Abstract
Congenital Heart Defects (CHDs) are the most common form of birth defects, observed in 4-10/1000 live births. CHDs result in a wide range of structural and functional abnormalities of the heart which significantly affect quality of life and mortality. CHDs are often seen in patients with mutations in epigenetic regulators of gene expression, like the genes implicated in Kabuki syndrome – KMT2D and KDM6A, which play important roles in normal heart development and function. Here, we examined the role of two epigenetic histone modifying enzymes, KMT2D and KDM6A, in the expression of genes associated with early heart and neural crest cell (NCC) development. Using CRISPR/Cas9 mediated mutagenesis of kmt2d, kdm6a and kdm6al in zebrafish, we show cardiac and NCC gene expression is reduced, which correspond to affected cardiac morphology and reduced heart rates. To translate our results to a human pathophysiological context and compare transcriptomic targets of KMT2D and KDM6A across species, we performed RNA sequencing (seq) of lymphoblastoid cells from Kabuki Syndrome patients carrying mutations in KMT2D and KDM6A. We compared the human RNA-seq datasets with RNA-seq datasets obtained from mouse and zebrafish. Our comparative interspecies analysis revealed common targets of KMT2D and KDM6A, which are shared between species, and these target genes are reduced in expression in the zebrafish mutants. Taken together, our results show that KMT2D and KDM6A regulate common and unique genes across humans, mice, and zebrafish for early cardiac and overall development that can contribute to the understanding of epigenetic dysregulation in CHDs.
Introduction
Heart disease is the leading cause of deaths in the United States and worldwide, and congenital heart defects (CHDs) are a leading cause of birth defect-associated infant illness and death (Benjamin et al., 2019; Gilboa, Salemi, Nembhard, Fixler, & Correa, 2010; Oster et al., 2013; Ottaviani & Buja, 2017). CHDs are present at birth, and they have a profound impact on the structure and function of an infantile heart. CHDs span a diverse spectrum of defects ranging from mild small atrial septal defects which heal on their own, to severe large ventricular septal defects which require medical intervention. In some cases, CHD-associated complications arise in adulthood, while for critical CHDs, surgery is required in the first year of life.
Significant gaps exist in our knowledge of the causes of CHDs, which may involve genetic, epigenetic, and environmental factors. Although epigenetic signatures like DNA methylation, histone modifications, chromatin remodeling, and microRNA have begun to be explored in CHDs and embryonic development in general, their underlying mechanisms in cardiogenesis are incompletely understood. Nonetheless, epigenetic modifications significantly contribute to the etiology and prognosis of CHDs and other developmental disorders because patients with same genotypes and syndromes often have large variations in phenotype, penetrance, and severity. Indeed, mutations in chromatin remodeler Chd7 and histone acetyltransferase Kat6a/b are associated with birth defects leading to Charge and Ohdo syndromes, respectively (Vissers et al., 2004) (Campeau et al., 2012).
Several studies have reported links between early heart development and two epigenetic modifiers - KMT2D, which methylates the histone H3K4 (histone 3 lysine 4), and KDM6A, which demethylates H3K27 (histone 3 lysine 27) (Ali, Hom, Blakeslee, Ikenouye, & Kutateladze, 2014; Hong et al., 2007; Koutsioumpa et al., 2019; Lan et al., 2007). Mutations in KMT2D and/or KDM6A result in Kabuki Syndrome (KS) where 28-80 % patients are afflicted with CHDs (Digilio et al., 2017) like atrial and ventricular septal defects, bicuspid aortic valves, aortic coarctation, double outlet right ventricle, transposition of great arteries, infundibular pulmonary stenosis, dysplastic mitral valve, Tetralogy of Fallot, etc. (Cocciadiferro et al., 2018; Digilio et al., 2017; Gazova, Lengeling, & Summers, 2019; Luperchio, Applegate, Bodamer, & Bjornsson, 2019; Shangguan et al., 2019; Tekendo-Ngongang, Kruszka, Martinez, & Muenke, 2019; Yap et al., 2020). Several variants and missense mutations of KMT2D and KDM6A have been identified in KS and other novel syndromes (Cocciadiferro et al., 2018; Cuvertino et al., 2020; Xin et al., 2018). In addition to CHDs, KS patients have typical dysmorphic features like long palpebral fissures with eversion of lateral third of the lower eyelid, arched and broad eyebrows with notching or sparseness of lateral third, short columella with depressed nasal tip, prominent or cupped ears, and persistent fingertip pads (Adam et al., 2019).
Recent studies have significantly contributed to our understanding of cardiac development and function as regulated by KMT2D and KDM6A (Ang et al., 2016; Lee, Lee, & Lee, 2012; Schwenty-Lara, Nurnberger, & Borchers, 2019; Serrano, Demarest, Tone-Pah-Hote, Tristani-Firouzi, & Yost, 2019). KMT2D or lysine-specific methyltransferase 2D, is also known as MLL2 or myeloid/lymphoid or mixed-lineage leukemia 2. It belongs to a family of 7 SET1-like histone methyltransferases (Ali et al., 2014). KMT2D contains a SET domain for histone methylation, PHD and coiled-coil domains, and zinc finger domains. KDM6A (and kdm6l in zebrafish) or Lysine Demethylase 6A, is also known as UTX or Ubiquitously transcribed tetratricopeptide repeat, X chromosome. UTX, UTY and JMJD3 comprise a subfamily of proteins containing JmjC-domain (Jumonji C) and zinc fingers, and they are evolutionarily conserved from Caenorhabditis elegans to human (Klose, Kallin, & Zhang, 2006; Lan et al., 2007; Shi & Whetstine, 2007). The JmjC-domain is associated with histone demethylation (Accari & Fisher, 2015), while the tetratricopeptide repeats at N-terminal regions of UTX and UTY are predicted protein interaction motifs (Lan et al., 2007). It is important to note that KDM6A or UTX is an X-linked gene, with 2 copies in females (XX) and 1 copy in males (XY), where the Y chromosome has the homolog, UTY (Itoh et al., 2019). Studies indicate that UTY and UTX diverged from a common ancestor, yet the functional role of UTY in histone demethylation needs further investigation, because some studies show that it has a lower histone demethylation activity than UTX (Faralli et al., 2016; Hong et al., 2007; Itoh et al., 2019; Lan et al., 2007; Shpargel, Starmer, Wang, Ge, & Magnuson, 2017; Shpargel, Starmer, Yee, Pohlers, & Magnuson, 2014; Walport et al., 2014).
