Subtype-specific cardiomyocytes for precision medicine: where are we now?

Abstract

Patient-derived pluripotent stem cells (PSCs) have greatly transformed the current understanding of human heart development and cardiovascular disease. Cardiomyocytes derived from personalized PSCs are powerful tools for modeling heart disease and performing patient-based cardiac toxicity testing. However, these PSC-derived cardiomyocytes (PSC-CMs) are a mixed population of atrial-, ventricular-, and pacemaker-like cells in the dish, hindering the future of precision cardiovascular medicine. Recent insights gleaned from the developing heart have paved new avenues to refine subtype-specific cardiomyocytes from patients with known pathogenic genetic variants and clinical phenotypes. Here, we discuss the recent progress on generating subtype-specific (atrial, ventricular, and nodal) cardiomyocytes from the perspectives of embryonic heart development, and how they will expand our current knowledge on molecular mechanisms of cardiovascular disease and the future of precision medicine.

Introduction

Human pluripotent stem cells (PSCs), including embryonic stem cells (ESCs) and induced pluripotent stem cells (iPSCs) have greatly revolutionized modern cardiovascular medicine due to their ability to generate any cell type in the cardiovascular system. Human ESCs are derived from the inner cell mass of developing embryos (blastocysts), thus they can’t be patient-specific [1]. Patient-specific PSCs can be generated by two reprograming mechanisms: somatic cell nuclear transfer (SCNT) [2] and iPSC reprograming [3]. SCNT is the process of cellular reprogramming of somatic cells by injecting the nucleus of a somatic cell into an enucleated oocyte which is both technically and ethically challenging. In contrast, human iPSCs can be derived by transient overexpression of four transcription factors (OCT4/SOX2/CMYC/KLF4), thus circumventing the ethical barriers for translational medicine. Although differentiated cells from isogenic SCNT-ESCs and iPSCs are equivalent in the aspects of molecular and functional properties [4], patient-specific iPSCs have become prevalent in current biomedical research. Large banks of disease-specific iPSCs have been created and distributed around the world, which are transforming our understanding of cardiovascular science and health.

Human iPSCs have been extensively employed to model prevalent heart disease, such as long QT syndrome [5–7], Brugada syndrome [8], arrhythmogenic right ventricular dysplasia [9], dilated cardiomyopathy (DCM) [10,11], and hypertrophic cardiomyopathy (HCM) [12]. In addition, human iPSC-derived cardiomyocytes (iPSC-CMs) are considered powerful platforms for novel drug screening and cardiac toxicity testing [13]. Recent CiPA (Comprehensive In Vitro Proarrhythmia Assay) studies, initiated by the US Food and Drug Administration (FDA), have highlighted the importance of using human iPSC-CMs to evaluate drug-induced arrhythmic effects [14]. Moreover, drug-induced cardiac toxicity, as seen with the chemotherapeutic agent doxorubicin, can be recapitulated in patient-derived iPSC-CMs, where genetic variance responsible for cardiac toxicity may be inferred for clinical references [15,16]. This has paved the way to an innovative high-throughput approach to screen novel drugs that can prevent and reverse cardiac toxicity caused by anti-cancer chemotherapy drugs [17]. Though human cardiomyocytes can now be derived in a patient-specific manner through iPSC reprogramming, most of these studies have been carried out with a mixed population of atrial-, ventricular-, and pacemaker-like cardiomyocytes. The mixed subtypes of cardiomyocytes may confound the disease phenotypes in the dish and compromise the therapeutic outcomes of transplanted cells in vivo [18–21]. Therefore, it is becoming increasingly important to develop subtype-specific differentiation protocols for precise generation of atrial-, ventricular-, and nodal-like cardiomyocytes in the petri dish.

The protocols for cardiac differentiation from human PSCs have been greatly improved in the past decade, transitioning from embryoid body (EB) and growth factor-based differentiation to small molecule-mediated monolayer differentiation. Numerous excellent review articles on cardiac differentiation methods have been well documented and readers can refer therein [22,23]. In this Review, we will discuss how to generate subtype-specific cardiomyocytes using strategies mimicking embryonic heart chamber development, and how the precise production of chamber-specific cardiomyocytes will impact the future of cardiovascular disease modeling and precision medicine. While epicardial- and endocardial-derived cells also make important contributions to the mature heart, a comprehensive discussion of these cellular contributors is beyond the scope of this review and excellent reviews on human PSC-derived endothelial and epicardial cells (cardiac fibroblasts) have been recently published elsewhere [20,24]. Here we will focus on different cardiomyocyte subtypes (atrial, ventricular, and nodal) generated by human PSC differentiation.

