Introduction

Hypoplastic left heart syndrome (HLHS) accounts for 2-3% of all cases of congenital heart disease (CHD) [1, 2] and is characterized by underdevelopment of the left ventricle, mitral and aortic valves, and aortic arch [3]. HLHS has a recognized genetic component based on its familial association with left-sided obstructive congenital heart diseases [4–6]. However, segregation analyses in multiplex HLHS-CHD families [6] and genome-wide association studies (GWAS) of large cohorts [5] have lacked sufficient power to conclusively identify candidate HLHS-susceptibility genes with small to moderate effect sizes. Defects in a small number of genes involved in cardiogenesis, such as NOTCH1 [7], NKX2–5 [8], MYH6 [9], and GATA4 [8, 10] have been implicated as contributors to HLHS, as well as multiple other CHDs. However, the variety of phenotypic manifestations of HLHS, together with numerous studies linking it to diverse genetic loci [7–9, 11], suggests that HLHS is genetically heterogeneous and has a multigenic etiology [12, 13].

It has also been hypothesized that the underdevelopment of the left ventricular myocardium includes restricted blood flow across the mitral valve and its hemodynamic effect during ventricular growth and development (“no flow – no grow” [14]), as well as endocardial defects [15]. Therefore, HLHS manifestation may be due to a combination of cardiomyocyte autonomous and non-autonomous genetic as well as mechanical effects, all of which likely affect proliferation and differentiation in the developing heart.

In a digenic HLHS mouse model, loss of both Sin3A Associated Protein 130 (Sap130) and the protocadherin Pcdha9 cause decreased CM proliferation resulting in left ventricular (LV) hypoplasia [16]. In addition, LVs from HLHS patients have cardiac damage associated with CM proliferation defects compared to healthy subjects [17]. These studies suggest that intrinsically defective cardiac differentiation and impaired CM proliferation are likely contributing to HLHS-associated heart defects. Further strengthening this hypothesis, we found that HLHS-patient-derived hPSC-CMs exhibited reduced proliferation compared to the parents [18]. There is a need to identify the potentially causative genes, which contribute to those processes described above and to elucidate their role in CHD/HLHS pathogenesis.

To advance CHD gene discovery, we have developed a multi-model systems platform enabling to both rigorously prioritize candidate genes from whole-genome sequencing (WGS) based on rare, predicted-damaging variants and their mode of inheritance, and to functionally characterize gene function in complementary and genetically tractable model systems: human induced pluripotent stem cells (hPSC), the fruit fly Drosophila melanogaster and zebrafish Danio rerio. The genetic basis of cardiac development, originally uncovered in Drosophila, is fundamentally conserved across species [19–22] and ∼80% of human disease genes have fly orthologs [23]. In addition, the non-redundant fly genome together with the simple structure and function of the fly heart allow straightforward genotype-phenotype correlations. Zebrafish [24–26], which has a two-chambered heart and can be easily manipulated using morpholino injections and CRISPR technologies to affect heart development and function. In addition, phenotypes can be directly observed in the developing larva for several days post fertilization [27]. Finally, cardiomyocytes (CMs) differentiated from hPSCs (hPSC-CMs) enable to identify and quantify cellular phenotypes associated with human diseases, including HLHS [7, 28–30] and are amenable to large scale functional screens [31, 32], thereby enabling to serve both as a discovery and validation platform. This integrated approach allows to rapidly identify novel HLHS-associated gene candidates and to characterize their ability to regulate developmental processes associated with disease (e.g., hypoplasia) [18].

Defective CM proliferation is likely a hallmark of HLHS [16–18], thus to uncover novel regulators of human CM proliferation with potential links to HLHS, we performed a whole genome siRNA screen (18,055 genes) using an EdU assay in hPSC-CMs, and identified ribosomal proteins (RPs) as the major class of genes controlling CM proliferation. In parallel, whole genome sequencing (WGS) of 25 poor-outcome HLHS proband-parent trios followed by unbiased variant filtering, revealed an enrichment for rare, predicted-damaging variants in RP genes. Moreover, analysis of a high-value family (75H) comprised of a HLHS proband, his phenotypically normal parents, and a fifth degree relative born with left-sided CHD led to the identification of a predicted- damaging variant in RP protein gene RPS15A that segregated with disease. Consistent with a central role for RPs in the regulation of heart development, heart-specific knock-down (KD) of HLHS-associated RPs in Drosophila caused severe phenotypes, including near complete heart-loss and lethality. Similarly, in zebrafish, rps15a CRISPR mutants and morphants had greatly reduced heart size and CM number, looping defects and diminished contractility. In this context, a multi-model system evaluation (hPSC-CMs, flies and zebrafish) of RP function show their important role on heart growth and differentiation and CM proliferation by modulating the p53, Myc and Hippo pathways. Moreover, we find evidence that RPs may regulate heart development and CM proliferation in a cardiac-specific manner by synergistically interacting with core cardiac transcription factors such as T-box, Nkx and Gata factors. In sum, here we show that RP genes are potent regulators of CM proliferation and cardiac growth and differentiation, consistent with a potentially critical role in the hypoplastic phenotype of HLHS. We suggest that RP genes are an emerging class of novel genetic effectors in HLHS/CHD.

Results

Whole genome siRNA screen identifies ribosomal proteins as central regulators of cardiomyocyte proliferation

Based on previous and our recent data [16–18], impaired cardiomyocyte (CM) proliferation is emerging as an important mechanism in HLHS pathogenesis. Thus, to comprehensively map the human genome for novel regulators of CM proliferation, we screened a library of 18,055 siRNAs for their ability to modulate proliferation in human iPSC-derived CMs (hPSC-CMs). To assess CM proliferation, we used a dual read-out and quantified both DNA synthesis using EdU incorporation [31], and total number of CMs, 3 days after treatment (Figure 1A-C). Briefly, siRNAs were transfected into 25- day old hPSC-CMs [29, 33]. After 2 days, EdU was added to the wells for 24 hours. On day 3, cells were fixed and co-stained for sarcomeric protein alpha-Actinin (ACTN2), EdU, and DAPI. Next, the number of EdU/alpha-Actinin double-positive cells, and the total number of CMs were determined using a commercially available high-throughput image analysis software (MetaXpress, Molecular Devices)[31]. Note that alpha-Actinin-, EdU+ cells which typically express fibroblast markers (i.e POSTN)[34], were excluded from this analysis, to focus on CM biology only. In total, we found 153 siRNAs that decreased proliferation (< 0.5-fold EdU incorporation) and number of CMs (<0.8-fold), and 162 siRNAs that increased EdU (>1.2-fold) and CM number (>1.1-fold; Figure 1B, C and Supplemental Table 1). Consistent with previous studies [35, 36], KD of cell cycle checkpoint agonist, CHEK1 and MDM2, the negative regulator of TP53, both were among top hits that reduced CM proliferation. As expected, we identified TP53 and CDKN1A among the genes that significantly increased CM cell number upon KD (Figure 1B, C). Next, to gain insight into the molecular pathways regulating proliferation in hPSC-CMs, we performed Gene Ontology (GO) term analysis of hits decreasing EdU incorporation or CM numbers and found an enrichment for genes associated with “translation/ribosome” and p53 signaling as top hits (Figure 1D, E). Consistent with these observations, we had previously found that the p53 pathway is dysregulated in hPSCs differentiated from a HLHS family trio [18]. Remarkably, the most represented gene family are ribosomal proteins (RPs), KD of which caused consistently the strongest inhibition of cell proliferation (Figure 1B). Next, since, the primary screen was performed as a single data point, we next sought to validate these findings by re-testing all 80 RP genes for function [37] using a distinct set of siRNAs in biological quadruplicate conditions. Remarkably and consistent with the primary screen, KD of 59/80 RPs significantly reduced proliferation as compared to siControl (p<0.05; Supplemental Figure 1A, B). Collectively, these findings indicate that RP function is essential for hPSC-CM proliferation and thus, suggest that RPs might represent a new class of regulators of ventricular growth during embryonic development.

