Moorman, Antoon F. M., and Vincent M. Christoffels. Cardiac Chamber Formation: Development, Genes, and Evolution. Physiol Rev 83: 1223-1267, 2003; 10.1152/physrev.00006.2003.—Concepts of cardiac development have greatly influenced the description of the formation of the four-chambered vertebrate heart. Traditionally, the embryonic tubular heart is considered to be a composite of serially arranged segments representing adult cardiac compartments. Conversion of such a serial arrangement into the parallel arrangement of the mammalian heart is difficult to understand. Logical integration of the development of the cardiac conduction system into the serial concept has remained puzzling as well. Therefore, the current description needed reconsideration, and we decided to evaluate the essentialities of cardiac design, its evolutionary and embryonic development, and the molecular pathways recruited to make the four-chambered mammalian heart. The three principal notions taken into consideration are as follows. 1) Both the ancestor chordate heart and the embryonic tubular heart of higher vertebrates consist of poorly developed and poorly coupled “pacemaker-like” cardiac muscle cells with the highest pacemaker activity at the venous pole, causing unidirectional peristaltic contraction waves. 2) From this heart tube, ventricular chambers differentiate ventrally and atrial chambers dorsally. The developing chambers display high proliferative activity and consist of structurally well-developed and well-coupled muscle cells with low pacemaker activity, which permits fast conduction of the impulse and efficacious contraction. The forming chambers remain flanked by slowly proliferating pacemaker-like myocardium that is temporally prevented from differentiating into chamber myocardium. 3) The trabecular myocardium proliferates slowly, consists of structurally poorly developed, but well-coupled, cells and contributes to the ventricular conduction system. The atrial and ventricular chambers of the formed heart are activated and interconnected by derivatives of embryonic myocardium. The topographical arrangement of the distinct cardiac muscle cells in the forming heart explains the embryonic electrocardiogram (ECG), does not require the invention of nodes, and allows a logical transition from a peristaltic tubular heart to a synchronously contracting four-chambered heart. This view on the development of cardiac design unfolds fascinating possibilities for future research.
Theory without fact is fantasy, but fact without theory is chaos.
C. O. Whitman (334)
In birds and mammals, the four chambers of the formed heart are arranged in parallel (Fig. 1, A and C). The right atrium is exclusively connected to the right ventricle and the left atrium exclusively to the left ventricle. The right half of the heart drives blood from the body through the lungs to the left half that drives the blood through the rest of the body. The two halves can only beat in synchrony, the right half pushing the pulmonary circulation and the left half the systemic circulation. This parallel-arranged four-chambered heart develops from a single circuited tubular heart. The current concept is that this heart tube is composed of a linear array of segments (21, 56, 254, 330) (Fig. 1, B and D). In this model the atrium is not connected to the right ventricle and the left ventricle not to the outflow tract. The conversion of such a serial arrangement of segments into the proper parallel arrangement has remained one of the most difficult concepts of heart development (156), for it seems illogical first to make a serial arrangement of cardiac segments and then a parallel one. Based on morphological and flow direction considerations, Steding and Seidl (292) described this concept as “one of the most fatal assumptions.” Recent functional and molecular data also challenged the concept (51, 64, 118, 210), and a new concept was formulated, “the ballooning model of chamber formation.” In this review we discuss the possibility that during the evolution and ontogenesis of the heart the chambers develop from a linear tube by local differentiation and that this local differentiation and expansion sets the scene for the electrical configuration of the chamber heart. From the outset, the right and left halves of the heart are brought into parallel position by looping of the heart tube. A fundamental question is thus whether indeed all compartments of the formed heart are represented as a linear array in the tubular heart.
In a stimulating review, Fishman and Olson (90) suggested a model of the vertebrate heart as an assembly of separable genetic modules that were added to the primitive ancestor chordate heart during evolution. The modules are supposed to be component parts or functions of the heart that can be selectively affected by single gene mutations. Examples of such modules would be the cardiac chambers and the localized pacemaker activity. Challenging though these ideas are, this concept can only fully be exploited if the molecular data are integrated morphologically and functionally in a unifying model of chamber formation. The deep evolutionary conservation of molecular pathways in the hearts from fruitfly to human is intriguing and instrumental in finding general developmental mechanisms. On the other hand, evaluation of evolutionary diversity is fundamental to revealing both conserved and novel modules of the building plan of the present-day mammalian heart.
De la Cruz and Markwald (66) wrote an exhaustive review on cardiac development in general. The roles of endocardium (186), epicardium (191), and neural crest (85) in heart development, as well as molecular cardiac genetics (48, 121, 285) were reviewed elsewhere. In this review, we try to define the basic cardiac building blocks and examine their developmental and evolutionary origins. This, in turn, requires evaluation of the identity of the atrial and ventricular chambers and of the conduction system of the avian and mammalian hearts. We need to determine whether these structures are indeed basic cardiac building blocks, or are composed of smaller component parts developed during evolution and integrated during embryonic development. It is our goal to set out a logical model of cardiac chamber formation from these elementary components. In this review we largely restrict ourselves to the myocardial component of the heart.
B. The Cardiac Cell Types: Origin and Diversity
The heart essentially has a mesodermal origin (Fig. 2). The minor contribution of neural crest cells is largely confined to the formation of the aorto-pulmonary septum at the arterial pole of the heart (86, 157), and the endoderm only plays a regulatory role (186). Committment of mesodermal cells to the cardiogenic lineage is established during and shortly after gastrulation (274, 297), upon which the cardiogenic cells are organized in bilateral epithelial sheets that are part of the visceral mesoderm lining the developing coelomic cavity. The cardiogenic cells are in close association with the endoderm. The endocardial cells are formed from the cardiogenic mesoderm in between the endodermal and mesodermallayers. During the subsequent process of folding of the embryonic disc, the bilateral mesodermal cardiac plates migrate into the embryo, fuse in the midline, and form the cardiac tube (64).
The three principal cardiac cell types comprise 1) the endocardium that forms the endothelial lining of the heart (see Ref. 186 for a review), 2) the myocardium or cardiac muscle cells, and 3) the epicardium that is the source of cardiac fibroblasts and coronary arteries (see Ref. 191 for a review). The endocardium is formed early in cardiogenesis because, after retroviral labeling, endocardial cells were shown to share hardly any lineage relationship with myocardial cells (201). However, the two types of cell must have separated after recruitment of the mesodermal cells into the myocardial lineage, because initially the endocardial cells display a cardiac muscle phenotype as well (62, 184, 317). Upon signals from the myocardium, the endocardial cells undergo a transformation from epithelial to mesenchymal cells, by which the mesenchymal cardiac cushions are formed. These cushions fuse and form part of the cardiac septa (30, 31, 83, 205, 244). The mesenchyme of these cardiac cushions, in turn, can be transformed into cardiac muscle, by which process these parts of the cardiac septa become muscularized (211, 225, 308, 309, 339).
The epicardium develops from the so-called proepicardial organ that is located caudally from the heart. The proepicardial organ forms cauliflower-like mesothelial structures protruding into the coelomic cavity that will cover the entire heart and form the epicardium. Mesenchymal cells derived from the epicardium contribute to the cardiac-cushion mesenchyme, to the cardiac fibroblasts, and to the endothelial and smooth muscle cells of the coronary arteries (72, 203, 238, 242).
Interestingly, endocardium, epicardium, and myocardium share a common lineage as they are all derived from a group of mesodermal cells that express the basic helix-loop-helix transcription factor Mesp1 (266, 267). The role of this transcription factor in the formation of the cardiovascular system remains to be resolved. It is of great interest that endothelial cells proved to be capable of transdifferentiating into cardiac muscle when they were cocultured with cardiomyocytes (54). Preliminary results from studies from the author's lab indicate that proepicardium is able to transdifferentiate into cardiac muscle in vitro as well.
C. The Different Manifestations of Cardiomyocytes: Contraction, Conduction, and Automaticity
The vertebrate heart is myogenic, which implies that all cardiomyocytes are in principle capable of producing an intrinsic cycle of electrical activity that leads to contraction (80). This phenomenon is shared with smooth muscle cells and is called automaticity, intrinsic rhythmicity, or pacemaker activity. Because all cardiomyocytes are electrically coupled, the cells with the fastest intrinsic rate, in the mammalian heart usually the cells of the sinus node, stimulate the whole heart to contract and determine its rate. The cells with the fastest intrinsic rate are called the (leading) pacemakers; the others can be categorized as followers or latent pacemakers. If the pacemakers were to stop, other cardiomyocytes take over, albeit at a slower pace. Automaticity requires poor coupling of the cardiomyocytes, which is achieved by a specific composition and density of gap junctions and ion channels, as well as by the size and the arrangement of the cardiomyocytes involved. Poor intercellular coupling permits loading of the cells to a threshold value of electrical charge, resulting in depolarization of the adjacent myocardium (145). Hence, pacemaker (nodal) cells share automaticity and slow conduction (Table 1).
Automaticity and slow conduction velocity are also features of the peristaltically contracting hearts of primitive chordates and vertebrate embryonic tubular hearts, because slow conduction of the depolarizing impulse is an absolute prerequisite for peristalsis (see sect. ii). The development of synchronously contracting chambers required higher conduction velocities of the depolarizing electrical impulse. For this reason, and because unambiguous morphological criteria are lacking, automaticity and conduction velocity became important characteristics to distinguish the different areas in the developing heart.
In the formed heart, atrial and ventricular working myocardium are conventionally distinguished from the so-called “conduction system” (see Table 1). This notion is supported by the fact that the atrial and ventricular muscle cells have well-developed sarcomeric and sarcoplasmic reticular structures (39) but are also well coupled, allowing fast conduction of the depolarizing impulse and efficient synchronous contractions. On the other hand, and in contrast to the cells of the working myocardium, the cells of the conduction system all share a poorly developed contractile and sarcoplasmic reticular apparatus. In this respect, these cells are reminiscent of the embryonic cells of the tubular heart. In contrast to what might be suggested by its name, the conduction system is not just composed of fast-conducting cells, but also of pacemaking or nodal cells. These cells are slowly conducting and poorly coupled and display high automaticity. Only the cells of the atrioventricular bundle and bundle branches are well coupled, permitting fast conduction of the depolarizing impulse. Because the bulk of the cardiac muscle cells comprise cells of the working myocardium, the embryonic-like cells of the developing conduction system are often called “specialized.”
II. EVOLUTIONARY ASPECTS OF CHAMBER FORMATION
The morphologically complex mammalian four-chambered heart has evolved from a simple chordate tubular heart (Fig. 3) (257). Modification of existing elements and the addition of new elements accomplished this evolution. This in turn required continuous adaptation of the genomic regulatory programs that control cardiac development. In general genetic terms, molecular pathways for regional specification of morphological structures had to be selected, which resulted in the activation of sets of downstream genes. Subsequently, cycles of regional specification were required to achieve the final adult configuration (240).
Many features that seemed unique to vertebrates were already present in the early chordate ancestors, and new functions became associated with modified duplicated sets of genes (114). What we thus need to understand both at the morphological and genomic level is what cardiac elements were added during evolution. It is equally important to integrate conceptual insights derived from the cardiac anatomy and physiology of lower vertebrates that underlie mammalian cardiac development. According to the rules of Von Baer (318), the embryos of different vertebrate groups should be compared rather than the adult stages, because the general features of a vertebrate group appear earlier in development than the specialized features. Consequently, the mammalian embryo heart is never like the heart of an adult lower vertebrate but only like that of its embryo. A practical problem is that most experimental data are from extant adult lower vertebrates and not from embryos. Thus one should be cautious in formulating generalities and specializations derived from these adult specimens.
Recent molecular data indicate that some essential aspects of the regulation of heart development in Drosophila resemble those in vertebrate hearts, despite the long evolutionary distance between both groups (25). Drosophila has become the paradigm for many developmental studies in vertebrates, so we start our description with the Drosophila heart.
The heart of Drosophila lies within a dorsal pericardial cavity and is also called the dorsal vessel. It is an elongated tube that is closed posteriorly and drains into the head anteriorly. The anterior part is narrowed and called the aorta. In fact, the Drosophila heart is a blind sac from which hemolymph is expelled into the body by rhythmic contractions. During relaxation, hemolymph enters the heart tube via ostia in its wall, which are closed when the tube contracts. The wave of contractions initiates caudally and moves rapidly in the anterior direction, indicating localized pacemaker activity (253). The Drosophila heart is myogenic, because the heart continues to beat outside the body for several hours. As yet, no specialized pacemaker cells have been reported, and it is unknown whether the localized pacemaker activity is intrinsic or imposed by innervation.
The wall of the Drosophila heart tube consists of two cell layers. The inner layer lining the lumen is composed of a single type of myocardial cells forming a contractile layer (263). The myocardial cells are not fused but interconnected via adherent junctions where the myofilaments also attach. The outer layer lining the pericardial cavity consists of pericardial cells called nephrocytes. They do not have a contractile phenotype but a filtration and secretion function and should not be confused with the pericardial cells in mammals. The pericardial layer is attached to the body wall via so-called alary muscle connections at the segmental boundaries.
C. Primitive Chordates
All chordates have a notochord, a neural tube, and metameric lateral muscle blocks. In the early representatives, a closed cardiovascular system has developed. The chordate phylum comprises the subphyla of the urochordates, or tunicates, and related forms, the cephalochordates with Amphioxus as a well-known representative, and the vertebrates (257).
The tunicate heart is a tubular structure that lies in a pericardial cavity. It consists of a single layer of cardiac muscle cells that are electrically coupled and are not lined by endocardial cells (247). These cardiac muscle cells are also called myoendothelial cells. Although all regions of the heart possess intrinsic automaticity, the dominant pacemaker is active at one end of the cardiac tube for a number of beats, after which a pacemaker at the other end of the tube takes over (3). The resulting peristaltic contraction waves travel along the heart tube with velocities of ∼0.5 cm/s and propel the blood in either of the two directions (164).
The vascular plan of Amphioxus is basically similar to that of the vertebrates (207), but a unique feature is the lack of a true heart. However, several vessels are capable of contractile movements: the subintestinal vein, the portal (afferent) and hepatic (efferent) veins, and the ventral aorta (endostylar artery), which acts as the principal pump. A myoepithelial layer surrounds the vessels (207). The vessels are not lined by endothelium, although scattered cells called hemocytes are sometimes closely associated with the wall of the vessels (245). These cells might be comparable to the ancestral cell population that gave rise to the vertebrate endothelium or vertebrate blood cells or both (294). The vessels contract in succession, although coordination is poor and a single pacemaker area seems to be lacking (247, 319). Very slow, long-lasting contraction waves (∼0.03 cm/s) travel along the vessels, which results in movement of the blood in these valveless vessels, indicating some kind of polarity. However, this polarity of pacemaker activity is not well developed because reversal of flow and consequent extinction of the meeting waves of contraction were observed (319).
D. Lower Vertebrates
The vertebrates constitute the principal group of the chordate phylum, and the development of the chambered heart is a new phenomenon in this group (Fig. 4A). It is first seen in the jawless fishes the cyclostomes, such as the hagfish (127, 247, 268). Their hearts consist of a sinus venosus, atrium, ventricle, and conus arteriosus. The common atrial chamber bulges out over the right and the left sides of the ventricle. The conus arteriosus is poorly developed in hagfish but is clearly present in jawed cartilaginous fishes, such as sharks and rays, and also in lungfishes. The conus arteriosus is the most distal part of the primitive fish heart and forms the connection between the ventricle and the ventral aorta.1 At the sinoatrial, the atrioventricular, and the ventriculoconal junctions, valves developed to prevent backflow of blood during relaxation of the preceding compartment. The cardiac tube is S-shaped with a dorsally positioned atrium and a ventrally positioned ventricle pointing caudally. This S-shape, by which the ventricular inlet and outlet become positioned in almost the same plane, is a general feature of all vertebrate hearts. It has dynamic advantages because it redirects the momentum of the bloodstream, which flows from the atrium into the ventricle, toward the arterial pole. In a linear chamber heart, the inflowing bloodstream would be rebounded from the closed semilunar valves toward the atrium (153).
The sinus venosus became the drainage pool of the venous system to facilitate atrial filling. However, the wall of the sinus venosus is poorly muscularized, and its development as a separate cardiac compartment as well as the development of the sinoatrial valves that physically separate sinus venosus from atrium during contraction have allowed the development of the atrium as the principal driving force for ventricular filling (142, 143). This role has become less important in mammalian hearts where collection of blood is the principle function of the atria. In line with this, in mammals the sinus venosus has become part of the right atrium. Although the lungfish branch does not give rise to the amphibian group, the lungfish heart already has an atrial septum comparable to the amphibian condition. The ventricular septum of the lungfish is considered to be a derived structure that is not homologous to the avian and mammalian ventricular septum (130).
