Introduction

Building organs, such as the heart, starts with the specification of a pool of progenitor cells and segregation towards the different lineages that will form the tissue. The mammalian heart is a complex organ, with different chambers, that ensure the oxygenation of the blood and transport of oxygen and nutrients throughout the organism during both fetal and post-natal life. The adult heart contains diverse cell types including myocardial and endocardial cells that appear progressively during embryonic development, which are essential to maintain its function. Heart development is a complex but ordered process that is spatially and temporally regulated [1]. Studies in chick embryos have first defined the cardiogenic mesoderm as the area of the lateral plate mesoderm that has the potential to form myocardium. In the mouse, the earliest cardiac progenitors originate from the primitive streak during gastrulation, around embryonic day (E) 6.5, and then migrate rostrally (anteriorly) to form the first differentiated structure called the cardiac crescent at E7.75. The earliest molecular regulator of these cardiac progenitor cells (CPCs) is the transcription factor Mesp1 (Mesoderm posterior 1) [2]. Mesp1 is expressed in gastrulating mesoderm but then is rapidly downregulated. Cells in the cardiac crescent constitute the first heart field (FHF) and mainly form the left ventricle (LV) [3]. Convergence of the bilateral structures results in the formation of the linear heart tube. The heart tube is a transient structure composed of an inner layer, the endocardium, and covered by a myocardial layer. The rapid growth of the heart tube is driven by progressive addition of second heart field (SHF) progenitor cells from adjacent splanchnic or so called pharyngeal mesoderm [4–6]. These cells ultimately contribute to the cardiac outflow tract (OFT), right ventricle (RV), and a major part of atrial myocardium [4, 7]. Impaired SHF development has been shown in animal models to result in a spectrum of congenital heart defects (CHDs). As the tube forms, it undergoes asymmetric morphogenesis (heart looping), which corresponds to the formation of a rightward helix at E8.5. After rightward looping, the heart is shaped by expansion of the myocardium, which leads to the formation of four cardiac chambers. Separation of the left and right ventricular chambers depends on the interventricular septum, which has dual contribution from the two heart fields [8, 9]. The heart is connected by the OFT to the aortic sac and the pharyngeal arch arteries. At its maximal extension, the OFT is a tensioned myocardial cylinder lined with endocardial cells. Interactions between multiple cell types including myocardial, endocardial, and neural crest cells are involved in septation of the OFT. In addition to myocardial (CMs) and endocardial lineages (ECs), the adult heart contains smooth muscle cells (SMCs), fibroblasts, and vascular cells that derive from the epicardium, which itself originates from the proepicardium located at the base of the looping heart [10].

Our knowledge on the early steps of heart development has increased considerably since the discovery of the SHF, 20 years ago, and the emergence of new approaches that allow investigation of cardiac specification and differentiation at single cell resolution. We will discuss in this review what these single cell techniques have taught us on how CPCs can build a functional heart, with a special focus on clonal analyses and single-cell transcriptomics.

Clonal Analyses as Tools to Investigate Cardiac Progenitor-Derived Lineages

The first description of the two sources of cardiac progenitors was based on a mouse cell lineage study using β-galactosidase labelling in a retrospective clonal analysis [4, 11••]. This approach relies on a rare and random recombination event in the αcardiac-actin-nlaacZ cassette that reverts the non-functional nlaacZ cassette into a functional nlacZ gene that will be transmitted in all descendant cells. The spatial distribution of clonally related β-gal positive cells at E8.5 revealed two categories of clones with distinct patterns of regionalization in the heart tube [11••]. One lineage contributes to LV, atrioventricular canal (AVC), and the atria, while the other contributes to the OFT and all other heart regions, except the LV. This second lineage corresponds to the contribution of the SHF, in which the transcription factor Isl1 (LIM homeodomain transcription factor Islet1) is expressed [12]. This retrospective clonal analysis demonstrated that the first and second lineages segregate from a common progenitor at the onset of gastrulation. Interestingly, SHF cells also give rise to multiple cell types including CMs, ECs, SMCs, and skeletal muscle cells that contribute to the heart, pharyngeal arches, and head/neck muscle cells [13, 14]. Indeed, retrospective clonal analysis performed at E14.5 showed that masticatory muscles, which derive from the first pharyngeal arch, share a common progenitor with the RV myocardium, whereas facial muscles deriving from the second pharyngeal arch are clonally related to myocardium at the base of the great arteries [14]. Moreover, non-somite-derived skeletal muscles of the neck, which derive from the most posterior pharyngeal arches, showed lineage relationship with the myocardium at the venous pole of the heart [13]. These findings revealed a lineage relationship between neck/head muscles and SHF myocardial derivatives with left and right segregation very early during embryogenesis (Fig. 1a).

