Molecular pathways in myocardial development: a stem cell perspective

Abstract

The heart has long been considered to adapt to increased work or pathology through the cellular growth process of hypertrophy. However, recent evidence for the existence of endogenous stem cells and regenerative capacity in the adult heart has given new impetus to the quest for cell therapies for heart failure, which remains the number one killer in Western cultures. The molecular cues driving cardiac development are now being explored in detail and will come into sharp focus as regimes for stem cell differentiation and efforts to augment endogenous regeneration are trialed. This review briefly outlines the current state of knowledge on the molecular basis of the four modalities of myogenesis that have been identified in the developing vertebrate heart. Stem cell-mediated myogenic repair in the heart represents a fifth modality, and an exciting frontier with basic and practical implications that remain to be explored.

Stem cells, Cardiac stem cells, Heart development, Cell therapy, Cardiomyopathy

Time for primary review 33 days.

1 Introduction

Stem cells are the essential building blocks of metazoans. Until recently, it has been assumed that embryonic stem cells become more and more restricted in their developmental potential during ontogeny, and that a limited number of stem cell populations persist in the adult as dedicated servants of single regenerative organ systems such as blood, skin, gut, liver etc. Findings that stem cells are much more widespread in adult tissues than anticipated, and that embryonic and adult stem cells have unexpectedly broad developmental potential, even able to cross germ-layer boundaries, has challenged our ingrained beliefs about developmental processes [1].

Biologists and clinicians have been quick to grasp the great potential of stem cells or their differentiated progeny for cellular therapies in the treatment of degenerative diseases and diseases of the aged. Differentiation of embryonic or adult stem cells in vitro into organ-specific progenitors or more mature cells has potential to revolutionise therapies for ischaemic heart disease, diabetes, stroke, Parkinson's disease, skeletal myopathies and perhaps a host of other conditions. It may also be possible to re-stimulate organ-specific regenerative processes that are present in lower vertebrates but have become attenuated in mammals [2].

The heart has long been thought to adapt to increased work and compensate for disease exclusively through the cellular growth process of hypertrophy. However, recent evidence for the existence of endogenous stem cells and regenerative capacity in the adult heart has given new impetus to the quest for cell therapies for the diseased or failing heart [3]. A number of cell grafting procedures have already been trialed in animals, including the use of embryonic, foetal and adult cardiomyocytes, cardiomyocyte tumour cells, smooth muscle cells, dermal fibroblasts, skeletal myoblasts, ES-derived cardiocytes and adult stem cells [3]. Practical, biological and ethical issues are associated with some of these approaches. However, the rational guided development of stem cell populations into cardiac muscle lineages in vitro remains a promising way forward for this field. Mobilisation or augmentation of endogenous stem cell populations capable of multi-lineage heart regeneration also appears feasible [4]. Jessell and colleagues have recently reported the orchestrated production of spinal motor neurons from mouse ES cells in vitro [5], an undertaking that relied on a sophisticated knowledge of neurogenic and patterning pathways in the embryo. A similar understanding of cardiac developmental pathways may be of enormous strategic benefit in devising or refining cellular therapies for the heart. The purpose of this review is to briefly outline what is currently known about the genes and developmental pathways that specify the cardiac myocytes of the vertebrate heart. Our understanding of these events is evolving rapidly and we are progressing towards a blueprint for the cardiac muscle program found in the embryo that will be the basis of future work in molecular cardiac bio-engineering.

2 Overview of cardiac morphogenesis

Cardiac myocytes are derived from the mesodermal germ layer, and the position of their progenitor cells can be mapped within the primitive ectodermal layer prior to gastrulation. The bilaterally-arranged progenitors migrate through the node and primitive streak at gastrulation [6], and move to the anterior and anterior/lateral aspects of the embryo to form the so-called paired heart fields [7] which then converge anteriorly over the foregut lip to form the cardiac crescent (Fig. 1). Heart field cells are defined as those that have cardiac developmental potency when explanted and placed into culture [8]. In lower vertebrates such as amphibia, heart fields may include cells that do not ultimately contribute to the heart, but can do so in a regenerative mode if definitive precursors are removed or damaged [9]. In a progression associated with foregut pocket formation, heart progenitor cells move ventrally and fuse at the midline to create the linear heart tube. This consists of an inner endothelial tube shrouded by a layer of myocardium that remains attached to the ventral aspect of the foregut. The elongating heart tube then largely loses its connection to the foregut and adopts a rightward spiral in a process termed cardiac looping morphogenesis. It is during looping that the primitive cardiac chambers become evident as a pattern of swellings and constrictions along the heart tube [10]. Looping is guided by an embryonic left/right axial pathway that determines the rightward direction of ventricular bending and distinct morphological identities of left and right atria. At the completion of looping (around E12.5 in the mouse), the heart has assumed a form that closely resembles that of the adult organ. There is extensive remodelling of the internal structures of the heart during looping, with septation and valvulogenesis completing the separation and connectivity of the chambers.