A typical amphibian heart has two atria, a single ventricle, and a bulbus cordis (the amphibian equivalent of the fish conus arteriosus) (127, 143, 342). The sinus venosus drains exclusively to the right atrium. The pulmonary vein bypasses the sinus venosus and directly drains to the left side. The left atrium is often considered to be the evolutionary novel chamber, because it is a component of the pulmonary circulation. Obviously, the atrium of the fish heart has a right and a left atrial myocardial wall as well, but has a common chamber because it has no atrial septum. Hence, rather than the left atrium, one should consider the atrial septum and possibly additional myocardium that surrounds the pulmonary vein as the novel components of the amphibian heart. Despite the presence of a single ventricle, amphibians share a highly effective separation of oxygenated and deoxygenated blood with the related lungfishes (143). Most of the left atrial bloodstream flows to the systemic arteries and most of the right atrial bloodstream to the pulmo-cutaneous arteries. The morphofunctional adaptations for this amazing performance are derived features typical for the extant amphibian specimens. However, these are not dealt with in this review.
E. The Electrical Configuration of the Chamber Heart
The heart tube of tunicates and amphioxus display peristaltic contraction waves, and all regions of the tube display intrinsic automaticity. This is so because peristaltic contraction requires slow propagation of the wave of depolarization, which, in turn, requires poor intercellular coupling. This characteristic also is a prerequisite of automaticity. However, polarity of pacemaker activity is not well developed.
The fish heart displays clear polarity of contraction in a posterior-to-anterior direction. The contraction waves originate in the sinus venosus and terminate in the conus arteriosus. The nodal phenotype persists in the inflow region of the heart, varying from the venosinus to the sinoatrial junctional areas in different species (246, 268). Similar to the mammalian situation, pacemaker tissue with a lower intrinsic rhythmicity is also found at the atrioventricular junction. Interestingly, conus tissue displays pacemaker activity as well (268). In the fish heart, the chamber myocardium has acquired a faster conduction of the depolarization wave (268). For instance, in the ventricular chamber of the shark, the conduction velocities of the depolarization waves are ∼50 cm/s, whereas velocities of ∼3 cm/s were measured in the conus myocardium (300). In ECGs these slow depolarization waves of the conus are recognizable as a peak between the QRS complex and the T peaks of the ventricular depolarization and repolarization, respectively (300). The long contraction that results from this slow conduction in combination with the length of the conus supports valve function and prevents backflow of blood from the ventral aorta into the ventricle during diastole (142, 269).
The electrical configuration of the amphibian heart is remarkably similar to that of the fish heart (2, 193). The fastest conduction velocities are observed in the atria and the slowest in the bulbar region. The atrioventricular junction propagates the depolarizing impulse slowly as well. In line with this, ultrastructural analyses of the atrioventricular region display a phenotype of nodal myocardial cells with the typical absence of a transverse tubular system, few myofibrils, few mitochondria, the presence of glycogen particles, small cell diameter, and scarcity of gap junctions (39). The bulboventricular junction would display the slowest conduction time of the impulse between ventricle and bulbus. This might be apparent because of the presence of a sulcus at the bulboventricular junction; the real traveling distance of the impulse over the sulcus myocardium may be longer than the direct distance between the measuring points in ventricle and bulbus. Also in amphibians the myocardium of the bulbus cordis assists valve function (142).
Whereas the peristaltically contracting hearts of the lower vertebrates do not need valves, the synchronously contracting chamber hearts of the lower vertebrates do need one-way valves at both ends of their chambers to prevent backflow from a downstream compartment during relaxation and to an upstream compartment during contraction. The electrical configuration described above shows that the entry and exit of a chamber are flanked by slowly conducting myocardium guaranteeing a sphincter function (prolonged contraction) at both orifices, as has been demonstrated for the myocardium surrounding the semilunar valves in fish and amphibian hearts (142, 269). Because slow conduction is also a prerequisite for nodal function, it may not be coincidental that in these areas nodes have developed, as discussed in section iv.
F. Tales of the Rings: the Emerging Concept of Ballooning Chambers
In the beginning of the previous century, Benninghoff (21) described regions in the hearts of fish, amphibians, reptiles, birds, and mammals that do not participate in the growth of chambers from the heart tube. Because these regions hardly grow, they define the openings of the expanding chambers as rings (Fig. 4B). He ascribed a sphincterlike function to these areas and associated them with the conduction system. As we discussed in the previous section, these areas indeed have a sphincter function and contribute to the slow components of the cardiac conduction system (the nodes, or pacemaking tissues). Later, the rings were retrieved from oblivion in studies on the development of the conduction system in human embryos, but their sphincter function was denied (330) and the rings were not explicitly restricted to the slowly conducting components of the conduction system, but to so-called “cardiac specialized tissue” (7, 330). The latter term does not distinguish unambiguously between the slow and the fast components of the cardiac conduction system. This use of this term is extremely confusing because this term, as well as the term conduction system, is often linked implicitly to fast conduction. Moreover, what is “specialized”? In a paper on the embryonic development of cardiac impulse conduction, Lieberman (179) expresses this ambiguity as follows:
Perhaps, concomitant with an increase in myofibrillar content, a process is triggered in the genetically determined working myocardial cells which causes the myocardial cell membrane to lose this “embryonic” property to pacemake. Implicit in this notion is the idea that only those cells that develop an abundance of contractile proteins (that is, working myocardium) should, in fact, be referred to as the “specialized tissue of the heart.” Or, the corollary may also be implied: all cells that lose the ability to synthesize myofibrils and maintain automaticity and conductivity should be considered specialized.
The studies of Benninghoff unfortunately led to the abstractive representation of the cardiac tube composed of a linear array of chambers with rings of conduction tissue at the transitions, which has become the paradigm of the development of the mammalian heart and the conduction system (21, 56, 254, 330) (compare Figs. 1B and 4C). However, the atrial chambers balloon out at the dorsal side and the ventricular chamber at the ventral side of the heart tube and lead to the typical S-folded heart of fishes and amphibians as represented in Figure 4 (127, 254).
This morphological process is typical for the developing mammalian heart as well and has been described in great detail for the developing human heart (59). It also explains why the expanding cardiac chambers are flanked by myocardium of the original heart tube and why this myocardium displays features of the slow components of the conduction system.
G. Overview of the Component Parts of the Lower Vertebrate Heart
The hearts of the lower fish and amphibian vertebrates have often been used as prototypes for the development of the avian and mammalian hearts, particularly to understand the origin of the cardiac conduction system (21, 56, 89, 139, 227, 254). Yet, it has remained a highly controversial topic, clearly illustrating the difficulties in uncovering the essentialities of cardiac design. It is thus of paramount importance to try to assess the essence of the design of the fish and amphibian hearts, because these hearts are considered to be the ancestor avian and mammalian hearts. This basic “prototypic” cardiac building plan will enable us to evaluate the development of the parallel-arranged hearts of birds and mammals in section iii. We will not discuss the reptilian heart, as the heart of the extant species is considered to be highly derived. The ventricular septum is considered to have evolved in the primitive reptilian ancestor that gave rise to the mammalian branch on the one hand and to the avian/crocodilian branch on the other hand; its evolution is still controversial (130, 327).
During evolution, the high-volume low-pressure cardiovascular system of the tunicates (blood volume equals ∼40% of their body weight) developed into a low-volume high-pressure system in vertebrates (blood volume comprises ∼6% of their body weight). This change was accompanied by a concurrent transformation of a peristaltically contracting straight tubular heart into the more efficient and powerful synchronously contracting looped chamber heart. This type of change allowed reduction of the circulation time of the blood from ∼6 min in tunicates to ∼1 min in humans. Vessels covered with a contiguous endothelial sheet replaced the largely leaky vessels in amphioxus.
1. Novel cardiac components
Three major adaptations, or “novel cardiac components,” that were not present in the ancestor chordate heart tube can be distinguished in the lower vertebrate heart: the atrium, ventricle, and possibly the muscular sinus venosus. Furthermore, within the ventricular component a compact outer myocardial component and an interiorly localized extensive trabecular component can be distinguished. The specific activation of the ventricle adds to its complexity as follows. The depolarizing impulse travels rapidly from the atrioventricular node toward the apex and then toward the conal region, achieving activation from apex to base. However, preferential conduction tracts have not been identified histologically as yet (268).
The largest changes in the amphibian heart compared with that of fish occurred at the venous pole (Fig. 5). In evolutionary retrospect, the fish atrial wall can be well described as being composed of a right auricular wall and a left auricular wall that bulge out from the common atrial portion of the heart tube. Similar to the mammalian embryonic condition, a single auricle bulges over the ventricular component at the right and left side. First, the sinus venosus became largely incorporated into the right auricular wall close to the midline. Second, an atrial septum and possibly accompanying pulmonary myocardium developed, by which the single atrial chamber was divided into a right and a left component. The systemic venous system drains to the right (systemic) atrium and the pulmonary system drains to the left (pulmonary) atrium. Thus the atrial myocardial compartment of the amphibian heart is composed of the following components: 1) right and left atrial wall or auricular myocardium; 2) sinus venosus myocardium or caval myocardium; 3) septal myocardium, possibly including pulmonary myocardium that surrounds the entrance of the pulmonary veins; and 4) the myocardium of the atrioventricular canal that forms the so-called atrial vestibulum, which is the smooth-walled lower atrial wall just above the atrioventricular junction. Support for the generality of this description comes from studies on the developing mouse heart, in which these component parts have been recognized on the basis of their unique transcription patterns (91).
2. Conserved cardiac components
Three components of the lower vertebrate heart might be considered to be minor adaptations, remnants, or “conserved components” of the ancestor cardiac tube. These components are the sinoatrial junction, the atrioventricular junction, and the conus, because these three areas display nodal activity and phenotype. In the absence of clear distinguishing features, sinus node and atrioventricular node remain difficult to identify unambiguously in the lower vertebrates. It is worth noting that nodal cells are virtually devoid of so-called specific myocardial granules, whereas these granules are abundant in atrial and ventricular myocardium (342, 343). In mammals, myocardial granules are the storage sites of atrial natriuretic factor (ANF) precursor (283) and the nodes are devoid of ANF, similar to the fish nodal tissue (282). Because fish hearts do contain ANF, it is assumed that these granules are the storage sites of ANF precursor protein in fish as well. It is of great interest that ANF is not expressed in the early murine embryonic heart tube (51, 347). For that reason its expression has been used to identify the differentiating chamber myocardium (51), as discussed in section ixA.
So far, we tried to extract the essentialities of the body plan of the vertebrate heart from the examination of adult hearts of lower vertebrates, rather than that of their embryos. In section iii we examine whether these essentialities can indeed be considered to be unifying features of the vertebrate heart.
III. DEVELOPMENT OF THE PARALLEL-ARRANGED FOUR-CHAMBERED HEART
A. Evolutionary Aspects and Terminological Problems
The parallel-arranged avian and mammalian hearts originate from the heart of primitive fish that gave rise to the lungfishes and the amphibian/reptilian branch (Fig. 3) (257). By that time, the development of the left atrium had already been realized in the lungfishes and amphibians and, therefore, the atrial part of their hearts is already essentially similar to its avian/mammalian counterparts. This subject has been dealt with in the previous sections. The right ventricle is the new component of the four-chambered heart and is therefore the focus of the discussion on the development of the four-chambered heart.
The right ventricle is thought to have developed from the proximal part of the conus arteriosus, which is the homolog of the bulbus cordis of amphibians and of avian and mammalian embryos. Trabecularization of the bulbus “ventricularized” this region that subsequently became easily accessible from the right atrioventricular canal by reduction of the so-called bulbo-auricular ridge at the right side (127). The use of the term bulbo-auricular ridge rather than the term ventriculo-auricular ridge may indicate that the early cardiac embryologists saw no ventricular tissue at the inner curvature. In human embryos, the proximal part of the bulbus cordis becomes trabeculated and is then called primitive right ventricle (314). However, trabecules develop at the ventral, outer curvature side of the bulbus only; at the inner curvature smooth-walled myocardium covered with endocardial cushions remains (73, 314). In our opinion, the question can justifiably be asked, whether these authors thought that the entire proximal bulbus would become the primitive right ventricle or merely the trabeculated part at the outer curvature? Others dispute the origin of the right ventricle in the bulbus cordis and argue that it is not a bulbar structure in its entirety (7). In fact, the latter authors may have called the proximal part of the bulbus mentioned by the authors of references (73, 314), right ventricle. Lack of a clear terminology may have led to misunderstanding. Again, the same question can be asked, whether the entire segment of the heart tube was meant to be the primitive right ventricle or merely the trabeculated ventral part. The conflicting terminology originating from both phylogenetic and ontogenetic considerations (241) has undoubtedly contributed to the confusion in the field, but the severe underexposure of dorsoventral patterning has mystified the field most.
Separate cardiac compartments are clearly recognizable at the outer curvature of the developing four-chambered heart, but one feels uncomfortable when one is forced to distinguish atrioventricular canal, left ventricle, right ventricle, and outflow tract at the inner curvature. Indeed, lumen reconstructions of this area in human embryonic hearts made more than half a century ago do not lend support to true segments at the inner curvature (Fig. 6) (57, 73, 218, 228). Based on the development of trabeculated myocardium, unanimity exists that the right ventricle develops later on in development, downstream and visibly separated from the left ventricle as is clear from Figure 6. These observations indicate that the right ventricle develops from a part of the heart tube, whatever the name of that part, that is downstream from the left ventricle, and argue against development from a common ventricular chamber by septum formation within a single ventricular chamber as occurs in the atrial part of the heart. In addition, the reconstructions clearly showed impressions of the ventricular trabecules at the outer curvature and impressions of smooth-walled myocardium at the inner curvature.
The significance of the above-mentioned description is that it unequivocally demonstrates dorsoventral (inner-outer curvature) patterning. One way to describe these factual observations of the cardiac configuration at the ventricular part of the heart tube is that the looping heart tube is a single compartment or module, from which two additional modules have grown, the embryonic ventricles. The smooth-walled myocardium is covered with endocardial cushions and ventrally where the endocardium approaches the myocardium, trabeculated myocardium develops. In the past, the smooth-walled myocardium was also called primary myocardium (73, 228), and recently, this name was reintroduced on the basis of functional and molecular characteristics (63, 209).
A comparable description could be made of the development at the venous side of the atrioventricular canal (73). Here, the primary heart tube covered with endocardial cushion tissue also extends toward the venous side. At the dorsolateral sides, where the endocardium approaches the myocardium, the auricular components develop.
The description given above implicitly considers the primary heart tube as a single entity, although in certain aspects the tube is highly patterned, as evidenced by the local differentiation of the atrial and ventricular chambers and the localized pacemaker activity at the intake of the heart (270, 313). Without implying the presence of strict boundaries (in particular not at the inner curvature) but to facilitate dialogue, we propose that the primary heart tube can be divided into an inflow tract part, an atrial part, an atrioventricular canal, a ventricular part, and an outflow tract (bulbus cordis). The patterning of the tube in the ventricular and outflow tract parts deserves additional attention, because it plays a crucial role in the formation of the right ventricle. After the development of this chamber, an anatomically complete parallel-arranged heart had been achieved.
B. Cardiac Looping, Changing Blood Flows, and Chamber Formation
A most conspicuous feature of cardiac looping is that the ventricular chambers that are positioned along the antero-posterior axis of the embryo become positioned in a left-right orientation (32, 190, 291). Looping starts at a stage in which only the ventricular component and part of the atrioventricular canal has been formed (60, 190). The linear heart tube grows at both poles, bends toward the ventral side, and subsequently turns toward the right side (190). By this process, the original left lateral wall becomes the ventral ventricular wall, the dorsal side the inner curvature, and the ventral side the outer curvature (38, 67, 190). The entire dorsal half including the left and right portions of the linear heart tube ends up at the inner curvature of the looped heart in later stages; a label positioned at the ventral midline ends up at the outer curvature (Fig. 7). This indicates that the original ventral side of the heart tube expands and bends toward the right. During this process, the ventricular chambers are formed and the flow of blood reaches its adult pattern (Fig. 8). All experimental evidence so far supports the notion that looping of the heart is intrinsic to the myocardial tube itself (185).
Because the high viscosity of the embryonic blood results in low Reynolds numbers2 (Re: 0.05-2.0 compared with an adult value of ∼1,000), flow in the embryonic vascular system is laminar (197). This flow can be considered to consist of numerous parallel streaming layers, which can be marked and followed by application of indian ink. Changing flow patterns have been documented by microcinematography during development (Fig. 8) (292, 293). Watching a beating embryonic heart during an early stage of development gives the impression that two separate bloodstreams flow through the heart tube (344), because contraction is not circular but asymmetric. During contraction, the lumen becomes ovoid, and transitory ridges arise at both long-sided walls of the ovoid, as a consequence of which the central part of the lumen is emptied earlier than the lateral parts (278, 293). At the long-sided wall of the lumen the cardiac cushions develop later. During further development of the ventricles, the decrease in the lumen during systole gradually becomes concentric, and thus the pattern of flow becomes adultlike. The blood-flow pattern changes from an initially right flow at the inner curvature and an initially left flow at the outer curvature to the adult configuration, with the left flow directed toward the inner curvature (Fig. 8) (129, 278, 293). These flow patterns have been meticulously documented during chicken development, and the change proved to take place between Hamburger and Hamilton stages (HH stage, Ref. 119) 26 and 28. This is at about Carnegie stage 16, comparable to about 38 days of human development or 12 days of mouse development. At this stage, the peristaltic contraction form disappears from the ventricle (292).
