Cardiopharyngeal Progenitor Specification: Multiple Roads to the Heart and Head Muscles

Abstract

During embryonic development, the heart arises from various sources of undifferentiated mesodermal progenitors, with an additional contribution from ectodermal neural crest cells. Mesodermal cardiac progenitors are plastic and multipotent, but are nevertheless specified to a precise heart region and cell type very early during development. Recent findings have defined both this lineage plasticity and early commitment of cardiac progenitors, using a combination of single-cell and population analyses. In this review, we discuss several aspects of cardiac progenitor specification. We discuss their markers, fate potential in vitro and in vivo, early segregation and commitment, and also intrinsic and extrinsic cues regulating lineage restriction from multipotency to a specific cell type of the heart. Finally, we also discuss the subdivisions of the cardiopharyngeal field, and the shared origins of the heart with other mesodermal derivatives, including head and neck muscles.

Throughout the animal kingdom, organogenesis takes place early during embryonic development. This developmental process consists of the formation of organs through the progressive commitment and specification of progenitor cells, their migration toward the final location of the organ, and their signaling interactions with surrounding tissues. Organogenesis starts with the commitment of multipotent and plastic progenitors that will progressively acquire the molecular program of more restricted progenitors and finally differentiate into mature functional cells.

The heart is the first organ to form in mammals and arises from the specification of cardiac progenitor cells, which can be identified as early as during gastrulation (Kinder et al. 1999) and which represent the building blocks of cardiac development. Cardiac progenitors are included in a larger mesodermal population termed the cardiopharyngeal field (CPF) that contributes to both the heart and skeletal muscles of the head and neck (Diogo et al. 2015). In vertebrates, the first specified cardiac cells generally express Nkx2-5 (Harvey 2002), which notably also marks certain skeletal muscle progenitors of the head (Diogo et al. 2015). Nevertheless, cardiopharyngeal progenitor cells (CPCs) can be identified earlier during development, before the onset of Nkx2-5 expression, and there are now several lines of evidence showing that the fate of CPCs is specified during or prior to gastrulation.

In mammals, cardiac progenitors will build a four-chambered heart with two atria and two ventricles, consisting of different cell types including cardiomyocytes (CMs), endocardial cells (ECs) forming the inner layer, epicardial-derived cells (EPDCs) at the outer layer, and smooth muscle cells (SMCs) (Lescroart and Meilhac 2012). It is now well established that vertebrate cardiogenesis occurs through progressive addition of several populations of progenitors to the forming linear heart tube. An early differentiating population of progenitors, termed the first heart field (FHF), gives rise to the left ventricle (LV) and some atrial cells, whereas a second wave of differentiation occurs from the second heart field (SHF) with a distinct contribution to the four-chambered heart, including the right ventricle (RV), the outflow tract (OFT), and the venous pole (Buckingham et al. 2005). Further subdivisions of the SHF distinguish the contribution of the anterior SHF (aSHF) to the OFT and RV from that of the posterior SHF (pSHF) to the OFT and venous region (Meilhac and Buckingham 2018). Subdivision of the FHF has not been clearly established, but, based on lineage tracing (Bardot et al. 2017), we could wonder whether the FHF can also be patterned along the anteroposterior axis.

Throughout the evolution of vertebrates, the heart has acquired a complex multichambered structure, although conserved FHF and SHF populations can already be identified in tunicates, which are the closest living relatives of an early chordate precursor. Therefore, CPC specification has been studied in different models, ranging from the simple chordate Ciona robusta, which presents a simple beating heart tube, to the mammalian mouse model. Moreover, differentiation of mouse and human pluripotent stem cells (m- or hPSCs) in vitro into different cardiac lineages has been shown to recapitulate the first steps of CPC specification, including the molecular pathways involved in this process, and are thus an attractive model to determine the minimal cues necessary for the formation of specific cardiac lineages.

In this review, we present currently known markers of early CPCs and their differentiation potential as studied by single-cell approaches. We also discuss insights into the regulation of CPC specification toward specific cardiac or pharyngeal lineages that have recently been discovered through the multiplication of genetic techniques and reduction in costs of next-generation sequencing. Although cardiac neural crest cells also make important contributions to the OFT and valves (Kirby et al. 1983; Kirby and Waldo 1995), we choose here to focus only on the mesodermal contribution to the forming heart.

MARKERS OF EARLY CPCs

Several key studies have isolated and characterized CPCs and more specifically cardiac progenitors from mouse embryos and differentiating PSCs, which are capable of proliferating and giving rise to multiple cell types of the heart.

As the majority of cardiac and pharyngeal muscle cells derive from the mesoderm, genes that are expressed in the nascent mesoderm such as Brachyury, also called T, encoding a T-box transcription factor (TF) (Kispert and Herrmann 1993; Showell et al. 2004) and Mesp1 (mesoderm posterior 1) encoding for a basic helix–loop–helix (bHLH) TF (Saga et al. 1996), are characteristic of early CPCs. Mesp1 is thereafter rapidly down-regulated once the mesoderm is formed. During differentiation of both m- and hPSCs, the expression of Mesp1 is also transient and precedes the expression of other cardiac markers (Liu et al. 2007; Ueno et al. 2007; Bondue et al. 2008; Lindsley et al. 2008; Den Hartogh et al. 2015). Mesp1 has spurred great interest in the last two decades as a potential marker of early cardiac progenitor specification of both heart fields (Bondue and Blanpain 2010). Moreover, in the chordate Ciona, Mesp, the unique homolog of Mesp1, is expressed transiently in the B7.5 cells pair, which will give rise to the Ciona heart, supporting a conserved role of Mesp genes in CPC specification throughout chordates (Satou et al. 2004).

