The Genetic Landscape of Hypoplastic Left Heart Syndrome

Abstract

Hypoplastic left heart syndrome (HLHS) is one of the most lethal congenital heart defects, and remains clinically challenging. While surgical palliation allows most HLHS patients to survive their critical heart disease with a single-ventricle physiology, many will suffer heart failure, requiring heart transplantation as the only therapeutic course. Current paradigm suggests HLHS is largely of hemodynamic origin, but recent findings from analysis of the first mouse model of HLHS showed intrinsic cardiomyocyte proliferation and differentiation defects underlying the left ventricular (LV) hypoplasia. The findings of similar defects of lesser severity in the right ventricle suggest this could contribute to the heart failure risks in surgically palliated HLHS patients. Analysis of 8 independent HLHS mouse lines showed HLHS is genetically heterogeneous and multigenic in etiology. Detailed analysis of the Ohia mouse line accompanied by validation studies in CRISPR gene-targeted mice revealed a digenic etiology for HLHS. Mutation in Sap130, a component of the HDAC repressor complex, was demonstrated to drive the LV hypoplasia, while mutation in Pcdha9, a protocadherin cell adhesion molecule played a pivotal role in the valvular defects associated with HLHS. Based on these findings, we propose a new paradigm in which complex CHD such as HLHS may arise in a modular fashion, mediated by multiple mutations. The finding of intrinsic cardiomyocyte defects would suggest hemodynamic intervention may not rescue LV growth. The profound genetic heterogeneity and oligogenic etiology indicated for HLHS would suggest that the genetic landscape of HLHS may be complex and more accessible in clinical studies built on a familial study design.

Congenital heart disease (CHD) is one of the common birth defects, affecting up to 1% of live births and is a leading cause of pre- and postnatal death. With recent advances in medical and surgical management, the population of CHD survivors is growing and there are now more adults with CHD in the US than infants born each year with CHD [1]. Hypoplastic left heart syndrome (HLHS), is a severe CHD involving hypoplasia of the left ventricle (LV), aorta, and mitral valve. Without treatment, it is uniformly fatal, but it is now survivable with a three-stage surgical palliation that recruits the right ventricle (RV) to be the systemic pumping chamber [2, 3]. This single-ventricle circulation now allows more HLHS patients to survive their critical heart disease, but the 5-year survival rate is still only 50–70% [4, 5].

The current clinical paradigm assumes HLHS is a valve disease with LV hypoplasia arising from the hemodynamic effects of restricted aortic blood flow. As most HLHS diagnosis is now made prenatally, pre-conception risk assessments may provide opportunities for in utero intervention [6-8]. Indeed, a clinical trial has been pursued with in utero aortic valvuloplasty performed to restore aortic blood flow in fetuses with evolving HLHS. However, this 10-year clinical trial has yielded little evidence of fetuses with evolving HLHS undergoing conversion from single to biventricular physiology [9]. Substantive progress in the development of therapeutic intervention will require deeper understanding of the underlying causes for the cardiac maldevelopment and the pathogenic mechanism causing the LV hypoplasia.

One of the continuing challenges underlying the poor prognosis for HLHS patients is clinical management of the single RV [10]. This RV serving as the systemic pumping chamber performs well for some patients, others will develop heart failure, resulting in the need for cardiac transplantation. The reason for the differing outcomes remains unknown and as more HLHS patients survive their critical heart disease, there is increasing urgency to understand the factors driving heart failure risk in HLHS [11-13]. Such insights are needed to guide the development of therapeutic interventions to improve the long-term prognosis for patients with HLHS.