At E7.5, the first differentiated cells in the cardiac crescent are marked by Tbx5 and Hcn4. Genetic or lineage tracing experiments, based on the use of a recombinase (Cre or Dre, for example) under the regulatory sequences of a gene marker and a conditional reporter that will be activated when the marker is expressed, are particularly helpful to trace cells that express a marker of interest [15]. Studies using Tbx5CreERT2 and Hcn4CreERT2 transgenic mice confirmed that FHF derivatives contribute to cardiomyocyte in the LV and parts of the atria [16, 17]. Lineage tracing experiments using Mesp1-Cre mice showed that all cardiac cells, including derivatives of the FHF and SHF, derive from CPCs expressing Mesp1 [2, 18••]. Similarly, the study of Flk1-Cre mouse line, which marks early mesodermal cells, showed that Flk1 + CPCs contribute to both CMs and ECs development [19, 20]. In vitro experiments using mouse or human embryonic stem cells (ESCs) have further validated the multipotentiality of CPCs. Flk1 + or Mesp1 + CPCs were able to differentiate into CMs, SMCs and ECs [21–24] and even to skeletal muscles [25]. However, all these experiments did not rule out whether a single CPC, in its niche, has the ability to contribute to multiple cardiac lineages. Only clonal analysis can address whether there is collective or individual multipotency, and whether FHF and SHF progenitor cells arise from a common Flk1 + /Mesp1 + progenitor. Thus, Lescroart et al. (2014) used clonal analysis of the earliest prospective CPCs in a temporal controlled manner (using tetracycline-inducible Mesp1-rtTA transgenic mouse with low dose induction) during early gastrulation to show that Mesp1 progenitors consist of two temporally distinct pools of progenitors restricted to either the FHF or SHF. Indeed, FHF and SHF cells correspond to the early and late Mesp1 + populations, respectively [18••]. The clonal analysis of single Mesp1 + CPCs revealed that FHF progenitors are unipotent since they are restricted to either CMs or ECs fates at the time of their specification (Fig. 1a). In contrast, Mesp1-derived SHF progenitors can be unipotent or bipotent (Fig. 1a). Interestingly, the ultimate fate of the progenitors can be regulated by the environmental factors that the different progenitors encounter during cardiac morphogenesis. Devine et al. (2014) used in vivo clonal analysis, by generating mosaic mice in which very few Mesp1 + cells were labeled, to confirm that during gastrulation, CPCs are already split into two distinct populations. Results from this study also suggest that these two populations are separated by a boundary that is established very early during development to form the septum that separates the LV and RV in the matured heart. A recent study addressed the location of distinct FHF and SHF progenitors in the early gastrulating embryos. This study used Hand1CreERT2 mice to perform a lineage tracing analysis and revealed a Hand1-expressing population at the boundary of the extraembryonic and intraembryonic regions of the gastrulating embryo that contributes to FHF derivatives [26••]. Interestingly, these Hand1-labeled cardiomyocytes reside largely within the dorsal region of the LV and AVC. Since LV cardiomyocytes have been shown to derive from the FHF, these results suggest that cells marked by Hand1 represent a subset of the FHF. Zhang et al. (2021) used Hand1CreERT2 and Rosa26-Confetti mice to perform clonal analysis and showed that this subset of FHF-derived cardiomyocytes derive from a common multipotent Hand1 + progenitor, which also contributes to the proepicardium (Fig. 1a).