Fig. 1

Heart development. Stylised pictures (above) and sections (below) depict the major transitions in early mammalian heart development. Dark shading represents myocardial tissue. AV, atrioventricular; LA, left atrium; LV, left ventricle. Reproduced with permission from Harvey et al., 2003 [149].

The genetic pathways underlying cardiogenesis are complex and interconnected (Fig. 2). There are several modalities of cardiac myogenesis that may differ in mechanism. Furthermore, inductive interactions between the endocardium and myocardium, and epicardium and myocardium, are essential for correct growth and differentiation of chamber muscle. The morphological and molecular aspects of these processes will be discussed in further detail below.

Fig. 2

Lineage map of cardiogenesis. A brief summary of the complex lineage relationships (solid lines) and interactions (dashed lines) that occur during vertebrate heart development. See text for details. AV, atrioventricular; OFT, outflow tract.

3 Induction of cardiogenesis in the heart fields: the role of bone morphogenetic proteins

Numerous cardiac-restricted transcription factor genes are induced in the heart fields [11]. The earliest known gene encodes the zinc finger transcription factor, GATA4. Gata4 is expressed in anterior endoderm and mesoderm and may help restrict mesoderm to a cardiac fate [12]. Later, this gene is vital for activation of many myocardial and endodermal differentiation genes [13]. Nkx2-5, encoding a homeodomain factor, is also expressed early in the cardiogenic program and is often used to delineate cardiac progenitors [9,14,15]. Induction of these genes occurs before the first appearance of endocardium and before imposition of the intra-embryonic coelom, which separates heart precursors from pericardial mesoderm (Fig. 2). Thus, the initial heart precursor cells are the common progenitors of myocardium, endocardium and pericardial mesoderm [14]. A large body of evidence shows that factors secreted from endoderm immediately juxtaposed to heart mesoderm play a key role in cardiac induction [12,16] (Fig. 3). Several members of the bone morphogenetic protein (BMP) family of secreted signalling molecules are expressed in endoderm of the cardiogenic region as well as in ectoderm and cardiac mesoderm itself [17–19]. Application of Bmp2 or Bmp4 to explants of cardiac or non-cardiac regions of chick embryos induces expression of early cardiac markers such as GATA4, Nkx2-5, Tbx2 and MHC, and a beating phenotype, whereas inhibition of BMP signalling blocks expression of Nkx2-5 and cardiac differentiation [17,20–22]. The dose of BMP seen by cardiac progenitors may be critical for correct induction [12,23]. In contrast to explant studies, ectopic placement of BMP-soaked beads or BMP-expressing cells into head mesoderm can induce cardiac markers such as Nkx2-5, but does not induce myosin expression or a beating phenotype [17,20]. Full cardiogenesis in explants may therefore be due to the interaction of BMP signals with other components, perhaps fibroblast growth factors (FGFs; see below), present in serum or embryo extract used in the culture medium [23]. Genetic dissection of the role of BMPs in mice appears complicated by genetic redundancy. However, some embryos lacking Bmp2 do not form a heart at all [24]. Furthermore, the hearts of most BMP2 mutant embryos, and those lacking its downstream effector gene Smad5, are abnormally placed and develop poorly [22,25,26]. Dominant inhibition of BMP signalling in frog and chick embryos inhibits cardiogenesis [27–29].

Fig. 3

Positive and negative signalling inputs which shape the cardiac progenitor fields. Figure shows a stylised transverse section through the posterior region of the cardiac crescent of a mouse embryo at E7.5–8.0 (Fig. 1). Arrows indicate positive inductive signals from tissue layers neighbouring the myocardium. Bars indicate negative influences. Note that the intraembryonic coelom is shown as formed in this figure. This cavity separates the cardiac mesoderm (ventral) from pericardial mesoderm (dorsal). However, early cardiogenic signals act prior to coelom formation. Adapted with permission from Harvey, 2002 [150].

Smad proteins are transcription factors that are positively regulated by phosphorylation as a result of signalling downstream of TGFβ receptor family members. Smad1, 3 and 5 are regulated by BMPs and during cardiac induction, BMP-Smads appear to directly activate early cardiac transcription factor genes, including Nkx2-5[30–32]. It is now well established that heart formation in the fruitfly, Drosophila, is guided by genetic pathways homologous to those used in vertebrates [33–35] and the direct action of BMP-Smads on cardiac transcription factor genes is one of the features of these pathways that has been conserved in evolution [36].