Genetic Etiology of HLHS

The current poor understanding of the disease mechanism in HLHS in part reflects the paucity of knowledge regarding the genetic causes of HLHS and the lack of a genetic animal model of HLHS. HLHS is a CHD that is particularly well documented to have a genetic etiology. Family studies have shown a high recurrence risk for CHD among siblings [14-20], a high relative risk of 36.9 reported for first-degree relatives, and an overall heritability of 0.71–0.90 [21]. HLHS is also highly associated with various syndromes such as Turner and Jacobsen [22, 23], as well as trisomy 13, 18, and 9 [24]. Analysis in small kindreds or limited HLHS cases has identified several candidate genes including heterozygous mutations in NKX2.5 [25], GJA1 [26], NOTCH1 [27], HAND1 [28], and RBFOX2 [29], while two recent reports support a recessive inheritance of HLHS [30, 31]. Findings from two family-based genetic studies suggest HLHS and associated CHD lesions are genetically heterogeneous with an oligogenic etiology. In one study, statistical modeling suggested HLHS most likely has a digenic etiology [19, 32]. Although human epidemiologic studies support a genetic origin of HLHS, until recently, no HLHS mouse model had ever been generated. This deficiency has recently been addressed with the recovery of HLHS mutant mice from a large-scale mouse mutagenesis screen. This has led to new insights into the genetic and developmental etiology of HLHS [33].

Recovery of HLHS Mutant Mice

A genetic model of HLHS was generated for the first time with recovery of HLHS mutant mice from a large-scale ENU mutagenesis screen [29, 33, 34]. The previous failure to generate HLHS mutant mice was thought to reflect the inadequacy of the short gestation period of mice (18.5 days) to orchestrate the hemodynamic remodeling required for the evolution of HLHS. However, the difficulty in generating HLHS mutant mice was actually due to the obligate multigenic etiology of HLHS and its profound genetic heterogeneity [19, 33-35]. Thus in a large-scale mouse forward genetic screen, we screened 100,000 fetal mice from 3000 G1 pedigrees using fetal echocardiography. We recovered nine HLHS mutants from among 100,000 fetuses scanned. They are derived from eight independent mutant pedigrees, with HLHS being the rarest CHD phenotype recovered in the screen.

HLHS mutants were diagnosed by fetal hemodynamic assessments, which typically showed a narrowed aorta (blue, Fig. 1b, left panel), and with 2D imaging showing a small LV with poor contractility (Fig. 1b) (see Supplementary Movies 1, 2 in Liu et al. [34]). Confirmation by necropsy and histopathology showed a small LV with a hypoplastic aorta and mitral valve (MV)—the three essential features of HLHS (Fig. 1c, d; Supplementary Movie 3, 4 in Liu et al. [34]). The LV is often muscle-bound with obliteration of the chamber lumen (Fig. 1d). Analysis of the HLHS mutant mice from the eight lines showed the extent of LV and aortic/MV hypoplasia spanned a spectrum (Fig. 2), similar to the findings clinically. Unlike other CHD phenotypes recovered in the screen, HLHS was observed with a low transmission ratio that indicated non-Mendelian or complex genetic inheritance.

Fig. 1.

HLHS phenotype in Ohia mutants. a, b Ultrasound color-flow imaging of normal fetus (a), showing robust flow from the aorta (Ao) and pulmonary artery (PA). In the HLHS mutant (b), the aorta showed only a narrow flow stream, whereas the pulmonary artery showed robust flow. 2D imaging revealed hypoplastic LV (b), as compared with the normal-sized LV in the control (a). c Newborn (P0) or E16.5 hearts from wild-type and HLHS mutants. Hypoplastic aorta and LV are visible in the HLHS mutant. d Histopathology showing the cardiac anatomy of HLHS mutant and littermate control at birth (P0) and E14.5. Compared with controls, the HLHS mutant exhibited hypoplastic aorta and aortic valve atresia, hypertrophied LV with no lumen, and MV stenosis and patent foramen ovale (FO), arrowhead. LA left atrium, RA right atrium, A anterior, P posterior, L left, R right, Cr cranial, Cd caudal. Scale bars: a, b 0.5 mm; c 1 mm; d 0.5 mm. Adapted with permission from Liu et al. [34]

Fig. 2.

HLHS phenotypes observed in six of the eight HLHS mutant lines recovered from the mutagenesis screen. Top row shows necropsy photo and bottom, confocal histopathology of the same heart, showing a wide spectrum of disease associated with HLHS in these different mutant lines. Ao aorta, MV mitral valve, LV left ventricle, RV right ventricle, PA pulmonary artery, RA right atrium, LA left atrium. Adapted with permission from Liu et al. [34]

Hemodynamic Uncoupling of Aortic Flow and the Small LV Phenotype

The mutant line Ohia, which yielded two HLHS mutants from the initial screen, was further analyzed by extensive breeding and phenotype/genotype analyses. This confirmed the heritability of the HLHS phenotype and in addition, yielded mutants with related CHD phenotypes, such as isolated hypoplastic LV, aortic arch anomalies and hypoplastic aorta, bicuspid aortic valve (BAV), or outflow tract malalignment defects. It is of significance to note that some Ohia mutants had double outlet RV with hypoplastic LV with normal-sized aorta, or hypoplastic aorta with normal-sized LV (Fig. 3a-c). These phenotypes demonstrated the hemodynamic uncoupling of the aortic flow from the small LV phenotype, indicating HLHS in the Ohia mutant line is not driven by hemodynamic factors.

