The rate and rhythm of heart muscle contractions are coordinated by the cardiac conduction system (CCS), a generic term for a collection of different specialized muscular tissues within the heart. The CCS components initiate the electrical impulse at the sinoatrial node, propagate it from atria to ventricles via the atrioventricular node and bundle branches, and distribute it to the ventricular muscle mass via the Purkinje fibre network. The CCS thereby controls the rate and rhythm of alternating contractions of the atria and ventricles. CCS function is well conserved across vertebrates from fish to mammals, although particular specialized aspects of CCS function are found only in endotherms (mammals and birds). The development and homeostasis of the CCS involves transcriptional and regulatory networks that act in an embryonic-stage-dependent, tissue-dependent, and dose-dependent manner. This Review describes emerging data from animal studies, stem cell models, and genome-wide association studies that have provided novel insights into the transcriptional networks underlying CCS formation and function. How these insights can be applied to develop disease models and therapies is also discussed.
The cardiac conduction system (CCS) and atrial and ventricular working myocardium are derived from shared precursor cells that diverge during heart formation owing to localized signalling cues.
Differentiation of CCS components is controlled by a network of core cardiac transcription factors and CCS-specific transcription factors; the latter also maintain phenotypic homeostasis of the adult CCS.
CCS-specific transcription factors suppress the working myocardial gene programme, maintain embryonic myocardial properties, and activate a pacemaker gene programme.
The atrioventricular bundle and bundle branches acquire fast-conducting properties during cardiac development, on top of their pacemaker-like properties.
The Purkinje fibre network is derived from the embryonic trabecular chamber myocardium, which acquires fast-conducting properties from the onset of its development.
Insights from developmental biology are being applied to develop novel cardiac disease models and additional translational efforts aimed at regeneration of the CCS.
Control of the rate and rhythm of heart chamber contractions is essential for ensuring an adequate blood circulation in the animal body. The rhythmic contractions result from coordinated depolarizations of electrically coupled heart muscle cells (cardiomyocytes) in the atria and ventricles. These depolarizations are initiated and distributed throughout the heart by components of the cardiac conduction system (CCS). Disturbances of this system can be caused by progressive cardiac conduction diseases (such as Lenegre–Lev syndrome or age-related fibrosis of the CCS), tissue destruction (such as valve replacement surgery), congenital heart block (such as that caused by maternal autoantibodies) or congenital malformations (such as cardiac isomerism) that result in heart rate slowing, or brady-arrhythmias1,2,3,4. Moreover, ectopic pacemaker and conduction pathways can be formed during embryogenesis as a result of errors in the developmental patterning of CCS components, as seen in patients with ventricular pre-excitation (Wolff–Parkinson–White syndrome)4,5. Generally, pharmacological treatment alone is not sufficient to control heart rhythm in patients with a disturbed CCS. Therefore, management of cardiac function in patients with CCS abnormalities relies on an implantable electronic pacemaker, which is the only available long-term treatment so far. However, use of such devices has disadvantages, including an increased risk of inflammation, insensitivity to autonomic modulation, limited battery capacity, and (in children) the inability to adapt to the growing heart. The generation of biological CCS components could overcome these problems and yield less-invasive treatments6,7. Especially during the past decade, several research groups have attempted (with variable success) to generate pacemaker cardiomyocytes by either differentiation of pluripotent stem cells or reprogramming of working cardiomyocytes8,9. However, further insights into the molecular mechanisms that control the differentiation and homeostasis of CCS components are required to realize the potential of these approaches.
In this Review, we provide a brief outline of the developmental origin of the various components of the CCS and describe the transcriptional networks underlying their development and homeostasis. We highlight some of the pathophysiological consequences of errors in these processes. Finally, we discuss the results of studies in which insights from CCS development have been used to generate pacemaker cells.
The cardiac conduction system
The functional CCS components can be broadly divided into the slow-conducting nodes, which include the sinoatrial node (SAN) and atrioventricular node (AVN), and the fast-conducting ventricular conduction system (VCS), which includes the atrioventricular bundle, the right and left bundle branches (BBs), and the Purkinje fibre network (PFN).
The SAN is the dominant pacemaker of the heart and controls its rate of contraction. The SAN is located at the junction of the right atrium and the superior vena cava, near the crista terminalis, where blood from the systemic circulation enters the right atrium10 (Fig. 1). The SAN comprises a small and heterogeneous population of cells, including pacemaker cardiomyocytes, which can depolarize spontaneously. These pacemaker cells are innervated by the autonomic nervous system, which regulates the heart rate and enables it to adapt to changes in cardiac output and metabolic demand11. From the SAN, action potentials are rapidly propagated through the atria to the AVN, which is positioned caudally in the heart at the junction of the atria and ventricles. The slow conduction velocity of impulses in the AVN enables the ventricles to fill with blood before the impulse is propagated to the ventricles. The impulse then rapidly propagates through the atrioventricular bundle (bundle of His) and its branches in the ventricular septum. Under normal conditions, this pathway is the only electrical connection through the connective tissues between the atria and ventricles (annulus fibrosus and central fibrous body). Once the impulse reaches the PFN, it is rapidly distributed to the ventricular cardiomyocytes, causing them to contract. The propagation of action potentials through the heart compartments can be visualized in an electrocardiogram (ECG): the P wave indicates atrial activation, the PR interval is the time taken for the impulse to reach the ventricles, the QRS complex indicates activation and depolarization of the ventricles, and the T-wave shows repolarization of the ventricles (Fig. 2). Thus, the function of the SAN, AVN, atrioventricular bundle, BBs, and PFN, which together comprise the CCS, can be observed on an ECG.