Developmental origin of chamber-specific cardiomyocytes

The human heart has four chambers: two atria and two ventricles. Multiple cell types comprise the four-chambered adult heart, where cardiomyocytes are the essential components responsible for maintaining the biochemical, mechanical, and electrical functions of the beating heart. The myocardium forms the muscular walls of the cardiac chambers and is primarily composed of atrial and ventricular cardiomyocytes. Pacemaker cells and Purkinje fibers are specialized cardiomyocytes capable of generating and conducting electrical impulses. Cardiomyocytes work together with other cell types in the heart, such as the endocardial cells which line the inner surface of the cardiac chambers, the endothelial cells and smooth muscle cells in the coronary vasculature, and the epicardial-derived fibroblasts, to pump blood to the entire body. Recent single-cell transcriptomic analysis of developing human heart further defines chamber-specific cell identity and reflects the developmental trajectories of various cell types in a spatiotemporal manner [25]. This work has provided potential clues to refine transcriptional signatures of subtype-specific cardiomyocytes derived from PSCs at the single-cell level [26] (Figure 1).

Figure 1. Single-cell transcriptomic profiles of atrial and ventricular chambers of human fetal heart.

(A) Anatomy of human heart chambers (image by courtesy of Centers for Disease Control and Prevention, National Center on Birth Defects and Developmental Disabilities). LA: left atrium, RA: right atrium, LV: left ventricle, RV: right ventricle. (B) Visualization of the clusters of single-cell transcriptomes from human fetal ventricles (LV and RV) and atria (LA and RA) using the UMAP algorithm. The single-cell RNA-seq data were extracted from the study by Cui et al., Cell Reports 2019 [25]. (C) Rankings of chamber-specific genes in one cluster compared to the rest of clusters. Notably, single-cell transcriptomic analysis has identified ventricular-specific genes MYL2 and MYH7, and atrial-specific genes MYH6 in human fetal heart.

During the initial stages of heart development, the precursors of subtype-specific cardiomyocytes (atrial, ventricular, and pacemaker cells) originate from distinct cellular populations in the first heart field (FHF) and second heart field (SHF) [27] (Figure 2). Mammalian cardiogenesis starts with the formation of cardiac mesoderm shortly after gastrulation. Mesoderm induction is triggered by multiple growth factors secreted from the adjacent endoderm, including bone morphogenetic protein 4 (BMP4) and Nodal. Nodal signaling in the proximal epiblast triggers Bmp4 expression in the extraembryonic ectoderm adjacent to the epiblast. BMP4 then induces Wnt3 expression in the epiblast, which further leads to the commitment of mesoderm marked by the expression of T (brachyury) and Eomes [28]. Cardiac commitment is initiated by the induction of the transcription factors NKX2.5 (NK2 transcription factor related, locus 5) and MESP1 [29,30]. The homeobox protein, NKX2.5, is expressed in early cardiac mesoderm, cardiac progenitors, and early cardiomyocytes in the left ventricular and atrial chambers where it cooperates with the zinc finger transcription factors of the GATA family to activate cardiac gene expression. MESP1 (mesoderm posterior 1), one of the earliest molecular markers of cardiac progenitors, is then transiently expressed in the newly formed cardiac mesoderm. MESP1 is a master regulator of multipotent cardiac progenitor specification for both FHF and SHF during early gastrulation [31]. MESP1⁺ progenitors consist of two temporarily distinct ancestors: FHF MESP1⁺ cells contribute to myocardium whereas SHF MESP1⁺ progenitors contribute to both the endocardium and myocardium [32]. Recent studies have shown that SHF progenitors from human PSCs can generate cardiomyocytes, smooth muscle cells, and endothelial cells [33].

Figure 2. Schematic view of the developmental scenarios of cardiomyocyte commitment.

