Decoding Cardiac Development and Maturation at Single-Cell and Spatial Transcriptomic Resolution

Abstract

Recent advances in single-cell transcriptomics have revolutionized our understanding of cardiac development and maturation by resolving cellular heterogeneity, delineating lineage trajectories, and uncovering gene regulatory networks and intercellular signaling at unprecedented resolution. The heart develops through a tightly coordinated spatiotemporal process that extends from early organogenesis through postnatal maturation. Despite major progress, key questions remain unresolved, including the localization and function of rare progenitor populations and the mechanisms that guide cardiomyocyte maturation and loss of regenerative capacity. Spatial transcriptomics has emerged as a powerful complement to single-cell profiling because it preserves the native tissue architecture and reveals how gene expression is organized within anatomical context. While many spatial studies to date have focused on cardiac disease and injury, emerging developmental datasets spanning embryonic to postnatal stages now enable reconstruction of spatially resolved trajectories of heart formation. Here, we review key findings and limitations from recent single-cell and spatial transcriptomic studies of heart development and maturation and discuss how integrative approaches and advanced computational tools are redefining the molecular and spatial logic of cardiogenesis. These insights are expected to greatly accelerate future regenerative and translational research.

1. INTRODUCTION

For several decades, extensive efforts have been devoted to understanding the factors and gene regulatory networks that influence development of the heart, the first functional organ to form in life. The process is highly complex and dynamic,¹ continuing through fetal development and into postnatal maturation. Heterogeneous progenitor cells that originate from diverse embryonic sources migrate, interact, and differentiate to form the structural and functional components of the heart. As development progresses, distinct cell populations organize into specialized microenvironments, where local interactions guide maturation and confer region-specific physiological properties.

Until the mid-2010s, heart researchers relied primarily on microarray and qPCR analysis, along with low-throughput in situ hybridization, to unravel these events at the transcriptomic level. The emergence of single-cell RNA sequencing² (scRNA-seq) has transformed our understanding of cardiac development by enabling high-throughput gene expression profiling at single-cell resolution. Combined with advanced computational tools, this technology has revealed remarkable cellular diversity within the developing and mature heart, allowing researchers to refine cell type classifications, uncover rare or transitional populations, and infer regulatory networks and lineage trajectories (Figure 1). However, it lacks spatial context. Although scRNA-seq captures cellular heterogeneity, the tissue dissociation required for single-cell profiling erases information about each cell’s physical location, spatial relationships, and local signaling environment. Because morphogenesis is inherently a spatial process shaped by morphogen gradients and cell–cell interactions, understanding how gene expression is organized within anatomical context is essential to fully elucidate how a heart is built.

Figure 1.

Comparison of major single-cell technologies. scRNA-seq, single-cell RNA sequencing; snRNA-seq, single-nuclei RNA sequencing. Created with BioRender.com

Newly developed spatial transcriptomics (ST) technologies now offer a means to bridge the gap between gene identification and cell localization.³ ST preserves the architecture of the tissue while capturing transcriptomic information, enabling researchers to visualize gene expression in situ and reconstruct cell type organization in cardiac structures. Integrating scRNA-seq with ST datasets has enabled anchoring of high-resolution cell type annotations to precise spatial coordinates, thereby revealing how transcriptional programs map onto developmental patterning and functional zones.

In this review, we highlight recent progress in applying single-cell transcriptomics and ST to study heart development, from early mesodermal specification and cardiac progenitor patterning to chamber morphogenesis and postnatal maturation. We discuss the major biological insights emerging from these datasets and the challenges and limitations inherent to profiling a structurally complex and dynamic organ. Finally, we outline future opportunities for using multi-omics spatial data and advancing computational approaches to construct a comprehensive, spatiotemporal blueprint of heart development and maturation.

2. SINGLE-CELL TRANSCRIPTOMICS IN HEART DEVELOPMENT AND MATURATION

2.1. Single-Cell RNA Sequencing Technologies

Rapid evolution of scRNA-seq technologies has enabled increasingly precise dissection of cellular heterogeneity across cardiovascular tissues. The most common approaches can be categorized into plate-based, droplet-based, and split-pool combinatorial indexing platforms, each offering distinct advantages in sensitivity and throughput.

Plate-based methods isolate one cell and one barcode per well, providing high sensitivity but limited scalability. Droplet-based methods such as Drop-seq^4,5 and 10x Genomics Chromium⁶ encapsulate cells in nanoliter sized droplets with microbeads, enabling high-throughput profiling across large cell numbers. However, these systems are restricted by microfluidic channel and droplet size. Therefore, their compatibility with large, fragile cell types such as mature cardiomyocytes is limited. The development of split-pool methods such as split-seq,⁷ sciRNA-seq,⁸ and Parse Evercode⁹ further increased the capacity of scRNA-seq. These methods use combinatorial barcoding by mixing and distributing cells repeatedly to ligate new barcodes. Split-pool methods are compatible with large cell types and can be more cost effective when preparing a library for more than 100,000 cells. Libraries generated by these platforms then can be sequenced by synthesis from Illumina¹⁰ or nanopore long-read sequencing from PacBio¹¹ and Oxford Nanopore.¹² More detailed comparisons of each single-cell platforms are discussed elsewhere.^13,14

2.2. Use of scRNA-seq in Heart Development and Cardiomyocyte Maturation

Formation of the heart from a simple linear heart tube to a four-chambered organ with coordinated electrical and mechanical function requires precisely orchestrated morphogenetic movements and signaling cues that are spatially restricted and temporally regulated. Numerous reviews have detailed this dynamic process from different angles.^1,15–18 Now, single-cell transcriptomics combined with computational advancements have revolutionized our understanding of heart development by both validating established lineage relationships, and revealing previously unrecognized rare populations, transitional cell states, and complex differentiation hierarchies. Yet, understanding how these cell types collectively assemble into a functional heart requires contextualizing them within their tissue microenvironment. Spatial information that reveals where cell types reside, how morphogen gradients are distributed, and how neighboring cells communicate is critical for linking transcriptional identity to developmental function (Figure 1). Several recent articles have comprehensively reviewed scRNA-seq studies related to cardiac development.^14,19–22 Notably, most early developmental single-cell studies employ standard dissociation and sequencing workflows, as embryonic cardiac cells are readily compatible with conventional platforms. In contrast, profiling adult cardiomyocytes presents substantial technical challenges due to their large size, structural fragility, and multinucleated nature, necessitating specialized isolation and sequencing strategies, which are discussed in the Challenges and Limitations section.

Here, we provide a concise overview of the major cellular lineages and signaling pathways that drive murine heart formation, while highlighting where ST can add the anatomical context required to interpret these transcriptional programs in vivo. In particular, we emphasize recent progress in resolving postnatal maturation, a stage that has been technically challenging to profile but is now being illuminated by emerging single-cell technologies and advanced computational frameworks.