4 Upstream of bone morphogenetic proteins

The organiser, or node-equivalent in Xenopus, has long been known to be critical for cardiac induction [37]. Several recent studies have noted that BMP signalling, while maintaining cardiac regulatory gene expression, is not involved in the very earliest phases of induction [27–29]. Furthermore, in the mouse, an early phase of Nkx2-5 expression is lost in embryos lacking the Smoothened (Smo) gene, which encodes a membrane-bound signalling protein for members of the Hedgehog family of secreted factors [38]. Expression recovers to normal levels by E9.0. Although these findings need extending, data suggest that the earliest stages of Nkx2-5 expression are initially regulated by a Hedgehog signal. Cardiac progenitors may be exposed to Hedgehog signals from the node (Sonic and Indian Hedgehog) and/or visceral endoderm (Indian Hedgehog) [39]. Indian Hedgehog can induce expression of the Bmp4 gene [39], possibly connecting early and later inductive events in the forming heart. The maintenance role played by BMPs in early cardiac induction is also seen in Drosophila. Here, the fly homologues of BMP2 and 4, called Dpp and Screw, are required to maintain expression of tinman, the homologue of Nkx2-5, in dorsal mesoderm that carries the heart progenitors [35].

5 A requirement to block Wnt/β-catenin signalling

The cardiac crescent is shaped not only by the positive influences of BMPs and other inducers, but also by repressive signals emanating from axial mesoderm and the neural plate (Fig. 3). BMP induction of cardiac markers in chick anterior mesendodermal explants is almost completely blocked by juxtaposed explants of neural tube and notochord [40]. However, Nkx2-5 expression is still induced in these experiments, suggesting that signals from axial tissues prevent propagation of the cardiac program downstream of initial Nkx2-5 expression. Signals from the notochord may include BMP antagonists such as noggin and chordin [41]. Furthermore, two members of the Wnt family of secreted glycoproteins, Wnt-1 and Wnt-3a, are expressed in the dorsal neural plate at the time of cardiac induction and can mimic its repressive effects in culture [42]. Indeed, ectopic expression of a Wnt antagonist can overcome the repressive effects of the neural plate on cardiogenesis [42]. Wnt antagonists of the Dickkopf and Frizzled-like families are expressed in the organizer/node as well as in endoderm neighbouring the heart mesoderm [43–45]. Importantly, these antagonists can activate cardiogenesis in non-cardiac mesoderm of frog and chick embryos, leading, in the frog, to ectopic beating heart tubes lined with endothelial cells [43,44]. Thus, elimination of Wnt/β-catenin signalling is essential for cardiogenesis and the balance of Wnts and their antagonists appears to play a critical role in shaping the cardiac fields.

β-Catenin is one of the best characterised effectors of the Wnt pathway [46]. Stabilisation of cytoplasmic β-catenin within the ‘canonical’ Wnt signalling pathway allows it to enter the nucleus, associate with transcription factors of the TCF/LEF family, and activate transcription of target genes. The serine/threonine kinase, GSK3β, whose normal role is to destabilise β-catenin in the absence of Wnt signalling, can also initiate cardiogenesis in non-cardiac mesodermal explants from Xenopus when over-expressed [43]. Remarkably, targeted inactivation of β-catenin in the node, notochord, and endoderm of the developing mouse results in the formation of ectopic hearts along the anterior/posterior axis [47], consistent with the need to block Wnt/β-catenin signalling for vertebrate cardiogenesis.

6 The Wnt/JNK pathway

Recent observations suggest that certain Wnt proteins activate a non-canonical cytoplasmic signalling pathway [48]. Wnt-11, which is expressed in mesoderm, has been implicated in regulating cell polarity and movement during gastrulation through activation of c-Jun-N-terminal kinase (JNK) and the small GTPases RhoA and cdc42 [49]. In contrast to the Wnt/β-catenin pathway discussed above, which is inhibitory for cardiogenesis, the non-canonical Wnt/JNK pathway is essential for cardiac induction in the frog and chick embryo systems [50,51]. In frogs, ectopic expression of Wnt-11 or activation of both protein kinase C (PKC) and JNK induces a cardiogenic phenotype [50]. Wnt-11 also induces cardiogenesis in mouse pluripotent P19 teratocarcinoma cells which normally only form beating cardiac muscle after treatment with DMSO [50]. Inhibition of Wnt11, or of PKC or JNK, inhibits endogenous or induced cardiogenesis. An interesting twist is that Wnt-11 also inhibits the canonical Wnt/β-catenin pathway, and that Dickkopf and Frizzled family Wnt inhibitors actually stimulate the Wnt/JNK pathway at the same time as blocking Wnt/β-catenin signalling [50]. Thus, activation of Wnt/JNK signalling by factors expressed in both endoderm and mesoderm promotes cardiogenesis, while inhibitory Wnts expressed in the neural plate help to limit the extent of the cardiac fields (Fig. 3).