Cardiac mesoderm is induced in response to signals (BMP4 and WNT) originated from the adjacent endoderm shortly after gastrulation. Cardiac commitment is initiated by early cardiogenic transcription factors such as MESP1 and cardiac mesoderm is marked by MESP1, KDR, and PDGFRA [120,121]. Early cardiovascular progenitors (ISL1⁺) migrate and contribute to the first heart field (FHF) and the second heart field (SHF) progenitors [122]. The FHF progenitors (NKX2.5⁺ISL1⁻) contribute to the left ventricle and portions of the atria, whereas the SHF progenitors (NKX2.5⁺ISL1⁺) generate the right ventricle, outflow tract, and parts of atria [123,124]. Other sources of progenitors may also contribute to the final cardiac subtype lineages. For example, HCN4⁺ FHF cells contribute to SAN conduction cells [111]; in chicks SAN pacemaker cells are derived from the tertiary heart field [112]. The cardiac subtype commitment is regulated by the cooperative interactions of the extrinsic signaling pathways (BMP, RA, and NOTCH) and the intrinsic transcriptional regulatory network (NKX2.5, GATA4, MEF2C, and TBX5).

Descendants of MESP1⁺ cardiac progenitors colonize and form all of the myocardium [34]. The cardiac mesoderm gives rise to the endocardium, the FHF, the SHF, and the proepicardial mesenchyme. The FHF forms the left ventricle and parts of the atria, and the SHF contributes to the right ventricle, outflow tract, and portions of the atria. Further myocardial differentiation is initiated by the transcriptional activation of key myocardial transcription factors, such as NKX2.5, GATA4, MEF2C, TBX5, HAND2, and ISL1 [35]. Mutations in many of these cardiac transcription factors have been linked to congenital heart defects in humans [36]. MESP1-mediated cardiac transcriptional activation is dependent on the signaling pathways induced by BMPs and FGFs [27], and MESP1 promotes cardiovascular differentiation by blocking Dkk-1 mRNA expression in the canonical Wnt signaling pathway [37].

Cardiac mesoderm progenitors converge and form a primitive heart tube that includes an interior layer (endocardium) and exterior layer (myocardium). Consequently, FHF- and SHF-derived cardiac progenitor cells incorporate and proliferate within the heart tube. Cardiomyocyte differentiation and proliferation is initiated in response to NOTCH signaling from the endocardium. Retinoic acid (RA) from the epicardium regulates myocardial proliferation by inducing FGF signaling [38–40]. These two signaling pathways play essential roles in the subsequent cardiovascular development, including endocardial-to-mesenchymal transition/endocardial cushion formation, valve development, epithelial-to-mesenchymal transition, ventricular trabeculation, and cardiac outflow tract morphogenesis [41,42]. Recent studies have demonstrated that they are prospective signal pathways that could be manipulated to derive subtype-specific cardiomyocytes from human PSCs [18]. Thus, in the following sections, we will discuss the mechanistic insights into the pivotal functions of the NOTCH and RA signaling in cardiomyocyte differentiation and heart development, and how we can manipulate these signals to coax PSCs into a more defined cardiomyocyte subtype.

NOTCH signaling pathway and cardiac development

NOTCH signaling plays an essential role in the cell fate determination and morphogenesis of cardiac chambers in the developing heart. NOTCH modulates cardiomyocyte proliferation and differentiation and is required for ventricular chamber formation and coronary vessel specification [41,43]. NOTCH signaling is also a key regulator of artery formation during arteriovenous specification in the vessel development. The core components of the NOTCH pathway include 4 receptors (NOTCH1–4) and 2 families of ligands (JAG1–2, DLL1, DLL3, and DLL4). Receptor-ligand interactions lead to a series of proteolytic cleavages and ultimately the release of the NOTCH intracellular domain (NICD) from the cell membrane (Figure 3A). This allows for NICD to translocate to the nucleus where it forms an activation complex with the RBPJ (Recombining binding protein suppressor of hairless) protein by evicting a histone deacetylase corepressor (CoR). The activation complex further recruits histone acetyltransferase (HAc) and MAML1 (Mastermind-like protein 1), and triggers the transcriptional activation of NOTCH target genes including the bHLH transcriptional repressors HEY1 (Hairy/enhancer-of-split related with YRPW motif protein 1) and HEY2 [44,45].

Figure 3. NOTCH signaling pathway and roles in ventricular chamber development and heart disease.

(A) Schematic illustration of NOTCH signal transduction. Membrane-bound NOTCH ligands (JAG1–2, DLL1, DLL3, and DLL4) bind to transmembrane NOTCH receptors (NOTCH1–4) in neighboring cells. The interaction between NOTCH receptors and ligands leads to a series of proteolytic cleavages of the receptor. The cleavage by γ-secretase results in the release of NOTCH intracellular domain (NICD) from the membrane. The NICD then translocates to the nucleus and forms an activation complex with the RBPJ by replacing a histone deacetylase corepressor (CoR). The activation complex further recruits the histone acetyltransferase (HAc) and initiates the transcriptional activation of the downstream NOTCH target genes including the transcriptional repressors, HEY1 and HEY2. (B) During embryonic heart development, the crosstalk between endocardial NOTCH receptors and myocardial ligands is required for ventricular trabeculation, cardiac valve formation, and outflow tract development. Dysregulation of cardiac NOTCH signaling in the developing heart leads to congenital heart defects and cardiomyopathy.

