Among >4000 CHD patients enrolled and studied by WES through National Heart, Lung, and Blood Institute’s Pediatric Cardiac Genomics Consortium (PCGC) (Homsy et al., 2015; Jin et al., 2017), we identified nine heterozygous de novo variants: four LoF (LoF) and five damaging missense variants in GATA6 (Figure 1A). The congenital anomalies in these patients were consistent with previously recognized roles for GATA6 in developing the cardiac outflow tract, arterial-ventricular valves, and posterior brachial arches that form aorta and pulmonary vessels (Gharibeh et al., 2018; Laforest and Nemer, 2011; Losa et al., 2017). Among PCGC CHD patients and 61 previously reported CHD patients with pathogenic GATA6 variants (Figure 1B) there were a preponderance of outflow tract malformations, including persistent truncus arteriosus, double-outlet right ventricle, tetralogy of Fallot, as well as aortic and pulmonary valve and septation defects (Kelly, 2012). Some but not all of these patients also had extra-cardiac phenotypes including pancreatic agenesis, congenital diaphragmatic hernia, and neurodevelopmental deficits.

We considered whether the distribution of these 70 damaging GATA6 variants (61 published, 9 PCGC) across the 595 encoded GATA6 amino acids correlated with clinical phenotypes (Figure 1; Supplementary file 1A, B). All variants (43 missense, 27 LoF, including eight recurrent variants) caused CHD. Extra-cardiac phenotypes occurred in 29/70 (41%) patients, and more often with LoF (18/27) than missense (11/43) variants (p<0.001). Neurocognitive dysfunction occurred in 13 patients (18.5%). Pancreatic agenesis/hypoplasia or congenital diaphragmatic hernia occurred in 20/70 patients (28.5%) and more frequently with GATA6 LoF (14/27) than missense (9/43) variants (p=0.001).

Among 43 GATA6 missense variants, 11 variants altered residues in the DNA-binding zinc finger (ZF) domain encoded by exon 4 (amino acids 435–477), significantly more than expected by chance (p=0.0004). Nine of these 11 patients with exon four missense variants had pancreatic agenesis or congenital diaphragmatic hernia, but none of 32 patients with missense variants located elsewhere (p=1.1e-6). Within exon 4, recurrent missense mutations altered the basic arginine residue 456. Computational modeling of this domain (PyMOL software) positioned residue 456 alongside a polar residue (asparagine 466) in close proximity to DNA (Bates et al., 2008; Figure 1C). Substitution of a non-polar glycine at residue 456 (R456G) is predicted to disrupt these interactions, and potentially alter GATA6 binding to DNA.

We created GATA6 LoF hiPSCs using two independent guide RNAs (gRNAs) targeting exon 2 (Figure 2—figure supplement 1A) that were transfected with Cas9 endonuclease into an early passage healthy hiPSC line PGP1 (Lee et al., 2009). Targeted hiPSCs were subcloned and GATA6 variants were confirmed by next-generation and Sanger sequencing of PCR-amplified fragments (METHODS). Four independent mutant hiPSC lines were obtained: two carry a heterozygous 1 bp insertion (GATA6^+/-, chr18:19,752,124–19,752,124, A:TA) and two have a homozygous 1 bp deletion (GATA6^-/-, chr18:19,752,123–19,752,124, TA:T). Using similar strategies, we transfected a gRNA targeting exon four with a single-stranded DNA oligonucleotide to serve as a template for homology-directed repair and generated two hiPSC lines with homozygous missense variant R456G (GATA6^R456G/R456G; Figure 2—figure supplement 1B). No lines with a heterozygous R456G variant were obtained. In parallel we produced GATA6^+/-, GATA6^-/-, and GATA6^R456G/R456G variants in a PGP1 hiPSC line that carried expressed green fluorescent protein (GFP) fused to endogenous cardiac troponin T alleles (TNNT2-GFP; Figure 2—figure supplement 1C).

