Epithelial- and Endothelial- to Mesenchymal Transition: from Cardiovascular Development to Disease

Introduction

Cellular switching from an epithelial to mesenchymal phenotype, and conversely from a mesenchymal to epithelial phenotype, are important biologic programs that are operative from conception to death in mammalian organisms. Indeed, the capacity of cells to switch between these states has been fundamental to the generation of complex body patterns throughout evolution. Phenotypic switching from an epithelial to mesenchymal cell, termed epithelial to mesenchymal transition (EMT), was a paradigm that evolved from numerous observations on early embryonic development, the foundations of which date back to 1920's and the pioneering work of Johannes Holtfreter on embryo formation and differentiation.^1,2 By the late 1960's, seminal chick embryo studies by Elizabeth Hay³ led to the first formal description that epithelial cells can undergo a dramatic phenotypic `transformation' and give rise to embryonic mesoderm.<sup>4</sup> Subsequent studies have revealed that this process is reversible (mesenchymal to epithelial transition [MET]) and gradually, the term `transition' has come to replace `transformation'.

Given that EMT/MET was initially identified and described by developmental biologists, it is perhaps not surprising that these processes are best understood during embryonic implantation and development. As explored in this review, it is now known that successive waves of cellular transition, from an epithelial to mesenchymal and then back to epithelial state, are required for normal embryonic patterning and organ formation. In addition, numerous studies which span a broad spectrum of physiologic and pathologic conditions have expanded our knowledge of EMT/MET and now provide evidence for the important role played by these processes in various adult conditions including fibrosis, wound repair, inflammation and malignancy. Indeed, our conceptual framework now also encompasses several variations and sub-categories of cellular phenotypic switching, including endothelial to mesenchymal transition (EndMT).

In this review, epithelial, endothelial and mesenchymal phenotypic cellular switching will be explored in the cardiovascular system, spanning cardiovascular development through to adult end organ disease. Key areas of recent scientific progress will be examined, including recent developmental and pathological insights, which potentially may lead to novel therapeutic opportunities.

EMT: a key role in early development

Within days of conception and during very early embryonic implantation, the process of EMT is already operative. Typically at the blastocyst stage, following initial adherence to the uterine lining (decidua), the outer trophoblast sends forwards columns of epithelial cells to penetrate the uterine wall.⁵ At the leading edge of these embryonic cellular columns epithelial trophoblast cells undergo EMT and invade the underlying maternal decidual interstitum and vessels. These invading embryonic cells ultimately go on become mesenchymal placental giant cells, participating in the remodeling of the maternal uterine vasculature and securing a functional placental blood supply.⁵ This embryonic execution of the EMT program establishes cellular phenotypic switching as a key biologic paradigm and interestingly, sets an early precedent for the vascular involvement of this process.

Soon after these events EMT plays a pivotal role in germ layer specification (ectoderm, mesoderm, endoderm). Epithelial cells from the primitive epiblast layer migrate to the midline and undergo EMT to give rise to mesoderm and endoderm.⁶ This process is highly ordered in time and space, and generates pools of primitive stem/progenitor cells at precise anatomical locations within the embryo, which constitute the primordia of the developing organs. For example, lateral plate mesoderm gives rise to the heart and hematopoietic cells, the paraxial mesoderm to the musculo-skeletal system, and the intermediate mesoderm to the urogenital tract. These events, occurring in tissues that have not previously undergone cellular phenotypic switching, are termed primary EMT (Figure 1).

Figure 1.

Organ formation occurs via EMT/MET. Embryonic cells from the epiblast layer (a single layer of epithelial cells) give rise to mesodermal cells via primary EMT. Definitive organ formation then ensues via secondary EMT and successive rounds of EMT-MET. Differing regions of the mesodermal layer give rise to differing organs/structures: the heart, circulatory system and hematopoietic cells arise from lateral plate mesoderm; the urogenital system from intermediate mesoderm; the axial skeleton and skeletal muscle from paraxial mesoderm, and; the primitive notochord (which becomes the nucleus pulposus of intervertebral discs in the adult human) from the axial (midline) mesoderm. This figure depicts cardiac formation. EMT-MET is also involved with endoderm and ectoderm tissue/organ formation.

Having established these primitive mesodermal populations, successive waves of EMT/MET then typically occur before final organ formation. In the case of the heart, this can involve recurrent waves of EMT/MET before the heart begins to assume a recognizable 4-chambered form (see below). For other organs, such as the kidney, successive waves of EMT/MET are required to ultimately give rise to epithelial structures such as nephrons and nephric ducts.⁷

Genetic, molecular and cellular basis of EMT/MET

At the core of EMT/MET, fundamental molecular and architectural rearrangements occur to bring about the dramatic cellular changes necessary to switch phenotypes. Underpinning this is a complex network of gene activation and repression programs that are required for the initiation, execution and maintenance of EMT or MET (Figure 2).

Figure 2.

Key aspects of the molecular and cellular changes that occur with EMT. Abbreviations not previously defined: BM, basement membrane.