In mouse embryos, the Notch receptor (Notch1) and ligands Dll4 and Jag1 are first detected in primitive E7.5 endocardium, suggesting a regulatory role of Notch signaling in early cardiogenesis [46]. In E8.0 embryos, with the separation of the inner endocardium and outer myocardium, Jag1 is expressed in the primitive myocardium, whereas Dll4, Notch1, Notch2, and Notch4 are restricted to the endocardium (Figure 3B). At E9.5, the epicardium, which differentiates from the proepicardial organ, emerges and expresses several Notch components, including Notch1, Dll4, Jag1, Notch2, Notch3, and Hey1, and subsequently migrates to cover the entire myocardial surface by E11.0 [47,48]. Epicardium, myocardium, and endocardium mutually interact, and cooperate in a series of coordinated patterns of proliferation and differentiation to form a mature 4-chambered heart. The endocardium lines the lumen of the cardiac chambers and contributes to the atrioventricular valves and part of outflow tract (semilunar) valves; the myocardium consists of contractile cardiomyocytes; and the epicardium gives rise to smooth muscle cells and vasculature structure. Notch signaling is required for normal ventricular chamber development and spatial allocation of cardiomyocytes to their proper morphological position in the ventricular wall. Notch1 mutant embryos display defective trabeculation with abnormal cardiomyocyte proliferation and differentiation [38,49]. The loss-of-function of Notch1 results in extensive expansion of cardiac progenitor cells (CPCs) and impairs their differentiation through regulation of cardiac transcription factor, Isl1, which is required for CPCs differentiation into cardiomyocytes and smooth muscle cells [50]. Recent studies demonstrate that Notch signaling first incorporates cardiac endocardium and myocardium to sustain trabeculation, and then coordinates ventricular patterning and compaction with coronary artery development to generate a mature ventricular chamber [51]. In addition, pathogenic variants in NOTCH signaling pathway members have been associated with congenital cardiac malformations, such as bicuspid aortic valve and tetralogy of Fallot, along with cardiomyopathy in humans [52–56].

Retinoic acid, cardiogenesis, and congenital heart defects

Retinoic acid (RA), a derivative of vitamin A, is the first diffusible morphogen identified during vertebrate development [57]. RA binds to retinoic acid receptors (RARα, β, and γ), which are nuclear receptors and can act as transcription factors (Figure 4). RARs act as ligand-inducible activators by forming heterodimers with any of three retinoid X receptors (RXRα, β, and γ) [58]. The RAR-RXR heterodimer complex regulates transcription by binding to RA response elements (RAREs) near target genes. In the absence of RA signaling (a repressive state), RAR-RXR heterodimer recruits co-repressors such as nuclear receptor co-repressor 1 (NCOR1) and NOCR2 that further recruit histone deacetylase (HDAC) and Polycomb repressive complex 2 (PRC2), leading to condensed chromatin and gene silencing deposited by H3 lysine 27 trimethylation (H3K27me3) [59,60]. The RA binding triggers a functional change in the RAR-RXR heterodimer, which evicts repressive factors and recruits co-activators such as nuclear receptor co-activator 1 (NCOA1) or NCOA2. These co-activators initiate the recruitment of histone acetyltransferase (HAC) complexes and Trithorax proteins that modulate H3K4me3 and result in permissive chromatin and gene activation [61].

Figure 4. Retinoic acid signaling and its regulatory roles in atrial cardiomyocyte lineage commitment.

Retinol binds to retinol binding protein 4 (RBP4) and is transported to the cytosol by the cooperation of a cell-surface RBP receptor, STRA6 (stimulated by retinoic acid gene 6). With the assistance of cellular retinol bind protein (CRBP), retinol is transformed to retinaldehyde that is catalyzed by retinol dehydrogenase (RDH). Retinaldehyde is further converted to retinoic acid (RA) by the retinaldehyde dehydrogenase (RALDH). RA then binds to the cellular retinoic acid binding protein (CRABP) and translocates into the nucleus, where RA binds to RA receptors (RARs). RARs are ligand-inducible activators by forming heterodimers with any of three retinoid X receptors (RXR). The RAR-RXR heterodimer complex binds to the RA response elements (RAREs) and regulates nearby gene expression. Alternatively, excessive RA is either secreted outside the cytosol as paracrine signal or degraded by the cytochrome P450 family 26 (CYP26) enzyme. During embryonic cardiogenesis, RA signal determines the atrial chamber morphogenesis and mediates atrial cardiomyocyte differentiation, possibly through interacting with the transcription factor COUP-TFII that is a RA-activated nuclear receptor [69] and could potentially suppress the RXR-mediated RA signaling pathway [125,126].