Wildtype (WT), GATA6^+/-, GATA6^-/-, and GATA6^R456G/R456G hiPSCs were processed for differentiation into cardiomyocytes (hiPSC-CMs) by modulation of the Wnt signaling pathway and subsequent metabolic selection via glucose deprivation (Sharma et al., 2018b). This protocol yields hiPSC-CMs that express first and second heart field genes (Zhang et al., 2019). As previous single-cell RNA- sequencing (RNA-Seq) of cardiomyocytes isolated from developing mouse hearts (DeLaughter et al., 2016) identified peak Gata6 expression in cardiac progenitors and early cardiomyocytes, we studied hiPSC-CMs at differentiation days 4 and 8, which approximate these in vivo developmental stages. GATA6 protein expression and nuclear localization were reduced in GATA6^+/- compared to isogenic WT lines and absent in GATA6^-/- lines (Figure 2—figure supplement 1D,E). GATA6^+/- hiPSCs, unlike WT hiPSCs, exhibited mono-allelic GATA6 expression, suggesting nonsense-mediated decay of RNAs transcribed from targeted alleles.

We assessed sarcomere production as a measure of cardiomyocyte differentiation in the TNNT2-GFP lines. GATA6^+/- hiPSC-CMs had weaker fluorescent signal than WT cells, while GATA6^-/- lines expressed no fluorescence at baseline or during the differentiation protocol (Figure 2A,B). Consistent with these data, at differentiation day eight when contracting sarcomeres were present in WT cells, fewer independent differentiation rounds of GATA6^+/- hiPSC-CMs contained beating sarcomeres and no GATA6^-/- cells showed sarcomeres or spontaneous beating (Figure 2C). Analogous studies of GATA6^R456G/R456G hiPSC-CMs (Figure 2D–F) showed a fluorescence signal comparable to that of GATA6^+/- hiPSC-CMs and spontaneous beating cells in approximately ~60% of differentiation rounds.

We assessed transcriptional responses to altered GATA6 levels throughout hiPSC-CM differentiation by bulk RNA-Seq analyses of cultures at days 0, 4, 8, 12, and 30 (Figure 3, Figure 2—figure supplement 1F, Figure 3—figure supplement 1, Supplementary file 2), and by single-cell RNA-Seq on days 4 and 8 (Figure 4) that precedes metabolic selection and enriches cultures for cardiomyocytes. All cells were differentiated, processed, and harvested in parallel. RNA-Seq data was aligned and processed to limit potential batch effects, and clustered using methods implemented in DESEQ2 (bulk) or Seurat (single cell) (METHODS). We observed consistency between bulk and single-cell expression data. In addition, principal component analyses (PCA) (Figure 3A, Supplementary file 3) of independent, genotype-identical lines demonstrated close clustering of RNA expression, indicating that GATA6 genotype and differentiation stage largely accounted for differences in gene expression.

At day 4 of differentiation, RNA-Seq of WT cells showed expression of both pluripotent stem cell markers (e.g., POU5F1 encoding OCT4) and transcriptional modulators associated with early cardiomyocyte differentiation (Figures 3B,C and 4A,B). These primordial cardiomyocytes expressed SMYD1 (SET and MYND domain-containing protein-1), a nuclear histone methyl-transferase involved in remodeling chromatin and sarcomere assembly (Li et al., 2011), Tbox transcription factors TBX5 and TBX20, two of the earliest markers of cardiac development (Bruneau et al., 2001; Takeuchi et al., 2005), and MEF2C, an essential transcription factor for sarcomere assembly and function (Lin et al., 1997). From day eight onward, the expression of stem cell marker genes was extinguished while expression of cardiomyocyte differentiation transcripts increased (Figures 3C and 4C,D).

Our protocol for cardiomyocyte differentiation also yielded subpopulations of cells expressing endodermal genes. Day 4 WT cells expressed hepatocyte nuclear factor family members (HNF1, HNF4A, HNF4B) that are depleted in Gata6-null mice (Morrisey et al., 1998), FOXA1 and FOXA2, transcription factors that activate expression of the pancreas/duodenum homeobox-1 gene PDX1, which is essential for pancreas development and β islet cell differentiation (Gao et al., 2008; Gerrish et al., 2001; Lee et al., 2019; Zhou et al., 2008) and other endodermal markers such as SOX17 (Wang et al., 2011; Figure 4A,B and Supplementary file 2). PDX1 was expressed through differentiation day 12 in WT lines, but not after metabolic enrichment for cardiomyocytes.

GATA6^+/- cells at day 4 of differentiation had lower transcript levels of primordial cardiomyocyte genes compared to WT cells (Figure 3B,C) and gene ontology (GO) analyses inferred that pathways involved in cardiac muscle contraction, development, and chamber organization were reduced. Notably, the expression of SMYD1, which regulates the expression of HAND2, a critical transcriptional regulator of the second heart field (Gottlieb et al., 2002; Laurent et al., 2017) that gives rise to the outflow tract, was strikingly diminished (Figure 3B–D, Supplementary file 2). Similarly, expression of multiple HAND2 network genes was also lower than in WT cells (Figure 3C,D), including GATA gene family members, ISL1, MEF2C, and prototypic cardiac transcription factors MEIS-1, NKX2-5, TBX5, and TBX20.