Figure 1:

Enrichment of RP gene variants in an HLHS patient cohort with poor clinical outcome

To identify candidate genes involved in HLHS pathogenesis, we performed whole genome sequencing (WGS), variant filtering, and enrichment analysis (see also [18], and Methods: WGS and bioinformatic strategies) in a cohort of 25 HLHS patient-parent trios with poor clinical outcome defined as either prenatal – restrictive atrial septal defect (n=2); postnatal – reduced right ventricular ejection fraction following stage II or III surgical palliation (n=19); protein-losing enteropathy (n=2); or cardiac transplantation/failed surgical palliation (n=2). The 25 probands were comprised of 18 males and 7 females, 24 of whom were of white ancestry. The HLHS phenotype with respect to valve morphology (mitral atresia, MA; mitral stenosis, MS; aortic atresia, AA; aortic stenosis, AS) was: MA/AA in 10, MS/AS in 9, MA/AA in 4, MA/AS in 1 and unknown in 1. CHD in the parents was excluded by echocardiography in all but two families, in which a bicuspid aortic valve was present in the mother. Genomic sequences were filtered for uncommon (MAF < 1%) de novo and recessive variants in coding and regulatory regions of genes expressed in the developing heart and predicted to alter protein structure or expression, yielding 292 unique HLHS candidate genes that primarily fit a recessive mode of inheritance (Figure 2A and Supplemental Table 2). To determine whether certain gene networks were over-represented among these variants, we used two online bioinformatics tools (STRING and PANTHER [38, 39]). After applying Fisher’s exact test and false discovery rate corrections, RPs were the most enriched class of proteins when compared to the reference proteome, which includes data annotated by protein class (5.24-fold, p=0.0028), and cellular component (7.27- fold, p=0.011). In total, 14 variants found in 9 RP genes among 6 HLHS probands were identified, most fitting a recessive inheritance disease model (Supplemental Table 3). These RP variants encompass mutations in upstream promoter regions that potentially affect transcription factor binding sites, as well as non-synonymous substitutions inside the protein coding regions.

Figure 2:

Next, all 292 prioritized candidate genes were systematically evaluated for effects on the proliferation of hPSC-CMs in vitro and for cardiac differentiation of fly hearts in vivo. Again, loss of RP function (RPL39, RPL26L1, RPS15, RPS17, RSPS28) caused the most severe phenotypes on proliferation in hPSC-CMs (Figure 2B), which was consistent with our genome-wide proliferation screen (see Figure 1B). In the fly, KD of RPs in the cardiac lineage and throughout embryonic development was achieved using a Hand^4.2-Gal4 driver (see [40] and methods), and led to partial or complete heart loss (RpL26, RpL36A, RpL3) or caused lethality (RpS17, RpL39, RpS15, RpL26, RpL36A) (Figure 2C). Also, KD of RpL10 gave rise to severely constricted hearts, while KD of RpS28b caused inflow tract defects (Figure 2C). Finally, functional testing of RPs in zebrafish embryos revealed that KD of rpl39 leads to reduced ventricle size, whereas rps17 CRISPR mutants exhibit systolic dysfunction (Figure 2D, E). In sum, our patient centric approach (cohort of 25 probands) identifies RPs as the most enriched gene category affected by HLHS-associated variants (STRING, PANTHER analysis). In this context, systematic testing in multi-model systems (fly, zebrafish, hPSC-CMs) identifies RPs as most potent regulators of cardiac differentiation, proliferation and contractility among HLHS-associated genes. Collectively, these findings highlight that RPs regulate critical steps of cardiogenesis that are also found to be defective in HLHS probands [41, 42], and thus suggest a potential link between RP biology and HLHS phenotypic etiology.

RPS15A variant associated with HLHS in familial CHD

To further establish a phenotypic link between RP function and HLHs associated phenotypes, we selected a rare familial CHD case (75H) (Supplemental Figure 2A), that is comprised of a young teenager with HLHS (MS/AS) and normal ventricular function following Fontan operation, his phenotypically normal parents, and a 5^th degree female relative born with left-sided CHD (bicuspid aortic valve and coarctation of the aorta). Based on these observations, we hypothesized the presence of a heterozygous driver variant exhibiting incomplete penetrance and variable expression. To investigate further, we filtered whole-genome sequencing (WGS) data for rare, predicted-damaging variants in coding and regulatory regions. This analysis identified six prioritized variants (Supplemental Table 4), including one located in the RPS15A gene locus, shared by the HLHS proband, the unaffected mother, and the proband’s fifth-degree relative with congenital heart disease (CHD).

We next investigated whether family members carrying these variants (the mother and proband) exhibit defects in cardiomyocyte (CM) proliferation compared to the father. Notably, analysis of EdU incorporation in CMs from the parent-proband trio revealed significantly reduced proliferation in both the proband and mother compared to the father. However, the proband displayed a markedly more severe phenotype than the mother (Figure 3 A,B). Interestingly, this reduction in proliferation appeared to be specific to cardiomyocytes (CMs), as EdU incorporation in the non-CM cell population showed no significant differences among the family members (Supplemental Fig. 2B). Collectively, these findings suggest a phenotypic association between the presence of the variants and impaired CM proliferation.

Figure 3:

Finally, to determine the potential relative contribution of the six genes harboring rare and damaging variants in the regulation of CM proliferation, we performed functional KD experiments in hPSC-CMs in vitro. Remarkably, while KD of 5 of the other prioritized genes from the 75H family also resulted in moderate reduced CM proliferation, RPS15A KD produced the most pronounced effect, with over an 85% reduction in proliferation, emerging as the top candidate (Figure 3C, D). Together, these findings indicate that RPS15A plays a critical role in regulating cardiomyocyte (CM) proliferation in human pluripotent stem cells (hPSCs) and may influence heart developmental processes that are disrupted in HLHS.

