes was more common in patients with LOF variants than in those with damaging missense variants^17,19 (Fig. 5a). Notably, these analyses confirmed that in genes related to chromatin modification, de novo LOF variants were enriched in patients with CHD with extracardiac abnormalities, whereas transmitted LOF variants were enriched in patients with isolated CHD (Fig. 5b). These observations might imply cardiac-specific functions for some chromatin-modifying proteins.

Table 2 |.

Functional assignment of 132 definitive and candidate genes for CHD

Category	Functional class	CHD-associated genes enriched for variant type (%)		Example genes associated with syndromic and non-syndromic CHD
		Loss of function	Damaging missense	Gene	Phenotypea
Gene regulation	Chromatin modification	15	19	KMT2D	Kabuki syndrome
				CHD7	CHARGE syndrome
				KAT6B	Genitopatellar syndrome
	Transcription factor	8	8	GATA4	Isolated atrial, ventricular or atrioventricular septal defect, tetralogy of Fallot
				TBX5	Holt–Oram syndrome
	mRNA processing	8	3	CDK13	Distinctive faces, intellectual disability
	DNA binding	3	1	CLSPN	Xeroderma pigmentosa
Signalling	Extracellular matrix	8	12	COL3A1	Polymicrogyria
	Cilia	0	1	AMBRA1	Corpus callosum agenesis, seizures, cataracts, pigmentation defects, cardiomyopathy, variable immune defects
	Cytoskeleton	12	13	NF1	Neurofibromatosis
	Signal transduction	29	23	PTPN11	Noonan syndrome
				JAG1	Alagille syndrome
Other	Ion binding, transmembrane transport, oxidoreductase activity	18	21	CLCN7	Hypopigmentation, organomegaly, delayed myelination, developmental delays

Genes associated with congenital heart disease (CHD) identified in the analysis shown in FIGS 3 and 4 were mapped to gene ontology terms found in the Gene Ontology database. The gene ontology categorization of 132 genes and the identified variants are provided in Supplementary Table 1.

^aPhenotypes are detailed at Online Mendelian Inheritance in Man.

Fig. 5 |. Relationship between extracardiac features, genotypes and functions of CHD-associated genes.

a | In a cohort of 2,871 patients with congenital heart disease (CHD) for whom whole-exome sequencing and clinical data were available¹⁷, extracardiac abnormalities are more common among patients with CHD with loss-of-function (LOF) variants (blue bars) in genes encoding proteins with gene-regulatory functions than among patients with LOF variants in genes related to signalling or other categories (P=6.36 × 10⁻⁵) or among patients with damaging missense variants (red bars) in regulatory genes (P=1.88 × 10⁻³)^17,19. b | For LOF variants in genes encoding a protein involved in chromatin modification, extracardiac abnormalities are more common among patients with CHD with de novo variants than in patients with an inherited variant (P=4.04 × 10⁻⁵).

Analyses of larger cohorts of patients with CHD and reference cohorts composed of individuals of diverse races and ethnicities and studied using newer techniques^17,19,130 and with functional assessment of deleterious missense variants are expected to reclassify some candidates genes for CHD as definitive, improve genotype-phenotype correlations and increase our knowledge of their precise roles in cardiac development.

Beyond rare monogenic coding variants

Aneuploidy, CNVs, indels and damaging coding variants have been identified in only 45% of patients with CHD. These data, combined with the recurrence rates of 3–7% for CHD in the offspring of patients with CHD^131–135, indicate that monogenic variants cannot fully account for human CHD^17,18,116. Other genotypes, including common and mosaic variants as well as variants in non-coding regulatory regions, are have been predicted to cause or contribute to CHD.

Common variants.