The change in pattern of flow associated with cardiac looping and formation of the ventricles, as represented in Figure 8, makes immediately clear that during development the antero-posterior serial arrangement of the right and left ventricles at the outer curvature had transformed into a parallel arrangement. Thus, via the primary heart tube, the parallel-arranged ventricles are connected with the atria that were arranged in parallel from the very outset. Usually this is presented as a problem because the atrioventricular canal would be entirely positioned above the left ventricle. However, as we described above, functionally this is not entirely true. Anatomically this is not entirely true either: when describing developing cardiac morphology one is confronted with the changing position of the organ within the body and of the cardiac component parts in relation to each other. This easily results in confusion: people with a medical background tend to use the body axes as the points of reference and those with a biological background use the axes of the organ, while often terminology is used indifferently (156).
Scanning electron microscopic photographs of the atrioventricular canal of the embryonic heart disclose a narrow entrance, not only to the right ventricle but also to the left ventricle (Fig. 9); the atrioventricular cushions separating the right and left blood flows occupy most of the volume of the atrioventricular canal. The scanning electron microphotograph of the atrioventricular canal represented in Figure 9 clearly demonstrates that both the right and the left atrioventricular canal halves have to expand almost equally to achieve the adult configuration. Hence, the expansion of the right atrioventricular canal to the right seems to be overemphasized to match the segmental model in which the common atrium is supposed to be directly connected to the left ventricle only. When the axes of the developing heart tube are taken as prime reference points, this notion is not true. Viewed in the longitudinal (antero-posterior) direction of the primary heart tube, the right side of the atrioventricular canal is directly connected to the right ventricle. Similarly, the left side of the atrioventricular canal is directly connected to the left ventricle. Both sides expand during further development. In line with this view, the so-called dorsal (inferior or caudal) atrioventricular cushion is not limited to the classical atrioventricular canal but extends toward the ventricular septum (Fig. 9) and guides the right blood flow directly toward the right ventricle and the left blood flow to the left ventricle.
C. The Primary Ring: Development of the Right Ventricle
A crucial landmark in the development of the right ventricle of the four-chambered heart is the so-called primary ring, also called interventricular or bulboventricular ring (Figs. 8, 9, 10) (43, 212, 333). This ring was recognized in the beginning of the previous century and associated with the developing conduction system (see sect. iiF) (7, 21, 330). Lack of molecular markers had prevented unambiguous delineation of its developmental fate and hence full appreciation of its morphological significance. The primary ring comprises the myocardium of the primary heart tube at the position where the ventricular septum develops. In human embryos, the ring is characterized by the expression of a protein epitope called GlN2 (18). In chickens, it exclusively expresses the Drosophila muscle segment related homeobox transcription factor Msx-2 (43). The complex morphological expression patterns of these factors during further development were described in great detail, and three-dimensional reconstructions were made (43, 154, 173). Both in humans and in chickens, remarkably similar patterns of expression appeared. Figure 8 depicts the position of the primary ring relative to the embryonic blood flow and clearly shows that the ring demarcates the entrance to the forming right ventricle and the outlet of the forming left ventricle. Concordantly, during further development the primary ring becomes positioned at the lower rim of the right atrium just above the right atrioventricular junction and at the left ventricular outlet just below the left ventriculoarterial junction (173, 331). Although this morphological analysis is not strictly speaking a lineage analysis, the precise description of the changing patterns of expression of Msx-2 and GlN2 in a large developmental series in two species (chickens and humans) allows assessment of the contiguity of the changing pattern from stage to stage and makes this morphological analysis almost similar to a lineage analysis.
Considering Msx-2 and GlN2 as lineage markers, their developmental patterns of expression unambiguously demonstrate that right atrial myocardium develops at the inner curvature upstream from the primary ring, and left ventricular outlet myocardium downstream from the ring. The simplest explanation of the previous statement is that these areas develop directly from the primary heart tube rather than from parts of the tube that had already developed a (left and right) ventricular identity, as is assumed in the segmental model. This notion is supported by molecular data (51, 210) (see sect. ix). The ring divides the primary heart tube into two parts, an upstream or atrioventricular part and a downstream or bulboventricular part. For the right blood flow the ring is atrioventricular, demarcating the right ventricular inlet, and for the left flow of blood the ring is bulboventricular, demarcating the left ventricular outlet (Fig. 8). Similarly, one could divide the cushions of the primary heart tube into two types of cushions as well: 1) upstream cushions rather than atrioventricular cushions, and 2) downstream cushions rather than truncus cushions (Fig. 9). We will conform to conventional terminology, but it is important to appreciate that large parts of the cushions are in fact localized in parts of the primary heart tube that are conventionally called right or left ventricle. The atrioventricular cushions also extend into the atrial part of the heart tube upstream from the atrioventricular canal, although this is less pronounced (Fig. 9).
D. Completion of Septation
Chamber formation in the avian and mammalian heart is completed with the physical separation of the blood flows (Fig. 9). For the sake of clarity, we only consider septation of the three principal components of the heart: 1) septation of the atrial chamber by formation of the atrial muscular septum, 2) septation of the ventricular chamber by formation of the ventricular muscular septum, and 3) septation of the primary heart tube by formation of the cardiac cushions. For precise descriptions of the changes in these complex morphological processes we refer to detailed accounts from our laboratory and others (59, 64, 73, 154, 171-173, 228, 292, 314). Only some very basic principles are given here.
The atrial chamber is divided by a muscular septum (primary septum) that is formed by local active proliferation of the myocardial cell layer. Its free rim is covered with a mesenchymal cap (100) that fuses with the atrioventricular cushions and the dorsal mesenchymal protrusion (vestibular spine), as shown in Figure 9 (154, 299, 328, 332). Whether the mesenchymal contribution from three distinct origins (atrioventricular cushion, atrial septal cap, and extracardiac mesenchyme) to the formation of the lower rim of the primary atrial septum is of developmental significance is as yet unknown. Folding of the atrial wall at the right side of this primary septum forms the secondary atrial septum (8, 64, 314). Note that the secondary atrial septum is a unique mammalian structure and is not present in avian and amphibian hearts.
The muscular ventricular septum is formed by apposition of cardiomyocytes at the outer side concomitantly with the outgrowth of the ventricles at the outer curvature (314). In vivo labeling studies have shown that the septum is derived from ventricular cells located at the ventral fusion line of the bilateral mesodermal heart epithelia (65).
The primary heart tube is physically separated into left and right components by fusion of the so-called cardiac cushions. In the primary heart tube, the endocardium is separated from the myocardium by an extracellular matrix layer or cardiac jelly (58, 59). The cardiac jelly disappears from the chamber-forming regions of the cardiac tube, but remains covering the smooth-walled myocardial tube. Subsequently, upon myocardial signaling, a subpopulation of endocardial cells overlying the cushions undergoes a transformation from epithelial to mesenchymal cells, by which the cardiac cushions are formed (31, 83, 264, 336). Neural crest cells also contribute to the mesenchyme of the outflow tract ridges (55, 243, 320), while epicardium-derived cells contribute to the atrioventricular cushions (103). However, the specific functions of these cells are still unclear.
Although the molecular mechanisms of the formation of cushion mesenchyme are well-described (30, 31, 83,205, 244) and the position of cushion formation is clear from the contraction pattern mentioned above, the molecular cues that determine the precise anatomical positions and extents of cushion formation are as yet unknown. Upstream from the primary ring, the two atrioventricular cushions (upstream cushions) develop (Fig. 9); note that they extend both in the ventricular part and in the atrial part of the tube. Only the dorsal (inferior or caudal) atrioventricular cushion extends toward the ventricular septum, guiding the two blood flows separately into left and right ventricular directions. Two cushions are also present downstream from the primary ring, usually called ridges. We call the ridge that is contiguous with the ventricular septum the septal outflow tract ridge (64) (other authors have called this ridge the left conal swelling, Ref. 314), and the other ridge the parietal outflow tract ridge (also called the right conal swelling). The septal ridge guides the blood from the left ventricle toward the aortic outlet. The parietal ridge extends into the atrioventricular part of the heart tube, and because of that, it also guides the right flow of blood into the right ventricle.
Fusion of the cushions affects the physical separation of the right and left blood flows, but the precise contributions of the individual cushions in the ventricular part of the heart tube are uncertain. It should be appreciated that the cartoon in Figure 9 overemphasizes the relative dimensions of the heart tube at the inner curvature. The outflow tract of the heart was hinged to the right to permit visualization of the cushions at the inner curvature.
Most of the cardiac cushion mesenchyme becomes muscularized during further development (68, 211, 225, 308, 309, 339), but the extent of muscularization varies among the different species. For instance, in humans, the atrioventricular cushion mesenchyme approaching the top of the ventricular septum remains membranous and forms the so-called atrioventricular part of the membranous septum (172).
IV. DEVELOPMENT OF CARDIAC FUNCTION AND CONDUCTION SYSTEM
Some topics in the field of cardiac development have always attracted special interest, for example, the neural crest and the cardiac conduction system. The combination is irresistible but has not particularly clarified the field (for reviews, see Refs. 107, 112, 170, 202, 208, 271). Controversy often appears to be based on misunderstanding, due to a highly confusing terminology. The pacemaking tissues (sinus node and atrioventricular node) are well-recognized components of the cardiac conduction system. As we have seen, pacemaking activity is already present in the hearts of the chordate ancestor in which neural crest cells have not yet developed (114). An ECG (232, 277, 313), indicating coordinated depolarization of the distinct cardiac compartments, can be recorded concomitantly with chamber formation at a stage in which neural crest cells (106, 107, 158, 159) as well as cells derived from the epicardium (103, 191) have not arrived in the heart. Although a role for these cells in the maturation of the conduction system can be envisaged, the essential electrical configuration of the heart has already been laid down before these cells arrive at the heart and before the different components of the conduction system can be morphologically identified. Recently, this notion was confirmed independently by both elegant lineage studies of Cheng and co-workers using retroviral and adenoviral cell lineage tracing constructs (49), and clonal analysis of the fate of the atrioventricular canal (60). A sensible approach to examining how the essential electric configuration is achieved is to relate its development with the development of the basic components of the heart.
A. Development of Polarity
The first key issue to resolve with regard to the development of polarity in the heart is the origin of the sinus node or leading pacemaker activity. Its development is intimately associated with the development of anteroposterior polarity of the heart. Patten (234) already noticed this and wrote,
One of the most startling and significant facts about the developing embryonic heart is that its myocardium at different cephalo-caudal levels exhibits different inherent rates of contraction. Each new part of the heart that is added behind the first formed cono-ventricular part has a higher intrinsic contraction rate than the part of the heart tube already formed. The part of the cardiac tube with the highest contraction rate at any given phase of development sets the rate for the entire heart.
Patten referred to experiments in which fragments of chicken myocardium are cocultured or grafted. As soon as the fragments have become connected, the faster beating fragment takes the lead and causes the more slowly beating fragment to beat at a faster rate. He concludes that the pacemaker is always the part of the heart that was added most recently. Because electromechanical coupling is first established in the ventricular region, the first beats are observed in the ventricle, while the pacemaker is present in the upstream region that does not yet beat (313). These observations and experiments demonstrate antero-posterior polarity in pacemaker activity and contraction rate along the linear heart tube. Due to the slow conduction of the impulse (13, 63, 126), contraction is peristaltic (234). Taken together, in the early chicken embryo all regions of the linear heart tube possess intrinsic automaticity and the peristaltic contraction waves travel along the tube with low conduction velocities. This condition is reminiscent of the hearts of the primitive chordates, in which polarity evolved with the dominant pacemaker at the intake of the heart, but the mechanism that imposes antero-posterior polarity upon the heart tube is not settled yet (see sect. vi).
Functional measurements of early mammalian cardiac development are rare, because the mammalian embryo is not easily accessible for experimental manipulation. Furthermore, early mammalian cardiac development is rapid compared with early chicken cardiac development. For example, in rats the first contractions are already observed before the paired primordia have fused (108); the myocardial cells are poorly coupled, which would suggest that also in the early mammalian heart tube the propagation of the impulse is slow (94, 115, 310).
B. Development of the ECG
Already before the onset of overt chamber formation, an ECG can be recorded in embryonic chicken hearts from HH stage 11 onwards (232, 277, 289). Initially, the ECG is sinusoidal in shape, as reflected by the peristaltic waves of contraction. A delay between the atrial and ventricular parts of the tube can be recorded, when the atrioventricular sulcus becomes visible. This development is accompanied by an “adultlike” electrical organization, comprising the typical P wave of atrial depolarization, the QRS complex representing ventricular depolarization, and the T wave representing ventricular repolarization. Interestingly, a conal wave (C wave) in between the QRS complex and the T wave is present in embryonic chicken hearts as well (231). Thus the chicken embryonic ECG is basically similar to that of the formed heart of the shark, as discussed in section iiE. Although in the early embryonic heart (chicken stage HH 14) no specialized/distinct cells of a conduction system can be recognized, it displays coordinated propagation of the impulse, implying a pacemaker at the intake of the heart, rapid conduction in the atrial region, a delay in the atrioventricular region, rapid conduction in the ventricular region, and slow conduction in the outflow tract. And, indeed, slow conduction velocities were measured in the sinoatrial region (63), atrioventricular region (13, 14, 63, 180, 181, 313), and outflow tract (28, 63, 231), and it was concluded that the atrioventricular ring tissue is the functional counterpart of the atrioventricular node (181). In line with this, the primary heart tube consisting of atrioventricular canal, inner curvature, and outflow tract is virtually devoid of the two main gap junctional proteins connexin (Cx) 40 and Cx43 (69, 208, 310).
Studies in chicken demonstrated that the linear heart tube contracts peristaltically and has a matching sinusoidal ECG (277). Also, during further development when chambers have formed, the pattern of contraction matches the electrical configuration (28, 230). In the linear heart tube the layer of cardiac jelly, in between the endocardium and the myocardium, plays a crucial role by supporting the unidirectional propulsion of the blood: the limited sarcomeric shortening would not permit closure of the myocardial tube without such a layer (19, 235). During further development, a progressive lag in contraction duration along the cardiac tube becomes apparent as follows. The atrium displays short, brisk contractions that are finished before the ventricle contracts, the ventricle contracts at a slower pace, and the sluggish, prolonged contractions of the outflow tract prevent backflow of blood while the ventricle relaxes (28). This situation is highly reminiscent of that in the adult shark heart, in which the prolonged contraction of the conus (outflow tract) supports the function of the semilunar valves (142, 269). Such a function is also essential for the embryonic heart, which has no valves; the slow conduction of the depolarizing impulse in the atrioventricular canal and outflow tract, and the subsequent prolonged contraction duration substitute valve function in these areas (63). The pattern of contraction is in line with the regional differences in calcium-triggered calcium release and with the mRNA and protein levels of sarco(endo)plasmic calcium ATPase (210). High levels were observed in the atrium, intermediate levels in the ventricles, and low levels in the primary heart tube, that is in the atrioventricular canal, the inner curvature, and in the outflow tract (Fig. 9C).
In summary, the molecular, functional, and electrical configurations of the looping heart tube at the onset of chamber formation fully explain the coordinated contraction as reflected in the ECG.
V. LINEAGE ANALYSIS AND CHAMBER DEVELOPMENT
Cell-tagging (fate-mapping), grafting, and explantation experiments in several species have yielded a detailed insight into the contribution of cell populations to the heart at consecutive stages of development (199, 258, 291, 297). Precursor cells of myocardium and endocardium are colocalized in the epiblast. During gastrulation, the epiblast heart precursors ingress through the primitive streak until late-streak stages [chicken HH3; mouse embryonic day (E) 6.5-7] to form the bilateral heart-forming regions lateral of the anterior gut endoderm (chicken HH4-8; mouse E7-7.5).
A. Formation of the Linear Heart Tube
The tubular heart is formed by migration and fusion of the two heart-forming regions at the midline [chicken HH9-10 (291); mouse E7.5-8(71)]. Labeling experiments in chicken showed that the cells of the cardiac crescent fuse in the antero-posterior direction (258, 291). The ventral side of this tubular heart will eventually contribute to the apical part of the right ventricle of the formed heart (67). As development proceeds, myocardium is continuously recruited at the posterior side of the elongating tube. This myocardium will form the apical part of the left ventricle, the atrioventricular canal, and the sinoatrial region, respectively (67, 291). At the anterior side of the tube, myocardium is recruited from the pharyngeal mesoderm that will contribute to the outflow tract (67, 150, 204, 316, 321). From anterior to posterior serially aligned primordia are thus formed. As a result of the continuous elongation of the heart tube, the morphological inflow tract and outflow tract of the linear heart tube of HH9-10 chicken embryos are formed by myocardium fated to contribute to the left ventricle and right ventricle in the mature heart, respectively. While recruitment of myocardium at both heart poles takes place, the prospective ventricular portion has already begun to loop and to change its molecular and morphological phenotype (see below).