Cardiac conduction system development
The atrioventricular canal
The atria and ventricles differentiate from the embryonic myocardium of the primitive heart tube when it starts to loop. Contraction of these chambers is activated sequentially from the very beginning of heart development. This coordination is possible because the fast-conducting myocardium of the atrial and ventricular chambers differentiates within a slow-conducting primitive heart tube, whereas the area in between these regions (the atrioventricular canal or ‘heart of the heart’) maintains its slow-conducting phenotype. The dominant pacemaker is always present at the inflow tract12. This arrangement generates a unidirectional contraction pattern, including an appropriate atrioventricular delay, that resembles the activation patterns of the mature heart13. Indeed, distinct P-wave and QRS complexes can be identified in the electrical activity of the embryonic avian and mammalian heart soon after the onset of chamber differentiation12.
The atrioventricular canal myocardium originates from the limbs of the cardiac crescent in the embryonic day (E) 7.5 mouse embryo and from the inflow tract of the stage 9 chicken embryo14,15,16,17 (Fig. 1). The atrioventricular canal, the first component to form during heart development, is largely derived from the first heart field and is highly conserved among vertebrates. Ectothermic vertebrates (fish, amphibians, and reptiles) maintain an embryonic-like atrioventricular canal throughout adulthood18. In mammals and birds, by contrast, the atrioventricular canal is fated to form the AVN, atrioventricular ring bundles, and the base of the left ventricle15,19 (Fig. 1). The atrioventricular canal lies at the centre of the developing heart and is the structure around which the atria, sinus venosus, ventricles, and outflow tract subsequently develop. The atrioventricular canal remains connected to these structures at their inner curvatures, and tissues derived from the atrioventricular canal maintain their pacemaker-like phenotype (that is, spontaneous depolarization and slow conduction) into adulthood. The AVN becomes the principal unidirectional electrical conduit between the atria and ventricles: this node defines the delay in ventricular activation and can act as a secondary pacemaker if the SAN fails. In 2017, resident macrophages in the AVN were shown to be electrically coupled to myocardial cells in the AVN and to facilitate electrical conduction in a manner dependent on expression of Gja1 (the gene encoding gap junction α1 protein; also known as connexin 43)20.
Our insights into the regulation of CCS development are mostly derived from transgenic mouse models. The embryonic atrioventricular canal expresses Bmp2 (the gene encoding bone morphogenetic protein (BMP) 2). BMP2 activates the expression of Tbx2 and Tbx3 genes, which encode the T-box transcription factors TBX2 and TBX3, respectively. These transcription factors actively repress the chamber myocardial gene programme21,22,23,24 (Fig. 3). BMP2 is involved in formation of the cardiac jelly and atrioventricular cushions within the atrioventricular canal, which prevent the direct contact of the atrioventricular canal with factors secreted by the endocardium25. Furthermore, Bmp2 gene expression is associated with both endocardial and epicardial epithelial-to-mesenchymal transition (EMT)21,26. EMT is required for formation of the annulus fibrosus, the structure that provides a conduction barrier between the atria and ventricles. Atrioventricular canal-specific deletion of Bmpr1a (which encodes BMP receptor type 1A) leads to impaired development of the AVN and annulus fibrosus27,28.
The function and homeostasis of the atrioventricular conduction system are highly sensitive to changes in the levels of Tbx2 and Tbx3 expression29. TBX2 and TBX3 have redundant functions in the atrioventricular canal, as both transcription factors act to repress the chamber myocardial gene expression programme (which involves expression of Gja1, Gja5, Nppa, and Scn5a) and to stimulate the pacemaker and neuronal gene programme23,30,31. Deletion of both Tbx2 and Tbx3 results in increased expression of chamber myocardial genes, loss of atrioventricular cushion development, and reduced BMP2 levels in the atrioventricular canal, which indicate the loss of atrioventricular canal specification23. Conversely, overexpression of Tbx2 or Tbx3 induces pacemaker gene expression, blocks chamber gene expression, and causes ectopic EMT of the adjacent endocardium23. These factors interact with homeobox protein MSX2 in the atrioventricular canal to suppress Gja1 expression32. Myocardium-specific loss of function of Tbx2 results in ectopic (chamber-myocardium-like) connections between the atrium and ventricle that conduct the impulse rapidly through the annulus fibrosus, causing ventricular pre-excitation reminiscent of that observed in Wolff–Parkinson–White syndrome24 (Fig. 3).