Specific combinations of surface markers expressed in CPCs can now facilitate their isolation, purification, and characterization. Flk1 (or Vegfr2) and Pdgfra are both expressed in nascent mesoderm (Kataoka et al. 1997; Motoike et al. 2003; Ema et al. 2006). Interestingly, cells coexpressing Flk1 and Pdgfra during differentiation of PSCs are enriched for cardiac potential (Hirata et al. 2007). Cxcr4 has also been used in combination with Flk1 (Nelson et al. 2008) or both Flk1 and Pdgfra (Bondue et al. 2011) to isolate CPCs with cardiac potential. The expression of these three signaling receptors coincides with that of Mesp1, and overexpression of Mesp1 directly activates their transcription (Bondue et al. 2011). Flk1/Pdgfra/Cxcr4 also mark early Mesp1⁺ CPCs in vivo (Lescroart et al. 2014). hPSCs undergoing differentiation into cardiac cell types also necessarily pass through an Flk1⁺ mesodermal stage (Yang et al. 2008). Expression of Flk1 and Pdgfra together is generally predictive of CM potential in hPSCs (Kattman et al. 2011). Several other markers have recently been proposed for the isolation of hPSC-derived cardiac progenitors, such as Ror2 and Cd13 (Drukker et al. 2012; Den Hartogh et al. 2015; Skelton et al. 2016), or Fzd4 (Yoon et al. 2018).

One subunit of a chromatin-remodeling complex, Smarcd3, has cardiac-specific expression. Smarcd3 is expressed in a subset of Mesp1⁺ CPCs, and a smaller 2.5-kb regulatory region of this gene can activate reporter gene expression in progenitors for both heart fields in mouse embryos (Devine et al. 2014). Interestingly, the same regulatory region marks zebrafish CPCs, even though this particular sequence was not conserved between both species (Yuan et al. 2018). This suggests that the same core transcriptional regulators are active in CPCs throughout vertebrates and can induce the expression of CPC-specific regulatory regions from other species. As regulatory regions of genes have readily been used to activate reporter gene expression in specific populations of CPCs for the whole heart (Wu et al. 2006) or of the SHF (Verzi et al. 2005), the advent of whole-genome predictions of cardiac enhancers and validation of their activity in vivo (He et al. 2011; Yuan et al. 2018) may accelerate the identification of cardiac-specific enhancers expressed as early as during gastrulation in the FHF, SHF, or whole CPF. It would then be interesting to investigate whether these enhancers are specifically active in multipotent pancardiac CPCs or rather independently activated in different lineages.

LINEAGE POTENTIAL OF CPCs

Although in Ciona the cardiac lineage tree is highly ordered and reproducible, starting from only two Mesp⁺ progenitors at the time of gastrulation (Satou et al. 2004), lineage diversification is more complex and plastic in vertebrates, with about 250 Mesp1⁺ progenitors contributing to the heart in the mouse model (Chabab et al. 2016).

The first experiments assessing the lineage potential of CPCs were performed in the chick and the mouse, using dye labeling or grafting experiments (Stalsberg and DeHaan 1969; Tam and Behringer 1997; Kinder et al. 1999; Lawson 1999; Tzahor et al. 2003). Grafting experiments have shown that cardiac progenitors are committed to a cardiac fate only after the formation of the cardiac crescent (Tam and Behringer 1997; Auda-Boucher et al. 2000), suggesting a certain degree of plasticity of CPCs at gastrulation stages.

Lineage tracing experiments, clonal analyses, and, more recently, single-cell RNA-sequencing (scRNA-seq) have allowed us to decipher the appearance of different cell lineages that arise from early CPCs.

Lineage Tracing and In Vitro Differentiation Demonstrates the Multipotency of CPCs

Lineage tracing experiments with a Flk1-Cre knockin mouse model showed that although originally involved in hematopoietic lineages, Flk1⁺ cells also contribute to the development of CMs and ECs (Motoike et al. 2003; Ema et al. 2006). Furthermore, during differentiation of PSCs, T⁺/Flk1⁺ cells present the ability to form colonies comprising CMs, SMCs, and ECs, suggesting that this early mesodermal cell population is at least tripotent (Kattman et al. 2006). Tracking the progeny of Mesp1-expressing cells revealed that they comprise the majority of cardiac cells including CM, ECs, and EPDCs (Saga et al. 1999). In addition, the Mesp1 lineage also contributes to noncardiac derivatives, including skeletal muscles of the head and neck (see Table 1; McBratney-Owen et al. 2008; Yoshida et al. 2008; Asahina et al. 2009; Harel et al. 2009). Interestingly, this observation seems consistent with findings in Ciona, where Mesp⁺ cells also produce the anterior and tail muscles, indicating their multipotency (Fig. 1; Stolfi et al. 2010). During mouse and hPSC differentiation, isolated Mesp1-expressing cells have the potential to become ECs, SMCs, and CMs in a clonal assay (Bondue et al. 2011; Den Hartogh et al. 2015). Furthermore, an in vitro clonal assay using mPSCs demonstrated the bipotential of Mesp1⁺ cells to generate skeletal and cardiac muscles, in agreement with their broad fate potential in vivo (Chan et al. 2016).

Table 1.

Lineage tracing in mouse has revealed the existence of different early CPC subpopulations

CPC subpopulation	Genetic construct used	Labeled cardiac regions	Labeled cardiac cell types	Noncardiac derivatives	Reference(s)
Mesp1+	Mesp1-Cre; R26R	OFT, RV, LV, RA, LA, IFT	SMCs, ECs, CMs, EPDCs	Some head bones and head muscles, dorsal aorta, intersomitic and cranial vessels, hematopoietic progenitors, amion, mesothelial cells of the liver	Saga et al. 1999; McBratney-Owen et al. 2008; Yoshida et al. 2008; Asahina et al. 2009; Harel et al. 2009; Lescroart et al. 2014
Smarcd3+	Smarcd3-F6-CreERT2; R26R (Tam E6.5)	OFT, RV, LV, RA, LA	ECs and CMs	Anterior forelimb	Devine et al. 2014
Foxa2+	Foxa2-Cre; R26R-tdTomato	RV, LV	CMs, ECs, EPDCs	Endoderm derivatives	Bardot et al. 2017
Mef2c-AHF-enh+	Mef2c-AHF-enh-Cre; R26R	OFT, RV, IS	CMs, ECs, SMCs	Head muscles	Verzi et al. 2005; Lescroart et al. 2010; Devine et al. 2014
Hcn4+	Hcn4-CreERT2; R26R (Tam E6.-E7.0)	LV	CMs	Cardiac specific	Liang et al. 2013
Hand1+	Hand1-Cre; R26R	OFT, LV	CMs, EPDCs	Postmigratory NCCs, forelimb and hindlimb mesenchyme	Barnes et al. 2010
Sfrp5+	Sfrp5-Cre; R26R	OFT, LV, RA, LA	CMs, ECs, EPDCs	Not determined	Fujii et al. 2017
Six2+	Six2-Cre-ERT2; R26R	OFT, RV	SMCs, CMs	Head muscles, urogenital tissues	Zhou et al. 2017
Hoxb1+	Hoxb1-Cre; R26R	OFT, RA, LA	CMs, ECs	Large contribution to the endoderm and mesoderm	Bertrand et al. 2011

Summary of the results obtained after lineage tracing of different CPC subpopulations.