Clonal analysis was also used to trace other cardiac cell lineages during early heart development. Red-Horse et al. (2010) used clonal analysis to investigate the origins and early development of coronary arteries. This study showed that coronary arteries develop from endothelial sprouts of the venous inflow tract of the forming heart [27]. Clonal analysis was also used to study the lineage relationships between non-venous SHF-derived cells at the base of the great arteries and the cardiac lymphatic vasculature. Lioux et al. performed a random, lineage-unrestricted clonal analysis of the developing heart to explore all lineage relationships in an unbiased manner. This study characterized the fate of the SHF progenitors at the base of the great arteries and found a contribution of SHF cells to the coronary lymphatic vasculature. This analysis identified a shared lineage for arterial mesothelial and sub-mesothelial cells and cells derived from the SHF like right ventricular cardiomyocytes, valve mesenchyme, and coronary ECs [28•].

The cardiac conduction system (CCS) coordinates atrial and ventricular contraction via generation and propagation of electrical impulses. The ventricular conduction system (VCS) is composed of central components (the atrioventricular node, bundle, and right and left bundle branches) and a peripheral Purkinje fiber network. Retrospective clonal analysis combined with genetic tracing experiments using Connexin40-GFP (Cx40-GFP) mice, that delineate the VCS, revealed a dual contribution of FHF and SHF progenitor cells during the formation of the VCS [29]. Clonal analysis using SmaCreERT2 and Rosa26-Confetti mice highlighted the early segregation of the VCS lineage within SMA + cardiomyocytes [30]. This study using SMA prospective clonal analysis suggested also that the atrioventricular bundle is specified earlier than the bundle branches or during a more restricted time window. These findings are in accordance with the retrospective clonal analysis indicating the existence of early common progenitors for conductive and working cardiomyocytes [29]. These studies suggested that clonal analysis is a strategy more appropriate to distinguish lineage segregation. More recently, the same group used Cx40CreERT2 and Rosa26-Confetti mice as a temporal clonal analysis of CPCs to identify distinct phases of peripheral Purkinje fiber network formation [31]. In addition, this study demonstrated the existence of bipotent progenitors that participate in the early and late phase of CSS development.

Photoconvertible assay was used in zebrafish as a clonal analysis to quantify the dynamics and regulation of myocardial cell addition during heart development. Similar to murine heart growth, Lazic and Scott used photoconvertible fluorescent protein to reveal gradual myocardial addition at the arterial pole of the zebrafish heart between 24 h post-fertilization (hpf) and 34hpf [32]. This study showed that the cells contributing to the early heart tube, and the later addition to the arterial pole, are derived from the same pre-gastrula region of the embryo, suggesting a shared progenitor. This result was confirmed by a lineage tracing analysis showing that latent TGF-β binding protein 3 (ltbp3) transcripts mark a field of CPCs with defining characteristics of the SHF in mammals [33].

In addition to their contribution to lineage inference, clonal analyses have also been determinant for investigating cell behavior, such as CPC’s mode of growth [34–37]. Ivanovitch and colleagues have also elegantly tracked single CPCs during heart morphogenesis and showed the dynamics of differentiation of FHF and SHF progenitors [38••]. Using live imaging of ex vivo cultured mouse embryos, they could demonstrate three phases during heart morphogenesis: First, FHF progenitors rapidly differentiate to form the cardiac crescent, then cardiac differentiation slows down while important morphogenesis processes occur, and finally SHF progenitors differentiate and participate in the heart tube closure.

All these studies have contributed to a better understanding of CPC specification and deployment. Clonal analyses have thus been really determinant to reconstruct the lineage history of CPCs and their contribution to the heart (Fig. 1a). These findings also revealed a high heterogeneity in CPCs. However, the molecular characteristics, such as gene expression and gene regulatory networks, cannot be addressed with classical clonal analysis. The recent development of single-cell transcriptomics has opened new avenues to address these questions.

