Abnormalities of cardiac development result in congenital heart defects (CHDs), the most common form of birth defects (Kirby, 2017). Although outcomes for patients with CHDs are improving with advances in patient care, CHDs continue to be a leading cause of death in children (Bouma and Mulder, 2017; Hoffman and Kaplan, 2002; Mandalenakis et al., 2020; Su et al., 2022). Adult survivors of CHDs now outnumber children with CHDs and remain at risk for life-limiting cardiac complications (Liu et al., 2023). Deconstructing how disruptions of normal heart development lead to CHDs is necessary to define ontogenies, improve prenatal screening and prognosis, and inform translational approaches for heart repair. However, little is known about human heart development and CHDs in vivo, partly because of limited access to fetal tissue. Two recent preprints by the teams of Sarah Teichmann and Sanjay Sinha present herculean efforts to assemble and analyze a temporal, single-cell- and spatial-level compendium of human fetal hearts during the first and second trimesters (Bayraktar et al., 2024 preprint; Cranley et al., 2024 preprint). Indeed, these datasets represent a precious large-scale human fetal resource for the cardiovascular community. By leveraging several computational tools to analyze these complementary datasets, the authors bring insight into previously underrepresented human cardiac cell types, gene regulatory networks (GRNs) and cell lineage relationships within the normal fetal heart, as well as for diseased hearts from Trisomy 21.
To date, transcriptomics studies have revealed the cellular composition of the human heart by single-cell RNA sequencing (scRNA-seq) from a few thousand cells of the normal human fetal heart (Cui et al., 2019; Miao et al., 2020) or of a diseased fetal heart affected by autoimmune-mediated complete heart block (Suryawanshi et al., 2020). In combination with scRNA-seq, analyses of spatial transcriptomics and in situ sequencing of 69 gene transcripts have demonstrated spatial relationships in the human fetal heart (Asp et al., 2019). Recently, a larger-scale scRNA-seq analysis of more than 100,000 cells, along with MERFISH imaging of 238 genes, has highlighted underrepresented anatomic regions of human cardiac gene expression, multicellular ‘cell communities’, and functional cell-cell signaling pathways (Farah et al., 2024). In addition, a single-cell epigenomics atlas of the human fetal heart was derived from an analysis of more than 30,000 cells by single-cell assay for transposable-accessible chromatin using sequencing (scATAC-seq) (Ameen et al., 2022). Although many insights into the human fetal heart were gained from these studies, a limited number of fetal hearts were evaluated in many cases, some disease-relevant or rare cell types were missing, or only one modality was evaluated.
A cell atlas of the human heart and great vessels during development
Bayraktar and colleagues assembled one of the largest specimen collections to date, with 21 human fetal heart samples from 4-20 post-conception weeks (PCW), and generated a comprehensive transcriptomic cell atlas of normal human heart and great vessel morphogenesis (Bayraktar et al., 2024 preprint). By analyzing transcriptomes of nearly 300,000 cells or nuclei, they deduced pseudotime trajectories, inferred cell-cell communications and integrated spatial gene expression (albeit lacking single-cell resolution) from an accompanying report (Cranley et al., 2024 preprint). Importantly, these analyses discerned several markers that distinguish cardiomyocytes by location as well as function (e.g. conduction system in the sinoatrial node, atrioventricular node and His-Purkinje fibers). The study also uncovered differences between coronary vessels and great vessels, and represents the most in-depth characterization so far of several poorly annotated human fetal cardiac cell types, including the pericardium, endocardium, the ductus arteriosus, neural cells (e.g. sympathetic and parasympathetic neurons), leukocytes (e.g. macrophages and monocytes) and lymphatic endothelial cells. Undoubtedly, this single cell-gene expression data addresses an unmet need, by providing an invaluable reference dataset that captures many previously missed cell types relevant for a range of CHDs.
A multiomics human cell atlas of the developing heart
In a complementary study, Cranley and colleagues used scRNA-seq and single-nucleus RNA sequencing (snRNA-seq) from Bayraktar et al. (2024 preprint) and sampled an expansive number of nuclei (>150,000) using snATAC-seq, which enabled simultaneous evaluation of gene expression and chromatin accessibility in the same cells. In addition, spatial gene expression data was derived from multiple tissue planes of several fetal hearts (25 sections in total from six hearts), which provided considerable tissue coverage of the developing fetal heart (Cranley et al., 2024 preprint).
Using SCENIC+ (González-Blas et al., 2023) with fetal heart multiomics data, the authors deduced enhancer-mediated GRNs modulated by transcription factors, termed regulons, specific to human fetal heart cells. In one example, genes2genes (Sumanaweera et al., 2023 preprint) was used to compare regulon activities between trajectories of human right atrial cardiomyocyte and left atrial cardiomyocyte development. The authors identified several regulons enriched in the left atrial cardiomyocyte trajectory, including PITX2. Pitx2 is essential for left atrial development in mice (Liu et al., 2001) and associated with atrial fibrillation (Gudbjartsson et al., 2007; Steimle et al., 2022), which is the most common arrhythmia in adults (Colilla et al., 2013; Brundel et al., 2022), including CHD survivors (Mandalenakis et al., 2018; Teuwen et al., 2018). Interestingly, ETS2 and MAF regulons were predicted to promote right atrial cardiomyocyte identity, and to suppress left atrial cardiomyocyte identity. As few transcription factors have been implicated in left-right atrial identity beyond PITX2, it is intriguing to speculate that ETS2 and MAF may regulate a developmental bifurcation for left-right atrial identity, potentially in concert or opposition to PITX2.