(CPC) Cardiopharyngeal progenitor cell, (OFT) outflow tract, (RV) right ventricle, (LV) left ventricle, (RA) right atrium, (LA) left atrium, (IFT) inflow tract, (CMs) cardiomyocytes, (ECs) endothelial/endocardial cells, (SMCs) smooth muscle cells, (EPDCs) epicardium-derived cells, (NCC) neural crest cell, (IS) interventricular septum.

Figure 1.

Steps from cardiopharyngeal progenitor cells (CPCs) to cardiac and skeletal muscle cell lineages: comparison between tunicates and mouse. (A) The Ciona CPC lineage tree is found on top. Two Mesp⁺ CPCs (green) are found at ∼4 hours postfertilization (hpf). After two cell divisions, these CPCs form four anterior tail muscle (ATM) cells and four multipotent trunk ventral cells (TVCs). These TVCs divide asymmetrically to form a second TVC (STVC) and a first heart precursor ([FHP], red). The STVC (dark gray) divides to form an atrial siphon muscle ([ASM], blue) progenitor cell, analogous to the vertebrate head muscles and a second heart precursor ([SHP], orange). A subset of the lineage branches is depicted from a single Mesp CPC, the other branches being identical in sequence and outcome. (Below) Layout of the tunicate larva. Black lines indicate sister cells. Time after fertilization is indicated in hpf. At ∼28 hpf, ASM progenitor derivatives form an ASM ring, whereas FHP- and SHP-derived cells form the developing heart, including pericardial cells and cardiomyocytes. (B) The mouse CPC lineage. On top, early CPCs (in green) are found in the gastrulating embryo and are marked by the expression of Mesp1, Flk1, Pdgfra, and Cxcr4 as soon as they leave the primitive streak (PS). At embryonic day (E) 8.5, the primary heart tube is mainly formed from first heart field ([FHF], red) progenitors, with additional contribution to the arterial and venous pole, respectively, from anterior and posterior second heart field ([aSHF], blue, and [pSHF], purple, respectively) progenitors that lie behind the heart tube, within the cardiopharyngeal mesoderm. On the right, the contribution of FHF and anterior and posterior pharyngeal mesoderm to the head and neck muscles and to the heart; below, the lineage tree deduced from retrospective clonal analysis. Interestingly, there are at least three reiterations of the Ciona ontogenetic motif along the anteroposterior axis of the embryo. The expression window of Mesp1 is indicated by green dotted lines. Mesp1 is first expressed in the FHF progenitors (red), and expressed later in the anterior (blue) and posterior (purple) progenitors of the cardiopharyngeal field. (EOMs) Extraocular muscles, (ao) aorta, (pt) pulmonary trunk, (RA) right atrium, (LA) left atrium, (RV) right ventricle, (LV) left ventricle.

Additional PSC studies have further shown that cardiac progenitors from slightly later stages, when Nkx2-5 is coexpressed with other markers such as the TF Isl1, are able to generate colonies with the potential to give rise to two or three cardiac cell types (Moretti et al. 2006; Wu et al. 2006). Therefore, early CPCs are multipotent and plastic. Nevertheless, we will discuss below that they are also specified very early toward a certain fate, and that when left in their endogenous environment, they are mostly uniproductive and contribute to specific cardiac regions and cell types.

Single Cells Analyses Reveal Heterogeneity among CPCs

Single-cell analyses have considerably changed our vision of CPC specification and have shown that the differential contribution of CPCs in the mouse model is complex, with many branching points and possible outcomes.

The first clonal analyses performed in mouse, using retrospective clonal analysis, demonstrated the existence of two myocardial cell lineages, corresponding to the FHF and SHF. On the basis of the number of β-gal⁺ cells included in clones covering both lineages, lineage segregation between the first and second myocardial lineages was estimated to occur at the onset of gastrulation (Meilhac et al. 2004). More recent clonal tracing of Mesp1⁺ CPCs using two independent strategies (Mesp1-rtTA/tetO-Cre/Rosa-Confetti or an Mesp1-Cre/MADM system) established that lineage segregation of the FHF and SHF occurs before the onset of Mesp1 expression, as shown by the absence of clones in both FHF and SHF derivatives (Devine et al. 2014; Lescroart et al. 2014). Similarly, clonal tracing of Smarcd3⁺ progenitors showed no evidence of a common progenitor for the FHF and SHF (Devine et al. 2014).

As in tunicates, multipotent progenitor cells giving rise to the SHF in the mouse also generate skeletal muscles of the head and neck, with at least three different populations of bipotent progenitors found along the anteroposterior axis of the pharyngeal region, contributing to the masticatory muscles and RV, facial expression muscles, and OFT or neck muscles and venous pole as well as OFT myocardium (Fig. 1; Lescroart et al. 2010, 2015). Furthermore, Mesp1 lineage tracing also identified clones labeling both skeletal muscles of the head and the myocardium, suggesting that segregation between cranial and cardiac muscles occurs after the onset of Mesp1 expression (Lescroart et al. 2014). Nevertheless, most Mesp1⁺ CPCs are uniproductive and contribute in the heart to either the epicardium, CMs, SMCs, or ECs, with only a few bipotent Mesp1-derived clones contributing to CMs and ECs, or SMCs and CMs, and only in SHF derivatives (Devine et al. 2014; Lescroart et al. 2014).