RPS15A KD in Drosophila and zebrafish causes severe cardiac proliferation and differentiation defects

To characterize the role of RPS15A during heart development, we first induced a heart-specific KD of RpS15Aa in flies (see methods), which caused a partial or complete loss of the heart (Figure 3E,F). Note that this phenotype is consistent with our results above and a previous study involving another RP gene (RPL13) [43].

Next, to evaluate the role of RP genes during heart development, we knocked- out rps15a in zebrafish (Danio rerio) using CRISPR [24–26]. We examined F₀ larva at 72 hpf (hours post fertilization) using high speed digital imaging (SOHA [44], see also methods) and confocal microscopy to monitor heart size and cardiomyocyte numbers respectively. Remarkably, rps15a mutants hearts were smaller in size (Supplemental Fig. 2C,D), with reduced CM numbers (Figure 3G,H). Interestingly, CM numbers were predominantly reduced in the ventricle compared to the atrium (Figure 3G,H). We observed similar effects in response to injection of a rps15a morpholino (MO) compared to uninjected larva [45, 46] (Supplemental Figure 2 C,D). In this context, we also noted that the overall body plan and morphology of rps15a CRISPR mutants or morphants were largely unaffected, except for a mild pericardial edema, a possible indicator of heart dysfunction (Figure 3H) [47]. Collectively, these data support the hypothesis that RPS15A is required for CM proliferation and heart morphogenesis (hPSC-CMs, fly and zebrafish), thus suggesting a potential role for RPs as genetic drivers of hypoplasticity observed in HLHS or defects in other CHDs.

Ribosomal proteins control CM cell cycle by regulating TP53 pathway activity

To investigate how RPs regulate cell cycle activity in hPSC-CMs, we performed RNA-seq upon KD of RPS15A and RPL39, which belong to the small and the large ribosomal subunits that harbor predicted-damaging variants in 75H and 151H probands, respectively. Comparison of differential gene expression between siRPS15A and siRPL39 KD revealed that ∼53% of genes were commonly dysregulated (Figure 4A), indicating the existence of a conserved RP-dependent transcriptional network. Consistent with a central role for RPs in the regulation of proliferation, GO term analysis for differentially expressed genes revealed an enrichment for genes involved in DNA replication, mitosis and DNA damage response mediated by p53, and included the downregulation of positive regulators of cell cycle such as CDK1, CCNA2, AURKA, PCNA and the upregulation of cell cycle arrest mediators such as CDKN1A, BTG2, GADD45A (Figure 4B and Supplemental Tables 5,6). To confirm these findings, we performed immunostaining for canonical p53 downstream transcriptional target [48], CDKN1A, and observed that the percentage of CDKN1A+ CMs was increased by ∼60% in response to RPS15A KD (Figure 4C, D), indicating that RP KD induces cell cycle arrest via the activation of the p53 signaling pathway in hPSC-CMs.

Figure 4:

Next, to evaluate if the reduction in proliferation caused by RP KD is mediated by the p53 signaling pathway in CMs, we co-KD TP53 or CDKN1A along with RPS15A, and observed a significant rescue of CM proliferation by EdU incorporation and CM number in comparison to RPS15A KD (Figure 4E, F and Supplemental Figure 4A, B).

Next, we tested if RP-mediated defects in zebrafish can also be rescued by inhibition of p53. To this aim, we used morpholinos to co-KD p53 with rps15a [49] and examined larval hearts at 72 hpf. p53 MO-mediated KD on its own had little effect on heart function but partially reversed the rps15a MO-mediated reduction in heart size and cardiomyocyte numbers (Figure 4G, H). Moreover, bradycardia, heart looping and reduced contractility phenotypes of rps15a morphants were also rescued by co-KD of p53 (Supplemental Figure 4C, D). Collectively, these observations support our hypothesis that RPs control cardiac development, including heart size and function, via a p53 pathway-mediated regulation of CM proliferation.

Heart loss in Drosophila by RpS15Aa KD is partially rescued by Hippo pathway activation

The fly heart develops similarly to the vertebrate heart at early stages [20] and a link between RPs and cell cycle-regulating pathways, such as p53 and Hippo pathways [50], has been proposed to be conserved between mammals and flies [51]. To test whether p53 has a function in fly heart development similar to hPSC-CMs and zebrafish, we performed cardiac co-KD of RpS15Aa and p53 and examined heart structure and function. Co-KD did not rescue the cardiac defects of RpS15Aa KD suggesting that p53 has a different role in flies, as compared to vertebrates, as has been reported previously [52].

In vertebrates, TP53 is known to act through the negative regulation of the Hippo pathway [53]. In this context, YAP, a downstream effector of Hippo, has been shown to promote CM proliferation in the mouse heart [54]. We therefore tested if YAP/yorkie can substitute the function of p53 in the fly heart and found that overexpressing the Hippo pathway gene yorkie (yki), the fly ortholog of YAP, in the developing heart considerably restored RpS15Aa KD-induced heart loss in Drosophila (Figure 5A’,B). We also observe formation of adult ostial structures, which indicates that the larval heart underwent partial remodeling to an adult heart during metamorphosis. Further, we found that yki-mediated rescue depends on the transcriptional co-factor encoded by scalloped (sd, Drosophila TEAD1/2/3/4), since upon sd KD, yki OE can no longer rescue the RpS15Aa KD-induced heart loss (Figure 5A”, B). Moreover, KD of sd in an RpS15Aa KD background (without yki OE) worsens the RpS15Aa KD phenotype (Figure 5A”, B).

Figure 5:

RPS15A genetically interacts with cardiac transcription factors nkx2.7/tinman, and TBX5/tbx5a/Dorsocross, nkx2.7/tinman, and Gata4,5,6/pannier in model systems

The profound importance of RPs for cell growth and proliferation in most cell types ([55] and Supplemental Figure 1), raises the question of how tissue specific phenotypes such as those observed in HLHS can arise from the defective function of ubiquitously expressed genes. To address this apparent paradox, we hypothesized that RPs might functionally interact with tissue-specific proteins such as cardiac transcription factors (cTFs), to control CM proliferation. Analysis of the impact of cTFs alone on hPSC-CM proliferation from our whole genome screen, revealed that KD of most cTFs did not cause major proliferation defects (except for e.g., HES4 and HOPX; Figure 6A). Among the cTFs, TBX5 is specifically expressed in the left ventricle at stages of intense CM proliferation [56] and thus we next asked if TBX5 could functionally interact with RPs. Remarkably, while hPSC-CM proliferation was not affected by increasing doses of siTBX5 (0-2nM), increasing siTBX5 dosage in siRP backgrounds led to proportional decrease in EdU incorporation (Two-way ANOVA; Figure 6B and Supplemental Figure 5A,B), thereby indicating that RPs and TBX5 genetically interact to regulate proliferation in hPSC-CMs.