Prompted by the transcriptional signatures in GATA6^+/- cells, we compared the heart malformations in 54 PCGC patients with damaging variants in 11 HAND2-network and second heart field genes (Supplementary file 1C) with patients who have GATA6 mutations. Forty of these 54 patients also had outflow tract malformations, including all (n = 5) CHD patients with damaging de novo variants in TBX20.

GATA6^+/- cells also identified dysregulation of other gene programs involved in forming the outflow tract. The transient peak expression in WT cells of HOXA1, HOXB1, and MSX1, which regulate the induction and expression of critical molecules involved in specifying neural crest cell development (Makki and Capecchi, 2011; Simões-Costa and Bronner, 2015; Tvrdik and Capecchi, 2006), remained low throughout differentiation of GATA6^+/- cells. Notably, HOXB1 also participates in specifying endoderm destined for pancreatic and other abdominal cell lineages (Huang et al., 2002). Expression was also diminished of key molecules that couple outflow tract myocardium with developing vascular beds, including KDR (encoding vascular endothelial growth factor (VEGF) receptor-2) and the VEGF co-receptor, NRPI (Supplementary file 2). Human mutations in HOXA1 (Tischfield et al., 2005) and KDR (Reuter et al., 2019) alleles cause cardiac outflow tract malformations. Our data infers a critical role for GATA6 in orchestrating SMYD1-HAND2, VEGF, and neural crest network interactions during outflow tract morphogenesis.

Day 8 GATA6^+/- cells expressed many cardiomyocyte transcripts at levels found in WT lines, but single-cell RNA-Seq analyses demonstrated some persistent differences including reduced expression of MYH7, a defining transcript of mature cardiomyocytes (Figure 3C, Figure 4C,D and Figure 2—figure supplement 1F). Genes encoding other sarcomere proteins were reduced through day 12 in GATA6^+/- cells and abnormalities remained after metabolic enrichment for cardiomyocytes (Supplementary file 2). Compared to WT, day 30 GATA6^+/- cardiomyocytes had 10-fold higher expression levels of ISL1 and 4-fold higher levels of fetal myosin transcripts (MYH6), suggesting developmental immaturity of GATA6 mutant cardiomyocytes.

Endodermal lineage genes were variably misexpressed in differentiating GATA6^+/- cells. In comparison to WT cells, transcript levels were lower for GATA4, HNF1, and HNF4A but normal for FOXA1 and FOXA2 (Supplementary file 2). The day 12 peak expression of PDX1 in WT cells was absent in GATA6^+/- cells.

GATA6^+/- cells also misexpressed transcription factors genes involved in diaphragm development. In day 8 cells transcripts encoding NR2F2 (encoding a retinoic acid (RA) responsive transcription factor) were 2 to 3-fold higher while expression of ZFPM2 (encoding a family member of the Friend of GATA (FOG) transcription factors) was half of WT levels. As damaging variants in these genes cause congenital diaphragmatic hernia (Kardon et al., 2017), these data imply that GATA6 haploinsufficiency causes diaphragmatic dysgenesis by disrupting NR2F2 and ZFPM2 gene programs (Supplementary file 2).

GATA6^-/- hiPSC had profound deficits in cardiomyocyte differentiation (Figure 3, Figure 2—figure supplement 1F, Supplementary file 2). Transcripts associated with early mesoderm specification including MESP1 and transcription factors associated with primordial cardiomyocytes were more depressed than in GATA6^+/- cells. These findings support prior observations that shRNA knockdown of GATA6 reduces hiPSC-CM differentiation capacity (Yoon et al., 2018). Given the absence of emerging cardiomyocytes, metabolic selection to further enrich for this lineage was not performed.

The expression of endodermal genes in GATA6^-/- hiPSCs was also abnormal. Throughout differentiation GATA4 transcripts were 10-fold reduced, FOXA1 and FOXA2 were transiently expressed only at day 4, and HNF4A and PDX1 expression were extinguished (Supplementary file 2). Dysregulated expression of NR2F2 (increased) and ZFPM2 (decreased) was also observed.