The heterogeneity of CPCs is exemplified by the different genetic lineage tracing experiments that have been performed and that label different compartments of the heart (Table 1; Meilhac and Buckingham 2018). For example, Foxa2⁺ cells, found very early during gastrulation, preferentially contribute to the ventricles and epicardium. Interestingly, Foxa2⁺ cells seem to express lower levels of Mesp1 (Bardot et al. 2017). Tracing with a particular enhancer of Mef2c, activated during gastrulation, labels only a subpopulation of cardiac tissue and specifically aSHF derivatives, comprising the OFT and RV, as well as some head muscles (Verzi et al. 2005; Lescroart et al. 2010; Devine et al. 2014). The anterior Hox genes (Hoxa1, Hoxa3, Hoxb1) distinguish pSHF progenitors with contributions to the inferior wall of the OFT and atria (Bertrand et al. 2011). Hand1⁺ cells only contribute to the CM of the LV (Barnes et al. 2010). Similarly, early Hcn4-expressing cells contribute specifically to the cardiac crescent and subsequently to LV CMs (Liang et al. 2013; Spater et al. 2013). Interestingly, Sfrp5, which encodes a secreted receptor for Wnt signaling, is expressed in progenitors of all cardiac regions including CMs, ECs, and EPDCs, with the exception of the RV (Fujii et al. 2017).

In addition, there is also clearly a dynamic temporal expression of markers during CPC specification such that CPCs expressing Mesp1 at different time points during gastrulation will have different contributions to the heart and head. Temporally controlled clonal analysis of Mesp1⁺ CPCs has thus shown that LV progenitors express Mesp1 early (induced with doxycycline at E6.25), whereas the OFT, RV, and atria and head muscles derive preferentially from CPCs that express Mesp1 slightly later (induced at E7.25) (Lescroart et al. 2014). Similarly, clonal analysis using a Six2-Cre-ERT2 line revealed that Six2 is expressed in a subset of late CPCs and that Six2 is successively activated in different populations of CPCs that contribute to the RV (induction with tamoxifen at E6) and then to the OFT (induction at E7.5) together with head muscles (Zhou et al. 2017). Interestingly, inducible lineage tracings of Hcn4⁺, Tbx5⁺, or Sma⁺ (smooth muscle actin, expressed in embryonic CMs) show a progressive contribution of cardiac progenitors to the heart with early differentiating cells contributing specifically to the left side of the interventricular septum (Spater et al. 2013; Devine et al. 2014; Choquet et al. 2016).

Finally, with the emergence of new technologies such as high-throughput scRNA-seq, it is now possible to study the heterogeneity of CPCs in detail as they arise and specify to a cardiac or muscular fate during gastrulation. A first study by Chan et al. (2016) applied scRNA-seq to Mesp1⁺ mPSCs during differentiation and revealed the existence of at least six subpopulations of Mesp1⁺ progenitors, with some primed for hematopoietic lineages, whereas others were primed toward the cardiac lineages. Single-cell expression profiling of nascent mesoderm at the late stage of mouse gastrulation demonstrated the existence of distinct populations of cardiac progenitors with different states of expression of Mesp1, Tbx5, Nkx2-5, and Isl1 (Kokkinopoulos et al. 2015). scRNA-seq of Mesp1⁺ CPCs in vivo (from E6.75 to E7.25) has shown a continuum of cell states along distinct cardiopharyngeal-related trajectories toward four destination cell types. These distinct trajectories likely correspond to CMs, ECs, anterior CPF (including aSHF) or posterior CPF (including pSHF) lineages, with different markers enriched in these different populations. RNA-fluorescent in situ hybridization (FISH) using specific markers of these four types further demonstrated that these populations of prespecified CPCs have distinct spatial localization in the gastrulating embryo (Fig. 2). EC-primed Mesp1⁺ progenitors are found in the outer layer of mesodermal cells, in the vicinity of the visceral endoderm, whereas CM-primed Mesp1⁺ progenitors are found closer to the epiblast. aSHF and pSHF-Mesp1⁺ markers were largely nonoverlapping, with Wnt2b (pSHF marker) being expressed close to the PS and FoxC2 (aSHF marker) being expressed more anterolaterally (Fig. 2; Lescroart et al. 2018). Tcf21, which labels cardiopharyngeal progenitors that contribute to head muscle formation, is expressed in the anterior CPF population.

Figure 2.

Heterogeneity of cardiopharyngeal progenitor cell (CPC) specification as revealed by single-cell analysis. (A) In single-cell RNA-sequencing (scRNA-seq) experiments, four distinct Mesp1⁺ CPC subpopulations have been identified that emerge from epiblast cells and that correspond to the endothelial/endocardial cell ([EC], black), cardiomyocyte ([CM], red), anterior cardiopharyngeal field ([aCPF], light green), and posterior CPF (pCPF, dark green) progenitors. These populations are enriched for some transcripts (in parentheses) and are found in particular locations in the embryo (B). (B) Drawing of the gastrulating embryo, where the different CPC subpopulations are found. The embryo is bilateral; on one side is depicted the localization of the Mesp1⁺ EC versus Mesp1⁺ CM populations found either in the outer layer or close to the epiblast, whereas on the other side the localization of aCPF and pCPF progenitors is represented along the anteromedial axis of the embryo. These distinct CPC subpopulations make different regional contributions to the definitive heart (C). (AA) Aortic arch, (PT) pulmonary trunk, (Ao) aorta, (PV) pulmonary vein, (SCV) superior caval vein, (ICV) inferior caval vein, (RA) right atrium, (LA) left atrium, (RV) right ventricle, (LV) left ventricle.

In Ciona, the two Mesp⁺ progenitors divide once in the early gastrula to give rise to four equivalent multipotent progenitors, which then divide asymmetrically at the end of gastrulation to generate anterior tail muscle (ATM) progenitors and trunk ventral cells (TVCs). Cardiac cells (CMs and pericardial cells) derive from TVCs (Davidson et al. 2005, 2006; Beh et al. 2007; Christiaen et al. 2008). TVCs are multipotent progenitors that divide asymmetrically a second time to form first heart precursors (FHPs) and second TVCs (STVCs) (Stolfi et al. 2010). STVCs divide asymmetrically again to form lateral atrial siphon muscle founder cells (ASMFs) and second heart precursors (SHPs) (Fig. 1; Stolfi et al. 2010; Wang et al. 2013). scRNA-seq of progenitor cells covering the early branches of the cardiopharyngeal lineage tree in Ciona intestinalis identified three distinct transcriptional clusters that correspond to the cardiac FHP and SHP populations as well as an Ebf⁺ pharyngeal muscle population. Interestingly, this analysis showed that specific transcriptional programs are activated in fate-committed progenitors. In addition, this single-cell transcriptomic study identified new specific-lineage markers, including Mmp21 for the FHP and Dach as an SHP-specific marker. Remarkably, in the mouse, Dach1 is first broadly expressed in pharyngeal mesoderm and then also restricted later to cardiac SHF progenitors (Wang et al. 2019).