The core cardiac transcription factors (homeobox protein NKX2.5, TBX5, GATA4, and GATA6) interact with and co-regulate each other33 and also interact with transcription factors that have more localized roles, such as TBX2 and TBX3. The core transcription factor network is involved in the development of most components of the heart, including the atrioventricular canal. Heterozygous loss-of-function mutations in the Nkx2-5 gene cause atrioventricular block in mice and AVN disease in humans (Table 1). Heterozygous loss of Tbx5 results in maintenance of an extensive fetal-like atrioventricular conduction system (atrioventricular canal phenotype) in adult MinK–lacZ mice34, suggesting that a finely tuned balance between differentiation-suppressive TBX3 and chamber-programme-activating TBX5 activity controls the homeostasis of the atrioventricular conduction system. Gata4+/− mice have accelerated atrioventricular conduction and reduced expression of Gjd3 (encoding gap junction δ3 protein; also known as connexin 30.2) compared with Gata4+/+ mice35. Homozygous deletion of Msc (encoding musculin, a transcriptional repressor that interacts with GATA4) results in prolonged atrioventricular conduction and increased expression of Gjd3 (ref.36). Gata6−/− mice show prolonged atrioventricular conduction and a hypoplastic AVN37 (Table 1).
Tbx20 is expressed in the entire heart, and its product, TBX20, normally suppresses Tbx2 expression in the chambers (possibly through interfering with BMPs and SMAD-mediated activation of Tbx2 expression), which helps to confine TBX2-dependent atrioventricular canal formation38. The atrioventricular canal boundary is additionally delimited by Notch signalling. Studies in chick hearts revealed that expression of HEY1 and HEY2 (genes encoding the transcriptional repressors hairy/enhancer-of-split related with YRPW motif proteins 1 and 2; HEY1 and HEY2, respectively) is activated by Notch signalling and inhibits BMP2 and TBX2 expression39,40. Induction of TBX2 expression by BMP2 decreases the levels of both HEY1 and HEY2, which suggests the existence of a feedback loop39. Ectopic Notch activation in the developing myocardium leads to atrioventricular canal mispatterning, formation of accessory conduction pathways, and ventricular pre-excitation. Conversely, inhibition of Notch signalling leads to a hypoplastic AVN41. The mechanism downstream of Notch has not been elucidated, but could involve dysregulation of Hey1 and Hey2 expression. Atrioventricular-canal-specific gene expression is also achieved through GATA4-dependent regulatory switches42. GATA4 and GATA6 act in concert with SMAD proteins and nuclear BMP signalling effectors to recruit histone acetyltransferases, thereby driving the atrioventricular canal gene programme. Conversely, in the chambers, GATA4 and GATA6 recruit HEY1 or HEY2 and histone deacetylases (HDACs), resulting in histone deacetylation and repression of the atrioventricular canal gene programme (Fig. 3).
Moreover, Wnt signalling regulates the expression of bmp4 and tbx2b in the atrioventricular canal of zebrafish43. In mice, canonical WNT signalling is active during development of the atrioventricular canal, and loss of function of WNT proteins leads to progressive loss of the atrioventricular canal phenotype during late gestation, which by the time of birth results in the near absence of Tbx3+ and Gja5− atrioventricular canal myocardium. Conversely, ectopic activation of WNT signalling induced the development of a structure that morphologically and phenotypically resembled the atrioventricular canal in the ventricles, in which Tbx3 expression was induced and expression of Gja1, Gja5, and Scn5a was repressed44.
The sinoatrial node
Spontaneous depolarization is a property of the embryonic myocardium. The first spontaneous calcium oscillations can already be observed in the crescent-shaped heart tube of the mouse embryo at around E7.75, rapidly followed by nascent contractions at E8 (ref.45). The first spontaneous action potentials can be observed in the caudal part of the early heart tube in the seven-somite stage chick embryo, and the first contractions occur at the nine-somite stage46. However, at this stage, the future SAN cells that will become the dominant pacemaker reside in the lateral plate mesoderm caudal to the heart-forming region and have not yet been incorporated into the heart tube47,48 (Figs 1,4a). As the heart tube elongates owing to the addition of posterior second heart field (SHF) progenitors, dominant pacemaker activity moves towards the caudal pole. Likewise, expression of Hcn4 (encoding potassium/sodium hyperpolarization-activated cyclic nucleotide-gated channel 4, which is required for pacemaker activity49 and expressed in the early heart tube) is activated in caudal cardiac progenitors upon their differentiation and downregulated in the ‘older’ part of the heart tube, thus effectively shifting its expression in the caudal direction50. In mice, cardiomyocytes and their progenitors express Nkx2-5 until E9.0–E9.5, at which point the future atrial compartment has been established51. However, from then until E12, the recruited progenitor cells express Tbx18 instead of Nkx2-5 and develop into the Nkx2-5lowTbx18+ myocardium, which includes the sinus venosus and the definitive SAN52 (Fig. 4). Similarly, Hcn4 expression is downregulated in cardiomyocytes that will form the atria, and becomes confined to the Nkx2-5lowTbx18+ sinus venosus and eventually the SAN53,54. Pacemaker activity shifts into the sinus venosus upon its formation, and from there to the anatomical SAN, during early fetal stages. The remainder of the sinus venosus myocardium starts to express the working myocardial genes Gja1, Gja5, and Nkx2-5, a process also referred to as atrialization of the sinus venosus53.