B. The Heart-Forming Regions Include Anterior and Posterior Mesenchyme
In a recent lineage study by Redkar et al. (248), a precise map of the mesodermal heart-forming region in chicken embryos was generated by DiI cell labeling. The stages of labeling were HH4-8, which represent the period until just before fusion of the heart-forming region at the midline to form a tubular heart (from HH9 onwards) and were followed by the analysis of the fate 20 h later (until stage HH12). The heart-forming region thus mapped appeared larger than the heart-forming region that was defined in a previous study by Ehrman and Yutzey (82), in which the fate was analyzed at HH10. This discrepancy may result from the difference in stage of analysis between both studies, because in the latter study cells that will contribute to the heart tube after HH10 were not mapped to the heart-forming region. Furthermore, in the analysis of Ehrman and Yutzey (82) Nkx2-5 and Vmhc were used as markers for the heart-forming region. At HH5, however, this gene was not expressed in all cells fated to become cardiomyocyte at HH12 (248). The posterior border of the Nkx2-5 expression domain within the heart-forming region was found to represent the posterior ventricular border of the heart tube at HH12. Because the entire heart tube of HH12 embryos expresses Nkx2-5, this finding indicates that between HH5 and HH12 at the posterior side of the heart ongoing recruitment of cells to the cardiogenic lineage (Nkx2-5 positive) has taken place.
In reality the heart-forming region is even larger, because at stage HH12 part of the sinoatrial region and the outflow tract have not yet been added to the tubular heart (67) and were thus not mapped to the heart-forming region. The entire heart-forming region would be found if the end-stage of analysis is performed at approximately stage HH18-22, when all myocardium has been recruited (67). Indeed, in two independent studies (204, 321) the outflow tract of the chicken heart at HH18-22 was shown to be formed from cells that are localized outside the “classical” heart-forming region, in a “novel” anterior or secondary heart field that got unrecognized before. It remains to be defined which cells in the HH4-8 embryos will give rise to the outflow tract. Kelly et al. (150) extended the observations in chicken to mouse and demonstrated that between E8.25 and E10.5 pharyngeal and anterior pericardial mesodermal cells are recruited to the heart, forming the myocardial outflow tract. These myocardial cells originate medially from the classical heart fields. Whether or not the right ventricle is part of the anterior heart field is not settled yet. The pattern of expression of the Fgf10 enhancer trap LacZ transgenic line suggests it is. However, a small Mlc2v promoter fragment that is active in the right ventricle and the outflow tract is expressed already in the anterior region of the cardiac crescent and linear heart tube (260), which may indicate that right ventricle precursors may not be part of the anterior heart field.
C. The Inner Curvature Connection
An important issue to resolve is the lineage relationship between the distinct regions of the linear heart tube and the component parts of the cardiac chambers. De la Cruz and Markwald have made relevant contributions as documented in detail in Reference 66. Some other authors interpreted these analyses as follows (192): “The segments of the tubular heart, like the somites, form progressively along an antero to posterior axis with each segment being an undifferentiated, morphogenetic unit separated by boundary interfaces.”
At present, this segmental view is commonly used, albeit in general without any conceptual connotation. Because De la Cruz and Markwald did not explicitly include dorsoventral regionalization in their descriptions, it seems to be a logical extension of their work, but nevertheless it is fundamentally wrong. De la Cruz and Markwald (66) explicitly state that “In the straight heart tube there is no region that gives origin either to the right or the left ventricle in their entirety,” and further that “the straight heart tube is constituted by two primordia, one cephalic, i.e., that of the apical trabeculated region of the anatomical right ventricle, and the other caudal, that of the apical trabeculated region of the anatomical left ventricle. Consequently there is no single primordium for each of the definitive cardiac cavities.”
This description is entirely in line with that of the early cardiac anatomists (57, 73, 218, 228) given in sectioniii. In the mature four-chambered heart, the atrioventricular junction and the right ventricle are directly connected without intervention by left ventricular myocardium (compare Figs. 1 and 11). Likewise, the left ventricle is directly connected to the outflow tract, without intervention of right ventricular myocardium. Even the atrioventricular junction and outflow tract are directly connected without intervention of ventricular cells at the inner curvature. It is clear that these connections cannot be made in a representation of the tubular heart composed of serially aligned segments, where atrioventricular canal and outflow tract are separated by cells fated to become left ventricle and right ventricle, respectively (Fig. 11).
The discrepancy between the concept of the segmented tubular heart and the adult condition is virtual and can entirely be solved by appreciating that from the very outset of heart-tube formation onward a direct connection exists between cells fated to become atrioventricular canal and those fated to become outflow tract. This connection is found at the dorsal side of the straight heart tube and at the inner curvature of the looped heart tube. In line with this assumption is the fact that only the apical components of the ventricles are formed from the “ventricular segments,” whereas the connected basis and conus parts are generated from the atrioventricular canal and outflow tract “segments” (Fig. 11) (67). The lineage analysis studies of De la Cruz and co-workers (67) do not resolve issues related to the fate of myocardium located at the cardiac inner curvature, simply because these areas were not labeled. Hence, the entire concept would be strengthened if some of the labeling experiments were repeated in conjunction with analyses of patterns of gene expression, such as those of the genes encoding Msx2 and ANF as markers for the primary ring and for the forming chambers, respectively (Fig. 10).
D. Origin of the Nodal Tissue
It has been a long-standing issue whether the nodal components of the cardiac conduction system are remnants of the primary myocardium and escaped differentiation into the chamber lineage (“escape” hypothesis) (208, 209, 241), or are recruited during later stages of development from a common precursor that also gives rise to the working myocardium (“recruitment” hypothesis, Ref. 49). Although placed opposed to each other (49), the two hypotheses are essentially similar. However, the escape hypothesis explicitly assumes that the primary myocardium is the common precursor of nodal and working myocardium. It also assumes that nodal function is already present in the embryonic heart, as we have discussed in section iv. The recruitment hypothesis, on the other hand, is indeterminate about the nature of the precursor cells, assumes that recruitment takes place until late stages of development, and implicitly suggests that the nodal conduction system is formed de novo in the developing heart.
Several arguments are in favor of the escape hypothesis, as follows.
The observations described in section iv reveal an amazing simplicity of cardiac design, in which the slowly conducting areas of the primary heart tube relatively decline at the expense of the developing fast conducting working myocardium of the chambers. This description explains the observed electrical configuration (ECG), without the need to recruit new “nodal” elements. In line with the escape hypothesis, the slowly dividing cells of the primary heart tube map to the conduction system (60, 304, 305). Further evidence comes from Gata-6 gene enhancer studies and analysis of the expression pattern of minK (60, 169). Burch and co-workers (60) observed that transgenic mice having the lacZ gene under control of a chicken Gata-6 gene enhancer expressed the transgene in the atrioventricular canal. At later stages of development expression became confined to the atrioventricular node, atrioventricular bundle, and atrioventricular ring bundle. By deploying the Cre-lox system, it was shown that these nodal components are clonally related to the embryonic atrioventricular canal cells (60). Similarly, when inserted into the minK gene, the lacZ gene was expressed in primary myocardium of the atrioventricular canal in the embryo, and in the region of the atrioventricular node in the mature heart (169). These observations indicate that already early in development part of the primary heart tube (atrioventricular region up to and including the primary ring) is predisposed not to participate in chamber formation, but fated to atrioventricular node and bundle, and is not recruited late in development (49).
Birth-dating experiments by label dilution of [3H]thymidine and lineage tracing experiments using replication-incompetent viral lacZ-expressing constructs have been interpreted by Cheng et al. (49) in terms of the recruitment hypothesis. However, the results of these birth-dating experiments also match the escape hypothesis and confirm the observations of Thompson and co-workers (304, 305) that the conduction system is recruited from slowly proliferating myocardium. Lineage tracing studies that started in primary heart tube stages of necessity reveal a lineage relationship of cells of the conduction system with neighboring cells of the working myocardium, as both are formed from this source. Finally, it is still unknown why some cells of the primary myocardium in the sinoatrial and atrioventricular region escape differentiation into chamber myocardium and form the nodes, whereas the primary myocardium of the outflow tract eventually differentiates in the ventricular direction. It also is unclear why morphologically distinct sinus and atrioventricular nodes are present in the formed heart of mammals only, whereas in other vertebrates the nodes remain morphologically quite diffuse (39).
E. Fate of the Primary Ring
The fate of the primary ring is of great interest. In section iiiC we described that this ring can be identified molecularly by the expression of Msx-2 and GlN2. It is a part of the primary heart tube that is localized at the crest of the ventricular septum and that after further development surrounds the right atrioventricular junction and the left ventricular outlet (43, 212, 333). In adult avian hearts, these regions were morphologically identified as parts of the conduction system that still form a ring and comprise the atrioventricular bundle, septal branch, retroaortic branch, and right atrioventricular ring bundle (170, 315). Thompson and co-workers (304, 305) identified these regions as proliferation-quiescent areas derived from the slowly dividing tubular heart, which was recently confirmed by the above-mentioned lineage studies of Cheng et al. (49). In the adult mammalian heart, none of these components can be recognized except the atrioventricular bundle, although in embryonic and neonatal hearts of various species (including humans) some elements are still present (4, 5, 9, 11, 208, 333).
Based on its location and origin the ring is expected to have a “primary heart tube” phenotype. Morphologically this is indeed the case; similar to the nodal tissues these cells are always described as “more undifferentiated” relative to the working myocardium of the atrial and ventricular chambers. We mentioned (see sect. iiF) the paradoxical naming of the conduction system as cardiac specialized tissue, which is at odds with the poorly developed myofibrillar and t-tubular system (39). However, it has been argued that the conduction system is specialized because it has almost withdrawn from the cell cycle (304). These cells were considered to be terminally differentiated, analogous to skeletal muscle precursor cells that withdraw from the cell cycle and differentiate into muscle cells. It is therefore interesting to speculate whether the derivatives of the ring have retained the slowly conducting feature as well. Arguëllo et al. (14) mapped this area in great detail in embryonic chicken hearts after 6 days of incubation. In the atrioventricular bundle, as well as in the atrioventricular node, atrioventricular canal, and lower interatrial septum, they measured low conduction velocities compared with the velocities in the atrial auricle and ventricle. In line with this, the gap junctional proteins Cx40 and Cx43 are generally found to be absent from the atrioventricular bundle and proximal part of the bundle branches in fetal and neonatal mammalian hearts (311). In adult hearts of both mouse and human, the atrioventricular bundle is fast conducting (128, 161). Thus, although the cells of the conduction system withdraw from the cell cycle relatively early in development, they acquire their mature phenotype relatively late in ontogenesis, indicating that the maturation of the conduction system is an ongoing process.
An intriguing aspect of the primary ring is that it constitutes an almost separate compartment. At the right side of the atrioventricular canal, the insulating fibrous tissue will interpose the myocardium downstream from the ring. By this process the ring forms the lower rim of the right atrium that dorsally penetrates the insulating fibrous plane as the atrioventricular bundle (331). The atrioventricular bundle will become surrounded rather than interrupted by fibrous tissues. In mammalian development, the ventral penetration disappears concomitantly with the retraction of the myocardium of the outflow tract by which process the part of the ring that surrounds the left ventricular outlet disappears as well (326, 339). The atrioventricular bundle thus connects the cardiac specialized primitive tissue of the atrioventricular node with the cardiac specialized primitive tissue of the ventricular conduction system (see below), and in itself is also a cardiac specialized primitive tissue.
We know very little about the development of the function of the primary ring, its molecular identity, its interaction with the neighboring compartments, and about the maturation of the atrioventricular bundle to a fast-conducting bundle. The role of Msx-2 is still unexplained. We also do not know whether the ring is present in amphibian or fish hearts as a phenotypic distinct entity. Unraveling the regulatory pathways will add a fascinating story to the understanding of cardiac morphogenesis.
F. Origin of the Ventricular Conduction System
The ventricular conduction system comprises the right and left bundle branches and the subendocardial peripheral ventricular conduction system. In rodents, the peripheral ventricular conduction system is poorly developed, but in chickens and ungulates it is well developed. In chickens, an additional well-developed periarterial Purkinje system has developed (39, 170, 208, 315). Studies by Mikawa's laboratory revealed that in chickens 1) the periarterial Purkinje cells are clonally related to surrounding working myocardium, 2) these cells develop in close association with and are induced by the developing coronary arteries, and 3) their formation can be induced by endothelin-1 in vitro and in vivo (49, 113, 136, 296). The periarterial Purkinje system develops late in chicken development (second half of the incubation time), which is well after the development of the ECG and of the preferential conduction of the impulse toward the apex of the ventricle (53, 63, 232, 277, 289). The periarterial Purkinje system expresses a slow tonic myosin (105) and a distinct gap junctional protein Cx42 (110), and it displays a unique transcription factor expression profile (295, 301). The periarterial Purkinje system has not yet been characterized functionally.
The bundle branches and the peripheral ventricular conduction system are thought to develop from the trabecular ventricular component (208, 315). The early ventricular wall consists of a compact layer of myocardial cells and protrusions into the lumen called trabecules. Before and during looping, the wall of the heart consists of only a few layers of epithelial-like myocytes and is lined by a cellular cardiac jelly and endocardium. Just after looping (chicken HH16/17 and mouse E10), trabecules become evident at the entire outer curvature between the atrioventricular and outflow tract cushions (276). The cells of the ventricular trabecules and of the bundle branches display a poorly developed contractile apparatus and t-tubular system but well-developed gap junctions (39). The trabecules display a high abundance of Cx40 and Cx43 protein (94, 111, 310) and the impulse is preferentially conducted via the trabecules (53, 63). Retroviral lineage analyses revealed that individual myocytes generate a series of myocyte clusters, forming one or a few trabecules (199, 202). These trabecules fuse and, because the cells at the epicardial side proliferate faster than those at the endocardial side (304, 305), cone-shaped clonally derived structures are formed. Part of the trabecules becomes incorporated in the compact layer of myocardium. Consistently, label particles placed on the ventral wall of the unlooped heart were found in the trabecular regions of the ventricles (67). The sequence of events implies that the cells in the trabecules at the endocardial side are the “oldest” cells directly derived from the embryonic ventricular wall of the heart tube. Indeed, Anf (51, 133, 347), Cx40 (70, 111, 312) and Irx3 (52) initially are expressed in the epithelial-like ventricular wall of the embryonic heart. During chamber expansion their expression becomes restricted to the trabecules and subsequently to the subendocardial ventricular conduction network. A similar process, but less extensive, is observed in the atrial appendages, where the trabecules become the pectinated muscles.
G. Overview of the Component Parts of the Higher Vertebrate Heart
The formed heart comprises the right and left atrial and ventricular chambers and a conduction system encompassing the pacemaking tissues (sinus node and atrioventricular node) and the fast-conducting atrioventricular bundle and bundle branches. The definitive cardiac chambers are anatomical units. They are built up out of several embryonic component parts that often constitute distinct transcriptional domains (88, 91, 93, 151, 152).
1. The atrial and ventricular chambers
Compared with the fish/amphibian ancestor heart, the essence of the design of the atrial part of the avian and mammalian heart has not significantly changed (see sect.iiG, Fig. 5), whereas in the ventricular part a right ventricle has developed and the outflow tract has been incorporated into the ventricular component. This part of the heart is still a distinct transcriptional domain primarily within the right ventricle (151). Because the fibrous insulation occurs at the lower boundary of the atrioventricular canal (331), the atrium can be considered to be composed of four component parts (see Fig. 5), i.e., 1) sinus venosus, 2) primary atrial septum including the pulmonary myocardium, 3) auricles (atrial appendages), and 4) atrioventricular canal (atrial vestibule). Each of these component parts has a left and a right domain (91).
Whereas the heart has a left and a right atrium in terms of identity, this is not so for the ventricles. We described above that the ventricles develop at the outer curvature along the antero-posterior axis. After looping, the ventricular chambers become localized in left-right orientation; the process of looping positions the left side of the heart tube ventrally. Thus, in terms of identity, each ventricle has a left component (located at the ventral side) and a right component (located at the dorsal side) (38, 67, 190). Consequently, it might be clearer to call the left ventricle the systemic ventricle and the right ventricle the pulmonary ventricle.
Finally, two components, which we have called “evolutionarily conserved” elements, are now incorporated into the chambers, namely, the atrioventricular canal into the atrial chambers (171, 172, 331) and the proximal part of the outflow tract into the ventricular chambers; the distal part loses its myocardium to the semilunar valves (306, 326, 339). The myocardium of the atrioventricular canal has retained some of its primary features (196), but the outflow tract has retained hardly any primary feature (208). It is of great interest that in patients ventricular tachycardias originating from the right or left ventricular outflow tract were described that sometimes originated from the stem of the pulmonary artery (146, 255). These observations suggest that in these patients the myocardium of the outflow tract has retained some of its primitive characteristics and may even partially persist in the arterial trunk.
2. The conduction system
In the formed heart, three distinct components of the conduction system can be distinguished morphologically, functionally, and molecularly (Table 1): 1) the sinus node and atrioventricular node, consisting of poorly coupled cells that do not express ANF; 2) the atrioventricular bundle, consisting of well-coupled cells that do not express ANF; and 3) the bundle branches and the subendocardial extensions, consisting of well-coupled cells that do express ANF. Other differences exist as well (208), but those mentioned here are highly distinctive and have been discussed in the previous chapters.