Although single-cell approaches have helped us investigate in more detail the heterogeneity of the CPC population at different stages of development mostly through gene expression, the combination of these techniques with rapidly advancing systems approaches to functional genomic analysis (single-cell CITE-seq, single-cell ATAC-seq) will further improve our understanding of the molecular mechanisms implicated in early cell fate decisions.

REGULATION OF CPC SPECIFICATION

Early Induction of Cardiogenic Mesoderm

The extrinsic signals given to cells during development in vivo can be used as guidance cues during PSC differentiation from pluripotency toward specific cell types. It is therefore important to understand the stage-specific requirements of signaling pathways and their downstream TFs during early differentiation (Noseda and Schneider 2010; Loh et al. 2016).

CPCs arise from epiblast cells that form the primitive streak (PS). Their specification therefore depends on correct positioning and formation of the PS; this process relies on several signaling pathways, including canonical Wnt/β-catenin and Tgfβ/Nodal pathways (Conlon et al. 1994; Haegel et al. 1995; Noseda and Schneider 2010). The same pathways are required to induce formation of PS-like cells and subsequent cardiogenic mesoderm from PSCs (Xu et al. 1999; Gadue et al. 2006). Canonical Wnt signaling has a biphasic role during cardiogenesis, as it promotes early steps of cardiogenesis but then inhibits CM differentiation (Marvin et al. 2001; Schneider and Mercola 2001; Tzahor and Lassar 2001; Naito et al. 2006; Liu et al. 2007; Qyang et al. 2007; Ueno et al. 2007; Paige et al. 2010). Bone morphogenetic protein (BMP) signaling is also required for subsequent gastrulation and formation of the cardiac crescent (Mishina et al. 1995; Miura et al. 2006). The requirement of these signaling pathways for cardiac development is well conserved throughout chordates (Noseda and Schneider 2010). Modulation of these pathways and others has guided most protocols of differentiation of PSCs toward cardiac lineages (Laflamme et al. 2007; Yang et al. 2008; Kattman et al. 2011). Briefly, formation of cardiac mesoderm can be induced from PSCs by BMP4, Vegf, Activin A, and Fgf2, and subsequent differentiation is enhanced by Wnt inhibition. These are the minimal requirements for the formation of cardiac lineages in vitro in serum-free conditions. These extrinsic signals induce an intrinsic cascade of molecular events, in which the activation of core transcriptional regulators thereafter progressively promotes cardiac cell fates during differentiation. We will now discuss these intrinsic events.

In Drosophila, tinman, the ortholog of Nkx2-5, is considered to be the master regulator of cardiogenesis (Azpiazu and Frasch 1993; Bodmer 1993; Frasch 1995). Whereas mutations in Nkx2-5 are associated with congenital heart disease (Schott et al. 1998), Nkx2-5 mouse mutants form an early heart tube (Lyons et al. 1995; Tanaka et al. 1999) showing that during evolution, Nkx2-5 lost its central role in cardiac specification, likely as a result of the expansion of the vertebrate Nkx gene family. A key role for Mesp1 has clearly been established for cardiac specification in chordates (Saga et al. 2000; Bondue and Blanpain 2010; Liu 2017). Mesp1 mutant mouse embryos present cardia bifida, with differentiated CMs and ECs in separated heart fields (Saga et al. 1999). Double mutants for Mesp1 and its closest homolog, Mesp2, on the other hand, have severe defects of gastrulation. Chimeras with Mesp1/2 null and wild-type cells showed contributions of mutant cells to most of the embryo excluding the heart, indicating a cell-autonomous requirement of Mesp1/2 for CPC specification (Kitajima et al. 2000). Mesp1^Cre/Cre knockout cells present a developmental block similar to that of double-mutant embryos, with mesodermal precursors being stuck in the gene program of the epiblast, as shown by scRNA-seq. Transcripts of several regulators of pluripotency such as Nanog, Oct4, and Eras or markers of the epiblast including E-cadherin and Epcam, were thus up-regulated in Mesp1-null cells (Lescroart et al. 2018).

Interestingly, during mPSC differentiation, overexpression of human or murine Mesp1 accelerates and increases cardiac differentiation, as shown by the appearance of beating areas and enhanced troponin T expression (Bondue et al. 2008; David et al. 2008; Lindsley et al. 2008). However, continuous expression of Mesp1 inhibits cardiac differentiation (Bondue et al. 2008). Mesp1 has been shown to directly bind to the promoter of key cardiac TFs, such as Mef2c, Tbx20, Gata4, and Nkx2-5, thereby up-regulating their expression and promoting further up-regulation of downstream cardiac genes (Fig. 3). Nevertheless, in the context of reprogramming, forced Mesp1 expression in cardiac fibroblasts is not sufficient to induce cardiac specification. In contrast, Smarcd3, Tbx5, and Gata4 have been shown to be sufficient for cardiac reprogramming (Takeuchi and Bruneau 2009; Ieda et al. 2010).

Figure 3.

Early intrinsic regulation of cardiac specification. Representation of the molecular hallmarks of cardiac progenitor cell commitment. When cardiac progenitors express Mesp1 in the nascent mesoderm, they turn off the genetic program of pluripotency and turn on the gene regulatory network of cardiac specification and migration. Green arrows represent direct activation; red arrows represent repression. Genes grouped together are regulated in a similar manner by Mesp1. Although Linc1405 was shown to interact with Eomes, it was not the case for Braveheart, another cardiogenic long noncoding RNA. (FHF) First heart field, (SHF) second heart field, (CPC) cardiopharyngeal progenitor cell, (EMT) epithelial-to-mesenchymal transition.