To elucidate the roles of KMT2D and KDM6A in heart and overall development, we performed an inter-species comparison of common gene targets of KMT2D and KDM6A from zebrafish, mice, and human RNA-seq datasets, and an in vivo analysis of mutations in zebrafish genes kmt2d, kdm6a and its paralog kdm6al, during early development. We mutated kmt2d, kdm6a and kdm6al by CRISPR/Cas9 mutagenesis, while the human data are derived from RNA-sequencing analysis of two patients with Kabuki Syndrome (KS). One of the KS patients has a T to C substitution at position 4288 in KMT2D cDNA, resulting in C to R amino acid change at position 1430, while the other patient has a 3.2-MB microdeletion on the X chromosome (chrX:43620636-46881568) leading to a deletion of KDM6A (Van Laarhoven et al., 2015). Our results reveal that some of the genes which are involved in cardiogenesis, cardiac function, vasculature, neurogenesis, and overall development, are regulated by KMT2D and KDM6A across zebrafish, mice, and humans. Our in vivo analysis in CRISPR-mutated zebrafish indicate that kmt2d and kdm6a regulate genes involved in early embryonic, cardiac, and neural crest cell (NCC) development. Although the mechanism of the phenotypes are yet to be elucidated, our results corroborate current findings in the field, and present a pan-species perspective of the roles of KMT2D and KDM6A in heart and overall development. The study strengthens the establishment of zebrafish as a model to study CHDs resulting from KS and epigenetic dysregulation, and identifies genes targeted by KMT2D and KDM6A across humans, mice, and zebrafish, which can serve as either potential biomarkers or therapeutic targets in heart disease.
Materials and Methods
Animals
Zebrafish were maintained as previously described (Westerfield, 2000). Embryos were raised in defined Embryo Medium at 28.5°C and staged developmentally following published standards as described (Kimmel, Ballard, Kimmel, Ullmann, & Schilling, 1995). The Institutional Animal Care and Use Committee of the University of Colorado Anschutz Medical Campus approved all animal experiments performed in this study, and conform to NIH regulatory standards of care.
Genotyping
For fish, fin clips, single whole embryos or single embryo tails were lysed in Lysis Buffer (10 mM Tris-HCl (pH 8.0), 50 mM KCl, 0.3% Tween-20, 0.3% NP-40, 1 mM EDTA) for 10 minutes at 95° C, incubated with 50 ug of Proteinase K at 55°C for 2 hours, followed by another incubation at 95°C for 10 minutes. For genotyping kmt2d mutants, the following primers were used (F) 5’-AACAGAGGCAGTACAAAGCCAT-3’ and (R) 5’-TTGTTGTTGCATCAATAGGTCC-3’. The 300 bp PCR product was run out on a high percentage (4%) agarose gel to distinguish homozygous mutant embryos from heterozygous and wildtype embryos. This 300 bp PCR product was diluted 1:100 in nuclease-free water. A second set of primers was designed using dCAPS Finder 2.0 (http://helix.wustl.edu/dcaps/) to generate a primer containing a mismatch that creates a restriction endonuclease site based on the identified SNP mutation in the kmt2d mutant allele. Using above forward primer and (R) 5’-GCATTCCTACTGCCGAAGCATCAAGAGAAG −3’, a PCR product was generated that was then digested using the restriction enzyme, MboII, which digests the wildtype sequence but is unable to cut the mutant sequence, allowing for homozygous mutant embryos to be distinguished from heterozygous and wildtype embryos. For kdm6a, the genotyping primers are (F) 5’-ATGAAATCGTGCGGAGTGTCGGT-3’ and (R) 5’-ATAGAGGACAACACAGGCGGTTC-3’. For dCAPS, above forward primer and (R) 5’-CTGCAATCATTTGCACCGCTTTGAGCAT-3’ are used, followed by restriction enzyme SphI. For kdm6al, the genotyping primers are (F) 5’-GGCTTTTATTGTGGCTTTTGGAC-3’ and (R) 5’-AACTATCAAAAGCGCAGAAGCC-3’. For dCAPS, (F) 5’-GAGGAGAAGAAAATGGCGGCG GGAAAAGCGAGTGAACTCGA −3’, above reverse primer, and restriction enzyme is XhoI are used.