2.2.1. Cardiac specification and heart field patterning

Heart development begins remarkably early in the mammalian embryo, emerging from the mesodermal cells specified during gastrulation. By embryonic day (E) 6.5–7.0 in mice, precardiac mesodermal cells exiting the primitive streak migrate into the anterior lateral plate mesoderm, where Wnt, BMP, FGF, and Nodal signaling gradients establish positional cues that distinguish cardiac from noncardiac mesodermal fates.²³ The transient expression of Mesp1 marks the committed cardiovascular progenitors, initiating epithelial-to-mesenchymal transition (EMT) and anterior migration.^24,25 Single-cell transcriptomics has further confirmed that the Mesp1-lineage diversifies into the major myocardial, endocardial, and epicardial lineages of the heart.^26,27

As these cells migrate, they organize into distinct but continuous progenitor fields.^15,16 The first heart field (FHF) emerges and forms the cardiac crescent by E7.5 and rapidly differentiates into early cardiomyocytes that contribute primarily to the left ventricle (LV) and portions of the atria. Dorsal to the FHF, the second heart field (SHF) remains undifferentiated and proliferative longer, contributing to the right ventricle (RV), outflow tract (OFT), inflow tract, and remaining atrial myocardium. Recently, scRNA-seq analysis identified a juxta-cardiac field (JCF) adjacent to the FHF that produces both myocardial and epicardial lineages, adding new complexity to early heart progenitor topography.²⁸ Each heart field exhibits spatiotemporally dynamic expression of distinct and shared molecular markers, such as Hcn4 and Tbx5 in the FHF; Tbx1, Tbx5, and Isl1 in the SHF; and Mab21l2 and Hand1, which mark the JCF.^17,28

By E8.0, the bilateral FHF merges at the midline to form a linear heart tube, suspended from the pericardial wall. Meanwhile, SHF progenitors remain within the pharyngeal mesoderm, organized into anterior and posterior SHF domains that connect to the outflow and inflow poles of the growing tube.²⁹ Retinoic acid (RA) signaling contributes to the spatial segregation of these transcriptionally distinct domains by suppressing anterior Tbx1 and activating posterior Tbx5 expression.³⁰

Building on extensive clonal studies, single-cell transcriptomics has been instrumental in resolving lineage contributions and fate decisions within these heterogeneous populations. A combined scRNA-seq and single-cell assay for transposase-accessible chromatin with sequencing (scATAC-seq) study of Isl1⁺ and Nkx2.5⁺ progenitors confirmed heterogeneity within these two populations and highlighted Hox transcription factors as key regulators of cardiomyocyte differentiation.³¹ Another recent multi-omic study identified a multilineage-primed transitional progenitor population and demonstrated that Tbx1 governs its chromatin accessibility landscape.³² More recently, Wen et al.³³ combined lineage tracing and scRNA-seq to resolve lineage contribution of under-characterized Wnt2⁺ cardiopulmonary progenitor cells to both heart and lung development. A comprehensive review by Robert Kelly²⁰ summarizes how scRNA-seq has refined our understanding of the transcriptional heterogeneity and lineage trajectories of heart field specifications.

Morphogenic signals, including Wnt, BMP, Notch, FGF, RA, and Hedgehog, from neighboring tissues such as the foregut endoderm and pharyngeal endoderm and found between cardiac fields themselves guide the spatial segregation of these progenitor pools and instruct their specification, differentiation, and migration.^16,23,34 Computationally inferred information from single-cell studies are beginning to help resolve these interactions with finer granularity. For example, our recent work utilizing trajectory analysis demonstrated that canonical Wnt ligands originating from FHF cardiac progenitors are required for anterior SHF expansion and proper formation of the RV.³⁵ Such cross-field communication underscores how localized signaling environments shape lineage outcomes and structural asymmetry within the early heart.

2.2.2. Morphogenesis: from a linear heart tube to a four-chambered heart

From E8.0 to E10.5, continued addition of SHF progenitors to both poles drives elongation of the heart tube, while asymmetric proliferation, ciliary signaling, and cytoskeletal remodeling mediate rightward looping, establishing the spatial foundation for chamber alignment and left-right asymmetry.³⁶ Beginning around E9.0, localized expansion of the myocardium initiates ballooning morphogenesis, which generates the primitive atrial and ventricular chambers. Concurrently, trabeculation generates luminal myocardial ridges that enhance oxygen diffusion and contractile performance prior to coronary circulation. The myocardium subsequently undergoes compaction from E14.5 until birth to form the mature ventricular wall.³⁷ These processes depend on reciprocal endocardial–myocardial signaling, including NOTCH1, NRG1–ERBB2/ERBB4, BMP10, and Wnt/planar cell polarity pathways.^38–41 Spatially resolved methods are critical for understanding these localized interactions and patterning of the developing myocardium, as demonstrated by Wu et al.⁴² who conducted the first study integrating scRNA-seq and ST in mouse embryonic heart. They reported that Prdm16 activates compact genes and suppresses trabecular genes in the LV to maintain region-specific identity. More recently, coupling single-cell transcriptomics with biomechanical modeling uncovered tension-dependent cardiomyocyte heterogeneity, highlighting interaction between mechanical forces and transcriptional regulation during trabeculation.⁴³

In parallel with chamber formation, atrioventricular canal (AVC) and OFT endocardial cushions form through endocardial-to-mesenchymal transition to initiate valve and septal formation. The muscular septum grows toward the AVC and fuses with the cushions at around E13.5, physically dividing the chambers. This process requires coordinated contributions from multiple lineages, including endocardium, SHF, epicardium, and cardiac neural crest cells, and is tightly regulated by signaling pathways such as Hedgehog, BMP, Wnt, and Notch.^17,44–46 A particularly intriguing but understudied structure during this process is the dorsal mesenchymal protrusion (DMP), a transient septal component that contributes to atrioventricular (AV) septation. Lineage tracing studies have established its origin from the SHF,^47–49 yet the absence of well-defined molecular markers has made it difficult to resolve DMP-specific progenitors within scRNA-seq datasets to dissect their developmental trajectories and gene regulatory programs. Because its annotation depends primarily on anatomic localization relative to adjacent tissues, ST presents a powerful opportunity to identify molecular signatures, reconstruct the lineage relationships, and elucidate the signaling environment underlying DMP formation. Recent single-cell profiling of other non-myocyte populations within the cushion endocardium and mesenchyme has begun to reveal transitional cell states and gene regulatory drivers of endocardial-to-mesenchymal transition and valve morphogenesis.^50,51 Integrating these high-resolution molecular datasets with spatial context will be critical for fully elucidating how the AVC and OFT integrate multiple progenitor sources to generate the mature septa and valves of the four-chambered heart.

The cardiac conduction system (CCS) emerges from specialized myocardial domains in the inflow tract and AVC, where transcriptional regulators such as Tbx3, Tbx18, and Shox2 suppress the working myocardial program and promote pacemaker identity.^52–55 As looping progresses, discrete CCS domains, including the sinoatrial node, atrioventricular node, His bundle, bundle branches, and Purkinje fibers, emerge through spatially restricted signaling cues from neighboring fibroblasts, endocardial cells, and autonomic innervation.^56,57 Because CCS subpopulations are sparse, deeply embedded, and lack definitive surface markers, they are difficult to isolate and profile with conventional scRNA-seq. Goodyer et al.⁵⁷ addressed this challenge by combining microdissection of conduction regions with scRNA-seq, providing a comprehensive transcriptional map of developing CCS cell types. More recently, Ren et al.⁵⁸ expanded profiling into the postnatal window, uncovering further insights into CCS subtype maturation.

The heart’s critical functions are precisely modulated by its nervous system, which consists of sympathetic, parasympathetic, and sensory nerves primarily derived from neural crest cells (NCCs), with an additional non-NCC source for cholinergic fibers.^59,60 Trunk NCCs form sympathetic ganglia, while cranial NCCs contribute to parasympathetic nerves.⁶¹ Migrating neurons typically reach the heart in mice at around E13.5, and cardiac intricate innervation patterning is meticulously controlled and initiated by a balance of neurotrophic factors such as NGF and BMP, which guide axon growth, induce differentiation, and trigger stop-growth signals upon cardiomyocyte contact.^62,63 Moreover, signaling through Trk/p75 receptors, repulsive cues like SEMA3A, and specific cardiac fibroblast populations contribute to this local microenvironment by influencing sympathetic growth, axon bundling along vessels, and synaptic maturation.^64–67 Collectively, early studies revealed that cardiac innervation is remarkably heterogeneous. Its variations in cell types, spatial patterning, and developmental dynamics profoundly impact cardiac physiology and disease.