Polymorphic loci across the genome are analysed in genome-wide association studies to define risk alleles for traits of interest with the use of imputation of genomic sequences from low-cost, single-nucleotide polymorphism (SNP) arrays to provide genome-wide data or directly by WGS. Genome-wide association studies in CHD have identified loci associated with tetralogy of Fallot¹³⁶, left-sided CHD lesions^137–139 and atrial septal defects¹⁴⁰. Protection from CHD is associated with a variant in the FIGN locus¹⁴¹ (encoding fidgetin), perhaps reflecting the reported association between increased fidgetin expression and increased folate transport and utilization and that folate deficiency is associated with developmental defects. Moreover, SNPs in GPR98 (encoding an adhesion G protein-coupled receptor) are enriched in patients with tetralogy of Fallot with the 22q11 deletion¹⁴², indicating that common variants might modify the malformations caused by pathogenic genotypes.

Mosaic variants.

Damaging de novo variants that arise after initial zygote formation result in mosaicism, with the variant genotype present in only a subset of cells. Mosaic variants are identified by an alternative allele fraction significantly lower than the 50% expected for a heterozygous genotype¹⁴³. Interrogation of data from WES of saliva or blood-derived samples have identified likely pathogenic mosaic variants in <1% of patients with CHD (25 of 2,530), indicating a fairly small role for mosaic variants in CHD^144,145. Similar rates of mosaicism occur in autism spectrum disorder and neurodegenerative disorders¹⁴⁴. Emerging strategies to better define CHD-associated damaging mosaic variants include high-depth WGS, which provides more uniform genome coverage than WES, single-cell sequencing to detect low-level mosaicism, particularly when combined with phasing of nearby SNPs to exclude polymerase-generated errors^146,147, and analyses of CHD tissues.

Probability of LOF intolerance.

(pLI). A measure of evolutionary constraint estimated by the ratio of observed LOF variant alleles in the gnomAD cohort compared with the expected number of LOF variants on the basis of mutation rate, cohort size and sequence context.

Genome-wide association studies.

A statistical genetic analysis approach that associates common genetic variants, often with a minor allele frequency of ≥5%, with quantitative and qualitative traits. Given the number of genomic loci, a commonly accepted significance threshold of P< 5 × 10⁻⁸ is used regardless of how many loci are included in the analysis.

Alternative allele fraction.

The proportion of total nucleotide reads at a particular genomic position that represent a non-reference allele. For example, in a sequencing library containing ten reads of the reference allele A and ten reads of the alternative allele T, the alternative allele fraction would be 0.5 (10/20).

Whole-genome sequencing

Contemporary WGS provides technical advantages over WES, including avoiding the need for PCR amplification, a more uniform coverage of the genome and superior validation rates for SNPs and CNVs^148–149. In addition to exome sequences, WGS allows the interrogation of rare or common mitochondrial sequences, microRNAs, non-coding RNAs, and promoter and regulatory sequences. Successful WGS analyses of patients with CHD include the identification of the genome-wide association of a MYH6 variant with aortic coarctation¹⁵⁰ and the identification of increased rates of damaging variants among VEGF pathway genes in adults with tetralogy of Fallot with a pulmonary valve absent compared with adults with tetralogy of Fallot with a pulmonary valve present⁷⁴.

Non-coding sequence variation via WGS.

Approximately 1% of the genome encodes protein and the remaining 99% is non-coding sequences that comprise promoters, enhancers, elements that influence chromatin topology or binding site accessibility for RNA or protein factors^151–154 and other unknown functions. Damaging variants in some of these non-coding sequences might cause or influence the expression of CHD-associated genes. This hypothesis is based on the topological association between non-coding regions and gene promoters during cardiomyocyte differentiation in vitro^155–157 and evidence that assessing non-coding sequences within CNVs can help define crucial regions involved in CHD¹⁵⁸. Indeed, focused analyses of non-coding sequences flanking TBX5 identified three putative enhancers, one of which harboured a homozygous variant in a patient with CHD¹⁵⁹. Compared with the normal sequence of the putative enhancer, the presence of the homozygous non-coding variant abrogated cardiac expression in model organisms¹⁵⁹.