Fig. 3.

HLHS and other CHD phenotypes in the Ohia mutant line. a–c Ohia mutants with isolated hypoplasia of the aorta or LV. Some Ohia mutants exhibit DORV with hypoplasia of both the aorta and LV (a), while others can exhibit DORV with hypoplastic LV, but normal-sized Ao (b), or DORV with hypoplastic aorta, but normal LV (c). These independent occurrences of hypoplasia of the LV vs. Ao would suggest these anatomical defects are not hemodynamically driven in Ohia mutants. Scale bar, 1 mm. d–g Delineating cardiac development in Ohia HLHS mutants using fetal echocardiography. d, e Developmental profile of the growth of cardiac valves and chambers in normal vs. HLHS embryos (n = 15 embryos per group). Data shown are mean ± s.e.m. LA left atrium, LV left ventricle, RA right atrium, RV right ventricle, MV mitral valve, TV tricuspid valve, AV aortic valve, PV pulmonary valve. f Fetal ultrasound measurements show LV contractility is decreased in HLHS mutant fetuses (n = 15 embryos per group). Data shown are median with interquartile range. Wilcoxon rank-sum test, P = 0.000003. g Measurements of the foramen ovale (FO) showed FO size was not restrictive in HLHS mutants. Data shown are mean ± s.e.m. Unpaired Student’s t test: E13.5, n = 12 (control = 6, HLHS = 6) P = 0.385, t = −0.908, df = 10; E14.5, n = 18 (control = 12, HLHS = 6), P = 0.000084, t = 5.222, df = 16; E16.5, n = 27 (control = 15, HLHS = 12), P = 0.001, t = 3.7, df = 25; E17.5, n = 21 (control = 12, HLHS = 9), P = 3.4633E – 9, t = 10.262, df = 19; E18.5, n = 12 (control = 6, HLHS = 6), P = 0.453, t = −0.782, df = 10. NS not significant. Adapted with permission from Liu et al. [34]

This is supported by further longitudinal interrogation conducted with fetal echocardiography at E13.5–18, which showed only right-sided heart structures grew with development of the HLHS mutant fetuses (Fig. 3d), while normal littermates showed steady growth of all four cardiac valves/chambers (Fig. 3e) (see Supplementary Movie 1, 2 in Liu et al. [34]). The HLHS mutants also showed poor LV contractility (Fig. 3f), but with sustained left-to-right flow through the foramen ovale (FO) (see Supplementary Movie 1). The previous hypothesized premature closure or restriction of the FO driving the LV hypoplasia in HLHS was not observed in Ohia HLHS mutants (Fig. 3g) [36]. Indeed quantification using the ultrasound imaging data showed enlargement rather than restriction in the opening of the FO. Together these findings indicate that LV hypoplasia in Ohia HLHS mutants is not due to hemodynamic effects from restricted blood flow from the aorta or the FO.

Complex Genetic Etiology for HLHS

Unlike other CHD phenotypes recovered in the screen, HLHS was observed with low transmission ratio indicating non-Mendelian inheritance. Whole exome sequencing analysis for mutation recovery showed that none of the eight HLHS mutant lines have mutations in the same genes. Together these findings indicate HLHS is multigenic and genetically heterogeneous, consistent with findings from human studies. This likely explains why HLHS had not been seen in mice previously, as any single gene knockout (KO) would not suffice, nor multiple KO gene combinations unless they are gene combinations known a priori to cause HLHS. In comparison, every pedigree in the mouse mutagenesis screen by design harbored 50–100 deleterious mutations, and with the screening of 3000 pedigrees, this greatly increased the odds of having mutation combinations that can cause HLHS, accounting for our success in recovering HLHS mutant mice. Nevertheless, we note HLHS is the rarest CHD phenotype recovered in our screen, with only nine HLHS mice recovered from over 100,000 fetal mice screened.