Recent high-resolution single-cell profiling has led to the discovery of previously unappreciated intrinsic cardiac nervous system cell types crucial for cardiac regulation. Notably, scRNA-seq analyses identified a novel neuroendocrine chromaffin cell population within the developing human heart, hypothesized to mediate sympathetic modulation in response to hypoxia.⁶⁸ Concurrently, unique cardiac nexus glia, neural crest–derived cells predominantly found in the OFT, have been characterized as critical regulators of cardiac innervation and autonomic tone.⁶⁹ Furthermore, an intrinsic catecholaminergic system within the murine heart has been proposed based on the identification of dopamine beta-hydroxylase–expressing cardiomyocytes localized near CCS cells.⁷⁰ Also, studies that have traced NCC derivatives in mice indicate an earlier commitment to neuronal lineage than previously recognized, with neurons and Schwann cells present in the heart from E10.5.⁷¹ Beyond novel cell identification, single-cell transcriptomics of rat and human ganglionated plexi revealed diverse neuronal subtypes.^72,73 Striking evidence suggests co-expression of cholinergic and catecholaminergic markers in individual neurons, a stark contrast to the central nervous system.⁷⁴ Glial cells, including Schwann cells and various progenitors, are consistently implicated in supporting other neuronal subtypes^73,75 and exhibit regional specificity and functional diversity.^69,72 This body of work collectively established a comprehensive, holistic view of the intrinsic cardiac nervous system, highlighting its intricate cellular interactions and dynamic regulatory roles in modulating cardiac rhythm and function. Importantly, facilitating a comprehensive understanding of sparse cardiac neuronal populations requires further technological refinement of current methodologies, with a focus on enhanced depth, capture capabilities, and spatial resolution.

The epicardium makes essential contributions to chamber development throughout heart morphogenesis. Originating from the proepicardial organ near the venous pole, epicardial cells spread across the myocardial surface beginning at E9.5 and subsequently undergo EMT to generate epicardium-derived cells (EPDCs). These EPDCs invade the myocardium and differentiate into cardiac fibroblasts, coronary smooth muscle cells, and perivascular stromal populations, essential for ventricular wall growth and coronary vasculature formation.^76,77 Epicardial signaling, including FGF, PDGF, RA, and Wnt pathways, regulate myocardial proliferation and compaction.^77–79 A recent single-cell study by Weinberger et al.⁸⁰ identified multiple transcriptionally and functionally distinct EPDC subpopulations during zebrafish heart development, suggesting that the epicardium extends specialized contributions to myocardial patterning and repair.

2.2.3. Postnatal maturation

During postnatal maturation, cardiomyocytes undergo profound structural, functional, and transcriptional remodeling to support the increasing physiological demands of the adult heart.^81,82 Morphologically, immature cardiomyocytes transition from small, round cells with disorganized sarcomeres to elongated, rod-shaped cells exhibiting highly organized myofibrillar alignment and mature intercalated discs that enable efficient force transmission. Functionally, cardiomyocytes withdraw permanently from the cell cycle, coinciding with binucleation and the acquisition of polyploidy in a subset of nuclei. These transformations collectively limit proliferative capacity while promoting hypertrophic growth. At the metabolic level, there is a hallmark switch from glycolytic energy production, predominant in the hypoxic neonatal environment, to mitochondrial oxidative phosphorylation and fatty acid β-oxidation. These coordinated transitions establish the metabolic efficiency, contractile specialization, and electrophysiological stability characteristic of the mature myocardium, while simultaneously constraining its regenerative potential beyond the neonatal stage.

Single-cell transcriptomic studies have begun to profile these dynamic postnatal transitions. While many have employed single-nuclei RNA sequencing (snRNA-seq) for technical feasibility,⁸³ true single-cell resolution remains challenging because cardiomyocytes are large and fragile. Using LP-FACS,⁸⁴ Kannan et al.⁸⁵ performed scRNA-seq on left ventricular cardiomyocytes from E14 to postnatal day (P) 84 and compared their maturation trajectory with that of pluripotent stem cell–derived cardiomyocytes (PSC-CMs). A key strength of this study was the high sequencing depth (~500,000 reads/cell), which captured ~4,000 genes per cell, enabling the identification of a critical maturation window during which ~80% of genes exhibited differential regulation. This analysis further revealed genes that are dysregulated in PSC-CMs, elucidating the transcriptomic basis for their developmental arrest. This dataset not only provides insights into strategies for promoting PSC-CM maturation in vitro but also serves as a valuable resource for enhancing our understanding of the molecular mechanisms underlying cardiomyocyte maturation in vivo.

Notably, this study used a scRNA-seq–based metric called entropy score⁸⁶ to address the special challenge of integrating maturation datasets across different sources to compare the maturation status. Traditional pseudotime-based trajectory analysis relies on comparing the transcriptional similarity between cells to align them in the order of developmental progression and can be used to effectively compare samples within the same experiment. However, they are highly sensitive to batch effects arising from differences in tissue dissociation, library preparation, and sequencing parameters. Entropy score is grounded in the idea that early-stage cells display more diverse transcription activities and a broad gene expression profile, whereas more advanced cells exhibit a narrower gene expression profile and thus have lower entropy scores. This metric has proven to be a robust tool for quantifying and comparing the maturation status across different studies and species. Murphy et al.⁸⁷ employed entropy score to examine the role of PGC1α during maturation. Transcriptomic analysis showed that PGC1α-deficient postnatal cardiomyocytes were developmentally stalled at around P14, consistent with their structural and functional maturation arrest. The study also identified YAP1 and SF3B2 as key transcriptional and splicing regulators that mediate cardiomyocyte maturation.

Heart maturation is fundamentally a multicellular process that depends on coordinated crosstalk between cardiomyocyte and non-myocyte populations. One single-cell study mapped the transcriptional maturation trajectories of cardiomyocytes and non-cardiomyocytes from neonatal to adult stages, revealing a postnatal switch in fibroblast subtypes from neonatal to adult states that secrete ligands promoting cardiomyocyte maturation.⁸⁸ Co-culture experiments showed that neonatal cardiomyocytes grown with adult cardiac fibroblasts exhibit enhanced morphological and electrophysiological maturation mediated by extracellular matrix (ECM)-receptor, chemokine, and STAT3 signaling. Notably, human embryonic-stem-cell–derived cardiomyocytes co-cultured with adult human cardiac fibroblasts also showed accelerated maturation, underscoring the conserved, instructive role of fibroblast cues. In another single-cell study that covered the E18.5 to P28 period, cellular cross talk analysis showed immune cell communication in the first week after birth and vascular and T cell communication at later time points.⁸⁹ The authors also found that Agrin and Notch signaling decreased gradually and that FGF signaling increased over time, as it is important for homeostasis. These single-cell studies showed that maturation of the myocardium is orchestrated by both intrinsic transcriptional programs and spatially organized intercellular communication. Understanding how these niches evolve and how they vary across anatomical regions of the heart represents a key frontier for ST approaches.