A genome-wide analysis of de novo variants in non-coding sequences in CHD focused on variants in non-coding elements that were likely to participate in the regulation of gene expression with the use of two related approaches¹⁶⁰. WGS was used to identify variants by scores derived from machine learning analysis of diverse genomic annotations or by location within regions of active transcription during human induced pluripotent stem cell (hiPSC) differentiation into cardiomyocytes (hiPSC-CMs). Functional assays validated the candidate non-coding variants prioritized by both strategies. Additional analyses showed that non-coding variants within binding sites of RNA-binding proteins were increased among patients with CHD compared with control individuals. Collectively, these data indicate the potential for de novo non-coding variants, particularly within enhancers and RNA-binding protein sites, to contribute to 4–12% of CHD. Although a first step, the landscape of enhancer elements and RNA-binding protein sites across the genome is huge and considerably more analyses of larger cohorts will be needed to validate and extend these initial observations.

Structural variation analyses using WGS.

WGS enables the identification of structural variants with greater resolution than that of previous methodologies and expands the repertoire of identified LOF variants in human disease^{129,161–164}. Previous studies have shown that small structural variants can cause CHD, including an 8-kb deletion in FLT4 that has been associated with tetralogy of Fallot⁷⁴ and a 10-kb deletion mediated by a retrotrans-poson mobile element (Alu element) insertion in CHD7 that has been associated with CHARGE syndrome¹⁶⁵. WGS can also detect genomic inversions and complex rearrangements. These structural variants were identified in patients with unexplained Alagille syndrome and affected the expression of the JAG1 and NOTCH2 genes¹⁶⁶. Advances in structural variant algorithms to interrogate WGS are expected to increase substantially the identification of deletions, insertions, translocations and other genomic rearrangements in CHD.

Future genomic applications.

Unexplored genome features provide ongoing opportunities to improve our understanding of CHD genetics and to explain the 55% of unexplained cases of CHD. WGS affords the opportunity to genotype more completely an individual’s SNPs and structural variants and, therefore, the investigation of oligogenic causes of CHD (encompassing both rare and common damaging variants) is now possible. Expanded genome-wide variant analyses might contribute insights into CHD penetrance and phenotypic diversity, both in familial cases and among unrelated patients with shared CHD genotypes. How exposures influence the consequences of a CHD genotype remains an untapped and fruitful area for future research, given the prominence of CHD-associated genes related to chromatin remodelling and the evidence for environmental influences on chromatin structure^167,168. Exploring these complexities might provide information on enigmatic features of CHD, including epidemiological evidence that the recurrence rates of CHD in families (3–7%)^131–135 are lower than those predicted from the current knowledge of CHD genotypes, which would anticipate 25% and 50% recurrence for autosomal recessive and dominant disorders, respectively

Integrating biology to infer causality

The discovery of genes and variants with a potential causal link to CHD combined with analyses of gene expression during development and functional studies in model systems provide crucial evidence for assigning causality in the clinical setting^169,170.

Insights from single-cell transcriptomics.

The advent of single-nuclear RNA sequencing and single-cell RNA sequencing (scRNA-seq) has enabled a detailed characterization of the many cells that populate the mouse developing heart¹⁷¹ and human mature heart^172,173. Data from these studies have identified transcriptional heterogeneity between different cell lineages and distinct cell states within each lineage that were unappreciated in analyses of bulk RNA sequencing. For example, the human healthy adult heart contains four subsets of ventricular cardiomyocytes and five subsets of atrial cardiomyocytes¹⁷². These data raise new questions regarding the developmental origins, plasticity and physiological functions of distinct cell subsets, which will propel ongoing studies. scRNA-seq studies of human fetal hearts, although more limited in scope than studies on adult hearts, have identified transcriptional signatures of cardiac muscle cells (TNNI3 and TNNT2), fibroblast-like cells (COL1A1, COL1A2 and POSTN), endothelial cells (PECAM1 and KDR) and valvular cells (SOX9)¹⁷⁴. Genes of the Notch signalling pathway are highly enriched in endocardial cells at gestational week 7, when compaction of the myocardium occurs, whereas genes related to the BMP pathway are expressed in endocardial and fibroblast-like cells from gestational week 5 to week 25, reflecting periods of endocardial-to-mesenchymal transition¹⁷⁴.