TBX18 does not seem to be required for regulation of the sinus venosus or SAN gene programme54. However, the sinus venosus is formed from Tbx18+ precursors in the dorsal pericardial wall. The cells around the endothelial veins differentiate into Nkx2-5low myocardium while being released into the pericardial cavity54. In Tbx18−/− mice, the pericardial cells fail to form sinus horns. Instead, the endothelial veins remain positioned within the pericardial wall, and the heart remains attached to the pericardial wall at the level of the atria. As a consequence, the veins and part of the SAN remain hypoplastic. The Tbx18−/− precursor cells that stay in the pericardial membrane differentiate much later into the sinus venosus and the SAN myocardium.
The SAN anlage forms in the right sinus horn, adjacent to the atrial chamber myocardium, and can be recognized immediately by the expression of Tbx3. This restriction of SAN development to the right sinus horn is ensured by Pitx2 (encoding pituitary homeobox 2 protein (PITX2)), which is expressed only in the left sinus horn where it suppresses right-sided sinus venosus morphogenesis and gene expression55. The mechanism underlying how Pitx2 expression mediates right-sided SAN morphogenesis is not entirely clear. Direct suppression of Shox2 (encoding short stature homeobox protein 2 (SHOX2)) and Tbx3 expression by PITX2 has been indicated56,57, but given that expression of Shox2 is bilateral in normal sinus horns and that Pitx2−/− mice have complete morphological right isomerism of the left side of the sinus venosus (including the formation of extra venous valves)55, PITX2 probably acts upstream of Shox2 expression and perhaps also through different mechanisms (Fig. 4).
The phenotypic boundary between the SAN and atrial myocardium is maintained by NKX2.5 (ref.55). Indeed, the absence of Nkx2-5 expression is required for the sinus venosus to maintain its pacemaker gene programme53,55,58,59,60. TBX5, a transcriptional activator expressed in the sinus venosus and its precursors, regulates expression of the pacemaker genes Bmp4, Shox2, and Tbx3 (refs61,62). Shox2 expression is crucial for SAN development and is restricted to the sinus venosus, where it prevents the formation of working myocardium by inhibiting Nkx2-5 expression and activating Hcn4 Isl1, and Tbx3 expression58,59,63 (Fig. 4).
Insulin gene enhancer protein ISL1 (encoded by Isl1) is a LIM-homeobox transcription factor that is also associated with SAN development64,65,66. Initially, Isl1 is expressed by cardiac progenitor cells (as ISL1 mediates SHF development), but thereafter Isl1 expression becomes restricted to the SAN where it is maintained until adulthood66,67,68. ISL1 acts upstream in the SAN signalling cascade to regulate Bmp4, Hcn4, Shox2, and Tbx3 expression66,69. Isl1 itself is a direct target of SHOX2, which implies the existence of a positive feedback loop between ISL1 and SHOX2 (ref.64).
Tbx3 expression is specifically maintained in the SAN, where it is required (both before and after birth) to suppress the expression of genes in the working myocardial programme, including Gja1, Gja5, and Scn5a29,70. Moreover, ectopic Tbx3 expression in prenatal and postnatal atrial tissue resulted in acute downregulation of genes related to action potential conduction and calcium handling, activation of expression of Hcn4 and other SAN markers, and the formation of ectopic functional pacemaker cells31,70. Activation of Tbx3 expression requires AT-rich interactive domain-containing protein 1A (ARID1A, also known as Baf250a) and TBX3, which, together with HDAC3, then suppress Nkx2-5 expression in the SAN71.
WNT signalling is not only required for atrioventricular canal formation and myocardialization of the sinus horns72, as discussed above, but has also been implicated in SAN development. WNT signalling might act to inhibit Nkx2-5 expression, which would contribute to activation of the SAN gene programme47. However, mice with Tbx18+-cell-specific deletion of Ctnnb1 (which encodes catenin β1, a crucial downstream component of the canonical WNT signalling pathway) do not exhibit any changes in SAN size or morphology, which suggests that WNT signalling is not directly involved in the formation of the SAN myocardium72. The role of WNT signalling in SAN development remains elusive.
The ventricular conduction system
The atrioventricular bundle and proximal bundle branches
The atrioventricular bundle is conserved across endothermic vertebrates (mammals and birds) but is not found in ectothermic vertebrates (fish, amphibians, and reptiles) — except for crocodilians, the only reptiles that have a full ventricular septum73. The atrioventricular bundle is derived from the ventricular septal part of the G1N2+Tbx3+ primary interventricular ring, and the BBs are derived from the G1N2+Tbx3+ subendocardial trabecules, which develop into the ventricular septum and are connected to the interventricular ring74,75. The Tbx3+ ring has been observed in mice as early as E9, just after the onset of chamber differentiation75. Genetic lineage-tracing experiments have indicated that the atrioventricular bundle and right BB are derived from cardiomyocyte progenitors that express SHF markers, such as the anterior SHF (AHF)-specific regulatory DNA sequence of Mef2c (termed Mef2c–AHF-enhancer )76,77. However, the domains derived from the first heart field (Tbx5+ lineage) and the SHF (Mef2c–AHF-enhancer+ lineage) overlap in the interventricular septum at the level of the atrioventricular bundle78. A sharp lineage boundary has been detected between the Mef2c–AHF-enhancer+Tbx2− atrioventricular bundle and the anatomically and functionally connected AVN, which is derived from Mef2c–AHF-enhancer−Tbx2+ myocytes from the atrioventricular canal77. Initially, the developing atrioventricular bundle and proximal BBs have a nodal (both SAN and AVN)-like gene expression profile associated with slow conduction (namely, Tbx3+Gja1−) until expression of Gja5 is activated during the fetal period34,79,80,81. Nonetheless, Scn5a is highly expressed in the atrioventricular bundle from its early development onwards, suggesting that this structure enters the fast-conduction gene programme earlier than is indicated by the induction of Gja5 expression82. Hcn4 expression becomes detectable during development of the VCS, including in the atrioventricular bundle and BBs.