Exome sequencing of the eight HLHS mutant lines yielded 355 coding/splicing mutations, with 61 being homozygous, and none shared between the eight lines. Interestingly, five genes related to Notch signaling were recovered—Rbpj, Ep300, Nap1l1, Ncoa1, and Tspan12, consistent with previous studies indicating an important role for the Notch pathway in the genetic etiology of HLHS. We observed that mutations in five of the HLHS mouse lines are in genes situated in 10 of the 13 human chromosome intervals linked to HLHS (Fig. 4) [35, 37, 38]. Significant enrichment is observed when two or more of the mouse HLHS genes are interrogated across these human HLHS chromosome linkage intervals (P = 5.6 × 10⁻¹⁰) [34]. Together these findings demonstrate the clinical relevance of the mutations recovered from the HLHS mutant mice. They point to a multigenic etiology for HLHS that involves a profound degree of genetic heterogeneity. This would suggest that the clinical studies into the genetic landscape of HLHS may be difficult with the use of a standard case–control study design.

Fig. 4.

Mutations recovered from HLHS mutant mice and HLHS patients. a Mouse HLHS mutations in human chromosome regions linked to HLHS and left ventricular outflow tract obstruction (LVOTO). Genes recovered from mouse HLHS mutant lines found in human chromosome regions linked to HLHS/LVOTO are shown [8, 9]. The mouse chromosome location of each gene is indicated in parenthesis. The mouse mutant line is identified as L with three or four digit line identifier. b ToppCluster-generated interactome network of multi-hit genes recovered from subjects with HLHS and HLHS mutant mice. Genes (hexagons) associated with heart development and various gene ontology biological processes (color rectangles) are shown with corresponding color. Dark blue-outlined hexagons are NOTCH-related genes. Protein–protein interactions are indicated as solid dark blue lines between genes. Adapted with permission from Liu et al. [34]

Digenic Etiology of HLHS in the Ohia Mutant Line

Extensive breeding and phenotype/genotype analysis of the Ohia mutant mice identified HLHS arising from mutations in two genes, Sap130 and Pcdha9. Sap130, sin3A-associated protein 130, is a component of the Sin3A histone deacetylase (HDAC) repressor complex, while Pcdha9 is a protocadherin mediating cell–cell adhesion well described in the development of the nervous system [39-41]. The Sap130 mutation is a point mutation that causes an anomalous splice causing an in-frame 35 amino acid deletion, while the Pcdha9 mutation is a missense (P793L) mutation. Antibody staining showed the truncated Sap130 protein is expressed and is translocated into the nucleus. Analysis of a Sap130 KO mouse revealed that Sap130 deficiency causes peri-implantation lethality, while the Ohia Sap130 mutant can survive to term. These findings suggest the Ohia Sap130 mutation is likely hypomorphic.

Analysis of the Ohia mutant mice showed that the small LV phenotype can be elicited by the Sap130 mutation alone, while Pcdha9 played an essential role in the aortic hypoplasia and valvular defects in HLHS. Interestingly, the Pcdha9 mutation alone can cause BAV (Fig. 5a-c), a phenotype well described as highly associated with HLHS and often seen among family members of patients with HLHS [19, 42]. Sufficiency of Sap130/Pcdha9 mutations in causing HLHS was validated with replication of the HLHS phenotype in transgenic mice harboring Sap130/Pcdha9 mutation generated by CRISPR gene editing [34]. The CRISPR-targeted Sap130 allele expressed the same anomalously spliced Sap130 transcripts as seen in the Ohia mutants. The Pcdha9 CRISPR-targeted allele harbored a missense mutation near the missense mutation found in the Ohia mutant, suggesting both likely exert loss-of-function (LOF) effects.

Fig. 5.