2.3. scRNA-seq–based Screening Methods

Emerging high-throughput screening methods such as Perturb-seq, CRISPR-seq, and CROP-seq enable researchers to track many perturbations such as knockouts, knockdowns, and overexpression in the same dish.^90–94 With these powerful tools, researchers can systematically probe gene regulatory networks during cardiomyocyte differentiation and maturation. By applying Perturb-seq to differentiating human PSC-CMs, Sivakumar et al.⁹⁵ linked disease-associated enhancers to transcriptional programs governing cardiac lineage commitment. Their study identified a previously unrecognized regulatory loop between NKX2.5 repression and NRG1 activation, suggesting intricate crosstalk between myocardial and endocardial signaling pathways. Recently, Wang et al.⁸³ introduced a novel in vivo perturbation platform called Probe-based Indel-detectable Perturb-seq (PIP-seq), enabling high-throughput, single-nucleus capture of sgRNA expression, perturbation status, and transcriptomic profile in the native postnatal heart environment. Applying PIP-seq, they identified 21 previously unknown regulators of cardiomyocyte maturation, including Rreb1, Mef2c, and Ppargc1a, whose perturbation disrupted metabolic and contractile maturation programs. Rreb1 loss impaired the postnatal metabolic shift toward oxidative phosphorylation, whereas Mef2c and Ppargc1a perturbations altered sarcomeric gene expression and mitochondrial biogenesis. Together, these regulators form nodal points that link transcriptional control with metabolic remodeling, offering mechanistic insights into congenital cardiomyopathies and potential targets for promoting regenerative maturation in engineered heart tissue.

2.4. Challenges and Limitations

2.4.1. Interpreting fate decisions and gene regulation from computational analysis

Single-cell transcriptomics enables reconstruction of putative differentiation hierarchies through pseudotime analysis and RNA velocity modeling (Figure 1). This type of analysis can be especially helpful in understanding fate decisions and gene regulation of the cardiac progenitor cells at early developmental stages. However, they infer developmental progression solely based on transcriptional similarity and splicing dynamics. In cardiac development where progenitor populations are spatially segregated and shaped by morphogen gradients, transcriptional proximity does not necessarily reflect direct lineage progression. Similarly, cell–cell communication analyses predict ligand–receptor interactions based on gene expression levels without accounting for physical proximity or protein-level activity. Thus, conclusions from these computational predictions require cautious interpretation and experimental validation.

Incorporating spatial information partially mitigates these limitations by restoring anatomical context and enabling assessment of whether proposed intermediate states or interactions could happen in space (Figure 1). Integrating lineage tracing, single-cell profiling, and spatial mapping therefore provides a more rigorous framework than computational modeling alone.

2.4.2. Applying scRNA-seq to postnatal and adult cardiomyocytes

Applying scRNA-seq to postnatal and adult cardiomyocytes presents unique challenges during cell isolation and library preparation (Figure 2). Mature cardiomyocytes are large, elongated, and fragile. Additionally, they are susceptible to shear stress during enzymatic dissociation, which results in low viability and biased cell recovery. Specialized dissociation approaches such as Langendorff perfusion⁹⁶ can improve cardiomyocyte yield but are technically demanding, time-consuming, and difficult to scale. In addition, conventional fluorescence-activated cell sorting (FACS) platforms are incompatible with large cardiomyocytes, as narrow flow channels and high pressures damage cells and lead to RNA leakage, reflected in abnormally high mitochondrial read percentages.

Figure 2.

Challenges and limitations of applying single-cell RNA sequencing (scRNA-seq) and spatial transcriptomics (ST) to heart development studies. CM, cardiomyocytes; MT, mitochondria. Created with BioRender.com.

In early cardiac RNA-seq methods, researchers handpicked cells,⁹⁷ which, despite allowing visual selection of viable or fluorescently labeled cells, was labor-intensive and prone to ambient RNA contamination. Delaughter et al.⁹⁸ used Fluidigm integrated flow chips to study ~1200 embryonic and early postnatal cells. They showed the transcriptional progression of ventricular cardiomyocytes and were able to determine the age of stem cell–derived cardiomyocytes. While groundbreaking, these early efforts underscored the need for higher-throughput technologies.

Recently, platforms tailored for large cells such as the iCell8⁹⁹ and large particle (LP)-FACS⁸⁴ have been introduced to address the issue. LP-FACS, also known as the “worm sorter,” features a larger channel size and uses slow, low-pressure flow, making it ideal for sorting large myocytes. Adult cardiomyocytes isolated by LP-FACS display a normal range of mitochondrial transcripts (~40%) and are also suitable for single-cell functional analysis. For downstream library preparation, postnatal/adult cardiomyocytes remain incompatible with droplet-based platforms owing to their large size but can be successfully processed using plate-based or split-pool methods.

An alternative solution to circumvent the challenge of isolating high quality adult cardiomyocytes is to use single nuclei for RNA-seq¹⁰⁰. Single-nuclei RNA sequencing (snRNA-seq) can be applied to sample types where whole-cell isolation is difficult, including heart neurons and frozen samples (Figure 1). Furthermore, it is compatible with droplet-based library preparation methods. A major drawback of snRNA-seq is that it does not capture cytoplasmic transcripts, including structural, mitochondrial, and metabolic mRNAs that are abundant and functionally important in mature cardiomyocytes. A direct comparison of scRNA and snRNA with human induced pluripotent stem cell–derived cardiomyocytes showed that snRNA-seq detects 30% to 50% fewer transcripts than scRNA-seq.¹⁰¹ Additionally, because adult hearts contain multinucleated cardiomyocytes (~95% in rodents and ~25% in humans)¹⁰², snRNA-seq does not fully reflect single-cell resolution. The number of genes detected per nuclei is often lower, and longer transcripts are more likely to be detected, overrepresenting unspliced pre-mRNAs, making snRNA-seq less suitable for analyses of translation-level regulation or post-transcriptional dynamics.¹⁰³

3. SPATIAL TRANSCRIPTOMICS IN HEART DEVELOPMENT AND MATURATION

3.1. Spatial Transcriptomics Platforms

ST enables transcriptome-wide RNA profiling directly within intact tissue sections, preserving spatial architecture and allowing cell identity and function to be interpreted within native microenvironmental context. This is particularly critical for cardiovascular development, where positional cues govern progenitor deployment, chamber morphogenesis, and region-specific maturation programs. ST has therefore emerged as an essential complement to scRNA-seq, which captures transcriptional heterogeneity but loses spatial information during tissue dissociation. Current ST technologies can be broadly categorized into two classes: sequencing-based platforms, which capture transcripts labeled with spatial barcodes followed by next-generation sequencing, and imaging-based platforms, which visualize and quantify selected transcripts in situ through multiplexed fluorescence hybridization. Sequencing-based approaches typically offer unbiased, transcriptome-wide detection with compromised spatial resolution and depth, whereas imaging-based approaches achieve subcellular localization of targeted gene panels. Therefore, selection of the appropriate method depends on the specific biological questions posed, whether the goal is discovery of novel biomarkers or high-resolution mapping of known signaling interactions.

Among sequencing-based approaches, Visium³ remains the most widely used. Its first-generation array utilizes 55-μm capture spots that typically encompass multiple cells, whereas the newly developed Visium HD replaces spots with 2-μm barcoded pixels to achieve near single-cell spatial resolution. The earlier probe-based chemistry permitted detection of ~18,000 genes, whereas the most recent 3′ chemistry now supports unbiased transcriptome-wide capture. Visium HD is compatible with fresh frozen, fixed frozen, and formalin-fixed paraffin-embedded (FFPE) tissues and can be overlaid with hematoxylin & eosin or immunofluorescent images for visualization of histology or selected protein markers.

Other sequencing-based platforms include Slide-seq¹⁰⁴ and Slide-seqV2,¹⁰⁵ which use spatially barcoded 10-μm beads to transfer mRNA from tissue sections. These techniques have been successfully used to map early organogenesis and reconstruct three-dimensional (3D) embryo architecture.¹⁰⁶ Stereo-seq¹⁰⁷ pushes resolution further by employing densely patterned DNA nanoballs with 220-nm spacing, supporting both subcellular precision and large capture areas that enable creation of whole-embryo spatial atlases such as the mouse organogenesis spatiotemporal transcriptomic atlas (MOSTA)¹⁰⁷ and high-resolution heart datasets that span multiple developmental stages.^108,109 Another commercially available system, GeoMx Digital Spatial Profiler,¹¹⁰ quantifies targeted RNAs or proteins in user-defined regions using photocleavable tags and has been used to study regional gene regulation in the postnatal CCS.¹¹¹ Additional technologies such as Seq-Scope,¹¹² Open-ST,¹¹³ and LCM-seq¹¹⁴ further expand user flexibility for high-resolution or anatomically targeted gene expression profiling.