At the cellular level, gross changes in polarity, morphology, functionality and cell-cell interactions are requisite steps. Epithelial cells are arranged on a basement membrane, exhibiting apico-basal polarity and abundant expression of inter-cellular adhesion complexes such as E-Cadherin and integrins. In order to adopt a mesenchymal phenotype these cells must lose cell adhesion by E-Cadherin downregulation and degradation.^8,9 Other epithelial proteins such as zonula occludens, cytokeratin and desmoplakin must also be repressed.^8–11 Transitioning cells then progressively lose polarity while eroding the basement membrane by metalloproteinase (MMP) production (MMPs 2, 3 and 9).^12–13 Cytoskeletal changes mediated by Rho GTPases induce apical constriction and further structural rearrangements to permit passage through the degraded basement membrane, culminating in delamination from the epithelial layer.¹⁴ Completing the transition, cells activate the expression of additional mesenchymal genes and proteins, such as alpha-smooth muscle actin (αSMA), smooth muscle protein 22α (SM22α), collagen I and III, vimentin, fibronectin or fibroblast specific protein 1 (FSP1; also known as S100A4).^8,15–17

Orchestrating these processes, the most widely described regulator of EMT/MET is the TGF-β superfamily of signaling molecules (TGF-β, Nodal, bone morphogenic proteins [BMPs], growth and differentiation factors [GDFs]).^18–21 Downstream of the receptors for the various TGF-β superfamily members, the Smad family of signal transducers are also of key importance, and in particular Smad2 and Smad3 appear to control EMT program activation.¹⁸ In turn, TGF-β stimulation with Smad activation leads to transcription of Snail 1 and 2 (the latter formerly known as Slug) and Twist, further triggering a cascade of signaling pathways that culminate in EMT.²² Snail1 is considered a key organizer of EMT, with one of its pivotal functions being to downregulate E-Cadherin transcription.^22,23 The downregulation of E-Cadherin, and other tight junction components,⁹ directly facilitates the loss of epithelial intercellular adhesions and cellular delamination from the epithelial layer. The TGF-β signaling network is a key mediator of early developmental EMT, and while other non-EMT effects are also operative, the genetic mutation of any number of these TGF-β family members often results in a non-viable embryo with major tissue, cardiovascular or other organ specification defects.^24–28 Importantly, not all members of the TGF-β superfamily stimulate EMT, with BMP-7 antagonizing TGF-β signaling and inhibiting EndMT in cardiac (see below)¹⁹ and renal²⁰ fibrosis models.

Several other signaling pathways also modulate EMT and MET, such as Notch and Wnt. For example, the canonical Wnt signaling pathway is crucial for events in early embryonic development, with chick embryos deficient in Wnt unable to properly form the ectoderm, mesoderm and endodermal layers.²⁹ The Notch pathway is an evolutionarily conserved and complex signaling network which also plays an important role in embryonic development. While not generally considered a `master' regulator of EMT, at least in certain scenarios it appears to act upstream of TGF-β signaling.³⁰ As a whole, the cellular rearrangements and changes that occur with EMT are extensive, and it is estimated that EMT in human cells changes the expression of about 4000 genes, representing ~10% of the entire genome.³¹ Additional signaling pathways are discussed below as relevant to cardiovascular development and disease, and interested readers are referred to any of the several excellent reviews on the molecular basis of EMT.^32–34

Classification and types of EMT

EMT is classified into 3 types depending on its biologic (or pathologic) role and the time-window in which it occurs. The embryonic and developmental EMT programs described above are classified as type 1 EMT, being distinct in that they do not generally cause fibrosis or give rise to mesenchymal cells with an invasive phenotype. As shown in Figure 3, the remaining types (2 and 3) are operative after birth and are concerned with fibrosis and malignant cellular transformation respectively. Type 2 EMT is extensively described in the literature, and it appears that chronic inflammation may be the sovereign inciting injury which triggers this form of EMT and sets the stage for end organ disease. Although not without controversy,^35,36 evidence exists to support type 2 EMT in numerous adult conditions, including those affecting the kidney,^17,20,37 liver,³⁸ skin,³⁹ intestines,⁴⁰ lung,⁴¹ eye⁴² and heart (the latter will be considered separately below).¹⁹ Type 3 EMT is an important step in malignant cell transformation. Of particular relevance with respect to carcinoma (epithelial cell tumors), EMT is proposed as the critical mechanism for the acquisition of malignant characteristics, with loss of E-Cadherin expression facilitating the delamination and metastasis of transformed epithelial cells.^34,43

Figure 3.

Classification of EMT. Type 1 EMT is highly regulated and is associated with embryonic implantation an organ formation. Type 1 EMT may be sub-classified into primary (cells giving rise to EMT for the first time), secondary and tertiary; with these latter forms involved in the successive waves of EMT-MET leading to definitive organ formation. Type 2 EMT is associated with inflammation and fibrosis, and is now increasingly recognized in adult pathologic conditions. Type 3 EMT is involved with malignant cell transformation, including the acquisition of invasive metastatic cellular properties. As distinct from type 1, neither type 2 or type 3 EMT adhere to any higher-order program of spatial or temporal restriction.

A further aspect of this classification system requiring clarification is the place of EndMT. The endothelium is a specialized form of squamous epithelial tissue, and as such, EndMT is a sub-category of EMT. Accordingly, EndMT may be observed in each of the 3 categories of EMT. For example, the endothelium gives rise to hematopoietic cells during embryonic development (type 1 EMT),⁴⁴ and to both fibrosis and malignancy in the adult (types 2 and 3 EMT respectively).^19,45

EMT during cardiac development: Valves, Cushions, Neural Crest and Epicardial-derived cells