7 The role of fibroblast growth factors

In chick embryos, anterior endoderm can induce cardiogenic mesoderm in posterior primitive streak explants [44]. However, this does not occur with BMPs, Wnt/β-catenin antagonism and Wnt/JNK activation alone. Other positive factors are clearly required. The FGF family of signalling molecules is also present in endoderm implicated in modulating cardiogenesis [52] and explant experiments reveal that FGF-4, in the presence of BMP-2, can induce cardiac differentiation in certain non-cardiac mesodermal explants [21,53]. Each of these signals, particularly that of FGF, is required only transiently to induce non-cardiac cells to a cardiac fate [23]. Both are required to sustain expression of Nkx2-5 and the serum response factor gene (SRF). In a recent study, ectopic application of FGF-8, expressed in the endoderm of the heart-forming region of the chick [12] and mouse [54], can prevent the loss of cardiac gene expression in embryos in which inducing endoderm has been resected [12]. Furthermore, the expression of cardiac markers is expanded when FGF-8-soaked beads are placed lateral to the heart fields [12]. In keeping with previous explant experiments [21,53], FGF-8 only has a cardiogenic effect in regions where BMP signalling is also present. Furthermore, fgf8 gene expression appears to be downstream of BMPs [12].

The above findings strongly suggest that BMP and FGF-specific pathways interact to specify the cardiac lineage. However, support for an essential role for FGFs in cardiac induction is still lacking. Studies in the chick system using inhibitors has thus far failed to show that FGF signalling is essential for inductive events [12,55]. The fgf8 gene is needed for normal cardiac morphogenesis and gene expression in zebrafish [56], although relative roles in mesodermal migration, a clear function for the mouse gene [57], and induction, have not been delineated.

8 Signalling downstream of BMPs and FGFs

Cardiogenic signalling pathways downstream of BMPs and FGFs are now being elucidated. Recent studies in the P19 teratocarcinoma system as reviewed elsewhere in this issue [58], in which cardiogenesis is induced by DMSO, show that myogenesis is dependent upon BMPs, BMP-Smads and induction of cardiac transcription factors Nkx2-5 and GATA4 [59,60]. However, also required are the mitogen-activated protein kinase kinase kinase (MAPKKK), TAK1, and the ubiquitously-expressed ATF/CREB family transcription factor ATF2. TAK1 can be activated by TGFβ pathways and, in turn, activates p38 MAP kinase, which phosphorylates ATF2. ATF2, which has intrinsic histone acetyl-transferase activity, can then associate with Smad hetero-oligomers and stimulate transcription. Raf1, a MAPKK normally associated with signalling pathways stimulated by FGFs, cannot substitute for TAK1 in BMP-mediated cardiogenesis. However, like Raf1, TAK1 can activate JNK and p38 kinase leading to activation of transcription factors of the AP-1 family, which can also associate with TGFβ-Smad proteins and co-activate transcription [61].

9 Other cardiogenic inducers

Numerous other signalling factors have been implicated in cardiogenesis in vivo or in vitro, although their roles have been less extensively studied. Cripto-1 belongs to the EGF-CFC family of extracellular membrane-tethered proteins, members of which serve as obligatory co-receptors for signalling by the TGFβ-family member Nodal [62]. Targeted disruption of the mouse Cripto-1 gene causes embryonic lethality after gastrulation [63]. Mutant embryos do form some mesoderm, although no cardiac markers are expressed. Furthermore, differentiating mutant ES cells can express brachyury (a marker of mesoderm) and Nkx2-5, although other markers of the cardiogenic program are absent.

Activin and platelet-derived growth factor have been shown to promote cardiogenesis in various contexts [21,64]. Activin potentially promotes anterior character in mesendoderm, thereby creating a permissive environment for BMP/FGF-mediated cardiogenesis [21]. Cerberus is a TGFβ-superfamily member that can antagonise both BMP and Nodal signalling by direct ligand binding [65]. Despite its anti-BMP activity, it has been shown to induce ectopic head-like structures sometimes containing hearts in Xenopus embryos. In Xenopus ectodermal explants, it can induce Nkx2-5 expression although not full cardiogenesis. The opioid prodynorphin B has also been found to promote cardiogenesis in P19 cells in the absence of DMSO, possibly through nuclear opioid receptors [66]. Opioid receptor antagonists inhibit cardiogenesis in this system. Interestingly, dynorphin B induces activation and intra-cellular translocation of PKC isoforms [67], perhaps connecting this system to the Wnt/JNK pathway.