An active transcriptional network composed of several tissue-specific transcription factors — NKX2.5, TBX3, TBX5, DNA-binding protein inhibitor ID2 (encoded by Id2), ETS translocation variant 1 (ETV1, encoded by Etv1), homeodomain-only protein (HOPX, encoded by Hopx), transcription factor SP4 (encoded by Sp4), and Iroquois-class homeodomain protein IRX3 (encoded by Irx3) — specifies and controls the gene programme and phenotype of cells forming the atrioventricular bundle and proximal branches from the onset of their development31,34,81,83,84,85,86,87,88 (Fig. 5). Data mainly derived from transgenic mouse models have uncovered important aspects of the function of nodes in this network and transcriptional output, which underlies the atrioventricular bundle and proximal BB phenotype. TBX5 is required for the development and homeostasis of the atrioventricular bundle and BBs81,83. This transcription factor directly activates expression of Gja5 and Scn5a in the atrioventricular bundle, which accounts for the high expression of these two genes and the high conductivity of the atrioventricular bundle. Nkx2-5+/− mice develop atrioventricular bundle hypoplasia and atrioventricular block, which is largely rescued by Prox1 haploinsufficiency, possibly through alleviating transcriptional repression of Nkx2-5 mediated by HDAC3 and/or prospero homeobox protein 1 (PROX1)89. In Tbx5+/−Nkx2-5+/− mice, cells of the prospective atrioventricular bundle fail to exit from the cell cycle and do not activate expression of Id2. ID2, in turn, is also required for normal atrioventricular bundle and BB development83.
TBX3 is essential for the appropriate development of the atrioventricular bundle and BBs. In the early atrioventricular bundle anlage, TBX3 suppresses transcription of Gja1, Gja5, and other genes related to fast conduction, controls the reduction in cell proliferation rate, and stimulates pacemaker-like properties (Fig. 5). Interestingly, although Gja1 expression remains suppressed in the atrioventricular bundle and BBs, expression of Gja5 (a target of both TBX3 and TBX5 (refs70,90)) is initiated in the fetal atrioventricular bundle, and Scn5a (another direct target of both TBX3 and TBX5 (ref.91)) is highly expressed in both the atrioventricular bundle and BBs from the outset of their development. These observations can be explained by assuming that the transcriptional activator TBX5 and the transcriptional repressor TBX3 compete both for interaction with Nkx2-5 and for binding to the regulatory elements of their shared target genes91,92. The balance between TBX5 and TBX3 activity might shift towards TBX5 during early fetal development, which could explain the induction of Gja5 expression. Alternatively, Gja1 could be more sensitive to TBX3 than to TBX5 and Scn5a more sensitive to TBX5 than to TBX3. Sensitivity of target gene expression to Tbx5 gene dose has been observed previously61. Other important genes activated by TBX5 are Ryr2 and Atp2a2, which encode the calcium-handling proteins ryanodine receptor 2 and sarcoplasmic/endoplasmic reticulum calcium ATPase 2, respectively93.
Genome-wide association studies of human genes associated with ECG parameters identified variants in SCN5A, SCN10A, TBX3, and TBX5 associated with PR interval and QRS interval durations, strongly implicating these genes in CCS function (Table 1). Functional analyses revealed that TBX5 activates and TBX3 suppresses the expression of SCN5A in the VCS and atrioventricular canal via transcriptional enhancers in the SCN10A–SCN5A locus (including one within SCN10A) that interact with the SCN5A promoter. A common variant in its T-box binding site inactivates this intronic SCN10A enhancer91,94. Tbx3 and Tbx5 are paralogous genes that form an evolutionarily conserved gene cluster. However, despite the dose-sensitive functional balance and interplay between TBX3 and TBX5, their encoding genes are each regulated by their own physically separated set of regulatory sequences95, indicating they are affected independently by genetic variation.
The Purkinje fibre network
The postnatal atrioventricular bundle, BBs and PFN cardiomyocytes (which collectively comprise the VCS) express genes for fast conduction (Gja5 and Scn5a), pacemaker properties (Hcn4), neuronal neurofilament proteins (Nefm and Nefl), and the axonal glycoprotein contactin 2 (Cntn2)34,81,96,97,98. Hcn4 is gradually turned on in the VCS during fetal development, whereas Cntn2 expression is activated around birth, indicating that the VCS undergoes further specialization and maturation in the perinatal period97,99. The PFN is derived from cells of the early ventricular chamber myocardium that initiate the expression of Bmp10, Gja5, Irx3 and Nppa as soon as they differentiate from the primary heart tube (at around E9 in the mouse)85,100,101,102,103,104. At this stage, the embryonic ventricular wall is only one to a few cell layers thick. The trabecular ventricular wall grows and maintains transmural expression of Bmp10, Gja5, Irx3, and Nppa, and other genes (Fig. 5). Soon after E11.5, the subepicardial compact wall zone starts to form and subsequently expands rapidly. This compact zone loses the expression of Gja5, Irx3, and Nppa, whereas the trabecular layer of the ventricular wall remains Gja5+Irx3+Nppa+ (Fig. 5). Irreversible genetic labelling of Gja5+ cells during mouse development has revealed that definitive establishment of the VCS occurs at mid to late fetal stages: Gja5+ cells irreversibly labelled before E14.5 contribute to both compact wall myocardium and the VCS, confirming the common ancestry of these cardiomyocyte subtypes, whereas Gja5+ cells labelled after E16.5 are exclusively found in the VCS103. Whereas in mammals and birds the trabecular layer develops into a specialized PFN, ectothermic vertebrates, including crocodiles, maintain the trabecular layer into adult stages, which is responsible for activation of the ventricle or ventricles73.