Early cardiac formation

The heart forms via a remarkable series of sequential waves of EMT/MET. As the definitive germ layers emerge in the developing embryo, cardiac progenitors are amongst the first epiblast cells to undergo EMT and to migrate out from the primitive streak.^46,47 This population of mesenchymal cardiac precursors migrates bilaterally within the lateral plate towards the anterior pole of the embryo,^47–49 to coalesce in mammals into an anterior cardiac crescent.⁴⁷ The formation of the celomic cavity divides the lateral plate to give rise to the somatic and splanchnic mesodermal layers. Within the splanchnic layer, primitive cardiac progenitor cells organize into a bilayered epithelium via MET.⁵⁰ Next, either via another round of EMT/MET involving these mesodermal cardiac progenitor cells or from a separate cell population, the endocardial cells which will line the cardiac structures are formed.^51–53 The primitive heart tube soon emerges by folding and remodeling of the cardiac crescent cells. Genetic and direct labelling of cardiac precursor cells at various stages of heart tube development has shown that the primary heart tube only contains the precursors of the left ventricle, and that the remaining chambers are formed by the progressive infiltration and incorporation of new cardiac precursors to the outflow and inflow poles of the heart tube.^54–56 The area of the cardiac crescent that gives rise to the initial heart tube is named the `primary heart field', while the area that remains behind in the pharyngeal region that is added later is called the `secondary heart field'. Primary and secondary heart field precursors occupy adjacent areas in the cardiac crescent but differ in the mechanism by which they are added to the heart tube (folding versus migration) and in the timing of addition. At the current time, the role of EMT in the migration and incorporation of second heart field precursors remains under investigation.

Endothelial to mesenchymal transition contributes to valve formation and heart septation

Soon after the primitive heart tube appears, endothelial cells from the region of the forming atrio-ventricular (A-V) canal and of the outflow tract (OFT) region undergo another round of EMT, or more specifically EndMT, as the cells undergoing phenotypic switching are endothelial. At this time the endocardium and the myocardium are separated by a thick acellular matrix termed the cardiac jelly. As the cardiac chambers start to form, the cardiac jelly gets thicker in the A-V canal and OFT regions, where endothelial-derived mesenchymal cells invade the adjacent cardiac jelly to form endocardial cushion tissue.⁵⁷ The OFT cushions are the precursors of the semi-lunar valves, while the A-V cushions give rise to the A-V septum, the membranous part of the ventricular septum and the mitral and tricuspid valves.⁵⁸ The extent to which EndMT contributes to these two cushion areas differs significantly, and while most of the A-V cushion mesenchyme derives from EndMT,⁵⁷ most of the OFT cushion derives from pharyngeal mesodermal cells. Subsequently, both cushions receive further specific mesenchymal cellular contributions involving EMT. The OFT region receives an important third mesenchymal population arising from the neural crest, which delaminates by EMT from the neural tube.⁵⁹ This neural crest-derived population is essential for OFT septation into the aortic and pulmonary trunks; however, it represents a transient population that does not contribute significantly to the definitive heart structures. In the region of the A-V canal, a third mesenchymal population derives from EMT of the epicardium (see below).^60,61

Epithelial to mesenchymal transition and the epicardium

In parallel with endocardial EndMT and cushion formation, the outermost epicardial layer of the heart is also coming in to existence.⁶² Like the primitive early cardiac progenitor cells, epicardial progenitor cells also arise from the splanchnic mesoderm (likely also via MET). Initially, the cells destined to form the epicardium assemble to create a transitory body of cells termed the proepicardial organ; consisting of an accumulation of pericardial progenitor cells lying adjacent to the sinus venosus (the venous pole of the heart) (Figure 4A). These proepicardial cells migrate, or in some species float freely within the pericardium, and attach to the myocardial surface.^63–67 There, they proliferate and flatten to cover the embryonic heart as the epicardial sheet. Concomitantly, some epicardial cells undergo EMT and generate a mesenchymal population of epicardial-derived cells (EPDCs). While a population of EPDCs remain to occupy the extracellular matrix-rich region between the epicardium and myocardium named the subepicardial space, some migrate further to invade the myocardium.

Figure 4.

Origin and fate of epicardial-derived cells (EPDCs). (A) mRNA in situ hybridization (ISH) for tbx18 on zebrafish sagittal heart sections marking epicardial cells (blue staining). Additional immunostaining against myosin heavy chain was performed to identify myocardium (brown staining) and is shown in the lower panels (e – g). These panels show the progressive developmental sequence of epicardial formation in zebrafish, with vertically matched panels taken from the same developmental stage (a, 2 days postfertilization (dpf); b and e 3 dpf; c and f, 4 dpf; d and g, 5 dpf). Proepicardium (PEO), Epicardium (Epi), Heart Tube (HT), Ventricle (V), Atrium (A). (B) Epicardial EMT during zebrafish regeneration. (a, c) anti-GFP immunohistochemistry on adult zebrafish heart sections from the transgenic ET-27 line, expressing GFP constitutively in all epicardial cells (brown staining). (b, d) ISH on sections of cryoinjured Tg(wt1b:GFP) zebrafish hearts, in which GFP is reactivated in the epicardium upon damage (blue staining). Dorsal is to the top (a) The control heart reveals a single layer of GFP-positive cells on the ventricular surface. (b) Upon cryoinjury, the epicardium thickens as a consequence of injury-induced EMT. (c) 20× enlargement of control heart. (d) 20× enlargement of injured heart. Images acquired at Centro Nacional de Investigaciones Cardiovasculares (Madrid, Spain) by the group of Nadia Mercader.