Generation of Zebrafish Mutant Lines
Zebrafish mutant lines for kmt2d and kdm6a and kdm6al (kdm6a/l) were generated via CRISPR-based mutagenesis. CRISPR target sites were identified using Zifit Targeter http://zifit.partners.org/ZiFiT/ and CHOPCHOP https://chopchop.cbu.uib.no/. Selected targets were used to generate guide RNAs (gRNAs) specific for kmt2d or kdm6a. gRNAs were prepared as previously described (Gagnon et al., 2014). Prepared gRNAs were then co-injected with 600 ng/ul of Cas9 Protein (PNA bio) and 200 mM KCl. Cas9/sgRNA complexes were formed by incubating Cas9 protein with sgRNA at room temperature for 5 minutes prior to injecting into the cytoplasm of wild-type zebrafish embryos at the 1-cell stage. Mosaic germline founders were identified by screening the progeny of CRISPR injected fish grown to adulthood. Two separate alleles were identified for each single mutant gene, and we focused on the alleles schematized in Figure 1. The kmt2d and kdm6a/l mutant alleles are predicted frameshift mutations that interrupt the coding sequence upstream of the SET and JmjC domains, respectively, which are responsible for functional histone methyltransferase and demethylase activities in the corresponding genes.
In situ hybridization
Zebrafish embryos were collected at 48 hours post fertilization and fixed in 4% paraformaldehyde overnight and dehydrated in methanol. Following rehydration in methanol gradients, embryos were bleached (3% hydrogen peroxide and 0.5% potassium hydroxide) until body pigment cells and the eyes were visibly cleared (10 minutes), digested with proteinase K for 8 minutes and fixed in 4% paraformaldehyde for 25 mins at 4°C. Embryos and DIG-labeled probes (sox10, foxd3, crestin, hand2a, mef2ca, kmt2d, kdm6a, kdm6al) were prehybridized for 24 hours at 65°C before incubating embryos with the indicated probes overnight at 65°C. Embryos were subjected to a series of stringency washes in a wash buffer containing 50% formamide, 2X SSC pH 4.5, and 1% SDS before blocking in a buffer containing 2% Blocking Reagent (Roche) and 20% Heat-Inactivated sheep serum diluted in 1X MABT (100 mM maleic acid, 150 mM NaCl, 0.1% Tween-20) and anti-DIG antibody (Roche) incubation overnight at 4°C. After another set of stringency washes in NTMT (100 mM NaCl, 100 mM Tris pH 9.5, 50 mM MgCl2, 1% Tween-20), embryos were developed with BM purple reagent (Roche) and stored in PBS.
RNA isolation and qPCR
Total RNA was isolated from kmt2d-/-, kdm6a-/-, and wildtype 48 hpf embryos with TRIzol reagent (Invitrogen) and pheno/chloroform. RNA (1-1.5 μg) was reverse transcribed to complementary DNA (cDNA) with SuperScript III First-Strand Synthesis cDNA kit (Invitrogen). Using Roche FastStart Universal Probe Master (Rox) (Catalog# 4914058001), realtime semiquantitative PCR (qPCR) was performed in Applied Biosystems StepOne Plus, with the following primers: sox12 Pr#6 F: 5’ tgacttgcaccacaactgct 3’ R: 5’ gcaactcaccaattcctgct 3’ slc22a23 Pr#56 F: 5’ ggcagtgatagtctgccctct 3’ R: 5’ gcagagactcgggaaaagc 3’ foxp4 Pr#68 F: 5’ gacaaggcctggtgaaccta 3’ R: 5’ tgctgcaggctctgaatg 3’ nr2f6a Pr#2 F: 5’ aaggccatagcgctcttct 3’ R: 5’ ctcggtaagagccacctgag 3’ six3a Pr#67 F: 5’ aacaggagacaacgagacagg 3’ R: 5’ gccattctgccctattgct 3’ id3 Pr#25 F: 5’ gctatgtgccattaggatgga 3’ R: 5’ ggctcaccccttcactctct 3’ prdm1a Pr#66 F: 5’ ccacattgccaccacaacta 3’ R: 5’ gtagcccttgaggtggaacc 3’ anxa4 Pr#68 F: 5’ agaggatgatgcccagaaaa 3’ R: 5’ atgatagtcgcctcgttggt 3’ Pr followed by the number represents probe numbers from Roche Universal Probe Library. Transcript levels were normalized to actb1 as reference gene, (F): 5’-GGCAGCGATTTCCTCATC-3’, and (R):5’-TCACTCCCCTTGTTCACAATAA-3’. Transcript abundance and relative fold changes in gene expression were quantified using the 2-ΔΔCt method relative to control.
Western Blot
For western blotting, 30 to 40 zebrafish embryos of each genotype were collected and pooled at 48 hpf. Embryos were incubated on ice for 5 minutes. Calcium-free Ginzberg Fish Ringer’s solution (NaCl, KCl, NaHCO3) was added to the embryos for deyolking. Samples were centrifuged for 1.5 minutes at 5000 rpm. Pellets were washed in deyolking wash buffer (NaCl, Tris (pH 8.5), KCL and CaCl2). Samples were again centrifuged for 1.5 minutes at 5000 rpm. All liquid was removed, and pellet was resuspended and lysed in SDS sample buffer (0.1% glycerol, 0.01% SDS, 0.1 M Tris (pH 6.8)) for at least 10 minutes on ice. Embryos were homogenized in lysis buffer and briefly centrifuged. Total protein concentrations were determined with the Bio-Rad Dc Protein assay. Proteins (20 ug) were separated by SDS-PAGE (12%) and transferred to polyvinylidene difluoride membranes. Membranes were blotted using antibodies for H3K4me3 (Abcam, ab8580), H3K27me3 (Abcam, ab6002) and H3 (Cell Signaling Technology, 9715S), and secondary antibodies i.e. goat anti-rabbit for H3K4me3 and H3, and goat anti-mouse for H2K27me3, conjugated with HRP (horseradish peroxidase). Chemiluminescent detection was performed with Luminata Classico Western HRP Subtrate (Millipore) and imaged on a BioRad Chemidoc multiplex imager.