Imaging-based ST platforms directly visualize and quantify RNA molecules within the tissue context through multiplexed fluorescence imaging. These methods provide subcellular resolution but only target a predefined set of transcripts. MERFISH¹¹⁵ uses combinatorial barcoding and error-correction strategies to detect thousands of transcripts with 100-nm precision and has revealed laminar organization and conduction-system patterning in the human fetal heart.¹¹⁶ seqFISH+¹¹⁷ takes a similar approach but performs multiple rounds of hybridization and imaging to sequentially detect up to 10,000 genes. Although technically demanding, seqFISH+ has proven powerful for characterizing spatial gradients of gene expression during early organogenesis in the mouse.¹¹⁸ Other commercial imaging-based systems, such as Xenium and CosMx Spatial Molecular Imager,¹¹⁹ combine highly multiplexed RNA and protein detection with ~100-nm spatial resolution, positioning them as key platforms for future multi-omic heart studies. More details of the above ST platforms are summarized in Table 1.

Table 1.

Summary of Common Spatial Transcriptomic Platforms.

Platform/Method (manufacturer)	Type	Resolution	Transcriptome coverage	Compatible samples	Image overlay
Visium (10x Genomics)	Sequencing-based	55 μm spots	V1 3’: Whole transcriptome V2 WT: Probe-based	V1(3’): FF V2 (WT): FF/FFPE/FxF	H&E
Visium HD (10x Genomics)	Sequencing-based	2 μm × 2 μm pixels	V1 WT: Probe-based (~18k genes) V2 3’: Whole transcriptome	V1 (WT): FF/FFPE/FxF V2 (3’): FF/FFPE	H&E/IF
Slide-seq (Takara Curio Bioscience)	Sequencing-based	10 μm beads	TREKKER: nuclear transcripts only SEEKER: whole transcriptome	FF	Incompatible, computationally decode spatial location
Stereo-seq (STOmics/BGI)	Sequencing-based	220 nm DNA nanoballs	Whole transcriptome	V1: FF V2: FF/FFPE	H&E/IF
GeoMx DSP (NanoString)	Sequencing-based	5 μm × 5 μm AOI	Probe-based (~18k genes) + >570 proteins	FF/FFPE	H&E/IF
Xenium (10x Genomics)	Imaging-based	30–100 nm	50–5000 genes	FF/FFPE	H&E/IF
MERSCOPE (Vizgen)	Imaging-based	100 nm	Predesigned/custom-designed panel (up to 1000 genes)	FF/FFPE	Cell boundary and DAPI staining
seqFISH+	Imaging-based	100 nm	Multiple imaging rounds to profile up to 10k genes	FF/FxF	IF/Cell boundary and DAPI staining
CosMx SMI (NanoString)	Imaging-based	50–120 nm	Predesigned/custom-designed panel (up to 19k genes)	FF/FFPE	H&E/IF
Seq-Scope	Sequencing-based	0.5–0.8 μm	Whole transcriptome	FF	H&E/IF
Open-ST	Sequencing-based	0.6 μm	Whole transcriptome	FF	H&E
LCM-seq	Sequencing-based	Isolate whole region of interest	Whole transcriptome	FF/FFPE/FxF	H&E/IF

AOI, area of interest; DAPI, 4’,6-diamidino-2-phenylindole; DSP, Digital Spatial Profiler; FF, fresh frozen; FFPE, formalin-fixed paraffin-embedded; FxF, fixed frozen; H&E, hematoxylin & eosin; IF, immunofluorescence; SMI, Spatial Molecular Imager; WT, wild type; V, version.

Recent advances in ST technologies reflect a clear trajectory toward higher spatial resolution, greater molecular sensitivity, and broader compatibility with diverse sample types and multi-omic assays. Earlier platforms prioritized transcriptome breadth over spatial resolution, but newer systems approach single-cell and even subcellular resolution while maintaining nearly whole-transcriptome coverage. Enhanced RNA capture chemistry and imaging throughput have improved detection of low-abundance transcripts, enabling the characterization of transient cell states, signaling niches, and fine-scale developmental processes. Notably, expanded compatibility with FFPE and fixed frozen tissues has opened opportunities to study clinical samples that were previously inaccessible to spatial profiling. Meanwhile, simultaneous processing of RNA, protein, and additional molecular modalities within the same tissue section greatly increases the biological questions that can be addressed, offering deeper insights into how gene regulation, microenvironmental cues, and tissue architecture collectively drive heart development.

3.2. Application to Heart Development and Maturation

A growing number of studies are applying ST approaches to investigate cardiovascular development and diseases. Recent reviews have summarized the implementation of ST in cardiac biology, including technical considerations and emerging insights into regional cellular heterogeneity and microenvironmental signaling.^120–123 Owing to strong clinical relevance and dynamic cellular interactions, ST methods have been widely deployed to study ischemic and fibrotic remodeling^122,123; however, these disease-focused studies fall outside the scope of this review. Here, we examine the application of ST across the continuum of heart development, from early cardiac specification to postnatal maturation. Notably, few studies have focused on the earliest stages, such as cardiac progenitor specification and heart tube morphogenesis, likely reflecting the technical challenges of capturing small, 3D embryonic structures that require multiple sections and precise alignment to recover full spatial information. In contrast, an increasing number of recent studies have applied ST to cardiac maturation, a developmental window characterized by complex multicellular organization and microenvironmental diversity that aligns well with the strengths of spatial profiling. In the following section, we review representative studies that span these developmental stages (Figure 3, Table 2) and discuss how ST is reshaping our understanding of the spatial logic that underlies heart formation and functional maturation.

Figure 3.

Key spatial transcriptomic datasets spanning heart developmental time points. Triangles in the figure point to the timeline axis (mouse/human) used. A, atrium; E, embryonic day; FHF, first heart field; LA, left atrium, LV, left ventricle; OFT, outflow tract; P, postnatal day; pcw, postconceptional week. RA, right atrium; RV, right ventricle; SHF, second heart field; SV, sinus venosus; V, ventricle. Created with BioRender.com.

Table 2.

Summary of Key Spatial Transcriptomic Studies Mentioned in This Review.