Single-cell RNA-Seq (Figure 4C,D) identified two GATA6^-/- populations. One population maintained high expression of the stem cell (SC) marker POU5F1 throughout differentiation day 8, while the other (FB) had increased expression of fibroblast markers (COL3A1, IGFBP7), epithelial to mesenchymal transition markers (SNAIL1/2, MMP2, VIM) (Figure 3—figure supplement 1A), and neural differentiation genes (SHH, ZEB1/2, NCAM2) (Figure 3—figure supplement 1A,B). Together these data inferred that GATA6^-/- cells, lacking the normal signals involved in specifying cardiomyocyte and endoderm lineages, adopted alternative differentiation programs.

Parallel analyses of isogenic GATA6^R456G/R456G lines showed shared and distinct transcriptional profiles from GATA6^+/- or GATA6^-/- cells. At days 4–12, primordial cardiomyocyte transcripts (GATA4, HAND2, and SMYD1) in GATA6^R456G/R456G lines were expressed at levels midway between GATA6^+/- and GATA6^-/- cells, but markedly below levels in WT cells (Figures 3C and 4, and Supplementary file 2).

Distinctive transcription profiles in differentiating GATA6^R456G/R456G cells (days 4–12) suggested aberrant RA signaling (Figure 4C,D, Supplementary file 2). ALDH1A2 transcripts (encoding the enzyme that converts RA from retinaldehyde) were 50-fold higher than WT cells and the expression of RARA, RARB (RA receptors A and B), and STRA6 (receptor for retinol uptake) were increased 2 to 3-fold. Transcripts encoding targets of RA signaling were also increased. HOXB1, which contains two RA-responsive elements, including one 6.5 kb 3’ of coding sequences that is critical for foregut expression (Huang et al., 2002), was 100-fold increased at differentiation day 4 – a marked contrast to the depressed levels observed in GATA6^+/- cells.

RA signaling modulates neural crest cell development (Martínez-Morales et al., 2011) and at day 4 GATA6^R456G/R456G cells had increased PAX3 and PAX7 expression (100-fold and 3-fold, respectively; Supplementary file 2, Figure 3—figure supplement 1B). HOXA1 (6-fold increased) and HOXB contribute to activating expression of ZIC1 (8-fold increased) (Jaurena et al., 2015; Makki and Capecchi, 2011; Tvrdik and Capecchi, 2006) that with PAX3/7 induce and specify a neural crest gene program. PAX3 also promotes neural crest migration into the cardiac outflow tract development (Conway et al., 1997) and like STRA6, (Coles and Ackerman, 2013) and contributes to diaphragm development (Stuelsatz et al., 2012) by specifying paraxial mesoderm (Magli et al., 2019). Aberrant RA signaling may contributed to dysregulation of other developmental regulators of diaphragm formation, including NR2F2 (~8-fold increased) and ZFPM2 (5- to 12-fold decreased).

We further assessed activation of RA signaling by comparing RNA-Seq analyses of day 4 differentiating WT and GATA6^R456G/R456G cells that were cultured with or without an inhibitor of the ALDH1A2 enzyme (METHODS). Treated WT cells (Figure 4—figure supplement 1A) had significantly altered expression (p<0.05, 1.5-fold), decreasing the expression of transcripts assigned to Gene Ontology terms for heart and muscle cell differentiation and increasing endodermal fate specification and Wnt signaling transcripts. Treated GATA6^R456G/R456G cells (Figure 4—figure supplement 1B) had increased expression of transcripts assigned to Gene Ontology terms for myoblast differentiation, circulatory system development and anatomical structures. Notably, the expression of 25% of genes in treated GATA6^R456G/R456G cells (Figure 4—figure supplement 1C–F) had levels found in untreated WT cells (HOXB1 and FOXH1) or normalized levels, approaching those in untreated WT cells (HOXA1, HAND2, RARB, TBX20).

GATA6^R456G/R456G cells had aberrant expression of other endoderm genes, including markedly lower transcripts of HNF gene family members and striking loss (≥100-fold below WT) of FOXA1 and FOXA2 expression (Supplementary file 2). Biallelic deletion of FOXA2 in human stem cells prevents pancreatic specification (Lee et al., 2019). GATA6^R456G/R456G cells also had increased expression of SOX6, which represses PDX1-dependent transcriptional activation of the insulin gene, INS (Iguchi et al., 2005). The cumulative effect of these deficits could account for the extinguished expression of PDX1 and contribute to the high association of pancreatic agenesis with heterozygous GATA6^R456G mutations (Figure 1).