Fig. 1
figure 1

Clonal analyses and scRNAseq in the early stages of mouse heart development. a Model of the early steps of CPC specification and lineage commitment during mouse development as shown by clonal analyses. Clonal analysis of Mesp1 + and Hand1 + progenitors showed the existence of temporally distinct Mesp1 progenitors that contribute to different regions (top) and to different cell lineages (bottom). Early CPCs are either bi/multipotent (dark black circles) or unipotent. Early Mesp1 progenitors (labelled at E6.75, red) contribute preferentially to the LV, RV while late Mesp1 progenitors contribute to the RV, RA, LA, inflow region (RSCV, LSCV, and PV) and to some skeletal muscles of the head or neck. LV, left ventricle; AVC, atrioventricular canal; RV, right ventricle; pt, pulmonary trunk; ao, aorta; LSCV, left superior caval vein; RSCV, right superior caval vein; PV, pulmonary vein; fac. Exp. Muscles, facial expression muscles; Epi, epicardium; Peri, pericardium; CM, cardiomyocyte; EC, endothelium/endocardium; SMC, smooth muscle cell; skM, skeletal muscle. b In scRNA-seq experiments of Mesp1 + cells at E6.75 and E7.25, four distinct Mesp1 + subpopulations have been identified that emerge from epiblast cells and that correspond to the endothelial (EC, green), cardiomyocyte (CM, red), anterior SHF (aSHF, pink), and posterior SHF (pSHF, purple) progenitors. These populations are enriched for some transcripts (in brackets). On the right panel, map of the gastrulating embryo, where the different CPC subpopulations are found. Mesp1 + EC population (green) is found in the outer layer of the mesodermal wing, while Mesp1 + CM populations are found close to the epiblast. Populations of the FHF (red), aSHF (pink), and pSHF (purple) are represented along the antero-medial axis of the embryo. This summarizes the results from Lescroart et al. and Ivanovitch et al. [54••, 55••]. PS, primitive streak; A, anterior, P, posterior. c The left panel represents the scRNA-seq experiments of CPCs from E7.75 to E8 by Tyser et al. [57••]. Six distinct cardiac subpopulations have been identified (Me3–Me8). These populations are enriched for different transcripts (in brackets). JCF, juxta-cardiac field; CrM, cranial mesoderm. The trajectories identified among these cell clusters are shown in the middle. These subpopulations have been mapped to the developing embryo at the cardiac crescent stage (left panel). L, left; R, right. d The left panel represents the scRNA-seq experiment of Mesp1-derived cardiac cells from E7.25 to E8.25 by Zhang et al. [26••]. Seven distinct cardiac subpopulations have been identified (cardiomyocytes or CM1-3 and cardiac progenitors or CP4-7). Lineage inference (in the middle) can link the different clusters in an inferred lineage tree. LEM, late extra-embryonic mesoderm; Al, allantois. These subpopulations have been mapped to the embryo at an early heart tube stage (right panel). D, dorsal; V, ventral

Full size image

Single-Cell Transcriptomics as a Novel Platform to Investigate CPC Specification and Differentiation at a Single Cell Resolution

While bulk transcriptomic analyses provide an average gene expression, ignoring cell heterogeneity and stochasticity of gene expression, transcriptomics at the single cell level is a powerful approach for characterizing individual cells. Analyzing the transcriptome in single cells indeed allows the description of tissue cell heterogeneity, the identification of rare cell types, and comparison of cell heterogeneity in healthy and pathological contexts. In addition, new tools now allow elucidation of dynamics transition states during development [39], to some extent to infer lineage trajectories [40] and to hypothesize upon cell–cell communications via ligand-receptor binding [41, 42]. Protocols are detailed in a review by Kolodziejczyk et al. [44].