Analyses of scRNA-seq from the developing mouse embryo complement and extend the limited datasets available from human tissues. These datasets revealed the cellular complexity and transcriptional profiles of lineages derived from three germ layers: early lateral plate mesoderm, paraxial mesoderm and neural crest cells¹⁷¹. Expression of Mesp1 in mouse cells at embryonic day (E) 6.75 to E7.5 indicates a loss of pluripotency and the acquisition of cardiac lineage commitment¹⁷⁵. Hand2 expression initially demarcates the SFF but sub-sequently specifies cells of the outflow fract⁴¹. At E9.5 and beyond, cardiomyocytes exit the cell cycle and begin to show chamber-specific gene expression patterns^174,176.

These datasets also demonstrate the considerable crosstalk between cell lineages that promotes cardiac differentiation and morphogenesis. The use of these resources to interrogate the temporal and lineage-specific expression of genes that harbour damaging variants linked to CHD provides more granular detail than is feasible from transcriptional analyses of bulk cardiac tissues. FIGURE 6 depicts scRNA-seq data from embryonic germ-layer derivatives that contribute to heart development in mice¹⁷¹ (FIG. 6a), showing the temporal emergence of cardiomyocytes, endothelial cells and neural crest cells (FIG. 6b). Definitive and candidate genes for CHD identified in the meta-analysis described in the previous sections show diverse lineage-specific and temporal expression in embryonic cells that are crucial for normal cardiovascular development (FIG. 6c). As expected, genes with damaging variants identified in patients with CHD with tetralogy of Fallot⁷⁴, including Kdr, Flt1 and Flt4, have high expression in E8.25–E8.5 endothelial cells, consistent with the known functions of the encoded proteins in the VEGF pathway. By contrast, three genes, Rpl19, Rps24 and Rps26, that have been associated with Diamond–Blackfan anaemia and CHD¹⁷⁷ have variable cellular and temporal expression in the mouse developing heart. Rpl19 and Rps26 are expressed in mesenchyme at E7.5–E7.75, and Rps24 is enriched in cardiomyocytes at E8.5; therefore, these genes might have differing roles in cardiac development. Finally, genes that encode chromatin-modifying proteins with integral functions in cardiac development and that have been show to harbour damaging mutations in patients with CHD (Chd4, Chd7, Kat6a, Kat6b, Kmt2 and Kmt2d) show diverse cellular and temporal expression profiles in the mouse developing heart, implying that altered transcriptional dynamics in many cell types probably results in CHD. We anticipate that the resolution provided by scRNA-seq datasets will provide the context to understand the role of poorly studied genes that are enriched for damaging variants in patients with CHD and will aid in identifying the pathways that are perturbed in CHD. These insights are expected to improve our understanding of the morphological consequences of damaging variants and the cellular deficits that contribute to lifelong adverse events in patients with CHD.

Fig. 6 |. Cardiac lineage-specific and temporal expression of definitive and candidate genes for CHD.

2D (UMAP) representations of transcriptional profiles of ceils involved in cardiogenesis labelled by cell lineage (part a) and mouse embryonic day (E7.0–E8.5; part b). Expression profile in the mouse embryos of definitive and candidate genes for congenital heart disease (CHD) that were identified in a meta-analysis of whole-exome sequencing data from patients with CHD^17,19 (part c), classified by lineage and mouse embryonic stage. Definitive and candidate genes for CHD with a probability of loss-of-function intolerance (pLI) of >0.5 and the highest enrichment for loss-of-function or damaging missense variants in patients with CHD versus the Genome Aggregation Database are shown (84 genes in total). Definitive genes identified by statistical enrichment in case–control studies are shown in red boxes. Data adapted from a study on single-cell RNA sequencing of whole mouse embryos during organogenesis¹⁷¹.