Despite multiple transcriptional abnormalities, metabolic enrichment yielded GATA6^R456G/R456G cardiomyocytes (day 30) with prototypic gene expression. However, like GATA6^+/- hiPSC-CMs, GATA6^R456G/R456G hiPSC-CMs exhibited a higher MYH6:MYH7 transcript ratio than WT, indicating immaturity.

We studied potential mechanisms by which GATA6 regulated gene transcription by performing Assays for Transposase-Accessible Chromatin sequencing (ATAC-seq), GATA6 chromatin immunoprecipitation sequencing (ChIP-seq) and Hypergeometric Optimization of Motif EnRichment (HOMER) analyses at day 4 (Figure 5, Figure 6) and day 8 (Supplementary file 4). ChIP-seq of GATA6^-/- cells yielded very few peaks above background (data not shown), confirming antibody specificity.

At day 4, GATA6 bound ~21,000 open chromatin peaks genome-wide, of which ~ 21% are near promoters. Notably, only 12% of GATA6-bound open chromatin peaks overlapped with ATAC peaks and were markedly enriched for a GATA-binding motif (Figure 5A). ChIP-seq data from day 8 cells were similar (Supplementary file 4). These data inferred that GATA6 directly participates in regulating transcriptional activation of EOMES, GATA4, MEIS1/2, NKX2.5, TBX5, and TBX20 (Supplementary file 4).

By contrast, 88% of GATA-bound peaks occurred in closed chromatin (Figure 5B). Sequences within these peaks were not enriched for the GATA motif but were highly enriched for the MEF2 binding motif. MEF2 proteins contain a DNA-binding domain (MADS) and transcriptional activating domain (TAD) (Backs and Olson, 2006) that promote differentiation gene programs for cardiac, skeletal, and smooth muscle myocytes (Potthoff and Olson, 2007). During cardiomyocyte differentiation of WT hiPSCs MEF2 transcripts increased 120-fold between days 4–8 (Supplementary file 2). In comparison, at day 8–12, MEF2A and MEF2C levels were decreased in GATA6^+/- cells. As together, these data implied that GATA6 functioned as a cardiac pioneer factor, which has been proposed for other GATA proteins, we examined the correlation between GATA6 bound to closed chromatin at day 4 and gene expression at day 5 in WT cells. There were 2878 genes differentially expressed between day 4 and day 5 WT cells. Of those, 1049 were associated with GATA6-bound closed chromatin (36.4%). Among these, 583 (56%) genes had increased expression at day 5, including key cardiac developmental transcription factors, GATA4, SMYD1, KDR, and TBX5 (Figure 5—figure supplement 1, Supplementary file 5). Based on these data, we propose that GATA6 is a cardiac development pioneer factor.

In comparison to WT, day 4 GATA6^+/- cells had~6500 diminished GATA6 ChIP-seq peaks and ~5800 diminished ATAC peaks (Figure 6A and Figure 6—figure supplement 1). Diminished ATAC peaks in proximity to promoters were enriched for GATA, HAND, HNF, and SOX motifs (Figure 6—figure supplement 1, Figure 3, Figure 4 and Supplementary file 2), indicating that attenuated expression of these transcription factor family members could have direct functional consequences. Notably however, few diminished GATA-bound peaks resided within open chromatin (10.7%), of which only 35% were enriched for GATA motifs (Figure 5—figure supplement 1B, Figure 6C). Instead, and consistent with properties of a pioneer factor, GATA6^+/- cells had prominent loss (89%) and gain (97%) of GATA-bound peaks within closed chromatin (Figure 6B–D). From these data we infer that GATA interactions with closed chromatin are highly sensitive to protein dosage, while interactions with open chromatin at promoter regions remain relatively intact despite reduced GATA6 protein levels.

By contrast, day 4 GATA6^-/- cells contained~29,000 diminished ATAC-seq peaks of which ~ 31% resided in proximity to promoters (Figure 6—figure supplement 1). Profiles at day 8 cells were similar (Figure 6—figure supplement 2). Notably, diminished ATAC peaks showed enrichment in the binding motif for NFY that encodes the ubiquitous promoter element binding factor of CCAAT-boxes. These data implied that extinguishing GATA6^-/- expression caused widespread deficits in promoter activation, which likely accounted for failed mesoderm specification and cardiomyocyte differentiation.