Study	Sample	Stage	ST Platform	Integration	Findings
Asp et al.124	Human heart	4.5, 6.5, 9 pcw	ST + ISS	6.5–7 pcw scRNA-seq	Resolved spatiotemporal gene expression pattern in myocardium, epicardium, AV mesenchyme, and OFT. Spatially mapped neural crest cells and Schwann progenitor cells across developmental stages. Identified three cardiomyocyte subtypes in the human embryonic heart, including a MYOZ2-enriched population previously described only in adult mouse myocardium.
Mantri et al.125	Chicken heart	HH21-HH24, HH30-HH31, HH35-HH36, and HH40	Visium	HH21-HH24, HH30-HH31, HH35-HH36, and HH40 scRNA-seq	Mapped differentiation trajectory to spatial context. Confirmed that EPDC fate decision occurs after they migrate into the myocardium. Showed that upregulation of ECM genes AGRN, EGFL7, POSTN, and FN1 might guide EMT of EPDCs. Discovered that TMB4X is highly enriched in coronary vasculature cells.
Farah et al.116	Human heart	12–13 pcw	Merfish	13 pcw scRNA-seq	Identified the interactions among PLXNA2+PLXNA4+ ventricular cardiomyocytes, SEMA3C+SEMA3D+ fibroblasts, and SEMA6A+SEMA6B+ endothelial cells in patterning the multilayer ventricular wall during compaction.
Lázár et al.68	Human heart	6–12 pcw	Visium+ISS	5.5–14 pcw scRNA-seq	Generated the most comprehensive spatiotemporal atlas of first-trimester human heart development to date. Spatially mapped previously understudied populations in developing human heart, including CCS, cardiac innervation, cardiac endothelial cells, fibroblasts, and mesenchymal cells. First mapped chromaffin cells in developing human heart that are absent in mouse hearts.
Oh et al.111	CCS-tdT mouse heart	P4	GeoMx	CCS-tdT P1, P2, P4 scRNA-seq	Generated a spatially resolved map of the CCS and identified region-specific molecular markers validated by RNA FISH and immunofluorescence. Integrated E16.5 embryonic dataset to compare prenatal and postnatal CCS, identified transcriptomic consistency between these developmental stages, as well as significant changes that indicate CCS maturation. Investigated the individual and combined actions of Tbx3 and Irx3 on CCS gene regulation with published ATAC-seq datasets and in vitro models.
Zheng et al.109	Mouse heart	P0, P7, P56	Stereo-seq	No integration	Provided a high-resolution mouse ST dataset that included key maturation time points.
Kang et al.108	Mouse heart	E20, P0, P4, P14	Stereo-seq	No integration	Investigated cellular composition of maturing mouse heart and observed decreased cell–cell interactions during maturation. Performed trajectory analysis and identified left-right–specific genes.
Wang et al.83	Mouse heart	P0, P7, P14, P21	Xenium	P0, P7, P14, P21 snRNA-seq	Identified cell niches with distinct spatiotemporal patterns during postnatal maturation. Analyzed cell communications and gene regulatory networks within each niche. Uncovered interactions between capillary endothelial cells and cardiomyocytes in regulating cell cycle exit. Developed high-throughput in vivo Perturb-seq system PIP-seq to validate novel regulator genes that promote maturation. Identified 23 regulators of maturation, including the previously uncharacterized target Rreb1.

ATAC-seq, assay for transposase-accessible chromatin with sequencing; AV, atrioventricular; CCS, cardiac conduction system; E, embryonic day; ECM, extracellular matrix; EMT, epithelial-to-mesenchymal transition; EPDC, epicardium-derived cells; FISH, fluorescence in situ hybridization; HH, Hamburger Hamilton; ISS, In situ sequencing; OFT, outflow tract; P, postnatal day; pcw, postconceptional week; PIP-seq, protein interaction profile sequencing; scRNA-seq, single-cell RNA sequencing; snRNA-seq, single-nuclei RNA sequencing; ST, spatial transcriptomics; tdT, tdTomato.

3.2.1. ST studies in heart morphogenesis

Asp et al.¹²⁴ generated the first spatiotemporal, organ-wide transcriptomic atlas of the developing human heart. Using the original ST method³ with a resolution of ~30 cells per capture spot, they profiled human embryonic hearts at 4.5, 6.5, and 9 postconceptional weeks (pcw) and integrated scRNA-seq data from 6.5 to 7 pcw hearts to resolve cellular heterogeneity. The authors then validated key developmental markers at subcellular precision using in situ sequencing, a method that uses rolling-circle amplification to directly amplify transcripts on the tissue section. Interestingly, their analysis revealed that region-specific gene programs in the myocardium, AV mesenchyme, and epicardium were already established by 4.5 to 5 pcw and remained spatially restricted, whereas epicardial, neural crest, and Schwann progenitors exhibited dynamic spatiotemporal redistribution. They also identified a MYOZ2-enriched cardiomyocyte cluster that had been reported previously in adult mouse hearts. Despite limited spatial resolution, this landmark study provided the first 3D molecular map of early human cardiogenesis and established that spatially patterned transcriptional programs underlie early heart chamber formation.

Mantri et al.¹²⁵ generated a spatiotemporal atlas of ventricular development by integrating Visium profiling with scRNA-seq of embryonic chicken hearts, spanning early septation to the four-chambered stage. By projecting single-cell trajectories onto spatial sections, the authors reconstructed epicardial, endocardial, and myocardial lineage transitions and revealed region-specific gene programs underlying ventricular differentiation. A key finding was that EPDCs migrate into the myocardium while maintaining progenitor-like transcriptional signatures, implying that fate specification occurs after migration. Spatial maps also revealed enrichment of extracellular matrix components (AGRN, EGFL7, POSTN, FN1) in regions corresponding to EPDC invasion, suggesting that matrix-derived cues orchestrate temporally and regionally restricted EMT events critical for chamber morphogenesis. The study further identified spatiotemporal expression of congenital heart defect–associated genes (GATA5, IRX4, PITX2, TBX5, ACTC1) and discovered a TMSB4X-high population enriched in coronary vascular cells with elevated cytoskeletal and calcium signaling activity. These findings establish a dynamic spatial framework for ventricular morphogenesis and highlight TMSB4X as a potential regulator of coronary vasculature and cardiac maturation.

Farah et al.¹¹⁶ profiled the human fetal heart during ventricular and conduction system development by integrating MERFISH imaging (12–13 pcw) with a stage-matched scRNA-seq dataset (13 pcw). Their analysis revealed spatially continuous but transcriptionally distinct cardiomyocyte layers within the ventricular wall; the LV exhibited greater cellular complexity and progressive gene expression gradients across wall depth. Importantly, they mapped rare BMP2⁺ non-chambered cardiomyocytes and conduction system progenitors (SHOX2⁺, TBX3⁺, HCN4⁺, ISL1⁺) that formed continuous transitions from atrial to nodal regions. Cell–cell interaction analyses further revealed the role of PLXN–SEMA signaling in coordinating myocardial lamination. These insights were validated in a 3D bioprinted hPSC–derived ventricular wall model, confirming functional conservation of the identified signaling interactions in vitro. This study provides the most detailed spatial framework of human ventricular and conduction system development and reveals how coordinated signaling and spatial patterning drive cardiac morphogenesis.

Building on their earlier spatial atlas,¹²² Lázár et al.⁶⁸ recently used an updated Visium platform and expanded the temporal profiling window to generate the most comprehensive spatiotemporal transcriptomic atlas of human fetal heart development to date. The authors performed Visium for 6 to 12 pcw hearts and integrated the data with 5.5 to 14 pcw scRNA-seq datasets. A major advance of this study was the high-resolution spatial reconstruction of the cardiac pacemaker–conduction system, which demonstrated molecular specialization and electrophysiological coupling with adjacent myocardium and fibroblast networks. Importantly, the authors identified and precisely localized resident neuroendocrine chromaffin cells and associated innervation trajectories, providing the first evidence of neuronal–vascular co-patterning and autonomic regulation emerging during early human heart development. Compared with their pioneering 2019 atlas, which established early regional identity in the human heart, this updated dataset provided finer-grained delineation of conduction system, valvular, neural, and neuroendocrine territories and extended developmental staging into the critical window of functional specialization. These advances mark a key step toward a complete spatial blueprint of human cardiogenesis.