Whole Mount Immunofluorescence
Zebrafish embryos were collected at the indicated timepoint and fixed in 4% paraformaldehyde overnight at 4°C embryos were washed and dehydrated through a graded methanol series. Samples in 100% methanol were incubated at least overnight in methanol prior to the next steps. Following rehydration through the reverse methanol gradient, embryos were washed in 1x Phosphate Buffered Saline (PBS, pH 7.3) with 0.1% Tween 3 times for 10 minutes at room temperature, then equilibrated in 150 mM Tris pH 9.5 for 5 minutes at room temperature before a 20 minute incubation in the Tris solution at 70°C. Following this retrieval step, embryos were washed 3 times for 10 minutes each in PBS with 0.1% Tween at room temperature then 2 times for 5 minutes in distilled water before blocking at room temperature in blocking solution containing 10% normal goat serum and 1% bovine serum albumin in PBS. Embryos were then incubated in primary antibodies for 3 days at 4°C. Primary antibodies anti-MF20 and anti-GFP (Life Technologies, Conjugate Alexa Fluor® 488 A-21311), were diluted in blocking solution. For anti-MF20, we acknowledge the Developmental Studies Hybridoma Bank (DSHB) at the University of Iowa, and Dr. Donald Alan Fischman for depositing MF20 to DSHB (DSHB Hybridoma Product MF 20, Antibody Registry ID: AB_2147781). Following primary antibody incubation, samples were washed thoroughly in PBS with 0.1% Triton X-100. Next, samples were incubated in fluorescently tagged goat anti-mouse 594 secondary antibody (Invitrogen, A-21044) for 2 days at 4°C. After secondary antibody incubation, embryos were washed thoroughly in PBS with 0.1% Triton X-100 before incubating in DAPI diluted in PBS for 1 hour at room temperature. Samples were quickly washed in PBS and passed through a glycerol/PBS gradient before mounting on glass slides. Embryos were imaged on a Leica TCS SP8 confocal microscope. Images were processed in LASX software and quantification of cell numbers were completed on max projections of z-stack images using Image J.
Human Cell Samples
Patient-derived lymphoblastoid cell lines (LCLs) used in this study were previously generated from KS patients with mutations in KMT2D/MLL2 and KDM6A (Van Laarhoven et al 2015). Control LCLs from age and demography-matched subjects, were obtained from the Coriell Institute for Medical Research https://www.coriell.org/
RNAseq of human Kabuki Syndrome patients and comparison across taxa
Human LCLs from KMT2D and KDM6A patients and controls, as mentioned above, were grown to a count of 106 and pelleted. RNA was extracted by Zymo Research Direct-zolTM RNA MiniPrep kit (Catalog #R2050). Two cDNA libraries for each genotype (e.g. wildtype, KMT2D mutant, KDM6A mutant) were prepared from separate RNA MiniPreps using the Nugen mRNA kit for Illumina sequencing. Sequencing was perfomed on an Illumina NovaSEQ6000 (150bp paired-end reads) at the Genomics and Microarray Shared Resource Core Facility at University Of Colorado Denver Cancer Center. Libraries were sequenced to an average depth of 42 million reads. Reads were aligned to the human genome (hg38) using STAR Aligner (v2.7.3a) (Dobin et al., 2013), and gene counts were computed from STAR using quant mode (Dobin et al., 2013). Differential expression was performed in R using Deseq2 (Love, Huber, & Anders, 2014).
We compared our human RNAseq data to previously published datasets of KMT2D and KDM6A mutants from mice and zebrafish (Ang et al., 2016; Fahrner et al., 2019; Itoh et al., 2019; Lei & Jiao, 2018; Serrano et al., 2019; Shpargel et al., 2017). RNAseq datasets were downloaded from the NCBI Gene Expression Omnibus. Datasets included GSE129365 (Kmt2d mouse chondrocytes, (Fahrner et al., 2019)), GSE103849 (Kdm6a mouse neural crest cells, (Shpargel et al., 2017)), GSE121703 and GSE128615 (mouse Cd4 cells, (Itoh et al., 2019)), GSE110392 (mouse neural stem cells, (Lei & Jiao, 2018)). In addition an RNAseq dataset of whole embryo zebrafish Kmt2d mutants was provided by Serrano et al. (Serrano et al., 2019). Where necessary, downloaded RNAseq datasets were re-analyzed using DESeq2 for differential expression.
To identify potential core genes regulated by Kmt2d and Kdm6a we found the intersection set of genes differentially expressed at a nominal p-value of 0.05 in each dataset for all Kmt2d and Kdm6a RNAseq datasests respectively. To do this, we used BioMart as implemented in R to convert non-human gene symbols into orthologous human ensembl identifiers, while allowing multiple mappings to account for gene duplications. Intersection sets were computed from these human ensemble identifiers.
Statistical Analysis
Data shown are the means ± SEM from the number of samples or experiments indicated in the figure legends. All assays were repeated at least three times with independent samples. P values were determined with Student’s t tests using GraphPad Prism.