As a vitamin A metabolite, RA plays an important role in early heart development [42]. Its function primarily depends on a tight control of RA distribution within the embryo through the spatiotemporal regulation of the RA synthesizing (RALDH1–3) and metabolizing enzymes (CYP26). Many insights have been gained from studying Raldh mutant embryos. Early studies in zebrafish embryos indicate that RA signaling is necessary to restrict the pool of cardiac progenitor cells and maintain a developmental balance between cardiac and non-cardiac identities [62]. Homozygous knockout of Raldh2 in mouse embryos leads to abnormal development of the SHF with an expansion of Isl1⁺ cardiac progenitor populations and compromised deployment of SHF cells within the heart tube [63]. In addition, RA signaling specifies Tbx5⁺ progenitors of the FHF to a venous and atrial cell type, as blockade of RA synthesis results in an abnormal heart that lacks the atrial chamber. Further investigations have suggested that atrial-specific gene expression is modulated by local synthesis of RA, and RA exclusion is required for the correct specification of ventricles in the heart [42,64].

In vitro studies imply that RA signal can modulate the differentiation of atrial and ventricular cardiomyocytes from mouse and human pluripotent stem cells. RA signaling promotes atrial-specific gene expression to generate atrial-like cardiomyocytes in differentiating PSCs [65–67]. RA-induced atrial specification is mediated by the upregulation of the transcription factor COUP-TFII, a RA-activated receptor and key regulator that determines atrial identity [68,69]. Cardiomyocyte-specific COUP-TFII ablation in mice results in ventricularized atria that display ventricle-like action potentials, increased cardiomyocyte size, and development of extensive T tubules. Nevertheless, the effect of RA on the development of atrial-like myocytes at the expense of ventricular lineage is restricted to a narrow developmental window corresponding to the cardiac mesoderm stage, when distinct mesodermal populations are committed to atrial and ventricular cardiomyocytes [18,19]. The combination of RA, BMP, and WNT signaling directs human PSCs towards epicardial lineage, including cardiac fibroblasts and vascular smooth muscle cells [70,71]. Though RA treatment does not influence the expression of cardiac conduction tissue marker Hcn4 during cardiac differentiation of PSCs [72], the combination of RA and BMP signaling promotes the differentiation of sinoatrial node-like pacemaker cells from human PSCs [73].

Generation of atrial-, ventricular-, and nodal-like cardiomyocytes from human PSCs and implications in cardiovascular disease modeling and precision medicine

With the discovery of human iPSCs, patient-specific cardiomyocytes have been employed to model cardiovascular disease, screen novel drugs, and perform cardiac toxicity testing [14,74]. Cardiomyocyte differentiation protocols have been extensively developed in the past decade: from embryoid body (EB) based 3-D differentiation to small molecule-induced monolayer differentiation [22,23]. Robust cardiomyocyte generation can now be achieved in chemically defined conditions by sequential modulation of WNT signaling [75,76]. Despite an immature and fetal-like phenotype, iPSC-CMs display a mixture of atrial-, ventricular-, and nodal-like electrophysiological features [77]. Subtype-specific cardiomyocytes have been generated by multiple differentiation protocols that employ a cocktail of chemicals and growth factors (Table 1). A number of subtype-specific and genetically engineered reporter PSC lines have been created to facilitate the purification and characterization of subtype-specific cardiomyocytes (Supplemental Table 1). Together with tissue engineering technologies, atrial- and ventricular-specific cardiac tissue has been recently constructed, which is capable of recapitulating chamber-specific electrophysiological spectrum and drug responses [21]. In the following section, we will review the strategies to generate subtype-specific cardiomyocytes from human PSCs (Figure 5) and how they impact current cardiac cell therapy and disease modeling.

Table 1. Subtype-specific cardiomyocyte differentiation by sequential manipulation of developmental signaling pathways during heart development.

This summary is primarily based on the 2-D monolayer differentiation. RA can be added as early as Day 3 for atrial CM differentiation in an EB- based differentiation protocol [18].

