The scale of the underlying dataset is formidable. After quality control, the team retained nearly 300,000 cells and nuclei from 26 euploid fetal heart samples, spanning single-cell RNA sequencing, paired single-nucleus RNA and ATAC sequencing, and spatial transcriptomic platforms including Visium standard definition at 55-micrometre resolution, Visium HD at 2 micrometres, and Xenium subcellular imaging. Chromatin accessibility data from more than 167,000 nuclei yielded over 508,000 open chromatin peaks, two-thirds of which overlapped known regulatory elements catalogued by the ENCODE project. A statistical comparison of the transcriptomic and chromatin landscapes, using a Mantel test on cell-type distance matrices, showed strong concordance, confirming that gene expression and the underlying regulatory architecture tell a coherent story about how cardiac cell identities are specified.

From these data the researchers delineated 21 distinct tissue niches — multicellular neighbourhoods in which specific combinations of cell types, signalling molecules and extracellular environments cooperate to build and refine cardiac structures. These ranged from the sinoatrial node, the heart’s natural pacemaker, to valve-forming regions, the ductus arteriosus with its distinctive smooth muscle population, and the layered ventricular wall where compact and trabeculated cardiomyocytes occupy separate territories. Crucially, the computationally defined niches matched expert histological annotations of paired tissue sections with high agreement, giving the atlas a firm anchor in physical anatomy rather than purely abstract clustering.

To make this resource usable by others, the team built a machine-learning tool called TissueTypist, a hierarchical logistic regression classifier that assigns niche labels to new spatial transcriptomic data at almost any resolution. The model was trained not only on each spatial unit’s own gene expression but also on the expression of its neighbours and its distance from the tissue edge, reflecting the intuition that identity in a tissue depends on context. In leave-one-section-out cross-validation across fourteen sections from three Visium platforms, TissueTypist achieved weighted F1 scores of 0.88 to 0.98 at the coarse anatomical level. When applied to independent MERFISH and Xenium datasets it recovered major cardiac compartments and even subdivided regions that earlier annotations had merged, resolving the sinus horn and sinoatrial node head as separate structures supported by marker genes such as TBX3, BMP2, HCN4 and SHOX2. The framework also performed well on a human lung dataset, suggesting it may generalise beyond the heart.

One of the most striking biological discoveries concerns the developing pacemaker. High-resolution spatial profiling resolved three subtypes of pacemaker cell — those of the sinus horn, the sinoatrial node head and the sinoatrial node tail — arrayed along a linear axis through the sinoatrial region. Parasympathetic neurons, which slow the heartbeat, were found clustered preferentially around the sinus horn and node head populations and largely absent from the tail. Gene regulatory network analysis identified the canonical pacemaker transcription factors ISL1, SHOX2, TBX3, TBX5 and TBX18 as hubs controlling not only pacemaker identity and ion channels but also a suite of axon-guidance ligands — SEMA3A, SLIT2, SLIT3, EFNB2 and CXCL12 — that are strongly expressed in the first trimester and then downregulated. In other words, pacemaker cells appear to actively recruit their own nerve supply, broadcasting molecular beacons before the neurons arrive.

Cell-to-cell interaction modelling reinforced this picture. Using CellPhoneDB on spatially constrained niches, the researchers found significantly enriched semaphorin, ephrin and SLIT–ROBO signalling between head-region pacemaker cells and parasympathetic neurons, far exceeding the interactions seen between ordinary working atrial cardiomyocytes and the same neurons. The team also spotted CX3CR1-positive macrophages with a microglia-like molecular profile accumulating in the sinoatrial node, apparently drawn there by CX3CL1 — the ligand for CX3CR1 — which was expressed specifically by parasympathetic neurons, mirroring the mechanism by which microglia home to neurons in the brain. At early stages, around five to six weeks, neuronal signals were sparse despite robust axon-guidance gene expression in pacemaker cells, suggesting the pacemaker first builds a permissive molecular environment and innervation follows later.

The atlas also tracks how working heart muscle matures. Left and right atrial cardiomyocytes, initially transcriptionally similar, diverge progressively through development, with the left population marked by PITX2 — a gene adjacent to the well-known 4q25 atrial fibrillation risk locus — and the right population upregulating axon-guidance and innervation genes such as NTM, SLIT2, SLIT3 and SEMA5A, consistent with the denser autonomic innervation the right atrium acquires after birth. In the ventricle, the team defined a transmural coordinate axis running from the outer epicardium to the inner endocardium using the OrganAxis method, revealing that compact cardiomyocytes expand inward through the second trimester in tandem with coronary capillary endothelial cells and pericytes. Compact cardiomyocytes were inferred to promote their own vascularisation by secreting VEGF, and their maturation involved a pronounced lipid-metabolic switch, with genes such as CD36, OSBPL2 and SREBF2 pointing toward fatty acid uptake and sterol synthesis. Experimental induction of the transcription factor PRDM16 in stem-cell-derived cardiomyocytes produced a trend toward higher maturation scores and activated mitochondrial and redox genes, supporting its role as a driver of metabolic maturation.