The developmental cellular contributions of EPDCs are controversial and potentially species-dependent. An important contribution of EPDCs is made to the A-V canal cushion mesenchyme, where EPDCs merge with endocardium-derived cells to form the mitral and tricuspid valves and cardiac septa.^60,61,68 Interestingly, EPDCs do not colonize the OFT cushions, perhaps implying a specific role for these cells in the formation of the tricuspid and mitral but not the pulmonary and aortic valves. EPDCs also play a role in coronary vascular formation. Studies in avian species, which included labeling the proepicardium with replication deficient virus^62,69 or the generation of quail-chick chimeras,^60–62,70 indicated that EPDCs are the primary source of coronary endothelial cells, coronary vascular smooth muscle cells (cVSMCs) and cardiac fibroblasts. More recently, genetic fate mapping experiments have investigated the fate of EPDCs in the mouse. T-box transcription factor 18 (Tbx18) and Wilms tumor suppressor 1 (Wt1) are broadly expressed in the proepicardium and epicardium during development and represent appropriate markers to trace EPDCs.^71–74 Using Cre-based technology, the fate of Tbx18⁺ and Wt1⁺ cells was analyzed,^68,75 confirming that EPDCs differentiate into cVSMCs and cardiac fibroblasts. However in mice, in contrast to the chick, no contribution to coronary endothelial cells was identified for Tbx18⁺-derived cells and only a minor contribution was found from Wt1⁺ cells. Very recently, Kikuchi et al have also demonstrated that in zebrafish, while EPDCs contribute to perivascular cells, they do not give rise to endothelial cell populations.⁷⁶ Potentially, this discrepancy regarding the contribution of EPDCs to the endothelium in avian versus other species may be explained by species-specific differences or may reflect differing experimental approaches used to study EPDC fate. It is also possible that the initial observations performed in the chick were not correctly interpreted, or that contamination by non-EPDC endothelial precursor cells migrating together with the proepicardium during grafting/labeling may have occurred.^77,78

Consistent with this, genetic evidence from the mouse supports the classical notion that the coronary endothelium derives from the sinus venosus.⁷⁹ Using inducible VE-cadherin-Cre mice to trace endothelial cell clones, it was determined that most coronary veins and arteries derive from sprouts arising at the sinus venosus, with a lesser proportion potentially arising from the endocardium. Importantly, no endothelial clones were linked to the proepicardium, suggesting that this structure does not contribute to endothelial progenitors, although, it remains possible that epicardial cells may commit to the endothelial lineage at a later developmental stage, after colonization of the myocardial surface.

Interestingly, analysis of Tbx18⁺ and Wt1⁺ epicardial-derived lineages also revealed an unexpected myocardial contribution. Thus, around 4% of total cardiomyocytes were found to have arisen from Wt1⁺cells, contributing mostly to the intraventricular septum (IVS) (10%) and atria (18%) and to a lesser extent to the ventricular walls (7%).⁷⁵ Similarly, the Tbx18⁺-lineage was found to contribute to cardiomyocytes in the ventricular septum and in scattered areas within the ventricular walls and atria.⁶⁸ However, these findings remain under scrutiny as Tbx18 expression has been detected in maturing cardiomyocytes, potentially suggesting epicardial-independent Tbx18 activation in these cells.⁸⁰ Furthermore, studies in zebrafish have also refuted the ability of EPDCs to give rise to cardiomyocytes.⁷⁶

Most recently, yet another contribution of EPDCs was revealed by Harvey and co workers, who described a population of mesenchymal stem cell-like cells that occupy a perivascular, adventitial niche in the adult mouse.⁸¹ While the precise extent of their normal and pathological cellular contributions remain to be defined, these proepicardial/epicardial-derived stem cells exhibit trans-germ layer potency in vitro and in vivo, and appear distinct from previously described resident cardiac stem cell populations.

In summary, it is clear that EMT-derived EPDCs support coronary artery development by supplying vascular pericytes and cVSMCs, and that they make a major contribution to fibrous cardiac tissues/populations, including resident perivascular fibroblasts and the fibrous cardiac skeleton.^{60–62,68,69,75,77,78,82} Although the ability of EPDCs and/or other epicaridal cells to give rise to cardiomyocytes and endothelium during development remains controversial, it is clear that the EMT and MET programs are of key importance during cardiac formation and for epicardial provisioning of specific cell populations.

Signaling pathways governing EndMT/EMT during cardiovascular development

Signaling via the TGF-β superfamily, as previously described in this review, is the major regulator of EMT during cardiac formation, including TGF-β2, TGF-β3 and BMP-2, and the downstream transcription factors Snail1 and Snail2.⁸³ The precise role of these TGF-β isoforms differs between species, and at least in the chick, TGF-β2 mediates EndMT via endothelial cell activation and separation, while TGF-β3 mediates cell invasion into the extracellular matrix.⁸⁴

In addition to TGF-β, numerous other pathways modulate EMT during cardiac formation and development. Notch signaling is of particular significance during murine cardiac development, functioning to promote endocardial EndMT and with Notch deficient embryos displaying atrophic valve formation.⁸⁵ Expanding our understanding of these pathways, Luna-Zurita et al⁸⁶ recently showed that Notch1 is sufficient to activate a cell-autonomous, promesenchymal gene expression program in endocardial cells. BMP-2 was found to drive endocardial EndMT and mesenchymal cell invasion into the cardiac jelly. Further, myocardial BMP-2 inactivation impaired Notch1 activity,⁸⁶ suggesting a model in which the interplay between myocardial BMP-2 and endocardial Notch signaling restricts EndMT to prospective valve territory.^86,87