Subtypes	Mesoderm induction Day 1–3	Cardiac progenitors Day 3–5	CM specification Day 5–7	References
Atrial CMs	Low Nodal	WNT inhibition	Retinoic acid (RA)	Zhang et al., 2011 [65]
	BMP	RA		Devalla et al., 2015 [66]
	FGF			Lee et al., 2017 [18]
	WNT			Cyganek et al, 2018 [19]
Ventricular CMs	High Nodal	WNT inhibition	RA inhibition	Zhang et al., 2011 [65]
	BMP		Noggin	Lian et al., 2012 [75]
	FGF			Burridge et al., 2014 [76]
	WNT			Karakikes et al., 2014 [93]
Pacemaker CMs	Low Nodal	BMP	WNT	Protze et al., 2017 [73]
	BMP	RA		Liang et al., 2019 [102]
	FGF	WNT inhibition		Ren et al., 2019 [103]
	WNT	FGF inhibition

Figure 5. Strategic protocols for generating subtype-specific cardiomyocytes and their implications in modeling cardiovascular disease and precision medicine.

Pluripotent stem cells are induced to cardiac mesoderm by activation of WNT and BMP signaling. Cardiac progenitors are committed by the administration of WNT inhibitors such as IWR-1. Sequential manipulation of WNT signaling gives rise to robust cardiomyocyte differentiation in a monolayer culture system under chemically defined conditions. For cardiac progenitors, RA signaling directs towards atrial subtype commitment whereas inhibition of RA signaling promotes ventricular lineage differentiation. In addition, RA and BMP cooperate to instruct cardiac progenitors to SAN pacemaker cell fate. Canonical Wnt signaling also promotes pacemaker cell specification from cardiac mesoderm cells. Precision production of subtype-specific cardiomyocytes has the potential to elucidate cardiovascular disease mechanisms and facilitate the development of individual-based therapies. For example, atrial-like cardiomyocytes can be used to model atrial disease for drug discovery, large-scale production of ventricular-like cardiomyocytes may be utilized as a cell therapy for myocardial infarction, and patient-derived pacemaker cells are promising source for biological pacemakers.

Generation of atrial-like cardiomyocytes from human PSCs

Inspired by the elucidation of molecular regulators of chamber formation in the developing heart, scientists have invested much effort on manipulating key signaling molecules to coax human PSCs to atrial-like cardiomyocytes in vitro. One of the crucial players is RA signaling (Figure 4). Using atrial CM-specific reporter cell lines, several groups have found that RA promotes cardiac progenitor cells toward an atrial-like subtype in both mouse and human PSCs [18,19,65–67]. Regardless of monolayer or embryoid body (EB)-mediated differentiation, RA should be administered in a narrow window to direct progenitor cells into an atrial-like lineage. Excessive RA signaling exposure may cause severe cardiac malformations such as enlarged atrial chambers and small ventricles in mouse and chicken embryos [64]. This restrictive temporal window corresponds to mouse embryonic day E7.5 to E8.5 when cardiac mesoderm progenitors are migrating from the primitive streak. Therefore, RA is usually added for a short period of time (2–3 days) to stimulate atrial specification after the cardiac mesoderm is committed during human PSC differentiation [18,19]. In contrast, RA inhibition leads to ventricular cardiomyocyte specification, which will be discussed in detail in the following section.

Recent studies have demonstrated that atrial and ventricular subtypes are derived from distinct mesoderm progenitors that can be distinguished by different cell surface markers [18]. Using different concentrations of BMP4 and Activin A, human PSCs can be induced to distinct mesoderm populations: CD235a⁺ CYP26A1⁺ and RALDH2⁺ ALDH⁺ mesoderm progenitors. In the absence of RA signaling, CD235a⁺ CYP26A1⁺ mesoderm gives rise to ventricular cardiomyocytes whereas RALDH2⁺ ALDH⁺ mesoderm tends to generate atrial subtype. However, in the presence of RA, RALDH2⁺ ALDH⁺ mesoderm efficiently gives rise to atrial-like cardiomyocytes whereas CD235a⁺ CYP26A1⁺ mesoderm inefficiently produces atrial-like cardiomyocytes [18]. Therefore, the balance between RA synthesis and degradation appears to modulate the generation of atrial cardiomyocytes.