ChIP-seq analyses of GATA6^R456G/R456G cells (Figure 5C–F, Figure 5—figure supplement 1, Supplementary file 4) identified ~8000 diminished peaks. In comparison to other mutant lines, GATA6^R456G/R456G cells had a significantly higher proportion of diminished peaks (21%) in open chromatin that contained a GATA or HAND binding motif and 19% of these diminished peaks were associated with differential gene expression (Figure 5—figure supplement 1C). Additionally, missense cells had a remarkable number (~44,000) of augmented and ectopic peaks. We suggest that these aberrant chromatin interactions reflected direct promiscuous GATA6^R456G binding activity or perhaps occurred in response to aberrant RA signaling.

Among the upregulated peaks,~4% of these peaks had GATA-bound to open chromatin, with enrichment of the CTCF binding motif. Furthermore,~12% were associated with differential gene expression, including MSX1, FOXA1, MEIS2, HCN4, and HOXA2 (Supplementary file 4). However, the vast majority of increased or ectopic peaks (96%) resided within closed chromatin and lacked the GATA6 binding motif. Even so, 14% of the differential closed peaks were associated with differential gene expression, including peaks in TBX20, FOXA1, and FOXH1. Our data suggest that the GATA6^R456G variant impairs binding to the GATA motif, reducing normal GATA6 function, and also promotes promiscuous binding with either repressive or activating transcriptional effects.

We examined ATAC-seq and GATA6-ChIP-seq data to identify potential mechanisms for altered gene transcription with relevance to clinical phenotypes in CHD patients with pathogenic GATA6 variants. GATA6-bound peaks were universally decreased in GATA6^+/- cells, including those associated many cardiac developmental transcription factors genes (e.g., HAND2, KDR, and TBX5 Figure 7A–C) and these changes were associated with normal or modest reduction in transcript levels (Supplementary file 2). While many of these GATA6-bound peaks were also diminished in GATA6^R456G/R456G cells, the missense cells also had many enhanced and augmented peaks which were associated greater differential (decreased or increased) gene expression. For example, only GATA6^R456G/R456G cells had diminished GATA6-bound peaks identified in WT cells associated with GATA4 (Figure 7D) and MEF2A and also new GATA6-bound peaks in MEF2C, ZIC1 and ZIC3 (Figure 7E). The distinct epigenetic profiles in GATA6^R456G/R456G may contribute to the greater dysregulation of cardiac gene expression and cardiomyocyte maturation in comparison to GATA6^+/-cells (Figures 2 and 3 A-C). However regardless of whether epigenetic changes in GATA6^R456G/R456G cells decreased normal gene expression or erroneously activated the expression of other genes, the associated cardiac consequence of altered transcription was similar to that of GATA6^+/-cells – disruption of cardiac outflow tract development (Figure 1).

By contrast, altered GATA6 binding and chromatin accessibility with accompanying dysregulated transcripts likely contributed to the extra-cardiac phenotypes that disproportionately affect patients with heterozygous GATA6 R456G and other exon four missense variants. Pancreatic development requires cooperative GATA4 -GATA6 interactions and compound Gata4-null and Gata6-haploinsufficient mice have reduced Pdx1 expression, pancreatic agenesis or profound hypoplasia (Carrasco et al., 2012). In addition to reduced occupancy of GATA6-motifs in GATA4 with striking attenuated expression (Figure 7D, Supplementary file 2, Supplementary file 4), GATA6 R456G also aberrantly bound chromatin associated with FOXA2, DEANR1 (LINC00261), and PDX1 and diminished transcription of these genes (Figure 7—figure supplement 1). DEANR1 encodes a long noncoding RNA that specifies endoderm and pancreatic lineages by regulating the expression of FOXA2 (Jiang et al., 2015) which in turn regulates PDX1 expression (Gao et al., 2008).

Aberrant RA signaling may have contributed to the abnormal epigenetic profiles in GATA6^R456G/R456G cells. In missense cells, prominent ATAC peaks were identified in association with binding motifs for GATA, SOX6, HOX, and RARa in the promoter and exon 1 of ALDH1A2 and transcript levels were increased. SOX6 also had augmented ATAC peaks with RARa and GATA motifs, and GATA6^R456G bound an intron one motif (Figure 7—figure supplement 2A–B). GATA6^R456G/R456G cells also showed increased ATAC peak heights with RARa binding motifs in close proximity to the reported RA-responsive element in HOXB1 that promotes foregut expression. We also observe GATA6^R456G protein aberrantly bound to the STRA6 locus, potentially affecting expression of this gene (Figure 7—figure supplement 2C) These data suggest multiple direct and indirect epigenetic mechanisms by which the GATA6^R456G variant distinctly perturbed gene programs and resulted in maldevelopment of the diaphragm, pancreas and other abdominal organs in addition to cause cardiovascular malformations.