Bicuspid aortic valve phenotype in homozygous Pcdha9 mutant mice. a Echocardiography of adult Pcdha9 mutant vs. C57BL6 wild-type mice. Color-flow (top) and spectral Doppler (bottom) imaging, showing a high-velocity jet (> 3 m/s) across the aortic valve, thus indicating aortic stenosis (Supplementary Video 5). b Functional magnetic resonance imaging (MRI) showing the bicuspid aortic valve. MRI of homozygous Pcdha9 mutant mouse from (a), showing BAV with doming (double arrows) during systole (see Supplementary Videos 6, 7). LV hypertrophy, as determined by M-mode imaging (LV anterior wall/posterior wall: 0.45/0.64 mm in control vs. 0.91/0.91 mm in mutant). c Histopathology of newborn Pcdha9m/m vs. wild-type mice. Compared with wild type, the mutant mouse showed immature thickened aortic valves with increased trichrome staining, thus indicating abnormal matrix deposition. Adapted with permission from Liu et al. [34]

Regarding phenotype–genotype correlation, it is worth pointing out that in the Ohia HLHS mutant mice, we observed a spectrum of left-sided lesions as described above. This phenotype variability was replicated in the CRISPR gene-targeted mice, indicating it is inherent to the Sap130/Pcdha9 mutations. This is not unlike the often observed phenotype variability associated with most KO and mutant mouse models of CHD and likely reflects the stochastic processes regulating cardiovascular development. These findings would suggest that detailed subclassification of structural phenotypes that may have clinical utility for the management of patient care might not yield significant insights into phenotype–genotype correlations for advancing our understanding of the genetic etiology of HLHS.

PCDHA Mutation and Bicuspid Aortic Valve

BAV can be observed in Ohia HLHS mutants harboring the Sap130/Pcdha9 mutations, and also in isolation in mice harboring only the Pcdha9 mutations. BAV is the most common congenital anomaly observed in humans, affecting 1–2% of the adult human population [43]. Interestingly, a 16.7-kb deletion spanning the PCDHA9 and PCDHA10 genes is found in multiple human populations and reported to have a 1–8% allele frequency [44]. While this deletion is said to be present in unaffected individuals, we note BAV is typically not diagnosed until late adult life and would not affect reproductive fitness. Notably, African Americans have a lower frequency of the PCDHA deletion allele compared to Europeans, and they also have a lower BAV incidence (P = 0.001) [45]. Mice with deletion of Pcdha2 through Pcdha11 are reported to be viable, but have not been examined for cardiac phenotype [46, 47]. Our studies showed some Pcdha9 mutant mice with BAV are adult viable, but can succumb to sudden death from aortic insufficiency from BAV pathology. However, this is observed with low penetrance, suggesting other epigenetic or environmental factors may contribute to the emergence of pathology. Further studies are warranted to determine the risk of BAV in human subjects with the PCDHA deletion allele.

Intrinsic Cardiomyocyte Proliferation and Differentiation Defects

We investigated the causes for the HLHS-LV hypoplasia using the Ohia HLHS mutant embryos. For these studies, hearts at different development stages (E14.5 to term) were analyzed for cardiomyocyte cell proliferation using phosphohistone H3 (PH3) and Ki67. Interestingly, while cardiomyocytes labeled with PH3 increased, Ki67 positive cardiomyocytes decreased in the HLHS-LV tissue, findings reminiscent of a previous study of human HLHS fetal heart tissue [48]. In contrast, in the HLHS-RV, Ki67 was also decreased, but no change was observed for PH3. This was associated with a reduction in cardiomyocytes in anaphase–telophase in the HLHS-LV tissue, indicating a block in mitotic cell cycle progression, but this was not observed in the HLHS-RV tissue. TUNEL labeling also showed an increase in apoptosis in the HLHS-LV but not RV tissue. Electron microscopy revealed a cardiomyocyte differentiation defect, with the HLHS-LV exhibiting short myofilament bundles with more branching and poorly defined Z-bands. This was associated with a mitochondrial maturation defect indicated by a low-density matrix, sparse cristae, and a reduction in mitochondrial size and shape. While similar changes were observed in the RV, the changes were of lesser magnitude. Overall, these findings suggest the LV hypoplasia in HLHS is associated with intrinsic cardiomyocyte cell proliferation and differentiation defects linked to a mitochondrial maturation defect. These changes are not likely secondary to hemodynamic effects caused by the structural heart defects, since they can be observed at early stages of development (E14.5, equivalent to 7–8 week gestation human fetus). In addition, we recently observed these same cell proliferation and cardiomyocyte differentiation defects in cardiomyocytes generated in vitro from the differentiation of induced pluripotent stem cells (iPSC) derived from the Ohia HLHS mutant fibroblasts. Together these findings suggest that the small LV phenotype in Ohia HLHS mutants is elicited by intrinsic cardiomyocyte defects.