Notch signalling and neuregulin signalling are two principal systems that control ventricular trabecular development105,106 (Fig. 5). In chick embryos (but not in mouse embryos107), endothelin signalling also influences VCS development108. Endocardial Notch signalling stimulates expression of Bmp10 in trabecular myocardium, which in turn suppresses the expression of genes that encode inhibitors of the cell cycle104. Ectopic induction of canonical Notch signalling (through overexpression of Notch 1 intracellular domain) in the ventricles induces a Purkinje fibre-like phenotype in ventricular myocytes109, associated with expression of Gja5 and Scn5a. These myocytes also express CCS–lacZ (a transgenic reporter that can identify embryonic trabecules and mature VCS110) and Cntn2–EGFP (a transgenic reporter that identifies the VCS97). Notch signalling stimulates endocardial expression of Nrg1 (which encodes neuregulin 1 (NRG1)). NRG1 and its myocardial receptor tyrosine-protein kinases ERBB2 and ERBB4, in turn, control a gene regulatory network that controls graded differentiation and cell-cycle activity across the ventricular wall111. NRG1 is sufficient to induce and upregulate the expression of various genes involved in fast conduction88,112 (namely, Gja5, Irx3, Nkx2-5, and Scn5a) and the transgenic reporter CCS–lacZ, via RAS–mitogen-activated protein kinase (MAPK) signalling and ETV1. Etv1 is expressed in atrial myocardium, ventricular trabecular myocardium, and the atrioventricular bundle and BBs from the onset of their differentiation onwards (D. Park, personal communication). Expression of Etv1 subsequently becomes confined to the VCS in a pattern that resembles (both spatially and temporally) that of Gja5, a target gene of ETV1 (ref.88). Etv1−/− mice have a hypoplastic VCS (which can be identified by the Cntn2–EGFP transgenic reporter) that demonstrates reduced expression of Gja5, Nkx2-5, and Scn5a. VCS hypoplasia in these mice might involve the reduction of Nkx2-5 expression. Nkx2-5+/− mice also have impaired postnatal maturation of the VCS, and their Purkinje fibres are hypoplastic despite normal prenatal trabeculation113. Etv1−/− mice also show increased expression of Tbx5 in the VCS, which is noteworthy because Scn5a and Gja5 are downregulated in these mice despite being well-established target genes of TBX5 (ref.81). Etv1−/− mice also exhibit lengthening of the P wave, PR interval, and QRS duration88. A single nucleotide polymorphism in the ETV1 locus has been associated with conduction abnormalities (BB block and heart block) in humans88.
In mice, expression of Irx3 is first activated specifically in the differentiating ventricular myocardium at E9, after which it becomes confined to the trabecular myocardium and finally to the VCS85,114. Postnatally, expression of Irx3 is required for VCS maturation, as IRX3 induces expression of Gja5 and suppresses Gja1 expression in the VCS components85. Irx3−/− mice display ventricular conduction slowing, QRS prolongation, and ventricular tachyarrhythmias85. Moreover, polymorphisms in IRX3 have been associated with ventricular fibrillation in humans115.
A question that remains to be settled is how the Gja5+Nppa+ embryonic trabecular zone becomes confined to the mature subendocardial PFN. The trabecular zone is maintained throughout development because it proliferates very slowly. Expression of Irx3, Irx5, and Tbx5 shows a graded pattern (that is, expression is higher in the trabecular region on the endocardial side). The transcription factors encoded by these genes are involved in the regulation of Gja5, Scn5a, and other transmurally patterned genes related to conduction and repolarization81,85,115,116,117. In mouse embryos from approximately E12 onwards, the compact ventricular wall expands massively in response to signals from the epicardium118 and loses expression of Gja5 and Nppa. The expanding compact myocardial layer specifically expresses Nmyc (a cell proliferation marker) and Hey2. HEY2 is a transcriptional repressor that suppresses Gja5, Nppa, and Scn5a expression in the compact zone and (possibly indirectly) activates expression of Kcnd2 and Kcnip2 (which both encode proteins involved in the transient outward current) in the subepicardial domain119,120,121,122. Postnatally, the trabecular zone remodels to become the mature PFN, which is only a few cells thick. This remodelling process requires IRX3, ETV1, and normal levels of NKX2.5, but the underlying mechanism is not yet fully understood85,88,113,123.