CHD subjects were recruited to the Congenital Heart Disease Network Study of the Pediatric Cardiac Genomics Consortium (CHD GENES: ClinicalTrials.gov identifier NCT01196182) after approval from Institutional Review Boards as previously described (Jin et al., 2017; Gelb et al., 2013). Written informed consent was received from subjects or their parents prior to inclusion in the study. Any subject with CHD, regardless of age, race, or ethnicity was eligible for enrollment. Subjects with variants in GATA6 and 2^nd heart field genes including CHD diagnoses, and extra-cardiac phenotypes are provided in Supplementary file 1.

DNA was extracted peripheral blood samples and captured using the Nimblegen v.2 exome capture reagent (Roche) or Nimblegen SeqxCap EZ MedExome Target Enrichment Kit (Roche) followed by Illumina DNA sequencing as previously described (Jin et al., 2017). Reads were mapped to the reference genome (hg19), and further processed using the GATK Best Practices workflows as previously described (Jin et al., 2017; McKenna et al., 2010). Single nucleotide variants (SNVs) and small indels were called with GATK HaplotypeCaller and filtered for rarity (ExAC allele frequency ≤10–5) and quality as previously described (Jin et al., 2017). The MetaSVM algorithm was used to predict deleteriousness of de novo missense mutations (annotated as ‘D-Mis’) using software defaults (Dong et al., 2015; Sulahian et al., 2014).

All hiPSCs and hiPSC-CMs used here are derived from the male parent iPSC line PGP1 (Akerman et al., 2017; RRID-GM23338) and are mycoplasma-free. The TNNT2-GFP iPSC line was generated by homology-directed repair, using a plasmid with the endogenous TNNT2 sequence connected to a GSSSS linker region, attached to the GFP gene, and the entire construct was flanked by 500 bp homology arms. PGP1 iPSCs were transfected as described (Sharma et al., 2018b) to obtain homozygous integration of the GFP tag in both TNNT2 alleles. Edited clones were selected using puromycin and expanded and genotyped as previously described (Sharma et al., 2018b) and differentiated into cardiomyocytes as described (Sharma et al., 2018a).

GATA6^+/- and GATA6^-/- hiPSCs were generated from TNNT2-GFP and PGP1 iPSCs by non-homologous end joining using a 2 µg plasmid expressing Cas9 (PX459v2 from Addgene) that was co-transfected with 2 µg plasmid expressing guide RNA (provided in Key Resource Table) using a stem cell nucleofector kit (Amaxa). Edited clones were selected using puromycin and expanded and genotyped as previously described (Sharma et al., 2018b) and differentiated into cardiomyocytes as described (Sharma et al., 2018a).

GATA6^R456G/R456G hiPSCs were generated by homology-directed repair using a 2 µg plasmid expressing Cas9 (PX459v2 from Addgene) that was co-transfected with 2 µg plasmid expressing guide RNA (provided in Key Resource Table) and 2 µg single-stranded oligonucleotide (HDR template) using a stem cell nucleofector kit (Amaxa). Edited clones were selected using puromycin and expanded and genotyped as previously described (Sharma et al., 2018b) and differentiated into cardiomyocytes as described (Sharma et al., 2018a).

All edited lines sub-cloned, by dissociating 1000 genome-edited hiPSCs cells, filtering through a 60 µm strainer, and evenly distributing them onto a 15 cm dish containing Matrigel and mTeSR+ rho kinase inhibitor. Individual monoclonal hiPSC colonies were picked when colony size reached approximately 200 cells and placed into individual separate wells of a 96 well plate. Clones were allowed to grow to 80% confluency, at which time a sample was obtained for PCR amplification to verify the GATA6 variant or GFP-tagged TNNT2 and to assess zygosity. PCR-amplified fragments (primers provided in Key Resource Table) containing putative variants were submitted for Sanger sequencing and next-generation sequencing. Sanger sequencing traces were deconvoluted using TIDE software to confirm zygosity. Clones carrying GATA6 mutations were further expanded for subsequent culture and differentiation.