10 The secondary heart field

Recent papers describe a novel population of cardiac precursor cells termed the anterior or secondary heart field, that give rise to myocardium of the outflow tract [68–70]. Fate mapping studies using vital dyes and retrovirus lineage tags in chick and mouse embryos, as well as the expression of a mouse nlacZ-transgene inserted near the Fgf10 locus, reveal that cells dorsal and anterior to the heart migrate into the cranial part of the formed heart tube to build the outflow tract and possibly (in the mouse) right ventricular myocardium (Fig. 4). Although the extent of the secondary heart field has not been mapped in detail, it may include pericardial and head mesoderm, and/or branchial arch mesenchyme of neural crest origin [68–70]. Non-myogenic cells dorsal to the heart tube express cardiac transcription factors genes such as Nkx2-5 and GATA4, as well as growth factor genes fgf8 and fgf10[14,69,70]. In vitro analyses demonstrate that signals from existing outflow tract myocardium can recruit cells from the secondary heart field to a myocardial fate [68,69], an effect that is blocked by addition of the BMP antagonist Noggin [69]. Patterns of fgf10-LacZ expression suggest that secondary heart field cells arise, at least in part, from cells that lie medial to the primary heart fields that form the initial heart tube [70] (Fig. 4). These cells are apparently held in reserve before being deployed to make a major contribution to heart morphogenesis. The presence of FGFs and dependence on BMPs suggest that the primary and secondary heart fields share common mechanisms of myogenesis.

Fig. 4

Relative contributions of the primary and secondary heart fields to the developing mouse heart. Drawings depict the dynamic relationship between the primary (light-shaded) and secondary (dark-shaded) heart fields from the cardiac crescent through looping stages of heart development (see text). Secondary heart field cells appear to contribute to the right ventricle in mouse [70], although this has not been described in birds [68,69]. Adapted with permission from Kelly et al., 2001 [70].

11 Notch and early cardiogenesis

In Xenopus, the heart field occupies a ventral mesodermal region that closely corresponds to the expression domain of Nkx2-5[9]. The Xenopus heart field is ‘regulative’ in the sense that cells located laterally in the field can be recruited to form a normal heart if definitive progenitors are removed. Regulating cells normally form the dorsal mesocardium and dorsal pericardial mesoderm [9] (Fig. 1). Over time, heart forming potential in the field becomes restricted to the definitive progenitors without a corresponding reduction in the expression of Nkx2-5. Thus, restriction in potency occurs downstream of Nkx2-5.

Recently, Notch-1 and its ligand Serrate have been implicated in the progressive loss of cardiac potency in the Xenopus heart field [71]. Notch signalling mediates numerous cell fate decisions in vertebrates and invertebrates, including roles in repression of differentiation and allocation of alternative cell fates [72]. Initially, Notch-1 and Serrate are co-expressed with Nkx2-5 throughout the cardiac field, although Serrate-1 becomes restricted to cells of the future pericardial roof and dorsal mesocardium, which do not ultimately become myogenic [71]. Activation of the Notch pathway inhibits myogenesis and increases markers of dorsal non-myogenic tissues. Inhibition of the pathway has the converse effect.

The lateral regulative cells of the heart field in Xenopus may be homologous to cells of the secondary heart field in amniotes. Although the extent of secondary heart field cells has not been mapped in detail, it is appealing to believe that a regulative heart field in lower vertebrates has been coopted to drive one of the major morphogenetic innovations that has occurred during evolution of mammals. The implication is that Notch signalling inhibits differentiation of both the primary and secondary heart fields until their staged deployment in the heart.

12 Specialisation of chamber myocardium

Anatomical, electrophysiological and gene expression data suggest that the working muscles of the cardiac chambers are formed as a specialisation of a more primitive type of muscle found in the primary heart tube [10]. Chamber specialisation occurs in discreet zones at the outer curvature of the looping heart tube, implicating anterior/posterior and dorsal/ventral patterning processes in their definition [10]. These zones are marked by expression of several genes, such as those encoding transcription factors Hand1, Cited1 and Irx1/2/3, gap junction proteins connexin 40 and 43, the secreted peptide atrial natriuretic factor (ANF), and the cytoskeletal protein Chisel [10,73–75]. Other genes not strictly chamber-specific in their expression, including some myofilament genes, become up-regulated in chamber myocardium as development proceeds [10]. Thus, formation of definitive chamber muscles requires a regional myogenic specialisation leading to unique contractile, conduction and cytoskeletal properties appropriate for function.