Results
Disruption of kmt2d and kdm6a/l in zebrafish causes cardiac defects
We disrupted kmt2d, kdm6a and kdm6al in zebrafish by CRISPR/Cas9 mutagenesis to understand their roles in cardiac development. Guide RNAs were designed to target a Protospacer Adjacent Motif (PAM) site in exon 6 of kmt2d, and exon 1 of kdm6a and kdm6al (Figure 1A-C). Mutations in ORF (open reading frame) sequences for kmt2d, kdm6a, and kdm6al are shown in Figure 1A-C. kmt2dCO1015, kdm6aCO1016, and kdm6alCO1017 are used in the study, hereafter referred to as kmt2d-/-, kdm6a-/-, and kdm6al-/- mutants, respectively. Although mutations in kmt2d alleles are located in a downstream exon, the resulting mutations are predicted to be loss-of-function alleles, similar to mutations in kdm6a and kdm6al. The mutations were further verified by reduced expression of the transcript for each allele in each of the corresponding mutants at 48 hpf, following in situ hybridization with probes made from cDNA for kmt2d, kdm6a and kdm6al. In wildtype embryos, kmt2d, kdm6a, and kdm6al are strongly expressed in the brain, eye and broadly in the rest of the body, including the heart consistent with previous reports (Van Laarhoven et al., 2015) (Figure 1D). The expression for kmt2d, kdm6a and kdm6al in the corresponding mutants is reduced, however, expression in the notochord appears to be non-specific. To examine the effect of each mutation on histone modifications, we probed for H3K4me3 and H3K27me3 by western blot in 48 hpf whole embryos as compared to total histone3 (H3) levels. Consistent with the role of Kmt2d in promoting H3K4me3 (Koutsioumpa et al., 2019), we observed a reduction in the overall levels of H3K4me3 in kmt2d-/- zebrafish at 48 hpf (Figure 1E). Likewise, mutation of kdm6a and its zebrafish paralog kdm6al, led to elevated H3K27me3 at 48 hpf as compared to total H3 (Figure 1E) (Hong et al., 2007).
To study the roles of kmt2d and kdm6a/kdm6l in heart development, we analyzed zebrafish heart formation in the cmlc2:GFP (coldheart) background at 48 hpf. cmlc2 is the cardiac myosin light chain 2 gene which marks the myocardium (Huang, Tu, Hsiao, Hsieh, & Tsai, 2003), and its expression results in a fluorescent heart. We used an antibody against sarcomeric myosin heavy chain (MF20), which is expressed in cardiomyocytes (George, Colombo, & Targoff, 2015) in the cmlc2:GFP background to analyze cardiac morphology. We observed altered cardiac morphology of mutants (Figure 1 F), with kmt2d and kdm6a mutants presenting with altered chamber morphology as compared to wildtype. In addition, all have a small but statistically significant reduction in heart rates (Figure 1G) but not in the total area of the heart (Supplemental Figure 1). Hence, as previously observed in KS patients and the previous animal studies, single mutations in kmt2d, kdm6a and kdm6al cause defects in cardiac development to varying levels.
Cardiac and neural crest cell gene expression is reduced in kmt2d and kdm6a/l mutant embryos
Development of the heart occurs earliest among all other functional organs when the cardiac crescent begins forming with contributions from cardiogenic progenitor cells post-gastrulation (Herrmann, Gross, Zhou, Kestler, & Kuhl, 2012). Subsequently, the cardiogenic mesoderm gives rise to the first heart field (FHF) and second heart field (SHF) (Herrmann et al., 2012; Moretti et al., 2006; L. Yang et al., 2008). The FHF contributes to the primary heart tube, left ventricle, and most of the atria, while SHF contributes to right ventricle, outflow tract and atria (Buckingham, Meilhac, & Zaffran, 2005; Herrmann et al., 2012; Laugwitz, Moretti, Caron, Nakano, & Chien, 2008). Orchestrated signals from mesoderm-derived FHF and SHF, as well as endoderm, are crucial for cardiogenesis (Herrmann et al., 2012). Hence, we tested for the expression of hand2 and mef2ca which are expressed in both FHF and SHF (Herrmann et al., 2012), in kmt2d, kdm6a and kdm6al mutants by in situ hybridization at 48 hpf. hand2a, or heart and neural crest derivatives expressed 2a, is a basic helix-loop-helix transcription factor that regulates cardiac morphogenesis, cardiomyocyte differentiation and cardiac-specific transcription and is expressed in the FHF and neural crest in the pharyngeal arches (McFadden et al., 2005; Schindler et al., 2014). hand2a belongs to an evolutionarily conserved cardiac transcription program which specifies cardiac progenitors in the anterior lateral plate mesoderm (ALPM), which migrate to the midline of the embryo for expression of cardiac sarcomere proteins, ultimately fusing to generate a beating heart tube (Lu, Langenbacher, & Chen, 2016). hand2a expression domains demarcate the heart forming region in ALPM (Lu et al., 2016; Schoenebeck, Keegan, & Yelon, 2007), and it specifies major domains which pattern pharyngeal arches of zebrafish dorsoventral axis (Talbot, Johnson, & Kimmel, 2010). In zebrafish heart primordium, hand2a expression domains appear anterior-lateral to hindbrain rhombomere 3 (Maves, Tyler, Moens, & Tapscott, 2009). Expression of hand2a are reduced in the kmt2d, kdm6a and kdm6al mutants (Figure 2). hand2a is normally expressed broadly in the heart is reduced especially in the midline in kmt2d and kdm6a mutants, but lesser extent in kdm6al-/-. A similar pattern of expression is observed for mef2ca, or myocyte enhancer factor 2ca, expressed primarily in the forming heart region, is a transcription factor that regulates myocardial differentiation and cardiac development (Hinits et al., 2012). In zebrafish hearts, mef2ca is expressed in the ALPM from 6 somite stage and persists during the fusion of the two heart fields into the primitive heart tube (Hinits et al., 2012). Expression of mef2ca, is similarly downregulated in the heart region of kmt2d, kdm6a and kdm6al mutants. Hence, kmt2d, kdm6a and kdm6al regulate genes involved in early events of heart development.