The comparative arm of the study delivers its most clinically consequential finding. Analysing trisomy 21 heart samples from eleven to fourteen weeks of gestation alongside age-matched euploid controls, the researchers found that hearts with the extra chromosome 21 were depleted of compact ventricular cardiomyocytes and showed reduced activity of maturation and compaction gene programmes. Differential abundance testing with Milo revealed substantial transcriptional divergence in cardiomyocytes, and pathway analysis of trisomy-enriched neighbourhoods showed disruption of myogenesis, hypoxia response and oxidative phosphorylation, together with upregulation of apoptosis and senescence signatures, particularly in cycling cardiomyocytes. Signalling rewiring was broad: cardiomyocytes sent less VEGF, received less IGF and VEGF signalling, and showed a marked perturbation of the semaphorin–plexin axis, with SEMA3C upregulated across cell types while the receptor PLXNA4 was downregulated and PLXNA2 upregulated in cardiomyocytes — a receptor–ligand mismatch that may blunt the semaphorin responsiveness previously shown to be required for myocardial compaction.

Converging lines of evidence strengthened the case that these molecular changes have structural consequences. In the Dp1Tyb mouse model of Down syndrome, high-resolution episcopic microscopy at embryonic day 15.5 revealed a thinner compact myocardial layer than in wild-type littermates, and the reduction was even greater in embryos that also had congenital heart defects. Mouse single-cell data mirrored the human findings, with shared enrichment of impaired oxidative phosphorylation and heightened apoptosis. Most tellingly, when trisomy 21 induced pluripotent stem cells and their isogenic euploid counterparts were differentiated into cardiomyocytes in the laboratory, TUNEL staining showed a significantly higher proportion of dying cells in the trisomic cultures — an odds ratio of 1.94 — indicating an intrinsic, cell-autonomous susceptibility to apoptosis. Gene regulatory network analysis pointed to candidate upstream regulators of this programme, including NRF1, FOXO3 and SMAD5, none of which are encoded on chromosome 21, implying that the apoptotic shift is a downstream consequence of altered gene dosage rather than direct dosage of the regulators themselves.

The authors are careful about interpretation. The trisomy 21 samples were taken at a stage when septal defects cannot yet be confidently assessed in two-dimensional sections, so the observed changes are best understood as early perturbations that plausibly raise the risk of congenital heart disease — which affects roughly 40 to 50 percent of live-born children with Down syndrome — rather than as proof of causation for specific malformations. Even so, the study delivers a foundational resource: a spatially resolved, multimodal framework of human cardiogenesis, a generalisable niche-annotation tool, and a mechanistic hypothesis linking chromosome 21 dosage, impaired cardiomyocyte survival and failed ventricular compaction. If increased apoptosis proves to be a targetable process, the window in which it acts — the first and early second trimester — may define when future interventions would need to operate, and the atlas itself provides the benchmark against which any such therapeutic ideas will be tested.

Subject of Research: Spatial and single-cell multiomic mapping of human heart development and its disruption in trisomy 21

Article Title: Dynamics of human cardiogenesis and its disruption in trisomy 21

Article References: Cranley, J., Kanemaru, K., Bayraktar, S., Knight-Schrijver, V., Hulbert, R., Lana-Elola, E., Aoidi, R., Pett, J. P., Wilbrey-Clark, A., Polanski, K., Dabrowska, M., Mulas, I., Johnson, H., Combemorel, N., Yu, Y., Palmer, J. A., Lee, W., Van Wauwe, J., Ng-Blichfeldt, J.-P., … Teichmann, S. A. (2026). Dynamics of human cardiogenesis and its disruption in trisomy 21. Nature. https://doi.org/10.1038/s41586-026-11125-y

Image Credits: AI Generated

DOI: 10.1038/s41586-026-11125-y

Keywords: cardiogenesis, single-cell transcriptomics, spatial transcriptomics, sinoatrial node, pacemaker cells, trisomy 21, Down syndrome, congenital heart disease, cardiomyocyte maturation, apoptosis, gene regulatory networks, TissueTypist