Insights into the mechanisms underlying CCS development have been applied to generate CCS-like cells from pluripotent stem cells or differentiated somatic cells. These cells are being made for three purposes: to gain further insight into the mechanisms of CCS differentiation and the function of human CCS cells; to examine the efficacy, specificity, and adverse effects of antiarrhythmic compounds; and to generate biological pacemakers as an adjunct or an alternative to electronic pacemakers. This body of work has been extensively reviewed elsewhere6,7,124,125, and only a few key studies are briefly discussed here.
Differentiation of embryonic stem cells (ESCs) or induced pluripotent stem cells (iPSCs) to cardiomyocytes typically results in a heterogeneous mixture of spontaneously contracting nodal-like, atrial-like, and ventricular-like cells126. Interestingly, some features of genuine SAN cells (such as beating rate variability, which demonstrates fractal behaviour) have been reported in human ESC-derived and iPSC-derived cardiomyocytes127. Indeed, their complex phenotype resembles an important property of SAN cells128. Substantial efforts have been made to guide the differentiation process towards a pacemaker cell population. To this end, several approaches based on transcription factor manipulation have been explored70. Overexpression of Tbx3 in mouse ESCs can increase the percentage of nodal-like cells after cardiac differentiation. The removal of non-myocyte cells (using MYH6-promoter-based antibiotic selection) led to further enrichment of pacemaker cells, resulting in embryoid bodies comprising ~80% nodal cells129,130. Other groups have used a variety of different antibiotic selection cassettes under the control of various putative nodal-specific promoters (including a SHOX2 promoter fragment or a minimal GJD3 enhancer) to select nodal cells from mixed populations58,131. In a similar approach, Shox2 overexpression in differentiating mouse ESCs efficiently increased beating rates and the percentage of spontaneously active embryoid bodies compared with wild-type controls. The enrichment of pacemaker activity achieved by Shox2 overexpression seemed, at least in part, to result from temporal changes in WNT signalling132. In line with these findings, a 2013 study that compared Shox2+/+ and Shox2−/− ESC clones showed that differentiating precursors that would give rise to SAN-like cells could be isolated on the basis of their CD166 expression133. Subsequent studies of these SAN-like cells demonstrated decreased expression of Cacna1h, Gjc1, Hcn4, and Tbx18 in the Shox2−/− cells, further underlining the crucial role of SHOX2 in activating the pacemaker gene programme134.
Another approach to facilitate the differentiation of human pluripotent stem cells towards CCS phenotypes has employed multiple small molecules and ligands to (transiently) activate transcriptional pathways. An important proof of concept for this approach came from a study that used human ESCs with doxycycline-inducible MYC expression in combination with growth factors and small molecules to aid the clonal expansion of pre-NKX2-5+ multipotent cardiovascular progenitor cells135. Inhibition of BMP and fibroblast growth factor (FGF) signalling in these cells resulted in an NKX2-5low cell population that expressed high levels of podoplanin, a posterior SHF marker. Upon differentiation to cardiomyocytes, these cells displayed molecular and functional hallmarks of SAN cells. Another study also showed efficient differentiation of human pluripotent stem cells towards the SAN phenotype8. Their approach relied on a two-stage protocol, in which mesoderm induction was guided by low concentrations of BMP4 and activin A, followed by treatment with BMP4 and retinoic acid, then inhibition of FGF, transforming growth factor-β and WNT signalling to induce a SAN phenotype. These myocytes consistently expressed ISL1, SHOX2, and TBX3, but not MYL2 or NKX2-5 (ref.8). Enrichment for SAN-like cells relied on surface marker selection to identify SIRPA+ (CD172a-expressing) cardiomyocytes136 and exclude THY1+ (CD90-expressing) non-myocytes. The researchers also demonstrated the potential applicability of these cells to act as a biological pacemaker in a rat model of complete heart block8. Furthermore, sodium nitroprusside (identified by small-molecule screening in mouse ESCs expressing the CCS–lacZ and Cntn2–EGFP reporter transgenes) promoted ESC differentiation to a Purkinje-like phenotype by activating cyclic AMP signalling137.
Overexpression of Tbx3 was among the first approaches used to test the possibility of direct conversion of cardiac myocytes to pacemaker cells. Tbx3 overexpression, either in mature cardiomyocytes in vivo or in cultured neonatal cardiomyocytes, resulted in effective downregulation of the working myocardial gene programme. Nevertheless, pacemaker genes were not induced, and isolated single cells did not show increased automaticity31. These results contrast with the potent effect of Tbx3 overexpression during embryonic development, which can effectively reprogramme atrial cells to develop into bona fide pacemaker cells31,70. These disparate findings indicate either a need for cofactors that are absent in mature cardiomyocytes or the inability of Tbx3 overexpression to epigenetically reprogramme mature working cells. In an effort to optimize the efficiency of direct reprogramming, researchers who screened several transcription factors in neonatal rat ventricular myocytes found TBX18 to be the most promising in terms of increasing spontaneous activity9. Interestingly, although TBX3 and SHOX2 were also screened, both failed to increase automaticity. Follow-up experiments demonstrated that overexpression of Tbx18 in guinea pig myocardium enforced a pacemaker phenotype in a subset of the transduced cells9. In a clinically relevant pig model of complete heart block, Tbx18 overexpression resulted in effective biological pacing that almost completely eliminated the need for back-up electronic pacing138. From a developmental biology perspective, these potent effects of Tbx18 overexpression in biological pacemaker engineering are somewhat surprising because TBX18 seems to be redundant for development of a functional SAN54. Moreover, in a 2016 study, Tbx18 overexpression in mouse atrial and ventricular cardiomyocytes caused partial suppression of the working myocardium gene programme but failed to drive upregulation of pacemaker genes or to induce ectopic pacemaker activity139. One possible explanation is that under particular circumstances, TBX18 can mimic the repressive function of TBX3 (ref.140).