Metabolic Pathways Recovered by RNA-seq and ChIP-seq Analyses

Gene expression changes in the Ohia HLHS heart tissue was profiled using RNA-seq analysis of RV and LV heart tissue obtained from embryos at E13.5, 14.5, and 17.5. Gene enrichment analysis identified metabolic and mitochondria-related pathways, growth factor signaling, cancer pathway, calcium signaling, cardiac muscle contraction, and dilated/hypertrophic cardiomyopathy (Fig. 6a). Similar pathways were recovered in the HLHS-LV and RV tissues, but the HLHS-RV showed lower fold changes. Using antibody to Sap130, we also conducted chromatin immunoprecipitation sequencing analysis to identify Sap130 target genes in embryonic heart tissue. This analysis yielded many of the same pathways identified by RNA-seq, with the top pathway being metabolic pathways, followed by cell cycle, spliceosome, and pathways in cancer (Fig. 6c). We note that metabolic pathways and pathways in cancer were shared in common between the RNA-seq and ChIP-seq analyses. Also observed among Sap130 target genes are Notch pathway components such as Rbpj, Mib1, Ep300, and Notch1. Particularly notable is recovery of Meis1, a gene that regulates mitochondrial metabolism [49] and postnatal cardiomyocyte cell cycle arrest [50], and Rbfox2, a RNA-binding protein recently identified as a HLHS candidate gene [29] (Fig. 6c). Rbfox2 is known to regulate alternative splicing and RNA processing, consistent with the studies by Ricci et al. showing the dysregulation of RNA splicing and mRNA processing in human HLHS-RV tissue [51]. In the study by Ricci et al., spliceosome and various metabolic pathways were also identified as significantly affected pathways in the newborn HLHS-RV tissue, findings that overlap the pathway enrichment analysis of the Sap130 ChIP-seq data [52].

Fig. 6.

RNA-seq and Sapq130 ChIP-seq Pathway Enrichment Analysis. a–c Differentially expressed genes identified by RNA-seq analysis of Ohia mutant HLHS-LV (a) vs. RV (b) heart tissue, and ChIP-seq (c) analysis of Sap130 target genes in embryonic heart tissue were analyzed by WebGestalt KEGG pathway enrichment analysis. Interestingly, all three datasets yielded similar gene-enriched pathways. d ToppCluster network analysis of genes (hexagons) that are differentially expressed in the hearts of HLHS mice based on RNA-seq analysis, or identified as Sap130 target genes by ChIP-seq analysis of E12.5 mouse heart. Genes highlighted in yellow are associated with cardiac development terms. Associations of these genes are indicated by red edges into other functional categories that include GO biological process and cellular component, pathways, or mouse knockout allele phenotype associations as determined by ToppGene. Genes that are Sap130 targets are shown surrounded with green edges. Genes associated with other enriched functional categories such as growth factor signaling, Wnt signaling, or mitochondrial function are colored individually with matching edges, except when they are linked to cardiac morphogenesis, in which case the edges are in red. Adapted with permission from Liu et al. [34]

ToppGene Network Analysis

ToppGene network analysis [53] showed that many down-regulated genes are related to cardiac differentiation, including contractile fiber, mitochondria, cation transport and ion channels, and calcium binding/signaling. In contrast, genes up-regulated were involved in mesenchyme determination, heart development, and tube/branching morphogenesis, many overlapping with Wnt and Tgfβ signaling (Fig. 6d). The combined pathway enrichment analysis using data generated from exome sequencing, RNA-seq, and ChIP-seq analyses showed convergence of pathways indicating HLHS may arise from the disruption of early events in mesenchyme/heart development involving perturbation of Notch, Tgfβ, and Wnt signaling. The mitochondrial and cardiomyocyte differentiation defects, and premature cell cycle arrest with abnormal activation of the hypertrophy program, could underlie the muscle-bound phenotype seen in Ohia HLHS mutants. Mitochondrial fusion and calcium signaling, processes that regulate cardiomyocyte differentiation, are also known to regulate Notch signaling [54]. Regarding the restriction of defects to left-sided heart structures in HLHS, it is interesting to note Notch signaling is known to play a role in left–right patterning [55] and Brg1, a component of the SWI/SNF chromatin-modulating complex implicated in left–right patterning [56], is also known to interact with the Sin3A complex, suggesting a potential mechanistic basis for the restriction of lesions to left-sided heart structures [51-53, 55-58].