3. Engrailed-2/lacz transgene expression: speculations on cardiac design
The distinct components of the cardiac conduction system have much in common. The sarcomeric and ttubular systems generally are poorly developed (Table 1) (39, 208), and the entire conduction system originates from proliferation-quiescent areas (304, 305). In this context the pattern of a recently described cardiac transgene is of great interest (249). The pattern originates from a singular transgenic mouse line (engrailed-2/lacZ) harboring a fusion construct of the lacZ reporter gene and the enhancer region of the homeodomain-containing transcription factor engrailed-2 (Fig. 12). LacZ gene expression was first observed in a widespread pattern in the (primary myocardial) tubular heart (E8.5-9). During further development, the entire area of the conduction system, including the nodal regions, atrioventricular bundle, bundle branches, and subendocardial ventricular myocardium expressed the transgene. Interestingly, left and right atrioventricular junctional myocardium (also called the smooth-walled atrial vestibular myocardium that is derived from the atrioventricular canal region of the primary heart tube, Refs. 154, 331) expressed the transgene as well. In addition, expression was observed in a broad dorsal region of the right atrium roughly in between the nodes.
An intriguing possibility is that the transgene discloses the myocardium that has not fully differentiated into atrial or ventricular working myocardium and identifies the area of slowly proliferating embryonic myocardium (304, 305). The pattern of expression of the transgene is highly reminiscent of that of the proliferation-quiescent areas described in the chicken embryonic heart by Thompson et al. (304), who proposed that the early conduction system would serve as organizational tissue for myocardial differentiation. Interestingly, the area of engrailed-2/lacZ transgene expression can be enlarged by the addition of neuregulin-1 to whole mouse E8.5-9.5 embryo cultures (250). The transgene is not induced in the free ventricular wall of mouse E10.5 embryos, when overt compact myocardium has differentiated. Optical mapping of the neuregulin-1-treated embryos demonstrated ectopic preferential activation pathways that were attributed to the recruitment of additional cardiomyocytes to the conduction system. Taken together, the transgene may thus identify the primary myocardium and the trabecular ventricular component. Large parts of this myocardium do indeed develop into components of the conduction system. However, not all parts will do so, and these are not called conduction system. For instance, the myocardium at the left atrioventricular junction is not considered to be a part of the conduction system, although nodal action potentials and low abundance of gap junctions have been observed in the lower rim of the left atrium in the formed heart of various species (196). In addition, although internodal tracts are not considered to be significant components of the conduction system, their existence is an amazingly persistent, but highly controversial issue (16, 56, 138, 182). One of the reasons for this persistency might again be confusion about terminology. If this area would be a direct derivative of the primary myocardium, it would not be surprising to find cardiac specialized primitive cells in the internodal myocardium, i.e., in the right dorsal atrial wall (56). Such an observation is not necessarily linked to the existence of internodal tracts, certainly not if these tracts are supposed to conduct rapidly and to be insulated by fibrous tissue (10). Linkage with slow (nodal) conduction seems more appropriate but would not have much physiological significance, because the activated atrial cells conduct the depolarization wave much faster to the atrioventricular node than the internodal myocardium would do.
To fully exploit the challenging data of Rentschler and co-workers it is of great interest to relate the pattern of transgene expression with the expression of relevant molecular markers, such as the genes encoding connexins (69), Anf (51), chisel (233), and other genes. Definition of the site of engrailed-2/lacZ transgene integration into the genome might reveal which endogenous gene regulatory sequences control the specific lacZ expression pattern, and thus might provide clues for the regulatory mechanism underlying the restricted pattern of transgene expression.
VI. ANTERO-POSTERIOR PATTERNING AND CHAMBER FORMATION
Cardiac chambers are formed at specific sites of the heart tube. To understand the underlying mechanisms we need to know which factors provide the myocardium with positional information for the site-specific formation of chambers, and how these factors regulate the gene programs required for chamber formation. To delineate the complicated process of local chamber formation it is useful to distinguish antero-posterior (AP), dorso-ventral (DV), and left-right (LR) patterning and the interpretation of these patterns by genes that effectuate chamber formation.
A. AP Patterning During and After Gastrulation
The position of cardiac progenitor cells in the primitive streak along the AP axis correlates roughly to the AP position of their descendants in the tubular heart (97, 258), although this relation is not strict (97). Retroviral tagging of anterior primitive streak at HH3 resulted in tagged progenitors mostly in the anterior region of the embryo including the entire heart tube at HH12 (329), indicating that the AP position within the streak is largely conserved after gastrulation. In zebrafish, single cells injected in the midblastula stage formed single colonies in the atria or ventricles, indicating separation of atrial and ventricular lineages at that stage (288). Cardiogenic mesodermal cells of HH4-6 chicken embryos infected with recombination incompetent retrovirus formed single colonies in the chambers (201). These observations indicate lineage separation of the chambers or, alternatively, limited spatial rearrangement and/or a limited number of cell divisions between the stage of labeling and analysis (288). Fate-map experiments, however, do not reveal whether and when cells have interpreted positional information involved in their phenotypic specification.
Upon transplantation, cells of the anterior streak (HH3) were able to form posterior heart structures, and vice versa, indicating that at this stage AP patterning is not fixed (137). Grafting experiments in chicken embryos after gastrulation revealed that distinct regions along the AP axis of the tubular heart of HH12 embryos mapped to extensively overlapping regions in the precardiac areas of HH5 embryos, although some AP positional correlation was observed (291). In line with these observations, recent DiI labeling experiments (248) showed that the position of labeled cells in the tubular heart did not correlate well with positions on either the AP or the mediolateral axis in the heart-forming region of HH4-7 embryos. Only at HH7-8 a definitive pattern was formed as the AP position of progenitors in the cardiac crescent correlated with the AP position of the cardiomyocytes in the heart tube. These results indicate that myocytes have not interpreted AP patterning information until HH7. De Haan and co-workers (270) analyzed developing beat rates in cultured explants of HH5-7 heart-forming regions and in situ, in “multi-heart embryos” in which the heart-forming region was cut along the AP axis (270). They observed that at HH12 posterior segments had developed higher beat rates than anterior segments, indicating that the heart-forming region at these stages has interpreted AP positional information. However, when HH5-7 precardiac mesoderm from the anterior heart-forming region was transplanted to more posterior positions, it developed beat rates appropriate for their new location (270). These results indicate that at these stages the cells seem not to be determined to their AP fate but that the positional information is imposed by regional cues. Consistently, the expression of Amhc1 could be induced by retinoic acid treatment in anterior explants before but not after HH8, the stage at which the expression of differentiation markers in the anterior portion of the heart-forming region is initiated (345). Therefore, the phenotypic diversification between anterior and posterior cells is stable after HH8. Indeed, when rotating the entire “classical” heart-forming region (comprising mesoderm and endoderm) of HH4-6 embryos 180° along the longitudinal axis and placing it back, judged by their ability to express Amhc1, originally posterior cells adapted to their new anterior position and vice versa (236). From stage HH7 onward, however, the originally posterior cells initiated Amhc1 expression at their new anterior position.
Knowledge on commitment of cardiac cells in mammals is limited. Gruber et al. (117) demonstrated that in mice after looping and initiation of chamber formation (E10.5), the downregulation of Mlc2a and αMhc gene expression in ventricular cells when grafted into the atria appeared irreversible. Taken together, from HH7-8 onward, when anterior cardiac cells start to express differentiation markers, the cells within the heart-forming region are committed to their AP fate. Mechanisms that pattern the myocardium must therefore act between HH3 (gastrulation) and HH7-8.
B. Retinoic Acid: a Cardiac AP Patterning Molecule
Retinoic acid (RA) is an important morphogen during embryogenesis (261). Excess or deficiency in RA has dramatic effects on cardiogenesis, and RA has been used extensively to study processes related to AP patterning, chamber specification, and morphogenesis. Transient exposure of zebrafish to RA specifically truncates the heart tube, starting with the outflow tract and ventricle at low doses, and further along the AP axis with higher doses (287). The effects on cardiac development were already observed when embryos were treated during and shortly after gastrulation, indicating that AP polarity can be affected during or shortly after mesoderm has acquired its cardiogenic fate. In chicken, quail, and mouse, excess RA causes “posteriorization” of the heart tube (229, 337, 346). Genes normally expressed at higher levels in the posterior region expand their domain toward the anterior region. This suggests an anterior expansion of cells fated to the posterior phenotype at the expense of anterior myocytes. RA treatment of HH5-8 explants of the anterior part of the heart-forming region comprising meso- and endoderm results in the induction of Amhc1, suggesting that the cells are not yet committed to their anterior fate. Conversely, deficiency of RA in quail causes underdevelopment of the posterior structures of the heart, most notably the sinus venosus and atria (163, 307). RA is not required for specification or differentiation of myocardium, because a heart tube expressing Nkx2-5 and sarcomeric genes including Amhc1 is formed in this model.
Retinaldehyde dehydrogenase type 2 (Raldh2) is a key enzyme for the synthesis of RA in the embryo (214, 221, 349). In quail and chicken, Raldh2 gene expression was observed exclusively in posterior mesoderm at HH8, and in the posterior heart precursors from HH9 onward (338). In HH9-chicken embryos the most anterior extension of Raldh2 expression within the embryo colocalizes with the expression of the Amhc1 gene. In mouse, Raldh2 expression is initiated in the posterior mesoderm shortly after gastrulation (220). In E7.5-7.75 mouse embryos, just before fusion of the cardiogenic mesoderm, Raldh2 expression and the transcriptional response to RA [using a retinoic acid response element (RARE) hsplacZ transgene as a read-out] colocalize and are restricted to the most posterior cardiac progenitors (214). At E8.5 (4-5 somite pairs), when a tubular heart has formed, expression and response were restricted to the posterior region of the heart, the sinus venosus. Mouse mutants that lack the Raldh2 gene, and therefore lack the ability to produce endogenous RA (221), show severe cardiac malformations, which include severely impaired atria and sinus venosus (222). Like in the quail RA deficiency model, expression of cardiac actin and myosin heavy and light chain genes were not affected.
To further explore the role of RA in AP patterning of the heart, Xavier-Neto et al. (337) generated transgenic mice with a 0.8-kbp Smyhc3 promoter, which is derived from the quail homolog of the chicken Amhc1 gene (325), coupled to the human alkaline phosphatase (hAP) reporter gene. In transgenic mice, the reporter gene was active only in the posterior heart precursors of E8.25 mouse embryos, and subsequently in the inflow tract, atria, and atrioventricular canal of E9 embryos. RA administered to pregnant females at 7.5 days post coitum, but not after 8.5 days post coitum, induced cardiac dysmorphogenesis including reduced development of the outflow tract. The ventricles coexpressed Smyhc3-hAP and Mlc2v, a gene marking the anterior part of the heart tube. When, conversely, endogenous RA synthesis was blocked, ventricles were enlarged and tapered off in the sinus venosus. Smyhc3-hAP and RARE-hsplacZ transgene expression was reduced. The observed cardiac malformations were similar to the RA deficiency models in quail and Raldh2 mutant mice. The coexpression of Smyhc3-hAP and Mlc2v in the anterior myocytes can be explained by several mechanisms, which range from the respecification of anterior heart precursors to posterior ones, to simply the activation of the RA-sensitive Smyhc3 promoter in cells that otherwise have acquired the anterior phenotype. The coexpression of these two genes also implies the presence of a distinct pathway, not under control of RA, which specifies anterior cells.
Rosenthal and Xavier-Neto (259) proposed a model for the RA mediated specification of the cardiac progenitor cells. In this model posteriorly localized Raldh2 expression is initiated shortly after gastrulation (mouse E7.5, chicken HH6), resulting in the escape of the most anterior cells of the migrating cardiac precursors from contact with RA (259). These anterior cells, which form the early tubular heart, are specified to the anterior (including ventricular) fate. The posterior cardiac mesodermal precursors will contact an RA-producing region or express Raldh2 themselves. RA triggers the posterior phenotype (inflow tract, atria) when these cells are recruited at the posterior side of the heart tube. The model has several attractive features: it fits with the defined stage (HH7) at which irreversible programming of cardiogenic cells along the AP axis takes place; it explains the short window of RA sensitivity of heart development; it includes cardiac AP patterning in a more general AP patterning mechanism that is also involved in patterning of the neural tube and axial skeleton, where RA acts as an important regulator of AP identity (261); it uncouples formation of the cardiac lineage and patterning; and it is testable. Several transcription factors have been implicated in mediating AP patterning (see below), but it remains to be established what mechanism effectuates phenotypic AP polarization of the heart tube.
RA may also be involved in the earlier differentiation of anterior (ventricular) cells compared with posterior cells during early stages of heart formation (see above). Vitamin A deficiency in mice mimicked by blocking retinol binding protein expression at E7.5, or by mutation of retinoic acid receptor genes (Rxrα, β or Rarα), causes early differentiation of subepicardial ventricular myocytes (148). Furthermore, ultrastructural analysis showed that the differentiation of the ventricular myocytes of E8.5 Raldh2-deficient embryos was more advanced than that of wild-type embryos (222). Thus RA signaling is required for a delay in differentiation of these cells. Exposure of Xenopus embryos to low levels of RA blocks expression of differentiation markers in the heart (74). It is therefore tempting to speculate that selective contact of posterior cells with RA in early cardiogenesis is causal to the AP gradient in differentiation along the linear heart tube.
C. Factors Involved in AP Patterning and Morphogenesis
1. Gata factors
Gata4, -5 and -6 are expressed in, and posterior to, the heart forming region of HH7 (1 somite) embryos and in the tubular heart of HH9+ (8 somites) embryos. The three genes, particularly Gata4, are expressed at levels higher in the posterior inflow tract compared with the anterior heart tube (141, 163). In E8 mouse embryos,
Gata4 is expressed in the anterior intestinal portal and in the inflow tract region of the fusing heart tube just anterior to the intestinal portal (206). In quail, RA deficiency resulted in specific downregulation of Gata4 expression. Administration of exogenous RA at the 1-4 somite stages (HH7-8) completely rescues the cardiac phenotype and the expression of Gata4. RA treatment of early (HH5-6) chicken embryos resulted in induction of Gata4 expression in the lateral plate mesoderm including the heart-forming region (178). In RA-treated HH7-8 embryos, Gata4 expression was expanded and upregulated in the anterior heart tube and in the lateral plate mesoderm posterior to the heart. Also in RA-treated Xenopus embryos, Gata4, -5, and -6 are specifically upregulated (140). These observations suggest that Gata family members, especially Gata4, are components of the RA signaling pathway for specification or formation of the posterior heart.
Cardiomyocyte differentiation as defined by expression of Nkx2-5 and other cardiac markers is normal under RA-deficient conditions, indicating that Gata4 is not required for specification or differentiation of cardiac precursors. Indeed, although heart formation is severely affected in Gata4-deficient mice, the defect does not result from specification of the cardiac cell lineages, but from affected folding of the embryo and by that failure of fusion of the heart fields at the midline (167, 206). The unfused heart regions of Gata4 mutant mice express the chamber marker Anf (206), indicating that initiation of a chamber-specific program of gene expression occurs (see section ixA). Chimeric mice made of Gata4-deficient embryonic stem cells and wild-type blastocysts incorporated descendants of these cells into the formed heart (167). Also when Gata4, -5, and -6 were depleted simultaneously with antisense oligomers in chicken, cardiomyocyte differentiation was not affected (141).
Expression of a dominant negative isoform of Gata4 in Xenopus resulted in the expansion of the Nkx2-5 expression domain into posterior direction, whereas in RA-treated embryos the Gata expression domain was increased at the expense of the Nkx2-5 expression domain (140). These results indicate that Gata factors restrict Nkx2-5 to a more anterior region of the heart-forming region where it regulates myocardial differentiation. However, the Nkx2-5 regulatory region is stimulated by Gata factors (reviewed in Ref. 275) and Nkx2-5 and Gata4 in synergy activate cardiac promoters in cell culture systems (77, 175). Taken together, Gata4 may play a role in posterior cardiac morphogenesis, but its precise role is still unclear.
Tbx5 is expressed in a postero-anterior gradient in the heart-forming region and tubular heart (35, 51, 178; Fig. 13) and is, like Gata4, upregulated in RA-treated chicken embryos. Tbx5 gene expression was abnormal and downregulated in the posterior part of the tubular heart of Raldh2-deficient embryos (222). Therefore, Tbx5 may be involved in AP patterning or morphogenesis linked to RA signaling (178). The function of Tbx5 in cardiogenesis has been investigated in several species. In Xenopus, an inducible dominant negative isoform of Tbx5 (Tbx5 coupled to the repressor domain of engrailed) blocked normal cardiogenesis nearly completely (131), although interference with the function of other T-box factors cannot be ruled out in this study. In zebrafish, Tbx5 deficiency (heartstrings mutation) resulted in failure of cardiac looping and deterioration of both atrium and ventricle (98). Tbx5 haplo-insufficiency in human and mouse causes relatively mild cardiac defects, which include defects in the septa and conduction system (20, 36, 177). Tbx5 deficiency in mice caused severe hypoplasia of posterior heart structures in the linear heart tube from E8 onward, showing that Tbx5 is required for formation of the posterior heart, compatible with its expression profile (36). Expression of Nkx2-5 and Gata4, and of the anteriorly expressed genes for Irx4, Mlc2v, and Hey2 and of the chamber-specific genes for Cx40 and Anf was reduced in mutant embryos (36).