Thus far, gene therapies based on overexpression of Tbx18 or genes encoding ion channels and/or components of the β-adrenergic signalling cascade138,141,142 have been more effective than biological pacing approaches based on transplantation of ESC-derived or iPSC-derived cardiomyocytes. However, the transplanted ESC-derived or iPSC-derived cardiomyocytes used in these large-animal studies were not enriched for the SAN phenotype143,144. In addition, the longevity of the above-mentioned gene therapies still needs further optimization and validation, whereas human iPSC-derived cardiomyocytes have demonstrated stable biological pacing throughout a 3-month study. Therefore, the real comparisons of these different strategies will come from additional long-term, large-animal studies that are currently ongoing.
Careful morphological and developmental cell lineage-tracing studies, along with the identification of key transcriptional networks, have provided a concept for the development of the CCS145,146,147,148,149,150,151,152,153,154,155 (Fig. 1). In mouse embryos, the primitive heart tube forms between E7.5 and E8.5 from the so-called first heart field precursors, which eventually develop into the left ventricle, including (part of) the atrioventricular bundle, left BB and left ventricular PFN, and the atrioventricular canal, including the AVN. Subsequently, between about E8.5 and E10, the so-called SHF progenitors are added to the heart tube, which will eventually form the atria, right ventricle, part of the atrioventricular bundle, right BB and right ventricular PFN, and the outflow tract. Finally, at between E9.5 and approximately E12, the sinus venosus (including the SAN) is added to the heart tube from distinct SHF progenitors. A network of cardiac core transcription factors and CCS-restricted transcription factors suppresses both the working myocardial gene programme and cell proliferation of the embryonic CCS components, while stimulating a different gene programme that provides pacemaker-like properties. As a consequence, the CCS components remain physically small compared with the rapidly growing atria and ventricles, and the CCS cells maintain some aspects of the embryonic myocardial phenotype (a glycogen-rich cytoplasm with few mitochondria and poorly developed myofibrils, sarcoplasmic reticulum, and T-tubules145,146) and increasingly diverge phenotypically from the surrounding working myocardium. A spatially restricted transcriptional network induces a gene programme that provides fast conductive properties to the atrioventricular bundle, BBs, and PFN. Subsequently, the CCS components terminally differentiate and become confined to discrete regions, a process that continues for a few days after birth. The factors identified by studying CCS development have been deployed to direct stem cell differentiation towards CCS cell fates or to impose pacemaker-like properties on cardiac or non-cardiac cells. These approaches are useful not only to study the mechanisms of CCS development, but also to generate models for studying CCS diseases and for drug testing. Eventually, these models could become key assets in regenerative medicine.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
V.M.C. is supported by Fondation Leducq grant 14CVD01, Netherlands Heart Foundation grant COBRA3, and CVON grant ConcorGenes. G.J.J.B. is supported by personal grants from the Netherlands Foundation for Scientific Research (ZonMW Veni 016.156.162), the Dutch Heart Foundation (Dr Dekker grant no. 2014T065), and the European Research Council (ERC; Starting Grant no. 714866).
Despite great inconsistency in the literature, in this article we have used the currently accepted nomenclature: thus, NKX2.5 is the protein encoded by the Nkx2-5 gene, with capitalization indicating human (all caps) or mouse (first letter capitalized only) gene names.
- MinK–lacZ mice
The MinK gene (now named Kcne1) encodes potassium voltage-gated channel subfamily E member 1, also known as MINK. MINK-deficient mice in which the bacterial lacZ gene has been substituted for the MINK coding region express β-galactosidase under the control of Kcne1 regulatory elements. Thus, β-galactosidase staining in postnatal Kcne1−/− hearts is highly restricted to the sinoatrial node, caudal atrial septum, and proximal cardiac conducting system.
The target of a monoclonal antibody (and also the name of the monoclonal antibody itself) initially identified by binding to an extract from the chick ganglion nodosum and later found to bind to the cardiomyocyte subpopulation that develops into the ventricular conduction system, specifically the atrioventricular bundle and bundle branches.
An intronic regulatory element from the mouse Mef2c gene (encoding myocyte enhancer factor 2C) identified in transgenic mice. When used to drive either lacZ or cre expression, this enhancer was found to be active specifically in the anterior second heart field (AHF) during early cardiogenesis and was termed the Mef2c–AHF-enhancer.
- MYH6-promoter-based antibiotic selection
In this technique, a plasmid containing an antibiotic selection cassette (an antibiotic resistance gene controlled by the MYH6 promoter) is inserted into the stem cell population of interest. Administration of this antibiotic during T-box transcription factor TBX3-induced cardiac differentiation results in enrichment of pacemaker-like cells.