Single cell transcriptomics has started to be used to investigate early cardiac specification in the early 2010s. To explore the potential of Mesp1 + early CPCs in vitro, Bondue et al. have performed single-cell PCR using key marker genes of the different cardiovascular lineages (Pecam1 for ECs, Acta2 for SMCs, Tnnt2 for CMs, Isl1 as a pan SHF marker, and Tbx5 for the FHF) and showed the pluripotency of single Mesp1 + ESC-derived CPCs [21]. Similarly, in vivo, cell sorting of Mesp1 + CPCs at gastrulation, followed by single-cell PCR, have demonstrated the molecular heterogeneity of CPCs and supported the results of uni- and bipotency revealed by clonal analysis [18••]. With the development of techniques for single cell encapsulation, Li et al. have then further investigated and compared in vitro and in vivo cardiac specification using a microfluidic-enabled multiplex PCR platform. They showed the utility of single cell transcriptome profiling to study cardiac specification and cell fate choices [46•]. These experiments showed, in particular, that ESC-derived cardiac lineages are close to cells isolated from the embryonic heart. This approach has been also used to describe the molecular heterogeneity of cells in the heart tube and has allowed the definition of the different gene expression profiles of different subpopulations in the early heart tube [47]. In a parallel set of experiments, Delaughter and colleagues have used the fluidigm platform to perform single cell RNA sequencing (scRNAseq) of cardiac cells from E9.5 to the neonatal stage [48]. scRNAseq has, this time, allowed a non-biased study of the whole transcriptome. These studies have provided an atlas of the transcription expression profile of all major embryonic cardiac cell types, which is fundamental and represent the basis for further studies on progenitor cell fate decisions toward these different characterized cardiac cell types. In addition, atlases have been generated for different embryonic stages in the mouse, focusing on different cardiac cell types [49,50,51], and for other species such as tunicates or even human [52, 53] (Table 1).

Table 1 scRNAseq analyses of CPCs specification
Full size table

Further scRNAseq studies performed on CPCs have revealed the transcriptional heterogeneity of these cells during early heart development, as well as, at later time points within SHF progenitors. The analysis of early Mesp1-expressing cells at E6.25 and E7.25 revealed a relative heterogeneity among CPCs, at the time of gastrulation, with a continuum of cells to at least 4 distinct cardiac-related trajectories [54••]. These 4 final subpopulations likely correspond to ECs (marked by the expression of Sox17 and Etv2), CMs (Hand1 + , Bmp4 + , Tnnt2 +) and to anterior (Tbx1 + , Foxc2 +) and posterior SHF progenitors (Wnt2b + , Hoxb1 + , Hoxa1 +) (Fig. 1b). These results have further demonstrated the heterogeneity of the CPCs, reflecting their different lineage and regional contributions to the adult heart and suggested that CPCs segregate early in the different lineages. Furthermore, high-resolution single molecule fluorescent in situ hybridization (smFISH) has shown that markers of the 4 Mesp1 + subpopulations were expressed in specific locations in the embryo. CMs progenitors are, for example, located close to the epiblast, while EC progenitors are located in the vicinity of the visceral endoderm (Fig. 1b). These findings were further confirmed by Ivanovitch et al., with recovery of clusters for FHF and anterior and posterior SHF from scRNAseq of E7.5 embryos, which could be further subdivided into subclusters that likely correspond to different cardiac regions [55••]. The combination of lineage tracing with scRNAseq has allowed to conclude that FHF and anterior and posterior SHF derive from distinct groups of cells, with different transcriptional programs, and are found in particular spatial and temporal regions of the primitive streak at gastrulation [55••]. At slightly later time points, de Soysa et al. have shown that the Isl1 + CPCs could be divided into at least 3 subclusters with distinct transcriptional profiles: the anterior and posterior SHF, as well as the head muscle progenitors [56••]. This conclusion is consistent with findings obtained by clonal analysis experiments. Cells from the heart tube could be also subdivided into different transcriptional clusters, representative of the LV and RV, atrial, sinus venosus, AVC, and OFT myocardium, confirming previous reports on the significant heterogeneity in the transcriptional profile of cardiomyocytes in the heart depending on their location [47, 48, 56••].