The key role of Mesp genes during cardiac progenitor specification appears to be highly conserved in chordates, as shown by the induction of ectopic cardiac differentiation upon expression of a constitutive form of Mesp ortholog in Ciona and human Mesp1 in Xenopus (Davidson et al. 2005; David et al. 2008). Knockdown experiments using morpholinos also demonstrated the importance of Mesp for cardiac specification in ascidians (Satou et al. 2004). In fish, the master regulator role of Mesp is not clear. Although there are four mesp genes, only mespaa has a procardiogenic role after animal cap transplants with mespaa overexpressing cells. Surprisingly, mespaa or even quadruple mespaa/ab/ba/bb CRISPR/Cas9 mutants did not show any cardiac defects (Deshwar et al. 2016).

Mesp1 is activated in nascent mesoderm by key regulators of endoderm and mesoderm development Eomesodermin (Tbr2) and Brachyury (T) (Costello et al. 2011; David et al. 2011; van den Ameele et al. 2012). In Ciona embryos, Mesp is activated by Tbx6 and Lhx3, only in B7.5 cells coexpressing these factors (Davidson et al. 2005; Christiaen et al. 2009). Consistent with these findings, a recent report searching for mesoderm-inducing factors in mouse embryonic fibroblasts found that Tbx6 was sufficient to activate Mesp1 and subsequent CPC gene expression. During PSC differentiation, Tbx6 promoted mesoderm and endoderm induction in the absence of serum and growth factors in both a cell-autonomous and non-cell-autonomous manner (Sadahiro et al. 2018). Further studies in mouse will be required to determine how the precise domain of Mesp1 expression is defined within the nascent mesoderm, although other factors are also implicated in early specification of cardiac progenitors, such as long noncoding RNAs (Klattenhoff et al. 2013; Ounzain et al. 2015; Guo et al. 2018), which have even been used to promote transdifferentiation to cardiac cell types (Fig. 3; Hou et al. 2017).

Finally, the dynamic gene expression changes involved in transitions from pluripotency to cardiogenic mesoderm are accompanied by modifications of the epigenetic landscape or chromatin accessibility, with regulatory regions of genes being first primed then activated, and other regions involved in alternative fates being decommissioned (Paige et al. 2012; Wamstad et al. 2012; Racioppi et al. 2019). TFs cooperate to induce these changes, through interactions with each other and specific chromatin modifiers (Chang and Bruneau 2012; Luna-Zurita et al. 2016). Combining data from single-cell expression and broad epigenetic studies, such as repressive histone mark ChIP-seq data from many tissues, may help us to identify new genetic drivers of cell fate specification (Shim et al. 2019).

CPC Specification to the Different Cardiac Lineages

In vitro, mPSCs cultured in serum conditions can form three populations with differential expression of T and Flk1 (Kouskoff et al. 2005; Kattman et al. 2006). Hemangioblast specification usually precedes cardiac specification, as the former derives from a first wave of Flk1-expresssing Brachyury⁺ cells, whereas the cardiogenic mesoderm arises from a Brachyury⁺ population, with later expression of Flk1 (Kattman et al. 2006). Interestingly, if Mesp1 is induced earlier during mPSC differentiation, there is an increase in the hematopoietic progenitor population through direct activation of Tal1, encoding a vascular TF, at the expense of cardiac fates (Chan et al. 2013). Conversely, in the absence of Tal1, extraembryonic tissue (also expressing Mesp1) can differentiate into CMs (Van Handel et al. 2012).

CM and EC lineages of the FHF segregate early during gastrulation (Mikawa and Fischman 1996; Wei and Mikawa 2000; Lescroart et al. 2014). scRNA-seq of Mesp1⁺ CPCs showed that several Notch signaling actors, including Notch1, are specifically down-regulated in CM-fated cells and conversely up-regulated in EC-fated cells (Lescroart et al. 2018). The Notch pathway has been implicated in several key processes of cardiac development, including trabeculation and valve formation (MacGrogan et al. 2010, 2018). In vitro activation of Notch reduces CM differentiation (Nemir et al. 2006) and repression of Notch increases CM differentiation (Schroeder et al. 2003). Lineage tracing of early Notch1⁺ cells preferentially marks the endocardium, suggesting a potential role for the Notch pathway in early CPC fate decisions between CMs and ECs (Lescroart et al. 2018). Moreover, further cross talk between these two lineages happens in vitro and in vivo, suggesting paracrine mechanisms governing CPC specification and differentiation (Saba et al. 2019; Saint-Jean et al. 2019). Interestingly, ECs with “valve-like” properties can be derived from hPSCs through the activation of other signaling pathways. Vascular endothelial growth factor (VEGF), together with FGF2 and FGF8, can induce EC differentiation at the expense of CMs, thus favoring the specification of a prevalvular cell type (Neri et al. 2019).

The PSC model has also been used to specify PSCs into other cardiac cell lineages (Fig. 4). Epicardial cells, for example, can be produced from hPSCs. Indeed, stage-specific activation of BMP, retinoic acid (RA), and Wnt/β-catenin signaling pathways after induction of CPCs induces the specification of epicardial-like progenitors expressing TBX18 and WT1, two markers of the proepicardial organ (Iyer et al. 2015; Bao et al. 2016; Guadix et al. 2017). In line with these findings, mouse embryos lacking the Wnt inhibitors Dkk1 and Dkk2 have a thicker epicardium and thinner myocardium (Phillips et al. 2011), showing again the necessity and usefulness of translating our current in vivo knowledge for in vitro generation of specific cell types. Epicardial cells can differentiate into cardiac fibroblasts or vascular SMCs through epithelial-to-mesenchymal transition, which can be induced in vitro by addition of Tgf-β1 to the cells. Nevertheless, as the epicardium gives rise to multiple cell lineages, the resulting cell population is heterogeneous. To obtain a pure population of cardiac fibroblasts without going through an epicardial-like stage, another protocol has recently been developed in which activation of the FGF pathway in hPSC-derived CPCs induces formation of periostin- and vimentin-expressing cardiac fibroblasts, presenting many functional similarities with their in vivo counterparts (Zhang et al. 2019).

Figure 4.

Extrinsic regulation of cardiac specification. By differentiating mouse and human pluripotent stem cells (PSCs) under serum-free conditions, the minimal cues necessary for their specification and differentiation to a particular cardiopharyngeal cell lineage were defined. Arrows indicate different steps of differentiation. All protocols pass through a stage in which cells present features of multipotent cardiopharyngeal progenitor cells (CPCs), then diverge in terms of given signaling modulators and cell markers. (AA) Ascorbic acid, (BMP) bone morphogenetic protein, (CFs) cardiac fibroblasts, (CMs) cardiomyocytes, (FGF) fibroblast growth factor, (IGF-1) insulin-like growth factor-1, (HGF) hepatocyte growth factor, (SAN) sinoatrial node, (SHF) second heart field, (SMCs) smooth muscle cells, (VEGF) vascular endothelial growth factor.