Sap130 and Sin3A Regulation of Pcdha9 and EMT

To interrogate the Sap130–Pcdha9 interactions driving the pathogenesis of HLHS, we generated a cardiomic expression atlas using the combined microarray data obtained from 118 heart-related samples at different stages of development available from public databases [57-64]. This analysis showed Pcdha9 expression in E9.5 embryos was restricted to the cardiac cushion, which we confirmed by real-time PCR analysis. In contrast, Sap130 was more broadly expressed. Using the microarray data, Pcdha9-co-regulated genes were identified. This yielded many genes associated with cardiac valve development, including Notch pathway components, genes mediating epithelial–mesenchymal transformation (EMT), and mesenchymal cell differentiation. Genes identified as Sap130 co-regulated included many associated with the mitotic cell cycle, mitotic checkpoint control, and chromatin-regulatory components (Eed, Ezh2, and Suz12) of the Polycomb repressive complex 2 (PRC2). Previous studies showed SIN3A together with HDAC, PRC2, and lysine-specific demethylase 1 (LSD1) plays important roles in the repression of cadherin to promote EMT [65-67]. Together, these findings suggest Sap130 epigenetically regulates EMT mediating cardiac cushion development. Consistent with this, we observed the PCDHA gene cluster has multiple enhancers with SIN3A-binding sites. SIN3A-binding sites also have been mapped within the PCDHA cluster by ChIP-seq analysis.

Most unexpected was the finding in the original Ohia HLHS mutant mouse, not only Sap130 and Pcdha9 mutations, but also mutations in Sin3A and Snai1 among other mutations identified by exome sequencing analysis. While Sap130 and Pcdha9 mutations were homozygous, the Sin3A and Snai1 mutations were heterozygous. Snai1 is a transcription factor known to promote EMT by forming a complex with Sin3A and HDAC to repress E-cadherin expression. The presence of mutations in Snai1, Sin3A, Sap130, and Pcdha9 all in the same HLHS mutant is unlikely to have occurred by chance. While only the Sap130 and Pcdh9 mutations are required to generate HLHS, the original Ohia mutant harboring all four mutations survived to term. In contrast, the Ohia mutant with only the Sap130/Pcdha9 mutations dies at mid-gestation (E13.5–14.5). Such mid-gestation lethality is also observed in the CRISPR Sap130/Pcdha9 gene-targeted mice. Thus the Sin3A and Snai1 mutations may act as protective modifiers suppressing the more severe deleterious effects of the Sap130/Pcdha9 mutations, thereby allowing HLHS mutants to survive to term. These findings suggest that the complex genetics of HLHS may involve the interplay of pathogenic mutations and protective modifiers that together determine the final phenotypic disease outcome.

Role of Sin3A/Sap130 and Rbfox2 in Stem Cell Survival, Muscle Differentiation, and EMT Regulation

It is notable that Sin3A/Sap130 and Rbfox2 share many biological functions in common, suggesting insights into the mechanism of HLHS pathogenesis may emerge from a better understanding of the functions of these two genes. Thus Sin3A/Sap130 and Rbfox2 are both known to regulate muscle differentiation [68, 69] and metabolism [51, 70] and both are required for stem cell survival. Hence, deficiency in either of these genes may contribute to LV hypoplasia via disturbance of stem cell survival and muscle differentiation. Additionally, as both genes play important roles in EMT [71, 72], this could drive the valve defect phenotypes associated with HLHS, with presence of mutations in Pcdha9 in combination with Sap130 increasing the penetrance of valve defects in the Ohia mutant mice.