Endocardial EndMT is also dependent on receptor tyrosine kinase (RTK) signaling via the phosphoinositide-3 kinase (PI3K)-phosphoinositide-dependent protein kinase 1 (PDK1)-Akt/protein kinase B (PKB) cascade, which is upstream of Snail. Genetic ablation of PDK1 in endothelial cells leads to embryonic lethality with abnormal vascular remodeling and a failure of endocardial cushion development due to defective EndMT.⁸⁸ Gata4, an upstream regulator of an Erbb3-Erk pathway, is expressed in the endothelium and mesenchyme of the embryonic A-V valves and is also required for endocardial EndMT. Selective Gata4 inactivation in endothelial-derived cells results in a failure of EndMT and hypocellular endocardial cushions in animal models.⁸⁹ This phenotype corresponds with that seen in humans, where heterozygous Gata4 mutation is associated with defects in the inter-atrial or inter-ventricular septae.^90,91 Endocardial expression of the protein tyrosine phosphatase SHP2, encoded by the gene PTPN11, also regulates EndMT and endocardial cushion formation. Gain of function PTPN11 mutations are responsible for a significant proportion of cases of Noonan syndrome and cause cardiac valve and septal defects by increasing Erk-MAPK activation, probably downstream of ErbB family RTKs, thus extending the time-window during which endocardial EndMT is operative.⁹² Endocardial EndMT is also restrained by inhibitory influences such as endocardial cell expression of Nfatc1 (nuclear factor of activated T cells, cytoplasmic 1), which functions to inhibit EndMT and promote semi-lunar valve elongation in a cell-autonomous manner.⁹³ In addition, cushion mesenchymal cells are especially sensitive to reactive oxygen species-mediated injury, suggesting that oxidative stress may contribute to the formation of congenital cardiac defects.⁹⁴

Additional specific factors are operative during epicardial EMT and EPDC migration. For example, Wt1 mutant mice display disrupted epicardial formation, and reduced numbers of EPDCs and their derivatives.^73,95 Consistent with this, conditional ablation of Wt1 in Gata5-Cre lineage cells (marking proepicardium- and epicardial-derived cells) has been reported to reduce epicardial Snail1 and increased E-Cadherin levels, thus inhibiting the EMT program.⁹⁶ However, the influence of Wt1 on epicardial E-Cadherin is disputed.⁹⁷ Interestingly, during kidney development Wt1 regulates MET, rather than EMT, thereby highlighting the importance of Wt1 in controlling the switch between a mesenchymal and epithelial state.^95,98

Potentially upstream of Wt1, epicardial EMT is known to be modulated by thymosin β 4 (Tβ4), an actin-binding peptide. Although not required for proper cardiac formation,⁹⁹ during development Tβ4 is expressed by the myocardium, providing a paracrine stimulus for EPDCs to migrate and differentiate into cVSMCs and facilitating vascular development.¹⁰⁰ This observation supports the concept that cross-talk between different regions of the heart during development via non cell-autonomous signaling may exert a major influence on EMT and EPDCs. As a further example Van Gogh-like 2 (Vangl2), a component of the planar cell polarity signaling pathway, is expressed exclusively in myocardial cells of the developing heart. However, Vangl2 mutant mice with disruption of the planar cell polarity pathway exhibit cardiac defects that involve the epicardium and EPDCs. In these mice, among other defects, reduced fibronectin deposition in the subepicardial space is associated with reduced EPDC migration into the ventricular myocardium, where the coronary vessels fail to develop an intact cVSMC layer.¹⁰¹

β-catenin also regulates epicardial EMT. Conditional epicardial β-catenin deletion using the Gata5-Cre driver line leads to incomplete invasion of EPDCs into the myocardium and disruption of cVSMC formation.¹⁰² Wu et al showed that β-catenin affects epicardial EMT by regulating cell spindle orientation and stabilization of adherens junctions.¹⁰³ These investigators also demonstrated that proliferation is a prerequisite for epicardial cells to become mesenchymal, and that the mitotic division plane should be orthogonal to the myocardial surface in order to enable EMT.¹⁰³ A direct interaction between Wt1 and β-catenin has also recently been suggested based on the observation that β-catenin is downregulated in Wt1 mutant epicardium.⁹⁷

Platelet derived growth factor (PDGF) signaling is yet another factor that is able to induce epicardial EMT.¹⁵ While PDGFRβ is required for EMT and the subsequent differentiation of cVSMCs,¹⁰⁴ PDGFRα is required for the subpopulation of EPDCs giving rise to intracardiac fibroblasts.¹⁰⁵ Additional factors that regulate EMT/MET during cardiac development are shown in Table 1.

Table 1.

Genes and proteins known to be of importance for EMT/MET and EPDCs during cardiac development and in cardiovascular disease.