The hiPSC-CMs were generated from GATA6 mutant hiPSCs using a small-molecule mediated differentiation approach that modulates Wnt signaling (Sharma et al., 2018a). Cells began beating at approximately day 7 post-differentiation. Cardiomyocytes were metabolically selected from other differentiated cells by using glucose deprivation as previously described (Sharma et al., 2018a). hiPSC-CM Western blots: GATA6 protein expression and nuclear localization were performed using ab175927.

RNA-seq experiments were performed on at least two independent cultured cell samples per time point, with the exception of GATA6^R256G/R256G DMSO control in the RA-signaling experiment. Trizol (Thermo Fisher) was used to harvest differentiating GATA6 mutant hiPSC-CMs (days 0, 4, 8, 12, 30 of differentiation) designated for RNA-sequencing analysis. Samples harvested in Trizol were stored at −80°C until RNA was extracted. All RNA samples had an RNA integrity number of ≥8. Library preparation was conducted using the Nextera library preparation method. RNA-Seq library samples were pooled and run on four lanes (one flow cell) using the Illumina NextSeq500 platform. All data was combined into a single fastq file. Samples typically had 30–50M reads each. The raw reads were aligned by HISAT2 (v.2.1.0) to human genome (hg38). The aligned reads were quantified by FeatureCounts, counts were normalized, and differentially expressed genes were identified using DESeq2 (v1.24.0). DESeq2 data was analyzed and visualized in R using the ggplot2 (v3.1.0), pheatmap (v1.0.12), gProfiler (v0.6.7), and VennDiagram (v1.6.20) packages. Single-cell RNA-seq analysis for hiPSC-CMs was conducted using a 10x Genomics Chromium platform and analyzed in R using the Seurat (v. 3.5.1) pipeline. Seurat was used to regress out cell-cell variation in cell complexity, gene expression driven by batch, cell alignment rate, the number of detected molecules, and mitochondrial gene expression. Starting with Seurat v2.0, Seurat implements this regression as part of the data scaling process.

ATAC-seq was performed as previously described (Buenrostro et al., 2015; Corces et al., 2017). Briefly, 50,000 cells were harvested and lysed to isolate nuclei. Nuclei were treated with Tn5 transposase (Nextera DNA Sample Prep Kit, Illumina) and DNA was isolated. Fragmented DNA was amplified using bar-coded PCR primers (defined in Buenrostro et al., 2015) and libraries were pooled. Pooled libraries were sequenced (Illumina Next-Seq) to a depth of 100 million reads per sample. Reads were aligned to the hg19 reference genome using BWA-MEM and peaks were called using HOMER (v4.10.3) (http://homer.ucsd.edu/homer/index.html). Functional analysis of ATAC-seq peaks was performed using ChIP-Seeker (v.1.14.1). Motif enrichment was performed using HOMER (v4.10.3). Differential peaks were called using HOMER (v4.10.3).

Cells (6-well plates) were grown to approximately 80% confluency and then fixed with 1% fresh formaldehyde diluted in media. Cells were quenched with glycine, washed, and harvested. Cell pellets were stored at −80°C. Cells were thawed and nuclei were prepared using the Covaris Tru ChIP Chromatin Shearing Kit (Covaris, MA, USA). Chromatin was sheared using the Covaris E210 to an average fragment size of 200–700 bp and an input sample was purified. Chromatin (5 µg) was incubated with GATA6 antibody (RRID:AB_5851S) at 4°C O/N and then added to Protein G beads for 2 hr at 4°C. Beads were washed and bound chromatin was eluted at 65°C for 30 min on a heated vortex. DNA was then purified and quantified. ChIP-seq libraries were prepared using Nextera XT DNA sample prep (Illumina). Sequences were aligned to hg19 using BWA-MEM. Peaks were called using MACS2 using default parameters with a q value of 0.001. Differential peaks were called using HOMERv4.10.3.

hiPSCs were seeded and differentiated to day 4, and then treated for 24 hr with either DMSO or 1 µM WIN 18446 (Tocris), an inhibitor of the ALDH1A2 enzyme. Cells were harvested and processed for RNA-seq.

The distribution of damaging human variants across the GATA6 gene was analyzed using the binominal test, implemented in R. The association of GATA6^+/- and GATA6^R456G/+ variants and pancreatic agenesis or congenital diaphragmatic hernia was analyzed using the Fisher Exact test, implemented in R. Transcriptional responses of iPSC-CMs with or without a GATA6 variant was compared using the Student’s t-test was used for comparison. All error bars refer to standard deviation, unless otherwise specified. A p value of < 0.05 was considered significant.