As single-cell transcriptomics allow the analysis of pseudotime and the reconstruction of lineage trajectories, lessons on cardiac lineage contribution could be learnt from scRNAseq of CPCs. With such approaches, two recent scRNAseq studies have recently highlighted the contribution of a population of CPCs that expresses Mesp1 early and is marked by the expression of Hand1 and Mab2l12. This population contributes specifically to the LV and was called the JuxtaCardiac Field (JCF), likely a subdivision of the FHF [26••, 57••]. In their study, Tyser and colleagues showed that, at early stages of heart development (between E7.75 and E8.5), heart derivatives could be divided into 5 clusters with specific transcriptional signatures. Analysis of the transcriptional trajectories showed that differentiated cardiomyocytes could be produced though transition states corresponding to SHF and FHF progenitors, as well as through a population expressing a newly identified marker, Mab2l21 [57••] (Fig. 1c). Zhang and colleagues have applied lineage inference (URD) to reconstruct lineages from scRNAseq of Mesp1-derived cells between E7.25 and E8.25. This analysis revealed a close relationship between clusters defined as late extraembryonic mesoderm (LEM) and lateral plate mesoderm and a common contribution to a cluster of CMs. The LEM cluster is characterized by high Hand1 expression and Mab2l12 is expressed along the trajectory from this cluster toward differentiated CMs [26••] (Fig. 1d). Lineage tracing of Hand1 and Mab2l12 derived cells showed that this population contributes to the epicardium as well as to the CMs of the LV and AVC. Once again, with high-resolution smFISH, the clusters defined by scRNAseq were mapped to the embryo, showing that the JCF or Hand1-derived Mab2l12 + population is found at the interface between the cardiac crescent and the extraembryonic tissue [26••, 57••] (Fig. 1c–d). These results are in total agreement with temporal Mesp1 clonal analyses, which showed that early Mesp1 CPCs contribute to the epicardium and CMs of the LV [18••]. To investigate the multipotency of CPCs, other groups have analyzed the transcriptome of single Nkx2-5 + vs Isl1 + progenitors between E7.75 and E9.5. Trajectory inference suggested that Nkx2-5-expressing cells were committed to a unidirectional CM fate, whereas Isl1-expressing CPCs could diverge into multiple developmental branches toward a CM or EC fate, with evidences for an intermediate cell state [58•]. However, in another study, transcriptional profiling of Nkx2-5-derived CPCs has showed that the Nkx2-5 lineage seems to have a lineage trajectory toward both CM and ECs [59]. The discrepancies between these analyses could be due to the difference between Nkx2-5-derived CPCs and Nkx2-5-expressing CPCs or might exemplify the limits of lineage inference using single-cell transcriptomics that will be further discussed later in this review.

Single cell transcriptomics of control versus mutant cells also offers a way to identify genes involved in cardiac specification or in lineage bifurcation. Those analyses have confirmed, for example, that Mesp1/Mesp2 are critical for the exit of pluripotency and for early cardiac specification as cells with decreased level of Mesp1/Mesp2 showed a transcriptional profile similar to epiblast cells [54••]. Computational methods can also be used to predict lineage specifiers from scRNAseq. With such approach, De Soysa et al. have identified Hand2 as a potential lineage specifier with important role for OFT but not RV development, despite the failure of RV formation in Hand2 null mutant mice [60]. Upon loss of Hand2, the cluster of OFT myocardial cells disappeared showing the critical role of Hand2 for OFT specification [56••]. Using scRNAseq during human ESC differentiation and in human fetal hearts, LGR5 has been identified as a human-specific OFT myocardium marker and is expressed together with MESP1 in some clusters. Further analysis in hESC has showed that LGR5 is required for cardiac specification through expansion of the ISL1 + TNNT2 + intermediates [61]. Different labs have similarly shown that Nkx2-5 is essential for progenitor specification/maturation toward a CM fate [47, 48, 58•] but also that Nkx2-5 is required for the proper expression of Cxcr4 and Cxcr2, thus regulating SHF progenitor migration [59]. Using scRNAseq, Isl1 has been shown to be critical for cardiac specification, as Isl1-knock-out cells appeared to be stuck in a transition state, unable to differentiate into the CM or EC lineages in vivo [58•]. However, profiling of ISL1 knock-out hESC-derived cells showed that ISL1 KO cells are redirected toward a neural program, but that CM specification is not affected in such model [62]. This can potentially reveal discrepancies between mouse and human or between the in vivo and in vitro situation. In silico analysis of scRNAseq data can also allow the identification of transcription factors that are involved in CPC specification. Friedman et al. have used a probabilistic method for constructing regulatory networks from scRNAseq during iPSC (induced pluripotent stem cell) cardiac differentiation. They showed different bifurcation points with a segregation between definitive endoderm and mesendoderm lineages and another bifurcation from a CPC state toward a CM lineage and a non-muscular lineage. Their in silico analyses notably highlighted a role for YY1 in the transition toward a CM fate and PBX1 for non-muscular cardiac fate [63].