Gene/protein	Expression	Function
BMP family (see also TGF-β)	Myocardial	BMP-2 promotes endocardial EndMT and mesenchymal cell invasion into the cardiac jelly, and also interacts with endocardial Notch1 to spatially govern endocardial EndMT.86 BMP-7 attenuates EMT-related cardiac fibrosis.19
Connexin 43	Ubiquitous	Governs cell polarity and cytoskeletal integrity. Required for epicardial EMT, EPDC migration and OFT myocardialization. Likely also involved with endocardial EndMT.106,107
FGF family	Differs among family members	Broadly promote epicardial EMT and regulate other developmental events such as myocardial invasion and endothelial differentiation of EPDCs.108–109 Epicardial EMT during heart regeneration is dependent on FGF signaling.110,111
Gata4	Proepicardium, endothelium and mesenchyme of A-V valves	Upstream regulator of Erbb3-Erk pathway, which is required for formation of the proepicardium112 and endocardial EndMT.89 Human Gata4 mutations are associated with atrial and ventricular septal defects.90,91
Nfatc1	Proepicardium, epicardium, valve endocardium	Inhibits EndMT by inhibiting Snail1 and Snail2; promotes cell proliferation and semi-lunar valve elongation.93 Promotes invasion of EPDCs into the myocardium by induction of extracellular matrix-degrading enzyme gene expression.113
Notch pathway	Epicardium and endocardium	Promotes endocardial EndMT. Notch deficient embryos have atrophic valve formation.85 Notch pathway includes upstream protein kinase D2, Histone deacetylase 5, Kruppel-like factors, and downstream Hey1, Hey2, and HeyL.148 Also involved in epicardial/proepicardial development and EPDC differentiation.115,116
PDGF-PDGFR signaling	PDGFRs by EPDCs. PDGF by endothelial and other cardiac cells.	PDGFRβ required for epicardial EMT and cVSMC differentiation. PDGFRα required for EPDC contributions to cardiac fibroblasts.15,104,105 Inhibition of PDGF signaling disrupts epicardial EMT during cardiac injury repair.117
PDK1	Ubiquitous	Endothelial deletion leads to abnormal vascular remodeling and defective endocardial EndMT with decreased Snail expression and embryonic lethality.88
Periostin	Fibrous cardiac skeleton and endocardial cushions	Required for maturation and extracellular matrix stabilization of non-cardiomyocyte cardiac lineages, with knockout mice exhibiting minor leaflet abnormalities. Marks EndMT/EMT-derived mesenchymal cells in the developing heart.82,118
PTPN11 (encodes SHP2)	Ubiquitous	Gain of function mutations cause human valve and septal defects (Noonan syndrome) by increasing Erk-MAPK activation and increasing the extent of endocardial EndMT.92
Reactive oxygen species	Ubiquitous presence	Affect cushion mesenchymal cells, which may lead to congenital cardiac defects via aberrant EndMT/EMT.94
Retinoid X receptor alpha	Ubiquitous	Promotes epicardial EMT via a Wnt signaling pathway.119
Tβ4	Myocardium	Dispensable for cardiac development,99 but regulates EPDC migration and differentiation into cVSMCs and facilitates cardiac vascular formation.100 Stimulates epicardial EMT after injury and enhances cardiac repair.120–122
TGF-β superfamily	Ubiquitous	Critical role in cardiac EndMT/EMT. Includes multiple isoforms, other TGF-β family members (see BMP) and downstream factors.83 The exact role of each isoform and family member differs between species.84 TGF signaling is implicated in EndMT-related cardiac fibrosis.19,123 ALK2, a TGF-β/BMP receptor, is implicated in EndMT-related vascular calcification.123 In vitro, while TGF-β induces EMT in proepicardial explants,124 it may have inhibitory effects in avian epicardial explants.108
Vangl2	Myocardial	Vangl2 mutants have reduced subepicardial fibronectin, reduced EPDC migration, enlarged ectopic coronary vessels with deficient cVSMC coverage and disrupted cardiomyocyte organization.101
VEGF	Dynamic temporal expression during development	Promotes endocardial EndMT125 and possibly epicardial EMT.108,126,127
Wnt/β-catenin	Epicardium	Regulates EPDC migration and cVSMC formation via effects on cell spindle orientation and stabilization of adherens junctions.102,103
Wt1	Epicardium	Required for epicardial EMT and EPDC formation.73,97 Wt1 is upstream of canonical Wnt/β-catenin signaling.97

Periostin, a secreted protein, is associated with the endocardial EndMT and epicardial EMT programs. While its role as a true mediator of EndMT/EMT seems less likely, it appears to mark EndMT/EMT-derived mesenchymal cells in the developing heart. Periostin is expressed throughout the cardiac fibrous skeleton and emerging endocardial cushions but is absent from normal and/or pathological mouse cardiomyocytes.^82,118 Periostin knockout mice exhibit minor leaflet abnormalities in the viable majority. Assessment of collagen production, 3D lattice formation ability, and TGF-β responsiveness indicate that periostin is required for maturation and extracellular matrix stabilization of non-cardiomyocyte lineages of the heart.¹¹⁸ The utility of manipulating periostin and several of the abovementioned pathways to attenuate cardiac fibrosis, regeneration and remodeling is currently under investigation.

EMT and EndMT in the adult cardiovascular system

Epicardial EMT in the adult heart

In the adult heart, the epicardium constitutes a unicellular layer of epithelial cells. Several animal models of myocardial infarction or cardiac damage have shown that an early response to injury consists of the reactivation of epicardial-specific genes usually expressed only during development such as Wt1.^{121,128–131} While this epicardial genetic reactivation is initially generalized in nature, it subsequently becomes restricted to the site of damage, suggesting a role for the epicardium during cardiac wound healing. Concomitantly, epicardial cells undergo EMT, proliferate and accumulate in the subepicardial space to form a thickened epicardial cap comprised of EPDCs (Figure 4B).^{110,116,117,120,128,130,132} In zebrafish, inhibition of fibroblast growth factor (FGF) signaling by genetic manipulation results in a lack of epicardial marker gene expression within wounds induced by partial apical cardiac amputation, suggesting that during heart regeneration epicardial EMT is dependent on FGF signaling.¹¹⁰ Similarly, pharmacological inhibition of PDGF signaling was also shown to disrupt epicardial EMT during zebrafish cardiac injury repair.¹¹⁷ In the mouse, both myocardial infarction and ventricular pressure-overload induce Wnt1- and Notch-regulated activation of epicardial cells, which then undergo EMT and contribute to fibrosis repair.^116,133 Together these studies indicate that, at least in modified form, the cardiac developmental EMT programs may become reactivated in the adult.