3.2.2. ST studies in maturation

Spatial transcriptomics provides a major advantage for studying cardiomyocyte maturation by bypassing the tissue dissociation challenges (Figure 2) and directly profiles various cell types within their local signaling niche. A few recent studies have used Stereo-seq to extend spatial profiling into the postnatal period. Zheng et al.¹⁰⁹ generated a high-resolution spatial atlas of P0, P7, and P56 mouse hearts and captured dynamic changes in regional transcriptional states. Kang et al.¹⁰⁸ similarly analyzed late embryonic to early postnatal hearts (E20–P14), identifying spatiotemporal changes in cell–cell interactions and genetic signatures that contribute to atrial asymmetry. Oh et al.¹¹¹ spatially mapped the maturation of the CCS. They utilized a CCS lineage tracing mouse model (Cntn2^Cre/+;Rosa26^tdTomato/+) and integrated MULTI-seq and GeoMx spatial profiling to reconstruct region-specific gene regulatory modules defining the sinoatrial node, atrioventricular node, His bundle, and Purkinje network. These analyses identified key transcription factors, including Tbx3, Irx3, Gnao1, Bmp2, and Prdm6, that coordinate electrical specialization across CCS compartments. Functional validation showed that Tbx3 maintains nodal identity by repressing chamber-like gene programs, whereas Irx3 activates ventricular conduction pathways to promote His–Purkinje differentiation. Spatial mapping further revealed zonated transcriptional boundaries that distinguish CCS subdomains. This work established the first integrated regulatory framework of the postnatal conduction system, linking transcriptional control, spatial organization, and functional maturation.

In the PIP-seq study, Wang et al.⁸³ performed Xenium experiments on P0 to P21 mouse hearts for a selected list of maturation-related genes and integrated the results with their snRNA-seq data to build a spatial maturation map. Their analysis delineated distinct cellular niches with unique temporal and spatial maturation trajectories, revealing key intercellular signaling hubs such as capillary endothelial–cardiomyocyte interactions mediated by the PTPRM-MEIS1 axis that drives cardiomyocyte cell cycle exit and metabolic reprogramming. Combined with the PIP-seq validation results, this study represents a major advance in linking spatial and single-nucleus transcriptomics with functional perturbation screening. The findings provide an integrated framework for dissecting the signaling and transcriptional checkpoints that coordinate postnatal cardiac maturation.

3.3. Whole-Embryo Atlases

Most ST studies of heart development have focused on organ-level analysis during mid-to-late gestation, leaving the earliest phases of mesoderm specification and heart-field formation largely underexplored. During these stages, the heart is not yet morphologically distinct and relies on inductive cues from surrounding tissues, including the foregut endoderm, pharyngeal mesoderm, and neural crest, to establish cardiac identity and spatial organization. Whole-embryo spatial atlases therefore provide a powerful opportunity to resolve organism-scale patterning, long-range signaling gradients, and extracardiac regulatory networks that are inaccessible to organ-restricted studies.

Harland et al.¹¹⁸ integrated seqFISH maps at E7.25–E7.5 with existing E8.5 spatial and E6.5–E9.5 single-cell datasets to capture the earliest transcriptional transitions from mesoderm to nascent cardiac mesoderm within the full embryonic geometry. Sampath Kumar et al.¹⁰⁶ employed Slide-seq at E8.5–E9.5 and introduced a 3D reconstruction method that merges serial sections into virtual embryos, enabling volumetric visualization of regionally restricted expression of cardiac markers such as Nppa and Myh6. The MOSTA Stereo-seq atlas¹⁰⁷ profiled E9.5–E16.5 embryos with near-subcellular resolution, capturing chamber formation and early coronary vasculature. Using Visium at E13.5, Qu et al.¹²⁶ generated a whole-embryo spatial map that delineated distinct cardiac subdomains, including ventricular (Myl2), OFT (Eln), and epicardial (Nsrp1) regions. Spatial deconvolution further identified eight cardiac cell types, providing a valuable reference for chamber patterning, epicardial development, and cardiac–extracardiac signaling during mid-gestation. Although constrained by limited gene coverage, discontinuous sampling, and suboptimal resolution of small cardiac structures, these whole-embryo atlases illustrate the feasibility of mapping cardiac lineage at the earliest stages and at the whole-body scale. Future advancement in computational tools for 3D reconstruction, multimodal alignment, and cross-scale data integration will be key to extending heart-focused studies toward a truly systems-level understanding of cardiac development.

3.4. Challenges and Limitations

Applying spatial transcriptomic profiling to heart development studies faces several technical and biological barriers. From a technical standpoint, current ST platforms are constrained by a fundamental trade-off between spatial resolution and molecular depth (Figure 1). Imaging-based methods achieve subcellular precision but are restricted to targeted gene panels, typically detecting hundreds to thousands of transcripts. In contrast, sequencing-based approaches offer near–whole-transcriptome coverage, detecting more than 10,000 genes, but at lower spatial resolution, with each capture spot encompassing multiple cells. Hence, these platforms are well suited for mapping large-scale morphogen gradients and tissue domains, but they struggle to resolve rare, transient, or intermediate cell states whose transcriptional signatures are easily diluted across neighboring cells. ST dataset analysis still relies heavily on integration with scRNA-seq references to achieve sufficient depth and statistical robustness. Therefore, improving both spatial resolution and molecular sensitivity remains a central technical goal for advancing ST in cardiac research.

Biologically, the heart’s size variability and dynamic morphological transformations across developmental stages complicate standardized analysis (Figure 2). During early morphogenesis, the heart is small, curved, and hollow, and its fragile structures such as cardiac cushions, valves, and trabeculae are easily lost or distorted during sectioning, compromising spatial integrity. The limited RNA yield from these small tissues demands high sequencing depth and resolution to capture fine lineage relationships. Moreover, early cardiac morphogenesis occurs rapidly within narrow developmental windows, and many transient structures or intermediate cell states are difficult to capture in a single section. Achieving full 3D reconstruction of these stages requires serial sectioning and image registration. However, such an undertaking is computationally complex and limited by small capture areas and the high cost of available ST platforms. At later stages, postnatal maturation progresses gradually over extended time windows, and subtle transcriptional changes require high sensitivity and sequencing depth. The increasing size of postnatal hearts further necessitates platforms with large capture areas to achieve whole-organ coverage.

Owing to these challenges, current ST datasets display both developmental and species-related biases (Figure 2). Most studies have concentrated on mid-to-late gestation or postnatal maturation, whereas the earliest stages, such as gastrulation and heart field specification, remain largely underexplored. Sampling biases also exist between species. Datasets that capture early heart morphogenesis derive primarily from human embryonic hearts, whose larger size and structural integrity are more compatible with current spatial platforms. In contrast, maturation-focused studies rely predominantly on mouse models because mouse scRNA-seq references are abundant, genetic manipulation is efficient, and postnatal tissue is easy to access. Hence, improved ST methods are needed that would enable researchers to conduct consistent profiling across matched developmental windows of different species and capture the full temporal and anatomical continuum of cardiac formation.

4. EMERGING TOOLS FOR SPATIAL PROFILING OF THE HEART

4.1. Profiling Beyond the Transcriptome: Spatial Multi-omics

Integration of ST data with scRNA-seq datasets has proven highly effective for overcoming the limitations in spatial resolution and molecular depth inherent to current ST approaches. Recent advances now extend this integration and joint profiling beyond the transcriptome to include additional molecular layers such as the proteome, epigenome, and metabolome, within the same tissue section. These spatial multi-omic assays have already demonstrated their potential in complex cardiovascular research.¹²⁷

Multi-omic profiling can be incorporated directly into the experimental workflow and performed on the same tissue section, thereby minimizing batch effects and eliminating inaccuracies introduced by downstream data integration. However, such protocols often sacrifice transcriptomic depth or spatial resolution to accommodate additional modalities, as each omics layer requires distinct chemistries and reaction conditions. Recent innovations such as dual spatial profiling of transcriptomes and epigenomes¹²⁸ and SPACE-seq¹²⁹ now allow co-detection of chromatin accessibility and mRNA expression within intact tissues. Likewise, DBiT-seq¹³⁰ enables simultaneous mapping of mRNA, proteins, and open chromatin states, while commercial platforms such as Xenium and GeoMx Digital Spatial Profiler extend transcriptomic assays with targeted protein panels. Although not yet widely applied to cardiac samples, these emerging technologies provide a robust foundation for uncovering how multi-omic landscapes govern the patterning, signaling, and structural maturation of the developing heart.