Insights from induced pluripotent stem cell models of CHD.

hiPSCs, derived from either patients or CRIPPR–Cas9 editing of isogenic cell lines, provide an experimental approach in human cells to ascertain the functional consequences of damaging variants identified in patients with CHD. The differentiation process of hiPSC-CMs demonstrates properties consistent with the FHF^178,179 or SHF, cardiac neural crest or pre-valvular endocardial cells^180–182 that enable analyses of early developmental stages, such as the differentiation of mesoderm progenitor cells into early cardiomyocytes¹⁷⁸. When combined with epigenetic and transcriptional profiling, differentiating hiPSC-CMs provide a window into early cardiogenesis that is difficult to achieve with model organisms^43,180–182.

The molecular characterization of mutant hiPSCs can reveal the functional consequences of CHD-associated variants on gene expression, chromatin status and differentiation states. Although these in vitro assays lack the influences of in vivo morphology and haemodynamics, hiPSC-CMs have provided insights into CHD pathogenesis. hiPSCs with GATA6 LOF variants show disruption of crucial downstream regulators associated with outflow tract development (KDR and HAND2), accounting for the association between GATA6 LOF variants and outflow tract malformations in patients⁴³. By contrast, hiPSCs with a GATA6 missense variant that is recurrently identified in patients with both CHD and pancreatic agenesis have a broadly disrupted epigenetic landscape and augmented retinoic acid signalling⁴³. Similarly, transcriptomic analyses of hiPSCs with TBX5 haploinsufficiency have revealed that gene regulatory networks controlling cardiomyocyte differentiation are sensitive to TBX5 dosage and elucidated novel genetic interactions that might influence CHD phenotypes¹⁸³. hiPSCs with a GATA4 missense variant that is associated with septal defects and pulmonary stenosis show loss of GATA4–TBX5 interactions and inappropriate expression of endothelial cell genes in iPSC-CMs⁸⁴.

The use of hiPSCs also shows promise for the assessment of non-coding variants in CHD, overcoming the barriers of the limited conservation of non-coding sequences between species that impede analyses in experimental models. Massively parallel assays designed for hiPSCs can distinguish probable contributory non-coding variants from the vast numbers of benign polymorphisms and can help identify the associated genes¹⁸⁴. Moreover, epigenetic and transcriptional profiling of hiPSCs during cell lineage differentiation provides a window into early cardiogenesis that is difficult to achieve in model organisms¹⁵⁶.

Translating genomics into the clinic

The landscape of clinical genetic testing in patients with CHD is rapidly evolving due to technical advances in testing methodologies and identified associations between genotypes and longitudinal clinical phenotypes. Traditional genetic testing strategies have largely focused on syndromic CHD or selected CHD phenotypes that are strongly associated with a particular genotype (such as the 22q11 deletion)¹⁸⁵. However, with the increasing availability and cost-effectiveness of clinical WES, gene-based diagnosis of both syndromic and non-syndromic CHD is emerging. Given that WES provides comprehensive analyses that provide information on the causes of CHD as well as detect incidental findings that are clinically actionable and/or of unclear significance, informed consent and guidance of genetic counsellors or experienced clinicians is essential. Moreover, with ongoing discoveries, the evidence for pathogenicity of damaging variants in genes associated with CHD requires criteria similar to those proposed by ClinGen and the American College of Medical Genetics and Genomics^{169,170,186,187}.

The identification of definitive causal CHD variants influences patient care. Both the family of a newborn baby with CHD and adults with repaired or palliated CHD benefit when the genotype helps to inform the risk of recurrence, prenatal genetic counselling and pre-implantation genotyping¹⁸⁸. Increasingly, the genotype can indicate the risk of associated extracardiac anomalies or comorbidities such as neurodevelopmental delay¹⁸ and highlight the need for surveillance for deleterious postoperative outcomes¹⁸⁹ and the future development of cancers^190–192.

CHD associated with extracardiac anomalies or neurodevelopmental delay.