When Tbx5 was ectopically expressed in the tubular heart of mice under control of the βMhc promoter, loss of anterior gene expression (Mlc2v) and retardation of ventricular chamber morphogenesis were observed (178). These observations are compatible with a role of Tbx5 in patterning, imposing posterior identity on the heart tube. Viral expression of Tbx5 in chicken hearts caused inhibition of myocardial growth and of the formation of the trabecules, consistent with the observed phenotype in the βMhc-Tbx5 transgenic mice (123). The experiments in chicken indicate that Tbx5 is involved in the downregulation of cell proliferation. Its expression was indeed found to colocalize with regions of low proliferation in several tissues. This is not necessarily the case in the heart where Tbx5 expression is observed in both the rapidly expanding atria and left ventricle and in the slowly proliferating inflow tract, atrioventricular canal, and ventricular trabecules. Taken together, Tbx5 is a component of a pathway controlling AP specification and morphogenesis. However, its position in the AP signaling pathway and the mechanism by which it imposes its activity on the forming heart still need to be defined.
3. Other factors
Mice with a targeted deletion of the Coup-TFII gene have underdeveloped atria and sinus venosus (237). Coup-TFII gene expression is normally restricted to these posterior structures. The mechanism for the failure of the posterior structures to develop is not understood but may involve affected endothelial-mesenchymal interactions, as angiopoietin-1, a proangiogenic soluble factor thought to mediate this interaction, is downregulated in mutant mice. Targeted deletion of genes for Nkx2-5, Hand2, Mef2C, and Bob revealed their importance for heart formation, which in mutants was affected more at the anterior than at the posterior side. These mutants are discussed in section ixC. The homeobox transcription factor Irx4 has been implicated in the anterior (outflow tract, ventricles) or posterior (inflow tract, atria) regulation of myosin genes in chicken and mouse (see sect. ixB). Based on their expression profiles, a number of genes have been implicated in specification or differentiation along the AP axis. Genes for the hairy related basic helix-loop-helix factors Hey1/HRT1 and Hey2/HRT2 are heterogeneously expressed along the AP axis throughout cardiac development (176, 216). Their role in AP patterning and morphogenesis remains to be established.
A significant contribution to AP patterning of the developing heart may come from the continuous recruitment of myocardium at both poles of the tubular heart during development. The recruited myocardium has been subject to conditions and signals that change during development, such as RA at the posterior pole of the heart (see above). At the anterior pole, recruited cells may have been exposed to signals like fibroblast growth factors (FGFs) and bone morphogenetic proteins (BMPs) in a manner different from the myocardium of the tubular heart (150, 204, 321). The differential exposure may contribute to the distinct phenotype of the anterior cells. The role of FGFs and BMPs in these patterning events and in chamber and vessel morphogenesis is not clear yet.
VII. CARDIAC PATTERNING ALONG THE DORSOVENTRAL AXIS
The dorsal side of the forming linear heart tube is connected to the body wall by the dorsal mesocardium, a morphological hallmark for the difference between the dorsal and ventral side of the heart tube. During looping the ventral side of the anterior ventricular region and the posterior dorsal side of the atrial region will become the outer curvatures from where the chambers will expand (Figs. 7 and 9). Therefore, chambers form in response to integrated DV and AP patterning. To date, only a very limited number of genes are known to discriminate between the ventral and dorsal sides of the heart tube. Hand1 is specifically expressed at the ventral side of the linear heart tube at E8-8.5 (23, 51, 303; Fig. 13). After looping, its expression is confined to the left side of the atrioventricular canal (derived from the ventral side of the linear heart tube; Refs. 38, 60), the outer curvature of the ventricles, and the outflow tract. The right ventricle expression diminishes by E8.5-9.5 (88, 303), reflecting AP as well as DV patterning. The function of Hand1 in heart formation is not clear. Development of Hand1-deficient mice is arrested at E7.5, and embryos die between E8.5 and E9.5 (88, 251). However, a tubular heart is formed that expresses cardiac markers. Notably, Anf is expressed in Hand1 null-mice (88), suggesting that a chamber-specific program of gene expression is initiated despite the failure of the heart to loop and expand. Mlc2v expression was also observed, indicating that AP patterning was not severely affected either. In tetraploid rescue experiments, embryos died at E10.5 due to cardiac failure (251). Development of the hearts of these embryos was arrested at the onset of looping, and trabecules were not formed, similar to the Nkx2-5 null mice that lack Hand1 expression in the heart (see below). When chimeric mice were made of Hand1 mutant ES cells and wild-type (Rosa26-lacZ) cells, the outer curvature of the heart tube was generated invariably of descendants of the wild-type cells, demonstrating that Hand1 mutant cells cannot contribute to this part of the heart in E9.5 embryos (252). It is therefore likely that Hand1 plays an important role in the morphogenesis of the ventral side/outer curvature of the heart tube.
In mouse and chicken, Anf gene expression is first observed at the ventral side of the putative ventricular region of the fused heart tube, just after the onset of looping (51, 133, 347). A small region at the dorsal side does not express Anf. The ventral side of the tubular heart is derived from the lateral sides of the bilateral heart-forming region. In HH9 chicken, just posterior to the site of fusion of the crescent the Anf gene was expressed lateral in the cardiac mesoderm (133). This indicates that mediolateral signaling precedes the DV pattern. Transgenic mice with a chicken Gata6 gene enhancer driving the lacZ reporter gene showed AP-restricted expression in heart-forming region (E7.5) and in the inflow tract of the tubular heart (E8.5; Ref. 60) that was restricted to the lateral sides. Expression became confined to the dorsal and ventral myocardial walls of the atrioventricular canal that are lined by cushions. These results indicate that in chicken and mouse, mediolateral patterning in the heart-forming region is present that precedes DV patterning in the tubular heart.
Expression of the genes for Chisel and Cited1 (Msg1) confirms the presence of DV patterning in the tubular heart, as they are expressed selectively at the ventral side of the future ventricular region of the linear heart tube of E8.25 embryos (76, 233). After looping, expression is confined to the chambers. Also, the onset of Irx3 expression in the myocardium and Irx5 expression in the endocardium at the ventral side of the E9 mouse heart (52) indicates DV patterning in the linear heart tube.
In summary, expression patterns of Hand1, Anf, Chisel, Cited1, Irx3, and Irx5 demonstrate the presence of DV differences in transcription regulation in the early tubular heart, in line with the morphological observations dealt with in section iii. These genes are specifically expressed at the outer curvature and, except for Hand1, are restricted to the forming chambers (see sect. ix), underlining the relation between early DV patterning and chamber formation at the outer curvature.
VIII. LEFT-RIGHT SIGNALING IN CHAMBER SPECIFICATION AND MORPHOGENESIS
The LR identity of the heart-forming regions and the direction of looping of the atrial and ventricular regions of the tubular heart are evolutionary conserved, and both depend on LR signaling in the embryo (reviewed in Refs. 40, 120). Incorrect decisions regarding laterality may result in serious cardiac malformations (32, 149). LR signaling, however, is not required for chamber formation per se and is discussed here only briefly. A cascade of signaling molecules that regulate the establishment of LR identity of the embryo (reviewed in Ref. 40) culminate into the expression of the homeodomain factor Pitx2 in the left side of the visceral organs, including the heart. Asymmetric Pitx2 expression seems to be sufficient for establishing LR identity of the heart.
During fusion of the left and right heart-forming region, a tubular heart is formed that consists of two halves with a left and right identity, respectively. The contribution of each heart-forming region to the fused heart tube is essentially equal. However, the left heart-forming region contributes more to the anterior portion of the tube and the right heart-forming region more to the posterior part (248, 290), indicating that LR contributions are influenced by AP patterning. In models of heterotaxy, the heart is normally patterned and formed in relation to its AP and DV axis, indicating that AP and DV specification can act independently from LR specification. In contrast, RA, involved in AP patterning (see above), is required for the LR asymmetry pathway, as randomized cardiac situs occurs under conditions of RA excess or deficiency (45, 350). RA appears to control the level and location of expression of components of the LR signaling pathway (lefty, nodal, Pitx2).
The left and right atria are largely derived from the corresponding side of the tubular heart. This can be shown by following Pitx2 expression, a bona fide marker for the fate of the left portion of the heart tube (38), in the developing heart (Fig. 7). Morphologically, isomerism leads to two left atrial appendages, which do express Pitx2, or two right atrial appendages, which do not express Pitx2. Isomerism of the ventricular chambers is less obvious. The atrioventricular canal, the ventricles, and the outflow tract are aligned along the AP and DV axis before looping, and each component is composed of descendants of the left and right heart-forming regions. As mentioned in section iiiB, the AP arrangement of the ventricular chambers of the heart tube is converted with looping into a LR arrangement (38, 51, 303). Due to this process, the ventricular part derived from the left half is found at the ventral side of the heart, the part derived from the right side is found at the dorsal side (38) (Fig. 5). Indeed, in mutant mouse strains that have heterotaxy, such as the iv/iv mouse, both double-sided absence and double-sided presence of Pitx2 expression in the ventricles are observed. This phenomenon is associated with double-outlet right ventricle, a condition of improper alignment of the ventricles and great arteries. Therefore, for isomerism of the ventricles, typical features of the dorsal side or the ventral side of the ventricular region are expected to be either present or missing at both sides. These features have not been described, which may indicate that the condition is incompatible with life.
In corrected transposition of the great arteries (CTGA), the morphologically left and right ventricles are found at the right and left side, respectively, and each fulfills the physiological function of the other. This suggests that only the ventricular region has been inverted, by left-sided looping of the ventricular loop, whereas the remainder of the heart is correctly specified. Indeed, in patients with CTGA, the presence of a (rudimentary) atrioventricular bundle at the ventral (anterior) side of the ventricles has been reported (6). Furthermore, in the iv/iv mouse model of cardiac heterotaxy, symmetry in presence or absence of Pitx2 expression was sometimes observed only in atria or in ventricles. Together with the CTGA, this suggests that regulation of LR identity of different components of the heart may be modular and may occur relatively late in development.
IX. CHAMBER-SPECIFIC AND REGIONALIZED TRANSCRIPTIONAL PROGRAMS
A. Chamber-Specific Patterns of Gene Expression
As outlined in previous sections, morphological and electrophysiological criteria distinguish chamber myocardium from primary myocardium. As a consequence, gene products that discriminate between these types of myocardium are present as well. The Anf gene is not expressed in the primary linear heart tube before the onset of looping (E8.5). First expression is detected at the ventral side that will form the trabeculated portion of the left-ventricular chamber. During and after looping, Anf gene expression gradually increases in the right ventricular chamber. Atrial expression of Anf is first observed at the end of looping, E9.25, at the dorsolateral side of the atrial region. Expression is never observed in the primary myocardium of the inflow tract, atrioventricular canal, inner curvatures, and in the outflow tract (Fig. 11) (51, 195, 347), making this gene a suitable marker for the chambers. The expression in the right ventricle is weak.
From E11 onward, Anf no longer is expressed in the rapidly proliferating compact myocardium at the epicardial side of the ventricles and becomes restricted to the trabecules that will form the subendocardial ventricular conduction system (Purkinje network, see sect. vF and Ref. 208). Anf expression is not observed in the sinoatrial node and atrioventricular node that are derived from the inflow tract and atrioventricular canal, respectively. The pattern of Cx40 gene expression is highly reminiscent of that of Anf. In contrast to Anf, Cx40 is expressed in the atrial septum and in the caval veins (but not in sinoatrial node), indicating that these component parts form distinct transcriptional domains. Another gap junction gene, Cx43, shows a cardiac pattern very similar to that of Anf and Cx40, albeit that atrial chamber myocardial expression is initiated later (69, 312). Furthermore, Cx43 is expressed at high levels in the compact ventricular myocardium.
Expression of Chisel, encoding a small muscle-specific protein (233), was first detected at the ventral side of the fusing heart tube at E8.25. It has an expression profile very similar to that of Anf. However, it is expressed at comparable levels in left and right ventricles and at higher levels in the compact layer of the ventricles than in the trabecules (51, 233). Initiation of Cited1 (Msg1) gene expression (76) is at the ventral side of the linear heart tube (E8-8.25). Thereafter its expression profile is very similar to that of Anf, albeit that atrial expression is weak. Serca2a, which encodes sarcoplasmic reticulum Ca2+-ATPase, is expressed at moderate levels in the entire cardiac crescent and linear heart tube in a postero-anterior gradient. The gene is upregulated in the atrial and ventricular chamber myocardium (colocalized with Anf, Chisel, and Cx40 expressing myocardium) that form at the outer curvatures (210), but retains its original pattern in the remnants of the primary myocardium at the inner curvatures, atrioventricular canal, and outflow tract. The homeobox gene Irx5 is expressed in the endocardium lining the trabecular myocardium of the ventricles and the atrial myocardium (52). Endocardium lining the cushions, present in the inflow tract, atrioventricular canal, inner curvatures, and outflow tract (Fig. 9), does not express Irx5. Its expression therefore coincides with that of Anf, Cx40, and Chisel. The gene for Tbx2, a member of the T-box transcription factor family (44), was found to be strictly associated with the primary myocardium. It is expressed both in myocardium and cushions of the inflow tract, the inner curvature of the atrial loop, the atrioventricular canal, the inner curvature of the ventricular loop, and the outflow tract (118, 340). Its pattern is strikingly complementary to that of Anf (Fig. 13). The significance of this factor in the primary myocardium is discussed in section ixB. Several genes show regional expression in the tubular and subsequently looped chamber heart. Mlc2v gene expression is first observed at E8, in the paired lateral regions of the fusing heart-forming regions, just before the formation of the tubular heart (188, 219). This pattern provides one of the first indications that the mammalian heart is effectively patterned along the AP axis before formation of the tubular heart. When the tubular heart is formed at E8, Mlc2v and Irx4 are expressed in a large portion of the tube that includes dorsal and ventral sides but excludes the inflow and outflow tract (Fig. 13) (223). After looping, the atrioventricular canal, the ventricles, the inner curvature of the ventricular loop, and the proximal outflow tract express Mlc2v and Irx4 (34, 51, 52, 223). Therefore, neither gene is restricted exclusively to the ventricular chambers during development, and both exceed the morphological and phenotypic boundaries of the ventricular chambers. Because the ventricles rapidly proliferate and expand, they will form the bulk of the Mlc2v-expressing domain. Therefore, Mlc2v and Irx4 expression will become more and more “ventricle specific” during development. Using a Cre-lox based method to follow the fate of myocytes and their progenitors that expressed Mlc2v (Cre gene under control of the Mlc2v locus; Ref. 47), we found labeled cells outside the ventricles (e.g., in the lower rim of the atria). This shows that the Mlc2v-expressing domain in the embryonic heart is broader than that fated to form the ventricular chambers (unpublished observations). The patterns of Mlc2v and Irx4 constitute an AP-defined segment (from atrioventricular canal to proximal outflow tract) throughout cardiac development (Fig. 13). The ventricular chambers form at the ventral side and outer curvature within the Mlc2v- and Irx4-expressing segment (Fig. 13).
A number of genes, including Amhc1, Smyhc3, αMhc, Mlc2a, and others (51, 92, 187, 325, 346), are selectively expressed in, or become confined to, the posterior portion of the tubular and chamber-forming heart. Here, again, the patterns are restricted along the AP axis only, because also the inflow tract, inner curvature of the atria, and the atrioventricular canal express these genes. Within this posterior portion (or segment), atrial chamber myocardium will form and expand from the dorsolateral sides (Fig. 13).
In conclusion, several chamber-specific genes are restricted to the morphological and physiological chambers at embryonic stages, whereas others are expressed in a broader domain along the AP axis, forming segments of expression within the tubular and looped heart. Positional information along both the AP and DV axes is required for the site-specific initiation of expression of the chamber-specific genes.
B. Regulation of Chamber-Specific Gene Expression
To gain insight into the transcriptional program underlying the regionalized chamber formation, two strategies appeared to be informative: 1) transgenic analysis to study regulatory DNA sequences that provide regional activity (152) and 2) inactivation or ectopic activation of genes for cardiac transcription factors. In the next sections, studies that have provided insight into the control of regionalized gene expression or chamber formation are reviewed.