The spatial specificity of chamber formation is guided, in part, by transcriptional repression within the myocardium. The T-box family transcriptional repressor, Tbx2, is expressed in an evolving pattern in the forming heart tube that is mutually exclusive of the zones that become chamber myocardium [76]. Furthermore, Tbx2 acts inter-dependently with Nkx2-5 on a repressive element in the proximal promoter of the ANF gene (a marker of chamber myocardium). Such repression results in down-regulation of ANF gene expression in non-chamber myocardium.

The endocardial layer of the heart also appears to be the source of a key determinant for chamber myocardium. Trabeculae are the spongiform layer of myocytes that form on the inner surface of the developing chambers next to the endocardial layer. Trabeculae are ultrastructurally more differentiated and less proliferative that other layers of the heart [77] and are likely to be the force-generating component of the embryonic ventricle. Trabeculae also have privileged conduction properties and in mice appear to be the precursors of the Purkinje fibers of the conduction system, which are a specialised form of chamber-specific myocyte [78,79]. In mouse, deletion of genes encoding neuregulin-1, an EGF-related signalling molecule expressed in endocardium, or its myocardial receptors ErbB2 or ErbB4, leads to virtually complete elimination of trabecular myocardium [80–82]. Exogenously-delivered neuregulin-1 induces excessive trabeculation and up-regulation of a transgenic marker of the Purkinje system [79,83]. These studies highlight the key role played by endocardium in promotion of chamber muscle and Purkinje fibre differentiation. Recent studies from this laboratory show that neuregulin-1 acts by maintaining expression of a host of cardiac transcription factor genes (unpublished data). The neuregulin-1/ErbB signalling axis is also a homeostatic system for cardiac muscle, its loss in the adult leading to dilated cardiomyopathy and drug susceptibility in myocytes [84,85]. Interestingly, formation of trabeculae in the embryo and their differentiation can be uncoupled. In C3H/HeJ mice lacking the jumonji gene, which encodes a nuclear protein normally expressed in the trabecular layer of the heart, ventricles become filled with trabeculae but their differentiation is inhibited [86].

The endocardium becomes regionally induced by myocardium and contributes profoundly to heart valve and septal morphogenesis [87]. The endocardium associated with chamber myocardium is also a specialised entity, selectively expressing the homeodomain factor gene Irx5[74]. The cardiac extracellular matrix, termed cardiac jelly, elaborated mainly by myocardium, is essential for correct propagation, integration and/or stabilisation of endocardial signals. This is exemplified in mice lacking the major hyaluronic acid synthase gene, Has2, in which trabeculae and endocardial cushions are completely lacking [88]. ErbB2 signalling in endocardial cushion explants of Has2 mutant embryos is rescued by treatment with neuregulin-1 or hyaluronic acid, suggesting an intimate link between neuregulin signalling and the cardiac matrix [89]. In mice lacking the matrix protein versican, endocardial cushions do not form and the right ventricle and conus are severely under-developed [90].

13 Control of myocardial growth by epicardium

The epicardium is the outermost layer of the heart which gives rise to the coronary circulation, including its smooth muscle cells, and interstitial fibroblasts [91]. This cardiac layer is derived from the proepicardial organ, an out-pocketing of the septum transversum mesenchyme lying adjacent to the venous tributaries of the heart. While cardiomyocyte proliferation in the early phase of chick heart development is dependent on FGF signalling [92], later proliferation of chamber myocardium occurs principally in the sub-epicardial layer [93], implicating epicardium in chamber growth. A host of genes have been implicated in growth and differentiation of chamber myocardium through analysis of knockout mice, including those encoding retinoic acid and retinoid-X receptors (RARs and RXRs, respectively), VCAM-1, α4-integrin, TEF-1, erythropoietin, erythropoietin receptor, WT-1 and gp130. How most of these genes act is unclear, especially in the light of evidence that the roles of RXRα, gp130 and erythropoietin receptor in chamber growth are non-cell autonomous for myocytes [94–96]. Nevertheless, retinoic acid (RA) signalling may be a key component of epicardium-mediated chamber growth. Retinaldehyde dehydrogenase 2 (RALDH2), the rate-limiting enzyme in RA synthesis in the embryo, is highly expressed in epicardium [97]. Furthermore, a RA-responsive-LacZ transgene is expressed during chamber growth in both epicardium and myocardium. Mouse embryos lacking Raldh2, or various RAR and RXRs, display thin-walled ventricles and precocious differentiation of myocytes [98].