In parallel, computational integration of multi-omic datasets is becoming a powerful strategy to reconstruct multilayered spatial landscapes. In a notable example, Kanemaru et al.⁷³ performed paired snRNA-seq and snATAC-seq and integrated the results with scRNA-seq, snRNA-seq, and Visium datasets to generate a comprehensive spatial gene regulatory map of the adult human heart. Their analysis revealed distinct chromatin accessibility patterns in different cardiac cell types. They also identified various cardiac niches with previously unknown cell composition. In another recent study, authors developed SEU-TCA,¹³¹ a computational tool for integrating scRNA-seq and spatial data via transfer component analysis. The authors applied this tool to gastrulation stage datasets and resolved the spatial localizations of heart field progenitors. Specifically, they identified the spatial enrichment of Irx1 in anterior SHF cells and validated its function in regulating anterior SHF contribution to ventricular septation.

Recent progress in machine learning tools has further expanded the analytical frontier of multimodal spatial biology. Li et al.¹³² developed SWITCH, a deep-generative model that integrates unpaired spatial omics, including transcriptomics, epigenomics, and proteomics, using graph attention networks to encode spatial relationships among cells. Through a novel cycle-mapping mechanism, SWITCH performs cross-modal translation and enforces reconstruction consistency, effectively generating pseudo-paired datasets to bridge modality gaps. When applied to developing embryo and adult brain datasets, it achieved integration fidelity and spatial resolution superior to that of existing tools. As similar frameworks are adapted to cardiac tissues, such models will be key to unifying transcriptomic, epigenetic, and proteomic information within a coherent, spatially resolved framework of heart development.

4.2. Predicting the Missing Information: Artificial Intelligence

The development of artificial intelligence (AI) is revolutionizing how transcriptional and spatial information can be inferred, reconstructed, and predicted from existing datasets. A novel machine learning framework called iSCALE¹³³ addresses the challenge of scaling ST mapping to large organs such as the heart. By aligning localized spatial captures with histological images, iSCALE applies deep learning to predict gene expression across unmeasured regions. Through hierarchical feature learning, it achieves super-resolution reconstruction (~8-μm per pixel), enabling accurate cell type annotation and segmentation across millimeter-scale tissue sections. Benchmark analyses showed that iSCALE outperforms existing tools for reconstructing spatial organization in complex tissues such as gastric cancer. Thus, it offers a scalable framework for developing whole-heart molecular atlases.

Other emerging transformative tools for single-cell and spatial omics are the foundation models. These AI models create “virtual cells,” which are in silico representations that simulate, predict, and reason about cellular behavior. These models, analogous to large language models in natural language processing, are typically based on transformer architectures and trained on tens to hundreds of millions of single-cell transcriptomes spanning diverse tissues, species, and perturbations. These large models are pretrained with self-supervised objectives such as masked gene expression prediction, contrastive learning, or next-cell prediction to capture generalizable representations of transcriptional programs. They are then fine-tuned on downstream tasks, including cell type annotation, perturbation response prediction, and disease-state classification. Examples of such models include scGPT,¹³⁴ CellFM,¹³⁵ Geneformer,¹³⁶ and scFoundation.¹³⁷ These models utilize architectures with over a billion parameters to learn gene–gene relationships and predict perturbational outcomes, enabling mechanistic inference at the network level. Notably, integrative frameworks such as HEIST¹³⁸ extend beyond transcriptomics by incorporating spatial and proteomic data through hierarchical graph representations. HEIST captures both intracellular and intercellular communication within tissue microenvironments and is moving the field closer to comprehensive, context-aware modeling of cellular systems.

5. CONCLUSIONS AND FUTURE DIRECTIONS

In this review, we summarized recent advances in single-cell and ST technologies and highlighted how they have been applied across species to dissect the cellular and molecular mechanisms governing heart development. These approaches have enabled the discovery of previously unrecognized cell states, developmental trajectories, and spatially restricted signaling programs that collectively drive cardiac morphogenesis and maturation.

Despite these advances, several major challenges remain. The uneven distribution of datasets across developmental stages and species continues to limit our understanding of the complete developmental continuum. In particular, the scarcity of datasets that capture gastrulation and early cardiac specification constrains insights into the spatial origins and lineage diversification of cardiac progenitors. Most available datasets from these stages are derived from whole-embryo sections, which provide only coarse spatial resolution of cardiac domains. As ST platforms now offer greater flexibility with image overlay, generating early embryo datasets overlayed with heart progenitor lineage tracing marker will be essential for uncovering the earliest patterning events. In parallel, integrating cardiac lineage-specific scRNA-seq datasets with these spatial whole-embryo references provides a great opportunity to link lineage contributions to a spatial context.

Spatial studies now increasingly focus on heart maturation, a dynamic process that involves coordinated interactions among cardiomyocytes, fibroblasts, vascular cells, and immune cells within spatially defined signaling niches. However, generating high-quality scRNA-seq reference datasets of adult cardiomyocytes for integration remains technically challenging. Moreover, many aspects of maturation, such as structural remodeling and metabolic switching, extend beyond the transcriptome. Integrating ST with multi-omic layers will therefore be critical for achieving a more comprehensive understanding of the molecular logic underlying postnatal heart maturation and function.

Future progress will depend on innovations that bridge these biological and technical barriers. Advances in high-resolution capture chemistry, 3D spatial reconstruction, and lineage-aware multimodal integration will be essential for mapping progenitor dynamics within the full embryonic context. Improved compatibility with diverse sample preparations, including fixed frozen and FFPE tissues, will broaden clinical and translational applications. In parallel, computational innovations such as probabilistic deconvolution, deep-learning–based cell fate prediction, and integrated spatiotemporal modeling will enable finer-grained, mechanistic interpretation of heart morphogenesis. Combining ST with complementary modalities such as epigenomics, proteomics, metabolomics, and live imaging will further link positional information to regulatory and functional states, enabling true multidimensional reconstruction of cardiac development.

These technological and analytical advances will move the field toward a comprehensive, multiscale understanding of how a functional four-chambered heart emerges from simple progenitor populations. This knowledge will not only refine fundamental developmental biology but also lay the groundwork for addressing congenital heart disease and informing regenerative and bioengineering strategies aimed at rebuilding or repairing the human heart.

Non-standard Abbreviations and Acronyms

AV

atrioventricular

AVC

atrioventricular canal

BMP

bone morphogenic protein

CCS

cardiac conduction system

CNG

cardiac nexus glia

DMP

dorsal mesenchymal protrusion

E

embryonic day

ECM

extracellular matrix

EMT

epithelial-to-mesenchymal transition

EPDC

epicardium-derived cell

FACS

fluorescence-activated cell sorting

FFPE

formalin-fixed paraffin-embedded

FGF

fibroblast growth factor

FHF

first heart field

JCF

juxta-cardiac field

LP-FACS

large particle FACS

LV

left ventricle

MOSTA

mouse organogenesis spatiotemporal transcriptomic atlas

NCC

neural crest cell

NGF

neural growth factor

OFT

outflow tract

P

postnatal day

pcw

postconceptional week

PSC-CMs

pluripotent stem cell–derived cardiomyocytes

RA

retinoic acid

RV

right ventricle

scATAC-seq

single cell assay for transposase-accessible chromatin with sequencing

scRNA-seq

single-cell RNA sequencing

SHF

second heart field

snRNA-seq

single-nuclei RNA sequencing

ST

spatial transcriptomics