1. Regulation of atrial gene expression: Irx4 and the Smyhc3 gene
Bao et al. (17) identified a novel member of the vertebrate Iroquois family of homeobox genes, Irx4, which is specifically expressed in the chicken heart from HH10 onward (17). Expression is confined to an AP-restricted segment of the tubular heart (see above). In later stages Irx4 expression is largely confined to the ventricles. Vmhc1 and Amhc1 gene expression were detected in the entire tubular heart from HH9+ and HH10, respectively. This Amhc1 pattern is different from the posterior-restricted pattern found by Yutzey et al. (346). The probes used in these studies contain regions homologous to other myosin heavy chain genes expressed in the heart tube. This and differences in the applied sensitivity of the in situ hybridization account for the discrepancy in observed patterns (unpublished observations). With further development Amhc1 and Vmhc1 expression became restricted to posterior and anterior portions of the looping tubular heart, respectively. The region of Irx4 expression coincides with the Vmhc1 expression domain and is complementary to the Amhc1 expression domain. Viral expression of Irx4 in the heart caused downregulation of the Amhc1 gene and upregulation of the Vmhc1 gene in the atria. Ectopic expression of the homeodomain of Irx4 coupled to the repressor domain of engrailed caused the opposite effect. Therefore, Irx4 was implicated in chamber-specific gene regulation and maintenance or even determination of cardiac chamber formation in chicken. The mechanism by which Irx4 activates the Vmhc1 gene in the ventricle is not clear but may be indirect, since functional studies so far indicated that members of the Iroquois family of factors primarily act as repressors (41, 104, 165). Irx4 expression does not respond to RA signaling (34, 222), indicating that Irx4 is not directly responsible for the RA-inducible expression of Amhc1. However, Irx4 and Rxrα cooperate in the repression of the quail homolog of Amhc1 (see below). In mouse, Irx4 is expressed in a pattern similar to that of Irx4 in chicken (34, 52). Therefore, Irx4 function may be conserved between vertebrate species. Irx4 deficiency in mouse caused no cardiac malformation but resulted in postnatal derepression of Anf and α-skeletal actin, and in cardiomyopathy after several months (33). Before birth, Hand1 was downregulated in the ventricles of Irx4 null mice, whereas the Smyhc3-hAP transgene, active in the inflow tract, atria, and atrioventricular canal in mouse (see sect. viB), was upregulated. Therefore, Irx4 is involved in some aspects of the ventricle-specific program of gene expression. The relatively mild phenotype of the Irx4-deficient mice may result from compensation by other Irx family members. Using the conserved chicken Irx4 homeodomain as a probe, we isolated five other Irx family members from a mouse cardiac cDNA library, which all are expressed in highly specific patterns in the myocardium (Irx1, -2, -3) or endocardium (Irx5, -6) (52, 215). Their spatial and developmental patterns are markedly different from Irx4, indicating that they can compensate Irx4 deficiency only partially. The function of the other Irx genes in the specification or formation of the separate cardiac components has yet to be determined.
The Smyhc3 gene, homolog of the chicken Amhc1 gene, is initially expressed in the entire tubular heart and becomes downregulated in the ventricles during development (325). Wang and co-workers (323, 325) showed that the first 840 bp of its promoter region are sufficient to mediate atrial activity and ventricular repression. Functional analysis of the regulatory fragment revealed the presence of a functional GATA factor binding element and a vitamin D response element (VDRE). The GATA binding element is required for cardiac activity of the promoter. However, mutation of the VDRE completely abolished ventricular repression. Further studies showed that a heterodimer of retinoid X receptor α (Rxrα) and vitamin D receptor (Vdr) binds the VDRE. Because the receptors are expressed in the entire heart, an unidentified ventricle-specific corepressor involved in the ventricle-specific repression was postulated (325). The observation that the Smyhc3 transgene is upregulated in the ventricles of Irx4-deficient mice suggests that Irx4 could be the corepressor of Smyhc3. Indeed, Irx4 was found to repress the promoter in a VDRE-dependent manner but did not interact with the DNA (324). The Rxrα receptor within the Vdr/Rxrα heterodimer was found to interact with Irx4. Together these findings have revealed a mechanism for the ventricle-specific repression of Smyhc3, in which an inhibitory complex composed of Vdr/Rxrα receptor and Irx4 binds to the VDRE and represses transcription. It is not clear whether the chamber-specific repression mechanism found for the Smyhc3 gene applies to other genes as well. However, a number of cardiac genes that are expressed in a chamber-restricted pattern, are initially expressed in the entire tubular heart (51, 92, 187). It is therefore tempting to speculate that repression during chamber formation is a general mechanism by which the expression domain of genes becomes restricted.
2. Regulation of ventricular gene expression: the Mlc2v gene
The mechanism controlling Mlc2v gene expression has been subject to extensive study (219). Transgenic mice in which the proximal (250 bp) Mlc2v promoter was coupled to lacZ showed expression in a steep AP gradient that after looping results in expression in the outflow tract and right ventricle, tapering of in the left ventricle (260). This pattern of transgene expression differs from the expression profile of the endogenous Mlc2v gene (both ventricles, the ventricular inner curvature, atrioventricular canal, and proximal outflow tract; see above). A pattern similar to that of the 250-bp promoter was observed when a dimerized 28-bp HF-1 fragment of the Mlc2v promoter containing HF1a and HF1b/MEF2 elements was coupled to the minimal Mlc2v promoter fragment. This core element therefore contains sufficient information to drive expression in the anterior myocardium of the linear and looped heart tube. Different lines with the same 250-bp promoter construct and dimerized element constructs showed expression that to a variable degree penetrated into the left ventricle. This indicates that the promoter senses an AP-graded signal along the heart tube and that the sensitivity of its response depends on the site of integration of the transgene construct into the genome. One of the lines has been used to monitor right ventricular specification in various mouse-mutant backgrounds with affected cardiogenesis (122, 341). In light of the pattern of the Mlc2v -lacZ transgene, it may however be that the promoter only senses a transcriptional signal in the anterior region of the heart tube within which the outflow tract and right ventricle will develop.
The patterns of the Mlc2v-lacZ transgene and endogenous gene differ, and the activity of the Mlc2v promoter is weak compared with the endogenous gene (47) and in contrast to the endogenous gene is not downregulated in Nkx2-5 mutant mice (188, 298, 341). Therefore, expression of the endogenous gene depends on sequences outside the proximal promoter (260). Although within the proximal Mlc2v promoter some specific cardiac activity is contained, its main function may be to relay transcriptional input from distal regulatory elements to the transcriptional machinery. These distal elements drive very strong expression and restrict Mlc2v expression within its boundaries (from atrioventricular canal to proximal outflow tract) along the AP axis. Evidence that supports the notion that the proximal Mlc2v promoter primarily relays activity from distal elements was obtained from chimeric promoter studies in transgenic mice (118 and unpublished data). When the Anf regulatory sequences (position -638 to -138), which are active in atria and ventricles but not in atrioventricular canal and outflow tract, were coupled to the Mlc2v promoter, the Anf pattern was observed, indicating that the Anf regulatory sequences impose their activity on this promoter. Similar observations were done when the Mlc2v promoter was replaced by a small cardiac troponin I (cTnI) promoter fragment. However, when the Mlc2v promoter was placed upstream of the cTnI promoter, the pattern typical for the cTnI promoter was observed, indicating that the Mlc2v promoter itself has only weak regulatory activity indeed.
The anterior activity of the proximal Mlc2v promoter has been used as a tool to gain insight into the regulation of localized gene expression within the heart. The proximal Mlc2v promoter contains binding sites for a number of cardiac enriched as well as ubiquitous transcription factors (219). The expression patterns of these factors, however, do not correlate with the anterior activity of the Mlc2v promoter. Because the dimerized 28-bp HF-1 regulatory unit confers anterior activity to a reporter gene, the factors that interact with the sites within this unit have been studied most extensively in relation to the specific anterior activity. The ubiquitously expressed factor EFIA, the rat homolog of human YB-1, was found to bind to the HF-1a site within the 28-bp HF-1 unit and to activate the Mlc2v promoter in cardiac cells (351). With the use of YB-1 as bait in a yeast two-hybrid screening, the factor Carp (cardiac ankyrin repeat protein) was isolated (352). Carp, a cofactor for YB-1, functions as a repressor. The gene is expressed specifically in the heart and at E10 is reduced in the ventricle compared with atria and outflow tract, compatible with the hypothesis that Carp negatively regulates endogenous Mlc2v. The Carp pattern, however, is not compatible with the pattern observed for the 250-bp Mlc2v promoter or the dimerized HF-1 elements that Carp is thought to interact with via YB-1. Further analysis of Carp regulation (168) showed that the first 0.3 kbp of its promoter drives reporter gene expression in the outflow tract and right ventricle from E9 onward, in a pattern similar to the Mlc2v-lacZ transgenes. When a dimerized fragment of the upstream 128-bp promoter was used, expression was restricted to the outflow tract. Therefore, consecutive restriction of the promoter results in a restriction along the AP axis. In conclusion, both Mlc2v and Carp regulatory fragments are sensitive to AP patterned activity, but the underlying mechanism is still enigmatic.
3. Regulation of chamber-specific gene expression
In section ixA, the Anf and Cx40 genes were introduced as specific markers for the forming chambers. Both genes are therefore excellent model genes to study mechanisms that control chamber-specific expression. The Anf gene encodes ANF, a hormone that in the formed heart is secreted by the atria and involved in blood-pressure regulation. It serves as a marker for cardiac hypertrophy, during which the gene is strongly upregulated in the ventricles (50). The promoter region of the Anf gene has been used as a paradigm for regulatory pathways that control cardiac-specific gene expression (79). The upstream 0.6 kbp of the Anf regulatory region are sufficient to drive expression in transiently transfected pre- and perinatal atrial and ventricular cells (12). Among the factors that interact with the Anf promoter are Gata family members, Nkx2-5, Mef2C, Tbx5, Tbx2, Srf, and Myocardin (36, 77, 78, 118, 124, 125, 175, 213, 279, 322). Several of these factors were shown to interact with each other and with sites within the promoter. In transient transfection studies, Nkx2-5 was reported to activate the promoter in synergy with Gata4, Srf, or Tbx5, respectively. From mutant mice it was deduced that Anf requires Nkx2-5, Mef2C, and Tbx5 directly or indirectly for its expression (36, 122, 183, 298). Irx4 null mice show higher Anf expression in the ventricles after birth (33), which suggests that Irx4 may be a component of the pathway responsible for ventricular downregulation of Anf in the compact ventricular myocardium after birth (12, 347). Together, these studies have provided valuable insights into the mechanism of cardiac-specific expression of Anf, and into the mechanisms underlying the cooperative action of different classes of transcription factors in the regulation of cardiac gene regulation. They do not, however, provide a mechanism for its relatively late onset of expression, shortly after the onset of looping. Furthermore, the spatial expression patterns of the factors involved do not explain the restriction of Anf gene expression to the chamber myocardium of the embryonic heart and to the atrial appendages in the postnatal heart.
In transgenic mice, the upstream 0.6-kbp promoter fragment of the Anf gene appeared to be sufficient to recapitulate the spatial and developmental pattern of the endogenous gene (87, 118, 162). Sequences required for the strong response to hypertrophy in the ventricles, observed for the endogenous Anf gene, are not present in this fragment in vivo (162). This observation indicates that the regulatory region of the Anf gene is modular and that the hypertrophy pathways are different from the regulatory pathways for the spatial and developmental regulation. Furthermore, transient transfection analysis of Anf promoter parts in ventricular, atrial, and noncardiac cells suggest that the promoter is composed of modules (79). The most upstream region was found to be required for activity in isolated ventricular cells and the middle portion for activity in atrial cells. Such modularity has been found for several other genes and seems a common feature of regulatory sequences (15, 275). Each regulatory module has a specific contribution to the overall activity of the gene (15). The part most proximal to the transcription start site of the Anf gene is required to relay the transcriptional activities from the upstream parts, reminiscent of the function of the proximal promoters of the Mlc2v and cTnI genes (see previous section). Indeed, Mlc2v or cTnI promoters can freely exchange the proximal part of the Anf promoter without loss in activity or pattern in vivo (118). Whether the ventricular and atrial modules indeed confer compartment-specific activities in vivo remains to be established.
In homozygous Tbx5 mutant mice, several genes were found to be downregulated, including the chamber markers Anf and Cx40 (36). Surprisingly, these two genes were specifically reduced in heterozygous mutants, indicating that Tbx5 is a rate-determining factor in the activity of both genes. Consistent with the observed requirement of Anf and Cx40 for Tbx5, careful analysis of expression patterns revealed that Anf and Cx40 are never observed outside the Tbx5 expression boundaries. After looping, expression of Tbx5 is absent from the most anterior components, the outflow tract and most of the right ventricle (35, 51, 178). Also, in the left ventricle expression disappears gradually from the compact layer. Anf and Cx40 are not expressed in these parts of the heart. Therefore, Tbx5 represents one of the first examples of a factor that directly determines several aspects of the regionalized patterns of two downstream cardiac genes.
The mechanism by which Tbx5 regulates Anf and Cx40 was studied in in vitro assays (36, 101, 125). Both Anf and Cx40 contain T-box binding elements (TBE), which are homologous to one-half of the palindromic brachyury (T) binding site (160). Both sites are functional targets of Tbx5. In a yeast two-hybrid screen, Tbx5 was found to interact with Nkx2-5 (125). Nkx2-5, which is implicated in the positive regulation of Anf through NK factor binding elements (NKE), binds with Tbx5 to a combined TBE-NKE within the Anf promoter. These factors synergistically activate Anf and Cx40 promoter-driven gene expression in cultured cardiac and noncardiac cells.
To gain insight into the chamber-restricted expression of Anf, we generated a series of chimerical regulatory constructs in which the upstream region of the Anf gene was placed upstream of different cardiac promoters (118). A small cTnI gene promoter fragment that is predominantly active in the atrioventricular canal was specifically extinguished in the atrioventricular canal by Anf sequences, whereas the Mlc2v promoter, active in the outflow tract and right ventricle, was extinguished in the outflow tract. However, the typical chamber expression pattern of the Anf promoter was retained irrespective of the proximal promoter fragment used. Therefore, upstream Anf sequences are stimulating Anf transcription in the chambers and are repressing Anf transcription in the myocardium of the atrioventricular canal and outflow tract. When we inactivated the TBE or NKE (see above) within the chimeric Anf-cTnI construct by mutation, the chamber expression was not abolished. In contrast, the repression in the atrioventricular canal was relieved. These observations clearly demonstrate that in vivo these sites are not essential for activation by either Nkx2-5 or by Tbx5 but that factors interacting with these sites are components of an inhibitory pathway. Analysis of the expression patterns of other T-box genes revealed that the transcriptional repressor Tbx2 is expressed selectively in the myocardium and cushion mesenchyme of the inflow tract, atrioventricular canal, inner curvature, and outflow tract, strictly complementary to the expression of Anf and Cx40 (Fig. 11). Functional studies provided further evidence that Anf and Cx40 are functional targets for Tbx2 and that Tbx2 forms a stable complex with Nkx2-5 on the combined TBE-NKE of the Anf promoter (118). In a simple model for chamber-restricted expression, Anf and Cx40 are activated by Tbx5 in the chamber myocardium and repressed by Tbx2 in primary myocardium (Fig. 14). Nkx2-5 functions as a “cardiac” accessory factor for both T-box factors, restricting their action to specific cardiac genes. The T-box factors provide the first clue of a mechanism underlying the localized differentiation of chamber myocardium within the tubular heart by repression of a program in the primary myocardium. Results of transgenic embryos expressing Tbx2 in the heart favor this hypothesis. Heart development is blocked in the tubular heart stage, and expression of Anf is inhibited (unpublished observations). Preliminary studies have disclosed a similar role for Tbx3. Its expression becomes confined to the components of the conduction system.
4. Localized activity of other regulatory DNA fragments
A number of cardiac promoters in addition to the Smyhc3, Mlc2v, Carp, and Anf promoters have localized activity in the heart (see Ref. 152 for a review). Expression of the gene for Hand2, a factor that is required for the formation of the anterior region of the tubular heart (see below), is initially expressed in the entire tubular heart but becomes restricted to the anterior region during looping (286). A 1.5-kbp fragment of the regulatory region of the mouse Hand2 gene drives lacZ reporter gene expression in a pattern similar to that of the endogenous gene (194). Two conserved Gata factor-binding sites within the fragment are essential for the regionalized activity, although Gata factors themselves are not chamber restricted. These findings suggest that positive or negative coregulators cooperate with Gata factors to control regionalized gene activity. Similar observations were done for a Desmin gene enhancer fragment. Here, the right ventricle-outflow tract specific expression was dependent on a binding site for Mef2 factor family members (166), which are present and active in the entire developing heart (81, 217). Further analysis of the regulatory regions of Hand2, Desmin, and other genes is required to understand the mechanisms that are involved in their regionalized activity.