Figure 6:

Conservation of genetic interactions between human and fly cardiac transcription factors and constitutive genes have been identified before [57]. We therefore specifically tested for genetic interaction between RpS15Aa and the cardiac TFs tinman/NKX2-5, pannier/GATA4/5/6 and Dorsocross/TBX5 in adult Drosophila hearts. A heterozygous deficiency covering RpS15Aa causes moderate dilation of hearts compared to controls (Supplemental Figure 5C). However, when RpS15Aa was placed in trans to loss of function alleles of cardiac TFs tinman, pannier or Dorsocross1/2/3, these hearts exhibited significantly deformed hearts, which is only rarely observed in the single heterozygotes (Figure 5C, D). In addition, RpS15Aa heterozygous flies showed a prolonged heart period when combined with a tinman or pannier heterozygous mutations (Supplemental Figure 5C).

Next, we tested whether similar genetic interactions are conserved in zebrafish. Zebrafish expresses the NKX2-5 ortholog nkx2.7 in the heart and functional studies have shown that both nkx2.5 and nkx2.7 play critical roles in cardiac development [58]. To determine whether rps15a and nkx2.7 or tbx5a genetically interact in zebrafish, we injected low doses of rps15a [45] MO, in combination with nkx2.7 [59] or tbx5a MO. nkx2.7 MO alone had little effect on heart size, cross-sectional area, or contractility, whereas double morphants exhibited cardiac dysfunction, with contraction virtually abolished in some animals (Supplemental Figure 5D). Furthermore, we observed a significant prolongation of the heart period in double morphants (rps15a + nkx2.7 or rps15a + tbx5a) in comparison to each MO alone (Two-way ANOVA; Supplemental Figure 5D, E) consistent with the effects observed in Drosophila. Together, these results highlight the existence of evolutionarily conserved (fly, zebrafish and hPSC- CMs) genetic interactions between RPs and cardiac TFs genes in the regulation of cardiogenesis. These data also illustrate how the activity of ubiquitously expressed genes such as RPs, can be modulated by cell type-specific genes, like cardiac TFs, to achieve tissue-specific outcomes.

RP-regulated HLHS-associated genes control proliferation in a CM-specific manner

Previous studies and this work show that reduced RP function can lead to tissue- specific developmental defects, that in turn can be rescued by loss of p53 [60, 61]. While these findings highlight a downstream role for p53 in the development of RP- dependent phenotypes, they also imply that RPs can regulate cell cycle and/or p53 activity in a tissue-specific manner. Thus, to delineate how RPs might control proliferation in a CM-specific manner, we compared differential gene expression profiles upon the two RP KDs in hPSC-CMs and undifferentiated hPSCs (Figure 7A). Consistent with our hypothesis, this analysis revealed that RPs regulate the expression of 493 genes in a CM-specific manner (Supplemental Tables 5-8), thereby suggesting the existence of a cell type-specific (cardiac) transcriptional response to RP loss of function.

Figure 7:

Next, to evaluate whether HLHS-associated genes were part of this lineage- specific transcriptional response, we selected those differentially expressed genes harboring rare and predicted damaging variants in probands from the poor outcome cohort and 75H family (Supplemental Tables 2, 4). Remarkably, this prioritization strategy identified 12 genes potentially associated with HLHS (Supplemental Table 9), 6 of which were found to also regulate cell cycle activity in hPSC-CMs (Figure 7B, C). Thus, collectively this approach led us to identify an RP-dependent and cardiac-specific HLHS-associated gene network that supports CM proliferation via the upregulation of EWSR1, MDC1, PKMYT1 and KIF1C, and downregulation of FAT4 and EPHA4 (Figure 7D, E).

To further explore if RP-regulated HLHS-associated genes control proliferation in a cell type-specific manner, we asked if EWSR1 KD, which elicited the strongest proliferation phenotype in hPSC-CMs, could regulate proliferation in non-cardiac cell types such hPSCs. In contrast to hPSC-CMs, siEWSR1 did not affect the number of OCT4+ cells (Figure 7F, G), which suggests that EWSR1 is able to regulate proliferation CM-specifically. EWSR1 is an RNA/DNA binding protein that regulates transcription and RNA splicing and plays a major role as oncogenic driver when fused to ETS transcription factors [62]. In a next step, we tested if EWSR1 regulates CM proliferation by modulating the p53 pathway. Consistent with this hypothesis, immunostaining for p53 downstream target, CDKN1A, revealed that EWSR1 KD both increased the percentage of CDKN1A-positive CMs and decreased EdU incorporation as compared to siControl (Figure 7H-J). Consistent with these observations, RpS15Aa deficiency fly line placed in trans to loss of function allele for cabeza (caz), the ortholog of EWSR1, caused prolongation of systolic interval lengths (Supplemental Figure 5F) as compared to controls, suggesting a genetic interaction between these two genes. Collectively, these results support the conclusion that RPs can modulate proliferation in a cell type-specific manner, by controlling the expression of downstream effectors (i.e EWSR1), with a cell type-specific ability to modulate the p53 pathway activity (Figure 7K and Figure 8).

Figure 8:

Discussion

Implication of ribosomal proteins as genetic modulators in HLHS/CHD pathogenesis

Recent work, including our own, strongly suggest that defective cardiac differentiation and impaired CM proliferation contribute to HLHS-associated heart defects [16–18]. In this study, we find that ribosomal proteins are critical for maintaining CM proliferation, as well as cardiac structure and function in fish and fly heart models. We show a potential role for RP function in cardiac development and HLHS/CHD pathogenesis as : (1) RPs are central regulators of cardiomyocyte proliferation, and (2) RP variants are enriched in a cohort of poor-outcome HLHS patients; (3) a rare predicted-damaging promoter variant in RPS15A was associated with HLHS in familial CHD (75H); (4) KD of RPs in vivo results in cardiac deficiencies in hPSC-CMs, Drosophila and zebrafish, and (5) induces TP53 signaling and blocks proliferation that can be alleviated by inhibition of TP53 in the vertebrate models. Interestingly, the transcriptomic response of a previously reported HLHS proband [18] paralleled these findings and raises the possibility of an underlying ribosomopathy condition in that HLHS family. While universally important for cellular growth, an increasing number of variants in RP genes have been linked to CHD in humans indicating tissue-specific differences in the penetrance of RP gene mutations. Most notably, ∼30% of patients with Diamond-Blackfan anemia [63], a ribosomopathy characterized by hypo- proliferative, proapoptotic erythropoiesis, also have CHD. In agreement with another study [46] rps15a KD in fish caused a noticeable reduction in circulating red blood cells.

In addition to RPS15A, RpL13 was recently identified as a potential candidate gene involved in CHD from a screen for de novo copy number variations (CNVs) in 167 patients with CHD [43]. Furthermore, the Pediatric Cardiac Genomics Consortium (PCGC) exome dataset of 2871 patients with CHD identified several rare, predicted- damaging de novo variants in RP genes, 2 of which were found in patients with HLHS [64]. Our finding that RPs genetically interact with core cardiac transcription factors provides a potential mechanism for the cardiac specificity of RP phenotypes. In this context, we hypothesize that RP gene variants might play a role as genetic modifiers (or sensitizers) in the oligogenic pathogenesis of HLHS.