An emerging pathway to promote the atrial-specific cell fate is NOTCH signaling. As discussed earlier, NOTCH signaling is required for ventricular chamber specification during heart development [38,41,51]. NOTCH activation inhibits COUP-TFII activity that is essential for atrial chamber formation in the developing heart [68]. Conversely, COUP-TFII suppresses NOTCH activity and may indirectly promote atrial cardiomyocyte specification [78–80]. Loss-of-function of HEY2, a NOTCH pathway downstream transcription factor, dramatically boosts atrial specification by increasing the percentage of atrial-like cardiomyocytes without exogenous RA exposure [80]. As NOTCH signaling also regulates ion channel activity and electrophysiological properties in the heart chamber [81,82], further investigation of its roles in the atrial and ventricular cardiomyocyte specification is warranted.

Generation of ventricular-like cardiomyocytes from human PSCs and use in cell therapy and drug discovery

Human iPSC-derived cardiomyocytes are potential donor cells for regenerating the injured heart caused by ischemic myocardial infarction, as evidenced by functional restoration in left ventricular chamber in non-human primates after the injection of PSC-CMs [83]. However, ventricular arrhythmias have been observed in the human PSC-CM-grafted primates, possibly due to the mixed atrial- and ventricular-like cardiomyocytes in the grafts [84,85]. Therefore, enriching ventricular cardiomyocytes while eliminating other subtype cells is paramount for clinically implementing human PSC-CM mediated cardiac regeneration.

Current cardiomyocyte differentiation protocols generate populations of cells in which more than half is ventricular-like cardiomyocytes, but this is mixed with small numbers of atrial- and pacemaker-like cells. Multiple strategies have been proposed to purify ventricular-like cardiomyocytes for cell therapy purpose (Figure 5). First, a number of ventricular-specific reporter lines have been created to enrich the ventricular cardiomyocyte subtype (Supplemental Table 1) [86]. In these reporter lines, a fluorescence protein is cloned into the endogenous MLC2v locus or driven by an MLC2v promoter for selection of ventricular-like cardiomyocytes [87,88]. This reporter can also be combined with a second cardiac marker such as NKX2.5 for double selection [89]. A dual reporter line NKX2.5^EGFP/+ COUP-TFII^mCherry/+ is generated to segregate atrial from ventricular cardiomyocytes based upon COUP-TFII expression [90]. The EGFP⁺/mCherry⁻ cells exhibit transcriptional and electrophysiological characteristics of ventricular subtype. Recently, NKX2–5^TagRFP and TBX5^Clover2 dual reporters have been engineered to extract FHF-like and SHF-like progenitor cells. TBX5⁺ NKX2–5⁺ FHF-like cells give rise to ventricular-like cardiomyocytes whereas TBX5⁻ NKX2–5⁺ SHF-like progenitors are prone to generate atrial-like cardiomyocytes [33]. In addition, cell surface markers have been sought for the purification of ventricular-like cardiomyocytes. Though SIPRA is a general surface marker for isolating TNNT2⁺ beating cardiomyocytes from human PSCs [91], a CD77⁺/CD200⁻ cell-surface signature is recently identified for the enrichment of ventricular-like cardiomyocytes [92]. Also, a combination of small molecules and growth factors is used to stimulate ventricular subtype specification during cardiac differentiation [93,94]. Sequential administration of BMP4, Rho kinase inhibitor, Activin A, and IWR-1 in cardiogenic embryoid bodies generates over 92% of ventricular-like cardiomyocytes, which is confirmed via ventricular-specific action potential duration and ionic current. Lastly, inhibition of RA signaling indirectly promotes ventricular subtype specification [65]. A pan-retinoic acid receptor antagonist BMS-189453 together with Noggin significantly directs the cardiac progenitors toward a ventricular-like subtype. Alternatively, ventricular-like cardiomyocytes can be derived from CD235a⁺ CYP26A1⁺ mesoderm progenitors in the absence of RA signaling [18].

Generation of nodal-like cardiomyocytes from human PSCs for development of biological pacemaker cells to treat cardiac arrhythmias

The sinoatrial node (SAN) pacemaker cells possess pivotal roles in the cardiac conduction system and control heart rate throughout life. Dysfunction of SAN pacemaker cells leads to the reduction of heart rate and insufficient blood circulation, which currently can be managed by the implantation of an artificial (electronic) pacemaker. Generating biological SAN pacemakers in vitro may be an alternative cell therapy for the failing SAN (sinus arrhythmias) and may potentially replace an electronic pacemaker. In the developing heart, SAN pacemaker lineage specification is modulated by the coordinated interaction of multiple cardiogenic transcription factors that include TBX18, TBX3, SHOX2, NKX2.5, and ISL1 [95–97]. The SAN lineage originates from TBX18⁺ NKX2.5⁻ mesoderm progenitors, and the lack of NKX2.5 expression distinguishes the SAN pacemaker cells from the NKX2.5⁺ atrial/ventricular cardiomyocytes and the atrioventricular node (AVN) pacemakers [98,99]. In addition, hyperpolarization-activated cyclic nucleotide-gated potassium/sodium channel 4 (HCN4) is one of the ion channels that is specifically expressed in SAN pacemaker cells. HCN4 is initially expressed in the FHF, and gradually confined in the SAN during embryonic heart development [100].