In addition to the cardiogenic mesoderm, cardiac development is also regulated by cardiac NCCs which contribute to outflow tract septation, aortic arch patterning, and semilunar valves (Brade, Pane, Moretti, Chien, & Laugwitz, 2013). Previous studies have indicated that kmt2d and kdm6a plays a role in regulating NCC formation (Schwenty-Lara, Nehl, & Borchers, 2019; Shpargel et al., 2017). To further investigate the roles of kmt2d and kdm6a we analyzed the expression of early neural crest-specific markers like sox10, foxd3, and crestin in zebrafish by in situ hybridization. sox10 regulates the formation of NCCs from the time of their emergence, and specifies several derivatives of NCCs (Aoki et al., 2003; Dutton et al., 2001; Southard-Smith, Kos, & Pavan, 1998). foxd3 is expressed in NCCs and maintains NCCs progenitors (Labosky & Kaestner, 1998; Teng, Mundell, Frist, Wang, & Labosky, 2008). crestin is characterized by pan-neural crest expression during zebrafish embryogenesis including premigratory and migrating NCCs (Luo, An, Arduini, & Henion, 2001; Rubinstein, Lee, Luo, Henion, & Halpern, 2000). In kmt2d, kdm6a and kdm6al mutant zebrafish, the expression of sox10 is slightly reduced in all three mutants at 12 hpf suggesting a reduction in NCC induction. The expression of foxd3 at 15 hpf is also reduced specifically in the anterior most population migrating to the heart in kmt2d and kdm6a but not in kdm6al mutants. The overall expression of crestin at 24 hpf is not markedly reduced in any of the mutants which indicates normal NCC migratory streams (Figure 3). These results show that kmt2d and kdm6a regulate early stages of NCCs development in zebrafish.
Comparative analysis reveal common targets of KMT2D and KDM6A
To translate our results into the context of human health, we investigated the transcriptome of 2 human Kabuki Syndrome patients carrying mutations in KMT2D and KDM6A, by RNA sequencing of their lymphoblastoid cells. The patient with KMT2D mutation has a transition c.4288T>C (p.C1430R) (Figure 4A) and the patient with KDM6A mutation has a microdeletion in chromosome X resulting in a deletion of KDM6A (Figure 4A Table). As controls, we used LCLs from age and ethnicity matched individuals. To ascertain differential gene expression, generated reads from RNA-seq were mapped to the human genome using STAR aligner (Dobin et al., 2013) and differential expression was performed in R with DESeq2. Differentially expressed genes (nominal p value <= 0.05) from our KMT2D human dataset were compared to RNA sequencing datasets for Kmt2d-mutant mice and zebrafish generated by other researchers (see methods). For mice, the RNA sequencing dataset is derived from embryonic hearts (Ang et al., 2016), and the zebrafish data are obtained from whole embryos (Serrano et al., 2019). Likewise, our KDM6A human RNA seq datasets were compared to three distinct RNA seq datasets of mice that are mutant for Kdm6a. The mice datasets are derived from NCCs (Shpargel et al., 2017), neural stem cells (Lei & Jiao, 2018), and CD4+ cells (Itoh et al., 2019). For KDM6A, the comparison is only between humans and mice, since zebrafish RNA sequencing for kdm6a mutation are not available.
A comparison between datasets from the 3 species revealed 76 common targets for KMT2D (Figure 4B). 30 of the 76 common targets are either upregulated or downregulated in a similar pattern across the 3 species, while 46 do not share a similarity in expression patterns among the 3 species, as shown by heat maps (Figure 4C). For KDM6A, 7 common targets were found between human and the 3 mice datasets (Figure 5A). The variation in the number of common targets between KMT2D and KDM6A datasets could partly be due to the differences in the roles and targets of KMT2D and KDM6A, as well as functional redundancy of other family members across species (Akerberg, Henner, Stewart, & Stankunas, 2017; Crump & Milne, 2019; Fellous, Earley, & Silvestre, 2019; Nottke, Colaiacovo, & Shi, 2009). Heat maps show that out of the 76 common targets between the 3 species, expression patterns of 49 genes are similar between mouse and human datasets, while 27 genes are not (Figure 5B). The dissimilarity is probably due to differences in cell type and age. Hence, like KMT2D, KDM6A also likely regulates gene expression through a tight temporal and spatial orchestration.
We selected common targets for both KMT2D and KDM6A datasets, based on their roles in cardiac and overall development as summarized in Tables 1 and 2, to validate whether these genes are similarly regulated by kmt2d or kdm6a during early development in zebrafish. The genes selected from analysis of KMT2D datasets are SOX12, SLC22A23, FOXP4, NR2F6, and SIX3. The genes selected from analysis of KDM6A datasets are ID3, PRDM1, and ANXA4. For validation, we performed qRT-PCR for sox12, slc22a23, foxp4, nr2f6, and six3 in 48 hpf kmt2d-/- mutant zebrafish. Likewise, we performed qRT-PCR for id3, prdm1a, and anxa4 in 48 hpf kdm6a-/- mutant zebrafish. Our results show decreased expression for sox12, foxp4, nr2f6, and six3 in the kmt2d-/- mutant (Figure 4C), and id3, prdm1a, and anxa4 in the kdm6a-/- mutant (Figure 5D).
Our qRT-PCR results for kmt2d-/- mutant zebrafish show a downregulation of sox12, foxp4, nr2f6, and six3, thus displaying greater consistency with the zebrafish RNA seq where these genes are downregulated (Figure 4C). Sox12, Foxp4, and Six3 are not downregulated, rather upregulated in varying degrees, in the mice RNA-seq (Figure 4C). SOX12, FOXP4, and NR2F6 are not downregulated, rather upregulated in varying degrees, in the human RNA-seq (Figure 4C). Expression of slc22a23 is not significantly changed in the qRT-PCR, though it is moderately upregulated in the zebrafish and mice RNA-seq, and highly in the human RNA seq (Figure 4C). The KMT2D and KDM6A datasets show variations in the expression patterns of same genes across different species and cell types which indicate a tight spatial and temporal orchestration of the roles of KMT2D and KDM6A in gene regulation.