Cardiac-specific and RP-dependent regulation of proliferation in the context of HLHS

Defective proliferation in left ventricular CMs is a phenotypic hallmark of HLHS [16, 17], thereby implying the existence of HLHS-associated mechanisms that regulate proliferation in a CM-specific specific-manner. In this study, we identified RPs as novel candidate genetic regulators of hypoplastic phenotypes observed in HLHS, as 1) rare, predicted damaging variants affecting RP genes are enriched in HLHS probands as compared to healthy controls, 2) KD of most RPs impairs proliferation in hiPSC-CMs (59/80 tested) and heart development in flies (6/6 tested), and 3) systemic loss of rps15A function causes heart-specific hypoplastic phenotypes in zebrafish. However, in this context, the evaluation of RPs function in multiple cell types revealed that they are generally required for proliferation (hPSCs, hPSC-CMs and dermal fibroblasts), thus, indicating that RP loss of function triggers a non-cell type specific cell cycle block.

Thus, an appealing mechanism by which RP function could be invoked in a spatially restricted manner would involve the regulation of RP expression by cell type- specific TFs. Consistent with this model, all probands harboring RP variants contain mutations affecting their promoter regions (Supplemental Table 1). Also, several of these mutations are predicted to disrupt the binding site of previously identified HLHS- associated TF, ETS1 [15, 65]. Consistent with these observations, a recent study from the Bruneau lab [66] show that 38 RP genes are downregulated in TBX5-null hiPSC- CMs and ChIP analysis [67] reveals that a third of these dysregulated RPs (14/38 genes) are bound by TBX5 at their core promoter regions. Thus, we speculate that cell type specific TFs controlling RP expression represent a promising gene class enabling to modulate RP-dependent proliferation in a tissue-specific manner.

A second mechanism by which RPs might control proliferation and differentiation in a CM-specific manner, is based on the hypothesis that CMs may be molecularly more sensitive to RP loss-of-function than other cell types. Consistent with this model, RP deficiency predisposes flies, fish and hPSC-CM for interactions with core cardiogenic TFs, including TBX5/Doc, GATA4/pnr and NKX2-5/tin, the latter being the most cardiac restricted. (Fig. 6). Moreover, our comparative transcriptomics analysis (Figure 4) revealed that RPs regulate gene expression in a CM-specific manner. Among these RP regulated genes, EWSR1, MDC1, PKMYT1, KIF1C, FAT4 and EPHA4, were found to be mutated in HLHS families and regulate CM proliferation (Figure 7). Remarkably, EWSR1 KD did not affect proliferation in hPSCs, while it strongly reduced EdU incorporation in hPSC-CMs, thus suggesting that EWSR1 regulates proliferation in a CM-specific manner. Interestingly, in flies RpS15Aa genetically interacts with cabeza (caz, the fly ortholog of EWSR1). Consistent with our observations, a recent study has found that overexpression of a fusion Ewsr1-Fli1 protein in mice specifically induces dilated cardiomyopathy by promoting apoptosis in cardiac myocytes [68]. Similarly, Fat4, a negative regulator of the hippo pathway, which is upregulated upon RP, KD was found to specifically regulate heart size via the regulation of CM size and proliferation [69]. Thus, collectively we propose that RPs represent a central regulatory hub for cell proliferation and differentiation during embryonic development, which can be modulated in a cell type-specific manner by tissue-restricted modulators of RP expression and/or activity.

An RP–MDM2–p53/YAP surveillance network in heart development

Defects in ribosome biogenesis caused, for example, by reduction in RP gene copy numbers induce a stress response program that activates TP53 (p53) and a cascade of cellular events resulting in apoptosis, inhibition of the cell cycle or DNA damage response (reviewed in e.g. [70]), commonly referred to as nucleolar stress response [71] with activation of TP53 being a key hallmark [72]. A key mechanism involves the release of RPs to the nucleoplasm, where several RPs bind and inhibit MDM2, which again results in p53 stabilization and pathway activation [73, 74]. A recent study showed that RP haploinsufficiency in the developing mouse limb bud also activates a common TP53 cascade but results in TP53-dependent tissue-specific changes of the translatome, which might confer the specificity often observed in ribosomopathy [60]. The TP53-MDM2 feedback loop is the central molecular node in response to a wide variety of stress signals, including in the human diseased adult heart [75, 76]. As mentioned above, the transcriptomic response of the HLHS family 5H [18] also showed activation of the TP53 pathway potentially due to an underlying ribosomopathy. Since low RP levels induce a proliferation blockade that can be overridden by p53-p21/CDKN1A KD, we hypothesize that RP levels may act as signaling components sensing cellular fitness and with a functional outcome either being p53 activation or non-activation. Such an RP–MDM2–p53 surveillance network was previously proposed to be important in response to nutrient availability changes and inhibition of oncogenic activity [77, 78]. Thus, the inhibition of the RP-MDM2-p53 axis might be a therapeutic avenue to consider for functional intervention, although further studies are needed, to identify the upstream mechanisms leading to the p53 pathway activation observed in HLHS proband cells.

Heart-specific KD of the RpS15Aa in Drosophila causes constriction of larval hearts and atrophy in adult hearts, due to heart loss during metamorphosis, which could not be rescued by p53 reduction, unlike in hPSC-CMs and zebrafish. This might be due to a different, context-dependent role of p53 in flies previously reported to cause a dwarfing, Minute-like phenotype [79]. We hypothesized that the loss of RPs in flies, in conjunction with reduced protein synthesis, might cause nucleolar stress, triggering a cell-intrinsic signaling cascade that prevents the heart to further differentiate and grow.

The role of the Hippo pathway to regulate cell growth/organ size, proliferation, and survival, and its importance in cardiac biology is conserved between humans and fly [80–83]. The Hippo-YAP pathway is a cell-intrinsic pathway that regulates cardiomyocyte proliferation and thus heart size during development, as demonstrated by the ability of activated Yap to induce postnatal CM regeneration [84]. Interestingly, we find here that overexpression the Drosophila ortholog of YAP/TAZ, Yorkie, rescues RpS15Aa KD-induced heart loss, dependent on its downstream factor scalloped (the Drosophila ortholog of TEAD1/2/3/4) (Figure 5). Of note, YAP1 is not only an important regulator of cardiomyocyte proliferation in the embryo but also promotes cardiomyocyte survival and growth in the postnatal heart [85, 86], which is in line with our findings. Interestingly, a proband among the PCGC HLHS patients was transheterozygous for mutations in RPL15 and TEAD4 [64], making this an interesting disease candidate pair.