Pacemaker CMs can also be generated from mPSC by induction of Tbx3 (Jung et al. 2014) or Shox2 (Ionta et al. 2015). Sinoatrial node (SAN)-like pacemaker cells arise from Nkx2.5^–, Tbx18⁺ CPCs (Mommersteeg et al. 2010). As a first demonstration of generating SAN-like pacemaker cells from hPSCs without recurring to overexpression, inhibition of the FGF pathway with activation of BMP and RA signaling generated a NKX2-5- but SIRPA⁺ cardiac cell type, presenting pacemaker properties similar to SAN cells. These cells were able to display pacemaker activity when grafted into rat hearts (Protze et al. 2017).

Regional Specification

Multiplication of the cardiac chambers has significantly increased the efficiency but also complexity of the heart during evolution. The specification of early CPCs toward a particular cardiac region is not yet fully understood and there may be a lot to glean from in vitro PSC models and in vivo studies.

One of the main regional distinctions is between atrial and ventricular cardiac tissue. As discussed previously, lineage-tracing experiments in the mouse have shown the existence of distinct CPCs that contribute to different cardiac regions. For example, Foxa2⁺ CPCs specifically label ventricular derivatives. Formation of chimeric embryos from wild-type and Foxa2^−/− mutant mPSCs have shown that Foxa2 is cell-autonomously required for ventricular specification such that there is no contribution of mutant cells to the ventricles. Similarly, Foxa2^−/− mPSCs have a decreased ability to form ventricular CMs (Bardot et al. 2017). RA signaling is also a known player for regional specification, such that exogenous RA treatment patterns the CPCs along the anteroposterior axis and promotes atrial identity (Xavier-Neto et al. 1999; Hochgreb et al. 2003; Huang et al. 2011; Stefanovic and Zaffran 2017; De Bono et al. 2018). The specific addition of RA at the cardiac mesoderm-CPC stage during hPSC differentiation favors specification toward an atrial-like phenotype (KCNJ3⁺) at the expense of ventricular fate (MYL2⁺) (Zhang et al. 2011; Lee et al. 2017). A study has shown that ventricular hPSC-derived CPCs are marked by the expression of CYP26A1, which encodes an enzyme responsible for RA degradation. In contrast, atrial hPSC-derived CPCs express RALDH2 (an enzyme required for the synthesis of RA). RALDH2⁺ CPCs will preferentially generate atrial CMs (Lee et al. 2017). Consistently, in the mouse model, early Mesp1-expressing CPCs, which contribute preferentially to the LV, express Cyp26a1, whereas atrial Mesp1⁺ CPCs that express Mesp1 at later stages preferentially express Raldh2 (Lescroart et al. 2014). Interestingly, ISL1 knockout hPSCs resemble PSC cultures exposed to RA signaling and lead to the preferential specification of atrial CMs. Further investigation showed that ISL1 suppresses the atrial specifier NR2F1. Specification of hPSCs into atrial CMs through RA signaling appears to be mediated by activation of the TALE TF MEIS2, which antagonizes ISL1 expression (Quaranta et al. 2018).

In fish, RA signaling also indirectly counteracts the activity of Isl1 through Ajuba, an LIM domain protein (Witzel et al. 2012). In addition, in vitro studies have demonstrated that the proportion of atrial versus ventricular progenitors can be modulated by different concentrations of Activin A and Bmp4 early during hPSC differentiation (Fig. 4; Lee et al. 2017). It would be very informative to evaluate the levels of BMP4 and Activin A in vivo to see whether this reflects a distinct environmental context for atrial and ventricular progenitors. Furthermore, differentiation protocols established so far do not yet allow us to differentiate between LV versus RV or between atrial cardiac cell types, such as right versus left atrial CMs or pacemaker cells. These might be achieved by understanding the specification of the FHF and anterior and pSHF as they contribute to different regions of the heart.

The establishment of spheroids from HCN4-GFP/Tbx1-Cre; Rosa-tdTomato mPSCs has allowed the study of FHF versus SHF specification, despite the fact that Tbx1 is only expressed in a subpopulation of the SHF. Transcriptomic analysis demonstrated that Bmp signaling is enriched in HCN4⁺ FHF progenitors, whereas Wnt signaling appears to be specific to Tbx1-derived SHF cells. Modulating levels of Bmp4 or Wnt3A demonstrated that Bmp4 can promote FHF or Tbx1⁺ SHF specification depending on its concentration, whereas Wnt3a specifically promotes the appearance of the Tbx1-derived SHF population. Interestingly, the same study showed that Cxcr4, which initially marks early CPCs, is not expressed later in the FHF and specifically labels the Tbx1⁺ SHF population (Andersen et al. 2018).

Finally, scRNA-seq has enabled prediction of lineage specifiers for specific cardiac regions. For example, Hopx, Chd3, and Ncor1 and Shox2, Sox4, and Hoxb4 were among the major genes predicted to play a role in the specification of the atrial or sinus venosus lineages, respectively. Similarly, Nr1d1 could be a cell fate determinant for the RV, whereas Hand2, Msx1 may specify the OFT (de Soysa et al. 2018). Indeed, scRNA-seq of Hand2-null cells shows dysregulation of the RV and OFT gene program with abnormal activation of posterior genes, but these data suggest a failure of RV cells to migrate to their final location rather than a defect in their specification (de Soysa et al. 2018). Similarly, ectopic expression of Hoxb1 in aSHF CPCs causes deregulation of the aSHF gene program, suggesting that homeodomain genes play a key role in the anteroposterior patterning and cell identity of CPCs (Stefanovic et al. 2019). Relatedly, in tunicates, knockout of the SHP-specific gene Dach in multipotent CPCs led to ectopic expression of the FHP marker Mmp21, suggesting a change in cell fate identity (Wang et al. 2019).