The presence of an extracardiac anomaly or neurodevelopmental delay increases the likelihood of detecting a pathogenic CHD genotype from 8% to 22%^18,19, particularly in genes encoding proteins involved in modifying chromatin¹⁷. WES studies of cohorts with neurocognitive disorders have also identified enrichment of damaging variants in chromatin-modifying genes^18,193, indicating a strong genetic contribution to the association between CHD and neurodevelopmental delay. Therefore, the clinical identification of chromatin-related damaging genotypes in a newborn baby with CHD can inform pre-emptive interventions that might improve the baby’s learning and social skills.

De novo CNVs influence perioperative outcomes.

CHD genotypes can help identify high-risk patients for cardiac surgical interventions. Patients with CHD with 22q11.2 deletions have longer cardiopulmonary bypass times, more re-operations and longer intensive care stays than patients with CHD without this deletion^194,195. Patients with CHD with other large CNVs (>300 kb) have a higher risk of death or need for transplantation than patients with CHD without large CNVs¹⁹⁶. WES studies are extending these observations. Patients with syndromic or non-syndromic CHD with de novo CNVs or damaging single-nucleotide variants have increased time to extubation and decreased transplantation-free survival¹⁸⁹. These adverse outcomes are most pronounced among patients with CHD without extracardiac anomalies. Given that some of these risk-associated genomic regions are also linked to cardiac trabeculation and myocardial performance¹⁹⁷, the relationships between genotype and morphological structure might underlie the risk of heart failure and transplantation in some patients with CHD.

Risk of cancer in patients with CHD.

Adults with CHD have a higher cancer prevalence than the general population (1.4-fold to 2-fold)^190–192. In addition to increased radiation exposure from therapeutic interventions¹⁹⁸, certain CHD genotypes might also increase the risk of cancer¹¹⁸, because many genes with damaging variants associated with CHD are also recognized as genes associated with the risk of cancer. This association might be intertwined with the fundamental and broad developmental role of some CHD-associated genes, because patients with CHD with damaging variants in genes linked to cancer risk often have extracardiac manifestations.

Future clinical applications.

Investigation of the relationship between causal CHD-associated variants and clinical outcomes remains an area of enormous opportunity. How do variants in specific genes or defects in signalling pathways relate to the variability observed in CHD phenotypes? Do variants in genes linked to CHD or other genetic variants affect survival after surgical palliation or repair in patients with CHD? How do damaging variants that result in structural heart disease relate to measures of long-term cardiovascular function? To address these areas of inquiry, large genome-wide sequencing datasets harmonized with comprehensive, high-quality clinical data will be necessary. Given the expense and magnitude of the datasets necessary to address these questions, multicentre collaborative efforts are likely to be crucial in accelerating these areas of research.

Conclusions

Technological advances have propelled the discovery of the genetic architecture of CHD, which encompasses damaging germline and mosaic variants, copy number and structural variants that encompass many genes and variants in non-coding sequences. The influence of these discoveries extends well beyond the elucidation of the molecular basis for this common birth defect. These insights define crucial developmental genes and pathways that participate in the formation and function of the heart and many other organs. CHD-associated variants also identify genetic variants that promote neurocognitive dysfunction and cancer and confer distinct risks on patients who need surgical intervention for CHD.

Despite our current understanding of the genetic architecture of CHD, there is still much to learn. Aetiologies remain unknown in >50% of patients with CHD. New directions are needed to address this challenge. The identification of aberrant gene expression from direct analyses of human CHD tissues, comprehensive assessments of non-coding sequences and oligogenic variants, and deciphering the effect of environmental exposures might be productive. The success of these and other strategies holds great opportunities not only to improve the diagnosis, classification and clinical care of patients with CHD, but also to enrich the fundamental knowledge in developmental and genome biology.

Key points.

• A genetic risk contributes substantially to congenital heart disease (CHD), and variantsin >400 genes are estimated to cause human CHD.

• Analyses of large cohorts of patients with CHD have empowered the identification of novel genes associated with human CHD.

• Non-coding variants contribute to CHD, in part by altering the activity of cardiac developmental enhancers.

• Despite the genetic insights gained so far, a definitive cause is not identified in half of the cases of CHD.

• Integration of genomic data with developmental biology promises to increase our understanding of the pathogenesis of CHD.