Origins of some human cardiac cell types
The authors leveraged SComatic (Muyas et al., 2024), a tool originally developed to identify somatic mutations in cancer, to carry out a cell lineage analysis of developing cardiomyocytes. They adapted SComatic to use sequencing reads from snRNA-seq and snATAC-seq of the large number of cells from each fetal heart specimen. Remarkably, a cell lineage tree from an aggregate of fetal heart samples predicted that pacemaker cardiomyocytes represent a distinctive cell lineage from working cardiomyocytes, before segregation of atrial and ventricular cardiomyocytes. In mice, a Tbx18+ progenitor population contributes to the morphology of the sinoatrial node (Mommersteeg et al., 2010; Wiese et al., 2009). Whether the human pacemaker cell lineage uncovered by Cranley and colleagues corresponds to the mouse Tbx18+ lineage or represents a different progenitor population remains to be established. Given that sinus node dysfunction can occur in patients with CHDs after heart surgery (Greenwood et al., 1975) or with aging (Jensen et al., 2014), such cell lineage information from the human fetal heart could inform efforts to develop reprogramming approaches to transform the working myocardium into pacemaker tissue (Greulich et al., 2016; Hu et al., 2014; Kapoor et al., 2013). As researchers collect many more cells for snRNA-seq and snATAC-seq from donations of fetal heart samples in the future, I expect that this bioinformatics approach will identify other lineage relationships in the developing heart and, possibly, aberrant cell lineages in the context of some CHDs.
TissueTypist and Down syndrome
One challenge was inferring multicellular niches in fetal heart tissue from spot-based spatial transcriptomics. To overcome this hurdle, the authors used cell2location (Kleshchevnikov et al., 2022) to deconvolute per-spot spatial data using cardiac cell type signatures from the accompanying snRNA-seq atlas. By doing so, the authors annotated 19 fetal heart tissue structures and developed a classifier named ‘TissueTypist’, which is based on a logistic regression CellTypist model (Conde et al., 2022). A second challenge was to compare, as consistently as one can, the coordinates of tissue layers among samples of variable size. For example, to collate the relative locations of the trabecular layer and the compact layer of the ventricular myocardium across time points, the authors defined a transmural axis across the ventricular wall using OrganAxis (Yayon et al., 2023 preprint), which provided a spatial framework to assess attributes such as gene expression or cell density from histology.
The authors then applied TissueTypist with OrganAxis to data of two hearts from fetuses at 13-14 PCW from Trisomy 21 (Down syndrome), the most common chromosomal abnormality and a condition in which half of those with the syndrome have some form of CHD (Dimopoulos et al., 2023). CHDs from Trisomy 21 most commonly include atrial septal defects, ventricular septal defects or atrioventricular canal defects, many of which require surgical repair. In this analysis of Trisomy 21 specimens, anomalies of cardiac septation or endocardial cushion development were not noted. Per the preprint authors, information regarding structural defects were not available beyond the histology that underwent spatial transcriptomics. Consequently, it is hard to know whether these specimens displayed any CHDs. Instead, TissueTypist revealed a reduction of compact layer cardiomyocytes, coronary endothelial cells and pericytes. In addition, abnormal cell-cell interactions between left ventricular compact myocardium and other cell types were predicted, along with aberrant gene regulatory networks in compact myocardium. Interestingly, a thinner compact layer in the right ventricle might be present in at least two mouse disease models of Trisomy 21 (Kazuki et al., 2020; Lana-Elola et al., 2024). However, only limited tissue planes are shown in each of these studies, and this phenotype is less apparent in a related mouse study (Lana-Elola et al., 2016). It is, therefore, possible that a thin compact layer might be a feature of Trisomy 21 that has so far not been described and deserves further study. Regardless, the datasets from these two Trisomy 21 fetal hearts are a remarkable genomics resource that I anticipate will facilitate our understanding of what underlies this common genetic cause of CHDs.
Together, these preprints present a tremendous resource for understanding the building blocks of human heart development. Importantly, the datasets will provide crucial benchmarks for evaluating mouse models and human cell-based models of disease, and for analyzing human specimens of CHDs, the majority of which have unknown etiologies.
Acknowledgements
The author is grateful to Swetansu Hota, members of the Kathiriya laboratory and Francoise Chanut for comments and feedback, as well as preprint authors for fact-checking.
Footnotes
Funding
The author is funded by the Saving Tiny Hearts Society, University of California, San Francisco Pediatric Heart Center and University of California, San Francisco Anesthesia Research Support.
References
Competing interests
The authors declare no competing or financial interests.