Other single cell sequencing approaches are available and can be used to investigate CPCs specification at the single cell level. Single-cell ATACseq (scATAC-seq), for example, is designed to profile chromatin accessibility and represents an alternative way to investigate progenitor cell heterogeneity. Jia et al. have shown that scATAC-seq is indeed a powerful tool which is equal or superior to scRNAseq for revealing cell heterogeneity [58•]. Additionally, profiling chromatin accessibility can allow the investigation of potential molecular mechanisms for lineage segregation. With such analysis, Jia et al. could find the enrichment of particular transcription factor motif accessibility along the trajectories from an Isl1 + CPC state toward a CM or EC fate. The FOX motif seems, for example, more accessible in the CPC state, whereas TBX5/GATA/HOX/HAND motifs are closely associated with a CM fate and SOX motifs with an EC fate, thus indicating some potential transcription factors critical for cardiac cell fate decisions.

It is important to note that scRNAseq approaches alone are not sufficient to draw conclusions. Most of the studies published and discussed in this review have required additional validations, using lineage tracing [26••, 54••, 55••, 57••], clonal analysis [26••], and single cell resolution smFISH [26••, 54••, 56••, 57••] (see also Table 1). One main caveat of scRNAseq is the loss of spatial information, ultimately linked to the cell dissociation protocols that disrupt cell–cell communication. In silico tools have been developed to compute and hypothetize potential receptor-ligand pairs between cell populations. Xiong et al. have performed such kind of in silico single-cell receptor-ligand pairing screen and found an interesting ligand-receptor paring between CM, expressing the ligand Mif, and SHF progenitors that express its receptor Cxcr2. These results, further validated by in situ hybridization and functional assays, have showed a role for FHF CM in the guidance of SHF progenitor migration through the MIF/Cxcr2 chemotaxis [59]. While Li et al. have identified, with scRNAseq, the AVC and CM trabeculae region as cardiac regions with reduced proliferation activity, they also applied ligand-receptor pairing to find signal pathways that could drive activation or repression of proliferation and found in silico that the expression of Tgfß1 from the endocardium and Rspo1 from the epicardium might control the proliferation of cardiomyocytes [64]. Spatial transcriptomics are alternative approaches that are now emerging quite fast with different degrees of resolution [65]. So far, this technique has not been used for early stages of heart development but only at fetal human stages [66•]. This approach opens new avenues and could be really important to resolve questions regarding the spatial molecular heterogeneity of CPCs. While the antero-posterior heterogeneity has been well described [56••, 67], it would be really interesting to further investigate, for example, the differences that are found along the left–right axis of the embryo.

Single cell sequencing approaches have thus significantly increased our understanding of CPCs specification. It has demonstrated the high transcriptomic heterogeneity of CPCs and has provided some insights into the mechanisms that lead to cardiac specification and lineage specification.

Conclusion and Perspectives

Clonal analyses and single cell transcriptomic approaches have redefined the field and have clearly provided a deeper understanding of CPCs specification at higher resolution. In some studies, some discrepancies could however be found between clonal analysis and scRNAseq or between the trajectories inferred from different scRNAseq studies. These discrepancies highlight some caveats of trajectories directly deduced from scRNAseq analyses. For example, Weinreb and colleagues have already discussed the limitations of inferring trajectories from snapshots of scRNAseq [68]. However, new tools have been recently developed to reconstruct lineage history from single cell sequencing, using CRISPR/Cas9 genomic editing or recombination for generation of lineage barcodes [40, 69]. If applied to CPCs, these approaches could help reconstruct the full lineage history of CPCs to investigate if early CPCs, at gastrulation, have a broader contribution than what is accepted and to investigate the time of lineage segregation, which is still not completely resolved. Recent techniques allow recording of both the transcriptome and lineage history of single cells, as it has been performed for example in the hematopoietic system [70, 71].

Single cell multi-omics approaches in general are emerging very fast [72], and these will allow, in combination, future investigations of the transcriptome (scRNAseq), chromatin accessibility (scATCseq), protein expression (CITE-seq), or the methylome (Me-seq) of single CPCs. This should help uncover the molecular mechanisms driving CPC specification and lineage segregation and ultimately to better understand CHDs.