Generation of SAN-like pacemaker cells (SANLPCs) from human PSCs has been achieved by using transgene dependent and independent methods in the past decade (Figure 5). By mimicking the developmental scenario of SAN lineage specification, SANLPCs are generated from human PSCs through stage-specific activation of BMP and RA signaling pathways or from expandable cardiac progenitor cells induced by enforced expression of the c-MYC oncogene [101]. The SIRPA⁺ CD90⁻ SANLPCs are derived from NKX2.5⁻ mesoderm progenitors by blocking FGF signaling [73]. These SANLPCs show gene expression and electrophysiological characteristics of pacemaker cells, and successfully function as a biological pacemaker for the host rat ventricular cardiomyocytes upon transplantation. Recent studies have demonstrated that canonical WNT signaling promotes the differentiation of cardiac mesoderm toward pacemaker cardiomyocytes in human and mouse PSCs [102,103].

Forced expression of pacemaker-specific transcription factors such as SHOX2 promotes differentiation of human PSC to pacemaker-like cells [104], and SHOX2 voltage-sensitive fluorescent protein (SHOX2-VSFP) positive cells display nodal-like action potential with high level of HCN4 expression [87]. Forward overexpression of a nodal cell inducer TBX3 in human PSCs can facilitate the generation of functional nodal tissue [105]. Similarly, overexpression of TGF-beta activated kinase (TAK1/MAP3K7) directs cardiac embryoid bodies toward the SAN lineage, suggesting a specific role of MAP3K7 in the differentiation of cardiac conduction system [106]. Recently, Schweizer et al. reported a direct pacemaker cell differentiation protocol by co-culturing with the visceral endoderm-like cell line END-2 and subsequent exposure to FBS-enriched medium [107]. A few SAN-specific markers have been identified to purify and enrich SAN cardiomyocytes from mouse and human PSCs [108]. CD166⁺ cardiac precursors derived from PSCs are thought to develop into functional SAN pacemaker cells [109]. In the developing heart, HCN4 marks the FHF at cardiac crescent stage and distinct cardiac conduction precursors at different developmental stages, then labels the entire conduction system by late fetal stages [110,111]. In contrast, avian cardiac pacemaker cells are not derived from Hcn4⁺ cells in the FHF [112]. HCN4 is dynamically expressed in PSC-derived pacemaker cells and may be insufficient to identify pacemaker-like cardiomyocytes [113]. Additional chamber-specific structural and functional features should be carefully considered for the phenotypic characterization of differentiated human PSC-CMs [114].

Conclusions and perspectives

Precise generation of subtype-specific (atrial-, ventricular-, and nodal-like) cardiomyocytes will significantly improve the translational applications of patient-derived PSCs for disease modeling, cell therapy, and drug discovery. Atrial-like cardiomyocytes are particularly suitable for modeling atrial arrhythmias, such as atrial fibrillation. Large-scale production of pure ventricular-like cardiomyocytes has the potential to facilitate the restoration of cardiac function after myocardial infarction and abrogate possible ventricular arrhythmias in the graft after transplantation. Biological SAN pacemaker cells represent a promising alternative to artificial pacemakers that have serious drawbacks especially for newborns and children with heart block. Efficient methods of deriving atrial-, ventricular-, and nodal-like cardiomyocytes into homogenous populations will be a future challenge and may be circumvented by the identification of cardiac progenitors committed to a specific subtype lineage. Direct reprogramming of somatic cells [115] and forward programming of human PSCs by lineage-specific transcription factors [116] will advance our understanding on how diverse cardiac subtype identity is established and maintained during heart development. Recent emerging technologies, such as single-cell genomics [117], cardiac organoids [118], and CRISPR/Cas9-mediated genome editing [119] will propel the implementation of precision medicine in cardiovascular disease using more refined subtype-specific cardiomyocytes derived from individual patients.