Consistent with their fate during cardiac development, lineage tracing of epicardial cells following myocardial infarction in mice has shown that EPDCs give rise to fibroblasts, myofibroblasts and smooth muscle cells but, at least in the absence of additional stimulation, not to endothelial cells nor cardiomyocytes.¹³⁰ Interestingly, after myocardial infarction, epicardial cells were also found to produce trophic factors which promote capillogenesis such as VEGFA.¹³⁰ Confirming this paracrine effect, supplementation of the epicardium with transplanted EPDCs or EPDC-conditioned media after myocardial infarction improved subsequent cardiac function.^130,134

Riley and co-workers have provided significant evidence to demonstrate that Tβ4 can stimulate epicardial EMT after injury and enhance myocardial repair. Using explanted tissues, these investigators provided proof of concept that with Tβ4 stimulation, adult murine epicardial cells undergo EMT in vitro and differentiate into smooth muscle cells, fibroblasts, cardiac progenitors and endothelial cells.¹³⁵ In vivo studies of Tβ4 stimulation have shown that this peptide is able to induce a dramatic EPDC contribution to neovascularization following myocardial infarction.¹²⁰ Most recently, Riley and co-workers reported that in Tβ4-pretreated mice, a small subset of around 1% of Wt1⁺ cells can give rise to cardiomyocytes following myocardial infarction.¹²¹ However, while other salutary effects are present, when initiated soon after myocardial infarction Tβ4 treatment does not reprogram epicardial cells into cardiomyocytes.¹³⁶ Nevertheless, based also on other data suggesting that the epicardium may harbor adult cardiac progenitor cells,^137,138 intense research efforts are currently underway investigating the possibility that augmenting the normal post-injury response by activating epicardial EMT may be of therapeutic utility following cardiac injury or infarction in humans.

EndMT contributes to cardiac fibrosis in the adult heart

Evidence implicating EndMT in cardiac fibrosis has been mounting for several years. In addition to the above studies and as important corroborating data, it is well described that TGF-β plays a major causal role in myocardial fibrosis and diastolic dysfunction through fibroblast activation in pressure-overload rodent models, with systemic administration of TGF-β-neutralizing antibody preventing this process.¹³⁹ In a landmark publication in 2007, Kalluri and co-workers¹⁹ demonstrated that EndMT makes a significant contribution to myocardial fibrosis in the adult heart. These investigators used murine models of cardiac fibrosis/heart failure in association with a Cre-lox genetic marking system to identify cells that have ever expressed Tie1 (an endothelial marker), revealing that 27 – 33% of all cardiac fibroblasts are of endothelial origin in these models (Figure 5A). Further, TGF-β1 induced cultured endothelial cells to undergo EndMT in vitro, whereas BMP-7 preserved the endothelial phenotype. The use of mice genetically deficient in Smad3, or the systemic administration of recombinant human BMP-7, inhibited EndMT and the progression of cardiac fibrosis in vivo.¹⁹ Treatment with BMP-7 was associated with a decrease in the number of fibroblasts of endothelial origin, providing strong evidence for the importance of EndMT in cardiac fibrosis. These provocative findings suggest that the inhibition of EndMT/EMT with BMP-7 or other means may be a promising target for clinical therapeutic translation in settings such as cardiac fibrosis where EndMT/EMT is injury-causing.

Figure 5.

EndMT contributes to cardiac fibrosis as demonstrated by confocal microscopy with immunofluorescence double-label staining. (A) Aortic banding was used as a rodent model of cardiac hypertrophy/fibrosis. Mice used in these studies were Tie1Cre;R26RstoplacZ, in which all cells of endothelial origin are permanently marked by β-gal expression. Staining was performed with antibodies against β-gal (red) and S100A4, a fibroblast marker (green). Arrows indicate colocalization of β-gal and FSP1 expression involving both microvessels (left) and arterioles (right), indicative of endothelial-derived cells that have undergone EndMT and adopted a fibroblast-like phenotype. Scale bars, 20 μm. Figure reproduced with permission from Zeisberg et al.¹⁹(B) EndMT in cardiac fibrosis shown by co-expression of endothelial (CD31 - red) and fibroblast (S100A4 - green) markers. Z-stack images showed specific overlay of double immunostaining for CD31 and S100A4. Nuclear staining is shown in blue (DAPI). Figure reproduced with permission from Widyantoro et al.¹⁴⁰

While yet to be replicated with robust endothelial lineage tracing systems, supporting evidence for EndMT/EMT has since arisen from other models of cardiac fibrosis including hypertrophic cardiomyopathy,¹⁴¹ diabetes-induced cardiac fibrosis,¹⁴⁰ and perhaps more controversially with genetic deficiency of plasminogen activator inhibitor-1 in aged but not young mice.^142–144 For example, in mice with streptozotocin-induced diabetes, evidence for EndMT was provided by the identification of cells co-positive for endothelial (CD31) and mesenchymal (αSMA and FSP1) markers (Figure 5B). These investigators also showed that EndMT could be induced in vitro by endothelial cell culture under high glucose conditions, and that this process was mediated by the endothelial expression of endothelin-1 and TGF-β.¹⁴⁰ Angiotensin II, which is known to have pro-fibrotic effects in the heart and elsewhere,¹⁴⁵ may also promote cardiac EndMT. While evidence is limited, Tang et al¹⁴⁶ suggested the administration of irbesartan (an angiotensin II receptor antagonist) reduces interstitial cardiac fibrosis and left ventricular dysfunction by inhibiting EndMT in a rodent model of diabetic cardiomyopathy. These effects on cardiac EndMT and fibrosis are consistent with the clinical response to angiotensin II receptor blockade and offer novel insights into potential mechanisms of action of this class of drug.¹⁴⁷