Rapid Gene Discovery and Prioritization Approach for Complex Genetic Diseases

Uncovering the genetic basis of polygenic and heterogeneous diseases, such as HLHS, remains a significant challenge. This difficulty arises from the lack of experimental approaches capable of rapidly determining the role and contribution of genetic variants, as well as their epistatic relationships, in generating disease-associated phenotypes. The current state-of-the-art approach involves using CRISPR-mediated editing or correction of gene variants [87], which is effective for analyzing single-family pedigrees. However, this method is impractical for prioritizing the large number of candidate genes identified through cohort-wide analyses. Consequently, progress in gene discovery for HLHS has been limited in recent years [12, 88]. In this study, we aimed to address these limitations by employing a HT and unbiased exploration of genes regulating a conserved HLHS-relevant phenotype [41, 89–91], such as reduced CM proliferation, in hiPSCs. Notably, results integration from a whole-genome functional screen in hiPSC-derived CMs (Figure 1, RPs are top hits from the screen), genomic data from a HLHS parent-proband cohort (Figure 2, RPs represent the most enriched gene class containing rare damaging variants), hiPSC-CM phenotyping from a high value HLHS family (75H), and functional validation in two independent in vivo model systems (Figure 3), identified RPs as a novel class of HLHS-associated genes. While our study highlights the potential of this approach for gene prioritization, additional research is needed to confirm the causal roles of the identified genetic variants. In this context, future research could also incorporate unbiased morphological assessments of disease phenotypes [92] to identify additional disease-associated traits and regulatory pathways. In conclusion, we propose that the approach outlined in this study offers a novel framework for rapidly prioritizing genes in the context of complex genetic diseases, paving the way for deeper insights into their underlying mechanisms.

Materials and methods

Key resources table

Study subjects

Written informed consent was obtained for HLHS probands and family members under a research protocol approved by the Mayo Clinic Institutional Review Board. Cardiac anatomy was assessed by echocardiography. Candidate genes were identified and prioritized by WGS of genomic DNA and RNA sequencing of patient-specific iPSC and cardiomyocytes. Methods for genomic analyses, RNA Sequencing, iPSC technology, bioinformatics and statistics are described in the Online Appendix. Data are available at NCBI GEO database.

WGS and bioinformatic strategies

Our methods for family phenotyping, WGS, variant filtering, and candidate gene prioritization in the rare HLHS-CHD family and 25 HLHS proband-parent trios have been previously described [7, 9, 18]. In brief, WGS of DNA isolated from blood or cheek swabs was performed on a HiSeq 4000 platform at the Mayo Clinic Medical Genome Facility. BAM files underwent primary and secondary analyses using an established workflow with standard data quality metrics, and reads were aligned to the human hg38 reference genome. Variant call format files with single nucleotide variants and small insertion-deletion (indel) calls were analyzed using Ingenuity Variant Analysis software and knowledge database (Qiagen). To retain only high-confidence data, variants were required to have a base call quality of at least 20 and to pass Variant Quality Score Recalibration. Functionally annotated coding and regulatory variants underwent primary filtering and prioritization by an iterative approach with variants required to have a minor allele frequency (MAF) <0.01 across all races in gnomAD v2.1.[93] Variants predicted to impact protein structure were retained and included missense, frameshift, stop-gain, stop-loss, canonical splice and non-canonical variants predicted to disrupt splicing based upon MaxEntScan[94]. Regulatory variants were defined as those that (a) impact a microRNA or microRNA binding site (b) reside in an enhancer binding site annotated in the VISTA database (http://enhancer.lbl.gov/) or (c) disrupt a predicted promoter or transcription factor binding site informed by the position-weighted matrices available in JASPAR (http://jaspar.cgb.ki.se/). Genes harboring a rare functional variant were further required to have upper quartile cardiac expression during mouse embryonic (e14.5) (Homsy et al., 2015) or human fetal heart development (ranked percentiles from RNASeq experiments available in ENCODE: ENCSR047LLJ (120 day male); ENCSR863BUL (91 day female); ENCSR000AEZ (28 week female and 19 week female) (Davis et al., 2018).

For the HLHS-CHD family, a secondary segregation filter was applied to model autosomal dominant inheritance with incomplete penetrance, which required sharing of prioritized rare variants between affected family members. For the poor-outcome HLHS proband-parent trios, a secondary Mendelian filter was applied to identify major-effect driver variants that arose de novo or fit recessive modes of inheritance (i.e., homozygosity, hemizygosity, compound heterozygosity).

To determine whether certain gene networks were over-represented, two online bioinformatics tools were used. First, STRING[39] was used to investigate experimental and predicted protein-protein and genetic interactions, and clustering of RP genes was demonstrated when the highest stringency filter was applied. In addition, PANTHER[38] was employed to identify potential enrichment of genes by ontology classification.

Generation of hPSC-VCMs

Id1 overexpressing hPSCs [33, derived from, 95] were dissociated with 0.5 mM EDTA (ThermoFisher Scientific) in PBS without CaCl₂ and MgCl₂ (Corning) for 7 min at room temperature. hPSC were resuspended in mTeSR-1 media (StemCell Technologies) supplemented with 2 µM Thiazovivin (StemCell Technologies) and plated in a Matrigel-coated 12-well plate at a density of 3 x 10⁵ cells per well. After 24 hours after passage, cells were fed daily with mTeSR-1 media (without Thiazovivin) for 3-5 days until they reached ≥ 90% confluence to begin differentiation. hPSC-VCMs were differentiated as previously described [95]. At day 0, cells were treated with 6 µM CHIR99021 (Selleck Chemicals) in S12 media [96]. for 48 hours. At day 2, cells were treated with 2 µM Wnt-C59 (Selleck Chemicals), an inhibitor of WNT pathway, in S12 media. 48 hours later (at day 4), S12 media is fully changed. At day 5, cells were dissociated with TrypLE Express (Gibco) for 2 min and blocked with RPMI (Gibco) supplemented with 10% FBS (Omega Scientific). Cells were resuspended in S12 media supplemented with 4 mg/L Recombinant Human Insulin (Gibco) (S12+ media) and 2 µM Thiazovivin and plated onto a Matrigel-coated 12-well plate at a density of 9 x 10⁵ cells per well. S12+ media was changed at day 8 and replaced at day 10 with RPMI (Gibco) media supplemented with 213 µg/µL L-ascorbic acid (Sigma), 500 mg/L BSA-FV (Gibco), 0.5 mM L-carnitine (Sigma) and 8 g/L AlbuMAX Lipid-Rich BSA (Gibco; CM media). Typically, hPSC-ACMs start to beat around day 10. At day 15, cells were purified with lactate media (RPMI without glucose, 213 µg/µL L-ascorbic acid, 500 mg/L BSA-FV and 8 mM Sodium-DL-Lactate (Sigma), for 4 days. At day 19, media was replaced with CM media.