14 Myocardialization

A fourth modality of myogenesis in the heart has been termed ‘myocardialization’ (Fig. 5). This refers to the relatively late muscularisation of endocardial cushion tissues of the atrium, atrioventricular canal and outflow tract, and of the caval and pulmonary veins [99,100]. Its degree is species-specific, more extensive in endocardial cushions of the chick, for example, compared to mammals. Competence to form cushion myocardial networks is restricted to the more primitive (primary) myogenic zones of the heart [10]. It does not occur adjacent to, nor can it be directed by, specialised chamber myocardium. Within cushions, myocardial networks always form in continuity with existing myocardium (Fig. 3), implying a migratory mechanism. However, elegant use of explant assays and conditioned media show that cushion mesenchyme of endothelial origin can be recruited directly to the myogenic lineage, suggesting that de novo myogenesis is at least one mechanism for myocardialisation [99,100]. The molecular details of this mode of myogenesis are completely unknown.

Fig. 5

Myocardialization of the outflow septum in the developing chicken heart. Brightfield micrographs of sections from chick embryos taken at the level of the outflow tract stained with a monoclonal antibody recognizing myosin heavy chains. (A,B) At Hamburger and Hamilton Stage (H/H) 30, the endocardial ridges are being invaded by cardiomyocytes. (C,D) Muscularization encompasses the entire cushion by H/H 32. Boxes in A,C represent areas shown in B,D. AO, aorta; LA, left atrium; OTC, outflow tract cushion tissue; RA, right atrium; RV, right ventricle. Adapted with permission from van den Hoff et al., 1999 [99].

15 Transcription factor programs in the myocardium

Cardiac myogenic and morphogenic programs are regulated by interconnected networks of transcription factor genes [11]. Included in these networks are the ancient factors that play central roles in all muscle lineages, such as the MADS-box proteins SRF and Mef2A/B/C/D [101]. More cardiac-restricted factors include Nkx2-5/2-6, GATA4/5/6, Tbx2/5/20, Hand1/2, Irx1/2/3/4, CITED1/2, COUPTFII, Hey1/2, Pitx2 and myocardin. Rapid accumulation of data over the past decade has given us a flavour of the complexity of these circuits although an integrated model is still lacking. Transcription factor programs are subject to complex inputs and are interpreted and integrated at the level of individual target gene enhancers. The details of individual cardiac transcription factors and their mutant phenotypes have been recently reviewed [11,102–104], and we will restrict comments here to some general features relevant to myocyte heterogeneity.

15.1 Positive transcription factor circuits in cardiomyogenesis

Cardiac induction in the embryo is accompanied by a positive feed-forward and mutually cross-regulatory circuit [11]. Thus, both Nkx2-5 and Hand2 are regulated by GATA factors [105,106], and gata6 is regulated by Nkx2-5 [107]. Likewise, Nkx2-5 is required for full expression of Hand1 and Irx4 in vivo [108,109], Irx4 is required for expression of Hand1[110], Hand2 and Tbx5 for Irx4[108,111], and so on [11]. Furthermore, many cardiac transcription factors interact directly, allowing greater stability of transcriptional complexes on DNA, synergistic activation of target genes and recruitment of individual factors to enhancers that lack their specific binding sites [112–114]. Nkx2-5 has been shown to interact with SRF, GATA4, and Tbx5 [114–116], while GATA4 can also interact with Mef2C, FOG2 and NFATc [112,117,118] and SRF with myocardin and GATA4 [113,119]. Few of these interactions have been tested genetically, although one study characterised mice carrying a point mutation in GATA4 that eliminates its association with FOG2 [120]. These mice have defects similar to those lacking FOG2, and a subset of those seen in mice lacking GATA4 completely. While there does not appear to be a single cardiac master regulatory gene, the combinatorial mode of regulation described above may account for why individual factors, such as Nkx2-5, GATA4 or Mef2C, can activate the cardiogenic program in pluripotent P19 cells without DMSO [121]. Complex post-translational inputs also regulate the activity of cardiac transcription factors. Both Nkx2-5 and Mef2 proteins are positively regulated by phosphorylation [122,123], and GATA4 is phosphorylated by p38 MAP kinase acting downstream of Rho-family GTPases, promoting transcriptional activation of the RhoA gene itself and sarcomere assembly [124].

15.2 Negative regulatory circuits

Cardiac transcriptional programs will also be subject to complex negative regulatory circuits, which may dampen positive pathways and guide region-specific myogenesis and morphogenesis. The importance of such pathways is perhaps highlighted by the exquisite dose-sensitivity of cardiac transcription factors [125]. The inhibitory Smad protein, Smad6, is expressed in the developing heart and may modulate BMP and/or TGFβ-pathways [29,126]. Numerous other transcriptional repressors implicated in heart development have been described [76,110,127–131]. MEF2 proteins may be universally repressed by direct association with histone deacetylases (HDACs), becoming de-repressed by Ca²⁺-dependent signalling pathways that export HDACs from the nucleus [132]. Nkx2-5 activates two negative feedback cascades in vivo. Acting through a GATA factor-dependent promoter element, it induces the ankyrin-repeat protein CARP, which represses multiple cardiac promoters [127,133]. It also directly induces HOP, a minimal homeodomain protein that associates with and represses SRF-dependent gene expression [129,130].