Other contributions of EndMT and EMT to the adult cardiovascular system

The origins of vascular calcification, and in particular the origins of the cells that cause heterotopic vascular ossification, is a matter of controversy.¹⁴⁸ Importantly, osteoblasts and chondrocytes, the cells responsible for ossification, are mesenchymal cells. Therefore, the question arises as to whether osteoblasts and chondrocytes which are operative in adult pathological conditions may be derived from endothelial cells via EndMT. This possibility was investigated using samples from humans suffering fibrodysplasia ossificans progressiva, a condition in which heterotopic ossification occurs due to a gain of function mutation in the gene encoding ALK2, a member of the TGF-β/BMP family of receptors.¹²³ In this study, using a combination of in vivo murine lineage tracing and in vitro cell characterization experiments, Medici et al¹²³ provided extensive evidence of an endothelial origin of osteoblasts and chonrdocytes via EndMT. Provocatively, in addition to causing EndMT, these investigators identified that the expression of constitutively active ALK2 in endothelial cells leads to the acquisition of a mesenchymal stem cell-like phenotype. Similar results were obtained by treatment of untransfected endothelial cells with TGF-β2 or BMP-4 in an ALK2-dependent manner. These mesenchymal stem-like cells could be triggered to differentiate into osteoblasts, chondrocytes or adipocytes.¹²³ While it remains to be shown if EndMT contributes to vascular calcification in persons without this genetic mutation, this study lays the foundations for a wider exploration of the role of EndMT in the adult vasculature. Moreover, the possibility that EndMT generates mesenchymal stem-like cells is consistent with other emerging data indicating that MET is involved with the reprogramming of fibroblasts to become pluripotent stem cells.¹⁴⁹ Although not a core feature of the stem cell repertoire,¹⁵⁰ these data and other material presented in this review suggest that the ability to undergo cellular phenotypic switching via EMT/MET is involved with aspects of the generation and functionality of certain stem or progenitor cell populations.

While definitive lineage tracing studies are yet to be performed, EndMT/EMT is potentially implicated in several other adult pathological cardiovascular conditions. Foremost, pulmonary hypertension, regardless of its etiology, has been speculated to involve EndMT/EMT for several years.¹⁵¹ This is based on the fact that increased numbers of αSMA-expressing cells is a near universal finding in remodeled pulmonary vessels,¹⁵² and that the major signaling systems associated with EndMT/EMT are also operative in pulmonary hypertension, including the TGF-β, BMP, Smad pathways.¹⁵³ In addition, it has been demonstrated that adult pulmonary artery endothelial cells undergo EndMT in vitro.¹⁵⁴ As another chronic inflammatory condition affecting the vasculature, it is believed that EndMT is involved in the microvascular changes seen in Systemic Sclerosis.¹⁵⁵ Of relevance, Systemic Sclerosis bears many of the hallmarks of conditions involving EndMT/EMT, including mesenchymal cell proliferation and TGF-β signaling. EndMT/EMT may also be involved in mitral valve pathology¹⁵⁶ and transplant vasculopathy,¹⁵⁷ although again, we reiterate that at the current time the definitive involvement of EndMT/EMT in these abovementioned conditions remains to be demonstrated using rigorous lineage tracing systems.

Conclusions

Throughout life, EndMT, EMT and MET are conserved and integral cellular process that play a diverse role in biologic organism formation and pathologic end organ disease. Until recently, research in this field was hampered by a lack of specific endothelial and epithelial lineage tracing systems to follow the fate of these cells as they transition to a mesenchymal phenotype. However, this problem is being progressively overcome and accurate animal models for permanent genetic cell labeling are now increasingly being used for this purpose. These models differentiate themselves from all prior systems in their specificity, fidelity and accuracy, and enhance our capacity to address the scientific questions proposed. In addition, the possibility of tracking EMT/MET in vivo in humans may soon be a reality, with techniques such as fluorodeoxyglucose-positron emission tomography (FDG-PET), or PET-magnetic resonance imaging (PET-MRI) continuing to evolve and now approaching the resolution required for cell tracing studies.

As discussed, EndMT/EMT appears likely to play a role in numerous chronic cardiovascular disease states such as heart failure, pulmonary hypertension and various forms of chronic vasculopathy. Furthermore, as reviewed elsewhere, chronic inflammatory and malignant diseases are a particularly common finding in older persons.^158,159 Given that types 2 (inflammatory/fibrosing) and 3 (malignancy-associated) EMT appear to play an important role in the pathology of these conditions, EMT may prove to be a key aspect of age-related morbidity and mortality. The prospect that the therapeutic manipulation of EndMT/EMT may be used in the treatment of these conditions is particularly attractive. As a result, interest and research in this field is expanding rapidly. An increased understanding of the function and contribution of EndMT/EMT during development and in the adult should provide the foundation for the subsequent clinical exploitation of these pathways.