siRNA transfection, proliferation assay and immunostaining in hPSC-VCMs

At day 25 of differentiation, hPSC-VCMs were dissociated with TrypLE Select 10X (Gibco), 10 min and neutralized with RPMI supplemented with 10% FBS. Cells were resuspended in RPMI with 2% KOSR (Gibco) and 2% B27 50X with vitamin A (Life Technologies) supplemented with 2 μM Thiazovivin and plated at a density of 5,000 cells/well in a Matrigel-coated 384-well plate. hPSC-VCMs were transfected with siRNA (Dharmacon: ON-TARGETplus, custom RNAi cherry-pick libraries 0.1 nmol). For the whole genome screening, siRNAs directed to 18,000 human genes were purchased from the Genomic Center Facility at SBP at a final concentration of 25 nM using lipofectamine RNAiMax (ThermoFisher) and opti-MEM (Gibco). 48 hours post- transfection, cells were labelled with 10 μM EdU (ThermoFisher). After 24h of EdU incubation, cells were fixed with 4% paraformaldehyde for 30 min and blocked in blocking buffer (10% Horse Serum, 10% Gelatin and 0.5% Triton X-100) for 20 min. EdU was detected according to protocol and cells were stained with cardiac specific marker ACTN2 (A7811, Sigma, dilution 1/800), secondary antibody Alexa Fluor 568 (Invitrogen, 1/500) and DAPI (1/1000) in Blocking Buffer. Cells were imaged with ImageXpress Micro XLS microscope (Molecular Devices) and custom algorithms were used to quantify % of EdU+ ACTN1+ hPSC-VCMs and the number of ACTN1+ cells. To quantify TP53 levels in hPSC-VCMs, cells were stained with Phospho-p53 (700439, ThermoFisher, 1/500), counterstained with ACTN1 (anti-mouse A7811, Sigma, or anti- rat ab50599, Abcam), and imaged and quantified with ImageXpress microscope. Whole genome screening was performed in one-replicate. All the other siRNAs experiments were performed in quadruplicates.

RNA-seq and data analysis

hPSCs and hPSC-derived CMs were transfected with 25nM final concentrations of siRNA against RPS15A, RPL39 and with scrambled control siRNAs as above. 2 days after siRNA transfection RPs KD were verified and confirmed by proliferation assay (see above), and cells were pelleted and resuspended in 500µl TRIzol reagent followed by total RNA extraction. Library preparation and sequencing of the samples was done at La Jolla Institute of Immunology (La Jolla, CA). FASTQ files were processed using nf- core/rnaseq (version 21.03.0.edge [97]). Differential gene expression was determined using R/DESeq2 [98] and GO term enrichment was done using gprofiler2[99]. Analysis scripts can be downloaded at github.com/gvogler. Raw reads and counts tables are available at GEO accession number GSE207658.

Proliferation assay in parent/proband iPSCs

hPSCs were dissociated as described previously and plated in a Matrigel-coated 384-well plate at a density of 3,000 cells per well. Cells were transfected with siRNA at a final concentration of 25nM using lipofectamine RNAiMax and opti-MEM. After 3 days, cells were labelled with 10 μM EdU for 1 hour. Cells were fixed with 4% paraformaldehyde for 30 min and blocked in blocking buffer for 20 min. EdU was detected according to the protocol and stained with stem cell specific marker OCT4 (P0082, Sigma, 1/500), secondary antibody Alexa Fluor 568 and DAPI in Blocking Buffer. Cells were imaged with ImageXpress microscope and % of EdU+ cells and number of OCT4+ cells were quantified.

Quantitative Real Time PCR (RT-qPCR) in parent/proband iPSCs and hPSC-CMs

Total RNA was extracted using TRIzol and chloroform. 1ug of RNA was converted in cDNA using QuantiTect Reverse Transcription kit (Qiagen). qRT-PCR was performed using SYBR green (Biorad). Human primers sequences for qRT-PCR were obtained from Harvard Primer Bank: TP53 (Primer Bank ID: 371502118c1), CDKN1A (Primer Bank ID: 310832423c1), CCNB1 (Primer Bank ID: 356582356c1), CCNB2 (Primer Bank ID: 332205979c1), CDK1 (Primer Bank ID: 281427275c1), MCM2 (Primer Bank ID: 33356546c1), RPS15A (Primer Bank ID: 71772358c2). All values were normalized to GAPDH (Primer Bank ID: 378404907c1). At least 3 independent biological replicates were performed for each experiment.

Statistical analysis

To determine any statistical significance between experimental and control groups in hPSC-VCMs and iPSCs from parent/proband trio, we calculated two-sided p values with Student’s t-test using GraphPad Prism 8.1.2 software.

Drosophila heart function studies

Drosophila orthologs were determined using the DIOPT database [100], and fly stocks were obtained from the Vienna Drosophila Resource Center (VDRC) stock center or Bloomington Drosophila Stock Center (BDSC) as indicated in the key source table. For in vivo functional heart analysis, we developed a high-throughput method based on genetically modified flies with cardiomyocyte-specific RFP fluorescence (Klassen et al.). The reporter line by itself or combined with the heart-specific Hand^4.2- Gal4 driver (including all cardioblasts/cardiomyocytes, pericardial cells and wing hearts throughout development starting at post-mitotic, mid-embryonic stages through adulthood [101–103] was crossed to UAS-lines or mutant flies. For interaction studies a fly line harboring the fluorescent reporter, Hand^4.2-Gal4 driver and a Deficiency for RpS15Aa was generated. Adult progeny flies were immobilized and exposed to fluorescence light to record 5 sec high frame rate movies of the beating heart. Movies were analyzed by fully automated quantification of contractility and rhythmicity parameters of the heart [104]. Semi-intact adult fly hearts were filmed and analyzed according to standard protocol [27, 105].

Immunostainings of the fly heart

The immunostaining of fly adult hearts was performed as described previously [106]. In short, adult flies were dissected in a semi-intact fashion to expose the heart according to protocol [27, 107]. Myofibrils were relaxed using 10mM EGTA followed by fixation in 4% formaldehyde for 15 min. Hearts were washed with PBS + Triton (PBT; 0.03 % triton X-100) and stained using Alexa 647 phalloidin (Life Technologies) 2h at room temperature (RT). For Mhc labelling, hearts were stained using anti-Mhc antibody (1:50, incubation over night at 4°C) and after three washes with PBT, secondary antibody was applied for 2 h at RT. Hearts were washed with PBT three times and PBS one time and mounted using ProLong Gold mountant with DAPI (Life Technologies). Heart preparations were imaged with the Imager.Z1 with an Apotome (Carl Zeiss), Hamamatsu Orca Flash 4.0 and ZEN imaging software (Carl Zeiss).

Zebrafish Husbandry

All zebrafish experiments were performed in accordance with protocols approved by IACUC. Zebrafish were maintained under standard laboratory conditions at 28.5°C. In addition to Oregon AB wild-type, the following transgenic lines were used: Tg(myl7:EGFP)^twu277 [108] and Tg(myl7:H2A-mCherry)^sd12 [109].

In zebrafish, gene expression was manipulated using standard microinjection of morpholino (MO) antisense oligonucleotides [110]. In addition, we performed targeted mutagenesis using CRISPR/Cas9 genome editing[111–113], to create insertion/deletion (INDEL) mutations in relevant ribosomal genes (F0). Subsequently, zebrafish were raised to 72 hours post fertilization (hpf), immobilized in low melt agarose and the hearts were filmed and analyzed according to standard protocol[27].

MO sequence information

All MOs were purchased from Gene Tools, LLC, synthesized at 300nmol, except zebrafish p53 oligos at 100nmol.