Cardiomyocyte Versus Skeletal Muscle Fate

As discussed above, the CPF is able to differentiate into both cardiac tissue and skeletal muscles of the head and neck. Several studies in mice, chick, and fish have shown that there is a conserved regulatory network (including Tbx1, Pitx2, Nkx2-5, and Isl1) that governs cardiac and skeletal muscle progenitor specification (Tirosh-Finkel et al. 2006; Nathan et al. 2008; Sambasivan et al. 2009; Nevis et al. 2013; Paffett-Lugassy et al. 2017; Guner-Ataman et al. 2018). Interestingly, fundamental insights on how the CPF is able to differentiate into cardiac or skeletal muscle lineages arise from studies in C. robusta. It is now clear that in tunicates, there is an antagonism between Nk4 (homolog of Nkx2-5) and Tbx1/10 to regulate cardiac versus skeletal muscle specification. Thus, in committed cardiac progenitors, NK4 inhibits Tbx1/10-dependent activation of Ebf, whereas in atrial siphon muscle (ASM) progenitors, NK4 is repressed, thereby allowing Tbx1/10 to activate the skeletal muscle program (Wang et al. 2013). Ebf is then required for differentiation of ASMs by promoting the pharyngeal myogenic program (Razy-Krajka et al. 2014).

Fgf/MAPK signaling also plays a critical role in Ciona in skeletal versus cardiac specification. Activated in multipotent progenitors, Fgf signaling is then maintained only in the lateral-most daughter cells during cell division within the CPF lineage. Fgf signaling is thereafter restricted to ASM progenitors, thus promoting skeletal muscle identity (Razy-Krajka et al. 2018). Even if a role for Fgf signaling in cardiac versus skeletal muscle specification has not yet been shown in vertebrates, it is required for development of the CPF (Abu-Issa et al. 2002; Vitelli et al. 2002; Brown et al. 2004; Hu et al. 2004; Aggarwal et al. 2006; Ilagan et al. 2006; Park et al. 2006, 2008; Kelly and Papaioannou 2007; de Pater et al. 2009; Watanabe et al. 2010, 2012; Lazic and Scott 2011).

scRNA-seq in Ciona has further enabled precise investigation of temporal CPC specification. It led to the characterization of 10 discrete regulatory states along specification pseudotime, with identification of “primed” and “de novo” markers that are, respectively, expressed either throughout these states or only in more committed progenitors. About 50% of cardiac or skeletal muscle genes are “primed” and therefore already expressed in multipotent progenitors. Cross comparison with other published scRNA-seq of developing mouse hearts showed high conservation of the transcriptional network governing cardiac versus skeletal muscle specification (Wang et al. 2019).

Tbx1 mutant embryos show defects in both craniofacial skeletal muscles and heart (Lindsay et al. 2001; Kelly et al. 2004; Lescroart et al. 2015). Through RNA sequencing of wild-type versus tbx1 mutant zebrafish embryos, gdf3-ALK4, a TGFµ superfamily ligand has been identified as a downstream effector of tbx1 involved in CPC specification (Guner-Ataman et al. 2018). In fish, tbx1 is also required to specify the nkx2.5⁺ CPF. Moreover, recent studies have also shown a role for Wnt signaling in the balance between cardiac and skeletal muscle lineages, as Wnt activation during or prior to gastrulation inhibits pharyngeal muscle specification at the expense of cardiac lineages (Mandal et al. 2017). Such a role for Wnt has not been found in tunicates (Kaplan et al. 2019). Other factors have been implicated in pharyngeal muscle specification. NR2F or COUP-TF proteins, downstream targets of RA signaling, have been shown, in fish, to promote posterior pharyngeal muscle specification (muscles analogous to the vertebrate neck muscles) while restricting ventricular CM progenitors (Dohn et al. 2019).

BMP signaling, when applied to differentiating mPSCs, leads to the down-regulation of skeletal muscle markers while promoting cardiac genes. RNA-seq showed that Podxl and Cdh4 are surface markers that could be used to distinguish between committed skeletal and cardiac progenitors, respectively (Chan et al. 2016). After mesoderm induction, dual inhibition of Wnt and Nodal promotes CPC specification, and cells thus express Tbx1, Isl1, and Nkx2-5. With specific culture conditions, it is then possible to specify the CPF into skeletal muscle fibers with a cocktail of signaling molecules including a Bmp inhibitor, Hgf, Fgf2, and Igf1 or into beating CMs when supplemented with ascorbic acid, Vegf, Fgf2, and Fgf10 (Nandkishore et al. 2018).

CONCLUDING REMARKS

The specification of CPCs is a complex developmental process that occurs concomitantly with their migration. Mesp1, a key transcription master regulator of CPC specification, has also a critical role in CPC migration by controlling the epithelial-to-mesenchymal transition, cell polarity, and migration speed (Bondue et al. 2008; Lindsley et al. 2008; Chiapparo et al. 2016). Therefore, it would be important to integrate both CPC migratory trajectories and lineage behavior to investigate the influence of each process on the other. In addition, it is well established that left–right asymmetry is critical for proper heart development (Desgrange et al. 2018), but it is not clear how this affects CPC specification.

With the emergence of single-cell approaches, the model of CPC specification has changed, and it is now clear that CPCs represent a very heterogeneous and dynamic population of progenitors, reflecting distinct physical positions in space and time but also some stochastic factors. In addition, recent research has shown that early CPCs are “primed” toward a particular lineage very early during embryonic development. The determinants of this early “primed” transcriptional heterogeneity are not yet understood. Comprehending the regulation of CPC specification will help to decipher the developmental mechanisms leading to congenital heart disease, which affect ∼1% of newborns worldwide (Srivastava 2006).

In addition, investigating the embryonic regulation of CPC specification has already significantly improved differentiation protocols of PSCs toward a specific cell type. This is particularly relevant in the context of cell therapy after cardiac injury. The identification of multipotent CPCs has brought significant hope to the field of heart repair and further research into the early steps of CPC specification will undoubtedly lead to the design of strategies to specify multipotent CPCs into a specific mature cardiac cell type that could then be used for cell therapy. As a recent example, hPSC-derived CMs have recently been grafted into infarcted pig hearts and showed significant engraftment, although accompanied by ventricular tachyarrhythmia (Romagnuolo et al. 2019). Further improvements are thus required to successfully repair infarcted hearts in patients.