15.3 Modular gene regulation in the heart

Analysis of cardiac phenotypes in zebrafish and mice, and the patterns of multiple mouse LacZ transgenes carrying cis-regulatory elements of cardiac-expressed genes, suggests a modular basis for gene regulation in the heart [106,133–136]. The implication of these findings is that the complete expression pattern of at least some cardiac genes will be a composite of sub-patterns controlled by different region-specific regulatory modules (reviewed in this issue by Habets et al. [137]). This may reflect the regional diversity of regulatory mechanism and functionality that likely occurred during addition of new regulatory ‘modules’ to the heart in the course of evolution [134]. Stable expression of most transgenes occurs only in the right ventricle and outflow tract, likely reflecting regulation of enhancers active in cells of the secondary heart field. The mechanisms underlying this modular regulation are unknown, although bHLH transcription factors Hand1 and Hand2, and the T-box factor, Tbx5, which show differential expression and function between left and right ventricles, are suggested to be involved [11].

15.4 Atrial and ventricular transcriptional programs

The structural and functional characteristics of the atria and ventricles are distinct. Transcriptional repression and mutually antagonistic pathways appear to be mechanisms utilised to establish chamber-specific identity. There is now considerable evidence that the morphogen, RA, is critical for the initial diversification of the cardiac lineage into atrial and ventricular phenotypes. In the chick and mouse, excess RA causes ventricular chambers to acquire atrial characteristics [138,139]. These teratogenic effects occur only if RA is administered very early in heart development, suggesting a role for RA in the earliest stages of heart tube patterning. Expression of the Raldh2 gene, encoding the enzyme responsible for virtually all RA synthesis in the embryo [97], is initially restricted to the sinuatrial region (primitive atrium and inflow tributaries) of the forming heart, and mouse embryos lacking Raldh2 show unlooped hearts without a distinct atrial chamber [98]. This is also true if RA signalling is inhibited by other means [139,140]. The T-box transcription factor gene Tbx5, mutated in Holt-Oram syndrome in humans, may mediate some of the effects of RA [111]. Tbx5 is initially expressed throughout the cardiac crescent and linear heart tube, but rapidly adopts a graded distribution, high in the atria, low in the left ventricle and largely absent in the right ventricle and outflow tract [141]. Tbx5 is induced by RA and its graded distribution in the heart is flattened in embryos lacking Raldh2[98]. The hearts of Tbx5-null embryos have a severely hypoplastic sinuatrial region [111], and enforced mis-expression of Tbx5 in the ventricles induces abnormal ventricular morphology and down-regulation of ventricle-specific markers [142]. The orphan nuclear receptor and transcriptional repressor, COUP-TFII, is also important for sinuatrial development [143].

In the ventricles, the homeodomain factor Irx4 appears to be involved in negatively regulating atrial chamber identity. Irx4 is expressed in the right and left ventricles and its deletion in mice leads to inappropriate activation of a transgene carrying an avian slow myosin heavy chain gene promoter, which in mice is atrial-specific [110]. In vitro characterisation of this promoter shows that repression in the ventricles requires an interaction between Irx4 and the RXRα component of an RXRα/vitamin D receptor dimer. The complex acts on a vitamin D-responsive promoter element to repress transcription [144]. Enforced mis-expression of Irx4 in chick atria induces ventricular gene expression [145], suggesting that mutual antagonism between a RA-directed atrial program, and an Irx4-mediated ventricular program establishes chamber identity.

16 Summary and conclusions

Four modes of cardiac myogenesis are described in the paragraphs above. Three of them lead to formation of a primitive myocardium that contributes to the primary myogenic scaffold of the heart and myogenic investments to endocardial cushion tissues. Another induces the specialised myocardium of the heart chambers and Purkinje system. There is much to learn about the mechanism underlying these processes. A fifth modality of cardiac myogenesis is stem cell-mediated cardiac repair, and has been demonstrated by findings that the adult heart has a resident stem cell pool, and that mobilised or injected bone marrow stem cells can participate in and/or induce myogenic repair of a cardiac infarct [4,146,147]. Indeed, an innate capacity for cardiac regeneration is present in certain lower vertebrates and strains of mice [2,148]. Remarkably, these findings show that cues for the cardiac developmental program are maintained in the adult organ to some extent. How this program is written and maintained is a new problem for cardiac developmental biologists, and a wonderful opportunity in the stem cell field.

Acknowledgments

We thank C. Biben and F. Stennard for helpful comments and A. Tung for artwork. M.J.S. holds a postdoctoral fellowship from the NIH (F32-HL10389).