Zebrafish rpl13 5’-UTR MO: TTGTTCACTCCGTCCTTAGCGGAAA Zebrafish rps15a 5’-UTR MO: CGCACCATGATGCCAGTTCTGCAAT Zebrafish rpl39 5’-UTR MO: GGATCGCAATCCGTTCACCACTATG Zebrafish p53 MO: GCGCCATTGCTTTGCAAGAATTG

Zebrafish Control MO:

Zebrafish SOHA (Semi-automated Optical Heartbeat Analysis)

Larval zebrafish (72 hpf) were immobilized in a small amount of low melt agarose (1.5%) and submerged in conditioned water. Beating hearts were imaged with direct immersion optics and a digital high-speed camera (up to 200 frame/sec, Hamamatsu Orca Flash) to record 30 seconds movies; images were captured using HC Image (Hamamatsu Corp.). Cardiac function was analyzed from these high-speed movies using semi-automatic optical heartbeat analysis software[27, 114], which for zebrafish quantifies heart period (R-R interval), cardiac rhythmicity, as well as chamber size and fractional area change. All hearts were imaged at room temperature (20-21°C). Statistical analyses were performed using Prism software (Graphpad). Significance was determined using two-tailed, unpaired student t-test or one-way ANOVA and Dunnett’s multiple comparisons post hoc test as appropriate.

Zebrafish Cardiomyocyte Cell Counts and Cardiac immunofluorescent imaging

To count cardiomyocytes, we used the expression of H2AmCherry in the nuclei (Tg(myl7:H2A-mCherry)[109] to qualify as an individual cell, performed the “Spot” function in Imaris to distinguish individual cells in reconstructions of confocal z- stacks.[115, 116] To compare data sets, we used Prism software (GraphPad) to perform Student’s t-test with two-tail distribution. Graphs display mean and standard deviation for each data set.

Whole-mount immunofluorescence was performed as previously described[18, 115–117] (see also key resources table). Confocal imaging was performed on an LSM 710 confocal microscope (Zeiss, Germany) with a 40x water objective. Exported z- stacks were processed with Imaris software (Bitplane), Zeiss Zen, and Adobe Creative Suite software (Photoshop and Illustrator 2020). All confocal images shown are projection views of partial reconstructions from multiple z-stack slices, except where noted that images are views of a single slice.

Zebrafish CRISPR/Cas9 experiments

Detailed steps were previously described[118] and we followed IDT manufacture instruction for Complexes preparation. crRNA:tracrRNA Duplex Preparation: Target- specific Alt-R® crRNA ( sequence information see below) and common Alt-R® tracrRNA were synthesized by IDT and each RNA was dissolved in duplex buffer (IDT) as 100 μM stock solution. Stock solutions were stored at -20°C. To prepare the crRNA:tracrRNA duplex, equal volumes of 100 μM Alt-R® crRNA and 100μM Alt-R® tracrRNA stock solutions were mixed together and annealed by heating followed by gradual cooling to room temperature by manufacture instruction: 95°C, 5 min on PCR machine; cool to 25°C; cool to 4°C rapidly on ice. The 50 μM crRNA:tracrRNA duplex stock solution was stored at -20°C. Preparation of crRNA:tracrRNA:Cas9 RNP Complexes: Cas9 protein (Alt-R® S.p. Cas9 nuclease, v.3, IDT) was adjusted to 25 μM stock solution in 20mM HEPES-NaOH (pH 7.5), 350 mM KCl, 20% glycerol, dispensed as 8 ul aliquots, and stored at -80°C. 25μM crRNA:tracrRNA duplex was produced by mixing equal volumes of 50 μM crRNA:tracrRNA duplex stock and duplex buffer (IDT). We used 5 μM RNP complex. To generate 5μM crRNA:tracrRNA:Cas9 RNP complexes: 1 μl 25 μM crRNA:tracrRNA duplex was mixed with 1 μl 25 μM Cas9 stock, 2 μl H2O, and 1 μl 0.25% phenol red solution. Prior to microinjection, the RNP complex solution was incubated at 37°C, 5 min and then placed at room temperature. Approximately one nanoliter of 5 μM RNP complex was injected into the cytoplasm of one-cell stage embryos to generate F0 larva.

IDT crRNA sequence information

All crRNAs were purchased from Integrated DNA Technologies, Inc., synthesized at 2 nmol with standard desalting condition.

Dr.Cas9.RPS15A.1.AC: /AltR1/rUrU rGrUrU rGrUrC rArArU rCrUrC rArCrA rGrGrG rGrUrU rUrUrA rGrArG rCrUrA rUrGrC rU/AltR2/

Dr.Cas9.RPS15A.1.AB: /AltR1/rGrC rGrUrA rCrUrA rUrGrA rCrUrU rUrArG rArGrC rGrUrU rUrUrA rGrArG rCrUrA rUrGrC rU/AltR2/

Dr.Cas9.RPS17.1.AC:

/AlTR1/rUrGrArCrUrUrCrCrArCrArUrUrArArCrArArGrCrGrUrUrUrUrArGrArGrCrUrArUrG rCrU/AlTR2/

Dr.Cas9.RPS28.1.AA:

/AlTR1/rCrUrGrGrGrArArGrArArCrUrGrGrCrUrCrCrCrArGrUrUrUrUrArGrArGrCrUrArUr GrCrU/AlTR2/

Data Availability

All data produced in the present study are available upon reasonable request to the authors.

Supplementary figures

Table 1:

Supplemental Figure 1:

Supplemental Figure 2:

Supplemental Figure 3:

Supplemental Figure 4:

Supplemental Figure 5:

Acknowledgements

This work was supported by a grant from the Wanek Foundation at Mayo Clinic in Rochester, M.N., to J.L.T., T.J.N., T.M.O., R.B. and A.R.C. We gratefully acknowledge the patients and families who participated in this study. We thank Marco Tamayo for excellent technical assistance. This work was supported by National Institutes of Health (R01 HL054732 to R.B.; R01 HL153645, R01 HL148827, R01 HL149992, R01 AG071464 to A.R.C.); by California Institute for Regenerative Medicine (DISC2-10110 to A.R.C.); by the American Heart Association (AHA Predoctoral Fellowship 18PRE33960593 to K.B.).

Additional files

Supplemental Table 1: Whole genome siRNA screen results table.

Supplemental Table 2: 292 filtered variants from poor-outcome HLHS probands. MOI – mode of inheritance

Supplemental Table 3: Segregating variants of 75H proband.

Supplemental Table 4: Subset from all 292 genes that cause CM proliferation defects and/or fly heart defects.

Supplemental Table 5: Differential gene expression lists for siRPL39 and siRPS15A in hPSCs and CMs.

Supplemental Table 6: Differential gene expression lists for siRPL39 and siRPS15A in hPSCs and CMs.

Supplemental Table 7: Differential gene expression lists for siRPL39 and siRPS15A in hPSCs and CMs.

Supplemental Table 8: Differential gene expression lists for siRPL39 and siRPS15A in hPSCs and CMs.

Supplemental Table 9: 12 genes prioritized after RNAseq and WGS analysis.