As Rbfox2 and Sap130/Sin3A are both known to regulate splicing, this also would point to the important role of altered splicing in HLHS disease pathogenesis. Indeed alternative splicing has been shown to regulate cardiac precursor proliferation and differentiation [73], with mice deficient in Rbm24 (RNA-binding motif 24) or Srsf10 (serine–arginine-rich splicing factor 10) exhibiting early embryonic lethality with impaired cardiogenesis [74, 75]. The observation of genome wide alternative splicing in the human fetal heart further suggests a role for splicing in the regulation of heart development and the pathogenesis of CHD [76]. In addition, various cardiac diseases are observed with mutations in splicing factors [77], such as dilated cardiomyopathy in patients and mice with RBM20 (RNA-binding motif protein 20) mutations [78, 79]. This may reflect the essential role of alternative splicing in the physiological regulation of cardiac muscle differentiation and function, such as in the critical regulation of calcium handling [75, 76].

Clinical Study of the Genetic Etiology of HLHS

To gain further insights into the genetic etiology of HLHS, we conducted exome sequencing of 68 HLHS patients. Rare LOF variants were recovered and compared to LOF variants in 95 ethnically matched CEU 1000 genomes control subjects (MAF < 0.01). Genes harboring these mutations were compared to those recovered from the HLHS mutant mice. This analysis showed more genes were shared between the 68 HLHS patients and eight HLHS mutant mice than expected at random (P = 0.013; Table 1). More genes were found to be shared among the 68 HLHS patients (P = 9.1 × 10⁻²⁴), but not between the control CEU subjects [80] (Table 1). ToppGene network analysis of genes shared between the HLHS patients and mice showed enrichment of pathways that regulate heart development, chromatin, cell proliferation/cell cycle, mitochondrial associated, TGFβ/BMP and Notch signaling, and centrosome/cilia (Fig. 4b). The cilia finding is of particular interest given we previously observed a significant enrichment for cilia-related genes in our large-scale mouse forward genetic screen for CHD-causing mutations [29].

Table 1.

Sharing genes with loss-of-function mutations in HLHS patients and mice (adapted with permission from Liu et al. [34])

Comparisona		No. shared genes with unique LOF			P valueb	Inference
Group 1	Groups 2	Observed	Expected	Observed/ expected
hHLHS	mHLHS	33	22	1.50	0.013	More sharing than by chance
CEU	mHLHS	28	22	1.27	0.21	Chance sharing
hHLHS		247	69	3.57	9.1 × 10−24	More sharing than by chance
CEU		116	265	0.44	4.5 × 10−16	Less sharing in CEU than expected
hHLHS	CEU	306	453	0.67	4.2 × 10−7	Less sharing in CEU than HLHS

^aComparison includes 68 HLHS (hHLHS) patients, and 95 1000 genomes ethic-matched CEU subjects, and eight independently derived HLHS mutant mice (mHLHS). The HLHS patients had 1278 genes with 1736 unique LOF, while the 1000 genomes CEU subjects had 1372 genes with 1550 unique LOF. Unique LOF are seen in only a single subject in the HLHS or CEU cohort. The eight HLHS mutant mice in total had 329 mutations in 329 genes

^bComparison between groups was performed with χ² goodness of fit tests

Summary and Future Prospects

We showed HLHS has a multigenic etiology, which likely accounts for the previous failure to obtain HLHS mutant mouse models. Our analysis revealed a digenic etiology for HLHS in the Ohia HLHS mutant mouse line, consistent with the 2-loci model previously proposed for HLHS based on human clinical studies [32]. In Ohia mutants, Sap130 was shown to drive the LV hypoplasia, likely by perturbing cardiomyocyte proliferation and differentiation, while protocadherin mutation may drive the aortic valve defects. A paradigm shift is indicated in which the complex phenotype of HLHS may arise in a combinatorial fashion, with intrinsic cardiomyocyte defects underlying the LV hypoplasia. These findings suggest a familial study design might be better suited for interrogating the genetic etiology of HLHS. As our findings also indicated HLHS is not solely a valve disease as previously proposed [19], this has important implications for therapeutic interventions, suggesting in utero aortic angioplasty will not be effective [81]. Also of interest is the indication of the disruption of RNA processing, findings that align with the notion that cell intrinsic defects underlie the pathogenesis of HLHS. Overall, these findings suggest that novel therapeutic intervention will emerge as new insights are obtained into the molecular mechanism of disease pathogenesis in HLHS, interventions that may help prevent heart failure or reduce heart failure risk among HLHS patients.