The qRT-PCR results for the kdm6a-/- mutant zebrafish, show reduction in id3, prdm1a, and anxa4 (Figure 5C), which are consistent with RNA-seq datasets from mouse NCCs and neural stem cells (Figure 5B). Additionally, Prdm1 from mouse CD4+ RNA-seq, and ANXA4 from human RNA seq, are downregulated as in the qRT-PCR data. The upregulation of Id3 and Anxa4 in RNA seq of mouse CD4+ cells, and ID3 and PRDM1 in RNA seq of humans may be attributed to differences in source cell types used, and differences in species-specific temporal and spatial regulation of gene expression by KDM6A which is not fully understood. It is interesting to note the similarities and differences in transcriptomic profiles of similar types of cells i.e. human lymphoblastoid and murine CD4+, respectively.
Discussion
Cardiac development is orchestrated by progenitor cells from the cardiogenic mesoderm, proepicardium, and cardiac neural crest cells (Brade et al., 2013). The mesoderm contributes to cardiac crescent and linear heart tube, the proepicardium contributes to chamber maturation and coronary vessel formation, and the cardiac neural crest cells contribute to outflow tract septation, aortic arch patterning, semilunar valves, cardiomyocytes in amniotes, and heart regeneration in zebrafish (Brade et al., 2013; Jain et al., 2011; Tang, Martik, Li, & Bronner, 2019). Consequently, CHDs arise from disruption of cardiac progenitor cell development which is tightly orchestrated in a temporal and spatial manner by a complex network of genetic and epigenetic factors. Single cell-based high throughput assays further corroborate the transcriptional regulation of cardiogenesis (de Soysa et al., 2019). Epigenetic factors like KMT2D and KDM6A have been studied by several groups to identify how their mutations lead to heart defects. KMT2D and KDM6A are associated with crucial histone modifications – H3K4me3 and H3K27me3, respectively – which are linked to various aspects of cardiac development and disease.
Here we present, for the first time, a comparison between the transcriptome of zebrafish, mice, and humans with mutations in KMT2D, and a comparison between transcriptomes of humans and three distinct cell populations of mice with mutations in KDM6A. A subset of genes involved in cardiac development and physiology, vasculature, and overall development, appeared as common targets across all datasets for each mutation. Specifically, our analysis revealed 76 genes as common targets of KMT2D upon comparing our human datasets with the zebrafish and mouse datasets. 30 out of 76 genes show a similar expression pattern across the 3 species, while 46 genes do not. The dissimilar expression patterns of the 46 genes can partly be explained by the fact that the transcriptomes of the 3 species are derived from different cell populations derived during different stages of life. Hence, epigenetic regulation by KMT2D is tightly orchestrated spatially and temporally. On the other hand, a similar expression pattern in 3 species for the 30 genes by KMT2D indicates that its role is partly conserved. For KDM6A, we identified 7 common targets upon comparing human RNA-seq datasets with 3 mice datasets obtained from neural crest, neural stem, and CD4+ cells. Our in vivo analysis of kmt2d and kdm6a mutations in zebrafish corroborate the findings of previous studies investigating these two genes in cardiac development and disease (Ang et al., 2016; Lee et al., 2012; Schwenty-Lara, Nurnberger, et al., 2019; Serrano et al., 2019).
Serrano et al. have shown that kmt2d mutations in zebrafish cause defects in cardiovascular development, angiogenesis and aortic arch formation, as well as cardiac hypoplasticity (Serrano et al., 2019). Kmt2d associates with genes in cardiomyocytes, and myocardial deletion of Kmt2d reduces cardiac gene expression during murine heart development (Ang et al., 2016). In xenopus, morpholino-based knockdown of kmt2d causes cardiac hypoplasticity, reduced expression of early cardiac developmental genes like tbx20, isl1, nkx2.5 (Schwenty-Lara, Nurnberger, et al., 2019), and impaired formation and migration of NCCs by reducing the expression of early NCC-specific markers like pax3, foxd3, twist, and slug (Schwenty-Lara, Nehl, et al., 2019).
On the other hand, KDM6A or Utx promotes the differentiation of mouse embryonic stem cells (ESCs) into a cardiac lineage, and serves as a co-activator of core cardiac transcription factors like Srf, Nkx2.5, Tbx5, and Gata4 (Lee et al., 2012). KDM6A or Utx associates with cardiac-specific genes like Anf and Baf60c, to help recruit Brg1, an ATP-dependent chromatin remodeler that activates transcription of core cardiac transcription factors (Hang et al., 2010; Lee et al., 2012; Lickert et al., 2004). A neural crest-specific conditional deletion of KDM6A or Utx allele with a Wnt1-Cre transgene causes CHDs like patent ductus arteriosus (Shpargel et al., 2017).
While we observe a modest reduction in the early specification of NCCs (sox10 and foxd3, Figure 3) in zebrafish mutants of kmt2d and kdm6a, migration appears normal for the most part since we do not see drastic reduction of the number of cells in the developing heart (Figures 2 and 3). Likewise, our in vivo analysis of zebrafish shows a decrease in expression of NCCs-specific markers of early cardiac development like hand2a and mef2ca (Figure 2). Together, our study underscores several recent publications on the roles of KMT2D and KDM6A during early developmental stages of the heart and cardiac NCCs. Our cross-species transcriptomic analysis reveals conserved and unique gene expression patterns for KMT2D and KDM6A.