An intriguing aspect of a number of promoter fragments, most of which have been mentioned above, is their localized activity either anterior (outflow tract and right ventricle) or posterior (left ventricle and further posterior) to the interventricular connection. Fragments of Mlc2v, Carp, Nkx2-5, Gata6, Sm22α, Desmin, Dystrophin, Hand2, and Hop are active anterior to this connection, whereas fragments of cTnI, Mlc3f (2E), and α-cardiac actin are active posterior to this connection (46, 61, 118, 152, 168, 194, 275). The interventricular border coincides with the border of expression of the Fgf10 enhancer trap LacZ transgene that marks the contribution of cells to the tubular heart between E8 and E10 (150; the “anterior heart field,” see sect. vB). A possible implication is that the anterior heart field forms a conserved transcriptional domain, added during evolution of the heart of higher vertebrates, and that cardiac genes have developed regulatory fragments to ensure their activation in this domain. Whether the cells of the anterior heart field form a distinct population in which the regulatory elements become active once the cells have differentiated into cardiac cells, or whether the cells move into an AP patterning field resulting in the activation of the regulatory elements by anterior signals remains to be established.
5. Regulation of regionalized transcription factors
The role of Tbx2, -3, and -5, but also of Irx4 in the compartment-specific regulation of genes such as Anf, invokes the question of what mechanisms are involved in the regionalized expression of these regulating factors. Although little is known about their regulation, several studies indicate that both groups of genes are downstream of BMPs. BMPs are members of the transforming growth factor-β superfamily of extracellular signaling peptides of which multiple members are expressed in highly localized patterns in the chicken and mouse heart during development (1, 75, 144, 189, 284, 340, 348). Functional studies indicate that BMPs are obligatory for early cardiac development (273, 274). Genetic inactivation of components of BMP-signaling pathways revealed roles in multiple steps of heart development, including early heart morphogenesis and affected cushion formation and outflow tract septation (42, 96, 155, 284, 348).
In the chamber-forming heart (mouse E9.5; chicken HH16) Bmp4 is expressed in the inflow tract and outflow tract, whereas Bmp2 is expressed in the atrioventricular canal (1, 340, 348). The genes for Tbx2 and -3 are expressed in highly specific patterns in the chamber-forming heart (102, 118, 340, 340; unpublished observations). The patterns largely coincide with those of Bmp2 and -4 together (inflow tract, atrioventricular canal, inner curvature, outflow tract). Ectopic application of BMP2 via soaked beads specifically induced Tbx2 and -3 expression in the cells surrounding the beads (340). Tbx5 was not upregulated. Moreover, Tbx2 expression was specifically downregulated in the hearts of Bmp2-deficient mouse embryos. These observations suggest that BMPs are part of the regulatory pathway that controls the localized expression of Tbx2 and -3, which in turn repress the activation of chamber-specific genes in the primary myocardium of the inflow tract, atrioventricular canal, inner curvature, and outflow tract (see previous section).
The Irx4 gene is a member of a family of six Iroquois homeobox genes found in vertebrates (27, 239). The genes are localized in two paralogous clusters of three genes each (Irx1, -2, -4 on chromosome 13, Irx3, -5 and -6 on chromosome 8; Refs. 27, 132, 239). The genes of each cluster show strong similarity in expression pattern, indicating that they are regulated by shared regulatory elements (132). In Xenopus, the Irx1 homolog Xiro1 was shown to be activated by Wnt signaling and to be negatively regulated by BMP4, whereas Xiro1 itself was able to inhibit Bmp4 gene expression (104). Also Xiro3 is activated by anti-BMP signals and able to repress Bmp expression (165). Therefore, a role for BMPs in the regulation of the Irx genes in other vertebrates is anticipated. Although no experimental evidence has been presented for reciprocal Irx/Bmp regulation in the heart, cardiac expression of Bmp2 and -4 does not colocalize with any of the six Irx genes, with the exception of Irx4 that colocalizes with Bmp2 in the atrioventricular canal.
C. Gene Deficiency Reveals Pathways Involved in Chamber Formation
Targeted deletion of genes in embryonic stem cells, and the subsequent generation of mouse lines or chimeric mice, has revealed the importance of several factors for cardiac morphogenesis. Overviews of the cardiac defects in various species that result from gene mutations are given in References 89, 262, 285. Disruption Rxr and Rar genes, N-myc, Tef-1, Nf-1, and others results in thin ventricular walls, possibly as a result of proliferation defects of the myocardium. Disruption of Nf-atc, Smad-6, and Bmp6/7 leads to defects in septation or valve formation. In the next paragraphs the consequences of deficiency of factors involved in chamber formation are discussed.
Signaling between endocard, myocard, and cardiac jelly, the extracellular matrix in between the endocard and myocard, is important for the formation of trabecules at the outer curvature and of cardiac cushions that line the wall of the nonchamber myocardium. Mice in which the genes for neuregulin or its tyrosine kinase receptors, Erbb2 or Erbb4, are inactivated do not form trabecules at the ventricular outer curvature, whereas the compact outer layer of the ventricles is thinner or unaffected (84, 99, 174, 198). The lack of trabecules may result from defects in signaling between neuregulin, expressed in endocardial epithelium and Erbb2 or -4, expressed in myocardium. Whether any of these components is responsible for the localization of the formation of the trabecules is not clear. Expression of the genes for neuregulin or its receptors in the heart is not restricted to the regions where trabecules form. Furthermore, expression of the Erbb2 receptor in the whole heart of Erbb2 mutants did not result in expansion of the trabecular zone or of the chamber myocardium, although it rescues the cardiac phenotype (335). Disruption of the serotonin 2B receptor resulted in a phenotype similar to that of neuregulin or Erbb mutant mice, possibly because Erbb2 is downregulated in these mutants or because signaling through heteromeric GTPases Gαq and Gα11 is affected. E10.5 mouse embryos that lack both GTPase encoding genes have hypoplastic ventricular wall and trabecules (224). Hyaluronan synthetase-2-deficient mouse embryos are unable to produce hyaluronan, an important component of cardiac jelly (37). Mutant mice have no trabecules and atrioventricular cushions. Hyaluronan and other cardiac jelly components are ligands for CD44, which is associated functionally with Erbb2 (29), providing a possible link between hyaluronan and formation of the trabecules. The endocardial cushions appeared to be reduced in size in neuregulin- and Erbb3-deficient mice (84), and to a lesser extent in Erbb2- and Erbb4-deficient mice, indicating that signaling between neuregulin and Erbb3, which is expressed in cushion mesenchyme (198) and presumably heterodimerizes with Erbb2, is important for cushion formation.
Drosophila lacking the NK class homeodomain gene tinman does not form a dorsal vessel (26). Expression of dominant negative isoforms of XNkx2-3 or XNkx2-5, Xenopus homologs of tinman, can cause complete elimination of cardiac gene expression and cardiogenesis (95, 116). In contrast, mice deficient in Nkx2-5 do form a tubular heart that is arrested during looping (188, 298). Therefore, Nkx2-5 is not required for establishment of the cardiac lineage, possibly as a result of functional redundancy. The Nkx2-5-deficient embryos do not show affected proliferation or enhanced apoptosis, but the atrial and ventricular chambers fail to expand. The right ventricle that normally forms later than the left ventricle does not visibly form, endocardial cushions are not formed, and the ventricular trabecules do not form either, suggesting a general inhibition of cardiogenesis after the establishment of the tubular looping heart. A number of genes are specifically downregulated. These genes include Anf, Bnf, and Chisel, markers for the forming chamber myocardium; Hand1, required for the formation of the outer curvature; and homeobox gene Hop, involved in modulation of Srf activity and required for normal cardiomyocyte proliferation and differentiation (23, 46, 233, 252, 280, 298). The expression of Anf and Bnf is affected more in the ventricle than in the common atrium, suggesting that Nkx2-5 is more important for anterior than for posterior gene regulation. Irx4 and Mlc2v, two genes that are patterned along the AP axis, are expressed in Nkx2-5-deficient embryos, although their levels fail to reach the levels that are normal for age-matched wild-type litter-mates (188, 298). Furthermore, the Mlc2v promoter-lacZ transgene that is normally active selectively in the anterior part of the linear and tubular heart is expressed normally in tubular hearts of Nkx2-5-deficient embryos (122). Taken together, in Nkx2-5 mutant mice a primary heart tube is formed that is patterned along the AP axis but fails to further differentiate due to a failure in downstream cardiac gene expression, resulting in arrest during looping and impaired chamber morphogenesis.
Not only is Nkx2-5 important for early cardiogenesis, its expression at the right dose at later stages is also required for septation of the chambers and formation of a functional conduction system. In patients with septal defects and atrioventricular conduction defects, mutations have been found in Nkx2-5 (22, 272). Even though mice with only one intact Nkx2-5 allele do not show defects as severe as those in patients (24), haploinsufficiency is likely to be an important mechanism for the phenotype in patients (147). In line with these findings, detailed analysis of mice deficient for polycomb-group gene Rae28 showed that Nkx2-5 is required for septation of the cardiac chambers (281). Rae28-deficient mice have septal defects, and Nkx2-5 expression is not sufficiently sustained during later development. The cardiac phenotype was rescued by transgenic expression of Nkx2-5 in Rae28-deficient embryos showing that Nkx2-5 expression later in development is crucial for normal cardiogenesis and septation. Why low levels of Nkx2-5 result in defects in specific structures is unclear. Although higher levels of Nkx2-5 mRNA and protein were observed in components of the conduction system (301), Nkx2-5 is expressed broadly in the (developing) heart. Nkx2-5 cooperates with Tbx5 and Tbx2. the latter two interacting through the same sites in the promoter of Anf. Haploinsufficiency for Tbx5 causes septal defects and atrioventricular conduction defects (20, 36, 177), resembling the phenotype of Nkx2-5 mutants. Tbx2 is expressed in the atrioventricular and outflow tract myocardium and cushions, which will form the septa and atrioventricular conduction system, and together with Nkx2-5 represses Anf gene expression in these areas (118). These findings suggest that in vivo the interplay between Nkx2-5, Tbx5, and Tbx2 is important in the correct septation of the chambers and formation of the atrioventricular canal.
Mice with a targeted deletion of Hand2 have severely hypoplastic anterior heart structures and a malformed aortic sac (286). Targeted deletion of Bob, a histone deacetylase-dependent transcriptional repressor, results in a similar although more severe cardiac phenotype (109). In Bop mutants Hand2 expression is abolished specifically in the heart, indicating that the cardiac phenotype of Bob mutant mice is caused by absence of Hand2 in the heart. Rescue experiments by expression of Hand2 in Bob-deficient mice may prove this point. In Hand2 mutant embryos, increased apoptosis was observed in the branchial arches and the anterior region of the tubular heart, indicating that the hypoplastic structures are a result of programmed cell death (302, 341). The regions that are affected correlate well with the expression pattern in the looped heart of E9.5. Hand2 is initially expressed in the entire cardiac crescent and tubular heart but is downregulated in the posterior components (atria, atrioventricular canal, and left ventricle) of the looped heart. Hand2 deficiency does not affect the anterior expression of the Mlc2v-lacZ transgene in the tubular heart, indicating that Hand2 is not required for the AP pattern that regulates this promoter fragment. Endogenous Mlc2v gene expression is also not affected (34, 341). Hand2 deficiency does however interfere with Irx4 expression (34). Whether this downregulation is due to specific interference with Irx4 gene regulation has not been established. In double mutants of Nkx2-5 and Hand2, heart formation is arrested in the tubular stage, and Mlc2v is strongly downregulated (341). However, some cells at the ventral side of the small tubular heart that is formed express Mlc2v. Coup-TFII, Hey1, and Tbx5 expression was observed in the posterior region of this tube. These results suggest that a primary heart tube is formed in compound null mice that is robustly patterned along AP and DV axes, but that maturation and expansion of the chamber myocardium, especially the ventricular chambers, is severely affected. Irx4 and Hand1 expression were undetectable in these mutants. The double mutants seem to have an additive phenotype of the Nkx2-5 or Hand2 mutant phenotypes, indicating that the factors act at least in part independently in the maturation/proliferation of the myocardium of the anterior region.
Homozygous disruption of the myocyte enhancer factor (Mef) 2C gene causes severely affected cardiac morphogenesis (183). The heart tube forms but does not loop and anterior heart structures are malformed and hypoplastic, whereas posterior structures are relatively unaffected. Several cardiac genes, including Anf, are down-regulated, but Mlc2v and Hand1 are expressed normally. This phenotype suggests that a primary heart tube forms and is patterned along the AP and DV axis, but that due to a failure in downstream cardiac gene expression the further differentiation of myocardium is blocked, resulting in arrest during looping and lack of ventricular chamber morphogenesis and expansion. Mef2C is expressed in the entire heart (81). Its ventricular-specific requirement therefore indicates that Mef2C is a necessary cofactor for ventricle-specific factors. Hand2 is downregulated in Mef2C mutant mice at the time of looping. As Hand2 null-mice also have hypoplastic/malformed anterior heart structures and as Mef2 factors cooperates with other basic helix-loop-helix factors in skeletal muscle development, Hand2 has been implicated in cooperating with MEF2C in formation of the anterior region of the heart (226).
X. CONCLUSIONS AND PERSPECTIVES
The descriptions of the developing heart by the early anatomists and physiologists in the beginning of the previous century and recent molecular data have now culminated in a concept that is in harmony with the meticulous observations. In this review, anatomic details have been omitted whenever possible to give emphasis to essentialities. Although many of our ideas are not novel, they have not been presented in an integrated view before.
Both phylogenetic and ontogenetic considerations have led us to conclude that the atrial and ventricular chambers and the cardiac conduction system cannot be considered to be single genetic modules of the chordate heart, but are assembled from smaller component parts. In the formed heart, these components are still recognizable as unique transcriptional domains.
By considering the chambers as local expansions from the tubular heart at ventral (ventricular) and dorsal (atrial) sides (the ballooning model), two long-standing issues have been resolved that remained controversial in the classical segment concept: 1) the formation of the proper connections between the distinct compartments and 2) the formation of the conduction system. Within the ballooning model, the proper connections between the distinct compartments are present from the beginning onward. Furthermore, the formation and positioning of the conduction system is a logical consequence of the regionalized process of chamber formation. The ballooning model also accounts for the gradual developmental transition of the peristaltically contracting tubular heart into the synchronously contracting chambered heart. The tubular heart does not need valves, but the chambered heart requires valves at both ends of the chambers to prevent backflow of the blood. As long as valves have not developed, the original peristaltic type of contraction of the original heart tube could fill this requirement. To this end, it was essential to repress the chamber-specific program of gene expression at the entries and exits of the chambers to guarantee a valvular function. The topographical disposition of the chambers allowed the remaining primary myocardium to largely conserve its original features and to use these features for “novel” functions as component parts of the conduction system. These novel functions are essential for the coordinated contraction of the chambered heart. It is intriguing that atrial and ventricular working myocardium are both activated and connected to one another by “conserved” or embryonic-like myocardium. It is quite ironic that this myocardium has been called “specialized.”
We have committed ourselves to an opinion that we have reduced to the scheme presented in Figure 14, which may provide an important focus for future research. We consider the developmental and evolutionary formation of the chambered heart principally as a two-step process. The first step is the formation of a primary myocardial tube. The second step involves localized differentiation of chambers. This step occurs while recruitment of primary myocardium at both poles of the tubular heart is still ongoing. Patterning of the heart tube along the AP and DV axes provides crucial positional information for the site-specific process of chamber formation in higher vertebrates. Like the site-specific formation of chambers, the localized conservation of the primary myocardium requires positional information, as is evident from the function of Tbx2 in the repression of chamber-specific genes. Retinoic acid was shown to play an important role in the pattern formation along the AP axis of the heart tube, but other signals and pathways that regulate this patterning remain to be determined. We hope that our analysis and attempt to integrate the molecular, morphological, and physiological data into a unifying concept are a help and a stimulus for both teaching and research purposes. It is challenging to examine the nature of the patterning process of the heart tube, which results in the precisely localized formation of the cardiac chambers and of the conduction system. Research on these very basic principles may have an important impact on the understanding of many cardiac malformations and may give inroads into therapeutic strategies.
We are grateful to our colleagues of the Experimental and Molecular Cardiology Group and to Drs Jaques de Bakker, Judith Goodship, and Gertien Smits for valuable suggestions that have much improved the manuscript. We thank Lara Laghetto and Cees Hersbach for the illustrations and Wendy van Noppen for linguistic advice.
The authors are supported by the Netherlands Heart Foundation Grant M96.002.
↵1 The conus arteriosus is considered a component part of the heart because it has a myocardial wall and lies within the pericardial cavity. It is a feature of the evolutionary primitive state. In amphibians it is called the bulbus cordis, a term that is also used for its equivalent in mammalian embryos. The more derived extant bony fish, like the zebrafish, do not have this cardiac compartment. They have a so-called bulbus arteriosus, which is not enclosed by cardiac muscle, but by elastic tissue and smooth muscle, and therefore is considered to be a specialization of the proximal part of the ventral aorta (256). However, similar to the mammalian condition (306, 326, 339), the bulbus arteriosus in zebrafish embryonic hearts is surrounded by myocardium that disappears with development (134, 135). The bony fish bulbus arteriosus might thus be homologous to the shark conus arteriosus and amphibian/mammalian bulbus cordis.
↵2 Reynolds number (Re) is a unitless number representative of the tendency of a liquid (or gas) to become turbulent. It is proportional to the velocity of flow and to the density, and inversely proportional to the viscosity.
- Copyright © 2003 the American Physiological Society