Our work on KMT2D and KDM6A can potentially contribute to future studies on cardiac regeneration, reprogramming, and disease where H3K4me3 and H3K27me3 are implicated (Ben-Yair et al., 2019; Delgado-Olguin et al., 2012; Engel & Ardehali, 2018; He et al., 2012; Ieda et al., 2010; Jonsson et al., 2016; Kaneda et al., 2009; Lee et al., 2012; C. F. Liu & Tang, 2019; L. Liu, I. Lei, H. Karatas, et al., 2016; L. Liu, Lei, & Wang, 2016; Z. Liu et al., 2016; Mathiyalagan, Keating, Du, & El-Osta, 2014; Monroe et al., 2019; Papait et al., 2013; Rosa-Garrido, Chapski, & Vondriska, 2018; Stein et al., 2011; Q. J. Zhang & Liu, 2015). In zebrafish, H3K27me3 has recently been shown to silence structural genes in proliferating cardiomyocytes during wound invasion and regeneration of injured hearts (Ben-Yair et al., 2019). ChIPsequencing studies on primary neonatal murine cardiomyocytes revealed that H3K4me3 is enriched, while H3K27me3 is reduced, at promoters of cardiac-specific genes like Mef2c, Gata4, and Tbx5 (Z. Liu et al., 2016). This study further showed that early re-patterning of H3K4me3 and H3K27me3 occurs during reprogramming of induced cardiomyocytes (iCMs), which are used for modeling cardiac diseases and regeneration (Z. Liu et al., 2016).
A study on the integration of ChIP- and ATAC-sequencing of adult and fetal human cardiac fibroblasts (HCFs) revealed that open chromatin or ATAC-seq peaks overlaid active promoters marked by H3K4me3 but not the repressive mark of H3K27me3 (Jonsson et al., 2016). However, both H3K4me3 and H3M27me3 were enriched around transcriptional start sites of their corresponding target genes (Jonsson et al., 2016). Moreover, ChIP-sequencing of adult murine cardiomyocytes followed by Gene Ontology (GO) analysis with the Genomic Regions Enrichment of Annotations Tool (GREAT) revealed that sites with differential distribution of H3K4me3 and H3K27me3 are associated with hypertrophic murine heart phenotypes, like thick ventricular wall, altered heart contraction, impaired skeletal muscle contractility, and abnormal cardiac stroke volume (Papait et al., 2013). During embryogenesis, enrichment of germ layerspecific H3K27me3 has been identified at sites of high DNA methylation in undifferentiated states (Gifford et al., 2013). Hence, the complexities of epigenetic modifications associated with KMT2D and KDM6A in cardiac biology are gradually being understood.
Mutations in KMT2D and KDM6A also contribute to complications during pregnancy (Rosenberg et al., 2020). Polyhydramnios or excessive amniotic fluid in the amniotic sac, reduced placental weights, and abnormal quadruple marker tests are reported from an analysis of 49 individuals with KS and their mothers (Rosenberg et al., 2020). Building on clinical data, (Adam et al., 2019; Adam, Hudgins, & Hannibal, 1993; Y. R. Wang, Xu, Wang, & Wang, 2019), interventions via genetic counselling and ongoing research on small molecule-based epigenetic inhibitors can provide potential diagnostic and preventive solutions for Kabuki Syndrome-based obstetric disorders.
Interestingly, missense KMT2D variants have also been correlated with a novel multiple malformations syndrome that is distinct from KS (Cuvertino et al., 2020), while KDM6A mutations are implicated in autoimmunity and drug resistance in leukemia (Itoh et al., 2019; Stief et al., 2020). Frequent immunopathological manifestations like immune thrombocytopenic purpura, vitiligo, etc. have also been reported from a registry study on 177 individuals with mutations in KMT2D or KDM6A (Margot et al., 2020).
To conclude, several features of the roles of KMT2D and KDM6A remain to be deciphered, especially other factors which cross-talk with H3K4me3/K27me3 and are linked to CHDs like KDM1A/LSD1, KDM4A, KDM4C, EZH2, and cilia (Agger et al., 2007; Ahmed & Streit, 2018; L. Chen et al., 2012; Lee et al., 2012; Nicholson et al., 2013; Rosales, Carulla, Garcia, Vargas, & Lizcano, 2016; Willaredt, Gorgas, Gardner, & Tucker, 2012; Wu et al., 2015; Y. Yang et al., 2019; Q. J. Zhang et al., 2011). FDA-approved drugs targeting inhibitors of LSD1 and EZH2 are used in treating cancers, which provide hope for expansion towards developing epigeneticsbased therapies against heart disease. Interestingly, KMT2D and KDM6A mutations are also implicated in cancers (Kantidakis et al., 2016; L. Wang & Shilatifard, 2019), and cardiotoxicity is reported as a side-effect in some but not all patients who receive cancer therapies which has evolved to a new field of study called cardio-oncology. Although at a nascent stage, the links between epigenetic factors in cardiac development and health, and cancers, can potentially address important questions in cardio-oncology. Taken together, a comprehensive understanding of epigenetic regulation of heart development and disease, in combination with large nextgeneration sequencing datasets from patients that are continuously being generated, combined with animal models can contribute to the emerging field of personalized medicine to translate epigenetics-based research from the bench to the bedside.
Conflicts of Interest
The authors declare no conflicts of interest
Acknowledgements
We acknowledge Maria de los Angeles Serrano and H. Joseph Yost for providing the complete dataset of kmt2d mutant zebrafish RNA sequencing as referenced in (Serrano et al., 2019). The study was supported by Summer 2018 Postdoctoral Fellowship to R. Sen from the American Heart Association (18POST33960371), funding from Dr. K.B. Artinger (R01DE024034). E.L is supported by a National Institutes of Health Ruth L. Kirschstein National Research Service Award T32 Postdoctoral Fellowship (T32CA17468).